Departments of 1Pediatric Oncology and 2Pediatrics, Academic Medical Centre, University of Amsterdam, Amsterdam, The Netherlands
Received 28 June 2001; revised 2 November 2001; accepted 16 November 2001.
Abstract
Background
Anthracyclines are essential for the treatment of the children with cancer. We performed a systematic review to evaluate the existing evidence of the frequency and risk factors of anthracycline-induced clinical heart failure (A-CHF) in children.
Design
Medline was searched for articles reporting the frequency of A-CHF, published from 1966 to December 2000. Information about study features, risk factors and frequency were abstracted, and a validity score was given for each study. The potential predictive factors of A-CHF were analysed both within and across the studies.
Results
The frequency of A-CHF in children was estimated in 30 studies described in 25 articles. All studies have serious methodological limitations. The frequency varied between 0% and 16%. In the analysis across the studies the type of anthracyclines and the maximal dose in 1 week explain a considerable part of the variation of the frequency of A-CHF.
Conclusions
Doxorubicin and a dose above 45 mg/m2 within 1 week seemed to increase the frequency of A-CHF. Well designed and executed studies are needed to accurately estimate the frequency of A-CHF and reliably assess the importance of potential risk factors.
Key words: anthracycline, cardiotoxicity, children, heart failure, systematic review
Introduction
Anthracycline-induced clinical heart failure (A-CHF), clinical cardiotoxicity, has been acknowledged since the introduction of anthracyclines [1]. Despite this adverse effect, anthracyclines form an important part of the treatment of childhood cancer. An accurate estimation of the frequency of A-CHF and the associated risk factors is needed, since decisions about treatment with other chemotherapeutic treatment options or cardioprotective strategies during chemotherapy depend on the impact of the adverse effects. Moreover, for the follow-up of patients treated with anthracyclines it is crucial to know the frequency of A-CHF and to understand which patients are at greatest risk. The reported frequency of A-CHF varies between the different studies. It was hypothesised that this wide variation could be explained by differences in study design, such as composition of the study group, duration of follow-up or validity of the studies. An evaluation was made of the existing evidence to obtain more insight into the frequency and risk factors of A-CHF after anthracycline treatment for childhood cancer.
Materials and methods
Search for studies
The objective of the literature search was to identify all studies reporting on the frequency of A-CHF after anthracycline treatment for childhood cancer. The selection process involved four steps. First, Medline was searched for potentially relevant articles published from 1966 to December 2000. The sensitive search strategy is presented in Table 1. Secondly, relevant articles which possibly met the following inclusion criteria were selected on the basis of the title and abstract by two researchers (L.K., E.v.D.) and retrieved for more detailed examination: (i) original report; (ii) published in English; (iii) study population of >50 children who were treated with doxorubicin, daunorubicin or epirubicin; and (iv) reported clinical heart failure as outcome. Thirdly, the bibliographies of all relevant articles and reviews were searched for additional references. Finally, all retrieved articles were screened by the two researchers to ensure that they described original research and met the inclusion criteria. Interobserver agreement was calculated for the second and fourth step of the selection process. In cases of disagreement, the abstracts and articles were re-examined and discussed until consensus was achieved.
|
To determine the validity of the selected studies, two researchers (L.K., E.v.D.) assessed the methods, design and execution of each study. The validity was assessed according to the criteria for prognostic studies described by the Evidence-Based Medicine Working Group [2]. A sample was considered to be well defined if the mean, the median or the range of the cumulative anthracycline dose was mentioned. For studies in which patients were included from the start of therapy, a sample was defined as representative for the underlying population if it consisted of >95% of the whole cohort or a random sample of the cohort of patients from the start of therapy with respect to the cumulative dose. For studies investigating the survivors of childhood cancer, a sample was defined as representative if >95% of the survivors or a random sample of the survivors were studied. It was also assessed whether the study patients entered the cohort at a similar well defined point in the course of the disease. A minimal follow-up of >1 year after the end of treatment or a median follow-up of 2 years was considered to be adequate to determine the frequency of A-CHF. Follow-up was considered to be complete if the outcome was assessed at the end date of the study for >95% of the study patients. It was also assessed whether objective and unbiased outcome criteria were applied, and whether there was adjustment for important prognostic factors. Each study was graded on the basis of meeting or not meeting the first five criteria, and a validity score was calculated (range 05).
