Non-genetic risk factors of Parkinson’s disease: A large meta-analysis and systematic review

To the Editor: In the recent decades, significant global endeavors have been undertaken to investigate the risk factors associated with Parkinson's disease (PD) and to enhance the evidence on PD prevention.[1] However, the prevailing key issues in this field are the frequent occurrence of inconsistent or contradictory conclusions and the varying levels of credibility, which are primarily due to the diverse study designs employed. Observational prospective studies (OPSs) and ambidirectional cohort studies (ACSs) have fewer sources of bias and confounding factors compared with retrospective studies.[2] Therefore, we conducted a meta-analysis on prospective studies to obtain more robust results. Considering that the majority of genetic mutations or chromosomal abnormalities occur randomly during DNA replication and are challenging to prevent, this meta-analysis aims to investigate non-genetic risk factors for PD. The non-genetic risk factors encompass the aspects such as environment, lifestyle, and socio-economic that are not directly determined by an individual's genes.[3] These factors may independently influence an individual's health status or interact with genetic factors. We conducted a comprehensive meta-analysis in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020 guidelines.[4] From inception until July 1, 2023, articles published in English were searched in PubMed, EMBASE, Cochrane Library, and Web of Science databases. Our search strategy included relevant keywords, including "Parkinson's disease", "risk", "cohort", "prevention or prevent", and "prospective". This systematic review and meta-analysis were registered in PROSPERO (https://www.crd.york.ac.uk/PROSPERO/, No. CRD42022365049). The inclusion criteria for the studies were as follows: (1) clear statement of the diagnostic criteria of PD by either the UK biobank or Movement Disorder Society; (2) conduction of prospective research, including OPSs and ACSs; and (3) provision of hazard ratio (HR) values with corresponding 95% confidence intervals (CIs) or availability of primary data that could be converted. Exclusion criteria encompassed non-prospective study types, review articles, abstract-only publications, case reports, editorials, opinion pieces, and book chapters. A conceptual diagram illustrating the study selection is shown in Supplementary Figure 1, https://links.lww.com/CM9/C44. The following information was extracted from the eligible literature: (1) name of the first author; (2) year of publication; (3) geographical location, including country, and region; (4) source of study population; (5) recruitment period; (6) follow-up duration; (7) exposure factors and outcome indicators; (8) statistical model employed; and (9) effect size with corresponding 95% CI (if the literature provided both the crude HR and corrected HR, the corrected HR value should be used) [Supplementary Table 1, https://links.lww.com/CM9/C43]. Quality assessment was performed using the Newcastle–Ottawa Quality Assessment Scale (NOS), where studies scoring 7–9, 4–6, and <4 were classified as high-, medium-, and low-quality, respectively. Low-quality studies were also excluded as shown in Supplementary Table 2, https://links.lww.com/CM9/C43. Factors with three or more studies were included in the pooled analysis, and HR and 95% CI were combined using the random effect model. The combined effect was visually represented using a forest map. Heterogeneity was assessed using the Q test and I2 statistics, where I2 ≥75% indicated significant heterogeneity. Publication bias was evaluated through Begg's and Egger's tests and visualized using a funnel chart. In case of asymmetry in the funnel plot, the "trim and fill" method was employed to investigate potential publication bias. The robustness of the results was evaluated by sensitivity analysis using the leave-one-out method. Subgroup analyses were performed based on study type, follow-up duration, and country grouping to investigate the potential sources of heterogeneity. Meta-regression analysis was conducted when more than ten studies were available. Statistical significance was defined as a P-value <0.05. The meta-analysis was conducted using R4.2.1 software (R Foundation for Statistical Computing, Vienna, Austria) and Stata16 software (Stata-Corp LP, College Station, TX, USA). Furthermore, based