Long-Term Use of Selective Serotonin
Reuptake Inhibitors
and Risk of Glaucoma in Depression Patients
Hsin-Yi Chen, MD, Cheng-Li Lin, MSc, and Chia-Hung Kao, MD
INTRODUCTION
D
epression is a highly prevalent mood disorder that can lead to serious disabilities and functional impairment.1,2 Onestudy indicated that from 1997 to 2005, the prevalence of
antidepressant usage among elderly people increased substantially in Taiwan.3 Currently, available selective serotonin
reuptake inhibitors (SSRIs) are the most widely prescribed type of medication for depression patients.4 Short-termSSRI exposure
induces acute angle-closure glaucoma (AACG),4,5 which is a
potentially blinding ocular emergency, and is relatively common in Asians, especially those of Chinese ethnicity.6–8 We recently
reported that patients with short-term SSRI use are at a 5.8-folds increased risk of AACG.9 It, however, remains unclear whether
long-term SSRI use influences intraocular pressure (IOP) or increases the risk of glaucoma.4,10 Glaucoma comprises a set
of ocular disorders that lead to optic nerve damage that is often associated with increased IOP.11 It is also the leading cause of
irreversible blindness worldwide.12 Primary glaucoma can be
divided into 2 major types, primary open-angle glaucoma (POAG), and primary angle-closure glaucoma (PACG), which are the 2 most common types in the Chinese ethnic population of Taiwan.6,7,13 Furthermore, previous studies have reported a
strong association between glaucoma and depression.14–16
Therefore, to evaluate whether long-term SSRI use influences the risk of POAG and POAG in patients diagnosed with depression, we conducted this study by using a population-based dataset from the National Health Insurance (NHI) program of Taiwan. According to a review of relevant literature, this study is the first
to address this crucial problem by using a large claims database.
METHOD
Data Source
The data for analysis in this retrospective cohort study
were retrieved from the Longitudinal Health Insurance Database 2000 (LHID2000), an electronic claims database of the
NHI program. The NHI program, which started on March 1, 1995, provides comprehensive medical coverage for people residing in Taiwan.17 The LHID2000 was established by the
National Health Research Institutes and contains all the original claims data of 1000,000 patients (approximately 5% of the
Taiwan population), who were randomly sampled from the 2000 Registry of Beneficiaries of the National Health Insurance
Research Database. The diagnostic codes in the LHID2000 are based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). This study was exempted from informed consent by the Institutional Review Board of China Medical University (CMU-REC-101–012).
Sample Selection
This study included patients aged >20 years who were diagnosed with depression (ICD-9-CM codes 296.2, 296.3, 300.4, and 311), had complete information regarding age and sex, and had no history of glaucoma (ICD-9-CM code 365) from 2000 to 2010. The depression patients were divided into 2 cohorts on the basis of their SSRI use: the SSRI cohort included patients who had undergone SSRIs therapy for at least 1 year (365 days), whereas the comparison cohort included patients who had not receivedSSRI therapy. The index date for the SSRI cohort as well as the comparison cohort was day 365. Patients in the SSRI and comparison cohorts were selected through 1:1 matching based on a propensity score.18 The propensity score was calculated using
logistic regression to estimate the probability of treatment assignment on the basis of baseline variables, namely the year of SSRI
treatment, age, sex, the comorbidities of diabetes mellitus (ICD-9-CM code 250), hypertension (ICD-(ICD-9-CM codes 401–405), hyperlipidemia (ICD-9-CM code 272), coronary artery disease (ICD-9-CM codes 410–414), anxiety (ICD-9-CM code 300.00), and non-SSRI medication for treating depression. The C-statistic
of the logistic regression model was 0.53.
Outcome Measurement
The main outcome was a diagnosis of POAG (ICD-9-CM code 365.1) or PACG (ICD-9-CM code 365.2) during followup. All patients were followed from the index date until a new diagnosis of POAG or PACG, censorship because of death, withdrawal from the insurance system, loss to follow-up, or the end of follow-up on December 31, 2011.
Statistical Analysis
The distributions of demographic variables, namely age, sex, non-SSRI use, and comorbidities, were compared between the 2 cohorts. The baseline characteristics of the SSRI cohort and comparison cohort were compared using standardized mean differences, calculated as the difference in the mean or proportion of a variable divided by a pooled estimate of the
standard deviation of the variable. A value of the standardized mean differences equal to 0.1 or less indicates a negligible difference in the mean between the 2 cohorts. The cumulative incidence of POAG and PACG between the SSRI and comparison cohorts was assessed using the Kaplan–Meier method
and the differences between the curves were evaluated by conducting a log-rank test. The incidence density rates of POAG and PACG were calculated according to age, sex, non-SSRI antidepressant use, and comorbidity. Univariable and multivariable Cox proportional hazards regression models were used to assess the effects of SSRIs on the risk of POAG and PACG. In addition, hazard ratios and 95% confidence intervals (CIs) were estimated in the Cox models. In the multivariable Cox model, only diabetes mellitus and hyperlipidemia were found to be significant. Further data analysis was performed to determine the interaction effect between SSRIs, diabetes mellitus, and hyperlipidemia. SAS version 9.4 for Windows (SAS
Institute, Cary, NC) was used to conduct all statistical analyses, and all statistical testing was performed at a 2-tailed significance level of 0.05.
