Association of topical glaucoma medications with lacrimal drainage obstruction and eyelid malposition

Abstract

Background/objectives

Adverse effects of topical glaucoma medications (TGMs) may include development of ocular adnexal disorders. We undertook a study to determine the effect of TGMs on the risk of developing lacrimal drainage obstruction (LDO) and eyelid malposition.

Subjects/methods

All patients 66 years of age and older in Ontario, Canada initiating TGM and all patients diagnosed with glaucoma/suspected glaucoma but not receiving TGM from 2002 to 2018 were eligible for inclusion in this retrospective cohort study. Using validated healthcare administrative databases, cohorts were identified with TGM and no TGM patients matched 1:2 on sex and birth year. The effect of TGM treatment on risk of surgery for LDO and lid malpositions was estimated using Kaplan–Meier and Cox proportional hazards models.

Results

Cohorts included 122,582 patients in the TGM cohort and 232,336 patients in the no TGM cohort. Among the TGM cohort there was decreased event-free survival for entropion (log-rank P < 0.001), trichiasis (P < 0.001), and LDO (P = 0.006), and increased ectropion-free survival (P = 0.007). No difference in ptosis-free survival was detected (P = 0.78). For the TGM cohort there were increased hazards for entropion (hazard ratio [HR] 1.24, 95% confidence interval [CI] 1.12–1.37; P < 0.001), trichiasis (HR 1.74, 95% CI 1.57–1.94; P < 0.001), and LDO (at 15 years: HR 2.39, 95% CI 1.49–3.85; P = 0.004), and a decreased hazard for ectropion (HR 0.89, 95% CI 0.81–0.97; P = 0.008). No association between TGM treatment and ptosis hazard was detected (HR 0.99, 95% CI 0.89–1.09; P = 0.78).

Conclusions

TGMs are associated with an increased risk of undergoing surgery for LDO, entropion, and trichiasis.

Introduction

Glaucoma is a leading cause of irreversible vision loss [1]. Therapy is directed towards reduction of intra-ocular pressure (IOP) to lower the risk of progressive retinal ganglion cell loss [2]. While the management of glaucoma is complex, with multiple treatment modalities available to lower IOP, topical glaucoma medications (TGMs) have a prominent role [3]. These medications are usually prescribed as first-line therapy for patients with glaucoma [4], and treatment may be indefinite.

All classes of TGM have clinically important side effects, with considerable overlap between classes in this regard [5]. In cases when medications are not tolerated, or if anticipated adverse effects are unacceptable, then an alternative therapy may be considered, such as laser trabeculoplasty, microinvasive glaucoma surgery, or traditional incisional surgery [6, 7]. Awareness of TGM adverse effects is critical to patient care, with implications including adherence to therapy, control of IOP, and disease progression [8].

Lacrimal drainage obstruction (LDO) has been observed in patients using TGMs, including canalicular stenosis and nasolacrimal duct obstruction (NLDO) [9–11]. Additionally, reports have described entropion, ectropion, and trichiasis in patients with TGM exposure [12–18]. The generalizability of these small studies is limited. LDO and eyelid malposition were not evaluated in the landmark clinical trials of glaucoma therapy [19–23]. Similarly, longitudinal population-level data are lacking.

Given the ubiquitous use of TGMs in the management of glaucoma, and considering the increasing availability of alternative therapies, it is important for providers to be knowledgeable of adverse effects associated these medications. Hence, we evaluated the effect of treatment with TGM on the risk of undergoing surgery for LDO and eyelid malposition in a population-based setting.

Methods

Study overview

We conducted a population-based matched cohort study using linked healthcare databases to investigate the association between TGM and LDO and eyelid malposition. We compared patients starting TGM to those followed for glaucoma or suspected glaucoma but not prescribed TGM and evaluated the subsequent occurrence of LDO or eyelid surgery. The study protocol was approved by the Research Ethics Board at Queen’s University, Kingston, Canada and follows the RECORD guidelines for reporting observational studies [24]. Individual patient-level consent was not required.

