Based on clinical evidence which has serious limitations (see 8.3.1.1)

Prostaglandin analogues are more cost-effective for this group (see Table 8-92). This comparison refers to those patients for whom Prostaglandin analogues are contraindicated.

Method of randomisation is not reported.

Not clear who was masked to treatment.

There were too few patients in the study to show a clear estimate of effect.

Very small sample size.

The same eye could be included in more than one group when the treatment was change.

Not statistically significant.

See also clinical evidence (Table 8-84).

Partially based on clinical evidence which has serious limitations (see 8.3.1.1)

Significant heterogeneity found in overall result. No specific cause for heterogeneity identified.

The confidence intervals are wide making the estimate of harm uncertain.

Limited time horizon (5 years).

Clinical outcomes were not derived from a systematic search.

Possible underestimation in the utilisation of ophthalmologic resources.

Patients had advanced COAG. Discount of costs was 5%

Only one study reported method of randomisation, neither mentioned allocation concealment.

Patients were not masked to treatment although observers were.

Some heterogeneity in the result with one study showing a greater than 2mmHg difference in mean change in IOP from baseline with prostaglandins and the other showing less than 2mmHg. This could be due to the different follow up periods (one study - 12 months, the other - 6 months).

Method of randomisation is not reported and there is no mention of allocation concealment.

Not reported how patients were randomised or if there was allocation concealment.

Not reported whether the clinicians and observers were masked to treatment.

Outcomes not reported properly. One study⁹² does not report the standard deviations associated with the mean reductions, nor the IOP at the end of the study.

Not enough data provided to calculate the pooled weighted mean difference. Beta-blockers were significantly better than carbonic anhydrase inhibitors in both studies. In one⁹² the difference was 2mmHg (confidence intervals not available), in the other 1.3mmHg (0.38, 2.22)¹⁴⁵.

An absolute effect calculation is not possible as there are no events in the control arm of the study.

The reporting of the methods within the studies was poor and the studies were not placebo controlled.

The wide confidence intervals make the estimate of effect imprecise

The method of randomisation was not reported. There was no mention of allocation concealment.

Neither patients nor observers were masked to treatment.

An absolute effect calculation is not possible as there are no events in the control arm of the study.

Very small sample size.

The same eye could be included in more than one group when the treatment was change.

Additional patients whose IOP is controlled with no severe adverse events

Not statistically significant

Method of randomisation is not described and there is no mention of allocation concealment.

The studies do not provide standard deviations for IOP change from baseline and although visual field testing results are reported they are not valid as miotics constrict the pupil..

One study¹⁴¹ was very old.

Unable to provide a pooled estimate. The mean change in IOP from baseline between arms is similar suggesting no difference between miotics and beta-blockers.

Very small sample size.

The same eye could be included in more than one group when the treatment was changed.

Not significant

See also clinical evidence (7.3.2.1)

The study does not describe the method of randomisation nor whether there was allocation concealment.

Only assessors of IOP measurements were masked to treatment.

The confidence intervals are broad making the effect size imprecise.

An absolute effect calculation is not possible as there are no events in the control arm of the study.

One study did not report the method of randomisation

Allocation concealment was not reported

There is significant unexplained statistical heterogeneity within the results. In one study the fixed combination is statistically more effective than prostaglandin analogues in reducing IOP[HIGGINBOTHAM2002A}, in the other there is no statistical difference and the point estimate favours prostaglandin analogues¹¹⁶.

The confidence intervals are broad making the effect size imprecise.

One study did not report the method of randomisation.

Allocation concealment was not reported.

There is significant unexplained statistical heterogeneity within the results.

In one study the fixed combination is statistically and clinically more effective than beta-blockers in reducing IOP⁶¹, in the other there is no statistical difference¹¹⁶. The confidence intervals do not overlap.

The confidence intervals are broad making the effect size imprecise.

Outcomes are not reported properly. Mean diurnal IOP pressures are not reported. Standard deviations for each mean are not reported.

Method of randomisation is not mentioned.

Allocation concealment is not mentioned.

Masked outcome assessment was not mentioned in one study¹¹⁷

Serious statistical heterogeneity was observed between studies which may have been due to different dosages of CAI applied. One study¹²¹ applied CAI at a dosage of 3/day rather than the recommended 2/day for use alongside a beta-blocker.

The confidence intervals are broad making the effect size imprecise.

Only one study reports the method of randomisation. This study has a 90% weighting on the estimate of effect.

Allocation concealment is not mentioned in either study.

Only observers were masked to treatment.

The confidence intervals are broad making the effect size imprecise.

Method of randomisation is not reported.

An absolute effect calculation is not possible as there are no events in the control arm of the study.

Not based on RCT clinical evidence.

Short follow-up.

Small sample size

Not UK cost figures.

Patients were previously prescribed a topical beta-blocker as monotherapy.

Second-line treatment

Not statistically significant.

Significance not reported.

Outcomes not reported properly. Mean diurnal IOP pressures are not reported. Standard deviations for each mean are not reported.

Only 77% of those randomised were included in the analysis.

Not based on RCT clinical evidence.

Short follow-up.

Not UK cost figures.

Second-line treatment

Significance not reported.

Studies are supplemented by data from the Cochrane systematic reviews Rolim 2007¹²⁴.

Randomisation and allocation concealment are adequate but masking of outcome assessment is not reported.

Wide confidence interval making estimate of effect uncertain.

All patients were maintained on current IOP lowering medications throughout study and some patients previously received ALT treatment.

Studies are supplemented by data from the Cochrane systematic review Rolim 2007¹²⁴.

