Sildenafil

What sildenafil is and how it works

Sildenafil is a small-molecule inhibitor of phosphodiesterase type 5. The US register does not recognize the bare molecule name at all: enumerating the distinct ingredient strings inside the Drugs@FDA result set for this compound on 1 September 2026 returns one value on every product, SILDENAFIL CITRATE, and label strengths are printed as base equivalents. That is the name this page uses when it is talking about the register, and "sildenafil" when it is talking about the molecule.

The mechanism the approved labeling describes is downstream rather than direct. Sexual stimulation releases nitric oxide locally; nitric oxide activates guanylate cyclase; guanylate cyclase raises cyclic guanosine monophosphate, which relaxes smooth muscle and allows blood to flow in. Sildenafil does not start that chain — it blocks the enzyme that ends it, so cyclic GMP persists. The same labeling states plainly that the drug has no direct relaxant effect on isolated human erectile tissue and no effect at all in the absence of sexual stimulation. That precondition is built into the mechanism, and it is why the drug is described as enabling a response rather than producing one.

The selectivity numbers are the most molecule-specific thing on this page, and one of them is unusually small. The approved labeling reports sildenafil as roughly 10-fold more potent against PDE5 than against PDE6, more than 80-fold against PDE1, and more than 700-fold against PDE2, PDE3, PDE4, PDE7, PDE8, PDE9, PDE10 and PDE11 — with the PDE3 margin about 4,000-fold, which the label notes because PDE3 is involved in the control of cardiac contractility. PDE6 is a retinal enzyme in the phototransduction pathway, and the label attributes color-vision abnormalities to that 10-fold margin by name.

The pharmacokinetics are equally specific. Both sildenafil and its active N-desmethyl metabolite have terminal half-lives of about four hours; in children weighing 10 to 70 kg the estimated half-life range is 2.9 to 4.4 hours. A high-fat meal delays the time to peak concentration by a mean of 60 minutes and lowers peak concentration by a mean of 29%. Roughly four hours of exposure is what the labeled dosing instruction is built around, and it is why that instruction is written as a window before an event rather than as a standing daily regimen.

One molecule, two therapeutic destinations. The pulmonary indication exists because the same enzyme is present in lung vascular smooth muscle, so inhibiting it enhances cyclic-GMP-mediated relaxation and growth inhibition of those cells as well.^[5] Those destinations sit on separate applications, at separate amounts, by separate routes, and this page keeps them separate throughout.

What the research actually shows

The evidence landscape here is large enough that a citation count means nothing. A PubMed search on 1 September 2026 for sildenafil returned 9,433 records; sildenafil citrate[MeSH Terms] returned 6,183 — the descriptor is the salt name, not the bare molecule; sildenafil AND randomized controlled trial[pt] returned 726; sildenafil AND meta-analysis[pt] returned 145; sildenafil AND erectile dysfunction returned 3,275; and sildenafil AND pulmonary arterial hypertension returned 1,250. What follows is a selection from 726 indexed randomized trials, made by one rule: for each claim area, the largest or most recent randomized trial or systematic review reporting that outcome for this molecule at a labeled amount, by the route the indication names, in the population it names — plus its strongest published counterweight.

Erectile dysfunction. A 2026 systematic review and dose-response network meta-analysis pooled 83 double-blind placebo-controlled randomized trials in 6,029 participants and reported sildenafil 100 mg (odds ratio 9.06) and 50 mg (odds ratio 7.90) as the highest-ranked on-demand options for reaching a satisfactory erectile-function score, ahead of tadalafil 20 mg (7.13), vardenafil 10 mg (7.78) and avanafil 200 mg (3.42); a threshold analysis put sildenafil 100 mg at odds ratio 2.20 (95% CI 1.78 to 2.63) and 50 mg at 2.10 (95% CI 1.69 to 2.50) against placebo.^[1] Both halves. The same analysis reports the odds of treatment-related adverse events as highest for vardenafil and lowest for tadalafil, with sildenafil in between, and its authors note that wide confidence intervals leave real uncertainty in some estimates. And the disclosure that goes with it: two of the twelve authors give a departmental affiliation at the company that is the current sponsor of record for the US erectile-dysfunction and pulmonary-hypertension applications for this molecule. That match was made by comparing the affiliation strings printed in the record's own author-information block against the sponsor field of the Drugs@FDA application records, not by matching surnames, which would not have found it.

