Response to Yao's Letter

Dear Editor:

First of all, we would like to thank you for offering us the opportunity to clarify or refine our published article (Cui et al., 2011). We addressed the comments made by Yao et al. below.

It is indeed the case that two other studies (Shin et al., 2004; Cattaneo et al., 2006) can be easily found in PubMed, but the patients in the Shin study were all diagnosed with familial gastric cancer (28 cases from 27 familial gastric cancer families), being completely inconsistent with the other included patients in our analysis. Thus, Shin study was and should be excluded in our meta-analysis in order to avoid high heterogeneity (Mayrbaeurl et al., 2010; Mastoraki et al., 2011; Sereno et al., 2011; Yamada et al., 2011). In our meta-analysis, diffuse-type gastric cancer is sporadic.

As for Cattaneo study, this study was initially overlooked due to the lack of full text for free online. We recently obtained the full text of this article through NML catalog Electronic Links (http://ar.iiarjournals.org/content/by/year). We have conducted the meta-analysis again. When Cattaneo study is included, our original main conclusions are the same. (i) For all gastric patients regardless of ethnicity, OR₁ (p-value), OR₂ (p-value), and OR₃ (p-value) of C-160A SNP were 1.17 (p=0.377), 1.03 (p=0.612), and 1.07 (p=0.662), respectively, hardly suggesting a recessive effect of the putative susceptible A allele. But for Asians, since the participants of Cattaneo were Caucasians, OR₁ (p-value), OR₂ (p-value), and OR₃ (p-value) of C-160A SNP among Asians are the same, that is to say, 0.67 (p=0.032), 0.97 (p=0.701), and 0.69 (p=0.050), respectively, indicating a recessive effect of the A allele. The recessive genetic model of the susceptible A allele is still applicable in all our meta-analyses. As shown in Figure 1 (odds ratios [ORs] for associations between CDH1 C-160A SNP and gastric cancer risk [AA vs. CA+CC] among different ethnicity populations), those conclusions are still the same; no significant association of C-160A SNP was found with the overall risk of developing gastric cancer, but the opposite tendency could be noted between Caucasians and Asians, and a statistically significant finding could be witnessed among Asians. (ii) Given that no further information about anatomic or pathologic classification of gastric cancer was provided, those conclusions are still the same; an opposite association was seen between noncardia and cardia types or between sporadic diffuse and intestinal types of gastric cancer, but no statistically significant findings could be noted.

OPEN IN VIEWER

FIG. 1.

Odds ratios (ORs) for associations between CDH1 C-160A SNP and gastric cancer among different ethnicities.

In response to the comment that the data reported in the meta-analysis for the study by Humar et al. (2002) did not seem in line with the data provided in the original literature, the incorrect information was reported in Table 1 due to file conversion issues. However, all the statistical calculations were conducted in the original file with the correct numbers. The numbers for CC, CA, and AA in the controls should be listed as 40, 27, and 3 in Table 1.

Cui et al. mentioned that they calculated the Hardy–Weinberg equilibrium (HWE) in controls for all eligible studies. For brevity and uniformity, Cui et al. removed the study by Kuraoka et al. (2003) from the meta-analysis, as the controls in the study were deviated from HWE (p=0.013). However, we found that the distribution of genotypes in the controls was not consistent with HWE in the study reported by Zhang et al. (2008) (p=0.042). According to the exclusion criteria, they should remove this study from the analysis.

In Zhang's report, the recruited samples were successfully genotyped for the three CDH1 polymorphisms. None of the three SNP genotype distributions was correlated with age and gender in both the patients and controls (data not shown). The genotype distribution in esophageal squamous cell carcinoma (ESCC) and gastric cardia adenocarcinoma (GCA) patients and the healthy controls did not obviously deviate from that expected by the HWE (p>0.05). In our meta-analysis, HWE in controls was calculated again. The chi-square goodness of fit was used to test deviation from HWE (significant at the 0.05 level). After our uniform recalculation, p-value=0.042245. Generally speaking, any study that deviated from HWE through our calculation should have been removed; however, considering that (i) the p-value was close to 0.05 (though a little lower), the discrepancy possibly caused by different statistical software, (ii) the number of participants in the Zhang study was very large, and (iii) sensitivity analyses would be conducted in our meta-analysis, we determined after careful deliberation that Zhang study should remain in our meta-analysis. Whereas in the Kuraoka study, the p-value (0.013363) was much lower than 0.05, and the total sample size was very small so we finally decided to exclude the Kuraoka study from our meta-analysis.

Apart from ethnicity (Caucasians and Asians), the influence of factors such as study design, gender, country, smoking behavior, Helicobacter pylori (HP) infection, tumor location, and the Lauren's classification of gastric cancer would also have been interesting to investigate. Especially, tobacco use and HP infection are widely recognized as risk factors for gastric cancer. Therefore, it would be valuable if the authors could provide a stratification analysis for these risk factors in the meta-analysis.

We agree that, however, information for more specific stratification could not be obtained, so those subgroup analyses cannot be done in this meta-analysis.