Mortality in ETHOS: A Question of “Power”

To the Editor:

We read with great interest the paper by Martinez and colleagues (1) concerning the additional analyses of all-cause mortality of the ETHOS (Efficacy and Safety of Triple Therapy in Obstructive Lung Disease) trial (2). However, we are puzzled that although mortality was a prespecified secondary endpoint of the ETHOS trial and a large section of the discussion was focused on the reduced risk of death in patients treated with budesonide/glycopyrrolate/formoterol fumarate (BGF) 320/18/9.6 μg compared with glycopyrrolate/formoterol fumarate (GFF) 18/9.6 μg (2), Martinez and colleagues (1) inform the scientific community that data on the vital status of 384 patients (4.51% of the enrolled population) were not included in the primary analysis of the ETHOS trial (2). However, this hasty approach in analyzing an important clinical endpoint such as mortality is somewhat questionable when applied to independent research that has no access to patient-level data of sponsored trials. Furthermore, doubts may arise about whether other data of prespecified secondary endpoints in the ETHOS trial may have been roughly analyzed (2).

Indeed, we recognize that the publication by Martinez and colleagues (1) provides extremely interesting and important findings concerning all-cause mortality in the ETHOS trial, compensating for the flaws of the primary analysis (2). In this respect, the statistically significant superiority in terms of the risk of death of BGF 320/18/9.6 μg over GFF 18/9.6 μg resulted from the analysis of 4,257 patients with chronic obstructive pulmonary disease (COPD) (2,137 plus 2,120, respectively) (1).

Interestingly, the post hoc analysis of the power concerning the total adjudicated deaths from the retrieved dataset (1) showed that because of the low mortality prevalence ratio of 0.56 between BGF 320/18/9.6 μg and GFF 18/9.6 μg, data on vital status from at least 5,140 patients with COPD (2,570 each arm) should have been analyzed to have 80% power for observing a statistically significant result (1 − α = 0.95) if a truly beneficial effect was present for BGF 320/18/9.6 μg versus GFF 18/9.6 μg (sample-size calculation performed by using OpenEpi [Emory University] [3]). Thus, 883 additional patients with COPD are needed in the current analysis of the retrieved dataset to exclude the possibility that statistical errors (type I or II) may have affected the results published by Martinez and colleagues (1). Definitely, the ETHOS trial was not adequately powered to detect a statistically significant difference between BGF 320/18/9.6 μg and GFF 18/9.6 μg with respect to the risk of all-cause mortality.

Moreover, looking at the problem from a different point of view, the current evidence (1) resulting from the limited number of events does not allow ruling out that BGF 160/18/9.6 μg may also protect against the risk of all-cause mortality when compared with GFF 18/9.6 μg, precluding a potential therapeutic option.

In any case, the IMPACT (Informing the Pathway of COPD Treatment) trial (4) also goes in the same direction with respect to the effect of triple therapy versus dual-bronchodilation therapy on all-cause mortality, with data from the ETHOS trial suggesting that such an effect is dose dependent and that the most protective effect against mortality seems to be related to protection against cardiovascular events (1).

Overall, data on the risk of death resulting from underpowered studies in which the mortality rates are as low as those in the ETHOS trial (2) should be interpreted with caution, while also being considered in light of the fact that selected populations with COPD enrolled in clinical trials are generally only partially representative of real-life populations (5). In this regard, the retrieved analysis of Martinez and colleagues (1) has the unquestionable advantage of providing a solid base to correctly design long-term clinical trials to definitely assess whether triple therapy may really reduce the risk of death in COPD. In conclusion, we should never forget the lessons from the TORCH (Towards a Revolution in COPD Health) trial (6).