Twenty-four double-blind RCTs were included in the review (n=9,988 patients from table 2). Jadad scores ranged from 2 to 5 of 5 points. The duration of active treatment ranged from 14 to 84 days; the duration of the run-in period ranged from none to 14 days.
The authors stated that significant heterogeneity was found (p=0.0001; I2=83%), but it was not clear which analysis this referred to.
The overall mean placebo response rate was 18.85% (SD 12.72%; range 2.94 to 47.06). The overall mean response rate for active treatment groups was 43.49% (SD 13.80%, range 21.59 to 70.22). Active treatment was associated with a significantly higher response rate than placebo (OR 3.71, 95% CI 2.78 to 4.96).
Variables associated with placebo response rates
Placebo response rates were lower (p=0.05) in trials of proton-pump inhibitor treatment (14.51%) compared with trials of H2-receptor antagonist treatment (24.69%). These findings were also suggested by the subgroup analyses (OR 2.07, 95% CI 1.66 to 2.57,10 H2-receptor antagonist RCTs; OR 5.27, 95% CI 3.77 to 7.37, 14 proton-pump inhibitor RCTs). Meta-regression appeared to suggest no significant association between type of treatment and placebo response (p=0.16).
Placebo response rates were lower (but not statistically significant, p=0.246) in patients with erosive disease (11.87%) compared with non-erosive disease (18.31%). Meta-regression also suggested no significant association (p=0.896). Placebo response rates were not significantly correlated with other variables tested.
Cumulative meta-analysis showed that the odds ratio for treatment response declined from 1983 to 1991, then increased in 2009.
There was some evidence of publication bias (Begg and Mazumder’s test p=0.094; Egger’s test p=0.041).