If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Division of Cardiovascular Sciences, School of Medical Sciences, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre, University of Manchester, Oxford Road, Manchester, M13 9PLManchester University NHS Foundation Trust, Southmoor Road, Wythenshawe, Manchester, M23 9LT
Division of Cardiovascular Sciences, School of Medical Sciences, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre, University of Manchester, Oxford Road, Manchester, M13 9PLManchester University NHS Foundation Trust, Southmoor Road, Wythenshawe, Manchester, M23 9LT
Liverpool Clinical Trials Centre, Clinical Directorate, Faculty of Health and Life Sciences, University of Liverpool (a member of Liverpool Health Partners), Alder Hey Children's NHS Foundation Trust, Liverpool, L12 2AP
Department of Health Data Sciences, Institute of Population Health, Faculty of Health and Life Sciences, University of Liverpool (a member of Liverpool Health Partners), Block F, Waterhouse Boulevard, 1-5 Brownlow Street, Liverpool, L69 3GL
School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania, USA.Minneapolis Heart Institute East, Saint Paul, Minnesota, USA.Clinical and Translational Science Institute, University of Pittsburgh, Pittsburgh, PA, USA
Address for Correspondence Prof. Christopher A. Miller, Division of Cardiovascular Sciences, School of Medical Sciences, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre, University of Manchester, Oxford Road, Manchester, M13 9PL, Telephone: 0044 161 291 3244. Fax: 0044 161 291 2389
Division of Cardiovascular Sciences, School of Medical Sciences, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre, University of Manchester, Oxford Road, Manchester, M13 9PLManchester University NHS Foundation Trust, Southmoor Road, Wythenshawe, Manchester, M23 9LTWellcome Centre for Cell-Matrix Research, Division of Cell-Matrix Biology & Regenerative Medicine, School of Biology, Faculty of Biology, Medicine & Health, Manchester Academic Health Science Centre, University of Manchester, Oxford Road, Manchester, M13 9PT
Medication adherence in patients with heart failure with preserved ejection fraction (HFpEF) is unclear. This study sought to evaluate treatment adherence in the Pirfenidone in Patients with Heart Failure and Preserved Left Ventricular Ejection Fraction (PIROUETTE) trial.
Methods and Results
Adherence was evaluated through pill counts and diary cards. Univariable and multivariable regression models were used to assess the relationship between adherence and baseline characteristics. Instrumental variable regression was used to estimate the causal effect of pirfenidone treatment duration on myocardial fibrosis.
Complete adherence data was available in 54 of 80 participants completing the trial. Mean adherence with study medication was 94.7% and 96.9% in the pirfenidone and placebo groups respectively. Each additional day of treatment with pirfenidone resulted in a significant decrease in myocardial extracellular volume (–0.004%; 95% confidence interval: -0.007% to -0.001%; p=0.007). Associations with adherence included older age, higher symptom burden, lower body weight and smaller right ventricular size.
Conclusion
Adherence with study medication in the PIROUETTE trial was very high among patients for whom complete adherence data was available. Importantly, each additional day of treatment reduced myocardial fibrosis. Potential predictors of adherence were identified. Implementation of improved methods for assessing adherence is required.
Non-adherence with guideline-directed medications for reduced ejection fraction heart failure (HF) is common and is associated with worsening symptoms and higher rates of hospitalisation and death (
Non-adherence to heart failure medications predicts clinical outcomes: assessment in a single spot urine sample by liquid chromatography-tandem mass spectrometry (results of a prospective multicentre study).
European Journal of Heart Failure.2021; 23: 1182-1190
). Medication adherence in patients with heart failure with preserved ejection fraction (HFpEF) is less clear, in part because of the paucity of effective medical therapies (
The phase II, double-blind, placebo-controlled, randomised trial of Pirfenidone in Patients with Heart Failure and Preserved Left Ventricular Ejection Fraction (PIROUETTE), focussed on patients with HFpEF and myocardial fibrosis. Administration of pirfenidone for 52 weeks significantly reduced myocardial fibrosis (
This study aimed to evaluate treatment adherence in PIROUETTE, identify factors that predict adherence in HFpEF that could be used to direct support for adherence, and understand the impact that missed doses have on treatment efficacy.
