Cardiac Function in 5-Year Survivors of Childhood Cancer:  A Long-term Follow-up Study FREE

Long-term survival of childhood cancer has improved considerably, from 20% in the 1940s,¹ to 70% to 80%^2- 5 at present, owing to chemotherapy, radiotherapy, supportive therapy, and combined modality treatment. Unfortunately, improved survival is accompanied by the occurrence of late treatment effects.^6- 8 Cardiovascular disease (CVD) and cardiac mortality are among the most serious late effects. Several population-based studies observed a 6- to 8-fold increased mortality owing to CVD among childhood cancer survivors (CCSs) compared with the general population.^9- 11

Anthracycline- and radiation-induced CVD are recognized as important long-term complications after childhood cancer.^7- 8 Anthracyclines are among the most effective antineoplastic drugs for childhood cancer and are still widely used in about half the patients. The use of radiotherapy as part of childhood cancer treatment has decreased over time. Heart damage can become manifest as either subclinical or clinical cardiac dysfunction. The frequency of anthracycline-induced cardiac dysfunction varies across studies, depending on follow-up time, attained age, population studied, and definitions used, ranging from 0% to 57% for subclinical cardiac dysfunction and from 0% to 16% for clinical CVD.^12- 17 Reported frequencies of radiation-induced clinical CVD vary by follow-up time and concurrent chemotherapy, whereas the frequency of subclinical radiation-induced damage is largely unknown. Most studies focus on radiation-induced cardiac mortality, which is increased 22- to 68-fold compared with the general population.¹⁸ Suspected risk factors for treatment-induced CVD are cumulative dose and type of anthracycline, radiation dose, age at diagnosis, and female sex.^12,14,18 However, results are not univocal across studies.^19- 24 Furthermore, the impact of other potentially cardiotoxic drugs, like high-dose cyclophosphamide or ifosfamide, has not been examined in large studies.^25- 27

In this study we evaluated the prevalence of echocardiographic left ventricular (LV) dysfunction and associated risk factors in a large cohort of long-term CCSs treated with different potentially cardiotoxic therapies.

In 1996, the Outpatient Clinic for Late Effects of Childhood Cancer for adult CCSs was established in the Emma Children's Hospital/Academic Medical Center (EKZ/AMC), Amsterdam, the Netherlands. All adult 5-year CCSs were traced using the Childhood Cancer Registry of the EKZ/AMC, which was established in 1966 and maintains data on diagnosis, treatment, last known medical status, and follow-up of all patients treated for childhood cancer in the EKZ/AMC. All CCSs surviving 5 years or more were invited to participate and enrolled into prospective follow-up study protocols tailored to previous diagnosis and treatment. These protocols were developed by health care professionals based on consensus. The CCSs gave informed consent for data collection. The CCSs who received treatment with anthracyclines, cardiac irradiation (thorax, spine, left or whole upper abdomen, total body irradiation [TBI]), and/or high-dose cyclophosphamide or ifosfamide (ie, total cumulative dose >10 g/m² or ≥1 g/m² per course) were enrolled in the cardiotoxic effects screening protocol. Five-year CCSs were eligible for the present study if they were treated between January 1966 and August 1997, qualified for the cardiotoxic effects screening protocol, and visited our outpatient clinic for adult CCSs for the first time after the age of 18 years between January 1996 and April 2004.

Data were extracted directly from the Childhood Cancer Registry database or collected from the medical records and the digital hospital information system, including: birth date, sex, tumor diagnosis, date of start of cardiotoxic treatment, cumulative doses of anthracyclines (doxorubicin, daunorubicin hydrochloride, epirubicin hydrochloride), cyclophosphamide, ifosfamide, cardiac irradiation, date of last follow-up, date and cause of death, signs and symptoms of heart failure, and echocardiography data.

