This was a retrospective observational cohort study using patient data from a large medical and pharmacy claims database that identified male patients ≥18 years old with ≥1 diagnosis of ED between January 1, 2006, and October 31, 2020 (study period). From this population, we identified an exposed group consisting of men who filled 1 or more prescriptions for an approved PDE-5i (sildenafil, vardenafil, tadalafil, or avanafil) after their diagnosis of ED during the patient identification period (January 1, 2007, to October 31, 2020), without any PDE-5i claim in the baseline period (12 months before the index date). The first claim date of PDE-5i was defined as the index date of the exposed group. From the same population, we identified an unexposed group of patients (controls) without any claims for PDE-5i during the study period. To ensure that patients in the exposed and unexposed groups had similar amounts of time between their first ED diagnosis date and their index date, the index date of the unexposed group was the date of any pharmacy claim closest to a randomly selected date based on the distribution of days between first ED diagnosis and the index date among the exposed group.

We opted for covariate matching in combination with multivariable regression adjustment on matched-pair differences to control for any inherent biases within treatment and comparison groups. PDE-5i–exposed men were matched to controls on the basis of age; index date; diagnoses of CAD, T2DM, hypertension, and hypercholesterolemia/dyslipidemia; and antiplatelet, statin, and antihypertensive therapy using a matching ratio of up to 1:4. These matching variables were selected from a prespecified list of potential matching variables as the most clinically important variables that could be used for matching, while still maintaining the desired sample size. A single variable indicating “current smoking” was initially considered as a potential matching variable, although matching on this variable was impractical due to an insufficient number of matches with this indicator present. To address potential imbalances in smoking that remained after matching, a broader assessment algorithm was used to identify any smoking history (eg, current, past), and this modified smoking-related variable was included in subsequent multivariable-adjusted analyses.

Continued eligibility and exclusion Continuous eligibility was required for the 12-month baseline period and between the date of the initial ED diagnosis claim and the index date. Patients were excluded from analysis if they had any claim for MACE in the baseline period, if they received an index PDE-5i product specified for treatment of pulmonary hypertension, or if they had a claim for diagnosis of pulmonary hypertension and/or related treatment use during the baseline period. Patients were followed until they were censored, which was defined as the end of continuous enrollment, death, claim of outcomes (MACE), or study end date, whichever occurred earlier. Diagnoses were determined using International Classification of Diseases, Clinical Modification Ninth and Tenth Revision diagnosis codes (ICD-9-CM; ICD10-CM). Procedures were determined using ICD-9 and ICD10 procedure codes, Current Procedural Terminology codes, or Healthcare Common Procedure Coding System codes,17,18 and drug therapies were classified using Generic Product Identifier codes.

Cohorts were created for comparisons of study outcomes in patients with vs without PDE-5i exposure in all patients (main study), and in 3 subgroups: patients without a baseline diagnosis of CAD but with 1 or more baseline risk variables (age ≥45 years, smoking, T2DM, hypertension, hypercholesterolemia/dyslipidemia, peripheral arterial disease, chronic kidney disease, or statin therapy), patients with baseline T2DM, and patients with baseline diagnosis of CAD. In addition, we explored the relationship between the level of PDE-5i exposure and outcomes in the main study cohort, the PDE-5i–exposed group. Quartiles of exposure were determined by the cumulative number of tablets received based on the quantity dispensed from pharmacy claims extending from the index date until the date of censoring, with the lowest quartile as the reference.

The HealthCore Integrated Research Database contains healthcare claims and other information from over 50 million members of commercial health plans in the United States and includes health maintenance organizations, point of service plans, preferred provider organizations, Medicare Advantage and consumer-directed health plans and indemnity plans since January 1, 2006. Death and cause of death data were obtained from the National Death Index (NDI), the federal database for all recorded deaths in the United States, maintained by the Centers for Disease Control (CDC).19

The primary outcome was MACE, defined as a composite of CV death or hospitalization (≥1 inpatient claim) for myocardial infarction, coronary revascularization (percutaneous coronary intervention or coronary artery bypass graft), stroke, heart failure, or unstable angina pectoris during the follow-up period. Cardiovascular death was assessed by linkage to CDC-based NDI data, with utilization of NDI-defined underlying cause of death based on ICD-10-CM diagnosis codes. Cardiovascular death included death due to myocardial infarction, sudden cardiac death, heart failure, stroke, CV hemorrhage (including nontraumatic subarachnoid hemorrhage, aortic rupture, arterial rupture, or cardiac tamponade), or other CV causes (including peripheral arterial disease, coronary artery disease, pulmonary embolism, cardiogenic shock, ventricular tachycardia, or dilated cardiomyopathy), in alignment with consensus CV and stroke endpoint definitions.20 The secondary outcomes were overall mortality and the individual components of MACE.

