Methodological remarks concerning the recent meta-analysis on CYP1A1 polymorphisms-smoking interaction and hepatocellular carcinoma risk

Abstract

We read with great interest the recent meta-analysis by Yu et al. [1], which evaluated the association between Cytochrome P-450 1A1 (CYP1A1) polymorphisms, cigarette smoking and hepatocellular carcinoma risk. Nine studies were included in their meta-analysis and their study found that the MspI and Ile–Val polymorphisms of CYP1A1 may play important roles in increasing susceptibility to smoking-related hepatocellular carcinoma (HCC). Nevertheless, close inspection of the studies analyzed by the authors revealed some methodological issues that are worth mentioning and clarifying. Adopting the same search strategy and end-of-search date as Yu et al. [1], we have located one relevant case–control study in CNKI with a total number of 102 HCC cases and 102 controls, which have not been included in the meta-analysis [2]. Moreover, a methodological issue needs to be addressed with respect to the data provided in this meta-analysis by Yu et al. The data reported by Yu et al. for the study of Yuan et al. [3] (about HBV carriers for MspI polymorphism) do not seem to be in line with the data actually provided by Yuan et al. [3]. Yu et al. reported that the m1m1 genotype frequency in cases/controls among HBV carriers was 76/49 and the m2m2 was 75/50 respectively, whereas the m1m1 was 75/50 and m2m2 was 76/49 in Yuan et al.’s study. Also, the data reported by Yu et al. for the study of Silvestri et al. [4] are not consistent with the data provided by Silvestri et al. In the study of Silvestri et al., there were only 91 HCC cases and 99 blood donors as controls; however, Yu et al. reported 95 HCC cases and 103 controls for the Ile–Val polymorphism and 107 HCC cases and 115 controls for MspI polymorphism in CYP1A1 gene. Meanwhile, the genotype frequency in cases/controls for the two polymorphisms loci were incorrect; Yu et al. taken the total number for cases and controls in Silvestri et al’s study as the wild homozygous genotype for cases and controls (Table 1 and Table 2 in Yu et al.’s study). Moreover, we can only obtain the genotype frequency of Ile/Val ?Val/Val and Ile/Ile for the Ile–Val polymorphism, and m1m2 ? m2m2 and m1m1 for MspI polymorphism in the study of Silvestri et al., while the separate genotype frequency of Ile/Ile, Ile/Val, Val/Val for the Ile–Val polymorphism (m1m1, m1m2, m2m2 for MspI polymorphism) were unreported. Therefore, Silvestri et al’s study can only be included in the pooled analysis under dominant genetic model, but not under the other three genetic models. However, Yu et al. did not address this in rtmfthe statistical analysis section in their meta-analysis. Moreover, some other methodological issues seem worth commenting. First, the genotype distribution of MspI polymorphism among control subjects disobeyed the law of Hardy–Weinberg equilibrium (HWE) in two studies [5, 6]. The deviation from HWE presents the probability of genotyping errors, selection bias or other bias [7], and the results of genetic association studies might be spurious if the distribution of genotypes in the control groups were not in HWE [8, 9]. Therefore, the authors should carry out subgroup analysis by HWE in controls. Second, as the authors said in the article, the definitions of cigarette smoking were different for the included studies. It is not very suitable to pool the studies with different definitions of cigarette smoking together. Even if the definitions of cigarette smoking were the same for the 3 included studies, there still was a very high risk of reporting bias for F. Liu B. Li (&) Y.-G. Wei W.-T. Wang Department of liver and Vascular Surgery, West China Hospital of Sichuan University, 37 Guo Xue Road, Chengdu 610041, Sichuan Province, China e-mail: cdlibo688@163.com