Statistics from Altmetric.com
Our interpretation of the important data on corticosteroid responsiveness in chronic obstructive pulmonary disease (COPD) presented by Burge et al1 is quite different from that of the authors. They have shown that those with an improvement in forced expiratory volume in 1 second (FEV1) of >20% following short term treatment with prednisolone had a more significant reduction in exacerbation frequency with longer term treatment with inhaled fluticasone than those without. Thus, arguably, the only important positive clinical outcome of long term treatment of COPD with inhaled corticosteroids is largely confined to those who are identified by a positive response to prednisolone. This is a powerful validation of the suggestion in current guidelines2 that decisions on the use of long term inhaled corticosteroid therapy are based on a careful assessment of the response to short term treatment with prednisolone.
Clinicians will continue to use short term prednisolone trials in patients with unclassified, apparently fixed, airflow obstruction since this is the only practical way to identify the minority with markedly corticosteroid responsive disease. The data presented by Burge et al are important in that they show that smaller responses to prednisolone can provide additional useful clinical guidance. They also suggest that the mechanism of the bronchodilator response to prednisolone might have features in common with the mechanism of inhibition of COPD exacerbations with corticosteroids. We have recently shown that the improvement in FEV1 and quality of life scores seen with prednisolone compared with placebo in COPD increases progressively with increasing eosinophilic airway inflammation, and is associated with amelioration in eosinophilic airway inflammation as reflected by a sixfold reduction in the induced sputum eosinophil count.3 Future studies should investigate whether treatment targeting eosinophilic airway inflammation in COPD results in a reduction in exacerbations, as it has clearly been shown to do in asthma.4
We thank Dr Pavord and colleagues for their letter.1 We were asked to add the section relating the effect of prednisolone response to fluticasone related exacerbation prevention by the papers’ referees. We originally omitted this as there was no relationship between the effect of exacerbation reduction and prednisolone response when prednisolone response was expressed as a continuous variable (p = 0.84), and we were worried that the analysis based on responders and non-responders would result in misinterpretation, which unfortunately has been the case. The main conclusion from the paper is that any separation of the patients into responder and non- responder groups is statistically flawed. The response is unimodally distributed and individual changes between two visits (that is, before and after prednisolone) suffer from the regression to the mean effect. We showed good evidence that this was in fact occurring, with “responders” declining by a mean 127 ml in the 4 weeks before prednisolone, while those in whom forced expiratory volume in 1 second (FEV1) was reduced by >20% after prednisolone had a 47 ml increase in FEV1 in the preceding 4 weeks. The most likely explanation for the spurious relationship between exacerbation reduction and prednisolone response related to confounding by a low FEV1. Responders were classified as those with a 20% increase in FEV1 as a percentage of the starting value favouring a “response” in those with the lowest FEV1 (a change from 0.8 to 0.96 l in the lowest group) and exacerbations are also more frequent in those with a lower FEV1. A clinical decision made on a steroid trial is not a reliable, reproducible, or valid way of separating out “responders” and so cannot be recommended.