1 GENETIC, PHARMACOGENETIC, AND PHARMACOTHERAPEUTIC RISK FACTORS FOR THIAZIDE INDUCED DYSGLYCEMIA By JASON HANSEN KARNES 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 2012
2 2012 Jason Hansen Karnes
3 To my mother and father
4 ACKNOWLEDGMENTS I would like to express my sincere st gratitude to Dr. Rhonda Cooper DeHoff and Dr. Julie Jo hnson for their mentorship and support throughout the past four years f or the incredible training experience they have given me and for the future opportunities this has provided I would also like to thank Dr. Yan Gong, Dr. Lauren McIntyre, Dr. Marian Limacher and Dr. Taimour Langaee for their help, encouragement, and expertise. I would like to thank the present and former graduate students and postdocs in the department of Pharmacotherapy and Translational Research I would not have survived the grad uate experience without their mentorship, compassion, and friendship. A special thanks to Dr. Martin Brunner, Benjamin Burkley Cheryl Galloway, Dr. Maximilian Lobmeyer, Dr. Caitrin McDonough, Dr. Mike Pacanowski and Lynda Stauffer, who facilitated so muc h of the research included in this dissertation. I would also like to thank all the summer students who have contributed so much to th e completion of this research. I would like to thank the UF C linical and Translational Sciences Institute for their suppo rt facilitation of our original clinical study and their role in PEAR and PEAR 2. I would also like to thank the J Craig Venter Institute Resequencing and Genotyping Service particular ly Sam uel Levy Tim Stockwell, and Ewen Kirkness, for their large pa rt in the sequencing of TCF7L2 Finally, I would like to thank my family, Mom, Dad, and Emily, who have supported me constantly, and my friends, who have occasionally provided a welcome distraction from pharmaceutical research. Far more individuals than I can list here deserve gratitude for supporting me and making this research and training possible
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 11 LIST OF ABBREVIATIONS ................................ ................................ ........................... 13 ABSTRACT ................................ ................................ ................................ ................... 17 CHAPTER 1 INTRODUCTION AND BACKGROUND ................................ ................................ 19 Hypertension and Type 2 Diabetes ................................ ................................ ......... 19 Blood Pressure Reduction with Thiazide Diuretics ................................ ................. 20 Thiazide Induced Dysglycemia ................................ ................................ ............... 21 Mechanisms of Thiazide Induced Dysglyc emia ................................ ...................... 23 Genetics of Type 2 Diabetes ................................ ................................ ................... 24 The Transcription Factor 7 Like 2 Gene ( TCF7L2 ) ................................ ................. 26 Pharmacogenetics of Thiazide Induced Dysglycemia ................................ ............. 27 Short and Long Term Thiazide Induced Dysglycemia ................................ ............ 29 2 SEQUENCING, DETERMINATION OF LINKAGE DISEQUILIBRIUM STRUCTURE, AND ASSOCIATION ANALYSIS IN TCF7L2 ................................ .. 37 Introduction ................................ ................................ ................................ ............. 37 Methodo logy ................................ ................................ ................................ ........... 40 INVEST Study Design and Study Population ................................ ................... 40 INVEST GENES Study Design and Population ................................ ................ 41 TCF7L2 Sequencing ................................ ................................ ........................ 41 TCF7L2 Polymorphism Discovery and Visual Display ................................ ...... 42 In Silico Functional Pr ediction of TCF7L2 Polymorphisms ............................... 42 Race/Ethnicity and Linkage Disequilibrium Structure in Sequenced Samples 43 Identification o f T2D Predictor SNPs from Sequenced Samples ...................... 44 SNP Genotyping in the INVEST GENES New Onset Diabetes Case Control .. 46 Baseline Ch aracteristic and NOD Association Analysis in the INVEST GENES NOD Case Control Cohort ................................ ............................... 47 Pharmacogenetic Analysis in the INVEST GENES NOD Case Control Cohort ................................ ................................ ................................ ........... 48 Results ................................ ................................ ................................ .................... 49 Sequence and Genotype Data Quality Control in Sequenced Samples ........... 49 Characteristics o f Sequenced TCF7L2 Variation ................................ .............. 49 In Silico Functional Prediction of Sequenced TCF7L2 Variants ....................... 50
6 LD Structure of Sequenced TCF7L2 Variants ................................ .................. 50 Baseline Characteristics and PCA of Sequenced Samples .............................. 51 Validation of Candidate T2D Predictor SNP Selection in Whi te Sequenced Samples ................................ ................................ ................................ ........ 51 Candidate T2D Predictor SNP Identification in Hispanic Sequenced Samples ................................ ................................ ................................ ........ 52 Candidate T2D Predictor SNP Identification in Black Sequenced Samples ..... 52 Baseline Characteristics and PCA for the INVEST GENES NOD Case Control ................................ ................................ ................................ ........... 53 NOD Assoc iation in White INVEST GENES NOD Case Control Patients ........ 53 NOD Association in Hispanic INVEST GENES NOD Case Control Patients ... 54 NOD Association in Black INVEST GENES NOD Case Control Patients ........ 55 TCF7L2 SNP*HCTZ Treatment Pharmacogenetic Interactions ........................ 56 Dis cussion ................................ ................................ ................................ .............. 57 Summary and Significance ................................ ................................ ..................... 65 3 ASSOCIATION OF TAG SNPS IN KCNJ1, ADD1, ACE, AND AGTR1 WITH CHANGE IN FG AND NOD DURING THIAZIDE TREATMENT ............................. 80 Introduction ................................ ................................ ................................ ............. 80 Methods ................................ ................................ ................................ .................. 83 PEAR Study Des ign and Population ................................ ................................ 83 INVEST Study Design and Population ................................ ............................. 84 Genotyping and Quality Control ................................ ................................ ....... 85 Definition and Treatment of Race/Ethnicity ................................ ...................... 85 Statistical Analysis ................................ ................................ ............................ 86 PEAR ................................ ................................ ................................ ......... 86 INVEST ................................ ................................ ................................ ...... 87 Linkage Disequilibrium and Haplotype Design ................................ ................. 88 Results ................................ ................................ ................................ .................... 88 Baseline Characteristics and Clinical Predictor of Outcome Variables ............. 88 PEAR ................................ ................................ ................................ ......... 88 INVEST ................................ ................................ ................................ ...... 89 Hardy Weinberg Equilibrium for Candidate Gene SNPs ................................ ... 90 KCNJ1 and Increased FG during HCTZ Treatment in PEAR ........................... 90 KCNJ1 and NOD Risk after HCTZ Treatment in INVEST ................................ 92 ADD1 and Increased FG during HCTZ Treatment in PEAR ............................. 93 ADD1 NOD Risk after HCTZ Treatment in INVEST ................................ ......... 95 ACE and Increased FG during HCTZ Treatment in PEAR ............................... 96 ACE NOD Risk after HCTZ Treatment in INVEST ................................ ........... 97 AGTR1 and Increased FG during HCTZ Treatment in PEAR ........................... 97 AGTR1 NOD Risk after HCTZ Treatment in INVEST ................................ ....... 97 Discussion ................................ ................................ ................................ .............. 98 Summary and Significance ................................ ................................ ................... 106
7 4 LONG TERM ANTIHYPERTENSIVE EXPOSURE AND ADVERSE METABOLIC EFFECTS: PEAR FOLLOW UP STUDY ................................ ......... 119 Introduction ................................ ................................ ................................ ........... 119 Met hods ................................ ................................ ................................ ................ 122 PEAR and PEAR 2 Study Designs and Populations ................................ ...... 122 PEAR Follow Up Study Design and Population ................................ ............. 123 Statistical Analysis ................................ ................................ .......................... 125 Results ................................ ................................ ................................ .................. 126 PEAR Follow Up Study Population Characteristics a t Baseline versus Follow Up ................................ ................................ ................................ .... 126 Characteristics of PEAR Follow Up Study Population at Follow Up ............... 127 Change in FG during Short Te rm versus Long Term Thiazide Treatment ...... 128 Stepwise Linear Regression of Change in Lab Measures during Long Term Thiazide Treatment ................................ ................................ ..................... 129 Stepwise Linear Regression of Lab Measures at Follow Up Visit after Long Term Thiazide Treatment ................................ ................................ ............ 130 Correlation of Change FG and Change in Serum Potassium during Follow Up ................................ ................................ ................................ ............... 130 Evaluation of IFG, IGT, EGI, and T2D ................................ ............................ 130 Discussion ................................ ................................ ................................ ............ 131 Summary and Sig nificance ................................ ................................ ................... 137 5 SUMMARY AND CONCLUSIONS ................................ ................................ ........ 153 APPENDIX A ADDITIONAL ANALYSIS OF PHARMACOGENETIC PREDICTORS OF THIAZIDE INCUDED DYS GLYCEMIA ................................ ................................ 159 B ADDITIONAL ANALYSIS OF PEAR FOLLOW UP STUDY DATA ....................... 168 LIST OF REFERENCES ................................ ................................ ............................. 175 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 192
8 LIST OF TABLES Table page 1 1 Overview of genetic variants associated wit h type 2 diabetes ............................ 35 2 1 Study populations and study design for TCF7L2 SNP discovery, LD characterization, and statistical analyses ................................ ........................... 67 2 2 Strongest putative functional variants from TCF7L2 sequence data determined in silico ................................ ................................ ............................. 67 2 3 Characteristics of new onset diabetes cases and controls at baseline in INVEST sequenced samples ................................ ................................ .............. 68 2 4 Validation of candidate SNP method in INVEST sequenced whites ................... 69 2 5 Identification of candidate SNPs by race/e thnic groups in INVEST sequenced Hispanics and blacks ................................ ................................ .......................... 70 2 6 Characteristics of new onset diabetes cases and controls at baseline and during INVEST ................................ ................................ ................................ .... 71 2 7 Association of top candidate SNPs by race/ethnic groups in sequenced samples and INVEST GENES new onset diabetes case control cohort ............. 72 3 1 Summary of candidate gen es investigated as pharmacogenetic predictors ..... 108 3 2 Baseline characteristics of PEAR patients by randomized treatment arm ........ 109 3 3 Association of linear regression model covariates from primary analysis with change in fasting glucose in PEAR ................................ ................................ ... 110 3 4 Characteristics of new onset diabetes cases and controls at baseline and during INVEST ................................ ................................ ................................ .. 111 3 5 Association of logistic regression model covariates and new onset diabetes in INVEST ................................ ................................ ................................ ......... 112 3 6 Significa nt associations for candidate gene tag SNPs on change in fasting glucose in PEAR correction ................................ ................................ .............. 113 3 7 Odds ratios for KCNJ1 SNPs and haplotypes for new onset diabetes during HCTZ treatment by r ace/ethnicity in INVEST ................................ ................... 114 4 1 Characteristics of PEAR Follow Up Study participants at baseline and at follow up ................................ ................................ ................................ ........... 139
9 4 2 Chara cteristics of PEAR Follow Up Study participants during follow up period. ................................ ................................ ................................ .............. 140 4 3 Fasting glucose levels at baseline and at follow up by drug treatment status .. 141 4 4 Variables associated with FG changes during long term thiazide treatment .... 141 4 5 Variables associated change in HOMA changes during long term thiazide tre atment ................................ ................................ ................................ .......... 142 4 6 Variables associated change in insulin changes during long term thiazide treatment ................................ ................................ ................................ .......... 142 4 7 Variables associated w ith triglyceride changes during long term thiazide treatment ................................ ................................ ................................ .......... 143 4 8 Variables associated with uric acid changes during long term thiazide treatment ................................ ................................ ................................ .......... 143 4 9 Variables associated with FG at follow up ................................ ........................ 144 4 10 Variables associated with two hour OGTT glucose at follow up ....................... 145 4 11 Variables associated with HbA1c at follow up ................................ .................. 145 4 12 Variables associated with OGTT AUC at follow up ................................ .......... 146 4 13 Variables associated with one hour OGTT glucose at follow up ....................... 146 A 1 Candidate gene SNPs which deviated from Hardy Weinberg Equilibrium in at least one race/ethnic group in PEAR and INVES T ................................ ........... 159 A 2 SNP effects on change in fasting glucose in PEAR for SNPs previously associated with thiazide induced dysglycemia ................................ ................. 161 A 3 INVEST NOD odds ratios for SNPs previously associated with thiazide induced dysglycemia ................................ ................................ ........................ 162 A 4 Stepwise multivariate models for including genetic and pharmacogenetic predictor SNPs in P EAR ................................ ................................ ................... 163 A 5 Stepwise multivariate models for including genetic and pharmacogenetic predictor SNPs in INVEST ................................ ................................ ................ 163 B 1 Variables as sociated with LDL changes during long term thiazide treatment ... 168 B 2 Variables associated with HDL changes during long term thiazide treatment .. 169
10 B 3 Variables associated with total cholesterol changes during long term thiazide treatment ................................ ................................ ................................ .......... 170 B 4 Variables associated with serum potassium changes during long term thiaz ide treatment ................................ ................................ ............................. 171
11 LIST OF FIGURES Figure page 1 1 Theoretical framework of dissertation research aims ................................ ........ 36 2 1 Summary of Chapter 2 meth odology by patient population ................................ 74 2 2 Haploview generated linkage disequilibrium (LD) plot of sequenced TCF7L2 SNPs in INVEST. ................................ ................................ ................................ 75 2 3 Venn diagrams for TCF7L2 polymorphisms by race/ethnicity in sequenced s amples. ................................ ................................ ................................ ............. 77 2 4 Plot of principal components one and two in sequenced samples by self reported race/ethnicity. ................................ ................................ ....................... 78 2 5 Odds ratios per copy of allele and 95% confidence intervals for TCF7L2 SNPs and new onset diabetes in INVE ST patients by race/ethnicity. ................ 79 3 1 Physiological role of candidate genes in development of hy perglycemia after thi azide diuretic administration. ................................ ................................ ........ 115 3 2 Assessment of fasting glucose in the PEAR study design. ............................... 116 3 3 Chan ge in fasting glucose during hydrochlorothiazide treatment by KCNJ1 SNP rs17137967 genotype in black PEAR patients ................................ ......... 117 3 4 Odds ratios per copy of allele and 95% confidence intervals for KCNJ1 SNPs nominally associated (p<0.05) with new onset diabetes during hydrochlorothiazide treatment in IN VEST patients by race/ethnicity ................ 118 4 1 Progression of subjects for PEAR Follow U p Study e nrollment and analysis ... 147 4 2 Change in fasting plasma glucose during short term versus long term thiazid e diuretic treatment ................................ ................................ ................. 148 4 3 Mean fasting plasma glucose at baseline, end of short term thiazide treatment, and end of long term thiazide treatmen t by antihypertensive therapy. ................................ ................................ ................................ ............ 149 4 4 Change in fasting plasma gluc ose during long term thiazide diuretic treatment versus duration of follow up. ................................ ............................. 150 4 5 Change in fasting plasma glucose versus change in serum potassium during long term thiazide diuretic trea tment.. ................................ ............................... 151 4 6 Venn diagram of participants with IFG, IGT, and/or EGI. ................................ 152
12 A 1 Haploview generated linkage disequilibrium (LD) plot of KCNJ1 SNPs in INVEST whites. ................................ ................................ ................................ 164 A 2 Haploview generated linkage disequilibrium (LD) plot of nominally significant ADD1 SNPs in PEAR non blacks ................................ ................................ ..... 165 A 3 Haploview generated linkage disequilibrium (LD) plot of ADD1 SNPs in INVEST whites. ................................ ................................ ................................ 166 A 4 Area under the receiver operating characteristic curve for INVEST HC TZ treated white patients.. ................................ ................................ ..................... 167 B 1 Mean fasting glucose after short term and long term thiazide treatment including patients treated with anti diabetic medications. ................................ 172 B 2 Mean fasting glucose after short term and long term thiazide treatment by add on antihypertensive treatment including patients treated with anti diabetic medications. ................................ ................................ ........................ 173 B 3 Mean fasting glucose after short term and long term thiazide treatment by thiazide and statin therapy. ................................ ................................ ............... 174
13 LIST OF ABBREVIATION S 2logL negative two natural log of the likelihood function 95%C I 95% Confidence Interval Micro international units per milliliter ACE Angiotensin I converting enzyme gene ACE Angiotensin I converting enzyme ACE I Angiotensin II converting enzyme inhibitor ADA American Diabetes Association ADD1 Alpha adducin 1 ge ne AGTR1 Angiotensin II type 1 receptor gene AIM Ancestry informative marker ALLHAT Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial AME Adverse metabolic effect ARB Angiotensin II type 1 receptor blocker ARIC Atherosclerosis R isk In Communities AUROC A rea under the receiver operator characteristic BMI B ody mass index BP Blood pressure CAD Coronary artery disease CCB Calcium channel blocker CHF Congestive heart failure CV C ardiovascular dbSNP Database for single nucleotide polym orphisms DNA Deoxyribonucleic acid EGI Elevated glucose intolerance
14 ENaC Epithelial sodium channel ESE Exonic splice enhancers ESRD End stage renal disease ESS Exonic splice site FDR F alse discovery rate GCRC G eneral clinical research center GenHAT Geneti cs of Hy pertension Associated Treatment GERA Genetic Epidemiology of Responses to Antihypertensives GLP1 Glucagon like peptide 1 GNB3 guanine nucleotide binding protein b eta polypeptide 3 G WAS Genome wide association study HbA1c Percent glycated hemoglobin HCTZ Hydrochlorothiazide HDL High density lipoprotein HOMA H omeostatic model assessment HWE Hardy Weinberg Equilibrium I/D Insertion/deletion IBD Identity by descent IDF International Diabetes Federation IFG Impaired fasting glucose IGT Impaired glucose t olerance INVEST INternational Verapamil SR Trandolapril STudy INVEST GENES INternational Verapamil SR Trandolapril STudy GENEtic Substudy IQR Interquartile range IR I mmediate release
15 JCVI J Craig Venter Institute JNC6 Sixth report of the Joint National Com mittee on prevention, detection, evaluation, and treatment of high blood pressure JNC7 Seventh report of the Joint National Committee on prevention, detection, evaluation, and treatment of high blood pressure Kb Kilo (thousand) base pairs KCNJ1 P otassium i nwardly rectifying channel, subfamily J, member 1 gene LD Linkage disequilibrium LDL Low density lipoprotein LVH Left ventricular hypertrophy MAF Minor allele frequency mEq/L milliequivalents per liter mg/dL Milligrams per deciliter MI Myocardial infarctio n mmHg Millimeters of mercury mRNA Messenger ribonucleic acid NCBI National Center for Biotechnology Information NHLBI National Heart, Lung and Blood Institute NIH National Institutes of Health NOD N ew onset diabetes OGTT O ral glucose tolerance test OMB O ffice of Management and Budget OR Odds ratio PC Principal component PCA Principal components analysis PCR Polymerase chain reaction
16 PEAR Pharmacogenomic Evaluation of Antihypertensive Responses PEAR 2 Pharmacogenomic Evaluation of Antihypertensive Response s 2 PHARMO RLS Pharmaco Morbidity Record Linkage System QC Quality control RAS Renin a ngiotensin s ystem ROMK1 Renal outer medullary potassium channel 1 RS&G Resequencing and genotyping service SD Standard deviation SE Standard error SNP Single nucleotide polymorphism SR Sustained release SSRI Selective serotonin reuptake inhibitor T2D Type 2 diabetes TCA Tricyclic antidepressant TCF7L2 Transcription factor 7 like 2 gene TFBS Transcription factor binding site UCSC University of California at Santa Cruz UF University of Florida US United States UTR Untranslated region WGA Whole genome amplified
17 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of D octor of Philosophy GENETIC, PHARMACOGENETIC, AND PHARMACOTHERAPEUTIC RISK FACTORS FOR THIAZIDE INDUCED DYSGLYCEMIA By Jason Hansen Karnes August 2012 Chair: Rhonda M Cooper DeHoff Cochair: Julie A Johnson Major: Pharmaceutical S ciences Hypertension and type 2 diabetes (T2D) are major contributors of morbidity and mortality T hiazide diuretics are first line antihypertensive agents but are associated with T2D in some individuals. Little knowledge currently exists to identify individuals at risk for thiazide induced dysg lycemia defined as alterations in glucose homeostasis T his research utilize s several phenotypes along a continuum of hyper glycemia to determine genetic, pharmacogenetic, and pharmacotherapeutic risk factors for hydrochlorothiazide (HCTZ) induced dysglycemia First we sought to identify genetic risk factors in the TCF7L2 gene for T2D in African and Hispanic ethnic/race groups. After sequenc ing TCF7L2 i n new onset diabetes (NOD) case s and age, gender and race/ethnicity matched con trols from the INternational V E rapamil SR Trandolapril STudy ( INVEST ), w e identified 910 novel variants and genotyped potential T2D predictors in a larger INVEST NOD case/control cohort. We f ound no novel T2D risk predictor single nucleotide polymorphisms ( SNPs ) in African or Hispanic race/ethnic groups We found nine TCF7L2 SNPs with significant pharmacogenetic effects on thiazide induced NOD with the strongest SNP*HCTZ
18 treatment interaction for rs7917983 ( p= 3.7x10 4 p FDR = 0.02 ) suggest ing that TCF7L2 SNPs influence NOD risk during HCTZ treatment W e then investigated the candidate gene s KCNJ1 ADD1 ACE and AGTR1 to determine pharmacogenetic risk factors for thiazide induced dysglycemia in the Pharmacogenomic Evaluation of Antihypertensive Responses (PEAR) study and INVEST In PEAR, we found a significant association between the KCNJ1 rs17137967 C allele (beta=8.47, p=0.0008 [p FDR =0.009]) and the ACE rs4303 A allele (beta= 6.39 p= 6.80x10 4 [p FDR =0.03]) with change in fasting glucose (FG) during th iazide treatment. In INVEST, multiple significant SNP*HCTZ treatment interactions were found for KCNJ1 in each race/ethnic group. P harmacogenetic risk factors remained significant after adjustment for TCF7L2 SNPs from T2D genome wide association studies suggest ing that pharmacogenetic effects of candidate gene variation were present regardless of an Finally, we conducted a n original clinical study to investigate thiazide treatment duration as a risk factor for thiazide induced dysglycemia The PEAR F ollow Up Study enrolled previous PEAR and Pharmacogenomic Evaluation of Antihypertensive Responses 2 ( PEAR 2 ) study p articipant s continuously treated with HCTZ or chlorthalidone for more than six months W e observ ed t hat increased thiazide treatment duration (beta=0.34, p=0.008) and decreased baseline FG (beta= 0.46, p=0.02) were associated with increased FG during long term thiazide treatment Our results suggest that thiazide induced FG increases p ersist during long term treatment but are not predicted by short term changes in FG
19 CHAPTER 1 INTRODUCTION AND BAC KGROUND Hypertension and Type 2 Diabetes One third of adults in the United States (US) have hypertension, defined as a systolic blood pressure (BP) greater than or equal to 1 40 millimeters of mercury ( mmHg ) or diastolic BP 1 2 For an individual who is normotensive at age 55 years, the lifetime probability of developing hypertension is 90 % 3 Hypertension is a major underlying c ause of cardiovascular ( CV ) disease as a strong relationship exists between BP and CV risk that is independent of other risk factors. 2 E very 20 mmHg incremental increase in systolic BP or 10 mmHg in diastolic BP doubles the risk of CV death, including death due to stroke ischemic heart disease, and other vascular causes 4 5 Hypertension also has a major economic impact, as the estimated direct and indirect cos t for hypertension was $73.4 billion in 2009. 1 H ypertensive patients are at a two fold greater risk of dev eloping Type 2 Diabetes ( T2D ) compared to non hypertensive patients and t he presence of both hypertension and T2D can increase CV risk up to three fold 6 7 T2D constitutes a major health problem in the US and the rest of the world T2D is a leading cause of CV eye, kidney, and neurological disease and the global healthcare expenditure on diabetes is expecte d to total $490 billion in 2030. 8 The World Health Organization predicts that by the year 2030, 3 66 million individuals worldwide will have diabetes 9 and the International Diabetes Federation (IDF) estimate is much higher at 552 million. 10 Worldwide diabetes estimates represent up to a 69 percent increase in adults with diabetes in developed countries between 2010 and 2030. 11
20 T2D increa ses CV risk at any level of BP and has been designated as a CV disease risk equivalent according to the National Cholesterol Education Program Adult Treatment Panel III 7 12 14 Even pre diabetes ( impaired fasting glucose [ IFG, fasting glucose 100 125 mg/dL ] or impaired glucose tolerance [ IGT, 2 hour oral glucose tolerance te st ( OGTT ) 140 199 mg/dL ]) increases the risk of CV disease and has a prevalence of nearly 37 percent in the US. 15 20 An increased CV risk with pre diabetes suggests that CV risks begin at a threshold of glucose that is lower than what is used to diagnose diabetes. Individuals with IFG and IGT are also at an increase d r isk of developing diabetes. 21 25 Since T2D is a preventable condition in up to two thirds of high risk cases, 26 27 e arly identification of at risk patients has the potential to reduce or delay progression to T2D and related CV and microvascular disease. 28 Blood Pr essure Reduction with Thiazide Diuretics Lowering BP with antihypertensive medications is an important means of CV risk reduction 2 A 10 year, 12 mmHg decrease in BP results in the prevention of 1 death for every 11 patients treate d. 29 Antihypertensive therapy is associated with 35 40% reduction in stroke, 20 25% decrease in myocardial infarction (MI), and greater than 50% reduction in c ongestive heart failure ( C HF). 30 The thiazide diuretic hydrochlorothiazide ( HCTZ ) is one of the most commonly prescribed antihypertensive s in the US, with over 90 million HCTZ prescriptions in 2010, either as a single agent or in combination formulations. 31 Thiazide diuretics have been the gold standard of antihypertensive therapy in most clinical trials and they remain a first line recommendation for treating uncomplicated hypertension. 2 The Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) and a meta analysis by Psaty et al suggest the superiority of
21 thiazide diuretics over other agents in terms of adverse CV outcome reduction. 32 33 However, the majority of clinic al trial evidence shows that, in uncomplicated hypertension, the benefit of antihypertensive agents with respect to overall CV outcomes is driven by BP reduction rather than the drug class of the antihypertensive agent 34 41 Several trials have observed that thiazide diuretics a re inferior to other antihypertensive drug classes in terms of CV risk reduction whether used as first line therapy or in combination wit h other agents 42 43 E vidence is available to support the use of thiazide diuretics in many patient populations, in head to head comparisons with other antihypertensives, and in combination with other antihypertensive agents. For these reasons and because of low cost, clinical guidelines recommend thiazide diuretics as initial therapy for most patients with uncomplicated hypertension. 2 44 Thiazide Induced Dysglycemia Thiazide s are first line agents in the treatment of hypertension, but this commonly prescribed antihypertensive class is associated with adverse m etabolic effects (AMEs) such as dysglycemia F or the purposes of this dissertation, dysglycemia is defined as disru ption of glucose metabolism, includ ing both hyperglycemia and T2D development. 45 51 Thiazide diuretics can adversely affect glucose homeostasis, which has led to a reproducible association between thiazides and new onset diabetes ( NOD ) i n large s cale clinical trials. 50 53 Many large, prospective, clinical trials have shown significant differences in the rates of NOD for those treated with thiazide and thiazide like diuretics compared with angiotensin I converting enzyme (ACE) inhibitors (ACEIs), angiotensin II type 1 receptor blockers (ARBs), and calcium channel blockers (CCBs) 32 35 42 50 A meta analysis by Elliott et al observed that thiazide diuretics confer the greatest risk of NOD compared to all other major antihypertensive classe s. 47
22 Several other meta analyses have shown increased incidences of NOD with thiazide diuretics compared to other antihypertensive classes. 54 58 Differences in study design, follow up period, sample size, patient populations, drug dosage, and comparator agents make direct comparisons of thiazide induced NOD between trials and interpretation of results difficult. Antihypertensive trials also lack con sistency with respect to the diagnostic criteria of diabetes 33 Despite these difficulties, prospective clinical trials, meta analyses and retrospective studies have demonstrated a reproducible association for increases in NOD with thiazide diuretics. In addition, major controversy exists with respect to the clinic al implications of thiazide induced NOD. Some argue that T2D developed as a direct result of thiazide diuretic treatment does not carry the same CV risk as T2D through other risk factors, 49 59 60 while others report non significant trends of increased CV risk associated with NOD. 61 62 In addition, most studies suffer from insufficient follow up periods since they were not designed a priori to determine differen ces in T2D risk between antihypertensive classes. 53 The majority of studies of clinical implications of NOD have observed significant increases in CV risk with antihypertensive induced NOD. 6 63 67 Currently, the prognostic impact of thiazide induced NOD is not clear. The lack of direct supporting evidence of a drug:NOD:CV outcome relationship may be due to a lack of a priori study design and inadequate statistical power. However, there is little reason to expect that T2D occurring during antihypertensive therapy carries any less CV risk than T2D developed as a result of other factors, especially considering the wealth of evidenc e supporting CV risk for T2D.
