PA394

Modelling peripheral blood eosinophils to identify response to

budesonide in COPD: a post-hoc analysis
Mona Bafadhel,1 Stefan Peterson,2 Miguel A De Blas,3 Peter MA Calverley,4 Stephen I Rennard,3,5
Kai Richter,6 Malin Fagerås7
Respiratory Medicine Unit, Nuffield Department of Medicine, University of Oxford, Oxford, UK; 2StatMind, Lund, Sweden; 3AstraZeneca, Cambridge, UK; 4School of Ageing and
1

Chronic Disease, University of Liverpool, Liverpool, UK; 5Department of Internal Medicine, Division of Pulmonary, Critical Care, Allergy, and Sleep Medicine, University of Nebraska
Medical Center, Omaha, NE, USA; 6AstraZeneca, Wedel, Germany; 7AstraZeneca, Mölndal, Sweden

Introduction
• Peripheral blood eosinophil (PBE) count has been suggested as an
easily measured biomarker to predict the benefit of adding inhaled
Figure 1. Exacerbations according to PBE count. (A) Annual exacerbation rate with
95% CI for BUD/FORM 160/4.5 µg and FORM 4.5 µg. (B) Exacerbation rate reduction
Figure 3: Effect of BUD/FORM 160/4.5 µg versus FORM 4.5 µg on
(A) mean pre-bronchodilator FEV1 and (B) SGRQ total score, by PBE count
(rate ratio) for BUD/FORM 160/4.5 µg versus FORM 4.5 µg A

Difference in mean pre-bronchodilator

A 0.25
BUD/FORM 160/4.5 µg FORM 4.5 µg
corticosteroid (ICS) to long-acting 2-agonist (LABA) treatment in

FEV1 at end of treatment (L)

0.20
patients with chronic obstructive pulmonary disease (COPD). 2.5 0.15

• The relationship between PBE and exacerbation risk in patients 0.10 50 mL

improvement
treated with ICS/LABA or LABA alone is still unclear; previous studies 2.0
0.05

Annual exacerbation rate

No effect
have used different, arbitrary PBE cut-offs and have not investigated 0

PBE count as a continuous variable in its relationship with –0.05

1.5
exacerbation risk and treatment response to ICS, or its interaction –0.10

with other patient characteristics.1–3 –0.15

1.0 0 0.20 0.40 0.60 0.80 1.00

Objective PBE count (×109 cells/L)

0.5 B
• To investigate the clinical effect of budesonide/formoterol (BUD/FORM)
versus formoterol (FORM) alone in patients with COPD, modelled by 12
PBE count at study entry, and additional patient characteristics that 0

Difference in SGRQ total score

0 0.20 0.40 0.60 0.80 1.00 8
may determine risk of exacerbations and response to ICS treatment.
PBE count (×109 cells/L)

at end of treatment
4
Methods B
No effect
0
Study selection and design 2.0 100
4-point
improvement
–4
• Studies from the AstraZeneca clinical trial database were reviewed to
identify randomised trials of BUD/FORM fixed-dose combination in Statistically significant –8
treatment effect
patients with COPD, with PBE counts collected at screening. 1.5 75
Exacerbation rate ratio

–12
• Three studies were identified; Table 1 describes their design and 0 0.20 0.40 0.60 0.80 1.00

