Citation
Cannabis Use and Coronary Disease

Material Information

Title:
Cannabis Use and Coronary Disease
Creator:
Crooke, Hannah R
Publisher:
University of Florida
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Epidemiology
Committee Chair:
STRILEY,CATHERINE L
Committee Co-Chair:
COTTLER,LINDA B
Committee Members:
PEARSON,THOMAS A
NIXON,SARA J

Subjects

Subjects / Keywords:
angina
artery
cannabis
cardiovascular
coronary
cvd
depression
epidemiology
health
heart
marijuana
mi
substance
thc

Notes

General Note:
About 44% of Americans report lifetime cannabis use, and about 60% of Americans support legalization. Identifying negative health outcomes related to cannabis use such as coronary disease is of great public health importance. The association between cannabis use and coronary disease was explored through a SLR and secondary data analyses examining the 1) direct association between cannabis use and coronary disease, and 2) association between latent classes of substance use, including cannabis use, and coronary disease. The SLR (Chapter 2) employed a structured electronic search resulting in sixteen articles identified for review. Secondary data analyses used data from HealthStreet, a community engagement model. The association between cannabis use and coronary disease was investigated through multiple logistic regression (Chapter 4), and latent class analysis with multinomial logistic regression (Chapter 5). The results of the SLR were equivocal. Both positive and negative associations between cannabis use and obesity, hypertension, dyslipidemia, and metabolic syndrome were found. Negative associations between cannabis use and diabetes and positive associations between cannabis use and angina and MI were identified. More research focused on the association between cannabis use and coronary disease is needed. Analyses revealed a significant relationship between cannabis use and coronary disease that was attenuated after controlling for other substance use, mental health conditions, and coronary disease risk factors. Results suggest any direct association between cannabis use and coronary disease may be the result of confounding. The second analysis revealed three latent classes of substance user: 1) mono-substance use and abstainer, 2) cannabis use and tobacco use, and 3) polysubstance use. The association between coronary disease and class membership was attenuated in fully adjusted models. When considering the association between cannabis use and coronary disease in the context of other substance use, other confounding factors such as mental health disorders cannot be ignored. This dissertation fills a gap in the literature and contributes to the growing public health interest in outcomes related to cannabis use.

Record Information

Source Institution:
UFRGP
Rights Management:
All applicable rights reserved by the source institution and holding location.
Embargo Date:
5/31/2018

Downloads

This item has the following downloads:


Full Text

PAGE 1

CANNABIS USE AND CORONARY DISEASE By HANNAH RENEE CROOKE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2017

PAGE 2

2017 Hannah Renee Crooke

PAGE 3

To my loving parents and husband without whom I would have never completed my schooling

PAGE 4

4 ACKNOWLEDGMENTS I would like to acknowledge my committee cha ir, Dr. Catherine Striley, who se teaching, support, and encourage ment have been invaluable. I would also like to thank the rest of my committee, Dr. Linda Cottler, Dr. Thomas Pearson, and Dr. Sara Jo Nixon for their continued support in my pursuit of a PhD. I would like to thank Dr. David Sheps for providing personalized tr aining in cardiovascular epidemiology. Finally, I would like to acknowledge the administrative assistance of Becca Pieters, and the administrative assistance of Erica Boyd, Tamara Millay and Betsy Jones. Hannah Crooke is a pre doctoral trainee supported through the University of Florida Substance Abuse Training Center in Public Health (T32DA035167 Cottler PI ). The content of this dissertation is solely the responsibility of authors and does not necessarily represent the official views of the National Ins titutes of Health.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWL EDGMENTS ................................ ................................ ................................ ......... 4 LIST OF FIGURES ................................ ................................ ................................ ................ 12 LIST OF ABBREVIATIONS ................................ ................................ ................................ 13 ABSTRA CT ................................ ................................ ................................ ........................... 15 CHAPTER 1 BACKGROUND ................................ ................................ ................................ ............. 17 Epidemiology of Cannabis Use ................................ ................................ ........................ 17 Results from the 2015 National Survey on Drug Use and Health ................................ 17 Results from the 2015 Monitoring the Future Survey ................................ .................. 18 Results from the National Epidemiologic Survey on Alcohol and Related Conditions III ................................ ................................ ................................ ........ 19 Risk Perception of Cannabis ................................ ................................ ...................... 21 Epidemiology of Coronary Disease and Coronary Disease Risk Factors ............................. 22 Association between Cannabis Use and Coronary Disease ................................ ................. 27 Possible Direct Mechanisms for the Association between Cannabis Use and Coronary Disease ................................ ................................ ................................ .. 28 Possible Indirect Mechanisms for the Association between Cannabis Use and Coronary Disease ................................ ................................ ................................ .. 29 Other Substance Use and Mental Health Condit ions ................................ ................... 30 Public Health Significance ................................ ................................ ............................... 31 2 THE ASSOCIATION BETWEEN CANNABIS USE AND CORONARY DISEASE AND CORONARY DISEASE RISK FACTORS: A SYSTEMATIC REVIEW ................. 35 Introduction ................................ ................................ ................................ ..................... 35 Methods ................................ ................................ ................................ .......................... 38 Sources of Evidence ................................ ................................ ................................ .. 38 Study design ................................ ................................ ................................ ...... 38 Criteria for study selection ................................ ................................ .................. 39 Data Collection and Analysis ................................ ................................ ..................... 39 Search strategy ................................ ................................ ................................ ... 39 Study selection ................................ ................................ ................................ ... 39 Data extraction ................................ ................................ ................................ ... 40 Assessment o f risk of bias ................................ ................................ ................... 40 Synthesis of results ................................ ................................ ............................. 41 Results ................................ ................................ ................................ ............................ 41 Included Study Summaries ................................ ................................ ........................ 41 Stud y Results by Outcome ................................ ................................ ......................... 44 BMI ................................ ................................ ................................ ................... 45

PAGE 6

6 Hypertension ................................ ................................ ................................ ...... 45 Dyslipidemia ................................ ................................ ................................ ...... 46 Diabetes ................................ ................................ ................................ ............. 46 Metabolic Syndrome ................................ ................................ .......................... 47 Myocardial Infarction ................................ ................................ ......................... 47 Angina ................................ ................................ ................................ ............... 47 Risk of Bias ................................ ................................ ................................ .............. 48 Discussion ................................ ................................ ................................ ....................... 49 3 GENERAL METHODS ................................ ................................ ................................ ... 77 HealthStreet Model ................................ ................................ ................................ .......... 77 Recruitment and Sampling ................................ ................................ ............................... 77 Health Intake Form ................................ ................................ ................................ .......... 78 Specific Aims ................................ ................................ ................................ .................. 78 Measures ................................ ................................ ................................ ......................... 78 Cannabis Use ................................ ................................ ................................ ............ 78 Coronary Disease ................................ ................................ ................................ ...... 79 Covariates ................................ ................................ ................................ ................. 79 Socio demographics ................................ ................................ ........................... 79 Coronary disease risk factors ................................ ................................ .............. 80 Other substance use ................................ ................................ ............................ 80 Mental health conditions ................................ ................................ ..................... 81 Methods of Analysis ................................ ................................ ................................ ........ 81 Univaria te Analyses ................................ ................................ ................................ .. 82 Multiple Logistic Regression ................................ ................................ ..................... 82 Identifying relevant covariates ................................ ................................ ............ 82 Multicollinearity ................................ ................................ ................................ 83 Assessing confounding ................................ ................................ ....................... 83 Model building ................................ ................................ ................................ ... 84 Model fit ................................ ................................ ................................ ............ 84 Latent Class Analysis ................................ ................................ ................................ 85 Multinomial Logistic Regression ................................ ................................ ............... 85 4 THE ASSOCIATION BETWEEN CANNABIS USE AND CORONARY DISEASE ........ 92 Introduction ................................ ................................ ................................ ..................... 92 Methods ................................ ................................ ................................ .......................... 96 Sample ................................ ................................ ................................ ..................... 96 Measures ................................ ................................ ................................ .................. 96 Exposure ................................ ................................ ................................ ............ 9 6 Outcome ................................ ................................ ................................ ............ 97 Covariates ................................ ................................ ................................ .......... 97 Analysis ................................ ................................ ................................ ................... 99 Results ................................ ................................ ................................ .......................... 100 Discussion ................................ ................................ ................................ ..................... 103

PAGE 7

7 5 ASSOCIATION BETWEEN CANNABIS USE AND CORONARY DISEASE IN THE CONTEXT OF OTHER SUBSTANCE USE: A LATENT CLASS ANALYSIS .............. 113 Introduction ................................ ................................ ................................ ................... 113 Methods ................................ ................................ ................................ ........................ 116 Sample ................................ ................................ ................................ ................... 116 Measures ................................ ................................ ................................ ................ 117 Manifest variables ................................ ................................ ............................ 117 Coronary disease ................................ ................................ .............................. 117 Covariates ................................ ................................ ................................ ........ 117 Statistical Analysis ................................ ................................ ................................ .. 118 Results ................................ ................................ ................................ .......................... 120 Class Membership ................................ ................................ ................................ ... 121 Descriptio n of Total Sample ................................ ................................ .................... 121 Description of Class Membership ................................ ................................ ............ 122 Multinomial Logistic Regression ................................ ................................ ............. 123 Discussion ................................ ................................ ................................ ..................... 124 6 CONCLUSIONS ................................ ................................ ................................ ........... 133 Main Findings ................................ ................................ ................................ ............... 133 Mental Health Conditions ................................ ................................ .............................. 137 St rengths ................................ ................................ ................................ ....................... 138 Limitations ................................ ................................ ................................ .................... 139 Public Health Consequence & Future Research ................................ ............................... 140 APPENDIX A SYSTEMATIC REVIEW PROTOCOL ................................ ................................ .......... 145 Title ................................ ................................ ................................ .............................. 145 Date of Protocol ................................ ................................ ................................ ............. 145 Introduction ................................ ................................ ................................ ................... 145 Rationale ................................ ................................ ................................ ................ 145 Objectives ................................ ................................ ................................ ............... 145 Methods ................................ ................................ ................................ ........................ 145 Eligibility Criteria ................................ ................................ ................................ ... 145 Types of studies ................................ ................................ ............................... 145 Population studied and language of publication ................................ ................. 145 Expo sure and outcome measurement ................................ ................................ 146 Search Methods and Information Sources ................................ ................................ 147 Electronic searching ................................ ................................ ......................... 147 Pubmed /Medline ................................ ................................ .............................. 147 Web of Science ................................ ................................ ................................ 147 Data Collection and Analysis ................................ ................................ ......................... 147 Study Selection ................................ ................................ ................................ ....... 147 Data Extraction and Management ................................ ................................ ............ 148

PAGE 8

8 Assessment of Risk Bias ................................ ................................ ......................... 148 Data Synthesis ................................ ................................ ................................ ........ 148 B SYSTEMATIC REVIEW DATA ABSTRACTION TABLES ................................ ......... 152 Ross, 2016 ................................ ................................ ................................ ..................... 152 Yankey, B 2016 ................................ ................................ ................................ ............. 154 Waterreus, 2016 ................................ ................................ ................................ ............. 157 Vidot, 2016 ................................ ................................ ................................ ................... 160 Blackstone, 2016 ................................ ................................ ................................ ........... 162 Bancks, 2015 ................................ ................................ ................................ ................. 164 Racine, 2015 ................................ ................................ ................................ .................. 166 Dube, 2015 ................................ ................................ ................................ .................... 169 Thompson, 2015 ................................ ................................ ................................ ............ 171 Penner, 2013 ................................ ................................ ................................ .................. 173 Rajavashisth, 2012 ................................ ................................ ................................ ......... 175 Hayatbak hsh, 2010 ................................ ................................ ................................ ........ 177 Rodondi, 2006 ................................ ................................ ................................ ............... 179 Mittleman, 2001 ................................ ................................ ................................ ............ 181 Huang, 2013 ................................ ................................ ................................ .................. 183 Aronow, 1974 ................................ ................................ ................................ ................ 185 LIST OF REFERENCES ................................ ................................ ................................ ...... 186 BIOGRAPHICAL SKETCH ................................ ................................ ................................ 206

PAGE 9

9 LIST OF TABLES Table page 1 1 2015 prevalence estimates of cannabis use by national study. ................................ ........ 33 2 1 Inclusion criteria for studies in the systematic review ................................ ................... 55 2 2 Exclusion criteria for studies in the systematic review ................................ .................. 55 2 3 Overview of the studies included in the systematic review ................................ ............ 57 2 4 Lead author, mean sample age, cannabis use measure, BMI measure, and main results for all included studies with BMI as an outcome ................................ ................ 60 2 5 Lead author, mean sample age, cannabis use measure, hypertension measure, and main results for all included studies with hypertension as an outcome ........................... 62 2 6 Lead author, mean sample age, cannabis use measure, dyslipidemia measure, and main results for all included studies with dyslipidemia as an outcome ........................... 65 2 7 Lead author, mean sample age, cannabis use measure, diabetes measure, and main results for all included studies with diabetes as an outcome ................................ ........... 69 2 8 Lead author, mean sample age, cannabis use measure, metabolic syndrome measure, and main results for all included studies with metabolic syndrome as an outcome .......... 71 2 9 Lead author, mean sample age, cannabis use measure, myocardial infarction measure, and main results for all included studies with myocardial infarction included as an outcome ................................ ................................ ................................ 73 2 10 Lead author, mean sample age, cannabis use measure, angina measure, and main results for all included studies with angina included as an outcome ............................... 73 2 11 Risk of bias classifications for each study included in the review by category of possible bias ................................ ................................ ................................ ................ 74 3 1 HealthStreet Intake Form questions used in analyses ................................ .................... 88 3 2 HealthStreet Intake Form question number, variable name, and coding strategy used in analyses ................................ ................................ ................................ .................. 90 4 1 Socio demographic characteristics and physiological risk factors for coronary disease in the total HealthStreet sample, and by lifetime self reported history of coronary disease ................................ ................................ ................................ ........ 110 4 2 Cannabis use, other substance use and history of mental health conditions in the total HealthStreet sample, and by lifetime self reported history of coronary disease ............ 111

PAGE 10

10 4 3 Odds ratios and 95% confidence intervals for the associations between cannabis use and coronary disease ................................ ................................ ................................ 112 5 1 Percentage of HealthStreet members positively endorsing lifetime substance use ........ 129 5 2 Item response probabilities ................................ ................................ ........................ 129 5 3 Socio demographics, lifetime history of mental health conditions, and coronary disease risk factors for the entire sample and by latent class ................................ ........ 131 5 4 Odds ratios and 95% confidence intervals for the association between lifetime history of coronary disease and latent class membership. ................................ ............ 132 6 1 Summary of state of the research or future research needed to follow Bradford Hill criteria. ................................ ................................ ................................ ..................... 144 A 1 Data to be extracted from each identified study ................................ .......................... 149 A 2 Domains considered in risk of bias assessments ................................ .......................... 150 A 3 Assessment of bias forms ................................ ................................ .......................... 151 B 1 Data abstraction for Ross, 2016 ................................ ................................ ................. 15 2 B 2 Risk of bias judgment for Ross, 2016 ................................ ................................ ......... 153 B 3 Data abstraction for Yankey, 2016 ................................ ................................ ............. 154 B 4 Risk of bias judgment for Yankey, 2016 ................................ ................................ ..... 156 B 5 Data abstraction for Waterreus, 2016 ................................ ................................ ......... 157 B 6 Risk of bias judgment for Waterr eus, 2016 ................................ ................................ 159 B 5 Data abstraction for Vidot, 2016 ................................ ................................ ................ 160 B 8 Risk of bias judgment for Vidot, 2016 ................................ ................................ ........ 161 B 9 Data abstraction for Blackstone, 2016 ................................ ................................ ........ 162 B 10 Risk of bias judgment for Blackstone, 2016 ................................ ................................ 163 B 11 Data abstraction for Bancks, 2015 ................................ ................................ .............. 164 B 12 Risk of bias judgment for Bancks, 2015 ................................ ................................ ..... 165 B 13 Data abstraction for Racine, 2015 ................................ ................................ .............. 166 B 14 Risk of bias judgment for Racine, 2015 ................................ ................................ ...... 168

PAGE 11

11 B 15 Data abstraction for Dube, 2015. ................................ ................................ ................ 169 B 16 Risk of bias judgment for Dube, 2015 ................................ ................................ ........ 170 B 17 Data abstraction table for Thompson, 2015 ................................ ................................ 171 B 18 Risk of bias judgment for Thompson, 2015 ................................ ................................ 172 B 19 Data abstraction for Penner, 2013 ................................ ................................ .............. 173 B 20 Risk of bias judgment for Penner, 2013 ................................ ................................ ...... 174 B 21 Data abstraction for Rajavashisth, 2012 ................................ ................................ ...... 175 B 22 Risk of bias judgment for Rajavashisth, 2012 ................................ ............................. 176 B 23 Data abstraction table for Hayatbakhsh, 2010 ................................ ............................. 177 B 24 Risk of bias judgment for Hayatbakhsh, 2010 ................................ ............................. 178 B 25 Data abstraction for Rodondi, 2006 ................................ ................................ ............ 179 B 26 Risk of bias judgment for R odondi, 2006 ................................ ................................ ... 180 B 27 Data abstraction for Mittleman, 2001 ................................ ................................ ......... 181 B 28 Risk of bias judgment for Mittleman, 2001 ................................ ................................ 182 B 29 Data abstraction for Huang, 2013 ................................ ................................ ............... 183 B 30 Risk of bias judgment for Huang, 2013 ................................ ................................ ...... 184 B 31 Data abstraction for Aronow, 1974 ................................ ................................ ............ 185 B 32 Risk of bias judgment for Aronow, 1974 ................................ ................................ .... 185

PAGE 12

12 LIST OF FIGURES Figure page 1 1 Relationship between selected coronary disease risk factors and coronary disease ........ 34 2 1 Systematic review flow log ................................ ................................ .......................... 56 2 2 Model of the genesis of coronary disease with exposure, outcomes, and direction of statistically significant associations from this literature review shown in red ................. 76 3 1 Infographic depicting the HealthStreet mission and aims to community members interested in becoming HealthStreet members ................................ ............................. 87 5 1 Latent class membership by percentage of members with lifetime use of each type of substance ................................ ................................ ................................ .................. 130 6 1 Directions for future research in the context of the etiology of corona ry disease .......... 143

PAGE 13

13 LIST OF ABBREVIATIONS ABM Agent Based Modeling BIC Bayesian Information Criterion BMI Body Mass Index CHW Community Health Worker CI Confidence Interval CAD Coronary Artery Disease DBP Diastolic Blood Pressure DSM IV Diagnostic and Statistical Manual of Mental Disorders, 4 th Edition DSM 5 Diagnostic and Statistical Manual of Mental Disorders, 5 th Edition CB1 Endocannabinoid Receptor 1 CB2 Endocannabinoid Receptor 2 HDL C High Density Lipoprotein Cholesterol HbA1c Hemoglobin A1c ICD 9 International Classification of Disease 9 ICD 10 International Classification of Disease 10 LCA Latent Class Analysis LDL C Low Density Lipoprotein Cholesterol MUSP Mater University Study of Pregnancy and its Outcomes MI Myocardial Infarction; commonly referred to as heart attack MTF Monitoring the Future Survey NESARC III National Epidemiologic Survey on Alcohol and Related Conditions III NHANES National Health and Nutrition Examination Survey NSDUH National Survey on Drug Use and Health NAU Navigation as Usual OR Odds Ratio RR Relative Risk SAM HSA Substance Abuse and Mental Health Services Administration SBP Systolic Blood Pressure

PAGE 14

14 SLR Systematic Literature Review UF CTSI University of Florida Clinical and Translational Science Institute US A United States of America VIF Variance Inflation Factor VLMR Vuong Lo Mendell Rubin THC tetrahydrocannabinol; the primary psychoactive constituent of cannabis

PAGE 15

15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CANNABIS USE AND CORONARY DISEASE By Hannah Renee Crooke May 2017 Chair: Catherine W Striley Major: Epidemiology About 4 4 % of Americans report lifetime cannabis use and a bout 60% of Americans s upport legalization I dentifying negative health outcomes related to cannabis use such as coronary disease is of great public health importance. T he association between cannabis use and coronary disease was explored through a SLR and secondary data analyses examining the 1) direct association between cannabis use and coronary disease and 2) association between latent classes of subst ance use, including cannabis use, and coronary disease. T he SLR (C hapter 2) employed a st ructured electronic search resulting in s ixteen articles identified for review Second ary data analyses used data from HealthStreet, a community engagement model T he association between cannabis use and coronary diseas e was investigated through m ultiple logistic regression ( C hapter 4 ) and latent class analys i s with multinomial logistic regression (C hapter 5) The results of the SLR were equivocal Both positive and negative associations between cannabis use and obesity, hypertension, dyslipidemia, and metabolic syndrome were found N egative associations between cannabis use and diabetes and positive associations between

PAGE 16

16 cannabis use and angina and MI were identified M ore research focused on the association between cannabis use and coronary disease is needed Analyses revealed a significant relationship between cannabis use and coronary disease that was attenuated after controlling for other substance use, mental health conditions and coronary disease risk factors R esults suggest any direct association between cannabis use and coronary disease may be the result of confounding The second analysis revealed three latent classes of substance u ser : 1) m on o substance use and abstainer 2) cannabis use and tobacco use, and 3) polysubstance use The association between coronary disease and class membership was attenuated in fully adjusted models W hen c onsidering the association between cannabis use and coronary disease in the context of other substance use, other confounding factors such as mental health disorders cannot be ignored. This dissertation fills a gap in the literature, and contributes to the growing public health interest in outcomes related to cannab is use.

PAGE 17

17 CHAPTER 1 BACKGROUND Epidemiology of Cannabis Use Cannabis is a generic term used to denote multiple psychoactive preparations of the Cannabis sativa plant. However, in some countries, includ ing the United States, the term (World Hea lth Organization, 2017) Cannabis is the most commonly used federally cont rolled, and therefore federally illicit drug in the United States (Bose et al., 2016; Farmer et al., 2015) Use of most illici t drugs in the United States has either stabilized or decreased in the past decade, but cannabis use has increased and is the primary contributing factor to the overal l increase in illicit drug use (Bose et al., 2016) Since 2002, there has been a three fold increase in cannabis use among adults aged 55 to 59 years old (Vidot et al., 2014) Table 1 1 gives an overview of the nationwide prevalence rates of cannabis use based on the results of the three national surveys outlined below. Results from the 2015 National Survey on Drug Use and Health The 2015 National Survey on Drug Use and Health (NSDUH) sponsored by the Substance Abuse and Mental Health Services Administration (SAMHSA) within the U.S. Department of Health and Human Services (HHS), is a nationally representative survey with the sampling frame covering the civilian, non institutionalized population age 12 and older residing within the United States The survey uses a multist age area probability sample designed to be representative of the United States and each of the 50 states and the District of Columbia. Within each state, state samplin g regions were formed to yield roughly the same number of interviews per region. Census tracts within the state sampling regions constituted the primary sampling units If necessary, census tracts were aggregated to comprise 100 dwelling units in rural are as or 150 dwelling units in urban areas at minimum. In 201 5, 132,210 people were screen ed and

PAGE 18

18 68,073 interviews completed. The overall response rate for those 12 and older was 55.2% (Bose et al., 2016) Questions used to assess lifetime, past year, an d past 30 day cannabis use were last used marijuana or hashish? [within the past 30 days, more than 30 days ago but within the past 12 months or more than 12 months ago] The 2015 survey estimate d the lifetime prevalence rate of cannabis use among those age 1 2 and older at 44.0 % the past year prevalence rate of cannabis use at 13.5% and past 30 day prevalence rate of cannabis use at 8.3% (Center for Behavioral Health Statistics and Quality, 2016) The 2015 NSDUH also found that males age 12 and older are more likely to be lifetime cannabis users compared to females (48.9% vs 39.4%). Among lifetime cannabis users, males were also more likely to be current (past 30 day) users than females (10.6% vs 6.2%). Additi onally, Non Hispanic Whites had a higher prevalence of lifetime cannabis use compared to Blacks or African Americans (49.1 % vs 43.3% ) However, Non Hispanic Whites have a lower prevalence of current use compared to Blacks or African Americans (8.4% vs 10.7%) (Center for Behavioral Health Statistics and Quality, 2016) Results from the 2015 Monitoring the Future Su rvey Monitoring the Future (MTF) is a long term, national survey of American adolescents, college students, and high school graduates through age 55. The survey has been conducted l Research. The 2015 MTF survey included approximately 44,900 8 th 10 th and 12 th grade students from 382 secondary schools throughout the United States (Johnston et al., 2016b) A multistage random sampling procedure generate s a nationally representative sample each year at each grade level. The sample is selected by geographic area, the n by school, and final ly by specific class The

PAGE 19

19 survey is administered during a normal class period whenever possible. For a selected sample from each senior class, follow up surveys are mailed to their home addresses biannually every year since graduation. Data from follow up su rveys are presented by modal age. For example, the graduating class of 2015, surveyed in 2015, represents modal age 18 while the graduating class of 2002, surveyed in 2015, represents modal age 30. The survey inclu des the following questions to assess cannabis use: 1) 2) marijuana (weed, pot) or hashish (ha 3) many occasions (if any) have you used marijuana (weed, pot) or hashish (hash, hash oil) during for each question were : 0 occasions, 1 2 occasions, 3 5, occasion s, 6 9 occasions, 10 19 occasions, 20 39 occasions, and 40 or more occasions. Data from the 2015 MTF indicate that the lifetime prevalence rate of cannabis use among those modal age 18 is 45% and the lifetime prevalence rate of cannabis use among those mo dal age 55 is 81% (Johnston et al., 2016a) The l ifetime prevalence rate of cannabis use for combine d 8 th 10 th and 12 th gra ders was 30.0%, past year (12 months) prevalence rate was 23.7%, and past 30 day prevalence rate was 14.0% (Johnston et al., 2016b) Males in all age groups surveyed were more li kely to report lifetime cannabis use compared to females (Johnston et al., 2016b, 2016b) Results from the National Epidemiologic Survey on Alcohol and Related Conditions III The National Epidemiologic Surve y on Alcohol and Related Conditions III (NESARC III) is the fourth and most recent national survey done by the National Institute on Alcohol Abuse and Alcoholism (NIAAA). NESARC III is a cross sectional, nationally representative survey based on the civili an, non institutionalized population within the United States sponsored by the National Institute on Alcohol Use and Alcoholism. Information on alcohol and drug use

PAGE 20

20 disorders for 36,309 individuals ages 18 and older Use Disorder and Associated Disabilities Interview Schedule (AUDADIS 5) during the years 2012 2013 NESARC III employed a multistage probability sampling strategy with primary sampling units as counties or groups of contiguous counties, secondary sampling units as census defined blocks, and tertiary sampling units as households. The overall response rate for those years w as 60.1% Data are weighted to represent the US civilian population. The survey used na, weed, pot, answer choice s T he national prevalence rate of past year cannabis use based on NESARC III data was estimated at 9.5% (Hasin et al., 2015) Hasin and colleagues found the p ast year prevalence r ate of DSM IV (American Psychiatric Association, 1994 ) cannabis abuse or dependence among the entire sample was 2.9% Among only past year cannabis users, prevalence of DSM IV cannabis use disorder was 30.6% (Hasin et al., 2015) NESARC III, in agreement with NSDUH and MTF found that c annabis use was more prevalent in men compared to women (12.3% vs 6.9%) (Hasin et al., 2015) The Diagnostic and Statistic al Manual (DSM) 5 presents the most recent diagnostic guidelines for cannabis use disorder, however, there is limited information on prevalence of cannabis use disorder using these guidelines. Hasin and colleagues used DSM IV cannabis abuse which requires endorsement of one or more of four criteria in a 12 month period and DSM IV dependence which requires endorsement of three or more of seven criteria in a 12 month per iod Contrastingly, DSM 5 includes only one cannabis use disorder ( with abuse being eliminated ), which requires endorsement of at least two of 11 criteria in a 12 month period, with the ability to specify mild, moderate or severe use disorder based on numb er of criteria endorsed

PAGE 21

21 (American Psychiatric Association, 2013, 1994) Specific changes in the criteria between the edition IV and 5 include dropping l egal problems d ue to use (formerly for abuse) and adding craving (an ICD 10 criterion) (American Psychiatric Association, 2013) Risk Perception of Cannabis P ublic opinion surrounding cannabis use o ver the past 15 years has changed as evidenced by the legalization of both medical and recreational cannabis use in numerous states (Davis et al., 2015) Currently, 29 states and the District of Columbia have passed laws legalizing some form of cannabis use, with 8 states and the District of Columbia hav ing legalized the recreational use of cannabis (MacCoun, 2017) Despite state level changes, the federal Drug Enforcement Agency in conjunction with the Food and Drug Administration decided i n 2016 not to remove c annabis from its listing as a Schedule I drug (Department of Justice, 2016) However, f ederal prosecutors have been instructed not to spend time prosecuting those who use or sell cannabis when legal by state law (Ogden, 2009) With large amounts of taxable profit in this legal climate, large medical and recreational cannabis dispensaries have grown acr oss the nation resulting in increased access to cannabis While many states were leg alizing cannabis use, bis use was changing. By October, 2016, nationwide support for legalization of cannabis reached 60%, an all time high (Swift, 2016) C oloradans for example, perceived a lower risk of cannabis use and increased (Schuermeyer et al., 2014) R egardless of state of reside nce, users perceive cannabis to be less likely to cause social and academic harm compared to the perceptions of never users (Kilmer et al., 2007) Data from the NSDUH indicate a significant decrease in perceived great risk of cannabi s use, and a significant increase in d aily cannabis use from 2002 2012 (Pacek et al.,

PAGE 22

22 2015) D ata from MTF show that among 12 th graders, the percentage of those who disapprove d of those who try cannabis once or twice decreased almost 10 percentage points fro m 2005 to 2015 (55.0% vs 45.5%). T he percentage of those who disapprove d of those using ca nnabis regularly fell almost 12 percentage points in that same time frame (82.0% to 70.7%) (Johnston et al., 2016b) The national increase in people who use cannabis coupled with their decreased risk perception, suggests cannabis use will continue to rise. Epidemiology of Coronary Disease and Coronary Disease Risk Factors The continued increase in cannabis use has led researchers and policy makers to question if and how cannabis use is associated with chronic conditions that have high levels of morbid ity and mortality such as coronary disease Coronary disease is the leading cause of death in the United States (Heron, 2015) A ngina, coronary artery disease, and myocardial infarction (MI) are all conditions covered by coronary disease These conditions are not only prevalent, that also affect minority groups with a higher disease burden. The Heart Disease and Stro ke Statistics 2016 Upd ate reports that c oronary disease is most prevalent in African American s, and, among African American s, more prevalent in males than females (Mozaffarian et al., 2015) Coronary disease is also linked to low er socio economic status, including specifically low education l evel (Eng et al., 2002; Galea et al., 2011) Thi s highly prevalent disease has a significant negative impact on the national economy. Corona ry disease is responsible for 14 % of the US national health expenditures in 2012 and 2013, and total costs related to coronary disease are expected to triple to $ 30 0 billion by the year 2030 (Benjamin et al., 2017; Heidenreich et al., 2011; Mozaffarian et al., 2015) Figure 1 1, adapted from Pearson et al (1993) gives the rela tionship between coronary disease risk factors and coronary disease. There are three different types of risk factors listed in

PAGE 23

23 the model; non modifiable, behavioral, and physiological. Non modifiable risk factors listed in the figure include age, sex, and family history. It is well documented that coronary disease risk increases with age (Labarthe, 2011) Additionally, risk of coronary disease differs by sex, with a compared to men (Labarthe, 2011) Unsurprisingly results from the 2009 to 2012 NHANES indicate that prevalence of coronary heart disease is least among those ages 20 39, and greatest among those age 80+. Additionally, the results indicate the same prevalence of coronary heart disease for men and women between the ages of 20 39 ( 0.6% of the population) B etween the ages of 40 to 59, the ages of 60 to 79 and beyond age 80+ prevalence is higher for men compared to women (6.3% vs 5.6%, 19.9% vs 9.7%, and 32.2% vs 18.8% respectively) (Mozaffarian et al., 2015) Family history of coronary disease represents the contributions of genetic and ecologic factors Family history of coronary disease explain s a sign ificant portion of prevalent and future coronary disease (Kardia et al., 2003; Labarthe, 2011) Behavioral risk factors for coronary disease listed in Figure 1 1 include sedentary lifestyle, diet or dietary imbalanc e, binge drinking, tobacco smoking, and cocaine use. About 20 years ago, a report by the Surgeon General of the United States was released that summarized, in part, 7 studies of physical activity and coronary disease (U S Department of Health and Human Services, 1996) Despite being published nearly 2 decades ago, presented clear evidence of an association between sedentary lifestyle and coronary disease that holds today (Benjamin et al., 2017; Labarthe, 2011) In fact, the literatu re indicates a dose response relationship between increased physical activity and lower risk of coronary disease among both men and women (Benjamin et al., 2017; Carnethon, 2009) In addition to physical inactivity, t he importance of diet in relation to coronary disease has also long been established

PAGE 24

24 (Walker, 1969) The most recent iteration of the Heart Disease and Stroke Statistics Report from the American Heart Association review ed a cohort s t udy of 72,113 US female nurses It found a 28% lower risk of coronary disease related mortality among those with a diet high in vegetables, fruits, legumes, fish, poultry, and whole grains (Benjamin et al., 2017; Heide mann et al., 2008) Similar findings were reported in a number of other cohort studies reviewed in the Heart Disease and Stroke Statistics Report (Benjamin et al., 2017; Brunner et al., 2 008; Fitzgerald et al., 2012; Osler et al., 2001) The remaining behavioral risk factors associated with coronary disease listed in Figure 1 1 are binge drinking, tobacco smoking, and cocaine use all of which have been associated with coronary disease (Chiva Blanch et al., 2013; De Giorgi et al., 2012; Dodani et al., 2016; McBride, 1992; Thylstrup et al., 2015) While low to moderate alcohol consumption has been associated with lower risk of coronary disease, binge drinking, defined as four or more drinks in a single day for men and three or more drinks in a single day for women, has been found to incr ease risk of coronary disease. The J shaped curve sho wing lower risk of coronary disease for low and moderate alcohol use, and high risk for binge drinking has been found in multiple studies (Thompson, 2013) Tobacco use rep resents a major risk factor for coronary disease, and never having smoked or having quit smoking >12 months ago is listed by the American Heart Association as one of the ideal components of cardiovascular health (Benjamin et al., 2017; Lloyd Jones et al., 2010) Smoking is an independent risk factor for coronary disease (Centers for Disease Co ntrol and Prevention 2010) and a recent meta analysis showed that female smokers have a 25% gr eater risk of coronary disease compared to male smokers (Huxley and Woodward 2011) A dditionally, a systematic review of the literature revealed that cocaine use is

PAGE 25

25 associated with increased risk of angina, myocardial infarction, and sudden death from cardiac causes (Lange and Hillis, 2001) Physiological r isk factors for coronary dis ease include d in Figure 1 1 a re obesity hypertension dyslipidemia, diabetes, metabolic syndrome, and depression Obesity is considered a major risk factor coronary disease by the American Heart Association Obesity is classified as having a BMI of greater than 30 kg/m 2 According to NHANES 2013 2014, about 37.7% of US adults were obese (Benjamin et al., 2017) Rates of higher BMI, specifically obesity are greater in women (36%) compared to men (34%), particularly among non Hispanic Black women (58%) and Hispanic women (43%) compa red to non Hispanic white women (33%) (Mozaffarian et al., 2015) Hypertension is currently defined as a blood pressure of 140/90 mmHg in those age 60 or younger, and 150/90 mmHg in those over 60 years old (James et al., 2014 p. 8) Hypertension is considered a major risk factor for coronary disease by the American Heart Association. Despite the current definition of hypertension given above, the AHA considers blood pressure less than 120/80 mmHg in those age 20 and over t o be one of the seven components of ideal cardiovascular health (Benjamin et al., 2017; Chobanian et al., 2003; Lloyd Jones et al., 2010) Based on data from NHANES, t he national prevalence of hyper tension among adults 20 years and older is 32.6%. There is a disproportio nately high prevalence rate of hypertension among Black s (44.9% for men and 46.1% for women ) compared to Non Hispanic Whites (32.9% for men and 32.1% for women) (Mozaffarian et al., 2015) Dyslipidemia defined as elevated total or LDL cholesterol levels, and/or low HDL cholesterol levels is an important risk factor for coronary disease (Fodor, 2015) Prevalence of dyslipidemia in the United States varies by BMI classification, with prevalen ce of dyslipidemia

PAGE 26

26 in obese adults at 49.7%, overweight adults 44.2% and normal weight adults 28.6% (Mozaffarian et al., 2015) It is estimated that about 30.9 million US adults (13.1%) have diagnosed hypercholesterolemia or elevated total cholesterol Mozaffarian and colleagues (2015) also note that 6.2% of US adults are estimated to have undiagnosed hypercholesterolemia. Diabetes a major risk factor for coronary disease, is defined by the American Diabetes Association as having an untreated fasting plasma glucose of 126 mg/dl or higher, or having an HbA1C of 6.5% or higher (American Diabetes Association, 2016) Diabetes affects 1 in 10 US adults, with the nearly all cases being t ype II (90% 95%). Among males, Non Hispanic Blacks have the highest pre valence of physician diagnosed diabetes (14.1%) c ompared to Non Hispanic Whites (8.0%), Hispanics (12.6%), and Non Hispanic Asians (11.8%). Among females, the same pattern held, with Non Hispanic Blacks having the highest prevalence of physician diagnosed diabetes (13.6%) compared to Non Hispanic Whites (7.4%), Hispanics (12.7%), and Non Hispanic Asians (9.1%) (Benjamin et al., 2017) Data from the Framingham Heart Study suggest that the adjusted relative risk for coronary disease is 38% higher for each 10 year increase in duration of diabetes (Goldstein et al., 2011) Metabolic syndrome is a multicomponent risk factor for coronary diseas e, reflect ing the clustering of individual cardiometabolic risk factors for coronary disease (Benjamin et al., 2017) Metabolic syndrome is defined by the American Heart Association as meeting 3 or more of the following criteria: 1) fasting plasma glucose 100 mg/dl, 2) HDL C <40 mg/dL in males or <50 mg/dL in females or undergoing treatment for low HDL C, 3) triglycerides undergoing treatment for elevated triglycerides, 4) waist circumference > 102 cm in males or > 88cm in females, or 5) blood undergoing treatment for hypertension. The age specific prevalence of metabolic syndrome was

