Residential Greenspace and Domestic Gardens, Genetic Predisposition, and Incident Glaucoma: A Prospective Cohort Study Authors: Xinyu Zhang, Changjing Han, Xiaojing Liu, Ting Lu, Zixuan Zhang, Xinyu Zhu, Yikeng Huang, Chenxin Li, Xun Xu, Yong Sun, Shuzhi Zhao, Zhi Zheng Published: 2026 ABSTRACT AbstractWhile greenness has been linked to enhanced physiological and psychological health, its association with glaucoma remains unexplored. This study aims to investigate the effect of residential greenspace and domestic gardens on incident glaucoma. 357, 920 participants from the UK Biobank were included. Land use coverage percentages of greenspace and domestic gardens within 300 and 1,000 m buffers of participants' residential locations were estimated using land use databases. Cox regression models were used to estimate the association between greenness exposures and incident glaucoma. Four‐way decomposition models were used to explore potential mediators or moderators in the pathway. Gene‐environment interaction effects were tested using multi‐trait genetic polygenic risk score of glaucoma and Myocilin p.Gln368Ter genotype. During a median (IQR) follow‐up of 13.5 (12.8, 14.1) years, 4, 046 glaucoma cases occurred. After full adjustment, the fourth quartile of greenspace was associated with 17.7% (HR = 0.823, 95%CI 0.723–0.937, p = 0.003) and 19% (HR = 0.810, 95%CI 0.708–0.927, p = 0.002) lower risk of incident glaucoma in 300 and 1000 m buffer. Domestic gardens showed a similar trend (4th quartile vs. 1st quartile: 300 m HR = 0.843, 95%CI 0.742–0.957, p = 0.008; 1,000 m HR = 0.827, 95%CI 0.727–0.941, p = 0.004), though restricted cubic spline analysis showed a nonlinear association in 300 m buffer. Greenspace in 300 m buffer was associated with lower cornea‐compensated intraocular pressure. No significant mediating or interactive effects were identified for air pollution, recreational screen time, physical activities, or genetic predisposition. Residential greenspace and domestic gardens may be protective against glaucoma. Therefore, greenspace coverage should be incorporated into land use planning policies for public ocular health promotion. Introduction Glaucoma, the leading cause of irreversible blindness worldwide, is currently affecting approximately 76 million individuals, with the figure estimated to rise to 112 million by 2040 (Bourne et al., 2017). Current treatment strategies can only decelerate but not restore vision loss in glaucoma (Jayaram et al., 2023), and yet approximately 40% of affected individuals progress to unilateral blindness over their lifetime (Tribble et al., 2023). Therefore, finding new potential targets for this complex and multifactorial disease for early prevention is critical. Apart from unmodifiable risk factors, such as age, ethnicity, and genetic predisposition (Hollands et al., 2013; Wang et al., 2022), a few recent studies have demonstrated the potential role of modifiable environmental determinants in glaucoma pathogenesis (Lee et al., 2024). For example, air pollutants were found to confer excess risk to primary open‐angle glaucoma (Huang et al., 2024). In contrast, enriched environments (EE), a complex combination of social and inanimate stimulation, have demonstrated neuroprotective and anti‐inflammatory properties to protect against glaucoma development and progression in mice (González Fleitas et al., 2021). Residential greenness, including public greenspaces and domestic gardens, not only provides individuals with spaces for recreation and social interaction, but also has been shown to offer multiple psychological and physiological health benefits over the past decades (Yang, Zhao, et al., 2021). For instance, vegetative presence and public greenspaces were associated with decreased risk of developing neurodegenerative diseases, including Parkinson's disease and Alzheimer's dementia (Astell‐Burt et al., 2020; Hu et al., 2024; Klompmaker et al., 2024; Paul et al., 2020). Moreover, greenness exposure was also associated with lower incident cardiovascular events and mortality (Chen et al., 2020; Roscoe et al., 2022), cerebrovascular diseases (Hu et al., 2023; Paul et al., 2020), and better retinal microvascular function (Bauwelinck et al., 2024). Given that glaucoma is a retinal neurodegenerative disease characterized by neurovascular dysfunction (Alarcon‐Martinez et al., 2023; Mancino et al., 2018; Marshall et al., 2021; Wareham et al., 2022), it is plausible that greenspace may also exert a protective effect against glaucoma. However, this association has not been investigated yet. Furthermore, residential greenness has been found to improve public health through reducing air pollution, promoting healthy lifestyle, and other pathways (Bao et al., 2024; Markevych et al., 2017), but the potential mechanisms underlying any association between greenness and incident glaucoma have not been explored either. Using a prospective cohort of 357, 920 participants with over 12 years of follow‐up from the UK Biobank study, we aim to investigate the association between greenspace, domestic gardens, and incident glaucoma. Additionally, we will explore the association between greenness and intraocular pressure (IOP) to elucidate potential underlying physiological mechanisms. Furthermore, we will also use four‐way decomposition models to examine potential mediating and interactive factors, including air pollutants, recreational screen time, and physical activities. Finally, we will test the potential gene‐environment interaction using polygenic risk score (PRS) for glaucoma and Myocilin p.Gln368Ter, the most common