Advertisement

Magnetic resonance imaging criteria at onset to differentiate pediatric multiple sclerosis from acute disseminated encephalomyelitis: A nationwide cohort study

      Highlights

      • MRI at onset can distinguish pediatric MS and ADEM by up to 89%.
      • Callen, KIDMUS, and IPMSSG criteria performed well in distinguishing MS and ADEM.
      • The McDonald 2017, Barkhof, MAGNIMS, and Verhey criteria had poorer performance.

      Abstract

      Background

      MRI of the nervous system is the critical in distinguishing pediatric MS from acute disseminated encephalomyelitis (ADEM). Our aim was to propose MRI criteria to distinguish MS from monophasic ADEM based on the first MRI and to validate previously proposed MRI criteria.

      Methods

      A neuroradiologist undertook retrospective evaluation of the MRI at the first demyelinating event in children (<18 years) with medical record-validated MS and ADEM in Denmark during 2008–15. We used forward stepwise logistic regression to identify MRI categories that differed significantly between MS and ADEM. We estimated accuracy statistics for all MRI criteria to distinguish MS from ADEM.

      Results

      The monophasic ADEM cohort (n=46) was nationwide and population-based during 2008–15; the median age at onset of 5.3 years (range 0.8‒17.2) and children had at least five years of follow-up to ensure a monophasic disease course. Children with MS (n=67) had a median age at onset of 16.3 years (range 3.3‒17.9). Having at least two categories best distinguished MS from monophasic ADEM by an area under the curve of 83% to 89%: (a) corpus callosum long axis perpendicular lesion; (b) only well-defined lesions; (c) absence of basal ganglia or thalamus lesion OR, (a) corpus callosum long axis perpendicular lesion; (b) only well-defined lesions; (c) absence of diffuse large lesions; (d) black holes. The Callen, KIDMUS, and IPMSSG criteria performed well. The McDonald 2017, Barkhof, MAGNIMS, and Verhey criteria had poorer performance.

      Conclusion

      This study provides Class II evidence that MRI has good performance in differentiating MS from monophasic ADEM at onset.

      Keywords

      Abbreviations:

      ADEM (acute disseminated encephalomyelitis), ICD (International Classification of Diseases), IPMSSG (International Pediatric MS Study Group criteria), KIDMUS (Kids with MS study (French Study Group)), MOG (myelin oligodendrocyte glycoprotein), MRI (magnetic resonance imaging), MS (multiple sclerosis)

      1. Introduction

      In children with a first episode of acquired CNS demyelination, MRI is used to help distinguish monophasic from relapsing disease. Acute disseminated encephalomyelitis (ADEM) is the most commonly acquired demyelinating syndrome in the first decade of life, and it is usually monophasic. In contrast, multiple sclerosis (MS) is a relapsing disease and becomes increasingly more frequent after age 11 years (
      • Boesen MS
      • Magyari M
      • Koch-Henriksen N
      • Uldall PV
      • Thygesen LC
      • Blinkenberg M
      • et al.
      Implications of the International Paediatric Multiple Sclerosis Study Group consensus criteria for paediatric acute disseminated encephalomyelitis: a nationwide validation study.
      ;
      • Boesen MS
      • Magyari M
      • Koch-Henriksen N
      • Thygesen LC
      • Born AP
      • Uldall PV
      • et al.
      Pediatric-onset multiple sclerosis and other acquired demyelinating syndromes of the central nervous system in Denmark during 1977-2015: A nationwide population-based incidence study.
      ). Accordingly, it is paramount to distinguish the two diseases early because they differ considerably regarding treatment and prognosis for relapse (
      • Krupp LB
      • Tardieu M
      • Amato MP
      • Banwell B
      • Chitnis T
      • Dale RC
      • et al.
      International Pediatric Multiple Sclerosis Study Group criteria for pediatric multiple sclerosis and immune-mediated central nervous system demyelinating disorders: revisions to the 2007 definitions.
      ;
      • Thompson AJ
      • Banwell BL
      • Barkhof F
      • Carroll WM
      • Coetzee T
      • Comi G
      • et al.
      Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
      ;
      • Polman CH
      • Reingold SC
      • Banwell B
      • Clanet M
      • Cohen JA
      • Filippi M
      • et al.
      Diagnostic criteria for multiple sclerosis: 2010 Revisions to the McDonald criteria.
      ).
      MRI is the most important investigation to distinguish ADEM from MS. The following MRI features seem to increase the risk of relapse: no diffuse bilateral lesions, black holes, periventricular lesions (in particular, corpus callosum long axis perpendicular lesion), sole presence of well-defined lesions, optic nerve lesion, occipital lesions, presence of contrast and non-contrast enhancing lesions, and more than nine T2 lesions (
      • Thompson AJ
      • Banwell BL
      • Barkhof F
      • Carroll WM
      • Coetzee T
      • Comi G
      • et al.
      Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
      ;
      • Callen DJA
      • Shroff MM
      • Branson HM
      • Li DK
      • Lotze T
      • Stephens D
      • et al.
      Role of MRI in the differentiation of ADEM from MS in children.
      ;
      • Mikaeloff Y
      • Adamsbaum C
      • Husson B
      • Vallée L
      • Ponsot G
      • Confavreux C
      • et al.
      MRI prognostic factors for relapse after acute CNS inflammatory demyelination in childhood.
      ;
      • Barkhof F
      • Filippi M
      • Miller DH
      • Scheltens P
      • Campi A
      • Polman CH
      • et al.
      Comparison of MRI criteria at first presentation to predict conversion to clinically definite multiple sclerosis.
      ;
      • Verhey LH
      • Branson HM
      • Shroff MM
      • Callen DJ
      • Sled JG
      • Narayanan S
      • et al.
      MRI parameters for prediction of multiple sclerosis diagnosis in children with acute CNS demyelination: a prospective national cohort study.
      ;
      • Boesen MS
      • Blinkenberg M
      • Born AP
      • Magyari M
      • Chitnis T
      • Thygesen LC
      • et al.
      Magnetic resonance imaging at baseline and follow-up to differentiate between pediatric monophasic acquired CNS demyelination and MS.
      ). Several criteria have been proposed to distinguish ADEM from MS but few have been validated (
      • Ketelslegers IA
      • Neuteboom RF
      • Boon M
      • Catsman-Berrevoets CE
      • Hintzen RQ
      Dutch Pediatric MS Study Group. A comparison of MRI criteria for diagnosing pediatric ADEM and MS.
      ). To date, the most useful are the Callen, KIDMUS and McDonald criteria (
      • Thompson AJ
      • Banwell BL
      • Barkhof F
      • Carroll WM
      • Coetzee T
      • Comi G
      • et al.
      Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
      ;
      • Callen DJA
      • Shroff MM
      • Branson HM
      • Li DK
      • Lotze T
      • Stephens D
      • et al.
      Role of MRI in the differentiation of ADEM from MS in children.
      ;
      • Mikaeloff Y
      • Adamsbaum C
      • Husson B
      • Vallée L
      • Ponsot G
      • Confavreux C
      • et al.
      MRI prognostic factors for relapse after acute CNS inflammatory demyelination in childhood.
      ).
      In a nationwide retrospective cohort study of children (<18 years) with a first demyelinating event as part of ADEM or MS, our aim was to develop MRI criteria to differentiate the two diseases and to test previously proposed criteria.

