Meta-analysis of the Efficacy and Safety of α2-Adrenergic Agonists, β-Adrenergic Antagonists, and Topical Carbonic Anhydrase Inhibitors With Prostaglandin Analogs FREE

Prostaglandin analogs (PGAs) are the most commonly prescribed medications to lower intraocular pressure (IOP) because of their potency and safety. However, in some eyes, the magnitude of IOP reduction with a PGA is insufficient and a second agent may be required to achieve the low target pressures that minimize the risk of progressive visual field loss or conversion to glaucoma.^1- 4 In the Ocular Hypertension Treatment Study, 39% of patients required 2 or more medications to achieve target IOP.⁴ In the Collaborative Initial Glaucoma Treatment Study, a median of 2 medications was required to achieve target IOP in the medical therapy group at 5 years (P. Lichter, MD, written communication, April 3, 2007).

α₂-Adrenergic agonists (AAs), β-adrenergic antagonists or β-blockers (BBs), and topical carbonic anhydrase inhibitors (TCAIs) are the ocular hypotensive agents most commonly used in combination with PGAs. Despite the existence of a considerable body of literature, controversy remains regarding the relative efficacy and safety of these agents when used adjunctively with PGAs. Accordingly, a systematic review and meta-analysis of relevant randomized clinical trials published through April 2009 was conducted.

The QUOROM guidelines were followed in conducting this meta-analysis.⁵ All published prospective, randomized, parallel or crossover, single- or double-masked clinical trials reporting the adjunctive ocular hypotensive efficacy of any AA (administered 2 or 3 times daily), BB (administered 1 or 2 times daily), or TCAI (administered 2 or 3 times daily) in combination with any PGA (administered 1 time daily) were identified. A comprehensive literature search was performed on April 2, 2009, using MEDLINE, Embase, and the Cochrane Controlled Trials Register. The key words used for adjunctive agents were adrenergic α-agonists, adrenergic β- antagonists, betaxolol, brimonidine, brinzolamide, carbonic anhydrase inhibitors, carteolol, dorzolamide, levobunolol, and timolol. The key words used for PGAs were bimatoprost, latanoprost, travoprost, prostaglandin analogs, and synthetic prostaglandins. Additionally, the tables of contents and lists of articles in press were manually reviewed on the Web sites of 8 journals with the most frequently identified relevant citations using the previously listed online databases on June 15, 2009.

A list of 10 predetermined criteria, summarized in Table 1, were used to select the publications included in the final meta-analysis. The 9 Delphi criteria,⁶ 5 criteria selected from van der Valk et al,⁷ and 4 additional criteria, summarized in Table 2, were used to evaluate the quality of the studies included in the meta-analysis. The quality score was defined as the percentage of the 18 criteria met.

Peak, intermediate, and trough effect were defined to occur 1 to 4, more than 4 and less than 9, and 9 to 12 hours after administration of the adjunctive agent, respectively. Mean diurnal IOP reduction was captured for studies that measured IOP at 2 or more times spanning at least 8 hours and reported the mean IOP of all times measured for both baseline while receiving PGA monotherapy and after addition of the adjunctive agent. Peak, intermediate, trough, and mean diurnal IOP reduction were recorded for each study that reported these times. All adjunctive agents were administered 1 time daily (for BB) or 2 times daily, except in 1 AA and 4 TCAI study arms. One of these comprised 15 subjects in whom dorzolamide, 2%, was administered 3 times daily.¹⁰ That study only reported data on mean diurnal IOP. The other comprised 64 subjects in whom dorzolamide, 1%, was administered 3 times daily and only reported data on IOP reduction at peak.⁸ Two other studies reported peak, trough, intermediate, and mean diurnal IOP data in a total of 35 subjects treated with a TCAI^11- 12 and 20 subjects treated with brimonidine purite, 0.1%.¹¹

Three studies reported IOP measurements at multiple peak, trough, and intermediate times, including evening and nocturnal times.^12- 13,16 The evening and nocturnal times (defined as 8 PM to 5 AM, inclusive) were excluded from the meta-analysis. In 2 studies,^12,16 2 different diurnal (daytime) peak and trough times were reported; in these studies, the single times closest to those used for peak and trough in the majority of the other studies were captured for analysis. In 1 study,¹² 2 intermediate times were reported, and the same approach was used. In 1 study, the 8 AM (trough) time was evaluated twice in one 24-hour period.¹³ The mean of these 2 measurements was used as the primary data point for trough.