on our previous investigation, the identified risk factors can be categorized as follows: Class I (strong risk factors) with an HR value ≥1.5; Class II (weak risk factors) with an HR value <1.5; and an HR value <1 is considered a protective factor for PD. Moreover, those factors, which were reported by fewer than three studies or for which no HR was provided, were included in the systematic review. A total of 59,111 articles were retrieved; out of these, 209 articles were deemed eligible, including 174 OPSs and 35 ACSs [Supplementary Figure 1, https://links.lww.com/CM9/C44]. In the pooled analysis, a comprehensive evaluation was conducted on a total of 58 factors classified into six categories. Ultimately, we identified 28 factors associated with PD, comprising of 23 risk factors and five protective factors [Figure 1]. Notably, within the lifestyle category, three protective factors emerged: smoking (HR = 0.69, 95% CI = 0.63–0.77, P <0.001), coffee consumption (HR = 0.74, 95% CI = 0.62–0.88, P = 0.001), and physical activity (HR = 0.78, 95% CI = 0.68–0.89, P <0.001). Conversely, long-term consumption of whole dairy products was identified as a risk factor for PD (HR = 1.31, 95% CI = 1.12–1.53, P = 0.001). Additionally, in the environmental category, the only significant risk factor observed was pesticide exposure (HR = 1.59, 95% CI = 1.22–2.09, P = 0.001) [Supplementary Figures 2–8, https://links.lww.com/CM9/C44]. The medical history encompassed the highest number of risk factors, totaling 18 [Supplementary Figures 9–32, https://links.lww.com/CM9/C44, https://links.lww.com/CM9/C45], while the long-term medication only featured one protective factor (long-term use of ibuprofen: HR = 0.70, 95% CI = 0.58–0.85, P <0.001). Furthermore, within the category of biochemical indicators, one protective factor was identified (low density lipoprotein cholesterol [LDL-C]: HR = 0.91, 95% CI = 0.83–1.00, P = 0.040), while the other factors category contained three risk factors, namely the duration of estrogen therapy (HR = 1.13, 95% CI = 1.08–1.17, P <0.001), hyperechoic substantia nigra (SN) (HR = 8.16, 95% CI = 4.10–16.2, P = 0.00), and family history (HR = 1.80, 95% CI = 1.42–2.30, P <0.001). The remaining 30 factors in our study, however, did not exhibit statistical significance [Supplementary Figure 33, https://links.lww.com/CM9/C45]. Although our study identified SN hyperechogenicity as having the highest HR for the risk of PD, this conclusion is drawn from the analysis of merely three articles. Future research with more extensive datasets is required to validate this observation.Figure 1: Twenty-eight identified modifiable factors associated with PD in the meta-analysis. The following risk factors are classified into three categories: coffee drinking, including coffee, male and female coffee intake; full fat dairy products, including whole dairy products and whole milk; and viral hepatitis, including HBV, HCV, and HBV + HCV. HBV: Hepatitis B virus; HCV: Hepatitis C virus; IBD: Inflammatory bowel disease; IBS: Irritable bowel syndrome; LDL-C: Low density lipoprotein cholesterol; PD: Parkinson's disease.To provide prioritized guidance for the development of subsequent strategies in preventing PD, we identified nine Class I factors: hyperechoic SN and bipolar disorder (BD) (HR = 3.25, 95% CI = 1.31–8.07, P = 0.011), constipation (HR = 2.66, 95% CI = 1.20–5.87, P = 0.015), cognitive impairment (HR = 2.15, 95% CI = 1.86–2.48, P <0.001), depression (HR = 2.13, 95% CI = 1.76–2.59, P <0.001), family history of PD and pesticides (HR = 1.59, 95% CI = 1.22–2.09, P = 0.001), migraine (HR = 1.52, 95% CI = 1.27–1.82, P <0.001), and sleep disorders (HR = 1.51, 95% CI = 1.11–2.05, P = 0.009). Among these, pesticides, constipation, depression, BD, sleep disorders,[5] cognitive impairment, and migraines are intervenable. The remaining 14 factors were classified as Class II. Furthermore, all five identified protective factors, including physical activity, smoking, coffee drinking, ibuprofen, and LDL-C, are also deemed to be modifiable [Figure 1 and Supplementary Table 3, https://links.lww.com/CM9/C43]. Although smoking, physical activity, LDL-C, type 2 diabetes, hypertension, depression, BD, Sjogren's syndrome, and cardiovascular diseases exhibit heterogeneity (I2 ≥ 75%), no evidence of publication bias was observed (Begg's test, P >0.05) [Supplementary Figures 2–32, https://links.lww.com/CM9/C44, https://links.lww.com/CM9/C45]. The