NPC Archive Item: Post-hoc TORCH re-analysis doesn’t light a beacon for ICS in COPD

NOTE – This is an archive post from the NPC and has not been updated since first publication. Therefore, some hyperlinks may no longer be working.
MeReC Rapid Review NPC Logo

Celli B et al.  Effect of pharmacotherapy on rate of decline of lung function in chronic obstructive pulmonary disease.  Am J Respir Crit Care Med 2008; 178: 332-338

The blog below gives an example of a paper that does not add to the evidence-base on the management of COPD, even though reading the paper’s conclusion and commentary may suggest otherwise. We outline briefly the reasons why it does not justify a change to current guidance.

What does current guidance suggest?
Current NICE guidance recommends adding an inhaled corticosteroid (ICS) to treatment with a long-acting bronchodilator (salmeterol, formoterol or tiotropium) only in patients with moderate or severe COPD (FEV1 < 50% predicted) who have had two or more exacerbations requiring treatment with antibiotics or oral corticosteroids in a 12-month period.  Adding an ICS could also be considered in patients with moderate or severe COPD who are still breathless despite monotherapy with a long-acting beta-agonist, but the ICS should be discontinued if there is no benefit after four weeks.

What does this study say?
This post-hoc analysis (i.e. it was not part of the original trial design) of the TORCH study showed no significant difference in rate of decline in FEV1 when fluticasone was added to salmeterol. In the placebo group, the average decline in FEV1 was 55 ml/year (in people without COPD the annual decline in FEV1 due to aging is about 30 ml/year). Active therapy reduced the annual decline in FEV1 by 13-16 ml, compared with placebo (P<0.003).  However, decline in FEV1 in the salmeterol and fluticasone monotherapy groups was the same and only 3 ml/yr less than the combination group.  This difference was not statistically significant (P=0.44)

The TORCH study found that after three years of treatment, combination therapy was not significantly better than salmeterol alone in reducing the risk of death from any cause (P=0.48). Combination therapy did significantly reduce the annual rate of exacerbations but, importantly, not the rate of severe exacerbations requiring hospitalisation, compared with salmeterol alone. For more information about the original TORCH study, which was funded by the manufacturers of these drugs, see the MeReC review here.

So why is this study not useful for clinicians?
An important improvement change in the management of COPD has been a shift in focus on the patient outcomes considered to be important. In the past, the effect of drug therapy on the rate of decline in FEV1 was used to assess its value as a treatment option, whereas reducing exacerbations and breathlessness, and improving quality of life and exercise performance are now considered more appropriate patient-oriented outcomes (POOs), i.e. does the patient live longer or have a better quality of life?  The original TORCH paper findings with regard to these POOs are outlined above and the observed effects on FEV1 add little of practical value to patients.  Even if we accept the observed benefits of drug therapy over placebo with regard to FEV1 (but see the methodological issues below), most people with moderate to severe COPD already use long acting bronchodilators.   At best, this study suggests that adding an ICS to a long acting bronchodilator does not provide additional benefits in terms of reducing the decline in FEV1 compared to treatment with long acting bronchodilator alone.

There are also methodological problems with this study.  Firstly, this was not a pre-specified outcome in the TORCH trial, TORCH was not formally powered for the analysis of rate of decline in FEV1, and the original primary outcome of TORCH did not achieve statistical significance, so there are good statistical reasons for regarding the comparison between active therapy and placebo as hypothesis generating, not hypothesis testing.  Secondly, as the accompanying editorial points out, this analysis was not really a true intention to treat analysis as the authors claim: missing data may have biased the results.

More information about management of COPD can be found on the relevant section of NPC

Please comment on this blog in the NPC discussion rooms, or using our feedback form