Association Between Statin Use and Dry Eye Disease: A Systematic Review and Meta-Analysis

Topic

This systematic review and meta-analysis evaluated the literature-pooled association between statins and dry eye disease (DED).

Clinical Relevance

Statins, a common treatment modality for dyslipidemia, have been proposed as a potential contributor to DED via their activity in meibomian gland epithelial cells. However, single studies show mixed evidence, and there remains an unmet clinical need to clarify whether statin exposure is associated with DED.

Methods

This review was reported in accordance with the Preferred Reporting Items for Systematic Reviews of Interventions (PRISMA) 2020 statement and was registered a priori on PROSPERO (CRD420251238004). Ovid MEDLINE, Embase, CINAHL, Web of Science, CENTRAL, and the reference lists of relevant reviews were searched from inception to November 2025 for studies reporting the association between statin use and DED. Random-effects meta-analysis using inverse-variance weighting was conducted to pool effect estimates as odds ratios with 95% confidence intervals (CIs). Study risk of bias was appraised using the ROBINS-E tool, and the certainty of the evidence was reported using the GRADE framework.

Results

Six observational studies were included in the meta-analysis ( n = 560,821; 356,012/559,141 [63.7%] statin users). The pooled analysis revealed a significant positive association between statins and DED (odds ratio 1.09, 95% CI 1.05-1.13, P <.001), with an absolute risk difference of 10.2 more DED cases per 1000 (95% CI 5.5 more to 15.2 more). This result was derived from very low-certainty evidence given limitations in study design and serious inconsistency. Subgroup and sensitivity analyses for risk of bias ( P =.123), method of outcome ascertainment ( P =.737), type of effect estimate ( P =.496), and leave-one-out analyses showed no evidence of effect modification and demonstrated consistent direction of association across studies.

Conclusion

Statin use was associated with a small but statistically significant increase in DED, limited by very low-certainty evidence. Physicians should monitor for and educate patients on ocular surface symptoms in patients using statins with pre-existing DED risk factors. Future studies should use standardized DED diagnostic criteria to investigate the impact of statin dose, type, and duration to better characterize this potential association.

INTRODUCTION

D ry eye disease (DED), or keratoconjunctivitis sicca, is a common disorder of the ocular surface characterized by a loss of homeostasis of the tear film, hyperosmolarity, increased ocular inflammation, ocular damage, and neurosensory abnormalities. ^, DED is a disturbance of the lacrimal functional unit, consisting of the ocular surface, lacrimal glands, and central nervous system, resulting in reduced tear production or excessive tear evaporation. The global prevalence of DED is estimated to be 11.6%, although estimates vary between 5% and 50% depending on the diagnostic criteria. ^, DED is associated with decreased quality of life, as patients may report problems with reading, working, computer use, watching television, and driving. In the United States alone, the estimated annual burden of direct costs for DED in 2013 was over $3.8 billion USD, in addition to indirect costs from lost productivity. ^, Female sex, increased age, East Asian ethnicity, digital screen exposure, and systemic inflammatory conditions are known risk factors for DED. ^, Further, systemic medications, such as antihistamines and antidepressants, have been associated with the onset or progression of DED due to the disruption of lacrimal gland function and meibomian gland integrity. ^,

Statins, a common treatment modality for dyslipidemia and cardiovascular disease prevention, have also been implicated with DED. Statins are among the most commonly prescribed medications worldwide, with an estimated global prevalence of over 68 defined daily doses per 1000 adults aged ≥40 years. Their mechanism of action primarily involves the inhibition of 3‑hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase, the rate-limiting enzyme in the mevalonate pathway of cholesterol synthesis. ^, Since HMG-CoA reductase is also expressed in human meibomian gland epithelial cells, it has been posited whether statins may inadvertently inhibit endogenous meibomian lipid synthesis, alter meibum composition, and induce evaporative dry eye. Conversely, other hypotheses assert that statins exhibit pleiotropic anti-inflammatory and immunomodulatory effects, suggesting a possible protective role against blepharitis and ocular surface inflammation, common contributing factors to DED. Accordingly, the overall impact of statins on DED remains uncertain, creating a clinical dilemma for the management of patients with both dyslipidemia and ocular surface disease. To date, there has been no comprehensive synthesis of the association between statin use and DED and its symptoms. Thus, this systematic review and meta-analysis aims to quantify the literature-pooled association between statin use and DED.