PAGE 27

27 20.3% among people age 20 to 39 years, 40.8% among people age 40 to 59 years, and 51.5% for of age according to data from the 2003 to 2006 NHANES (Benjamin et al., 2017) In 2010, the age adjusted prevalence of metabolic syndrome was highest among Mexican Americans (34.8%) compared to Whites (22.9%), and Blacks (19.0%) according to NHANES (Benjamin et al., 2017) Th e last physiologic risk factor for coronary disease covered in F igure 1 1 is depression. L iterature shows a strong association between depression and coronary disease with d epression showing an increase in the risk of cardiovascular disease including coro nary disease by 1.5 3 times in otherwise healthy individuals (Xian et al., 2010; Lett et al., 2004; Baune and Tully, 2016) In 2015, the results of the NSDUH indicated that overall 6.7% of US adults experienced a depressive episode (depression for 2 weeks or more) in the last year. Last 12 month prevalence of depression is greatest amon g those age 18 25 (10.3%), and among females (8.5%) (National Institute on Mental Health, 2016) Association b etween Cannabis Use and Coronary Disease An association between cannabis use and coronary disease is hypothesized in a number of articles; t wo literature r eviews of particular relevance to this dissertation synthesize the results of these articles First, Franz and Frishman published a review article in 2016, giving a comprehensive look at the numerous case reports citing acute coro nary events after cannabis use, thought to be due to coronary arterial vasospasm Second, a report published in January 2017 by the National Academies of Sciences, Engineering, and Medicine reviewed the literature linking cannabis use to acute MI, metabolic syndrome and diabetes. The report found that there was not sufficient evidence to support or to refute the hypothesis of a positive association between cannabis use and acute MI (National Academies of Sciences, Engineering, and Medicine, 2017) The report also found that there was limited evidence supporting a negative association between

PAGE 28

28 cannabis u se and metabolic syndrome and diabetes. Contrastingly, there was limited evidence supporting a positive association between cannabis use and prediabetes (defined as an HbA1C 5.7% and <6.5%, or a fasting blood glucose 100mg/dl and <126mg/dl). To date, 16 epidemiologic studies have been conducted assessing outcomes such as hypertension, elevated BMI (obesity), metabolic syndrome, and cardiovascular mortality associated with cannabis use, with equivocal results These studies are reviewed systematically in C hapter 2. Possible Direct Mechanisms for the Association between Cannabis Use and Coronary Disease While there is limited epidemiologic evidence, n umerous researchers have suggested a possible direct relationship between cannabis use and coronary disease using biologic and animal model studies Both endocannabinoids ( ligands to cannabinoid receptors produced naturally in the body ) and phytocannabinoids ( produced by the marijuana plant) bind to, and can activate the cannabinoid receptors (CB1 and CB2) in the endocannabinoid system (Gordon et al., 2013) The system as a whole helps to regulate many functions, including appetite, stress response, immune function and sleep (Hall and Degenhardt, 2009) THC; the primary psychoactive component of cannabis) (Pagotto et al., 2006) appears especially important in regulati ng cardiovascular changes associated with cannabis use. When the CB1 receptor is activated, atherosclerotic mediators are released that interfere with vasodilation (Montecucco and Di Marzo, 2012; Rajesh et al., 2010) Additionally, patients with coronary artery disease show increased endocannabinoid s ystem activation, and CB1 receptor blockade show s anti inflammatory effects within the cardiovascular system (Sugamura et al., 2009) Cannabis induces production of catecholamines (Gordon et al., 2013; Jones, 2002) Their release (including epinephrine, norepinephrine, and dopamine) from the adrenal gland activates

PAGE 29

29 platelet formation (Musselman et al., 1998) which induce s clotting. Clotting can lead to thrombotic events, including myocardial infarctions. Previous research has confirmed that cannabis use is associated with increased platelet activation (Dahdouh et al., 2011; Thomas et al., 2014) As mentioned previously, t he endocannabinoid system has both CB1 receptors and CB2 receptors While activation of the CB1 receptor appears to cause harmful changes to the cardiovascular system, activation of the CB2 receptor may help prevent harmful cardiovascular changes; atherosc lerotic inflammation is reduced with CB2 receptor activation (Montecucco and Di Marzo, 2012) CB2 receptor activation has also been sh own to reduce pro inflammatory mediators (Toguri et al., 2014) Possible Indirect Mechanisms for the Association between Cannabis Use and Coronary Disease Along with biological reasons for a direct association, there are biological reasons for an indirect association between cannabis use and coronary disease. Cannabis use may lead to weight gain and poor metabolic function which in turn can lead to coronary disease (Nissen et al., 2008; Pi Sunyer et al. 2006; Van Gaal et al., 2005) Activation of CB1 increases appetite, increases motivation to eat palatable food, increases fatty acid synthesis, increases fat storage, decreases glucose induced insulin secretion, decreases satiety, and decreases glucos e uptake in muscle (Cota, 2007; Foltin et al., 1988) E vidence points to increased average caloric intake with chronic cannabis use (Smit and Crespo, 2001) Given the relationship between cannabis use and factors that contribute to obesity, it has been hypothesized that cannabis use may be indirectly associated with coronary disease through an increased risk of obesity and other cardio metabolic risk factors. This hypo thesis is bolstered by research showing a significant decrease in weight, and improved lipid profiles through CB1

PAGE 30

30 receptor antagonism (Cota, 2007; Desprs et al., 2005; Scheen et al., 2006) suggesting that receptor activation may be harmful. Another possible mechanism for an indirect association between cannabis use and coronary disease is the CB1 recep tor mediated analgesic effects of cannabis P articularly evident in chronic users, these effects may result in delayed seeking of medical care for cardiovascular issues. (Gordon et al., 2013; Pacher et al., 2006) Finally, a n association between cannabis use and coronary d isease may be the result of confounding by unhealthy behaviors such as a high calor ic diet, tobacco use, and other drug use independently associated with both cannabis use and coronary disease (Rodond i et al., 2006; Yankey et al., 2016) Other Substance Use and Mental Health Conditions In addition to considering possible direct and indirect mechanisms it is important to consider how other substance use and mental health disorders are related to t he association between cannabis use and coronary disease Cannabis use is often initiated in adolescence (Chen et al., 2016) and seldom used alone Frequently substances such as tobacco, alcohol and cannabis are initiated concurrently (within the same time period) with subsequent initiation of other illicit drugs (Agrawal et al., 2006; Haberstick et al., 2014; Redonnet et al., 2012) Gateway theory strongly suggests that cannabis use can lead to initiation of other drug use (Fergusson et al., 2006; Kandel, 2003; Kandel et al., 2006) Additionally simultaneous use of tobacco and cannabis has been reported in adolescents and adults (Schauer et al., 2017) as well as simultaneous use of tobacco, alcohol, and cocaine (Lange and Hillis, 2001) Given the likelihood of concurrent initiation of cannabis use with other substances that are risk factors for coronary disease, it is important to consider cannabis use in th e context of other substance use.

PAGE 31

31 F igure 1 1 shows depression as a physiologic risk factor for coronary disease. As discussed previously, t here is ample evidence to suggest a strong association between depression and coronary disease (Ariyo et al., 2000; Baune et al., 2012; Carney and Freedland, 2016; Elderon and Whooley, 2013; Frasure Smith N et al., 1993; Gonzlez and Tarr af, 2013; Hare et al., 2013; Joynt et al., 2003) and some evidence suggesting an association between depression and cannabis use (Degenhardt et al., 2001; Feingold et al., 2015; Lev Ran et al., 2013; Manrique Garcia et al., 2012) S tudies indicate an association between cannabis use and anxiety disorders (Buckner et al., 2012; Patton et al., 2002) as well as other mental health conditions such as personality disorders in adults (Copelan d, Rooke, & Swift, 2013) and behavior disorders in young adults and adolescents (Conway et al., 2016) A nxiety and borderline personality disorder are also associated with coronary disease (Powers and Oltmanns, 2013; Thurston et al., 2013) As such, it is important to consider mental health comorbidities of depression and cannabis use. To further emphasize the ne ed to consider comorbid mental health conditions and substance use a 2016 article looking at the results of the 2012 NSDUH found that individuals with a lifetime history of any substance abuse or dependence h ad e 3.4 times the odds of a lifetime history of any mental health condition (Walker and Druss, 2016) Public Health Significance About 60% of the United States population supports legalization of cannabis ( Swift, 2016) and cannabis use across the US continues to rise. Despite this, the re is a gap in the literature (Vidot et al., 2016) L ittle information is available on chronic disease outcomes associated with cannabis use, a highly prevalent modifiable risk factor Additionally, despite the fact that cannabis use is increasing sequelae associated with chron ic cannabis use are just now being considered by researchers. Given the complex genesis of coronary disease, cannabis use may only be associated with a small amount of increased risk, however, the high prevalence of

PAGE 32

32 coronary disease means the attributable risk from cannabis use may potentially be substantial (National Academies of Sciences, Engineering, and Medicine, 2017; Ros e, 2008) T he high morbidity and mortality associated with coronary disease and the biologically plausible association between cannabis use and coronary disease gives ample reason to explore the association between cannabis use and coronary disease usin g epidemiologic techniques.

PAGE 33

33 Table 1 1. 2015 p revalence estimates of cannabis use by national study. National Study Prevalence Type Prevalence Estimate Age Group NSDUH a Lifetime Use 44.0% 12 and older NSDUH Past Year Use 13.5% 12 and older NSDUH Past 30 Day Use 8.3% 12 and older MTF b Lifetime Use 45.0% Modal age 18 MTF Lifetime Use 81.0% Modal age 55 MTF Lifetime Use 30.0% 8 th 10 th and 12 th grade students MFT Past Year Use 23.7% 8 th 10 th and 12 th grade students MTF Past 30 Day Use 14.0% 8 th 10 th and 12 th grade students NESARC III c Past Year Use 9.5% US Adults 18 and older NESARC III Past Year DSM IV C ombined A buse and D ependence 2.9% US Adults 18 and older a National Survey on Drug Use and Health b Monitoring the Future c National Epidemiologic Survey on Alcohol and Related Conditions

PAGE 34

34 *Increased risk for coronary disease with binge drinking d efined as 4 or more drinks in a single day for men, and 3 or more drinks in a single day for women Any tobacco use is a risk for coronary disease (National Heart Lung and Blood Institute, 2016) b Any cocaine use increases risk for coronary disease (Lange and Hillis, 2001; Schwartz et al., 2010) Figure 1 1. Relationship between selected coronary disease risk factors and coronary disease ; a dapted from Pearson et al (1993) Cannabis Use? Sedentary lifestyle Diet Saturated Fat Salt Cholesterol Total energy content Binge drinking* Tobacco smoking Cocaine use b Behavioral Risk Factors Age Sex Family History Non modifiable Risk Factors Obesity (Elevated BMI) Hypertension (Elevated SBP, DBP) Dyslipidemia (Adverse LDL C, HDL C, Total Cholesterol, Triglycerides) Diabe tes (Elevated Plasma Glucose, HbA1c ) Metabolic Syndrome Physiological Risk Factors Coronary Artery Disease Angina Myocardial Infarction Coronary Disease Outcomes Depression

PAGE 35

35 CHAPTER 2 THE ASSOCI ATION BETWEEN CAN N A BIS USE AND CORONARY DISEASE AND CORONARY DISEASE RISK FACTORS: A SYSTEMATIC REVIEW Introduction Cannabis is the most commonly used illicit drug in the United States (Bose et al., 2016; Farmer et al., 2015) C annabis use has increased over the past decade, and is the primary contributing factor to the overall increase in illicit drug use in the United States (Bose et al., 2016) Data from the National Survey on Drug Use and Health (NSDUH) indicate that the lifetime prevalence rate of cannabis use for those age 12 and older is 44.0% (Center for Behavioral Health Statistics a nd Quality, 2016) P ublic opinion in the US surrounding cannabis use is also changing, with growing support for recreational use, medical use, and decriminalization (Bostwick, 2012; Hoffmann and Weber, 2010; Yankey et al., 2016) I n October 2016 the results of a Gallup poll indicated nationwide support for legalized recreational cannabis use was at 60% (Swift, 2016) Despite about 60% of the US population supporting cannabis l egalization there remains little information about the relations hip between cannabis use and its short and long term health effects. I ncreased support for recreational cannabis use in the United States calls for an understanding and documentation of any a ssociations between cannabis use and outcomes with high rates of morbidity and mortality (Yankey et al., 2016) Of particular interest are possible associations between cann abis use and coron ary disease p articularly given that c oronary disease is the leading cause of death in the United States (Heron, 2015) Further, evidence on the association between cannabis and common diseases such as hypertension will be useful in health care decision making and policy making (Thompson and Hay, 2015a) Numerous scientists have suggested a possible association between cannabis use and coronary disease and/or coronary disease risk factors. Cannabis smoking is associated with a

PAGE 36

36 dose dependent increase in resting heart rate of 20% to 100% (Beaconsfield, 1974; Johnson and Domino, 1971; Roth et al., 1973) There is also a net increase in myocardial oxygen demand coupled with a decrease in oxygen supp ly associa ted with cannabis use (Hollister, 1986) Immediately after cannabis use, tachycardia is often seen (Hollister, 1986) Franz and Frishman (2016) wrote a review of the literature looking at the relationship between cannabis use and cardiovascular disease (including stroke) that included animal mode l studi es, case reports and a few epidemiologic studies. Their review found that daily cannabis use increases the annual risk of MI from 1.5% to 3.0% per year, but cardiovascular mortality did not appear to be associated with cannabis use (Franz and Frishman, 2016) There are biological reasons for a possible direc t association The endogenous cannabinoid system is a neuromodulatory system consisting of ca nnabinoid receptors (type 1 and type 2), and endocannabinoids (endogenous ligands for the receptors) (Cota, 2007) The cannabinoid type 1 (CB1) receptor is highly expressed within the central nervous system, in particular in the forebrain, basal ganglia, cerebellum, hippocampus, and cerebral cortex (Cota, 2007; Herkenham et al., 1991; Mailleux et al., 1992) CB1 receptors are also expressed on peripheral nerves including those that innervate the gastrointestinal tract, and on organs such as the liver, adipose tissues, and pancreas (Bensaid et al., 2003; Cota, 2007; Cota et al., 2003; Croci et al., 1998; Osei Hyiaman et al., 2005) A ctivation of CB1 either by endocannabinoids or via phytocannabinoids such as those found in cannabis (specifically, THC) has been hypothesized a s a direct link between cannabis use and coronary disease. Activation of CB1 results in vasodilation (Montecucco and Di Marzo, 2012; Rajesh et al., 2010) It also induces production of catecholamines such as epinephrine and norepinephrine (Gordon et al., 2013; Jones, 2002) and results in increased platelet activation (Dahdouh et al., 2011; Thomas et al., 2014)

PAGE 37

37 Along with biological reasons for a direct association, there are biological reasons for an indirect association betw een cannabis use and coronary disease. Activati on of CB1 also increases appetite, increases motivation to eat palatable food, increases fatty acid synthesis, increases fat storage, decreases glucose indu ced insulin secretion, decreases satiety, and decrea ses glucose uptake in muscle (Cota, 2007) Given this direct relationship between cannabis use and factors that contribute to obesity, it has been hypothesized that cannabis use may be indirectly associated with coronary disease through an increased risk of obesi ty and other cardio metabolic risk factors. This hypothesis is bolstered by research showing a significant decrease in weight, and improved lipid profiles through CB1 receptor antagonism (Cota, 2007; Desprs et al., 2005; Scheen et al., 2006) Despite the hypothesized association betwee n cannabis use and coronary disease and coronary disease risk factors suppor ted by biological evidence, epidemiologic evidence is necessary to ascertain an association between cannabis use and coronary disease/risk factors. As such, the objectives of this qualitative systematic review were to : 1) summarize the current epidemiologic literature on association s between cannabis use and coronary disease and coronary disease risk factors, and 2) make recommendations for clinical practice and the future of resea rch in this field. A number of literature reviews have been previously undertaken on the health effects of cann abis specific to cardiovascular, coronary, and cardiometabolic diseases (Aryana and Williams, 2007; Cota, 2007; Franz and Frishman, 2016; Gordon et al., 2013; National Academies of Sciences, Engineering, and Medicine, 2 017; Sidney, 2002) This review, however, represents one of the first to focus specifically on epidemiologic evidence levels III I while excluding case studies non controlled studies (level IV) and animal model studies (Wang and

PAGE 38

38 Attia, 2010) Additionally, this review focuses not only on the association between cannabis use and coronary disease, but also on the association between cannabis use and coronary disease risk factors (elevated BMI, hypertension, dyslipidemia, metabolic syndrome and diabetes ). The re are 3 primary reasons for including these specific coronary disease risk factors in the review: 1) these are risk factors closely related to the genesis of coronary disease that are downstream from the behavioral and non modifiable risk factors (Figure 1 1) (Alberti et al., 2009; Kirk and Klein, 2009; Mozaffarian et al., 2015) 2) previous research indicates plausible biological r elationships between cannabis use and these coronary disease risk factors, and 3) a cursory review of the literature prior to this systematic review revealed few epidemiologic studies specific to the relationship between cannabis use and coronary disease o nly. Methods To conduct this review the Preferred Reporting Items for Systematic Reviews and Meta Analyses (PRISMA) statement was followed (Moher et al., 2009) PRISMA criteria were the result of a history of efforts to define rigor in health research; epide miologists such as Ken Rothman (1986) and Peter Goldschmidt (1986) along with seminal articles by Mu lrow (1987) and Sacks (1987) helped to define quality systematic review criteria. The complete PRISMA based protocol for this qualitative review is included in Appendix A. Below is a summary of the methods used. Sources of Evidence Study design The study designs considered for inclusion in this review were randomized controlled trials, prospective and retrospective cohort studies, case control studies, and cross sectional studies.

PAGE 39

39 Criteria for study selection Table 2 1 gives the inclusion criteria for consideration in this review in the PICO (Population, Intervention/exposure of interest, Comparator, Outcomes) format. Studies were included in this review if they met the inclusion criteria and fulfilled the study design criteria listed above An additional condition of inclusion was reporting of an effect estimate. Table 2 2 lists t he exclusion criteria. It is important to note that only studies with the pre determined outcomes were included in this review. As noted in Figure 1 1, the genesis of coronary disease is complex, and a review of the association between cannabis use and ea ch possible risk factor for coronary disease including behavioral and non modifiable risk factors is outside the scope of the aims of this review. Data Collection and Analysis Search strategy The search strategy for this review employed an electronic sear ch of the Pubmed/Medline database and Web of Science database. Additionally, a manual search of the references from studies found via electronic searching was undertaken. The specific keyword search strategy is outlined in the study protocol in Appendix A Study selection Titles and abstracts obtained from the electronic literature search were collected and de duplicated. The initial exclusions were made based on keywords (Pubmed/Medline) or study design labels (Web of Science). S cr eening of remaining arti cles continued via title, with artic les containing a relevant title then screened based on relevance of abstract. Articles with relevant abstracts were pulled for full text review. References of articles chosen for inclusion in the review were manually searched for other relevant studies.

PAGE 40

40 Data extraction Data from 16 included studies were extracted: authors, title, year of publication, study design, study period, sample population, sample size, mean sample age, cannabis/marijuana use measure, coronary disease or coronary disease risk factor measure, measure of association and effect size for the association between the exposure and outcome of interest For each included study, the fo rm outlined in T able A 1 was used to collect and organize data. In some instances, required information was not found in the t able. Appendix B contains t he data extracted from included studies. Many studies included informat ion on more than one outcome of interest. S tudies assessing metabolic syndrome tended to also report data on BMI, hypertension, dyslipidemia and diabetes. Studies were counted as measuring a particular outcome if 1) the au thors specifically state that was the mea sured outcome or 2) biologic measures associat ed with the specific outcome were included and adjusted effect estimates for the association between cannabis use and the specific outcome are included For example, Penne r and colleagues (2013) did not suggest they are measur ing diabetes in the text, but did report HbA1C, which is an accepted measure for diagnosing diabetes (American Diabetes Association, 2016) so that study was counted as measuring diabetes as an outcome. Contrastingly, Vidot and colleagues (2016) provide d measures of hypert ension, dyslipidemia, and diabetes as component parts o f metabolic syndrome but they did not present effect estimates for the association between cannabis and hypertension, dyslipidemia, or diabetes; only o utcomes for metabolic syndrome were reported Ass essment of risk of bias Each study was assessed for quality and risk of bias within the context of specific criteria adapted from the Cochrane C ollaborations tool for assessing risk of bias in studies (Higgins and Green, 2011) Risk of bias was assessed for the following criteria: 1) generalizability ( select ion

PAGE 41

41 bias ) 2) classification of exposure ( information bias ) 3) classification of outcome ( information bi as ) 4) confounding, and 5 ) temporality. Specifics for ho w each criterion was assessed are outlined in the review protocol (Appendix A ). For each of t he 16 included studies, criteria. Two reviewers independently rated each included study on each of the 5 risk of bias criteria. There was no disagreement between reviewers, and so no need to arbitrate. Results are reported for each included study (Appendix B ) and are tabulated in the results section of this chapter. Synthesis of results For each included study, extracted data and judgment of risk of bias were summarized in text and tables Results After full text review 15 articles met inclusion criteria The references of those 15 articles were manually searched for other relevant studi es. The manual search revealed one other st udy that was appropriate for inclusion in the review. Figure 2 1 gives the systematic review flow diagram. In total, 16 studies are included in this review with publication dates ranging from 1974 to 2016. Included Study Summaries In total, 10 of the 16 s tudie s were cross sectional studies one included a cross sectional component and a prospective coh ort component, two were solely prospective cohort studies one study was a case crossover study and one was a crossover study Table 2 3 gives the name of the lead author, year of publication, journal of publication, study design, study period, data set, sample size, and outcomes of interest measured for the 16 included studies.

PAGE 42

42 A 2016 cross sectional study of 238 adolescents of mean age 15.62 in Miami/Dade County, FL conducted by Ross and colleagues on the association between BMI and cannabis use. A 2016 cross sectional study by Yankey and colleagues examine d the relationship between marijuana use and metabolic syndrome using data from the National Health and Nutrition Examination Survey for the years 2011 2012. The study sample consisted of 3,051 participants. P ublished by Waterr eus and colleagues in 2016 us ing data from the Australian based Study of High Impact Psychosis this cross sectional study (n=1,813) aimed to estimate the association between none, occasional, and frequent cannabis use and metabolic syndrome, dyslipidemia (elevated triglycerides and e levated HDL), diabetes (elevated fasting glucose), and hypertension among indivi duals with psychotic disorders. Published in 2016, the fourth cross sectional study included in this review by Vidot and colleagues, used data from the 2005 2010 National Heal th and Nutrition Examination Survey (NHANES) to examine the association between never, past and current cannabis use and metabolic syndrome among 8,478 individuals Blackstone and colleagues published a cross sectional study in 2016 assessing the associat ion between lifetime marijuana use (dichotomous Y/N) and self reported BMI among 10,925 United States adolescents in grades 6 10 using data from the Health Behavior in School Aged Children Series. Using data from the Coronary Artery Risk Development in Yo ung Adults (CARDIA) study, Bancks and colleagues (2015) used both a cross sectional and prospective cohort design to understand the association between cannabis use and diabetes in a sample of 3,151 individuals

PAGE 43

43 In 2015, Racine and colleagues published th e results of a cross sectional study of 100 Black patients (mean age 46.3 years) from a family practice clinic at University Hospital of Brooklyn, NY assessing the association between never, former, and current (past 180 days) cannabis use and BMI, hyperte nsion (systolic blood pressure and diastolic blood pressure), and dyslipidemia. Representing one of the few prospective cohort studies found for this review, a 2015 study by Dube and colleagues used data from 590 individuals included in the Nicotine Depend ence in Teens study (Montreal, Canada) to estimate the association between frequency of cannabis use in the past year and BMI (mean change in BMI from age 17 24). Using data from the 2005 2012 NHANES, Thompson and colleagues assessed the association betwe en never, past and current cannabis use and hypertension (SBP and DBP), BMI, diabetes and dyslipidemia (triglycerides and H DL C) among 6,281 individuals Another study that used data from NHANES (2005 2010) was conducted by Penner and colleagues, and publ ished in 2013. This study also examined the association between never, past, and current cannabis use and BMI, hypertension (systolic blood pressure and diastolic blood pressure) and dyslipidemia among 4,657 individuals. In 2013, Huang and colleagues publ ished a prospective cohort study using a subset of the child sample of the 1979 National Longitudinal Survey of Youth (n=5,141) The study, in part, assessed the association between cannabis use trajectories (low use, sporadic use, or increasing use) and o besity trajectories (low risk of obesity, increasing risk of obesity, sustained high risk of obesity).

PAGE 44

44 Another study that used data from NHANES (1988 1994) was published in 2012 by Rajavashisth and colleagues. With a sample size of 8,127 adults, the study assessed the association between cannabis use and diabetes In a cross sectional analysis of data from the Mater University Study of Pregnancy and its Outcomes (MUSP), Hayatbakhsh and colleagues (2010) examined the association between duration of cannabis use and BMI among 2,566 young adults The second study included in this review that used the CARDIA data was a prospective cohort study conducted by Rodondi and colleagues (2006) to examine the relationship between total days of cannabis use and BMI, hypertension (SBP and DBP), dyslipidemia (triglycerides), and diabetes (fasting glucose) among 3,617 individuals In 2001, Mittleman and colleagues used a case crossover design to estimate the relative risk of myocardial infarction (MI) within 1 hour of cannabis use using data from 3,882 participants enrolled in the Determinants of Myocardial Infarction Onset Study. Interestingly, the case crossover design allowed for patients to serve as their own controls, thus handling confounding by variables that are stable over time but may differ between individuals. The final study included in this review was published in 1974 by Aronow and colleagues. The study used a cross over design to estimate the association between cannabis cigarette smoking versus cont rol compared to placebo cigarette smoking versus control and time to onset of angina among 10 men with classical exertional angina and greater than 75% narrowing at major coronary vessels. Study Results by Outcome Tables 2 4 through 2 10 give the lead aut hor, year of publication, mean sample age, cannabis use measure, outcome measure, and results of the included studies presented in descending chronological order by publication date.

PAGE 45

45 BMI S tudies by Ross and colleagues (2016) and Dube and colleagues (2015) found significant, positive associations between cannabis use and BMI (OR 1.001, p value 0.02; = 0.09 95% CI ( 0.00, 0.17 ) in males and = 0. 09 95% CI ( 0.01 0.17 ) in females, respectively ) One study by Huang et al (2013) found a significant positive ass ociation (OR 1.6, p value=0.05) between increasing cannabis use over time and increasing obesity over time. The same study found a significant negative association ( OR 0.1, p value<0.01 ) between low cannabis use and increasing obesity over time Two st udies found significant, negative associations between cannabis use and BMI Hyatbakhsh and colleagues (2010) found significantly decreased odds of being overweight or obese among those who initiated use before 16 years and used at least once in the last m onth (OR 0.3, 95% CI 0.2, 0.5), and those who initiated use at 16 years or older and used at least a few days a week in the last month (OR 0.4, 95% CI 0.2, 0.7) compared to those who never used cannabis. Thompson and colleagues (2015) found significantly d ecreased mean BMI among current cannabis users compared to non users (multivariable adjusted mean difference = 0.771, p value<0.05). Five studies found no significant association between cannabis use and BMI (Blackstone and Herrmann, 2016; Penner et al., 2013; Racine et al., 2015; Rodon di et al., 2006; Thompson and Hay, 2015a) Hypertension Two studies found significant positive association s between cannabis use and hypertension; o ne was between never, former and current (past 180 day) cannabis use and increased mean diastolic blood pressure ( p value=0.0245; Racine et al., 2015) and the other was between increasing years of cannabis use and hypertension ( OR 1.05, 95% CI 1.02, 1.09;

PAGE 46

46 Yankey et al., 2016) Two studies found significant negative associations between cannabis use and hypertension; o ne was between frequent cannabis use and hypertension ( measured as SBP 130 mmHg and/or DBP reported use of antihypertensive medication) compared to never use ( OR 0.56, 95% CI 0.39, 0.80; Waterreus et al., 2016) and the other was between regular cannabis use and hypertension ( measured as 85 mmHg) compared to never use ( OR 0.50, 95% CI 0.1 0, 0.67; Yankey et al., 2016) Three studies found no significant association between cannabis use and hypertension (either combined SBP and DBP, or SBP alone, or DBP) (Penner et al., 2013; Rodondi et al., 2006; Thompson an d Hay, 2015a) Dyslipidemia One study found a significant, positive association between increasing years of cannabis use and elevated triglycerides ( OR 1.03, 95% CI 1.01, 1.06; Yankey et al., 2016) One study found a significant negative association between frequent cannabis use and elevated triglycerides as well as elevated HDL C compared to never use ( OR 0.61, 95% CI 0.45, 0.83 and OR 0.68, 95% CI 0.43, 0.85 respectively; Waterreus et al., 2016) Four studies found no significant association between cannabis use and dyslipidemia (Penner et al., 2013; Racine et al., 2015; Rodondi et al., 2006; Thompson and Hay, 2015a ) Diabetes Three studies found significant negative associations between cannabis use and diabet es; o ne between frequent cannabis use as well as occasional cannabis use and diabetes (measured as elevated fasting glucose) compared to never use ( OR 0.60, 95% CI 0.45, 0.75 and OR 0.72, 95% CI 0.54, 0.96 respectively; Waterreus et al., 2016) one between past cannabis use and diabetes (measured as elevated fasting glucose) compared to never use (multivariable adjusted mean differ ence 0.07, 95% CI 4.22, 0.11 ; Penner et al., 2013) and one between lifetime cannabis

PAGE 47

47 use and diabetes (measured as self reported dia betes or elevated fasting glucose) compared to never users ( OR 0.36, 95% CI 0.24, 0.55; Rajavashisth et al., 2012) Five studies found no si gnificant association between cannabis use and diabetes (Bancks et al., 2015; Penner et al., 2013; Racine et al., 2015; Rodondi et al., 2006; Yankey et al., 2016) Metabolic Syndrome One study found a significant, positive associat ion between increasing years of cannabis use and metabolic syndrome ( OR 1.05, 95% CI 1.01, 1.05; Yankey et al., 2016) O ne study found a significant, negative association between frequent cannabis use and metabol ic syndrome compared to never users ( OR 0.56, 95% CI 0.39, 0.80; Waterreus et al., 2016) One study found no significant association between cannabis use and metabolic syndrome (Vidot et al., 2016) Myocardial Infarction Only one study assessed the association between cannabis use and myocardial infarction. Results indicated that within 1 hour after smoking cannabis risk of MI was significantly increased compared to periods of non use (RR 4.8 95% CI 2.9, 9.5). Additionally, within 2 hours after smoking cannabis, risk of MI was still significantly higher than periods of non use (RR 1.7 95% CI 0.6, 5.1; Mittleman et al., 2001) Angina Only one study assessed the association between cannabis use and angina. Results showed a 48% decrease in time to angina onset after cannabis cigarette use compared to control Contrastingly, there was only a 4% dec rease in time to angina onset after placebo cigarette use compared to control. The difference in time to angina onset between cannabis cigarette versus control and placebo cigarette versus control was statistically significant (p<0.001) (Aronow and Cassidy, 1974)

PAGE 48

48 Risk of Bias Table 2 11 shows the designations for risk of bias for each study included by category. Table 2 12 gives th e overall frequency of each risk of bias rating by risk of bias category. Support for the risk of judgment s for eac h study are listed in Appendix B S election bias and generalizability of the findings were judged based on the sampling method used, and th e target population for which statements of results were made. Overall 60% of the studies included were judged to have a low risk of selection bias. This was in part due to the number of studies using NHANES data (n=5 out of 16 ), which is nationally repres entative. Judgment s on information bias related to the classification of exposure were made based on the type of cannabis use measure used, and any validity or reliability information reported about the measure. Measures that account ed for duration of use frequency of use, quantity of use, or cumulative lifetime use were Most of the included studies ( 68.8 %) were judged to have a high risk of bias. The most frequent classification of cannabis exposure used was never, former/past, and current. I nformation bias related to classification of outcome was assessed based on the type of outcome use measure used, and any validity or reliability information reported about th e measure. Measures that used standard diagnostic tests (e.g. non invasive imaging, blood tests such as HbA1c clinic or trained interviewer measured height and weight ) or validated measures such as ICD9/ICD10 codes accompanied by chart review were conside Most outcome measures were judged to have low risk of bias ( 75.0 %), primarily due to height and weight measurements and blood samples taken on site or at the time of the interview. Judgment s on risk of bias due to confounding were made based on the methods taken to control for confounding, in particular information on confounding factors previously discussed were considered Two thirds of studies ( 68.8 %) included were judged to have low risk of bias

PAGE 49

49 due to confounding. These studi es frequently controlled for demographics, measures of socio economic status, measures of other substance use including tobacco and alcohol use, measures of physical activity and/or caloric intake, and measures of coronary diseas e risk factors where appro priate. While only two studies (Dube et al. 2015 and Hayatbakhsh et al. 2010) controlled for mental health conditions such as anxiety and depression, the exclusion of mental health conditions alone was not deemed enough to consider a study to have a high r isk of bias. R isk of bias due to temporality was judged by determining if the exposure necessarily occurred prior to the outcome. While the results of cross sectional studies cannot be used to make causal assertions, it is possible that the exposure neces sarily came before the outcome. For example, if the exposure measure was never, former, and current cannabis use, and the outcome was BMI measured at the time of the interview, the exposure would have necessarily pre dated the outcome. In total, a little o ver half of the studies included ( 56.2 %) were considered low risk of bias due to temporality. Discussion This systematic review represents one of the first literature reviews to focus on the association between cannabis use and coronary disease and corona ry disease risk factors using only epidemiologic evidence Figure 2 2 shows a model of the genesis of coronary disease adapted from Pearson and colleagues ( 1993 ; Figure 1 1) with the findings of this review highlighted in red. The direction of the significant association (positive or negative) is shown underneath each arrow, where a positive symbol indicates that cannabis use is significantly associated with increa sing outcome, and a negative symbol indicates that cannabis use is significantly assoc iated with decreasing outcome. The majority of findings included in this review were not statistically significant. Of the findings that were statistically significant, the direction of association was contradictory for

PAGE 50

50 the association between cannabis use and obesity, hypertension, dyslipidemia and metabolic syndrome. In contrast a ll significant findings indicated that cannabis use was negatively associated with diabetes, and that cannabis use was positively associated with myocardial infarction and angin a. The current body of evidence for associations between cannabis use and coronary disease or coronary disease risk factors is severely limited as evidenced by the numerous contradictory findings. The limitations of the extant literature include overuse o f datasets, limited ability to determine causal associations bias in the exposure measure, limited data on coronary disease as opposed to coronary disease risk factors, and potential problems controlling for confounding. One of the first limitations of t he extant literature was that many studies utilized the sa me datasets. While there were 16 studies included in this review, there were only 11 unique datasets represented. In particular, data from NH ANES was used by five studies (Penner et al., 2013; Rajavashisth et al., 2012; Thompson and Hay, 2015a; Vidot et al., 2016; Yankey et al., 2016) When data from these studies is counted as significant or non significant five separate times it may l ead to bias; allowing it to appear that there is more evidence for an association than exists here Unfortunately there are not many large publicly available datasets with sufficient information to determine associations between cannabis use and coronary disease. To researchers, NHANES represents one of the only viable dataset s with which to explore these questions because it has data on both cannabis use and coronary disease outcomes. As a direct result of the limited data available causal associations have not been adequately studied. While nine of the studies included did have exposure prior to outcome, more longitudinal data is needed to help determine causal associations.

PAGE 51

51 A second limitation of the extant literature is the bias resulting from the measures of cannabis used. Studies employed a range of cannabis use measures including binary measures (never use, lifetime use), various categorical measures, and continuous measu res. Even within the subgroup of categorical and continuous measures there was no common standard for measures of use making it im possible to pool exposure data. Most frequently categorical measures of cannabis use were used. Even when data on cumulative lifetime use, or use frequency were available they were categorized. Categorizing continuous data results in a loss of information and consequentially results in less precise effect estimates For example, Yankey and colleagues (2016 ) found significantly decreased odds of hypertension among regular cannabis users compared to never users, but when the measure of cannabis use was increasing years of use they found significantly increased odds of hypertension for every year of use. Additionally, categorizing cannabis use measures may result in researchers missing the true shape of the distribution of use For example, non normal distributions such as J or U shaped curves use may be important when considering risk. Dube and colleagues (2016) found that the significant, positive, association between past year frequency of cannabis use and change in BMI from ~20 to ~24 years of age was U shaped in both males and females. S tudies that used categorical measures of cannabis use found negati ve associations between cannabis use and coronary disease, while studies that used continuous measures accounting for quantity/frequency/duration of use found a positive association These contradictory findings might be due to the loss of information inhe rent in classifying non normal distributions. If this is true, it would be important to better understand how use at inflection points in the distribution change, and then how interventions can target reducing use below such levels.

