Br J Ophthalmol 92:7-12 doi:10.1136/bjo.2007.123737
  • Review

The efficacy and harm of prostaglandin analogues for IOP reduction in glaucoma patients compared to dorzolamide and brimonidine: a systematic review

  1. W G Hodge1,
  2. J Lachaine2,
  3. I Steffensen2,
  4. C Murray2,
  5. D Barnes2,
  6. V Foerster2,
  7. T Ducruet2,
  8. A Morrison3
  1. 1
    University of Ottawa Eye Institute, Ottawa, Canada
  2. 2
    Meta-Business Advantage Ltd, Kanata, Canada
  3. 3
    Canadian Agency for Drugs and Technologies in Health (CADTH), Ottawa, Canada
  1. W G Hodge, University of Ottawa Eye Institute, The Ottawa Hospital General Campus, 501 Smyth Road, Tower III, Ottawa Ontario, K1H 8L6, Canada; whodge{at}
  • Accepted 21 June 2007


Aim: To systematically review the literature on the efficacy and harm of prostaglandin analogues (PGAs) compared to brimonidine and dorzolamide in treating elevated intraocular pressure (IOP).

Methods: Keywords were searched in major literature databases to identify relevant randomised clinical trials (RCTs) of PGAs for ophthalmic use. The study quality of RCTs was assessed using the Jadad scale. Outcomes assessed included reduction in IOP in individual patients, adverse events (AEs) and withdrawals due to AEs.

Results: Eight unique RCTs evaluating a total of 1,722 individuals were included in this systematic review. Analysis did not show a significant reduction in the mean IOP from patients receiving latanoprost compared with those receiving brimonidine (WMD = −1.04; p = 0.30). On the other hand, the latanoprost group showed a significant reduction in mean IOP compared to the dorzolamide group (WMD = −2.64; p<0.00001). The number of ocular AEs (excluding hyperaemia) was significantly higher in the brimonidine group compared with the latanoprost group (RR = 0.66; p = 0.0005).

Conclusion: Latanoprost was found to be significantly superior to dorzolamide but not brimonidine. However, ocular adverse events were significantly fewer in latanoprost users than in brimonide users. Neither travoprost nor bimatoprost was compared to dorzolamide or brimonidine in the present literature.

Glaucoma is a chronic ocular disorder characterised by the slow, progressive degeneration of the optic nerve leading to total and irreversible blindness over time. In industrialised countries, glaucoma is one of the most common causes of blindness.1

Although not always associated with glaucoma, high intraocular pressure (IOP) is recognised as the most important risk factor contributing to the development and progression of glaucoma.2 Open-angle glaucoma (OAG) is thought to be caused by impaired outflow of aqueous humour, increasing IOP, although the exact mechanism underlying this increase in resistance to flow is not completely understood.3 Glaucoma treatment is aimed at lowering IOP to preserve visual field and vision, and even a 1 mm Hg change in IOP has been associated with clinically significant differences.47 Recently, Maier and colleagues reported results of a meta-analysis confirming that lowering of IOP in patients with ocular hypertension or manifest glaucoma reduced the risk of visual field loss in the long term.8

Several approaches have been advocated for setting appropriate IOP treatment goals. The American Academy of Ophthalmology, for example, currently recommends lowering of at least 20% from the baseline IOP,9 whereas The European Glaucoma Society recommends lowering of at least 30% from baseline.10 An IOP of 18 mm Hg has been recommended as an upper IOP limit, based on results of Advanced Glaucoma Intervention Study (AGIS), which showed that patients with IOP always less than 18 mm Hg had minimal visual field progression.11

Pharmacotherapy is usually the first line of treatment for elevated IOP and chronic OAG. There are currently five major classes of drugs used to manage glaucoma and elevated IOP (table 1).2 12

Table 1 Intraocular-pressure-lowering agents currently available

In general, beta-adrenergic antagonists and carbonic anhydrase inhibitors reduce IOP by decreasing production of aqueous humour, whereas the cholinergics increase trabecular outflow.3 Alpha2-adrenergic agents decrease IOP by both decreasing aqueous production and increasing uveoscleral outflow.3 Similarly, the PGAs increase uveoscleral outflow.3 PGAs are lipophylic agents derived from arachidonic acid and lower IOP by increasing outflow of aqueous humour, primarily through the uveoscleral pathway. Although their exact mechanism of action is still not clear, they are thought to exert their effect by binding to the prostanoid FP receptors of the ciliary body, upregulating production of various metalloproteinases.12 Metalloproteinases are zinc-dependent endoproteinases that are involved in both normal and pathogenic remodelling of the extracellular matrix. PGAs are thought to induce metalloproteinase remodelling rendering the extracelular matrix more permeable to aqueous humour, which leads to enhanced outflow of aqueous humour and reduction of IOP.13 PGAs have also been shown to relax the ciliary muscle, further facilitating uveoscleral outflow.14

