Cholesterol and glaucoma: a systematic review and meta‐analysis

Abstract Purpose Intraocular pressure is the main risk factor for glaucoma; however, additional risk factors may also matter. This systematic review and meta‐analysis were conducted to summarize the evidence regarding the association of cholesterol parameters (total cholesterol, low‐density lipoprotein (LDL) and high‐density lipoprotein (HDL) levels) and glaucoma. Methods Four electronic databases were searched for all publications containing ‘glaucoma’ and one of various forms of ‘cholesterol’ or ‘lipoprotein’. Two independent reviewers screened abstracts and potentially full texts of identified articles for eligibility. Risk of bias was assessed with the Newcastle–Ottawa Scale. A random‐effects meta‐analysis was used to investigate the differences in total cholesterol, LDL and HDL levels between patients with and without glaucoma. Results Overall, 29 observational studies were included in the systematic review and 26 reported quantitative information to investigate differences in cholesterol parameters between patients with glaucoma (N = 7196) and patients without glaucoma (N = 350 441). Patients with glaucoma had significantly higher total cholesterol levels than patients without glaucoma (Mean Difference (MD) 7.9 mg/dl, 95% CI 3.3 to 12.5, p = 0.001) and lower mean HDL levels (MD −2.0 mg/dl, 95% CI: −3.1 to −0.9, p = 0.001). Patients with glaucoma had higher mean LDL levels than patients without glaucoma, albeit not statistically significant (MD 6.1 mg/dl, 95% CI: −4.3 to 16.4, p = 0.251). Conclusion This systematic review and meta‐analysis of observational studies found an association of glaucoma and high total cholesterol and low HDL levels, respectively. Although this supports the hypothesis that lipid levels pose an additional risk for glaucoma development, heterogeneity was substantial and causality cannot be presumed from identified observational studies.


Introduction
Glaucoma is the most frequent cause of irreversible blindness worldwide. The main risk factor is intraocular pressure (IOP). Lowering of IOP may preserve the visual field in patients with´glaucoma and thus remains the most important treatment strategy in glaucoma (Jonas et al. 2017).
However, in some patients IOP lowering does not seem sufficient to stop progression of visual field loss. Therefore, other risk factors are currently investigated (Roddy 2020).
One of these additional risk factors is hypothesized to be lipid levels as polymorphisms in genes encoding proteins important for lipid metabolism such as ABCA1, GAS7 and ATXN2 have been associated with glaucoma (Wiggs & Pasquale 2017). Since a large case-control study found that longterm statin use was associated with a reduced risk of glaucoma interest in lipid levels as an additional risk factor has risen (McGwin et al. 2004). However, statins do not only lower lipid levels, but also seem to have an antiinflammatory and consequently neuroprotective effect (Xu et al. 2017). This further obscures the relationship between glaucoma, lipid levels and lipid-lowering drugs.
The available data on the association between cholesterol and glaucoma are conflicting. Some studies found an association between cholesterol and glaucoma (Kim et al. 2014b), while other studies could not confirm this (Modrzejewska et al. 2015). Therefore, we conducted a systematic review and meta-analysis to summarize the evidence regarding the association of cholesterol parameters (total cholesterol, low-density lipoprotein and high-density lipoprotein levels) and glaucoma.

Registration
We registered our systematic review in International prospective register of systematic reviews (PROSPERO registration number CRD42017067748). The reporting in this publication follows the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement (see Table S1).

Literature search
An experienced medical information specialist (BW) systematically searched the electronic databases: MEDLINE, Embase, Cochrane Central Register of Controlled Trials and Science Citation Index Expanded for all publications from database inception until October 2020. Additionally, all bibliographies of identified articles were scanned to identify potentially relevant manuscripts missed by our search in the databases. Using free term and controlled term formulations the following keywords were searched for in the databases: 'glaucoma' AND 'cholesterol', 'glaucoma' AND 'low-density lipoprotein', 'glaucoma' AND 'highdensity lipoprotein', 'glaucoma' AND 'dyslipidemia' and 'glaucoma' AND 'lipoprotein'. We limited our search to articles published in English.