Underneath that sits the registration-era safety programme: a pooled analysis of double-blind placebo-controlled studies and ten open-label extensions, with 4,274 men aged 19 to 87 receiving double-blind treatment for up to six months and 2,199 continuing for up to a year.^[2] In a defined organic cause, 27 men with erectile dysfunction from spinal cord injury between T6 and L5 were randomized for 28 days; 9 of 12 on sildenafil against 1 of 14 on placebo reported improved erections (P = 0.0043).^[4]

And the trial a PHL reader should see first. Sixty healthy men aged 20 to 40 with no reported erectile dysfunction were randomized to a single 25 mg tablet or placebo before intercourse. Reported improvement in erection quality was 12 of 30 against 10 of 30 (P = 0.79) — no difference. The one significant result was a reduction in the postejaculatory refractory time (12 of 30 against 4 of 30, P = 0.04), and the authors' stated conclusion is that the drug does not improve erections in young healthy men.^[3]

Pulmonary arterial hypertension in adults. The 278-patient registration trial reported placebo-corrected six-minute walk distance gains of 45, 46 and 50 metres at 20, 40 and 80 mg three times daily (P < 0.001 for all), with improvements in mean pulmonary-artery pressure and WHO functional class — and, in the same abstract, the incidence of clinical worsening did not differ significantly from placebo, and the trial was explicitly not powered for mortality.^[5] That record carries a published erratum and three published comments with author replies, which is part of reading it honestly. The regulator-mandated follow-up randomized 385 adults to 5, 20 or 80 mg three times daily; it was halted at the first interim analysis on meeting noninferiority of 80 mg against 5 mg for all-cause mortality (hazard ratio 0.51, 99.7% CI 0.22 to 1.21), with time to clinical worsening favouring 80 mg (hazard ratio 0.44, 99.7% CI 0.22 to 0.89) and a 18.9 metre walk-distance advantage (95% CI 2.99 to 34.86, P = 0.0201).^[6] Both halves again: no significant difference was found between 80 mg and 20 mg on mortality, clinical worsening or walk distance, and adverse-event-related discontinuations were numerically higher at 80 mg. Its ClinicalTrials.gov record NCT02060487 carries posted results, a status of terminated, and a stated reason — stopped by the sponsor on a data monitoring committee recommendation after the first interim analysis because the primary objective was met.

Pulmonary arterial hypertension in children. In 235 treatment-naive children aged 1 to 17 weighing at least 8 kg, the primary comparison — percent change in peak oxygen consumption for the three doses combined against placebo — was 7.7 plus or minus 4.0 percent (95% CI -0.2 to 15.6, P = 0.056), which the authors describe as only marginally significant; functional class and hemodynamics improved at the medium and high doses and the low dose was ineffective.^[7] The extension reported 37 deaths, three-year Kaplan-Meier survival of 94%, 93% and 88% by ascending dose randomization, and a mortality hazard ratio of 3.95 (95% CI 1.46 to 10.65) for high versus low dose — while stating in the same abstract that multiple analyses raised uncertainty about the survival-dose relationship and that all dose groups showed favorable survival for this condition.^[8] Both registry records carry posted results.

Where the evidence is weak

The population gap is the first weakness and it is the one that matters most to the reader of this page. The 726 randomized trials above enrolled men with diagnosed erectile dysfunction, adults with symptomatic pulmonary arterial hypertension, and children with the same disease. The one randomized trial on this page that enrolled men who did not have erectile dysfunction found no improvement in erection quality.^[3] A wellness reader is much more likely to resemble that trial's population than the registration programme's, and no efficacy figure from the latter transfers to the former.

