Analysis of the role of rs2031920 and rs3813867 polymorphisms within the cytochrome P450 2E1 gene in the risk of squamous cell carcinoma

Five electronic databases, including PubMed, Web of Science, Cochrane, Scopus and Chinese National Knowledge Infrastructure (CNKI), were searched prior to January 2018. We used a group of keywords: Carcinoma, Squamous Cell; Carcinomas, Squamous Cell; Squamous Cell Carcinomas; Squamous Cell Carcinoma; Carcinoma, Squamous; Carcinomas, Squamous; Squamous Carcinoma; Squamous Carcinomas; Carcinoma, Epidermoid; Carcinomas, Epidermoid; Epidermoid Carcinoma; Epidermoid Carcinomas; Carcinoma, Planocellular; Carcinomas, Planocellular; Planocellular Carcinoma; Planocellular Carcinomas; esophageal squamous cell carcinoma head and neck squamous cell carcinoma; lung squamous cell carcinoma; skin squamous cell carcinoma; oral squamous cell carcinoma; cervix squamous cell carcinoma; vagina squamous cell carcinoma; SCC; ESCC; HNSCC; LSCC; SSCC; OSCC; Cytochrome P-450 CYP2E1; Cytochrome P 450 CYP2E1; Cytochrome P-450-J; Cytochrome P 450 J; 4-Nitrophenol-2-Hydroxylase; 4 Nitrophenol 2 Hydroxylase; Dimethylnitrosamine N-Demethylase; Dimethylnitrosamine N Demethylase; Cytochrome P450 2E1; N-Nitrosodimethylamine Demethylase; N Nitrosodimethylamine Demethylase; CYP2E1; Cytochrome P-450 IIE1; Cytochrome P 450 IIE1; CYP IIE1; CYPIIE1; Cytochrome P-450 (ALC).

The retrieved studies were then reviewed and screened with the following exclusion criteria: (1) data based on animal experiments; (2) case reports, cohort studies or meeting abstracts; (3) without SNP data; (4) meta-analyses or reviews; (5) no SCC or CYP2E1 data; (6) duplicate studies; (7) no pathological typing data; (8) no genotype data. The data of genotype frequencies in cases and controls must have been provided in the selected studies.

Based on the eligible articles, the authors extracted and summarized the usable information, including the first author’s name, year, country, race, SNP, genotype frequency, SCC type, control source, genotyping assay, and HWE (Hardy–Weinberg equilibrium), in a table. The Newcastle–Ottawa Scale (NOS) system was also used to assess the methodological quality of individual studies. Only the studies with NOS score > 5 were ultimately included in our meta-analysis.

STATA software (Stata Corporation, College Station, TX, USA) was used for our heterogeneity and association tests. In the case of heterogeneity, the P value of Cochran’s Q statistic < 0.05 or I² value > 50% were considered to represent high heterogeneity among studies, which led to the use of a random effects model (DerSimonian and Laird method). Otherwise, the fixed effects model (Mantel–Haenszel statistics) was used. In the association test, odds ratio (OR), 95% confidence interval (CI) and P value were computed to assess the association strength in the allele, carrier, homozygote, heterozygote, dominant, and recessive models. In addition, based on the factors of race, SCC type, control source and HWE, a series of subgroup analyses were performed as well.

After our initial database retrieval, a total of 393 records [PubMed (n = 89), Web of Science (n = 161), Cochrane (n = 1), Scopus (n = 116) and CNKI (n = 26)] were obtained, as presented in Fig. 1. Then, 113 duplicate records were excluded. Based on the exclusion criteria, 223 records were removed. Moreover, the lack of confirmed pathological typing data or genotype frequency distribution resulted in the exclusion of another 25 articles. Finally, our meta-analysis involved a total of 32 articles [14‐45] containing 7435 cases and 10,466 controls. The characteristics of each study are presented in Table 1. No study had poor quality; the NOS score of all studies was greater than five (Table 1).

