Use of Fibrates and Cancer Risk: A Systematic Review and Meta-Analysis of 17 Long-Term Randomized Placebo-Controlled Trials

Background Fibrates comprise a class of well-established antilipidemic agents that significantly reduce cardiovascular events. Given the concerns of cancer with fibrate therapy, we undertook a systematic review and meta-analysis to investigate the effects of fibrates on cancer outcomes. Methods We systematically searched Medline, Scopus, SCI Expanded, and the Cochrane Library for studies published up to 2012. We included randomized controlled trials (RCTs) that evaluated a fibrate therapy compared with placebo, had a minimum duration of two years, and reported data on the incidence of and/or deaths from cancer during the trial. Reviews of each study were performed and the relative data were abstracted. Pooled relative risk estimates (RR) and 95% confidence intervals (CIs) were calculated using the inverse variance weighted approach. Subgroup, sensitivity and meta-regression analyses were also conducted. Results Seventeen RCTs, involving 44,929 participants with an average follow-up of 5.2 years, contributed to the analysis. The degree of variability between trials was consistent with what would be expected to occur by chance alone. The quantitative synthesis of data retrieved from the RCTs was not indicative of a fibrate effect on cancer incidence (780 [fibrate] vs 814 [control]; RR = 1.02, 95% CI 0.92–1.12) or cancer death (385 [fibrate] vs 377 [control]; RR = 1.06, 95% CI: 0.92–1.22). When the analysis was restricted to major RCTs, the results did not substantially change. Similarly, we found no evidence of differential effects by length of follow-up or type of fibrate. Insignificant results were also obtained for the role of fibrates in cancers of the respiratory tract, breast, colon, gastrointestinal tract, prostate, genitourinary tract, or in melanoma. Conclusion Our findings demonstrate that fibrates have a neutral effect on cancer outcomes. However, it is important to continue monitoring their long-term safety profiles.


Introduction
Fibrates comprise an important class of therapeutic agents for the management of dyslipidemia. They are agonists of the peroxisome proliferator-activated receptors alpha (PPAR-a), which are mainly expressed in liver, heart and skeletal muscle [1]. Fibrates have been shown to stimulate the expression of genes involved in fatty acid and lipoprotein metabolism, resulting in a shift from hepatic fat synthesis to fat oxidation [2]. This leads to a substantial reduction in serum triglycerides and an increase in high-density lipoprotein cholesterol levels. Though cardiovascular protection using antilipidemic agents has largely been dominated by the 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins) since the 1990s, fibrates have been particularly useful in patients with primary hypertriglyceridemia, mixed hyperlipidemia, and in patients with type 2 diabetes and dyslipidemia characterized by high triglyceride and low high-density lipoprotein cholesterol concentrations.
However, in the current era of intensive lipid-lowering therapies to reduce the risk of cardiovascular disease, there is still debate regarding the potential relationship between fibrates use and cancer. Concerns about a possible increase in cancer-related deaths arose initially due to fibrates' contribution to the total mortality observed during the ''in-trial'' period of the World Health Organization (WHO) cooperative study [3], although subsequent extended follow-up showed a smaller difference in incidence of and death rates from cancer between treatment and control arms [4,5]. On the other hand, a review of rodent carcinogenicity tests reported that lipid-lowering drugs, including fibrates, initiate or promote cancer in rats and mice [6]. However, in most of the reviewed studies the doses used were substantially higher than the recommended doses for humans and the employed bioassays were criticised for being inadequate to predict carcinogenicity in humans [7].
Given that the use of fibrates has steadily increased during the past decade [8], more knowledge is needed on the relationship between these medications and cancer. To address this issue, we conducted a systematic review and meta-analysis of randomized placebo-controlled trials published in the peer-reviewed literature.

Search Strategy
To identify the studies of interest, we systematically searched the following databases: (i) Medline, (ii) Scopus, and (iii) Science Citation Index Expanded, from the date of inception of each database to January 2012. Search terms included: ''fibrate'' or ''fibric acid'' or ''fenofibrate'' or ''bezafibrate'' or ''ciprofibrate'' or ''clofibrate'' or ''gemfibrozil''. The search was limited to randomized controlled trials (RCTs), human subjects. The Cochrane Central Register of Controlled Trials was also reviewed. The title and abstract of studies identified in the computerized search were scanned to exclude any that were clearly irrelevant. The full text of the remaining articles was read to determine whether it contained information on the topic of interest. The reference lists of the articles were reviewed to identify citations to other studies of the same topic. No language restrictions were imposed.