PAGE 52

52 Unfortunately, catego rical measures of lifetime cannabis use (e.g. never, former, current) are often the only available measures in datasets that also provide both sufficient sample size and coronary disease or coronary disease risk factors measures. To sufficiently explore po ssible adverse effects of cannabis use, investigators need to collect detailed information on cannabis use including duration of use, frequency of use, and quantity of use in enough detail to standardize units Further, none of the studies assessed potency of cannabis used. Between 1995 and 2014, the ratio of THC to cannabidiol has increased from about 14:1 to about 80:1 (ElSohly et al., 2016) THC is the primary psychoactive component of cannabis. The primary hypothesized mechanism for an association between cannabis use and coronary disease is CB1 receptor activation by THC. As such, change in cannabis potency over time is an important factor that was excluded from consideration in these studies. A third key limitation of the extant literature is the paucity of studies measuring coronary disease status as an outcome. One possible reas on for this may be that researchers are reluctant to include coronary disease as the outcome if it self reported as in NHANES, t hough the CARDIA data provide adequate diagnostic measures of coronary disease. It may also be that there has been insufficien t interest in pursuing this line of research. Finally, researchers must control for relevant confounders Given the complex genesis of coronary disease and the many non modifiable behavioral factors that contribute to the development of risk factors and d isease itself (Figure 1 1), adequate controlling for confounding is imperative. Most studies included in this review adequately controlled for socio demographic characteristics physical activity, and other substance use. However, mental health conditions were infrequently controlled for despite being associated with both the exposure and outcome.

PAGE 53

53 Regardless of the measured factors controlled for, i nconsistent findings suggest that there may be other unmeasured factors affecting the observed associations. Thompson and colleagues (2016) used regression models to assess the association between cannabis use and coronary disease risk factors including BMI The results of their study showed a significant negative association between cannabis use and metabolic sy ndrome. Considering this paradoxical, the authors replaced cannabis use with alcohol use, and then with carbohydrate intake. Despite adequately controlling for confounding, including physical activity, the results also showed a significant negative associa tion between both cannabis use and alcohol use and cannabis use and carbohydrate intake. The authors suggest that regression estimates may be biased by the correlation between cannabis use and unobserved characteristics also affecting BMI (Thompson and Hay, 2015a, 2015b) Further, the authors suggest that NHANES does not adequately capture environmental or behavioral variables necessary to control for differences between cannabis users and non user s (Thompson and Hay, 2015b) This may mean the standard epidemiological approach to risk factor epidemiology has failed to adequately address factors above the individual level such as social support, neighborhood environment and home environment (Glanz et al., 2008) There are a few limitations that should be considered when interpreting the results of this qualita tive systematic review. First non English language articles were excluded, so studies on this subject included in this rev iew were biased towards populations in English speaking countries. Second the grey literature was not included. It is expected that there is limited grey literature on this topic conducted with the required rigor. However, it is possible that study resul ts were available outside of traditional academic publishing mechanisms and that these studies were missed

PAGE 54

54 Overall the extant literature on the association between cannabis use and coronary disease or coronary disease risk factors is inadequate to make any determinations about association. Clinicians should use caution when discussing the possible effects (either positive or negative) of cannabis on coronary disease because there is simply not enough evidence to make any conclusions about risk. A recent report on the health effects of cannabis published by the National Academies of Sciences, Engineering and Medicine (2017) similarly suggests caution, as the evidence is unclear as to whether there is an association between cannabis use and cardio metabolic disease. A dequate funding and support for a representative, longitudinal, properly powered study that includes data on frequency/quantity/duration of cannabis use and diagnosis of coronary disease is imperative. Cannabis use is increasing, and will cont inue to increase in light of nationwide support for legalization and decriminalization (Bostwick, 2012; Hoffmann and Weber, 2010; Yankey et al., 2016) The charge of public health is to conduct sound research that identifies the protective and harmf ul effects of cannabis as it relates to coronary disease

PAGE 55

55 Table 2 1. Inclusion criteria for studies in the systematic review Category Criteria Population (P) Any age group or gender breakdown is acceptable Intervention/exposure of interest (I) Cannabis or marijuana use Comparator (C) Individuals with no history of cannabis or marijuana use Outcomes (O) Coronary Disease (myocardial infarction, coronary artery disease, angina) OR Coronary disease risk factors (Hypertension, Diabetes, BMI, Dyslipidemia, Metabolic Syndrome) Table 2 2. Exclusion criteria for studies in the systematic review Category Criteria Study design Case reports and case series Reviews Conference abstracts Policy papers Editorials/Commentaries/Letters Studies using animal models Language Non English language Exposure Synthetic cannabinoids Comparator No control group used Outcomes Subclinical measures of coronary disease such as carotid intima media thickness and coronary calcium scores Outcomes not listed in Table 2 1 Association No measure of association between exposure and outcome reported

PAGE 56

56 Figure 2 1. Systematic review flow lo g Full text articles excluded (n = 23 ) Reasons for exclusion Full text not located (n=2 ) Ineligible exposure measure (e.g. not cannabis use alone) (n=3) Ineligible outcome measure (n=10 ) No measure of association between exposure and outcome reported (n=2) Ineligible comparator (n=1) Review Paper (n=4) Commentary (n=1) Records identified through Pubmed/Medline and Web of Science (n = 1009) Records after duplicates removed (n = 750) Article Titles Screened (n = 404) Records excluded (n = 226) Full text articles assessed for eligibility (n = 38) Studies included in qualitative synthesis (n = 16 ) Article Abstracts Screened (n = 178) Records excluded (n = 140) Full text articles included after manual search of references (n = 1) Records excluded (n = 346 )

PAGE 57

57 Table 2 3. Overview of the studies included in the systematic review Lead Author Year of Publication Journal Study Design Study Period (years) Data Set a Sample Size Outcome Measured Ross, MJ 2016 Journal of the International Neuro psychological Society Cross Sectional Not reported Adolescents age 14 17 from Miami/Dade Country 238 BMI Yankey, B 2016 Diabetes & Metabolic Syndrome: Clinical Research & Reviews Cross Sectional 2011 2012 NHANES 3,051 BMI, Hypertension, Dyslipidemia Diabetes, Metabolic Syndrome Waterreus, A 2016 Psychological Medicine Cross Sectional 2010 2011 Survey of High Impact Psychosis 1,813 Hypertension, Dyslipidemia, Diabetes, Metabolic Syndrome Vidot, D 2016 American Journal of Medicine Cross Sectional 2005 2010 NHANES 8,478 Metabolic Syndrome a NHANES=National Health and Nutrition Examination Survey

PAGE 58

58 Table 2 3. Continued Lead Author Year of Publication Journal Study Design Study Period (years) Data Set a Sample Size Outcome Measured Blackstone, S 2016 Health Education and Behavior Cross Sectional 2009 2010 Health Behavior in School Aged Children Series 10,925 BMI Bancks, M 2015 Diabetologia Cross Sectional & Prospective Cohort 1992 2011 CARDIA 3,151 Diabetes Racine, C 2015 Journal of Disease and Global Health Cross Sectional 2014 Black patients from a Family Practice clinic at University Hospital of Brooklyn, NY 100 BMI, Hypertension, Dyslipidemia, Diabetes Dube, E 2015 Pharmacology Biochemistry and Behavior Prospective Cohort 2005 2012 Nicotine Dependence in Teens 590 BMI Thompson, CA 2015 Annals of Epidemiology Cross Sectional 2005 2012 NHANES 6,281 BMI, Hypertension, Dyslipidemia, Diabetes NHANES=National Health and Nutrition Examination Survey ; CARDIA=Coronary Artery Risk De velopment in Young Adults Study

PAGE 59

59 Table 2 3. Continued Penner, EA 2013 American Journal of Medicine Cross Sectional 2005 2010 NHANES 4,657 BMI, Hypertension, Dyslipidemia, Diabetes Huang, DYC 2013 Addictive Behavior Prospective Cohort 1986 2008 National Longitudinal Survey of Youth 5,141 BMI Rajavashisth, TB 2012 BMJ Open Cross Sectional 1988 1994 NHANES 8,127 Diabetes Hayatbakhsh, MR 2010 American Journal of Drug and Alcohol Abuse Cross Sectional 1981 2002 MUSP 2,566 BMI Rodondi, N 2006 American Journal of Cardiology Prospective Cohort 1985 2000 CARDIA 3,617 BMI Hypertension, Dyslipidemia Diabetes Mittleman, MA 2001 Circulation Case Crossover 1989 1996 Determinants of Myocardial Infarction Onset Study 3,882 MI Aronow, WS 1974 New England Journal of Medicine Crossover 1974 Men with classic stable exertional angina and >75 percent narrowing at major coronary vessels 10 Angina a NHANES=National Health and Nutrition Examination Survey; CARDIA=Coronary Artery Risk Development in Young Adults Study; MUSP= Mater University Study of Pregnancy

PAGE 60

60 Table 2 4. Lead author, mean sample age, cannabis use measure, BMI measure, and main results f or all included studies with BMI as an outcome Lead Author Year of Publication Mean Sample Age Cannabis Use Measure BMI Measure Results Ross, MJ 2016 15.62 years Lifetime use in grams calculated from quantity and frequency data Height and weight measured at time of interview; standardized into BMI z score and clinical classifications Adjusted odds ratios: 1.001* Blackstone, S 2016 Not Reported; Majority of participants were in grade 8 Lifetime cannabis use [Y/N] Self reported height and weight Racine, C 2015 46.3 years Never, Former (lifetime but not past 180 days), and Current (past 180 days) use Abstracted from medical records Mean BMI by Cannabis Use Status Current: 26.5 Former: 31.1 Never: 29.6 Dube, E 2015 24 years at follow up Weekly frequency of use over the past year Change in mean interviewer measured height and weight from follow up cycle 19 to 22 Males: 0.09* Females: 0.10* Thompson, CA 2015 Not Reported Never, past (lifetime but not past 30 days), and current (past 30 days) use Interviewer measured height and weight Multivariable adjusted mean difference Never Use: Reference Past Use: 0.051 Current Use: 0.771* Penner, EA 2013 Not Reported; Majority of participants were 30 44 years old Never, past (lifetime but not past 30 days), and current (past 30 days) use Interviewer measured height and weight Multivariable adjusted mean difference Never Use: Reference Past Use: 0.08 Current Use: 0.61

PAGE 61

61 Table 2 4. Continued Lead Author Year of Publication Mean Sample Age Cannabis Use Measure BMI Measure Results Huang, DYC 2013 Not Reported Use trajectory groups from age 12 years to 18 years: Low Use Trajectory Sporadic Use Trajectory Increasing Use Trajectory BMI trajectory groups from age 20 24: Low Obesity Trajectory Increasing Obesity Trajectory Sustained Obesity Trajectory Adjusted odds ratios for being in the increasing obesity trajectory: Low Use: Reference Sporadic Use: 0.1* Increasing Use: 1.6* Adjusted odds ratios for being in the sustained obesity trajectory: Low Use: Reference Sporadic Use: 0.2* Increasing Use: 1.1* Hayatbakhsh, MR 2010 20.56 years at last follow up N ever use, use once or less in the last month, use at least every few days in the last month with onset before 16 years, and use at least every few days in the last month with onset at 16 years or later Interviewer measured height and weight Adjusted Odds Ratios Never Used: Ref Once or less in last month: 1.0 Started before 16 years, used at least once in the last month: 0.3* Started at 16 years or above and used at least few days a week in the last month: 0.4* Rodondi, N 2006 40.1 years Cumulative days of use categorized as never use, <180 days use, 180 1799 days, and Interviewer measured height and weight Adjusted Mean Values Never User: 28.8 <180 Days: 28.6 180 1799 Days: 28.8 1800 Days: 28.9 Statistically significant at p 0.05

PAGE 62

62 Table 2 5 Lead author, mean sample age, cannabis use measure, hypertension measure, and main results for all included studies with hypertension as an outcome Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Hypertension Measure Results Yankey, B 2016 38.7 years Never, non regular, and regular users; total years of use Average blood pressure above 130/85 mmHg or self reported use of antihypertensive medication Adjusted odds ratio by use status Never: Reference Regular: 0.26* Non regular: Not Reported Adjusted odds ratio by years of use: 1.05* Waterreus, A 2016 Not Reported; Majority of participants 25 34 years old Past year cannabis use categorized as non user, occasional user (less than once per week), or frequent user Interviewer measure blood pressure with systolic blood and/or diastolic or self reported use of antihyper tensive medication Adjusted odds ratio Non Users: Reference Occasional Users: 0.72* Frequent Users: 0.58* Racine, C 2015 46.3 years Never, Former (lifetime but not past 180 days), and Current (past 180 days) use Systolic and diastolic blood pressure abstracted from medical records Mean SBP by Cannabis Use Status Current: 126.0 Former: 129.5 Never: 127.7 Mean DBP by Cannabis Use Status* Current: 73.3 Former: 80.0 Never: 73.4 Statistically significant at p 0.05

PAGE 63

63 Table 2 5 Continued Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Hypertension Measure Results Thompson, CA 2015 Not Reported Never, past (lifetime but not past 30 days), and current (past 30 days) use Interviewer measured systolic and diastolic blood pressure Multivariable adjusted mean difference for SBP Never Use: Reference Past Use: 1.045 Current Use: 0.251 Multivariable adjusted mean difference for DBP Never Use: Reference Past Use: 0.080 Current Use: 0.513 Penner, EA 2013 Not Reported; Majority of participants were 30 44 years old Never, past (lifetime but not past 30 days), and current (past 30 days) use Interviewer measured systolic and diastolic blood pressure Multivariable adjusted mean difference for SBP Never Use: Reference Past Use: 1.04 Cur rent Use: 0.64 Multivariable adjusted mean difference for DBP Never Use: Reference Past Use: 0.01 Current Use: 0.49 Statistically significant at p 0.05

PAGE 64

64 Table 2 5 Continued Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Hypertension Measure Results Rodondi, N 2006 40.1 years Cumulative days of use categorized as never use, <180 days use, 180 1799 days, S ystolic and diastolic blood pressure Adjusted Mean SBP Values Never User: 113.6 <180 Days: 112.9 180 1799 Days: 112.1 1800 Days: 112.9 Adjusted Mean DBP Values Never User: 74.8 <180 Days: 74.2 180 1799 Days: 74.0 1800 Days: 73.9 Statistically significant at p 0.05

PAGE 65

65 Table 2 6 Lead author, mean sample age, cannabis use measure, dyslipidemia measure, and main results for all included studies with dyslipidemia as an outcome Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Dyslipidemia Measure Results Yankey, B 2016 38.7 years Never, non regular, and regular users; total years of use Plasma with HDL C less than 50 mg/dl for females and less than 40 mg/dl for males Plasma triglycerides of 150 mg/dl and above. Adjusted odds ratio for low HDL C Never Use: Reference Regular Use: 0.54 (0.18, 1.55) Non regular Use: Not Reported Adjusted odds ratio for low HDL C by years of us e: 1.02 (0.99, 1.05) Adjusted odds ratio for hypertriglyceridemia Never: Reference Regular: 0.76 (0.15, 3.97) Non regular: Not Reported Adjusted odds ratio for hypert riglyceridemia by years of use: 1.03 ( 1.01, 1.06) Statistically significant at p 0.05

PAGE 66

66 Table 2 6. Continued Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Dyslipidemia Measure Results Waterreus, A 2016 Not Reported; Majority of participants were 25 34 years old Past year cannabis use categorized as non user, occasional user (less than once per week), or frequent user Blood samples with mmol/l Blood samples with HDL C < 1.0 mmol/l for men and < 1.3 mmol/l for women. Adjusted odds ratio for elevated triglycerides Non Users: Reference Occasional Users: 0.75 Frequent Users: 0.61* Adjusted odds ratio for adverse HDL C Non Users: Reference Occasional Users: 0.95 Frequent Users: 0.68* Racine, C 2015 46.3 years Never, Former (lifetime but not past 180 days), and Current (past 180 days) use Total cholesterol, LDL, HDL, and triglycerides abstracted from medical records Mean total cholesterol by Cannabis Use Status Current: 156.9 Former: 189.0 Never: 181.8 Mean LDL by Cannabis Use Status Current: 92.5 Former: 105.9 Never: 103.8 Mean HDL by Cannabis Use Status Current: 48.3 Former: 57.6 Never: 54.6 Mean triglycerides by Cannabis Use Status Current: 85.9 Former: 133.0 Never: 120.3 Statistically significant at p 0.05

PAGE 67

67 Table 2 6. Continued Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Dyslipidemia Measure Results Thompson, CA 2015 Not Reported Never, past (lifetime but not past 30 days), and current (past 30 days) use Serum triglycerides (mg/dL) and serum HDL C (mg/dL) Multivariable adjusted mean difference for triglycerides Never Use: Reference Past Use: 0.009 Current Use: 0.022 Multivariable adjusted mean difference for HDL C Never Use: Reference Past Use: 0.110 Current Use: 0.726 Penner, EA 2013 Not Reported; Majority of participants were 30 44 years old Never, past (lifetime but not past 30 days), and current (past 30 days) use Serum triglycerides (mg/dL) and serum HDL C (mg/dL) Multivariable adjusted percent difference for triglycerides Never Use: Reference P ast Use: 0.29% Current Use: 1.2% Multivariable adjusted mean difference for HDL C Never Use: Reference Past Use: 0.14 Current Use: 1.22 Statistically significant at p 0.05

PAGE 68

68 Table 2 6. Continued Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Dyslipidemia Measure Results Rodondi, N 2006 40.1 years Cumulative days of use categorized as never use, <180 days use, 180 1799 days, Fasting plasma levels of total cholesterol (mg/dL), HDL C (mg/dL) and triglycerides (mg/dL) Adjusted mean total cholesterol values Never User: 184.2 <180 Days: 184.2 180 1799 Days: 184.6 1800 Days: 186.1 Adjusted mean HDL C values Never User: 50.6 <180 Days: 50.6 180 1799 Days: 50.6 1800 Days: 51.4 Adjusted mean triglyceride values Never User: 86.7 <180 Days: 88.5 180 1799 Days: 87.6 1800 Days: 92.9 Statistically significant at p 0.05

PAGE 69

69 Table 2 7 Lead author, mean sample age, cannabis use measure, diabetes measure, and main results for all included studies with diabetes as an outcome Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Diabetes Measure Results Yankey, B 2016 38.7 years Never, non regular, and regular users; total years of use Fasting plasma glucose of 100mg/dl or on antidiabetic treatment including insulin Adjusted odds ratio Never Use: Reference Regular Use: 0.50 Non regular Use: Not Reported Adjusted odds ratio by years of use:1.02 Waterreus, A 2016 Not Reported; Majority of participants were 25 34 years old Past year cannabis use categorized as non user, occasional user (less than once per week), or frequent user Fasting blood mmol/l. Adjusted odds ratio Non Users: Reference Occasional Users: 0.59* Frequent Users: 0.60* Bancks, M 2015 Not Reported Never, past (lifetime but not past 30 days), and current (past 30 days) use; cumulative lifetime use Fasting plasma glucose, 2 hour oral glucose tolerance test, or HbA1c Cross sectional Adjusted ORs: Never users: Reference Former users: 1.23 Current users: 1.16 Prospective adjusted HRs: Never use: Reference 1 9 times use 0.93 10 99 times use 1.30 >= 100 times use 1.16 Statistically significant at p 0.05

PAGE 70

70 Table 2 7. Continued Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Diabetes Measure Results Racine, C 2015 46.3 years Never, Former (lifetime but not past 180 days), and Current (past 180 days) use Plasma glucose and HgA1c abstracted from medical records Mean glucose by Cannabis Use Status Current: 108.7 Former: 108.7 Never: 112.6 Mean HgA1c by cannabis use Status Current: 7.2 Former: 6.4 Never: 6.5 Thompson, CA 2015 Not Reported Never, past (lifetime but not past 30 days), and current (past 30 days) use Fasting plasma glucose Multivariable adjusted mean difference Never Use: Reference Past Use: 1.503 Current Use: 1.324 Penner, EA 2013 Not Reported; Majority of participants were 30 44 years old Never, past (lifetime but not past 30 days), and current (past 30 days) use Fasting plasma glucose, HgA1c Multivariable adjusted mean glucose difference Never Use: Reference Past Use: 2.16* Current Use: 0.47a Rajavashisth 2010 Not Reported Lifetime use [Y/N] Self reported diabetes or fasting 126mg/dL Adjusted odds ratio: 0.36* Rodondi, N 2006 40.1 years Cumulative days of use categorized as never use, <180 days use, 180 1799 days, and Fasting plasma glucose (mg/dL) Adjusted Mean Values Never User: 86.7 <180 Days: 86.3 180 1799 Days: 86.8 1800 Days: 87.4 Statistically significant at p 0.05 a It is not clear whether this was statistically significant as the mean difference falls outside the reported confidence inter val; Mean Difference 0.47 95% CI 2.51, 1.57

PAGE 71

71 Table 2 8. Lead author, mean sample age, cannabis use measure, metabolic sy ndrome measure, and main results for all included studies with metabolic syndrome as an outcome Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Metabolic Syndrome Measure Results Yankey, B 2016 38.7 years Never, non regular, and regular users; total years of use Three or more of the following components: 1) Hypertension based on blood pressure or use of antihypertensive medication 2) Diabetes based on fasting plasma glucose of 100mg/dl or on antidiabetic treatment 3) Abdominal obesity based on waist circumference of more than 88 cm for women and 102 cm for men 4) Low HDL C based on 50mg/dl for males and 40 mg/dl for females 5) Hypertriglyceridemia based on plasma triglycerides of 150mg/dl Adjusted odds ratio Never Use: Referenc e Regular Use: 0.25 Non regular Use: Not Reported Adjusted odds ratio by years of use: 1.05

PAGE 72

72 Table 2 8. Continued Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Metabolic Syndrome Measure Results Waterreus, A 2016 Not Reported; Majority of participants were 25 34 years old Past year cannabis use categorized as non user, occasional user (less than once per week), or frequent user Three or more of the following components: 1) waist circumference 2) blood pressure 3) triglycerides 4) glucose (fasting plasma) 5) HDL OR prescribed medication for hypertension, hyperlipidemia, or hyperglycemia OR self reported diabetes Adjusted odds ratio for elevated triglycerides Non Users: Reference Occasional Users: 0.75 Frequent User s: 0.56* Vidot, D 2016 Not Reported; Majority of participants were 45 to 59 years old Never, past (lifetime but not past 30 days), and current (past 30 days) use Three or more of the following components: 1) waist circumference 102 cm in men, 88 cm in women 2) hypertension SBP 130mmHg, DBP 85mmHg 3) HDL cholesterol 40 mg/dL for men, 50 mg/dL women 4) triglycerides 150 mg/dL 5) fasting glucose 100 mg/dL Odds Ratio Never Use: Reference Past Use: 0.76 Current Use: 0.69 Statistically significant at p 0.05

PAGE 73

73 Table 2 9 Lead author, mean sample age, cannabis use measure, myocardial infarction measure, and main results for all included studies with myocardial infarction included as an outcome Lead Author Year of Publicati on Mean Sample Age Cannabis Use Measure MI Measure Results Mittleman, M 2001 44 years for cannabis users 62 years for non cannabis users Cannabis use in the year prior to MI used to estimate the expected frequency of use in an average 1 hour period Standard diagnostic codes and chart review Relative risk for MI within 1 hour cannabis use compared to periods of non use: 4.8* Relative risk for MI within 2 hours cannabis use compared to periods of non use: 1.7 Statistically significant at p 0.05 Table 2 10. Lead author, mean sample age, cannabis use measure, angina measure, and main results for all included studies with angina included as an outcome Lead Author Year of Publication Mean Sample Age Cannabis Use Measure Angina Measure Results Aronow, WS 1974 47.3 10 puffs of cannabis cigarette Time to angina pectoris in seconds after exercise Mean time to Angina Control 1: 244.3 Cannabis Cigarette: 129.4 Control 2: 240.5 Placebo Cigarette: 220.1* Statistically significant at p 0.05

PAGE 74

74 Table 2 11 Risk of bias classifications for each study included in the review by category of possible bias Lead Author Selection Bias Generalizability Information Bias Classification of Exposure Information Bias Classification of Outcome Confounding Tempora lity Ross, MJ High Low Low High Low Yankey, B Low High Low Low High Waterreus, A High Low Low Low High Vidot, D Low High High High High Blackstone, S High High High High High Bancks, M High Low Low Low Low Racine, C High High Low High High Dube, E High Low Low Low Low Thompson, CA Low High Low Low High Penner, EA Low High Low Low Low Huang, DYC Low High High High Low Rajavashisth, TB Low High High Low High Hayatbakhsh, MR High High Low Low Low Rodondi, N High High Low Low Low Mittleman, MA High High Low Low Low Aronow, WS High Low Low Low Low

PAGE 75

75 Table 2 12 Frequency of high versus low risk of bias for each category of possible bias Selection Bias Generalizability Information Bias Classification of Exposure Information Bias Classification of Outcome Confounding Temporality Bias Rating n (%) n (%) n (%) n (%) n (%) High Risk of Bias 10 ( 62.5 %) 11 ( 68.8 %) 4 ( 25.0 %) 5 ( 31.2 %) 7 ( 43.8 %) Low Risk of Bias 6 (37.5 %) 5 (31.2 %) 12 ( 75.0 %) 11 ( 68.8 %) 9 (56.2 %)

PAGE 76

76 Figure 2 2. Model of the genesis of coronary disease adapted from Pearson et al. 1993 (Figure 1 1) with exposure, outcomes, and direction of statistically significant associations from this literature review shown in red Metabolic Syndrome ( ) Cannabis Use Sedentary lifestyle Diet Saturated Fat Salt Cholesterol Total energy content Heavy alcohol consumption Tobacco smoking Other substance use Behavioral Risk Factors Age Sex Family History Non modifiable Risk Factors Obesity (Elevated BMI) Hypertension (Elevated SBP, DBP) Dyslipidemia (Adverse LDL C, HDL C, Total Cholesterol, Triglycerides) Diabetes (Elevated Plasma Glucose, HgA1c ) Depression Physiological Risk Factors Coronary Artery Disease Angina Myocardial Infarction Coronary Disease Outcomes (+) ( /+) ( /+) (+) ( /+) ( /+)

PAGE 77

77 CHAPTER 3 GENERAL METHODS HealthStreet Model The sample for the analyses included in this dissertation comes from HealthStreet an innovative community engagement model and outreach program with a mission to improve the health of our community by reducing disparities in health care and health research. In Florida, HealthStreet is Science In (Cottler, LB PI). HealthStreet at the University of Florida began in November 2011 and is based on HealthStreet in St. Louis at Washington University, which began in 1989 Both were founde d by Linda B. Cottler, PhD MPH. HealthStreet has engaged over 8, 8 00 commun ity members in the Gainesville, and Jacksonville areas through a community health worker model (Cottler et al., 2012, 2010) HealthStreet assesses community member health conditions and concerns, and links community members to health research opportunities and medical and social services. T he mission and aims of HealthStreet are depicted in F igure 3 1 below. Recruitment and Sampling HealthStreet members are recruited within the communities in which they live, work, and recreate. CHWs are sent to places such as parks, bus stops, shelters, churches, grocery stores, laundromats, fitness centers, and health fairs to engage community membe rs. This method of recruitment is culturally sensitive and engages diverse populations. Data included in these analyses come fro m HealthStreet Gainesville, HealthStreet Jacksonville and HealthStreet Miami The majority of HealthStreet members are African American (61.5%), and female (58.5%).

PAGE 78

78 Health Intake Form After an informed consent is signed, each new HealthStreet member is interviewed by a CHW using the HealthStreet Intake Form. The HealthStreet Intake Form is an in person interview that takes about 3 0 minutes, and assesses health conditions including mental health conditions medical health concerns, history of substance use, and willingness to participate in health research. The HealthStreet Intake Form questions used i n these analyses are listed in Table 3 1. CHWs then link HealthStreet members to health research, and medical and social services. All intakes are entered into a REDCap database for analysis. Specific Aims The HealthStreet dataset offers comprehensive self reported information from la rge sample of North Floridians with which to examine the association between cannabis use and coronary disease through the following aims: 1. Evaluate differences in coronary disease status by risk factors such as socio demographic factors, mental health pro blems, and substance use among HealthStreet members then analyze the association between cannabis use and coronary disease using multiple logistic regression. 2. Identify subgroups ( latent classes ) among HealthStreet members based on lifetime substance use, and evaluate the association between coronary disease and latent class membership using multinomial logistic regression while controlling for socio demographic factors and mental health conditions Me asures Table 3 2 below lists Intake F orm question number, variable name and variable coding strategy for each measure used in the analyse s. Cannabis Use C annabis use served as the primary exposure of interest for Chapter 4, and as one of the manifest vari ables in Chapter 5. Cannabis use was categorized as 1) never use, former use (lifetime use but not in the p ast 30 days) or current use (use in the p ast 30 days) or as 2) never

PAGE 79

79 and lifetime use, depending on the analysis. History of cannabis use was elicite d by asking: (marijuana) were H ave you used marijuana in the p answe r no t o having ever used cannabis (marijuana) were cate gorized as never cannabis users. Those who answer no to having used cannabis in the last 30 days were categorized as former users, and those who answer ed yes were categorized as current users. Lifetime use was coded as positive for those who answer yes to having ever used cannabis. Coronary Disease Self reported lifetime history of coronary disease was the outcome of interest in Chapter 4, and a predictor of latent class membership in Chapter 5 Coronary disease was defined as endorsing a lifetime history of one or more of the following: Angina, Coronary Artery Disease (CAD), or Myocardial Infarction (MI). The Health Street Intake Form question asks ever been told you have, or have you ever had a pr answer yes to one or more of these condition questions were coded as having a lifetime history of coronary disease. Covariates Socio demographics Socio demographic information for these analyses include d self reported sex, race, age, health insurance status, education level, food insecurity status, and employment status Socio demographic variables were used in descriptive analyses and as relevant covariates in Chapters 4 6. Race was categorized as Black or African American, White, or Other. Age was left as a continuous variable. Education level was initially recorded as the las t gr ade completed, and then categorized as 12 or fewer years of education and greater than 12 years of education. Food insecurity was elicited by aski

PAGE 80

80 have enough money to buy food that you or your family neede yes coded as food insecure. Health insurance status and employment status was elicited by asking part yes coded as insured, and employed. Coronary disease risk factors Coronary disease risk factors were included as confo unding factors in the regression analyses in Chapters 4 and 5 Coronary disease risk factors include d BMI, history of hypertension, history of type II diabetes and history of hypercholesterolemia. T he HealthStreet Intake Form question asks been told you had, or ha ve you ever had a problem with high blood pressure ? had a problem with di Those who answer ed yes were coded as having a l ifetime history of hypertension, hypercholesterolemia and/or type II diab e tes respectively. reported height and weight, an d then categoriz ed as underweight, normal weight, overweight, or obese acco r ding to World Health Organization guidelines (World Health Organization, 2000) Other substance use Use of other substances (both illicit and not illicit) were included as confounding factor s in the regression analysis in Chapt er 4, and as manifest variables in Chapter 5. Non cannabis s ubstance use covariates were measured using 1) three categories : n ever use, former use (lifetime but not l ast 30 day use ) or current use (last 30 day use) or 2) two categories: never use, or lifetime use HealthStreet members were first asked yes the follow up question was asked ave you Those who answer no to having e ver used the

PAGE 81

81 substance are coded as never use. Lifetime use is coded as those who answer yes to having ever used the substance. Those who answer no to having used the substance in the p ast 30 days were categorized as former users, and those who answer yes were categorized as current users. Th e tobacco use covariate combine d the Health Intake Form substance use questions asking about cigarette u se, cigar or cigarillo use, hookah and/ or bong use. In the same way that cannabis and other illicit substanc es were coded, tobacco use was coded as 1) never, fo rmer, or current, or 2) never or lifetime use. The Health Intake Form assess ed p ast 30 day binge drinking four drinks for men, three dri ed yes were coded positive for binge drinking. Mental health conditions Mental health condition covariates assessed were lifetime self reported history of depression, anxiety, and other mental health conditions (ADD/ADHD, autism, bipolar disorder, eating disorders, mania, personality disorder, schizophrenia) These were elicited using the For depression and anxiety, those who answer ed yes to the associated question were coded as having a lifetime history of depression or anxiety. For other mental health conditions those who answer e d yes to one or more of the remaining disorders were coded positive for lifetime history of other mental health conditions Methods of Analysis The univariate and regression analyses conducted in this dissertation were generated using SAS software, ve rsion 9.4 of the SAS System for Windows. Copyright 2002 2012 SAS Institute Inc. The latent class analyses conducted in this dissertation were generated using Mplus Version 6.0 (Muthen and Muthen, 2007)

PAGE 82

82 Univariate Analyse s The univariate a n alyses conducted in Chapters 4 5 were chi square analyse s exact tests. These tests assess the relationship between the socio demographic, substance use, mental health, and coronary disease risk factor variables and lifetime history of coronar y disease. Multiple Logistic Regression In Chapter 4, the primary analysis assess e d the association between cannabis use and coronary disease using multiple logistic regression. Regression is a modeling technique by which one variable is explained on the basis of one or more other variables (Hilbe, 2009) T h e Chapter 4 analysis uses logistic regression due to the binary nature of the outcome of interest (positive lifetime coronary disease vs negative lifetime coronary disease), and the interpretability of the odds ratios produced by the model (as compared to probit or log log models, for example) (Hilbe, 2009) The analytic process involves multiple steps: 1) identifying relevant covariates to be included, 2) assessing multicollinearity, 3) assessing confounding, 4 ) model building, and 5) assessing model fit. Identifying relevant covariates Given the complexity of the genesis of coronary disease it was important to account for many factors other than just cannabis use (see Figure 1 1). Relevant covariates for possi ble model inclusion were identified via literature searches. The HealthStreet dataset provide d the unique opportunity to include many relevant covariates in the analysis, including socio demographic characteristics other substance use, mental health condi tions and traditional coronary disease risk factors like elevated BMI, hypertensio n, hypercholesterolemia and type II diabetes

PAGE 83

83 Multicol l inearity Given the number of covariates considered when assessing the relationship between cannabis use and coronary disease it was important to consider multico l linearity. Multico l linearity occurs when two covariates are highly correlated thus add ing redundant information. For example, if two variables x 1 and x 2 are highly correlated, their effect estimates may indicate that neither relate to outcome y well, while the fit diagnostics would indicate that the model fits the data. This happens because the coefficient of x 1 is the expected change in log odds of y whi le holding x 2 constant. If x 1 and x 2 are highly correlated then x 1 more information about y, thus a misleading non significant result occurs. Each covariate was tested prior to the model building process against the other covariates for mu ltico l linearity by creating a correlation table using proc corr. If a pair of covariates were highly correlated (r > 0.75), only one was included in the final model. After the model building process, the variance inflation factor (VIF) was assessed for eac h covariate left in the model using the VIF option for proc reg in SAS. While model building was done using proc logistic, proc logistic does not offe r the option to compute the VIF; thus proc reg with the VIF option was necessary to assess multico l linear ity. The VIF is a metric for the severity of multico l linearity. If the SAS output indicate d a VIF>10 for any given predictor variable, meaning there was severe co l linearity, it was investigated and possibly removed from the model. Assessing confounding Af ter assessing for multico l linearity using the correlation table, each remaining covariate was assessed for confounding Confounding was assessed by first determining if the covariate has a significant association with the independent variable ( cannabis use ), and then if it ha d a significant association with the dependent variable ( coronary disease ) in the unexposed.

