Characteristics and outcomes of patients with a history of cancer recruited to heart failure trials

Heart failure (HF) therapy trials usually exclude cancer patients. We examined the association between cancer history and outcomes in trial participants with HF and reduced (HFrEF) or preserved ejection fraction (HFpEF).


Introduction
More than 50% of patients now survive longer than 10 years following a diagnosis of cancer. 1 As these patients are living longer, they are at risk of developing age-related conditions such as heart failure (HF). Furthermore, the risk of developing HF may be amplified by agents used to treat cancer. 2 In patients with cancer, the subsequent lifetime risk of cardiovascular (CV) disease may be greater than the residual risk of the initial malignancy or its recurrence. 3 Therefore, these patients are an increasingly relevant group for whom a clear understanding of the competing risks of CV and cancer outcomes is essential to inform treatment decisions.
In a large Danish administrative database, 17% of patients with HF also had a history of cancer. Malignancy was less frequent in age-and sex-matched controls without HF. 4 In a cohort of almost 10 000 American patients, 19.5% had a history of cancer. 5 Despite this, patients with a history of recent or active cancer have either been excluded from or under-represented in clinical trials in HF. This is because the competing risk of death from cancer has been viewed as being excessively large and not modifiable by therapies designed to improve CV outcomes. Although this argument has been used to justify the exclusion of patients with a history of cancer from trials, it is unclear whether this might now be inappropriate. This may be especially relevant to HF trials. Outcomes for patients with HF remain poor, if not poorer, than for many common cancers. 6,7 We examined the association between prior cancer and outcomes in trials enrolling patients with HF and reduced (HFrEF) or preserved ejection fraction (HFpEF). We also examined the reporting of new cancer diagnoses in these trials.

Study populations
Patients from two large, randomized trials in patients with HFrEF and HFpEF were included. The design and results of PARADIGM-HF (Prospective comparison of ARNI with ACEI to Determine Impact on Global Mortality and morbidity in Heart Failure), ATMOSPHERE (Aliskiren Trial to Minimise Outcomes in Patients with Heart Failure), PARAGON-HF (Prospective comparison of ARNI with ARB Global Outcomes in Heart Failure with Preserved Ejection Fraction) and CHARM-Preserved (Effects of Candesartan in Patients with Chronic Heart Failure and Preserved Ejection Fraction) are published. [8][9][10][11] A total of 8442 and 7016 patients were randomized in the PARADIGM-HF 9 and ATMOSPHERE 8 trials, respectively. The patients enrolled were in New York Heart Association (NYHA) functional class II-IV and had a left ventricular ejection fraction (EF) ≤35% (changed from ≤40% initially in PARADIGM-HF by amendment) and elevated natriuretic peptide levels. In PARADIGM-HF, patients were randomly assigned 1:1 to enalapril or sacubitril/valsartan. In ATMOSPHERE, patients were assigned 1:1:1 to enalapril, aliskiren or enalapril plus aliskiren. Data from PARADIGM-HF and ATMOSPHERE were pooled for these analyses.
A total of 3023 and 4796 patients with HFpEF were enrolled in the CHARM-Preserved 11 and PARAGON-HF 10 trials, respectively. In CHARM-Preserved, patients were in NYHA functional class II-IV and had an EF ≥40%. They were randomized 1:1 to candesartan or placebo. For this analysis, we excluded 450 patients from CHARM-Preserved who had a left ventricular EF <45% to ensure a consistent EF threshold with PARAGON-HF. In PARAGON-HF, patients were in NYHA class II-IV and had an EF ≥45%, structural heart disease, and elevated natriuretic peptide levels. Patients were randomized 1:1 to treatment with either sacubitril/valsartan or valsartan. Data from PARAGON-HF and CHARM-Preserved were then pooled for these analyses.
For each of the trials, exclusion criteria with malignancy explicitly mentioned or implicitly relevant in relation to a life-expectancy threshold were as follows -PARADIGM-HF: life expectancy <5 years; ATMOSPHERE: life expectancy <5 years; PARAGON-HF: life expectancy <3 years or any history of cancer within the previous 5 years; CHARM-Preserved: life expectancy <2 years.
Case report forms for each of these trials recorded the presence or absence of a previous cancer diagnosis. Data on the time from diagnosis, site, stage, or histology of cancer were not captured. Patients with a prior history of cancer were compared with patients with no history of cancer. To determine the incidence of cancer during follow-up, we examined new cancer diagnoses recorded as adverse events and serious adverse events in those without a history of cancer at baseline. Data were extracted and grouped based on the primary cancer site. Only patients with a clearly documented diagnosis of malignancy were included to avoid the inclusion of non-malignant tumours or non-melanomatous skin cancers.