Statistical methodology All demographic data, baseline clinical characteristics, and outcomes of interest were described in bivariate statistics. Frequencies and percentages were provided for categorical variables. Relevant measures of centrality such as means and medians were presented for continuous measures, as well as variance measures such as SDs and percentiles.

Hazard ratios and 95% CIs from marginal Cox proportional hazards models were used to report associations between PDE-5i use and time-to-event outcomes, including MACE and mortality, for matched and multivariable-adjusted analyses.21,22

Exact matching plus multivariable adjustment was selected given its ability to account for residual imbalances that remain after matching alone in claims-based analyses.23 Multivariable-adjusted models for MACE controlled for age on index date, insurance type, year of index drug claim, months between ED diagnosis and index date, baseline conditions (smoking, CAD, T2DM, hypertension, hypercholesterolemia/dyslipidemia, peripheral arterial disease, and hypogonadism), and baseline medications (antiplatelet therapy, short- and long-acting nitrates, high- and moderate-low– intensity statins, ezetimibe/cholesterol absorption inhibitors, eicosapentaenoic acid/docosahexaenoic acid, angiotensin receptor blockers, beta-blockers, calcium channel blockers, aldosterone receptor modulators, diuretics, thiazolidinediones, insulin, and androgen/testosterone replacement therapy). We defined “hypogonadism” as at least 1 medical claim with an ICD-9 or ICD-10 diagnosis code for hypogonadism (ICD-9: 257.2; ICD-10: E29.1), hypogonadotropic hypogonadism/hypopituitarism (ICD-9: 253.2, 253.7; ICD10: E23.0, E23.1), or postprocedural testicular hypofunction (ICD-9: 257.1; ICD-10: E89.5).

Multivariable-adjusted models for mortality controlled for age on index date, baseline Quan-Charlson Comorbidity Index, and baseline conditions (smoking, CAD, T2DM, and hypertension).24,25

The same statistical approach was used to establish the association between quartiles of PDE-5i exposure and outcomes, with the exception that the multivariableadjusted models for mortality additionally controlled for months between ED diagnosis and index date and baseline medications (antiplatelets, high- and moderate-low–intensity statins, angiotensin receptor blockers, beta-blockers, calcium channel blockers, aldosterone receptor modulators, and diuretics), since this analysis was conducted only among men exposed to PDE-5is and was thus not matched on the key variables that were used to match exposed and unexposed men in the main analysis. Kaplan-Meier curves were plotted to visualize event-free survival over the follow-up period. The study protocol with incorporated statistical analysis plan was specified a priori before review of the results. Analyses were completed using software from Instant Health Data (IHD; Panalgo) and SAS Enterprise Guide 8.3 (SAS Institute).

This study was conducted in accordance with legal and regulatory requirements, as well as with scientific purpose, value, and rigor, and followed Good Practices for Outcomes Research issued by the International Society for Pharmacoeconomics and Outcomes Research.26 A Health Insurance Portability and Accountability Act Waiver of Authorization was applied for from an Institutional Review Board (IRB) in order to access the NDI data. The protocol and applicable study materials were submitted to an IRB for review, and a waiver of authorization was granted (IRB # 20210336).

In men with known baseline CAD, there were no significant differences in adverse cardiac outcomes between men exposed vs not exposed to PDE-5, but there was a nonsignificant trend toward less cardiac mortality (HR 0.41; 95% CI 0.161.05; P = .063) in the PDE-5i exposed group. The lack of a statistically significant difference in this cohort may have been in part due to the lower number of patients in this subgroup.

Key findings from the matched and multivariable adjusted data are shown in Figure 2, demonstrating survival using Kaplan-Meier Curves and Figure 3, showing Forest plots of HRs.

Our findings confirm and extend the findings of recent studies that investigated the effects of PDE-5is in subpopulations of men with ED, including those who had T2DM,9-11 known CAD including past myocardial infarctions,12,13 and rheumatoid arthritis.27 Our findings also extend the findings from P valuea — — — <.001 ED, erectile dysfunction; EPA/DHA, eicosapentaenoic acid/docosahexaenoic acid; IQR, interquartile range; MUSE, medicated urethral system for erections; NA, not applicable; PCSK9, proprotein convertase subtilisin/kexin type 9 serine protease; PDE-5i, phosphodiesterase-5 inhibitor. a P values calculated using McNemar’s tests. b Based on Desai et al. published algorithm. c Amylin analogs, meglitinide analogues, alpha-glucosidase inhibitors, dopamine receptor agonists. 7 5 7 4 9 7 5 7 70 8 .-170 3 22 021 .-174 3 76 6 .-123 1 76 8 .-222 3 99 3 .-160 6 68 9 .-176 5 10 1 .-160 3 12 8 .-134 4 n T f .7 .7 .5 2 .5 .9 .5 9 .1 .7 .4 6 .0 .1 .6 2 .0 .6 .3 9 .0 .7 .3 3 .0 .4 .1 6 .0 .8 .5 2 3 0 0 .1 1 0 0 .6 0 0 0 .3 0 1 0 .6 1 0 0 .0 0 0 0 .5 0 0 0 .0 1 0 0 .6 ito lloo ign ito , t : d n n m , e r s a

