1 PHOSPHODIESTERASE TY PE 5 INHIBITORS AND RISK OF SUDDEN SENSORINEURAL HEARIN G LOSS: A POST MARKETING SAFETY STU DY By WEI LIU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2014
2 Â© 2014 W ei Liu
3 To my f amily !
4 ACKNOWLEDGMENTS The completion of the thesis would not have been possible without the help and support of many people . First , I am indebted to my supervisor, Dr. Almut G. Winterstein , for her consistent support and guidance in supervising the thesis research . Thank you so much for sharing your extensive knowledge on pharmacoepidemiology and for always finding time to answer my questions which helped to the growth of my professional attitude. I would like to acknowledge the contribution of my other dissertation committee me mbers : Drs. Patrick J Antonelli and Philipp Dahm for your clinical input and guidance, and for encouraging me to undertake this challenging yet very excit ing project; Dr. Tobias Gerhard who helped me get access to the MarketScan data through your home inst itute at the Rutgers University, for which I am especially grateful; Dr. Joseph AC Delaney for your friendly advice, encouragement and insightful discussion about the project; and f inally Dr. Richard Segal for your constant encourage ment and support throug h my graduate training . I would also like to thank the faculty members and staff at the Department of Pharmaceutical Outcome and Policy. In particular, thank you to Paul Kubilis, Dandan Xu, and Carl Henriksen for your patience to teach me how to work on a dministrative claims data and how to run my analyses i n a serve r environment . Thank you to Jill Hunt, Linda Orr, and Nicole Corwine for everything you did to make my study at the UF such a rewarding experience. To my fellow graduate students, thank you for all your friendship and support throughout the years. I am particularly grateful to Mr. Lawrence J DuBow ; this work woul d not have been possible without your fellowship.
5 Finally, special thank s t o my wife and my lovely children , who continued to provide me with their support and belief in my ability to make it . Special thanks to my parents, my parents in law, without your love this wor k would not have been possible.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 11 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 Background ................................ ................................ ................................ ............. 15 Study Objectives ................................ ................................ ................................ ..... 18 2 LITERATURE REVIEW ................................ ................................ .......................... 20 Epidemio logy of Erectile Dysfunction ................................ ................................ ...... 20 Disease Burden of ED ................................ ................................ ...................... 20 Causes of ED ................................ ................................ ................................ ... 22 Treatment Option s for ED ................................ ................................ ................. 23 Oral Phosphodiesterase Type 5 (PDE5) Inhibitors ................................ ................. 26 Mechanisms of Action ................................ ................................ ...................... 26 Pharmacology of PDE5 Inhibitors ................................ ................................ ..... 28 Efficacy, Tolerability, Patient Preference ................................ .......................... 29 Sudden Sensorineural Hearing Loss (SSHL) ................................ .......................... 30 Risk Factors of SSHL ................................ ................................ ....................... 30 Diagnosi s and Treatment for SSHL ................................ ................................ .. 32 PDE5 Inhibitor Use and Risk of SSHL ................................ ................................ .... 33 Critical Appraisal of Existing Evidence ................................ ............................. 33 Possible Mechanisms of PDE5 Inhibitor Induced SSHL ................................ ... 36 3 METHODOLOGY ................................ ................................ ................................ ... 44 Overview of Study Design ................................ ................................ ....................... 44 Data Source and Data Files Linkage ................................ ................................ ...... 47 Part 1: Patterns of PDE5 Inhibitor Use in a Commercially Insured ......................... 49 Study Population ................................ ................................ .............................. 49 Data Analyses ................................ ................................ ................................ .. 49 Prevalence and incidence of ED ................................ ................................ 49 Prevalence and incidence of PDE5 inhibitor use ................................ ....... 51 Longitudinal pattern of PDE5 inhibitor use ................................ ................. 53
7 ................................ 54 Study Design and Population ................................ ................................ ........... 54 PDE5 Inhibitor Exposure ................................ ................................ .................. 56 Study Endpoint ................................ ................................ ................................ . 58 Covariates ................................ ................................ ................................ ........ 59 Data Analyses ................................ ................................ ................................ .. 60 Sensitivity An alyses ................................ ................................ .......................... 61 Part 3: Comparative Safety of PDE5 Inhibitors and Risk of SSHL .......................... 62 Study Cohort and Design ................................ ................................ ................. 63 Characterization of PDE5 Inhibitor Exposure ................................ ................... 64 Study Endpoint ................................ ................................ ................................ . 65 Baseline covariates ................................ ................................ .......................... 65 Data Analysis ................................ ................................ ................................ ... 65 Sensitivity Analyses ................................ ................................ .......................... 67 Bias Rel ated Issues ................................ ................................ ................................ 67 4 RESULTS ................................ ................................ ................................ ............... 78 Part 1: Patterns of PDE5 Inhibitor Use in a Commercially Insured Population ....... 78 Prevalence and Incidence of Diagnosed ED ................................ .................... 78 Prevalence and Incidence of PDE5 Inhib itor Use ................................ ............. 79 Longitudinal pattern of PDE5 inhibitor use ................................ ....................... 79 ................................ 81 Cohort Selection ................................ ................................ ............................... 81 Patient Characteristics ................................ ................................ ...................... 81 Risk of SSHL in Study Cohorts ................................ ................................ ......... 82 Part 3: Comparative Safety of PDE5 Inhibitors and Risk of SSHL .......................... 84 Cohort Selection ................................ ................................ ............................... 84 Patient Characteristics ................................ ................................ ...................... 84 Risk of SSHL in Study Cohorts ................................ ................................ ......... 85 5 DISCUSSION ................................ ................................ ................................ ....... 1 13 Part 1: Patterns of PDE5 Inhibitor Use in a Commercially Insured Population ..... 113 .............................. 118 Part 3: Comparative Safety of PDE5 Inhibitors and Risk of SSHL ........................ 124 Recommendation for future research ................................ ................................ ... 126 Summary and conclusion ................................ ................................ ...................... 127 APPENDIX A INCLUSION EXCLUSION CRITERION, APPLIED AS TO T 0 , THE PRESCRIPTION FILL DATE OR COMPARABLE DATE FOR NONUSER CONTROLS ................................ ................................ ................................ .......... 130 B EXCLUSION ILLNESSES APPLIED AS TO T 0 ................................ .................... 131
8 C DIAGNOSIS AND PROCEDURE CODES USED IN THE STUDY ....................... 132 TO IDENTIFY STUDY OUTCOME ................................ ................................ ....... 132 LIST OF REFERENCES ................................ ................................ ............................. 133 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 145
9 LIST OF TABLES Table page 2 1 Causes of Erectile Dysfunction ................................ ................................ .............. 38 2 2 Pharmacodynamics and Pharmacokinetics of PDE5 Inhibitors .............................. 39 2 3 Risk Factors for Sudden Sensorineural Hearing Loss ................................ ............ 40 2 4 Case Reports of PDE5 Inhibitor Use and SSHL ................................ ..................... 41 3 1 Baseline covariates: socio demographics ................................ .............................. 71 3 2 Baseline covariates: comorbid conditions ................................ .............................. 72 3 3 Baseline covariates: medications ................................ ................................ ........... 74 3 4 Baseline covariates: healthcare utilization during 6 months look back period ....... 75 3 5 Summary of secondary analyses in Part 2 ................................ ............................. 76 4 1 Prevalence of erectile dysfunction ................................ ................................ .......... 86 4 2 Incidence of erectile dysfunction ................................ ................................ ............ 88 4 3 Prevalence of PDE5 inhibitor use ................................ ................................ ........... 90 4 4 Incidence of PDE5 inhibitor use ................................ ................................ ............. 92 4 5 PDE5 inhibitor utilization pattern during 6 months after initiation ........................... 94 4 6 Prescription pattern during 12 months aft er PDE5 inhibitor initiation ..................... 95 4 7 Baseline characteristics of cohort members ................................ ........................... 96 4 8 HRs for the association between PDE5 inhibitor use and risk of SSHL ................. 99 4 9 Baseline patient characteristics of initiators of sildenafil and tadalafil .................. 101 4 10 Risk of SSHL comparing sildenafil with tadalafil (primary analysis) ................... 103 4 11 Risk of SSHL comparing sildenafil with tadalafil (secondary analyses) .............. 103 4 12 Baseline patient characteristics of initiators of sildenafil and vardenafil ............. 104 4 13 Risk of SSHL comparing sildenafil with vardenafil (primary analysis) ................ 106 4 14 Risk of SSHL comparing sildenafil with vardenafil (secondary analyses) .......... 106
10 LIST OF FIGURES Figure page 2 1 Mechanism of Action of PDE5 Inhibitors. ................................ ............................... 43 3 1 Study Flow Chart for Part 2 of the Thesis. ................................ ............................. 77 4 1 Annual prevalence and incidence of erectile dysfunction ................................ ..... 107 4 2 Annual prevalence and incidence of PDE5 inhibitor use ................................ ...... 108 4 3 Market share of total PDE5 inhibitor products dispensed in 1998 2007 ............... 109 4 4 Flow chart (Part 2) ................................ ................................ ................................ 110 4 5 Risk of SSHL and PDE5 inhibitor use ................................ ................................ .. 111 4 6 Study f lowchart (Part 3) ................................ ................................ ........................ 112
11 LIST OF ABBREVIATIONS ABR Auditory brainstem response AERS Adverse Events Reporting System AMP Adenosine monophosphate AUA American Urology Association CCAE Commercial Claims and Encounters cAMP C yclic adenosine 3,5 monophosphate cGMP C yclic guanosine 3,5 monophosphate CI Confidence interval CMV Cytomegalovirus CPT Current Procedural Terminology CVD Cardiovascular disease CYP Cytochrome P450 EAU European Association of Urology ED Erectile dysfunction FDA Food and Drug Administration GMP Guanosine monophosphate HIV Human immunodeficiency virus HPFS Health Profes sional Follow up Study HR Hazard ratio ICD 9 CM International Classification of Diseases, 9 th Revision, Clinical Modification IIEF International Index of Erectile Function iNOS Inducible nitric oxide synthase IPTW Inverse probability of treatment weighting
12 IRB Institutional Review Board LUTS Lower urinary tract syndrome MEPS Medical Expenditure Panel Survey MMAS Massachusetts Male Aging Study NAION Non arteritic anterior ischemic optic neuropathy NDC National Drug Code NHANES National Health and Nutrition Examination Survey NO Nitric oxide NSAIDS Nonsteroidal anti inflammatory drugs OAE Otoacoustic emission PAH Pulmonary arterial hypertension PDE5 Phosphodiesterase type 5 PRN Pro re nata PS Propensity score QOL Quality of life SHL Sudden hearing loss SNHL Sensorineural hearing loss SSHL Sudden sensorineural hearing loss US United States VA Veteran Affairs WHO World Health Organization
13 Abstract of Dissertation Presented to the Graduate School of th e University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy PHOSPHODIESTERASE TY PE 5 INHIBITORS AND RISK OF SUDDEN SENSORINEURAL HEARIN G LOSS: A POST MARKETING SAFETY STU DY By Wei Liu August 2014 Chair: Almut Winterstein Major: P harmaceutical Science s Several c ase report s published around 2007 suggest that phosphodiesterase type 5 ( PDE5 ) inhibitor use might increase the risk of sudden sensorineural hearing loss (SSHL). We c onducted a post marketing observational study to evaluate the association in a large cohort of privately insured male adults in the U S . In study 1, we analyzed 4.4 million pharmacy claims of PDE5 inhibitor s collected during 1998 2007. The annual prevalence of use inc reased by 44% from 16.4 (1998) to 29.5 (2007) per 1,000 subjects . The incidence of use increased from 11.5 ( 1998 ) to 18.6 per 1,000 person years ( 2007 ) . About 2/3 rd of new users refilled their prescription within 6 months after their index prescription. M e dication switching and dose escalation were not common in this population . In Study 2, 377,722 patients who initiated treatment with a PDE5 inhibitor and 1,957,233 nonusers we re followed for 4,652,265 person years. The crude incidence for SSHL was 4.35 an d 2.38 per 1,000 person years for current use and nonuse , respectively. In the primary analysis (where patients were assumed to take one PDE5 inhibitor dose per week ), the adjusted hazards ratio (HR) was 1.2 5 ( 95% CI, 1.01 1.55) , and the excess risk was 1. 97 SSHL per 1,000 person years , both comparing current
14 use with nonuse . We conducted sensitivity analyses to test the robustness of our findings. The results were not meaningfully different compared with the primary analysis . In Stud y 3, we examined the c omparative safety of the three PDE5 inhibitors . The analysis sugg ested that new use of vardenafil ( HR=0.80, 95%CI, 0.42 1.51 ) and vardenafil ( HR=1.02, 95%CI, 0.51 2.02 ) did not significant ly increase or decrease the risk of SSHL compared with new use of sildenafil (reference group) , although the small sample size resulted in limited statistical power of the analyses . In conclusion , we found an increased risk of SSHL associated with current use of PDE5 inhibitors . There was no significant difference in t he risk of SSHL comparing initiators of the different PDE5 inhibitor s. Given the widespread use of these medication s, clinicians and patients need to be aware of the potential ototoxic effect associated with the use of PDE5 inhibitors.
15 CHAPTER 1 INTRODUCTION Background an erection, or both, sufficient for , 1 is highly prevalent in the United States (US) and worldwide. 2 The National Institute of Health (NIH) Consensus Conference estimates that ED may affect as many as 30 million American men over the age of 20. 1 According to the Massachusetts Mal e Aging Study (MMAS), about 6 1 0,000 new cases of ED occur annually in the US. 3 Age is the most important risk factor of ED. T he prevalence of ED increase s significantly after a man turns 40 years . 4 In the past, ED has often been considered a s a psychological problem or a part of the natural aging process. It is now widely acknowledged that ED is primarily organic resulting from vascular, hormonal, or neurological complications. 5 In addition, ED is considered by many physicians as a marker for undiagnosed conditions such as hypertension, diabetes, and ischemic heart disease. 6 Oral phosphodiesterase type 5 (PDE5) inhibitors including sildenafil citrate (Via gra Â® , Pfizer Inc.), vardenafil (Levitra Â® , Bayer Inc.), and tadalafil (Cialis Â® , Eli Lily) are the recommended first line t reatment option for ED. 7 In March of 1998, sildenafil was the first oral PDE5 inhibitor approved by the Food and Drug Administration (FDA) to treat ED. Vardenafil and tadalafil were approved in 2003, providing more treatment options if sildenafil is either ineffective or undesirable. According to Pfizer, more than 35 million men worldwide used sildenafil between 1998 and 2007 8 and about 17 million PDE5 inhibitors are prescribed each year to an estimated 5 million men. 9
16 PDE5 is a cGMP specific binding protein, 10 and is the predominant cGMP hydrolyzing PDE in corpus cavernosum. Inhibition of PDE5 leads to increase in the level of cGMP, relaxa tion of smooth muscle, increased blood flow, enlargement of cavernosal tissue, and finally penile erection. 11 Taken approximately 30 60 minutes before sexual activity, sildenafil and vardenafil work for about 6 8 hours; tadalafil lasts for up to 36 hours. 12 Direct head to head comparative trials of the three PDE5 inhibitors are not available , but a systematic review of randomized controlled trials comparing PDE5 inhibitors to placebo suggests that the three medications have similar efficacy and safety profiles. 13,14 The first case report of PDE5 inhibitor us e and sudden sensorineural hearing loss (SS HL) was pu blished in the April 2007 issue of the Journal of Laryngology and Otology. 15 The authors described a 44 year old man who developed bilateral progressive hearing loss aft er taking sil denafil 50 mg continuously for 15 days. Management of the patient with inhaled oxygen and corticosteroid did not reverse the symptoms and his hearing was not improved at 6 months. The case was soon followed by a search of additional verse Events Reporting System (AERS) database , which resulted in a total of 29 cases that had a strong temporal association with use of PDE5 inhibitor s . In October of 2007, the FDA announced a labeling change for all PDE5 inhibitor product s available on the market alerting to the potential increased risk of SSHL in patients using these drugs . 16 As of today, a caus al association between PDE5 inhibitor exposure and risk of SSHL has not been established and underlying biological mechanism for the potential ototoxic effect is not totally clear. It has been postulat ed that the increased activity of nitric oxide (NO) , a
17 chemical that has been implicated in a wide variety of otologic diseases, including hearing loss secondary to bacterial meningitis 17 , gentamicin 18 and cisplatin ototoxicity 19 , might be the cause of SSHL in PDE5 inhibitor users . 20 A lthough case reports cannot be used to establish causality, the FDA believed that a strong temporal relationship between the use of PDE5 inhibitor and the subsequent occurrence of SS HL warrant ed r evisions of the product labels . 21 Men with SSHL have a lower quality of life (Q O L). Sudden hearing loss (SHL) , frequencies that occurs in less than 72 hours, 22 is often accompanied by an emergency visit to an otolaryngologist or audiologist . SSHL is one of many cau ses of SHL. About 10 15% of SSHL cases have identifiable causes including trauma, syphilis, autoimmune , t he remaining are idiopathic. 23 SSHL affect s roughly 5 to 20 individuals per 100,000 p opulation, with an expected 4,000 new cases per year in the US . 22 However, c urrent statistics likely underestimate the true significance of the problem because patients with spontaneous recovery may not seek health care. Ototoxic drugs or d rug induced hearing impairment ha ve not received significant attention in pharmacoepidemiol ogy , unless the drug class ( e.g. , aminoglycosides) i s alrea dy known to be ototoxic. T he usual pharmacoepidemiologic armamentarium to evaluate emerging safety concerns involving ototoxicity is therefore limited . In this thesis, a large healthcare administra tive data base (i.e., physician generated diagnos e s and procedure codes used for reimbursement) with the statistical power delivered by tens of millions of subjects was used to study drug induced ototoxicity. We based our methods on our previous study of th e ototoxicity of neomycin ear drops in pediatric
18 patients with non intact tympanic membranes us ing 2 9 state funded Medicaid program data . 24 We hope that these two studies will build a foundation for a pharmacoepidemiologic research program on drug induced ototoxicity. Finally, t he efficacy of marketed PDE5 inhibitors appears quite similar, with approximately 70% of men having successful intercourse after drug administration , but the W eekend P vardenafil. Th e increased flexibility in timing intercourse may influence and preference when they choose a PDE5 inhibitor . Furthermore, access to PDE5 inhibitor for ED patients is strictly regulated by health plans , 25 typically with restrictions on d rug choic e. Thus, prescribers and patients need information on comparative safety of these agents to help them make an informed decision to choose the optimal PDE5 inhibitor . Study Objectives T he ov erall objective of this thesis wa s to evaluate the associa tion between the use of PDE5 inhibitor s and the risk of S S HL in male adult s in a large commercial insurance database. We hypothesize d that PDE5 inhibitor users have an increased risk of SS HL compared to unexposed subjects. Other specific objectives include d : Study Objective 1 : To describe the trend s of PDE5 inhibitor use among male adult patients identified in the Thomson Reuters MarketScan Â® Commercial Claims and Encounters (CCAE) health insurance database for calendar years from 1998 to 2007. Study O bjective 1a : To estimate the annual prevalence and incidence of ED among male patients (18 64 years) in the CCAE data base . Study Objective 1b : To estimate the annual prevalence and incidence of P DE5 inhibitor use , overall and by predetermined age groups .
19 Study Objective 1 c : Among new users of PDE5 inhibitors, to describe the initial and longitudinal pattern of use including medication refills, switching and dose titration. Study Objective 2 : To evaluate the risk of S S HL comparing patients started on PDE5 inhibitors to nonuser s of PDE5 inhibitors, controlling for other confounders . Hypothesis: H 0 : Current use of PDE5 inhibitor s wa s not associated with an increased risk of S S HL relative to nonuse of PDE5 inhibitor s . H a : Current use of PDE5 inhibitor s wa s associated with an increased risk of S S HL relative to nonuse of PDE5 inhibitor s . Study Objective 3 : To e xamine the incidence rate and risk of S S HL among new users of sildenafil compared with vardenafil or tadalafil users , after controlling for other established and potential confounders . Hypothesis 3a: H 0 : There wa s no difference in the risk of S S HL comparing new users of sildenafil to new users of vardenafil . H a : The risk of S S HL wa s different comparing new users of sildenafil to new users of varden afil . Hypothesis 3b: H 0 : There wa s no difference in the risk of inciden t S S HL comparing new users of sildenafil to new users of tadalafil. H a : The risk of S S HL wa s different comparing new users of s ildenafil to new users of tadalafil .
