Clinically Important Improvements and Disease Stability with Fluticasone Furoate/Umeclidinium/Vilanterol Once-Daily Single-Inhaler Triple Therapy in the ELLITHE Trial: A Post Hoc Responder Analysis

Abstract

Introduction

Responder analyses provide information about characteristics associated with therapeutic benefits. Short-term responses may predict long-term benefits. We evaluated responders, clinically important improvement (CII), disease stability (DS), and the relation of short- to long-term responses in patients with chronic obstructive pulmonary disease (COPD) in ELLITHE.

Methods

ELLITHE was a multicenter, open-label, non-interventional effectiveness study between 2020 and 2022 evaluating the effects of treatment initiation with once-daily single-inhaler triple therapy (odSITT) FF/UMEC/VI (100/62.5/25 µg via ELLIPTA) on COPD Assessment Test (CAT), forced expiratory volume in 1 s (FEV₁), and exacerbations over 12 months. Post hoc responder analyses for CAT (≥ 2 units improvement), FEV₁ (≥ 100 ml change), and exacerbations (no event) were performed. Composite endpoints CII and DS (CII = response to at least two outcomes; DS = absence of clinically important deterioration for all outcomes) were also evaluated.

Results

A total of 786 patients had available data for any analysis. At study completion, 53.3% of patients were CAT, 36.7% FEV₁, and 90.2% exacerbation responders, with 22.1% responding to all outcomes; 64.3% had a CII, and 52.7% showed DS. CII and DS were more frequent in subjects with higher baseline CAT score, and DS in patients  on prior ICS/LABA therapy (all p < 0.05). Early (3 months) CAT, FEV₁ and CII response strongly predicted respective responses at study end (odds ratios = OR ranging from 6.3 to 7.4), and DS (OR from 3.0 to 4.2). In the patient subset with available baseline eosinophil counts, response was generally similar at < 150 versus ≥ 150 cells/μl.

Conclusions

Despite overlapping responses to single and composite outcomes with odSITT, individual patterns support a multidimensional approach to evaluate benefits in COPD. Responders had higher baseline CAT scores and frequency of prior dual therapies. Short-term responses of FEV₁ and/or CAT were reasonable predictors of long-term responses, including DS. DS was achievable for the majority of patients and may represent a useful outcome for future COPD research and management.

Supplementary Information

The online version contains supplementary material available at 10.1007/s41030-025-00306-1.

Key Summary Points

Why carry out this study?

Responder analyses are important to understand characteristics associated with therapeutic outcomes in patients with COPD. Composite endpoints are currently discussed as future outcomes. We performed responder analyses to single (CAT, FEV1, exacerbations) and composite (clinically important improvement, disease stability) outcomes in a post hoc evaluation of the ELLITHE non-interventional study with inhaled triple therapy.

What was learned from the study?

Most patients showed a clinically meaningful response to at least one outcome, with reasonable overlap between responses. Defining treatment success in COPD should be based on multiple outcomes. Short-term responses of FEV1 and/or CAT are reasonable predictors of long-term responses. Disease stability is achievable for a large group of patients after 1 year and may represent a feasible outcome for future research and COPD management.

Introduction

The main goals of chronic obstructive pulmonary disease (COPD) treatment are symptomatic relief and future risk reduction [1]. Due to the complex nature of COPD and its clinical manifestations, potential benefits of interventions (pharmacological or non-pharmacological) are usually assessed in a multidimensional way, combining patient-reported outcomes (PROs), objective measures of function (typically the forced expiratory volume in 1 s, FEV₁) and events of prognostic significance, most importantly moderate and/or severe exacerbations [2]. Although these outcomes are interconnected (e.g., worse FEV₁ increases the risk of exacerbations [3], which in turn diminish health status [4], and promote further FEV₁ decline [5] and so forth), discordance in response to either of these outcomes may be observed on the individual level [6, 7]. In some large-scale clinical trials, surprisingly little agreement existed between clinical improvements in FEV₁ and symptoms like dyspnea [8, 9]. It is therefore important to be aware of differential responses to interventions, and the identification of possible underlying reasons for response disparities may help to tailor treatments to specific patients’ needs as part of treatment personalization [10]. Responder analyses can be informative in this regard by identifying subgroups of patients with particularly beneficial (or poor) responses to intervention, either to single or multiple outcomes. As one recent example, data from the EMAX study demonstrated that patients with COPD on dual bronchodilation with umeclidinium/vilanterol (UMEC/VI) have higher likelihood to experience a “clinically important improvement” (CII), ie a combined FEV₁ and PRO response, compared to patients on single bronchodilator (UMEC or salmeterol) [11]. Moreover, an early response to therapy was predictive of long-term benefits in the same outcome categories in this study [11, 12]. The latter is of importance, as in clinical practice response to a treatment or treatment change is often evaluated at time intervals pre-determined by the organization of healthcare, e.g., at monthly, quarterly, or annual follow-up visits. It is important for physicians to understand, if short-term responses to therapy (usually assessed by symptoms, general health status or lung function during follow-up), are reliable predictors of more relevant long-term outcomes of COPD (exacerbation, progression, mortality). In this regard, Kostikas et al. showed that an “early” CII (ECII), a composite endpoint defined by achieving minimally clinically important improvements in FEV₁ and one PRO (COPD assessment test = CAT, or St. George’s Respiratory Questionnaire = SGRQ) after 4 or 12 weeks treatment was able to predict subsequent exacerbation risk over 1 year [13].

