The Evidence Doesn’t Say That: On the New E-Cigarette Carcinogenicity Review

The review, led by Stewart and colleagues, draws on case reports, biomarker studies, rodent bioassays, and mechanistic evidence to conclude that nicotine-based e-cigarettes cause an “indeterminate burden” of oral and lung cancer. What it openly concedes, however, is that no epidemiological study has yet compared tumour incidence in e-cigarette users with nonusers. The paper builds its case from indirect evidence—and that matters for how far the conclusion can responsibly reach.

The problem appears immediately.

The central problem surfaces almost immediately. The authors openly concede that there are no epidemiological studies contrasting tumour incidence in e-cigarette users and nonusers. They then proceed anyway, arguing that carcinogenicity can be inferred from case reports, biomarkers, rodent bioassays, and mechanistic studies. That is the crux of the paper’s logic—and the crux of its overreach.

A claim about carcinogenicity in humans should rest, at some meaningful point, on evidence in humans showing disease in humans. This paper acknowledges the absence of that evidence and substitutes a chain of inference in its place.

E-cigarettes have been on the market for over a decade, used by tens of millions of adults, yet long-term cancer epidemiology remains strikingly thin. To date, the available longitudinal human evidence has not demonstrated increased incident oral or lung cancer among exclusive never-smoker e-cigarette users compared with never-users. A handful of cross-sectional analyses have reported associations with prevalent self-reported cancer, but those designs cannot establish temporality and cannot show that vaping caused the disease.

A 2025 systematic review put the position plainly: no significant incident or prevalent risk of lung cancer or other cancers was found in the never-smoker current vaper population. The same review identified only two longitudinal human studies and noted that follow-up remains far too short for cancers with long latency. The authors of the Carcinogenesis review treat this absence as a timing issue rather than what it is: the central evidentiary deficiency of the entire case.

Most vapers were smokers first.

There is a further problem the review never seriously confronts: the overwhelming majority of adult vapers are former or current smokers. Exclusive never-smoker vapers are a small minority of the total vaping population, and they are precisely the group whose data would actually answer the carcinogenicity question.

This matters enormously. Any biomarker signal, any DNA damage finding, any epigenetic change, and certainly any case report in a vaper with prior smoking history is confounded by tobacco exposure. Combustible tobacco is a confirmed Group 1 carcinogen whose biological effects persist for decades after cessation. You cannot cleanly attribute oxidative stress, DNA adducts, or oral lesions to vaping in someone who spent years inhaling cigarette smoke.

The Stewart paper’s own framing is revealing here. The authors explicitly say they want to assess e-cigarettes “in their own right” and exclude dual-use findings. Yet they then lean on biomarker studies and case reports drawn from populations where prior smoking is common and often poorly characterised. The 51-year-old lung cancer case has a 10-pack-year smoking history built in. The paper itself states that causality cannot be established in that case. A determination of “likely carcinogenic to humans” cannot rest on a body of evidence that has not, in most cases, cleanly separated the carcinogen we know causes cancer from the one being investigated.

Hazard is not risk.

The review tries to shield this move by describing itself as a “qualitative risk assessment.” It says this approach goes beyond hazard identification, yet also admits that it “does not address quantitative differences” such as how much aerosol people inhale or how patterns of use alter dose. That is not a small footnote. In toxicology, dose matters. The presence of a carcinogen, or a biomarker associated with carcinogenic processes, is not the same as exposure at a level that causes cancer.

“Objectively definitive”?

The paper also claims unusual certainty about its own method. Because it restricted itself to reviews and studies published since 2017, it describes the assessment as “objective” and even “objectively definitive.” That is an extraordinary phrase for a field where the authors themselves admit that cancer epidemiology is not yet available. No literature review becomes definitive merely by declaring itself so.

Case reports are not causation.

The case reports section relies on two oral cancers in older men with long vaping histories, one oral squamous cell carcinoma in a 19-year-old, and one lung cancer in a 51-year-old man with an 11-year vaping history and a 10-pack-year smoking history. The paper itself states that in the lung cancer case, causality cannot be established. It then ends the section by calling for the very thing it does not have: longitudinal epidemiological studies linking e-cigarette use to cancer incidence.

That is effectively an admission that the current evidence cannot answer the question the paper claims to answer.

Biomarkers are surrogates, not diagnoses.

Evidence of oxidative stress, DNA adducts, epigenetic change, and inflammation is real and should not be brushed aside. But biomarkers are surrogate indicators, not cancer diagnoses. A signal of oxidative stress is not oral cancer. DNA adducts are not lung adenocarcinoma. Plausibility is not proof, and the cited literature is in several places described as limited or based on small studies.

One mouse study, doing too much work.

The animal evidence does even more heavy lifting than the human evidence. The review’s overall assessment states that its conclusion is primarily based on carcinogenicity of e-cigarette aerosol in mice together with mechanistic findings. One rodent study, however provocative, should not bear the weight of a conclusion this sweeping.

Why the overreach matters.

None of this means vaping is harmless. E-cigarette aerosols expose users to potentially harmful constituents, long-term surveillance is urgently needed, and adolescents and never-smokers should not start. But that is a different claim from saying the current evidence already justifies the conclusion that nicotine e-cigarettes are likely carcinogenic to humans.

The overreach matters. Tobacco use causes more than 7 million deaths each year worldwide, with roughly 1.6 million more deaths from second-hand smoke exposure. If smokers come to believe vaping is essentially as carcinogenic as smoking, they have less incentive to switch—and a premature verdict amplified through headlines can entrench combustible tobacco use.

If the evidence supports concern, say concern. If it supports surveillance, call for surveillance. But if direct epidemiological evidence is absent, case reports are sparse, biomarkers are indirect, and animal evidence carries outsized influence, then the honest conclusion is not certainty. It is uncertainty.

This review presents an inference as if it were a determination, and treats plausibility as though it were demonstration. In doing so, it says more than the evidence—at least for now—can truthfully bear.

Other independent experts have also criticised the paper’s methods, evidentiary standards, and strength of inference. Their reactions via the Science Media Centre are worth reading alongside the paper itself.

Another retracted review.

It is also worth noting that the Stewart Carcinogenesis paper is not the only recent review to make strong claims about e-cigarette carcinogenicity on a thin evidentiary base. A systematic review by Guedes de Oliveira and colleagues, published in the Journal of Cancer Policy in September 2025 (doi:10.1016/j.jcpo.2025.100615), which claimed to present evidence on vaping e-cigarettes as a risk factor for cancer, has since been retracted. The retraction underscores a recurring problem in this field: reviews that reach beyond the evidence they actually present. When papers asserting carcinogenic risk are subsequently withdrawn, it should prompt greater scrutiny of other reviews making similarly strong claims on similarly limited foundations.

Yusuff Adebayo Adebisi

Pharmacist and epidemiologist advancing equity-driven, policy-relevant public health research. PhD Researcher at the University of Glasgow. Director of Research at Global Health Focus.

Learn more →