Machine Learning-based Prediction Model for Treatment of Acromegaly With First-generation Somatostatin Receptor Ligands

Abstract Context Artificial intelligence (AI), in particular machine learning (ML), may be used to deeply analyze biomarkers of response first-generation somatostatin receptor ligands (fg-SRLs) the treatment acromegaly. Objective To develop a prediction model therapeutic acromegaly fg-SRL. Methods Patients with not cured by primary surgical and who had adjuvant therapy fg-SRL for at least 6 months after surgery were included. considered controlled if they presented growth hormone (GH) <1.0 ng/mL normal age-adjusted insulin-like factor (IGF)-I levels. Six AI models evaluated: logistic regression, k-nearest neighbor classifier, support vector machine, gradient-boosted random forest, multilayer perceptron. The features included analysis age diagnosis, sex, GH, IGF-I levels diagnosis pretreatment, subtype 2 5 (SST2 SST5) protein expression cytokeratin granulation pattern (GP). Results A total 153 patients analyzed. Controlled older (P = .002), lower GH .01), pretreatment < .001), more frequently harbored tumors that densely granulated .014) or highly expressed SST2 .001). performed best was SST2, SST5, GP, age, It an accuracy 86.3%, positive predictive value 83.3% negative 87.5%. Conclusion We developed ML-based high has potential improve medical management acromegaly, optimize biochemical control, decrease long-term morbidities mortality, reduce health services costs.