      2. Method and materials

      2.1 Data sources

      The Danish Civil Registration System

      The Danish Civil Registration System was established in 1968 as a register of residents in Denmark (
      • Schmidt M
      • Pedersen L
      • Sørensen HT.
      The Danish Civil Registration System as a tool in epidemiology.
      ;
      • Pedersen CB.
      The Danish Civil Registration System.
      ). Every resident at birth or on immigration receives a unique personal identification number that is used to link registers (
      • Thygesen LC
      • Daasnes C
      • Thaulow I
      • Brønnum-Hansen H.
      Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving.
      ).

      The National Patient Register

      The National Patient Register is a nationwide register with routinely collected administrative and health-related data on all hospital admissions in Denmark since 1977 (
      • Lynge E
      • Sandegaard JL
      • Rebolj M.
      The Danish National Patient Register.
      ). Data include date of admission and diagnoses according to the International Classification of Diseases (ICD)-8 (1977–1993) and ICD-10 (1994 until today). On hospital discharge, physicians code each patient by diagnosis with at least one diagnosis. In Denmark, hospital admissions and outpatient visits are tax funded and free of charge. Patients are coded at each hospital visit, giving multiple registrations for patients with chronic diseases (
      • Andersen TF
      • Madsen M
      • Jørgensen J
      • Mellemkjoer L
      • Olsen JH.
      The Danish National Hospital Register. A valuable source of data for modern health sciences.
      ). Private consultant physicians (e.g., general practitioners, neurologists, and ophthalmologists) and private hospitals in Denmark play a minor role in the diagnostic procedure and refer children with suspected demyelinating disease to public hospitals. We have validated the National Patient Register regarding pediatric acquired demyelinating syndromes and found that agreement between diagnostic codes and a medical-record validated diagnosis was excellent for MS but unacceptable for ADEM (
      • Boesen MS
      • Magyari M
      • Born AP
      • Thygesen LC.
      Pediatric acquired demyelinating syndromes: a nationwide validation study of the Danish National Patient Register.
      ).

      2.2 Children with ADEM and MS

      Children with ADEM and MS were identified by hospital ICD-10 codes in the Danish National Patient Register for acquired demyelinating syndromes during 2008–15 (Fig. 1, flow chart) (
      • Boesen MS
      • Magyari M
      • Koch-Henriksen N
      • Thygesen LC
      • Born AP
      • Uldall PV
      • et al.
      Pediatric-onset multiple sclerosis and other acquired demyelinating syndromes of the central nervous system in Denmark during 1977-2015: A nationwide population-based incidence study.
      ). We reviewed all medical records to ensure a correct diagnosis of ADEM and MS.
      Fig. 1
      Fig. 1Flow chart.
      Abbreviations: ADEM: acute disseminated encephalomyelitis; ICD-10: International Classification of Diseases, revision 10; MRI: magnetic resonance imaging; MS: multiple sclerosis.
      Children with MS fulfilled the International Pediatric MS Study Group criteria (IPMSSG) for MS (
      • Krupp LB
      • Tardieu M
      • Amato MP
      • Banwell B
      • Chitnis T
      • Dale RC
      • et al.
      International Pediatric Multiple Sclerosis Study Group criteria for pediatric multiple sclerosis and immune-mediated central nervous system demyelinating disorders: revisions to the 2007 definitions.
      ).
      The ADEM cohort was nationwide and population-based from 2008 to 2015 in children younger than 18 years in Denmark, as described previously (
      • Boesen MS
      • Magyari M
      • Koch-Henriksen N
      • Uldall PV
      • Thygesen LC
      • Blinkenberg M
      • et al.
      Implications of the International Paediatric Multiple Sclerosis Study Group consensus criteria for paediatric acute disseminated encephalomyelitis: a nationwide validation study.
      ). A diagnosis of ADEM was based on clinical criteria (“clinical ADEM”) as diagnosed by pediatricians including a review committee consisting of highly experienced pediatric neurologists (Fig. 1). A thorough description of the definition of “clinical ADEM” has previously been published (
      • Boesen MS
      • Magyari M
      • Koch-Henriksen N
      • Uldall PV
      • Thygesen LC
      • Blinkenberg M
      • et al.
      Implications of the International Paediatric Multiple Sclerosis Study Group consensus criteria for paediatric acute disseminated encephalomyelitis: a nationwide validation study.
      ;
      • Boesen MS
      • Langkilde A
      • Born AP
      • Magyari M
      • Blinkenberg M
      • Chitnis T
      • et al.
      School performance and psychiatric morbidity 6 years after pediatric acute disseminated encephalomyelitis: A nationwide population-based cohort study.
      ). All children considered as having “clinical ADEM” had an abnormal baseline MRI, but encephalopathy or polyfocal neurological deficits were not prerequisite. We defined a “monophasic ADEM group” after follow-up of the ADEM cohort for least 5 years (median follow-up: 9.3 years, range: 5.5–13.0 years). We excluded the following children from the “monophasic ADEM group”: one child who had previously experienced a longitudinal-extensive transverse myelitis; one child who experienced a new ADEM event during follow-up fulfilling the criteria for multiphasic ADEM; and one child who experienced new MRI lesions and a non-encephalopathic event fulfilling the IPMSSG criteria for MS (
      • Krupp LB
      • Tardieu M
      • Amato MP
      • Banwell B
      • Chitnis T
      • Dale RC
      • et al.
      International Pediatric Multiple Sclerosis Study Group criteria for pediatric multiple sclerosis and immune-mediated central nervous system demyelinating disorders: revisions to the 2007 definitions.
      ). This child was relocated to the MS group (Fig. 1). Further, we defined a “monophasic IPMSSG ADEM group” consisting of children with monophasic ADEM including encephalopathy and polyfocal neurological deficits as proposed by the IPMSSG.