METHODS

This work abided by the tenets of the Declaration of Helsinki. As this work relied entirely on published primary research, the requirement for institutional ethics board review was waived. This systematic review and meta-analysis were conducted in accordance with the Cochrane Handbook and used the Preferred Reporting Items for Systematic Reviews of Interventions (PRISMA) 2020 statement to report the findings. Tables S1 and S2 show the completed PRISMA and PRISMA Abstract checklists, respectively. The protocol for this review was prospectively registered on the International Prospective Register of Systematic Reviews (PROSPERO: CRD420251238004). All decisions regarding eligibility criteria, search strategy, study selection, data extraction, risk of bias assessment, and analysis were established a priori .

Study identification

A systematic search of MEDLINE (Ovid), Embase and Embase Classic (Ovid), CINAHL Plus (EBSCOhost), Web of Science (Core Collection), and CENTRAL from database inception to November 22, 2025 was conducted. Tables S3 to S7 present the complete search strategies for each database. The reference lists of relevant reviews were also manually searched for additional eligible studies.

Eligibility criteria

Cohort, case-control, cross-sectional, and randomized controlled trials that investigated current or former use of any statin medication for cholesterol management in human patients were eligible for inclusion. Studies investigating lipid-lowering drugs without providing additional information on drug type were excluded due to ambiguity with nonstatin lipid-lowering drugs. Eligible studies reported on the development of DED, which was ascertained via review of medical records, ophthalmologist clinical diagnoses, or patient-reported questionnaires; DED could be classified using clinical signs, symptoms, or both. Studies were also required to report associations between statin use and DED using odds ratios (ORs), risk ratios (RRs), or hazard ratios (HRs), comparing a statin-exposed group with an unexposed group. Noncomparative studies were excluded. No restrictions on language or grey literature were imposed.

Study selection

Three reviewers (D.G., W.Q., V.C.) screened all titles and abstracts independently and in duplicate, and subsequently screened relevant full-texts using Covidence (Veritas Health Innovation, Melbourne, Australia). Non-English articles were translated into English when necessary. Disagreements were resolved by discussion or consultation with a third reviewer (B.K.T.) when necessary.

Data extraction

Data extraction was performed independently and in duplicate by three reviewers (D.G., W.Q., V.C.) using prepiloted extraction sheets on Microsoft Excel. Data related to the study designs, recruitment, participant demographics, comedication use, comorbidities, and exposure to statins, including dose, type, and duration, were extracted. For each outcome, its definition, method of ascertainment, prevalence, effect estimate, confidence interval (CI), and regression model covariates were also documented. In cases of unclear or missing data, the corresponding authors were contacted to seek clarification or to obtain unpublished data.

Risk of bias assessment and certainty of evidence

The same reviewers (D.G., W.Q., V.C.) assessed the risk of bias of the included studies independently and in duplicate using the Cochrane Risk Of Bias In Nonrandomized Studies of Exposures (ROBINS-E) tool for nonrandomized studies and the Cochrane RoB 2 tool for randomized studies. ^, The robvis shiny web app was used to visualize the risk of bias assessments. Any disagreements were resolved through discussion or consultation with a third reviewer (B.K.T.) to reach consensus. The Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) framework was used to assess the certainty (quality) of evidence in the meta-analysis.

Data synthesis and data analysis

The included studies reported point estimates and 95% CIs as ORs or RRs. All estimates were converted to ORs using the baseline risk estimates from each study if the rare disease assumption was met (<10%) and if OR was sufficiently close to the null. ^,^, Assuming that heterogeneity exists between studies, effect estimates were pooled using random-effects meta-analysis through inverse-variance weighting. Studies reporting stratified statin exposure groups without an overall effect estimate were excluded from the meta-analysis due to the shared reference group. The median study-level baseline risk of DED was used to convert the relative measure of association (OR) to an absolute risk difference. If meta-analysis was not possible, the results of individual studies were presented narratively.

