Patients' characteristics
Forty-three patients with gastric cancer and 119 patients without gastric cancer who achieved HP eradication were analyzed in the present study. Patients' characteristics in the groups with/without gastric cancer are shown in Table 2. ESD was successfully performed for the patients with early gastric cancer (Supplementary Table 2). The sex ratio was significantly different between the groups (P = 0.0006). There was no significant difference in the duration after HP eradication between the groups, and patients without gastric cancer were younger than those with gastric cancer (P = 0.0305). There was no significant difference in Brinkman's index between the groups.
Differences in gastric mucosal endoscopic findings between patients with and without gastric cancer after HP eradication
The risk score according to the Kyoto classification of gastritis (total Kyoto risk sore) and scores for each endoscopic finding (atrophy, intestinal metaplasia, enlarged folds, nodularity, and diffuse redness) were compared between the groups with/without gastric cancer by univariate analyses (Table 3). The group with gastric cancer had the higher total Kyoto risk score (4.163 ± 0.1049) compared with the group without gastric cancer (2.361 ± 0.1263) (P < 0.0001). Among the five endoscopic findings, scores for atrophy (2.000 ± 0.000 vs. 1.496 ± 0.04755; P < 0.0001), intestinal metaplasia (1.814 ± 0.08323 vs. 1.076 ± 0.08619; P < 0.0001), and diffuse redness (0.2326 ± 0.08040 vs. 0.04167 ± 0.01832; P = 0.0027) were significantly higher in patients with gastric cancer compared with those without cancer, respectively. The prevalence of map-like redness was significantly higher in patients with gastric cancer than in patients without gastric cancer (60.5% vs. 31.1%, respectively; P = 0.0010). Next, we performed a logistic regression analysis with forward selection based on the likelihood ratio using the endoscopic findings that showed significant differences in the univariate analyses, namely, atrophy, intestinal metaplasia, diffuse redness, and map-like redness. As shown in Table 4, we developed a model with atrophy and diffuse redness (Hosmer–Lemeshow test: P = 1.000). The odds ratios of atrophy (p = 0.997) and diffuse redness (P = 0.035) were 1013681190.83195 and 3.988, respectively.
Gastric mucosal atrophy indicates the presence of gastric cancer after HP eradication.
Based on the regression equation (Table 4), the probability of the presence of gastric cancer was calculated for all 163 patients (Supplementary Table 1). We evaluated the predictive performance of this calculated probability for the presence of gastric cancer using a receiver operating characteristic curve analysis. The area under the curve was 0.7828 (95% confidence interval (CI): 0.7131–0.8524), and Youden's index suggested the threshold as 0.1892 (Table 5). With the threshold set at 0.1892, sensitivity was 100.0% (95% CI: 91.8%–100.0%), specificity was 49.58% (95% CI: 40.75%–58.43%), positive predictive value was 41.74%, and negative predictive value was 100.0% (Table 6). There are nine possible combinations of the atrophy score (0–2) and diffuse redness score (0–2), and the calculated probability was larger than 0.1892 only when the atrophy score equaled 2, regardless of the diffuse redness score (Table 7). These results indicated that severe gastric mucosal atrophy (atrophy score = 2) is an endoscopic marker to predict the presence of gastric cancer.