This study was performed in accordance with the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) program. This protocol was registered in INPLASY (registration number: INPLASY202280088).

The PubMed, Embase, Web of Science Core Collection, and MEDLINE databases were searched for literature describing a link between PD and AIDs. From the day that each database was created until December 12, 2022, each database was searched. PD and AID-related free-text phrases were combined with restricted vocabulary terms that were particular to the database. To capture other potentially relevant articles, we also combined search terms representing PD with search terms for 34 AIDs (i.e., chronic inflammatory demyelinating polyneuropathy, Guillain–Barre syndrome, multiple sclerosis [MS], myasthenia gravis [MG], Addison’s disease, type 1 diabetes mellitus [T1D], Graves’ disease [GD], Hashimoto disease, autoimmune hepatitis, inflammatory bowel disease [IBD], Crohn’s disease [CD], ulcerative colitis [UC], coeliac disease [CLD], pernicious anemia [PA], primary biliary cirrhosis [PBC], alopecia areata, primary sclerosing cholangitis, antiphospholipid syndrome, autoimmune hemolytic anemia [AIHA], immune thrombocytopenic purpura, polymyositis [PM], rheumatoid arthritis [RA], polyarteritis nodosa, giant cell arteritis, bullous pemphigoid [BP], discoid lupus erythematosus, vitiligo, Behcet’s disease [BD], scleroderma, SS, systemic lupus erythematosus [SLE], dermatitis herpetiformis, narcolepsy, and rheumatic fever) that were reported to be most prevalent (i.e., with a worldwide prevalence rate of ≥10/100,000 people based on a review that included a comprehensive list of AIDs) (18). To be as comprehensive as possible, the search was not restricted to any study type. An example search strategy for the PubMed database is described in Supplementary Table 1 .

Peer-reviewed publications that presented population-based research showing a link between PD and any form of AIDs were required for the study’s inclusion. Case reports and case series were not included since it was unclear how well they sampled from among those groups. To prevent duplication and incorrect weighting toward more frequently referenced or discussed articles, review papers, meta-analyses, organizational recommendations, editorial letters, and professional views were eliminated. Additionally, conference abstracts were disregarded since their full study reports could not be obtained for evaluation and their scientific rigor had not undergone peer assessment. Additionally, only research that appeared in English-language publications was considered.

The four databases’ search results yielded articles that were all imported into Endnote for review. For study inclusion, there were two rounds of screening. Two reviewers (ML and JW) independently performed screening in the first round by examining titles, abstracts, and key terms for relevance to both AIDs and PD. The entire contents of the articles found during the first round of screening were collected and carefully read to determine eligibility in the second round of screening. To establish agreement, potential disagreements throughout the study selection process were discussed with a third reviewer (ZX).

Data from included studies were extracted into a standard table, detailing authors, year of publication, country, study period, study design, selection of control and comparison groups, number of study participants, reported risk estimates for PD associated with AIDs, and any adjustments for confounding factors in producing effect estimates.

Study quality was evaluated on the Newcastle–Ottawa Scale (NOS). Studies that achieve seven or more stars on the NOS are considered high quality, while four to six stars indicate moderate quality, and less than four stars indicate poor quality.

Meta-analysis was performed for each study reporting the correlation between PD and AID. Statistical heterogeneity between studies was assessed for each outcome by examining study-specific effect size and heterogeneity (I²) statistics. I² values of <25%, 25–50%, 51–75%, and >75% were considered to denote no, mild, moderate, and large heterogeneity, respectively (19). In meta-analyses of multiple studies for a specific outcome, a fixed-effect estimate was calculated if the I² value was <50%; a random-effect estimate was calculated if the I² value was ≥50%. Forest plots were constructed to present risk estimates and meta-analysis results for each AID reported by at least three studies to be associated with PD. In meta-analyses of cohort studies, the hazard ratio (HR) and standardized incidence ratio (SIR) were treated as being equal to the odds ratio (OR). Publication bias was assessed by observing the symmetry of the funnel plot and performing Egger’s and Begg’s tests.

Statistical analysis was performed using STATA software (Version 16.0, StataCorp, College Station, TX, USA). Statistical significance was set at P<0.05.

In the original search, 4,535 results from MEDLINE, 4,237 from PubMed, 5,676 from the Web of Science Core Collection, and 6,296 from Embase were found. The four databases’ search results yielded articles that were all imported into Endnote for screening. There were 10,379 papers uploaded to Endnote for first-stage screening following the removal of duplicates. Following the initial screening of publications by looking at their titles, abstracts, and key words relating to PD or AIDs, 10,058 publications were disqualified for not addressing both PD and AIDs, and 321 papers were found to be potentially relevant and subsequently evaluated for eligibility. After the first round, 275 articles were excluded for being a review article or meta-analysis (n=100), not having a control group (n=7), being a comment (n=12), a case report (n=129), or a meeting (n=27). Finally, 46 articles met all the inclusion criteria and were included in this review, 38 of which were included for quantitative synthesis for having calculable risk estimates ( Figure 1 ).

