Comparative analysis of malnutrition diagnosis methods in lung cancer patients using a Bayesian latent class model.

There are no consensus criteria for malnutrition diagnosis in clinical settings, the Global Leadership Initiative on Malnutrition (GLIM) criteria were developed to facilitate global comparisons of malnutrition prevalence, interventions and outcomes. Validation to assess usefulness in clinical practice is essential, however, the imperfect nature of reference standards used in concurrent validation may result in biased estimates of diagnostic accuracy. The Bayesian latent class model (BLCM) can assess the diagnostic performance when a "gold standard" is absent. This study's objective was to assess the diagnostic performance of the GLIM criteria in comparison with the Nutritional Risk Screening 2002 (NRS-2002) and the Patient Generated Subjective Global Assessment (PG-SGA) in lung cancer patients using a BLCM. We hypothesized that the GLIM criteria are more sensitive and specific for malnutrition diagnosis in lung cancer patients.1,384 patient records retrospectively obtained from the "Investigation on Nutrition Status and its clinical outcome of common Cancers" (INSCOC) study were used to determine the prevalence of malnutrition, sensitivity (Se) and specificity (Sp) by applying a BLCM.The prevalence of malnutrition was 0.56. The sensitivity and specificity of the GLIM criteria were Se: 0.85 and Sp: 0.88; Se: 0.74 and Sp: 0.85 for NRS-2002 and Se: 0.96 and Sp: 0.89 for PG-SGA.Although the GLIM criteria were acceptable for malnutrition diagnosis, PG-SGA is superior for determining cancer-associated malnutrition. Because of its fair sensitivity, NRS-2002 was best equipped to screen out patients not at nutritional risk.