Incidence, Etiology, and Long-Term Outcome of Acute Myelitis in Stockholm County, Sweden

Abstract

Background and Objectives

Myelitis is a relatively common clinical entity for neurologists, with diverse underlying causes. The aim of this study was to describe the incidence of myelitis, its causes, clinical presentation, and factors predicting functional outcomes and relapses.

Methods

Using the Swedish National Patient Registry, we identified all adult patients in Stockholm County between 2008 and 2018 using International Classification of Diseases, 10th Edition (ICD-10) codes likely to include myelitis. We collected medical records and classified patients using a modification of the 2002 Transverse Myelitis Consortium Group criteria. Long-term follow-up data were collected for patients not diagnosed with multiple sclerosis (MS) or neuromyelitis optica spectrum disorder as a result of the initial myelitis.

Results

We identified 2,321 individuals, of whom 461 were patients with myelitis. The crude mean incidence of all-cause myelitis was 24.9 (95% CI 16.7–33.9) cases per million person-years, of which idiopathic myelitis had an incidence of 8.0 (95% CI 3.8–12.1) cases per million person-years. Partial myelitis was found in 80% of patients. Poor functional outcome was found in 11% of the cohort and correlated, in a multivariate logistic model, with age older than 50 years (OR 4.26, 95% CI 1.75–10.40), transverse spinal cord lesions (odds ratio [OR] 6.85, 95% CI 2.68–17.52), elevated CSF count of polymorphonuclear cells (OR 6.09, 95% CI 1.56–23.72), and elevated CSF/serum albumin ratio (OR 3.17, 95% CI 1.23–8.17). The median follow-up time was 5.4 years. Relapses occurred in 27% of patients with idiopathic myelitis and 72% of patients with unspecified demyelinating disease of the CNS. An increased relapse rate after idiopathic myelitis was found to be associated, in a multivariate model, with the presence of oligoclonal bands (incidence rate ratio [IRR] 4.47, 95% CI 1.70–11.73), transverse spinal cord lesions (IRR 2.81, 95% CI 1.11–7.12), and multifocal spinal cord lesions (IRR 2.82, 95% CI 1.03–7.69). Around half (48%) of all patients with myelitis received MS diagnosis during the study period.

Discussion

This large population-wide study describes a relatively high incidence of myelitis and low risk of relapses after idiopathic myelitis. A complete diagnostic workup of myelitis, including MRI of the entire CNS and collection of CSF, is essential in evaluating underlying causes and prognosis.

Introduction

Myelitis is a clinical entity of inflammatory myelopathy that presents with a variable combination of motor, sensory, and autonomic (usually bladder) dysfunction.¹ Most cases of myelitis are associated with an immune-mediated disease or an infection.² Among disease-associated myelitis, multiple sclerosis (MS) is the most frequent underlying disease, followed by neuromyelitis optica spectrum disorder (NMOSD) and myelin oligodendrocyte glycoprotein antibody–associated disease (MOGAD), among others.³ Multiple infectious agents have been associated with myelitis, i.e., Borrelia (Lyme disease), HIV, mycoplasma, enterovirus D68, and Herpes simplex type 2.⁴ In the absence of a definite cause for the myelitis, it is commonly termed as idiopathic.

Few previous studies have assessed the incidence of myelitis, with reported incidence rates varying between 1.3 and 24.6 cases per million person-years.^5-10 Studies that specifically addressed idiopathic myelitis have shown an incidence between 0.9 and 8.6 per million person-years.^5,6,8,9,11

The prognosis of neurologic recovery in patients with myelitis varies greatly, ranging from complete recovery to severe disability. Previous studies have reported substantial disability in approximately 13%–45% of the patients.^7,12-20 The risk of recurring neuroinflammation after myelitis and conversion to a diagnosis of MS is generally believed to be low for idiopathic myelitis, with reported rates between 13% and 30%.^13,21-23

While there is abundant literature on many of the diseases that can cause myelitis, a broader perspective with myelitis as the starting point has been less explored. Only a small number of studies have reported the incidence of myelitis, its presentation, and functional outcome. Furthermore, there is limited understanding of risk factors of neuroinflammatory recurrence and/or the development of a chronic neuroinflammatory condition after idiopathic myelitis.

In this study, we described the incidence of myelitis in Stockholm County, Sweden, between 2008 and 2018, as well as the clinical characteristics and findings from the diagnostic workup of myelitis cases. We also analyzed potential factors associated with poor functional outcomes at 6 months after myelitis onset and the risk of relapsing neuroinflammation in patients with idiopathic myelitis.

