Serious Adverse Events of Oral and Topical Carbonic Anhydrase Inhibitors

Key Points

Question  What is the risk of a severe complicated adverse reaction for patients following prescription of an oral or topical carbonic anhydrase inhibitor?

Findings  In this population-based matched cohort study of 128 942 individuals, patients prescribed an oral carbonic anhydrase inhibitor had an absolute risk of a severe complicated adverse reaction that was small and just marginally higher than the risk for patients prescribed a topical carbonic anhydrase inhibitor.

Meaning  Given the low risk of severe adverse reactions, this population-level analysis may support reconsidering the reluctance toward prescribing an oral carbonic anhydrase inhibitor.

Importance  Some ophthalmologists may be reluctant to prescribe oral carbonic anhydrase inhibitors, given the potential for life-threatening systemic adverse reactions.

Objective  To conduct a population-based analysis of the safety of oral or topical carbonic anhydrase inhibitors in clinical care.

Design, Setting, and Participants  This matched longitudinal cohort study took place in Ontario, Canada. Consecutive patients older than 65 years who were prescribed an oral or topical carbonic anhydrase inhibitor in Ontario, Canada, between January 1, 1995, and January 1, 2020, were identified. Patients were matched 1-to-1 based on age, sex, and diabetes status. Time zero was defined as the date of the first identified prescription for the medication, and the primary analysis focused on the first 120 days of follow-up.

Main Outcomes and Measures  The primary end point was a severe complicated adverse event of either Stevens-Johnson syndrome, toxic epidermal necrolysis, or aplastic anemia.

Results  Overall, 128 942 matched patients initiated an oral or topical carbonic anhydrase inhibitor during the 25-year study period. The mean (SD) age was 75 (6.6) years, 71 958 (55.8%) were women, and 25 058 (19.4%) had a diagnosis of diabetes. The oral and topical carbonic anhydrase inhibitor groups had similar baseline demographics. Patients prescribed an oral carbonic anhydrase inhibitor had an absolute risk of a severe complicated adverse event of 2.90 per 1000 patients, whereas patients prescribed a topical carbonic anhydrase inhibitor had an absolute risk of 2.08 per 1000 patients. This difference was equivalent to a risk ratio of 1.40, with a number needed to harm of 1 in 1220 patients (95% CI, 1.12-1.74; P = .003). This generally low risk was replicated in multivariable regression controlling for confounding factors. Additional risk factors for a severe complicated adverse event included patients with more comorbidities and those with more frequent clinic contacts.

Conclusions and Relevance  The risk of a serious adverse reaction following prescription of an oral or topical carbonic anhydrase inhibitor was low and similar between agents. Given the low risk of severe adverse reactions, this population-level analysis supports reconsidering the reluctance toward prescribing an oral carbonic anhydrase inhibitor.

Glaucoma is a progressive optic neuropathy that affects more than 60 million patients worldwide.¹ Carbonic anhydrase inhibitors (CAIs) decrease aqueous humor production, thus lowering intraocular pressure.² Topical CAIs are associated with a 20% intraocular pressure reduction, whereas oral CAIs are associated with a 30% intraocular pressure reduction.^3,4 When there are concerns about efficacy or patient adherence, oral CAIs are sometimes preferred over topical agents.

Topical CAIs are often prescribed early in glaucoma treatment, whereas oral CAIs are reserved as a temporizing measure for acute, significant intraocular pressure elevation or for patients who are refractory to topical medications and laser trabeculoplasty. A reluctance to prescribe oral CAIs is attributed to the risk of life-threatening complications, systemic adverse events, and cross-reactivity in patients with a sulfonamide allergy.^2,5-9 CAIs can cause systemic adverse reactions, including fatigue, paresthesias, nausea, dizziness, hypokalemia, kidney stones, and weight loss.^2,9-11 Rare but serious complications include Stevens-Johnson syndrome (SJS), toxic-epidermal necrolysis (TEN), and aplastic anemia. The symptoms from CAIs have been described as intolerable, debilitating, serious, and sometimes worse than the disease itself. A prior report has noted that oral CAIs have fallen into disfavor because of the concern for serious adverse events.⁹ The same reluctance is not found with topical CAIs.

