Evaluating the Impact of Triple Therapy on Mortality in Copd: The End is the Beginning?

Mortality has long been one of the two “Holy Grails” of COPD management, the second being the change in the natural history of the disease, as expressed by the rate of decline of FEV₁. The hype around the role of inhaled corticosteroids (ICS) in mortality reduction has been long-standing, as there is evidence that these drugs reduce exacerbations [1] and the rate of decline of FEV₁ [2]. However, the two mega-trials of combinations of ICS/long-acting β-agonist (LABA) designed to evaluate all-cause mortality as the primary endpoint, were not able to show a mortality benefit for these combinations vs. placebo: the TORCH trial showed a 17.5% reduction in mortality with salmeterol/fluticasone propionate that did not reach statistical significance with the notorious p-value of 0.052 [3], whereas the SUMMIT trial did not show any survival benefit for vilanterol/fluticasone furoate in COPD patients with history of cardiovascular disease or with cardiovascular risk factors [4]. Other trials that were not powered to evaluate mortality have provided contradicting results, with the 2-year INSPIRE trial in exacerbating COPD patients showing a mortality benefit for salmeterol/fluticasone vs. tiotropium [5], whereas that was not the case for the FLAME trial in the comparison of salmeterol/fluticasone and the LABA/long-acting muscarinic antagonist (LAMA) combination of indacaterol/glycopyrronium [6], not allowing for firm conclusions. Recently the topic was revisited in the large IMPACT [7] and ETHOS [8] trials, that showed a 28% and 49% reduction in mortality with the triple (ICS/LABA/LAMA) therapy vs. the respective LABA/LAMA combinations, in carefully performed analyses using final retrieved vital status data [9,10]. Although the IMPACT and ETHOS mega-trials possibly mark the end of the data needed for the evaluation of the exacerbation reduction benefit for triple vs. dual bronchodilator therapy in exacerbating COPD patients, the mortality signals that originated from prespecified analyses are not conclusive, since these studies were not designed to evaluate that endpoint. Given the absence of large randomized controlled mortality trials of appropriate duration, possibly more than a year-long, the best possible way to evaluate the mortality benefit of triple therapy vs. dual bronchodilation would be the use of real-life effectiveness data.

In this issue of the COPD journal, Suissa and colleagues have performed an elegant comparative effectiveness analysis of triple therapy vs. dual bronchodilation on mortality, as well as on severe COPD exacerbations and severe pneumonia, in a large cohort of COPD patients in the UK Clinical Practice Research Datalink (CPRD) primary care database [11]. The authors showed a modest 17% increase in mortality in initiators of triple therapy compared to those initiating LABA-LAMA, that was accompanied by a 19% higher risk of severe COPD exacerbations and a 29% increased risk of hospitalized pneumonia. The increased mortality risk was evident in the subgroups of patients without a prior diagnosis of asthma and those without a history of exacerbations in the previous year, suggesting a potential harmful effect in patients who are unlikely to benefit from triple therapy, according to our current understanding for the positioning of ICS in COPD. These observations are useful for our understanding of the real-life effectiveness of triple therapy in COPD, but need to be evaluated in the context of the specific study design of the analysis of Suissa and colleagues [11].

Although not directly comparable with the recent triple therapy randomized controlled trials, we need to focus on some details of the analysis of Suissa and colleagues to understand the observed mortality differences. A major differentiation is the population of the current analysis, using an incident new users cohort design, that involved exclusively patients initiating triple or LABA-LAMA therapy at the index date. A significant proportion of patients were receiving ICS and/or triple therapy in the IMPACT and ETHOS studies, and the mortality benefit was evident in these patients and less or not at all in those who were not receiving ICS or triple therapy at screening [2,9]. A potential “ICS-withdrawal” effect in patients who needed these drugs in the first place has been previously extensively discussed [12]. At the same time, in observational studies, the selection of treatments is based exclusively in the treating physicians’ judgment, and indeed 43.8% of the patients initiated on triple therapy had a history of no exacerbations in the previous year, whereas 34.5% of patients initiated on LABA-LAMA had a diagnosis of asthma, setting under question either the history or the treatment selection. Moreover, significant proportions of patients had been exposed to various treatments for COPD, including 84.8% and 38.7% of triple therapy and LABA-LAMA patients, respectively, having received ICS, thus not being treatment-naive in the year prior to the inclusion in the cohort. Therefore, potential biases in the allocation of treatments exist also in the current CPRD analysis. This is rather expected in “real-world” analyses, where factors like physicians’ judgment cannot be controlled for. Indeed, one would expect physicians to treat more severe patients with more intensified treatment, and this may be reflected on the analysis by Suissa and colleagues.

An additional difference of the current CPRD analysis and the triple therapy trials may lie in the different components, as the proportions of patients using specific LABA, LAMA and ICS substances suggest that the majority has used open triple, or even LABA-LAMA combinations. The mortality benefit signal observed in ETHOS and IMPACT [9,10] may represent the effect of the specific fixed triple combinations vs. their respective LABA-LAMA components. At the same time, the adherence to medication in randomized controlled trials is high, in contrast to the “real-world” treatment duration in the Suissa analysis, where the mean treatment duration on triple therapy was 5.5 months (with >70% of the patients discontinuing one of the components) and 4.8 months for the LABA-LAMA patients (with 19% initiating ICS and >50% discontinuing at least one medication). This likely explains the choice of the authors to evaluate 1-year outcomes and not longer, that would have been more appropriate to evaluate a mortality signal and would have further strengthened the importance of the data.