Definition and frequency of A-CHF
A-CHF in this systematic review was defined as heart failure with clinical symptoms in patients treated with anthracyclines reported by the authors. The frequency of A-CHF was calculated as the number of patients with A-CHF divided by the number of patients in the study group. The confidence interval (CI) was calculated with statistical programme Statxact, version 3. The cumulative incidence of A-CHF was expressed as this ratio at the end of follow-up for each study.
Risk factors for A-CHF
The risk factors for A-CHF were analysed both within and across the studies. For the analysis within the studies, a 2 x 2 table was constructed if the number of patients with and without a certain risk factor and the number of patients with and without A-CHF was reported. The relative A-CHF risk with the 95% CI for a certain risk factor was calculated with the statistical programme CIA [3]. A risk factor was defined as significant if the lower limit of the 95% CI of the relative risk was >1. Across the various studies, a multivariate linear regression model with a backward selection strategy (P <0.10) was used to examine the relation between potential predictive factors and A-CHF. Categorical variables included the following possible predictive factors: study design (i.e. studies of survivors versus studies of cohorts of patients from the start of therapy); the validity of the study (2+ or >2+ on the validity score); type of anthracycline treatment in the study group (doxorubicin or daunorubicin); cumulative dose of anthracycline (mean dose in the study population >450 mg/m2 or
450 mg/m2); and maximal dose of anthracyclines within 1 week (dose of >45 mg/m2 for some or all patients or <45 mg/m2 for all the patients). Only studies in which all these risk factors were known were included in this model. For the linear regression, a logistic transformation of the cumulative incidence was performed to obtain a normal distribution. Before this transformation, 0.5 was added to incidence figures because two studies reported an incidence of 0%.
Results
Selection of articles
The search strategy yielded 201 potentially relevant refer-ences from the Medline search. Sixty-eight articles were retrieved for more detailed examination. The interobserver agreement was 96% for this selection. Three additional articles were retrieved after searching the bibliographies. Twenty-five of the 71 articles retrieved met all the inclusion criteria, with an interobserver agreement of 95%. The other 46 articles were excluded for the following reasons: not original research, study patients were adults, impossible to differentiate between adults and children in the study group, <50 children, not published in English, not reporting on clinical heart failure.
Description of the selected studies
The frequency of A-CHF in children was estimated in 30 studies, described in 25 articles [428]. A total of 12 507 children treated with anthracyclines were studied. The 30 studies were divided into 19 studies of children from the start of anthracycline treatment [420] and 11 studies of survivors [2128]. Table 2 presents details of studies of patients from the start of therapy, and Table 3 presents studies of survivors in chronological order. Seventeen studies described the frequency of A-CHF after treatment with doxorubicin, four studies after treatment with daunorubicin and nine studies after a combination anthracycline therapy. Fourteen of the 19 studies of patients from the start of therapy described the entire cohort of patients from the start of therapy [46, 8, 1116, 18, 20]. Three of the 19 studies of patients from the start of therapy described a subgroup of the original cohort from the start of therapy [9, 10, 17]. Two studies probably described a subgroup, because they reviewed the medical records or studied evaluable patients without mentioning the number of patients in the original cohort [7, 19]. Four of the 11 studies of survivors described an entire cohort of patients from the moment of complete remission, and one study described an entire cohort from the completion of therapy [21, 22]. Six of the 11 studies of survivors described a subgroup of the original cohort of the survivors [2328].
|
|
|
|
Two of the six studies of survivors in which a subgroup of the original cohort was studied mentioned that there were no patients with A-CHF in the group not studied [26, 27]. The frequency for these two studies was re-calculated as the number of patients with A-CHF divided by the number of patients in the original cohort. Five studies of a subgroup of a cohort of patients from the start of therapy and four of the six studies of survivors that studied a subgroup of patients did not give the number of patients with A-CHF in the group not studied [7, 9, 10, 16, 19, 2325, 28]. In these studies, the frequency of A-CHF was calculated by dividing the number of patients with A-CHF by the number of patients in the study group.