Study eligibility criteria
All observational studies (cross-sectional, case-control, cohort, survey and surveillance reports) reporting the association between glaucoma and cholesterol including cross-sectional, case-control, cohort, survey and surveillance reports were included. Studies had to report on adult patients (≥18 years) and had to be published in English.
Abstracts and conference proceedings that are not published in peer-reviewed journals were not included. Furthermore, any publication without original data for the quantitative analysis was excluded for the quantitative meta-analysis.
Studies had to ascertain diagnosis of glaucoma by any one or more of the following: fundus photo, fundus examination, retinal nerve fibre layer (RNFL) thickness evaluation, visual field defects, medical records, self-report and/or glaucoma treatment.

Study selection
Two reviewers (LP, MM) independently screened references for inclusion. After pilot testing, we performed dual abstract screening based on the eligibility criteria. Included references underwent subsequent dual full-text review to decide on final inclusion or exclusion of the study. Disagreements were resolved by consensus. The online software 'Covidence' (Van der Mierden et al. 2019) was used for abstracts and full-text screening.

Data extraction
Two investigators (LP, MM) independently extracted the title, name of authors, year of publication, study design, sample size, type of glaucoma, definition of glaucoma and control patients, demographic data and outcome variables (i.e. total cholesterol, LDL, HDL). These data were recorded in a Microsoft Excel (Microsoft Cooperation) spreadsheet.

Risk of bias assessment
The Newcastle-Ottawa Scale (NOS) was used to assess potential risk of bias in each individual study (Hartling et al. 2013). The NOS was developed to assess the quality of nonrandomized studies. Two reviewers (LP, MM) independently assessed the included studies. Disagreements were solved by consensus.

Quantitative synthesis
Differences in total cholesterol, LDL and HDL levels between patients with and without glaucoma were analysed using random-effects meta-analysis. We extracted means and standard deviations from the studies whenever available. For one study reporting only the range, the standard deviation was estimated as (maxÀmin)/6 due to a sufficiently large sample size (Wan et al. 2014). For studies with more than one glaucoma study arm (e.g. normaltension glaucoma (NTG)), we used the weighted mean and pooled standard deviation to combine the arms. For studies with a healthy control group as well as a control group consisting of patients with PEX syndrome without glaucoma, only the healthy controls were considered. Mean differences (MD) between glaucoma and non-glaucoma patients are displayed in forest plots together with their 95% confidence intervals (CI). Possible publication bias was assessed using Egger's linear regression test and Begg's rank correlation test.
Furthermore, we performed leaveone-out analyses to assess the influence of each individual study on the overall results. Heterogeneity was assessed by means of the I 2 value. We tried to explain the heterogeneity through subgroup analyses by grouping studies into whether or not patients using lipid-lowering drugs or patients with diabetes were excluded. Finally, we performed meta-regression analyses to account for metric influential parameters; mean triglyceride levels, age, and BMI of patients as well as percentage of female patients were considered. Weighted means between the glaucoma and control groups were used to obtain one value per study.
All statistical analyses were performed using R version 3.5.1 (R Project for Statistical Computing, Vienna, Austria). In particular, the package 'meta' was used.

Results
Our search yielded 1496 references after removal of duplicates. Figure 1 shows details of the study selection process. After abstract screening, 167 full-text articles were assessed for eligibility. Most were deemed unfit due to different outcome parameters or study populations as well as 'ineligible type of publication'. Finally, we included 29 studies in the qualitative synthesis and out of those 26 studies contained relevant information for the quantitative analysis.

