<%BANNER%>

Risk of Tendon Rupture after Quinolone Use in a U.S. Military Population

Permanent Link: http://ufdc.ufl.edu/UFE0021357/00001

Material Information

Title: Risk of Tendon Rupture after Quinolone Use in a U.S. Military Population
Physical Description: 1 online resource (134 p.)
Language: english
Creator: Garman, Patrick M, II
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: analysis, cephalosporin, fluoroquinolone, function, hazard, military, quinolone, rupture, survival, tendon, tendonopathy
Pharmacy Health Care Administration -- Dissertations, Academic -- UF
Genre: Pharmaceutical Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Non-battle related injuries are the greatest cause of morbidity and mortality in the US military. Within the category of non-battle injuries, musculoskeletal injuries are the leading cause of disability. Tendon ruptures in particular, require long rehabilitation programs and can result in permanent disability. The consequence of tendon rupture is degradation in unit readiness and operational effectiveness. Case reports and case studies have linked tendon ruptures with quinolone antibiotic usage in the civilian population. This retrospective cohort study utilized a Department of Defense administrative database to estimate the risk of tendon rupture in active duty personnel associated with quinolone use compared with the use of cephalosporin antibiotics. It identifies an induction period from quinolone exposure to tendon rupture while also identifying risk factors that are associated with an increased risk of tendon ruptures. Data from military personnel June 2005 through May 2006 were collected. Internal validity was confirmed using macro-level assessment techniques. Survival analysis was used to estimate the hazard function and the Cox proportional hazards model was employed to produce hazard ratios for the main treatment and other independent risk factors. There was a significant increase in the risk of tendon ruptures over a 60-day period between active duty personnel who used quinolones compared to those who used cephalosporins (HR=1.65, 95%CI 1.33-2.04). The risk was highest during the 26 to 35 day window which marked the average induction period. Risk factors were days of supply (HR=1.02, 95%CI 1.018-1.022); military occupation (HR=1.53, 95%CI 1.24-1.89); age (HR=1.02, 95%CI 1.01-1.03); provider specialty (HR=2.40, 95%CI 1.92-2.99); Marine Corps service (HR=1.65, 95%CI 1.20-2.28); GI/GU antibiotic indication (HR=1.25, 95%CI 0.99-1.57). This is the first study to identify an increased risk of tendon rupture from quinolone use in a demographically diverse population. The risk is elevated upon the first day of therapy and increases incrementally until it reaches its maximum between days 26-35. The risk of tendon rupture is augmented by advancing age, length of treatment, and working in a physically demanding occupation. These findings are important as they provide evidence that a significant portion of this debilitating injury is avoidable by identifying specific risk factors and incorporating them into the patient care plan.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Patrick M Garman.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Ried, Lyle D.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2008-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021357:00001

Permanent Link: http://ufdc.ufl.edu/UFE0021357/00001

Material Information

Title: Risk of Tendon Rupture after Quinolone Use in a U.S. Military Population
Physical Description: 1 online resource (134 p.)
Language: english
Creator: Garman, Patrick M, II
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: analysis, cephalosporin, fluoroquinolone, function, hazard, military, quinolone, rupture, survival, tendon, tendonopathy
Pharmacy Health Care Administration -- Dissertations, Academic -- UF
Genre: Pharmaceutical Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Non-battle related injuries are the greatest cause of morbidity and mortality in the US military. Within the category of non-battle injuries, musculoskeletal injuries are the leading cause of disability. Tendon ruptures in particular, require long rehabilitation programs and can result in permanent disability. The consequence of tendon rupture is degradation in unit readiness and operational effectiveness. Case reports and case studies have linked tendon ruptures with quinolone antibiotic usage in the civilian population. This retrospective cohort study utilized a Department of Defense administrative database to estimate the risk of tendon rupture in active duty personnel associated with quinolone use compared with the use of cephalosporin antibiotics. It identifies an induction period from quinolone exposure to tendon rupture while also identifying risk factors that are associated with an increased risk of tendon ruptures. Data from military personnel June 2005 through May 2006 were collected. Internal validity was confirmed using macro-level assessment techniques. Survival analysis was used to estimate the hazard function and the Cox proportional hazards model was employed to produce hazard ratios for the main treatment and other independent risk factors. There was a significant increase in the risk of tendon ruptures over a 60-day period between active duty personnel who used quinolones compared to those who used cephalosporins (HR=1.65, 95%CI 1.33-2.04). The risk was highest during the 26 to 35 day window which marked the average induction period. Risk factors were days of supply (HR=1.02, 95%CI 1.018-1.022); military occupation (HR=1.53, 95%CI 1.24-1.89); age (HR=1.02, 95%CI 1.01-1.03); provider specialty (HR=2.40, 95%CI 1.92-2.99); Marine Corps service (HR=1.65, 95%CI 1.20-2.28); GI/GU antibiotic indication (HR=1.25, 95%CI 0.99-1.57). This is the first study to identify an increased risk of tendon rupture from quinolone use in a demographically diverse population. The risk is elevated upon the first day of therapy and increases incrementally until it reaches its maximum between days 26-35. The risk of tendon rupture is augmented by advancing age, length of treatment, and working in a physically demanding occupation. These findings are important as they provide evidence that a significant portion of this debilitating injury is avoidable by identifying specific risk factors and incorporating them into the patient care plan.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Patrick M Garman.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Ried, Lyle D.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2008-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021357:00001


This item has the following downloads:


Full Text

PAGE 1

1 RISK OF TENDON RUPTURE AFTER QUINOLONE USE IN A U.S. MILITARY POPULATION By PATRICK M. GARMAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007

PAGE 2

2 2007 Patrick M. Garman

PAGE 3

3 To my wife Kim

PAGE 4

4 ACKNOWLEDGMENTS I thank my chair and supervisory committee. My chair L. Douglas Ried is a model of patience and support. Abraham Hartzema is a consummate professional while Almut Winterstein embodies commitment to excellence. I extend my appreciation to Nabih Asal for his reassuring council during this pr ocess and to Mike Daniels for his insight into the analytic process. I express gratitude to the members of th e Department of Defense Pharmacoeconomic Center for help in acquiring the study data se t and to Christine Bono for her assistance in structuring the data set for analysis.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........8 LIST OF FIGURES................................................................................................................ .......10 ABSTRACT....................................................................................................................... ............11 CHAPTER 1 INTRODUCTION..................................................................................................................13 Background..................................................................................................................... ........13 Need for Study................................................................................................................. .......15 Purpose of the Study........................................................................................................... ....16 Main Research Question.........................................................................................................17 Objectives..................................................................................................................... ..........17 2 LITERATURE REVIEW.......................................................................................................18 Injury......................................................................................................................... ..............18 Incidence of Tendon Ruptures in the U.S. Military................................................................21 Pathogenesis of Tendonopathy...............................................................................................22 Tendon Rupture Risk Factors.................................................................................................23 Quinolones..................................................................................................................... .........23 Mechanism of Action......................................................................................................24 Drug Safety Overview.....................................................................................................25 Connective Tissue Toxicity.............................................................................................25 Animal Chondrotoxicity..................................................................................................25 Human Chondrotoxicity..................................................................................................26 Animal Tendon Toxicity.................................................................................................27 Etiology of Connective Tissue Toxicity..........................................................................27 Human Tendonopathy.....................................................................................................29 Case reports..............................................................................................................29 Cohort studies...........................................................................................................31 Case-control studies.................................................................................................32 General Safety Review....................................................................................................36 Gastrointestinal toxicity...........................................................................................36 Neurotoxicity............................................................................................................36 Phototoxicity............................................................................................................37 Nephrotoxicity..........................................................................................................37 Opthalmotoxicity......................................................................................................38 Cardiotoxicity...........................................................................................................38 Reproductive Toxicity..............................................................................................38

PAGE 6

6 Genotoxicity.............................................................................................................39 Large Administrative Health Care Databases.........................................................................39 Database Validation Methods..........................................................................................40 External Validity.............................................................................................................40 Measuremen t Validity.....................................................................................................41 Data vs. Source information consistency.................................................................41 Internal data consistency..........................................................................................41 3 METHODS........................................................................................................................ .....47 Population Description......................................................................................................... ..48 U.S. Military Demographics............................................................................................48 Overview..................................................................................................................48 Active Duty..............................................................................................................48 Branches:..................................................................................................................48 Patient Selection.............................................................................................................. .......50 Inclusion Criteria.............................................................................................................50 Exclusion Criteria............................................................................................................51 Data Source.................................................................................................................... .........51 The Executive Information and Decision Support (EIDS) Program Office....................51 The Military Health System Manageme nt Analysis and Reporting Tool (M2)..............52 Data Validity.................................................................................................................. .........52 Database Characteristics..................................................................................................53 Internal Validity.............................................................................................................. .53 Incomplete Data...............................................................................................................53 Longitudinal Data Integrity.............................................................................................53 Diagnosis vs. Demographics...........................................................................................54 Diagnosis and Drugs........................................................................................................54 Drugs and Diagnosis........................................................................................................54 The Variables.................................................................................................................. ........54 Dependant Variable.........................................................................................................54 Predictor Variable............................................................................................................55 Additional Covariates:.....................................................................................................55 Analysis....................................................................................................................... ...........56 Descriptive Statistics.......................................................................................................56 General Measures of Association....................................................................................56 Hazard Function..............................................................................................................57 Relative Hazard...............................................................................................................58 Modeling the Hazard Function........................................................................................59 Time-Dependent Exposure..............................................................................................61 Power.......................................................................................................................... .....64 General.....................................................................................................................64 Survival Analysis.....................................................................................................65 Objectives, Research Qu estions, and Hypothesis...................................................................65 Main Research Question..................................................................................................66 Objective 1.................................................................................................................... ...66 Research Question 1........................................................................................................66

PAGE 7

7 Research Question 1 Hypothesis.....................................................................................66 Objective 2.................................................................................................................... ...66 Research Question 2........................................................................................................67 Research Question 2 Hypothesis.....................................................................................67 Objective 3.................................................................................................................... ...67 Research Question 3........................................................................................................67 Research Question 3 Hypothesis.....................................................................................67 Objective 4.................................................................................................................... ...68 Research Question 4........................................................................................................68 Research Question 4 Hypothesis.....................................................................................68 4 RESULTS........................................................................................................................ .......73 Data Validity.................................................................................................................. .........73 Missing Data................................................................................................................... .73 Longitudinal Validity......................................................................................................73 Diagnosis versus Demographics Validity.......................................................................74 Diagnosis versus Drugs Validity.....................................................................................74 Drug versus Diagnosis Validity.......................................................................................75 Patient Demographics........................................................................................................... ..75 Dependent Variable............................................................................................................. ...76 Other Independent Variables..................................................................................................76 Research Question 1............................................................................................................ ...77 Research Question 2............................................................................................................ ...78 Research Question 3............................................................................................................ ...79 Research Question 4............................................................................................................ ...81 5 DISCUSSION..................................................................................................................... ..112 Objective 1.................................................................................................................... ........112 Objective 2.................................................................................................................... ........114 Objective 3.................................................................................................................... ........115 Objective 4.................................................................................................................... ........117 Limitations.................................................................................................................... ........120 Recommendations................................................................................................................ .123 Conclusion..................................................................................................................... .......123 LIST OF REFERENCES.............................................................................................................125 BIOGRAPHICAL SKETCH.......................................................................................................133

PAGE 8

8 LIST OF TABLES Table page 2-1 Tendonopathy risk factors..................................................................................................44 2-2 Common quinolones grouped by PEG@ classifications....................................................45 2-3 Quinolone toxicities leading to market w ithdrawal or decreased clinical significance.....45 2-4 Clinical trial quinolone adverse reactions..........................................................................46 3-1 Key active duty demographic summary............................................................................71 3-2 Conditional probabilities.................................................................................................. ..72 4-1 Normal pregnancy delivery in th e DoD (ICD-9-650.00) by gender and year*.................89 4-2 Benign prostate hyperplasia in the DoD (ICD-9-600.00 600.91) by gender and year*.......................................................................................................................... .........89 4-3 Diabetes mellitus type 1 versus in sulin in the DoD (ICD-9-250.01) by year*..................90 4-4 Insulin use versus diabetes mellitus in the DoD (ICD-9-250.xx) by year*.......................91 4-5 Sample demographics........................................................................................................92 4-6 Case demographics.......................................................................................................... ..92 4-7 Description of tendon rupture cases...................................................................................93 4-8 Independent variables...................................................................................................... ..94 4-9 Model 1 regression parameters for demographic variables...............................................95 4-10 Treatment variable stratified by occupation......................................................................95 4-11 Treatment variable stratified by occupation......................................................................96 4-12 Treatment variable strati fied by provider specialty...........................................................96 4-13 Treatment variable strati fied by provider specialty...........................................................96 4-14 Model 2 full regression parameters for treatment and covariate effects............................97 4-15 Regression parameters stratified by treatme nt, adjusted for covariate effects, and reported based on study relevance.....................................................................................98

PAGE 9

9 4-16 Regression parameters stratified by treatment adjusted for all covariate effects, and reported based on study relevance.....................................................................................98 4-17 Regression parameters stratified by provi der specialty, adjusted for all covariate effects, and reported ba sed on study relevance..................................................................99 4-18 Regression parameters stratified by provi der specialty, adjusted for all covariate effects, and reported ba sed on study relevance..................................................................99 4-19 Regression parameters stratified by occupati on, adjusted for all covariate effects, and reported based on study relevance...................................................................................100 4-20 Regression parameters stratified by occupati on, adjusted for all covariate effects, and reported based on study relevance...................................................................................100 4-21 Model selection for objective 3........................................................................................101

PAGE 10

10 LIST OF FIGURES Figure page 3-1 DoD population description...............................................................................................69 3-2 DoD active duty trend 1990-2003......................................................................................69 3-3 Patient inclusion schematic................................................................................................70 3-4 Illustration of RH(t), RR(t), and OR(t) where the hazard functions for both exposed and unexposed are constant...............................................................................................70 4-1 Total DoD prescriptions by month and year......................................................................88 4-2 Normal pregnancy delivery in th e DoD (ICD-9 650.00) by gender and year...................88 4-3 Benign prostate hyperplasia in the DoD (ICD-9-600.00 600.91) by gender and year...89 4 Diabetes mellitus type 1 versus insu lin use in the DoD (ICD-9-250.01) by year.............90 4 Insulin use versus diabetes mellitus in the DoD (ICD-9_250.xx) by year........................91 4-6 Hazard function stratif ied by treatment group*...............................................................101 4-7 Plotted treatment hazard ratio at 10-day intervals...........................................................102 4-8 Treatment hazard ratio at three selected intervals............................................................103 4 9 Hazard function stratifie d by military occupation*.........................................................104 4 Hazard function stratified by branch*..............................................................................105 4 11 Hazard function stratifie d by provider specialty *...........................................................106 4 Most frequent days of quinolone da ys of supply supply risk estimates..........................107 4 Age interval risk estimates...............................................................................................108 4 Hazard function stratified by antibiotic indication*........................................................109 4-15 Hazard function stratif ied by military grade*..................................................................109 4-16 Hazard function stratified by gender*..............................................................................110 4-17 Hazard function stratified by race*..................................................................................111 4-18 Hazard function stra tified by steroid use*.......................................................................111

PAGE 11

11 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy RISK OF TENDON RUPTURE AFTER QUINOLONE USE IN A U.S. MILITARY POPULATION By Patrick M. Garman August 2007 Chair: Lyle D. Reid Major: Pharmaceutical Sciences Non-battle related injuries are the greatest cause of morbidity and mortality in the US military. Within the category of non-battle injuries, musculoske letal injuries are the leading cause of disability. Tendon ruptures in particular, require long rehabilitation programs and can result in permanent disability. The consequence of tendon rupture is degradation in unit readiness and operational effec tiveness. Case reports and case studies have linked tendon ruptures with quinolone antibiotic usage in the civilian population. This retrospective cohort study utilized a Department of De fense administrative database to estimate the risk of tendon rupture in activ e duty personnel associated with quinolone use compared with the use of cephalosporin antibio tics. It identifies an induction period from quinolone exposure to tendon rupture wh ile also identifying risk fact ors that are associated with an increased risk of tendon ruptures. Data from military personnel June 2005 through May 2006 were collected. Internal valid ity was confirmed using macro-level assessment techniques. Survival analysis was used to estimate the ha zard function and the Cox proportional hazards model was employed to produce haza rd ratios for the main treatment and other independent risk factors.

PAGE 12

12 There was a significant increas e in the risk of tendon ruptures over a 60-day period between active duty personnel who used quinolones compared to those who used cephalosporins (HR=1.65, 95%CI 1.33-2.04). The risk was high est during the 26 to 35 day window which marked the average induction period. Risk factors were days of supply (HR=1.02, 95%CI 1.0181.022); military occupation (HR=1.53, 95%CI 1.24-1.89); age (HR=1.02, 95%CI 1.01-1.03); provider specialty (HR=2.40, 95%CI 1.92-2.99); Marine Corps service (HR=1.65, 95%CI 1.202.28); GI/GU antibiotic indi cation (HR=1.25, 95%CI 0.99-1.57). This is the first study to iden tify an increased risk of tendon rupture from quinolone use in a demographically diverse populatio n. The risk is elevated upon the first day of therapy and increases incrementally until it reaches its ma ximum between days 26-35. The risk of tendon rupture is augmented by advancing age, length of treatment, and working in a physically demanding occupation. These findings are important as they provide eviden ce that a significant portion of this debilitating in jury is avoidable by identify ing specific risk factors and incorporating them into the patient care plan.

PAGE 13

13 CHAPTER 1 INTRODUCTION Background The United States Army Medical Department (AMEDD) has a three-pronged mission in todays military. Its leaders must ensure that the United States Army (USA) can project and sustain a healthy and medically protected force. The AMEDD must be able to deploy a trained and equipped medical force that supports the Army transformati on and it must manage the care of the Soldier and the military family. Diseas e and non-battle injury ( DNBI) is the one of the largest threats to these objectives(1). The contribution of disease to DNBI has cont inued to fall due to immunization programs, appropriate chemoprophylaxis and a heightened command emphasis on hygiene and other public health initiatives(1). N on-battle injuries (NBI) are now the greatest cause of morbidity and mortality in the US military(1). Within NBI, the leading cause of disability in the U.S. military is musculoskeletal injuries. Among MS injuries, tendonopathie s are a common group of conditions ranging from tendonitis to a comp lete tear or rupture of the tendon(2). The occurrence of tendonopathies in the U.S. military has been difficult to quantify, but is believed to be quite common. Tendon ruptures, in particul ar, require long rehabilitation programs and can result in a permanent disabili ty. This lengthy rehabilitation time has the unwanted consequence of degradi ng unit readiness and ope rational effectiveness. Case reports and case-control studies have linked tendon ruptur es with quinolone anti biotic usage in the civilian population (3). This risk factor has not been evaluated in a military population. The use of quinolone antibiotics is of particular importance as it is an avoidable risk factor. This novel class of antibacterial medications is widely used throughout the world. As more quinolones have been synthesized, their spectrum of actively has widened. As their bacterial

PAGE 14

14 coverage has increased, so has their inclusion in treatment re commendations for many kinds of infections, ranging from anaerobic, skin/soft tissue, and upper/lower respiratory tract infections. These new treatment indications have increased th e frequency of their use, and the potential for increases in previously poorly recognized adverse events such as connective tissue disorders. Quinolones are contraindicated in children under the age of tw elve due to connective tissue toxicity seen in juven ile animal studies. Connective tissue disorders include chondrotoxicity and tendonopathies. The first report of quinolone associated tendonitis in adults was published in 1983, while case reports of tendon r uptures began appearing in th e early 1990s(4-6). It is recognized that most tendonopathy is seldom traceab le to a single factor(7). The degenerative process that precedes tendon rupture probably emanat es from a variety of different pathways and causative factors, many of which are unavoidable to the active duty soldier. The term induction period is the epidemiologic term used to explain the period of time from causal action until disease initiation(8). Qu inolone use is one postula ted causal factor that may lead to the initiation of tendon rupture. Th ere is a cascade of intr insic steps post quinolone exposure, that may lead to the ch aracteristic tendon lesion, but at the end of this relatively quick process (days) we do not see an immediate occurr ence of tendon rupture. Data pertaining to a detailed induction period for tendon ruptures related to quinolone use, which includes the induction periods for the other component causes is lacking. For clarific ation, the epidemiologic term latent period is the time interval between disease occurrence and detection(8). The cellular process that leads to a weakened tendon non-with standing, the occurrence of tendon rupture, and its detection is rapid, thus the latent period at the end of the inducti on period is minor. The induction period plus the la tent period is sometimes referred to as the lag time. The lag time is the time from quinolone usage unt il a tendonopathy has been diagnosed.

PAGE 15

15 Traditionally epidemiologic studies try to id entify and induction period by selecting a set of induction periods and calculating the correspo nding set of risk estimates. The estimates would ideally reach a peak when the selected induction period matches the actual induction period. If a pattern emerges, (r ising estimates then falling estimates) that would be evidence that the induction period with the highest estimate would be the best estimate. If no pattern emerges then this method may not be reliable since any peak may be random variability. Instead of selecting a set of arbitrary induc tion periods and looking for a patter n in their risk estimates, the use of the hazard function within the survival analysis model framework, may provide better estimates of the induction period. This study of quinolone antibiotics identifies an avoidable expos ure risk factor, establishes a defined average induction peri od, and describes various other important risk factors for developing a tendon rupture. Need for Study Ensuring the U.S. military is ready to execute its mission is the responsibility of leaders at all levels. Maintaining or in creasing military readiness is of paramount concern and must continually be re-evaluated to identify possi ble opportunities for impr ovement. Non-traumatic musculoskeletal injuries effect military readine ss more than any other medical condition (9-12). Musculoskeletal injuries are frequent occurrences in all branches of the military. They are often found in combat units due to the physical na ture of their occupatio n(13-15). Economically speaking, musculoskeletal injuries account for the largest impact on direct health care costs to the active duty force(10). Te ndon ruptures are musculoskeletal injuries that 1) impact unit readiness because they require longterm rehabilitation, and 2) affect senior leaders more so than younger military personnel (3, 16).