Surrogates are not outcomes. Six-minute walk distance and peak oxygen consumption are the endpoints the pulmonary approvals were built on; they are measures of exercise capacity, not of survival. The registration trial said so itself, and the trial that was actually designed to test mortality tested it between doses rather than against placebo, so neither establishes that treatment extends life against no treatment.^[5][6]

Class evidence is not molecule evidence. The 2025 Cochrane review of a PDE5 inhibitor added to an endothelin receptor antagonist reports a high-certainty reduction in clinical worsening against the antagonist alone (risk ratio 0.53, 95% CI 0.41 to 0.68) and very-low-certainty evidence for the mirror comparison (risk ratio 0.68, 95% CI 0.33 to 1.39) — but its included regimens pool sildenafil with tadalafil, so no row in it is a sildenafil row.^[9] The melanoma meta-analysis has the same shape and the divergence is sharper: the class estimate reaches significance (odds ratio 1.60, 95% CI 1.13 to 2.27, P = 0.009) while the sildenafil-specific estimate does not (odds ratio 1.85, 95% CI 0.97 to 3.51, P = 0.059), on I-squared of 99%, in pooled observational data — and the authors nonetheless close by recommending avoidance for people with a skin-cancer history, which their own confidence intervals do not support.^[12] The VigiBase analysis is a case-non-case disproportionality study indexed by PubMed as a journal article rather than as a systematic review or meta-analysis; it is hypothesis-generating, and its authors say so.^[13]

Outside the approved populations the evidence thins immediately. A 2025 systematic review of nitric-oxide-pathway drugs in pulmonary hypertension secondary to chronic obstructive pulmonary disease found 14 studies in 567 adults, randomized and non-randomized mixed, and opens by stating that no therapy is approved for that group.^[10] Below the pediatric approval's lower age bound, a 2024 randomized trial compared oral sildenafil with bosentan in newborns with persistent pulmonary hypertension — a different disease in a different age band, cited here only to mark where the approval stops.^[11]

Two dated searched negatives. A PubMed search on 1 September 2026 for sildenafil AND sudden sensorineural hearing loss AND randomized controlled trial[pt] returned 0 records, so the labeled hearing-loss warning described below rests on post-marketing reporting rather than on a randomized trial. A PubMed search on the same date for sildenafil AND pulmonary arterial hypertension AND child[MeSH Terms] AND randomized controlled trial[pt] returned 10 records, which is the whole indexed randomized pediatric base for a population the label affirmatively indicates.

The adulteration thread is real, current, and is a finding about products rather than about the molecule. FDA's Health Fraud Product Database, retrieved from fda.gov on 1 September 2026 and parsed as a table of 2,197 data rows rather than searched as text, returns 491 rows naming sildenafil in the Subject column, spanning 2007 to 2026, split by action into 376 public notifications, 102 recalls, 6 warning letters, 6 news releases and 1 further row whose action cell carries a misspelling of "Public Notification" in the source, counted separately here rather than silently merged. Counting the raw page text instead of the table returns 530, which is why the figure is a table parse and not a text search. Of those 491 rows, 457 name sildenafil itself and 34 name only a structurally modified analogue. The split is made by one rule, applied identically on this page and on the tadalafil page: a row counts as naming the ingredient itself only where the Subject cell carries the bare ingredient name as its own word, with no chemical modifier attached to it or standing immediately in front of it — so hydroxythiohomosildenafil reads as an analogue name, and so do the space-separated dimethyl sildenafil thione and propoxyphenyl sildenafil, and a row reading only "undeclared sildenafil analogues", "undeclared analogue of sildenafil" or "undeclared substances that are similar in chemical structure to sildenafil" names no parent either. Of the 34, 24 name a specific analogue: hydroxythiohomosildenafil in 7; thiosildenafil, sulfosildenafil and thiomethisosildenafil in 3 each, one of the thiomethisosildenafil rows spelling it thio-methisosildenafil; dimethylsildenafil, hydroxyhomosildenafil and dimethyl sildenafil thione in 2 each; and hydroxylthiohomosildenafil, propoxyphenyl sildenafil, sulfohydroxyhomosildenafil and sulfohomosildenafil in 1 each. That is 26 analogue mentions across 24 rows, because two of the rows name two analogues apiece, which is why the per-analogue figures sum higher than the row count. The remaining 10 rows name an analogue class without naming a molecule. All of these are chemically distinct molecules and are not sildenafil. Separately, openFDA's drug enforcement endpoint returned 193 records on an unfielded full-text search for sildenafil on the same date, with report dates from 29 August 2012 to 19 August 2026 — 165 of them Class I, all 193 firm-initiated, and 154 whose stated reason names an undeclared or unapproved active ingredient. The same endpoint on the fielded openfda.generic_name path returned HTTP 404 for sildenafil on that date, against a same-session positive control on that same field — openfda.generic_name for tadalafil, HTTP 200, 9 records — so the zero is a fact about that field and not about the route: these records are not annotated as approved drug products at all, which is the clearest available measure of what that thread is. Every one of these rows is an action against a finished article marketed without an approved application; none of them is a safety finding about the sildenafil in an approved product, and a Class I classification describes the hazard of an undisclosed ingredient rather than the record of the disclosed one. The analogue names are the part that matters for a reader: a modified molecule is designed to be missed by an assay looking for the parent one, and nothing on the label of an unapproved article tells anyone which is present or at what amount.

