Classification, prevalence, and outcomes of anticancer therapy-induced cardiotoxicity: the CARDIOTOX registry

Abstract

Aim

Cardiotoxicity (CTox) is a major side effect of cancer therapies, but uniform diagnostic criteria to guide clinical and research practices are lacking.

Methods and results

We prospectively studied 865 patients, aged 54.7 ± 13.9; 16.3% men, scheduled for anticancer therapy related with moderate/high CTox risk. Four groups of progressive myocardial damage/dysfunction were considered according to current guidelines: normal, normal biomarkers (high-sensitivity troponin T and N-terminal natriuretic pro-peptide), and left ventricular (LV) function; mild, abnormal biomarkers, and/or LV dysfunction (LVD) maintaining an LV ejection fraction (LVEF) ≥50%; moderate, LVD with LVEF 40–49%; and severe, LVD with LVEF ≤40% or symptomatic heart failure. Cardiotoxicity was defined as new or worsening of myocardial damage/ventricular function from baseline during follow-up. Patients were followed for a median of 24 months. Cardiotoxicity was identified in 37.5% patients during follow-up [95% confidence interval (CI) 34.22–40.8%], 31.6% with mild, 2.8% moderate, and 3.1% with severe myocardial damage/dysfunction. The mortality rate in the severe CTox group was 22.9 deaths per 100 patients-year vs. 2.3 deaths per 100 patients-year in the rest of groups, hazard ratio of 10.2 (95% CI 5.5–19.2) (P < 0.001).

Conclusions

The majority of patients present objective data of myocardial injury/dysfunction during or after cancer therapy. Nevertheless, severe CTox, with a strong prognostic relationship, was comparatively rare. This should be reflected in protocols for clinical and research practices.

Open in new tabDownload slide

See page 1730 for the editorial comment on this article (doi: 10.1093/eurheartj/ehaa013)

Introduction

Cardiotoxicity (CTox) has long been recognized as a major side effect of chemotherapy in patients with cancer.^1–4 New, more effective therapies and some forms of radiotherapy may also have a variety of cardiovascular (CV) secondary effects, in particular, left ventricular dysfunction (LVD) and heart failure (HF).^5–10 Early diagnosis and treatment of HF increase the chance of complete LV function recovery,^{11  ,  12} highlighting the relevance of developing new techniques and protocols for early management of cancer therapy-induced LVD, commonly named CTox, and a multidisciplinary clinical approach throughout the cancer process.^13–15 Several definitions of CTox have been proposed,^16–18 somehow different from the standard definition of LVD and HF in major guidelines.^{19  ,  20}

To better understand the relationship among cancer, cancer therapy, and CV disease, we conducted a prospective, multicentre registry (CARDIOvascular TOXicity induced by cancer-related therapies; CARDIOTOX registry). The main objective of the registry was the risk assessment and early diagnosis of CTox. The aim of the present study is to determine (i) the prevalence of common clinical, biochemical, and echocardiographic (ECHO) parameters currently accepted as indicative of CTox and identified after initiation of cancer therapies and (ii) their relationship with HF criteria and treatment recommendations in current guidelines.^{19  ,  20}

Patients and methods

Study design

The CARDIOTOX registry (Clinical trials identifier NCT02039622) is a prospective multicentre study aiming at identifying the factors related with risk of cancer therapy CTox and assessing the utility of clinical, biochemical, and ECHO parameters for the early detection of CV disease in patients treated with cancer therapies as well as the possible factors related with prognosis and the recovery of LV function.

The study was approved by the ethics committee at La Paz Hospital and collaborative hospitals and certified by the Health Authority of the Madrid Autonomic Community. All patients signed an informed consent.

Patients

A total of 1324 adult patients receiving cancer therapies previously associated with moderate/high risk for CV toxicity (reported incidence of cardiac toxicity ≥2%) (Supplementary material online, Table S1) and with an expected life survival of >6 months were prospectively included in the registry, from April 2012 to October 2017. Patient selection and oncologic treatment were determined by the responsible oncologist or haematologist trained as investigator in the CARDIOTOX registry. Seven hospitals participated in the study including patients according to a pre-established protocol. Patients with a previous history of cancer, chemotherapy, radiation therapy, or with known CV disease were not excluded. Patients were treated following cancer and HF guidelines according to the best criteria of the patient responsible physician. Special care was given to the identification and control of CV risk factors. To improve the precision of CTox definitions, patients with pre-existing symptomatic HF or left ventricular ejection fraction (LVEF) <40%, patients in whom baseline biomarkers and/or baseline ECHO data were not complete and those who died before 3 months were excluded. A total of 865 patients were considered for this analysis (Figure 1).