PAGE 84

84 Associations were tested using chi square tests, with significant associations set at a p value 0.05 Model building After determining which covariates were likely significant confounders of the association between cannabis use and coronary disease a manual backward elimination model building process was undertaken to garner the best possible model This was done using proc logistic in SAS. Initially, a ll potential confounders were included in the logistic regression model, then each covariate was removed one by one based on level of significance. If the removal of a covariate cause d a change of 10% or more in the odds r atio it w as kept in the final model. If the covariate cause d a change of less than 10% in the odds ratio it may have be en kept in the model if it was determined to be of theoretical importance Model fit One benefit of logistic regression is that it does not require many assumptions to be met unlike other regression models. However, it remains important to assess model fit. R squared (coefficient of determination) was used to measure how well the regression model fits the data. It was also used to determi ne how much of the variation in the model is explained by the variables in the model. The Hosmer Lemeshow test was used to assess goodness of fit to determine if the model could be improved with the addition of other variables or interaction terms (Hosmer and Lemeshow, 2000) The Hosmer Lemeshow test groups cases together based on their predicted value from the logistic regression model, arrays them from lowest to highest, and puts them in 10 groups of equal size. For each group, the number of observed events and the number of expected square is used to compare observed and expected event counts. A low p value suggests rejection of the logistic regression model

PAGE 85

85 Latent Class Analysis Chapter 5 utilize d latent class analysis (LCA). LCA was conducted using Mplus to identify subgroups of those endorsing a lifetime history of coronary disease based on their lifetime history of substance use A one class model was created initially, where it was assumed that there were no differences by substance use for the entire sample. Then, an iterative model building process commence d where each new model include d one more class than the model prior to it. This model building process continue d until an optimal fit was achieved. Optimal fit was measured through the use of the Bayesian Information Criterion (BIC), Vuong Lo Menedel Rubin Likelihood Ratio Test (VLMR), and Entropy. BIC is a Bayesian based measure of model fit, with a better fit indicated by a smaller value than the value in the single group model. The VLMR is also a measure of model fit that looks specifically at whether a k class solution is superior to a k 1 class solution, and where the best fitting models have a VLMR p value of <0.05. Entropy is a measure of how well the model is classifying each individual, with higher values indicating better classification. Each of these statistical measures help ed to identify the best number of latent classes, but they did no t replace the use of logic and theory. A combined statistical and theoretical approach determine d the final number of latent classes that were assigned. Multinomial Logistic Regression Multinomial logistic regression was employed in C hapter 5 to understand the relationship between the latent classes and lifetime history of coronary disease. Multinomial logistic regression, as opposed to multiple logistic regression, is used when the dependent variable has multiple, nominal categories such that a traditional logistic regression cannot be used (Engel, 1988) Similar to multiple logistic regression, multinomial logistic regression does not assume

PAGE 86

86 linearity, normality, or homoscedasticity. It does, however, assume independence of irrelevant alternatives (IIA), meaning that the odds of be ing in one class over another are not dependent on the presence or absence of other alternatives. The analytic process for multinomial logistic regression mirrors that of multiple logistic regression: 1) identifying relevant covariates to be included, 2) assessing multicollinearity, 3) assessing confounding, 4) model building, and 5) assessing model fit. 2 a pseudo R 2 was used to estimate model fit (Allison, 2014; McFadden, 1974) It is important to remember that a pseudo R 2 R 2 statistic, so it is not possible to say how much of the variance in the model is explained by the variables included. However it is true that the higher the value, the better the fit (McFadden, 1974) However, McFadden (1979) cautions users of the pseudo R 2 that its values tend to be much lower than those of a traditional R 2 analysis do not ap ply when using the psudeo R 2

PAGE 87

87 Figure 3 1. Infographic depicting the HealthStreet mission and aims to community members interested in becoming HealthStreet members ; published in 2016

PAGE 88

88 Table 3 1. HealthStreet Intake Form questions used in analys es Q uestion Number Question Answer Choices (if any) 9 Sex 1 Male; 2 Female; 3 Transgender 11 Race/Ethnicity 1 American Indian/Alaskan Native; 3 Asian; 4 Black/African American; 6 Native Hawaii a n/Pacific Islander; 7 White; 9 Other 14 Age Continuous; in years 43 Last Grade Completed Continuous; numeric grade 44 Are you employed full or part time? 1 No; 5 Yes 47b Have there been times in the last 12 months when you did not have enough money to buy food that you or your family needed? 1 No; 5 Yes 49b BMI Continuous; calculated based on self reported height and weight 53 Do you have any type of medical insurance? 1 No; 5 Yes 53a If yes, what kind of medical insurance do you have? 62b Have you ever been told you had, or have you ever had a problem w ith Type 2 Diabetes? 1 No; 5 Yes 65b Have you even been told you had, or have you ever had a problem with chest pain or angina? 1 No; 5 Yes 65d Have you even been told you had, or have you ever had a problem with coronary artery disease? 1 No; 5 Yes 65e Have you even been told you had, or have you ever had a problem with heart attack? 1 No; 5 Yes 65h Have you even been told you had, or have you ever had a problem with high blood pressure? 1 No; 5 Yes 65i Have you even been told you had, or have yo u ever had a problem with high cholesterol? 1 No; 5 Yes 68a Have you even been told you had, or have you ever had a problem with ADD/ADHD? 1 No; 5 Yes 68b Have you even been told you had, or have you ever had a problem with anxiety? 1 No; 5 Yes 68c Have you even been told you had, or have you ever had a problem with autism? 1 No; 5 Yes 68d Have you even been told you had, or have you ever had a problem with bipolar disorder? 1 No; 5 Yes 68e Have you even been told you had, or have you ever had a problem with depression? 1 No; 5 Yes 68f Have you even been told you had, or have you ever had a problem with eating disorders? 1 No; 5 Yes 68g Have you even been told you had, or have you ever had a problem with mania? 1 No; 5 Yes

PAGE 89

89 Table 3 1. Continued Question Number Question Answer Choices (if any) 68h Have you even been told you had, or have you ever had a problem with personality disorder? 1 No; 5 Yes 68i Have you even been told you had, or have you ever had a problem with schizophrenia? 1 No; 5 Yes 68j Have you even been told you had, or have you ever had a problem with any other mental health problem? 1 No; 5 Yes 81a FOR MEN: Within the last 30 days have you had more than 4 drinks like beer, wine, liquor in a single day? 1 No; 5 Yes 81b FOR WOMEN: Within the last 30 days have you had more than 3 drinks like beer, wine, liquor in a single day? 1 No; 5 Yes 82 Have you ever used a club drug like Ecstasy, GHB, or Ketamine? 1 No; 5 Yes 82a If yes, have you used a club drug in the last 30 days? 1 No; 5 Yes 83 Have you ever used cocaine or crack? 1 No; 5 Yes 83a If yes, have you used cocaine or crack in the last 30 days? 1 No; 5 Yes 84 Have you ever used marijuana? 1 No; 5 Yes 84a If yes, have you used marijuana in the last 30 days? 1 No; 5 Yes 85 Have you ever used heroin? 1 No; 5 Yes 85a If yes, have you used heroin in the last 30 days 1 No; 5 Yes 86 Have you ever used speed or amphetamines? 1 No; 5 Yes 86a If yes, have you used speed or amphetamines in the last 30 days? 1 No; 5 Yes 91 Have you ever used hallucinogens? 1 No; 5 Yes 91a If yes, have you used hallucinogens in the last 30 days? 1 No; 5 Yes 92 Have you ever smoked cigarettes? 1 No; 5 Yes 92a If yes, have you smoked cigarettes in the last 30 days? 1 No; 5 Yes 92aa Have you ever smoked cigars, or cigarillos like Black & Mild, Swisher Sweets? 1 No; 5 Yes 92aa_a If yes, have you smoked cigars or cigarillos in the last 30 days? 1 No; 5 Yes 92b Have you ever used a hookah, bong, or other pipe to smoke tobacco? 1 No; 5 Yes 92b_a If yes, have you used a pipe to smoke in the last 30 days? 1 No; 5 Yes

PAGE 90

90 Table 3 2 HealthStreet Intake Form question number, variable name, and coding strategy used in analyses Intake Form Question Number Variable Name Coding Strategy 9 sex 0 Male 1 Female 11 race2 1 Black/African American 2 White 3 Other 14 age continuous 43 education 1 > 12 years 44 employment 0 unemployed 1 employed 47b food insecure 0 not food insecure 1 food insecure 49b BMI continuous 53 insured 0 uninsured 1 insured 53a medical insurance 0 no insurance 1 public insurance 2 private insurance 62b type_2_diabetes 0 no history 1 positive history 65b angina 0 no history 1 positive history 65d CAD 0 no history 1 positive history 65e MI 0 no history 1 positive history 65h bp 0 no history 1 positive history 65i chol 0 no history 1 positive history 68a, c, d, f j othermh 0 no history 1 positive history 68b anxiety 0 no history 1 positive history 68e depression 0 no history 1 positive history 81a,b binge 0 no history 1 positive history

PAGE 91

91 Table 3 2. Continued Intake Form Question Number Variable Name Coding Strategy 82, a club_drug 0 never use 1 former use 2 current use 82, a club_drug2 0 never use 1 lifetime use 83, a cocaine 0 never use 1 former use 2 current use 83, a cocaine2 0 never use 1 lifetime use 84, a cannabis 0 never use 1 former use 2 current use 84, a cannabis2 0 never use 1 lifetime use 85, a heroin 0 never use 1 former use 2 current use 85, a heroin2 0 never use 1 lifetime use 86, a speed 0 never use 1 former use 2 current use 86, a speed2 0 never use 1 lifetime use 91, a hall 0 never use 1 former use 2 current use 91, a hall2 0 never use 1 lifetime use 92, a, aa, aa_a, b, b_a tob 0 never use 1 former use 2 current use 92, a, aa, aa_a, b, b_a tob2 0 never use 1 lifetime use

PAGE 92

92 CHAPTER 4 THE ASSOCIATION BETWEEN CANNABIS USE AND CORONARY DISEASE Introduction Cannabis is the most commonly used federally controlled, and therefore federally illicit drug in the United States according to the 2015 National Survey on Drug Use and Health (NSDUH) (Bose et al., 2016) While u se of most illicit drugs in the United States has either stabilized or de creased in the past decade, cannabis use has increased and is the primary contributing f actor to the overall increase in illicit drug use in the United States (Center for Behavioral Health Statistics and Quality, 2016) Estimates of prevalence of cannabis use vary depending on the national survey used. Data from the 2015 NSDUH a nationally representative survey of civilian, non institutionalized individuals age 12 and older sponsored by the Substance Abuse and Mental Health Services Administration show the nationwide lifetime prevalence rate of cannabis use in those age 12 and older is 44 % (Center for Behavioral Health Statistics and Quality, 2016) D ata from the 2015 Monitoring the Future Survey (MTF) a long term national survey of American adolescents, college students and high school graduates through age 55, s how lifetime prevalence rate of cannabis use among those age 18 at 45% and lifetime prevalence rate of cannabis use among those modal age 55 (or the graduating class of 1978) at 81% (Johnston et al., 2016a) With one in two having used marijuana in their lifetime, past year prevalence is expected to be high. Indeed, t he national prevalence rate of past year cannabis use is estimated at 9.5% or almost one in ten based on data from the National Epidemiologic Survey on Alcohol and Related Conditions III (NESARC III) a cross sectional, nationally representative survey based on the civilian, non institutionalized population (Hasin et al., 2015) Past 30 day use rates are estimated by the NSDUH at 8.4% among those 12 years of age and older (Bose et al., 2016).

PAGE 93

93 Given these rates, about 22.2 million Americans age 12 and older currently use marijuana. Due to the high rates of current use among teens and young adults, this number falls to about 13.6 mi llion age 26 or older (Bose et al., 2016) Based on data from the 2015 NSDUH, males age 12 and older are more likely to be lifetime cannabis users and current users compared to females (48 .9% vs 39.4 and 10.6% vs 6.2% respectively ). Additionally, Non Hispanic Whites have a higher prevalence rate of lifetime cannabis use compared to Non Hispanic Blacks or African Americans (43.3% vs 49.1%), however they have a lower prevalence rate of current cannabis use compared to Blacks or African Americans (8.4% vs 10.7%) (Center for Behavioral Health Statistics and Quality, 2016) Cannabis use is also associated with low socioeconomic status (Lemstra et al., 2008) other drug use, and a variety of mental health conditions For instance, the in itiation of cannabis use is frequently concurrent with alcohol and tobacco use du ring adolescence ; initiation of other drugs often follows (Agrawal et al., 2006; Haberstick et al., 2014; Kandel, 2003; Redonnet et al., 2012) Additionally, simultaneous use of tobacco and cannabis has been reported in adolescents and adults (Schauer et al., 2017) as well as simultane ous use of tobacco, alcohol, and cocaine ( Lange and Hillis, 2001) The literature is equivocal on the existence of an association between depres sion and cannabis use (Degenhardt et al., 2001; Feingold et al., 2015; Lev Ran et al., 2013; Manrique Garcia et al., 2012) Literature also indicates an association between cannabis use and anxiety disorde rs (Buckner et al., 2012; Patton et al., 2002) personality disorders and schizophrenia (Agrawal, Nurnberger Jr., & Lynskey, 2011; Copeland, Rooke, & Swift, 2013; McGrath, Welham, Scott, & et al, 2010) Regardless of its association with mental health conditions and other substance use, public opinion surrounding cannabis use has changed drastically over the past 15 years with the

PAGE 94

94 legalization of both medical and recreational cannabis use in numerous states (Davis et al., 2015) Currently, 29 states and the District of Columbia have laws legalizing some form of cannabis use, with 7 states and the District of Columbia having legalized recreational use of cannabis (MacCoun, 2017) In addition to the increased number of states legalizing cannabis for recreational use, perception of cannabis use over the past 15 years has changed Coloradans perceived a lower risk of cannabis use and medical (Schuermeyer et al., 2014) Data from the NSHDUH also indicates a significant decrease in perceived great risk of cannabis use, and a significant increase in daily cannabis use from 2002 2012 (Pacek e t al., 2015) The nat ional increase in cannabis use coupled with decreased risk perception, suggests that cannabis use will continue to rise Against this backdrop of increased cannabis use, coronary disease continues to be a leading cause of morbidit y and mortality in the United States (Heron, 2015) Though rates of coronary disease are similar among males and females, females tend to develop the disease later in life and report different manifestations of first events (Leening et al., 2014; Maas and Appelman, 2010) Coronary disease is linked to low soci o economic status, low education level, and unemployment (Albert et al., 2006; Eng et al., 2002; Naimi et al., 2009) Coronary disease is als o associated with a lcohol use (particularly binge drinking), tobacco use, and cocaine use (Chiva Blanch et al., 2013; De Giorgi et al., 2012; McBride, 1992; Thylstrup et al., 2015) as well as depression, anxiety and other mental health conditions T here is strong evidence for an association between depression and coronary disease (Ariyo et al., 2000; Baune et al., 2012; Carney and Freedland, 2016; Elderon and Whooley, 2013; Frasure Smith N et al., 1993; Gonzlez and Tarraf, 2013; Hare e t al., 2013; Joynt et al., 2003) Additionally, t here is evidence

PAGE 95

95 that anxiety and borderline personality disorder are associated with coronary disease (Powers and Oltmanns, 2013; Thurston et al., 2013 ) Cardiovascular diseases, including coronary disease, were responsible for 14 % of national health expenditures in the years 2012 2013 (Benjamin et al., 2017) and dir ect medical costs related to these conditions are expected to triple to $818 billion by the year 2030, making the burden of disease in the United States substantial (Heidenreich et al ., 2011) Given the prevalence of coronary disease, identification of any behavioral risk factors associated with even a small degree of increased risk is a priority (National Academies of Sciences, Engineering, and Medicine, 2017) A number of researchers have suggested an association between cannabis use and coronary disease (Aryana & Willi ams, 2007; Gordon et al., 2013; Sidney, 2002) based on evidence that activation of the cannabinoid 1 receptor (CB1) results in vasodilation (Montecucco and Di Marzo, 2012; Rajesh et al., 2010) that cannabis induces production of catecholamines such as epin ephrine and norepinephrine (Gordon et al., 2013; Jones, 2002) and that cannabis use results in increased platelet activation (Dahdouh et al., 2011; Thomas et al., 2014) Franz and Frishman, ( 2016) reviewed numerous case reports purporting acute coronary events after cannabis use. Both human and animal models suggest these acute eve nts may be due to coronary arterial vasospasm (Franz and Frishman, 2016) To date, only 16 studies have been identified assessing the association between cannabis use and outcomes such as obesity, hypertension, dyslipidemia, diabetes, metabolic syndrom e, and coronary disease (systematically reviewed in C hapter 2). (Aronow and Cassidy, 1974; Bancks et al., 2015; Blackstone and Herrmann, 2016; Dube et al., 2015; Hayatbakhsh et al., 2010; Huang et al., 2013a; Mittleman et al., 2000; Pen ner et al., 2013; Racine et al., 2015;

PAGE 96

96 Rajavashisth et al., 2010; Rodondi et al., 2006; Ross et al., 2016; Thompson and Hay, 2015a; Vidot et al., 2016; Waterreus et al., 2016; Yankey et al., 2016) Additionally, the association between cannabis use and cardiovascular mortality has been considered by two teams (Frost et al., 20 13; Sidney et al., 1997) We seek to fill a gap in the literature by assessing the association specifically between can nabis use and coronary disease Methods Sample The sample for these analyses comes from HealthStreet a community engagement model an d outreach program which is (Cottler, LB PI) HealthStreet assesses community member health condition s and concerns, and links community members to health research opportunities and medical and social services. HealthStreet at the University of Florida began in November 2011, and has engaged over 8, 8 00 commun ity members in the Gainesville, Jacksonville and Miami areas through a community health worker model (Cottler et al., 2012, 2010) Each new HealthStreet mem ber is intervi e wed by a CHW using the Health Intake Form. The Health In take Form is an in person questionnaire that after informed consent, takes about 3 0 minutes and assesses health conditions including mental health conditions medical health concerns, history of substance use, and willingness to participate in health research. Measures Exposure The exposure of interest was history of cannabis use categorized as never use, former use (lifetime use but not in the last 30 days) and current use (use in th e last 30 days) The instrument used to gather this data was the Health Intake Form History of cannabis use was elicited by

PAGE 97

97 Have you ever used marijuana ? [Y/N] Those who answer yes to having ever used categorized as former users, and th This categorization of cannabis use is frequently seen in the literature as the exposure measure in cross sectional analyses (Penner et al., 2013; Racine et al., 2015; Thompson and Hay, 2015a; Vidot et al., 2016; Waterreus et al., 2016) Outcome Self reported lifetime history of coronary disease was assessed using the Health Intake Form. Coronary disease was defined as endorsing a lifetime history of one or more of the following: Angina, Coronary Artery Disease (CAD), or Myocardial Infarction (MI). The Health Intake Form question reads oblem coded as having a lifetime history of coronary disease. Covariates Socio demographic information for this analysis was recorded through the Health Intake Form and included self reported sex race, age, health insurance status, education level, food insecurity status, and employment status Race was categorized as Black o r African American, White, or Other. Age was a continuous variable for this analysis. Education level was initially recorded as the last grade completed, and then was categorized as 12 or fewer years of education and greater than 12 years of education. Foo d insecurity is times in the last 12 months when you did not have enough money to buy food that you or your family neede Health insurance

PAGE 98

98 status and employmen t status are In addition to socio demographic covari ates, measu res of hypertension, hypercholesterolemia type II diabetes, and body mass index (BMI) were considered important covariates. T he Health Intake Form question reads you ever had a problem wi th high blood pressure? [ Those answering yes were coded as having a lifetime history of that condition. The BMI covariate was created using the reported height and weight, and then categoriz ed as underweight, normal weight, overweight, obese according to World Health Organization standards (World Health Organization, 2000) Non cannabis s ubstance use covariat es were measured using 3 categories: never use, former use (lifetime but not l ast 30 day use ) and current use (last 30 day use). Lifetime For those respondents answering yes the follow up question was asked combined Health Intake Form substance use questions asking about cigarette use, cigar or cig arillo use, and hookah or bong use. The Health Intake Form does not assess never, former and current alcohol use, however the last 30 days, have you had more than (4 drinks for men, 3 drin ks for women) like beer, wine,

PAGE 99

99 Mental health disorder covariates assessed were lifetime self reported history of depression, anxiety, or other mental health conditions (ADD/ADHD, autism, bipolar disorder, eating disorders, mania, personality disorder, schizophrenia) These were elicited using the Analysis The data analysis for this paper was generated using SAS software, version 9.4 of the SAS System for Windows. Copyright 2002 2012 SAS Institute Inc. First, univariate analyses were conducted using chi the relationship between the soci o demographic, substance use, mental health, and coronary disease risk factor variables and lifetime history of coronary disease. Second, i nitial crude odds ratios were generated for the relationship between cannabis use and coronary disease. Second poss ible confounding by each of the covariates was assessed by determining if there was a statistically significant (p 0.05) association between the covariate and independent variable, as well as the covariate and the dependent variable in the unexposed. Th ird, t hree logistic regression models were then built, the first controlling for socio demographic factors (age, sex, race, education level). The second model controlled for socio demographic factors as well as mental health conditions and substance use (d epression, anxiety, other mental health conditions, tobacco use, and cocaine use). The third and final model controlled for socio demographic factors, mental health conditions, substance use, and coronary disease risk factors (hypertension, hypercholestero lemia, type II diabetes and BMI). The third model is considered the fully adjusted model. A ll models were generated using a manual backward elimination process where all possible variables were included at first, then removed one at a time an d evaluated for their effect on the odds ratio Parsimony was achieved by

PAGE 100

100 elimina ting covariates that impacted the odds ratio by less than 10%. For example, in Model 2, the removal of amphetamine use, club drug use, and hallucinogen use from the model resul ted in almost no change to the odds ratio compared to the full model, thus they were eliminated from the model. As state d previously, age, sex, race, and education level were left in all models. The f ully adjusted model was assessed for model fit using R square values, the c statistic, and the Hosmer and Lemeshow Goodness of fit test. Values for the c statistic between 0.7 and 0.8 are considered to show acceptable discrimination, values between 0.8 and 0.9 are considered to show excellent discrimination an d values greater than or equal to 0.9 are considered to show outstanding discrimination (Hosmer and Lemeshow, 2000) The p value of the Hosme r and Lemeshow Goodness of fit t est should be non signif icant, which indicates the model is well calibrated so that probability predictions from the model reflect the occurrence of events in the data (LaValley, 2008) Multicolinearity was assessed using the variance inflation factor (VIF), where a VIF >10 indicated collinearity problems. When the final model was assessed for multicollinearity using the VIF, no problems were identified. Power analysis was conducted using proc power in SAS, where the ordered categorical was the test predictor and had response probabilities 0.51 for never use, 0.32 for past users, and 0.17 for current users. Given the available sample size of 8,453 and a desir ed alpha of 0.05, the analysis had 80% power to detect a statistically significant odds ratio of 1.08 for Results Of the 8,453 total respondents, 1 07 1 respondents ( 12.7 %) indicated a lifetime histor y of coronary disease. Of those 1 071 respondents 899 indicated a lifetime history of angina (83.4%) 1 14 indicated a lifetime history of CAD (10.6%) and 2 68 indicated a lifetime history of MI (25.0%) All three conditions were reported by 48 individuals (4.5%) Fewer than ha lf the sample

PAGE 101

101 rep orted any cannabis use ( 48.8 %) U nder one third (32.6%) report ed former cannabis use, and 16.2 % reported current cannabis use. Table 4 1 gives the socio demographic characteristics of the entire sample and by coronary disease status. The average age of the total sample was 43 .6 years, with 58.5 % of respondents identifying as female and 61.5 % of respondents identifying as Black or African American. An equal proportion had more than 12 years of education ( 59.9 %) and some form of health insurance (59 .4%: 26.8% public and 32.6 % private). Only 34.5 % of the sample was employed either full or part time. About one third of the sample ( 34.8 %) endorsed a lifetime history of hypertension, 18.7 % endorsed a lifetime history of hypercholesterolemia and 11.2% endorsed a lifetime history of type II diabetes Four percent of the sample was underweight, 29.8 % had normal weight, 28.7 % was overweight and 37.3 % of the sample was obese based on the calculated BMI. HealthStreet members reporting a lifetime h istory of coronary disease were older than members who did not report ever having coronary disease (47.3 years vs 43.0 years) There were no significant differences in sex or race by history of coronary disease. Those with a lifetime history of c oronary di sease were significantly more frequently covered by private insurance ( 41.0 % vs 31.4 %), and were less frequently employed ( 21.2 % vs 36.4 %). They also reported a lifetime history of hypertension ( 59.9 % vs 31.2 %), hypercholesterolemia (39. 6 % vs 15. 7 %), diabe tes (20.3% vs 9.9%), and obesity ( 45.8 % vs 36.1 %) significantly more frequently. Table 4 2 gives the substance use and mental health conditions for the entire sample and by coronary disease status. As mentioned above, former cannabis use was reported by 32.6% of the sample, and current cannabis use was reported by 16.2% of the sample meaning 48.8% of sample reported a lifetime history of cannabis use. A little under half the sample (46.5%) never

PAGE 102

102 used tobacco, while 18.5% reported former tobacco use. A li ttle more than one third of the total sample current ly (in the last 30 days) used tobacco (35.1 %) Most of the sample (81.8%) never used cocaine while only 1.9 % were current cocaine users. The majority of the sample never used heroin (97.0%), amphetamines (92.6%), club drugs (93.5%), or hallucinogens (93.0%). A little more than one quarter ( 27.4 %) endorsed a lifetime history of depression, 22.6 % endorsed a lifetime history of anxiety, and 20.6 % endorsed a lifetime history of other mental health conditions Of HealthStreet members reporting ever having depression, 58.7% also reported ever having anxiety (data not shown). Those with a lifetime history of coronary disease reported former and current cannabis use significantly more frequently than those without a lifetime h istory of coronary disease (36.3% vs 32.1% and 18.1% vs 16.0 %). Additionally, there were significant differences in tobacco use, cocaine use heroin use, amphetamine use and hallucinogen use by coronary disease status. There were, however, no significant differences in club drug use or recent binge drinking by coronary disease group. Those with coronary disease were significantly more likely to report a lifetime history of depression (47.0% vs 24.1%) anxiety (37.8% vs 19.6%) or other mental health conditions ( 35.0% vs 18.6% ). Table 4 3 gives the odds ratios and 95% confidence intervals for the association between cannabis use and coronary disease for each of the three logistic regression models In M odel 1 (controlled for age, sex race, and education level), those who endorsed former cannabis use, and current cannabis use had significantly increased odds of endorsing a lifetime history of coronary disease compared to those who endo rsed never cannabis use (OR 1.35, 95% CI 1.16, 1.56 and OR 1.43 95% CI 1.18, 1.73 respectively). Though sex and race did not vary significantly by

PAGE 103

103 coronary disease status in univariate analyses, they were left in the logistic regression models because of previous research showing sex and race differences in both cannabis use and coronary disease (Center for Behavioral Health Statistics and Quality, 2016; Leening et al., 2014; Maas and Appelman, 2010; Mozaffarian et al., 2015) A fter controlling for mental health conditions and other substance use (Model 2) as well as coron ary disease risk factors (Model 3) these significant associations were attenuated Adjusted analyses for the association between cannabis use and each component of coronary disease (angina, CAD, MI) were also non significant (data not shown) Discussion T his is one of the first analys e s of the association between cannabis use and coronary disease. The prevalence rate of angina among HealthStreet respondents (10.6%) was slightly higher than expected given the average age of respondents, while the overall p revalence rate of CAD (1.4%) among HealthStreet respondents was slightly lower than expected, and the prevalence rate of MI among respondents (3.2%) was approximately in line with the national prevalence rate where national prevalence rate is based on data from the National Health and Nutrition Examination Survey presented in the Heart Disease and Stroke Statistics 2016 Update, (Mozaffarian et al., 2015) The prevalence rate of lifetime cannabis use in our sample of those 18 and older was about 48.8 % (former and current use combined) which is slightly higher than the national estimate o f 44 % from the 2015 NSDUH sampl e of those age 12 and older (Center for Behavioral Health Statistics and Quality, 2016) The higher prevalence rate in our sample may be because the median age of our sample was 45, which is greater than the median age of 37.3 in the NSDUH sa mple (as defined by the US census bureau) thus HealthStreet members had a longer period of time in which they could have used cannabis.

PAGE 104

104 The prevalence rate of cu rrent use in our sample was 16.2 %, which is higher than the national estimate of 8.3% for past month use also based on data from the 2015 National Survey but within the past 12 months, or 3) more than 12 months ago (Center for Behavioral Health Statistics and Quality, 2016) A higher percent of HealthStreet members were out of work compared to NSDUH participants which may have partially influenced patterns of current cannabis use Overall, our sample was more apt to endorse food insecurity compar ed to the national average (47.2 % vs 14.0%) reported by the United States Department of Agriculture based on data collected in a supplement to the Current Population Survey (Coleman Jensen et al., 2014) Our sample was also more apt to endorse having a high school education or less compared to the national average ( 40.2 % vs 11.6%) reported by the United States Census Bureau (Coleman Jensen et al., 2014; Ryan and Bauman, 2015) Additionally, our sample over represented females ( 58.5 %) and Black or African Americans ( 61.5 %) compared to national estimates from the United States Census B ureau (50.8% female; 13.3% Black or African American) (U.S Census Bu reau, 2015) The HealthStreet sample is a community sample recruited by community health workers in places where individuals congregate thus i ndividuals with the time to engage in the study may over represent unemployed or part time workers Difference s seen in socio demographic characteristics between the HealthStreet sample and national statistics may be due in part to this recruitment method Prevalence rate of current (past month) binge drinking in the HealthStreet sample was 23.5%, similar to the national prevalence rate estimate of 24.9% for past month binge drinking

PAGE 105

105 R : our question asks about 4 or more drinks in a single day for men and 3 or more drinks in a single day for women ; the 2015 NSDUH asks about 5 or more drinks for males on the same occasion and 4 or more drinks for females on the same occasion (Center for Behavioral Health Statistics and Quality, 2016) T he lifetime prevalence rate of any tobacco pr oduct use in our sample was also lower ( 53.6 %) compared to the NSDUH based national estimate of 63.9% (Bose et al., 2016; Center for Behavioral Health Statistics and Quality, 2016 ) Similarly, combined lifetime and current tobacco use in Florida rank in the bottom half of the United States (United Health Foundation, 2015) The lifetime prevalence rate of co caine use in our sample was 18.2 %, the lifetime prevalence rate of heroin use was 3.0%, and the lifetime prevalence rate of amphetamine use was 7.4%, all of which were only slightly higher than the NSDUH 2015 national prevalence rate estimates (14.5%, 1.9%, and 5.4% respectively) R esults for the unadjusted associations between socio demographic factors and coronary disease were generally consistent with previous literature. For instance, previous research indicates an inverse association between employment status and coronary disease (Naimi et al., 2009) W e did not find that lower educational level was associated with coronary disease Contrary to previous literature (Albert et al., 2006) our findings indicated significantly higher education level among those with a l ifetime history of coronary disease compared to those without a lifetime history of coronary disease. However, the study by Albert and colleagues (2006) only included women, so the results of that study are not directly comparable to our findings. Former and current use of cannabis, tobacco, cocaine, heroin, and amphetamines was significantly greater among those with a lifetime history of coronary disease compared to those

PAGE 106

106 without coronary disease, in line with previous literature and our expectations (Chiva Blanch et al., 2013; De Giorgi et al., 2012; McBride, 1992; Thylstrup et al., 2015) Also in line with our expectations were the results of the first multiple logistic regression model adjusted for age, sex, race and education level (Model 1) This model showed t he odds of endorsing a lifetime history of coronary disease were significantly greater in those who e ndorsed former cannabis use, or current cannabis use compared to those who never used cannabis (OR 1.29, 95% CI 1.12, 1.51 and OR 1.33 95% CI 1.10, 1.62 respectively) ; however after controlling for mental health conditions, other substance use (Model 2) and coronary disease risk factors such as hypertension, hypercholesterolemia and BMI (Model 3) this significant association was attenua ted There are a few possible reasons for this result. I t was hypothesized that cannabis use is directly related to coronary disease, even after confounding is controlled for. The results of this analysis do not support that hypothesis. I t may be that can nabis use is not independently associated with coronary disease and that the association seen in univariate analyses and Model 1 was a result of confounding due to factors known to be associated with both cannabis use and coronary disease Mental health co nditions important confounding factors, attenuated our results when added to the model (Model 2) In C hapter 2 it was noted that the extant literature has failed to adequately control for mental health conditions but our results combined with evidence fo r associations between mental health conditions and both cannabis use and coronary disease suggest it is imperative to consid er mental health conditions Additionally, any direct relationship between cannabis and coronary disease may be confounded by relat ionships between cannabis and co occurring substance use. This supports findings from Rodondi and colleagues (2006), who noted that associations between cannabis use and BMI, hypertension and dyslipidemia were

PAGE 107

107 attenuated in adjusted analyses primarily con founded by other substance use such as alcohol and tobacco. Another possible reason for the non signifi cant findings seen in Models 2 and 3 is that any relationship between cannabis use and coronary disease is the result of indirect effects or mediation Figure 1 1 gives a model of the genesis of coronary disease. Arrows go directly from behavioral risk factors to coronary disease, but they also go from behavioral risk factors to physiological risk factors to coronary disease. The results of the systematic review presented in C hapter 2 indicate d that cannabis use may be associated with, for example, obesity, and obesity is in turn associated with coronary disease. Molecular studies show that CB1 activation may result in increased weight and poor metabolic function, which contribute to obesity and coronary disease (Nissen, Nicholls Wolski, & et al, 2008; Pi Sunyer et al., 2006; Van Gaal, Rissanen, Scheen, Ziegler, & Rssner, 2005) If it is true that factors like obesity mediate the relationship between cannabis use and corona ry disease, then controlling for a mediating factor would render the model non significant, as is seen with Models 2 and 3. There are some important limitations to this analysis that need to be considered when interpreting the results. First this analysis had 80% power to find an odds ratio of 1.08, but in Model 3, former cannabis users had 1.05 times the odds of endorsing a lifetime history of coronary disease compared to never users, and current cannabis users had 1.0 5 times the odds of endorsing a lifet ime history of coronary disease compared to never users. As such, it could be that non significant findings resulted from an underpowered analysis. Second our community sample may not be representative. W e have an oversampling of Black/African Americans, women, and low income populations in our sample Additionally, the average age of our sample is 43.6 years. As such many individuals in our sample may still be too young to have detectable

PAGE 108

108 negative health effects related to continued cannabis exposure, or to have experienced coronary disease outcomes. Nonetheless, the sample provides a baseline understanding of the association between cannabis use and coronary disease. Third our sample is cross sectional, meaning we cannot make any conclusions about tempor ality, but instead can only report associations. Fourth cannabis use, coronary disease, and all covariates were measured through self report. S elf report may be subject to increased bias, for example, Johnson and Fendrich (2005) found that social desirability bias was predictive of drug use under reporting, while memory difficulties were predictive of drug use over reporting. However, research has found g ood reliability for self report of cannabis (Mennes et al., 2009) and during the consent process for HealthStreet CHWs clarify the risks of disclosure are minimized through internal controls and a federal certificate of confidentially. T he literature also indicates agreement betw een self report and medical records for M I among both genders, and agreement for angina among women (Okura et al., 2004; Simpson et al., 2004) alt hough Okura colleagues (2004) note that agreement is worse among males, those who are older than 65 and those with less than 12 years of education. Fifth our questions did not capture frequency, quantity or duration of cannabis use. As such, we were unable to examine possible dose response relationships between cannabis use and coronary disease. Additionally, we were unable to assess potency. Between 1995 and 2014, the ratio of THC to cannabidiol has increased from about 14:1 to about 80:1 (ElSohly et al., 2016) Given the proposed biological mechanism for the association between cannabis use and coronary dis ease is activation of the endocannabinoid receptors by THC, change in cannabis potency over time is an important factor we were unable to consider. Sixth, the R square value for the final model (Model 3) shows that the model explains less than 10% of the v ariance However, the model fits well based on the Hosmer Lemeshow goodness of fit statistics, and shows acceptable

PAGE 109

109 discrimination of subjects based on the c statistic (Austin and Steyerberg, 2012; Hosmer and Lemeshow, 2000; LaValley, 2008) Although our parsimonious model cap tured common c ovariates such as hypertension, hypercholesterolemia, BMI, depression, anxiety, and tobacco use were statistically significant, factors like family history physical inactivity, and diet were not included and would have likely increased the a mount of variance explained (LaValley, 2008) Coronary disease is multi faceted (see Figure 1 1), with risk factors spanning multiple domains such as socio e conomic status, demographics, environment, substance use, and genetics (Kessler et al., 2013; Khot et al., 2003; Roberts, 2014) Future research should strive to assess more of the multiple domains of risk. In conclusion, our fully adjusted model (Model 3) did not show a significant relationship between cannabis use and coronary disease. This is one of the first analyses to asses this relationship and confidence in the analysis is strengthened by the number of relevant covariates included. Given the push for legalization of both medicinal and recreational cannabis use it is important to understand the health consequences likely to follow increased use This r esearch fills a gap in the current literature and addresses a growing public health concern. Future re search should focus on improving our understanding of how cannabis use impacts chronic disease outcomes using a comprehensive approach

PAGE 110

110 Table 4 1. Socio demographic characteristics and physiological risk factors for coronary disease in the total HealthStr eet sample, and by lifetime self reported history of coronary disease Total Sample Coronary Disease n= 8453 n= 7382 (87.3%) + n= 1071 (12.7%) p value Mean Age in years (SD) 43.6 (16.2 ) 43.0 (16.3) 47.3 (14.9) <0.0001 Sex Female 4942 (58.5 %) 4317 (58.5 %) 625 (58.5 %) 0. 9996 Race Black 5194 (61.5 %) 4535 (61 .4%) 659 (61 .5%) 0.9 483 White 2679 (31.7 %) 2338 ( 31.7 %) 3 41 (31 .8%) Other 580 (6.7 %) 509 (6.9 %) 71 (6.6 %) Education Level 3387 (40.2 %) 3027 (41.1 %) 360 (33.7 %) <0.0001 > 12 Years 5048 (59.9 %) 4340 (58.9 %) 708 (66.3 %) Food Insecure Yes 3965 (47.2 %) 3300 (45.0 %) 665 (62.3 %) <0.0001 Health Insurance None 3349 (40.6 %) 2913 (40.4 %) 436 (41.5 %) <0.0001 Public 2215 (26.8 %) 2031 (28.2 %) 184 (17.5 %) Private 2693 (32.6 %) 2262 (31.4 %) 431 (41.0 %) Employed Yes 2899 (34.5 %) 2673 (36.4 %) 226 (21.2 %) <0.0001 Hypertension Yes 2930 (34.8 %) 2297 (31.2 %) 633 (59 .9 %) <0.0001 Hypercholesterolemia Yes 1563 (18.7 %) 1147 (15.7 %) 416 (39.6 %) <0.0001 Type 2 Diabetes Yes 946 (11.2%) 730 (9.9%) 216 (20.3%) <0.0001 Body Mass Index <18.5 359 (4.3 %) 323 (4.4 %) 3 6 (3.4 %) <0.0001 18.5 24.9 2515 (29.8 %) 2267 (30.7 %) 248 (23 .2%) 25 29.9 2428 (28.7 %) 2131 (28.9 %) 297 ( 27.7 %) 30 + 3151 (37.3%) 2661 (36.1 %) 490 ( 45.8 %) *T test used

