Fourteen randomised controlled trials (RCTs) were included in the review (n=2,149 patients), seven parallel group (n=1,882 patients) and seven-crossover (n=267 patients). Quality scores (out of 5 points) were 5 points (two RCTs), 3 (five RCTs), 2 or 1 (three RCTs each), and 0 (one RCT). Six RCTs used adequate methods of allocation concealment, nine used some form of blinding and six used intention-to treat analysis. Withdrawal rates ranged from nil to 17% (where reported).
Diurnal intraocular pressure: There was no statistically significant difference in percentage reduction in diurnal mean intraocular pressure between the two treatment groups at any period of follow-up. In subgroup analyses, latanoprost was significantly more effective for this outcome at three months than combined dorzolamide plus timolol among patients insufficiently treated with timolol alone (WMD 3.12% intraocular pressure reduction, 95% CI 0.47 to 5.78; three RCTs). Latanoprost was also significantly more effective for this outcome at three months than concomitantly administered dorzolamide and timolol (WMD 3.07% intraocular pressure reduction, 95% CI 0.64 to 5.50; one RCT).
Intraocular pressure at 10.00 am: There was no statistically significant difference between the treatment groups at any period of follow-up in percentage reduction intraocular pressure at 10.00 am. In subgroup analysis by baseline treatment, combined dorzolamide plus timolol was significantly more effective for this outcome at one month than latanoprost among patients insufficiently treated with timolol (WMD -4.14% intraocular pressure, 95% CI -5.78 to -2.50; one RCT). Also at one month, a fixed combination of dorzolamide plus timolol was significantly more effective than latanoprost (WMD -4.67% intraocular pressure, 95% CI -6.44 to -2.91; four RCTs).
Tolerability: Latanoprost was associated with significantly fewer withdrawals from treatment due to adverse events than combined dorzolamide plus timolol (RR 0.34, 95% CI 0.13 to 0.84; four RCTs). There was no statistically significant difference between the treatment groups in ocular adverse event rates (three RCTs).
There was significant statistical heterogeneity for several meta-analyses (χ2 p =0.05 to 0.007) and funnel plots for efficacy were asymmetrical: this was attributed largely to two individual trials.
Sensitivity analyses by quality did not substantially affect the results.