Predictive model of surgical remission in acromegaly: age, presurgical GH levels and Knosp grade as the best predictors of surgical remission.

PURPOSE: To identify presurgical clinical, hormonal and radiological variables associated with surgical remission in acromegaly and develop a predictive model for surgical remission.
METHODS: Ambispective study of acromegaly surgical patients followed in two Spanish tertiary hospitals. Patients operated by the same neurosurgeon by endonasal endoscopic transsphenoidal approach (n = 49) were included to develop the predictive model, and patients operated by other neurosurgeons (n = 37) were used for external validation of the predictive model. The predictive model was developed with a multivariate logistic regression model based on the 2000 criteria.
RESULTS: 86 acromegalic patients were included. 49 patients, 83.7% with macroadenomas and 32.7% with Knosp grade > 2, were included for the development of the predictive model. The overall rate of surgical remission with the 2000 criteria was 73.5% and 51.0% with the 2010 criteria. Using the 2000 criteria, variables associated with surgical remission were: older age (OR = 1.1, p = 0.001), lower basal presurgical GH levels (OR = 0.9, p = 0.003), Knosp 0-2 (OR = 34.1, p < 0.0001) and lower maximum pituitary adenoma diameter (OR = 0.9, p = 0.019). The model with the best diagnostic accuracy to predict surgical remission combined age, Knosp 0-2 and presurgical GH levels (AIC = 29.7, AUC = 0.95) with a sensitivity of 93.8% and a specificity of 75.0%. The estimated loss of prediction with the external validation (n = 37) was 4.2%.
CONCLUSION: The predictive model with the best diagnosis accuracy for surgical remission combined age, Knosp 0-2 and presurgical GH levels, with a sensitivity of 93.8% and a specificity of 75.0%. This model could be very useful to select candidates to preoperative medical treatment and planning the follow-up.