The Effect of Psychotropic Medications on Glaucoma Risk and Intraocular Pressure: A Bayesian Meta-Analysis

Abstract

Background and Objectives

Glaucoma is one of the leading causes of irreversible blindness worldwide and is increasingly recognized as a potential adverse effect of various pharmacological agents. It has been suggested that psychotropic medications influence glaucoma risk, but findings across studies have remained inconsistent. We aimed to clarify the association between psychotropic drug use and glaucoma through a Bayesian meta-analysis.

Methods

We conducted a systematic literature search up to December 2024. Studies that examined the relationship between psychotropic medications and glaucoma or intraocular pressure (IOP), and reported odds ratios (ORs), relative risks (RRs), or mean differences, were eligible. Bayesian random-effects models were applied using informative priors based on existing evidence for specific compounds. Estimates were reported as pooled ORs and Hedges’ g with corresponding 95% credible intervals (CrIs).

Results

A total of 22 observational studies, including 293,228 users of psychotropic medications, met the inclusion criteria. Selective serotonin reuptake inhibitors (SSRIs) were associated with a modestly reduced risk of open-angle glaucoma (OR = 0.832, 95% CrI: 0.753–0.921) and a small but consistent reduction in IOP (Hedges’ g = −0.332, 95% CrI: −0.487 to −0.179). Although tricyclic antidepressants were expected to have a direct causative effect, results did not show a significant association with glaucoma risk (OR = 1.466, 95% CrI: 0.700–3.338). Benzodiazepines were associated with a significantly increased risk of glaucoma (OR = 1.550, 95% CrI: 1.436–1.674), with consistent effects across both short- and long-acting compounds. Topiramate demonstrated a strong association with acute angle-closure glaucoma (OR = 3.930, 95% CrI: 1.784–11.465), in accordance with its known mechanism of inducing anterior displacement of the lens-iris diaphragm. Studies on methylphenidate, limited to pediatric populations, suggested a modest but non-significant reduction in IOP compared with untreated individuals. Evidence on antipsychotics was inconsistent, precluding any quantitative synthesis.

Conclusions

While some drug classes (e.g., benzodiazepines, topiramate) show a strong association with glaucoma, results for other compounds must be taken judiciously. The high level of heterogeneity, and the presence of special populations suggest caution when moving to real-life scenarios. Nonetheless, our results highlight the importance of ophthalmologic monitoring in patients prescribed with psychiatric drugs (e.g., benzodiazepines or topiramate), at risk for angle closure.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40263-025-01249-6.

Key Points

Psychotropic medications have been anecdotally linked to the risk of developing glaucoma, a leading cause of blindness globally.

This analysis found that certain antidepressants, such as selective serotonin reuptake inhibitors (SSRIs), may lower the risk of open-angle glaucoma, while drugs like benzodiazepines and topiramate were linked to a higher risk of angle-closure glaucoma.

Patients prone to angle closure using psychotropic medications, especially those prescribed benzodiazepines and topiramate, should undergo ophthalmology monitoring due to potential glaucoma risk.

Introduction

Glaucoma, a leading cause of blindness globally, encompasses several conditions characterized by progressive degeneration of optic nerve fibers. Projections estimate that the number of affected individuals will reach 111.8 million by 2040 [1, 2]. Open-angle glaucoma (OAG) is the most prevalent type worldwide, particularly in Africa, with an overall prevalence of 2.4% expected to increase due to population ageing [2, 3]. OAG is defined as a progressive optic neuropathy associated with an open anterior chamber angle and risk factors including family history, elevated intraocular pressure (IOP), advanced age, male sex, myopia, and optic nerve hypoperfusion [4]. Angle-closure glaucoma (ACG), less common and more prevalent in Asia, results from mechanical obstruction of the anterior chamber angle, impeding aqueous humor outflow. The underlying mechanisms include pupillary block, angle crowding, plateau iris, lens abnormalities, aqueous misdirection, or choroidal effusion [5].

The association between medications and glaucoma risk has gained significant attention, particularly concerning psychotropic drugs. Psychotropic medications, including antidepressants, antipsychotics, anxiolytics, stimulants and mood stabilizers, are widely prescribed for the treatment of various mental health disorders, such as depression, anxiety, bipolar disorders, and schizophrenia [6, 7]. These medications can induce glaucoma via mechanisms such as pupillary dilation, ciliary body effusion, and aqueous humor dynamics dysregulation [8] (Fig. 1). A key mechanism is anticholinergic blockade, which interferes with parasympathetic control of the iris, causing mydriasis. In individuals with narrow angles, this can impede aqueous outflow, triggering acute angle closure [9–11]. Medications with high anticholinergic burden, including many antipsychotics and tricyclic antidepressants (TCAs), are particularly relevant, as are those acting via noradrenergic α1 stimulation or β2 receptor blockade (SNRIs) or type A gamma-aminobutyric acid (GABA-A) receptors allosteric modulators (benzodiazepines [BDZ] and Z-drugs) [8, 12–14]. Drugs inducing ciliary body effusion, like topiramate, can cause anterior rotation of the ciliary body, shallowing the anterior chamber, and triggering secondary angle closure [15, 16]. Conversely, serotonin 5-HT_1A and 5-HT_2A receptor stimulation in the ciliary body may reduce aqueous humor production, potentially explaining the observed ocular hypotensive properties of selective serotonin reuptake inhibitors (SSRIs) [17, 18].