rare mutation in European ancestry for primary open angle glaucoma and increased IOP (Craig et al., 2020; Han et al., 2019). Collectively, this study aims to investigate greenspace as a potential modifiable risk factor for glaucoma and explore underlying pathways, thereby providing evidence for public health strategies for glaucoma prevention. Methods Study Population The UK Biobank recruited approximately 500, 000 participants aged 40–69 years at 2006–2010 and collected deep phenotypic data. Touchscreen questionnaire collected sociodemographic characteristics, personal lifestyle, and medical history. Verbal interview assisted by nurses further recorded medical information for better accuracy and details. Physical and eye measures were also performed in recruitment centers to record anthropometrics and physical conditions objectively. Hospital inpatient records with diagnosis and operations were collected with linkage to NHS England, Information and Statistics Division, Scotland, and Secure Anonymized Information Linkage, Wales. Primary care data with diagnosis, prescriptions and operations were available in approximately 45% of the cohort. Details for the UK Biobank study were described elsewhere (Sudlow et al., 2015). Participants with available greenness records were included. Those with prevalent glaucoma at baseline according to touchscreen questionnaire, verbal interview, hospital inpatient record and general practice record were excluded. Participants who changed home location within 3 years before recruitment were also excluded to avoid bias. Those without complete record of covariates were further excluded for complete case analysis. Details for study cohort creation were presented in Figure S1 of Supporting Information S1. The last follow‐up date was defined as the earliest of: first occurrence of incident glaucoma, date of death, date of loss of follow‐up, and the last date of hospital inpatient record update, which was 31 October 2022 in England, 31 August 2022 in Scotland, and 31 May 2022 in Wales. Exposure Assessment Residential greenness exposures were derived using the Generalized Land Use Database 2005 (GLUD 2005) for England based on each participant's home location. Center for Ecology and Hydrology (CEH) Land Cover Map 2007 data were used for participants in Wales and Scotland (Benedict, n.d.). Residential locations were based on the participants' postcodes as held by UK Biobank or provided at the assessment visit, and geocoded by Experian QAS Batch software to ensure accurate verified addresses of high precision. Greenspaces were defined as publicly accessible natural areas, such as parks and recreational fields, and domestic gardens were referred to private green areas, including backyards or gardens. The land use percentage coverage of greenspaces and domestic gardens in 300 and 1,000 m buffer were used to indicate near or wider area greenness relative to participant's home location, and 300 m buffer was adopted as the primary analysis as it was considered the accessibility of residential greenspace, which was in line with previous studies (Hu et al., 2024; Zhang et al., 2022). As the main land use types and proportions remained stable during 2000–2018 in Europe (European Environment Agency, 2019), the baseline greenness measurements served as a proxy for long‐term exposure. Outcome Ascertainment Glaucoma ascertainment was in line with previous studies: (a) hospital inpatient diagnosis: the 9th (365, excluding 3650) and 10th International Disease Classification (H40, excluding H40.0 and H42) of Diseases codes; (b) hospital inpatient operations: Office of Population Censuses and Surveys Classification of Interventions and Procedures 4th edition (OPCS‐4) (C601, C605, C611, C613, C614, and C615); (c)self‐reported medical history: self‐reported glaucoma in touchscreen questionnaire (Field 6148: 2) or verbal interview (Field 20002: 1277); (d) self‐reported operation: self‐reported glaucoma surgery (Field 20004: 1436); (e) self‐reported medication: Read v2 code; and (f) general practice operations, diagnosis and medications: corresponding Read v2 and v3 codes (Table S1 in Supporting Information S1). Intraocular pressure (IOP) lowering medications during follow‐up were excluded from control. Only hospital inpatient records were adopted as incident cases in the primary analysis to avoid potential bias, as only ∼45% of the participants had available general practice record. We further performed a sensitivity analysis incorporating general practice linkage data for incident cases due to the importance of primary care diagnosis (Clifton et al., 2024). In glaucoma subtype analysis, we further defined hospital admissions of primary angle closure glaucoma (PACG: ICD‐10: H402; ICD‐9: 3652) and primary open angle glaucoma (POAG: ICD‐10: H401, H409; ICD‐9: 3651, 3659; excluding PACG and other specified glaucoma). Cornea‐compensated IOP (IOPcc) was adopted in our analyses as such measurement was more associated with glaucoma progression and accounted for corneal biomechanical properties (Susanna et al., 2019). IOPcc was measured for 111, 845 participants for both eyes at initial assessment visit using the Ocular Response Analyzer noncontact tonometer (Reichert Cop.) (Chua et al., 2019). We used mean IOPcc of two eyes or the IOP for the only available one eye in our analysis. Covariates and Potential Mediators/Moderators The selection of confounding covariates and potential mediators was guided by prior knowledge and literature and visually represented in a Directed Acyclic Graph using DAGitty (https://www.dagitty.net) (Figure S2 in