Repurposing clinical trial data for skin cancer aetiology

In this issue of the BJD, Passarelli and colleagues provide randomized, placebo-controlled, clinical trial data on the effects of aspirin and folic acid and the risk of basal cell carcinoma (BCC).¹ The main findings were a modest nonsignificant increased risk of BCC associated with any aspirin use and no association with folic acid. However, subgroup analyses suggested that aspirin use may lead to a reduced risk of BCC among individuals with a history of skin cancer and an increased risk of BCC among those with no history of skin cancer, although these associations were also not significant. Despite relying on a small sample size and reporting largely null associations, the analysis adds high-quality data on the relationship between aspirin use, folic acid supplementation and subsequent risk of BCC.

Increased BCC risk has been convincingly associated with skin cancer screening, use of photosensitizing medications, tanning bed use and ambient ultraviolet radiation exposure (e.g. occupational or leisure-time outdoors).² For nonsteroidal antiinflammatory drugs (NSAIDs), Passarelli and colleagues point out that many of the previous observational studies set in the general population have found either no risk or a slightly decreased risk associated with regular NSAID use.³ This is important to put in the context of study design, because people regularly taking NSAIDs may differ from the general population in their exposure to other BCC risk factors. For example, if the population under study regularly taking NSAIDS also refrain from spending time outdoors, this group may appear to have lower risk of BCC. This could lead observational studies to report risks of BCC associated with NSAID use lower than studies using an experimental design without this source of bias. In practice, predicting the direction of such biases may be quite difficult. Thus, randomized trials such as this one provide not only a window into the true BCC risks associated with aspirin and folic acid, but hints as to the possible directions (and sources) of bias in observational studies as well.

As the name ‘Aspirin/Folate Polyp Prevention Study’ implies, the current study was originally intended to examine the outcome of colon polyps, not skin cancer, and did not have individual information on sun exposure.¹ However, the investigators leveraged the randomized design to examine aspirin use, folate supplementation and BCC without the need for detailed lifetime history of ultraviolet radiation exposure. The Women’s Health Initiative, which includes a set of randomized placebo-controlled trials designed to examine heart disease, breast and colorectal cancer, and osteoporotic fractures, has also recently examined the association between medication use (menopausal hormone therapy), supplementation (vitamin D and calcium) and skin cancer risk.^4,5 Despite the frequency of BCC in the U.S. population, sample size in these studies remains a limitation, particularly for subgroup analysis. In spite of this, the current study represents an example of successfully repurposing an existing randomized trial to both address a new question and examine the magnitude and direction of possible biases reported in existing, highly powered observational studies.