Elevated serum high-density lipoprotein is positively associated with incident glaucoma among middle-aged and older adults: a national cohort in China Authors: Zilin Chen, Hanming Lin, Xianqi Zheng, Xi Song, Haitong Feng, Hui Guo, Xiaoyi Yu Published: 2026 ABSTRACT PurposeEmerging evidence challenges the traditional view of high-density lipoprotein (HDL) as uniformly protective in ocular health. This study aimed to evaluate the association between serum HDL levels and glaucoma in a middle-aged and older Chinese population.MethodsWe conducted a retrospective cohort study using data from the China Health and Retirement Longitudinal Study (CHARLS). A total of 7,510 participants aged over 45 years without baseline glaucoma were included. Serum HDL was measured at baseline (2011), and incident glaucoma was ascertained via self-report during follow-up (2013–2018). Cox proportional hazards models and restricted cubic spline (RCS) analyses were used to assess the association, with adjustment for demographic and clinical covariates.ResultsOver a 7 years’ follow-up, 224 participants (2.98%) developed glaucoma. After multivariable adjustment, each 1 mmol/L increase in HDL was associated with a 57% higher risk of glaucoma (HR = 1.57; 95% CI: 1.10–2.26; p = 0.014). Participants with low HDL level (< 1.0 mmol/L) had a significantly lower glaucoma risk compared to those with normal HDL level (HR = 0.63; 95% CI: 0.38–0.89; p = 0.012). RCS analysis indicated a steeper increase in risk at serum HDL ≤ 1.501 mmol/L. Subgroup analyses revealed consistency across most strata, with a stronger association observed in participants with liver disease (p for interaction = 0.035).ConclusionElevated serum HDL levels were associated with an increased risk of glaucoma in Chinese middle-aged and older adults, particularly at lower HDL concentrations. These findings challenge the conventional “good cholesterol” paradigm in glaucoma and underscore the need for further research into lipid-related pathways in its pathogenesis.Graphical Abstract Introduction Glaucoma is a heterogeneous group of progressive optic neuropathies characterized by irreversible loss of retinal ganglion cells, leading to visual field defects and permanent vision loss if left untreated [1, 2]. It is the second leading cause of blindness among individuals aged over 50 years worldwide, responsible for approximately 3.6 million cases of blindness in 2020 according to a Global Burden of Disease Study [3], Established risk factors for glaucoma include elevated intraocular pressure (IOP), advanced age, African or Asian ancestry, family history, and high myopia [1, 4], Among these, IOP remains the only modifiable risk factor which is consistent with large-scale randomized controlled trials [5, 6], Identifying additional modifiable risk factors is critical for developing novel preventive strategies and for reducing the global burden of glaucoma. HDL has traditionally been regarded as “good cholesterol” in contrast to low-density lipoprotein cholesterol (LDL), total cholesterol (TC), and triglycerides (TG), which are often labeled as “bad cholesterol” [7–9]. However, this binary classification is increasingly being recognized as overly simplistic, particularly in the context of ophthalmic diseases. Emerging evidence challenges this traditional view: Elevated HDL levels have been associated with retinal nerve fiber layer thinning [10], increased risk of age-related macular degeneration [11] and glaucoma in European populations [12]. In contrast, cross-sectional studies in Asian populations have reported no significant association between HDL levels and glaucoma [13, 14], highlighting potential ethnic heterogeneity. Given the inconsistent findings and differences in lipid metabolism across populations, we leveraged data from CHARLS and to investigate the association between HDL and glaucoma in Chinese population. Methods Study design and population We conducted a retrospective cohort study using data