RSS-Feed abonnieren
DOI: 10.1055/s-0044-1780278
Machine Learning-Based Model to Predict Long-Term Tumor Control and Additional Interventions following Pituitary Surgery for Cushing’s Disease
Objective: Patients with adrenocorticotropic hormone (ACTH)-dependent Cushing’s syndrome (i.e., Cushing’s disease) may suffer from severe hypertension, diabetes, and central obesity, which can ultimately lead to death if left untreated. Surgical resection is the mainstay of treatment; yet, there is a clinical need to identify predictors of persistent or recurrent hypercortisolism in the postoperative setting. This study aimed to establish a supervised machine learning model based on multiple tree-based algorithms to predict long-term biochemical outcomes and additional intervention-free survival after endonasal transsphenoidal surgery (ETS) for CD patients.
Methods: The authors reviewed the medical records of patients who underwent ETS for CD patients at our institution between 2013 and 2022. Data on patients’ baseline characteristics, intervention details, histopathology, surgical outcomes, and postoperative neurological and endocrine functions were collected. The primary outcome of the study was intervention-free survival (IFS) rate, and the therapeutic outcomes were labeled as “under control” or “treatment failure,” depending on whether additional therapeutic interventions were required. The secondary outcomes included overall survival (OS) and neurological and endocrine outcomes. The decision tree and random forest classifiers were trained and tested to predict long-term tumor recurrence based on unseen data using 80/20 split of the cohort.
Results: Data on 189 ETSs for 150 patients with a median (range) follow-up period of 55 (1–261) months were extracted. In the entire cohort, 42 patients (28%) required additional interventions for persistent or recurrent CD. Consequently, IFS rates following ETS alone were 83% at 3 years and 78% at 5 years in the entire cohort ([Fig. 1]).


Multivariable Cox proportional hazard analysis demonstrated that only Knosp–Steiner (KS) grade 4 cavernous sinus (CS) extension (vs. KS grades 0–3) was the significant risk associated with worsen IFS (hazard ratio, 8.12; 95% confidence interval, 2.50–26.37; p = 0.001). In the decision tree analysis, when patients had a KS grade 4 CS extension, they had a 90% risk of additional intervention. In addition, when patients had KS grades 1 and 2 CS extension and were younger than 34 years, they had a 62% risk of additional intervention (accuracy 78%; [Fig. 2]).


Random forest analysis (n = 500 trees) revealed that KS grade (mean minimal depth = 1.14), tumor size (1.47), and patient age (1.50) were the three most significant predictors of long-term tumor control and intervention needs ([Fig. 3]).


Conclusions: Based on supervised machine learning models, KS grade, age, and tumor size were the three most important predictors of intervention-free survival. These data provide insight into at-risk CD patients who may require addition interventions on follow-up.
Publikationsverlauf
Artikel online veröffentlicht:
05. Februar 2024
© 2024. Thieme. All rights reserved.
Georg Thieme Verlag KG
Rüdigerstraße 14, 70469 Stuttgart, Germany