Clinical data, blood samples, and ECHO parameters were prospectively collected according to protocol, at baseline before cancer therapy and then at 3 weeks, 3 months, 6 months, 1 year, 1.5 years, and 2 years after initiation of cancer therapy.

Clinical variables included age, sex, CV risk factors, previous CV history, previous cancer treatments, and current cancer diagnosis and treatment. The cumulative dose of chemotherapy and radiotherapy was measured in every visit.

Blood samples to determine plasma levels of high-sensitivity troponin T (hs-cTnT) and N-terminal natriuretic pro-peptide (NT-proBNP) were processed in a central laboratory (La Paz Hospital) using standardized methods. Upper normal limit for hs-cTnT was the 99th percentile (women 9 pg/mL; men 16 pg/mL). For NT-proBNP, the upper normal limit was 125 pg/mL, according to European Society of Cardiology (ESC) HF guidelines¹⁹; however, as NT-proBNP concentration correlates more strongly with age, for patients older than 75 years this value was considered as 450 pg/mL.²¹

Echocardiograms were obtained at each participating hospital by a cardiologist with experience in advanced echocardiography and trained for the requirements of the study, using local echocardiographers [GE Vivid E9 (Vingmed Ultrasound, Horten, Norway) or iE33 or EPIC 7 (Philips Medical Systems, Andover, MA, USA)] and following current recommendations for cardiac chamber quantification in adults.^22–24

Patients were followed in each centre by a dedicated cardio-oncology team. Data were centrally reviewed for quality by participating cardiologists/oncologist at La Paz Hospital. All cases with suspected myocardial toxicity were reviewed and centrally adjudicated. For LVEF, several cut-offs were considered following the ESC-HF guidelines¹⁹ and current consensus documents for the evaluation of CTox.^16–18 (Table 1).

Study data were collected and managed using REDCap electronic data capture tools hosted at IdiPaz Research Institute, Madrid.²⁵

Grading of myocardial injury/dysfunction

Myocardial injury/dysfunction was defined by the presence of abnormal values of cardiac biomarkers, LV function parameters, or clinical symptoms of HF. Several mutually exclusive degrees of myocardial injury/dysfunction that may require different treatment recommendations according to clinical practice guidelines were defined as illustrated in Figure 2.^{19  ,  20}

Normal

No evidence of myocardial injury/dysfunction. Asymptomatic patients with normal biomarkers and LV function parameters.

Mild

Asymptomatic patients with LVEF ≥50% with elevated biomarkers or at least one additional abnormal echo parameter (increased LVESV, LAA >30 cm², 10% decrease of LVEF to an LVEF <53%, average E/Eʹ >14, global longitudinal strain (GLS) more than −18%, 15% relative reduction of GLS from baseline).

Moderate

Asymptomatic patients with LVEF ≥40% and <50% with or without biomarker increase or other LV function abnormalities.

Severe

Patients with asymptomatic LVEF <40% or clinical HF. Heart failure was defined as following: HF with reduced ejection fraction (HFrEF): HF symptoms/signs and LVEF <40%; HF with mid-range ejection fraction (HFmrEF): symptoms/signs of HF with elevated NT-proBNP, LVEF 40–49% and at least one additional criterion (enlarged LA, LV hypertrophy, or other relevant diastolic function parameters); and HF with preserved ejection fraction (HFpEF): in presence of symptoms/signs of HF, elevated NT-proBNP, LVEF ≥50%, and at least one additional criterion (enlarged LA, LV hypertrophy, or other diastolic dysfunction parameters).¹⁹

Definition and grading of cardiotoxicity

Cardiotoxicity was defined in presence of new or worsening myocardial damage/dysfunction during follow-up, from normal to mild-moderate or severe, mild to moderate or severe, or moderate to severe. Cardiotoxicity was classified as mild, moderate, or severe according to the worst myocardial injury/dysfunction observed during follow-up.