PAGE 111

111 Table 4 2. C annabis use, other substance use and history of mental health conditions in the total HealthStreet sample, and by lifetime self reported history of coronary disease Total Sample Coronary Disease n=8453 n=7382 (87.3%) + n=1071 (12.7%) p value Cannabis Use Never 4319 (51.2 %) 3831 (52.0 %) 488 (45.6 %) 0.00 05 Former 2754 (32.6%) 2365 (32.1 %) 389 (36.3 %) Current 1370 (16.2 %) 1176 (16.0 %) 194 (18.1 %) Tobacco Use Never 3923 (46.5%) 3539 (48.0 %) 384 ( 35.9 %) <0.0001 Former 1560 (18.5 %) 1324 (18.0 %) 236 (22.1 %) Current 2962 (35.1 %) 2512 (34.1 %) 450 (42.1 %) Cocaine Use Never 6905 (81.8 %) 6102 (82.7%) 803 (75.0 %) <0.0001 Former 1378 (16.3 %) 1150 (15.6%) 228 (21.3 %) Current 163 (1.9 %) 123 (1.7%) 40 (3.7 %) Heroin Use Never 8195 (97.0%) 7169 (97.2%) 1026 (95.9 %) 0.0473 Former 242 (2.9%) 200 (2.7%) 42 (3.9 %) Current 10 (0.1%) 8 (0.1%) 2 (0.2%) Amphetamine Use Never 7813 (92.6 %) 6864 (93.1 %) 949 (88.7 %) <0.0001 Former 597 (7.0%) 483 (6.5%) 114 (10.7 %) Current 30 (0.4%) 23 (0.3%) 7 ( 0.7 %) Club Drug Use Never 7893 (93.5%) 6889 (93.4%) 1004 (94.0 %) 0.2992 Former 506 ( 6.0%) 444 (6.0%) 62 (5.8 %) Current 43 (0.5%) 41 (0.6%) 2 (0.2%) Hallucinogen Use Never 7855 (93.0 %) 6884 (93.4 %) 971 (90.7 %) 0.001 8 Former 560 (6.6 %) 468 (6.4 %) 92 (8.6%) Current 31 (0.4%) 23 (0.3%) 8 (0.8 %) Binge Drinking Current 1 987 (23.5 %) 1 736 (23.5 %) 251 (23.5 %) 0. 9803 Depression Yes 2 307 (27.4 %) 1 803 (24.5 %) 504 (47.3 %) <0.0001 Anxiety Yes 1906 (22.6 %) 1497 (20.3 %) 409 ( 38.4 %) <0.0001 Other Mental Health Conditions Yes 1742 (20.6 %) 1 364 (18.5 %) 3 78 (35.3 %) <0.0001

PAGE 112

112 Table 4 3. Odds ratios and 95% confidence intervals for the associations between can nabis use and coronary disease Model 1 A Coronary Disease OR (95% CI) Model 2 B Coronary Disease OR (95% CI) Model 3 C Coronary Disease OR (95% CI) Never Ref Ref Ref Former 1.35 (1.16, 1.56) 1.05 (0.90, 1.24) 1.06 (0.90, 1.26) Current 1.43 (1.18, 1.73) 0.99 (0.80, 1.24) 1.05 (0.84, 1.31) A Controlled for age, sex, race, and education level c statistic 0.607, R 2 0.0144, Hosmer and Lemeshow Goodness of Fit p value=0.0006 B Controlled for age, sex, race, education level, depression, anxiety, other mental health conditions tobacco use, and cocaine use; c statistic 0.685, R 2 0.0454, Hosmer and Lemeshow Goodness of Fit p value=0.0754 C Controlled for age, sex, race, education level, depression, anxiety, other mental health conditions tobacco use, cocaine use, hypertension, hypercholesterolemia, and BMI; c statistic 0.740, R 2 0.079, Hosmer and Lemeshow Goodness of Fit p value=0.3053

PAGE 113

113 CHAPTER 5 ASSOCIATION BETWEEN CANNABIS USE AND CORONARY DISEASE IN THE CONTEXT OF OTHER SUBSTANCE USE: A LATENT CLASS ANALYSIS Introduction Data from the 2015 National Survey on Drug Use and Health (NSDUH) indicate that cannabis is the most commonly used federally deemed illicit drug in the United States (Bo se et al., 2016) Cannabis use has steadily increased over the past decade, and is the primary contributing factor to the overall increase in illicit drug use in the United States (Hedden et al., 2015) Nationally representative su rveys such as the NSDUH and Monitoring the Future Survey (MTF) indicate l ifetime prevalence of cannabis use ranges from 44% to 81%. In those 12 years of age and older lifetime prevalence of cannabis use is estimated at 44 % ( NSDUH; Center for Behavioral Health Statistics and Quality, 2015) and among those age 55, the lifetime prevalence of cannabis use is estimated to be 81% (MTF; Johnston et al., 2016a) Additionally, data from the national ly representative National Epidemiologic Survey on Alcohol and Related Conditions III (NESARC III) indicate the rate of past year cannabis use to be 9.5% (Hasin et al., 2015) There have been substantial changes in public opinion of cannabis over the past 15 years which has resulted in the legalization of both medical and recreat ional cannabis use in n umerous states. Currently, 8 states and the District of Columbia have legalized recreational use of cannabis, 28 states have legalized medicinal use of cannabis, and there is a national push to decriminalize cannabis use (MacCoun, 2017) Data from NSDUH also indicate a significant decrease in perceived great risk of cannabis use, and a significant increase in daily cannabis use from 2002 2012 (Pacek et al., 2015)

PAGE 114

114 Cannabis is often used concurrently with substa nces such as tobacco; they are initiated within the same time period, with subsequent initiation of other illicit drugs (Agrawal et al., 2006; Haberst ick et al., 2014; Redonnet et al., 2012) It is important, then, to consider lifetime cannabis use in the larger context of patterns of lifetime substance use. Previous research has identified between 2 and 5 latent classes of substance user, depending on the manifest variables (Baggio et al., 2016; Lynskey et al., 2006; Monga et al., 2007) Lynskey and colleagues (2006) used data from the young adult cohort of the Australian Twin Registry, and found 5 latent classes of substance user within that popula tion Individuals in each class identified had high probabilities of cannabis use. This included one latent class that w as comprised of respondents who only used cannabis Both alcohol and tobacco use were significant predictors of membership in this class (Lynskey et al., 2006) The national increase in cannabis use has led researchers to question how cannabis may be associated with outcomes that have high rates of morbidity and mortality such as coronary disease (Yankey et al., 2016) Coronary disease is a leading cause of morbidity and mortality in the United States (Heron, 2015) with t he national prevalence rate of diagnosed coronary disease being about 11.5% (National Center for Health Statistics, 2014) Rates of coronary disease are similar among males and females, however, females tend to develop the disease later in life and report different manifestations of first events (Leenin g et al., 2014; Maas and Appelman, 2010) Coronary disease is linked to a number of socio demographic factors such as low socio economic status, low education level, and unemployment (Albert et al., 2006; Eng et al., 2002; Naimi et al., 2009) as well as tobacco use, and cocaine use (Chiva Blanch et al., 2013; De Giorgi et al., 2012; McBride, 1992; Thylstrup et al., 2015) It is also s trongly linked to depression (Ariyo et al., 2000; Baune et al., 2012; Elderon and Whooley, 2013; Frasure Smi th N

PAGE 115

115 et al., 1993; Gonzlez and Tarraf, 2013; Hare et al., 2013; Joynt et al., 2003; Lim et al., 2012) where d epression independently increases risk for coronary disease by an average of 1.5 times compared to those without depression. (Baune et al., 2012) Coronary disease is also associated with anxiety (Powers and Oltmanns, 2013; Thurston et al., 2013) Previous research suggests a direct relationship between cannabis use and coronary disease (Aryana & Willi ams, 2007; Gordon et al., 2013; S idney, 2002) There is e vidence that activation of the cannabinoid 1 receptor (CB1) results in vasodilation (Montecucco and Di Marzo, 2012; Rajesh et al., 2010) that cannabis induces production of catecholamines such as epinephrine and norepinephrine (Gordon et al., 2013; Jones, 2002) and that cannabis use results in increased platelet activation (Dahdouh et al., 2011; Thomas et al., 2014) C ase reports also indicate an increased risk of acute coronary events after cannabis use. Both human and animal models suggest these acute events may be due to coronary arterial vasospasm (Franz and Frishman, 2016) Additionally, the resu lts of the systematic literature review presented in C ha pter 2 indicates a significant relationship between cannabis use and both angina and myocardial infarction. However, the result s of the analyses presented in C hapter 4 showed that the association between cannabis use and coronary disease is attenuated after adjustment for other substance use (tobacco and cocaine) and mental health conditions (depression and anxiety). These results may suggest that higher rates of coronar y disease among cannabis users may be a result of concurrent cannabis, tobacco and other substance use. Research has already shown that patterns of substance use (cannabis, stimulants, hallucinogens, sedatives, inhalants, cocaine, opioids and solvents) ar e significantly associated with socio demographic factors like age and sex and with psychiatric comorbidities like depression, anxiety and other mental health conditions (Lynskey et al., 2006; Monga et al.,

PAGE 116

116 2007) However, there is no information on possible associations between patterns of substance use and coronary disease Depending on the patterns of substances used, different coronary effects could be expected. Given the continued rise in cannabis use, and the co ncurrent use of cannabis with other illi cit and non ill icit substances it is important to understand how coronary disease is associated with typologies of substance user s This analysis aimed to identify latent classes of HealthStreet members based on lifetime substance use, and to evaluate the association between latent class membership and coronary disease using multinomial logistic regression while controlling fo r socio demographic factors, mental health conditions and coronary disease risk factors Methods Sample HealthStreet, a community health worker (CHW) based community engagement and outreach model first developed by Dr. Linda Cottler, PhD, MPH, at Washington University in St. Louis, is funded by the U and Clinical and Translational Science Institute. HealthStreet CHWs assess community member health conditions and concerns and link c ommunity members to health research opportunities and medical and social services. H ealthStreet has engaged over 8,800 commun ity members in the Gainesville, Jacksonville and Miami areas since it began at the University of Florida in November 2011 (Cottler, Striley, O'Leary, Ruktanonchai, & Wilhelm, 2012; The Health Intake Form is a 30 minute in person survey used by the CHW s to assess each Following the asses sment, referrals for needed services and opportunities to participate in health research are provided by the CHW.

PAGE 117

117 Measures Manifest v ariables Lifetime illicit substance use variables included in the LCA were cannabis use, club drug use, cocaine or crack u se, amphetamine use, hallucinogen use, and tobacco use. Illicit substance use was captured through the Health Intake Form question determined by combining the Health Intake Form binge dr inking is captured b y Health Intake Form, it was not considered a manifest variable for this analysis because binge drinking is only assessed for the last 30 days, and not for lifetime. Heroin use is also measured in the Health Intake Form. It was not incl uded as a manifest variable because only 2.9% of HealthStreet members positively endorsed lifetime use, and its inclusion might have resulted in an artificial latent category Coronary disease Coronary disease was defined as endorsing a lifetime history o f one or more of the following: Angina, Coronary Artery Disease (CAD), or Myocardial Infarction (MI). Associations between coronary disease and the latent classes were calculated based on answers to the HealthStreet Intake Form question, Have you ever bee n told you have, or have you ever had a pr oblem with (condition)? [Y/N] Covariates Socio demographic factors, lifetime history of mental health conditions and lifetime history of coronary disease risk factors were included as covariates Socio demograp hic factors included age, sex race health insurance status, education level, food insecurity status, and

PAGE 118

118 employment status. Age was measured as a continuous variable. Race was categorized as Black/African American, White, or Other. Education level was ca tegorized as 12 or fewer years of education and greater than 12 years of education. Food insecurity was measured via the Health enough money to buy food that you or yo Health insurance status was Mental heal th conditions were categorized as self reported lifetime history of depression [Y/N] a lifetime history of anxiety [Y/N], or a lifetime history of other mental health conditions [Y/N] Other mental health conditions included a positive endorsement of ADD/ADHD, autism, bipolar disorder, eating disorders, mania, personality disorder, or schizophrenia. Coronary disease risk factors include d in these analyses were BMI, history of hypertension, history of type II diabetes and history of hypercholesterolemia. The Health Intake (risk factor) ? [Y/N Those who answer yes are coded as having a lifetime history of hypertension, hyperchol esterolemia and/or type II diabetes respectively. The BMI covariate was created using reported height and weight, and then categoriz ed acco r ding to World Health Organization standards as underweight, normal weight, overweight or obes e (World Health Organization, 2000) Statistica l Analysis Latent class analysis (LCA) (Hagenaars and McCutcheon, 2002) was applied using Mplus Version 6.0 (Muthen and Muthen, 2007) to investigate the patterns of lifetime substance use among HealthStreet members. LCA is a person centered approach (Muthen and Muthen,

PAGE 119

119 2000) that allows for the identification of heterogeneous subgroups (latent classes) based o n the observed patterns of lifetime substance use. The results of LCA provide the probability that an individual will fall into a given latent class. LCA offers advantages over other person centered methods such as cluster analysis because it is less restr ictive (does not require equal indicator variances across classes), and it estimates multiple indices of model fit (Nylund et al., 2007; Rosellini et al., 2014) As noted above, 6 m anifest variables were input into the LCA. First, a one class model, or the baseline model, was created where no differences by substance use for the entire sample were assumed. Then, an iterative model building process commenced where each new model inclu ded one more class than the model prior to it. The optimal number of classes was determined through the use of the Bayesian Information Criterion (BIC), Vuong Lo Menedel Rubin Likelihood Ratio Test (VLMR), and Entropy (Henson et al., 2007; Lo et al., 2001; Nylund et al., 2007; Ramaswamy et al., 1993; Sclove, 1987; Vuong, 1989) BIC is a Bayesian based measure of model fit, with a better fit indicated by a smaller value than the baseline model (Nylund et al., 2007; Sclove, 1987) The VLMR is also a measure of model fit that looks specifically at whethe r a k class solution is superior to a k 1 class solution, and where the best fitting models have a VLMR p value of <0.05 (Henson et al., 2007; Lo et al., 2001; Vuong, 1989) Entropy is a measure of how well the model is classifying each individual, with higher values indicating better classification (Henson et al., 2007; Ramaswamy et al., 1993) While, each of these stati stical measures will be used to help identify the best number of latent classes, but they cannot replace the use of logic and theory. A combined statistical and theoretical approach determined the final number of latent classes assigned.

PAGE 120

120 After the optimal latent class model was built, descriptive statistics and multinomial logistic regression analysis (Engel, 1988) were conducted to understand how coronary disease w as associated with the latent classes using SAS software, version 9.4 of the SAS System for Windows. Copyright 2002 2 012 SAS Institute Inc. Three multinomial logistic regression models were generated. The first regression model controlled for sex race, age, insurance status, food insecurity, and education level; the second regression model controlled for the same factor s, but added lifetime history of depression, lifetime history of anxiety, and lifetime history of other mental health conditions ; the third regression model controlled for the same factors but added lifetime history of hypertension, hypercholesterolemia, t ype II diabetes, and BMI After fitting 2 a pseudo R 2 was used to estimate model fit (Allison, 2014; McFadden, 1974) It is important to remember that a pseudo R 2 does not have the same interpretation as a standard R 2 statistic, so it is not possible to say how much of the variance in the model is explained by the variables included. However it is true that the higher the value, the better the fit (McFadden, 1974) Previous literature indicates an association between cannabis use and psyc hotic illness including schizophrenia (Marconi et al., 2016; Murray and Forti, 2016) As such, analyses were also conducted without the inclusion of lifetime hallucinogen use as a manifest variable, and without controlling for other mental health disorders. Removing these variables from these analyses did not change the significance of results, thus the results presented below include them both. Results Table 5 1 shows that about half of the 8,530 HealthStreet members included in this analysis reported a lifetime history of cannabis use (48.8%) and 53.6% reported tobacco use There were 1, 559 members who reported lifetime cocaine or crack use (18.3%). Similar numbers

PAGE 121

121 of HealthStreet members reported lifetime club drug use, lifetime amphetamine use, and lifetime hallucinogen use (6.5 %, 7.5%, and 6.0 % respectively). Class Membership A 3 class model was selected, first, because it had a lower BIC ( 38069.114 ) compared to the baseline (1 class) model ( 44716.824 ). The 3 class model also had the highest relative e ntropy at 0.706 compared to the baseline, 2 class, and 4 class models. Additionally, the 3 class model had a statistic ally significant VLMR (p value < 0.0001), indicating the 3 class model was superior to the 2 class model. In conjunction with the statistical tests indicating the superiority of th e 3 class model compared to the 2 or 4 class models, the 3 class model was also chosen for ease of interpretability. Table 5 2 gives the results of the LCA, with a 3 class model. Class 1 accounted for 58.3 % of the sample (n=4 973 ) and was distinguished from other classes by low probabilities of lifetime use of any substance. Additionally, there were no individuals in the class with a lifetime history of use of more than one substance. Class 1 is the comparison group, and is referred to as the mono substance users and abstainers Class 2 accounted for 33.7% of the sample (n=2 873 ) and was distinguished from the other classes by having high lifetime use of cannabis and tobacco but lower use of all other substances They will be referred to as the cannabis and to bacco users Class 3 accounted for 8.0 % of the sample population (n=684 ), and was distinguished from other classes as the poly substance use group. Figure 5 1 gives the percentage of HealthStreet members with lifetime use of each type of substance by latent class. Description of Total Sample The first column of T able 5 3 gives the socio demographic characteristics lifetime history of mental health conditions coronary disease risk factors and lifetime history of coronary disease for the total sample. The a verage age of the total sample was 43.7 years. The

PAGE 122

122 majority of the sample was female ( 58.6 %), and Black ( 61.2 %). Most HealthStreet members had greater than 12 years of education (59.9 %) O nly about one third were employe d either full or part time (34.4 %), and about half reported being food insecure ( 47.1 %). Not having health insurance was reported by 40.5 % of the sample Similar proportions of the sample reported a lifetime history of depression ( 27.5 %), anxiety ( 22.7 %) or oth er mental health conditions (20 .6 %). About one third of the sample endorsed a lifetime history of hyper tension (35.0%), while only 18.8 % of the sample endorsed a lifetime history of hypercholesterolemi a. A little over one third (37.4 %) of the entire sample was obese (BMI 30+) and 11.2% of the sample reported a lifetime history of type II diabetes Overall, 12.7% of the sample reported a lifetime history of coronary disease. Description of Class Membership The final 3 c olumns of T able 5 3 give the socio demographic characteristics life time history of mental health conditions coronary disease risk factors and coronary disease by latent class. There was a significant difference in average age across latent classes, with polysubstance users being the oldest on average (46.7 years) While t he majo rity of polysubstance users were male (58.6 %), the majority of mono substance users and abstainers were f emale (65 .7%). About half of cannabis and tobacco users w ere female ( 50.6 %). Most polysubstance users were White ( 69.4 %), however most mono substance users and abstainers, and cannabis and tobacco users were Black ( 66.0% and 62.6 % respectively). There were significant differences across latent classes in education level, employment status, and food insecurity with polysubstance users havi ng the greatest percentage of low education level, unemployment, and food insecurity compared to the other two groups. Unsurprisingly, polysubstance users had the greatest number of uninsured members ( 51.6 %) followed by the cannabis and tobacco users (48.1 %) and then the comparison group (mono substance users and abstainers) ( 34.6 %). For all mental health

PAGE 123

123 conditions (depression, anxiety, other mental health conditions ), polysubstance users reported the highest frequency (52.3%, 48.7%, 45.0 % respectively), followed by cannabis and tobacco users with th e second highest frequency (34.2%, 27.3%, and 27.9 % respectively), and mono substance users and abstainers having the lowest frequency ( 20.2 %, 16.6%, and 13.1 % respectively). There were no significant differenc es in hypertension and type II di abetes by latent class. Polysubstance users had the highest frequency of those with lifetime history of hypercholesterolemia (22.5%). The highest frequency of obesity was seen among mono substance users and abstainers (40.5 %). Differences in lifetime history of coronary disease across latent classes were statistically significant There was a linear trend, with the lowest percentage of HealthStreet members endorsing a lifetime history of coronary disease being part of the ab staining group (10.7%), and the highest percentage of HealthStreet members endorsing a lifetime history of coronary disease being part of the polysubstance group (17.9%). Multinomial Logistic Regression Table 5 4 gives the odds ratios and 95% confidence intervals for the association between lifetime history of coronary disease and latent class membership while controlling for socio demographic factors (Model 1), socio demographic factors plus mental health condit ions (Model 2) and socio demographic factors, mental health conditions and coronary disease risk factors (Model 3) Model 1 shows significantly increased odds of membership in the cannabis and tobacco use class versus the comparison class (mono substance users and abstainers) for those reporting a lifetime history of coronary disease (OR 1.37 95% CI 1.18 1.58) There were also significantly increased odds of membership in the polysubstance users class versus the comparison class (mono substance users a nd abstainers) for those reporting a lifetime history of coronary disease (OR 1.66 95% CI 1.30, 2.12 ). However, after controlling for depression, anxiety and other

PAGE 124

124 mental health conditions (Model 2) as well as coronary disease risk factors (Model 3) th e s ignificant relationship in both the cannabis and tobacco user class and the polysubstance user class versus the comparison class (mono substance users and abstainers) are attenuated Discussion To our knowledge this is the first analysis looking at the a ssociation of l atent classes of lifetime substance use and lifetime history of coronary disease The overall prevalence rate of lifetime can nabis use in our sample was 48.8%, which is higher than national estimates for the lifetime prevalence rate of cannabis use (44 %) found by the NSDUH (Center for Behavioral Health Statistics and Quality, 2015) This may be because the median age in our sample is higher than the median age of the NSDUH sample (45 years vs 37.3 years ) giving HealthStreet members a longer timespan to have use cannabis It may also be linked to the socio demographic breakdown of our sample. For example, our sample is primarily unemployed, and unemployment has previously b een associated with substance use (Compton et al., 2014) HealthStreet members reported a lifetime prevalence of tobacco use at 53.4%, slightly under the national lifetime prevalence rate for tobacco use (66.2%) (Center for Behavioral Health Statistics and Quality, 2015) The lifetime prevalence of club drugs and hallucinogens in our sample were slightly lower than national prevalence rate estimates, while l ifetime cocaine use and lifetime amphetamine use in our sample was slightly higher than national estimates (Center for Behavioral Health Statistics and Quality, 2015) Our analysis identified 3 classes of lifetime substance user s First, the mon substance users and abstainers who had relatively low probabilities of any substance use. Second, the cannabis and tobacco use group, who had low probabilities for use of club drugs, amphetamines, hallucinogens, and cocaine or crack, but who had high probabilities for lifetime use of cannabis and tobacco, and finally the polysubstance use group, who had high probabilities of using

PAGE 125

125 numerous illicit and non illicit substances. Our results are logically congruent with literature suggesting concurrent initiation and use of cannabis and tobacco, with a subset then initiating use of other illicit drugs (Agrawal et al., 2006; Haberstick et al., 2014; Redonnet et al., 2012) Our overall sample was primarily Bla ck, and primarily female. Interestingly, the polysubstance use group was overwhelmingly male and White, while the cannabis and tobacco user group was about half male and primarily Black. Other research indicates a higher lifetime prevalence rate of any ill icit drug use among both males and Whites, including a higher prevalence rate of lifetime cannabis use among both males and Whites (Center for Behavioral Health Statistics and Quality, 2015) Ou r sample over represents Blacks and females, possibly leading to the differences seen in the demographic makeup of the cannabis and tobacco use group. The lifetime prevalence of depression in our sample was 27.5 %, and the lifetime prevalence of anxiety i n our sample was 22.7 %. Prevalence of depression and prevalence of anxiety varied significantly by latent class, where polysubstance users had the highest lifetime prevalence of both depression and anxiety, cannabis and tobacco users had the second highest prevalence of depression and anxiety, and mono substance users and abstainers reported a low lifetime prevalence of depression and anxiety. These results are in line with pre vious analyses indicating an association between depression and anxiety and any illicit substance use, with increased odds of depression and anxiety among polysubstance users (Degenhardt and Hall, 201 2) Additionally, Lynskey and colleagues (2006) found increased odds of depression and anxiety in their polysubstance latent class compared to their cannabis only latent class. Though research provides no one clear mechanism for the comorbidity, some re search indicates a strong association between substance use and mental health conditions due to self medication, wh ile

PAGE 126

126 other research indicates mental health conditions may follow substance use (Compton et al., 2000; Swendsen and Merikangas, 2000) P revalence of coronary diseas e in our sample was 12.7%, which is slightly higher than the national prevalence of adults diagnosed with heart disease (1 1.5% ) (National Center for Health Statistics, 2014) Again, this may be due to a higher average age in our sample (43.7 years) than the national average. In univariate analysis the association between lifetime history of coronary disease risk factors and latent class membership varied significantly by class membership, with mono substance users and abstainers having the lowest frequency of coronary disease ( 10.7 %), cannabis and tobacco users having the second highest frequ ency ( 15 .0%), and polysubstance users having the highest frequency of coronary disease ( 17.9 %). This relationship held in the first multinomial logistic regression model but was attenuated in the second and third model s (controlling for mental health cond itions and coronary disease risk factors respectively) Increased odds of coronary disease among the cannabis and tobacco user class a nd the polysubstance user class versus the comparison class (mono substance users and abstainers) in Model 1 is unsurpri sing based on previous literature showing strong associations between tobacco use and coronary disease (Chiva Blanch et al., 2013; De Giorgi et al., 2012; McBride, 1992; Thylstrup et al., 2015) Additionally, evidence exists for a relationship betwee n cannabis use and coronary disease ( Aryana & Willi ams, 2007; Gordon et al., 2013; Sidney, 2002 ) cocaine use and coronary disease, and amphetamine use and coronary disease (Ghuran and Nolan, 2000) Attenuation in Model 2 indicates the importance of considering confounding by mental health conditions when assessing the relationship between coronary disease and substance use latent class membership The lack of statistical significance after adjustment for mental health conditions could mean no independent association between the latent classes and coronary

PAGE 127

127 disease. However, that explanation seems unlikely given the strong evidence for the association between tobacco use, cocaine use, and coronary disease. It could be that instead of acting as a true confounding factor through direct association with both coronary disease and latent classes of substance user, mental health conditions increase the risk of substance use, which in turn increases the risk of coronary disease. However, this analysis was cross sectional, meaning we are unable to clarify the direction of the relationship between substance use, mental health conditions and coronary disease. La stly, it could be that mental health conditions such as depression manifest after the diagnosis of coronary disease. Depression has previously been linked to coronary disease (Ariyo et al., 2000; Baune et al., 2012; Carney and Freedland, 2016; Elderon and Whooley, 2013; Frasure Smith N et al., 1993; Gonzlez and Tarraf, 2013; Hare et al., 2013; Joynt et al., 2003) In fac t, d epression independently increase s the risk for coronary disease 1.5 times (Baune et al., 2012) and is a risk fac tor for poor prognosis in current cardiac patient populations (Lichtman et al., 2014) Additionally, anxiety and borderline personality disorders have been associated with coronary disease (Powers and Oltmanns, 2013; Thurston et al., 2013) As noted above, they have also been associated with latent classes of substance use (Lynskey et al., 2006) There are two important limi tations to this analysis. First, the HealthStreet sample is not nationally representative. HealthStreet is a community engagement model that aims to increase the enrollment of under represented groups into health research. As a result, our sample ove r repr esents females, Blacks, and cannabis users. However, the oversampling of these groups provides the unique opportunity to conduct analyses using a sample with commonly underrepresented groups T his sample of over 8 8 00 HealthStreet members provides a baseli ne

PAGE 128

128 understanding of patterns of substance use, and of coronary disease risk factors associated with those patterns. The second important limitation is that all measures used were self reported. Self report may be subject to increased bias. However, the re is agreement between self report of coronary disease and medical record diagnoses in the literature including substantial agreement for M I among both genders, and substantial agreement for angina among women (agreement among men was not examined in that a nalysis ) (Okura et al., 2004 ; Simpson et al., 2004) The literature also suggests self report is a valid measure of tobacco, alcohol and illicit drug use and (Del Boca and Darkes, 2003; Ledgerwood et al., 2008; Yeager and Krosnick, 2010) T his analysis represents one of the first to assess the association between lifetim e history of coronary disease and latent classes of substance use. In particular, we were able to examine the odds of being in a class of primarily cannabis and tobacco users by coronary disease status This analysis importantly considers how cannabis is a ssociated with coronary disease in the context of overall patterns of substance use. Increases in cannabis use nationwide have made identifying health risks associated with cannabis use of great public health importance Additionally the comorbid use of c annabis and other substances makes it important to consider overall patterns of use. Future research should focus on further understanding how cannabis use is related to coronary disease, and on how to use this information to develop personalized intervent ions.

PAGE 129

129 Table 5 1. Percentage of HealthStreet members positively endorsing lifetime substance use Total Sample n=8533 Lifetime Cannabis Use 4160 (48.8 %) Lifetime Tobacco Use 4572 (53.6 %) Lifetime Club Drug Use 554 (6.5%) Lifetime Cocaine Use 1559 (18.3%) Lifetime Amphetamine Use 637 (7.5 %) Lifetime Hallucinogen Use 598 (7.0 %) Table 5 2. Item response probabilities Class 1 Mono Substance Users & Abstainers (n=4973 ) Class 2 Cannabis & Tobacco Users (n=2873 ) Class 3 Polysubstance Users (n=684 ) Cannabis 0.179 0.835 0.981 Tobacco 0.26 3 0.85 3 0.921 Club Drug Use 0.001 0.089 0.37 2 Cocaine or Crack 0.01 5 0.2 78 0.85 1 Amphetamines 0.006 0.040 0.667 Hallucinogens 0.00 0 0.03 6 0.661

PAGE 130

130 Figure 5 1. Latent class membership by percentage of members with lifetime use of each type of substance 0 10 20 30 40 50 60 70 80 90 100 Percentage Substance Abstainers Cannabis and Tobacco Users Polysubstance Users Low Users

PAGE 131

131 Table 5 3. Socio demographics, lifetime history of mental health conditions and coronary disease risk factors for the en tire sample and by latent class Total Sample Class 1 Mono Substance Users & Abstainers Class 2 Cannabis & Tobacco Users Class 3 Polysubstance Users n=8533 n=4973 n=2873 n= 684 Mean Age (SD) A 43.7 (16.2) 43.5 (17.5 ) 43.2 (14.5) 46.7 (13.4 ) Sex A Female 4 999 (58 .6%) 3265 ( 65.7 %) 1 451 ( 50.6 %) 282 ( 41.4 %) Race A Black 5222 ( 61.2 %) 3 284 (66.0 %) 1799 ( 62.6 %) 138 (20.2 %) White 2 724 ( 31.9 %) 1352 ( 27.2 %) 896 ( 31.2 %) 475 ( 69.4 %) Other 587 (6.9 %) 337 (6.8 %) 178 (6.2 %) 71 (10.4 %) Education Level A 3418 (40.2 %) 2158 (43.5 %) 920 (32.1 %) 339 (49.6 %) > 12 Years 5096 (59.9 %) 2800 (56.5 %) 1950 (68.0 %) 344 (50.4 %) Employed A 2919 (34.4 %) 1854 (37.6 %) 852 (29.8 %) 211 (31.0 %) Food Insecure A 4000 (47.1 %) 1996 (40.4 %) 1594 (55.7 %) 409 (60.0 %) Insurance A None 3376 (40.5 %) 1673 (34.6 %) 1354 (48.1 %) 348 (51.6 %) Public 2232 (26.8 %) 1531 (31.6 %) 533 (18.9 %) 167 (24.8 %) Private 2727 (32.7 %) 1635 (33.8 %) 931 (33.0 %) 160 (23.7 %) Depression A 2339 (27.5 %) 999 (20.2 %) 979 (34.2 %) 361 (53.2 %) Anxiety A 1935 (22.7 %) 823 ( 16.6 %) 781 (27.3 %) 331 (48.7 %) Other MH Disorders A 1761 (20.6 %) 6 51 (13.1 %) 802 ( 27.9 %) 308 (45.0 %) Hypercholesterolemia A 1588 ( 18.8 %) 905 (18.3%) 533 (18.8%) 150 (22.5%) Hypertension 2969 ( 35.0 %) 1734 (35.0%) 1012 (35.4%) 223 (33.0%) Type II Diabetes 946 (11.2%) 626 (12.7%) 257 (9.0%) 63 (9.3%) Body Mass Index A Underweight 362 ( 4.2 %) 241 (4.9%) 99 (3.5%) 22 (3.2%) Normal Weight 2533 ( 29.7 %) 1327 (26.7%) 953 (33.2%) 253 (37.0%) Overweight 2450 ( 28.7 %) 1389 (27.9%) 842 (29.3%) 219 (32.0%) Obese 3185 ( 37.4 %) 2016 (40.5%) 979 (34.1%) 190 (27.8%) Coronary Disease A 1084 (12.7%) 533 ( 10 .7%) 429 ( 15.0 %) 122 ( 17.9 %) A Statistically significant difference across classes at p<0.05

PAGE 132

132 Ta ble 5 4. Odds ratios and 95% confidence intervals for the association between lifetime history of coronary disease and latent class membership. Model 1 A Model 2 B Model 3 C Lifetime History of Coronary Disease Cannabis & Tobacco vs Mono Substance & Abstainers Polysubstance vs Mono Substance & Abstainers Cannabis & Tobacco vs Mono Substance & Abstainers Polysubstance vs Mono Substance & Abstainers Cannabis & Tobacco vs Mono Substance & Abstainers Polysubstance vs Mono Substance & Abstainers Yes 1.37 (1.18, 1.58)* 1.66 (1.30, 2.12)* 1.12 (0.86, 1.44) 1.12 (0.96, 1.30) 1.15 (0.98, 1.34) 1.20 (0.92, 1.57) No Ref Ref Ref Ref Ref Ref A Model adjusted for sex race, age, insurance status, food insecurity and education l square= 0.1351 B Model adjusted for sex race, age, insurance status, food insecurity, education level, lifetime history of depression, lifetime history of anxiety, and lifetime history of other mental health conditions square= 0.1790 C Model adjusted for sex race, age, insurance status, food insecurity, education level, lifetime history of depression, l ifetime history of anxiety, lifetime history of other mental health conditions hypercholesterolemia, hypertension, BMI and type II diabetes square= 0.1850 Stat istically significant odds ratio

PAGE 133

133 CHAPTER 6 CONCLUSIONS This dissertation represents one of the first attempts to characterize the relationship between cannabis use and coronary disease First, a systematic literature review w as undertaken to determin e the state of the extant literature on the association between cannabis use and coronary disease as well as coronary disease risk factors. Then, data from HealthStreet was used to analyze the hypothesized direct association between cannabis use and corona ry disease via multiple logistic regression. Next, latent class es of HealthStreet members were identified based on lifetime substance use, and multinomial logistic regression was used to determine the association between the latent classes and lifetime his tory of coronary disease. Main Findings The results of the literature review (presented pictographically in Figure 2 2), were equivocal. The two included studies that addressed direct associations between cannabis use and coronary disease found significan t, positive associations. However, the results of included studies on cannabis use and coronary disease risk factors were mixed with most finding non significant associations some finding significant and positive associations and others finding significant and negative associations The current body of evidence for associations between cannabis use and coronary disease or coronary disease risk factors is severely limited as evidenced by the numerous contradictory findings. First, o ut of the 16 included studies, only 11 unique datasets were represented. In particular, data from NHANES was used by 5 studies (Penner et al., 2013; Rajavashisth et al., 2012; Thompson and Hay, 2015a; Vidot et al., 2016; Yankey et al., 2016) When data from these studies are counted as significant or non significant five separate times it may lead to bias; allowing it to appear that there is more evidence for an association than there truly is.

