Epidemiological findings

Thirty-seven of the 50 MOG-IgG-positive patients were female, corresponding to a sex ratio of 1:2.8 (m:f) (Fig. 1a). Median age at onset was 31 years (35.5 years in patients presenting with isolated ON [N = 32] and 28.5 years in the remainder [N = 18]; p < 0.04) with a broad range of 6 to 70 years. 3 patients were > =60 years of age at onset, and 8 patients were under 18 at first attack (including 4 ≤ 12 years) (Fig. 1b). Fourty-nine of the 50 patients (98%) were of Caucasian and 1 of Asian descent. Symptoms had started between Jul 1973 and Apr 2016. The mean observation period since disease onset was 75 ± 46.5 months (range 1-507 months). In line with the fact that many MOG-IgG-positive patients develop ON and myelitis only successively, the mean observation period was longer in patients with a history both of ON and of myelitis at last follow-up (88.6 months; N = 22) than in patients with a history of either ON but no myelitis or myelitis but not ON (64.6 months; N = 28).

Disease course

Fourty of 50 MOG-IgG-positive patients (80%) had a relapsing disease course. In the remaining 10 cases only a single attack had occurred at last follow-up. The proportion of patients with a monophasic course declined with increasing observation time (Fig. 2, upper panel). If only patients with a very long observation period (≥8 years) are considered, 93% (13/14) had a recurrent course (Fig. 2, lower panel). In line with this finding, the median observation time was shorter in the ‘monophasic’ than in the relapsing cases (26 vs. 52.5 months). The proportion of patients with a relapsing disease course did not differ significantly between female (83.8% [31/37]) and male (69.2% [9/13]) patients.

Symptoms developed acutely or subacutely in the vast majority of cases; progressive deterioration of symptoms was very rare (at least once in 3/46 or 7%) and reported only in patients with myelitis.

Clinical presentation during acute attacks

Overall, 276 clinically apparent attacks in 50 patients were documented. 205 attacks clinically affected the optic nerve, 73 the spinal cord, 20 the brainstem, 3 the cerebellum, and 9 the supratentorial brain. 44/50 (88%) patients developed at least once acute ON, 28/50 (56%) at least once acute myelitis, 12/50 (24%) at least once a brainstem attack, 2/50 (4%) acute cerebellitis, and 7/50 (14%) acute supratentorial encephalitis (Fig. 3, upper panel).

At last follow-up, 26/50 (52%) patients had developed at least two different clinical syndromes (i.e., combinations of ON, myelitis, brainstem encephalitis, cerebellitis, and/or supratentorial encephalitis), either simultaneously or successively. Of these, 22 (84.6%) had experienced attacks both of ON and of myelitis at last follow-up (corresponding to 44% [22/50] of the total cohort). Another 22 (44%) had a history of ON but not of myelitis (recurrent in 15 or 68.2%), and 6 (12%) had a history of myelitis but not ON (recurrent in 4; LETM in all) at last follow-up (Fig. 3, lower panel).

Myelitis and ON had occurred simultaneously (with and without additional brainstem or brain involvement) at least once in 9/22 (40.9%) patients with a history of both ON and myelitis at last follow-up (and in 18% or 9/50 in the total cohort).

Overall, 16/50 (32%) patients presented at least once with more than one syndrome during a single attack (more than once in 10/16). While 15 attacks of myelitis (without ON) in 11 patients were associated with clinical signs and symptoms of simultaneous brain or brainstem involvement, only 1 attack of ON (without myelitis) in 1 patient had this association. Clinically inapparent spinal cord, brain, or brainstem involvement was detected in further patients by MRI (see Brain MRI findings below and part 3 of this article series [31] for details).

Symptoms associated with acute myelitis

Symptoms present at least once during attacks of myelitis included tetraparesis in 8/29 (27.6%) patients, paraparesis in 14/29 (48.3%), hemiparesis in 2/29 (6.9%), and monoparesis in 2/29 (6.9%). Paresis was severe (BMRC grades ≤2) at least once in 6/29 (20.7%) patients. Attacks included at least once pain and dysesthesia in 19/28 (67.9%) patients and were purely sensory in 15/29 (51.7%). Sensory symptoms included also Lhermitte’s sign. Bladder and/or bowel and/or erectile dysfunction occurred at least once in 20/29 (69%) patients (Fig. 4).

Symptoms associated with acute ON

In 36/39 (92.3%) patients ON was associated with reduced high-contrast visual acuity (VA) as determined using a Snellen chart. In one patient, low-contrast but not high-contrast VA was reduced; in another patient with hazy vision but normal high-contrast VA, low-contrast VA was not tested. In a third patient, impaired color perception and papilledema were the only clinical symptoms.

Most patients with ON reported retrobulbar pain and/or pain on eye movement. Disturbed color vision including color desaturation was reported in some patients, but was not systematically examined in all patients.

Attack-related functional blindness (defined as VA ≤0.1) in one or both eyes occurred at least once in 27/39 (69.2%) patients and VA ≤0.5 was present at least once in 33/39 (84.6%) during acute ON attacks (Fig. 5). Both eyes were affected simultaneously (‘bilateral ON’) at least once in 22/43 (51.2%) patients, and scotoma was noted at least once in 23/35 (65.7%) with available data.

Other symptoms

Brainstem symptoms occurred in 12 MOG-IgG-positive patients. A detailed analysis can be found in part 3 of this article series [31]. Respiratory insufficiency due to brainstem encephalitis (2 ×) or myelitis (1 ×) occurred at least once in 3/48 (6.3%) patients with available data (median observation time 50.5 months; range 1-507) and was fatal in one of these two cases. Two patients had clinical signs and symptoms indicating cerebellar involvement. These included limb, gait, and stance ataxia with or without accompanying dysarthria. Sensory ataxia was noted in others.

Supratentorial brain lesions were symptomatic in 7 patients. These patients showed (sometimes severe) headache, fatigue, psychomotor slowing, disorientation, impaired consciousness/somnolence, hemihypesthesia, meningism, and photophobia.

Of note, several further patients had brainstem, cerebellar, and/or supratentorial brain lesions (see section Brain MRI findings below and Appendix as well as part 3 of this article series [31]) but no clinical symptoms attributable to those lesions.

Presentation at onset

ON was clearly the most common manifestation at disease onset (present in 37/50 [74%] patients), followed by myelitis (17/50 [34%]), brainstem encephalitis (4/50 [8%]) and symptoms attributable to brain (3/50 [6%]) or cerebellar lesions (1/50 [2%]). While in some patients only one site was clinically affected, multiple manifestations were noted in others: thirty-two of 50 patients (64%) initially presented with isolated ON; 9 (18%) with isolated myelitis; 5 (10%) with simultaneous ON and myelitis (additional brainstem involvement in 2); 1 (2%) with simultaneous myelitis, rhombencephalitis, and supratentorial encephalitis; 2 (4%) with myelitis and supratentorial encephalitis; and 1 (2%) with isolated brainstem encephalitis (Fig. 6). Accordingly, clinical evidence for dissemination in space (here understood as involvement of more than one of the following anatomical sites: optic nerve, spinal cord, prosencephalon, brainstem, and/or cerebellum) was present at onset in 8/50 (16%) patients (compared to 16/50 (32%) if the entire observation period is considered).

In the subgroup of patients with multiple manifestations at follow-up (including NMO and any other combinations of ON, myelitis, brainstem encephalitis, cerebellitis, and/or supratentorial encephalitis) (N = 26), disease had started with an isolated syndrome in 17 (65.4%) (isolated ON in 12 [46.2%] and isolated myelitis in 5 [36.4%]); with simultaneous ON and myelitis in 4 (15.4%); with simultaneous ON, myelitis, and brainstem encephalitis in 1 (3.8%); with simultaneous myelitis, rhombencephalitis and supratentorial encephalitis in 1 (3.8%); and with simultaneous myelitis and supratentorial encephalitis in 2 (7.7%).

In the subgroup of patients meeting Wingerchuk’s 2006 criteria at last follow-up, 3/14 (21.4%) had simultaneous ON and myelitis at onset (exclusively or in combination with brain, brainstem or cerebellar symptoms) and 3/8 (37.5%) of those with ON at disease onset, including 2 of the 3 cases with simultaneous ON and myelitis – presented with bilateral ON.

The initial attack affected both eyes in 15/37 (40.5%) of all patients with ON at onset and in 11/32 (34.4%) of all patients with isolated ON at onset; overall, 15/50 (30%) patients had bilateral ON at onset (partly in combination with other manifestations).

The first attack of myelitis was clinically characterized by tetraparesis in 5 patients and by paraparesis in 6; in 5 patients, myelitis was associated with purely sensory and/or autonomous symptoms at onset. In 2 patients, respiratory dysfunction was among the presenting symptoms.

Time to second attack

Among the MOG-IgG-positive patients with more than one documented attack and available data, the median time between the first and the second attack was just 5 months (range, 1-492; N = 38) (Fig. 7). There was no significant difference between patients with ON at onset (median of 6 months to next relapse; range 1-492) and patients with myelitis at onset (median 4 months; range 1-23). The median interval between first and second attack was slightly longer among patients with full recovery from the first attack (n = 17) than in the remaining patients (6 vs. 3.5 months; p = n.s.).

Presentation at second attack

The most common manifestation (isolated [N = 22] or in combination with other syndromes) at second attack was ON (21/23 [91.3%], which was mostly unilateral (21/23 [91.3%]; no data in one case). Other presentations at second attack included isolated myelitis (N = 12), isolated supratentorial encephalitis (N = 1), myelitis with brain or brainstem involvement (N = 2), and simultaneous ON and myelitis with brain involvement.

The initial presentation had high predictive value for the second attack: in 18 of 25 patients (72%) initially presenting with isolated ON, the second event was isolated ON again (and in 19/25 or 76% patients, ON was among the presenting manifestations); similarly, in 6/8 (75%) patients with isolated myelitis the second event was also isolated myelitis. Overall, at least one manifestation present at onset (ON, myelitis, brainstem encephalitis, cerebellitis, supratentorial encephalitis) was present also at the second attack in 31/40 (78%) patients with a recurrent disease course.

Of note, both optic nerves were affected clinically early in the disease course: in 6/10 (60%) patients with available data who experienced a unilateral ON at disease onset and ON at first relapse, the second attack affected the previously unaffected eye (or both eyes). Overall, 21/34 (62%) patients had a history of ON in both optic nerves (simultaneously or subsequently) already after the second event.

