Discussion
The results of this population-based cohort study suggest that, compared with sulfonylureas, DPP-4 inhibitors were associated with a reduced risk of melanoma, with evidence of a duration-response relationship. In contrast, these drugs were not associated with non-melanoma skin cancer. Overall, these results remained consistent in several sensitivity analyses.
To our knowledge, no clinical study has examined the use of DPP-4 inhibitors and skin cancer as a stand-alone outcome. One meta-analysis of 72 RCTs including 35 768 patients on DPP-4 inhibitor and 33 319 on comparison drugs/placebo reported a numerically lower risk of malignant melanoma (relative risk 0.87, 95% CI 0.48 to 1.59) but a numerically higher risk of skin cancer overall (relative risk 1.79, 95% CI 0.86 to 3.71).15 A larger meta-analysis of 115 RCTs with 65 740 patients in the DPP-4 inhibitor group and 56 221 in the control group reported a lower overall risk of cancer (OR 0.91, 95% CI 0.85 to 0.97), all skin cancers (OR 0.85, 95% CIs 0.72 to 0.99), and malignant skin cancer (OR 0.86, 95% CIs 0.73 to 1.00), although the type of skin cancer was not specified.14 In another meta-analysis of 157 RCTs with 66 825 patients on DPP-4 inhibitor treatment and 61 524 patients in the control group, DPP-4 inhibitor use was not associated with melanoma (OR 1.13, 95% CI 0.73 to 1.00).16 However, all meta-analyses included short-term studies (<52 weeks), which may be inadequate to examine cancer risk.14 15 Our study found that DPP-4 inhibitor use was associated with a 23% lower risk of melanoma skin cancer, the risk reduction occurring after 2 years of use. We did not find any consistent association of DPP-4 inhibitor use with non-melanoma skin cancer.
The DPP-4 enzyme has a complex role in the melanocyte malignant transformation.11 36 Potential benefit of DPP-4 inhibitors in melanoma can be explained by biological findings in murine models where sitagliptin delayed melanoma development, tumor growth, and metastasis by trafficking antitumor CXCR3+lymphocytes to the tumor site by locally elevating the chemokine CXCL10, a substrate of the DPP-4 enzyme.12 Sitagliptin also improved tumorous response to checkpoint blockade.12 Given that cutaneous lymphocyte trafficking is a clinically well-characterized effect of DPP-4 inhibitor use mediating skin reactions such as bullous pemphigoid,37 38 this could be a potential mechanism behind a reduced melanoma incidence with DPP-4 inhibitors.
Interestingly, we did not find a change in the risk of non-melanoma skin cancer with DPP-4 inhibitor use. Indeed, DPP-4 enzyme activity in non-melanoma skin cancer is variable, with high activity noted in basal cell carcinomas while both high and low activity in squamous cell carcinomas.3 Even the role of CXCR3 in non-melanoma skin cancer is unclear:39 imiquimod, an immunomodulator effective in non-melanoma skin cancer, recruits CXCR3+lymphocytes.40 In contrast, CXCR3 gene deletion lowers the incidence of skin tumors in mice, and CXCR3+lymphocyte recruitment promotes keratinocyte proliferation.41 These variable roles of DPP-4 and CXCR3 on keratinocytes might explain the relatively null association between DPP-4 inhibitors and non-melanoma skin cancer.
Overall, our study has several strengths. First, we used the CPRD as our data source, a database that is largely representative of the UK population. Using this database also allowed us to examine well-validated outcome definitions. Second, we accrued one million person-years of follow-up in both cohorts, with a potential follow-up of 13 years for each cohort, making our study well powered to determine whether DPP-4 inhibitors are associated with skin cancer. Finally, we used a new-user, active comparator design which likely minimized confounding and detection bias at the design stage, as well as bias from the inclusion of prevalent users.42
Our study also has some limitations. First, exposure misclassification is possible because the CPRD does not directly record prescriptions written by specialists. However, this is unlikely to be an important source of misclassification since general practitioners almost entirely manage type 2 diabetes in the UK.43 Second, it was not possible to assess sun exposure at the patient level, an important skin cancer risk factor. Reassuringly, sun exposure is not considered when prescribing an antihyperglycemic drug versus another. Nonetheless, we included several proxies for variations in sun exposure in the propensity score models, including various photodermatoses, calendar year and region. Third, there was substantial missingness in the variable ethnicity, an important risk factor for skin cancer. However, given it was not differentially distributed between exposure groups, it is unlikely to have been a major source of residual confounding. Fourth, the possibility that undiagnosed skin cancer was present in the study cohort before commencing the study drugs cannot be ruled out. However, to minimize this possibility we applied a lag period, counting cases at 1 (primary) and 3 or 5 (sensitivity) years after the beginning of follow-up, with similar findings. Finally, residual confounding is a possibility given the observational nature of this study. However, given that clinical studies on the DPP-4 inhibitors-skin cancer association were published after the study period, any channelling related to this outcome is unlikely.
In summary, DPP-4 inhibitor use was associated with a reduced risk of melanoma but not with the incidence of non-melanoma skin cancer. Given the high mortality associated with melanoma, and dearth of preventive strategies related to this malignancy, more research should be conducted to confirm our findings.