High-sensitivity cardiac troponin I-guided combination angiotensin receptor blockade and beta blocker therapy to prevent anthracycline cardiotoxicity: the Cardiac CARE RCT

Monitoring and identification of anthracycline-induced cardiotoxicity

Existing research

Previous trials have investigated whether the administration of medications established for the treatment of heart failure can prevent systolic dysfunction in patients receiving chemotherapy. These studies are limited by (1) prescribing therapy to all patients resulting in substantial overtreatment and (2) using either an angiotensin-converting enzyme inhibitor (ACEi)/angiotensin II type I receptor antagonist (ARB) or a B-blocker rather than co-prescription, which has the most robust evidence base for improving function and survival among patients with left ventricular systolic dysfunction. 17

The combination of inhibiting the renin-angiotensin system and blocking β-adrenoreceptors has shown significant benefits in reducing morbidity and mortality in heart failure patients with reduced ejection fraction, including those with chemotherapy-related heart muscle disease. These therapies have also proven effective for asymptomatic left ventricular systolic dysfunction, the most common form of anthracycline cardiotoxicity. For instance, enalapril has been shown to decrease the risk of death and hospitalisation for heart failure in patients with asymptomatic left ventricular dysfunction. 18 Additionally, the combination of carvedilol and an ACE inhibitor has been found to reduce overall mortality in patients with left ventricular dysfunction following acute myocardial infarction. 19 However, randomised controlled trials (RCTs) investigating the potential of neurohormonal blockade for preventing anthracycline cardiotoxicity have yielded mixed results.

In the PRADA study, breast cancer patients receiving anthracycline treatment, with or without trastuzumab, were randomly assigned to receive the ARB candesartan, the beta-blocker metoprolol, or a placebo. 7 Whereas candesartan showed protection against myocardial dysfunction measured by LVEF on cardiac magnetic resonance imaging (MRI), metoprolol did not demonstrate the same effect. The decline in LVEF immediately after chemotherapy was 2.6 percentage points in the placebo group compared with 0.8 percentage points in those receiving candesartan (p = 0.021). However, in a subsequent extended follow-up study, a small decline in LVEF persisted, but there was no significant difference between the treatment groups. 20

A recent meta-analysis21 of 17 trials involving patients receiving anthracycline-based chemotherapy and randomised to neurohormonal blockade showed a 4% higher LVEF in the blockade group, along with a non-significant trend towards fewer clinical events. However, these trials were often single-centre, exhibited significant heterogeneity and displayed evidence of publication bias. Moreover, patient inclusion and randomisation did not take into account stratification for elevated risk of cardiotoxicity, and the trials frequently examined single therapeutic agents. Therefore, it remains unclear whether the potential treatment effect was diluted by including lower-risk patients exposed to different treatments.

The International Cardio oncology Society-One (ICOS-ONE) multicentre trial22 focused on patients treated with anthracycline and randomised them to receive enalapril either upfront or triggered by elevated cardiac troponin levels measured using either a traditional or a high-sensitivity assay. The trial lacked a placebo group, and changes in cardiac function were evaluated based on the development of cardiotoxicity defined by a decrease in LVEF of > 10% and an overall LVEF of < 50% on surveillance echocardiography up to 1 year after chemotherapy. This categorical definition of cardiotoxicity using echocardiography may not capture smaller yet clinically significant changes in LVEF following chemotherapy. The study found no significant difference between the two treatment approaches, but the most notable observation was the low rate of cardiotoxicity, with only three cases among the 273 patients in the trial and no instances of congestive heart failure or cardiovascular death. It remains uncertain whether initiating cardioprotective medications solely based on changes in high-sensitivity cTn concentrations affects the development of left ventricular dysfunction or heart failure.

Rationale for choice of intervention

Candesartan has an established role in the treatment of patients with left ventricular dysfunction, and in the PRADA study it demonstrated an early protective effect on LVEF in this patient population. 7 The avoidance of cough as a side effect is considered an advantage of angiotensin receptor blockade (candesartan) over ACE inhibition in this immunocompromised cancer population. The target dose and dose titration schedule for candesartan were identical in Cardiac CARE and the PRADA study. Carvedilol was tested previously in a similar population of anthracycline-treated breast cancer patients in the Carvedilol for Prevention of Chemotherapy-Related Cardiotoxicity (CECCY) trial. 23 There was no difference in the primary cardiotoxicity end point compared with placebo. Carvedilol treatment was associated with lower circulating cTnI concentrations and less diastolic dysfunction. In the Cardiac CARE trial, the target carvedilol dose of 25 mg twice-daily was the same but the interval between dose titration was much shorter, at 3 days compared with 3 weeks in the CECCY trial. The key difference in approach between the Cardiac CARE trial and previous cardioprotection studies is the focus in the former on maximum neurohormonal blockade on high-risk patients with coprescription of candesartan and carvedilol rather than a single agent.

Risks

Candesartan (ARB) and carvedilol (B-blocker) are widely used in the NHS with an established safety profile and cost-efficacy. No toxicity was reported in the PRADA study, which examined ARB and B-blocker combination in an identical study population to ours. ARBs and B-blockers are in widespread use for hypertension and other serious conditions including heart failure and following myocardial infarction. Patients in the study who were randomised to cardioprotection had their renal function and blood pressure monitored at dose titration clinics supervised by oncology research nurses. We followed a dose titration protocol used in the PRADA study. B-blockers may exacerbate psoriasis and asthma. Common side effects include lethargy and cold peripheries. For these reasons we believed that identifying an at-risk population of breast cancer and non-Hodgkin lymphoma (NHL) patients to target cardioprotection and closer monitoring was key.