Legal and regulatory status

Sildenafil is FDA-approved, and the approval is not one thing. Each of the sentences below is scoped to a product, a route, an amount and a population, because none of them generalizes to the others.

The census, with its field named. On 1 September 2026 the products.active_ingredients.name field in Drugs@FDA, searched for "SILDENAFIL", held 58 applications covering 106 products; the openfda.generic_name field in the same database on the same day held 40 applications for "sildenafil" — a third fewer, for a reason that is about the index rather than about the register. The row sets are nested, not disjoint: all 40 of the smaller set's application numbers appear in the larger, none of the larger's are absent from it in the other direction, and of the 18 applications visible only to the ingredient field, 17 omit the openfda key from the returned record entirely and the remaining one returns it as the empty object {} — read by key presence rather than by searching for a string. The smaller field under-counts by about 31% and the larger, ingredient-field figure is the defensible one. Enumerating the distinct ingredient values inside those 106 products returns exactly one string — SILDENAFIL CITRATE, on 106 of 106 — so unlike some molecules in this library this count is not inflated by a longer ingredient name that merely contains the word. One route note: the exact-match query products.active_ingredients.name.exact:"SILDENAFIL" returns HTTP 404, while the same field with the full salt string returns 200 and the same 58 — the correct exact string is the salt name, and a 404 there is a fact about the string, never about the register. A same-session negative control on that same exact field, products.active_ingredients.name.exact:"ZZZNOTADRUGZZZ", also returned 404, which is what establishes that this endpoint answers an empty set with a 404 rather than with a zero total.

Inside those 58 applications: 6 NDAs and 52 ANDAs. The six NDAs are NDA 020895 (the erectile-dysfunction tablet), NDA 021845, NDA 022473 and NDA 203109 (the pulmonary-hypertension tablet, intravenous solution and powder for oral suspension respectively), NDA 210858 (an oral film) and NDA 214952 (an oral suspension). By route, 104 products are oral and 2 are intravenous — the intravenous presentation is real and is not a footnote here. Every one of the 106 products carries a populated marketing_status, and the three values it takes are: 79 Prescription, 18 Discontinued and 9 None (Tentative Approval) printed in full, which is the register's way of saying the application is tentatively approved and the product is not marketed, not an empty field; NDA 203109 and NDA 214952 are both Discontinued and must not be described as currently marketed. Zero products are over the counter.

The indications, verbatim and each on its own application. For the erectile-dysfunction product the entire indication section is one sentence: "VIAGRA is indicated for the treatment of erectile dysfunction." There are no subsections and no limitation of use. For the pulmonary product, section 1 reads "REVATIO is indicated for the treatment of pulmonary arterial hypertension (PAH) (World Health Organization [WHO] Group I) in adults to improve exercise ability and delay clinical worsening" and, separately, "REVATIO is indicated in pediatric patients 1 to 17 years old for the treatment of pulmonary arterial hypertension (PAH) (WHO Group I) to improve exercise ability and, in pediatric patients too young to perform standardized exercise testing, pulmonary hemodynamics thought to underlie improvements in exercise." Both labels were read in this session sorted by effective date rather than by version number, which matters here: the current erectile-dysfunction label is version 6 with an effective date of 17 November 2023, while a superseded document for the same product carries version 32 from 15 December 2017. A higher version number on an older document is exactly how a stale label gets quoted as current. The pulmonary label read is version 5, effective 18 December 2024, covering all three of its applications and both routes.