Data sources

In Ontario, universal healthcare insurance covers the population of approximately 14.5 million. Data evaluated in this study are sourced from population-level healthcare databases that have been used in previous studies [4, 25–27]. These databases were accessed through the secure, privacy-protected environment at ICES (formerly known as the Institute for Clinical Evaluative Sciences). ICES is an independent, non-profit research institute whose legal status allows it to collect and analyse healthcare and demographic data for health system evaluation and improvement. Datasets were linked using unique encoded identifiers and analysed at ICES.

The Ontario Health Insurance Plan database provides accurate data about inpatient and outpatient physician services [28]. The Ontario Drug Benefit database contains accurate records of all outpatient prescriptions filled by patients 65 years of age and older [29]. The Canadian Institute for Health Information discharge abstract database contains reliable population-wide hospitalization records [30]. The National Ambulatory Care Reporting System Database provides data for all outpatient surgical procedures [31, 32]. The Registered Persons Database provides detailed demographic information for all insured persons in Ontario. The ICES provider database provides data on all practicing physicians and optometrists in Ontario [33].

Study cohorts and exposures

We identified all patients who started a TGM or were diagnosed with glaucoma or suspicion of glaucoma but not treated with medication in Ontario between January 1, 2002 and December 31, 2018 and met the following criteria: were aged 66 years or older (provincial drug insurance in Ontario covers drug costs for medications beginning at 65 years of age, hence the design allowed a 1-year look back for all drug exposures); had no exposure to any surgical or laser therapy for glaucoma in the previous 10-year period; and had no exposure to glaucoma medications in the previous year. Patients were excluded if there was no follow up with an eye care provider after study entry (first diagnosis).

Two matched cohorts were created, a TGM cohort and a no TGM cohort. Patients were matched on birth year and sex, with up to 2 untreated patients matched to every treated patient. The TGM cohort included all patients who initiated a TGM and subsequently used any TGM continuously for at least 1 year (continuous use was defined as a period with no gap greater than 6 months between prescriptions being filled). The index date for TGM patients was defined as the date of TGM initiation, and their untreated matches were assigned the same index date [34]. Matched patients in the no TGM cohort were required to have an established glaucoma diagnosis before their assigned index date. Further details of study exposures and cohorts are available in Appendix 1 (Supplementary Information).

Study outcomes and follow-up

We evaluated several outcomes, selected a priori, including LDO repair (lacrimal stenting or dacryocystorhinostomy [DCR]), entropion repair, ectropion repair, trichiasis repair (hyfrecation or cryotherapy), and ptosis repair. Details of study outcome definitions are available in Appendix 1 (Supplementary Information). Punctoplasty was not included as an outcome due to concerns that this procedure may in some instances be offered to patients with epiphora in the absence of punctal stenosis, and therefore is not a specific marker of disease. Instead, the selected study outcomes are more invasive and/or resource intensive. Patients were followed until December 31, 2019 to allow a minimum follow up period of 1 year. Patients were censored at the first of the following: end of the study period; last visit with an eye care provider (in order to avoid bias introduced by loss to follow up); 12 months plus 30 days following the last medication refill for treated patients (medications are typically dispensed in 30-day quantities, so this allowed for 1 year of follow up after the putative last TGM exposure); or, death.

Predictors

We examined variables that might vary between cohorts and that might be relevant to the requirement for lacrimal or eyelid surgery. Patient-level factors included demographics (age at index date, sex, socioeconomic status as measured by neighborhood income quintile, and rural versus urban residence); prior surgery defining the study outcomes within 5 years before index date; other ocular surgeries within 5 years before index date (cataract, cornea, retina, strabismus surgeries); a series of general medical diagnoses and procedures; medications; and, total unique medications prescribed in the year before index date. Covariate definitions are available in Appendix 2 (Supplementary Information). These 36 covariates were used to estimate a propensity score for inverse probability of treatment (IPT) weighting (see below).

Analyses

Baseline characteristics were compared between cohorts using standardized differences, which are commonly used in large observational studies. In very large studies, common statistical tests can be misleading and suggest important differences between groups when differences are clinically trivial [35, 36]. The standardized difference is independent of variable scale and is calculated as the difference between variable means divided by the pooled estimate of the standard deviation [35]. Values of less than 0.1 are considered to be insignificant differences and indicate close balance between groups [35, 36].

