Ненормальный МРТ мозга в энцефалите против NMDA: клинические и прогнозистические последствия.

Introduction

Anti-N-methyl- d -aspartate encephalitis (NMDARE) is the most frequent autoimmune encephalitis; it predominantly affects young women and children and leads to long-term neurologic sequelae in 20%–30% of the patients. 1-4 Although complex, NMDARE is a highly recognizable syndrome composed of 6 typical symptoms comprising cognitive, psychiatric, and behavioral manifestations; seizures; speech impairment; movement disorders; decreased consciousness; and autonomic dysfunction. 1-4 These symptoms are occasionally associated with others, defined as nontypical symptoms, such as cerebellar ataxia or motor impairments. 2,5 Brain MRI can show T2-weighted hyperintensities in mesial temporal lobe structures including the hippocampus and amygdala, suggesting limbic inflammation, in approximately 25% of the patients. 1,6,7 In addition, several studies found in 23–30% of the cases the involvement of extralimbic regions, such as the cortical or subcortical regions, basal ganglia, brainstem, or cerebellum; the morphology and distribution of these lesions, as well as their association with the clinical symptoms, are not well described. 2,5-12 Of note, 3%–4% of NMDARE cases also have clinical and radiologic features of demyelinating diseases such as multiple sclerosis (MS), neuromyelitis optica spectrum disorder (NMOSD), or myelin oligodendrocyte glycoprotein–associated disorder (MOGAD). However, these overlapping demyelinating syndromes cannot explain all the extralimbic lesions. 5,13-17 At baseline, abnormal brain MRI (all abnormalities taken into account) is associated with poor clinical outcomes and is a criterion of the anti-NMDAR Encephalitis One-Year Functional Status (NEOS) score that predicts poor outcomes. 7,17,18 It is still unclear whether all types of brain MRI lesions are equally important to predict NMDARE prognosis. However, the recognition of lesions on baseline MRI associated with poor outcomes could be important to the therapeutic decision making, especially regarding escalation to second-line therapy. 19 In this study, we aimed to characterize the lesions on brain MRI associated with NMDARE through a retrospective, nationwide cohort of patients with NMDARE. In addition, we also aimed to assess the association of extralimbic lesions with (1) the diagnosis of overlapping demyelinating syndromes, (2) the prevalence of typical and nontypical symptoms of NMDARE, and (3) the clinical recovery at 2 years.

Methods

Patients

This study included all adult and pediatric patients diagnosed with NMDARE at the French Paraneoplastic Neurologic Syndrome Reference Center, between October 2007 and December 2021, and who had available MRI data. A minimal 1-year follow-up after symptom onset was required to prevent selection bias interference in the diagnosis of overlapping demyelinating syndromes. Patients with NMDARE secondary to herpes simplex virus encephalitis were excluded ( Figure 1 ). Diagnosis of NMDARE was based on the association of compatible clinical symptoms 3 with presence of IgG antibodies against the NMDAR NR1 subunit in the CSF, detected using rat brain immunochemistry and cell-based binding assay, as previously reported. 1,20 Data were retrospectively collected from patient medical records. Neurologic symptoms were classified as typical or nontypical of NMDARE, based on authoritative publications. 1,3,4 MRI abnormalities at diagnosis were extracted from the radiologic reports and/or medical charts. In case of extralimbic abnormalities, MRI scans were then collected from the referral clinician. The images were then reviewed by 2 investigators (L.K., J.P.); 1 was blinded to the clinical data (J.P.) to characterize and classify the extralimbic lesions. In case of discrepancy between the 2 investigators, an expert radiologist (F.C.), blinded to the clinical data, was consulted and his opinion retained. When the images were not available, the lesions were classified according to the brain MRI radiologic report. Hyperintensities of the corticospinal tract and other lesions considered as noninflammatory (comprising lesions compatible with chronic stroke, post-traumatic lesions, seizure-related lesions, cerebral small vessel disease lesions, or cytotoxic lesions of the corpus callosum) were not included in the analysis. All available serum samples were tested for anti-myelin oligodendrocyte glycoprotein IgG antibodies (MOG-Abs) and anti-aquaporin 4 IgG antibodies (AQP4-Abs), as previously described. 21,22 For MOG-Abs, only clear positive results (titer ≥ 1:640) were considered; 23 international guidelines were used to identify patients with overlapping demyelinating syndromes such as MS, NMOSD, and MOGAD. 24–26 Longitudinal outcomes were assessed retrospectively using the modified Rankin Scale (mRS). 27

Flowchart of the Study

NMDARE = anti-N-methyl-d-aspartate receptor encephalitis.

