The CuPeR model: A dynamic online tool for predicting Cushing's disease persistence and recurrence after pituitary surgery.

OBJECTIVE

Predicting postoperative persistence and recurrence of Cushing's disease (CD) remains a clinical challenge, with no universally reliable models available. This study introduces the CuPeR model, an online dynamic nomogram developed to address these gaps by predicting postoperative outcomes in patients with CD undergoing pituitary surgery.

METHODS

A retrospective cohort of 211 patients treated for CD between 2010 and 2024 was analyzed. Key patient and tumor characteristics, imaging findings, and treatment details were evaluated. Multivariate logistic regression identified independent predictors of postoperative persistence or recurrence of CD (PoRP-CD), which were then incorporated into the CuPeR model using stepwise selection based on Akaike Information Criterion. Internal validation was performed using a testing dataset, and a user-friendly online nomogram was developed to facilitate immediate, patient-specific risk estimation in clinical practice.

RESULTS

The final predictive model identified four key factors: symptom duration, MRI Hardy's grade, tumor site, and prior pituitary surgery. Longer symptom duration and a history of prior surgery significantly increased the risk of recurrence, while bilateral tumor location reduced this risk. The model demonstrated an area under the receiver operating characteristic curve (AUC-ROC) of 0.70, with 83% accuracy, specificity of 96%, and sensitivity of 33%.

CONCLUSIONS

The CuPeR model may offer a practical tool for predicting PoRP-CD, enhancing preoperative decision-making by providing personalized risk assessments.