Study characteristics
We included 26 studies investigating the difference in cholesterol levels (i.e. total cholesterol, LDL and/or HDL) between patients with glaucoma (N = 7196) and patients without glaucoma (N = 350 441). Seven studies included all glaucoma, six specifically NTG, five pseudoexfoliation (PEX) glaucoma, and four studies had two glaucoma arms (e.g. NTG and PEX glaucoma). Controls were recruited at the hospital in four studies, from the community in ten studies, and no recruitment method is given in eight studies (see Table 1 and 2).

Study quality
The NOS score ranged from 1 to 6 showing overall high risk of bias for all 22 studies included in the meta-analysis.

Total cholesterol
Twenty-three studies reported sufficient information to investigate differences in total cholesterol levels in patients with glaucoma (N = 6751) and patients Patients with glaucoma had significantly higher total cholesterol levels than patients without glaucoma (MD 7.9 mg/dl, 95% CI 3.3 to 12.5, p = 0.001; Fig. 2). However, there was substantial heterogeneity (I 2 = 86%, p < 0.01).
To detect publication bias, Eggers test and Beggs test were performed. Both were not significant (p = 0.271, respectively 0.653) suggesting unbiased results.
Subgroup analysis revealed that studies excluding patients taking lipid-lowering drugs had a smaller mean absolute difference of total cholesterol levels between glaucoma patients and controls. In detail, studies excluding patients on lipid-lowering drugs showed an MD of 3.7 mg/dl (95% CI À2.3 to 9.6) between glaucoma and control patients, and studies including patients on lipid-lowering drugs an MD of 8.8 mg/dl (95% CI 3.4 to 14.3).
Another subgroup analysis showed that studies excluding diabetics also had a smaller mean absolute difference. In detail, studies excluding diabetics had a MD of 4.6 mg/dl (95% CI À0.5 to 9.7) between glaucoma patients and controls, studies with diabetics a MD of 5.6mg/dl (95% CI À2.1 to 13.4) and studies with no information of diabetics 13.9 mg/dl (95% CI À5.6 to 33.3).
In meta-regression analysis, triglycerides (p = 0.277), age (p = 0.854), BMI (p = 0.518) and gender (p = 0.745) did not appear to modify the association between glaucoma and total cholesterol, but were also not able to explain the large heterogeneity.

Low-density lipoproteins
Fourteen studies provided information to investigate the difference in LDL levels in patients with glaucoma (N = 627) and patients without glaucoma (N = 822). The mean LDL levels strongly varied between studies and ranged from 104.1 mg/dl in the study by Rasoulinejad et al. (2015) to 155.5 mg/dl in Pavljasevic & Asceric (2009).

High-density lipoproteins
Nineteen studies contained information to investigate the difference in HDL levels in patients with glaucoma (N = 1713) and patients without glaucoma (N = 43 259). The mean HDL levels strongly varied between studies and ranged from 41.1 mg/dl in the study by Modrzejewska et al. (2015) to 72.1 mg/dl in Bossuyt et al. (2015). Patients with glaucoma had lower mean HDL levels (MD À2.0 mg/dl, 95% CI: À3.1 to À0.9, p = 0.001; Fig. 4) than patients without glaucoma and the heterogeneity was substantial (I 2 = 69%, p < 0.01).
In meta-regression, triglycerides modified the association between HDL and glaucoma (p = 0.007). Age (p = 0.291), BMI (p = 0.235) and gender (p = 0.238) did not modify the association between HDL and glaucoma.

