- Email: [email protected]
Testosterone Therapy and Risk of Acute Myocardial Infarction in Hypogonadal Men: An Administrative Health Care Claims Study
ORIGINAL RESEARCH
ENDOCRINE
Testosterone Therapy and Risk of Acute Myocardial Infarction in Hypogonadal Men: An Administrative Health Care Claims Study Hu Li, MBSS, PhD,1 Lucy Mitchell, BS,2 Xiang Zhang, PhD,1 Darell Heiselman, DO,1 and Stephen Motsko, PharmD, PhD1
ABSTRACT
Background: There are some ongoing debates on the potential link between testosterone therapy (TT) and risk of acute myocardial infarction (MI). Aim: To investigate the association between acute MI and TT use compared with non-use in men having documented hypogonadism (diagnostic International Classification of Diseases, Ninth Revision codes 257.2, 257.8, 257.9, 758.7) in patient claims records. Methods: This retrospective cohort study used a real-world US-based administrative health care claims database (MarketScan 2004e2013; Truven Health Analytics, Ann Arbor, MI, USA) to compare MI rates between TT-treated men and a cohort of untreated hypogonadal men matched by a calendar time-specific propensity score. Subgroup analyses were performed by route of administration, age, and prior cardiovascular disease (CVD). Outcomes: Incidence rates of MI (per 1,000 person-years) and hazard ratio. Results: After 1:1 calendar time-specific propensity score matching, 207,176 TT-treated men and 207,176 untreated hypogonadal men were included in the analysis (mean age ¼ 51.8 years). Incidence rates of MI were 4.20 (95% CI ¼ 3.87e4.52) in the TT-treated cohort and 4.67 (95% CI ¼ 4.43e4.90) in the untreated hypogonadal cohort. Cox regression model showed no significant association between TT use and MI when comparing TTtreated with untreated hypogonadal men overall (hazard ratio ¼ 0.99, 95% CI ¼ 0.89e1.09), by age, or by prior CVD. A significant association was observed when comparing a subgroup of injectable (short- and long-acting combined) TT users with untreated hypogonadal men (hazard ratio ¼ 1.55, 95% CI ¼ 1.24e1.93). Clinical Implication: In this study, there was no association between TT (overall) and risk of acute MI. Strengths and Limitations: Strengths included the use of a comprehensive real-world database, sophisticated matching based on calendar blocks of 6 months to decrease potential bias in this observational study, carefully chosen index dates for the untreated cohort to avoid immortal time bias, and implemented sensitivity analysis to further investigate the findings (stratification by administration route, age, and prior CVD). Key limitations included no information about adherence, hypogonadism condition based solely on diagnosis (no information on clinical symptoms or testosterone levels), lack of information on disease severity, inability to capture diagnoses, medical procedures, and medicine dispensing if corresponding billing codes were not generated and findings could contain biases or fail to generalize well to other populations. Conclusion: This large, retrospective, real-world observational study showed no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men overall, by age, or by prior CVD; the suggested association between injectable TT and acute MI deserves further investigation. Li H, Mitchell L, Zhang X, et al. Testosterone Therapy and Risk of Acute Myocardial Infarction in Hypogonadal Men: An Administrative Health Care Claims Study. J Sex Med 2017;14:1307e1317. Copyright 2017, International Society for Sexual Medicine. Published by Elsevier Inc. All rights reserved.
Key Words: Myocardial Infarction; Testosterone Therapy; Hypogonadism
Received March 3, 2017. Accepted September 16, 2017. 1
Eli Lilly and Company, Indianapolis, IN, USA;
2
Eli Lilly and Company Limited, Erlwood, Windlesham, Surrey, UK Trial Registration: EU PAS, number ENCEPP/SDPP/9151.
J Sex Med 2017;14:1307e1317
The findings of this study were presented in part at the 32nd International Society of Pharmacoepidemiology; Dublin, Ireland; August 25e28, 2016. Copyright ª 2017, International Society for Sexual Medicine. Published by Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.jsxm.2017.09.010
1307
1308
INTRODUCTION The incidence of low serum testosterone in men increases with age and is associated with cardiovascular (CV) risk factors such as hyperglycemia, abdominal obesity, insulin resistance, adverse lipid profiles, and hypertension.1 Studies have shown that men with low total or free serum testosterone levels are at increased risk for coronary artery disease, more severe coronary artery disease, and potentially fatal CV events.2e5 Recent observational studies, systematic reviews and metaanalyses, and commentaries have cited conflicting findings on whether exogenous testosterone therapy (TT) use provides CV disease (CVD) protection including decreased CV-related mortality,6e10 shows neutral net findings of TT on CVD risk,11e19 or increases the risk of acute myocardial infarction (MI), mortality, or risk factors associated with CVD.20e26 For example, studies have shown a decrease in risk or no increased risk when TT users with normalized serum testosterone levels are compared with TT-treated men without normalized testosterone levels.6,7,11 Other studies have shown an increased risk of CV-related adverse events in patients treated with TT vs placebo or phosphodiesterase type 5 inhibitors (PDE5is).22,24,25 However, the article by Vigen et al24 has been widely criticized by Traish et al27 for multiple reasons including their misreporting of the absolute rate of CV events after TT and data errors. The inappropriateness of comparing MI risk between TT and PDE5i users22 is another concern because PDE5is were used mostly by patients with healthy heart conditions.28 The aim of the study was to assess the association between TT use and acute MI in a real-world clinical setting.
METHODS Study Design This retrospective cohort analysis used the 2004 to 2013 Truven Health Analytics (Ann Arbor, MI, USA) MarketScan database, which includes individual-level, de-identified, health care claims information (diagnoses, procedures, and prescriptions) from health plans and Medicare Part D and Medicaid programs. Specifically, the database includes information for employees, dependents, and retirees with commercial or Medicare insurance whose employers license health care data to Truven Health Analytics. A new-user design (ie, men newly exposed to TT) was used.29 This study was designed to provide information to address a public health concern regarding TT use in the real-world setting. Hypogonadism is the only indication for TT; a diagnosis of hypogonadism was not an inclusion criterion for the TT-treated cohort because real-world TT users at the population level are treated for hypogonadism. Because previous literature has indicated that a low baseline testosterone level is associated with an increased risk of CV outcomes,1e5 to adjust for confounding by indication, this study used men with hypogonadism who did not receive TT treatment as a comparator group, and statistical methods were applied to balance the baseline characteristics and CV risk factors of the 2 cohorts.
Li et al
Study Population The study population included men at least 18 years old who received a new prescription for TT or had a diagnosis of hypogonadism based only on claims codes (ie, International Classification of Diseases, Ninth Revision [ICD-9] codes 257.2, 257.8, 257.9, 758.7). During the baseline period (the year before the index date), TT-treated and untreated men had not received TT. All men had at least 365 days of continuous enrollment in a health plan before the index date, with continuous enrollment defined as no enrollment gap longer than 31 consecutive days. Subjects were excluded if they were women or of dual sex, received their first prescriptions of TT and a PDE5i concomitantly (±3 days), or had a diagnosis of pulmonary arterial hypertension.
Cohort Identification and Calendar Time-Specific Propensity Score Matching Calendar time-specific propensity score (CTPS) matching was used to adjust for potential changes in patterns of standard care of hypogonadism over time and to avoid immortal time bias.30 Briefly, the CTPS for each patient was defined by the predicted probability of TT initiation given the patient’s measurable baseline characteristics.31 The CTPS was constructed at discrete 6-month periods of calendar time, and TT-treated men were matched 1:1 with untreated hypogonadal men using the estimated CTPS. The TT-treated cohort included men who received at least 1 new prescription for a testosterone product after the baseline period. The untreated cohort included men who had a diagnostic code for hypogonadism but did not receive a TT prescription before the calendar block. By design, untreated patients who initiated TT later would have been included in the analysis but censored at the prescription date. TT-treated and untreated cohorts were balanced for relevant baseline characteristics, including demographic characteristics, comorbid diagnoses, prior CVD diagnoses or procedures, concomitant medications, and health care use, which were assessed by standardized differences.32,33 These covariates were selected a priori based on the plausibility of having an association with risk for acute MI and their availability in the MarketScan databases. Any imbalanced baseline characteristics were included in the statistical model to be adjusted further.
Baseline, Index Date, and Follow-Up The index date for TT-treated patients was identified as first prescription dispensing of TT. The index date for untreated patients was based on the equal probability of them receiving a TT prescription, which was the date of hypogonadism diagnosis during the first 6-month calendar block (if they were matched initially) or the randomly assigned date of any clinical or hospital visit within the subsequent 6-month calendar block (ie, if the patient was not matched initially and rolled over to the subsequent 6-month calendar block and remained untreated).34 J Sex Med 2017;14:1307e1317
1309
Testosterone Therapy and Acute MI in Hypogonadal Men
Follow-up after the index date concluded at the first occurrence of any of the following events: acute MI, death from any cause, discontinuous enrollment (gap > 31 days),35 discontinuation of TT use or gap in TT use (90 days), any change in exposure status (ie, TT-treated patients discontinuing TT [change in route of administration was not considered a change in status, except for subgroup analysis] or untreated patients beginning TT), or end of data period (December 31, 2013). The 90-day window was added as a washout period after discontinuation of TT based on findings from a previous study, which reported a 3-month recovery period after TT discontinuation to allow hemoglobin to return to pretreatment levels.36 The average follow-up time for the TT-treated cohort was approximately 8.8 months (including the 3-month washout period) and approximately 19.2 months for the untreated hypogonadal cohort.
Outcome Measures Acute MI during the follow-up period was defined by an inpatient ICD-9 diagnosis code (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 410.x0, 410.x1) or record of death within 24 hours of an emergency department visit for ischemic heart disease (ICD-9CM codes 410.x0, 410.x1, 411.1, 411.8x, and 413.x). The positive predictive value of MI in a claims database has been found to be approximately 95%.37
Statistical Analysis Baseline characteristics before and after CTPS matching are presented using descriptive statistics. Balance was assessed with standardized differences (standardized difference 0.1 indicated a good balance). Incidence rates of acute MI were calculated per 1,000 person-years (PY) with 95% CI. A Cox proportional hazard model was used to assess the association between acute MI and TT use. The assumption of proportional hazard was checked and no severe violation was observed, and baseline covariates that remained imbalanced after CTPS were adjusted in the multivariate regression model. A Kaplan-Meier plot was constructed to estimate the time to the first new acute MI. Subgroup analyses also were performed: (i) route of administration (gel or topical, transdermal patch, injectable [short- and long-acting combined], or non-specified), (ii) prior CVD, and (iii) age (18e65 vs >65 years). A separate analysis evaluated the incidence of acute MI for the TT-treated cohort compared with a CTPS-matched PDE5i-treated cohort and will be reported in a separate publication. All analyses were performed using SAS 9.2 (SAS Institute, Cary, NC, USA).
RESULTS Before applying the exclusion criteria, the study population consisted of 1,622,115 men, which included men who received TT and/or had a diagnosis of hypogonadism. After applying all exclusion criteria (Figure 1), there were 356,695 eligible J Sex Med 2017;14:1307e1317
TT-treated men and 331,785 eligible untreated hypogonadal men. After CTPS matching, each cohort was composed of 207,176 men (58.1% matched for TT-treated men and 62.4% matched for untreated men).
Demographic and Baseline Clinical Characteristics Baseline demographic and clinical characteristics, acute MI risk factors, and prior CVD (pre-matched and CTPS-matched) are listed in Tables 1 and 2. Before CTPS matching, the TT-treated cohort had more associated comorbidities and CVD risks than the untreated cohort as evidenced by increased reporting of concomitant medications (eg, significantly more TT-treated men reported taking antihypertensive, antihyperlipidemic, antidiabetic, and psychotropic medications, etc; Table 1). After CTPS matching, the TT-treated and untreated hypogonadal cohorts had a mean age of 51.8 years; underlying comorbidities and concomitant medications were well balanced (Table 1). CTPS-matched cohorts were comparable in risk factors for acute MI (Table 2). The most commonly reported prior CVD events were unspecified ischemic heart disease and other heart disease; prevalence of prior acute MI was no higher than 0.7% in the 2 cohorts. In the subgroup analysis by route of administration, the propensity score generated from the overall cohort was applied to subgroup patients.38 Although balance was achieved for most conditions, patients treated by intramuscular injection had numerically higher rates of prior CVD conditions compared with untreated and other treated groups (eTable 1).
