Association of Statin Use and High Serum Cholesterol Levels With Risk of Primary Open-Angle Glaucoma

Key Points

Question  Are self-reported elevated cholesterol level and statin use associated with risk of primary open-angle glaucoma?

Findings  In this population-based cohort study of 887 incident cases of primary open-angle glaucoma from 136 783 participants who contributed person-time for up to 15 or more years and provided repeated updated data on statin use and cholesterol levels, 5 or more years of statin use was not associated with risk of primary open-angle glaucoma, and higher total serum cholesterol was also not associated with risk of primary open-angle glaucoma.

Meaning  Statin use, which is highly prevalent among older persons at risk for primary open-angle glaucoma, was not associated with risk of primary open-angle glaucoma.

Importance  The use of statins (hydroxymethylglutaryl coenzyme A inhibitors) has been associated with a lower risk of primary open-angle glaucoma (POAG); however, results have been conflicting, and little is known about the association between high cholesterol levels and POAG.

Objective  To assess the association of elevated cholesterol levels and statin use with incident POAG.

Design, Setting, and Participants  This study used data collected biennially from participants aged 40 years or older who were free of glaucoma and reported eye examinations, within 3 population-based cohorts: the Nurses’ Health Study (N = 50 710; followed up from 2000 to 2014), the Nurses’ Health Study 2 (N = 62 992; 1999-2015), and the Health Professionals Follow-up Study (N = 23 081; 2000-2014). Incident cases of POAG were confirmed by medical record review. The analyses were performed in January and November 2019.

Exposures  Biennially updated self-reported information on elevated cholesterol level status, serum cholesterol levels, and duration of statin use.

Main Outcomes and Measures  Multivariable-adjusted relative risks (RRs) and 95% CIs were estimated using Cox proportional hazards regression models on pooled data, with stratification by cohort.

Results  Among the 136 783 participants in the 3 cohorts (113 702 women and 23 081 men), 887 incident cases of POAG were identified. A 20-mg/dL increase in total serum cholesterol was not associated with risk of POAG (RR, 1.02 [95% CI, 0.98-1.07]; P = .32). Any self-reported history of elevated cholesterol was not associated with risk of POAG (RR, 1.11 [95% CI, 0.94-1.31]). A history of any statin use was not associated with risk of POAG (RR, 0.93 [95% CI, 0.80-1.10]). Use of statins for 5 or more years vs never use of statins was not associated with risk of POAG (RR, 0.93 [95% CI, 0.75-1.15]; P = .49 for linear trend).

Conclusions and Relevance  Among adults aged 40 years or older, higher self-reported total serum cholesterol levels and statin use compared with never use of statins were not associated with risk of POAG.

Statins (hydroxymethylglutaryl coenzyme A inhibitors) are cholesterol-lowering drugs prescribed for the prevention and treatment of cardiovascular disease.^1,2 The mechanisms by which statins may lower the risk of cardiovascular disease include those independent of cholesterol-lowering properties.^3,4 Statins have pleiotropic effects^5,6 including antioxidant and immunomodulatory effects, effects on thrombosis formation, and effects on the nitric oxide synthase system. Statins may also have beneficial effects on cancers,^5,7 inflammatory diseases,⁸ and neurologic disorders.^9-11 In vitro studies have suggested that statins may lower intraocular pressure (IOP)¹² and protect retinal ganglion cells against glaucomatous damage,^11,13-18 indicating statins’ potential role in mitigating glaucoma pathogenesis.

