We read with great interest the recent paper by Lajin and Alachkar (2013). The authors performed a comprehensive meta-analysis of 92 case–control studies involving 21 178 cancer cases and 25 157 controls to examine the association between NQO1 C609T polymorphism and cancer susceptibility. Their comprehensive meta-analysis results suggest that NQO1 C609T polymorphism is an important genetic factor in the overall risk for developing cancer, especially in Caucasian populations. It is an interesting study. Nevertheless, we would like to raise several concerns related to this article.

First, sensitivity analysis may need to be routinely performed by excluding and including the Hardy–Weinberg equilibrium (HWE)-violating studies in meta-analyses of genetic association studies, which is a good approach to heterogeneity (Mao et al, 2010). We also assessed deviation from HWE in controls for all the included studies, and the results demonstrated that most genotype distributions for the control group were well goodness-of-fit except for five studies. However, the authors only performed the meta-regression analysis to identify three possible sources of heterogeneity including ethnicity, tumour site, and minor allele frequency (MAF). We would recommend that in their meta-analyses, the authors should conduct the meta-regression analysis including HWE, not only excluded these five studies deviated from HWE. Therefore, we believe that the bias would be introduced into the results of the meta-analysis due to this shortage.

Second, in the meta-analysis, the authors have retrieved data on the source of control groups (hospital- or population-based controls), but the definitions for the population-based study and hospital-based study were not clear in this meta-analysis. This point greatly influenced the results of this meta-analysis. For example, if the authors defined the population-based study as controls from healthy population, and the hospital-based study as controls from patients, we could clearly ascertain that at least the report by Zhang et al (2003) was not a population-based study. Furthermore, the authors should perform a stratified analysis by source of control groups.

Finally, the data reported by Lajin and Alachkar (2013) do not seem in line with the data provided by Malik et al (2011) in their original publication. The numbers reported by Lajin and Alachkar (2013) for CC, CT, and TT, in cases and controls, respectively, are 51–38–18 and 112–68–15. Interestingly enough, after carefully studying the data presented by Malik et al (2011), the frequencies that we have retrieved on the 108 cases and 195 controls were 51–39–18 and 112–68–15, respectively. Therefore, this similar error may exist in other included studies in the meta-analysis. It would be valuable if the authors could provide a more careful checking for genotype data in previously published studies.

In conclusion, the above comments may reveal that the association between the NQO1 C609T polymorphism and cancer susceptibility was conflicting. We believe that this remark will contribute to further, more accurate elaboration and substantiation of the original results presented by Lajin and Alachkar (2013).