A prediction model for reactivation of Langerhans cell histiocytosis based on machine-learning algorithms.

Abstract

Langerhans cell histiocytosis (LCH) is a rare inflammatory myeloid neoplasm characterized by the clonal proliferation of myeloid progenitor cells. The reactivation rate of LCH exceeds 30%. However, an effective prediction model to predict reactivation is lacking. To select potential prognostic factors of LCH and construct an easy-to-use predictive model based on machine-learning algorithms. Clinical records of LCH inpatients in the Second Xiangya Hospital of Central South University, from 2008 to 2022, were retrospectively studied. Seventy-six patients were classified into a reactivated/progressive group or a stable group. Clinical features and laboratory outcomes were compared, and machine-learning algorithms were used for building prognostic prediction models. Clinical classification (single-system LCH, multiple-system LCH, and central nervous system/lung LCH), level of anemia, bone involvement, skin involvement, and elevated monocyte count were the best performing factors and were finally chosen for the construction of the prediction models. Our results show that the above-mentioned five factors can be used together in a prediction model for the prognosis of LCH patients. The major limitations of this study include its retrospective nature and the relatively small sample size.