Annualized relapse rate

If all patients with an observation time of ≥12 months are considered, the median annualized relapse rate (ARR) was 0.83 (range 0.05-6.92) in the total group (n = 39) and 0.92 (range 0.05-6.92) among patients with a recurrent disease course (n = 34). It was higher among female than among male patients both in the total cohort (0.92 vs. 0.535; N = 29 and 10, respectively) and in the relapsing subgroup (0.92 vs. 0.83; N = 27 and 7, respectively), but the differences were not statistically significant.

The median ARR was highest (1.17; range 0.05-4.2; N = 19) in relapsing patients with a history of both ON and myelitis (n = 21), compared with 0.8 (range 0.5-6.92) among patients with recurrent isolated ON but no myelitis (n = 12) and 0.57 and 0.83 in the two only patients with recurrent isolated LETM but no ON and an observation time ≥12 months.

Brain MRI findings

Supratentorial MRI abnormalities were present at onset in 17/48 (35.4%) MOG-IgG-positive patients and infratentorial MRI lesions in 7/48 (14.6%). Supratentorial MRI lesions at onset included periventricular lesions; lesions in the corpus callosum (some of them confluent); frontal, parietal, temporoparietal, and occipital deep white matter lesions; subcortical or juxtacortical lesions (including insular lesions); and, in one case, lesions in the thalamus (pulvinar) and in the basal ganglia (putamen) (Fig. 8). In one patient leptomeningeal enhancement was noted at onset (Fig. 8, panel d), and in one both optic tracts were affected (Fig. 8, panel c).

Infratentorial lesions at onset included lesions in the cerebral peduncles, the pons (incluing tegmentum), medulla oblongata, cerebellar hemispheres, and cerebellar peduncles (see part 3 of this series [31] for details).

Taking not only the first but all MRIs into account, 22/47 (46.8%) patients had supratentorial brain lesions at least once; brainstem lesions occurred at least once in 14/48 (29.2%); and cerebellar lesions were noted at least once in 6/48 (12.5%) (see part 3 of this series for details [31]). Lesions affected the periventricular white matter, deep white matter (in some cases large and confluent) and corona radiata, sub- or juxtacortical white matter, corpus callosum, thalamus (pulvinar), basal ganglia, cerebral peduncles, pons (ventral, median, tegmentum), medulla oblongata (including the area postrema and the periaqueductal gray), cerebellar hemispheres, and cerebellar peduncles and were partly Gd-enhancing. Lesions were found in the frontal, parietal, temporal, and occipital lobes and in the insula. Taking the entire course of disease into account, callosal lesions were present at least once in 8/48 (16.7%) patients and periventricular lesions in 12/47 (25.5%). Callosal lesions were longitudinally extensive (more than half the length of the corpus callosum), as considered typical for AQP4-IgG-positive NMOSD [29], in 1/8 (12.5%).

Optic nerve MRI findings

MRI signs of ON were present in at least 24/44 (54.5%) patients with available data, all of whom had a history of clinical ON (Fig. 9). Intraorbital swelling of the optic nerve was noted at least once in 13/21 (61.9%) patients, and contrast enhancement in 20/21 (95.2%). A longitudinally extensive optic nerve lesion (more than half the length of the nerve) (n = 6) and/or involvement of the optic chiasm (n = 4), two findings previously considered typical for AQP4-IgG-positive NMO [29], were present during acute ON in 8/26 (30.8%) cases with available data. Signs of optic nerve atrophy were noted in at least 5 patients and involved the optic chiasm in at least one of them. However, post-chiasmatic parts of the optic pathway were also affected in individual patients: as mentioned above, one patient had optic tract lesions, and occipital lobe white matter lesions were documented in four cases.

Of particular note, in 11/28 (39.3%) patients with available data, perioptic contrast enhancement, i.e. gadolinium enhancement within the nerve sheath and the immediately surrounding orbital tissues, was present during acute ON (Fig. 9). The remaining patients had either no history of ON or no or no suitable post-contrast orbital MRI was performed or retrospectively available for re-analysis and the presence of absence of perioptic enhancement was not mentioned in their MRI reports.

Spinal cord MRI findings

MRI signs of spinal cord inflammation were present in 29/44 (65.9%) patients with available data, including 27/28 (96.4%) with a history of clinical myelitis (Fig. 10).

Spinal MRI was performed also in 16 patients without a history of clinically apparent myelitis and showed a spinal cord lesion extending over 2 segments in 2 of them.

In 20 out of the 28 (71.4%) patients with a history of clinical myelitis and available data, two or more lesions were present simultaneously (i.e., in the same MRI) at least once.

Spinal cord lesions on MRI extending over three or more vertebral segments (VS), i.e., so-called LETM lesions, were documented in 21/29 (72.4%) patients at least once. LETM lesions were present during the first attack in 11/17 (64.7%) patients initially presenting with acute myelits.

By contrast, in 8 patients exclusively short lesions (<3 VS), i.e., so-called non-longitudinally extensive transverse myelitis (NETM) lesions, were documented over the entire observation period. Of potential differential diagnostic importance, spinal cord MRI showed one or more NETM lesions but no LETM lesions at disease onset in 6/17 (35.3%) patients initially presenting with acute myelitis (alone or in combination with other syndromes). If all available MRIs are considered, MRI lesions extended over fewer than three segments during acute attacks of myelitis in 12/27 (44.4%) patients.

The median length of all documented LETM lesions (n = 32) was 4 VS (range 3-20) and that of all documented NETM lesions (n = 44) was 1.5 VS (range 1-2). If all spinal cord lesions with available data are considered (n = 76), i.e., both LETM and NETM lesions (including NETM lesions present in addition to LETM lesions in the same MRI), the median longitudinal extension was 2 VS (range 1-20). Finally, the median length of the longest spinal cord lesion (LETM or NETM) ever observed in each patient was 5 VS (range, 1-20; N = 27) if all patients with available MRI data were considered and 5 (range, 1-20; N = 26) if only patients with clinical evidence for myelitis were considered.

Swelling of the spinal cord was noted at least once in 19/27 (70.4%) patients and contrast enhancement in 19/28 (67.9%). Signs of necrosis of the spinal cord were noted in 0/23 (0%) patients with available data.

Spinal cord lesions were located in the cervical spinal cord at least once in 23/28 (82.1%) patients and in the thoracic spinal cord at least once in 21/28 (75%). Lumbar and conus lesions were documented only in 3/27 (11.1%) and 3/27 (11.1%) patients, respectively. Taking all available spinal cord MRIs into account, cervical lesions were present in 44/81 (54.3%) MRIs, thoracic lesions in 31/81 (38.3%), lumbar lesions in 4/81 (4.9%), and the conus was affected in 3/81 (3.7%). However, as a limitation, not all MRIs performed showed the entire spinal cord, and spinal cord MRI data were absent for 18 myelitis attacks in 8 patients.

Information on intramedullary lesion location was available for 34 lesions in 20 MOG-IgG-positive patients. Lesions were located predominantly in the central portion of the spinal cord in 17 MRIs and predominantly in the peripheral portion in another 17 MRIs.

The spinal cord MRI was normal during 2 attacks; in both cases, symptoms were purely sensory (paresthesia and hyp- and dysesthesia, respectively). Of note, a total of five asymptomatic spinal cord lesions were noted in two patients (in addition to brainstem lesions in one) with a history of ON but no clinical evidence of myelitis over the course of disease.

Evaluation of Barkhof’s and Paty’s MRI criteria for MS

Seven of 46 (15.2%) MOG-IgG-positive patients with a history of myelitis and/or ON and 7/26 (26.9%) of those with brain lesions met Barkhof’s MRI criteria for MS at least once [38]. However, at least 2 of the 7 patients meeting Barkhof’s criteria also had one or more NMOSD-typical lesions at least once.

The revised 2006 diagnostic criteria for NMO [28] required a brain MRI at disease onset that does not meet Paty’s MRI criteria for MS [39] if either no LETM lesion is present or NMO-IgG is negative. Accordingly, Paty’s criteria were evaluated only at disease onset. In the present cohort, the initial MRI of 12 out of 48 (25%) MOG-IgG-positive patients with available data met Paty’s criteria.

Intrathecal IgG synthesis

Data on CSF-restricted oligoclonal IgG bands (OCB) were available from 45/50 (90%) MOG-IgG-positive patients. Pattern 2 or 3 OCB [40] indicative of intrathecal IgG synthesis were positive at least once only in 6/45 (13.3%). A second lumbar puncture was performed in 2 out of the 6 OCB-positive patients, in both of whom OCB remained positive.

Patients with classical MS display a polyspecific, intrathecal humoral immune response to neurotropic viruses such as measles, rubella, and varicella zoster virus (the so-called MRZ reaction, MRZR) [41–44]. MRZR was tested in 11 MOG-IgG-positive patients (2 x ON + myelitis; 1 x ON + myelitis + brainstem encephalitis; 1 x myelitis + brainstem encephalitis; 3 x LETM; 5 x ON) and was negative in all of them.

CSF white cell counts

White cell counts (WCC) in the CSF were documented at least once in 46 MOG-IgG-positive patients and were elevated (>5/μl) in 32 (69.6%). In those patients with pleocytosis, WCC ranged between 6 and 306 cells/μl (median 33; quartile range 13-125). WCC ≥100 cell/μl were present at least once in 9/32 (28.1%) patients. Neutrophil granulocytes were present at least once in 9/14 (64.3%) patients with pleocytosis and available data (median 22% of all white cells; range 3-69%).

Blood-CSF barrier function

An increased albumin CSF/serum ratio (QAlb) reflects a disturbed blood-CSF barrier (BCSFB) function caused by structural damage and/or a reduced CSF flow rate [45]. QAlb was determined in 37 MOG-IgG-positive patients and was elevated in 12 (32.4%). Blood-CSF barrier dysfunction was present both among patients with a history of isolated ON (2/15; 13.3%) and, more frequently, in patients with a history of spinal cord and/or brain/brainstem involvement (10/21; 47.6%).

Visual evoked potentials

Data on visual evoked potentials (VEP) were available from 47 MOG-IgG-positive patients. A delayed P100 latency was noted at least once in 34 (72.3%); in another 6 (12.8%) patients latencies could not be determined since potentials were lost due to severe optic nerve damage.

Only 41 (78.7%) of the 47 patients examined had a history of clinically manifest ON; in 31 of these 41 patients (75.6%) P100 latency was delayed, and in 6 further patients (14.6%) latencies could not be determined. The remaining 6 patients had a history of myelitis (LETM in all cases) but no history of clinically manifest ON. 3 of those 6 had delayed P100 latencies in at least one eye, indicating that subclinical optic nerve damage might be relatively frequent in MOG-IgG-positive patients with myelitis.

In 23/41 (56.1%) patients, all of whom had a history of clinical ON, VEP amplitudes were reduced (n = 16) or lost (n = 7) at least once. In all but one patient with reduced amplitudes, P100 latencies were also delayed at some point in time, but not vice versa.