We evaluated Kaplan–Meier curves and Cox proportional hazards models to compare the survival to different study outcomes between cohorts. In all analyses we employed IPT weights based on treatment propensity scores [36–38]. This approach minimizes the impact of confounding when estimating the relative effect of treatment on time-to-event outcomes using observational data [39]. Briefly, we created a propensity score using a logistic regression model to estimate the probability of treatment assignment from baseline variables [36]. Using the propensity score, weights were then calculated such that each patient’s weight was equal to the inverse of the probability of receiving the treatment that the patient received [37]. These weights were applied to the sample to create a synthetic sample in which observed baseline variables are not confounded with treatment assignment [39]. Robust standard errors were used to account for the matched study design. We tested the proportional hazard assumption using graphical methods and, based on this analysis, included a linear treatment-time interaction where appropriate to account for changing HRs over time [40]. Number needed to treat (NNT) with TGM to cause or prevent 1 outcome for different study outcomes were calculated from the Kaplan–Meier estimates as previously described for time-to-event data [41, 42]. Analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC).

Results

A total of 437,392 patients were eligible, including 123,043 patients with TGM treatment and 313,349 patients without TGM treatment. Of those, a total of 354,918 patients were matched and included in subsequent analyses including 99.6% of TGM patients (122,582) and 73.9% of no TGM patients (232,336). The lower percentage of age- and sex-matched no TGM patients is a result of the larger total number of untreated patients eligible for inclusion; hence not all such patients were needed when creating the matched study cohorts.

Patients in the matched TGM and no TGM cohorts were very similar at baseline (Table 1). The mean age was 75.7 years (standard deviation [SD] 6.9) within the TGM cohort and 75.3 years (SD 6.7) within the no TGM cohort. Among TGM patients, 57.2% (70,122/122,582) were female while among no TGM patients, 56.6% (131,579/232,336) were female. The level of comorbidity was similar in both groups, with the median number of unique medications used in the previous year of 7 (interquartile range 4–11) in each cohort. Within the TGM cohort, 15.1% (18,555/122,582) had undergone cataract surgery within 5 years before study entry, and within the no TGM cohort 26.6% (61,805/232,336) had undergone cataract surgery. Otherwise, rates of ophthalmic surgeries in the 5 years before study entry were similar in each cohort, including surgeries for entropion (TGM: 0.2% or 277/122,582, no TGM: 0.3% or 696/232,336), ectropion (TGM: 0.4% or 450/122,582, no TGM: 0.5% or 1,241/232,336), trichiasis (TGM: 0.2% or 219/122,582, no TGM: 0.3% or 670/232,336), ptosis (TGM: 0.3% or 353/122,582, no TGM: 0.4% or 865/232,336), and LDO (TGM: 0.1% or 163/122,582, no TGM: 0.2% or 491/232,336).

Table 1.

Baseline characteristics of matched cohorts with and without topical glaucoma medication treatment^a.

