International consensus diagnostic criteria for neuromyelitis optica spectrum disorders

Consensus diagnostic criteria.

Neuroimaging and neurophysiologic testing.

MRI lesion patterns are a major arbiter of CNS demyelinating disease differential diagnosis. Several brain, optic nerve, and spinal cord patterns are characteristic or highly suggestive of NMOSD (table 3; figures 1–3). Detection of a LETM spinal cord lesion (figure 1) associated with acute myelitis is the most specific neuroimaging characteristic of NMOSD and is very uncommon in adult MS.² Such lesions typically involve the central gray matter and may be associated with cord swelling, central hypointensity on T1-weighted sequences, and enhancement following IV gadolinium administration; extension of a cervical lesion into the brainstem is characteristic.^{4,32,33,e35,e64–e67} In contrast, MS cord lesions are usually about 1 vertebral segment long or less, occupy peripheral white matter tracts such as the dorsal columns, and may be asymptomatic (see table 3 for neuroimaging red flags relevant to MS and other diseases).^{34,–,36,e68–e70} Although the LETM pattern is characteristic of NMOSD, 7%–14% of initial and 8% of subsequent myelitis attacks in AQP4-IgG-seropositive patients do not meet the LETM definition.^32,37 Therefore, NMOSD must be considered in the differential diagnosis in patients presenting with short myelitis lesions. The timing of an individual MRI scan must be correlated with clinical status. Occasionally, lesions of less than 3 segments are detected in NMOSD because the MRI was performed early in the evolution of acute myelitis or in clinical remission, during which a LETM lesion may fragment into discontinuous lesions.^37,e71 Some patients with progressive MS have coalescent cord lesions that can superficially suggest a LETM pattern^e72; both axial and sagittal plane images should be used to judge lesion extent. A LETM MRI pattern may also occur in patients with infectious, granulomatous, neoplastic, and paraneoplastic diseases, acute disseminated encephalomyelitis (ADEM), spinal cord infarction, and dural arteriovenous fistula.^e63

View this table:

• View inline
• View popup
• Download powerpoint

Table 3

Neuroimaging characteristics of NMOSD

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Spinal cord and optic nerve MRI patterns in neuromyelitis optica spectrum disorder

Spinal cord imaging in the context of acute myelitis in neuromyelitis optica spectrum disorders (NMOSD) usually reveals a longitudinally extensive transverse myelitis (LETM) lesion extending over 3 or more vertebral segments. Sagittal T2-weighted MRI of the thoracic spinal cord (A) demonstrates a typical LETM lesion involving most of the thoracic spinal cord (arrows). LETM lesions have a predilection for the central cord, as shown by axial T2-weighted (B; arrowhead) and T1-weighted MRI with gadolinium (C; arrowhead). Cervical LETM may extend into the medulla, a characteristic NMOSD pattern demonstrated in D (arrows; sagittal T2-weighted MRI) and E (arrows; sagittal T1-weighted MRI with gadolinium). Acute LETM lesions can be associated with intralesional hypointensity as shown by sagittal T1-weighted MRI (F; arrow); in this example, a rim of gadolinium enhancement surrounds the hypointense region. Chronic sequelae of LETM may include longitudinally extensive segments of spinal cord atrophy as shown by T2-weighted MRI using sagittal (G; the 2 arrowheads indicate the atrophic segment and the top arrow indicates the normal diameter of unaffected cervical spinal cord) and axial planes (H; arrowhead shows an atrophic spinal cord). Fast spin echo fat-suppressed T2-weighted MRI in the axial (I) and coronal (J) planes shows increased signal throughout most the length of the left optic nerve, especially its posterior portion (arrows). Axial T1-weighted MRI with gadolinium shows enhancement of the optic chiasm (K; arrows). These images are from 2 different patients experiencing acute optic neuritis in the setting of NMOSD.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Dorsal medulla, area postrema, and other brainstem lesions in neuromyelitis optica spectrum disorder

Sagittal T2-weighted fluid-attenuated inversion recovery (FLAIR) MRI shows a lesion in the dorsal medulla (A; arrow). Sagittal T2-weighted (B) and T1-weighted MRI with gadolinium (C) each demonstrate an acute lesion (arrows) associated with area postrema clinical syndrome. Axial T2-weighted FLAIR (D; arrows) and T1-weighted MRI with gadolinium (E; arrowheads) show dorsal medulla involvement in a patient with acute area postrema clinical syndrome. Axial T2-weighted FLAIR MRI shows periependymal lesions involving the pons (F; arrows) and dorsal midbrain (G; arrow). Sagittal T2-weighted FLAIR MRI shows increased signal surrounding the fourth ventricle (H; arrows).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Diencephalic and cerebral lesions in neuromyelitis optica spectrum disorder