Methods
Study design
The PIROUETTE trial design and main results have been published previously (
). Briefly, 94 participants with HFpEF and myocardial fibrosis were randomised to receive pirfenidone or placebo for 52 weeks. The primary outcome was change in myocardial fibrosis, measured using cardiovascular magnetic resonance (CMR) extracellular matrix volume (ECV), from baseline to 52-weeks. The study protocol was approved by a research ethics committee.
Investigational medicinal product
The active treatment was pirfenidone (Esbriet) 2403 mg daily, taken orally as three 267 mg capsules three times per day. The comparator was placebo, taken as three capsules three times per day. Pirfenidone and placebo were both opaque, hard, white to off-white, gelatin capsules, manufactured by Roche Products Limited. Capsules were supplied in 250 ml white high density polyethylene bottles, dispensed to participants every 13 weeks. Treatment dose was titrated, as tolerated, to the target dose of three capsules three times a day over the initial 14 days, resulting in 3,213 capsules in total over 52 weeks.
Adherence assessment
Study treatment adherence was evaluated at each visit through pill counts from returned medication bottles. Adherence information was also provided by participants in diary cards; however, the quality of data provided, as a result of inconsistent diary return and poor completion, precluded meaningful diary card analysis.
Statistical analysis
Only participants with complete pill count data i.e., participants who returned medication bottles at all visits, were included in the adherence analysis. Adherence was defined as the proportion of prescribed capsules that were taken.
Univariable and forward stepwise multivariable regression models were used to assess the relationship between adherence and baseline characteristics. Instrumental variable regression was used to estimate the causal effect of pirfenidone on treatment response, by appropriately allowing for the duration of treatment received (i.e., adjusting for premature treatment discontinuation). All analyses were performed in SAS (Version 9.4, SAS Institute, Inc.; Cary, NC).
Results
Patients
Ninety-four patients were randomised. Baseline characteristics have been described previously (
). Mean age of patients was 78 years, and 46% were female. Mean LVEF was 64%. Median NT-proBNP was 1,104 pg/ml. Mean myocardial ECV was 30.1%. No patients were lost to follow up.
Tolerance
In the pirfenidone group (n=47), 18 patients prematurely discontinued treatment, including 8 patients who withdrew from the trial. In the placebo group (n=47), 9 patients prematurely discontinued treatment, including 2 patients who died, and 4 patients who withdrew from the trial. Treatment-related issues were the reason for premature treatment discontinuation or study withdrawal in 10 patients in the pirfenidone group, and 3 patients in the placebo group.
Adherence
Of the 80 patients who completed the study, 54 had complete pill count data (i.e., 54 patients returned medication bottles at every visit) and were included in the adherence analyses. Selected baseline characteristics split according to completeness of adherence data are summarised in Table 1. Characteristics were very similar in both groups.