Clinical outcomes
The primary outcome in CHARM-Preserved, PARADIGM-HF and ATMOSPHERE was a composite of CV death or HF hospitalization in a time-to-first event analysis, while in PARAGON-HF it was a composite of CV death and total (first and repeat) HF hospitalizations.
In the current analysis, we examined the composite of CV death or HF hospitalization in each trial as well as its components, CV death and HF hospitalization, individually. We also examined death from any cause and non-CV death. HF hospitalization and causes of death were adjudicated by a Clinical Events Committee according to standardized, pre-specified criteria in each trial. [8][9][10][11]

Statistical analyses
Baseline characteristics are presented as means with standard deviations or medians with interquartile ranges for continuous variables, and frequencies with percentages for categorical variables. Patients were analysed according to history of cancer (baseline characteristics were compared using t-tests, Wilcoxon rank-sum tests, and chi-squared tests where appropriate). The rates of all outcomes are presented per 100 patient-years. Cox regression for time-to-first event was used to analyse each outcome and reported as the number of events and hazard ratios (HRs), with robust standard error accounting for clustering within trials, and the resulting 95% confidence intervals (CIs). Competing risk regression using the Fine-Grey method was used to analyse outcomes (to account for the risk of multiple potential competing events) and these outcomes are reported as sub-distribution HRs (SHRs) with 95% CIs. The composite primary outcome and CV death were examined in the presence of the competing risk of non-CV death, and first HF hospitalization was examined in the presence of the competing risk of all-cause death. Non-CV death was examined in the presence of the competing risk of CV death. HRs and SHRs are adjusted for the following variables, which were chosen based on clinical judgement: trial, randomized treatment, age, sex, region, heart rate, systolic blood pressure, body mass index, NYHA functional class, left ventricular EF, previous HF hospitalization, previous myocardial infarction, previous diabetes, smoking history, estimated glomerular filtration rate (eGFR), and N-terminal pro-B-type natriuretic peptide (NT-proBNP) (with missing indicator method used to handle missing eGFR and NT-proBNP values). 12 Poisson regression was used to analyse age-and sex-adjusted incident cancer diagnoses, which are expressed per 100 patient-years.
All analyses were performed using Stata version 16.0 (Stata Corp., College Station, TX, USA). Two-sided p-values <0.05 were considered significant.

Baseline characteristics
Baseline characteristics of patients with and without a history of cancer are shown in Table 1. Overall, there were 15 415 patients in the HFrEF trials (mean age 63.5 years) and 7363 patients in the HFpEF trials (mean age 70.6 years). There were 658 patients in the HFrEF trials with a history of cancer and 624 patients with a history Table 1 Baseline

characteristics according to history of cancer in patients with heart failure with reduced ejection fraction (PARADIGM-HF and ATMOSPHERE trials) and heart failure with preserved ejection fraction (PARAGON-HF and CHARM-Preserved trials) Characteristic
Heart failure with reduced ejection fraction Heart failure with preserved ejection fraction     Table 1).

Heart failure characteristics and investigations
There were no differences in HF symptom burden or markers of congestion between HFrEF patients with and without a history of cancer. HFrEF patients with a history of cancer had a higher EF than those with no cancer history (43.2% vs. 37.7%). By contrast, when compared to HFpEF patients without a history of cancer, those with a history of cancer were less likely to .