Statistical heterogeneity was assessed by visually inspecting forest plots for the consistency of point estimates and the overlap of CIs as well as the I ² statistic and Cochran’s Q . Using guidance from the Cochrane handbook, heterogeneity thresholds for the I ² statistic were classified as 0% to 40% (might not be important), 30% to 60% (moderate), 50% to 90% (substantial), and 75% to 100% (considerable). A two-sided P value of <.05 indicated statistical significance for all comparisons. All analyses were performed in R statistical software (version 4.5.2), and the primary statistical synthesis was conducted using the metafor package (version 4.8-0). ^,

A priori subgroup analyses were conducted only if at least two subgroups contained more than one study based on statin medication type, statin medication dose, duration of statin use, and method of outcome ascertainment. An a priori subgroup analysis using a mixed-effects meta-regression based on the overall risk of bias of each paper (low risk vs some concerns vs high risk vs very high risk) and a posthoc analysis for the original measures of effect reported by studies (OR vs RR vs HR) were also performed. Publication bias was assessed using funnel plots and visual inspection of symmetry.

In addition to the planned analyses, a posthoc exploratory synthesis was performed for continuous ocular outcomes among the studies identified in the initial literature search. Owing to the exploratory nature of this synthesis, we present the findings narratively.

RESULTS

The literature search identified 1704 studies, 955 of which were screened after removing duplicates. Forty-nine full-text studies were retrieved for further screening, and seven studies were included. ^,^,^,^,^,^, One study included in the review was not amenable to meta-analysis. An additional 293 studies were screened following deduplication from our reference list search of relevant reviews; 19 full-texts were reviewed, and none were included. One additional study from the initial search was initially included, but we were unable to obtain the required data despite contacting the authors several times. Result S1 presents a list of citations excluded during the full-text review phase. The PRISMA flow diagram for the study selection process is depicted in Figure 1 .

Of the six studies included in the meta-analysis, three (50.0%) were longitudinal cohorts, ^,^, two (33.3%) were cross-sectional, ^, and one (16.7%) employed a case-control design. Three studies (50.0%) were from Taiwan, ^,^, and one (16.7%) each from Australia, the Netherlands, and the US. These studies included 560,821 adult patients with a median study-level mean age of 54.0 years (range 50.4-74.0) and a median of 59.0% women (range 42.1% to 68.4%). To ascertain a diagnosis of DED, three (60.0%) studies used data from electronic medical records (EMRs), ^,^, two (40.0%) used symptom-based assessments, ^, and one used clinical assessment via the Schirmer test. The median duration of follow-up was 2 years (range 0.5-10). All studies used patients as the unit of analysis, and there were no significant baseline differences reported in any study between those who received or did not receive statins. Table 1 presents the characteristics of the included studies. Details about statin use, comorbidities, and comedication are presented in Table S8.

TABLE 1

Characteristics of the Included Studies.

Study	Country	Study Design	Dates	Sample Size	Statin Use (%)	Mean ± SD Age (Y)	Female (%)	Method of Outcome Ascertainment
Aldaas	USA	Case-control	2008-2018	39,336	11.3	NR	53.9	EMR (algorithm using diagnosis codes)
Chien	Taiwan	Case-control	2002-2016	87,045	54.9	54.0 ± 10.4	51.8	EMR (algorithm using diagnosis codes, prescriptions, and punctum occlusion procedures)
Ooi	Australia	Cohort	2002-2004	1680	NR	74.0 ± 9.3	59.0	Symptoms (moderate or severe dryness, discomfort, grittiness, itchiness, discomfort, or watering)
Wang	Taiwan	Cohort	2001-2018	83,862	50.0	52.9 ± 13.0	42.1	EMR (algorithm using diagnosis codes)
Wolpert	Netherlands	Cross-sectional	2014-2017	79,606	6.6	50.4 ± 12.6	59.2	Symptoms (Women’s Health Study dry eye questionnaire)
Wu	Taiwan	Cross-sectional	2016-2019	98	48.9	67.3 ± 10.3	68.4	Clinical assessment (Schirmer’s test)
Yu	USA	Cohort	2002-2019	308,530	84.6	NR	NR	EMR (algorithm using diagnosis codes and Restasis or Xiidra prescription)

EMR = electronic medical records; NR = not reported; SD = standard deviation.