Several observational studies have evaluated the association between statin use and the risk of primary open-angle glaucoma (POAG). However, the results have been inconsistent: some studies have observed inverse associations,^19-21 whereas others have observed null²² or even adverse associations.²³ Also, studies reporting inverse associations observed stronger associations with at least 2 years of use,^19-21 but data are lacking on the association with longer use of statins.²⁴ Furthermore, for some studies, confounding by indication due to hyperlipidemia (and concomitant hypertension and diabetes) has been raised as a concern,^25,26 and the association between hyperlipidemia and POAG has been unclear.^25,26 Our primary a priori hypothesis was that longer duration of statin use is associated with a lower risk of POAG, and our secondary hypothesis was that higher cholesterol levels are associated with a higher risk of POAG. Therefore, we examined the association of hyperlipidemia as well as statin use with risk of POAG among 50 710 women in the Nurses’ Health Study (NHS), 62 992 women in the NHS2, and 23 081 men in the Health Professionals Follow-up Study (HPFS) followed up for 15 or more years.

The NHS began in 1976 with 121 700 US female registered nurses (age, 30-55 years) who completed mailed health and lifestyle questionnaires.²⁷ The NHS2 started in 1989 with 116 429 US female registered nurses (age, 30-55 years) who returned similar questionnaires.²⁸ The HPFS began in 1986 with 51 529 US male health care professionals (age, 40-75 years) who also returned health questionnaires.²⁹ Participants were followed up biennially with mailed questionnaires that asked about diseases such as cancer, cardiovascular disease, and glaucoma; the rate of follow-up has been high (>85%). The Human Research Committees of Brigham & Women’s Hospital, Massachusetts Eye and Ear Infirmary, and Harvard T. H. Chan School of Public Health approved this study and allowed participants’ completion of questionnaires to be considered as implied consent.

Because participants were first asked about current and lifetime statin use in 1999 in the NHS2 and in 2000 in the NHS and HPFS, 1999-2000 is considered the baseline for our study. Follow-up ended with a glaucoma diagnosis, incident cancer (as cancer causes profound changes in behavior), death, loss to follow-up, or the study’s end (2014 in NHS and HPFS and 2015 in NHS2), whichever occurred earliest. The analyses were performed in January 2019.

At baseline, we excluded the following participants: (1) 29 233 women in the NHS and 1596 men in the HPFS who did not complete the food frequency questionnaire (in 1980 for NHS and in 1986 for HPFS), because for the NHS and HPFS, the association between diet and glaucoma was the primary aim at initiation and case confirmation occurred only for those with diet data (in the NHS2, there was no such restriction for case confirmation); (2) 16 378 women in the NHS, 8847 men in the HPFS, and 4701 women in the NHS2 who had cancer (except nonmelanoma skin cancer); (3) 5132 women in the NHS, 2321 men in the HPFS, and 543 women in the NHS2 who had prevalent glaucoma; (4) 5083 women in the NHS, 4322 men in the HPFS, and 435 women in the NHS2 who died before baseline; (5) 4591 women in the NHS, 3554 men in the HPFS, and 6523 women in the NHS2 whose last completed questionnaire was at baseline (who were lost to follow-up); (6) 564 women in the NHS, 778 men in the HPFS, and 29 895 women in the NHS2 who never reported an eye examination (in the NHS2, the eye examination question was asked once in 2013); and (7) 7270 women in the NHS, 4452 men in the HPFS, and 7471 women in the NHS2 who did not complete baseline questionnaires or had missing baseline statin use information. After these exclusions, 53 449 women in the NHS, 25 659 men in the HPFS, and 66 861 women in the NHS2 were potentially eligible to ever contribute person-time to the analysis. Among these participants, at each 2-year period of observation, we allowed only those who were 40 years or older, who reported a recent eye examination, and who had information on serum cholesterol levels. Participants who did not meet these provisional criteria at any time were allowed to contribute person-time at later periods when they met these criteria. Overall, 50 710 women in the NHS, 23 081 men in the HPFS, and 62 992 women in the NHS2 ever contributed person-time in this analysis.

