Unraveling the link: Is Parkinson's disease a causal factor for glaucoma? Insights from Mendelian randomization

Abstract

Background

Glaucoma is recognized as the second-leading cause of complete blindness in developed countries and a significant contributor to irreversible vision loss worldwide. Understanding the potential genetic links between neurodegenerative diseases, such as Parkinson's disease, and glaucoma is crucial for developing preventive strategies.

Methods

This study utilized data from Genome-Wide Association Studies databases, focusing on European populations without gender restrictions. Single nucleotide polymorphisms (SNPs) were selected as instrumental variables based on criteria ensuring no direct association with outcomes and independence from confounding factors. Mendelian randomization analyses were conducted to assess the relationship between SNP effect sizes for Parkinson's disease and the risk of developing glaucoma. Finally, detailed information on high-risk SNP loci was queried and compared in the Ensembl gene database to further explore the relationship between the two diseases.

Results

The analysis revealed that for every one standard deviation increase in the SNP effect size for Parkinson's disease, the risk of developing glaucoma increased by 0.3% to 11.9%. Pleiotropy testing indicated a p-value of 0.464, suggesting that genetic pleiotropy is unlikely to affect the results. Sensitivity analyses confirmed the stability of the findings. A notable observation is the proximity between the loci of both diseases on chromosome 7. However, this positional overlap alone, without functional validation, does not confirm biological relevance.

Conclusion

The study provides suggestive evidence of a potential causal relationship, suggesting that genetic factors associated with Parkinson's disease may potentially contribute to an increased risk of glaucoma. These findings tentatively highlight the need to consider neurodegeneration-related mechanisms in glaucoma management and prevention, though further validation is required.

Keywords

Glaucoma Parkinson's disease Mendelian randomization single nucleotide polymorphisms neurodegenerative diseases GWAS

Background

In developed countries, glaucoma is considered the second most common cause of complete blindness and a major cause of irreversible vision loss globally.¹ Glaucoma can impact patients’ economic status and educational opportunities, ultimately diminishing their quality of life and increasing their mortality risk.² Therefore, early detection and treatment of glaucoma are crucial.³ As of 2020, the number of glaucoma patients was 76 million, with projections indicating that this number will nearly double to 112 million over the next 20 years.⁴

Parkinson's disease (PD) is a progressive neurodegenerative disorder marked by the selective degeneration of dopaminergic neurons in the nigrostriatal pathway. Retrospective studies have suggested that PD patients may present with glaucomatous-like visual field defects.^5,6 Optical coherence tomography has revealed that peripapillary retinal nerve fiber layer (RNFL) thickness can be reduced in PD patients to varying degrees.⁷ Historical postmortem examination indicated that PD patients exhibit low dopamine levels in the retina, analogous to the brain's pathology. Additionally, retinal amacrine cells are lost in animal models of PD, reflecting changes similar to those in glaucoma.^8,9 However, large-scale clinical studies have yet to establish a definitive link between PD and an increased risk of glaucoma. Researchers in this field have noted challenges in obtaining patient consent and managing sensitive data due to personal information concerns.¹⁰

The advent of Mendelian randomization (MR) has significantly reduced the difficulty of assessing causal relationships and risk evaluation in disease research¹¹ while circumventing ethical challenges.¹² This method allows for a more rigorous examination of causal relationships between two or more factors by minimizing confounding variables.¹³ We have previously utilized this approach to explore complex disease relationships successfully.¹¹ Therefore, this study aims to investigate whether PD increases the risk of developing glaucoma through MR, seeking to understand the underlying mechanisms and provide better theoretical guidance for future glaucoma prevention in Parkinson's patients. The detailed procedure is illustrated in Figure 1.

Figure 1.

The causal relationship between Parkinson's disease and glaucoma can be further validated through Mendelian randomization studies, which can help exclude the impact of confounding factors. These confounding factors may include the direct effects of medications such as Levodopa on retinal neurons.

Method

We conducted an MR analysis to investigate the potential causal relationship between PD and the risk of developing glaucoma. The MR approach utilizes genetic variants as instrumental variables to assess the causal effect of PD on glaucoma, effectively controlling for confounding variables. All statistical analyses were performed using R programming language (version 4.3.3), with the aid of specialized packages including TwoSampleMR (version 0.5.11) and ggplot2 (version 3.5.0). The study protocol and details were not preregistered.

