Validation of the scored Patient-Generated Subjective Global Assessment Short Form as a prognostic tool for incurable cancer patients.

Abstract

The Patient-Generated Subjective Global Assessment Short Form (PG-SGA SF) is a standardized tool for assessing nutrition risk in patients with cancer. The aim of this study was to propose and validate a cutoff point for the PG-SGA SF related to the prognosis of patients with incurable cancer in exclusive palliative care. This is a prospective cohort study of patients with incurable cancer at the National Cancer Institute in Brazil. A total sample (n = 2,144) was randomly divided into groups: (1) training (n = 1,072), to determine the most accurate PG-SGA SF cutoff, and (2) validation (n = 1,072), to test the predictive accuracy of this cutoff point. The receiver operating characteristic curve was plotted to determine the best cutoff point of the PG-SGA SF related to death. Concordance statistics (C statistic) were used to test the predictive accuracy of the models. Kaplan-Meier curve and the Cox hazard model were used to verify a prognostic value of the cutoff point. PG-SGA SF score ≥15 was found to be the best cutoff based on 90-day mortality with good accuracy discrimination (C statistic ≥ 0.74). Patients whose PG-SGA SF score was ≥15 had a shorter survival of 32 (interquartile range [IQR], 12-75) vs 83 days (IQR, 31-90) (p-value < .001) and higher risk of death (hazard ratio: 2.20; 95% CI, 1.64-2.95). The proposed PG-SGA SF cutoff score is valid and, alongside its usefulness in nutrition triage, could provide prognostic value for patients with incurable cancer.