3.3 Heterogeneity and publication bias
Egger’s test showed no evidence of publication bias in the studies analysed except ABCB1 c.3435C>T polymorphism in the(CC vs CT+TT) and (CC+CT vs TT) model. The distribution of the genotype in the CBZ responsive group of each study was consistent with HWE except 2 studies [22, 29] for ABCB1 c.3435C>T and 1 study [7] for EPHX1 c.337T>C and c.416A>G . The heterogeneity and publication bias of included studies on each gene polymorphism are presented in Table 2. Considering that significant heterogeneity was identified in overall studies and the ineligible studies for HWE, we performed a sensitivity analysis by repetitive calculation of the combined OR estimate and a heterogeneity test after removing one study at a time. However, while remove 1 study [22] in the (CC vs TT) model of ABCB1 c.3435C>T which is not conformed with HWE, the results of association changed with significant difference (OR = 1.781 (95% CI: 1.166~2.722), P = 0.008) (Table 4).