Date source

We obtained genome-wide association study (GWAS) datasets for Glaucoma (GWAS ID: GCST90011766) and PD (GWAS ID: ieu-b-7) from the Integrative Epidemiology Unit's OpenGWAS (IEU OpenGWAS) project website (https://gwas.mrcieu.ac.uk), which provides datasets with the largest available sample sizes. The raw data can be accessed through the associated publications listed on PubMed. Parkinson's disease data retrieval was conducted on 4 December 2024, and the Glaucoma data was obtained on 1 August 2024. Both datasets were comprised of European populations and did not impose gender restrictions. The two GWAS datasets were derived from distinct research cohorts; thus sample independence is defaulted. The glaucoma dataset included 14,219,919 single nucleotide polymorphisms (SNPs), while the PD dataset contained 17,891,936 SNPs.

Mendelian randomization assumptions

Mendelian randomization analyses rely on three core assumptions to infer causal relationships, all of which were validated in the present study.

Instrumental variable criteria (relevance)

Instrumental variables must be independent of confounding factors that affect both the exposure (PD) and outcome (glaucoma). The criteria for selecting SNPs as instrumental variables were as follows:

No Direct Association with Outcome: The instrumental variables were not directly linked to the outcome but exerted their effect exclusively through the exposure, thereby ensuring the absence of genetic pleiotropy. A pleiotropy test was conducted, with a result of p ≥ 0.05 signifying no genetic pleiotropy.

Independence from Confounding Factors: The instrumental variables were not associated with unmeasured confounding variables. Given that MR-selected SNPs are based on random allele inheritance from parents to offspring, their susceptibility to environmental and postnatal confounding factors are minimized. The presence of confounding factors can be assessed using the intercept of the MR-Egger regression line with the y-axis. If the intercept value is less than 0.05, it indicates that there is no significant confounding interference in the study.

Strong Correlation with Exposure: The instrumental variables were highly correlated with the exposure, with an F = beta²/se² > 10 as a strong correlation criterion. This ensures that the selected SNPs have sufficient statistical power to proxy the exposure (PD) in the MR analysis.¹⁴

Single nucleotide polymorphism selection (independence)

Instrumental variables must be independent of confounding factors that affect both the exposure (PD) and outcome (glaucoma). Meaningful SNPs were selected from the PD GWAS summary data using a significant threshold of P < 5 × 10⁻⁸. To ensure SNP independence, a linkage disequilibrium (LD) threshold of r² < 0.001 and an LD window of 10,000 kb were applied, reducing the potential for genetic pleiotropy.¹⁵ Single nucleotide polymorphisms associated with glaucoma were subsequently cross-referenced with the PD GWAS summary data, using an r² threshold > 0.8 to confirm result accuracy. Single nucleotide polymorphisms that were missing from the datasets were excluded to ensure the accuracy of this research.¹⁶ Figure 2 displays the Manhattan plot for the selected Glaucoma GWAS data.

Figure 2.

The Manhattan plot for the selected glaucoma Genome-Wide Association Study (GWAS) data. The x-axis denotes chromosomal positions (1–22 and 23 representing the X chromosome), with each dot representing a single nucleotide polymorphism (SNP). The y-axis shows -log10(P-values) from GWAS associations, where higher values indicate stronger statistical evidence for association with glaucoma risk. These significant SNPs help identify genetic variants relevant for studying causal relationships in Mendelian randomization (MR) analysis.