We included 887 individuals with confirmed cases of incident POAG (522 women in the NHS, 209 men in the HPFS, and 156 women in the NHS2). Glaucoma cases were first ascertained in biennial questionnaires that asked participants about physician diagnoses of glaucoma. For those self-reporting glaucoma diagnoses, we obtained permission to contact their eye care professionals. We asked the eye care professionals to send visual fields (VFs) with either medical records or a completed glaucoma questionnaire with items on maximal IOP, filtration apparatus status, optic nerve structural information, ophthalmic surgery, and earliest date of VF loss. A glaucoma specialist (L.R.P.) reviewed records to confirm cases using standardized criteria. After medical record review, we excluded participants with isolated elevated IOP or optic disc cupping and other forms of glaucoma (eg, exfoliation, closed-angle, or secondary), as well as unconfirmed reports (eg, supporting medical records could not be obtained or the participant did not give permission to review records).³⁰ Those who self-reported glaucoma but were not considered to have definite or probable POAG were censored as of the date of self-reported diagnosis.

For confirmation of POAG, we required (1) results of gonioscopy indicating that the filtration angle was not occludable in either eye (516 cases [58.2%]) or slitlamp biomicroscopy demonstrating open angles and pharmacologic dilation without adverse effects (371 cases [41.8%]); (2) results of slit-lamp biomicroscopy showing no signs in either eye of pigment dispersion syndrome, uveitis, exfoliation syndrome, trauma, or rubeosis iridis; and (3) reproducible VF defects consistent with POAG on 2 or more reliable tests. For VF defects, the type of perimetry was not restricted. However, full static threshold testing was documented in 869 cases (98.0%), and kinetic VFs were documented in 5 cases (0.6%). For static threshold or suprathreshold tests, we used the reliability definitions of fixation loss of 33% or less, false-positive rate of 20% or less, and a false-negative rate of 20% or less; for kinetic VFs, VFs were considered reliable unless there were examiners’ notes to the contrary.

Participants were asked biennially from when the cohorts began about any clinician-diagnosed elevated cholesterol. In addition, participants were asked to report the date when they received a diagnosis of elevated cholesterol. Once a participant reported such a history, she or he was considered to always have a history of elevated cholesterol.

Participants were asked about their most recent (≤5 years prior) total serum cholesterol (10 categories from <140 to ≥330 mg/dL [to convert to millimoles per liter, multiply by 0.0259]; asked in 1988, 1990, 1994, 2000, 2004, and 2008 in the NHS; 1986, 1990, 2000, 2004, and 2008 in the HPFS; and 1989, 2005, and 2009 in the NHS2). Self-reported serum cholesterol levels in the NHS were previously validated against actual measured serum cholesterol levels.^31,32 We evaluated cumulative mean serum cholesterol values, which may better represent long-term exposure and have less random measurement error.³³ With cumulative averaging, the mean of all available information as of a risk period was used (eg, in the NHS for the risk period 2008-2010, the mean of the 1988, 1990, 1994, 2000, 2004, and 2008 values was used).

In the NHS and HPFS, we asked about current use of any cholesterol-lowering drugs in the 1988 through 1998 questionnaire cycles; in 2000, the use of any statins (including lifetime cumulative duration of use) was queried separately from use of nonstatin cholesterol-lowering drugs. In 2004-2012, the use of specific statin types was queried. In the NHS2, the use of statins (including cumulative duration of use) was queried in 1999-2003; in 2005-2013, specific types of statins used were queried. In the NHS2, we also asked about use of nonstatin cholesterol-lowering drugs from 1999 to 2013.

For deriving the status and duration of statin use, if information about statins was missing, we carried forward information for 1 cycle only for never users and current users, as their status remained unchanged more than 75% of the time from cycle to cycle; past users with missing data in the current cycle were skipped from contributing person-time and allowed to contribute in a later cycle when information was available. Participants with missing statin data on 3 consecutive questionnaire cycles were censored as of the date of the first cycle from which data were missing.