      2.3 MRI re-evaluation

      Senior neuroradiologist AL scored all baseline MRIs. AL is a board-certified neuroradiologist with special interest and extensive experience in MRI evaluation of acquired demyelinating syndromes. For blinding purposes, the MRI review included children with MS, ADEM, and clinically isolated syndromes such as monophasic optic neuritis; accordingly, AL was aware that the MRIs belonged to children with acquired demyelinating syndromes but was blinded to all other data, including the acquired demyelinating syndrome phenotype (
      • Boesen MS
      • Blinkenberg M
      • Born AP
      • Magyari M
      • Chitnis T
      • Thygesen LC
      • et al.
      Magnetic resonance imaging at baseline and follow-up to differentiate between pediatric monophasic acquired CNS demyelination and MS.
      ).
      The MRI evaluation scheme was designed prior to the MRI review. Poor quality MRI scans (i.e., missing obligatory sequences or excessive patient movement) were not included in the review. Individual lesions were identified on axial T1, FLAIR or T2 weighted images and categorized according to location, size, demarcation, and gadolinium enhancement. Lesions were defined as proposed by the McDonald 2017 criteria (for definitions of specific MRI lesions, see Supplementary Table 1) (
      • Thompson AJ
      • Banwell BL
      • Barkhof F
      • Carroll WM
      • Coetzee T
      • Comi G
      • et al.
      Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
      ).

      2.4 Previously proposed criteria to differentiate MS from ADEM

      • 1)
        Callen (
        • Callen DJA
        • Shroff MM
        • Branson HM
        • Li DK
        • Lotze T
        • Stephens D
        • et al.
        Role of MRI in the differentiation of ADEM from MS in children.
        ): Fulfill ≥2/3:
        • a)
          no diffuse bilateral lesions
        • b)
          ≥1 black hole
        • c)
          ≥2 periventricular lesions
      • 2)
        KIDMUS (
        • Mikaeloff Y
        • Adamsbaum C
        • Husson B
        • Vallée L
        • Ponsot G
        • Confavreux C
        • et al.
        MRI prognostic factors for relapse after acute CNS inflammatory demyelination in childhood.
        ): Fulfill ≥1/2:
        • a)
          corpus callosum long axis perpendicular lesion
        • b)
          sole presence of well-defined lesions
      • 3)
        Barkhof (
        • Barkhof F
        • Filippi M
        • Miller DH
        • Scheltens P
        • Campi A
        • Polman CH
        • et al.
        Comparison of MRI criteria at first presentation to predict conversion to clinically definite multiple sclerosis.
        ): Fulfill ≥3/4:
        • a)
          ≥1 gadolinium-enhancing lesion or ≥9 T2 lesions
        • b)
          ≥3 periventricular lesions
        • c)
          ≥1 juxtacortical lesion (we used juxtacortical or cortical)
        • d)
          ≥1 infratentorial lesion
      • 4)
        McDonald 2017 (
        • Thompson AJ
        • Banwell BL
        • Barkhof F
        • Carroll WM
        • Coetzee T
        • Comi G
        • et al.
        Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
        ):
        • Dissemination in space: 1 or more T2 lesion in at least 2 of 4 MS-typical regions of the CNS (periventricular, juxtacortical/cortical, infratentorial, or spinal cord).
        • Dissemination in time: Gadolinium-enhancing and non-enhancing lesions at any time or new lesion on follow-up MRI.
      • 5)
        IPMSSG (
        • Krupp LB
        • Tardieu M
        • Amato MP
        • Banwell B
        • Chitnis T
        • Dale RC
        • et al.
        International Pediatric Multiple Sclerosis Study Group criteria for pediatric multiple sclerosis and immune-mediated central nervous system demyelinating disorders: revisions to the 2007 definitions.
        ):
        • a)
          Diffuse, poorly demarcated, large (>1–2 cm) lesions involving predominantly the cerebral white matter
        • b)
          T1 hypointense lesions in the white matter are rare
        • c)
          Deep grey matter lesions (e.g., thalamus or basal ganglia) can be present.
        • We invented the “reversed IPMSSG” to favor MS as: Absence of diffuse, poorly demarcated, large (>1–2 cm) lesions involving predominantly the cerebral white matter.
      • 6)
        MAGNIMS (
        • Filippi M
        • Rocca MA
        • Ciccarelli O
        • De Stefano N
        • Evangelou N
        • Kappos L
        • et al.
        MRI criteria for the diagnosis of multiple sclerosis: MAGNIMS consensus guidelines.
        ):
        • Dissemination in space: fulfill ≥2/5: (a) periventricular: ≥3 lesions, (b) cortical-juxtacortical: ≥1 lesions, (c) infratentorial: ≥1 lesions, (d) spinal cord: ≥1 lesions, (e) optic nerve: ≥1 lesions.
        • Dissemination in time: Gadolinium-enhancing and non-enhancing lesions at any time or new lesion on follow-up MRI.
      • 7)
        Verhey (
        • Verhey LH
        • Branson HM
        • Shroff MM
        • Callen DJ
        • Sled JG
        • Narayanan S
        • et al.
        MRI parameters for prediction of multiple sclerosis diagnosis in children with acute CNS demyelination: a prospective national cohort study.
        ): Fulfill 2/2:
        • a)
          ≥1 periventricular lesion
        • b)
          ≥1 T1-hypointense lesion

      2.5 Statistical analysis

      Two raters (AL and JL) both blinded to clinical data reviewed MRIs. One rater (AL) scored all images. The second rater (JL) independently scored 30 randomly selected baseline MRIs (15 children with ADEM and 15 children with MS, Fig. 1). We assessed interrater scores for MRI categories using the percent agreement and Cohen kappa scores. The interrater agreement was high for all MRI categories, apart from the presence of occipital lesion, having more than 9 T2 lesions, and the number of periventricular lesions, presumably because it is difficult to determine in children with a diffuse lesion pattern (Supplementary Table 2).
      We used forward stepwise conditional logistic regression to determine which combination of MRI categories could best separate children with MS from those with ADEM. In each step, the following binary MRI categories were considered for addition (pin<0.10) or subtraction (pout>0.10) from the regression model: simultaneous enhancing and non-enhancing lesion, T2 brain lesions (-8 vs. +9), frontal, parietal, temporal, occipital, long axis perpendicular callosal body, periventricular (0 vs. +1, -1 vs. +2, -2 vs. +3), cerebellum, brainstem, sole well-defined lesions, juxtacortical/cortical, black hole, diffuse and poorly demarcated lesions, and basal ganglia/thalamus lesion. Children with missing values were excluded from the model. First, we compared MS and ADEM (“Criteria A”). Second, we excluded MRI categories requiring contrast administration (“Criteria B”). Third, we compared MS and ADEM including encephalopathy and polyfocal neurological deficits (IPMSSG ADEM, “Criteria C”). The combination of MRI categories found to be significantly different between ADEM and MS were then used to develop our proposed criteria. We estimated accuracy statistics for our criteria and all previously proposed MRI criteria to distinguish MS from ADEM. An area under the curve of 100% represents a perfect test, and an area under the curve of 50% represents a worthless test; area under the curve can be described as 90%‒100%=excellent, 80%‒90%=good, 70%‒80%=fair, and 60%‒70%=poor.
      Analyses were performed using SAS, version 9.4 (SAS Institute Inc., Cary, NC, USA).