23 Mechanisms o f Thiazide Induced Dysglycemia Although thiazides have the potential to raise fasting glucose ( FG ) levels, the mechanisms by which they cause dysglycemia are not fully understood. Several theories stipulate that thiazide induced T2D is related to potassiu m, uric acid, skeletal muscle and pancreatic perfusion, and/or the Renin Angiotensin System (RAS). 46 68 Thiazide diuretics cause volume loss resulting in RAS stimulation, which has been associated with impairment of glucose metabolism. 69 70 The RAS is further implicated in NOD by the fact that RAS blocking antihypertensive agents are associated with decreased NOD incidence in clinical trials. 48 50 51 71 Thiazide diuretics also interfere with renal excre tion of uric acid and an elevated level of uric acid has been associated with obesity and T2D 72 74 Some evidence supports that thiazide induced dysglycemia is mediated by potassium. 75 76 Thiazide diuretics may increase FG through increases in potassium excretion, which may blunt insulin release or decrease glucose uptake into skeletal muscle. 46 77 An inverse relationship between serum potassium and F G levels was observed in a meta analysis of 59 clinical studies 76 A recent anal ysis from the Atherosclerosis Risk In Communities (ARIC) Study observed a greater increase in NOD risk in patients with low (< 4.0 mEq/L) and moderate (4.0 4.5 mEq/L) serum potassium levels compared to high potassium levels (4.5 5.0 mEq/L). 78 However, these studies are s econdary analyses of population averages rather than individual patient data. An analysis of the Pharmacogenomic Evaluation of Antihypertensive Responses ( PEAR ) study used individual patient data and found n o correlation between FG increase and serum potassium decrease. 79
24 The association between thiazide diuretics and dysglycemia is strong considering the reproducibil ity of this association from antihypertensive clinical trials the physiologic plausibility of proposed mechanisms and the paucity of directly conflicting evidence. Since thiazide diuretics are first line antihypertensives, the ir potential to induce T2D and associated CV risk is concerning. Further knowledge regarding the mechanisms of thiazide induced NOD could help identify more effective predictors of thiazide induced NOD Genetics of Type 2 Diabetes Although development of T2D is affected by a number of well described environmental factors, T2D has strong genetic influences. 80 81 Major differences in disease prevalence among different racial/ethnic groups, including a disproportionately high T2D prevalence in Hispanics and blacks, 1 suggest a genetic component to T2D. Genetic factors are further implicated by increased rates of T2D among persons who have first degree relatives with the disease. 82 83 Offspring of one parent with T2D have a 40 percent lifetime risk of deve loping T2D, whereas offspring of two parents with T2D have an almost 70% lifetime risk of developing T2D 84 In addition, concordance rates for the presence of T2D are higher in monozygotic twins (up to 70%) than in dizygotic twins (20 30%). 85 90 T2D heritability is estimated to exceed 50 percent and sibling relative risk 82 88 The large number of single nucleotide polymorphisms ( SNPs ) that have now been associated with T2D also supports a genetic basis for T2D development. Since initial T2D genome wide association studies ( GWAS s) in 2006 and 2007, 91 92 a deluge of T2D GWAS and large scale meta analyses of GWAS data have been published to elucidate SNPs that confer T2D susceptibility. 80 91 93 94 Over 40 loci have been associated with
25 T2D with over 30 of these loci reproducibly associated with T2D. 93 (Table 1 1) T2D associated SNPs are usually common in studied populations and have modest effect sizes with most odds ratios ranging from 1. 1 1.3. Many of these T2D risk alleles have also been associated with T2D related traits such as percent glycated hemoglobin (HbA1c), body mass index (BMI), and obesity Although the involvement of a number of the associated genes in the pathophysiology of T2D is unclear, many associated genes play roles in insulin secretion, insulin resistance, and obesity that can be easily linked to the T2D phenotype. Several T2D SNP associations were found in gen es known to cause monogenic forms of diabetes and in genes encoding direct targets of anti diabetic drugs ( KCNJ11 and PPARG ). Further knowledge of genes involved in T2D physiology may prove useful in the identification of possible drug targets as well as the prediction of individual drug response. GWAS has contributed to our knowledge of the genetics and pathophysiology of T2D, but our ability to translate GWAS results into clinical T2D prediction algorithms suffers from several major limitations. While a considerable number of SNPs have been reproducibly associated with T2D, SNPs from GWAS currently explain only 10 15% of T2D heritability. 95 T2D risk scores that include GWAS associated SNPs have not been shown to improve T2D risk prediction algorithms when family history of T2D is known. 96 97 In addition, T2D GWAS has been performed almost exclusively in populations with European and Asian ancestry, leaving populations such as Hispanics and blacks largely uninvestigated. Considering the strong genetic component of T 2D development, further study of genetic associations with T2D, especially in black and
26 Hispanic populations, may prove beneficial in identifying those at an increased risk for developing the disease The Transcription Factor 7 Like 2 Gene ( TCF7L2 ) Of the SNPs associated with T2D in GWAS, variants in the Transcription Factor 7 Like 2 gene ( TCF7L2 ) specifically the SNP rs7903146, h a ve been the strongest and most reproducible signals yet identified. 92 98 100 A recent meta analysis in 42,542 individuals with T2D and 98,912 controls of European descent confirmed that rs7903146 has the strongest association with T2D of any k nown genetic marker. 95 The relative risk of the rs7903146 T allele for T2D is approximately 1.46 (meta analysis p=1x10 140 ), making the TCF7L2 SNP rs7903146 t he genetic risk factor most strongly associated with T2D and conferring the greatest risk of T2D 101 Although GWAS have identified SNPs in TCF7L2 as consistent and reliable risk factors for T2D in European populations, TCF7L2 SNPs identified in European populations may not be genetic risk factors for T2D in African and Hispanic populations 91 99 102 A robust association of TCF7L2 SNPs and T2D is observed from GWAS in European populations, 99 102 but s tudies conducted in groups with non European ancestry have not yielded similarly robust association s between TCF7L2 SNPs an d T2D. 98 103 The relatively small number of s tudies in po pulations with African and Hispanic descent have inconsistently observed genetic associations. 101 103 108 A recent study including sequencing of the rs7903146 region in African American concluded that rs7903146 was the most highly associated TCF7L2 SNP. 109 However, other studies in populations of African descent have observed no association between the rs7903146 T allele and T2D. 99 103 104 A study by Waters et al. observed significant associations with rs7903146 in Europeans ( odds ratio [OR] 1.55 [ 1.29 1.87 ]) Af rican
27 Americans (OR 1.32 [ 1.16 1.51 ]), and Hispanics (OR 1.31 [ 1.19 1.45 ]). 110 Other studies in Hispanic populations have yielded non significant ORs for rs790314 6 and T2D, primarily in individuals of Mexican descent. 106 107 Diversity and admixture among Hispanic popul ations make interpretation of genetic associations especially difficult. Difficulty in interpretation is exacerbated by the fact that few association studies have been published and individuals of Mexican descent do not provide an adequate model of geneti c diversity for the Hispanic population. In the US, the prevalence of diagnosed T2D in 2012 was estimated at 7.1% for non Hispanics whites, 11.8% for Hispanics, and 12.6% for non Hispanic blacks. 1 Despite the higher T2D prevalence in African and Hispanic Americans, studies in these race/ethnic groups have attempted to replicate TCF7L2 associations from studies in Europeans rather than perform GWAS or acquire complete coverage of the TCF7L2 gene in those populations In stark contrast to data available in Europeans, few T2D GWAS ha ve been published in populations of African or Hispanic descent. Th e lack of publish ed GWAS suggests that alternate TCF7L2 SNPs could more accurately predict T2D in African and Hispanic populations. GWAS data are particularly important in African and Hispanic race/ethnic groups, who suffer from much higher rates of T2D. The identificati on of genetic risk factors for T2D in African and Hispanic populations forms the focus of Aim 1 of this dissertation, described in Chapter 2. (Figure 1 1) Pharmacogenetic s of Thiazide Induced Dysglycemia A priori identification of patients who will develop hyperglycemia during thiazide treatment may identify patients who would benefit from treatment with alternate antihypertensives Consistent evidence from clinical trials supports that thiazide s increase glucose levels and risk for T2D however this adver se metabolic effect does
28 not occur in all patients and clinical factors such as BMI and elevated BP are also important predictors of NOD 49 51 111 Inter individual variability in thiazide induced dysglycemia and strong genetic associations with T2D suggest that thiazide induced dysglycemia may also h ave pharmaco genetic influences In addition to known clinical predictors of NOD, i dentification of patient specific genetic risk factors for thiazide induced dysglycemia has the potential to delay or prevent T2D through preferential prescription of altern ative antihypertensive agents. 46 Pharmaco genetic associations with thiazide induced dysglycemia have been observed, suggesting that pharmacogenetic risk factors could be used to predict thiazide induced glucose increases 112 114 Bozkurt et al. found significant associations between the angiotensin II type 1 receptor ge ne ( AGTR1 ) + A1166C, the ACE gene ( ACE ) G4656C, and the guanine nucleotide binding protein b eta polypeptide 3 ( GNB3 ) C825T polymorphisms and NOD in individuals treated with thiazides and thiazide like diuretics compared with those not treated with these age nts. 112 An investigation from the Genetics of Hy pertension Associated Treatment (GenHAT) study observed that the amiloride sensitive epithelial sodium channel ( SCNN1A ) SNP rs2228576 A allele was associated with increased FG in individuals treated with a CCB compared with those treated with the thiazide like diuretic chlorthalidone. 113 Supporting the role of potassium depletion in thiazide induced changes in glucose metabolism, the non synonymous SNP rs59172778 (M338T) in the potassium inwardly rectifying channel subfamily J member 1 (ROMK1) gene ( KCNJ1 ) has been associa ted with decreased FG after four weeks of HCTZ treatment. 114
29 The three pharmacogenetic studies on thiazide induced dysglycemia have left the majori ty of genetic variability in candidate genes uninvestigated. P harmacogenetic studies on thiazide induced dysglycemia have focus ed on a small number of putative functional SNPs in genes involved in electrolyte homeostasis and the R A S In addition, despite racial/ethnic differences in T2D prevalence, in genetic associations with T2D, and in linkage disequilibrium ( LD ) structure, the impact of r ace/ethnicity on genetic and pharmacogenetic risk factors for thiazide induced dysglycemia remains largely uninvest igated. 98 115 Further study of the influence of can didate genes with roles in electrolyte homeostasis and the RAS on thiazide induced dysglycemia, as well as replication of previous pharmacogenetic associations, might help determine genetic markers for identification of patients who will develop dysglycemi a during thiazide treatment The identification of pharmacogenetic risk factors for dysglycemia using a candidate gene approach forms the basis for Aim 2 of this dissertation research, described in Chapter 3. (Figure 1 1) Short and Long Term Thiazide Indu ced Dysglycemia The long term effects of antihypertensive medications on dysglycemia are well studied, but randomized BP reduction trials consider antihypertensive related AME s as a secondary outcome or in secondary analyses. 53 The results of investigations into antihypertensive related AME s are thereby subject to bias and have not provide d conclusive evidence. Clinical trials investigating thiazide diuretics for BP reduction typically monitor FG every six months or every year, which may be adequate for clinical monitoring but not for characterization of short term AME s of these drug classes. To date, only limited data are available compar ing short term (1 2 months) and long er term
30 ( between 6 months and five years ) AME s of antihypertensive therapy in the same patient population 116 Th e PEAR study was designed to test patient BP response and AME s after administration of the thiazide diuretic HCTZ and/or the beta blocker atenolol. 117 The PEAR study quantified changes in BP FG, insulin, serum potassium and homeostatic model assessment (HOMA) during 9 18 weeks of antihypertensive therapy with atenolol 100 mg, HCTZ 25 mg, and their combination. During HCTZ monotherapy and atenolol plus HCTZ combination therapy in PEAR patients with abdominal obesity, the proportion of pa tien ts with new onset IFG increased and the proportion of pa rticipants with high density lipoprotein ( HDL ) mg/dL was significantly increased compared to patients without abdominal obesity 118 The observation of increased AMEs during HCTZ treatment confirmed that AME s of thiazide diuretics could be observed in a short term period and supports the utility of patient specific characteristics such as abdominal obesity in predicting AMEs of thiazide diuretics Howe ver, the PEAR study duration ( approximately 18 weeks) is insufficient to characterize th iazide effects on glucose over a longer duration of time ( between six months and five years ). Since hypertensive treatment typically requires lifetime therapy, charact erization of long term effects of antihypertensive medications on glucose and lipid status is important. T he relative severity of glucose alterations during short and long term antihypertensive therapy and characteristics that make patients more susceptib le to hyperglycemia during long term antihypertensive treatment warrants further investigation
31 T hiazide diuretic induced changes in glucose metabolism are typically evaluated using FG and T2D is not diagnosed 119 Howeve r, CV risk does not begin when a patient FG becomes greater than 126 mg/dL IFG carr ies an adverse prognostic impact on CV outcomes although evidence is inconclusive. 15 17 120 In addition, several studies suggest that IGT or the two hour glucose measurement during an OGTT, may be a better predictor of CV disease and mortality than FG. 121 126 HbA1c has also been shown to better predict CV risk than FG. 127 Despite the CV risk associated with IFG, IGT, and increased HbA1c, analyses of dysglycemia in BP reduction trials focus 126 mg/dL. 119 A ddition ally identification of a patient in early stages of impaired beta cell function or insulin sensitivity can be difficult. FG may not be a very sensitive measure for metabolic abnormality since T2D prevalence determined using FG underes timates T2D prevalence obtained using the two hour OGTT glucose as FG values often appear normal in patients with IGT. 119 126 In individuals with IGT, hyperglycemia may only manifest when the individual is challenged by an OGTT a procedure involving a 75 gram oral glucose load and plas ma glucose measurements at baseline, one hour, and two hour time point s 128 Guidelines from the American Diabetes Association (ADA) suggest a 2 hour post load plasma glucose less than 140 mg/dL indicates normal glucose tolerance, 140 to 199 mg/dL indicates IGT, and greater than 200 mg/dL indicates a positive T2D diagnosis. 119 The OGTT has not been adopted in clinical practice nor is OGTT utilized in controlled trials primarily due to increased cost, time, and inconvenience of the procedure. However utilization of OGTT may be clinically valuable in patients who are
32 suspected of having metabolic abnormalities, especially considering the utility of the OGTT for CV risk prediction. In contrast, HbA1c has several advantages over FG and has been recomme nded by the ADA 6.5%). 119 The gold standard for in vivo studies of insulin sensitivity is the hyperinsulinemic euglycemic clamp, however this method is not used clinically due to an increased time and risk associated with the procedure. 129 HOMA defined as the product of fasting insulin and FG divided by 405, is a convenient, noninvasive indicator of insulin sensitivity that correlates well with insulin sensitivity determined by the hype rinsulinemic euglycemic clamp. 130 Despite the utility of HbA1c, the OGTT, and HOMA in diagnosing metabolic abnormalities and pr edicting CV risk, the effects of thiazide diuretics on HbA1c, OGTT and HOMA are not well understoo d. The identification of thiazide treatment duration as a risk factor for AMEs and the need to acquire detailed glycemic characteristics of patients after l ong term thiazide treatment forms the basis of Aim 3 of this dissertation research, described in Chapter 4. (Figure 1 1) Summary and Significance Hypertension and T2D are major contributors of morbidity and mortality. Thiazide diuretic s are first line age nts for the treatment of hypertension, but can induce T2D in certain individuals. Increasing prevalence of T2D in the US makes T2D prevention a globa l public health priority. 131 133 Data are lacking to allow identif ication of individuals at risk for thiazide induced FG increases and T2D The ability to identify at risk individuals would enable physicians to avoid thiazide s in these patients and preferentially prescribe alternate antihypertensive drugs. A proactive approach to
33 identifying individuals at risk for T2D may allow early reduction of modifiable risk factors, such as thiazide pharmacotherapy, and eventually lead to T2D prevention. 53 The research outlined in this dissertation utilizes several phenotypes along the diabetes continuum including change in FG and NOD, which allows evaluation of both a surrogate and a clinical endpoint for metabolic and CV risk. NOD carries greater clinical significance but variability in this phenotype is more susceptible to environme ntal effects, confounding genetic and pharmacogenetic associations. Change in FG is a surrogate for eventual T2D but variability in this endpoint is more likely to be directly related to gene products and thiazide therapy. Th e research described under Ai m 1 seeks to identify SNPs in TCF7L2 as risk factors for T2D in under represented ethnic/race groups and to investigate the impact of race/ethnicity on SNP associations Identification of these genetic risk factors in TCF7L2 m ay help improve our risk asse ssment for T2D and increase our understanding of T2D pathophysiology A ccess to TCF7L2 sequencin g data in genetically diverse race/ethnic groups from the INternational VErapamil SR Trandolapril STudy ( INVEST ) represents a unique opportunity to study diffe rences in LD structure and identify alternate T2D predictor SNPs in African and Hispanic populations from the US and Puerto Rico. Th e research described under Aim 2 utilizes tag SNPs for candidate genes involved in the RAS and electrolyte homeostasis chos en based on their potential impact on thiazide induced changes in FG and NOD Identification of pharmacogenetic risk factors for thiazide induced dysglycemia is aided by clinical and genetic data for two
34 large antihypertensive study populations which inco rporate thiazide treatment, provide adequate statistical power, and enable replication of genetic associations. Th e research described under Aim 3 seeks to refine the role of thiazide treatment duration as a pharmacotherapeutic risk factor for thiazide i nduced dysglycemia to compare s hort and long term effects of thiazide pharmacotherapy on FG and to acquire detailed glycemic characteristics after long term thiazide treatment Such an investigation might implicate duration of thiazide treatment in T2D risk and support more rigorous monitoring of FG during short term thiazide treatment to predict long term changes in FG Th e PEAR Follow Up Study will enroll previous PEA R participants after long term thiazide therapy, allowing a comparison of short and l ong term dysglycemia. The PEAR F ollow U p S tudy is unique in assessment of FG and detailed glycemic characteristics after long term thiazide therapy in a population for which FG response data are currently available after short term thiazide treatment. Th e aims of this research proposal are designed to determine genetic (Aim 1), pharmacogenetic (Aim 2), and pharmacotherapeutic (Aim 3) risk factors for thiazide induced dysglycemia. (Figure 1 1) Identification of genetic, pharmacogenetic, and pharmacotherap eutic risk factors may lead to a m ore comprehensive, accurate and effective assessment of individuals at risk for T2D during thiazide therapy. Improved T2D risk assessments may contribute to a clinical paradigm shift away from reactive long term FG monit oring during thiazide therapy and towards proactive prevention of T2D through avoiding thiazides in patients at risk for NOD and rigorous FG monitoring. P revention of T2D has the potential to reduce the disease burden of T2D, including a negative eco nomic and public health impact.
35 Table 1 1. Overview of genetic variants associated with type 2 diabetes Gene Locus/Loci Variant(s) Associated phenotype(s) Odds ratio for T2D ADAMTS9 rs4607103, rs6795735 T2D 1.09 1.1 ARAP1 (CENTD2) rs1552224 T2D 1.13 1.14 BC L11A rs243021 T2D 1.08 1.09 CAPN10 rs2975760, rs3792267 T2D 1.17 CDC123/CAMK1D rs12779790 T2D 1.09 CDKAL1 rs7754840, rs10440833 T2D 1.12 1.25 CHCHD9/TLE4 rs13292136 T2D 1.11 1.20 CDKN2A/2B rs10811661, rs10965250 T2D, CAD 1.20 DUSP9 rs5945326 T2D 1.27 FTO rs8050136 T2D, BMI, obesity 1.07 1.26 HHEX/IDE/KIF11 rs1111875, rs5015480 T2D 1.13 1.18 HMGA2 rs1531343 T2D 1.10 1.20 HNF1A rs7957197 T2D 1.07 1.14 HNF1B rs1920792, rs7501939, rs757210, rs4430796 T2D 1.1 1.17 HCCA2 rs2334499 T2D 1.35 IGF2BP2 rs 4402960, rs1470579 T2D 1.14 IRS1 rs2943641, rs7578326 T2D, CAD 1.09 1.12 JAZF1 rs864745, rs849134 T2D 1.12 1.13 KCNJ11 rs5219 T2D 1.09 1.14 KCNQ1 rs2237892, rs163184, rs231362 T2D 1.08 1.23 KLF14 rs972283 T2D 1.07 1.1 NOTCH2/ADAM30 rs10923931 T2D 1.1 4 PPARG rs1801282, rs13081389 T2D 1.14 1.24 PRC1 rs8042680 T2D 1.07 1.1 PTPRD rs17584499 T2D 1.57 RBMS1 rs7593730 T2D 1.11 SLC30A8 rs13266634, rs11558471, rs3802177 T2D, HbA1c 1.12 1.15 SRR rs391300 T2D 1.28 TCF7L2 rs7903146, rs12255372, rs7901695, rs4506565, rs11196218 T2D, FG, HbA1c 1.4 THADA rs7578597, rs11899863 T2D, cholesterol 1.15 1.17 TP53INP1 rs896854 T2D 1.06 1.1 TSPAN8 LGR5 rs7961581, rs4760790 T2D 1.11 WFS1 rs10010131, rs6446482, rs1801214 T2D 1.11 1.13 ZBED3 rs4457053 T2D 1.08 1.16 ZFAND6 rs11634397 T2D 1.06 1.11 BMI indicates body mass index ; CAD coronary artery disease ; FG fasting glucose ; HbA1c percent glycated hemoglobin ; T2D type 2 diabetes
36 Figure 1 1. Theoretical framework of dissertation research aims. Red arrows indi cate the modification of risk for dysglycemia by genetic, pharmacogenetic, and pharmacotherapeutic risk factors. Green arrows indicate the interaction of genetic risk factors and drug therapy to produce a pharmacogenetic interaction. TCF7L2 indicates Tra nscription factor 7 Like 2 gene ; SNP, single nucleotide polymorphism.
37 CHAPTER 2 SEQUENCING, DETERMINATION OF LINKAGE DISEQUILIBRIUM STRUCTURE, AND ASSOCIATION ANALYSIS IN TCF7L2 Introduction Genetic influences on T2D are well established primarily bas ed on T2D GWAS performed most ly in individual s with European ancestry. 81 82 S NPs in TCF7L2 have b e en identi fied as robust predictors of T2D risk in European populations but TCF7L2 SNPs inconsistently predict T2D in other race/ethnic groups. 91 99 102 Despite a higher prevalence of T2D in individuals of African or Hispanic descent the association of TCF7L2 SNPs and T2D in these populations remains unclear 98 99 101 103 108 110 134 S tudies conducted in groups of African or Hispanic descent have not identified TCF7L2 SNPs consistently associated with T2D risk. Although some evidence suggests that the TCF7L2 SNP rs7903146 is functional, 135 a SN P that is functionally responsible for the association between TCF7L2 and T2D has not yet been identified. A relatively small number of s tudies in African and Hispanic populations have observed significant associations between T2D with TCF7L2 SNPs and rep ort variable point estimates for TCF7L2 SNPs 101 103 108 A recent study including sequencing of the rs7903146 region in African American s concluded that rs7903146 which is the strongest SNP in European populations, 95 was the most highly associated TCF7L2 SNP in th e African Americans studied 109 However, other studies in populations of African descent have observed no asso ciation between the rs7903146 T allele and T2D. 99 103 104 S tudies in Hispanic populations have yielded non significant ORs, primarily in individuals of Mexican descent 106 107 Diversity and admixture among Hispanic populations make interpretation of genetic associations especially difficult. Interpretation of findings in Hispanic populations is also confounded by the fact that few
38 association studies have been published and that individuals of Mexican descent do not provide an adequate mo del of genetic diversity among other Hispanic population s Studies in populations of African or Hispanic descent have attempted to replicate T2D associations with TCF7L2 SNPs from studies in Europeans rather than explore TCF7L2 more broadly In addition, few T2D GWAS are currently available in African or Hispanic populations. 102 A recently published GWAS in African Americans observed a s trong signal in a gene previously unassociated with T2D and only nominal significance for the TCF7L2 SNP rs7903146. 134 Th e nominally significant association of TCF7L2 in African Americans contrasts with the reproducibly significant association s of rs7903146 and T2D observed in populations of European ancestry. In addition, t he T2D phenotyp e from th e GWAS in African Americans was complicated by end stage renal disease (ESRD) The presence of ESRD in all T2D cases in this study might explain the fact that the st r ong est genetic signals observed in the discovery cohort were related to diabetic nephropathy rather than T2D 134 C urrent GWAS chip genotyping methods are inadequate to capture low frequency variation and population specific variation in ind ividuals of African or Hispanic descent 102 S equence data in African and Hispanic populations m ay be useful in identifying novel T2D genetic signals and determining causal variants for T2D i n TCF7L2 Currently available sequence data are limited in TCF7L2 in African and Hispanic populations and catalogued v ariation in TCF7L2 is not comprehensive. 109 A ssociation studies in ethnically diverse populations have investigated only a small portion of TCF7L2 variation and alternate TCF7L2 SNPs may be better predictors of T2D in African and Hispanic populations. Additional TCF7L2 sequence data in balcks or
39 Hispanics might aid in the i dentification of genetic risk factors in diverse race/ethnic groups, where T2D prevalence is high and may aid in determin ing baseline T2D risk. More effective T2D predic tors could be used to improve T2D risk assessments in Af rican or Hispanic populations and to a ccount for underlying genetic T2D risk whe n investigating pharmacogenetic risk factors for NOD. In addition to genetic risk factors, many environmental factors are well known risk factors for T2D. 50 97 136 Consistent evidence supports that thiazide diuretic s are an environmental risk factor for T2D, as they have been associated with NOD in many randomized clinical trials 47 53 Strong g enetic influences on T2D and inter individual variability in NOD suggest pharmacogenetic s might play a role in thiazide induced T2D. In addition, TCF7L2 SNPs might act as phar macogenetic risk factors if the combination of TCF7L2 risk alleles and thiazide treatment increased NOD risk in a synergistic fashion The strength of the association between TCF7L2 and T2D makes TCF7L2 a candidate gene for the pharmacogenetic s of thiazid e induced NOD. To our knowledge, TCF7L2 has not been investigated with respect to pharmacogenetics of thiazide induced NOD. The purpose of th e research presented in this chapter is t o identify new variation in TCF7L2 define LD structure and investigat e TCF7L2 SNP associations with NOD in African and Hispanic populations We accomplished this using DNA samples from the INternational Verapamil SR Trandolapril STudy GENEtic Substudy ( INVEST GENES ) which compared CV outcomes and NOD following treatment w ith a CCB or beta blocker based antihypertensive treatment strategy in an ethnically diverse cohort of patients with hypertension and coronary artery disease (CAD). In addition, we
40 investigated the impact of TCF7L2 polymorphisms on the development of NOD by HCTZ treat ment in INVEST GENES. Methodology INVEST Study Design and Study Population INVEST randomized patients to either atenolol or verapamil sustained release (SR) based antihypertensive treatment strateg ies and followed patients for ad verse CV outco mes and NOD A total of 2 2,576 patients at least 50 years of age with hypertension and CAD were enrolled between September 1997 and February 2003 at 862 sites in 14 countries. All patients enrolled in INVEST provided written informed consent, and the ins titutional review boards of participating study centers approved the study protocol. The design, primary outcome, and NOD results have been previously published in detail 50 137 138 NOD was determined by site investigators from a review of all availab le patient data, including use of diabetic medication and available lab oratory data 50 Briefly, the CCB based strategy consisted of verapamil SR 240 mg da ily (Step 1), addition of trandolapril 2 mg daily (Step 2), dose titration to verapamil SR 240 mg / trandolapril 2 mg twice daily (Step 3), and HCTZ 25 mg daily add on treatment (Step 4) to achieve Sixth Report of the Joint National Committee on Prevention Detection, Evaluation, and Treatment of High Blood Pressure (JNC VI ) BP goals. 139 The beta blocker based stra tegy consisted of atenolol 50 mg daily (Step 1 ), addition of HCTZ 25 mg daily (Step 2), titration to atenolol 50 mg / HCTZ 25 mg twice daily (Step 3), and trandolapril 2 mg daily add on treatment (Step 4) for BP control as necessary. Both strategies were optimized to provide end organ protection with trandolapril in patients with diabetes and/or renal insufficiency.
41 INVEST GENES Study Design and Population A total of 5,979 genomic DNA samples were collected from participants in INVEST from June 2001 to Feb ruary 2003 from 184 sites in the US and Puerto Rico. INVEST GENES participants provided additional informed consent for genomic studies. The INVEST GENES population is similar to the entire INVEST and includes an ethnically diverse population with a larg e population of Hispanic individuals mostly from Puerto Rico INVEST genomic DNA was collected using buccal cells from mouthwash samples using commercially available kits (PureGene, Gentra Systems Inc orporated Minneapolis, M N ) as previously described 140 TCF7L2 S equencing A total of 150 INVEST GENES samples were sequenced for TCF7L2 including 25 white Hispanic and black NOD cases and controls Our sample size provided 99% power to detect polymorphisms of >5% frequency in the o verall population (n=150) and 95% power to detect such polymorphisms with in each race/ethnic group (n=50). Polymerase chain reaction (PCR) amplification, s equencing, and SNP discovery was performed by the National Heart, Lung, and Blood Institute (NHLBI) Resequencing and Genotyping (RS&G) service at the J Craig Venter Institute (JCVI) on whole genome amplified (WGA) INVEST GENES deoxyribonucleic acid ( DNA ) The target sequence region included the entire TCF7L2 gene, including 2 kb upstream and downstream of gene ends (219,874 base pairs total ) Amplicons w ere PCR amplified using T m matched primers designed from a masked reference sequence ( Human Genome NCBI36 Assembly) to exclude problematic regions Amplicons were and reverse sequencing primer. Sanger Big Dye Terminator sequencing and detection with capillary based sequencing machines was
42 used to determine DNA sequences o n forward and reverse strands to provide double stranded coverage TCF7L2 Polymorphism Disco very and Visual Display Sequences were base called and assembled on the reference sequence. Sequence data for each amplicon w ere used in SNP discovery only if amplicons were 75 % of samples. SNPs and small inser tion/deletion variants were computationally identified, confirmed on both forward and reverse strands, and chromatograms were manually reviewed by specialized data analysts at the JCVI A batch inquiry of the region sequenced was performed using the Natio nal Center for Biotechnology Information (NCBI) database for single nucleotide polymorphisms ( dbSNP [ www.ncbi.nlm.nih.gov ]) to determine previously identified polymorphisms. TCF7L2 SNP data w ere uploaded i nto th e publicly available dbSNP on September 21, 2010 under Batch ID TCF7L2 R122_ISDP1 20100921 with population data displayed under the population ID R122_ISDP1. Amplicons and polymorphisms were visually displayed in the University of California at Santa Cruz ( UCSC ) genome browser 141 Sequenced genotype data were also visually displayed using the Genome Variation Server. 142 In Silico Functional Predictio n of TCF7L2 Polymorphisms Polymorphisms were first tested for putative functional status. Polymorphisms were considered to have putative functional consequences if located in exons, the promoter region n factor binding sites (TFBS), predicted miRNA sites, regions of high mammalian conservation, or within known structural variants. Exonic splice enhancers (ESE) and exonic splice silencers (ESS) were also predicted. F unctional consequences of TCF7L2 poly morphisms were
43 determined in silico using SNPNexus and Pupasuite 143 144 Functional effects of radical amino acid changes were predicted for non synonymous polymorphisms using SIFT. 145 Race /Ethnicity and Linkage Disequilibrium Structure in Sequenced Samples INVEST race/ethnic groups were determined by patient report with interaction by the study investigator and confirmed through princip a l components analysis ( PCA) PCA was performed using JMP Genomics version 5.0 (SAS, Cary, NC) in which the EIGENSTRAT method was implemented. 146 PCA was performed on an LD pruned (r 2 >0.5) HumanCVD BeadChip data set (n=2,214 INVEST GENES participants and n=32,235 SNPs ) INVEST GENES included primarily white, Hispanic, and black race/ethnic groups If self reported race/ethnic category disagreed strongly with PCA clustering results patients wer e re categorized to reflect PCA which w as considered to better reflect genet ic ancestry. PCA was also performed with in each race/ethnic group for principal components ( PC s) adjustment in analyses performed by race/ethnicity PCA and race re categorization was performed by the UF Center for Pharmacogenomics. In addition, 87 ances try informative markers (AIMs) were genotyped on 149 (99%) sequenced individuals Greater than 65 loci were acquired in 135 (91%) sequenced samples and self reported and PCA derived race/ethnic groups were also visually confirmed with AIMs data using a te rnary plot. To determine LD structure of TCF7L2 in each race/ethnic group LD plots of pairwise r 2 for all sequenced SNPs were produced by race/ethnicity in Haploview version 4.2 147 Haplotype blocks were defined using the method proposed by Gabriel e t al. 148 Haplotype blocks were visually examined for degree of LD and number of haplotype blocks within each race/ethnic group. In addition, pairwise LD was d etermined for GWAS associated TCF7L2 SNPs, individually
44 genotyped T2D predictor SNPs, and putative functional SNPs within each race/ethnic group Identification of T2D Predictor SNP s from Sequenced Samples TCF7L2 polymorphisms were selected as candidate T 2D predictor for individual SNP genotyping for Hispanic and black race/ethnic groups in a larger case control cohort (described below) Candidate T2D predictor SNPs were identified in sequenced whites and compared to previous GWAS results in Europeans to validate SNP selection procedures for Hispanic and Black race/ethnic groups. No candidate SNPs in whites were genotyped in the larger case control since T2D predictor SNPs are known from GWAS in Europeans. Selection of candidate SNPs was based on NOD as sociation analysis described below, SNP call rate Hardy Weinberg Equilibrium ( HWE ) minor allele frequency ( MAF ) putative functional consequences, and pairwise LD. SNPs were not considered candidates for genotyping if they were in high LD (r 2 >0.8) with the GWAS index SNPs rs7903146 or rs12255372 or any SNP on the HumanCVD Beadchip, since HumanCVD Beadchip data had already been acquired. SNPs were also excluded as candidates if SNPs had insufficient call rates or higher than expected departure from HWE a nd heterozygosity. Variants were considered good candidates if they had a high frequency (MAF>0.10) in black or Hispanic race/ethnic groups or they were newly identified by the sequencing effort. Selection of candidate T2D predictor SNPs for individual SNP genotyping was based primarily on the strength of association (p value) and point estimate (odds ratio) in an unadjusted logistic regression in a dominant model by race/ethnicity The dominant model was used in the original T2D GWAS 91 and is appropriate considering
45 the limited power available with in each race/ethnic group of sequenced samples to detect allelic ORs Additionally, several NOD association analyse s were carried out by race/ethnicity in sequenced samples using a variety of other statistical models with the intent to determine T2D predictors that were robust to changes in statistical models. However, no correction for multiple comparisons was appli ed to SNP analyses in sequenced samples since sufficient power to detect associations by race/ethnicity after multiple comparisons adjustment was unlikely. Consistency of associations in various statistical models was assessed, indicated by the addition of the negative log of the p value for the four mod els described below to create a log(p) score. The validity of the unadjusted dominant model and the log(p) score to indicate candidate T2D predictor SNPs was confirmed in sequenced whites Sequenced samples were tested for association with NOD in logis tic regressions using both additive and dominant models, including 1) an unadjusted model, 2) a model adjusted for PCs one, two, and three, 3) a model adjusted for AIMs, and 4) a model adjusted for variables clinically associated with T2D. Clinical varia bles for adjustment were chosen based on previous association with NOD in INVEST 50 and included age, BMI, gender, average on treatment systolic BP, left ven tricular hypertrophy ( LVH ) hypercholesterolemia (defined as history of or currently taking lipid lowering medications) smoking history antihypertensive treatment strategy, and atenolol, HCTZ and trandolapril treatment and duration. HCTZ, atenolol, and trandolapril drug treatment and duration were used as covariates since they belong to drug classes known to affect NOD incidence. 47 Individual drug treatment was defined as any prescription before diagnosis of NOD during INVEST. Deviations from HWE
46 act Test in control samples by race/ethnicity. All statistical analyses were performed using SAS version 9.2 and JMP Genomic s version 5.0 (SAS, Cary, NC). SNP Genotyping in the INVEST GENES New Onset Diabetes Case Control Candidate T2D predictor SNPs iden tified using criteria stated above were individually genotyped in a larger case control cohort from INVEST GENES. We conducted a nested case control study in INVEST GENES including cases that developed NOD and age, race, and gender matched controls, who r emained free of T2D over a mean 2.8 years follow up. Patients taking anti diabetic medication or with diabetes history at baseline were excluded from the study G enomic DNA was normalized to 1 0 ng/ identified from seq uenced samples Genotyping in the NOD case control was accomplished using the TaqMan 7900HT real time PCR system (Applied Biosystems, Foster City, CA). The Taqman allelic discrimination assay was performed using a man Genotyping Master Mix (Applied using the following conditions: 95 C for 10 minutes, foll owed by 45 50 cycles of 92 C for 15 seconds and 58 62 C for one minute, followed by 4 C until fluorescent signal detection. A total of 101 additional TCF7L2 cosmopolitan tag SNPs, selected for r 2 >0.5 and MAF>0.05, were genotyped using the HumanCVD Bea dChip and Infinium II Assay (Illumina, San Diego, CA) at the University of Florida (UF) Center for
47 Pharmacogenomics The HumanCVD BeadChip contains approximately 50,000 cosmopolitan tag SNPs for 2,100 CV and metabolic related genes 149 Genotype calling was performed using GenomeStudio Software version 2011.1 and Genotyping Module version 1.9 calling algorithm (Illumina, San Diego, CA). Genotyping of sequenced samples and NOD case control samples using various platforms and related analyses are summarized in Table 2 1. For the HumanCVD BeadChip, i ndividual patient s were excluded if sample call rates were below 95% and SNPs were excluded if genotype call rates we re below 90%. Genotype data quality was further ensured in PLINK using concordance rates for 87 blind duplicates, gender confirmation using X chromosome genotype data, cryptic relatedness using pairwise identity by descent (IBD) and estimation of heteroz ygosity using the inbreeding coefficient F. 150 QC procedures involving HumanCVD BeadChip data were performed by the UF Center for Pharmacogenomics. Af ter QC, all v ariants were included in association analyses regardless of departure from HWE, but variants with low HWE p values were flagged and departure from HWE was considered in identification of candidate T2D predictor SNPs and interpretation of resul ts. Baseline Characteristic and NOD Association Analysis in the INVEST GENES NOD Case Control Cohort B aseline differences in patient characteristics between cases and controls were determined using t tests and chi square tests as appropriate. In the full NOD case control cohort, l ogistic regression was used to assess TCF7L2 SNP effects on NOD. ORs and 95%CIs were calculated using allelic trend tests. NOD association in the NOD case control cohort included candidate SNPs identified in sequenced samples an d TCF7L2 SNPs from the HumanCVD BeadChip. SNP association analyses for TCF7L2
48 SNPs were performed by race/ethnicity and adjusted for PCs one, two, and three t o minimize confounding by population stratification. PCs one, two and three provided sufficient separation of ancestry clusters in the NOD case control and were chosen as covariates in NOD case control analys e s to adjust for ancestry. Logistic regressions were also adjusted for clinical covariates as stated above. The false discovery rate (FDR) met hod was used to correct for multiple comparisons in association tests in the NOD case control. A ssuming an OR of 1.7 and MAF of 0.3, t he full NOD case control cohort provided 87% power in Whites, 36% power in Blacks, and 87% power in Hispanics to detect a dditive effects of TCF7L2 polymorphisms on T2D. Deviations from Pharmacogenetic Analysis in the INVEST GENES NOD Case Control Cohort We tested pharmacogenetic effects of T CF7L2 SNPs, including HumanCVD BeadChip SNPs and the three TCF7L2 candidate SNP s, on thiazide induced NOD. To test for pharmacogenetic effects, we calculated ORs and 95%CIs for NOD in HCTZ treated and non HCTZ treated individuals and tested for HCTZ SNP i nteractions on NOD in logistic regression models. HCTZ treatment was defined as any HCTZ prescription before new diagnosis of T2D. We also performed sensitivity analyses by test ing for pharmacogenetic interactions with alternative HCTZ treatment defin iti ons, including continuous HCTZ and daily dose of HCTZ 25mg ORs and 95% CI s in pharmacogenetic analyses were adjusted for variables clinically associated with T2D as stated above and PCs one, two, and three. Interaction p valu es in pharmacogenetic analyses were corrected for multiple comparisons using the FDR method. 151 All statistical analyses were performed using SAS version 9.2 and JM P Genomic s version 5.0 (SAS, Cary, NC).
49 Results Sequence and Genotype Data Quality Control in Sequenced Samples A total of 156,000 of 219,874 base pairs (70.95%) were successfully sequenced by the JCVI on both forward and reverse strands in the target regi on in TCF7L2 A total of 329 of 552 amplicons (60%) passed QC procedures, as they were successfully sequenc ed on both strands in at least 75 % of samples Of successfully sequenced amplicons, there was a n average 92.8% double stranded success and a 97.7% PCR success. Polymorphisms were not observed in s even of 329 (2%) reported amplicons. When a double stranded amplicon sequence success cutoff of 75% was implemented, 1,78 4 polymorphisms in the target TCF7L2 region were observed. When an amplicon success cutoff of 80% was implemented, 1,764 polymorphisms were identifi ed, suggesting validity for the vast majority of polymorphic sites despite more stringent QC cutoffs. Notable departures from HWE among sequenced SNPs included rs290490 in Whites (p=4.6x10 7 ) and rs7919152 in Hispanics (p=2.6x10 8 ). Ninety percent of sequenced samples (135/150) were successfully genotyped on the HumanCVD BeadChip after QC. Characteristics of Sequenced TCF7L2 Variation We observed 1,78 4 variant s in the target TCF7L2 region including 1, 701 SNPs with overall MAF s ranging from 0.003 0.492 ( Figure 2 1 ) A total of 666 variants were identified in whites, 951 variants were identified in Hispanics, and 1,017 variants were identified in blacks. Non SNP variation included 26 dele tions, 8 insertion deletion polymorphisms, and 12 mixed polymorphisms. Of the 1,78 4 polymorphic sites, 910 were private mutations, occurring as heterozygous in a single individual, and 1,110 polymorphisms were considered rare v ariants (MAF<0.01).