Risk of bias of the included studies

Using the ROBINS-E tool, one study (14.3%) had a low risk of bias, two studies (28.6%) had some concerns, ^, and four studies (57.1%) were deemed to be at a high risk of bias. ^,^,^, The most frequent source of bias was insufficient control or adjustment of confounding. Other sources of bias arose from participant selection, missing data, and selection of the reported result. The full risk of bias ratings for each study are presented in Figure S1. Publication bias was not quantitatively assessed as fewer than 10 studies were included, although a funnel plot is shown in Figure S2. The GRADE summary of findings table is presented in Table 2 .

TABLE 2

Summary of Findings Table for the Association Between Statin Use and DED.

Association	No. of Participants (Studies)	Relative Effect Estimates (95% CI)	Prevalence of Statin Use	Absolute Effect Estimates (95% CI)		Certainty of the Evidence (Quality of Evidence)	Plain Language Summary
				Baseline	With Predictor
Statin use and DED	560,821 (6)	1.09 (1.05-1.13)	65.9%	138.3 per 1000	148.5 per 1000	Very low Due to limitations in observational study designs and serious inconsistency	The evidence is very uncertain about the effect of statin use on DED
Statin use and DED	560,821 (6)	1.09 (1.05-1.13)	65.9%	Difference: 10.2 more per 1000 (5.5 more to 15.2 more)

In our calculation of absolute effect estimates, we calculated the prevalence of statin use, excluding case-control studies that artificially selected for a predetermined proportion of statin users.

Meta-analysis results

Among six comparative studies, 356,012 (63.5%) participants used statins for cholesterol management while 203,129 (36.2%) did not. One study did not report the proportion of statin use in 1680 (0.3%) participants. There were 50,698 patients who had DED (9.1%) recorded among the five studies that reported event counts. Among the four studies providing event counts stratified by statin exposure, there were 15,485 patients with DED in the statin groups (4.4%) and 11,310 in the nonstatin groups. Figure 3 presents the forest plot for the pooled odds of DED among patients with and without statin use for cholesterol management. With very low-certainty evidence, the pooled OR of DED in patients using statins was 1.09 (95% CI 1.05-1.13, P <.001, I ² = 35.3%, P for heterogeneity =.036), with an absolute risk difference of 10.2 more per 1000 (95% CI 5.5 more to 15.2 more) ( Figure 2 ).

Subgroup and sensitivity analyses

To investigate the impact of study risk of bias, a subgroup analysis using a mixed-effects meta-regression was conducted. There was no evidence supporting a difference between risk of bias subgroups ( P for interaction =.123; Figure S3).

A subgroup analysis based on the method of outcome ascertainment used in the individual studies (EMR or clinical assessment vs symptom-based assessments) was also performed. Both subgroups suggested a positive association between statin use and DED, and there was no statistical evidence of a difference between subgroups ( P for interaction =.737; Figure S4). No other subgroup analyses were possible due to a paucity of data in the included studies.

A posthoc sensitivity analysis was conducted to investigate the impact of the initial measure of effect reported in individual studies (ie, OR vs RR), and no difference was detected between these groups ( P for interaction =.496; Figure S5).

The leave-one-out analysis demonstrated consistency between studies, with a preservation of the direction of association and statistical significance when each study was removed (Figures S6-S12). The resultant associations ranged from OR 1.06 (95% CI 1.03-1.10) when removing Chien to OR 1.10 (95% CI 1.04-1.16) when removing Wang.

Narrative synthesis

Several studies reported on the effects of various statin medications dosages and durations of use ( Table 3 ). Two studies did not appear to find any meaningful trends between increasing duration of statin use and DED, with all associations remaining statistically insignificant (91-180 days and >180 days vs ≤90 days OR = 1.05 [95% CI 0.97-1.14] and 0.97 [95% CI 0.90-1.05], respectively; second, third, and fourth quartiles of cumulative defined daily doses of statin use vs first quartile OR = 0.97 [95% CI 0.90-1.04], 0.99 [95% CI 0.91-1.06], 0.98 [95% CI 0.90-1.05]). ^, Two studies investigated different intensities of statin therapy, defined by the individual studies based on a combination of different statins and doses. One study compared different statin intensities to controls (low, moderate, and high vs no statins, OR = 1.39 [95% CI 1.13-1.72], 1.47[95% CI 1.30-1.65], and 1.46 [95% CI 1.21-1.75], respectively) and one study compared different statin intensities amongst each other (moderate and high vs low, OR = 1.00 [95% CI 0.92-1.08] and 0.92 [95% CI 0.72-1.19], respectively), and both did not appear to show meaningful differences between intensities. ^, One study comparing lipophilic to hydrophilic statins found no difference (OR 0.99, 95% CI 0.93-1.06). Another investigated specific statin medications, identifying a significant association between rosuvastatin use and DED (OR 1.22, 95% CI 1.05-1.42) but not for atorvastatin, pravastatin, or simvastatin.