Patients, %
treatment allocation. Full study design, inclusion/exclusion criteria No effect PBE count (×109 cells/L)
1.0 50
and primary findings have been reported previously.4–6 Data shown are (A) pre-bronchodilator FEV1 mean difference over whole study and (B) SGRQ total score at end of study.
Black dashed line denotes no treatment effect. Red dashed line denotes (A) a 50 mL improvement in FEV1 or (B) the
Table 1. Study design and treatment allocation 50% minimal clinically important difference of –4 in SGRQ total score. Shaded area represents 95% CIs.
reduction
Study Duration Treatment allocation 0.5 25
(two inhalations, twice per day) Table 4: Simplified risk score for COPD exacerbations in the next 12 months
Tashkin et al.6
6 months BUD/FORM pMDI 160/4.5 µg
(NCT00206154) BUD/FORM pMDI 80/4.5 µg Variable Category Score
BUD pMDI 160 µg + FORM DPI 4.5 µg 0 0
BUD pMDI 160 µg Sex Male
0 0.20 0.40 0.60 0.80 1.00
FORM DPI 4.5 µg Female 10
Placebo PBE count (×10 cells/L)
9
ICS use at entrya No
Rennard et al. 4
12 months BUD/FORM pMDI 160/4.5 µg
Histogram demonstrates distribution of PBE counts. Black dashed line denotes no treatment effect (exacerbation rate Yes 10
(NCT00206167) BUD/FORM pMDI 80/4.5 µg
FORM DPI 4.5 µg ratio of 1.0); values 1.0 represent improvement with BUD/FORM 160/4.5 µg versus FORM 4.5 µg. A decrease of 0.5 in LAMA use at entry No
Placebo exacerbation rate ratio (red dashed line) denotes a 50% reduction in exacerbation rate. Shaded area represents 95% CIs.
Yes 10
Sharafkhaneh et al. 5
12 months BUD/FORM pMDI 160/4.5 µg SABA rescue use, inhalations/dayb 6
(NCT00419744) BUD/FORM pMDI 80/4.5 µg Table 3. Exacerbation rate reduction with BUD/FORM 160/4.5 µg versus
FORM DPI 4.5 µg FORM 4.5 µg, according to PBE count 6 10
DPI, dry powder inhaler; pMDI, pressurised metered-dose inhaler. Breathlessness, BCSS score 2
Difference in exacerbation rates with 2 9
Mean PBE count, 109 cells/L
BUD/FORM 160/4.5 µg vs. FORM 4.5 µg
• In all studies, a full blood count with cell differential was collected and Pre-bronchodilator FEV1, % predicted 30

analysed centrally, prior to randomisation. Non-significant 25% increase to 22% reduction 0.01–0.09 30 9
Pack-year history, years 30
• For these post-hoc analyses of the pooled patient populations, the 25% reduction a
0.10–0.19
30 8
free combination of BUD 160 µg and FORM 4.5 µg group was pooled 26–50% reduction 0.20–0.34 Post-bronchodilator FEV1/FVC ratio, % 60
with the fixed-dose combination BUD/FORM 160/4.5 µg group. Data 51–60% reduction 0.35–0.63
60 10
Exacerbations in previous year, n 1
for the pooled BUD/FORM 160/4.5 µg and FORM 4.5 µg groups are a
Mean reduction for 0.10–0.19 109 cells/L. 2–4 10
presented here.a
4 17

Efficacy evaluations Independent predictors of response to ICS/LABA PBE count, 109 cells/L 0.15
0.15 7
• Exacerbations, defined as worsening of COPD that required • Only PBE count (splines) and smoking history (categorical) showed AUC for best fit of prediction model is 0.64 for pooled analysis. Individual AUCs for the studies of Tashkin et al.,
treatment with oral corticosteroids and/or hospitalisation. significant interaction with the treatment effect of BUD/FORM Rennard et al. and Sharafkhaneh et al. are 0.57, 0.63 and 0.60, respectively.
a
Prescribed medication including an ICS component (single-agent ICS not approved in COPD). bRecorded in the evening.
• Pre-bronchodilator forced expiratory volume in 1 second (FEV1) (L). 160/4.5 µg versus FORM 4.5 µg as predictors of exacerbation rate AUC, area under the curve; BCSS, Breathlessness, Cough, and Sputum Scale; SABA, short-acting 2-agonist.
reduction (p=0.0129 and 0.0152, respectively).
• St George’s Respiratory Questionnaire (SGRQ) total score. Figure 4: Risk (%) of at least 1, 2 or 3 exacerbations during 12 months of treatment
• In former smokers, exacerbation rate was independent of PBE count, with BUD/FORM 160/4.5 µg or FORM 4.5 µg by risk score
Statistical analyses and was reduced by 34–39% across PBE counts with BUD/FORM
BUD/FORM 160/4.5 µg FORM 4.5 µg
• The primary outcome of this post-hoc analysis was annual 160/4.5 µg versus FORM 4.5 µg (Figure 2).
1 exacerbation 1 exacerbation
exacerbation rate with treatment and study as factors, with • In current smokers, increasing PBE count was associated with a 2 exacerbations 2 exacerbations
3 exacerbations 3 exacerbations
adjustments made for time (exposure; as a log-transformed offset). higher exacerbation rate in patients treated with FORM 4.5 µg only
100 100
• Treatment effects were evaluated using negative binomial regression and with a greater exacerbation rate reduction in patients treated with
analysis (exacerbation rate) or linear models (FEV1 and SGRQ score), BUD/FORM 160/4.5 µg versus FORM 4.5 µg; with the reduction
Risk of at least 1, 2 or 3 exacerbations

using splines to model continuous variables including PBE counts. reaching 68% at PBE counts 0.6 109 cells/L (Figure 2). 80 80

