Development and validation of a predictive model for poor initial outcomes after Gamma Knife radiosurgery for...

The present study aimed to develop a reliable predictive model for identifying preoperative predictors of poor initial outcomes in patients with primary trigeminal neuralgia (PTN) treated with Gamma Knife radiosurgery (GKRS) and further elucidate the clinical significance of these predictors in initial outcomes and long-term pain recurrence.

A total of 217 PTN patients were divided into a training set (n = 167) and a validation set (n = 50). The initial outcomes of GKRS treatment were assessed based on the Barrow Neurological Institute pain intensity scale. A predictive model was developed through multivariate regression and validated with repeated sampling. The differences in predictors of long-term pain recurrence were assessed using Kaplan-Meier analysis. The association between predictors was tested using chi-square tests, and subgroup analyses were performed to compare initial outcomes and long-term pain recurrence between two clinically significant correlates.

The training and validation sets showed areas under the curve of 0.85 and 0.88, respectively. Calibration curves and decision curve analysis indicated significant clinical benefits in both sets. Independent risk factors for poor initial outcomes included hyperglycemia, absence of neurovascular contact, carbamazepine insensitivity, and atypical pain (trigeminal neuralgia type 2 [TN2]). Carbamazepine insensitivity was moderately associated with TN2 and predicted long-term pain recurrence. Patients with both phenotypes had significantly worse initial outcomes compared with other subgroups (adjusted p = 0.0125).

Patients with both TN2 and carbamazepine insensitivity have the poorest initial treatment outcomes and face an increased risk of recurrence. Furthermore, this predictive model is highly accurate and useful, offering a comprehensive method of identifying PTN patients likely to experience poor initial outcomes based on clinical characteristics and imaging perspectives. The authors believe that the nomogram presented in this model enables clinicians to calculate multiple variables and predict the probability of adverse events.