Supporting Information S1). Age at recruitment (years), sex (female or male), ethnicity (white or other), education qualification (college/university or below), Townsend deprivation index (1st tertile, 2nd tertile, and 3rd tertile), time spent outdoor (hours per day), smoker status (previous, current or never), alcohol consumption (units per day), physical activity (per SD increase in Metabolic Equivalent Task minutes per week, MET‐min/w), healthy diet (yes or no), body mass index (per SD increase in kg/m2), prevalent diabetes mellitus (yes or no), systolic blood pressure (mmHg), time spent in current address (years), and visual acuity (mean spherical equivalent refractive error averaged between eyes, avMSE, Diopters) were adopted as potential confounders and adjusted in our primary multi‐variable models. Townsend deprivation index was used to describe socioeconomic status, and a higher value denotes more materially deprived. While there was no definitive healthy diet for glaucoma, we defined dietary quality according to a widely recognised criterion for overall health (Zhang et al., 2021). A healthy diet was determined if one meeting at >5 in 10 recommendations, including adequate consumption of fruit, vegetables, whole grains, fish, shellfish, dairy products, vegetable oils, as well as reduced consumption of refined grains, processed meats, unprocessed meats, and sugar sweetened beverages. Physical activity, time spent outdoors, leisure screen time, and exposure to air pollutants were adopted as potential mediators/moderators. MET‐min/week was derived from self‐reported physical activity using an adapted version of the validated International Physical Activity Questionnaire at baseline (Craig et al., 2003). Time spent outdoors was averaged between self‐reported time spent outdoors in summer (Field 1050) and winter (Field 1060). Leisure screen time was the sum of time spent watching television (Field 1070) and time spent using a computer (Field 1080) per day. Prefer not to answer or do not know were recoded as missing, and less than an hour a day was recoded as 0.5 hr a day. Air pollutant estimates were modeled based on each verified residential locations using a Land Use Regression model developed in the European Study of Cohorts for Air Pollution Effects (ESCAPE), and were representative for the year 2010. Nitrogen dioxide (NO2), nitrogen oxides (NOx), particulate matter 10 (PM10), particulate matter 2.5 absorbance (PM2.5 absorbance), and particulate matter 2.5 (PM2.5) were investigated. Genotype Analysis The Affymetrix UK BiLEVE Axiom Array and the Affymetrix UK Biobank Axiom Array generated the genotype array data, and imputation was based on the UK 10K and the Haplotype Reference Consortium reference panels (Bycroft et al., 2018). The PRS for glaucoma was constructed using summary statistics from a recent multi‐trait genome‐wide association analysis of glaucoma (Craig et al., 2020). PRS was calculated using the effect estimates from the original multi‐trait GWAS analysis and a standard weighted sum of individual SNPs as provided by Craig et al. (Craig et al., 2020; Madjedi et al., 2023). Myocilin p. Gln368Ter (rs74315329) genotype was detect from the whole genome sequencing (WGS) data. The WGS data in the UK Biobank was generated via paired‐end sequencing (∼32.5x coverage in average) on Illumina NovaSeq6000 instruments provided by deCODE (Halldorsson et al., 2022). Variants were aligned to GRCh38, and single variant nucleotide polymorphisms were called using GraphTyper (Eggertsson et al., 2017). Myocilin was the first identified pathogenic gene for primary open angle glaucoma, and rs74315329 GA was the most common rare mutation of MYOC gene. It has been found to also be associated with increased IOP as well as normal tension glaucoma (Han et al., 2019). We identified 606 (0.23%) heterozygous carriers of the p.Gln368Ter allele and no homozygous carriers in the cohort of 260, 516 participants with available exposure and all covariate records, which was equivalent to genotype frequency based on imputation data in the UK Biobank (Zebardast et al., 2021). Statistical Analysis Mean (Standard deviation, SD) and number (percentage) were used to describe continuous and categorical variables, respectively. For skewed‐distributed continuous variables, median (interquartile range, IQR) was also used to describe data distribution. Multivariable cox regression model was used to evaluate the association between greenspace/domestic gardens and incident glaucoma. Multivariable linear regression models were used to estimate the association between greenness exposures and cornea‐compensated intraocular pressure. We performed several sensitivity analyses to verify the robustness of our results. All models performed mutual adjustment for greenspace/domestic gardens as greenness exposures were coverage percentages in 300/1000 m buffer. Model 1 was the minimally adjusted model, adjusting for age, sex, and ethnicity. Model 2 further adjusted for years spent in the current address, education, Townsend index, time spent outdoor, smoker status, alcohol consumption, physical activity, healthy diet, body mass index, prevalent diabetes mellitus, systolic blood pressure, and avMSE (for IOPcc). Schoenfeld residuals showed there was no violation to proportional hazards assumption, with p for global test ranging from 0.382 to 0.474. Restricted cubic splines were used to test potential nonlinear association. Four‐way decomposition models were used to estimate the total relative risk ratio and the effect was decomposed into four proportions: (a) neither mediation nor