Development and Validation of a Diagnostic Nomogram to Predict the Anthracycline-Induced Early Cardiotoxicity in Children with Hematological Tumors

As an effective chemotherapeutic agent, anthracyclines are widely used in the treatment of various cancers, especially children’s hematological malignancies [1]. However, the cardiotoxicity caused by anthracyclines severely affects the quality of life of surviving cancer children and is significantly associated with a poor prognosis in these children [2]. Additionally, the clinical utility of anthracyclines is also significantly limited by their life-threatening cardiotoxic side effects [3]. Cardiovascular complications induced by anticancer therapy have become the main cause of morbidity and mortality in childhood cancer survivors [4].

To date, the specific mechanism of anthracycline-induced cardiotoxicity remains unclear. The mechanism of cardiotoxicity was mainly based on the hypothesis of cardiomyocyte oxidative stress, mitochondrial dysfunction, topoisomerase II β-dysfunction, and iron metabolism disorder [5‐8]. Recent studies [9, 10] have demonstrated that early cardiac injury after anthracycline chemotherapy is transient and reversible. Early detection and prompt treatment may effectively lower long term of cardiovascular disease by preventing LV remodeling and progression to heart failure [11, 12]. However, it is more difficult to diagnose cardiotoxicity in clinical treatment owing to the unclear specific mechanism. Although the incidence of severe cardiotoxicity is dose dependent, even safely cumulative doses of anthracyclines may lead to permanent myocardial injury [13, 14]. Once clinical symptoms appear in children, even appropriate treatment cannot effectively prevent the progression of cardiotoxicity. Almost all children with heart failure need heart transplantation or die from cardiac complications [15]. Therefore, early identification and diagnosis of anthracycline-induced cardiotoxicity are essential, especially before permanent myocardial injury occurs.

Endomyocardial biopsies are still the gold standard for the diagnosis of myocardial injury induced by anthracycline. However, an invasive examination is not suitable for the large-scale detection of all children with anticancer therapy. At present, echocardiography is widely used in the clinical diagnosis of anthracycline cardiomyopathy [16]. Nevertheless, its sensitivity to screening at the early stage of heart injury is not high. Recently, some cardiac markers, such as N-terminal pro-B-type natriuretic peptide (NT-proBNP) and Troponin I (cTnT), are the focus of research and have the potential to diagnose cardiotoxicity before permanent heart injury. However, the diagnostic value of these cardiac markers in cardiotoxicity is obscure and even the conclusions of some studies are contradictory [17‐19]. Considering the limited diagnostic ability of a single index, a multi-index summary model should be established for the early diagnosis of cardiotoxicity.

Nomogram is a prediction tool that can provide personalized disease-related risk estimation [20]. It constructs a multi-factor regression model, integrates the prediction indicators according to the regression coefficient of each influencing factor in the model, and then presents a visual prediction result. It has been broadly adopted to assist clinical professional decision-making, such as cancer prognosis, trauma treatment, and surgery [21‐23]. To date, nomograms have not been applied for patients with anthracycline-induced cardiotoxicity. Our study aims to develop and validate a credible nomogram for predicting anthracycline-induced early myocardial injury during the treatment of hematological tumors in children.

In previous studies, there were several models for predicting cardiotoxicity in patients with anthracycline chemotherapy, although none of them were developed for hematological malignancies or children [33, 34]. At present, studies focus more on the assessment of risk factors and the exploration of predictors for cardiotoxicity. Nonetheless, the models set for predicting cardiotoxicity are different due to the different types of tumors and the difference in the study population. Although many studies have reported the related predictors of anthracycline chemotherapy-induced cardiotoxicity in children’s hematological malignancies, there is little research on the development and validation of a prediction model of cardiotoxicity based on these predictors. In this single-center, retrospective, and population-based study, a novel and practical nomogram was presented to predict the risk of cardiotoxicity in children. Compared with a single clinical predictor, this model can improve the prediction ability of early cardiac toxicity after anthracycline chemotherapy. Our study is the first nomogram to predict anthracycline chemotherapy-induced cardiotoxicity.

Nomogram is an intuitive visual tool for predicting disease and can convert a complex regression equation with clinical variables into a simple and visual graph [21]. Because of its readability and practicability, the nomogram is widely applied to the prediction of cancer prognosis and has attracted much attention in medical research and clinical decision-making in recent years [21‐23]. In our study, 11 predictors were considered to narrow the bias of nomograms, such as demographic information, biomarkers, and imaging features. These variables were screened by LASSO regression analysis and then the final four predictors for cardiotoxicity were identified. Demographic variables mainly consist of age at cancer diagnosis, gender of children, type of tumor, and cumulative dose of anthracyclines related to body surface area. The age of children was divided into three grades: < 6 years old, 6–10 years old, and > 11 years old. Similarly, the dose–response relationship of anthracycline-induced cardiotoxicity is unclear in children. Therefore, the classified variable of cumulative dose of anthracyclines was used, instead of the continuous variable for regression analysis. According to the research results [28‐30], the cumulative dose of anthracyclines of body surface area was divided into three grades: < 300 mg/m², 300–550 mg/m², and > 550 mg/m². The regression analysis results revealed that sex, age at diagnosis, and type of tumor were not the predictors of cardiotoxicity, consistent with the study of Spewak et al. [35]. Subsequently, our results demonstrated that the cumulative dose of anthracyclines had a potential correlation with the occurrence of cardiotoxicity. It confirms observations from previous studies [9, 36]. Hence, children with a high cumulative dose of anthracycline chemotherapy may have a higher risk of cardiotoxicity than children with a low cumulative dose.

