Development and internal validation of a prediction model for the malnutrition in lung cancer patients based on...

The present study aimed to establish a new predictive model based on mPG-SGA and compared it with the GLIM criteria to evaluate its accuracy and applicability in diagnosing malnutrition among Chinese lung cancer (LC) patients, further exploring its potential advantages in clinical practice.

The study is a prospective cohort of 354 LC patients. Blood test indexes, baseline characteristics and human body composition (BIA) were collected and analyzed. The data set was split as training and testing sets in a 7:3 ratio. Univariate and multivariate logistic regression analysis were taken step by step to select meaningful variables, which was further combined with mPG-SGA to build the model. The discrimination of the model was evaluated using R², Brier score, C-index, goodness-of -fit, net reclassification index. The model was assessed for clinical utility by decision curve analysis (DCA).

The median age of LC patients was 64 y (56, 70), and 261 (73.73 %) were male. There were 313 (88.42 %) patients with advanced clinical stage. Malnutrition was detected using GLIM and mPG-SGA tools with prevalence of 70.34 % and 57.91 %, respectively. There was no significant difference between all variables between training and testing sets. Hemoglobin was finally selected to combine with mPG-SGA as a new predictive model, with R²0.121, Brier score 0.167 in the training set and R²0.015, Brier score 0.182 in the testing set. The C-index was 0.686, 95 % CI (0.610-0.761) in the training set and 0.6197, 95 % CI (0.491-0.7483) in the testing set. Goodness-of -fit test revealed that X-squared = 1.0237, df = 2, p-value = 0.5994 in the training set and X-squared = 0.19973, df = 2, p-value = 0.905 in the testing set. Compared with GLIM, NRI for new model was 0.4052, 95 % CI (0.1226-0.6878), p-value: 0.00495. The prediction model produces greater net clinical benefit in both two sets as revealed by the clinical decision curve.

The predictive efficacy of mPG-SGA combined with hemoglobin for the nutritional status and survival period of lung cancer patients is basically equivalent to that of the GLIM criteria, indicating that this new model has the potential to be an alternative replacement for nutritional assessment among LC patients.