In addition to this new CTox definition and classification, we explored the prevalence of abnormal biomarkers and LVD parameters using a recently published classification, based on the cardio-oncology consult experience at the Royal Brompton Hospital.¹⁵ In brief, they proposed a practical clinical strategy aimed to improve CV prognosis and cancer treatment continuation in high-risk patients. They described six different CTox risk categories: (i) early biochemical cardiotoxicity: new BNP or troponin-I rise but with normal cardiac imaging; (ii) early functional cardiotoxicity: new reduction in GLS or grade III–IV diastolic dysfunction and normal biomarkers; (iii) early mixed cardiotoxicity: normal LVEF with abnormal biomarkers and GLS/diastolic dysfunction; (iv) symptomatic HFpEF; (v) asymptomatic LVD: new LVEF reduction to <50%, or a reduction in LVEF >10% to an LVEF <55%; and (vi) symptomatic LVD: symptomatic reduction in LVEF <50%, or a reduction in LVEF >10% to an LVEF <55%.

Cardiotoxicity hard endpoint

Generally accepted as a composite endpoint in HF morbidity/mortality trials,²⁶ defined by all-cause death, CV death, or HF hospitalization receiving intravenous diuretics or other intravenous HF therapies.

Statistical analysis

Baseline characteristics, including demographics, medical history, type of cancer, previous chemotherapy, and radiotherapy were summarized using mean and standard deviation for continuous variables and frequencies and percentages for categorical ones. Non-normal distributed continuous variables were summarized with medians and interquartile range. Confidence intervals of proportions of different CTX groups were calculated using Wilson’s method. Sankey diagrams were developed depicting the flow of patients through categories between baseline and follow-up. Differences between all groups were tested with analysis of the variance or Kruskal–Wallis rank test for continuous variables and with the Pearson’s χ² or Fisher’s exact test for categorical data. Event rates of myocardial injury/dysfunction and mortality were estimated using Kaplan–Meier methods. A Cox-proportional hazards model was used to compare the all-cause-mortality rates between myocardial injury/dysfunction groups. An overall alpha-level of 0.05 was used as a cut-point for statistical significance and all statistical tests were two-sided. All data were analysed using STATA v.15 statistical software (StataCorp LLC. 2017, College Station, TX, USA).

Results

A total of 865 patients were included with a median follow-up of 24.1 months (CI 22.1-24.9 months), accounting for a total of 5058 ECHO studies. Mean age was 54.7 ± 13.9 and 16% were men. Patient demographic characteristics, cancer diagnosis, and type of cancer therapy are detailed in Table 1. Eleven patients (1.3%) had a previous history of myocardial infarction; 42 (4.9%) received previous anticancer drugs; and 4 (0.5%) were previously diagnosed of CTox, (none with LVEF <50% at inclusion). Forty-three (4.9%) had already received at least one dose of chemotherapy before the first echocardiogram and blood sampling. Breast cancer was the most common type of cancer (65.7%), followed by non-Hodgkin’s lymphoma (15.4%) and Hodgkin’s disease (5.1%). Myocardial damage/dysfunction abnormalities at baseline are also detailed in Table 1. In nine patients (1%), LVEF was ≥40% <50%, GLS was more than-18% in 147 (28.3%), and none presented an LVEF <40%.

Cancer therapy and HF-related medications during follow-up are detailed in Table 2. In total, 731 patients (84.5%) received anthracyclines, 177 patients (20.5%) anti-HER2 therapy, and 140 patients (16.2%) both treatments. About 27.3% received left breast radiotherapy and 2.7% mediastinal radiotherapy.

Patients with CTox were older and more likely to have a previous history of cancer or previous anticancer drugs or radiotherapy and presented with a lower LVEF and GLS at baseline (Table 1).

Biomarkers abnormalities and echocardiographic parameters during follow-up

Detailed information of studied parameters is included in Supplementary material online, Table S2. Many of the biomarker and ECHO abnormalities were transient, with a peak around 6 months of follow-up. Abnormal values of biomarkers at any time point during follow-up was found in 78.4% of the patients; 2 dimensional echocardiography (2D) abnormalities in 64.6%; advanced ECHO abnormalities in 79.5%. Supplementary material online, Figure S1 illustrates the prevalence of abnormal values of myocardial injury and LVD during follow-up. Figure 3 illustrates the cumulative prevalence of representative myocardial damage/dysfunction parameters during follow-up.

