Treatment regimens for neuromyelitis optica spectrum disorder attacks: a retrospective cohort study

Neuromyelitis optica spectrum disorder (NMOSD) attacks are more damaging than multiple sclerosis (MS) attacks and require more aggressive treatments [1‐3]. According to the most recent diagnostic criteria [4], NMOSD may be associated with anti-aquaporin 4 (AQP4) antibodies targeting astrocytes, with anti-myelin oligodendrocyte glycoprotein (MOG) antibodies targeting oligodendrocytes or with double seronegative status (NMOSD–DN, i.e., seronegative for AQP4 and MOG antibodies). However, AQP4-Ab-positive patients and MOG-Ab-positive patients have different immunopathologic and clinical courses and the nosography is currently evolving rapidly. Indeed, patients positive for MOG-Ab included in the NMOSD set of criteria have recently been individualized as having MOG-associated diseases (MOGAD) [5, 6], and there is some debate about the classification of MOGAD in NMOSD [7]. Despite this, AQP4-Ab-positive and MOG-Ab-positive patients can have a similar initial presentation, corresponding to an NMOSD phenotype; Such phenotypes have been well described [4]: longitudinally extensive transverse myelitis, severe optic neuritis (ON), area postrema/diencephalic/brainstem or encephalic syndrome. The first attack of these patients must, therefore, be considered diagnostically uncertain until the results of the serological status are known. During this period, if an NMOSD phenotype is observed, there is an urgent need to instigate a powerful anti-inflammatory therapeutic strategy. Consequently, despite MOGAD having a broader and less clearly defined spectrum of manifestations than classical NMOSD as defined by the IPND criteria [4, 8‐10], we focused on NMOSD phenotypical first presentations, and thus used the term NMOSD–MOG+ in this report.

Attack treatment strategy in NMOSD relies on various combinations of high-dose intravenous corticosteroids and apheresis methods including plasma exchanges (PE), immunoadsorption (IA) or intravenous immunoglobulins (IVIG). As inadequately treated attacks result in disability [11, 12], there is a need to identify the optimal attack treatment regimen. There is no randomized clinical trial for the treatment of NMOSD relapses but there have been efforts to reach a consensus [13‐15]. Previous retrospective studies suggest the superiority of PE in AQP4-Ab-positive patients (NMOSD–AQP4+), with better outcomes when initiated early [11, 16‐21]. However, most of these studies did not address the specificities of NMOSD–DN or NMOSD–MOG+ and suffered from potential confusion with the effects of immunosuppressive treatments at the time of the attacks. Therefore, we investigated in naïve-treatment patients the first attacks for any given location related to an NMOSD presentation and looked for predictive factors to identify the best treatment strategy.

Among the French national NMOSD cohort NOMADMUS, 374 patients met the inclusion criteria. Our on-site data collection found a complete data set including posology and time to attack treatment, EDSS score at nadir and at 6 months or found enough data to score them retrospectively in 183 patients from 11 centers (Table 1). This represents 211 attacks; 179 attacks were the first one in the patient’s history (84.8%). The others affected an independent location. Almost all attacks were confirmed on MRI (n = 202; 95.7%). A majority of patients were female (n = 121; 68.9%). The mean age at first attack was 38.5 years (range: 13–80). Among the 183 patients, 104 had anti-AQP4 antibodies (56.8%), 60 had anti-MOG antibodies (32.8%), and 19 were NMOSD double seronegative (10.4%). Attacks were mostly isolated myelitis (n = 93; 44%) and isolated ON (n = 74; 35%), with 16 combined attacks of neuromyelitis optica (NMO; 8%) (Table 2). The median nadir EDSS score was 5.00 (range: 1–9.5). Among the myelitis attacks, 98 were longitudinally extensive and transverse at MRI (76%). Among the ON attacks, 43 were associated with papilledema (44.8%), 39 were bilateral (40.6%), and 33 were extensive at MRI (34.4%).

The majority of patients in the NMOSD–AQP4+ and NMOSD–DN groups were female (85% and 61%, respectively) but the sex ratio was more balanced in the NMOSD–MOG+ group (45% female). The mean age at first attack was higher in the NMOSD–AQP4+ group (40.8 years, versus 35.6 years in the NMOSD–MOG+ group and 33.8 years in the NMOSD–DN group; p = 0.021). Myelitis was the most frequent location in the NMOSD–AQP4+ and NMOSD–DN groups (Table 2; 68% and 78%, respectively). ON was the most frequent presentation in the NMOSD–MOG+ group. Three cases of acute disseminated encephalomyelitis (ADEM) were reported (1.4%), all in the NMOSD–AQP4+ group. The nadir EDSS score was similar for all serogroups (Table 1; p = 0.106).

