Fan Y, Li Y, Bao X, et al. - Researchers sought to generate and validate machine learning (ML) models that will allow prediction of delayed remission (DR) in non-immediate remission (IR) patients with Cushing disease (CD). A total of 201 CD patients were enrolled and divided randomly into training and test datasets. Among these patients, 88 (43.8 %) had DR. Rate of DR were lower among patients who were younger, had low BMI, Knosp grade III–IV and a tumor not found by pathological examination. After recursive feature elimination (RFE) feature selection, the Adaboost model, which incorporated18 features, exhibited the highest discriminatory ability; It had significantly better predictive ability than using Knosp grade and postoperative immediate morning serum cortisol (PoC). Results from permutation importance and local interpretable model–agnostic explanation (LIME) algorithms indicated that the most important features were preoperative 24-hour urine free cortisol, PoC and age and showed the reliability and clinical practicability of Adaboost model in DC prediction. The findings are suggestive of the effectiveness of ML-based models as a non-invasive approach to prognosticate DR, and their possible utility in determining individual treatment and follow-up strategies for CD patients.