Seventeen RCTs were included for the review (n=1,491 patients), including four crossover designs (n=156 patients) and 13 parallel group designs (n=1,335 patients). Two trials scored 5 points in the quality assessment, four scored 4 points, seven scored 3 points, three scored 2 points and one scored 1 point. Withdrawal rates ranged from 0 to 39%.
Diurnal mean intraocular pressure reduction: Travoprost significantly reduced diurnal mean intraocular pressure compared to latanoprost at two weeks (WMD 1.47mmHg, 95% CI 0.33 to 2.62; four RCTs, n=505 patients) and two months (WMD 0.71mmHg, 95% CI 0.04 to 1.38; RCTs; n=146 patients), but not at one, three, six or twelve months. When the four RCTs with a quality score of 2 points or less were excluded, there were no significant differences between the groups. There was evidence of significant statistical heterogeneity at two weeks (I2=60.7%) and one month (I2=65.3%).
Mean intraocular pressure reduction at 9am and 5pm: There were no significant differences between travoprost and latanoprost in mean 9am intraocular pressure reduction at any times points (five RCTs). Travoprost significantly reduced the mean 5pm intraocular pressure compared to latanoprost at two weeks (WMD 0.87mmHg, 95% CI 0.40 to 1.33; three RCTs; n=726 patients) but not at one, two, three, six months or twelve months. There was no evidence of significant statistical heterogeneity except for mean 9am intraocular pressure reduction at two months (Q=18.29, p>0.001).
Adverse events: The adverse events reported were generally mild and resolved without treatment. The most common adverse events were hyperaemia (42.2% travoprost and 25.3% latanoprost) and eyelash growth (travoprost 8.0% and latanoprost 5.2%). Travoprost significantly increased the risk of hyperaemia compared to latanoprost (RR 1.72, 95% CI 1.33 to 2.23; nine RCTs; n=1,268 patients). There were no differences between treatments in the risk of any other adverse events. Three patients discontinued travoprost due to conjunctival hyperaemia.
There was no evidence of publication bias