After confirming that the results from the 3 cohorts did not show significant heterogeneity (eg, P = .31 for heterogeneity for statin use duration association), we pooled the data for greater statistical power. We calculated POAG incidence by dividing incident cases by person-years for each exposure category. For multivariable analyses, we conducted Cox proportional hazards regression analysis,³⁴ while simultaneously controlling for potential time-varying glaucoma risk factors. To control as finely as possible for confounding by age, calendar time, cohort, and any possible interactions, we stratified the analysis jointly by cohort, age in months, and calendar year of each questionnaire cycle. We also adjusted for the following updated potential covariates: body mass index (calculated as weight in kilograms divided by height in meters squared); cigarette smoking (pack-years); hypertension; history of using β-blockers, diuretics, other blood pressure–lowering medications, or other nonstatin cholesterol-lowering medicines; type 1 or 2 diabetes; physical activity (quartiles of metabolic equivalents of task–hours per week); any cardiovascular or cerebrovascular disease (myocardial infarction, stroke, transient ischemic attack, or coronary artery bypass graft); cumulatively averaged serum cholesterol levels; cumulatively averaged intake of alcohol and caffeine; family history of glaucoma; race/ethnicity; and, among women, age at menopause and postmenopausal hormone use.

We estimated hazard ratios as measures of relative risks (RRs) and 95% CI. We examined the possibly nonlinear association between total serum cholesterol and RR of POAG nonparametrically with restricted cubic splines.³⁵ Tests for nonlinearity used the likelihood ratio test, comparing the model with only the linear term with the model with the linear and the cubic spline terms. All significance tests were 2-sided, and results were deemed statistically significant at P < .05. Analysis was performed with SAS statistical software, version 9.4 (SAS Institute Inc).

We conducted sensitivity analyses where prevalent statin users as of baseline were excluded, and we included only those who first reported statin use at baseline or later to minimize selection bias and to mimic a clinical trial setting.³⁶ Finally, we evaluated whether associations may differ by age, hyperlipidemia diagnosis, sex, or family history of glaucoma. For interaction testing, a product term of the effect modifier and duration of statin use were added to models with the 2 main effects, and the significance of this term was evaluated with Wald tests.

During 1 485 182 person-years of follow-up, we identified 887 individuals with incident POAG (mean [SD] age, 68.6 [9.6] years; mean [SD] mean deviation in the worse eye, –5.6 [5.2] dB). Individuals who used statins for 5 or more years were more likely to have reported a history of elevated cholesterol levels, higher total serum cholesterol levels, and a higher prevalence of cardiovascular disease and risk factors (Table 1). Longer statin use was not associated with known risk factors for glaucoma such as African American race/ethnicity or having a family history of glaucoma.

Any history of elevated cholesterol (defined as a self-report of a physician’s diagnosis of high cholesterol or ever reporting a total serum cholesterol of ≥240 mg/dL) was not associated with risk of POAG (multivariable-adjusted RR, 1.11; 95% CI, 0.94-1.31) (Table 2). We observed that a 20-mg/dL increase in total serum cholesterol was not associated with risk of POAG (RR, 1.02; 95% CI, 0.98-1.07; P = .32) (Figure). When those with no history of elevated cholesterol levels were divided into 2 categories (with and without cholesterol-lowering treatment) and those with elevated cholesterol were divided into 4 categories (no treatment, history of nonstatin drug use only, history of statin use only, and history of having used both statins and nonstatins), we did not observe major differences in risk of POAG compared with those with no elevated cholesterol, likely because of the limited power (Table 2).

In models adjusted for current use of nonstatin cholesterol-lowering drugs, cumulatively updated total serum cholesterol levels, and other covariates, any statin use (eTable in Supplement 1) was not associated with POAG (RR, 0.93; 95% CI, 0.80-1.10). The RR was 0.91 (95% CI, 0.63-1.32) for past statin use and was 0.94 (95% CI, 0.80-1.10) for current use. We did not observe major differences by type of statin currently used; for the most commonly used statin (atorvastatin calcium), the RR was 0.94 (95% CI, 0.71-1.24).