The cardiotoxic effects screening protocol comprised a full medical assessment and an echocardiography on the same day. All eligible CCSs underwent a clinical cardiac assessment by the medical oncologist to evaluate signs and symptoms of heart failure. Clinical heart failure (CHF) was defined as congestive heart failure not attributable to other known causes, like direct medical effects of the tumor, septic shock, or renal failure. We defined congestive heart failure as the presence of the following: dyspnea, pulmonary or peripheral edema, and/or exercise intolerance, which were treated with anticongestive therapy (according to the New York Heart Association criteria).²⁸ Patients were examined by echocardiography at least once. The LV systolic function was assessed by means of standard M-mode echocardiography with a concurrent electrocardiogram. The LV shortening fraction (LVSF) was calculated using the formula: LVSF (%) = [(LVDD–LVSD)/LVDD] × 100%, where LVDD is the LV diastolic diameter measured at the start of the QRS complex and LVSD is the LV systolic diameter. The LVSF correlates to ejection fraction (EF) as follows: an LVSF of 30% equals an EF of 50%, an LVSF of 25% equals an EF of 45%.²⁹ The measurements of LVSF were performed by trained specialists from our department of cardiology, blinded to the patient's treatment, according to a follow-up protocol specific for CCSs. Measurements of LVSF were performed multiple times, diminishing the risk of intraobserver variability. For each patient, the first evaluable echocardiogram was used as the outcome measurement for the LVSF. We chose the first echocardiography as the most unbiased measurement possible, to avoid both selection bias and confounding by treatment. Survivors with an LVSF of less than 30% received a different additional follow-up schedule and possible treatment of CVD than those with an LVSF within normal limits. Treatment could even influence the prevalence and course of cardiac dysfunction (ie, treatment with angiotensin-converting enzyme [ACE] inhibitors could improve cardiac function and slow progression to overt clinical CVD).³⁰

The main outcomes of interest were the mean values of the LVSF across all available first measurements and the number of patients with an abnormal first LVSF (defined as an LVSF <30%) according to the Common Terminology Criteria for Adverse Events (CTCAE), version 3.0.²⁹ The following potential determinants of the LVSF were evaluated: sex, age at cancer diagnosis, time since cancer diagnosis, body mass index (BMI) at the time of the echocardiogram, cumulative doses of doxorubicin, epirubicin, daunorubicin, all anthracyclines combined, cyclophosphamide, and ifosfamide and total radiation dose to each of the 4 fields involving the heart.

Using multivariate linear regression models, changes in LVSF (percentage points) were estimated for categories of each variable compared with a reference category. We generally used the low-dose category as the reference instead of the zero-dose category to avoid confounding because all patients were exposed to potentially cardiotoxic treatments; hence, patients not exposed to anthracyclines were always exposed to radiotherapy or high-dose cyclophosphamide or ifosfamide.

We also examined determinants of an abnormal LVSF using multivariate logistic regression models to estimate the odds ratio (OR). Tests of trend were based on the corresponding continuous variable.

We evaluated whether the relationship between a certain treatment and LVSF was modified by another factor (eg, calendar year of treatment or time since diagnosis). Heterogeneity of effects across categories of the potential effect modifier was assessed based on goodness-of-model fit. Differences in radiation- or anthracycline-related risks by calendar year of treatment could indicate calendar period effects, while differences by time since diagnosis would indicate latency. Because detailed information on anthracycline and radiation dose was available and included in the analysis, any residual effect of calendar period of treatment was considered small. We therefore focused on modification of radiation- or anthracycline-related risks by time since diagnosis.

To address interaction between anthracyclines and radiotherapy, we evaluated whether the anthracycline dose effect was homogeneous across groups of patients whose radiotherapy fields involved different parts of the body. Where we observed strong evidence of an association (ie, significant trends among all patients and among exposed patients only), we evaluated the shape of the dose-response curve parametrically by adding a quadratic term to the linear regression model, as well as nonparametrically by using relatively narrow categories of dose.