PAGE 16

16 Tendon ruptures have been associated with qui nolone antibiotic us e in case reports and case-control studies(5, 6, 17-22). Identification of an avoidable risk factor for tendon ruptures will help increase unit readiness and decrease the resources currently utilized treating musculoskeletal injuries. Additionally, the iden tification of a specific tendon rupture average induction period, post treatment init iation, will enable be tter counseling of patients in order to reduce tendon rupture risk. The id entification of other risk fact ors for tendon rupture will enable clinicians to pinpoint patients that are at an increased risk and tailor their an tibiotic therapy accordingly. Information from active duty soldiers is the best way to provide answers to these issues. Purpose of the Study This study utilized a large Department of Defense com puter database, which contains encounter and pharmacy data, to estimate the risk of tendon rupture in active duty military personnel associated with quinolone antibiotic usage compared w ith the use of cephalosporin antibiotics. Connective tissue disorders have not been described in patients using the cephalosporin class of antibiotics. It also identifies the averag e induction period from quinolone exposure to tendon rupture using time to event analyt ic techniques. Lastly, risk factors that are associated with an increased risk of tendon ruptures, and are relevant to the active duty force, are identified and their risk estimated. One of the major recommendations of the DoDs Injury Surveillance and Prevention work group for preventing and controlling injuries in th e military services, was to ensure adequate injury research to support preven tion programs(1, 22). Treating this epidemic of injuries in a scientific way allows researchers to employ study designs and methods that yield reliable and valid results, which can be turned into action plans by military leaders. It takes this command involvement to bring about the needed changes to reduce lost duty time and shortened careers.

PAGE 17

17 Main Research Question Estimate the risk of tendon rupture from using a quinolone antibacterial medication relative to an antibacterial medica tion, in a military population Objectives This study uses formal Pharmacoepidemiologic and time at risk methods to estimate the risk of tendon rupture outcomes due to quinolone exposure relative to cephalosporin exposure in an active duty military population. It also estimates the hazard f unction in order to identify the average induction period in the exposure to ou tcome pathway while determining important risk factors for an individual using quinolone antibiotics. This study addresses four specific objectives: 1) Estimate the risk of tendon rupture from qui nolone use, relative to a cephalosporin medication, while adjusting for relevant risk factors and accounting for time at risk. 2) Investigate whether the risk of tendon ruptures from qui nolone use varies over time. 3) Identify the average induction pe riod within the study interval. 4) Identify other major risk factors that affect an active duty military individuals estimated risk for tendon rupture.

PAGE 18

18 CHAPTER 2 LITERATURE REVIEW Injury The realization after Operations Desert Shield/Storm (the Persian Gulf War) that within DNBI, non-battle injuries were becoming the major source for mo rbidity and mortality in the U.S. Armed Forces, resulted in a request in 1994 from The U.S. Army Office of the Surgeon General (OTSG). The OTSG asked the Armed Forces Epidemiological Board (AFEB) for guidance on surveillance, prevention, and control of injuries in military populations. In response, the AFEB formed the Department of Defense (DoD) Injury Surveillance and Prevention Work Group (ISP) to gather information on injuries in the military and make recommendations for future surveillance and prev ention based on their findings. The final report of their findings was published in 1996 and was sent to the Surgeons General of the three military medical departments for implementation in 1997. Ultimately, in 1999, the ISPs work was collected and published as the Atlas of Injuries in the U.S. Armed Forces This extensive text showed that injuries, not illnesses, have th e largest negative effect on the health of military personnel. The Atlas of Injuries in the U.S. Armed Forces graphically illustrates three major facts in the identification of the injury problem. First, injuries are the leading cause of death, disabilities, hospitalizations, and outpatient visits in the military services, relative to other causes of morbidity and mortality. Second, sports, fall s, training-related injuries, and motor vehicle accidents, are the leading causes of injury-rel ated morbidity. Third, unintentional injuries (accidents/mishaps), in particular motor vehicle crashes, are the leading cause of death for all services. The DoD ISP, after studying the issue, concluded, among other findings, that musculoskeletal (MS) injuries are the leading cause of disability in the military(1, 14). They are

PAGE 19

19 responsible for 51% of diagnoses resulting in di sability discharge from the USA(9). These musculoskeletal injuries and their resulting chro nic conditions account for a large proportion of DoD disability cost. MS conditions are the l eading cause of Veterans Administration (VA) disability payments (l ifetime costs of $485 million to newl y disabled Army personnel in 1993) (9). Injuries and their MS sequelae are the l eading causes of hos pitalization in the DoD (14). Wojik, et al. (2004) published a paper in the Ameri can Journal of Industria l Medicine that looked at disease and non-battle injuri es based on Persian Gulf War admission rates(23). They found that musculoskeletal and connective tissue di agnoses accounted for 12.6% (1,718) of the total non-battle related injuries duri ng operation Desert Shield and ope ration Desert Storm. Gordon, et al. (2000) in their article enti tled Hospitalization Due to Injuri es in the Milita ry published in the American Journal of Preventative Medicine evaluated the current data and found combat injuries represent a small part of the injury probl em in the U.S. Military. Most injuries in the military occur in similar ways to those in the ci vilian world, where injuries are recognized as a leading cause of morbidity, disabil ity, and death in society as a whol e. Data from the first Gulf War suggest that injuries and musculoske letal conditions accounted for 39% of all hospitalizations during the operation and less than 5% of all hospitaliz ations were combat related. Musculoskeletal and connective tissue disorders (IC D code group 710-739) comprised 14% of all hospitalizations, many of which were the chronic or recu rrent effects of injuries that occurred before deployment. In 1992, the mu sculoskeletal diagnosis group accounted for 28.1 (Army), 9.7 (Navy), 12.0 (Airforce) per 1000 pers on-years, this represented the number one reason for hospitalization in the Army, second in the Navy and second in the Airforce. Results from studies concerning NBI injuries in other conflicts support thes e MS injury trends (24, 25).

PAGE 20

20 MS injuries associated with physical training and athletic/spor ts injuries are a major cause of training time loss and are the major contributor to the occurrence of non fatal injuries(9). In the Army and Airforce athletic/physical training injuries are more common that motorvehicle injuries and correspond to an incidence rate of 3.2 / 10 00 in the Army and 2.2 in the Airforce from hospitalization data(12). Lauder et al. (2000) studied spor ts and physical training injury hospitalizations in the U.S. Army and pub lished their findings in th e American Journal of Preventative Medicine. Since injuries are the l eading health problem in the military service, Lauder looked at the amount of injuries caused by sports and physical training in the Army. They found that during the six-year study pe riod there were 13,861 hospital admissions for injuries resulting from sports and Army physi cal training. The rate s were 38 and 18 per 10,000 person-years for men and women resp ectively. Sports injuries resu lted in 29,435 lost duty days each year. Musculoskeletal injuries were respon sible for 82% of all sports or physical training injuries(12). This finding dem onstrates a direct impact on military readiness. Risk factors for injury during military training are, female gende r, BMI, smoking, ethnicity, flat feet, illness or injury in the past year, no previous military expe rience and baseline fitness level. Ankle injuries, which include Achilles te ndon ruptures, were the 3rd most common injury diagnosis. The studys major finding was that sports and physical trai ning are responsible for a large number of lost duty days per year. It is of inte rest that the majority of these injuries were musculoskeletal in nature and resulted in a direct reduction of military readiness. Musculoskeletal injuries are frequent occurrences in all branches of the military. They are often found in combat units due to the physical na ture of their occupation(13-15). Non-traumatic tendon ruptures are musculoskeletal injuries that 1) impact unit readiness because they require

PAGE 21

21 long-term rehabilitation, and 2) affect senior leaders more so than younger military personnel (16). Incidence of Tendon Ruptures in the U.S. Military The Department of Army (DA) through the Ce nter for Health Promotion and Prevention Medical (CHPPM) published a repo rt in its Medical Surveill ance Monthly Report (MSMR), on spontaneous ruptures of the Achilles tendon, (16) This study chronicles incidence rates and trends for non-traumatic Achilles tendon rupt ures among active duty service members between January 1998 and May 2001. In 2001, non-traumatic Ac hilles tendon ruptures occurred at a rate of 51.3 per 100,000 person-years(3) In this sa mple, non-traumatic tendon ruptures from all anatomic sites summed to 165 per 100,000 person year s (3). This incidence rate is higher than the 6 to 18 per 100,000 person years seen in studies utilizing a civilian population sample(21, 2629). This study identifies older individuals, Blac k race, and male gender as subgroups at highest risk. The increased incidence in older milita ry personnel compared to more junior ones illustrates a disproportionate impact on senior lead ers. An important finding of this study is the identification of a trend of in creasing rates of tendon ruptures of all kinds during the study period. Achilles tendon ruptures had the shar pest increase starti ng at 22.7 per 100,000 person years in 1998. Also of import is the observed higher incidence rate in a military population compared to a civilian one. Now that the general toll of MS and in par ticular non-traumatic tendon rupture injuries in the military has been reviewed, it is time to study specific causes of MS injuries and apply preventative measures to redu ce their occurrences. Injuries must not be thought of as unpreventable random events, but as a predictable re sult of many risk factors. Important risk factors can be identified and controlled, whic h will lead to a reduction in injuries and the resulting morbidity and mortality. Tendon ruptures are important injuries that are associated

PAGE 22

22 with many risk factors. Identification of an avoidable risk factor for non-traumatic tendon rupture will help increase unit r eadiness and decrease the resources currently utilized treating non-traumatic musculoskeletal injuries. Pathogenesis of Tendonopathy Tendon tissue is a matrix that is well suited to perform its major f unction of transferring force from muscle to bone. It is not a static tissue and is consta ntly being remodeled to adapt to the frequency and force applied. The term tendonopa thy refers to disorders primarily affecting tendons, including chronic pain, inflammation, and tendon rupture. Tendonitis is characterized by chronic pain and the assumption of an in flammatory response to the tendon injury. Spontaneous tendon rupture describes ruptures that occur without a ny noticeable preceding clinical symptoms. This is a misnomer as he althy tendons have much higher tensile strength than is required for normal activities. Because of this, tendon ruptures are rarely spontaneous and must be related to at leas t some degree of tenocyte degeneration. The cause and nature of this pathological degeneration is not settled. Many systemic diseases may compromise tendon strength and elasticity, or result in an inflammatory process. There are two main theories that attempt to explain tendon ruptures. The first is the mechanical hypothesis, which says that repetitiv e micro-trauma or a single physiological load that strains the tissue above a certain percentage of length causes tendon ruptures. The second theory relies on the tendon vasculature. Accordi ng to this theory, injury occurs as a result of diminished blood flow to the tendon because of aging, vascular disease, physical disuse, or trauma. In reality, most tendonopathies are probably a combination of mechanical and vasculature in nature and are init iated by a variety of different causa tive factors that lead to the degenerative process.

PAGE 23

23 Tendon Rupture Risk Factors Identification and avoidance of a risk factor for non-trauma tic tendon ruptures will help increase unit readiness and decrease the reso urces currently utilized treating non-traumatic musculoskeletal injuries. Overall, risk factors for tendon injuries ar e listed in table 1. Increasing age, male gender, and Afro-American ethnicity, have been associated with an increased risk of non-traumatic tendon ruptures in military personnel (3). Case reports and observational studies of non-military personnel treated with quinolone antib iotics have demonstrated an increased risk of tendon rupture(19-22, 30-34). Other a ssociated risk factors include renal dysfunction/dialysis/transplantation, rheumatic disease, gout, diabetes mellitus, sports participation, hyperparathyroidism, and hypoparat hyroidism(18). Case-control studies have reinforced these factors linked to an increased risk of non-traumatic te ndon ruptures (5, 6, 19-22, 30, 35, 36). Quinolones Quinolone antibiotics were first developed in the early 1960s with the introduction of nalidixic acid (2). The first-generation quinolone s have bactericidal activity over most common Gram-negative bacteria(37). In the 1980s, w ith the introduction of norfloxacin, the first fluorinated quinolone derivatives began to a ppear(38). These second-generation quinolones include norfloxacin, ciprofloxaci n, and ofloxacin, among others, and exhibit an increased, yet still mainly Gram-negative, antibacterial spec trum. In the following years, more secondgeneration quinolones have been created with more balanced broad-spectrum activity that encompass some Gram positive as well as Gram negative coverage. Temafloxacin, grepafloxacin, and sparfloxacin, are a few of th e more recent second-generation quinolones. During the 1990s, a third generation was synthesi zed with levofloxacin being the most popular member. This generation is characterized by an enhanced Gram-positive and atypical pathogen

PAGE 24

24 coverage, which makes it suitable for infections of the respiratory tract, while still maintaining the basic quinolone Gram-negative activity. Mo st recently, a fourth generation has been designated. This group is distinguished by an in creased anaerobe coverage. This coverage is added to the third generation, which includes m oxifloxacin and gatifloxacin. There are several organizational methods used to classify quinolone antibiotics based on thei r structure and activity relationships (SARs). The four-generation cate gories used here were developed by the PaulEhrlich-Society or Chemotherapy (PEG) and ar e commonly employed to reflect both the drugs chemical and functional characteristics. Some researchers have utilized different category representations to accentuate th e distinction between either th e original quinolone chemical entities and their subsequent fluorinated offspri ng or the unique functionalities of the different SARs. The former distinction is becoming obsolete with the synt hesis of newer quinolones with fluorine atoms at different posit ions or the absence of fluorin ated derivatives altogether. Throughout this study, the term qui nolone will be used to desc ribe all non-fluorinated and fluorinated quinolone derivatives. Mechanism of Action Quinolones exert their antibacterial activ ity through the inhibition of bacterial topoisomerase II. This interference of this DNA gyrase enzyme leads to inhibition of DNA synthesis in nuclei and mitochondria. The specifi c binding reactions to topoisomerase II and the subsequent damage to bacterial DNA are beyond th e scope of this text, but this mechanism of action makes quinolones rapidly bactericidal. As a class, they exhibit concentration-dependent killing, which makes them highly ef fective and clinically useful in treating a range of bacterial infections. Quinolones are comm only used in the treatment of diseases ranging from urinary tract infections, sexually transmitted diseases, resp iratory infections, gastrointestinal infections,

PAGE 25

25 skin and soft tissue infections, a nd bone infections. They are also used empirically to treat fever in neutropenic cancer patients and for prophylaxis in many surgical procedures. Drug Safety Overview Quinolones have an important toxicity a nd adverse effect profile that has been demonstrated in animals as well as in humans. Pharmaceutical manufacturers routinely do precli nical toxicity studies in animals. The results of these studies are usually not publishe d in peer-reviewed journals, which make them difficult to evaluate. The relevance of these studies becomes smaller as human information, from clinical studies becomes available. Still, animal studies are important as they produce a range of possible toxicities that gi ve the clinical investigator fore warning of what to look for in human patients. Human trials have established quinolones to be a relatively safe and well-tolerated class of antimicrobials. The overall adverse event rate in most clinical trials is very similar to the control group whether the group was a placebo or another antim icrobial. This should not be implied to mean that they do not have a unique safety risk profile that is conse quential when weighted against the possible benefits of quinolone therapy. Connective Tissue Toxicity Connective tissue toxicity is a well-known adverse effect that has been noted to occur in all quinolones so far tested. Effects on connective ti ssue are broken down into chrondrotoxicity and tendonopathies. Animal Chondrotoxicity Ingham et al. (1977) first descri bed chrondrotoxicity in quinol ones, consisting of articular lesions in the cartilage of juven ile dogs. Young beagle dogs were given pipemidic acid orally for 1-15 days at doses of 200-1000 mg/kg. Lameness o ccurred by way of gait abnormalities within

PAGE 26

26 the first 3-7 days after starting the treatment. Upon inspection, articular cartilage lesions were identified in major joints such as the hip, wrist, elbow, ankle, and shoulde r. Lesions were also detected in the distal growth plates (epiphyses) of the humerus, femur, proximal tibia, and tarsal bone in rats. These lesions started as blisters before progressing to ulcerative erosions, which were considered irreversible. When these lesion s were examined histologically they revealed a general necrosis of chondrocytes. It has been postulated that quinolones di rectly or indirectly damage collagen fibrils, which leads to hyperhyd ration and osmolality changes in the cartilage matrix. These changes predispose the cartilage to blister formation from the mechanical pressure of body weight while ambulating. Th e clinical recovery, in the fo rm of a reduction in lameness, took place over a 2 to 3 week pe riod after cessation of the drug, although cartilage lesions were still present up to 3 months. D ogs between 2.5 to 5 months old were most at risk for the toxic effect. Following this study, othe r investigations repli cated the toxicity of this compound as well as uncovering the same adverse effects in all other available quinolones across many different species of laboratory animals. Because of the propensity of quinolones to induce toxic lesions in the articular cartilage of distinctly juvenile anim als, quinolones have been restricted from use by children and growing adolescents. Human Chondrotoxicity Chondrotoxicity caused by quinolon es has only been directly demonstrated in juvenile animal models. Because of the resulting arthropa thies identified in animals, these drugs are restricted in children under the age of twelve. Consequent ly, clinical evidence of this side effect is small due to their limited use in children as well as the fact that cartil age lesions are not always associated with clinical symptoms. The general de ficit of clinical evidence should not be used to minimize the effect of irreversib le cartilage toxicity, as clinic al symptoms may arise over a long period of time. The best evidence of clinical arthropathies in humans comes from the use of

PAGE 27

27 quinolones in children with cystic fibrosis. In this internat ional review of 2,030 courses of ciprofloxacin treatment in 1,795 children, the inci dence of arthralgia wa s 1.5% (39). All cases resolved without intervention. Case reports of arthralgia have been re ported with quinolones other than ciprofloxacin, in cluding nalidixic ac id, pefloxacin and norfloxacin(4, 40, 41). Animal Tendon Toxicity Tendon tissue is also effected by quinolone use. Starting with Kato et al. (1997) researchers have consistently f ound structural changes in tendon a nd tendon associated tissues in rats. Edema and mononuclear cell infiltration in the inner sheath of the Achilles tendon, and infiltration surrounding the adjacent synovial membrane and joint space marked the tendon lesions induced by quinolones early trials. Later, Sendzik et al (2005) showed that the final event in the pathogenisis of qui nolone induced tendonapathies is in fact apoptosis of tenocytes, which leads to necrosis of the tendon tissue. In a study comparing the effects of 10 different quinolones on rat Achilles tendon tissue, Kashida and Kato, (1997) showed that there is a difference in the toxic potential within the drug cl ass. Pefloxacin and fler oxacin were the most toxic while sparfloxacin, norfloxaci n, and ciprofloxacin, resulted in less lesion formations. The researchers suggested that this difference in toxicity is more a result of pharmacokinetics, as pefloxacin and fleroxacin have higher absorp tion rates than the ot her quinolones. Etiology of Connective Tissue Toxicity Early reports attempting to describe the mechanism for quinolone induced connective tissue toxicity centered on thei r inhibition of DNA synthesis in nuclei and mitochondria via their interference with the topoisomer ase II enzyme. Even though quinol ones have a low affinity for eukaryotic DNA several studies showed that qui nolones my target chondrocytes and tenocytes more so than other cell types thus initiating the typical lesions. Later, different evidence pointed toward the inhibition of synthesi s, rather than the stimulated degradation of collagen. This

PAGE 28

28 hypothesis, first put forth by Schroter-Kermani et al. (1992), and reinforced by several others, was evidenced by several factors. These resear chers showed that quino lones were responsible for compositional changes in indi vidual proteoglycans and glycos aminoglycans in rats and dogs. These changes along with the inhibition of soluble prostaglandins work to reduce chondrogenesis. Original resear ch done by Stahlman et al. (1993) diverged from these two hypotheses by showing that the prim ary event in quinolone chrondrot oxcity is the negative effect they have on magnesium-dependent integrins in articular cartilage. This brings the basis of quinolone toxicity to a long know n characteristic of quinolones known as chelation. Chelation occurs when metallic and nonmeta llic substances combine to form stable chelate complexes. This attribute is well known for its effect on the pharmacokinetics of quinolones as their absorption is reduced if dior trivalent-metal ca tions are present in the GI tract when quinolones are orally administered. Typically, magnesium (Mg2+), aluminium (Al3+), or calcium (Ca2+), containing agents such, as anta cids or vitamin supplements, form poorly absorbed chelated compounds probably consisting of quinolone molecu les and metal cations. Stahlmans findings hinted that by forming stable chelate complexes, systemically within th e connective tissue, with magnesium, quinolones affected the el ectrolytic balance, impaired a specific family of integrins, and eventually caused degeneration of the collagen matrix and irreversible articular damage. This hypothesis has become the accepted m echanism for quinolone induced connective tissue toxicity for several reasons. Both tendon and cartilage tissues ar e characterized by low vascularization and similar matrix components. Because of the poor vascularization any change in nutrient or electrolyte balan ce in the surrounding e nvironment can not be corrected quickly. This vascular isolation enhances the chances that chelate complexes formed by quinolones will deprive important integrin proteins of magnesium which will lead to chrondocytes and tenocytes

PAGE 29

29 losing integrity and releasing harmful inflammatory radicals into the local cellular area. This notion has been reinforced by the fact that an imals fed a magnesium deficient diet develop similar lesions as the ones induced by quinolones. Shakibael et al. (1999) and Lozo, et al. (2002) demonstrated that rats fed a magnesium deficien t diet while treated with quinolones had a higher occurrence of degenerative alterations in teno cytes than rats fed a normal magnesium diet. Recently Pouzaud, et al. (2003) showed that th e toxic effects of quinol ones on tendon cells were partially related to reac tive oxygen species production (ROS). Increased ROS contributed to the damage of cartilage and tendon microstructure se en in rabbit tendon cells. This increased ROS production may be directly caused by quinolones, or may result from cellular proteolytic activity caused by the magnesium deficient degenerative pr ocess. Quinolone induced connective tissue toxicity is mostly likely a degenerative pro cess that results from a variety of different microenvironment pathways. Its pathological fe atures are similar in both cartilage and tendon tissues indicating that quinol one induced chrondrotoxicity and quinolone-induced tendonopathy are probably different clinical manifestations of the same toxic effect on cellular components of connective tissue structures. Animal and human models have shown it to be an accepted, identifiable, toxicity that diffe rs in intensity across different age groups and individual quinolone compounds. Human Tendonopathy Case reports Tendonopathies linked to quinolone use usually involve tendonitis that may or may not progress to a tendon rupture. The first report of tendonitis related to quinolone use was in 1983 and involved two post renal transplantation pa tients who developed tendonitis while being treated with norfloxacin(4). Sin ce then, there have been over a thousand case reports to various