PAGE 134

134 Unfortunately there are not many large, publicly available datasets with sufficient information to determine associations between cannabis use and coronary disease. The second key limitation was an inability to determine causal associations; adequate longitudinal data is needed. The third key limitation to the extant body of literature was bias in the exposure measure. Studies employed a range of cannabis use measures including binary measures (never use, lifetime use), various categorical measures, and continuous meas ures. Measures of cannabis use including quantity of use, frequency of use, and duration of use are the gold standard, however categorical measures of lifetime cannabis use (e.g. never, former, current) are often the only available measures in datasets tha t provide sufficient sample size and coronary disease or coronary disease risk factor measures. The fourth key limitation of the extant literature was the lack of evidence for associations between cannabis use and coronary disease as compared to coronary d isease risk factors. Of the 16 included studies, only two, Aronow and colleagues (1974) and Mittleman and colleagues (2001), addressed coronary disease as an outcome. The final key limitation of the extant literature was the inability to account for unmeas ured factors The inconsistent findings mean there are likely unmeasured factors such as home and neighborhood environment that contribute to the association between cannabis use and coronary disease or coronary disease risk factors. Such factors are seldo m measured in large, self report datasets. To help fill the gap in the literature, two sets of analyses (C hapters 4 and 5) were done with data from HealthStreet. The first (C hapter 4) analysis showed that the association between cannabis use and coronary disease was statistically significant in univariate analyses and after controlling for socio demographic characteristics (Model 1) However, the significant association was attenuated after controlling for mental health conditions and substance use (Model 2) and coronary disease risk factors (Model 3) Theses analyses contribut e to the field by

PAGE 135

135 adding to the few studies that have examined the association between cannabis use and coronary disease as opposed to coronary disease risk factors. Additionally, th e HealthStreet data used for those analyses allowed for adequate controlling of confounding factors. There were a few possible re asons for the results found in C hapter 4. First, it may be that any relationship between cannabis use and coronary disease is the result of indirect effects, where other factors such as obesity, hypertension, or dyslipidemia mediate the association between cannabis use and coronary disease. If it is true that factors like obesity mediate the relationship between cannabis use and coronary disease, then controlling for important mediating factors would render the model non significant. Second it may be that cannabis use is not independently associated with coronary disease and that the association seen in univariate analyses and Model 1 was a result of confounding due to factors known to be associated with both cannabis use and coronary disea se. It may be that any relationship between cannabis use and coronary disease is confounded by mental health conditions In the systematic review it was noted that the extant literature has failed to adequately control for mental health conditions but ou r results combined with evidence for associations between mental health conditions and both cannabis use and coronary disease suggest they are important confounders. Additionally, a ny direct association between cannabis and coronary disease may be eclipsed by the stronger relationships between cannabi s and co occurring substance use. This is in line with findings from Rodondi and colleagues (2006), who noted that associations between cannabis use and BMI hypertension and dyslipidemia were attenuated in adjusted analyses, primarily confounded by other substance use such as alcohol and tobacco. The possibility of confounding by other substances is bolstered by previous literature

PAGE 136

136 showing that cannabis is s eldom used alone, but is often used concurrently with alcohol and tobacco, both of which are known risk factors for coronary disease. To that end, a latent class analys i s account ed for patterns of substance use and their association with coronary disease. The latent class analysis done in C hapter 5 revealed three distinct classe s of lifetime substance user: 1) mono substance users and abstainers 2) cannab is and tobacco users and 3) polysubstance users. These results are consistent with previous literature showing that cannabis and tobacco use is often concurrent and that a subset of individuals who use cannabis and tobacco use other illicit substances as well (Agrawal et al., 2006; Haberstick et al., 2014; Redonnet et al., 2012) The mu ltinomial regression models in Chapter 5 mirrored those in C hapter 4, with the association between latent classes of substance user and coronary disease being signif icant in unadjusted analyses and analyses adjusted for socio demographic characteristics However, after the addition of mental health conditions (Model 2) and coronary disease risk factors (Model 3) the significant association was attenuated. This sugges ts at least two things. First, that, as in previous analyses, mental health conditions such as depression and anxiety may play a key role in coronary disease among those who use marijuana. Second, that while tobacco, alcohol, cocaine and other substance us e are associated with coronary disease in the literature (Chiva Blanch et al., 2013; De Giorgi et al., 2012; McBride, 1992; Thylstrup et al., 2015) they may not be the primary confounding factors Instead, as mentioned in the C hapter 4 discussion, a ny relationship between cannabis use and coronary disease may be primarily confounded by mental health conditions The association may also be the result of indirect effects or mediation with other known risk factors This is line with findings from the sy stematic literature review conducted in C hapter 2 suggesting possible

PAGE 137

137 associations between cannabis use and coronary disease risk factors such as obesity, hypertension, and dyslipidemia. Other substances used concomitantly may be mediating the association between cannabis use and coronary disease. In particular tobacco use is associated with hypertension, obesity, dyslipidemia and metabolic syndrome (Balhara, 2012) Additionally, a lcohol use and cocaine use are both associated with hypertension (Beilin and Puddey, 2006; Lange and Hillis, 2001) Mental Health Conditions A theme that emerged from each analysis was the importance of considering mental health conditions when looking at associations between cannabis use and coronary disease. Depression, anxiety and other mental heal th conditions varied significantly by coronary disease status a nd by latent class membership (T ables 4 2 and 5 3). In the latent class analysis, depression and anxiety varied significantly by latent class membership with polysubstance users reporting a lif etime history of a given mental health disorder with greater frequency than the other groups. These results fit with previous literature showing an association between depression and anxiety and any illicit substance use, with increased odds of depression and anxiety among polysubstance users (Degenhardt and Hall, 2012) Additionally, Lynskey and colleagues (2006) found increased odds of depression and anxiety in their polysubstance latent class com pared to their cannabis only latent class. Depression is also known to be associated with polysubstance use (Brook, Brook, Zhang, Cohen, & Whiteman, 2002; Hasin, Goodwin, Stinson, & Grant, 2005; Kessler, Berglund, Demler, et al, 2003; Leatherdale & Ahmed, 2010) These findings are bolstered by the large body of prior research showing signifi cant associations between depression and coronary disease (Ariyo et al., 2000; Baune et al., 2012, 2012; Carney and Freedland, 2016; Elderon and Whooley, 2013; Frasure Smith N et al., 1993; Gonzlez and Tarraf, 2013; Hare et al., 2013; Joynt et al., 2003; Powers and Oltmanns, 2013;

PAGE 138

138 Thurston et al., 2013) Anxiety is also associated with coronary disease (Buckner et al., 2012; Patton et al., 2002) O ther me ntal health conditions including personality disorders (Copeland, Rooke, & Swift, 2013) have been shown to be associated with coronary disease and substance use as well (Copeland, Rooke, & Swift, 2013) Thus our finding of the important role of mental health problems in coronary disease is not surprising, although underappreciated in ecological models. Strengths This dissertation represents one of the first pieces of literature to assess the association between cannabis use and coronary disease. The systematic literature review was one of the first to look at the association between cannabis use and coronary disease as well as coronary disease risk factors using only epidemiologic literature (cross sectional, case control, cohort, and randomized trial studies only). The systematic literature review conducted in C hapter 2 identified key areas of w eakness in the body of literature addressing possible associations between cannabis use and coronary disease. The January 2017 report by the National Academies of Sciences, Engineering, and Medicine on the health effects of cannabis also reviewed literatur e on the association between cannabis use and MI as well as diabet es. In line with findings from C hapter 2, they note that the current body of literature is inadequate to make any assessment about the risk of coronary disease associated with cannabis use (National Academies of Sciences, Engineering, and Medicine, 2017) A nalyses in C hapters 4 and 5 serve to address those areas of weakness by specifically looking at coronary disease as an outcome, by controlling for confounding with relevant covariates, and by utilizing a dataset with both a large sample size and that had not previously been used to a ddress these research questions.

PAGE 139

139 Limitations Even though this dissertation represents one of the first times the association between cannab is use and coronary disease was explored, there are some important limitations relevant to all analyses. First, the HealthStreet sample over represents females, Blacks, and cannabis users (Table 4 1) As noted, HealthStreet is a community engagement model that aims to increase the enrollment of under represented groups into h ealth research Consequently, traditionally u nderrepresented groups may be over represent ed here. However, HealthStreet presented a unique opportunity to explore the cannabis/coronary disease relationship across these groups of traditionally underrepresented individuals The HealthStreet sample is ro bust in size and provided data with which to create a foundational understanding of the association between cannabis use and coronary disease and patterns of substance use and associated coronary disease risk factors Second, cannabis use, coronary diseas e, and all covariates were measured through self report. While self report may be subject to increased bias, the re is agreement between self report of coronary disease and medical records in the literature including substantial agreement for MI among both genders, and substantial agreement for angina among women (this analysis did not consider men) (Okura et al., 2004; Simpson et al., 2004) Natarajan and colleagues (2002) f ound that self reported hypercholesterolemia had high specificity (89%) but fairly l ow sensitivity (51%). Additionally, self report of BMI is often slightly under reported (Gorber et al., 2007) However, literature has shown agreement between self report of hypertension and hypertension in medical records (Okura et al., 2004) S elf report of substance use like tobacco use and illicit drug use has been shown to be valid (Del Boca and Darkes, 2003; Ledgerwood et al., 2008; Shillington et al., 1995; Yeager and Krosnick, 2010) The literature also indicates self report of tobacco, alcohol cannabis and illicit drug use to be valid (Del Boca and Darkes, 2003;

PAGE 140

140 Ledgerwood et al., 2008; Mennes et al., 2009; Yeager and Krosnick, 2010) Additionally, we were unable to capture measures of quantity, frequency, duration of cannabis use or potency of the cannabis used Public Health Consequence & Future Research To date, the extant literature on the association between cannabis use and coron ary disease as well as coronary disease risk factors is inadequate to make any determinations about specific associations. Overall, we are unable to say with certainty whether or not cannabis use is a harmful or protective factor for coronary disease. Act ivation of the CB1 receptor is the hypothesized mechanism of action for cannabis on coronary disease, but the equivocal results of the systematic review suggest that cannabis use may provide a protective effect, possibly through activation of the CB2 recep tor. Despite this gap in the literature as a whole, support for legalization of cannabis use in the US is great (60% of the US population according to the Gallop Pole ; Swift, 2016) an d cannabis use across the US continues to rise. This, along with the high morbidity and mortality associated with coronary disease gave ample reason to explore the association using epidemiologic techniques. The results of these analyses suggest that cann abis use may not be directly associated with coronary disease outcomes. Instead, findings suggest that the relationship is the result of either confounding or mediation. In combination, results of the systematic literature review, logistic regression and l atent class analyses provide findings that lead to a number of testable hypotheses useful to tease apart the role of cannabis in coronary disease. F uture research should examine the direction of the possible pathways linking cannabis use to coronary disea se. Figure 6 1 can be used to direct this type of research. There is strong evidence showing an association between demographic factors like age and sex, and cannabis use as evidence d by the solid, red arrow However the directionality of the association between

PAGE 141

141 cannabis and other behavioral risk factors is not well established. Additionally, more research is needed to clarify directional associations between cannabis use and physiological risk factors for coronary disease as well as coronary disease itsel f. To determine causation, however, temporal ity or direction of association is not the only thing that should be considered. The criteria for causation put forth by Sir Austin Bradford Hill offer a way to contextual ize the direction of future research (Table 6 1; Fedak et al., 2015; Hill, 1965) Prior research indicates a biologically plausible association; however the remaining causal criteria need to be addressed. The first causal criterion to be addressed is strength of association. While a strong association between cannabis use and coronary disease may not exist the high morbidity and mortality associated with coronary disease means even small causal association s can affect the attributable risk. As such, it is imperative to have studies that are sufficiently powered to detect small effect sizes. Furthermore, studies should be sufficiently powered to examine associations within sub groups. For example, researcher s should examine whether or not cannabis use is associated with coronary disease among those age 50 years and older compared to those younger than 50 years, thus assessing specificity. To examine consistency, multiple well designed studies that utilize sta ndardized measures of cannabis use and coronary disease will be needed Having standard measures of cannabis use that address quantity, freq uency and duration of use will allow researchers to assess the possible biological gradient. Understanding biologica l gradient may also mean understanding how cumulative lifetime cannabis use changes the risk of coronary disease as an individual ages. As more and more states legalize cannabis for recreational and medicinal use, our current understanding of its health ef fects will change. It is important, then, that future research is carefully designed to be coherent, or to fit within the current model of the health effects of cannabis. While randomized

PAGE 142

142 controlled trials with cannabis use as the exposure are not always e thical the experiment criterion may be addressed with modern epidemiologic methods such as agent based modeling. Agent based modeling allows for the simulation of counterfactual outcomes in the context of complex etiologies such as those associated with c oronary disease (Marshall and Galea, 2015) Lastly, to assess the analogy criterion, researchers should look to see if synthetic cannabino ids show similar associations with coronary disease as cannabis. The changing legal climate surrounding cannabis use combined with increased use presents the perfect opportunity to make highly consequential contributions to the field of public health. Thi s dissertation presents the beginning of that work, and offers a way forward.

PAGE 143

143 Figure 6 1. Directions for future research in the context of the etiology of coronary disease (adapted from Pearson et al., 1993) Cannabis Use Sedentary lifestyle Diet Saturated Fat Salt Cholesterol Total energy content Binge drinking Any tobacco use Any cocaine use Behavioral Risk Factors Age Sex Family History Non modifiable Risk Factors Obesity (Elevated BMI) Hypertension (Elevated SBP, DBP) Dyslipidemia (Adverse LDL C, HDL C, Total Cholesterol, Triglycerides) Diabetes (Elevated Plasma Glucose, HbA1c ) Depression/Anxiety Physiological Risk Factors Coronary Artery Disease Angina Myocardial Infarction Coronary Disease Outcomes Metabolic Syndrome

PAGE 144

144 Tab le 6 1. Summary of state of the research or future research needed to follow Bradford Hill criteria. Criterion State of the Research or Future Research Direction to Fulfill Criterion Strength Adequately powered study to detect small associations Consistency More than one rigorous ly designed study on this topic should be funded Specificity Population, dosage and delivery specific studies Temporality Longitudinal study design; Clarify direction of association between cannabis, mental health conditions and coronary disease Biological Gradient I ncreasing frequency, quantity, or dose of cannabis and their associat ion with coronary disease Plausibility Completed r esearch indicates biologically plausible mechanisms Coherence Systematic reviews help coalesce findings into a body of research Experiment Not possible with cannabis use but modern techniques like ABM exist and could be used Analogy Tests of association between synthetic cannabinoids and coronary disease

PAGE 145

145 APPENDIX A SYSTEMATIC REVIEW PROTOCOL Title The association between cannabis use and coronary disease risk factors: A systematic review Date of Protocol October 2016 (Updated December 2016) Introduction Rationale While there have been numerous case reports published citing acute cor onary events after cannabis use it is un clear how many epidemiologic studies have been conducted looking at the association between cannabis use and coronary disease or coronary disease risk factors. A systematic approach is needed to define the gap in the literature. Objectives 1. To summarize t he current epidemiologic literature on the association between cannabis use and coronary disease or coronary disease risk factors. 2. To make recommendations for the future of research in this field. Methods Eligibility Criteria Types of s tudies Prospective cohort Retrospective cohort Case control Randomized Controlled Trial Cross sectional Population s tudied and language of publication Inclusion Criteria:

PAGE 146

146 Any age group or gender break down was acceptable for inclusion in the review. Exclusion Criteria: Non English Language Case Reports Case Series Conference Abstracts Policy Papers Editorials/Commentaries/Letters Studies using animal models Exposure and o utcome measurement Exposure of Interest: Cannabis or marijuana use Outcomes of Interest: Coronary disease o Myocardial Infarction o Coronary Artery Disease o Angina Coronary Disease Risk Factors (Labarthe, 2011) o Hypertension o Type II Diabetes o Elevated BMI o Dyslipidemia o Metabolic Syndrome Exclusion Criteria: Synthetic cannabinoids (e.g. K2, spice, etc) measured as the exposure Exposure to cannabis not compared against a control (e.g. never use) group Subclinical measures of coronary disease such as carotid intima media thickness and coronary calcium scores measured as the outcome No information on the association between the exposure of interest and outcome/s o f interest

PAGE 147

147 Search Methods and Information Sources Electronic s earching Searches of the Pubmed/Medline and Web of Science databases will be carried out from the database inception to October 31, 2016. Pubmed/Medline A systematic search of the Pubmed/Medline and Web of Science databases will be performed. The keyword strategy is outline below: ( marijuana ) OR ( medical marijuana ) AND heart disease ( coronary disease ) OR ( myocardial infarction ) OR ( angina ) OR ( coronary artery disease ) OR ( BMI ) OR ( diabetes ) OR ( hyp ertension) OR (metabolic syndrome) Web of Science (cannabis OR marijuana OR medical marijuana ) AND ( heart disease OR coronary disease OR myocardial infarction OR angina OR coronary artery disease OR BMI OR diabetes OR hyp ertension OR metabolic syndrome) Data Collection and A nalysis Study S election First, titles of the de duplicated records will be screened, with relevant articles then having their abstracts screened for inclusion. Relevant abstracts will then be pulled for full text review. The papers identified for full text review will be collecte d and carefully reviewed for eligibility. Reasons for exclusion will be recorded in the PRISMA flow chart.

PAGE 148

148 Data Extraction and M anagement Data will be extracted from included studies on authors, title, year of publication, study design, study period, sam ple population, sample size, mean sample age, cannabis/marijuana use measure, coronary disease or coronary disease risk factor measure, and measure of association between the exposure and outcome of interest The data items to be extracted are detailed in Table A 1. Assessment of R isk B ias Each study will be individually assessed for risk for risk of bias. The criteria listed in Table A 2 below will be used to assign risk of bias to each study by component. The components considered to assign risk of bias are derived from the Cochrane collaborations tool for assessing inclusion in this review, the tool has been adapted. The author will make a judgment using Table A 3 Data S ynthesis Data extracted from eligible studies will be summarized by tabulation and in short paragraphs.

PAGE 149

149 Table A 1. Data to be extracted from each identified study Data Item Details Lead Author Title Year of Publication Study Design e.g. cross sectional, case control Study Period Sample Population Either a known study (e.g. NHANES) or description of study sample Sample Size Mean Sample Age Exposure Measure e.g. cumulative lifetime cannabis use, frequency of use over the past 12 months % of Sample Exposed Outcome Measure e.g. self reported MI % of Sample with Outcome (or Mean Outcome) e.g. 20% had MI, mean BMI of 28.0 Measure of Association e.g. Odds Ratio, Risk Ratio, Chi Square Analysis Association Present actual number

PAGE 150

150 Table A 2. Domains considered in risk of bias assessments Domain Support for judgment judgment Selection Bias: Generalizability Describe the sampling method used, and the target populatio n for which statements are made Assessing lack of generalizabi lity. Classification of exposure Describe the type of cannabis use measure used, and any validity or reliability information reported about the measure. Measures that account for duration of use, frequency of use, quantity of use, or cumulative lifetime use will b Information bias due to misclassification of the exposure. Information bias: Classification of outcome Describe the type of outcome use measure used, and any validity or reliability information reported about the measure. Measures that use standard diagnostic tests (e.g. non invasive imaging, blood tests such as HbA1c) or validated measures such as ICD9/ICD10 codes accompanied by chart review will be Information bias due to misclassifi cation of the outcome. Confounding of the association between the exposure and the outcome Describe methods taken to control for confounding and any unmeasured confounders Assessing residual confounding. Temporality: Confirming the exposure precedes the outcome Describe the study de sign in relation to temporality Examining lack of temporality.

PAGE 151

151 Table A 3. Assessment of bias forms Entry Judgment Support for Judgment Selection bias generalizability Low risk of bias, high risk of bias, unknown risk of bias Information bias classification of exposure Low risk of bias, high risk of bias, unknown risk of bias Information bias classification of outcome Low risk of bias, high risk of bias, unknown risk of bias Confounding Low risk of bias, high risk of bias, unknown risk of bias Temporality Low risk of bias, high risk of bias, unknown risk of bias

PAGE 152

152 APPENDIX B SYSTEMATIC REVIEW DATA ABSTRACTION TABLES Ross, 2016 Table B 1. Data abstraction for Ross, 2016 Data Item Details Lead Author J. Megan Ross Title Decision Making Does not Moderate the Association between Cannabis Use and Body Mass Index among Adolescent Cannabis Users Year of Publication 2016 Study Design cross sectional Study Period Not reported Sample Population Adolescents age 14 17 from Miami/Dade county Sample Size 238 Mean Sample Age 15.62 Exposure Measure Lifetime frequency and quantity of cannabis use over the lifetime in grams % of Sample Exposed 84.0% (ever used cannabis); mean amount of lifetime cannabis used 22.80 grams Outcome Measure Body Mass Index z score (relative weight adjusted for child age and sex), and clinical classification of BMI (overweight or obese, underweight or normal weight) % of Sample with Outcome (or Mean Outcome) M ean BMI percentile 0 .64; mean BMI z score 1.06 Measure of Association Standardized regression coefficient and odds ratio Association SRC: 0.187 p value 0.01; OR: 1.001 p value 0.02

PAGE 153

153 Table B 2. Risk of bias judgment for Ross, 2016 Entry Judgment Support for Judgment Selection bias generalizability High risk of bias Sample was 76% Hispanic/Latino all from Miami Dade Information bias classification of exposure Low risk of bias Exact questions/questionnaire used to determine lifetime frequency/quantity of cannabis use in grams was not able to be located; however quantity/frequency of use and cumulative life use generally represent a low risk of bias Information bias classification of outcome Low risk of bias Height and weight recorded by trained personnel Confounding High risk of bias Models were not controlled for any relevant covariates because univariate correlations between BMI z score and nicotine use, alcohol use, MDD, race/ethnicity, education, and estimated IQ were non significant Temporality Low risk of bias BMI measured at time of sampling, all cannabis use was prior to BMI measurement

PAGE 154

154 Yankey, B 2016 Table B 3 Data abstraction for Yankey 2016 Data Item Details Lead Author Barbara N.A. Yankey Title A cross sectional analysis of the association between marijuana and cigarette smoking with metabolic syndrome among adults in the United States Year of Publication 2016 Study Design cross sectional Study Period 2011 2012 Sample Population NHANES; individuals aged 20 years and above Sample Size 3051 Mean Sample Age 38.7 years Exposure Measure Never, non regular and regular marijuana users; years of use % of Sample Exposed 47.0% were regular, 27.0% were non regular, and 26.6% were never users Outcome Measure Metabolic Syndrome having three or more of the following components: 1) hypertension based on blood pressure or use of antihypertensive medication 2) diabetes based on fasting plasma glucose of 100mg/dl or on antidiabetic treatment 3) Abdominal obesity based on waist circum ference of more than 88 cm for women and 102 cm for men 5) low HDL C based on 50mg/dl for males and 40 mg/dl for females 6) Hypertriglyceridemia based on plasma triglycerides of 150mg/dl % of Sample with Outcome (or Mean Outcome) 23.5% with metabolic syndrome Measure of Association Odds ratios

PAGE 155

155 Table B 3. Continued Data Item Details Association Metabolic Syndrome: OR for each year increase in cannabis use 1.05 (1.01, 1.05); regular compared to never users 0.25 (0.06, 1.02 ) Hypertension: Adjusted odds ratio by use status Never: Reference Regular: 0.26 (0.10, 0.67) Non regular: Not Reported Adjusted odds ratio by years of use: 1.05 (1.02, 1.09) Hyperglycemia: Adjusted odds ratio by use status Never: Reference Regular: 0.50 (0.13, 1.89) Non regular: Not Reported Adjusted odds ratio by years of use: 1.02 (0.98, 1.06) Hypertriglyceridemia: Adjusted odds ratio by use status Never: Reference Regular: 0.76 (0.15, 3.97) Non regular: Not Reported Adjusted odds ratio by years of use: 1.03 (1.01, 1.06) Low HDL Cholesterolemia Adjusted odds ratio by use status Never: Reference Regular: 0.54 (0.18, 1.55) Non regular: Not Reported Adjusted odds ratio by years of use: 1.02 (0.99, 1.05)

PAGE 156

156 Table B 4. Risk of bias judgment for Yankey, 2016 Entry Judgment Support for Judgment Selection bias generalizability Low risk of bias Nationally representative dataset Information bias classification of exposure High risk of bias Self reported cannabis use categorized as never, non regular and regular along with self reported number of years used. No information on reliability or validity provided. Information bias classification of outcome Low risk of bias Outcome based on biol ogical samples and standard diagnostic tests, not self report. Confounding Low risk of bias Control variables included age, gender, race/ethnicity, education, and physical activity, marital status, poverty to income ratio, alcohol intake, cocaine use, he roin use, methamphetamine use, and health insurance. No information on mental health conditions provided. Temporality High risk of bias Cross sectional study lacking temporal association between exposure and outcome

PAGE 157

157 Waterreus, 2016 Table B 5. Data abstraction for Waterreu s 2016 Data Item Details Lead Author A. Waterreus Title Metabolic syndrome in people with a psychotic illness: is cannabis protective? Year of Publication 2016 Study Design Cross sectional Study Period April 2010 April 2011 Sample Population Study of High Impact Psychosis (Australian national survey) Sample Size 1,813 Mean Sample Age Majority of participants were in the 25 34 age range (n=566) Exposure Measure Past year cannabis user categorized as non user, occasional user (less than once per week), and frequent user % of Sample Exposed 33.0% with past year cannabis use Outcome Measure Metabolic syndrome (3 out of 5 of the following): 1) waist circumference 2) blood pressure 3) triglycerides 4) glucose (fas ting plasma) 5) HDL OR prescribed medication for hypertension, hyperlipidemia, or hyperglycemia NOTE: to meet glucose criteria participants could self report diabetes % of Sample with Outcome (or Mean Outcome) Prevalence of metabolic syndrome: 57.8% Measure of Association Odds Ratio

PAGE 158

158 Table B 5. Continued Data Item Details Association The adjusted odds ratio for metabolic syndrome among frequent users compared to non users was 0.56 (95% CI 0.39, 0.80), and the adjusted odds ratio for metabolic syndrome among occasional users compared to non users was 0.75 (95% CI 0.49, 1.13). The adju sted odds ratio for hypertension among frequent users compared to non users was 0.58 (95% CI 0.45, 0.75), and the adjusted odds ratio for hypertension among occasional users compared to non users was 0.72 (95% CI 0.54, 0.96). The adjusted odds ratio for elevated fasting glucose among frequent users compared to non users was 0.60 (95% CI 0.43, 0.85), and the adjusted odds ratio for elevated fasting glucose among occasional users compared to non users was 0.59 (95% CI 0.40, 0.88). The adjusted odds ratio for adverse HDL among frequent users compared to non users was 0.68 (95% CI 0.51, 0.92), and the adjusted odds ratio for elevated HDL among occasional users compared to non users was 0.95 (95% CI 0.60, 1.34). The adjusted odds ratio for elevated triglyce rides among frequent users compared to non users was 0.61 (95% CI 0.45, 0.83), and the adjusted odds ratio for elevated triglycerides among occasional users compared to non users was 0.75 (95% CI 0.53, 1.06).

PAGE 159

159 Table B 6. Risk of bias judgment for Waterreus, 2016 Entry Judgment Support for Judgment Selection bias generalizability High risk of bias Study was conducted only among Australian individuals with psychotic illness Information bias classification of exposure low risk of bias Self reported cannabis use over the last 12 months, but accounts for frequency of use over those 12 months Information bias classification of outcome Low risk of bias Generally speaking, a low risk of bias is assessed here because of the use of standard diagnostic/biologic tests to determine metabolic syndrome. However, risk of biased is increased by including self report in the outcome measure. Confounding Low risk of bias Adjusted for age, sex, psychotic diagnosis, antipsychotic use, current smoking, alcohol use, physical activity, SES, cognitive function Temporality High risk of bias There was no temporality; cannabis use was measured for past year only, while metabolic syndrome could have been at point of interview or lifetime depending on whether b iologic measurement or self reported drug use or self reported condition was used to meet criteria

PAGE 160

160 Vidot, 2016 Table B 7 Data abstraction for Vidot, 2016 Data Item Details Lead Author Denise C. Vidot Title Metabolic Syndrome Among Marijuana Users in the United States: An Analysis of National Health and Nutrition Examination Survey Year of Publication 2016 Study Design Cross sectional Study Period 2005 2010 Sample Population NHANES Sample Size 8478 Mean Sample Age Majority of the sample were in age group 45 59 years (n=2947) Exposure Measure Never, past (lifetime) and current (30 day) marijuana users % of Sample Exposed 45.3% never use, 42.7% past use, 11.8% current use Outcome Measure Metabolic Syndrome (3 or more of the following): 1 ) waist circumference 102 cm in men, 88 cm in women 2) hypertension SBP 130mmHg, DBP 85mmHg 3) HDL cholesterol 40 mg/dL for men, 50 mg/dL women 4) triglycerides 150 mg/dL 5) fasting glucose 100 mg/dL NOTE: Other outcomes not included in review becaus e effect estimate (OR or AOR) not provided for component parts of metabolic syndrome individually % of Sample with Outcome (or Mean Outcome) Metabolic syndrome among: 1) Never users 19.5% 2) past users 17.5% 3) current users 13.8% Measure of Association Adjusted Odds Ratio Association Past use compared to never use 0.76 (0.57, 1.02), current use compared to never use 0.69 (0.47, 1.00)

PAGE 161

161 Table B 8. Risk of bias judgment for Vidot, 2016 Entry Judgment Support for Judgment Selection bias generalizability Low risk of bias Nationally representative survey Information bias classification of exposure High risk of bias Self reported measures of cannabis use and no information on quantity/frequency were reported Information bias classification of outcome High risk of bias Standard diagnostic measurements were used, however there was not controlling for individuals who are on medication for things like diabetes, hypertension, cholesterol thus is it is likely individuals with metabolic syndrome were classified as no metabolic syndrome Confounding High risk of bias Associations were controlled for sex, age, race/ethnicity, poverty to income ratio, survey cycle year and cigarette use Other physical activity, substance use and m ental health conditions were not considered leading to a high risk of bias rating Temporality High risk of bias Lifetime measures of cannabis use were assessed and compared to metabolic syndrome at the time of the survey

PAGE 162

162 Blackstone, 2016 Table B 9. Data abstraction for Blackstone, 2016 Data Item Details Lead Author Sarah R. Blackstone Title Relationships Between Illicit Drug Use and Body Mass Index Among Adolescents Year of Publication 2016 Study Design Cross sectional Study Period 2009 2010 Sample Population Health Behavior in School Aged Children Series; United States Adolescents Grades 6 10 Sample Size 10,925 Mean Sample Age Majority of participants were grade 8 (n=2,475) Exposure Measure Dichotomous lifetime marijuana use % of Sample Exposed 14.3% with ever marijuana use Outcome Measure BMI based on self reported height and weight % of Sample with Outcome (or Mean Outcome) Not reported Measure of Association Linear regression (beta coefficient) Association Beta = 0.288; non significant

PAGE 163

163 Table B 10. Risk of bias judgment for Blackstone, 2016 Entry Judgment Support for Judgment Selection bias generalizability High risk of bias Nationally representative survey for adolescents only Information bias classification of exposure High risk of bias Dichotomized use/no use variable that does not account for frequency or quantity of use Information bias classification of outcome High risk of bias Self reported height and weight used to calculate BMI Confounding High risk of bias Adjusted for gender, age and race Not controlled for SES, education, other substance use, mental health conditions or other factors that are associated with BMI such as diabetes or metabolic syndrome Temporality High risk o f bias Cross sectional design does not provide temporality

PAGE 164

164 Bancks, 2015 Table B 11. Data abstraction for Bancks, 2015 Data Item Details Lead Author Michael P. Bancks Title Marijuana use and risk of prediabetes and diabetes by middle adulthood: the Coronary Artery Risk Development in Young Adults (CARDIA) study Year of Publication 2015 Study Design Cross sectional and prospective cohort Study Period Y7 (1992 1993) to Y25 (2010 2011) Sample Population CARDIA Sample Size 3,151 Mean Sample Age Not reported Exposure Measure Ever marijuana use, number of days of use in past 30 days, and cumulative lifetime use % of Sample Exposed 70.1% at Y25 Outcome Measure Diabetes; measured using fasting glucose, 2 h oral glucose tolerance test, HbA1c % of Sample with Outcome (or Mean Outcome) At follow up year 25, 11.3% of participants were classified as diabetic Measure of Association OR (Y25 MJ to Y25 diabetes); HR (cumulative MJ to incident diabetes) Association Cross sectional Adjusted ORs: Former users vs Never Users 1.23 (0.85, 1.78) Current users vs Never Users 1.16 (0.64, 2.08) Prospective adjusted HRs: 1 9 times use vs never use 0.93 (0.68, 1.29) 10 99 times use vs never use 1.30 (0.91, 1.86) >= 100 times use vs never use 1.16 (0.77, 1.7 4) p value for trend 0.32

PAGE 165

165 Table B 12. Risk of bias judgment for Bancks, 2015 Entry Judgment Support for Judgment Selection bias generalizability High risk of bias Participants for CARDIA come from 4 metropolitan populations in the United States and only included Black and White participants, no other racial/ethnic minorities were included Information bias classification of exposure Low risk of bias Accounted for cumulative lifetime marijuana use Information bias classification of outcome Low risk of bias Standard diagnostic/biologic tests were used to determine the presence of diabetes Confounding Low risk of bias Models adjusted for age, sex, race, smoking, alcohol, education, field center, SBP, CRP, physical activity, and use of other illicit drugs. While mental health conditions are not accounted for, many factors also associated with mental health conditions are included. Temporality Low risk of bias Given the prospective nature of the analysis, temporality was adequately addressed

PAGE 166

166 Racine, 2015 Table B 13. Data abstraction for Racine, 2015 Data Item Details Lead Author C. Racine Title Metabolic Effects of Marijuana Use Among Blacks Year of Publication 2015 Study Design Cross sectional Study Period Jan March 2014 Sample Population Black patients from a Family Practice clinic at University Hospital of Brooklyn, NY Sample Size 100 Mean Sample Age 46.3 years Exposure Measure Never, Former (lifetime) & Current (past 180 days) marijuana use; frequency of marijuana use % of Sample Exposed Not Reported Outcome Measure Hypertension, Dyslipidemia, Elevated BMI, Diabetes % of Sample with Outcome ( or Mean Outcome) Hypertension 47%; Dyslipidemia 22%; mean BMI 29.6 Measure of Association t tests and chi square tests

PAGE 167

167 Table B 13. Continued Association No significant differences by never/lifetime use or by never/former/current use Comparison of metabolic markers between current users, former users and never users Current users (SEM) Former users (SEM) Never users (SEM) p value Cholesterol (mg/dL) 156.9 9.42 189.0 8.10 181.8 6.97 0.0569 Triglycerides (mg/dL) 85.9 9.58 133.0 15.00 120.3 12.30 0.1245 LDL cholesterol (mg/dL) 92.5 7.75 105.9 7.92 103.8 6.64 0.5527 HDL cholesterol (mg/dL) 48.3 3.65 57.6 5.05 54.6 3.90 0.4737 Glucose (mg/dL) 108.7 13.27 108.7 6.82 112.6 8.39 0.9287 Hemoglobin A1c (%) 7.2 1.13 6.4 0.39 6.5 0.30 0.6252 BMI (Kg/m 2 ) 26.5 2.06 31.1 1.17 29.6 1.00 0.0906 Waist circumference (inches) 32.1 1.35 35.9 0.88 33.4 0.74 0.0175 Systolic Blood Pressure (mm Hg) 126.0 4.40 129.5 2.52 127.7 2.79 0.7593 Diastolic Blood Pressure (mm Hg) 73.3 2.25 80.0 2.13 73.4 1.58 0.0245

PAGE 168

168 Table B 14. Risk of bias judgment for Racine, 2015 Entry Judgment Support for Judgment Selection bias generalizability High risk of bias Non representative sample of only Black participants Information bias classification of exposure High risk of bias While frequency of marijuana use was measured, it was not used in any of the comparative analyses. Only never, former, current and never vs lifetime use were used in comparative analyses Information bias classification of outcome Low risk of bias Information on metabolic risk factors was gathered via medical record abstraction within the 3 months prior to exposure questionnaire Confounding High risk of bias Given the univariate nature of the analyses, potential confounding was not controlled for Temporality High risk of bias Cross sectional study with the possibility of exposure coming after outcome

PAGE 169

169 Dube, 2015 B 15. Data abstraction for Dube, 2015. Data Item Details Lead Author E. Dube Title Cigarette smoking may modify the association between cannabis use and adiposity in males Year of Publication 2015 Study Design Prospective cohort Study Period 2005 2012 Sample Population Nicotine Dependence in Teens study; Montreal, Canada Sample Size 590 Mean Sample Age 24 at cycle 22 Exposure Measure Past year cannabis use frequency reported as weekly use over the past year % of Sample Exposed Males: 0 times per week 56% 0.1 times per week 17% 0.4 times per week 8.1% 3 times per week 9.2% 7 times per week 9.6% Females: 0 times per we ek 56.7% 0.1 times per week 20.6% 0.4 times per week 7.2% 3 times per week 9.4% 7 times per week 5.9% Outcome Measure Mean change in BMI % of Sample with Outcome (or Mean Outcome) Males: Mean change from age 17 to 24 was 2.0 Females: Mean change from age 17 to 24 was 1.5 Measure of Association Beta coefficient for the association between cannabis use and change in BMI Association Males: 0.09 (0.00, 0.17) Females: 0.09 (0.01, 0.17)

PAGE 170

170 Table B 16. Risk of bias judgment for Dube, 2015 Entry Judgment Support for Judgment Selection bias generalizability High risk of bias Sample consists only of adolescents from 10 schools in Montreal, Canada Information bias classification of exposure Low risk of bias Cannabis use frequency was measured Information bias classification of outcome Low risk of bias BMI was measured using standard techniques. Inter rater reliability for repeat measures for 1 in 10 participants was 0.99 for height and 0.99 for weight Confounding Low risk of bias Sex, sedentary behavior, physical activity, alcohol use, anxiety, depression, cigarette use and maternal education were explored as potential confounders Temporality low risk of bias prospective cohort study design, the exposure was measured in cycle 21 the ou tcome was measured as change from cycle 19 to cycle 22

PAGE 171

171 Thompson, 2015 Table B 17. Data abstraction table for Thompson, 2015 Data Item Details Lead Author CA Thompson Title Estimating the association between metabolic risk factors and marijuana use in U.S. adults using data from the continuous National Health and Nutrition Examination Survey Year of Publication 2015 Study Design Cross sectional Study Period 2005 2012 Sample Population NHANES Sample Size 6281 Mean Sample Age Not Reported Exposure Measure Never, past and current marijuana use % of Sample Exposed 54.4% with any marijuana use Outcome Measure Hypertension, elevated BMI, dyslipidemia, diabetes % of Sample with Outcome (or Mean Outcome) Not Reported Measure of Association Linear regression Association p < .05; ** p < .01. The text indicates what is labeled as mean is truly a regression coefficient, so they have been interpreted as such. Dependent variables Past use Current use Mean SE Mean SE Triglycerides (mg/dL) 0.022 0.022 0.035 HDL C (mg/dL) 0.503 0.726 0.621 BMI (kg/m2) 0.051 0.223 0.340 SBP (mm Hg) 0.560 0.251 0.785 DBP (mm Hg) 0.080 0.448 0.513 0.672 Glucose (mg/dL) 1.503 0.866 1.324 1.472

PAGE 172

172 Table B 18. Risk of bias judgment for Thompson, 2015 Entry Judgment Support for Judgment Selection bias generalizability Low risk of bias Nationally representative sample Information bias classification of exposure High risk of bias No information on quantity/frequency; discussion specifically mentions this as a limitation Information bias classification of outcome Low risk of bias Standard diagnostic/biologic tests were used to determine outcome variables Confounding Low risk of bias Regression models were adjusted for age, sex, race, education, income, BMI, tobacco use, alcohol use and physical activity; discussion specifically notes the lack of other substance use variables as a limitation The dataset lacks proper information on other behavior al and environmental variables necessary to control for unobserved differences per Thomas and Hays, 2016 Temporality High risk of bias Study design did not ensure that exposure preceded outcome

