In some countries inhaled corticosteroids are widely prescribed for patients with chronic obstructive pulmonary disease (COPD), despite the lack of good studies to support their use. In the last 12 months these three important large, parallel group, placebo controlled studies have reported at scientific meetings but, at the time of going to press with this article, they have not been published. This review will give an individual view of what has been presented, and provide a basis for the assessment of the trials when they are published.

All three studies used similar definitions of COPD and excluded patients with a clinical diagnosis of asthma or significant bronchodilator responsiveness. The Copenhagen study also excluded those with a prednisolone response, which was found in only 5% of their otherwise unselected population. The Copenhagen study started with a random population survey which identified all those with an FEV₁/VC ratio of <70%, irrespective of their FEV₁. They have the least diseased group with a mean FEV₁ of 85% predicted and include many subjects whose FEV₁ was within the normal range; indeed, only 39% had an FEV₁ of <80% predicted. The subjects in the ISOLDE study were mostly recruited from respiratory clinics and have the most

severe COPD with a mean FEV₁ of 50% predicted. The EUROSCOP group is intermediate in severity with a mean FEV₁ of 77% predicted.

The EUROSCOP subjects were all current smokers, having failed to quit in a three month period during the run in. The ISOLDE subjects had all been smokers, but only 48% were smoking at trial entry. The Copenhagen study did not have any entry criteria relating to smoking; 76% were current smokers.

The principal outcome measure for all three studies was longitudinal decline in FEV₁. It was thought that the pathology of COPD was largely irreversible, and that untreated patients with COPD deteriorate more quickly than normal, leading to premature disability and death. All three studies set out to include data over three years for each subject. It is not possible to establish individual rates of decline of FEV₁ with any certainty within this time, as the short term reproducibility of FEV₁measurements is around five times the normal annual decline in FEV₁. A reduction in FEV₁ slope can be difficult to show, as demonstrated by the Lung Health Study of smoking cessation in which subjects were followed up for five years and yet significant effects were only found with subgroup analysis.1 Although some have suggested that more than five years are required for such studies, the problem is that, even with a three year follow up, 46% of the ISOLDE subjects withdrew before the end of the study period, making longer studies with FEV₁ as an outcome difficult in the more diseased group. All three studies used the mixed effects model to estimate the FEV₁ slope with time. This is the best method available at present, but weights the estimates in favour of those reaching the end of the study, who are likely to be the least affected. The model is therefore conservative and will tend to underestimate any effect. The model was not applied as planned in the EUROSCOP and ISOLDE studies as there was a small increase in FEV₁ in the first 3–6 months, precluding a linear model using the initial data points.

No study showed an unequivocal difference in the FEV₁slope between treatment groups. The Copenhagen study showed no evidence of any difference at all between groups, whilst the EUROSCOP and ISOLDE studies both showed reductions in the FEV₁ slope which were not statistically significant when analysed in the whole study group.

The mixed effects model in each study produced estimates for FEV₁ decline that were not more than twice the predicted values for normal subjects. Those in the EUROSCOP and ISOLDE studies would not have reached their pre-trial FEV₁ if they had, at some time in their lives, had measurements close to 100%. It is therefore important to know the rate of decline in FEV₁before trial entry. The Copenhagen study has the best data, the majority of subjects having measurements taken 13 years previously. The estimates from the mixed effects model and the 13 year observations were similar. Subjects in the EUROSCOP study had a six month run in period and the FEV₁ decline in this six months was much larger than that estimated from the mixed effects model during the trial. Few, if any, subjects had been taken off inhaled corticosteroids before entry to the trial. The ISOLDE study has the greatest difficulty in estimating pre-trial decline in FEV₁. The run in period was only eight weeks, during which those withdrawn from inhaled corticosteroids declined faster than those who were steroid naïve. A tentative estimate of decline can be made from the steroid naïve subjects who were randomised to placebo. Their observed FEV₁ decline in the 5.5 months from recruitment was more than twice that estimated from the mixed effects model during the trial.

Exacerbations of COPD are related to the severity of the disease and to increasing age. They were only common in the ISOLDE group and were significantly reduced by active treatment. The Copenhagen study showed that current sputum production increased the risk of an exacerbation requiring hospital admission fivefold,2 and the ISOLDE study showed that exacerbations were increased in the eight weeks after stopping inhaled corticosteroids in the 55% taking them prior to the run in period. Exacerbations are a clinically relevant outcome with substantial costs. One other shorter study has confirmed the reduction of exacerbations with inhaled fluticasone propionate.3

Showing small changes in FEV₁ slope (or failing to show such changes) is difficult to interpret in clinical terms. Health effects measures (quality of life) are important in aiding interpretation and as an outcome in their own right. The ISOLDE study used the St George’s respiratory questionnaire and showed reduced rates of decline in the scores in each domain. The effects were linear with time, the difference between active and placebo groups increasing with time. The Copenhagen study used a less sensitive measure which showed no impairment in most of their subjects and was therefore not a useful outcome measure. The EUROSCOP study did not incorporate a health effects questionnaire.

Overall, the Copenhagen study showed no benefit from inhaled budesonide 800 μg daily (with 1.2 mg for the first six months) on any outcome measure. The EUROSCOP study showed non-significant benefit in terms of FEV₁ decline with budesonide 800 μg daily, whilst the ISOLDE study showed benefit in terms of quality of life, along with non-significant improvement in FEV₁ decline, with fluticasone propionate 1 mg daily. These differences could be due to the differences in severity of the disease, inhaled corticosteroids working best for those with the most severe disease, or it could be a dose related effect, the ISOLDE study using a significantly higher relative dose than the other studies. A meta-analysis of three previous small studies of inhaled corticosteroids in COPD suggests that beclomethasone dipropionate in a dose of 800 μg was significantly less effective than budesonide in a dose of 1.6 mg or beclomethasone dipropionate at 1.5 mg/day (this estimate was based on very small numbers), and also showed that the decline in FEV₁ was greater in those with lower starting values of FEV₁.4 It is therefore probable that the two budesonide studies were suboptimally dosed. Lack of compliance with the study inhalers is an unlikely reason for the differences since compliance was measured in each study and exceeded 80%.

Safety of relatively high doses of inhaled corticosteroids is an important issue and was best studied in the EUROSCOP trial where a significant small increase in skin bruising was seen with active treatment. No study showed an increase in fractures. Bone density was measured in a subset of subjects in the EUROSCOP trial and those on budesonide had less bone loss than those on placebo. There was also a small increase in dysphonia and oral candidiasis with active treatments.

COPD has mixed pathology, including emphysema, small airways disease, and changes in mucous glands and goblet cells. It is likely that different pathologies respond differently to inhaled corticosteroids. The studies are likely to be analysed with such subgroups; none has yet been presented. There is a large and conflicting literature on predictive factors for short term steroid response and, as yet, no known relationship between the short term effects and longitudinal decline in FEV₁. The EUROSCOP study can investigate this by relating the improvement in the first six months of treatment with subsequent decline; the ISOLDE study included an open steroid trial after randomisation and before active or placebo treatments. Help with the usefulness of short term steroid trials (or lack of it) should be available soon.

COPD is emerging from the backwaters of respiratory medicine. These three trials, when published, will provide good evidence for the place of inhaled corticosteroids in disease management and will suggest that they are unlikely to be the ideal drugs for this disease. One positive aspect of this is that it now leads us to look for alternative treatments for COPD. The three studies have produced important guidance on how such treatments could be evaluated.