Table 1 shows a detailed description of the key characteristics of the 46 included studies. In brief, there were 16 national studies; 18 (39.1%) were conducted in European populations, 5 (10.9%) were conducted in South America, and 23 (50.0%) were conducted in Asia. In terms of study design, 4 (8.5%) were cross-sectional studies, 18 (38.3%) were case-control studies, and 25 (53.2%) were cohort (including 8 prospective cohort and 17 retrospective cohort) studies, with the study by Brick et al. (29) reporting both case-control and retrospective cohort studies.

The search identified clinical studies with 17 AIDs and PD. A total of 14,276,464 individuals, 873,643 patients, and 13,402,821 controls were included in the studies, including 752,488 patients with AIDs and 121,155 with PD. For the reported AIDs, Rugbjerg et al. (13) conducted a national case-control study in Denmark and reported 32 AIDs before diagnosis in 13,695 patients with PD. Li et al. (12) conducted a national epidemiological study in Sweden and reported a total of 33 AIDs in 310,522 patients, with 932 subsequently presenting with PD. In addition, 17 (37.0%) studies reported on BP, 12 (26.1%) on IBDs (including UC and CD), 7 (15.2%) on SS, 7 (15.2%) on RA, 4 (8.7%) on SLE, 5 (10.9%) on MS, 3 (6.5%) on CLD, 3 (6.5%) on GD, 2 (4.3%) on T1D, 2 (4.3%) on BD, 2 (4.3%) on MG, 2 (4.3%) on PM, 2 (4.3%) on scleroderma, 2 (4.3%) on Addison’s disease, 2 (4.3%) on AIHA, 2 (4.3%) on PA, and 2 (4.3%) on PBC. Six studies included patients with PD as the test group, and 39 used patients with AIDs as the test group. Most studies matched 1–10 controls per case based on age and sex. Table 2 lists the studies included in the meta-analysis that included the correlation between PD and AIDs.

The number of clinical studies on T1D, BD, MG, PM, scleroderma, Addison’s disease, AIHA, PA, and PBC with PD was <3; therefore, those diseases were not included in the meta-analysis. Supplementary Table 2 lists the studies that were not included in the meta-analysis. Among them, results from Klimek et al.’s study (32) suggested a significantly higher risk of PD combined with T1D (relative risk [RR]=2.30, 95% CI: 1.90–2.70). Results from Park et al.’s study (49) suggested a significantly higher risk of PD combined with BD of 2.47 (1.65–3.68). Rugbjerg et al. (13) and Li et al. (12) included studies of multiple AIDs and risk of PD; however, a meta-analysis was not performed for some of these diseases due to the small number of studies that could corroborate any associated findings.

Both Hsu et al.’s (20) and Lin et al.’s (34) studies are reports on IBD with patients originating from the same database; therefore, the study with the larger sample size (i.e., Lin et al.’s study) was selected, excluding that by Hsu et al. (20). Studies by Ju et al. (46), Hsu et al. (51), and Chang et al. (14) all originate from a study on SS, with patients from the same database. The study with the larger sample size (i.e., Chang et al.’s study) was selected synthetically, excluding those by Ju et al. (46) and Hsu et al. (51). Studies by Nielsen et al. (30) and Thormann et al. (36) are Danish reports on MS with patients derived from the same database; because of which, Thormann et al.’s study (36) was selected as it included more complete information, excluding Nielsen et al.’s report (30). A retrospective cohort study in South Korea by Noh et al. (52) compared conventional treatment and combination treatment for IBD and found a reduced risk of PD in the combined treatment group without normal controls, because of which it was excluded. Paakinaho et al.’s study (23) examined the association between RA and PD by improving anti-rheumatic drugs, which did not match the data of this study; thus, it was excluded from the meta-analysis. Thirty-seven studies were eventually included in the meta-analysis.

The NOS scores of the included studies ranged from 6 to 9, indicating a high level of overall quality. The studies had clear definitions of exposure and outcome, appropriate adjustment for confounders, and sufficiently long follow-up ( Supplementary Table 3 ).

