PDE5 inhibitors: a primary-source fact-check

Status: draft compiled 2026-04-28. Scope: sildenafil, tadalafil, vardenafil, avanafil. Each numbered claim from the libido-research draft is tested against FDA labels, registrational pharmacokinetic studies, and the recent observational cohort and Mendelian-randomization work.

The PDE5-inhibitor class shares a single, narrow mechanism. These molecules competitively inhibit phosphodiesterase type 5, the enzyme that hydrolyses cyclic GMP in vascular and corpus-cavernosum smooth muscle. With PDE5 blocked, the cGMP that nitric-oxide-driven sexual stimulation generates persists; cGMP-dependent kinase keeps myosin light-chain phosphorylation low; smooth-muscle relaxation is sustained; the helicine arteries dilate; venous outflow is mechanically compressed; and an erection forms and persists. None of that pathway is dopaminergic, melanocortinergic, or noradrenergic in the brain. Bremelanotide, flibanserin, apomorphine and bupropion are the libido-relevant CNS agents; the PDE5 class is, in labeling terms, peripheral-vascular. The drugs cannot cause an erection without upstream sexual stimulation, because they do nothing about the nitric-oxide signal that has to start cGMP production. That detail is what the draft is gesturing at when it says sildenafil "just makes erections easier." Sources: PDE5 inhibitors StatPearls overview, Sildenafil StatPearls, Tadalafil StatPearls. Confidence: C1.

The pharmacokinetic numbers are tighter than community writing usually treats them. Sildenafil's terminal half-life on the FDA label is "about 4 hours" for parent and the active N-desmethyl metabolite, with maximum plasma concentrations reached in 30 to 120 minutes (median 60 minutes) on an empty stomach and a Tmax delay of roughly 60 minutes after a high-fat meal. Absolute bioavailability is around 41 percent. Tadalafil's terminal half-life in healthy adult men is 17.5 hours (5th–95th percentile 11.5–29.6 hours), rising to about 21.6 hours in elderly men; Tmax is around 2 hours; food does not meaningfully alter exposure; clearance is about 2.5 L/h and protein binding is 94 percent. Vardenafil sits between them, with a terminal half-life of approximately 4 to 5 hours, Tmax 30–120 minutes, low absolute bioavailability around 15 percent, and a high-fat meal that delays absorption. Avanafil is the fastest of the four: median Tmax 30–45 minutes fasted, terminal half-life around 5 hours (extended to roughly 10 hours when co-administered with amlodipine, which matters in older hypertensive men). The draft's "17.5 hours" for tadalafil and "4 hours" for sildenafil match the labels exactly. Sources: Viagra label, FDA accessdata 2014, Forgue et al. tadalafil PK PMC1885023, Levitra label, FDA accessdata 2014, Stendra label, FDA accessdata 2015, Tadalafil StatPearls. Confidence: C1.

Calling tadalafil "basically a longer-lasting version" of sildenafil is a useful first approximation but understates two real differences. The first is the action curve. Sildenafil and vardenafil's 4-hour half-life produces a clinical window of 4 to 6 hours, which is why both are event-driven and why sildenafil's absorption is vulnerable to fatty meals. Tadalafil's 17.5-hour half-life keeps plasma concentrations in the therapeutic range for up to 36 hours, the basis of the "weekend pill" framing and the natural daily-dosing route. The second difference is selectivity. Tadalafil is structurally distinct (a methylenedioxyphenyl-pyrazinopyrido-indoledione rather than a sulfonyl-substituted pyrazolopyrimidinone) and is the most PDE6-selective of the marketed agents, with a PDE6/PDE5 ratio around 550-fold versus sildenafil's roughly 16-fold and vardenafil's roughly 21-fold; avanafil sits at about 121-fold. PDE6 is the retinal isoform whose inhibition explains the blue-tinted vision and color disturbances historically reported with sildenafil. Tadalafil also has weaker PDE11 selectivity, which has been linked speculatively to its back-pain and myalgia signal. Tadalafil is more than a long-acting sildenafil clone; it is a structurally different molecule with a different selectivity profile and an unrelated dosing logic. Sources: Wang et al. avanafil selectivity, PMID 22759639, Cialis label, FDA accessdata 2018, PDE5 inhibitors StatPearls. Confidence: C2.