Variable	Topical glaucoma medication	No topical glaucoma medication	Standardized differenceb
Number of patients	122,582	232,336
Age
Median (IQR)	75 (70–80)	74 (70–80)	0.05
Mean (SD)	75.7 (6.9)	75.3 (6.7)	0.05
66–70 yrs	34,403 (28.1%)	68,151 (29.3%)	0.03
71–75 yrs	30,913 (25.2%)	60,482 (26.0%)	0.02
76–80 yrs	27,070 (22.1%)	50,717 (21.8%)	0.01
≥81 yrs	30,196 (24.6%)	52,986 (22.8%)	0.04
Sex
Female	70,122 (57.2%)	131,579 (56.6%)	0.01
Male	52,460 (42.8%)	100,757 (43.4%)	0.01
Socioeconomic status quintilec
1	23,695 (19.3%)	43,921 (18.9%)	0.01
2	25,683 (21.0%)	48,016 (20.7%)	0.01
3	24,146 (19.7%)	46,321 (19.9%)	0.01
4	23,643 (19.3%)	44,851 (19.3%)	0.00
5	25,044 (20.4%)	48,591 (20.9%)	0.01
Missing/unknown	371 (0.3%)	636 (0.3%)	0.01
Residence
Urban	105,469 (86.0%)	202,685 (87.2%)	0.04
Rural	17,015 (13.9%)	29,462 (12.7%)	0.04
Missing/unknown	98 (0.1%)	189 (0.1%)	0.00
Charlson comorbidity scored
0	110,881 (90.5%)	207,241 (89.2%)	0.04
1–2	8600 (7.0%)	17,771 (7.6%)	0.02
≥3	3101 (2.5%)	7324 (3.2%)	0.04
Median no. medications (IQR)e	7 (4–11)	7 (4–11)	0.11
Ophthalmic surgeries within 5 years
Entropion repair	277 (0.2%)	696 (0.3%)	0.01
Ectropion repair	450 (0.4%)	1241 (0.5%)	0.02
Trichiasis repairf	219 (0.2%)	670 (0.3%)	0.02
Ptosis repair	353 (0.3%)	865 (0.4%)	0.01
LDO repairg	163 (0.1%)	491 (0.2%)	0.02
Cataract surgery	18,555 (15.1%)	61,805 (26.6%)	0.28
Corneal transplantation	758 (0.6%)	332 (0.1%)	0.08
Retina surgery	8116 (6.6%)	10,287 (4.4%)	0.10
Strabismus surgery	70 (0.1%)	126 (0.1%)	0.00

IQR interquartile range, SD  standard deviation, LDO  lacrimal drainage obstruction (canalicular obstruction or nasolacrimal duct obstruction).

^aAll data are presented as number of patients (percentage) unless otherwise indicated. Percentages correspond to column percentages for each variable.

^bStandardized differences greater than 0.1 are considered significant

^cNeighborhood income quintile (1 = lowest, 5 = highest).

^dCalculated over the 2 years before study entry.

^eTotal unique medications prescribed in the 1 year before study entry.

^fIncludes hyfrecation and cryotherapy.

^gIncludes lacrimal stenting and dacryocystorhinostomy.

IPT-weighted Kaplan–Meier plots for study outcomes were constructed (Fig. 1) [38]. In the TGM cohort, there was decreased event-free survival for entropion (Fig. 1a, log-rank test P < 0.001) and trichiasis (Fig. 1b, P < 0.001) compared to the no TGM cohort. Among patients using TGMs, there was increased ectropion-free survival (Fig. 1c, P = 0.007). In the TGM cohort, there was decreased LDO-free survival (Fig. 1d, P = 0.006); the curves became increasingly divergent at later time points, suggesting possible violation of the proportional hazard assumption. No difference in ptosis-free survival between cohorts was detected (Supplementary Fig. 1, P = 0.78).

Fig. 1. Weighted Kaplan–Meier curves for lacrimal drainage obstruction and eyelid malpositions.

Cohorts included patients with exposure to topical glaucoma medications, and patients followed for glaucoma/suspected glaucoma without topical glaucoma medication exposure. Study outcomes included repair of (a) entropion, (b) trichiasis, a composite of cryotherapy and hyfrecation, (c) ectropion, and (d) lacrimal drainage obstruction, a composite of canalicular stenting and dacryocystorhinostomy. Curves are adjusted using inverse probability of treatment weights.

IPT-weighted Cox survival models supported Kaplan–Meier results (Table 2). TGM use was associated with increased hazards of undergoing entropion repair (HR 1.24, 95% confidence interval [CI] 1.12–1.37; P < 0.001) and trichiasis repair (HR 1.74, 95% CI 1.57–1.94; P < 0.001), and a decreased hazard of undergoing ectropion repair (HR 0.89, 95% CI 0.81–0.97; P = 0.008). The proportional hazard assumption was evaluated using graphical methods and, based on this analysis, a linear treatment-time interaction term was incorporated for LDO repair surgery to account for the changing HR over time [40]. TGM use was associated with an increased hazard at all time points beyond 3 years and the HR increased at longer follow up time points (at 15 years: HR 2.39, 95% CI 1.49–3.85; P = 0.004). No association between TGM therapy and ptosis repair was detected (HR 0.99, 95% CI 0.89–1.09; P = 0.78).