interaction; (b) just interaction; (c) both interaction and mediation; and (d) just mediation. (Discacciati et al., 2019) All potential mediators/moderators were recategorized into binary categorical factors according to WHO guidelines (Geneva: World Health Organization, 2020, 2021) (Physical activity: 1500 MET‐min per week; PM2.5: 10 μg/m3; PM10: 20 μg/m3; NO2, 40 μg/m3) or median levels (leisure screen time: 3.5 hr; PM2.5 absorbance: 1.12 μg/m3; NOx: <41.67 μg/m3). Cox regression was adopted for the outcome model, and logistic regression was adopted for the mediator model. Total effect was estimated comparing the fourth quartile to the first, and proportions of four‐way decomposed effect were computed setting mediator/moderators at optimized levels, that is, more physical activity, and less leisure screen time and air pollutants. Models adjusted for age, sex, ethnicity, years spent in the current address, education, Townsend deprivation index, time spent outdoors, smoker status, alcohol consumption, physical activity, healthy diet, body mass index, prevalent diabetes mellitus, and systolic blood pressure, with covariates set at mean levels when estimating decomposition. All data cleaning and statistical analysis procedures were conducted in Stata (version 18, StataCorp, College Station, TX) and R (version 4.2.1). Results Characteristics of Study Population A total of 357, 920 participants were involved in the current study, and 265, 736 with complete record of covariates were involved in the complete case analysis (Figure S1 in Supporting Information S1). With a median (IQR) follow up of 13.5 (12.8, 14.1) years, 4, 046 glaucoma cases occurred during follow‐up. Baseline characteristics were demonstrated in Table 1. In brief, those with incident glaucoma were older, more likely to be male or a previous smoker, less likely to be white, had lower education qualification and socioeconomic status, drank less alcohol and had a healthier diet, possessed higher BMI, more likely to have diabetes and had higher systolic blood pressure, and spent longer time outdoors and more years in the current location. The median (IQR) levels of green space and domestic gardens at 300 m buffer was 29.78% (17.39%, 48.64%) and 32.34% (21.25%, 42.24%), and were 42.12% (27.81%, 60.32%) and 24.48% (16.65%, 32.16%) at 1,000 m buffer (Table S2 in Supporting Information S1). Table 1Baseline Characteristics of Study PopulationOverallIncident glaucomaP valueNoYesN = 357, 920N = 352, 274N = 5, 646Age at recruitment (mean [SD])57.01 (7.90)56.94 (7.90)61.52 (6.16)<0.001Male (%)160856 (44.9)158311 (44.9)2545 (45.1)0.849Non‐white ethnicity (%)19037 (5.3)18695 (5.3)342 (6.1)0.014Below University/College (%)246166 (68.8)241967 (68.7)4199 (74.4)<0.001Townsend Index (mean [SD])−1.47 (2.97)−1.47 (2.97)−1.40 (3.10)0.097Missing (%)18(0.01)17(0.005)1(0.02)Time spent outdoor, hour/day, (mean [SD])3.93 (2.33)3.93 (2.33)4.22 (2.33)<0.001Missing (%)37, 769(10.55)37, 155(10.55)614(10.87)Smoking status (%)<0.001Never196996 (55.0)194037 (55.1)2959 (52.4) Previous124620 (34.8)122443 (34.8)2177 (38.6) Current34326 (9.6)33859 (9.6)467 (8.3)Missing (%)1978 (0.6)1935 (0.5)43 (0.8)Alcohol consumption, unit/day, (median [IQR])1.21 [0.08, 2.57]1.21 [0.08, 2.57]1.07 [0.00, 2.43]<0.001Alcohol consumption, unit/day, (mean [SD])1.78 (2.15)1.78 (2.15)1.67 (2.12)<0.001Physical activity, MET‐min/week, (median [IQR])1786.50 [813.00, 3600.00]1786.50 [813.00, 3599.00]1836.00 [798.00, 3786.00]0.175Physical activity, MET‐min/week, (mean [SD])2677.90 (2715.26)2676.92 (2714.79)2741.57 (2745.08)0.119Missing (%)71, 933(20.1)70, 629(20.0)1, 304(23.1)Healthy diet (%)33920 (9.5)33265 (9.4)655 (11.6)<0.001Body mass index, kg/m2, (median [IQR])26.70 [24.11, 29.83]26.70 [24.11, 29.83]26.86 [24.26, 30.02]0.009Body mass index, kg/m2, (mean [SD])27.38 (4.75)27.37 (4.75)27.57 (4.87)0.003Missing (%)1, 947(0.54)1, 920(0.55)27(0.48)Prevalent diabetes mellitus (%)21228 (5.9)20681 (5.9)547 (9.7)<0.001Systolic blood pressure (mean (SD))138.16 (18.67)138.10 (18.66)141.72 (18.50)<0.001Missing (%)779(0.22)766(0.22)13(0.23)Years living in current location, (median [IQR])12.15 [8.28, 18.58]12.13 [8.27, 18.55]12.82 [8.92, 20.14]<0.001Years living in current location, (mean [SD])14.66 (9.86)14.64 (9.84)16.07 (10.79)<0.001 Note. SD, Standard deviations. IQR, interquartile range; MET‐min, Metabolic Equivalent Task minutes. Baseline Characteristics of Study Population Note. SD, Standard deviations. IQR, interquartile range; MET‐min, Metabolic Equivalent Task minutes. Prospective Association Between Greenspace, Domestic Garden and Incident Glaucoma The association between greenness and incident glaucoma was demonstrated in Table 2. After full adjustment, per 10% increase in green space (aHR = 0.966, 95%CI 0.943, 0.989, p = 0.004) as well as domestic garden (aHR = 0.948, 95%CI 0.913, 0.983, p = 0.004) were found to be negatively associated with incident glaucoma. Compared with the first quartile, the fourth quartile of greenspace proportion in 300 m buffer was associated with 17% decreased risk of incident glaucoma (aHR = 0.810, 95%CI 0.708, 0.927, p = 0.002), and a significant trend exists for increase in greenspace and decreased glaucoma risk (p for trend = 0.003). A similar association was found for greenspace in 1000 m buffer. Restricted cubic spline showed that there was no evidence for nonlinear association (P for nonlinear: 300 m buffer 0.6929, 1,000 m buffer 0.6115). Compared with the first quartile, the fourth quartile of domestic garden was associated with 16% decreased risk (aHR = 0.843, 95%CI 0.742, 0.957, p = 0.008), and the