Investigation of health care use and a possible prodrome before the first attack in NMOSD and MOGAD

Abstract

Background:

Prodromal phases are well recognized in many inflammatory and neurodegenerative diseases, including multiple sclerosis. We evaluated the possibility of a prodrome in aquaporin-4 antibody positive (AQP4+) neuromyelitis optica spectrum disorder (NMOSD) and myelin oligodendrocyte glycoprotein antibody disease (MOGAD) using health administrative data.

Methods:

We investigated individuals with AQP4 + NMOSD and MOGAD, confirmed by medical chart review, in Ontario, Canada. Each NMOSD and MOGAD participant was matched 1:5 to general population controls by sex, birth year, immigrant status, and region. Total outpatient visits and hospitalizations were compared in the 5 years preceding the incident attack in multivariable negative binomial models.

Results:

We identified 96 people with AQP4 + NMOSD, matched to 479 controls, and 61 people with MOGAD, matched to 303 controls. In the 5 years preceding the incident attack, health care use was elevated for outpatient visits and hospitalizations for the NMOSD cohort (adjusted rate ratio (aRR): 1.47; 95% confidence interval (CI): 1.25–1.73; aRR: 1.67; 95% CI: 1.19–2.36, respectively) but not for MOGAD. Rate ratios steadily increased in NMOSD for outpatient visits in the 2 years preceding the incident attack.

Conclusion:

Our findings support a prodromal phase preceding clinical onset of AQP4 + NMOSD. Earlier recognition and management of NMOSD patients may be possible.

Background

A prodrome refers to an early period of symptoms, often non-specific, that precedes the typical clinical presentation of a disease and indicates underlying disease onset. ¹ Prodromal phases are now well recognized in multiple inflammatory and neurodegenerative diseases, including multiple sclerosis (MS). ² Understanding of the prodrome can provide critical insights into disease pathogenesis and approaches to disease prevention. In MS, recognition of a prodrome emerged from studies of health administrative data which demonstrated increased health service use in the 5–10 years predating MS diagnosis or the first attack.^3,4

To date, the possibility of a prodrome has not been systematically investigated in two other inflammatory, demyelinating disorders of the central nervous system (CNS): aquaporin-4 antibody positive (AQP4+) neuromyelitis optica spectrum disorder (NMOSD) and myelin oligodendrocyte glycoprotein antibody disease (MOGAD). These conditions are both rare, each with an estimated global prevalence of 1 per 100,000.^5,6 Interestingly, in both diseases, pathobiology has been inferred to commence immediately before the first attack.^7,8 However, other findings have suggested a possible prodromal phase. In a study from the US Department of Defense Repository, more than 60% (20/32) of individuals with NMOSD had blood samples from before clinical onset that tested positive for the AQP4 antibody. ⁹ Other instances of AQP4 antibody detected in serum samples banked from individuals’ years before the first NMOSD attack have been reported.^10,11 In addition, there have been case reports of AQP4 + NMOSD patients with symptoms or signs preceding the first attack, including neuropathic pain, ¹² myalgias and myositis, ¹³ and asymptomatic optic disk edema. ¹⁴ There is less evidence to suggest earlier disease onset in MOGAD, but there have been rare reports of cranial and peripheral neuropathies preceding the first attack. ¹⁵

We aimed to evaluate the possibility of a prodrome in AQP4 + NMOSD and MOGAD using prospectively collected health administrative data. We separately assessed health care use in the NMOSD and MOGAD cohorts, each matched and compared to its own set of general population controls.

Methods

This was a retrospective matched cohort study in Canada’s largest province, Ontario (population of 15.5 million in 2023). Ontario has a publicly funded universal health insurance system that covers > 98% of the population without user fees. Information regarding health system use is captured securely in coded datasets held by ICES (formerly known as Institute for Clinical Evaluative Sciences). The use of ICES data is authorized under section 45 of the Ontario Personal Health Information Protection Act. This study was approved by the research ethics board at each participating institution.