Table 2.

Association between topical glaucoma medication exposure and eyelid malposition or lacrimal drainage obstruction^a.

Outcome	Hazard ratio	95% CI	P value
Entropion repair	1.24	1.12 to 1.37	<0.001
Ectropion repair	0.89	0.81 to 0.97	0.008
Trichiasis repairb	1.74	1.57 to 1.94	<0.001
Ptosis repair	0.99	0.89 to 1.09	0.78
LDO repairc			0.004
3 years	1.14	0.98 to 1.33
5 years	1.29	1.11 to 1.49
10 years	1.76	1.32 to 2.33
15 years	2.39	1.49 to 3.85

CI confidence interval, LDO lacrimal drainage obstruction (canalicular obstruction or nasolacrimal duct obstruction), DCR dacryocystorhinostomy.

^aConditional models for study outcomes with inverse probability of treatment weighting.

^bComposite outcome, trichiasis repair by hyfrecation or cryotherapy.

^cComposite outcome, lacrimal stenting or DCR. We tested the proportional hazard assumption using graphical methods and, based on this analysis, included a linear treatment-time interaction where appropriate to account for changing HRs over time [40]. Hence, hazard ratios for LDO repair reported at multiple time points (treatment-time interaction term included in model).

Evaluation of incidence rates adjusted with IPT weights (Table 3) corroborated the preceding analyses. Among TGM patients, there were higher rates per 100,000 person-years of follow up of entropion (TGM: 95, no TGM: 74; P < 0.001) and trichiasis (TGM: 105, no TGM: 57; P < 0.001), and a lower rate of ectropion (TGM: 114, no TGM: 125; P = 0.04). There was a higher rate of LDO in the TGM cohort (TGM: 45, no TGM: 36; P = 0.003). No difference between rates of ptosis between cohorts was detected (TGM: 92, no TGM: 90; P = 0.60). Crude incidence rates were very similar to adjusted rates (Supplementary Table 1).

Table 3.

Adjusted incidence rates per 100,000 person-years of follow up for study outcomes among patients with and without topical glaucoma medication exposure^a.

Outcome	TGM	no TGM	P value
Entropion repair	95 (87–103)	74 (70–78)	<0.001
Ectropion repair	114 (105–123)	125 (119–130)	0.04
Trichiasis repairb	105 (97–114)	57 (54–61)	<0.001
Ptosis repair	92 (85–100)	90 (85–94)	0.60
LDO repairc	45 (40–51)	36 (34–40)	0.003

TGM topical glaucoma medication, LDO lacrimal drainage obstruction (canalicular obstruction or nasolacrimal duct obstruction).

^aData are presented as rate (95% confidence interval).

^bComposite outcome, trichiasis repair by hyfrecation or cryotherapy.

^cComposite outcome, lacrimal stenting or dacryocystorhinostomy.

NNT with TGM to cause (if NNT > 0) or prevent (if NNT < 0) an additional event over 10 years were calculated (Table 4). Estimates were statistically significant for entropion (NNT: 609), trichiasis (NNT: 245), ectropion (NNT: −685), and LDO (NNT: 684).

Table 4.

Number needed to treat with topical glaucoma medications to either cause or prevent 1 outcome over 10 years.

Outcome	NNTa	95% CI
Entropion repair	609	377 to 1587
Ectropion repair	−685	−3340 to −382
Trichiasis repairb	245	197 to 322
Ptosis repair	5997	−∞ to −1121; 821 to ∞c
LDO repaird	684	439 to 1551

NNT number needed to treat, CI confidence interval, LDO lacrimal drainage obstruction (canalicular obstruction or nasolacrimal duct obstruction), DCR dacryocystorhinostomy.

^aPositive and negative NNT values indicate the number of patients needed to treat with TGM to respectively cause or prevent an additional outcome over 10 years [41].

^bComposite outcome, trichiasis repair by hyfrecation or cryotherapy.

^cStatistically insignificant CIs for NNT are considered to cover two separate numerical regions, ranging from the upper confidence limit to infinity, and from negative infinity to the lower confidence limit [42].

^dComposite outcome, lacrimal stenting or DCR.

