Identifiers of CAE by Echocardiographic Parameters
Cox regression showed that among the significant univariate factors of 3D Echocardiography after the completion chemotherapy, LVEF, LVESV, LVGLS, LVGCS, RVEF, and RVESV had P values < 0.05 associated with CAEs. Neither RVEDV nor RV strains were associated with CAEs (Table 2).
Based on the receiver operating characteristic curve (Figure 2), a cutoff RVEF value of less than 46.33% (sensitivity, 72.2%; specificity, 82.1%; AUC=0.811; p<0.001) after the completion anthracycline treatment would have correctly identified 13 of 18 of the study patients (72.2%) who developed CAE and 64 of 78 patients (82.1%) who did not have CAE. While LVGLS<17.29% (sensitivity, 66.7%; specificity, 82.1%; AUC=0.787; p<0.001) was able to discriminate between patients with and without CAE (Figure 2A). When LVGLS and RVEF were combined, the sensitivity increased to 88.9%, the specificity was 82.1%, and the AUC increased to 0.882 (Figure 2B).
Figure 3 demonstrates the Kaplan-Meier survival curve comparing normal RVEF (≥46.33%) with abnormal RVEF (<46.33%). Of note, 72% of the patients with abnormal RVEF (13 of 18) suffered from CAE during the follow-up, compared with only 18% of the group with normal RVEF (14 of 78).
A second multivariate analysis was done starting with the significant independent variable (GLS). RVEF was added via stepwise block analysis to these clinical parameters and was found to be significantly important (p<.001), as detailed in Figure 4.
Intraclass correlation coefficients (ICCs) showed interobserver and intra-observer reproducibility for our facility were 0.827 and 0.844 for LVGLS, respectively, and 0.939 and 0.859 for RVEF, respectively (Supplemental Table 2).