OC01 - The predictors of mortality in patients with acromegaly: an old debate revisited by machine learning

Background/Introduction Mortality studies on acromegaly face two fundamental problems: large registries suffer from heterogeneous approaches among centers, while homogeneous single-center studies are limited by sample size. Furthermore, conventional statistical methods did not produce consistent results in mortality predictors. These challenges fueled the debate about which factors most significantly influence mortality in this orphan disease. Systematic analysis of complex interactions between multiple clinical variables in a large homogenous dataset may overcome these concerns. In this setting, machine learning (ML) might be useful. Purpose To evaluate predictors of mortality in patients with acromegaly using ML models. Methods We analyzed 607 patients with acromegaly. Grid search was applied to compare five types of prediction models in 1,200 configurations. Each model predicted mortality based on 40 disease characteristics, assessed using Matthews’ correlation coefficient (MCC) and area under curve (AUC) metrics. To ensure robust and reliable predictions, we constructed a virtual ensemble model by generating 100 Explainable Boosting Machine (EBM) variants and evaluating their performance across 50 random train-test splits. Results The single EBM model fitted to all data achieved an MCC of 0.88 and an AUC of 0.99. Feature importance analysis identified hypertension, first-line transsphenoidal surgery (TSS), disease duration, baseline insulin-like growth factor-1 levels, hypopituitarism, and repeat TSS as top-tier predictors of mortality. While combination drug therapy was correlated with improved survival; repeat TSS, radiotherapy, and cavernous invasion were correlated with death. A virtual ensemble prediction model showed good generalization performance: MCC of 0.64±0.08 and an AUC of 0.96±0.015 across 50 random data splits. The clinical context included 58% female gender and 66.9% macroadenoma. The first-line treatment was TSS in 83.2%. Additional treatments were repeat TSS (16.6%), somatostatin analogs (57%), cabergoline (35.6%), pegvisomant (10.9%), and radiotherapy (16.6%), achieving a 77.8% success rate. Hypertension (37.2%) and diabetes (36.4%) were the most common comorbidities. The mortality rate was 8.4%, mainly from cardiovascular disease (41.2%) and cancer (21.6%). Conclusion(s) ML showed promising performance in classifying survival status and identifying mortality predictors in acromegaly. The ensemble model offers a reliable and interpretable tool, balancing accuracy with the robustness required for clinical decision-making in patients with acromegaly.