Evaluation of diagnostic criteria and red flags of myelin oligodendrocyte glycoprotein encephalomyelitis in a clinical routine cohort

Abstract Aims Myelin oligodendrocyte glycoprotein antibodies (MOG‐IgG) have been proposed to define “MOG encephalomyelitis” (MOG‐EM), with published diagnostic and “red flag” criteria. We aimed to evaluate these criteria in a routine clinical setting. Methods We retrospectively analyzed patients with borderline/positive MOG‐IgG and applied the diagnostic and red flag criteria to determine likelihood of MOG‐EM diagnosis. Para‐/clinical parameters were described and analyzed with chi‐square test. Results In total, 37 patients fulfilled MOG‐EM diagnostic criteria (female‐to‐male ratio: 1.6:1, median onset age: 28.0 years [IQR 18.5‐40.5], n = 8 with pediatric onset). In 24/37, red flags were present, predominantly MOG‐IgG at assay cutoff and/or MRI lesions suggestive of multiple sclerosis (MS). As proposed in the consensus criteria, these patients should rather be described as “possible” MOG‐EM. Of these, we classified 13 patients as “unlikely” MOG‐EM in the presence of the red flag “borderline MOG‐IgG” with negative MOG‐IgG retest or coincidence of ≥1 additional red flag. This group mainly consisted of patients diagnosed with MS (n = 11). Frequency of cerebrospinal fluid (CSF‐)—specific oligoclonal bands (OCB) is significantly lower in definite vs possible and unlikely MOG‐EM (P = .0005). Conclusion Evaluation of diagnostic and red flag criteria, MOG‐IgG retesting (incl. change of assay), and CSF‐specific OCB are relevant in clinical routine cohorts to differentiate MOG‐EM from MS.


| INTRODUC TI ON
The association of immunoglobulin G (IgG) autoantibodies directed against myelin oligodendrocyte glycoprotein (MOG) to different disease entities has been discussed 1,2 : MOG-IgG was initially described in multiple sclerosis (MS), neuromyelitis optica spectrum disorders (NMOSD), relapsing inflammatory optic neuritis (RION), acute demyelinating encephalomyelitis (ADEM), or limbic encephalitis. [3][4][5][6][7][8][9] Yet, disorders associated with MOG-IgG have recently been proposed to form their own disease entity, MOG encephalomyelitis (MOG-EM). 2,10 Diagnostic criteria and a list of atypical conditions for MOG-EM ("red flags") have been suggested. 2 However, these criteria have not yet been validated, and MOG-EM is associated with considerable phenotypic variability among different age groups. 7,9,[11][12][13][14][15][16] Differentiation of MOG-EM from other inflammatory central nervous system (CNS) disorders, especially the most common one MS, is crucial as it harbors therapeutic implications. In contrast to MS, treatment (and even necessity of long-term treatment) of MOG-EM is thus far unclear as there are no established predictive markers to differentiate monophasic vs relapsing disease.
Historically, MOG-IgG has been described as a potential part of MS pathophysiology 6 demonstrating histopathological similarities between MOG-EM and MS. 17 Differentiation between the two diseases remains challenging in routine clinical practice due to overlapping features and test-inherent issues such as false-positive findings. 2,18,19 We retrospectively applied the proposed diagnostic criteria to a monocentric cohort of patients with borderline and positive results for MOG-IgG as detected in clinical practice in order to critically evaluate the likelihood of MOG-EM vs MS.

| Study design
In this retrospective study, clinical records of patients of the Dept.  Table 2), cerebrospinal fluid (CSF), and MRI results. Results of CSF oligoclonal band (OCB) analysis were obtained from routine diagnostics (n = 36, isoelectric focusing, silver staining).

| Autoantibody testing
The presence of MOG-IgG in the serum of all patients was assessed using two different cell-based assays: (a) until 04/2017, a FACS assay was performed at the University Hospital Basel, Prof.
Derfuss, Basel, Switzerland, abbreviated as FACS. These samples were sent overnight and further processed according to the published protocol. 5 borderline, >1:10 positive). In order to assess the qualitative robustness of the two different methods, only a small number of samples (n = 6) taken at the same date were tested in both assays, n = 3 negative in both assays, n = 2 positive in both assays ( All test results are summarized for all patients in Table 2. MOG-IgG was available for each patient from 1 up to 4 times. All other autoantibodies (ANA, p-and c-ANCA, and paraneoplastic antibodies) were measured in the clinical routine laboratory.

| MRI
Cerebral and spinal MRI scans were performed with in-house standard protocols for demyelinating disorders: MR images were acquired on 3 Tesla (T) and 1.5T MR scanners (Magnetom Verio 3T; Magnetom Trio 3T; Magnetom Avanto 1.5T and Magnetom 1.5T Aera; Siemens Healthcare) with the standardized MS protocol. Generally, for follow-up scans, the same scanner was used as in the first MRI scan. All images were routinely evaluated by trained neuroradiologists. For this analysis, all scans were manually re-evaluated by the investigators for the presence of MRI red flags as defined in Table 1.