Methods

Study Design, Patient Inclusion, and Data Set

We performed a cohort study of patients with incident myelitis as an adult (age 18 years or older) in Stockholm County, Sweden, between 2008 and 2018, using retrospectively collected data from the Swedish National Patient Register (NPR) and medical records. The NPR includes all specialized inpatient and outpatient contacts during the study period identified with a national personal identification number. Because there is no single International Classification of Diseases, 10th Edition (ICD-10) code for myelitis, the diagnostic codes G04.0 (acute disseminated encephalitis), G04.8 (other encephalitis, myelitis, and encephalomyelitis), G04.9 (encephalitis, myelitis, and encephalomyelitis, unspecified), G05.0 (encephalitis, myelitis, and encephalomyelitis in diseases classified elsewhere), G05.1 (encephalitis, myelitis, and encephalomyelitis in viral diseases classified elsewhere), G05.8 (encephalitis, myelitis, and encephalomyelitis in other diseases classified elsewhere), G36.0 (neuromyelitis optica disorder), G36.9 (acute disseminated demyelination, unspecified), G37.3 (acute transverse myelitis in demyelinating disease of the CNS), G37.8 (other specified demyelinating diseases of the CNS), and G37.9 (demyelinating disease of the CNS, unspecified) were used to identify patients in the NPR and collect medical records from the relevant health care providers.

The review of medical reports was divided between 2 neurologists (D.I.J., E.I.) and a neurology resident (N.M.). First, we excluded patients without myelopathy, those with a known diagnosis of MS or NMOSD, those with onset of myelitis symptoms outside of Stockholm County or study period, or those with onset before the age of 18 years. The next step was to apply the 2002 Transverse Myelitis Consortium Working Group (TMCWG) criteria²⁴ with later modification,²⁵ allowing for partial myelitis to be included. According to the criteria, patients were classified as having complete myelitis when there was moderate-to-severe bilateral weakness, a symmetric sensory level, and autonomic dysfunction. In patients classified as having partial myelitis, the symptoms were either minor, unilateral, or bilateral, or when more severe, symptoms were unilateral or highly asymmetric.²⁵ Furthermore, myelitis can be classified as definitive when CNS inflammation is demonstrated by detection of CSF pleocytosis, an elevated IgG index, or detection of gadolinium (Gd) enhancement on MRI.²⁴ When CNS inflammation was not demonstrated, the myelitis was included but classified as possible, as proposed in the 2002 TMCWG criteria.²⁴ Further inclusion criteria included progression to nadir between 4 hours and 21 days,²⁴ but cases with progression to nadir >21 days after disease onset were included if both the recorded medical record diagnosis and our medical record review agreed with a diagnosis of myelitis. As recommended by the criteria, patients with a history of radiation of the spinal cord within 10 years before the clinical onset of myelitis, or with clinical findings consistent with affection of the anterior spinal artery, or with MRI findings indicating a spinal vascular lesion, were excluded.²⁴

From all patients with myelitis, we collected data on the date of symptom onset, demographics (sex, age, and ethnicity), previous comorbidities, possible triggers of inflammation within 6 weeks before onset (infection and/or vaccination), time to nadir, clinical presentation, laboratory findings, MRI findings, acute treatment, initial diagnosis, and functional outcome at six-month follow-up. A standardized protocol was applied during medical record data retrieval. The diagnosis was based on the initial neurologist evaluation and categorized into idiopathic myelitis, unspecified demyelinating disease (myelitis along with additional signs of an underlying demyelinating disease without meeting MS diagnostic criteria), MS, NMOSD (including MOGAD because it was not an established separate entity during the study period), acute demyelinating encephalomyelitis (ADEM), myelitis associated with systemic autoimmune disease (SAA), infection-associated myelitis, paraneoplastic myelitis, and iatrogenic myelitis (drug-associated and vaccine-associated).

For the follow-up part of the study, we excluded patients with the initial diagnosis of MS or NMOSD. All other patients were followed until a diagnosis of MS or NMOSD was established, or until the last notification in the medical chart. Information on death was retrieved from the medical charts. The study was reported according to the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines.²⁶

Functional Outcome Measures

Functional outcome at 6 months after the onset of the myelitis was assessed using the modified Rankin Scale (mRS) and classified as favorable if the mRS score was 0–1 and poor if the mRS score was ≥2.

Definition of Population Data

Population data with yearly age and sex distribution in Sweden and Stockholm County for the relevant years were extracted from an open official database.²⁷

Statistical Analysis

Statistical analysis was performed in Stata 18.²⁸ Descriptive data were presented as medians and interquartile range (IQR), proportions, or number of patients, as appropriate. CIs for incidence rates were calculated from the Poisson distribution. For analysis of trends of incidence rates, a linear regression model was used. Incidence rates standardized to the Swedish population with 95% CIs were calculated and stratified by sex and age groups.

For regression models, age was dichotomized at 50 years and the presence of mononuclear cells (MNCs) in the CSF was categorized as normal (0–5 × 10⁶/L), low to moderate (6–49 × 10⁶/L), or high (≥50 × 10⁶/L). In all patients with myelitis, a logistic regression model was applied to assess risk factors of poor functional outcome, first testing each variable in a univariate model and then creating a multivariate model in a stepwise exclusion manner using log-likelihood and goodness of fit to find the best-fitting model. We further tested the effects of the underlying diagnosis (with idiopathic myelitis as baseline) on the risk of poor functional outcome, first only adjusting for age 50 years or older and sex and then including the variables from the best-fitting multivariate model. Analysis of variables potentially associated with relapse risk after idiopathic myelitis was performed using negative binomial regression and presented as an incidence rate ratio (IRR) with 95% CI. We first performed univariate analysis, and thereafter created a multivariate model in a stepwise excluding manner using log-likelihood and goodness of fit. Kaplan-Maier survival estimations were performed at 2, 5, and 10 years from baseline. Statistical significance level was set at p < 0.05 throughout.