from CHARLS, a nationally representative survey of community-dwelling adults aged over 45 years in mainland China [15]. The baseline survey was conducted from June 2011 to March 2012, enrolling 17,596 participants through face-to-face computer-assisted personal interviews. A multistage, probability-proportional-to-size sampling strategy was used to select 450 primary sampling units (villages or neighborhoods) across 150 counties in 28 provinces. All the participants provided written informed consent. The study protocol was approved by the Biomedical Ethics Review Committee of the Peking University (IRB00001052-11015). We used 2011 as the baseline and follow-up data from 2013, 2015, and 2017 to ascertain the incidence of glaucoma. Participants were excluded if they (1) had a self-reported diagnosis of glaucoma at baseline; (2) lacked follow-up glaucoma status; (3) had missing serum lipid data; (4) were aged under 45 years, had a history of cancer, cataract surgery, or use of lipid-lowering medications; or (5) had missing data on key covariates including sex, BMI, hypertension, diabetes, liver disease, smoking, alcohol use, or education. In total, 7,510 participants were included in the final cohort (Fig. 1). Fig. 1Flowchart outlining eligible participants for this study in the CHARLS Flowchart outlining eligible participants for this study in the CHARLS Serum lipid assessment Fasting serum lipids (including HDL, LDL, TG and TC) were measured at baseline (2011) by using an enzymatic colorimetric assay at the Youanmen Center for Clinical Laboratory, Capital Medical University. Serum HDL was evaluated as the primary exposure in three ways: (1) as a continuous variable (per 1 mmol/L increase); (2) using a clinical cutoff of < 1.0 mmol/L versus ≥ 1.0 mmol/L, based on the 2024 Chinese Expert Consensus on Community-Based Adult Dyslipidemia Management [16]; and (3) by quartiles (Q1: ≤1.02 mmol/L; Q2: 1.02–1.27 mmol/L; Q3: 1.27–1.54 mmol/L; Q4: ≥1.54 mmol/L). Additionally, to explore potential non-linear relationships, we fitted restricted cubic spline (RCS) models and performed two-piece wise Cox regression to estimate HRs in different HDL ranges. Ascertainment of incident glaucoma The incident of glaucoma was identified through self-reporting. In each wave, participants were asked “Have you ever been diagnosed with glaucoma by a doctor or other healthcare professional?” Those who answered “yes” were considered to have glaucoma. Participants with a self-reported diagnoses at baseline were excluded. All remaining participants were followed up at least once between 2013 and 2018. Covariates Based on previous literature and clinical relevance [17, 18], we adjusted for the following potential confounders: age (years), sex, body mass index (BMI, kg/m²), education level (categorized as junior high school or above vs. below junior high), marital status (married or cohabiting vs. not), hypertension (defined as self-reported diagnosis, systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, or use of anti-hypertensive medication), diabetes (defined as self-reported diagnosis, fasting plasma glucose ≥ 7.0 mmol/L, HbA1c ≥ 6.5%, or use of glucose-lowering medications or insulin), liver disease (self-reported), smoking status (ever smoked vs. never) and alcohol consumption (ever drank vs. never). Statistical analysis Baseline characteristics were compared between participants with low HDL level (< 1.0 mmol/L) and those with normal HDL level (≥ 1.0 mmol/L). Continuous variables with normal distribution were expressed as mean ± standard deviation (SD) and compared using Student’s t-test. Categorical variables are presented as percentages and analyzed using the chi-square test. Time-to-event analyses were conducted using Cox proportional hazards models with follow-up time (in years) from baseline as the time-scale to estimate HR and 95% CI for the association between baseline HDL levels and glaucoma. Participants were censored at the earliest of: glaucoma diagnosis, death, loss to follow-up, or the end of the study period (2018). Three models were constructed: (1) a crude model without adjustment; (2) Model I: adjusted for age, sex, BMI, education level, marital status, hypertension, diabetes, liver disease, smoking status, and alcohol consumption; and (3) Model II: additionally adjusted for LDL and TG. To explore potential non-linear relationships, we fitted restricted cubic spline (RCS) models and performed two-piecewise Cox regression using the rcssci R package [19], which integrates the segmented package. The breakpoint was determined based on the likelihood-ratio test combined with bootstrap resampling methods to identify the data-driven point that best separates two distinct linear segments. We emphasize that this threshold analysis is an exploratory heuristic procedure providing heuristic guidance for cutoff selection rather than precise cutoff determination. Subgroup analyses were conducted to assess the effect modification by age (< 60 vs. ≥60 years), sex, hypertension, diabetes, liver disease, smoking status, and alcohol consumption. Interaction terms were tested using likelihood ratio tests to compare models with and without interaction terms. All statistical analyses were performed using the R software (version 4.5.1). Statistical significance was defined as a two-side p-value of < 0.05. Results Baseline characteristics After applying the inclusion and exclusion criteria, 7,510 participants were included in the final cohort after applying inclusion and exclusion criteria (Fig. 1). The mean age was 59.2 ± 9.2 years, and 3,526 (47.0%) were male. Mean serum HDL was 1.3 ± 0.4 mmol/L; 1,519 participants (20.2%) had low HDL level (< 1.0 mmol/L) at baseline. Table 1Baseline characteristics of participants included in the studyVariableLow HDL(< 1.0 mmol/L)Normal HDL(≥ 1.0 mmol/L) P N15195991Age, mean (SD)58.9 ± 9.359.3 ± 9.20.11Gender, n (%)< 0.001Male806 (53.1%)2,720 (45.4%)Female713 (46.9%)3,271 (54.6%)BMI, mean (SD)24.9 ± 3.823.0 ± 3.7< 0.001Education, n (%)< 0.001Junior and above517 (34.0%)1,722 (28,7%)Under junior1,002 (66.0%)4,269 (71.3%)Married, n (%)1,360 (89.5%)5,294 (88.4%)0.218Hypertension, n (%)442 (29.1%)1,263 (21.1%)< 0.001Diabetes, n (%)368 (24.2%)795 (13.3%)< 0.001Liver Disease, n (%)56 (3.7%)230 (3.8%)0.84Smoking History, n (%)666 (43.8%)2,295 (38.3%)< 0.001Drinking History, n (%)481 (31.7%)2,023 (33.8%)0.128HDL, mean (SD)0.9 ± 0.11.5 ± 0.3< 0.001LDL, mean (SD)2.7 ± 0.93.1 ± 0.9< 0.01TC, mean (SD)4.8 ± 1.05.1 ± 1.0< 0.01TG, mean (SD)2.4 ± 1.61.2 ± 0.7< 0.01Note: Data are presented as n, n(%) or the mean ± standard deviation. N, number of individuals; BMI, body mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; TG, triglyceride; TC, total cholesterol Table 1Baseline characteristics of participants included in the studyVariableLow HDL(< 1.0 mmol/L)Normal HDL(≥ 1.0 mmol/L) P N15195991Age, mean (SD)58.9 ± 9.359.3 ± 9.20.11Gender, n (%)< 0.001Male806 (53.1%)2,720 (45.4%)Female713 (46.9%)3,271 (54.6%)BMI, mean (SD)24.9 ± 3.823.0 ± 3.7< 0.001Education, n (%)< 0.001Junior and above517 (34.0%)1,722 (28,7%)Under junior1,002 (66.0%)4,269 (71.3%)Married, n (%)1,360 (89.5%)5,294 (88.4%)0.218Hypertension, n (%)442 (29.1%)1,263 (21.1%)< 0.001Diabetes, n (%)368 (24.2%)795 (13.3%)< 0.001Liver Disease, n (%)56 (3.7%)230 (3.8%)0.84Smoking History, n (%)666 (43.8%)2,295 (38.3%)< 0.001Drinking History, n (%)481 (31.7%)2,023 (33.8%)0.128HDL, mean (SD)0.9 ± 0.11.5 ± 0.3< 0.001LDL, mean (SD)2.7 ± 0.93.1 ± 0.9< 0.01TC, mean (SD)4.8 ± 1.05.1 ± 1.0< 0.01TG, mean (SD)2.4 ± 1.61.2 ± 0.7< 0.01Note: Data are presented as n, n(%) or the mean ± standard deviation. N, number of individuals; BMI, body mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; TG, triglyceride; TC, total cholesterol Baseline characteristics of participants included in the study Note: Data are presented as n, n(%) or the mean ± standard deviation. N, number of individuals; BMI, body mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; TG, triglyceride; TC, total cholesterol Table 1 presents the baseline characteristics stratified by HDL status. Participants with low HDL level were more likely to be male, have a higher BMI, have a higher education level, and be current smokers. They also had a higher prevalence of hypertension and diabetes than compared to those with normal HDL levels. HDL and incident glaucoma Figure 2 illustrates the incidence of glaucoma over 7 years’ follow-up according to the baseline HDL status and quartiles. The incidence of glaucoma in participants with low HDL level (< 1.0 mmol/L) was 2.1% (32/1,519) compared to 3.2% (192/5,991) in those with normal HDL level. A significant upward trend in glaucoma risk was observed across increasing HDL quartiles (p for trend = 0.026). The incidence rates were 2.1% in Q1, 2.9% in Q2, 3.6% in Q3, and 3.2% in Q4. Fig. 2Prevalence of incident glaucoma according to baseline HDL levels and HDL quartiles. Note: All of the study participants were divided into two groups according to whether it is in dyslipidemia: Low HDL(< 1.0mmol/L), Normal HDL(≥ 1.0mmol/L); (B) All of the study participants were divided into four groups four groups according to quartiles of HDL(quartile 1 [Q1]: ≤ 1.02mmol/L; quartile 2 [Q2]:1.02–1.27; quartile 3 [Q3]: 1,27–1.54; quartile 4 [Q4]: ≥ 1.54) Prevalence of incident glaucoma according to baseline HDL levels and HDL quartiles. Note: All of the study participants were divided into two groups according to whether it is in dyslipidemia: Low HDL(< 1.0mmol/L), Normal HDL(≥ 1.0mmol/L); (B) All of the study participants were divided into four groups four groups according to quartiles of HDL(quartile 1 [Q1]: ≤ 1.02mmol/L; quartile 2 [Q2]:1.02–1.27; quartile 3 [Q3]: 1,27–1.54; quartile 4 [Q4]: ≥ 1.54) Table 2 presents both crude and adjusted Cox proportional hazards models examining the association between baseline HDL levels and glaucoma over 7 years of follow-up. In the crude model, each 1 mmol/L increase in HDL was associated with a higher risk of glaucoma (HR = 1.40, 95% CI: 1.03–1.91, p = 0.034). Participants with low HDL level (< 1.0 mmol/L) had a 34% lower risk of glaucoma compared to those with normal level (HR = 0.66, 95% CI: 0.45–0.95, p = 0.027). Table 2Association of baseline HDL with glaucoma according to three Cox regression modelsHDL levelsNon-adjusted ModelModel IModel IIHR(95% CI) P HR(95% CI) P HR(95% CI) P HDL per 1mmol/L increase1.4(1.03–1.91)0.0341.49(1.07–2.08)0.0191.57(1.10–2.26)0.014Low HDL level(< 1.0mmol/L)0.66(0.45–0.95)0.0270.63(0.43–0.92)0.0180.58(0.38–0.89)0.012HDL in quartilesQ1(≤ 1.02mmol/L)ReferenceQ2(1.02-1.27mmol/L)1.37(0.9–2.09)0.1391.42(0.93–2.17)0.1051.54(0.98–2.40)0.06Q3(1.27-1.54mmol/L)1.73(1.15–2.61)0.0081.8(1.18–2.7)0.0071.97(1.25–3.11)0.003Q4 (≥ 1.54mmol/L)1.56(1.03–2.35)0.0361.64(1.07–2.53)0.0251.84(1.14–2.98)0.012Note: Model I: adjusted for age, sex, BMI, education, marital status, hypertension, diabetes, liver disease, smoking history, drinking history; Model II: additionally adjusted for LDL and TG; HR, hazard ratio; CI, confidence interval Table 2Association of baseline HDL with glaucoma according to three Cox regression modelsHDL levelsNon-adjusted ModelModel IModel IIHR(95% CI) P HR(95% CI) P HR(95% CI) P HDL per 1mmol/L increase1.4(1.03–1.91)0.0341.49(1.07–2.08)0.0191.57(1.10–2.26)0.014Low HDL level(< 1.0mmol/L)0.66(0.45–0.95)0.0270.63(0.43–0.92)0.0180.58(0.38–0.89)0.012HDL in quartilesQ1(≤ 1.02mmol/L)ReferenceQ2(1.02-1.27mmol/L)1.37(0.9–2.09)0.1391.42(0.93–2.17)0.1051.54(0.98–2.40)0.06Q3(1.27-1.54mmol/L)1.73(1.15–2.61)0.0081.8(1.18–2.7)0.0071.97(1.25–3.11)0.003Q4 (≥ 1.54mmol/L)1.56(1.03–2.35)0.0361.64(1.07–2.53)0.0251.84(1.14–2.98)0.012Note: Model I: adjusted for age, sex, BMI, education, marital status, hypertension, diabetes, liver disease, smoking history, drinking history; Model II: additionally adjusted for LDL and TG; HR, hazard ratio; CI, confidence interval Association of baseline HDL with