As COPD is by definition a not fully reversible disease, any intervention, however beneficial, will reach a ceiling effect. Instead of returning to “normal”, it may be more relevant for patients to achieve and experience a stabilization of their chronic disease, as observed e.g., after smoking cessation [14]. Indeed, COPD progresses in many patients over time, with gradual worsening of both objective parameters like airflow obstruction, hyperinflation, physical performance, and oxygenation, and subjective PROs like symptoms or health status [15–17]. In patients experiencing frequent exacerbations, the risk for future exacerbations is heightened, and exacerbation-free intervals may shorten over time, further increasing the risk for repeated hospitalizations and ultimately, death [18]. Therefore, a novel concept of establishing “disease stability” as an important outcome in COPD has recently been proposed [19]. Disease stability (DS) was defined as composite endpoint, e.g., absence of exacerbations, stable lung function (FEV₁) and health status over a given time period (usually 6–12 months). Noteworthy, a similar concept of long-term COPD “control” has been proposed by Soler-Cataluna et al. [20]. Recent post hoc analyses from clinical trials indicated that more patients achieved DS with inhaled triple therapy fluticasone furoate/umeclidinium/vilanterol (FF/UMEC/VI) than dual comparators budesonide/formoterol (BUD/FOR) [21], or FF/VI and UMEC/VI [22] after 6 or 12 months of treatment, respectively. Triple therapy with inhaled corticosteroid (ICS), long-acting β₂-agonist (LABA) and long-acting muscarinic antagonist (LAMA) is currently approved as maintenance treatment option for patients with moderate-to-severe COPD uncontrolled on dual combination therapies [23–25], and recommended for patients experiencing exacerbations [1]. However, using DS as a key outcome in a COPD trial could help identifying patients that might benefit from more intense early intervention with e.g., pharmacotherapies, as compared to the usual ex post “step-up” approach proposed in guidelines and management strategies. Moreover, in everyday practice, the treatment goal of DS may not only be more realistic, but also more relevant to patients [19].

For the time being, data on DS are scant and have only been published in abstract form. We therefore performed a post hoc analysis from the ELLITHE non-interventional trial with once-daily single inhaler triple therapy (odSITT) FF/UMEC/VI over 1 year [26]. Briefly, the study demonstrated clinically relevant improvements in CAT score, FEV₁, exacerbations, and adherence at study completion, confirming and extending evidence from randomized controlled clinical trials (RCTs) [27, 28]. Here, we sought to a) describe response frequency and responder populations for single (CAT, FEV₁, exacerbations) and composite outcomes (CII and DS); and b) study the relationship of short-term (3 months) to long-term outcomes including DS after initiation of odSITT in patients with COPD.

Methods

Trial Design and Oversight

A detailed description of the trial design has been previously published [26]. In brief, ELLITHE was a multicenter, non-interventional, open-label, effectiveness study evaluating once-daily single-inhaler FF/UMEC/VI delivered by the ELLIPTA inhaler in uncontrolled patients with COPD in a usual clinical practice setting over 12 months. The primary objective was to evaluate the effectiveness of odSITT on health status in patients with COPD after 12 months of treatment versus baseline. The ELLITHE study was conducted from June 2020 to July 2022 in pulmonology specialist (N = 111) and internal/general medicine (N = 8) practices in Germany. Five study visits were planned: one at baseline/enrollment (Visit 1, V1) and one after 3, 6, 9, and 12 months on treatment (V2-V5). If part of the routine, blood eosinophil counts (BEC, actual or historical value) were collected.

Ethical Approval

The ELLITHE study was registered under the German Clinical Trials Register (identifier: DRKS00031897) The study was carried out in accordance with Good Clinical Practice guidelines under the provisions of the latest version of the Declaration of Helsinki (2013) and received approval from of the State Chamber of Physicians of Hesse (“Ethikkommission der Landesärztekammer Hessen”) as the coordinating ethics committee of the national chief investigator. The study was registered at the German Clinical Trials Register (DRKS00031897). All patients provided signed informed consent.

Study Population

The inclusion and exclusion criteria were minimal to align with the use according to the EU label. Briefly, all patients had a confirmed COPD diagnosis by spirometry in the medical records of their treating physician. They were symptomatic with a CAT score ≥ 10 at screening and had at least one exacerbation in the year before study entry. [26].

Responder Analysis of Effectiveness Outcomes

The primary outcome of the original trial was the change in CAT score between baseline and study end. Key secondary outcomes were FEV₁ and exacerbations. We used established criteria to define response, CII and DS by applying thresholds of minimal important changes [29–32] from baseline. For CAT, a ≥ 2 units improvement signified “response”, while improvements of twice the minimal clinically important difference (MCID) were labelled as “large response”. For FEV₁ response, increases of ≥ 100 ml and ≥ 200 ml, respectively, were used. Exacerbation responders had no event (versus any event in non-responders) during the observation period of 12 months. A CII was achieved when a subject experienced at least two of the above-listed three responses during the observation period. We also defined a group of total responders that achieved responses in all three components after initiation of odSITT. Finally, DS was attributed to all subjects that neither experienced a clinically relevant worsening in CAT (+ 2 units increase or more) and FEV₁ (− 100-ml decrease or more), nor experienced any exacerbation until study end.