RESULTS
Eligible study patients comprised 13,093 patients who had used SSRIs for more than 1 year, and 13,093 patients who had
never undergone SSRI therapy, matched according to the propensity score (Table 1). The mean ages in the SSRI cohort and the comparison cohort were 49.3 years (SD¼16.2) and 49.4 years (SD¼16.5), respectively. Certain patients in the
SSRI cohort did not have a medicinal history for non-SSRI antidepressant use (53.7%), diabetes mellitus (9.91%), hypertension
(35.9%), hyperlipidemia (26.7%), coronary artery disease (22.7%), and anxiety (37.1%). The Kaplan–Meier graph in Figure 1A-B illustrates that the cumulative incidence of POAG and PACG between the SSRI cohort and the comparison cohort exhibited nonsignificant differences (log-rank test P¼0.52 in Figure 1A, and log-rank test P¼0.32 in Figure 1B).
During the mean follow-up periods of 5.55 years
(SD¼2.97) and 5.71 years (SD¼3.11) for the SSRI cohort and the comparison cohort, respectively, the overall POAG incidence was nonsignificantly higher in the SSRI cohort than in the comparison cohort (1.51 versus 1.39, respectively, per 1000 person-years), with an adjusted hazard ratio (aHR) of 1.07 (95% CI¼0.82–1.40; Table 2). In both cohorts, the POAG incidence increased with the presence of comorbidity. The relative risk of POAG between the SSRI and comparison cohorts, however, was higher among patients without comorbidity (aHR¼1.76, 95% CI¼1.03–3.02).
The overall incidence of PACG in the SSRI cohort was
nonsignificantly lower than that in the comparison cohort (0.95 versus 1.11, respectively, per 1000 person-years), with an aHR of 0.85 (95% CI¼0.62–1.18). Furthermore, the risk of PACG in the SSRI cohort varied nonsignificantly relative to that in the comparison group for age subdivisions.
Table 3 lists the joint effects of SSRI use and comorbidity on the risk of POAG or PACG. Compared with patients with SSRI use and nondiabetes mellitus, patients who used an SSRI and had diabetes mellitus were 3.16-folds more likely to develop POAG (95% CI¼2.05–4.87), followed by patients without SSRI use and with diabetes mellitus (aHR¼1.87, 95% CI¼1.11–3.16). Compared with patients without SSRI use and hyperlipidemia, those without SSRI use but with hyperlipidemia were 2.66-folds more likely to develop POAG (95%
CI¼1.79–3.97), followed by patients who used an SSRI and had hyperlipidemia (aHR¼1.85, 95% CI¼1.20–2.86). We observed a significantly higher risk of PACG in patients without SSRI use and with hyperlipidemia (aHR¼2.43, 95%
CI¼1.55–3.80) and in patients who used SSRIs and had hyperlipidemia (aHR¼1.80, 95% CI¼1.11–2.90) compared with patients without SSRI use and hyperlipidemia.
DISCUSSION
A literature review revealed a relatively high prevalence of anxiety and depression among glaucoma patients.19 The level of
depression was found to be significantly higher in PACG patients than in POAG patients and controls.14 Depression is
more common in patients with increasing glaucoma severity.20
Currently, available SSRIs are the most widely used
medications for treating depression patients.4,5 Acute angleclosure
glaucoma is the most severe SSRI-related ocular complication.