20 CHAPTER 2 LITERATURE REVIEW This chapter i s divided into four sections. We first review s evidence o f disease burden, causes, and treatment option s for ED. Established risk factors for ED are review ed next . Second, the p harmacology of PDE5 inhibitors is reviewed brie fly . Topics include pharmacodynamics and pharmacokinetics, efficacy and side effects, a s well as factors that determine patient preference for different PDE5 inhibitors. T he section also contain s a summary of published drug utilization studies of PDE5 inhibitors. Third, common causes and diagnos e s of S S HL are reviewed. Finally , existing literature that discusse s t he use of PDE5 inhibitor s and S S HL are appraised . The potential biological mechanism linking the use of PDE5 inhibitor to SS HL is briefl y discussed. Epidemiology of Erectile Dysfunction Disease Burden of ED Disease burden of ED has been est ablished in studies mainly using community based surveys and administrative claims data . The landmark Massachusetts Male Aging Study (MMAS) was a po pulation based survey of 1,709 healthy men aged 40 70 years who lived in the Boston area. 5 A self administered sexual activity questionnaire was d elivered from 1987 1989 to characterize sexual function among participants of the study . The prevalence of combined minor, moderate, and severe ED estimated in this an erection sufficient for sexual i f or men at age 40 years to 25% by the age of 75. 5 A follow up survey of the study population 8 years later foun d that the crude annual incidence rate of ED was about 26 cases per 1,000 men. This translates to as many as 610,000 new cases of ED annually in the U S . 3 Other
21 studies have also tried to estimate the prevalence of ED. The Health Professional Follow up Study (HPFS) which included 31,742 mal e health professionals (dentists, optometrists, osteopaths, podiatrists, pharmacists, and veterinarians) aged 53 90 years found that 33% of survey respondents had some difficulty t o have and maintain an erection for intercourse in the previous 3 months . 26 The condition was more common among elderly men . 26 Using the 2001 2002 National Health and Nutrition Examination Survey (NHANES), Selvin et al. estimated that the prevalence of ED in male s age d 20 years and older was 18.4% (95% CI: 16.2 20.7). 27 In 2001 2002, Laumann et a l. conducted a survey among a national representative sample of men aged 40 and older and found the prevalence of moderate or severe ED was 22.0% (95% CI: 19.4 24.6) overall and the prevalence increased with increasing age. 28 Interestingly, studies that used healthcare administrative databases t ended to report a much lower rate of ED. For example, u sing 1999 2002 VA data , Sohn et al. found the prevalence of ED was only 2,884/100,000 veterans >17 years. The authors of the study defined ED using ICD 9 CM (302.71, 302.72, 302.72, 607.84, 607.82, 607.89, 607.9x) and CPT codes. 29 Using 2007 2009 VA administrative data (ICD9 CM codes 607.84 or 302.72), Hosain et al. estimated the prevalence of ED among 4,755 Iraq and Afghanistan veterans > 17 years old to be 5.5%. The prevalence of E D was 3.6% and 15.7% in men aged between 18 40 years and >40 years respectively. 30 Studied largely agree that ED is associated with a significa ntly reduced quality of life (QO L), loss of self confidence, self esteem, and damage to couple relationships , 31 as well as a significant medical cost. 32,33 In summary, the existing evidence suggests that ED is a prevalent condition among adult male i n the
22 US with a significant impact on personal wel l being, public health, and healthcare cost and utilization . Causes of ED In the past, ED has been thought to be of psychological in origin , because it is a common symptom of anxiety, depression, stress, or relationship problems. However, it is now widely acknowledged that ED is fundamentally organic resulting from vascular, hormonal, or neurological complications. In addition, ED is associated with many other pathological conditions, 4 and can serve as a marker for undiagnosed organic disease such as hypertension, diabetes, hyperlipidemia, kidney disease, atherosclerosis, neurological and endocrine disorders. 34 Other causes of ED include medical interventions such as prostate surg ery or medication use (Table 2 1). A n ormal erection depends on blood vessels in the penis. Thus, it is not surprising that vascular disease such as clogged arteries (atherosclerosis) can cause ED. Previous s tudies suggest that almost 90 percent of men with ED have at least one risk factor for cardiovascular disease , 26,35,36 which makes ED a marker for asymptomatic cardiovascular disease. 34,36 One hypothesis for the association between vascular disease and risk of ED is related to endothelial dysfunction. According to th is theory, endothelial dysfunction results in impaired smooth muscle relaxation by interfering with nitric oxide (NO) synthesis or release. 37 Thus, common CVD risk factors, including cigarette smoking, obesity, low physical activity, chro nic alcohol abuse, high blood pressure, coronary arterial disease, peripheral vascular disease, kidney disease, and penis, preventing or worsening proper erectile functio n . Another hypothesis that link s
23 vascular problem s with ED is the so called venous leak the blood seeps out of the penis during an erection instead of remaining trapped inside. 38 The neural system also plays a pivotal role in achieving or maintaining a normal erection. Damage to any nerves (i.e., due to diabetes) that produce sensation to arteries in the penis can also cause ED. Other degenerative disorders of the nervous system such as multiple sclerosis or . 39,40 In addition, spinal cord injury or treatment for prostate cancer (i.e., radical prostatectomy 27,35 , radiation therapy for prostate cancer 41 ) and bladder cancer can damage nerves and arteries needed for erection and lead to ED. Importantly, t he complex neurovascular process outlined above is susceptible to interference by various psychological problems such as anxiety, stress, or depression. A recent systematic review summarized studies that link lower urin ary tract symptoms (LUTS) to ED and concluded that this is likely a cause and effect relationship. 42 PDE5 inhibitor s are therefore recommended as either prophylaxis or as primary treatment for LUTS. 43 Finally, several prescription drugs have been associated with the development of ED. These include antihypertensive medications (thiazide diuretics, calcium channel b lockers, and beta blockers ), antipsychotics, antidepressants, and antihistamines. 44,45 Treatment Options for ED ED can be treated at any age. The goal of ED treatment is to restore a natural and normal sexual life and ultimately improve quality of life. 46 Several treatment options are available, most of which offer high rates of efficacy and a favorable safety profile. ED patients play a key role in c hoosing the treatment and evaluating the treatment effect. Important considerations to select the treatment include the underlying cause
24 with mild ED, simple lifestyl e change (i.e., weight loss, increased physical activity, or smoking cessation ), behavioral modification, psychological therapy (talk therapy for patients with stress related ED, addressing relationship difficulties), or treatment of underlying health prob lems may improve erectile dysfunction. However, as treatment of underlying organic risk factors does not always improve erectile function, medication therapy or other direct treatments are oftentimes necessary . 4 As most men prefer less invasive treatments, pharmacotherapy is currently the mainstay of treatment for ED. Pharmacological management of ED is available in oral dosage forms, as sublingual preparation, topica l gel, or as an injection. Currently, the guidelines by the World Health Organization (WHO) 47 , the European Association of Urology (EAU) 48 , and the American Urology Association (AUA) 49 support o ral phosphodiesterase type 5 (PDE5) inhibitors as the first choice for ED treatment. PDE5 is the predominant cGMP metabolizing enzyme in the penile tissue. By blocking PDE5, and thus the breakdown of the erection producing chemical cGMP, these drugs enable the penis to fill with blood and to stay erect for intercourse. In randomized clinical trials, the efficac y of PDE5 inhibitors wa s about 70 80% as assessed by the International Index of Erectile Function (IIEF) EF domain scores or global efficacy assessme nt response rates. 50 PDE5 inhibitors can be used by almost most ED patients regardless of the unde rlying cause of the ir disease, but best efficacy has been reported in ED cases with psychological causes . The efficacy of PDE5 inhibitors is lower in patients with severe ED (IIEF EF domain score <10), patients with diabetes and diabetic complications, and those who have undergone radical prostatectomy. 51
25 Apomorphine is a dopamine receptor agonist and is administered sublingually on demand in 2 or 3 mg dose. The centrally acting drug improves erectile function by enhancing the natural central erection si gnals. It has limited efficacy compared to PDE5 inhibitors and is therefore limited to patients with mild ED. 51 Intracavernous injection of vasoactive medications is considered as a second line therapy for ED, although it is associated with high efficacy and treatment satisfaction rate. 52 It is thus often used by patients for whom the PDE5 inhibitor is either not tolerable or ineffective. For example, t he injection therapy is more effective in patien ts whose ED is caused by diabetes. Its main disadvantage is the high discontinuation rate and limited compliance due to fear of needles and lack of spontaneity. 51 Intraurethral and other topical therapies are noninvasive, but the efficacy of these therapies is low compared to PDE5 inhibitors , and they have been associated with local adverse effects. They can be offered to patients not responding to PDE5 inhibitors. If none of t he medications is helpful, other options to treat ED include vacuum erection devices, hormone therapy, vascular surgery, and surgical implants. Among the non pharmacological options for ED, vacuum constriction devices (penile pump or VCD s) , are offered mai nly to elderly patients who engage in occasional sexual intercourse. Young patients show limited preference to VCD s because of the unnatural erection association with VCD s . VCDs are also offered to patients who are nonresponsive to oral drugs and who do no t want to use other more invasive methods such as intracavernous injections. Testosterone supplements are appropriate only if ED is caused due to testosterone deficiency. This cause has been showed to be less
26 prevalent in the majority of men. If testostero ne has been prescribed, it is often administered in gel form or as skin patches. Finally, penile implant s (semirigid rods) and blood vessel surgery are generally the last treatment option, because of the invasiveness, cost, and non reversibility, despite t he fact that they are associated with highest patient satisfaction rates in properly selected patient populations. 51 Oral Phosphodiesterase Type 5 (PDE5) Inhibitors PDE5 inhibitors are the recommended first line therapy for ED. Sildenafil was the first PD E5 inhibitor to be approved by the FDA as an oral pharmacotherapy for ED. Two closely related drugs vardenafil and tadalafil, were introduced to the ED market 5 years later. The three PDE5 inhibitors are administered orally and used on an as needed basis . The recommended initial dose for sildenafil is 50 milligrams (mg), which can be adjusted to 100 mg or 25 mg. 53 The recommended starting dose for tadalafil or vard enafil is 10 mg, and it can be titrated to 20 mg. According to the product labeling, PDE5 inhibitors are not recommended in men who have recent history of stroke or myocardial infarction (within 6 8 weeks), or who have si gnificantly low blood pressure, uncontrolled high blood pressure, unstable angina, severe cardiac failure, severe liver impairment or end stage renal disease requiring dialysis. 54 56 Sexual stimulation is a prerequisite for all PDE5 inhibitors. No PDE5 inhibitors should be taken more than once a day. 54 56 Mechanisms of A ction The human genome contains at least 21 genes involved in determining the intercellular levels of cyclic nucleotides (cyclic adenosine 3,5 monophosphate [cAMP], and cyclic guanosine 3,5 monophosphate [cGMP] ) by the expression of phosphodiesterases (PDEs). 57 At least 11 distinct PDE isoforms have been identified and their func tional properties determined. PDEs are enzymes that hydrolyze cAMP and
27 cGMP into AMP and GMP , respectively . Cyclic nucleotides are intracellular secondary messengers that regulate many important physiological processes, such as vascular smooth muscle contr action, apoptosis, bone formation, and insulin response. 58 The cAMP is formed from ATP by the enzy me adenylyl cyclase and cGMP is formed from GTP by the enzyme guanylyl cyclase , both of which are either membrane bound or soluble in the cytosol . The PDEs differentiate from each other based on sequence homology, enzymati c properties, and selectivity to inhibitors. But the most distinguishing feature between the 11 PDEs is their substrate specificity. Some PDEs specifically hydrolyze cAMP (PDEs 4, 7, 8); others are highly specific for cGMP (PDEs 5, 6, 9), whereas some are able to hydrolyze both cAMP and cGMP (PDEs 1, 2, 3, 10, 11). PDE5 is a cGMP specific PDE and is particularly abundant in vascular smooth muscle of the penis and lung. The PDE5 also presents in platelets, visceral smooth muscle (gastrointestinal tracts), br ain, and kidney. Competitive inhibitors of PDE are analogs of cyclic nucleotides that bind to the substrate binding site and directly decrease the rate of substrate catalysis. Drugs that can counteract against these enzymes can regulate the intracellular c oncentrations of cAMP and cGMP, thereby influencing a broad variety of physiological functions. Penile erectile tissue consists of the corpus cavernosum and the corpus spongiosum, which are composed of sponge like, interconnected trabecular spaces lined by vascular epithelium and smooth muscle. During sexual stimulation, nitric oxide (NO) is released from nerves and vascular endothelial cells in the penis. NO then passes through the smooth muscle cell membrane and stimulates the enzyme guanylate cyclase to convert guanylate triphosphate (GTP) to cGMP. The increased
28 concentration of cGMP causes vascular smooth muscle relaxation in the corpus cavernosum and in penile arterioles, leading to an inflow of blood, which can produce penile erection. PDE5 is the pred ominant PDE isozyme in penile tissue , which is responsible for metabolizing cGMP. All PDE5 inhibi tors selectively inhibit PDE5, hence preventing the breakdown of PDE5, enhancing the vasodilatory effect of NO, resulting in increased intracellular cGMP level s and restoring the erectile function (Figure 2 1). The selective induction of vasorelaxation in the penile tissue via the NO cGMP signaling pathway occurs only during sexual stimulation. Other tissues do not have this mechanism. Pharmacology of PDE5 Inhib itors As competitive inhibitor of PDE5, the PDE5 inhibitors have a chemical structure that resembles that of cGMP. Of the three PDE5 inhibitors, sildenafil possesses a molecular structure very similar to vardenafil, whereas tadalafil has a slightly differe nt structure. This structure difference determines a slight difference in the pharmacokinetic properties of the three agents. After oral administration, the three agents are rapidly absorbed from the gut. Time to peak plasma concentration (T max ) for silden afil and vardenafil is about 1 hour. Peak plasma concentration is reached approximately 2 hours after intake for tadalafil. The PDE5 inhibitors show marked differences in their elimination half life. According to the package insert, the elimination half li fe for sildenafil and vardenafil is about 3 5 hours, and tadalafil has a half life of about 17.5 hours. 54 56 Related to the half life data, the duration of action of sildenafil and vardenafil is about 4 6 hours , 54,55 but the duration of action of tadal afil is up to 36 hours. 56 The long duration of action for tadalafil allo ws ED patients more freedom to choose the timing of sexual encounter s (spontaneity). Another advantage of tadalafil is that fatty food has no effect
29 on its absorption, whereas it can delay the absorption of sildenafil and vardenafil, which in turn decrease s its maximum plasma concentration and delays the mean time to maximum plasma concentration. The three PDE5 inhibitors are metabolized hepatically, mainly by cytochrome P450 (CYP) isozyme 3A4. All three PDE5 inhibitors are excreted predominantly via the fe cal route and to a lesser extent in the urine (Figure 2 1). Efficacy, Tolerability, Patient Preference There are no published head to head trials to evaluate the efficacy of different PDE5 inhibitors. Clinical trials comparing one active drug to placebo pr ovide indirect evidence with regard to the comparative efficacy and safety of these agents. 59 In these trials, patient reported outcomes are widely used in evaluating the treatment success of PDE5 inhibitors. For example, the self administered 15 item questionnaire International Index of Erection Function (IIEF) has been validated in 10 different languages and is found to address the relevant domains of male sexual function. 60 Overall, the efficacy of all 3 agents is similar, with about 70 80% of men reporting that the PDE5 inhibitors produce an improvement in erections and about 75% of men being able to have intercourse successfully. 50,53 In addition, studies have demonstrated that PDE5 inhibitors are also effective in specific subpopulations, such as those with ED and stable cardiovascular disease, diabetes mellitus, depression, spinal cord injuries, or who h a ve undergone radical prostatectomy or renal transplantation. 51,61 Current treatment guideline s recommend that patients should not be deemed true treatment failure s unless they have tried the maximum dose of the drugs for at least eight continuous occasions. 61 In addition, patients who were previously non responding to a PDE5 inhibitor may become successful after switching to either another PDE5 inhibitor , a regular dosing schedule, or by increas ing the dose.
30 The selectivity of an inhibitor for PDE5 is the key determinant of the potential side effects of PDE5 inhibitors. Systematic review of clinical trials and FDA AER S records show s that all 3 PDE5 inhibitors are generally well tolerated and have similar contraindications and warnings. 9,53,62 The most commonly reported class specific side effects include flushing, headache, dyspepsia, gastrointestinal upset, nasal congestion, and myalgia. 9,62 These are generally mild and of little long term consequence. Rare post marketing adverse reports of non art eritic anterior ischemic optic neuropathy (NAION) causing sudden vision loss in patients taking PDE5 inhibitors led the FDA to request a labeling change for all three products in 2005. 63 D ouble blinded, case crossover trials have been conducted to evaluate patient preference s for the three PDE5 inhibitors. After 12 weeks, patients were crossed over to the other treatment. Following another 12 weeks, the patients made a blinded choice of which of the two t these trials were rather conflicting; some studies suggesting that tadalafil has the best patient preference; others failed to demonstrate significant superiority between th e three agents. 51,64 Sudden Sensorineural Hearing Loss (S S HL) Risk Factors of S S HL In sensorineural hearing loss (SNHL) , the damage to the hearing organ takes place in the inn er ear (cochlea), nerve pathways that help transmit the sound messages from cochlea to the brain, or both. P atients with SNHL usually complain that they have difficulty understanding what is being said to them when there is a lot of background noise. Other signs and symptoms include difficulty hearing high pitched sound, experience of dizziness or a ringing sound in their ears .
31 Common cause s for SNHL include natural aging, exposure to excessive loud noise, congenital deafness, exposure to some medications (i.e., loop diuretics, aminoglycoside, NSAIDS, chemotherapies such as cisplatin, vincristine, and vinblastine), illnesses such as infectio n (i.e., syphilis, mumps, measles, CMV, HIV, varicella zoster, Lyme disease, rubella, Epstein Barr virus), meningitis, cancer, vascular problems (i.e., smoking, diabetes mellitus, hyperlipidemia, stroke), neurological conditions such as multiple sclerosis, and head injuries (traumas such as skull fractures). 65 In most cases, SNHL cannot be medically or surgically corrected. SNHL is the mo st common type of permanent hearing loss and significantly affects the quality of life for subjects who experience the condition. When SNHL is caused by natural aging or ongoing exposure to loud noise, it comes on gradually; sudden onset of SNHL can be cau sed by head injury or ototoxic drug s . Table 2 3 summari z es the risk factors for SNHL. Sudden sensorineural hearing loss (SS HL), or sudden deafness, is defined as a hearing loss that is greater than 30 decibels (dB) in three connected frequencies that occu rs in less than 3 days. 23 It i s estimated that SS HL affects 5 to 20 per 100,000 population, with about 4,000 new cases per year in the US. 22 Most SS HL occurs within minutes to several hours. Hearing loss affects only one ear in 9 out of 10 people who experience SS HL. SS HL is a frightening symptom and is considered a n otologic emergency that requires immediate medical attention. Some patients may recover completely without medical intervention, often within the first 3 days. Others may get better slowly over a 1 or 2 week period. Ab out 15 percent of those with SS HL experience worsening hearing loss. 23
32 The pathophysiology for SS HL includes 4 possible causal pathways. 22,66 First, many viral infections including cytomegalovirus, mumps, rubella, HIV, varicella zoster, and EBV can be the cause for SS HL. The viral infection or viral reactivation within the inner ear causes cochlear inflammation and/or damage to critical inner ear structures. Second, cochlear function is extreme ly sensitive to changes in blood supply. Vascular compromise of the cochlea due to thrombosis, embolus, reduced blood flo w, or vasospasm seems to be a likely etiology for idiopathic SS HL. A vascular cause for SS HL has long been presumed because of its sudden or abrupt clinical course and accompanying symptoms. Third, intra cochlear membrane rupture can produce a leak of peri lymph fluid into the middle ear and thus allow mix ing perilymph and endolymph, affecting endocochlear potential . Fourth, disorders affecting the immune system such as Cogan syndrom e may have a role in causing SSHL. SS HL has been known to occur in establish ed autoimmune disease s such as rheumatoid arthritis and diabetes. Finally, certain ototoxic drugs such as aminoglycoside antibiot ics can be the cause for SS HL. 22,23 Diagnosis an d Treatment for SS HL D octor s can determine whether a person has experienced S S HL by conducting a standard hearing test. The evaluation of the patient with S S HL includes a review of his or her medical history and symptoms, an otologic and neurologic examina tion, audiologic testing, and laboratory studies. Most commonly, S S HL affects the high tones, with relatively good hearing at the lower frequencies. S S HL is confirmed using a battery of audiologic tests. A comprehensive hearing assessment usually includes acoustic admittance measurements such as acoustic reflex test, acoustic reflex decay, otoacoustic emission (OAE), auditory brainstem response (ABR), tympanometry, and
33 behavioral audiometry testing including pure tone audiometry, air and bone conducti on tes ting, and speech audiometry. 67 No treatment has been proved absolutely effective for SSHL . Thus, treatment regimens can be varied , including vasodilators, anti inflammatory agents, antiviral agents, and hyperbaric oxygen. 68 Oral steroids treatment is the current standard treatment for SSHL. Current data suggested that intratympanic steroid injection might offer another treatment option for SSHL. Some otolaryngologists believe this will reduce side effects by avoiding systemic exposure, although the cost for intratympanic steroid injection is higher. 23,69 P DE5 Inhibitor Use and Risk of S SHL Critical Appraisal of Existing Evidence The PubMed database and FDA website were searche d for reports of PDE5 inhibitor and hearing loss published since 2007 . After reviewing the contents, two reviews of premarketing RCTs , 8,16 two animal stud ies, 70,71 five case reports or case series , 15,20,72 74 one cross sectional database analysis , 75 and a n FDA newsletter 21 were found for in depth analysis. Table 2 4 summarizes the major findings from the se studies . Giuliano et al. reported a large review of 67 double blinded placebo controlled (DBPC) trials and post marketing safety data from t he manufacturer of sildenafil citrate (Pfizer internal database) . The DBPC database contained more than 14,000 men who took at least one dose of sildenafil. Only one case of mild hearing loss was reported in these trials. The patient had prior history of h earing impairment, which would be expected to be a risk factor for the current episode. In the post marketing database, the reporting rate for sudden hearing loss and impaired hearing was 0.01% (3/39,277) and
34 0.07% (26/39,277) respectively. Of note, this f igure may be underestimated because many patients suffering from sudden hearing loss may recover spontaneously and it has been noticed that AE outcomes deemed to be more severe (i.e., cardiovascular outcomes and death) may be reported with greater frequenc y than less severe AEs . 9 Another issue is that most cases also had other risk factors that are known to predispose for hearing loss. 8 Prompted by the fi rst case report, the FDA independently reviewed RCTs data for the 3 PDE5 inhibitor agents. The incidence of patients experiencing sudden hearing loss from these clinical trials was about 5/25,000 for sildenafil, 3/16,000 for vardenafil, and 4/18,000 for ta dalafil. 16 The ototoxic effect of sildenafil was further evaluated in a mouse model, demonstrating dose response hearing im pairment over time in mice treated with sildenafil. 70 Auditory function tests were performed on mice that were orally administered high doses of sildenafil for up to 135 days. Hearing loss was evaluated by recording auditory middle ear latency responses and otoacoustic emissions. High dose of sildenafil incr eased hearing thresholds as measured by auditory brainstem responses. The authors suggested that high dose and long term sildenafil administration can induce hearing impairment in mice. 70 In total, 5 case reports or case series provide sufficient information and are included in the analysis. In 3 of the five reports, only grouped data was presented. 20,73,74 After excluding the duplicate cases, a total of 53 cas es were reviewed. The summary of these 53 cases is presented in Table 2 4 .
35 T hirty (57%) of the 53 cases of PDE5 inhibitor associated SNHL were found in the AERS database of the US FDA. Other countries that contributed to the case series include Germany (n= 8), United Kingdom (n=8), Canada (n=5), Australia (n=1), and India (n=1). The se patients were between 36 and 85 years of age, with a male to female ratio of 49:4. Only four cases in the AERS database were treated with PDE5 inhibitors for pulmonary arterial hypertension ( PAH ) , in the remaining cases, the drug was used to treat ED. No detailed information on potential alternative causes of SNHL w as reported in the other four case reports of PDE5 inhibitor induced SNHL , 20,73,74 , 15 although some risk factors (smoking, hypertension, diabetes, hyperlipidemia) were present ed by the authors in one case report ( 57 year old m a n who had audiologic confirmed unilateral sudden SNHL after exposure to vardenafil ) . Use of s ildenafil was most frequently linked to the development of S S HL (n=33; 58%). Tadalafil and tadalafil were implicated and suspected to be associated with S S HL in 12 ( 21%) reports each . Only 4 (10%) patients had bilateral hearing loss. The interval between the last dose of a PDE5 inhibitor and the onset of hearing loss was documented in 32 patients (60%). Hearing loss occurred within 24 hour s of PDE5 inhibitor ingestion in 75% (24/32) of these patients, although the lag times between exposure and onset were reported to be as long as 18 months. 74 . Information on prognosis was documented in 14 patients. Thirteen patients r eported a complete resolution or significant improvement in hearing. C omplete and persistent deafness was recorded in only one patient .
36 After November 2007, the raw FDA data yielded an additional 223 logged reports submitted after the public advisory was released. 74 Although little detail is available about these cases , the FDA Drug Safety Newsletter did record two patients who had a positive re challenge following the initial episode. 21 Until now, only one epidemiologic study on PDE5 induced hearing loss has been conducted using a sample of men 40 years and older from the 2003 2006 Medical Expenditure Panel Survey (MEPS) . 75 P articipants of the survey provided information on medication use inclu ding PDE5 inhibitor in a total of five interviews. Hearing loss was self reported. The study found that the overall incidence of hearing impairment was 17.9% and increased with age. Two percent of the subjects had taken a PDE5 inhibitor, 80% of them used s ildenafil. In this study, m en who reported having hearing impairment ha d a 2 fold increased odds of us ing PDE5 inhibitor s compared to men without hearing impairment , after adjusting for demographics and other potential confounders. T he study found the asso ciation was confined to sildenafil only; no meaningful relationship to hearing impairment was seen with vardenafil and tadalafil. 75 However, as acknowledged by the author s , t his might be due to sample size issues . 75 Of note, a lthough th is study included a large population of PDE5 inhibitor (especially sildenafi l) users , the cross sectional desig n is rather weak for causal inference . Other limitations of the study include the unknown validity of the self reported hearing impairment, recall bias for drug exposure, and residual confounding. Possibl e Mechanisms of P DE5 Inhibitor Induced S S HL Based on diagnostic information in the case reports, SNHL caused by PDE5 inhibitors appears to be predominantly cochleotoxic leading to hearing loss and tinnitus. H earing loss appears to be unilateral and mild to moderate in severity . Reversibility of
37 the hearing loss appears inconsisten tly in the published case reports. Although some patients recovered completely, others experienced no improvement. Among those whose hearing recovered, time to recovery ranged from a few days t o weeks after cessation of the medication. As such, the potential mechanism of PDE5 inhibitor induced S S HL is thought to be more related to biochemical or metabolic changes in the cochlea rather than to morphologic abnormalit ies . 76 Especially in light of the limited evidence, biological plausibility of a causal association between PDE5 inhibitors and S S HL is an important consideration . One conceivable mechanism for the ototoxi c effect may be a pathway related to nitr ous oxide (NO) and cGMP mediated hearing loss. NO/cGMP is critical in erection health, and NO has been implicated to cause a variety of otologic disease s including hearing loss associated with bacterial meningitis 17 , gentamicin induced hearing loss 18 , and cisplatin ototoxicity. NO is produced by endothelial nitric oxide synthase or inducible nitric oxide synthase (i NOS ) . NO has been measured in the auditory nerve, lateral well, and neuroepithelium of the guinea pig cochlea. 77 NO can lead to inner ear dysfunction and inhibition of i NOS lowers the cochlea damage in animal models. 78,79 In addition, increased production of NO has been demonstrated in animals with hearing loss caused by drugs. It is therefore biologically plausible that PDE5 inhibitors might increase the risk of S S HL through the mediation of NO. In summary, although cause and effect remains unproven based o n current lines of evidence, an association between PDE5 inhibitor therapy and S S HL in patients taking these drugs is plausible .
38 Table 2 1 . Causes of Erectile D ysfunction Disease Lifestyle factors Cigarette smoking, heavy alcohol drink, inactive lifestyle Psychological disorder Anxiety, depression, stress Vasculargenic Hypertension, coronary arterial disease, peripheral vascular disease, kidney disease Endocrine disease Diabetes mellitus, obesity Neurologenic Multiple sclerosis, spinal cord injury, disease Prostate cancer, bladder cancer treatment Radical prostatectomy, pelvic radiotherapy Nephology Lower urinary tract symptoms Medication use Antihypertensive medications (beta blockers, calcium channel blockers, thiazide diuretics), antipsychotics, antidepressants * Modified from 3 reference papers 22,23,41
39 Table 2 2. Pharmacodynamics and Pharmacokinetics of PDE5 I nhibito rs Sildenafil citrate Sildenafil citrate Vardenafil hydrochloride Tadalafil Drug name Viagra (Pfizer) Revatio (Pfizer) Tadalafil (Lilly) Cialis (Bayer) Indication Erectile dysfunction (ED) Pulmonary arterial hypertension (PAH) Erectile dysfunction (ED) Erectile dysfunction (ED) Available doses (mg) 25, 50, 100 25, 50, 100 5, 10, 20 5, 10, 20 FDA approval March 27 th , 1998 June 19 th , 2005 August 19 th , 2003 November 21 st , 2003 Dosing 50 mg initial dose, range 25 100 mg Max: 20 mg, tid 10 mg initial dose, range 5 20 mg 10 mg initial dose, range 5 1 20 mg Special populations Use low dose (25 mg) in elderly and consider in those with hepatic/renal dysfunction N.A. A 5mg starting dose should be used in elderly patients and those with mild to moderate hepatic or severe renal impairments Dose adjustment is not necessary in elderly patients Time to take before sexual activities About 60 minutes N. A. 30 60 minutes 30 minutes Duration of action 4 6 hours 4 6 hours 4 6 hours Up to 36 hours Protein binding 96 96 95 94 C max 31.8 31.8 378 327 T max , hours Mean 1 hour Mean 1 hour Mean 1 hour Mean 2 hours T 1/2 , hours 3 5 hours 4 hours 3 5 hours 17.5 hours Metabolism CYP3A4 CYP3A4 CYP3A4 CYP3A4 Percent excreted in faces/urines 80/13 80/13 92/5 61/36 Footnote: C max : Peak plasma concentration after administration; T max : Time to reach peak plasma concentration; T 1/2 : Elimination half life * Modified from 3 reference papers 12,61,80
40 T able 2 3 . Risk Factors for Sudden Sensorineural Hearing L oss Type Example Viral infection Viral cochleitis associated with herpes viruses, parainfluenza virus, influenza, mumps, measles, rubella, or HIV; bacterial meningitis; Mycoplasma pneumoniae infection; Lyme disease ; tuberculosis, syphilis, or fungal infection Neoplasms Acoustic neurinoma; meningeal carcinomatosis; lymphoma, leukemia, or plasma cell dyscrasia Trauma Head injury, barotraumas; noise exposure Autoimmune disease Autoimmune inner ear disease; Cogan's syndrome; Susac syndrome; systemic lupus erythematosus; antiphospholipid antibody sydrome; rheumatoid arthritis ; SjÃ¶gren's syndrome; relapsing polychondritis; vasculitides (polyarteritis nodosa, BehÃ§et's disease, Kawasaki disease, granulomatosis with polyangiitis (Wegener's), temporal arteritis, or primary c entral nervous system vasculititis) Vascular disorder Vertebrobasilar cerebrovascular accident or transient ischemic attack; cerebellar infarction; inner ear hemorrhage Drugs Aminoglycosides, vancomycin, erythromycin, loop diuretics, antimalarials, cisplatin, sildenafil, cocaine Other causes Meniere disease, otosclerosis; Paget's disease; multiple sclerosis ; sarcoidosis; hypothroidism; idiopathic SNHL * Modified from 3 refer ence papers 32 34
41 Table 2 4 . Case Reports of PDE5 Inhibitor Use and SSHL No. of patients Age range, years Sex Drug Dose Duration of use Indication Time to onset a Laterality b 1 44 M Sildenafil 50mg 15 days ED <24 hours Bilateral 25 ED: 44 85 PAH: 36 63 M (22) F (3) c Sildenafil (17) Vardenafil (3) Tadalafil (6) d NA NA ED (21) PAH (4) <24 hours in 15/17 patients Unilateral (24) Bilateral (1) 4 45 65 M Vardenafil 10mg 1 st dose ED <24 hours Unilateral 1 57 M Vardenafil 2.5mg 1 st dose ED <24 hours Unilateral 22 f 37 72 M (18) F (1) NA (3) Sildenafil (15) Vardenafil (4) e Tadalafil (6) e NA NA ED <24 hrs (3) <2 wks (2) NA (13) Unilateral (8) Bilateral (2) NA (12) PDE5, phosphodiesterase type 5 ; SSHL, sudden sensorineural hearing loss; AERS, adverse event reporting system; ED, erectile dysfunction; PAH: pulmonary arterial hypertension; NA, not available a: Defined as time between onset of S S HL and date for the last pill ingestion b: Unilateral, bilateral, or unknown c: Al l three female patients were treated by sildenafil for PAH d: One patient used both sildenafil and tadalafil to treat ED e: One patient used both vardenafil and Sildenafil; two patients used both sildenafil and tadalafil f: Include only the 22 cases from Canada, Germany, United Kingdom, and Australia
42 Table 2 4. Continued Audiology confirmed Prognosis Comorbid condition Reference Yes Profound hearing loss; no improvement No Mukherjee et al. (2007) 15 NA Complete resolution (5) Partial improvement (3) NA Maddox, et al. (2009) 20 Yes Complete resolution NA Okuyucu, et al (2009) 73 Yes Complete resolution CKD, Type 2 diabetes, hypertension, hyperlipidemia Snodgrass et al (2010) 72 NA NA NA Khan, et al (2011) 74 PDE5, phosph odiesterase type 5 ; SSHL, sudden sensorineural hearing loss; AERS, adverse event reporting system; ED, erectile dysfunction; PAH: pulmonary arterial hypertension; NA, not available a: Def ined as time between onset of SS HL and date for the last pill ingestion b: Unilateral, bilateral, or unknown c: All three female patients were treated by sildenafil for PAH d: One patient used both sildenafil and tadalafil to treat ED e: One patient used both vardenafil and Sildenafil; two patients used both sildena fil and tadalafil f: Include only the 22 cases from Canada, Germany, United Kingdom, and Australia
43 Figure 2 1 . Mechanism of Action of PDE5 Inhibitors . (Reprint permission obtained from Nature Review Drug Discovery ).