The importance of treatment effectiveness according to blood eosinophil analyses in a “real-world” setting may carry a significant bias due to the lack of knowledge of the timing of the sampling and the condition of the patient. As the authors also acknowledge, blood eosinophils present significant day-to-day variability and also between steady state and exacerbation [13]. A more important observation is that the 66% higher mortality risk in the COPD patients with triple therapy was observed in patients with an FEV1 < 30% predicted. This may reflect, as mentioned, the greater severity of these patients, the need for a multidisciplinary approach with complex management schemes in that group [14], or even the potential futility for ICS use in very advanced airflow limitation, where lung function loss and – potentially – the room for improvement is minimal [15]. The optimal treatment in patients with very severe airflow limitation deserves further study, as such patients are generally underrepresented in clinical trials.

Overall, the data from Suissa and colleagues build on our current knowledge on the efficacy of triple therapy vs. dual bronchodilation by adding a large body of real-world effectiveness data to the overall picture. The authors may have missed the opportunity for a longer evaluation of mortality, as the follow-up of up to 1 year is likely not enough to evaluate the long-term effectiveness of the two combinations of inhaled medications, yet the treatment changes happening in real-life for various reasons may have anyway prevented them from performing a longer analysis. They may have also missed an opportunity to evaluate any treatment effect for ICS on top of dual bronchodilation in patients with “early” COPD, since they have – rightfully – excluded patients younger than 50 years old. An interesting observation is that the crude mortality rates in this analysis (8.0 vs. 5.5 events per 100 patients per year for triple therapy and dual bronchodilation, respectively) are higher to the quite low 1-year mortality rates in all the patient groups that received effective treatments, ranging from 1.7% to 3.2% in the FLAME, IMPACT and ETHOS trials (Table 1) [2,6,9] that did not include a placebo arm, suggesting that in the context of high adherence within a clinical trial the current combination treatments are very effective. An improvement in the overall survival of COPD has already been observed in the post-millennial years, attributed plausibly to the better treatment of COPD (especially the use of long-acting bronchodilators) and the appropriate management of comorbidities [16]. Having reached the end of a cycle of studies for modern triple and LABA/LAMA combinations, we are concluding our understanding of the positioning of these drugs for exacerbation prevention [17]. For more robust conclusions on the potential mortality benefits of triple therapy, this needs to be the beginning for the building of a properly designed mortality trial, ideally in triple naive patients, in order to avoid any ICS withdrawal effect in patients who needed this treatment [12].

Table 1. Major trials involving inhaled corticosteroids combinations and mortality.

	TORCH [3]6,112 pts	INSPIRE [5]1,323 pts	SUMMIT [4]16,485 pts	FLAME [12]3,358 pts	IMPACT [7,9] 10,355 pts		ETHOS [8,10]8,509 pts
Follow up	3 years	2 years	median 1.8 years (≥1,000 deaths)	1 year	1 year		1 year
No. of deaths	875	59	1,037	48	138		170
Primary endpoint	All-cause mortality	Exacerbations	All-cause mortality	Exacerbations	Exacerbations		Exacerbations
Comparators	SFC vs. placebo	SFC vs. TIO	FF/VI vs. placebo	IND/GLY vs. SFC	FF/UMEC/VI vs. UMEC/VI	FF/UMEC/VI vs. FF/VI	BUD320/GLY/Form vs. GLY/Form	BUD320/GLY/Form vs. BUD320/Form
Deaths per arm	193 vs. 231	21 vs. 38	246 vs. 275	24 vs. 24	66 vs. 98	66 vs. 109	37 vs. 64	37 vs. 46
Rates	15.2 vs. 12.6%	3 vs. 6%	6.0 vs. 6.7%	1.4 vs. 1.4%	2.4 vs. 3.2%	2.4 vs. 2.6%	1.7 vs. 3.0%	1.7 vs. 2.2%
HR (95%CI)	0.825 (0.681 to 1.002)	0.48 (0.27–0.85)	0.88 (0.74–1.04)	Not provided	0.72 (0.53–0.99)	0.89 (0.67–1.16)	0.51 (0.33–0.80)	0.72 (0.44–1.16)
% Reduction	−17.5%	−52%	−12%	0%	−28%	−11%	−49%	−28%
p-value	p = 0.052	p = 0.032	p = 0.137	Not provided	p = 0.042	p = 0.387	p = 0.0035	p = 0.1721

BUD, budesonide; CI, confidence intervals; FF, fluticasone furoate; Form, formoterol; GLY, glycopyrronium; HR, hazard ratio; SFC, salmeterol/fluticasone propionate; TIO, tiotropium; UMEC, umeclidinium; VI, vilanterol.

Declaration of Interest

Dr. Kostikas has received honoraria for presentations and consultancy fees from AstraZeneca, Boehringer Ingelheim, Chiesi, CSL Behring, ELPEN, GSK, Menarini, Novartis, Sanofi Genzyme and WebMD. His department has received funding and grants from AstraZeneca, Boehringer Ingelheim, Chiesi, Innovis, ELPEN, GSK, Menarini, Novartis and NuvoAir. He is a member of the GOLD Assembly.

Dr. Kyriakopoulos kas nothing to disclose.

Dr. Gogali has received honoraria for presentations and consultancy fees from AstraZeneca, Boehringer Ingelheim, Chiesi, ELPEN, GSK, Novartis.