Fig. 1.

This schematic illustrates three pharmacological pathways by which psychotropic medications may affect IOP and glaucoma risk. The first mechanism (top row) shows pupillary block and mydriasis, commonly induced by drugs with anticholinergic, adrenergic, or GABA-related properties (e.g., tricyclic antidepressants, SNRIs, BDZs), leading to impaired aqueous humor outflow and increased IOP. The second (middle row) highlights ciliary body swelling and anterior displacement of the lens-iris diaphragm, a mechanism associated with topiramate, also resulting in elevated IOP and risk of angle-closure glaucoma. The third (bottom row) shows increased aqueous humor outflow and reduced production, attributed to serotonergic antidepressants (e.g., SSRIs), resulting in a net decrease in IOP. BDZs benzodiazepines, GABA gamma-aminobutyric acid, 5-HT 5-hydroxytryptamine, IOP intraocular pressure, SNRIs serotonin-norepinephrine reuptake inhibitors, SSRIs serotonin selective reuptake inhibitors. Created in BioRender. Armentano, M. (2025) https://BioRender.com/zortckv

This study aims to quantitatively synthesize available evidence on the association between psychotropic medications (antidepressants, BDZs, mood stabilizers, antipsychotics, and stimulants), the risk of glaucoma, and IOP changes, utilizing a Bayesian meta-analytic approach. This allows for the integration of prior knowledge and explicitly accounts for between-study heterogeneity, providing probabilistic estimates of effect sizes and quantifying the certainty surrounding each estimate, guiding clinicians in balancing the psychiatric benefits of these medications against potential ocular risks [19–21].

Materials and Methods

Search Strategy and Registration

M.A., MD, F.G., MD, T.B.J., MD, and L.A., MD, conducted the last literature review on 15 December 2024. The review protocol was registered in PROSPERO (CRD42024605860). The present study was conducted following the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines.

Searches were conducted on PubMed, Scopus, and Web of Science covering the literature up to December 2024, using the combination of the following keywords:

(glaucoma OR intraocular pressure) AND

1. ("benzodiazepines" OR "Midazolam" OR "Triazolam" OR "Alprazolam" OR "Lorazepam" OR "Oxazepam" OR "Temazepam" OR "Lormetazepam" OR "Brotizolam" OR "Diazepam" OR "Clonazepam" OR "Chlordiazepoxide" OR "Flurazepam" OR "Clobazam" OR "Nitrazepam" OR "Prazepam" OR "Flunitrazepam" OR "Clorazepate" OR "Estazolam" OR "Ketazolam" OR "Mexazolam") OR ("Z-drugs" OR "Zolpidem" OR "Zaleplon" OR "Eszopiclone" OR "Zopiclone"), for benzodiazepines and Z-drugs

2. ("antipsychotic agents" OR "Antipsychotics" OR "Chlorpromazine" OR "Haloperidol" OR "Risperidone" OR "Olanzapine" OR "Clozapine" OR "Quetiapine" OR "Brexpiprazole" OR "Aripiprazole" OR "Paliperidone" OR "Lurasidone" OR "Ziprasidone" OR "Fluphenazine" OR "Thioridazine" OR "Perphenazine" OR "Trifluoperazine" OR "Sulpiride" OR "Amisulpride" OR "Pimozide"), for antipsychotics

3. ("antidepressive agents" OR "Antidepressants" OR "Fluoxetine" OR "Sertraline" OR "Paroxetine" OR "Citalopram" OR "Escitalopram" OR "Fluvoxamine" OR "Venlafaxine" OR "Duloxetine" OR "Desvenlafaxine" OR "Bupropion" OR "Amitriptyline" OR "Nortriptyline" OR "Imipramine" OR "Clomipramine" OR "Doxepin" OR "Mirtazapine" OR "Trazodone" OR "Vilazodone" OR "Vortioxetine" OR "Phenelzine" OR "Tranylcypromine" OR "Isocarboxazid" OR "Selegiline" OR "Moclobemide") for antidepressants

4. ("mood stabilizers" OR "Lithium" OR "Valproic Acid" OR "Carbamazepine" OR "Lamotrigine" OR "Oxcarbazepine" OR "Gabapentin" OR "Topiramate" OR "Lamotrigine" OR "Pregabalin"), for mood stabilizers

5. ("Central Nervous System Stimulants" OR "Stimulants" OR "Methylphenidate" OR "Amphetamine" OR "Dextroamphetamine" OR "Lisdexamfetamine" OR "Modafinil" OR "Armodafinil" OR "Dexmethylphenidate"), for stimulants.

Eligibility Criteria and Data Collection

This meta-analysis included studies that evaluated the association between psychotropic medications and glaucoma risk, as well as their potential impact on IOP. Observational studies (e.g., cohort, case–control, and cross-sectional designs) and randomized controlled trials (RCTs) were considered to be eligible. Studies were included if they: (1) reported quantitative data on psychotropic medications (e.g., antidepressants, antipsychotics, BDZs, mood stabilizers, or stimulants) and glaucoma or IOP; (2) provided effect size measures such as odds ratios (ORs), relative risks (RRs), or mean differences, along with corresponding confidence intervals (CIs) or standard errors; and (3) were published in peer-reviewed journals in English. Articles were excluded if they lacked quantitative relevant data (narrative reviews/case reports/meeting abstracts), focused on animal or in vitro studies, or included participants with pre-existing glaucoma unrelated to psychotropic medication use. The search strategy excluded studies that did not involve humans and were not written in the English language (see Supplementary Table 1).