PAGE 30

30 governmental surveillance agencies associa ting most quinolones with tendonopathies (2, 18, 4245). Depending upon the source of case reports used, re view articles have show that roughly half of these tendonopathies are tendon ruptures (18, 44, 45). Characterizations of spontaneous tendon ruptures post quinolone therapy are numer ous, dating back to 1992 when seven patients with Achilles tendonitis, includi ng 3 with rupture following quinolone therapy were reported (5). Since then tendon rupture has been accepted as an infrequent but important adverse reaction to quinolone therapy. The most common injury site is the Achilles occurring in about 90% of the cases and 40% of these occurred bilaterally(18, 35, 43, 44, 46). The Achilles tendon is probably affected the most because it is the main weight-bearing tendon. Ruptures have also occurred at the triceps tendon, flexor tendon (finger), thumb, patellar, s upraspinal tendon, quadriceps, subscapularis and rotator cuff tendon(2, 18, 21, 44). Overall, the mean time of onset of tendonopath ies has been estimated to be 17 days with half occurring within the first six days post initiation of quinol one therapy. Rupture of a tendon had a mean onset of 25 days but the median was 6 days. Case reports show that tendonopathies have occurred as soon as 2 hours post treatment initiation and as late as 6 months after the medication has been discontinued (2, 18, 21, 31, 44, 47). Pefloxacin induced tendonapathies account fo r 37% of case reports while ciprofloxacin was implicated in 25%. Norfloxacin is res ponsible for 11%, levofloxa cin 8%, and ofloxacin 6%(44). In one case tendonopa thy symptoms were eliminat ed by lowering the dose of norfloxacin, and upon rechallenge at the higher dose, the symptoms reappeared(4). Diagnosis was mostly by physical examination and tr eatment usually involved discontinuing the

PAGE 31

31 medication. Anti-inflammatory pain medications physical therapy, immobilization, and surgery were all documents interventions to treat the tendonopathy. Rec overy took a mean of 59 days and lingering sequelae of swelli ng, bruising, difficulty in walki ng, decreased flexion, and pain were reported(44). The mean age for case reports is 59 years with a range of 28 to 92 years. Men seem to be affected more than women by 2:1. Interestingly, reviews in the Netherlands, Sw eden, and Switzerland all show sharp national increases in tendonopathies when a newer third or fourth generation quinolone was added to the national formulary (18, 29, 45). The increases were characterized by large jumps in tendon ruptures that were not attributed to young active males but to older patients that were administered a corticosteroid in conjunction with the quinolone fo r a respiratory tract problem(18, 29). The new generations of quinol ones, with their increased Gram-positive coverage, have become the drug of choice for upper respiratory tract infections such as community acquired pneumonia. This increase in cases may be due to the increased prescribing of quinolones because of their increased inclusion in treatment recommendations or because of a yet undetected, enhanced connective tissue toxicity of the newer quinolones. The exposure rate to quinolones was estimated at 2% in the Dutch case review study. It is worth noting here that th e incidence of tendon rupture al so increased, mostly due to a doubling of the occurrence of Achilles tendon ruptur es, in the U.S. military from 1999 to the end of the Medical Surveillance Monthly Repor t (MSMR) incidence study in 2001 (3). Cohort studies Two cohort studies helped solidify the esti mated incidence rate of quinolone induced tendon disorders in otherwise healthy individuals. The rate is characteri stically low at 0.14 and 0.4% respectively(35, 48). Van der Linden et al. (1999) employed a retrospective cohort study design using data from the Integrated Primary Care Information system (IPCI). The IPCI

PAGE 32

32 contained computerized patient records from general practitione rs throughout the Netherlands monitoring about 250,000 patient reco rds. In the end, this study was woefully underpowered to find an association as after patients were excl uded for various reasons they were left with n=1,841 in the quinolone group and n= 9,406 in th e control group. There was a significant nested finding for ofloxacin and Achilles te ndonitis of RR 10.1 (95%CI: 2.2-46.0). Ofloxacin treatment was associated with an increased risk of 15 cases pe r 100,000 days. The increased risk of tendonopathy attributed to qui nolone therapy was 0.4%(35). Another study completed in England compared the results of five observatio nal cohort studies that reported the safety of ciprofloxacin, norfloxacin, oflox acin, azithromycin, and cefixime. The three quinolones were compared to the macrolide and cephalosporin safety findings. All samples were less than n=10,000. The average rate was 0.14 % in the quinolone group and 0.03% for the other antibiotics. Of note, there were four tendon ruptures in the qu inolone group and zero in the nonquinolone group(48). Case-control studies More recently, three case control studies have been completed. The first centered on the risk of Achilles tendonopathy in general and the other two on Achilles tendon rupture in particular. In the first study, researchers from the Nether lands conducted a nested case-control design of quinolone users whose records were contained in the IMS Health database (UK MediPlus). This database contains data from general prac tice (GP) visits on a sour ce population of 2 million. The researchers established four categories of exposure to quinolones: current use, recent use, past use, and no use. Current use was defined as day 1 of drug therapy through 30 days past the calculated end date. Recent use was between 30 to 90 days post therapy end date, and past use was greater than 90 days after the calculated e nd date. Analyzing the data using logistic

PAGE 33

33 regression they adjusted for age, sex, number of GP visits, use of a cort icosteroid, calendar year, obesity, and history of musculos keletal disorders. The cohort contained 46,776 patients of which 704 were cases. Cases were 61% female and had a mean age of 56 years. Significant covariates included age, number of GP visits gout, obesity, and use of cortic osteroids. The overall adjusted relative risk (RR) for Achilles tendon disorders for current use of quinolones was a modest but significant 1.9 (95% CI: 1.3-2.6). Recent and pa st use was comparable to no use. Other important results were that patients over the age of 60 had an Achilles tendon disorder current use RR of 3.2 (95%CI: 2.1-4.9). When the tendon disorder was stratified the RR of Achilles tendon rupture in current use patients over 60 ye ar old was 7.1 (95% CI : 1.7-29.1) and similarly the RR for tendonitis was 3.1 (95%CI: 2.0-3.8). This study reinforces results from case reports and smaller cohort studies that this adverse event is relatively small with an excess risk of 3.2 cases per 1,000 patient years in th is study. The risk was highest in patients over the age of 60 with concomitant use of corticosteroids(19). In the second population-based case-control st udy, the same research group utilized the General Practice Research Database (GPRD) fr om the United Kingdom to carry out a similar research project. This studys main objective was to quantify the risk of Achilles tendon rupture from quinolone use, and to report on concomitant risk factor(20). The GPRD contains GP entered medical information on approximately ei ght million residents of the United Kingdom, of these 50,000 persons were randomly selected to ac t as controls. This study design and methods are similar to the previous research article with a few differences. Adjustments were used for age, sex, cortical steroid use, history of musculoskeletal relate d disorders, disorders of lipid metabolism, organ transplants or hemodialysis, a nd number of GP visits. They also performed a stratification analysis by age, se x, and concomitant use of cortic al steroids. The researchers

PAGE 34

34 identified 1,367 occurrences of Achilles tendon rupture from January 1, 1988 to January 1, 1999 that after review were included in the study. The identification of exposure was broken into current, recent, and past use of quinolones defined the same as in the previous study. The cases were 69.4% male and had a mean age of 48 years. The use of quinolones was significantly lower in males compared to females and higher in patients between 60 and 79 and over 80 years old compared to patients 59 and younger. Exposure to any quinolone was 4.5% of cases and 2.0% of controls. The overall Odds Ratio (OR) for Ac hilles tendon rupture was 4.3 (95% CI: 2.4-7.8) for current exposure and 2.4 (95% CI: 1.5-3.7) for recen t exposure. When the results were stratified by age the OR was 6.4 (95% CI: 3.0-13.7) for patients between 60 to 79 years and 20.4 (95% CI: 4.6-90.1) in patients over 80 years. This st udy found no cases of Achilles tendon rupture in patients under the age of 60 after exposure to qu inolones. The OR in the non-cortical steroid stratification associated with current exposure was 5.3 (95% CI: 1.8-15.2) and ballooned to 17.5 (95% CI: 5.0-60.9) and 18.4 (95% CI: 1.4-240.2) in the current a nd recent quinolone and steroid users group. This analysis utilized only oral co rtical steroid use, even though injectable steroids are sometimes used in tendonitis tr eatment. Lastly, the overall absolute risk of Achilles tendon rupture was 5.5 in patients 60-79 years and 3.5 in patients 80 years and older per 100,000 personyears. The attributable risk pe rcent was 2.2% and 6.3% respectively. The third case-control study completed by Seeg er, DS et al. (2005), investigated the association between Achilles tendon rupture (ATR) and quinolone exposure(36). Secondarily, they attempted to quantify and account for other ri sk factors related to ATR. The researchers utilized the Ingenix Research Database, sourced from United Healthcare to conduct a nested case-control study. The population included pati ents with commercial as well as Medicare supplement health insurance. Case selection wa s stringent in order to reduce misclassification

PAGE 35

35 bias. Cases were persons with both a diagnosis and a procedure (s urgical or non-surgical) related to ATR occurring between 1 January 1997 and 30 June 2001. All cases were affirmed through medical record review. Controls were randomly sampled from the same database at a 20:1 control to case ratio. The cont rols were categorized by age (017, 18-59, 60+ years) to achieve nearly the same distribution as the cases. Expos ure and risk factors were created from medical and pharmacy claims data over six months prior to the index or case date. When evaluating cases, this study found that 55% of potential case s were confirmed to have been ATR. The potential cases that were not ATR tended to be diagnostic rule-outs or cases of trauma or procedure related to the Achille s tendon but not associated with ATR. In all there were 947 cases and 18,940 controls. The association be tween quinolone antibiotics and ATR was not significant (OR=1.2; 95% CI=0.9-1.7). The associ ation was stronger but s till not significant, with higher cumulative exposure to quinolones (O R=1.5;95% CI=1.0-2.3). Risk factors for ATR included; trauma (OR=17.2; 95% CI=14.0-20.2), male gende r ((OR=3.0; 95% CI=2.6-3.5), injected corticosteroid inj ection (OR=2.2; 95% CI=1.6-2.9), ob esity (OR=2.0; 95% CI=1.2-3.1), rheumatoid arthritis (OR=1.9; 95% CI=1.0-3.7), sk in or soft tissue infections (OR=1.5; 95% CI=0.9-2.3), oral corticosteroids (OR=1.4; 95% CI=1.0-1.8), and non-quinolone antibiotics (OR=1.2; 95% CI=1.1-1.5)(36). This study suggested that the risk of ATR after quinolone use is not different among quinolones or from that associ ated with corticosteroids. The researchers also found no potentiation of ATR risk among persons exposed to corticosteroids and quinolones. The main result of this study was to id entify trauma, obesity, arthritis, male gender, and injected steroids as risk factors for AT R. The associations in this study between tendonopathies (tendonitis, TR) are consistent with a causal pathway that put s tendonitis or other tendonopathies as an intermediate stage that ev entually progresses to tendon rupture(21, 36).

PAGE 36

36 General Safety Review Gastrointestinal toxicity The most common adverse effects reported by hu mans are gastrointestinal in nature. Studies testing gastric emptying and intestinal motility in anim als suggest that quinolones do not have a direct effect on the gastro intestinal (GI) tract. Even af ter long-term use, no histological toxicity has been shown in the GI tract of animals taking quinolones. Vomiting has been demonstrated in animal and human models and is believed to be a central effect, as opposed to a direct irritation of the gastrointestinal mucosa. Diarrhea is a common side effect in animals and humans given quinolones. This effect is believed to be from the increased release of toxins from pathogenic bacteria whose con centration tends to grow when normal intestinal flora are destroyed by the antibiotic medication. Naus ea, vomiting, abdominal pain, and diarrhea have been described in all quinolones. The frequency of GI side effects, ranging from 0.8 to 6.8% in clinical trials, is not generally higher than other antimicrobials. Neurotoxicity Neurotoxicity is a common finding in animal models. While the mechanism of action is unknown, the administration of quinolones to rats ha s led to convulsive seiz ures. Studies have advanced two possible explanations that entail the inhibition of the -aminobutyric acid (GABA) inhibitory neurotransmitter or a decrease in CNS magnesium concentration probably caused by the chelating of quinolones and meta l cations. There is currently no model to predict the seizure potential of each individual quinolone. In humans neurotoxicity can be divided into minor Central Nervous System CNS side effects and the more severe events that requir e discontinuation of ther apy. Minor reactions consist of headache, dizziness, tiredness, and sleeplessness along with some reports of vision distortion, bad dreams, and restlessness. The more serious reactions occur at a rare >.05% rate

PAGE 37

37 and include depression, grand ma l seizures, psychotic reactions and hallucinations. The medication dose is related to CNS adverse effects as higher doses increase the incidence in safety studies (bowie et al. 1989). Phototoxicity Phototoxicity has been shown, in animal models, to occur in all quinolones. This toxicity is related to the compounds rela tive photoinstability and degrad ation process that induces the creation of tissue damaging free radicals. This degradation process caused by UV radiation, has also been shown to be photomutagenic and photocarcinogenicic in some quinolones. Clinical manifestations of phot otoxicity include mild erythema to severe bullous eruptions in areas of the skin exposed to UV radiation. A ll quinolones have exhibited this side effect to one degree or another. The interesting point in the evaluation of th e likelihood for individual quinolones to cause phototoxicity is that it can be estimated by measuring the rates of compound degradation when exposed to UV radiation. The higher the degradation ra te the more extensive the cellular damage will be. Quinolones such as fleroxacin, lomefloxacin, and sparfloxacin have the highest phototoxic pote ntial. Hypersensitivity reactions ha ve an incidence rate of 0.4 to 2.1% and include eythema, pruritus, uticaria, rash, among other skin reacti ons. These are reported separately in clinical trials but may be a result of low level phototoxic reactions. Nephrotoxicity Nephrotoxicity in the form of crystalluria in neutral or al kaline pH conditions has been demonstrated in rats and monkeys administered norfloxacin or ciproflo xacin. These two drugs are slightly soluble in neutral or high pH environments found in these animalss urine whereas human urine has a lower pH, which reduces the risk.

PAGE 38

38 Opthalmotoxicity Ocular toxicity has been demonstrated in fe line retinas when treated with nalidixic acid and to a lesser extent when treated with norfloxa cin. No other quinolone has been reliably found to have ophthalmotoxicity. Cardiotoxicity Cardiotoxicity in the form of hypotension, tach ycardia, bradycardia, and dysrhythmias, has been induced after intravenous injection in cat s and dogs. A more recently identified adverse effect related to newer quinolone s is the prolongation of the QT in terval. This effect has been produced in dogs administered newer quinol one compounds such as sparfloxacin or moxifloxacin. The most common cardiovascular side effect, in humans, of most quinolones is either hypotension or tachycardia upon init iation of systemic treatment. One of the most severe consequences of quinolone therapy is the i nducement of a pronounced QT-interval prolongation, which can lead to a dangerous polym orphic ventricular tachycardia called torsade de pointes. Grepafloxacin was withdrawn from the market due to the inducement of this cardiovascular adverse event, while sparfloxacin and moxifloxaci n both carry warning. Generally, the safety of any quinolone not reported to cause QT-prol ongation should not be assumed pending ongoing post marketing surv eillance studies. Reproductive Toxicity Reproductive and developmental toxicity has been accessed in animals with mostly negative results. While there are toxic changes that have been observed in rats, rabbits, dogs, and monkeys, during the organogenesis period, these changes have not resulted in any irreversible overt teratogenicity either perior postnatal. Due to these toxic changes in animal studies, quinolones are not considered safe to use during human pregnancy. In rat models,

PAGE 39

39 several quinolones induced fertility disorders in ma les. This male reproductive toxicity consists of histopathological changes in the testes that led to impaired spermatogenesis. Genotoxicity Quinolones exhibit some genotoxicity th rough their affinity, although at several magnitudes less than bacterial enzymes, for e ukaryotic topoisomerase s,. Many studies in animals have shown that quinolones with increas ed Gram-positive antibacterial coverage are generally more toxic to mammalian DNA. Wh ile these results reveal DNA damage and induction of mutations and chromosomal aber rations there has been no indication of a carcinogenic effect even after li fe-long animal drug exposure. Large Administrative Health Care Databases Large administrative health care databases have been developed primarily for financial accountability and planning reasons. Health care administrative databases have recorded huge amounts of clinical information in diverse settings throughout the world. The ability of these databases to link multiple aspects of care for an individual patient in a longitudinal fashion have solidified them as the main source of populatio n-based pharmacologic res earch and health policy evaluation. These databases have different features, such as the number of variables collected, size of data, and length of follow-up, depending u pon its intended use. Examples of common health care administrative database s used for drug utilization a nd quality improvement research are the Group Health Cooperati ve of Puget Sound, the Manitoba Health Service Commission, the Medicaid Management Information System, and th e Medicare database(49). The advantages of using a health care database are quick access to a large well defined population and the ability to select both cases and controls, or exposures to the drug of interest (Hartzema in press). Disadvantages stem from the fact that the normal everyday health care that is captured does not take place in a controlled environment. During patient encounters and subsequent coding of

PAGE 40

40 diagnostic information, there are many opportunities for error, biases, an d misunderstandings to occur. The use of administrative databases fo r health research is wide spread, but less consideration has been paid to the validity of the health care in formation stored within them. Database Validation Methods When evaluating the validity of health car e databases there are two central components that must be examined. First, the external valid ity of the database shoul d be explored to show the extent that results abstracted from a database can be generalized to ot her settings. Next, the internal validity ought to be scrutinized to judge if the inform ation measures what it purports to measure. When studying databases this quality of internal validity is more specifically referred to as measurement validity. Evaluating measurement validity is accomplished by investigating the consistency of the subject database compared to either external clinical records, or particular data files internal to the subject database. External Validity Database external validity is a subjective area where the question of whether the information obtained from a health care databa se is representative of different health care settings and different health care environments in the world at large. Th is evaluation is usually referred to as generalizability a nd is lacking in objective met hods to calculate a quantifiable appraisal. The most voiced concern generally centers on the differences in the various demographics covered by the database population. How generalizable is the Medicaid database, when its population is poor, unemployed, and less educated, when compared to a large health insurance database where their population is more middle-class, employed, and better educated. Carson, et al. (2000), postulated that this seeming deficiency in generalizability can be a problem in descriptive studies but less of one in more analytical researc h. Their rationale is that in analytical studies the researcher is comparing treatment and control groups taken from the same

PAGE 41

41 population and so are subject to the same biases and unknown confounders(50). Because of this, comparing incidence rates between different he alth care databases can be problematic but relative statistics such as the risk ratio or odds ratio are generally considered more generalizable (50). Measurement Validity Data vs. Source information consistency The agreement between the source information and the computer data that it came from, is the benchmark for gauging measurement validity. Th is analysis rests on the assumption that the source information is accurate and complete. Th e results of this comparison of data external from the database to data within the database should be in good agreement with each other. Many evaluations of health care databases using this method have shown that sociodemographic data, names of dispensed drugs, and other non-clinical record s are very consistent with what is in the medical r ecord (50-57). Agreement of >90% should be expected. Clinical data, on the other hand, can be more erratic. Co ncordance has ranged from 58% to near 100% in assessments of diagnosis and surgical procedures in several large admi nistrative health care databases(54, 58, 59) Overall agreement between computerized health data and the records from which they were derived should be good, but just because they are broadly consistent does not mean individual diagnosis codes can not vary markedly(60-62). Other external information besides the patient record can also be used to examine external consistency. Patient interviews have been used in several assessments as the source information. Like its use in other areas, patient recall bias can cause substantial discre pancies and result in und ependable agreement(59). Internal data consistency When an external source of information, such as the patient record, is not available, internal validation methods have been used. Th is method of evaluating the internal consistency

PAGE 42

42 of a database entails cross-comparing informa tion from different fields to identify any inconsistencies. This is done in a cross-sec tional way or by comparing entries over time in a longitudinal fashion. This approach to validati on has been utilized e ffectively in research concerning diagnosis verification, disease incidence, and adverse drug event appraisals(63-66). An illustration of this approach to validate the diagnosis code of Type II Diabetes may be the presence of an elevated FBG or HbA1c labor atory result. Likewi se, the existence of hypoglycemic medication prescriptions in the pa tients pharmacy profile would reinforce the Diabetes diagnosis entry. These three para meters, diagnosis code, laboratory result, and pharmacy entry, are all in different subsystems of the same larger health care database and taken together reflect the consistency of the entered data. Pharmacoepidemiologic studies using this method have shown an increase in the predictive value of case ascertainment. In their study, Gerstman et al. (1990) demonstrated an increase from 42% to 65% for probable deep venous thromboembolism and from 70% to 97% for possible deep venous thromboembolism in the predic tive value of case ascertainment. They did this by requiring evidence of outpatient an ticoagulant use within six months of hospitalization(67). Internal descriptive research done on the Medicaid databases has proved useful in assessing validity. Hennessy et al. (2002) looked at four categories of pot ential data errors in datasets provided by the Computerized On -line Medical Pharmaceutical Analysis and Surveillance System (COMPASS) from six Me dicaid states and cove ring a defined time period(68). COMPASS is a Medica id program data vendor that pr ovides software and services related to drug utilization reviews (DURs) within the Medicaid program. The categories included: incomplete claims for certain time pe riods; absence of an accurate indicator of

PAGE 43

43 inpatient hospitalizations; missing hospitalizatio ns for those aged 65 years and older; and diagnostic codes in demographic groups in which those conditions should be rare. By reviewing the data over time the researchers were able to see if there were missing blocks or gaps in claims data. Next, they compared clai ms for inpatient hospitalizations in a defined time-period with the number reported by the US Cent ers for Medicare & Medicaid Services (CMS). Thirdly, they looked at the consistenc y of the data sets over time by exam ining the trend of the number of hospitalizations per en rollee stratified by age group. Specifically, they hypothesized that hospitalizations should increase amo ng adults as their age increases. An evaluation of this trend as adults reached the age of 65 and beyond was used as an indicator of consistency of claims in this age group. Lastly, an explor ation of the overall validity of diagnosis and demographic data was done by identifying disorders that were expect ed to be found predominantly in a particular demographic group. They specified female speci fic disorders in females, complications of childbirth and pregnancy, and l ung cancer in those aged 40 year s and older. Their hypothesis was that the number of disease-demographi c matches would far exceed the number of mismatches. The presentation of this analysis shows the quality of the data using longitudinal graphs to depict missing data and trends in the data using stratifi cation by age group. The accuracy of the diagnosis can be judged by disp laying the code along with patient demographics over time and comparing to expected results (68). The authors cal led this kind of internal data assessment macro-level validity. Administrative health care data bases are often made up of separately maintained data files that are linkable to an individua l patient file. This allows th e examination of consistency of diagnostic information, which is especially importa nt in a cohort studies that seek to identify specific medical disorders. Preferable there s hould be a high level of concordance between the

