A Biomarker-Based Diagnostic Model for Cardiac Dysfunction in Childhood Cancer Survivors

Abstract

Background: Childhood cancer survivors at risk for heart failure undergo lifelong echocardiographic surveillance. Previous studies reported the limited diagnostic accuracy of N-terminal pro-B-type natriuretic peptide (NT-proBNP) and high-sensitivity cardiac troponin T (hs-cTnT) in detecting left ventricular (LV) dysfunction. However, potential enhanced diagnostic accuracy through the combination of biomarkers and clinical characteristics has been suggested.

Objectives: The aim of this study was to develop and internally validate a diagnostic model that combines cardiac biomarkers with clinical characteristics for effectively ruling in or ruling out LV dysfunction in childhood cancer survivors.

Methods: A multicenter cross-sectional study included 1,334 survivors (median age 34.2 years) and 278 siblings (median age 36.8 years). Logistic regression models were developed and validated through bootstrapping, combining biomarkers with clinical characteristics.

Results: Abnormal NT-proBNP levels were observed in 22.1% of survivors compared with 5.4% of siblings, whereas hs-cTnT levels exceeding 10 ng/L were uncommon in both survivors (5.9%) and siblings (5.0%). The diagnostic models demonstrated improvement upon the addition of NT-proBNP and hs-cTnT to clinical characteristics, resulting in an increased C statistic from 0.69 to 0.73 for LV ejection fraction (LVEF) <50% and a more accurate prediction of more severe LV dysfunction, with the C statistic increasing from 0.80 to 0.86 for LVEF <45%. For LVEF <50% (prevalence 10.9%), 16.9% of survivors could be effectively ruled out with high sensitivity (95.4%; 95% CI: 90.4%-99.3%) and negative predictive value (97.5%; 95% CI: 94.6%-99.7%). Similarly, for LVEF <45% (prevalence 3.4%), 53.0% of survivors could be ruled out with moderate to high sensitivity (91.1%; 95% CI: 79.2%-100%) and high negative predictive value (99.4%; 95% CI: 98.7%-100%).

Conclusions: The biomarker-based diagnostic model proves effective in ruling out LV dysfunction, offering the potential to minimize unnecessary surveillance echocardiography in childhood cancer survivors. External validation is essential to confirm these findings. (Early Detection of Cardiac Dysfunction in Childhood Cancer Survivors; A DCOG LATER Study; https://onderzoekmetmensen.nl/nl/trial/23641).

Keywords: biomarkers; cardio-oncology; childhood cancer survivors; diagnostic accuracy.

Graphical abstract

Figure 1

Receiver-Operating Characteristic Curves Receiver-operating characteristic curves showing the discrimination of diagnostic models with cardiac biomarkers alone, clinical characteristics alone, and cardiac biomarkers combined with clinical characteristics for left ventricular dysfunction in childhood cancer survivors. hs-cTnT = high-sensitivity cardiac troponin T; LVEF = left ventricular ejection fraction; NT-proBNP = N-terminal pro–B-type natriuretic peptide.

Figure 2

Calibration Plots Calibration plots showing agreement between observed and predicted risk for left ventricular dysfunction in childhood cancer survivors using the diagnostic model incorporating cardiac biomarkers and clinical characteristics. Model calibration was assessed using the Spiegelhalter test, where P < 0.05 indicates inadequate calibration. LVEF = left ventricular ejection fraction.

Central Illustration

Biomarker-Based Model to Rule Out Cardiac Dysfunction in Childhood Cancer Survivors In a cross-sectional cohort study including 1,334 childhood cancer survivors treated with anthracyclines and/or chest-directed radiotherapy, a diagnostic model combining cardiac biomarkers (N-terminal pro–B-type natriuretic peptide [NTproBNP] and high-sensitivity cardiac troponin T [hs-cTnT]) with clinical characteristics was useful in ruling out the presence of left ventricular ejection fraction (LVEF) <50% on echocardiography for 16.9% of the population, demonstrating high negative predictive value and sensitivity. The considered clinical characteristics include sex, age at diagnosis, age at study, anthracycline dose, chest-directed radiotherapy dose, and heart rate (HR). AUC = area under the receiver-operating characteristic curve; NPV = negative predictive value.