For the sake of group comparisons, responders for composite outcomes were labelled (+) vs. (−) for non-responders for the respective category. We calculated the percentage of CAT, FEV₁, and exacerbation responders and compared responder rates according to specified baseline characteristics including age, sex, GOLD spirometry grade, smoking status, historic blood eosinophil levels and prior treatment regime in responders versus non-responders, CII (+) vs. CII (−), and DS (+) vs. DS (−) subjects. We also analyzed the relation of early responses to treatment (at first follow-up visit after approximately 3 months) to late treatment response (study end).

Statistical Considerations

For this responder analysis, we included all patients with an in-label prescription of odSITT, i.e., those on prior LAMA/LABA, ICS/LABA or LAMA/LABA/ICS. In addition, only patients with available data for all three response outcomes CAT score, FEV₁ and exacerbations at V1 and during follow-up were included for the analysis of total responders. If patients discontinued prematurely from study treatment or follow-up, the last-observation-carried-forward (LOCF) approach was used [33]. For the prediction of late response at study end by early response at V2, only patients with values at V2 and available V5 (or LOCF if applicable) were taken into account. V2 values were not considered for LOCF. No other imputation methods were applied.

The post hoc analysis was purely exploratory in nature. No confirmatory approach, hypothesis testing and/or multiplicity adjustment was undertaken. Data were analyzed descriptively. Arithmetic data were presented as mean values with 95% confidence intervals (CI) or standard deviation (SD). Group comparisons between responders and non-responders were performed exploratively with a significance level of α = 5% by non-parametric or parametric statistics, where applicable. Logistic regression analyses were performed for the prediction of late response at study end by early response at V2. Odds ratios (OR) were calculated and reported together with the 95% CI. All statistical analyses were carried out by means of the SAS^® package (version 9.4).

Results

Responder Analysis Population

Between June 2020 and July 2021, a total of 931 patients were enrolled; 906 patients were included in the full analysis set (FAS). A total of 786 patients with in-label treatment (dual LAMA/LABA, ICS/LABA, or free or fixed triple LAMA/LABA/ICS combinations) prior to switch to odSITT were included for the present post hoc analyses and had available responder data. Varying N numbers for each analysis, if applicable, are due to missing data on single or multiple responder outcomes for individual subjects. Baseline characteristics of this subgroup have been published in the primary analysis [26]. Briefly, patients were highly symptomatic with a mean CAT score of 21.7 ± 6.7 and all had a mean exacerbation rate in the year prior to enrolment of 1.4 events.

Responder Analyses for the Effectiveness Outcomes

CAT Score

Following initiation of odSITT, 53.3% of subjects were CAT-responders at V5/LOCF. In addition, 40.7% of subjects had a large response (twice MCID) in CAT score. Mean baseline CAT scores were higher in CAT responders, and more CAT responders had GOLD spirometry grade II. There were no differences between CAT responders and non-responders for age, sex, smoking status, BEC and prior maintenance regime (Table 1). However, CAT response rates were numerically higher in patients on prior dual therapies (57.1% for ICS/LABA, 54.4% for LAMA/LABA) versus prior triple therapy (48.4% for non-odSITT).

Table 1.

Baseline patient demographics and baseline characteristics depending on CAT response at V5/LOCF

Baseline demographics and characteristicsa		CAT responder Yes (N = 329)	CAT responder No (N = 288)
Sex	Male, n (%)	179 (54.4)	160 (55.6)
	Female, n (%)	150 (45.6)	128 (44.4)
Age (year), mean ± SD		66.7 ± 9.8	66.7 ± 9.5
BMI (kg/m2), mean ± SD		27.4 ± 5.6	26.8 ± 5.4
Smoking status	Active smoker, n (%)	121 (36.8)	111 (38.5)
	Former smoker, n (%)	163 (49.5)	152 (52.8)
	Non-smoker, n (%)	45 (13.7)	25 (8.7)
Atopic patient, n (%)		31 (9.4)	17 (5.9)
Asthma diagnosis before 40th year of age, n (%)		36 (10.9)	17 (5.9)
Chronic bronchitis at first diagnosis, n (%)		154 (47.0)	137 (47.7)
GOLD spirometry grade	GOLD II, n (%)	195 (59.3)*	142 (49.5)
	GOLD III, n (%)	134 (40.7)*	145 (50.5)
Rate of COPD exacerbations in the prior 12 months	Total, mean ± SD	1.4 ± 0.7	1.4 ± 0.8
	Mild, mean ± SD	0.4 ± 0.7	0.4 ± 0.8
	Moderate, mean ± SD	0.9 ± 0.8	0.8 ± 0.8
	Severe, mean ± SD	0.1 ± 0.3	0.1 ± 0.4
Total CAT score, mean ± SD		23.0 ± 6.5**	19.7 ± 6.1
FEV1 (L), mean ± SD		1.446 ± 0.491	1.422 ± 0.504
mMRC grading	Grade 0-I, n (%)	93 (28.6)	72 (25.0)
	Grade II-IV, n (%)	233 (71.5)	216 (74.9)
Peripheral blood eosinophil count (%), mean ± SD#		3.1 ± 2.8	2.7 ± 1.7
Peripheral blood eosinophils (cells/µl), mean ± SD##		243.0 ± 225.1	212.9 ± 149.5
Prior treatment	LAMA/LABA/ICSb, n (%)	77 (23.4)	82 (28.5)
	ICS/LABAb, n (%)	60 (18.2)	45 (15.6)
	LAMA/LABAb, n (%)	192 (58.4)	161 (55.9)