Standard Protocol Approvals, Registrations, and Patient Consents

The study was approved by the Swedish Ethical Review Authority (Dnr. 2020-01398), which permitted the waiver of individual consent collection.

Data Availability

The data collected in this study cannot be shared because of Swedish data privacy laws. Any researcher can access the registered data by obtaining ethical approval from the Swedish Ethical Review Authority and requesting the data from the Swedish National Board of Health, Welfare and Statistics Sweden, and Stockholm Center for Health Data.

Results

Patient Inclusion

The NPR search identified 2,321 individuals with an ICD-10 code of G04, G05, G36.0, G36.9, G37.3, G37.8, or G37.9 during the study period (Figure 1). Medical records were obtained for 2,280 (98%) of the patients. The initial ICD-10 code screening identified 461 potential myelitis cases (20%). Roughly two-thirds (68%) of the patients were excluded because myelopathy was not present. Other reasons for exclusion were myelitis onset outside of the designated period, county, or age range (7%); a previously established diagnosis of MS or NMOSD (2%); and fulfillment of exclusion criterion or another cause deemed more likely (2%).

Figure 1. Patient Inclusion.

Flowchart of the patient inclusion and data collection. ICD = International Classification of Diseases; MS = multiple sclerosis; NMOSD = neuromyelitis optica spectrum disorder.

Characteristics of Patients With a Diagnosis of Myelitis

Table 1 provides the characteristics of patients with myelitis, categorized by the initial diagnosis. Around one-third (34%) of the myelitis cases were classified as idiopathic, 30% received an MS diagnosis during the diagnostic workup, 15% were categorized as unspecified demyelinating disease of the CNS, and 11% were identified as infection-associated.

Table 1.

Characteristics of Patients With Myelitis

	Total	Idiopathic myelitis	Demyelinating disease of CNS, unspecified	MS	NMOSD	ADEM	SAA myelitis	Paraneoplastic myelitis	Iatrogenic myelitis	Infection-associated myelitis
N (%)	461	157 (34)	71 (15)	139 (30)	14 (3)	10 (2)	6 (1)	3 (1)	11 (2)	50 (11)
Median age at onset, y (IQR)	39 (30–50)	43 (33–53)	37 (30–45)	34 (27–47)	42 (26–60)	34 (23–53)	52 (45–63)	65 (50–74)	42 (32–50)	44 (30–54)
Definitive myelitis, %	82	73	76	91	93	100	67	100	73	88
Complete myelitis, %	19	19	11	9	57	80	67	0	18	32
Female, %	62	61	61	66	86	50	50	0	45	58
Caucasian, %	87	90	84	85	79	80	67	100	91	90
Autoimmune comorbidity, %	12	12	8	8	36	10	50	0	55	8
Infection/vaccination within 6 wk of onset, %	17	12	7	7	7	40	0	0	55	66
Time to nadir, %
12–24 h	4	7	4	1	0	0	0	0	0	4
1–7 d	52	55	43	48	62	50	17	33	64	65
8–21 d	34	29	45	40	8	50	67	0	27	22
>21 d	10	9	7	10	31	0	17	67	9	8
Motor symptoms, %	56	58	50	49	71	80	83	100	55	66
Sensory symptoms, %	96	97	97	97	93	100	100	67	91	90
Sensory level, %	50	46	44	49	64	90	67	0	27	66
Bladder/bowel dysfunction, %	41	39	27	28	79	100	83	100	36	72
Transverse lesion, %	34	36	23	16	100	89	67	67	30	64
LETM, %	22	20	10	6	93	78	50	33	20	58
Multifocal spinal lesions, %	32	18	33	44	21	78	67	67	0	38
Gd enhancement, % (n testeda)	66 (393)	64 (137)	60 (58)	70 (124)	71 (14)	67 (9)	60 (5)	67 (3)	70 (10)	70 (33)
Brain lesions (≥1), % (n testeda)	53 (448)	5 (154)	87 (71)	99 (137)	43 (14)	80 (10)	33 (6)	33 (3)	50 (10)	23 (43)
Elevated CSF MNC count, % (n testeda)
Minor/moderateb	41 (434)	32 (151)	44 (62)	53 (132)	33 (12)	30 (10)	0 (6)	33 (3)	50 (10)	43 (48)
Highc	10 (434)	7 (151)	3 (62)	1 (132)	50 (12)	60 (10)	17 (6)	0 (3)	0 (10)	30 (48)
Elevated CSF PMN count, % (n testeda)	5 (434)	3 (151)	2 (62)	1 (132)	17 (12)	30 (10)	0 (6)	0 (3)	0 (10)	20 (48)
Elevated CSF/serum albumin ratio, % (n testeda)	39 (417)	25 (143)	38 (64)	65 (127)	38 (10)	0 (9)	17 (6)	0 (3)	20 (10)	22 (45)
Elevated IgG index, % (n testeda)	30 (409)	27 (142)	23 (64)	18 (130)	60 (8)	78 (9)	33 (6)	67 (3)	40 (10)	60 (37)
Oligoclonal bands, % (n testeda)	61 (420)	49 (144)	68 (66)	87 (136)	64 (11)	10 (10)	33 (6)	0 (3)	44 (9)	29 (35)
Poor functional outcome at 6 mo (mRS ≥2), %	11	9	7	2	38	33	33	100	0	29