negative trend also existed (p for trend = 0.002). Restricted cubic spline demonstrated that the relationship was nonlinear in the 300 m buffer (p for nonlinear = 0.0038), rising slightly before 17.48%, decreasing significantly before 45.15%, and increasing again slightly after. The association was linear in the 1000 m buffer (p for nonlinear = 0.2994) (Figure S3 in Supporting Information S1). Table 2The Association Between Green Space, Domestic Garden, and Incident Glaucoma Based on Cox Regression Model300 m buffer1000 m bufferModel 1Model 2Model 1Model 2HR (95%CI) P valueHR (95%CI) P valueHR (95%CI) P valueHR (95%CI) P valueGreen spacePer 10% increase0.964 (0.945–0.984) <0.001 0.966 (0.943–0.989) 0.004 0.963 (0.943–0.985) 0.001 0.966 (0.941–0.991) 0.008 Quartile 11[REF]–1[REF]–1[REF]–1[REF]–Quartile 20.983 (0.898–1.076)0.7070.975 (0.890–1.068)0.5840.972 (0.887–1.065)0.5400.967 (0.881–1.061)0.477Quartile 30.914 (0.827–1.011)0.0810.908 (0.817–1.009)0.0740.876 (0.792–0.969) 0.010 0.873 (0.783–0.974) 0.015 Quartile 40.820 (0.730–0.920) 0.001 0.823 (0.723–0.937) 0.003 0.803 (0.712–0.906) <0.001 0.810 (0.708–0.927) 0.002 P for trend 0.001 0.003 <0.001 0.001 Domestic gardenPer 10% increase0.945 (0.916–0.976) 0.001 0.948 (0.913–0.983) 0.004 0.946 (0.908–0.986) 0.009 0.951 (0.909–0.996) 0.033 Quartile 11[REF]–1[REF]–1[REF]–1[REF]–Quartile 21.016 (0.924–1.116)0.7441.016 (0.924–1.117)0.7470.935 (0.844–1.036)0.1970.936 (0.845–1.038)0.209Quartile 30.837 (0.749–0.934) 0.002 0.841 (0.750–0.942) 0.003 0.879 (0.782–0.988) 0.031 0.887 (0.788–0.999) 0.048 Quartile 40.836 (0.745–0.939) 0.003 0.843 (0.742–0.957) 0.008 0.817 (0.722–0.923) 0.001 0.827 (0.727–0.941) 0.004 P for trend <0.001 0.002 0.001 0.004 Note. HR, hazards ratio; CI, confidence interval; IQR, interquartile range. P for trend was the association between greenness and tinnitus across interquartiles of greenness distribution percentage. All models included mutual adjustment for greenspace or domestic garden percentage at corresponding buffer. Model 1, adjusted for age at recruitment, sex, and ethnicity. Model 2, Model 1 + duration of year living in the current address, education, townsend index, time spent outdoor, smoker status, alcohol consumption, physical activity, healthy diet, body mass index, prevalent diabetes status, and systolic blood pressure. Values in bold indicate P < 0.05. The Association Between Green Space, Domestic Garden, and Incident Glaucoma Based on Cox Regression Model Note. HR, hazards ratio; CI, confidence interval; IQR, interquartile range. P for trend was the association between greenness and tinnitus across interquartiles of greenness distribution percentage. All models included mutual adjustment for greenspace or domestic garden percentage at corresponding buffer. Model 1, adjusted for age at recruitment, sex, and ethnicity. Model 2, Model 1 + duration of year living in the current address, education, townsend index, time spent outdoor, smoker status, alcohol consumption, physical activity, healthy diet, body mass index, prevalent diabetes status, and systolic blood pressure. Values in bold indicate P < 0.05. We performed several sensitivity analyses. Analysis including all participants with greenness record, with missing covariates imputed based on MICE algorithm showed similar association as our main analysis (Table S3 in Supporting Information S1). Including available medical record from general practice for glaucoma case identification did not change the association (Table S4 in Supporting Information S1). Excluding incident cases within 3 years since baseline also exhibited consistent trend and significance (Table S5 in Supporting Information S1). Analysis for overall greenspace and overall greenspace stratified by greenspace subtype showed there was cumulative effect of overall greenspace, and there was no significant multiplicative or additive interaction effect of two subtypes (Tables S6 and S7 in Supporting Information S1). Cross‐Sectional Association Between Greenspace, Domestic Garden, and Cornea‐Compensated IOP Only greenspace within 300 m buffer was found to be consistently associated with decreased IOP in both basically and fully adjusted models [Table 3]. Both per 10% increase in greenspace and the fourth quartile (beta −0.168, 95%CI −0.295, −0.040, p = 0.010) was associated with decreased IOPcc. Exclusion of prevalent glaucoma cases from the analysis demonstrated similar but stronger association (Table S8 in Supporting Information S1). Restricted cubic spline shows that there was no nonlinearity for greenspace in 300 or 1,000 m buffer (p for nonlinear: 300 m 0.8522, 1,000 m 0.4876). However, there was significant nonlinearity for domestic garden in 1000 m, with the spline decline stably before 31.9% and substantially after (Figure S4 in Supporting Information S1). Table 3The Association Between Green Space, Domestic Garden, and IOPcc Based on Linear Regression Model300 m buffer1000 m bufferModel 1Model 2Model 1Model 2beta(95%CI) P valuebeta(95%CI) P valuebeta(95%CI) P valuebeta(95%CI) P valueGreen spacePer 10% increase−0.027 (−0.048, −0.006) 0.013 −0.041 (−0.066, −0.016) 0.001 −0.034 (−0.057, −0.011) 0.003 −0.054 (−0.08, −0.027) <0.001 Quartile 11[REF]–1[REF]–1[REF]–1[REF]–Quartile 2−0.035 (−0.111, 0.041)0.364−0.043 (−0.118, 0.033)0.271−0.062 (−0.137, 0.013)0.107−0.077 (−0.152, −0.001) 0.048 Quartile 3−0.078 (−0.168, 0.012)0.088−0.107 (−0.200, −0.014) 0.025 −0.048 (−0.138, 0.043)0.300−0.096 (−0.193, 0.001)0.052Quartile 4−0.122 (−0.239, −0.004) 0.042 −0.168 (−0.295, −0.040) 0.010 −0.085 (−0.216, 0.046)0.204−0.130 (−0.270, 0.010)0.070P for trend 0.031 