This review was prepared by the Canadian agency for drugs and technology in health in conjunction with the University of Ottawa Eye Institute as part of a more comprehensive report. The review question to be addressed in this paper is: what are the efficacy and harm of prostaglandin analogues compared to brimonidine and dorzolamide for lowering intraocular pressure in patients with glaucoma or ocular hypertension?


Literature search

Published literature was obtained by cross-searching MEDLINE® (1951–present), BIOSIS Previews® (1969–present), ToxFile (1964–present) and EMBASE® (1974–present) databases, with no language restrictions, through the Dialog interface. A broad search strategy with appropriate descriptors and keywords was used, in combination with a filter to restrict results to controlled trials, meta-analyses and systematic reviews. A parallel search on PubMed and the Cochrane database was also conducted. In addition, a number of trial registries were searched. Searches were not restricted by publication date, but were restricted by language (ie, reports written in languages other than English were not included).15 16 The original search was performed in November 2005. Regular alerts were established on MEDLINE®, BIOSIS Previews®, ToxFile and EMBASE® databases to capture new studies until 3 October 2006 and updated searches on the Cochrane databases regularly. The last Cochrane updates for this report were performed on issue 3, 2006.

Grey literature was obtained by searching the websites of regulatory agencies, health technology assessment and near-technology assessment agencies. Specialised databases such as the University of York NHS Centre for Reviews and Dissemination and the Latin American and Caribbean Center on Health Sciences Information (LILACS) were also searched. The internet was searched using the Google™ and Yahoo!® search engines to obtain information. Websites of professional associations such as the Association of Research in Vision and Ophthalmology (ARVO), American Optometric Association, the International Glaucoma Association, the Canadian Glaucoma Society and the Association of International Glaucoma Societies and their associated conference sites were searched for additional information. Trial registries were searched for completed and ongoing trials. Manufacturers of relevant pharmaceutical agents were also contacted for additional materials.

Selection method

Selected studies included published and unpublished RCTs, and RCTs published in abstract form. However, reports of RCTs that were deemed to be confidential, and that were not publicly available, were not included. As well, reports written in languages other than English were not included.10 12 RCTs using both a parallel group and crossover design were included; however, for crossover studies, only data from the first treatment period were extracted to avoid confounding from possible carry-over effects. Two reviewers (IS and CM) independently screened the clinical search results according to established selection criteria (table 2).

Table 2 Inclusion and exclusion criteria for literature screening

Potentially relevant bibliographic records were imported into a reference database, with duplicates removed manually. The relevant references were exported to SRS 3.0 (TrialStat Corporation, Ottawa, Ontario Canada), a web-based data-management software system for systematic reviews. The software presented each record along with selection criteria, recording and comparing the reviewers’ ratings and determining the record’s eligibility in accordance with selection criteria. At Level 1 screening, the two reviewers (IS and CM) independently screened the title and abstract for relevance. If both reviewers agreed that the citation title or abstract met all the criteria, or if there was uncertainty or disagreement between reviewers, the publication was obtained in full text. The reviewers then proceeded to Level 2 screening and applied the selection criteria to the full text articles to select the relevant articles to be included in the review. Disagreements were resolved through consensus.

Data extraction

One reviewer (VF) independently extracted data from the studies, and then completed evidence tables were independently checked for accuracy by a second reviewer (DB). Data regarding study design, patient population, interventions, comparators, results, harm and adverse events (AEs), and limitations were entered into the evidence tables. The primary outcome of interest was reduction in IOP in individual patients, at a minimum of 3 months follow-up. Secondary outcomes investigated (at a minimum of 3 months follow-up) included AEs, slow/stop in the progression of visual field defects, slow/stop in the increase in cupping of the optic nerve, lack of response or loss of therapeutic effect, and withdrawals due to AEs.