The primary data point used from each study was the mean IOP reduction from baseline while receiving PGA monotherapy and its standard deviation. If the standard deviation of the reduction in IOP was not reported or could not be directly calculated, then the formula for the standard deviation of the difference, as described by van der Valk et al,⁷ was used with a correlation coefficient of 0.5. For each drug class and time, mean reductions were pooled across studies using a random-effects model to obtain the pooled mean and standard deviation.¹⁷ This method assumes that within each class of drugs, the difference between the study-specific mean and the overall class mean is a normally distributed variable with mean zero and a specified variance. To assess heterogeneity for each drug class and each time, a test was done to ascertain whether this specified variance was zero.

One-sample t tests for each class of adjunctive agent/time were used to compare the mean reduction with zero. The pooled means and standard deviations became the raw data for comparisons across the 3 classes of adjunctive agent at each time, which was also done using random-effects models. For each class and time, 3 comparisons using t tests from the random-effects model were made across the 3 classes, and a Bonferroni correction was used for these, with statistical significance indicated when P < .02 (0.05/3). Percentage of IOP reduction from baseline was calculated for each study arm. Parallel analyses were conducted using random-effects models to compare percentage of reduction in IOP with each class of adjunctive agent/time.

For the 4 clinical trials that directly compared the TCAI and AA groups and the 4 clinical trials that directly compared the TCAI and BB groups, a separate series of analyses of IOP reduction were conducted using fixed-effects models when there was no significant heterogeneity and random-effects models when significant heterogeneity was present. In these analyses, peak comparisons were based on all studies, while the other time comparisons were based on 3 studies for the TCAI/AA comparisons and on 2 studies for the TCAI/BB comparisons.

The frequencies of various adverse events as well as the frequency of discontinuation of the study drug because of an adverse event were captured from the original publications. Only adverse events and discontinuations that occurred after initiation of therapy with the adjunctive agents were captured. Adverse events that occurred only once among all study arms were not included. Abnormalities of eyelash length and color are known to occur as a result of PGA use and were excluded from this analysis. Discontinuations that occurred as a result of protocol violations or that were due to adverse events specifically stated to have been considered unassociated with the study medications were excluded from this analysis. All reported adverse events from each publication were captured and classified into 1 of 17 categories, including discontinuation of the study drug. The classification of adverse events into the various categories was performed independently with differences reconciled by 2 investigators (A.P.T. and R.M.F.). Adverse events are reported as number of events divided by the number of enrolled subjects (percentage of subjects at risk). Because the duration of exposure to the adjunctive agents varied widely from study to study, the rate of adverse events per person-month of exposure to adjunctive agent was also calculated for each drug class and each category of adverse event (number per person-months at risk). Only 1 study⁸ reported the specific time at which subjects discontinued participation in the study, and in this case, person-months of exposure was calculated using this information. For the other studies, for the purpose of calculation of person-months of exposure to study medications, subjects who discontinued participation for any reason were assumed to have continued therapy for half of the study period. Fisher exact test with Bonferroni correction (P < .02) was used to compare the percentage of subjects and the number of subjects per person-months at risk with adverse events.

After elimination of duplicate citations, 1134 potentially relevant articles were identified. A review of the titles and abstracts resulted in 44 possible articles of interest. After a complete review of these articles by 3 of us (A.P.T., R.M.F., and W.C.S.), a total of 10 articles were judged eligible for inclusion in the meta-analysis based on the predefined criteria listed in Table 1.^8- 16,18 The flowchart and reasons for exclusion of articles are shown in Figure 1.

The methods used in the 10 articles selected for inclusion in the meta-analysis are summarized in Table 3 and Table 4. The medication dosing times and IOP measurement times are listed in Table 4. There were 4, 5, and 9 articles that contributed study arms to the AA,^9,11,13,18 BB,^{8- 9,14- 16} and TCAI groups,^8- 14,16,18 respectively.

