Should Spinal MRI Be Routinely Performed in Patients With Clinically Isolated Optic Neuritis?

Abstract

Clinically isolated optic neuritis (ON) can be associated with a variety of autoimmune disorders. In addition to multiple sclerosis (MS), it is associated with antibodies to aquaporin-4 and myelin oligodendrocyte glycoprotein (MOG). Spinal MRI may help increase diagnostic yield for these disorders, but at the cost of time, expense, and exposure to contrast. Experts in ON and demyelinating disease debate the utility of routine spinal imaging in patients with isolated ON. Pro: Ethan Meltzer, MD, Elliot M. Frohman, MD, PhD, FAAN, FANA, Teresa C. Frohman, MPAS, PA-C, FANA The question posed to us to debate is the utility of spinal imaging in patients who present with ON. However, this question is inextricably linked to our understanding of what ON is, and what diagnostic tests (if any) are needed when making the diagnosis. The past decade has seen not only advancements in how we recognize and treat ON, but also new advances have demonstrated that not all ON is created equally. Unfortunately, there is no widely accepted definition of ON. However, for the purposes of our argument, we will define it in clinical terms. We choose to use the clinical definition, because that is what is relevant to the patient and the clinician. In the acute presentation, the clinician does not have the luxury of knowing what the course of disease will be and is faced with a diagnostic conundrum. We define ON as inflammation of the optic nerve causing subacute vision loss over hours to days, and is generally associated with pain during eye movements (present in 91% of cases according to the ONTT) (1). This can be further differentiated into typical ON, related to MS, and atypical ON, related to other immunologic diseases—including neuromyelitis optic spectrum disorder (NMOSD) and ON associated with antibodies to MOG. There are few rules that govern how a clinician distinguishes typical from atypical ON. Therein lies one of the core problems that clinicians face when they evaluate a patient with ON. Is this an isolated event or is this a harbinger of MS? What evidence do we have for management of isolated ON? What are the risks of treatment delay? Could this be a manifestation of NMOSD, which requires early aggressive treatment not otherwise routinely done in “run-of-the-mill” ON? The clinician is now charged with not only recognizing ON, but putting it into the broader context of neuroimmunologic disease. These core problems relate to what we must do as clinicians when a patient presents with a first episode of ON. How deep we search for evidence of inflammation within the central nervous system (CNS), even in otherwise asymptomatic individuals? This is a societal question, one with diagnostic, prognostic, pharmacologic, cultural, economic, and, even, medicolegal confounding factors. We will attempt to break down each of these facets in an argument in favor of spinal cord imaging as a necessary part of the workup for patients who present with ON. Optic neuritis as initial manifestation of multiple sclerosis The incidence of MS varies among ethnicities. The worldwide range is approximately 1–6 persons per 100,000 per year (2–6). Large population studies have shown that ON is the presenting symptom of MS in 25% of cases (5). However, in patients who present with ON, the risk of conversion to clinically definite MS varied by ethnicity because the rate of MS is not identical in all ethnic groups. For example, in Japan, a country with overall lower rates of ON compared with other populations, the rate of MRI findings suggestive of MS in patients who present with ON is only 14%, compared with lesions seen on MRI in >50% of patients from many Western European countries (7). To help the clinician in making a diagnosis of MS, the 2010 McDonald criteria incorporated MRI findings into establishing dissemination in time (DIT) and space, which was further revised in 2017 (8,9). The Magnetic Resonance Imaging in MS (MAGNIMS) study group published new consensus guidelines on the use of MRI to establish the diagnosis of MS, which were considered in the most recently revised McDonald criteria. They recommend spinal MRI in patients with clinically isolated syndrome (CIS) who do not fulfill brain MRI criteria for MS to identify patients at risk of developing clinically definite multiple sclerosis (10,11). One reason that spinal MRI might be helpful is that even normal brain MRIs may have nonspecific T2 hyperintense white matter lesions, which can be difficult to distinguish between the typical MS periventricular plaques. The clinician can be left in a difficult position as to counseling a patient on overall risk of