4,5 Several instances of AACG associated with the use
of SSRIs, such as paroxetine,21–25 fluvoxamine,26 citalopram, 27,28 escitalopram,29 and sertraline30 have been reported.5
We also recently reported that patients using SSRIs immediately are at a 5.8-folds increased risk of AACG.9 Acute angleclosure
glaucoma is a subtype of PACG, which is an ocular
emergency characterized by a sudden spike in IOP. By contrast, chronic angle-closure glaucoma consists of slow and progressive angle closure with an elevated IOP,6,7,9,13 and is a more
common occurrence than AACG in PACG. By comparison, POAG is painless, tends to advance slowly over time, and often exhibits no symptoms until the disease has progressed substantially. Pupil dilatation mediated by 5-hydroxytryptamine, serotonin
receptors and noradrenaline receptors has been proposed as the mechanism underlying SSRI-induced AACG.10,30 The
presence of serotonin and its metabolites in the aqueous humor, in addition to its availability on 5-hydroxytryptamine receptors in the iris–ciliary body complex, are involved in regulating aqueous humor dynamics.4 Although SSRIs might influence
IOP in humans,31 the literature presents conflicting data on
whether the activation of serotonin receptors induces changes in IOP. One study group expressed the concern that long-term
SSRI use might raise the risk of POAG.31 In the premarketing
and subsequent clinical trials for fluoxetine, 585 adult patients were assessed by an ophthalmologist.31 In these trials, 63 cases
of glaucoma were reported as a suspected adverse effect of fluoxetine from an estimated patient population of 21 million
(Dista Products Limited, personal communication).31 The manufacturers
of paroxetine are aware of 4 cases of AACG, 6 cases
of glaucoma (unspecified), and a single case of increased IOP in a UK patient population of more than one million people
(SmithKline Beecham Pharmaceuticals, personal communication).
31 Moreover, no long-term study regarding the influence
of SSRIs on IOP has been reported. All of these data indicate that the understanding of long-term SSRI effects on IOP remains relatively limited, although the manufacturers’ own data suggest that less than 1% of patients exhibited any IOP
changes after SSRI treatment. In the current study, the cumulative incidence of POAG and PACG exhibited nonsignificant differences between the
SSRI and comparison cohorts. Furthermore, the overall incidences of POAG and PACG in the SSRI cohort were found to
be nonsignificantly higher than those in the comparison cohort after using a nationally representative dataset. This result offers new insight supporting the belief that long-term SSRI use does not influence the risk of POAG and PACG in the Taiwan population. Although we cannot understand the real effect of SSRIs on IOP on the basis of the claims database, we believe that this result provides indirect support to the notion that depression patients who use SSRIs are not at a risk of glaucoma.
According to recent epidemiologic studies conducted in Taiwan, patients with open-angle glaucoma are
significantlymore likely to have comorbidities.16,32 Therefore,
we considered the interaction effect of comorbidities and SSRI usage on the risk of glaucoma. The results revealed that diabetes mellitus and hyperlipidemia are 2 strong risk factors for POAG in patients with SSRI use as well as in those without. This finding confirms the previously held notion that POAG patients are significantly more likely to have comorbidities.16
PACG in patients without SSRI use (aHR¼2.43, 95% CI¼1.55–3.80) as well as in patients with SSRI use (aHR¼1.80, 95% CI¼1.11–2.90) compared with patients without SSRI use and without hyperlipidemia. In the literature, few studies have evaluated the medical comorbid condition in PACG patients. This study is the first to report that hyperlipidemia is a critical risk factor for PACG in patients with
depression. Plausible reasons for this finding remain unclear,
and further observation is warranted to address this finding. Despite obtaining promising results, our study had the
following limitations. First, we defined glaucoma by relying entirely on claims data (ICD-9-CM coding from clinicians), an approach that may be less accurate than determining diagnoses individually through a standardized procedure. In this type of claims database study, clinical information, such as that on IOP, central corneal thickness, visual field findings, and optic nerve evaluations is unavailable. Second, this study had a selection bias. Because theNHI database is composed only of data frompatients who underwent treatment, patients who had not received treatment for depression might have been recruited in the comparison
cohort. Third, because of the lack of laboratory and imaging data in individual chart records, the National Health Insurance
Research Database is used primarily for insurance purposes
and has not been validated entirely for research; thus, uncontrolled confounding factors, such as visual field severity, IOP
reading, and potential biases may have affected our retrospective case-control study. Fourth, despite the large sample, the study cohort consisted of Taiwanese patients, and thus these findings are not easily generalizable to other population groups. Finally, the SSRI use reported in the LHID2000 does not necessarily correspond to actualSSRI use. This is because of the possibility of poor medication compliance and the ease with which such medication use can be initiated and ceased.33
Our study had the following strengths. First, the strength of
the database is excellent because of the large sample randomization, and we could follow patient cases over time to assess the
relationship between SSRI exposure and the subsequent onset of POAG or PACG. Second, the database includes data on a broad
range of people with different sociodemographic profiles, unlike those used in smaller studies, in which patients are recruited from specific regions that might not be representative of the entire population. Third, our study is the first to evaluate the relationship between long-term SSRI use and glaucoma risk in a purely Asian depression population by using a large claims database. Our findings can provide a strong foundation for further longitudinal research.
In conclusion, long-term SSRI use does not influence the risk of POAG or PACG in depression patients. Clinicians, however, should still consider the potential risk of glaucoma in certain high-risk groups.