Among the 4 publications with AA study arms, 2 used brimonidine purite, 0.15%, 2 times daily (n = 105),^13,18 1 used brimonidine purite, 0.1%, 3 times daily (n = 20),¹¹ and 1 used brimonidine tartrate, 0.2%, 2 times daily (n = 16).⁹

The TCAI group includes 10 study arms from 9 publications. In 6 study arms, 277 subjects received dorzolamide (n = 84)^12- 13,16 or brinzolamide (n = 193)^9,14,18 therapy 2 times daily. In 3 study arms, 94 subjects received dorzolamide 3 times daily,^8,10,12 and in 1 study arm, 20 subjects received brinzolamide 3 times daily.¹¹

The BB group is composed of 5 articles with 6 study arms comprising 275 subjects^{8- 9,14- 16} (1 article includes 2 different BB formulations, timolol hemihydrate, 0.5%, [n = 30] and timolol maleate, 0.5%, gel-forming solution [n = 30], each administered 1 time daily).¹⁵ The other 4 BB study arms used timolol maleate, 0.5%, or carteolol, 2%, administered 2 times daily (n = 215).

Eighteen criteria were used to judge the quality of the articles included in the meta-analysis. Publications that failed to meet specific criteria are summarized in Table 2. The mean quality score was 88.9% (range, 78.8%-100%).

Of the 12 pooled data subsets (3 drug classes × 4 time categories), the tests for heterogeneity indicated that 2 of these (BB peak and TCAI peak) had P < .001 and 3 (AA peak, BB mean diurnal, and TCAI mean diurnal) had .02<P < .05. The other 7 tests were nonsignificant (P > .05), indicating homogeneity of effect. Data were pooled regardless of the test for heterogeneity. To account for heterogeneous effects, the random-effects model takes into account both the between-study and within-study variance, rather than only the within-study variance, when calculating the standard error for the pooled mean. This conservative approach allows for the pooling of data to meet the objective of comparing effects across drug classes.

All classes of adjunctive ocular hypotensive agents significantly lowered the IOP compared with baseline (while receiving PGA monotherapy) at all times evaluated (Figure 2 and Table 5). The number of pooled subjects for each medication class and time category, the magnitude of IOP reduction, and statistical comparisons are summarized in Table 5. At trough, the additional mean IOP reduction observed with the TCAIs (2.98 mm Hg) and BBs (3.12 mm Hg) was significantly greater than that observed with AAs (2.01 mm Hg; P < .001 vs BB and TCAI). There was no significant difference between TCAIs and BBs. At peak, there were no significant differences among the 3 classes of agents. At intermediate times, mean IOP reduction was greater in the TCAI group (2.96 mm Hg) and the BB group (2.97 mm Hg) than in the AA group (1.98 mm Hg; P = .003 vs TCAI and P = .008 vs BB). Mean diurnal IOP reduction from baseline was statistically similar in the 3 groups. The analyses of percentage of reduction in IOP from baseline disclosed similar findings in that the statistical conclusions were the same (data not shown).

Because of concerns about methods in some studies, the random-effects model that was used to evaluate efficacy at peak was repeated after exclusion of both arms of Maruyama and Shirato⁸ (carteolol, 2%, 2 times daily and dorzolamide, 1%, 3 times daily) as well as 2 of the 3 arms of Reis et al⁹ (brimonidine tartrate, 0.2%, 2 times daily and brinzolamide, 1%, 2 times daily). Both of these studies contributed data with regard to IOP reduction at peak only. The resultant analysis (after exclusion of these study arms) also disclosed no significant differences among the 3 classes of adjunctive agents at peak. The model was also repeated after excluding all study arms in which AA or TCAI agents were administered 3 times daily. A separate analysis was conducted in which both study arms of Maruyama and Shirato, the brimonidine and brinzolamide study arms of Reis et al, and all 3-times-daily AA or TCAI dosing arms were simultaneously excluded. In each of these analyses, there was no change with respect to the statistical significance of the comparisons among groups summarized in Table 5. Although the P values varied, the basic statistical conclusions remained unchanged.