Group comparison p value (χ²-test or t test) *p < 0.05; **p < 0.0001, ^#N = 120; ^## N = 115

BMI body mass index, CAT COPD Assessment Test, COPD chronic obstructive pulmonary disease, FEV₁ forced expiratory volume in 1 s, GOLD Global initiative of chronic obstructive pulmonary disease, ICS inhaled corticosteroid, LABA long-acting β₂-agonist, LAMA long-acting muscarinic receptor antagonist, mMRC modified Medical Research Council Dyspnea Scale, SD standard deviation

^aNumber of missing values varied between the described patient demographics and characteristics

^bFixed or free combination

Lung Function

At study completion, 36.7% of subjects were FEV₁ responders, while 23.5% had large (twice MCID) responses. FEV₁ responders were younger and the proportion of subjects with FEV₁-response and large response was higher in those on prior ICS/LABA therapy compared to subjects on prior LAMA/LABA and triple therapy (48.0%, 34.5%, and 34.4%; p = 0.0337, and 37.3%, 19.8%, and 23.1%; p = 0.0012, respectively). No differences were found for other baseline characteristics, including BMI, smoking status, atopy, Asthma/bronchitis comorbidity, GOLD grade, exacerbation history, baseline CAT and FEV₁, mMRC grade and eosinophils (Columns 3 and 4 of Supplementary Table 1).

COPD Exacerbations

During the 1-year observation period, 90.2% of subjects remained exacerbation-free (exacerbation responders). The proportion of exacerbation-free subjects differed between those on former ICS/LABA therapy versus prior LAMA/LABA or triple combinations other than odSITT (97.2%, 90.4%, and 85.3%; p = 0.0045). Exacerbation responders had fewer exacerbations prior to study enrollment versus non-responders (mean total exacerbations 1.4/year vs. 1.7/year; p = 0.014), were less frequent active smokers, had a higher proportion of GOLD II vs. GOLD III, and were slightly older. No differences were found for other baseline characteristics including BMI, atopy, Asthma/bronchitis comorbidity, baseline CAT and FEV₁, mMRC grade and eosinophils (Columns 5 and 6 of Supplementary Table 1).

Overlap Between Clinical Response Outcomes and CII

Figure 1 shows response distribution of the selected three outcomes and their overlap in the study population at study end. In total, 22.1% (n = 131) of patients were total responders (all three outcomes). Total response was more frequent in subjects on prior ICS/LABA therapy than LAMA/LABA or triple therapy (30.8%, 19.7%, and 17.8%; p = 0.0251).

Fig. 1.

Overlap of responders for CAT, FEV₁ and exacerbations at study end. For this responder analysis, data from N = 594 patients were available who had valid data for each outcome at V5/LOCF; 564 patients were responders in at least one of the three outcomes (predominantly absence of exacerbations, followed by CAT-score improvement by ≥ 2 units, and FEV₁ improvement ≥ 100 ml vs. baseline). Isolated CAT and/or FEV₁ responses were very rare. Most subjects had at least one outcome response in addition to exacerbations, and 22.1% of patients were responders for all three outcomes. CAT COPD assessment test, FEV₁ forced expiratory volume in 1 s

A CII (response to at least two outcomes) was achieved by 64.3% of subjects (including the 22.1% total responders), either by a combination of CAT/Exacerbations (30.0% of responders), FEV₁/Exacerbations (13.3%), or CAT/FEV₁ (0.7%). Subjects with CII (+) had higher CAT scores at baseline than those without CII, had a higher proportion of GOLD II subjects, and fewer active smokers (Table 2).

Table 2.

Baseline patient demographics and baseline characteristics depending on clinically important improvements or disease stability at V5/LOCF

