Emphasis on clinical staging has been increasing in recent years(10), and obtaining precise and reliable clinical staging is a challenging topic in the real world(11). EUS was used as a standard procedure in preoperative T staging of patients with gastric carcinomas. For GC, as EUS can visualize the different layers of the gastric wall as corresponding sonographic layers, it is generally considered an effective tool for T staging(12, 13), especially in detailed staging——early (T1) vs advanced cancers; T1a vs T1b cancers(13–17). However, in previous studies, the accuracy rate of EUS ranged from 43% to more than 90%(14, 18). In addition, several meta-analyses have also identified remarkable heterogeneous results(5, 19). Therefore, the validity of EUS for staging remains controversial.
Our previous prospective study confirmed the efficacy of EUS for T staging(6). The deepest point determined by EUS showed great consistency with pathological findings. We tried to elucidate the answer for the heterogeneity. One reason might be related to the procedure. As mentioned in previous reports, the image quality and standardization no-omitting EUS scanning may be of great significance in increasing the accuracy of EUS(9, 15). The experience of operator also affect the accuracy(20). In our center, a great number of patients with GC undergo EUS before any treatment for staging. And quality of EUS procedure and image is routinely evaluated. The other one reason might be related to the tumor. We found some interesting cases with representative discrepancies and these inconsistencies were reported. Some tumors tend to infiltratively invade the layers without destruction, which could result in incorrect pre-T staging and inappropriate treatments.
In this study, we retrospectively analyzed the data of 1736 patients, of whom 1302 patients were evaluated as having resectable tumors and underwent resection (endoscopic or surgical) as primary treatment. A comparison was made between the EUS and pathological T stages. Determining a case as early or advanced was the first step in staging tumors. Our results suggested that the specificity and positive predictive value of EUS were 96.13% and 90.94%, respectively. These findings, together with the previous studies, highlight the value of EUS in determining T staging, especially in distinguishing between early and advanced GC. Subsequently, treatment was decided. T3 and T4a are difficult to distinguish, as the serosa is significantly thin to detect invasion. However, it does not have a significant effect on treatment choices. Generally, either cT3 or cT4a is appropriate for neoadjuvant therapy before surgery.
For EGC cases, the most important issue is the accurate identification of ESD candidates which is mainly based on T staging. Therefore, we aimed to distinguish between the “fit” subgroup for EUS staging and the “un-fit” subgroup for EUS staging in EGC cases. Therefore, we performed a multivariate logistic regression analysis between the “right” and “wrong” groups. When we examined pT1a cases (uT1a and uT1b in pT1a), potential factors reported in previous studies were analyzed. Of note, lesion diameter, tumor differentiation, and ulcer showed statistical significance.
Underestimation in larger size is not surprising(9, 21–25).Interestingly, the treatment choice showed significant statistical significance in uT1a and uT1b in pT1a cases (P < 0.001). In our center, ESD specimens are evenly spread and sectioned continuously in pathological processes. However, the surgical specimen was not. The specimens could not be sectioned thoroughly. Whether the deepest point is “caught” remains uncertain, especially for large ones. This may have led to the discrepancy. To the best of our knowledge, in most clinical centers in China, the pathological process for surgical specimens is not as thorough as that for ESD specimens, because of the large size of the specimen and the shortage of pathologists. In advanced cases, pathologists find the deepest invasion part easier, but in early cases, this is not the case. Because the specimens are not spread evenly when fixed, identifying the deepest point by the naked eye seems relatively difficult.
When we examined uT1a and uT1b cases (pT1a and pT1b in uT1a, pT1a and pT1b in uT1b), ulcer status and location showed statistically significance. Ulcers might mimic tumors in some EUS images and lead to incorrect estimation, as previously reported(23, 26–28). However, in our center, early cancers with ulcers are treated with proton pump inhibitor before staging to accelerate healing(29), possibly making up for this deficiency. After healing, the ulcer subgroup showed equally or better accuracy for pretreatment staging. It is also reported, even EUS’s accuracy was poor, it was still superior to that of conventional endoscopy in ulcerative EGC(28).
According to our results, tumor location matters. If the lesion is located in the antrum, uT1a is more likely to be pT1a, and ESD is recommended. uT1b might also be pT1a eventually, which we recommend further evaluation or diagnostic ESD. When a lesion in the upper stomach is found to be uT1b, surgery is recommended; otherwise, we could perform diagnostic ESD or further investigation. Similar findings are also mentioned by some other scholars in that lesions located in the upper stomach tend to be underestimated (9, 25, 30). Based on this finding, we proposed a decision-making algorithm, as shown in Fig. 4. This finding is consistent with those of previous reports and our clinical impressions.
In recent years, patients with locally advanced cancers mostly receive neoadjuvant therapy before resection in China, so we cannot compare the pathological result with EUS in this new era. But treatment choice and response assessment are of unprecedented importance in this new era, thus we need a reliable tool for detailed staging. Also as mentioned in a previous study, EUS may provide a highly informative assessment of gastric wall invasion(11). Therefore, we selected a series of consecutive patients during the period from 2015 to 2017. In our data, the proportion of advanced GC is remarkable. For patients with early cancers, we present a large proportion who underwent surgery. However, this study has limitations in that only a single center’s data were analyzed in this study and no mini-probe EUS was used. Additionally, this was a retrospective study, and prospective studies should be conducted in the future to validate or revise this potential algorithm.