Development and validation of a prognostic model incorporating patient reported outcomes for advanced gastric and esophageal carcinoma (AGOC) using individual patient data from two AGITG randomized clinical trials.

Background: We developed and validated a prognostic model incorporating readily accessible clinicopathological data and specific patient-reported outcomes (PROs).

Methods: We used data from two randomized trials comparing regorafenib to placebo: AGITG INTEGRATE IIa (n = 251) for model development and AGITG INTEGRATE (n = 152) for validation. Candidate variables were chosen from a systematic literature review and expert consultation. Significant prognostic factors in the multivariable model were identified using univariable Cox proportional hazards models with a p-value of < 0.1. Multivariable Cox proportional hazards models were developed using clinicopathological and PRO variables, with model selection refined using least absolute shrinkage and selection operator (LASSO). The model's discrimination and calibration were assessed using concordance indices (C-statistics) and calibration plots.

Results: Univariable analysis identified 9 clinicopathological variables and 4 PRO domains that were prognostic for overall survival: body mass index (BMI), ECOG performance status, number of metastatic sites, liver involvement, treatment with regorafenib, neutrophil-lymphocyte ratio (NLR), LDH, albumin, CA 19-9, appetite loss, constipation, fatigue, and pain. The initial multivariable model (M1) incorporated geographic region (Asia vs non-Asia), performance status, number of metastatic sites, treatment with regorafenib, NLR, BMI, LDH, CA 19-9, and albumin. The preferred multivariable model (M2), including the abovementioned variables plus the 4 PROs, demonstrated superior discriminative ability with higher C-statistic values than models without PROs. Plots supported the model's calibration.

Conclusions: Incorporating PROs into prognostic models for AGOC improved the accuracy of survival predictions. Further research is needed to validate its use in routine clinical practice.