Methods for therapeutic trials in COPD: lessons from the TORCH trial

Abstract

The TORCH (Towards a Revolution in COPD Health) trial has highlighted some important issues in the design and analysis of long term trials in chronic obstructive pulmonary disease. These include collection of off-treatment exacerbation data, analysis of exacerbation rates and the effect of inclusion of patients receiving inhaled corticosteroids (ICS) prior to randomisation. When effective medications are available to patients who withdraw, inclusion of off-treatment data can mask important treatment effects on exacerbation rates. Analysis of on-treatment data avoids this bias but it needs to be combined with careful analysis of withdrawal patterns across treatments. The negative binomial model is currently the best approach to statistical analysis of exacerbation rates, while analysis of time to exacerbation can supplement this approach. In the TORCH trial, exacerbation rates were higher among patients with previous use of ICS compared to those with no prior use on all study treatments. Retrospective subgroup analysis suggests ICS reduced exacerbation rates compared with placebo, regardless of prior use of ICS before entry to the study. Factorial analysis provides an alternative analysis for trials with combinations of treatments, but assumes no interaction between treatments, an assumption which cannot be verified by a significance test. No definitive conclusions can yet be drawn on whether ICS treatment has an effect on mortality.