Benefits

The long-term follow-up of breast cancer and NHL survivors demonstrates an increase in late cardiac events among this population, including symptomatic heart failure. 1 Clinical studies that record cardiac events during the period of chemotherapy therefore underestimate the magnitude of the problem. Improved survival has led to the recognition of the late impact of cardiac disease related to breast cancer therapies, and consensus statements from the European Society of Cardiology highlight the need for improved monitoring and preventive treatment. 5,12 Current clinical protocols for cardiotoxicity monitoring are suboptimal, aiming to identify cardiac muscle dysfunction after it has become established. Cardiotoxicity is identified by assessing cardiac function, using imaging to measure ejection fraction with echocardiography or radionuclide scans. By demonstrating that early cardiac injury can be detected with a high-sensitivity cardiac troponin I (hs-cTnI) assay, it will be possible to screen out a population of low-risk patients who do not exhibit elevation of this marker during chemotherapy and so do not require surveillance imaging.

Rationale for study

In this study, we used surveillance with high-sensitivity hs-cTnI blood testing in patients receiving cardiotoxic systemic therapy to identify early muscle injury, enabling targeted protective treatment with ARB and B-blocker. Our group’s prospective cohort studies including 6304 patients presenting to hospital with suspected acute coronary syndrome and 155 patients with moderate to severe aortic stenosis have confirmed that low cTnI concentrations (< 5 ng/l) are associated with low risk of future cardiac events. 9,24 In two independent validation cohorts of patients presenting with suspected acute coronary syndrome, we demonstrated that cTnI concentrations of < 5 ng/l had a negative predictive value of 99.4% for myocardial infarction or death at 30 days. 9 We further demonstrated that using the high sensitivity cTnI assay to define a gender-specific upper reference limit (99th centile: ≥ 16 ng/l for women and ≥ 34 ng/l for men) in patients presenting with chest pain identifies a population of women at increased risk of cardiac events who would be missed using older contemporary cTnI assays. 25 Our pilot study for the Cardiac CARE trial in anthracycline-treated breast cancer patients demonstrated an increase in hs-cTnI concentrations with progressive cycles of treatment, with many patients developing circulating concentrations of > 16 ng/l. 10 There is a continuum of heart muscle injury, and the PRADA study illustrated the potential for cardiac MRI to detect smaller, less severe changes in left ventricular function in these patients. 26 To capture patients with lesser degrees of cardiac dysfunction, we selected a hs-cTnI concentration threshold that randomised at least 33% of patients recruited into the study. Patients with hs-cTnI concentrations above the threshold were randomised to receive candesartan and carvedilol or to continue with routine clinical care.

Our research examined the clinical efficacy of ARB and B-blockade in preventing the development of heart muscle failure in breast cancer and NHL patients receiving anthracycline-containing chemotherapy. These treatments have an established role in the treatment of patients with left ventricular dysfunction. ARB and B-blockade have an additive treatment effect, and a strong treatment response has been demonstrated in patients with different heart failure aetiologies, including chemotherapy-related heart muscle disease. Response to ARB and B-blockade includes improved survival, improved symptoms and recovery of LVEF. LVEF is a potent prognostic indicator of heart failure,27,28 and changes resulting from therapy or disease progression are closely associated with outcomes. 29,30 All patients had LVEF monitored with serial cardiac MRI scans. Cardiac MRI is the most precise measure of cardiac function31,32 and provides additional measures of systolic volume and cardiac strain that will inform early mechanisms of chemotherapy-induced cardiac muscle injury. We therefore chose change in LVEF recorded 6 months following completion of anthracycline chemotherapy as the primary end point and surrogate marker of future heart failure events. The hypothesis was that carvedilol and candesartan will prevent the development of cardiac dysfunction in at-risk patients identified by elevated plasma hs-cTnI concentrations. Additional outcomes included treatment effect on ongoing cardiac injury (persistence of cTnI elevation), death and heart failure (definition is provided in version 11.0 of the protocol; link to publication is in Additional information) and a provisional health economic analysis of this selective intervention strategy to prevent chemotherapy-related heart failure.

The event rate was low in the PRADA study, with an average LVEF decline of only 2.6 (95% CI 1.5 to 3.8) percentage points in the placebo group. 7 Sixty per cent of patients received low-dose anthracycline (cumulative epirubicin dose of 240 mg/m²). Around 20% of patients also received trastuzumab, and the final MRI scan was conducted variably according to the end of adjuvant therapy: either immediately following 4–5 months’ treatment with anthracycline or at 15 months following treatment with anthracycline and trastuzumab. We set out to ensure a higher event rate by (1) only approaching patients scheduled for higher anthracycline doses (≥ 300 mg/m² cumulative dose of epirubicin) and (2) restricting randomisation to high-risk patients identified by hs-cTnI elevation.

The hypothesis was that hs-cTnI monitoring would improve the detection of heart muscle injury compared with previous studies with the more insensitive contemporary assays. Increased concentrations of this plasma marker appear before the development of reduced left ventricular function and heart failure. Patients recruited into the study who exhibited a plasma hs-cTnI concentration above a threshold defined by our cTnI monitoring study were randomised to receive both carvedilol and candesartan (B-blocker and ARB) or to continue with standard care.

Our research plan was relevant to NHS cancer care pathways, recruiting patients receiving anthracycline and taking a precision medicine approach by targeting cardioprotective treatments only to at-risk patients.

Primary objective

The primary objective was to determine whether known treatments for heart failure can prevent or reduce myocardial injury and the development of left ventricular systolic dysfunction.

Secondary objectives

The secondary objectives were to establish whether a novel highly sensitive plasma marker of myocardial injury can anticipate the development and monitor the progression of left ventricular systolic dysfunction.

