Usual nutritional scores have acceptable sensitivity and specificity for diagnosing malnutrition compared to GLIM criteria in hemodialysis patients

Diagnosing malnutrition by the recently published Global Leadership Initiative on Malnutrition (GLIM) criteria requires using modern techniques for body composition measurements. We hypothesized that the prevalence of malnutrition identified by usual nutritional scores and according to GLIM criteria may be close to each other due to the number of components shared between them. Our aim was to compare the concurrent validity of four nutritional scores, malnutrition-inflammation score (MIS), objective score of nutrition on dialysis, geriatric nutritional index (GNRI), and nutritional risk index against the GLIM criteria for malnutrition in maintenance hemodialysis patients. This prospective observational study was performed on 318 maintenance hemodialysis outpatients (37% women) with a mean age of 68.7 ± 13.1 years and a median dialysis vintage of 21 months. According to the GLIM criteria, 45.9% of these patients were diagnosed with malnutrition. Nutritional scores, dietary intake and body composition parameters were measured. All nutritional scores showed a strong association with malnutrition in multivariable logistic regression models. In discriminating the nutritional risk, the ROC AUC was largest for GNRI (0.70, 95% CI: 0.65-0.75; P< .001). Nutritional risk index and MIS showed high specificity but lower sensitivity compared to GNRI and objective score of nutrition on dialysis. Compared to MIS, GNRI had better concurrent validity (higher sensitivity and acceptable specificity) but was inferior to MIS in terms of relation to certain etiologic and phenotypic components of the GLIM criteria (specifically, to dietary intake and decrease in dry weight). In summary, of the nutritional scores tested, GNRI is the most sensitive score in identifying malnutrition diagnosed by GLIM criteria, but MIS is more specific and better in predicting the individual components of the GLIM criteria.