Twenty-two RCTs (n=1,279) were included.
Twelve of the 22 included trials used intention to treat analysis, and only one attempted to blind both the observers and participants.
Analysis of pooled data.
The pooled cure rate was 88.5% (SD 7.4) for conventional length antimicrobial therapy and 78.8% (SD 13.5) for single-dose/short duration therapy, based on all 22 studies. The pooled difference was 9.8% (SD 10.3). The cure rates, and the difference in cure rates between the two regimens, were not affected by the inclusion of studies that did not match antimicrobial agents in both treatment groups.
Meta-analysis.
When all 22 trials were pooled there was significant heterogeneity between the studies (P=0.01). A random-effects meta-analysis showed an absolute difference in the cure rate of 6.38% (95% confidence interval, CI: 1.88, 10.89), in favour of longer treatment. The NNT was 16 (95% CI: 9, 53).
There was significant heterogeneity between the 17 trials (n=832) that used the same drug in both treatment groups (P=0.01). A random-effects meta-analysis showed an absolute difference in the cure rate of 7.92% (95% CI: 2.09, 13.8), in favour of longer treatment. The NNT was 13 (95% CI: 6, 35).
There was no difference between the treatment groups in the subgroup meta-analyses of studies wherein a single-dose was compared with conventional length therapy (9 trials, n=383), or more than a single dose was compared with conventional length therapy (13 trials, n=896).
Heterogeneity between the 5 trials (n=193) that used amoxicillin in both treatment groups was not statistically significant (P=0.6). A fixed-effect meta-analysis showed an absolute difference in the cure rate of 13.0% (95% CI: 4.0, 24.0), in favour of longer treatment. The NNT was 8 (95% CI: 5, 25).
There was significant heterogeneity between the 6 trials (n=310) that used trimethoprim/sulfamethoxazole in both treatment groups (P=0.004). A random-effects meta-analysis showed no significant difference in the cure rate (6.24%, 95% CI: -3.74, 16.2).