The reviewers (MA, FG, TBJ, and LA) independently screened and selected the articles employing Rayyan [22], to facilitate the blinded screening process.

Data Extraction and Risk of Bias Assessment

Data extraction from the included studies was conducted by four authors, LA, MA, FG, and TBJ, using Microsoft Excel for organized data management. The process involved systematically entering information into an Excel spreadsheet, including authors, year of publication, study design, country, mean age, main effect size measure with its relative standard deviation, pharmaceutical compound, and drug category. Efforts were made to consider all available studies, including direct contact with authors when data were missing. The authors did not come across unpublished articles or abstracts relevant to the study, so they were not included. The risk of bias was evaluated in line with the guidelines from the Cochrane Handbook for Systematic Reviews [23]. Specifically, for non-randomized studies, the Risk of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool was applied [24]. MA and LA independently rated the bias as low, moderate, serious, critical, and no information. Following any disagreements, a third author (TBJ) was consulted to resolve the issue. The risk of bias was visually represented using the Robvis tool [25]. The tool provided an overall serious risk of bias if at least one of the seven assessed domains was at serious risk.

Outcome Measures

The primary outcome measures included estimates of statistical correlation (such as ORs and RRs) for the association between BDZs, antidepressants, antipsychotics, mood stabilizers, and stimulants with glaucoma. In line with established methods for Bayesian meta-analysis, we synthesized these measures under the assumption that, given the low prevalence of glaucoma in the general population (typically <2–5%) [2], ORs and RRs converge to similar values and can thus be pooled [26]. To this end, reported RRs were converted to ORs using the following formula, which corrects for baseline risk in the unexposed population:

$$\text{OR}=\frac{\text{RR}\times (1-\text{P}0)}{1-(\text{P}0\times \text{RR})}$$

where P0 represents the baseline prevalence of glaucoma in the non-exposed group [26]. This transformation allowed for consistent modeling across studies reporting different effect measures.

Furthermore, the Bayesian framework is particularly well-suited for combining different effect measures, as it is more tolerant of between-study differences in study design, population characteristics, and outcome definitions. By explicitly incorporating prior information and modeling heterogeneity as a probability distribution rather than a fixed parameter, the Bayesian approach allows for a more flexible and probabilistic synthesis of the available evidence, even when data sources are diverse and study-level uncertainty varies widely [19–21].

Given the methodological differences between included studies, Hedges’ g was calculated separately for cross-sectional and prospective designs, ensuring appropriate standardization across study types. In cross-sectional studies, Hedges’ g reflected differences in IOP between independent groups (e.g., long-term users, short-term users, and non-users). In prospective studies, Hedges’ g quantified pre-post changes in IOP within the same individuals before and after drug exposure. Since within-subject designs generally exhibit lower variance due to the correlation between repeated measures, we adjusted the standard deviation (SD) for prospective studies by applying a correction factor incorporating an assumed intra-subject correlation of r = 0.5, as suggested in prior meta-analyses of repeated-measures designs [27]. This correction accounted for reduced variability in within-subject comparisons.

Data Synthesis and Analysis

All statistical analyses were conducted using R (version 4.4.1, R Foundation for Statistical Computing, Vienna, Austria). Data wrangling was performed using tidyverse, version 2.0 [28]. Bayesian meta-analyses were performed using the bayesmeta package, version 3.4, which incorporates hierarchical models to estimate pooled effects, heterogeneity (τ²), and credible intervals for effect sizes [29]. We calculated pooled estimates as mean effects (μ) with 95% credible intervals (CrIs) and evaluated heterogeneity using the posterior median of the heterogeneity parameter (τ) and its credible interval. The degree of heterogeneity was further quantified using the Bayesian I² statistic. The Bayesian framework allows for the contrasting of evidence supporting the null hypothesis and an alternative hypothesis based on the actual data [30]. Unlike the frequentist approach, which only allows for the rejection or non-rejection of a given hypothesis, the Bayesian approach provides evidence in favor of one hypothesis over another [19]. This feature is particularly advantageous for nuanced inference. Moreover, the Bayesian method mitigates variability underestimation, a limitation inherent to the frequentist approach, which treats the point estimate of heterogeneity variance as a fixed quantity [20]. Instead, the Bayesian framework incorporates uncertainty in heterogeneity by modeling it as a random variable with a prior distribution.

To determine a suitable prior for the heterogeneity parameter (τ), we assumed a half-normal distribution. Given the potential variability introduced by differences in study designs—including both case–control and case-crossover studies—we set a more permissive scale parameter of 2, rather than the conventional value of 1, to accommodate the broader uncertainty in heterogeneity [31, 32]. This adjustment was made to prevent underestimation of between-study variance, a risk when pooling effect estimates from methodologically diverse studies, and was guided by prior research on heterogeneity in meta-analyses of observational data [33]. For the effect size parameter (μ), we used a normal prior informed by the findings of Wang et al, setting the mean and SD to the pooled estimates reported in their meta-analysis for SSRIs when these values were available [34]. In cases where prior data were not available and literature was not well-established, we adopted a weakly informative prior with a mean of 1 (if the outcome variable was an OR) or 0 (if the outcome variable was continuous) and a SD of 4, reflecting a high degree of uncertainty about the effect size.