Baseline demographics and characteristicsa		CII Yes (N = 397)	CII No (N = 220)	DS Yes (N = 322)	DS No (N = 289)
Sex	Male, n (%)	223 (56.2)	116 (52.7)	168 (52.2)	169 (58.5)
	Female, n (%)	174 (43.8)	104 (47.3)	154 (47.8)	120 (41.5)
Age (year), mean ± SD		66.6 ± 9.9	66.8 ± 9.4	67.5 ± 9.4*	65.5 ± 9.7
BMI (kg/m2), mean ± SD		27.1 ± 5.5	26.7 ± 5.5	27.2 ± 5.4	26.9 ± 5.5
Smoking status	Active smoker, n (%)	148 (37.3)*	88 (40.0)	107 (33.2)*	127 (43.9)
	Former smoker, n (%)	193 (48.6)*	117 (53.2)	171 (53.1)*	137 (47.4)
	Non-smoker, n (%)	56 (14.1)*	15 (6.8)	44 (13.7)*	25 (8.7)
Atopic patient, n (%)		32 (8.1)	16 (7.3)	22 (6.9)	26 (9.0)
Asthma diagnosis before 40th year of age, n (%)		32 (8.1)	21 (9.5)	28 (8.7)	25 (8.7)
Chronic bronchitis at first diagnosis, n (%)		177 (44.8)	113 (51.4)	146 (45.6)	145 (50.2)
GOLD spirometry grade	GOLD II	230 (57.9)*	105 (47.9)	173 (53.7)	158 (54.9)
	GOLD III	167 (42.1)*	114 (52.1)	149 (46.3)	130 (45.1)
Rate of COPD exacerbations in the prior 12 months	Total, mean ± SD	1.4 ± 0.8	1.4 ± 0.7	1.4 ± 0.8	1.4 ± 0.7
	Mild, mean ± SD	0.4 ± 0.8	0.3 ± 0.7	0.4 ± 0.8	0.3 ± 0.6
	Moderate, mean ± SD	0.8 ± 0.8	0.8 ± 0.8	0.8 ± 0.7	0.9 ± 0.8
	Severe, mean ± SD	0.1 ± 0.3**	0.1 ± 0.4	0.1 ± 0.3	0.1 ± 0.4
Total CAT score, mean ± SD		22.3 ± 6.6***	20.1 ± 6.3	22.0 ± 6.2*	20.7 ± 6.8
FEV1 (L), mean ± SD		1.455 ± 0.497	1.395 ± 0.493	1.396 ± 0.493*	1.476 ± 0.507
mMRC grading	Grade 0-I, n (%)	109 (27.7)	55 (25.0)	82 (25.7)	78 (27.1)
	Grade II-IV, n (%)	285 (72.3)	165 (75.0)	238 (74.4)	210 (72.9)
Peripheral blood eosinophil count (%), mean ± SD#		3.0 ± 2.7	2.7 ± 1.7	2.9 ± 2.5	2.9 ± 2.2
Peripheral blood eosinophils (cells/µl), mean ± SD##		236.2 ± 219.0	217.6 ± 152.7	235.8 ± 228.2	221.7 ± 159.8
Prior treatment	LAMA/LABA/ICSb, n (%)	92 (23.2)	66 (30.0)	73 (22.7)**	87 (30.1)
	ICS/LABAb, n (%)	74 (18.6)	29 (13.2)	68 (21.1)**	35 (12.1)
	LAMA/LABAb, n (%)	231 (58.2)	125 (56.8)	181 (56.2)**	167 (57.8)

Group comparison p value (χ²-test or t test) *p < 0.05; **p < 0.01; ***p < 0.0001; ^#N = 123–124; ^##N = 118

BMI body mass index, CAT COPD Assessment Test, CII clinically important improvement, COPD chronic obstructive pulmonary disease, DS disease stability, FEV₁ forced expiratory volume in 1 s, GOLD Global initiative of chronic obstructive pulmonary disease, ICS inhaled corticosteroid, LABA long-acting β₂-agonist, LAMA long-acting muscarinic receptor antagonist, mMRC modified Medical Research Council Dyspnea Scale, SD standard deviation

^aNumber of missing values varied between the described patient demographics and characteristics

^bFixed or free combination

Disease Stability

At study completion, 52.7% of subjects achieved disease stability with odSITT. Reasons for failing DS were equally distributed between CAT (worsening by ≥ 2 units: 25.6%) and FEV₁ (worsening by ≥ 100 ml: 25.2%), while 9.8% of subjects had an exacerbation. Most characteristics of subjects achieving DS were similar to those failing to achieve DS (Table 2). However, baseline CAT score and age was higher in DS ( +) patients, while FEV₁ was slightly lower compared to DS (−) patients. Fewer active smokers and more frequent use of prior ICS/LABA treatment was observed in clinically stable subjects.

Early Improvement as Predictors of Long-Term Outcomes

The rate of CAT responders at the first follow-up visit (V2) was 53.2%. The majority of early responders also had CAT response at V5/LOCF (Fig. 2A). The odds ratio (OR) for a long-term CAT response in early responders was 6.3 (95% CI 4.2–9.3) vs. early non-responders. Similar observations were made for FEV₁ (OR = 7.4, 95% CI 4.8–11.4) (Fig. 2B). 69.8% of subjects had a CII at V2, and 60.5% had CII at V5/LOCF (OR = 7.3, 95% CI 4.7–11.4) (Fig. 2, C). Early FEV₁ response was also associated with late CAT response (OR = 1.7, 95% CI 1.2–2.4). DS at V5/LOCF was associated with early CAT response (OR = 3.1, 95% CI 2.1–4.5), early FEV₁ response (OR = 3.0, 95% CI 2.0–4.5) and a combination of early CAT and early FEV₁ response (OR = 5.2, 95% CI 3.1–8.9). The latter was similar to the association with a combination of large early FEV₁ and CAT response (2-times MCID) (OR = 4.5; 95% CI 2.3–9.0). DS at V5/LOCF was also associated with early CII response (OR = 4.2, 95% CI 2.7–6.5).

Fig. 2.

