In the current study, we developed an EGC index to differentiate EGC from non-cancer controls based on the serum levels of four miRNAs (miR-4257, miR-6785-5p, miR-187-5p, and miR 5739). In the validation set, the EGC index demonstrated a sensitivity of 0.996 and a specificity of 0.953, with an AUC of 0.998. The sensitivities of the EGC index did not differ significantly among the clinical stages of EGC or between the three sets of non-cancer control samples.
The updated version of the Japanese Guidelines for Gastric Cancer Screening recommends performing an upper gastrointestinal series and gastroscopy for population-based and opportunistic gastric cancer screenings . Some large-scale cohort studies have reported that both of these screening modalities contribute to the reduction of gastric cancer mortality [14,15,16], although there were several inconsistent results among the studies. Hamashima et al. reported that the sensitivities of radiographic and endoscopic screening methods for EGC detection were 0.893 (95% C.I. 0.718–0.977) and 0.955 (95% C.I. 0.875–0.991), respectively, with specificities of 0.856 (95% C.I. 0.846–0.865) and 0.851 (95% C.I. 0.843–0.859), respectively . In the study by Hamashima et al., the false-negative rates in the first round were 10.7% and 4.5% for radiographic and endoscopic screening, respectively, and the false-positive rates in the first round were 14.4 and 14.9%, respectively . Notably, EGC is easily missed during screening, even when it is performed by qualified endoscopists [18,19,20]. Despite its inability to visualize the target, the EGC index developed here could be used as an alternative non-invasive screening modality for the detection of EGC. In the validation set of our study, the AUC for the EGC index was 0.998 (95% C.I. 0.995–1.000), with a sensitivity of 0.996 (95% C.I. 0.991–1.000) and a specificity of 0.953 (95% C.I. 0.938–0.969). When we consider the prevalence of gastric cancer as 0.742% in a screening population in Japan according to Hamashima et al. , the positive predictive value (PPV) and negative predictive value were 0.138 (95% C.I. 0.103–0.182) and 1.00 (95% C.I. 0.999–1.00), respectively. The PPV of the EGC index was higher than those of the endoscopic screening (0.055) and the radiographic screening (0.031) .
Helicobacter pylori infection is a well-known risk factor for gastric cancer, and intestinal metaplasia is one of the most common pre-cancerous lesions that may lead to the disease . A combination of detecting serum antibodies against H. pylori and measuring the level of serum pepsinogens is a method of screening for EGC [22, 23]. Although this combination method is non-invasive, it is designed for the risk stratification of gastric cancer rather than its detection. The Japanese Guidelines for Gastric Cancer Screening does not recommend the combination method for population-based screening because there is insufficient evidence that it reduces the mortality of gastric cancer . The EGC index could be utilized not as a risk stratification method, but as a sensitive screening modality with high specificity. The use of non-invasive diagnostic biomarkers could contribute to the detection of EGC detection and hence improve medical management of the disease.
A variety of serum or plasma miRNAs are frequently upregulated or downregulated in gastric cancer . Although several groups have investigated the use of serum miRNAs to detect gastric cancer and predict the recurrence and prognosis of the disease [25,26,27]. To our knowledge, our current study is the largest cohort analysis of the use of serum miRNAs to detect EGC with the highest sensitivity and specificity reported to date. So et al. recently developed a clinical assay for the detection of gastric cancer based on a 12-miRNA Biomarker panel with AUCs of 0.93 and 0.92 in the discovery and verification cohorts, respectively. Although the sample size in the training set was smaller and AUC was lower in their study than in our study, the 12-miR assay was validated and cost-effectiveness was analyzed in a large prospective validation cohort consisting of 5282 participants .
The EGC index developed here includes four miRNAs: miR-4257, miR-6875-5p, miR-187-5p, and miR-5739. The serum levels of miR-4257 and miR-187-5p were higher in the EGC samples than in the non-cancer control samples, whereas the level of miR-6785-5p was lower in the EGC samples than in the control samples. Although the levels of miR-5739 were comparable in the EGC and control samples (AUC 0.463), AUC of the EGC index was higher when this miRNA was included than when it was excluded (Supplementary Table 2). Although the roles of these miRNAs in carcinogenesis remain unclear, some previous reports support the results of our study. Notably, miR-187-5p has already been described as a serum biomarker for the early detection of gastric cancer . In a study by Wang et al., the expression level of miR-187-5p was significantly lower in diffuse-type gastric cancer tissue than in normal gastric tissue , suggesting that the damaged gastric tissue surrounding the cancer site could release miR-187-5p. Notably, exosomes derived from normal gastric epithelial cells function to inhibit the progression of gastric cancer [30, 31]. miR-187-5p has a tumor-suppressive effect in non-small-cell lung cancer ; therefore, the active release of miR-187-5p from the tumor microenvironment might play a role in suppressing tumor growth. It was difficult to explain why the miR-187-5p level was higher in the EGC samples than in the non-cancer control samples in this study. Furthermore, Shuai et al. reported that miR-6785-5p suppresses tumor growth by targeting BCL2  and demonstrated that the long non-coding RNA MNX1-AS1, which is highly expressed in gastric cancer tissue, can suppress the function of miR-6875-5p in gastric cancer cells. This mechanism could possibly explain why the serum levels of miR-6875-5p were lower in the EGC samples than in the control samples in the present study. We were unable to find any publications related to the roles of miR-4257 and mir-5739 in gastric cancer; therefore, further studies of their functions and methods of regulation are warranted.
It is important to analyze the diagnostic performance of the EGC index in advanced gastric cancer and other malignancies for further discrimination of EGC. Exploratory data were analyzed in ten patients with serum samples of Stage III locally advanced gastric cancer, 50 esophageal squamous cell carcinomas (ESCC), and 50 colorectal cancers (CRC). The AUCs of the EGC index in Stage III gastric cancer, ESCC, and CRC were 1.00, 0.640, and 0.440, respectively (Supplementary Fig. 1A, B). Further discrimination models should be established in a large-scale cohort study.
Our current study has several limitations. First, the study was a retrospective analysis using archival samples, and an external validation cohort for patients with EGC was not available. Although the reproducibility of 3D-Gene® in the diagnostic index of prostate cancer was reported previously by our study group , further investigations are warranted to confirm the reproducibility of the EGC index. Second, variations in sample collection and storage may have influenced the EGC index because samples were collected from three different institutions. To confirm that the EGC index can discriminate not only external controls but also internal controls, our control set included serum samples from patients with benign diseases from the National Cancer Center Hospital. Third, although we performed a comprehensive analysis of miRNAs using age- and gender-matched EGC and non-cancer control samples, we were not able to evaluate other well-known risk factors of gastric cancer, such as H. pylori infection, atrophic gastritis, and smoking, which could have influenced the levels of circulating miRNAs [22, 34], because the data of these risk factors were unavailable owing to the retrospective nature of data collection in this study. To overcome these limitations, we have recently conducted a prospective confirmatory study using serum samples from multiple institutions.
In conclusion, the novel combination of serum miRNAs comprising miR-4257, miR-6785-5p, miR-187-5p, and miR 5739 could be a useful diagnostic biomarker to detect EGC with high accuracy.