Session Item

A substantial proportion of gastric cancer patients develop treatment-related toxicity. Since these side effects have a significant impact on health-related quality of life, the avoidance or mitigation of toxicity development is critical, hence optimal strategies that address this need, are warranted. The primary aim of this study was to identify clusters of locally advanced non-metastatic gastric cancer patients with distinct toxicity profiles and assess the prognostic potential of these clusters early during the first course of preoperative chemotherapy within the CRITICS study. Subsequently, we assessed whether membership of the identified clusters was associated with (1) treatment compliance and (2) with overall (OS) and event-free (EFS) survival.

For this study, we included all patients who participated in the CRITICS trial (NCT00407186) and who started the first course of preoperative chemotherapy. Toxicity was assessed according to the National Cancer Institute Common Toxicity Criteria for Adverse Events (CTCAE; version 3). Agglomerative hierarchical cluster modelling was used to identify clusters of patients with distinct toxicity profiles during the first chemotherapy course. For this, we randomly split the data into a training (80%) and a test (20%) set. Cox proportional hazards analysis was performed to assess the association between class membership and EFS and OS.

In this study 689 of 788 patients (87%) in the CRITICS study were included. A four-cluster model with four distinct clusters (i.e. mild toxicity (blue), neutropenia toxicity (green), poly toxicity (red), and gastrointestinal toxicity (orange)) had the best model fit (Figure 1). The percentage of patients completing the preoperative treatment was highest in the neutropenia toxicity cluster (95%) and lowest in the poly toxicity cluster (76%; p=0.003). Membership of the different clusters was significantly associated with OS: patients in the neutropenia toxicity cluster had the longest median OS, with 58 months (95% confidence interval (CI) 22-93) while patients within the gastrointestinal toxicity cluster had the shortest median OS of 22 months (95% CI 6-38; p=0.005). This was also observed for EFS, where the neutropenia toxicity cluster had the longest median EFS of 44 months (95% CI 1-87) compared to 13 months (95% CI 8-18) in the gastrointestinal toxicity cluster (p=0.018).

In patients with locally advanced gastric cancer and treated in the CRITICS trial, we identified four distinct toxicity clusters during preoperative chemotherapy. Membership of the different clusters was highly prognostic for both EFS and OS, with very large differences between clusters and patients in the neutropenia toxicity cluster having the best EFS and OS. Therefore, this novel approach in which we identified distinct toxicity clusters already during the first chemotherapy course, could potentially play a role in clinical and shared decision-making during this intense treatment.