0.010 0.132 0.026 Domestic gardenPer 10% increase−0.022 (−0.052, 0.009)0.169−0.043 (−0.078, −0.007) 0.018 −0.028 (−0.067, 0.011)0.159−0.057 (−0.100, −0.014) 0.009 Quartile 11[REF]–1[REF]–1[REF]–1[REF]–Quartile 2−0.028 (−0.133, 0.076)0.596−0.035 (−0.138, 0.068)0.5070.057 (−0.063, 0.176)0.3520.051 (−0.067, 0.168)0.398Quartile 30.059 (−0.057, 0.176)0.3160.026 (−0.09, 0.143)0.6560.073 (−0.059, 0.205)0.2790.048 (−0.083, 0.179)0.473Quartile 4−0.057 (−0.177, 0.063)0.351−0.102 (−0.228, 0.025)0.1150.062 (−0.071, 0.196)0.3610.019 (−0.117, 0.155)0.782P for trend0.4960.2020.5420.795 Note. CI, confidence interval; IQR, interquartile range. P for trend was the association between greenness and tinnitus across interquartiles of greenness distribution percentage. Model 1, adjusted for age at recruitment, sex, and ethnicity. Model 2, Model 1 + duration of year living in the current address, education, townsend index, time spent outdoor, smoker status, alcohol consumption, physical activity, healthy diet, body mass index, prevalent diabetes status, systolic blood pressure, and avMSE. Values in bold indicate P < 0.05. The Association Between Green Space, Domestic Garden, and IOPcc Based on Linear Regression Model Note. CI, confidence interval; IQR, interquartile range. P for trend was the association between greenness and tinnitus across interquartiles of greenness distribution percentage. Model 1, adjusted for age at recruitment, sex, and ethnicity. Model 2, Model 1 + duration of year living in the current address, education, townsend index, time spent outdoor, smoker status, alcohol consumption, physical activity, healthy diet, body mass index, prevalent diabetes status, systolic blood pressure, and avMSE. Values in bold indicate P < 0.05. Additional Mediation and Interaction Analysis We performed exploratory analysis to investigate the potential role of leisure screen time, physical activity, and air pollution, including NO2, NOx, PM10, PM2.5, and PM 2.5 absorbance. The total excess relative risk was decomposed into four proportions for each factor: (a) to neither mediation nor interaction (proportion of controlled direct effect, CDE %); (b) to just interaction (proportion of reference interaction, IntRef %); (c) to both interaction and mediation (proportion of mediated interaction, IntMed %); and (d) to just mediation (proportion of pure indirect effect, PIE %). After full adjustment, there were significant total effect risk ratio and direct effect of both green space and domestic garden for incident glaucoma in all models, in both 300 and 1,000 m buffer (Table 4; Table S9 in Supporting Information S1). However, there was no evidence for significant interaction nor mediation role of reduced leisure screen time, increased physical activity, or reduced air pollution in the pathway. Table 4Proportion Mediated and/or Interacted of Potential Mediator/Moderator Factors in the Pathway of Greenness and Incident Glaucoma Within 300 m BufferGreenspaceBeta (95%CI) P valueDomestic gardenBeta (95%CI) P valueLeisure screen time (<3.5 hr/d)TERiRa82.6 (72.41, 92.8) <0.001 TERiRa82.76 (72.85, 92.66) <0.001 CDE%139.96 (90.21, 189.72) <0.001 CDE%64.07 (15.53, 112.61) 0.010 IntRef%−39.88 (−89.57, 9.8)0.116IntRef%35.93 (−11.78, 83.64)0.140IntMed%−0.1 (−0.76, 0.56)0.775IntMed%1.84 (−0.67, 4.35)0.151PIE%0.01 (−0.1, 0.13)0.817PIE%−1.84 (−3.79, 0.11)0.064Physical activity (>1500 MET‐min/week)TERiRa83.05 (72.81, 93.3) <0.001 TERiRa82.46 (72.6, 92.31) <0.001 CDE%75.81 (35.63, 116) <0.001 CDE%109.78 (73.05, 146.51) <0.001 IntRef%24.1 (−15.36, 63.55)0.231IntRef%−9.07 (−44.88, 26.74)0.620IntMed%1.19 (−0.81, 3.19)0.244IntMed%−0.51 (−2.52, 1.51)0.621PIE%−1.1 (−2.59, 0.39)0.148PIE%−0.2 (−1.57, 1.17)0.777NO2 (<40 μg/m3)TERiRa82.62 (72.19, 93.06) <0.001 TERiRa81.96 (71.45, 92.46) <0.001 CDE%105.71 (86.54, 124.88) <0.001 CDE%102.62 (93.69, 111.54) <0.001 IntRef%−20.12 (−74.14, 33.9)0.465IntRef%−3.7 (−16.3, 8.89)0.564IntMed%19.79 (−33.35, 72.93)0.465IntMed%3.2 (−7.69, 14.1)0.564PIE%−5.38 (−24.03, 13.27)0.572PIE%−2.12 (−9.62, 5.39)0.581NOx (<42.13 μg/m3)TERiRa82.98 (72.65, 93.32) <0.001 TERiRa82.83 (72.83, 92.82) <0.001 CDE%97.41 (29.21, 165.6) 0.005 CDE%97.74 (44.62, 150.86) <0.001 IntRef%−15.29 (−111.36, 80.77)0.755IntRef%−2.58 (−64.43, 59.27)0.935IntMed%13.14 (−69.39, 95.67)0.755IntMed%1.13 (−25.89, 28.14)0.935PIE%4.75 (−51.87, 61.36)0.869PIE%3.71 (−16.82, 24.25)0.723PM10 (<20 μg/m3)TERiRa82.84 (72.52, 93.16) <0.001 TERiRa83.33 (73.26, 93.4) <0.001 CDE%77.65 (33.71, 121.58) 0.001 CDE%74.35 (35.27, 113.44) <0.001 IntRef%−53.87 (−136.2, 28.47)0.200IntRef%6.49 (−36.99, 49.98)0.770IntMed%47.32 (−25.01, 119.65)0.200IntMed%−3.78 (−29.1, 21.54)0.770PIE%28.9 (−10.59, 68.39)0.152PIE%22.93 (−1.12, 46.98)0.062PM2.5 (<10 μg/m3)TERiRa82.7 (72.39, 93.02) <0.001 TERiRa82.95 (72.94, 92.96) <0.001 CDE%122.49 (56.42, 188.57) <0.001 CDE%78.91 (26.3, 131.53) 0.003 IntRef%−53.25 (−156.19, 49.68)0.311IntRef%22.93 (−39.86, 85.71)0.474IntMed%48.36 (−45.11, 141.83)0.311IntMed%−13.16 (−49.19, 22.88)0.474PIE%−17.6 (−75.79, 40.59)0.553PIE%11.32 (−16.53, 39.17)0.426PM2.5 absorbance (<1.13 μg/m3)TERiRa82.71 (72.32, 93.09) <0.001 TERiRa82.48 (72.44, 92.52) <0.001 CDE%97.41 (37.68, 157.14) 0.001 CDE%112.03 (61.83, 162.24) <0.001 IntRef%−20.32 (−101.56, 60.92)0.624IntRef%−22.18 (−81.53, 37.18)0.464IntMed%15.13 (−45.36, 75.61)0.624IntMed%8.62 (−14.45, 31.69)0.464PIE %7.78 (−34.03, 49.59)0.715PIE%1.52 (−14.77, 17.82)0.855 Note. TERiRa, total effect risk ratio; CDE%, proportion of controlled direct effect; IntRef%, proportion of reference interaction; IntMed%, proportion of mediated interaction; PIE%, proportion of pure