glaucoma according to three Cox regression models 1.4 (1.03–1.91) 1.49 (1.07–2.08) 1.57 (1.10–2.26) Low HDL level (< 1.0mmol/L) 0.66 (0.45–0.95) 0.63 (0.43–0.92) 0.58 (0.38–0.89) 1.37 (0.9–2.09) 1.42 (0.93–2.17) 1.54 (0.98–2.40) 1.73 (1.15–2.61) 1.97 (1.25–3.11) 1.56 (1.03–2.35) 1.64 (1.07–2.53) 1.84 (1.14–2.98) Note: Model I: adjusted for age, sex, BMI, education, marital status, hypertension, diabetes, liver disease, smoking history, drinking history; Model II: additionally adjusted for LDL and TG; HR, hazard ratio; CI, confidence interval After adjusting for age, sex, BMI, education, marital status, hypertension, diabetes, liver disease, smoking and alcohol consumption in Model I, each 1 mmol/L increase in HDL remained significantly associated with glaucoma (HR = 1.49, 95% CI: 1.07–2.08, p = 0.019). Low HDL level (< 1.0 mmol/L) remained associated with a 37% lower risk (HR = 0.63, 95% CI: 0.43–0.92, p = 0.018). In Model II, additionally adjusted for LDL and TG based on Model I, each 1 mmol/L increase in HDL remained significantly associated with glaucoma (HR = 1.57, 95% CI: 1.10–2.26, p = 0.014). Low HDL level (< 1.0 mmol/L) remained associated with a 42% lower risk (HR = 0.63, 95% CI: 0.38–0.89, p = 0.012). Compared with Q1, HRs were 1.54 (95% CI: 0.98–2.40, p = 0.06) in Q2, 1.97 (95% CI: 1.25–3.11, p = 0.003) in Q3, and 1.84 (95% CI: 1.14–2.98, p = 0.012) in Q4. We examined the shape of the association using multi-variable adjusted RCS models (Fig. 3). Although the test for non-linearity wasn’t significant (p = 0.183), the hazard ratio for glaucoma was significantly elevated at HDL levels ≤ 1.501 mmol/L (HR = 2.77, 95% CI: 1.17–6.55, p = 0.02), but not at levels > 1.501 mmol/L (HR = 1.35, 95% CI: 0.63–2.9, p = 0.44) (Table 3). Fig. 3The nonlinear Relationship between serum HDL and glaucoma within 7 years. Note: The solid and dashed lines represent the predicted values and 95% CIs, respectively. They were adjusted for age, sex, BMI, education, marital status, hypertension, diabetes, liver disease, smoking history, drinking history, LDL level and TG level The nonlinear Relationship between serum HDL and glaucoma within 7 years. Note: The solid and dashed lines represent the predicted values and 95% CIs, respectively. They were adjusted for age, sex, BMI, education, marital status, hypertension, diabetes, liver disease, smoking history, drinking history, LDL level and TG level Table 3Association between HDL and glaucoma using two-piece multivariate Cox regressionHDL levelHR95% CIP value≤ 1.501 mmol/L2.771.17–6.550.02> 1.501 mmol/L1.350.63–2.90.44Note: Adjusted for age, sex, BMI, education, marial status, hypertension, diabetes, liver disease, smoking history, drinking history, LDL level, and TG level. HRs represent the effect of a 1 mmol/L increase in HDL within each respective segment, with the breakpoint at 1.501 mmol/L Table 3Association between HDL and glaucoma using two-piece multivariate Cox regressionHDL levelHR95% CIP value≤ 1.501 mmol/L2.771.17–6.550.02> 1.501 mmol/L1.350.63–2.90.44Note: Adjusted for age, sex, BMI, education, marial status, hypertension, diabetes, liver disease, smoking history, drinking history, LDL level, and TG level. HRs represent the effect of a 1 mmol/L increase in HDL within each respective segment, with the breakpoint at 1.501 mmol/L Association between HDL and glaucoma using two-piece multivariate Cox regression Note: Adjusted for age, sex, BMI, education, marial status, hypertension, diabetes, liver disease, smoking history, drinking history, LDL level, and TG level. HRs represent the effect of a 1 mmol/L increase in HDL within each respective segment, with the breakpoint at 1.501 mmol/L Subgroup analyses by adjusted potential effect con-founders We conducted subgroup analyses to assess whether the association between HDL and glaucoma differed according to age, sex, educational background, hypertension, diabetes, liver disease, smoking status, alcohol consumption, TG level and LDL