BACKGROUND: Assessment of left ventricular systolic function is necessary during trastuzumab-based chemotherapy because of potential cardiotoxicity. Deformation indices have been proposed as an adjunct to clinical risk factors and ejection fraction (EF), but the optimal parameter and optimal cutoffs are undefined. The aim of this study was to determine the best means of early detection of subsequent reduction of EF in patients with breast cancer treated with trastuzumab. METHODS: Eighty-one consecutive women (mean age, 50 ± 11 years) receiving trastuzumab were prospectively studied, 37 of whom received concurrent anthracyclines. Conventional echocardiographic indices (mitral annular systolic [s'] and diastolic [e'] velocities) and myocardial deformation indices (global longitudinal peak systolic strain [GLS], global longitudinal peak systolic strain rate [GLSR-S], and global longitudinal early diastolic strain rate [GLSR-E]) were measured at baseline and at 6 and 12 months. Cardiotoxicity was defined as a >10% decline as a percentage of baseline EF in 12 months. RESULTS: In the 24 patients (30%) who later developed cardiotoxicity, myocardial deformation indices decreased at 6 months (GLS, P < .001; GLSR-S, P = .009; GLSR-E, P = .002 vs baseline), but e' was unchanged. The strongest predictor of cardiotoxicity was ΔGLS (area under the curve, 0.84); an 11% reduction (95% confidence interval, 8.3%-14.6%) was the optimal cutoff, with sensitivity of 65% and specificity of 94%. In sequential models, the clinical model (χ(2) = 10.2) was improved by GLSR-S (χ(2) = 14.7, P = .03) and even more so by GLSR-E (χ(2) = 18.0, P = .005) or GLS (χ(2) = 21.3, P = .0008). Discrimination improvement by adding GLS was confirmed by an integrated discrimination improvement of 18.6% (95% confidence interval, 8.6%-28.6%; P = .0003). A net 29% of the patients without events were reclassified into lower risk categories, and a net 48% of the patients with events were reclassified into higher risk categories, resulting in a total continuous net reclassification improvement (>0) of 0.77 (95% confidence interval, 0.33-1.22; P = .036). CONCLUSIONS: GLS is an independent early predictor of later reductions in EF, incremental to usual predictors in patients at risk for trastuzumab-induced cardiotoxicity.