We performed a comprehensive literature search of the PubMed, Cochrane, Embase, and Web of Science databases up to 1 March 2024. Autoimmune neurological diseases involve the immune system’s erroneous assault on the nervous system, sparking inflammation and neural damage. These conditions can affect the brain, spinal cord, and peripheral nerves, resulting in various symptoms depending on the affected region (17). The study focused on a range of such conditions, including MS, AIE, MG, NMOSD, neuro-Behcet’s disease (NBD), acute disseminated encephalomyelitis, and Guillain–Barré syndrome. Comprehensive search details are provided in Supplementary Table S1 .

Two independent investigators, XL Deng and X Gong, used Endnote to screen studies for inclusion in this evaluation. Initially, studies were screened by title and abstract to exclude those not meeting the eligibility criteria, followed by a full-text review for further evaluation. Any discrepancies between investigators regarding study inclusion were discussed to establish consensus. We included studies that fulfilled the following criteria: (1) implementation of observational designs including case–control studies, cross-sectional studies, cohort studies, or baseline data of intervention studies; (2) analysis of gut microbiota using available diversity or abundance data; (3) inclusion of populations diagnosed with autoimmune neurological diseases; and (4) publication of full-text articles in English in peer-reviewed journals. In cases where data were duplicated or shared between studies, the most recent and comprehensive dataset was used for analysis.

XL Deng and X Gong extracted data from full-text articles using a pre-designed template, which was then cross-checked for accuracy. The primary outcomes and metrics were α-diversity, β-diversity, and the relative abundance of intestinal microbes. α-diversity provides an overview of the microbial community within a single sample and is used to compare across different groups assessing richness (number of taxa) and evenness (the degree to which each taxon is represented). β-diversity serves as a metric for assessing the diversity among samples and examining the phylogenetic composition of communities relative to other analyzed samples (1, 18). Additionally, other pertinent information such as publication details, participant demographics, and methodologies was extracted.

The quality of the included research was assessed using the Newcastle–Ottawa Quality Assessment Scale (NOS) (19). Scores of 5 or lower were considered low quality, scores of 6 or 7 signified moderate quality, and scores of 8 or 9 indicated high quality. XL Deng and X Gong independently conducted assessments to determine the quality of the research involved in the analysis. In cases of score disagreement, the two reviewers collaboratively re-evaluated the articles to reach a consensus (2).

A meta-analysis was conducted to compare the α-diversity between individuals with autoimmune neurological diseases and healthy controls. The analysis included indices containing data from at least 10 research studies. The pooled standardized mean difference (SMD) and related 95% confidence interval (CI) were computed via inverse-variance random-effects meta-analysis for each index. The effect sizes were classified based on magnitude: trivial (SMD ≤ 0.2), small (0.2 < SMD < 0.5), moderate (0.5 ≤ SMD < 0.8), or large (SMD ≥ 0.8). Means and standard deviations were derived from medians and interquartile ranges. If numerical data were only presented graphically, they were extracted using WebPlotDigitizer V.4.6 (2). The DerSimonian-Laird estimator was utilized to assess heterogeneity, with significant heterogeneity defined as I ² > 50%. Publication bias was assessed using funnel plots and the Egger’s test.

Subgroup analysis was conducted to investigate the heterogeneity among studies, examining various factors such as the type of autoimmune neurological diseases, the geographical distribution of study populations (comparing countries in the East and West), and the use of immunotherapy (including immunosuppressants, immunomodulators, and immune-reconstitution therapy), distinguishing between patients receiving treatment and those without it. Eastern countries comprise East and South Asian nations, while Western countries encompass North America, Europe, and the Middle East. Studies where at least 80% of patients received immunotherapy were classified as involving treated patients. To assess the reliability of the results, two sensitivity analyses were carried out. Studies rated as low quality (with NOS scores of 5 or less) and those without matching characteristics (including age, sex, and body mass index) were excluded due to their susceptibility to confounding bias (1, 2). All analyses were conducted using the STATA/MP 17.0, with statistical significance defined as p-values less than 0.05.

When significant variations in β-diversity between patients with autoimmune neurological diseases and healthy individuals were observed across all studies focusing on a particular condition, these differences were considered consistent. A qualitative synthesis was undertaken due to the extensive findings and low overlap in the relative abundance of microbial species. Intra-diagnostic and inter-diagnostic comparisons were undertaken to delineate disease-specific and common alterations. Disease-related findings reported by at least two studies per taxon were summarized, with taxa categorized as exhibiting increased, decreased, or inconsistent patterns. Inconsistency was defined as less than 75% concordance between studies. Concordant findings from two studies were deemed notable for future validation, while those from three or more studies conducted by at least two independent research teams were deemed possibly relevant to the disease. A taxon exhibiting consistent change in only one disease was identified as a potential disorder-specific response, whereas changes observed in at least three disorders were categorized as common changes (2).

