Reliability of MRI for predicting characteristics of neurovascular conflicts in trigeminal neuralgia: implications for surgical decision making

OBJECTIVE The choice of microvascular decompression (MVD), among the several other surgical options, for treating refractory classical trigeminal neuralgia (TN) relies mostly on preoperative imaging, but the degree of reliability of MRI remains a matter of debate. The authors approached the question of predictability of neurovascular conflict (NVC) in a series of 100 protocolized MRI studies from patients with TN who underwent MVD, by reexamination of MR images, blinded to the clinical data and surgical findings, including the side of the neuralgia. METHODS Patients included in the study were those who underwent MVD after surgical indication had been determined based on a protocolized imagery workup (3D high-resolution T2-weighted cisternography centered on the trigeminal nerve, 3D time-of-flight angiography, and 3D gadolinium-enhanced T1-weighted imaging) performed at our institution. All MR images were blindly reexamined, and neurovascular relationships were described on both sides, noting the existence of compression, vessels involved, situation along the root, and degree of compression. The results of MRI evaluation were then compared with actual surgical findings. The extent of agreement and quality of the prediction were expressed with Cohen's kappa coefficient (kappa) and receiver operating characteristic (ROC) statistics. RESULTS A conflict had actually been found during surgery in 94 of 100 patients. The sensitivity of MRI to detect a conflict was 97% and the specificity was 50%. Vessel type was identified with high reliability (kappa = 0.80), while the grade of the conflict and its situation along the root showed poor to average reliability (kappa = 0.38 and kappa = 0.40, respectively). The area under the ROC curve for predicting the presence of a conflict according to the grades of conflict seen on MRI was 0.93, which is considered very good. The positive predictive value was differentiated according to the grade of conflict, with a very high value for high grades of vascular conflict. CONCLUSIONS This study shows an overall good reliability of MRI to predict the existence of an NVC. The prediction value is excellent for high grades of compression. Some apparent low-grade compressions on MRI may be revealed as false positives in surgical exploration. This raises the question of what other imaging methods might be used to determine not only the existence of a conflict but also its degree of compression. The degree of compression is of paramount importance to predict the probability of long-term pain relief, and therefore in the decision to propose MVD as the first choice of surgical treatment.