2025 Poster Presentations
P207: MULTI-FACTOR PREDICTION MODEL FOR ESTIMATING SURGICAL REMISSION IN ACROMEGALY PATIENTS
Biren K Patel, MBBS, MS, MCh; Leonardo Tariciotti; Alejandra Rodas; Youssef Zohdy; J. Manuel Revuelta Barbero; Erion Jr De Andrade; Justin Maldonado; Samir Lohana; Rodrigo Uribe-Pacheco; Hanyao Sun; Roberto Soriano; Tomas Garzon-Muvdi; C.Arturo Solares; Gustavo Pradilla; Emory University
Background: Growth hormone-secreting adenomas are difficult to manage due to their complex and variable biological behavior. Surgical resection is the primary treatment, with somatostatin receptor ligands being the first-line medical treatment. Predicting treatment outcomes in this tumor is complex due to variable remission rates (24-65%) influenced by tumor size, invasiveness, and preoperative hormone levels. The concept of "difficult" or "aggressive" GH-secreting pituitary adenoma includes tumors resistant to standard treatments, exhibiting invasive growth, high proliferation rates, and recurrence. This heterogeneity in tumor behavior and treatment response necessitates a multidisciplinary approach to achieve complete remission. Traditional predictive models often fall short in capturing these complexities, necessitating the exploration of advanced methodologies such as machine learning (ML).
Objective: This study aimed to develop and validate a machine learning-based prediction model to predict surgical remission in patients with acromegaly.
Methods: A retrospective study was conducted involving approximately 80 acromegaly patients who underwent surgery at our institution between July 2014 and July 2024. Key variables which were collected include demographic information, clinical features, preoperative biochemical markers (GH and IGF-1 levels), histopathological markers, and radiological characteristics (tumor size, cavernous sinus invasion). Surgical details and postoperative outcomes, specifically remission status at 3 months and 1-year post-surgery, were recorded. Various machine learning algorithms were tested, and the most accurate model was used as the prediction model. Model performance was evaluated using metrics such as the area under the receiver operating characteristic curve (AUC-ROC), accuracy, sensitivity, and specificity.
Results: The machine learning model demonstrated superior predictive accuracy compared to traditional statistical methods. The model's performance were validated on an independent test set, ensuring its generalizability and robustness.
Conclusion: Our machine learning-based prediction model for surgical remission in acromegaly patients incorporated a comprehensive set of preoperative, intraoperative as well as postoperative variables. This model has the potential to significantly improve preoperative decision-making and patient counseling, ultimately enhancing clinical outcomes in acromegaly management.
Keywords: Acromegaly, pituitary adenoma, Surgery, predictive model, surgical remission, machine learning