Initially, a total of 3,042 studies were identified ( Figure 1 ). Following the removal of 1,403 duplicates, 1,550 studies that failed to fulfill the inclusion criteria were screened based on their titles and abstracts. Subsequently, 89 studies underwent full-text review, of which 27 were excluded ( Supplementary Table S2 ). Finally, 62 studies focusing on six autoimmune neurological disorders (including AIE, MG, NMOSD, MS, and DON) were included (4–7, 9, 11–13, 20–73). MS was the most extensively studied disease, followed by MG, NMOSD, and AIE ( Supplementary Table S3 ). The total number of participants was 3,126, ranging from 13 for NBD to 2,393 for MS. The mean age of participants varied from 32.93 years for AIE to 42.10 years for NBD ( Table 1 ).

More than half of the studies [39 (62.90%)] were conducted in Western countries (e.g., USA, Germany, UK, Italy, and Israel), 21 (33.87%) were conducted in Eastern countries (China and Japan), and only 2 (3.23%) were conducted in Africa (Egypt). Most studies on MS were conducted in Western countries [37 (84.09%)], while the majority of studies on other diseases were conducted in Eastern countries ( Supplementary Table S3 ). Sample sizes for most studies were small to moderate, ranging from 7 to 500 participants. Although the exclusion criteria were generally consistent, the use of immunotherapy varied widely. Out of the studies analyzed, 22 studies (35.48%) were conducted without immunotherapy. In contrast, six studies (9.68%) included only patients who had received immunotherapy. The remaining studies did not control for this, resulting in a wide variation in the proportion of patients on immunotherapy, ranging from 9% to 95.45%. Stool processing methods and compositional analyses also varied significantly. Sequencing of 16S ribosomal RNA (16S rRNA) was the most widely applied method, used in 54 studies, followed by shotgun metagenomics in 12 studies (5 of which used metagenomics exclusively), real-time quantitative polymerase chain reaction in 2 studies, and fluorescence in situ hybridization in 1 study ( Supplementary Table S5 ). Based on the NOS score, two studies were rated as high quality, 43 studies were rated as moderate quality, and 17 studies were rated as low quality ( Supplementary Table S4 ).

Data on α-diversity were contributed by 42 studies, encompassing metrics of richness [observed species, abundance-based coverage estimator (ACE), Chao1, richness, and Good’s coverage), evenness (Shannon index, Simpson index, Pielou’s evenness, and evenness), and biodiversity (Faith’s phylogenetic diversity). The most frequently used measures were the observed species, ACE, Chao1, Shannon index, and Simpson index. Inspection of funnel plots revealed some asymmetry ( Supplementary Figure S1 ), but no statistically significant results were observed using Egger’s test ( Supplementary Table S10 ).

In terms of richness, data on observed species in patients (n = 644) compared to healthy individuals (n = 773) were provided by 16 studies (11, 12, 21, 23, 25–27, 29–33, 35, 44, 49, 50). The aggregated assessment indicated no significant differences between the groups (SMD = −0.16, 95% CI = −0.40 to 0.08, p = 0.20), with considerable heterogeneity (I ² = 76%). When analyzing individual diseases, a significant reduction in gut microbiota richness was found exclusively in MG patients (SMD = −0.74, 95% CI = −1.38 to −0.09, p = 0.025) ( Figure 2A ). ACE data were available from 11 studies (4, 5, 11, 21, 25, 29, 31, 32, 41, 49, 55), comparing patients with autoimmune neurological diseases (n = 390) to controls (n = 378). The overall estimate for combined autoimmune neurological diseases showed a decline in ACE compared to controls, albeit with a small effect size (SMD = −0.26, 95% CI = −0.58 to 0.07, p = 0.12; I ² = 80%), which did not reach statistical significance. However, a moderate significant decrease was observed in MG patients individually (SMD = −0.79, 95% CI = −1.53 to −0.06, p = 0.03; I ² = 74%) ( Figure 2B ). Chao1 data were reported in 22 studies (4–6, 11, 12, 21, 23, 25, 27, 29–33, 41, 44, 49, 55, 61, 63, 72, 73), encompassing patients with autoimmune neurological diseases (n = 942) and controls (n = 866). Overall, there was a significant reduction in Chao1, though the effect size was small (SMD = −0.26, 95% CI = −0.45 to −0.07, p < 0.01), with high heterogeneity (I ² = 73%). When focusing on specific diseases, a significant reduction was also only found in MG (SMD = −0.73, 95% CI = −1.19 to −0.27, p < 0.01; I ² = 69%) ( Figure 2C ). Regarding evenness, the Shannon index was reported in 38 studies (4–6, 11–13, 20, 21, 23–27, 29–33, 35, 41, 42, 44, 49, 51–53, 55, 60, 61, 63–69, 72, 73), comparing patients (n = 2,101) to controls (n = 1,754). The combined assessment revealed no significant difference (SMD = −0.09, 95% CI = −0.23 to 0.05, p = 0.21). When examining specific diseases, a significant decrease was only observed in the AIE group (SMD = −0.39, 95% CI = −0.74 to −0.04, p = 0.03; I ² = 57%) ( Figure 3A ). Data on the Simpson index were supplied by 22 studies (4, 5, 11, 20, 21, 23, 25–27, 29–31, 33, 41, 44, 49, 51, 52, 55, 62, 63, 73), comparing patients (n = 712) to controls (n = 664). No significant change between the groups was observed (SMD = 0.07, 95% CI = −0.15 to 0.28, p = 0.48; I ² = 75%) ( Figure 3B ).