It was possible to assess the relative risk of possible risk factors in 10 studies (Table 6) [6, 8, 1113, 17, 1921, 26]. The following risk factors were associated with a higher risk of A-CHF: a higher cumulative dose (four studies), radiation therapy in the heart region (three studies), age <4 years (one study), children (one study), a higher maximal dose in 1 week (two studies), daunorubicin (one study), amsacrine (one study), black race (one study), female (one study) and trisomy 21 (one study). Oberlin et al. [15] suggested that ifosfamide was a risk factor for A-CHF.
|
Discussion
The frequency of A-CHF has been assessed in a series of studies that have used various designs: cohort studies that monitored patients from the start of anthracycline treatment, cohort studies of patients who survived childhood cancer, and subgroups of these cohorts. The reported frequency of A-CHF in these studies varies between 0% and 16%. The type of anthracycline used and the maximal dose given within 1 week explain a considerable proportion of the variation in the reported frequency of A-CHF. The association between the type of anthracycline and the maximal dose given within 1 week with the frequency of A-CHF could be a true causal relationship. A higher risk of A-CHF after a higher maximal dose within 1 week has also been suggested by Krischer et al. [20] and Sallan et al. [21]. However, the association between type of anthracycline and the maximal dose within 1 week could also be the result of confounding, because non-randomised comparisons were analysed. It is conceivable that the choice of anthracycline type and the choice for a higher dose could have been related to a clinical factor that also predisposes to an increased risk of A-CHF. An associated limitation of this review is that no evaluation was made of the influence of age, sex distribution, radiation therapy or duration of follow-up on the frequency of A-CHF, because only a few studies reported adequate information of these variables. It was remarkable that only half of the studies mentioned the range, the median or the mean cumulative dose of anthracyclines. As a result, the risk factor analysis was performed on a small subgroup of studies, which reported type of anthracycline used, the cumulative dose and the maximal dose given within 1 week.
Variation in studies that evaluate the frequency of A-CHF can be explained by several factors. First, differences in frequency of A-CHF can be a reflection of the different risk profiles in the study population. It is quite conceivable that factors other than the reported type and dose of anthracycline may have varied across the studies, such as age at diagnosis, number of patients receiving radiation therapy or co-treatment with other cardiotoxic drugs. These potential risk factors were not well defined in the studies, and some risk factors are still unknown. In the present review, eight studies described a non-representative sample of patients from the original cohort. This selection of patients can lead to an overestimation of the true frequency of A-CHF if only patients with a higher risk profile were selected. For example, Steinherz et al. [24] studied patients who were referred by their paediatric oncologists during follow-up visits. This probably led to an overestimation of the true frequency of A-CHF, because patients with a higher risk of developing A-CHF are more likely to be seen at a follow-up clinic. On the other hand, the true frequency of A-CHF will be underestimated if patients with A-CHF were missed in the study group. Secondly, selection of patients at different points during the course of their disease could influence the results. The frequencies reported in studies of cohorts from the start of treatment represent the true cumulative incidence of A-CHF, whereas in studies of survivors, only early cases of A-CHF could be missed if patients died before the onset of follow up. Thirdly, it is also possible that differences in the duration of follow-up and completeness of the follow-up can explain the variation. Only four of 17 studies that reported the duration of the follow-up reported that the outcome for all patients included in the study was assessed at the end of the study. With a longer follow-up period more patients will have the opportunity to develop A-CHF. Incompleteness of follow-up can lead to an underestimation of the cumulative incidence if the missed cases ares associated with poor health, or to an overestimation of the cumulative incidence if the missed cases are associated with good health. As a last point, none of the studies reported a blinded assessment of the outcome for the information on risk factors. Presence of presumed risk factors can, in some cases, lead the clinician to classify a patient as having A-CHF and thus lead to overestimation of the frequency in these type of studies. The lack of a precise and consistent definition of A-CHF in the different studies can also lead to a variation in the reported frequency of A-CHF.