indirect effect. THe TERiRa was estimated at the fourth quartile, referring to the first quartile as reference. Mediator level for the decomposition was set at optimal level in each model. Models were adjusted for age at recruitment, sex, ethnicity, duration of year living in the current address, education, townsend index, time spent outdoor, smoker status, alcohol consumption, physical activity, healthy diet, body mass index, prevalent diabetes status, systolic blood pressure, and mutual adjustment for greenspace/domestic gardens. Values in bold indicate P < 0.05. Proportion Mediated and/or Interacted of Potential Mediator/Moderator Factors in the Pathway of Greenness and Incident Glaucoma Within 300 m Buffer Note. TERiRa, total effect risk ratio; CDE%, proportion of controlled direct effect; IntRef%, proportion of reference interaction; IntMed%, proportion of mediated interaction; PIE%, proportion of pure indirect effect. THe TERiRa was estimated at the fourth quartile, referring to the first quartile as reference. Mediator level for the decomposition was set at optimal level in each model. Models were adjusted for age at recruitment, sex, ethnicity, duration of year living in the current address, education, townsend index, time spent outdoor, smoker status, alcohol consumption, physical activity, healthy diet, body mass index, prevalent diabetes status, systolic blood pressure, and mutual adjustment for greenspace/domestic gardens. Values in bold indicate P < 0.05. Subgroup analysis stratified by PRS or MYOC genotype showed there was no significant interaction effects between residential greenspace or domestic gardens and genetic predisposition (Figure 1). Adjusted‐association association between residential greenspace and domestic gardens and incident glaucoma. HR, hazard ratios. HRs were estimated at the fourth quartile, with the first quartile as reference. Discussion In this prospective cohort, we found an independent negative association between residential greenness and incident glaucoma. Both public greenspace and domestic gardens showed a negative association with incident glaucoma, with public greenspace showing a linear dose‐response relationship. The beneficial effect of greenspace may be partially explained by a decrease in IOP, especially if greenspace was easily accessible within a 300 m buffer. Notably, we found that increased physical activities, decreased recreational screen time, or reduced air pollution did not interact with greenspace or mediate the relationship. Furthermore, genetic susceptibility did not interact with greenspace on the risk of incident glaucoma. To our knowledge, this is the first study to demonstrate the potential protective role of greenspace and domestic garden in glaucoma. This may be supported by studies on the beneficial association between greenness and visual acuity in children. A higher normalized Difference Vegetation Index (NDVI) was associated with decreased myopia prevalence and progression (Bao et al., 2024; Yang, Li, et al., 2021; Zhang et al., 2023), and the proportion, aggregation and connectivity of greenspace were morphological factors showing greater impacts (Yang et al., 2024). Green views were associated with mitigating eye fatigue (Labib et al., 2020), and the tendency to shift from near work to distant greenery due to higher proportion of greenspace may further alleviate myopic progression (Yang et al., 2024). Additionally, an increase in NDVI was also associated with a decrease in diabetic retinopathy prevalence in China. However, there is scarce mechanistic basis for such association. Only one study demonstrated the potential retinal protective role of green light exposure, showing that green light emitting diodes exposure was associated with decreased apoptotic cells detected in outer nuclear layers in goldfish (Song et al., 2016). Future laboratory studies are warranted to explore the underlying physiobiological mechanisms for the association between greenspace and retinal diseases. Restricted cubic splines showed that there was a non‐linear relationship between domestic garden in the 300 m buffer and incident glaucoma. To be specific, the hazard ratio slightly increased after domestic garden coverage percentage reaching 45.15%, yet the fourth quartile (>41.85%) still showed significant protective associations against incident glaucoma (HR = 0.843, 95% CI 0.742–0.957) compared to the first quartile. On the other hand, greenspace coverage showed a linear beneficial effect. This suggested a saturation phenomenon for domestic garden coverage after reaching a high level, a finding similar to effects of greenness on other diseases (Jia et al., 2023; Nie et al., 2020). The difference might be explained by different morphologies of these two greenness types. Trees were more likely to be present in public greenspace areas, and only tree canopy but not open grass was associated with reduced incident dementia and Parkinson's disease risk (Astell‐Burt et al., 2020; Klompmaker et al., 2024). This was echoed by an eye movement analysis, which demonstrated that partially open greenspace like under‐forest space had the highest positive restorative effects among greenspace types (Liu et al., 2022). Therefore, the type and structure of vegetation found in domestic gardens may reach saturation point sooner than the potentially more diverse and tree‐rich environments of public greenspace. The specific morphological mechanisms still warrant future studies. While the benefit of domestic gardens appears to saturate, our analysis of overall greenspace and stratification analysis incorporating