Onset is not interchangeable across the class. Sildenafil is taken 30 to 60 minutes before sex on an empty stomach because food delays Tmax by an hour and blunts the peak; vardenafil behaves similarly; tadalafil is taken 30 minutes to 2 hours before activity, can be taken with food, and the long half-life makes timing imprecise; avanafil is the only agent positioned as a 15-minute-onset product on the basis of post-marketing studies. A patient who eats a steak after a sildenafil dose and then fails 60 minutes later is demonstrating the published food-effect, not a treatment failure. Sources: Viagra label, Stendra label, Levitra label, Cialis label. Confidence: C1.

The libido-versus-physiology claim in the draft is mostly defensible but should not be treated as airtight. The labels are clear that PDE5 inhibitors require sexual stimulation to act and do not cause arousal in the absence of it; the original Steidle and Glina-style work in men without ED found no improvement in self-reported arousal with sildenafil, only a modest reduction in postejaculatory refractory time. Tadalafil's three off-label trials in women with FSAD and antidepressant-associated sexual dysfunction generated some genital-blood-flow signals but no consistent improvement in subjective arousal or satisfaction, which is why no PDE5 inhibitor is approved for any female sexual indication in the United States and why the recent FSAD work has shifted toward topical sildenafil formulations rather than systemic tadalafil. So far, the "no central effect" framing holds. Where it bends is at the edges. There is a real preclinical literature suggesting sildenafil crosses the blood-brain barrier and modulates dopaminergic and serotonergic activity in the medial preoptic area and nucleus accumbens of male rats during sexual arousal; there is human work suggesting subjective improvement in sexual function in some men without ED is greater than placebo on certain endpoints; and the pragmatic reality is that successful erections in a previously failing relationship reduce performance anxiety and partner stress in ways that can secondarily lift desire. The honest summary is that PDE5 inhibitors do not have a clinically reliable libido-augmenting effect in either sex, but the strict statement "no central effects at all" is too strong; the draft's softer framing of "does not generally affect libido directly" is closer to the evidence. Sources: Steidle et al. sildenafil in men without ED, Kyratsas et al. 2013, PMID 23157427, tadalafil/sildenafil for FSAD review, Berman et al. sildenafil women trial, PMID 12150499. Confidence: C2.

The contraindication map is narrow but emphatic, which is the opposite of "some few" if "few" is taken to mean trivial. The absolute contraindication is co-administration with any nitrate or nitric-oxide donor in any form, including isosorbide mononitrate/dinitrate, sublingual nitroglycerin, transdermal nitroglycerin patches, recreational alkyl nitrites ("poppers"), and nicorandil. Nitrates raise cGMP production via guanylate cyclase activation while PDE5 inhibitors block its degradation, and the multiplicative result is a cGMP excursion the vasculature cannot buffer, producing severe and sometimes fatal hypotension. The nitrate-free interval should be at least 24 hours after sildenafil or vardenafil and at least 48 hours after tadalafil. The second absolute contraindication is riociguat, a soluble guanylate cyclase stimulator hitting the same final pathway from another direction; the PATENT-PLUS trial demonstrated unacceptable hypotension. Severe hepatic impairment is a relative contraindication. PAH-specific contraindications include pulmonary veno-occlusive disease, PAH with sickle cell anemia, and left-ventricular outflow obstruction. The list is short, dominated by nitrates and riociguat, and absolute rather than discretionary. Sources: Viagra label, contraindications section, Adempas (riociguat) label, FDA accessdata 2017, Drug-drug interactions in PAH management, PMC9773230, Kloner cardiovascular review. Confidence: C1.

Drug interactions short of contraindication still matter. Doxazosin co-administered with tadalafil 20 mg dropped 28 percent of healthy normotensive men below 85 mmHg standing systolic versus 6 percent on doxazosin plus placebo. Tamsulosin was much better tolerated because it is more uroselective. Guidance is that PDE5 inhibitors should be initiated only after the alpha-blocker is at a stable dose, at the lowest PDE5i dose, with doxazosin and terazosin treated most cautiously. Strong CYP3A4 inhibitors (ritonavir, ketoconazole, itraconazole) raise PDE5i exposure substantially; ritonavir prolongs vardenafil's half-life to about 26 hours. CYP3A4 inducers (rifampin) reduce exposure. Grapefruit juice, HIV protease inhibitors, and systemic azole antifungals need dose adjustment. Sources: Kloner et al. 2004, doxazosin-tadalafil interaction PMID 15540759, Levitra label drug interactions, Cialis label drug interactions. Confidence: C2.