Some patients may not tolerate oral CAIs; however, the risk of serious adverse reactions is unclear. Given the paucity of evidence and the potential benefits, a critical analysis can support or refute prevailing concerns associated with oral CAIs. To examine the safety of CAIs in a universal health care system in Canada, we identified patients prescribed an oral CAI and compared their adverse event rates with matched controls prescribed a topical CAI. We hypothesized the risk of a serious adverse reaction following prescription of an oral CAI was low.

We performed a matched longitudinal cohort study by identifying consecutive patients older than 65 years who were prescribed an oral or topical CAI in Ontario, Canada, between January 1, 1995, and January 1, 2020. Patients were excluded if they lacked valid health care insurance or were younger than 66 years, as the latter group did not have universal prescription drug coverage. Patients were not required to be under the care of an ophthalmologist or to have a diagnosis of glaucoma. All patients in Ontario meeting the prespecified inclusion criteria and time frame were included. Time zero was defined as the date of the first prescription for an oral or topical CAI, and patients who had a prior prescription of a CAI in the 120 days before time zero were excluded. An oral CAI was defined as either acetazolamide or methazolamide, whereas a topical CAI was defined as dorzolamide, brinzolamide, or a combination agent. The oral agent zonisamide was not available and was excluded. Topiramate is not used in glaucoma but is sometimes prescribed off-label in refractory idiopathic intracranial hypertension.¹² Given the rarity of topiramate use and that the evidence for topiramate in this context is weak, this drug was excluded. The study was approved by the Research Ethics Board of Sunnybrook Health Sciences Centre and was conducted using patient privacy safeguards at the Institute for Clinical Evaluative Sciences. A waiver for individual consent was granted, and patients were not offered incentives to participate.

Baseline demographics, drug prescriptions, physician clinic visits, and hospital discharge summaries were collected and analyzed through health services databases at the Institute for Clinical Evaluative Sciences using secure linkage procedures based on unique encoded patient identifiers.¹³ These databases have been validated.^14-17 Included baseline demographic data were age (years), sex (binary), socioeconomic status (quintile), and home location (urban, rural). Data on race or ethnicity were not available.

A patient’s general health was defined by the 1-year preceding study enrollment, which was available for all patients residing in Ontario during this time frame. Medical comorbidities included a previous diagnosis of diabetes, hypertension, and depression (eTable in the Supplement).^18,19 Ocular comorbidities included a diagnosis of glaucoma, recent intraocular surgery, nonsurgical procedures, additional ocular medications, and repeated visual field tests (eTable in the Supplement). In addition, the following 6 general clinical indicators were collected: number of prescribed medications, emergency department visits, hospitalizations, outpatient physician visits, outpatient ophthalmologist visits, and Charlson Comorbidity Index (CCI) scores.²⁰ The CAI characteristics included the specific agent and days of medication supplied.

The primary analysis compared patients prescribed an oral CAI with control patients prescribed a topical CAI. Individual patients were matched 1-to-1 based on age (within 1 year), sex (binary), and diabetes status (binary) identified through the same database during the same baseline time frame. Diabetes was chosen as a matching characteristic given that it is a leading cause of kidney failure and a surrogate for kidney function, which may affect the excretion of oral CAIs. This created pair-matched sets whereby 1 patient was prescribed an oral CAI and 1 patient was prescribed a topical CAI. Patients initiating both oral and topical CAIs were included in the oral CAI cohort only. For patients who were prescribed CAIs at multiple time points, only the first medication initiation was considered.

The primary end point was a severe complicated adverse reaction (SCAR), representing a physician diagnosis of at least 1 of 3 idiosyncratic adverse events: SJS, TEN, or aplastic anemia based on International Classification of Diseases, Ninth Revision (ICD-9) codes (eTable in the Supplement).²¹ A previous systematic review documented that of all patients with the ICD-9–Clinical Modification (ICD-9-CM) code 695.1, between 53% and 60% had validated cases of SJS, TEN, or erythema multiforme.²² No similar analysis exists for the more general ICD-9-CM code of 695. To our knowledge, similar data are not available for the diagnostic code of aplastic anemia; however, the ICD-9 code of 284 has been used for case identification in this setting.^23,24

Minor or nonserious adverse drug reactions without the corresponding ICD code were not available for analysis. Secondary analyses included aplastic anemia, SJS, TEN, kidney failure, kidney stone, cardiac dysrhythmia, delirium, or physician diagnosis of a serious adverse drug reaction as individual outcomes. In addition, total hospital admissions and emergency department visits were investigated. The primary analysis focused on the first 120 days following prescription of therapy, while a sensitivity analysis considered the following intervals: 1 week, 1 month, 6 months, and 1 year. The first 120 days following initial prescription was chosen given that most cases of SJS/TEN have been reported to occur within the first 8 weeks and cases of aplastic anemia following CAI therapy typically occur within the first 3 months of treatment.^25,26

