A comprehensive literature search was conducted in the electronic databases PubMed and Embase through September 10, 2024, to identify potential literature. The search terms used were (parkinson OR parkinsonian OR parkinsonism OR parkinson disease OR parkinson’s disease OR paralysis agitans OR parkinsonian disorders OR parkinsonian syndromes OR parkinsonian diseases) AND (appendectomy OR appendectomy OR appendicitis OR appendix OR append*). Inclusion criteria encompassed case–control studies, prospective cohort studies, and retrospective cohort studies published in English, of high quality, with a matched control group, and reporting measurable outcomes.

HubMeta, a free web-based data entry system, was used in the data extraction process. Two independent reviewers (HLC and YST) screened titles and abstracts of extracted data after removing duplicates. Full-text articles were then assessed independently by the same reviewers to determine eligibility. Disagreements were resolved through discussion with a third reviewer (HS) until consensus was reached.

The quality of the collected literature was assessed using the Newcastle-Ottawa Scale (NOS). Studies with a score > = 7 were considered high quality studies. Two researchers (HLC and YST) independently conducted the quality assessments, with any disagreements resolved by a third reviewer (HS) after discussion.

We pooled the data and calculated adjusted relative risks (RR) with 95% Confidence Interval (95% CI). Odd ratios (OR) and Hazard ratios (HR) were treated as RR in this study, given that the prevalence of PD in the general population is less than 10% (10). The meta-analysis study employed the random-effects model, and statistical analyses were conducted using Review Manager 5.4 (Nordic Cochrane Centre, Copenhagen, Denmark). A p-value <0.05 was considered statistically significant. Heterogeneity was evaluated using the I² statistic, with I² > =50 indicating significant heterogeneity. Subgroup analyses were conducted using a fixed-effects model to assess differences between groups. Initial subgroup analyses included maximum follow-up years and study design. Additional subgroup analyses based on geographic region and appendectomy assessment method were conducted in response to reviewers’ feedback. No adjustment for multiple testing was applied for subgroup analyses. Sensitivity analysis was also performed to determine the robustness of the results. Meta-regression, Egger’s test and Begg’s test were conducted using STATA/SE version 17.0 (StataCorp, College Station, TX, USA). Meta-regression was performed as a random-effects meta-regression model with restricted maximum likelihood (REML) method. The moderators included follow-up years, study design, geographic region, and appendectomy assessment method.

The initial literature search retrieved 764 articles, with 532 remaining after removing duplicates. Title and abstract screening excluded 513 articles, and 19 full-text articles were assessed for eligibility. Of these, three articles had only abstract available without further data published in full text. Three articles were abstracts that later published as full articles which were included in the analysis. Four studies were excluded based on quality criteria assessed by NOS. Ultimately, 9 studies met the inclusion criteria for the systematic review and meta-analysis (11–19) (Figure 1). The quality assessment of the included studies using the Newcastle-Ottawa Scale is depicted in Table 1.

The included studies comprised a total population of 8,297,621, with sample sizes ranging from 49,248 to 3,224,650. The studies were published between 2016 and 2024 and included participants from Canada (11), Denmark (12), Sweden (13, 15), United States (14, 16, 17), Korea (18), United Kingdom (19). Of the 9 included studies, 7 studies were cohort studies (11–14, 16, 18, 19), 1 was case–control (15), and 1 employed a case–control design with complementary cohort (17). Assessment of appendectomy included self-report and recorded codes. Assessment of PD included recorded codes and history of antiparkinson drug prescription. Maximum follow-up time ranged from 13 years to 52 years. All included studies scored highly on the NOS, with scores between 7 and 9. The characteristics of included studies are depicted in Table 2.

Pooled results from the 9 included studies demonstrated no statistically significant association between appendectomy and risk of PD (Pooled RR: 1.01, 95%CI: 0.90–1.12, p = 0.89) (Figure 2). Significant heterogeneity was observed (I² = 88%, p < 0.01). The funnel plot appeared asymmetrical, supported by a positive Egger’s test (p < 0.01), while Begg’s test was not significant (p = 0.18), suggesting the presence of potential small-study effects (Figure 3). Meta-regression analyses were conducted to evaluate potential moderators, including follow-up years, study design, geographic region, and appendectomy assessment method; none of these variables sufficiently explained the heterogeneity observed (all p > 0.05).

Given that PD is a chronic disease that becomes more common with age, and all effect estimates were treated as RR due to PD prevalence being <10% in the general population, two subgroup analyses were decided to be performed before the beginning of the study (Table 3).

For maximum follow-up years, studies were divided into two subgroups: >30 years and <30 years. A statistically significant subgroup differences p-value was observed, suggesting a possible presence of subgroup effect. However, substantial amount of heterogeneity was noted within both subgroups (>30 years: I² = 93%, p < 0.01; <=30 years: I² = 77%, p < 0.01), making the validity of effect estimate for each subgroup uncertain (Table 3).

For study design, studies were divided into two subgroups: cohort and case–control. One of the included studies used a design of case–control with complementary cohort. This study was treated as a case–control design in our study. A statistically significant subgroup differences p-value was observed, suggesting a possible presence of subgroup effect. However, substantial amount of heterogeneity was noted within both subgroups (cohort: I² = 83%, p < 0.01; case–control: I² = 87%, p < 0.01), making the validity of effect estimate for each subgroup uncertain (Table 3).

Additional subgroup analyses based on geographic region and appendectomy assessment method were conducted in response to reviewers’ feedback, which also demonstrated high heterogeneity within geographic region subgroups (Asia-Pacific: I² = 70%, p = 0.02; Europe: I² = 93%, p < 0.01) as well as within appendectomy assessment method subgroups (ICD codes: I² = 92%, p < 0.01; Non ICD codes: I² = 50%, p = 0.11). A statistically significant subgroup difference was observed in the appendectomy assessment method analysis (Table 3).

Sensitivity analyses were performed to evaluate the robustness of findings to changes. Each study was omitted one by one in performing the sensitivity analyses. Since the two Swedish studies included had a potential partial overlap of populations with variations in ascertainment, a model excluding both studies was also performed to assess the potential effect of oversampling on skewing the results (13, 15). Results indicated that removing any single study did not significantly alter the conclusion that no association was observed between appendectomy and the risk of PD (Table 4).