To explore the reasons for the higher heterogeneity, subgroup analyses were conducted. In the subgroup analyses of Eastern and Western studies, a minor difference in the Shannon index was noted solely in the Eastern population (SMD = −0.27, 95% CI = −0.47 to −0.07, p < 0.01; I ² = 72%) ( Supplementary Table S6 ). Most α-diversity indices exhibited no significant difference, and substantial heterogeneity persisted (I ² ranged from 46% to 85%). In the subgroup analyses of treatment-naive patients versus those on treatment, a decrease in most α-diversity indices was evident only in treatment-naive patients. Nevertheless, substantial heterogeneity remained (I ² ranged from 44% to 82%) ( Supplementary Table S7 ). Additionally, sensitivity analyses were conducted by excluding low-quality studies ( Supplementary Table S8 ) and studies without variable matching ( Supplementary Table S9 ). Consistent with the overall analysis, only Chao1 remained significantly decreased in individuals with autoimmune neurological diseases, indicating that the data are robust.

A comparison of β-diversity between patients with autoimmune neurological disorders and healthy individuals was conducted across 53 studies encompassing six distinct diseases (AIE, MG, NMOSD, MS, NBD, and DON) using varied methodologies ( Figure 4 ). Significant changes in β-diversity between patients with autoimmune neurological disorders and healthy individuals were reported in 4 out of 5 studies on anti-NMDAR encephalitis, 1 study on anti-LGI1 encephalitis, 5 out of 7 studies on MG, all 7 studies on NMOSD, 21 out of 34 studies on MS, 1 study on NBD, and 1 study on DON. Notably, among these disorders, only NMOSD showed consistently distinct β-diversity.

A total of 55 studies compared the relative abundance of gut microbiota at the genus, family, or phylum levels between patients with autoimmune neurological disorders and healthy individuals. These alterations encompassed 11 phyla, 32 families, and 88 genera after excluding non-replicated findings ( Supplementary Figure S2 ). Within-disease and between-disease comparisons were summarized for diseases with sufficient research (AIE, MG, NNMOSD, and MS reported by two or more studies). The findings suggest that specific microorganisms undergo common changes across multiple autoimmune neurological diseases, although most consistent disease-related changes were replicated by only two studies. At the phylum level, the most consistent alterations were an increase in Proteobacteria (8 of 8 studies) and a decrease in Firmicutes (8 of 8 studies) in AIE, MG, and NMOSD, and an increase in Verrucomicrobia (9 of 10 studies) in AIE, NMOSD, and MS. At the family level, Ruminococcaceae (11 of 13 studies) and Lachnospiraceae (14 of 16 studies) consistently showed decreases, while Streptococcaceae showed an increase (10 of 11 studies) in AIE, MG, NMOSD, and MS. At the genus level, the most consistent alterations were an increase in Streptococcus (16 of 18 studies) and Escherichia-Shigella (12 of 13 studies), and a decrease in Roseburia (17 of 18 studies) and Faecalibacterium (23 of 24 studies) ( Figure 5 ) in AIE, MG, NMOSD, and MS.

In subgroup analyses, microbial changes were found only in studies from Eastern countries, including an increase in genus Acinetobacter and decreases in genera Clostridium_XIV, Granulicatella, and Megasphaera. In contrast, decreases in the genus Slackia and increases in the genera Holdemania, Megasphaera, and Olsenella were observed in studies from Western countries. In subgroup analyses of immunotherapy use, comparisons were made between studies without treatment (n = 22) and those with 80% or more of patients on medication (n = 13). Increases in the genera Acinetobacter, Collinsella, Enterococcus, Escherichia, Fusicatenibacter, Succinivibrio, and Sutterella, along with decreases in the family Rikenellaceae and the genera Gemella and Haemophilus, were exclusively reported in treatment-naive studies. However, these findings should be regarded as preliminary due to the uneven distribution of studies by region (e.g., AIE, MG, and NMOSD were predominantly studied in Eastern populations) and immunotherapy use (e.g., AIE and MG were predominantly studied in treatment-naive patients, while NMOSD was predominantly studied in patients receiving immunotherapy).