A comparative evaluation of different neuromyelitis optica spectrum disorder sets of criteria

Three different sets of criteria have been proposed for the diagnosis of neuromyelitis optica spectrum disorder (NMOSD). The objective was to compare the specificity, sensitivity and diagnostic accuracy of the three different sets of NMOSD criteria in patients presenting with inflammatory disorders of the central nervous system suggestive of NMOSD.


Introduction
Since 1999, three different sets of diagnostic criteria have been proposed, mainly to distinguish neuromyelitis optica (NMO) from multiple sclerosis (MS) and other central nervous system inflammatory disorders. Initially, NMO diagnosis was based essentially on a probabilistic combination of clinical signs, brain or spinal cord magnetic resonance imaging (MRI) and cerebrospinal fluid (CSF) analysis [1]. The identification of aquaporin-4 immunoglobulin G antibody (AQP4-IgG) [2] led to redesigning of the diagnostic criteria in 2006 [3]. In 2015, the International Panel for NMO Diagnosis (IPND) revised the diagnostic criteria for NMOSD [4], taking into account the broadening clinical spectrum. This new set of criteria allowed an earlier diagnosis by the inclusion of positive AQP4-IgG (AQP4-IgG + ) clinically limited forms, and patients with lesions outside the optic nerve and the spinal cord. Analysis of the impact of the IPND criteria showed a dramatic increase of the diagnostic yield, in comparison to 2006 [5][6][7], concerning both AQP4-IgG + and seronegative patients. In addition, the recent identification of myelin oligodendrocyte glycoprotein (MOG) IgG in NMOSD [8] and less frequently in MS [9] questions the link between NMOSD and MOG-IgG associated disease/disorder (MOGAD). No direct comparison between the three sets of NMOSD criteria has been performed to date, and studies evaluating the impact of the 2015 IPND criteria were not designed to assess specificity. Here the diagnostic accuracy and specificity of the three sets of NMOSD criteria were compared.

Study objective
The main objective of this observational retrospective study was to compare the diagnostic value of three different sets of NMOSD criteria, as indexed by specificity, sensitivity and accuracy, in patients with clinical symptoms suggestive of NMOSD.

Study population
The data were evaluated of 756 patients (from Lyon, Paris, Strasbourg) referred to the Lyon laboratory for serum AQP4-IgG testing because of central nervous system inflammatory disorders suspected of being NMOSD (called NMO until 2007) by the treating clinician, between 2012 and 2014. Medical records were available for 520 patients. Incomplete files were excluded from the analysis. Finally, 236 files were evaluated for which all the clinical, imaging and biological information necessary to evaluate the three sets of NMOSD criteria were available. For each patient, the presence of the following core clinical characteristics was determined: optic neuritis, myelitis, area postrema syndrome, brainstem syndrome, diencephalic and encephalic syndrome. Brain, spinal cord and orbital MRI performed for diagnosis purposes during the attacks were reviewed when available, as well as CSF analyses. All samples were tested for AQP4-IgG using the same technique based on live-cell-based assay.

Data collection
Clinical, imaging and biological data were collected centre by centre in 2017 and reviewed concomitantly by three neurologists, experts in NMOSD, from the three participating referral centres. The treating neurologist was defined as the neurologist of the centre evaluated and the two evaluating neurologists were defined as external neurologists at the evaluated centre.
Blind to the diagnosis proposed by the treating neurologist, the two evaluating neurologists, fully agreed, applied retrospectively the 1999, 2006 NMO criteria and 2015 IPND NMOSD criteria [1,3,4]. The final diagnosis was based in 2017, on the conclusion of the treating neurologist.

Statistical analysis
Statistical analyses were performed with R Core Team 2019 (3.6.0 version). The performance of the NMOSD criteria with regard to clinical diagnosis was expressed as sensitivity, specificity, positive predictive value, negative predictive value and accuracy.

Ethical concerns
According to the French legislation regarding observational studies, patients were informed of the study but did not have to provide a formal and written informed consent to have their anonymized data collected. Data confidentiality and safety were ensured in keeping with the recommendations of the French Commission Nationale Informatique et Libert es, which also provided approval.