A variety of brain lesion patterns are associated with neuromyelitis optica spectrum disorder. Axial T2-weighted fluid-attenuated inversion recovery (FLAIR) MRI from 2 patients demonstrates lesions involving the right thalamus (A; arrow) and the hypothalamus (B; arrows). Axial T2-weighted FLAIR MRI shows an extensive subcortical white matter lesion (C; arrow) that enhances after gadolinium administration on T1-weighted sequences (D; arrow). Chronic longitudinally extensive and linear corpus callosum lesions are depicted on sagittal T2-weighted FLAIR MRI (E; arrows). Coronal T2-weighted FLAIR MRI shows longitudinal involvement of the corticospinal tract extending to the cerebral peduncle and pons (F; arrows). Acute periependymal cerebral lesions from one patient are depicted using sagittal (G; arrow) and axial (H; arrows) T2-weighted FLAIR MRI and axial T1-weighted MRI with gadolinium (I; arrows).

During optic neuritis attacks, increased signal within the optic nerve may be detected with fat-suppressed T2-weighted orbital MRI sequences, typically with gadolinium enhancement seen on T1-weighted sequences (figure 1). Bilateral optic nerve involvement, posterior nerve predominance (especially with extension into the optic chiasm), or extensive lesions of the optic nerve (more than half of its length) are all suggestive of NMOSD (table 3).^e51,e73,e74

According to the 2006 diagnostic scheme, a normal brain MRI or detection of only nonspecific white matter lesions was a key supportive criterion.² Sixty percent of patients with NMOSD accumulate asymptomatic white matter lesions on longitudinal study and as many as 16% fulfill Barkhof MS MRI criteria.^{33,38,39,e75–e76} Detection of brain MRI white matter lesions compatible with MS does not exclude the diagnosis of NMOSD but is considered a red flag, indicating that additional evidence may be required to confidently distinguish NMOSD from MS in individual cases.^{33,38,e64,e76,e77} Analysis of qualitative features of lesions assists in distinguishing NMOSD from MS. First, at least 7% of patients have NMOSD-typical patterns that, although not pathognomonic for NMOSD, are considered exceptional for MS (table 3),³⁹ including lesions of the dorsal medulla/area postrema (figure 2)^{22,33,40,e36,e73,e78} and periependymal regions in the brainstem (figure 2)^{33,40,e36,e73} and diencephalic structures,^39,40,e79 or cerebral hemispheres^e80 (figure 3), or long lesions spanning much of the length of the corpus callosum^39,41 or corticospinal tracts (figure 3).^33,34 Large, confluent, or tumefactive cerebral lesions may suggest NMOSD but in isolation may not be discernable from atypical MS lesions, especially in AQP4-IgG-seronegative patients.^{27,33,39,42,e46,e81−e83} Second, some MRI lesion patterns considered typical of MS are rarely seen in NMOSD, including perpendicular orientation of periventricular lesions (Dawson fingers), periventricular lesions located in the inferior temporal lobe, and cortical lesions.^33,38,43,e64 Recent 7T MRI studies revealed differences between NMOSD and MS specifically regarding cortical lesions (frequent in MS but absent in NMOSD) and white matter lesions (MS plaques are periventricular and traversed by a central venule whereas NMOSD lesions are subcortical and lack a central venule).^e84,e85 However, this technology is not widely available and further validation is needed.

The Panel considered evidence for use of other imaging modalities, such as nonconventional MRI techniques and optical coherence tomography, and neurophysiologic tests such as visual evoked potentials. None of these modalities were included in the revised diagnostic criteria because of concerns regarding lack of evidence, specificity, or reliability.

Considerations for AQP4-IgG serologic and other laboratory testing.

Technological advances in AQP4-IgG assays have improved diagnostic sensitivity without compromising specificity.⁴⁴ The Panel recommended testing with cell-based serum assays (microscopy or flow cytometry-based detection) whenever possible because they optimize autoantibody detection (mean sensitivity 76.7% in a pooled analysis; 0.1% false-positive rate in a MS clinic cohort).^{12,–,14,44,45} However, cell-based assays are not yet widely available.

