Clinical and body composition parameters as predictors of response to chemotherapy plus PD-1 inhibitor in gastric cancer.

Abstract

Background: Predicting the treatment efficacy of programmed cell death protein 1 (PD-1) inhibitors is crucial for guiding optimal treatment plans and preventing unnecessary complications for cancer patients. We aimed to develop a prediction model using clinical and body composition parameters to identify gastric cancer (GC) patients who would respond to chemotherapy plus PD-1 antibody.

Methods: Clinical data of GC patients treated with chemotherapy plus PD-1 antibody (immunotherapy cohort, n = 120) or chemotherapy alone (chemotherapy cohort, n = 82) following surgical resection were reviewed as the training set. Patients treated with chemotherapy plus PD-1 antibody at an external center were included as the validation set (n = 43). Tumor regression grade (TRG) was recorded and classified as TRG0/1 or TRG2/3 during analysis. Body composition parameters were assessed on computed tomography images at the third lumbar vertebral level using the SliceOmatic software. Univariate and multivariate analyses were performed to identify parameters associated with TRG0/1, and then a logistic regression model was developed to stratify patients into the good and poor response groups.

Results: In the training set, clinical and body composition parameters between the immunotherapy cohort and chemotherapy cohort were similar. Skeletal muscle radiation attenuation (SMRA), neutrophil-to-lymphocyte ratio (NLR), and weight loss were associated with TRG0/1 in the immunotherapy cohort. Subcutaneous adipose tissue index (SATI) and metastasis were identified in the chemotherapy cohort. A logistic regression model was developed to stratify immunotherapy cohort patients into two response groups with an area under the receiver operating characteristic curve (AUC) value of 0.728. In the immunotherapy cohort, patients stratified as good responders showed a higher TRG0/1 rate (37/55, 67.3%) than poor response patients (18/65, 27.7%, p < 0.001) and had better overall survival (p = 0.001). In the external validation set, patients stratified using the clinical model as good responders also showed a higher TRG0/1 rate (14/18, 77.8%) than poor response patients (9/25, 36.0%, p = 0.012).

Conclusion: The prediction model consisting of SMRA, NLR, and weight loss could help identify GC patients who respond well to chemotherapy plus PD-1 antibody.