PAGE 173

1 73 Penner, 2013 Table B 19. Data abstraction for Penner, 2013 Data Item Details Lead Author Elizabeth A Penner Title The Impact of Marijuana Use on Glucose, Insulin, and Insulin Resistance among US Adults Year of Publication 2013 Study Design Cross sectional Study Period 2005 2010 Sample Population NHANES Sample Size 4657 Mean Sample Age Majority of individuals were age group 30 44 (n=1732) Exposure Measure Never, past and current (30 day) cannabis use % of Sample Exposed 12.2% are current users, 47.7% past users Outcome Measure Elevated BMI, hypertension, Dyslipidemia, Diabetes % of Sample with Outcome (or Mean Outcome) Mean BMI: Never Users: 29.1, Past Users: 28.5, Current Users: 27.2 Mean SBP: Never Users: 117.4, Past Users: 117.0, Current Users: 118.8 Mean D BP: Never Users: 70.4, Past Users: 70.5, Current Users: 69.3 Triglycerides geometric mean: Never Users: 108.5, Past Users: 111.1, Current Users: 110.8 Measure of Association Multivariable Adjusted Mean/Percent Differences Association Mean BMI: Never Users: Ref, Past Users: 0.08 ( 0.6, 0.47), Current Users: 0.61 ( 1.31, 0.09) Mean SBP: Never Users: Ref, Past Users: 1.04 ( 2.55, 0.47), Current Users: 0.64 ( 1.11, 2.39) Mean DBP: Never Users: Ref, Past Users: 0.01 ( 1.06, 1.04), Current Users: 0.49 ( 0.98, 1.96) Triglycerides: Never Users: Ref, Past Users: 0.29% ( 5.1%, 6.0%), Current Users: 1.2% ( 6.9%, 8.8%) HDL C: Never Users: Ref, Past Users: 0.14 ( 1.10, 1.38), Current Users: 1.22 ( 0.25, 2.70) Fasting Glucose: Never Users: Ref, Past Users: 2.1 6 ( 4.22, 0.11), Current Users: 0.47 ( 2.51, 1.57) HgA1c: Never Users: Ref, Past Users: 0.07 ( 0.17, 0.02) Current Users: 0.02 ( 0.11, 0.15)

PAGE 174

174 Table B 20. Risk of bias judgment for Penner, 2013 Entry Judgment Support for Judgment Selection bias generalizability Low risk of bias Nationally representative study Information bias classification of exposure High risk of bias No measures of quantity or frequency were assessed Information bias classification of outcome Low risk of bias Blood tests were used to measure biologic outcomes, standard diagnostic methods were used to collect information o n BMI, BP and WC Confounding Low risk of bias Controlled for age, sex, race/ethnicity, education, income, marital status, tobacco use, physical activity, and alcohol use. No information on mental health conditions or other substance use Letter to the editor published by Thompson about this problem in the American journal of medicine Temporality Low risk of bias All outcome measurements taken at the time of the survey, while exposure variables were necessarily prior to outcome measurements

PAGE 175

175 Rajavashisth, 2012 Table B 21. Data abstraction for Rajavashisth, 2012 Data Item Details Lead Author Tripathi B Rajavashisth Title Decreased prevalence of diabetes in marijuana users: cross sectional data from the National Health and Nutrition Examination Survey (NHANES) III Year of Publication 2012 Study Design Cross sectional Study Period 1988 1994 Sample Population NHANES Sample Size 8,127 adults Mean Sample Age Not reported Exposure Measure Never use, lifetime use, light current use (past 30 days with use <= 4 days per month), heavy current (past 30 days with use >= 5 days per month) For adjusted analyses only never/lifetime was used % of Sample Exposed Never Use n=6667, past users n=3346, light current n=557, heavy current n=326 Outcome Measure Type 2 Diabetes Mellit us, based on self report and/or fasting blood glucose level (did not exclude individuals thought to have Type 1 diabetes) % of Sample with Outcome N=719, 8.8% Measure of Association Adjusted Odds Ratio Association OR 0.36 (0.24, 0.55) p<0.0001

PAGE 176

176 Table B 22. Risk of bias judgment for Rajavashisth, 2012 Entry Judgment Support for Judgment Selection bias generalizability Low risk of bias Nationally representative sample Information bias classification of exposure High risk of bias While there is minimal accounting for frequency of use, it does not span the entire exposure measure Information bias classification of outcome High risk of bias Of the 719 individuals with DM, 525 were identified through self report only, and 194 were identified thro ugh fasting blood glucose levels greater than or equal to 126 Confounding Low risk of bias Final model adjusted for race/ethnicity, physical activity, alcohol use, interaction of alcohol and marijuana use, BMI, total cholesterol, triglycerides, CRP, and hypertension Temporality High risk of bias Cross sectional nature of the study combined with exposure and outcome measures means there was no assurance of temporality

PAGE 177

177 Hayatbakhsh, 2010 Table B 23. Data abstraction table for Hayatbakhs h, 2010 Data Item Details Lead Author Mohammad R Hayatbakhsh Title Cannabis Use and Obesity and Young Adults Year of Publication 2010 Study Design Cross sectional (while overall study design is prospective cohort, these analyses only used data from 21 year follow up of offspring) Study Period Enrollment at Birth 1981 1983; F/U at offspring age 14 and 21 Sample Population Mater University Study of Pregnancy and its Outcomes (MUSP) Brisbane Australia Sample Size 2,518 Mean Sample Age 20.56 at last follow up Exposure Measure Duration of cannabis use (never used, used once or less in the last month, use at least every few days in the last month with onset before 16 years, and used at least every few days in the last month with onset at 16 years or later) % of Sample Exposed Lifetime cannabis use 50.9% Outcome Measure Measured BMI % of Sample with Outcome (or Mean Outcome) 65.7% with normal BMI, 21.4% overweight, 12.9% obese (in multivariate analyses overweight and obese were combined) Measure of Association Adjusted Odds Ratios Association Never Used: Ref Once or less in last month: 1.0 (0.8, 1.4) Started before 16 years, used at least once in the last month: 0.3 (0.2, 0.5) Started at 16 years or above and used at least few days a week in the last month: 0.4 (0.2, 0.7)

PAGE 178

178 Table B 24. Risk of bias judgment for Hayatbakhs h, 2010 Entry Judgment Support for Judgment Selection bias generalizability High risk of bias Participants came from one hospital in Brisbane, Australia Information bias classification of exposure High risk of bias No information on quantity, limited information on frequency (i.e. frequency within the last month at the 21 year follow up only), age of onset only measure of duration. Information bias classification of out come Low risk of bias BMI at follow up visits were calculated from measured height and weight on a validated scale Confounding Low risk of bias Final model adjusted for gender, age, cigarette smoking, alcohol consumption, anxiety/depression, aggression/delinquency, BMI at age 14, physical activity Temporality low risk of bias Marijuana use necessarily came before measured BMI at follow up visit

PAGE 179

179 Rodondi, 2006 Table B 25. Data abstraction for Rodondi, 2006 Data Item Details Lead Author Nicolas Rodondi Title Marijuana Use, Diet, Body Mass Index, and Cardiovascular Risk Factors (from the CARDIA Study) Year of Publication 2006 Study Design Prospective cohort Study Period 1985 2000 (enrollment to Y15 follow up) Sample Population CARDIA Sample Size 3617 Mean Sample Age 40.1 years Exposure Measure Total days of use over the 15 year study period % of Sample Exposed 37.7% lifetime use 62.3% Never use 16.8% <180 days 16.6% 180 1799 days 4.3% >=1800 days Outcome Measure BMI, hypertension (SBP/DBP), triglycerides % of Sample with Outcome (or Mean Outcome) Not Reported Measure of Association Adjusted mean values Association BMI: Never User: 28.8, <180 Days: 28.6, 180 1799 Days: 28.8, >=1800 Days: 28.9 p value 0.65 SBP: Never User: 113.6, <180 Days: 112.9, 180 1799 Days: 112.1, >=1800 Days: 112.9 p value 0.34 DBP: Never User: 74.8, <180 Days: 74.2, 180 1799 Days: 74.0, >=1800 Days: 73.9 p value 0.28 Triglycerides mg/dl: Never User: 86.7, <180 Days: 88.5, 180 1799 Days: 87.6, >=1800 Days: 92.9 p value 0.08 Adjusted Mean Values Fasting glucose Never User: 86.7 <180 Days: 86.3 180 1799 Days: 86.8 1800 Days: 87.4

PAGE 180

180 Table B 26. Risk of bias judgment for Rodondi, 2006 Entry Judgment Support for Ju dgment Selection bias generalizability High risk of bias Not nationally representative; only has 4 metropolitan areas considered and also lacks information on individuals who are not Black or White Information bias classification of exposure High risk of bias While days of use over the 15 years is superior to never, past, current user status number of days was determined by extrapolating past 30 day use to cover 3 and 5 year increments, presuming the same patterns of use every month for five years is a stretch Information bias classification of outcome Low risk of bias Standard measurements were used to determine BMI, measurements of SBP and DBP were measured according to plasma levels were used to gather triglycerides Confounding Low risk of bias Alcohol use, race, gender, age, education, annual income, smoking status, illicit drug use, and physical activity were all controlled for Study center was also controlled for Mental health conditions were not considered Te mporality Low risk of bias Outcomes were measured at Y15 follow up, exposure was cumulative over the course of the study

PAGE 181

181 Mittleman, 2001 Table B 27. Data abstraction for Mittleman, 2001 Data Item Details Lead Author Murray A Mittleman Title Triggering Myocardial Invarction by Marijuana Year of Publication 2001 Study Design Case crossover Study Period 1989 1996 Sample Population Onset Study; Patients interviewed at 64 medical centers a median of 4 days after MI Sample Size 3,882 Mean Sample Age 44 years for users 62 years for nonusers Exposure Measure MJ use in year prior to MI used to estimate expected frequency in an average 1 hour period % of Sample Exposed 3.2% Outcome Measure MI based on standard diagnostic criteria and chart review % of Sample with Outcome (or Mean Outcome) 100% (case crossover design) Measure of Association Relative Risk Association Within 1 hour after smoking marijuana relative risk 4.8 (2.9, 9.5) p value <0.001 compared with periods of nonuse. Second hour after smoking marijuana relative risk was 1.7 (0.6, 5.1) p value=0.34

PAGE 182

182 Table B 28. Risk of bias judgment for Mittleman, 2001 Entry Judgment Support for Judgment Selection bias generalizability High risk of bias Participants all came from medical centers in Maryland Data on study design from Mittleman et Information bias classification of exposure High risk of bias Use or no use in the year prior to MI was used to estimate the expected frequency in an average 1 hour period for this particular population Information bias classification of outcome Low risk of bias Standard diagnostic tests accompanied by chart review Confounding Low risk of bias The case crossover design allows cases to be their own controls, handling confounding by variables that are stable over time but differ between individuals Temporality Low risk of bias Case crossover design ensured that the exposure occurred prior to the outc ome

PAGE 183

183 Huang, 2013 Table B 29. Data abstraction for Huang, 2013 Data Item Details Lead Author David HC Huang Title Association between Adolescent Substance Use and Obesity in Young Adulthood: A Group based Dual Trajectory Analysis Year of Publication 2013 Study Design Prospective Cohort Study Period 1986 2008 Sample Population Subset of the child sample of the 1979 National Longitudinal Survey of Youth Sample Size 5,141 Mean Sample Age Not Reported; Exposure Measure Ever marijuana use at each wave of the study, measured at 12 times points and put into trajectory groups % of Sample Exposed 77.9% were in the low use group, 16.3% in the sporadic use group, and the 5.8% in the increasing use group Outcome Measure Obesity (elevated BMI) calculated from self reported weight and height, measured 12 times and put into trajectory groups % of Sample with Outcome (or Mean Outcome) 72% in the low risk of obesity, 15% increase in obesity, 13% sustained high risk of obesity Measure of Ass ociation Odds of obesity trajectories given substance use trajectories Association 1) Adolescents in the increasing marijuana trajectory had 1.6 p value 0.05 times the odds of being in the increased obesity trajectory compared to adolescents in the low marijuana trajectory but exhibited similar risk of belonging to high obesity trajector y compared to low marijuana 2) sporadic marijuana trajectory compared to low marijuana trajectory OR 0.2 p value <0.01 for high obesity and for increased obesity OR=0.1; p value0.01

PAGE 184

184 Table B 30. Risk of bias judgment for Huang, 2013 Entry Judgment Support for Judgment Selection bias generalizability Low risk of bias This is a nationally representative sample Information bias classification of exposure High risk of bias Though the study had available information on quantity/frequency and date of most recent use low prevalence of marijuana use in the sample meant authors used self reported yes/no use at each wave Information bias classification of outcome High risk of bias Height and weight were self reported Confounding High risk of bias Adjusted analyses controlled for gender, ethnicity, adolescent obesity status, and average family income. Not controlled for other substance use, mental health conditions or physical activity Temporality Low risk of bias Substance use was examined from ages 12 18 and obesity was examined from ages 20 to 24

PAGE 185

185 Aronow 1974 Table B 31 Data abstraction for Aronow, 1974 Data Item Details Lead Author Wilbert S Aronow Title Effect of marihuana and placebo marihuana smoking on angina pectoris Year of Publication 1974 Study Design Crossover Study Period 1974 Sample Population Men with classic stable exertional angina and >75 percent narrowing at major coronary vessels Sample Size 10 Mean Sample Age 47.3 Exposure Measure 10 puffs of marihuana cigarette % of Sample Exposed 100%, individuals served as their own controls Outcome Measure Time to angina pectoris while riding exercise bike % of Sample with Outcome (or Mean Outcome) 100%, all patients had to have angina Measure of Association t test Association 48% decrease in time to angina greater than placebo cigarette which caused 8.6% decrease p values<0.001 Table B 32. Risk of bias judgment for Aronow, 1974 Entry Judgment Support for Judgment Selection bias generalizability High risk of bias Small sample of 10 patients Information bias classification of exposure Low risk of bias Patients smoked marihuana in front of researchers Information bias classification of outcome Low risk of bias Diagnostic tests appropriate for the time were used to determine angina Confounding low risk of bias Patients smoked an actual cigarette and then smoked placebo, so they were their own controls Temporality Low risk of bias Use certainly occurred prior to onset of angina pectoris

PAGE 186

186 LIST OF REFERENCES Agrawal, A., Grant, J.D., Waldron, M., Duncan, A.E., Scherrer, J.F., Lynskey, M.T., Madden, P.A.F., Bucholz, K.K., Heath, A.C., 2006. Risk for initiation of substance use as a function of age of onset of cigarette, alcohol and cannabis use: Findings in a Midwestern female twin cohort. Prev. Med. 43, 125 128. doi:10.1016/j.ypmed.2006.03.022 Agrawal, A., Nurnberger, J.I., Lynskey, M.T., 2011. Canna bis involvement in individuals with bipolar disorder. Psychiatry Res. 185, 459 461. doi:10.1016/j.psychres.2010.07.007 Albert, M.A., Glynn, R.J., Buring, J., Ridker, P.M., 2006. Impact of Traditional and Novel Risk Factors on the Relationship Between Socio economic Status and Incident Cardiovascular Events. Circulation 114, 2619 2626. doi:10.1161/CIRCULATIONAHA.106.660043 Alberti, K.G.M.M., Eckel, R.H., Grundy, S.M., Zimmet, P.Z., Cleeman, J.I., Donato, K.A., Fruchart, J. C., James, W.P.T., Loria, C.M., Smit h, S.C., International Diabetes Federation Task Force on Epidemiology and Prevention, Hational Heart, Lung, and Blood Institute, American Heart Association, World Heart Federation, International Atherosclerosis Society, International Association for the St udy of Obesity, 2009. Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Feder ation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation 120, 1640 1645. doi:10.1161/CIRCULATIONAHA.109.192644 Allison, P.D., 2014. Measures of Fit for Logistic Regression (No. 1485 2014), SAS Global Forum. American Diabetes Association, 2016. Classification and Diagnosis of Diabetes. Diabetes Care 39, S13 S22. doi:10.2337/dc16 S005 American Psychiatric Association, 2013. Substance Related and Addictive Disorders, in: Diagnostic and Statistical Manual of Mental Disorders. American Psychiatric Association. American Psychiatric Association, 1994. Diagnostic and statistical manual of mental disorders, 4th ed. Washington, DC, US. Ariyo, A.A., Haan, M., Tangen, C.M., Rutledge, J.C., Cushman, M., Dobs, A., F urberg, C.D., Group, for the C.H.S.C.R., 2000. Depressive Symptoms and Risks of Coronary Heart Disease and Mortality in Elderly Americans. Circulation 102, 1773 1779. doi:10.1161/01.CIR.102.15.1773 Aronow, W.S., Cassidy, J., 1974. Effect of marihuana and placebo marihuana smoking on angina pectoris. N. Engl. J. Med. 291, 65 67. doi:10.1056/NEJM197407112910203 Aryana, A., Williams, M.A., 2007. Marijuana as a trigger of cardiovascular events: Speculation or scientific certainty? Int. J. Cardiol. 118, 141 144. doi:10.1016/j.ijcard.2006.08.001

PAGE 187

187 Austin, P.C., Steyerberg, E.W., 2012. Interpreting the concordance statistic of a logistic regression model: relation to the variance and odds ratio of a continuous explanatory variable. BMC Med. Res. Methodol. 12, 82. doi:10.1186/1471 2288 12 82 Baggio, S., Spilka, S., Studer, J., Iglesias, K., Gmel, G., 2016. Trajectories of drug use among French young people: Prototypical stages of involvement in illicit drug use. J. Subst. Use 21, 485 490. doi:10.3109/14659891.2 015.1063720 Balhara, Y.P.S., 2012. Tobacco and metabolic syndrome. Indian J. Endocrinol. Metab. 16, 81. doi:10.4103/2230 8210.91197 Bancks, M.P., Pletcher, M.J., Kertesz, S.G., Sidney, S., Rana, J.S., Schreiner, P.J., 2015. Marijuana use and risk of predia betes and diabetes by middle adulthood: the Coronary Artery Risk Development in Young Adults (CARDIA) study. Diabetologia 58, 2736 2744. doi:10.1007/s00125 015 3740 3 Baune, B.T., Stuart, M., Gilmour, A., Wersching, H., Heindel, W., Arolt, V., Berger, K., 2012. The relationship between subtypes of depression and cardiovascular disease: a systematic review of biological models. Transl. Psychiatry 2, e92. doi:10.1038/tp.2012.18 Baune, B.T., Tully, P.J., 2016. Cardiovascular Diseases and Depression: Treatment and Prevention in Psychocardiology. Springer. Beaconsfield, P., 1974. Some cardiovascular effects of cannabis. Am. Heart J. 87, 143 146. Beilin, L.J., Puddey, I.B., 2006. Alcohol and Hypertension. Hypertension 47, 1035 1038. doi:10.1161/01.HYP.0000218586.2 1932.3c Benjamin, E.J., Blaha, M.J., Chiuve, S.E., Cushman, M., Das, S.R., Deo, R., Ferranti, S.D. de, Floyd, J., Fornage, M., Gillespie, C., Isasi, C.R., Jimnez, M.C., Jordan, L.C., Judd, S.E., Lackland, D., Lichtman, J.H., Lisabeth, L., Liu, S., Longene cker, C.T., Mackey, R.H., Matsushita, K., Mozaffarian, D., Mussolino, M.E., Nasir, K., Neumar, R.W., Palaniappan, L., Pandey, D.K., Thiagarajan, R.R., Reeves, M.J., Ritchey, M., Rodriguez, C.J., Roth, G.A., Rosamond, W.D., Sasson, C., Towfighi, A., Tsao, C .W., Turner, M.B., Virani, S.S., Voeks, J.H., Willey, J.Z., Wilkins, J.T., Wu, J.H., Alger, H.M., Wong, S.S., Muntner, P., 2017. Heart Disease and Stroke Statistics 2017 Update: A Report From the American Heart Association. Circulation 135, e146 e603. doi: 10.1161/CIR.0000000000000485 Bensaid, M., Gary Bobo, M., Esclangon, A., Maffrand, J.P., Le Fur, G., Oury Donat, F., Soubri, P., 2003. The cannabinoid CB1 receptor antagonist SR141716 increases Acrp30 mRNA expression in adipose tissue of obese fa/fa rats a nd in cultured adipocyte cells. Mol. Pharmacol. 63, 908 914. Blackstone, S.R., Herrmann, L.K., 2016. Relationships Between Illicit Drug Use and Body Mass Index Among Adolescents. Health Educ. Behav. 43, 21 24. doi:10.1177/1090198115579414

PAGE 188

188 Bose, J., Hedden, S., Lipari, R., Park Lee, E., 2016. Key Substance Use and Mental Health Indicators in the United States: Results from the 2015 National Survey on Drug Use and Health. Substance Abuse and Mental Health Services Administration. Bostwick, J.M., 2012. Blurred Boundaries: The Therapeutics and Politics of Medical Marijuana. Mayo Clin. Proc. 87, 172 186. doi:10.1016/j.mayocp.2011.10.003 Brunner, E.J., Mosdl, A., Witte, D.R., Martikainen, P., Stafford, M., Shipley, M.J., Marmot, M.G., 2008. Dietary patterns and 1 5 y risks of major coronary events, diabetes, and mortality. Am. J. Clin. Nutr. 87, 1414 1421. Buckner, J.D., Heimberg, R.G., Schneier, F.R., Liu, S. M., Wang, S., Blanco, C., 2012. The relationship between cannabis use disorders and social anxiety disorde r in the National Epidemiological Study of Alcohol and Related Conditions (NESARC). Drug Alcohol Depend. 124, 128 134. doi:10.1016/j.drugalcdep.2011.12.023 Carnethon, M.R., 2009. Physical Activity and Cardiovascular Disease: How Much Is Enough? Am. J. Life style Med. 3, 44S 49S. doi:10.1177/1559827609332737 Carney, R.M., Freedland, K.E., 2016. Depression and coronary heart disease. Nat. Rev. Cardiol. advance online publication. doi:10.1038/nrcardio.2016.181 Center for Behavioral Health Statistics and Quality 2016. Results from the 2015 National Survey on Drug Use and Health: Detailed Tables. Substance Abuse and Mental Health Services Administration. Center for Behavioral Health Statistics and Quality, 2015. 2014 National Survey on Drug Use and Health: Detail ed Tables. Substance Abuse and Mental Health Services Administration, Rockville, MD. Centers for Disease Control and Prevention (US), National Center for Chronic Disease Prevention and Health Promotion (US), Office on Smoking and Health (US), 2010. How Tob acco Smoke Causes Disease: The Biology and Behavioral Basis for Smoking Attributable Disease: A Report of the Surgeon General, Publications and Reports of the Surgeon General. Centers for Disease Control and Prevention (US), Atlanta (GA). Chen, X., Yu, B., Lasopa, S.O., Cottler, L.B., 2016. Current patterns of marijuana use initiation by age among US adolescents and emerging adults: implications for intervention. Am. J. Drug Alcohol Abuse 0, 1 10. doi:10.3109/00952990.2016.1165239 Chiva Blanch, G., Arranz, S., Lamuela Raventos, R.M., Estruch, R., 2013. Effects of Wine, Alcohol and Polyphenols on Cardiovascular Disease Risk Factors: Evidences from Human Studies. Alcohol Alcohol 48, 270 277. doi:10.1093/alcalc/agt007 Chobanian, A.V., Bakris, G.L., Black, H.R., Cushman, W.C., Green, L.A., Izzo, J.L., Jones, D.W., Materson, B.J., Oparil, S., Wright, J.T., Roccella, E.J., Committee, the N.H.B.P.E.P.C., 2003. Seventh Report of the Joint National Committee on Prevention,

PAGE 189

189 Detection, Evaluation, and Treatment of High Blood Pressure. Hypertension 42, 1206 1252. doi:10.1161/01.HYP.0000107251.49515.c2 Coleman Jensen, A., Gregory, C., Singh, A., 2014. Household Food Security in the United States in 2013 (SSRN Scholarly Paper No. ID 2504067). Social Science Research Networ k, Rochester, NY. Compton, W.M., Cottler, L.B., Ben Abdallah, A., Phelps, D.L., Spitznagel, E.L., Horton, J.C., 2000. Substance dependence and other psychiatric disorders among drug dependent subjects: race and gender correlates. Am. J. Addict. 9, 113 125. Compton, W.M., Gfroerer, J., Conway, K.P., Finger, M.S., 2014. Unemployment and substance outcomes in the United States 2002 2010. Drug Alcohol Depend. 142, 350 353. doi:10.1016/j.drugalcdep.2014.06.012 Conway, K.P., Swendsen, J., Husky, M.M., He, J. P., Merikangas, K.R., 2016. Association of Lifetime Mental Disorders and Subsequent Alcohol and Illicit Drug Use: Results From the National Comorbidity Survey Adolescent Supplement. J. Am. Acad. Child Adolesc. Psychiatry 55, 280 288. doi:10.1016/j.jaac.2016.01 .006 Copeland, J., Rooke, S., Swift, W., 2013. Changes in cannabis use among young people: impact on mental health. Curr. Opin. Psychiatry 26, 325 329. doi:10.1097/YCO.0b013e328361eae5 Cota, D., 2007. CB1 receptors: emerging evidence for central and peripheral mechanisms that regulate energy balance, metabolism, and cardiovascular health. Diabetes Metab. Res. Rev. 23, 507 517. doi:10.1002/dmrr.764 Cota, D., Marsicano, G., Tschp, M., Grbler, Y., Flachskamm, C., Schubert, M., Auer, D., Yassouridis, A. Thne Reineke, C., Ortmann, S., Tomassoni, F., Cervino, C., Nisoli, E., Linthorst, A.C.E., Pasquali, R., Lutz, B., Stalla, G.K., Pagotto, U., 2003. The endogenous cannabinoid system affects energy balance via central orexigenic drive and peripheral lipog enesis. J. Clin. Invest. 112, 423 431. doi:10.1172/JCI17725 Based Approach to Include Mental Health in Public Health Research, in: Mental Health in Public Health: The next 100 Years Oxford University Press, New York, pp. 247 273. Cottler, L.B., Striley, C.W., O Leary, C.C., Ruktanonchai, C.W., Wilhelm, K.A., 2012. Engaging the Community in Research with the HealthStreet Model: National and International Perspectives, in: Alving, B., Dai, K., Chan, S.H.H. (Eds.), Translational Research in Biomedicine. S. KARGER AG, Basel, pp. 98 109. Croci, T., Manara, L., Aureggi, G., Guagnini, F., Rinaldi Carmona, M., Maffrand, J.P., Le Fur, G., Mukenge, S., Ferla, G., 1998. In vitro functional evid ence of neuronal cannabinoid CB1 receptors in human ileum. Br. J. Pharmacol. 125, 1393 1395. doi:10.1038/sj.bjp.0702190

PAGE 190

190 Dahdouh, Z., Roule, V., Lognon, T., Sabatier, R., Grollier, G., 2011. Cannabis and coronary thrombosis: What is the role of platelets? Platelets 23, 243 245. doi:10.3109/09537104.2011.601824 Davis, J.M., Mendelson, B., Berkes, J.J., Suleta, K., Corsi, K.F., Booth, R.E., 2015. Public Health Effects of Medical Marijuana Legalization in Colorado. Am. J. Prev. Med. doi:10.1016/j.amepre.2015.0 6.034 De Giorgi, A., Fabbian, F., Pala, M., Bonetti, F., Babini, I., Bagnaresi, I., Manfredini, F., Portaluppi, F., Mikhailidis, D., Manfredini, R., 2012. Cocaine and Acute Vascular Diseases. Curr. Drug Abuse Rev. 5, 129 134. Degenhardt, L., Hall, W., 2012 Extent of illicit drug use and dependence, and their contribution to the global burden of disease. The Lancet 379, 55 70. doi:10.1016/S0140 6736(11)61138 0 Degenhardt, L., Hall, W., Lynskey, M.T., 2001. The relationship between cannabis use, depression a nd anxiety among Australian adults: findings from the National Survey of Mental Health and Well Being. Soc. Psychiatry Psychiatr. Epidemiol. 36, 219 227. Del Boca, F.K., Darkes, J., 2003. The validity of self reports of alcohol consumption: state of the sc ience and challenges for research. Addiction 98, 1 12. doi:10.1046/j.1359 6357.2003.00586.x Department of Justice, 2016. Denial of Petition to Initiate Proceedings to Reschedule Marijuana; Proposed Rules and Applications to Become Registered Under the Cont rolled Substances Act to Manufacture Marijuana to Supply Researchers in the United States; Policy Statment. Fed. Regist. 81, 53688 53766. Desprs, J. P., Golay, A., Sjstrm, L., Rimonabant in Obesity Lipids Study Group, 2005. Effects of rimonabant on meta bolic risk factors in overweight patients with dyslipidemia. N. Engl. J. Med. 353, 2121 2134. doi:10.1056/NEJMoa044537 Dodani, S., Ruktanonchai, C.W., Kaeley, G.S., Vaddiparti, K., Striley, C.W., Cottler, L.B., 2016. Clinical Comorbidities among Cocaine Us ers Screened in the Community through HealthStreet. Health Behav. Policy Rev. 3, 54 61. doi:10.14485/HBPR.3.1.6 Aswad, D., 2015. Cigarette smoking may modify the association between cannabis use and adiposity in m ales. Pharmacol. Biochem. Behav. 135, 121 127. doi:10.1016/j.pbb.2015.04.018 DW, B., JS, B., C, Z., P, C., M, W., 2002. Drug use and the risk of major depressive disorder, alcohol dependence, and substance use disorders. Arch. Gen. Psychiatry 59, 1039 1044 doi:10.1001/archpsyc.59.11.1039 Elderon, L., Whooley, M.A., 2013. Depression and Cardiovascular Disease. Prog. Cardiovasc. Dis., Symposium on Psychosocial Factors in Cardiovascular Disease 55, 511 523. doi:10.1016/j.pcad.2013.03.010

PAGE 191

191 ElSohly, M.A., Mehmed ic, Z., Foster, S., Gon, C., Chandra, S., Church, J.C., 2016. Changes in Cannabis Potency Over the Last 2 Decades (1995 2014): Analysis of Current Data in the United States. Biol. Psychiatry, Cannabinoids and Psychotic Disorders 79, 613 619. doi:10.1016/j. biopsych.2016.01.004 Eng, P.M., Rimm, E.B., Fitzmaurice, G., Kawachi, I., 2002. Social Ties and Change in Social Ties in Relation to Subsequent Total and Cause specific Mortality and Coronary Heart Disease Incidence in Men. Am. J. Epidemiol. 155, 700 709. doi:10.1093/aje/155.8.700 Engel, J., 1988. Polytomous logistic regression. Stat. Neerlandica 42, 233 252. doi:10.1111/j.1467 9574.1988.tb01238.x Farmer, R.F., Kosty, D.B., Seeley, J.R., Duncan, S.C., Lynskey, M.T., Rohde, P., Klein, D.N., Lewinsohn, P.M., 2015. Natural course of cannabis use disorders. Psychol. Med. 45, 63 72. doi:10.1017/S003329171400107X Fedak, K.M., Bernal, A., Capshaw, Z.A., Gross, S., 2015. Applying the Bradford Hill criteria in the 21st century: how data integration has changed causal inference in molecular epidemiology. Emerg. Themes Epidemiol. 12. doi:10.1186/s12982 015 0037 4 Feingold, D., Weiser, M., Rehm, J., Lev Ran, S., 2015. The association between cannabis use and mood disorders: A longitudinal study. J. Affect. Disord. 172, 2 11 218. doi:10.1016/j.jad.2014.10.006 Fergusson, D.M., Boden, J.M., Horwood, L.J., 2006. Cannabis use and other illicit drug use: testing the cannabis gateway hypothesis. Addiction 101, 556 569. doi:10.1111/j.1360 0443.2005.01322.x Fitzgerald, K.C., Chiuve S.E., Buring, J.E., Ridker, P.M., Glynn, R.J., 2012. Comparison of associations of adherence to a Dietary Approaches to Stop Hypertension (DASH) style diet with risks of cardiovascular disease and venous thromboembolism. J. Thromb. Haemost. JTH 10, 189 1 98. doi:10.1111/j.1538 7836.2011.04588.x Fodor, G., 2015. Primary Prevention of CVD: Treating Dyslipidemia. Am. Fam. Physician 83, 1207 1208. Foltin, R.W., Fischman, M.W., Byrne, M.F., 1988. Effects of smoked marijuana on food intake and body weight of hum ans living in a residential laboratory. Appetite 11, 1 14. Franz, C.A., Frishman, W.H., 2016. Marijuana Use and Cardiovascular Disease: Cardiol. Rev. 24, 158 162. doi:10.1097/CRD.0000000000000103 Frasure Smith N, Lesprance F, Talajic M, 1993. Depression f ollowing myocardial infarction: Impact on 6 month survival. JAMA 270, 1819 1825. doi:10.1001/jama.1993.03510150053029

PAGE 192

192 Frost, L., Mostofsky, E., Rosenbloom, J.I., Mukamal, K.J., Mittleman, M.A., 2013. Marijuana Use and Long Term Mortality among Survivors of Acute Myocardial Infarction. Am. Heart J. 165, 170 175. doi:10.1016/j.ahj.2012.11.007 Galea, S., Tracy, M., Hoggatt, K.J., DiMaggio, C., Karpati, A., 2011. Estimated Deaths Attributable to Social Factors in the United States. Am. J. Public Health 101, 145 6 1465. doi:10.2105/AJPH.2010.300086 Ghuran, A., Nolan, J., 2000. The cardiac complications of recreational drug use. West. J. Med. 173, 412 415. Glanz, K., Rimer, B.K., Viswanath, K. (Eds.), 2008. Health Behavior and Health Education: Theory, Research, an d Practice, 4th ed. Jossey Bass, San Francisco, CA. Goldschmidt, P.G., 1986. Information synthesis: a practical guide. Health Serv. Res. 21, 215 237. Goldstein, L.B., Bushnell, C.D., Adams, R.J., Appel, L.J., Braun, L.T., Chaturvedi, S., Creager, M.A., Cul ebras, A., Eckel, R.H., Hart, R.G., Hinchey, J.A., Howard, V.J., Jauch, E.C., Levine, S.R., Meschia, J.F., Moore, W.S., Nixon, J.V. (Ian), Pearson, T.A., 2011. Guidelines for the Primary Prevention of Stroke. Stroke 42, 517 584. doi:10.1161/STR.0b013e3181f cb238 Gonzlez, H.M., Tarraf, W., 2013. Comorbid cardiovascular disease and major depression among ethnic and racial groups in the United States. Int. Psychogeriatr. 25, 833 841. doi:10.1017/S1041610212002062 Gorber, S.C., Tremblay, M., Moher, D., Gorber, B., 2007. A comparison of direct vs. self report measures for assessing height, weight and body mass index: a systematic review. Obes. Rev. 8, 307 326. doi:10.1111/j.1467 789X.2007.00347.x Gordon, A.J., Conley, J.W., Gordon, J.M., 2013. Medical Consequence s of Marijuana Use: A Review of Current Literature. Curr. Psychiatry Rep. 15, 1 11. doi:10.1007/s11920 013 0419 7 Haberstick, B.C., Young, S.E., Zeiger, J.S., Lessem, J.M., Hewitt, J.K., Hopfer, C.J., 2014. Prevalence and correlates of alcohol and cannabis use disorders in the United States: Results from the national longitudinal study of adolescent health. Drug Alcohol Depend. 136, 158 161. doi:10.1016/j.drugalcdep.2013.11.022 Hagenaars, J.A., McCutcheon, A.L., 2002. Applied Latent Class Analysis. Cambridg e University Press, Cambridge, United Kingdom. Hall, W., Degenhardt, L., 2009. Adverse health effects of non medical cannabis use. The Lancet 374, 1383 1391. doi:10.1016/S0140 6736(09)61037 0 Hare, D.L., Toukhsati, S.R., Johansson, P., Jaarsma, T., 2013. D epression and cardiovascular disease: a clinical review. Eur. Heart J. eht462. doi:10.1093/eurheartj/eht462

PAGE 193

193 Hasin, D., Goodwin, Stinson, F., Grant, B., 2005. Epidemiology of major depressive disorder: Results from the national epidemiologic survey on alcoh olism and related conditions. Arch. Gen. Psychiatry 62, 1097 1106. doi:10.1001/archpsyc.62.10.1097 Hasin DS, Saha TD, Kerridge BT, et al, 2015. Prevalence of marijuana use disorders in the united states between 2001 2002 and 2012 2013. JAMA Psychiatry 72, 1235 1242. doi:10.1001/jamapsychiatry.2015.1858 J.M., 2010. Cannabis use and obesity and young adults. Am. J. Drug Alcohol Abuse 36, 350 356. doi:10.3109/00952990.201 0.500438 Hedden, S., Kennet, J., Lipari, R., Medley, G., Tice, P., 2015. Behavioral Health Trends in the United States: Results from the 2014 National Survey on Drug Use and Health. Substance Abuse and Mental Health Services Administration. Heidemann, C., Schulze, M.B., Franco, O.H., van Dam, R.M., Mantzoros, C.S., Hu, F.B., 2008. Dietary Patterns and Risk of Mortality from Cardiovascular Disease, Cancer, and All Causes in a Prospective Cohort of Women. Circulation 118, 230 237. doi:10.1161/CIRCULATIONAHA.1 08.771881 Heidenreich, P.A., Trogdon, J.G., Khavjou, O.A., Butler, J., Dracup, K., Ezekowitz, M.D., Finkelstein, E.A., Hong, Y., Johnston, S.C., Khera, A., Lloyd Jones, D.M., Nelson, S.A., Nichol, G., Orenstein, D., Wilson, P.W.F., Woo, Y.J., 2011. Forecas ting the Future of Cardiovascular Disease in the United States A Policy Statement From the American Heart Association. Circulation 123, 933 944. doi:10.1161/CIR.0b013e31820a55f5 Henson, J.M., Reise, S.P., Kim, K.H., 2007. Detecting Mixtures From Structural Model Differences Using Latent Variable Mixture Modeling: A Comparison of Relative Model Fit Statistics. Struct. Equ. Model. Multidiscip. J. 14, 202 226. doi:10.1080/10705510709336744 Herkenham, M., Lynn, A.B., Johnson, M.R., Melvin, L.S., de Costa, B.R., Rice, K.C., 1991. Characterization and localization of cannabinoid receptors in rat brain: a quantitative in vitro autoradiographic study. J. Neurosci. Off. J. Soc. Neurosci. 11, 563 583. Heron, M., 2015. Deaths: Leading Causes for 2012. Natl. Vital Stat. Rep. 64. Higgins, J., Green, S. (Eds.), 2011. Cochrane Handbook for Systematic Reviews of Interventions, 5.1.0. ed. The Cochrane Collaboration. Hilbe, J.M., 2009. Logistic Regression Models. CRC Press. Hill, A.B., 1965. The Environment and Disease: Associ ation or Causation? Proc. R. Soc. Med. 58, 295 300. Hoffmann, D.E., Weber, E., 2010. Medical Marijuana and the Law. N. Engl. J. Med. 362, 1453 1457. doi:10.1056/NEJMp1000695