The results of the meta-analysis of 17 case-control, 4 cross-sectional, and 18 cohort studies showed a significantly higher overall risk of PD with AIDs (OR=1.55, 95% CI: 1.33–1.81, P=0.000, I^{2 =} 95.1%, large heterogeneity), with case-control (OR=1.83, 95% CI: 1.24–2.69, P=0.000, I^{2 =} 96.5%, large heterogeneity), cohort (OR=1.42, 95% CI: 1.22–1.65, P=0.000, I^{2 =} 92.7%, large heterogeneity) and cross-sectional studies (OR= 1.72, 95% CI: 1.06–2.79, P=0.000, I^{2 =} 90.2%, large heterogeneity) all suggesting an increased risk of PD with AIDs ( Figure 2 ). The findings suggested high heterogeneity, and a subgroup analysis was performed to find the source of heterogeneity. However, we found no significant differences in risk by study type, gender, age, race, and study design ( Table 3 ). Therefore, we performed separate analyses to determine the relationship between different types of AID and PD.

Seventeen studies reported the risk between BP and PD, and the results showed a significantly higher risk for PD combined with BP (OR=2.67, 95% CI: 2.15–3.31, P=0.000, I^{2 =} 63.3%, moderate heterogeneity). According to the grouping of study types, the results showed no heterogeneity in the effect values of the 12 case-control studies combined (OR=3.36, 95% CI: 2.98–3.79, P=0.206, I^{2 =} 24.4%) ( Figure 3 ).

Ten studies reported the risk between IBD and PD and showed a significantly higher risk of PD combined with IBD (OR=1.24, 95% CI: 1.04–1.47, P=0.001, I^{2 =} 91.0%), with large heterogeneity. However, when Camacho-Soto et al.’s study (21) was removed from this analysis, heterogeneity was significantly lower after combining effect values and the association between PD and IBD was stronger. We therefore included nine studies, excluding Camacho-Soto et al.’s report (21), which showed a significantly higher risk of PD with IBD (OR=1.30, 95% CI: 1.18–1.45, P=0.024, I^{2 =} 54.6, moderate heterogeneity). Subtypes of IBD, UC (OR=1.31, 95% CI: 1.14–1.50, P=0.035, I^{2 =} 51.7%, moderate heterogeneity), and CD (OR=1.30, 95% CI: 1.20–1.42, P=0.220, I^{2 =} 25.2%, mild heterogeneity) were also associated with a significantly higher risk of PD ( Figure 4 ).

Five studies reported the risk between SS and PD, and the results showed an increased risk of PD combined with SS (OR=1.61, 95% CI: 1.24–2.09, P=0.033, I^{2 =} 61.9%, moderate heterogeneity) ( Figure 5 ). However, sensitivity analysis showed fewer stable results, and two studies, i.e., those by Chang et al. (14) and Wu et al. (17), would have a greater impact on the results ( Supplementary Table 4 , Supplementary Figure 1F ). Three studies reported the risk between GD and PD, and the results showed an increased risk of PD combined with GD (OR=1.45, 95% CI: 1.24–1.70, P=0.136, I^{2 =} 50.0%, mild heterogeneity) ( Figure 5 ).

Six studies reported the risk between RA and PD, and the results showed that the risk of PD combined with RA was not significant (OR=0.79, 95% CI: 0.61–1.03, P=0.000, I^{2 =} 92.4%, large heterogeneity) ( Figure 5 ). Notably, after removing this study by Chang et al. (14) from the sensitivity analysis, the combined effect values became meaningful, showing a negative association between PD and RA ( Supplementary Figure 1I ).

Four studies reported the risk between SLE and PD, and the results showed that the risk for PD combined with SLE was not significant (OR=0.82, 95% CI: 0.66–1.03, P=0.155, I^{2 =} 42.8%, large heterogeneity) ( Figure 5 ). Four studies reported the risk between MS and PD, and the results showed that the risk of PD combined with MS was not significant (OR=2.02, 95% CI: 0.87–4.70, P=0.001, I^{2 =} 89.5%, large heterogeneity) ( Figure 5 ). Three studies reported the risk between CLD and PD and showed that the risk for PD combined with CLD was not significant (OR=1.16, 95% CI: 0.79–1.69, P=0.944, I^{2 =} 0.0%, no heterogeneity) ( Figure 5 ).

The P-values for both Begg’s and Egger’s tests were >0.05, indicating a low likelihood of potential publication bias ( Supplementary Table 5 ).

Sensitivity analyses removed each included study individually and performed a pooled analysis of the remaining studies to assess whether the individual included reports had a greater impact on the results of the overall meta-analysis. Of the several analyses with positive results, the remaining studies analyzed did not have a disproportionate effect on the results of the meta-analysis, except for the less stable results of SS, indicating that the results of the remaining studies were stable and reliable ( Supplementary Table 4 , Supplementary Figure 1 ).