Background: Some laboratory and clinical features are associated with a probability of recurrence after transnasal adenomectomy for Cushing disease (CD). However, there is no consensus on a set of predictors. Rules for prediction of recurrence were not proposed earlier.
Aim: To develop prediction model of recurrence/remission after successful neurosurgical treatment for CD.
Methods: Retrospective single-site comparative study included 349 patients (52 men and 297 women) with a verified diagnosis of CD who underwent effective endoscopic transsphenoidal adenomectomy between 2007 and 2014. Clinical and laboratory parameters were evaluated. Laboratory tests were performed using immunochemiluminescent method. Time-to-event analysis and ROC-analysis were applied. Multivariate models were developed using logistic regression and artificial neural network (ANN).
Results: Postoperative cortisol and ACTH levels and their combinations cannot be used for prediction of recurrence. ANN for prediction of recurrence within 3 years after successful surgery was developed. Input variables are age, duration of the disease, MRI data on adenoma, morning postoperative levels of ACTH and cortisol, output variable is binary (recurrence/remission). Predictive value for remission is 93%, 95% CI [89%; 96%], and predictive value for recurrence is 85%, 95% CI [71%; 94%]. Web-calculator based on the model is developed and free for use.
Conclusion: Effective method for prediction of recurrence and long-term remission within 3 years after successful endoscopic transsphenoidal adenomectomy is proposed.
Keywords: Cushing disease; Neural network; Neuroendoscopy; Pituitary ACTH hypersecretion; Prediction; Recurrence; Remission; Web calculator.