conversion to MS if they are unsure if the nonspecific white matter changes they are seeing are just normal changes or if they represent MS plaques. This is in contrast to spinal cord lesions, which are very specific for demyelinating disease (12). If spinal cord lesions are present, the clinician is able to offer a more definitive prediction of risk of progression to MS. Although an MRI is not needed to make a diagnosis of ON, MRI of the orbits with contrast often is obtained along with MRI of the brain. In patients with CIS, 50%–70% of patients will have an abnormal brain MRI. Abnormalities on MRI convey an elevated long-term risk of developing MS: 60%–82% risk compared with 8%–25% risk in patients without lesions on MRI (13–16). The population that we are then looking at is theoretically a group of “lower-risk” patients. However, it is our job to ensure that they are in fact lower-risk patients. Importantly, up to a third of patients with CIS may have asymptomatic spinal cord lesions. There are no large studies that look at the risk of development of MS based on spinal cord lesions alone, but we could reasonably infer that these patients are at similarly elevated risk to those patients with typical MS lesions on brain MRI (11–13). One could argue that the likelihood of spinal cord lesions to be asymptomatic is much lower than supratentorial lesions. However, a study of MS patients found that 83% of patients with recent diagnosis of MS had an asymptomatic spinal cord lesion present on MRI (17). We do not have that same information for patients with ON as the first episode of MS, but, there is no indication that the previous population were enriched compared with the group of CIS patients. Thus, the absence of symptoms does not fundamentally lessen the ability for these lesions to be present. Clinically isolated syndrome risk stratification and varied treatment At this point, we must do some extrapolation from the data we have regarding risk of conversion to MS from CIS. We do have 6 randomized controlled trials to address the effect of conventional MS medications (interferons, glatiramer acetate, and teriflunomide) on preventing conversion of high risk-CIS patients, those with various degrees of typical MS lesions on brain MRI, to clinically definite multiple sclerosis (18–23). These studies showed a range of reduction in developing MS of 28%–45% over the 2–3-year follow-up period using MS medications. These studies did not use presence of spinal cord lesions as part of their criteria. However, it is the opinion of the authors that a typical MS plaque seen on spinal cord MRI should be treated as equivalent to typical brain parenchymal lesions. We would expect to see similar reduction in the rates of developing clinically definite MS. Finding these asymptomatic lesions is thus not just helpful in determining prognosis, but can be an important finding that changes treatment. There are some studies with indirect evidence that the natural history of MS is less severe when the number of early relapses are decreased with disease-modifying therapy (24,25). Population studies show prolongation of disability-free survival with increased time between initial relapses (24). Longitudinal cohort studies predict that approximately 75% of patients with high lesion volume on MRI will advance to an Expanded Disability Status Scale (EDSS) >6 within 5 years (14–26). Conversely, some longitudinal data sets have shown that decreasing the burden of accumulated T2 lesions on MRI in the first 5 years of MS does lead to a delay or prevention of secondary progressive MS (14). Given this information, we postulate that it is the task of the clinician at first symptom to look for any evidence, be it CSF markers such as oligoclonal bands or MRI findings, that might trigger initiation of treatment for MS. Unfortunately, these studies are plagued by lack of data for long-term decrease in overall disability, and it is not yet proven that the overall course of disease for an individual patient is altered by immunotherapy. However, we postulate that lack of evidence of efficacy is not a reason to withhold a potential treatment with few adverse effects. There is discordance in the available literature regarding asymptomatic spinal cord lesions taken together with MRI lesions. One study looked at individuals with asymptomatic spinal cord lesions and found that only 12% had no coexistent asymptomatic supratentorial lesions on MRI (27,28). It was noted that spinal cord MRI only led to 3 additional diagnoses of MS using McDonald criteria out of a population of 113 patients initially presenting with isolated ON (27). This would suggest that there is no role of additional spinal cord imaging in this population