44 CHAPTER 3 METH O DOLOGY This chapter first provides an overview of study design and data sources to be used for this thesis. It is followed by a description of methods f or the 3 study parts. Detail of the methods include s identification of study cohorts, def inition of expo sure, outcome, covariates, and plan for statistical analysis . This investigation is approved by the Institutional Review Board (IRB) of University of Florida and a data use agreement with the Rutgers University is in place. Special measures are undertaken to ensure the data security and integrity . All patient data is de identified to ensure full protection of confidentiality of study subjects. Overview of Study D esign Medical service and pharmacy claims data were obtained from the MarketScan Â® Commercial C laims and Encounters (CCAE) database, which compiles claims data from private health insurance plans of large employers in the US. During the study period (1998 2007) , three PDE5 inhibitors (sildenafil, varde nafil, tadalafil) were commercially available fo r on demand treatment of erectile dysfunction (ED) . PDE5 inhibitors are considered to be a ; because of controversial opinion about the medical necessity of ED treatment and concern about rising pharmaceutica l cost of managing ED , 81,82 health plans usually restrict the reimbursed supply to a maximum of 4 8 table t s per month. 81 83 R ecent research suggest s potential be nefits of continuous use of PDE5 inhibitor as a mean s for the management of ED caused by nerve sparing radical prostatectomy 84 86 and lower urinary tract syndrome (LUTS) related to benign prostatic hyperplasia (BPH), 87 89 and health plans usuall y do not impose restrictions on the number of tablets allowed per
45 prescription for patients with LUTS or in men who have been treated for prostate cancer. The FDA approved sildenafil (20mg tablet) to be used to treat pulmonary arterial hypertension (PAH) i n June of 2005. According to the approved drug label, patients with PAH are instructed to take sildenafil three times daily. 21 In Part 1 , a retrospective claims analysis was conducted to explore the pattern of PDE5 inhibitor use among patients enrolled in large employer based health insurance programs. 90 Annual prevalence and incidence of PDE5 inhibitor use were estimated. In stratified analyses, the prevalence and incidence was estimated in subgroups defined by age (18 29, 30 39, 40 49, 50 59, and 6 0 64 years) . While estimates of prevalence and incidence give an indication of the number of patients currently taking or starting PDE5 inhibitors, these estimates do not describe actual patterns of PDE5 inhibitor use at the daily level. Additional paramet ers including number of refills , distinctions between on demand versus chronic use, switching to another PDE5 inhibitor agent , and dose titration were also estimated. In Part 2 , a population based, incident user cohort study was undertaken to assess the association between PDE5 inhibitor use and risk of S S . The retrosp ective cohort design is considered appropriate for the study objective because it allows a concurrent control group and allows estimation of the absolute risk as well as risk difference , which are important parameters for risk assessment and risk management . Furthermore, b ecause treatment in our study population was not randomly assigned to patients by investigator s as in randomized trials , differences in baseline patient level characteristics in treatment groups that are also associated with the outcome can bias the effect estimates of treatment . 91 T h e proposed s tudy is more
46 vulnerable to this con founding bias, because ED and SS HL share many common risk factors ( e.g., vasculargenic and neurogenic causes , me dication use ). In order to adjust for confounding , we used propensity score ( PS ) adjustment to account for baseline differences between treatment groups. 92,93 Because the PS method model s the probability of exposure, the temporality of a cohort design (first exposure then ou t come) makes it logically and conceptually straightforward to con struct the PS model in a cohort stud y framework . 94 The incident user design is recommended for comparative effectiveness and safety studies of pharmaceutics . 95 The design mimics randomized trials by comparing ding under ascertainment of study events that could where patients susceptible to a drug side effect may drop out early because they suffer from event s early in the course of therapy, whereas those remaining in the study cohort (i.e., prevalent users) are actually less susceptible to the study outcome . Hence, studies that include only prevalent users may underestimate the drug effect. In addition, the incident user design ensures that baseline covariates are ascertained before treatment initiation, which avoids accidental conditioning on covariates that are in the causal pathway, thus leading to a bias towards masking a true positive association. 91,94 In Part 3, an incident user cohort design was u sed to study the r isk of SS HL comparing ne w users of three PDE5 inhibitors. Since an active drug (sildenafil) w as used as the reference group in this study p art, this active control design can help reduce confounding by indication further . 96
47 Data S ource an d Data File s L inkage The study cohorts were established from the Thomson Reuters MarketScan Â® Commercial Claims and Encounters (CCAE) database. The computerized CCAE database is an insurance claims database on approximately 15 million unique patients that include s active employees, early retirees, and their dependents who are covered under a varie ty of employer sponsored health insurance programs (i.e., preferred provider organizations, health maintenance organization, point of service plans with or without capitation). The MarketScan CCAE database represents more than 100 payers within the US. For this thesis, data is available on enrollees for a 10 year period (1998 2007). The database contains fully adjudicated de identified claims data for inpatient services, outpatient services, outpatient prescription drugs, and monthly enrollment data provide d in the following tables: 1. The enrollment table provides individual level enrollment records (i.e., monthly enrollment indicators) with demographics (i.e., age, gender, birth year, geographic region, industry type, and employment status), and health plan information (i.e., plan type, deductible amount, and copayment amount). In order to protect confidentiality, information on age is limited to the year of birth. 2. The inpatient service table contains both facility and physician services associated with an in patient admission. It contains data on date of admission, date of discharge, princip a l diagnosis (codes with the International Classification of Diseases, Ninth Revision, Clinical Modification [ICD 9 CM] codes), secondary diagnosis, procedure codes, proced ure code type (ICD 9 CM, CPT, HCPC, NAB SP, UB92 Revenue Code ), and date of service incurred . 3. The outpatient service table includes encounters and claims for services that are other outpatient facilities. Like the inpatient service table, the outpatient service table c ontains data on date of service, diagnoses codes, and procedures. 4. The outpatient pharmaceutical claims table contains data on prescription drugs dispensed from community based retail pharmacies and mail order pharmacy services, including the drug name def ined by National Drug Code (NDC), date of dispensing, quantity dispensed (i.e., refill number), dosage form, and days of supply dispensed (i.e., 30 or 90 days of supply). In the outpatient pharmaceutical
48 claims table, there is no field to indicate if the p rescription is pro re nata (take as needed). Drugs not reimbursed by the payer (e.g., purchased on the internet or available over the counter) are not recorded in the outpatient pharmaceutical claims table. Each individual data table in the CCAE databas e includes a unique encrypted identifier for each patient. All data tables are linked using this encrypted identifier. As data comes from a number of different health insurance programs, they undergo a series of cleaning procedures and quality checks to en sure a standardized format. Advantage s of using the MarketScan CCAE data for the investigation proposed in this thesis include coverage of a large insurance population from 1998 (when Viagra was first introduced to the ED marketed) to 2007 (when the FDA a nnounced the labeling change for all PDE5 inhibitors due to the potential increased risk of S S HL) ; coverage of a large proportion of adults who may consider use of PDE5 inhibitors for management of ED ; and coverage of private pharmacy benefit organizations , many of which including PDE5 inhibitors in their formulary (in contrast to public insurance programs, which do not) . T he CCAE database was previously used in a drug utilization study that examined the preference and adherence to treatment among ED patients following radical prostatectomy (RP) or radiotherapy for localized prostate cancer. 97 After receiving MarketScan CCAE data files from the Ru tgers University, we carried out further data cleaning and data integrity check. Specifically, we checked for duplicate patient IDs in the Annual Summary Enrollment Tables. We found no duplicate records in these files. In addition, we checked missing data and invalid values of key variables to get an initial image of the quality of the MarketScan data. We found the data sets have fairly low percent of missing data and invalid values. Finally, we counted the volume of in and out patient claims, prescriptio n drug claims by calendar year and
49 by types of health plan (e.g., exclusive provider organization, health maintenance organization, non capitated point of service, preferred provider organization) in the MarketScan CCAE data. A capitated provider is at hig her financial incentive for patient referral, and patients may receive suboptimal care through underutilization of services. Thus, the completeness and validity of claims data submitted by capitated providers might be questionable. We found there is no evi dence of underreporting that is associated with a particular type of health plan. Part 1 : Patterns of PDE5 I nhi bitor U se in a Commercially Insured Study Population The base population for Part 1 consists of male enrollee s aged 18 to 64 years who were co ntinuously enrolled for at least 6 months in a health plan that contributed data to the MarketScan Â® CCAE database from January 1, 1998 through December 31, 2007 . PDE5 inhibitor c laim s are abstracted from outpatient pharmac y table s by using the National Drug Index (NDC) codes. PDE5 inhibitor s that are ordered through internet or free drug samples are not included in the CCAE database. Data A nalyses Prevalence and incidence of ED Using the in patient and out patient service tables, p atients with ED were identified based on the presence of at least one in patient or outpatient medical service claim with a primary or secondary diagnosis for ED (ICD 9 CM codes 607.84 (organic impotence) or 302.72 (psychogenic impotence ). Calculation of prevalence and inciden ce of ED was defined calendar year specific. To be included in the population for the prevalence estimate in Year X (i.e., the denominator of the prevalence estimate), the patient had to be continuously enrolled for at least 6 months during Year X. To be i ncluded as a case
50 of ED in Year X (the numerator of the prevalence estimat e ), the subject had to have at least one in patient or o utpatient medical service claim with ICD 9 r was then calculated as the numerator divided by the denominator, as defined above, and expressed as the number of cases per 1,000 subjects. The 95% confidence interval (CI) was calculated assuming a binomial probability distribution. To be included as an incident ED case (the numer ator of the incidence estimate ) in Year X, the subject had to have at least one inpatient or outpatient ED diagnosis in Year X that was preceded by at least 365 days continuous enrollment without a diagnose code for ED. The deno minator for the incidence estimate includes person time during year X for all subjects who were eligible in January of the measurement year and who had 12 months continuous enrollment in the previous year without a diagnosis of ED. The incidence of ED , expressed as the number of cases per 1000 subject years) was then calculated as the sum of incident cases divided by the sum of the months subjects were enrolled during the measurement year . We required an incident case to have at least 12 months continuou s health plan enrollment preceding their first ED diagnosis in a measurement year, we cannot estimate the incidence of ED for calendar y ear 1998. In secondary analyses, the annual prevalence and incidence of ED was measured in subgroups defined by age cate gories. Age was measured at the beginning of each calendar year and all variables were retrieved from the member annual enrollment file.
51 Prevalence and incidence of PDE5 inhibitor use Estimations of the annual prevalence and incidence of PDE5 inhibitor us e are also calendar year specific. From the base population as described above , all males between the age of 18 64 years who were enrolled in a drug benefits program as indicated by the program eligibility file were eligible for th e analysis. Because we re quire d incident user s to have no PDE5 inhibitor dispensing record for at least 12 months before their first PDE5 inhibitor prescription, we cannot estimate the incidence of PDE5 inhibitor use for Year 1998 . To be included in the population for the prevalence estimate in Year X (i.e., the denominator of the prevalence estimate), subjects had to have at least 6 months continuous enrollment in a calendar year of interest and be between 18 64 years at the beginni ng of that year . To be consider ed a current user of PDE5 inhibito r s (the numerator for the prevalence es timate ) in Year X , individuals had to have filled at least one prescription for a ny PDE5 inhibitor in that calendar year of interest . The prevalence of PDE5 inhibitor use in a given calendar year was then calculated as the numerator divided by the denominator, as defined a bove . Individuals included in the denominator for the incidence calculation include all individuals who satisfy the same eligibility c riteria as described above and were continuously enrolled in January of the measurement year and the entire preceding year where the preceding year had no PDE5 inhibitor dispensing records . The n umerator for the incidence calculation include s incident PDE5 inhibitor users defined as subjects whose first PDE5 inhibitor claim in a given calendar year was preceded by at least 12 months of continuous eligibility without a prescribed PDE5 inhib itor claim. The incidence of PDE5 inhibitor use was then calculated as the sum of all incident users divided by the
52 sum of all months patients were enrolled during the measurement year, expressed as the nu mber of cases per 1,000 subject years . The annual incidence of use was estimate d for overall and for i ndividual PDE5 inhibitor s . In secondary analyses, the annual prevalence and incidence of PDE5 inhibitor use was measured in subgroups defined by age categories. During the study period , the three PDE5 inhibitors were approved to treat ED on an as needed (PRN) basis. D aily dos e sildenafil was approved to treat pulmonary arterial hypertension (PAH) in 2005 . Routine dos age of PDE5 inhibitor s has been investigated in more recent years a s a mean s for long term management of PAH, 98 lower urinary tract symptoms (LUTS) related to beni gn prostatic hyperplasia (BPH), 87,99 penile rehabilitation after prostatectomy, 84 86 100 and urinary i ncontinence . 101 Possible indication s for PDE5 inhibitor use were identified among new users of PDE5 inhibitors (i.e., no PDE5 inhibitor filled in the 12 mon ths before the index prescription in the study period ) . Medical service e ncounters occurring during the 90 days before the index PDE5 inhibitor prescription were searched to identify outpatient o r inpatient encounters with ICD 9 CM diagnoses for conditions frequently treated with PDE5 inhibitors. The list of potential diagnoses was developed a priori based on a review of FDA approved drug labels, 54 56 and listed indications in the Clinical Pharmacology as well as published review article s. 7,50,102 We include d only indications that are either FDA approved or have evidence for efficacy fr om a randomized controlled trial. Th e final list of indication included ED ( 607.84 , 302.72 ) , prostate cancer ( 185.x ) , pulmonary hypertension ( 416.0 , 416.8 , 416.9 ) , LUTS/BPH ( 222.2 , 600.x ) , and ( 607.81 ) . Subjects without any such indication in the
53 90 days preceding the incident pharmacy dispensing record were classified as having an unknown indication. Longitudinal pattern of PDE5 inhibitor use To gain insi ght into longitudinal pattern of PDE5 inhibitor use, the total number of refills, total number of tablets dispensed , and the percent age of subjects who switched to a different PDE5 inhibitor agent or to a different dosage form during the first 6 month s after the in dex PDE5 inhibitor prescription were es timated. The date of first use of a PDE5 inhibitor preceded by 12 months continuous enrolment for each subject was flagged . All subsequent PDE5 inhibitor prescriptions filled by the subject were then used to analyze longitudinal pattern of use . In the prim ary analysis, a 6 month follow up period was used . This is because previous utilization studies suggest that the percent age of patients who discontinue or switch to another agent was highe st during the first few months after a patient initiated PDE5 inhibitor treatment . 103 In the secondary analy se s, we estimate d the se parameters in subjects who had at least 12, 18, or 24 months of continuous follow up . Thus, prescription refill was measured as the number of subjects who refilled at least one PDE5 inhibitor prescription during the 6 months post index date period divided by the total number of eligible subjects who me t the cohort inclusion criteria. The medication switching proportion was calculated by dividing the number of subjects who had switched to another PDE5 inhibitor during the 6 months period after the index date, by total number of eligible subjects. The pro portion of patient with dose titration was defined as the proportion of subjects who had increased or decreased their dose during the 6 months period after the index date, divided by total number of eligible subjects.
54 Part 2: PDE5 Inhibitor Use and Risk o f S S Study Design and Population For p art 2 of the thesis we employed a retrospective cohort study to evaluate the risk of S S HL comparing PDE5 inhibitor use with nonuse . For exposed group, we included all male beneficiaries who were new users of PDE5 inhibitors and also enrolled in a health plan that contributed data to the MarketScan CCAE databases between 1998 and 2007. The used new user design to avoid underestimation of outcomes that might have occurre d soon after initiation of therapy, 91 which is important because most cases rep orted to AERS system had a sudden onset of hearing loss that was usually within 1 3 days after initiation of PDE5 inhibitor treatment . 16 To be considered as a new user of PDE5 inhibitor s , we first identified all patients who filled a prescription for any PDE5 inhibitor after not having filled a PDE5 inhibitor medication in the preceding 183 days (index date ) . We retained only subjects who had 183 days of continuous enrollment in both health plan and drug benefits programs before the index date in order to ensure a complete ascertainment of all baseline comorbid conditions and prior drug use. For this study, w e included only the first qualifying index PDE5 inhibitor prescription for each patient, even though a patient may have multiple index prescription episodes that all satisfied the cohort inclusion criteria, because past experience of using PDE5 inhibitor c and the occurrence of outcome in the f uture. 91,95 Patients <18 years or >64 years at the index date were excluded, as were patient who had been diagno sed with comorbid conditions that can cause SS HL . These conditions were measured during the 183 day washout period and included hearing loss of any form (conductive, sensorineural, mixed, unspecified), cancer (except prostate
55 cancer and bladder cancer), or gan transplantation, HIV/AIDS, cytomegalovirus infection, rubella, syphilis, bacterial meningitis, viral encephalitis, severe head injury, and head or neck radiation. We removed the original restriction to require eligible patients to have a diagnosis of E D before their PDE5 inhibitor index prescription, because only about a quarter of new users of PDE5 inhibitor in our study had an ED diagnosis before index date . We excluded individuals who received PDE5 inhibitors for pulmonary arterial hypertension (PAH) because the PDE5 inhibitor dosing schedule for these patients is different, 50 and because the pop ulation was too small to allow separate inferences ( 1,016 patients with PAH in our cohort ) . Finally, we excluded patients who had filled prescriptions for other known ototoxic drugs including aminoglycosides, interferon, cisplatin, cyclosporine, vinblastin e and vincristine during the look back period and during any time during the follow up. All inclusion and exclusion criteria along with their operationalization are summarized in Appendix A. A similar definition was applied to the reference group which consisted of all male adults who were not dispensed any PDE5 inhibitor during the study period . For each control, the index date was a random date selected from their continuous enrollment period that was pr e ceded by 183 days health plan membe r eligibility. All other cohort inclusion criteria were identical for PDE5 inhibitor users and controls. Because PDE5 inhibitor users and nonusers were ex pected to differ in various baseline characteristics that may be directly related to the outcome , in t he secondary analysis, we selected users and controls from only individuals who had an ED diagnosis to control for confounding bias.