Patients were treated with various regimens of corticosteroids, PE, or IVIG (Fig. 2). The mean delay between the onset of symptoms and the start of treatment was 14.3 days (SD = 17.9 days). Corticosteroids were the most frequent first-line treatment (n = 193; 91.9%), with a mean delay of 14.3 days, and PE was the most frequent second-line treatment (n = 66; 70.2%), with a mean delay of 25.0 days (overall mean = 26.3 days, overall median = 16 days, whatever the treatment line). Fourteen attacks were untreated (6.6%). Ninety-four patients had a combined regimen (44.5%), corticosteroids followed by PE being the most frequent combination (Fig. 2). A second line of treatment was required for 44.5% of attacks and a third line for 5.7%. The treatment regimen was similar whatever the location of the attack (Fig. 2) and the serogroup (Table 1), except for a lower mean number of treatment lines for the NMOSD–MOG + group (mean = 1.24, versus 1.65 for the NMOSD–DN group and 1.53 for the NMOSD–AQP4+ group; p = 0.013).

At 6 months, the median EDSS score was 2.50 (range: 0–10), with a median improvement of 2.00 points (range: 3.0-point deterioration–8.0-point improvement). NMOSD–MOG+ patients showed the greatest improvement (Table 1), with a median 6-month EDSS score of 1.50 (p < 0.001) and a median improvement of 3.00 points (p = 0.001). Recovery was complete after 40 attacks (19.0%), according to the relative definition (Tables 3 and 4, Fig. 3), and the response was good after 134 attacks (63.5%), according to the absolute definition (Tables 3 and 5, Fig. 3). According to both definitions, there was no improvement (absence of recovery or response) after 50 attacks (23.7%). Patients in the NMOSD–MOG+ group reached complete recovery or a good response more frequently than patients in the NMOSD–AQP4+ and NMOSD–DN groups (respectively: 31.3% versus 14.0% and 8.7% for recovery; 82.1% versus 55.4% and 52.2% for treatment response; Fig. 3). We found these differential associations when corticosteroids or PE were included in the therapeutic regimen (Table 3). At 1-year follow-up, we recorded the introduction of a disease modifying treatment after 138 attacks of all 211 attacks (65.4%): by serogroup, after 91 of 121 attacks (75.2%) in the NMOSD–AQP4+ group, after 32 of 67 attacks (47.8%) in the NMOSD–MOG+ group and after 15 of 23 attacks (65.2%) in the NMOSD–DN group. Among the 73 attacks without the subsequent introduction of a disease modifying treatment, 65 (89.0%) were the patient’s first attack: by serogroup these comprised 26 of 30 attacks in the NMOSD–AQP4+ group, 32 of 35 attacks in the NMOSD–MOG+ group and 7 of 8 attacks in the NMOSD–DN group.

Using the relative classification of improvement as recovery (Table 4), our inferential statistical analysis showed that better outcomes were significantly associated with NMOSD–MOG+ patients (OR = 0.29 versus NMOSD–AQP4+ patients; 95% CI [0.15; 0.53]; p = 0.0042). Poorer recoveries were significantly associated with the requirement for more treatment lines (OR = 1.69; 95% CI [1.15; 2.51]; p = 0.009), delayed PE (OR = 1.04 per day; 95% CI [1.02; 1.07]; p = 0.0023) and extensive ON (OR = 3.51; 95% CI [1.46; 8.8]; p = 0.007). Recovery did not significantly correlate with bilateral ON (OR = 1.68; 95% CI [0.78; 3.66]; p = 0.19), nor ON associated with papillary edema (OR = 0.58; 95% CI [0.26; 1.28]; p = 0.18), nor longitudinally extensive transverse forms of myelitis (OR = 1.78; 95% CI [0.69; 4.68]; p = 0.24). At the multivariate analysis, MOG-Ab-positivity was significantly and independently associated with better recoveries (OR = 0.058 versus AQP4-Ab+ patients; 95% CI [0.0078; 0.35]; p = 0.0042) and delayed PE with poorer recoveries (OR = 1.04 per day; 95% CI [1.01; 1.07]; p = 0.012). Combined ON and transverse myelitis attacks (NMO) were independently associated with poorer recovery in comparison to the other presentations (OR = 8.82 versus isolated transverse myelitis; 95% CI [1.25; 80.2]; p = 0.041).