The "few side effects" framing is roughly right for an outpatient drug class, but the specifics deserve more precision than typical patient-facing copy gives them. In the registrational and post-marketing literature, headache (10–16 percent), flushing (10 percent), dyspepsia (4–7 percent), nasal congestion (4 percent), back pain (more pronounced with tadalafil at roughly 6 percent), myalgia (5 percent with tadalafil), and dose-dependent visual disturbances (color vision changes, blue tinge, photophobia) are the consistent common adverse events, and most are dose-related, transient, and reversible on discontinuation. The visual disturbances are highest with sildenafil because of its lower PDE6 selectivity, lower with vardenafil, lowest with tadalafil and avanafil; about 3 to 11 percent of users report some visual symptom on sildenafil and below 1 percent on avanafil. Pharmacovigilance work on the WHO global database confirms the same headline order. The serious complications are rare but real. Non-arteritic anterior ischemic optic neuropathy (NAION), a sudden, painless, monocular optic-nerve infarction associated with small cup-to-disc ratio and other vasculopathic risk factors, has been reported in temporal association with PDE5 inhibitors; the meta-analytic literature is split, with several systematic reviews unable to demonstrate a clear class-level increase in NAION risk and others suggesting a modest excess in heavy users with predisposing optic-disc anatomy. Sudden sensorineural hearing loss has been reported and prompted FDA label updates in 2007; the absolute incidence is very low and causality is debated. Priapism is uncommon at therapeutic doses, with pooled sildenafil RCT data showing about 0.1 percent and FDA AERS data identifying a relatively stronger tadalafil signal, but the absolute risk remains under 1 in 10,000 in standard prescribing. None of these are reasons to avoid the class in healthy men, but the cumulative profile is more textured than "few side effects." Sources: Lui et al. WHO pharmacovigilance, PMID 36905319, Penedones et al. NAION meta-analysis 2020, Burnett et al. priapism analysis, PMID 32622767, PDE5 inhibitors StatPearls, visual side effects narrative review PMC8126729. Confidence: C1-C2.

Tachyphylaxis is one of the few areas where reality is more reassuring than community lore. Six-month randomized double-blind tadalafil work specifically tested for tolerance and did not find it, and broader long-term cohorts have not shown a true pharmacological loss of effect; what patients perceive as "Cialis stopped working" is generally either progression of the underlying vascular disease, deteriorating sleep or testosterone status, or a return to baseline anxiety after an initial confidence boost. Some older sildenafil work reported reductions in efficacy of 15 to 50 percent over 1 to 18 months, but most authors attribute that to organic disease progression rather than tachyphylaxis. The clinical implication is that an apparent loss of response should trigger a search for new comorbidities, not a dose escalation by default. Sources: Tadalafil 6-month tolerance RCT, J Sex Med 2006, El-Meliegy long-term sildenafil, PMID 11490248. Confidence: C2.

Tadalafil's 5 mg-daily benign-prostatic-hyperplasia indication is one of the most underappreciated facts in this field, and it is the lever that explains a large fraction of the "older men live longer on tadalafil" observational signal. The FDA approved daily 5 mg tadalafil for BPH on October 6, 2011, on the basis of randomized trials showing improvement in International Prostate Symptom Score by week 4 and continuing through week 12. A combination tadalafil-finasteride product was later approved for short-term BPH initiation. That regulatory move pushed tadalafil into a long-term, daily, older-male prescribing pattern that sildenafil never developed. The "tadalafil cohort" is therefore older men with cardiovascular risk factors and BPH, exposed continuously over years rather than episodically. Any observational mortality comparison between tadalafil and "no PDE5 inhibitor" must confront that the BPH cohort is a different cohort, with different prescribing pathways and probably different baseline activity. Sources: Lilly Cialis BPH FDA approval press release, Cialis label, BPH indication, tadalafil for BPH review PMC4054509. Confidence: C1.

The pulmonary-arterial-hypertension indications round out the regulatory map: sildenafil 20 mg three times daily as Revatio (FDA-approved 2005) and tadalafil 40 mg once daily as Adcirca (approved 2009). These trials established the cardiovascular-safety database for the class outside the ED literature, because PAH patients are by definition cardiovascularly fragile. They are also the cleanest evidence that long-term, daily, high-dose PDE5 inhibition is tolerable in patients with serious vascular disease. Sources: Revatio (sildenafil) FDA label 2017, Adcirca (tadalafil) label PAH, Pfizer IV Revatio approval press release. Confidence: C1.