Causal relationship verification (exclusion restriction)

Instrumental variables must influence the outcome (glaucoma) exclusively through the exposure (PD), with no direct effect or pleiotropic pathways. To investigate the causal relationship between PD and Glaucoma using SNPs as instrumental variables, we applied four regression models: MR-Egger regression, the weighted median estimator (WME), the inverse-variance weighted (IVW) random-effects model, and weighted mode regression. The IVW method calculates causal effect estimates directly from summary statistics, eliminating the need for individual-level data.¹⁷ MR-Egger regression assesses the correlation between each SNP and PD, as well as between each SNP and Glaucoma, by fitting a linear model.¹⁸ The WME regression provides a causal effect estimate that is robust to up to 50% invalid instrumental variables by taking the median of the ratios weighted by the inverse of their variances.¹⁹ The weighted mode regression estimates the causal effect by identifying the most frequent causal estimate among SNPs, being robust to the majority of invalid instruments, with weights based on each SNP's precision.²⁰ The IVW method utilizes the information from all instrumental variables, weighting their effects by the inverse of their variances, and generally offers higher statistical efficiency and lower standard errors (SEs) under the assumption that all instrumental variables are free from pleiotropy.²¹ A p-value <0.05 from the IVW regression indicates a statistically significant causal relationship between the two factors investigated. Specifically, a regression slope > 1 suggests an increased disease risk, whereas a slope < 1 implies a reduced risk. In addition, we will also retest for pleiotropy using MR-PRESSO. The number of repetitions will be set to 3000, with an increase when the final number of SNPs obtained is small. If no outlier data emerges, it suggests that there is no significant pleiotropy in the results of this study.²² And heterogeneity assessments will be conducted. For the degrees of freedom corresponding to the final acquired data, a resulting p-value greater than 0.05 will be taken as evidence indicating the absence of heterogeneity in the results. Finally, sensitivity analysis was conducted using the leave-one-out method.

Results

Single nucleotide polymorphism information screening results

In the GWAS data for PD, we initially obtained 17,891,936 SNPs. After applying the aforementioned selection criteria, 23 SNPs remained. When these were combined with the 14,219,919 SNPs from glaucoma data, the number of SNPs remained at 23. Detailed information is provided in Table 1.

Table 1.

Overview of filtered GWAS data.

Number	SNP	CHR	BP	A1	Beta	SE	F
1	rs10451230	17	16035225	T	−0.0259	0.013	3.969
2	rs10513789	3	182760073	G	−0.0094	0.0161	0.341
3	rs10847864	12	123326598	T	0.0352	0.0145	5.893
4	rs12934900	16	30923602	T	−0.003	0.0134	0.050
5	rs144814361	10	121410917	T	0.0519	0.0585	0.787
6	rs329647	11	133764666	C	−0.0035	0.0141	0.062
7	rs34311866	4	951947	C	−0.0312	0.0181	2.971
8	rs35265698	6	32561334	G	−0.0447	0.0182	6.032
9	rs356203	4	90666041	T	−0.0056	0.0133	0.177
10	rs35749011	1	155135036	A	−0.0443	0.0658	0.453
11	rs4488803	3	58218352	A	0.0047	0.0132	0.127
12	rs4588066	18	40672964	A	0.0094	0.0136	0.478
13	rs4613239	2	169119609	G	−0.0127	0.0196	0.420
14	rs4698412	4	15737348	A	−0.0013	0.0129	0.010
15	rs4774417	15	61993702	A	0.0173	0.0143	1.464
16	rs58879558	17	44095467	C	−0.052	0.0162	10.303
17	rs620490	8	16697579	G	−0.0088	0.0142	0.384
18	rs6741007	2	135537119	G	−0.0123	0.0131	0.882
19	rs75505347	12	40885549	T	0.0218	0.052	0.176
20	rs75646569	5	60345424	G	0.0138	0.0214	0.416
21	rs7695720	4	77183300	C	−0.0267	0.0157	2.892
22	rs823106	1	205656453	C	0.0055	0.0187	0.087
23	rs858295	7	23245569	G	0.0065	0.0132	0.242

SE: standard error; SNP: single nucleotide polymorphism number; GWAS: Genome-Wide Association Study; CHR: chromosome number; BP: location; A1: effector allele.