To assess the cardiotoxic potency of the different anthracycline derivates, we analyzed the effect of the cumulative doses of specific anthracyclines separately from each other. We tested the heterogeneity of the results for the different derivates. We corrected the dose of epirubicin compared with doxorubicin and daunorubicin, because 90 mg/m² of epirubicin hydrochloride is as effective against cancer as 60 mg/m² of doxorubicin or daunorubicin hydrochloride. We evaluated the change in LVSF (linear model) and the OR for low LVSF (logistic model) per 60 and per 90 mg/m² of epirubicin hydrochloride.^10,31- 32 Statistical analyses were performed using SPSS software for Windows (version 16.0.2; SPSS Inc, Chicago, Illinois), SAS software (version 9.1; SAS Institute Inc, Cary, North Carolina), and Epicure software (version 1.4; Hirosoft, Seattle, Washington).

Between January 1966 and August 1997, 3310 patients were treated for childhood cancer in the EKZ/AMC. Of those, 1535 survived their cancer for more than 5 years, but 229 were not yet 18 years old during the study period and were excluded. Of the remaining 1306 CCSs, 735 patients were treated with anthracyclines, cardiac irradiation, high-dose cyclophosphamide, and/or ifosfamide. Sixty 5-year CCSs died before the start of our outpatient clinic (ie, the start of our study), 1 of CHF. Thus, 675 CCSs were eligible for inclusion, and 601 visited the outpatient clinic (89%). The other 74 CCSs did not visit the outpatient clinic for the following reasons: the CCS moved abroad (n = 9), visited another Dutch hospital (n = 6), died before the visit (n = 22), was lost to follow-up (n = 6), or did not show up at follow-up (n = 31). None of the 22 CCSs who died before they could visit the outpatient clinic died because of CVD.

Table 1 lists the characteristics of the study population. Of the 601 patients, 525 had an echocardiogram (87.4%). Seventy-six patients did not have an echocardiogram for various reasons: they had had echocardiography elsewhere (n = 20), had planned to have the echocardiogram but not yet undergone it (n = 39), refused it (n = 15), or had died (n = 2). Patients with and without an echocardiogram did not differ or did not differ substantially by cumulative anthracycline dose, age at diagnosis, sex, cardiac irradiation, cumulative cyclophosphamide or ifosfamide dose, time since diagnosis, and tumor diagnoses. Six CCSs died after the first cardiac assessment owing to causes other than CVD.

Of the 525 echocardiograms, 514 were evaluable for assessment of the LVSF. The median overall LVSF was 33.1% (range, 13.0%-56.0%). An LVSF of less than 30% was identified in 139 patients (27.0%) at an attained age of 23 years (Table 2). Two patients (1.4%) had an LVSF of less than 15% (CTC-AE grade 3), 29 (22.3%) had an LVSF of 15% to 23% (CTC-AE grade 2), and 108 (77.7%) had an LVSF of 24% to 29% (CTC-AE grade 1). An abnormal LVSF was most common among patients who received combination therapy: anthracyclines with either cyclophosphamide or ifosfamide (34.1%), or cardiac irradiation (28.1%) and in the group with cardiac irradiation and either cyclophosphamide or ifosfamide (30.8%). Seven CCSs (1.3%) were diagnosed as having CHF during cancer treatment. All were treated with anticongestive therapy and fully recovered. At the time of evaluation, none of these patients had symptoms or signs of CHF. Three patients did not use medication; 4 were treated with ACE inhibitors, digoxin, or carnitine. These patients were included in the echocardiographic analyses; all but 1 had an abnormal LVSF.

Results of the multivariate linear regression model are shown in Table 3. Age at diagnosis, time since diagnosis, cumulative anthracycline dose, and thoracic radiotherapy were significantly associated with a change in LVSF. The use of vincristine was not significantly associated with a decrease in LVSF (P = .07). Adding BMI to the regression model did not substantially improve the goodness-of-fit, nor did it substantially change the regression coefficients for other variables in the model (results not shown). Using multivariate logistic regression to model the effect of these potential determinants on subclinical cardiac dysfunction led to largely similar results as the linear regression analyses on continuous LVSF (Table 3).