However, for BDZs and topiramate, whose effects on the ACG are more consolidated, we applied informative priors to account for the well-documented association between these drug classes and ACG. Specifically, we set a prior μ of 2 and a prior SD of 0.8 to reflect both the clinical plausibility of a substantial effect and the uncertainty surrounding the exact magnitude. This approach was justified by consistent case reports, pharmacological plausibility [9–11, 35–37], and the precautionary guidance issued by the American Academy of Ophthalmology (https://www.aao.org/eyenet/article/medication-induced-acute-angle-closure-glaucoma).

The 95% credible intervals were used to make inferences about the parameters of interest. These intervals provided a direct probability statement regarding the plausible range of the effect size (μ) and heterogeneity (τ). For example, a 95% CrI for μ indicates that there is a 95% probability that the true effect size lies within the specified range, given the data and the priors. This approach not only accounts for the uncertainty in parameter estimation but also allows for robust conclusions, especially in datasets with high heterogeneity or small sample sizes.

Sensitivity analysis using the “one study removed at a time” technique was performed to test whether a potential outlier within included studies could have influenced the results of the meta-analysis [38]. Additionally, we assessed the impact of studies employing different methodological designs by running sensitivity analyses, allowing us to determine whether these studies disproportionately influenced the overall pooled effect. Lastly, we performed subgroup analyses whenever data from at least two independent datasets were available. Subgroup analyses were conducted to determine whether results would differ as a function of the type of glaucoma (i.e., OAG or ACG), according to drug class (i.e., SSRIs or SNRIs), or according to the compound half-life (e.g., short-acting or long-acting molecules). Regarding BDZs, half-lives were determined according to reference textbooks in psychopharmacology [39]. This distinction was made assuming that the duration of action in the human body might influence the probability of presenting ocular side effects, with longer half-life molecules possibly leading to more sustained effects on intraocular pressure or angle anatomy.

Results

Study Selection

The study selection process is detailed in the PRISMA flow chart in Fig. 2 [40]. A comprehensive literature search across electronic databases initially identified 3034 articles. Following the application of eligibility criteria and full-text review of 93 articles, 22 observational studies were ultimately included in the final analysis (see Supplementary Table 1 for the excluded studies). The key characteristics of these studies are summarized in Table 1.

Fig. 2.

Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) diagram including the detailed report of the excluded studies and the reason behind the exclusion

Table 1.

Characteristics of the included studies classified by medication category and target outcome

Author (year)	Country	Design	Medication	Comparison	Numerosity	Mean age (± SD)	Female %	Target outcome
Antidepressants
Qiao et al. (2024) [64]	Canada	Case–control study	Bupropion	Users	6	71.9 ± 14.2	65.4	ACG
				Nonusers	7759
Ik Na et al. (2022) [65]	Korea	Case-crossover study	SSRIs/SNRIs/TCAs	Amitriptyline users	421	68.0 ± 9.6	81.0	ACG
				Duloxetine users	193	69.9 ± 7.6	94.2
				Escitalopram users	171	68.7 ± 8.3	90.9
				Nortriptyline users	149	68.3 ±7.9	85.7
				Nonusers	443	n/a	n/a
Özer et al. (2022) [66]	Turkey	Prospective study	SNRIs	Duloxetine users	38	48.5 ± 13.9	47.36	IOP
Karaküçük et al. (2019) [67]	Turkey	Prospective study	SSRIs	Sertraline users	10	24.3 ± 7.4	87.1	IOP
				Escitalopram users	11
				Fluoxetine users	10
Gündüz et al. (2018) [68]	Turkey	Case–control study	SSRIs	Users > 6 months	53	37 ± 9.7	73.6	IOP
				Users < 6 months	58	35.6 ± 8.6	75.9
				Nonusers	55	35.1 ± 10.2	81.8
Zheng et al. (2018) [69]	Taiwan	Case–control study	SSRIs/SNRIs/TCAs/Bupropion	SSRI users	7163	72 ± 10.8	52	OAG
				SNRI users	2507
				TCA users	2679
				Bupropion users	1651
				Nonusers	22780
Gündüz et al. (2018) [14]	Turkey	Case–control study	SNRIs	Users > 6 months	19	40.0	21.1	IOP
				Users < 6 months	22	36.5	18.2
				Nonusers	44	39.0	15.9
Chen et al. (2017) [57]	Taiwan	Case–control study	SSRIs	Users	4238	n/a	53.5	OAG/ACG
				Nonusers	88641	n/a
Chen et al. (2016) [58]	Taiwan	Case–control study	SSRIs	Immediate users	15	61.8 ± 13.9	58.1	ACG
				Long users	29
				Nonusers	7133
Masís et al. (2016) [70]	USA	Case–control study	SSRIs/Bupropion/TCAs	SSRI users	482	56.5 ± 1.3	53.0	Glaucoma
				Bupropion users	116
				TCA users	32
				Nonusers	6130
Chen et al. (2015) [71]	Taiwan	Case–control study	SSRIs	Users	13093	49.3 ± 16.2	65.0	OAG/ACG
				Nonusers	13093	49.4 ± 16.5	64.7
Stein et al. (2015) [43]	USA	Case–control study	SSRIs/Bupropion/TCAs	SSRI users	142765	58.9 ± 12.0	59.5	OAG
				Bupropion users	45337
				TCA users	50312
				Nonusers	386304
Symes et al (2015) [72]	Canada	Case–control study	Bupropion	User	331	54.6 ± 8.3	69.2	ACG
				Nonuser	16763	54.6 ± 8.3	56.3
Seitz et al. (2012) [73]	Canada	Case-crossover study	SSRIs	Users	82	74.3 ± n/a	66	ACG
				Nonusers	6388
Benzodiazepines and Z-drugs
Park et al. (2019) [55]	Korea	Case–control study	BDZ	Short-acting users	57	71.8 ± 5.7	77.4	ACG
				Long-acting users	74
				Nonusers	4809
Ho et al. (2015) [74]	Taiwan	Case–control study	Zolpidem	Users	952	40.5 ± 20.5	42.8	Glaucoma
				Nonusers	43538
Kim et al. (2020) [44]	Korea	Case-crossover study	BDZ	Short-acting users	4951	n/a	31.7	ACG
				Long-acting users	4674
Ik Na et al. (2022) [65]	Korea	Case-crossover study	BDZ	Short-acting users (alprazolam)	858	68.0 ± 9.4	85.4	ACG
				Long-acting users (diazepam)	1259	69.4 ± 9.6	80.9
				Nonusers	1383	n/a	n/a
Mood stabilizers
Qiao et al. (2024) [64]	Canada	Case–control study	Topiramate	Users	26	71.9 ± 14.2	65.4	ACG
				Nonusers	7739
Ik Na et al. (2022) [65]	Korea	Case-crossover study	Topiramate	Users	51	50.8 ± 9.3	81.8	ACG
				Nonusers	10	n/a
Symes et al. (2015) [72]	Canada	Case–control study	Topiramate	Users	96	54.6 ± 8.3	69.2	ACG
				Nonusers	16998		56.3
Ho et al. (2013) [75]	Taiwan	Retrospective cohort study	Topiramate	Users	1956	44.1	58.2	Glaucoma
				Nonuser	15648
Etminan et al. (2012) [45]	Canada	Case–control study	Topiramate	Current users	1533	65.8 ± 15.6	57.9	Glaucoma
				First-time users	181
				Prevalent current	1352
				Past users	3137
				Nonusers	1064914
Stimulants
Larranaga-Fragoso et al. (2015) [76]	Spain	Prospective cohort study	Methylphenidate	Users	14	11 ± 2.79	47	IOP
Guvenmez et al. (2020) [77]	Turkey	Prospective cohort study	Methylphenidate	Users	42	10.1 ± 1.8	33.3	IOP
				Nonuser	36	10.2 ± 1.6	36.2
Bingöl Kiziltunç et al. (2022) [78]	Turkey	Prospective cohort study	Methylphenidate	Users	22	7.6	18.2	IOP