      2.6 Standard Protocol Approvals, Registrations, and Patient Consents

      Our study was approved by the Danish Data Protection Agency (case number: 30-1423/03567) and the Danish Health Data Authority (FSEID: 00003621/DST: 707103). The Danish Health and Medicines Authority waived the requirement for patient informed consent to access medical records and undertake MRI re-evaluation (case number: 3-3013-896/1), and chief physicians approved access to patient records from their hospital departments.

      2.7 Data sharing statement

      Any qualified investigator is welcome to contact our research group for purposes of replicating procedures and results.

      3. Results

      3.1 Demographics

      We included 67 children with MS, and 46 children with monophasic ADEM (16 children with ADEM had both encephalopathy and polyfocal neurological deficits). Children with MS were older than children with ADEM at onset, and more girls presented with MS (Table 1). Only 3% of children with MS presented before 11 years of age, and 15% had onset before 14 years of age.
      Table 1Characteristics of children with monophasic ADEM and MS at onset.
      Monophasic ADEM (n=46)MS (n=67)P-value
      Age at onset, years, median (range)5.3 (0.8‒17.2)16.3 (3.3‒17.9)<0.0001
      Age at onset before 11 years, n (%)36 (78%)2 (3%)<0.0001
      Follow-up, years, median (range)9.2 (5.5‒12.8)6.7 (5.4‒7.9)
      Girls, n (%)18 (39%)47 (70%)0.002
      Cerebrospinal fluid specific oligoclonal bands, positive/tested (% positive)0/19 (0%)39/43 (91%)<0.0001
      Myelin oligodendrocyte globulin (MOG) antibodies,

      positive/tested (% positive)
      2/4 (50%)

      1/14 (7%)

      0.11

      3.2 MRI features in children with ADEM and MS

      The median time from the first neurological symptom to the baseline MRI for the ADEM group was 7 days (interquartile range: 4 to 11 days) and for the MS group 15 days (interquartile range: 7 to 38 days). MRIs were either 1.5 or 3.0 tesla magnets with slice thicknesses ranging 3–5 mm and an interslice gap of up to 2.5 mm.
      Table 2 shows baseline MRI features in children with ADEM and MS. The following MRI features favored MS over ADEM: contrast enhancement, absence of basal ganglia/thalamus lesion, sole well-defined lesions, having two or more periventricular lesions, lesions perpendicular to the corpus callosum, absence of a diffuse lesion pattern, simultaneous enhancing and non-enhancing lesion, or optic nerve lesion. At baseline, the McDonald 2017 MRI criteria were fulfilled in 7/43 (16%) of children with ADEM and in 26/49 (53%) of children with MS. We have previously published a thorough description of MRI features at baseline and follow-up in these children (
      • Boesen MS
      • Blinkenberg M
      • Born AP
      • Magyari M
      • Chitnis T
      • Thygesen LC
      • et al.
      Magnetic resonance imaging at baseline and follow-up to differentiate between pediatric monophasic acquired CNS demyelination and MS.
      ).
      Table 2Selected baseline MRI features in children with ADEM or MS. Our proposed criteria (Criteria A, B and C) are explained in the manuscript and in Table 3, Table 4.
      Monophasic ADEM (n=46)MS (n=67)P-value
      ≥1 gadolinium-enhancing brain lesion8/39 (21%)30/49 (61%)0.0002
      ≥9 T2 lesions20/46 (43%)38/67 (57%)0.18
      ≥1 T1-hypointense lesion32/46 (70%)42/61 (69%)1.00
      ≥1 gadolinium-enhancing lesion or ≥9 T2 lesions20/46 (43%)38/67 (56%)0.18
      Basal ganglia or thalamus26/46 (57%)19/67 (28%)0.003
      Infratentorial30/46 (65%)36/67 (54%)0.25
      Sole well-defined lesions4/46 (9%)47/65 (72%)<0.0001
      Periventricular lesion (0 vs ≥1)33/46 (72%)55/67 (82%)0.25
      Periventricular lesion (0–1 vs ≥2)25/46 (54%)49/67 (73%)0.05
      Periventricular lesion (0–2 vs ≥3)16/46 (35%)42/67 (63%)0.004
      Lesion perpendicular to long axis of corpus callosum11/46 (24%)47/67 (70%)<0.0001
      Juxtacortical or cortical lesions23/46 (50%)41/67 (61%)0.25
      Diffuse, poorly demarcated, large (>1–2 cm) WM lesions31/46 (67%)12/67 (18%)<0.0001
      Simultaneous enhancing and non-enhancing lesion7/43 (16%)28/48 (58%)<0.0001
      Optic nerve lesion
      Few children were evaluated with optic nerve specific MRI sequence.
      2/26 (8%)14/41 (34%)0.02
      Black hole5/46 (11%)15/63 (24%)0.13
      Spinal cord T2 lesion11/14 (79%)29/43 (67%)0.52
      Callen criteria7/46 (15%)41/63 (65%)<0.0001
      KIDMUS criteria13/46 (28%)60/65 (92%)<0.0001
      Barkhof criteria18/46 (39%)38/67 (57%)0.09
      McDonald 2017 criteria
      Spinal MRI was not a prerequisite for the analyses.
      7/43 (16%)26/49 (53%)0.0004
      Dissemination in space (McDonald 2017)
      Spinal MRI was not a prerequisite for the analyses.
      35/46 (76%)53/67 (79%)0.82
      Dissemination in time (McDonald 2017)7/43 (16%)29/49 (59%)<0.0001
      IPMSSG criteria (REVERSED to favor MS)15/46 (33%)55/67 (82%)<0.0001
      MAGNIMS criteria
      Spinal MRI was not a prerequisite for the analyses.
      4/24 (17%)14/31 (45%)0.04
      Verhey criteria26/46 (57%)38/61 (62%)0.56
      Criteria A (fulfill ≥2/3)5/43 (12%)35/46 (76%)<0.0001
      Criteria B (fulfill ≥2/3)7/46 (15%)53/65 (82%)<0.0001
      Criteria C (fulfill ≥2/4)8/46 (17%)51/61 (84%)<0.0001
      Criteria C (fulfill ≥3/4)4/46 (10%)34/61 (56%)<0.0001
      Few children were evaluated with optic nerve specific MRI sequence.
      Spinal MRI was not a prerequisite for the analyses.