50 A total of 91 0 SNPs were considered novel at the time of SNP data upload in dbSNP (September 21, 2010) including 765 novel SNPs in blacks, 683 in Hispanics, and 420 in whites ( dbSNP Population ID: R122_ISDP1; Batch ID: TCF7L2 R122_ISDP1 20100921 ) A total of 197 polymorphisms had been previously identifi ed in dbSNP and w ere validated by our sequencing effort A total of 733 p olymorphisms identified remain currently un validated in dbSNP identified solely by our sequencing effort in the INVEST GENES population A total of 853 SNPs had been previously validate d by Hapmap and the 1000 Genomes project 100 152 In Silico F unctional Prediction of Sequenced TCF7L2 Variants Four polymorphisms were identified in the TCF7L2 promoter region. (Table 2 2) Two novel synonymous SNPs were observed in exons 6 and 15 (rs146872546 and rs142903496 respectively) Four non synonymous SNPs (rs148523217, rs148050954, rs147841431, and rs77673441) were observed in Exon 15 of TCF7L2 Two non synonymous SNPs, including one novel (rs148523217) and one previously identified SNP (rs77673441) were predicted to be protein damaging mutations by the S IFT program with a SIFT score less than 0.05. 145 One previously identified SNP (rs1056877) w as Fifty two novel and nine previously identified SNPs were found in conserved TFBSs predicted in silico The remaining 1,767 sequenced mutations were identified in TCF7L2 introns and were not predicted in silico to have functional consequences TCF7L2 S NPs with putative functional consequences were found primarily in black and Hispanic sequenced individuals. LD Structure of Sequenced TCF7L2 Variants TCF7L2 LD structure was similar among individuals of European and Hispanic descent, in which the large L D block in intron 1 containing T2D associated SNPs from
51 GWAS, characteristic of European individuals, can be observed. (Figure 2 2 ) The characteristic European LD block, which includes the GWAS associated variant rs7903146, is not evident in sequenced ind ividuals with African ancestry. The greater number of identified SNPs in b lack s and Hispanics indicat ed greater genetic diversity in TCF7L2 among African and Hispanic populations. The distribution of polymorphisms within and between race/ethnic groups is represented in Figure 2 3 Baseline Characteristics and PCA of Sequenced Samples Characteristi cs of cases and controls from sequenced samples are summarized in Table 2 3 In sequenced samples, differences in baseline characteristics between cases and con trols included a higher mean diastolic BP (88 versus 84 mmHg, p=0.02) and BMI (30.9 versus 28.9 kg/m 2 p=0.01). PCA analysis indicated that patient reported race/ethnicity was cons istent with PCA clustering in sequenced individuals. (Figure 2 4 ) Two sequ enced samples which had self reported white race were re categorized as Hispanic based on PCA results for association anal yse s Validation of Candidate T2D Predictor SNP Selection in White Sequenced Sampl es The rs7903146 SNP was most highly associated wi th NOD in white sequenced samples using an unadjusted dominant model (OR 4.50 [1.17 17.37], p=0.03) suggesting the validity of this model in determining T2D predictor SNPs from sequence data (Table 2 4 ) The rs7903146 SNP is the SNP most strongly and rep roducibly variant associated with T2D from GWAS. 95 After ranking of candidate SNPs in whites based on the log(p) score from a dominant model, the rs7903146 SNP was the most highly associated SNP in sequenced whites using multiple statistical models, suggesting that the log(p) score is also effective in identifying candidate T2D predictor
52 SNPs despite limited power in sequenced samples by race/ethnicity. In addi tion, the top three SNPs identified in whites based on the log(p) score have also been associated with T2D in studies in Europeans. 95 153 (Table 2 4 ) Candidate T2D Predictor SNP Identification in Hispanic Sequenced Samples In Hispanics, no novel SNPs had sufficient MAF in sequenced Hispanics to be considered candidate SNPs for further genotyping in the INVES T GENES NOD case control. However the SNPs rs7895307 and rs7901275 showed association with NOD in sequenced Hispanics and were chosen as candidate SNPs. To our knowledge, t he SNPs rs7895307 and rs7901275 have not been previously associated with T2D risk The SNPs rs7895307 and rs7901275 showed ass ociation with NOD in Hispanic sequenced samples in an unadjusted dominant model (rs7895307 OR 7.50 [ 95%CI 2.04 27.59], p=0.002 and rs7901275 OR 4.22 [ 95%CI 1.32 13.47], p=0.02). (Table 2 5 ) In addition, rs7895 307 and rs7901275 SNPs had high log(p) scores in sequenced Hispanics, indicating consistent associations across statistical models. The SNPs rs7895307 and rs7901275 also had high MAFs, did not depart from HWE had call rates above 75% and were not in hi gh LD with T2D SNPs from GWAS or SNPs on the HumanCVD BeadChip. (Tabl e 2 5 ) Both SNPs a re located upstream of the LD block containing the GWAS ind ex SNP rs7903146 in Hispanics. Candidate T2D Predictor SNP Identification in Black Sequenced Samples No nov el SNPs with sufficient MAF showed association with NOD in sequenced Blacks. In Blacks, rs74159629 was chosen as a candidate T2D predictor SNP. Th e SNP rs74159629 was recently validated by the pilot phase of the 1000 Genomes and i s thereby unlikely to be represented on currently available GWAS genotyping arrays. The SNP rs74159629 has not been previously associated with T2D risk. The rs74159629
53 SNP was associated with NOD in Blacks in an unadjusted dominant model (OR 0.09 [ 0.01 0.85 ], p=0.04) with a MAF of 0.14 (Table 2 5 ) Rs74159629 is located upstream of rs7903146, but is not located in any haplotype blocks in sequenced Blacks Baseline Characteristics and PCA for the INVEST GENES NOD Case Control For the NOD case control, we identified 446 NOD ca ses in INVEST GENES over a mean 2.8 years. (Table 2 6 ) As in sequenced samples a t baseline, NOD cases had higher BMI (31 versus 29 kg/m 2 p<0.0001) and higher diastolic BP (87 versus 86 mmHg, p=0.04) than controls in the full NOD case control cohort. A h igher percentage of LVH was also observed in cases compared to age, race/ethnicity and gender matched controls (17% versus 13%, p=0.02) During follow up the proportion of patients who were treated with HCTZ was higher in cases than in controls (74% ver sus 62%, p<0.0001). An association between HCTZ treatment and NOD was observed in the NOD case control (OR 1.70 [95%CI 1.31 2.20], p<0.0001). Consistent associations were observed for HCTZ and NOD in white patients (OR 1.97 [95%CI 1.32 2.95], p=0.001), i n black patients (OR 2.85 [95%CI 1.11 7.30], p=0.03), and trended towards association in Hispanic patients (OR 1.37 [95%CI 0.96 1.95], p=0.08). The average duration of HCTZ treatment was 87.1 (SD 73.6) weeks and the average dose was 28mg daily in diabetic cases and 26mg daily in controls (p=0.17) A higher mean atenolol dose was also observed in NOD cases (72 mg versus 67 mg, p=0.02) among patients treated with atenolol NOD Association in White INVEST GENES NOD Case Control Patients All candidate T2D p redictor SNPs genotyped with Taqman had SNP call rates greater than 95% and no departure from HWE was observed in candidate SNP in any race/ethnic group. Fourteen of the 101 TCF7L2 SNPs on the HumanCVD BeadChip
54 were out of HWE (p<0.05) in at least one rac e/ethnic group, but the SNPs deviating from HWE were not significantly associated with NOD. Nine SNPs were observed to be nominally associated with NOD in whites after analysis of HumanCVD BeadChip in the NOD case control cohort, including eight SNPs that were nominally associated with an increased NOD risk and one SNP associated with a decreased NOD risk. However, none of these associations remained significant after multiple comparisons adjustment. The strongest association observed in whites was rs381 4572 (OR 1.74 [95%CI 1.15 2.64], p=0.01), (Figure 2 5) although rs3814572 has not been previously associated in T2D genetic association studies. Although we observed a strong association between NOD and rs7903146 in sequenced samples, we observed showed n o association in the full NOD case control in whites (OR 1.00 [95%CI 0.71 1.40], p=0.97). NOD Association in Hispanic INVEST GENES NOD Case Control Patients After individual genotyping of the candidate T2D predictor SNPs rs7895307 and rs7901275 in the NO D case control cohort, neither SNP departed from HWE. The candidate T2D predictor SNP rs7895307 was not associated with NOD in Hispanics in a dominant or additive model, before or after covariate adjustment (OR 1.03 [95%CI 0.78 1.35], p=0.85). (Table 2 7) The SNP rs7901275 also was not associated with NOD in Hispanics (OR 1 .01 [95%CI 0.76 1.34], p=0.94). Five SNPs were observed to be associated with NOD using HumanCVD BeadChip in the full NOD case control cohort, including two SNPs that were nominally ass ociated with an increased NOD risk and three SNPs associated with a decreased NOD risk. (Figure 2 5 ) None of these associations remained significant after multiple comparisons adjustment. The strongest associations observed in Hispanics were rs7089262 (O R 0.32 [95%CI 0.15 0.70], p=0.02) and rs7899644 (OR 0.32 [95%CI 0.15
55 0.70], p=0.02) Interestingly, t he SNP rs11196213, which has been previously associated with T2D in Europeans, 154 showed an increased NOD risk in Hispanics (OR 1.32 [95%CI 1.01 1.72], p=0.04) and a decreased NOD risk in whites (OR 0.67 [95%CI 0.49 0.92], p=0.02). The GWAS SNP rs7903146 showed no association with NOD in Hispanics (OR 0.95 [95%CI 0.71 1.27], p=0.75). NOD Association in Black INVEST GENES NOD Case Control Patients After genotyping in the NOD case control, the candidate SNP rs74159629 did not depart from HWE. (Table 2 7) Rs74159629 showed a nominally significant association with dec reased NOD risk in Blacks (OR 0.28 [95%CI 0.08 1.03], p=0.05) after covariate adjustment, but was not associated in an unadjusted model (OR 0.63 [95%CI 0.24 1.67], p=0.35). Interestingly, the rs74159629 SNP also showed a trend towards association with NOD in Hispanics, although the direction of the point estimate in Hispanics was opposite of that observed in Blacks (2.56 [0.96 6.84], p=0.06). The candidate SNPs we identified for Hispanics showed no association with NOD in Blacks ( rs7895307 OR 1.47 [95%CI 0.77 2.82], p=0.25 and rs7901275 OR 1.77 [95%CI 0.87 3.59], p=0. 12 ) although the direction of point estimates was similar (Table 2 7) Four SNPs were nominally associated with NOD using HumanCVD BeadChip in blacks, including two SNPs (rs 10885410 and rs6585 202 ) that were associated with increased NOD risk and two SNPs ( rs7896091 and rs 12573128 ) associated with a decreased NOD risk. (Figure 2 5 ) The strongest association observed in blacks was rs12573128 (OR 0.48 [95%CI 0.26 0.89], p=0.02) after adjustment f or covariates None of these associations remained significant after multiple comparisons adjustment. Interestingly, rs12573128 which was associated with an increased NOD risk in whites (OR 1.55 [95%CI 1.00 2.39], p=0.04) was associated with a decrease d NOD risk in
56 blacks (OR 0.5 [95%CI 0.27 0.91], p=0.02). The SNP rs12573128 h as been previously associated with insulin sensitivity and glucose tolerance during an OGTT. 155 The GWAS SNP rs7903146 showed no association with NOD in blacks (OR 0.91 [95%CI 0.48 1.72], p=0.74). TCF7L2 SNP *HCTZ Treatment Pharmacogenetic Interactions No SNPs with significant SNP*HCTZ interactions in pharmacogenetic analyses deviating f rom HWE. I n whites, nine out of 10 1 TCF7L2 SNPs showed a significant pharmacogenetic interac tion with hydrochlorothiazide treatment on NO D after FDR correction. Seven of the SNPs with significant pharmacogenetic interactions showed an increased risk of NO D in HCTZ treated patients and two showed a decreased risk of NOD in non HCTZ treated patients. (Table 2 8) The strongest interaction was observed for the rs7917983 T allele (p inx =3.7x10 4 p FDR =0.02) with a significantly increased risk of NOD in HCTZ tre ated patients (OR 1.50 [95%CI 1.00 2.25]) and a decreased risk of NOD in non HCTZ treated patients (OR 0.47 [95%CI 0.26 0.84]). To our knowledge, n o association between rs7917983 and T2D has been previously reported. An additional 15 TCF7L2 SNPs showed n ominally significant pharmacogenetic interactions. Interestingly, TCF7L2 SNPs that had been previously associated with T2D in GWAS, including rs7901695 and rs4506565, 92 153 showed significant pharmacogenetic associations with an increased risk of NOD with HCTZ with T2D risk alleles and decreasing NOD with T2D risk allele s in non HCTZ treated individuals. The SNP rs 11196228, which has also been associated with T2D, 154 also showed a significant interaction with a decreased risk of NOD with HCTZ with the T2D risk allele and in creasing NOD with the T2D risk allele in non HCTZ treated individuals. The strongest GWAS SNP from the literature, rs7903146, 95 showed similar trends although on ly
57 nominally significant (p inx =0.01, p FDR =0.09). The SNPs rs12243326 and rs11196213, which have been previously associated with two hour post OGTT glucose and T2D respectively, 154 156 also showed nominally significant pharmacogenetic interactions. Results were similar when HCTZ and daily HCTZ dose 25mg No significant pharma cogenetic associations were observed in Hispanic s or blacks after FDR correction. In blacks, a trend toward a significant SNP*HCTZ treatment interaction was observed for the rs290490 G allele (p inx =0.005 [p FDR =0.09]), with an increased risk for NOD in HCT Z treated patients (OR 1.49 [95%CI 1.06 2.10] p=0.02) D iscussion I n the present study, we observed a large amount of previously unreported variation in TCF7L2 in our population, including 910 novel TCF7L2 variants and many novel variants w ith potential f unctional significance. The present study adds to existing literature by characterizing novel variation in TCF7L2 in populations of African and Hispanic descent, who have high T2D prevalence. 1 We observed several TCF7L2 SNPs that were nominally associated with NOD in each race/ethnic group, but were not associated after correction for multiple comparisons. Nominally significant associations with NOD indicated larg e differences in point estimates between race/ethnic groups and support the need for further research in Hispanic and black race/ethnic groups Significant pharmacogenetic interactions between TCF7L2 SNPs and HCTZ treatment on the development of NOD were observed in white s. Previously associated TCF7L2 SNPs from T2D GWAS were well represented among TCF7L2 pharmacogenetic interactions from the present study TCF7L2 pharmacogenetic predictors were observed when HCTZ treatment was defined as any exposure or continuous treatment
58 for an extended duration. Our study implicates TCF7L2 in thiazide associated NOD and provides evidence for TCF7L2 as a candidate gene for the pharmacogenetics of thiazide induced NOD. Since the initial T2D GWAS in 2006, 91 many T2D GWAS have been published that establish TCF7L2 SNPs as the strongest and most reproducible genetic risk factors for T2D. 92 95 99 101 103 153 Although GWAS have identified SNPs in TCF7L2 as consistent risk factors for T2D in European populations, TCF7L2 SNPs are not as well studied in African and Hispanic populati ons. In the present study, we sequenced an ethnically diverse group of individuals in order to identify new variation better define LD structure, and identify SNP s in TCF7L2 associated with T2D risk in Hispanic and African race/ethnic groups. The additi on of novel TCF7L2 SNPs, including several predicted to have functional consequences, to publicly available databases should improve our understanding of LD structure of the gene in populations of Hispanic and African descent and improve our ability to per form fine mapping of GWAS signals in TCF7L2 Our s equenc ing of TCF7L2 in individuals with European, Hispanic, and African descent provides identification of detailed variant information on these populations which was not available in dbSNP or deep sequenc ing efforts in the 1000 Genomes Project. 100 Another benefit of our deep sequencing in a diverse population is the ability to find causative alleles through ex amination of differences in LD and in associations between race/ethnic groups. Due to insufficient frequency of functional SNPs and insufficient sample size, establish ment of causative alleles solely responsible for the association between TCF7L2 SNPs and T2D was unlikely in the present study
59 We were also unable to identify any novel polymorphisms as potential T2D predictor SNPs in Hispanics and blacks The lack of novel T2D predictor SNPs may be part ly due to the fact that the vast majority of novel v ariation observed was of insufficient frequency to observe statistically significant associations with T2D. We observed strong associations in sequenced samples for three SNPs, but these SNPs did not show significant association in their respective race/e thnic groups in the larger NOD case control cohort. Strong associations in sequenced samples by race/ethnic group may have been due to false positive associations in the relatively small sequenced population. Differences in observed associations between sequenced samples and the full NOD case control might also be explained by differences in drug treatment in each cohort. Whereas HCTZ was associated with NOD in the larger case control, no difference in HCTZ treatment was observed between cases and contro ls in sequenced samples. Differences in HCTZ treatment between cohorts may have affected TCF7L2 associations as significant SNP*HCTZ treatment pharmacogenetic interactions were observed in our INVEST population The lack of association between rs7903146 and NOD could also be due to a unique genetic architecture of the INVEST population. A distinct genetic architecture in INVEST is supported by the fact that the L D block containing the GWAS associated variant rs7903146 characteristic of individuals with European ancestry, wa s not evident in sequenced individuals with African ancestry Th e absence of th e characteristic LD block possibly explain s the lack of association of rs7903146 in individuals with African ancestry but not in whites or Hispanics Ho wever, the rs7903146 SNP was not significantly associated with NOD in INVEST whites either. A possible explanation for
60 the lack of association is a confounding effect of CAD, which was present in all INVEST patients and has been previously associated with T2D. 157 The lack of significant association in whites may be due to a modest sample size in this race/ethnic group compared to previously published association studies, which typically investigate rs7903146 associations in over 1,000 cases and controls. 91 99 101 103 153 Such a limitation contrasts with our ability to identify a strong rs7903146 association in sequenced samples with only 23 NOD cases and 25 controls and supports the validity of our T2D SNP predictor selection approach when limited sample sizes are available. The implementation of an unadjusted dominant model and observation of consistent asso ciations in various statistical models using a log(p) score did identify the most reproducible TCF7L2 association from the literature despite a very limited sample size in sequenced whites. Despite the apparent validity of this procedure, we did not obse rve significant effects of identified T2D predictor SNPs in individuals of Hispanic and African descent. Our lack of association is potentially explained by a limited sample size of sequenced individuals within each race/ethnic group or limitations of our statistical approach in Hispanics and blacks. The lack of association may also reflect limitations of currently available statistical methods to determine causative alleles for complex traits. The lack of association suggests limitations of the common d isease common variant hypothesis, which hypothesizes that the heritability of common diseases can be explained by several common alleles with large effects sizes. The common disease common variant hypothesis is being increasingly rejected in favor of cumu lative contribution of rare alleles to the heritability of complex disease. 115 Analysis involving rare variants is an alternative statistical approach that m ight yield meaningful
61 results, although currently available rare variant analysis techniques are subject to similar limitations in populations with limited sample size. Further investigation of potential T2D predictor SNPs is warranted in populations of A frican and Hispanic descent. Despite the lack of significant associations between TCF7L2 SNPs and NOD after FDR correction, we identified many SNPs nominally associated with NOD, which we re observed to have striking differences between race/ethnic groups. The SNP rs12573128, which has been previously associated with insulin sensitivity and glucose tolerance during an OGTT, 155 was nominally associated with an incr eased NOD risk in whites and a decreased NOD risk in blacks. The SNP rs11196213, which has been associated with T2D without genome wide significance, 154 was nomi nally associated with decreased NOD risk in whites, but increased NOD risk in Hispanics. Differences in NOD associations between whites, Hispanics, and blacks suggest that TCF7L2 SNPs associated in European populations may not be appropriate for more dive rse racial/ethnic groups. F urther investigation of TCF7L2 SNP associations with T2D is warranted in sufficiently powered populations of Africa n and Hispanics descent T CF7L2 is a transcription factor involved in the WNT signaling pathway. TCF7L2 has be en implicated in incretin signaling pathways since it has been shown to regulate transcription of the glucagon gene, which encodes glucagon like peptide 1 (GLP1) in the L cells of the gut. 150 The rs7903146 T allele has been associated with increased TCF7L2 expression and has been implicated as a functional variant, being mapped to open chromatin sites in pancreatic islet cells. 158 159 Whether rs7903146 is a functional variant is unclear Less robust associations in non Europeans would imply that
62 rs7903146 is not functional and GWAS associated TCF7L2 SNPs may be in LD with yet undetermined functional variants that influence amino acid sequence s, mRNA expression or mRNA stability Functional impairment of TCF7L2 by SNPs could cause changes in gene expressi on and influence T2D development. However, the results of th e present study suggest that rs7903146 is not functional, since a consistent association with T2D was not seen across multiple race/ethnic groups. Despite a lack of associations with T2D, we dete cted pharmacogenetic interactions between TCF7L2 SNPs and HCTZ treatment on NOD risk We found significant pharmacogenetic associations in whites, which is the race group in which the strongest and most reproducible TCF7L2 disease associations are found. Furthermore, several of the TCF7L2 SNPs showing a pharmacogenetic interaction are significantly associated with T2D in previous GWAS, including rs7901695 and rs4506565. 9 2 153 However, the strongest TCF7L2 SNP from the literature, rs7903146, 95 showed only a nominal pharmaco genetic association, although the directions of point estimates by HCTZ treatment groups were consistent with rs7901695 and rs4506565. A significant pharmacogenetic interaction was observed for rs11196228, which was associated with T2D, although this stud y was not a GWAS. 154 Among the SNPs with nominally significant pharmacogenetic interactions was rs12243326, which was previously associated with two hour glucose after glucose challenge. 156 The strongest pharmacogenetic interaction observed was for rs7917983, which has not bee n previously associated with T2D. To our kn owledge, th e present study is the first to investigate TCF7L2 SNPs for pharmacogenetic influences on thiazide induced NOD
63 and the first to observe significant pharmacogenetic interactions between TCF7L2 SNPs and HCTZ treatment on NOD risk Our results sugg est that TCF7L2 variation affects the influence of HCTZ on the incidence of NOD. HCTZ might affect expression of TCF7L2 or the ability of the T he majority of SNPs associated with NOD during HCTZ treatment were primarily intronic with few p redicted functional consequences. Furthermore, rs7903146 did not show the strongest pharmacogenetic interaction, suggesting that other SNPs may be responsible for the observed ph armacogenetic association D ifferences in associations between race /ethnic groups suggest differences in LD and the need to i d entify functional variants Such differences in LD between race/ethnic groups are obser vable in the INVEST population in LD anal yses Since the role TCF7L2 and its SNPs plays in T2D development remains unclear, it is difficult to speculate on the physiology of a pharmacogenetic interaction. TCF7L2 is primarily thought of as a T2D genetic risk factor with little or no role in the pharmacology of HCTZ However, HCTZ tr eatment may precipitate T2D in a patient who is otherwise at risk for T2D development based on their TCF7L2 genotype. If T2D GWAS SNPs from TCF7L2 could be considered pharmacogenetic risk factors many of the T2D in dex SNPs from GWAS (Table 1 1) may be pharmacogenetic risk factors as well To our knowledge, p ublished research investigating th e pharmacogenetic impact of SNPs from T2D GWAS on thiazide induced NOD is not currently available. The primary strength of th e presented research is the ethnic diversity of our INVEST population. The diversity in INVEST enables us to investigate differing disease
64 genetic and pharmacogenetic associations among different race/ethnic groups which expands the generalizability of ou r results and increases our ability to discern functional consequences of genetic variation In addition, the availability of sequence data in the entire TCF7L2 gene enables us to investigate variation and LD structure among European, Hispanic, and Africa n populations. We also have detailed clinic, outcomes, and drug exposure data for each member of our NOD case control cohort, enabling us to detect pharmacogenetic interactions, build multivariable logistic regressions, and adjust for potentially confound ing variables. Our study has several limitations worthy of mention. One limitation is the small number of NOD cases in our cohort when divided by race/ethnicity. Our sample size by race/ethnicity is small relative to most disease genetics studies inves tigating T2D. However, TCF7L2 is the strongest known T2D risk factor and has been observ ed in smaller disease genetics studies. In light of our sample size, we must conclude that our lack of observed associations may be due to limited sample size, especi ally in sequenced samples Another limitation of the present study is the possibility of alpha error, which is increased by the number of variants uncovered in the sequencing project. Alpha error no s ignificant associations were found with the exception of pharmacogenetic analyses. Significant pharmacogenetic associations after an FDR correction and multiple pharmacogenetic TCF7 L 2 SNP associations lend credence to the validity of pharmacogenetic findin gs. We do recognize the potential for false positive results in pharmacogenetic analyses and our observations need to be independently replicated.
65 I n addition antihypertensive therapy in INVEST may confound TCF7L2 SNP associations with NO D. TCF7L2 poly morphisms were tested for associations with NOD, which may have been affected by antihypertensive treatment or other environmental factors. To minimize confounding by antihypertensive treatment we adjus ted statistical models for treatme nt and duration of certain antihypertensive medications. HCTZ atenolol, and trandolapril have been shown to affect T2D risk and were all used in INVEST antihypertensive treatment strategie s. 47 50 51 We adjusted for atenolol and trandolapril treatment and duration in order to minimize confounding of antihypertensive pharma cotherapy in pharmacogenetic analyses. Confounding due to other environmental factors known to affect T2D incidence was also addressed through statistical adjustment of clinical characteristics previously associated with NOD. Summary and Significance In summary, our results add previously unknown polymorphisms to available data on TCF7L2 variation and further describe LD structure particularly in populations of Hispanic and African descent To our knowledge, th e described sequencing project represents the most comprehensive evaluation of TCF7L2 variation in black and Hispanic (derived primarily from Puerto Rican ) individuals that is currently available. We observed several TCF7L2 SNPs that were nominally associated with NOD in each race/ethnic group, but were not associated after correction for multiple comparisons. Nominally significant associations with NOD indicated large differences in point estimates between race/ethnic groups, which supports that TCF7L2 SNPs associated with T2D in Europeans may differ in populations of African and Hispanic descent. Further research of TCF7L2 SNP associations with T2D in individuals with African and Hispanic descent is warranted in populations with sufficient samples sizes.
66 In addition, o ur results also suggest that genetic variation in TCF7L2 particularly SNPs associated with T2D from GWAS, may influence the effect of HCTZ on NOD risk Our observation of significant SNP effects only in White s suggests differences in LD or that TCF7L2 SNPs have pharmacogenetic influences only in individuals with European ancestry Functional studies and replication of pharmacogenetic associations are needed to confirm our observed association and define the potential role of TCF7L2 SNPs in predicting NOD during HCTZ treatment TCF7L2 is a compelling candidate gene in the pharmacogenetic study of thiazide induced dysglycemia.
67 Table 2 1. Study populations and study design for TCF7L2 SNP discovery, LD characterization, and statistical analyses INVEST GENES Cohort Platform for da ta acquisition Analyses p erformed S equenced samples (n=150) Sanger Sequencing ( 1,7 84 variant s ) SNP discovery LD characterization Determination of T2D predictor SNPs HumanCVD BeadChip (101 SNPs) Comparison of PCA and self reported race LD of novel SN Ps and BeadChip SNPs NOD Case Control (n=1,435) Taqman (3 SNPs) T2D predictor SNP association with T2D HCTZ pharmacogen e tic associations HumanCVD BeadChip (101 SNPs) Association with T2D HCTZ pharmacogenetic associations Principal components ana lysis LD indicates linkage disequilibrium; NOD indicates new onset diabetes; SNP, single nucleotide polymorphism; T2D, type 2 diabetes Table 2 2 Strongest putative functional varian ts from TCF7L2 sequence data determined in silico rs number Novel* Posit ion Region Alleles AA Change MAF Race/ethnic group(s) Â§ rs10885396 114701745 Promoter G/A 0.344 B H, W rs138659283 Yes 114701766 Promoter TCTC/ 0.003 B rs140632597 Yes 114701867 Promoter A/G 0.003 B rs10885397 114701873 Promoter G/A 0. 114 B, H, W rs138272435 Yes 114701897 Promoter G/A 0.003 B rs146872546 Yes 114890959 Exon 6 C/T 0.004 B rs148523217 Yes 114895779 Exon 15 C/A P247T** 0.004 H rs142903496 Yes 114902087 Exon 15 T/C 0.004 W rs148050954 Yes 114910413 Exon 15 G/A R47 2Q 0.004 H rs77673441 114915359 Exon 15 C/A P477T** 0.014 W rs147841431 Yes 114915665 Exon 15 T/C S579P 0.010 B rs1056877 114915748 T/C 0.213 B, H W AA indicates amino acid; MAF, minor allele frequency; B, blacks; H, Hispanics; W, whites; U TR, untranslated region Polymorphism not described in dbSNP before s ubmission of sequence data P osition on chromosome 10 in human genome Build 3 7 (GRCh37) Â§ Race/ethnic group in which the polymorphism was identified ** P redicted to be protein damaging amino acid substitutions by SIFT
68 Table 2 3 C haracteristics o f new onset diabetes c ases and c ontrols at baseline in INVEST sequenced samples Characteristic Diabetes Cases (n=73) Controls (n=77) p value At b aseline Age (years) 64.0 (10) 64.1 (9) 0.96 Female n (%) 39 (54%) 42 (55%) 0.89 BMI (mean) 30.9 (5) 28.9 (5) 0.01 Race/Ethnicit y 0.74 Black n (%) 22 (54%) 19 (46%) Hispanic n (%) 28 (46%) 33 (54%) White n (%) 23 (48%) 25 (52%) Blood pressure (mmHg) Systolic 151 (18) 147 (18) 0.14 Diastolic 88 (10) 84 (10) 0.02 Hypercholesterolemia n (%) 35 (48%) 47 (61%) 0.11 History of LVH n (%) 11 (15%) 9 (12%) 0.54 History of MI, n (%) 13 (18) 14 (9) 0. 95 History of s moking n (%) 27 (37 %) 35 (45%) 0.29 During INVEST Verapamil SR s trategy n (%) 41 ( 56 %) 4 3 ( 56 %) 0.97 Atenolol treatment n (%) 30 (41%) 33 (43 %) 0.83 HCTZ treatment n (%) 51 (70 %) 53 (69%) 0.89 Trandolapril treatment n ( %) 45 (62 %) 52 (66%) 0.45 Verapamil SR treatment, n (%) 41 (56%) 43 (56%) 0.97 INVEST indicates INternational VErapamil SR and Trandolapril Study; BMI body mass index; mmHg, millimeters of mercury; LVH, left ventricular hypertrophy; SR, sustained releas e; HCTZ hydrochlorothiazide Values are mean standard deviation unless otherwise noted. for t test or chi square test where appropriate lowering medications
69 Table 2 4 Validation of c andidate SNP method in INVEST sequenced whites SNP MAF Call Rate HWE log(p) score (dominant) log(p) score (additive) Odds ratio (dominant) rs7903146 0.33 0.83 0.37 6.21 3.15 4. 50 (1.17 17.36) rs4506565 0.30 0.92 0.43 5.64 3.05 3.11 (0.91 10.69) rs35198068 0.32 0.8 5 0.33 5.57 3.58 3.79 (1.03 13.91) rs146426147 Â§ <0.01 1.00 5.20 3.58 3.37 (0.87 13.12) rs2296783 0.10 0.94 0.83 5.17 5.17 5.50 (1.00 30.36) rs4132670 0.35 1.00 0.97 5.06 3.08 3.07 (0.91 10.37) rs140658764 Â§ <0.01 1.00 4.85 4.23 2.83 (0.80 10.04) rs 7901695 0.33 1.00 0.19 4.80 2.49 2.48 (0.76 8.10) SNP indicates single nucleotide polymorphism; INVEST INternational VErapamil SR and Trandolapril Study; MAF, minor allele frequency; HWE, Hard y Weinberg Equilibrium p value SNPs are r anked in order of log(p) score from the dominant model D etermined by the sum of four negative log(p values) for four d escribed statistical models S ignificantly associated with T2D in at least one publication Â§ N ovel at t ime of sequence data delivery
70 Table 2 5 Identif ication of candidate SNPs by race/ethnic groups in INVEST sequenced Hispanic s and blacks* SNP MAF GWAS LD ( r 2 ) Chip LD ( r 2 ) Call Rate HWE log(p) score (dom ) Â§ log(p) score (add ) Â§ Odds ratio ** (dominant model ) Hispanics r s7895307 Â§Â§ 0.34 0.43 0.31 0.79 0.71 7.98 7.74 7.5 (2.0 27.6) rs73358278 0.02 0.05 0.15 0.95 6.54 4.20 5.1 (1.4 19.0) rs116449537 0.02 0.06 0.15 0.90 0.93 6.41 4.54 4.9 (1.3 18.4) rs150064718 0.03 0.01 0.74 0.98 0.78 6.21 6.21 0.3 (0.1 0.97) rs7901275 Â§Â§ 0.41 0.14 0. 23 0.92 0.98 6.04 7.87 4.2 (1.3 13.5) rs7895340 0.34 0.35 0.67 0.82 0.38 6.14 4.91 0.2 (0.1 0.7) rs12243578 0.26 0.22 0.65 0.75 0.50 5.00 6.50 0.2 (0.1 0.8) rs10787471 0.49 0.39 0.97 0.95 0.47 5.94 5.21 3.9 (1.2 12.4) Blacks rs741596 29 Â§Â§ 0.14 0.11 0.22 0.71 0.23 5.02 5.02 0.09 (0.01 0.85) rs10885410 0.17 0.05 1.00 0.85 0.47 3.76 2.48 3.6 (0.6 22) rs11196180 0.17 0.01 0.69 0.80 0.78 3.57 3.82 0.2 (0.04 1.4) rs73360216 0.09 0.35 0.86 0.93 0.68 0.91 4.66 1.7 (0.4 8.4) SNP indicat es single nucleotide polymorphism; INVEST, INternational VErapamil SR and Trandolapril Study; MAF, minor allele frequency; HWE, Hardy Weinberg Equilibrium p value ; dom, dominant model; add, additive model SNPs are r anked in order of log(p) score from th e dominant model R epresents LD (r 2 ) of SNP with nearest GWAS index SNP including rs7903146, rs12255372, rs4506565, or rs7901695 R epresents highest LD (r 2 ) value with a ny SNP in cluded on the HumanCVD BeadChip Â§ D etermined by the sum of four negative log(p values) for four d escribed statistical models ** U nadjusted odds ratio and 95% confidence interval for NOD N ovel at t ime of sequence data delivery Â§Â§ SNPs were chosen as NOD predictor SNPs in their corresponding race/ethnic groups
71 Table 2 6 Characteri stics of new onset diabetes cases and controls at baseline and during INVEST Characteristic* NOD Cases(n=446) Controls (n=1,025) p value At baseline Age (years) 65 (10) 65 (9) 0.73 Female, n (%) 250 (56) 573 (56) 0.96 BMI (kg/m 2 ) 31 (6) 29 (5) <0.0001 Race/ethnicity, n (%) 0.66 White 176 (40) 409 (40) Black 51 (11) 121 (12) Hispanic 217 (49) 492 (48) Blood pressure (mm Hg) Systolic 149 (19) 148 (18) 0.26 Diastolic 87 (10) 86 (10) 0.04 Hypercholesterolemia n (%) 242 (54) 548 (54) 0.78 History of LVH, n (%) 77 (17) 128 (13) 0.02 History of MI, n (%) 99 (22) 194 (19) 0.15 History of smoking, n (%) 184 (41) 405 (40) 0.53 During INVEST Verapamil SR strategy, n (%) 208 (47) 519 (51) 0.16 Blood pressure (mm Hg) Â§ Systolic 135 (11) 134 (11) 0.07 Diastolic 79 (6) 79 (6) 0.49 HCTZ treatment, n (%) 328 (74) 640 (62) <0.0001 320 (72) 627 (61) <0.0001 HCTZ dose (mg) 28 (12) 26 (12) 0.17 Atenolol treatment, n (%) 229 (51) 479 (47) 0.10 Atenolol dose (mg) 72 (32) 67 (31) 0.02 Trandolapril treatment, n (%) 267 (60) 664 (65) 0.07 T randolapril dose (mg) 3.3 (2.6) 3.4 (2.6) 0.49 Verapamil SR treatment, n (%) 208 (47) 519 (51) 0.16 Verapamil SR dose (mg) 234 (75) 238 (75) 0.71 INVEST indicates INternational VErapamil SR and Trandolapril Study; NOD, new onset diabetes; BMI body mass index; SR, sustained release; LVH, left ventricular hypertrophy; HCTZ, hydrochlorothiazide Values are mean standard deviation unless otherwise noted. lowering medications Â§ Average of clinic blood pressure measurements during study
72 Table 2 7 Association of top candidate SNPs by race/ethnic gr oups in sequenced samples and INVEST GENES new onset diabetes case control cohort SNP, single nucleotide polymorphism ; NOD indicates new onset diabetes; MAF, minor allele frequency; HWE, Hardy Weinberg Equilibrium p value; OR, odds ratio ; 95%CI, 95% Confidence Interval Race/ethnic group in which SNP was identified as a candidate T2D predictor SNP Odds ratios 95% confi dence intervals and p values presented in a dominant model Odds ratios 9 5% confidence intervals and p values for NOD in an allelic trend test adjusted for BMI, gender, age, on treatment SBP, history of high cholesterol, history of smoking, treatment with and duration of HCTZ, ACE inhibitor, and atenolol, and PCs one, two and three. Sequenced s amples (n=150) INVEST GENES NOD case control cohort (n=1,435) SNP Race/Ethni c Group MAF HWE Unadjusted OR (95%CI) p MAF HWE Unadjusted OR (95%CI) p Adjusted O R p rs7895307 Blacks 0.31 0.58 1.04 (0.62 1.74) 0.88 1.47 (0.77 2.82) 0.25 Hispanics* 0.34 0.71 7.5 (2.0 27.6) 0.002 0.38 0.43 1.08 (0.85 1.36) 0.53 1.03 (0.78 1.35) 0.85 Whites 0.37 0.10 0.94 (0.71 1.24) 0.65 0.91 (0.65 1.27) 0.58 rs7901275 Blacks 0.29 0.65 1.07 (0.61 1.87) 0.82 1.77 (0.87 3.59) 0.12 Hispanics* 0.41 0.98 4.2 (1.3 13.5) 0.02 0.49 0.36 0.90 (0.71 1.14) 0.38 1.01 (0.76 1.34) 0.94 Whites 0.57 0.50 0.92 (0.69 1.22) 0.55 0.96 (0.67 1.37) 0.82 rs74159629 Blacks* 0.14 0.23 0.09 (0.01 0.85) 0.04 0.08 0.30 0.63 (0.24 1.67) 0.35 0.28 (0.08 1.03) 0.05 Hispanics 0.01 0.83 2.15 (0.84 5.49) 0.11 2.56 (0.96 6.84) 0.06 Whites 0.00
73 Table 2 8 Significant and n ominally significant p harmacogenetic interaction s for TCF7L2 SNPs and hydrochlorothiazide treatment on new onset diabetes in INVEST whites SNP Allele Freq. HWE OR (95%CI) (HCTZ Treated) OR (95%CI) (Not HCTZ Treated) p inx p FDR Â§ SNPs associated with higher NOD risk in HCTZ treated patients rs7917983 T 0.53 0.20 1.50 (1.00 2.25) 0.47 (0.26 0.84) 3.7x10 4 0.02 rs7901695 ** C 0.32 0.82 1.54 (0. 99 2.40) 0.52 (0.27 0.97) 9.7x10 4 0.02 rs4506565 ** T 0.31 0.91 1.54 (0.99 2.39) 0.52 (0.28 0.99) 0.001 0.02 rs4132670 A 0.32 0.73 1.47 (0.95 2.30) 0.52 (0.27 0.97) 0.001 0.02 rs4074720 T 0.54 0.49 1.27 (0.85 1.91) 0.45 (0.25 0.82) 0.001 0.02 rs6585202 T 0.53 0.69 1.16 (0.77 1.76) 0.47 (0.26 0.84) 0.003 0.03 rs7924080 T 0.53 0.62 1.18 (0.77 1.78) 0.47 (0.26 0.84) 0.003 0.03 rs11196174 G 0.26 1 1.10 (0.71 1.69) 0.41 (0.20 0.87) 0.007 0.06 rs7903146 ** T 0.28 0.54 1.40 (0.89 2.19) 0.62 (0.33 1.15) 0.01 0.09 rs6585195 C 0.13 0.63 2.68 (1.46 4.92) 0.86 (0.38 1.93) 0.02 0.09 rs12243326** C 0.28 0.46 1.32 (0.84 2.07) 0.60 (0.32 1.10) 0.02 0.10 rs10885399 A 0.21 0.16 1.92 (1.18 3.13) 0.73(0.36 1.49) 0.02 0.10 rs7094463 G 0.54 0.37 1.19 (0.79 1.77) 0.50 ( 0.29 0.88) 0.02 0.11 rs11196213 ** T 0.45 0.01 0.82 (0.55 1.22) 0.36 (0.19 0.68) 0.03 0.12 rs4918789 T 0.55 0.01 0.82 (0.55 1.22) 0.36 (0.19 0.68) 0.03 0.12 rs7087006 A 0.55 0.02 0.80 (0.54 1.19) 0.36 (0.19 0.68) 0.03 0.13 rs7079711 A 0.20 1 1.15 (0.71 1.87) 0.49 (0.22 1.11) 0.04 0.15 SNPs associated with lower NOD risk in HCTZ treated patients rs11196228 ** C 0.08 0.34 0.36 (0.15 0.86) 2.57 (1.09 6.02) 9.7x10 4 0.02 rs176632 T 0.15 0.46 0.47 (0.25 0.87) 2.40 (1.13 5.08) 0.002 0.02 rs3814572 G 0.15 0.52 0.91 (0.64 1.31) 1.63 (0.92 2.90) 0.009 0.06 rs7082458 G 0.16 0.72 0.42 (0.22 0.78) 1.38 (0.71 2.68) 0.02 0.09 rs7079673 A 0.15 0.35 0.47 (0.25 0.87) 1.45 (0.74 2.85) 0.01 0.09 rs12354626 A 0.03 0.23 0.64 (0.20 2.08) 3.54 (1.08 11.60) 0.02 0.11 r s12184389 A 0.16 0.72 0.43 (0.23 0.79) 1.25 (0.64 2.45) 0.03 0.13 SNP indicates single nucleotide polymorphism; Freq, frequency of allele in INVEST whites; OR, odds ratio; 95%CI, 95% confidence interval; HCTZ hydrochlorothiazide Hardy Weinberg equilibr average on treatment systolic blood pressure, hypercholesterolemia, history of smoking, potassium supplementation, princip al components one, two, and three, and trandolapril or atenolol treatment and treatment duration. HCTZ treatment and SNP after adjustment Â§ p value for interaction of HCTZ treatment and SNP after adjustment and correction for m ultiple testing using FDR ** SNPs previously associated with type 2 diabetes or diabetes related traits in genome wide association studies
74 Figure 2 1. Summary of Chapter 2 methodology by patient population Boxes represent steps in methodology as d ia gram progresses from top of figure to bottom of figure. Analyses in shaded blue box were performed in sequenced samples. Analyses in shaded pink box were performed in the new onset diabetes case control cohort.
