Statin use and non-melanoma skin cancer risk: a meta-analysis of randomized controlled trials and observational studies

Background Existing evidence of the association between statin use and non-melanoma skin cancer (NMSC) risk has been inconsistent. Objective To maximize statistical power to synthesize prospective evidence on this relationship. Materials and Methods PubMed, EMBASE, Web of Science, Cochrane Central Register of Controlled Trials, and ClinicalTrial.gov were systematically searched up to December 11, 2016. A random-effects meta-analysis was conducted to calculate summary estimates. Results Our meta-analysis of 14 randomized controlled trials (RCTs) including 63,157 subjects showed no significant association between statin use and NMSC risk (RR = 1.09, 95%CI = 0.85–1.39). However, meta-analysis of four observational studies including 1,528,215 participants showed significantly increased risk of NMSC among statin users compared to non-users (RR = 1.11, 95%CI = 1.02–1.22). Furthermore, ever using lipophilic statins (RR = 1.14, 95%CI = 1.04–1.24) or lower-potency statins (RR = 1.14, 95%CI = 1.03–1.26), as well as usage of any statin longer than one year (RR = 1.14, 95%CI = 1.09–1.18) were significantly associated with increased NMSC risk based on observational studies. Conclusions Evidence from observational studies supported an association between statin use and increased NMSC risk. This finding should be interpreted with caution due to modest number of included studies, possible between-study heterogeneity and inherent limitations of observational studies.


INTRODUCTION
Non-melanoma skin cancer (NMSC), including basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) of the skin, is the most common cancer in the United States [1,2]. Statins (HMG-CoA reductase inhibitors), potent cholesterol-lowering medications, have proven effective in the prevention of adverse cardiovascular events. Moreover, some studies also indicated a potential chemopreventive action of statins against overall cancer risk [3], as well as risks of cancers such as gastric [4], colorectal [5], breast [6], and prostate cancers [7]. However, evidence for risk of NMSC has been conflicting. Compared to controls, more NMSC cases were observed among the treatment group in some large randomized controlled trials (RCTs) of statins such as the Scandinavian Simvastatin Survival Study (4S) [8] and the Heart Protection Study (HPS) [9], but not others [10][11][12]. One previous meta-analysis found no significant association between statins and skin cancer [13]. A few prospective observational studies have specifically examined this hypothesis. Recently, Wang et al. [14] found that statins were associated with increased NMSC risk in the Women's Health Initiative (WHI) cohort. A nested

Meta-Analysis
Oncotarget 75412 www.impactjournals.com/oncotarget case-control study using Denmark national registries data also identified a significant association between statin use and slightly increased risk of BCC but not SCC [15]. Therefore, to maximize statistical power to synthesize prospective evidence on the relationship between statin use and NMSC risk, we conducted this meta-analysis of all eligible prospective observational studies and randomized controlled trials.

Study selection and study characteristics
Among the 6,420 citations retrieved from the electronic databases, a total of seven reports that covered 14 RCTs including 63,157 subjects and 1,211 NMSC events [8-12, 16, 17], as well as four observational studies including 1,528,215 participants and 55,793 NMSC cases [14,15,18,19] met our eligibility criteria and were included in this meta-analysis. The mean duration of included RCTs varied from 2 to 5.4 years while the mean/ median follow-up time of included cohort studies varied from 3.5 to 10.5 years. A diagram of the study selection process is shown in Figure 1; basic characteristics of all included studies are shown in Table 1, and quality assessment results are shown in Supplementary Table 1.

Subgroup analysis of statin use and NMSC risk
The results of subgroup analyses are shown in Table 2. Between-group heterogeneity was not statistically significant in most subgroups (P > 0.05). None of the subgroup analyses detected a significant association based on RCTs (P > 0.05). However, subgroup analyses of observational studies revealed significantly increased risks of NMSC among ever-users of lipophilic statins (RR = 1.14, 95% CI = 1.04-1.24) or statins with lower potency (RR = 1.14, 95% CI = 1.03-1.26), compared to neverusers. Additionally, users of statins for more than one year had significantly higher NMSC risk than never-users (RR = 1.14, 95% CI = 1.09-1.18).

