Thirty-one studies with 34 experimental treatment groups were included (n=1,860): 18 trials used random allocation to treatment groups, 6 used a non-random method of assignment, and 7 were of pre-post design.
The mean effect size values are for:
asthmatic attacks, 0.56 (95% confidence interval, CI: 0.40, 0.72; n=11);
dynamic respiratory volume, 0.34 (95% CI: 0.10, 0.58; n=10);
peak expiratory flow rate, 0.29 (95% CI: 0.01, 0.57; n=6);
functional status, 0.46 (95% CI: 0.09, 0.83; n=4);
adherence to treatment, 0.78 (95% CI: 0.44, 1.12; n=7);
utilisation of health care, 0.29 (95% CI: 0.13, 0.45; n=10);
use of pro re nata medications, 0.62 (95% CI: 0.26, 0.98; n=8);
psychological well-being, 0.53 (95% CI: 0.21, 0.84; n=6); and
psychomotor knowledge, 1.02 (95% CI: 0.76, 1.28; n=4).
All effect sizes were statistically significant (p<0.05) and homogeneous.
Fail-safe N values were calculated and found to be greater than 50 in all but 2 outcomes, i.e. functional status and peak expiratory flow rate. Values this large provide reasonably strong evidence against the threat to validity of publication bias.