because the authors found no independent increased risk of developing MS with spinal cord lesions. However, a separate study looked at expanded EDSS for patients 5 years after an initial CIS even of ON. This study found that asymptomatic spinal cord lesions, seen in 26% of their patients, led to worse disability at 5 years, and that no other asymptomatic lesions within the CNS had that same association (28,29). We would argue that given the controversy regarding the role of spinal cord lesions in assessing of risk of developing MS, that taken together, that is still proof of utility of spinal cord imaging. Learning from radiologically isolated syndrome There also is controversy regarding the prognostic value of asymptomatic spinal cord lesions in radiologically isolated syndrome (RIS) vs those seen in CIS (30). One retrospective review of RIS found that 21 of 25 patients with asymptomatic spinal cord lesions went on to develop CIS or primary progressive MS within 1–6 years (31). Although the population this study looked at is different than the question posed to us, the presence of asymptomatic spinal cord lesions had a significantly higher risk of progression to MS with an odds ratio of 75 (32). This argues for the high predictive value of spinal cord lesions and the importance of their presence in treating these patients. The population of patients with ON and normal brain MRI has to be one in which we consider non-MS entities such as NMO and ON associated with antibodies directed against MOG. Although it would be unlikely for a patient with NMOSD to have an asymptomatic spinal cord lesion, the presence of typical, small, ovoid, eccentrically placed spinal lesions would make MS the more likely diagnosis (6,33). Disease associated with anti-MOG antibodies, commonly presents with ON, and can have a normal brain MRI. At least in the adult population, these patients rarely have asymptomatic spinal cord lesions (34). ON in the pediatric population can be due to a heterogeneous group of disorders, including ADEM, NMOSD, anti-MOG, as well as other systemic inflammatory processes. Obtaining a spinal cord MRI can be helpful in distinguishing among these different disorders, which is especially important in children as they are much less likely (as few as ∼2% over the next 3 years) to eventually be diagnosed with MS if the brain MRI is normal (35). Ultimately, the reason it is important to distinguish between these diseases quickly is that initial treatment differs dramatically. If a patient presents with severe ON, the clinician and patient do not have the luxury of waiting for antibody testing to come back. For example, any ancillary data that help to risk stratify the patient between ON secondary to NMOSD rather than MS can make a large impact on long-term disability for the patient (36,37). Clinical trial design Besides the above clinical considerations, an additional benefit to obtaining spinal cord MRI is enrollment in clinical trials. Although not applicable to all practice environments, many patients with ON may receive care either in or are affiliated with MS centers. Given the number of ongoing clinical trials, having a full set of initial data can help facilitate potential clinical trial enrollment (38). Economic, social, and cultural influences These authors cede that the preceding arguments are not made in a cultural vacuum, but instead exist within our particular health care system. We are members of a health care system that places great importance as well as reliance on technology within all aspects of medical care. Our utilization of advanced technology, such as MRI, surpasses by a factor of several fold other similar Western countries (39). However, there are also cultural influences that we operate under that incentivize using any available resource for patient care. We acknowledge that this practice of medicine is not applicable to all health care systems, cultures, or geographical locations. Con: Fiona E. Costello, MD, FRCP and Jodie M. Burton, MD, FRCP, MSc ON is a common CIS heralding the diagnosis in over 20% of MS patients (40). In the era of the Optic Neuritis Treatment Trial, ON patients needed to have a minimum of 2 discrete clinical events over time to satisfy Poser criteria for the diagnosis of MS (41). Over the past 2 decades, however, a constellation of clinical, imaging, and laboratory parameters have been used to establish diagnosis, with the goal of facilitating earlier treatment options for MS patients. In 2017, the most recent iteration of the McDonald criteria was unveiled, with specific emphasis on the requirement that no better explanation than MS be overlooked to account for the clinical presentation among patients with CIS (9). In this