75 Fig ure 2 2 Haploview generated l inkage dis equilibrium (LD) plot of sequenced TCF7L2 SNPs in INVEST Regions of higher LD are shaded darker according to higher r 2 values. Monomorphic SNPs, SNPs with MAF<0.05, and SNPs with call rate <75% are not included. 1a. INVEST sequenced individuals of Euro pean Ancestry (n= 48 ). 1b. INVEST sequenced individuals of Hispanic descent (n= 61 ). 1c. INVEST sequenced individuals of African ancestry (n= 41 ).
76 A B C
77 Figure 2 3 Venn diagrams for TCF7L2 polymorphisms by race/ethnicity in sequenced samples.
78 Fig ure 2 4 Plot of princi pal components one and two in sequenced samples by self reported race/ethnicity. PCA was performed using HumanCVD BeadChip data in 2,305 INVEST GENES participants. The two white individuals which cluster with Hispanics were re cate gorized as Hispanic for association analysis.
79 Figure 2 5 Odds ratios per copy of allele and 95% confidence intervals for TCF7L2 SNPs and new onset diabetes in INVEST patients by race/ethnicity. Candidate diabetes predictor SNPs nominally significant HumanCVD BeadChip SNPs and rs7903146 are included. All odds ratios are adjusted for age, gender, body mass index, average on treatment systolic blood pressure, left ventricular hypertrophy, hypercholesterolemia, history of smoking, principal components one, two, and three, and treatment and duration of treatment with trandolapril, atenolol, and hydrochlorothiazide. SNP indicates single nucleotide polymorphism, NOD new onset diabetes.
80 CHAPTER 3 ASSOCIATION OF TAG S NPS IN KCNJ1 ADD1 ACE, AND AGTR1 WITH CHANGE IN FG AND NOD DURING THIAZIDE TREATMENT Introduction The importance of identifying predictors of thiazide induced dysglycemia was emphasized by a working group from the NHLBI 46 A priori identification of patients who will develop dysglycemia during thiazide treatment could guide thiazide prescribing to reduce the risk of NOD S trong genetic predictors of T2D development have been observed in European popu lations 91 95 and pharmaco genetic associations with thiazide induced dysglycemia have been observed, 112 114 suggesting that pharmacogenetic risk factors could be used to predict T2D and thiazide induced dysglycemia. Despite the potential utility of personalized medicine in reducing the potential for AME s of thiazides, few studies have identifie d important pharmacogenetic risk factor s for thiazide induced hyper glycemia The f ew published studies that examine pharmacogenetic effect s of SNPs observe d significant associatio ns between SNPs and T2D or change in FG during thiazide treatment support ing SNP influences on thiazide induced dysglycemia 112 114 However, the impact of the majority of genetic variation in candidate genes is uninvestigated due to inadequate gene coverage in these studies D espite racial/ethnic differences in T2D prevalence, 1 in genetic associations with T2D, 98 and in LD structure, 115 the impact of r ace/ethnicity on genetic and pharmacogene tic risk factors for T2D and thiazide induced dysglycemia i s also largely uninvestigated. Furthermore the phenotype studied varies between change in FG and NOD, making interpretation of results problematic and resulting in an inability to draw conclusion s regarding the existence and implication of pharmacogenetic effect s The research presented in this
81 chapter seeks to i dentif y SNPs as risk factors for T2D and thiazide induced hyper glycemia using a candidate gene approach and to investigate the impact of rac e/ethnicity on SNP associations Whereas previous studies focus on a small number of well studied, primarily functional SNPs, 112 114 the research outlined in Chapter 3 investigates the effect of comprehensive variation within candidate genes on thiazide induced dysglycemia Thiazide s increase potassium excretion, which may blunt insulin release and contrib ute to thiazide induced hyper glycemia. 76 79 Thiazides also cause RAS activation 112 which contributes to hyper glycemia 70 160 Thereby, c andidate genes investigated in this research are involved in either electrolyte homeostasis or the RAS, which is con sistent with previous studies. (Figure 3 1) In each candidate gene studied, we attempted to replicate findings for the SN P previously associated with thiazide induced dysglycemia. Although increased FG levels have been observed with thiazide and thiazide like diuretics the mechanisms of thiazide induced hyper glycemia are not fully understood. Supporting the role of potassi um depletion in thiazide induced dysglycemia, the non synonymous SNP rs59172778 in the potassium inwardly rectifying channel, subfamily J, member 1 gene ( KCNJ1 ) has been associated with change in FG during four weeks of HCTZ treatment. 114 Since the protein coded by KCNJ1 the renal outer medullary potassium channel (ROMK1), plays an important role in potassium homeostasis 161 and a K CNJ1 SNP has been associated with change in FG after thiazide therapy 114 KCNJ1 was chosen as a candidate gene for this research. Another candid ate gene for the pharmacogenetic s of thiazide induced dysglycemia is the alpha adducin 1 gene ( ADD1 ) 162 Adducin is a ubiquitously
82 expressed cytoskeletal prote in that is involved in electrolyte homeostasis. 163 Variants in ADD1 ha ve been associated with hypertension, decreased BP response to diuretics, and increased CV outcomes with diuretic treatment. 164 167 The ADD1 variant rs4961 (Gly460Trp) previously associated with BP response to diuretic therapy 165 has been studied in investigati ons of pharmacogenetic s of thiazid e induced dysglycemia with inconsistent results. 112 114 The physiological role of ADD1 and a previous as sociation of an ADD1 SNP with thiazide induced NOD make ADD1 a compelling candidate gene for this research. Candidate genes of interest for this research also include key genes in the RAS 53 68 Reduced blood volume caused by thiazides results in a reduced perfusion of the juxtaglomerular apparatus, which subsequently releases renin. Renin activat es the RAS whic h has potentially unfavorable consequences inclu ding sympathetic activation and inflammation, which may adversely affect glucose homeostasis. 160 Stimulation of the RAS also results in a ldosterone release, causing potassium excretion which may further contribute to dysglycemia. In addition, RAS blocking agents, such as ACE I s and ARBs have been associated with reduced risk of NOD. 47 50 53 The genes that encode the two directly targeted proteins of ACEIs and A RBs are ACE and A GTR1 respectively and were chosen as candidate genes for this research. While t h e ACE insertion/deletion (I/D) polymorphism and the AGTR1 rs5186 SNP ( + A1166C ) have been associated with thiazide induced NOD 112 they have not been associated with thiazide induced FG changes to date 113 114 W e investigated the association of KCNJ1 ADD1 ACE, and AGTR1 tag SNPs with change in FG during short term HCTZ treatment in the PEAR study and NOD
83 during long term HCTZ treatment in INVEST We first sought to replicate pharmacogenetic associations with SNPs from previous studies and then investigated tag SNPs within these candidate genes for pharmacogenetic effects In addition, multivariate models including TCF7L2 SNPs from Chapter 2 were created to test genetic risk for T2D. Methods All patient s enrolled in both studies provided voluntary, written informed consent, and the institutional review boards of participating s tudy centers approved the study protocol s PEAR and INVEST are registered at ClinicalTrials.gov (NCT00246519 and NCT00133692 respectively). PEAR Study Design and Population PEAR is a prospective, randomized, open label parallel group study to evaluate the pharmacogenetic effects of the thiazide diuretic HCTZ the beta blocker atenolol, and their combination on BP response and AME s. Details of the PEAR study design ha ve been previously published 117 PEAR patient s aged 17 to 6 5 had mild to moderate essential hypertension without a history of heart disease secondary forms of hypertension, renal disease, or diabetes (type 1 diabetes or T2D). After a 3 8 week washout period, patient s were randomized to receive HCTZ 12.5 mg or atenolol 50 mg daily followed by dose titration to HCTZ 25 mg or atenolol 100 mg daily for 6 9 weeks (Figure 3 2 ) The other agent was then added, with similar dose titration for 6 9 wee ks of c ombination treatment FG, plasma insulin, total cholesterol, low density lipoprotein (LDL) cholesterol, high density lipoprotein (HDL) cholesterol, triglycerides, serum potassium, uric acid, and
84 urinary potassium were also acquired at response asses sment and safety check study visits 117 Patients were asked to fast for at least eight hours before study visits that included response assessments F asting s erum levels of glucose, lipids, and uric acid were determined using an Hitachi 911 Chemistry Analyzer (Roche Diagnostics, Indianapolis, IN) at a central laboratory at the Mayo Clinic. BP and electrolytes were regularl y monitored with determination of ser um potassium every 3 4 weeks at a local laboratory in patients taking HCTZ. Study physicians could elect to prescribe oral potassium supplementation with a protocol mandated 40 mEq potassium chloride daily by mouth i f serum potassium was below 3.2 mEq/L. INVEST Study Design and Population INVEST and INVEST GENES study designs are d escribed in detail in the Methods section of Chapter 2. Briefly, INVEST evaluat ed adverse CV outcomes and NOD occurring during randomized treatment with either a n atenolol bas ed or a verapamil SR based antihypertensive strategy in patients with hypertension and CAD years of age. The design, exclusion criteria, primary outcome, and NOD results have been previously published in detail 50 137 138 168 INVEST GENES collected DNA samples from 5,979 INVEST patient s at 187 sites in the US and Puerto Rico, who provided additional written informed consent for genetic studies. We conducted a nested case control study including those who developed NOD during follow up (cases) and a ge, race, and gender matched participants who remained diabetes free over a mean 2.8 years follow up (controls) NOD was determined by site investigators from a review of all av ailable patient data, including the use of anti diabetic medication 50 Patients taking anti diabetic medication or with diabetes history at baseline were excluded from this analysis.
85 Genotypin g and Quality Control In both PEAR and INVEST genotyping for candidate gene tag SNPs was accomplished using the HumanCVD BeadChip and Infinium II Assay (Illumina, San Diego, CA). Genotyping and QC methods for the HumanCVD Be adChip are further described in the Methods section of Chapter 2. Gene coverage for candidate genes was based on priority, with all candidate genes from Chapter 3 being either priority one (tag SNPs selected for r 2 >0.8 and MAF>0.02) or priority two (tag S NPs selected for r 2 >0.5 and MAF>0.05) based on Hapmap and Seattle SNP s genotype data. 142 152 (Table 3 1) Genotypi ng for the KCNJ1 SNP rs59172778 (hcv632615) was accomplished using the TaqMan 7900HT real time PCR system using c onditions specified in Chapter 2. 169 Rs59172778 genotypes were confirmed in five percent of genotyped samples, including all heterozygous individuals F unctional consequences of SNPs wer e determined in silico as described in Chapter 2. Definition and Treatment of Race/Ethnicity PEAR race/ethnicity was self described by study patient according to guidelines set forth by the NIH Office of Management and Budget (OMB) minimum standards for maintaining, collecting, and presenting data on race and ethnicity Race/ethnicity was confirmed by PCA performe d using LD pruned data from the HumanOmni1 Quad BeadChip (Illumina, San Diego, CA). In PEAR, any patient that did not report black ancestry du ring screening was considered n on black in statistical analyses. INVEST race/ethnic groups were determined by methods described in Chapter 2, including patient report with interaction by the study investigator and confirm ation through PCA T o minimize co nfounding by population stratification, pharmacogenetic analyses were performed by race/ethnicity and adjusted for PCs one, two, and three. PCs one, two,
86 and three were generated separate ly for each race/ethnic group in both PEAR and INVEST Statistical A nalysis We first sought to replicate pharmacogenetic associations with SNPs in candidate genes from previous studies which are summarized in Table 3 1. We t hen investigated effects of other tag SNPs within candidate genes. For previously associated SNPs significance was determined at p<0.05. For other tag SNPs within candidate genes, s ignificance was determined using an FDR correction for all SNPs within each candidate gene by race/ethnic group in each trial D eviations from HWE were assessed u sing Fi sher E xact Test by race/ethnicity with alpha =0.05 SNPs departing from HWE at p<0.05 were flagged, but in cluded in statistical analyses SNPs with significant departures from HWE in several race/ethnic groups or with HWE p<1x10 3 were excluded from pha rmacogenetic analyses. All statistical a nalys e s w ere performed using SAS version 9.2 and JMP Genomics version 5.0 (SAS, Cary, NC). PEAR Differences in change in FG and serum potassium during HCTZ monotherapy versus HCTZ add on therapy were tested using Wilcoxon rank sum. Laboratory values were tested for normality using Kolmogorov Smirnov and variables were log transformed if non normal. L inear regression was used to model SNP effects on change in log( FG ) during HCTZ using allelic trend test s in an add itive model P values were determined using log transformed data whereas betas were calculated using non transformed data to provide clinically interpreta ble genotype effects on change in FG Variables for adjustment were selected based on previously pub lished studies 118 170 and for potential impact on FG and included log(FG) at start of HCTZ age, ge nder, waist
87 circumference, potassium supplementation during the study, treatment arm, average baseline systolic and diastolic BP measured at home HCTZ dose, duration of HCTZ treatment, and PCs one, two, and three. KCNJ1 SNP effects on change in serum pot assium during HCTZ treatment were tested. We determined p harmacogenetic effects us ing a multivariate linear regression of change in FG during HCTZ treatment using combined data from HCTZ monotherapy and HCTZ add on therapy. ( Figure 3 2 ) Change in FG durin g HCTZ monotherapy was defined as the difference in FG from the start of HCTZ monotherapy to the end of HCTZ monotherapy. Change in FG during HCTZ add on therapy was defined as the difference in FG from the start of HCTZ to the end of the trial. An FDR a djust ment was used for all SNPs within each candidate gene by race/ethnicity. Consistency of SNP effects was also assessed in each race/ethnic group and for HCTZ monotherapy and HCTZ add on therapy. Finally, for significantly associated SNPs, a pharmacog enetic effects model was performed with adjustment for rs7903146 genotype in addition to other covariates. Assuming a MAF of 0. 06 we have 99 % power in Black s (n=3 04 ) and Non b lacks (n= 464 ) to detect an effect size of 0. 5 INVEST Basel ine differences in patient characteristics between cases and controls were determined using t tests, Wilcoxon Rank Sum, and chi square tests as appropriate. ORs per allele copy and 95% CIs were calculated using allelic trend tests. Multi variable logisti c regression models were used to assess SNP effects on NOD risk in patients treated with HCTZ as described in Chapter 2. Area under the receiver operating characteristic curves (AUROCs) were also generated for clinical variables used as covariates by race /ethnicity. SNP*HCTZ treatment interaction p values were
88 determined for significant SNPs. Significant SNPs were then tested for association with NOD in HCTZ treated patients with adjustment for rs7903146 genotype in addition to other covariates. In INVE ST, assuming an OR of 1.7 and MAF of 0.15, we have 8 9 % power in Whites, 40 % power in Blacks, and 94 % power in Hispanics to identify a SNP *HCTZ treatment interaction on NO D risk (two sided alpha =0.05). Linkage Disequilibrium and Haplotype Design LD analysis and pairwise LD (r 2 ) were performed within race/ethnic groups using Haploview. 147 We used SAS (SAS, Cary, NC) to create phased haplotypes within race/ethnic groups from SNPs reaching nominal significance (p<0.05) for candidate genes with multiple nomi nally significant associations. Common haplotypes (frequency>0.05) were tested for association with change in FG during HCTZ treatment and NOD risk in previously described statistical models. If two nominally associated SNPs were in very high LD (r 2 >0.9) one SNP was included in haplotypes. Results Baseline Characteristics and Clinical Predictor of Outcome Variables PEAR A total of 768 patient s were included in PEAR analysis, including 382 who received HCTZ monotherapy and 386 who received HCTZ add on the rapy. (Table 3 2 ) The variables FG, serum potassium, urine potassium, and plasma insulin were considered non normal using the Kolmogorov Smirnov test ( all p<0.01 ) and were log transformed before linear regression modeling. The average duration of HCTZ tr eatment was 9.3 (SD 1.9 ) weeks during HCTZ monotherapy and 9.6 (SD 1.9 ) weeks during HCTZ add on therapy (p=0.07).
8 9 Parameter estimates SEs, and p values for model covariate effects on change in FG are presented in Table 3 3 Strong predictors of increase d FG were lower baseline FG (beta= 0.39 [standard error (SE) 0.04 ] p<0.0001) and greater waist circumference (beta= 0.18 [ SE 0.03 ] p<0.0001). Median increase in FG was 2.50 mg/dL (interquartile range [IQR] 3.5 7.5) during HCTZ monotherapy and 2.00 mg/d L (IQR 4.0 8.0) during HCTZ add on therapy (p=0.69 for comparison between treatment arms ). The model containing clinical predictor variables explained 20.8% of the variability in the change in FG endpoint in blacks and (R 2 =0.208) and 20.6% of the variabi lity in the change in FG endpoint in non blacks (R 2 =0.206). Serum potassium decreased a median 0. 27 mEq/L (IQR 0.04 0.57) during HCTZ monotherapy and 0.31 mEq/L (IQR 0.03 0.59) during HCTZ add on therapy (p=0.60). These results support the similarity of distributions in FG data from HCTZ monotherapy and HCTZ add on therapy and support pooling change in log(FG) data. INVEST O ver a mean 2.8 years, 446 patients in the INVEST GENES cohort developed NOD 410 of which were successfully genotyped on the HumanC VD BeadChip. Differences in baseline characteristics between NOD cases and controls are described in detail in Chapter 2. Briefly, NOD cases had higher baseline BMI ( p<0.0001) and diastolic BP (p=0.04) and a higher percentage of LVH ( p=0.02) than age, r ace/ethnicity, and gender matched controls. (Table 3 4 ) During INVEST, the proportion of patients who were treated with HCTZ was higher in cases than in controls (p<0.0001). Mean atenolol dose was also higher in NOD cases than in controls (p=0.02) among patients treated with atenolol. Interestingly, significantly more NOD cases than controls received potassium supplementation during the study ( p=0.0001).
90 Associations between model covariates and NOD are presented in Table 3 5 In a multivariate logist ic regression, NOD was predicted by BMI (OR 1. 24 [95%CI 1.0 9 1. 42 ] per 5 kg/m 2 p = 0.00 2 ), history of LVH (OR 1.58 [95%CI 1.06 2.35], p=0.02) HCTZ treatment duration (OR 1. 13 [ 95%CI 1.0 7 2.35 ] per 6 month s p<0.0001), trandolapril treatment (OR 0.68 [95%CI 0.53 0.88], p <0.0001 ) trandolapril treatment duration (OR 0. 53 [ 95%CI 0. 47 0. 59 ] per 6 month s p<0.0001) and potassium supplementation during the study (OR 2.19 95%CI [1.3 9 3.43], p=0. 0007 ). (Table 3 5) The model containing clinical predictor variable s had an AUROC of 0.839 in INVEST whites treated with HCTZ, 0.791 in INVEST Hispanics treated with HCTZ and 0.855 in INVEST blacks treated with HCTZ. Hardy Weinberg Equil i br i um for Candidate Gene SNPs A total of 23 SNPs departed from HWE at p<0.05 in at least one race/ethnic group, which is less than that expected by chance for the 201 SNPs tested in two race/ethnic groups in PEAR and three race/ethnic groups in INVEST. However, several SNPs were out of HWE in multiple race/ethnic groups or had HWE p<1x 10 3 ( Appendix A: Table A 1 ) S even SNPs were excluded from pharmacogenetic analyses based on HWE p values including t he ADD1 SNP s rs1263347 and rs16843458 t he ACE SNP rs12709436 and t he AGTR1 SNP s rs2640543 rs12721286, rs12695918, and rs5183 KCNJ1 and Increased FG during HCTZ Treatment in PEAR The previously reported missense KCNJ1 SNP rs59172778 which was associated with a decrease in FG, 114 occurred only in non blacks (MAF=0.01), was not in LD with any other SNP tested, and was not associated with change in FG during HCTZ treatment (beta = 2.10 [SE 3.34], p=0.58). PEAR r esults for previously associated SNPs from our candidate genes are repo rted in Appendix A. (Table A 2 ) W e observe d
91 increased serum potassium during HCTZ treatment in non blacks, increasing a n average 0.2 3 (SD 0.44) mEq/L in six rs59172778 A/G heterozygotes and decreas ing 0.34 (SD 0.23) mEq/L in 222 A/A homozygotes (p=0.00 1 [ p FDR =0.02 ] ). Increased potassium with the G allele may be consistent with the decreased FG during HCTZ treatment observed previously with this allele. 114 The nonsynonymous rs59172778 SNP, coding a methionine to threonine substitution at amino acid 357, was predicted to be tolerated with a SIFT score of 0.37. In PEAR blacks (n=304), th e intro nic KCNJ1 SNP rs17137967 ( MAF 0.05) w as associated with an increased FG during HCTZ treatment after FDR correction (beta=8.47 [SE 2.45], p=0.0008, p FDR =0.009 ). SNPs that were significantly associated with change in FG after FDR correction in PEAR are pres ented in Table 3 7 FG increased a n average 2.0 (SD 13.9) mg/dL among 271 T/T homozygotes, 10.4 (SD 16.6) mg/dL in 25 T/C heterozygotes, and 15.8 (SD 15.8) mg/dL in two C/C homozygotes (Figure 3 3 ) Th e association remained significant after adjustment f or baseline and change in serum potassium (beta=7.94 [SE 2.46], p=0.002), urinary potassium (beta=8.17 [SE 2.47], p=0.001 ), and plasma insulin (beta=4.96 [SE 1.79], p=0.002 ) The KCNJ1 SNP rs17137967 association also remained significant (beta=7.54 [SE 2. 40], p=0.002) after adjustment for TCF7L2 rs7903146 genotype, which was nominally associated with change in FG in PEAR blacks (beta=3.05 [SE 1.29], p=0.02). A n ominal (p<0.05) association with increased FG w as also observed for the rs2846680 A allele (beta =3.55 [SE 1.44], p=0.03, p FDR =0.13 ) in blacks. The two SNPs rs17137967 and rs2846680 ( r 2 =0.01 ) were used to create three common haplotypes for
92 black PEAR patients. Haplotype findings in blacks were driven by rs17137967 and therefore were not considered f urther. No SNPs in KCNJ1 were observed to have significant effects on change in FG after FDR correction in n on blacks. The intronic SNP rs7933427 ( MAF=0.04) was nominally associated with increased FG (beta=3.46 [SE 1.55], p=0.02, p FDR =0.29) although this SNP deviated from HWE in non blacks (p=0.01) The SNP rs17137967 was monomorphic in PEAR non blacks and so statistical analysis for change in FG during HCTZ was not possible. KCNJ1 and NOD Risk after HCTZ Treatment in INVEST The KCNJ1 SNP rs59172778 wh ich was previously reported to be associated with changes in FG 114 was not associated with NOD in any race/ethnic group and showed no SNP*HCTZ tre atment interactions. INVEST results for previously associated SNPs in our candidate genes are reported in Appendix A. (Table A 3) In HCTZ treated whites (n=371), two KCNJ1 SNPs (rs12795437 and rs11600347) were significantly associated with a greater tha n two fold increase in NOD risk per variant allele (p=0.006 [p FDR =0.04] and p=0.008 [p FDR =0.04] respectively). (Table 3 7 ) These two SNPs were also associated in HCTZ treated Hispanics (n=464) at nominal (p<0.05) significance. ( Figure 3 4 ) One SNP (rs658 903) was associated with a more than 60% reduced risk of NOD (p=0.002 [p FDR =0.04]) in Hispanics. In HCTZ treated black patients (n=131), one SNP (rs675388) was associated with a 3.13 fold increased NOD risk with each allele copy (p=0.004 [p FDR =0.03]). We found 10 other SNPs associated with NOD in HCTZ treated patients at nominal (p<0.05) significance in at least one race/ethnic group. (Figure 3 4 )
93 After LD analysis in whites (Appendix A: Figure A 1 ) three SNPs (rs2238009, rs12795437, and rs11600347) were used to construct three haplotypes with frequency>0.05. The haplotype HapW1 (GCA) was associated with significantly increased risk for NOD during HCTZ treatment (p=0.006 [p FDR =0.04]). (Table 3 7 ) In Hispanics, haplotypes were inferred from five nominall y associated SNPs (rs675388, rs1148058, rs658903, rs12795437, and rs3016774), resulting in five haplotypes with frequency>0.05. HapH1 (CATCT) was associated with an approximately two fold increased risk for NOD per haplotype copy ( p=0.003 [ p FDR =0.03 ] ) and the opposite haplotype HapH2 (TGAGC) was associated with a 57% reduction in NOD risk per haplotype copy ( p=0.007 [ p FDR =0.04 ] ). In blacks, two nominally associated SNPs ( rs675388 and rs1148059 ) were used to infer three haplotypes with frequency>0.05. Hap B1 (GC) was associated with a 72% decreased NOD risk during HCTZ treatment ( p=0.003 [ p FDR =0.02 ] ) Consistent associations were observed for SNPs and haplotypes in all three ethnic 7 ) Consistent associations were also observed for each SNP and haplotype after adjustment for TCF7L2 rs7903146 genotype, although rs7903146 was not associated with NOD in any race/ethnic group. Importantly, no KCNJ1 SNPs or haplotypes were associated wi th NOD in patients not treated with HCTZ in any race/ethnic group. This resulted in significant SNP*HCTZ treatment interactions for NOD. ADD1 and Increased FG during HCTZ Treatment in PEAR The ADD1 SNP rs4961 (MAF 0.21) which was associated with NOD dur ing thiazide treatment in a European population, 112 was associated with change in FG in PEAR non blacks (beta = 1.99 [SE 0.81], p=0.015 [ p FDR = 0.10]). (Table 3 6 ) For rs4961,
94 FG increased an average 1.04 (SD 10.6) mg/dL in 279 C/C homozygotes, 3.08 (SD 11.0) mg/dL in 141 C/A heterozygotes, and 6.26 (SD 9.0) mg/dL in 21 A/A homozygotes. The direction of effect for rs4961 is consistent with the previously publishe d association, with the A allele increasing the risk for NOD during thiazide treatment 112 The association remained significant after adjustment for TCF7L2 rs7 903146 genotype although rs7903146 was not associated with change in FG in PEAR non blacks. T he rs4961 A allele was also associated with new onset IFG in non black PEAR patients (OR 1.68 [95%CI 1.06 2.66], p=0.027). In PEAR blacks, rs4961 (MAF 0.06) was not associated with change in FG in PEAR blacks (beta = 0.05 [SE 2.32], p=0.90 [p FDR =0.99]). Functional effects for the non synonymous SNP rs4961 were also predicted in silico with the guanine to thymine substitution at amino acid 460 predicted to be d amaging with a SIFT score of 0.03. N o ADD1 SNPs were significantly associated with change in FG after FDR correction in non blacks However, 11 of 32 ADD1 SNPs were observed to have nominally significant associations with change in FG during HCTZ treatmen t. The majority of these nominally associated SNPs were in high LD (r 2 >0.90) with rs4961 ( Appendix A: Figure A 2 ) In PEAR blacks, no ADD1 SNPs were associated with change in FG during HCTZ treatment after FDR correction. The ADD1 SNP rs17777307 G alle le was nominally associated ( beta = 24.97 [SE7.77], p=0.003 [ p FDR =0.09 ] ) with FG changing an average 2.62 (SD 14.0) mg/dL in 301 A/A homozygotes and 24.67 (SD 38.5) mg/dL in three heterozygotes.
