The Viagra Cancer Study Is Firmer Than Expected

Two things happened in the same paper.

The first is cell biology. In Cancer Research, a team led by Yarden Ariav in Ayelet Erez's lab at the Weizmann Institute reports that sildenafil traps cholesterol inside lysosomes. The route is odd enough to be interesting. Sildenafil blocks PDE5, which raises levels of the signalling molecule cGMP, and cGMP turns out to bind NPC1, the transporter that ferries cholesterol out of the lysosome. Block the export and the cell runs short of a molecule it needs to build membranes. The authors describe the outcome as "phenocopying Niemann–Pick type C pathology", the lysosomal storage disease caused by NPC1 mutations. Cancer cells were more vulnerable than normal cells, which the paper attributes to their reduced lysosomal gene expression.

That is a checkable claim about a molecule an estimated 90 million men have taken since 1998, and it arrives with mouse models, patient-derived cell cultures, and microscopy showing cholesterol stranded in the lysosomes of sildenafil-treated breast cancer tissue.

An image of breast cancer tissue after treatment with sildenafil, Source: Weizmann Wander Wonder

Working with the lab, Clalit Health Services mined 20 years of records covering roughly five million members, including an observational cohort of 40,000 men diagnosed with cancer. Men who had filled sildenafil prescriptions before diagnosis had better overall survival than men who had not. Men who had taken both sildenafil and statins did better still. Erez's framing in the Weizmann release is that tumour biology is shaped by the patient's metabolic state and by whatever else they happen to be taking, and that oncologists should be "treating the whole patient – not just the cancer".

The study's own epidemiologist is careful about what that does and does not mean. Samah Hayek of the Clalit Research Institute told Healthline the analysis covered medication use before diagnosis only, never after, and that the results "do not mean cancer patients should start taking Viagra". She also noted the cohort was almost entirely male, because sildenafil is almost entirely prescribed to men, so the data say nothing about women. Mike Lattanzi, a medical oncologist at Texas Oncology who was not involved, told the same outlet that the mechanistic work is preclinical and would need human validation before it changes how anyone is treated.

Those caveats are correct and insufficient. The deeper problem with the Clalit analysis is not sample size or follow-up. It is the comparison group.

Men who fill Viagra prescriptions are not a random sample of men

A sildenafil prescription is a marker of several things at once: a man who sees a doctor, who is well enough and interested enough to want sex, who has insurance and disposable income, and who is therefore also more likely to have his cancer caught early. Compare that population against everyone else and you are measuring healthcare engagement at least as much as pharmacology.

This is not a hypothetical objection. It has already happened twice with this exact drug.

In 2021, Feixiong Cheng's group at the Cleveland Clinic ran a computational screen across 1,600 approved drugs, landed on sildenafil, and then mined insurance claims for 7.23 million people. The result, published in Nature Aging, was a hazard ratio of 0.31 for Alzheimer's disease: a 69 per cent reduction in risk. They backed it with iPSC-derived neurons showing increased neurite growth and reduced phospho-tau. Mechanism plus database, exactly the structure of the Weizmann paper.

Ten months later the NIA's Drug Repurposing for Effective Alzheimer's Medicines study tested the same hypothesis in Brain Communications and found nothing. The design difference is the entire story. Rather than comparing sildenafil users to everyone else, Madhav Thambisetty's team restricted to patients with pulmonary arterial hypertension and compared PDE5 inhibitor initiators against endothelin receptor antagonist initiators, an alternative treatment for the same condition, matching 2,888 pairs across 76 confounding variables. Same indication, same clinical population, same reason to be on a drug. The signal disappeared. The NIH announced the null result in October 2022.

The evidence has stayed unsettled since. A 2025 meta-analysis in Aging pooled five studies covering 885,380 patients and reported a hazard ratio of 0.47, still favouring sildenafil. Four years and roughly nine million patient records after the first claim, nobody can say whether it is real.

The melanoma episode ran the other way and ended more decisively. A 2014 prospective cohort in JAMA Internal Medicine found recent sildenafil use associated with an 84 per cent increase in melanoma risk. Alarming, mechanistically plausible, and wrong. Later work found no dose-response relationship, no elevated risk for longer-acting tadalafil or vardenafil, and a pattern of risk concentrated in early-stage disease and reduced in advanced disease. Parallel case-control studies in Danish and Kaiser Permanente data, covering more than 10,000 melanoma cases, found odds ratios of 1.22 and 0.95, attenuating further once markers of healthcare utilisation were adjusted for. A 2017 meta-analysis of 866,049 men put the relative risk at 1.11 and attributed it to detection bias and to sun exposure tracking socioeconomic status.

Sildenafil users get diagnosed with things. That is what the databases keep detecting.

The statin interaction compounds the concern rather than easing it. Men taking both sildenafil and a statin before a cancer diagnosis are, almost by construction, the most medically supervised subgroup available. The dose-response pattern the researchers report is worth something, but the melanoma literature is a reminder that dose-response can also track how much medicine a person is consuming overall.

What would actually settle it

The mechanism does not depend on any of this. NPC1-mediated cholesterol export is a lever, and if it is a real vulnerability in metastasising cells then it is worth pulling regardless of whether the specific drug that revealed it turns out to be useful. The lab result and the survival result should be graded separately.

For the clinical claim, the design that resolved the Alzheimer's question is available here too. Compare men on sildenafil against men on a different erectile dysfunction treatment, or against men with the same indication managed differently, rather than against the general population. If the effect survives an active comparator, it deserves a trial.

PDE5 inhibitors have already been through oncology once. A randomised, placebo-controlled phase II trial published in Clinical Cancer Research in 2015 found that tadalafil increased ex vivo T-cell expansion 2.4-fold in head and neck squamous cell carcinoma patients against 1.1-fold in controls, and reduced circulating myeloid-derived suppressor cells. That was eleven years ago, through a completely different mechanism, and combination trials with pembrolizumab are still running. Real biology, no approval.

There is a reason this pattern keeps recurring in the longevity literature specifically, and Erez's funders illustrate it: the work is supported in part by the Sagol Institute for Longevity Research. Geroscience runs on cheap generics with plausible mechanisms and large retrospective footprints, because that is the only way to get a signal on a slow disease without waiting thirty years. Metformin has been the field's flagship candidate for two decades on exactly this evidence base. The method is sound as a way of generating hypotheses. It is unreliable as a way of settling them, and the gap between those two functions is where most of the press coverage lands.

The Weizmann team found a cholesterol transporter that cancer cells appear to need and a drug that jams it. That is the finding. The 40,000 men are a reason to run the trial.