Using the absolute classification of improvement as treatment response (Table 5), better outcomes were significantly associated with MOG-Ab positivity (OR = 0.26 versus AQP4-Ab positivity; 95% CI [0.12; 0.52]; p < 0.001). Poorer responses were significantly associated with delayed PE (OR = 1.04 per day; 95% CI [1.01; 1.06]; p = 0.00431) and extensive ON (OR = 4.82; 95% CI [1.80; 13.8]; p = 0.00315). Treatment response did not significantly correlate with bilateral ON (OR = 1.68; 95% CI [0.71; 3.97]; p = 0.24), nor ON associated with papillary edema (OR = 0.55; 95% CI [0.21; 1.39]; p = 0.22), nor longitudinally extensive transverse forms of myelitis (OR = 0.93; 95% CI [0.41; 2.2]; p = 0.86). At the multivariate analysis, MOG-Ab positivity was significantly and independently associated with better responses (OR = 0.048 vs AQP4-Ab positivity; 95% CI [0.0041; 0.35]; p = 0.008) and delayed PE with poorer responses (OR = 1.03 per day; 95% CI [1.01; 1.07]; p = 0.024).

We studied 211 NMOSD attacks in 183 long-term immunotherapy-naïve patients, their treatment, and their outcome. Most patients had a good response to the attack treatments at 6 months (63.5%), but a marked proportion did not respond (23.7%), which confirms the severity of NMOSD attacks and the need to optimize their treatment. Our analysis highlighted two associations of predictive interest, namely, the serogroup and the PE initiation delay.

This study investigated the first attacks of patients with an NMOSD presentation according to their serostatus. The strength of our study lies in the nationwide coverage of patient recruitment, the MRI confirmation of almost all attacks and the high time resolution of EDSS scoring, enabling a precise measure of recovery (relative definition) and treatment response (absolute definition). The significant result of our study was the prognostic implication of the serogroup. We showed that patients in the NMOSD–MOG+ group had a better prognosis than patients in the NMOSD–AQP4+ group. This is in line with the MOGADOR study, which showed that MOG+ patients had an overall better outcome, fewer relapses, and less frequently reached an EDSS score ≥ 3 than NMOSD–AQP4+ patients [8]. The differences between serogroups (age at onset, sex ratio, attack characteristics, evolution) likely result from pathophysiological specificities, both in terms of antigenic (MOG or AQP4) and cell type targeting (oligodendrocytes for NMOSD–MOG+ and astrocytes for NMOSD–AQP4+). The pathways involved in the pathogenesis associated with NMOSD–MOG+ damage CNS myelin in a complement-independent manner, without leading to immune cell infiltrates, and appear to yield more reversible neural dysfunctions [24, 25]. These observations could have practical implications for personalizing attack treatment regimens based on the serogroup. However, this information is not currently available in the days immediately after the attack onset.