Standard Protocol Approvals, Registrations, and Patient Consents

This study was approved by the institutional review board of the Hospices Civils de Lyon. According to the French law, written informed consent was not necessary because of the retrospective and noninterventional design of the study; however, we provided a notice of nonopposition to each patient. Biological samples were stored in the NeuroBioTec Hospices Civils de Lyon BRC (France, AC-208-73 NFS96-900).

Data Availability

Any unpublished data within the article are available and will be shared anonymously on reasonable request from all qualified investigators.

Statistical Analysis

Continuous variables are expressed as median and range and discrete variables as count and percentage. For the analysis of nontypical symptoms, patients without inflammatory extralimbic lesions on MRI were used as a control group and compared with patients with extralimbic lesions. Owing to skewed distributions of the continuous data, medians were compared using the Wilcoxon-Mann-Whitney test. Frequencies were compared using the Fisher exact test. Recovery was studied 2 years after symptom onset, the time point at which most patients with NMDARE have stopped improving. 4 To study the parameters associated with recovery, the outcome was dichotomized into complete or near-complete recovery (mRS scores 0–1) and nonrecovery (mRS score 2 or higher) at 2 years after symptom onset. A multivariable logistic regression model was built to include the different types of brain MRI lesions identified. All variables found associated with clinical recovery at 2 years in the univariable model were also included, and those not contributing to the model were then eliminated in a stepwise fashion. All continuous variables with skewed distribution were dichotomized based on the median value found in patients with mRS scores 0–1 at 2 years; age was dichotomized at 45 years because of a bimodal distribution (eFigure 1). Multiple imputation was used to reduce bias because of missing data (3.4% of the total data set). Odds ratios and 95% confidence intervals were reported, and all p values were 2-tailed. p values <0.05 were considered statistically significant. Statistical analyses were performed using RStudio, version 3.4.0 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Patients

A total of 255 patients were included in this study; 201 of 255 (78.8%) were female, and the median age was 20 years (range 1–80). Twenty-one patients (8.2%) were older than 45 years. A teratoma was identified in 55 of 255 patients (21.6%) and a malignant tumor in 12 of 255 patients (4.7%; including 5/12 neuroendocrine carcinomas). Most of the patients with a malignant tumor (11/12) were older than 45 years (eFigure 1). Among the 183 tested patients, 5 (2.7%) were positive for anti-MOG (4 patients) or anti-AQP4 (1 patient) antibodies; in total, 10 of 255 patients (3.9%; median age 35.5 years, 3 men) with overlapping demyelinating syndromes were identified using the criteria from international guidelines for MS (4 patients), MOGAD (4 patients), and NMOSD (2 patients; eTable 1). In 7 of 10 patients with overlapping demyelination (70%), the first demyelinating event was concomitant with NMDARE onset; among them, 3 subsequently had recurrent demyelinating events. By contrast, 3 of 10 patients had their first demyelinating event from 4 to 16 years before NMDARE onset, and among them, 1 had recurrent demyelinating events after NMDARE onset.