Myocardial dysfunction and cardiotoxicity during follow-up

The prevalence and different grades of myocardial impairment through the study are detailed in Table 3. Sixteen patients (1.8%, 95% CI: 1.1–3%) did not present any clinical, analytical, or ECHO abnormality during the follow-up; 792 patients (91.6%, 95% CI: 89.5–93.2%) presented only mild myocardial damage; 30 patients (3.5%, 95% CI: 2.4–4.9%) moderate; and 27 patients (3.1%, 95% CI: 2.2–4.5%) severe myocardial damage through the follow-up (Figure 4A). The number of patients in each Royal Brompton Hospital myocardial toxicity class at baseline and through the follow-up is shown in Figure 4B.

Defining cardiotoxicity as a worsening of myocardial damage, 541 (62.2%, 95% CI 59.3–65.7%) did not suffer an impairment during follow-up (No CTox); 273 (31.6%, 95% CI 28.6–34.7%) fulfilled criteria for mild CTox; 24 (2.8%, 95% CI 1.9–4.1%) moderate CTox; and 27 (3.1%, 95% CI 2.2–4.5%) severe CTox.

Mortality

Fifty-four patients (6.2%) died of any cause during follow-up. Cardiovascular death occurred in four patients (0.4%), all caused by HF. All-cause mortality was higher in patients with severe CTox (48.15%) than patients with none (5.5%), mild (3.7%), or moderate (4.2%) CTox (P < 0.001). The odds ratio for all-cause death for severe CTox was 15.8 (95% CI 6.8–36.6). The mortality rate in the severe CTox group was 22.9 deaths per 100 patients-year vs. 2.3 deaths per 100 patients-year in the rest of groups, hazard ratio of 10.2 (95% CI 5.5–19.2) (P < 0.001) (Figure 5A). All-cause mortality rates by the Royal Brompton Hospital CTox classes are shown in Figure 5B.

Discussion

Incidence of cardiotoxicity

Using a new definition of CTox, we found a high incidence (37.5%) of patients with worsening ventricular function during high-risk chemotherapy (Table 3 and Figure 4). However, functional abnormalities considered as clear targets for HF evidence-based treatment recommendations^{19  ,  20} were much less frequent; severe CTox with asymptomatic LVEF <40% was only present in 6 (0.7%) and only 21 (2.4%) fulfilled the ESC clinical HF criteria.¹⁹

In previous studies, the incidence of CTox varied depending on the selection of patients, cancer therapies, the methodology for identifying LVD and, in particular, the definition of CTox.^{1  ,  2  ,  11  ,  12  ,  15  ,  16}

The relatively low prevalence of severe CTox in this moderate/high-risk population study may be partially explained by the exclusion of patients with previous HF and severe LVD. Besides, the management and follow-up of cancer patients at risk for CTox in the context of an integrated cardio-oncology service may considerably improve clinical outcomes.¹⁵ It is plausible that early identification and treatment of CV risk factors as well as milder, asymptomatic forms of CTox may delay further deterioration of ventricular function and severe CTox might have been higher with a longer follow-up period, but the majority of ventricular function abnormalities were identified during the first months after initiation of chemotherapy, as in other series.^{11  ,  12} Unfortunately, the nature of this registry does not allow determining the value of HF therapies used during follow-up.

Identification of myocardial damage and dysfunction

All variables of myocardial injury/dysfunction used have been applied in contemporary studies and proposed by guidelines.^{16–20  ,  22  ,  23} Abnormal values of cardiac troponins have been related with poor prognosis in cancer patients.^{11  ,  12  ,  27–30} The ECHO parameters included in this study have also been related with prognosis in different HF scenarios, including cancer.^{19  ,  31–36}

Recent consensus statements in cardio-oncology recommend serial ECHO monitoring of LVEF with the best available method to identify changes in LV function leading to subsequent therapeutic decisions.^16–18 2DE LVEF has low sensitivity for the detection of small changes in LV function, and a high test–re-test variability^{23  ,  37  ,  38} and, in experienced hands, 3 dimensional echcocardiography (3D) is the preferred technique for the longitudinal monitoring of cancer patients^{16–18  ,  23} due to the poor availability of cardiac magnetic resonance imaging outside academic centres.