A potential limitation of this systematic review is that only studies published in the English language were selected, and therefore some studies could have been missed. Another limitation is that this review did not included studies after January 2001. Recently, a cohort study of 607 children, followed from their first dose of anthracylcine, was published [29]. The study attempted to avoid all the limitations discussed above, and estimated a cumulative incidence of A-CHF, defined as congestive heart failure not attributable to known causes other than anthracyclines, of 5% at 15 years after treatment onset. However, the influence of the different types of anthracyclines and the maximal dose within 1 week was not evaluated in this study.
This systematic review evaluates the frequency of clinical cardiotoxicity after anthracycline therapy, which varies between 0% and 16%. Although many explanations for this variation are possible, the type of anthracycline used and the maximal given dose of anthracycline in 1 week can explain a considerable part of this variation. Subclinical cardiotoxicity has been reported in up to 56% of survivors of childhood cancer [23]. Patients with subclinical cardiotoxicity should be monitored over a long period of time to obtain insight into the clinical consequences of this subclinical cardiotoxicity.
In light of the serious methodological limitations of the current evidence, we advocate that new, well designed and well executed studies are needed to accurately estimate the frequency of A-CHF in these children and to reliably assess the importance of potential risk factors. These may include well designed cohort studies and randomised trials. As more reliable data become available, clinicians will be able to weigh the risk and impact of A-CHF against the benefits of high-dose anthracyclines to increase the chances of survival after childhood cancer, and this will enable them to make better informed proper clinical decisions.
Acknowledgements
We wish to thank Dr R. J. Scholten (Dutch Cochrane Centre) for helping with analyzing the data.
Footnotes
+ Correspondence to: Dr L. C. M. Kremer. Emma Kinderziekenhuis, Academic Medical Centre, University of Amsterdam, Department of Pediatrics, G8-262, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands. Tel: +31-20-5662453; Fax: +31-20-6917735; E-mail: L.C.Kremer@amc.uva.nl
References
1. Bonadonna G, Monfardini S. Cardiac toxicity of daunorubicin. Lancet 1969; 1: 837.
2. Laupacis A, Wells G, Richardson WS, Tugwell P. Users guides to the medical literature. V. How to use an article about prognosis. Evidence-Based Medicine Working Group. JAMA 1994; 272: 234237.[ISI][Medline]
3. Gardner MJ, Altman DG. Statistics with Confidence. BMJ Press 1989.
4. Tan C, Etcubanas E, Wollner N et al. Adriamycinan antitumor antibiotic in the treatment of neoplastic diseases. Cancer 1973; 32: 917.[ISI][Medline]
5. Halazun JF, Wagner HR, Gaeta JF, Sinks LF. Proceedings: daunorubicin cardiac toxicity in children with acute lymphocytic leukemia. Cancer 1974; 33: 545554.[ISI][Medline]
6. Gilladoga AC, Manuel C, Tan CT et al. The cardiotoxicity of adriamycin and daunomycin in children. Cancer 1976; 37: 10701078.[ISI][Medline]
7. Cangir A, Morgan SK, Land VJ et al. Combination chemotherapy with adramycin (NSC-123127) and dimethyl triazeno imidazole carboxamide (DTIC) (NSC-45388) in children with metastatic solid tumors. Med Pediatr Oncol 1976; 2: 183190.[Medline]
8. Von Hoff DD, Rozencweig M, Layard M et al. Daunomycin-induced cardiotoxicity in children and adults. A review of 110 cases. Am J Med 1977; 62: 200208.[ISI][Medline]