Somatosensory evoked potentials

Data on somatosensory evoked potentials (SSEP) were available from 39 MOG-IgG-positive patients, including 24 with a history of clinically manifest myelitis. SSEP were delayed, reduced in amplitude, or lost in 19/39 (46.2%), including in 16/24 (66.7%) with a history of clinical myelitis and available data. Of note, 3 patients with no clinical history of myelitis had SSEP abnormalities suggestive of subclinical spinal cord damage (none of them displayed unequivocal spinal cord MRI abnormalities).

Ophthalmoscopic findings

Fundoscopy revealed uni- or bilateral papillitis or papilledema in at least 15 patients with acute ON, suggesting inflammation of the anterior part of the optic nerve. The true prevalence of papillitis could be higher, however, since ophthalmoscopic data were not available from all patients. In case 6 (see Appendix), papilledema was described as marked (3 dpt) at first ON and as mild at second and third ON, while later on the optic disk was described as atrophic and pale. Optic atrophy as detected by fundoscopy was noted at last follow-up in 13/22 (59.1%) patients with available data.

Evaluation of the 2010 McDonald criteria for MS

Evaluation of the 2006 criteria for NMO

63.6% (14/22) of all MOG-IgG-positive patients with a history of both ON and myelitis met Wingerchuk’s 2006 revised diagnostic criteria for NMO [28] (Table 1). Of the 8 patients with ON and myelitis who did not meet Wingerchuk’s 2006 criteria, two had an LETM lesion but the first brain MRI met Paty’s criteria for MS; five did not meet Paty’s criteria at onset but spinal cord lesions extended over fewer than three vertebral segments; and one met Paty’s criteria at onset and had no LETM lesion.

Twenty eight patients had a history of ON but not myelitis or a history of myelitis but not ON (both with and without brain involvement) and did therefore not meet the 2006 diagnostic criteria. Taking the total cohort into account, 28% (14/50) of all patients met the 2006 criteria for NMO. Seven out of 43 (16%) patients with available data fulfilled both the clinicoradiological 2006 criteria for NMO [28] and the clinicoradiological 2010 McDonald criteria for MS [46].

Evaluation of the 2015 criteria for NMOSD

On the understanding that MOG-IgG seropositivity does not per se constitute an “alternative diagnosis”, i.e., based solely on clinical and radiological criteria, 16/50 (32%) patients met the 2015 international consensus criteria for NMOSD [29] (Table 1). Of those, 15 had a history of both ON and myelitis and 1 a history of ON but not of myelitis (this patient fulfilled the criteria despite the lack of myelitis due to the presence of brainstem encephalitis with periependymal lesions around the fourth ventricle and of symptomatic, extensive white matter lesions); none had a history of myelitis but not of ON. Of those patients who met the 2006 criteria, 12 (85.7%) also met the 2015 criteria. Conversely, 12 (75%) of those who met the 2015 criteria also met the 2006 criteria. 8 out of 43 (19%) patients with available data fulfilled both the clinicoradiological 2015 criteria for NMOSD and the clinicoradiological 2010 McDonald criteria for MS. If only patients with a relapsing course of disease are considered, 16/40 (40%) met Wingerchuk’s 2015 criteria.

Previous diagnoses

As reliable tests for MOG-IgG became available only relatively recently, most of the patients initially received diagnoses other than MOG-IgG-positive encephalomyelitis (EM). In 16/45 (35.6%) patients with available data, a diagnosis of MS was suspected at least once. Other suspected diagnoses included acute disseminated EM (ADEM), multiphasic disseminated EM, AQP4-IgG-negative NMO according to Wingerchuk’s 2006 criteria [28], AQP4-IgG-negative NMOSD according to the 2015 international diagnostic consensus criteria [29], viral encephalitis, bacterial encephalitis, paraneoplastic encephalitis, isolated vasculitis of the CNS, chronic relapsing inflammatory optic neuropathy (CRION), CNS lymphoma, sarcoidosis, spinal stenosis, “spinal tumor of unknown dignity”, suspected spinal ischemia, para- or postinfectious ON, and myelitis; some patients were diagnosed with ON, rON, (longitudinally extensive transverse) myelitis, brainstem encephalitis or EM “of unknown origin”.

Coexisting autoimmunity

Coexisting autoantibodies were present in 19/45 (42.2%) MOG-IgG-positive patients. These included antinuclear antibodies (ANA) in at least 14 patients and cardiolipin antibodies or phospholipid/glycoprotein beta-2 antibodies (2 ×), anti-tissue transglutaminase IgA (1 ×), rheumatoid factor (1 ×), anti-thyroid peroxidase (2 ×), anti-thyreoglobulin (1 ×), anti-thyroid-simulating hormone receptor (1 ×), perinuclear anti-neutrophil cytoplasmic antibodies (ANCA) (1×). None of the 50 patients was positive for AQP4-IgG [30].

Concomitant autoimmune disorders were present only in 4/47 (8.5%) patients and included rheumatoid arthritis (RA) (2 ×), Hashimoto thyroiditis (1 ×), Grave’s disease (1 ×). A further patients had atopic dermatitis and asthma bronchiale.

Preceding infections

Disease onset was preceded by infection in at least 11 patients. Diagnoses included common cold, sore throat, tonsillitis, sinusitis, bronchitis, “respiratory infection”, “feverish infection”, and, in one case, a gastrointestinal infection with positive Yersinia serology (species not determined).

Taking not only the first but all attacks into account, attacks were preceded by infection at least once in at least 15/37 (40.5%) patients; the infections included, in addition to those already mentioned above, “mycoplasma pneumonia,” one case each of a non-specified “respiratory” or “bronchopulmonary” infection, a “feverish common cold”, “fever and fatigue”, and a non-specified “feverish infection”. In at least one patient, both the first and the second attack were preceded by infection.

One further patient reported a history of two episodes of “borreliosis with meningitis” 20 and 19 years before onset.

Preceding vaccinations

Disease onset was preceded by revaccination against diphtheria, tetanus, pertussis, polio, and influenza 2 weeks prior to symptom onset in one patient (for details of this case see part 3 of this series [31]), and by vaccination against diphtheria, tetanus, and pertussis 13 days prior to symptom onset in a second case; the latter patient developed fever 2 - 3 days before symptoms started. Both patients (1 × male, 1 × female) were vaccinated at adult age (19 and 47 years) and both developed recurrent disease. While the first patient experienced seven relapses involving the the optic nerves (4 ×), spinal cord (5 ×), brain (2 ×), and brainstem (1 ×) within 20 months, which fully responded to IVMP or combined IVMP and plasma exchange (PEX), the second patient developed three attacks (2 × ON, 1 × myelitis and ON) within 6 months, which only partially responded to IVMP, PEX and IA and resulted in an EDSS of 8 at discharge; two of the attacks occurred despite treatment with rituximab.

Pregnancy-associated attacks

Seventeen percent (5/30) of all female patients aged ≥15 years at last follow-up experienced at least one attack of ON or myelitis during pregnancy or post partum. This corresponded to 50% (5/10) of all patients with a documented pregnancy (no data in 9) and, importantly, included all 5/5 women of reproductive age with available data who were pregnant shortly before (i.e., within the last 18 months), at, or after disease onset. Of a total of seven attacks, three had occurred during pregnancy and four post partum. These included the first attack ever in 3 patients: Disease started with simultaneous ON and LETM and accompanying brainstem and brain lesions occurring just 6 weeks after the delivery of the first child in one case; with an attack of unilateral ON 3 months post partum and during breast-feeding in a second patient; and with an attack of bilateral ON 8 months after delivery and while still breast-feeding in a third (as a limitation, however, ON was also preceded by a common cold with mild fever in this last case). In a fourth patient, an attack of LETM occurred during week 6 of pregnancy and an attack of bilateral ON a few weeks after delivery; however, the disease had started 8 years earlier in this patient and several ON attacks had occurred in the meanwhile. A fifth patient experienced at least two attacks of ON during pregnancy, which responded well to IVMP; disease had started 2 years before. While 3 patients had a relapsing course, 2 have not developed further attacks so far, although the follow-up time is short (6 and 3 months, respectively). Overall, 7/23 attacks in the 3 relapsing patients were associated with pregnancy or delivery, while the majority of attacks were not.

Tumor associations

In a single patient presenting with post-infectious whole-spine myelitis and severe brainstem and brain inflammation, a mature cystic ovarian teratoma had been removed 2 months before onset of the neurological symptoms, but no signs of malignancy had been found; NMDAR antibodies were negative. In the same patient, a ganglioneuroma was found and resected at a later date. MOG-IgG were not associated with malign tumors also in all other patients studied.

Treatments for acute attacks

Acute attacks were treated with high-dose IVMP at least once in 47/48 (97.9%) MOG-IgG-positive patients, with PEX at least once in 19/48 (39.6%), and with immunoadsorption (IA) in two. Other treatments included oral steroids or dexamethasone i.v. followed by oral steroids in single patients as well as acyclovir and/or antibiotics for pragmatic treatment of initially suspected CNS infection.

Overall, 136 documented attacks were treated with IVMP, 15 with PEX, and 25 with both IVMP and PEX or – in five of them - IA; 18 were not treated at all. PEX or IA were used to treat 20 ON attacks, 16 myelitis attacks (with or without brain and/or brainstem and/or cerebellum involvement), 3 attack of simultaneous ON and myelitis (with or without additional clinical brain involvement), and 1 pure brainstem attack.

Overall outcome of acute attacks

Outcome data were available for 134 ON attacks in 39 MOG-IgG-positive patients and for 46 myelitis attacks in 23 MOG-IgG-positive patients. Complete or almost complete recovery from acute ON was noted after 70 (52.2%) ON attacks, partial recovery after 54 (40.3%), and no or almost no recovery after 10 (Fig. 11b). Complete or almost complete recovery from acute myelitis was noted in 16 (34.8%) attacks, partial recovery in 30 (65.2%), and no or almost no recovery in none (Fig. 11a).

At last follow-up, 38/48 (79.2%) patients had experienced complete or almost complete recovery from at least one attack. In contrast, 22/48 (45.8%) had experienced at least one attack that was followed by no or almost no recovery. While 62.2% (28/45) of the patients’ initial attacks remitted completely or almost completely, the proportion was lower for all subsequent attacks (40.6% or 69/170) and dropped to 26.4% or 19/72 after the fifth relapse.