95 ADD1 NOD Risk after HCTZ Treatment in INVEST The previously associated ADD1 SNP rs4961 was not associated with NOD during HCTZ treatment in any INVEST race/ethnic group. ( Appendix A: Tab le A 3 ) No ADD1 SNPs showed significant associations with NOD after FDR correction or significant SNP*HCTZ treatment interaction s after FDR correction in any INVEST race/ethnic group. In INVEST whites, nominally significant SNP*HCTZ treatment interactions were observed for rs12509447 (p=0.01 [p FDR =0.08]) rs12503220 (p=0.01 [p FDR =0.12]) and rs3775067 ( p=0.006 [p FDR =0.08]). T hes e nominally significant interaction p values in INVEST whites were driven by significant ly increased ORs for NOD in the non HCTZ treated patients and non significant ORs in HCTZ treated patients. The rs3775067 T allele ( MAF 0.39 ) was associated with incre ased NOD risk in patients not treated with HCTZ (OR 2.40 [ 95%CI 1.35 4.24], p=0.003), but no association was observed in HCTZ treated patients (OR 1.14 [ 95%CI 0.85 1.52], p=0.38). Similar effects were seen with rs12509447 and rs12503220 in non HCTZ treate d patients (OR 2.52 [ 95%CI 1.35 4.70], p=0.003 and OR 2.29 [ 95%CI 1.23 4.26], p=0.009 respectively) with no association observed in HCTZ treated patients The SNPs rs12509447 and rs12503220 were in high LD (r 2 =0.92) and these SNPs were in moderate LD with rs3775067 (r 2 =0.33 and 0.36 respectively) (Appendix A: Figure A 3) but no haplotypes were generated since these SNPs were not associated with NOD in HCTZ treated patients In INVEST Hispanics, the rs7689864 A allele (MAF 0.02 ) was associated with increa sed NOD in HCTZ treated patients (OR 3.63 [95%CI 1.24 10.56 ] p=0.018 [p FDR =0.57] ). In INVEST blacks, the rs16843169 A allele (MAF 0.05 ) was associated with increased NOD in HCTZ treated patients (OR 4.25 [95%CI 1.04 17.35 ] p=0.04
96 [p FDR =0.7 8 ]). Although these ORs were consistent when high dose and long term treatment definitions were used, no significant SNP*HCTZ interaction p values were observed for these SNPs in either Hispanic or black patients. ACE and Increased FG during HCTZ Treatment in PEAR Th e ACE SNP rs4341, which was previously associated with NOD during thiazide treatment, has been observed to be in LD with the ACE I/D polymorphism and the effect of rs4341in the previous study was attributed to the effect of the ACE I/D polymorphism. 112 The SNP which most closely tags the I/D polymorphism on the HumanCVD Beadchip is rs4343. 171 The SNP rs4343 was not significantly associated with change in FG in PEAR blacks or non blacks. (Appendix A: Table A 2) In PEAR blacks, the non synonymous ACE SNP rs4303 A allele (MAF 0.10) was associated with decreased FG during HCTZ therapy (beta= 6.39 [SE1.92] p= 6.80x10 4 [p FDR =0.03]). (Table 3 6 ) FG decreased 37.3 (SD 54.8) mg/dL in two A/A homozygous variants, decreased 1.4 (SD 13.5) mg/dL in 55 A/C heterozygotes, and increased 4.2 (SD 13.6) mg/dL in 234 C/C wild type homozygotes. The ACE rs4303 associatio n remained significant (beta= 5.88 [SE 1.92], p=0.003) after adjustment for TCF7L2 rs7903146 genotype, which was nominally associated with change in FG (beta=3.05 [SE 1.29], p=0.02) in PEAR blacks. The rs4303 SNP encodes an Alanine to Serine substitution at amino acid 261, predicted to be damaging with a SIFT score of 0.08, supporting a true effect of the SNP. The rs4303 A allele was monomorphic in PEAR non blacks and so this association could not be assessed In PEAR non blacks, the intronic ACE rs127094 36 A allele ( MAF 0.0 02 ) was significantly associated with increased FG ( beta= 34.51 [SE 6.71] p=1.08x10 6 [ p FDR =4.22x10 5 ]) However, this polymorphism occurred in only two individuals with an
97 average FG increase of 40. 3 (SD 8.1) mg/dL and t h e association was considered spurious N o functional consequences of this intronic polymorphism were predicted in silico ACE NOD Risk after HCTZ Treatment in INVEST The tag SNP for the ACE I/D polymorphism (rs4343) was not associated with NOD risk during HCTZ treat ment and was not observed to have a significant SNP*HCTZ treatment interaction in any INVEST race/ethnic group and. (Appendix A: Table A 3) No SNPs showed significant associations with NOD after FDR correction or SNP*HCTZ treatment interactions in any rac e/ethnic group in INVEST. AGTR1 and Increased FG during HCTZ Treatment in PEAR The promoter AGTR1 SNP rs5186 ( + A1166C) C allele, which was previously associated with decreased NOD incidence during thiazide treatment, 112 was not associated with change in FG in any race/ethnic group in PEAR (Appendix A: Table A 2) No AGTR1 SNPs showed significant associations with change in FG during HCTZ treatment in either ra ce/ethnic group. In PEAR blacks, the intronic rs12721280 G allele (MAF 0.06) was nominally associated with increased FG during HCTZ treatment ( beta= 10.82 [SE4.58] p=0.018 [p FDR =0.49] ) AGTR1 NOD Risk after HCTZ Treatment in INVEST The AGTR1 rs5186 C all ele which was previously associated with decreased NOD incidence during thiazide treatment in a European population 112 was not associated with NOD during HCTZ treatment in any race/ethnic group although a weak trend towards significantly increased NOD risk was observed in HCTZ treated blacks (OR 2.69 [0.72 10.12], p=0.14). (Appendix A: Table A 3) Rs5186 was not observed to
98 have any significant SNP*HCTZ treatm ent interactions in any INVEST race/ethnic group. In INVEST whites, the AGTR1 rs9682137 A allele (MAF=0.01) was nominally associated with an increased risk of NOD in HCTZ treated patients (OR 31.26 [2.77 352.12], p=0.005 [p FDR =0.31] ), but did not show a si gnificant SNP*HCTZ treatment interaction. In INVEST blacks, the rs12695901 T allele (MAF=0.07) was nominally associated with an increased risk of NOD in HCTZ treated patients (OR 14.82 [2.43 90.23], p=0.002 [p FDR =0.29]), although a significant SNP*HCTZ tr eatment interaction was not observed. Discussion In the research presented in Chapter 3 we examined the effects of tag SNPs in four different candidate genes on two phenotypes across the continuum of T2D risk. For candidate gene SNPs previously associate d with thiazide induced dysglycemia, the nonsynonymous ADD1 SNP rs4961 was associated with increased FG in PEAR non blacks. Significant associations after FDR correction for candidate gene tag SNPs were observed for KCNJ1 SNP s and haplotypes in every race /ethnic group in INVEST and in PEAR blacks (rs17137967) A significant association was also observed for a nonsynonymous ACE SNP (rs4303) in PEAR blacks Alternate SNPs from both KCNJ1 and ACE have been previously associated with thiazide induced dysglyc emia, suggesting the importance of these two genes in predicting thiazide induced hyperglycemia. Significant associations from KCNJ1 ADD1 and ACE remained significant when adjusted for TCF7L2 rs7903146 genotype, suggesting that these pharmacogenetic pre dictors are robust to adjustment for baseline genetic T2D risk
99 In the present study, we observed a significant association between KCNJ1 variation and increased FG in blacks treated with short term HCTZ. We also observed significant increases in serum p otassium with a KCNJ1 SNP previously associated with decreased glucose during HCTZ treatment. Our study also observed that KCNJ1 variation was associated with NOD in patients treated with long term HCTZ in all race/ethnic groups. These associations remai ned significant in patients treated with HCTZ for an extended duration and in patients taking higher daily doses of HCTZ. We observed no associations achieving even nominal significance in non HCTZ treated patients, suggesting effects of KCNJ1 variation o n dysglycemia are specific to HCTZ treatment. KCNJ1 variation was associated with change in FG in PEAR and NOD in INVEST, during both short and long term HCTZ treatment, and in most race/ethnic groups studied, suggesting that variability in KCNJ1 affects HCTZ induced dysglycemia. Although KCNJ1 variation was associated with change in FG in PEAR and NOD in INVEST S NP associat ions with each endpoint differed between the two studies. In addition, t he ADD1 SNP rs4961 and ACE SNP rs4303, which w ere significan tly associated with change in FG in PEAR, w ere not associated with NOD in INVEST. Disparate results in PEAR and INVEST suggest different pharmacogenetic markers for FG in the short term versus NOD over the long term during HCTZ treatment These observati ons are consistent with literature from GWAS, which suggest that genetic associations with T2D are distinct from genetic associations with FG. 95 172 PEAR and INVEST also enrolled different study populations and used varied durations of HCTZ treatment These differences in study populations and study design might have contributed to disparate genetic associati ons with the dysglycemia
100 phenotype of focus in each study. The majority of INVEST and PEAR patients were treated with a 25 mg dose of HCTZ suggesting that dose likely had a negligible effect on differences in associations. T he duration of HCTZ treatment was v ery different between trials Patients received HCTZ for an average 9.5 weeks in PEAR and 87.1 weeks in INVEST. The difference in HCTZ treatment duration may partly account for conflicting results in PEAR and INVEST. Disparate results in PEAR and I NVEST might also be a reflection of racial/ethnic diversity. D isparate associations between race /ethnic groups suggest differences in LD and the need to i d entify functional variants in KCNJ1 S ignificantly associated variants from KCNJ1 (rs17137967) and ACE (rs4303) in PEAR blacks were monomorphic in non blacks in PEAR explainin g the lack of association in non blacks T he rs12795437 C allele, which increased NOD risk in INVEST whites and trended towards an increased risk in Hispanics, was not associated with changes in FG in any PEAR race/ethnic group. The lack of observation of an increased NOD risk with the rs12795437 C allele in INVEST blacks is potentially explained by the limited sample size in this race/ethnic group. Point estimates suggest ed an increased NOD risk in HCTZ treated patients with the rs12795437 C allele in all INVEST race/ethnic groups. The two KCNJ1 SNPs rs17137967 and rs12795437 were associated with different glucose related phenotypes within different race/ethnic groups in PEAR a nd INVEST. T he two SNPs rs17137967 and rs12795437 have been associated with mean 24 hour systolic BP, 173 hypertension, 174 and rosiglitazone induced edema 175 although they do not necessarily represent the main findings of each pape r. Alt hough these associations are only in directly related to T2D and both KCNJ1 SNPs are intronic the previous
101 association of both SNPs to phenotypes related to hypertension and electrolyte homeostasis suggest s a functional role for these SNPs and adds credence to our pharmacogenetic findings. This research adds to existing literature by attempting to replicate a previous KCNJ1 SNP association and investigating the effect of additional KCNJ1 variation on both FG changes and NOD. An analysis from the Gen etic Epidemiology of Responses to Antihypertensives (GERA) study observed decreased FG in rs59172778 (M338T) G allele carriers (n=8, mean 4.6 mg/dL) and increased FG in A/A homozygotes (n=532, mean 3.8 mg/dL) after four weeks of HCTZ treatment. 114 Similar to PEAR the GERA study quantified changes in FG from baseline during over four weeks of HCTZ monotherapy. 117 In the present study, a significant association between th e non synonymous KCNJ1 SNP rs59172778 was not observed for change in FG or NOD during HCTZ However, the rs59172778 G allele was assoc iated with increase d serum potassium during HCTZ treatment in PEAR This increase in serum potassium with the rs59172778 G allele may be consistent with decreased FG association in GERA, since potassium depletion during thiazide treatment has been implica ted in hyperglycemia 76 However, the GERA study reported a lack of association between th is SNP and change in serum potassium. 114 The rs59172778 SNP was predicted to be tolerated in silico but has a clear functional role being previously associated with Antenatal Bartter Syndrome. 176 Additional studies are necessary to refine the role of potassium depletion and KCNJ1 variation in thiazide induced dysglycemia.
102 The relationship between potassium and glucose is described in previous chapters and de creased serum potassium has been implicated in thiazide induced dysglycemia. 46 K CNJ1 encodes the r enal outer medullary potassium channel ( ROMK1 Kir1.1) that is r esponsible for potassium excretion in exchange for sodium absorption through the epithelial sodium channel (ENaC). 161 This action occurs in the collecting duct distal to the thiazide sensitive sodium /chloride co transporter, the direct target of thiazide diuretics SNPs may potassium homeostasis and affecting glucose dependent insulin secretion from pancreatic beta cells or glucose uptake into skeletal muscle. 77 161 KCNJ1 potassium excretion and the previous association of a KCNJ1 SNP with thiazide induced dysglycemia ma ke KCNJ1 a particularly compelling candidate gene for further study of thiazide induced dysglycemia pahrmacogenetics The present study observe d a significant effect of the non synonymous ADD1 SNPs rs4961 on FG in PEAR non blacks, although no significant a ssociati on was observed in INVEST Th e rs4961 association in PEAR is considered a complementation of the previous rs4961 association with NOD rather than a true replication, since the association observed with the phenotype in the present study, change i n FG, was different from the NOD phenotype in the previous study 112 The previous study implicating rs4961 in thiazide induced dysglycemia was an observational case control study from the Pharmaco Morbidity Record Linkage System (PHARMO RLS) and observed a significant SNP*thiazide treatment interaction, with a n increased NOD risk in rs4961 T allele carriers who were thiazide treated ( OR 1.88 [95%CI 1.36 2.59]). 112 This risk was increased in patients treated with higher daily doses of thiazides
103 (OR 2.36 [95%CI 1.64 3.39]). However, t he GERA study did not observe a sig nificant association between rs4961 and increased FG during four weeks of HCTZ treatment. 114 In addition, the GenHAT s tudy observed no association of rs4961 in a repeated measures, mixed models regression of FG levels during chlorthalidone treatment. 113 Our results suggest that the ADD1 rs4961 T allele ma y predict increases in FG in non black patients but not in black patients This contrasts with observations from GenHAT and GERA which investigated the effect of rs4961 and changes i n FG during thiazide treatment This discrepancy may be due to the pres ence of large black populations in both GERA and GenHAT. In addition, PHARMO RLS, which observed positive results was conducted almost entirely in white i ndi viduals consistent with our significant findings in PEAR non blacks The effect of rs4961 may h ave been confounded by black individuals, who tend to respond better to diuretic treatment especially when it is added on to a beta blocker 177 178 Our results add to the present body of evidence suggesting that rs4961 may be an important predictor of hyperglycemia during thiazide administration in non black individuals Adducin is a ubiquitously expressed he terodimeric cytoskeleton protein that promotes the binding of spectrin with actin and may be involved in such cell functions as ion transport. The ADD1 variant rs4961 (Gly460Trp) has been associated with hy pertension 164 and response to diuretic therapy. 165 The potential physiologic role of ADD1 in thiazide diuretic mechanism and the previous association of the ADD1 non synonymous SNP rs4961, a guanine to thymine non synonymous SNP at nucleotide 614 in exon 10 of ADD1 with NOD during thiazide treatment make it a compelling candidate gene for further study of thiazide induced hyperglycemia.
104 For ACE we observed a signif icant association between the non synonymous SNP rs4303 A allele and decreased FG which was consistent during both HCTZ monotherapy and HCTZ add on therapy. The rs4303 SNP was not associated with NOD risk, further supporting that pharmacogenetic predictor s of change in FG during HCTZ are distinct from predictors of NOD risk during HCTZ therapy. Although candidate SNPs within ACE have been tested in pharmacogenetic studies of thiazide induced dysglycemia, 112 114 none of these studies include investigation of rs4303. This SNP encodes an Alanine to Serine substitution at amino acid 261 and may inhibit function or efficiency of ACE, leading to decreased sympathetic activation and decrease d FG levels 160 Considering the potential functional impact of this SNP on ACE and the significant association in PEAR, further study of rs4303 and other ACE SNPs as a predictor of thiazide induced hyperglycemia is warranted. Previous pharmacogenetic studies investigating ACE polymorphisms have included the ACE I/D polymorphism or a tag SNP for this polymorphism. T he GenHAT study obs erved no effect of ACE I/D on FG during chlorthalidone treatment The PHARMO RLS study observed an increased NOD risk in thiazide treated patients with the rs4341 C allele (OR 2.25 [95%CI 1.53 3.29]), which the authors attributed to the effect of the I/D polymorphism which they report to be in complete LD with rs4341. 112 The functional SNP in this association remains unclear as rs4341 is itself a missense SNP a nd the authors reported only limited LD 0.91 r 2 not reported ) between rs4341 and the ACE I/D polymorphism 179 T he HumanCVD BeadChip does not include rs4341, but includes the rs4343 SNP known to tag the ACE I/D polymorphism. 171 Our results suggest that the ACE I/D polymorphism does not
105 influence thiazide induced hyperglycemia ob serving no significant associations in either PEAR or INVEST. AGTR1 is been implicated in thiazide induced hyperglycemia through its central role in the RAS and the decreased incidence of NOD observed with administration of ARBs 47 160 The GenHAT and GERA studies found no association of the AGTR1 rs5186 SNP (+ A 1166C) with FG change s during thiazide diuretic treatment. 113 114 The PHARMO RLS study observed a decreased NOD risk in the rs5186 CC genotyp e group who were treated with thiazide diuretics. 112 We observed no association between rs5186 or any other AGTR1 SNP and change in FG or NOD risk during HCTZ treatment. Our results agree with several previous studies, 113 114 suggesting that AGTR1 poly morphisms do not influence FG levels during thiazide treatment. Our study has several limitations worthy of mention. We recognize the potential for false positive results in our analyses and our observations need to be independently replicated. We attem pted to reduce false positives by using an FDR correction for all p values acquired in SNP analyses within each candidate gene and with in each race/ethnic group. We also sought to minimize false positive results by monitoring MAF and confirming consistenc y of results in different treatment arms. Furthermore, our candidate genes were all previously associated with glucose related phenotypes and all significantly associated SNPs had either putative functional consequences or published associations with dise ase, lending credence to our results. Another limitation is that important pharmacogenetic risk factors in candidate genes may not have be en identified due to low MAF or low r 2 with ge netic variation genotyped on the HumanCVD BeadChip This is particularl y true in INVEST blacks, for
106 which we had the most limited power to detect pharmacogenetic associations. W e cannot conclude a lack of important pharmacogenetic predictor SNPs in these candidate genes in any race/ethnic group, particular ly in INVEST blacks However, t he t ag SNPs included on our array were designed for black and white populations and likely provid ed good coverage of candidate genes in each race/ethnic group. Furthermore, all candidate genes except for KCNJ1 were priority one genes and ge netic variability was well covered on the HumanCVD BeadChip In addition, our analyses adjusting for TCF7L2 rs7903146 genotype must be interpreted with caution, as rs7903146 was not observed to be associated with NOD in any race/ethnic group in INVEST, and was only associated with change in FG in PEAR blacks. Finally, c hange in FG in PEAR and NOD in INVEST may be confounded by the use of other antihypertensive agents that affect FG levels, including atenolol and trandolapril treatment, ongoing environmenta l factors, or potassium supplementation. We attempted to reduce confounding in PEAR by controlling for metabolically important variables, drug arm, HCTZ dose and duration, as well as potassium supplementation. We also controlled for metabolically importa nt variables, treatment strategy, HCTZ treatment duration, atenolol or trandolapril treatment and treatment duration and potassium supplementation in INVEST. Summary and Significance In summary our results add to available evidence suggesting that g ene tic variation in the candidate genes KCNJ1 ADD1 and ACE influences effect s of HCTZ on glucose homeostasis In addition, we complimented a previous association of rs4961 with NOD through observations in the present study in PEAR non blacks. Our results suggest that pharmacogenetic associations with change in FG during short term HCTZ
107 treatment are distinct from pharmacogenetic associations with NOD after long term HCTZ treatment. Our observation of different SNP effects between race/ethnic groups sugges ts differences in LD and highlights the need to identify functional SNPs. Functional studies and replication of these associations are needed to better define the potential role of KCNJ1 ADD1 and ACE SNPs in predicting AME s of HCTZ. Significant associa tions from KCNJ1 ADD1 and ACE remained significant when adjusted for TCF7L2 rs7903146 genotype, suggesting that these pharmacogenetic predictors are robust to adjustment for baseline genetic T2D risk. Whether these findings are specific to HCTZ or can b e generalized to all thiazide diuretics is also unclear, but KCNJ1 ADD1 and ACE remain compelling candidate gene s for further study of pharmacogenetic s of thiazide induced dysglycemia.
108 Table 3 1 Summary of candidate genes investigated as pharmacogenet ic predictors Candidate Gene Protein Product Priority SNPs Candidate SNP (s) Evidence for candidate SNP association with thiazide induced dysglycemia ACE Angiotensin I 1 n=54 rs4341 A ssociated with NOD during thiazide treatment 112 converting enzyme rs4343 113 ADD1 A lpha adducin 1 1 n=32 rs4961 A ssociat ed with NOD during thiazide treatment ; 112 not 113 114 AGTR1 Angiotensin II type 1 receptor 1 n=90 rs5186 A ssociated with NOD during thiazide treatment; 112 n ot 113 114 KCNJ1 P otassium inwardly recti fying channel 2 n=2 5 rs59172778 A 114 SNP indicates single nucleotide polymorphism; ing glucose ; NOD, new onset diabetes; *Priority of candidate genes on the HumanCVD BeadChip: Priority 1 selected for minor allele frequency >0.01 and r 2 >0.5; Priority 2 selected for minor allele frequency >0.05 and r 2 >0.5 Number of SNPs investigated in the present study using HumanCVD BeadChip and Taqman T ag SNPs for the ACE insertion/deletion polymorphism
109 Table 3 2 Baseline characteristics of PEAR patients by randomized treatment arm Characteristic* HCTZ (n=382) Atenolol (n=386) Age (years) 49 (9) 49 (9) Female, n (%) 188 (49) 217 (56) BMI (kg/m 2 ) 31 (5) 31 (6) Waist circumference (cm) 98 (13) 98 (13) Race/ethnicity, n (%) Black 147 (38) 146 (38) White 219 (57) 221 (57) Asian 3 (1) 5 (1) Other/multiracial 13 (3) 14 (4) Race/ethnicity for analysis n (%) Black 152 (40) 152 (39) Non black 230 (60) 234 (61) Current smoker 61 (16) 50 (13) Home blood pressure ( mmHg ) Systolic 147 (11) 145 (10) Diastolic 94 (6) 93 (6) Fasting glucose (mg/dL), median (IQR) 91 (85 98) 89 (84 96) Serum potassium (mEq/L), median (IQR) 4.2 (4.0 4.5) 4.2 (4.0 4.5) Urine potassium (mEq/L), median (IQR) 53.8 (31.4 73.7) 55.0 (34.3 83.7) Plasma insulin ( U/mL), median (IQR) 7.2 (4.7 10.9) 6.8 (4.5 11.0 Serum Creatin ine (mg/dL) 0.87 (0.2) 0.87 (0.2) Triglycerides (mg/dL) 132 (106) 139 (100) HDL (mg/dL) 48 (14) 47 (14) Uric Acid (mg/dL) 5.72 (1.5) 5.76 (1.3) HCTZ indicates hydrochlorothiazide; BMI, body mass index; HDL, high density lipoprotein; IQR, interquartile range; mg/dL, milligrams per deciliter; mEq/L, milliequivalents per liter. Values are mean standard deviation unless otherwise noted. Race/ethnicity based on principal components analysis. A verage of home blood pressure values.
110 Table 3 3 Associat ion of linear regression model covariates from primary analysis with change in fasting glucose in PEAR Covariate Parameter Estimate (standard error) p value Baseline fasting glucose (mg/dL) 0.39 (0.04) <0.0001 Age (years) 0.09 (0.05) 0.09 Gender (mal e) 0.28 (0.93) 0.76 Waist circumference (cm) 0.18 (0.03) <0.0001 Potassium supplementation 0.37 (1.48) 0.80 Randomized drug arm (HCTZ) 0.44 (1.01) 0.67 Systolic blood pressure (mmHg) 0.03 (0.05) 0.62 Diastolic blood pressure (mmHg) 0.06 (0.09) 0. 52 Time of HCTZ treatment (days) 0.07 (0.03) 0.05 Race/ethnicity (Black) 0.62 (5.97) 0.92 Principal component 1 26.73 (80.37) 0.74 Principal component 2 3.02 (11.98) 0.80 Principal component 3 7.74 (13.00) 0.55 HCTZ indicates hydrochlorothiazide; mg/dL, milligrams per deciliter. Generated using linear regression of change in fasting glucose for all PEAR patients. Represents average of home blood pressure values.
111 T able 3 4 C haracteristics o f new onset diabetes c ases and c ontrols at baseline and during INVEST Characteristic NOD Cases (n=446) Controls (n=1,025) p value Baseline Age 65 (10) 65 (9) 0.73 Female, n (%) 250 (56) 573 (56) 0.96 BMI (kg/m 2 ) 31 (6) 29 (5) <0.0001 Race/ethnicity, n (%) 0.66 White 176 (40) 409 (40) Black 51 (11) 121 (12) Hispanic 217 (49) 492 (48) Blood pressure (mm Hg) Systolic 149 (19) 148 (18) 0.26 Diastolic 87 (10) 86 (10) 0.04 Verapamil SR strategy, n (%) 208 (47) 519 (51) 0.16 Hypercholesterolemia, n (%) 242 (54) 548 (54) 0.7 8 History of LVH, n (%) 77 (17) 128 (13) 0.02 History of prior MI, n (%) 99 (22) 194 (19) 0.15 History of smoking, n (%) 184 (41) 405 (40) 0.53 During INVEST Blood pressure (mm Hg) ** Systolic 135 (11) 134 (11) 0.07 Diastolic 79 (6) 79 ( 6) 0.49 HCTZ treatment, n (%) 328 (74) 640 (62) <0.0001 HCTZ 320 (72) 627 (61) <0.0001 HCTZ dose (mg) 28 (12) 26 (12) 0.17 Atenolol treatment, n (%) 229 (51) 479 (47) 0.10 Atenolol dose (mg) 72 (32) 67 (31) 0.02 Trandolapril treatment, n (%) 267 (60) 664 (65) 0.07 Trandolapr il dose (mg) 3.3 (2.6) 3.4 (2.6) 0.49 Verapamil SR treatment, n (%) 208 (47) 519 (51) 0.16 Verapamil SR dose (mg) 234 (75) 238 (75) 0.71 Potassium supplementation, n (%) 75 (17) 100 (10) 0.0001 BMI indicates body mass index; HCTZ hydrochlorothia zide; INVEST, INternational VErapamil SR and Trandolapril Study; LVH, left ventricular hypertrophy; MI, myocardial infarction; NOD, new onset diabetes; SR, sustained release. Values are mean standard deviation unless otherwise noted. P values repres ent t tests and chi square tests where appropriate. History of or currently taking lipid lowering medications. ** Average of clinic blood pressure measurements during study.
112 Table 3 5 Association of logistic regression model covariates and new onset diabetes in INVEST Characteristic Odds ratio (95% Confidence Interval) p value Baseline Age ( per 5 years) 0.99 (0.9 2 1.0 8 ) 0.88 Gender (female) 1.0 2 (0.7 5 1 .4 0) 0. 88 BMI ( per 5 kg/m 2 ) 1. 24 (1.0 9 1. 42 ) 0.00 2 Hypercholesterolemia 1.06 (0. 78 1. 42 ) 0.72 History of smoking 1. 17 (0.8 6 1. 61 ) 0.32 History of LVH 1.58 (1.06 2.35) 0.02 Principal component 1 172.85 (0. 14 >999 .99 ) 0. 16 Principal component 2 0.09 (<0.001 456.31 ) 0. 28 Principal component 3 <0.001 (<0.001 844.87 ) 0. 22 During INVEST Systolic blood pressure (mmHg) 1.0 6 (0 .99 1. 13 ) 0. 10 HCTZ treatment duration ( per 6 months) 1. 13 (1.0 7 2.35 ) <0.0001 Atenolol treatment 0.63 (0.26 1.53) 0. 31 Atenolol treatment duration ( per 6 months) 1.0 4 (0.9 0 1.2 1 ) 0.57 Trandolapril treatment 0.68 (0.53 0.8 8) <0.0001 Trandolapril treatment duration ( per 6 months) 0. 53 (0. 47 0. 59 ) <0.0001 Potassium supplementation 2.19 (1.39 3.43) 0.0007 INVEST indicates INternational VErapamil SR and Trandolapril Study ; BMI, body mass index; kg/m 2 kilograms per meter squ ared; LVH, left ventricular hypertrophy; MI, myocardial infarction; mmHg, millimeters of mercury History of or currently taking lipid lowering medications. Average of clinic blood pressure measurements during study.
113 Table 3 6 Significant associa tions for candidate gene tag SNP s on change in fasting glucose in PEAR correction SNP Race/ethnic group Allele (frequency) Beta (SE) HCTZ monotherapy p value Beta (SE) HCTZ add on p value Beta (SE) combined p value KCNJ1 rs17137967 Blacks T (0.05 ) 7.06 (3.49) 0.04 10.97 (3.49) 0.002 8.47 (2.45) 0.0008 Â§ Non blacks T (0.00) ADD1 rs4961 Blacks A (0.06) 1.17 (2.66) 0.59 2.32 (4.62) 0.57 0.05 (2.32) 0.90 Non blacks A (0.21) 2.24 (1.28) 0.06 1.74 (1.06) 0.13 1.99 (0.81) 0.02 ACE rs4303 Blacks A (0.10) 2.40 (2.62) 0.18 7.70 (3.03) 0.02 6.39 (1.92) 0.0007 Â§ Non blacks A (0.00) SNP indicates single nucleotide polymorphism; HCTZ, hydrochlorothiazide; SE, standard error. *All parameter estimates (betas ) and p values represent change in FG per copy of allele adjusted prespecified covariates p values determined using change in log(FG) in linear regressions Previously associated with change in fasting glucose during thiazide treatment Â§ p value signi ficant after FDR correction for all SNP associations within race/ethnic group p value significant at alpha=0.05 representing complementation of previous association
114 Table 3 7 Odds ratios for KCNJ1 SNPs and haplotypes for new onset diabetes during HC TZ treatment by race/ethnicity in INVEST SNP or Haplotype Allele(s) (frequency) OR (95%CI) HCTZ treated OR (95%CI) HCTZ treated OR (95%CI) HCTZ treated OR (95%CI) Not HCTZ treated Interaction p value Whites rs12795437 C (0.07) 2.36 (1.28 4.37) p=0.006 2.43 (1.11 5.34) p = 0.03 2.41 (1.30 4.47) p = 0.005 0.40 (0.13 1.27) p = 0.12 p = 0.002 r s11600347 A (0.07) 2.29 (1.24 4.21) p = 0.008 2.31 (1.06 5.02) p = 0.03 2.33 (1.26 4.30) p = 0.007 0.54 (0.19 1.50) p = 0.23 p = 0.003 HapW1 GCA (0.08) 2.35 (1.27 4.34) p = 0.006 3.09 (1.49 6.41) p = 0.002 2.39 (1.29 4.44) p=0.006 0.42 (0. 13 1.32) p = 0.14 p = 0.002 Hispanics rs658903 A (0.12) 0.38 (0.21 0.69) p = 0.002 0.31 (0.15 0.62) p = 0.0009 0.41 (0.22 0.74) p = 0.003 1.17 (0.55 2.50) p = 0.69 p = 0.03 HapH1 Â§ CATCT (0.09) 2.14 (1.31 3.53) p = 0.003 2.06 (1.23 3.45) p = 0.006 2.06 (1.24 3.41) p = 0.005 0.86 (0.35 2.11) p = 0.74 p = 0.07 HapH2 Â§ TGAGC (0.10) 0.43 (0.24 0.79) p = 0.007 0.39 (0.20 0.76) p = 0.005 0.48 (0.26 0.86) p = 0.01 0.94 (0.42 2.10) p = 0.69 p = 0.18 Blacks rs675388 T (0.18) 3.1 3 (1.45 6.75) p = 0.004 4.16 (1.72 10.03) p = 0.002 3.06 (1.42 6.60) p = 0.004 NE p = 0.63 p = 0.06 HapB1 ** GG (0.70) 0.28 (0.12 0.66) p = 0.003 0.25 (0.11 0.60) p = 0.002 0.29 (0.13 0.67) p = 0.004 NE p = 0.73 p = 0.10 95%CI indicates 95% confidence interval; HCTZ, hydrochlorothiazide; NE, odds ratio not estimable due to sample size; OR, odds ratio; SNP, single nucleotide polymorphism. All point estimates and p values represent increased risk of NOD per copy of allele and are adjusted for age, gender, body mass index, average on treatment systolic blood pressure, history of smoking and hypercholesterolemia, principal components one, t wo, and three, and treatment with trandolapril, atenolol, or potassium supplementation. Interaction p val ues for allele HCTZ treatment. Haplotype HapW1 for INVEST whites inferred from SNPs rs2238009, rs12795437, and rs11600347. Â§ Haplotypes HapH1 and HapH2 inferred from SNPs rs675388, rs1148058, rs658903, rs12795437, and rs3016774. ** Haplotypes inferred from SNPs rs675388 and rs1148059. p value significant after FDR correction for all SNP and haplotype associations within race/ethnic group
115 Figure 3 1. Physiological role of candidate genes in development of hyperglycemia after thiazide diuretic a dministration Red arrows indicate inhibition and blue arrow indicate stimulation. Candidate genes are represented in red boxes. ADD1 indicates alpha adducin 1 gene; DCT, distal convoluted tubule; KCNJ1 potassium inwardly rectifying channel, subfamily J, member 1 gene; ACE angiotensin II converting enzyme gene; AGTR1 angiotensin II type 1 receptor gene.
116 Figure 3 2 FG 1 indicates change in fasting glucose during HCTZ monotherapy; FG 2 change in fasting glucose during HCTZ add on therapy to atenolol; FG, fasting glucose; HCTZ hydrochlorothiazide
117 Figure 3 3 Change in fasting glucose during hydrochlorothiazide treatment by KCNJ1 SNP rs17137967 genotype in black PEAR patient s Bars represent medians and p trend indicates p value for change in log(fasting glucose) during hydrochlorothiazide using an allelic tre nd test adjusted for log ( fasting glucose ) at start of hydrochlorothiazide, age, gender, waist circumference, potassium supplementation, drug arm, average home systolic and diastolic blood pressure, duration of HCTZ treatment, and principal components one, two, and three. P FDR indicates p trend FDR corrected for all SNP s in PEAR blacks. HCTZ indicates hydrochlorothiazide; mg/dL, milligrams per deciliter; SE, standard error.
118 Figure 3 4 Odds ratios per copy of allele and 95% confidence intervals for KCNJ1 SNPs nominally associated (p<0.05) with new onset diabetes during hydrochlorothiazide treatment in INVEST patients by race/ethnicity. All odds ratios are adjusted for age, gender, body mass index, average on treatment systolic blood pressure, hypercholes terolemia, history of smoking, potassium supplementation, principal components one, two, and three, and trandolapril or atenolol treatment. NOD indicates new onset diabetes; SNP, single nucleotide polymorphism.