both greenspace subtypes demonstrated the potential benefits of maximizing overall green exposure. We did not find evidence that increased physical activities or decreased recreational screen time or air pollution mediated the association. While other studies showed that these factors played an important role in the effect of greenness on other diseases in the elderly and even visual impairment in children (Bao et al., 2024; Liu et al., 2023; Yang, Li, et al., 2021; Yang et al., 2023), this was not the case for glaucoma. Therefore, there may be other underlying mechanisms. However, as physical activity, and recreational screen time were self‐reported in the current study, recall bias may obscure the true effect. The true effect may also be diluted due to misclassification of individual‐level greenness exposure due to unmeasured aspects, for example, exact location and greenery environmental morphologies during these activities. There are some other possible indirect mechanisms for the association between greenness and glaucoma. First, as mental health plays an important role in glaucoma (Han et al., 2023; Zhang et al., 2024), greenspace exposure may reduce glaucoma risk through promoting mental well‐being. A systematic review of randomized controlled trials showed that greenspace was associated with decreased fatigue, reduced stress, and increased vitality (S. Song et al., 2022). Furthermore, greenspace exposure could significantly alleviate anxiety symptoms and reduce risk of incident anxiety (Lan et al., 2022; Liu et al., 2024) through impacts on emotional regulatory brain regions (Xu et al., 2023). Second, engagement with greenspace was associated with immune system regulation. Biodiversity of the natural environment provided evolutionarily determined diverse microbial inputs to the immune system, and was associated with reduced inflammation and less cytokine response to stress (Rook, 2013). Exposure to green spaces has been associated with enrichment in IL‐3, IL‐6 and macrophage‐stimulating protein signaling pathways in plasma proteomic signatures in the Coronary Artery Risk Development in Young Adults (CARDIA) study (As et al., 2024), and a reduction in circulating IL‐6 was also identified with increased greenness in an elderly cohort with a mean age of 71.4 years (Scheer et al., 2024). As abnormal microbiota and inflammatory imbalance were deeply associated with glaucoma pathophysiological pathways (Baudouin et al., 2021; Huang et al., 2023), greenspace may benefit glaucoma through immune‐regulation. Third, greenspace exposure was also associated with neurological function and degeneration, and glaucoma is a neurodegenerative disease (Mancino et al., 2018; Saccà et al., 2020). Greenspace exposure, represented by tree canopy or NDVI, reduced risks of neurodegenerative diseases, ranging from dementia and dementia mortality, to Parkinson's disease (Astell‐Burt et al., 2020; Hu et al., 2023; Klompmaker et al., 2024; Paul et al., 2020; Rodriguez‐Loureiro et al., 2022). This is echoed by a prospective study using structural brain imaging in healthy individuals, which showed that greenness was associated with better cortical thickness in AD‐associated brain regions (Crous‐Bou et al., 2020). In Barcelona school children, a greener neighborhood was also found to be associated with positive brain structure development for better cognitive function and attentiveness (Dadvand et al., 2018). Therefore, greenspace may reduce glaucoma risk through its potential neuroprotective function (Lawlor et al., 2018; Tribble et al., 2023; Wareham et al., 2022). Future studies are warranted to distangle direct and indirect mechanisms underlying the association. Strengths of our study include prospective design of a long follow‐up period and a large cohort, objectively computed near and wider range of residential greenspace coverage, and investigation of potential mediating mechanisms. However, our study still has several limitations. First, the population of our study was mostly white; therefore, generalization to other ethnicities should be made with caution. Second, misclassification of individual‐level greenness exposure may occur since we only considered residential greenspace and domestic gardens, and dynamic exposure in other places, such as the workplace, was not available in the current data set. Third, due to data set resource availability, other greenspace indicators, including NDVI and in‐depth greenspace morphology, were not analyzed in the current study, which are still warranted to help inform optimal greenspace formation for urban planning (Labib et al., 2020). Fourthly, potential residual confounding might remain despite our adjustment for various potential covariates. Conclusions This study demonstrated that public greenspace and domestic gardens within 300 and 1,000 m buffers of home location were associated with lower risk of incident glaucoma, with public greenspace showing a linear dose‐response association. These findings emphasize the importance of integrating greenspace near neighborhoods into urban planning for ocular health promotion and highlight the need to encourage public engagement with greenspaces. Future in‐depth observational and interventional studies are needed to investigate the most cost‐effective qualities and morphologies of greenspaces for ocular health. Laboratory studies will also be crucial for understanding underlying psychological, physiological, and biological mechanisms. Conflict of Interest The authors declare no conflicts of interest relevant to this study. Supporting information Supporting Information S1