Heterogeneity and sensitivity analyses
Significant heterogeneity was observed, as shown in Tables 2 and 3. We observed that ethnicity, HWE, and quality score were not source of source of heterogeneity. However, when we excluded Pereira Serafim PV et al. (supplemental Table 1), I2 value dropped and significant association was found in overall analysis (G vs. A: OR = 1.15, 95% CI: 1.02–1.30) for CYP1A1 A2455G polymorphism. Furthermore, after BFDP correction, no significant association was observed in G vs. A (BFDP= 0.998)
Sensitivity analysis was conducted using two methods. First, the results did not change when we applied one by one exclusion in the present study. Further, when we only involved high-quality and HWE in controls studies, a significant association was yielded in Asians (C vs. T: OR = 0.92, 95% CI: 0.85–1.00), Asia countries populations (C vs. T: OR = 0.92, 95% CI: 0.85–1.00), rectal cancer (CC vs. TT: OR = 0.73, 95% CI: 0.55–0.97; CC vs. (TT + TC): OR = 0.75, 95% CI: 0.57–0.98; C vs. T: OR = 0.89, 95% CI: 0.80–1.00), and females (TC vs. TT: OR = 1.27, 95% CI: 1.00–1.62) for the CYP1A1 T3801C polymorphism, as also shown in Table 2; concerning the CYP1A1 T2455G polymorphism, a significant association was observed in overall analysis, Asians, Caucasians, Asia countries populations, Europe countries populations, and males, as also shown in Table 3. However, after BFDP correction, no significant associations were found in the above all analyses (all BFDP > 0.8), as also listed in Tables 2-3.