From SGAP-Model to SGAP-Score: A Simplified Predictive Tool for Post-Surgical Recurrence of Pheochromocytoma

Abstract

A reliable prediction of the recurrence risk of pheochromocytoma after radical surgery would be a key element for the tailoring/personalization of post-surgical follow-up. Recently, our group developed a multivariable continuous model that quantifies this risk based on genetic, histopathological, and clinical data. The aim of the present study was to simplify this tool to a discrete score for easier clinical use. Data from our previous study were retrieved, which encompassed 177 radically operated pheochromocytoma patients; supervised regression and machine-learning techniques were used for score development. After Cox regression, the variables independently associated with recurrence were tumor size, positive genetic testing, age, and PASS. In order to derive a simpler scoring system, continuous variables were dichotomized, using > 50 mm for tumor size, ≤ 35 years for age, and ≥ 3 for PASS as cut-points. A novel prognostic score was created on an 8-point scale by assigning 1 point for tumor size > 50 mm, 3 points for positive genetic testing, 1 point for age ≤ 35 years, and 3 points for PASS ≥ 3; its predictive performance, as assessed using Somers’ D, was equal to 0.577 and was significantly higher than the performance of any of the four dichotomized predictors alone. In conclusion, this simple scoring system may be of value as an easy-to-use tool to stratify recurrence risk and tailor post-surgical follow-up in radically operated pheochromocytoma patients.

Keywords

• pheochromocytoma
• chromaffin system
• predictive score
• recurrence prediction
• machine learning