119 CHAPTER 4 LONG TERM ANTIHYPERT ENSIVE EXPOSU RE AND ADVERSE METABOLIC EFFECTS: PEAR FOLLOW UP STUDY Introduction T2D is a major cause of mortality and morbidity and t he IDF predicts that 552 million individuals worldwide will have T2D by the year 2030 10 Furthermore, a patient with both T2D and hypertension is at a two to three fold risk of an adverse CV outcome compared to a patient with hypertension alone 1 61 Hypertension is also positively associated with hyperlipidemia 180 181 and increases in serum cholesterol increases a 182 184 Hypertension, dyslipidemia, and I FG are components of the metabolic syndrome, which is estimated t o affect 34% of US 1 The presence of metabolic syndrome further increases risk for CV disease compared to patients without metabolic syndrome 185 BP control is an important means of CV risk reduction. 2 Strong evidence fr om randomized clinical trials as summarized in meta analyses and literature reviews, shows that thiazide diuretics contribute to hyperglycemia and hypertriglyceridemia 47 53 186 In addition, thiazide diuretics have been associated with increases in serum uric acid, which has been associated with T2D, 74 the metabolic syndrome, 187 and CV risk. 72 73 188 189 Since thiazide diuretics contribute to hyperglycem ia, hypertriglyceridemia, and hyperuricemia their benefit in a hypertensive patient, including CV risk reduction could be offset by AME s. The long term effects of thiazide diuretics on measures of glucose and lipid metabolism are well studied, but rando mized BP reduction trials consider thiazide induc ed AME s only as a secondary outcome or in secondary analyses. 51 53 190 Clinical trials investigating thiazide diuretics for BP reduction typically do not monitor FG or
120 cholesterol or only monitor FG and cholesterol biannually or annually which may be adequate for clinical monitoring but not for characterization of the short term AME s of thiazides Existing studies have investigated short term effects over a matter of weeks, but their study durations are not sufficient to describe AMEs during long term thiazide tre atment. 114 118 Characterization of AMEs in the short term (after 1 2 months ) might b e useful in predicting AMEs during long term (greater than six months) thiazide treatment. Only limited data are available compar ing short and long term AME s specifically for thiazide diuretic therapy in the same patient population 190 and one such study utilized HCTZ doses up to 200mg, w hich are no longer clinically appropriate. 116 Comparison of AMEs during short and long term thiazide therapy could also clarify whether duration of th iazide treatment is a risk factor for AMEs. In addition, data are lacking for examin ation of the effect of concomitant pharmacotherapy on AMEs during long term thiazide administration in an observational setting. The dysglycemic effects of thiazide diuret ics are typically evaluated using FG and/or T2D diagnosis. However, CV risk may not begin when a patient becomes diabetic. IFG may carr y an adverse prognostic impact 191 192 and several studies have shown that IGT is a better predictor of CV disease 193 and mortality 125 than FG. E GI has also been observed to be a better T2D predictor than IFG and IGT 194 195 and HbA1c has been observed to better predict CV risk than FG. 127 In addition, HOMA is a convenient, noni nvasive indicator of insulin sensitivity that correlates well with insulin sensitivity determined by the hyperinsulinemic euglycemic clamp t he gold standard for in vivo studies of insulin sensitivity 129 130 Despite the potential predictive power for T2D and
121 CV risk with IFG, IGT, EGI, HOMA, and HbA1c analyses of AMEs in BP reduction trials focus on T2D status determined by FG 12 6 mg/dL. In addition, the diagnosis of impaired glucose metabolism or pre diabetes is often difficult. FG may not be a very sensitive measure for metabolic abnormality since values often appear normal in patients with IGT. 119 In individuals with IGT, hyperglycemia may only manifest when the individual is challenged by an OGTT. 128 The OGTT has not been adopted in clinical practice and many randomized controlled trials because of increased cost, time, and inconvenience of the procedure. However, uti lization of OGTT may be clinically valuable in patients who are suspected to have metabolic abnormalities, especially considering the utility of the OGTT for CV risk prediction. 196 In contrast, HbA1c has several advantages over FG and has been recommended by the ADA as a criterion for the diagnosis of T2D. 119 Despite clinical advantages of HbA1c and the OGTT in diagnosing metabolic abnormalities and predicting CV risk, the effects of thiazide diuretics on HbA1c and OGTT values a re not well studied. We designed the PE AR Follow U p S tudy enrolling previous PEAR and PEAR 2 participants who completed either study over six months ago and were continuously treated with a thiazide diuretic during follow up. The PEAR Follow Up Study was designed to determine the effect of dur ation of thiazide treatment and concomitant antihypertensive pharmacotherapy on AMEs during long term thiazide administration in an observational setting. We were also able to test association of AMEs during short term versus long term thiazide treatment. The PEAR Follow Up Study also gathered detailed glycemic characteristics of patients after long term thiazide treatment allowing
122 a comparison of glycemic characteristics obtained in the fasting state and those obtained after OGTT. In addition, we teste d the correlation of change in FG and change in serum potassium during long term thiazide treatment. Methods PEAR and PEAR 2 Study Designs and Populations Details of t he P EAR study which investigated ge netic influences of HCTZ atenolol and their combina tion on BP and AME s are previously published and have been described in detail in Chapter 3. 117 Briefly, p articipants age 17 through 65 years were are randomiz ed to either atenolol or HCTZ, with one dose titration step, followed by assessment of response to therapy after approximately nine weeks on the target dose. The second agent wa s then added followed by similar dose titration and response assessment proced ures for a total study duration of approximately 18 weeks. Biological samples we re collected in the fasting state at baseline at a response assessment after monotherapy, and at a response assessment after combination therapy The PEAR study design allow ed evaluation of change in FG, triglycerides, uric acid, insulin, and HOMA during nine weeks of initial HCTZ administration C hanges in FG and other lab measures after nine weeks of HCTZ treatment were determined during PEAR f ollowing both HCTZ monotherap y and HCTZ add on therapy Change in FG during HCTZ monotherapy was defined as the difference in FG from the baseline visit, prior to the start of HCTZ monotherapy to the end of HCTZ monotherapy. Change in FG during HCTZ add on therapy was defined as th e difference in FG from the start of HCTZ to the end of the trial. PEAR 2 similarly investigated genetic influences on BP and AMEs after administration of the thiazide like diuretic chlorthalidone and the beta blocker
123 metoprolol. PEAR 2 included no random ization and all participants underwent a 3 4 week washout of antihypertensive medication. Patients were then given metoprolol 50 mg i mmediate release (IR) twice daily for two weeks, with dose titration to metoprolol IR 100 mg twice daily (if BP was greate r than 120/70) for at least six weeks. Participants then underwent a 3 4 week washout period followed by chlorthalidone 15 mg once daily monotherapy for two weeks with similar dose titration to chlorthalidone 25 mg once daily for at least six weeks. Inc lusion and exclusion criteria were similar to PEAR. The PEAR 2 study design allowed evaluation of change in FG, triglycerides, uric acid, insulin, and HOMA during approximately eight weeks of chlorthalidone administration. C hange in FG and other lab meas ures after approximately eight weeks of chlorthalidone was defined as the difference in FG from the baseline visit, at start of chlorth a lidone monotherapy to the end of the trial. Neither PEAR n or PEAR 2 incorporated an OGTT or HbA1c so these data w ere not available for any PEAR Follow Up Study participants from their original PEAR or PEAR 2 study period PEAR Follow Up Study Design and Population The PEAR Follow U p S tudy is a n observational, non randomized, o pen label, follow up study of the PEAR and PE AR 2 trials. P revious PEAR and PEAR 2 participants were contacted for willingness to participate in the PEAR Follow U p S tudy if they agreed to be contacted for future research studies in the original PEAR or PEAR 2 informed consent s or during subsequent c orrespondence regarding future research studies Patients were eligible for PEAR F ollow U p S tudy participation if they 1) previously participated in PEAR or PEAR 2 during which HCTZ or chlorth a lidone response data was collected, 2) participated in their f inal PEAR or PEAR 2 study visit at least six months prior to the follow up study visit, and 3) were treated with a thiazide or
124 thiazide like diuretic continuously during the follow up period. Figure 4 1 displays the progression of previous PEAR and PEAR 2 subjects to PEAR Follow Up Study enrollment. Participants were eligible if they were between 17 and 75 years of age and p regnancy was an exclusion All study interventions were approved by the UF IRB and all participants provided written informed consen t for study procedures. The PEAR F ollow U p S tudy is registered on clinicaltrials.gov ( NCT01409434 ). The PEAR Follow U p S tudy consisted of a single study visit for which participants were asked to be in the fasting state not having consumed food or bever ages other than water within eight hours prior to the visit. The study visit include d collection of a medical history and detailed medication history designed to assess the dose and duration of therapy of thiazide diuretics and other antihypertensive med ications t hat might alter metabolic status The interview also included an adherence assessment for antihypertensive treatment. D ata were collected for other medications that alter metabolic status including statin s or other lipid lowering agents, alpha adrenergic agonists, tricyclic antidepressants, corticosteroids, anti diabetic and glucose lowering medications birth control, and potassium supplementation A social history including weekly alcohol consumption and cigarette smoking status was obtain ed A nthropomorphic measurements acquired during the visit includ ed height, weight, and waist and hip circumference Three BP measurements were acquired using an automated sphygmomanometer and averaged for follow up SBP and DBP values A baseline blood d raw was obtained to acquire whole blood FG plasma insulin, HbA1c, a lipid panel, uric acid, and serum potassium. Each participant then drank a 75 gram glucose solution (Azer Scientific, Morgantown, PA) and whole blood glucose
125 measurements were acquired a t one hour and two hour time points af ter ingestion of the solution. PEAR Follow U p S tudy visits were performed at the UF general clinical research center ( GCRC ) or UF Department of Community Health and Family Medicine Clinic s. Whole blood glucose measur ements were analyzed using a n YSI 2300 STAT Plus (YSI, Yellow Springs, OH). Statistical Analysis Patient characteristics at baseline defined as start of thiazide treatment, and at the follow up visit were and paired t tests as appropriate. Whole blood glucose measurements from the PEAR F ollow U p S tudy were converted to plasma adjusted glucose measurements using multipli cation by a factor of 1.11, 119 for comparison to fasting plasma glucose measurements acquired during PEAR and PEAR 2. Long term change in FG was defined as difference between FG at start of thiazide treatment (during PEAR or PEAR 2) and FG at the follow up visit. Linear regression was used to determine variables associated with change in FG, tri glycerides, uric acid, insulin, and HOMA during long term thiazide treatment in univariate analyses L aboratory measures included in univariate model s for each phenotype were baseline and short term change s in FG, LDL, HDL, and total cholesterol, triglyce rides, HOMA, plasma insulin serum uric acid, and serum potassium Independent variables for univariate regressions also included BMI, age, gender, race, baseline systolic and diastolic BP, heart rate, smoking status at follow up alcohol consumption at f ollow up, degree relative), treatment with beta blockers, ACEIs, ARBs, statins o ther lipid lowering
126 agents, alpha adrenergic agonists, selective serotonin reuptake inhibitors (SSRIs), tricyclic antidepressants (TCAs) corticosteroids, anti diabetic medications, birth control, and potassium supplementation and duration of thiazide diu retic therapy. Variables associated with each phenotype in univariate analyses were utilized in stepwi se stepwise regressions variables entered the model at p=0.20 and were retained in the model at p P atients treated with anti diabetic medication were excluded from analyses in order to eliminate confounding of anti diabet ic treatment and T2D on AME phenotypes Similar stepwise linear regression s were used to test association of independent variables with HbA1c, one and t wo hour OGTT glucose and OGTT area under the curve (AUC) at follow up Similar s tepwise logistic reg ressions following univariate regressions were used to model dichotomous variables at follow up, including IFG, IGT, EGI, and T2D Spearman partial correlations were used to test correlation of long term change in FG versus duration of thiazide trea tment, adjusted for baseline FG. Spearman partial correlations were also used to determine correlation of short term versus long term change in FG adjusted for baseline FG and to determine long term change in FG versus long term change in serum potassiu m adjusted for baseline FG serum potassium and potassium supplementation All s tatistical analys e s were performed u sing SAS software (SAS, Cary, NC) and all data was collected and stored using REDCap software. 197 Results PEAR Follow Up Study Population Characteristics at Baseline versus Follow Up A total of 4 4 patients participated in PEAR Follow Up Study interventions f our of which were excluded from analys is due to treat ment with anti diabetic medication.
127 (Figure 4 1) A total of 40 participants were included in analysis, including 29 patients (73%) who participated in PEAR and 1 1 patients (27%) who participated in PEAR 2. (Table 4 1) For those that previo usly participated in PEAR, 1 6 (5 5 %) were originally randomized to the atenolol strategy and 13 (4 5 %) were randomized to the HCTZ strategy The PEAR F ollow U p S tudy population who were an average 51 years of age at follow up included 2 4 (6 0 %) females and 29 (7 3 %) non black individuals. Compared to baseline values, participants at follow up had increased age ( p<0.0001 ), waist circumference (p=0.002), HDL cholesterol (p<0.0001), uric acid (p<0.0001), consumed more alcoholic beverages per week in participa nts who consumed alcoholic beverages (p=0.04), and had a greater percentage of patients with abdominal obesity (p=0.03) (Table 4 1) At follow up, patients had decreased systolic BP (p=0.003), diastolic BP (p=0.02), and serum potassium (p=0.00 3 ). N on sig nificant increase s w ere observ ed in FG (p=0.16) and triglycerides (p=0.59) compared with baseline Characteristics of PEAR Follow Up Study Population at Follow Up The mean duration of follow up was 2 9 (SD 19) months ranging from 8 to 72 months of thiazid e diuretic treatment which included thiazide treatment during PEAR or PEAR 2 in addition to the follow up period (Table 4 2) All patients were treated with a thiazide or thiazide like diuretic, with t en participants (25%) t reated with chlorthalidone and 3 0 (75%) t reated with HCTZ during the observational follow up period The median dose of both HCTZ and chlorthalidone was 25 mg T en participants (2 5 %) were taking thiazide monotherapy with either HCTZ or chlorthalidone 1 2 (3 0 %) were also taking a beta blocker, 1 3 (3 3 %) were also taking an ACEI and 5 ( 13 %) were also taking a beta blocker and an ACEI. Two participants (5%) were taking a CCB at follow up.
128 With regard to other pharmacotherapy, 5 participants (13%) were treated with potassium supplementat ion during follow up. (Table 4 2) Nine participan ts (25%) were taking a statin and one participant was treated with a fibrate Three participants (8%) were taking a n SSRI, two (5%) were taking a TCA, and one (3%) was taking long term corticosteroid treat ment. Following OGTT, the mean one hour plasma adjusted glucose was 157 (SD 47) mg/dL and the mean two hour plasma adjusted glucose was 132 (SD 51) mg/dL. (Table 4 2) S ix teen participants ( 40 %) had IGT and 18 participants had EGI (4 0 %) M ean HbA1c was 5 .68 (SD 0.44) % in PEAR Follow U p S tudy participants Change in FG during Short Term versus Long Term Thiazide Treatment Mean FG increased from 91 (SD 12) mg/dL to 97 (SD 16) mg/dL during short term thiazide treatment in PEAR and PEAR 2, constituting a s ignificant 6.5 (SD 13.0) mg/dL FG increase (p=0.005). PEAR F ollow U p S tudy participants had a mean 91 [SD 12] mg/dL at baseline and 94 [SD 13] mg/dL a t follow up representing a non significant increase in FG (3.6 [SD 15.3] mg/dL, p=0.16) during long term thiazide treatment on average. No significant correlation was observed between short term change in FG and long term change in FG (r=0.16, p=0.38). (Figure 4 2) When PEAR F ollow U p S tudy participants were divided according to antihypertensive drug treat ment status a significant 8.8 (SD (14.8) mg/dL increase in FG was observed during follow up in beta blocker treated patients (p=0.02) The long term change in FG in beta blocker treated participants was also significantly different from the change in FG in participants not treated with a beta blocker (p=0.05). (Table 4 3) N o significant change was observed in FG in pa rticipan ts treated with ACEIs Long term change in FG was not significantly different by type of thiazide treatment (p=0.29), ACEI treatme nt status (p=0.18), or statin treatment status (p=0.30).
129 Change in FG during the observational follow up period is displayed by concomitant antihypertensive treatment in Figure 4 3. Figure 4 3 indicates that, after short term treatment with thiazides, p atients treated with a thiazide plus a beta blocker displayed an increased mean FG compared to patients treated wi th a thiazide alone, a thiazide plus an ACEI, or triple therapy with a thiazide, beta blocker, and ACEI. Stepwise Linear Regression of Chang e in Lab Measures during Long Term Thiazide Treatment Univariate associations of change in FG during long term thiazide treatment are listed in Table 4 4 Stepwise linear regression modeling of long term FG changes indicated a significantly increased FG with longer duration of thiazide treatment (p=0.008) and lower baseline FG (p = 0.0 2 ) (Table 4 4 ) The full model R 2 for change in FG during long term thiazide treatment was 0. 4 5 C hange in FG was positively correlated with duration of thiazide treatment ( r=0.4 7 p=0.00 4 ), which i s consistent with a positive parameter estimate in stepwise regression (Figure 4 4 ) Increased HOMA during long term thiazide treatment was associated with longer duration of thiazide diuretic treatment (p=0.009) with model R 2 =0. 19 (Table 4 5 ) Increased insulin was associated with lower baseline insulin (p<0.0001) and white race (p=0.01) with model R 2 =0.59. (Table 4 6 ) Increased triglycerides during long term thiazide treatment w as associated with decreased baseline triglycerid es (p<0.0001) with model R 2 =0.67. (Table 4 7 ) After stepwise regression, n o significant associations were found for change in uric acid during long term thiazide treatment. (Table 4 8 ) No patients were treated with uric acid lowering medications such as allopurinol during the follow up period. Stepwise regressions of change in serum potassium, LDL, HDL, total
130 cholesterol and serum potassium during long term thiazide treatment are presented in the Appendix B (Table B 1, Table B 2, Table B 3, and Table B 4 respectively) Stepwise Linear Regression of Lab Measures at Follow Up Visit after Long Term Thiazide Treatment Variables associated with higher adjusted plasma adjusted FG at follow up were family history of T2 D in a first degree relative (p=0.03) and chlorthalidone treatment (p=0.05) with full model R 2 =0.22. (Table 4 9 ) Increased two hour OGTT plasma adjusted glucose at follow up was associated with increased D BP during short term thiazide treatment (p=0.004), older age (p=0.02), and concomitant beta blocker treatment ( p =0.0 3) with R 2 =0.34. (Table 4 10 ). Beta blocker treatment was associated with a 33 mg/dL increase in two hour OGTT plasma adjusted glucose compared to participants without beta blocker treatment N o significant associations were ident ified for HbA1c (Table 4 1 1 ) OGTT AUC (Table 4 1 2 ) or one hour OGTT glucose (Table 4 1 3 ) at follow up Correlation of C hange FG and Change in Serum Potassium during Follow Up A trend towards a significant inverse correlation between change in FG and chan ge in serum potassium was observed after controlling for baseline FG, baseline serum potassium, and potassium supplementation (r= 0. 33 p=0.0 6 ). (Figure 4 5 ) Non significant negative correlations were also observed by potassium supplementation status, chl orthalidone and HCTZ treatment, and by ACEI treatment status. Evaluation of IFG, IGT, EGI, and T2D At follow up, IFG was present in 1 1 participants ( 28 %) (Table 4 1) Five of 11 also had IFG at baseline and six had newly developed IFG (Table 4 2) IGT wa s present in 1 6 participants ( 40 %) and 18 participants had EGI (4 0 %) Agreement between IFG and
131 IGT was 73 %, agreement between IFG and EGI was 73%, and agreement between IGT and EGI which both use post load glucose measurements rather than FG was 8 5 %. Seven PEAR F ollow U p S tudy participants were observed to have only one of the three classifications and e ight were observed to have comorbid IFG, IGT, and EGI (Figure 4 6 ) A fter stepwise logistic regression modeling, no significant associations were iden tified for IFG, IGT, EGI, or T2D Discussion In t he PEAR Follow Up Study longer duration of thiazide treatment and lower baseline FG were associated with increas ed FG after long term thiazide treatment. Change in FG during short term thiazide treatment was not associated or correlated with change in FG during long term thiazide treatment. B eta blocker treated patients had larger mean FG increases during follow up and increased two hour glucose following OGTT A significant increase in uric acid was ob served after long term thiazide treatment, however no associat ions with increased uric acid were observed We also observed no significant increase in t riglycerides after long term thiazide treatment Few associations were observed for OGTT values and ag reement between IGT and EGI was higher than between IFG and IGT or IFG and EGI, suggesting that values acquired after glucose challenge are distinct from those acquired in the fasting state. In addition, change in FG was not significantly correlated with ch ange in serum potassium during long term thiazide treatment. The strong association and significant correlation between longer thiazide treatment duration and long term c hange in FG during long term thiazide treatment suggest s that longer treatment durat ion with a thiazide results in increasing FG levels. Our observations occurred over a large range of follow up periods ranging from eight to
132 72 months. PEAR Follow Up Study data suggest that longer treat ment is a risk factor for increased FG in an observ ational setting. An increasing FG with longer thiazide treatment is consistent with previous literature, in which increasing FG is observed during thiazide treatment 49 51 The impact of duration of thiazide treatment is further supported by the association of duration of thiazide treatment with increas ed HOMA, indicating worsening insulin sensitivity w ith longer thiazide treatment. In the PEAR Follow Up Study, increased FG during long term thiazide treatment was strongly associated with lower baseline FG. A similar negative parameter estimate was obser ved for change in FG and baseline FG after four weeks of HCTZ treatment in GERA (n=585). 114 In GERA, similar stepwise linear regression models wer e used to determine associations with change in FG during four weeks of HCTZ. The inverse relationship between baseline FG and change in FG has now been observed in analysis of both short and long term thiazide treatment A n inverse relationship was obse rved in both GER A and the PEAR Follow Up Study between baseline triglycerides and change in triglycerides during thiazide treatment. In the GERA study, stepwise models explained relatively small amounts of variability in FG during thiazide treatment (R 2 = 0.11). 114 The PEAR F ollow U p S tudy model had a high er R 2 (0.45) despite a smaller sample size. Similarly high model R 2 values were observed for c hanges in triglycerides in the PEAR Follow Up Study versus GERA. 114 The main difference in study design is the greater duration of thiazide treatm ent in the PEAR Follow Up Study versus GERA. Since thiazide treatment duration was a significant predictor of FG changes, our results suggest that a greater amount of variability in FG can be explained over long term thiazide therapy, but must
133 be interpre ted with caution considering the relatively small sample size of the PEAR Follow Up Study population. In the PEAR Follow Up Study, s hort term changes in FG during thiazide treatment were not associated with or correlated with long term changes in FG during thiazide treatment. The lack of association suggests that short term monitoring of FG during initial thiazide administration are insufficient to determine whether long term thiazide induced dysglycemia will develop. Baseline FG and duration of thiazide therapy may be more important determinants of AMEs than short term monitoring during thiazide treatment. PEAR Follow Up Study data suggest that acquiring a FG level approximately nine weeks after start of thiazide treatment would not be clinically informa tive or determine severity of long term changes in FG. The PEAR Follow Up Study observed a 6.5 (SD 13.0) mg/dL FG increase during nine weeks of HCTZ treatment and a 3.6 mg/dL FG increase after long term ( average 29 months ) HCTZ treatment The observation of a large FG increase during short term HCTZ treatment compared to a FG increase that is disproportionate in terms of treatment duration i s consistent with previous literature. One previous study observed a 4.0 mg/dL FG after 10 weeks of HCTZ treatment a nd a 4.7 mg/dL increase after one year of HCTZ treatment although this study utilized up to 200mg daily doses of HCTZ 116 Another study observed a 0. 59 mmol/L (10.6 mg/dL) FG increase after 4 6 months of HCTZ treatment and a 0.46 mmol/L (8.3 mg/dL) FG increase after 24 months of HCTZ treatment. 190 These data suggest that an exaggerated change in FG occur s during short term thiazide treatment. Despite exaggerated FG increases in the short term, HCTZ treatment duration wa s associated with long term FG increases in the
134 PEAR Follow Up Study However, our results suggest that patients with large FG increases during short term HCTZ treatment will not necessarily experience large FG increa ses during long term HCTZ treatment PEAR Follow Up Study results support an effect of concomitant antihypertensive pharmacotherapy on AMEs P articipants treated with thiazides and beta blockers had greater long term increase s in FG Beta blocker treatme nt was also associated with increased two hour OGTT glucose. Beta blocker treatment has been previously associated with T2D 47 and may impair insulin release through its action on the beta 1 receptor, 68 causing increases in fasting and post prandial glucose. In addition, chlorthalidone was associated with higher FG at follow up, suggesting that chlorthalidone treatment may be associated with greater FG increases than the much more commonly prescribed HCTZ. T he potency of chlorthalidone versus HCTZ has not been well studied in terms of AMEs, but chlorthalidone treatment has been associated with new onset diabetes 49 and recent evidence suggests that chlorthalidone is more potent than HCTZ regarding BP and CV event reduction, hypokalemia, and hyperuricemia 1 98 199 Although two hour glucose following an OGTT was associated with older age, beta blocker treatment, and increased diastolic BP, n o signific ant associations were found for OGTT AUC or one hour OGTT glucose, suggesting that available baseline data in PEAR and PEAR 2 do not model these phenotypes after long term thiazide treatment The paucity of associations for OGTT related measure s might hav e been due to the fact that no OGTT data was acquired at baseline in PEAR or PEAR 2 The lack of associations with OGTT values from baseline data support s the notion that
135 measurements obtained during glucose challenge are distinct from glucose m easurement s in the fasting state 122 125 Th e limitation of fasting measurements is clinically concerning considering evidence suggesting improvement in T2D prediction and CV risk using glucose challenge measurements. 123 125 196 A greater sensitivity of glucose challenge measurements is also suggested by the greater percentage of PEAR F ollow U p S tudy patients that EGI and IGT versus IFG The PEAR F ollow U p S tudy observed a l a ck of correlation between FG changes and potassium changes during long term thiazide treatment. The relationship between potassium and glucose is well described and has been discussed at length in previous chapters 46 75 While some evidence suggest s that potassium depletion is associated with increased FG during thiazide treatment 76 78 a n analysis from the PEAR study observed no correlation between change FG a nd change in serum potassium during short term thiazide treatment, using individual patient data rather than aggregate pooled data. 79 While further research is re quired, data from PEAR during both short and long term thiazide treatment suggest a lack of correlation between serum potassium and FG The lack of correlation calls into question the idea that thiazide induced hyperglycemia is directly related to potassi um depletion and suggests that t he pathophysiology of thiazide induced dysglycemia is more complex than a simple inverse relationship between potassium and glucose. Th e strengths of this study are the inclusion of thiazide response data, including glucose, lipid, and electrolyte related measurements, after both short term and long term thiazide treatment in a contemporary hypertensive cohort The PEAR F ollow U p S tudy also included detailed information regarding concomitant drug therapy, including
136 antihyper tensive, anti diabetic, statin therapy and other medications known to affect glucose homeostasis Detailed pharmacotherapy data allowed evaluation of concomitant drug treatment effects on AMEs during long term thiazide treatment in an observational setti ng. The majority of studies of thiazide induced AMEs come from large scale hypertension treatment trial data. Our analysis of data during non randomized treatment outside of a clinical trial may better reflect actual participant practices and be more gen eralizable to a broader patient population We were also able to evaluate patients using an OGTT to evaluate detailed glycemic characteristics of PEAR F ollow U p S tudy participants after long term thiazide treatment. Our study has several limitations worth y of mention. The limited sample size may have contributed to a lack of significant differences in FG and lipid measures during long term thiazide treatment and may also have resulted in inflated model R 2 values in stepwise regressions Although our R 2 v alues were high, predictors of long term FG and triglycerides were consistent with previously published models in a much larger patient population. 11 4 Sample size may also have affected our ability to detect a ssociations with dichotomous outcome variables such as IFG, EGI, and IGT Furthermore, sample size limited our ability to determine effects of all class es of medications known to affect glucos e homeostasis. However, we were able to detect effects of beta blocker and chlorthalidone treatment on two hour OGTT glucose and FG at follow up respectively. Whole blood FG values at follow up were also adjusted to reflect plasma FG measurements at fol low up which might contribute to variability in the change in FG phenotype. However, the utilized 11% plasma adjusted value corresponds well with
137 plasma measurements and is endorsed by the ADA 119 No OGTT data was available from PEAR and P EAR 2 which prevented analysis comparing short and long term effects of thiazides on glucose tolerance In addition, compliance to drug therapy during follow up was assessed via interview, which may not reflect actual compliance with chronic medications However, changes in several objective parameters strongly suggest a compliant population in terms of thiazide treatment and other antihypertensive therapies including a significant reduction in systolic and diastolic BP, decrease in serum potassium and increase in serum uric acid. Finally, detailed data regarding diet and exercise was not collected for PEAR Follow Up Study patients. Therefore, we were not able to detect potential effects of changing habits in diet and exercise on glucose and lipid rela ted measurements in our population. I ncreases in participant weight and BMI, which might indicate worsening diet and exercise habits, were observed although not significantly different over the follow up period Summary and Significance In the PEAR Follo w Up Study, we observed a strong association and correlation between duration of thiazide therapy and increasing FG during long term thiazide treatment. The association of duration of treatment with increasing FG and HOMA may ha ve clinical implications si nce antihypertensive therapy with thiazide diuretics is typically lifelong. Concomitant antihypertensive treatment with beta blockers was associated with larger mean FG increases during follow up and increased two hour OGTT glucose at follow up, suggestin g that the antihypertensive combination of thiazides and beta blockers should be avoided for patients in whom glucose levels are of concern
138 PEAR Follow Up Study data suggest that change in FG during short term thiazide treatment i s not associated nor co rrelated with change in FG during long term thiazide treatment. The lack of association suggests that acquiring a FG level approximately nine weeks after start of thiazide treatment may not be as clinically informative as monitoring glucose consistently t hroughout therapy The PEAR Follow Up Study results also supported the distinction between measures of glucose tolerance versus measures of fasting glucose a nd the lack of correlation between change in FG and change in serum potassium during long term thi azide treatment.
139 T ab le 4 1. Characteristics of PEAR F ollow U p S tudy participants at baseline and at follow up Characteristic Baseline (n= 40 ) Follow up (n= 40 ) P value PEAR participants 29 ( 73 %) PEAR 2 participants 1 1 (27%) Gender (female) 2 4 ( 60 %) Race (Nonblack) 29 (7 3 %) Age (years) mean (SD) 49 (10) 51 (10) <0.0001 Weight (kg), mean (SD) 89.8 (17.7) 91.7 (17.3) 0.0 9 Body mass index (kg/m 2 ) mean (SD) 31. 1 ( 5.6 ) 31.8 (5.4) 0.0 9 Waist circumference (cm) mean (SD) 9 8 (1 4 ) 10 2 (16) 0.00 2 Abdominal Obesity Â§ 2 2 (5 5 %) 27 ( 68 %) 0.03 SBP, mean (SD) ** 1 42 (16) 131 (1 4 ) 0.0 0 3 DBP, mean (SD) ** 8 9 (9) 83 (10) 0. 0 2 Consumes alcoholic beverages 18 (45%) 22 (55%) 0.18 Drinks per week 5 ( 7 ) 8 ( 10 ) 0.0 4 Current smoker 10 (23%) 7 (1 6%) 0. 08 Statin treatment 4 (9%) 9 (2 3 %) 0.0 6 Fasting Glucose (mg/dL) 91 (12) 94 (13) 0.16 Total cholesterol (mg/dL) 198 ( 40 ) 200 ( 40 ) 0. 63 LDL cholesterol (mg/dL) 12 2 (36) 11 7 (36) 0. 35 HDL cholesterol (mg/dL) 4 5 (1 4 ) 52 (17) <0.0001 Triglycerides (mg/dL) 15 4 (1 21 ) 15 7 (8 9 ) 0.5 9 Serum potassium (mEq/L) 4.4 9 (0.4 8 ) 4.1 8 (0. 49 ) 0.00 3 Uric acid (mg/dL) 5.9 (1.5) 6.6 (1.6) <0.0001 1 5.8 ( 12.6 ) 15.2 (10.2) 0. 85 HOMA 2. 73 (1. 7 9) 3. 3 2 (2.58) 0.15 Impaired fasting glucose 1 1 (28 %) 1 1 (28 %) 0.99 SD indicates standard deviation; kg, kilograms; kg/m 2 kilograms per meter squared; cm, centimeters, SBP, systolic blood pressure; DBP, diastolic blood pressure; mg/dL, milligrams per deciliter; LDL, low density lipoprotein; HDL, high density lipoprotein; mEq/ L, milliequivalents per liter; microunits per milliliter, ; HOMA, homeostatic model assessment Values indicate n (%) unless otherwise stated. Baseline is defined as start of thiazide diuretic treatment during PEAR or PEAR 2. aseline and follow up. Â§ Abdominal obesity inches for males ** Average of home BP measurements for baseline and three clinic BP measurements for follow up study visit Impaired fasting glucose defined as fasting glucose 100 mg/dL
140 Table 4 2. C haracteristics of PEAR F ollow U p S tudy participants during follow up period. Characteristic at follow up Number of participants* (n=4 0 ) Drug Treatment Characteristics Duration of thiazide treatment (m onths), mean (SD) 2 9 (19) Hydrochlorothiazide treatment 30 (75%) Chlorthalidone treatment 1 0 (25%) Hydrochlorothiazide dose, median (IQR) 25 (25 25) Chlorthalidone dose, median (IQR) 25 (25 25) Thiazide monotherapy 10 (2 5 %) Beta blocker + thiazide 1 2 ( 30 %) ACEI + thiazide 1 3 (3 3 %) Beta blocker + ACEI + thiazide 5 ( 13 %) Calcium channel blocker treatment 2 (5%) Statin treatment 9 (2 3 %) Fibrate treatment 1 (3%) Potassium supplementation 5 (13%) SSRI treatment 3 (8%) Tricyclic antidepressant t reatment 2 (5%) Corticosteroid treatment 1 (3%) Anticonvulsant treatment 2 (5%) Glycemic Characteristics HbA1c (%), mean (SD) 5. 68 (0. 44 ) One hour OGTT glucose (mg/dL), mean (SD) 15 7 ( 47 ) Two hour OGTT glucose (mg/dL), mean (SD) 13 2 (5 1 ) OGTT AUC (mg/dL h), mean (SD) 27 3 ( 74 ) IGT (2 h our OGTT glucos 1 6 ( 40 %) EGI (1 h our 1 8 (4 5 %) New onset IFG 6 (15%) SD indicates standard deviation; ACEI, Angiotensin I converting enzyme inhibitor; SSRI, selective serotonin reuptake inhibitor; IQR, interquartile range ; OGT T, oral glucose tolerance test; mg/dL, milligrams per deciliter; AUC, area under the curve; IGT, impaired glucose tolerance; EGI, elevated glucose intolerance ; IFG, impaired fasting glucose R epresented as number (percentage) unless otherwise not ed Data acquired during 2 hour oral glucose tolerance test after 75 gram glucose load follow up who did not have impaired fasting glucose at baseline
141 Table 4 3. Fasting glucose levels at baseline and at follow up by drug treatment status Drug treatment* FG at baseline FG at follow up p value Type of thiazide HCTZ (n= 30 ) 90 (12) 92 (12) 0.29 C hlorthalidone (n= 10 ) 93(12) 100 (14) Beta blocker Beta blocker treated (n=17) 90 (13) 98 (12) 0.05 No beta blocker (n=23) 92 (12) 91 (13) ACEI ACEI treated (n=18) 90 (14) 90 (10) 0.18 No ACEI (n=22) 92 (10) 97 (15) Statin Statin treated (n=9 ) 87 (17) 95 (13) 0.30 No statin (n=3 1 ) 92 (11) 94 (13) FG indicates fasting glucose; HCTZ, hydrochlorothiazide; ACEI, angiotensin I converting enzyme inhibitor. Excludes individuals treated with anti diabetic drugs with t he exception of the anti diabetic drug heading p value for paired t test for difference between groups in change in FG between baseline and follow up Table 4 4 Variables associated with FG change s during long term thiazide treatment Ind ependent variabl e Parameter p value Univariate associations (p<0.20) Baseline FG, mg/dL 0.7 0 0.00 3 Change in FG during short term thiazide treatment, mg/dL 0.28 0.16 Gender (female) 6.56 0.20 Duration of thiazide treatment months 0.48 0.0002 Beta bl ocker treatment 9.76 0.05 ACEI treatment 6.74 0.18 Current smoker 9.37 0.15 Family history of T2D 8.11 0.11 Baseline SBP mmHg 0.23 0.17 Stepwise Results (R 2 =0.45) Duration of thiazide treatment months 0.34 0.008 Baseline FG, mg/dL 0.46 0 .02 FG indicates fasting glucose; mg/dL, milligrams per deciliter; ACEI, angiotensin I converting enzyme inhibitor; T2D, type 2 diabetes; SBP, systolic blood pressure. p values determined using linear regression e xclud ing patients with anti diabetic tre atment
142 Table 4 5 Variables associated change in HOMA changes during long term thiazide treatment Independent variable Parameter p value Univariate associations (p<0.20) Race (white) 1.51 0.08 Duration of thiazide treatment months 0.06 0.009 Current smoker 1.90 0.07 Baseline Insulin, U/mL 0.07 0.04 Stepwise Results (R 2 =0.19) Duration of thiazide treatment months 0.06 0.009 HOMA indicates homeostatic model ass essment; U/mL, microunits per milliliter *p values determined using linear regression excluding patients with anti diabetic treatment. Table 4 6 Variables associated change in insulin changes during long term thiazide treatment Independent variable Pa rameter p value Univariate associations (p<0.20) Race (white) 8.18 0.06 Duration of thiazide treatment months 0.37 0.001 Beta blocker treatment 5.89 0.17 Chlorthalidone treatment 6.70 0.18 Current smoker 8.82 0.11 Statin treatmen t 12.28 0.01 Change in potassium during short term thiazide treatment, mEq/L 6.62 0.06 Baseline uric acid, mg/dL 2.51 0.08 Baseline insulin, U/mL 0.72 <0.0001 Change in insulin during short term thiazide treatment, U/mL 0.66 0.009 Stepwise R esults (R 2 =0.59) Baseline insulin, U/mL 0.73 <0.0001 Race (white) 8.42 0.01 HOMA indicates homeostatic model assessment; U/mL, microunits per milliliter p values determined using linear regression excluding patients with anti diabetic treatment.
143 Table 4 7 Variables associated with triglyceride changes during long term thiazide treatment Independent variable Parameter p value Univariate associations (p<0.20) Beta blocker treatment 38. 59 0.16 Abdominal obesity 48.22 0.08 Cha nge in LDL during short term thiazide treatment mg/dL 0.89 0.17 Baseline H DL, mg/dL 1.63 0.10 Change in HDL during short term thiazide treatment, mg/dL 4.18 0.13 Change in uric acid during short term thiazide treatment, mg/dL 27.87 0.05 Baseline tri glycerides mg/dL 0.4 6 <0.0001 Change in triglycerides during short term thiazide therapy mg/dL 0.4 5 0.00 2 Stepwise Results (R 2 =0. 45 ) Baseline triglycerides, mg/dL 0.46 <0.0001 LDL indicates low density lipoprotein; mg/dL, milligrams per decil iter; T2D, type 2 diabetes; OGTT, oral glucose tolerance test; AUC, area under the curve; BP, blood pressure; U/mL, microunits per milliliter *p values determined using linear regression excluding patients with anti diabetic treatment. Table 4 8 Varia bles associated with uric acid changes during long term thiazide treatment Independent variable Parameter p value Univariate associations (p<0.20) Baseline triglycerides, mg/dL 0.002 0.19 Stepwise Results: no significant associations Mg/dL indicates m illigrams per deciliter. p values determined using linear regression excluding pati ents with anti diabetic treatment.