levels. Overall, the positive association between HDL (per 1 mmol/L increase) and glaucoma was consistent across most subgroups (Fig. 4). However, a statistically significant interaction was observed for liver disease (p for interaction = 0.035), with a stronger association observed noted in participants with liver disease. Fig. 4The relationship between baseline serum HDL and glaucoma in subgroup analysis. Note: They were adjusted for age, sex, BMI, education, marital status, hypertension, diabetes, liver disease, smoking history, drinking history, LDL level and TG level The relationship between baseline serum HDL and glaucoma in subgroup analysis. Note: They were adjusted for age, sex, BMI, education, marital status, hypertension, diabetes, liver disease, smoking history, drinking history, LDL level and TG level Discussion This large-scale, population-based retrospective cohort study using CHARLS data demonstrated that higher HDL levels were associated with an increased risk of glaucoma among middle-aged and older adults in China over a 7-year follow-up period. The RCS curve suggested a more pronounced increase in glaucoma risk at HDL levels ≤ 1.501 mmol/L. Subgroup analyses revealed consistency across most strata. Notably, a UK study observed this association only among participants aged ≥ 55 years, whereas we found it primarily among those aged < 60 years, highlighting potential age and ethnic differences. Three national cohort studies in UK and Republic of Korea reported positive associations between HDL and IOP [20, 21], and inverse associations with retinal nerve fiber layer thickness [10]. These findings suggest that elevated HDL levels may contribute to glaucoma pathogenesis by increasing IOP and reducing optic nerve resilience to IOP-related stress. Similarly, the association of HDL and dementia also attracted our attention [22]. We propose that shared neurodegenerative mechanisms may link elevated HDL to both conditions. While glaucoma and Alzheimer’s disease (AD) differ in their primary neuropathological hallmarks— progressive retinal ganglion cell (RGC) and axonal degeneration versus amyloid-β plaques and phosphorylated tau tangles—both are increasingly recognized as neurodegenerative disorders of the central nervous system that share critical molecular vulnerabilities [23]. The retina is an anatomical and functional extension of the central nervous system (CNS), and RGCs are CNS neurons with exceptionally high metabolic demands, rendering them susceptible to the same stressors that affect cerebral neurons. A key convergent pathway is lipid metabolism dysregulation mediated by apolipoprotein E (APOE) and ABCA1. APOE is the principal lipid carrier in the CNS, and the APOE ε4 allele is the strongest genetic risk factor for late-onset AD; it has also been implicated in primary open-angle glaucoma [24]. ABCA1, which catalyzes cholesterol efflux and HDL biogenesis, plays a critical role in both conditions: rare ABCA1 variants increase AD risk, and ABCA1 deficiency in mice causes progressive RGC loss through aberrant oxidative phosphorylation, mTORC1 activation, and autophagy suppression [25]. Thus, elevated HDL may not be universally protective; rather, dysfunctional HDL or impaired cholesterol efflux capacity may simultaneously compromise lipid homeostasis in the retina and brain. Emerging evidence suggests that systemic or local lipid dysregulation may contribute to ocular diseases, including corneal degeneration, cataract, glaucoma, and pathological myopia [26, 27]. A large-scale proteome-wide mendelian randomization study identified SEL1L, a plasma protein involved in lipid metabolism, as a potential therapeutic target for glaucoma [28]. Additionally, lipid-related genes such as ABCA1 and Caveolin-1 have been implicated in glaucoma susceptibility [29, 30]. These findings underscore the need for further investigation of lipid metabolic pathways as potential targets for glaucoma prevention and treatment. It is imperative to