Causal relationship verification

The results of the four regression models are presented in Table 2. The p-value for the IVW method is 0.039, which is below the 0.05 threshold. In contrast, the p-values for both the WME and Weighted Mode regression methods are above 0.05. This discrepancy may be attributed to the assumptions of the IVW method, which presumes that all instrumental variables are free of pleiotropy and are valid. This suggests the possibility of underlying pleiotropy or insufficient instrumental variable efficacy. However, pleiotropy testing conducted using R software yielded a p-value of 0.464, which is significantly greater than 0.05, indicating that pleiotropy is unlikely to affect the results. And when the number of iterations was set to 10,000 in the MR-PRESSO test, no outlier data was detected. This could potentially be attributed to the relatively small number of SNPs ultimately obtained in each group. Nevertheless, it also provides evidence that there is no significant pleiotropy in the results of this study. Heterogeneity among instrumental variables was evaluated, which yielded a p-value of 0.149 with 22 degrees of freedom, indicating no statistically significant heterogeneity across the selected SNPs. This finding is consistent with results from leave-one-out sensitivity analysis, which revealed no substantial deviations in the overall causal estimate when individual SNPs were sequentially excluded. Therefore, the IVW results are considered reliable. Based on the IVW results, there is suggestive evidence that for every one standard deviation increase in the SNP effect size for PD, the risk of developing glaucoma may increase by 0.3% to 11.9%. However, this finding is not corroborated by other MR models, indicating preliminary and inconsistent support. The scatter plot is shown in Figure 3.

Figure 3.

Scatter plot illustrating the relationship between instrumental variable effects on Parkinson's disease (exposure) and glaucoma (outcome) in Mendelian randomization analysis. Each dot represents a single nucleotide polymorphism (SNP), with the x-axis denoting the SNP-specific effect size on Parkinson's disease and the y-axis representing the corresponding effect size on glaucoma. Lines of different colors correspond to results from different Mendelian randomization methods, where the slope of each line quantifies the causal effect of Parkinson's disease on glaucoma: a slope greater than 1 indicates that Parkinson's disease may increase the risk of glaucoma. In contrast, a slope less than 1 suggests that Parkinson's disease may reduce the risk of glaucoma. The blue line represents the Mendelian randomization (MR)-Egger regression result, which is nearly parallel to the x-axis and intersects the y-axis at approximately 0.01, a value below the conventional significance threshold of 0.05. This indicates that the study is unlikely to be significantly affected by genetic pleiotropy.

Table 2.

Four methods MR regression model results.

Four methods MR regression model results
Method	Β	SE	OR(95% CI)	p
MR-Egger	0.009	0.070	1.009(0.879–1.159)	0.895
WME	0.054	0.038	1.056(0.980–1.137)	0.152
IVW	0.058	0.028	1.060 (1.003–1.119)	0.039
Weighted mode	0.038	0.055	1.039 (0.933–1.156)	0.500

SE: standard error; WME: weighted median estimator; IVW: inverse-variance weighted; MR: Mendelian randomization.

Sensitivity analysis

Through sensitivity analysis by sequentially removing individual SNPs, it is observed that the overall effect size is positioned to the right of the y-axis, indicating a value greater than 0. Most outcome indicators are also to the right of the y-axis, suggesting that the conclusions of this study are relatively stable. The sensitivity analysis is shown in Figure 4.

Figure 4.

Results of the sensitivity analysis. The aggregated results are situated to the right of the y-axis and are generally consistent with the IVW regression line results.

Heatmap

The heatmap uses SNPs as the x-axis, beta values as the y-axis, and SEs as the filling criteria, where blue indicators represent higher precision and reliability. Based on the previously discussed regression results, we focus on the blue indicators above the x-axis. These SNPs, ranked from highest to lowest beta value, are rs10847864, rs4774417, rs4588066, rs858295, and rs4488803. These 5 out of 23 loci may be key points associated with an increased risk of glaucoma in PD patients, and further investigation into their detailed information will be conducted. The heatmap is shown in Figure 5.

Figure 5.

Mendelian randomization studies heatmaps. Single nucleotide polymorphisms (SNPs) are plotted on the x-axis, Beta values on the y-axis, with standard error (SE) as the filling criterion. Our main focus is on the blue markers and the results above the x-axis in conjunction with the regression analysis.