results of the sensitivity analysis are given in Supplementary Figures 34–42, https://links.lww.com/CM9/C45. Despite the Begg's test suggesting a potential bias for cognitive disorder, the results remained stable after applying the shear compensation method [Supplementary Table 4, https://links.lww.com/CM9/C43]. Sensitivity analysis also demonstrated the relative stability of cognitive disorder results [Supplementary Figure 42, https://links.lww.com/CM9/C45]. Furthermore, to identify the sources of heterogeneity among the factors reported in more than ten studies, including hyperlipidemia, depression, cerebrovascular diseases, hypertension, diabetes, and chronic kidney disease, subgroup and regression analyses were conducted based on countries/regions, research types, and follow-up durations. The subgroup analyses results indicated that the countries/regions, research types, and follow-up durations contributed to the heterogeneity of these factors to a certain extent [Supplementary Figures 43–54, https://links.lww.com/CM9/C45]. However, the regression analyses did not pinpoint the specific source of this heterogeneity [Supplementary Table 5, https://links.lww.com/CM9/C43]. Additionally, a total of 53 studies, including 46 factors, were systematically reviewed [Supplementary Figure 55, https://links.lww.com/CM9/C45]. Restless leg syndrome, hypothyroidism, and β receptor blockers exhibited an increased risk for PD, with two supporting studies validating their respective risk effects. Behçet's disease, ankylosing spondylitis, and open-angle glaucoma also demonstrated an elevated risk for PD; however, for each of these conditions, one study supported the increased risk, while another study did not confirm this association. Furthermore, additional risk factors such as post-traumatic stress disorder (PTSD), elevated fibrinogen levels, and Helicobacter pylori infection were each supported by only one study in their association with an increased risk of PD, suggesting that further validation is necessary. Further details are given in Supplementary Figure 55, https://links.lww.com/CM9/C45. To our knowledge, our study represents the comprehensive synthesis of prospective studies on PD, characterized by reduced biases and confounding factors compared with case-control and cross-sectional designs. A notable contribution of our research lies in the systematic categorization of risk factors, enabling identification of the most significant risk and protective factors. This prioritized direction facilitates subsequent development of PD prevention strategies. However, we acknowledge the limitations of our study. Given the extensive literature incorporated into our study, there is a possibility that some relevant articles may have been inadvertently omitted, potentially impacting the robustness of our results. Our findings may be influenced by heterogeneity in factors such as age, gender, follow-up duration, diagnostic criteria for PD, and study region; thus affecting the applicability and precision of the results. Additionally, we recognize that solely classifying risk factors based on HR magnitude without adequately considering differences in exposure units may constrain our ability to deeply interpret the study findings. In conclusion, through this meta-analysis and system review, we have identified nine class I risk factors and five protective factors for PD. These finding provide clinicians and stakeholders with comprehensive prevention recommendations. We strongly advocate the critical necessity of personalized strategies for PD prevention, which should be meticulously designed considering the diverse range of risk and protective factors. The potential for early intervention, facilitated by prompt identification and management of PD risk factors, has the capacity to delay or even arrest disease progression. Acknowledgments We would like to express our gratitude to all the researchers who have provided assistance for our study. We sincerely thank all the participants and researchers involved in the prospective study. Funding This study was supported by grants from the National Natural Science Fund of Sichuan (No. 2022NSFSC0749), the 1·3·5 project for disciplines of excellence–Clinical Research Fund, West China Hospital, Sichuan University (No. 2023HXFH032), and the Science and Technology Bureau Fund of Sichuan Province (No. 2023YFS0269). Conflicts of interest None