Analyses of the 4 clinical trials that directly compared TCAI and AA efficacy at peak and the 3 that directly compared intermediate, trough, and mean diurnal efficacy demonstrated that the basic statistical conclusions were unchanged for peak, trough, and mean diurnal IOP reduction. However, for intermediate times, relying only on the 3 clinical trials in which direct comparisons were made between TCAI and AA, the difference in efficacy was not statistically significant (P = .07).

Analyses of the 4 clinical trials that directly compared TCAI and BB at peak and the 2 that directly compared intermediate, trough, and mean diurnal efficacy demonstrated that the basic statistical conclusions were unchanged for peak, trough, and intermediate IOP reduction. However, for the mean diurnal comparison, relying only on the 2 clinical trials in which direct comparisons were made between TCAI and BB, there was a significant difference in efficacy, favoring the TCAI group (P = .01).

The proportions of subjects exposed to an adjunctive agent who experienced an adverse event and the rates of adverse events (per person-month of therapy with an adjunctive agent) are summarized in Table 6. The proportion of subjects who experienced a treatment-related adverse event resulting in discontinuation from a study was 5.40%, 0.40%, and 2.34% in the AA, BB, and TCAI groups, respectively. The only statistically significant difference was between the AA and BB groups (P = .002).

The use of an AA was associated with a significantly increased risk of eye or eyelid pain or burning (P < .001 vs BB and TCAI). The use of a TCAI was associated with an increased risk of taste disturbance (P = .007 vs AA and P < .001 vs BB). Xerostomia was significantly more likely in the AA group compared with both the BB (P = .007) and TCAI (P = .002) groups. Fatigue, weakness, or dizziness were more likely to occur in the AA group (P = .007) and BB group (P = .02) compared with the TCAI group. There was no significant difference between the BB and TCAI groups when evaluating this adverse effect category based on the incidence per person-month of exposure as opposed to the percentage of subjects affected (P = .02). The threshold of significance with Bonferroni correction is P < .02.

To include only high-quality studies, prospective, parallel group or crossover, double- or observer-masked randomized clinical trials published in the peer-reviewed literature were included in this meta-analysis. Studies also had to report baseline IOP and standard deviation while receiving PGA monotherapy in the same cohort of subjects who were randomized to an adjunctive agent. Because the most common clinical indications for the use of the adjunctive agents studied are open-angle glaucoma and ocular hypertension, at least 80% of subjects in each trial were required to have these diagnoses.

The TCAIs and AAs are commonly administered 2 times daily in clinical practice despite the fact that they are only approved by the US Food and Drug Administration (FDA) for 3 times daily use. Accordingly, studies in which these agents were administered either 2 or 3 times daily were eligible for inclusion. Additionally, multiple formulations and concentrations of brimonidine are FDA approved, all of which were eligible for inclusion. There were 4 TCAI study arms in which the adjunctive agent was administered 3 times daily. One of these studies only contributed peak efficacy data and another only contributed mean diurnal efficacy data. There was 1 study arm that used 3 times daily dosing of an AA. When the relevant random-effects models were repeated after excluding all study arms that used 3 times daily dosing of an AA or TCAI, there were no changes in the basic statistical comparisons of IOP-lowering efficacy among the 3 classes of adjunctive agents.

The BBs are commonly used either 1 or 2 times daily; therefore, either dosing frequency was eligible. The dosing frequency in all of the 6 BB study arms was consistent with FDA labeling.

This systematic review and meta-analysis demonstrates that therapy with each class of ocular hypotensive agent evaluated (AA, BB, and TCAI) results in a statistically significant reduction in the IOP when added to a PGA. This holds true for the daytime trough, peak, and intermediate times as well as the mean diurnal IOP.

Although in the main analysis there were no significant differences among the 3 classes of adjunctive ocular hypotensive agents with respect to mean diurnal IOP reduction, statistically significant differences exist among the therapies for intermediate and trough times. The AA group had the least trough and intermediate IOP reduction from baseline, when added to a PGA. When percentage of reduction in IOP was analyzed, as opposed to the magnitude of IOP reduction, the basic statistical conclusions were unchanged.