The all-cause-mortality literature is where the draft most needs interrogation, because "weak evidence" is doing a lot of work and the field has moved fast since 2021. The pivotal observational study is Andersson et al., JACC 2021, "Association of Phosphodiesterase-5 Inhibitors Versus Alprostadil With Survival in Men With Coronary Artery Disease." That study used the Swedish Patient Register and Prescribed Drug Register to compare 16,548 men prescribed PDE5 inhibitors with 1,994 prescribed alprostadil after a prior myocardial infarction or revascularization between 2006 and 2013, with mean follow-up of 5.8 years. PDE5i exposure was associated with lower mortality versus alprostadil (hazard ratio 0.88, 95% CI 0.79–0.98), with a dose-response: men in the upper three quintiles of PDE5i prescriptions had progressively lower all-cause mortality than men in the lowest quintile, and similar associations were observed for myocardial infarction, heart failure hospitalization, and cardiac revascularization. Andersson et al. were explicit that the observational design forbids causal inference and called for randomized trials. Sources: Andersson et al. JACC 2021, PMID 33766260, Maas & Rodionov accompanying editorial PMID 33766261. Confidence: C1.

Anderson et al., Heart 2017, ran the Cheshire diabetic-male cohort: 5,956 men aged 40–89 with type 2 diabetes followed for up to 7.5 years, in which PDE5-inhibitor users had 19.1 percent mortality versus 23.8 percent in non-users and a 31 percent lower adjusted all-cause mortality hazard, with effects most pronounced in subgroups with prior heart failure, transient ischemic attack, or peripheral vascular disease. Kloner and colleagues then published a US commercial-claims cohort in Clinical Cardiology in 2024 with 8,156 tadalafil-exposed and 21,012 unexposed men, propensity-matched, in which all-cause mortality was 44 percent lower with tadalafil exposure (HR 0.56, 95% CI 0.43–0.74) and mortality fell further with higher quartile exposure (HR 0.40 for the top quartile). The most aggressive recent paper, Jehle et al. in the American Journal of Medicine 2025, used the TriNetX database with 509,788 men with ED and 1,075,908 with lower-urinary-tract symptoms, and reported tadalafil/sildenafil relative risks of 0.66/0.76 for all-cause mortality, 0.73/0.83 for myocardial infarction, 0.66/0.78 for stroke, 0.79/0.80 for venous thromboembolism, and 0.68/0.75 for dementia at three years of follow-up, with tadalafil consistently showing larger associations than sildenafil. A systematic review and meta-analysis of 16 studies in European Heart Journal-Cardiovascular Pharmacotherapy 2024 (Mostafa et al.) summarized the same direction: PDE5i exposure associated with significantly lower MACE and lower all-cause mortality in pooled analysis, with tadalafil-specific subgroups in line with the class. Sources: Anderson et al. type 2 diabetes Heart cohort, PMID 27465053, Kloner et al. Clinical Cardiology 2024, PMID 38377018, Jehle et al. AJM 2024 PMID 39532245, Mostafa et al. systematic review and meta-analysis PMC11323371. Confidence: C1 for the existence and effect sizes, C3 for clinical interpretation.

The interpretation is harder than the headline numbers. Every one of these studies is observational. Confounding by indication is severe in both directions. Men who fill PDE5-inhibitor prescriptions tend to be sexually active, sufficiently mobile and alert to seek and use the drug, and well-enough engaged with primary care to receive the prescription in the first place; they are also more likely to be on statins, antihypertensives, antiplatelets, and other secondary-prevention medication that independently lowers cardiovascular and all-cause mortality. Healthy-user bias is therefore not a small caveat; it is the default explanation that any cohort design has to defeat before it can claim a treatment effect. The Andersson et al. design partially addresses this by using alprostadil as the active comparator, since alprostadil-treated men also have ED but cannot benefit from any systemic PDE5 effect; that design narrows the bias substantially, which is one reason JACC accepted it for publication despite the field's previous skepticism. The Kloner and Jehle designs are weaker on this point because they compare PDE5i users to general non-users, which reintroduces the very healthy-user bias the comparator-design was meant to remove. The Mendelian randomization analysis in European Heart Journal-Cardiovascular Pharmacotherapy 2025 (Xiao et al.) is the most causal-inference-friendly piece of work in the literature: genetically predicted increased PDE5A expression correlates with elevated coronary heart disease and myocardial infarction risk, and the inverse relationship is consistent with a real protective effect of PDE5 inhibition that runs partly through blood-pressure and LDL-cholesterol pathways. Sources: Xiao et al. PDE5 MR analysis EHJ-CVP 2025, Mostafa et al. EHJ-CVP meta-analysis 2024, JACC 2023 nitrate-PDE5i interaction work. Confidence: C2-C3.