Discussion

As previously mentioned, PD is a progressive neurodegenerative disorder of the central nervous system that can also affect the retina.²³ Given that the brain and retina share a common embryonic origin, over 80% of PD patients experience declines in spatial contrast, visual acuity, and color perception, which are associated with disease progression.^24–26 With advancements in optical imaging technologies, researchers can now more easily observe changes in the fine structure of the retina in both PD and glaucoma patients. However, distinguishing retinal changes in PD from those in glaucoma can still be challenging, and the causal relationship between the two is difficult to elucidate through simple trials, particularly as the clinical staging of PD alone is already complex.^27,28 Additionally, the use of levodopa in PD patients complicates clinical studies further due to its effects on the optic nerve.^29–31 Nevertheless, the advent of MR methods has made this research more feasible and the conclusions more reliable.

Our MR analysis provides potential, albeit weak and inconsistent, evidence that PD may be associated with increased glaucoma susceptibility: the IVW method suggests a 0.3%–11.9% increase in glaucoma risk per one standard deviation increase in PD-related SNP effect size, but this is not supported by other MR models. We also identified 5 out of 23 SNP loci involved: rs10847864, rs4774417, rs4588066, rs858295, and rs4488803. Detailed information was obtained through the Ensembl Genome Browser (http://asia.ensembl.org/index.html). Notably, the rs858295 locus (from PD GWAS data) and the glaucoma-associated SNP rs113432289 are both mapped to chromosome 7p15.3, with close chromosomal positions as shown in Figure 6. The rs858295 SNP was sourced from our PD GWAS data (GWAS ID: ieu-b-7), while rs113432289 was identified in a prominent genetic study by Zhou et al.³² in 2018. However, without functional annotations, colocalization evidence, or Expression Quantitative Trait Locus data, this proximity alone does not confirm biological relevance, and their potential role as “key nodes” remains speculative. It is worth noting that these loci do not appear to play a major role in their respective diseases, which seems to corroborate the 0.3%–11.9% causal relationship range found in our MR study.

Figure 6.

Specific information on single nucleotide polymorphism (SNP) loci identified in this study compared to common glaucoma-associated SNPs obtained from the Ensembl website reveals that they are located in the 7p15.3 region of chromosome 7 and are proximate to each other.

Previous research has shown that PD patients exhibit thinning of the macular RNFL, thinning of the ganglion cell layer, and reduced superficial and deep vascular density,^33–35 which closely resembles the retinal changes observed in glaucoma patients.³⁶ In PD, dopamine dysregulation affects the transmission in retinal amacrine and ganglion cells, leading to retinal damage.³⁷ Conversely, glaucoma is often associated with elevated intraocular pressure,³⁸ and whether there is a direct link between the two conditions may warrant further investigation. Notably, some researchers have identified similarities between the disease processes of PD and normal-tension glaucoma,³⁹ which may help to contextualize the aforementioned observations.

For further research, we observed that Wang et al.⁴⁰ in a 2024 MR study found that the increased levels of inflammation are associated with an increased risk of PD, and identified Galectin-3, Haptoglobin (HP), and Holo-Transcobalamin-2 as potential predictors for PD onset. We obtained data from the provided GWAS IDs and conducted a study on their correlation with glaucoma. The GWAS IDs for Galectin-3, HP, and Holo-Transcobalamin-2 are ebi-a-GCST90012009, prot-a-1369, and prot-a-2939, respectively. The specific procedures for SNP selection and validation were consistent with those employed in the present study. Our analysis found that none of these markers were associated with an increased risk of glaucoma, as detailed in Table 3. The results do not support a shared neuroinﬂammatory mechanism based on the markers tested, though this does not preclude other forms of overlap. Therefore, understanding the precise mechanism by which PD might lead to glaucoma may require larger-scale MR studies in the future.

Table 3.

The results of the MR analysis on inflammation levels.

Results of the MR analysis on inflammation levels
Method	Exposure	OR(95% CI)	p
MR-Egger	Galectin-3	1.008(0.916–1.108)	0.880
	HP	0.924(0.852–1.001)	0.149
	Holo-Transcobalamin-2	1.014(0.908–1.132)	0.829
WME	Galectin-3	0.997(0.935–1.062)	0.922
	HP	0.975(0.941–1.012)	0.181
	Holo-Transcobalamin-2	1.022(0.986–1.059)	0.225
IVW	Galectin-3	0.985 (0.929–1.043)	0.604
	HP	0.982(0.947–1.017)	0.307
	Holo-Transcobalamin-2	1.018(0.985–1.052)	0.281
Weighted mode	Galectin-3	0.999 (0.937–1.066)	0.986
	HP	0.975(0.940–1.011)	0.247
	Holo-Transcobalamin-2	1.025(0.987–1.065)	0.286

HP: haptoglobin; WME: weighted median estimator; IVW: inverse-variance weighted; MR: Mendelian randomization.