Clinical data: cases

Cases were identified from participants in the Canadian Neuromyelitis Optica Spectrum Disorder and other atypical demyelinating diseases Cohort Study (CANOPTICS) in Ontario who had consented to linkage of their medical records with health administrative data (Supplement). Participants included in the NMOSD cohort had a positive serum test for the AQP4 antibody on the cell-based assay (CBA) or ELISA and met 2015 international consensus criteria for the diagnosis of NMOSD. ⁵ Participants in the MOGAD cohort had a positive serum MOG antibody test on the fixed cell-based assay and a clinical syndrome consistent with MOGAD as determined by the treating neurologist. Cases with an incident demyelinating health administrative code between 1 January 1997 or after 31 December 2022 were included. See Supplement for power calculation.

Administrative data and index date

NMOSD and MOGAD cases were linked deterministically with administrative data at ICES using their unique, lifelong Ontario Health Insurance Program (OHIP) numbers. See Supplement for detailed description of databases used. Briefly, these data comprised outpatient physician visits and hospitalizations including date of encounter and the diagnosis using International Classification of Diseases (ICD)-9/10-Canadian version (CA) codes.

The date of the incident demyelinating attack (index date) was identified using the first inpatient or outpatient record for the following ICD 9/10 codes: optic neuritis (377/H46), transverse myelitis or encephalomyelitis (323/G36 or G37), or multiple sclerosis (340/G35), as some individuals with NMOSD and MOGAD are initially misdiagnosed as MS. No specific ICD code for NMOSD or MOGAD was available. The first health administrative demyelinating code was used as the index date to avoid recall bias and allow for complete ascertainment of dates as several participants lacked specific attack dates in their medical records. Date of the incident demyelinating attack from the medical record was used as the index date in the sensitivity analysis.

Controls

Each case was matched to five controls randomly chosen from the general population based on sex, birth year (exact), immigrant status (yes/no), and region of residence (first 3 digits of postal code). Immigrant status was determined using linked data from the Permanent Resident Database of Immigration, Refugees and Citizenship Canada. We matched on immigrant status because a large share of people living with NMOSD in Canada is comprised of immigrants. ¹⁶ Controls were required to have no demyelinating disease ICD code (377/H46, 340/G35, 323/G36, G37) and were assigned the same index date as their matched case.

Outcomes

The primary outcomes were all-cause use of outpatient physician care and hospitalizations in the 5-year period preceding the first demyelinating disease code. We evaluated the entire 5-year period given sample size considerations. As secondary outcomes, we assessed outpatient physician visits and hospitalizations in each of the 5 years preceding the first attack or index date.

Comorbidities

Eighteen comorbidities were identified using case algorithms applied at the index date and grouped into cardiovascular conditions, pulmonary conditions, other immune-mediated conditions and mental health disorders (Supplementary Appendix Table 1).

Analysis

The NMOSD and MOGAD cohorts were analyzed and described separately. International diagnostic criteria for MOGAD had not been published when this study was commenced; however, after the study concluded, we reported the number (%) of included MOGAD participants meeting 2023 diagnostic criteria (see “Results” section). ⁶ Crude outcomes were evaluated in each cohort and compared to controls using Poisson distributions with 95% confidence intervals (CIs). To assess the association between disease group and each outcome (outpatient physician visits, hospitalizations), we used unadjusted and adjusted negative binomial regression models. Covariates in the adjusted models included sex, birth year, immigrant status (immigrant/non-immigrant), index year (1997–2010, 2011–2016, 2017–2022), and neighborhood income (low vs high income). We included the logarithm of health insurance (OHIP) eligibility time as an offset. Annual trends in rate ratios were evaluated by inclusion of an interaction term between NMOSD/MOGAD diagnosis and year pre-index date (“Year 0”). To determine whether comorbid immune-mediated conditions modified the findings, we also ran models including interaction terms for NMOSD/MOGAD diagnosis × other immune-mediated conditions, and NMOSD/MOGAD diagnosis × year pre-index × other immune-mediated conditions.

A sensitivity analysis was pre-planned for statistically significant associations observed in the analyses where the same models were repeated using the date of the incident attack from the medical record.

All statistical analyses were performed in SAS enterprise guide 8.3.

Results

NMOSD cohort

Of 101 AQP4 + NMOSD consented participants, 69 (71.1%) had AQP4 antibody tests performed by CBA, 27 (27.8%) by ELISA, 1 (1.0%) by tissue-based indirect immunofluorescence, and 4 had unknown assay type. Ninety-six (95.0%) met the inclusion criteria (Figure 1) and were matched to 479 controls. Cases and controls were demographically similar; both had a mean age at the index date of 43 years and 86.7% were female (Table 1).