Abbreviations: ADEM = acute demyelinating encephalomyelitis; IQR = interquartile range; LETM = longitudinally extensive transverse myelitis; MNC = mononuclear cell; mRS = modified Rankin Scale; MS = multiple sclerosis; NMOSD = neuromyelitis optica spectrum disorder; PMN = polymorphonuclear cell; SAA = systemic autoimmune-associated.

^aNumber of patients tested.

^bMinor/moderate 6–49 × 106/L.

^cHigh ≥50 × 106/L.

Most of the patients (80%) presented with partial symptoms, and most (82%) had demonstrable intrathecal inflammation (fulfilling the definition of definitive myelitis).

The cohort had a female dominance (60%), with a prominent sex difference in younger patients that disappeared in patients aged older than 60 years (Figure 2). The median age was 39 years, and there was a bimodal age distribution with peaks around the age of 30 years and 45 years for both men and women (Figure 2). A previous comorbid autoimmune disease was found in 12% of all patients, with a higher proportion observed in patients with NMOSD (36%) compared with patients with MS (8%) (Table 1). An infection or vaccination was noted within 6 weeks before symptom onset in 17% of the myelitis cases, with a notably high occurrence in patients with ADEM (40%) compared with 7% of patients with MS, NMOSD, or demyelinating disease of the CNS and 12% of patients with idiopathic myelitis (Table 1).

Figure 2. Distribution of Age and Sex in Patients With Myelitis.

Histogram of the distribution of female patients (red) and male patients (blue) and age with 5-year intervals.

The progression of myelitis symptoms from symptom onset to nadir occurred within 1–7 days in approximately half (52%) of the patients and within 8–21 days in approximately one-third (34%) (Table 1). Only a small group of patients (4%) reached nadir between 4 and 24 hours while 10% continued to progress after 21 days. Motor symptoms were found in 56% of the cases while sensory disturbance occurred in almost all patients (96%), with nearly half (49%) presenting with a unilateral or bilateral sensory level (Table 1). Bladder and/or bowel disturbance was observed in 41% of patients, and around a quarter (24%) experienced new-onset pain in the neck or back.

MRI spinal cord lesions were localized to the cervical spine (65%), the thoracic spine (53%), and the conus (10%). Few patients (3%) had no detectable lesion on MRI. A transverse spinal cord lesion was found in 34% of the patients, and 22% had longitudinally extensive transverse myelitis (LETM). Almost a third (32%) had multifocal spinal cord lesions, and brain lesions were found in around half (53%) of the patients. Gd enhancement in the spinal cord was observed in two-thirds (66%) of all cases (Table 1).

In the CSF, elevated MNC counts were found in 40% of the patients at low to moderately increased levels (6–49 × 10⁶/L) and in 10% at high levels (≥50 × 10⁶/L) (Table 1). Elevation of polymorphonuclear cell (PMN) count was detected in 5% of the patients. Elevated CSF/serum albumin ratios were found in 30% of patients, 39% had an increased IgG index, and oligoclonal bands (OCBs) were observed in 61%.

During the initial diagnostic workup, aquaporin-4 antibodies (AQP4-IgG) were found in the serum of 7 patients (of 143 tested) and myelin oligodendrocyte glycoprotein antibodies (MOG-IgG) in the serum of 3 (of 46 tested). Specifically, 3 (of 10) patients with ADEM and 1 (of 6) postvaccination patients tested negative for MOG-IgG while none of the postinfectious patients were tested. For AQP4-IgG, 3 (of 6) patients with SAA myelitis tested negative.

Almost two-thirds (63%) of the patients received acute treatment with high-dose corticosteroids, and treatment escalation to IV immunoglobulins was present in 4% and to plasmapheresis in 5%.

Underlying Diseases of the Myelitis Cohort

We categorized 6 patients as having SAA myelitis: 2 with systemic lupus erythematosus (SLE), 1 with scleroderma, 1 with ankylosing spondylitis, 1 with sarcoidosis, and 1 with hypereosinophilic syndrome. Three myelitis cases were classified as paraneoplastic: 1 with B-cell lymphoma, 1 with myeloma, and 1 with acute systemic mastocytosis. Only one of the patients tested negative for paraneoplastic antibodies. In 11 patients, an iatrogenic trigger was considered causative, including 4 with ongoing TNF-α inhibitor treatment, 6 patients with vaccination shortly before symptom onset, and 1 with graft-versus-host disease.