Acute MI Incidence Rates and Cox Regression Model A total of 639 TT-treated men had an acute MI, with an incidence rate of 4.20 per 1,000 PY (95% CI ¼ 3.87e4.52), and 1,546 untreated hypogonadal men had an acute MI, with an incidence rate of 4.67 per 1,000 PY (95% CI ¼ 4.43e4.90; Table 3). The Cox regression model showed no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men overall (ie, all administration routes; adjusted hazard ratio [HR] ¼ 0.99, 95% CI ¼ 0.89e1.09, P ¼ .80; Table 4). A Kaplan-Meier plot showed no significant difference in acute MI risk between cohorts (Figure 2).
Subgroup Analysis: By Age or Prior CVD at Baseline Older men (>65 years) had higher incidence rates of acute MI than younger men: 11.12 per 1,000 PY (95% CI ¼ 9.36e12.93) vs 3.52 per 1,000 PY (95% CI ¼ 3.21e3.83) for the TT-treated cohort and 10.92 per 1,000 PY (95% CI ¼ 9.89e11.95) vs 3.81 per 1,000 PY (95% CI ¼ 3.59e4.04) for the untreated hypogonadal cohort (Table 3). Men with a history
1310
Li et al
1,622,115 total cohort paƟents before exclusion criteria applied 226 paƟents excluded due to female or dual gender
1,621,889 paƟents 41,467 paƟents excluded due to taking TRT during baseline period
1,580,422 paƟents 13,010 paƟents excluded due to first TRT/PDE5i ± 3 days
1,567,412 paƟents 8,676 paƟents excluded due to prior PAH
1,558,736 paƟents 869,835 paƟents excluded due to lack of 1-year baseline conƟnuous drug coverage and enrollment
688,901 paƟents 421 paƟents excluded due to ambiguity in paƟents’ death date before Index
688,480 total eligible paƟents Propensity Score Matching 414,352 propensity score-matched paƟents
Figure 1. Flow diagram summarizing the selection of patients eligible for inclusion in the TT-treated vs untreated hypogonadal cohort. PAH ¼ pulmonary arterial hypertension; PDE5i ¼ phosphodiesterase type 5 inhibitor; TT ¼ testosterone therapy.
of CVD had higher incidence rates of acute MI than those without prior CVD: 10.83 per 1,000 PY (95% CI ¼ 9.63e12.02) vs 2.63 per 1,000 PY (95% CI ¼ 2.34e2.92) for the TT-treated cohort and 11.59 per 1,000 PY (95% CI ¼ 10.77e12.40) vs 2.90 per 1,000 PY (95% CI ¼ 2.69e3.10) for the untreated hypogonadal cohort (Table 3). There was no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men by age or by prior CVD (Table 4).
DISCUSSION
Subgroup Analysis: By Routes of Administration
This large, retrospective cohort study (>400,000 men after CTPS matching) used a US-based administrative health care claims database (2004e2013). Results showed no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men. Subgroup analyses also showed no significant association between TT use and acute MI for older men and those with a history of CVD. There was a significant association observed between acute MI and injectable (short- and long-acting combined) TT use, but not between acute MI and other routes of administration.
The incidence rate of acute MI was higher for the subgroup of men who received injectable (short- and long-acting combined) TT (6.09 per 1,000 PY, 95% CI ¼ 5.13e7.04) vs men who received transdermal gel TT (4.06 per 1,000 PY, 95% CI ¼ 3.68e4.45) or transdermal patch TT (5.09 per 1,000 PY, 95% CI ¼ 3.69, 6.48). There also was a significant association between injectable TT use and acute MI when comparing injectable TT-treated with untreated hypogonadal men (adjusted HR ¼ 1.55, 95% CI ¼ 1.24e1.93, P < .0001; Table 4).
Whether TT use increases the risk of adverse CV events is under debate. Some observational studies have found a lower risk or neutral association in treated men who achieved normal testosterone levels vs those who did not: (i) Sharma et al7 reported a significantly lower risk of acute MI in TT-treated men who achieved normal testosterone levels (HR ¼ 0.82, 95% CI ¼ 0.71e0.95); (ii) Anderson et al6 found that TT-treated men who achieved normal testosterone levels (90% gel, 9% injection, 1% oral) had a numerically decreased 3-year risk for MI (HR ¼ 0.73, 95% CI ¼ 0.40e1.34, P ¼ .32) vs largely untreated men who J Sex Med 2017;14:1307e1317
1311
Testosterone Therapy and Acute MI in Hypogonadal Men
Table 1. Baseline characteristics in TT-treated and untreated hypogonadal cohorts before and after CTPS matching
Characteristics Age at index (y), mean (SD) Health care use Patients with hospitalizations within prior 30 d, n (%) Office visits per patient (n), mean (SD) Drug classes per patient (n), mean (SD) Total health care cost per patient (US$), mean (SD) Charlson Comorbidity Index, mean (SD) Hypogonadism-related comorbidities, n (%) Sleep disturbance Malaise or fatigue Klinefelter syndrome Pituitary disorders Testicular cancer Prostate disease Sexual dysfunction Cognitive impairment Prostate cancer Concurrent medications, n (%) Antihypertensives Antihyperlipidemics Diabetes medications Hematological agents Sleep medications Opiates Psychotropic drugs
Pre-matched population (n ¼ 688,480)
CTPS-matched population (n ¼ 414,352)
TT treated (n ¼ 356,695)
TT treated (n ¼ 207,176)
Untreated (n ¼ 207,176)
Untreated (n ¼ 331,785)
52.2 (11.4)
51.5 (12.8)
51.8 (11.4)
51.8 (12.6)
3,040 (0.9)
2,910 (0.9)
1,820 (0.9)
1,796 (0.9)
7.1* (6.3) 6.8* (4.9) 2,552 (5,725.4)
6.4* (6.2) 5.7* (4.7) 2,087 (5,221.9)
7.2 (6.0) 6.4 (4.7) 2,360 (5,515.7)
7.2 (6.5) 6.4 (4.8) 2,322 (5,186.7)
1.0 (1.7)
1.0 (1.7)
1.0 (1.7)
1.0 (1.7)
66,299 119,562 272 11,748 1,397 45,098 174,035* 530 6,219*
(18.6) (33.5) (0.1) (3.3) (0.4) (12.6) (48.8) (0.2) (1.7)
53,111 99,574 485 6,823 1,439 51,617 297,730* 583 11,736*
(16.0) (30.0) (0.2) (2.1) (0.4) (15.6) (89.7) (0.2) (3.5)
38,998 68,873 255 6,338 944 30,004 173,628 339 5,264
(18.8) (33.2) (0.1) (3.1) (0.5) (14.5) (83.8) (0.2) (2.5)
38,113 68,710 338 5,772 891 30,068 173,261 344 5,510
(18.4) (33.2) (0.2) (2.8) (0.4) (14.5) (83.6) (0.2) (2.7)
185,487* 158,180* 66,129* 29,897 45,085 156,367* 135,146*
(52.0) (44.4) (18.5) (8.4) (12.6) (43.8) (37.9)
143,677* 121,320* 48,933* 23,797 31,889 125,038* 99,960*
(43.3) (36.6) (14.8) (7.2) (9.6) (37.7) (30.1)
101,576 86,224 34,702 16,102 24,287 87,421 74,436
(49.0) (41.6) (16.8) (7.8) (11.7) (42.2) (35.9)
98,919 85,331 34,152 15,950 23,838 87,213 72,500
(47.8) (41.2) (16.5) (7.7) (11.5) (42.1) (35.0)
CTPS ¼ calendar time-specific propensity score; TT ¼ testosterone therapy. *Standardized difference greater than 0.1 represents a statistically significant imbalance between the 2 comparator cohorts.
did not achieve normal testosterone levels. Furthermore, of studies that did not assess serum testosterone levels, neutral or protective effects also were reported when compared with untreated hypogonadal men (or general population): (i) Baillargeon et al12 reported no increased risk of acute MI with intramuscular TT use (adjusted HR ¼ 0.84, 95% CI ¼ 0.69e1.02); (ii) Maggi et al17 reported that in a large diverse cohort of European men with hypogonadism, rates of new-onset CV events were not statistically different between TT-treated and untreated men; (iii) Shores et al9 reported a protective effect of TT use on mortality using a Veterans Affairs database; and (iv) Wallis et al39 reported that the risks of overall mortality and CV events were progressively lower with increased TT exposure, with a significant decrease in men with long-term exposure (median ¼ 35 months). However, other studies have reported increased risks with TT: (i) using a nested case-control study, Etminan et al21 found a significant association between first-time TT exposure and acute MI (rate ratio ¼ 1.41, 95% CI ¼ 1.06e1.87),
J Sex Med 2017;14:1307e1317
although there was no statistically significant association between acute MI and current TT use (gel, patch, injection, and others; rate ratio ¼ 1.01, 95% CI ¼ 0.89e1.16); (ii) Vigen et al24 reported that in men who underwent coronary angiography and had low serum testosterone levels, TT use was associated with an increased risk of all-cause mortality, MI, and stroke; (iii) Finkle et al22 reported a 2-fold increase in MI risk in older men shortly after initiation of TT vs older men shortly after initiation of PDE5i; and (iv) Wallis et al39 reported an increased risk of mortality and CV events after short-term exposure to TT vs a protective effect in long-term users. In addition, Xu et al25 published a systemic review and meta-analysis of placebocontrolled randomized trials and found an association between TT use and increased risk (odds ratio ¼ 1.54) of CV-related events. An updated review of 7 previously conducted reviews and meta-analyses,18 including the one by Xu et al,25 did not find any significant associations between TT use and CV-related events for 6 of the meta-analyses.13e16,19,20 It also is noteworthy
1312
Li et al
Table 2. Baseline characteristics of acute MI risk factors and prior CVD in TT-treated and untreated hypogonadal cohorts before and after CTPS matching
Characteristics Acute MI risk factors Hypertension Hyperlipidemia or lipid disorder Diabetes mellitus (mild to moderate) Smoking Osteoporosis Fracture Renal insufficiency End-stage renal disease Hypoglycemia Dementia Cancer HIV and/or AIDS Asthma or COPD Depression Peripheral neuropathy Prior CVD Prior acute MI Other ischemic heart disease CABG or PCI Other heart disease Stroke Carotid revascularization procedures Lower extremity amputation Lower extremity revascularization
Pre-matched population (n ¼ 688,480)
CTPS-matched population (n ¼ 414,352)
TT treated, n (%)
Untreated, n (%)
TT treated, n (%)
Untreated, n (%)
163,880 170,138 79,182 17,122 6,346 9,854 11,451 1,926 5,263 2,499 17,731 4,163 29,116 36,848 21,231
(45.9) (47.7) (22.2) (4.8) (1.8) (2.8) (3.2) (0.5) (1.5) (0.7) (5.0) (1.2) (8.2) (10.3) (6.0)
143,890 160,140 66,296 18,548 5,933 8,499 11,504 1,796 5,114 2,508 21,338 2,819 26,764 30,702 18,622
(43.4) (48.3) (20.0) (5.6) (1.8) (2.6) (3.5) (0.5) (1.5) (0.8) (6.4) (0.9) (8.1) (9.3) (5.6)
94,788 102,829 43,798 11,242 3,882 5,643 6,948 1,088 3,222 1,520 11,718 1,875 17,103 22,232 12,432
(45.8) (49.6) (21.1) (5.4) (1.9) (2.7) (3.4) (0.5) (1.6) (0.7) (5.7) (0.9) (8.3) (10.7) (6.0)
94,382 102,777 43,364 11,194 3,969 5,740 6,913 1,036 3,233 1,529 11,854 1,962 17,269 21,750 12,486
(45.6) (49.6) (20.9) (5.4) (1.9) (2.8) (3.3) (0.5) (1.6) (0.7) (5.7) (1.0) (8.3) (10.5) (6.0)
2,298 38,715 8,939 46,220 10,827 914 281 378
(0.6) (10.9) (2.5) (13.0) (3.0) (0.3) (0.1) (0.1)
2,137 33,826 8,214 43,382 10,156 774 232 376
(0.6) (10.2) (2.5) (13.1) (3.1) (0.2) (0.1) (0.1)
1,356 21,938 5,270 27,285 6,337 519 150 215
(0.7) (10.6) (2.5) (13.2) (3.1) (0.3) (0.1) (0.1)
1,303 21,646 5,188 27,156 6,278 508 146 219
(0.6) (10.5) (2.5) (13.1) (3.0) (0.3) (0.1) (0.1)
CABG ¼ coronary artery bypass grafting; COPD ¼ chronic obstructive pulmonary disease; CTPS ¼ calendar time-specific propensity score; CVD ¼ cardiovascular disease; MI ¼ myocardial infarction; PCI ¼ percutaneous coronary intervention; TT ¼ testosterone therapy.