Each study may appear more than once as some studies addressed multiple molecular classes

ACG angle-closure glaucoma, BDZ benzodiazepines, IOP intraocular pressure, OAG open-angle glaucoma, SD standard deviation, SNRI serotonin-norepinephrine reuptake inhibitors, SSRI selective serotonin reuptake inhibitors, TCAs tricyclic antidepressants

Quality of the Included Studies

The risk of bias assessment, as shown in the ROBINS-I diagram (Supplementary Fig. 1), indicated that most studies were classified as having a moderate risk of bias. While some concerns were noted, these primarily reflected challenges inherent to observational study designs, such as potential confounding and the selection of participants. In general, interventions and outcomes were well documented, although occasional instances of missing data or unclear reporting introduced some uncertainties. For example, in a few studies, the reporting of outcomes lacked granularity or omitted important details, such as follow-up timing or secondary outcomes. Despite these limitations, the studies were methodologically sound in key areas, including classifying interventions and measuring outcomes, where the risk of bias was consistently low. Overall, the included studies demonstrated a solid level of methodological rigor, and the findings can be interpreted with moderate confidence while accounting for the noted limitations.

Characteristics of the Included Studies

The final analysis included 22 articles on antidepressants, BDZs, Z-drugs, mood stabilizers, and stimulants, with a total of 293,228 drug users (Table 1). However, no articles regarding antipsychotics were part of this selection due to inconsistency in outcomes and endpoints. A total of 14 articles focused on antidepressants, including SSRIs, SNRIs, TCAs, and bupropion, were covered in the final meta-analysis. These articles comprehended a total of 271,993 antidepressant users and 555,533 nonusers. Case–control studies were the majority (10), followed by case-crossover studies (2) and prospective studies (2); no RCT was available for this drug class. For BDZs (long- and short-acting) and Z-drugs, four studies were included with a total of 12,825 users and 49,730 nonusers. Studies were case–controls (2) and case-crossover (2). Among mood stabilizers, only studies involving topiramate were included in the meta-analysis, due to the paucity of eligible studies for the other compounds in this class (e.g., lithium, gabapentin, pregabalin, etc.). The number of users was 8332, while nonusers were 1,105,309. A total of 5 studies were included: case–control studies (3), case-crossover studies (1), and a retrospective cohort study (1). Lastly, stimulants included only three studies on methylphenidate in the pediatric population. There were 78 users, and 36 nonusers. All studies were prospective cohort studies. Overall, the geographic distribution of the included studies was uneven, with a major prominence of USA/Canadian, Turkish, and Eastern Asian countries. Only one small study was conducted in Europe (Spain), and no available study was conducted in third-world countries.