Inclusion and Exclusion Criteria
The studies considered in this meta-analysis were RCTs that evaluated exposure to fibrates and cancer risk. They were considered eligible if they had evaluated a fibrate therapy compared with placebo, they had a minimum duration of two years, and reported data on the incidence of and/or deaths from cancer during the trial. We excluded trials that evaluated multiinterventional therapies where the effect of the fibrate could not be separated out.
We did not assess the methodological quality of the primary studies as quality assessment in meta-analysis is controversial and results can be highly misleading [9,10]. Instead, we performed subgroup and sensitivity analyses according to study characteristics.

Data Extraction
Two reviewers (SB, GN) abstracted the data independently. The following information was collected from each study: (i) publication data: first author's last name, year of publication and geographical location of the study; (ii) study design; (iii) number of participants; (iv) population characteristics; and (v) interventions' parameters including type of drug, dose and duration. Study-level risk ratios and their 95% confidence intervals (CIs) were estimated by reconstructing contingency tables based on the number of patients randomly assigned and the number of patients who had experienced cancer events (intention-to-treat analysis). When no cancer events occurred in either or both arms, a continuity correction of 0.5 was added to each cell of the respective contingency table. Non-melanoma skin cancers were not included in the analysis because they were neither recorded nor routinely reported in the primary studies. Differences in data extraction were resolved by consensus, referring back to the original article.

Quantitative Data Synthesis
We used the inverse variance weighted approach to calculate summary effect-estimates. Outcome reporting bias was evaluated using the Begg-Mazumdar adjusted rank correlation test [11] and the Egger regression asymmetry test [12]. To evaluate whether the results of the studies were homogeneous, we used the Cochran's Q test [13]. We also calculated the quantity I 2 [14,15] that describes the percentage variation across studies that is due to heterogeneity rather than chance. We regarded an I 2 value less than 40% as indicative of ''not important heterogeneity'' and a value higher than 75% as indicative of ''considerable heterogeneity'' [16].
Subgroup analyses by fibrate type were also performed to investigate potentially different effects on risk. We, then, conducted a sensitivity analysis by restricting the meta-analysis: (i) to trials with a minimum duration of 5 years, (ii) to trials that enrolled at least 1,000 subjects, and (iii) to major trials that fulfilled both previous criteria. Last, we performed site-specific analyses to evaluate the association between fibrate use and type of cancer diagnosis (cancer subtypes: respiratory, breast, genitourinary, prostate, gastrointestinal, colorectal and melanoma). In those analyses, only RCTs reporting at least one site-specific new cancer diagnosis were included.
We also conducted a meta-regression analysis [17] to investigate the impact of certain study characteristics on the study estimates of relative risk. We first converted all risk ratios by logarithmic transformation to achieve more symmetrical distributions. The natural logarithm of the risk ratio was the dependent variable, and (i) the mean age of participants at enrollment, and (ii) the mean duration of follow-up, were entered as covariates. This analysis was an indirect way to deal with aspects such as the possibility of effect modification by age, and to examine for increasing or decreasing risks with increasing duration of drug use, a feature often associated with causal relationships. We applied a weighted regression model, so that the more precise studies have more influence in the analysis. The weight for each trial was equal to the inverse of the sum of the within-trial variance and the residual between-trial variance, which corresponds to a random effects meta-regression analysis [18]. Estimation of the residual betweentrial variance was based on a restricted maximum likelihood method [19].
This work was performed in accordance with the PRISMA statement for the conduct of meta-analyses of intervention studies [20]. For all tests, a probability level lower than 0.05 was considered statistically significant. All statistical tests were twosided. Stata 9 software was used for the statistical analyses (Stata Corp., College Station, Texas, USA).
Meta-regression analysis, using the mean age of participants and the mean duration of follow-up as covariates, did not reveal any significant association (Table 4).
The subgroup analysis, according to the type of fibrate, did not demonstrate any significant association between bezafibrate, clofibrate, fenofibrate or gemfibrozil use and cancer incidence (Table 3, Figure 5).
No statistically significant differences were observed between patients receiving fibrate vs placebo for any of the prespecified cancer subtypes (Table 3, Figure 5). Heterogeneity or reporting bias were not observed in any of these site-specific analyses (p.0.14 for all) with the exception of gastrointestinal cancer (Egger's p = 0.033; Table 3).
Last, in the meta-regression analysis, the results did not show any impact of either the mean age of participants or the duration of follow-up on the study estimates of relative risk (Table 4).
In a re-analysis, excluding all trials reporting zero cancer events in either or both arms, no statistically or clinically meaningful differences occurred vs the initial analyses (both for cancer mortality and cancer incidence).