Outcome of attacks treated with IVMP

Outcome data were available for 122 attacks treated with IVMP but not PEX (including attacks of ON; myelitis; brainstem encephalitis; cerebellitis; supratentorial encephalitis; simultaneous ON and myelitis; simultaneous ON, myelitis, and brainstem encephalitis; simultaneous ON, myelitis, and supratentorial encephalitis; simultaneous myelitis and brainstem encephalitis; simultaneous myelitis and supratentorial encephalitis; and simultaneous myelitis, brainstem, and brain inflammation). In 61 (50%) of those relapses, IVMP treatment was followed by complete or almost complete recovery, in 54 (44.3%) by partial recovery, and in 7 (5.7%) by no or almost no recovery.

Of particular note, symptoms flared up after withdrawal or tapering of steroids at least once in 21/47 (44.7%) patients (see Appendix and Discussion for details). To control symptoms, IVMP was combined with or escalated to PEX or IA in 17/48 (35.4%) patients at least once and for 9.1% (25/276) of all documented attacks. If those attacks that were subsequently treated with PEX are also taken into account and on the understanding that the use of PEX after IVMP implies partial or full IVMP failure, 86/147 (58.5%) attacks initially treated with IVMP responded only partially or not at all to IVMP, while IVMP was followed by complete or almost complete recovery in 41.5%.

Outcome of attacks treated with PEX or IA

Outcome data were available for 40 attacks treated either with PEX/IA alone or with both IVMP and PEX/IA; IA instead of PEX was used to treat five of those attacks.

Stand-alone PEX/IA was used for treating attacks (N = 15) of ON and/or myelitis with and without brain or brainstem involvement and attacks of isolated brainstem encephalitis. The median number of PEX/IA cycles used per attack was 5 (range, 3-11). In 3 (20%) of those 15 attacks, PEX treatment was followed by complete or almost complete recovery, in 11 (73.3%) by partial recovery, and in 1 by no or almost no recovery.

In addition, 25 attacks of ON and/or myelitis (with and without brain and/or brainstem involvement) were treated with both IVMP and, subsequently, PEX/IA. In 10 (40%) of these attacks, PEX/IA treatment was followed by complete or almost complete recovery, in 14 (56%) by partial recovery, and in 1 by no or almost no recovery.

If all attacks treated with PEX/IA (with or without IVMP) are considered, PEX/IA treatment was followed by complete or almost complete recovery in 13 (32.5%) attacks, by partial recovery in 25 (62.5%), and in 2 (5%) by no or almost no recovery.

IA was used instead of PEX for two attacks of ON in case 11. While treatment with four courses of IA was followed by almost complete recovery from an ON attack that had responded only transiently to a first IVMP cycle (and not at all to a second one) and by a relapse-free period of 3 years, the next ON attack responded only partially to IVMP and four courses of IA. The reason for the differential response to IA during those two relapses is unknown, but, as with PEX, differences in antibody titers as well as timing issues might have played a role. In case 28, IA resulted only in partial recovery when used after IVMP to treat three attacks of isolated myelitis, simultaneous ON and myelitis, and of isolated ON, respectively.

Outcome of untreated attacks

Only 14 attacks in 11 patients were not treated with steroids or PEX/IA. Among those attacks, no or almost no recovery was noted in 2 cases (acute ON and brainstem encephalitis in one patient, ON in a second), one of which was fatal, partial recovery in 3 (acute ON in all), and full or almost full recovery in 9 (acute ON in 7, acute encephalitis/brainstem encephalitis in 2 ). The reasons for not treating patients for acute attacks were not specified in all cases. IVMP treatment was declined by at least two patients once each (no recovery in one and full recovery in the other one), and a decision in favor of palliative care had been previously made in another patient; in at least one case, ON was considered mild and therefore left untreated.

Long-term treatments

Long-term immunosuppressive (IS) or immunomodulatory (IM) treatments were used at least once in 35/49 (71.4%) patients and included azathioprine (AZA) in 18, methotrexate (MTX) in 8, rituximab in 16, glatiramer acetate (GLAT) in 5, interferon-beta (IFN-beta) in 4, natalizumab (NAT) in 3, ofatumumab in 1, intravenous immunoglobulins (IVIG) in 1, mitoxantrone in 2, ciclosporin in 1, mycophenolate mofetil in 1, and oral steroids in 5; 14 patients (including 8 with a so far monophasic disease course) never received any IS/IM treatment.

Breakthrough attacks were noted in 21/31 (67.7%) patients treated with IS/IM at least once.

Response to AZA treatment

Data on acute attacks during AZA therapy were available from 17/18 patients treated. The median treatment period was 10 months (range 2-101). Of these 17 patients, 14 (82.4%) experienced at least one attack under treatment with AZA. In total, 34 attacks occurred under AZA over a cumulative treatment period of 412 months (cumulative ARR 0.99) with a median of 1 attack/patient (range 0-6) in the total AZA group and of 1.5 attacks/patients (range 1-6) in those who had breakthrough relapses.

Of particular note, 14 of the 34 attacks (41%) took place during the first 6 months, i.e., during the drug-specific latency period. Of these, 11 attacks developed during the first 3 months and only 3 during months 4-6. If all patients are taken into account, the median of all individual ARRs was 2 during the 6-month AZA latency period and 0.92 after the latency period.

Cotreatment with oral steroids or, in a single case, regular PEX was administered only in 9/23 (39.1%) patients (no data in 2), either for 3 or for 6 months or for the entire treatment period. Importantly, most attacks (12/14) observed during the AZA latency period occurred in patients who were not cotreated. Relapses occurred in only 1 of 14 cotreated patients during the latency period, but in 6/9 patients who were not cotreated. Similarly, 4/5 patients who developed relapses after the latency period were not cotreated, and 14/17 attacks occurring during that period affected non-cotreated patients. Taking the total treatment period into account, 10/12 patients with relapses under AZA were not cotreated at the time of the attack and 26/31 attacks occurred in non-cotreated patients.

Response to MTX treatment

Data on acute attacks before and during MTX therapy were available from six patients. In case 13 (see Appendix for case reports), a single (though severe and non-remitting) relapse occurred under MTX within 134 months compared with 3 attacks in an 11-month period including 9 months of combined treatment with AZA and oral steroids. As a possible limitation, it remains unclear whether further attacks in the affected right eye went unrecognized due to the pre-existing severe visual deficit. Patient 3 experienced two attacks (both with complete recovery) within a period of 5.5 years of MTX treatment. Of note, however, this included the patient’s first attack ever, which occurred under active MTX treatment for pre-existing RA. MTX was used as treatment for RA also in patient 6 described in part 3 of this article series [31]; in that patient, temporary discontinuation of MTX after 5 years due to severe infection was followed by the first relapse for 40 years. MTX was continued and no further attack occurred over the following 12 months. Similarly, patient 12 in part 3 of this series [31] suffered no attacks during 21 months of MTX treatment, although, three attacks had occurred within 7 months prior to commencement of MTX. Finally, combined treatment with MTX and oral steroids (plus ciclosporin A during the initial 7 months) resulted in disease stabilization in case 6, with only two relapses (with only partial recovery though) in almost 7 years; by contrast, 14 attacks had occurred in the preceding 5 years in this patient (including during treatment with IFN-beta, GLAT, AZA, or rituximab).

Overall, 5 attacks took place in 22.5 years in these patients under treatment with MTX. This corresponds to a cumulative ARR of 0.22, which is lower than the cumulative ARR of 0.95 found among all patients (n = 34) with a relapsing disease course. Patient 1, in whom three breakthrough attacks occurred within 8 months of MTX therapy, was the only patient with apparent MTX failure.

Response to IFN-beta treatment

No decrease in relapse rate was observed under treatment with various IFN-beta preparations, which were given for suspected MS. In case 6, commencement of therapy with i.m. IFN-beta-1a (Avonex®) was followed by two ON relapses 1 and 4 months later. Similarly, s.c. IFN-beta-1a (Rebif®) was followed by an ON relapse less than 2 months after treatment was started. Finally, treatment with s.c. IFN-beta-1b (Betaferon®) was associated with another ON relapse after 2 months. Overall, four relapses occurred within around 16 months of IFN-beta treatment (ARR 3.0). This is in strong contrast to just two ON relapses within 71 months under therapy with MTX and oral steroids in that patient (ARR 0.33). Of interest, both IFN-beta-1a and -1b led to leukopenia. In another patient (see case 5 in part 3 of this article series [31] for details), further relapses occurred and marked disease exacerbation on MRI was noted after the initiation of i.m. IFN-beta-1a treatment, with new spinal and brainstem lesions. In a third patient (case 12), two relapses occurred within 11 months and led to discontinuation of s.c. IFN-beta-1a therapy. A fourth patient experienced an attack of mild ON and myelitis after 8 months of IFN-beta (Rebif®) therapy. When the same patient was again treated with IFN-beta 5 years later (now with Avonex®), an attack of severe unilateral ON occurred two months after treatment initiation and an attack of ON in the opposite eye with simultaneous myelitis after a further 2 months. In total, she experienced three attacks during a total IFN-beta treatment period of 19 months.

Response to GLAT treatment

Five patients were treated with GLAT for suspected MS. In case 6, no relapse occurred over a period of 6 months; by contrast, four relapses had occurred under IFN-beta over a period of 16 months in the same patient. However, considering the GLAT-specific latency period of 3-6 months observed in MS it remains uncertain whether that decline in relapse rate was due to GLAT treatment or to discontinuation of (potentially disease-exacerbating) IFN-beta treatment. GLAT treatment had to be stopped due to leukopenia in that patient. In case 8, no relapses occurred over a period of 36 months on therapy with GLAT and remission of spinal cord lesions was detected by MRI. However, this patient had previously experienced a relapse-free interval of more than 5 years, rendering it uncertain also in this case whether GLAT was effective. A third patient (see case 1 in part 3 of this article series [31] for details) was relapse-free for almost a year under GLAT, but experienced two relapses (1 x ON, 1 x myelitis) 11 and 13 months after initiation of therapy, leading to discontinuation of GLAT. Previously, one to two relapses per year had occurred over a period of around 6 years, and three relapses within the last 10 months prior to GLAT. A fourth patient (case 14 in part 3 [31]) experienced three ON attacks during 8 months of GLAT treatment; moreover, a further relapse of severe ON leading to transient unilateral blindness occurred a few weeks after GLAT therapy was discontinued. In a further patient (case 13 in part 3 [31]), two attacks occurred during 7 months of treatment with GLAT (3 and 7 months after the first injection). When treated a second time with GLAT more than 3 years later, she experienced a protracted attack of myelitis with paresis, impaired coordination, and impaired ambulation 1 months after commencement of therapy (and thus during the drug’s latency period), which lasted over 2 months and required a total of three cycles of high-dose IVMP therapy.