      3.3 MRI criteria to differentiate monophasic ADEM from MS

      Using all MRI variables, we identified the following three MRI categories to best distinguish monophasic ADEM (n=46) from MS (n=67): a) corpus callosum long axis perpendicular lesion, b) sole presence of well-defined lesions, c) simultaneous enhancing and non-enhancing lesion (labeled Criteria A in Table 2, Table 3, Table 4). Having at least two of these features favored MS over ADEM by 76% sensitivity, 88% specificity, 88% positive predictive value, 78% negative predictive value, and 82% area under the curve (Table 3).
      Table 3MRI criteria to favor MS (n=67) compared with all children with ADEM (“monophasic ADEM”, n=46).
      Correct classification
      Correct classification means that children with MS fulfill the criteria, and children with ADEM do not fulfill the criteria.
      Sensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)AUC (95% CI)
      ADEM (n=46)MS (n=67)
      Criteria A (fulfill ≥2/3):

      (a) sole presence of well-defined lesions

      (b) corpus callosum long axis perpendicular lesion

      (c) simultaneous enhancing and non-enhancing lesion




      38/43




      35/46




      76% (64%–88%)




      88% (79%–98%)




      88% (77%–98%)




      78% (66%–89%)




      82% (74%–90%)
      Criteria A (fulfill 3/3)42/4315/4633% (19%–46%)98% (93%–100%)94% (82%–100%)57% (46%–67%)65% (58%–72%)
      Criteria B (fulfill ≥2/3):

      (a) sole presence of well-defined lesions

      (b) corpus callosum long axis perpendicular lesion

      (c) absence of basal ganglia or thalamus lesion




      39/46




      53/65




      82% (72%–91%)




      85% (74%–95%)




      88% (80%–96%)




      76% (65%–88%)




      83% (76%–90%)
      Criteria B (fulfill 3/3)45/4623/6535% (24%–47%)98% (94%–100%)96% (88%–100%)52% (41%–62%)67% (60%–73%)
      Criteria C (fulfill ≥2/4):

      (a) sole presence of well-defined lesions

      (b) corpus callosum long axis perpendicular lesion

      (c) absence of diffuse, poorly demarcated, large lesions

      (d) black holes




      38/46




      51/61




      84% (74%–93%)




      83% (72%–94%)




      86% (78%–95%)




      79% (68%–91%)




      83% (76%–90%)
      Criteria C (fulfill ≥3/4)42/4634/6156% (43%–68%)91% (83%–99%)89% (80%–99%)61% (49%–72%)74% (66%–81%)
      Callen39/4641/6365% (53%–77%)85% (74%–95%)85% (75%–95%)64% (52%–76%)75% (69%–84%)
      KIDMUS33/4660/6592% (86%–99%)72% (59%–85%)82% (73%–91%)87% (76%–98%)82% (75%–89%)
      Barkhof28/4638/6757% (45%–69%)61% (47%–75%)68% (56%–80%)49% (36%–62%)59% (49%–68%)
      McDonald 2017
      Spinal MRI was not a prerequisite for the analysis. Accordingly, children without a spinal MRI had zero in this category the dissemination in space calculations.
      36/4325/4852% (38%–66%)84% (73%–95%)78% (64%–92%)61% (49%–73%)68% (59%–77%)
      IPMSSG (REVERSED to favor MS)31/4655/6782% (73%–91%)67% (54%–81%)79% (69%–88%)72% (59%–86%)75% (66%–83%)
      MAGNIMS
      Spinal MRI was not a prerequisite for the analysis. Accordingly, children without a spinal MRI had zero in this category the dissemination in space calculations.
      20/2414/3145% (28%–63%)83% (68%–98%)78% (59%–97%)54% (38%–70%)64% (53%–76%)
      Verhey20/4638/6162% (50%–74%)43% (29%–58%)59% (47%–71%)47% (32%–61%)53% (43%–62%)
      Our proposed criteria (Criteria A, B, C) are followed by all previously proposed MRI criteria to differentiate MS and ADEM. First, we modelled all MRI categories (“Criteria A”). Second, we excluded MRI variables that need contrast administration (“Criteria B”). Third, we compared MS and ADEM including encephalopathy and polyfocal neurological deficits (IPMSSG ADEM, “Criteria C”). Differences between our criteria are highlighted in italic.
      Abbreviations: AUC, area under the curve; CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value.
      Correct classification means that children with MS fulfill the criteria, and children with ADEM do not fulfill the criteria.
      Spinal MRI was not a prerequisite for the analysis. Accordingly, children without a spinal MRI had zero in this category the dissemination in space calculations.
      Table 4MRI criteria to favor MS (n=67) compared with “IPMSSG ADEM” (n=16).
      Correct classification
      Correct classification means that children with MS fulfill the criteria, and children with ADEM do not fulfill the criteria.
      Sensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)AUC (95% CI)
      IPMSSG ADEM (n=16)MS (n=67)
      Criteria A (fulfill ≥2/3):

      (a) sole presence of well-defined lesions

      (b) corpus callosum long axis perpendicular lesion

      (c) simultaneous enhancing and non-enhancing lesion




      13/15




      35/46




      76% (64%–88%)




      87% (69%–100%)




      95% (87%–100%)




      54% (34%–74%)




      81% (71%–92%)
      Criteria B (fulfill ≥2/3):

      (a) sole presence of well-defined lesions

      (b) corpus callosum long axis perpendicular lesion

      (c) absence of basal ganglia or thalamus lesion




      15/16




      53/65




      82% (72%–91%)




      94% (82%–100%)




      98% (95%–100%)




      56% (37%–74%)




      88% (80%–95%)
      Criteria C (fulfill ≥2/4):

      (a) sole presence of well-defined lesions

      (b) corpus callosum long axis perpendicular lesion

      (c) absence of diffuse, poorly demarcated, large lesions

      (d) black holes




      15/16




      51/61




      84% (74%–93%)