A potential limitation of this study is that PD, as a complex neurodegenerative disorder, lacks well-defined, modifiable interventions in a conventional sense. As noted by Hernán and Taubman,⁴¹ causal inference requires clear interventional interpretations, which are challenging here given the absence of specific strategies to alter PD status. This may restrict the direct translational utility of our MR findings for clinical interventions.

Another limitation is that the core assumption of MR, that genetic instruments for PD are unrelated to confounders of its relationship with glaucoma, may be challenged. Both conditions share neurodegenerative pathways and complex etiologies, as indicated by adjacent SNP loci on chromosome 7. This overlap raises concerns that genetic instruments might be associated with unmeasured confounders, requiring more extensive validation to fully justify the assumption.

A third limitation is the absence of a preregistration protocol and formal study prespecification. Key methodological decisions, including SNP inclusion thresholds, data harmonization approaches, and outlier handling procedures, were determined during the analytical process rather than being predefined. While we strived to adhere to standardized MR practices, the lack of preregistration introduces potential for selection bias in methodological choices, which may affect the reproducibility and interpretability of our findings. This underscores the need for caution when generalizing the results, as prespecified protocols help mitigate such biases in observational research.

A further limitation lies in the divergent results across different MR models. Only the IVW model indicated a significant association, whereas no consistent findings were observed in other robust models. This discrepancy may stem from the strict assumptions underlying the IVW model, including the absence of pleiotropy and confounding, alongside potential limitations in the strength of our instrumental variables. As shown in Table 1, the F-statistics for most selected SNPs are below 10, suggesting that these genetic variants may have a less robust correlation with PD than initially expected. While this does not entirely invalidate the relevance assumption, it indicates that the instrumental variables might be less powerful in proxying the exposure, which could influence the stability of causal estimates—particularly in the IVW model, which is more sensitive to such nuances. If unrecognized pleiotropic SNPs, residual confounding factors, or the aforementioned limitations in instrument strength are present, they could collectively affect the IVW results. Conversely, the null results from robust models suggest that the conclusion that PD constitutes a high-risk factor for glaucoma should be interpreted with caution, and the strength and reliability of the causal association would benefit from further validation.

Finally, the results are generalizable to other European populations, yet their applicability to non-European populations requires validation in future studies. The current findings, based on specific exposure data, may need further evaluation across different exposure periods and levels to confirm consistency.

Conclusion

This MR study provides preliminary, inconsistent evidence suggesting that PD may potentially contribute to the risk of developing glaucoma. The IVW method indicates a 0.3%–11.9% increase in glaucoma risk per standard deviation increase in PD-related SNP effects—this association, observed after excluding factors such as age, PD medications, and glaucoma medications, though not validated by other MR models. This preliminary finding tentatively raises the question of whether glaucoma's pathogenesis involves degenerative neurological pathways, though this requires further exploration before clinical reevaluation can be considered. However, it is important to note that the pathogenesis of glaucoma is not a direct replication of PD, as there are some differences between the two. These preliminary findings may generate hypotheses for future research, though they currently lack sufficient evidence to inform clinical diagnosis or treatment strategies.

Footnotes

Acknowledgements

Thanks to Jing Tang for her help in data collection in this study.

ORCID iD

M.M. Bin Lin

Consent for publication

All authors agree to publish their findings in this journal.

Authors contributions

Under the guidance of DKL, BL aligned the research focus with contemporary trends and jointly determined the research direction. Subsequently, MYZ and LLC independently conducted searches in GWAS databases for data on PD and glaucoma. With input from BL and DKL, they selected the most reliable database, taking into account factors such as ethnicity, gender, and literature credibility. After confirming the database, BL executed three rounds of calculations using R programming to ensure result accuracy. The manuscript was drafted under DKL's supervision, while LLC and MYZ reviewed relevant literature to interpret the computational results. Finally, the manuscript was revised based on DKL's feedback before submission.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Xiamen Municipal Bureau of Science and Technology (3502Z20244ZD1192).