Antidepressants

SSRIs and SNRIs

The Bayesian random-effects meta-analysis investigating the association between SSRIs and SNRIs and the risk of OAG and ACG yielded a pooled OR of 1.065 (95% CrI: 0.840–1.350), indicating a high degree of uncertainty surrounding the association between antidepressant use and glaucoma risk (Fig. 3a). This high heterogeneity was confirmed by the Bayesian I² of 99.6%, indicating that nearly all variability was due to differences between studies rather than random sampling error.

Fig. 3.

Forest plot showing the association between a antidepressants and glaucoma risk, and b serotonin selective reuptake inhibitors (SSRIs) and open-angle glaucoma (OAG). In particular, panel b highlights the putative protective role of selective serotonergic antidepressants in the development of OAG. Estimates are expressed as odds ratios. 95% CrI 95% credible intervals

When performing subgroup analyses, consistently with previous evidence [36], a significantly lower risk of development of OAG (Fig. 3b) in participants receiving SSRIs treatment compared to those not exposed to SSRIs was observed (k = 4, OR = 0.832, 95% CrI: 0.753–0.921, I² = 91.3%). However, the same class of drugs did not show a significant effect on the risk of ACG (k = 4, OR = 1.529, 95% CrI: 0.590–4.009, I² = 94.9%; Supplementary Fig. 2). Sub-group analysis for SNRI drugs was not possible, as only two studies on two different conditions (OAG and ACG) were found in the literature.

The Bayesian random-effects meta-analysis evaluating the impact of serotonergic antidepressants (SSRIs and SNRIs) on IOP suggested a slight reduction in pressure (k = 14, Hedges' g = −0.332, 95% CrI: −0.487 to −0.179). Between-study heterogeneity was low (I² = 31.3%), indicating reasonably consistent findings across studies (Supplementary Fig. 3). When considering single classes, SSRI showed a significant hypotensive effect (k = 11, Hedges’ g = −0.384, 95% CrI: −0.619 to −0.094), with a relatively low heterogeneity (I² = 19.5%), whereas SNRI did not reach the significance threshold (k =3, Hedges’ g = −0.143, 95% CrI: −0.600 to 0.329, I² = 49.6%). See Supplementary Figs 4a–b.

TCAs

The analysis investigating the association between TCAs and glaucoma did not show significant results and reported a high level of heterogeneity (k = 5, OR = 1.466, 95% CrI: 0.700–3.338, I² = 98.2%; Supplementary Fig. 5).

Bupropion

The analysis assessing the relationship between bupropion and glaucoma showed a trend toward reduced risk, but credible intervals crossed the null (k = 6, OR = 0.757, 95% CrI: 0.506–1.011), with low to moderate heterogeneity (I² = 55.6%; Supplementary Fig. 6).

Benzodiazepines and Z-Drugs

The Bayesian meta-analysis on overall BDZs and glaucoma risk showed a pooled OR of 1.550 (k = 3, 95% CrI: 1.436–1.674), suggesting a moderately increased risk associated with BDZ use. Between-study heterogeneity was low (I² = 20.8%), indicating consistent findings across studies (Fig. 4a).

Fig. 4.

Forest plot shows the association between a benzodiazepines and glaucoma risk, and b topiramate and glaucoma risk. Both pooled analyses show a significant effect of these molecules in increasing the risk of glaucoma. Estimates are expressed as odds ratios. 95% CrIs 95% credible intervals

When considering only molecules with short half-lives, the effect remained significant with OR = 1.647 (k = 3, 95% CrI: 1.887–2.658), although in the face of a moderate degree of heterogeneity (I² = 63.4%; Supplementary Fig. 7a).

Similarly, the effect of compounds with longer clearance times was deemed to be significant, with an OR = 1.538 (k = 3, 95% CrI: 1.141–2.488). Alongside short-acting BDZs, longer half-life molecules presented a moderate degree of heterogeneity (I² = 63.9%; Supplementary Fig. 7b).

Mood Stabilizers

Topiramate

The random-effect model showed a significant overall effect of topiramate on glaucoma, regardless of the glaucoma type, although the level of heterogeneity was high (k = 10, OR = 1.768, 95% CrI: 1.213–2.769, I² = 94.8%; Fig. 4b). When performing the subgroup analysis, topiramate reported a significant effect on the risk of ACG (k = 2, OR = 3.930, 95% CrI: 1.784–11.465, I² = 70.6%; Supplementary Fig. 8).

Stimulants

Methylphenidate

Studies on methylphenidate were conducted on a population of patients aged < 18 years. The pooled outcome measure highlighted a very modest reduction of IOP in patients treated with methylphenidate compared with untreated individuals, although statistically non-significant (k = 8, Hedges' g = −0.162, 95% CrI: −0.407–0.084, I² = 12.3%; Supplementary Fig. 9).

Sensitivity Analysis

The removal of included studies in each class of pharmacological compound, one at a time, did not change the direction or significance level of this Bayesian meta-analysis.