PAGE 44

44 different data files within a data base since the source information should be the same. Even if the source data is unavailable, by using these cr oss-comparison and longitudi nal techniques, it is still possible to describe the internal validity of large administrative he alth care databases. Table 2-1. Tendonopathy risk factors Extrinsic Intrinsic Systemic Disease Occupation Age Inherited disorders Sports Vascular perfusion Endocrine disorders Physical load Nutrition Metabolic disorders Training errors Anatomical variants Rheumatological diseases Shoes and equipment Joint laxity Medication Environment (temp, surface) Muscle weakness Quinolones Gender Race/ethnicity Body weight Systemic Disease Medication

PAGE 45

45 Table 2-2. Common qui nolones grouped by PEG@ classifications Generation Generic name Spectrum/use First norfloxacin Mainly Gram-negative / urinary tract infections Second ciprofloxacin Gram-negative, moderate Gram pos itive / broad oral and systemic use Third levofloxacin Improved Gram-positive, atypical / respiratory tract and many others Fourth moxifloxacin Improved Gram-positive and atypical plus anaerobic / respiratory tract and many others @Paul-Ehrlich-Society of Chemotherapy(69) Table 2-3. Quinolone toxicities le ading to market withdrawal or decreased clinical significance Quinolone Reason for Discontinued De velopment or Market Withdrawal enoxacin inhibition of cytochrome p450 pefloxacin phototoxicity, tendonapathies, etc. fleroxacin phototoxicity, CNS effects sitafloxacin phototoxicity temafloxacin hemolytic uremic syndrome lomefloxacin phototoxicity sparfloxacin phototoxici ty, QT prolongation tosufloxacin thrombocytopenia, nephritis trovafloxacin hepatotoxicity, CNS effects grepafloxacin QT prolongation, arrhythmia, nausea clinafloxacin phototoxicity, i nhibition of cytochrome p450 gatifloxacin hypoglycemia

PAGE 46

46 Table 2-4. Clinical trial quinolone adverse reactions Type Incidence Range (%) Gastrointestinal 0.8-6.8 Central Nervous System 0.9-4.7 Serious reactions <0.5 Skin/hypersensitivity 0.4-2.1 Phototoxicity 0.5-2.0 Cardiovascular 0.5-2.0 Renal 0.5-4.5 Hematological 0.5-5.3 Musculosketetal/rheumatological 0.5-2.0 Cumulative incidences 4.4-20 Modified from Christ and Esch (1994)

PAGE 47

47 CHAPTER 3 METHODS This study utilizes a retros pective cohort design employing one study drug group and one comparator drug group. Statistical analysis includes the estima tion of general epidemiologic measures of association between the two groups expressed as re lative hazards. The proportional hazards (Cox regression) model is used to evalua te the overall relative hazar d and time at risk of tendon ruptures from quinolone exposure in U.S. active duty military personnel, compared to a cephalosporin exposed group(70, 71). Exposure, outcome, and demographic informati on were abstracted from the relevant data fields in the eligibility, pharmacy, and enc ounter databases containe d under the M2 query system. Exposure to the quinolone class of antibiotics was selected because of its demonstrated toxicity to connective tissues a nd specifically, its propensity to weaken tendons. Exposure to the cephalosporin class of antibiotics was selected as a comparator group due to its similarity of clinical use to quinolones and because of the absence of any documented connective tissue toxicity. By using this design, the risk of tendon rupture from quinolones was estimated while the distribution of unmeasured confounding factors associated with the use of an antibiotic drug was roughly equal in each group. Lastly, the survival analysis model was used for two main reasons. The first rationale relates to the effect of the primary predictor va riable on the shape of th e hazard function. Next, it allows determination of which combination of explanatory variables effect the shape of the hazard function. This elucidates the effect that exposure has on the hazard of tendon rupture, as well as the effects of other co -variables. Secondly, the modeli ng of the hazard function provides the estimation of the instantaneous risk for a noti onal individual. This produces an estimated risk of tendon rupture after starting a quinolone, which is a function of the explanatory variables in

PAGE 48

48 the model. This process will also shed light on the induction time where the hazard of tendon rupture continues onward for an unknown time after discontinuance of the drug. Population Description U.S. Military Demographics This information describes members in the military community for fiscal year 2003 and before. Active Duty service branches include the Department of Defenses (DoDs) Army, Navy, Marine Corps, and Air Force. Overview The total number of military personnel is over 3.2 million strong, including Active Duty military personnel (1,419,061); Department of Ho meland Security (DHS) Active Duty Coast Guard members (38,389), DoD Ready Reserve and DHS Coast Guard Reserve members (1,167,101); and DoD appropriated-fund civilia n personnel (650,714). DoD and DHS Coast Guard Active Duty members comprise the largest portion of the military force (44.5%), supplemented by Ready Reserve members ( 35.6%) and DoD civilian personnel (19.9%). Active Duty Branches: The Army has the largest number of Active Duty members (493,563) followed by the Navy (376,970), the Air Force (370,945), and the Marine Corps (177,583). There are also 38,389 Active Duty members in the DHSs Coas t Guard. The total 1,419,061 DoD Active Duty service contingent in 2003 is 30.1% smaller than it was in 1993, when there were 2,029,300 Active Duty members. Since 1990, the number of DoD Active Duty service members has declined by 9.6% Marine Corps; 30.1% Air Force; 32.2% Army, and 34.3% Navy. Ranks: The Active Duty force has one officer for every 5.2 enlisted personnel. In order, the Air Force has one officer for every 4.0 enlisted personnel, Army has one for every 5.2, the Navy has one for every 5.9, and the Marine Corps has one for every 8.5.

PAGE 49

49 Women: Women comprise 213,059, or 15% of the Do D Active Duty force. The percent of women in the Active Duty popul ation has continued to grow si nce 1990. Is has gone from 11.5% officers and 10.9% enlisted to 15.3% officers and 15% enlisted in 2003. Minorities: Over one-third (35.8%), or 507,418, of Active Duty members identify themselves as a minority. Classifications incl ude African American, Hispanic American, Native American, Alaskan Native, Asian Ameri can, Pacific Islander, or multi-racial. Location: While the Active Duty population is spread throughout the world, it can be divided into three basic areas. Active Duty members are assigned to the United States and its territories (84.5%), Europe (8.0%), and East Asia (6.6%). The states with the largest Active Duty populations are California (166, 397), Virginia (140,575), and Texas (116,638). Age: Close to half of the Active Duty force is 25 years old or younger (47.4%). The next largest group is 26 to 30 (18.1%), followed by 31 to 35 (13.8%), 36 to 40 (12.3%) and 41 and above (8.4%). The average age of the Active Duty force is 28.2. The average age for officers is 34.5 and 27.0 for enlisted personnel. Education: 86.1% of officers have a Bachelors degree or higher while only 3.7% of enlisted members do. Most (94.0%) enlisted memb ers have a high school diploma and/or some college experience. Marital status: Over half (52.3%) of Active Duty members are married. A majority (68.8%) of officers are married while slightly less then half (49 .2%) of enlisted personnel are married. In addition, 6.7% of the DoD Active Duty force is in dual-military marriages. Dependents: There are fewer Active Duty member s (1,419,061) than th eir associated family members (1,924,174). Just over one-third (37.3%) of Active Duty members are married with children while 6.1% are single parents.

PAGE 50

50 Patient Selection Inclusion Criteria Information on study measures was abstracted from the M2 database for patients 18 years and older and serving on active duty in the U.S. military at the time of receiving medical care. All visits for outpatient services from June 1, 2005 to May 31, 2006 were reviewed for study inclusion. Study patients were identified in th e Pharmacy Detail Transaction Service (PDTS) data file by the annotation of a quinolone antibiotic in the medicati on brand name field and/or the generic name field. Identification also incl uded The American Hospital Formulary Service (AHFS) classification number field for quinolone antibiotics and/or AHFS therapeutic class description field. Lastly, the Ge neric Code Number (GCN) field, wh ich is specific to the generic ingredient, dosage form, and dr ug strength, was used for patient recognition. The GCN is the same regardless of manufactures and/or package si ze. Identified patients were followed forward from the day the quinolone pres cription was dispensed (day 0) to day 60. Quinolone exposed patients had their records reviewed sixty days pre-drug exposure and sixt y days past the 60-day study period cutoff. Review identified any pr evious quinolone exposure or co-morbid disease states related to connective tissue disorders. The cephalosporin group in this study cons isted of patients fitting the same inclusion criteria as above. The singular difference is th at this group wass selected based on their exposure to any cephalosporin antibiotic medication. Cephalosporin exposure was chosen as the comparator group because it is an antibiotic class with similar treatment uses as quinolones. Cephalosporins have not been implicated in a ny connective tissue toxi cities in animals or humans. There are no cases in the literature of cephalosporin-induced tendonopathies or tendon ruptures. It is intended that by choosing to compare patients that have been exposed to quinolones to patients exposed to cephalosporins that unknown or unmeasured differences in our

PAGE 51

51 sample patients will be similar in both the study and control group thus reducing biases inherent in a nonrandomized study. Cephalosporin exposed patients had their records reviewed 60-days pre-drug exposure and 60-days past the 60-day st udy cutoff. Review dentified any previous quinolone exposure or co-morbid disease states related to connectiv e tissue disorders. Exclusion Criteria Patients possessing the E code in the injury and poisoning category of the International Classification of Disease, Ninth Revision, Clin ical Modification (ICD-9-CM) defined disease codes, signifying an external cause of the injury (ICD-9-E800-E999) were excluded from selection into any of the study groups. There are diseases and health states that ha ve been shown to, possibly, predispose patients to connective tissue di sorders. After initial pa tient selection, both study groups were screened to allow identification and exclusion of patients pos sessing any of these conditions. Exclusion of patients with arthropathies and related di sorders (ICD-9-710-719), ne oplasms (ICD-9-140-239), presence of HIV (ICD-9-042), alcohol depende ncy (ICD-9-303), drug abuse (ICD-9-304), organ transplant (ICD-9-V42), hemodialysis (ICD-9V42), diabetes mellitus (ICD-9-250) to reduce confounding the outcome. As expected the occurrenc e of these conditions were low or absent, as all but alcohol dependency is reason for a change from active to inactive service. This study will utilized the c oncept of a New User design posited by Ray and associates (72, 73) Data Source The Executive Information and Decisi on Support (EIDS) Program Office EIDS is the centralized data store for the Mil itary Health System (MHS). EIDS collects, processes, and manages nearly 100 terabytes of m ilitary health data thro ugh a powerful suite of decision-support tools that enable effective ma nagement of MHS health care operations. EIDS

PAGE 52

52 has active interfaces around the globe in order to manage the receipt, processing, and storage of billions of health care re cords that characterize MHS operations and performance. The Military Health System Management Analysis and Reporting Tool (M2) The M2 is a powerful ad-hoc query tool used to obtain summary and detailed views of population, clinical, and fina ncial data from all MHS regions. M2 includes Military Treatment Facility (MTF) and commercial network claims data integrated with eligibility and enrollment data. This integrated data enha nces support to healthcare managers across the MHS. M2 allows users to perform trend analyses, conduct pati ent and provider profili ng studies, and conduct business case analyses to maximize health plan efficiency. In a recent pilot information pull, taken from the M2 system for the FY 2005, there were 111-million prescriptions for the entire DoD population, 10-million prescriptions for activ e duty military, 1.3-million prescriptions for quinolones, and 0.9-million for cephalosporins. Data Validity The M2 databases contain eligibility, diagnos is, dispensing, and administrative information for the Military Health Service (MHS). The repo sitory presently includes only outpatient data for both institutional (military hea lth facility) and non-in stitutional (commercial network) health care providers. The goal of the M2 is to provide a system that is easily queried to enhance the decision making process of the DoD health care managers. It has not been validated in a meaningful way through a published peer-rev iewed validation study, as such, this study undertooke a limited validation study to build conf idence in the use of the M2 for research purposes. Since the M2 has global coverage a nd researchers are not gi ven access to individual patient identification information, the ability to validate M2 through comparison with external data i.e the patient medical record, is not possi ble. Due to this, the present study performed an internal consistency validation process along the sa me lines as the Medicaid descriptive analysis

PAGE 53

53 performed by Hennessy, 2003(68). This process u tilized the three diffe rent databases, or Universes in Microsoft Object terminology, to perform cross-checks of pertinent data fields. Database Characteristics The M2 is the query module for the MHS medical data repository. Data contained in the repository comes from clinical databases used in military treatment facilities worldwide and from commercial health care facilities that are part of the MHS network. The M2 links eligibility, demographic, diagnosis, and prescription data, th rough the patient specific military eligibility identification number. Because of this, an individual patients eligibility and medical information can be retrieved from each of the th ree universes and combined in a common report. Internal Validity Macro-level data quality assessment was used in this study to judge the consistency and validity of data in the absence of external records for comparison(68). Incomplete Data Descriptive statistics were used to depict the magnitude of missing data in several important fields across all thr ee universes. The following variables were examined for missing data: 1) patient demographics (gender, age, race) in the eligibility, clinical care, and pharmacy universes, 2) military status ( active or dependent), 3) diagnosis codes (ICD-9-CM), 4) drug quantity, 5) days of supply 6) military occupation specialty code. Missing data was handled in the analytic phase by exclusion a nd description if necessary. Pa tients with exposure information but missing diagnosis data were necessarily excluded from this study. Longitudinal Data Integrity In order to examine the data over time the to tal number of prescriptions in the MHS was abstracted from M2 and graphe d, by month, over a five-year peri od (FY 2002-06). This gave a gross indication of the consistency of M2 over that period.

PAGE 54

54 Diagnosis vs. Demographics One disease diagnose was selected and crosschecked with demographic information that one would commonly expect to occur together For the fiscal years 2002-06 the percent coherence between normal pregnancy deliver y in females (ICD-9-CM 650.00) and benign prostate hyperplasia ICD-9-CM 600. 00 600.91), was generated and plotted. Diagnosis and Drugs One disease state was cross-checked versus the medication usually prescribed to treat it. Coherence levels were generated yearly for fiscal years 2002-06 for the following ICD-9-CM drug combination: Type I diabetes (ICD-9250.01 and any insulin product on the DoD uniform formulary (UF). Drugs and Diagnosis The algorithm from the previous paragraph was reversed for the examination of drug data. All patients with a prescription for insulin in M2 during the FYs 200206 were cross-checked with the ICD-9-CM code 250.00 for any type diabetes. All cross-checks generated the percent matche d or coherence of the information between the different data fields. In the study by Hennessy et.al, 2003, the data was considered valid if the percent coherence was equal or greater than 80 %(68). This study utilized this rule in judging the validity of the M2 information system. The Variables Dependant Variable Tendon Rupture: The outcome variable of interest is the occurrence of a tendon rupture in the study group at a specified time (t) from drug exposure until the end of a sixty-day study interval. All patients are censore d at 60 days. The occurrence wa s expressed in a binary fashion by the presence or absence of the appropriate ICD9-CM entry in the patients M2 computerized

PAGE 55

55 record. The ICD-9-CM codes of interest are; 727.6 rupture of tendon, non-traumatic, 727.60 non-traumatic rupture of unspecified tendon, 727. 61 rupture of rotator cuff, 727.62 tendon of biceps, 727.63 extensor tendons of hand and wr ist, 727.64 flexor tendons of hand and wrist, 727.65 quadriceps tendon, 727.66 patellar tendon 727.67 Achilles tendon, 727.68 other tendons of foot and ankle, 727.69 other tendon ruptures. Predictor Variable Treatment: Exposure to a quinolone medication in the study group, or exposure to a cephalosporin medication in the comparator group is the main explanatory variable of interest. The quinolone group exposure status, is the recorded dispensing of a valid prescription for an oral quinolone antibacterial medication in the ou tpatient PDTS sub-database of M2. Exposure time in this cohort was calculated in person-days. Qu inolone antibiotics used in exposure data include: moxifloxacin, ciprofloxacin, gemifloxa cin, levofloxacin, norfloxacin, and gatifloxacin. The comparator group exposure was the recorded di spensing of a valid prescription of an oral cephalosporin antibacterial medication in the ou tpatient PDTS sub-database of M2. Exposure time was calculated in person-days starting at the date of a valid prescription dispensing. Time at risk continues for 60-days post prescrip tion dispensing for quinolone and cephalosporin groups. Cephalosporin medications used in e xposure data include: cefazolin,cefuroxime, ceftazidime, cefepime, cefdinir, ceftr iaxone, cefditoren, and cefixime proxetil. Additional Covariates: Gender : Male or female Race: Expressed as White, Black, Other Military grade : Categorized as officer or enlisted Military occupation specialty (MOS): Taken from DoD operational MOS codes and collapsed into two main groups; 1) Direct 2) support

PAGE 56

56 Age: Years at time medication was dispensed Concomitant corticosteroid: outpatient use, grouped as yes or no Days of mediation supply: calculated locally at the time of dispensing = #of pills / daily direction for use Service: Branch of military, Army, Na vy, Air Force, Marine Corps Provider specialty: Specialty of the provider that wrot e the prescription for the treatment medication divided into prim ary care and specialty care. Medication diagnosis: Recorded diagnosis leading to antib iotic use. Categories are; 1) upper respiratory infection 2) ga strointestinal/genitour inary and 3) soft tissue/bone infections. Analysis Descriptive Statistics Demographic and related explanatory variable s expressed in a continuous fashion were analyzed via means and their standard deviations while categorical variables are presented as proportions. General Measures of Association Due to the cohort study design, only probabilitie s that condition on expos ure status can be estimated. Conditional probabilities for tendon rupture that is categor ized as yes or no and occur within a sixty-day period, given exposure (P^(TR | E)) and no n-exposure (P^(TR | )) were used to calculate unadjusted Relative Rate Ratio (RR), Excess Risk (ER), and A ttributable Risk (AR) in quinolone versus cephalosporins groups. Using the conditional baseline probabilities from Table 3-2 the following unadjusted measures of association may be estimated. .Estimated Relative Risk (RR^) = ) /( ) /( d c c b a a (3-1)

PAGE 57

57 Excess Risk (ER^) = d c c b a a (3-2) Attributable Fraction (AFe^)) = RR RR 1 (3-3) Hazard Function When either the population at risk or the in cidence rate changes substantially over the study time-period it becomes necessary to stratify time into shorter intervals to adequately capture this effect. The incidence proportion wh ich is a cumulative measure used in closed cohorts and the more flexible incidence rate which is an average rate over the risk time interval and can be used in both closed or open cohorts are both summary measures for the entire time interval and thus do not explain within-interval risk changes. If a study has a large population with plentiful outcomes, it is possible to measure the outcomes over smaller and smaller time intervals. A plot of the incidence rates over th e associated interval on which the incidence rate was based reveals a graph of the changes in th e incidence rate over time. By reducing the intervals to their hypothetical li mit it yields th e hazard function, h(t) The hazard function is interpreted as the instantaneous incidence rate. The hazard function can be linked mathematically to the cumulative incidence proportion through the following equation(74).. )) ( 1 ( ) ( / ) ( t I t d t dI t h (3-4) Here the total interval of inte rest is [0,T] where T equals 60days in this study. The link between the hazard function at h (t) for 0 t T and the incidence proportion I( t), over the interval [0, t ] is demonstrated. This connection assumes th at an incident case is no longer at risk after incurring the outcome of interest and that th is is the only way that an individual can cease to be at risk. In the above equation d I( t ) / d ( t ) represents the slope of I( t ) at time t The

PAGE 58

58 denominator (1-I( t )), accounts for the proportion of the population still at risk at time t The benefit of using the hazard functi on is its ability to extract dynamic information from a plot of the function as compared to other methods th at measure association over preconceived time intervals assuming a linear ri sk over time. Whereas I( t ) measures the cumulative incidence from time = 0 to time = t h ( t ) measures the incidence rate at exactly time t Relative Hazard As has been previously describe d the exposure variable of interest in this study is binary in nature. Thus in each sample there are two in cidence proportions of interest. These are IE( t ) and I( t ) reflecting the interval [0, t] within [0,T], for the exposed and unexposed respectively. This is also true for the hazard func tions for the two groups e.g. ( hE( t ) and h( t )). Since the RR and Odds Ratio (OR) are measures of relative di fference in incidence between the exposed and unexposed, they can be written as: RR( t ) OR( t ) ) ( ) ( t I t IE E (3-5) The hazard function can also be used to m easure the relationship between exposure and incidence by: RH( t ) = ) ( ) ( t h t hE E (3-6) RH( t) stands for the Relative Hazard at time t Using the hazard function, we can produce a relative hazard that provides an instantaneous measure of how exposure to the study medication affects the incidence of tendon ruptures. The RH( t ) can vary widely at different t s within [0,T], so that at certain times RH( t ) can be larger than one, reflecting an increasing risk, or below one, signi fying a decreasing risk. It is

PAGE 59

59 these swings in risk that are captured by utilizi ng the hazard function and masked when using the RR or O R If the assumption is made that the relativ e impact of exposure on the instantaneous incidence rate remains constant over the interval [0,T], then the hazard functions in the exposed and unexposed will be proportional to each other. Thus the RH( t ) will remain constant throughout the [0,T] interval and RH( t ) = RH. This assumption is called the proportional hazards model which allows comparison of the RH( t ), RR( t ), and OR( t ). Importantly in this study, since the incidence of tendon rupture at baseline over [0,T] is likely small then RR( t ) OR( t ) RH( t ) RH. If the proportional hazard assump tion is valid in this study then there should be little difference between RH, RR, or OR as a measure of the relative difference in tendon rupture between those exposed to quinolone therapy and those exposed to cephalosporins or in the unexposed control group. Figure 3 uses notional data with a constant RH of two and a lo w incidence rate to show the relationship between the three meas ures of association over time. This study utilized the hazard function to gra ph the change in incidence over time after exposure to the study medication. This allows id entification of the time, post therapy initiation, of the greatest risk of tendon rupt ure. This also allows the calc ulation of the RH as the main measure of relative association. Modeling the Hazard Function In the current study, we are modeling two distinct groups or subpopulations. The two groups have exposure (X) at tw o different levels (X = x0, X = x1,) corresponding to x0 = cephalosporin exposure, x1 = quinolone exposure. The dependent variable is the hazard of tendon rupture at time t for each level of exposure. The simplest model includes only two groups and has the form:

PAGE 60

60 ) | ( ) | (0 1x t ch x t h (3-7) The hazard of tendon rupture when exposed to quinolones at time t equals the hazard of tendon rupture with no exposure multiplied by some constant c This model assumes that the hazard at time t for both groups is proportional. The constant c is the previously discussed relative hazard (RH). If RH is less than one then the hazard of tendon rupture at time t is less in the quinolone group than in the cephalosporin group. Conversely, if the RH is greater than one the hazard of tendon rupture at time t is greater in the quinolo ne group compared to the cephalosporin group. A more generalized form of equation 3-7 can be written to include other relevant covariates. This allows the simplest twogroup proportional hazard model, which allows the addition of other independent variables that affect the shape of the hazard function: ) exp( ) ( ) / (0x t h X t h (3-8) The exponentiated component of this equati on is a linear combination of explanatory variables and can be interpreted as the risk score for the individual repr esented by the model. The risk score contains variates and factors. Variates are stated in numerical form and are usually measured on the continuous scale, e.g. ag e. Factors are represen ted by a limited set of values called levels e.g. gender. The advantages of utilizing the proportional haza rds model in this study are threefold: 1) It allows estimation of the ba seline hazard function, ho(t), (all explanatory variables = 0). 2) It allows estimation of Relative Hazard coefficients, which are similar to the Relative Risk/Rate ratios, or Odds Ratio, in measuring association between exposure and outcome. 3) Provides an insight into how the hazard changes over time in any exposure group during the time interval.

