Predictive model for the diagnosis of ectopic Cushing syndrome

The purpose of our study was to develop and externally validate a multivariable diagnostic model to distinguish ectopic ACTH secretion (EAS) from Cushing's disease (CD) using routine clinical parameters. The model was derived from a Spanish multicenter retrospective cohort and externally validated in a Colombian cohort. Predictors were selected through a multivariable logistic regression model with penalized logistic regression (LASSO) using the Spanish cohort. Discriminative performance was assessed using the area under the ROC curve (AUC) and calibration using the slope, the origin, and the Brier score. The derivation cohort included 253 patients from Spain (199 with CD and 54 with EAS). The external validation cohort comprised 72 Colombian patients (53 with CD and 19 with EAS). In the derivation cohort, multivariable modeling identified four independent predictors: ACTH concentration, 24-h urinary free cortisol (UFC), serum potassium, and maximum pituitary tumor diameter. The final model demonstrated good discrimination in the derivation cohort (mean AUC of 0.987), with excellent calibration (calibration slope ranged from 0.998 to 5.22, intercept -0.2 to 0.99, Brier score 0.035). In external validation, model performance remained robust (AUC ranging from 0.9885 to 0.9890; Brier score 0.0381). The model achieved a sensitivity of 95.3%, specificity of 93.6%, positive predictive value of 82.3%, and negative predictive value of 98.7% for detecting EAS. Thus, the diagnostic model developed, based on routinely available clinical variables, shows high accuracy and calibration in differentiating between EAS and CD, supporting its potential use in diverse clinical settings and integration into diagnostic workflows.