With regard to specific anthracycline derivates, doxorubicin and epirubicin, but not daunorubicin, were significantly associated with a reduced LVSF (Table 4). Doxorubicin seemed to be the most potent determinant of LVSF, although the magnitude of effects did not differ significantly. However, taking into account that epirubicin is usually administered at a 1.5-fold higher dose than doxorubicin and daunorubicin, we observed about a 1.3-fold increased OR for cardiac dysfunction associated with either doxorubicin, 60 mg/m², or epirubicin hydrochloride, 90 mg/m². The observed association for daunorubicin was weaker with a significant trend among patients who received daunorubicin but not among all patients. There was no evidence of heterogeneity of the magnitude of effects of specific anthracycline derivates.

We observed no evidence that the effect of anthracycline dose was different among patients whose radiotherapy field involved a certain body part compared with others (thorax, P = .92; abdomen, P = .38; spine, P = .40; TBI, P = .30, based on linear regression; and thorax, P = .36; abdomen, P = .80; spine, P = .93; TBI, P = .26 based on logistic regression). Departure from linearity was mild for doxorubicin and suggested little decline of LVSF for doses of less than 100 mg/m² and a linear decline for higher doses (Figure).

Abnormal cardiac function, defined as an LVSF of less than 30%, was observed during long-term follow-up (median duration of follow-up, 15.4 years) in 27% of CCSs. It was most common in the combined treatment groups. Important determinants of decreased LVSF were younger age at diagnosis, higher cumulative anthracycline dose, and radiotherapy to the thorax. We found some, albeit weak, evidence that vincristine sulfate could decrease cardiac function. Epirubicin was as cardiotoxic as doxorubicin when corrected for tumor efficacy; daunorubicin seemed less cardiotoxic. We found no evidence that female sex, high-dose cyclophosphamide, or ifosfamide were risk factors for cardiac dysfunction.

The overall prevalence of 27% of CCSs with cardiac dysfunction is alarmingly high in this young population. Although large prospective longitudinal cohort studies with complete follow-up are currently lacking, these patients are expected to be at greater risk of developing CHF in the future. Lipshultz et al¹⁵ showed that cardiac abnormalities were persistent and progressive with time in ALL survivors. Brouwer et al¹⁶ also found evidence of progressive cardiac disease in survivors of malignant bone tumors, with 1 of 22 patients developing CHF. Because both studies were relatively small, these results need to be confirmed in a large cohort with complete follow-up. Furthermore, subclinical cardiac abnormalities need to be validated as surrogate markers for clinically important end points, like CHF. In adult patients with asymptomatic cardiac dysfunction from causes other than anthracyclines, the LVSF has been validated as a surrogate marker for CHF and death from CHF.^33- 34 However, because the etiology of cardiac damage in these patients was different, it is important to confirm these findings in patients treated with anthracyclines.

Our results showed higher LVSF measurements in patients with longer time since diagnosis, in contrast to most other studies.^15- 16 However, appropriate longitudinal data are still missing, so it is difficult to interpret these results. Lipshultz et al^15,30 described a recovery phase of the LVSF after treatment, so our findings might reflect this temporary improvement of the LVSF, owing to a compensatory or remodeling mechanism of the heart. However, the suggestion of an improvement in LVSF over time should not encourage a wait-and-see policy. The measurement of LVSF has limitations. It is a sum of different pathophysiologic mechanisms of the heart. If we had examined our cohort with more sensitive and sophisticated markers of the LV function, we might have found more deterioration of cardiac parameters as described by Lipshultz et al¹⁵ and Brouwer et al.¹⁶

To our knowledge, this is the first study in CCSs with information on the cardiotoxic effect of different anthracycline derivates. The decrease in LVSF per 100 mg/m² of doxorubicin in the linear regression analysis showed an apparent but not significant difference with the decrease per 100 mg/m² of daunorubicin hydrochloride and epirubicin hydrochloride. When we corrected the epirubicin dose for tumor efficacy, the cardiotoxic effects of doxorubicin and epirubicin were almost identical, and that of daunorubicin was weaker. Therefore, our data suggest that epirubicin is as cardiotoxic as doxorubicin at the same tumor efficacy, whereas daunorubicin seemed to be less cardiotoxic.