PAGE 61

61 Time-Dependent Exposure The need to stratify by time when using logi stic regression highlights the main advantage of using the proportional hazard model. The C ox regression model assumes proportional hazards and controls for the effect of time at risk, before considering the associat ion between risk factors and outcome(70, 71). In order for the results from logistic regres sion and Cox regression to differ it is necessary for 1) time at risk be associated with the outcome a nd 2) time at risk be associated with exposure. In simpler terms, this means that the risk is ch anging over time. In the present study, as in most epidemiologic studies, the lengt h of exposure is assumed to ef fect the risk of the outcome occurring. The use of the proportional hazards model in this study rests on the time at risk effecting exposure. Since the exposure levels will change during the follow-up period as a person first takes the medication then stops taki ng it, often after a 7-10 day period, but the individual will remain at risk as the drug levels decline in the body and onward to some, as yet, unspecified time. Since exposure levels do change in this study, then by definition there will be association between time at risk and exposure, thus time at risk confounding will occur. Variables whose values change over time are know n as time-dependent variables. In this study, the exposure variable will change over time, but because of the inability to specifically track this change, through blood le vels etc., the use of a time-va rying coefficient will be used. In the proportional hazards model the time*exposure variable is assumed to be constant. This allows the interpretation of this coefficient as the log-hazard ratio where it is constant over time. In this instance, if this ratio is in fact a function of time then the coefficient of the exposure variable is called a time-varying coefficient. In other words, this model allows the hazard to change over time in a way that is not proportional to the baselin e group. Time is affecting exposure and outcome in the quinolone group di fferently than in the cephalosporin group.

PAGE 62

62 This assertion of a time-varying coefficien t results in a non-proportional hazard model and can be difficult to fit. This difficulty can be overcome by 1) specify ing a unique distribution chosen to model the variable; 2) stratification by the identified variable or 3) use piece-wise regression. This study first stratified by the chosen variab le in order to graph the resulting hazard function. If the variable was found to vary w ith time, by way of the tim e*variable interaction term being significant, then the piece-wise regression technique was employed to model the variable. This technique keeps the proportional hazard assumption intact and results in more accurate risk estimates (hazard ratios) being pr oduced from the different regression equations that model the different pieces of the study interval. The resulting piece-wise regression models was tested for selection using the Akaike ;s information criterion (AIC) on the basis of there resulting -2logL test statistic. The Cox Proportional Hazards Regression Mode ls utilized for this study were the following: Study model 1 Cox regression equation for demographic variables hi(t) = h0(t) exp( 2x2 + 3x3 + 4x4) (3-9) Study model 2 Cox regression equation for 60-day interval hi(t) = h0(t) exp( 1x1 + 2x2 +.+ 11x11) (3-10) Study model 3 Cox regression equati on with treatment*time variable hi(t) = h0(t) exp( 1x1 + *x1(t) +.+ 11x11) (3-11) Study model 4 Cox regression equa tion with three risk windows hi(t) = h0(t) exp( Ax1 (1) + Bx1 (2) + Cx1 (3) 2x2 +.+ 11x11) (3-12) Equation components are defined below:

PAGE 63

63 h(t): Hazard of tendon rupture at time t conditioned on quinolone e xposure and controlling for covariates. X: Treatment variable. ho(t): Baseline hazard at time ( t) x1: Days 1-60, treatment, a factor with two levels. Cephalosporin=0, quinolone=1. x2: Age, a continuous variable. x3: Gender, a factor with two levels. Male=0, female=1. Race, a factor with 3 levels. x4: White=0, Black=1. x5: White =0, Other=2. Military Occupation Specialty (MOS), a factor with 2 levels. x6: Support = 0, Direct = 1. Grade, a factor with two levels, x7: Officer=0, Enlisted=1. x8: Days of Supply, a continuous variable. Provider Specialty, a factor with two levels. x9: Primary care (PC)=0, Secondary care (SC)=1. Oral concomitant corticosteroid use, a factor with two levels. x10: no=0, yes=1. Diagnosis leading to antibiotic treat ment, a factor with three levels. x11; Upper respiratory infection (URI) = 0, Ga strointestinal/Genitourinary (GI/GU) = 1, Soft tissue/Bone infection (ST/BI) = 3. Branch of Service, a factor with four levels.

PAGE 64

64 x12; Army = 0, Air Force (AF) = 1, Navy = 2, Marine Corps = 3. x1(t) = interaction term consisting of the treatment variable and time (60 days). x1 (1) = Days 1-25, treatment, a factor with two levels. Cephalosporin=0, quinolone=1. x1 (2) = Days 26-35, treatment, a factor with two levels. Cephalosporin=0, quinolone=1. x1 (3) = Days 36-60, treatment, a factor with two levels. Cephalosporin=0, quinolone=1. Power General For general epidemiologic pow er calculations the incide nce rates of tendon rupture calculated in a recent incidence study of active m ilitary were used (3). The rates ranged from 30.9 per 100,000 person-years for Achilles tendo n ruptures to 120 per 100,000 person-years for all site tendon ruptures (3). Th e probability of making a type I error (alpha) equals 0.05 while the probability of making a type II error (beta) equals 0.20 (power equals 0.80). The estimated incidence in the quinolone group equals 0.004, an d the estimated the effect size taken from a cohort of otherwise healthy Dutc h subjects is RR=1.9 (35). Th e minimum required sample size is estimated at 22,520 person-years of follow-up time. The maximum required sample size is estimated at 52,135 person-years of follow-up time. The sample size calculation was performed using Epi Info(TM), database and statistics so ftware for public health professionals, 4/26/2004 (75). Secondary objectives involving analysis of subgroup populations, such as race or military occupation specialties, may or may not have su fficient power to adequately address their hypotheses. This is a retrospective study, the nu mber of cases is static, as result, power was determined in a post-hoc fashion.

PAGE 65

65 Survival Analysis In survival analysis, it is the number of events that is most important. Accordingly, the first step in sample size calculati on utilizing a survival analysis analytic method is to calculate the number of events that must be observed. ez z d2 2 2 /) ( 4 (3-13) d=required number of events, = risk estimate (effect size) Using a conservative risk estimate of RR=1. 5 (Dutch cohort was 1.9(19)), and alpha = 0.05, beta = .20 (power equals .80), th e number of events required, d= 79. A simplified equation for calculating the required number of subject to ensure the 79 cases are identified is the following. ) ( event P d n (3-14) d = number of events from Equation 3-7. Using a conservative estimate of P(event) from a recent study on Achilles tendon ruptures only, events is 0.309%, gives a maximum sa mple size requirement of 25,566, which is commiserate with the minimum sample size calcul ated based on a general measure of association (3). These calculations are based on having an 80% chance of rejecting th e null hypothesis if in fact it is false, e.g. having a power probability of 0.80 in de tecting a statistically significant difference in outcome between groups if one truly exists. Objectives, Research Questions, and Hypothesis All hypothesis use = 0.05

PAGE 66

66 Main Research Question Estimate the risk of tendon rupture from using a quinolone antibacterial medication relative to a cephalosporin antibacterial medication, in a military population Objective 1 Estimate the risk of tendon rupture from qui nolone use, relative to a cephalosporin medication, while adjusting for relevant ri sk factors and accounting for time at risk. Research Question 1 Is there a difference in risk of tendon ruptur e in active duty U.S. military personnel, from quinolone antibiotic use relative to cephalosporin an tibiotic use, when adjusting for age, gender, race, military occupation, grade, provider specialt y, days of medication supply, diagnosis related to antibiotic dispensing, branch of servi ce, oral steroid status, and time at risk? Research Question 1 Hypothesis The null hypothesis is the regr ession coefficient in the quino lone exposed group relative to the cephalosporin exposed group will equal zero, after adjusting for age, gender, race, military occupation, grade, provider spec ialty, days of medication supply diagnosis relate d to antibiotic dispensing, branch of service, or al steroid status, and time at risk The alternative hypothesis is the regression coefficient is different from zero. Ho: 1= 0 Ha: 1 0 Note: 1 = treatment variable from equation 3-10. Objective 2 Investigate whether the risk of tendon rupture from quinolone use varies over the 60-day study interval.

PAGE 67

67 Research Question 2 Does the adjusted hazard function for each antibiotic treatment, support the proportional hazards model assumption? Research Question 2 Hypothesis The null hypothesis states that the interacti on regression coefficient, created from the treatment and time variables, is not different from zero during the 60-day post exposure period, after adjusting for treatment, age, gender, race, military occupation, grade, provider specialty, days of medication supply, diagnos is related to antibiotic dispensi ng, branch of service, oral steroid status, and time at risk. The alternative hypothesis states that the adjusted interaction regression coefficient is different from zero over the 60-day period. Ho: = 0 Ha: 0 Note: 1 = treatment variable from equation 3-11. Objective 3 What is the average induction period fr om quinolone exposure to tendon rupture? Research Question 3 Does model 4 (piece-wise model) fit better than Model 2 or 3? If so, at what interval, in days, is the hazard ratio of tendon ruptur e post-quinolone therapy at its highest? Research Question 3 Hypothesis The null hypothesis is that all of the time inte rval regression coeffici ents in the quinolone exposed group relative to the cephalosporin ex posed group will equal zero after adjusting for age, gender, race, military occupation, grade, pr ovider specialty, days of medication supply, diagnosis related to antibiotic dispen sing, branch of service, oral st eroid status, and time at risk.

PAGE 68

68 The alternative hypothesis is at least one of the time intervals regression coefficient is different from zero. Ho: A = B = C = 0 Ha: at least one 0 Objective 4 Identify other major risk factors that affect an active duty military individuals estimated risk for tendon rupture. Research Question 4 What are the significant risk factors, other than quinolone treatment, for tendon rupture relevant to active duty military personnel? Research Question 4 Hypothesis The null hypothesis states that the age, gender, race, military occupation, grade, provider specialty, days of medication supply, diagnosis related to antibiotic di spensing, branch of service, and oral steroid status regression coefficients are eq ual to zero. The alternative hypothesis is that one or more covariate regr ession coefficients are not equal to zero. Ho: 2 = 3 = = 11 = 0 Ha: at least one 0 This study has been approved by the Universi ty of Florida Institutional Review Board #303-06, and the Brooke Army Medical Cent er (BAMC) Institutional Review Board C.2007.088d.

PAGE 69

69 Figure 3 1.650,714 1,167,101 38,389 1,419,061 DoD Civilian Ready Reserve DoD Active Duty Coast Guard Active Duty Figure 3-1. DoD popul ation description 0 500000 1000000 1500000 2000000 2500000 1990199520002003 Army Navy Marine Air Force Total Figure 3-2. DoD ac tive duty trend 1990-2003

PAGE 70

70 a Tendon rupture; b Diagnosis Figure 3-3. Patient inclusion schematic 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 5101520 Time[years]Measures of Association RH RR OR Figure 3-4. Illustration of RH(t ), RR(t), and OR(t) where the hazard functions for both exposed and unexposed are constant. Index date (dispensing date of quinolone or cephalosporin) 60 days 60 days 60 days end date all censor .1st TRa DXb

PAGE 71

71 Table 3-1. Key active duty demographic summary Demographic Variable Active Duty Total Number 1,419,061 Ratio of officers to enlisted 1 to 5.2 Percent Women 15.0 Percent minorities 35.8 Percent located in the United States 84.5 Percent 25 years old or younger 47.4 Percent with bachelors degree or higher16.9 Percent married 52.3 Percent in dual-military marriage 6.7 Number of dependents 1,924,174 Percent with children 43.4 Percent single parents 6.1

PAGE 72

72 Table 3-2. Conditional probabilities Tendon Rupture (TR) TR No-TR Exposed a b a + b Unexposed c d c + d Medication a + c b + d n Conditional probabilities P^(TR | E) = b a a p 1 P^(TR | ) = d c c p 2

PAGE 73

73 CHAPTER 4 RESULTS Data Validity Missing Data The military occupation specialty (MOS) variab le utilized in the proportional hazards regression model had an apprec iable amount of missing data. Nearly 6.1% (6,518) of the military personnel in the study were without MO S observations out of the 107,334 total from both groups combined. There were 3,914 (6.6 %) missing in the quinolone group and 2,606 (5.4%) in the cephalosp orin sample. All 382 cases had a va lue included for the MOS variable (210 support and 172 direct). Ten variables came from a combination of th e medication and encounter databases. The medication database was almost bere ft of missing data. The reason fo r this is the finding that the Pharmacy Data Transaction Service (PDTS) medication database prompts the pharmacy dispensing the medication to complete missing data fi elds prior to moving to the next screen. In some instances, the prescription cannot be filled unless all medication an d demographic fields are completed. The MOS information came solely from the DEERs eligibility database, which resulted in the missing data. Longitudinal Validity The integrity of the data in the M2 databa se can be assessed by graphing the number of prescriptions per month and per year. This long itudinal analysis reveal s any unusual spikes or downward trends that are irregula r. These irregularities may re present blocks of missing data that constitute poor database integrity. In Figure 4-1, the normal trend during the same month in different ye ars steadily increases which signifies an increasing volume of prescripti ons over time. It also shows the increase in

PAGE 74

74 prescriptions within the DoD by year This graph presents no identifiable blocks of missing data that would impugn the longitudinal validity this database, and lead to a conclusion of reduced integrity of the data. Diagnosis versus Demographics Validity This is a particularly important evaluation as the diagnosis information comes from two separate databases representing military purchased car e, received in civilian facilities, and direct care, received on a military base. The gender information comes from another database file containing demographic information and functioni ng as a membership database. Essentially, the validity assessed here is actually calculated from the cohe rence coming from the linking of three separate databases. The coherence is very high for females and nor mal pregnancy deliveries (Table 4-1). The coherence remained stable and high over time as demonstrated by the stability of normal pregnancy deliveries over time (Figure 4-2). Both of these data presentations help reinforce the coherence of the sources within the M2 database. Coherence was high when the diagnosis of beni gn prostate hyperplasia was cross-checked with gender (Table 4-2). The coherence rema ins consistently high ove r the five-year span. Diagnosis versus Drugs Validity The coherence characterized by Figure 4-4 and Table 4-3 is formulated from the three separate databases described above. The diabetes mellitus diagnosis codes were abstracted from the purchased care and direct care clinical databases within M2, while the prescription information was gathered from the prescription de tail transaction service (PDTS) database. The coherence for the five years measured stays well above the a priori 80% threshold s tipulated in this study.

PAGE 75

75 These results are cautiously relayed in the ne xt two sections with the understanding that the reliability of these results is affected by the appropriatene ss of prescribing and diagnosis coding. These terms (insulin and type 1 diabetes) were selected to minimized the misclassification error as it is a requirement to su stain life that exogenous in sulin be administered to truly type I diabetes sufferers. Drug versus Diagnosis Validity Coherence between a prescription for insulin and a diagnosis for diabetes mellitus (any type) is above the 80% threshold for each of the five years evaluated. The greater amount of non-matching, drug versus diagnosis observations shows the difficulty for large administrative databases to have absolute cohe rence. Since all patients using insulin should have a diagnosis for some type of diabetes, this measure of inte rnal consistency and reli ability is crucial in interpreting results gained from using th is database (Figure 4-5, Table 4-4). Patient Demographics Upon the completion of the cohort selection process, 107,334 active duty patients were included in this study. The quinolone st udy group contained 59, 264 patients while the cephalosporin control group counted 48,070. In 2006, for active duty members only, there were 116,479 outpatient prescriptions dispensed for quinolones and 91,785 for cephalosporin antibiotics. There were 10,370, 568 prescriptions dispensed DoD wide for active duty members. On average, the quinolone treatment group was slightly older than the cephalosporin group (29.1.8 versus 28.7.4; p-value<.001) and th e population of the DoD as a whole (28.2 years) (Table 4-5). The quinolone group containe d a higher proportion of females than the cephalosporin group (28.3 versus 21.8; p-value<. 001) and both had considerably more than the DoD population at large (15.0) (T able 4-5). The quinolone group, while still mostly White, has a

PAGE 76

76 higher percentage of Black patients when comp ared to the cephalosporin group (18.2 versus 16.1; p-value<.001) (Table 4-5). Dependent Variable The dependent variable is the number of cases of tendon rupture from the dispensing date of the prescription for either a quinolone or cepha losporin, censored at 60-days. There were 255 cases of all-site tendon rupture in the quinolone group and 127 in the cephalosporin control group. Median time to rupture was twice as long in the quinolone group at 27 vs. 12 days. (Table 4-6) There were fewer male cases of tendon rupt ures when compared to the cephalosporin group, as well as a racial disp arity. More Whites had tendon r uptures in the quinolone group, although this difference proved to be stat istically non-significa nt (Table 4-6). Table 4-7 gives an overview of the nature of the cases of tendon rupt ures in both groups. Significant differences between cas es in the treatment groups are highlighted by the median time to rupture (27 days versus 12 days; p-value<.001) and pr ovider specialty (2 = 27.02; pvalue<.001). Other Independent Variables Military occupation was divided into direct and support categori es. There were predictably twice as many patients in the support occupations as in the direct occupations. There were missing data on this variable ranging from 6.1% in the quinolone group to 5.5% in the control group. Military rank was broken in to officer and enlisted for the purpose of these analyses. The two treatment groups had similar proportions whil e having more officers and less enlisted than the military population as a whole. A small perc entage of each group used outpatient steroids

PAGE 77

77 during the study (7.5% quinolone group versus 7.0% cephalosporin group; (p-value 0.002, Table 4-8). The only continuous measure in this group is medication days of supply. The variable averaged just over 11 days in the quinolone grou p and just below 10 days in the control group. The most frequent days of supply, was 10 days followed by 30 in each treatment group. Research Question 1 Is there a difference in risk of tendon ruptur e in active duty U.S. military personnel from quinolone antibiotic use relative to cephalosporin an tibiotic use, after adjusting for age, gender, race, military occupation, grade, provider specialt y, days of medication supply, diagnosis related to antibiotic dispensing, branch of servi ce, oral steroid status, and time at risk? Due to the significant p-value from Table 4-14 fo r the treatment regression coefficient, the null hypothesis for Research Question 1 was rejected. Ho: 1 = 0 Ha: 1 0 The adjusted hazard ratio for the occurrence of tendon ruptures within sixty days of using a quinolone antibiotic compared to a cephalosporin antibiotic is 1.65 (95% confidence interval: 1.33 2.04) (Table 4-9b). Active duty soldiers, sailors, airmen, and marines have a 65% greater likelihood of incurring a tendon rupture during th e first sixty day post quinolone exposure compared to the same period for cephalosporin use. The absolute impact of quinolones compared to cephalosporins reflected by the excess risk is an additional 166 cases of tendon ruptur es per 100,000 active duty military members. Since the excess risk includes individuals who ruptured their tendons and were not exposed to quinolones it is important to estimate the risk to individuals characte rized by the presence of

PAGE 78

78 solely quinolone use. By using the attributable risk or more pr ecisely the attributable fraction, the answer to the question of how many cases in the quinolone gr oup can be attributed to the quinolone exposure is answered. Th e attributable fraction is 0.39 in this study. Thus, 39% of the tendon ruptures, or 100 of 255 cases, in the quinolone group can be attributed to quinolone use. Stated differently, 100 cases of tendon rupture may have been avoided if they had been dispensed a cephalosporin instead of a quinolone. Research Question 2 Does the adjusted hazard function for each antibiotic treatment support the proportional hazards model assumption? The graph of the hazard function stratified by tr eatment and adjusted for model covariates illustrates the dynamic nature of the risk of te ndon rupture over the 60-day post exposure period (Figure 4-6). This graph also shows the appare nt violation of the propor tion hazards assumption, which underpins the use of the Cox regression model. In other words, it a ppears that the risk of tendon rupture is not in relative proportion over time between the two treatment groups. The appearance of a proportiona l hazard violation is reinforced by adding a term representing the treatment and time interaction. The time-depe ndant covariate represen ted by this interaction term can then be tested for significance. The treatment by time interaction in this study is significant (p-value<.001). Th ese two methods show that th e effect of quinolones on tendon rupture is changing over time when compared to the cephalosporin group. Since exposure is dichotomous in this study, it is more exact to sa y that the quinolone coefficient varies over time differently than the cephalosporin coefficient. Based on the significant p-value of the inte raction term the Research Question 2 null hypothesis was rejected.

