ACCURACY OF MULTIPLE SCLEROSIS DAIGNOSTIC CRITERIA FOR DETECTING PERIVENULAR DEMYELINATION VISUALIZED BY MRI AND FREQUENCY OF MS-MIMICKING DISEASES

Objective: Perivenular lesions (PVL) are a cardinal pathological feature of multiple sclerosis (MS) that now can non-invasively be detected by brain MRI. High PVL frequency/patient (PVL-f) is specific of MS and can therefore be used as a gold standard to evaluate accuracy and predictive values of the MS diagnostic criteria in distinct types of MS and the frequency of possible MS misdiagnosis, in particular in MS patients fulfilling the MS diagnostic criteria but carrying also red flags (clinical laboratory or MRI features) suggesting better explanation but not formally allowing other diagnosis (MS-plus patients). The aims of this study were - To identify an accurate PVL-f threshold that discriminates MS from other conditions, generated with a ROC (receiver operating characteristic curve) analysis including true positive cases (typical MS) and true negative cases (patients with other definite CNS diseases). - To evaluate McDonald MS criteria accuracy in MS-plus patients categorized as true or false MS cases according to the PVL-f threshold previously identified with ROC analysis. - To evaluate differences among MS-plus patients subgroups stratified according to the PVL-f threshold previously identified with ROC analysis. Methods: Typical relapsing remitting (RR)MS (n= 28), RRMS-plus (n=59) fulfilling the DIS and DIT based diagnostic criteria and Not-MS neurological syndromes with MS-like brain white matter lesions (WML) (n=32), received one brain MRI scan including conventional and FLAIR* sequences. PVL-f and conventional brain MRI characteristics were evaluated in each MS patient. The PVL-f threshold that best discriminates MS from other neurological conditions was obtained with ROC analysis including true MS (typical MS cases) and true negative cases (patients with not-MS definite neurological conditions). MS-plus patients fulfilling or not the PVL-f threshold generated by ROC analysis were categorized in two groups for evaluating the accuracy of the MS diagnostic criteria for detecting fulfilment of this threshold. Data concerning clinical, demographic, conventional MRI, OCT and OCT-angiography were also collected and analyzed to find potential differences among patients’ groups. Results: The threshold-value of PVL-f identified with ROC analysis to discriminate true-MS cases resulted >51%. Typical MS patients had a median PVL-f = 91% (range 67–100%), the MS-plus = 55% (range 8–100%; p=0.001) and the non-MS = 23% (range 0-89%, p< 0.00001). The 51% PVL-f threshold - selected by ROC analysis - was fulfilled by 100% (28/28) of the Typical MS and by 3% (1/33) of non-MS (p< 0.00001) indicating 0.98 accuracy of the MS diagnostic criteria in this population. However only 52.5% (31/60) of the MS-plus patients fulfilled this threshold (p= 0.001), indicating in this patient population 0.68 accuracy of the MS diagnostic criteria. Conventional MRI measure did not contribute to the accuracy of the MS diagnostic criteria, but MS-atypical lesions resulted more frequent in MS-plus, representing the most reliable red flag for its identification. Presence of cerebrovascular comorbidities, high frequency of small lesions and in non-typical MS locations, segregated with the MS-plus patients who did not reach the >51% PVL-f threshold suggesting that in MS-plus these diseases may represent an explanation better than MS. Interpretation: In MS-plus patients categorized by PVL-f>51% (the MRI in vivo hallmark of MS pathology), the DIS/DIT based MS diagnostic criteria have low performances, indicating that atypical MS cases (MS-plus) represent a group of patients at high risk of misdiagnosis and therefore need PVL-f evaluation.