that TT provided a protective effect against CV-related events, including mortality, in men with cardiometabolic disease.13 Although our study replicates the findings of some previous studies that show no increased risk between TT use and acute MI,6,7,17 the validity of our findings compared with others must be interpreted in the context of the nature of the data sources and different study designs and analytical methods. Compared with studies showing a protective effect with normalization of testosterone levels after TT initiation,6,7 our study found no association between any TT use and acute MI. Our findings did not include laboratory measurements of serum testosterone levels, which resulted from a lack of representative laboratory measurements in the MarketScan database (ie, the treated cohort in our study could include men who reached the therapeutic target and those who did not). The inability to account for baseline testosterone levels and possible population differences between patients who did and did not achieve testosterone normalization during the follow-up period might have resulted in a null finding in our analysis. Furthermore, unlike our study that censored patients at 90 days after discontinuation of a TT prescription, previous studies that showed a protective effect of TT
in men followed for longer periods (eg, 3 years) assumed longterm use of TT6,7 and imply a different etiology (ie, atherosclerosis) vs short-term exposure. Different methods chosen to adjust for acute MI risk factors (CTPS matching, disease score, and multivariate regression) and different comparator groups (hypogonadal men, non-TT general population, and PDE5i users) could explain differences among the results from various studies.6,7,12,21,22 We acknowledge that a diagnosis of hypogonadism might not indicate the type or degree of severity of hypogonadism, and our study lacks information on laboratory testosterone measurements. To our knowledge, no prior study has validated hypogonadal diagnosis using electronic claims data. We conducted an exploratory analysis of patients who had baseline testosterone levels (n ¼ 13,259) and found, on average, a larger percentage of men with lower baseline testosterone levels among TT-treated (59.1%) vs untreated (41.9%) patients. A previous study using a different database found that TT-treated men had significantly lower baseline testosterone levels than untreated men with hypogonadism (baseline testosterone level ¼ 160 vs 193 ng/dL, respectively; P < .001).9 Thus, treated men are likely to have J Sex Med 2017;14:1307e1317
0.26 0.78
0.76 0.20
489 150
315 324
J Sex Med 2017;14:1307e1317
CTPS ¼ calendar time-specific propensity score; CVD ¼ cardiovascular disease; IR ¼ incidence rate (per 1,000 person-years); MI ¼ myocardial infarction; PY ¼ person-years; TT ¼ testosterone therapy. *Non-specific ¼ testosterone product that had no specified route of administration in the records.
11.59 (10.77e12.40) 2.90 (2.69e3.10) 67,407 263,893 781 765 29,091 10.83 (9.63e12.02) 41,520 123,193 2.63 (2.34e2.92) 165,656
1.88 0.46
291,552 3.81 (3.59e4.04) 39,748 10.92 (9.89e11.95) 1,112 434 3.52 (3.21e3.83) 11.12 (9.36e12.93) 138,825 13,459
186,173 21,003
0.60 2.07
(3.99e5.06) (4.41e4.96) (4.14e5.84) (3.02e5.48) 4.53 4.69 4.99 4.25 60,521 233,502 26,448 10,829 274 1,094 132 46 0.35 0.29 0.36 0.20 156 420 51 12
(5.13e7.04) (3.68e4.45) (3.69e6.48) (1.29e4.67) 6.09 4.06 5.09 2.98 25,621 103,341 10,027 4,026
44,703 142,418 14,049 6,006
0.61 0.77 0.94 0.77
4.67 (4.43e4.90) 331,299 0.75 1,546 207,176 4.20 (3.87e4.52) 152,284 0.31 639
All matched patients 207,176 By routes of administration Injection 44,703 Transdermal gel 142,418 Transdermal patch 14,049 Non-specific* 6,006 By age group 65 y 187,902 >65 y 19,274 By prior CVD at baseline Prior CVD 41,520 No prior CVD 165,656
Patients, n Outcome, n Outcome, % PY Patients, n Outcome, n Outcome, % PY Cohort
IR (95% CI)
Untreated hypogonadal men TT-treated men
Table 3. Acute MI incidence rates in TT-treated and CTPS-matched untreated hypogonadal men
IR (95% CI)
Testosterone Therapy and Acute MI in Hypogonadal Men
1313
lower baseline testosterone levels than an untreated population. Although this study attempted to balance the baseline risk factors using CTPS matching, the lack of direct measurement of testosterone levels could have led to overestimated HRs secondary to lower endogenous testosterone levels (translated into a higher predisposed risk at baseline) in the TT cohort vs controls. Nevertheless, a meta-analysis and study by Araujo et al40,41 suggested that low endogenous testosterone levels in men are associated with increased risk of all-cause and CVD mortality. Healthy user bias is usually raised as a concern for studies that compare patients using drugs with a non-drug group. However, based on baseline comorbidity in this study (for the matched and unmatched populations), there is no indication of healthy user bias in our study. Before CTPS matching, baseline characteristics indicated that TT-treated patients were slightly older and had similar CV risk factors (eg, prevalence of prior acute MI ¼ 0.6% vs 0.6%) compared with the untreated group. After CTPS matching, the patients’ baseline characteristics were further matched on health care use, medication use, and comorbidities. It is important to note that although this study used the same commercial claims database (ie, MarketScan) as Finkle et al,22 there were contrasting findings. The difference in association between TT use and acute MI could be due to the choice of comparator groups. In the study by Finkle et al,22 a group of men using PDE5is was used as the comparator group, which can introduce selection bias because PDE5i drugs are more likely to be prescribed to patients without existing heart conditions.42 Another explanation for the discrepant findings is that different methods were used. We used a cohort design that is known for studying temporal associations, whereas Finkle et al22 used a selfcontrol study design that is known to be inappropriate for studying severe outcomes with a high mortality rate. The significant increase in acute MI risk among men using intramuscular TT vs untreated hypogonadal men in our study corroborated the findings of Layton et al,43 but not those of other studies.9,12,21 Layton et al43 used a study design similar to our study (eg, retrospective cohort study using the same US claims database, a new-user design, and as-treated analysis) and observed an increased risk of acute MI with injectable vs gel TT (combined HR ¼ 1.30, 95% CI ¼ 1.17e1.45). In contrast, Baillargeon et al12 reported that intramuscular TT use did not increase acute MI risk compared with a general population of non-users. Similarly, Shores et al9 found that TT use (88.6% intramuscular) was associated with a significant decrease in all-cause mortality (HR ¼ 0.61, 95% CI ¼ 0.42e0.88, P ¼ .008) vs propensity score-matched untreated men with documented low testosterone levels (<250 ng/dL) but unconfirmed hypogonadism. In contrast to our study and the one by Layton et al,43 the studies by Baillargeon et al12 and Shores et al9 used long follow-up periods (ie, >3 years). Conversely, Etminan et al21 reported an increased risk of acute MI with first-time gel TT use vs controls (rate ratio ¼ 1.49, 95%
1314
Li et al
Table 4. Treatment differences in risk of acute MI between TT-treated and untreated hypogonadal men in the CTPS-matched population Outcome, n
Unadjusted
Adjusted
Cohort
TT treated
Untreated
HR
95% CI
P value
HR
95% CI
P value
All matched patients* By route of administration Injection Transdermal gel Transdermal patch Non-specific† By age group 65 y >65 y‡ By prior CVD status With prior CVD§ Without prior CVD
639
1,546
0.91
0.83e1.01
.07
0.99
0.89e1.09
.80
156 420 51 12
274 1,094 132 46
1.48 0.86 1.11 0.63
1.19e1.83 0.77e0.97 0.78e1.57 0.32e1.24
.0004 .02 .58 .18
1.55 0.94 1.09 0.77
1.24e1.93 0.83e1.06 0.76e1.55 0.39e1.54
<.0001 .28 .64 .46
489 150
1,112 434
0.94 0.98
0.84e1.05 0.80e1.20
.28 .85
0.96 1.05
0.86e1.08 0.86e1.29
.49 .64
315 324
781 765
0.90 0.97
0.78e1.03 0.85e1.12
.13 .69
0.94 1.03
0.81e1.08 0.90e1.18
.37 .70
CTPS ¼ calendar time-specific propensity score; CVD ¼ cardiovascular disease; HR ¼ hazard ratio; MI ¼ myocardial infarction; TT ¼ testosterone therapy. *Cox regression model included treatment (dependent variable) and the following baseline patient characteristics that did not reach balance between the 2 cohorts after CTPS matching as independent variables (standardized differences > 0.10): age at index date, index year, risk factors (alcohol use, asthma or chronic obstructive pulmonary disease, cancer, chronic kidney disease, dementia, depression, end-stage renal disease, fracture, HIV or AIDS, hyperlipidemia, hypertension, hypoglycemia, obesity, osteoporosis, peripheral neuropathy, and tobacco use), and concomitant medications (antidiabetic, antihyperlipidemic, antihypertensive, hematologic, opiate, psychotropic, and sleep agents). † Non-specific ¼ testosterone product that had no specified route of administration in the records. The following additional covariates were added to the Cox regression model: number of office visits, number of drug classes, total health care cost, Charlson Comorbidity Index, number with mild to moderate diabetes, number with malaise or fatigue, number with sexual dysfunction, number taking antihypertensives, number taking antihyperlipidemics, and number taking antidiabetic medications. ‡ The following additional covariates were included in the Cox regression model: number of office visits and number with sexual dysfunction. § The following additional covariate was added to the Cox regression model: number of office visits.
CI ¼ 1.02e2.18), but not injectable routes vs controls (RR ¼ 1.16, 95% CI ¼ 0.68e1.99), although the sample was small (acute MI events in gel users [n ¼ 39] and injection users [n ¼ 18]). The higher risk of MI after intramuscular TT in our study might have been related not to TT but simply to slightly numerically higher rates of prior CVD conditions (eg, prior
acute MI or stroke). Because our study was conducted using a US-based database, we recognize that the CV adverse event profile might differ after the use of long-acting injections, which are frequently used in European and other countries. A limitation of our analysis is that the injectable group was not homogeneous and consisted of men taking short-and long-acting TT.