Figure 1.

Flow diagram for selection of the NMOSD and MOGAD cohorts: (a) NMOSD cohort and (b) MOGAD cohort. NMOSD: neuromyelitis optica spectrum disorder; MOGAD: myelin oligodendrocyte glycoprotein antibody disease.

Table 1.

Baseline characteristics of NMOSD and MOGAD cohorts and their respective matched general population controls.

Characteristic	Value	NMOSD	NMOSD control	p value	MOGAD	MOGAD control	p value
		N = 96	N = 479		N = 61	N = 303
Sex	Females	83 (86.5%)	414 (86.4%)	0.99	36 (59.0%)	180 (59.4%)	0.96
	Males	13 (13.5%)	65 (13.6%)		25 (41.0%)	123 (40.6%)
Age at index date (years)	Mean ± SD	43.59 ± 16.64	43.66 ± 16.56	0.97	37.54 ± 15.62	37.64 ± 15.59	0.96
	Median (IQR)	43 (30-56)	43 (30-55)	0.95	38 (28-48)	38 (29-49)	0.97
Age group at index date (years)	⩽20	6 (6.3%)	30 (6.3%)	1.00	10 (16.4%)	50 (16.5%)	1.00
	21-30	19 (19.8%)	93 (19.4%)		8 (13.1%)	43 (14.2%)
	31-40	17 (17.7%)	83 (17.3%)		16 (26.2%)	75 (24.8%)
	41-50	19 (19.8%)	98 (20.5%)		13 (21.3%)	64 (21.1%)
	51+	35 (36.5%)	175 (36.5%)		14 (23.0%)	71 (23.4%)
Income quintile	1-3 (less affluent)	50 (52.1%)	284 (59.3%)	0.19	36 (59.0%)	168 (55.4%)	0.61
	4-5 (more affluent)	46 (47.9%)	195 (40.7%)		25 (41.0%)	135 (44.6%)
Urban residence		90 (93.8%)	450 (93.9%)	0.94	58 (95.1%)	285 (94.1%)	0.76
Index year	1998-2010	28 (29.2%)	139 (29.0%)	1.00	11 (18.0%)	55 (18.2%)	1.00
	2011-2016	29 (30.2%)	145 (30.3%)		16 (26.2%)	80 (26.4%)
	2017-2022	39 (40.6%)	195 (40.7%)		34 (55.7%)	168 (55.4%)
Immigrant status	Non-immigrant	60 (62.5%)	300 (62.6%)	0.981	52 (85.2%)	260 (85.8%)	0.91
	Immigrant	36 (37.5%)	179 (37.4%)		9 (14.8%)	43 (14.2%)
First attack localization	Unilateral optic neuritis	18 (18.8%)			21 (34.4%)
	Bilateral optic neuritis	<6			8 (13.1%)
	Transverse myelitis	46 (47.9%)			16 (26.2%)
	Area postrema syndrome	8 (8.3%)			<6
	Brainstem or cerebellar syndrome	<6			<6
	Cerebral syndrome or ADEM	<6			8 (13.1%)
	Diencephalic syndrome	<6			<6
	Polyfocal	8 (8.3%)			7 (11.5%)
	Unknown	<6			<6
Cardiovascular conditions: hypertension, diabetes, and/or CHF		23 (24.0%)	107 (22.3%)	0.73	13 (21.3%)	42 (13.9%)	0.14
Pulmonary conditions: COPD and/or asthma		16 (16.7%)	68 (14.2%)	0.53	10 (16.4%)	50 (16.5%)	0.98
Other immune-mediated conditions		7 (7.3%)	11 (2.3%)	0.01	<6	<6	>0.10*
Mental health disorder		<6	10 (2.1%)	1.00	<6	12 (4.0%)	>0.10*

NMOSD = neuromyelitis optica spectrum disorder, MOGAD = myelin oligodendrocyte glycoprotein antibody disease, ADEM = acute disseminated encephalomyelitis, CHF = congestive heart failure, COPD = chronic obstructive pulmonary disease.

^*Precise p-values and numbers < 6 not shown as per ICES regulations to minimize risk of loss of confidentiality and identification of participants.

