Most trials do not follow patients for the outcome of exacerbations after drug discontinuation (not ITT) |
Loss to follow-up frequent and related to both treatment received and outcome, causing bias |
Exacerbations variably defined |
Differences between groups possibly exaggerated by withdrawal of ICS in placebo and LABA groups |
Incorrect calculation of NNT |
Incorrect adjustment for heterogeneity of number of exacerbations between patients in early studies |