Menu Back to Poster-Presentations-Details

P103: Risk Prediction Models for Cardiotoxicity of Antineoplastic Agents for Breast Cancer: A Systematic Review

Poster Presenter

Honey Joseph

PGY2 Drug Information Pharmacy Resident
University of Illinois at Chicago
United States

Objectives

The objective of this systematic review is to identify and critically appraise models that predict the risk of cardiotoxicity of antineoplastic agents for treatment of breast cancer.

Method

A protocolized, systematic strategy was used to search PubMed and EMBASE to identify published multivariable models of development or validation that described model performance. For eligible studies, we will extract data and assess the risk of bias (ROB) and applicability.

Results

The literature search identified a total of 2,937 citations. After deduplication, 2,816 citations were screened by multiple authors. In total, 9 publications were identified that developed or validated a risk prediction model for cardiotoxicity of a US Food and Drug Administration-approved antineoplastic agent used for treatment of breast cancer. Years of publication ranged from 2008 to 2021. Most publications defined cardiotoxicity as a left ventricular ejection fraction (LVEF) decline of =10%, decline to <45% to 50%, and/or diagnosis of heart failure, measured between 1 and 2 years. Models were developed for trastuzumab (n=6), anthracyclines (n=2), and anthracyclines with or without trastuzumab (n=1). Most models used prospectively collected data (n=8) and described model development (n=8). The most common predictor variable was baseline LVEF (n=6). The majority of models were developed using logistic regression (n=8). Reported measures of model performance by area under the receiver operating characteristic curve ranged from 0.56 to 0.92. Additional data from included studies will be extracted based on guidance from the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies (CHARMS) checklist to describe model characteristics (including data source, predictor variables, and sample size) and performance (including measures of sensitivity, specificity, calibration, and discrimination). The ROB and applicability will be assessed using the Prediction Model Study Risk of Bias Assessment (PROBAST) tool. The review is ongoing and final results are pending. To our knowledge, this will be the first systematic review to assess risk prediction models for cardiotoxicity of antineoplastic agents used in treatment of breast cancer.

Conclusion

We identified 9 risk prediction models for cardiotoxicity of antineoplastic agents used in treatment of breast cancer. These most commonly assessed risk associated with trastuzumab therapy, used prospectively collected data and logistic regression modeling, and indicate potential for fair to good model performance. The ROB and applicability of these models will be reported based on guidance from PROBAST.