Prediction of response to outcomes from V2 to V5/LOCF. A Linked data from V2 and V5/LOCF for CAT response (≥ 2 units improvement) were available from N = 491 patients. At V2 already half of the patients were CAT responders. At study end, a similar proportion of patients were CAT responders. OR = 6.3, 95% CI 4.2–9.3. B Linked data from V2 and V5/LOCF for FEV₁ response (≥ 100 ml improvement) were available from N = 464 patients. While 40.3% of patients were FEV₁ responders at V2, the proportion slightly reduced until V5/LOCF to 35.8%. OR = 7.4, 95% CI 4.8–11.4. C Clinically important improvement (CII) was defined as response to at least two outcomes (CAT improvement ≥ 2 units, FEV₁ increase ≥ 100 ml or exacerbation-free). Linked data from V2 and V5/LOCF for CII were available from N = 453 patients. Most of the patients with CII at V2 were still CII (+) at V5/LOCF. OR = 7.3, 95% CI 4.7–11.4. Discrepancies in % responders to the total population are due to lower N numbers for patients with available linked data for V2 and V5/LOCF. CAT COPD assessment test, FEV₁ forced expiratory volume in 1 s

Response by Baseline BEC

Baseline characteristics for patients with BEC of 150 vs. ≥ 150 cells/µl are shown in Supplementary Table 2 and were similar in both groups. There were no differences in the rates of CAT, FEV₁ and exacerbation response, total response, CII, and DS in the subgroup of patients with baseline BEC of < 150 vs. ≥ 150 cells/µl (Fig. 3).

Fig. 3.

Response to outcomes depending on blood eosinophils at baseline. Values for eosinophilic granulocytes were available for N = 115–120 patients. Outcomes were determined at V5/LOCF. CAT responders were defined as improvement by ≥ 2 units, FEV₁ responders as increase by ≥ 100 ml, total responders were CAT and FEV₁ responder without exacerbations. CII was defined as response to at least two of the mentioned outcomes, disease stability was attributed to all subjects that neither experienced a clinically relevant worsening in CAT (+ 2 units increase or more) and FEV₁ (-100 ml decrease or more), nor experienced any exacerbation until V5/LOCF. Response to outcomes, total responders as well as CII and DS did not significantly differ between patients with low (< 150 cell/µl) vs. high (≥ 150 cell/µl) blood eosinophils at baseline. FEV₁ and total response were numerically lower in patients with low blood eosinophils. Group comparison: p value (χ²-test) not significant. CAT COPD assessment test, CII clinically important improvement, DS disease stability, FEV₁ forced expiratory volume in 1 s

Discussion

In this post hoc analysis from the ELLITHE non-interventional real-world study with FF/UMEC/VI odSITT over 12 months, we describe distinct patterns of treatment response, defined by established COPD outcomes, and provide evidence that shorter-term responses to therapeutic interventions may be indicative of long-term improvements and/or disease stability. The results are significant, as they support and extend prior knowledge derived from post hoc analyses of well-designed RCTs.

Responder analyses of early and sustained improvements in single or composite endpoints have been previously published. However, some of them were based on relatively short observation periods, (e.g., 12 weeks in the CRYSTAL study [8], or 24 weeks in the EMAX study [11, 12]) and/or included selective patient populations as required for pivotal RCTs [13]. Therefore, some of these analyses lacked data on long-term outcomes like disease progression or exacerbations, and the extension and confirmation of these findings in longer-term real-world studies were considered to be highly relevant by the authors [11]. Our post hoc analysis from ELLITHE clearly contributes to filling this evidence gap by providing responder analyses to single and composite outcomes in a broad COPD population with exacerbation background in real-world clinical practice. Although this non-interventional trial did not include a control group, the option to compare treatment responders and non-responders nonetheless provides important clinical information about responder characteristics, response overlap and predictive value of early versus sustained responses.

In ELLITHE, we observed considerable discordance between treatment responses, with 53.2% early CAT responders, 40.8% early FEV₁ responders and only 22.0% overlapping response. The responder rates for these outcomes were in line with observations in the CRYSTAL study [8], but considerably higher than those reported for the FLAME study with a similar trial population [13]. However, these RCTs included a run-in period and treatment blinding, so different baseline conditions may have contributed to the somewhat lower responder rate in FLAME versus ELLITHE. It is also likely, that in an uncontrolled, observational treatment setting, Hawthorne effects and expectations become more relevant, thus increasing the likelihood of surpassing a response threshold for a defined outcome [34, 35]. Consistent with prior publications, early treatment response to either CAT or FEV₁remained relatively stable and predicted a large proportion of the longer-term response to the same outcome at study end. Hence, assessment of short-term responses appears a reasonable and rational way to evaluate the efficacy of therapeutic interventions as part of the management cycle of COPD in clinical practice [11, 12]. However, on an individual level it is important to note limitations of this approach: on the one hand, early non-responders may change the responder status over time, on the other hand, early response of e.g., FEV₁ may vanish over time due to the progressive nature of COPD [30]. In ELLITHE, the FEV₁ responder population at study end was lower than at first follow-up, lending some support to the latter argument. Mechanistically, other factors than smooth muscle relaxation underlying early response for lung function and symptoms need to be taken into account, e.g., a reduction of mucus plugging [36]. The CAT score also contains several items that are only weakly correlated with changes in airway calibre, which could explain the observed disconnect of response in ELLITHE.

Interestingly, early FEV₁ response in ELLITHE was also associated with a high likelihood of CAT response at study end, indicating a certain general suitability of this “biomarker” to predict improvements in a PRO. Indeed, post hoc analyses of large, well-designed RCTs have confirmed that incremental changes in FEV₁ are linked to incremental improvements in PROs and exacerbations [6, 7], at least on a group level.