Allocation concealment and randomisation methods are not reported in one study⁴⁵ and masking of outcome assessment is not reported in any of the studies.

One study¹⁰⁴ included 51% OHT patients.

Wide confidence interval making estimate of effect uncertain.

Although there was no statistical heterogeneity observed other differences between studies were noted in length of follow up, IOP failure criteria, laser modality, laser degrees of treatment, class of medications, mean baseline IOP and COAG population (previously untreated or treated). One study¹⁰⁴ tested different in laser degrees of treatment against prostaglandin analogues. For the purposes of comparison the 360 degree was selected.

Studies are supplemented by data from the Cochrane systematic reviews Rolim 2007¹²⁴.

Allocation concealment, randomisation methods and masking of outcome assessment are not reported in one study¹⁰².

I-squared value of 81% indicates high statistical heterogeneity which may have been due to the studies being from very different populations. One study¹⁰² is exclusively in Afro-Caribbean patients. Variations between studies are also noted in laser degrees of treatment and mean baseline IOP.

Wide confidence interval making estimate of effect uncertain.

Studies are supplemented by data from the Cochrane systematic reviews Rolim 2007¹²⁴.

One study⁹⁸ does not report masking of outcome assessment.

Although there is no statistical heterogeneity observed at 0 – 6 months follow up, the I-squared value is high (51%) for 3 – 24 months follow up.

Differences between studies are noted in IOP failure criteria, laser degrees of treatment and mean baseline IOP.

Studies are supplemented by data from the Cochrane systematic review Burr 2004¹⁵.

Randomisation and allocation concealment are adequate for all studies but masking of outcome assessment is not attempted. Attrition bias is noted for 2 studies^65,98 where treatment failures are excluded from the analysis.

Statistically significant heterogeneity possibly due to differences in types of medications, classification methods for visual field changes and length of follow up.

For visual field progression in the medium term and IOP failure at 12 months wide confidence intervals make estimate of effect uncertain. For mean change in IOP from baseline in the medium and long term the lower confidence interval is clinically insignificant.

Other differences in study populations are noted in baseline IOP, severity of COAG and race.

Figures taken from the systematic review¹⁵. Data not provided for individual studies consequently no forest plot is provided in this guideline’s appendices.

Based on the RCT Jay1988⁶⁵ – see clinical evidence in 8.6.1.1.

Not a full economic evaluation.

Average length of stay after surgery was 7.6 days and therefore longer than the current average.

In bilateral COAG patients.

Studies are supplemented by data from the Cochrane systematic reviews Wilkins 2005¹⁶¹ and Wormald 2001¹⁶².

For the antimetabolite MMC: 3 studies do not report details of randomisation method^26,123,147. 3 studies do not report details of allocation concealment^94,118,147. 3 studies do not report masking of outcome assessment^26,118,147. Only 2 studies were placebo controlled^26,147. For the antimetabolite 5-FU: 2 studies do not report details of randomisation method^39,113. 3 studies do not report details of allocation concealment, masking of outcome assessment and are not placebo controlled^39,49,113. One study⁸⁸ is a placebo controlled double blind design.

Wide confidence intervals making estimate of effect uncertain.

Although there is no statistical heterogeneity observed other differences between studies are noted in type of antimetabolite (MMC or 5-FU) used and dosage, delivery method of 5-FU (intraoperative or postoperative injections), IOP failure criteria, length of follow up, reporting of complications, proportion of patients with closed-angle glaucoma of <50%, mean baseline IOP and whether patients received previous laser treatment. One study³⁹ is exclusively in Afro-Caribbean patients and one study¹²³ is exclusively in patients from the Indian sub-continent.

One study¹³⁸ reports adequate randomisation methods but neither study reports allocation concealment. Masking of outcome assessment is only performed in one study¹⁶⁵.

Wide confidence intervals make estimate of effect uncertain.

Although there no statistical heterogeneity is observed other differences between studies are noted in antimetabolite dosage, delivery method of 5-FU (intraoperative or postoperative injections), IOP failure criteria, length of follow up, reporting of complications and mean baseline IOP. One study¹³⁸ was exclusively in Afro-Caribbean patients.

Randomisation method, allocation concealment and masking of outcome assessment are not reported.

Confidence intervals are wide making estimate of effect uncertain.

Only 3 studies report adequate randomisation methods^22,77,164 and only 2 studies report allocation concealment^19,164. Only 2 studies report masking of outcome assessment^20,22, but all studies report low or zero dropout rates.

Some statistical heterogeneity is noted in mean change in IOP from baseline at 6 and 12 months which is not satisfactorily explained by subgroup analysis for type of non-penetrating surgery, use of augmentation or presence of PXF in population.

For mean change in IOP from baseline from baseline at 6 and 12 months the lower confidence interval is clinically insignificant. For complications: wound leak wide confidence intervals make estimate of effect uncertain.

Other differences between studies are noted in non-penetrating surgery type (viscocanalostomy or deep sclerectomy with or without implant); use of augmentation; study design where 3 studies^20,77,164 randomised fellow eyes to treatment; IOP failure criteria; length of follow up from 6 months to 2 years; reporting of complications and mean baseline IOP. 5 studies^{19,22,40,90,164} included a proportion of patients diagnosed with PXF and one study¹⁶⁴ included some CACG patients but <50%.

One study⁴⁰ included 3 arms, viscocanalostomy, deep sclerectomy and trabeculectomy. The data for trabeculectomy is added twice meaning there is some double counting. The overall effect to the weighted mean difference is around 0.1mmHg.