What an approval does not settle is whether the trials behind it describe the person reading about it. The erectile-dysfunction indication rests on randomized trials in men who had the disorder; the one randomized trial on this page that enrolled men who did not found no improvement in erection quality against placebo.^[3]

The labeled amounts, which are regulatory facts and not protocol ranges. The erectile-dysfunction label's section 2.1 describes an amount timed to an event rather than a standing regimen: "For most patients, the recommended dose is 50 mg taken, as needed, approximately 1 hour before sexual activity. However, VIAGRA may be taken anywhere from 30 minutes to 4 hours before sexual activity." It states a ceiling — "The maximum recommended dosing frequency is once per day" — which is a limit, not a once-daily regimen, and this product has no once-daily indication of any kind. Section 2.3 gives a single undifferentiated instruction for one interacting class, with a named number: patients should be stable on alpha-blocker therapy first, and the highlights state it as "With concomitant use of alpha-blockers, initiate VIAGRA at 25 mg dose." On the pulmonary side the labeled amounts are wholly different: 20 mg by mouth three times a day in adults, titratable to a maximum of 80 mg three times a day; 10 mg three times a day as an intravenous bolus, which the label says "does not need to be adjusted for body weight" and is "predicted to provide pharmacological effect… equivalent to that of a 20-mg oral dose"; and in children a weight-banded schedule of 10 mg three times a day at 20 kg or below and 20 mg three times a day from 20 kg upward.

The contraindications are absolute in current labeling, and this molecule's class statement is wide. Section 4.1 contraindicates administration to patients using "nitric oxide donors such as organic nitrates or organic nitrites in any form", whether used regularly or intermittently. The same section then declines to give a post-dose interval, in words worth carrying exactly: "After patients have taken VIAGRA, it is unknown when nitrates, if necessary, can be safely administered. Although plasma levels of sildenafil at 24 hours post dose are much lower than at peak concentration, it is unknown whether nitrates can be safely co-administered at this time point." Section 4.2 contraindicates known hypersensitivity to sildenafil or any component, and section 4.3 contraindicates concomitant guanylate cyclase stimulators such as riociguat. The pulmonary label carries the same nitrate and riociguat contraindications, and adds that hypersensitivity reactions including anaphylaxis have been reported.

The labeled warnings, each with the section it comes from and each with the label's own hedge intact. Section 5.2 records prolonged erection greater than four hours and priapism as infrequent post-approval reports and instructs that an erection persisting beyond four hours be met with immediate medical assistance. Section 5.3 covers non-arteritic anterior ischemic optic neuropathy and reports observational case-crossover risk estimates of 2.15 (95% CI 1.06 to 4.34) and 2.27 (95% CI 0.99 to 5.20) — then states that neither those studies nor the rare post-marketing reports "substantiate a causal relationship". Section 5.4 covers sudden decrease or loss of hearing and states that "it is not possible to determine whether these events are related directly to the use of PDE5 inhibitors or to other factors". Section 5.5 covers hypotension when combined with alpha-blockers or antihypertensives. Section 12.2 reports the molecule-specific finding the selectivity numbers predict: "At single oral doses of 100 mg and 200 mg, transient dose-related impairment of color discrimination was detected using the Farnsworth-Munsell 100-hue test", with subjects reporting it as difficulty discriminating blue and green.

There is no boxed warning, stated as a field-scoped finding rather than as a verdict. A query of the openFDA drug labeling endpoint on 1 September 2026 for openfda.generic_name:"sildenafil" AND _exists_:boxed_warning returned HTTP 404 — an empty set — against a base query on the same field, the same endpoint and the same date returning 167 sildenafil labels, of which 167 populate contraindications and 167 populate warnings_and_cautions. A same-session positive control run on that same openfda.generic_name field for estradiol returned 426 labels carrying a populated boxed_warning field, which is what distinguishes an empty register from a malformed query. This is a statement about one coded SPL field across one label set on one date. It is not a statement that this drug carries no serious warnings — the warnings above are in the labeling, in named sections, and one of them is an absolute contraindication.

