Study inclusion and characteristics
Through the planned search strategy, 86 documents were identified; after a review of titles and abstracts, 40 studies that appeared to be potentially inclusive were searched and evaluated (Fig. 1). In these, 2,755 patients were included from 9 articles reporting relevant data from 8 individual studies.18-26 After being diagnosed with GC and undergoing ER, these patients were followed for more than two years at the attending hospital. The patients were evaluated for the presence of IM or degree of AG at the time of ER.
The characteristics of studies included by appropriate criteria are presented in Table 1. The proportions of men in the studies ranged from 63% to 82%. At the time of ER, about 67% were negative for H. pylori or had successful eradication treatment of the infection within one year. The shortest and longest median follow-up period was 30 and 62 months, respectively.
In this study, the diagnosis of severe atrophy was basically based on the contents defined by the authors of each paper. This study analyzed severe atrophy based on the definition of each paper's authors. Among the studies that defined atrophy endoscopically, five studies were classified based on the Kimura-Takemoto classification. In three studies, open type II and III were defined as severe, one was open type III, and the other was all open type severe. One study did not specifically describe the Kimura classification and only mentioned severe AG. Two studies classified severe atrophy with a low serum pepsinogen I/II ratio, and three studies defined pathologic severe atrophy by updated Sydney classification. In the study where only open type III was evaluated as more extensive atrophy, the serum PG I/II ratio and histology were also investigated.21 It was evaluated that open type III reflects the most severity.
The risk of bias assessments for all included studies was evaluated by the Newcastle-Ottawa Scale for Cohort Study Quality (Supplementary Table S1).
Effects of the degree of mucosal atrophy on the occurrence of MGC after ER for GC
Data obtained from 8 studies were pooled. 18-26 Of the 2,755 cases, 1,295 cases were classified as severe AG, and 1,250 cases were classified as non-severe AG. The incidence of MGC in GC patients with severe AG and non-severe AG was 186 (14.4%) and 86 (6.9%), respectively (absolute risk difference 7.5%; Table 2). The incidence rate of MGC was 36.9 (95% CI 31.8–42.6) per 1,000 person-years in severe AG and 19.4 (95% CI 15.5–23.9) per 1,000 person-years in non-severe AG, and the difference in incidence rate was 17.5 per 1,000 person-years. Patients with severe atrophy of the background mucosa at the ER had a higher risk of MGC occurrence than patients without severe AG (RR 1.64, 95% CI 1.29–2.07, I2=45%; Figure 2).
Subgroup analysis was performed to estimate the MGC occurrence according to the parameter for judging the degree of mucosal atrophy. Endoscopically evaluated severe AG had significantly higher MGC occurrence than non-severe AG (RR 1.90, 95% CI 1.25–2.90, I2=52%; Figure 3a). In addition, serological (PG I, II) and histological (corpus and antrum) diagnostic methods were also analyzed. Testing for subgroup differences revealed no statistically significant subgroup effect (P=0.58). However, it was shown that severe AG favored MGC occurrence more than non-severe AG in all subgroups (P=0.58).
Countries were also sub-analyzed to examine the possibility of geographically different outcomes (Figure 3b). Five Japanese studies18,20-22,25,26 and three Korean studies19,23,24 were included in this study. In the Japanese study, the number of severe AG patients (n=905, 57.2%) was higher than the number of non-severe AGs, and in the Korean study, the number of severe AGs (n=352, 35.2%) was less classified. The RR for MGC in severe AG was 1.75 (95% CI 1.11–2.73, I2=55%) in the Japanese study and 2.09 (95% CI 1.09–4.02, I2=26%) in the Korean study. Test for subgroup difference showed no subgroup effect (P=0.65).
Effects of the presence or absence of IM on the occurrence of MGC after ER for GC
We investigated whether the presence of IM affects the occurrence of MGC in patients who underwent ER for GC. In four studies, 2,755 patients were included. The median follow-up period after ER was 2.5 to 4.2 years. IM was diagnosed histologically in 3 studies,19,23,24, and 1 study was diagnosed grossly during endoscopy.21,22 In the presence of IM, the absolute risk difference of MGC was higher at 9.2%, and the incidence rate difference per 1,000 person-years was 22.5 (Table 3).
In GC patients with IM, the RR of MGC occurrence was significantly higher than in patients without IM (RR 7.08, 95% CI 3.63–13.80, I2=0%; Figure 4). Only three studies with histological evaluation of IM out of 4 were performed for subgroup analysis. In patients with IM, the incidence of MGC was significantly increased by 6.94 (95% CI 3.48–13.82, I2=0%) compared with those without IM (Supplementary Figure S1). There were two studies in which the biopsy site was divided into antrum and corpus, but the meta-analysis results were the same.
Relationship between Helicobacter pylori infection status and MGC
A meta-analysis was performed to evaluate the relationship between H. pylori infection status and MGC occurrence (Figure 5). Since many people have not been clearly evaluated for H. pylori eradication in the past, we compared 660 patients with persistent infection and 1,912 patients with negative or eradicated infections. However, there was no significant difference between the risk of MGC occurrence in those with H. pylori-negative or eradicated, and those with persistent infection (RR 1.08, 95% CI 0.85–1.39, I2=0%).
Analyzes of risk factors affecting the occurrence of MGC
We performed an analysis to determine whether the risk of MGC occurrence increases according to demographic factors and characteristics of primary cancer. The risk of MGC occurrence according to demographic factors such as gender, age, drinking history, and smoking history were presented (Supplementary Table S2, Supplementary Figure S2). The risk of developing MGC was significantly higher in men than in women (total number of patients = 2579, RR 1.54, 95% CI 1.07–2.22, I2=22%), and there was also a greater incidence of MGC in patients with a history of smoking (total number of patients = 412, RR=2.10, 95% CI 1.25–3.53, I2=0%). The risk of developing MGC was higher in alcoholics, but no significant findings were found. There was only one study on the elderly.
A forest plot for the occurrence of MGC according to the characteristics of primary cancer was presented (Supplementary Table S3). The primary cancer location, gross type, Lauren's classification, the degree of differentiation, and the depth of invasion were investigated. Supplementary S3 is a meta-analysis for each variable, and no risk factors showed a statistically significant effect on the occurrence of MGC.