Risk Prediction Models for Cardiotoxicity of Chemotherapy Among Patients With Breast Cancer: A Systematic Review

• Published in: JAMA network open 2023-02, Vol.6 (2), p.e230569-e230569
• Main Authors: Kaboré, Elisé G, Macdonald, Conor, Kaboré, Ahmed, Didier, Romain, Arveux, Patrick, Meda, Nicolas, Boutron-Ruault, Marie-Christine, Guenancia, Charles
• Format: Article
• Language: English
• Subjects: Breast Neoplasms - drug therapy; Cardiotoxicity - epidemiology; Cardiotoxicity - etiology; Female; Humans; Life Sciences; Oncology; Online Only; Original Investigation; Reproducibility of Results; Stroke Volume; Ventricular Function, Left
• ISSN: 2574-3805; 2574-3805
• DOI: 10.1001/jamanetworkopen.2023.0569

Cardiotoxicity is a serious adverse effect that can occur in women undergoing treatment for breast cancer. Identifying patients who will develop cardiotoxicity remains challenging. To identify, describe, and evaluate all prognostic models developed to predict cardiotoxicity following treatment in women with breast cancer. This systematic review searched the Medline, Embase, and Cochrane databases up to September 22, 2021, to include studies developing or validating a prediction model for cardiotoxicity in women with breast cancer. The Prediction Model Risk of Bias Assessment Tool (PROBAST) was used to assess both the risk of bias and the applicability of the prediction modeling studies. Transparency reporting was assessed with the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) tool. After screening 590 publications, we identified 7 prognostic model studies for this review. Six were model development studies and 1 was an external validation study. Outcomes included occurrence of cardiac dysfunction (echocardiographic parameters), heart failure, and composite clinical outcomes. Model discrimination, measured by the area under receiver operating curves or C statistic, ranged from 0.70 (95% IC, 0.62-0.77) to 0.87 (95% IC, 0.77-0.96). The most common predictors identified in final prediction models included age, baseline left ventricular ejection fraction, hypertension, and diabetes. Four of the developed models were deemed to be at high risk of bias due to analysis concerns, particularly for sample size, handling of missing data, and not presenting appropriate performance statistics. None of the included studies examined the clinical utility of the developed model. All studies met more than 80% of the items in TRIPOD checklist. In this systematic review of the 6 predictive models identified, only 1 had undergone external validation. Most of the studies were assessed as being at high overall risk of bias. Application of the reporting guidelines may help future research and improve the reproducibility and applicability of prediction models for cardiotoxicity following breast cancer treatment.