A growing body of literature supports the use of myocardial deformation analysis in patients receiving cancer therapy for early detection of myocardial damage throughout the cancer process.^{32  ,  33  ,  36} Technologies such as speckle tracking provide accurate information in the early phases of myocardial diseases by measuring GLS²³ and recent industry standardization process helps in minimizing inter-vendor differences, leading to non-significant differences between GE and Philips vendors.^{39  ,  40}

Cardiotoxicity grading and prognosis

Cardiotoxicity was defined as a new or worsening in myocardial damage or functional parameters after initiation on chemotherapy and we propose a simple grading CTox definition based on current HF guidelines classification^{19  ,  20} (Figure 1).

Severe cardiotoxicity

Severe CTox as defined in this study seems to be related with all-cause mortality, with a 10-fold increase in total mortality as compared with patients without or milder forms of CTox (Figure 5A). Left ventricular ejection fraction <40% with or without biomarker abnormalities is a poor ventricular function parameter but has been the gold standard to select patients for clinical trials. For these reasons, our label of severe CTox includes symptomatic HF and asymptomatic LVEF <40.

Moderate cardiotoxicity

Left ventricular ejection fraction ≥40% and <50% (with or without the presence of abnormal values of other LV function parameters) is related with subsequent development of severe and potentially irreversible LVD and HF¹²; however, we could not find a relationship between moderate CTox and outcomes (Figure 5A). Although more information is needed in this area the fact that many of these patients were on HF drugs could prevent further deterioration of ventricular function.

Mild cardiotoxicity

Another grade of myocardial damage may be represented by abnormal biomarkers with normal LVEF (≥50%) with or without any other LV function abnormalities. We explored troponin, NT-proBNP, and some advanced ECHO parameters; however, we did not find a relationship with poor outcomes at 2 years follow-up in patients who remain in this stage (Figure 5). Nevertheless, the identification of mild forms of CTox seems to be critical for the long-term follow-up of cancer survivors. Cardinale et al.¹¹ proved that patients who experience a persistent increase in cardiac troponins during therapy have a high risk of further LVD in the long-term follow-up,^{12  ,  29  ,  30  ,  41} and early troponin-guided treatment with enalapril changes the natural history of CTox and minimizes this risk.^{11  ,  30} Although natriuretic peptides are considered extremely useful for HF diagnosis and prognosis,^19–21 a threshold value related with subsequent significant ventricular dysfunction has not been determined in cancer patients and smaller increments in our population may not be relevant.

Another critical clinical question is whether GLS-guided management would prevent subsequent deterioration in LVEF. Although several small studies have shown some benefit in terms of LVD prevention, larger randomized trials are needed before incorporating this strategy in clinical guidelines.^{16–18  ,  36} Recently, Santoro et al.  ⁴² followed 116 patients with breast cancer receiving epirubicine who presented a GLS drop >15% or reduction of LVEF below 50%. Treatment with perindopril and carvedilol and in some of them interruption of cancer treated resulted in an improvement of ventricular function.⁴² The ongoing Strain Surveillance of Chemotherapy for Improving Cardiovascular Outcomes (SUCCOUR) trial will be the first randomized controlled trial to address the role of advanced imaging-guided cardiotoxicity prevention.⁴³

The Royal Brompton hospital classification of cardiotoxicity

The Royal Brompton Hospital classification of CTox is different,¹⁵ focusing on more specific biomarker, clinical, or ECHO abnormalities, not considering the degree of LVEF reduction below normal, whereas the CARDIOTOX classification was aligned to the HF guideline-based definitions of HFmrEF and HFrEF (LVEF <40%). Grade 6 in the Royal Brompton Hospital classification found in about 2% cases, is the only class related with higher all-cause mortality (Figure 5B). The other grades of CTox have no apparent relationship with poor outcomes in the middle-term follow-up when managed by a co-ordinate cardio-oncology team.

