Our study of the association between the use of photosensitizing agents and NMSC provides some evidence that commonly used photosensitizing medications may enhance risk of both SCC and BCC. For BCC, associations were strong for early-onset disease (≤50 years), and for SCC among those with a sun sensitive phenotype (e.g., tendency to sunburn).
While the literature is sparse, as in our study, a study from Denmark likewise found an association between use of tetracyclines and BCC (incidence rate ratio (IRR) = 1.3, 95% CI = 1.3–1.4) and additionally found a relation with SCC (IRR = 1.5, 95% CI = 1.4–1.7) (Kaae et al., 2010
) that we did not detect. The study based in Denmark involved linkage between cancer registry and national pharmacy data, which limited their ability to assess sun sensitivity or sun exposure as potentially confounding or modifying factors. One possible explanation for our finding of an association with BCC but not SCC is that tetracyclines are commonly used for acne treatment during the teenage years, a period when UV exposure has been related to an increased risk of BCC (Gallagher et al., 1995
) whereas chronic UV exposure appears to more strongly relate to risk of SCC (Armstrong and Kricker, 2001
). In our study, the majority of BCC cases and controls reported antimicrobial use for acne and other skin conditions, which raises the possibility that acne or another form of treatment for acne, such as ionizing radiation, may be confounding the association between tetracycline use and BCC. While acne is not a risk factor for BCC to our knowledge, there are reported associations with psoriasis and atopic dermatitis (Frentz and Olsen, 1999
; Olesen et al., 2005
). The possibility of confounding by ionizing radiation treatment used for treatment of acne is unlikely because its use was uncommon after 1960, and therefore would not have contributed much to the incidence of early-onset BCC. Further, adjustment for ionizing radiation therapy did not appreciably alter our results.
Another category we studied was photosensitizing CV medications. While prior studies have investigated this topic, the results have varied for the different medications. A recently published U.S. based case-control study found longer durations of use of thiazide diuretics to be associated with increased odds ratios for development of carcinoma of the lip (Friedman et al., 2012
). A case-control study in Denmark found that users of the diuretic therapy combination of amiloride and hydrochlorothiazide had an increased risk of SCC with an IRR of 1.79 (95% CI = 1.45–2.21), but did not find statistically significant results for other diuretics or BCC (Jensen et al., 2008
). When we restricted our analysis to thiazide diuretics (i.e. hydrochlorothiazide or hydrochlorothiazide combination medications), we found a similar modest association with SCC. A population-based follow-up study based in The Netherlands found an association between loop diuretics (e.g. furosemide, bumetanide) and BCC with a hazard ratio of 1.07 (95% CI = 1.01–1.13) (Ruiter et al., 2010
). Another population-based cohort study from Denmark found a significant association between methyldopa and BCC, and an association between furosemide and SCC (Kaae et al., 2010
). These findings, along with our own, raise the possibility of an enhanced NMSC risk with certain CV medications, specifically diuretics and namely thiazide diuretics that will need further confirmation.
A major limitation of our study is that we relied on self-reported medication use. The phrasing of questions differed between the early and late phases of the study that we were able to take into account by including a random effect term for study period. We did not detect any statistically significant differences between the odds ratios by study phase. For our main finding on BCC and antibiotic use for example, the odds ratio in the early phase was 1.96 (95% CI = 1.11–3.45) and for the later phase was 1.83 (95% CI = 1.01–3.32). However, it is possible that we missed photosensitizing drugs that were used but not reported by subjects, especially in the early phase when the question relied on the subject supplying the name of the medication rather than choosing from a predetermined list. This misclassification was likely non-differential and thus could have biased our results towards the null as differential misclassification would have resulted in an increased risk associated with other similar medications where we observed a reduced risk (e.g., with NSAIDS)(Torti et al 2011
). Another bias inherent in this study design is selection bias from non-participation, although it is unlikely that those who chose not to participate were different in meaningful ways (e.g., in terms of photosensitizing drug use, and/or NMSC risk) for the purposes of this study. Controls were matched to cases on age and sex to represent the population from which the cases arose. Overall 97% of cases 65 years of age or older were enrolled in Medicare, and 96% of cases <65 years held a valid driver’s license. Additionally, despite our detailed personal interview, we consider that there could be residual confounding from unmeasured factors contributing to the modest odds ratios we observed. We note however that given the magnitude of the skin cancer problem worldwide, use of the photosensitizing agents causing even small increases in the relative risks of basal cell or squamous cell carcinomas could translate into an appreciable number of new cases of these malignancies.
One advantage to our approach is that it allowed us to evaluate several types of medications in a large group of BCC and SCC cases and controls. The study was population-based and we had an active surveillance system in place to ensure that we were informed of new cases during the study period. To our knowledge, there is limited prior research on these medications in relation to NMSC in the United States population. Additionally, the personal interview allowed for collection of demographic information as well as baseline sun sensitivity (i.e. propensity to tan or burn), cumulative lifetime sun exposure, indication for use, and prior radiation treatment, among others. As expected, we observed stronger associations among those with a sun sensitive phenotype for SCC and to a lesser extent multiple BCCs, which is consistent with a finding reported from the Rotterdam cohort study for BCC (Ruiter et al., 2010
). As to why our findings were stronger for early-onset BCC is unclear, but may relate to the fact that antibiotics used to treat acne are taken at a young age. Further studies will need to evaluate this possibility.
In conclusion, we found evidence that use of certain photosensitizing medications may enhance risks of SCC and BCC in the United States population, particularly among individuals with a sun sensitive phenotype, and for BCC, those with a young age at onset.