Figure 2. Kaplan-Meier plot of risk of AMI for calendar time-specific propensity score-matched testosterone therapy-treated and untreated hypogonadal cohorts with 95% Hall-Wellner bands. AMI ¼ acute myocardial infarction. J Sex Med 2017;14:1307e1317
1315
Testosterone Therapy and Acute MI in Hypogonadal Men
Moreover, early discontinuation of injectable TT is common and was not assessed in our study; such discontinuations might reflect undertreatment of patients with lower baseline testosterone levels and lead to increased CV events. Differences in pharmacokinetic disposition across dosage forms and routes of administration could have influenced the safety profile of TT.36 In fact, pharmacokinetic studies have reported immediate spikes in serum testosterone levels after the initial injection and an exponential decline resulting in low serum testosterone levels before the next injection.44 Such fluctuations in serum testosterone levels to low pretreatment levels might account for the observed acute MI risk in short-term studies of injectable TT, whereas stabilization of testosterone levels over time might explain a lack of such an association in long-term studies. Residual confounding factors (including differences in baseline endogenous testosterone levels and differential normalization) also might explain the inconsistent findings across routes of TT administration and risk of CV adverse events. Our study did not use exact matching for routeof-administration subgroups; CTPS matching of the subgroup analysis was inherited from the overall study cohort and the multivariate regression model was used to adjust for any imbalanced baseline covariates, which is considered a valid approach.38 Overall, the significant associations between injectable TT use and acute MI in our study and those of Layton et al43 need to be interpreted with caution because normalization of serum testosterone could not be assessed. Strengths of our study include use of a comprehensive realworld database, use of precise matching (ie, CTPS) to ensure that patients in the 2 cohorts had a similar baseline risk of developing acute MI, carefully chosen index dates for the untreated cohort to adjust for a time-related bias (eg, immortal time bias), and use of subgroup analyses to investigate potential differences by route of administration, age, and prior CVD. Our study also has several limitations inherent to claims data, including that only prescribed medicines are recorded in the MarketScan database, with no information about adherence or over-the-counter drugs. Presence or absence of disease might not be accurate, which could be due to missing data records in the administrative data source. For example, diagnosis of hypogonadism is based solely on ICD-9 diagnostic codes without considering clinical symptoms or laboratory measurements (eg, testosterone levels; however, these levels could have biased against the treated cohort if available [see above]) because they were infrequently captured in the database. Our study did not intend to study the association of TT duration of use and MI risk, and future studies that investigate long-term effect will likely observe a protective effect as shown by others.6,7,9 Furthermore, our study did not intend to differentiate on- from off-label use of TT; instead, the purpose was to investigate drug effect in users in a real-world setting, which provides significant public health information. Several important covariates were missing in the claims database (eg, body weight, blood pressure, and fasting
J Sex Med 2017;14:1307e1317
glucose levels). However, the study did include a number of pretreatment covariates in the CTPS model to balance and minimize measured differences between cohorts. Claims data also are hampered by the inability to measure and control for disease severity. Additional database limitations include an inability to capture diagnoses, medical procedures, and medicine dispensing if corresponding billing codes are not generated. Likewise, use of ICD-9-CM codes, Current Procedural Terminology codes, or national drug codes is subject to incompleteness or inaccuracies of coding in the database. Further, MarketScan databases are based on a large convenience sample, with the data captured mostly from large employers, Medicare, and Medicaid (medium and small firms are not represented); thus, the sample is not random and might contain biases or fail to generalize well to other populations.
CONCLUSIONS The present observational study, which used a large real-world database, new-user study design, and a number of sensitivity analyses, showed no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men overall, by age, or by prior CVD. In the subgroup analysis by route of administration, there was a significant association between injectable TT use and acute MI when comparing TT-treated with untreated hypogonadal men. The association between injectable TT and acute MI deserves further investigation.
ACKNOWLEDGMENTS Writing support was provided by Teresa Tartaglione, PharmD (Synchrogenix, a Certara Company, Wilmington, DE, USA). Corresponding Author: Hu Li, MBSS, PhD, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN 46285, USA. Tel: 317-655-9951; fax: 317-433-0268; E-mail: [email protected] Conflicts of Interest: All authors are employees and stockholders of Eli Lilly and Company. The authors have no other relevant declarations or other competing interests. Funding: Work was conducted and funded by Eli Lilly and Company, Indianapolis, IN, USA. Eli Lilly and Company played an indirect role, through the participation of the co-authors, in the design, data collection, analysis, and preparation of the article. Data analyses were performed by Eli Lilly and Company.
STATEMENT OF AUTHORSHIP Category 1 (a) Conception and Design Hu Li; Stephen Motsko (b) Acquisition of Data Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko
1316 (c) Analysis and Interpretation of Data Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko Category 2 (a) Drafting the Article Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko (b) Revising It for Intellectual Content Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko Category 3 (a) Final Approval of the Completed Article Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko
REFERENCES 1. Tsujimura A. The relationship between testosterone deficiency and men’s health. World J Mens Health 2013;31:126-135. 2. Morgentaler A, Miner MM, Caliber M, et al. Testosterone therapy and cardiovascular risk: advances and controversies. Mayo Clin Proc 2015;90:224-251. 3. Kloner RA, Carson C III, Dobs A, et al. Testosterone and cardiovascular disease. J Am Coll Cardiol 2016;67:545-557. 4. Corona G, Monami M, Boddi V, et al. Low testosterone is associated with an increased risk of MACE lethality in subjects with erectile dysfunction. J Sex Med 2010;7:1557-1564. 5. Hyde Z, Norman PE, Flicker L, et al. Low free testosterone predicts mortality from cardiovascular disease but not other causes: the Health in Men Study. J Clin Endocrinol Metab 2012;97:179-189. 6. Anderson JL, May HT, Lappé DL, et al. Impact of testosterone replacement therapy on myocardial infarction, stroke, and death in men with low testosterone concentrations in an integrated health care system. Am J Cardiol 2016;117:794-799. 7. Sharma R, Oni OA, Gupta K, et al. Normalization of testosterone level is associated with reduced incidence of myocardial infarction and mortality in men. Eur Heart J 2015;36: 2706-2715. 8. Ruige JB, Ouwens DM, Kaufman JM. Beneficial and adverse effects of testosterone on the cardiovascular system in men. J Clin Endocrinol Metab 2013;98:4300-4310. 9. Shores MM, Smith NL, Forsberg CW, et al. Testosterone treatment and mortality in men with low testosterone levels. J Clin Endocrinol Metab 2012;97:2050-2058.
Li et al 13. Corona G, Maseroli E, Rastrelli G, et al. Cardiovascular risk associated with testosterone-boosting medications: a systematic review and meta-analysis. Expert Opin Drug Saf 2014;13:1327-1351. 14. Fernández-Balsells MM, Murad MH, Lane M, et al. Adverse effects of testosterone therapy in adult men: a systematic review and meta-analysis. J Clin Endocrinol Metab 2010; 95:2560-2575. 15. Haddad RM, Kennedy CC, Caples SM, et al. Testosterone and cardiovascular risk in men: a systematic review and meta-analysis of randomized placebo-controlled trials. Mayo Clin Proc 2007;82:29-39. 16. Calof OM, Singh AB, Lee ML, et al. Adverse events associated with testosterone replacement in middle-aged and older men: a meta-analysis of randomized, placebo-controlled trials. J Gerontol A Biol Sci Med Sci 2005;60:1451-1457. 17. Maggi M, Wu FCW, Jones TH, et al; RHYME Investigators. Testosterone treatment is not associated with increased risk of adverse cardiovascular events: results from the Registry of Hypogonadism in Men (RHYME). Int J Clin Pract 2016; 70:843-852. 18. Onasanya O, Iyer G, Lucas E, et al. Association between exogenous testosterone and cardiovascular events: an overview of systematic reviews. Lancet Diabetes Endocrinol 2016;4:943-956. 19. Borst SE, Shuster JJ, Zou B, et al. Cardiovascular risks and elevation of serum DHT vary by route of testosterone administration: a systematic review and meta-analysis. BMC Med 2014;12:211. 20. Albert SG, Morley JE. Testosterone therapy, association with age, initiation and mode of therapy with cardiovascular events: a systematic review. Clin Endocrinol (Oxf) 2016;85:436-443. 21. Etminan M, Skeldon SC, Goldenberg SL, et al. Testosterone therapy and risk of myocardial infarction: a pharmacoepidemiologic study. Pharmacotherapy 2015;35:72-78. 22. Finkle WD, Greenland S, Ridgeway GK, et al. Increased risk of non-fatal myocardial infarction following testosterone therapy prescription in men. PLoS One 2014;9:e85805. 23. Schooling CM. Testosterone and cardiovascular disease. Curr Opin Endocrinol Diabetes Obes 2014;21:202-208. 24. Vigen R, O’Donnell CI, Barón AE, et al. Association of testosterone therapy with mortality, myocardial infarction, and stroke in men with low testosterone levels. JAMA 2013; 310:1829-1836. Erratum: JAMA 2014;311:967.
10. Corona G, Rastrelli G, Monami M, et al. Hypogonadism as a risk factor for cardiovascular mortality in men: a meta-analytic study. Eur J Endocrinol 2011;165:687-701.
25. Xu L, Freeman G, Cowling BJ, et al. Testosterone therapy and cardiovascular events among men: a systematic review and meta-analysis of placebo-controlled randomized trials. BMC Med 2013;11:108.
11. Snyder PJ, Bhasin S, Cunningham GR, et al; Testosterone Trials Investigators. Effects of testosterone treatment in older men. N Engl J Med 2016;374:611-624.
26. Basaria S, Coviello AD, Travison TG, et al. Adverse events associated with testosterone administration. N Engl J Med 2010;363:109-122.
12. Baillargeon J, Urban RJ, Kuo YF, et al. Risk of myocardial infarction in older men receiving testosterone therapy. Ann Pharmacother 2014;48:1138-1144.
27. Traish AM, Guay AT, Morgentaler A. Death by testosterone? We think not! J Sex Med 2014;11:624-629. 28. Viagra (sildenafil) package insert. New York: Pfizer; 2015. J Sex Med 2017;14:1307e1317
1317
Testosterone Therapy and Acute MI in Hypogonadal Men 29. Velentgas P, Dreyer NA, Nourjah P, et al, eds. Developing a protocol for observational comparative effectiveness research: a user’s guide. AHRQ publication no 12(13)-EHC099. Rockville, MD: Agency for Healthcare Research and Quality; 2013. 30. Mack CD, Glynn RJ, Brookhart MA, et al. Calendar timespecific propensity scores and comparative effectiveness research for stage III colon cancer chemotherapy. Pharmacoepidemiol Drug Saf 2013;22:810-818. 31. Schneeweiss S, Rassen JA, Glynn RJ, et al. High-dimensional propensity score adjustment in studies of treatment effects using health care claims data. Epidemiology 2009;20: 512-522.
estimating positive predictive value on the basis of review of hospital records. Am Heart J 2004;148:99-104. 38. Rassen JA, Glynn RJ, Rothman KJ, et al. Applying propensity scores estimated in a full cohort to adjust for confounding in subgroup analyses. Pharmacoepidemiol Drug Saf 2012; 21:697-709. 39. Wallis CJ, Lo K, Lee Y, et al. Survival and cardiovascular events in men treated with testosterone replacement therapy: an intention-to-treat observational cohort study. Lancet Diabetes Endocrinol 2016;4:498-506.
32. D’Agostino RB Jr, D’Agostino RB Sr. Estimating treatment effects using observational data. JAMA 2007;297:314-316.
40. Araujo AB, Dixon JM, Suarez EA, et al. Clinical review: Endogenous testosterone and mortality in men: a systematic review and meta-analysis. J Clin Endocrinol Metab 2011; 96:3007-3019.
33. Rubin DB. The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials. Stat Med 2007;26:20-36.
41. Araujo AB, Kupelian V, Page ST, et al. Sex steroids and all-cause and cause-specific mortality in men. Arch Intern Med 2007;167:1252-1260.
34. Seeger JD, Walker AM, Williams PL, et al. A propensity scorematched cohort study of the effect of statins, mainly fluvastatin, on the occurrence of acute myocardial infarction. Am J Cardiol 2003;92:1447-1451.
42. Cialis (tadalafil) package insert. Indianapolis, IN: Lilly USA, LLC; 2015.
35. Fireman B, Toh S, Butler MG, et al. A protocol for active surveillance of acute myocardial infarction in association with the use of a new antidiabetic pharmaceutical agent. Pharmacoepidemiol Drug Saf 2012;21(Suppl 1):282-290.
43. Layton JB, Meier CR, Sharpless JL, et al. Comparative safety of testosterone dosage forms. JAMA Intern Med 2015; 175:1187-1196. Erratum: JAMA Intern Med 2015;175:1248. 44. Behre HM, Wang C, Handelsman DJ, et al. Pharmacology of testosterone preparations. In: Nieschlag E, Behre HM, eds. Testosterone: action, deficiency, substitution. 3rd ed. New York: Cambridge University Press; 2004. p. 405-444.
36. Jick SS, Hagberg KW. The risk of adverse outcomes in association with use of testosterone products: a cohort study using the UK-based general practice research database. Br J Clin Pharmacol 2013;75:260-270.
SUPPLEMENTARY DATA
37. Kiyota Y, Schneeweiss S, Glynn RJ, et al. Accuracy of Medicare claims-based diagnosis of acute myocardial infarction:
Supplementary data related to this article can be found at https://doi.org/10.1016/j.jsxm.2017.09.010.