Treatment arm

Standard care plus oral candesartan and carvedilol. Candesartan was started at 8 mg once-daily and increased at a minimum of 3-day intervals to 16 mg and 32 mg once-daily. Carvedilol was initiated simultaneously at 6.25 mg twice-daily and increased to 12.5 mg twice-daily and 25 mg twice-daily. The IMPs were dispensed as close as possible and ideally within 14 days of randomisation and continued until completion or withdrawal from the study. Drug prescription and dose titration visits in all sites were coordinated by oncology research nurses and supervised by oncologists.

Standard care

Standard care alone.

Primary outcome

Change in LVEF on cardiac MRI scan conducted 6 months after final anthracycline dose compared with baseline cardiac MRI scan conducted before anthracycline therapy compared between randomised groups.

Secondary outcomes: efficacy of candesartan and carvedilol treatment

The following biomarker, cardiac imaging and clinical end points were compared between randomised groups:

• hs-cTnI concentrations

  • hs-cTnI concentration change from baseline to 2 months following the completion of anthracycline chemotherapy.

• cardiac MRI

  • change in global longitudinal strain (GLS) and global circumferential strain (GCS) measured with feature tracking on cardiac MRI

  • change in left ventricular mass (LVM), left ventricular end-diastolic volume (LVEDV) and left atrial area (LAA).

Specificity of hs-cTnI assay for cardiotoxicity

The trial aimed to identify patients at high risk of cardiotoxicity from early changes in hs-cTnI concentration on anthracycline treatment. The following comparisons were made within the low-risk non-randomised group. An additional exploratory comparison was made between the non-randomised and the high-risk randomised-to-standard-care groups.

Health economics analysis

Health economics for all enrolled patients. Confirm the feasibility of data capture and assess the quality of data obtainable in this patient population. Provide information that can inform the design of further research including sample size calculation and/or value-of-information analysis.

Investigational medicinal product safety end points

The following were compared between the randomised groups:

• Hypotension: SBP of < 90 mmHg.

• Bradycardia: HR of < 50 bpm.

• Hyperkalaemia (K+ ≥ 5.0 mmol/l).

• Worsening renal function: decrease in eGFR of > 25% from baseline or an increase in creatinine of > 30% from baseline.

• Acute kidney injury: an eGFR drop to < 45 ml/minute/1.73 m².

• Fatigue grade of ≥ 2 using Common Terminology Criteria for Adverse Events (CTCAE) classification.

• New diagnosis of atrial fibrillation.

The protocol did not require all adverse events (AEs) to be recorded and reported in the eCRF. The safety assessments carried out for the trial and the pharmacovigilance reporting requirements are detailed in the protocol.

Primary end point

Patients randomised to cardioprotection or standard care had mean (SD) LVEF 6 months after completion of anthracycline chemotherapy of 65.7% (6.6%) and 64.9% (5.9%), respectively. Adjusted estimated mean change in primary end-point and secondary cardiac MRI measures are shown in Table 5. After adjusting for age, pre-treatment LVEF and planned anthracycline dose, the estimated mean difference in 6-month LVEF between the cardioprotection and standard care groups was –0.4 percentage points (95% CI –3.6 to 2.8 percentage points; p = 0.82). The outcome was no different using a non-adjusted linear regression model (estimated mean difference 0.2 percentage points, 95% CI –3.1 to 3.4 percentage points; p = 0.93).

We examined the per-protocol primary efficacy outcome between randomised groups in a post hoc sensitivity analyses. When only 19 cardioprotection patients who were adherent to treatment were included, there was no change in the primary outcome. The estimated mean difference in the change in 6-month LVEF between the cardioprotection and standard care groups was –0.7 percentage points (95% CI –4.3 to 2.9 percentage points; p = 0.70).

Main secondary end point

In non-randomised patients the baseline and 6-month LVEF (SD) were 69.3% (5.7%) and 66.4% (6.3%), respectively. The estimated mean difference was 2.9 percentage points (95% CI 1.4 to 4.3 percentage points; p = 0.92). The main secondary objective of demonstrating zero percentage-point change with equivalence of ± 2% was not met.

Additional secondary end points

Secondary analysis (see Table 5) identified a difference between the cardioprotection and standard care groups in adjusted LV end-diastolic volume indexed for body surface area of 6.0 ml/m² (95% CI 0.6 to 11.4 ml/m²; p = 0.03). There was no difference between the groups for global longitudinal and circumferential strain, LVM or LAA.

hs-cTnI concentrations from baseline through chemotherapy to 6 months post chemotherapy are presented in Table 6. hs-cTnI concentrations were higher in the randomised groups. Adjusted change in hs-cTnI concentration from baseline to 2 months in cardioprotection and standard care groups was 27.3 ng/l (7.4 ng/l) and 28.8 ng/l (8.8 ng/l) [estimated mean (SE)]. The estimated mean difference was –1.6 ng/l (95% CI –17.6 to 14.4 ng/l; p = 0.85).

Additional analyses

Exploratory comparisons were conducted between the high-risk standard care group and the low-risk non-randomised group. Figure 3 illustrates the increased average hs-cTnI concentrations from baseline to 6 months in the standard care group compared with the non-randomised group. Estimated mean difference in area-under-the-curve quantification of hs-cTnI concentrations from anthracycline treatment cycles 3–6 between groups are given in Table 7. Patients in the non-randomised group had mean (SD) LVEF 6 months after completion of anthracycline chemotherapy of 66.4% (6.3%). The mean (SD) change in LVEF was –2.9% (6.1%) compared with –4.3% (4.4%) in the standard care group. The difference in LVEF decline between groups was not significant. Adjusted estimated mean difference in 6-month LVEF and other cardiac MRI measures of cardiotoxicity between standard care and non-randomised groups are given in Table 8. There was no difference in these measures between the groups.

FIGURE 3.