      94% (82%–100%)




      98% (94%–100%)




      60% (41%–79%)




      89% (81%–96%)
      Criteria C (fulfill ≥3/4)16/1634/6156% (43%–68%)100% (100%–100%)100% (100%–100%)37% (23%–52%)78% (72%–84%)
      Callen15/1641/6365% (53%–77%)94% (82%–100%)98% (93%–100%)41% (25%–56%)79% (71%–88%)
      KIDMUS12/1660/6592% (86%–99%)75% (54%–96%)94% (88%–100%)71% (49%–92%)84% (72%–95%)
      Barkhof8/838/6757% (45%–69%)50% (26%–75%)83% (72%–94%)22% (8%–35%)53% (39%–67%)
      McDonald 2017
      Spinal MRI was not a prerequisite for the analysis. Accordingly, children without a spinal MRI had zero in this category the dissemination in space calculations.
      11/1625/4852% (38%–66%)73% (51%–96%)86% (74%–99%)32% (17%–48%)63% (49%–76%)
      IPMSSG (REVERSED to favor MS)13/1655/6782% (73%–91%)81% (62%–100%)95% (89%–100%)52% (32%–72%)82% (71%–92%)
      MAGNIMS
      Spinal MRI was not a prerequisite for the analysis. Accordingly, children without a spinal MRI had zero in this category the dissemination in space calculations.
      5/714/3145% (28%–63%)71% (38%–100%)88% (71%–100%)23% (5%–40%)58% (38%–78%)
      Verhey11/1638/6162% (50%–74%)69% (46%–91%)88% (79%–98%)32% (17%–48%)66% (52%–79%)
      “IPMSSG ADEM” means that all children had encephalopathy and polyfocal neurological deficits and no prior demyelinating events. Our proposed criteria (Criteria A, B, C) are followed by all previously proposed MRI criteria to differentiate MS and ADEM. Differences between our criteria are highlighted in italic.
      Abbreviations: AUC, area under the curve; CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value.
      Correct classification means that children with MS fulfill the criteria, and children with ADEM do not fulfill the criteria.
      Spinal MRI was not a prerequisite for the analysis. Accordingly, children without a spinal MRI had zero in this category the dissemination in space calculations.
      Excluding contrast dependent variables from the model, “simultaneous enhancing and non-enhancing lesion” was exchanged for “absence of basal ganglia or thalamus lesion” (labeled Criteria B in Table 2, Table 3, Table 4). Criteria A and B showed similar performance. Previously proposed Callen, KIDMUS, and IPMSSG criteria also performed well with areas under the curve ranging 75%–82%, whereas the McDonald 2017, Barkhof, MAGNIMS, and Verhey criteria had poorer performance with areas under the curve ranging 53%–68% (Table 3).

      3.4 MRI criteria to differentiate “IPMSSG ADEM” from MS

      In children with ADEM including encephalopathy and polyfocal neurological deficits (“IPMSSG ADEM”, n=16) versus MS (n=67), the two categories “corpus callosum long axis perpendicular lesion” and “sole presence of well-defined lesions” remained in the model but now “absence of diffuse and poorly demarcated lesions” and “black holes” entered the model (labeled Criteria C in Table 2, Table 3, Table 4). Having at least two of these four features had 84% sensitivity, 94% specificity, 98% positive predictive value, 60% negative predictive value, and 89% area under the curve to distinguish MS from IPMSSG ADEM (Table 4). The following criteria had better performance in distinguishing MS from “IPMSSG ADEM” than MS versus “monophasic ADEM”: Callen, KIDMUS, and our “Criteria B” (fulfill ≥2/3). The McDonald 2017, Barkhof, MAGNIMS, and Verhey criteria continued to have poorer performance.