After one year with odSITT, > 90% of subjects in ELLITHE remained exacerbation-free, despite a documented history of moderate/severe exacerbations prior to initiation of FF/UMEC/VI. The reasons for this, although encouraging, are not fully clear and have been discussed elsewhere [26]. Briefly, a considerable impact by the COVID-19 pandemic is likely, as exacerbation rates decreased by > 50% in many countries, while showing an increase again from 2021 onward [37]. However, this almost-complete response clearly represents a limitation to the value of this outcome to differentiate exacerbation responders in our study population. Nonetheless, exacerbation response (i.e., absence of) was included in both our definitions of the composite endpoints CII and DS. Including exacerbations in the composite CII is somehow different to approaches by authors from CRYSTAL [8], EMAX [11], or FLAME [13] post hoc analyses. However, we consider the absence of any exacerbation as relevant in the context of CII, in as much as it is hardly conceivable that a clinically important improvement with a treatment can be present when at the same time an exacerbation has occurred. Nonetheless, by analyzing responses of multiple outcomes, we were able to discern the main drivers of response achievement in the composite CII. Indeed, the largest proportion of CII responders were defined by CAT/exacerbations (30.0%), followed by CAT/FEV₁/exacerbations (total responders = 22.1%), FEV₁/exacerbations (13.3%) and finally, CAT/FEV₁ (0.7%).

In ELLITHE, we relied on a single PRO (CAT) to define CII. Other post hoc analyses included multiple PROs (with or without FEV₁) to define CII [11, 13]. However, although these PROs reflect distinct aspects of a multidimensional disease like COPD, and are not interchangeable, surprisingly small differences in responder rates were observed in most publications when defining CII by different PROs. For example, early CII response in the FLAME study was present in 20.0% (week 4) and 18.4% (3 months) of subjects in both study arms when CII was defined by FEV₁and CAT, while 17.9% (week 4) and 19.3 (3 months) were responders when CII was defined by FEV₁and SGRQ [13]. We did not collect symptom or health status PRO data other than CAT in ELLITHE due to the real-world design of the study. Indeed, most validated PRO instruments apart from the CAT are rarely used in clinical practice, mainly due to their complexity and time-consumption.

Our post hoc analysis is the first real-world study to provide long-term data on the novel COPD concept of “disease stability” (DS). Earlier post hoc publications of RCTs have used a composite endpoint of “clinically important deterioration” (CID), which is essentially a “negative” blueprint of the DS concept [32, 38, 39]. In contrast to CID, DS is defined by the absence of any deterioration in typical COPD outcomes (FEV₁, exacerbations, PROs). However, although technically similar, the term “stability” may be more reflective of a relevant, desirable COPD long-term outcome in future research and clinical practice. Our data from ELLITHE indicate that achievement of DS is possible in a large group of at-risk COPD patients with high symptom burden, exacerbation history and moderate-to-severe lung function impairment. Our 12 months of data also extend those from a recent post hoc analysis from the FULFIL study, where 49% of patients achieved DS with odSITT (CAT-based definition) versus 26% with dual ICS/LABA (BUD/FOR) therapy, albeit after 6 months treatment [21]. DS at study end in ELLITHE was more likely in early CII responders than in CII non-responders, supporting the predictive value of this composite outcome for long-term stability. That said, the feasibility of using a composite outcome like DS in real-world clinical practice remains elusive. Although simple by design, it would nonetheless require availabilty of spirometry to follow-up individual FEV₁ values, willingness to use CAT score consistently in the majority of patients, and should ideally incorporate a more formal/standardized way to capture occurrence of exacerbations. Further, the impact of ageing on the “natural course” of disease stability clearly requires better understanding [40].

Of note, we did not identify any baseline characteristic consistently associated with response or large response to single or composite outcomes, except for prior treatment regime. In particular, FEV₁ response or large response was more frequent in patients on prior ICS/LABA maintenance therapy prior to switching to odSITT, reflecting the add-on effect of a LAMA compound on bronchodilation. However, we also noted significant CAT responses in patients escalated from LAMA/LABA dual bronchodilation, supporting the evidence that adding ICS to bronchodilators may also confer relevant benefits in symptom control and/or health status [27, 41]. In clinical practice, these data confirm that COPD patients uncontrolled on either dual therapy may benefit from odSITT.

Our data also do not provide evidence of treatment response depending on baseline eosinophil count (BEC). Although BECs were available only in a small subset of patients (approx. 15% of the population), no trends for treatment response with odSITT were observed for any single or composite outcome, when applying established cutoffs for eosinophils (< 150 cells/µl vs ≥ 150 cells/µl). Our data may therefore support the notion that the number of blood eosinophils should rather be interpreted as a linear predictor of response-likelihood with ICS-containing therapies than a strict yes/no cutoff [1, 42]. In this regard, it is also important to prospectively evaluate the role of eosinophils (or other markers of type 2 inflammation, e.g., FeNO [43]) as predictive biomarker in the context of other established COPD outcomes, beyond exacerbations [44, 45]. However, our lack of response modification by BEC should be interpreted with caution, owing to the small number of subjects with BEC.