Brain MRI Findings

Brain MRI was performed within 3 months after symptom onset in most patients (242/255; 95%) and between 3 months and 1 year in 10 of 255 patients (3.9%), for whom the diagnosis was not immediately suspected. In 3 of 255 patients (1.2%), brain MRI was delayed because of an initial misdiagnosis (pure psychiatric presentation) and was performed more than 1 year after symptom onset, during a relapse. The brain MRI was performed before any immunologic treatment in 248 of 255 patients (97.2%), and the median number of brain MRI scans per patient during the study period was 2 (range 1–16). Brain MRI scans showed lesions consistent with inflammation in 78 of 255 patients (30.6%), including T2/fluid-attenuated inversion recovery (FLAIR) hyperintensities in the limbic regions (hippocampus, amygdala) in 56 of 78 patients (71.9%), 18 of whom were bilateral (32.1%), and in extralimbic regions in 41 of 78 patients (52.6%). There was no difference between the frequency of extralimbic lesions in the reference center of Lyon and its surroundings (6/34 patients, 17.6%) and in the rest of France (35/221 patients, 15.8%; p = 0.8). Moreover, there was also no difference regarding the frequency of extralimbic lesions in MRI scans performed more than 3 months after symptom onset (3/13, 23.1%) compared with the MRI scans performed earlier (38/242, 15.7%; p = 0.48). Among the 41 patients with extralimbic lesions, 19 (46.3%) also had limbic lesions ( Figure 1 ). The brain MRI files could be reviewed in 28 of 41 patients (68.3%) with extralimbic lesions; in the remaining 13 patients (31.7%), the lesions were classified according to the radiologic report. Three mutually exclusive patterns of extralimbic lesions were identified: MS-like lesions (14/41, 34.1%); extensive FLAIR lesions (5/41, 12.2%); and poorly demarcated, fluffy lesions (22/41, 53.7%). The 214 of 255 patients (83.9%) without extralimbic lesions (eTable 2) composed the control group used to study the prevalence of symptoms nontypical for NMDARE ( Table 1 ). Compared with controls, patients with extralimbic lesions more frequently presented with nontypical symptoms (23/41, 56.1%, vs 22/214, 10.3%; p < 0.001), including cerebellar ataxia (17/41, 41.5%, vs 13/214, 6.1%; p < 0.001), motor impairment (12/41, 29.3%, vs 11/241, 5.1%; p < 0.001), optic neuritis (3/41, 7.3%, vs 1/214, 0.5%; p = 0.01), and ocular (7/41, 17.1%, vs 5/214, 2.3%; p = 0.001) and other cranial nerve palsies (8/41, 19.5%, vs 10/214, 4.7%; p = 0.003).