Discussion

TMGs are commonly prescribed, and associations between their use and outcomes including LDO and eyelid malposition have not been examined outside of small cross-sectional observational studies. In this longitudinal, population-based cohort study, we found that patients exposed to TGMs were at increased risk of undergoing LDO repair (canalicular stenting or DCR), entropion repair, and repair of trichiasis (hyfrecation or cryotherapy).

The absolute differences in survival to these outcomes between cohorts were small and most apparent after several years of therapy. Outcomes were uncommon in both cohorts, and NNT with TGM to cause/prevent study outcomes were large. Nevertheless, the observed effects are important for several reasons. First, glaucoma is a common disease and large numbers of patients are treated with TGM; second, use of TGMs is frequently long-term; third, therapies that avoid topical drug delivery are increasingly available; and fourth, the question of an association between TGM exposure and incidence of eyelid malposition/LDO remains unanswered in the literature.

Investigators have reported on the potential association between TGMs and LDO. Kashkouli et al. examined patients with and without TGM use for LDO; signs of obstruction were present in 20% eyes with TGM use compared to 9% of control eyes [11]. In a non-comparative series, McNab described canalicular obstruction arising in 14 patients with exposure TGM [9]. Seider et al. found that previously-diagnosed glaucoma was present in 23% of patients with symptomatic NLDO in comparison to 6% of eyes undergoing cataract surgery and concluded that TGM exposure may mediate the link between glaucoma and NLDO [10]. In contrast, Ohtomo et al. found no difference in a history of glaucoma among patients undergoing repair of NLDO compared to patients undergoing cataract surgery, nor in history of timolol use [43]. Our study provides population-level evidence that first, TGM use is associated with subsequent receipt of LDO surgery, and second, such obstruction may arise years after therapy initiation.

Previous studies of TGM use and eyelid malposition include small descriptive studies. Golan et al. reported the prevalence of previously-diagnosed glaucoma to be 13% among patients undergoing repair of entropion or ectropion, versus 4% among control patients undergoing blepharoplasty, with an average of 2.7 classes of TGM used by patients with glaucoma [15]. Non-comparative reports have described entropion, ectropion, or trichiasis associated with TGM [12–14, 16–18]. In many cases, drug-associated ectropion resolved after discontinuing the putative causal agent [16–18]. In one case, TGM-related ectropion progressed to cicatricial entropion [12]. These reports indicate that, while eyelid malpositions arise in the setting of TGM use, ectropion specifically may resolve without surgery. Indeed, we found that surgical repair of entropion and trichiasis were increased among TGM patients, whereas such an effect was not seen for ectropion.

Upregulation of inflammatory pathways underlies ocular surface disease, which is common among patients with long-term use of TGMs [5]. More specifically, preservatives used in many TGMs may be an important contributing factor; inflammatory symptoms and signs are more prevalent among patients using preserved TGMs [44]. In our study, the pattern of risks for different eyelid malpositions mirrors what is observed in conjunctival cicatrization, where entropion and trichiasis are commonly described but ectropion is not [45]. While overt TGM-induced cicatricial conjunctivitis may be uncommon, comparatively subtle signs on that spectrum include forniceal shortening and subepithelial fibrosis which do occur in the setting of TGM use [46, 47]. In patients with chronic exposure to TGM, these tissue changes may lead to entropion and trichiasis while effectively protecting against ectropion through vertical shortening and mechanical tethering of the posterior lamella of the lower eyelid. Considering the nasolacrimal system is contiguous with the conjunctiva, its mucosa will be exposed to topically-administered agents; similar inflammatory changes may result in fibrosis and obstruction of that system [9, 11]. Hence, chronic inflammation incited by TGMs may be a unifying etiologic mechanism for ocular surface disease, eyelid malposition, and LDO in patients with glaucoma. Mechanical stretching of eyelids associated with long-term drop instillation may increase eyelid laxity and contribute to eyelid malposition.

We did not detect any relative increase in risk of receiving ptosis repair associated with TGM treatment. Shah et al. reported use of topical prostaglandin analogues was associated with a four-fold increased risk for ptosis compared to no such exposure [48]. Our finding of no increased risk of ptosis repair among patients using TGM would indicate that if ptosis develops, then its clinical significance may not warrant repair.