Response to NAT treatment

Three patients were treated with NAT for suspected MS. In one of them (see case 1 in [31]), two infusions of NAT were followed by three relapses by 2, 3 and 5 months, which only partially responded to PEX. Treatment with NAT was not continued after the second infusion due to recurrent headache. In the second patient (see case 5 in part 3 [31]) an attack of brainstem encephalitis occurred and MRI showed a new LETM lesion 9 months after commencement of NAT therapy. The third patient (case 13 in part 3 [31]) experienced two myelitis attacks 1 and 4 months after initiation of NAT treatment, followed by a relapse-free interval of 21 months. When NAT was re-initiated 11 months later, she developed two further attacks of myelitis after 4 and 5 months, followed by a relapse-free interval of 9 months; treatment was discontinued due to John Cunningham virus (JCV) seroconversion. In total, four attacks occurred during 29 months of NAT treatment.

Response to rituximab and ofatumumab

Of 16 patients treated with rituximab at least once, observation periods under rituximab therapy were sufficiently long to allow meaningful analyses of the drug’s efficacy only in 9 patients.

Treatment with rituximab was followed by a decline in relapse rate in 3/9: In one patient (see case 18 in the Appendix), no relapse occurred in 12 months under rituximab compared with four relapses of ON within 6 months beforehand. In another patient (see case 7 in part 3 of this series [31]) one minor relapse with spontaneous remission took place in 28 months, compared with three attacks within the previous 4 months). Finally, in case 12, no relapses occurred during 8 months of rituximab treatment compared with three relapses in the preceding 14 months (two of which, however, took place under treatment with IFN-beta, which was reported to cause disease exacerbation in NMO and which was associated with ongoing or increasing disease activity also in our patients).

Of note, in the other six patients one or more attacks were noted during therapy with rituximab, most of which occurred shortly after rituximab infusion. This is reminiscent of early attacks observed in AQP4-IgG-positive NMO patients treated with rituximab. Two relapses of ON occurred 3 and 7 weeks after the first rituximab infusion (2 × 1000 mg i.v., days 1 and 15) in case 6 (see Appendix). Similarly, patient 1 in part 3 of this series [31] developed severe clinical and radiological deterioration 4 weeks after the first and 2 weeks after the second infusion of rituximab. The latter patient had been treated with PEX 1 month before rituximab was started, indicating that even pretreatment with PEX may not be sufficient in all cases to prevent the risk of rituximab-related attacks. A further patient developed two relapses of ON one months after the first and 2 months after the second infusion, respectively. The fourth patient (case 11 in part 3 [31]) developed severe bilateral ON three months after the second infusion (i.e. four months after the first infusion) of rituximab. A fifth patient developed two attacks of myelitis and of ON 2 months after the first and three months after the second infusion. Finally, one patient who was treated with rituximab for a first attack of myelitis, developed ON just five months after the first infusion of 1000 mg rituximab. By contrast, no early relapses were noted in ten cases.

Of note, two end-of-dose relapses in rituximab-treated patients were documented. One patient (see case 7 in part 3 [31]) relapsed immediately after reappearance of B cells 9 months after the first infusion. Similarly, a relapse occurred in case 6 12 months after the first rituximab infusion. By contrast, CD19 cells were still undetectable and no new relapse has occurred 14 months after onset in case 12.

In one patient (case 13 in part 3 [31]), therapy with rituximab had to be discontinued due to an allergic exanthema.

A single patient was treated with ofatumumab (18 months, four cycles to date). While eight attacks of ON and three attacks of myelitis (one with accompanying brainstem encephalitis) had taken place over a period of 63 months under various previous therapies (ARR 2.1), only a single attack of ON occurred during 18 months (ARR 0.66) of ofatumumab treatment in this patient.

Response to mitoxantrone and other rare therapies

In the only patient with available data (case 1 in part 3 of this article series [31]), three infusions of mitoxantrone (1 × 12 mg/m² and 2 × 8 mg/m²) did not prevent three relapses of myelitis and two of ON within around 5 months, with some of the relapses occurring just a few weeks after infusion. A further patient (case 13 in part 3 [31]) experienced a relapse of sensory myelitis 1 month after initiation of fingolimod. Discontinuation of fingolimod after 3 months due to lymphopenia was immediately followed by a relapse of myelitis with impaired ambulation, paresthesia and dysesthesia below T5, and two flare-ups over the next 2 months, requiring a total of three cycles of (escalating) high dose IVMP therapy.

Ciclosporin was used in combination with MTX and oral steroids in a single patient (see case 6 in the Appendix) for a period of 6 months; no relapses occurred under this regimen. One patient (see case 13 in part 3 [31]) was treated for 4 months with dimethylfumarate. While no relapses occurred during that period, treatment had to be discontinued due to reflux, pharyngitis and laryngitis.

Another patient (case 2 in part 3 [31]) was treated with IVIG over 11 months (and tapering of oral steroids during the initial 3 months). No new relapses occurred during that period and IVIG treatment was temporally associated with clinical improvement and resolution of MRI lesions; the patient was still relapse-free 12 months after discontinuation of IVIG.

Long-term outcome

At last follow-up, VA was impaired in at least one eye in 21/38 (55.3%) patients with a history of clinical ON (median observation time 53.5 months, range 1-507) and around one third (14/38; 36.8%) of all patients were either functionally blind at last follow-up in one eye or both or had a severe visual impairment (VA >0.1 and ≤0.5). Functional blindness (VA ≤0.1) in at least one eye was noted in 10/38 (26.3%) patients, severe visual impairment but no functional blindness (VA >0.1 and ≤0.5) in 4/38 (10.5%), moderate impairment (VA >0.5 and ≤0.75) in 2/38 (5.3%), and mild impairment (VA >0.75 and <1.0) in 5/38 (13.2%). Both optic nerves had been affected at least once at last follow-up (median observation time 54 months, range 1-394), either clinically or subclinically (i.e., based on MRI, VEP, fundoscopy, and/or OCT findings only) and either simultaneously or successively, in 35/42 (83.3%) patients, while only one optic nerve had been affected in the remainder (median observation time 29.5 months, range 5-507).

Severe paresis was present at last follow-up in 1/28 (3.6%) patients with a history of clinical myelitis (median observation time 41.5 months, range 6-102), moderate paresis in 4/28 (14.3%), and mild paresis in 6/28 (21.4%). Ambulation was impaired at last follow-up due to paresis and/or gait ataxia in 25%.

If the total cohort is considered, VA was reduced at last follow-up in 23/47 (48.9%) patients with available data (median observation time 49 months, range 1-507) and paresis was present in 14/48 (29.2%) (median observation time 50.5 months, range 1-507).

EDSS at last follow-up

The expanded disability status scale (EDSS) was developed for use in classical MS and strongly focuses on ambulation deficits [47]. When interpreting EDSS results, it should be taken into consideration that complete bilateral visual loss corresponds to an EDSS score of just 4 and that patients with isolated ON can reach no higher scores. In accordance with that well-recognized underrepresentation of visual deficits – the main long-term sequelae in our patients (with functional blindness or severe visual loss present in 36%) – EDSS scores were nominally low at last follow-up in most cases (median 2.5 [range 0-10] in the total cohort, N = 47, and 3 [range 0-10] among patients with relapsing disease, N = 40). A median EDSS ≥3.5 was reached after more than 60 months (Fig. 12). The median EDSS was 3 (range 1-10) among patients with an observation period of ≥100 months (n = 12) and 3.25 [range 1.5-10] among patients with an observation period of ≥120 months (n = 8). A higher median EDSS at last follow-up was noted in women (3, range 0 -10; n = 35) than in men (1, range 0-6; n = 12; p < 0.05) despite a longer median observation period in the male subgroup (72 months [range 1-127] vs. 50.5 months [range 1-507]).

Survival rate

After a median follow-up period of 52 months (range 1-507), 49/50 (98%) patients were still alive. One patient died from severe brainstem encephalitis leading to respiratory insufficiency 123 months after disease onset and after a total of 27 attacks, including attacks of ON, myelitis, encephalitis and/or brainstem encephalitis (see case 1 in part 3 of this article series [31]).

Substantial phenotypic overlap with AQP4-IgG-positive NMOSD and MS

MOG-IgG-positive myelitis and ON showed a significant overlap with AQP4-IgG-positive NMOSD in clinical and radiological presentation, with more than 60% patients with a history of both ON and myelitis meeting Wingerchuk’s 2006 criteria for NMO [28] and around a third of all patients fulfilling the revised 2015 criteria [29]. Even manifestations considered relatively typical for AQP4-IgG-positive NMOSD, such as medulla oblongata lesions and intractable nausea and vomiting or ON with involvement of the optic chiasm, were noted in some cases. Moreover, MS was initially suspected in more than a third of all patients, and every fourth patient with MOG-IgG-positive ON and/or myelitis presenting with brain and/or brainstem lesions met Barkhof’s criteria for MS, demonstrating a substantial phenotypic overlap between these two conditions.

With the discovery of AQP4-IgG [1, 6, 48], MOG-IgG [10], N-methyl-D-aspartate receptor-IgG [49], and a plethora of often non-paraneoplastic autoantibodies identified in acute CNS inflammation over the past decade [50–54], including in patients with primary or secondary demyelination, it becomes increasingly clear that not all patients presenting with relapsing CNS disease of putative autoimmune etiology have classical MS – even if they formally meet the ‘positive’ clinicoradiological criteria for MS [46]. MOG-IgG-positive patients in whom the disease starts with isolated brain or brainstem involvement are particularly challenging. Thus more and more importance attaches to carefully considering the ‘negative’ criterion of ruling out other diagnoses (“no better explanation”) included in the current diagnostic consensus criteria for MS [46].

Of note, 11 patients who met the clinicoradiological criteria for MS and 11/14 patients in whom a diagnosis of MS was initially suspected by their then treating physicians were negative for CSF-restricted OCB. Similarly, many patients with AQP4-IgG-positive NMO who were falsely diagnosed with MS in the past were negative for OCBs in a previous study [34]. This suggests that CSF analysis should be re-included in the diagnostic criteria for MS as an important tool to exclude alternative diagnoses, as previously recommended by us and others [55]. Moreover, 11/16 patients in whom MS had been initially suspected, later developed LETM lesions, which are not typically present in classical MS. In total, 15 out of the 16 patients were either negative for OCBs or had LETM lesions.

Most patients have relapsing disease

The relatively long observation time is a particular strength of the present study, since it allows assessment of disease course and outcome in the long run. While previous studies with shorter observation periods (12 months in [13], 18 months in [11], 2 years in [12]) and smaller sample sizes (4 patients in [13], 9 in [11], 16 in [12]) suggested that MOG-IgG-positive patients might often have monophasic disease, our series demonstrates that most MOG-IgG-positive patients with ON or myelitis have a relapsing disease course. Moreover, a very short median time to first relapse of just 5 months was noted in this cohort, indicating an overall high risk of early relapse in MOG-IgG patients.