Our results are nevertheless already applicable to a clinical situation corresponding to a first attack in a patient with an NMOSD presentation. It should be noted that in such cases a decision on the therapeutic strategy needs to be made even though information on the patient’s serostatus is not immediately available. Although the treatment regimen in our French population was relatively homogeneous, with corticosteroids as first-line therapy and PE as second-line therapy, our study highlighted that delayed PE initiation was detrimental. PE are currently considered as an adjunctive therapy for NMOSD attacks. We argue that PE should be used systematically and combined with corticosteroids as soon as possible right from the first attack if its characteristics are suggestive of NMOSD. Currently, some an expert opinion [14] and a set of guidelines [26] suggest the use of PE right away for subsequent attacks in patients who did not respond to corticosteroids in the past. However, apart from our study, several reports have shown that early PE are associated with better recoveries, regardless of disease modifying treatments. In 2018, Bonnan et al. found that the probability of achieving a complete recovery continuously decreased as the delay to PE increased: from 50% when started at day 0 to 1–5% when started after day 20 [20]. In their cohort of 115 severe NMOSD attacks (AQP4+ or seronegative), the median PE delay was 7.0 days, in most cases associated with corticosteroids, and 63% of patients received PE within 10 days after attack onset, making this population different from ours (median delay to PE in our study: 16 days). Likewise, investigating a subgroup of 207 NMOSD attacks treated by PE, Kleiter et al. found 40% of complete recovery when PE was started within 2 days after symptoms onset, 29% within 6 days, with a rapid decrease beyond as there were no cases of complete recovery when PE delay exceeded 20 days [18]. Despite longer delays to PE in our cohort the effect of the delay to PE remained significant. This suggests that PE should be considered for severe attacks even at a late stage after the onset of symptoms. No such associations with the other lines of treatments have been found. A retrospective comparison further encouraged the combination of PE with corticosteroids as an attack treatment rather than corticosteroids alone [27]. IA could not be studied in our population. IA is an alternative apheresis therapy, which selectively depletes immunoglobulins without requiring transfusion of another patient’s plasma [28]. This extracorporeal technique has been less studied than PE, but shows a similar efficacy in small heterogeneous populations with neuroinflammatory diseases [29, 30]. We did not find predictive associations for recovery with age or attack location, in contrast to the recent study of NMOSD attacks in the LATAM registry [21]. Some discrepancies with our study might be due to geographic population differences and patients not always being treatment naïve. The different serostatuses were analyzed separately, whereas we considered them as predictors. In our study, NMOSD–AQP4+ correlated with greater ages at onset (Table 1; p = 0.021) and with poorer outcomes, which suggests a confusion bias between age at onset and outcome. Finally, the delays to treatments were pooled, whereas we analyzed them individually. Considering the rapid distribution of intravenous methylprednisolone and its short half-life of 0.25 h [31], infusions could be planned at return of PE cycles sessions or every other day between PE cycles to avoid its clearance through apheresis. Thus, early initiation of PE in combination with corticosteroids seems to be a valuable therapeutic strategy for the first NMOSD attack but should be assessed more deeply in a clinical trial. The early adjunction of PE to treat subsequent attacks is more a matter of discussion given the better prognosis of NMOSD–MOG+ and the potential contributory effect of disease modifying treatments.

This retrospective cohort study has limitations inherent in its design, including the retrospective EDSS scoring, when the EDSS score was not available prospectively, indications bias in the choice of attack treatments and confusion bias. However, by selecting treatment-naïve patients, we avoided confusion with the anti-inflammatory effect of long-term immunotherapy treatments. By selecting the first attack or the next in an independent location, we avoided the risk of confusion with pre-existing disability. Contrary to recent randomized clinical trials, the attacks were physician-defined rather than adjudicated [32‐34]. Strict attack definitions with severity thresholds and adjudications are better suited to prospective time-to-event studies. However, most of these definitions are based on the functional system scores or MRI data, which were the modalities used in our study. Physician-defined attacks are more representative of real-world practice and better suited to retrospective studies. In addition, as in phase 3 clinical trials, the diagnosis and the clinical course of each patient was reviewed by a panel of NMOSD experts belonging to the NOMADMUS scientific committee. The EDSS score has some limitations in assessing disability in NMOSD. It was designed to evaluate disability in patients with MS. It is non-linear and the dominance of the motor function may mask impairments of other systems when EDSS score ≥ 5.5. Unfortunately, there is no better assessment of disability available today for NMOSD and this approach is still used in phase 3 trials. The stratification into three outcome classes attenuated classification bias related to the retrospective EDSS scoring. The risk of potential misclassification was low according to the relative definition, likely to reduce the whole recovery class, and moderate according to the absolute definition, which was conversely sensitive to slight changes in EDSS score. This stratification aimed to evaluate more precisely the functional improvement in the “partial recovery” class (EDSS score at 6 months between the EDSS score at the nadir of the attack and a complete return to the previous state). The absolute definition is intended to create a more relevant evaluation tool to assess the treatment response. Finally, the early use of recently approved treatments, such as eculizumab, satralizumab and inebilizumab [32‐34], could not be assessed as the disease onset in our population was their investigation and approval. The relationship between the recovery of the first attack and the long-term prognosis is another question of interest.

This national collaborative study sheds light on the therapeutic management of NMOSD presentations. We identified two prognostic factors: serostatus (with better outcomes among MOG-Ab-positive patients) and the delay to PE. We, therefore, argue in favor of more aggressive anti-inflammatory management, requiring early combination of PE with corticosteroids. Furthermore, the lower severity and better treatment response of NMOSD–MOG+ support its nosologic reclassification as MOGAD.