Study quality assessment

The quality of each included trial was independently assessed by two reviewers (VF and DB) using the Jadad scale.17 The Jadad scale assesses the reporting of randomisation, double blinding, and the inclusion of data for withdrawals and dropouts. Total scores ranged from 0 to 5, with a score less than 3 indicating low quality. Concealment of allocation to treatment was also categorised by the reviewer as adequate, inadequate or unclear.18

Data analysis

Analyses were performed using the RevMan software (release 4.2.5) provided by the Cochrane collaboration group and SAS statistical software (release 9.1.3, SAS Institute, Cary, NC). The software was used to compute statistics and generate forest plots to compare outcomes in the different treatment arms.

For continuous data, the weighted mean difference (WMD) and 95% confidence interval (CI) were calculated. A chi-square test (using n–1 degrees of freedom and a p value <0.05) and an I2 test were used to test for statistical heterogeneity between studies. If significant heterogeneity was detected, it was investigated using sensitivity analyses based on trial characteristics; if significant heterogeneity remained, or if the I2 value indicated moderate to high heterogeneity (ie, >25%),19 the data were summarised using a random effects analysis. Test for overall effect, studies’ sample sizes and per cent weights were also presented. Studies were grouped according to the length of follow-up. Sensitivity analyses were undertaken to evaluate the effect of study quality and intent-to-treat (ITT) analysis on the primary outcome measure. The ITT population was defined as patients who received at least one treatment and had at least one evaluation following treatment. A descriptive summary was provided for data that were unsuitable for pooling.



Of the 195 potentially relevant reports identified for retrieval, seven reports2023 of nine unique randomised controlled trials (RCTs), reporting on 1131 individuals, were included in this systematic review (fig 1). The range of mean ages was 56 to 67 years. Of the data available, 78% were white, and 47% were male. Of the studies reporting on diagnosis, of the 1033 subjects, 418 had POAG, 555 had isolated ocular hypertension, two had pigmentary glaucoma, and the rest were a mixture of exfoliation glaucoma and mixed conditions. Kampik et al24 studied latanoprost as a second-line therapy thus excluded in the meta-analysis for IOP reduction.

Figure 1 Report screening and selection procedure.

Of the eight included RCTs, all were reported in peer-reviewed journals. Camras and Sheu21 was previously reported as a conference abstract.25 The Kappa score was 0.90, indicating a good level of agreement between reviewers. Two of the seven studies (37.5%) were of a high quality (Jadad score = 5);21 22 the rest all received a Jadad score of 2. Details of each study, including study design, participant characteristics, interventions, outcomes and funding source(s), are summarised in table 3. All of the studies were of parallel design. All eight trials assessed the change in IOP from baseline, and diurnal IOP (at least two measurements, one in the morning and one in the evening) was the main outcome in six trials. All studies reported that IOP was measured using Goldmann applanation tonometry.

Table 3 Characteristics of included studies

Reduction in IOP

Meta-analyses comparing data on PGAs and comparators on the reduction of IOP are summarised in table 4 and corresponding forest plots.

Table 4 Summary of meta-analysis: IOP outcome

Latanoprost versus brimonidine

A meta-analysis of data from three studies with 471 patients did not show a significant reduction in the mean IOP at 3 months post-treatment in the latanoprost group compared with the brimonidine group (fig 2).2123 This result did not change when only higher-quality studies were pooled for the meta-analysis (p = 0.41).21 22 A sensitivity analysis for ITT was not performed, since only one study used an ITT analysis.23

Figure 2 Pooled 3-month data for latanoprost versus brimonidine: reduction in IOP.

Latanoprost versus dorzolamide

A meta-analysis of data from three studies with 328 patients showed a significant reduction in the mean IOP at 3 months post-treatment in the latanoprost group compared with the dorzolamide group (fig 3).20 26 27 A sensitivity analysis for quality was not performed, since all studies were of a lower quality (Jadad score = 2). A sensitivity analysis for ITT was not performed, since only one study used an ITT analysis.26

Figure 3 Pooled 3-month data for latanoprost versus dorzolamide: reduction in IOP.

Adverse events

Table 5 summarises the differences in occurrences of serious AEs among treatment groups. Pooled results from three studies of 803 patients showed a significant increase in the number of ocular AEs (excluding hyperaemia) in the brimonidine group compared with the latanoprost group (fig 4).21 22 24 The NNH was 9 (6, 20) relative to latanoprost. There were no other statistically significantly differences between PGAs and brimonidine or dorzolamide with regard to withdrawals due to AEs and incidences of hyperaemia.