Table 1Selected baseline characteristics, split according to participants with complete and incomplete adherence data
Characteristic
Complete (n=54)
Incomplete (n=40)
Overall (n=94)
Allocation, n (%)
Pirfenidone
22 (40.74)
25 (62.5)
47 (50)
Placebo
32 (59.26)
15 (37.5)
47 (50)
Age (years)
78.15 ± 7.67
77.36 ± 6.99
77.82 ± 7.36
Female, n (%)
25 (46.3)
18 (45)
43 (45.74)
Ethnicity, n (%)
White
51 (94.44)
37 (92.5)
88 (93.62)
Asian
3 (5.56)
1 (2.5)
4 (4.26)
Other ethnic group
0 (0)
2 (5)
2 (2.13)
Weight (kg)
83.24 ± 16.48
87.45 ± 16.86
85.03 ± 16.68
BMI (kg/m2)
30.26 ± 5.35
31.80 ± 5.97
30.91 ± 5.64
Atrial fibrillation, n (%)
29 (53.7)
17 (42.5)
46 (48.94)
Hypertension, n (%)
44 (81.48)
35 (87.5)
79 (84.04)
Diabetes, n (%)
14 (25.93)
14 (35)
28 (29.79)
Dyslipidaemia, n (%)
16 (29.63)
6 (15)
22 (23.40)
Previous MI, n (%)
6 (27.27)
7 (30.43)
13 (28.89)
Current or Ex-smoker, n (%)
17 (31.48)
16 (40)
33 (35.11)
Chronic kidney disease, n (%)
30 (55.56)
21 (52.5)
51 (54.26)
6-minute walk test (m)
26.64 ± 11.49
26.25 ± 11.57
26.47 ± 11.46
NYHA Class, n (%)
I
3 (5.56)
2 (5.00)
5 (5.32)
II
25 (46.30)
20 (50.00)
45 (47.87)
III
26 (48.15)
18 (45.00)
44 (46.81)
KCCQ Physical Limitation Score a
52.66 ± 21.25
53.25 ± 23.49
52.91 ± 22.10
KCCQ Symptom Stability Score
49.54 ± 19.11
51.28 ± 15.12
50.27 ± 17.48
KCCQ Total Symptom Score
60.44 ± 23.19
58.23 ± 23.26
59.50 ± 23.12
KCCQ Self-Efficacy Score
76.89 ± 22.12
68.75 ± 28.59
73.39 ± 25.29
KCCQ Quality of Life Score
54.55 ± 24.58
54.38 ± 25.39
54.48 ± 24.79
KCCQ Social Limitation Score
52.83 ± 28.32
53.21 ± 25.71
52.99 ± 27.10
KCCQ Symptom Burden Score
62.27 ± 25.24
61.04 ± 24.27
61.75 ± 24.71
KCCQ Symptom Frequency Score
58.60 ± 23.34
55.42 ± 23.91
57.25 ± 23.51
KCCQ Clinical Summary Score
56.29 ± 19.20
55.58 ± 21.06
55.99 ± 19.90
KCCQ Overall Summary Score
55.31 ± 20.12
54.56 ± 19.65
54.99 ± 19.82
Log transformed NT-proBNP (pg/ml)
7.10 ± 0.92
6.93 ± 0.95
7.03 ± 0.93
Myocardial ECV (%)
30.62 ± 3.03
29.36 ± 2.13
30.08 ± 2.74
Global Longitudinal Strain (%) b
-16.25 ± 3.54
-15.61 ± 3.52
-15.98 ± 3.53
LA volume index (ml/m2)
73.48 ± 18.99
66.32 ± 17.54
70.43 ± 18.63
LA strain - reservoir (%) c
16.29 ± 8.07
17.69 ± 6.89
16.88 ± 7.58
LA strain - conduit (%) c
10.21 ± 4.01
10.75 ± 3.60
10.44 ± 3.83
LA strain - booster (%) d
12.88 ± 4.82
11.78 ± 3.59
12.36 ± 4.26
RV end diastolic volume index (ml/m2)
68.30 ± 15.41
69.45 ± 17.87
68.79 ± 16.42
RV ejection fraction (%)
51.61 ± 9.49
52.18 ± 9.52
51.85 ± 9.46
Data shown as mean ± SD unless specified. BMI – body mass index; ECV – extracellular volume; KCCQ – Kansas City Cardiomyopathy Questionnaire; LA – left atrial; MI – myocardial infarction; NT-proBNP – N-terminal pro B-type natriuretic peptide; NYHA – New York Heart Association; RV – right ventricular. a KCCQ scores range from 0 to 100, with higher scores indicating fewer symptoms (if patients answered ‘Limited for other reasons or did not do’ for a specified number of responses the scores are set to “missing value”). b Measurements were unobtainable in one patient at baseline (in the placebo group). c Measurements were unobtainable in 2 patients at baseline (one in placebo group and one in pirfenidone group). d Measurements unobtainable in 46 patients with atrial fibrillation.
,213), mean number of capsules prescribed per participant was 2,543 ± 887 in the pirfenidone group and 2,983 ± 463 in the placebo group. Mean number of capsules taken was 2,391 ± 832 in the pirfenidone group and 2,888 ± 474 in the placebo group. Mean adherence was 94.7% and 96.9% in the pirfenidone and placebo groups respectively.