Previous studies have shown that when brimonidine is administered 2 times daily, as was done in all 3 brimonidine arms in this meta-analysis, there was little IOP-lowering efficacy at trough attributable to the drug when given as monotherapy or as adjunctive therapy with timolol.^19- 23 In contrast, 3-times-daily dosing improves efficacy at the end of the dosing cycle.²⁰ The findings of diminished efficacy of adjunctive AA therapy compared with that achieved with BB or TCAI therapy at trough are consistent with previous studies on 2-times-daily dosing and cannot be extrapolated from this study to 3-times-daily dosing.

The results also showed that TCAIs were as effective as timolol when added to a PGA. Timolol monotherapy has been reported previously to reduce IOP more effectively than TCAIs.^24- 26 This suggests that BBs and TCAIs behave differently when used as monotherapy than as adjunctive therapy with a PGA. The reasons for this are unknown.

To demonstrate that the statistical analysis is robust, the random-effects model for IOP reduction at peak was repeated after excluding both arms of Maruyama and Shirato⁸ (BB at peak and TCAI at peak) and 2 of the 3 arms of Reis et al⁹ (AA at peak and TCAI at peak). Although both clinical trials met our initial criteria for inclusion in the meta-analysis, there were factors in both studies that had the potential to alter the outcomes of the meta-analysis. In the case of Maruyama and Shirato, dorzolamide, 1%, and carteolol, 2%, were used, as opposed to the usual dorzolamide, 2%, and carteolol, 1%, used in the United States and Europe. Second, although only 19% of subjects had normal-tension glaucoma, the mean baseline IOP (without any therapy) was 19.1 mm Hg.⁸ Eyes with lower baseline IOPs are known to experience a smaller magnitude of IOP reduction with ocular hypotensive therapy. In Reis et al,⁹ eyes, rather than subjects, were randomized to adjunctive therapy with brimonidine, 0.2%, brinzolamide, 1%, or timolol, 0.5%. Accordingly, it may have been possible for a subject to have been receiving timolol in one eye and another agent in the fellow eye. In light of the potential fellow-eye IOP-lowering effect with timolol, the meta-analysis was repeated after excluding the brimonidine and brinzolamide arms from this study. To clarify, the meta-analysis of peak efficacy was repeated, excluding both arms of Maruyama and Shirato and simultaneously excluding the brimonidine and brinzolamide arms of Reis et al. Both of these studies only contributed data regarding efficacy at peak. After exclusion of these study arms, the statistical significance of comparative efficacy analyses at peak were unchanged.

In another analysis, the random-effects models were repeated after excluding all study arms with 3 times daily dosing of an AA or TCAI. Again, the basic statistical conclusions remained unchanged.

A separate series of analyses that relied only on the 4 clinical trials that directly compared the TCAI and AA groups and the 4 clinical trials that directly compared the TCAI and BB groups disclosed findings similar to those of the main analysis; however, there were 2 differences. The intermediate time comparison between the TCAI and AA groups (based on 3 studies) disclosed no significant difference (P = .07) and the mean diurnal comparison between the TCAI and BB groups (based on 2 studies) did show a statistically significant difference, favoring the TCAI group (P = .01). We believe that these findings are less reliable than those of the main analysis given the small number of studies that were used in these models.

The 3 different commercially available formulations and concentrations of brimonidine (0.1%, 0.15%, and 0.2%) were pooled in the main analysis because most previously published studies have shown that they have similar IOP-lowering efficacy.^27- 29 One study, however, showed that brimonidine, 0.2%, 2 times daily was more efficacious than brimonidine purite, 0.15%, 2 times daily at trough.³⁰