• Univariate and multiple regression analyses were performed to Figure 2. Annual exacerbation rate for BUD/FORM 160/4.5 µg and FORM 4.5 µg across
in 12 months, %

investigate the independent factors associated with treatment effect smoking status and PBE count. 60 60
Patients, %

interaction and exacerbation rate. BUD/FORM 160/4.5 µg FORM 4.5 µg

– Dominant predictors were identified by backward selection and Former smokers (n=1541) Current smokers (n=1052)
40 40

used to construct a simplified scoring system, weighted for each

2.0
independent predictor to sum to 100,7 to predict exacerbation risk. –68% 20 20

Results
1.5 –55%
Patient baseline characteristics 0 0
Annual exacerbation rate

–39%
• A total of 4528 patients (excluding those allocated to BUD 160 µg) –34% –37% 0 20 40 60 80 100
–34% –31%
–37% Score
were randomised across the three studies4–6 and had available –39%
1.0 –11% –14%
baseline PBE counts with efficacy data. Of these, 2593 patients who –7%
Histogram demonstrates distribution of exacerbation risk scores.

received BUD/FORM 160/4.5 µg or FORM 4.5 µg were included in the

pooled analysis reported here.
0.5
• Patient baseline characteristics are presented in Table 2. Conclusions
Table 2. Baseline demographics and clinical characteristics, by pooled treatment • In patients with COPD, higher PBE count was
groups, in patients included in the analysis 0 associated with a greater clinical response to
Variable BUD/FORM 160/4.5 µg FORM 4.5 µg BUD/FORM over FORM alone in terms of exacerbation
2

6
0.

0.

0.

0.

0.

0.

0.

0.

0.

0.

0.

0.