Determinants of adherence
Univariable analysis identified older age and lower right ventricular (RV) end diastolic volume (EDV) to be significantly associated with higher adherence (Table 2). There were also non-significant trends towards a higher burden of symptoms and higher adherence, and between lower body weight and higher adherence. In the stepwise selection model, only lower RV EDV was deemed to be associated with adherence.
Table 2Univariable and multivariable regression analyses comparing adherence (%) with baseline characteristics
Baseline Covariate
Univariable model
Multivariable model
β-coefficient (SE)
95% CI
P-value
β-coefficient (SE)
95% CI
P-value
Allocation (Pirfenidone vs Placebo)
-2.21 (1.59)
-5.39, 0.98
0.170
-2.25 (1.53)
-5.31, 0.82
0.147
Age (years)
0.21 (0.10)
0.01, 0.41
0.043
Gender (Female vs Male)
-0.44 (1.59)
-3.63, 2.75
0.782
Ethnicity (White vs Asian)
2.89 (3.44)
-4.02, 9.80
0.405
Weight (kg)
-0.09 (0.05)
-0.18, 0.01
0.068
BMI (kg/m2)
-0.23 (0.15)
-0.52, 0.06
0.121
Atrial fibrillation (Yes vs No)
0.97 (1.59)
-2.21, 4.15
0.544
Hypertension (Yes vs No)
2.43 (2.02)
-1.62, 6.47
0.234
Diabetes (Yes vs No)
0.42 (1.81)
-3.21, 4.06
0.815
Dyslipidaemia (Yes vs No)
0.76 (1.73)
-2.73, 4.24
0.665
Previous MI (Yes vs No)
2.24 (3.01)
-4.05, 8.52
0.467
Smoking status (Current/Ex-smoker vs Never Smoker)
1.50 (1.70)
-1.91, 4.90
0.381
Chronic kidney disease (Yes vs No)
-0.10 (1.60)
-3.31, 3.10
0.948
6-minute walk test (m)
-0.02 (0.07)
-0.16, 0.12
0.774
NYHA Class (II vs I)
-3.37 (3.57)
-10.53, 3.79
0.349
NYHA Class (III vs I)
-3.11 (3.56)
-10.25, 4.04
0.387
KCCQ Physical Limitation Score
0.04 (0.04)
-0.04, 0.12
0.288
KCCQ Symptom Stability Score
0.05 (0.04)
-0.04, 0.13
0.273
KCCQ Total Symptom Score
0.06 (0.03)
-0.00, 0.13
0.068
KCCQ Self-Efficacy Score
0.04 (0.04)
-0.04, 0.11
0.309
KCCQ Quality of Life Score
0.01 (0.03)
-0.06, 0.07
0.837
KCCQ Social Limitation Score
0.02 (0.03)
-0.04, 0.07
0.575
KCCQ Symptom Burden Score
0.05 (0.03)
-0.01, 0.11
0.094
KCCQ Symptom Frequency Score
0.06 (0.03)
-0.01, 0.13
0.070
KCCQ Clinical Summary Score
0.07 (0.04)
-0.01, 0.16
0.087
KCCQ Overall Summary Score
0.04 (0.04)
-0.04, 0.12
0.293
Log NT-proBNP (pg/ml)
0.36 (0.87)
-1.40, 2.11
0.684
Myocardial ECM volume (%)
0.39 (0.26)
-0.13, 0.91
0.139
Global Longitudinal Strain (%)
-0.07 (0.23)
-0.53, 0.38
0.742
LA volume index (ml/m2)
-0.04 (0.04)
-0.12, 0.05
0.394
LA strain - reservoir (%)
-0.13 (0.10)
-0.33, 0.07
0.196
LA strain - conduit (%)
-0.28 (0.20)
-0.68, 0.12
0.161
LA strain - booster (%)
-0.10 (0.27)
-0.66, 0.46
0.715
RV end diastolic volume index (ml/m2)
-0.11 (0.05)
-0.21, -0.01
0.031
-0.11 (0.05)
-0.21, -0.01
0.028
RV ejection fraction (%)
0.01 (0.08)
-0.16, 0.18
0.870
Patients with complete adherence data were included (n=54). Variables for which p-value <0.3 in univariable regression model included within stepwise forward selection multivariable regression model. Abbreviations and missing data as per Table 1.