update from the previous 2010 criteria, several changes were implemented to help define requirements for dissemination of CNS lesions in space and time, which are germane to the diagnosis of MS (Table 1). For example, in the case of a patient with a typical CIS, such as ON, and fulfillment of clinical or MRI criteria for the dissemination in space, the finding of CSF-specific oligoclonal bands may be used to render the diagnosis of MS. In addition, cortical lesions captured by MRI can be used to establish dissemination in space. Finally, per the revised 2017 criteria, symptomatic and asymptomatic MRI lesions (excluding the optic nerve location in the setting of ON) can be used to demonstrate dissemination in space or time. The reason for this is that ON in the setting of a normal MRI implicates a broader diagnostic differential than would be generated in the presence of typical CNS demyelinating lesions.TABLE 1.: The 2017 McDonald criteria for the diagnosis of MSThe specificity of the diagnostic evaluation is particularly relevant in an era when MS treatment is becoming increasingly more aggressive. Consequently, although there may be drawbacks to delayed MS diagnosis, there may be equal if not greater perils for misdiagnosed patients who are wrongly initiated on therapies with deleterious side effects. It is within this evolving framework that there is debate regarding the “value-added” of routine spinal imaging for CIS patients presenting with ON, who have normal cranial MRI findings. To better understand the role of routine spine imaging in this clinical setting, several cogent questions need to be addressed: How will the spinal MRI scan increase the diagnostic certainty for MS? How will the spinal MRI scan help predict MS-related disability? How will the spinal MRI scan help exclude a diagnostic mimic for MS, such as neuromyelitis optica spectrum disorder (NMOSD)? Will routine spinal MRI improve the diagnostic yield for definite multiple sclerosis in a clinically isolated syndrome presentation (including optic neuritis)? Asymptomatic spinal cord lesions are found in 30%–40% of patients with CIS (42,43). Therefore, in the case of patients with ON who lack evidence, historically or by examination, of disseminated CNS inflammation, spinal cord imaging may offer diagnostic value. Short segmented spinal lesions are a relatively specific finding for MS, (42) and may show proof of dissemination in space (DIS); yet, the diagnosis of MS would currently still require one or more T2-hyperintense characteristic lesions in 2 or more of 4 areas of the CNS, including: (1) periventricular, (2) cortical or juxtacortical, and (3) infratentorial brain regions, and the (4) spinal cord (9). Thus, spinal cord imaging alone would not change the diagnostic yield in a monosymptomatic ON patient with normal brain MRI and a normal neurological examination. To fulfill criteria for DIT, the simultaneous presence of contrast-enhancing and nonenhancing lesions at any time, or a new T2-hyperintense or enhancing lesion on follow-up MRI would be required, compared with a baseline scan. A patient who has brain MRI findings already satisfying DIS will not gain any diagnostic yield by the detection of an asymptomatic spinal cord lesion. Furthermore, the search for asymptomatic spinal cord lesions to satisfy DIT is of low yield and is not recommended by the Magnetic Resonance Imaging in MS (MAGNIMS) research group (10). Finally, the additional gadolinium dye administration may be hard to justify for routine spinal imaging, given the uncertain effects of gadolinium retention that have been reported (44). Implicit to this discussion is that appropriate spinal cord sequences are obtained in the evaluation process. To satisfy the tenets of the 2017 McDonald criteria, a relevant spinal cord finding refers to a hyperintense lesion in the cervical, thoracic, or lumbar spinal cord seen on T2 plus short tau inversion recovery, proton-density images, or other appropriate sequences, or in 2 planes on T2 MRI (9). To obtain accurate imaging results, Rovira and Tintore (43) pointed out that spinal cord MRI studies ideally should be obtained using high-resolution sagittal T2-weighted sequences using at least 2 different algorithms (double-echo T2, or T2 and short-tau inversion recovery, in combination with transverse T2-weighted sequences). These protocol requirements substantially increase the time required within the scanner and medical costs. Arguably, the time and costs issues could be ameliorated by reducing scan times (43), but this may lead to the acquisition of suboptimal studies, adding more confusion than clarity to the diagnostic process. In a previously published debate, Barkhof (45) suggested that there are “hard” indications for spinal cord imaging in patients with CIS suspected of having MS, which include (1) any patient with CIS—including ON—not fulfilling McDonald. Instead, we uphold the recommendations of The International Panel on Diagnosis of Multiple Sclerosis (“The Panel”), which state that cranial MRI be obtained in all patients being considered for a diagnosis of MS (9). According to “The Panel,” spinal cord imaging is advisable, when the clinical presentation suggests spinal cord involvement, when there is evidence of a primary progressive MS course, when considering MS in a patient population in which the disease is relatively uncommon (e.g., older individuals or non-white populations), or when additional data are needed to increase diagnostic yield (9). Will routine spinal imaging improve the ability to predict multiple sclerosis–related disability? The role of spinal cord lesions for predicting long-term disability is a matter of debate. Previous studies have suggested that asymptomatic spinal cord lesions detected with MRI in patients with CIS predict future neurological disability (43,46,47). Arguably, early knowledge of such predictors could impact treatment decisions for patients with CIS flagged as being at increased risk. Yet, Dekker et al (48) performed a recent longitudinal cohort study on CIS and early relapsing–remitting patients with MS to investigate the influence of asymptomatic spinal cord lesions on the time to MS-related disability. Their study included 178 patients, among whom 42 patients (24%) had asymptomatic spinal cord lesions. These investigators observed no significant differences in the time to disability development, or the time to a second event with detection of asymptomatic spine lesions. In related work, Tummala et al (49) studied 62 CIS (n = 3), RRMS (n = 56), and SPMS secondary progressive MS (n = 2) patients to determine the utility of brain and spinal cord 3T MRI in the 1-year evaluation of NEDA (“no evidence of disease activity” as a therapeutic goal in MS). In this study, achieving NEDA by spinal or brain MRI was defined as no new or enlarging T2 hyperintense lesions at 1 year. No evidence of disease activity was detected in 48% of patients at 1 year, and no differences were observed with the addition of spinal cord MRI (49). These investigators concluded that spinal cord MRI demonstrated low diagnostic yield as an adjunct to brain MRI in monitoring patients with MS for NEDA over 1 year. Future research with longer follow-up intervals and larger patient cohorts will be needed to reconcile the role of spinal cord MRI in predicting future disability for CIS (and RRMS) patients. Based on the available evidence, gauging future disability cannot be used as a rationale to justify routine spinal imaging to detect asymptomatic lesions, for the patient with ON in the proposed case scenario, or patients with CIS in general. Will spinal MRI increase the diagnostic yield for multiple sclerosis mimics, such as neuromyelitis optica spectrum disorder? Suspicion regarding the diagnosis of NMOSD should definitely prompt additional investigations, including spinal MRI in a CIS patient presenting with ON. According to Wingerchuk et al (50), specific clinical features that are suggestive of NMOSD in this context include ON that is simultaneously bilateral, involves the optic chiasm, causes altitudinal visual field defects, or leads to severe vision loss (acuity worse than 20/200). MRI evidence of bilateral optic nerve involvement, posterior nerve predominance (with extension into the optic chiasm), or extensive lesions of the optic nerve (more than half of its length) also are suggestive of NMOSD (50,51). Because diagnostic requirements are more stringent for patients in whom aquaporin-4 IgG antibodies are not detected, an argument could be made for additional spinal MRI testing for the ON patient in question, if NMOSD is a diagnostic concern. That being said, a patient with “typical” ON who does not demonstrate any clinical, demographic, or laboratory features of NMOSD is not someone in whom NMOSD is high in the differential diagnostic possibilities. Detecting a short-segment spinal cord lesion in a patient who meets no other criteria of NMOSD is still not likely to increase suspicion for this diagnosis. As for anti-MOG demyelinating disease, a relatively recent diagnostic entity characterized by patients who have serum anti-MOG IgG antibodies in the absence of aquaporin-4 antibodies, there is no consensus as of yet on the yield of spinal MRI in asymptomatic patients. In fact, in one large case series from the Neuromyelitis Optica Study Group, only 2 of 29 patients (with available spinal MRI data) had asymptomatic spinal cord lesions (52). Summary