Using a multi-institutional retrospective dataset, we investigated the oncological similarities between large types 3 and 4 tumors. Significant differences in patient age, prognosis, and initial recurrence patterns were observed between the two groups of patients. However, these oncological outcomes were similar when the comparisons were made between large type 3 tumors with undifferentiated phenotype and type 4 tumors.

Type 4 tumors have the poorest prognosis among all macroscopic tumor types [6] and are closely associated with peritoneal dissemination [22]. Type 3 tumors are the most common macroscopic type of resectable advanced gastric cancer [6], and large type 3 tumors often present with peritoneal dissemination and positive lavage cytology [23, 24]. Clinical trials have been conducted to develop treatments targeting these high-risk macroscopic cancer types [12, 13, 25, 26], and challenges persist in developing intensive perioperative chemotherapy regimens [14]. The efficacy of perioperative chemotherapy reportedly differs based on macroscopic or histological types [15, 16]; however, concerns have been raised as to whether it is appropriate to develop such a treatment strategy with the conventional target population consisting of large type 3 and type 4 tumors. Our study showed significant differences in prognosis and disease recurrence patterns between the large type 3 and type 4 tumors in a full cohort analysis. Type 4 tumors were associated with a shorter OS and a higher peritoneal disease recurrence rate, whereas large type 3 tumors had a higher lymph node disease recurrence rate. These results suggest that large type 3 and 4 tumors may have different oncological characteristics and mechanisms of progression. We observed a significant difference in prognosis between the differentiated and undifferentiated histological types in the large type 3 tumors. Therefore, we designated large type 3 tumors with undifferentiated phenotype and type 4 tumors as high-risk macroscopic types and conducted further analyses of survival and disease recurrence patterns. After adjusting for the patient background and treatment course, we concluded that there was an oncological similarity between large type 3 with undifferentiated phenotype and type 4 tumors, although some differences remained.

We found that differentiated large type 3 tumors have a notably better prognosis. The reason for this improved prognosis remains unclear. Given the rarity of differentiated large type 3 tumors and limited literature on this subgroup, we did not explore the underlying factors contributing to this outcome. However, based on our findings, we suggest that differentiated large type 3 tumors should not be grouped with type 4 tumors as a novel treatment strategy.

The macroscopic types of gastric cancer are defined by the JGCA guidelines and are considered one of the essential factors to be recorded. In addition, the discrimination between macroscopic types is not always an easy task. For example, some type 4 tumors are difficult to distinguish from the superficial tumor type, whereas others are classified as type 5 tumors. Similarly, large type 3 ulcers range from large ulcers (similar to type 2 ulcers) to diffuse infiltration with central ulcers (similar to type 4). In this study, a multi-institutional retrospective analysis was conducted, and the macroscopic type classification was performed at each institution without a central review. However, a central review of the macroscopic types is not routinely performed, even in prospective clinical trials. Large type 3 tumors with undifferentiated phenotype can be objectively classified based on the information obtained before treatment. In an attempt to identify effective treatments using large type 3 tumors and type 4 tumors as targets, it may be possible to eliminate bias by including histology in the eligibility criteria for clinical trials so that large type 3 tumors with differentiated phenotype and significantly more favorable outcomes can be excluded.

Surgery followed by postoperative adjuvant chemotherapy with S-1 monotherapy for 1 year was the standard treatment for the patient population included in this study. In addition, some patients have undergone neoadjuvant chemotherapy as part of clinical trials or practice. A subgroup analysis revealed that patients with undifferentiated large type 3 tumors had shorter OS than that of those with type 4 tumors among patients who received neoadjuvant chemotherapy, which is consistent with the findings of the JCOG0210 trial [12]. We found that the proportion of patients with a poor pathological response (grade 0, 1a) to neoadjuvant chemotherapy was higher in the group with undifferentiated large type 3 tumors than that in the group with type 4 tumors. The difference in response to neoadjuvant chemotherapy between the two groups of patients may have influenced the prognosis. Our results suggest the need for even more intensive neoadjuvant treatment to target these high-risk macroscopic cancers.

Surgical outcomes showed significant differences, with longer operative times and more postoperative complications in the group with undifferentiated large type 3 tumors than in those with type 4 tumors. The differences in tumor shrinkage due to neoadjuvant chemotherapy might have influenced the difficulty of surgery because patient characteristics, tumor size, and gastrectomy type were well balanced by propensity-score matching. However, all surgeries in our study were performed using open gastrectomies, and it is unclear whether the results would be similar in the era of minimally invasive surgery.

Our study included a relatively high number of differentiated type 4 tumors (12%), which increased to 15% after propensity score matching. This might be influenced by the multi-institutional nature of the study and the lack of a central review of macroscopic classification. Currently, there is no conclusive evidence regarding the impact of differences in histological phenotypes on the prognosis of patients with type 4 tumors. To address this, we stratified patients after propensity score matching into three groups, namely, differentiated type 4, undifferentiated type 4, and undifferentiated large type 3, and analyzed their survival outcomes. The survival curves did not show a significant difference between the differentiated and undifferentiated type 4 groups (Online Resource 3), suggesting that the high ratio of the differentiated phenotype in type 4 did not significantly affect the overall outcomes.

Our study had some limitations. First, although we performed propensity-score matching to eliminate the effects of confounding factors, differences in perioperative treatment between the two groups of patients could not be eliminated completely. Moreover, propensity-score matching reduced the sample size in both groups. Second, molecular analyses, which could provide new insights into the optimization of treatment strategies and improvement of patient outcomes, were not performed. Third, the current standard postoperative adjuvant chemotherapy for stage III gastric cancer is S-1 +  docetaxel, which was not delivered in a majority of patients in the current study. Fourth, owing to insufficient information on the histological types at preoperative biopsy, the histological types obtained from the resected specimens were used for analysis. Fifth, the classification of large type 3 tumors is unique to the Japanese study group and may not be well understood, particularly in Western countries. Since this study aimed to determine whether large type 3 and type 4 gastric cancers should be treated as a single group and not redefine the criteria for large type 3 tumors, establishing a cut-off value for large type 3 tumors was beyond the scope of this study. Future research should investigate and validate the appropriate cut-off value for large type 3 tumors to ensure the applicability of our findings across different clinical settings. Finally, resectable large type 3 and 4 gastric cancer tumors are rare and account for approximately 5% of all resectable gastric cancers. Although their specific natures may limit its usefulness in clinical practice, the analysis of large-scale multi-institutional datasets is considered valuable.

In conclusion, large type 3 tumors with undifferentiated phenotype and type 4 tumors have oncological similarities in patients with resectable gastric cancer. When developing the same novel treatment strategy for type 4 and large type 3 tumors, avoiding large type 3 tumors with differentiated phenotype may lead to less-biased results.