J Sex Med 2017;14:1307e1317
ENDOCRINE
Testosterone Therapy and Risk of Acute Myocardial Infarction in Hypogonadal Men: An Administrative Health Care Claims Study Hu Li, MBSS, PhD,1 Lucy Mitchell, BS,2 Xiang Zhang, PhD,1 Darell Heiselman, DO,1 and Stephen Motsko, PharmD, PhD1
ABSTRACT
Background: There are some ongoing debates on the potential link between testosterone therapy (TT) and risk of acute myocardial infarction (MI). Aim: To investigate the association between acute MI and TT use compared with non-use in men having documented hypogonadism (diagnostic International Classification of Diseases, Ninth Revision codes 257.2, 257.8, 257.9, 758.7) in patient claims records. Methods: This retrospective cohort study used a real-world US-based administrative health care claims database (MarketScan 2004e2013; Truven Health Analytics, Ann Arbor, MI, USA) to compare MI rates between TT-treated men and a cohort of untreated hypogonadal men matched by a calendar time-specific propensity score. Subgroup analyses were performed by route of administration, age, and prior cardiovascular disease (CVD). Outcomes: Incidence rates of MI (per 1,000 person-years) and hazard ratio. Results: After 1:1 calendar time-specific propensity score matching, 207,176 TT-treated men and 207,176 untreated hypogonadal men were included in the analysis (mean age ¼ 51.8 years). Incidence rates of MI were 4.20 (95% CI ¼ 3.87e4.52) in the TT-treated cohort and 4.67 (95% CI ¼ 4.43e4.90) in the untreated hypogonadal cohort. Cox regression model showed no significant association between TT use and MI when comparing TTtreated with untreated hypogonadal men overall (hazard ratio ¼ 0.99, 95% CI ¼ 0.89e1.09), by age, or by prior CVD. A significant association was observed when comparing a subgroup of injectable (short- and long-acting combined) TT users with untreated hypogonadal men (hazard ratio ¼ 1.55, 95% CI ¼ 1.24e1.93). Clinical Implication: In this study, there was no association between TT (overall) and risk of acute MI. Strengths and Limitations: Strengths included the use of a comprehensive real-world database, sophisticated matching based on calendar blocks of 6 months to decrease potential bias in this observational study, carefully chosen index dates for the untreated cohort to avoid immortal time bias, and implemented sensitivity analysis to further investigate the findings (stratification by administration route, age, and prior CVD). Key limitations included no information about adherence, hypogonadism condition based solely on diagnosis (no information on clinical symptoms or testosterone levels), lack of information on disease severity, inability to capture diagnoses, medical procedures, and medicine dispensing if corresponding billing codes were not generated and findings could contain biases or fail to generalize well to other populations. Conclusion: This large, retrospective, real-world observational study showed no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men overall, by age, or by prior CVD; the suggested association between injectable TT and acute MI deserves further investigation. Li H, Mitchell L, Zhang X, et al. Testosterone Therapy and Risk of Acute Myocardial Infarction in Hypogonadal Men: An Administrative Health Care Claims Study. J Sex Med 2017;14:1307e1317. Copyright 2017, International Society for Sexual Medicine. Published by Elsevier Inc. All rights reserved.
Key Words: Myocardial Infarction; Testosterone Therapy; Hypogonadism
Received March 3, 2017. Accepted September 16, 2017. 1
Eli Lilly and Company, Indianapolis, IN, USA;
2
Eli Lilly and Company Limited, Erlwood, Windlesham, Surrey, UK Trial Registration: EU PAS, number ENCEPP/SDPP/9151.
J Sex Med 2017;14:1307e1317
The findings of this study were presented in part at the 32nd International Society of Pharmacoepidemiology; Dublin, Ireland; August 25e28, 2016. Copyright ª 2017, International Society for Sexual Medicine. Published by Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.jsxm.2017.09.010
1307
1308
INTRODUCTION The incidence of low serum testosterone in men increases with age and is associated with cardiovascular (CV) risk factors such as hyperglycemia, abdominal obesity, insulin resistance, adverse lipid profiles, and hypertension.1 Studies have shown that men with low total or free serum testosterone levels are at increased risk for coronary artery disease, more severe coronary artery disease, and potentially fatal CV events.2e5 Recent observational studies, systematic reviews and metaanalyses, and commentaries have cited conflicting findings on whether exogenous testosterone therapy (TT) use provides CV disease (CVD) protection including decreased CV-related mortality,6e10 shows neutral net findings of TT on CVD risk,11e19 or increases the risk of acute myocardial infarction (MI), mortality, or risk factors associated with CVD.20e26 For example, studies have shown a decrease in risk or no increased risk when TT users with normalized serum testosterone levels are compared with TT-treated men without normalized testosterone levels.6,7,11 Other studies have shown an increased risk of CV-related adverse events in patients treated with TT vs placebo or phosphodiesterase type 5 inhibitors (PDE5is).22,24,25 However, the article by Vigen et al24 has been widely criticized by Traish et al27 for multiple reasons including their misreporting of the absolute rate of CV events after TT and data errors. The inappropriateness of comparing MI risk between TT and PDE5i users22 is another concern because PDE5is were used mostly by patients with healthy heart conditions.28 The aim of the study was to assess the association between TT use and acute MI in a real-world clinical setting.
METHODS Study Design This retrospective cohort analysis used the 2004 to 2013 Truven Health Analytics (Ann Arbor, MI, USA) MarketScan database, which includes individual-level, de-identified, health care claims information (diagnoses, procedures, and prescriptions) from health plans and Medicare Part D and Medicaid programs. Specifically, the database includes information for employees, dependents, and retirees with commercial or Medicare insurance whose employers license health care data to Truven Health Analytics. A new-user design (ie, men newly exposed to TT) was used.29 This study was designed to provide information to address a public health concern regarding TT use in the real-world setting. Hypogonadism is the only indication for TT; a diagnosis of hypogonadism was not an inclusion criterion for the TT-treated cohort because real-world TT users at the population level are treated for hypogonadism. Because previous literature has indicated that a low baseline testosterone level is associated with an increased risk of CV outcomes,1e5 to adjust for confounding by indication, this study used men with hypogonadism who did not receive TT treatment as a comparator group, and statistical methods were applied to balance the baseline characteristics and CV risk factors of the 2 cohorts.
Li et al
Study Population The study population included men at least 18 years old who received a new prescription for TT or had a diagnosis of hypogonadism based only on claims codes (ie, International Classification of Diseases, Ninth Revision [ICD-9] codes 257.2, 257.8, 257.9, 758.7). During the baseline period (the year before the index date), TT-treated and untreated men had not received TT. All men had at least 365 days of continuous enrollment in a health plan before the index date, with continuous enrollment defined as no enrollment gap longer than 31 consecutive days. Subjects were excluded if they were women or of dual sex, received their first prescriptions of TT and a PDE5i concomitantly (±3 days), or had a diagnosis of pulmonary arterial hypertension.
Cohort Identification and Calendar Time-Specific Propensity Score Matching Calendar time-specific propensity score (CTPS) matching was used to adjust for potential changes in patterns of standard care of hypogonadism over time and to avoid immortal time bias.30 Briefly, the CTPS for each patient was defined by the predicted probability of TT initiation given the patient’s measurable baseline characteristics.31 The CTPS was constructed at discrete 6-month periods of calendar time, and TT-treated men were matched 1:1 with untreated hypogonadal men using the estimated CTPS. The TT-treated cohort included men who received at least 1 new prescription for a testosterone product after the baseline period. The untreated cohort included men who had a diagnostic code for hypogonadism but did not receive a TT prescription before the calendar block. By design, untreated patients who initiated TT later would have been included in the analysis but censored at the prescription date. TT-treated and untreated cohorts were balanced for relevant baseline characteristics, including demographic characteristics, comorbid diagnoses, prior CVD diagnoses or procedures, concomitant medications, and health care use, which were assessed by standardized differences.32,33 These covariates were selected a priori based on the plausibility of having an association with risk for acute MI and their availability in the MarketScan databases. Any imbalanced baseline characteristics were included in the statistical model to be adjusted further.
Baseline, Index Date, and Follow-Up The index date for TT-treated patients was identified as first prescription dispensing of TT. The index date for untreated patients was based on the equal probability of them receiving a TT prescription, which was the date of hypogonadism diagnosis during the first 6-month calendar block (if they were matched initially) or the randomly assigned date of any clinical or hospital visit within the subsequent 6-month calendar block (ie, if the patient was not matched initially and rolled over to the subsequent 6-month calendar block and remained untreated).34 J Sex Med 2017;14:1307e1317
1309
Testosterone Therapy and Acute MI in Hypogonadal Men
Follow-up after the index date concluded at the first occurrence of any of the following events: acute MI, death from any cause, discontinuous enrollment (gap > 31 days),35 discontinuation of TT use or gap in TT use (90 days), any change in exposure status (ie, TT-treated patients discontinuing TT [change in route of administration was not considered a change in status, except for subgroup analysis] or untreated patients beginning TT), or end of data period (December 31, 2013). The 90-day window was added as a washout period after discontinuation of TT based on findings from a previous study, which reported a 3-month recovery period after TT discontinuation to allow hemoglobin to return to pretreatment levels.36 The average follow-up time for the TT-treated cohort was approximately 8.8 months (including the 3-month washout period) and approximately 19.2 months for the untreated hypogonadal cohort.
Outcome Measures Acute MI during the follow-up period was defined by an inpatient ICD-9 diagnosis code (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 410.x0, 410.x1) or record of death within 24 hours of an emergency department visit for ischemic heart disease (ICD-9CM codes 410.x0, 410.x1, 411.1, 411.8x, and 413.x). The positive predictive value of MI in a claims database has been found to be approximately 95%.37
Statistical Analysis Baseline characteristics before and after CTPS matching are presented using descriptive statistics. Balance was assessed with standardized differences (standardized difference 0.1 indicated a good balance). Incidence rates of acute MI were calculated per 1,000 person-years (PY) with 95% CI. A Cox proportional hazard model was used to assess the association between acute MI and TT use. The assumption of proportional hazard was checked and no severe violation was observed, and baseline covariates that remained imbalanced after CTPS were adjusted in the multivariate regression model. A Kaplan-Meier plot was constructed to estimate the time to the first new acute MI. Subgroup analyses also were performed: (i) route of administration (gel or topical, transdermal patch, injectable [short- and long-acting combined], or non-specified), (ii) prior CVD, and (iii) age (18e65 vs >65 years). A separate analysis evaluated the incidence of acute MI for the TT-treated cohort compared with a CTPS-matched PDE5i-treated cohort and will be reported in a separate publication. All analyses were performed using SAS 9.2 (SAS Institute, Cary, NC, USA).
RESULTS Before applying the exclusion criteria, the study population consisted of 1,622,115 men, which included men who received TT and/or had a diagnosis of hypogonadism. After applying all exclusion criteria (Figure 1), there were 356,695 eligible J Sex Med 2017;14:1307e1317
TT-treated men and 331,785 eligible untreated hypogonadal men. After CTPS matching, each cohort was composed of 207,176 men (58.1% matched for TT-treated men and 62.4% matched for untreated men).
Demographic and Baseline Clinical Characteristics Baseline demographic and clinical characteristics, acute MI risk factors, and prior CVD (pre-matched and CTPS-matched) are listed in Tables 1 and 2. Before CTPS matching, the TT-treated cohort had more associated comorbidities and CVD risks than the untreated cohort as evidenced by increased reporting of concomitant medications (eg, significantly more TT-treated men reported taking antihypertensive, antihyperlipidemic, antidiabetic, and psychotropic medications, etc; Table 1). After CTPS matching, the TT-treated and untreated hypogonadal cohorts had a mean age of 51.8 years; underlying comorbidities and concomitant medications were well balanced (Table 1). CTPS-matched cohorts were comparable in risk factors for acute MI (Table 2). The most commonly reported prior CVD events were unspecified ischemic heart disease and other heart disease; prevalence of prior acute MI was no higher than 0.7% in the 2 cohorts. In the subgroup analysis by route of administration, the propensity score generated from the overall cohort was applied to subgroup patients.38 Although balance was achieved for most conditions, patients treated by intramuscular injection had numerically higher rates of prior CVD conditions compared with untreated and other treated groups (eTable 1).
