Sir,

The aim of the performed meta-analysis was mainly to investigate the association between the overall cancer risk and the NQO1 C609T polymorphism in the worldwide population as well as in individual ethnic groups, given the biological plausibility of combining several different cancer sites in the light of the common background implied by the known functions of the investigated enzyme. Most notably, our meta-analysis demonstrated a statistically significant association in the worldwide and Caucasian populations, as indicated by the resulting P-values. Power analysis was not published for most of the individual studies, and therefore it was not reported in our meta-analysis for each individual study. The main purpose of conducting the present meta-analysis was to overcome the expected low power in most of the individual studies due to the small sample sizes, and at least improve the power of detecting association by combining the large number of studies. The calculated power for our present meta-analysis was 97% for the worldwide population analysis and 98% for the Caucasian subgroup analysis (α=0.05). Calculations were performed using the genetic power calculator developed by Purcell et al (2003). Therefore, the conclusions of the meta-analysis are further supported by the obtained high power, which was not unexpected given the large total number of samples, the significant resulting odds ratios, and the low P-values. The power obtained far exceeds the standard, and rather arbitrary, value of 80% with respect to the worldwide and Caucasian population, which leaves no possibility of a false-positive association. Concerning the tumour site analysis, we had noted in the article that the results should be approached with caution due to the small sample sizes available, and we highlighted the need for more studies investigating individual cancer sites and involving less common ethnic groups in the conclusion. The main conclusion from the meta-analysis performed involved the total cancer risk combining all tumour sites.

It is noteworthy that post hoc power analysis is controversial and often misinterpreted (Hoenig and Heisey, 2001). The common well-accepted usage for power is in prospectively estimating a sufficient sample size to detect an association when it is present in order to avoid type II error and false-negative associations.

Finally, we believe that it is unlikely that the inadvertently missed single case sample had an impact on the results and conclusions of the meta-analysis that included 21 178 case samples.