Given that (i) the observation period among ‘monophasic’ patients was significantly shorter than in the relapsing subgroup and below the median time to relapse in around one third of the ‘monophasic’ patients, (ii) the proportion of relapsing patients increased with observation time (Fig. 2), and (iii) the interval between the first and the second attack was long in some of the relapsing cases (>12 months in eight; up to 492 months), it is conceivable that some of the few ‘monophasic’ patients will develop further attacks in the future. A monophasic course of disease might thus be even less common than suggested here.

On the other hand, time since onset was >5 years at last follow-up in 3 patients in the monophasic group, all of whom were not treated with immunosuppressants, so the disease may in fact follow a monophasic course at least in small proportion of cases.

Similarly, the significantly shorter observation time since onset in patients with a history of ON but no myelitis or of LETM but no ON than in patients with a history of both ON and myelitis (i.e., NMO) suggests that the differences in presentation between these groups are probably an effect of observation time and that some of our patients with isolated ON may develop myelitis in the future and some of those with isolated LETM may develop attacks of ON. Indeed, disease had started with either isolated ON or isolated myelitis (rather than simultaneous ON and myelitis) in around three fourths of patients with a diagnosis of NMO at last follow-up. Importantly, myelitis occurred only after several ON attacks in some of these patients and ON only after several myelitis attacks in others. Similarly, disease starts with isolated ON or myelitis rather than simultaneous ON and myelitis in the vast majority of AQP4-IgG-positive patients [34].

These findings are highly important when it comes to deciding whether to treat MOG-IgG-positive patients or not. The frequently relapsing course observed in the present cohort indicates that prophylactic long-term immunotherapy should be considered in MOG-IgG-positive patients. Given that a relapsing course was also noted in 5/8 (63%) patients with onset under the age of 18, this might possibly hold true also for children and adolescents. Studies systematically investigating the efficiency of long-term immunosuppression and/or immunomodulation in MOG-IgG-positive ON and myelitis are therefore strongly warranted. Moreover, given the lack of systematic long-term treatment data in MOG-IgG-positive disease, currently planned or ongoing treatment trials in NMO that include AQP4-IgG-negative patients should consider testing for MOG-IgG to allow subgroup analyses.

Severe attacks and unfavorable long-term outcome are relatively frequent

Severe attack-related disability was noted in many cases, including tetraparesis in around 30%, severe motor dysfunction with MRC grades ≤2 in around 25%, pain -and/or dysesthesia in around 70%, bladder and bowel disturbances in around 70%, functional blindness in almost 75%, bilateral optic nerve damage in 51%, scotomas in 66%, and brainstem encephalitis with, among other symptoms, ataxia, intractable nausea and vomiting or, of particular note, attack-related respiratory insufficiency in two patients, which was fatal in one. Importantly, long-term outcome was characterized by marked persisting visual impairment or blindness and/or significantly impaired ambulation in 40%. Moreover, inflammatory damage was noted in the entire CNS, with the spinal cord, optic nerves, brainstem, diencephalon, cerebellum, and telencephalon affected in individual patients. These findings, together with the mostly relapsing course observed in our patients, underline that MOG-IgG-related CNS autoimmunity is a severe condition that requires consistent treatment and care.

Although a favorable outcome was noted in several patients, our findings do not support the notion that MOG-IgG seropositivity generally denotes a mild disease course [11, 13]. Again, previous studies reporting such findings may have been unintentionally biased by short observation periods and small sample size.

It should be taken into account in both epidemiological and therapeutic studies in the future that optic nerve damage is the leading manifestation of MOG-IgG-positive autoimmunity and that MOG-IgG-positive patients may present for many years or decades with isolated ON. The EDSS, which was developed for use in classical MS and which largely focuses on ambulation, may not sufficiently reflect the high degree of disability resulting from persisting visual loss in a substantial number of MOG-IgG-positive patients. Other scales of disability may need to be used in addition.

IVMP was not always effective and flare-ups were frequent

In this context it is relevant that high-dose IVMP, though effective in many cases, was followed by only partial recovery or no recovery in 50% of all treated attacks. Moreover, IVMP lead to only temporary improvement in 44% of all patients at least once, resulting in flare-up of symptoms requiring repeat or ultra-high-dose IVMP therapy. In at least one case symptoms flared up not immediately after IVMP treatment but in a delayed fashion after tapering of subsequent oral steroid treatment. In some cases, even ultra-high-dose IVMP was ineffective or only transiently effective with a second flare-up occurring shortly after. Interestingly, in some patients IVMP was effective during initial attacks but not later in the disease course.

The occurrence of cerebral venous sinus thrombosis in one of our patients highlights the risks that repeat IVMP therapy and escalation to ultra-high-dose IVMP carry.

It is unknown why IVMP was effective during some attacks but not all. However, timing issues and differences in antibody titers, other immunological parameters (e.g., T cell activation), IVMP dosage, and previous or concomitant treatments might play a role.

Given the high frequency of flare-ups observed in our cohort, close clinical monitoring after acute attack therapy for MOG-IgG-positive ON and/or myelitis is recommended. Moreover, oral tapering of corticosteroid therapy as well as additional PEX treatment (see below) should be considered.

MOG autoimmunity may underlie CRION in a subset of patients

Given the high proportion of patients with flare-up of ON after steroid withdrawal, i.e., of steroid-dependent ON, we propose that a subset of patients previously diagnosed as having CRION [56] may in fact have MOG-IgG-positive ON. Indeed, at least 3 of our patients had received a diagnosis of CRION before MOG-IgG was detected. Testing of larger cohorts of patients with CRION for MOG-IgG is highly warranted.

PEX treatment was often followed by full or partial recovery

PEX was used in most cases as rescue therapy if steroids did not result in complete recovery; only in four patients was PEX used as first-line treatment for acute attacks. Of note, PEX treatment (as stand-alone therapy or following IVMP) was followed by complete or almost complete recovery in a substantial number of attacks (around 40%). For example, in case 2 ON symptoms flared up twice after high-dose and subsequent ultra-high-dose IVMP therapy; only PEX ended the attack and was followed by complete recovery. The efficacy of PEX in this and other cases of MOG-IgG-positive ON and/or myelitis has potentially important pathophysiological implications, since it suggests a direct pathogenic role of the antibody. Interestingly, PEX treatment stopped the progression of dysesthesia in case 9, one of only 3 cases in which a slowly progressive (yet also relapsing) course of disease was noted.

However, as a limitation, it should not be overlooked that in almost 60% of attacks treated with PEX, and thus in the majority of cases, only partial recovery was achieved, and in 2 cases there was no response to PEX. The variability in response to PEX may be linked to differences in PEX timing; MOG-IgG titers; intensity, extension, and site (e.g., ON vs. myelitis as seen in AQP4-IgG-positive NMOSD [35]) of inflammation; and, importantly, the number of PEX courses applied, which varied between 3 and 11 in the present cohort. Preliminary findings from our laboratory (S.J., unpublished data) show that AQP4-IgG and MOG-IgG may remain detectable even after five to seven plasma exchanges, raising the question of whether PEX treatment is discontinued too early in some cases. This is also supported by the early reoccurrence of attacks in cases 1 and 9 in part 3 of this series [31], just 1, 2 and 3 months after PEX. Alternatively, T cell-mediated mechanisms might play a more important role than antibody-mediated mechanisms in patients who do not sufficiently respond to PEX.

On the understanding that the use of PEX after IVMP implies previous IVMP failure, only incomplete recovery or no recovery at all was achieved in around 60% of all attacks treated with IVMP (N = 147). Twenty-five of those attacks were subsequently treated with PEX, and full recovery was achieved in 40% of them. This would suggest a beneficial role of PEX in MOG-IgG-positive patients with IVMP failure, similar to what has been observed in AQP4-IgG-positive NMOSD [35]).

The overall good response to escalatory PEX therapy, together with the risks associated with extensive cortisone pulse therapy as highlighted by the occurrence of sinus thrombosis with brain edema and seizures in case 2 might suggest that PEX treatment should be considered more often in patients with MOG-IgG-positive ON and/or myelitis. PEX may be considered as a substitute for escalatory ultra-high-dose IVMP therapy for severe attacks, particularly in patients who have responded well to PEX in the past. However, the observation of urosepsis in case 1 of part 3 [31] after several cycles of PEX illustrates that attention must be paid also to risks associated with PEX and IA, especially if those treatments are applied repeatedly and in combination with IVMP or IS treatment.

Breakthrough attacks despite long-term immunotherapy

Similarly, long-term IS and IM treatments were not always effective in preventing further relapses. Almost 70% of all patients treated with IS or IM drugs developed at least one attack during therapy. This included patients receiving AZA, MTX, NAT, IFN-beta, GLAT, rituximab, ofatumumab, and mitoxantrone. In case 6 at least 12 attacks of ON and myelitis occurred under various immunotherapies, and as many as 15 attacks occurred in case 1 in part 3 of this article series [31].

Complications of IS/IM therapy were rare in this cohort and included condylomata acuminata requiring surgical treatment, elevated liver enzymes under AZA treatment, and an allergic reaction to rituximab.

AZA failure was associated with latency period and lack of cotreatment

AZA, which has been previously reported to be partially effective in NMO [57–59], including in AQP4-IgG-positive NMOSD [60], was the most commonly applied IS therapy in our cohort. However, more than 80% of all AZA-treated patients experienced at least one attack while under therapy. As AZA has a latency period of 3-6 months during which cotreatment with oral steroids has been recommended [33], we analyzed the temporal pattern of AZA failure. Of 34 attacks during AZA treatment, 14 (41.2%) took place during the first 6 months (11 during months 1-3 and 3 during months 4-6). Furthermore, 12 of those 14 attacks (85.7%) occurred in patients (n = 6) not cotreated with oral steroids, PEX, or other immunosuppressants during that period. This suggests that AZA failure in MOG-IgG-positive patients may be caused in a substantial proportion of cases by the drug’s well-known latency in efficacy. Moreover, it may indicate that cotreatment with oral steroids during the initial 6 months of AZA treatment should not be abandoned in MOG-IgG-positive patients, provided contraindications have been excluded. However, it must be mentioned as a potential limitation that AZA was discontinued early in some patients after breakthrough attacks occurred, which may have introduced a considerable bias towards a higher proportion of attacks in the first 6 months. Larger studies are therefore needed before any treatment recommendations can be made.

Future studies on the efficacy of AZA and oral steroids as well as on that of oral steroids as stand-alone therapy in MOG autoimmunity should take into account that a recent retrospective analysis suggested a better response rate to high-dose azathioprine (2.5-3 mg/kg) than to standard treatment (1-1.5 mg/kg) in AQP4-IgG-positive NMOSD [58]. Whether such a high-dose regimen is also required in MOG-IgG-positive patients is currently unknown.

