Predictive modeling for pituitary adenomas: single center experience in 501 consecutive patients

Personalized postoperative management of patients with pituitary adenomas requires an early risk stratification system.

We reviewed 501 cases operated between 10/27/2011 and 5/5/2016 by a single neurosurgeon. We determined biochemical remission and tumor resection at 3 months, and biochemical recurrence, tumor recurrence, radiation and reoperation during follow-up. We considered age, gender, tumor diameter, cavernous sinus invasion (CSI) by MRI, diagnostic category (clinical, biochemical and immunohistochemical), and proliferation markers in a Cox proportional hazards model. We built predictive models with the significant parameters and used Kaplan-Meier survival curves for time-dependent analyses.

The 501 cases comprised 141 functional and 360 nonfunctional adenomas. Tumor diameter, CSI, and ki-67 index predicted long-term events. Model 1 (CSI, diameter ≥ 2.9 cm and ki-67 > 3%) identified 18 (3.6%) adenomas and predicted persistent hypersecretory syndrome and residual tumor with 98.7% specificity (OR 8.6; CI 3.0-24.7). Model 2 (ki-67 > 3% and CSI) identified 48 (9.6%) adenomas and had 93.1% specificity (OR 3.3; CI 1.8-6.0). Model 3 (ki-67 > 3%, mitoses and p53, former "atypical" adenoma) identified 26 (5.2%) adenomas and had 96.0% specificity (OR 2.3; CI 1.0-5.0). Model 1 best predicted the long-term event-free survival and was strengthened when Knosp 3-4 CSI grades were used. Model 2 better identified the smaller adenomas at risk. Among the WHO 2017 special PA subtypes, patients with silent corticotroph adenoma had a lower event-free survival than ACTH-negative nonfunctional adenomas.

Use of CSI, ki-67 and tumor diameter in prediction models facilitates tailored surveillance and management of patients with pituitary adenomas.