9. Hutchinson RJ, Bailey C, Wood D, Donaldson MH. Systolic time intervals in monitoring for anthracycline cardiomyopathy in pediatric patients. Cancer Treat Rep 1978; 62: 907910.[ISI][Medline]
10. Lewis AB, Pilkington R, Takahashi M, Siegel SE. Echocardiographic assessment of anthracycline cardiotoxicity in children. Med Pediatr Oncol 1978; 5: 167175.[ISI][Medline]
11. Praga C, Beretta G, Vigo PL et al. Adriamycin cardiotoxicity: a survey of 1273 patients. Cancer Treat Rep 1979; 63: 827834.[ISI][Medline]
12. Goorin AM, Borow KM, Goldman A et al. Congestive heart failure due to adriamycin cardiotoxicity: its natural history in children. Cancer 1981; 47: 28102816.[ISI][Medline]
13. Dearth J, Osborn R, Wilson E et al. Anthracycline-induced cardiomyopathy in children: a report of six cases. Med Pediatr Oncol 1984; 12: 5458.[ISI][Medline]
14. Petrilli AS, Gentil FC, Epelman S et al. Increased survival, limb preservation, and prognostic factors for osteosarcoma. Cancer 1991; 68: 733737.[ISI][Medline]
15. Oberlin O, Habrand JL, Zucker JM et al. No benefit of ifosfamide in Ewings sarcoma: a nonrandomized study of the French Society of Pediatric Oncology. J Clin Oncol 1992; 10: 14071412.[Abstract]
16. Hudson MM, Greenwald C, Thompson E et al. Efficacy and toxicity of multiagent chemotherapy and low-dose involved-field radiotherapy in children and adolescents with Hodgkins disease. J Clin Oncol 1993; 11: 100108.[Abstract]
17. BuLock FA, Mott MG, Oakhill A, Martin RP. Early identification of anthracycline cardiomyopathy: possibilities and implications. Arch Dis Child 1996; 75: 416422.[Abstract]
18. Wexler LH, Andrich MP, Venzon D et al. Randomized trial of the cardioprotective agent ICRF-187 in pediatric sarcoma patients treated with doxorubicin. J Clin Oncol 1996; 14: 362372.[Abstract]
19. Godoy LY, Fukushige J, Igarashi H et al. Anthracycline-induced cardiotoxicity in children with malignancies. Acta Paediatr Jpn 1997; 39: 188193.[Medline]
20. Krischer JP, Epstein S, Cuthbertson DD et al. Clinical cardiotoxicity following anthracycline treatment for childhood cancer: the Pediatric Oncology Group experience. J Clin Oncol 1997; 15: 15441552.[Abstract]
21. Sallan SE, Clavell LA. Cardiac effects of anthracyclines used in the treatment of childhood acute lymphoblastic leukemia: a 10-year experience. Semin Oncol 1984; 11: 1921.
22. Pastore G, Zurlo MG, Acquaviva A et al. Health status of young children with cancer following discontinuation of therapy. Med Pediatr Oncol 1987; 15: 16.[ISI][Medline]
23. Lipshultz SE, Colan SD, Gelber RD et al. Late cardiac effects of doxorubicin therapy for acute lymphoblastic leukemia in childhood. N Engl J Med 1991; 324: 808815.[Abstract]
24. Steinherz LJ, Steinherz PG, Tan CT et al. Cardiac toxicity 4 to 20 years after completing anthracycline therapy. JAMA 1991; 324: 16721677.
25. Sorensen K, Levitt G, Sebag-Montefiore D et al. Cardiac function in Wilms tumor survivors. J Clin Oncol 1995; 13: 15461556.[Abstract]
26. BuLock FA, Mott MG, Oakhill A, Martin RP. Left ventricular diastolic function after anthracycline chemotherapy in childhood: relation with systolic function, symptoms, and pathophysiology. Br Heart J 1995; 73: 340350.[Abstract]
27. Kakadekar AP, Sandor GG, Fryer C et al. Differences in dose scheduling as a factor in the etiology of anthracycline-induced cardiotoxicity in Ewing sarcoma patients. Med Pediatr Oncol 1997; 28: 2226.[ISI][Medline]
28. Sorensen K, Levitt G, Bull C et al. Anthracycline dose in childhood acute lymphoblastic leukemia: issues of early survival versus late cardiotoxicity. J Clin Oncol 1997; 15: 6168.[Abstract]
29.
Kremer LC, van Dalen EC, Offringa M et al. Anthracycline-induced clinical heart failure in a cohort of 607 children: long-term follow-up study. J Clin Oncol 2001; 19: 191196.