There were 50 patients regarded to have an infection-associated myelitis. These included; Borrelia burgdorferi (n = 10), enteroviruses (n = 2), Herpes simplex virus (n = 3), Varicella zoster virus (n = 1), HIV (n = 3), tick-Borne encephalitis (TBE) virus (n = 3), Listeria monocytogenes (n = 1), Campylobacter jejuni (n = 1), Mycoplasma pneumonia (n = 5), fungal (n = 1), and in addition, 17 were deemed postinfectious and 3 as parainfectious.

Incidence of Myelitis in Stockholm County (2008–2018)

The crude mean incidence of all-cause myelitis was 24.9 (95% CI 16.7–33.9) per million person-years, of which the incidence of idiopathic myelitis was 8.0 (95% CI 3.8–12.1) per million person-years (Figure 3). No change in incidence was found over the study period for all-cause myelitis (p = 0.42) or for the subgroup of idiopathic myelitis (p = 0.18).

Figure 3. Incidence of Myelitis in Stockholm County, 2008-2018.

Yearly incidence of all-cause myelitis (purple) and idiopathic myelitis (green). Gray areas represent 95% CIs, and the dashed lines refer to the average incidences over the period.

The age-standardized and sex-standardized incidence rates in Sweden were 23.3 (95% CI 16.6–32.6) per million person-years for all-cause myelitis and 8.1 (95% CI 4.3–14.3) per million person-years for idiopathic myelitis. Detailed incidence rates by age group and sex are presented in eTable 1.

Functional Outcome 6 Months After Myelitis

At 6 months after myelitis onset, most patients (60%) had no or negligible symptoms (mRS score = 0), 29% had minor remaining symptoms (mRS score = 1), and 11% had debilitating symptoms (mRS score ≥2) and were categorized as having poor outcomes.

Depending on the initial diagnosis, there were pronounced differences in functional outcome at 6-month follow-up. Only 2% of patients with MS and 9% of those with idiopathic myelitis had a mRS score ≥2 compared with 38% of patients with NMOSD and 32% of patients with infection-associated myelitis.

In a univariate logistic model, an increased risk of poor outcome was associated with age 50 years or older, complete myelitis, MRI findings of transverse spinal cord lesions or LETM, CSF findings of high MNC count, elevation of PMN count, and elevated CSF/serum albumin ratio, whereas the presence of OCBs correlated with a lower risk of poor outcome (Figure 4A). In the final multivariate model, an increased risk of poor functional outcome was associated with age 50 years or older, transverse spinal cord lesions, elevated CSF PMN count, and elevated CSF/serum albumin ratio (Figure 4B). When testing the effects of the underlying diagnosis on risk of poor functional outcome, we found, when only adjusting for age and sex, a lower risk for MS patients but higher risk for patients with NMOSD, ADEM, and infection-associated myelitis (eFigure 1 part A). Then, after including the variables from the final multivariate model (transverse spinal cord lesions, elevated count of PMNs in the CSF, elevated CSF/serum albumin ratio, and presence of OCBs), there was no significant effect of the diagnosis while the other variables maintained the effects as in the final multivariate model (eFigure 1 part B).

Figure 4. Univariate and Multivariate Models for Risk of Poor Functional Outcome at 6 Months After Myelitis.

(A) Univariate logistic models with odds ratios (ORs) for poor functional outcome (modified Rankin Scale score ≥2), 95% CIs. (B) Multivariate logistic model with odds ratios (ORs) for poor functional outcome (modified Rankin Scale score ≥2) with 95% CIs.

Long-Term Follow-Up

A total of 308 patients (67% of the total myelitis cohort) did not receive a diagnosis of MS or NMOSD during the initial workup and were followed for a median of 5.4 years (IQR 1.6–9.4). Table 2 presents the follow-up data. Overall, one-third (34%) of the follow-up cohort had 1 or more neuroinflammatory relapses with an incidence rate of 5.9% (95% CI 4.9–7.1) per year.

Table 2.