Discussion

Main Results

This Bayesian meta-analysis is the first to comprehensively review the association between psychotropic medications and glaucoma risk, as well as their potential modulatory effects on IOP. We integrated evidence from multiple observational studies, with the largest number of studies pertaining to antidepressant compounds. Results obtained from the analysis of available data on SSRIs and SNRIs yielded similar results to those of a previously published frequentist work [34]. Interestingly, patients receiving SSRIs were confirmed to show a significantly lower risk of developing OAG compared to controls (Fig. 3b). This is further supported by evidence indicating that SSRIs lower IOP (Supplementary Figs 3 and 4a–b). Through the binding of 5-HT_1A receptors, these drugs can indeed regulate aqueous dynamics, reducing its production and eventually leading to an overall reduction in the IOP [41] (Fig. 1). In contrast to anecdotal evidence, SSRIs did not show a significant influence on the development of ACG, despite their mild mydriatic effects (Supplementary Fig. 2) [16].

Tricyclic antidepressants (TCAs) have been historically linked to glaucoma onset, especially when used in at-risk populations [10, 11]. Although the pooled analysis showed no significant results, most of the glaucoma guidelines point toward the induction effect of these compounds on ACG due to their anticholinergic activity [42]. Therefore, caution should be advised when handling these compounds among patients with an increased risk of angle closure, such as in older adults.

Similar to TCAs, bupropion showed non-significant results, with patients on this drug reporting a trend toward reduced glaucoma risk. Although data were insufficient to allow subtype-specific analyses, this effect might be attributed to its anti-tumor necrosis factor-alpha activity that prevents ganglion cells' apoptosis [43].

Benzodiazepines and Z-drugs demonstrated an increased risk of overall glaucoma, with an OR of 1.550 (95% CrI: 1.436–1.674; Fig. 4a and Supplementary Figs 7a–b). This effect was still significant when dividing compounds depending on their mean half-lives, thus suggesting a class-wide effect. Furthermore, most of the included studies focused on ACG, which corroborates the idea that these drugs act through a GABA-A-mediated pupillary block, which impedes aqueous outflow [44]. Contrary to what we expected, short-acting BDZ showed a slightly higher OR (1.64) compared to long-acting BDZ (1.53). These results were characterized by the lowest level of heterogeneity (I² = 20.8%), which further endorses the possible effect of these drugs on the pathogenesis of secondary ACG. Hence, given the widespread use of BDZ, our analyses suggest careful ocular monitoring in those predisposed to acute ocular events.

The use of topiramate was associated with a higher risk of glaucoma, likely due to ciliochoroidal effusion and anterior displacement of the ciliary body, which resulted in a forward shift of the iridolenticular diaphragm [45]. Our results support this hypothesis, particularly highlighting the strong association between the use of topiramate and the onset of ACG (Supplementary Fig. 8). Given the clinical implications of this relationship and due to the role of this drug in glaucoma pathophysiology, high-risk patients on topiramate should be carefully monitored by an ophthalmologist.

Stimulant drugs are commonly linked to an increased risk of both OAG and ACG development [46]. These molecules are known to inhibit the reuptake of norepinephrine and dopamine. While the precise mechanism behind glaucoma induction remains unclear, it is hypothesized that the enhancement of noradrenergic activity through the stimulation of α1 receptor can induce mydriasis. However, our pooled analysis showed that methylphenidate minimally and non-significantly reduced IOP, potentially due to α2-adrenergic effects (Supplementary Fig. 9). Nonetheless, the studies included in this meta-analysis focused solely on pediatric populations and lacked sufficient statistical power, thus preventing meaningful conclusions regarding methylphenidate’s ocular effects.

Health-Economic Considerations and Clinical Recommendations

Population-wide screening for glaucoma in asymptomatic adults does not appear clearly justified on current evidence and guidance, with the US Preventive Services Task Force maintaining an “I statement” and National Institute for Health and Care Excellence favoring case-finding pathways over universal programs [47, 48]. From a health-economic perspective, most evaluations in high-income settings have not demonstrated favorable value for general population screening at typical prevalence, although selective pathways, such as tele-glaucoma or AI-assisted strategies in specific age-risk strata, may be reasonable in defined contexts [49–51]. Coverage and patient costs also vary substantially across jurisdictions (for example, US Medicare limits reimbursed screening to individuals deemed “high risk”), suggesting that, where feasible, opportunistic IOP checks could be embedded within routine eye care for those at higher prior probability rather than implemented as stand-alone tests [52]. Against this background, a cautious, risk-based approach may be considered: prioritizing individuals with recognized susceptibility to primary angle-closure disease (older age, female sex, Asian ancestry, hyperopia/short axial length, crowded anterior segments) for targeted assessment rather than pursuing mass screening [53]. With respect to psychotropic exposure, pragmatic monitoring could be tailored by class. For instance, topiramate-associated angle closure is typically precipitated within 2 weeks by initiation or dose escalation. Because the mechanism is non-pupillary-block, laser peripheral iridotomy (LPI) is not effective. These results may indicate the need for structured counselling about symptoms plus an early ophthalmic review (IOP and angle assessment) within 1–2 weeks of treatment start, or dose increase in at-risk patients. Acute management concerns drug cessation, cycloplegia, topical steroids, and IOP-lowering therapy [54]. In cases of BDZ- and Z-drug-related glaucoma: population studies suggest an early period of increased ACG risk after initiation, especially in older adults. A pragmatic approach could be to triage anatomic risk pre-prescription (history of narrow angles/plateau iris, hyperopia, prior angle symptoms), and to perform IOP and angle review at 1–2 weeks in high-risk individuals—or sooner if symptoms occur [44, 55]. Concerning antidepressants, long-term SSRI exposure has not been consistently linked with elevated primary open-angle glaucoma risk and may correlate with slightly lower IOP at the population level; yet short-term ACG around initiation has been described in susceptible eyes, so additional monitoring could be reserved for patients with known narrow angles or early symptoms. In case of TCA and SNRI drug regimens, a cautious selective angle evaluation and counselling may be appropriate in anatomically predisposed eyes [56–58].