Impact of treatment duration on treatment efficacy
Mean duration of study treatment (at any dose) for all randomised participants (n = 94) was 260 ± 135 days in the pirfenidone group, and 330 ± 77 days in the placebo group. Instrumental variable regression revealed that each additional day of treatment with pirfenidone resulted in a significant decrease in week 52 myocardial ECV (–0.004%; 95% confidence interval: -0.007% to -0.001%; p=0.007; Table 3).
Table 3Impact of treatment duration on 52 week myocardial extracellular volume
Covariate
Coefficient (95% CI)
Z statistic
p-value
Baseline myocardial ECV (%)
0.96 (0.80, 1.11)
12.24
<0.001
Gender (Female vs Male)
-0.71 (-1.57, 0.14)
-1.64
0.100
Days on pirfenidone
-0.004 (-0.007, -0.001)
-2.72
0.007
Instrumental variable regression of week 52 myocardial extracellular volume (ECV) (%) on number of days on pirfenidone, adjusted for baseline myocardial ECV (%) and gender. Eighty patients were included in the analysis (12 patients withdrew from the study and two died), including 39 in the pirfenidone group and 41 in the placebo group.
This analysis of data from the PIROUETTE trial demonstrates several key findings. Adherence with study treatment was very high among patients with complete adherence data, marginally more so in the placebo group. Determinants of adherence included older age, higher symptom burden, lower body weight and smaller RV size, although only the latter was independently predictive, albeit multivariable analysis was limited by sample size. Demonstrating the importance of treatment adherence, each additional day of treatment with pirfenidone reduced myocardial fibrosis. Finally, the study serves to highlight the difficulties of adherence assessment.
Medication adherence is, clearly, essential for treatment efficacy and safety. The instrumental variable regression analysis in the current study showed that each additional day of treatment with pirfenidone led to a significant reduction in myocardial fibrosis, which corroborates the causal analysis based on total dose received in the main trial analysis. Mean treatment duration in the pirfenidone group was considerably shorter than planned (see below), thus without the high level of adherence observed, it is possible that the intention to treat analysis may not have identified the antifibrotic efficacy, and the paradigm shift may have been lost.
In the Candesartan in Heart failure: Assessment of Reduction in Mortality and morbidity (CHARM) programme, 89% of patients were at least 80% adherent with study medication, with minimal difference between treatment groups (
Adherence to candesartan and placebo and outcomes in chronic heart failure in the CHARM programme: double-blind, randomised, controlled clinical trial.
). Adherence in the current study was similarly very high, likely reflecting the nature of the trial; the follow-up schedule included 11 visits, which is expected to have selected motivated participants, and resulted in regular participant contact. Furthermore, all visits took place at one site, which is likely to have facilitated strong participant-research team relationships.
Univariable associations with adherence included older age, higher symptom burden, lower body weight and smaller RV size. Older age is well known to be associated with increased adherence (
). In non-cardiovascular chronic conditions, higher symptom burden is often associated with lower adherence, although the confounding impact of chronic pain and depression, both of which are themselves associated with lower adherence, is unclear (
). Lower body weight has not previously been shown to associate with higher medication adherence, however, studies evaluating adherence in patients with obesity report a high rate of non-adherence (
). The multivariable analysis was limited by sample size. It is difficult to provide a physiological explanation for the identified association between smaller RV size and higher adherence, and it reflects multiple statistical testing.
Mean duration of study treatment at any dose for all randomised participants was considerably shorter in the pirfenidone group compared to the placebo group. This reflects the greater number of participants that prematurely discontinued treatment in the pirfenidone group, which was predominantly due to treatment-related issues. The side effect profile of pirfenidone, which most commonly includes gastrointestinal, skin and liver adverse events, is well described, and real-world experience in idiopathic pulmonary fibrosis demonstrates an annual discontinuation rate due to treatment-emergent adverse events of 30 to 68% (
). There is an urgent need for novel antifibrotic agents that are more tolerable and easier for patients to take.