Table 1 Clinical Features and Diagnostic Tests of the Patients According to the Patterns of Extralimbic Lesions No extralimbic lesion N = 214 Fluffy lesions N = 22 p Value Extensive FLAIR lesions N = 5 p Value MS-like lesions N = 14 p Value Male, n (%) 42 (19.6) 6 (27.3) 0.407 1 (20) 1.000 5 (35.7) 0.172 Age, range 20 (1–80) 20 (3–70) 0.863 20 (4–65) 0.756 22.5 (5–64) 0.363 Teratoma, n (%) 51 (23.8) 3 (13.6) 0.424 0 0.593 1 (7.1) 0.199 Overlapping demyelinating syndromes, n (%) 0 0 — 3 (60) — 7 (50) —  Multiple sclerosis — — — 0 — 4 (28.6) —  MOGAD — — — 1 (20) — 3 (21.4) —  NMOSD — — — 2 (40) — 0 (0) — Nontypical NMDARE symptoms, n (%) 22 (10.3) 11 (50) <0.001 4 (80) 0.001 8 (57.1) <0.001  Oculomotor disorders 5 (2.3) 3 (13.6) 0.029 1 (20) 0.131 3 (21.4) 0.009  Optic neuritis 0 1 (4.5) 0.093 1 (20) 0.023 1 (7.1) 0.0614  Other cranial nerve palsies 10 (4.7) 2 (9.1) 0.310 2 (40) 0.036 2 (14.3) 0.162  Cerebellar ataxia 13 (6.1) 9 (40.9) <0.001 2 (40) 0.039 6 (42.9) <0.001  Motor impairment 11 (5.1) 4 (18.2) 0.039 3 (60) 0.002 5 (35.7) 0.001  Sensory impairment 8 (3.7) 2 (9.1) 0.236 0 1.000 3 (21.4) 0.023  Myelitis a 1 (0.5) 1 (4.5) 0.179 1 (20) 0.045 3 (21.4) 0.001 Typical NMDARE symptoms, n (%) 214 (100) 22 (100) — 5 (100) — 14 (100) —  Typical symptom count, median (range) 4 (1, 6) 4 (2, 6) 0.969 4 (1, 5) 0.579 2.5 (1, 5) 0.003  Behavior disorder 206 (96.3) 22 (100) 1.000 4 (80) 0.191 13 (92.9) 0.441  Cognitive impairment 197 (92.1) 21 (95.5) 1.000 4 (80) 0.351 14 (100) 0.607  Seizures 184 (86) 19 (86.4) 1.000 4 (80) 0.537 6 (42.9) <0.001  Aphasia 150 (70.1) 15 (68.2) 0.812 3 (60) 0.639 8 (57.1) 0.371  Movement disorder 167 (78) 19 (86.4) 0.583 3 (60) 0.311 5 (35.7) 0.001  Dysautonomia 111 (52.1) 10 (45.5) 0.656 3 (60) 1.000 5 (35.7) 0.278  Sleep disorder 82 (38.3) 7 (31.8) 0.648 1 (20) 0.652 3 (21.4) 0.262  Decreased consciousness 134 (62.6) 14 (63.6) 1.000 3 (60) 1.000 8 (57.1) 0.778 Hippocampal T2 MRI hyperintensity, n (%) 37 (17.3) 11 (50) 0.001 4 (80) 0.005 4 (28.6) 0.287 CSF analysis, n (%)  WBC count >5/µL 172/199 (86.4) 15/21 (71.4) 0.100 5 (100) 1.000 9/13 (69.2) 0.103  Protein level >60 mg/dL 22/188 (11.7) 2/22 (9.1) 1.000 3/4 (75.0) 0.007 2/12 (16.7) 0.641  Oligoclonal bands 44/58 (75.9) 2/5 (40) 0.117 2/3 (66.7) 1.000 7/8 (87.5) 0.671 mRS scores 4–5 at nadir, n (%) 186 (86.9) 20 (90.9) 1.000 5 (100) 1.000 11 (78.6) 0.414 Admission in ICU, n (%) 144/212 (67.9) 16/21 (76.2) 0.622 4 (80) 1.000 9 (64.3) 0.774 First-line treatment b , n (%) 210/211 (99.5) 21/21 (100) 1.000 5 (100) 1.000 14 (100) 1.000 Second-line treatment b , n (%) 153 (71.5) 14 (63.6) 0.599 5 (100) 0.368 12 (85.7) 0.399 Steroid-sparing agents b , n (%) 75 (35) 2 (9.1) 0.025 1 (20) 0.823 6 (42.9) 0.762 mRS scores 0–1 at 2 y, n (%) 124/162 (76.5) 13/19 (68.4) 0.410 1/4 (25) 0.047 5/12 (41.7) 0.014 Relapse, n (%) 37 (17.3) 4 (18.2) 1.000 1/5 (20) 0.538 4 (28.6) 0.287 Death, n (%) 13 (6.1) 3 (13.6) 0.180 2 (40) 0.040 1 (7.1) 0.603 Abbreviations: FLAIR = fluid-attenuated inversion recovery; MOGAD = myelin oligodendrocyte glycoprotein antibody–associated disorder; mRS = modified Rankin Scale; MS = multiple sclerosis; NMDARE = anti-N-methyl- d -aspartate receptor encephalitis; NMOSD = neuromyelitis optica spectrum disorder; WBC = white blood cell. p Value: comparison with the control group (214 patients with no extralimbic lesions). a Spine MRI performed in 16 (6.3%) of 255 patients. b First-line treatments included steroids (208 patients), IV immunoglobulins (230 patients), and plasma exchanges (64 patients), and second-line treatments included rituximab (176 patients), cyclophosphamide (61 patients), tocilizumab (3 patients), bortezomib (1 patient), and mitoxantrone (1 patient). Steroid-sparing agents included azathioprine (28 patients), methotrexate (12 patients), and mycophenolate mofetil (49 patients).

Clinical Features and Diagnostic Tests of the Patients According to the Patterns of Extralimbic Lesions

Abbreviations: FLAIR = fluid-attenuated inversion recovery; MOGAD = myelin oligodendrocyte glycoprotein antibody–associated disorder; mRS = modified Rankin Scale; MS = multiple sclerosis; NMDARE = anti-N-methyl- d -aspartate receptor encephalitis; NMOSD = neuromyelitis optica spectrum disorder; WBC = white blood cell.

p Value: comparison with the control group (214 patients with no extralimbic lesions).

Spine MRI performed in 16 (6.3%) of 255 patients.

