Objective: To compare the performance of the 2017 McDonald criteria with that of the 2010 criteria for the diagnosis of multiple sclerosis (MS) in children in the clinical setting. Methods: In this retrospective, multi-centre study, we identified children who presented with symptoms suggestive of a clinically isolated syndrome (CIS) and were followed up for at least 2 years or until their second attack. Results: Of 156 children with CIS followed up for a median of 4.17 years, 94 (60.3%) were diagnosed with MS. In all, 83 (88.3%) of these fulfilled the 2010 dissemination in space (DIS) criteria at onset. Three additional children fulfilled the 2017 DIS criteria because of the inclusion of symptomatic lesions. Of the 59 children with MS who underwent post-gadolinium magnetic resonance imaging (MRI), 44 (74.6%) fulfilled the 2010 dissemination in time (DIT) criteria at baseline. When the presence of oligoclonal bands (OCBs) was used to substitute DIT, an additional 35 children (79/94, 84.0%) were diagnosed with MS according to the 2017 criteria. The 2017 criteria had higher accuracy (87.2% vs 66.7%), higher sensitivity (84.0% vs 46.8%), but reduced specificity (91.9% vs 96.8%) when compared to the 2010 criteria. Conclusion: The improved performance of the 2017 criteria when compared to the 2010 criteria was predominantly due to the inclusion of intrathecal OCBs.