Results
The clinical, imaging, CSF characteristics and AQP4-IgG sero-status of the 236 files analysed are detailed in Table 1. The final diagnosis was NMOSD in 66 cases, MS in 85 cases. Another diagnosis was made for the 85 other cases ( Table 2).
Of the 236 files analysed, 76 patients fulfilled at least one set of NMOSD criteria: 37, 38 and 70 respectively for the 1999, 2006 and 2015 NMOSD criteria. Forty-seven patients fulfilled only one set of criteria: 7, 10 and 30 respectively for the 1999, 2006 and 2015 NMOSD criteria. Twenty-eight patients fulfilled all NMOSD criteria (Fig. 1).   (n = 3), 'acute demyelinating encephalomyelitis like' lesions (n = 1), brainstem abnormalities (n = 1) and area postrema abnormalities (n = 2). Eight patients were tested positive for AQP4-IgG. The diagnosis of NMOSD was made in all of these 10 cases (Table 3). Nine patients fulfilled the 1999 but not the 2006 NMOSD criteria. All patients fulfilling the 1999 criteria but not the 2006 criteria met the major criteria (optic neuritis and acute myelitis and no evidence of clinical disease outside of optic neuritis and spinal cord). As supportive criteria of the 1999 NMOSD criteria, seven had a normal brain MRI, one had a longitudinal extensive transverse myelitis on spinal cord MRI and confluent bilateral subcortical and deep white matter lesions and one had severe and bilateral optic neuritis. None was tested positive for AQP4-IgG. The final diagnosis selected was NMO (n = 4), Epstein-Barr virus transverse myelitis and optic neuritis (n = 1) and optico-spinal MS (n = 4). Amongst these four patients with a final diagnosis of NMO, two fulfilled also the 2015 IPND criteria. The two patients with NMO and fulfilling only the 1999 criteria were tested negative for AQP4-IgG. They had bilateral optic neuritis with visual acuity worse than 20/200 (two minor supportive 1999 criteria) and short myelitis without other symptoms. Brain MRI fulfilled MacDonald 2010 criteria for one. Thirty patients met the 2015 IPND criteria but not the 2006 or 1999 NMOSD criteria. Twenty-three had a limited form of NMOSD with only optic neuritis (n = 4) or longitudinal extensive transverse myelitis (n = 19). Four cases also had a brainstem (n = 3) or area postrema (n = 4) syndrome. All limited forms were tested positive for AQP4-IgG. Five of the seven patients with negative AQP4-IgG had at least clinical and radiological area postrema and/or brainstem syndrome. In one case the MRI lesion extended from the cervical cord to the area postrema (Fig. 2a, patient 52) and in one other case a tumefactive brain lesion was observed in addition to brainstem lesion (Fig. 2b, c, patient 48).

Comparison between 1999 and 2006 NMOSD criteria
Amongst these 30 patients, five (16.6%) had a final diagnosis other than NMOSD. One had extended brain lesions and longitudinal extensive transverse myelitis with a final diagnosis of endovascular lymphoma (Fig. 3). Two patients were considered as MS fulfilling MacDonald 2010 criteria. One patient testing positive for AQP4 with extensive myelitis had a final diagnosis of glioma of the spinal cord. Another patient testing positive for AQP4-IgG with optic neuritis was diagnosed as paraneoplastic syndrome. The underlying tumour was squamous cell lung carcinoma associated with Hu-antibody associated encephalitis. The remaining 25 patients had a final diagnosis of NMOSD, 22 of whom were positive for AQP4-IgG.

Final diagnosis
Amongst the 66 cases of NMOSD, 38 were positive for AQP4-IgG and three for MOG-IgG (17/28 AQP4-IgG negative patients tested).  Of the 85 cases with a final diagnosis of MS, seven patients met at least one set of NMOSD criteria, including one patient who met all three sets of criteria. One patient with progressive MS was tested positive for AQP4-IgG.
Amongst the 85 cases without NMOSD or MS diagnosis (Table 2), six patients with isolated myelitis were tested positive for MOG-IgG. Finally, 34 patients still remained without diagnosis at the date of the evaluation: 16 had isolated or recurrent idiopathic optic neuritis, 15 isolated or recurrent idiopathic transverse myelitis, one isolated encephalitis and two multiple symptoms.