Background treatment
Patients with HFrEF and a history of cancer were less likely to receive treatment with beta-blockers, mineralocorticoid receptor antagonists (MRA) and digoxin compared to those without. However, those with a cancer history were more likely to have a pacemaker (21.0% vs. 11.3%) or an implantable cardioverter-defibrillator (23.1% vs. 14.5%) than those without ( Table 1). Patients with HFpEF with a history of cancer were more likely to be receiving diuretics (90.5% vs. 87.4%) and to have a pacemaker (13.0% vs. 8.4%) than those with no cancer history ( Table 1).

Clinical outcomes
The risk of the primary composite outcome of CV death and first HF hospitalization did not differ between patients with or without a history of cancer in either those with HFrEF (adjusted HR 1.03; 95% CI 0.89- 1.20, p = 0.682) or HFpEF (adjusted HR 0.83; 95% CI 0.67-1.03, p = 0.086, respectively) (Tables 2, 3, Figure 1). HFrEF patients with a history of cancer had a higher risk of a first hospitalization for HF with an adjusted HR of 1.28 (95% CI 1.07-1.52, p = 0.007) than those without ( Table 2). There was no difference in first HF hospitalization between HFpEF patients with     Hazard ratios are reported with 95% CIs within parentheses followed by p-value. CI, confidence interval; CV, cardiovascular; HF, heart failure. a Unadjusted analysis was adjusted for randomized treatment and region. b Adjusted for: age, sex, region, heart rate, systolic blood pressure, body mass index, N-terminal pro-B-type natriuretic peptide, New York Heart Association functional class, ejection fraction, estimated glomerular filtration rate, previous hospitalization for HF, prior myocardial infarction or diabetes, and smoking history. Missing indicator method used to handle missing estimated glomerular filtration rate and N-terminal pro-B-type natriuretic peptide. c CV death and HF hospitalization.  Table 3).
There was no association between cancer history and CV death in HFrEF or HFpEF patients (adjusted HR 0.85; 95% CI 0.70-1.04, p = 0.106 and adjusted HR 0.72; 95% CI 0.50-1.05, p = 0.085, respectively) (Tables 2, 3). The risk of non-CV death was higher in HFrEF patients with a history of cancer (adjusted HR 1.57; 95% CI 1. 16-2.12, p = 0.003) ( Table 2), but there was no difference in risk of non-CV death in HFpEF patients with a cancer history (adjusted HR 1.09; 95% CI 0.75-1.59, p = 0.639) ( Table 3). Consequently, there was no association between cancer history and all-cause mortality in either HFrEF patients (adjusted HR 1.00; 95% CI 0.85- 1.18, p = 0.983) or HFpEF patients (adjusted HR 0.87; 95% CI 0.67- 1.13, p = 0.306). The risk of cancer death was higher in HFrEF patients with a history of cancer than those without (adjusted HR 2.03; 95% CI 1.45-3.64, p < 0.001) ( Table 2), but there was no difference in the risk of cancer death in HFpEF patients with a cancer history (adjusted HR 1.16; 95% CI 0.59-2.27, p = 0.670) ( Table 3). These findings were similar when considering all HF patients together (online supplementary  Table S2). After accounting for the competing risks of non-CV death for the primary outcome and CV death, all-cause death for first HF hospitalization, and CV death for non-CV death, the associations were similar for HFrEF and HFpEF patients (online supplementary  Table S3).

Incident cancer diagnosis
The numbers of new cancers reported during trial follow-up are summarized in Figure 2. Those with a history of cancer were excluded from this analysis to ensure only new cancers were analysed. There were 789 reports of cancer: 538 in the HFrEF trials (PARADIGM-HF and ATMOSPHERE), and 251 in the HFpEF trials (PARAGON-HF and CHARM-Preserved The most common cancers in both trials were gastrointestinal, lung and prostate, with smaller proportions of patients diagnosed with renal, pancreatic, hepatocellular, and haematological cancer. Gynaecological cancers accounted for a greater proportion of incident cancers in HFpEF than HFrEF patients (6.8% vs. 0.7%), but the frequency of other malignancies was similar between groups ( Figure 2).