DISCUSSION
Our meta-analysis of four prospective observational studies involving 1,528,215 participants and 55,793 NMSC cases suggested a significant and positive association between statins and NMSC risk. However, our meta-analysis of 14 RCTs with 63,157 participants and 1,211 NMSC events did not observe such an   Oncotarget 75415 www.impactjournals.com/oncotarget may have some advantage of examining a long-term drug-safety issue.
One previous meta-analysis evaluated statins and all-cancer risk, among which only four RCTs were related to NMSC, and their pooled results indicated a significant association between statins and increased NMSC risk [20]. Another meta-analysis conducted by Li et al. [13] found no significant association between statins and risk of either melanoma or NMSC. However, that study also included skin cancer cases of unspecified type in the meta-analysis of NMSC. To our knowledge, ours is the first metaanalysis specifically examining the relationship between the risk of NMSC and statins, as well as statins of different potency, category, and duration of use.
Melanoma, SCC, and BCC are three major types of skin cancer. The underlying mechanisms of melanoma and non-melanoma skin cancers (SCC and BCC) are not the same. In terms of melanoma, which is the most serious type of skin cancer, most randomized controlled trials and observational studies published so far showed nonsignificant association between statin use and melanoma risk [13]. Recently, sixteen years of follow-up of the Longterm Intervention with Pravastatin in Ischemic Disease (LIPID) trial also reported no excess cases of melanoma in the intervention group, compared to the control group [21].
We still lack substantial understanding of the biological mechanisms by which statins affect cancer risk, and whether specific properties of statins make any difference in this process. Some studies have indicated that the immunomodulatory effects of statins may be responsible for its potential carcinogenic actions. A statininduced increase in CD4+CD25+ regulatory T cells (Tregs) may impair the host antitumor immune response through suppressing the tumor specific effector T cell response, which may lead to an increased risk of cancer [22]. Notably, immunosuppression is an emerging risk factor for NMSC [23], and may explain the high incidence of NMSC in the elderly populations who are susceptible to immunosenescence. In addition, immunosuppressant therapies constitute a major risk factor of developing post-transplantation skin cancers among organ transplant recipients [24]. In kidney transplant recipients, high Tregs levels within the peripheral circulation were associated with new SCC development [25]. On the contrary, some research has also suggested that statins exert a protective action against skin cancers by inhibiting the mevalonate and hedgehog signaling pathways, disrupting cancer cell growth and inducing keratinocyte apoptosis [26]. As statins have also been associated with reduced risks of several other types of cancers [7,20], it is possible that statins may play roles in affecting different cancer risks through various biological pathways.
Our findings should be interpreted with caution due to the modest number of studies we were able to include, possible between-study heterogeneity, as well as the residual confounding inherent in observational studies. Due to the unavailability of original data, we could not conduct meta-analysis separately for BCC and SCC. We also could not estimate dose-response effects as available data were too heterogeneous to pool.
In conclusion, our meta-analysis of observational studies supported an association between statin use and increased NMSC risk. In future, ongoing post-marketing surveillance and long-term follow-up studies of RCTs could provide ongoing monitoring of new NMSC cases that may be related to the use of statins. More prospective studies with large sample sizes could be conducted to further evaluate the association between statins and NMSC risk. In addition, basic research on cholesterol metabolism and tumor formation is needed to explore underlying mechanisms.

MATERIALS AND METHODS
This meta-analysis was conducted according to the Preferred Reporting Items for Systematic reviews and Meta-analyses (PRISMA) guidelines [27].

Systematic search
We conducted a comprehensive and systematical literature search in PubMed, EMBASE, Web of science, Cochrane Central Register of Controlled Trials (CENTRAL), and ClinicalTrials.gov registry to identify all prospective evidence (all from inception to December 11 th , 2016 with no language limitation). References list from all relevant review articles, metaanalysis, and the identified articles were manually checked. The following terms were used for literature search: ((hydroxymethylglutaryl coenzyme A reductase inhibitor) OR (hydroxymethylglutaryl-coa reductase inhibitors) OR (HMG-CoA reductase inhibitors) OR

Eligibility criteria
Both randomized controlled trials and prospective observational studies (cohort study and nested casecontrol study) were considered. Included studies should meet the following criteria: (1) Published and unpublished RCTs of statins that reported NMSC cases as adverse events (length of trial ≥ 24 weeks); (2) Cohort or nested case-control studies that examined the association between statin use and NMSC incidence; (3) All study participants were adults aged ≥ 18 years. NMSC cases referred only to Oncotarget 75416 www.impactjournals.com/oncotarget SCC and BCC in this meta-analysis. Retrospective casecontrol studies were excluded as they were subjected to various biases. Also, to minimize bias, we used safety data from only the original trials rather than prolonged posttrial follow-up (if available), as the majority of all subjects received open-label lipid-lowering treatment during the latter. Studies did not provide sufficient information or failed to meet eligibility criteria were also excluded.

Data extraction
Two reviewers (K.Y. and A.M.) independently performed the study selection, data extraction, and quality assessment. Numbers of NMSC cases were extracted from randomized trials and adjusted risk ratios were extracted from observational studies for data synthesis. As two included RCTs did not report NMSC events in the manuscript (JUPITER and AURORA) [11,16], the case numbers were extracted from the "Serious Adverse Events" section on the ClinicalTrials.gov registry website. Study characteristics including first author's name, publication/report year, trial name, study design, study location, sample size, study length, type of statins, exposure and outcome assessment, as well as population characteristics including age, sex and race of the study population in each included study were also extracted.

Quality assessment
The quality of randomized trials was evaluated by Cochrane Risk of Bias (RoB) Tool [28]. We used Cochrane RoB tool to assess potential study bias in 6 domains, including random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), and selective reporting (reporting bias). The quality of included observational studies was assessed by Newcastle-Ottawa quality assessment scale (NOS) [29]. Each cohort/case-control study was evaluated on three broad perspectives: the selection of the study groups, the comparability of the groups, and the ascertainment of the exposure and the outcome of interest.