As a simple, repeatable, and risk-free diagnostic tool, cardiac biomarkers have great potential in early identification, evaluation, and monitoring of cardiotoxicity risk caused by anthracyclines. In the past 10 years, many cardiac biomarkers related to cardiotoxicity have been developed. However, the current research still stresses troponin I/HS-troponin and NT-proBNP, although they are full of disputes and contradictions in the conclusions of some studies [17‐19]. Therefore, this study attempted to identify whether they have sufficient diagnostic ability in the early stage of cardiotoxicity. Notably, the reference values of these biomarkers may vary in childhood, and the standard values of children of different ages may be different. Additionally, the value of NT-proBNP decreased with age and increased in the second decade of girls [37, 38]. The NT-proBNP and cTnI were used as binary variables rather than continuous ones to reduce the possible bias of these confounding factors (such as age and race) on our research conclusions. LASSO regression analysis suggested that NT-proBNP played an important role in our nomogram, while the ability of troponin I as an early diagnostic cardiac biomarker was not enough. This result is consistent with the conclusion of a recent meta-analysis [19]. This meta-analysis included 27 related studies. The pooled data revealed that NT-proBNP is associated with cardiotoxicity in pediatric cancer patients receiving anthracycline therapy. However, the diagnostic benefit of troponin was unclear. In contrast, a study of 703 cancer patients demonstrated that elevated troponin in patients receiving cardiotoxic chemotherapy may be a sensitive indicator of early cardiotoxicity [39]. Other small studies have also confirmed the correlation between the increase of troponin and the decrease of LVEF [40, 41]. These contradictions may be correlated with the optimal timing of plasma marker assessment. NT-proBNP is mainly synthesized and secreted by left ventricular cardiomyocytes and regulated by left ventricular wall tension [42]. The severity of left ventricular dysfunction is positively correlated with its secretion. Diastolic dysfunction frequently occurs in children with cardiotoxicity [9], leading to more frequent abnormal secretion of NT-proBNP and more easily detected in peripheral blood. Troponin I is a specific index of myocardial injury. The peak time of cTnI after myocardial infarction is 10–24 h. The short detection time window usually induces false negatives in the detection of children’s peripheral blood. Owing to the exact timing of biomarker measurement and the variability of technology, their value in diagnosis has not been adequately affirmed. However, our nomogram fills these deficiencies to a great extent. Nomogram predicts cardiotoxicity by combining various variables, rather than relying on a single index or biomarker.

As a widely available, repeatable, and noninvasive tool for continuous evaluation of cardiac function, echocardiography is the most commonly used imaging technology to monitor cardiac function during and after chemotherapy. LVEF is a relatively insensitive tool in the diagnosis of the early stage of cardiotoxicity, since the decrease of LVEF does not occur until a critical number of myocardial injuries and the depletion of the cardiac compensation mechanism [43]. However, LVEF found almost all cardiotoxicity in the early stage in a recent large-scale study of breast cancer [9]. This study suggested that LVEF at the end of chemotherapy is an independent predictor of the development of cardiotoxicity. In patients with chemotherapy-induced cardiotoxicity, diastolic dysfunction may occur at an early stage [44]. Similarly, pericardial effusion, as a complication of anticycline, is related to cardiac dysfunction, which also implies pericardial disease and hemodynamic instability. This may be a sign of early adverse progression of cardiotoxicity [45]. Therefore, LASSO regression analysis was performed on LVEF, pericardial effusion, and diastolic dysfunction in this paper. The results demonstrated that LVEF and diastolic dysfunction were effective predictors of anthracycline-induced cardiotoxicity, with significant statistical significance.

Our study has several limitations. First, the number of factors leading to the study is limited because our study is retrospective. Some traditional variables, such as race and radiotherapy, and some novel predictors, such as GLS and tissue-type plasminogen activator, were not included in our study. Hence, a large prospective study is necessary. Second, this is a single-center study with data from a tertiary hospital. The number of cases is limited and not as representative as multicenter studies, although unified diagnostic criteria are beneficial to our study and analysis. Third, our study lacks external validation. Although the model exhibits good predictive performance in our study, it may not perform well in external validation in other child cohorts. Whether the model can be extended to primary hospitals should be further verified with more external data.