Sensitivity vs specificity

Pediatric multiple sclerosis (MS) is clearly underrecognized and is often misdiagnosed as acute disseminated encephalomyelitis (ADEM), or as one of many relapsing pediatric neurologic conditions. Up to 5% of MS cases have a pediatric onset,¹ and there are an estimated 10,000 cases in the United States alone.² Recently, the International Pediatric MS Study Group proposed new clinical criteria for pediatric MS and related demyelinating disorders.³ However, definitive MRI criteria for these conditions are lacking. Recent studies have shown that at the time of their MS-defining attack, children meet McDonald MRI criteria only 67% of the time, suggesting a lower lesion burden at the time of presentation compared to adults.⁴ Previously published studies have suggested that the presence of periventricular and corpus callosum well-defined lesions may be sensitive predictors of MS after a first attack in childhood.⁵ Thus, neuroimaging biomarkers are powerful tools that may assist in both diagnosis and prognostication for demyelinating disorders in children.

In a pair of articles published in this issue of Neurology®, Callen and a well-recognized team of authors^6,7 contrasted the MRI features of clinically definite MS in children with ADEM and with other neurologic diseases. In the first publication, they propose new criteria for differentiating MS from ADEM at first presentation.⁶ This is an important issue: while adult MS often starts with a distinct clinically isolated demyelinating syndrome (CIS), ADEM may be the first manifestation of pediatric MS in up to one fifth of cases.⁸ The authors studied 28 MS and 20 ADEM cases that had not converted to MS, defined using the International Pediatric MS Study Group consensus criteria.³ Although cases were followed for at least 2 years (mean 4.1 years), it is conceivable that some may still develop MS in the future. Multiphasic and recurrent ADEM cases were excluded and the study was not designed to clarify any potential confusion regarding MS and those entities. Only universally available MRI sequences were used, which makes the findings easily applicable at any medical center. However, spinal cord MRI, volumetric studies, and advanced MRI techniques addressing changes in normal-appearing gray and white matter may all have provided useful additional contribution to the differentiation of ADEM from MS.

In the second study, the authors compared MRI features of the first and second relapses in cases of clinically definite pediatric MS with cases of pediatric migraine and CNS lupus, and on this basis, propose revisions to the McDonald MRI criteria⁹ when applied to children.⁷ One of the motivations for these studies is the absence of definitive MRI criteria for childhood-onset MS, despite the development of consensus clinical criteria³; however, it is unclear how relevant the entities of migraine and CNS lupus are as comparison cohorts.

The two studies yielded some overlapping as well as divergent findings. Using a forward stepwise conditional logistic regression approach, the following qualitative MRI variables were found to be significant in distinguishing a first attack of pediatric MS from ADEM: 1) presence of T1 black holes; 2) presence of two or more periventricular lesions; 3) absence of diffuse bilateral lesions. Of note, diffuse bilateral lesions were the weakest of these predictors, were only seen below age 10, and mostly in ADEM. In contrast, MRI criteria that the authors propose as effective in distinguishing between pediatric MS at the time of the MS-defining relapse from migraine and CNS lupus were 1) presence of five or more lesions (as opposed to nine lesions in the McDonald criteria⁹); 2) two or more periventricular lesions (as opposed to three or more); 3) one brainstem lesion (as opposed to one infratentorial lesion). Juxtacortical lesions were not found to be significant, whereas contrast-enhancing lesions were not studied due to the low number of available scans. The reason for omitting postcontrast scans was likely related to the inconvenience of IV administration in children, but a perceived fear of triggering nephrogenic systemic fibrosis (NSF) or nephrogenic fibrosing dermopathy (NFD) may also have played a role. NSF and NFD have only been reported in patients with severe renal dysfunction, hepatorenal syndrome, or in the perioperative liver transplantation period.¹⁰ In general, it is not advised to administer gadolinium to children below the age of 1, but no other pediatric advisories exist. In the few patients with postgadolinium scans, one enhancing lesion could perhaps substitute for five T2 hyperintense lesions without any reduction of sensitivity, but this requires further validation.

The ADEM-MS MRI criteria were 81% sensitive and 95% specific in distinguishing ADEM from a first attack of MS. In contrast, the pediatric MS MRI criteria were only 75% sensitive and 25% specific when applied to the ADEM-first attack cohort. Thus, specificity may have to be sacrificed for sensitivity, depending on the question. The relative absence of black holes in ADEM compared to eventual MS cases suggests that ADEM may be a less destructive process; however, “black holes” in the MS literature usually refers to chronic, nonenhancing lesions, whereas in this article, most black holes were transient hypointensities in contrast-enhancing lesions. Longitudinal assessment for chronic black holes was not performed.

The studies by Callen et al. represent a positive step toward defining MRI criteria for pediatric MS, and in prognostication after an initial demyelinating event in childhood. However, larger prospective studies, which employ standardized protocols and include brain and spinal cord imaging, are required to validate their proposed MRI criteria for pediatric MS. Diagnostic and prognostic MRI criteria will have to balance the demands of sensitivity and specificity given the multiple mimics of pediatric MS, and the unclear immunopathogenetic relationship with ADEM. Given that children with MS are in close proximity to the potential inciting events of MS, definitive neuroimaging studies are not only essential for diagnosis and prognosis, but may hold the key to unlocking the earliest pathogenic events in MS.