Discussion
Fibrates comprise a class of lipid-lowering agents that significantly reduce cardiovascular events through a substantial reduction in serum levels of triglycerides and modest effects on levels of low-density and high-density lipoprotein cholesterol [45,46]. The concerns that lipid-lowering drugs (including fibrates) might increase the risk of cancer have been present for three decades [6,[47][48][49][50][51]. However, while for the 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins) several meta-analyses of existing data have convincingly demonstrated that they do not cause any substantial change in overall [52,53] or site-specific  cancer risk [54][55][56][57][58][59], this is the first meta-analysis that focuses specifically on the relationship between fibrates and cancer.
Our study encompassed 17 randomized placebo-controlled trials involving 45,000 individuals with a broad range of baseline characteristics, and accrued a total experience of 232,000 person-  years. The analysis did not provide any evidence that the use of fibrates significantly affects the incidence of cancer or cancerrelated death. When the analysis was restricted to major RCTs, the results remained practically unchanged. Similarly, when bezafibrate, clofibrate, fenofibrate and gemfibrozil were evaluated alone, we found no impact on cancer outcomes. No significant results were also noted for cancers of the respiratory tract, breast, colon, gastrointestinal tract, prostate, genitourinary tract, or skin (melanoma) when fibrates were used.
The neutral results of the present study are in line with several well-designed observational studies that have analysed overall cancer [60,61] and site-specific cancers [62][63][64] in fibrate users, and exclude the strong protective effect of fibrates found in the PRIME study [65]. This large observational cohort, suggesting that cancer mortality was significantly lower in fibrate users as compared with untreated dyslipidemic subjects (hazard ratio: 0.52, 95% CI: 0.28-0.97), may have been affected by uncontrolled (unmeasured) or residual confounding or other biases, problems known to plague even well-designed observational studies because they lack the experimental random allocation of the intervention necessary to test exposure-outcome hypotheses [66].
Our study has several merits. We have conducted an extensive literature search to retrieve all relevant eligible trials. Moreover, the absence of significant between-study heterogeneity, the small likelihood of important reporting bias, as well as the stability of results in subgroup and sensitivity analyses, reinforce our confidence in the validity of the conclusion that fibrate use has a neutral effect on cancer outcomes. The strengths of this quantitative synthesis should be, however, weighed against some limitations. First, the trials included in this meta-analysis were not designed to specifically analyze the relationship between fibrates and cancer risk. They have assessed cancer outcomes as secondary (safety) endpoints. Thus, problems in cancer detection and reporting may exist. However, the definition used and the surveillance intensity were consistent within each study for the fibrate and placebo groups, so the relative impact should still be accurate. Second, our search was restricted to published studies and we did not seek for unpublished/original data. However, we did not impose any exclusion criteria with regard to language, place of publication or study quality. Last, a main issue remaining beyond our control is cancer latency. As the exposure and followup times only lasted for nearly five years, estimates of cancer risk resulting from longer exposure to fibrates are not possible. Thus, our results should be interpreted with caution.
In conclusion, the synthesis of existing data from randomized placebo-controlled trials supports a neutral effect of fibrates on  cancer risk in the short term. We found no site-specific type of cancer whose risk was affected by fibrates, or subtype of fibrates that influenced the risk of cancer. However, given the steadily increasing use of fibrates during the past decade [8], and the indications for long-term and perhaps lifelong use, it is important to continue monitoring their long-term safety profiles.

Data Access and Responsibility
Stefanos Bonovas had full access to all the data in the study, takes responsibility for the integrity of the data and the accuracy of the data analysis and acts as guarantor of the paper.