Descriptive statistics were used to characterize patients initiating an oral or topical CAI. The primary analysis examined the risk of SCAR events in the 120 days following initial treatment using a conditional logistic regression to compare patients initiating oral CAI (cases) with those initiating topical CAI (controls) while accounting for pair matching. McNemar test accounting for pair matching was also performed in the primary analysis. Regular logistic regression was examined as a test of robustness, as well as the examination of the primary end point across relevant patient subgroups, including prescription date. We also examined other severe adverse events beyond those captured in the definition of a SCAR event in a secondary analysis. In addition, univariable logistic regression analyses were performed for baseline covariates as potential risk factors of adverse events. In turn, clinical variables were considered in a multivariable logistic regression analysis to control for risk factors. The odds ratio and 95% CIs were computed for each risk factor. A 2-tailed P value of .05 was used to define statistical significance for the primary outcome of interest, and there were no adjustments to P values for multiple analyses. SAS Enterprise Guide version 7.15 (SAS Institute) was used for analysis.

Overall, 64 477 patients were prescribed an oral CAI and 111 598 were prescribed a topical CAI during the 25-year study interval. After matching for age, sex, and diabetes diagnosis, we obtained a cohort of 128 942 total patients, of whom 64 471 were prescribed an oral CAI and 64 471 were prescribed a topical CAI. The 2 groups were similar in most baseline characteristics (Table 1). Differences between groups existed in the proportion of patients with glaucoma, repeated visual field test, recent eye surgery, and hospital admissions. The mean (SD) age was 75 (6.6) years, 71 958 (55.8%) were women, and 25 058 (19.4%) had a diagnosis of diabetes. The proportion of patients with a CCI score of 3 or higher was 3.7% (n = 2355) in the oral CAI group and 2.0% (n = 1259) in the topical CAI group. About half (64 394 [49.9%]) had a diagnosis of glaucoma within the year before initiating the CAI. The most common oral agent was acetazolamide (59 554 [92.4%]), prescribed for a mean (SD) initial duration of 8.4 (11.1) days. The most common topical CAI was dorzolamide or combination agents involving this agent (45 505 [70.6%]), prescribed for a mean (SD) initial duration of 23.0 (18.0) days.

The primary analysis was based on 15 317 879 total patient-days of follow-up (median [IQR], 120 [120-120] days) and identified 321 total SCAR events. Patients prescribed an oral CAI accounted for 187 SCAR events (absolute risk, 2.90 per 1000 patients; number needed to harm, 1 per 345 patients). Patients prescribed a topical CAI accounted for 134 SCAR events (absolute risk, 2.08 per 1000 patients; number needed to harm, 1 per 481 patients). The difference amounted to a 40% relative increased risk of a SCAR event for patients prescribed an oral CAI (unmatched χ² test: 95% CI, 12-74; P = .003). The number needed to harm for the difference in SCAR events between oral and topical CAIs was 1 in 1220 patients. McNemar test accounting for matching yielded a 40% relative increased risk of a SCAR event for patients prescribed an oral CAI (95% CI, 12-74; P = .004). The risk of a SCAR event was directly proportional with time for both the oral and topical CAI groups (Figure 1).

The low baseline risk and modest relative risk of a SCAR event associated with oral relative to topical CAIs extended to most relevant patient subgroups. The pattern was consistent for both younger and older patients, men and women, and throughout the range of socioeconomic status (Figure 2). Similarly, the results were consistent for patients regardless of a diagnosis of diabetes, hypertension, depression, or glaucoma. Further analyses based on CCI score, total medications, clinic contacts, emergency department visits, and prior hospital admissions also yielded findings that were consistent with the main analysis. No subgroup showed a markedly higher risk of a SCAR event following topical compared with oral CAIs. All subgroups with at least 160 total SCAR events showed findings that replicated the primary analysis. The overall pattern of results was similar for patients treated during remote years (1995 to 2007) and recent years (2008 to 2019).