PAGE 79

79 Ho: = 0 Ha: 0 Research Question 3 At what interval, in days, is the hazard rati o of tendon rupture post-quinolone therapy at its highest? To remedy the proportional hazard violati on a Cox piece-wise regr ession analysis was utilized to model the p ieces of the hazard function that appear to be proportional. In plotting the relative hazard between the quinolone and ce phalosporin groups at ar bitrary 10-day pieces or intervals it appears the risk of tendon rupt ure increases directly after being exposed to quinolones and becomes non-significant after day 30 (Figure 4-7). This estimation of risk is misleading because it ignores the unique ti me-varying nature of risk in the study. By using the hazard function, stratified by treatment and adjust ed for other model covariates, as a guide, an altern ative to the equal intervals can be formulated to provide an improved risk profile for exposur e to quinolones compared to cepha losporins. The intervals, post antibiotic dispensing, that best cap ture the effect modification of time were: 1) day 0 through 25; 2) day 26 through 35 and; 3) day 36 though 59. Since the hazard function remains generally proportional within these three inte rvals, it allows estimation of a less misleading hazard ratio for each interval. To reinforce the observation that the hazard function is proportional within each of the three intervals a treatment *time interaction term was tested for each of the intervals and was statistically non-significant. The first interval, day 0, has a signifi cant hazard ratio of 1.38 (95% confidence interval: 1.11.73), while the seco nd intervals (26 35) hazard ratio increases to 1.74 (95% confidence interval: 1.49 2.04). The last inte rval, formed from days 36 through 59, exhibited

PAGE 80

80 no significant difference between the two treatm ent groups 1.00 (95% confidence interval: 0.60 1.65) (Figure 4-8). The hazard function performs two important tasks in this analysis. First, it helps detect a treatment effect modified by time, which resulted in a violation in the proportional hazards assumption in this study and second, it helps determ ine the best way to overcome this violation and attain optimal estimates of the risk of tendon ruptures from quinolone use. Referring back to Figure 4-6 and 4-8, the risk progression over the 60day study period comes into focus. First, in Figure 4-6, it appears the risk of tendon rupture in the quinolone group is increased immediately after the index date. This is confirmed in Ta ble 4-8 by a significant hazard ratio of 1.38 (pvalue=0.004; 95% confidence interv al 1.11-1.73) for the first 25 days post exposure. Next, the hazard function shows a sizable increase in risk that diverges from the cephalosporin group around day 25. The ten day, 25 to 35, period measur es the highest hazard ratio estimate (1.74; pvalue<.001; 95% confidence interval 1.49-2.04) (F igure 4-8) in the quinolone group compared to the cephalosporin group. The hazard function then begins to diminish after day 35 where it approaches a comparable risk to the cephalosp orin group, until the end of the study period. Lastly, the non-significant haza rd ratio of 1.01 (p-value=0.886 ; 95% confidence interval 0.601.65) (Figure 4-8) from day 36 to 59 confirms th e reduction of risk at th e end of the study period. Indeed the model selection pro cess, which uses the -2logL statistic to test for model fit illustrates that model 3 is a more adequate combination of explanatory variables (Equation 3-12, Table 4-21). Model 3 contains three treatment predictor variables that represent windows of time at risk while Model 2 only includes one pred ictor variable or risk window, for treatment. These results point to an average induction pe riod that begins, immediately after exposure to quinolone antibiotics and ex tends through day 35 post exposure The maximum risk of tendon

PAGE 81

81 rupture from quinolone use occurs between days 26 through 35 compared to cephalosporin use. In other words, the risk of tendon rupture is incr eased by 38% from the first administration of the drug to around day 25, compared to cephalosporin us e. After this interval, the risk increases by 74% starting around day 25 and lasting until roug hly day 35. Finally, approximately 35 days after a patient is dispensed a pr escription for a quinolone antibiotic their increased risk of tendon rupture compared to the cephalospor in group, falls to nearly zero. Two of the three, interval regression coeffici ents, are shown to be significantly different from zero in Figure 4-8. Due to this, the nul l hypothesis for Research Question3 was rejected. Ho: Ax1 (1) = Bx1 (2) = Cx1 (3) = 0 Ha: at least one 0 The method of using Cox piece-wise regres sion to overcome the violation of the proportional hazards model assumption results in a be tter model fit for the effect of changing risk in this study. Research Question 4 What are the significant risk factors, other than quinolone treatment, for tendon rupture relevant to active duty military personnel? Several of the covariate regression coefficients are significantly different from zero (Table 4-14). Due to this, Research Ques tion 4 null hypothesi s was rejected. Ho: 2 = 3= = 11 = 0 Ha: at least one 0

PAGE 82

82 Each of the remaining covariates is analy zed for their effect on tendon ruptures when adjusting for the other covariates. The hazard function for the modeled Cox equation, stratified by each independent dichotomous variable, is revi ewed for the presence of any variable by time interaction. This both identifies the most im portant risk factors and optimizes the resulting hazard ratio. MOS: Military occupation was subdivided into direct involvement characterized by more physically demanding jobs, and support, dis tinguished by its admini strative and medical components. As can be ascertained from inspect ing the hazard function, direct versus support occupations are generally in proportion througho ut the study period (Figure 4-9). This is reinforced by a non-significant time dependent in teraction term (p-value =0.096). There is a spike of tendon rupture cases in the direct group around day 25, but the support group is also increasing over the same interval. The overall h azard ratio of MOS is significant at HR=1.53 (pvalue<.001; 95% confidence interval 1.24-1.89) (Table 4-7). An estimated 53% of tendon ruptures in this cohort are e xplained by having a military occupa tion that requires more physical exertion compared to supporting personnel. This makes intuitive sense as soldiers, sailors, airmen and marines who carry, lift, or patrol more should be more prone to orthopedic injuries. Branch: Branch of service is a four category vari able consisting of 1) Army 2) Air Force 3) Navy and 4) Marines. After adjusting for other covariates, there emerged one significant difference in risk between the Army and Marine Corps. The Marine Corps had a 65% higher risk of tendon rupture over th e 60-day study period compared to the Army (p-value=0.002; HR=1.65; 95% confidence interval 1.20-2.28) (Table 4-9b). The Marine Corps has very little support personnel and is generally considered to be the most phys ically demanding branch in the

PAGE 83

83 active duty force. This result coincides with th e significant military occupational findings that more physical occupations result in an increase risk of tendon ruptures regardless of treatment. Viewing the Marine Corps versus Army hazar d functions reveal a relative increase in tendon ruptures for the Marines, which occur late r when compared to the Army (Figure 5-10). The time-dependent interaction te rm is significant (p -value=0.028), meaning that the Marine Corps and Army hazard functions are not generally proportional over the 60-day study period. Further analysis of the risk prof ile for the Marine Corps when using the same interval as before, shows that the risk is highest from day 35 through day 50 (p-value 0.048; HR= 1.78; 95% confidence interval 1.04, 2.52) when co mpared to their Army counterparts. Provider Specialty: Patients being seen by specialty car e providers (internist, infectious disease, nephrologists, etc.) were almost tw o and a half times more likely to incur a tendon rupture than patients seeing a primary care provider (HR 2.40, p-value <.001; 95% confidence interval 1.92-2.99) (Table 4-9b). The hazar d curve is generally proportional over the study period giving reassurance that the Cox regressi on model is appropriately estimating the hazard ratio (Figure 4-11). The test of proportionality time-depe ndant interaction term is nonsignificant (p-value=0.105). This variable seems to help explain the belief that sicker patients require specialty care and that sicker patients tend to have an array of rela ted or unrelated health risk factors for tendon ruptures. The subgroup anal ysis of primary care versus orthopedic care was non-significant (p-value 0.457). Days of Medication Supply: The days of supply explanator y variable consists of the number of days a patient is expe cted to take the medication assuming the directions are followed. An example of days of supply would be a prescr iption for 30 capsules with instructions to take twice a day. The days supply in this case would be 15. The most frequently prescribed days of

PAGE 84

84 supply for antibiotics in this study were: 1(4.2%), 3(6.6%), 5(8.5%), 7(18%), 10(44%), 14(3.7), and 30(9.5). This is commensurate with prescribing guidelines for quinolones and cephalosporins. The days of supply variable is significan t (p-value<.001; HR=1.02; 95% confidence interval 1.018-1.022, Table 4-14). The result indi cates there is a 2% increase in tendon rupture risk for every additional day of quinolone treatment. When the days of supply variable was stratified by exposure the result was a significant 3% increase in tendon ruptures per day in the quinolone group (p-value<.001, HR=1.03; 95% c onfidence interval 1.028-1.039, Table 4-15). This stratification also include d a protective finding in the cephalosporin group (p-value<.001, HR=0.90; 95% confidence interval 0.854-0.957, Table 4-16) For further insight, days of supply was categor ized by the most frequent days of supply occurring in the quinolone sample. Inspection of the hazard ratios show th at the longer a patient is on quinolones the higher thei r risk of tendon ruptur e becomes compared to patients taking cephalosporins. Patients prescribed quinolone s for the most common 10-day regimen are 1.39 times more likely to experience a rupture, whil e patients on a 30-day supply are 2.7 times more likely to incur a rupture compared to patients prescribed cephalosporins (Figure 4-12). Age: The continuous variable age is statistica lly significant (p-value = 0.002; HR 1.02; 95% confidence interval 1.001-1.029) (Table 4-14). It can be interpreted as increasing the risk of tendon rupture by 2% as a patient ge ts 1-year older or that in th is sample every year older a person ages increases their risk by 2% compar ed to patients who received cephalosporins.. To illustrate the relationship between age and tendon ruptures the hazard ratio was calculated for four time intervals (Figure 4-13). The hazard ratio of in curring a tendon rupture over a 10-year span was 1.20, 20-year 1.43, 30year 1.72, and 40-years 2.05. Having a 2%

PAGE 85

85 increase in risk per year means that over a ty pical 20-year military car eer a soldier, sailor, airman, or marine has a 43% incr ease in their risk of tendon rupture merely from the effects of aging, which included reduced flexibility and a loss in vascular pe rfusion in tendon tissue. Diagnosis Leading to Antibiotic Treatment: Quinolone and cephalo sporin antibiotics have similar indications for use. These indicatio ns were divided into three main categories for analysis. The comparison group is upper re spiratory infections (URI), followed by gastrointestinal / genitourinar y (GI/GU) infections, and lastly soft tissue / bone infections (ST/BI). GI/GU use came closest to being significant (p-value=0.058; HR=1.25; 95% confidence interval 0.99-1.57) (Table 4-14), when compared to URI. Specifically, patients prescribed quinolones and cepha losporins for GI/GU infections were 25% more likely to experience a tendon rupture when compared to patie nts who were prescribed these antibiotics for URI infections. Notably, treatment of ST/BI with quinolones or cephalo sporins, was not an important explanation of tendon ruptures when co mpared to URI (p-value=0.590) (Table 4-14). The hazard functions stratified by the treatmen t groupings show the similarity of the URI and ST/BI curves (Figure 4-14). The graph also reveals the slight increase in events in the GI/GU group between day 10 and day 20, compared to the URI group. Grade: The explanatory variable grade is dichot omous, consisting of officers compared to enlisted military personnel. This variable is evenly distributed between the two treatment groups (enlisted 78.3 versus 77.5 and officer 21.7 versus 22.5,Table 4-8). Case s of tendon rupture are also evenly distributed between treatment groups (Table 5-7), indicated by the non-significant hazard ratio of 1.08 (p-value=0.529; 95% confid ence interval 0.85 1.37, Table 4-14) when comparing officers to enlisted members according to the occurrence of tendon ruptures over time. Grade does not vary with time as its tim e interaction variable is also non-significant (p-

PAGE 86

86 value=0.199). Officer and enlisted members of th e military are not significantly different at any interval post exposure when compared to each ot her, even after viewing the baseline hazard function (Figure 4-15) and tail oring the intervals accordingly (day 30 through 40: p-value = 0.163; HR = 0.71; 95% confidence interval 0.45 1.20). The occurrence of tendon ruptures does not seem to be affected by whether a person is an officer or enlisted member of the armed forces in this study Gender: Gender has a non-significant hazard ratio over the 60-day study period, (pvalue=0.165; HR=1.20; 95% confidence interval 0.93 1.54, Table 4-14). Upon viewing the hazard function for males and females it appears th ere may be a time-varying increase in risk occurring from day 25 to day 35 (Figure 4-16). Th e gender-time interaction term trends toward statistical significance but ultimately is non-si gnificant (p-value = 0.068) signifying that the occurrence of tendon ruptures according to ge nder is generally pr oportional over the study period. The hazard ratios for the intervals fr om index to day 25 (p-value=0.374), from day 26 through 36 (p-value=0.157and from day 36 to 60 (p-value=1.00), are non-significant. Gender does not affect the occurrence of tendon ruptures in this study. Race: The race variable is divided into three gr oups, 1) White, 2) Black, and 3) Other. Categorical comparisons utilizing the White categ ory as the reference group versus Black (pvalue=0.697) and Other group (p-value=0.218) base d on tendon rupture occurrence, proved to be statistically non-significant (Tab le 4-14). When Black and the Other categories were combined and compared to Whites, the results we re still non-significant (p-value=0.260). The hazard function stratified by race illustrates the similarity of the risk of tendon rupture between the different race categories (Figure 4-17).

PAGE 87

87 Steroid: The steroid explanatory variable desc ribes whether a DoD member used oral steroids prior to, during, or prom ptly after the study period. Th e distribution is similar between treatment groups (Table 4-8) and between cases stratified by treatment group (Table 4-7). As such, steroid use is not an impor tant independent variable in this study as revealed by its nonsignificant p-value (p-value=0.925; HR = 0.98) (Table 4-14), and hazard function which is nearly overlapping during the enti re 60-day period (Figure 4-18).

PAGE 88

88 0 2,000,000 4,000,000 6,000,000 8,000,000 10,000,000 12,000,000 JanMarMayJulSepNov 2002 2003 2004 2005 2006 Figure 4-1. Total DoD pres criptions by month and year 0 10,000 20,000 30,000 40,000 50,000 60,000 70,000 80,000 20022003200420052006 NPD Females Figure 4-2. Normal pregnancy delivery in the DoD (ICD-9 650.00) by gender and year

PAGE 89

89 Table 4. Normal pregnancy delivery in the DoD (ICD-9-650.00) by gender and year* 2002 2003 2004 2005 2006 NPD 63,184 71,805 68,88759,98159,800 Females 63,181 71,802 68,88159,97959,796 Coherence (%) 99 99 99 99 99 includes entire DoD population; NPD = normal pregnancy delivery 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 20022003200420052006 BPH Males Figure 4-3. Benign prostate hyp erplasia in the DoD (ICD-9-600. 00 600.91) by gender and year Table 4-2. Benign prostate hyperplasia in the DoD (ICD-9-600.00 600.91) by gender and year* 2002 2003 2004 2005 2006 BPH 8,011 8,487 10,72510,76715,792 Males 7,985 8,472 10,68110,75015,770 Coherence (%) 99 99 99 99 99 includes entire DoD population; BPH = benign prostate hyperplasia

PAGE 90

90 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 20022003200420052006 Type I DM Insulin RX Figure 4. Diabetes mellitus type 1 versus insulin use in the Do D (ICD-9-250.01) by year Table 4. Diabetes mellitus type 1 versus insulin in the DoD (ICD-9-250.01) by year* 2002 2003 2004 2005 2006 Type I DM 12,090 12,906 12,76814,03814,210 Insulin rxs 11,579 12,098 12,14513,10813,843 Coherence (%) 96 94 95 93 97 includes entire DoD populati on not just active duty; one physician encounter for t ype 1 DM per fiscal year

PAGE 91

91 0 100,000 200,000 300,000 400,000 500,000 600,000 700,000 800,000 20022003200420052006 Insulin rxs Diabetes dx Figure 4. Insulin use versus diabetes mellitus in the DoD (ICD-9_250.xx) by year Table 4. Insulin use versus diabetes mellitus in the DoD (ICD-9-250.xx) by year* 2002 2003 2004 2005 2006 Insulin rxs 593,529 638,242 662,790693,532721,029 Diabetes dx 552,335 603,485 604,589670,142701,287 Coherence (%) 92 94 90 96 97 includes entire DoD populati on not just active duty;

PAGE 92

92 Table 4-5. Sample demographics N=107,334 quinolone n=59,264 cephalosporin n=48,070 p-value DoDa n=1,419,061 male 70.7 78.2 85.0 Gender (%) female 28.3 21.8 <.001 c 15.0 Age, year (meanSDd) 29.1.8 28.7.4 <.001 b 28.2 White 72.7 75.1 64.2 Black 18.2 16.1 Race (%) Other 9.1 8.8 <.001 c 35.8 a Department of Defense; b independent t-test; c chi-squared test; d Standard deviation Table 4-6. Case demographics Na=382 quinolone n=255 cephalosporin n=127 p-value Case (%) 66.7 33.3 <.001e male 74.1 88.2 Gender (%) female 25.9 11.8 0.035c Age, year (meanSDb) 31.6.9 30.0.7 <.069d White 78.0 70.9 Black 13.3 22.0 Race (%) Other 8.6 7.1 0.706c a total cases; b Standard deviation; c Chi-squared test; d Independent t-test; e test of difference in proportions

PAGE 93

93 Table 4-7. Descripti on of tendon rupture cases Nb=382 quinolone n=255 cephalosporin n=127 p-value Case (%) 66.7 33.3 <.001e Median time to rupture (days) 27 95%CI(22,29) 12 95%CI(11,16) <.001f direct 44.3 46.5 Occupation (%) support 55.7 53.5 0.692c enlisted 77.6 74.0 Grade (%) officer 22.4 26.0 0.431c Steroid use (%) 7.8 6.3 0.585c Days of supply, days (meanSD) 16.9.4 8.2.6 0.001d Primary care 61.6 87.4 Provider Specialty (%) Specialty care38.4 12.6 <.001c URIb 65.5 66.9 GI/GUc 27.8 25.2 Diagnosis leading to abxa treatment (%) ST/BId 6.7 7.9 0.810c Army 31.4 34.6 Air Force 32.9 21.3 Navy 23.1 32.3 Branch of Service (%) Marines 12.5 11.8 0.082c a Antibiotic; b total cases; b Standard deviation; c Chi-squared test; d Independent t-test; e test of difference in proportions; f Kaplan-Meier estimator

PAGE 94

94 Table 4-8. Independent variables Nh = 107,334 quinolone n=59,264 cephalosporin n=48,070 p-value Treatment group (%) 55.2 44.8 <.001g Army 34.6 34.9 Air Force 32.5 30.4 Navy 19.8 21.3 Branch of Service (%) Marines 13.1 13.4 <.001f direct 32.5 33.9 support 60.9 60.6 Occupation (%) unknown 6.6 5.5 <.001f enlisted 78.3 77.5 Grade (%) officer 21.7 22.5 0.004f Steroid use (%) 7.5 7.0 0.002f Days of supply, days (meanSD) 11.1.2 9.8.4 <.001e Primary care 84.9 85.2 Provider Specialty (%) Specialty care15.1 14.8 0.173f URIb 70.4 69.9 GI/GUc 22.9 23.2 Diagnosis leading to abxa treatment (%) ST/BId 6.7 6.9 0.137f a Antibiotic; b Upper respiratory infection; c Gastrointestinal infection/Genital urinary; d Soft tissue/Bone infection; e Independent t-test; f Chi-squared test; g Test of difference in proportions; h Total number of patients

PAGE 95

95 Table 4. Model 1 regression para meters for demographic variables Model one Variable p-value HRa 95%CIb Age <.001 1.025 1.014, 1.036 Gender, malee 0.071 1.26 0.99, 1.61 Black 0.672 0.94 0.72, 1.24 Racef Other 0.537 0.89 0.61, 1.29 Model Summary -2logLc w/out covariates 8848.583 w/ covariates8823.352 L-ratio(df)d 25.23(4) p-value <.001 a HR = hazard ratio; b CI = Confidence interval; c -2 log likelihood; d Likelihood ratio (degrees of freedom); e reference female; f reference White Table 4-10. Treatment variab le stratified by occupation Support Na = 65,239 TRb No TR p-valueORc 95%CId quinolone 142 35,946 cephalosporin 68 29,083 <.001f 1.69f 1.27, 2.26 a total number of secondary patients; b tendon rupture; c odds ratio; d confidence interval; f chi-squared test

PAGE 96

96 Table 4-11. Treatment variab le stratified by occupation Direct Na = 35,577 TRb No TR p-valueORc 95%CId quinolone 113 19,149 cephalosporin 59 16,256 <.003f 1.63f 1.19, 2.23 a total number of secondary patients; b tendon rupture; c odds ratio; d confidence interval; f chi-squared test Table 4-12. Treatment variable stratified by provider specialty Primary Na = 91,259 TRb No TR p-valueORc 95%CId quinolone 157 50,152 cephalosporin 111 40,839 <.269f 1.15 0.90, 1.47 a total number of secondary patients; b tendon rupture; c odds ratio; d confidence interval; f chi-squared test Table 4-13. Treatment variable stratified by provider specialty Secondary Na = 16,075 TRb No TR p-valueORc 95%CId quinolone 114 8,857 cephalosporin 16 7,104 <.001f 4.91 2.89, 8.34 a total number of secondary patients; b tendon rupture; c odds ratio; d confidence interval; f chi-squared test