Relapses and Conversion to MS or NMOSD Diagnosis During Follow-Up

	Total	Idiopathic myelitis	Unspecified demyelinating CNS disease	ADEM	SAA myelitis	Paraneoplastic myelitis	Iatrogenic myelitis	Infection-associated myelitis
N (%)	308	157 (51)	71 (23)	10 (3)	6 (2)	3 (1)	11 (4)	50 (16)
Median follow-up time, y (IQR)	5.4 (1.6–9.4)	6.4 (2.7–9.7)	1.5 (0.6–5.8)	5.4 (1.9–6.3)	6.3 (4.9–9.4)	3.0 (1.4–7.5)	5.5 (0.7–10.5)	6.9 (2.8–10.5)
Relapsed, n (%)	104 (34)	42 (27)	51 (72)	2 (20)	1 (17)	0 (0)	2 (20)	6 (12)
Relapse rate per patient-y (95% CI)	0.059 (0.049–0.071)	0.041 (0.030–0.056)	0.209 (0.159–0.275)	0.043 (0.011–0.173)	0.021 (0.004–0.178)	N.A.	0.030 (0.007–0.119)	0.018 (0.008–0.040)
Relapse-free at 2 y (95% CI), %	78 (73–83)	90 (84–94)	40 (29–52)	78 (36–94)	100	100	80 (41–95)	93 (80–98)
Relapse-free at 5 y (95% CI), %	70 (65–75)	78 (70–84)	33 (22–44)	78 (36–94)	100	100	80 (41–95)	91 (77–96)
Relapse-free at 10 y (95% CI), %	63 (56–69)	70 (61–78)	24 (14–36)	78 (36–94)	75 (13–96)	N.A.	80 (41–95)	91 (77–96)
Conversion to a diagnosis of MS, n (%)	81 (26)	29 (18)	45 (63)	1 (10)	0 (0)	0 (0)	2 (18)	4 (8)
Conversion to a diagnosis of NMOSD, n (%)	9 (2)	3 (2)	4 (6)	1 (10)	1 (17)	0 (0)	0 (0)	0 (0)
Mortality, n (%)	21 (7)	9 (6)	3 (4)	1 (10)	2 (33)	3 (100)	0 (0)	3 (6)

Abbreviations: ADEM = acute demyelinating encephalomyelitis; IQR = interquartile range; MS = multiple sclerosis; N.A. = not applicable; NMOSD = neuromyelitis optica spectrum disorder; SAA = systemic autoimmune-associated.

Time to first relapse was estimated with Kaplan-Meier survival curves at 2, 5, and 10 years from the initial myelitis (Table 2). Among patients with unspecified demyelinating disease of the CNS, only 40% (95% CI 29%–52%) of the patients remained relapse-free at 2 years, 33% (95% CI 22%–44%) at 5 years, and 24% (95% CI 14%–36%) after 10 years. A slower relapse rate was observed in patients with idiopathic myelitis, and 90% (95% CI 84%–94%) were relapse-free after the first 2 years, 78% (95% CI 70%–84%) after 5 years, and 70% (95% CI 61%–78%) after 10 years (Table 2).

Risk factors of a neuroinflammatory relapse after an idiopathic myelitis were analyzed with univariate models (Figure 5A). An increased relapse risk was found in patients with MRI findings of transverse and/or multifocal spinal cord lesions and with CSF findings of low-to-moderate increase in MNCs, elevated IgG index, and presence of OCBs.

Figure 5. Univariate and Multivariate Models for Relative Relapse Risk in Idiopathic Myelitis.

(A) Univariate negative binomial model of incidence rate ratios (IRRs) for neuroinflammatory relapses with 95% CIs. (B) Multivariate negative binomial model of incidence rate ratios (IRRs) for neuroinflammatory relapses with 95% CIs.

In a multivariate model (Figure 5B), we found an increased IRR of relapses associated with the presence of OCBs and MRI findings of transverse and multifocal spinal cord lesions.

During the follow-up, conversion to MS diagnosis occurred in 45 patients (63%) initially diagnosed with unspecified demyelinating disease of the CNS and in 29 patients (18%) with idiopathic myelitis. This resulted in almost half (48%) of the total myelitis cohort receiving a diagnosis of MS. Similarly, 19 patients (4% of the total cohort) ended up with an NMOSD diagnosis, including 4 patients with unspecified demyelinating disease of the CNS (2 double seronegative and 2 AQP4-IgG negative), 3 with idiopathic myelitis (1 AQP4-IgG positive, 1 MOG-IgG positive, and 1 with unknown antibody status), 1 with ADEM (AQP4-IgG positive), and 1 with SLE (AQP4-IgG positive). At the end of the follow-up, 122 (26%) of the total cohort were classified as having idiopathic myelitis, of whom 8 had recurrent idiopathic myelitis (4 double seronegative and 2 AQP4-IgG negative).

We found that 21 patients (6%) had died during the follow-up, mostly (n = 12) due to a malignancy. Only 2 patients died directly as a result of the myelitis and concurrent neuroinflammation (1 with ADEM and 1 with TBE encephalomyelitis).

Discussion

The study encompasses one of the largest cohorts of myelitis cases studied, with patients included from the whole population over an eleven-year period. We describe the clinical presentation and functional outcome at 6 months, as well as long-term follow-up data, with a median duration of more than 5 years, which allowed us to analyze risk factors of relapsing neuroinflammation after idiopathic myelitis.

The incidence of around 25 myelitis cases per million person-years, of which 8 were idiopathic myelitis, is relatively high compared with older studies.^7-10 Still, similar numbers have been described in Canterbury, New Zealand,⁶ for both total and idiopathic myelitis, and in Olmsted County, US, for idiopathic myelitis.¹¹ Of interest, a recent Finnish study reported a lower incidence of 10.4 myelitis cases per million person-years.⁵ In contrast to our study, the Finnish study did not include possible cases of myelitis and had a much lower proportion of idiopathic myelitis compared with this study, possibly because of a more hospital-centered approach.