Increased average hs-cTnI concentrations from baseline to 6 months in the standard care group compared with the non-randomised group area under the curve.

Clinical Trial of Investigational Medicinal Product adherence

Twenty patients (69%) were adherent to cardioprotection treatment, although one stopped candesartan within 2 months and continued with carvedilol alone. Two patients randomised to cardioprotection did not receive any medication owing to intercurrent illness and COVID-19 infection. A further seven (24%) stopped both cardioprotection drugs within 2 months owing to symptoms of light-headedness and dizziness, possibly related to low blood pressure.

Pulse and blood pressure

Pulse and blood pressure from baseline and across all visits after chemotherapy out to 6 months are reported in Table 10. Blood pressure and HR were lower in the cardioprotection treatment group at 6 months. Post hoc analysis confirmed greater reduction in HR at 6 months in the cardioprotection group (estimated mean difference – 11 bpm, 95% CI –18 to –4 bpm; p = 0.003). Although reductions were observed, there was no significant difference in SBP (–7 mmHg, 95% CI –17 to 2.0 mmHg; p = 0.12) and diastolic blood pressure (DBP, –6 mmHg, 95% CI –12 to 0.2 mmHg; p = 0.06) in the cardioprotection treatment group.

Safety end points related to cardioprotection therapy

Safety end points relevant to cardioprotection treatment were recorded for all groups. There was no protocol-defined hypotension or bradycardia at baseline or at 2, 4 and 6 months after chemotherapy.

Hyperkalaemia occurred in 10.3% of non-randomised patients. Hyperkalaemia was more common in the randomised groups: 20.7% of the cardioprotection group and 17.9% of the standard care group. Worsening renal function at any point beyond baseline occurred in 2.7%, 6.9% and 7.1% of the non-randomised, cardioprotection and standard care groups. Two patients in the non-randomised and none in the randomised groups developed acute kidney injury. Fatigue was reported by 12.1%, 3.4% and 25% of the non-randomised, cardioprotection and standard care groups.

Adverse event reporting

Table 11 presents a summary of AE reporting. AEs were more commonly reported in the cardioprotection group, with 71.4% of patients having at least one AE compared with 12.7% non-randomised and 10.3% standard care patients. A total of 62.5% (20 out of 32) of AEs in the cardioprotection group were possibly related to CTIMP, with dizziness and syncope listed in 17 out 20 possibly related AEs and hypotension, palpitation and venous thromboembolism listed for the remaining three AEs that had a possible causal link with the CTIMP.

Quality of life (quality-adjusted life-year) analysis

Table 12 shows the results of a linear regression, where ‘beta’ represents the difference in expected 6-month and 1-year QALYs gained by the intervention group (hs-cTnI-guided cardioprotection) compared with the randomised standard care group, controlling for baseline utility. The regression model follows the methodology outlined by Manca. 37 Table 13 shows the expected adjusted 6-month and 1-year QALYs for each arm as predicted by the regression model. The detailed methodology behind these calculations is given in Appendix 1.

Cost analysis

Table 14 shows the mean values and 95% CIs of NHS-perspective and societal-perspective costs for both randomised groups. Detailed cost and HRU breakdowns, as well as cost data for the whole Cardiac CARE patient population, can be found in Appendix 1. The largest direct (NHS perspective) costs include the costs of inpatient stays, anthracycline therapy, radiotherapy and hospital doctor visits, with large differences between trial arms in each category.

Incremental cost-effectiveness ratio

Based on the study results, from the NHS perspective, hs-cTnI-guided cardioprotection has an ICER of £481,500 per QALY gained compared with standard care among patients marked as high risk by the troponin test with a time horizon of 1 year. From the societal perspective, however, hs-cTnI-guided cardioprotection ‘dominates’ standard care (i.e. the intervention’s costs are lower while the QALY benefits are higher). However, the ICER estimates are limited by the fact that the QALY and cost estimates are not statistically significant.

hs-cTnI concentrations and risk of anthracycline cardiotoxicity

The hs-cTnI concentration thresholds used in the trial protocol failed to define true high- or low- risk populations in this study. Anthracycline chemotherapy was associated with LVEF decline in breast cancer and NHL patients 6 months after completion of chemotherapy. The degree of LVEF decline was smaller than expected in the randomised groups and not significantly greater than in the hypothesised low-risk, non-randomised group. This finding suggests that the correlation between on-treatment hs-cTnI concentrations and subsequent LVEF decline is not strong. That patients in the non-randomised group also had a detectable decline in LVEF indicates future risk of further deterioration.

Prevention of LVEF decline with cardioprotection therapy

We did not find evidence that combined candesartan and carvedilol therapy prevents LVEF reduction or reduces other markers of cardiotoxicity such as GLS and circulating hs-cTnI concentrations. However, the mean deterioration in LVEF with chemotherapy was smaller than anticipated, and no participant’s LVEF deteriorated by > 10% points and the randomised sample was not at higher risk. Therefore, a larger trial would be needed to detect the prevention of this smaller LVEF decline, or higher-risk participants would have to be recruited in future trials.

Impact of patient and public involvement on Cardiac CARE trial

Patients and public representatives were engaged from the start and several protocol design features were influenced by patient recommendations. Our lead patient representative and co-applicant was Professor Abigail Marks. Professor Marks has experience of treatment for early breast cancer including chemotherapy and monitoring for cardiac toxicity. It was helpful to have affirming comments from the patient and public involvement (PPI) committee of the National Institute for Health and Care Research (NIHR) Diagnostic Evaluation Cooperative and the national patient advocacy charity Independent Cancer Patients’ Voice. Both groups recognised the concern that cancer patients have about cardiotoxicity from successful cancer treatment:

… I like the idea of having treatment that could provide protection against developing heart muscle problems. I also like the idea of being able to avoid a dose of radiation.