144 Table 4 9 Variables associated with FG at follow up Independent variable Parameter p value Univariate associations (p<0.20) Baseline FG, mg/dL 0.30 0.09 Duration of thiazide treatment months 0.18 0. 12 Beta blocker treatment 7.78 0.06 ACEI treatment 7.30 0.08 Chlorthalidone treatment 7.73 0.10 Family history of T2D 7.43 0.07 BMI kg/m 2 0.48 0.20 Baseline HDL mg/dL 0.22 0.17 Baseline DBP m mHg 0.37 0.12 Stepwise Results (R 2 =0.22) Fam ily history of T2D 9.31 0.03 Chlorthalidone treatment 9.59 0.05 FG indicates fasting glucose; mg/dL, milligrams per deciliter; ACEI, angiotensin I converting enzyme inhibitor; T2D, type 2 diabetes; BMI, body mass index; HDL, high density lipoprotein ; DB P, diastolic blood pressure p values determined using linear regression excluding patients with anti diabetic treatment in a first degree relative
145 Table 4 10 Variables associated with two hour OGTT glucose at follow up In dependent variable Parameter estimate p value Univariate associations (p<0.20) Age, years 1.24 0.13 Duration of thiazide treatment months 0.74 0.10 Beta blocker treatment 27.93 0.09 Abdominal obesity 22.22 0.17 Baseline potassium, mEq/L 31.53 0.08 Baseline L DL mg/dL 0.36 0.13 Baseline total cholesterol, mg/dL 0.39 0.06 Change in SBP during short term thiazide treatment, mmHg 1.22 0.09 Change in DBP during short term thiazide treatment, mmHg 2.16 0.10 Stepwise Results (R 2 =0.34) Change in DBP during short term thiazide treatment, mmHg 4.64 0.004 Age years 1.90 0.02 Beta blocker treatment 33.10 0.03 OGTT indicates oral glucose tolerance test; mEq/L, milliequivalents per liter; LDL, low density lipoprotein; mg/dL, milligrams per deciliter; mmHg, mi llimeters of mercury. p values determined using linear regression excluding patients with anti diabetic treatment Abdominal obesity inches for males Table 4 1 1 Variables associated with HbA 1c at follow up Independent variable Parameter p value Univariate associations (p<0.20) Beta blocker treatment 0.22 0.12 Family history of T2D 0.19 0.18 Statin treatment 0.23 0.18 Baseline LDL, mg/dL 0.003 0.16 Baseline total cholest erol, mg/dL 0.003 0.13 Baseline DBP mmHg 0.01 0.15 Stepwise Results: no significant associations T2D indicates type 2 diabetes; LDL, low density lipoprotein; mg/dL, milligrams per deciliter; DBP, diastolic blood pressure p values determined us ing linear regression excluding patients with anti diabetic treatment.
146 Table 4 1 2 Variables associated with OGTT AUC at follow up Independent variable Parameter p value Univariate associations (p<0.20) Baseline FG, mg/dL 1.39 0.18 Age, years 1.72 0.14 Duration of thiazide treatment months 1.26 0.06 Beta blocker treatment 42.70 0.07 Baseline potassium, mEq/L 36.13 0.17 Stepwise Results: no significant associations OGTT indicates oral glucose tolerance test; FG fasting glucose; mg/dL, milligrams per deciliter; mEq/L, milliequivalents per liter. p values determined using linear regression excluding patients with anti diabetic treatment. Table 4 1 3 Variables associated with one hour OGTT glucose at follow up Independent variable Parameter p value Univariate associations (p<0.20) Baseline FG, mg/dL 1.00 0.13 Age, years 1.05 0.16 Duration of thiazide treatment months 0.77 0.07 Beta blocker treatment 25.28 0.10 Chlorthalidone treatment 27.45 0.11 Baseline pota ssium, mEq/L 22.96 0.18 Change in potassium during short term thiazide treatment mEq/L 18.83 0.14 Stepwise Results: no significant associations OGTT indicates oral glucose tolerance test; FG fasting glucose; mg/dL, milligrams per deciliter; mEq/ L, milliequivalents per liter. p values determined using linear regression excluding patients with anti diabetic treatment.
147 Figure 4 1. Progression of subjects for PEAR Follow Up Study e nrollment and analysis UF indicates University of Florida; PCOS pol ycystic ovary syndrome. *Less than six months of follow up time prior to PEAR Follow Up Study enrollment
148 Figure 4 2. Change in fasting plasma glucose during short term versus long term thiazide diuretic treatment. P value and r calculated using Spe arman partial correlation adjusted for baseline fasting glucose.
149 Figure 4 3 Mean fasting plasma glucose at baseline, end of short term thiazide treatment, and end of long term thiazide treatment by antihypertensive therapy. Error bars represent stand ard deviations. ACEI indicates angiotensin I converting enzyme inhibitor.
150 Figure 4 4 Change in fasting plasma glucose during long term thiazide diuretic treatment versus duration of follow up. P value and r calculated usin g Spearman partial correlat ion adjusted for baseline fasting glucose
151 Figure 4 5 Change in fasting plasma glucose versus change in serum potassium during long term thiazide diuretic treatment. P value and r calculated using Spearman partial correlation adjusted for baseline fas ting glucose baseline serum potassium and treatment with potassium supplementation
152 Figure 4 6 Venn diagram of participants with IFG, IGT, and/or EGI IFG indicates impaired fasting glucose, IGT indicates impaired glucose tolerance, EGI indicates elevated glucose intolerance.
153 CHAPTER 5 SUMMARY AND CONCLUSI ONS Thiazide diuretics are first line agents for the treatment of hypertension, but are associated with T2D and hyperg lycemia Little knowledge currently exists to identify individuals at risk f or thiazide induced dysglycemia T his research utilize d several phenotypes along a continuum of dysglycemia including change in FG and NOD as well as multiple methodological approaches to identify risk factors for thiazide induced dysglycemia For this research, we focused on the identification of genetic risk factors from TCF7L2 pharmacogenetic risk factors from the candidate genes KCNJ1 ADD1 ACE and AGTR1 and pharmacotherapeutic risk factor s, including duration of thiazide treatment In the resea rch described in Chapter 2, we sought to identify SNPs in TCF7L2 as genetic risk factors for T2D in African and Hispanic ethnic/race groups and to investigate the impact of race/ethnicity on SNP associations with T2D Genetic risk factors for T2D are well studied from GWAS in European populations, but studies continue to be limited in individuals of African and Hispanic descent. We attempted to identify T2D predictor SNPs using TCF7L2 sequence data for NO D case s and age, gender, and race/ethnicity matched control s from INVEST GENES. After performing association analysis in sequenced individuals and genotyping potential predictor SNPs in a larger case/control cohort, w e found no novel T2D risk predictor SNPs in black or Hispanic individuals W e also did n ot confirm associations between NOD and existing T2D predictor SNPs from European populations, regardless of race/ethnicity in INVEST GENES.
154 Our result s do not preclude the existence of yet unknown T2D predictor SNPs from TCF7L2 in black and Hispanic popul ations since our ability to confidently identify true associations is limited by the small number of sequenced samples among each race/ethnic group. However, o ur results do add previously unknown polymorphisms to available data on TCF7L2 variation and fu rther describe d LD structure particularly in Hispanic and black s. To our knowledge, the described sequencing project represents the most comprehensive evaluation of TCF7L2 variation in black and Hispanic (primarily Puerto Rican ) individuals that is curren tly available. Considering the large number of variants identified and the low MAF of novel variants, type 2 error may have prevented identification of T2D predictor SNPs. Further research, including deep sequencing in diverse populations with larger sam ple sizes and investigation of cumulative contribution of rare variants, is needed to elucidate potentially important genetic risk factors in such race/ethnic groups. Pharmacogenetic analysis of TCF7L2 variants and thiazide induced NOD produced strong evid ence that genetic variation in TCF7L2 particularly SNPs associated with T2D from GWAS, influence the effect of HCTZ on NOD risk Previous evidence suggests that TCF7L2 SNPs from T2D GWAS increase T2D risk 95 and also that thiazide diuretic administration increases T2D risk. 47 Our research supports the interaction of these two factors, suggesting that individuals with TCF7L2 T2D risk SNPs are at a high risk of T2D when treated with HCTZ and at a relatively low risk when not treated with HCTZ. To our knowledge, the research from Chapter 2 represents the only investigation of TCF7L2 SNPs in the pharmacogenetics of thiazide induced dysglycemia. Functional studies and replication of pharmacoge netic associations are
155 ne cessary to confirm this pharmacogenetic effect and define the potential role of TCF7L2 SNPs in predicting NOD during HCTZ treatment The research described in Chapter 3 utilize d a candidate gene approach to determine pharmacogeneti c risk factors for thiazide induced dysglycemia. Previous studies investigating the pharmacogenetics of thiazide induced dysglycemia do not utilize a tag SNP approach and investigate different phenotypes, making interpretation of SNP effects difficult. 112 114 In the present research, tag SNPs in the candidate genes KCNJ1 ADD1 ACE and AGTR1 were investigated i n the two hypertensive cohorts PEAR and INVEST GE NES w h i c h both collected clinical and genetic data and included HCTZ as a study treatment. Inclusion of both PEAR and INVEST allowed a comparison of pharmacogenetic effects on two phenotypes including change in FG and NOD The results of research desc ribed in Chapter 3 suggest that g enetic variation in selected candidate genes, especially KCNJ1 influence s the effect of HCTZ on both NOD risk and hyperglycemia Our results add to available evidence suggesting that g enetic variation in KCNJ1 influences the dysglycemic effect of HCTZ 114 Our observation of different KCNJ1 SNP effects between race/ethnic groups suggests differences in LD and highli ghts the need to identify functional SNPs. We also observed associations between change in FG and non synonymous ACE and ADD1 SNPs, adding to existing evidence of a pharmacogenetic role of ACE and ADD1 polymorphisms in thiazide induced dysglycemia. 112 Furthermore, our results suggest that pharmacogenetic predictors for change in FG during short term HCTZ treatment are distinct from pharmacogenetic predictors fo r NOD during long term HCTZ treatment.
156 Functional studies and replication of pharmacogenetic associations are needed to better define the potential role of KCNJ1 ACE and ADD1 SN Ps in pr edicting the dysglycemic effect of HCTZ. After further study and re plication, SNPs from these candidate genes might be used to provide genotype guided thiazide prescribing to avoid NOD in patients with risk genotypes. Whether these findings are specific to HCTZ or can be generalized to all thiazide diuretics is also uncl ear, but KCNJ1 ACE and ADD1 remain compelling candidate gene s in the pharmacogenetic study of thiazide induced dysglycemia. Pharmacogenetic associations remained significant for both change in FG and NOD after adjustment for the T2D GWAS SNP r s 7903146 f rom TCF7L2 suggest ing that pharmacogenetic predictors a re present regardless of baseline genetic risk. However, our results must be interpreted with caution as these TCF7L2 SNPs were a ssociated with change in FG only in PEAR blacks and not as sociated with NOD in any race/ethnic group in INVEST R esearch presented in Chapter 4 s ought to refine the role of thiazide treatment duration and concomitant antihypertensive treatment as pharmacotherapeutic risk factor s for thiazide induced dysglycemia We conducted a n original clinical study of previous PEAR and PEAR 2 patients treated continuously with HCTZ or chlorthalidone for more than six months which included an OGTT for detailed glycemic characterization of patients after long term thiazide trea tment. Few clinical trials have assessed OGTT response data after thiazide administration 116 200 201 Furthermore, l imited data are available compar ing short term (1 2 months) and long term (greater than 6 months) effects of thiazide treatment on FG changes in the same population. 116 The PEAR F ollow U p S tudy allowed evaluation of effects of long term thiazide treatment
157 on OGTT response data in an observational setting as well as comparison of s hort and long t erm effects of thiazide pharmacotherapy on FG In the PEAR Follow Up Study, we observed a strong association and correlation between duration of thiazide therapy and increasing FG during long term thiazide treatment. Baseline FG and baseline triglycerid es strongly predicted long term change in FG and triglycerides respectively. Concomitant antihypertensive treatment with beta blockers was associated with larger mean FG increases during follow up and increased two hour OGTT glucose at follow up. The ass ociation of duration of treatment with increasing FG may have clinical implications since antihypertensive therapy with thiazide diuretics is typically lifelong. PEAR Follow Up Study results suggest that change in FG during short term thiazide treatment is not associated with or correlated with change in FG during long term thiazide treatment. The lack of association suggests that acquiring a FG level approximately nine weeks after start of thiazide treatment would not be clinically informative or deter mine severity of long term changes in FG. Although non significant increases in FG and triglycerides were observed during long term thiazide treatment in the PEAR Follow Up Study, t he effect of thiazide diuretics on AMEs was likely more subtle than our sa mple size was able to detect. However, the str e ng th of associations from linear regressions, existence of previous studies identifying similar predictors of change in FG and triglycerides, 114 and physiological plausibility of associations lend credence to reported results Future research is necessary to confirm the predictive powe r of baseline FG and duration of thiazide treatment, which may provide insight into those patients who are most likely to develop AMEs
158 The evaluation of genetic, pharmacogenetic, and pharmacotherapeutic risk factors for thiazide induced dysglycemia presen ted in this dissertation may eventually contribute to better determination of T2D risk. Our observations might be used in future studies which utilize genotyping and clinical laboratory and drug data to predict increases in FG and NOD. Such studies may l ead to eventual individualization of antihypertensive treatment to avoid thiazide induced AMEs allow ing early reduct ion of modifiable risk factors and potentially, T2D prevention. After further research, a clinical risk assessment incorporating TCF7L2 KCNJ1 and ADD1 polymorphisms, duration of thiazide therapy and baseline FG values might encourage a proactive approach to identification of individuals at risk for dysglycemia Improved T2D risk assessments may contribute to a clinical paradigm shift aw ay from reactive FG monitoring during thiazide therapy and towards proactive prevention of T2D through personalized thiazide prescribing. Effective prevention of T2D has the potential to reduce the disease burden of T2D, including a negative eco nomic and public health impact.
159 APPENDIX A ADDITIONAL ANALYSIS OF PHARMACOGENETIC PREDICTORS OF THIAZIDE INCUDED DYSGLYCEMIA Table A 1. Candidate gene SNPs which deviated from Hardy Weinberg Equilibrium in at least one race/ethnic group in PEAR and INV EST PEAR INVEST Candidate Gene SNP Race/ethnic group HWE p value* Race/ethnic group HWE p value* KCNJ1 rs581472 Blacks 0.02 Blacks 1.00 (25 SNPs) Non blacks 0.55 Hispanics 0.65 Whites 0.46 rs583352 Blacks 0.01 Blacks 1.00 Non blacks 0.60 Hispanics 0.78 Whites 0.46 rs7933427 Blacks 0.33 Blacks 1.00 Non blacks 0.01 Hispanics 0.50 Whites 1.00 ADD1 rs1263347 Blacks 0.005 Blacks 0.33 (32 SNPs) Non blacks 5.70x10 7 Hispanics 0.34 Whites 0.71 rs16843458 Blacks 0.28 Blac ks 0.55 Non blacks 0.23 Hispanics 5.85x10 5 Whites rs16843589 Blacks 0.64 Blacks Non blacks 0.03 Hispanics 0.99 Whites 0.34 rs6833874 Blacks 0.99 Blacks 0.04 Non blacks Hispanics 0.99 Whites rs7689864 Blacks 0.99 Black s 0.04 Non blacks 0.22 Hispanics 0.99 Whites ACE rs1800764 Blacks 0.58 Blacks 0.79 (54 SNPs) Non blacks 0.99 Hispanics 0.03 Whites 0.27 rs4354 Blacks Blacks 0.26 Non blacks 0.99 Hispanics 0.03 Whites 0.50 rs12709436 Blacks 0.28 Blacks 2.89x10 5 Non blacks 0.99 Hispanics Whites AGTR1 rs1492102 Blacks 0.01 Blacks 0.19 (90 SNPs) Non blacks Hispanics 0.08 Whites 0.76
160 Table A 1. Continued PEAR INVEST Candidate Gene SNP Race/ethnic group HWE p value* R ace/ethnic group HWE p value* AGTR1 rs12721286 Blacks Blacks 0.62 (continued) Non blacks 0.001 Hispanics 0.57 Whites rs12721221 Blacks 0.29 Blacks 0.99 Non blacks Hispanics 0.03 Whites 0.99 rs275649 Blacks 0.65 Blacks 0.05 Non blacks 1.00 Hispanics 0.17 Whites 0.04 rs2640543 Blacks 0.45 Blacks 0.008 Non blacks 0.29 Hispanics 0.01 Whites 1.00 rs1492099 Blacks 0.32 Blacks 1.00 Non blacks 0.32 Hispanics 0.003 Whites 0.85 rs12721280 Blacks 0.99 Blacks 0. 006 Non blacks 1.00 Hispanics 0.99 Whites rs12721211 Blacks 1.00 Blacks 0.04 Non blacks 1.00 Hispanics 0.99 Whites rs9682137 Blacks 1.00 Blacks 0.04 Non blacks 0.50 Hispanics Whites rs12721232 Blacks 0.33 Blacks 0.02 Non blacks Hispanics 0.85 Whites 0.23 rs12695918 Blacks 0.34 Blacks 1.22x10 5 Non blacks 0.87 Hispanics 0.52 Whites 1.00 rs5183 Blacks 0.99 Blacks 3.73x10 4 Non blacks 1.00 Hispanics 0.64 Whites 0.25 rs389566 Blacks 1.00 Blac ks 0.49 Non blacks Hispanics 0.10 Whites 0.03 PEAR indicates Pharmacogenomics Evaluation of Antihypertensive Responses; INVEST, INternational VErapamil SR Trandolapril STudy; SNP, single nucleotide polymorphism; HWE, Hardy Weinberg Equilibrium. P values less than 0.05 are in bold. Assay for SNP considered failed due to HWE p <0.001 or deviations from HWE in multiple race/ethnic groups page..
161 Table A 2 SNP effects on cha nge in fasting glucose in PEAR for SNPs previously associated with thiazide induced dysglycemia SNP Race/ethnic group Allele (frequency) Beta (SE) HCTZ monotherapy p value Beta (SE) HCTZ add on p value Beta (SE) combined p value KCNJ1 rs59172778 Blacks G (0.00) Non blacks G (0.01) 3.15 (4.49) 0.48 1.91 (5.60) 0.73 2.10 (3.34) 0.58 ADD1 rs4961 Â§ Blacks A (0.06) 1.17 (2.66) 0.59 2.32 (4.62) 0.57 0.05 (2.32) 0.90 Non blacks A (0.21) 2.24 (1.28) 0.06 1.74 (1.06) 0.13 1.99 (0.81) 0.02 ACE rs4343 Blacks G (0.24) 1.13 (1.92) 0.45 1.25 (2.23) 0.72 1.07 (1.43) 0.41 Non blacks G (0.53) 0.52 (0.99) 0.63 0.03 (0.88) 0.89 0.17 (0.66) 0.81 AGTR1 rs5186 Blacks C (0.05) 2.22 (3.38) 0.51 0.77 (3.83 0.84 1.77 (2.54 ) 0 .46 Non blacks C (0.29) 0.64 (1.14) 0.58 0.22 (0.89) 0.80 0.48 (0.71) 0.50 SNP indicates single nucleotide polymorphism; HCTZ, hydrochlorothiazide; SE, standard error. *All parameter estimates (betas) and p values represent change in FG per copy of alle le adjusted for FG at start of HCTZ, age, gender, waist circumference, potassium supplementation during the study, drug arm, average home systolic and diastolic BP, HC TZ dose, duration of HCTZ treatment, and PCs one, two, and three. p values determined using change in log(FG) in linear regressions Â§SNP considered to complement previous association with thiazide induced dysglycemia
162 Table A 3. INVEST NOD odds ratios for SNPs previously associated with thiazide induced dysglycemia Gene SNP Race/ethnic group Allele (frequency) OR (95%CI) HCTZ treated OR (95%CI) OR (95%CI) Not HCTZ treated Interaction p value KCNJ1 rs59172778 White G (0.01) 1.54 (0.22 10.68) p=0.66 1.68 (0.23 12.26 ) p=0.61 0.77 (0.07 8.47) p=0.83 p=0.80 Hispanic G (0.003) 2.31 (0.44 12.3) p=0.32 2.51 (0.47 13.37 ) p=0.28 NE p=0.95 p=0.87 Black G (0.01) NE NE NE ADD1 rs4961 White A (0.20) 0.91 (0.59 1.39) p=0.66 0.90 (0.58 1.37) p=0.61 0.93 (0.48 1.82) p=0.83 p=0.88 Hispanic A (0.15) 0.99 (0.66 1.49) p=0.97 0.97 (0.64 1.47) p=0.90 1.11 ( 0 .59 2.0 9) p=0.74 p=0.98 Black A (0.07) 0.50 (0.14 1.56) p=0.21 NE p=0.25 NE p=0.50 p=0.19 ACE rs4343 White G (0.55) 1.31 (0.93 1.85) p=0.13 1.29 (0.91 1.82) p=0.15 0.76 (0.46 1.27) p=0.30 p=0.07 Hispanic G (0.45) 1.22 (0.90 1.65) p=0.20 1.23 (0.91 1.8 2) p=0.19 1.42 (0.93 2.19) p=0.11 p=0.71 Black G (0.22) 0.96 (0.48 1.90) p=0.91 0.99 (0.50 1.96) p=0.97 NE p=0.69 p=0.25 AGTR1 rs5186 White C (0.30) 0.96 (0.63 1.46) p= 0.85 0.96 (0.63 1.46) p=0.84 0.84 (0.47 1.51) p=0.56 p=0.25 Hispanic C (0.2 3) 0.87 (0.59 1.29) p=0.93 0.82 (0.54 1.24) p=0.34 0.89 (0.50 1.60) p=0.70 p=0.59 Black C (0.07) 2.69 (0.72 10.12) p=0.14 2.65 (0.71 9.97) p=0.15 NE p=0.92 p=0.94 95%CI indicates 95% confidence interval; HCTZ, hydrochlorothiazide; NE, odds ratio not est imable due to sample size; OR, odds ratio; SNP, single nucleotide polymorphism. *All point estimates and p values represent increased risk of NOD per copy of allele and are adjusted for age, gender, body mass index, average on treatment systolic blood pres sure, left ventricular hypertrophy, history of smoking and hypercholesterolemia, principal components one, two, and three, and treatment with trandolapril, atenolol, or potassium supplementation. Interaction p values for SNP*HCTZ treatment adjusted for prespecified covariates.
163 Table A 4 Stepwise multivariate models for including genetic and pharmacogenetic predictor SNPs in PEAR SNP Gene Parameter Estimate* (standard error) p value* R 2 2logL Blacks 0.255 2169 rs17137967 KCNJ1 7.54 (2.40) 0.00 2 rs4303 ACE 5.88 (1.92) 0.003 rs7903146 TCF7L2 3.05 (1.29) 0.02 Non Blacks 0.224 3040 rs4961 ADD1 1.97 (0.84) 0.02 rs7903146 TCF7L2 0.39 (0.73) 0.59 PEAR indicates Pharmacogenomic Evaluation of Antihypertensive Responses; SNP, single nucleotide polymorphism ; 2logL, negative two log likelihood. Generated using linear regression of change in fasting glucose during hydrochlorothiazide for all PEAR patients with adjustment for covariates used in previous models. Indicates significan t improvement of model over clinical variables alone. Table A 5 Stepwise multivariate models for including genetic and pharmacogenetic predictor SNPs in INVEST SNP Gene OR (95%CI)* p value* AUROC p value Blacks 0.853 0.25 rs7903146 TCF7L2 0.95 (0 .45 2.03) 0.89 Hispanics 0.804 0.23 rs12795437 KCNJ1 1.87 (1.15 3.05) 0.01 rs7903146 TCF7L2 1.03 (0.72 1.48) 0.88 Whites 0.849 0.18 rs12795437 KCNJ1 2.44 (1.18 5.03) 0.02 rs11196228 TCF7L2 0.35 (0.14 0.59) 0.02 rs7903146 TCF7L2 1.29 (0.81 2.05) 0.29 OR indicates odds ratio; 95%CI, 95% confidence interval; AUROC, area under the receiver operating characteristic curve. Generated using logistic regression of new onset diabetes for HCTZ treated INVEST patients by race/ethnicity with adjustment for covariates used in previous models. p value for improvement in area under the receiver operating characteristic curve with genetic and pharmacogenetic predictors over model with clinical covariates alone
164 Figure A 1 Haploview generat ed l inkage disequilibrium (LD) plot of KCNJ1 SNPs in INVEST w hites Regions of higher LD are shaded darker according to higher r 2 values. The number within each box indicates the r 2 value. Monomorphic SNPs are not included.
165 Figure A 2 Haploview gene rated l inkage disequilibrium (LD) plot of nominally significant ADD 1 SNPs in PEAR non blacks. Regions of higher LD are shaded darker according to higher r 2 values. The number within each box indicates the r 2 value. Monomorphic SNPs are not included.
166 Figure A 3 Haploview generated l inkage disequilibrium (LD) plot of ADD 1 SNPs in INVEST whites. Regions of higher LD are shaded darker according to higher r 2 values. The number within each box indicates the r 2 value. Monomorphic SNPs are not included.
167 Figure A 4 Area under the receiver operating characteristic curve for INVEST HCTZ treated white patients. Blue (ROC1) indicates sensitivity versus 1 specificity of model containing clinical covariates. Red (ROC2) indicates sensitivity versus 1 speci ficity of model containing clinical covariates as well as genetic ( TCF7L2 rs7903146 and rs11196228) and the KCNJ1 pharmac ogenetic predictor SNP rs12795437
168 APPENDIX B ADDITIONAL ANALYSIS OF PEAR FOLLOW UP STUDY DATA Table B 1. Variables assoc iated with LDL changes during long term thiazide treatment Independent variable Parameter p value Univariate associations (p<0.20) Baseline FG, mg/dL 0.58 0.15 Race (white) 20.38 0.11 Waist circumference, cm 0.58 0.19 Duration of thiaz ide treatment months 0.46 0.16 Statin treatment 25.30 0.07 Baseline potassium m Eq/L 21.64 0.09 Baseline LDL mg/dL 0.52 0.002 Baseline total cholesterol, mg/dL 0.33 0.03 Baseline SBP m mHg 0.64 0.10 Change in SBP during short term thiazide tre atment m mHg 0.92 0.12 Baseline DBP m mHg 1.45 0.03 Change in DBP during short term thiazide treatment m mHg 2.30 0.03 Stepwise Results (R 2 =0.31) Baseline LDL, mg/dL 0.47 0.003 Baseline DBP, mmHg 1.23 0.04 FG indicates fasting glucose; mg/dL milligrams per deciliter; cm, centimeters; mEq/L, milliequivalents per liter; LDL, low density lipoprotein ; SBP, systolic blood pressure; mmHg, millimeters of mercury; DBP, diastolic blood pressure
169 Table B 2. Variables associated with HDL changes durin g long term thiazide treatment Independent variable Parameter p value Univariate associations (p<0.20) Beta blocker treatment 7.11 0.01 Potassium supplementation 5.78 0.20 ACEI treatment 3.95 0.18 Family history of T2D 5.85 0.05 Statin treatment 5.86 0.08 Baseline HDL, mg/dL 0.19 0 .03 Change in HDL during short term thiazide therapy m g/dL 0.52 0.07 Baseline SBP, mmHg 0.16 0.09 Baseline DBP, mmHg 0.40 0.01 Change in DBP during short term thiazide therapy mmHg 0.61 0.02 Stepwise Results (R 2 =0.79) Beta blocker treatment 6.40 0.0001 Change in HDL during short term thiazide therapy m g/dL 0.58 0.0003 ACEI indicates angiotensin I converting enzyme inhibitor; T2D, type 2 diabetes; HDL, high density lipoprotein; mg/dL, milligrams per deciliter; SBP, systolic blood pressure ; mmHg, millimeters of mercury; DBP, diastolic blood pressure. History of T2D for a first degree relative
170 Table B 3. Variables associated with total cholesterol changes during long term thiazide treatment Independent variable Parameter p va lue Univariate associations (p<0.20) Baseline FG mg/dL 0.48 0.17 Race (white) 17.82 0.11 Waist circumference, cm 0.60 0.12 Duration of thiazide treatment months 0.49 0.09 Potassium supplementation 22.60 0.16 Statin treatment 18.79 0.13 Basel ine serum potassium, mEq/L 16.81 0.14 Baseline LDL, mg/dL 0.37 0.01 Baseline total cholesterol, mg/dL 0.34 0.01 Baseline SBP, mmHg 0.51 0.15 Change in SBP during short term thiazide therapy, mmHg 0.98 0.06 Baseline DBP mmHg 1.16 0.05 Change in D BP during short term thiazide therapy, mmHg 2.10 0.03 Stepwise Results (R 2 =0.19) Baseline total cholesterol, mg/dL 0.39 0.005 FG indicates fasting glucose; mg/dL, milligrams per deciliter; cm, centimeters; mEq/L, milliequivalents per liter; mmHg millimeters of mercury; SBP, systolic blood pressure; DBP, diastolic blood pressure.
171 Table B 4. Variables associated with serum potassium changes during long term thiazide treatment Independent variable Parameter p value Univariate associations (p<0.20) Ge nder (female) 0.29 0.15 Age, years 0.02 0.11 Race (white) 0.61 0.002 Abdominal obesity 0.34 0.08 Current smoker 0.35 0.18 Statin treatment 0.31 0.17 Baseline serum potassium, m Eq/L 0.78 <0.0001 Change in serum potassium during short term thiazide therapy, mEq/L 0.23 0.16 Baseline LDL, mg/dL 0.006 0.04 Baseline HDL, mg/dL 0.01 0.11 Baseline uric acid, mg/dL 0.09 0.17 Baseline total cholesterol, mg/dL 0.004 0.08 Baseline DBP, mmHg 0.02 0.06 Change in DBP during short term thiazide therapy, mmHg 0.04 0.04 Stepwise results (R 2 =0.51) Baseline serum potassium, mEq/L 0.77 <0.0001 Current smoker 0.57 0.008 Baseline HDL, mg/dL 0.01 0.03 mEq/L indicates milliequival ents per liter; LDL, low density lipoprotein; mg/dL, milligrams per deciliter; HDL, high density lipoprotein; DBP, diastolic blood pressure.
172 Figure B 1. Mean fasting glucose after short term and long term thiazide treatment including patients treated with anti diabetic medications Error bars indicate standard deviations.
173 A B Figure B 2. Mean fasting glucose after short term and long term thiazide treatment by add on antihypertensive treatment including patients treated with anti diabetic medicatio ns A) Mean fasting glucose in beta blocker and no beta blocker treated patients. B) Mean fasting glucose in ACE inhibitor and no ACE inhibitor treated patients. Error bars indicate standard deviations.
174 A B Figure B 3. Mean fasting glucose after sho rt term and long term thiazide treatment by thiazide and statin therapy. A) Mean fasting glucose in chlorthalidone versus HCTZ treated patients. B) Mean fasting glucose in statin and no statin treated patients. Error bars indicate standard deviations.
175 LIST OF REFERENCES 1. Roger VL, Go AS, Lloyd Jones DM, et al. Executive Summary: Heart Disease and Stroke Statistics -2012 Update: A Report From the American Heart Association. Circulation 2012;125:188 97. 2. Chobanian AV, Bakris GL, Bl ack HR, et al. The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report. JAMA 2003;289:2560 72. 3. Vasan RS, Beiser A, Seshadri S, et al. Residual lifetime risk for deve loping hypertension in middle aged women and men: The Framingham Heart Study. JAMA 2002;287:1003 10. 4. Lewington S, Clarke R, Qizilbash N, Peto R, Collins R. Age specific relevance of usual blood pressure to vascular mortality: a meta analysis of individu al data for one million adults in 61 prospective studies. Lancet 2002;360:1903 13. 5. Psaty BM, Furberg CD, Kuller LH, et al. Isolated systolic hypertension and subclinical cardiovascular disease in the elderly. Initial findings from the Cardiovascular Hea lth Study. JAMA 1992;268:1287 91. 6. Verdecchia P, Reboldi G, Angeli F, et al. Adverse prognostic significance of new diabetes in treated hypertensive subjects. Hypertension 2004;43:963 9. 7. Vaccaro O, Eberly LE, Neaton JD, Yang L, Riccardi G, Stamler J. Impact of diabetes and previous myocardial infarction on long term survival: 25 year mortality follow up of primary screenees of the Multiple Risk Factor Intervention Trial. Arch Intern Med 2004;164:1438 43. 8. Zhang P, Zhang X, Brown J, et al. Global heal thcare expenditure on diabetes for 2010 and 2030. Diabetes Res Clin Pract 2010;87:293 301. 9. Wild S, Roglic G, Green A, Sicree R, King H. Global prevalence of diabetes: estimates for the year 2000 and projections for 2030. Diabetes Care 2004;27:1047 53. 1 0. Whiting DR, Guariguata L, Weil C, Shaw J. IDF Diabetes Atlas: Global estimates of the prevalence of diabetes for 2011 and 2030. Diabetes Res Clin Pract 2011;94:311 21. 11. Shaw JE, Sicree RA, Zimmet PZ. Global estimates of the prevalence of diabetes for 2010 and 2030. Diabetes Res Clin Pract 2010;87:4 14. 12. Intensive blood glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). UK Prospective Diabetes Stu dy (UKPDS) Group. Lancet 1998;352:837 53.
176 13. Effects of ramipril on cardiovascular and microvascular outcomes in people with diabetes mellitus: results of the HOPE study and MICRO HOPE substudy. Heart Outcomes Prevention Evaluation Study Investigators. La ncet 2000;355:253 9. 14. Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) final report. Circulation 2002;106:3143 421. 15. Sourij H, Saely CH, Schmid F, et al. Post challenge hyperglycaemia is strongly associated with future macrovascular events and total mortality in angiographied coronary patients. Eur Heart J 2010;31:1583 90. 16. Liu J, Grundy SM, Wang W, et al. Ten year r isk of cardiovascular incidence related to diabetes, prediabetes, and the metabolic syndrome. Am Heart J 2007;153:552 8. 17. Mozaffarian D, Kamineni A, Prineas RJ, Siscovick DS. Metabolic syndrome and mortality in older adults: the Cardiovascular Health St udy. Arch Intern Med 2008;168:969 78. 18. Barr EL, Boyko EJ, Zimmet PZ, Wolfe R, Tonkin AM, Shaw JE. Continuous relationships between non diabetic hyperglycaemia and both cardiovascular disease and all cause mortality: the Australian Diabetes, Obesity, and Lifestyle (AusDiab) study. Diabetologia 2009;52:415 24. 19. Hyvarinen M, Qiao Q, Tuomilehto J, et al. Hyperglycemia and stroke mortality: comparison between fasting and 2 h glucose criteria. Diabetes Care 2009;32:348 54. 20. Rijkelijkhuizen JM, Nijpels G, Heine RJ, Bouter LM, Stehouwer CD, Dekker JM. High risk of cardiovascular mortality in individuals with impaired fasting glucose is explained by conversion to diabetes: the Hoorn study. Diabetes Care 2007;30:332 6. 21. Nichols GA, Hillier TA, Brown JB. Pr ogression from newly acquired impaired fasting glusose to type 2 diabetes. Diabetes Care 2007;30:228 33. 22. Haffner SM, Mykkanen L, Festa A, Burke JP, Stern MP. Insulin resistant prediabetic subjects have more atherogenic risk factors than insulin sensiti ve prediabetic subjects: implications for preventing coronary heart disease during the prediabetic state. Circulation 2000;101:975 80. 23. Tai ES, Goh SY, Lee JJ, et al. Lowering the criterion for impaired fasting glucose: impact on disease prevalence and associated risk of diabetes and ischemic heart disease. Diabetes Care 2004;27:1728 34.
177 24. Ferrannini E, Nannipieri M, Williams K, Gonzales C, Haffner SM, Stern MP. Mode of onset of type 2 diabetes from normal or impaired glucose tolerance. Diabetes 2004;5 3:160 5. 25. Gerstein HC, Santaguida P, Raina P, et al. Annual incidence and relative risk of diabetes in people with various categories of dysglycemia: a systematic overview and meta analysis of prospective studies. Diabetes Res Clin Pract 2007;78:305 12. 26. Knowler WC, Barrett Connor E, Fowler SE, et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 2002;346:393 403. 27. Tuomilehto J, Lindstrom J, Eriksson JG, et al. Prevention of type 2 diabetes mel litus by changes in lifestyle among subjects with impaired glucose tolerance. N Engl J Med 2001;344:1343 50. 28. Garber AJ, Handelsman Y, Einhorn D, et al. Diagnosis and management of prediabetes in the continuum of hyperglycemia: when do the risks of diab etes begin? A consensus statement from the American College of Endocrinology and the American Association of Clinical Endocrinologists. Endocr Pract 2008;14:933 46. 29. Ogden LG, He J, Lydick E, Whelton PK. Long term absolute benefit of lowering blood pres sure in hypertensive patients according to the JNC VI risk stratification. Hypertension 2000;35:539 43. 30. Neal B, MacMahon S, Chapman N. Effects of ACE inhibitors, calcium antagonists, and other blood pressure lowering drugs: results of prospectively des igned overviews of randomised trials. Blood Pressure Lowering Treatment Trialists' Collaboration. Lancet 2000;356:1955 64. 31. 2010 Top 200 generic drugs by total prescriptions. Drug Topics, 2011. (Accessed July 2, 2012, at http://drugtopics.modernmedicine .com/drugtopics/data/ articlestandard//drugtopics/252011/727243/article.pdf. ) 32. Major outcomes in high risk hypertensive patients randomized to angiotensin converting enzyme inhibitor or calcium channel blocker vs diuretic: The Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial (ALLHAT). JAMA 2002;288:2981 97. 33. Psaty BM, Smith NL, Siscovick DS, et al. Health outcomes associated with antihypertensive therapies used as first line agents. A systematic review and meta analysis. JAM A 1997;277:739 45.