Figure 4 Pooled data for latanoprost vs brimonidine: all ocular AEs excluding hyperaemia.
Table 5 Summary of meta-analyses of harms data

Second-line therapy

Only one study examined a PGA in patients who could not use, or had failed, therapy with an IOP-lowering agent.24 This study compared latanoprost with brimonidine in patients with glaucoma or ocular hypertension inadequately controlled with monotherapy or dual therapy. Of the 375 patients included in the ITT analysis, 76% of the latanoprost-treated and 53% of the brimonidine-treated patients obtained a mean IOP reduction of at least 20% from baseline to month 6 of treatment (p = 0.013). The mean IOP reduction difference at 6 months was 1.9 mm Hg in favour of latanoprost (p<0.001).


With the exception of trials comparing latanoprost with brimonidine or dorzolamide, there was a significant lack of trials comparing the other IOP-lowering agents. This, combined with the short-term nature of the clinical data, is a major limitation for decision makers in this therapeutic area. Although long-term studies are lacking, current studies suggest that the PGAs are well tolerated. Of all three PGAs currently available on the market as IOP-lowering agents, only latanoprost was studied and compared to brimonidine and dorzolamide; neither travoprost nor bimatoprost was used.

Latanoprost was found to be significantly superior to dorzolamide, although the pooled data were from poor-quality studies (Jadad = 2).20 26 27 Latanoprost was not found to reduce IOP significantly compared with brimonidine; however, chi-square and I2 analyses demonstrated significant heterogeneity between the three pooled studies.2123 Underlying this heterogeneity could be the fact that only one of the three studies used diurnal IOP.21 Although one of the three studies was of poor quality,23 removing this study from the pooled analysis did not alter the results. The additional IOP reduction observed with the PGAs ranged from a mean of 0.92 mm Hg to 2.64 mm Hg over dorzolamide. Given that even a 1 mm Hg change in IOP has been associated with differences in glaucomatous damage,47 the small differences in IOP lowering achieved with the PGAs compared with dorzolamide may be clinically significant. The Early Manifest Glaucoma Trial, for example, demonstrated a 10% decrease in the risk of progression with each mm Hg reduction in IOP from baseline.6

Excluding hyperaemia, ocular AEs occurred less frequently in patients treated with latanaprost compared with those treated with brimonidine. Although ocular hyperaemia associated with ocular hypotensive agents has been shown to significantly impact patient satisfaction,28 few patients discontinue therapy because of intolerance to ocular hyperaemia.29 While evidence suggests that PGA induced-iris pigmentation is a purely cosmetic, although probably irreversible, side effect, ongoing surveillance is required to further understand long-term consequences.

There are some limitations observed in the reviewed studies. First, it is well established that IOP fluctuates during the course of a 24-h period, and a single IOP measurement may or may not reflect the IOP at other times of the day or night.30 Two of the eight trials reviewed indicated that only one IOP measurement was obtained per visit (usually in the morning). In addition, although all studies compared the absolute change in IOP from baseline, a clinically significant response, defined as either reaching a target IOP or achieving a ⩾20% reduction in IOP from baseline, was only evaluated in four studies.21 22 24 31 Finally, the short-term nature of the clinical data, combined with the use of a surrogate marker, makes it difficult to extrapolate results from these studies to long-term clinical outcomes. Although IOP is a well-established surrogate marker of glaucomatous damage,8 the overall goal of glaucoma therapy is to preserve vision, and none of the studies were of long enough duration to detect changes in either the optic nerve or the visual field. Thus, more long-term studies evaluating the impact of these agents on clinical outcomes are warranted.

With regard to applicability of the results of this review, the patient population evaluated was predominantly Caucasian, and is not representative of the glaucoma cohort, given that the prevalence of POAG is estimated to be six times more common in blacks than whites in the same age group.32 Furthermore, different racial groups may have differential responses to the various IOP-lowering agents, and so the findings of this study may not be applicable to these subgroups.3335 The patient population in this study was a mixed sample of open-angle glaucoma patients and those with isolated ocular hypertension. There were insufficient data to perform a subgroup analysis on these different disease cohorts, which may differ in disease progression and patient compliance.36 The majority of studies evaluated a mixed population of patients who were treatment-naïve, had received previous treatment with an IOP-lowering agent and had failed prior therapy. There were insufficient data to pool results from these individual patient subgroups. Few studies evaluated only patients who were receiving IOP-lowering agents as first-line therapy; however, it was not appropriate to pool data from these studies. Only one study evaluated patients who had failed prior therapy.


  • Competing interests: None.