The present study confirmed that younger age at diagnosis had a clinically significant role in the decrease of LVSF compared with the highest age group of 15 years or older at diagnosis.¹⁴ Numerous studies indicate that CCSs treated with higher cumulative anthracycline doses are at higher risk of developing subclinical and clinical cardiac dysfunction.^12- 14,35 Our data were consistent with a linear dose-response relationship with anthracycline dose; that is, there was a linear decrease of the LVSF with increasing cumulative anthracycline doses. This suggests an influence of anthracyclines on cardiac function also in the lower dose ranges. In our analyses we used the lowest dose range (1-150 mg/m²), not the zero-dose, as the reference category, because all patients had received some form of potentially cardiotoxic treatment. On the one hand, consistent with the findings of Hudson et al,¹⁹ we observed significantly decreased LVSF at relatively low doses of anthracyclines (150-300 mg/m²) (P  = .03), suggesting that there may be no safe cumulative anthracycline dose. On the other hand, some studies observed no deterioration of cardiac function in CCSs treated with low-dose anthracyclines.^20- 21,36 However, Rammeloo et al³⁶ evaluated the effect of low-dose daunorubicin, which may be less cardiotoxic, as observed in our study, and the relatively small size of their study may have prevented detection of an effect.

To investigate the influence of radiotherapy on cardiac function, the effect of different radiation fields was evaluated. Radiation equipment and techniques have changed considerably over the years. Although the regions irradiated were known for every patient, the individual cardiac radiation doses could not be calculated. Therefore, we used 4 categories of radiation (thorax, TBI, spine, and left or whole abdomen) and used the tumor dose as the maximum dose that (part of) the heart could have received. Although we observed a strong effect of any vs no radiotherapy to the thorax, there was no relationship between radiation dose to the thorax and LVSF among patients who received thoracic radiotherapy. We believe this does not provide evidence of absence of a dose-response relation between cardiac irradiation and LVSF but is rather due to the fact that a dose to the thorax may be inversely correlated with the irradiated heart volume in our study. Most patients treated with radiotherapy to the thorax had lymphomas, nephroblastoma, Ewing sarcoma, or osteosarcoma (89%). Radiotherapy for the latter 3 usually involves prophylactic radiation of the entire lung with a lower dose, including a different and mostly larger part of the heart, whereas radiotherapy for lymphomas usually involves smaller and different parts of the heart (eg, mantle field, mediastinal) but with higher doses. This shows the need for future studies to include detailed dosimetry with assessment of the volume of the heart irradiated as well as the dose to different parts of the heart. Furthermore, cardiac irradiation might induce more ischemic heart disease, and we did not evaluate that in this study.^18,37- 39

Our study has important strengths. We were able to prospectively evaluate a large cohort of CCSs treated for all types of childhood cancer at one institution with nearly complete follow-up, limiting the potential for selection bias. All echocardiographic measurements were performed using the same screening protocol, thus excluding observer bias. Complete treatment data were available. Cardiotoxic treatment was well characterized, including cumulative anthracycline doses, type of anthracycline derivate, and cumulative cyclophosphamide and ifosfamide doses. For radiotherapy, the administered tumor dose and fields were known. Furthermore, the cohort represents a heterogeneous group of diagnoses, treatments across a broad spectrum of different doses of anthracyclines and radiotherapy and a variety of ages at diagnosis. We were able to analyze the influence of different anthracycline derivates.

