Statin and the risk of hepatocellular carcinoma in patients with hepatitis B virus or hepatitis C virus infection: a meta-analysis

Systematic search of electronic databases of PubMed, Cochrane’s Library, and Embase were performed to identify potentially relevant studies, via the following terms: (1) “statin” OR “3-hydroxy-3-methyl-glutaryl CoA reductase inhibitor” OR “CS-514” OR “statin” OR “simvastatin” OR “atorvastatin” OR “fluvastatin” OR “lovastatin” OR “rosuvastatin” OR “pravastatin” OR “pitavastatin”; and (2) “chronic hepatitis B” OR “chronic hepatitis C” OR “hepatitis B virus” OR “hepatitis C virus” OR “HBV” OR “HCV”. We used this extensive search strategy to avoid missing of potentially relevant studies. The search was limited to human studies, and no language restriction was applied. Besides, we also studied the reference lists of related original studies and review articles using a manual approach. The final literature search was conducted on September 15, 2019.

The inclusion criteria were: (1) full-length articles reporting longitudinal follow-up studies, including randomized controlled trials (RCTs), cohort studies, and nested case-control studies; (2) enrolled at least 1000 adult patients with HBV or HCV infection and without HCC at baseline; (3) investigated the association between statin use and HCC risk during follow-up, with a minimal follow-up duration of 1 year; and (4) reported the relative risk for this association after adjustment of potential confounding factors. Review articles, preclinical studies, and studies irrelevant to the purpose of current meta-analysis were excluded.

Two authors indepdently performed database search, data extraction, and study quality assessment according to predefined criteria. If discrepancies occurred, they were solved by consensus between the two authors or discussion with the corresponding author. Data extracted included: (1) study information: name of first author, publication year, and study country; (2) study design characteristics; (3) patient characteristics: disease status, sample size, age, sex, prevalence of diabetes, and proportions of patients with cirrhosis at baseline; (4) definition of statin use; (5) follow-up durations; (6) strategy for HCC validation and number of HCC cases during follow-up; and (7) confounding factors adjusted. The Newcastle-Ottawa Scale was used as an instrument for study quality evaluation [28]. This scale ranges from 1 to 9 stars, and assesses study quality mainly regarding three domains, including study group selection, between-group comparability, and validation of the outcome of interest.

A risk ratio (RR) with corresponding 95% confidence interval (CI) was used as the main measure for the association between statin use and HCC risk in patients with HBV or HCV infection. Data of RRs and their corresponding stand errors (SEs) were calculated from 95% CIs or p values, and a logarithmical transformation was performed to stabilize variance and normalized the distribution [27]. The Cochrane’s Q test was performed to evaluate the heterogeneity, and the I² statistic was also estimated [29]. An I² > 50% indicates significant heterogeneity. We used a random-effect model for the meta-analysis of RR data because this model incorporates the potential heterogeneity among the included studies to calculate a more generalized result [27], By omitting one individual study at a time, we performed sensitivity analyses to test the robustness of the results [30]. We also performed stratified analyses to evaluate the influences of virus type, age, gender, diabetic status, with or without cirrhosis, lipophilic or hydrophilic statins, and accumulative dose of statin on the results. The potential publication bias was initially detected by visual inspection of the symmetry of funnel plots, then complemented with the Egger’s regression asymmetry test [31]. RevMan (Version 5.1; Cochrane Collaboration, Oxford, UK) software was used for the meta-analysis.

Figure 1 shows the literature search process. Briefly, 732 articles were obtained via initial search of the PubMed, Cochrane’s Library, and Embase databases, and 707 were excluded through screening of the titles and abstracts mainly because they were not relevant to the purpose of the meta-analysis. Subsequently, 25 records underwent full-text review. Of these, 12 were further excluded because four of them did not evaluate statin use as exposure, five did not report outcome of HCC risk, one did not provide available data for the multivariate adjusted association between statin use and HCC risk, and the remaining three were abstracts of already included studies. Finally, we included 13 studies in this meta-analysis [13‐25].

