MRI of the relevant domain should be performed to confirm whether clinical symptoms represent an attack of NMOSD: No

There is little doubt about the practical utility of magnetic resonance imaging (MRI) in diagnosing clinical attacks of central nervous system (CNS) neuroinflammatory events. Anatomical confirmation of inflammation at the sites suspected by the history and physical examination findings helps assure the clinician that their diagnostic impression is accurate. However, the notion that MRI is necessary to confirm a diagnosis of clinical attacks is over-stated. There are many examples in which an MRI scan is not necessary to confirm attacks in neuromyelitis optica spectrum disorder (NMOSD). For example, a patient who is already diagnosed with NMOSD and has no prior history of optic neuritis (ON), with a documented normal baseline vision examination, and who present with painful unilateral vision loss and a new afferent pupillary defect can be accurately diagnosed with ON without an MRI of the orbits. In this example, the history and physical examination provide sufficient confidence in diagnosis that treatment can be initiated without imaging. There are similar examples for cases of myelitis. MRI was not required for attack diagnosis in several recent clinical trials in NMOSD that led to regulatory approval of eculizumab and satralizumab. In fact, MRI outcomes were not assessed in these studies at all. ¹ The determination of clinical attacks made based on the treating physician’s acumen and was subsequently centrally adjudicated. Despite the absence of radiographic confirmation of neuroanatomical inflammation at the site corresponding to suspected clinical involvement, regulators were sufficiently convinced of differences in therapeutic efficacy between the treatment arms to approve these drugs. The clinical trial that led to inebilizumab’s approval defined 18 criteria for clinical attacks. Of the 11 ON criteria, only three required MRI confirmation, and of the four myelitis criteria, only two required MRI confirmation. ² MRI was used to confirm relatively mild clinical attacks or attacks with self-reported symptoms (e.g. hiccupping, nausea, and vomiting). In this trial, the concordance between clinical attacks and central adjudication was high demonstrating the robustness of the attack criteria, 72% of which of which did not require MRI. ³ Even the International Panel Diagnostic Criteria for NMOSD does not require MRI confirmation when patients have had a clinical attack consistent with NMOSD and test seropositive for anti-AQP4 antibodies. ⁴ As with determination of clinical attacks, MRI can be very useful in NMOSD when there is diagnostic uncertainty, for example, when patients are AQP4 seronegative.

Although MRI can visualize CNS inflammation, MRI also has important practical limitations. MRI is not always available in a timely manner. Even at prominent academic medical centers in the United States, MRI may not be available in a timely manner and often requiring days if not weeks to schedule as an outpatient. Given the often aggressive nature of NMOSD attacks, initiating timely anti-inflammatory treatment for acute attacks with high dose corticosteroids and or plasmapheresis is critical: delays in treatment are associated with worse outcomes. ⁵ Furthermore, MRI may not be even be available at all in some community settings or even in some countries. ⁶

In addition to these practical considerations, MRI also has some technical limitations. Although the concordance of new MRI lesions with clinical attacks is high, MRI is not 100% sensitive. MRI may sometimes be interpreted as normal or unchanged in some attacks when clinical worsening is apparent. ⁷ Recently, a post hoc evaluation of attack in the N-MOmentum study found that four adjudicated attacks had no MRI lesion correlates. ⁸ There is also a risk of false-negative MRI, if the scan is performed too early or without the appropriated sequences, especially for orbital MRI.

Conversely, an abnormal MRI does not necessary mean that an attack is occurring. In AQP4-IgG NMOSD, it can be difficult to interpret gadolinium-enhanced lesions, because chronic breakdown of the blood–brain barrier can occur due to aquaporin-4 water channel dysfunction, especially after ON. ⁹ It can also be difficult to distinguish between symptomatic and non-specific lesions on brain MRI in NMOSD patients with encephalic/brain symptoms. Finally, when comparing spinal cord MRI to assess for new lesions, it is important to be cautious when different sequences are used (i.e. short tau inversion recovery—STIR vs T2-weighted). All of these limitations highlight the need for international guidelines on the timing and sequences of MRI scans in NMOSD.

The wisdom that conventional MRI is the best marker for assessing NMOSD activity is probably outdated. Accumulating data suggest that fluid biomarkers, and more precisely serum glial fibrillary acidic protein (sGFAP), are more sensitive for detecting disease activity. ⁸ The N-MOmentum study reported two cases of patients with clinical symptoms, concomitant increase in sGFAP levels, but without MRI changes. This is logical since this biomarker directly reflects astrocyte injury. ¹⁰ In addition, there are clinical symptoms outside the classical core clinical presentation of NMOSD, including pain, fatigue, depression, cognitive impairment that are thought to be associated with a more global and smoldering process, that is not yet captured by conventional MRI.

Therefore, requiring MRI confirmation of clinical attacks may not always be feasible. It is essential that neurologists be able to diagnose suspected NMOSD attacks, primarily using clinical judgment, in established NMOSD patients. In the setting of diagnostic uncertainty, and when available, MRI can aid in both NMOSD diagnosis and attack determination.