XGBoost algorithm constructs a new model to predict the nutritional risk of multiple myeloma

Purpose Elderly patients of multiple myeloma (MM) were affected by frailty status, nutritional status, comorbidities, and other factors that influence disease prognosis. Currently, there were insufficient evidence to determine whether nutritional status like NRS2002 can as a prognostic factor in elderly patients of MM. The aim of this study was to assess the prognostic value of the NRS2002 in elderly MM and to develop a new model to predict nutritional risk. Methods We retrospectively investigated multiple clinical variables, including the NRS2002 score, age, sex, body mass index (BMI),Ability of daily living (ADL), M protein type, International Staging System (ISS) stage, Durie-Salmon (DS) stage, and blood test with newly diagnosed MM. Results The PFS and OS were significantly better in the group without nutritional risk (NRS2002<3) than those with nutritional risk (NRS2002 ≥ 3) (PFS: p value = 0.029<0.05; OS: p value = 0.021<0.05). The age in the group with nutritional risk was higher than the group without nutritional risk(63 vs 70, p value = 0.000<0.05). Compared with NRS2002, the specificity of the new model with Age, BMI, ADL, β2-microglobulin(β2-M) and Albumin(ALB) for nutritional risk prediction was increased by 10%.