First-line treatments included steroids (208 patients), IV immunoglobulins (230 patients), and plasma exchanges (64 patients), and second-line treatments included rituximab (176 patients), cyclophosphamide (61 patients), tocilizumab (3 patients), bortezomib (1 patient), and mitoxantrone (1 patient). Steroid-sparing agents included azathioprine (28 patients), methotrexate (12 patients), and mycophenolate mofetil (49 patients).

MS-Like Lesions

In 14 of 41 patients (34.1%), the extralimbic lesions corresponded to MS-like lesions, consisting in small, well-delineated ovoid lesions distributed in the periventricular, infratentorial, or juxtacortical regions, their axis perpendicular to that of the ventricles, resembling the MRI lesions of MS ( Figure 2, A and B ). Among the 14 patients with MS-like lesions, 7 (50.0%) had a definite demyelinating disorder in addition to NMDARE (3 with MOGAD and 4 with MS, including 4 women); these patients were older than those who did not fulfill the criteria for MS or MOGAD (median age 37 years [range 16–64] vs 17 years [range 5–32]; p = 0.035). Gadolinium enhancement was reported in 4 patients (3 with MOGAD and 1 with MS). During the study period, the MS-like lesion load increased in 4 of 13 patients (30.8%; 2 with MS, 1 with MOGAD, and 1 without demyelinating syndrome); a decrease in the lesion load was, however, reported in 5 of 13 patients (38.5%; 1 with MOGAD and 4 without demyelinating syndrome, Figure 2 ). The patients with MS-like lesions more often presented at least one symptom nontypical of NMDARE (8/14, 57.1%) compared with the control group (22/214, 10.3%; p < 0.001), including cerebellar ataxia (6/14, 42.9%, vs 13/214, 6.1%; p < 0.001), motor impairment (5/14, 35.7%, vs 11/214, 5.1%; p < 0.001), and oculomotor palsy (3/14, 21.4%, vs 5/214, 2.3%; p = 0.001). Three patients (21.4%) had myelitis concomitant with NMDARE (2 with MS, 1 with MOGAD). Compared with the control group, the patients with MS-like lesions had a lower median number of typical NMDARE symptoms (2.5 [range 1–5] vs 4 [range 1–6]; p = 0.003); they less frequently presented epilepsy (6/14, 42.9%, vs 184/214, 86%; p < 0.001) and hyperkinetic movement disorder (5/14, 35.7%, vs 167/214, 78%; p = 0.001). However, in both patients with MS-like lesions and the control group, the proportion of severe NMDARE (mRS scores 4–5 at nadir) was similar (11/14, 78.6%, vs 186/214, 86.9%, respectively; p = 0.414), as was the proportion of oligoclonal bands (7/8, 87.5%, vs 44/58, 75.9%, respectively; p = 0.67). At 2 years of follow-up, only 5 of 12 patients (41.7%) had recovered (mRS scores 0–1), compared with 124 of 162 (76.5%) in the control group ( p = 0.014).

Representative Images of the Extralimbic Lesions in NMDARE

Representative images of MS-like (A and B) and extensive FLAIR (C) lesions in patients with NMDARE overlapping with MS (A.a), myelin oligodendrocyte glycoprotein–associated disorder (MOGAD) (B.a), or NMOSD (C.a), and representative image of a fluffy lesion (D.a). Follow-up shows stable lesion load in the patient with MS (A.b; 6-month interval from A.a) and regression of the lesions in the patient with MOGAD (B.b, 7 months from B.a) and in the patient with NMOSD (C.b; 3 years from C.a), as well as of the fluffy lesion (D.b, 3 months from D.a). FLAIR = fluid-attenuated inversion recovery; MOGAD = myelin oligodendrocyte glycoprotein–associated disorder; NMDARE = anti-N-methyl- d -aspartate receptor encephalitis; NMOSD = neuromyelitis optica spectrum disorder.