For the entire 5-year period pre-index date, the probability of any outpatient visit, number of outpatient visits, probability of any hospitalization, and number of hospitalizations were higher in the NMOSD population versus controls (p < 0.02, Table 2). In the adjusted models, rate ratios were statistically significantly increased for NMOSD for outpatient visits (adjusted rate ratio (aRR): 1.47; 95% CI: 1.25–1.73) and hospitalizations (aRR: 1.67; 95% CI: 1.19–2.36) (Table 3).

Table 2.

Health care use in the NMOSD and MOGAD cohorts and their respective controls in the 5 years preceding the incident attack.

Variable	Value	NMOSD	NMOSD control	p value	MOGAD	MOGAD control	p value
Any outpatient visit		94 (97.9%)	442 (92.3%)	0.045	58 (95.1%)	276 (91.1%)	0.301
Number of outpatient visits	Mean ± SD	34.84 ± 29.68	26.11 ± 24.32	0.002	26.84 ± 24.01	23.64 ± 30.21	0.438
	Median (IQR)	27 (16-45)	20 (10-35)	0.001	20 (10-37)	16 (6-30)	0.037
Any hospitalization		50 (52.1%)	172 (35.9%)	0.003	22 (36.1%)	103 (34.0%)	0.756
Number of hospitalizations	Mean ± SD	1.03 ± 1.68	0.68 ± 1.20	0.015	0.62 ± 1.13	0.61 ± 1.67	0.944
	Median (IQR)	1 (0-1)	0 (0-1)	0.006	0 (0-1)	0 (0-1)	0.648

NMOSD = neuromyelitis optica spectrum disorder, MOGAD = myelin oligodendrocyte glycoprotein antibody.

Table 3.

Unadjusted and adjusted rate ratios for health care use in NMOSD and MOGAD cases versus their respective matched controls in the 5 years preceding the incident attack.

Matched cohort	Outcome	Unadjusted rate ratio for cases versus matched controls	95% Confidence interval (CI)	p value	Adjusted rate ratio for cases versus matched controls*	95% Confidence interval (CI)	p value
NMOSD	Number of outpatient visits	1.48	1.24-1.76	<.0001	1.47	1.25 – 1.73	<.0001
NMOSD	Number of hospitalizations	1.68	1.18-2.41	0.004	1.67	1.19 – 2.36	0.004
MOGAD	Number of outpatient visits	1.21	0.92-1.59	0.17	1.16	0.90 – 1.50	0.24
MOGAD	Number of hospitalizations	1.07	0.62-1.85	0.80	1.14	0.65 – 1.97	0.65

NMOSD = neuromyelitis optica spectrum disorder, MOGAD = myelin oligodendrocyte glycoprotein antibody.

^*Rate ratios were adjusted for sex, birth year, immigrant status, index year, and neighborhood income.

Evaluating annual trends, NMOSD cases had higher rates of outpatient visits in all 5 years before the incident attack (Figure 2(a)). Rate ratios steadily increased in the 2 years before the incident attack (rising from: aRR = 1.37, 95% CI: 1.10–1.71 to 2.24, 95% CI: 1.75–2.86 in the year before). For hospitalizations, aRR was elevated in most years, but with wide CIs, and, in particular, in the year before NMSOD onset (aRR = 2.98, 95% CI: 1.81–4.91). When interaction terms for other immune-mediated conditions were added to the model for the number of outpatient visits, there was no change in the observed associations or magnitude of the adjusted rate ratios (Supplementary Table 2); the model predicting numbers of hospitalizations was unstable, and the results did not converge.

Figure 2.

Annual adjusted rate ratios for health care use in NMOSD and MOGAD. Annual adjusted rate ratios are demonstrated for each of the 5 years preceding NMOSD or MOGAD onset, as follows: (a) number of outpatient visits in NMOSD cases versus controls; (b) number of hospitalizations in NMOSD cases versus controls; and (c) number of outpatient visits in MOGAD cases versus controls. The model for number of hospitalizations in MOGAD did not converge. Bars show 95% confidence intervals. NMOSD, neuromyelitis optica spectrum disorder; MOGAD, myelin oligodendrocyte glycoprotein antibody disease.

Given the findings observed in the NMOSD cohort, a sensitivity analysis was performed in the 89 (92.7%) NMOSD participants for whom there was a known date of the incident attack from the medical record. As in the primary analysis, NMOSD cases had statistically significant higher rates of both outpatient physician visits and hospitalizations versus controls for the 5-year period before the incident attack (Supplementary Table 3).