PAGE 194

194 Hollister, L.E., 1986. Health aspects of cannabis. Pharmacol. Rev. 38, 1 20. Hosme r, D.W., Lemeshow, S., 2000. Applied Logistic Regression. John Wiley & Sons. Huang, D.Y.C., Lanza, H.I., Anglin, M.D., 2013a. Association between adolescent substance use and obesity in young adulthood: a group based dual trajectory analysis. Addict. Behav 38, 2653 2660. doi:10.1016/j.addbeh.2013.06.024 Huang, D.Y.C., Lanza, H.I., Wright Volel, K., Anglin, M.D., 2013b. Developmental trajectories of childhood obesity and risk behaviors in adolescence. J. Adolesc. 36, 139 148. doi:10.1016/j.adolescence.2012.10.005 Huxley, R.R., Woodward, M., 2011. Cigarette smoking as a risk factor for coronary heart disease in women compared with men: a systematic review and meta analysis of prospective cohort studies. The Lancet 378, 1297 1305 doi:10.1016/S0140 6736(11)60781 2 James PA, Oparil S, Carter BL, et al, 2014. 2014 evidence based guideline for the management of high blood pressure in adults: Report from the panel members appointed to the eighth joint national committee (jnc 8). JAMA 311, 507 520. doi:10.1001/jama.2013.284427 Johnson, S., Domino, E.F., 1971. Some cardiovascular effects of marihuana smoking in normal volunteers. Clin. Pharmacol. Ther. 12, 762 768. Johnson, T., Fendrich, M., 2005. Modeling Sources of Self report Bias in a Survey of Drug Use Epidemiology. Ann. Epidemiol. 15, 381 389. doi:10.1016/j.annepidem.2004.09.004 Future National Survey Results on Drug Use 1975 2015 Volume 2. Institute for Social Research, The University of Michigan, Ann Arbor. Future national survey results on drug use, 1975 2015: Overview, key findings on adolescent d rug use. Institute for Social Research, The University of Michigan, Ann Arbor. Jones, R.T., 2002. Cardiovascular System Effects of Marijuana. J. Clin. Pharmacol. 42, 58S 63S. doi:10.1002/j.1552 4604.2002.tb06004.x 2003. Depression and cardiovascular disease: mechanisms of interaction. Biol. Psychiatry 54, 248 261. doi:10.1016/S0006 3223(03)00568 7 Kandel, D., 2003. Does marijuana use cause the use of other drugs? JAMA 289, 482 483. doi:10.1001/jama.289.4.482 Kande l, D.B., Yamaguchi, K., Klein, L.C., 2006. Testing the Gateway Hypothesis. Addiction 101, 470 472. doi:10.1111/j.1360 0443.2006.01426.x

PAGE 195

195 Kardia, S.L.R., Modell, S.M., Peyser, P.A., 2003. Family centered approaches to understanding and preventing coronary he art disease. Am. J. Prev. Med. 24, 143 151. doi:10.1016/S0749 3797(02)00587 1 Kessler RC, Berglund P, Demler O, et al, 2003. The epidemiology of major depressive disorder: Results from the national comorbidity survey replication (ncs r). JAMA 289, 3095 310 5. doi:10.1001/jama.289.23.3095 Kessler, T., Erdmann, J., Schunkert, H., 2013. Genetics of Coronary Artery Disease and Myocardial Infarction 2013. Curr. Cardiol. Rep. 15, 368. doi:10.1007/s11886 013 0368 0 Khot UN, Khot MB, Bajzer CT, et al, 2003. PReval ence of conventional risk factors in patients with coronary heart disease. JAMA 290, 898 904. doi:10.1001/jama.290.7.898 Kilmer, J.R., Hunt, S.B., Lee, C.M., Neighbors, C., 2007. Marijuana use, risk perception, and consequences: Is perceived risk congruent with reality? Addict. Behav. 32, 3026 3033. doi:10.1016/j.addbeh.2007.07.009 Kirk, E.P., Klein, S., 2009. Pathogenesis and Pathophysiology of the Cardiometabolic Syndrome. J. Clin. Hypertens. Greenwich Conn 11, 761 765. doi:10.1111/j.1559 4572.2009.00054. x Labarthe, D.R., 2011. Epidemiology and Prevention of Cardiovascular Diseases, 2nd ed. Jones and Bartlett, Sudbury, Massachusetts. Lange, R.A., Hillis, L.D., 2001. Cardiovascular Complications of Cocaine Use. N. Engl. J. Med. 345, 351 358. doi:10.1056/NEJ M200108023450507 LaValley, M.P., 2008. Logistic Regression. Circulation 117, 2395 2399. doi:10.1161/CIRCULATIONAHA.106.682658 Leatherdale, S.T., Ahmed, R., 2010. Alcohol, Marijuana, and Tobacco Use Among Canadian Youth: Do We Need More Multi Substance Prev ention Programming? J. Prim. Prev. 31, 99 108. Ledgerwood, D.M., Goldberger, B.A., Risk, N.K., Lewis, C.E., Kato Price, R., 2008. Comparison between self report and hair analysis of illicit drug use in a community sample of middle aged men. Addict. Behav. 33, 1131 1139. doi:10.1016/j.addbeh.2008.04.009 Leening, M.J.G., Ferket, B.S., Steyerberg, E.W., Kavousi, M., Deckers, J.W., Nieboer, D., Heeringa, J., Portegies, M.L.P., Hofman, A., Ikram, M.A., Hunink, M.G.M., Franco, O.H., Stricker, B.H., Witteman, J.C. M., Roos Hesselink, J.W., 2014. Sex differences in lifetime risk and first manifestation of cardiovascular disease: prospective population based cohort study. BMJ 349, g5992. doi:10.1136/bmj.g5992 Lemstra, M., Bennett, N.R., Neudorf, C., Kunst, A., Nannapa neni, U., Warren, L.M., Kershaw, T., Scott, C.R., 2008. A Meta analysis of Marijuana and Alcohol Use by Socio economic

PAGE 196

196 Status in Adolescents Aged 10 15 Years. Can. J. Public Health Rev. Can. Santee Publique 99, 172 177. Lett, H.S., Blumenthal, J.A., Babyak M.A., Sherwood, A., Strauman, T., Robins, C., Newman, M.F., 2004. Depression as a risk factor for coronary artery disease: evidence, mechanisms, and treatment. Psychosom. Med. 66, 305 315. Lev Ran, S., Roerecke, M., Le Foll, B., George, T., McKenzie, K., Rehm, J., 2013. 892 The association between cannabis use and depression: a systematic review and meta analysis of longitudinal studies. Eur. Psychiatry, Abstracts of the 21th European Congress of Psychiatry 28, Supplement 1, 1. doi:10.1016/S0924 9338(13 )76057 7 Lichtman, J.H., Froelicher, E.S., Blumenthal, J.A., Carney, R.M., Doering, L.V., Frasure Smith, N., Freedland, K.E., Jaffe, A.S., Leifheit Limson, E.C., Sheps, D.S., Vaccarino, V., Wulsin, L., 2014. Depression as a Risk Factor for Poor Prognosis A mong Patients With Acute Coronary Syndrome: Systematic Review and Recommendations. Circulation 129, 1350 1369. doi:10.1161/CIR.0000000000000019 Lim, S.S., Vos, T., Flaxman, A.D., Danaei, G., Shibuya, K., Adair Rohani, H., AlMazroa, M.A., Amann, M., Anderso n, H.R., Andrews, K.G., Aryee, M., Atkinson, C., Bacchus, L.J., Bahalim, A.N., Balakrishnan, K., Balmes, J., Barker Collo, S., Baxter, A., Bell, M.L., Blore, J.D., Blyth, F., Bonner, C., Borges, G., Bourne, R., Boussinesq, M., Brauer, M., Brooks, P., Bruce N.G., Brunekreef, B., Bryan Hancock, C., Bucello, C., Buchbinder, R., Bull, F., Burnett, R.T., Byers, T.E., Calabria, B., Carapetis, J., Carnahan, E., Chafe, Z., Charlson, F., Chen, H., Chen, J.S., Cheng, A.T. A., Child, J.C., Cohen, A., Colson, K.E., Co wie, B.C., Darby, S., Darling, S., Davis, A., Degenhardt, L., Dentener, F., Des Jarlais, D.C., Devries, K., Dherani, M., Ding, E.L., Dorsey, E.R., Driscoll, T., Edmond, K., Ali, S.E., Engell, R.E., Erwin, P.J., Fahimi, S., Falder, G., Farzadfar, F., Ferrar i, A., Finucane, M.M., Flaxman, S., Fowkes, F.G.R., Freedman, G., Freeman, M.K., Gakidou, E., Ghosh, S., Giovannucci, E., Gmel, G., Graham, K., Grainger, R., Grant, B., Gunnell, D., Gutierrez, H.R., Hall, W., Hoek, H.W., Hogan, A., Hosgood III, H.D., Hoy, D., Hu, H., Hubbell, B.J., Hutchings, S.J., Ibeanusi, S.E., Jacklyn, G.L., Jasrasaria, R., Jonas, J.B., Kan, H., Kanis, J.A., Kassebaum, N., Kawakami, N., Khang, Y. H., Khatibzadeh, S., Khoo, J. P., Kok, C., Laden, F., Lalloo, R., Lan, Q., Lathlean, T., Le asher, J.L., Leigh, J., Li, Y., Lin, J.K., Lipshultz, S.E., London, S., Lozano, R., Lu, Y., Mak, J., Malekzadeh, R., Mallinger, L., Marcenes, W., March, L., Marks, R., Martin, R., McGale, P., McGrath, J., Mehta, S., Memish, Z.A., Mensah, G.A., Merriman, T. R., Micha, R., Michaud, C., Mishra, V., Hanafiah, K.M., Mokdad, A.A., Morawska, L., Mozaffarian, D., Murphy, T., Naghavi, M., Neal, B., Nelson, P.K., Nolla, J.M., Norman, R., Olives, C., Omer, S.B., Orchard, J., Osborne, R., Ostro, B., Page, A., Pandey, K. D., Parry, C.D., Passmore, E., Patra, J., Pearce, N., Pelizzari, P.M., Petzold, M., Phillips, M.R., Pope, D., Pope III, C.A., Powles, J., Rao, M., Razavi, H., Rehfuess, E.A., Rehm, J.T., Ritz, B., Rivara, F.P., Roberts, T., Robinson, C., Rodriguez Portales J.A., Romieu, I., Room, R., Rosenfeld, L.C., Roy, A., Rushton, L., Salomon, J.A., Sampson, U., Sanchez Riera, L., Sanman, E., Sapkota, A., Seedat, S., Shi, P., Shield, K., Shivakoti, R., Singh, G.M., Sleet, D.A., Smith, E., Smith, K.R., Stapelberg, N.J., Steenland, K., Stckl, H., Stovner, L.J., Straif, K., Straney, L., Thurston, G.D., Tran, J.H., Van Dingenen, R., van Donkelaar, A.,

PAGE 197

197 Veerman, J.L., Vijayakumar, L., Weintraub, R., Weissman, M.M., White, R.A., Whiteford, H., Wiersma, S.T., Wilkinson, J.D., Williams, H.C., Williams, W., Wilson, N., Woolf, A.D., Yip, P., Zielinski, J.M., Lopez, A.D., Murray, C.J., Ezzati, M., 2012. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regio ns, 1990 2010: a systematic analysis for the Global Burden of Disease Study 2010. The Lancet 380, 2224 2260. doi:10.1016/S0140 6736(12)61766 8 Lloyd Jones, D.M., Hong, Y., Labarthe, D., Mozaffarian, D., Appel, L.J., Horn, L.V., Greenlund, K., Daniels, S., Nichol, G., Tomaselli, G.F., Arnett, D.K., Fonarow, G.C., Ho, P.M., Lauer, M.S., Masoudi, F.A., Robertson, R.M., Roger, V., Schwamm, L.H., Sorlie, P., Yancy, C.W., Rosamond, W.D., 2010. Defining and Setting National Goals for Cardiovascular Health Promotio n and Disease Reduction. Circulation 121, 586 613. doi:10.1161/CIRCULATIONAHA.109.192703 Lo, Y., Mendell, N.R., Rubin, D.B., 2001. Testing the number of components in a normal mixture. Biometrika 88, 767 778. doi:10.1093/biomet/88.3.767 Lynskey, M.T., Agrawal, A., Bucholz, K.K., Nelson, E.C., Madden, P.A.F., Todorov, A.A., Grant, J.D., Martin, N.G., Heath, A.C., 2006. Subtypes of Illicit Drug Users: A Latent Class Analysis of Data From an Australian Twin Sample. Twin Res. Hum. Genet. 9, 523 530. doi:10. 1375/183242706778024964 Maas, A.H.E.M., Appelman, Y.E.A., 2010. Gender differences in coronary heart disease. Neth. Heart J. 18, 598 603. doi:10.1007/s12471 010 0841 y MacCoun, R., 2017. How Medical Marijuana Smoothed the Transition to Marijuana Legalizati on in the United States. Annu. Rev. Law Soc. Sci. 13, null. doi:10.1146/annurev lawsocsci 110615 084851 Mailleux, P., Parmentier, M., Vanderhaeghen, J.J., 1992. Distribution of cannabinoid receptor messenger RNA in the human brain: an in situ hybridization histochemistry with oligonucleotides. Neurosci. Lett. 143, 200 204. Manrique Garcia, E., Zammit, S., Dalman, C., Hemmingsson, T., Allebeck, P., 2012. Cannabis use and depression: a longitudinal study of a national cohort of Swedish conscripts. BMC Psychia try 12. doi:10.1186/1471 244X 12 112 Marconi, A., Di Forti, M., Lewis, C.M., Murray, R.M., Vassos, E., 2016. Meta analysis of the Association Between the Level of Cannabis Use and Risk of Psychosis. Schizophr. Bull. 42, 1262 1269. doi:10.1093/schbul/sbw003 Marshall, B.D.L., Galea, S., 2015. Formalizing the Role of Agent Based Modeling in Causal Inference and Epidemiology. Am. J. Epidemiol. 181, 92 99. doi:10.1093/aje/kwu274 McBride, P., 1992. The health consequences of smoking. Cardiovascular diseases. Med. Clin. North Am. 76, 333 353.

PAGE 198

198 McFadden, D., 1979. Quantitative Methods For Analysing Travel Behavior of Individuals: Some Recent Developments, in: Hensher, D.A., Stopher, P.R. (Eds.), Behavioral Travel Modeling. Groom Helm, London, pp. 279 318. McFadden, D ., 1974. Conditional logit analysis of qualitative choice behavior, in: Frontiers in Econometrics. Academic Press, Berkley, California, pp. 105 142. McGrath J, Welham J, Scott J, et al, 2010. Association between cannabis use and psychosis related outcomes using sibling pair analysis in a cohort of young adults. Arch. Gen. Psychiatry 67, 440 447. doi:10.1001/archgenpsychiatry.2010.6 Mennes, C.E., Ben Abdallah, A., Cottler, L.B., 2009. The reliability of self reported cannabis abuse, dependence and withdrawal symptoms: Multisite study of differences between general population and treatment groups. Addict. Behav. 34, 223 226. doi:10.1016/j.addbeh.2008.10.003 Mittleman, M.A., Lewis, R.A., Maclure, M., Sherwood, J.B., Muller, J.E., 2001. Triggering Myocardial Inf arction by Marijuana. Circulation 103, 2805 2809. doi:10.1161/01.CIR.103.23.2805 Mittleman, M.A., Lewis, R.A., Maclure, M., Sherwood, J.B., Muller, J.E., 2000. Triggering of myocardial infarction by Marijuana. Circulation 101, 713 713. Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G., 2009. Preferred Reporting Items for Systematic Reviews and Meta Analyses: The PRISMA Statement. J. Clin. Epidemiol. 62, 1006 1012. doi:10.1016/j.jclinepi.2009.06.005 Monga, N., Rehm, J., Fischer, B., Brissette, S., Bruneau, J., El Guebaly, N., Nol, L., Tyndall, M., Wild, C., Leri, F., Fallu, J. S., Bahl, S., 2007. Using latent class analysis (LCA) to analyze patterns of drug use in a population of illegal opioid users. Drug Alcohol Depend. 88, 1 8. doi:10.1016/j.drugalcdep. 2006.08.029 Montecucco, F., Di Marzo, V., 2012. At the heart of the matter: the endocannabinoid system in cardiovascular function and dysfunction. Trends Pharmacol. Sci. 33, 331 340. doi:10.1016/j.tips.2012.03.002 Mozaffarian, D., Benjamin, E.J., Go, A.S., Arnett, D.K., Blaha, M.J., Cushman, M., Das, S.R., Ferranti, S. de, Desprs, J. P., Fullerton, H.J., Howard, V.J., Huffman, M.D., Isasi, C.R., Jimnez, M.C., Judd, S.E., Kissela, B.M., Lichtman, J.H., Lisabeth, L.D., Liu, S., Mackey, R.H., Magid, D.J., Mc Guire, D.K., Mohler, E.R., Moy, C.S., Muntner, P., Mussolino, M.E., Nasir, K., Neumar, R.W., Nichol, G., Palaniappan, L., Pandey, D.K., Reeves, M.J., Rodriguez, C.J., Rosamond, W., Sorlie, P.D., Stein, J., Towfighi, A., Turan, T.N., Virani, S.S., Woo, D., Yeh, R.W., Turner, M.B., 2015. Heart Disease and Stroke Statistics 2016 Update. Circulation CIR.0000000000000350. doi:10.1161/CIR.0000000000000350 Mulrow, C.D., 1987. The medical review article: state of the science. Ann. Intern. Med. 106, 485 488.

PAGE 199

199 Murray, R.M., Forti, M.D., 2016. Cannabis and Psychosis: What Degree of Proof Do We Require? Biol. Psychiatry 79, 514 515. doi:10.1016/j.biopsych.2016.02.005 Musselman DL, Evans DL, Nemeroff CB, 1998. The relationship of depression to cardiovascular disease: Epid emiology, biology, and treatment. Arch. Gen. Psychiatry 55, 580 592. doi:10.1001/archpsyc.55.7.580 Muthen, B., Muthen, L.K., 2000. Integrating Person Centered and Variable Centered Analyses: Growth Mixture Modeling With Latent Trajectory Classes 24, 882 89 1. Angeles, CA. Naimi, A.I., Paquet, C., Gauvin, L., Daniel, M., 2009. Associations between Area Level Unemployment, Body Mass Index, and Risk Factors for Cardiovascu lar Disease in an Urban Area. Int. J. Environ. Res. Public. Health 6, 3082 3096. doi:10.3390/ijerph6123082 Natarajan, S., Lipsitz, S.R., Nietert, P.J., 2002. Self report of high cholesterol: determinants of validity in U.S. adults. Am. J. Prev. Med. 23, 13 21. doi:10.1016/S0749 3797(02)00446 4 National Academies of Sciences, Engineering, and Medicine, 2017. The Health Effects of Cannabis and Cannabinoids: The Current State of Evidence and Recommendations for Research. doi:10.17226/24625 National Center for Health Statistics, 2014. Summary Health Statistics: National Health Interview Survey. National Heart Lung and Blood Institute, 2016. How Does Smoking Affect the Heart and Blood Vessels? [WWW Document]. Natl. Heart Lung Blood Inst. URL https://www.nhlbi.nih .gov/health/health topics/topics/smo (accessed 1.26.17). National Institute on Mental Health, 2016. Major Depression Among Adults [WWW Document]. URL https://www.nimh.nih.gov/health/statistics/prevalence/major depression among adults.shtml (accessed 3.24.17). Nissen SE, Nicholls SJ, Wolski K, et al, 2008. Effect of rimonabant on progression of atherosclerosis in patients with abdominal obesity and coronary artery disease: The stradivarius randomized controlled trial. JAMA 299, 1547 1560. doi:10.1001/j ama.299.13.1547 Nylund, K., Asparouhov, T., Muthen, B., 2007. Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study. Struct. Equ. Model. 14, 535 569.

PAGE 200

200 Ogden, D., 2009. Memorandum for Selected United State Attorneys: Investigations and Prosecutions in States Authorizing Medical Use of Marijuana. U.S. Department of Justice, Office of the Deputy Attorney General. Okura, Y., Urban, L.H., Mahoney, D.W., Jacobsen, S.J., Rodeheffer, R.J., 2004. Agreem ent between self report questionnaires and medical record data was substantial for diabetes, hypertension, myocardial infarction and stroke but not for heart failure. J. Clin. Epidemiol. 57, 1096 1103. doi:10.1016/j.jclinepi.2004.04.005 Osei Hyiaman, D., D ePetrillo, M., Pacher, P., Liu, J., Radaeva, S., Btkai, S., Harvey White, J., Mackie, K., Offertler, L., Wang, L., Kunos, G., 2005. Endocannabinoid activation at hepatic CB1 receptors stimulates fatty acid synthesis and contributes to diet induced obesit y. J. Clin. Invest. 115, 1298 1305. doi:10.1172/JCI23057 Osler, M., Heitmann, B.L., Gerdes, L.U., Jrgensen, L.M., Schroll, M., 2001. Dietary patterns and mortality in Danish men and women: a prospective observational study. Br. J. Nutr. 85, 219 225. Pacek L.R., Mauro, P.M., Martins, S.S., 2015. Perceived risk of regular cannabis use in the United States from 2002 to 2012: Differences by sex, age, and race/ethnicity. Drug Alcohol Depend. 149, 232 244. doi:10.1016/j.drugalcdep.2015.02.009 Pacher, P., Batkai S., Kunos, G., 2006. The Endocannabinoid System as an Emerging Target of Pharmacotherapy. Pharmacol. Rev. 58, 389 462. doi:10.1124/pr.58.3.2 Pagotto, U., Marsicano, G., Cota, D., Lutz, B., Pasquali, R., 2006. The Emerging Role of the Endocannabinoid Syst em in Endocrine Regulation and Energy Balance. Endocr. Rev. 27, 73 100. doi:10.1210/er.2005 0009 Patton, G.C., Coffey, C., Carlin, J.B., Degenhardt, L., Lynskey, M., Hall, W., 2002. Cannabis use and mental health in young people: cohort study. BMJ 325, 119 5 1198. doi:10.1136/bmj.325.7374.1195 Pearson, T.A., Jamison, D.T., Trejo Gutierrez, J., 1993. Cardiovascular disease, in: Disease Control Priorities in Developing Countries. Oxford University Press, New York, pp. 577 594. Penner, E.A., Buettner, H., Mittl eman, M.A., 2013. The Impact of Marijuana Use on Glucose, Insulin, and Insulin Resistance among US Adults. Am. J. Med. 126, 583 589. doi:10.1016/j.amjmed.2013.03.002 Pi Sunyer F, Aronne LJ, Heshmati HM, Devin J, Rosenstock J, RIO North America Study Group f, 2006. Effect of rimonabant, a cannabinoid 1 receptor blocker, on weight and cardiometabolic risk factors in overweight or obese patients: Rio north america: a randomized controlled trial. JAMA 295, 761 775. doi:10.1001/jama.295.7.761

PAGE 201

201 Powers, A.D., Oltma nns, T.F., 2013. Borderline personality pathology and chronic health problems in later adulthood: The mediating role of obesity. Personal. Disord. Theory Res. Treat. 4, 152 159. doi:10.1037/a0028709 Racine, C., Vincent, M., Rogers, A., Donat, M., Ojike, N. I., Necola, O., Yousef, E., Masters Israilov, A., Jean Louis, G., McFarlane, S.I., 2015. Metabolic effects of marijuana use among Blacks. J. Dis. Glob. Health 4, 9 16. Rajavashisth, T.B., Shaheen, M., Norris, K.C., Pan, D., Sinha, S.K., Ortega, J., Friedma n, T.C., 2012. Decreased prevalence of diabetes in marijuana users: cross sectional data from the National Health and Nutrition Examination Survey (NHANES) III. BMJ Open 2, e000494. doi:10.1136/bmjopen 2011 000494 Rajavashisth, T.B., Shaheen, M., Pan, D., Sinha, S.K., Ortega, J., Friedman, T.C., 2010. Decreased Prevalence of Diabetes in Marijuana Users: Data from NHANES III. Diabetes 59, A435 A435. Rajesh, M., Mukhopadhyay, P., Hask, G., Liaudet, L., Mackie, K., Pacher, P., 2010. Cannabinoid 1 receptor act ivation induces reactive oxygen species dependent and independent mitogen activated protein kinase activation and cell death in human coronary artery endothelial cells. Br. J. Pharmacol. 160, 688 700. doi:10.1111/j.1476 5381.2010.00712.x Ramaswamy, V., De sarbo, W.S., Reibstein, D.J., Robinson, W.T., 1993. An Empirical Pooling Approach for Estimating Marketing Mix Elasticities with PIMS Data. Mark. Sci. 12, 103 124. doi:10.1287/mksc.12.1.103 Redonnet, B., Chollet, A., Fombonne, E., Bowes, L., Melchior, M., 2012. Tobacco, alcohol, cannabis and other illegal drug use among young adults: The socioeconomic context. Drug Alcohol Depend. 121, 231 239. doi:10.1016/j.drugalcdep.2011.09.002 Roberts, R., 2014. Genetics of Coronary Artery Disease. Circ. Res. 114, 1890 1903. doi:10.1161/CIRCRESAHA.114.302692 Rodondi, N., Pletcher, M.J., Liu, K., Hulley, S.B., Sidney, S., 2006. Marijuana Use, Diet, Body Mass Index, and Cardiovascular Risk Factors (from the CARDIA Study). Am. J. Cardiol. 98, 478 484. doi:10.1016/j.amjcard. 2006.03.024 University Press. Rosellini, A.J., Coffey, S.F., Tracy, M., Galea, S., 2014. A person centered analysis of posttraumatic stress disorder symptoms follo wing a natural disaster: Predictors of latent class membership. J. Anxiety Disord. 28, 16 24. doi:10.1016/j.janxdis.2013.11.002 Ross, J.M., Graziano, P., Pacheco Coln, I., Coxe, S., Gonzalez, R., 2016. Decision Making Does not Moderate the Association bet ween Cannabis Use and Body Mass Index among

PAGE 202

202 Adolescent Cannabis Users. J. Int. Neuropsychol. Soc. 22, 944 949. doi:10.1017/S1355617716000278 Roth, W.T., Tinkleinberg, J.R., Kopell, B.S., Hollister, L.E., 1973. Continuous electrocardiographic monitoring dur ing marihuana intoxication. Clin. Pharmacol. Ther. 14, 533 540. Rothman, K.J., 1986. Modern Epidemiology, 1st ed. Little, Brown, Boston. Ryan, C.L., Bauman, K., 2015. Educational Attainment in the United States: 2015 (No. P20 578). U.S. Census Bureau. Sacks, H.S., Berrier, J., Reitman, D., Ancona Berk, V.A., Chalmers, T.C., 1987. Meta analyses of randomized controlled trials. N. Engl. J. Med. 316, 450 455. doi:10.1056/NEJM198702193160806 Schauer, G.L., Rosenberry, Z.R., Peters, E.N., 2017. Marijuana and tobacco co administration in blunts, spliffs, and mulled cigarettes: A systematic literature review. Addict. Behav. 64, 200 211. doi:10.1016/j.addbeh.2016.09.001 Scheen, A.J., Finer, N., Hollander, P., Jensen, M.D., Van Gaal, L.F., RIO Diabetes Study Grou p, 2006. Efficacy and tolerability of rimonabant in overweight or obese patients with type 2 diabetes: a randomised controlled study. Lancet Lond. Engl. 368, 1660 1672. doi:10.1016/S0140 6736(06)69571 8 Schuermeyer, J., Salomonsen Sautel, S., Price, R.K., Balan, S., Thurstone, C., Min, S. J., Sakai, J.T., 2014. Temporal trends in marijuana attitudes, availability and use in Colorado compared to non medical marijuana states: 2003 11. Drug Alcohol Depend. 140, 145 155. doi:10.1016/j.drugalcdep.2014.04.016 Sch wartz, B.G., Rezkalla, S., Kloner, R.A., 2010. Cardiovascular Effects of Cocaine. Circulation 122, 2558 2569. doi:10.1161/CIRCULATIONAHA.110.940569 Sclove, S.L., 1987. Application of model selection criteria to some problems in multivariate analysis. Psych ometrika 52, 333 343. doi:10.1007/BF02294360 Shillington, A.M., Cottler, L.B., Mager, D.E., Compton III, W.M., 1995. Self report stability for substance use over 10 years: data from the St. Louis Epidemiologic Catchment Study. Drug Alcohol Depend. 40, 103 109. doi:10.1016/0376 8716(95)01176 5 Sidney, S., 2002. Cardiovascular Consequences of Marijuana Use. J. Clin. Pharmacol. 42, 64S 70S. doi:10.1002/j.1552 4604.2002.tb06005.x Sidney, S., Beck, J.E., Tekawa, I.S., Quesenberry, C.P., Friedman, G.D., 1997. Mar ijuana use and mortality. Am. J. Public Health 87, 585 590. Simpson, C.F., Boyd, C.M., Carlson, M.C., Griswold, M.E., Guralnik, J.M., Fried, L.P., 2004. Agreement Between Self Report of Disease Diagnoses and Medical Record Validation in

PAGE 203

203 Disabled Older Wome n: Factors That Modify Agreement. J. Am. Geriatr. Soc. 52, 123 127. doi:10.1111/j.1532 5415.2004.52021.x Smit, E., Crespo, C.J., 2001. Dietary intake and nutritional status of US adult marijuana users: results from the Third National Health and Nutrition E xamination Survey. Public Health Nutr. 4, 781 786. doi:10.1079/PHN2000114 Sugamura, K., Sugiyama, S., Nozaki, T., Matsuzawa, Y., Izumiya, Y., Miyata, K., Nakayama, M., Kaikita, K., Obata, T., Takeya, M., Ogawa, H., 2009. Activated Endocannabinoid System in Coronary Artery Disease and Antiinflammatory Effects of Cannabinoid 1 Receptor Blockade on Macrophages. Circulation 119, 28 36. doi:10.1161/CIRCULATIONAHA.108.811992 Swendsen, J.D., Merikangas, K.R., 2000. The comorbidity of depression and substance use d isorders. Clin. Psychol. Rev. 20, 173 189. doi:10.1016/S0272 7358(99)00026 4 Swift, A., 2016. Support for Legal Marijuana Use Up to 60% in U.S. [WWW Document]. Gallup.com. URL http://www.gallup.com/poll/196550/support legal marijuana.aspx (accessed 1.21.17 ). Thomas, G., Kloner, R.A., Rezkalla, S., 2014. Adverse Cardiovascular, Cerebrovascular, and Peripheral Vascular Effects of Marijuana Inhalation: What Cardiologists Need to Know. Am. J. Cardiol. 113, 187 190. doi:10.1016/j.amjcard.2013.09.042 Thompson, C. Hay, J., 2015a. Estimating the association between metabolic risk factors and marijuana use in U.S. adults using data from the continuous National Health and Nutrition Examination Survey. Ann. Epidemiol. 25, 486 491. doi:10.1016/j.annepidem.2015.01.013 T hompson, C., Hay, J., 2015b. Marijuana use in models for health outcomes. Am. J. Med. 128, e23. doi:10.1016/j.amjmed.2014.07.039 Thompson, P.L., 2013. J curve revisited: cardiovascular benefits of moderate alcohol use cannot be dismissed. Med. J. Aust. 198 Thurston, R.C., Rewak, M., Kubzansky, L.D., 2013. An Anxious Heart: Anxiety and the Onset of Cardiovascular Diseases. Prog. Cardiovasc. Dis., Symposium on Psychosocial Factors in Cardiovascular Disease 55, 524 537. doi:10.1016/j.pcad.2013.03.007 Thylstru p, B., Clausen, T., Hesse, M., 2015. Cardiovascular disease among people with drug use disorders. Int. J. Public Health 60, 659 668. doi:10.1007/s00038 015 0698 3 Toguri, J.T., Lehmann, C., Laprairie, R.B., Szczesniak, A.M., Zhou, J., Denovan Wright, E.M., Kelly, M.E.M., 2014. Anti inflammatory effects of cannabinoid CB2 receptor activation in endotoxin induced uveitis. Br. J. Pharmacol. 171, 1448 1461. doi:10.1111/bph.12545

PAGE 204

204 U.S Census Bureau, 2015. ACS Demographic and Housing Estimates 2010 2014 American Community Survey 5 Year Estimates. US Department of Health and Human Services, 1996. Physical activity and health: a report of the Surgeon General. US Department of Health and Human Se rvices, Public Health Service, CDC, National Center for Chronic Disease Prevention and Health Promotion, Atlanta, Georgia. Van Gaal, L.F., Rissanen, A.M., Scheen, A.J., Ziegler, O., Rssner, S., 2005. Effects of the cannabinoid 1 receptor blocker rimonaban t on weight reduction and cardiovascular risk factors in overweight patients: 1 year experience from the RIO Europe study. The Lancet 365, 1389 1397. doi:10.1016/S0140 6736(05)66374 X Vidot, D.C., Prado, G., Hlaing, W.M., Arheart, K.L., Messiah, S.E., 2014 Emerging issues for Addict. Dis. 33, 1 8. doi:10.1080/10550887.2014.882718 Vidot, D.C., Prado, G., Hlaing, W.M., Florez, H.J., Arheart, K.L., Messiah, S.E., 2016. Metabolic Syndrome Among Marijuana Users in the United States: An Analysis of National Health and Nutrition Examination Survey Data. Am. J. Med. 129, 173 179. doi:10.1016/j.amjmed.2015.10.019 Vuong, Q.H., 1989. Likelihood Ratio Tests for Model Selection a nd Non nested Hypotheses. Econometrica 57, 307 33. Walker, E.R., Druss, B.G., 2016. Cumulative burden of comorbid mental disorders, substance use disorders, chronic medical conditions, and poverty on health among adults in the U.S.A. Psychol. Health Med. 0 1 9. doi:10.1080/13548506.2016.1227855 Walker, W.J., 1969. The National Diet Heart Study Final Report: American Heart Association Monograph. No. 19. Arch. Intern. Med. 123, 473 474. doi:10.1001/archinte.1969.00300140119031 Wang, J.J., Attia, J., 2010. Study Designs in Epidemiology and Levels of Evidence. Am. J. Ophthalmol. 149, 367 370. doi:10.1016/j.ajo.2009.08.001 Waterreus, A., Di Prinzio, P., Watts, G.F., Castle, D., Galletly, C., Morgan, V.A., 2016. Metabolic syndrome in people with a psychotic ill ness: is cannabis protective? Psychol. Med. 46, 1651 1662. doi:10.1017/S0033291715002883 World Health Organization, 2017. Cannabis [WWW Document]. WHO. URL http://www.who.int/substance_abuse/facts/cannabis/en/ (accessed 2.18.17). World Health Organization, 2000. Obesity: Preventing and managing the global epidemic (No. 894), WHO Technical Report Series. Geneva. Xian, H., Scherrer, J.F., Franz, C.E., McCaffery, J., Stein, P.K., Lyons, M.J., Jacobsen, K., Eisen, S.A., Kremen, W.S., 2010. Genetic Vulnerability and Phenotypic Expression of

PAGE 205

205 Depression and Risk for Ischemic Heart Disease in the Vietnam Era Twin Study of Aging. Psychosom. Med. 72, 370 375. doi:10.1097/PSY.0b013e3181d28125 Yankey, B.N.A., Strasser, S., Okosun, I.S., 2016. A cross sectional analysis of the association between marijuana and cigarette smoking with metabolic syndrome among adults in the United States. Diabetes Metab. Syndr. Clin. Res. Rev. doi:10.1016/j.dsx.2016.03.001 Yeager, D.S., Krosnick, J.A., 2010. The Validity of Self Reported Nic otine Product Use in the 2001 2008 National Health and Nutrition Examination Survey: Med. Care 48, 1128 1132. doi:10.1097/MLR.0b013e3181ef9948

PAGE 206

206 BIOGRAPHICAL SKETCH Hannah Renee Crooke was born in Olathe, Kansas to parents Steven and Sabrina Crooke. Han nah graduated from Texas Christian University in My, 2011 with a Bachelor of Science in biology and a minor in Spanish. The following summer she completed a certificate in business from the Neeley School of Business at Texas Christian University. She recei ved her Master of Public Health with an emp hasis in e pidemiology from Drexel University in June, 2013. While economic status and burn rela Hannah began her academic training at the University of Florida in fall 2013. There she worked as a graduate research assistant for Dr. Catherine W Striley before becoming a Pre Doctoral Fellow funded by the National In stitute of Drug Abuse through the University of Florida Substance Abuse Training Center in Public Health (PI Cottler). While at the University of Florida, Hannah was involved both within the university and within national organizations, serving as the stud ent representative to the Department of Epidemiology Faculty C ommittee, and as the Vice President of the Public Health and Health Professions Doctoral Student C ouncil In 2015, she was chosen as the Associate Member L iaison to the Education C ommittee for the American College of Epidemiology. Hannah completed her Doctor of Philosophy in e pidemiology in May 2017