emphasize that our observational findings do not support any clinical strategy to lower HDL levels for glaucoma prevention, given the well-established cardioprotective roles of HDL [31, 32]. Notably, genetic studies of cholesteryl ester transfer protein deficiency provide compelling evidence that lifelong elevation of HDL is associated with a markedly increased risk of age-related macular degeneration (HR = 2.33), directly mediated by higher HDL levels rather than lower non-HDL cholesterol [33]. This genetic paradigm supports the concept that high HDL is not uniformly protective across tissues and that lipid metabolic pathways may have divergent cardiovascular and ocular effects. Taken together, rather than advocating HDL reduction, our findings call for a shift in clinical awareness: patients with elevated HDL, traditionally reassured regarding their cardiovascular risk, should be recognized as a subgroup that may benefit from enhanced glaucoma screening. Our study has several notable strengths. First, data from CHARLS employed a multistage, probability-proportional-to-size sampling strategy across 150 counties in 28 provinces, enhancing the generalizability of our findings to community-dwelling middle-aged and older adults throughout China. Second, the prospective cohort design with up to 7 years of follow-up establishes the temporal sequence between HDL exposure and glaucoma onset, substantially mitigating reverse causation. Third, we adopted a comprehensive analytical framework, modeling HDL as a continuous variable, applying clinically relevant categorical cutoffs, and exploring potential non-linear dose-response relationships via restricted cubic spline and two-piece multivariate regression analyses. This multi-pronged approach offers a more granular characterization of the exposure-outcome relationship. Fourth, we extensively adjusted for demographic, lifestyle, and clinical confounders, and further controlled for other lipid parameters (LDL and TG), thereby isolating the independent association of HDL with glaucoma risk. Nevertheless, several limitations should be acknowledged. First, glaucoma was ascertained by self-reported physician diagnosis, which is subject to misclassification and under-ascertainment. In China, community-based screening indicates that over 40% of glaucoma cases aged ≥ 50 years remain undiagnosed, with undiagnosed individuals typically being younger, asymptomatic, and having better visual acuity and smaller cup-to-disc ratios [34]. We found that HDL levels were essentially unrelated to healthcare-seeking behavior (Spearman ρ = -0.053; epsilon-squared = 0.007), supporting non-differential misclassification that would bias our HR toward the null. Thus, our observed association likely represents a conservative estimate. Second, key ocular parameters (IOP, visual field, optic nerve assessment) were unavailable in CHARLS, precluding phenotypic subtyping and adjustment for these confounders. Third, the RCS non-linearity test was not significant (p = 0.183); the inflection point at 1.501 mmol/L is exploratory and requires independent validation. Conclusions In summary, this large-scale, nationally representative cohort study provides robust evidence that elevated serum HDL levels are associated with an increased risk of glaucoma in Chinese middle-aged and older adults. Notably, exploratory RCS analysis suggested a potentially steeper increase at lower HDL concentrations (≤ 1.501 mmol/L), although this non-linear pattern requires validation in independent cohorts. These findings challenge the conventional paradigm of HDL as a uniformly protective factor in ocular health and highlight the complex, context-dependent role of lipid metabolism in glaucoma pathogenesis. Further research is warranted to elucidate the underlying biological mechanisms, validate these findings in diverse populations, and explore whether modulating lipid pathways could offer novel strategies for glaucoma prevention and intervention.