MOG-EM and "red flags"
The recently published criteria and proposed "red flags" 2 were evaluated for the initial cohort of n = 40 patients. The red flags "bi-/trispecific MRZ reaction," "MOG-IgM/-IgA with negative MOG-IgG," and "MOG-IgG in CSF, but not in serum" could not be evaluated as these are not part of the routine clinical workup in our center (Table 1).
For patients fulfilling the diagnostic criteria, we assessed the presence of red flags and thus defined three groups: • Definite MOG-EM (as proposed by 2 ) • Possible MOG-EM (as proposed by 2 ) • Unlikely MOG-EM: presence of the red flag "MOG-IgG at assay cutoff" plus a negative retest for MOG-IgG or at least one additional red flag.

| Analysis and statistics
Descriptive statistics were used to summarize patient data. For categorical variables, absolute and relative frequencies are reported. For continuous variables, median and interquartile range (IQR), if deemed informative, also minimum (min.) and maximum (max.) are given. Frequencies of missing data are disclosed for each parameter. Group comparisons were performed with Kruskal-Wallis test as a nonparametric test. Chi-square test was used to calculate difference in OCB and clinical diagnosis distribution between MOG-EM groups. Statistical significance level was set to P < .05.
All analyses were performed with GraphPad Prism version 7.03 for Windows, GraphPad Software.

| Ethics approval
This study was approved by the responsible ethics committee (cantonal ethics committee Bern, registration no. KEK-BE 2017-01369).
For this retrospective analysis, pseudonymized patient data were included. A separate informed consent was waived by the committee. For patients seen after the introduction of the general consent (Feb-2015), the presence of the patients' general consent was checked before inclusion in the analysis. Pediatric patients and their legal representatives were individually asked for their consent to the pseudonymized use of data by their treating physician (SB), as these are not covered by the general consent.

| Characteristics of the cohort
We identified n = 40 patients with borderline or positive MOG-IgG. Of this initial group, n = 3 patients did not demonstrate clinical findings compatible with a demyelinating event (functional disorder, episodic vertigo, depression, each n = 1). Two of these three had normal brain MRI scans and were tested positive in FACS with no retest available and a negative retest in EUROIMMUN, respectively. For the third patient tested borderline in EUROIMMUN, neither a retest, nor further clinical or paraclinical data were available. These three presumably false-positive/borderline patients not fulfilling the diagnostic criteria were excluded from further analyses.
The remaining 37 patients fulfilling the diagnostic criteria were included in the analyses.
The syndrome at first manifestation was mainly spinal cord or optic nerve involvement ( Figure 3B).

| Evaluation of "red flags" and classification of likelihood of MOG-EM
A large proportion of patients (24/37) demonstrated at least one red flag (Table 2) and should thus be defined as "possible" MOG-EM. 2 Most frequently, the MRI red flag "lesions suggestive of MS" was present (n = 18/37). In 16/37 patients, MOG-IgG was detected at the assay cutoff. Progressive disease course was described in 5/37 patients. The MRI red flag "silent increase in lesion burden" was noted for 1/37 patient.
We classified one patient (pat. 26, Table 2  Re-evaluating the cohort characteristics per diagnosis likelihood groups, the female-to-male ratio shifts for "definite" and "possible" MOG-EM close to 1:1 ( Figure 1). Age at onset is not significantly dif-  For classification by EDSS functional system scores ( Figure 3A) and by syndrome ( Figure 3B) at onset, "definite" MOG-EM seems prone to involvement of the optic nerve (8 of 13) in our cohort. Due to the various groups of small size, no formal statistical analysis was run over this part.  Table 2). No ADEM manifestations were present in our incidental cohort.
TA B L E 2 Characteristics of the cohort and red flags per patient (as numbered in Table 1) with classification of patients in MOG-EM categories "definite," "possible," and "unlikely"
A patient with an atypical clinical manifestation (classified as autoimmune encephalomyelitis with reflex-myoclonus, classified "unlikely" MOG-EM) was extensively re-evaluated at the age of 67 years. A MOG-IgG titer of 1:10 was detected, combined with a low-titer SOX-1 antibody. A large panel of other autoantibodies in this patient was negative, including GAD and IA-1 antibodies (coexisting Diabetes mellitus). Due to repeated treatment with intravenous immunoglobulins after this first testing, retesting was not performed. This male patient experienced first symptoms at the age of 29 years. A neoplastic comorbidity was not found.