PAGE 97

97 Table 4-14. Model 2 full regression parame ters for treatment and covariate effects Model two Variable Parameter(SEa)2n p-valueHRb 95%CIc Treatment, quinoloned 0.501(0.110) 21.016 <.001 1.65 1.33, 2.04 Days supply 0.018(0.003) 90.305 <.001 1.02 1.018, 1.022 Gender, malee 0.179(0.130) 1.924 0.165 1.20 0.93,1.54 Grade, officerf 0.076(0.121) 0.397 0.529 1.08 0.85, 1.37 Steroid -0.019(0.197) 0.009 0.925 0.98 0.67, 1.44 Occupation, directg 0.424(0.107) 15.607 <.001 1.53 1.24, 1.89 Age 0.018(0.132) 9.584 0.002 1.02 1.01, 1.03 Provider, specialty careh 0.874(0.114) 59.310 <.001 2.40 1.92, 2.99 AF 0.035(0.132) 0.070 0.791 1.04 0.80, 1.34 N 0.292(0.135) 4.658 0.113 1.24 0.95, 1.63 Branchi M 0.362(0.173 4.398 0.002 1.65 1.20, 2.28 Black -0.055(0.142) 0.152 0.697 0.95 0.72, 1.25 Race j Other -0.234(0.190) 1.516 0.218 0.79 0.55, 1.15 GI/GUl 0.222(0.117) 3.595 0.058 1.25 0.99, 1.57 Antibiotic indication k ST/BIm 0.109(0.203) 0.291 0.590 1.12 0.75, 1.66 Model Two Summary -2logLo Model 1 8823.352 Model 28644.503 L-ratio(df)p 178.849(15) p-value <.001 a standard error; b hazard ratio; c confidence interval; d reference cephalosporin; e reference female; f reference enlisted; g reference support; h reference primary care; i reference army; j reference white; k reference upper resp iratory infection; l gastrointestinal/genitourinary; m soft tissue/bone infection; n Chi-squared statistic; o -2 log likelihood; p likelihood ratio(degrees of freedom)

PAGE 98

98 Table 4-15. Regression paramete rs stratified by treatment, adjusted for covariate effects, and reported based on study significance Quinolone strata Variable N=59,264 Parameter(SEa)2n p-valueHRb 95%CIc Days supply 0.033(0.003) 141.39<.001 1.03 1.028, 1.039 Occupation, direct g 0.429(0.131) 10.74 <.001 1.54 1.19, 1.99 Provider, specialty care h 1.27(0.130) 94.35 <.001 3.55 2.75, 4.58 Branch, Marinesd 0.626(0.202) 141.39<.002 1.87 1.26, 2.78 Age 0.017(0.007) 5.94 <.015 1.0171.003, 1.039 a standard error; b hazard ratio; c confidence interval; g reference support; h reference primary care; n Chi-squared statistic; d reference Army Table 4-16. Regression parameters stratified by treatment, adjusted for all covariate effects, and reported based on study relevance Cephalosporin strata Variable 48,070 Parameter(SEa)2n p-valueHRb 95%CIc Days supply -0.101(0.029) 12.15<.001 0.900.85, 0.96 Occupation, direct g 0.410(0.184) 2.21 <.026 1.511.05, 2.16 Gender, male 0.686(0.279) 6.04 <.014 1.991.15, 3.43 Branch, Marined 0.253(.281) 0.081<.369 1.290.74, 2.34 a standard error; b hazard ratio; c confidence interval; g reference support; n Chi-squared statistic; d reference Army

PAGE 99

99 Table 4-17. Regression paramete rs stratified by provider specialt y, adjusted for all covariate effects, and reported based on study relevance Secondary care strata Variable N=14,854 Parameter(SEa)2n p-valueHRb 95%CIc Tx, quinolone d 1.591(0.271) 34.61<.001 4.91 2.89, 8.34 Occupation, direct g 0.256(0.198) 1.68 <.196 1.29 0.88, 1.91 Days of Supply 0.024(0.005) 16.63<.001 1.0241.012, 1.036 Antibiotic indication, GI/GUe 0.468(0.204) 5.27 <.022 1.60 1.07, 2.38 a standard error; b hazard ratio; c confidence interval; g reference support; n Chi-squared statistic; d reference cephalosporin;e reference URI Table 4-18. Regression paramete rs stratified by provider specialt y, adjusted for all covariate effects, and reported based on study relevance Primary care strata Variable N=85,961 Parameter(SEa)2n p-valueHRb 95%CIc Tx, quinolone d 0.155(0.125) 1.54 <.215 1.17 0.91, 1.50 Occupation, direct g 0.501(0.127) 15.55<.001 1.65 1.29, 2.12 Branch, Marinese 0.535(0.193) 7.67 <.0056 1.71 1.17, 2.50 Days of Supply 0.155(0.125) 66.79<.001 1.0181.013, 1.022 Age 0.021(0.007) 8.79 <.003 1.0211.007, 1.035 a standard error; b hazard ratio; c confidence interval; g reference support; n Chi-squared statistic; d reference cephalosporin;e reference Army

PAGE 100

100 Table 4-19. Regression parameters stratified by occupation, adjust ed for all covariate effects, and reported based on significance Direct strata Variable N=35,576 Parameter(SEa)2n p-valueHRb 95%CIc Tx, quinolone d 0.473(0.164) 8.37 <.004 1.61 1.17, 2.21 Provider, specialty care h 0.786(0.172) 20.81<.001 2.20 1.57, 3.08 Branch, Marinese 0.677(0.227) 8.85 <.003 1.97 1.26, 3.07 Days of Supply 0.027(0.006) 20.86<.001 1.0271.016, 1.039 a standard error; b hazard ratio; c confidence interval; h reference primary care; n Chi-squared statistic; d reference cephalosporin; e reference Army Table 4-20. Regression parameters stratified by occupation, adjust ed for all covariate effects, and reported based on significance Support strata Variable N=65,239 Parameter(SEa)2n p-valueHRb 95%CIc Tx, quinolone d 0.526(0.148) 12.54<.001 1.69 1.26, 2.26 Provider, specialty care h 0.947(0.150) 39.85<.001 2.58 1.92, 3.46 Antibiotic Indication, GI/GUe 0.545(0.149) 13.36<.001 1.72 1.29, 2.31 Days of supply 0.018(0.002) 61.05<.001 1.0181.013, 1.022 Age 0.029(0.007) 16.46<.001 1.0291.015, 1.043 a standard error; b hazard ratio; c confidence interval; h reference primary care; n Chi-squared statistic; d reference cephalosporin; e reference URI

PAGE 101

101 Table 4-21. Model selection for objective 3 Model Selection Summary -2logLa Model 2 8644.503 Model 3 8523.213 L-ratio(df)b 121.290(18) p-value <.001 a -2 log likelihood; b likeli hood ratio(degrees of freedom) G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0.00012 0.00013 0102030405060 Adjusted for model covariates; group 1 = cephalosporin, group 2 = quinolone Figure 4-6. Hazard function stratified by treatment group*

PAGE 102

102 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 1011 2021 3031 4041 5051 60 Hazard Ratio Figure 4-7. Plotted treatment h azard ratio at 10-day intervals p-value HR 95%CI 0 10 <.001 1.62 1.31, 2.01 11 20 <.001 1.67 1.30, 2.14 21 30 0.015 1.43 1.07, 1.91 31 40 0.274 1.24 0.85, 1.81 41 50 0.889 1.04 0.60, 1.81 51 59 0.956 1.01 0.31, 3.21

PAGE 103

103 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 2526 3536 59 Hazard Ratio Figure 4-8. Treatment hazard rati o at three selected intervals p-value HR 95%CI 0 25 0.004 1.38 1.11, 1.73 26 35 <.001 1.74 1.49, 2.04 36 59 0.886 1.01 0.60, 1.65

PAGE 104

104 G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0.00012 0.00013 0.00014 0102030405060 1 = support, 2 = direct Figure 4 9. Hazard function stratified by military occupation*

PAGE 105

105 G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0.00012 0.00013 0102030405060 Group 1 = Marine Corps; Group 2 = Army G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0102030405060 G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0.00012 0.00013 0102030405060 Group 1 = Air Force ; Group 2 = Army Group 1 = Navy; Group 2 = Army Figure 4. Hazard function stratified by branch*

PAGE 106

106 G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0.00012 0.00013 0.00014 0.00015 0.00016 0.00017 0.00018 0.00019 0.00020 0.00021 0.00022 0102030405060 1 = primary care, 2 = specialty care Figure 4 11. Hazard function stratified by provider specialty

PAGE 107

107 0 0.5 1 1.5 2 2.5 3 1357101430 Hazard Ratio *Stratified by treatment, quinolone group Days 1 3 5 7 10 14 30 HR 1.03 1.11 1.18 1.26 1.39 1.59 2.70 Figure 4. Most frequent days of quinol one days of supply supply risk estimates

PAGE 108

108 0 0.5 1 1.5 2 2.5 10203040 Hazard Ratio 10 year interval; 20 year interval; 30 year interval; 40 year interval Interval (yrs) 10 20 30 40 HR 1.20 1.43 1.72 2.05 Figure 4. Age interval risk estimates

PAGE 109

109 G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0.00012 0.00013 0102030405060 G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0.00012 0102030405060 Group 1 = URI; Group 2 = ST/BI Group 1 = URI; Group 2 = GI/GU Figure 4. Hazard function stra tified by antibiotic indication* G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0.00012 0102030405060 Group 1 = enlisted, 2 = officers Figure 4-15. Hazard function stratified by military grade*

PAGE 110

110 G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0102030405060 1 = females, 2 = males Figure 4-16. Hazard function stratified by gender*

PAGE 111

111 G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0102030405060 G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0102030405060 Group 1 = Black; Group 2 = White Group 1 = Other; Group 2 = White Figure 4-17. Hazard function stratified by race* G R O U P12 0.00000 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006 0.00007 0.00008 0.00009 0.00010 0.00011 0102030405060 1 = no steroid, 2 = steroid Figure 4-18. Hazard function stratified by steroid use*

PAGE 112

112 CHAPTER 5 DISCUSSION Objective 1 Estimate the risk of tendon rupture from quinolone antibiotics use, relative to cephalosporin antibiotic use, while adjusting for relevant risk factors and accounting for time at risk. This is the first study to demonstrate in a military population an association between quinolone use and an increase in this physically debilitating injury. The incidence of tendon rupture was 0.43 % in the quinolone treated group and 0.26% in th e cephalosporin group. The resulting risk ratio of the two treatment groups in terms of tendon rupture occurrence, while adjusting for other relevant risk factors, re vealed that active duty personnel dispensed a prescription for a quinolone antibiotic were 1. 65 times more likely to incur a tendon rupture during the 60-day post-exposure study interval. It is helpful from a public health and prev ention standpoint to a ssess, in the quinolone exposed group, how much of the total risk of tendon rupture is actu ally due to quinolone exposure. In this study, 39% of the cases of te ndon rupture that occur in individuals prescribed quinolones may be due to quinolones. More specif ically, 100 of the 255 cas es of tendon rupture, in the quinolone-exposed group, can be directly attributable to being prescribed a quinolone medication. This further translates into 168 cases of tendon rupture pe r 100,000 quinolone users. The absolute difference between the tw o treatment groups is 166 cases per 100,000 personnel. In other words if all the quinolone users were instead prescribed cephalosporins 166 cases of tendon rupture might have been avoided. This information is particularly important because patients presenting with an infectious dis ease must be treated. The increase in the risk of tendon rupture seen in this study from quinol one use affects the decision to prescribe a

PAGE 113

113 quinolone or cephalosporin if both are equally indicated. The reduction in unit readiness and cohesiveness caused by 166 to 168 tendon injuries requiring surgery and followed by long-term rehabilitation may be prevented by treating pati ents with cephalosporins instead of quinolones where appropriate, and by paying more attention to the specific adverse event risk factors and time line presented in this study. Most of the studies examining the relati onship between quinolone use and tendon rupture have been conducted in older populations. In two case-control studies (Van der Linden 2002, 2003), the relative risk of tendon rupture from qui nolone use varied from 7.1 for those over the age of 60, to 4.3 for patients between 60 and 79, and 20.4 for those patients aged 80 and older (19, 20). The discovery of a robus t finding of an increase in risk for patients ranging from 18 to 60 years old has not been previously reported. La rger relative risks in other studies may emanate from two probable areas. The first concerns ad vancing age, which has consistently been found to increase the risk of tendon rupture. Conseque ntially, age as a risk factor for tendon injury, will be reiterated as a meaningful result of this study in discussion of Objective 4. Next, is the fact that previous studies used a randomly assigned non-treatment control group for comparison while the present study utilized personnel who were treat ed with a class of antibiotics with similar indications for use as quinolones. By definition, the presen t studys cephalosporin control group and quinolone group were both being treated for an infectio us disease of some type. This study may have resulted in a better ri sk estimate, when compared to other comparable studies because the infectious diagnosis was effectively contro lled by sample selection, while other studies ignored this possible effect modifier. The use of quinolone antibiotics cannot and shoul d not be eliminated from clinical use due to their importance in treating a variety of medi cally important conditions. What can be done is

PAGE 114

114 to modify their use based on a more patient specif ic approach. This patient specific approach, which will be outlined in the discussion to follow, has the potential to reduce the incidence of a serious injury that not only takes military personnel out of their unit for long periods of time, but also may inflict life long morbidity on the affected person. Objective 2 Investigate whether the risk of tendon rupture from quinolone use varies over the 60-day post exposure interval. Exposure to quinolone antibiotics was shown in Objective 1 to increas e the risk of tendon ruptures. Objective 2 seeks to analyze the mode l in order to provide the most accurate risk estimate. This is done by testing the proportional hazards assumption to determine if the risk of tendon rupture is varying over time. The proportiona l hazards model controls for the effect of time at risk prior to the consid eration of the association between the predictor variables and the outcome variable. This model assumes that ther e is no interaction between time at risk and the predictor variables. In this st udy, the variability of the risk is verified by a significant exposure by time interaction term. This informs us that the initial risk estimate may be misleading because the risk is changing over the study period. In order to increase the precision of the risk estimate the piecewise regression model, which entails breaking the 60-day study period in several pieces was utilized. This method restores the proportionality of the proportional hazar ds model. The resulting risk estimates, of 1.38, 1.74, and 1.01 occurring between days 1 to 25, 26-35, and 36 to 59 respectively, show that the risk of tendon rupture from quinolone use is changing over the 60-day post exposure interval and that the maximum risk is higher than 1.65. This endeavor to identify time at risk variabil ity is important for several reasons. First, it reveals that traditional risk estimates found in th e literature are likely conservative in nature

PAGE 115

115 because they fail to assess the tr ue effect of time on the risk of tendon rupture. Second, it shows that the effect of time at risk on exposure can be quantified in the absenc e of medication strength or blood levels even when exposure is defined in a dichotomous fashion. Third, this analysis allows the identification of a risk time-line that is beneficial in truly understanding how quinolone use effects the occurrence of tendon ruptur es that may be useful in informing clinical practice for the purpose of increasing troop readiness. Objective 3 Identify the period of maximum risk a nd average induction pe riod within the study interval. The term induction period is a more precise epid emiologic term to explain the lag or post exposure follow-up time in this study. Inducti on period is the time from causal action until disease initiation, thus it has a de finite biologic connotation. In th is study, it is defined by risk. It is the time from the dispensing of a prescrip tion for a quinolone to th e increase in risk of tendon rupture when compared to the cephalosporin group. The literature on this topic points to a sequence of action of various causes that can le ad to a ruptured tendon. Quinolone use is one postulated causal factor. Specifically, this study is not dealing with a general induction period for tendon rupture but with the induction period related to quinol one use, which includes the induction periods for the other component causes. If only one analysis is conducted, using a single exposure period where the time window is misclassified (e.g. too long or t oo short) and may or may not cont ain the relevant window where risk is maximized, information concerning the time relation between exposure and outcome may be misleading. In this study, the estimated hazard ratio is 1.65 over the entire 60-day postexposure period. The 60-day period is based on an assumption of the actual time at risk taken from previous studies on quinolone use. This study period surely contains portions of time when

PAGE 116

116 the risk of tendon rupture is small or absent. This misclassification of the time at risk leads to the attenuation of the risk estimate during the relevant time at risk period. Traditionally, epidemiologists will select a se t of induction time assumptions and estimate the exposure effect for each interval. The peak risk estimate would then be assumed the closest to the true effect. If the set of induction time a ssumptions provide a pattern of effects i.e rising then falling, the middle of the largest risk estim ate interval would be the best estimate of the induction period (76). The assumption of the time at risk from the beginning of exposure drives the selection of the study period. If a study period is selected that is diffe rent from the true time at risk, the resulting exposure measure is a misc lassified version of the true exposure. This exposure misclassification will reduce the size of the risk estimate and under estimate the risk of tendon rupture from quinolone use. By using the survival analysis technique of graphing the hazard function of the quinoloneexposed group, it allows for the more precise matching of the true time at risk with the assumed study time at risk period. This leads to a better maximum risk estimate and a better representation of the induction period. The hazard function allows the identification of the time in days, post quinolone exposure that the quinolone effect on tendon ruptures is at its zenith. Cr eating three time intervals using piecewise regression, restores the proportional hazard violation, iden tifies the best risk estimate, and reveals the induction period. It also adju sts for the effects of other model covariates. The average induction period starts at day one and lasts until appr oximately day 35 after receiving a prescription for a quinolone. This co incides with the hypothesi zed etiology related to quinolone induced tendon ruptures where tendon weakening begins to occur upon the initial dose and lasts for an unknown amount of time after the cessation of thera py. The best estimate of

PAGE 117

117 maximum risk, in a military population, is an increase of 74% compared to cephalosporin use, and occurs on average, between days 26 through 35. The risk model that includes three different intervals during the 60-d ay study period is statistically bett er than using a single 60-day post treatment model. The improvement in fit is modest when the 95% confidence intervals are scrutinized, as there is consid erable overlapping in the risk es timates for the 60-day and threepiece models. Additional analysis is needed in this area to pinpoint the induction period with more precision. Further research will utilize th e boot-strap statistical me thod to develop the most likely range of average induction periods which w ill a more statistical precision to estimation of the average induction period. The classification of an averag e induction period ha s great value when conveyed to health care providers as it lends itself to the patient counseling process. Patients who will be taking quinolones can be counseled that physical exer tion should be reduced for up to 35 days after exposure begins. This counseling may effectively help decrease tendon rupture cases. This can be particularly important because the negative effects of the infection for which they are receiving treatment will tend to subside before the time at risk for tendon rupture does. Soldiers, sailors, airmen, and marines will be back at work and undergoing normal physical fitness training after the infec tion related complications subside. Results from this study suggest a reduction in physical exertion for a longer period after quinolone therapy is may be indicated. Objective 4 Identify other major risk factors that increa se an active duty military individuals risk of tendon rupture. Tendon ruptures often require surgery and result in long periods of rehabilitation. The ability of this study to identify risk factors in a contemporary popul ation that is diverse in terms of demographics and physical demands is extrem ely important. These results give the health

PAGE 118

118 care provider more specific information that aids th em in treating their patients in a way that will help avoid this debilitating condition. This in tu rn will help increase readiness in military units across the Department of Defense. Taken as a whole, the important i ndependent risk factors from this study paint a picture of a physical demanding environment, advancing age, and longer exposure to quinolones, as contri buting to more tendon ruptures. The occupation of military personnel is an im portant contributor for tendon rupture. Military members who have more physically de manding jobs are 53% more likely to have a tendon rupture when compared to support personnel. This dove tails nicely with the finding that Marines are 65% more likely than Army personne l to develop a tendon rupture. The Marine Corps, have a higher proportion of their memb ers involved in more physically demanding work than the other services. Marines typically rely on the Navy for certain support, transportation, and medical services. These two situations support the notion that military personnel engaged in more physically demanding occupations are more susceptible to musculoskeletal injuries This supports the recommendation that military personnel prescribed a quinolone antibiotic to treat their infection, and are classified as having a physically dema nding occupation should be given a longer period of reduced physical exertion (up to 35 days) to avoid tendon ruptures or other musculoskeletal injuries. These high-risk patients should be screened for the use of an alternative antibiotic such as a cephalosporin, if possible. One of the key findings from Lum (2002) is the impact to the military from the finding that tendon ruptures are more frequent as military pers onnel advance in age(3). This study reinforces those findings and extends them by quantifying a 2% increase in risk for every year older a service member ages. A typical 20-year military career increases a service members risk of

PAGE 119

119 tendon rupture by 43%. This is a prominent finding due to the reality that senior leaders from both the NCO and Officer Corps, are at a higher risk than junior personnel when involved in similar occupations or physical exertion. Th e long recovery time required post tendon rupture requires that senior leaders be taken away from their units for long periods. This kind of disproportionate risk between juni or and senior personnel can reduce readiness, as there are less senior leaders available to step in when one is undergoing a long re habilitation phase during tendon rupture recovery. A question often asked by clinicians is whether a shorter length of treatment will lead to a lower risk of adverse events while longer treatme nt duration leads to a higher risk of adverse events. This study suggests that l onger length of treatm ent with a quinolone is associated with a higher risk for tendon rupture when compared to treatment with cephalosporins. Every day of medication therapy results in a 2% increase in ri sk. A typical 3-day regimen for a urinary tract infection has a 6% increase while a 30-day re gimen for prostatitis carries a 72% increase compared to treatment with cephalosporins. These findings should encourage health care providers to ensure they are prescribing the minimum length of treatment. Extra days of unneeded quinolone treatment may expose patien ts to an unnecessary increase in risk. Another category of patients that should be monitored more closely are those referred for secondary or specialty care. These patients are 2.4 times more likely to experience a tendon rupture when compared to patients who are seen and treated by primary care providers. This may be seen through the lens of a Berksons bias explanation where by people with multiple conditions or risk factors are over represented in the secondary health care providers patient population(77-79). Generally, the literature support s the notion that pe rsons referred to secondary health care providers ar e more likely to be sicker th an patients who are seen and

PAGE 120

120 treated by primary care providers(80) The result is that referred patients have more, and a wider array, of health problems, includ ing musculoskeletal injuries than non-referred patients. In this study, it is not related to longer treatment regimens or an older patient clientele as both primary and secondary patients have comparable days of supply and age values. This information should alert secondary care providers to be vigilant to the fact that rare adverse events, like tendon ruptures from quinolone use, are mo re likely to happen in their pa tient population. The risk is compounded if a secondary health care provider is seeing an olde r patient or one that has a physically demanding military occupation or one that may require treatm ent with a quinolone. The main infectious disease diagnoses with indications for quinolone use can be broadly broken into upper respiratory trac t infections (URI), gastrointestinal/genit ourinary (GI/GU), and soft tissue/bone infections (ST/BI). Of th ese three divisions, only GI/GU approaches the threshold of being an important high-risk group for tendon rupture. Patie nts who are prescribed a quinolone for a GI/GU infectious diagnosis have an approxim ately 25% increase in their likelihood of experiencing a tendon r upture compared to patients with URIs. While the increase in risk is low, it should be fo llowed-up in later studies because the type of infectious disease diagnosis may help identify patien ts who are at increased risk for musculoskeletal injuries. These patients may benefit from a different an tibiotic or longer peri ods of reduced physical exertion post quinolone treatment. Limitations The M2 query software system that was used to recall results from the Military Health System clinical data repository has not been ex tensively used for research purposes. Because of this, there have been no published peer-reviewed investigations into the validity of the data. A comparison of the data with pa per records was not possible due to the expansiveness of the Military Health System. Consequently, patient information that was not entered into the