e se = D n o n A

U p ( C , 4 e ) v i ed 5 t c s 0 e o 5 j , d it 4 n r 6 o a c u = e n S Q ( ) ir a h u =

n

T Q ( c 3

) M e r v le t) 8 fo tc trh itr se 36

i s e h 5 l j a e b u u ig = d o

h n o O F Q ( ( ) 1 9 3 - 9 x d 3 e se = n o n U p ( ) ) 2 8 6 - 8 x d 4 e se = n o n

U p ( e l b ab 2 5 ir e 0 a iv d 2 v t e it c s 9 l e o 1 u j

p = b x

n

O E ( a z a a

) h 1 6 d e 1 n iv d 8 a t e

c s 3 s e o 2 te j p = b x

n

O E ( w o l l . d o i r e p p u o f g n i r u d y t i l a t r o m d n a E C A m n o i s s e r g e r m m o r f s o i t a r d r a r e c n e d i c n i d e t s u j d a f o y r a m m u S . 2 e l b ce ce ce ce ce ce ce ce f n th i h o t iv ic d .1161 freeenR .4190 freeenR .4140 freeenR .2770 freeenR .3207 freeenR .2104 freeenR .0502 freeenR .5802 freeenR eoom ,reou fo1≥sohpdo i,tseanu i,ireapp teeann ceeanm jtrecbo lillscay frohm tc r h d ap lp o ty tt uo liu ith po ex em so re .F ec sp .320 .98 ..-41707 12 .634 .49 ..-01727 89 .704 .60 ..-91744 68 .432 .78 ..-21623 53 .942 .97 ..-61510 46 .252 .49 ..-21642 65 .130 .35 ..-11235 58 .273 .16 ..-30487 60 itsepom ,tfreaah adndw tsepod ieonnd /liisydpd i,itrseco trseeon itrsenon ligaudn /:/cdaea 1 0 0 .2 0 0 0 .6 0 1 0 .6 0 0 0 .4 0 0 0 .1 0 0 0 .7 0 0 0 .1 0 0 0 .0 m fllreavECAM /-ssrrsseeeaya001pnoC itrzaaadoRH I59C% levauP fliiIftrrccaayandono /-ssrrsseeeaya001pnoC itrzaaadoRH I59C% levauP iiltrrsreczvyaaaanooouR /-ssrrsseeeaya001pnoC itrzaaadoRH I59C% levauP fek /-ssrsrseeeaya001pno itrzaaadoR I59C% levauP flitrreeaauF /-ssrrsseeeaya001pnoC itrzaaadoRH I59C% levauP fliittsrsecegaanbnnoPA /-ssrsrseeeaya001pnoC itrzaaadoRH I59C% levauP lil-tsrrceeaaaavuoddoRM /-ssrsrseeeaya001pnoC itrzaaadoRH I59C% levauP glllittrreaavyoM /-ssrsrseeeaya001pnoC itrzaaadoRH I59C% levauP ,rcoAD rcyom ttseuh recohp inDMitfsconu illtsceay l-tcakob jftseodu ijtcevbo trecbop y e d io n d p a r c e n

t o / o d a a ry rc it ia e is h m n 2 te

n s s u i e a i o , i b tr fa l m l

e n id lc n d , a in up le s g c a o an sr ry la op reo ab ia a ,c eg 1 ek a i t e d ic s a s c n rd y s h n y re r e lo

l o a d le t C

H a nationwide cohort study in Denmark that also observed improved cardiac outcomes in men prescribed medicine for ED, although in that study the benefits persisted only for the first 3 years.28 Our findings expand the previous observations by showing that PDE-5is were associated with reduced MACE and mortality in a large general US population of men with ED, as well as men without overt CAD but with 1 or more CV risk factors, and that these associations persisted long term. In fact, Kaplan Meier curves (Figure 2) suggest wider separation in outcomes between exposed and unexposed groups over time. The results also show that within this general population of men with ED that the highest exposure levels of PDE-5is were associated with the largest magnitude of associations. A similar finding was observed in a previous study of men with known CAD treated with either PDE-5is or alprostadil.13 This dose-response observation can now be extended to a much broader population of men with ED in the US. Our findings are consistent with recent studies demonstrating safety of PDE-5is when combined with antihypertensive medicines29 and notably, no increase in CV events when patients received both PDE-5is and nitrates.30 Our findings in diabetic patients are consistent with recent reports that chronic tadalafil therapy improved cardiac function in men with diabetic cardiomyopathy.31

