RESUMO
BACKGROUND: The present study aims to assess the interrater reliability of the Global Leadership Initiative on Malnutrition (GLIM) criteria, a framework to provide a consensus diagnosis of malnutrition. We also aimed to investigate its concurrent and predictive validity in the context of patients with cancer admitted to the intensive care unit (ICU). METHODS: Individuals aged ≥19 years with cancer who were admitted to the ICU within 48 h of their initial hospital admission were included. Nutrition status was assessed with the Nutritional Risk Screening 2002, the Subjective Global Assessment (SGA), and the GLIM criteria. Interrater reliability was assessed by the kappa test (>0.80). The SGA served as the established benchmark for assessing concurrent validity. To evaluate predictive validity, the occurrence of mortality within 30 days was the outcome, and Cox regression models were applied. RESULTS: A total of 212 patients were included: 66.9% were at nutrition risk, and 45.8% were malnourished according to the SGA. According to the GLIM criteria, 68.4% and 66% were identified as malnourished by evaluators 1 and 2, respectively (κ = 0.947; P < 0.001). The GLIM combination incorporating weight loss and the presence of inflammation exhibited sensitivity (82.4%) and specificity (92%). In the multivariate Cox regression models, most GLIM combinations emerged as independent predictors of complications. CONCLUSION: The GLIM criteria demonstrated satisfactory interrater reliability, and the combination involving weight loss and the presence of inflammation exhibited noteworthy sensitivity and specificity. Most GLIM combinations emerged as independent predictors of 30-day mortality.