Discussion
In this analysis of patients with HF enrolled in clinical trials, we found that a history of cancer was associated with a higher risk of HF hospitalization and non-CV death in HFrEF trial participants. This association persisted even after adjusting for differences in characteristics between those with and without cancer. However, there was no difference in all-cause death because a higher risk of non-CV death was offset by a trend to a lower risk of CV death. A history of cancer was not associated with any difference in clinical outcomes in HFpEF patients after adjusting for baseline characteristics. These findings may have implications for future HF trial design. Compared to the general population, cancer survivors have a higher risk of CV disease, including HF. 13,14 As cancer survivorship grows, understanding the efficacy and safety of HF therapies in this population is increasingly important. Guidelines or position statements addressing the management of CV problems in cancer survivors are often based on limited evidence. 15,16 The number of patients with cancer in the trials underpinning guideline recommendations for angiotensin-converting enzyme inhibitors, beta-blockers and MRA in HFrEF has not been reported. [17][18][19][20][21][22] However, at the time those trials were conducted, it is likely that malignancy was widely considered to be a life-limiting condition, thus excluding those patients (online supplementary Table S4). In the context of modern cancer therapy and dramatic improvements in survival for a wide range of malignancies, 23 recent trials have refined exclusion criteria making many patients with a history of cancer eligible for participation. While there is some variability in the specifics of cancer-related inclusion and exclusion criteria in these trials, patients with a history of cancer remain under-represented: up to 20% of 'real-world' patients with HF also have a history of cancer 4,5 while 5.6% of patients recruited into the trials we examined here had a past history of cancer. Irrespective of formal exclusion criteria, it is conceivable that investigators are less likely to consider discussing HF trial participation with patients who have a history of cancer. Whether patients who have had cancer are more or less likely to consent to participation in HF trials is also not known. In addition to the under-representation of cancer patients in these trials, cancer mortality is not routinely reported. Furthermore, there is a ubiquitous lack of information relating to the class of anti-cancer therapies received by patients with previous cancer in the trials analysed here, as well as by those in other contemporary HF trials (online supplementary Table S4).
Given the recent rapid advances in the development of anti-cancer drugs which exert both anti-neoplastic and potential cardiotoxic effects via a broad range of mechanisms, recording this information is essential to refining our understanding of the implications of cancer treatments upon HF outcomes. The higher risk of first HF hospitalization in HFrEF trial participants with cancer may be due to differences in baseline .
demographics, but the association persisted after extensive adjustment, including for natriuretic peptides. However, HFrEF trial participants with a cancer history were less likely to receive HF therapies, including beta-blockers, MRAs and digoxin than those with no cancer history. This may explain why these patients were at higher risk of HF hospitalization. Additionally, there were no differences in HF therapies in HFpEF trial participants with and without cancer and no differences in HF hospitalization. Although the risk of non-CV death was only higher in HFrEF trial participants with cancer, the risk of non-CV death in all HFpEF trial participants was so high that it likely obscures any enhanced risk from a history of cancer in these patients. Although we demonstrated a potentially unsurprising higher risk of cancer death in HFrEF trial participants, this finding was not replicated in HFpEF trial participants. However, this analysis was under-powered. How might these findings be used to inform future trial design and analysis? Given the finding that there were higher rates of HF hospitalizations in trial participants with HFrEF and a history of cancer, better representation of patients with cancer may help with event accrual. This is particularly relevant in HFpEF where recurrent HF hospitalizations are a feature of the disease and the inclusion of recurrent events into the primary outcome of recent trials of therapies in HFpEF such as PARAGON-HF 10 has become commonplace and is increasingly used in HFrEF trials. 24,25 However, this would require HF treatments to be as effective at reducing HF hospitalizations in patients with cancer as those without, which is an area where data remain scarce. These findings may be counterintuitive to the previous view that a higher risk of non-CV death among patients with cancer would lower the rates of modifiable non-fatal and fatal events in trials. However, in our HFpEF cohort there was no association between cancer history and HF hospitalizations or CV death and the association with higher non-CV death disappeared after adjustment. It would