Primary open-angle glaucoma (POAG) is a chronic condition characterised by optic neuropathy and vision loss. Elevated intraocular pressure (IOP) can damage the optic nerve and is a risk factor for glaucoma, thus treatment usually comprises topical hypotensives. This analysis aims to address methodological issues associated with the synthesis of glaucoma clinical trial data, given variations in study methodology and IOP measurement. Meta-regression was used to estimate how IOP varies over time for patients receiving treatment. Relative treatment effects were assessed using a random-effects mixed treatment comparison (MTC) in order to preserve randomisation and avoid selection bias. To produce clinically meaningful outputs, these analyses were combined to obtain the mean on-treatment IOP and the proportion of patients achieving different IOP targets at different time points. A further MTC estimated the probability of hyperaemia events. The analysis showed that after 3 months' treatment, between 58 and 83% of patients will have a > or =20% reduction in IOP and 70-93% of patients will have an absolute IOP <20 mmHg. Latanoprost and bimatoprost were found to produce significantly lower on-treatment IOP compared with timolol (p < 0.05); the difference between latanoprost and bimatoprost was not significant. Travoprost produced a lower mean IOP compared with timolol (not significant). Latanoprost-timolol was found to produce significantly lower IOP than latanoprost alone or beta-blockers. The probability of hyperaemia-type events varied between treatments from 14.8 to 63.03%. Latanoprost had significantly lower odds of hyperaemia than travoprost, bimatoprost, travoprost-timolol, or bimatoprost-timolol. This analysis suggests that latanoprost and bimatoprost produce a statistically significant reduction in IOP compared with timolol, but are associated with a higher risk of hyperaemia. Out of all the prostaglandins, latanoprost may achieve a good balance between tolerability and IOP efficacy. As with all forms of meta-analysis, the results are based on the assumption that the studies and intervention groupings are sufficiently similar to be compared.