This study has important strengths. First, we examined a clinically relevant question with important implications for patient care, as LDO and eyelid malposition are associated with meaningfully decreased quality of life [49] and alternative therapies for IOP lowering are increasingly available [6, 7]. Second, consistent results were demonstrated across a multiplicity of analyses, including weighted Kaplan–Meier curves, Cox regression, and incidence rates. Third, the study’s large size and longitudinal design permit assessment of relatively uncommon outcomes that may present years after exposure to TGM. Fourth, study cohorts were well balanced with respect to baseline characteristics, and the use of propensity score weighting provided additional higher-dimensional control of potential confounders. Fifth, the population-based design allows for superior generalizability compared to previous studies.

This study has limitations. First, as the study objective was to evaluate risk of LDO and eyelid malposition outcomes among all patients receiving topical glaucoma therapy, we did not stratify patients with TGM exposure by specific medication class, or by exposure to medication preservatives. Of note, preservatives used in TGMs are felt to be an important contributor to inflammatory changes of the ocular surface, especially in patients with pre-existing ocular surface disease or poly-pharmacy [50]. Nevertheless, in current treatment algorithms patients frequently have exposure to multiple medication classes and preserved drops [3]; hence our approach provides meaningful data that are generalizable to clinical practice. Second, our methodology considers repair of LDO and eyelid malpositions as surrogates for these disorders; their true incidence is likely higher. Third, the study included only patients aged 66 and older; we would not expect this to impact our major findings. Fourth, there may be provider- and patient-level factors that correlate with preference for intervention in general. We cannot determine if all such factors were addressed, although important patient demographic factors were controlled for in the analyses. Fifth, multiple comparison correction was not performed. Although robust P values were generally found, this should be considered when interpreting study findings.

In this population-level study of patients starting TGM, we found that use of these medications is associated with a small increase in the long-term risk of undergoing surgery for LDO, entropion, and trichiasis and a decreased risk of ectropion surgery. These findings may support the growing role for therapies that reduce or delay TGM use, such as laser trabeculoplasty and novel drug delivery systems.

Summary

What was known before

• Topical glaucoma medications (TGMs) are a mainstay of glaucoma therapy. Knowledge of adverse effects is important considering the increasing availability of alternative therapies, including those that avoid topical drug delivery.

• Ocular adnexal disorders, including lacrimal drainage obstruction and eyelid malposition, have been observed among patients who use TGMs in cross-sectional reports. Longitudinal, population-level data regarding incident ocular adnexal disorders in patients using TGMs are lacking.

What this study adds

• Patients who use TGMs are at an increased long-term risk of undergoing surgery for lacrimal drainage obstruction (dacryocystorhinostomy and canalicular stenting), entropion, and trichiasis compared to patients who are followed for glaucoma/suspected glaucoma but who are not treated with TGMs.

• These findings may support the growing role for therapies that reduce or delay use of TGMs.

Author contributions

MPQ was responsible for conception and design of the study, data acquisition and interpretation, drafting and revising the paper, approval of the final version, and overall study accountability. MW was responsible for study design, data acquisition, critical revision of the paper, approval of the final version, and overall study accountability. VK, SSG, and MAM were responsible for study design, data interpretation, critical revision of the paper, approval of the final version, and overall study accountability. RJC was responsible for conception and design of the study, data acquisition and interpretation, critical revision of the paper, approval of the final version, and overall study accountability.

Funding

This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health and the Ministry of Long-Term Care. Dr. Campbell is supported by the David Barsky Chair in Ophthalmology and Visual Sciences, Queen’s University, Kingston, Ontario, Canada.

Data availability

The dataset from this study is held securely in coded form at ICES. While data sharing agreements prohibit ICES from making the dataset publicly available, access may be granted to those who meet pre-specified criteria for confidential access, available at www.ices.on.ca/DAS. The full dataset creation plan and underlying analytic code are available from the authors upon request, understanding that the computer programs may rely upon coding templates or macros that are unique to ICES and are therefore either inaccessible or may require modification.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41433-022-02322-w.