56 PDE5 Inhibitor Exposure All PDE5 inhibitor prescriptions dispensed by new users between index date and end of follow up wer e included in this analysis. Because PDE5 inhibitor s are lifestyle drug s , 104 health plans implement a strict quantity limit policy , typically allowing a maximum of 4 8 tablets per qualified member per month . In addition, patients cannot elapsed. 81,82,105 B e cause PDE5 inhibitors are used on a PRN or as needed basis, 54 56 patients have compl ete freedom t o decide when and how they would like to take the drug . Thus , exact time and frequency of PDE5 inhibitor exposure is not ascertainable from pharmacy dispensing data . In order to overcome th is limitation, some assumptions needed to be made in t erms of the usual patterns of PDE5 inhibitor use by ED patients in the study. Several authors looked into this issue by conducting surveys among ED patients. O verall, among men who responded to the surveys, most used PDE5 inhibitors 1 2 times per week foll owed by individuals who used it once or twice a month . 106 109 Accordingly , we estimated the duration of PDE5 inhibitor exposure by multiplying the dispensed number of tablets by the estimated frequency of use . For example, in the primary analysis, we assumed that patients used one tablet of PDE5 in hibitor per 7 days. If a patient was dispensed 4 tablets, we computed the intended days of supply to be 28 days. plus an added 30 day grace period . The 30 day grace period was added based on two assumptions: first, PDE5 inhibitor s are s are used by patients to enhance their sexual performance rather than to treat a particular medical necessity , hence patients may not use the ful l monthly supply 82 or may not refill their prescriptions on time . 103 S econd, although SSHL is characterized as a medical emergency with rapid
57 loss of hearing over a short period of time (usually within 72 hours) , 9 out of 10 patients have only unilateral hearing loss and sensation loss range s fro m mild to s evere. 23 Hence, many patients may not seek health care immediately after symptom onset or may first consult their primary care physician . Thus, audiologic exams and final diagnosis of SSHL may be delayed . In fact, a recent study suggested that the average delay in referral for an audiologic and otolaryngologi c exam was 20.8 days in patients with SSHL. 110 After calculating the duration of PDE5 inhibitor use for each prescription dispensed, each week of follow up was classified as exposed or unexposed according to the probability of use. The resulting consecutive drug exposure periods were used to deter mine the time dependent exposure periods indicating whether the patient was treated with PDE5 inhibitor at any given week during the follow up. Subsequently, the classification of exposure was divided into 3 mutually exclusive categories current use, rec ent use, and nonuse, based on the majority of days attributed to each category of exposure in each week of follow up. Current use included the time from date of prescription filling (t 0 ) to the end of the days of supply as determined by the number of PDE5 inhibitor use plus a grace period of 30 days. Recent use was defined as use after current use up to 3 65 days after the end of the current use period. The recent use category was introduced because of the potential to misclassify exposed person time as nonuse person time for drugs that are used intermittently, hence masking a positive safety signal that wo uld be detected otherwise. Furthermore, some PDE5 inhibitor users may split their pills to achieve cost saving s which contribute s further to the exposure
58 misclassification. 111 In this study, if the patient received a new dispensing of PDE5 supply was exhausted, the excess supply was not carried over . The n onuse period was defined as no PDE5 inhibitor use in the past 365 days . Thus, the nonuse period included follow up person time from the control group as well as the unexposed person time f rom PDE5 inhibitors users after their recent use period has elapsed. For this study, switching between different PDE5 inhibitor agents or a change in dosage did not affect current use or recent use designations (Figure 3 1) . I n summary , the following expo sure categories were constructed : Exposure c ategory 1: Current use ( time dependent; 1=yes, 0=no ) Exposure c ategory 2 : Recent use ( time dependent; 1=yes, 0=n o) Exposure c ategory 3 : Nonuse ( time dependent; 1=yes, 0=no ) Study Endpoint The study outcome was defined as the first SSHL diagnosed during follow up. We did not have data on audiologic testing results in the study database. The primary definition of SSHL required at least one in or out patient diagnosis of sensorineural hearin g loss (ICD 9 CM c odes: 389.1x, 389.2x, or 388.2) combined with at least one CPT code for an audiometric hearing test in each of the 30 day window before and after SSHL diagnosis, except for those hearing test s that were undertaken during the first 3 days after SSHL diagnosis. Although the computer based case definition has not been validated previously, we expect the positive predictive value (PPV) of the algorithm to be high because we required very specific procedure codes for audiologic hearing tests th at are used to determine the presence or absence of hearing loss and the type of hearing loss (e.g., behavioral audiometry, otoacoustic emission, auditory brainstem response). 23 Accurate diagnosis of SSHL is essential for proper management of
59 patients with SSHL and it also ensures that treatment is given to patients who meet audiometric criteria for diagnosis. In our definition of SSHL, we required that at least two audiometric tests were performed within 30 to +30 days of SNHL diagnosis, beca use most otologists and audiologists recommend that a follow up audiometric test needs to be ordered 2 4 weeks after diagnosis to evaluate prognosis and effectiveness of treatment. 23 Furthermore, because oral steroids or intratympanic steroid injection are the most popular treatment options for SSHL in North America, 68 in the sensitivity analysis, we required that a confirmed case of SSHL should have evidence of treatment with steroids after SSHL diagnosis. We would expe ct the PPV for this algorithm to improve while sensitivity would be lowered. Given our cohort design, this may reduce study power but decrease systematic bias. Covariates From enrollment files, we defined demographic variables including age, region, type o f insurance, and calendar year at the index date. Other baseline patient characteristics were defined using medical and pharmacy claims during the look back period. We defined a group of clinical conditions that are risk factors of SSHL. These conditions i ncluded myocardial infarction, atrial fibrillation, ventricular arrhythmia, congestive heart failure, hypertension, perivascular disease, kidney disease, depression, anxiety, alcoholism, and other psychotic disorders; concurrent use of medications such as angiotensin converting enzyme inhibitors, angiotensin receptor blockers, beta blockers, calcium channel clockers, loop diuretics, statins, fibrate lipid lowering agents, antidepressants, and antipsychotics; health service utilization variables such as numb er of hospitalizations, number of prescription drugs, and number of physician office visits. Since age is the major risk factor for SSHL in adults, we adjusted
60 age as a time dependent variable in the regression models. In the sensitivity analysis, to accou nt for confounding by concomitant use of ototoxic drugs that might cause SSHL, we also adjusted time dependent covariates for use of antidepressants, antipsychotics, loop diuretics, and anti neoplastic. Risk factors which are not available in the MarketSca n CCAE data and therefore not measureable, included smoking, body mass index (BMI), physical activity, occupational exposure to noise, and f amily history of hearing loss. Data Analyses We presented the distribution of demographic variables, clinical and drug utilization characteristics between users and nonusers of PDE5 inhibitors. We calculated crude incidence rates of SSHL and incidence rate differences (IRD) with 95% confidence intervals (CIs) as suming a binomial distribution . Cox proportional hazards (PH) models were used to estimate the hazards ratios ( H R) between exposure groups. Cohort members were followed from the index date until they had the first SSHL diagnosis, or were censored because o f loss of insurance coverage, turned 65 years old, or for reaching the end of the study period (December 31, 2007), whichever came first. The reference category for all hazards ratios was person time with nonuse of PDE5 inhibitors. The Cox PH models contai ned PDE5 inhibitor exposure, the logit transformed propensity score (PS), and age (continuous) as a time dependent variable. We used logistic regression to calculate the exposure propensity score. 92,112 The propensity score is the estimated probability of initiating PDE5 inhibitor versus not initiating a PDE5 exposure propensity score rather than to adjust for a large group of baseline covariates
61 across exposure groups because our outcome is a rare event relative to the number of covariates need ed for adjustment in the model. Previous theoretical work and simulation studies sugges t that PS methods can improve confounding control particularly when the outcome is rare. 93,112 The use of PS matching, trimming (restriction), or inverse probability of treatment weighting (IPTW) have become popular tools in pharmacoepidemiology. 92 All baseline confounders (demographics, diagnos e s, pharmacy dispensing) described above were included in the propensity score model regardless of any significance threshold. 113 Suc c ess in achieving balance of all covariates was assessed using by inverse propensity score weighing and calculation of the standardized mean difference. 92 As the association between the logit transformed propensity score and the outcome was assumed to be non linear, we used a restrict ed cubic spline with 5 knots placed at the 5 th , 25 th , 50 th , 75 th , and 95 th percenti le to explore the shape of the logit transformed propensity score. 114,115 The proportional hazards assumption was tested by plotting the Schoenfeld residual and parameter estimates against time, and by Kolmogorov supermen test, which showed there was no violation against this assumption (all p values >0.5) . 114,116 All analyses were conducted using SAS statistical package version 9.2 (SAS Institute) . Sensitivity Analyses We performed a range of sensitivity analyses to explore the robustness of our findings. First, we explored the impact of different utilization patterns on the effect estimate by changing the assumed frequency of PDE5 inhibitor use to include the following scenarios : two PDE5 inhibitor doses each week, one PDE5 inhibitor dose every 14, 21, or 30 days. Second, we conducted a propensity score matched cohort
62 analysis. Propensity score matching methods have been used i ncreasingly in the observational comparative effectiveness or safety study of medications. Some theoretical work in this field suggest s that p ropensity score matching excels multivariate model based approaches when the treated and control groups have a fai r amount of overlap , although matching has the limitation of reduced statistical power . 92,117 We used an algorithm of greedy nearest neighbor matching without replacement withi n specified caliper widths (0.05) to form the matched p airs . 118 Third, we defined individuals as new users of PDE5 inhibitor s if they had a 12 (instead of 6 ) month washout period before the index date without filling any PDE5 inhibitor prescription. Fourth, we considered a patient to be a true SSHL case if he had any evidence of being treated with oral steroids or intratym p anic steroid injection within 30 days after SSHL diagnosis. By using specific treatment for SSHL, we expected an increase in specificity and predictive value of t he case identification algorithm . Fifth, confounding by indication is a major threat to study validity in pharmacoepidemiology. 119 In an attempt to address confounding by indication , we required all subjects in nonuse control and user groups to have a diagnosis of ED. By comparing ED patients with and without PDE5 inhibitor, we expected the baseline characteristics would be better balanced across exposure groups. Finally, we did a sensitivity analysis by adjusting for other time de pendent variables for the following me dication use during the study follow up (antidepressants, antipsychotics, antineoplastic s , and loop diuretics). Part 3: Comparative Safety of PDE5 Inhibitors and Risk of S S HL T he three PDE5 inhibitors share the same mechanism of action, similar efficacy an d safety profiles . However, tadalafil has a slower onset and longer duration of action, whereas sildenafil and vardenafil have similar half life, onset and duration of action. 7,120
63 These differences may affect patients preference when choos ing a PDE5 inhibitor , which in turn might produce imbalance in the underlying disease severity and the general risk profiles between initia tors of different PDE5 inhibitors ( i.e., confounding by severity of underlying illness ). Without proper adjust ment for the differences, the effect estimates from a n observational comparative safety stud y may be biased. 121 Study C ohort and Design The study base population consisted of all male beneficiaries who initiated a PDE5 inhibitor therapy during December 2003 through November 2 007. We chose a cohort incident user design to avoid underestimat ion of outcome events that occur soon after initiation of therapy, and to ensure that all baseline covariates collected at cohort entry were not affected by treatment itself. For the present study, several mutually exclusive sub cohorts, each including only new users of a particular type of PDE5 inhibitor identified based on their first PDE5 inhibitor prescribed during follow up, were assembled. To ensure complete ascertainment of baseline com orbid conditions and prior drug use, we required that individuals be enrolled in both health plan and drug benefits program during 183 days before the PDE5 inhibitor initiation. We restricted the study cohort to include all male adults between 18 64 years at the time of their index PDE5 inhibitor prescription (index date). We did not require eligible adults to have a recorded diagnosis for ED because only 25% of new users of PDE5 inhibitor had one ED diagnostic claim prior to their treatment initiation. We excluded subjects with a diagnosis of malignant neoplasm, organ transplantation, bacterial meningitis, viral encephalitis, or use of aminoglycosides or cisplatin. We further excluded individuals who initiated sildenafil (RevatioÂ®) for mana gement of PAH bec ause their PDE5 inhibitor
64 utilization pattern is expected to be quite different from that of ED patients . We found 1,016 initiators of RevatioÂ® who had an indication for PAH. Characterization of PDE5 Inhibitor Exposure Data on medication use was obtained f rom pharmacy dispensing records which include drug name, date of dispensing, quantity supplied, and dosage. The first new prescription filled for a PDE inhibitor agent by each individual determined his exposure group. Initiation was defined as filling a pr escription for a PDE5 inhibitor medication without having filled one in the preceding 183 days (washout period). Subjects were allowed to enter the cohort only once , because prevalent users may be less susceptible e s Because exact date and frequency of PDE5 inhibitor use for individuals was unknown using pharmacy dispensing data in claims, we imputed those data based on hypothesized frequency of use. For instance, in the primary analysis, we assumed that ED patients used one PDE5 inhibitor pill every week. Thus, for a patien t who received 4 pills, we computed his intended days of supply to be 28 days plus a 30 day grace period to refill his prescription . Thus, duration of use was estimated from the day after the dispensing though 30 days after the imputed days of supply. In t he secondary analyses, we altered frequency of PDE5 inhibitor use to other pattern of use. If the patient received a new dispensing of the same medication before the pati ent was dispensed a different PDE5 inhibitor before the supply of the preceding drug ha d been exhausted, he was assumed to have switched therapy and his follow up was terminated at the day of switching .
65 Study Endpoint As in Part 2, the outcome of interest for the comparative safety study was incident SNHL. The same computer case definition was used to identify S S HL cases. In the secondary analysis, we further required that a confirmed case of SSHL had been treated with oral or int ratympanic steroids within 30 days after diagnosis . Baseline covariates Baseline patient characteristics were defined using medical and pharmacy claims during the 183 days before the index date. These included demographic variables such as age, region, an d type of health plan; health service utilization variables such as number of hospitalizations, number of prescription drugs, number of physician office visits; clinical conditions included myocardial infarction, atrial fibrillation, ventricular arrhythmia , congestive heart failure, hypertension, perivascular diseases, kidney diseases, depression, anxiety, alcoholism, and other psychotic disorders; and concurrent use of medications such as angiotensin converting enzyme inhibitors, angiotensin receptor block ers, beta blockers, calcium channel clockers, loop diuretics, statins, fibrate lipid lowering agents, antidepressants, and antipsychotics. Since age is a major risk factor for SSHL in adults and follow up could include several years , we adjusted age as a t ime dependent variable in all regression models. No information was available on smoking, body mass index (BMI), physical activity, occupational exposure to noise, or family history of hearing loss. Data Analysi s We followed each subject from the day foll owing their treatment initiation (index date) until they experienced an outcome event, initiated another PDE5 inhibitor
66 years, or for reaching the end of the study period (D ecember 31, 2007), whichever came first. To control for confounding by indication, we used propensity score adjustment to balance baseline covariates for each comparison with sildenafil (common reference group). The propensity score adjustment was chosen in this study because our outcome was rare relative to the number of prognostic factors that needed to be controlled. We constructed two separate propensity score models using multivariate logistic regression. The first model predicted exposure to tadalafi l versus sildenafil and the other predicted use of vardenafil versus sildenafil. We chose sildenafil as the reference group, because it was the most frequently used PDE5 inhibitor in this cohort. All baseline confounders (demographics, diagnos e s, pharmacy dispensing) descried above were included in each propensity score model. Success in achieving balance on propensity score was assessed using the standardized mean difference. We compared the distribution of demographic variables, clinical and drug utiliza tion characteristics among initiators of different PDE5 inhibitors. We computed crude incidence rates and incidence rate differences with 95% confidence intervals (CIs) in the three exposure categories. We used Cox proportional hazards models to estimate t he hazard ratio s comparing the risk for SSHL between exposure groups. Similar to Part 2, the final outcome models contained the PDE5 inhibitor exposures, logit transfor m ed propensity score, and age (modeled as a continuous variable). We tested the proportional hazards assumption by including interaction terms between exposure and study time, and we found no violation of the proportionality assumption.
67 Sensitivity Analyses To test the robustness of our findings, we performed sensitivi ty analyses by altering the frequency of PDE5 inhibitor use. In addition to the primary analysis, we further tested the following patterns of PDE5 inhibitor use: two PDE5 inhibitor tablets per week, one PDE5 inhibitor every 14, 21, or 30 days. Due to the s mall number of SSHL cases identified in the primary analysis, we did not propose more extensive sensitivity analyses involving matching or other restrictions that would have resulted in further loss of sample size . Bias Related Issues During the study pe riod, PDE5 inhibitors are approved only as on demand therapy for the treatment of ED. This increases the opportunity for exposure misclassification in the study since claims data does not have information on when and how frequent patients are taking the se drugs. Furthermore, s imilar to other nonrandomized studies, confounding is a major threat to the internal validity of the study . This section addresses several im portant types of bias that may be present in this study . Exposure misclassification. 25,122,123 In pharmacoepidemiologic studies , computerized pharmacy dispensing records are often used as a proxy measure for actual drug exposure. Because drug dispensing direct ly affects reimbursement and claims are typically closely audited, drug d ata in claims is considered of high quality . 119 However, even for drugs that are us ed to treat a clinical condition, exposure misclassification can occur in studies using claims data either due to patient s non adherence to the treatment plan , 119 free drug samples distributed by physicians, 124 and drugs that are used intermittently (e.g., benzodiazepines for episodic anxiety ) . 125 All
68 these potential mechanisms of exposure misclassification could be encountered in our PDE5 inhibitor safety study. In addition, because PDE5 inhibitor s are lifestyle drug s , many insurance plans have restrict ed coverage polic ies to control cost associated with PDE5 inhibitor use , 126 , 122 potentially resulting in out of pocket purchases not captured by claims data. For example, t he BlueCross BlueShield insurance plan allows a quantity of 8 tablets per month for any combination of PDE5 inhibitors for ED patie nts. 127 In UK, is considered to be appropriate for the majority of ED patients. 128 In the scenario whe re PDE5 inhibitor s are taken as needed, it is quite possible that patients would be considered exposed when they are actually not taking the drug. More likely, patie nts may use the remaining stock of pills for an extended period beyond the assumed end of the prescription. In that case, exposure would be overestimated in underestimated after t he end of the prescription when the subjects are considered to be unexposed. This bias would reduce the true difference be tween periods of exposure and non expose and would underestimate any risk due to exposure , regardless of which specific exposure measu re is assessed . We used a range of s ensitivity analyses to explore the influence of different PDE5 inhibitor utilization patterns. Outcome misclassification. Occurrence of S S HL is defined as ICD 9 CM codes 389.1x, 389.2x , and 388.2 at any diagnosis field in either outpatient or inpatient service files . In addition, diagnoses had to be accompanied by claims for audiometric testing within [ 30,+30] days of SSHL diagnosis. The validity of this approach relies on the assumption that (a) all physicians use audiometric testing for diagnosis of hearing
69 loss, (b) audiometric testing is accurate, and (c) both diagnosis and procedure are reliably captured with claims data. SNHL, particularly sudden deafness , is a severe condition that often re sult s in emergency care. However, many SNHL cases are rev ersible or symptoms may develop progressively . Hence, use of claims data to capture SNHL cases may lower the sensitivity of the approach compared to other survey based method s, because p atients may d elay seeking care or recover spontaneously, resulting in no healthcare encounter or no audiometric confirmation. However, the positive predictive value (PPV) of our computer algorithm to detect SNHL might be relatively high , since we require evidence of both medical diagnosis and audiology testing that occur s within a short period of time . A high specificity will allow an unbiased estimation of relative risk, regardless of sensitivity of the outcome definition. 129 Confounding. ED and SNHL share many common risk factors, including age, vascular, neurologic conditions, and certain medication use. These variables are important confounders and need to be properly controlled in non randomized studies in order to have unbiased estimates of the causal effect of treatment. Using multivariate adjustment that include s all measured confounders in a regressi on model is a classic approach to adjust for the confounding. Alternatively, t he propensity score (PS) me thod has become an efficient way to adjust for confounding in nonrandomized studies . 130 The propensity score is a summary score which can be used to balance covariates across treatment groups. 112 The limitation of th e PS method is that it cannot be used to control for confounders that are not measured in the database. 92,130 In this study, w e decided against PS matching in our primary analysis because the matching algorithms often
70 omit a significant amount of data, hence, reducing statistical power and generalizability . 131 Previous st ud ies suggest potential residual confounding when stratification of estimated PS is used to control for confounding. 132 Residual confounding due to unmeasured confounding factors . Using the diagnosis a nd prescription drug information collected in the MarketScan CCAE data, we control for a large number of confounders . Other confounding factors such as smoking, diet, and physical activity are not available in the CCAE data. These unmeasured confound ers ma y bias the risk estimate either towards or away from the null. Without further external data on these unmeasured confounders, from at least a portion of the study participants, it is not possible to adjust for the unmeasured confounders in this study. Thus, we conducted s ensitivity analysis to estimate the impact of unmeasured confounding .
71 Table 3 1 . Baseline covariates : s ocio demographics Variable Names Notes Geographic region Northeast, North central, South, West, Unknown Type of health plan Employer based, large health plan Age, 1 year increment Calendar year 1998 2007
72 Table 3 2 . Baseline covariate s: c omorbid conditions Comorbidity ICD 9 CM codes Notes Cardiovascular disease Hypertensive disease 401 405 Ischemic heart disease M yocardial infarction 410 Acute MI 412 Old MI Other ischemic heart disease 411 Other acute and sub acute forms of ischemic heart disease 413 Angina pectoris Coronary artery disease 414 except 414.1 and 414.3 Other forms of chronic ischemic heart disease 429.2 Cardiovascular disease, unspecified Cardiac arrhythmia 426 427 Congestive heart failure 428 Peripheral vascular disease 440 Atherosclerosis 441 Aortic aneurysm and dissection 442 Other aneurysm 443, except 443.0 Other peripheral vascular disease 447.1 Other disorder of arteries and arterioles stricture of artery Valvular disease 393 398, except for 398.91 Rheumatic heart disease 424 Non rheumatic valve Other cardiovascular disease 745 746 Congenital heart diseases 747 Other congenital anomalies of CV system 425 Cardiomyopathy 420 422 End , peri and myocarditis 390 392 Rheumatic fever Cerebrovascular disease Cerebral hemorrhage 430 Subarachnoid hemorrhage 431 Intracerebral hemorrhage 432 Other and unspecified intracranial hemorrhage Ischemic event 433 Occlusion and stenosis of precerebral arteries 434 Occlusion of cerebral arteries TIA 435 Transient cerebral ischemia Other 436 Acute, but ill defined, cerebrovascular disease 437 Other and ill defined, cerebrovascular disease 438 Late effects of cerebrovascular disease
73 Table 3 2 . Continued Comorbidity ICD 9 CM codes Notes Neurologic Multiple sclerosis 340 Dementia 290, 291 Seizure 345.8, 345.9, 780.32, 780.39 Other comorbidities Diabetes 250, 357.2 362.0 362.07 HIV infection 042 044 Prostate cancer 185, 233.4 Morbid obesity 278.01
74 Table 3 3 . Baseline covariates: m edication s Medication Notes Angiotensin converting enzyme inhibitors 1+ fill for benazepril, captopril, enalapril, fosinopril, lisinopril, moexipril, perindopril, quinapril, ramipril, trandolapri Angiotensin receptor blockers 1+ fill for candesartan (atacand), eprostartan (teveten), irebsartan (avapro) losartan (cozaar), olmesartan (benicar), telmisartan (micardis), valsartan (diovan) Calcium channel blockers 1+ fill for diltiazem HCL, diltiazem malate, mibefradil di HCL, verapamil HCL, amlodipine besylate, bepridil HCL, felodipine, isradipine, nicardipine HCL, nifedipine, nimodipine, nisoldipine blockers 1+ fill for acebutolol HCL, atenolol, betaxolol HCL, bisoprolol fumarate, carteolol HCL, esmolol HCL, metoprolol succinate, metoprolol tartrate, nadolol, oxprenolol, penbutolol sulfate, pindolol, propranolol HCL, sotalol HCL, timolol maleate Diuretics 1+ fill for bendroflumethiazide, benzthiazide, chlorothiazide, chlorothiazide sodium, chlorthalidone, cyclothiazide, hydrochlorothiazide, hydroflumethiazide , indapamide, methyclothiazide, metolazone, polythiazide, quinethazone, trichlormethiazide Non steroidal anti inflammatory drugs 1+ fill for Ibuprofen, Diclofenac, Naproxen, Flurbiprofen, Ketoprofen, Ketorolac, Sulindac, Piroxicam, Oxaprozin, Nabumetone, Mefanamic acid, Meclofenamate, Fenoprofen, Diflunisal, Etodolac, Meloxicam, Tolmetin, Indomethacin Statin 1+ fill for atorvastatin, cerivastatin, fluvastatin, lovastatin, pravastatin, simvastatin, rosuvastatin Fibrate lipid lowering agents 1+ fill for Ch olestyramine, Colestipol, Colesevelam, Bezafibrate, Fenofibrate, Ezetimibe, Gemfibrozil
75 Table 3 3. continued Medication Notes Benzodiazepines 1+ fill for Alprozolam, Chlordiazepoxide, Clonazepam, Clorazepate, Diazepam, Estazolam, Flurazepam, Halazepam, Lorazepam, Midazolam, Oxazepam, Quazepam, Temazepam,Trizolam, Buspirone, Zaleplon, Zolpidem Selective serotonin reuptake inhibitors 1+ fill for Paroxetine, Sertraline, Venlafaxine, Fluoxetine, Citalopram, Nefazodone, Mirtazapine, and Fluvoxamine Antipsychotics 1+ fill for Acetophenazine, Chlorpromazine, Fluphenazine decanoate, Fluphenazine hcl, Mesoridazine, Perphenazine, Prochlorperazine, Promazine, Thioridazine, Trifluoperazine, Triflupromazine, Chlorprothixene, Haloperidol, Haloper idol lactate, Haloperidol decanoate, Loxapine, Molindone, Pimozide, Thiothixene, Aripiprazole, Clozapine, Olanzapine, Quetiapine, Risperidone, Ziprasidone Table 3 4 . Baseline covariates: healthcare utilization during 6 months look back period Health care utilization Number of hospitalization for any reason Number of emergency room visits Number of different medication class es filled
76 Table 3 5. Summary of secondary analyses in Part 2 Sensitivity analyses Assumptions Analysis 1 Take 2 PDE5 inhibitor tablets every 7 days Change frequency of PDE5 inhibitor use Analysis 2 Take 1 PDE5 inhibitor tablet every 14 days Change frequency of PDE5 inhibitor use Analysis 3 Take 1 PDE5 inhibitor tablet every 21 days Change frequency of PDE5 inhibitor use Analysis 4 Take 1 PDE5 inhibitor tablet every 30 days Change frequency of PDE5 inhibitor use Analysis 5 Conduct a propensity score matched cohort analysis PS matching eliminates a greater proportion of confounding by baseline covariates compared with other methods Analysis 6 Define new users of PDE5 inhibitors if they had 12 months baseline period without filling any PDE5 inhibitor prescriptions A longer look back period may be necessary t o define new user status for a drug that is used on demand Analysis 7 Require evidence of oral or intra tympanic steroids treatment to define a SSHL case Further increase the predictive value of the algorithm for identifying SSHL in claims data by requiring evidence of treatment with steroids (oral/intra tympanic injection) Analysis 8 Subgroup of patients with at least 1 ED diagnosis claim Control for confounding by indication by restricting to subjects with 1 ED diagnostic claim Analysis 9 Adju st for additional time dependent covariates (use of anti depressants, anti hypertensive, anti psychotics, NSAIDs) Control for residual confounding by adjusting for other time dependent covariates
77 Figure 3 1 . Study F low C hart for Part 2 of the Thesis . Flow chart to illustrate how exposure categories are defined in primary and sensitivity analysis based on different assumptions of PDE5 inhibitor use. For the demonstration purpose, we assume an adult male patient filled 2 prescr iptions for PDE5 inhibitors that are 3 months apart. Each prescription contains 4 PDE5 inhibitor pills. In the primary analysis, we assume the patient starts the treatment on the day of dispensing. He takes one pill per week for the next 4 weeks. In second ary frequency of us PDE5 inhibitor use in the past 365 day.
78 CHAPTER 4 RESULTS P art 1 : Patterns of PDE5 Inhibitor Use in a Commercially Insured Population Prevalence and Incidence of Diagnosed ED Table 4 1 present s the annual number of eligible patients who had at least one ED diagnosis claim (ICD 9 CM code 607.84 or 302.72 ) and the annual prevalence of ED in 1998 2007 . W e observed a small decreasing prevalence of ED between 1998 and 2007 (Figure 4 1) , t he prevalence of ED was 16. 2 (95% CI: 16.0 16.5) per 1,000 subjects in 1998 and 13.3 (95 % CI: 13.2 13.4 ) per 1 ,000 subjects in 2007 (Table 4 1). However, it was hard to depict an overall secular trend of increasing or decreasing prevalence from these data, because the estimates fluctuated from year to year , possibly because the study population has changed considerably. This could have changed the composition of patient population, as well as the types of health plan. In this study, p revalence of ED was higher as subjects became older , and this age distribution was cons i stent in all calendar years. Approxima tely , two third s of ED patients were identified with ICD 9 CM code 607.84 ( ED with an organic o rigin ), and the remaining by ICD 9 CM code 302.72 (ED with a psychological origin). Table 4 2 present s the number of incident ED cases, number of person years and incidence rate of ED by calendar year and by age group . The annual incidence of E D increased from 9.2 cases per 1,000 person years in 199 9 to 1 2 .3 cases per 1,000 person years in 2007 . As we expected, the incidence of ED was higher with in creasing age . Similar to the prevalence estimates, the annual incidence estimates also fluctuated from year to year which might have been affected by the significant changes in study population.
79 Prevalence and Incidence of PDE5 Inhibitor Use The a nnual pr evalence of PDE5 inhibitor use in the study population increased from 16. 4 (95% CI: 16.1 16.6) per 1,000 subjects in 1998 to 29.5 (95% CI: 29.4 29.6) per 1,000 subjects in 2007 (Figure 4 2). Within each calendar year, the proportion of patients receiving P DE5 inhibitor was relatively similar across age grou ps . The m ajority (>90%) of PDE5 inhibitor users were 40 years of age. The p revalence of PDE5 inhibitor use was highest among subjects aged 60 64 years, but the most increase was seen in those patients aged 40 49 years (Table 4 3). Table 4 4 present s the number of incident PDE5 inhibitor user s and incidence of PDE5 inhibitor use in 1999 2007. The annual incidence of PDE5 inhibitor use increased from 11.5 (95% CI: 11.2 11.7) per 1,000 person years in 19 99 to 18.6 (95% CI: 18.5 18.8) per 1,000 person years in 2007. Similar to the pattern seen in the prevalence of PDE5 inhibitor use, there was a monotonic increas ing trend in the incidence of PDE5 inhibitor use by age, suggesting that elder people were more likely to initiate PDE5 inhibitor therapy. Men with the greatest percent age of increase from 1999 to 2007 were in the age group of 40 49 years (Table 4 4). Finally, the annual prevalence and incidence estimates went up and down from year to year. This ma y have been affected by the changes in health plan types and patient population over the study period. Longitudinal pattern of PDE5 inhibitor use For this analysis, w e identified a total of 216,412 eligible subjects who initiated a PDE5 inhibitor during 1998 2007 and who had at least 6 months of continuous eligibility before and after the index prescription. Among th is group , the majority (71.6%) received an initial PDE5 inhibitor prescription for sildenafil , 12.9% for vardenafil and 15.5%
80 received tadalafil. M ost patients started with either a mediate or high dose of any of the three PDE5 inhibitor s . Overall, about 60% of new users refilled their prescription at least once within the 6 months follow up time, regardless of which PDE5 inhibitor medica tion they received initially and the dose of their initial prescription . During the 6 month follow up time, t he average number of prescription s dispensed per patient was 3.5, 3.3, and 3.5 for those individuals who started on sil denafil, vardenafil, and tad alafil respectively (including the index prescription). The average number of pills dispensed was 23, 20, and 21 for subjects who started o n sildenafi l, vardenafil and tadalafil respectively , corresponding to an average of 3.8, 3.3, and 3.5 tablets per mon th . Neither th e initial type of drug nor the initial dose ha d any appreciable impact on refill frequency or the average number of tablets dispensed (Table 4 5). Among patients who had filled at least two prescriptions (include the index prescription) durin g the 6 month follow up time , 7%, 9%, and 7% of sildenafil , vardenafil, and tadalafil initiators changed their dose . As expected , most patients who changed their dose were on a low dose regimen at the beginning . To estimate the percent of patients who switched medication during the 6 month follow up time, we further restricted the cohort to those who initiated a PDE5 inhibitor between 2004 and 2007 , when all three drugs were marketed . We found a very small percent of medicat ion switching for all 3 PDE5 inhibitors in the study population (Table 4 5). F indings from the sensitivity analysis where the follow up was expanded to 12 months suggest a very similar utilization pattern (Table 4 6). The proportion of patients who refill ed their prescription increased slightly to about 70% while the average number of pills per months decreased.