MOGAD cohort

Of 63 MOGAD consented participants, 61 (96.8%) met the inclusion criteria and were matched to 303 controls (Figure 1). MOGAD cases were demographically similar to controls, with 59% females and a mean age at index of 38 years (Table 1). Retrospective review of the cohort revealed that 82% met the 2023 diagnostic criteria for MOGAD. ⁶

For the 5-year period preceding the first attack, we did not observe any statistically significant (p < 0.05) difference in hospitalizations or outpatient visits between the MOGAD population and controls except for the median number of outpatient visits which was slightly higher among those who developed MOGAD (Table 2). In the adjusted models, there was no statistically significant difference in rate ratios comparing MOGAD cases to controls (Table 3).

When annual rate ratios were evaluated, aRR was increased in the year before the first attack (aRR = 1.61, 9% CI: 1.22–2.14) for the number of outpatient visits, but was 1.0 demonstrating no effect in all other years (Figure 2(c)). The multivariable model did not converge when investigating the number of hospitalizations. The results were unchanged when controlling for other immune-mediated conditions (Supplementary Table 2).

Discussion

In this study using prospectively collected health administrative data, we found evidence suggesting a prodromal phase preceding the first clinical attack of AQP4 + NMOSD. Higher health care use was observed in the 5 years preceding the first attack date in NMOSD cases compared to general population controls. These findings were consistent across both outpatient visits and hospitalizations, in the primary and sensitivity analyses, and when controlling for presence of other immune-mediated conditions. Rates of outpatient visits increased steadily in the 2 years before the first attack. This pattern is similar to increases in health care use observed in the years preceding MS onset, which is also now considered to have a prodromal phase.^3,4 Our results raise the possibility, if validated, of underlying biologic onset of NMOSD years before the first typical CNS attack and suggest that NMOSD could be identified, monitored, and treated in some cases before a disabling attack occurs.

To our knowledge, this study is the first investigation of a possible prodromal phase in people with AQP4 + NMOSD and MOGAD. Interestingly, in recent work using the US Department of Defense Serum Repository, more than 60% of blood samples collected a mean of 940 days before NMOSD clinical onset were positive for the AQP4 antibody. ⁹ Serum AQP4 antibody testing has a reported specificity of >99.5% by CBA and the AQP4 antibody is considered to be pathogenic.^17,18 Other evidence suggests that CNS inflammation in NMOSD may not be an all-or-none phenomenon. For example, some studies have shown prolonged latency of visual evoked potentials in NMOSD patients without a history of optic neuritis in the same eye. ¹⁹ In the N-Momentum clinical trial, asymptomatic or mildly symptomatic gadolinium-enhancing lesions and more pronounced elevations in serum glial fibrillary acidic protein (GFAP), a marker of ongoing destruction of astrocytes, were observed in between attacks in some participants with NMOSD, particularly those not on therapy (e.g. on placebo).^20,21 It is possible that subclinical NMOSD disease activity in the CNS, or CNS disease activity manifesting with milder, non-specific symptoms, may be present before the first typical attack. Biomarker studies investigating AQP4 antibody status and GFAP and NfL levels from serum samples drawn before the first NMOSD attack are required to investigate for biologic evidence of CNS inflammation predating clinical onset; in this study, however, we did not find any serologic tests performed predating the index attack. Serum GFAP may be an especially useful biomarker since it is significantly elevated in people with AQP4 + NMOSD compared to unaffected controls and people with MS, whereas NfL is less specific. ²⁰ These biomarkers may allow us in the future to pinpoint when CNS damage begins in prodromal NMOSD, monitor patients, and flag those in need of treatment.

Although prodromal symptoms are non-specific, certain symptoms or conditions may help to distinguish NMOSD from MS and other diseases during the prodrome. Previously, we reviewed medical records of eligible CANOPTICS participants evaluating neurologic symptoms preceding the first NMOSD attack but not meeting criteria for an attack. ²² There were 17/116 (14.7%) with earlier neurologic symptoms occurring a median of 14 months (range: 1.5–245 months) before NMOSD onset. ²² The most commonly reported neurologic symptoms included sensory symptoms and neuropathic pain. In addition, other immune-mediated conditions may be more prevalent during the NMOSD prodrome and could be involved in epitope spreading. Notably, however, we did not observe any change in the magnitude of the adjusted rate ratios when we controlled for presence of other immune-mediated conditions in the NMOSD cohort. Recognition of characteristic symptoms could assist with differential diagnosis of prodromal NMOSD versus MS in the future and direct tests such as serum AQP4 antibody and GFAP testing which, if validated, may help to identify individuals in the prodromal stage of NMOSD.