| Para-/postinfectious disease onset
In our cohort, n = 4 patients demonstrated a para-or postinfectious disease onset ( Table 2, "definite" [n = 3] and "possible" Serum available from the acute EHEC sepsis phase was retrospectively tested and negative for MOG-IgG.

| Evaluation of CSF-specific oligoclonal bands in MOG-EM likelihood groups
Results of OCB testing were available for 36 patients (Table 2).
Altogether, type 2 OCB (CSF-specific) were documented in 23 patients of the whole cohort, type 1 OCB (polyclonal) in 10 and type 4 OCB (identical bands in CSF and serum) in 3 patients, respectively.
As described before, only type 2 OCB (and type 3 which did not occur in our cohort) are CSF-specific, whereas type 1 and type 4 OCB do not represent local IgG synthesis. 2,22 The frequency distribution of the presence of CSF-specific OCB vs absence of CSF-specific OCB is significantly different between the groups "definite," "possible," and "unlikely" (P = .0005, Table 3).

| CON CLUS ION
MOG encephalomyelitis has recently been proposed to form its own disease entity. 1,2,10 However, both the clinical spectrum and differential diagnosis are thus far not well determined and proposed diagnostic criteria need broader validation which is challenging in a rare condition. In clinical practice, different methods of MOG-IgG testing harbor relevant differences in sensitivity and specificity. 23 Occurrence of low-titer MOG-IgG in patients with MS may cause clinically relevant delays in therapeutic decision making and ultimately represent false-positive findings. 18 We present a monocentric cohort of 37 well-characterized patients with suspected MOG-IgG-associated disorder and retrospectively applied the proposed diagnostic criteria and red flags as well as OCB to differentiate MOG-EM from MS.
Regarding basic demographic and clinical results, our cohort is in line with existing data. 14,15,24 The shift of the female-to-male ratio to 1:1 after exclusion of "unlikely" MOG-EM might rather be related to sample size.
We retrospectively applied the proposed diagnostic criteria and However, given our small cohort, corroboration of our findings is necessary.
It has to be kept in mind that assay-related issues, that is, Single cases of para-/postinfectious disease manifestations of MOG-EM were described before, both in adults and in children. 14,28 In all our patients, signs of systemic infection were still present at the onset of MOG-EM and/or a distinct pathogen was identified.
Postinfectious disease onset is likely to be underdiagnosed, especially if MOG-EM occurs within weeks after a common trivial infection. With the negative MOG-IgG result of the patient during EHEC sepsis, disease manifestation of MOG-EM with RION phenotype 7 weeks later and sequential confirmed positive MOG-IgG, our work may imply a possible underlying pathogenesis that involves triggering a general immune response 29 or molecular mimicry processes. 30 After classification of MOG-EM likelihood groups, patients with coexisting ANA autoantibodies or another autoimmune disease were not present in the "definite" MOG-EM group. As in other cohorts, the small sample size limits wide interpretations. Yet, both ANA and coexistence of other autoimmune conditions have been described to be a potential help in differentiation vs AQP4-IgG positive NMOSD. 31 As we did not compare to an NMOSD cohort, we cannot validate this finding with our data. Detectable (mainly low-titer) autoantibodies including ANA and MOG-IgG may occur within an autoimmune condition that does not qualify for MOG-EM and even in healthy persons; these may thus be unspecific or false-positive findings 32,33 and argue for testing of MOG-IgG only in suspected cases of a demyelinating CNS disorder. 2 In addition to the application of the proposed red flags, we evaluated CSF-specific OCB. As addressed in the diagnostic criteria, the absence of CSF-specific OCB in patients with suspected MS should trigger MOG-IgG testing. 2 In turn, their presence may also be used to challenge the diagnosis of MOG-EM, especially if they occur in combination with the named red flags. This is supported by our and other data. 31 Due to the retrospective nature of our analysis without available retesting of MOG-IgG for all, particularly borderline-tested, patients, we cannot exclude that false-positive findings of MOG-IgG are present in the "possible," and maybe even in the "definite" group.
This situation will frequently occur in clinical practice. In addition, our study has the limitation of a small sample size. Nevertheless, for a rare and recently defined condition, our cohort, albeit small, contributes to a better understanding of MOG-EM and differential di-