Section 8.4 of the pulmonary label carries both halves of the pediatric question and both belong on this page. It records that during the pediatric studies "an imbalance in the number of deaths was noted: 5/55 (9.1%), 10/74 (13.5%), and 22/100 (22%) in the sildenafil low, medium, and high dose groups, respectively" — and, in the same section, that "this safety observation in pediatrics was not confirmed in a study conducted in adults designed to evaluate this risk (Study A1481324)", concluding that a causal association for the observed dose-related mortality effect is unlikely. The current label affirmatively indicates use in 1-to-17-year-olds; a page asserting that the agency recommends against pediatric use would be describing a document that is no longer current.

Anti-doping. Retrieved on 1 September 2026 through the United States Anti-Doping Agency's prohibited-list page, which returned HTTP 200 to a browser user agent and linked the WADA-hosted PDF of the World Anti-Doping Code International Standard Prohibited List 2026: a case-insensitive search of the full extracted text of that document for "sildenafil" returned 0 matches, and a search for "phosphodiesterase" returned 0, while a same-document control search for "testosterone" returned 32 matches — so the null is a fact about the document, not about the extraction. Sildenafil likewise does not appear anywhere in the 2026 Monitoring Program retrieved by the same route on the same date, where the same search returned 0 for "sildenafil" against in-document controls returning 1 line each for "tramadol", "caffeine" and "ecdysterone". Section S0 does not reach it, and the reason is explicit rather than inferred. S0 has two conjunctive conditions — a substance must be "not addressed by any of the subsequent sections of the List" and carry "no current approval by any governmental regulatory health authority for human therapeutic use". The second condition fails here: this molecule holds current US approvals in prescription marketing status on four of its six NDAs — NDA 020895, NDA 021845, NDA 022473 and NDA 210858 — the remaining two, NDA 203109 and NDA 214952, being Discontinued. Absence from the List is not a statement that a substance is permitted, and a tested athlete's therapeutic-use question belongs with the relevant sport body before treatment rather than after a test.

The compound-level evidence badge on this page is human RCT, and the rationale is worth saying out loud: the claim a reader typically arrives with — does this help an erection — is answered by randomized placebo-controlled trials in men with the disorder, not by animal work or case series. The badge grades the typical claim. It does not grade the claim a reader may actually be making about themselves.

Questions to bring to a provider

The single most consequential question is about other medicines, not about this one. Anything in the nitrate family in any form, and any guanylate cyclase stimulator, is an absolute contraindication in current labeling rather than a caution — and the label itself declines to say when a nitrate could be given after a dose. Alpha-blockers and other blood-pressure medicines are a separate conversation with a labeled starting-amount instruction attached. A full, honest list of everything currently taken is where that conversation starts.

Second: which evidence base is being invoked. An erectile-dysfunction question and a pulmonary-hypertension question reach two different applications, two different amounts, two different routes and two different trial literatures, and a figure from one says nothing about the other.^[5]

Third: whether the trials apply. If the question is about performance rather than about a diagnosed disorder, the closest randomized evidence is the 60-participant trial that found no improvement in erection quality in men without erectile dysfunction — a result worth putting in front of a clinician rather than around.^[3]

Fourth, the labeled caution list: a history of non-arteritic anterior ischemic optic neuropathy, a "crowded" optic disc, retinitis pigmentosa, sickle cell anemia, multiple myeloma, leukemia or an anatomical deformation is each named in current labeling as a reason for caution, and each is a reason to raise the subject with a clinician before anything is considered.

Fifth, cardiovascular status. The labeling frames the prior question as whether sexual activity itself is advisable, and lists conditions with no controlled safety data at all — recent myocardial infarction, stroke or life-threatening arrhythmia, resting hypotension or uncontrolled hypertension, and cardiac failure or unstable angina.

And last, for anyone considering an article sold for this purpose outside a pharmacy: an unfielded full-text search of the openFDA drug enforcement endpoint for sildenafil on 1 September 2026 returned 193 records, most of them Class I actions naming an undeclared or unapproved active ingredient, and some of them naming structurally modified analogs that a routine assay looking for the parent molecule would not report. The class-level melanoma signal and the priapism reporting signal are both worth raising in the same conversation, with the caveat that neither is a randomized finding.^[12][13]