In MOGAD, we observed elevated health care use only in the year before the first attack. This finding is consistent with previous studies suggesting a possible MOGAD prodrome in close temporal association to time of disease onset. Symptoms reported within 4–8 weeks of the first MOGAD attack include upper respiratory and gastrointestinal infectious symptoms, ²³ severe headache, ²⁴ and nausea/emesis. ²⁵ There is a high reported prevalence of viral signs and symptoms in as many as 40% before MOGAD onset; ²³ viral infections have been reported, as well, before the first NMOSD attack. ²⁶ Viral infection likely represents a risk factor for MOGAD, as opposed to a prodrome, and might explain increased health care use in the year immediately before MOGAD onset, although this requires further investigation.^23,27 Furthermore, recent vaccination has been associated with MOGAD onset and could possibly explain some of the increased health care use in the year before MOGAD onset if vaccination rates did not differ across cases and controls. ²⁸ It remains possible that there could be a more protracted prodrome in some people with MOGAD which was not captured in this study due to power limitations. There have been rare reports of cranial and peripheral neuropathy and subclinical optic neuritis based on optical coherence tomography findings preceding the first typical attack of MOGAD.^15,29

Strengths of this study include the use of comprehensive prospectively collected health administrative data from a universal health insurance system and a large group of controls closely matched to each cohort. Limitations include sample size particularly as these conditions are very rare. Our approach using health administrative data only allows for capture of prodromal symptoms when the patient sought out medical care. For the hospitalizations outcome, there was greater variability in rate ratios and CIs due to small number of events. For the same reason, the model investigating the annual number of hospitalizations in the MOGAD cohort was unstable. At the index date, prevalence of other immune-mediated conditions was greater in our NMOSD cohort compared to controls, although lower than that reported in some other cohorts. ³⁰ Controlling for the presence of other immune-mediated conditions did not change our results. We reported comorbidities around the time of the first attack based on prospectively collected data, whereas other studies have used retrospective analysis of those with established NMOSD and variable disease durations. ³⁰ However, we could not identify all possible immune-mediated conditions such as thyroiditis. NMOSD and MOGAD participants were identified through CANOPTICS from demyelinating disease clinics in Ontario. We cannot exclude the possibility of selection bias although most individuals diagnosed with these rare diseases, especially with NMOSD, are referred to specialized centers. About 28% of NMOSD cases were diagnosed using the ELISA AQP4 assay which has a slightly higher false-positive rate than CBA (~5%). Some discrepancies were observed between the first health administrative demyelinating disease code and the date of the first attack from the medical record; however, most were quite close. Using the first health administrative demyelinating code as the index date offered the advantage of complete ascertainment of dates for the whole cohort and avoidance of recall bias. In our sensitivity analysis using the date of the first attack from the medical record, we did not observe any substantial change in results. It is possible that increased health service use during the NMOSD prodrome could represent a missed attack, as has been argued with the MS prodrome. However, NMOSD attacks are typically more severe than MS attacks and do not go unnoticed. Recognition of mild disease activity during the prodrome before the first fulminant NMOSD attack would thus be an important observation, even if labeled with another term. This study is hypothesis generating; it would be valuable to investigate a possible prodromal phase of NMOSD in a larger cohort including indications for increased health care use during the prodromal phase, which we could not analyze in this study due to sample size limitations.

In conclusion, our findings suggest a prodromal phase to AQP4 + NMOSD, with increased health care use observed up to 5 years preceding the first clinical attack. In MOGAD, we did not find evidence of a prodromal phase but health care use was higher in the year before the first attack which may be explained by a brief prodrome or by risk factors for developing MOGAD such as infectious illness. Disease pathogenesis may commence in NMOSD years in advance of the first attack; further investigation is required including biomarker studies. Clinical implications of these findings include the possibility that some NMOSD patients could be diagnosed and monitored during the prodromal phase with biomarkers such as AQP4 serology and GFAP levels and treated to prevent the highly disabling attacks associated with this disease.