Indirect immunofluorescence assays and ELISAs have lower sensitivity (mean sensitivity 63%–64% each) and occasionally yield false-positive results (0.5%–1.3% for ELISA), often at low titer.^{12,–,14,44,45,e86} The Panel strongly recommended interpretative caution if such assays are used and when low-titer positive ELISA results are detected in individuals who present with NMOSD clinical symptoms less commonly associated with AQP4-IgG (e.g., presentations other than recurrent optic neuritis, myelitis with LETM, or area postrema syndrome) or in situations where clinical evidence suggests a viable alternate diagnosis. Confirmatory testing is recommended, ideally using 1 or more different AQP4-IgG assay techniques. Cell-based assay has the best current sensitivity and specificity and samples may need to be referred to a specialized laboratory.

Patients who fulfill NMOSD criteria but do not have detectable AQP4-IgG despite use of the best available assays, or for whom serologic testing is unavailable, sometimes represent a diagnostic challenge.⁴⁶ A greater proportion of patients with NMO who have monophasic disease appear to be AQP4-IgG-seronegative compared to those with established relapsing disease.³⁷ A minority of patients with clinical characteristics of NMO, almost all AQP4-IgG-seronegative, have been reported to have detectable serum myelin oligodendrocyte glycoprotein (MOG) antibodies and might have different characteristics (younger age, less frequently female, and less likely to relapse) from those with AQP4-IgG.^{47,–,49,e87} These findings might suggest that some AQP4-IgG-seronegative patients with clinical and neuroimaging features of NMOSD have a different underlying pathogenesis⁴⁶; the role of MOG or other antibodies in disease pathogenesis remains undetermined.^e88 The IPND nomenclature is designed to allow incorporation of cases of NMOSD associated with other validated biomarkers (e.g., NMOSD with a specific autoantibody).

Occasionally, patients without detectable serum AQP4-IgG are later found to be seropositive. There may be technical explanations in some cases but antibody levels also increase with clinical relapses and decrease with immunosuppressive therapy in some patients.⁵⁰ Therefore, retesting should be considered before B-cell or antibody-targeted therapies (plasma exchange, immunosuppressive drugs) are instituted and in seronegative patients who relapse.⁵¹ Cases of clinical NMO in which AQP4-IgG was detected in CSF, but not serum, have been reported but this appears to be rare.^52,e89–e91 Routine CSF testing of AQP4-IgG-seronegative patients is not recommended but might be considered in selected seronegative cases, especially those with additional confounding serum autoantibodies that may lead to uninterpretable or false-positive assay results.

The Panel considered absence of CSF oligoclonal bands as supportive evidence for NMOSD (although they are sometimes transiently detectable at the time of an attack) and presence of bands a red flag, but sensitivity and specificity are modest.^4,53 CSF pleocytosis >50 leukocytes/μL (incidence approximately 35% in NMOSD) or the presence of neutrophils or eosinophils (either >5/μL; incidence 44% and 10%, respectively, in attack-associated samples) are particularly useful in distinguishing NMOSD from MS.^4,53 CSF glial fibrillary acidic protein also shows promise as a diagnostic and prognostic biomarker but is elevated only for days to weeks following an attack.^e92,e93

Pediatric NMOSD criteria.

The Pediatric Working Group members noted that most clinical, neuroimaging, and laboratory characteristics of pediatric NMOSD are similar to those of adult-onset disease. The female preponderance may be of lower magnitude (∼3:1 female:male ratio compared with up to 9:1 for adults),^54,e94 a greater proportion of children may have monophasic disease,^54,55,e95 and acute CSF abnormalities in pediatric MS may mimic those ordinarily considered suggestive of NMOSD.^54,–,56 The Working Group identified caveats to application of the adult NMOSD criteria in children, noting especially that detection of a LETM MRI lesion associated with acute myelitis may be less specific for NMOSD. Approximately 15% of children with MS may have LETM during relapse, LETM can accompany monophasic ADEM, and AQP4-IgG is rarely detected in children with monophasic LETM.⁵⁵ In children diagnosed with ADEM according to international consensus criteria, which require a polyfocal demyelinating clinical presentation with encephalopathy,⁵⁷ the presence of AQP4-IgG favors a diagnosis of NMOSD, although prospective validation studies are required to determine risk of recurrent NMOSD events. In one study of AQP4-IgG-seropositive children, 45% experienced recurrent cerebral manifestations including encephalopathy.⁵⁶ Aside from these caveats, the currently proposed diagnostic criteria for NMOSD are appropriate for pediatric patients. Longitudinal observation of the clinical course for dissemination in time and retesting the AQP4-IgG status of some children, especially AQP4-IgG-seronegative individuals presenting with an ADEM-like event that includes optic neuritis and LETM, may be required to achieve confident diagnosis.

Monophasic NMOSD.