We did not observe associations with longer statin use and POAG risk (Table 3). Compared with never users, the RR was 0.95 (95% CI, 0.74-1.22) for those who used statins for less than 2 years, 0.93 (95% CI, 0.73-1.17) for those who used statins for 2 to 4 years, and 0.93 (95% CI, 0.75-1.15) for those who used statins for 5 or more years (P = .49 for trend). The similar lack of association was observed when 5 or more years of use was evaluated among current users. Each year of statin use was associated with a nonsignificant 1% lower risk of POAG (RR, 0.99; 95% CI, 0.97-1.01; P = .35). To allow for comparison with previous studies that evaluated shorter use durations, we evaluated the association of 2 or more years of statin use with POAG and found a multivariable-adjusted RR of 0.93 (95% CI, 0.78-1.11). When we evaluated even longer durations of statin use, the multivariable-adjusted RR for 5 to 9 years of statin use was 0.98 (95% CI, 0.78-1.22) and the multivariable-adjusted RR for 10 or more years of statin use was 0.79 (95% CI, 0.55-1.15).

The association with longer duration of statin use did not differ by family history of glaucoma (no history: RR, 0.83; 95% CI, 0.65-1.07; vs history: RR, 1.02; 95% CI, 0.59-1.78; P = .83 for interaction) or sex (female: RR, 0.99; 95% CI, 0.78-1.26; vs male: RR, 0.79; 95% CI, 0.51-1.22; P = .32 for interaction) (Table 4). There was no association between use of statins for 5 or more years compared with never use of statins and risk of POAG among those who were 65 years or older (≥65 years: RR, 0.87 [95% CI, 0.69-1.10]) or younger than 65 years (<65 years: RR, 1.13 [95% CI, 0.72-1.77]) (P = .02 for interaction).

In sensitivity analyses in which statin users were restricted to those who initiated statin use on or after 1999-2000 (402 individuals with POAG), we also did not observe associations with use of statins for 5 or more years and risk of POAG (RR, 0.91; 95% CI, 0.51-1.59; P = .74 for trend), providing support for the inverse associations in the main analyses. Because the use of nonstatin cholesterol-lowering drugs was specifically asked about from 1999 to 2000, we evaluated the association with duration of use of nonstatin cholesterol-lowering drugs in these analyses (although the statistical power was low), and we did not observe an association (≥5 years’ duration: RR, 1.17; 95% CI, 0.28-4.96; P = .72 for trend).

In this study of 136 783 participants followed for 15 or more years, among adults 40 years or older, self-reported higher serum total cholesterol levels were not associated with risk of POAG. Longer statin use (≥5 years), compared with never use, was also not associated with risk of POAG. Carefully determining the association with hyperlipidemia is important to help evaluate the possibility of confounding by indication³⁷ in studies that investigate associations with cholesterol-lowering drugs, such as statins. Other studies^26,38 have observed associations with higher cholesterol levels and higher risk of POAG (eg, a meta-analysis²⁶ reported an RR of 1.40 [95% CI, 0.73-2.68] for hyperlipidemia history); however, we observed no association with higher serum cholesterol levels and risk of POAG.

Although previous studies have reported inverse associations with statin use and risk of POAG,^19,22,23 a history of statin use was not associated with POAG in our study. Also, while McGwin et al¹⁹ (RR, 0.60; 95% CI, 0.39-0.92; 667 cases), Marcus et al²¹ (RR, 0.46; 95% CI, 0.23-0.92; 108 cases), and Stein et al²⁰ (RR, 0.92; 95% CI, 0.87-0.98; 10 266 cases) observed that longer duration of statin use compared with never use was associated with a lower risk of POAG, we did not observe an association with longer statin use and risk of POAG. Also, similar to findings in 2 other studies,^20,23 we did not observe notable associations with any of the 5 statin types.

