Meta-analyses are a mixed blessing; they can be helpful and they can be misleading,^1–3 and meta-analyses of studies in asthma are no exception.⁴ They are misleading when they fail to make clear that the conclusions only pertain to patients fulfilling the inclusion and exclusion criteria for the original studies. To suggest that they either confirm what everyone knows or provide conclusions, from somewhat dubious sources with spurious precision, may be unduly cynical. Nevertheless, it appears to be difficult to identify questions where a meta-analysis is able to provide important new insights into asthma management.

Professor Beasley’s group in New Zealand has used meta-analysis in the past to look at the dose-response relation for inhaled budesonide and fluticasone.^5,6 The two meta-analyses included all studies that had assessed at least two doses of the inhaled corticosteroid in addition to placebo. The conclusion of the analyses was that, in patients with mild to moderate asthma, 80% of the maximum benefit occurs with fluticasone 70–170 μg/day (depending on the end point studied), and 90% of the maximum benefit with a dose of 100–250 μg/day.⁵ Similarly, 80% and 90% of the maximum benefit from budesonide was seen with doses of 200–400 μg/day and 300–600 μg/day, respectively.⁶ Although the mean forced expiratory volume in 1 second (FEV₁) in these studies was 66% and 69% predicted, respectively, suggesting the patients had moderately severe asthma, the patients had to be considered to be safe on placebo since the meta-analyses only included placebo controlled trials. The findings are therefore only generalisable to patients for whom it was considered safe to give no inhaled corticosteroid for several weeks, usually 8 weeks or more. That will include a fair proportion of patients in general practice but fewer in secondary care. The authors recognised this in the conclusion of their second paper, but the caveat is sometimes lost when the data are discussed.

The most recent meta-analysis from the New Zealand group by Masoli and colleagues⁷ published in this issue of Thorax is asking an important question and appears to have circumvented many of the problems outlined above. The authors examine whether, in adults with symptomatic asthma, it is better to add salmeterol to a moderately low dose of inhaled corticosteroid (fluticasone 200 μg/day or equivalent) or to increase the dose of inhaled corticosteroid. The analysis included randomised double blind trials and the main outcome measure was the number of subjects withdrawn due to asthma or who had one or more moderate or severe exacerbations. The authors found more withdrawals due to asthma and severe exacerbations in patients taking higher doses of inhaled corticosteroids than in those given salmeterol with the lower dose of inhaled corticosteroid, with odds ratios of 1.58 and 1.35 respectively. The changes in the secondary end points were in the same direction. The differences between groups are not particularly large, but are worthwhile when talking about important end points such as exacerbations.

The strengths of this meta-analysis are that it covers over 4000 patients from 12 studies and the main end point of deteriorating asthma or exacerbations is clinically important and not directly related to bronchodilator activity. The patients had moderately severe asthma with a mean FEV₁ of 64% predicted and did not have to be deemed to be safe with no inhaled corticosteroids. The question is also important since, although there is reasonable agreement that, in general, adding a long acting β agonist provides more clinical benefit than doubling the dose of an inhaled steroid, the dose of inhaled steroid at which the addition of a long acting β agonist is beneficial has not been clearly determined.⁸ This study suggests that patients taking 200 μg fluticasone or 400 μg beclometasone or budesonide are likely to fare better if they add salmeterol rather than increasing the dose of inhaled steroid. Quadrupling rather than doubling the dose of inhaled steroid might provide more benefit, as shown in the FACET study with budesonide and formoterol,⁹ but this would involve relatively high doses of inhaled steroid.

Meta-analyses nearly always focus on efficacy rather than the balance of efficacy and adverse effects which is what, of course, a clinician has to take into account before prescribing. Both inhaled corticosteroids and long acting β agonists have a good safety record. There may be individual patients, however, where a physician might be influenced by potential adverse effects—patients with osteoporosis or a tendency to develop cardiac dysrhythmias, for example.

Masoli and colleagues limited their analysis to salmeterol because few studies with formoterol fulfilled the inclusion criteria for the meta-analysis. They argue that the effect of formoterol is likely to be similar to that of salmeterol, and this is supported by the findings of the FACET study.⁹ The conclusion that patients who are inadequately controlled on a relatively low dose of inhaled corticosteroid are more likely to benefit from a long acting β agonist rather than an increased dose of inhaled corticosteroid should enable firmer advice on this point to be incorporated into guidelines.