81 P a r t 2 : P DE5 Inhibitor Use and Risk of SS Cohort Selection We identified 611,016 subjects who filled at least one prescription for a PDE5 inhibitor and 18,299,768 nonusers of PDE5 inhibitors between 1998 and 2007. After applying the inclusion and exclusion criteria, a total of 377,722 initiators of PDE5 inhibitors and 1, 957,233 nonusers formed the final study cohort (Figure 4 4 ). Patient Characteristics Table 4 7 describ e s demographic and clinical characteristics of the study cohorts . Of all new users of PDE5 inhibitors, 60% of patients received sildenafil . A bout 20% each started on vardenafil and tadalafil. At the index date , the mean age was 52.9 years (SD, 7.8) for new users of PDE5 inhibitor and 41.7 years (SD, 12.7) for nonusers. Most subjects resided in the North Central and South regions . Compared with nonusers, n ew users of PDE5 inhibitors were more likely to have a diagnos is of hypertension, diabetes mellitus, and hyperlipidemia, and to have used one or more cardiovascular drugs , suggest ing an elevated risk of cardiovascular disease in this group . As one might expect, patients beginning to take PDE5 inhibitors were also mor e likely than nonusers to have a diagnosis of lower urinary tract syndrome (LUTS), prostate or bladder cancer during the baseline period . New users of PDE5 inhibitor were also more likely to have used statins and other lipid lowering drugs, NSAIDS, anti de pressants, and anti psychotic medications in the 6 months period prior to the index date. There were no appreciable differences with respect to other comorbid conditions , coexisting medications, and health care utilization variables between PDE5 inhib itor user and nonuser group s (Table 4 7 ).
82 Risk of SSHL in Study Cohorts (Table 4 8 ). After controlling for p ropensity score ( logit transformed ) and age, current use of PDE5 inhibitor s was associated with an elevated risk of SSHL compared with nonuse of PDE5 inhibitors ( HR=1.25 , 9 5% CI, 1.01 1.55 ). The adjusted relative risk of SSHL comparing periods of recent use with nonuse was also elevat ed ( HR=1.6 0 , 95% CI , 1.3 3 1. 94 ). Compared to n onuse, current use of PDE5 inhibitor s was associated with an excess risk of 1. 97 (95% CI, 1.12 2.82 ) cases per 1,000 person years . According to published surveys, the majority of ED patients report s use of PDE5 inhibitor s at least once every 1 2 weeks. 106 109 In the sensitivity analyses, we tested other assumptions with regard to the frequency of PDE5 inhibitor use, and we examined how these different patterns of use may change our effect estimates. When we assumed that patients used 2 PDE5 inhibitor tablets per week ( i.e., use PDE5 inhibitor more frequently) , we found no elevated risk for SSHL (HR=1 .01 , 95% CI, 0.77 1.34) comparing current use with nonuse. Interestingly, the HR was 1.66 (95% CI, 1.40 1.97) comparing recent use versus nonuse when patients took 2 P DE5 inhibitor pills per week . When we assumed that patients used one PDE5 inhibitor tablet every 14, 21, or 30 days, the adjusted HR for current use increased while the HR for recent use decrease somewhat due to re classification of cases to current or rec ent u se (Table 4 8 ). Results for the propensity score matched cohort were similar to the primary analysis as were results for the expanded definition of SSHL requiring steroid administration . Although no treatment has been proved to be absolutely effective, or al systemic or topical steroids were the most popular treatment for SSHL in the US during
83 the study period. 133 W e conducted a sensitivity analysis by requiring SSHL cases to be treated with steroids in addition to those ICD 9 CM and CPT codes as required in the primary analysis (Table 4 8) . W e conducted two sensitivity analyses in which we altered methods to assemb le the study cohorts. First, we defined new users of PDE5 inhibitor by requiring a 12 month look back period prior to the index date without filling any PDE5 inhibitor. In this new cohort, the adjusted HR for current and recent use versus nonuse was 1 .13 (0.92 1.39) and 1.2 0 (0.96 1.51) , respectively . The confidence bounds were wider because of the smaller SSHL number of events and cohort size . Second, we requir ed all patients (users and nonusers) to have at least one ED diagnosis with no specification for a temporal relationship between the ED diagnosis and PDE5 inhibitor initiation. The corresponding HR was 1. 26 (0.82, 1.94 ) and 1.85 (1.26, 2.72 ) comparing current use and recent use with nonuse. Finally, addition of several prescription drugs that have be en associated with SSH L in the Cox regression models as time dependent variables did not alter the results meaningfully (Table 4 8 ). Overall, a small significantly increased risk of SSHL associated with current PDE5 inhibitor use was noted in most analyse s performed (Figure 4 5). These analyses were base d on an untestable assumption that subjects who took these medications used them intermittently. Given the fact that, on average, patients in the study refilled one prescription every 1.5 2 months for 4 6 p ills of supply, the assumptions in the primary analysis are plausible .
84 P a r t 3 : Comparative Safety of PDE5 Inhibitors and Risk of S S HL Cohort Selection Figure 4 6 illustrat e s our cohort selection process for the primary and secondary analyses. Of 611,016 new users of PDE5 inhibitors, 398,917 subjects had 6 months continuous health plan enrollment prior to the index date. 395,706 subjects were aged 18 64 years and had access to full drug benefits program s . We excluded 17,984 subjects whose first PDE5 inhibitor prescription was dispensed before December of 2003 or who had other exclusion conditions including comorbid ities and medication use at baseline . T he final cohort consisted of 275,229 new users of PDE5 inhibitors, including 158,860 patients who started on sildenafil, 49,774 and 66,595 initiated vardenafil and tadalafil , respectively. No patients initiated more than one PDE5 inhibitor at the index date. Patient Characteristi cs The baseline characteristics of th e cohorts are listed in Table 4 9 and Table 4 1 2. The mean age was 52.5 years (SD 8.0) in new users of sildenafil , 51.8 years (SD 8.3) in tadalafil new users , and 52.1 years (SD 8.2) in v ardenafil new users. The baselin e characteristics including health care utilization, comorbidities and use of other medications were similar in the three groups. Compared with sildenafil new users, new users of tadalafil and vardenafil were slightly more likely to initiate the ir treatment in 2006 and 2007 and to have a diagnosis of hyperlipidemia, diabetes, or prostate cancer. Tadalafil users were also more likely to have a history of hypertension and lower urinary tract syndrome (LUTS) compared with new users of sildenafil . The inverse probability weighted IPTW weighted estimates demonstrat e that baseline characteristics were adequately balanced .
85 Risk of SSHL in Study Cohorts Only 61 incident cases of SSHL were observed during the follow up time. The crude incidence rate (IR) of SSHL in sildenafil, tadalafil and vardenafil users was 4.45, 3.62, and 3.69 SSHL per 1,000 person years respectively (Table 4 10, Table 4 12) . In the primary analysis (where we assumed patients used 1 PDE5 inhibitor dose per week) , tadalafil users were not at a significantly higher or lower risk of SSHL compared with sildenafil users (HR= 0.80, 95% CI, 0.42 1.51 ). The absolute difference in the IR between tadalafil and sildenafil users was also not significant ( IRD= 0.82, 95% CI, 3.27 1.60). Similarly, v ardenafil use was not associated with a significantly higher or lower risk of SSHL ( HR=1.02, 95% CI, 0.51 2.02 ; IRD=0.25, 95% CI, 2.90 3.34 ) compared with sildenafil use. However, owing to the wide confidence interval and relatively small number of SSHL events included in these analyses, we cannot rule out a significant r esult for any of these pairwise comparisons . We did a number of sensitivity analyses by assigning different PDE5 inhibitor use frequenc ies to these patients but results were not meaningfu lly different from the primary analysis. Overall, no significant differences were observed between the three PDE5 inhibitors (Table 4 11 , Table 4 14 ), although the wide confidence interval s limited formal inferences due to low statistical power .
86 Tabl e 4 1 . Prevalence of erectile dysfunction Year No. of ED cases No. of eligible subjects Prevalence rate (per 1,000 subjects) 1998 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 16,510 232 954 3,658 7,996 3,670 1,018,483 191,939 222,766 276,873 248,160 78,745 16.2 (16.0 16.5) 1.2 (1.1 1.4) 4.3 (4.0 4.6) 13.2 (12.8 13.6) 32.2 (31.5 32.9) 46.6 (45.1 48.1) 1999 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 13,097 246 892 3,086 6,058 2,815 1,046,789 198,061 219,054 275,441 263,866 90,367 12.5 (12.3 12.7) 1.2 (1.1 1.4) 4.1 (3.8 4.3) 11.2 (10.8 11.6) 23.0 (22.4 23.5) 31.2 (30.0 32.3) 2000 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 14,954 314 1,215 3,637 6,844 2,944 1,112,394 215,696 234,341 289,537 279,173 93,647 13.4 (13.2 13.7) 1.5 (1.3 1.6) 5.2 (4.9 5.5) 12.6 (12.2 13.0) 24.5 (23.9 25.1) 31.4 (30.3 32.6) 2001 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 21,821 481 1,615 5,152 10,330 4,243 1,780,168 350,157 334,989 448,415 482,635 163,972 12.3 (12.1 12.4) 1.4 (1.3 1.5) 4.8 (4.6 5.1) 11.5 (11.2 11.8) 21.4 (21.0 21.8) 25.9 (25.1 26.7) 2002 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 33,137 902 3,181 8,137 14,965 5,952 3,247,722 668,529 685,198 817,732 809,175 267,088 10.2 (10.1 10.3) 1.3 (1.3 1.4) 4.6 (4.5 4.8) 10.0 (9.7 10.2) 18.5 (18.2 18.8) 22.3 (21.7 22.9) ICD 9 CM code : 607 . 84 or 302 . 72) were be tween 18 64 years old and had 6 months continuous enrollment in a given calendar year
87 Table 4 1. Continued Year No. of eligible Prevalence rate (per 1,000 subjects) 2003 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 52,987 1,600 5,445 13,866 22,853 9,223 4,980,488 1,011,624 1,084,239 1,297,126 1,195,267 392,232 10.6 (10.5 10.7) 1.6 (1.5 1.7) 5.0 (4.9 5.2) 10.7 (10.5 10.9) 19.1 (18.9 19.4) 23.5 (23.0 24.0) 2004 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 54,728 1,674 5,559 14,334 23,256 9,950 5,111,258 1,031,461 1,085,710 1,332,458 1,238,949 422,707 10.7 (10.6 10.8) 1.6 (1.5 1.7) 5.1 (5.0 5.3) 10.8 (10.6 10.9) 18.8 (18.5 19.0) 23.4 (23.0 23.9) 2005 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 52,019 1,593 5,377 13,435 22,059 9,555 4,989,636 1,009,138 1,055,647 1,285,515 1,216,826 422,510 10.4 (10.3 10.5) 1.6 (1.5 1.7) 5.1 (5.0 5.2) 10.5 (10.3 10.6) 18.1 (17.9 18.4) 22.6 (22.2 23.1) 2006 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 54,066 1,726 5,795 14,117 22,823 9,605 4,630,321 911,795 951,845 1,192,046 1,171,756 402,879 11.7 (11.6 11.8) 1.9 (1.8 2.0) 6.1 (5.9 6.2) 11.8 (11.6 12.0) 19.5 (19.2 19.7) 23.8 (23.4 24.3) 2007 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 109,830 3,485 11,597 28,778 46,527 19,443 8,277,197 1,740,566 1,729,700 2,100,870 2,027,273 678,788 13.3 (13.2 13.4) 2.0 (1.9 2.1) 6.7 (6.6 6.8) 13.7 (13.5 13.9) 23.0 (22.7 23.2) 28.6 (28.2 29.0) CM code: 607.84 or 302.72) between 18 calendar year
88 Table 4 2 . Incidence of erectile dysfunction Year No. of incident ED cases No. of person years Incidence rate (per 1,000 person years) 1999 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 5,879 109 408 1,548 2,661 1,153 636,700 102,397 128,392 180,447 171,302 54,162 9.2 (9.0 9.5) 1.1 (0.9 1.3) 3.2 (2.9 3.5) 8.6 (8.2 9.0) 15.5 (14.9 16.1) 21.3 (20.1 22.5) 2000 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 9,304 151 717 2,325 4,351 1,760 770,125 122,288 149,749 211,239 216,229 70,640 12.1 (11.8 12.3) 1.2 (1.0 1.4) 4.8 (4.4 5.1) 11.0 (10.6 11.5) 20.1 (19.5 20.7) 24.9 (23.7 26.1) 2001 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 8,797 179 697 2,214 4,035 1,672 788,699 126,836 152,370 212,594 222,577 74,312 11.2 (10.9 11.4) 1.4 (1.2 1.6) 4.6 (4.2 4.9) 10.4 (10.0 10.8) 18.1 (17.6 18.7) 22.5 (21.4 23.6) 2002 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 14,093 287 1,112 3,279 6,648 2,767 1,317,988 209,539 227,753 342,257 399,194 139,244 10.7 (10.5 10.9) 1.4 (1.2 1.5) 4.9 (4.6 5.2) 9.6 (9.3 9.9) 16.7 (16.3 17.1) 19.9 (19.1 20.6) 2003 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 24,097 525 2,065 5,913 10,895 4,699 2,287,131 388,990 443,855 594,218 635,864 224,203 10.5 (10.4 10.7) 1.3 (1.2 1.5) 4.7 (4.5 4.9) 10.0 (9.7 10.2) 17.1 (16.8 17.5) 21.0 (20.4 21.6) presence of at least one medical claim for ED ( ICD9 CM code: 607 . 84 or 302 . 72) were 18 64 years old who were eligible in January of the measurement year and who had 12 months continuous enrollment in the previous year without a diagnosis of ED Â§ Incident ED case is defined as an ED patient whose first diagnosis of ED in a year is precede d by at least 12 months continuous enrollment without a diagnosis for ED
89 Table 4 2. Continued Year No. of incident ED No. of person Incidence rate (per 1,000 person years) 2004 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 32,136 820 3,017 8,113 13,977 6,209 3,308,968 559,486 654,436 890,280 891,192 313,575 9.7 (9.6 9.8) 1.5 (1.4 1.6) 4.6 (4.4 4.8) 9.1 (8.9 9.3) 15.7 (15.4 15.9) 19.8 (19.3 20.3) 2005 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 32,446 791 2,997 8,277 14,140 6,241 3,345,550 554,121 649,714 899,288 918,386 324,041 9.7 (9.6 9.8) 1.4 (1.3 1.5) 4.6 (4.4 4.8) 9.2 (9.0 9.4) 15.4 (15.1 15.7) 19.3 (18.8 19.7) 2006 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 30,368 746 2,985 7,885 13,194 5,558 2,697,045 434, 872 520,193 713,559 759,937 268,489 11.3 (11.1 11.4) 1.7 (1.6 1.8) 5.7 (5.5 5.9) 11.1 (10.8 11.3) 17.4 (17.1 17.7) 20.7 (20.2 21.2) 2007 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 30,583 710 2,828 7,721 13,353 5,971 2,481,051 403,278 470,314 649,679 700,113 257,667 12.3 (12.2 12.5) 1.8 (1.6 1.9) 6.0 (5.8 6.2) 11.9 (11.6 12.1) 19.1 (18.7 19.4) 23.2 (22.6 23.8) CM code: 607.84 or 302.72) were 18 64 years old who were eligible in January of the measurement year and who had 12 months continuous enrollment in the previous year without a diagnosis of ED Â§ Incident ED case is defined as an ED patient whose first diagnosis of ED in a year is preceded by at least 12 months continuous e nrollment without a diagnosis for ED
90 Table 4 3 . Prevalence of PDE5 inhibitor use Calendar year No. of PDE5 inhibitor users No. of eligible subjects Prevalence (per 1,000 subjects) 1998 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 16,672 115 639 3,080 8,279 4,559 1,018,483 191,939 222,766 276,873 248,160 78,745 16.4 (16.1 16.6) 0.6 (0.5 0.7) 2.9 (2.7 3.1) 11.1 (10.7 11.5) 33.4 (32.7 34.1) 57.9 (56.3 59.5) 1999 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 20,440 140 782 3,900 10,354 5,264 1,046,789 198,061 219,054 275,441 263,866 90,367 19.5 (19.3 19.8) 0.7 (0.6 0.8) 3.6 (3.3 3.8) 14.2 (13.7 14.6) 39.2 (38.5 40.0) 58.3 (56.7 59.8) 2000 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 26,055 192 1,195 5,227 13,277 6,164 1,112,394 215,696 234,341 289,537 279,173 93,647 23.4 (23.1 23.7) 0.9 (0.8 1.0) 5.1 (4.8 5.4) 18.1 (17.6 18.5) 47.6 (46.8 48.4) 65.8 (64.2 67.4) 2001 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 71,359 418 2,568 14,806 37,480 16,087 1,780,168 350,157 334,989 448,415 482,635 163,972 40.1 (39.8 40.4) 1.2 (1.1 1.3) 7.7 (7.4 8.0) 33.0 (32.5 33.5) 77.7 (76.9 78.4) 98.1 (96.7 99.6) 2002 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 115,763 798 5,166 24,557 60,055 25,187 3,247,722 668,529 685,198 817,732 809,175 267,088 35.6 (35.4 35.8) 1.2 (1.1 1.3) 7.5 (7.3 7.7) 30.0 (29.7 30.4) 74.2 (73.6 74.8) 94.3 (93.2 95.4) In each measurement year, eligible subjects had to have at least 6 months continuous enrollment in health plan that contributed data to the MarketScan CCAE database and were between 18 64 years at the beginning of that year.
91 Table 4 3. Continued Calendar year No. of user of No. of eligible Prevalence rate (per 1,000 subjects) 2003 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 151,287 1,229 7,556 32,794 76,900 32,808 4,980,488 1,011,624 1,084,239 1,297,126 1,195,267 392,232 30.4 (30.2 30.5) 1.2 (1.1 1.3) 7.0 (6.8 7.1) 25.3 (25.0 25.6) 64.3 (63.9 64.8) 83.6 (82.8 84.5) 2004 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 159,867 1,276 8,105 34,568 79,660 36,258 5,111,258 1,031,461 1,085,710 1,332,458 1,238,949 422,707 31.3 (31.1 31.4) 1.2 (1.1 1.3) 7.5 (7.3 7.6) 25.9 (25.7 26.2) 64.3 (63.9 64.7) 85.8 (84.9 86.6) 2005 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 161,296 1,309 8,520 34,993 79,879 36,595 4,989,636 1,009,138 1,055,647 1,285,515 1,216,826 422,510 32.3 (32.2 32.5) 1.3 (1.2 1.4) 8.1 (7.9 8.2) 27.2 (26.9 27.5) 65.6 (65.2 66.1) 86.6 (85.8 87.4) 2006 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 192,754 1,639 10,945 42,768 95,885 41,516 4,630,321 911,795 951,845 1,192,046 1,171,756 402,879 41.6 (41.4 41.8) 1.8 (1.7 1.9) 11.5 (11.3 11.7) 35.9 (35.5 36.2) 81.8 (81.3 82.3) 103.1 (102.1 103.9) 2007 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 244,622 2,417 15,049 56,257 118,831 52,068 8,277,197 1,740,566 1,729,700 2,100,870 2,027,273 678,788 29.5 (29.4 29.6) 1.4 (1.3 1.5) 8.7 (8.5 8.8) 26.8 (26.5 27.0) 58.6 (58.3 58.9) 76.7 (76.0 77.3) year, e ligible subjects had to have at least 6 months continuous enrollment in health plan that contributed data to the MarketScan CCAE database and were between 18 64 years at the beginning of that year.
92 Table 4 4 . Incidence of PDE5 inhibitor use Calendar year No. of new PDE5 inhibitor users No. of person years Incidence rate (per 1,000 person years) 1999 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 7,336 46 345 1,595 3,697 1,653 639,193 102,501 128,745 181,514 172,310 54,122 11.5 (11.2 11.7) 0.4 (0.3 0.6) 2.7 (2.4 3.0) 8.8 (8.4 9.2) 21.5 (20.8 22.1) 30.5 (29.1 32.0) 2000 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 9,933 59 471 2,099 5,056 2,248 768,636 122,415 150,185 211,813 214,716 69,498 12.9 (12.7 13.2) 0.5 (0.4 0.6) 3.1 (2.9 3.4) 9.9 (9.5 10.3) 23.5 (22.9 24.2) 32.3 (31.0 33.7) 2001 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 10,670 98 585 2,431 5,235 2,321 784,445 126,981 152,763 212,752 219,581 72,369 13.6 (13.3 13.9) 0.8 (0.6 0.9) 3.8 (3.5 4.1) 11.4 (11.0 11.9) 23.8 (23.2 24.5) 32.1 (30.8 33.4) 2002 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 30,998 225 1,413 6,864 15,968 6,528 1,283,905 209,718 227,722 337,088 379,516 129,861 24.1 (23.9 24.4) 1.1 (0.9 1.2) 6.2 (5.9 6.5) 20.4 (19.9 20.8) 42.1 (41.4 42.7) 50.3 (49.0 51.5) 2003 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 38,037 395 2,247 8,764 18,606 8,025 2,232,992 389,309 443,808 585,970 604,989 208,916 17.0 (16.9 17.2) 1.0 (0.9 1.1) 5.1 (4.9 5.3) 15.0 (14.6 15.3) 30.8 (30.3 31.2) 38.4 (37.6 39.3) alendar year, eligible subjects included male s 18 64 years who had to have at least 6 months continuous enrollment in h ealth plan that contributed data to MarketScan CCAE database and were eligible in January of the measurement year . Subjects were defined as new users of PDE5 inhibit ors in Year X, if they had not filled any PDE5 inh ibitor prescriptions in 12 months preceding the ir first PDE5 inhibitor prescription in Year X .
93 Table 4 4. Continued Calendar year No. of new users of PDE5 inhibitor No. of person years Incidence rate (per 1,000 person years) 2004 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 45,942 487 2,822 10,739 22,123 9,771 3,247,814 560,079 654,694 881,740 855,984 295,316 14.1 (14.0 14.3) 0.9 (0.8 1.0) 4.3 (4.2 4.5) 12.2 (11.9 12.4) 25.8 (25.5 26.2) 33.1 (32.4 33.7) 2005 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 44,063 495 2,863 10,657 21,044 9,004 3,278,130 554,702 649,714 889,301 880,155 304,258 13.4 (13.3 13.6) 0.9 (0.8 1.0) 4.4 (4.2 4.6) 12.0 (11.8 12.2) 23.9 (23.6 24.2) 29.6 (29.0 30.2) 2006 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 48,646 528 3,333 11,852 23,245 9,688 2,626,446 435,284 519,595 702,221 720,286 249,060 18.5 (18.4 18.7) 1.2 (1.1 1.3) 6.4 (6.2 6.6) 16.9 (16.6 17.2) 32.3 (31.9 32.7) 38.9 (38.1 39.7) 2007 Overall 18 29 years 30 39 years 40 49 years 50 59 years 60 64 years 44,978 554 3,073 10,685 21,284 9,382 2,413,692 403,582 469,412 638,040 663,084 239,572 18.6 (18.5 18.8) 1.4 (1.3 1.5) 6.5 (6.3 6.8) 16.7 (16.4 17.1) 32.1 (31.7 32.5) 39.2 (38.4 40.0) 64 years who had to have at least 6 months continuous enrollment in health plan that contributed data to MarketScan CCAE database and were eligible in January of the measurement year. inhibitor prescriptions in 12 months preceding their first PDE5 inhibitor prescription in Year X.