The occurrence of a second clinical attack, which defines a relapsing disorder, was often considered sufficient to revise the diagnosis to MS in otherwise typical NMO cases. Approximately 5%–10% of contemporary cases are described as monophasic, although the optimal definition for monophasic NMOSD remains elusive. It is unclear whether the occurrence of bilateral optic neuritis and myelitis at initial presentation are helpful or essential to distinguish such cases from relapsing NMOSD. Similarly, the interval between index clinical events that is compatible with monophasic NMO has not been consistently defined or adequately examined as a predictor of future clinical course. Although early risk of relapse is high in AQP4-IgG-seropositive cases (e.g., about 60% within 1 year after LETM),²⁰ cases have been documented in whom more than a decade elapsed between the index events and relapse.^e34

Several studies show a pattern of apparently monophasic NMO being associated with a more equitable sex distribution, relatively younger age at disease onset, tendency to present with simultaneous myelitis and bilateral optic neuritis (rather than unilateral optic nerve involvement), lower frequency of other autoimmune diseases, and lower prevalence of serum AQP4-IgG compared to relapsing NMO.^4,37,e96 A fraction of these patients may have other serum antibodies such as MOG-IgG.⁴⁷

The Panel concluded that monophasic NMOSD is a recognizable clinical entity but that criteria that accurately predict long-term adherence to a monophasic course cannot currently be defined. An interval longer than 4 weeks between index attacks indicates relapsing disease. The Panel also recommended that at least 5 years (preferably longer) of relapse-free clinical observation after the index events be required before a monophasic course is assumed with any degree of confidence. Patients who are AQP4-IgG-seropositive should be assumed to be at risk for relapse indefinitely and preventive treatment should be considered, even in the setting of a prolonged clinical remission.

Systemic autoimmunity associated with NMOSD.

Based on evidence from several observational studies, the Panel concluded that clinical diagnoses of SLE, SS, or myasthenia gravis may coexist with NMOSD clinical syndromes in AQP4-IgG-seropositive patients and, in fact, their presence strengthens confidence about a NMOSD diagnosis.^58,e97,e98 The underlying cause of CNS symptoms and signs is more likely to be co-associated NMOSD than a direct complication (e.g., associated vasculitis) of SLE or SS. In a patient suspected of having NMOSD, the presence of clinical myasthenia gravis or detectable serum acetylcholine receptor antibodies is considered supportive of NMOSD diagnosis.^e99–e101

Pathology.

Pathologic findings in biopsy or autopsy tissue obtained from patients with AQP4-IgG-seropositive NMOSD demonstrate loss of AQP4 immunoreactivity and evidence of perivascular complement activation in actively demyelinating lesions.^e20,e21 Additionally, findings supportive of astrocytopathy such as truncated astrocyte processes or cell loss may be detected by immunostaining for glial fibrillary acidic protein.^59,e20 These findings in active lesions distinguish AQP4-IgG-positive NMOSD from MS; data from seronegative individuals are not yet available.⁵⁹ Necrosis and lesion infiltration with neutrophils and eosinophils are supportive characteristics,^e102 but may not be present. Spinal cord lesions may differ from supraspinal lesions in this respect.^e103 Serum AQP4-IgG testing usually obviates the need for biopsy in severe myelitis and leukoencephalopathy syndromes.^{e19,e104,e105} Aquaporin-4 immunostaining is not a routine procedure in CNS biopsy processing and is available only in select centers. The Panel does not recommend CNS biopsy but recognizes that in atypical cases, expert pathologic review of biopsy tissue of brain or spinal cord might help establish NMOSD and exclude competing diagnoses.⁶⁰

Opticospinal MS.

The term opticospinal MS was introduced in Japan to refer to a pattern of MS that is relatively more common in Asian countries.^11,e106 It was defined by recurrent optic neuritis and myelitis attacks with no brain involvement, although some investigators allow occurrence of certain brainstem syndromes.^e107 The description of opticospinal MS represented an important milestone because it recognized that there was a relapsing illness distinct from conventional MS that selectively targeted the optic nerve and spinal cord at a time when relapses were deemed to be incompatible with a diagnosis of NMO in both Western countries and Japan.

Contemporary NMO criteria that include MRI criteria for length of spinal cord lesions and detection of AQP4-IgG are more specific than historical opticospinal MS criteria. Clinicians and investigators in many Asian countries now use the term NMO rather than opticospinal MS in research and practice. When similarly defined in Asia, NMO clinical syndromes and brain MRI lesions are both analogous to those encountered in Western countries.^{22,23,33,39,e108} The Panel concluded that opticospinal MS is a historically important but now superseded term, especially in light of the need to distinguish NMO from MS to guide treatment decisions.^{15,–,18,e24–e26}