Supervised machine learning to validate a novel scoring system for the prediction of disease remission of functional pituitary adenomas following transsphenoidal surgery

Abstract

Functional pituitary adenomas (FPAs) are associated with hormonal hypersecretion resulting in systemic endocrinopathies and increased mortality. The heterogenous composition of the FPA population has made modeling predictive factors of postoperative disease remission a challenge. Here, we aim to define a novel scoring system predictive of disease remission following transsphenoidal surgery (TSS) for FPAs and validate our process using supervised machine learning (SML). 392 patients with FPAs treated at one of the three Mayo Clinic campuses were retrospectively reviewed. Variables found significant on multivariate analysis were incorporated into our novel Pit-SCHEME score. The Pit-SCHEME score with a cut-off value ≥ 6 achieved a sensitivity of 86% and positive likelihood ratio of 2.88. In SML models, without the Pit-SCHEME score, the k-nearest neighbor (KNN) model achieved the highest accuracy at 75.6%. An increase in model sensitivity was achieved with inclusion of the Pit-SCHEME score with the linear discriminant analysis (LDA) model achieving an accuracy of 86.9%, which suggests the Pit-SCHEME score is the variable of most importance for prediction of postoperative disease remission. Ultimately, these results support the potential clinical utility of the Pit-SCHEME score and its prospective future for aiding in the perioperative decision making in patients with FPAs.