94 Table 4 5 . PDE5 inhibitor utilization pattern during 6 months after initiation Original treatment No. of patients Percent patient s with refill s Number of refill s among subjects with >1 PDE5 dispensing* Total number of tablets Subjects who switched to another PDE5** n, (%) Subjects who changed dose n, (%) Sildenafil 154,962 61 3.5 23 4 7 25mg 2,297 59 3.4 29 4 19 50mg 78,715 61 3.5 23 4 9 100mg 73,950 61 3.5 22 4 4 Vardenafil 27,908 59 3.3 20 6 9 5mg 373 51 3.3 19 5 16 10mg 11,058 59 3.4 20 6 12 20mg 16,477 59 3.3 19 6 7 Tadalafil 33,592 64 3.5 21 5 7 5mg 210 61 3.5 20 8 24 10mg 7,676 62 3.4 21 6 12 20mg 25,706 65 3.5 21 5 5 For new users of PDE5 inhibitors who had 6 months continuous enrollment after their index PDE5 inhibitor dispensing * Include s the index PDE5 inhibitor dispensing ** Dat a collected between 2004 and 2007 when all 3 PDE5 inhibitors were available
95 Table 4 6 . Prescription pattern during 12 months after PDE5 inhibitor initiation Original treatment No. of patients Percent patients with refills Number of refills among subjects with >1 PDE5 dispensing* Total number of tablets Subjects who switched to another PDE5** n, (%) Subjects who changed dose n, (%) Sildenafil 129,990 71 5.0 36 9 13 25mg 1,892 67 5.1 46 8 31 50mg 66,823 72 4.9 36 7 16 100mg 61,275 71 5.0 34 11 9 Vardenafil 21,107 67 4.9 31 11 16 5mg 299 62 4.3 28 10 25 10mg 8,891 68 4.8 32 11 20 20mg 11,917 67 4.9 30 12 13 Tadalafil 24,81 73 5.1 34 9 12 5mg 143 73 5.0 33 13 34 10mg 5,688 71 5.0 33 9 21 20mg 18,650 74 5.1 34 8 10 For new users of PDE5 inhibitors who had 12 months continuous enrollment after their index PDE5 inhibitor dispensing * Include the index PDE5 inhibitor prescription ** Date collected between 2004 and 2007 when all 3 PDE5 inhibitors were available
96 Table 4 7. Baseline characteristics of cohort members Characteristic Original cohort Cohort matched by propensity score PDE5 inhibitor nonusers PDE5 inhibitor new users PDE5 inhibitor nonusers PDE5 inhibitor new users n 1,957,233 377,722 302,067 302,06 7 Age, years, mean Â± SD 41.7Â±12.7 52.9Â±7.8 52.9Â±7.9 52.4Â±7.9 Age (%) 18 39 43.6 7.4 7.3 7.5 40 54 37.7 45.9 46.1 46.3 55 64 18.7 46.7 46.6 46.2 Health plan type (%) Employer based 76.8 88.4 86.9 87.7 Large health plan 23.2 11.8 13.1 12.3 Region (%) Northeast 11.4 10.2 10.8 10.5 North Central 26.2 34.9 31.8 33.6 South 39.0 36.6 38.1 37.1 West 22.3 17.3 18.4 17.9 Unknown 1.1 1.0 0.9 0.9 Year of index date (%) 1998 0.9 0.8 0.6 0.8 1999 3.3 2.1 2.4 2.0 2000 3.7 2.7 3.0 2.7 2001 5.0 5.0 4.8 5.0 2002 8.9 10.3 9.0 10.4 2003 13.9 13.1 12.9 13.3 2004 17.1 15.1 16.5 15.0 2005 17.0 15.0 16.3 15.7 2006 14.3 16.9 16.2 16.8 2007 15.7 19.1 18.8 18.8
97 Table 4 7. Continued Characteristic Original cohort Cohort matched for propensity score PDE5 inhibitor nonusers PDE5 inhibitor new users PDE5 inhibitor nonusers PDE5 inhibitor new users Comorbid conditions (%) Myocardial infarction 0.5 1.1 1.2 1.1 Atrial fibrillation 0.4 1.1 1.1 1.0 Ventricular arrhythmias 0.2 0.4 0.4 0.4 Congestive heart failure 0.3 0.7 0.8 0.7 Hypertension 9.3 26.9 25.6 24.7 Cerebrovascular disease 2.2 6.0 6.1 5.7 Peripheral vascular disease 0.2 0.6 0.5 0.5 Hyperlipidemia 10.0 26.3 26.0 24.4 Alcoholism 0.2 0.6 0.5 0.5 Smoking 0.4 0.9 0.8 0.8 Diabetes mellitus 4.5 15.4 13.6 13.8 Recorded obesity 0.4 0.8 0.8 0.7 Depression 2.2 4.9 4.4 4.3 Anxiety 1.0 1.8 1.8 1.6 Chronic kidney disease 0.3 0.7 0.7 0.7 Prostate cancer 0.4 3.6 1.5 2.2 Bladder cancer 0.1 0.2 0.2 0.2 Lower urinary tract syndrome 1.8 7.9 5.7 6.2 Erectile dysfunction 0.4 9.1 1.4 2.6 Co medications (%) Angiotensin converting enzyme inhibitors 6.4 18.6 17.9 17.4 A ngiotensin receptor blockers 1.8 5.6 5.1 5.0 Beta blockers 5.6 13.9 13.4 14.3 Calcium channel blockers 4.4 13.5 12.6 12.4 Diuretics 5.9 18.2 17.1 16.7 Statins 9.9 29.0 28.3 27.1 Other lipid lowering drugs 2.9 8.6 8.2 8.0 Benzodiazepines 3.5 9.9 8.8 8.8 SSRI 4.0 8.7 8.6 8.1 Anti psychotics 0.6 1.2 1.1 1.1 Non steroidal anti inflammatory drugs 9.0 19.9 18.9 18.3 Flu vaccination 2.1 5.1 4.9 4.7 Pneumococcal vaccination 0.3 0.9 0.8 0.8
98 Table 4 7. Continued Characteristic Original cohort Cohort matched for propensity score PDE5 inhibitor nonusers PDE5 inhibitor new users PDE5 inhibitor nonusers PDE5 inhibitor new users Health service utilization (%) Number of outpatient visits 1.2Â±1.9 2.6Â±2.6 2.4Â±2.6 2.3Â±2.3 Number of emergency department visits 0.1Â±0.3 0.1Â±0.4 0.1Â±0.3 0.1Â±0.4 Number of hospitalization 0.02Â±0.13 0.04Â±0.20 0.04Â±0.2 0.04Â±0.2
99 Table 4 8. HRs for the association between PDE5 inhibitor use and risk of SSHL No use Current use Recent use Primary analysis No. of events 994 104 135 Person years 4,171,208 239,001 242,056 Incidence rate Â¶ 2.38 4.35 5.58 HR (95% CI) Reference 1.25 (1.01, 1.55) 1.60 (1.33, 1.94) IRD Â¶ (95% CI) Reference 1.97 (1.12, 2.82) 3.19 (2.24, 4.14) Sensitivity analysis 1 (2 PDE5 inhibitors per week) No. of events 997 56 180 Person years 4,180,901 158,547 312,563 Incidence rate Â¶ 2.38 3.53 5.76 HR (95% CI) Reference 1.01 (0.77, 1.34) 1.66 (1.40, 1.97) IRD Â¶ (95% CI) Reference 1.15 (0.21, 2.08) 3.37 (2.52, 4.23) Sensitivity analysis 2 (take 1 PDE5 inhibitor every 14 days) No. of events 991 141 101 Person years 4,154,314 324,568 173,129 Incidence rate Â¶ 2.38 4.34 5.83 HR (95% CI) Reference 1.24 (1.02, 1.50) 1.66 (1.34, 2.06) IRD Â¶ (95% CI) Reference 1.96 (1.23, 2.69) 3.45 (2.30, 4.60) Sensitivity analysis 3 (take 1 PDE5 inhibitor every 21 days) No. of events 988 179 66 Person years 4,139,987 372,119 139, 069 Incidence rate Â¶ 2.39 4.81 4.75 HR (95% CI) Reference 1.37 (1.15, 1.63) 1.33 (1.03, 1.73) IRD Â¶ (95% CI) Reference 2.42 (1.70, 3.14) 2.36 (2.20, 3.51) Sensitivity analysis 4 (take 1 PDE5 inhibitor every 30 days) No. of events 985 198 50 Person years 4,124,917 413,627 113,467 Incidence rate Â¶ 2.39 4.79 4.41 HR (95% CI) Reference 1.35 (1.14, 1.60) 1.24 (0.92, 1.66) IRD Â¶ (95% CI) Reference 2.40 (1.72, 3.08) 2.02 (0.79, 3.25) Sensitivity analysis 5 (propensity score matched analysis) No. of events 249 88 112 Person years 1,042,802 203,151 209,748 Incidence rate Â¶ 2.39 4.33 5.34 HR (95% CI) Reference 1.38 (1.07, 1.74) 1.58 (1.18, 1.95) IRD Â¶ (95% CI) Reference 1.94 (1.36, 2.92) 2.95 (2.04, 3.87) CI, confidence interval; H R, hazard ratio; IRD, incidence rate difference; PY, person years Â¶ per 10,000 person years * Adjusted for age (continuous) as time dependent variable and logit of propensity score ** In the primary analysis, we assumed that patients used 1 PDE5 inhibitor per week *** Eligible subjects can enter study cohort once only
100 Table 4 8. Continued No use Current use Recent use Sensitivity analysis 6 (12 months look back period) No. of events 1,001 94 107 Person years 4,266,785 247,791 321,565 Incidence rate Â¶ 2.35 3.79 3.33 HR (95% CI) Reference 1.13 (0.92, 1.39 ) 1. 20 (0.96 , 1.51 ) IRD Â¶ (95% CI) Reference 1.45 (0.67, 2.23) 0.98 (0.33, 1.63) Sensitivity analysis 7 (require SSHL cases to have evidence of treatment with steroids) No. of events 670 68 89 Person years 2,752,997 156,740 161,240 Incidence rate Â¶ 2.43 4.34 5.52 HR (95% CI) Reference 1.36 (1.11, 1.72) 1.54 (1.17, 1.86) IRD Â¶ (95% CI) Reference 1.91 (1.01, 2.97) 3.09 (2.06, 4.34) Sensitivity analysis 8 (require an ED diagnosis claim for both groups) No. of events 78 31 44 Person years 213,158 68,553 64,601 Incidence rate Â¶ 3.66 4.52 6.81 HR (95% CI) Reference 1.26 (0.82, 1.94) 1.85 (1.26, 2.72) IRD Â¶ (95% CI) Reference 0.86 ( 0.92, 2.65) 3.15 (0.98, 5.32) Sensitivity analysis 9 (adjusting for additional time dependent variables) No. of events 994 104 135 Person years 4,171,208 239,001 242,056 Incidence rate Â¶ 2.38 4.35 5.58 HR (95% CI) Reference 1.19 (0.98, 1.47) 1.46 (1.25, 1.85) IRD Â¶ (95% CI) Reference 1.97 (1.12, 2.82) 3.19 (2.24, 4.14) CI, confidence interval; H R, hazard ratio; IRD, incidence rate difference; PY, person years Â¶ per 10,000 person years * Adjusted for age (continuous) as time dependent variable and logit of propensity score ** In the primary analysis, we assumed that patients used 1 PDE5 inhibitor per week *** Eligible subjects can enter study cohort once only
101 Table 4 9. Baseline patient characteristics of initiators of sildenafil and tadalafil Characteristic Unweighted Inverse PS weighted Sildenafil (n=158,860) Tadalafil (n=66,595) Sildenafil Tadalafil Age, years, mean Â± SD 52.5Â±8.0 51.8Â±8.3 51.9Â±8.0 51.6Â±8.0 Health plan type (%) Employer based 83.5 79.7 81.6 80.3 Large health plan 16.5 20.3 18.4 19.7 Region (%) Northeast 10.0 8.0 9.3 9.2 North Central 34.3 33.0 33.9 34.4 South 39.2 45.6 40.0 42.6 West 16.1 13.0 16.4 13.4 Unknown 0.4 0.4 0.4 0.4 Year of index date (%) 2003 4.2 0.5 3.1 2.9 2004 34.4 22.9 29.2 28.5 2005 21.9 24.7 22.8 23.3 2006 20.7 26.0 23.5 24.1 2007 18.8 25.9 21.4 21.2 Comorbid conditions (%) Myocardial infarction 1.2 0.9 1.0 1.0 Atrial fibrillation 1.3 1.0 1.1 1.0 Ventricular arrhythmias 0.5 0.4 0.5 0.5 Congestive heart failure 0.8 0.7 0.7 0.7 Hypertension 28.3 29.6 29.0 29.1 Cerebrovascular disease 6.5 5.3 5.8 5.6 Peripheral vascular disease 0.6 0.6 0.6 0.6 Hyperlipidemia 27.1 29.6 28.4 28.7
102 Table 4 9 . Continued Characteristic Unweighted Inverse of PS weighted Sildenafil (n=158,860) Tadalafil (n=66,595) Sildenafil Tadalafil Comorbid conditions (%) Alcoholism 0.6 0.5 0.6 0.6 Smoking 1.0 0.9 1.0 1.0 Diabetes mellitus 14.5 16.5 15.4 15.8 Recorded obesity 0.8 0.9 0.9 0.9 Depression 4.9 5.1 5.0 5.0 Anxiety 1.9 2.0 2.0 2.0 Chronic kidney disease 0.8 0.8 0.8 0.8 Prostate cancer 3.0 4.6 3.6 3.9 Bladder cancer 0.2 0.3 0.3 0.3 Lower urinary tract syndrome 7.5 9.6 8.4 8.8 Erectile dysfunction 8.6 12.6 10.6 10.9 Co medications (%) ACEI 18.7 17.1 18.2 17.9 A ngiotensin receptor blockers 5.6 6.3 5.9 6.0 Beta blockers 14.6 12.3 13.8 13.1 Calcium channel blockers 13.3 12.9 13.1 13.0 Diuretics 18.9 18.7 18.8 18.8 Statins 29.3 28.0 28.8 28.6 Other lipid lowering drugs 9.2 10.8 9.8 10.0 Benzodiazepines 10.2 10.8 10.5 10.6 SSRI 8.3 8.6 8.4 8.5 Anti psychotics 1.3 1.2 1.3 1.3 NSAIDS 18.6 19.0 18.9 18.9 Influenza vaccination 5.6 4.8 5.3 5.0 Pneumococcal vaccination 1.0 1.0 1.0 1.0
103 Table 4 9 . Continued Characteristic Unweighted Inverse of PS weighted Sildenafil (n=158,860) Tadalafil (n=66,595) Sildenafil Tadalafil Health service utilization (%) Number of outpatient visits 2.6Â±2.6 2.9Â±2.6 2.7Â±2.6 2.7Â±2.6 Number of emergency department visits 0.1Â±0.4 0.1Â±0.4 0.1Â±0.4 0.1Â±0.4 Number of hospitalization 0.04Â±0.20 0.04Â±0.21 0.04Â±0.20 0.04Â±0.20 PS, propensity score; SD, standard deviation; ACEI, Angiotensin converting enzyme inhibitors; SSRI, selective serotonin reuptake inhibitors; NSAIDS, non steroidal anti inflammatory drugs Table 4 10. Risk of SSHL comparing sildenafil with tadalafil (primary analysis) Sildenafil Tadala fil Events 37 13 Person years 83,226 35,882 Incidence rate a 4.45 3.62 HR (95% CI) Reference 0.80 (0.42, 1.51) IRD a (95% CI) Reference 0.82 ( 3.27, 1.60) H R, hazard ratio; IRD, incidence rate difference, CI, confidence interval a. Per 10,000 person years Table 4 11. Risk of SSHL comparing sildenafil with tadalafil (secondary analyses) Sensitivity analysis HR (95% CI) Tadalafil vs. Sildenafil (reference) Take 2 PDE5 inhibitor tablets every 7 days 0.80 (0.39, 1.65) Take 1 PDE5 inhibitor tablet every 14 days 0.79 (0.45, 1.40) Take 1 PDE5 inhibitor tablet every 21 days 0.95 (0.56, 1.60) Take 1 PDE5 inhibitor tablet every 30 days 0.95 (0.57, 1.59) H R, hazard ratio; CI, confidence interval
104 Table 4 12. Baseline patient characteristics of initiato rs of sildenafil and vardenafil Characteristic Unweighted Inverse of PS weighted Sildenafil (n=158,860) Vardenafil (n=49,774) Sildenafil Vardenafil Age, years, mean Â± SD 52.5Â±8.0 52.1Â±8.2 52.3Â±8.0 52.2Â±8.2 Health plan type (%) Employer based 83.5 86.0 84.6 84.9 Large health plan 16.5 14.0 15.4 15.1 Region (%) Northeast 10.0 7.9 9.3 8.8 North Central 34.3 25.4 30.2 29.4 South 39.2 35.4 38.0 37.6 West 16.1 31.0 22.2 23.8 Unknown 0.4 0.3 0.4 0.4 Year of index date (%) 2003 4.2 2.3 3.4 3.0 2004 34.4 24.0 29.5 28.8 2005 21.9 23.9 22.8 23.0 2006 20.7 25.9 23.0 22.1 2007 18.8 23.9 21.3 23.1 Comorbid conditions (%) Myocardial infarction 1.2 1.0 1.0 1.0 Atrial fibrillation 1.3 1.0 1.2 1.1 Ventricular arrhythmias 0.5 0.4 0.5 0.5 Congestive heart failure 0.8 0.8 0.8 0.8 Hypertension 28.3 28.4 28.3 28.3 Cerebrovascular disease 6.5 5.4 6.0 5.9 Peripheral vascular disease 0.6 0.6 0.6 0.6 Hyperlipidemia 27.1 25.8 26.7 26.4 PS, propensity score; SD, standard deviation; ACEI, Angiotensin converting enzyme inhibitors; SSRI, selective serotonin reuptake inhibitors; NSAIDS, non steroidal anti inflammatory drugs
105 Table 4 12. Continued Characteristic Unweighted Inverse of PS weighted Sildenafil (n=158,860) Vardenafil (n=49,774) Sildenafil Vardenafil Comorbid conditions (%) Alcoholism 0.6 0.5 0.6 0.6 Smoking 1.0 0.9 1.0 1.0 Diabetes mellitus 14.5 17.8 15.9 16.3 Recorded obesity 0.8 1.0 0.9 0.9 Depression 4.9 4.8 4.9 4.9 Anxiety 1.9 1.9 1.9 1.9 Chronic kidney disease 0.8 0.8 0.8 0.8 Prostate cancer 3.0 3.7 3.3 3.4 Bladder cancer 0.2 0.2 0.2 0.2 Lower urinary tract syndrome 7.5 7.4 7.4 7.4 Erectile dysfunction 8.6 12.3 9.8 10.2 Co medications (%) ACEI 18.7 20.5 19.2 19,6 A ngiotensin receptor blockers 5.6 6.1 5.8 5.9 Beta blockers 14.6 15.2 14.8 14.9 Calcium channel blockers 13.3 13.2 13.2 13.2 Diuretics 18.9 21.0 19.4 20.1 Statins 29.3 30.1 29.6 29.8 Other lipid lowering drugs 9.2 10.2 9.6 9.8 Benzodiazepines 10.2 9.9 10.1 10.1 SSRI 8.3 8.6 8.5 8.5 Anti psychotics 1.3 1.2 1.2 1.2 NSAIDS 18.6 19.1 18.8 18.9 Influenza vaccination 5.6 6.3 5.9 6.0 Pneumococcal vaccination 1.0 1.3 1.1 1.1 PS, propensity score; SD, standard deviation; ACEI, Angiotensin converting enzyme inhibitors; SSRI, selective serotonin reuptake inhibitors; NSAIDS, non steroidal anti inflammatory drugs
106 Table 4 12. Continued Characteristic Unweighted Inverse of PS weighted Sildenafil (n=158,860) Vardenafil (n=49,774) Sildenafil Vardenafil Health service utilization (%) Number of outpatient visits 2.6Â±2.6 2.8Â±2.6 2.7Â±2.6 2.7Â±2.6 Number of emergency department visits 0.1Â±0.4 0.1Â±0.4 0.1Â±0.4 0.1Â±0.4 Number of hospitalization 0.04Â±0.20 0.04Â±0.19 0.04Â±0.20 0.04Â±0.20 PS, propensity score; SD, standard deviation; ACEI, Angiotensin converting enzyme inhibitors; SSRI, selective serotonin reuptake inhibitors; NSAIDS, non steroidal anti inflammatory drugs Table 4 13. Risk of SSHL comparing sildenafi l with vardenafil (primary analysis) Sildenafil Vardenafil Events 37 11 Person years 83,226 23,440 Incidence rate a 4.45 4.69 HR (95% CI) Reference 1.02 (0.51, 2.02) IRD a (95% CI) Reference 0.25 ( 2.90, 3.34) H R, hazard ratio; IRD, incidence rate difference, CI, confidence interval a. Per 10,000 person years Table 4 14. Risk of SSHL comparing sildenafil with vardenafil (secondary analyses) Sensitivity analysis H R (95% CI) Vardenafil vs. Sildenafil (reference) Take 2 PDE5 inhibitor tablets every 7 days 1.11 (0.52, 2.36) Take 1 PDE5 inhibitor tablet every 14 days 1.01 (0.55, 1.85) Take 1 PDE5 inhibitor tablet every 21 days 1.07 (0.61, 1.91) Take 1 PDE5 inhibitor tablet every 30 days 1.14 (0.66, 1.96) H R, hazard ratio; CI, confidence interval
107 Figure 4 1 . Annual p revalence and incidence of erectile dysfunction 0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Incidence of ED (per 1,000 person years) Prevalence of ED (per 1,000 subjects) Year Prevalence of ED Incidence of ED
108 Figure 4 2 . Annual p revalence and incidence of PDE5 inhibitor use 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Incidence of PDE5 inhibitor use (per 1,000 person years) Prevalence of PDE5 inhibitor use (per 1,000 subjects) Year Prevalence of PDE5 inhibitor use Incidence of PDE5 inhibitor use
109 Figure 4 3 . Market s hare of total PDE5 inhibitor products dispensed in 1998 2007
110 Figure 4 4. Flow chart (Part 2)
111 Figure 4 5. Risk of SSHL and PDE5 inhibitor use . Supplemental analysis 1: take 2 PDE5 inhibitors per week; supplemental analysis 2: take 1 PDE5 inhibitor per 14 days; supplemental analysis 3 : take 1 PDE5 inhibitor per 21 days; supplemental analysis 4: take 1 PDE5 inhibitor per 30 days; supplemental anal ysis 5: propensity score matched analysis; supplemental analysis 6: 12 month look back period; supplemental analysis 7: require SSHL cases to have evidence of treatment with steroids; supplement 8: require ED diagnosis claim for both groups; supplement ana lysis 9: adjusting for additional time dependent variables
112 Figure 4 6 . Study flowchart (Part 3 )
113 CHAPTER 5 DISCUSSION Part 1 : Patterns of PDE5 Inhibitor Use in a Commercially Insured Population Our population based estimates using a large administrative claims database showed that prevalence and incidence of ED remained stable in this private insured population over the study period from 1998 to 2007. Overall , a 1.7 fold increase in the prevalenc e of PDE5 inhibitor use and a 1.8 fold increase in the incidence of PDE5 inhibitor use were observed in this population although the estimates seemed to fluctuate between calendar years of observation. Greater utilization was consistently observed among ol der age groups , but the greatest increase in the incidence of PDE5 inhibitor use was seen in the age group of 40 49 years old . About two third of PDE5 inhibitor initiators refilled prescriptions at least once during the 6 months period after their index pr escription. In this new user cohort, the frequency of dose escalation and medication switching was low regardless of which initial drug or dosage the patient had started on. Over the last decade, ED has become increasingly recognized as a common disorder i n male adults owing to the introduction of PDE5 inhibitor medications. ED is now considered by many physicians as a marker for undiagnosed diseases such as hypertension, hyperlipidemia, and diabetes. Most previous studies that estimated the prevalence of E D relied on patient self report . For example, Selvin et al. analyzed the NHANES 2001 years old was 18.4 %. The prevalence was 5.1%, 14.8%, 43.8%, 70.2% in men aged 20 39, 40 59, 60 69, and 7 0+ years respectively. 27 The Massachusetts Male Aging Study (MMAS) found the prevalence of mild ED in men 40 69 years old was 15 %, but the
114 prevalence of moderate and severe ED was 35%. 5 Up to today, only two published studies have used healthcare utilization databases to estimate the prevalence of ED in the population. Using 1999 2002 VA administrative files, Sohn et al. reported the overall prevalence of ED was 28.8 per 1 ,000 in adult VA patients . The prevalence was 2.6, 15.1, 29.8, and 35.9 per 1,000 in men aged 25 34, 35 44, 45 54, and 55 64 years, respectively. 29 Using 2007 2009 VA administrative files, another group of investigators reported the overall prevalence of ED was 5.5% among male veterans returning from Iraq and Afghanistan, and the prevalence was 3.6% and 15.7% in men a ged 18 40 and >40 years, respectively. 30 In both VA studies, ICD 9 CM diagnosis codes were used to identify ED patients from the administrative claims data, although the Sohn et al. study also used treatment procedure codes in combination of the diagnosis codes. 29,30 Our prevalence of ED estimates were similar to the findings in those VA studies, but were significantly lower than estimates obtained in the survey based studies. A possible explanation could lie in PDE5 inhibito rs reimbursement policies that focus on severe ED or presence of comorbid conditions , which might decrease overall healthcare encounters for ED. Furthermore, physicians might not code the diagnosis of ED if they did not prescribe PDE5 inhibitor medications or omit the potentially stigmatizing diagnosis altogether. Furthermore, only the first 2 diagnostic codes associated with outpatient and inpatient claims are captured in the MarketScan CCAE data. Because patients with ED usually have other comorbid condit ions such as diabetes, hyperlipidemia, and hypertension, the diagnosis of ED might be recorded in tertiary fields that were not included in the CCAE data. Third, use of a period prevalence of 6 months to capture claims data with ED diagnosis in our prevale nce definition might
115 result in smaller case ascertainment if healthcare encounters related to the condition occur infrequent. The same problem would not affect survey methods that rely on patient recall. Fourth, lack of an explicit definition of the clinic al diagnosis, lab or imaging tests, and very poor specificity of the behavioral diagnostic criteria for ED might further explain the differences between self reports and ICD 9 codes based algorithm to identify ED cases in the claims. In conclusion, the cla ims data probably have captured moderate or severe ED cases at the cost of missing mild cases, resulting in an underestimate of ED. T he MMAS is the only published study that reported incidence of ED . 134 During an average of 8.8 years f ollow up, the incidence of ED was found in the MMAS to be 12.4 (9.0 16.9), 29.8 (24.0 37.0), 46.4 (36.9 58.4) per 1,000 person years in men aged 40 49, 50 59, 60 69 years, respectively. However, our estimates in the older age groups seemed to be lower th an the findings from the MMAS. The MMAS survey involved a community based, random sample of non institutionalized men aged 40 70 years living in the Boston area, whereas we studied a group of men with employer based private insurance. It is likely that emp loyees in the MarketScan data represented a healthier subset than the general population in local communities in Boston. A study conducted by Goetzel et al. compared the prevalence of 8 behavioral and biometric risk factors in the MarketScan CCAE with that in the National Health Interview Survey (NHIS) and the National Health and Nutrition Examination Survey (NHANES), the two well known surveys conducted by the CDC representing the health status in the general US population. The study found that blood press ure, cholesterol, blood glucose, stress, alcohol use, and tobacco use risk prevalence rates were all lower in the MarketScan
116 sub sample than in the general US sample . 135 Finally, our estimation of a lower incidence of ED might also result from an under representative of the elderly people since our CCAE data does not include men >64 years old . Few studies have est imated the prevalence and incidence of PDE5 inhibitor use. Delate et al. estimated the prevalence of sildenafil use in commercially insured, adult claims from Express Scripts, I nc. The authors found the overall prevalence of sildenafil use increased from 0.77% in 1998 to 1.42% in 2002. Although the prevalence of use increased in older patients, the fastest growing segment of users was in males aged 18 45 years suggesting young ad ults were more likely to seek care or they were more likely to uptake a new therapy than the elderly population. 136 Cooke et al. reported the prevalence of sildenafil use was 54.1 per 1,000 male members of a large managed care organization with a coverage limit of 6 doses of PDE5 inhibitor for every 30 day supply. 82 The mean age of PDE5 inhibitor users in t hose two studies was comparable to the MarketScan population. Similar to these two and other studies, 82,136,137 we found the majority of subjects who filled PDE5 inhibitor prescriptions were between the ages of 40 64 years . Unlike the other claims based studies, we studied drug utilization behavior in more recent calendar years and found a slowing increase in prevalence and incidence of use . Furthermore, unlike previous studies that focused on a single in surer, we used the MarketScan data which contains multiple types of health insurance plans. Our study population grew dramatically over the 10 years period which may result in a different composition of the study cohort in terms of the types of health insurance. Different health plans may have different policy re garding accessibility of health plan
117 members to reimbursable PDE5 inhibitors, which may be a major driving factor for the different estimates. 138 We observed about one third of initiators of PDE5 inhibitor f illed their index PDE5 inhibitor prescription only. Using pharmacy claims data from a third party administrator of prescription drugs, Mulhall et al. reported that 52% of sildenafil initiators refilled their prescription during the 6 months after their ini tial PDE5 inhibitor prescription. 103 The refill rate was 30% and 29% for vardenafil and tadalafil initiators. Cooke e t al. found that 34.1% of male members aged 18 years or older filled only one sildenafil prescription in 2001 from their MCO where a strict quantity limit was implemented. 82 Simila r to what was found in Mulhall et al. study, we found a small percentage of patients who switched medication during the 6 months period after the index PDE5 inhibitor prescription, and the frequency of dose titration was very low for all three PDE5 inhibit or agents and initial dose of prescription. In Cooke et al. study, the number of sildenafil prescription filled was 3.3Â±2.7 in calendar year 2001. Correspondingly, the mean number of sildenafil tablets filled per health plan member was 29.4 per year, or 2. 5 tablets per months. 82 In our study, the mean number of tablets filled was 23, 20, 21 per the 6 months period, or 3.8, 3.3, or 3.5 tablets per month for patients who started on si ldenafil, vardenafil and tadalafil respectively. If follow up was increased to 12 months we estimated 3.0, 2.6, and 2.8 tablets per months for sildenafil, vardenafil and tadalafil respectively, similar to those estimates obser ved by Cooke et al . There were several limitations of the data presented. First, the MarketScan CCAE database reflects the commercially insured population and, therefore, the study over presents healthier patients and excludes individuals 65 years or older and d isabled
118 and the low income populations. Second, claims data reflect only dispensing events submitted for reimbursement in the outpatient setting and neglects medication that was purchased without reimbursement. However, the MarketScan CCAE data captures data from all retail and mail order dispensing regardless of which healthcare provider in the information were not available in the claims database. Therefore, it was not possi ble to know whether subjects stopped or switched treatment because of lack of efficacy or lack of tolerability. Part 2 : PDE5 Inhibitor Use and Risk of S S In this large cohort study that involved almost 2.5 million privately insured male adults in the US, we found the crude incidence rate for SSHL was 4.27 per 1 ,000 person years among current users of PDE5 inhibitor. The result of the primary analysis which revealed an adjusted HR of 1.22 (95% CI 1.01 1.49) was confirmed in most secondary analyses. T ogether, t hese results suggest a potential small but significantly increased risk of SSHL among current users of PDE5 inhibitors. However, the absolute excess risk of SSHL comparing current user with nonuse of PDE5 i nhibitor was only 1.97 cases per 1 ,000 person year s, hence a very large populati on would need to use the PDE5 inhibitor treatment in order to have one excess case of SSHL . Although this study was conducted using data from 1998 2007, the three PDE5 inhibito rs that were commonly used in our study cohort continued to be the most popularly used drugs today, and thus the estimated excess risk of SSHL associated with these medications are still applicable to current practice . Our study has two important implicat ions. First, we found a small but significantly increased risk of SSHL comparing current PDE5 inhibitor use with nonuse. Th is large