1 78 34. Hansson L, Lindholm LH, Niskanen L, et al. Effect of angiotensin converting enzyme inhibition compared with conventional therapy on cardiovascular morbidity and mortality in hypertension: the Captopril Prevention Project (CAPPP) rand omised trial. Lancet 1999;353:611 6. 35. Brown MJ, Palmer CR, Castaigne A, et al. Morbidity and mortality in patients randomised to double blind treatment with a long acting calcium channel blocker or diuretic in the International Nifedipine GITS study: In tervention as a Goal in Hypertension Treatment (INSIGHT). Lancet 2000;356:366 72. 36. Hansson L, Hedner T, Lund Johansen P, et al. Randomised trial of effects of calcium antagonists compared with diuretics and beta blockers on cardiovascular morbidity and mortality in hypertension: the Nordic Diltiazem (NORDIL) study. Lancet 2000;356:359 65. 37. Hansson L, Lindholm LH, Ekbom T, et al. Randomised trial of old and new antihypertensive drugs in elderly patients: cardiovascular mortality and morbidity the Swedi sh Trial in Old Patients with Hypertension 2 study. Lancet 1999;354:1751 6. 38. Malacco E, Mancia G, Rappelli A, Menotti A, Zuccaro MS, Coppini A. Treatment of isolated systolic hypertension: the SHELL study results. Blood Press 2003;12:160 7. 39. Zanchett i A, Rosei EA, Dal Palu C, Leonetti G, Magnani B, Pessina A. The Verapamil in Hypertension and Atherosclerosis Study (VHAS): results of long term randomized treatment with either verapamil or chlorthalidone on carotid intima media thickness. J Hypertens 19 98;16:1667 76. 40. Thompson AM, Hu T, Eshelbrenner CL, Reynolds K, He J, Bazzano LA. Antihypertensive treatment and secondary prevention of cardiovascular disease events among persons without hypertension: a meta analysis. JAMA 2011;305:913 22. 41. Turnbul l F. Effects of different blood pressure lowering regimens on major cardiovascular events: results of prospectively designed overviews of randomised trials. Lancet 2003;362:1527 35. 42. Wing LM, Reid CM, Ryan P, et al. A comparison of outcomes with angiote nsin converting -enzyme inhibitors and diuretics for hypertension in the elderly. N Engl J Med 2003;348:583 92. 43. Jamerson K, Weber MA, Bakris GL, et al. Benazepril plus amlodipine or hydrochlorothiazide for hypertension in high risk patients. N Engl J M ed 2008;359:2417 28.
179 44. Mancia G, De Backer G, Dominiczak A, et al. 2007 Guidelines for the management of arterial hypertension: The Task Force for the Management of Arterial Hypertension of the European Society of Hypertension (ESH) and of the European S ociety of Cardiology (ESC). Eur Heart J 2007;28:1462 536. 45. Rosendorff C, Black HR, Cannon CP, et al. Treatment of hypertension in the prevention and management of ischemic heart disease: a scientific statement from the American Heart Association Council for High Blood Pressure Research and the Councils on Clinical Cardiology and Epidemiology and Prevention. Circulation 2007;115:2761 88. 46. Carter BL, Einhorn PT, Brands M, et al. Thiazide induced dysglycemia: call for research from a working group from t he national heart, lung, and blood institute. Hypertension 2008;52:30 6. 47. Elliott WJ, Meyer PM. Incident diabetes in clinical trials of antihypertensive drugs: a network meta analysis. Lancet 2007;369:201 7. 48. Aksnes TA, Kjeldsen SE, Rostrup M, Storse t O, Hua TA, Julius S. Predictors of new onset diabetes mellitus in hypertensive patients: the VALUE trial. J Hum Hypertens 2008;22:520 7. 49. Barzilay JI, Davis BR, Cutler JA, et al. Fasting glucose levels and incident diabetes mellitus in older nondiabet ic adults randomized to receive 3 different classes of antihypertensive treatment: a report from the Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial (ALLHAT). Arch Intern Med 2006;166:2191 201. 50. Cooper Dehoff R, Cohen JD, Bak ris GL, et al. Predictors of development of diabetes mellitus in patients with coronary artery disease taking antihypertensive medications (findings from the INternational VErapamil SR Trandolapril STudy [INVEST]). Am J Cardiol 2006;98:890 4. 51. Gupta AK, Dahlof B, Dobson J, Sever PS, Wedel H, Poulter NR. Determinants of new onset diabetes among 19,257 hypertensive patients randomized in the Anglo Scandinavian Cardiac Outcomes Trial -Blood Pressure Lowering Arm and the relative influence of antihypertensiv e medication. Diabetes Care 2008;31:982 8. 52. Pepine CJ, Cooper Dehoff RM. Cardiovascular therapies and risk for development of diabetes. J Am Coll Cardiol 2004;44:509 12. 53. Karnes JH, Cooper DeHoff RM. Antihypertensive medications: benefits of blood pr essure lowering and hazards of metabolic effects. Expert Rev Cardiovasc Ther 2009;7:689 702. 54. Bangalore S, Parkar S, Grossman E, Messerli FH. A meta analysis of 94,492 patients with hypertension treated with beta blockers to determine the risk of new on set diabetes mellitus. Am J Cardiol 2007;100:1254 62.
180 55. Abuissa H, Jones PG, Marso SP, O'Keefe JH, Jr. Angiotensin converting enzyme inhibitors or angiotensin receptor blockers for prevention of type 2 diabetes: a meta analysis of randomized clinical tri als. J Am Coll Cardiol 2005;46:821 6. 56. Gillespie EL, White CM, Kardas M, Lindberg M, Coleman CI. The impact of ACE inhibitors or angiotensin II type 1 receptor blockers on the development of new onset type 2 diabetes. Diabetes Care 2005;28:2261 6. 57. O pie LH, Schall R. Old antihypertensives and new diabetes. J Hypertens 2004;22:1453 8. 58. Scheen AJ. Renin angiotensin system inhibition prevents type 2 diabetes mellitus. Part 1. A meta analysis of randomised clinical trials. Diabetes Metab 2004;30:487 96 59. Moser M. Is new onset diabetes of clinical significance in treated hypertensive patients? -Con. J Clin Hypertens (Greenwich) 2006;8:126 32. 60. Alderman MH. New onset diabetes during antihypertensive therapy. Am J Hypertens 2008;21:493 9. 61. Alderma n MH, Cohen H, Madhavan S. Diabetes and cardiovascular events in hypertensive patients. Hypertension 1999;33:1130 4. 62. Samuelsson O, Pennert K, Andersson O, et al. Diabetes mellitus and raised serum triglyceride concentration in treated hypertension -are they of prognostic importance? Observational study. BMJ 1996;313:660 3. 63. Aksnes TA, Kjeldsen SE, Rostrup M, Omvik P, Hua TA, Julius S. Impact of new onset diabetes mellitus on cardiac outcomes in the Valsartan Antihypertensive Long term Use Evaluation (VALUE) trial population. Hypertension 2007;50:467 73. 64. Almgren T, Wilhelmsen L, Samuelsson O, Himmelmann A, Rosengren A, Andersson OK. Diabetes in treated hypertension is common and carries a high cardiovascular risk: results from a 28 year follow up. J Hypertens 2007;25:1311 7. 65. Dunder K, Lind L, Zethelius B, Berglund L, Lithell H. Increase in blood glucose concentration during antihypertensive treatment as a predictor of myocardial infarction: population based cohort study. BMJ 2003;326:681. 66. Eb erly LE, Cohen JD, Prineas R, Yang L. Impact of incident diabetes and incident nonfatal cardiovascular disease on 18 year mortality: the multiple risk factor intervention trial experience. Diabetes Care 2003;26:848 54. 67. Kostis JB, Wilson AC, Freudenberg er RS, Cosgrove NM, Pressel SL, Davis BR. Long term effect of diuretic based therapy on fatal outcomes in subjects with isolated systolic hypertension with and without diabetes. Am J Cardiol 2005;95:29 35.
181 68. Duarte JD, Cooper DeHoff RM. Mechanisms for bl ood pressure lowering and metabolic effects of thiazide and thiazide like diuretics. Expert Rev Cardiovasc Ther 2010;8:793 802. 69. Conway J, Lauwers P. Hemodynamic and hypotensive effects of long term therapy with chlorothiazide. Circulation 1960;21:21 7. 70. Kurtz TW, Pravenec M. Antidiabetic mechanisms of angiotensin converting enzyme inhibitors and angiotensin II receptor antagonists: beyond the renin angiotensin system. J Hypertens 2004;22:2253 61. 71. Lindholm LH, Ibsen H, Borch Johnsen K, et al. Risk of new onset diabetes in the Losartan Intervention For Endpoint reduction in hypertension study. J Hypertens 2002;20:1879 86. 72. Masuo K, Kawaguchi H, Mikami H, Ogihara T, Tuck ML. Serum uric acid and plasma norepinephrine concentrations predict subseque nt weight gain and blood pressure elevation. Hypertension 2003;42:474 80. 73. Nakanishi N, Okamoto M, Yoshida H, Matsuo Y, Suzuki K, Tatara K. Serum uric acid and risk for development of hypertension and impaired fasting glucose or Type II diabetes in Japa nese male office workers. Eur J Epidemiol 2003;18:523 30. 74. Dehghan A, van Hoek M, Sijbrands EJ, Hofman A, Witteman JC. High serum uric acid as a novel risk factor for type 2 diabetes. Diabetes Care 2008;31:361 2. 75. Shafi T, Appel LJ, Miller ER, 3rd, K lag MJ, Parekh RS. Changes in serum potassium mediate thiazide induced diabetes. Hypertension 2008;52:1022 9. 76. Zillich AJ, Garg J, Basu S, Bakris GL, Carter BL. Thiazide diuretics, potassium, and the development of diabetes: a quantitative review. Hyper tension 2006;48:219 24. 77. Braun M, Ramracheya R, Bengtsson M, et al. Voltage gated ion channels in human pancreatic beta cells: electrophysiological characterization and role in insulin secretion. Diabetes 2008;57:1618 28. 78. Chatterjee R, Yeh HC, Shafi T, et al. Serum and dietary potassium and risk of incident type 2 diabetes mellitus: The Atherosclerosis Risk in Communities (ARIC) study. Arch Intern Med 2010;170:1745 51. 79. Smith SM, Anderson SD, Wen S, et al. Lack of correlation between thiazide indu ced hyperglycemia and hypokalemia: subgroup analysis of results from the pharmacogenomic evaluation of antihypertensive responses (PEAR) study. Pharmacotherapy 2009;29:1157 65.
182 80. Herder C, Roden M. Genetics of type 2 diabetes: pathophysiologic and clinic al relevance. Eur J Clin Invest 2011;41:679 92. 81. Barroso I. Genetics of Type 2 diabetes. Diabet Med 2005;22:517 35. 82. Meigs JB, Cupples LA, Wilson PW. Parental transmission of type 2 diabetes: the Framingham Offspring Study. Diabetes 2000;49:2201 7. 8 3. Pierce M, Keen H, Bradley C. Risk of diabetes in offspring of parents with non insulin dependent diabetes. Diabet Med 1995;12:6 13. 84. Groop L, Forsblom C, Lehtovirta M, et al. Metabolic consequences of a family history of NIDDM (the Botnia study): evi dence for sex specific parental effects. Diabetes 1996;45:1585 93. 85. Poulsen P, Kyvik KO, Vaag A, Beck Nielsen H. Heritability of type II (non insulin dependent) diabetes mellitus and abnormal glucose tolerance -a population based twin study. Diabetologi a 1999;42:139 45. 86. Poulsen P, Grunnet LG, Pilgaard K, et al. Increased risk of type 2 diabetes in elderly twins. Diabetes 2009;58:1350 5. 87. Medici F, Hawa M, Ianari A, Pyke DA, Leslie RD. Concordance rate for type II diabetes mellitus in monozygotic t wins: actuarial analysis. Diabetologia 1999;42:146 50. 88. Lyssenko V, Almgren P, Anevski D, et al. Predictors of and longitudinal changes in insulin sensitivity and secretion preceding onset of type 2 diabetes. Diabetes 2005;54:166 74. 89. Kaprio J, Tuomi lehto J, Koskenvuo M, et al. Concordance for type 1 (insulin dependent) and type 2 (non insulin dependent) diabetes mellitus in a population based cohort of twins in Finland. Diabetologia 1992;35:1060 7. 90. Newman B, Selby JV, King MC, Slemenda C, Fabsitz R, Friedman GD. Concordance for type 2 (non insulin dependent) diabetes mellitus in male twins. Diabetologia 1987;30:763 8. 91. Grant SF, Thorleifsson G, Reynisdottir I, et al. Variant of transcription factor 7 like 2 (TCF7L2) gene confers risk of type 2 diabetes. Nat Genet 2006;38:320 3. 92. Genome wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature 2007;447:661 78. 93. Ahlqvist E, Ahluwalia TS, Groop L. Genetics of type 2 diabetes. Clin Chem 2011;57:241 54.
183 9 4. Sladek R, Rocheleau G, Rung J, et al. A genome wide association study identifies novel risk loci for type 2 diabetes. Nature 2007;445:881 5. 95. Voight BF, Scott LJ, Steinthorsdottir V, et al. Twelve type 2 diabetes susceptibility loci identified throug h large scale association analysis. Nat Genet 2010;42:579 89. 96. van Hoek M, Dehghan A, Witteman JC, et al. Predicting type 2 diabetes based on polymorphisms from genome wide association studies: a population based study. Diabetes 2008;57:3122 8. 97. Talm ud PJ, Hingorani AD, Cooper JA, et al. Utility of genetic and non genetic risk factors in prediction of type 2 diabetes: Whitehall II prospective cohort study. BMJ 2010;340:b4838. 98. Tong Y, Lin Y, Zhang Y, Yang J, Liu H, Zhang B. Association between TCF7 L2 gene polymorphisms and susceptibility to type 2 diabetes mellitus: a large Human Genome Epidemiology (HuGE) review and meta analysis. BMC Med Genet 2009;10:15. 99. Florez JC, Jablonski KA, Bayley N, et al. TCF7L2 polymorphisms and progression to diabete s in the Diabetes Prevention Program. N Engl J Med 2006;355:241 50. 100. Helgason A, Palsson S, Thorleifsson G, et al. Refining the impact of TCF7L2 gene variants on type 2 diabetes and adaptive evolution. Nat Genet 2007;39:218 25. 101. Cauchi S, El Achhab Y, Choquet H, et al. TCF7L2 is reproducibly associated with type 2 diabetes in various ethnic groups: a global meta analysis. J Mol Med 2007;85:777 82. 102. Rosenberg NA, Huang L, Jewett EM, Szpiech ZA, Jankovic I, Boehnke M. Genome wide association studi es in diverse populations. Nat Rev Genet 2010;11:356 66. 103. Humphries SE, Gable D, Cooper JA, et al. Common variants in the TCF7L2 gene and predisposition to type 2 diabetes in UK European Whites, Indian Asians and Afro Caribbean men and women. J Mol Med 2006;84:1005 14. 104. Elbein SC, Chu WS, Das SK, et al. Transcription factor 7 like 2 polymorphisms and type 2 diabetes, glucose homeostasis traits and gene expression in US participants of European and African descent. Diabetologia 2007;50:1621 30. 105. Yan Y, North KE, Heiss G, et al. Transcription factor 7 like 2 (TCF7L2) polymorphism and context specific risk of impaired fasting glucose in African American and Caucasian adults: the atherosclerosis risk in communities (ARIC) study. Diabetes Metab Res Re v 2010;26:371 7. 106. Parra EJ, Cameron E, Simmonds L, et al. Association of TCF7L2 polymorphisms with type 2 diabetes in Mexico City. Clin Genet 2007;71:359 66.
184 107. Lehman DM, Hunt KJ, Leach RJ, et al. Haplotypes of transcription factor 7 like 2 (TCF7L2) gene and its upstream region are associated with type 2 diabetes and age of onset in Mexican Americans. Diabetes 2007;56:389 93. 108. Sale MM, Smith SG, Mychaleckyj JC, et al. Variants of the transcription factor 7 like 2 (TCF7L2) gene are associated with type 2 diabetes in an African American population enriched for nephropathy. Diabetes 2007;56:2638 42. 109. Palmer ND, Hester JM, An SS, et al. Resequencing and analysis of variation in the TCF7L2 gene in African Americans suggests that SNP rs7903146 is th e causal diabetes susceptibility variant. Diabetes 2011;60:662 8. 110. Waters KM, Stram DO, Hassanein MT, et al. Consistent association of type 2 diabetes risk variants found in europeans in diverse racial and ethnic groups. PLoS Genet 2010;6. 111. Davey S mith G, Bracha Y, Svendsen KH, Neaton JD, Haffner SM, Kuller LH. Incidence of type 2 diabetes in the randomized multiple risk factor intervention trial. Ann Intern Med 2005;142:313 22. 112. Bozkurt O, de Boer A, Grobbee DE, et al. Variation in Renin Angiot ensin system and salt sensitivity genes and the risk of diabetes mellitus associated with the use of thiazide diuretics. Am J Hypertens 2009;22:545 51. 113. Irvin MR, Lynch AI, Kabagambe EK, et al. Pharmacogenetic association of hypertension candidate gene s with fasting glucose in the GenHAT Study. J Hypertens 2010;28:2076 83. 114. Maitland van der Zee AH, Turner ST, Schwartz GL, Chapman AB, Klungel OH, Boerwinkle E. Demographic, environmental, and genetic predictors of metabolic side effects of hydrochloro thiazide treatment in hypertensive subjects. Am J Hypertens 2005;18:1077 83. 115. Altshuler DM, Gibbs RA, Peltonen L, et al. Integrating common and rare genetic variation in diverse human populations. Nature 2010;467:52 8. 116. Propranolol or hydrochloroth iazide alone for the initial treatment of hypertension. IV. Effect on plasma glucose and glucose tolerance. Veterans Administration Cooperative Study Group on Antihypertensive Agents. Hypertension 1985;7:1008 16. 117. Johnson JA, Boerwinkle E, Zineh I, et al. Pharmacogenomics of antihypertensive drugs: rationale and design of the Pharmacogenomic Evaluation of Antihypertensive Responses (PEAR) study. Am Heart J 2009;157:442 9. 118. Cooper DeHoff RM, Wen S, Beitelshees AL, et al. Impact of abdominal obesity o n incidence of adverse metabolic effects associated with antihypertensive medications. Hypertension 2010;55:61 8.
185 119. Diagnosis and classification of diabetes mellitus. Diabetes Care 2012;35 Suppl 1:S64 71. 120. Hu FB, Stampfer MJ, Haffner SM, Solomon CG, Willett WC, Manson JE. Elevated risk of cardiovascular disease prior to clinical diagnosis of type 2 diabetes. Diabetes Care 2002;25:1129 34. 121. Meigs JB, Nathan DM, D'Agostino RB, Sr., Wilson PW. Fasting and postchallenge glyc emia and cardiovascular disease risk: the Framingham Offspring Study. Diabetes Care 2002;25:1845 50. 122. Tominaga M, Eguchi H, Manaka H, Igarashi K, Kato T, Sekikawa A. Impaired glucose tolerance is a risk factor for cardiovascular disease, but not impair ed fasting glucose. The Funagata Diabetes Study. Diabetes Care 1999;22:920 4. 123. Glucose tolerance and mortality: comparison of WHO and American Diabetes Association diagnostic criteria. The DECODE study group. European Diabetes Epidemiology Group. Diabe tes Epidemiology: Collaborative analysis Of Diagnostic criteria in Europe. Lancet 1999;354:617 21. 124. Glucose tolerance and cardiovascular mortality: comparison of fasting and 2 hour diagnostic criteria. Arch Intern Med 2001;161:397 405. 125. Qiao Q, Pyo rala K, Pyorala M, et al. Two hour glucose is a better risk predictor for incident coronary heart disease and cardiovascular mortality than fasting glucose. Eur Heart J 2002;23:1267 75. 126. Will new diagnostic criteria for diabetes mellitus change phenoty pe of patients with diabetes? Reanalysis of European epidemiological data. DECODE Study Group on behalf of the European Diabetes Epidemiology Study Group. BMJ 1998;317:371 5. 127. Selvin E, Steffes MW, Zhu H, et al. Glycated hemoglobin, diabetes, and cardi ovascular risk in nondiabetic adults. N Engl J Med 2010;362:800 11. 128. Abdul Ghani MA, Tripathy D, DeFronzo RA. Contributions of beta cell dysfunction and insulin resistance to the pathogenesis of impaired glucose tolerance and impaired fasting glucose. Diabetes Care 2006;29:1130 9. 129. Matsuda M, DeFronzo RA. Insulin sensitivity indices obtained from oral glucose tolerance testing: comparison with the euglycemic insulin clamp. Diabetes Care 1999;22:1462 70. 130. Matthews DR, Hosker JP, Rudenski AS, Nayl or BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and beta cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985;28:412 9.
186 131. Sherwin RS, Anderson RM, Buse JB, et al. Prevention or de lay of type 2 diabetes. Diabetes Care 2004;27 Suppl 1:S47 54. 132. Weber MB, Narayan KM. Diabetes prevention should be a public health priority. Lancet 2008;371:473 4. 133. Wilson PW, Meigs JB, Sullivan L, Fox CS, Nathan DM, D'Agostino RB, Sr. Prediction o f incident diabetes mellitus in middle aged adults: the Framingham Offspring Study. Arch Intern Med 2007;167:1068 74. 134. Palmer ND, McDonough CW, Hicks PJ, et al. A genome wide association search for type 2 diabetes genes in African Americans. PLoS One 2 012;7:e29202. 135. Gaulton KJ, Nammo T, Pasquali L, et al. A map of open chromatin in human pancreatic islets. Nat Genet 2010;42:255 9. 136. de Miguel Yanes JM, Shrader P, Pencina MJ, et al. Genetic risk reclassification for type 2 diabetes by age below or above 50 years using 40 type 2 diabetes risk single nucleotide polymorphisms. Diabetes Care 2011;34:121 5. 137. Pepine CJ, Handberg Thurmond E, Marks RG, et al. Rationale and design of the International Verapamil SR/Trandolapril Study (INVEST): an Interne t based randomized trial in coronary artery disease patients with hypertension. J Am Coll Cardiol 1998;32:1228 37. 138. Pepine CJ, Handberg EM, Cooper DeHoff RM, et al. A calcium antagonist vs a non calcium antagonist hypertension treatment strategy for pa tients with coronary artery disease. The International Verapamil Trandolapril Study (INVEST): a randomized controlled trial. JAMA 2003;290:2805 16. 139. The sixth report of the Joint National Committee on prevention, detection, evaluation, and treatment of high blood pressure. Arch Intern Med 1997;157:2413 46. 140. Andrisin TE, Humma LM, Johnson JA. Collection of genomic DNA by the noninvasive mouthwash method for use in pharmacogenetic studies. Pharmacotherapy 2002;22:954 60. 141. Dreszer TR, Karolchik D, Zweig AS, et al. The UCSC Genome Browser database: extensions and updates 2011. Nucleic Acids Res 2012;40:D918 23. 142. Howie BN, Carlson CS, Rieder MJ, Nickerson DA. Efficient selection of tagging single nucleotide polymorphisms in multiple populations. H um Genet 2006;120:58 68. 143. Conde L, Vaquerizas JM, Dopazo H, et al. PupaSuite: finding functional single nucleotide polymorphisms for large scale genotyping purposes. Nucleic Acids Res 2006;34:W621 5.
187 144. Chelala C, Khan A, Lemoine NR. SNPnexus: a web database for functional annotation of newly discovered and public domain single nucleotide polymorphisms. Bioinformatics 2009;25:655 61. 145. Kumar P, Henikoff S, Ng PC. Predicting the effects of coding non synonymous variants on protein function using the SIFT algorithm. Nat Protoc 2009;4:1073 81. 146. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal components analysis corrects for stratification in genome wide association studies. Nat Genet 2006;38:904 9. 147. Barrett JC, F ry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005;21:263 5. 148. Gabriel SB, Schaffner SF, Nguyen H, et al. The structure of haplotype blocks in the human genome. Science 2002;296:2225 9. 149. Keat ing BJ, Tischfield S, Murray SS, et al. Concept, design and implementation of a cardiovascular gene centric 50 k SNP array for large scale genomic association studies. PLoS One 2008;3:e3583. 150. Purcell S, Neale B, Todd Brown K, et al. PLINK: a tool set f or whole genome association and population based linkage analyses. Am J Hum Genet 2007;81:559 75. 151. Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing. Stat Med 1990;9:811 8. 152. The International HapMap Project. Nature 2003;426:789 96. 153. Zeggini E, Weedon MN, Lindgren CM, et al. Replication of genome wide association signals in UK samples reveals risk loci for type 2 diabetes. Science 2007;316:1336 41. 154. Scott LJ, Bonnycastle LL, Willer CJ, et al. Association of t ranscription factor 7 like 2 (TCF7L2) variants with type 2 diabetes in a Finnish sample. Diabetes 2006;55:2649 53. 155. Ruchat SM, Elks CE, Loos RJ, et al. Evidence of interaction between type 2 diabetes susceptibility genes and dietary fat intake for adip osity and glucose homeostasis related phenotypes. J Nutrigenet Nutrigenomics 2009;2:225 34. 156. Saxena R, Hivert MF, Langenberg C, et al. Genetic variation in GIPR influences the glucose and insulin responses to an oral glucose challenge. Nat Genet 2010;4 2:142 8.
188 157. Sousa AG, Marquezine GF, Lemos PA, et al. TCF7L2 polymorphism rs7903146 is associated with coronary artery disease severity and mortality. PLoS One 2009;4:e7697. 158. Cowie CC, Rust KF, Ford ES, et al. Full accounting of diabetes and pre diab etes in the U.S. population in 1988 1994 and 2005 2006. Diabetes Care 2009;32:287 94. 159. Muendlein A, Saely CH, Geller Rhomberg S, et al. Single nucleotide polymorphisms of TCF7L2 are linked to diabetic coronary atherosclerosis. PLoS One 2011;6:e17978. 1 60. Macfarlane DP, Paterson KR, Fisher M. Cardiovascular drugs as antidiabetic agents: evidence for the prevention of type 2 diabetes. Diabetes Obes Metab 2008;10:533 44. 161. Welling PA, Ho K. A comprehensive guide to the ROMK potassium channel: form and function in health and disease. Am J Physiol Renal Physiol 2009;297:F849 63. 162. Lanzani C, Citterio L, Jankaricova M, et al. Role of the adducin family genes in human essential hypertension. J Hypertens 2005;23:543 9. 163. Matsuoka Y, Li X, Bennett V. Ad ducin: structure, function and regulation. Cell Mol Life Sci 2000;57:884 95. 164. Barlassina C, Schork NJ, Manunta P, et al. Synergistic effect of alpha adducin and ACE genes causes blood pressure changes with body sodium and volume expansion. Kidney Int 2 000;57:1083 90. 165. Psaty BM, Smith NL, Heckbert SR, et al. Diuretic therapy, the alpha adducin gene variant, and the risk of myocardial infarction or stroke in persons with treated hypertension. JAMA 2002;287:1680 9. 166. Davis BR, Arnett DK, Boerwinkle E, et al. Antihypertensive therapy, the alpha adducin polymorphism, and cardiovascular disease in high risk hypertensive persons: the Genetics of Hypertension Associated Treatment Study. Pharmacogenomics J 2007;7:112 22. 167. Gerhard T, Gong Y, Beitelshees AL, et al. Alpha adducin polymorphism associated with increased risk of adverse cardiovascular outcomes: results from GENEtic Substudy of the INternational VErapamil SR trandolapril STudy (INVEST GENES). Am Heart J 2008;156:397 404. 168. Cooper DeHoff RM, Handberg EM, Mancia G, et al. INVEST revisited: review of findings from the International Verapamil SR Trandolapril Study. Expert Rev Cardiovasc Ther 2009;7:1329 40.
189 169. De la Vega FM, Lazaruk KD, Rhodes MD, Wenz MH. Assessment of two flexible and compat ible SNP genotyping platforms: TaqMan SNP Genotyping Assays and the SNPlex Genotyping System. Mutat Res 2005;573:111 35. 170. Chapman AB, Schwartz GL, Boerwinkle E, Turner ST. Predictors of antihypertensive response to a standard dose of hydrochlorothiazid e for essential hypertension. Kidney Int 2002;61:1047 55. 171. Glenn KL, Du ZQ, Eisenmann JC, Rothschild MF. An alternative method for genotyping of the ACE I/D polymorphism. Mol Biol Rep 2009;36:1305 10. 172. Dupuis J, Langenberg C, Prokopenko I, et al. N ew genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nat Genet 2010;42:105 16. 173. Tobin MD, Tomaszewski M, Braund PS, et al. Common variants in genes underlying monogenic hypertension and hypotension and blo od pressure in the general population. Hypertension 2008;51:1658 64. 174. Ji W, Foo JN, O'Roak BJ, et al. Rare independent mutations in renal salt handling genes contribute to blood pressure variation. Nat Genet 2008;40:592 9. 175. Bailey SD, Xie C, Do R, et al. Variation at the NFATC2 locus increases the risk of thiazolidinedione induced edema in the Diabetes REduction Assessment with ramipril and rosiglitazone Medication (DREAM) study. Diabetes Care 2010;33:2250 3. 176. Simon DB, Karet FE, Rodriguez Soria no J, et al. Genetic heterogeneity of Bartter's syndrome revealed by mutations in the K+ channel, ROMK. Nat Genet 1996;14:152 6. 177. Gupta AK, Poulter NR, Dobson J, et al. Ethnic differences in blood pressure response to first and second line antihyperten sive therapies in patients randomized in the ASCOT Trial. Am J Hypertens 2010;23:1023 30. 178. Johnson JA, Gong Y, Bailey KR, et al. Hydrochlorothiazide and atenolol combination antihypertensive therapy: effects of drug initiation order. Clin Pharmacol The r 2009;86:533 9. 179. Villard E, Tiret L, Visvikis S, Rakotovao R, Cambien F, Soubrier F. Identification of new polymorphisms of the angiotensin I converting enzyme (ACE) gene, and study of their relationship to plasma ACE levels by two QTL segregation lin kage analysis. Am J Hum Genet 1996;58:1268 78. 180. Lloyd Jones DM, Evans JC, Larson MG, O'Donnell CJ, Wilson PW, Levy D. Cross classification of JNC VI blood pressure stages and risk groups in the Framingham Heart Study. Arch Intern Med 1999;159:2206 12.
190 181. MacMahon SW, Macdonald GJ, Blacket RB. Plasma lipoprotein levels in treated and untreated hypertensive men and women. The National Heart Foundation of Australia Risk Factor Prevalence Study. Arteriosclerosis 1985;5:391 6. 182. Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III). JAMA 2001;285:2486 97. 183. McMahan CA, Gidding SS, Fayad ZA, et al. Ri sk scores predict atherosclerotic lesions in young people. Arch Intern Med 2005;165:883 90. 184. Rajala U, Qiao Q, Laakso M, Keinanen Kiukaanniemi S. Antihypertensive drugs as predictors of type 2 diabetes among subjects with impaired glucose tolerance. Di abetes Res Clin Pract 2000;50:231 9. 185. Gami AS, Witt BJ, Howard DE, et al. Metabolic syndrome and risk of incident cardiovascular events and death: a systematic review and meta analysis of longitudinal studies. J Am Coll Cardiol 2007;49:403 14. 186. Nan deesha H, Pavithran P, Madanmohan T. Effect of antihypertensive therapy on serum lipids in newly diagnosed essential hypertensive men. Angiology 2009;60:217 20. 187. Lee JM, Kim HC, Cho HM, Oh SM, Choi DP, Suh I. Association between serum uric Acid level a nd metabolic syndrome. J Prev Med Public Health 2012;45:181 7. 188. Chuang HH, Li WC, Sheu BF, et al. Correlation between body composition and risk factors for cardiovascular disease and metabolic syndrome. Biofactors 2012. 189. Niskanen LK, Laaksonen DE, Nyyssonen K, et al. Uric acid level as a risk factor for cardiovascular and all cause mortality in middle aged men: a prospective cohort study. Arch Intern Med 2004;164:1546 51. 190. Lind L, Pollare T, Berne C, Lithell H. Long term metabolic effects of ant ihypertensive drugs. Am Heart J 1994;128:1177 83. 191. Yeboah J, Bertoni AG, Herrington DM, Post WS, Burke GL. Impaired fasting glucose and the risk of incident diabetes mellitus and cardiovascular events in an adult population: MESA (Multi Ethnic Study of Atherosclerosis). J Am Coll Cardiol 2011;58:140 6. 192. DeFronzo RA, Abdul Ghani M. Assessment and treatment of cardiovascular risk in prediabetes: impaired glucose tolerance and impaired fasting glucose. Am J Cardiol 2011;108:3B 24B. 193. Tamita K, Katay ama M, Takagi T, et al. Newly diagnosed glucose intolerance and prognosis after acute myocardial infarction: comparison of post challenge versus fasting glucose concentrations. Heart 2012;98:848 54.
191 194. Abdul Ghani MA, Abdul Ghani T, Ali N, Defronzo RA. O ne hour plasma glucose concentration and the metabolic syndrome identify subjects at high risk for future type 2 diabetes. Diabetes Care 2008;31:1650 5. 195. Strandberg TE, Pienimaki T, Strandberg AY, et al. One hour glucose, mortality, and risk of diabete s: a 44 year prospective study in men. Arch Intern Med 2011;171:941 3. 196. Sainaghi PP, Castello L, Limoncini AM, Bergamasco L, Bartoli E, Schianca GP. Poor specificity of fasting plasma glucose cut off values in ruling out glucose intolerance: the comple mentary usefulness of OGTT. Exp Clin Endocrinol Diabetes 2007;115:112 7. 197. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap) -a metadata driven methodology and workflow process for providing transla tional research informatics support. J Biomed Inform 2009;42:377 81. 198. Peterzan MA, Hardy R, Chaturvedi N, Hughes AD. Meta analysis of dose response relationships for hydrochlorothiazide, chlorthalidone, and bendroflumethiazide on blood pressure, serum potassium, and urate. Hypertension 2012;59:1104 9. 199. Roush GC, Holford TR, Guddati AK. Chlorthalidone compared with hydrochlorothiazide in reducing cardiovascular events: systematic review and network meta analyses. Hypertension 2012;59:1110 7. 200. Bak ris G, Molitch M, Zhou Q, et al. Reversal of diuretic associated impaired glucose tolerance and new onset diabetes: results of the STAR LET study. J Cardiometab Syndr 2008;3:18 25. 201. Stears AJ, Woods SH, Watts MM, et al. A double blind, placebo controll ed, crossover trial comparing the effects of amiloride and hydrochlorothiazide on glucose tolerance in patients with essential hypertension. Hypertension 2012;59:934 42.
192 BIOGRAPHICAL SKETCH Jason Hansen Karnes was born in Gainesville, Florida and grew u p in Richmond, International Studies in 2000. He then attended the College of William and Mary in Williamsburg, Virginia and graduated with a Bachelor of Arts degree in ancient Greek language and literature in 2004. Jason then returned to Gainesville to study clinical pharmacotherapy at the University of Florida and graduated with a Doctor of Pharmacy degree cum laude in 2008 Jason has authored multiple peer reviewed manuscrip ts and presented research at multiple national meetings. After defending his dissertation, Jason plans to move to Nashville, Tennessee in 2012 to complete a postdoctoral fellowship with the Vanderbilt University School of Medicine at the Division of Clini cal Pharmacology. Jason plans to pursue an academic career in clinical and translational pharmacogenomics.