Taken together, these recent studies on the effects of PDE-5is in diverse populations of men with ED have provided consistent signals that PDE-5is are not only safe but may have important cardioprotective properties, findings that suggest an urgent need for an adequately powered, prospective randomized placebo-controlled trial. In addition, studies might be designed to examine the effect of PDE-5is on a general population of men with CV risk factors but not necessarily those with ED, to determine if PDE-5is have beneficial effects in that population.

Our study has the limitations of it being a retrospective observational study with the potential for residual confounding by unaccounted-for variables. While we observed a clinically significant association between exposure to PDE-5i and reductions in MACE and mortality, we cannot establish causation. It may be postulated that men who have ED and are capable of taking PDE-5is and engaging in sexual activity may be healthier to begin with, and that unmeasured baseline differences may contribute to the effects observed. A second study limitation is the absence of an independent measure of sexual activity in these men, so it is not clear whether increased sexual activity may have contributed to the benefits seen that were independent of pharmacological effects of PDE-5i exposure. However, we attempted to address this issue in our study design by matching our control population for relevant risk factors and comorbidities; in addition, we conducted a preplanned, multivariable analysis that took into account other risk factors, concomitant illnesses, and differences in CV and other drug usage. Our observation of a “dose-response” effect with varying levels of PDE-5i exposure, estimated by the number of tablets dispensed, lends further support to this hypothesis. Patients in the highest quartile of PDE-5i exposure had noticeably greater reductions in MACE and mortality, compared to patients in the lowest quartile, in addition to significant reductions in myocardial infarction and stroke rates. These reductions in CV rates were consistent and clinically meaningful.

A major limitation of our study is that PDE-5i exposure was necessarily estimated from the number of tablets dispensed via the patient health plan. In addition, for the dose response analysis, the number of PDE-5i tablets dispensed was used as a proxy measure for drug exposure.

These measures do not take into account tablets that patients may have obtained from another source (e.g. internet, friends) or tablets dispensed, but not actually consumed, in the context of sexual activity. Since we anticipated a drug class effect, and to maintain adequate sample sizes for our subgroup analyses we did not differentiate effects of individual PDE-5is or actual dosages prescribed. We plan to investigate these differences in addition to the effects of PDE-5i treatment dropout or failure, and other subgroups of interest (any history of previous MACE) in future studies. Finally, we also have no measure of the frequency of sexual activity in either the PDE-5i–exposed or –unexposed groups.

If PDE-5is have CV protective effects as this and other studies suggest, what are the potential mechanisms? While the purpose of our study was not to investigate mechanisms by which PDE-5is may have CV-protective effects, there are existing data and hypotheses in the literature on this topic. In a study by Rosano et al,7 men with ED and CV risk factors were randomized to placebo vs 4 weeks of tadalafil and were then evaluated for brachial artery flow-mediated vasodilation. Tadalafil improved flow-mediated dilation compared to placebo and also increased nitrite/nitrate levels and decreased endothelin levels. The authors concluded that tadalafil significantly improved endothelial function in treated patients. In previous preclinical studies, our laboratory (RAK) observed that sildenafil in a rabbit model of coronary artery ischemia/reperfusion reduced specific vascular resistance during reperfusion within the risk region of the heart,32 which also suggested possible improvement in endothelial function. In addition, sildenafil reduced left ventricular end diastolic pressure during ischemia, suggesting an improvement in diastolic function of the heart. In a rodent model in our laboratory, tadalafil reduced the size of experimental myocardial infarction; it was also associated with lower blood pressure in this model. The reduction in cell death could have been due in part to reduced afterload and therefore less oxygen demand of the heart.33 Others have suggested that benefits of PDE-5is may involve reduced arterial stiffness, reduced systolic velocity, and reduced inflammation.13 An additional mechanism whereby PDE-5is may have a beneficial effect on the CV system is by a potential antiplatelet effect.34

In conclusion, our study found that in a general population of men with ED, exposure to PDE-5is was associated with a 13% lower rate of MACE, 25% lower rate of mortality, and 39% lower rate of CV mortality than nonexposure. The highest exposure levels were associated with the greatest reductions in MACE and death and were also associated with significant reductions in myocardial infarction and stroke. This study adds to our knowledge base of the potential for PDE-5is to provide CV protection and provides impetus for future randomized, controlled trials.