Extensive FLAIR Lesions

Extensive FLAIR lesions were observed in 5 of 41 patients (12.2%), consisting in confluent lesions of the cerebral cortex and white matter, on at least 2 cerebral lobes, sometimes involving the basal ganglia, cerebellum, or brainstem ( Figure 2C ). Three patients (60%) with overlapping demyelinating disorders were identified (2 with NMOSD and 1 with MOGAD, all female; aged 4, 20, and 64 years, respectively). Gadolinium enhancement was reported in the 3 patients with NMOSD or MOGAD. Hippocampal involvement was observed in 4 of 5 patients (80%). During the study period, the extensive FLAIR lesion load increased in 2 of 4 patients (50%; 2 with NMOSD); a decrease in the lesion load was, however, reported in 2 of 4 patients (50%; 1 with MOGAD, 1 with NMOSD; Figure 2 ). A total of 4 of 5 patients (80%) had at least one symptom nontypical of NMDARE, including motor impairment (3/5, 60%, vs 11/214, 5.1%; p = 0.002), cranial nerve palsies (2/5, 40%, vs 10/214, 4.7%; p = 0.036), and cerebellar ataxia (2/5, 40%, vs 13/214, 6.1%; p = 0.039). One patient (20%) had myelitis concomitant with NMDARE. In these patients with extensive FLAIR lesions, the number of typical NMDARE symptoms ranged from 3 to 5 (median was 4, similar to the control group) and all had severe NMDARE (mRS scores 4–5 at nadir). Only 1 of 4 (25%) recovered at 2 years of follow-up, compared with 124 of 162 (76.5%) in the control group ( p = 0.047).

Fluffy Lesions

In 22 of 41 patients (53.7%), the extralimbic lesions were classified as fluffy lesions, consisting in patchy, poorly demarcated lesions that involved the cerebral cortex (6 patients) or the cerebellum (6 patients; Figure 2D ) or were multifocal (at least 2 locations including the cerebral cortex, adjacent white matter, basal ganglia, cerebellum, and/or spinal cord; 10 patients). Gadolinium enhancement was observed in 4 of 22 patients (18.2%). Remarkably, limbic lesions were more frequent in patients with fluffy lesions than in the control group (11/22, 50.0%, vs 37/214, 17.3%; p = 0.001). During the study period, the fluffy lesion load increased in 3 of 19 patients (15.8%); a decrease in the lesion load was, however, reported in 14 of 19 patients (73.7%; Figure 2 ). None of these 22 patients had an overlapping demyelinating syndrome identified. Compared with the control group, the patients with fluffy lesions more often presented at least one symptom nontypical of NMDARE (11/22, 50%, vs 22/214, 10.3%, p < 0.001), including cerebellar ataxia (9/22, 40.9%, vs 13/214, 6.1%, p < 0.001), motor impairment (4/22, 18.2%, vs 11/214, 5.1%; p = 0.039), and oculomotor palsy (3/22, 13.6%, vs 5/214, 2.3%, p = 0.029). However, the median number of typical NMDARE symptoms was similar to that of controls (4, range 2–6), as was the proportion of patients with severe NMDARE (20/22, 90.9% had mRS scores 4–5 at nadir). At 2 years of follow-up, 13 of 19 patients (68.4%) had recovered (mRS scores 0–1), a rate similar to that of the control group (124/162, 76.5%; p = 0.41; Table 1 ).

Associations Between MRI Patterns and Outcomes

Most patients included herein received first-line (248/255, 97.2%) and second-line (184/255, 72.2%) treatments, and 84 of 255 (32.9%) received steroid-sparing agents ( Table 1 ). A relapse of NMDARE was reported in 46 of 255 patients (18.0%) after a median time of 3 years from symptom onset (range 2.8 months–12 years), regardless of MRI features at baseline (data not shown). At 2 years from disease onset, 54 of 197 patients (27.4%) had not recovered (mRS score 2 or greater), including 17 of 54 (31.5%) who died before reaching the 2-year end point (eFigure 2). The causes of death were sepsis (7/17), progression of a malignant tumor (4/17), severe dysautonomia (1/17), ventilator-associated hemoptysis (1/17), status epilepticus (1/17), and unknown (3/17). In univariable analysis, MS-like lesions and hippocampal lesions were negatively associated with recovery at 2 years (extensive lesions [n = 4] were not included), along with other factors such as decreased consciousness or CSF white blood cell count ≥20/µL. Conversely, age younger than 45 years was positively associated with recovery ( Table 2 provides unadjusted odds ratios). The multivariable logistic regression analysis found that MS-like lesions were independently associated with absence of recovery at 2 years (adjusted OR 0.1, 95% CI 0.03–0.42; p = 0.002) as opposed to hippocampal hyperintensities or fluffy lesions, which had no effect on the outcome ( Figure 3 ).