(n=1436) (n=1157) rate reduction (especially in current smokers), lung



Male, n (%) 950 (66.2) 721 (62.3) PBE count cut-off level (x109 cells/L) function and health status.
Mean age, years (range) 63 (40–86) 63 (40–89)
Former smokers, n (%) 871 (60.7) 670 (57.9) Percentage values show exacerbation rate reduction for BUD/FORM 160/4.5 µg versus FORM 4.5 µg. • These fi ndings suggest that PBE count as a
Mean pack-years (range) 48 (10–200) 48 (10–258) continuous variable could be used to guide treatment
Mean exacerbations
1.7 (0.0–13.0) 1.7 (0.0–12.0)
Lung function decisions with ICS/LABA.
in previous year (range)
COPD severity,a n (%) • Spline modelling corrected for baseline FEV1 demonstrated a
GOLD 2 247 (17.2) 204 (17.6)
signifi cant linear effect of PBE count and FEV1 (p0.001); a PBE
GOLD 3
GOLD 4
818 (57.0)
366 (25.5)
645 (55.7)
299 (25.8) count treatment interaction of mean FEV1 response with BUD/FORM Acknowledgements
Mean pre-bronchodilator FEV1, L (SD) 1.0 (0.4) 1.0 (0.4) 160/4.5 µg over FORM 4.5 µg did not reach statistical signifi cance The authors thank John Horton from AstraZeneca for his contributions to the statistical
55.7 (17.0) 56.0 (16.6) planning of the analyses reported in this poster. Nina Divorty, PhD, of Complete Medical
Mean SGRQ total score (SD)
[n=1355]b [n=1074]b
(p=0.067). A statistically signifi cant improvement of 32 mL occurred
Communications, Macclesfield, UK, funded by AstraZeneca, Cambridge, UK, provided
ICS use at entry, n (%) 924 (64.3) 746 (64.5) at a PBE count of 0.22 10 9 cells/L; a clinically important difference
medical writing support under guidance from the authors and in accordance with Good
LABA use at entry, n (%) 768 (53.5) 601 (51.9) of 50 mL was seen at counts above 0.27 10 9 cells/L (Figure 3A). Publication Practice (GPP3) guidelines (http://www.ismpp.org/gpp3).
PBE count geometric mean,
0.17 (0.03–0.86) 0.16 (0.03–0.75)
109 cells/L (95% CI) Health status
Based on GOLD 2007 guidance document. bn reported for variables with 5% of data missing.
a Disclosures
CI, confidence interval; GOLD, Global Initiative for Chronic Obstructive Lung Disease; SD, standard deviation. • PBE count was associated with a significant treatment effect of
MB: Funded by a National Institute of Health Research (NIHR) Post-doctoral Fellowship
BUD/FORM 160/4.5 µg versus FORM 4.5 µg on SGRQ total score (PDF-2013-06-052) and presents independent research funded by the NIHR; views
Interactions between exacerbation rate, PBE count and treatment (p=0.004). The SGRQ total score minimum clinically important expressed are those of the authors and not necessarily those of the NHS, the NIHR, or
• The pooled mean (95% confidence interval [CI]) annual exacerbation difference of –4 units occurred at a PBE count of 0.48 109 cells/L the Department of Health. Honoraria and travel expenses for attendance at educational
rates for BUD/FORM 160/4.5 µg and FORM 4.5 µg were 0.74 (Figure 3B). meetings from AstraZeneca (AZ), Chiesi, GlaxoSmithKline (GSK), Boehringer Ingelheim
(0.62, 0.87) and 1.05 (0.82, 1.34) [using a separate binomial model with (BI), Novartis and Pfizer. SP: Statistical advisor at StatMind, which received funding from
study as factor and time in study as offset for each treatment group].
Multiple regression analysis and exacerbation risk score AZ to complete the statistical analyses. PMAC: Speaker at meetings supported by AZ,
GSK and BI, and Principal Investigator of studies supported by these companies.
• Multiple regression analysis for dominant variables identified in the
• In patients treated with FORM 4.5 µg alone, exacerbation rate MADB, SIR, KR and MF are employees of AZ.
univariate analysis demonstrated that exacerbation history, female
increased with PBE count, increasing from 0.5 to 1.8 exacerbations
per year over the range of PBE counts (Figure 1A).
sex, ICS or long-acting muscarinic antagonist (LAMA) treatment at References
study entry, pack-years smoked, FEV1 % predicted, FEV1/FVC ratio 1. Pascoe S et al. Lancet Respir Med 2015; 3: 435–442.
• In patients treated with BUD/FORM 160/4.5 µg, exacerbation rate and PBE count were independent predictors of future exacerbation 2. Siddiqui SH et al. Am J Respir Crit Care Med 2015; 192: 523–525.
was consistent across the PBE range (Figure 1A). risk (all p0.001). 3. Watz H et al. Lancet Respir Med 2016; 4: 390–398.
• An exacerbation rate reduction was observed with BUD/FORM • A simplified risk score to calculate risk of exacerbation within 4. Rennard SI et al. Drugs 2009; 69: 549–565.
160/4.5 µg versus FORM 4.5 µg at low PBE counts, with a statistically 5. Sharafkhaneh A et al. Respir Med 2012; 106: 257–268.
12 months was constructed (Table 4), and used to calculate 6. Tashkin DP et al. Drugs 2008; 68: 1975–2000.
and clinically significant reduction of 25% observed at a PBE count of treatment effect of BUD/FORM 160/4.5 µg and FORM 4.5 µg 7. Make BJ et al. Int J Chron Obstruct Pulmon Dis 2015;
0.10 109 cells/L (Figure 1B; Table 3). The magnitude of protection based on exacerbation risk (Figure 4). 10: 201–209.
against exacerbations increased with PBE count, reaching >60% in
• BUD/FORM 160/4.5 µg treatment was associated with reduced risk
patients with the highest counts.
of experiencing 1, 2 and 3 exacerbations at risk scores 20, A PDF version of this poster can be accessed by scanning the
• The distribution of PBE counts is shown as a histogram in Figure 1B. versus FORM 4.5 µg alone. QR code.

Poster presented at the European Respiratory Society International Congress, Milan, Italy, 9–13 September, 2017
a
These analyses have been refined and extended since submission of the abstract, resulting in some differences in the data presented.

AZ87140_Bafadhel_Pst.indd 1 01/09/2017 15:48