The dementia and Alzheimer literature is a partly separate strand of the same conversation. Fang et al., Nature Aging 2021 (Cleveland Clinic, Cheng group), reported a 69 percent lower risk of incident Alzheimer's disease in sildenafil users in a 7.23-million-person retrospective insurance-claims study, with mechanistic support from iPSC-derived neuronal models showing reduced phospho-tau and increased neurite growth on sildenafil exposure. A 2024 follow-up in the Journal of Alzheimer's Disease replicated the association in a different claims database and added more patient-iPSC neuronal mechanistic data. However, a separate analysis using endothelin-receptor antagonists (used for the same PAH patients) as the active comparator failed to find a difference in incident Alzheimer's risk, which suggests the original 69 percent number is at least partly explained by confounding rather than a clean drug effect. The Jehle et al. paper above also reported a dementia signal (RR 0.68 tadalafil, 0.75 sildenafil). The honest summary is that the dementia literature is intriguing, mechanistically plausible (cGMP-mediated vascular and tau effects), and reasonably consistent in one direction, but it is not yet trial-grade and the strongest signals come from designs most vulnerable to confounding. Sources: Fang et al. Nature Aging 2021, PMID 35572351, Cleveland Clinic 2024 follow-up PMID 38427489, Jehle et al. AJM 2024. Confidence: C2-C3.

The "weak evidence" language in the draft therefore does not match the current shape of the literature. There is now multi-cohort, multi-database, multi-country observational evidence; a class-level meta-analysis; an active-comparator design that addresses the most obvious bias; and a Mendelian-randomization analysis that goes some way toward causal inference. None of those individually amounts to a randomized placebo-controlled mortality trial, and that gap is unlikely to be filled because no funder will randomize older men to placebo for years to test an off-label survival benefit. So the better framing is: there is consistent observational and quasi-causal evidence that PDE5-inhibitor exposure, particularly tadalafil, is associated with roughly 12 to 44 percent lower all-cause mortality across cohorts, that the effect is dose-related, that it is mechanistically plausible through endothelial and platelet effects, and that residual confounding remains real and probably explains some fraction of the signal. "Weak" is too dismissive; "strong but not RCT-grade" is closer to the mark, and the asymmetric reporting in the press tends to ignore the fact that healthy-user bias has not been definitively excluded. Confidence: C2-C3.

The newer agents still matter clinically. Avanafil (Stendra, FDA-approved 2012) is the most PDE5-selective of the marketed agents in functional terms (about 100-fold versus PDE6, minimal cross-reactivity with other PDE isoforms), with a 30-to-45-minute Tmax and a 5-hour half-life. The pitch is fewer visual disturbances, fewer myalgias, and faster onset than sildenafil; the limit is cost and thin cardiovascular outcome data. Vardenafil (Levitra, also as orally disintegrating Staxyn) occupies a niche where its slightly faster Tmax and slightly cleaner profile trump sildenafil familiarity. Tadalafil's combination of long half-life, food-independence, BPH-licensure, low visual side effects, and growing observational mortality data has made it the practical default in the men's-health stack, including the unregulated compounded tadalafil-finasteride-minoxidil products sold by direct-to-consumer telehealth. Sources: Stendra label 2015, avanafil clinical review PMC4542406, Wang et al. avanafil selectivity PMID 22759639. Confidence: C2.

The 5α-reductase-inhibitor / hair-loss niche deserves a brief, honest treatment. Finasteride and dutasteride are the only rigorously studied therapies for androgenetic alopecia; tadalafil has no high-quality human evidence for hair regrowth and is not approved for it anywhere. It appears alongside finasteride in compounded men's-health products partly because a single bundled prescription supports both ED and hair-loss markets, and partly because some men on finasteride report sexual side effects that low-dose daily tadalafil may offset. The combination tadalafil-finasteride product is approved for BPH, not hair loss. Sources: Finasteride StatPearls, PDE inhibitors for alopecia review PMID 30935254. Confidence: C2-C3.