Low relapse rate under MTX in most but not all cases

A recent study suggested that MTX might be effective in patients with AQP4-IgG-positive NMOSD [33, 61]. In our cohort, eight MOG-IgG-positive patients with ON and/or myelitis were treated with MTX, and exact data on attack dates were available from six. A lower relapse rate than in the total cohort and long attack-free intervals were observed in most MTX-treated patients. Based on these preliminary yet promising results, further retrospective studies seem warranted to assess the efficacy of MTX in MOG-IgG-positive ON and/or myelitis.

Attacks related to initial rituximab infusion and reappearance of B cells

Rituximab treatment was followed by a clear reduction in relapse rate in three out of nine patients. In the six remaining patients, relapses occurred 2, 3, 4, 4, 7, 8, 8, 12 and 20 weeks after the first or second infusion (in one case despite PEX treatment 1 month earlier). This is reminiscent of the transient deterioration reported in some patients with AQP4-IgG-positive NMO after commencement of rituximab, which is associated with an temporary increase in BAFF and autoantibody levels [62, 63]. Another MOG-IgG-positive patient who experienced postinduction relapses (three within 3 months) has recently been described [63]. Whether cotreatment with steroids can prevent such events still needs to be explored.

Of note, one patient relapsed immediately after reappearance of B cells. This is similar to what has been observed in AQP4-IgG-positive NMO patients [64, 65] and suggests that (i) B cells should be closely monitored in MOG-IgG-positive patients treated with rituximab and (ii) treatment intervals should be short and doses high enough to prevent B cell reappearance. Rituximab has also been found to be effective in AQP4-IgG-positive NMOSD in some studies [65, 66], though not in all [67].

Ofatumumab is a fully human anti-CD20 monoclonal antibody which targets an epitope distinct from that of rituximab [68]. The marked reduction in relapse rate in the single patient treated with ofatumumab in this study is promising. However, more data are needed before any recommendations can be made. To the best of our knowledge this is the first report on ofatumumab both in MOG-IgG- and in AQP4-IgG-associated EM.

Ongoing or increasing disease activity under IFN-beta

In the present cohort, 4 patients were treated with IFN-beta. All 4 showed ongoing or increasing disease activity. Although preliminary, these data suggest that IFN-beta, which has already been shown to be ineffective and to cause disease exacerbation in AQP4-IgG-positive NMOSD [69–72], may also be ineffective or even detrimental in MOG-IgG-positive patients. Given the substantial clinical overlap between MOG-EM and conventional MS, a condition often treated with IFN-beta, this would be of high clinical relevance. Larger retrospective studies evaluating the efficacy of IFN-beta in MOG-IgG-positive EM are therefore highly warranted.

Preliminary data do not support use of GLAT or NAT

Like IFN-beta, GLAT is frequently used to treat patients with conventional MS. With the efficacy of GLAT being equivocal in two patients and eight breakthrough attacks having occurred in another three, the use of GLAT cannot currently be recommended in MOG-IgG-positive EM. Of note, GLAT has also been suggested to be of no clear benefit in patients with AQP4-IgG-associated NMOSD [36].

NAT, another drug shown to be beneficial in MS, could not prevent relapses in three MOG-IgG-positive patients in our cohort. While these preliminary data are not supportive of the use of NAT in MOG-IgG-positive ON or myelitis, systematic studies are certainly needed before definite conclusions can be drawn. NAT has also been found to be ineffective or even detrimental in patients with AQP4-IgG-positive NMOSD in recent studies [73–75]. Numerous relapses also occurred in a patient treated with mitoxantrone, another agent considered effective in conventional MS.

The failure of the MS therapeutics IFN-beta, GLAT and NAT in many of our patients supports the view that classical MS and AQP4- or MOG-IgG-associated disorders differ in terms of immunopathogenesis. This stance is further supported by the lack of OCB, a hallmark of conventional MS, in most of our MOG-IgG-positive patients as well as in most patients with AQP4-IgG-positive NMOSD [34, 76].

MOG-IgG needs to be considered in children as well as in elderly patients

The median age at onset was around 30 years, which is similar to MS but differs from that in AQP4-IgG-positive NMOSD (~39) [34] by almost 10 years. However, the youngest patient in this cohort was just 6 years of age at onset and the oldest experienced his first attack at age 70, suggesting that MOG-IgG-positive ON and/or myelitis – just like AQP4-IgG-positive NMOSD [34] – can occur irrespective of age and need to be considered also in children and in the elderly.

All four patients with onset during childhood (at the ages of 6, 10, 12, and 12 years) initially presented with ON (bilateral in three), accompanied by myelitis in only one of them. Similarly, disease started with ON in the four oldest patients (onset at 58, 64, 66 and 77 years). Of note, 8/11 (73%) patients with onset at age <20 had a recurrent disease course at last follow-up, which led to relevant disability in 5 of them (EDSS 3.5, 3.5, 6.0, permanent unilateral blindness, and VA of 0.2, respectively, at last follow-up), and in one of the two remaining patients observation time was too short to rule out relapsing disease. This would argue against the notion that MOG-IgG-positive ON in young patients is generally a monophasic disease and suggests that long-term immunotherapy (e.g., with IVIG if immunosuppressants are to be avoided) should be considered also in children and adolescents; however, larger studies are certainly needed. A recurrent course was also present in three of the four oldest patients in this cohort.

Interestingly, the time between onset and first relapse was extraordinarily long in one of our patients (see case 6 in part 3 of this series [31]), who had suffered from a first attack of ON at age 12, followed by an LETM attack 41 years later. Of note, two further patients reported events during childhood that are compatible with a first attack of MOG-IgG-positive EM (two episodes of bulbar movement pain, diplopia, and headache at ages 10 and 11 in case 23, and “neurogenic diabetes insipidus” at age 7 in case 22). As it remains unclear whether those early events were caused by the same disorder as the patients’ more recent complaints, which started 18.5 and 22 years later, they were not considered for statistical analysis.

With an age at onset of 70 years, patient 20 is, to the best of our knowledge, the oldest Caucasian MOG-IgG patient reported to date. The youngest patient described in the previous literature was just 1 year of age [77] and the oldest, a Japanese patient, 70 years [12] at onset.

Similarly, AQP4-IgG-positive NMOSD has been described both in children and in elderly patients [78–81]. Accordingly, MOG-IgG-positive EM is an important differential diagnosis of AQP4-IgG-positive NMOSD irrespective of age.

Women are more often affected than men

Women outnumbered men by a factor of around 2.8 in this study. Female gender has also been identified as a risk factor for AQP4-IgG-positive NMOSD [82, 83]. However, a significantly higher preponderance of women (a male to female ratio of around 1:9) has been found in the latter condition in Caucasian patients [34]. The ratio found among MOG-IgG-positive patients in the present and previous cohort is more similar to that found in AQP4-IgG-negative NMOSD [34] and in classical MS [84]. Women might possibly be affected more severely than men as indicated by a higher median EDSS at last follow-up despite shorter median disease duration; however, confirmatory studies are needed to verify this finding.

Attacks may occur during pregnancy and post partum

The effect of pregnancy on MOG-IgG-positive EM has not yet been systematically investigated. A recent study (n = 16) indicated that pregnancy may negatively influence the disease course of NMO; however, no data on the patients’ AQP4-IgG or MOG-IgG status were given [85]. The authors found a significantly higher attack rate in the first trimester after pregnancy and greater disability progression 1 year after delivery [85]. In another cohort [86], 14/40 AQP4-IgG-positive patients developed the first symptoms of NMOSD either during pregnancy (n = 3) or within a year after delivery or abortion (n = 11). While the ARR during pregnancy did not differ from that before pregnancy, it increased significantly during the first and second trimesters after delivery; moreover, 77% of all deliveries were associated with post-partum relapses. In MS, pregnancy is thought to reduce the number of MS relapses, especially in the second and third trimesters; although attack rates tend to rise in the first 3-6 months post partum, no increased long-term disability has been found. In the present cohort, around a quarter of MOG-IgG-positive women experienced one or more attacks during pregnancy or post partum. Of special note, the disease started post partum in three of these patients. This could indicate that pregnancy- and/or delivery-related immunological changes may play a role both in triggering attacks and, possibly, in disease induction. However, given that most attacks in these patients as well as in the total female subgroup occurred irrespective of pregnancy and delivery, other risk factors may be more important. Based on these data, systematic prospective studies on the role of pregnancy and delivery in MOG-IgG-positive EM are warranted.

Attacks may follow infection or vaccination

Attacks were preceded by infection in around 40% of patients at least once, and disease started shortly after an infection in at least 11 cases and after vaccination in two cases. This is similar to AQP4-IgG-positive NMOSD, which has been reported to be preceded by infection in 20-30% of cases [34, 87]. Acute infections are also thought to trigger clinical attacks in classic MS. However, the exact relationship between infection or vaccination and MOG-IgG-positive EM is unknown. While there is no evidence yet for molecular mimicry, it is conceivable that infection-associated immunological changes and/or blood-brain barrier disruption could promote CNS lesion formation. In AQP4-IgG-positive NMOSD, acute relapses are indeed associated with an elevated QAlb, which can be caused by structural barrier damage [76, 88]. QAlb was also elevated in around one third of MOG-IgG-positive patients in the present study. As a limitation, QAlb likely also reflect changes in the CSF flow rate [45]. It is of potential interest and deserves further investigation that the two post-vaccinal cases both occurred after vaccination against tetanus, diphtheria and pertussis. Of note, both patients developed relapsing disease. This is different from conventional postvaccinal ADEM, which is usually monophasic.

CSF findings differ from MS but mimic AQP4-IgG-positive NMO

Examination of CSF harbors important potential for differentiating classical MS and MOG-IgG-positive EM but not MOG-IgG-positive EM and AQP4-IgG-positive EM: As in AQP4-IgG-positive patients [76], OCB and a positive IgG CSF/serum ratio, which are present in most patients with MS and which are thus considered a diagnostic hallmark of that disease, were missing in around 90% of our MOG-IgG-positive patients. Moreover, OCB disappeared later in 2 out of the 6 only OCB-positive patients; by contrast, OCB are considered to remain stable for decades in MS [89]. Finally, neutrophil granulocytes, which are also present in AQP4-IgG-positive NMO [34, 76, 90] (as well as in bacterial meningoencephalitis [91]), were found in the CSF at least once in 64% of cases, but are absent in classical MS.

Missing OCB or granulocytic pleocytosis should thus prompt physicians to challenge the diagnosis in patients with suspected MS and to consider MOG-IgG- or AQP4-IgG-positive encephalomyelitis.