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Availability of data and materials

The authors are willing to share all raw data files in this study if necessary.

References

Lin

. The pivotal role of inflammatory factors in glaucoma: a systematic review. Front Immunol 2025; 16: 1577200.

Fahmideh

Marchesi

Barbieri

, et al. Non-drug interventions in glaucoma: putative roles for lifestyle, diet and nutritional supplements. Surv Ophthalmol 2022; 67: 675–696. PubMed PMID: 34563531.

Hasan

Phu

Sowmya

, et al. Artificial intelligence in the diagnosis of glaucoma and neurodegenerative diseases. Clin Exp Optom 2024; 107: 130–146. PubMed PMID: 37674264.

Tham

Y-C

Wong

, et al. Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis. Ophthalmology 2014; 121: 2081–2090.

Bayer

Keller

Ferrari

, et al. Association of glaucoma with neurodegenerative diseases with apoptotic cell death: Alzheimer's disease and Parkinson's disease. Am J Ophthalmol 2002; 133: 135–137. PubMed PMID: 11755850.

Yenice

Onal

Midi

, et al. Visual field analysis in patients with Parkinson's disease. Parkinsonism Relat Disord 2008; 14: 193–198. PubMed PMID: 17888714.

Garcia-Martin

Satue

Fuertes

, et al. Ability and reproducibility of Fourier-domain optical coherence tomography to detect retinal nerve fiber layer atrophy in Parkinson's disease. Ophthalmology 2012; 119: 2161–2167. PubMed PMID: 22749083.

Harnois

Di Paolo

. Decreased dopamine in the retinas of patients with Parkinson's disease. Invest Ophthalmol Vis Sci 1990; 31: 2473–2475. PubMed PMID: 2243012.

Djamgoz

Hankins

Hirano

, et al. Neurobiology of retinal dopamine in relation to degenerative states of the tissue. Vision Res 1997; 37: 3509–3529.

10.

Hasan

Phu

Sowmya

, et al. Artificial intelligence in the diagnosis of glaucoma and neurodegenerative diseases. Clin Exp Optom 2024; 107: 130–146.

11.

Lin

Chen

L-l

D-K

. Mendelian randomization analysis reveals a causal relationship between preterm birth and myopia risk. Front Pediatr 2024; 12: 1404184.

12.

Smith

Ebrahim

. Mendelian randomization: prospects, potentials, and limitations. Int J Epidemiol 2004; 33: 30–42.

13.

Emdin

Khera

Kathiresan

. Mendelian randomization. JAMA 2017; 318: 1925–1926.

14.

Feng

, et al. Pulmonary embolism and 529 human blood metabolites: genetic correlation and two-sample Mendelian randomization study. BMC Genom Data 2022; 23: 69. PubMed PMID: 36038828; PubMed Central PMCID: PMCPMC9422150.

15.

Hemani

Zheng

Elsworth

, et al. The MR-base platform supports systematic causal inference across the human phenome. elife 2018; 7: e34408.

16.

Clarke

Mitchell

Sharp

, et al. Vitamin D levels and risk of juvenile idiopathic arthritis: a Mendelian randomization study. Arthritis Care Res (Hoboken) 2023; 75: 674–681.

17.

Burgess

Bowden

Fall

, et al. Sensitivity analyses for robust causal inference from Mendelian randomization analyses with multiple genetic variants. Epidemiology 2017; 28: 30–42.

18.

Slob

Groenen

Thurik

, et al. A note on the use of Egger regression in Mendelian randomization studies. Int J Epidemiol 2017; 46: 2094–2097.

19.

Bowden

Davey Smith

Haycock

, et al. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 2016; 40: 304–314.

20.

Hartwig

Davey Smith

Bowden

. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int J Epidemiol 2017; 46: 1985–1998.

21.

Wen

M-T

Liang

X-Z

Luo

, et al. Plasma lipids, alcohol intake frequency and risk of osteoarthritis: a Mendelian randomization study. BMC Public Health 2023; 23: 1327.