Discussion
This systematic review and meta-analysis of observational studies found that patients with glaucoma had higher mean total cholesterol levels and lower HDL levels than patients without glaucoma, whereas there was no significant difference regarding LDL. This may support the conjecture that total cholesterol levels and HDL levels pose an additional risk factor for glaucoma. However, no causality can be drawn from observational studies.
The studies on the association between glaucoma and lipid levels included in our systematic review found different results, which resulted in high heterogeneity across studies in the meta-analyses. There are several possible reasons for this. First of all, different types of glaucoma were included. Secondly, there were differences in inclusion criteria among the individual studies. Furthermore, the exclusion criteria of individual studies differed greatly. For example, some studies excluded patients with lipidlowering drugs, while others included them or did not report on the use of lipid-lowering medications (see Table 1).
Subgroup analysis showed that studies excluding patients on lipid-lowering drugs showed smaller mean differences in total cholesterol levels and HDL levels between glaucoma and control patients. It has been suggested that statin use reduces the incidence of glaucoma (McCann et al. 2016). Therefore, confounding by indication is a possibility in this meta-analysis and Fig. 2. Random-effects meta-analysis on total cholesterol level differences between glaucoma patients and controls. The differences are expressed on an absolute scale, that is mg/dl. our findings may be exaggerated. Otherwise it could be that patients taking lipid-lowering drugs have high cholesterol and thus studies excluding these patients had smaller mean differences. This would support our notion that cholesterol plays a role in the development of glaucoma.
Statin use has been shown to protect against glaucoma development and progression. It seems that the beneficial effect of statins is associated less with lowering lipid levels and more with other properties of statins such as possible anti-inflammatory properties (Talwar et al. 2017). In this metaanalysis, LDL levels were not associated with glaucoma, which might support the concept that statin lowers the glaucoma risk via other mechanism than lowering lipid levels. As our study found that HDL and total cholesterol levels were associated with glaucoma, it might be interesting to see whether statins, which additionally increase HDL levels such as pitavastatin (Pirillo & Catapano, 2017), may be even better in reducing glaucoma development than statins pre-eminently reducing LDL. Fig. 3. Random-effects meta-analysis on LDL level differences between glaucoma patients and controls. The differences are expressed on an absolute scale, that is mg/dl. Fig. 4. Random-effects meta-analysis on HDL level differences between glaucoma patients and controls. The differences are expressed on an absolute scale, that is mg/dl. Topical beta-blockers seem to lower HDL (Stewart et al. 1999) (Yamamoto et al. 1996). This was confirmed in the Blue Mountains Eye study, although the adverse effect on HDL was seen exclusively in men (Mitchell et al. 2000). Most of the included studies did not report on topical treatment used and thus no meta-regression could be performed. It may be assumed that a large number of patients use topical beta-blockers and the effect seen in this meta-analysis is due to this side effect.
We want to mention several limitations of this meta-analysis. Firstly, we only included observational studies. Observational data cannot prove causality. Further, there is some evidence that cholesterol influences IOP (Wang et al. 2019). Therefore, this study cannot answer whether cholesterol leads to glaucoma via IOP or whether this association is independent of IOP. Secondly, there was a large heterogeneity between studies concerning types of studies, inclusion and exclusion criteria and the selection of the control group. Additionally, Eggers test was significant in our meta-analysis on HDL and glaucoma suggesting publication bias. Thirdly, statin use is a potential confounder in this study, which could not be fully accounted for. No meta-regression could be performed due to the small number (n = 2) of studies reporting the percentage of glaucoma and control patients taking lipid-lowering drugs. Another potential bias are errors in the diagnosis of glaucoma, that is glaucoma cases were misidentified as normal and vice-versa. Finally, our results were statistically significant. However, mean absolute differences in total cholesterol and HDL were still small (MD 9.2 mg/dl, respectively À2.3 mg/dl) and thus, of unclear clinical relevance.
In conclusion, this meta-analysis of observational studies found an association of glaucoma and high total cholesterol and low HDL levels, respectively. Although this supports the hypothesis that lipid levels pose an additional risk for glaucoma development, we were unable to explain the large heterogeneity and causality can generally not be presumed from identified observational studies. PROSPERO registration number CRD42017067748

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1. Baujat-and Funnel-Plot for meta-analysis on total cholesterol and glaucoma. Figure S2. Baujat-and Funnel-Plot for meta-analysis on LDL and glaucoma. Figure S3. Baujat-and Funnel-Plot for meta-analysis on HDL and glaucoma.