Table 2 Association of the Variables With Complete or Near-Complete Recovery (mRS Scores 0–1) at 2 Years (Univariable Analyses) Characteristics OR 95% CI p Value Age younger than 45 years 12.2 4.11–44.8 <0.001 Age 12 years and younger 0.73 0.35–1.59 0.4 Female sex 1.91 0.90–3.97 0.087 Teratoma 2.43 1.06–6.30 0.047 Malignant tumor 0.04 0.00–0.19 0.002 Clinical features at baseline  Status epilepticus 0.53 0.27–1.03 0.06  Decreased consciousness 0.37 0.18–0.75 0.007  Nontypical symptoms 1.06 0.47–2.56 0.9  Overlapping demyelinating syndrome 0.21 0.04–0.89 0.037 Diagnostic tests  CSF WBC count >20/µL 0.4 0.2–0.79 0.008  Hippocampal lesions 0.38 0.19–0.76 0.006  MS-like lesions 0.24 0.07–0.80 0.020  Fluffy lesions 0.67 0.26–1.88 0.4 Treatments  Immunotherapy within 3 wk 0.65 0.32–1.33 0.24  Intensive care 0.46 0.22–0.98 0.044  Rituximab 0.47 0.22–0.95 0.042  Cyclophosphamide 0.37 0.18–0.73 0.004  Azathioprine 0.79 0.31–2.17 0.6  Mycophenolate mofetil 2.09 0.91–5.45 0.10 Abbreviations: mRS = modified Rankin Scale; OR = odds ratio.

Association of the Variables With Complete or Near-Complete Recovery (mRS Scores 0–1) at 2 Years (Univariable Analyses)

Abbreviations: mRS = modified Rankin Scale; OR = odds ratio.

Prognostic Factors of Recovery of Patients With NMDARE Identified by Multivariable Logistic Regression Analysis

The presence of MS-like lesions, CSF cellularity above 20/µL, and decreased consciousness were independently associated with decreased odds of recovery at 2 years. Conversely, age younger than 45 years increased the odds of recovery while hippocampal hyperintensities and fluffy lesions had no effect on the 2-year outcomes. Extensive FLAIR lesions were not included in the model because of the small number of patients (n = 4). Only the 197 patients with at least 2 years of follow-up were included. FLAIR = fluid-attenuated inversion recovery; NMDARE = anti-N-methyl- d -aspartate receptor encephalitis; WBC = white blood cell.

Discussion

This study identified 3 patterns of extralimbic involvement in NMDARE (MS-like, extensive, and fluffy lesions), in addition to the limbic hyperintensities typically found in NMDARE and other autoimmune encephalitides. Such an extralimbic involvement was found in less than 20% of this cohort while definite overlapping demyelinating disorders were identified in less than 4% of the patients. Of interest, these radiologic patterns were associated with clinical symptoms and prognostic features. MS-like and extensive FLAIR lesions were frequently associated with overlapping demyelinating disorders, such as MS, NMOSD, or MOGAD. In addition, MS-like lesions were significantly associated with the absence of neurologic recovery at 2 years, which was likely due to the disability added by MS and MOGAD. Of note, extensive FLAIR lesions could not be included in the regression analysis owing to the small number of cases; however, most of these patients had not recovered at 2 years, including all patients with overlapping NMOSD or MOGAD. This poor prognosis is also likely related to their association with overlapping demyelinating disorders. Overlapping demyelinating disorders were probably underdiagnosed in this cohort because of heterogeneous durations of follow-up and variable sampling time points. Possibly, most of the MS-like or extensive lesions identified herein corresponded to demyelinating lesions although all patients did not fulfill the criteria for MS, MOGAD, or NMOSD at the time of inclusion. Because demyelinating disorders require distinct treatment options, we advocate for a close clinical and radiologic monitoring of the patients with either MS-like or extensive FLAIR lesions.