The findings of this meta-analysis are mostly consistent with the conclusions of the individual studies that reported the comparative efficacy of these 3 classes of adjunctive therapies. Among the 10 clinical trials included in this meta-analysis, 7 directly compared the efficacy of the different adjunctive medication classes. Four studies compared AA vs TCAI.^9,11,13,18 One of these reported significantly greater mean diurnal, trough, and intermediate IOP reduction with the TCAI.¹⁸ One small study reported significantly greater IOP reduction with the TCAI at peak (no other times were reported).⁹ Another small study reported similar mean diurnal and trough IOP reduction and significantly greater peak and intermediate IOP reduction with the AA.¹¹ One reported no significant difference in IOP reduction between groups during the diurnal period.¹³ Four studies compared BB vs TCAI, all of which reported equal efficacy.^8,12,15,17 Only 1 study compared AA vs BB and reported greater IOP reduction at peak with BB.¹⁵

The frequency of adverse events and discontinuations were evaluated 2 ways: (1) by calculating the proportion of subjects randomized to each class of medication who experienced each category of adverse event and (2) by calculating the incidence of each category of adverse event per person-month of exposure to each class of adjunctive medication. The second approach was included because there was a broad range of duration of therapy with the adjunctive agents across studies.

The frequencies and types of specific adverse events reported in the 10 clinical trials are consistent with previous reports. There were only 4 specific categories of adverse events in which significant differences in the frequencies of occurrence among the classes of adjunctive therapies were observed. Taste disturbance was more common in the TCAI group, and xerostomia and eye or eyelid pain, burning, or discomfort were more common in the AA group. Fatigue, weakness, or dizziness was more common in the AA and BB groups compared with the TCAI group.

Although systematic reviews and meta-analyses are considered the strongest form of evidence, there are several limitations inherent in the process of conducting such studies. In this meta-analysis, the characteristics of the study subjects and the specific methods used in each of the studies varied, necessitating compromises in the pooling of data. For example, in some studies, peak IOP-lowering efficacy was assessed at 10 PM (2 hours after the adjunctive therapy dose), whereas in others it was assessed 4 hours after dosing, yet these were pooled into the single category of peak effect. Furthermore, in the main analysis, the statistical conclusions on peak efficacy are more reliable than those on trough, intermediate, and mean diurnal efficacy because the former is based on a larger number of studies and a larger number of total study subjects.

The evaluation of adverse events in a meta-analysis is limited because of the variable methods of data collection (physician observed vs patient reported vs patient queried). The reporting of adverse events was not required in the inclusion criteria or evaluated in the quality assessment. Additionally, adverse events were evaluated as a function of adjunctive drug class, regardless of concentration and formulation. The vast majority (89%) of subjects who received an AA received brimonidine purite, 0.15% or 0.1%. In the study that used brimonidine, 0.2%, there were no adverse events; however, direct comparisons of different formulations of brimonidine demonstrate that the incidence of adverse events is similar^27- 28 or lower^29- 30 in patients receiving the formulations with the lower concentrations. A variety of β-blockers were used in the included studies, but these were not analyzed separately with respect to adverse events. Two different TCAIs were used in the included studies, and similarly, these were not analyzed separately.

This meta-analysis disclosed that peak and mean diurnal IOP-lowering efficacy of adjunctive AA, BB, and TCAI therapy in combination with a PGA are similar. At trough and intermediate times, the BB and TCAI agents were more effective than AA agents. Although statistically significant, the long-term clinical importance of these differences is uncertain.

Correspondence: Angelo P. Tanna, MD, 645 N Michigan Ave, Ste 440, Chicago, IL 60611 (atanna@northwestern.edu).

Submitted for Publication: August 1, 2009; final revision received November 25, 2009; accepted December 22, 2009.

Author Contributions: Dr Tanna had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: Dr Tanna has received consultant, research, and educational grant support from Alcon Laboratories and Allergan Inc and has been a consultant for Apotex Inc. Dr Stewart has been on the speaker's bureau for Alcon Laboratories, Vistakon, Merck & Co, and Pfizer Inc, received research support from Alcon Laboratories, Merck & Co, and Pfizer Inc, and been a consultant for Vistakon and Merck & Co. Dr Feldman has been a consultant for Alcon Laboratories, Pfizer Inc, and Merck & Co and received research grants from Pfizer Inc, Alcon Laboratories, Allergan Inc, and Merck & Co.

Funding/Support: This work was supported by a grant from Alcon Laboratories and by an unrestricted grant from Research to Prevent Blindness.