JS01.4.A DIAGNOSTIC ACCURACY OF 2-HYDROXYGLUTARATE MAGNETIC RESONANCE SPECTROSCOPY FOR IDH MUTATION DETECTION IN GLIOMAS: VALIDATION IN A ROUTINE CLINICAL SETTING

Abstract BACKGROUND To evaluate the diagnostic performance of 2-hydroxyglutarate magnetic resonance spectroscopy (2HG-MRS) for non-invasive detection of isocitrate dehydrogenase (IDH) mutations in gliomas and to define its clinical utility in a routine clinical setting. MATERIAL AND METHODS In this retrospective single-center study, we initially investigated 135 glioma patients with suspected gliomas using MRS. Data from 81 patients who underwent resection and had histological confirmation of their lesions were analyzed. All patients underwent single-voxel 2HG-MRS using a PRESS sequence on a 3T MRI scanner. A positive MRS result was defined as either absolute 2HG concentration ≥1.0 mM with fitting error ≤15%, or a 2HG/creatine (2HG/Cr) ratio ≥0.252. Diagnostic sensitivity, specificity, positive and negative predictive values were calculated. ROC curve analysis was performed to compare both quantification strategies. We also documented technical and anatomical sources of error, including voxel placement near CSF or resection cavities. RESULTS The sensitivity of absolute 2HG concentration (≥1.0 mM, fitting error ≤15%) for detecting IDH-mutant gliomas was 36.6%, with specificity for excluding IDH-wildtype glioblastomas at 100%. The use of the 2HG/Cr ratio ≥0.252, a method suggested by others, improved sensitivity to 75.6%, although specificity decreased to 75.0%. Among WHO grade 4 IDH-mutant astrocytomas, sensitivity was higher: 71.4% using absolute 2HG and 85.7% using the 2HG/Cr ratio. ROC analysis demonstrated superior diagnostic accuracy for the ratio-based method (AUC = 1.000) compared to absolute quantification (AUC = 0.944). However, all misclassified cases were linked to technical or anatomical issues such as voxel misplacement near resection cavities or CSF. Six discordant cases were validated using LC-MS/MS on frozen tumor tissue, confirming misclassification due to acquisition errors rather than metabolic ambiguity. CONCLUSION While 2HG-MRS shows diagnostic potential for non-invasive IDH mutation detection, its sensitivity and specificity are not sufficient to use it universally across all glioma cases. The method is more reliable for large tumors or suspected gliomas but is prone to significant errors. Other modalities are needed to improve IDH mutation prediction rates.