… an excellent study and good use of drugs which although do have side effects, these are well known and can in most cases be dealt with. I am a great fan for looking at using drugs differently – it is not only economic, but sensible.

Professor Marks participated in some Trial Steering Committee meetings but found attendance challenging owing to her own professional commitments. Difficulties and challenges during the study centred on problems opening sites and access to research infrastructure during the COVID-19 pandemic. Whenever patients were approached they were keen to participate and there was no strong need for PPI during the study.

We hosted a webinar with the Cardiac CARE patients 5 days after presenting the results at the European Society of Cardiology congress on 1 September 2022. Fifteen trial patients signed in to the webinar. The message that cardiotoxicity levels were low in the study is good news for patients and this was well received. Patients asked about ongoing cardiac follow-up in their respective centres and were interested in the possibility of further follow-up cardiac imaging studies. They appreciated the opportunity to learn about the trial results. Several noted that they had not been offered this opportunity in previous clinical trials. A key comment was that results communication about research scan results should be expedient even when the scan is not being conducted to inform patient care. Patients receiving cancer care are, naturally, concerned to receive results even if these confirm that no surprising or concerning findings have been made.

Trial research teams

Edinburgh Cancer Centre and St John’s Hospital: Angus Broom, Peter Hall, Olga Oikonomidou, Fiona Scott, Sarah Beverstock, Clare Brown, Caroline Bruce, Jenny Buxton, David Cameron, Helen Creedon, Tze-en Ding, Fieke Froeling, Julie Gillies, Larry Hayward, Danka Lintott, Robbie McNeil, Desmond Maitland, Caroline Michie, Christopher Mullen, Rachel Nirsimloo, Robert Noble, Aista Oswald, Ashley Pheely, Karin Purshouse, Ahmed Rehan, Kelly Rust, Barbara Stanley, Mark Stares, Monica Szabo Xiangfei Yan, Jennifer Carruthers, Susan Crate, Stephanie Duncan, Lois Eddie, Heather Howie, Harriet Knott, Paula Little, Rachel MacAngus, Louise McFadyen, Jennifer MacPherson, Heather McVicars, Louise Maguire, Elaine Murray, Lorraine Primrose, Ruaridh Buchan, Annette Cooper, Ben Elliot, Carol Falcon, Hannah McKinlay, Claire McNicol and Mahéva Vallet.

Glasgow West of Scotland Cancer Centre and Queen Elizabeth University Hospital: Alan Cameron, Pam McKay, Iain MacPherson, Abdulla Alhasso, Sophie Barrett, Stephen Dobbin, Judith Fraser, Claire Glen, Ailsa Houroyd, Maria McGrath, Michael Manson, John Payne, Iffet Rabnawaz, Mark Rafferty, Matthew Wilson, Kate Beattie, Sophia Campbell, Annette Charlick, Alice Coy, Karen Ferguson, Gail Lynch, Linda McGarry, Pauline McIlroy, Rosie Morrison, Lynne Stirling, Marissa Varley, Martin Ball, Laura Dymock, Jacqui Gourlay, Tracey Hopkins, Evonne McLennan, Maria Nicol, Fiona Savage, Linda Taylor, Nicola Tynan and Rosie Woodward.

Velindre Cancer Centre: Jacinta Abraham, Tasia Aghadiuno, Maleeha Anwer, Peter Barrett-Lee, Claire Bartholomew, Clair Brunner, Charlie Crocker, Gwenllian Edwards, Nida Hassan, Theresa Howe, Ali Hussnain, Aida Kamarudin, Joanita Ocen, Helen Passant, Catherine Pembroke, Srijim Sashidharan, Anshu Wadhawan, Sarah Walters, Simon Waters, Richard Webster, Clare Boobier, Esther Bryant, Karen Davies, Maria Evans, Claire Lang, Katie Tanner, Rachel Williams, Michael Brown, Felix Cenal, Rob Henley, Stephanie Hughes, Colette Kemp, Jayne Richards, Emily Rumney and Zoe Lingham.

University Hospital Wales: Nagah Elmusharaf, Simona Gatto, Emily Hopkins, Zaheer Yousef, Hilary Benny, Charlotte Bloodworth, Jenni Cullen, Emma Williams, Katja Williams, Stuart Cunningham, Juliet Mort, Kathryn Murray and Catrin Truan.

Oxford Cancer and Haematology Centre: Simon Lord, Stephen Booth, Kieran Burton, Hugo De La Pena Gomez, Toby Eyre, Katrina Fordwor, Nicola Gray, Angela Hamblin, Catherine Hildyard, Sarah Jaafar, Rajesh Kharbanda, Murali Kesavan, Andrew King, Heather Leary, Paolo Polzella, Catherine Prodger, Richard Salisbury, Emma Weaver, Henna Wong, Xiao-Yin-Zhang, Gemma Austin, Cecilia Magallano, Sally Springett, Hazel Wynn, Cassandra Baker, Vanessa Ferreira, Aleema Iqbal, Joana Leal, Rebecca Mills, Raj Soundarajan, Joe Terry, Swapna Thummala and Shobana Yoganayagam.

Christie NHS Foundation Trust: Ankhi Barua, Fiona Britton, Rachel Broadbent, Adam Gibb, Jane Gibson, Tiana Kordbacheh, Kim Linton, Bence Nagy, Elizabeth Phillips, Manssor Shaikh, Rohan Shotton, Ann Tivey and Laura Woodhouse, Chiara Dalla Torre, Joanna Dash, Caroline Hamer, Helen Harbord, Nicola Heazell, Megan Jones, Rob Miller, Andrea Whitmore, Juliet Harris, Sophie Price, Abbey Walker and Marie Woolley.