Value of the three different sets of NMOSD criteria
The specificity of the criteria was highest (99%) in the 2006 NMO criteria. The specificity was similar in the first and the last set of criteria (96% in 1999 and in 2015 IPND criteria). The sensitivity was highest (97%) in the 2015 IPND NMOSD criteria and the accuracy significantly increased from the first to the last sets of NMOSD criteria (from 82% to 97%) ( Table 3).

Discussion
The accuracy of the NMO sets of criteria increases with the successive version of the criteria with a significant improvement of the sensitivity from 1999 to 2015 IPND criteria whilst the specificity remains similar.
The results of this study emphasize the limitation of the 1999 NMO criteria. The 1999 set of criteria allowed the inclusion of the optico-spinal form of MS, having therefore a low specificity. The integration of serological status for AQP4-IgG [2] in the revised 2006 criteria led the specificity of the NMO set of criteria to be improved drastically, allowing cases with extra optico-spinal presentation to be included, and also improved the sensitivity and enlarged the spectrum of the disease.
Our study confirms the increase of diagnostic yields of the 2015 IPND criteria in comparison to 2006, due to the inclusion of positive AQP4-IgG patients with a limited form of the disease. The 2015 IPND criteria also result in the inclusion of patients with either positive or negative AQP4 status, presenting with non optico-spinal syndrome (brain, brainstem or diencephalic lesions) in both. The IPND 2015 criteria also lead to the inclusion of patients with short myelitis [12] even though longitudinally extensive myelitis involving the thoracic cord and grey matter (H-shaped cord lesion) remains a key point in identifying NMOSD [13].
The study also raises some concerns about the 2015 IPND criteria with a lower specificity in comparison to 2006. Indeed, some patients fulfilling 2015 IPND criteria and not 2006 NMOSD criteria had an alternative NMOSD diagnosis. This issue concerns mainly patients with brain lesion and negative AQP4-IgG status. However, also two cases were found where AQP4-IgG seropositive status was eventually associated with an alternative diagnosis. It includes one spinal glioma. Our hypothesis was that the glioma (astrocytoma) might have induced the induction of autoimmunity against AQP4 by expressing an abnormal amount of AQP4, and AQP4-IgG is then only a secondary epiphenomenon. Indeed, AQP4 is upregulated and redistributed in human glioma [14]. Unfortunately, immunohistological staining of the malignant tissue for AQP4 was not performed. Such association between astrocytoma and the presence of AQP4-IgG has been previously described [14]. The second case was associated to paraneoplastic Sox 1 and Hu encephalitis secondary to small cell neuroendocrine bronchial cancer (with contingent adenocarcinoma) cTxN2M0. To our knowledge, such association of anti-Hu and anti-AQP4 has not been previously described. One can hypothesize a co-auto-immunity or a paraneoplastic AQP4-IgG production as already reported in some few cases, in particular with lung adenocarcinoma [15].
Moreover, if the 2015 IPND criteria are more accurate than the 1999 criteria to distinguish NMOSD from MS there is still a matter of concern in patients with 'MS like' lesions. The appearance of 'MS like' brain lesions during the evolution of the disease is particularly worrying in the limited form.
Despite the fact that the IPND 2015 allowed an early diagnosis, some patients with NMO related disorder, with either isolated or recurrent transverse myelitis and optic neuritis but testing negative for AQP4-IgG, are still not classified. As found in our study, some of these patients tested positive for MOG-IgG, highlighting the current need to establish a clear definition of MOGAD and its link with NMOSD.

Disclosure of conflicts of interest
The authors declare that they have no competing interests, no relevant disclosure to the paper and no funding statement to declare for this work.

Data availability statement
Because of national confidentiality requirements, only anonymized data, not pseudonymized data, can be shared. Whilst anonymization techniques might result in the impoverishment of data (article 29 of Directive 95/46/EC, Opinion 04/2014 on Anonymization Techniques À0829/14/EN WP 216) data used for this study were only pseudonymized.