Mortality in IMPACT: Confounded by Asthma?

To the Editor:

Lipson and colleagues’ expanded analysis of mortality from the IMPACT (Informing the Pathway of COPD Treatment) trial convey useful new data, yet they miss an opportunity to provide a more complete picture of the information from this large scale trial in chronic obstructive pulmonary disease (COPD) (1). Indeed, the IMPACT trial rather uniquely included patients with COPD who could have had asthma and selectively added patients with FEV₁ 50–80% predicted who were frequent exacerbators. This overrepresentation in the study population of patients with asthma or an asthma-like profile, together with the abrupt withdrawal of inhaled corticosteroids (ICS) at randomization, will influence the interpretation of the findings (2, 3).

Lipson and colleagues provide an important stratified analysis by prior ICS use. It shows that for the stratum of patients who were not on ICS before randomization, likely exclusively patients with COPD, mortality with triple therapy was no different than with dual long-acting β₂-agonist (LABA) and long-acting muscarinic antagonist (LAMA) bronchodilator therapy (their Figure 3D, with a corresponding hazard ratio of 1.25; 95% confidence interval [CI], 0.60–2.59). This is considerably different than among the stratum of patients with prior ICS use (hazard ratio, 0.63; 95% CI, 0.44–0.89). Other such stratified analyses on asthma diagnosis or profile are lacking.

The apparent reduction in mortality with triple therapy among the patients with prior ICS use could be viewed from the opposite perspective, namely that of a sudden and significant early surge in mortality among those randomized to the dual bronchodilator arm (their Figure 3C). This evidence is reinforced by the equal mortality between the triple and LABA–ICS arms. Indeed, depending on the number of patients with asthma in this large stratum of prior ICS users, the withdrawal of ICS at randomization can have a double impact on mortality. First, stopping ICS use in asthma was shown to increase asthma mortality over fourfold in the first 3 months after discontinuation, with a rate ratio of 4.6 (95% CI, 1.1–19.1) compared with patients continuing ICS use (3, 4). Second, treating asthma patients with a LABA, not combined with ICS, is contraindicated following the associated increase in asthma mortality shown for salmeterol in the SNS (Serevent Nationwide Surveillance) and SMART (Salmeterol Multicenter Asthma Research Trial) trials, with relative risks of asthma death with salmeterol of 3.0 (95% CI, 0.7–20.0) and 4.4 (95% CI, 1.2–15.3), respectively (5, 6).

Thus, the unknown subset of patients with asthma in IMPACT among the 70% of patients in whom ICS use before randomization was withdrawn and replaced by a LABA–LAMA inhaler need to be identified and their outcomes reported. Otherwise, this profile confounds the results, leading to the misleading conclusion that triple therapy reduces mortality in all patients with COPD. In the absence of this information, the more likely conclusion of the IMPACT trial is an increased mortality in the LABA–LAMA arm because of the abrupt withdrawal of ICS in patients who needed it and who were given a contraindicated LABA in a LABA–LAMA inhaler.

In the era of precision medicine, in which identifying the right treatment for the right patient is paramount, it is essential to better understand and dissect the heterogeneity of COPD with targeted trials and pertinent stratified analyses. The reanalysis of the IMPACT trial by Lipson and colleagues provides one important stratified analysis but is lacking other key ones. These could help identify the patients who benefit from triple therapy, thereby preventing unnecessary harms from ICS, including cataracts, pneumonia, and fractures, among those patients who do not benefit.