Nutritional features-based clustering analysis as a feasible approach for early identification of malnutrition in...

Malnutrition is prevalent that can impair multiple clinical outcomes in oncology populations. This study aimed to develop and utilize a tool to optimize the early identification of malnutrition in patients with cancer.

We performed an observational cohort study including 3998 patients with cancer at two teaching hospitals in China. Hierarchical clustering was performed to classify the patients into well-nourished or malnourished clusters based on 17 features reflecting the phenotypic and etiologic dimensions of malnutrition. Associations between the identified clusters and patient characteristics were analyzed. A nomogram for predicting the malnutrition probability was constructed and independent validation was performed to explore its clinical significance.

The cluster analysis identified a well-nourished cluster (n = 2736, 68.4%) and a malnourished cluster (n = 1262, 31.6%) in the study population, which showed significant agreement with the Patient-Generated Subjective Global Assessment and the Global Leadership Initiative on Malnutrition criteria (both P < 0.001). The malnourished cluster was negatively associated with the nutritional status, physical status, quality of life, short-term outcomes and was an independent risk factor for survival (HR = 1.38, 95%CI = 1.22-1.55, P < 0.001). Sex, gastrointestinal symptoms, weight loss percentages (within and beyond 6 months), calf circumference, and body mass index were incorporated to develop the nomogram, which showed high performance to predict malnutrition (AUC = 0.972, 95%CI = 0.960-0.983). The decision curve analysis and independent external validation further demonstrated the effectiveness and clinical usefulness of the tool.

Nutritional features-based clustering analysis is a feasible approach to define malnutrition. The derived nomogram shows effectiveness for the early identification of malnutrition in patients with cancer.