Background: Malnutrition is a condition with great impact in oncology patients. Poor nutritional status is often associated with increased morbidity and mortality, increased toxicity and reduced tolerance to chemotherapy, among other complications. The recent GLIM criteria for malnutrition aim to homogenize its diagnosis, considering the baseline disease status. Due to the few studies that assess the predictive capacity of these new criteria, we aimed to evaluate their performance for the prediction of complications and mortality in patients with cancer. Methods: Prospective, single-centre study. All outpatients under active treatment for head and neck, upper gastrointestinal and colorectal tumors between February and October 2020 were recruited. These patients were followed up for 6 months, assessing the occurrence of complications and survival, based on GLIM diagnosis of malnutrition. Results: We enrolled 165 outpatients 46.66% malnutrition. During 6-month follow-up, patients with malnutrition (46.7%, according to GLIM criteria) had an ̃3-fold increased risk of hospital admission (p < 0.001) and the occurrence of severe infections (considered as such those requiring hospitalization, intravenous antibiotics and/or drainage by interventional procedure) during follow-up (p = 0.002). Similarly, malnourished patients had a 3.5-fold increased risk of poor pain control and a 4.4-fold increased need for a higher dose of opioids (both p < 0.001). They also had a 2.6-fold increased risk of toxicity (p = 0.044) and a 2.5-fold increased likelihood of needing a dose decrease or discontinuation of cancer treatment (p = 0.011). 6-month survival of malnourished patients was significantly lower (p = 0.023) than non-malnourished patients. Conclusions: Diagnosis of malnutrition according to GLIM criteria in oncology patients on active treatment predicts increased complications and worse survival at 6-month follow-up, making them a useful tool to assess the nutritional status of oncology patients.

OR: Odds ratio and 95% confidence intervals. p: p-value in logistic regression