A meta-analysis of rs2031920 and SCC risk was conducted on the allele model (allele T vs. allele C), carrier model (carrier T vs. carrier C), homozygote model (TT vs. CC), heterozygote model (CT vs. CC), dominant model (CT + TT vs. CC), and recessive model (TT vs. CC + CT). As shown in Table 2, 18 case–control studies were enrolled for the allele, carrier, heterozygote models, 15 case–control studies were enrolled for the homozygote model, 21 case–control studies were enrolled for the dominant model, and 16 case–control studies were enrolled for the recessive model. Pooling results suggested that there was no statistically significant difference for the overall SCC risk between the case and control groups under any model (Table 2, P value of association test > 0.05).

Moreover, we conducted a statistical analysis of the subgroup of race (Asian/Caucasian), SCC type (HNSCC/ESCC/LSCC), control source (PB/HB), and HWE (Y/N). As shown in Table 2, in comparison with controls, a reduced ESCC risk was observed in the models of allele T vs. allele C (P = 0.025, OR = 0.67), carrier T vs. carrier C (P = 0.014, OR = 0.70), TT vs. CC (P = 0.029, OR = 0.65), CT vs. CC (P = 0.040, OR = 0.56), CT + TT vs. CC (P = 0.035, OR = 0.58), but not TT vs. CC + CT (P = 0.770). Figure 2a shows forest plot data in subgroup analysis by SCC type under the allele model. The “T” allele carrier of the rs2031920 polymorphism within the CYP2E1 gene seems to be linked to ESCC risk.

We also conducted the overall and subgroup meta-analysis of rs3813867 and SCC risk under the allele (10 case–control studies), carrier (10 case–control studies), homozygote (6 case–control studies), heterozygote (10 case–control studies), dominant (11 case–control studies), and recessive (6 case–control studies) models. The positive results regarding the association between CYP2E1 rs3813867 and SCC risk were detected in the overall SCC meta-analysis and subgroup analysis of “ESCC” and “Y” (P value of Hardy–Weinberg equilibrium > 0.05) under all genetic models (Table 3, all P < 0.05, OR < 1), only apart from the heterozygote model (P = 0.150). A decreased SCC risk was also detected in the subgroup analysis of “Asian” and “PB” under all genetic models (Table 3, all P < 0.05, OR < 1). Figure 3a shows the forest plot data of subgroup analysis by SCC type under the allele model. The “C” allele carrier of CYP2E1 rs3813867 polymorphism may be associated with the risk of SCC, especially the ESCC cases in Asian populations.

The results of overall and subgroup meta-analysis of the rs2031920/rs3813867 haplotype and SCC risk under the allele (five case–control studies), carrier (five studies), homozygote (three studies), heterozygote (five studies), dominant (seven studies), and recessive (three studies) models are shown in Table 4. We observed a decreased SCC risk in the overall SCC meta-analysis under the allele, carrier, homozygote, and dominant models (Table 4, all P < 0.05, OR < 1), and the subgroup analysis of “PB” under the allele, carrier, and dominant models (all P < 0.05, OR < 1). These results suggested a potential link between the c1/c2 or c2/c2 of rs2031920/rs3813867 haplotype and SCC risk, which still requires more case–control studies.

When assessing the heterogeneity level, the fixed model was used for the TT vs. CC model of rs2031920 due to the lack of high heterogeneity (Table 5, I² = 38.3%, P value of heterogeneity = 0.066), however, the random model was utilized for others. The fixed model was used for the allele, carrier, homozygote and recessive models of rs3813867 (Table 5, all I² < 50.0%, P value of heterogeneity > 0.05); and the allele, carrier, homozygote, dominant, and recessive models of the rs2031920/rs3813867 haplotype (Table 5, all I² < 50.0%, P value of heterogeneity > 0.05).

Begg’s and Egger’s tests did not provide confirmed evidence of obvious publication bias in the above comparisons (Table 5, all P value of Begg’s test and Egger’s test> 0.05) apart from the CT + TT vs. CC model of rs2031920 (P value of Egger’s test = 0.037). Figures 2b and 3b show the Egger’s publication bias plot of rs2031920 and rs3813867 under the allele model, respectively. Additionally, a relatively stable conclusion was obtained by sensitivity analysis results (Fig. 2c for allele model of rs2031920; Fig. 3c for allele model of rs3813867; data for others not shown).