The current study highlights the difficulties of assessing adherence. As is commonplace, adherence was measured via pill counts. Despite the aforementioned motivated study population, 32.5% of patients did not return all medication bottles, thus a limitation of the study is that the adherence analysis included only 67.5% of participants completing the study. It is likely that participants who were more compliant with returning medication bottles would also be more treatment compliant, and it is quite possible that participants who prematurely discontinued treatment, or withdrew from the study completely, may have had lower adherence still. Thus, adherence across the whole study population is likely to have been lower than that reported.
How best to measure adherence remains unclear. Medication Event Monitoring Systems, integrated into medication bottle caps to record the time and date that the bottle is opened or closed, are considered a gold standard. However, their associated cost prevents widespread use, and medication bottle opening does not mean the medication is taken. Direct measurement of drug metabolites in urine using liquid chromatography-tandem mass spectrometry does provide an objective assessment of adherence, albeit at a single time point, correlates with clinical measurements, and is predictive of clinical outcomes in patients with reduced ejection fraction HF (
Non-adherence to heart failure medications predicts clinical outcomes: assessment in a single spot urine sample by liquid chromatography-tandem mass spectrometry (results of a prospective multicentre study).
European Journal of Heart Failure.2021; 23: 1182-1190
Limitations of the study include the sample size and the multiple statistical testing in this context, and methods of adherence assessment, as discussed above.
In conclusion, treatment adherence in the PIROUETTE trial was very high among patients with complete adherence data, which was crucial, because each additional day of treatment with pirfenidone reduced myocardial fibrosis. Potential predictors of adherence included older age, higher symptom burden, lower body weight and smaller RV size. Implementation of improved methods for measuring adherence in heart failure is required.
F.S., and C.A.M. had full access to all the data and take responsibility for the integrity of the data and accuracy of data analysis. Study concept and design was provided by all authors. Acquisition, analysis or interpretation of data was carried out by all authors. Statistical analysis was carried out by A.R.H. and S.D. Drafting of the manuscript was performed by F.S. and C.A.M. All authors critically revised the manuscript for important intellectual content. Funding was obtained by C.A.M.
Funding
GAL was funded by a fellowship grant from the National Institute for Health Research. CAM, Advanced Fellowship, NIHR301338 is funded by the National Institute for Health Research (NIHR), and was previously funded by a Clinician Scientist Award (CS-2015-15-003) from NIHR. The views expressed in this publication are those of the authors and not necessarily those of the NIHR, NHS or the UK Department of Health and Social Care. CAM is also supported by a British Heart Foundation Accelerator Award to the University of Manchester (AA/18/4/34221). The investigational medicinal product was gifted by Roche Products Limited. Immunoassay testing equipment and materials were gifted by Roche Diagnostics International Limited. Roche Products Limited and Roche Diagnostics International Limited had no role in study design, and were not involved in the preparation, drafting or editing of this manuscript. Roche Products Limited and Roche Diagnostics International Limited conducted a factual accuracy check of this manuscript, but any decisions to incorporate comments were made solely at the discretion of the authors.
Conflict of interest statement
CAM has served on advisory boards for Novartis, Boehringer Ingelheim and Lilly Alliance, and AstraZeneca, serves as an advisor for HAYA Therapeutics and PureTech Health and has received research support from Amicus Therapeutics, Guerbet Laboratories Limited, Roche and Univar Solutions B.V.
References
Gupta P
Voors AA
Patel P
Lane D
Anker SD
Cleland JGF
et al.
Non-adherence to heart failure medications predicts clinical outcomes: assessment in a single spot urine sample by liquid chromatography-tandem mass spectrometry (results of a prospective multicentre study).
European Journal of Heart Failure.2021; 23: 1182-1190
Adherence to candesartan and placebo and outcomes in chronic heart failure in the CHARM programme: double-blind, randomised, controlled clinical trial.
00068-4&id=fx1.jpg){kind=link}