      4. Discussion

      We propose two sets of MRI criteria to distinguish the first demyelinating attack in MS from monophasic ADEM. The first criteria with high performance were having at least two of the following: (a) corpus callosum long axis perpendicular lesion; (b) sole presence of well-defined lesions; (c) absence of basal ganglia or thalamus lesion. The other MRI criteria with high performance were having at least two of the following: a) sole presence of well-defined lesions; (b) corpus callosum long axis perpendicular lesion; (c) absence of diffuse, poorly demarcated, large lesions; (d) black holes. Previously proposed Callen, KIDMUS, and IPMSSG criteria also performed well. However, the McDonald 2017, Barkhof, MAGNIMS, and Verhey criteria had poorer performance.
      The study has several strengths: 1) The ADEM and MS cohorts were large with precise case ascertainment by review of the medical records; 2) no children were lost to follow-up; 3) MRIs were scored blinded to clinical diagnosis; 4) we used two MRI raters to assess interrater scores; and 5) children with ADEM were followed for at least five years to ensure a monophasic disease course. The following limitations need to be addressed. 1) Few children with ADEM were tested for antibodies against myelin oligodendrocyte glycoprotein, but previous MRI studies have not stratified based on these antibodies (
      • Callen DJA
      • Shroff MM
      • Branson HM
      • Li DK
      • Lotze T
      • Stephens D
      • et al.
      Role of MRI in the differentiation of ADEM from MS in children.
      ;
      • Mikaeloff Y
      • Adamsbaum C
      • Husson B
      • Vallée L
      • Ponsot G
      • Confavreux C
      • et al.
      MRI prognostic factors for relapse after acute CNS inflammatory demyelination in childhood.
      ;
      • Verhey LH
      • Branson HM
      • Shroff MM
      • Callen DJ
      • Sled JG
      • Narayanan S
      • et al.
      MRI parameters for prediction of multiple sclerosis diagnosis in children with acute CNS demyelination: a prospective national cohort study.
      ). 2) Not all children with ADEM had encephalopathy, but results were similar when using criteria to distinguish MS from ADEM (Table 3) or IPMSSG ADEM (Table 4). 3) MRI protocols differed.
      Our proposed MRI criteria combine the KIDMUS, IPMSSG, and Callen criteria that also performed well. The KIDMUS group studied 116 children (<16 years) with a first episode of acute CNS demyelination (
      • Mikaeloff Y
      • Adamsbaum C
      • Husson B
      • Vallée L
      • Ponsot G
      • Confavreux C
      • et al.
      MRI prognostic factors for relapse after acute CNS inflammatory demyelination in childhood.
      ). During follow-up, 52/116 (45%) experienced a second demyelinating episode, and corpus callosum long axis perpendicular lesion and sole presence of well-defined lesions increased the hazard ratio for relapse by 2.89 (95% CI=1.65–5.06) and 1.71 (95% CI=1.29–2.27), respectively. The IPMSSG criteria have never been validated but they include diffuse, poorly demarcated, large (>1–2 cm) cerebral white matter lesions and deep grey matter lesions (e.g. thalamus or basal ganglia) (
      • Krupp LB
      • Tardieu M
      • Amato MP
      • Banwell B
      • Chitnis T
      • Dale RC
      • et al.
      International Pediatric Multiple Sclerosis Study Group criteria for pediatric multiple sclerosis and immune-mediated central nervous system demyelinating disorders: revisions to the 2007 definitions.
      ).
      • Callen DJA
      • Shroff MM
      • Branson HM
      • Li DK
      • Lotze T
      • Stephens D
      • et al.
      Role of MRI in the differentiation of ADEM from MS in children.
      compared 28 children with MS and at least two clinical relapses with 20 children with encephalopathic ADEM and a monophasic disease course at the 2-year follow-up. They found that having two of the following features distinguished pediatric MS from monophasic ADEM with 81% sensitivity and 95% specificity: 1) absence of diffuse bilateral lesions, 2) presence of black holes, and 3) presence of two or more periventricular lesions. We corroborate their findings, showing that children with ADEM more frequently have a diffuse bilateral lesion pattern and are less likely to have two or more periventricular lesions; however, presence of black holes did not differ between the groups in our study (11% in ADEM versus 24% in MS, p=0.13, Table 2). In addition, the MS group in Callen's study differed from our MS group by having a younger age at onset (39% had MS onset before 10 years of age versus 3% in our MS cohort) and a higher frequency of black holes at onset (58% had black holes versus 23% in our MS cohort). Accordingly, Callen's MS cohort may be inflicted by referral bias and is not representative of children with MS in Denmark (
      • Boesen MS
      • Magyari M
      • Koch-Henriksen N
      • Thygesen LC
      • Born AP
      • Uldall PV
      • et al.
      Pediatric-onset multiple sclerosis and other acquired demyelinating syndromes of the central nervous system in Denmark during 1977-2015: A nationwide population-based incidence study.
      ;
      • Boesen MS
      • Sellebjerg F
      • Blinkenberg M.
      Onset symptoms in paediatric multiple sclerosis.
      ). The Barkhof criteria did not perform well because ADEM and MS did not differ regarding juxtacortical/cortical and infratentorial lesions (Table 2) (
      • Barkhof F
      • Filippi M
      • Miller DH
      • Scheltens P
      • Campi A
      • Polman CH
      • et al.
      Comparison of MRI criteria at first presentation to predict conversion to clinically definite multiple sclerosis.
      ). In line with this, the McDonald 2017 criteria performed poorly because dissemination in space was seen in 76% of children with ADEM and in 79% of children with MS on baseline MRI (
      • Thompson AJ
      • Banwell BL
      • Barkhof F
      • Carroll WM
      • Coetzee T
      • Comi G
      • et al.
      Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
      ). The MAGNIMS criteria resemble the McDonald 2017 criteria, except that criteria for dissemination in space also include at least three periventricular lesions, and optic neuritis (
      • Filippi M
      • Rocca MA
      • Ciccarelli O
      • De Stefano N
      • Evangelou N
      • Kappos L
      • et al.
      MRI criteria for the diagnosis of multiple sclerosis: MAGNIMS consensus guidelines.
      ). Both MRI categories were more common in children with MS, but only 45% of children with MS fulfilled the MAGNIMS criteria on baseline MRI. Further, visualization of optic neuritis may require specific sequences that are not part of routine diagnostic work-up in children with ADEM or MS in Denmark. The Verhey criteria performed poorly because 72% of children with ADEM had at least one periventricular lesion and because 70% had at least one T1 hypointense lesion (
      • Verhey LH
      • Branson HM
      • Shroff MM
      • Callen DJ
      • Sled JG
      • Narayanan S
      • et al.
      MRI parameters for prediction of multiple sclerosis diagnosis in children with acute CNS demyelination: a prospective national cohort study.
      ).
      Taken together, the study adds important knowledge to the evaluation of MRI criteria in differentiating ADEM from MS and proposes criteria that increase the discriminative value. However, the best performing MRI criteria reached an area under the curve of only 83% (MS versus ADEM) and 89% (MS versus ADEM including encephalopathy and polyfocal neurological deficits). Two interesting observations are that age at onset and presence of oligoclonal bands differed considerably in the two groups (Table 1) (
      • Boesen MS
      • Magyari M
      • Koch-Henriksen N
      • Thygesen LC
      • Born AP
      • Uldall PV
      • et al.
      Pediatric-onset multiple sclerosis and other acquired demyelinating syndromes of the central nervous system in Denmark during 1977-2015: A nationwide population-based incidence study.
      ;
      • Neuteboom RF
      • Boon M
      • Catsman Berrevoets CE
      • Vles JS
      • Gooskens RH
      • Stroink H
      • et al.
      Prognostic factors after a first attack of inflammatory CNS demyelination in children.
      ;
      • Boesen MS
      • Born AP
      • Jensen PEH
      • Sellebjerg F
      • Blinkenberg M
      • Lydolph MC
      • et al.
      Diagnostic Value of Oligoclonal Bands in Children: A Nationwide Population-Based Cohort Study.
      ). Accordingly, in a future study, we will propose criteria to differentiate ADEM from MS based on MRI as well as other clinical and paraclinical data.

      Funding

      The study was supported by the Danish MS Society (grant numbers A29625 , A31526 , A33178 , A35179 , A38303 , A39760 ), Dagmar Marshalls Fond, Axel Muusfeldts Fond, Bent Bøgh og Hustrus Fond, and Helene og Viggo Bruuns Fond.

      Author contributions

      Author initials in parentheses: 1) conception and design of the study (MSB, MB, AL); 2) acquisition and analysis of data (MSB, AL, MB, JL); 3) drafting a significant portion of the manuscript (MSB); 4) revision of manuscript (MSB, MB, AL, LCT, JL).

      Conflicts of Interest

      Dr. Boesen has served on a scientific advisory board for Teva; has received speaker honoraria for lecturing from Novartis, and support for congress participation from Teva, Novartis and Roche. Dr. Blinkenberg has served on scientific advisory boards for Genzyme, Roche, Biogen, Merck, Novartis and Teva; has received speaker honoraria from Genzyme, Biogen, Merck, Novartis, Teva and Roche; has received consulting honoraria from the Danish Multiple Sclerosis Society, Biogen, Teva, Roche and Merck; and has received funding for travel from Genzyme, Roche and Biogen. Drs. Langkilde, Thygesen, and Ilginiene report no disclosures.