The term transverse myelitis has for several years been the preferred term for immune-mediated myelopathies but has met criticism in more recent years.²⁹ The clinical characteristics of our large cohort supports the criticism and highlights the limitation of the 2002 TMCWG criteria because fewer than 10% of patients with MS and just over half of patients with NMOSD in our cohort fulfilled the “transverse” criteria. Categorizing myelitis cases as complete or partial may still be clinically useful in guiding the initial diagnostic workup, as illustrated by the higher proportions of complete myelitis in NMOSD, ADEM, SAA myelitis, and infection-associated myelitis.

Most patients had observable signs of CNS inflammation (Gd enhancement, CSF pleocytosis, and/or elevated IgG index) while 18% had no signs of inflammation and were classified as possible myelitis. In some of these cases, objective signs of inflammation may have been missed because of early imaging and CSF sampling while others might have been missed because of diagnostic delays. We found that around one-third of the patients without objective inflammation had OCBs and 20% had elevated CSF-neurofilament light chain (NfL) levels. These and other more recent biomarkers, such as C-X-C motif chemokine ligand 13 and interleukin-6 (for which our data were too sparse for meaningful analysis), are also likely to increase the sensitivity for demonstrating inflammation and merit further investigation as possible diagnostic biomarkers.

The time to reach nadir of myelitis symptoms was found to be up to 21 days in most of the patients, which is in line with the present myelitis classification.²⁴ Our study included a subset of patients (10%) who had a clinical progression beyond 21 days, with gradual clinical deterioration up to 3 months after onset. This has been reported in other myelitis cohorts, in similar proportions,³⁰ and has specifically been described in neurosarcoidosis and paraneoplastic myelitis.³¹ In our cohort, prolonged progression to nadir was more common in NMOSD and paraneoplastic myelitis. Only around 4% of all patients had a rapid progression (4–24 hours) to nadir, and most of these patients were classified as having idiopathic myelitis. Of interest, in this small subgroup, none experienced poor outcomes or developed chronic autoimmune disease during follow-up. This suggests that these cases may have a different, more benign underlying pathophysiology.

A small proportion of patients with myelitis (3%) had no visible MRI lesions. Notably, two-thirds of these patients were diagnosed with infection-associated myelitis. Possible explanations for these cases include misdiagnosed polyradiculopathies or polyneuropathies, or differences in the diagnostic workup (e.g., earlier MRI scans and/or fewer follow-up MRI scans). In addition, MRI-negative myelitis has been reported, for example, in MOGAD³² and anti-glycine receptor antibody–associated disease.³³

Approximately 11% of the patients had marked disability (mRS score ≥2) at 6 months after the myelitis, which is a relatively lower proportion than previously reported.^7,12-20 Our population-based approach may have included a larger number of patients with mild myelitis. We identified several signs of aggressive myelitis (transverse spinal cord lesions, elevated CSF/serum albumin ratio, and elevated CSF PMN count) as well as older age (defined as ≥50 years) to be associated with poor functional outcome. When adjusting for these variables, the underlying diagnosis had no significant effect on the risk of poor functional outcome. Markers of neuronal injury (e.g., CSF NfL and glial fibrillary acidic protein) have been associated with worse functional outcomes in other studies.³⁴ Unfortunately, our data were too incomplete regarding these markers for purposeful interpretation.

During follow-up, we found that patients with unspecified demyelinating disease of the CNS had the highest yearly relapse rate (approximately 20% per year) and most received a diagnosis of MS as a result. This high rate of relapse and conversion to a diagnosis of MS is in line with previous reports of myelitis occurring alongside brain MRI lesions.^35,36 The newly proposed MS criteria might benefit this group, enabling earlier MS diagnosis.³⁷ Despite this, it is important to note a small group of patients (12% in our cohort) with OCBs and over 5 years of follow-up time without relapses or conversion to MS, which highlights the persistent challenges in accurately diagnosing MS.

We found a yearly relapse rate of 4% in the idiopathic myelitis group. The relapses were evenly distributed throughout the follow-up period, with some patients experiencing their first relapse a decade after the initial myelitis. Most patients who relapsed (29/42) received a diagnosis of MS. Increased relapse rates were found in patients with OCBs and transverse and/or multifocal spinal lesions in the multivariate model. Both OCBs and multifocal spinal lesions have, in previous studies, been associated with an increased risk of a relapse, contrary to transverse lesions.^13,21-23 Unexpectedly, we found no correlation between younger age and increased relapse risk in contrast to previously reported findings.^22,23

The strength of this study lies in broad inclusion of patients from an entire regional population and the long study period. In addition, most of the initial diagnostic evaluations and follow-up assessments have been conducted within specialized neurology clinics. Data were systematically extracted from the original medical records by a neurologist or neurology resident, ensuring a high level of clinical accuracy.