The diagnostic approach to MS continues to evolve, with increasing reliance on paraclinical measures such as MRI, and the results of CSF analysis. Although there are definitely circumstances in which spinal cord imaging may be indicated to consolidate the diagnosis of MS or exclude a mimic such as NMOSD, we do not recommend routine spinal MRI imaging for patients with CIS in general, or for the ON patient in question. Instead, we would reiterate the 2017 McDonald criteria recommendations, by emphasizing that additional testing, either with spinal MRI or CSF analysis, should be considered for certain patients with CIS. This is particularly the case when MS disease-modifying therapies are being considered, chiefly to aid in meeting the DIS component of the diagnostic criteria. The decision to initiate disease-modifying therapies would imply ample clinical suspicion for the diagnosis of MS, which is lacking for the ON patient in question. It would also be reasonable to consider spinal cord imaging if there is clinical suspicion for NMOSD as a diagnostic mimic, because interferon-b, natalizumab, and fingolimod may have detrimental effects for patients with NMOSD misdiagnosed as having MS. Thus, although we do not advocate routine for spinal imaging in patients with CIS, we do recommend that spinal MRI be used in patients with atypical CIS presentations, progressive course at onset, atypical imaging or clinical features for MS, and in populations in which MS is less common. Rebuttal: Drs. Frohman and Meltzer We appreciate the opinions of Drs. Costello and Burton in ON. We will briefly discuss a few points they raise in their discussion. We agree that the detection of an asymptomatic spinal cord lesion with a normal brain MRI does not make a diagnosis of MS by using the recently updated 2017 McDonald criteria. However, its presence, if found, would fulfill the updated McDonald criteria for MS if a new typical MS lesion (even nonenhancing and asymptomatic, in the presence of oligoclonal bands) were found on follow-up brain MRI, which is commonly done in these patients (9). Although the MAGNIMS consensus guidelines do not recommend routine spinal MRI for the purposes of fulfilling diagnostic criteria for DIT (i.e., surveillance spine MRI in patients without a new spinal cord clinical syndrome), the guidelines include a recommendation that spinal MRI be obtained at initial presentation for a patient with a nonspinal CIS to help define dissemination in space and time when brain MRI does not fulfill MRI criteria for a diagnosis of MS (10). We also reiterate the higher risk of progression to clinically definite MS in patients with asymptomatic spinal cord lesions in the RIS population (OR 75) (32). What this tells us is that spinal cord lesions are not equal to brain lesions, and although they are less common and not given different weight than brain lesions in the McDonald criteria, they should perhaps be weighed differently in the clinician's mind. Regarding the safety of gadolinium and potential adverse effects of gadolinium retention as identified by high signal in the dentate nucleus and globus pallidus, the clinical significance remains unknown (53). In addition, this was only seen with linear agents as opposed to macrocyclic agents, which are now used in most institutions (54,55). As of now, we have not changed our recommendations on surveillance imaging in our patients with MS, but we agree that it is a concerning finding that warrants discussion with our patients. Both of our discussions attempted to tackle the controversy regarding the utility of spinal MRI as a prognostic factor for disability in MS. We appreciate the points raised by Drs. Costello and Burton acknowledging the limitations of our current understanding and conflicting data in the literature. We would argue that although the majority of patients who present with ON will have a “typical” presentation unlikely to be confused with alternative diagnoses, a minority will have an “atypical” feature that raises the concern for an alternative diagnosis. From the ONTT data, 36% of patients presented with vision loss worse than 20/200, 15% were of non-white race, and even after excluding patients older than 46 years, the average age on presentation was 32 years (56). This highly selected population in the United States still had a significant percentage of patients who would have clinical features (such as race or severity of symptoms) that might make an alternative diagnosis more likely or possible confounding brain MRI features that might cause diagnostic uncertainty (older patients with nonspecific white matter changes). Although most of these patients will not have NMOSD, given the different treatments and