Acute MI Incidence Rates and Cox Regression Model A total of 639 TT-treated men had an acute MI, with an incidence rate of 4.20 per 1,000 PY (95% CI ¼ 3.87e4.52), and 1,546 untreated hypogonadal men had an acute MI, with an incidence rate of 4.67 per 1,000 PY (95% CI ¼ 4.43e4.90; Table 3). The Cox regression model showed no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men overall (ie, all administration routes; adjusted hazard ratio [HR] ¼ 0.99, 95% CI ¼ 0.89e1.09, P ¼ .80; Table 4). A Kaplan-Meier plot showed no significant difference in acute MI risk between cohorts (Figure 2).
Subgroup Analysis: By Age or Prior CVD at Baseline Older men (>65 years) had higher incidence rates of acute MI than younger men: 11.12 per 1,000 PY (95% CI ¼ 9.36e12.93) vs 3.52 per 1,000 PY (95% CI ¼ 3.21e3.83) for the TT-treated cohort and 10.92 per 1,000 PY (95% CI ¼ 9.89e11.95) vs 3.81 per 1,000 PY (95% CI ¼ 3.59e4.04) for the untreated hypogonadal cohort (Table 3). Men with a history
1310
Li et al
1,622,115 total cohort paƟents before exclusion criteria applied 226 paƟents excluded due to female or dual gender
1,621,889 paƟents 41,467 paƟents excluded due to taking TRT during baseline period
1,580,422 paƟents 13,010 paƟents excluded due to first TRT/PDE5i ± 3 days
1,567,412 paƟents 8,676 paƟents excluded due to prior PAH
1,558,736 paƟents 869,835 paƟents excluded due to lack of 1-year baseline conƟnuous drug coverage and enrollment
688,901 paƟents 421 paƟents excluded due to ambiguity in paƟents’ death date before Index
688,480 total eligible paƟents Propensity Score Matching 414,352 propensity score-matched paƟents
Figure 1. Flow diagram summarizing the selection of patients eligible for inclusion in the TT-treated vs untreated hypogonadal cohort. PAH ¼ pulmonary arterial hypertension; PDE5i ¼ phosphodiesterase type 5 inhibitor; TT ¼ testosterone therapy.
of CVD had higher incidence rates of acute MI than those without prior CVD: 10.83 per 1,000 PY (95% CI ¼ 9.63e12.02) vs 2.63 per 1,000 PY (95% CI ¼ 2.34e2.92) for the TT-treated cohort and 11.59 per 1,000 PY (95% CI ¼ 10.77e12.40) vs 2.90 per 1,000 PY (95% CI ¼ 2.69e3.10) for the untreated hypogonadal cohort (Table 3). There was no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men by age or by prior CVD (Table 4).
DISCUSSION
Subgroup Analysis: By Routes of Administration
This large, retrospective cohort study (>400,000 men after CTPS matching) used a US-based administrative health care claims database (2004e2013). Results showed no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men. Subgroup analyses also showed no significant association between TT use and acute MI for older men and those with a history of CVD. There was a significant association observed between acute MI and injectable (short- and long-acting combined) TT use, but not between acute MI and other routes of administration.
The incidence rate of acute MI was higher for the subgroup of men who received injectable (short- and long-acting combined) TT (6.09 per 1,000 PY, 95% CI ¼ 5.13e7.04) vs men who received transdermal gel TT (4.06 per 1,000 PY, 95% CI ¼ 3.68e4.45) or transdermal patch TT (5.09 per 1,000 PY, 95% CI ¼ 3.69, 6.48). There also was a significant association between injectable TT use and acute MI when comparing injectable TT-treated with untreated hypogonadal men (adjusted HR ¼ 1.55, 95% CI ¼ 1.24e1.93, P < .0001; Table 4).
Whether TT use increases the risk of adverse CV events is under debate. Some observational studies have found a lower risk or neutral association in treated men who achieved normal testosterone levels vs those who did not: (i) Sharma et al7 reported a significantly lower risk of acute MI in TT-treated men who achieved normal testosterone levels (HR ¼ 0.82, 95% CI ¼ 0.71e0.95); (ii) Anderson et al6 found that TT-treated men who achieved normal testosterone levels (90% gel, 9% injection, 1% oral) had a numerically decreased 3-year risk for MI (HR ¼ 0.73, 95% CI ¼ 0.40e1.34, P ¼ .32) vs largely untreated men who J Sex Med 2017;14:1307e1317
1311
Testosterone Therapy and Acute MI in Hypogonadal Men
Table 1. Baseline characteristics in TT-treated and untreated hypogonadal cohorts before and after CTPS matching
Characteristics Age at index (y), mean (SD) Health care use Patients with hospitalizations within prior 30 d, n (%) Office visits per patient (n), mean (SD) Drug classes per patient (n), mean (SD) Total health care cost per patient (US$), mean (SD) Charlson Comorbidity Index, mean (SD) Hypogonadism-related comorbidities, n (%) Sleep disturbance Malaise or fatigue Klinefelter syndrome Pituitary disorders Testicular cancer Prostate disease Sexual dysfunction Cognitive impairment Prostate cancer Concurrent medications, n (%) Antihypertensives Antihyperlipidemics Diabetes medications Hematological agents Sleep medications Opiates Psychotropic drugs
Pre-matched population (n ¼ 688,480)
CTPS-matched population (n ¼ 414,352)
TT treated (n ¼ 356,695)
TT treated (n ¼ 207,176)
Untreated (n ¼ 207,176)
Untreated (n ¼ 331,785)
52.2 (11.4)
51.5 (12.8)
51.8 (11.4)
51.8 (12.6)
3,040 (0.9)
2,910 (0.9)
1,820 (0.9)
1,796 (0.9)
7.1* (6.3) 6.8* (4.9) 2,552 (5,725.4)
6.4* (6.2) 5.7* (4.7) 2,087 (5,221.9)
7.2 (6.0) 6.4 (4.7) 2,360 (5,515.7)
7.2 (6.5) 6.4 (4.8) 2,322 (5,186.7)
1.0 (1.7)
1.0 (1.7)
1.0 (1.7)
1.0 (1.7)
66,299 119,562 272 11,748 1,397 45,098 174,035* 530 6,219*
(18.6) (33.5) (0.1) (3.3) (0.4) (12.6) (48.8) (0.2) (1.7)
53,111 99,574 485 6,823 1,439 51,617 297,730* 583 11,736*
(16.0) (30.0) (0.2) (2.1) (0.4) (15.6) (89.7) (0.2) (3.5)
38,998 68,873 255 6,338 944 30,004 173,628 339 5,264
(18.8) (33.2) (0.1) (3.1) (0.5) (14.5) (83.8) (0.2) (2.5)
38,113 68,710 338 5,772 891 30,068 173,261 344 5,510
(18.4) (33.2) (0.2) (2.8) (0.4) (14.5) (83.6) (0.2) (2.7)
185,487* 158,180* 66,129* 29,897 45,085 156,367* 135,146*
(52.0) (44.4) (18.5) (8.4) (12.6) (43.8) (37.9)
143,677* 121,320* 48,933* 23,797 31,889 125,038* 99,960*
(43.3) (36.6) (14.8) (7.2) (9.6) (37.7) (30.1)
101,576 86,224 34,702 16,102 24,287 87,421 74,436
(49.0) (41.6) (16.8) (7.8) (11.7) (42.2) (35.9)
98,919 85,331 34,152 15,950 23,838 87,213 72,500
(47.8) (41.2) (16.5) (7.7) (11.5) (42.1) (35.0)
CTPS ¼ calendar time-specific propensity score; TT ¼ testosterone therapy. *Standardized difference greater than 0.1 represents a statistically significant imbalance between the 2 comparator cohorts.
did not achieve normal testosterone levels. Furthermore, of studies that did not assess serum testosterone levels, neutral or protective effects also were reported when compared with untreated hypogonadal men (or general population): (i) Baillargeon et al12 reported no increased risk of acute MI with intramuscular TT use (adjusted HR ¼ 0.84, 95% CI ¼ 0.69e1.02); (ii) Maggi et al17 reported that in a large diverse cohort of European men with hypogonadism, rates of new-onset CV events were not statistically different between TT-treated and untreated men; (iii) Shores et al9 reported a protective effect of TT use on mortality using a Veterans Affairs database; and (iv) Wallis et al39 reported that the risks of overall mortality and CV events were progressively lower with increased TT exposure, with a significant decrease in men with long-term exposure (median ¼ 35 months). However, other studies have reported increased risks with TT: (i) using a nested case-control study, Etminan et al21 found a significant association between first-time TT exposure and acute MI (rate ratio ¼ 1.41, 95% CI ¼ 1.06e1.87),
J Sex Med 2017;14:1307e1317
although there was no statistically significant association between acute MI and current TT use (gel, patch, injection, and others; rate ratio ¼ 1.01, 95% CI ¼ 0.89e1.16); (ii) Vigen et al24 reported that in men who underwent coronary angiography and had low serum testosterone levels, TT use was associated with an increased risk of all-cause mortality, MI, and stroke; (iii) Finkle et al22 reported a 2-fold increase in MI risk in older men shortly after initiation of TT vs older men shortly after initiation of PDE5i; and (iv) Wallis et al39 reported an increased risk of mortality and CV events after short-term exposure to TT vs a protective effect in long-term users. In addition, Xu et al25 published a systemic review and meta-analysis of placebocontrolled randomized trials and found an association between TT use and increased risk (odds ratio ¼ 1.54) of CV-related events. An updated review of 7 previously conducted reviews and meta-analyses,18 including the one by Xu et al,25 did not find any significant associations between TT use and CV-related events for 6 of the meta-analyses.13e16,19,20 It also is noteworthy
1312
Li et al
Table 2. Baseline characteristics of acute MI risk factors and prior CVD in TT-treated and untreated hypogonadal cohorts before and after CTPS matching
Characteristics Acute MI risk factors Hypertension Hyperlipidemia or lipid disorder Diabetes mellitus (mild to moderate) Smoking Osteoporosis Fracture Renal insufficiency End-stage renal disease Hypoglycemia Dementia Cancer HIV and/or AIDS Asthma or COPD Depression Peripheral neuropathy Prior CVD Prior acute MI Other ischemic heart disease CABG or PCI Other heart disease Stroke Carotid revascularization procedures Lower extremity amputation Lower extremity revascularization
Pre-matched population (n ¼ 688,480)
CTPS-matched population (n ¼ 414,352)
TT treated, n (%)
Untreated, n (%)
TT treated, n (%)
Untreated, n (%)
163,880 170,138 79,182 17,122 6,346 9,854 11,451 1,926 5,263 2,499 17,731 4,163 29,116 36,848 21,231
(45.9) (47.7) (22.2) (4.8) (1.8) (2.8) (3.2) (0.5) (1.5) (0.7) (5.0) (1.2) (8.2) (10.3) (6.0)
143,890 160,140 66,296 18,548 5,933 8,499 11,504 1,796 5,114 2,508 21,338 2,819 26,764 30,702 18,622
(43.4) (48.3) (20.0) (5.6) (1.8) (2.6) (3.5) (0.5) (1.5) (0.8) (6.4) (0.9) (8.1) (9.3) (5.6)
94,788 102,829 43,798 11,242 3,882 5,643 6,948 1,088 3,222 1,520 11,718 1,875 17,103 22,232 12,432
(45.8) (49.6) (21.1) (5.4) (1.9) (2.7) (3.4) (0.5) (1.6) (0.7) (5.7) (0.9) (8.3) (10.7) (6.0)
94,382 102,777 43,364 11,194 3,969 5,740 6,913 1,036 3,233 1,529 11,854 1,962 17,269 21,750 12,486
(45.6) (49.6) (20.9) (5.4) (1.9) (2.8) (3.3) (0.5) (1.6) (0.7) (5.7) (1.0) (8.3) (10.5) (6.0)
2,298 38,715 8,939 46,220 10,827 914 281 378
(0.6) (10.9) (2.5) (13.0) (3.0) (0.3) (0.1) (0.1)
2,137 33,826 8,214 43,382 10,156 774 232 376
(0.6) (10.2) (2.5) (13.1) (3.1) (0.2) (0.1) (0.1)
1,356 21,938 5,270 27,285 6,337 519 150 215
(0.7) (10.6) (2.5) (13.2) (3.1) (0.3) (0.1) (0.1)
1,303 21,646 5,188 27,156 6,278 508 146 219
(0.6) (10.5) (2.5) (13.1) (3.0) (0.3) (0.1) (0.1)
CABG ¼ coronary artery bypass grafting; COPD ¼ chronic obstructive pulmonary disease; CTPS ¼ calendar time-specific propensity score; CVD ¼ cardiovascular disease; MI ¼ myocardial infarction; PCI ¼ percutaneous coronary intervention; TT ¼ testosterone therapy.