Limitations

The strengths of this study include the prospective collection of a broad range of commonly used clinical, biomarker, and ECHO data at pre-specified frequent intervals during a 2-year follow-up in patients with different forms of cancer receiving treatments previously related with a relatively high incidence of CTox.

Patients with abnormal biomarkers or ECHO values at baseline were included in the registry and this could be considered as a bias to determine the CTox of cancer therapies. However, this represents a more real population of patients in whom some degree myocardial injury or dysfunction may be present before cancer treatments.⁴⁴ We excluded, however, patients with an LVEF <40 or previous HF.

We used the left atrial area, as recommended when we planned the study. Currently, left atrial volume is more precise and recommended instead of the atrial area.

A number of missing visits or incomplete data collection during follow-up related to the investigational nature of the registry also includes a bias to estimate the prevalence of myocardial damage but was unavoidable due to priorities in the treatment of cancer patients.

What is more important, this study is in contradiction with other data supporting different opinions regarding the severity gradation of cardiotoxicity or the functional abnormality that should trigger a specific treatment, but this only supports the need for further clinical research.

Conclusions

A significant number of patients receiving high-risk cancer therapies present objective data of myocardial injury or LVD. Nevertheless, the number of patients with severe CTox is comparatively very low but is strongly related with all-cause mortality. Milder forms induced ventricular dysfunction were not found to be related with prognosis but represent an important warning to consider a closer follow-up, initiation of classic HF treatments, or even discontinue chemotherapy on an individual basis in spite of solid evidence-based data.

This issue is still controversial, and a comprehensive CV monitoring is critical to identify and treat HF risk factors and pre-clinical LVD when needed. This fact may explain the low percentage of cardiotoxicity hard endpoint in this registry and the urgent need to involve cardiologists in the design and monitoring of oncologic trials. We propose a classification of CTox (Take home figure) that could be used in protocols defining strategies for early identification, prevention, and treatment in patients receiving potentially cardiotoxic cancer therapies.

Future clinical research

Future clinical research is recommended to confirm the relationship of different grades of CTox with clinical outcomes. The elaboration of a practical clinical score to determine the risk of severe CTox during and long term after cancer therapies will help to refine current strategies to follow-up asymptomatic patients. The ultimate form of classifying CTox is considering the complete clinical profile of the patient; that is developing a clinical score able to more accurately predict the evolution of myocardial dysfunction and outcomes in this population.^{44  ,  45} There is also emerging evidence for genetic susceptibility (e.g. titin gene mutations) and screening targeted populations may be appropriate.⁷ This is beyond the scope of this study.

The other relevant area for clinical research is the robust identification of therapies that permit the prevention and recovery of myocardial damage/dysfunction. There is a clear need to explore strategies that provide benefit in these newly recognized forms of LVD. Heart failure has proved to be an elusive setting for demonstrating benefit, and most outcomes in clinical trials showed neutral results or increased mortality against all expectations.^{46  ,  47} We may use empirical treatments based on common sense, but further research is much needed.

Funding

This study was partially funded with a grant from the Instituto de Salud Carlos III (PI13/00559).

Conflict of interest: J.L.-S. reports grants from Novartis, Bayer, Merk, DalCore, and Sanofi; personal fees from Menarini and Servier, outside the submitted work; C.Á.-O., P.Z.A., A.B.S., A.R.L., D.C., I.R.R., A.A., G.M., J.R.G.-J., A.M.M., P.G.P., J.G.-C., J.M.S.A., R.C.C. have nothing to disclose; D.F. reports personal fees from Daiichi Sankyo, outside the submitted work. M.C.A. reports personal fees from Celgene, Janssen, Roche, Sandoz, Servier, Takeda, Novartis, Gilead, Amgen, and Karyopharm, outside the submitted work; J.F.B. reports personal fees from Amgen, Ipsen, Eissai, Merk, Roche, and Novartis, outside the submitted work. O.R.F. reports grants from Fundación Jose Luis Castaño-Spanish Society of Laboratory Medicine, outside the submitted work. T.L.F. reports personal fees from Janssen, Gilead, Pfizer, Novartis, Daiichi Sankyo, and TEVA, outside the submitted work. All other authors have nothing to disclose.

References