clinical outcomes of NMOSD vs MS, we would recommend spinal imaging. Finally, neurologists and ophthalmologists in the United States are more likely than their counterparts in Canada to order a brain MRI for patients who present with ON (87.4% and 76.5% to 72.1% and 53.9%, respectively) (57). In addition, practitioners in the United States are more likely to order subsequent yearly follow-up imaging on patients. We accept that our cultural opinions regarding the utility of neuroimaging will differ based on our practice environments. Rebuttal: Drs. Costello and Burton We respectfully thank our colleagues for their thoughtful and patient-centric argument in favor of obtaining spinal cord MRI studies in patients with ON. In many ways, we do not disagree with this assertion, with some caveats in place. In those patients with ON who do not demonstrate classic MS-type brain lesions on MRI, spinal cord imaging may very well serve several important purposes. The differential of ON is much more than simply a clinically isolated presentation of what may ultimately become MS, (40) and if the cranial MRI scan is not consistent with MS, then spinal imaging may provide clues about alternative diagnoses and/or confirm MS-like lesions (albeit only in the cord) that could expedite access to MS disease-modifying medications (41). However, although high-risk CIS patients may have silent cord lesions, they do not have “only cord lesions” and such patients would already have brain lesions to suggest MS using the current diagnostic criteria. The addition of cord lesions to MS-like brain lesions is unlikely to have a meaningful impact on the ultimate risk of MS (41–43). Given the current diagnostic criteria, there is limited utility of silent cord lesion because a silent lesion would rarely be gadolinium-positive (and hence not contribute to DIT) and would still require the presence of at least one MS-like brain lesion to achieve DIS. (41,42) Hence, if there are sufficient brain lesions to meet DIS with ON, there is no clear role for imaging the cord without clinical suspicion. Furthermore, the addition of a gadolinium-enhanced MRI spine for a potential silent lesion raises the current concern about the potential for gadolinium retention, for which most practitioners have elected to avoid unnecessary administration of gadolinium (44). With regards to the prognostic value of spinal cord lesions in patients with MS, several studies have shown that such lesions are common and do not portend a worse long-term prognosis in patients with CIS/MS (48,49). The putative value of spinal imaging in RIS patients notwithstanding, RIS patients are not interchangeable with ON patients; so, the utility of spinal cord imaging in an RIS cohort may not be applicable to the question at hand. Our colleagues also suggest that spinal cord imaging could assist with the diagnosis and management of NMOSD. Although small silent cord lesions are possible in NMOSD, the classic presentation is longitudinally extensive (greater than 3 vertebral bodies in length, often gadolinium-positive) and almost always symptomatic (58). Most practitioners would not use silent cord lesions to diagnose NMOSD without other “red flags” as outlined above (37). The newly recognized entity of ON associated with MOG is still being studied and, at present, we do not know about the prognostic value of spinal cord lesions in this cohort (52). In summary, we agree that spinal cord imaging has a role in certain patients with ON, specifically those who have no other features to suggest that this is a first demyelinating event of what may become MS, or patients in whom such lesions could allow for easier access to much needed therapy. But, in those patients who have “typical” ON with a unilateral presentation and good recovery, coupled with cranial MRI features consistent with MS-like lesions, the added value of spinal cord imaging in those without clinical spinal cord findings is of low yield and, we believe, not warranted. Conclusion The authors have done a superb job of summarizing the current data regarding the role of spinal imaging in clinically isolated ON. This is a continually evolving field and, over the next decade, we may have a better understanding of the role both AQP4 and MOG antibodies play in the pathogenesis of ON. The potential benefits of spinal imaging (increased diagnostic yield and better prognostic information) must be balanced against potential risks (time and expense, exposure to gadolinium, and potential overdiagnosis due to incorrect interpretation). Spinal imaging is warranted in all patients with “atypical” ON (with features noted by the authors), and should be at least considered in certain patients with more typical features.