Our post hoc analysis of ELLITHE also has some relevant limitations, the lack of a control group already aforementioned. However, unlike in most of the cited RCTs, the primary aim of our post hoc analysis was not to compare relative benefits of treatment arms. Instead, we sought to describe early and long-term response patterns with odSITT, and to compare characteristics of responders with non-responders. For this purpose, we firmly believe the main conclusions of our analysis are fully valid. Due to the non-interventional character of ELLITHE, it is of course difficult to compare absolute responder rates with those observed in other trials. In particular, the lack of a blinded control group and differences in baseline characteristics may impact the likelihood of achieving responses, as these rely on fixed cutoffs. Hawthorne effects may overinflate the proportion of responders in ELLITHE, although similar effects for other PROs have also been noted in well-designed RCTs [46, 47]. This would of course be less relevant to FEV₁ responders, a robust, objective marker of airflow limitation.

By combining responder analyses for multiple outcomes at linked early and longer-term timepoints, the study population available for this post hoc analysis was considerably smaller than the original population (FAS). While similar publications also inherently suffer from this limitation, it is unclear how this affects the findings and main conclusions of these analyses, including ELLITHE. We used a LOCF approach to account for early discontinuations; however, we needed to exclude some subjects from long-term responder analyses in cases where the early response was identical to the LOCF, due to discontinuation. To what extent these subjects potentially differed from the main population remains speculative.

We used the published MCID criteria of validated COPD outcomes to define response. While these isolated MCIDs have usually been validated in COPD populations, the evidence supporting the use of the same MCIDs (or their reciprocal value for deterioration) as thresholds for composite endpoints is far less established [30, 31]. In particular, the cited examples for publications on CII, DS and CID are retrospective in nature, including ELLITHE. Therefore, the prospective value of early and long-term response defined by these composite endpoints remains to be fully established, as well as the exact contribution of each single component to endpoint achievement. Finally, the clinical utility in decision-making needs to be defined, e.g., by using “loss of stability” as a trigger for treatment escalation beyond inhaled triple therapy. We also concede that our responder analysis suffers from the use of fixed cutoffs to define “response” in a disease that is variable in nature. Despite this, the proportion of e.g., CII responders was quite stable in our analysis. Fixed cutoffs may also be problematic when they are applied to broad patient populations, as baseline factors such as disease severity and medication may affect the likelihood of response. Composite outcomes should also ideally be compared to similar concepts of stability, such as COPD control [20].

Conclusions

In summary, this post hoc analysis from ELLITHE describes response patterns to single and combined outcomes with odSITT in COPD patients “at risk” in relation to baseline characteristics and treatments. Despite some overlap in clinical responses, our analysis highlights the justification of a multidimensional approach to evaluate treatment benefits in a holistic way, combining PROs, lung function, and exacerbations. Most patients showed a clinically meaningful response to at least one outcome. Treatment success should be based on multiple established outcomes, whenever practically possible. Importantly, our data clearly indicate that short-term responses of FEV₁ and/or CAT after 3 months odSITT are reasonable predictors of important single or composite long-term responses. Disease stability is achievable for a large group of patients after one year and may represent a useful and practical outcome for future research and practical COPD management. As the field of composite outcomes in patients with COPD is evolving, future research will even go beyond these proposed concepts to include e.g., cardiovascular events as part of a comprehensive endpoint assessment [48].

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Kai M. Beeh, Karl Scheithe, Heike Schmutzler, and Saskia Krüger are authors of the paper. All authors made a significant contribution to the work reported, including conception, study design, execution, acquisition of data, analysis, and interpretation. All authors engaged in drafting, revising or critically reviewing the article. They gave final approval of the version to be published and all versions before submission. All authors have agreed on the journal to which the article has been submitted and agree to be accountable for all aspects of the work.

Funding

The ELLITHE study and the journal’s Rapid Service Fee was funded by BERLIN-CHEMIE AG, Berlin, Germany.

Data Availability

Study synopsis and protocol are available at the BfArM-study registry DRKS. The data analyzed in this study are available from the corresponding author upon reasonable request.

Declarations

Conflict of Interest

Kai M. Beeh has received personal and/or institutional compensation for clinical research, consulting, and/or lecturing fees from AstraZeneca, Bosch Healthcare Solutions, Chiesi, Clario, GSK, Novartis, Menarini/Berlin Chemie, Orion, Sanofi, and Sterna. Kai M. Beeh is an Editor-in-Chief of Pulmonary Therapy. Kai M. Beeh was not involved in the selection of peer reviewers for the manuscript nor any of the subsequent editorial decisions.

Karl Scheithe declares no conflict of interest.

Heike Schmutzler is a current employee of the sponsor, Berlin Chemie.

Saskia Krüger was an employee of the sponsor Berlin Chemie at the time the study was conducted and evaluated.

Ethical Approval

The ELLITHE study was registered under the German Clinical Trials Register (identifier: DRKS00031897) The study was carried out in accordance with Good Clinical Practice guidelines under the provisions of the latest version of the Declaration of Helsinki (2013) and received approval from of the State Chamber of Physicians of Hesse (“Ethikkommission der Landesärztekammer Hessen”) as the coordinating ethics committee of the national chief investigator. The study was registered at the German Clinical Trials Register (DRKS00031897). All patients provided signed informed consent.

Thanking Investigators and Patient Participants

The authors would like to thank all study sites and patients involved in the ELLITHE study.