However, there are limitations inherent to the retrospective design of the study. The completeness of the data relies on the quality and thoroughness of the initial diagnostic workup and clinical documentation. The completeness of the initial workup ranged from 97% of patients with baseline brain MRI to 85% having Gd MRI sequences. Unfortunately, we did not keep an account of whether individual MRI scans covered the entire spinal cord, affecting the certainty of the data regarding distribution of spinal lesions. Approximately 2% of the medical records were unavailable for review. Some cases of myelitis may have been overlooked, either because they were assigned a diagnostic code not included in our search or because the symptoms were too mild to prompt the patient to seek medical attention or for clinicians to suspect a myelitis diagnosis.

Furthermore, there has been significant progress in the quality and accessibility of antibody testing during the study period. Our cohort had limitations in that only a small proportion of the patients underwent optimal testing for AQP4-IgG and MOG-IgG. Immunoblot assays were used for AQP4-IgG and MOG-IgG analysis from the beginning of the study period and until 2012 and 2014, respectively, whereafter the more accurate³⁸ cell-based assays were used. In addition, until 2014 for AQP4-IgG and 2018 for MOG-IgG, the analysis was conducted by an external laboratory (Wieslab, Malmö, Sweden). Nevertheless, as our follow-up period extended into a time when current gold standard antibody analyses were readily available, we would expect that most patients with NMOSD or relapsing MOGAD have been identified. However, some monophasic MOGAD cases and MOGAD cases with MS-like presentation have probably been incorrectly labeled.

Identification of the underlying cause of myelitis often needs an extensive diagnostic workup and expertise. It may require looking for signs of diseases outside the CNS (e.g., pulmonary granulomas in sarcoidosis³) and possible triggers of myelitis (e.g., vaccinations³). Furthermore, identifying typical MRI patterns of, e.g., LETM, and bright spots on MRI T2 sequences in AQP4-NMOSD³ or H-sign and conus involvement in MOGAD³⁹ can give additional clues to an early diagnosis. In our cohort, we found that 37% of patients with MS and 39% of patients with NMOSD received the diagnosis after a relapse. Hopefully, earlier diagnosis will be possible with updated MS criteria, the newly proposed MOGAD criteria,³⁹ improved MRI techniques, and the advancement of analyzing disease-specific antibodies.¹¹

In conclusion, we report a relatively high and stable incidence rate of myelitis, where the most common initial diagnoses were idiopathic myelitis and MS. A relatively small number of patients suffered from a poor functional outcome at 6 months. A relapse occurred in around one-third of the patients, and nearly half of the initial cohort received a diagnosis of MS.

Our study highlights the importance of a thorough diagnostic workup of myelitis, including MRI of the entire CNS, CSF analysis, and testing for disease-specific antibodies. We have also identified some remaining challenges, such as the limitations of the current myelitis diagnostic criteria and the substantial group of patients receiving a diagnosis of idiopathic myelitis for whom the cause of the myelitis remains unknown.

Glossary

ADEM

acute demyelinating encephalomyelitis

ICD

International Classification of Diseases

IQR

interquartile range

IRR

incidence rate ratio

LETM

longitudinally extensive transverse myelitis

MNC

mononuclear cell

MOGAD

myelin oligodendrocyte glycoprotein antibody–associated disease

MOG-IgG

myelin oligodendrocyte glycoprotein antibody

mRS

modified Ranking Scale

MS

multiple sclerosis

NfL

neurofilament light chain

NMOSD

neuromyelitis optica spectrum disorder

NPR

National Patient Register

OCB

oligoclonal band

OR

odds ratio

PMN

polymorphonuclear cell

SAA

systemic autoimmune-associated

SLE

systemic lupus erythematosus

TBE

tick-borne encephalitis

TMCWG

Transverse Myelitis Consortium Working Group

Author Contributions

D.I. Jonsson: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. O. Sveinsson: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. N. Moeini: major role in the acquisition of data. E Pivac: major role in the acquisition of data. K. Wirdefeldt: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. L. Brundin: study concept or design; analysis or interpretation of data. E. Iacobaeus: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data.

Study Funding

The study was supported by grants from the Swedish state under the agreement between the Swedish government and the county councils (the ALF agreement). K. Wirdefeldt was supported by Region Stockholm (clinical research appointment).

Disclosure

D. Jonsson, O. Sveinsson, N. Moeini, and E. Pivac report no disclosures. K. Wirdefeldt declares no conflict of interests. L. Brundin has received travel grants and lecturing fees from Sanofi/Genzyme, Biogen, Amirall, and MedDay; participated in advisory boards for Genzyme, Sanofi, Biogen, Amirall, and Merck; and has received grants from Swedish Research Foundation, the Brain Foundation, Stockholm Council, and Karolinska Institutet. E. Iacobaeus has received honoraria for serving on advisory boards for Merck KGaA and Sanofi-Genzyme, and speaker's fee from Merck. Go to Neurology.org/NN for full disclosures.