that TT provided a protective effect against CV-related events, including mortality, in men with cardiometabolic disease.13 Although our study replicates the findings of some previous studies that show no increased risk between TT use and acute MI,6,7,17 the validity of our findings compared with others must be interpreted in the context of the nature of the data sources and different study designs and analytical methods. Compared with studies showing a protective effect with normalization of testosterone levels after TT initiation,6,7 our study found no association between any TT use and acute MI. Our findings did not include laboratory measurements of serum testosterone levels, which resulted from a lack of representative laboratory measurements in the MarketScan database (ie, the treated cohort in our study could include men who reached the therapeutic target and those who did not). The inability to account for baseline testosterone levels and possible population differences between patients who did and did not achieve testosterone normalization during the follow-up period might have resulted in a null finding in our analysis. Furthermore, unlike our study that censored patients at 90 days after discontinuation of a TT prescription, previous studies that showed a protective effect of TT
in men followed for longer periods (eg, 3 years) assumed longterm use of TT6,7 and imply a different etiology (ie, atherosclerosis) vs short-term exposure. Different methods chosen to adjust for acute MI risk factors (CTPS matching, disease score, and multivariate regression) and different comparator groups (hypogonadal men, non-TT general population, and PDE5i users) could explain differences among the results from various studies.6,7,12,21,22 We acknowledge that a diagnosis of hypogonadism might not indicate the type or degree of severity of hypogonadism, and our study lacks information on laboratory testosterone measurements. To our knowledge, no prior study has validated hypogonadal diagnosis using electronic claims data. We conducted an exploratory analysis of patients who had baseline testosterone levels (n ¼ 13,259) and found, on average, a larger percentage of men with lower baseline testosterone levels among TT-treated (59.1%) vs untreated (41.9%) patients. A previous study using a different database found that TT-treated men had significantly lower baseline testosterone levels than untreated men with hypogonadism (baseline testosterone level ¼ 160 vs 193 ng/dL, respectively; P < .001).9 Thus, treated men are likely to have J Sex Med 2017;14:1307e1317
0.26 0.78
0.76 0.20
489 150
315 324
J Sex Med 2017;14:1307e1317
CTPS ¼ calendar time-specific propensity score; CVD ¼ cardiovascular disease; IR ¼ incidence rate (per 1,000 person-years); MI ¼ myocardial infarction; PY ¼ person-years; TT ¼ testosterone therapy. *Non-specific ¼ testosterone product that had no specified route of administration in the records.
11.59 (10.77e12.40) 2.90 (2.69e3.10) 67,407 263,893 781 765 29,091 10.83 (9.63e12.02) 41,520 123,193 2.63 (2.34e2.92) 165,656
1.88 0.46
291,552 3.81 (3.59e4.04) 39,748 10.92 (9.89e11.95) 1,112 434 3.52 (3.21e3.83) 11.12 (9.36e12.93) 138,825 13,459
186,173 21,003
0.60 2.07
(3.99e5.06) (4.41e4.96) (4.14e5.84) (3.02e5.48) 4.53 4.69 4.99 4.25 60,521 233,502 26,448 10,829 274 1,094 132 46 0.35 0.29 0.36 0.20 156 420 51 12
(5.13e7.04) (3.68e4.45) (3.69e6.48) (1.29e4.67) 6.09 4.06 5.09 2.98 25,621 103,341 10,027 4,026
44,703 142,418 14,049 6,006
0.61 0.77 0.94 0.77
4.67 (4.43e4.90) 331,299 0.75 1,546 207,176 4.20 (3.87e4.52) 152,284 0.31 639
All matched patients 207,176 By routes of administration Injection 44,703 Transdermal gel 142,418 Transdermal patch 14,049 Non-specific* 6,006 By age group 65 y 187,902 >65 y 19,274 By prior CVD at baseline Prior CVD 41,520 No prior CVD 165,656
Patients, n Outcome, n Outcome, % PY Patients, n Outcome, n Outcome, % PY Cohort
IR (95% CI)
Untreated hypogonadal men TT-treated men
Table 3. Acute MI incidence rates in TT-treated and CTPS-matched untreated hypogonadal men
IR (95% CI)
Testosterone Therapy and Acute MI in Hypogonadal Men
1313
lower baseline testosterone levels than an untreated population. Although this study attempted to balance the baseline risk factors using CTPS matching, the lack of direct measurement of testosterone levels could have led to overestimated HRs secondary to lower endogenous testosterone levels (translated into a higher predisposed risk at baseline) in the TT cohort vs controls. Nevertheless, a meta-analysis and study by Araujo et al40,41 suggested that low endogenous testosterone levels in men are associated with increased risk of all-cause and CVD mortality. Healthy user bias is usually raised as a concern for studies that compare patients using drugs with a non-drug group. However, based on baseline comorbidity in this study (for the matched and unmatched populations), there is no indication of healthy user bias in our study. Before CTPS matching, baseline characteristics indicated that TT-treated patients were slightly older and had similar CV risk factors (eg, prevalence of prior acute MI ¼ 0.6% vs 0.6%) compared with the untreated group. After CTPS matching, the patients’ baseline characteristics were further matched on health care use, medication use, and comorbidities. It is important to note that although this study used the same commercial claims database (ie, MarketScan) as Finkle et al,22 there were contrasting findings. The difference in association between TT use and acute MI could be due to the choice of comparator groups. In the study by Finkle et al,22 a group of men using PDE5is was used as the comparator group, which can introduce selection bias because PDE5i drugs are more likely to be prescribed to patients without existing heart conditions.42 Another explanation for the discrepant findings is that different methods were used. We used a cohort design that is known for studying temporal associations, whereas Finkle et al22 used a selfcontrol study design that is known to be inappropriate for studying severe outcomes with a high mortality rate. The significant increase in acute MI risk among men using intramuscular TT vs untreated hypogonadal men in our study corroborated the findings of Layton et al,43 but not those of other studies.9,12,21 Layton et al43 used a study design similar to our study (eg, retrospective cohort study using the same US claims database, a new-user design, and as-treated analysis) and observed an increased risk of acute MI with injectable vs gel TT (combined HR ¼ 1.30, 95% CI ¼ 1.17e1.45). In contrast, Baillargeon et al12 reported that intramuscular TT use did not increase acute MI risk compared with a general population of non-users. Similarly, Shores et al9 found that TT use (88.6% intramuscular) was associated with a significant decrease in all-cause mortality (HR ¼ 0.61, 95% CI ¼ 0.42e0.88, P ¼ .008) vs propensity score-matched untreated men with documented low testosterone levels (<250 ng/dL) but unconfirmed hypogonadism. In contrast to our study and the one by Layton et al,43 the studies by Baillargeon et al12 and Shores et al9 used long follow-up periods (ie, >3 years). Conversely, Etminan et al21 reported an increased risk of acute MI with first-time gel TT use vs controls (rate ratio ¼ 1.49, 95%
1314
Li et al
Table 4. Treatment differences in risk of acute MI between TT-treated and untreated hypogonadal men in the CTPS-matched population Outcome, n
Unadjusted
Adjusted
Cohort
TT treated
Untreated
HR
95% CI
P value
HR
95% CI
P value
All matched patients* By route of administration Injection Transdermal gel Transdermal patch Non-specific† By age group 65 y >65 y‡ By prior CVD status With prior CVD§ Without prior CVD
639
1,546
0.91
0.83e1.01
.07
0.99
0.89e1.09
.80
156 420 51 12
274 1,094 132 46
1.48 0.86 1.11 0.63
1.19e1.83 0.77e0.97 0.78e1.57 0.32e1.24
.0004 .02 .58 .18
1.55 0.94 1.09 0.77
1.24e1.93 0.83e1.06 0.76e1.55 0.39e1.54
<.0001 .28 .64 .46
489 150
1,112 434
0.94 0.98
0.84e1.05 0.80e1.20
.28 .85
0.96 1.05
0.86e1.08 0.86e1.29
.49 .64
315 324
781 765
0.90 0.97
0.78e1.03 0.85e1.12
.13 .69
0.94 1.03
0.81e1.08 0.90e1.18
.37 .70
CTPS ¼ calendar time-specific propensity score; CVD ¼ cardiovascular disease; HR ¼ hazard ratio; MI ¼ myocardial infarction; TT ¼ testosterone therapy. *Cox regression model included treatment (dependent variable) and the following baseline patient characteristics that did not reach balance between the 2 cohorts after CTPS matching as independent variables (standardized differences > 0.10): age at index date, index year, risk factors (alcohol use, asthma or chronic obstructive pulmonary disease, cancer, chronic kidney disease, dementia, depression, end-stage renal disease, fracture, HIV or AIDS, hyperlipidemia, hypertension, hypoglycemia, obesity, osteoporosis, peripheral neuropathy, and tobacco use), and concomitant medications (antidiabetic, antihyperlipidemic, antihypertensive, hematologic, opiate, psychotropic, and sleep agents). † Non-specific ¼ testosterone product that had no specified route of administration in the records. The following additional covariates were added to the Cox regression model: number of office visits, number of drug classes, total health care cost, Charlson Comorbidity Index, number with mild to moderate diabetes, number with malaise or fatigue, number with sexual dysfunction, number taking antihypertensives, number taking antihyperlipidemics, and number taking antidiabetic medications. ‡ The following additional covariates were included in the Cox regression model: number of office visits and number with sexual dysfunction. § The following additional covariate was added to the Cox regression model: number of office visits.
CI ¼ 1.02e2.18), but not injectable routes vs controls (RR ¼ 1.16, 95% CI ¼ 0.68e1.99), although the sample was small (acute MI events in gel users [n ¼ 39] and injection users [n ¼ 18]). The higher risk of MI after intramuscular TT in our study might have been related not to TT but simply to slightly numerically higher rates of prior CVD conditions (eg, prior
acute MI or stroke). Because our study was conducted using a US-based database, we recognize that the CV adverse event profile might differ after the use of long-acting injections, which are frequently used in European and other countries. A limitation of our analysis is that the injectable group was not homogeneous and consisted of men taking short-and long-acting TT.