Overall, these suggestions are intended as context-sensitive guidance aimed at aligning clinical vigilance with resource stewardship rather than prescriptive directives, and should be adapted to local service capacity, payer policies, and individual patient risk.

Limitations and Future Research

Several limitations of this study should be considered when interpreting its findings. First, the included studies employed diverse designs, with a few reporting the RR rather than the OR. While we assumed RR ≈ OR due to the low prevalence of glaucoma in the general population, this approximation introduces a degree of imprecision. Furthermore, heterogeneity exists within drug classifications; for example, the BDZ meta-analysis included Z-drugs like zolpidem, which may have slightly different pharmacological profiles. Second, although SSRIs appeared to exhibit a protective effect for OAG, our pooled analysis did not consider single molecules. Future research should investigate specific SSRIs with known anticholinergic activity, such as paroxetine. These agents are hypothesized to have implications for ACG, based on the established association between psychotropics with anticholinergic/sympathomimetic properties and narrower anterior chamber angles [59]. Although in our analysis, SNRIs and SSRIs were not significantly associated with the development of ACG, case reports have documented that SSRIs may precipitate ACG in susceptible individuals [60, 61]. This highlights the importance of cautious interpretation of our results, particularly in populations with a higher prevalence of ACG, such as Asians and South Asians, and the need for future prospective studies to clarify this risk [62]. Moreover, some studies on SSRIs and SNRIs, including the smaller Turkish cohorts, involved younger patients, which may limit the generalizability of the findings to the older populations in which glaucoma is most prevalent [63].

Third, several studies also relied on administrative databases, which may have suffered from misclassification or incomplete outcome data. This has led to the impossibility of conducting a meta-regression analysis, further limiting our ability to explore the potential sources of variability.

Fourth, some subgroup analyses were based on only two studies, reducing the statistical power and generalizability of those specific findings. Finally, interpretation is further limited by the lack of large prospective studies in populations requiring psychotropic medications. While case reports and pharmacovigilance databases such as FDA’s Adverse Events Reporting System (FAERS) and EUDRAVigilance contain valuable information on adverse events, these sources would require different study designs to analyze and interpret reliably. Future research should prioritize prospective designs, examine specific compounds within each drug class, incorporate dose and treatment duration, and employ meta-regression to explore heterogeneity and potential confounding factors.

Conclusions

This Bayesian meta-analysis provides valuable insights into the relationship between psychotropic medications and glaucoma risk, addressing an important gap in the current literature. Our findings suggest that SSRIs may provide a protective effect against OAG, potentially through modulation of aqueous humor production. In contrast, BDZs and topiramate appear to be significantly associated with an increased risk of ACG, likely due to their effects on pupillary dynamics and aqueous humor outflow. This research underscores the importance of considering potential ocular side effects when prescribing psychotropic medications and highlights the need for interdisciplinary collaboration in patient care. Given the widespread use of these medications, clinicians should remain vigilant about their potential impact on glaucoma risk, particularly in patients with anatomical predisposing factors.

Considering our findings and the implementation literature, we do not endorse mass screening for glaucoma in psychotropic-treated adults. Instead, we propose a risk-stratified strategy that aligns with guidelines and economic evidence. Regulatory monographs already acknowledge angle-closure risks with several psychotropics. For topiramate, US and other labels carry a dedicated warning on “acute myopia and secondary angle-closure glaucoma”, noting typical onset within the first month and advising prompt drug discontinuation and ophthalmic management. Frequency is not quantified in labeling. For SSRIs (e.g., sertraline) and SNRIs (e.g., venlafaxine), labels caution that drug-induced mydriasis may precipitate angle-closure in anatomically narrow angles; again, frequency of angle-closure per se is not quantified. Similarly, BDZs (e.g., diazepam) list acute narrow-angle glaucoma as a contraindication in US labeling (use is permissible in treated open-angle glaucoma). Taken together, current labeling supports a cautious, risk-based stance while acknowledging that formal frequency estimates for glaucoma in general are largely unavailable in regulatory documents and that real-world incidence remains uncertain. Future studies should focus on prospective research designs, incorporating detailed assessments of medication dosage, duration, and individual patient susceptibility factors to further elucidate the observed associations.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Open access funding provided by Università degli Studi di Roma La Sapienza within the CRUI-CARE Agreement. No funding was received. The open access fee was sponsored by Sapienza University of Rome.

Declarations

Conflict of interest

The authors have no competing interests to disclose.

Availability of data and material

The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Code availability

Not applicable.

Author contributions

T.B.J., L.A., M.A., and F.G. conceived the study and performed the literature search and data extraction. G.D.L. and C.N. contributed to the statistical analysis and interpretation of data. A.S. provided critical revisions and expert guidance throughout the project. All authors (L.A., M.A., F.G., T.B.J., G.D.L., C.N., and A.S.) contributed to drafting or revising the manuscript for important intellectual content. All authors have read and approved the final version of the manuscript and agree to be accountable for all aspects of the work.