Catechol-O-methyltransferase Val158Met polymorphism in breast cancer risk

Abstract

We read with great interest in two recent articles about meta-analyses of the association between catechol-Omethyltransferase (COMT) Val158Met polymorphism and breast cancer risk [1, 2]. The study by Ding et al. [1] included 26 case–control studies with 16,693 breast cancer patients and 18,261 healthy controls. The authors found no significant association between COMT Val158Met polymorphism and breast cancer risk in all genetic models in overall, Asian and European populations. After two studies were excluded as their controls genotype frequencies were not in Hardy–Weinberg equilibrium, a marginally significant association with breast cancer risk was observed in European population under a recessive genetic model (Met/ Met vs. Met/Val and Val/Val: odds ratio(OR) = 0.90, 95% confidence interval (CI) 0.90–1.00, P = 0.06). They concluded that COMT Val158Met polymorphism might be a risk factor in the development of breast cancer. Actually, as the authors found a marginally significant decreased breast risk for Met/Met vs Met/Val and Val/Val among Europeans, it should be concluded that COMT Val158Met polymorphism might be a protective factor rather than a risk factor for breast cancer development. If so, the conclusion was fallacious, as it was contrary to the biological view that COMT Met/Met homozygote was usually considered as the risk genotype for breast cancer development. The other one by Mao et al. [2] consisted of 41 studies with 25,627 cases and 34,222 controls. The results suggested that COMT Val158Met polymorphism was not significantly associated with breast cancer risk in all genetic models in both overall and subgroup analyses (by ethnicity or menopausal status). The sensitivity analysis further strengthened their conclusions. The partial discrepancy between two studies drove us to seek the potential reasons. We compared the two studies with each other carefully and found some methodological issues in the study by Ding et al. [1]. First, the authors did not state whether there was a language limitation in the ‘‘Search strategy’’ section. Indeed, recruiting all related published or unpublished studies with no language restriction might be a proper strategy, thereby minimizing the potential language and publication biases. In addition, just searching PubMed database would have omitted many important data, and then caused the estimated effect imprecisely when performing a meta-analysis. Other databases such as EMBASE, ISI web of science, as well as non-English databases, should also be searched to reduce the selection bias. Second, the data presented by Ding et al. were not in accord with those provided by the original publications. The genotypes of Val/Val, Val/Met, and Met/Met in cases and controls should be 68/37/13 and 66/55/4 for the study by Huang et al. [3], 16/64/46 and 13/61/43 for the study by Bergman-Jungestrom et al. [4], 84/156/73 and 58/144/60 for the study by Ahsan et al. [5], 645/1,360/845 and 448/926/534 for the study by Dunning et al. [6], 287/521/ 240 and 277/549/266 for the study by Gaudet et al. [7], 731/1,569/844 and 1,243/2,669/1,569 in the study by MARIE-GENICA et al. [8], 44/82/28 and 52/85/29 for the study by Yadav et al. [9]. The country of origin should be B. Xi (&) W. Liu Institute of Maternal and Child Health Care, School of Public Health, Shandong University, Jinan 250012, China e-mail: xibo2010@sdu.edu.cn