Spontaneous regression of Rathke's cleft cysts: MRI predictors and temporal dynamics.

PURPOSE: To identify objective MRI features predictive of spontaneous regression in Rathke's cleft cysts (RCCs) under active surveillance and to develop a predictive model based on these features. METHODS: This retrospective study included 161 patients with RCCs managed with active surveillance (30 with regression, defined as volume reduction ≥25% or disappearance; 131 without). MRI at 3.0 T (including T1-weighted, T2-weighted, and dynamic contrast-enhanced sequences) was independently evaluated by two neuroradiologists blinded to clinical data, time sequence, and group assignment. Multivariable logistic regression and Cox proportional hazards models were used to identify predictors, with model discrimination assessed using Harrell's C-index and time-dependent AUC (Area under the curve). RESULTS: Maximum cyst diameter was a significant predictor of regression (OR per mm = 1.24; HR = 1.27). Intracystic nodules were significant in logistic regression (OR = 3.80) but not in the time-dependent Cox model (HR = 1.92). Optic chiasm compression showed a negative association, significant in the Cox model (HR = 0.10). The combined model (diameter + nodules + compression) achieved significant discriminatory ability (C-index = 0.761), with higher discriminative performance during the 500-800 day follow-up interval. CONCLUSIONS: A maximum cyst diameter greater than 8 mm and intracystic nodules are associated with spontaneous RCC regression, suggesting that cysts with these features may warrant increased clinical attention within the first 500-800 days of surveillance, though specific follow-up strategies require prospective validation.