Mount Vernon Cancer Centre: Stephanie Sutherland, Helen Cladd, Elizabeth Lodge, Farhan Ahmed, Rachael Bowie, Nica Da Silva, Victoria Major, Michelle Pillay Achour, Elham Shadman Zanjans and Kirti Thakor.

Milton Keynes University Hospital: Diane Ritchie, Michele Burnham, Chloe Green, Rhiannon James, Amy Oakley, Cheryl Padilla-Harris, Diane Scaletta, Melanie Tucker, Lynn Wren, Cassandra Baker, Sam Bosompem, Snehal Shah, Jeannette Smith and Felicity Williams.

New Victoria Hospital: Louise Humphreys and Rebecca O’Neil.

Trial Manager: Dr Morag MacLean.

Edinburgh Clinical Trials Unit: Anna Foster, Garry Milne, Lynsey Milne, Hannah Rickman and Michelle Stevens.

Edinburgh Imaging and clinical MRI reporting: Suzanne Ashford, David Brian, Annette Cooper, Russell Everett, Alexander Fletcher, Alan Japp, Tom MacGillivray, Alexandra Nicolae, Giorgos Papanastasiou, Scott Semple, Nick Spath and Michelle Williams.

Edinburgh Imaging Core Analysis: Shruti Josh and Trish Singh.

Other: The co-applicants, Professor Nicholas Mills and Professor Dave Newby, provided valuable input throughout the trial.

Governance

Throughout the study oversight was provided from the Trial Steering Committee, Data Monitoring and Ethics Committee, sponsor and patient advisory group. We sincerely thank the following members of the Trial Steering Committee for their support and guidance during the trial. Their contribution was invaluable. We also thank the members of the Data Monitoring and Ethics Committee, for their helpful contribution and scrutiny of the data. The trial management group also thank the Patient Advisory Group for their involvement. Their feedback and input was not only helpful for improving trial documentation but also for reinforcing that patients are at the heart of research and their contribution is essential. This study was co-sponsored by the University of Edinburgh and NHS Lothian, through the ACCORD office, the Queen’s Medical Research Institute, 47 Little France Crescent, Edinburgh, EH16 4TJ. The sponsor reference number is AC16148.

Protocol

The study protocol is published on protocols.io: www.protocols.io/private/A47518A1A5A011EC9B000A58A9FEAC02

It is also available on the ECTU website: www.ed.ac.uk/usher/edinburgh-clinical-trials/our-studies/ukcrc-studies/cardiac-care/cardiac-care-trial

Summary of key variables and population groups

Summaries are provided in Tables 15 and 16.

Patient-reported outcome measure questionnaires

This section contains information about the observation period (in days) captured by the patient-reported outcome measures (PROMs). Baseline measurement was taken on day 1, corresponding to the first day of the first cycle of chemotherapy. Subsequent PROM measurements were conducted at the following time points:

• fourth cycle of chemotherapy

• post-anthracycline visit

• sixth cycle of chemotherapy

• follow-up (2 months)

• follow-up (4 months)

• follow-up (6 months).

The study period ended either on the date of the last available completed PROM form or on the day a patient’s status changes (e.g. from active to deceased or withdrawn). The planned study period captured by the study period captured by the questionnaires consisted of up to 15 weeks of chemotherapy and 6 months of follow-up. The actual period captured and the number of questionnaires completed at each time point are summarised in Tables 17 and 18.

Data cleaning and inspection

Health state utility and quality-adjusted life-years

The purpose of this section is to present 6-month and 1-year QALYs for each trial arm. QALYs are extrapolated from health state utility values (HSUVs), which in turn are derived from EQ-5D-5L profiles recorded in the PROM questionnaires.

Obtaining utility values

Utility values are obtained by converting EQ-5D-5L observation sets into HSUVs using the Crosswalk Index Value Calculator (CIVC). 39 The CIVC provides a ‘validated mapping function to derive utility values for the EQ-5D-5L from the existing EQ-5D (−3L)’, which matches specific 5L profiles with 3L-derived HSUV values and is the preferred method of EQ-5D-5L analysis by the NICE reference case. 36

Calculating quality-adjusted life-years

Daily QALY fractions (HSUV365.25) are calculated and summed for each patient from day 1 (first chemotherapy cycle) to the cut-off day based on the closest available utility observation; this ensures that calculations are consistent with the area-under-the-curve methodology. Furthermore, utility is set to 0 on and after the date of death and observations after the cut-off day (182 or 365) are ignored; utility between the last observation and the cut-off day is assumed not to change (last point carried forward). QALYs are calculated for each trial arm and summarised in Table 19.

Adjusting quality-adjusted life-years

To estimate the difference in QALYs between trial arms, it is important to control for baseline utility. The regression adjustment methodology and justification used in this analysis are outlined by Manca. 40 The formula provided in the paper is as follows:

Where:

• i= patient ID

• β0 = intercept

• β1 = adjusted differential QALY after controlling for imbalance in the mean utility at baseline

• ti = treatment arm dummy variable

• Qib = patient-specific baseline utility

EuroQol-5 Dimensions utility versus Visual Analogue Scale

As part of the EQ-5D-5L instrument, the Cardiac CARE study collected patient-reported health state scores using a Visual Analogue Scale (VAS), with results ranging from 0 (worst health) to 100 (perfect health).

The key difference between utilities derived from EQ-5D data and VAS scores in this study hinges on VAS scores being derived from patients’ subjective perceptions of health and quality of life (QoL), whereas EQ-5D utilities are based on general population preferences. Thus, the function of VAS is to provide ‘an alternative way to elicit an individual’s rating of their own overall current health’. 38

According to the NICE Reference Case, ‘EQ-5D is the preferred measure of health-related quality of life in adults’. 41 Therefore, VAS scores are not included in QALY calculations.

