Cross-sectional and longitudinal associations between PG-SGA criteria with survival of cancer patients.

Abstract

e23277 Background: Malnutrition is prevalent in cancer patients and is associated with treatment efficacy and clinical outcomes. Whether the available malnutrition assessment tools contain optimal survival prediction value for cancer patients, especially from developing countries, remains unclear. Methods: In this retrospective cohort study, cancer patients were selected from our center between Nov 2019 and Sep 2023. According to the PG-SGA criteria, PG-SGA scores < 4 referred to well-nourished, ≥ 4 PG-SGA scores < 9 referred to malnutrition, and ≥ 9 referred to severe malnutrition. Kaplan-Meier (KM) curves were used to estimate the survival patterns of cancer patients with varied levels of PG-SGA scores. The Cox proportional hazard model was used to evaluate the association between baseline and longitudinal changes of PG-SGA scores with the all-cause mortality of cancer patients. Subgroup analysis with interaction tests was used to determine the high-risk subgroups. A series of sensitive analyses were performed to validate the findings we determined. Results: There were 1,415 cancer patients included in our study, with a mean age of 46 years old. The mean PG-SGA score was 2.42 in non-malnourished patients, 10.10 in malnourished patients, and 12.65 in severely malnourished patients, respectively. The KM curves showed that cancer patients with severe malnourished status (PG-SGA ≥9) presented the lowest short- and long-term survival probability, compared with the other two subgroups. A linear trend was observed between the PG-SGA scores and cancer mortality (p for overall test < 0.001). The Cox regression analysis showed that baseline malnourished status was significantly associated with the survival of cancer patients [≥4 PG-SGA < 9: hazard ratio (HR): 1.46, 95% confidence interval (CI): 1.09 – 1.96, p = 0.012; PG-SGA ≥9: HR = 1.78, 95%CI: 1.34 – 2.37, p < 0.001]. Cancer patients with longitudinal increased PG-SGA scores ( > 2 points) were observed to have a high risk for mortality (HR = 1.69, 95%CI: 1.04 – 2.74, p = 0.033). Compared with longitudinal changes in PG-SGA scores, the baseline PG-SGA scores showed significantly higher predictive power in identifying the risk subgroup (concordance index: 0.646 vs. 0.586, p < 0.001). PG-SGA scores showed more pronounced predicting value among subgroups with features of younger age, normal and underweight body mass index, late cancer stage, without history of diabetes, smoking, radiotherapy, and immunotherapy. A series of sensitive analyses determined consistent correlations. . Conclusions: Our study validates the clinical value of baseline, as well as longitudinal changes of PG-SGA, scored status for predicting the survival of cancer patients. Our findings provide new insights into the identification and prevention strategies for malnourished conditions in cancer patients, resulting in improved cancer-related nursing in resource-limited regions.