Abstract
Dear Editor:
We have read with great interest the recent article titled “Meta-Analysis of Epidermal Growth Factor Polymorphisms and Cancer Risk: Involving 9,779 Cases and 15,932 Controls” published online on November 9, 2011, in the issue of DNA and Cell Biology (Li et al., 2011). The meta-analysis by Li et al. (2011) included 41 case–control studies involving 9779 cases and 15,932 controls to investigate the association between epidermal growth factor (EGF)-gene-related polymorphisms and cancer risk susceptibility, and the results suggested that the EGF +61G allele had a higher cancer risk in Mixed and European racial subgroups, whereas no significant association was found among EGF −1380A/G, −1744G/A, rs6983267T/G polymorphisms, and cancer risk. In conclusion, EGF +61A/G might act a risk factor for the development of cancer as well as a low-penetrance susceptibility cancer biomarker. It is an interesting study. Nevertheless, close inspection of the studies analyzed by the authors revealed some methodological issues that are worth mentioning and clarifying.
First of all, after we adopt the same search strategy and end-of-search date as Li et al. (2011), 3 relevant case–control studies, including renal clear cell carcinoma, hepatocellular carcinoma, and colorectal cancer with a total number of 684 cases and 626 controls, were not searched. Since these three articles were from China, and the language used was Chinese, so in the searched database section, the authors should add Chinese databases.
Second, evidence indicated that only the larger study should be included for the analysis when some publications contain the same or overlapping data (Little et al., 2002). Four studies have been performed by Araújo et al. (2009a, 2009b, 2011a, 2011b). In the four studies, all patients were recruited from the Portuguese Institute of Oncology-Porto (IPO-P), and the control subjects were randomly recruited from the Blood Donor Bank of IPO-P and Hospital de S. Marcos, Braga, and had no current or previous history of neoplastic disease. So, the largest case–control should be included in this meta-analysis. Meanwhile, Kang et al. (2007a, 2007b) have reported two studies, including the similar overlapping data.
Third, in the statistical analysis section, they stated that “A p>0.05 for the Q-test indicated a lack of heterogeneity among the studies.” Actually, a p-value>0.10 rather than >0.05 for the Q-test indicated a lack of heterogeneity among the studies (Kavvoura and Ioannidis, 2008).
Fourth, the distribution of genotypes in all the controls was in agreement with Hardy–Weinberg equilibrium (HWE; p>0.05), except for four studies in this meta-analysis. Actually, as a strict meta-analysis, studies whose distribution of genotypes is not consistent with HWE should not be included (Sen and Burmeister, 2008).
Finally, as a whole meta-analysis, the Egger's test should be included, which can assess funnel plot asymmetry and publication bias. A p<0.05 was considered statistically significant (Egger et al., 1997).
In conclusion, the results of the study by Li et al. (2011) should be interpreted with caution. To reach a definitive conclusion, further studies based on larger sample size are still needed to assess the association of EGF gene polymorphisms (+61A/G, −1380A/G, −1744G/A, and rs6983267) with susceptibility to cancer risk, especially in Mixed and African populations. We believe that our remarks will contribute to more accurate elaboration of the results presented by Li et al. (2011).