The risk of a SCAR event was mostly unrelated to patient characteristics. The pattern was consistent regardless of age, sex, and socioeconomic status (Table 2). Living in a rural location was associated with a marginally lower risk in both univariable and multivariable analysis. Patients diagnosed with diabetes had a marginally lower risk in multivariable analysis. In contrast, patients with a higher CCI score or those with more clinic contact had an increased risk. Total medication use was a risk factor for a SCAR event in the univariable analysis; however, no association was apparent after adjusting for other covariates. Otherwise, individual medical diagnoses and overall utilization were not predictors of a SCAR event. Adjustment for all measured covariates suggested a 51% relative increased risk of a SCAR event for patients prescribed oral compared with topical CAIs (95% CI, 15-98; P = .003).

The distinction between oral and topical CAIs was also associated with other adverse outcomes. As expected, the absolute risk of a dermatologic event (ie, SJS, TEN) exceeded the absolute risk of aplastic anemia; moreover, the increased risk with oral treatment extended to both (Table 3). Similarly, the absolute risk of acute kidney injury was substantial and an increased risk was observed with oral compared with topical treatment. Of the 2 oral medications, methazolamide was associated with a higher relative risk than acetazolamide. The duration of initial prescription was not a predictor of an adverse event. The associations with cardiac dysrhythmias, subsequent emergency department visits, and overall hospitalizations were modest. There was no significant difference between oral and topical CAIs in the risk of total adverse drug reactions, kidney stones, or delirium.

When considering alternate time horizons in the evaluation of SCAR events between groups, there was no significant difference between comparators when considering events within 1 week or 1 month of initial prescription. A modestly higher relative risk with oral CAIs was found by 6 months and 1 year (Table 3).

This matched longitudinal cohort study explored the incidence of serious adverse reactions following prescription of oral or topical CAIs. Overall, patients prescribed an oral CAI had an absolute risk of 2.90 per 1000 patients of developing a SCAR event. This was in comparison with an absolute risk of 2.08 per 1000 patients of developing a SCAR event following prescription of a topical CAI. These findings were replicated across a diverse set of analyzed patient subgroups. We also observed a directly proportional association in the risk of a SCAR event over time, contrary to the conventional thought that SCAR events are concentrated soon after therapy initiation. A possibility is that this directly proportional association reflects patients’ underlying baseline risk of sustaining these events.

The incidence of SJS, TEN, and erythema multiforme has been estimated at 0.04 cases per 1000 person-years when considering the specific ICD-9-CM code of 695.1.²⁷ The incidence of aplastic anemia has been estimated at 0.01 to 0.03 annual cases per 1000 for adults older than 60 years when using the ICD-9 code of 284.^28,29 Taken together, this suggests an annual incidence of 0.05 to 0.07 SCAR events per thousand person-years, corresponding with a cumulative incidence of 1.2-1.8 cases per 1000 persons throughout the 25-year period of this study. This suggests that 14% to 39% and 38% to 56% of SCAR events might be attributed to prescription of a topical and oral CAI, respectively. The risk of SCAR events in the topical CAI cohort is not a reflection of the population baseline risk. Topical CAIs have been implicated in cases of SJS/TEN³⁰ and can be detectable in the blood and give rise to adverse events, especially in those with poor kidney function.³¹

Past publications on adverse reactions from oral CAIs are laboratory physiologic studies, animal studies, case reports, and small case series.^32-35 A larger investigation of 86 patients found an increased risk of adverse events in the nervous, gastrointestinal, metabolic, and kidney organ systems in patients treated with acetazolamide compared with placebo for intracranial hypertension.³⁶ Acetazolamide was associated with an acceptable safety profile and tolerance at doses up to 4 g per day. Another study on 75 patients with glaucoma reported that patients taking acetazolamide had a higher risk of fatigue, diarrhea, vomiting, electrolyte abnormalities, and tinnitus compared with patients receiving topical dorzolamide.⁶ Neither study reported any SCAR event during follow-up. Per a recent systematic review, the only adverse events of acetazolamide that have been proven to be dose dependent are dysgeusia and paresthesia.³⁷ As such, our study is the first, to our knowledge, to allow for comparisons of SCAR events, which represent important idiosyncratic adverse events following CAI therapy.