22.

Lin

Chen

, et al. Deliberate dietary adjustments may not mitigate the progression of glaucoma: a two-sample Mendelian randomization study. Medicine (Baltimore) 2025; 104: e42944. PubMed PMID: 40696589.

23.

Chrysou

Jansonius

van Laar

. Retinal layers in Parkinson's disease: a meta-analysis of spectral-domain optical coherence tomography studies. Parkinsonism Relat Disord 2019; 64: 40–49. PubMed PMID: 31054866.

24.

Archibald

Clarke

Mosimann

, et al. Retinal thickness in Parkinson's disease. Parkinsonism Relat Disord 2011; 17: 431–436.

25.

Inzelberg

Ramirez

Nisipeanu

, et al. Retinal nerve fiber layer thinning in Parkinson disease. Vision Res 2004; 44: 2793–2797.

26.

Wang

Zhang

, et al. Retinal microvascular impairment in Parkinson's disease with cognitive dysfunction. Parkinsonism Relat Disord 2022; 98: 27–31.

27.

Chrysou

Heikka

van der Zee

, et al. Reduced thickness of the retina in de novo Parkinson’s disease shows A distinct pattern, different from glaucoma. J Parkinsons Dis 2024; 14: 507–519.

28.

van der Zee

Kanel

Gerritsen

, et al. Altered cholinergic innervation in de novo Parkinson's disease with and without cognitive impairment. Mov Disord 2022; 37: 713–723.

29.

Yavas

Yilmaz

Küsbeci

, et al. The effect of levodopa and dopamine agonists on optic nerve head in Parkinson disease. Eur J Ophthalmol 2007; 17: 812–816.

30.

Warren Olanow

Obeso

. Levodopa toxicity and Parkinson disease: still a need for equipoise. Minneapolis, MN and Hagerstown, MD: AAN Enterprises, 2011.

31.

Sen

Tugcu

Coskun

, et al. Effects of levodopa on retina in Parkinson disease. Eur J Ophthalmol 2014; 24: 114–119.

32.

Zhou

Nielsen

Fritsche

, et al. Efficiently controlling for case-control imbalance and sample relatedness in large-scale genetic association studies. Nat Genet 2018; 50: 1335–1341.

33.

Archibald

Clarke

Mosimann

, et al. Visual symptoms in Parkinson's disease and Parkinson's disease dementia. Mov Disord 2011; 26: 2387–2395.

34.

Hart de Ruyter

Morrema

den Haan

, et al. α-Synuclein pathology in post-mortem retina and optic nerve is specific for α-synucleinopathies. NPJ Parkinsons Dis 2023; 9: 124.

35.

Ortuño-Lizarán

Beach

Serrano

, et al. Phosphorylated α-synuclein in the retina is a biomarker of Parkinson's disease pathology severity. Mov Disord 2018; 33: 1315–1324.

36.

Nouri-Mahdavi

Nowroozizadeh

Nassiri

, et al. Macular ganglion cell/inner plexiform layer measurements by spectral domain optical coherence tomography for detection of early glaucoma and comparison to retinal nerve fiber layer measurements. Am J Ophthalmol 2013; 156: 1297–1307.e2.

37.

Garcia-Martin

Satue

Fuertes

, et al. Ability and reproducibility of Fourier-domain optical coherence tomography to detect retinal nerve fiber layer atrophy in Parkinson's disease. Ophthalmology 2012; 119: 2161–2167.

38.

Coleman

Miglior

. Risk factors for glaucoma onset and progression. Surv Ophthalmol 2008; 53: S3–S10.

39.

Lin

I-C

Wang

Y-H

Wang

T-J

, et al. Glaucoma, Alzheimer's disease, and Parkinson's disease: an 8-year population-based follow-up study. PLoS One 2014; 9: e108938.

40.

Wang

Shi

Yin

, et al. Association between neuroinflammation and Parkinson’s disease: a comprehensive Mendelian randomization study. Mol Neurobiol 2024: 1–11.

41.

Hernán

Taubman

. Does obesity shorten life? The importance of well-defined interventions to answer causal questions. Int J Obes (Lond) 2008; 32: S8–S14. PubMed PMID: 18695657.