Gastric cancer is one of the most common cancers globally [1]. And the highest incidence and mortality rates among both men and women are found in Asian countries, especially in Japan, Korea and China [2,3]. Several risk factors in gastric cancer including Helicobacter pylori infection, smoking, salt-preserved foods and dietary nitrite have been known well [4].

As obesity is the fastest growing disease worldwide [5,6], the evidences and mechanisms of obesity-induced gastrointestinal neoplasia have been suggested [7,8]. There were several meta-analyses for evaluating the association between body mass index (BMI, kg/m²) as level of body weight and risk of gastric cancer [9-12]. Among them, Renehan et al. in 2008 [9] reported the association between obesity and risk of gastric cancer had not a statistical significance. But Turati et al. [11] and Chen et al. [12] in 2013 reported that overweight as well as obesity were strongly related to risk of gastric cardia adenocarcinoma.

Interestingly, Yang et al. [10] concluded that excess body weight was associated with an increased risk of gastric cancer in non-Asians, but not in Asians. With reviewing of Yang et al. [10], following issues about the conclusion were draw. First, Yang et. al. [10] used the World Health Organization classification of weight with overweight(25.0 ≤ BMI ≤ 30.0) and obese (BMI ≥ 30.0). Thus, it is needed to re-analyze a meta-analysis using Asia-Pacific classification system with overweight (23.0 ≤ BMI < 24.9) and obese (BMI ≥ 25.0) [13]. Second, Park et al. [14] among selected studies by Yang et al. [10] should be excluded because it focused not on incidence but on survival in cancer patients. Last, Kuriyama et al. [15] published in 2005 ded not be included because of selecting studies that were published through to February 2009. Thus, the aim was to conduct a meta-epidemiological meta-analysis (MEMA) to evaluate association between excess body weight and the risk of gastric cancer in Asian adults.

Because the aim of this was to update and re-analysis the previous meta-analysis by Yang et al. [10], it was necessary to add relevant studies that were published till 30 April 2019. A search list was created through the citation discovery tools (CDT) of 'cited by' provided by PubMed [16] from 4 Asian studies [14, 17-19] selected by Yang et al. [10] as well as itself. The selection criteria were population-based prospective cohort studies that measured BMI of cohort participants and evaluated a risk of gastric cancer. As the materials were some published articles, it did not need an ethical consideration.

According to the proposed classification of weight by BMI in Asian adults [13], overweight group (OW) and obesity group (OB) were defined as 23.0-24.9 and ≥ 25.0, respectively. A group only showing results for BMI (18.5-22.9) as the reference, the inverse RR was calculated. The logarithm relative risk (logRR) and its standard error of logRR (SElogR) of each study was calculated from the extracted RR and 95% confidence intervals (CI).

Heterogeneity of selected articles was assessed with I² value (%). A random effect model was used when I² value was above 50% and if not, fixed effect model was used [20]. Subgroup analyses were conducted by three defined groups (OW, OB, and OWB) and sex (men and women). Publication bias was evaluated by Egger's test and Begg's test. The level of statistical significance was set to 0.05.

A total of 254 studies were retrieved using PubMed's CDT. Four studies among them were additionally selected [21-24]. While Oh et al. [17] and Jee et al. [21] shared the database constructed from the National Health Insurance Corporation in Korea, Jee et al. [21] was selected because it had a larger sample size with extended follow-up periods. Finally, 7 cohort studies with 21 datasets were selected for meta-analysis (Table 1) [15,18,19,21-24]. Twenty-one datasets were grouped into OW (i, k, and n in Table 1), OB (j, l, m, o, p, q, t, and u in Table 1) and OWB (a, b, c, d, e, f, g, h, r, and s in Table 1).

The I² value of OW, OB, and OWB were 76.1%, 83.5%, and 97.1%, respectively. The summary relative risk (sRR) of OB had a statistical significance but lost it with subgroup analyses by sex. Additional subgroup analyses suggested that only OWB in men had a I² value below 50% (22.5%) and showed a statistical significance with inverse association (sRR, 0.79; 95% CI, 0.77 to 0.81) (Fig. 1). The p-value of Egger’s test and Begg’ test for OWB in men were 0.059 and 0.497, respectively.

While all datasets in OW, OB and OWB showed a huge heterogeneity, BMI over than 23.0 (OWB) in Asian men having the lowest I² value was inversely associated with risk of gastric cancer. From these findings, the following three inferences could be draw.

Firstly, the different classification of BMI in Asian adults for overweight and obesity could cause a huge heterogeneity, and this might mask a real association between BMI level and risk of gastric cancer in Asian adults. Because only OWB in men having a I² value below 50% showed a statistical significance in this MEMA, but not in Yang et al. [10]. Of seven selected studies, three [15,21,22] did conduct to statistical analyses based on the Asian classification.

Secondly, the different classification of BMI for OWB could cause an opposing direction of sRR. While the sRR of OWB overall and OWB in men showed an under 1.0 in this MEMA, these sRRs in Yang et al. [10] were 1.17 (0.88-1.56) in overall and 1.10 (1.03-1.08) in men. The marginal BMI of OWB in Yang et al. [10] and this MEMA were over 25.0 and 23.0, respectively. Thus, it could be explained that a group having BMI between 23.0 and 24.9 might change the direction of sRR. Because this group was defined as OW in this MEMA, and the sRR of OW was less than 1.0 in Table 2, even though not having a statistical significance with I² values 76.1%. Based on these inferences, it could suggest a hypothesis that OW might be a protective factor in gastric cancer risk in Asian adults.

Lastly, men had higher sRR than women in OB group in Table 2, even though it was not a statistical significance. This finding was also found in Yang et al. [10] and Chen et al. [12] (Table 3). But additional studies are needed to suggest a hypothesis for risk difference between men and women because the I² values in men and women were 78.5% and 51.0%.

Thus, it will be necessary to conduct additional cohort studies with lengthening follow-up periods and re-analyzing the effect of overweight and obesity classified by the Asian criteria [25]. In addition, it will be necessary to perform an updated MEMA that adds information from further relevant studies by extending the end of search period.