The "few side effects" framing is reasonable for a low-risk daily medication in a healthy outpatient population but should not be confused with "no contraindications" or "no warnings." Headache, flushing, dyspepsia, nasal congestion, back pain (tadalafil), and reversible visual disturbances (sildenafil) are common-but-tolerated; nitrate co-administration is a hard contraindication because of fatal hypotension; alpha-blocker co-administration requires care; NAION, sudden hearing loss, and priapism are rare-but-serious; and the observational mortality signal probably reflects a real but partially confounded benefit rather than a randomized-trial-grade survival effect. Confidence: C2.

Verdict on each numbered claim

1. "Sildenafil/Viagra is a blood-flow enhancer; just makes erections easier." TRUE in everyday usage and consistent with the FDA label and StatPearls mechanism. Mild caveat: there is preclinical and limited human evidence of central effects (dopaminergic/serotonergic in MPOA and NAcc, modest fMRI signals), and a small but real reduction in postejaculatory refractory time, but the dominant clinical action is peripheral vasodilation. Sources: Viagra label, Kyratsas 2013, PMID 23157427.

1. "Sildenafil does NOT generally affect libido directly — only physiology." KIND-OF-TRUE. The strict claim "no effect on libido at all" is too strong; preclinical and some human data suggest modest CNS effects, and indirect effects via reduced performance anxiety are real. The framing "does not generally affect libido directly" is broadly defensible. The class is not a libido drug. Sources: Steidle PMID 12904810, PMC2643112 FSAD review.

1. "Tadalafil/Cialis is 'basically a longer-lasting version' of sildenafil." KIND-OF-TRUE. Same drug class, same mechanism, but tadalafil is a structurally different molecule with very different PDE6 selectivity, a different food-effect profile, daily-dosing biology, and a separate FDA-approved BPH indication. Useful first approximation, incomplete as a description. Sources: Cialis label, Wang et al. 2012.

1. "Tadalafil elimination half-life is 17.5 hours." TRUE. Matches the FDA label and Forgue et al. healthy-volunteer pharmacokinetic study (5th–95th percentile 11.5–29.6 hours, 21.6 hours in elderly men). Sources: Tadalafil PK PMC1885023, Cialis label.

1. "Sildenafil elimination half-life is 4 hours." TRUE. Matches the Viagra label exactly: parent and active metabolite both about 4 hours. Sources: Viagra label.

1. "PDE5 inhibitors are contraindicated with 'some few' medical conditions and other medications." KIND-OF-TRUE. The list is short but absolute on its critical entries. Nitrates and riociguat are absolute contraindications because the hypotensive interaction is potentially fatal; alpha-blockers (especially doxazosin) require care; severe hepatic impairment and certain PAH-specific conditions narrow eligibility further. The framing "some few" understates the seriousness of the nitrate interaction. Sources: Viagra label, Adempas (riociguat) label, Kloner 2004, PMID 15540759.

1. "PDE5 inhibitors generally have few side effects." KIND-OF-TRUE. For an oral outpatient drug used episodically or at low daily dose, the side-effect burden is modest. Common reactions (headache, flushing, dyspepsia, nasal congestion, back/myalgia, transient visual changes) are usually self-limited; rare serious complications (NAION, sudden hearing loss, priapism) exist and matter. "Few" is fair but should be paired with the FDA-labeled warnings rather than treated as "no side effects." Sources: WHO pharmacovigilance Lui 2023, PMID 36905319, PDE5i StatPearls.

1. "There is 'weak evidence' tadalafil might be correlated with reduced all-cause mortality in older people." KIND-OF-TRUE but understated. The evidence is now multi-cohort, multi-country, dose-responsive, with one active-comparator design (Andersson 2021) and one Mendelian-randomization analysis (Xiao 2025) that begin to address confounding. Effect sizes range from HR 0.88 (active-comparator) to HR 0.56 (general-non-user comparator) for all-cause mortality, with consistent direction across studies and stronger signals for tadalafil than sildenafil in head-to-head comparisons. Healthy-user bias and confounding by indication remain real, and no RCT will close the gap, but "weak" undersells the consistency. A more accurate framing would be "consistent observational evidence with residual confounding, no randomized confirmation." Sources: Andersson JACC 2021 PMID 33766260, Anderson Heart diabetic cohort PMID 27465053, Kloner Clin Cardiol 2024 PMID 38377018, Jehle AJM 2024 PMID 39532245, Xiao MR analysis EHJ-CVP 2025, Mostafa systematic review PMC11323371.