Oral CAIs have been avoided by some clinicians for patients with a sulfonamide allergy. A study by Strom and colleagues³⁸ concluded that individuals with a hypersensitivity to sulfonamide antibiotics had an association with allergic reactions to sulfonamide nonantibiotics owing to a general predisposition as opposed to cross-reactivity. In a retrospective case series of 34 patients with a documented sulfonamide allergy who received acetazolamide, only 2 (7%) experienced urticaria and none experienced a severe allergic reaction.³⁹ Sulfonamide allergies have been shown to be more often attributable to molecular structures other than the SO₂-NH₂ moiety in CAIs.⁴⁰ However, an association between topical and oral CAIs and SJS/TEN has been demonstrated, regardless of formulation.⁴¹

A strength of our analysis is that the population was heterogeneous for diagnosis, thereby improving the generalizability of the findings. Other indications for oral CAIs are an exacerbation of congestive heart failure, drug-induced edema, centrencephalic epilepsies, idiopathic intracranial hypertension, and altitude sickness.^42,43 Topical CAIs are indicated in the treatment of cystoid macular edema in retinitis pigmentosa and can reduce intraocular pressure in patients without glaucoma.⁴⁴ Our stratified analysis restricted to glaucoma (Figure 2) replicated the more inclusive primary analysis.

Limitations of our study should also be considered. Our analysis was restricted to patients older than 65 years given the available prescription data in Ontario. Given that specific ICD-9 codes were not available, the more general code of 695 was used to represent a case of SJS or TEN. A consequence of this is that the true incidence of SCAR events is likely lower than reported. In addition, certain potential confounders, the severity of adverse events, doses prescribed, specific end points and patient adherence were not available in the available databases. All adverse events were weighted in the same manner in the SCAR end point. We were unable to assess for confounding by indication whereby physicians may avoid prescribing oral CAIs to those at highest risk for SCAR events. In addition, estimates of SCAR events were not compared with the population baseline rate. Despite matching on 3 characteristics, there were differences in some baseline parameters between treatment cohorts. End points were analyzed up to 1 year of follow-up, and long-term complications remain uncertain. Given that oral CAIs are often administered for a short-term purpose, it was not possible to differentiate between treatment discontinuations owing to physician recommendation as opposed to patient intolerance. Drug reactions in those with a sulfonamide allergy and minor adverse drug reactions were not evaluated.

In closing, this population-level analysis demonstrated that oral CAIs were a safe treatment with a low incidence of serious adverse events. In comparison with topical CAIs, there was a modest difference in the risk of a SCAR event and no increased risk of a general adverse drug reaction. Oral CAIs should not be initiated casually, and appropriate informed consent discussions must be had with patients initiating these agents. At the same time, this population-level analysis supports reconsidering the reluctance toward prescribing oral CAIs, given the low risk of severe adverse events.

Accepted for Publication: December 2, 2021.

Published Online: January 27, 2022. doi:10.1001/jamaophthalmol.2021.5977

Corresponding Author: Donald A. Redelmeier, MD, MSHSR, Sunnybrook Health Sciences Centre, 2075 Bayview Ave, Room G1 51, Toronto, ON M4N 3M5, Canada (dar@ices.on.ca).

Author Contributions: Dr Redelmeier and Ms Thiruchelvam had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Popovic, Schlenker, Redelmeier.

Acquisition, analysis, or interpretation of data: Popovic, Thiruchelvam, Redelmeier.

Drafting of the manuscript: Popovic, Redelmeier.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Thiruchelvam, Redelmeier.

Obtained funding: Schlenker, Redelmeier.

Administrative, technical, or material support: Popovic, Redelmeier.

Supervision: Redelmeier.

Conflict of Interest Disclosures: Dr Popovic reported grants from the Glaucoma Research Society of Canada during the conduct of the study and the PSI Foundation outside the submitted work. Dr Schlenker reported personal fees from Alcon, Allergan, Bausch Health, Johnson & Johnson Vision, Light Matter Interaction, Théa-Labtician, and Santen outside the submitted work. No other disclosures were reported.

Funding/Support: This work was supported by a grant from the Glaucoma Research Society of Canada and a Canada Research Chair in Medical Decision Sciences. This study was supported by ICES, which is funded by the Ontario Ministry of Health and Ministry of Long-Term Care.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The analyses, conclusions, opinions, and statements expressed herein are those of the author, and not necessarily those of Canadian Institute for Health Information.

Additional Contributions: We thank IQVIA Solutions Canada Inc for use of their drug information file. Parts of this material are based on data and information compiled and provided by the Canadian Institute for Health Information.