119 post marketing observational safety study largely confirmed the safety signal detected in previous case reports sugge sting a n ototoxic effect of PDE5 inhibitor medications . In addition, w e showed that large healthcare administrative database is another useful tool that can be used to study drug related ototoxic reaction. The drug safety issue related to ototoxicity has not been addressed frequently in the pharmacoepidemiology literature. Our findings should be considered in the context of previous research. All three PDE5 inhibitors have been implicated as a cause of sudden hearing impairment in previous case report s. 15,20,72,74,139 141 In addition to the published case reports, researchers also analyzed pharmacovigilance data collected by drug safety regulato rs in North America, Australia and Europe, which revealed a number of unique features that seem to support the PDE5 inhibitor SSHL hypothesis. 74 First, a s trong temporal relationship between PDE5 inhibitor consumptio n and onset of hearing problem was observed in many previous case repo rts. 74 Of the 27 cases that underwent evaluation, 18 (66.7%) had experienced hearing loss within 24 hours after ingestion of a PDE5 inhibitor. Sec ond, the majority (88%) of cases was reported as having unilateral hearing impairment, which is a typical characteristic for idiopathic SSHL according to literature. 23 Third, two ED patients w ho used PDE5 inhibitor reported a positive re challenge which is considered a strong supporting evidence in any algorithm (e.g., WHO UMC system 142 ) for causal assessments of case reports. 16 However, a rguments put forward against the PDE5 SSHL association include the small number of case reports relative to the history of using these drugs , the relative low incidence of bilateral cases because other known ototoxic dru gs (e.g., antibiotic, antineoplastic agents induced ototoxicity) usually cause bilateral s ymmetrical symptoms . 143,144 Finally, some cases had one or
120 more SSHL risk factors such as CVD disease and concomitant use of other medications which could be the cause of their hearing impairment. 20 Besides those case reports, SSHL cases were also reported in phase III randomized trials of these drugs, although the number of cases was too small to allow any meaningful inference. 16 Finally, two mice studies provided further evidence in support of a causal association between PDE5 inhibitors and SSHL. In the first study, researchers from South Korea examined hearing in mice that were administered high dose sildenafil for up to 105 days. The authors found that high dose and long term administration can cause a hearing threshold shift of auditory brainstem response (ABR), latency delay of A BR and auditory middle latency response (AMLR), and changes in transient evoked otoacoustic emissions (TEOAE). Based on these results, the authors postulated that outer hair cells in the organ of Corti and auditory neural conductance were impaired by silde nafil use. 70 In the second study, Bakir et al . examined the inner ear biopsy specimens from mice administrated with sildenafil. The study provided histopathologic evidence of PDE5 inhibitor induced hearing impairment through apoptotic or necrotic cell deaths in the cochlea. The authors hypothesized t hat long term sildenafil administration could be the cause of hearing impairment through increased (hair) cell apoptosis . 71 Although the exact underlying biological mechanism is unclear, there are several plausible physiologic pathways which may explain the potential ototoxic effect of PDE5 inhibitors. 20,71,74 Previous studies demonstrate the presence of nitric oxide (NO) in the cochlea and the ro le of the NO/cGMP signaling pathway as a regulatory system in cochlea physiology. The cGMP can induce gene expression of nuclear factor B (NF B), a transcription factor that mediates many deleterious effects such as oxidative
121 stress and endothelial dysfun ction. It has been by localized inflammation or cochlea infarction secondary to the endothelial dysfunction or thrombosis is a potential mechanism for sudden deafness . Consistent with this s the fact that the majority of patients seen in the case reports had a unilateral hearing loss. The cellular stress etiology has also been used to explain other otologic disorders such as aminoglycoside ototoxicity. 77,145,146 An alternative pathway that involves the NO and downstream inflammatory cytokines including MAP kinase (MAPK) and c Jun N terminal kinase (JNK) have been found to cause loss of auditory hair cell via nec rosis and apoptosis . 20 Blocking the MAPK/JNK pathway can be otoprotective against aminoglycoside ototoxicity, neomycin ototoxicity, and acoustic trauma induced auditory hair cell death . 145,147 Although our primary finding of a small increased relative risk of SSHL was robust in most secondary analyses, the results must be interpreted with caution because of the potential limitations of the present study. We employed an electronic administrative claims database to study a drug adverse effect. PDE5 inhibitors are taken a s needed, thus actual time and dose of exposure is unknown based on pharmacy dispensing data. We simulated the duration of exposure by imputing the frequency of use from previous surveys . This approach, however, would systematically overestimate the intens ity of current use, if the patient did not consume all his prescribed pills (e.g., 4 6 tablets for a 30 day supply) at the designated frequency (e.g., use 1 PDE5 inhibitor dose per week) or if he splits the pills and used his stockpile later after the end have been considered as exposed while actually he was not, resulting in an
122 underestimation of current use and overestimation of recent use or nonuse. Therefore, the overall effect o f misclassification would have reduced the true difference between periods of current use and nonuse, and would result in underestimation of the true effect of the drug. Alternatively, if a patient split the pills but he still consumed all his prescribed d estimate for the current use would not be affected. In addition, PDE5 inhibitors could be obtained via the internet or purchased out of pocket leading to further misclassificatio. 148 We defined exposure groups based on imputed data o n duration of PDE5 inhibitor use . We found th at HR was larger during recent use than current use in the primary analysis (assuming patients took 1 PDE5 inhibitor dose every 7 days) . A possible explanation for this finding was th at th e frequency of PDE5 inhib itor use was incorrectly specified in the primary analysis . W e conducted several sensitivity analyses by assuming different patterns of drug use. Overall , we still found a small (20 40%) but significantly increased risk of SSHL comparing current use to non use in these analyses . Additionally, we found the HR for current use increased , and at the same time the HR for recent use decreased if we assumed a longer duration of current use. In fact, the HR for current use was higher than the HR for recent use when we assumed patients took 1 PDE5 inhibitor dose every 21, or 30 days, suggesting some person time might have been mistakenly assigned from current use to recent use in the prim ary analysis . Alternatively, a significant delay in seeking health care for SSHL might also result in a higher risk identified during recent use than current use. However, this explanation was unlikely because we added 30 days of follow up to the calculate d end of days supply in
123 current use already, this should have taken care of the issue properly especially some authors suggested the average delay in seeking specialist care for SSHL is 20 days. Furthermore, confounding is a major threat to study validity . ED and SSHL share many common risk factors including cardiovascular disease, diabetes, hyperlipidemia, smoking, and obesity. We considered a variety of risk factors of SSHL, which could act as potential confounders of the association, and included them i n the propensity score model. Adjusting for exposure propensity scores in outcome models is an effective and efficient method to control confounding bias and is especially attractive when the outcome of interest is rare. 96,149 Although we controlled for a large number of confounders, we cannot rule out the possibility of uncontrolled (e.g., smoking, physical activity, diet, OCT drug use) or inappropriately measured factors (e.g., obesity) resulting in residu al confounding that might affect our results. Another limitation of the study is the potential for outcome misclassification. The accuracy of the specific diagnosis and procedure codes to identify SSHL in claims data has not been formally validated. Ho wever, the selected diagnosis and procedure codes are highly specific for SSHL . It is particularly important to note that we required two CPT codes for audiologic testing, mimicking the clinical diagnostic process to increase the specificity of our definit ion . Mild SSHL cases may recover spontaneously, which may result in an under ascertainment of cases, but as long as the health seeking behavior of exposed and unexposed patients was similar, lower sensitivity would not bias our results. We further introduc ed the requirement for steroid treatment in the sensitivity analysis, which showed similar risk estimates, assuring us that misclassification bias is of limited concern.
124 Finally, our study finding may have limited generalizability as addressed earlier, bu t it is unlikely that the discovered PDE5 risk occurs differentially in healthier and younger subjects. Thus, similar or higher risk might exist in patients covered by public insurance plans. In conclusion, in this large cohort of privately insured male adults, subjects who were newly started on PDE5 inhibitor medications had a small but significantly increased risk of SSHL compared with subjects who were not receiving these medications. Together with other emerging evidence on the topic, our study suppor ts the notion that use of PDE5 inhibitor is associated with an elevated risk of SSHL . In this affluent population, the absolute risk of PDE5 inhibitor exposure was low. Thus, our of a more rigorous labeling but retaining these products on the market for ED treatment . 21 Finally, given the widespread use of these medications, even the low risk of ototoxicity might pose a potential public health concern. Thus, prescribers and patients need to weigh the potentially increased ototoxicity from PDE5 inhibitor agents against the benefits that arise from treatment. Part 3 : Comparative Safety of PDE5 Inhibitors and Risk of S S HL In this large cohort of ma le adults who initiated PDE5 inhibitors between 2003 and 2007, we did not find clinically and statistically meaningful differences in the risk of SSHL comparing the three PDE5 inhibitors. Occurrence of SSHL in either PDE5 inhibitor new user cohort was rare in this population with a total of only 61 cases, leading to wide confidence intervals of hazard ratios and limited power to compare the three agents. Only one previous study compared the risk of sensorineural hearing loss in patients taking different PD E5 inhibitors. The study , which involved 11,252 men aged 40
125 years or older who were enrolled in the Medical Expenditure Panel Survey (MEPS) between 2003 and 2006, found that men who reported having hearing impairment were more likely to have also reported the use of PDE5 inhibitors (OR=2.23, 9 5% CI, 1.36 3.66). In addition, a significant association was prominent in sildenafil users only, but not in vardenafil or tadalafil users. The authors admitted that the differential ototoxic effect may be due to the s mall sample size for vardenafil and tadalafil. The study was a cross report on drug exposure and occurrence of outcomes, thus recall bias and potential for exposure and outcome misclassification were additional limitations. 75 Sever al case reports and small follow up studies (n<50) have been published that describe a risk of SSHL associated with PDE5 inhibitor use. 8,15,20,72 74,139,140 Systematic reviews that pooled these studies, including the original randomized clinical trials, suggest that the three agents have similar efficacy and safety profile. 13,120,150 The three PDE5 inhibitors share the same mechanism of action and sildenafil and vardenafil have similar pharmacokinetics profiles. More importantly, the suggested biological . As in part 2, misclassification of exposure was a major limitation of this study and our ability to compare results of our sensitivity analysis varying exposure definitions was limited due to wide confidence intervals. As this was an observational study, residual confounding cannot be rule d out, but we adjusted propensity score for potential differences in baseline patient characteristics across different exposure groups , which should have taken care of the confounding bias to a large extent . Finally, the diagnosis of SSHL using claims reco rds has not been formally validated against -
126 audiology results . M isclassification of outcome should not differ according to the use of PDE5 inhibitor , and therefore such misclassification would be expected to bias ou r results towards masking a potential existing association. In a large cohort of male adults who were newly started on PDE5 inhibitor, we found no significant differences in the risk of SSHL between users of PDE5 inhibitors. However, despite the use of a l arge insurance claims database, we had limited power in estimating the risk of SSHL associated with use of PDE5 inhibitors. Given the widespread use of these drugs future studies conducted in a larger population would be beneficial to provide confirmation of our findings. Recommendation for future research A major deficiency in the literature is how the PDE5 inhibito rs are actually used in . Large electronic databases provide one useful tool to investigate the PDE5 inhibitor utilization patte rn. But PDE5 inhibitors could be purchased online , and over the counter sales are common. 151 Survey based research can supplement the claims based studies to better understand the pattern of PDE5 inhibitor use. Such information could be important in designing future safety studies for the se drugs . In the current study, the PDE5 inhibitor exposure frequency has been selected arbitrarily based on a few surveys that involved small number of patient population. As discussed in the thesis, we were unable to access to real audiologic data which was considered as a major limitation of the study. A validation study should be conducted to assess the accuracy of the computer based case identification algorithm used in the current study. The insight gained from such validation study will a llow investigators to better plan future research in this challenging area of drug induced ototoxicity.
127 In a comparative safety study, the greatest concern for a non significant finding was the potential for limited cohort size leading to imprecise estim ates, hence masking an existing safety signal. In addition, more than 50 different health plans contributed to the claims data in the MarketScan CCAE data. Different health plan has its own reimbursement policy for lifestyle drugs such as PDE5 inhibitors. We have shown in our analyses that in contrary to the notion that prevalence of ED increased in real life population after the introduction of PDE5 inhibitors, we actually found the prevalence and incidence of ED in our population to be fairly stable over the 10 year study period. One possible explanation is that this could be the impact of changing component of patient population and health plan type in the study. So i f feasible, future studies should repeat the analysis of the comparative risk of ototoxic ity for PDE5 inhibitor agents in a larger unique patient population. With several PDE5 inhibitor products already being on the US market and several other PDE5 inhibitors are currently being tested, the comparative safety and effectiveness data beco me more clinically relevant when there are more than two treatment options available. Summary and conclusion Our study on the association between PDE5 inhibitor use and the risk of SSHL at the population level revealed a number of important results: In part 1, we found that among privately insured male adults the annual prevalence and incidence of ED remained largely stables in 1998 2007, although prevalence and incidence of PDE5 inhibitor use increased slightly during the same time period. Higher utilization r ates were found when people become older. These estimates changed from year to year and were possibly affected by the change of component of
128 study population as well as health plans available in the MarketScan CCAE data. About two third of the new users of PDE5 inhibitors refilled their prescription after the index prescription, only a small population of these new users ever changed their dosage or switched to a different PDE5 inhibitor during the follow up. In part 2, we found a small ( about 20 to 40%) bu t significantly increased risk of SSHL comparing current use of PDE5 inhibitors with nonuse of PDE5 inhibitors , however, the absolute excess risk of SSHL due to PDE5 inhibitor use was small (1.97 cases per 1,000 exposed person years in the primary analysis ) indicating that a large number of male adults need to be exposed to the drugs in order to produce the ototoxic effects among patients . We tried different analytic approaches to mitigate the effects of exposure misclassification, outcome misclassification , confounding by indication, and residual confounding in a number of sensitivity analyses. In most of these sensitivity analyses, we still foun d a small increased risk of SSH . Finally in part 3, no meaningful differences were observed for the risk of SSHL associated with the three individual PDE5 inhibitor agents, although the analysis was limited by lack of statistical power. This is the first population based study that quantified the risk of SSHL in association with PDE5 inhibitor agents in a large coho rt of patients. Methods developed in this study provided new insights into the methodology to assess the risk of adverse events of PDE5 inhibitors as well as drug induced ototoxicity in large healthcare administration databases in the future. These findings may have important implicatio ns for physicians and patients . Sudden deafness although it is rare, it is considered as an otologic and audiologic emergency. If not properly diagnosed and treated, the hearing may remain impaired for a long time or i n some cases may even be permanent. 23 Our
129 finding of a small incr eased relative risk is convincing and biologically plausible. Hence, p rescribers of PDE5 inhibitors need to provide their patients with appropriate information on the benefits and risks of the treatment and how the se drugs may impact overall qual ity of life. We have addressed an otologic emergent condition in this thesis, other authors have shown a possible link between the use of PDE5 inhibitors and nonarteritic anterior ischaemic optic neuropathy (NAION) . 152,153 Since thousands and millions of patients are taking these drugs, the potential public health impact of these adverse events could be significant. Hence, the updated safety information needs to be properly disseminated to patients and physicians who are using these drugs frequently.
130 APPENDIX A INCLUSION EXCLUSION CRITERION, APPLIED AS TO T 0 , THE PRESCRIPTION FILL DATE OR COMPARABLE DATE FOR NONUSER CONTROLS 1. Aged 18 6 4 years at T 0 . Individuals <18 years have less chance to be prescribed PDE5 inhibitors and their risk of years. The upper age limit is present because people become eligible for Medicare after 65 years old. Thus, both exposure and outcome information are underrepre sented in our study database. 2. Enrolled in a health pl an that provides full pharmacy benefits. 3. At least 365 days continuous health plan enrollment. This means, subject needs to be enrolled on the 365 days prior to T 0 . This assures sufficient enrollment to use medical encounters as measure of comorbidity. 4. At l east one filled prescription and one in or out patient encounte r resulting a diagnosis in each of the two consecutive 6 months period preceding T 0 . This assures that cohort members have some ongoing medical surveillance, particularly near T 0 . 5. No exclusio n illness on T 0 or the preceding 365 days. This means, without well established medical conditions or medication use that can cause SNHL (i.e., conductive hearing loss, SNHL, mixed hearing loss, HIV, cancer, organ transplant, syphilis, aminoglycoside, anti retroviral therapy, chemotherapy) 6. Only a single s tudy drug prescription filled at the T 0 .
131 APPENDIX B EXCLUSION ILLNESSES APPLIED AS TO T 0 Unless otherwise specified, these are ICD9 CM diagnostic codes (3 rd or 4 th digit code implies inclusion of all sub codes) or National Drug Index (NDC) codes. Hence, from the pool of eligible patients, subjects were excluded if they had one of the following co nditions during the 183 baseline period (including the index date) that could have influenced their ris k of SSHL. Conditions Definition Conductive hearing loss At least one in or out patient claim with ICD9 CM diagnosis codes: 389.0x Sensorineural hearing loss At least one in or out patient claim with ICD9 CM diagnosis codes: 389.1x Mixed hearing loss At least one in or out patient claim with ICD9 CM diagnosis codes: 389.2x Hearing loss unspecified At least one in or out patient claim with ICD9 CM diagnosis codes: 389.7, 389.8, 389.9 Cancer (except prostate, bladder cancer) At least one in o r out patient claim with ICD9 CM diagnosis codes: 140 208, 209.0x 209.3, 230 234 Organ transplant At least one in or out patient claim with ICD9 CM diagnosis codes: 996.8x, 238.77, E878.0, V42.x, V43.2x, V43.3x, V43.4x, V43.6x, V45.87, V49.83, V58.44 HIV At least one in or out patient claim with ICD9 CM diagnosis codes: 279, 042, 795.71, V08 Cytomegalovirus At least one in or out patient claim with ICD9 CM diagnosis codes: 771.1, 078.5 Rubella At least one in or out patient claim with ICD9 CM diagnosis codes: 056.x, 771.0 Syphilis At least one in or out patient claim with ICD9 CM diagnosis codes: 090.x, 091.x, 092.x, 093.x, 094.x, 095.x, 096, 097.x Bacterial meningitis At least one in or out patient claim with ICD9 CM diagnosis codes: 320.x Viral encephalitis At least one in or out patient claim with ICD9 CM diagnosis codes: 062.x, 063.x, 064, 054.3, 055.0, 072.2 Severe head injury At least one in or out patient claim with ICD9 CM diagnosis codes: 800.x, 801.x, 802.x, 803.x, 804. x, 850.xx, 851.x, 852.x, 853.x, 854.x Head or neck radiation At least one in or out patient claim with ICD9 CM diagnosis codes: E926.x, V15.3 Pulmonary hypertension At least one in or out patient claim with ICD9 CM diagnosis codes: 415.0, 416.0, 41 6.8 Systemic aminoglycosides, interferon, cisplatin, cyclosporine, vinblastine, or vincristine At least one drug claim filled during the 183 days baseline period or at any time during their follow up
132 APPENDIX C DIAGNOSIS AND PROCEDURE CODES USED IN THE STUDY TO IDENTIFY STUDY OUTCOME Case definition ICD 9 CM codes CPT codes* Sudden sensorineural hearing loss At least 1 in or out patient claim with ICD 9 CM diagnosis codes for sensorineural hearing loss, plus 1 CPT code for any audiometric testing during [ 30 ,0] days and 1 CPT code for any audiometric testing during [+4, +30] days 389.1x, 389.2x, 388.2 92553, 92557, 92579, 92582, 92585, 92586, 92588 In the secondary analysis, require evidence of treatment with oral steroid or intra tympanic steroid injection 69799, 69801, 69433, 69436
133 LIST OF REFERENCES 1. NIH Consensus Conference. Impotence. NIH Consensus Development Panel on Impotence. JAMA 1993;270:83 90. 2. Ayta IA, McKinlay JB, Krane RJ. The likely worldwide increase in erectile dysfunction between 1995 and 2025 and some possible policy consequences. BJU Int 1999;84:50 6. 3. Johannes CB, Araujo AB, Feldman HA, Derby CA, Kleinman KP, McKinlay JB. Incidence of erectile dysfunction in men 40 to 69 years old: longitudinal results from the Massachusetts male aging study. J Urol 2000;163:460 3. 4. Jackson G, Betteridge J, Dean J, et al. A systematic approach to erectile dysfunction in the cardiovascula r patient: a Consensus Statement -update 2002. Int J Clin Pract 2002;56:663 71. 5. Feldman HA, Goldstein I, Hatzichristou DG, Krane RJ, McKinlay JB. Impotence and its medical and psychosocial correlates: results of the Massachusetts Male Aging Study. J Ur ol 1994;151:54 61. 6. Hackett G. The burden and extent of comorbid conditions in patients with erectile dysfunction. Int J Clin Pract 2009;63:1205 13. 7. Shindel AW. 2009 update on phosphodiesterase type 5 inhibitor therapy part 1: Recent studies on rout ine dosing for penile rehabilitation, lower urinary tract symptoms, and other indications (CME). J Sex Med 2009;6:1794 808; quiz 3, 809 10. 8. Giuliano F, Jackson G, Montorsi F, Martin Morales A, Raillard P. Safety of sildenafil citrate: review of 67 double blind placebo controlled trials and the postmarketing safety database. Int J Clin Pract 2010;64:240 55. 9. Lowe G, Costabile RA. 10 Year analysis of adverse event reports to the Food and Drug Administration for phosphodiesterase type 5 inhibitors. J Sex Med 2012;9:265 70. 10. Francis SH, Lincoln TM, Corbin JD. Characterization of a novel cGMP binding protein from rat lung. J Biol Chem 1980;255:620 6. 11. Schultheiss D, Stief CG. Physiology and pathophysiology of erection: consequences for present medical therapy of erectile dysfunction. Andrologia 1999;31 Suppl 1:59 64. 12. Gupta M, Kovar A, Meibohm B. The clinical pharmacokinetics of phosphodiesterase 5 inhibitors for erectile dysfunction. J Clin Pharmacol 2005;45:987 1003.