PAGE 121

121 database was not included in this study. Th e amount of missing data cannot be precisely known without external validation. In this study, the assessment of database validity was conducted by data integrity and inte rnal cross-checking methods and found to meet generally acceptable data coherence thresholds(54, 68, 81). Another limitation related to the inability to compare the database records with the paper patient record is the occurrence of rule-out entries for tendon r upture. The rupture of a tendon is a serious injury that would necessitate c ontinued management by the military health care system. Patients identified in this study by having a diagnosis for a tendon rupture were required to have more than one diagnosis entry to lim it possible misclassification error. There is no reason to assume that the occurrence of rule -out diagnoses would fa vor one treatment group over the other. A significant limitation of naturalistic observatio nal designs is the lack of randomization. The absence of randomization produces the potentia l for selection bias, which may lead to any differences between the two treatm ent groups to be the result of this systematic error that is unrelated to treatment. Because of the la rge sample size, the two groups differ in many characteristics. Characteristic s that are different at baseline between treatment groups and are related to tendon ruptures are da ys of supply, military occupa tion, age, provider specialty, military branch, and diagnosis leading to antibiotic treatment. Multivariate analysis was used to attenuate the confounding produced by selection bias on the primar y predictor variable. The measurement and inclusion of important aspects of this population related to their demographics and health care surrounding their being treated with antibiotics was accomplished in order to adjust for pre-existing differences between s ubjects in the two treatment groups. This

PAGE 122

122 multivariable analysis helps to identify and control for the affect of selection bias, which enhances the confidence in the results. Even though the implementation of an observati onal study design is a lim itation, it is also an asset. The results of an observational study still provide a realistic estimate of the risk of tendon rupture in the real world day-to-day milita ry, with potential effect modifiers identified and adjusted for in a multivariable model. Speci fically, randomized clinical trials comparing the risk cannot be conducted on quinol one induced tendon ruptures beca use of the large sample size required to find a noteworthy result from an outcome with such a small incidence in the population. To study adverse even ts, which occur in less than 1% of the population, like tendon ruptures, observational studies are uniquely well suited to pr ovide meaningful results. The two treatment groups in this study came from an active duty military population. Military Personnel are screened pr ior to entry for congenital or ch ronic maladies that may render them unable to perform active duty service. On average military personnel are more physically fit, than their civilian counterparts(82). Thes e and other differences between active duty military and civilian communities may limit the generalizabil ity of the obtained point estimates from this study to the US Military. Military members, who seek out medical care, usually present at the local medical facility during sick call hours in the early morning prior to the start of their duty day. If the treating physician decides that they are indeed ill, he or she may provide the patient with quarters paperwork. Quarters paperwork informs the pa tients unit that he/she is ill and requires a modified duty schedule. These modifications range from light duty (no physical training) to confinement to quarters for a specified length of time. This may affect the timing of the occurrence of tendon ruptures post quinolone exposure. Quinolone antibiotics likely weaken

PAGE 123

123 tendon tissue resulting in a higher probability of a rupture occurring during physical exertion. When a patient receives quarters he/she is less likely to participate in physically demanding actions. This may lower the incidence of te ndon rupture immediately following the patients diagnosis prompting quinolone therapy. The prac tice of giving quarters may reduce the risk of tendon rupture from quinolone use and result in a longer induction period. Recommendations 1 Health care providers should c onsider the use of cephalosporin antibiotics in place of quinolones if indicated or if th e clinical presentation allows. 2 If quinolones are required to treat a patients infection, make certain that the length of treatment is optimized to reduce the risk of tendon rupture. 3 When prescribing a quinolone, the health care provider shou ld understand a reduced duty profile might be necessary for up to 35 days after starting therapy. 4 Military members who are prescribed quinol ones should be asked about their military occupation specialty. Patients with physica lly intensive occupati ons should be given profiles to avoid extreme physical exertion fo r up to 35 days after starting therapy. 5 Secondary and specialty health care providers should be made aware of the increased vulnerability of their pati ent population to tendon ruptures. Information concerning alternate antibiotic medications with simila r indications and the need for a reduction in extreme physical exertion should be provided. 6 Care should be taken to deve lop an individual risk prof ile, including age, occupation specialty, indication for quinolone use, length of treatment, and branch of service, for each patient when prescribing quinolones. Patients at high risk should be prescribed a different antibiotic if possible or gi ven reduced duty status if quinolone use is necessary. Conclusion This is the first study to iden tify an increased risk of tendon rupture from quinolone use in a large demographically diverse pop ulation. The risk is elevated upon the first day of therapy and increases incrementally each day until it reach es its maximum between days 26-35. The risk of tendon rupture is augmented by advancing ag e, length of treatment, and working in a physically demanding environment. These findings are important as they provide evidence that a

PAGE 124

124 significant portion of this physic ally debilitating injury is po tentially avoida ble by identifying specific risk factors and incorporating th em into the over all patient care plan. Injuries are the single most significant medical impediment to the U.S. Armys ability to project and sustain a healthy and medically prot ected force(1). This dissertation identifies quinolone antibiotics and othe r independent characteristics as ri sk factors for a debilitating type of musculoskeletal injury. It also recommends ways to reduce the lost duty time and shortened military careers that result from tendon ruptures.

PAGE 125

125 LIST OF REFERENCES 1. Atlas of injuries in th e United States Armed Forces Mil Med 1999; 164(8 Suppl):633 pages 2. Stahlmann R: Clinical toxicological aspects of fluoroquinolones. Toxicol Lett 2002; 127(1-3):269-77 3. Lum GR: Spontatneous ruptures of th e achilles tendon,US Armed Forces, 1998-2001. Medical Surveilance Monthly Re port January 2002; 8(1):2-6 4. Bailey RR, Kirk JA, Peddie BA: Norfloxacin-induced rheumatic disease. N Z Med J 1983; 96(736):590 5. Ribard P, Audisio F, Kahn MF, De Bandt M, Jorgensen C, Hayem G, Meyer O, Palazzo E: Seven Achilles tendinitis including 3 complicated by rupture during fluoroquinolone therapy. J Rheumatol 1992; 19(9):1479-81 6. Huston KA: Achilles tendinitis and tendon r upture due to fluoroquinolone antibiotics. N Engl J Med 1994; 331(11):748 7. Riley G: The Pathogenesis of Tendinopat hy. A Molecular Perspective. Rheumatology 2003; 43:131-142 8. Rothman KJ: Induction and latent pe riods. Am J Epidemiol 1981; 114(2):253-9 9. Lincoln AE, Smith GS, Amoroso PJ, Bell NS : The natural history and risk factors of musculoskeletal conditions resulting in di sability among US Army personnel. Work 2002; 18(2):99-113 10. Lauder TD, Baker SP, Smith GS, Lincoln AE: Sports and physical training injury hospitalizations in the army. Am J Prev Med 2000; 18(3 Suppl):118-28 11. Lincoln AE, Smith GS, Amoroso PJ, Bell NS : The natural history and risk factors of musculoskeletal conditions resulting in di sability among US Army personnel. Work 2002; 18(2):99-113 12. Lauder TD, Baker SP, Smith GS, Lincoln AE: Sports and physical training injury hospitalizations in the army. Am J Prev Med 2000; 18(3 Suppl):118-28

PAGE 126

126 13. Smith TA, Cashman TM: The incidence of in jury in light infantry soldiers. Mil Med 2002; 167(2):104-8 14. Jones BH, Hansen BC: An armed forces epid emiological board evalua tion of injuries in the military. Am J Prev Med 2000; 18(3 Suppl):14-25 15. Potter RN, Gardner JW, Deuster PA, Jenkins P, McKee K, Jr., Jones BH: Musculoskeletal injuries in an Army air borne population. Mil Med 2002; 167(12):103340 16. Lum GR: Spontatneous ruptures of th e achilles tendon,US Armed Forces, 1998-2001. Medical Surveilance Monthly Re port January 2002; 8(1):2-6 17. Haddow LJ, Chandra Sekhar M, Hajela V, Gopal Rao G: Spontaneous Achilles tendon rupture in patients treated w ith levofloxacin. J Antimicr ob Chemother 2003; 51(3):747-8 18. Melhus A: Fluoroquinolones and tendon diso rders. Expert Opin Drug Saf 2005; 4(2):299309 19. Van der Linden PD, Sturkenboom MC, He rings RM, Leufkens HG, Stricker BH: Fluoroquinolones and risk of Achilles tendon disorders: case-control study. Bmj 2002; 324(7349):1306-7 20. Van der Linden PD, Sturkenboom MC, He rings RM, Leufkens HM, Rowlands S, Stricker BH: Increased risk of achilles te ndon rupture with quinolone antibacterial use, especially in elderly patie nts taking oral corticostero ids. Arch Intern Med 2003; 163(15):1801-7 21. Van der Linden PD, Van Puijenbroek EP Feenstra J, Veld BA, Sturkenboom MC, Herings RM, Leufkens HG, Stricker BH: Tendon disorders attributed to fluoroquinolones: a study on 42 spontaneous re ports in the period 1988 to 1998. Arthritis Rheum 2001; 45(3):235-9 22. Zabraniecki L, Negrier I, Vergne P, Arnaud M, Bonnet C, Bertin P, Treves R: Fluoroquinolone induced tendinopathy: report of 6 cases. J Rheumatol 1996; 23(3):51620

PAGE 127

127 23. Wojcik BE, Hassell LH, Humphrey RJ, Davi s JM, Oakley CJ, Stei n CR: A disease and non-battle injury model based on Persian Gu lf War admission rates. Am J Ind Med 2004; 45(6):549-57 24. Appenzeller GN: Injury patterns in p eacekeeping missions: the Kosovo experience. Mil Med 2004; 169(3):187-91 25. Bohnker BK, Bowman W, Dell D, Gutermuth F: Disease nonbattle in jury surveillance for commander, Joint Task Force Haiti, 2004. Mil Med 2005; 170(12):1032-3 26. Maffulli N, Waterston SW, Squair J, Reaper J, Douglas AS: Changing incidence of Achilles tendon rupture in Scot land: a 15-year study. Clin J Sport Med 1999; 9(3):157-60 27. Maffulli N: Rupture of the Achilles tendon. J Bone Joint Surg Am 1999; 81(7):1019-36 28. Waterston SW, Maffulli N, Ewen SW: Subc utaneous rupture of the Achilles tendon: basic science and some aspects of clinical practice. Br J Sports Med 1997; 31(4):285-98 29. Van der Linden PD, Nab HW, Simonian S, Stricker BH, Leufkens HG, Herings RM: Fluoroquinolone use and the change in incide nce of tendon ruptures in the Netherlands. Pharm World Sci 2001; 23(3):89-92 30. Lewis JR, Gums JG, Dickensheets DL: Levofloxacin-induced bilateral Achilles tendonitis. Ann Pharmacoth er 1999; 33(7-8):792-5 31. Jagose JT, McGregor DR, Nind GR, Ba iley RR: Achilles tendon rupture due to ciprofloxacin. N Z Med J 1996; 109(1035):471-2 32. Huston KA: Achilles Tendinitis and Tendon Rupture Due to Fluoroquinolone Antibiotics. N Engl J Med 1994; 331(11):74833. Carrasco JM, Garcia B, Andujar C, Garro te F, de Juana P, Bermejo T: Tendinitis associated with ciprofloxacin. Ann Pharmacother 1997; 31(1):120 34. Audisio F, Kahn MF, De Bandt M, Jorgense n C, Hayem G, Meyer O, Palazzo E: Seven Achilles tendinitis including 3 complicated by rupture during fluoroquinolone therapy. J Rheumatol Ribard, P. 1992; 19(9):1479-81

PAGE 128

128 35. Van der Linden PD, van de Lei J, Nab HW Knol A, Stricker BH: Achilles tendinitis associated with fluoroquinolones. Br J Clin Pharmaco l 1999; 48(3):433-7 36. Seeger JD, West WA, Fife D, Noel GJ, Johnson LN, Walker AM: Achilles tendon rupture and its association with fluoroquinolone antibiotics and other potential risk factors in a managed care population. Phar macoepidemiol Drug Saf 2006; 15(11):784-92 37. Lesher GY, Froelich EJ, Gruett MD, Bailey JH, Brundage RP: 1,8-Naphthyridine Derivatives. A New Class Of Chemotherapeutic Agents. J Med Pharm Chem 1962; 91:1063-5 38. Ito A, Hirai K, Inoue M, Koga H, Suzue S, Iri kura T, Mitsuhashi S: In vitro antibacterial activity of AM-715, a new na lidixic acid analog. Antimic rob Agents Chemother 1980; 17(2):103-8 39. Hampel B, Hullmann R, Schmidt H: Ciprof loxacin in pediatrics: worldwide clinical experience based on compassionate use--saf ety report. Pediatr Infect Dis J 1997; 16(1):127-9; discussion 160-2 40. Bailey RN, R Linton,AL: Nalidixic acid arthralgia. Canadian Medical Association Journal 1972; 107:60607 41. McDonald DS, HB: Usefulne ss of nalidixic acid in treatment of urinary infection. Antimicrob Agents Chemother 1964; 4:628-631 42. Stahlmann R: Children as a special populat ion at risk--quinolones as an example for xenobiotics exhibiting skeletal toxic ity. Arch Toxicol 2003; 77(1):7-11 43. Stahlmann R, Lode H: Fluoroquinolones in the elderly: safety considerations. Drugs Aging 2003; 20(4):289-302 44. Khaliq Y, Zhanel GG: Fluoroquinolone-associ ated tendinopathy: a crit ical review of the literature. Clin Infect Dis 2003; 36(11):1404-10 45. Fleisch F, Hartmann K, Kuhn M: Fluoroqui nolone-induced tendinopathy: also occurring with levofloxacin. Infe ction 2000; 28(4):256-7 46. Stahlmann R, Lode H: Toxicity of quinolones. Drugs 1999; 58 Suppl 2:37-42

PAGE 129

129 47. Casparian JM, Luchi M, Moffat RE, Hint horn D: Quinolones and tendon ruptures. South Med J 2000; 93(5):488-91 48. Wilton LV, Pearce GL, Mann RD: A comp arison of ciprofloxacin, norfloxacin, ofloxacin, azithromycin and cefixime examined by observational cohort studies. Br J Clin Pharmacol 1996; 41(4):277-84 49. Hartzema A, Porta,M., Tilson,HH.: Pharmacoepidemiology-An Introduction. In-Press; 4rd Edition 50. Carson JL RW, Strom BL: Medicaid databa ses, in Pharmacoepidemiology. Edited by BL S. Chichchester, Wiley, 2000, pp 307-324 51. Anderson DW, Bryan FA, Jr., Harris BS, 3rd, Lessler JT, Gagnon JP: A survey approach for finding cases of epilepsy. P ublic Health Rep 1985; 100(4):386-93 52. Kashner TM: Agreement between administrativ e files and written medical records: a case of the Department of Veterans Affairs. Med Care 1998; 36(9):1324-36 53. Newton KM, Wagner EH, Ramsey SD, McCullo ch D, Evans R, Sandhu N, Davis C: The use of automated data to identify compli cations and comorbidities of diabetes: a validation study. J Clin Epidemiol 1999; 52(3):199-207 54. Rawson NS, D'Arcy C: Assessing the validity of diagnostic information in administrative health care utilization data: experience in Saskatchewan. Pharmacoepidemiol Drug Saf 1998; 7(6):389-98 55. Osborne ML, Vollmer WM, Johnson RE, Buis t AS: Use of an automated prescription database to identify individuals with asthma. J Clin Epidemiol 1995; 48(11):1393-7 56. Christensen DB, Williams B, Goldberg HI, Martin DP, Engelberg R, LoGerfo JP: Comparison of prescription and medical record s in reflecting patient antihypertensive drug therapy. Ann Pharmacother 1994; 28(1):99-104 57. Christensen DB, Williams B, Goldberg HI, Martin DP, Engelberg R, LoGerfo JP: Assessing compliance to antihypertensive me dications using computer-based pharmacy records. Med Care 1997; 35(11):1164-70

PAGE 130

130 58. Malenka DJ, McLerran D, Roos N, Fisher ES, Wennberg JE: Usi ng administrative data to describe casemix: a comparison with the medical record. J Clin Epidemiol 1994; 47(9):1027-32 59. Virnig BA, McBean M: Administrative data for public health su rveillance and planning. Annu Rev Public Health 2001; 22:213-30 60. Rawson NS, Malcolm E: Validity of the reco rding of ischaemic heart disease and chronic obstructive pulmonary disease in the Saskatchew an health care datafiles. Stat Med 1995; 14(24):2627-43 61. Rawson NS, Malcolm E, D'Arcy C: Reliabili ty of the recording of schizophrenia and depressive disorder in the Saskatchewan hea lth care datafiles. Soc Psychiatry Psychiatr Epidemiol 1997; 32(4):191-9 62. Curkendall SM, DeLuise C, Jones JK, Lanes S, Stang MR, Goehring E, Jr., She D: Cardiovascular disease in patients with chronic obstructive pulmonary disease, Saskatchewan Canada cardiovascular dise ase in COPD patients. Ann Epidemiol 2006; 16(1):63-70 63. Ray WA, Griffin MR, Downey W: Benzodiaz epines of long and short elimination halflife and the risk of hip fr acture. Jama 1989; 262(23):3303-7 64. Movig KL, Leufkens HG, Lenderink AW, E gberts AC: Validity of hospital discharge International Classification of Diseases (IC D) codes for identifying patients with hyponatremia. J Clin Epidemiol 2003; 56(6):530-5 65. Wang PS, Walker AM, Tsuang MT, Orav EJ, Le vin R, Avorn J: Find ing incident breast cancer cases through US claims data and a state cancer regi stry. Cancer Causes Control 2001; 12(3):257-65 66. Ray WA, Griffin MR, Downey W, Melton LJ, 3rd: Long-term use of thiazide diuretics and risk of hip fracture. Lancet 1989; 1(8640):687-90 67. Gerstman BB, Freiman JP, Hine LK: Use of subsequent anticoagulants to increase the predictive value of Medicaid deep venous thromboembolism diagnoses. Epidemiology 1990; 1(2):122-7

PAGE 131

131 68. Hennessy S, Bilker WB, Weber A, Strom BL: Descriptive analyses of the integrity of a US Medicaid claims database. Pharm acoepidemiol Drug Saf 2003; 12(2):103-11 69. Naber KG, Adam D: Classification of fluoroquinolones. Int J Antimicrob Agents 1998; 10(4):255-7 70. Cox D: Regression Models and Life Tables (w ith Discussion). Journal of the Royal Statistical Society 1972; B(34):187-220 71. Cox D, Oakes D: Analysis of Survival Data. 1984 72. Guess HA: Exposure-time-varying hazard func tion ratios in case-control studies of drug effects. Pharmacoepidemiol Drug Saf 2006; 15(2):81-92 73. Ray WA: Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol 2003; 158(9):915-20 74. Jewell N: Statistics for Epidemiology. 2004 75. CDC: Epi Info(TM)Database and statistics software for public health professionals. 2004 76. Rothman KG, S: Modern Epidemiology. 1998; 2nd Edition:297-300 77. Schwartzbaum J, Ahlbom A, Feychting M: Berkson's bias reviewed. Eur J Epidemiol 2003; 18(12):1109-12 78. Berkson J: Limitations of the application of fourfold table analysis to hospital data. Biometrics Bulletin 1946; 2(3):47-53 79. Roberts RS, Spitzer WO, Delmore T, S ackett DL: An empirical demonstration of Berkson's bias. J Chroni c Dis 1978; 31(2):119-28 80. Sackett DL: Bias in analytic re search. J Chronic Dis 1979; 32(51-63) 81. Ray WA, Griffin MR: Use of Medicaid da ta for pharmacoepidemiology. Am J Epidemiol 1989; 129(4):837-49

PAGE 132

132 82. Population Representation in the Military Se rvices. U.S. Department of Defense Fiscal Year 2004:3-11

PAGE 133

133 BIOGRAPHICAL SKETCH Patrick M. Garman was born on March 15, 1968 in Troy, Ohio. The second of three boys, he grew up in Miami County Ohio and graduate d from Miami East High School in 1986. He earned his B.S. in pharmacy from Ohio Northe rn University (ONU) in 1991. He was also a member of the Bowling Green University Army Reserve Officer Training Corps (ROTC) and received a commission as a 2nd Lieutenant in the Medical Service Corps. Upon graduating in May 1991 with his degree in pharmacy, Patrick en tered the active duty Army. After completion of his Officer Basic Course, he spent his initial two assignments serving as the director of pharmacy at Fort Irwin, California and Ba umholder Health Clinic Baumholder Germany. During his tour of duty in Germany, he was selected to the Armys Long-Term Health Education and Training program (LTHET) which afforded him the opportunity to attend The Ohio State University and attain the Doctor of Pharmacy degree. His first assignment after graduation in July 2000 was as the Director of Pharmacy in the 121st General Hospital, Seoul Republic of South Kor ea. From there, he was assigned in July 2002 as the Pharmacy Consultant to the Commander of the United States Ar my Medical Material Agency (USAMMA), Fort Detrick Maryland. Duri ng this assignment, he was the point man for the Department of Defense in the distributi on of the Anthrax and Small Pox vaccines for Operation Enduring Freedom (Afghanistan) and Op eration Iraqi Freedom (Iraq). This tour of duty ended with another selection into the Ar mys LTHET program and his assignment at the University of Florida, College of Pharmacys graduate program in Pharmacy Health Care Administration to pursue Doctor of Philosophy degree. Upon completion of his Ph.D. pr ogram, Patrick will be assigned to the U.S. Army Military Vaccine Agency (MILVAX) as the Deputy Director for Scientific Affairs. MILVAX is located at the Army Surgeon Generals Office in Falls Church, Virginia. He has attained the rank of

PAGE 134

134 Lieutenant Colonel during his 16 years of active dut y service. Patrick has been married to Kim Garman for 15 years and they have four childr en: Keye, age 10; Kenna, age 9; Bella, age 19 months; and Kitrick, age 2 months.