As opposed to MS-like and extensive FLAIR lesions, fluffy lesions were not associated with poor outcomes or overlapping demyelinating syndromes but were associated with nontypical symptoms. Fluffy lesions consisted in patchy and poorly demarcated lesions, predominantly found in the cerebral cortex and in the cerebellum, along with the basal ganglia and brainstem in some patients, and represented approximately half of the nonlimbic lesions. The patients with fluffy lesions had a similar number of typical symptoms of NMDARE compared with patients from the control group, but more frequently presented with symptoms considered as nontypical, such as cerebellar ataxia and motor impairment. Of interest, half of the patients with fluffy lesions also had hippocampal hyperintensities, suggesting a pathophysiologic link between hyperintensities of the hippocampus and fluffy lesions of the cortex and cerebellum. Altogether, these results suggest that fluffy lesions, while adding to the clinical diversity of NMDARE, are not associated with overlapping disorders or poorer outcomes.

A previous study reported an association between hippocampal hyperintensities and poor outcomes at 2 years. 7 However, these findings suggest a confounding effect of age because age older than 45 years was associated with both a lack of recovery and hippocampal hyperintensities (which could result from a diagnostic bias in this age category). It is important to note that, after adjustment for age older than 45 years in this population, hippocampal hyperintensities did not have any significant association with the outcomes. It is noteworthy that the patients older than 45 years had more frequent malignant tumors, which likely contributed to the poor functional and survival outcomes. These findings indicate that when focusing on prognostic factors of NMDARE, age is an important confounder that is crucial to adjust for; they also confirm previous reports showing that a malignant tumor should be suspected in patients with NMDARE older than 45 years. 28,29

The study has limitations, mostly due to its retrospective design, the heterogeneity of MRI devices and of time of MRI from symptom onset, and the use of the mRS (which improperly reflects the impact of cognitive impairments) because of the lack of more appropriate clinical scales. 27 In the multivariable analysis, patient dropout led to exclude approximately a quarter of the patients and myelitis (2% of the cohort) was a potential confounding factor. We chose the cutoff of mRS score ≤1 (recovery without residual impairment) because the mRS tends to underestimate cognitive impairment in NMDARE 30 and because complete or near-complete recovery (mRS scores 0–1) is essential for the patients to return to work or academic studies. Although we were not able to retrieve all the initial MRI scans (thereby possibly underestimating the frequency of extralimbic lesions), a second reading was performed for most of the patients with extralimbic lesions by a neurologist specialized in neuroinflammation (J.P.) and using a framework made by an expert neuroradiologist (F.C.). This was paramount to select, characterize, and classify the extralimbic lesions. Moreover, the study included a large cohort with highly detailed clinical information and extensive follow-up, which allowed us to study the outcomes at 2 years from NMDARE onset, when most patients have improved and/or stabilized. 4 We also followed consensual guidelines to limit the risk of false-positive autoantibodies. 1,24,25 The frequencies of unremarkable MRI, hippocampal hyperintensities, and overlapping demyelinating syndromes, as well as the long-term outcomes (27% disability or death at 2 years), are in line with previous studies. 1,2,4,6,13,15 Conversely, the frequency of extralimbic lesions was lower herein (19% vs 23%–33% in other studies), which is likely due to the careful exclusion of nonspecific lesions in this study. Of note, the association of poor outcomes with increased CSF cell count is in line with previous studies. 17 While ICU admission has been found to increase the odds of poor outcomes in a previous study, 17 this effect was not found in our multivariable model. By contrast, we found an association between poor outcomes and decreased consciousness, which is to be confirmed in larger studies because the evaluation of the level of consciousness can be affected by multiple confounders, including severe cognitive impairment and seizures. Finally, the design of the study did not allow the evaluation of treatment efficacy because of a severity illness bias.

In conclusion, this study provides new information regarding the imaging patterns that can predict poor outcomes; we suggest that extralimbic lesions on brain MRI in NMDARE can be broadly classified into 3 patterns. MS-like and extensive FLAIR lesions are often associated with overlapping demyelinating syndromes and poor prognosis while fluffy lesions likely participate to the phenotypic variability of NMDARE and have no impact on disease progression or outcome. Therefore, radiologic patterns may contribute to the accurate estimation of NMDARE prognosis, which is essential to select optimal treatments, to provide appropriate information and guidance to the patient and their relatives, and to anticipate the long rehabilitation that is often required in the postencephalitic phase. Further prognostic studies involving large multicenter cohorts should be conducted because these findings can have practical implications on the monitoring of patients with NMDARE.