The potential mechanisms by which statins may lower the risk of POAG include IOP-lowering and neuroprotective mechanisms. Statins affect the activities of myosin II adenosine triphosphatase and ρ kinase in the trabecular meshwork that increase nitric oxide production and aqueous outflow facility, which may lead to some IOP lowering.^39-41 Also, greater production of nitric oxide would increase the blood flow to the optic nerve.^42,43 Other neuroprotective effects^{10,11,41,44-48} of statins include anti-excitotoxic,^11,49,50 anti-apoptotic,^45,46 and anti-inflammatory⁴⁷ properties that may protect retinal ganglion cells. Several genes involved in cholesterol metabolism (ABCA1, CAV1, ARHGEF12, and DGKG) have also been associated with IOP and POAG in large genome-wide association studies.^51,52 However, given the inconsistencies in results among studies of statins and glaucoma, more studies on hyperlipidemia, statins, and their role in glaucoma^53,54 are warranted.

Although our study was large, with a long follow-up, repeated assessment of statin use and hyperlipidemia, and the availability of key covariates, our study had several limitations. We likely had misclassification of statin use and cholesterol levels, as we did not measure participants’ recall errors. This misclassification would have biased associations toward the null. Our case ascertainment method had low sensitivity; however, methodologically, RR estimates are still valid if the case definition is highly specific and the case ascertainment is unassociated with exposure.⁵⁵ Also, as participants were mostly white health care professionals all free of cancer, our results may not be generalizable to other community-dwelling older populations with different underlying risks for POAG (eg, in predominantly African American populations). Despite our adjusting for many potential confounders, there may still have been some residual confounding. Also, to comprehensively evaluate the association between hyperlipidemia, cholesterol-lowering treatment, and POAG, we performed many secondary analyses; many of these results showed weak associations. The finding of a possible interaction with age for statin duration may have been owing to chance and must be interpreted with caution and confirmed in other studies.

In this large study of 136 783 participants with overall follow-up duration of 15 or more years, self-reported serum cholesterol levels and statin use were not associated with POAG risk.

Accepted for Publication: February 5, 2019.

Corresponding Author: Jae H. Kang, ScD, Channing Division of Network Medicine, Brigham and Women’s Hospital, Harvard Medical School, 181 Longwood Ave, Boston, MA 02115 (nhjhk@channing.harvard.edu).

Published Online: May 2, 2019. doi:10.1001/jamaophthalmol.2019.0900

Retraction and Replacement: This article was retracted and replaced on March 12, 2020, to fix errors in the results, discussion, Figure, and Supplement 1 (see Supplement 2 for the retracted article with errors highlighted and Supplement 3 for the replacement article with corrections highlighted).

Author Contributions: Dr Kang had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Kang, Boumenna, Khawaja, Pasquale.

Acquisition, analysis, or interpretation of data: Kang, Stein, Khawaja, Rosner, Wiggs, Pasquale.

Drafting of the manuscript: Kang, Pasquale.

Critical revision of the manuscript for important intellectual content: Boumenna, Stein, Khawaja, Rosner, Wiggs, Pasquale.

Statistical analysis: Kang, Rosner.

Obtained funding: Wiggs, Pasquale.

Administrative, technical, or material support: Boumenna, Pasquale.

Supervision: Pasquale.

Conflict of Interest Disclosures: Dr Stein reported receiving grants from the National Institutes of Health and grants from Research to Prevent Blindness during the conduct of the study. Dr Khawaja reported receiving personal fees from Allergan, Novartis, Thea, Grafton Optical, and Santen outside the submitted work. Dr Rosner reported receiving grants from the National Institutes of Health during the conduct of the study. Dr Kang reported receiving grants from the National Institutes of Health during the conduct of the study. Dr Wiggs reported receiving grants from the National Eye Institute during the conduct of the study and grants from the National Eye Institute outside the submitted work. Dr Pasquale reported receiving personal fees from Bausch & Lomb, Eyenovia, and Verily Inc outside the submitted work. No other disclosures were reported.

Funding/Support: This work was supported by grants UM1 CA186107, UM1 CA176726, UM1 CA167552, EY09611, and EY015473 from the National Institutes of Health (Dr Pasquale).

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.