      Appendix. Supplementary materials

      References

        • Andersen TF
        • Madsen M
        • Jørgensen J
        • Mellemkjoer L
        • Olsen JH.
        The Danish National Hospital Register. A valuable source of data for modern health sciences.
        Dan Med Bull. 1999 Jun; 46: 263-268
        • Barkhof F
        • Filippi M
        • Miller DH
        • Scheltens P
        • Campi A
        • Polman CH
        • et al.
        Comparison of MRI criteria at first presentation to predict conversion to clinically definite multiple sclerosis.
        Brain J Neurol. 1997 Nov; 120: 2059-2069
        • Boesen MS
        • Blinkenberg M
        • Born AP
        • Magyari M
        • Chitnis T
        • Thygesen LC
        • et al.
        Magnetic resonance imaging at baseline and follow-up to differentiate between pediatric monophasic acquired CNS demyelination and MS.
        Mult Scler Relat Disord. 2020 Nov 1; 46102590
        • Boesen MS
        • Born AP
        • Jensen PEH
        • Sellebjerg F
        • Blinkenberg M
        • Lydolph MC
        • et al.
        Diagnostic Value of Oligoclonal Bands in Children: A Nationwide Population-Based Cohort Study.
        Pediatr Neurol. 2019; 97: 56-63
        • Boesen MS
        • Langkilde A
        • Born AP
        • Magyari M
        • Blinkenberg M
        • Chitnis T
        • et al.
        School performance and psychiatric morbidity 6 years after pediatric acute disseminated encephalomyelitis: A nationwide population-based cohort study.
        Mult Scler Relat Disord. 2019 Nov; 36101425
        • Boesen MS
        • Magyari M
        • Born AP
        • Thygesen LC.
        Pediatric acquired demyelinating syndromes: a nationwide validation study of the Danish National Patient Register.
        Clin Epidemiol. 2018; 2018: 391-399
        • Boesen MS
        • Magyari M
        • Koch-Henriksen N
        • Thygesen LC
        • Born AP
        • Uldall PV
        • et al.
        Pediatric-onset multiple sclerosis and other acquired demyelinating syndromes of the central nervous system in Denmark during 1977-2015: A nationwide population-based incidence study.
        Mult Scler J. 2018 Jul; 24: 1077-1086
        • Boesen MS
        • Magyari M
        • Koch-Henriksen N
        • Uldall PV
        • Thygesen LC
        • Blinkenberg M
        • et al.
        Implications of the International Paediatric Multiple Sclerosis Study Group consensus criteria for paediatric acute disseminated encephalomyelitis: a nationwide validation study.
        Dev Med Child Neurol. 2018 Nov 1; 60: 1123-1131
        • Boesen MS
        • Sellebjerg F
        • Blinkenberg M.
        Onset symptoms in paediatric multiple sclerosis.
        Dan Med J. 2014 Apr; 61: A4800
        • Callen DJA
        • Shroff MM
        • Branson HM
        • Li DK
        • Lotze T
        • Stephens D
        • et al.
        Role of MRI in the differentiation of ADEM from MS in children.
        Neurology. 2009 Mar 17; 72: 968-973
        • Filippi M
        • Rocca MA
        • Ciccarelli O
        • De Stefano N
        • Evangelou N
        • Kappos L
        • et al.
        MRI criteria for the diagnosis of multiple sclerosis: MAGNIMS consensus guidelines.
        Lancet Neurol. 2016 Mar; 15: 292-303
        • Ketelslegers IA
        • Neuteboom RF
        • Boon M
        • Catsman-Berrevoets CE
        • Hintzen RQ
        Dutch Pediatric MS Study Group. A comparison of MRI criteria for diagnosing pediatric ADEM and MS.
        Neurology. 2010 May 4; 74: 1412-1415
        • Krupp LB
        • Tardieu M
        • Amato MP
        • Banwell B
        • Chitnis T
        • Dale RC
        • et al.
        International Pediatric Multiple Sclerosis Study Group criteria for pediatric multiple sclerosis and immune-mediated central nervous system demyelinating disorders: revisions to the 2007 definitions.
        Mult Scler. 2013 Sep; 19: 1261-1267
        • Lynge E
        • Sandegaard JL
        • Rebolj M.
        The Danish National Patient Register.
        Scand J Public Health. 2011 Jul; 39 (Suppl): 30-33
        • Mikaeloff Y
        • Adamsbaum C
        • Husson B
        • Vallée L
        • Ponsot G
        • Confavreux C
        • et al.
        MRI prognostic factors for relapse after acute CNS inflammatory demyelination in childhood.
        Brain J Neurol. 2004 Sep; 127: 1942-1947
        • Neuteboom RF
        • Boon M
        • Catsman Berrevoets CE
        • Vles JS
        • Gooskens RH
        • Stroink H
        • et al.
        Prognostic factors after a first attack of inflammatory CNS demyelination in children.
        Neurology. 2008 Sep 23; 71: 967-973
        • Pedersen CB.
        The Danish Civil Registration System.
        Scand J Public Health. 2011 Jul; 39 (Suppl): 22-25
        • Polman CH
        • Reingold SC
        • Banwell B
        • Clanet M
        • Cohen JA
        • Filippi M
        • et al.
        Diagnostic criteria for multiple sclerosis: 2010 Revisions to the McDonald criteria.
        Ann Neurol. 2011 Feb 1; 69: 292-302
        • Schmidt M
        • Pedersen L
        • Sørensen HT.
        The Danish Civil Registration System as a tool in epidemiology.
        Eur J Epidemiol. 2014 Aug; 29: 541-549
        • Thompson AJ
        • Banwell BL
        • Barkhof F
        • Carroll WM
        • Coetzee T
        • Comi G
        • et al.
        Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
        Lancet Neurol. 2018 Feb; 17: 162-173
        • Thygesen LC
        • Daasnes C
        • Thaulow I
        • Brønnum-Hansen H.
        Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving.
        Scand J Public Health. 2011 Jul; 39 (Suppl): 12-16
        • Verhey LH
        • Branson HM
        • Shroff MM
        • Callen DJ
        • Sled JG
        • Narayanan S
        • et al.
        MRI parameters for prediction of multiple sclerosis diagnosis in children with acute CNS demyelination: a prospective national cohort study.
        Lancet Neurol. 2011 Dec; 10: 1065-1073