Subclinical evidence for dissemination in space

Electrophysiological evidence for optic nerve damage was present in at least 3 patients with no history of clinically apparent ON, and for spinal cord damage in at least 3 patients with no history of clinically apparent myelitis in our cohort. Similarly, supratentorial, brainstem, or cerebellar MRI lesions were present in 21 patients who had never shown clinical signs of encephalitis or cerebellitis but only of ON and/or myelitis. Finally, spinal cord MRI lesions were detected in 2 patients with ON but no history of clinical myelitis. This indicates that subclinical inflammation occurs in some cases and that clinical examination needs to be complemented by electrophysiology and MRI to assess the real extent of CNS inflammation in MOG-IgG-positive patients.

If not only clinical attacks are taken into account but also clinically silent lesions as detected electrophysiologically or by MRI, evidence for dissemination in space (defined as involvement of more than one of the following structures: optic nerves, spinal cord, supratentorial brain, brainstem, cerebellum) was present at last follow-up in 37/50 or 74% of patients, compared with 26/50 or 52% based solely on clinical grounds.

VEP and SSEP were not considered in the 1999 and 2006 diagnostic criteria for NMO, which required clinically apparent attacks of myelitis and ON, and are still not considered in the 2015 criteria for NMOSD [29]. Systematic studies on the potential prognostic, diagnostic, and therapeutic implications of pathological EP and MRI findings suggesting dissemination in space in patients with MOG-IgG-positive isolated ON or isolated myelitis are warranted. Evidence for subclinical optic nerve damage has also been reported in AQP4-IgG-positive NMOSD [92].

Bilateral ON and simultaneous ON and myelitis are common at onset

More than 40% of all patients with a history of both ON and myelitis at last follow-up presented with simultaneous myelitis and ON at least once, which is not different from what has been described in AQP4-IgG-positive NMO (42% according to [34]). However, the frequency of simultaneous myelitis and ON at disease onset, i.e., as the initial presentation, was much higher in MOG-IgG-positive patients (23% of all patients with a history of ON and myelitis) than in AQP4-IgG-positive NMOSD patients (6.7% according to [34]; p < 0.03). Similarly, bilateral ON at onset was more frequent in MOG-IgG-positive patients with a history of ON (35%) than in AQP4-IgG-positive NMOSD patients with a history of ON (14.3% [34]; p < 0.04 ). Simultaneous ON and myelitis as well as bilateral ON at onset may thus be of diagnostic value and should prompt physicians to consider MOG-IgG testing.

Short spinal cord lesions do not preclude MOG-IgG positivity

Spinal cord MRI lesions extending over three or more vertebral segments (so-called LETM) are considered a hallmark of AQP4-IgG-positive NMOSD, but are usually not found in classical MS. The presence of an LETM lesion in addition to clinical myelitis was also listed as a supportive criterion in the 1999 diagnostic criteria for NMO and one of three minor characteristics, two of which had to be present in addition to a history of ON and myelitis before a diagnosis of NMO could be made, in the 2006 criteria [28]. The association of MOG-IgG with LETM found in this and in previous studies is therefore of differential diagnostic importance.

However, two recent studies could demonstrate that up to 15% of all MRIs of AQP4-IgG-positive patients show non-longitudinally extensive lesions [34, 93]. Similarly, lesions never exceeded two vertebral segments in 8 of our MOG-IgG-positive patients; in another 10 patients, at least one MRI showed only a non-LETM lesion but longitudinally extensive lesions were present in previous or later MRI examinations. The presence or absence of ‘short’ lesions in patients with AQP4-IgG- or MOG-IgG-positive myelitis is thought to depend, among other factors, on timing issues [94]. If MRI is carried out very early in the attack course or long time after an acute attack, lesions may be still evolving or be already in the process of resolution, respectively.

Similar to AQP4-IgG-positive myelitis, more than one lesion in the same MRI and swelling of the spinal cord were detected in many patients at least once. By contrast, necrotic lesions leading to spinal cord cavitation, as sometimes noted in AQP4-IgG-positive myelitis, were not reported in any of our MOG-IgG-positive patients.

Lesions may affect the entire visual pathway

While retrobulbar optic neuritis was highly common among our MOG-IgG-positive patients, lesions affecting other parts of the optic pathway should be taken into consideration as well in MOG-IgG-positive patients presenting with visual symptoms. Many patients had signs of papillitis as detected fundoscopically; evidence for inflammation of the anterior part of the optic nerve was also found by MRI (Fig. 9). However, some patients presented with lesions in the chiasm (Fig. 9) and/or with longitudinally extensive ON (LEON) affecting both the anterior and the posterior portion of the optic nerve. Both LEON lesions and chiasmatic lesions were previously thought to be indicative of (AQP4-IgG-positive) NMOSD [29]. Our findings are in line with a recent Australian study that reported greater optic nerve lesion lengths in MOG-IgG-associated ON and AQP4-IgG-associated ON than in MS-related ON [95]. In a single patient, visual disturbances were associated with lesions within the optic tract (Fig. 9). Finally, some patients had occipital white matter lesions.

Perioptic contrast enhancement warrants further investigation

As shown in Fig. 9, contrast enhancement was not only seen within the optic nerve but also in the perioptic nerve sheath and the immediately surrounding orbital tissue. This imaging pattern is of potential differential diagnostic relevance and, thus, deserves to be further investigated. In accordance with this finding, Kim et al. in a very recent study found perineural enhancement in 6 of 18 MOG-IgG-positive patients [96]. As is the case with other MRI features, it is likely that the presence or absence of that phenomenon depends on disease and treatment status: more than one third of all MRIs without perioptic enhancement in our cohort were performed during remission, and some of the remaining patients had been treated with high-dose IVMP before the MRI was performed.

Coexisting autoimmunity is rare in MOG-IgG-positive patients

Coexisting autoimmune disorders are present in more than one third of AQP4-IgG-positive NMOSD patients [34]. By contrast, only around 9% of our MOG-IgG-positive patients had a coexisting autoimmune disorder (2 × RA, 1 × Hashimoto thyroiditis, 1 × Grave’s disease). Systemic lupus erythematosus, Sjögren syndrome, and myasthenia gravis, which are common in AQP4-IgG-positive NMOSD [97–102], were absent in all of our MOG-IgG-positive patients. This is in line with a previous study that reported a lower frequency of concomitant autoimmune disorders in ‘AQP4-IgG-seronegative’ NMOSD patients [34]. Interestingly, first symptoms of anti-TPO-, anti-thyreoglobulin-, anti-TSH receptor-associated hyperthyreosis appeared just seven weeks after the first attack of MOG-IgG-positive myelitis in one of our patients, suggesting that MOG autoimmunity might have been part of a broader immune dysregulation in this case.

Nosological issues

The 2015 diagnostic consensus criteria for NMOSD demand that “alternative diagnoses” should be excluded [29]. However, it remains unclear whether MOG-IgG-positive ON or myelitis should be considered an “alternative diagnosis” or not [103–105]. AQP4-IgG-positive and MOG-IgG-positive EM differ in terms of target structures (astrocytes vs. oligodendrocytes) and, accordingly, immunohistopathology [18–20], but MOG-IgG is not explicitly mentioned as an exclusion criterion. This mainly reflects the fact that at the time the criteria were developed, data on MOG-IgG-positive patients were still scarce.

In the present study, only around one third of all MOG-IgG-positive patients met the clinical and radiological criteria for “NMOSD without AQP4-IgG” [29]. If MOG-IgG is not considered an exclusion criterion for NMOSD, this would result in a subset of MOG-IgG-positive patients being considered eligible for clinical studies and treatment trials, while others would be excluded based solely on phenotypic presentation and despite all of these patients belonging to the same immunopathogenetically defined disease spectrum. This could introduce a relevant inclusion bias given that almost all of the patients who met the criteria had a history of ON and LETM (15/16 or 94%), while most of those who did not meet the criteria had isolated ON or isolated LETM at last follow-up (27/34 or 79%). Conversly, inclusion of MOG-IgG-positive patients in NMOSD cohorts (which are predominantly AQP4-IgG-positive) would introduce bias as well.

We therefore believe that confirmed MOG-IgG seropositivity should be considered an exclusion criterion for NMOSD and that the term ‘NMOSD’ should be restricted to AQP4-IgG-positive patients and, possibly, double-seronegative patients meeting the criteria for ‘AQP4-IgG-negative NMOSD’ [29].

There are two potential limitations to such an approach. First, using MOG-IgG seropositivity as an exclusion criterion for NMOSD would require providing a reference assay for MOG-IgG testing with excellent specificity as established in an appropriately controlled multicenter setting. Alternatively, however, diagnostic criteria for MOG-IgG-positive NMOSD based both on serological and on supporting clinicoradiological criteria could be established in analogy to the current consensus criteria for NMOSD. Second, so-called ‘double-positive’ patients, i.e., patients positive for both AQP4-IgG and MOG-IgG, would pose a diagnostic dilemma. However, such patients should in any case be excluded from clinical trials as they may have two immunopathophysiologically distinct diseases. Moreover, ‘double-positive’ patients seem to be extremely rare (see part 1 of this article series [30] and Table 4 in [17]) and the few cases reported so far worldwide have not been independently confirmed.

Limitations

We acknowledge some obvious limitations of our study. First, the study design was retrospective, as in all previous studies in the field, and a high number of neurological centers were involved. However, due to the low prevalence of the condition, prospective single-center studies including sufficiently large numbers of patients are impracticable. Moreover, the multicenter design of this study, which included 11 academic centers, reduces the risk of referral bias, which was acknowledged as a possible limitation by the authors of previous large single-center studies in the field of NMO [28, 106]. Moreover, reliable assays for detecting MOG-IgG have only recently been developed; accordingly, only retrospective long-term data are currently available. Second, patients with a benign or monophasic long-term course are less likely to be admitted to hospital and might thus be under-represented in the present cohort. However, this type of potential bias is inherent in hospital-based studies and cannot be completely avoided. It is important in this context that all centers involved in the present study also have specialized outpatient clinics for patients with neuroinflammatory conditions and that participants were recruited among both inpatients and outpatients. Moreover, the threshold for admission is low in Germany and Italy, where public healthcare is free. Third, MOG-IgG has also been reported in patients with conditions classified as ‘ADEM’ based on clinical and radiological features, especially in children [107]. Such cases were not systematically included in the present cohort, which focused on patients with ON and/or myelitis. Given that our study found a relapsing disease course in most patients with MOG-IgG-positive CNS inflammation and that attacks did not develop until years after initial presentation in some cases, systematic follow-up studies on patients previously diagnosed with MOG-IgG-positive ‘ADEM’ seem warranted to confirm that rare association.