Our study has some limitations. The ability of our study to evaluate more subtle cardiotoxic effects of high-dose cyclophosphamide, high-dose ifosfamide, or vincristine compared with the much stronger cardiotoxic effect of anthracyclines and cardiac irradiation, may have been limited owing to the absence of a comparison group not exposed to any potentially cardiotoxic treatment. Second, we were able to focus only on the LVSF without looking at other echocardiographic measurements as afterload, wall thickness, and diastolic dysfunction. The added value of those more sophisticated measurements of cardiac dysfunction should be evaluated in new research protocols. However, for LVSF the predictive value for clinically relevant outcomes has been established in adults and children.^33- 34,40 Unfortunately, in this study we were not able to investigate the correlation of an abnormal LVSF with the development of clinical CVD later on. However, this will be the focus of our next study. Third, measurements of LVSF were performed multiple times, diminishing the risk of intraobserver variability, but interobserver variability was not measured. Fourth, we used the results of the first echocardiographic measurement at the outpatient clinic to avoid selection bias. As a result, patients entered the study cohort at different time points since diagnosis. In the multivariate analyses, we adjusted for time since diagnosis. Fifth, the generalizability of the results might be influenced by the fact that our study was a single-center study. However, the Netherlands is a small country with only 7 hospitals with a children's oncology unit, all united in the Dutch Childhood Oncology Group, so treatment across the Netherlands is quite uniform.

In conclusion, more than 25% of young adult CCSs had subclinical cardiac dysfunction at their first visit to the outpatient clinic for late effects of childhood cancer. Continued monitoring of all CCSs treated with potentially cardiotoxic therapy with or without subclinical cardiac dysfunction is necessary to identify CCSs who could possibly benefit from early treatment, which could avoid further deterioration of cardiac function. The most important risk factors for developing subclinical cardiac dysfunction are the cumulative anthracycline dose, radiotherapy to the thorax and younger age at diagnosis. There is the suggestion that both abdominal radiotherapy and vincristine might play a role in developing cardiac dysfunction; this has to be confirmed in a cohort of CCSs including those who did not receive cardiotoxic treatment. We found no evidence of a higher risk of cardiac dysfunction for female CCSs and high-dose cyclophosphamide and ifosfamide. Epirubicin was as cardiotoxic as doxorubicin when corrected for tumor efficacy, and daunorubicin seemed to be less cardiotoxic.

Correspondence: Helena J. van der Pal, MD, Departments of Medical Oncology and Pediatric Oncology, Room F4-224, Academic Medical Center, Meibergdreef 9, 1105 AZ Amsterdam, the Netherlands (h.j.vanderpal@amc.uva.nl).

Accepted for Publication: January 20, 2010.

Author Contributions: Drs van der Pal, van Dalen, Hauptmann, and Kremer had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: van der Pal, van Dalen, Caron, van Leeuwen, and Kremer. Acquisition of data: van der Pal, van Dalen, Kok, van den Bos, and Oldenburger. Analysis and interpretation of data: van der Pal, van Dalen, Hauptmann, Kok, van den Bos, Koning, van Leeuwen, and Kremer. Drafting of the manuscript: van der Pal, Hauptmann, and Kremer. Critical revision of the manuscript for important intellectual content: van der Pal, van Dalen, Hauptmann, Kok, Caron, van den Bos, Oldenburger, Koning, van Leeuwen, and Kremer. Statistical analysis: Hauptmann and van Leeuwen. Obtained funding: Caron. Administrative, technical, and material support: van der Pal, van Dalen, and van den Bos. Study supervision: Caron and Kremer.

Financial Disclosure: None reported.

Funding/Support: This study was supported by the Foundation of Pediatric Cancer Research, Amsterdam, the Netherlands.

Role of the Sponsor: The funding organization had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.

Additional Contributions: We thank the staff of our Outpatient Clinic for Late Effects of Childhood Cancer, EKZ/AMC. We also thank Richard C. Heinen, MSc, Department of Pediatric Oncology, EKZ/AMC, for his uncompensated help in identifying eligible patients. We are indebted to the patients for giving their permission to participate in the study.