EuroQol-5 Dimensions, five-level version profiles

Figures 4–8 depict the distribution of responses across EQ-5D-5L health domains, categorised by trial arm and time point. Each figure shows the percentage of responses at each level of the domain, which range from 5 (extreme problems) to 1 (no problems). Figures 4–8 show responses for ‘usual activities’, ‘anxiety/depression’, ‘mobility’, ‘pain/discomfort’ and ‘self-care’, respectively.

FIGURE 4.

EuroQol-5 Dimensions, five-level version profiles – usual activities, by time point and trial arm.

FIGURE 5.

EuroQol-5 Dimensions, five-level version profiles – anxiety/depression, by time point and trial arm.

FIGURE 6.

EuroQol-5 Dimensions, five-level version profiles – mobility, by time point and trial arm.

FIGURE 7.

EuroQol-5 Dimensions, five-level version profiles – pain/discomfort, by time point and trial arm.

FIGURE 8.

EuroQol-5 Dimensions, five-level version profiles – self-care, by time point and trial arm.

EuroQol-5 Dimensions health state utility

Figure 9 shows the mean HSUVs at each time point, stratified by randomisation group.

FIGURE 9.

EQ-5D utility scores, by time point and trial arm.

Visual Analogue Scale

Figure 10 shows the mean VAS scores at each time point, stratified by randomisation group.

FIGURE 10.

Visual Analogue Scale scores, by time point and trial arm.

Unadjusted quality-adjusted life-years

Table 19 shows the unadjusted QALYs accumulated by each trial arm during the study period.

Adjusted quality-adjusted life-years

Table 20 shows the regression coefficients, while Table 21 shows estimates of expected adjusted QALYs for each trial arm predicted by the regression model (assuming mean baseline utility). Intervention denotes a binary variable [0 = standard care, 1 = intervention (candesartan and carvedilol)]. Baseline utility refers to utility derived from the PROM questionnaires collected during cycle 1 of chemotherapy.

The p-values in Table 20 indicate that the results are not statistically significant. However, the results confirm the feasibility of data capture, a key objective outlined in the Cardiac CARE protocol.

Healthcare resource use

This section presents 6-month and 1-year costs of all HRU items recorded in the Cardiac CARE clinical trial, as well as differences in resource use patterns across trial arms.

Patient-reported outcome measure items

The Cardiac CARE PROM data include patient-reported resource use split into three categories: hospital based, community based and employment related. Each PROM observation contains a record of all resources used by a patient since their last questionnaire.

Six-month and 1-year costs for each item were calculated by multiplying the number of units used in the period of interest by the corresponding unit cost (see Table 22). For patients whose last PROM observation fell prior to the 6-month or 1-year cut-off, imputation was conducted based on calculating the mean daily cost between the penultimate and final PROM date and assuming that it stayed constant until death or cut-off. All unit costs in Table 22 were inflated to 2022 GBP using Consumer Price Index weights specific to medical goods published by the Office for National Statistics. 42

Candesartan and carvedilol

Candesartan and carvedilol costs were calculated by multiplying the cumulative dose taken of each drug during the study period (recorded in the Compliance SQL table) by its corresponding per-mg unit cost (see Table 22). For patients whose number of days prescribed exceeded the cut-off periods of 6 or 12 months, results were adjusted by the following formula, where unadjusted cost denotes the cost of the total dosage recorded over the study period:

Anthracycline therapy

As outlined in the Cardiac CARE protocol version 11.0, patients underwent 3, 4 or 6 anthracycline therapy cycles lasting 6, 9 or 15 weeks. Since the chemotherapy treatment lasted no longer than 15 weeks, 1-year costs were equal to 6-month costs, as all costs were incurred within the first 6 months.

Based on chemotherapy cost calculators used in the OPTIMA trial,58 all present calculations assumed:

• Patients followed the EC100 regimen for adjuvant and metastatic breast cancer, composed of epirubicin and cyclophosphamide – this assumption was based on expert advice.

• The AnthracyclineDose variable recorded in the SQL database corresponds to epirubicin (mg) but does not include cyclophosphamide.

• The cost of cyclophosphamide was calculated as a function of the epirubicin dosage, based on the proportion of cyclophosphamide in EC100 in the OPTIMA chemotherapy cost template.

• Per-cycle costs consist of:

  • delivery costs

  • specialist nurse review

  • blood tests.

• The supportive medications of EC100 are not included in the costs of anthracycline therapy; these were captured in the Steroids SQL table and calculated separately.

Steroids

Steroid dosage observations in the SQL database are recorded in one of two units:

• milligrams (mg)

• milligrams per metre squared (m²) (of surface area).

The conversion of mg/m² to mg assumes a surface area of 1.75 m² and a relative dose intensity of 0.92, based on the values in the OPTIMA chemotherapy cost template. 58

Radiotherapy

The radiotherapy costs consist of two elements:

1. radiotherapy preparation costs

2. radiotherapy fraction administration costs, multiplied by the number of administered treatments.

In the Cardiac CARE SQL database, dates of the first and last doses (fractions) are recorded for each patient who received radiotherapy, but the number of fractions itself is not recorded.

Radiotherapy cost calculations are based on an estimate of the number of patients’ fractions, which assumes five fractions per week (on weekdays) between the dates of the first and last fractions recorded in the database. The calculation’s assumptions are based on treatments described in the Royal College of Radiologists’ Radiotherapy Dose Fractionation, Third Edition, most of which involve five fractions per week. 59

For two patients, the period between the first and last fraction exceeded 2 months. These patients were assumed to have received 25 fractions, the maximum amount described in the RCR’s document.