134 13. Yuan J, Zhang R, Yang Z, et al. Comparative Effectiveness and Safety of Oral Phosphodiesterase Type 5 Inhibitors for Erectile Dysfunction: A Systematic Review and Network Meta analysis. Eur Urol 2013. 14. Tsertsvadze A, Fink HA, Yazdi F, et al. Oral phosphodiesterase 5 in hibitors and hormonal treatments for erectile dysfunction: a systematic review and meta analysis. Ann Intern Med 2009;151:650 61. 15. Mukherjee B, Shivakumar T. A case of sensorineural deafness following ingestion of sildenafil. J Laryngol Otol 2007;121:3 95 7. 16. FDA NEWS RELEASE. 2007. (Accessed January 6, 2013, at http://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInformationfo rPatientsandP roviders/ucm124841.htm. ) 17. Amaee FR, Comis SD, Osborne MP, Drew S, Tarlow MJ. Possible involvement of nitric oxide in the sensorineural hearing loss of bacterial meningitis. Acta Otolaryngol 1997;117:329 36. 18. Heinrich UR, Helling K, Sif ferath M, et al. Gentamicin increases nitric oxide production and induces hearing loss in guinea pigs. Laryngoscope 2008;118:1438 42. 19. Watanabe KI, Hess A, Bloch W, Michel O. Nitric oxide synthase inhibitor suppresses the ototoxic side effect of cispla tin in guinea pigs. Anticancer Drugs 2000;11:401 6. 20. Maddox PT, Saunders J, Chandrasekhar SS. Sudden hearing loss from PDE 5 inhibitors: A possible cellular stress etiology. Laryngoscope 2009;119:1586 9. 21. Phosphodiesterase type5 (PDE5) inhibitors: reports of sudden hearing loss associated with use of sildenafil citrate (marketed as Viagra and Revatio), vardenafil hydrochloride (marketed as Levitra), and tadalafil (marketed as Cialis). Drug Safety Newsletter2008:15 7. 22. Stachler RJ, Chandrasekhar SS, Archer SM, et al. Clinical practice guideline: sudden hearing loss. Otolaryngol Head Neck Surg 2012;146:S1 35. 23. Rauch SD. Clinical practice. Idiopathic sudden sensorineural hearing loss. N Engl J Med 2008;359:833 40. 24. Winterstein AG, Liu W, Xu D, Antonelli PJ. Sensorineural Hearing Loss Associated with Neomycin Eardrops and Nonintact Tympanic Membranes. Otolaryngol Head Neck Surg 2012. 25. Klein R, Sturm H. Viagra: a success story for rationing? Health Aff (Millwood) 2002;21:177 87.
135 26. Bacon CG, Mittleman MA, Kawachi I, Giovannucci E, Glasser DB, Rimm EB. Sexual function in men older than 50 years of age: results from the health professionals follow up study. Ann Intern Med 2003;139:161 8. 27. Selvin E, Burnett AL, Platz EA. Prevalence and ri sk factors for erectile dysfunction in the US. Am J Med 2007;120:151 7. 28. Laumann EO, West S, Glasser D, Carson C, Rosen R, Kang JH. Prevalence and correlates of erectile dysfunction by race and ethnicity among men aged 40 or older in the United States: from the male attitudes regarding sexual health survey. J Sex Med 2007;4:57 65. 29. Sohn MW, Zhang H, Taylor BC, et al. Prevalence and trends of selected urologic conditions for VA healthcare users. BMC Urol 2006;6:30. 30. Hosain GM, Latini DM, Kauth M, Goltz HH, Helmer DA. Sexual dysfunction among male veterans returning from Iraq and Afghanistan: prevalence and correlates. J Sex Med 2013;10:516 23. 31. Meuleman EJ, Donkers LH, Robertson C, Keech M, Boyle P, Kiemeney LA. [Erectile dysfunction: prevalen ce and effect on the quality of life; Boxmeer study]. Ned Tijdschr Geneeskd 2001;145:576 81. 32. Sun P, Seftel A, Swindle R, Ye W, Pohl G. The costs of caring for erectile dysfunction in a managed care setting: evidence from a large national claims databa se. J Urol 2005;174:1948 52. 33. Plumb JM, Guest JF. Annual cost of erectile dysfunction to UK Society. Pharmacoeconomics 1999;16:699 709. 34. Basson R, Schultz WW. Sexual sequelae of general medical disorders. Lancet 2007;369:409 24. 35. Bacon CG, Mitt leman MA, Kawachi I, Giovannucci E, Glasser DB, Rimm EB. A prospective study of risk factors for erectile dysfunction. J Urol 2006;176:217 21. 36. Basson R. Sexuality in chronic illness: no longer ignored. Lancet 2007;369:350 2. 37. Stein RA. Endothelial dysfunction, erectile dysfunction, and coronary heart disease: the pathophysiologic and clinical linkage. Rev Urol 2003;5 Suppl 7:S21 7. 38. Williams G, Mulcahy MJ, Hartnell G, Kiely E. Diagnosis and treatment of venous leakage: a curable cause of impote nce. Br J Urol 1988;61:151 5.
136 39. Safarinejad MR, Taghva A, Shekarchi B, Safarinejad S. Safety and efficacy of sildenafil citrate in the treatment of Parkinson emergent erectile dysfunction: a double blind, placebo controlled, randomized study. Int J Impo t Res 2010;22:325 35. 40. Keller JJ, Liang YC, Lin HC. Association between multiple sclerosis and erectile dysfunction: a nationwide case control study. J Sex Med 2012;9:1753 9. 41. Mendenhall WM, Henderson RH, Indelicato DJ, Keole SR, Mendenhall NP. Ere ctile dysfunction after radiotherapy for prostate cancer. Am J Clin Oncol 2009;32:443 7. 42. McVary K. Lower urinary tract symptoms and sexual dysfunction: epidemiology and pathophysiology. BJU Int 2006;97 Suppl 2:23 8; discussion 44 5. 43. KÃ¶hler TS, Mc Vary KT. The relationship between erectile dysfunction and lower urinary tract symptoms and the role of phosphodiesterase type 5 inhibitors. Eur Urol 2009;55:38 48. 44. Kupelian V, Hall SA, McKinlay JB. Common prescription medication use and erectile dysf unction: results from the Boston Area Community Health (BACH) survey. BJU Int 2013. 45. Manolis A, Doumas M. Antihypertensive treatment and sexual dysfunction. Curr Hypertens Rep 2012;14:285 92. 46. Hanson Divers C, Jackson SE, Lue TF, Crawford SY, Rosen RC. Health outcomes variables important to patients in the treatment of erectile dysfunction. J Urol 1998;159:1541 7. 47. Lue TF, Giuliano F, Montorsi F, et al. Summary of the recommendations on sexual dysfunctions in men. J Sex Med 2004;1:6 23. 48. Hir sch M, Donatucci C, Glina S, Montague D, Montorsi F, Wyllie M. Standards for clinical trials in male sexual dysfunction: erectile dysfunction and rapid ejaculation. J Sex Med 2004;1:87 91. 49. Montague DK, Jarow JP, Broderick GA, et al. Chapter 1: The man agement of erectile dysfunction: an AUA update. J Urol 2005;174:230 9. 50. Gur S, Kadowitz PJ, Serefoglu EC, Hellstrom WJ. PDE5 inhibitor treatment options for urologic and non urologic indications: 2012 update. Curr Pharm Des 2012;18:5590 606. 51. Hatzi mouratidis K, Hatzichristou DG. A comparative review of the options for treatment of erectile dysfunction: which treatment for which patient? Drugs 2005;65:1621 50.
137 52. Coombs PG, Heck M, Guhring P, Narus J, Mulhall JP. A review of outcomes of an intracavernosal injection therapy programme. BJU Int 2012;110:1787 91. 53. Campbell HE. Clinical monograph for drug formulary review: erectile dysfunction agents. J Manag Care Pharm 2005;11:151 71. 54. Viagra (Sildenafil Citrate) Prescribing Information. (Accessed April 1, 2013, at http://www.accessdata.fda.gov/drugsatfda_docs/label/2007/020895s027lbl.pdf. ) 55. L evitra (Vardenafil HCl) Prescribing Information. at http://www.accessdata.fda.gov/drugsatfda_docs/label/2007/021400s010lbl.pdf. ) 56. Cialis (Tadalafil) Prescribing Information. 57. Bender AT, Beavo JA. Cyclic nucleotide phosphodiesterases: molecular regulation to clinical use. Pharmacol Rev 2006;58:488 520. 58. Omori K, Kotera J. Overview of PDEs and their regulation. Circ Res 2007;100:309 27. 59. Hedelin H, Str Ã¶berg P. Treatment for erectile dysfunction based on patient reported outcomes: to every man the PDE5 inhibitor that he finds superior. Drugs 2005;65:2245 51. 60. Rosen RC, Riley A, Wagner G, Osterloh IH, Kirkpatrick J, Mishra A. The international index o f erectile function (IIEF): a multidimensional scale for assessment of erectile dysfunction. Urology 1997;49:822 30. 61. Wright PJ. Comparison of phosphodiesterase type 5 (PDE5) inhibitors. Int J Clin Pract 2006;60:967 75. 62. Shindel AW. 2009 update on phosphodiesterase type 5 inhibitor therapy part 2: updates on optimal utilization for sexual concerns and rare toxicities in this class. J Sex Med 2009;6:2352 64; quiz 65 6. 63. Hatzimouratidis K. Phosphodiesterase type 5 inhibitors, visual changes, and n onarteritic anterior ischemic optic neuropathy: is there a link? Curr Urol Rep 2007;8:482 90. 64. Mulhall JP, Montorsi F. Evaluating preference trials of oral phosphodiesterase 5 inhibitors for erectile dysfunction. Eur Urol 2006;49:30 7. 65. Sataloff RT , Sataloff J. Hearing loss: CRC Press; 2005.
138 66. Lin RJ, Krall R, Westerberg BD, Chadha NK, Chau JK. Systematic review and meta analysis of the risk factors for sudden sensorineural hearing loss in adults. Laryngoscope 2012;122:624 35. 67. Audiologic Management of Individuals Receiving Cochleotoxic Drug Therapy. at http://www.asha.org/policy/GL1994 00003/. ) 68. Conlin AE, Parnes LS. Treatment of sudden sensorineural hearing loss: I. A systematic review. Arch Otolaryngol Head Neck Surg 2007;133:573 81. 69. Rauch SD, Halpin CF, Antonelli PJ, et al. Oral vs intratympanic corticosteroid therapy for idiopathic sudden sensorineural hearing loss: a randomized trial. JAMA 2011;305:2071 9. 7 0. Hong BN, Yi TH, Kim SY, Kang TH. High dosage sildenafil induces hearing impairment in mice. Biol Pharm Bull 2008;31:1981 4. 71. Bakir S, Firat U, GÃ¼n R, et al. Histopathologic results of long term sildenafil administration on rat inner ear. Am J Otolar yngol 2012;33:667 72. 72. Snodgrass AJ, Campbell HM, Mace DL, Faria VL, Swanson KM, Holodniy M. Sudden sensorineural hearing loss associated with vardenafil. Pharmacotherapy 2010;30:112. 73. Okuyucu S, Guven OE, Akoglu E, UÃ§ar E, Dagli S. Effect of phosphodiesterase 5 inhibitor on hearing. J Laryngol Otol 2009;123:718 22. 74. Khan AS, Sheikh Z, Khan S, Dwivedi R, Benjamin E. Viagra deafness -sensorineural hearing loss and phosphodiesterase 5 inhibitors. Laryngoscope 2011;121:1049 54. 75. McGwin G. Phosphodiesterase type 5 inhibitor use and hearing impairment. Arch Otolaryngol Head Neck Surg 2010;136:488 92. 76. JA R. Physiology of the vestibular system. In: PS R, JA R, eds. Ototoxicity. London: BC Decker Inc; 2004:20 7. 77. Fessenden JD, Coling DE , Schacht J. Detection and characterization of nitric oxide synthase in the mammalian cochlea. Brain Res 1994;668:9 15. 78. Watanabe K, Hess A, Zumeger C, et al. Changes of the compound action potential (CAP) and the expression of inducible nitric oxide s ynthase (iNOS/NOS II) in the cochlea under the inflammatory condition(1). Hear Res 2000;145:149 55.
139 79. Watanabe K, Hess A, Bloch W, Michel O. Inhibition of inducible nitric oxide synthase lowers the cochlear damage by lipopolysaccharide in guinea pigs. F ree Radic Res 2000;32:363 70. 80. Bischoff E. Pharmacology of Phosphodiesterase Inhibitors. In: Broderick G, ed. Oral Pharmacotherapy for Male Sexual Dysfunction: A Guide to Clinical Management. New Jersey: Humana Press; 2005:43 64. 81. Phosphodiesterase 5 inhibitors quantity limit. (Accessed April 20, 2013, at http://www.azblue.com/pdfs/pharmacy/PDE 5InhibitorsQuantityLimitRxPolicy.pdf. ) 82. Co oke CE, Wong W, Lee H. Utilization and cost of sildenafil in a large managed care organization with a quantity limit on sildenafil. J Manag Care Pharm 2005;11:674 80. 83. Erectile dysfunction prior authorization criteria. at http://www.bcbsil.com/pdf/pharmacy/rx_criteria/erectile_dysfunction.pdf. ) 84. Bannowsky A, Schulze H, van der Horst C, Hautmann S, JÃ¼nemann KP. Recovery of erectile function after nerve sparing ra dical prostatectomy: improvement with nightly low dose sildenafil. BJU Int 2008;101:1279 83. 85. Padma Nathan H, McCullough AR, Levine LA, et al. Randomized, double blind, placebo controlled study of postoperative nightly sildenafil citrate for the preven tion of erectile dysfunction after bilateral nerve sparing radical prostatectomy. Int J Impot Res 2008;20:479 86. 86. Montorsi F, Brock G, Lee J, et al. Effect of nightly versus on demand vardenafil on recovery of erectile function in men following bilate ral nerve sparing radical prostatectomy. Eur Urol 2008;54:924 31. 87. McVary KT, Monnig W, Camps JL, Young JM, Tseng LJ, van den Ende G. Sildenafil citrate improves erectile function and urinary symptoms in men with erectile dysfunction and lower urinary tract symptoms associated with benign prostatic hyperplasia: a randomized, double blind trial. J Urol 2007;177:1071 7. 88. Roehrborn CG, McVary KT, Elion Mboussa A, Viktrup L. Tadalafil administered once daily for lower urinary tract symptoms secondary to benign prostatic hyperplasia: a dose finding study. J Urol 2008;180:1228 34. 89. Stief CG, Porst H, Neuser D, Beneke M, Ulbrich E. A randomised, placebo controlled study to assess the efficacy of twice daily vardenafil in the treatment of lower urinary t ract symptoms secondary to benign prostatic hyperplasia. Eur Urol 2008;53:1236 44.
140 90. Enthoven AC, Fuchs VR. Employment based health insurance: past, present, and future. Health Aff (Millwood) 2006;25:1538 47. 91. Ray WA. Evaluating medication effects o utside of clinical trials: new user designs. Am J Epidemiol 2003;158:915 20. 92. Austin PC. An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies. Multivariate Behav Res 2011;46:399 424. 93. Glynn R, Schneeweiss S, StÃ¼rmer T. Indications for propensity scores and review of their use in pharmacoepidemiology. Basic Clin Pharmacol Toxicol 2006;98:253 9. 94. Johnson ES, Bartman BA, Briesacher BA, et al. The incident user design in comparative effectivene ss research. Pharmacoepidemiol Drug Saf 2013;22:1 6. 95. Schneeweiss S. A basic study design for expedited safety signal evaluation based on electronic healthcare data. Pharmacoepidemiol Drug Saf 2010;19:858 68. 96. Walker AM. Confounding by indication. Epidemiology 1996;7:335 6. 97. Prasad MM, Prasad SM, Hevelone ND, et al. Utilization of pharmacotherapy for erectile dysfunction following treatment for prostate cancer. J Sex Med 2010;7:1062 73. 98. Singh TP, Rohit M, Grover A, Malhotra S, Vijayvergiya R. A randomized, placebo controlled, double blind, crossover study to evaluate the efficacy of oral sildenafil therapy in severe pulmonary artery hypertension. Am Heart J 2006;151:851.e1 5. 99. Giuliano F, Ãœckert S, Maggi M, Birder L, Kissel J, Viktrup L. The mechanism of action of phosphodiesterase type 5 inhibitors in the treatment of lower urinary tract symptoms related to benign prostatic hyperplasia. Eur Urol 2013;63:506 16. 100. Palmieri A, Imbimbo C, Creta M, Verze P, Fusco F, Mirone V. Tadalafil o nce daily and extracorporeal shock wave therapy in the management of patients with Peyronie's disease and erectile dysfunction: results from a prospective randomized trial. Int J Androl 2012;35:190 5. 101. Gacci M, Ierardi A, Rose AD, et al. Vardenafil ca n improve continence recovery after bilateral nerve sparing prostatectomy: results of a randomized, double blind, placebo controlled pilot study. J Sex Med 2010;7:234 43. 102. Vlachopoulos C, Terentes Printzios D, Ioakeimidis N, Rokkas K, Stefanadis C. PD E5 inhibitors in non urological conditions. Curr Pharm Des 2009;15:3521 39.
141 103. Mulhall JP, McLaughlin TP, Harnett JP, Scott B, Burhani S, Russell D. Medication utilization behavior in patients receiving phosphodiesterase type 5 inhibitors for erectile d ysfunction. J Sex Med 2005;2:848 55. 104. Atkinson T. Lifestyle drug market booming. Nat Med 2002;8:909. 105. Keith A. The economics of Viagra. Health Aff (Millwood) 2000;19:147 57. 106. Conaglen HM, Conaglen JV. Couples' reasons for adherence to, or di scontinuation of, PDE type 5 inhibitors for men with erectile dysfunction at 12 to 24 month follow up after a 6 month free trial. J Sex Med 2012;9:857 65. 107. Potts A, Grace V, Gavey N, Vares T. "Viagra stories": challenging 'erectile dysfunction'. Soc S ci Med 2004;59:489 99. 108. Bechara A, CasabÃ© A, De Bonis W, Helien A, Bertolino MV. Recreational use of phosphodiesterase type 5 inhibitors by healthy young men. J Sex Med 2010;7:3736 42. 109. StrÃ¶berg P, Hedelin H, BergstrÃ¶m A. Is sex only for the healt hy and wealthy? J Sex Med 2007;4:176 82. 110. Rasgon B, Schloegel LJ. Early and accurate diagnosis of sudden sensorineural hearing loss. Perm J 2009;13:61 3. 111. Stafford RS, Radley DC. The potential of pill splitting to achieve cost savings. Am J Manag Care 2002;8:706 12. 112. Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med 1997;127:757 63. 113. Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, StÃ¼rmer T. Variable selection for propensity scor e models. Am J Epidemiol 2006;163:1149 56. 114. Harrell F. Regression modeling strategies: with applications to linear models, logistic regression and survival analysis: Springer; 2001. 115. Herndon JE, Harrell FE. The restricted cubic spline as baseline hazard in the proportional hazards model with step function time dependent covariables. Stat Med 1995;14:2119 29. 116. Lin DY, Wei LJ, Ying Z. Model checking techniques based on cumulative residuals. Biometrics 2002;58:1 12. 117. Hade EM, Lu B. Bias ass ociated with using the estimated propensity score as a regression covariate. Stat Med 2014;33:74 87.
142 118. Austin PC. Optimal caliper widths for propensity score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat 2011;10:150 61. 119. Schneeweiss S, Avorn J. A review of uses of health care utilization databases for epidemiologic research on therapeutics. J Clin Epidemiol 2005;58:323 37. 120. Corona G, Mondaini N, Ungar A, Razzoli E, Rossi A, Fu sco F. Phosphodiesterase type 5 (PDE5) inhibitors in erectile dysfunction: the proper drug for the proper patient. J Sex Med 2011;8:3418 32. 121. Schneeweiss S, Gagne JJ, Glynn RJ, Ruhl M, Rassen JA. Assessing the comparative effectiveness of newly market ed medications: methodological challenges and implications for drug development. Clin Pharmacol Ther 2011;90:777 90. 122. Gamble JM, McAlister FA, Johnson JA, Eurich DT. Restrictive drug coverage policies can induce substantial drug exposure misclassifica tion in pharmacoepidemiologic studies. Clin Ther 2012;34:1379 86.e3. 123. Gamble JM, McAlister FA, Johnson JA, Eurich DT. Quantifying the impact of drug exposure misclassification due to restrictive drug coverage in administrative databases: a simulation cohort study. Value Health 2012;15:191 7. 124. Jacobus S, Schneeweiss S, Chan KA. Exposure misclassification as a result of free sample drug utilization in automated claims databases and its effect on a pharmacoepidemiology study of selective COX 2 inhibi tors. Pharmacoepidemiol Drug Saf 2004;13:695 702. 125. Stricker BH, Stijnen T. Analysis of individual drug use as a time varying determinant of exposure in prospective population based cohort studies. Eur J Epidemiol 2010;25:245 51. 126. Polinski JM, Kesselheim AS. Where cost, medical necessity, and morality meet: should US government insurance programs pay for erectile dysfunction drugs? Clin Pharmacol Ther 2011;89:17 9. 127. Erectile dysfunction agents prior authorization criteria. 2009. (Accessed M arch 12, 2013, at http://www.bcbsil.com/pdf/pharmacy/rx_criteria/erectile_dysfunction.pdf. ) 128. Erectile dysfunction (ED) and PDE5 inhibitor prescribing. (Accessed M arch 8, 2013 ) 129. Rothman KJ GS, and Lash TL. Modern Epidemiology. 3rd ed: Lippincott Williams & Wilkins; 2008.
143 130. Brookhart MA, Wyss R, Layton JB, StÃ¼rmer T. Propensity Score Methods for Confounding Control in Nonexperimental Research. Circ Cardiova sc Qual Outcomes 2013. 131. Curtis LH, Hammill BG, Eisenstein EL, Kramer JM, Anstrom KJ. Using inverse probability weighted estimators in comparative effectiveness analyses with observational databases. Med Care 2007;45:S103 7. 132. Lunceford JK, Davidia n M. Stratification and weighting via the propensity score in estimation of causal treatment effects: a comparative study. Stat Med 2004;23:2937 60. 133. Piccirillo JF. Steroids for idiopathic sudden sensorineural hearing loss: some questions answered, oth ers remain. JAMA 2011;305:2114 5. 134. Kaminetsky J. Epidemiology and pathophysiology of male sexual dysfunction. Int J Impot Res 2008;20 Suppl 1:S3 10. 135. Goetzel RZ, Pickens GT, Kowlessar NM. The Workforce wellness index: a method for valuing US work ers' health. J Occup Environ Med 2013;55:272 9. 136. Delate T, Simmons VA, Motheral BR. Patterns of use of sildenafil among commercially insured adults in the United States: 1998 2002. Int J Impot Res 2004;16:313 8. 137. Wysowski DK, Swann J. Use of medi cations for erectile dysfunction in the United States, 1996 through 2001. J Urol 2003;169:1040 2. 138. Titlow K, Randel L, Clancy CM, Emanuel EJ. Drug coverage decisions: the role of dollars and values. Health Aff (Millwood) 2000;19:240 7. 139. Barreto M A, Bahmad F. Phosphodiesterase type 5 inhibitors and sudden sensorineural hearing loss. Braz J Otorhinolaryngol 2013;79:727 33. 140. Thakur JS, Thakur S, Sharma DR, Mohindroo NK, Thakur A, Negi PC. Hearing loss with phosphodiesterase 5 inhibitors: a prospective and objective analysis with tadalafil. Laryngoscope 2013;123:1527 30. 141. Skeith L, Yamashita C, Mehta S, Farquhar D, Kim RB. Sildenafil and furosemide associated ototoxicity: consideration of drug drug interactions, synergy, and broader clin ical relevance. J Popul Ther Clin Pharmacol 2013;20:e128 31. 142. http://www.who umc.org . 143. Schweitzer VG. Ototoxicity of chemotherapeutic agents. Otolaryngol Clin North Am 1993;26:759 89.
144 144. de Jager P, van Altena R. Hearing loss and nephrotoxicit y in long term aminoglycoside treatment in patients with tuberculosis. Int J Tuberc Lung Dis 2002;6:622 7. 145. Wang J, Van De Water TR, Bonny C, de Ribaupierre F, Puel JL, Zine A. A peptide inhibitor of c Jun N terminal kinase protects against both aminoglycoside and acoustic trauma induced auditory hair cell death and hearing loss. J Neurosci 2003;23:8596 607. 146. Fessenden JD, Schacht J. The nitric oxide/cyclic GMP pathway: a potential major regulator of cochlear physiology. Hear Res 1998;118:168 76. 147. Eshraghi AA, Wang J, Adil E, et al. Blocking c Jun N terminal kinase signaling can prevent hearing loss induced by both electrode insertion trauma and neomycin ototoxicity. Hear Res 2007;226:168 77. 148. Kahan SE, Seftel AD, Resnick MI. Sildena fil and the Internet. J Urol 2000;163:919 23. 149. Brookhart MA, StÃ¼rmer T, Glynn RJ, Rassen J, Schneeweiss S. Confounding control in healthcare database research: challenges and potential approaches. Med Care 2010;48:S114 20. 150. Chrysant SG. Effective ness and safety of phosphodiesterase 5 inhibitors in patients with cardiovascular disease and hypertension. Curr Hypertens Rep 2013;15:475 83. 151. Shaeer O. The Global Online Sexuality Survey (GOSS): the United States of America in 2011 chapter II: phosp hodiesterase inhibitors utilization among English speakers. J Sex Med 2013;10:532 40. 152. Laties AM. Vision disorders and phosphodiesterase type 5 inhibitors: a review of the evidence to date. Drug Saf 2009;32:1 18. 153. Thurtell MJ, Tomsak RL. Nonarter itic anterior ischemic optic neuropathy with PDE 5 inhibitors for erectile dysfunction. Int J Impot Res 2008;20:537 43.
145 BIOGRAPHICAL SKETCH degree in medicine from Shanghai Medical University in 1997. H e was later trained as an infectious disease epidemiologist and worked at the department of epidemiology from the same university where he was engaged in phase III vacci ne trials in the field. He join ed the Inter national Vaccine Institute (IVI) in Seo ul, South Korean in 2001 as an a ssociate research scientist. At the IVI, he worked on translational research projects to hel p vaccine introduction in several Asian countries. He graduated from the Harvard school of Public Health with a master degree on epidemiology in 2009. He then joined conducted pha rmacoepidemiology studies us in g several large epidemiology cohort data owned by Harvard and National Cancer Institute ( NCI ) . H e received his PhD on pharmacoepidemiology from the University of Florida in 2014 . Wei Liu has authored and coauthored several peer reviewed publications and presented his research at national and international conferences. His recent research focuses on observational data analysis in large healthcare administrative databases for comparative eff ectiveness and safety of drugs .