The characteristics of the studies were presented in Table 1. All of them were observational studies, among which one was a prospective cohort study, another one was a nested case-control study, and the remaining 11 were retrospective cohort studies. Since two studies reported the association between statin use and HCC risk in patients with HBV and HCV infection separately, these datasets were included independently [22, 24]. Overall, 15 datasets from 13 studies, with 519,707 adult patients with HBV or HCV infection were included [13‐25]. These studies were performed in China [13, 14, 16, 17, 21, 24], Korea [19, 20], the US [15, 18, 23, 25], and Sweden [22]. The mean age of the patients varied from 35 to 64 years, with percentiles of male ranging from 49 to 98%. Statin use was validated by prescription records in all studies and defined by accumulative statin dose of more than 28~30 cumulative defined daily dose (cDDD) in most studies [13‐17, 20‐23]. The follow-up duration varied from 2.5 to 10.7 years. The International Classification of Diseases (ICD) version 9 or 10 codes were used to validate HCC cases, and a total of 40,588 patients with HCC were included. Potential confounding factors including age, sex, diabetic status, comorbidities, and concurrent medications, were adjusted when presenting the outcome. The NOS scores of the included studies ranged from seven to nine, indicating generally good study quality.

Pooled results of all included studies using a random-effect model showed that statin use was associated with a reduced risk of HCC in patients with HBV or HCV infection (RR: 0.54, 95% CI: 0.44 to 0.66, p < 0.001; Fig. 2a) with significant heterogeneity (p for Cochrane’s Q test < 0.001, I² = 86%). Sensitivity analyses by omitting one datasets at a time did not significantly change the results (RR: 0.50 to 0.56, p all < 0.05).

Stratified analyses showed that the association between statin use and reduced risk of HCC was consistent in patients with HBV (RR: 0.52, 95% CI: 0.40 to 0.68, p < 0.001) or HCV infection (RR: 0.58, 95% CI: 0.43 to 0.77, p < 0.001; Fig. 2b), in patients < 50 years (RR: 0.58, 95% CI: 0.44 to 0.76, p < 0.001) or ≥ 50 years (RR: 0.41, 95% CI: 0.30 to 0.57, p < 0.001; Fig. 3a), in males (RR: 0.51, 95% CI: 0.38 to 0.69, p < 0.001) or females (RR: 0.51, 95% CI: 0.35 to 0.75, p < 0.001; Fig. 3b), in diabetic (RR: 0.52, 95% CI: 0.39 to 0.69, p < 0.001) or non-diabetic patients (RR: 0.47, 95% CI: 0.31 to 0.70, p < 0.001; Fig. 4a), and in patients with (RR: 0.52, 95% CI: 0.35 to 0.79, p = 0.002) or without cirrhosis (RR: 0.50, 95% CI: 0.41 to 0.60, p < 0.001; Fig. 4b). The association between statin use and reduced risk of HCC were not significantly affected by the above patient characteristics (p for subgroup difference all > 0.05). However, stratified analyses with three datasets in each stratum showed that use of lipophilic statins was associated with reduced risk of HCC in patients with HBV or HCV infection (RR: 0.52, 95% CI: 0.44 to 0.62, p < 0.001), but not for hydrophilic statins (RR: 0.89, 95% CI: 0.73 to 1.07, p = 0.21; p for subgroup difference < 0.001; Fig. 5a). Nine studies reported the potential dose-response relationship between statin use and risk of HCC according to the cDDD of statins [13, 14, 16, 17, 19, 20, 22‐24]. However, difference cut-off values for cDDD were used, which prevented a dose-response analysis in our meta-analysis. Subsequently, we performed stratified analyses comparing the association between statin use and HCC risk in patients with highest and lowest cDDD categories in each study. Results showed that the association between statin use and reduced risk of HCC was more remarkable in patients with highest cDDD category for statin prescription (RR: 0.37, 95% CI: 0.27 to 0.51, p < 0.001) compared to those with lowest cDDD category (RR: 0.64, 95% CI: 0.55 to 0.75, p < 0.001; p for subgroup difference = 0.002; Fig. 5b).

The funnel plots for the meta-analysis of the association between statin use and HCC risk in patients with HBV or HCV infection were shown in Fig. 6. These plots were symmetry on visual inspection, suggesting low risk of publication bias. Results of Egger’s regression test also suggested low possibility of publication bias (p = 0.188).