Sunscreen & Skin Cancer: What the Study Data Actually Shows

A viral claim cites a 292% higher melanoma risk among sunscreen users from a UK Biobank study of 470,000 people. We went to the published paper, extracted the raw data, and examined the methodology. Here's what the numbers say - and what they can't say.

Primary source: Jeremian et al. (2023), Cancer Epidemiology, Biomarkers & Prevention, 32(11), 1599–1607 · PubMed Central full text

First: What Was This Study Actually About?

The study's title is: "Gene–Environment Analyses in a UK Biobank Skin Cancer Cohort Identifies Important SNPs in DNA Repair Genes That May Help Prognosticate Disease Risk."

It was a genetics paper. The researchers were looking for DNA repair gene variants (particularly in the FANCA gene) that interact with environmental factors to predict who is more likely to develop skin cancer. Sunscreen use was one of 11 variables they examined - not the study's focus.

Nicolas Hulscher is not an author on this paper. The actual authors are Richie Jeremian, Pingxing Xie, Misha Fotovati, Philippe Lefrançois, and Ivan V. Litvinov - all from McGill University, Montreal.

The Raw Numbers: What the Data Shows

These numbers come directly from Table 1 and Table S2 of the published paper. The relative risk (RR) values cited in the viral post do appear in the data:

The gradient is real. People who reported "always" using sun protection had a crude melanoma rate roughly 4x higher than those who "never/rarely" used it. The reported relative risk values across all cancer types:

So the 292% number is not fabricated. But here is where the methodology becomes critical - because what these numbers mean depends entirely on how they were calculated.

The Three Methodological Problems

These are not opinions or editorial judgments. They are structural limitations documented in the paper's own methods section.

Problem 1: The analysis is univariate - no confounders controlled

This is the critical issue. The researchers ran multinomial logistic regression on each variable separately. The sunscreen RR of 3.92 is not adjusted for:

• Time spent outdoors in sunlight
• Skin colour / skin type
• Hair colour
• History of sunburn
• Sunlamp / solarium use
• Geographic UV exposure
• Age, sex, or any other demographic

In epidemiology, an unadjusted relative risk from an observational study cannot support a causal claim. This is not a controversial position - it is the foundational rule of observational epidemiology, taught in every MPH programme including the one Hulscher attended.

To illustrate why this matters: the study also found that people with lighter skin and red/blonde hair had higher skin cancer rates. If you ran the same univariate logic, you could claim that being blonde causes cancer. The reason you can't is because blondness is a marker for fair skin and UV sensitivity - it's correlated with cancer but doesn't cause it. The same logic applies to sunscreen use.

Problem 2: The question isn't about sunscreen - it's about sun protection generally

The variable measures "use of sun/UV protection" - which includes hats, clothing, shade-seeking, and sunscreen. The viral claim specifically says "sunscreen use" causes cancer, but the data doesn't isolate sunscreen from other sun-protective behaviours. Someone who always wears a hat outdoors and someone who always applies SPF 50 are in the same category.

Problem 3: No temporal ordering - reverse causation is uncontrolled

The UK Biobank questionnaire was administered at a single point in time. A person diagnosed with skin cancer in 2015 who then started using sunscreen religiously appears in the data as both "frequent sun protection user" and "skin cancer patient." The cancer came first - the protection was a response to it, not a cause of it.

The study makes no attempt to exclude people diagnosed before enrollment or to establish whether sun protection use preceded or followed diagnosis.

What the Study's Own Authors Wrote

The authors were aware of the paradox in their data. This is what they wrote in the published paper - not in a press release, not in a tweet, but in the peer-reviewed text itself:

"This paradoxical finding was the increasing risk of skin cancers with increased sunscreen use, which we posit can be explained by greater exposure to UV light and/or a lack of reapplication of sunscreen throughout the day, or due to increased use of sun protection following skin cancer diagnosis." - Jeremian et al., Cancer Epidemiology, Biomarkers & Prevention, 2023

Their conclusion about the finding:

"These findings demonstrate the importance of adequate and frequent sunscreen use and minimisation of exposure to UV light." - Jeremian et al., same paper

The researchers saw the correlation, recognised it as a known confounding pattern, and explicitly concluded that the data supports more sunscreen use - not less.

What Would Actually Prove Sunscreen Causes Cancer?

To make a causal claim, you need one of two things:

An RCT on sunscreen and melanoma does exist

The Nambour trial (Green et al., 2011) is the only randomised controlled trial ever conducted on sunscreen and melanoma. In 1992, 1,621 Australians were randomly assigned to either daily sunscreen use (free, unlimited SPF 16 supplied) or discretionary use (their own choice).

After 15 years of follow-up:

Because participants were randomised, confounders were balanced between groups. The result: daily sunscreen users developed 50% fewer melanomas and 73% fewer invasive melanomas. This is the gold-standard evidence that the UK Biobank observational correlation cannot override.

Who Made This Claim?