Figure 2. Kaplan-Meier plot of risk of AMI for calendar time-specific propensity score-matched testosterone therapy-treated and untreated hypogonadal cohorts with 95% Hall-Wellner bands. AMI ¼ acute myocardial infarction. J Sex Med 2017;14:1307e1317
1315
Testosterone Therapy and Acute MI in Hypogonadal Men
Moreover, early discontinuation of injectable TT is common and was not assessed in our study; such discontinuations might reflect undertreatment of patients with lower baseline testosterone levels and lead to increased CV events. Differences in pharmacokinetic disposition across dosage forms and routes of administration could have influenced the safety profile of TT.36 In fact, pharmacokinetic studies have reported immediate spikes in serum testosterone levels after the initial injection and an exponential decline resulting in low serum testosterone levels before the next injection.44 Such fluctuations in serum testosterone levels to low pretreatment levels might account for the observed acute MI risk in short-term studies of injectable TT, whereas stabilization of testosterone levels over time might explain a lack of such an association in long-term studies. Residual confounding factors (including differences in baseline endogenous testosterone levels and differential normalization) also might explain the inconsistent findings across routes of TT administration and risk of CV adverse events. Our study did not use exact matching for routeof-administration subgroups; CTPS matching of the subgroup analysis was inherited from the overall study cohort and the multivariate regression model was used to adjust for any imbalanced baseline covariates, which is considered a valid approach.38 Overall, the significant associations between injectable TT use and acute MI in our study and those of Layton et al43 need to be interpreted with caution because normalization of serum testosterone could not be assessed. Strengths of our study include use of a comprehensive realworld database, use of precise matching (ie, CTPS) to ensure that patients in the 2 cohorts had a similar baseline risk of developing acute MI, carefully chosen index dates for the untreated cohort to adjust for a time-related bias (eg, immortal time bias), and use of subgroup analyses to investigate potential differences by route of administration, age, and prior CVD. Our study also has several limitations inherent to claims data, including that only prescribed medicines are recorded in the MarketScan database, with no information about adherence or over-the-counter drugs. Presence or absence of disease might not be accurate, which could be due to missing data records in the administrative data source. For example, diagnosis of hypogonadism is based solely on ICD-9 diagnostic codes without considering clinical symptoms or laboratory measurements (eg, testosterone levels; however, these levels could have biased against the treated cohort if available [see above]) because they were infrequently captured in the database. Our study did not intend to study the association of TT duration of use and MI risk, and future studies that investigate long-term effect will likely observe a protective effect as shown by others.6,7,9 Furthermore, our study did not intend to differentiate on- from off-label use of TT; instead, the purpose was to investigate drug effect in users in a real-world setting, which provides significant public health information. Several important covariates were missing in the claims database (eg, body weight, blood pressure, and fasting
J Sex Med 2017;14:1307e1317
glucose levels). However, the study did include a number of pretreatment covariates in the CTPS model to balance and minimize measured differences between cohorts. Claims data also are hampered by the inability to measure and control for disease severity. Additional database limitations include an inability to capture diagnoses, medical procedures, and medicine dispensing if corresponding billing codes are not generated. Likewise, use of ICD-9-CM codes, Current Procedural Terminology codes, or national drug codes is subject to incompleteness or inaccuracies of coding in the database. Further, MarketScan databases are based on a large convenience sample, with the data captured mostly from large employers, Medicare, and Medicaid (medium and small firms are not represented); thus, the sample is not random and might contain biases or fail to generalize well to other populations.
CONCLUSIONS The present observational study, which used a large real-world database, new-user study design, and a number of sensitivity analyses, showed no significant association between TT use and acute MI when comparing TT-treated with untreated hypogonadal men overall, by age, or by prior CVD. In the subgroup analysis by route of administration, there was a significant association between injectable TT use and acute MI when comparing TT-treated with untreated hypogonadal men. The association between injectable TT and acute MI deserves further investigation.
ACKNOWLEDGMENTS Writing support was provided by Teresa Tartaglione, PharmD (Synchrogenix, a Certara Company, Wilmington, DE, USA). Corresponding Author: Hu Li, MBSS, PhD, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN 46285, USA. Tel: 317-655-9951; fax: 317-433-0268; E-mail: [email protected] Conflicts of Interest: All authors are employees and stockholders of Eli Lilly and Company. The authors have no other relevant declarations or other competing interests. Funding: Work was conducted and funded by Eli Lilly and Company, Indianapolis, IN, USA. Eli Lilly and Company played an indirect role, through the participation of the co-authors, in the design, data collection, analysis, and preparation of the article. Data analyses were performed by Eli Lilly and Company.
STATEMENT OF AUTHORSHIP Category 1 (a) Conception and Design Hu Li; Stephen Motsko (b) Acquisition of Data Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko
1316 (c) Analysis and Interpretation of Data Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko Category 2 (a) Drafting the Article Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko (b) Revising It for Intellectual Content Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko Category 3 (a) Final Approval of the Completed Article Hu Li; Lucy Mitchell; Xiang Zhang; Darell Heiselman; Stephen Motsko
REFERENCES 1. Tsujimura A. The relationship between testosterone deficiency and men’s health. World J Mens Health 2013;31:126-135. 2. Morgentaler A, Miner MM, Caliber M, et al. Testosterone therapy and cardiovascular risk: advances and controversies. Mayo Clin Proc 2015;90:224-251. 3. Kloner RA, Carson C III, Dobs A, et al. Testosterone and cardiovascular disease. J Am Coll Cardiol 2016;67:545-557. 4. Corona G, Monami M, Boddi V, et al. Low testosterone is associated with an increased risk of MACE lethality in subjects with erectile dysfunction. J Sex Med 2010;7:1557-1564. 5. Hyde Z, Norman PE, Flicker L, et al. Low free testosterone predicts mortality from cardiovascular disease but not other causes: the Health in Men Study. J Clin Endocrinol Metab 2012;97:179-189. 6. Anderson JL, May HT, Lappé DL, et al. Impact of testosterone replacement therapy on myocardial infarction, stroke, and death in men with low testosterone concentrations in an integrated health care system. Am J Cardiol 2016;117:794-799. 7. Sharma R, Oni OA, Gupta K, et al. Normalization of testosterone level is associated with reduced incidence of myocardial infarction and mortality in men. Eur Heart J 2015;36: 2706-2715. 8. Ruige JB, Ouwens DM, Kaufman JM. Beneficial and adverse effects of testosterone on the cardiovascular system in men. J Clin Endocrinol Metab 2013;98:4300-4310. 9. Shores MM, Smith NL, Forsberg CW, et al. Testosterone treatment and mortality in men with low testosterone levels. J Clin Endocrinol Metab 2012;97:2050-2058.
Li et al 13. Corona G, Maseroli E, Rastrelli G, et al. Cardiovascular risk associated with testosterone-boosting medications: a systematic review and meta-analysis. Expert Opin Drug Saf 2014;13:1327-1351. 14. Fernández-Balsells MM, Murad MH, Lane M, et al. Adverse effects of testosterone therapy in adult men: a systematic review and meta-analysis. J Clin Endocrinol Metab 2010; 95:2560-2575. 15. Haddad RM, Kennedy CC, Caples SM, et al. Testosterone and cardiovascular risk in men: a systematic review and meta-analysis of randomized placebo-controlled trials. Mayo Clin Proc 2007;82:29-39. 16. Calof OM, Singh AB, Lee ML, et al. Adverse events associated with testosterone replacement in middle-aged and older men: a meta-analysis of randomized, placebo-controlled trials. J Gerontol A Biol Sci Med Sci 2005;60:1451-1457. 17. Maggi M, Wu FCW, Jones TH, et al; RHYME Investigators. Testosterone treatment is not associated with increased risk of adverse cardiovascular events: results from the Registry of Hypogonadism in Men (RHYME). Int J Clin Pract 2016; 70:843-852. 18. Onasanya O, Iyer G, Lucas E, et al. Association between exogenous testosterone and cardiovascular events: an overview of systematic reviews. Lancet Diabetes Endocrinol 2016;4:943-956. 19. Borst SE, Shuster JJ, Zou B, et al. Cardiovascular risks and elevation of serum DHT vary by route of testosterone administration: a systematic review and meta-analysis. BMC Med 2014;12:211. 20. Albert SG, Morley JE. Testosterone therapy, association with age, initiation and mode of therapy with cardiovascular events: a systematic review. Clin Endocrinol (Oxf) 2016;85:436-443. 21. Etminan M, Skeldon SC, Goldenberg SL, et al. Testosterone therapy and risk of myocardial infarction: a pharmacoepidemiologic study. Pharmacotherapy 2015;35:72-78. 22. Finkle WD, Greenland S, Ridgeway GK, et al. Increased risk of non-fatal myocardial infarction following testosterone therapy prescription in men. PLoS One 2014;9:e85805. 23. Schooling CM. Testosterone and cardiovascular disease. Curr Opin Endocrinol Diabetes Obes 2014;21:202-208. 24. Vigen R, O’Donnell CI, Barón AE, et al. Association of testosterone therapy with mortality, myocardial infarction, and stroke in men with low testosterone levels. JAMA 2013; 310:1829-1836. Erratum: JAMA 2014;311:967.
10. Corona G, Rastrelli G, Monami M, et al. Hypogonadism as a risk factor for cardiovascular mortality in men: a meta-analytic study. Eur J Endocrinol 2011;165:687-701.
25. Xu L, Freeman G, Cowling BJ, et al. Testosterone therapy and cardiovascular events among men: a systematic review and meta-analysis of placebo-controlled randomized trials. BMC Med 2013;11:108.
11. Snyder PJ, Bhasin S, Cunningham GR, et al; Testosterone Trials Investigators. Effects of testosterone treatment in older men. N Engl J Med 2016;374:611-624.
26. Basaria S, Coviello AD, Travison TG, et al. Adverse events associated with testosterone administration. N Engl J Med 2010;363:109-122.
12. Baillargeon J, Urban RJ, Kuo YF, et al. Risk of myocardial infarction in older men receiving testosterone therapy. Ann Pharmacother 2014;48:1138-1144.
27. Traish AM, Guay AT, Morgentaler A. Death by testosterone? We think not! J Sex Med 2014;11:624-629. 28. Viagra (sildenafil) package insert. New York: Pfizer; 2015. J Sex Med 2017;14:1307e1317
1317
Testosterone Therapy and Acute MI in Hypogonadal Men 29. Velentgas P, Dreyer NA, Nourjah P, et al, eds. Developing a protocol for observational comparative effectiveness research: a user’s guide. AHRQ publication no 12(13)-EHC099. Rockville, MD: Agency for Healthcare Research and Quality; 2013. 30. Mack CD, Glynn RJ, Brookhart MA, et al. Calendar timespecific propensity scores and comparative effectiveness research for stage III colon cancer chemotherapy. Pharmacoepidemiol Drug Saf 2013;22:810-818. 31. Schneeweiss S, Rassen JA, Glynn RJ, et al. High-dimensional propensity score adjustment in studies of treatment effects using health care claims data. Epidemiology 2009;20: 512-522.
estimating positive predictive value on the basis of review of hospital records. Am Heart J 2004;148:99-104. 38. Rassen JA, Glynn RJ, Rothman KJ, et al. Applying propensity scores estimated in a full cohort to adjust for confounding in subgroup analyses. Pharmacoepidemiol Drug Saf 2012; 21:697-709. 39. Wallis CJ, Lo K, Lee Y, et al. Survival and cardiovascular events in men treated with testosterone replacement therapy: an intention-to-treat observational cohort study. Lancet Diabetes Endocrinol 2016;4:498-506.
32. D’Agostino RB Jr, D’Agostino RB Sr. Estimating treatment effects using observational data. JAMA 2007;297:314-316.
40. Araujo AB, Dixon JM, Suarez EA, et al. Clinical review: Endogenous testosterone and mortality in men: a systematic review and meta-analysis. J Clin Endocrinol Metab 2011; 96:3007-3019.
33. Rubin DB. The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials. Stat Med 2007;26:20-36.
41. Araujo AB, Kupelian V, Page ST, et al. Sex steroids and all-cause and cause-specific mortality in men. Arch Intern Med 2007;167:1252-1260.
34. Seeger JD, Walker AM, Williams PL, et al. A propensity scorematched cohort study of the effect of statins, mainly fluvastatin, on the occurrence of acute myocardial infarction. Am J Cardiol 2003;92:1447-1451.
42. Cialis (tadalafil) package insert. Indianapolis, IN: Lilly USA, LLC; 2015.
35. Fireman B, Toh S, Butler MG, et al. A protocol for active surveillance of acute myocardial infarction in association with the use of a new antidiabetic pharmaceutical agent. Pharmacoepidemiol Drug Saf 2012;21(Suppl 1):282-290.
43. Layton JB, Meier CR, Sharpless JL, et al. Comparative safety of testosterone dosage forms. JAMA Intern Med 2015; 175:1187-1196. Erratum: JAMA Intern Med 2015;175:1248. 44. Behre HM, Wang C, Handelsman DJ, et al. Pharmacology of testosterone preparations. In: Nieschlag E, Behre HM, eds. Testosterone: action, deficiency, substitution. 3rd ed. New York: Cambridge University Press; 2004. p. 405-444.
36. Jick SS, Hagberg KW. The risk of adverse outcomes in association with use of testosterone products: a cohort study using the UK-based general practice research database. Br J Clin Pharmacol 2013;75:260-270.
SUPPLEMENTARY DATA
37. Kiyota Y, Schneeweiss S, Glynn RJ, et al. Accuracy of Medicare claims-based diagnosis of acute myocardial infarction:
Supplementary data related to this article can be found at https://doi.org/10.1016/j.jsxm.2017.09.010.
J Sex Med 2017;14:1307e1317