Perspectives

All HRU items were aggregated into totals and stratified into NHS perspective and non-NHS perspective costs. The societal perspective consists of the sum of costs in both categories. Each item was grouped according to the following guide:

The NHS perspective includes treatment costs such as medicine costs, administration and monitoring, other health service resource use costs associated with the managing the disease (e.g. GP visits, hospital admissions), and costs of managing adverse events caused by treatment. It does not include patients’ costs of obtaining care such as transportation, over the-counter purchases, co-payments or time off work. 60

Reproduced with permission from York Health Economics Consortium

Regression

To assess the statistical significance of differences in total costs between trial arms, a simple linear regression of trial arm on costs was conducted.

Unit costs

Where possible, costs were sourced from a published paper or database (e.g. NHS reference costs). Where published data were lacking or judged as inapplicable or unrealistic during a review by the Cardiac CARE chief investigator, price weights were supplied by a NHS Lothian finance manager using internal and unpublished data. Where no suitable data were identified, unit costs were based on assumptions (Table 22). All costs are reported in 2022 GBP.

Summary of resource units used

Table 23 summarises HRU units used per patient in each trial arm. Figures 11–14 show the mean per-patient HRU units used within various resource categories, split by trial arm. Figures 11–14 show on-site visits, home visits, telephone calls and scans, respectively. Figures 15 and 16 are histograms that show the distribution of the number of chemotherapy cycles (see Figure 15) and radiotherapy fractions (see Figure 16) per patient in each trial arm.

FIGURE 11.

Mean clinic/on-site visits per patient during the study period, by trial arm.

FIGURE 12.

Mean home visits per patient during the study period, by trial arm.

FIGURE 13.

Mean telephone calls per patient during the study period, by trial arm.

FIGURE 14.

Mean scans per patient during the study period, by trial arm.

FIGURE 15.

Histogram of chemotherapy cycles per patient, by trial arm.

FIGURE 16.

Histogram of radiotherapy fractions per patient, by trial arm.

Healthcare resource use item costs

Total costs

Total costs are summarised in Table 29. Figure 17 shows a breakdown of the largest components of the total costs for each trial arm.

FIGURE 17.

Mean 1-year HRU item cost breakdown, by trial arm.

Regression

In the regression analysis, Intervention is a binary variable where 0 represents standard care and 1 represents the intervention. Thus, the coefficient (Beta) shows the estimated difference in the mean total NHS-perspective per-patient costs (2021 GBP) between the intervention and standard care arms.

Table 30 shows that 6-month costs are estimated to be £67 lower for the intervention arm than the standard care arm by the regression model, while 1-year costs are estimated to be £1926 higher.

Incremental cost-effectiveness ratio

The ICER is a ‘ratio of the difference in the mean costs of a technology compared with the next best alternative to the differences in the mean outcomes’. 36 The formula for calculating the ICER for the candesartan and carvedilol intervention compared with standard care is:

Where:

• C denotes total NHS perspective costs

• E denotes total QALYs

• I denotes the intervention arm (candesartan and carvedilol)

• S denotes the standard care arm.

Estimates of ∆E and ∆C are given by the β₁ regression coefficients for the intervention in Tables 20 and 30, respectively, yielding a 1-year ICER of:

However, Tables 20 and 30 also show that estimates of ∆E and ∆C are not statistically significant, yielding an inconclusive ICER result. Even if the results are indeed accurate and the ICER calculated in this section is a good approximation of the intervention’s cost-utility, then it far exceeds the willingness-to-pay threshold (WTP) of £20,000–30,000 per QALY recommended by NICE. 36

Patient-reported outcome measure period

As shown in Table 17, the mean number of days tracked by PROM questionnaires (i.e. days between the first chemotherapy cycle and the last completed form) differs between trial arms: 269 days for standard care and 308.07 days for the intervention. Two patients in the intervention have over double the PROM days of the standard care arm’s maximum length of 362 days. As the 95% CIs for PROM days overlap (see Table 17), this difference may be due to chance. A future study could further investigate differing PROM response patterns between randomisation groups, which could also explain discrepancies in response rates at different study time points (see Table 18).

EuroQol-5 Dimensions utility and Visual Analogue Scale

In Figure 9, general population perception of health states (EQ-5D) is consistently higher for the intervention arm than for standard care across study time points, whereas the opposite is true for patient-perceived health state VAS scores in Figure 10. In both figures, 95% CIs overlap, suggesting that the differences could be due to random variation. However, one possible explanation for this discrepancy could be the presence of more AEs or serious adverse events (SAEs) in the intervention group, which, although not captured by the EQ-5D, may be perceived as a reduction in life quality by patients themselves.

Nurse visits

Figure 11 shows that, for patients in the standard care arm, the mean number of drop-in nurse clinic visits recorded over the study period (5.39) is greater than the mean number of hospital nurse visits (3.07). For the intervention arm, however, the opposite is true (3.45 drop-in nurse visits vs. 4.69 hospital nurse visits). However, as shown in Table 23, the 95% CIs for mean estimates of hospital versus drop-in nurse visits overlap, so variation between visit types and trial arms could be due to randomness. Nevertheless, a future study could further investigate differences between trial arms of nursing resource requirements.

Patient-reported outcome measure questionnaire design and recommendations

Overall, the results of this analysis confirm data capture feasibility. However, several elements of the PROM questionnaire could be amended to simplify data collection and remove some of the limitations in the current analysis.

The PROM questionnaire data contains information on a wide variety of cost variables, many of which are associated with negligible costs (e.g. non-GP telephone calls). A more streamlined questionnaire could reduce the burden on patients and improve the data collection process.

Another challenge to calculating costs from the Cardiac CARE PROM data stems from unscheduled assessments being recorded as binary variables. To estimate the costs, a questionnaire should collect the number of assessments in addition to binary information on whether or not a given patient had an unscheduled assessment.

References

See References.