TABLE 3

Summary of Findings in the Narrative Synthesis.

Study	Risk Factor	Comparator	Notes	N Analyzed	Odds Ratio (95% CI)
Duration/Quantity of Statin Use
Chien	Statin use 91-180 d	Statin use ≤90 d	Duration measured from initiation to DED onset	47,820	1.05 (0.97-1.14)
Chien	Statin use >180 d	Statin use ≤90 d	Duration measured from initiation to DED onset	47,820	0.97 (0.90-1.05)
Wang	Statin cDDD quartile 2	Statin cDDD quartile 1	Q1: <112 DDD; Q2: 112-351.5 DDD	83,862	0.97 (0.90-1.04)
Wang	Statin cDDD quartile 3	Statin cDDD quartile 1	Q3: 351.5-756 DDD	83,862	0.99 (0.91-1.06)
Wang	Statin cDDD quartile 4	Statin cDDD quartile 1	Q4: ≥756 DDD	83,862	0.98 (0.90-1.05)
Intensity of Statin Therapy
Aldaas	Low-intensity statin	No statin	Low: fluvastatin 20-40 mg QD, lovastatin 20 mg QD, pitavastatin 1 mg QD, pravastatin 10-20 mg QD, simvastatin 10 mg QD	39,336	1.39 (1.13-1.72)
Aldaas	Moderate-intensity statin	No statin	Moderate: atorvastatin 10-20 mg QD, fluvastatin 40 mg BID or 80 mg QD, lovastatin 40 mg QD, pitavastatin 2-4 mg QD, pravastatin 40-80 md QD, rosuvastatin 5-10 mg, simvastatin 20-40 mg QD	39,336	1.47 (1.30-1.65)
Aldaas	High-intensity statin	No statin	High: atorvastatin 40-80 mg QD, rosuvastatin 20-40 mg QD	39,336	1.46 (1.21-1.75)
Chien	Moderate-intensity statin	Low-intensity statin	Low: fluvastatin 20-40 mg QD, lovastatin 20 mg QD, pitavastatin 1 mg QD, pravastatin 10-20 mg QD, simvastatin 10 mg QD Moderate: atorvastatin 10-20 mg QD, fluvastatin 40 mg BID or 80 mg QD, lovastatin 40 mg QD, pitavastatin 2-4 mg QD, pravastatin 40-80 md QD, rosuvastatin 5-10 mg, simvastatin 20-40 mg QD	36,845	1.00 (0.92-1.08)
Chien	High-intensity statin	Low-intensity statin	High: atorvastatin 40-80 mg QD, rosuvastatin 20-40 mg QD	36,845	0.92 (0.72-1.19)
Lipophilicity of Statin
Chien	Lipophilic statin	Hydrophilic statin	Lipophilic: simvastatin, lovastatin, atorvastatin, fluvastatin, pitavastatin Hydrophilic: rosuvastatin, pravastatin Individuals on lipophilic and hydrophilic statins were excluded	47,820	0.99 (0.93-1.06)
Specific Statin Medications
Wolpert	Rosuvastatin	No rosuvastatin	NA	79,606	1.22 (1.05-1.42)
Wolpert	Atorvastatin	No atorvastatin	NA	79,606	1.06 (0.93-1.21)
Wolpert	Pravastatin	No pravastatin	NA	79,606	1.05 (0.82-1.35)
Wolpert	Simvastatin	No simvastatin	NA	79,606	1.00 (0.92-1.08)

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Tags: American Journal of Ophthalmology Volume 290

Sep 19, 2026 | Posted by drzezo in OPHTHALMOLOGY | Comments Off

Ento Key