In this study, involving two large cohorts of gastric cancer patients treated with either surgery alone (D1/D2 study) or with surgery and perioperative treatment (CRITICS trial), the concordance in histological subtype between pretreatment biopsies and resection specimens was investigated. The histological subtype is a commonly used stratification factor for clinical trials involving GC patients. However, the impact of preoperative systemic treatment on the different histological subtypes of GC is yet undetermined. To our knowledge, this is the first study to evaluate concordance between biopsies and resection specimens in patients treated with preoperative chemotherapy. We found an overall concordance in histological subtype of 72% in the surgery alone cohort and 74% in the chemotherapy cohort. This similarity in concordance rates indicates that the impact of systemic therapy on the histological subtype is negligible and that both pretreatment biopsies and surgical resection specimens can be used for the determination of the histological subtype in gastric cancer patients.

Histological subtypes of gastric cancer are widely recognized to represent distinct biological and clinical entities with regard to treatment response and prognosis, and can therefore be useful as prognostic markers [9, 16, 22]. The potential utility of histological subtype as a predictive marker to guide therapeutic decisions, however, remains a matter of debate. Although new therapies with corresponding molecular markers, such as HER2 for trastuzumab and PD-L1 expression for immune checkpoint inhibitors, show promise, they only apply for approximately 22% and 16% of patients, respectively [23, 24]. This means that personalized treatment options are as yet unavailable for the majority of GC patients. Meanwhile, the Lauren classification is readily available at low clinical costs. Therefore, accurate determination and stratification by histological subtype in clinical trials will remain crucial until novel markers arise.

Historically, we relied on pretreatment biopsies for stratification in clinical trials since the impact of chemotherapy on histological subtype was yet unknown. Our study, however, shows that chemotherapy has a negligible impact on histological subtype. Clinical trials in which patients are postoperatively randomized between treatment arms, for example, the VESTIGE trial that randomizes between chemotherapy and immunotherapy after surgical resection, could therefore stratify patients for histological subtype assessed on either the biopsy or resection specimen [25]. This could be relevant in case the biopsy specimen is not available. Similarly, studies investigating novel treatment modalities in patients with metachronous metastases can use histological subtype as assessed on the resection specimen for stratification. Since more material is available for evaluation in the resection specimen and it provides more insight into intratumor heterogeneity, we believe that histological subtype assessment for stratification in clinical trials does not have to be confined to pretreatment biopsies.

Our results are in line with those of previous studies that reported an overall concordance in histopathological subtype between biopsies and resection specimens of 65–75% [12,13,14]. A substantially higher concordance rate of 83–86% was seen in the diffuse subtype. This is in line with the study of Piessen et al., which showed a higher concordance of 90% in signet ring cell carcinomas [15]. As signet ring cells are often present in diffuse type GC, it is possible that the presence of signet ring cells also has contributed to the higher level of concordance in diffuse type, compared to the intestinal type in our study. In the mixed type, the limited amount of tissue available in pretreatment biopsies may have hampered the accurate distinction of this phenotype from the intestinal and diffuse subtypes, consequently resulting in low concordance. This finding is consistent with Flucke et al. who showed similar concordance rates of 79% in intestinal, 86% in diffuse, and 56% in mixed histological subtypes in a German cohort of 100 primary treatment-naive GC patients [12]. Likewise, a Chinese study with 116 pairs of biopsy and surgery samples reported that most discrepancies in histological subtyping could be attributed to the mixed subtype [14]. Taken together, these overall levels of concordances are suboptimal as the histological subtype is a commonly used stratification factor in clinical trials.

The number of (tumor-positive) biopsies and the tumor regression grade did not significantly impact the concordance of the histological subtype in our study. It is noteworthy that two tumors in the surgery-alone cohort exhibited features associated with near-complete tumor regression, despite not having undergone any preoperative treatment. This phenomenon of tumor regression-like changes in treatment-naïve GCs has previously been reported in a post-hoc analysis of the MAGIC trial, where 3 patients treated with surgery alone showed a near-complete pathological response as well [21]. Interestingly, both tumors in our study were restricted to the mucosal layer and showed fibrotic tumor regression-like changes in the submucosal layer. These findings may highlight the challenges that can be encountered in accurately assessing tumor regression in early-stage GCs.

It is noteworthy that several clinical trials using the Lauren classification assessed on pretreatment biopsies as a stratification factor reported that in approximately 30% of gastric cancers the histological subtype was not known.[9, 11]. Since the histological subtype of all included pretreatment biopsies for central review in our study could be determined, we believe this is most likely a matter of underreporting rather than of infeasibility to determine the histological subtype. As histological subtypes represent distinct biological and clinical subgroups, we would strongly recommend the (central) evaluation of histological subtype on pretreatment biopsies in clinical trials.

Nonetheless, it is possible that “other” histological subtypes that do not fit in Lauren classification [1] may have contributed to the high proportion of tumors of unknown Lauren classification in several clinical trials. The “other” histological subtype was first mentioned as a distinct category by Smyth et al. in 2020 and was diagnosed in 12% and 8% of cases in the D1/D2 study and CRITICS trial. Flucke et al. showed a higher concordance rate between biopsies and resection specimens using the more comprehensive WHO classification (distinguishing, e.g., the mucinous subtype) compared to the Lauren classification (84% vs. 74%, respectively)[12]. It supports the notion that concordance rates potentially improve with recognition and scoring of the “other” subtype. Further recognition of the “other” histological subtype should therefore be encouraged.

A limitation of this study is that interobserver agreement in GC histological subtyping could not be analyzed due to the design with central pathology review in the D1/D2 study and CRITICS trial. Interobserver variation could affect concordance rates. In previous studies, interobserver agreement varied between 70 and 90% [13, 26, 27]. Concordance rates in the surgery alone cohort, however, are similar to previous studies [12,13,14,15]. It is also important to acknowledge that the preferred chemotherapeutic regimen has shifted from an anthracycline-based treatment (as used in the CRITICS trial) to a taxane-based treatment following the publication of the results of the FLOT4-AIO trial [9]. Whether taxane-based treatment could potentially affect histological subtype remains therefore unclear.

In conclusion, accurate histological subtype determination on biopsies of gastric cancer is fairly good for diffuse-type GC, but suboptimal for intestinal GC and even less optimal in the mixed and “other” subtypes. We do however demonstrate that the impact of anthracycline-based preoperative chemotherapy on histological subtype in resection specimens is negligible. We propose that the assessment of histological subtype in clinical trials should not necessarily be confined to pretreatment biopsies, but can also be determined on resection specimens after preoperative chemotherapy.