The study was conducted at Houston Methodist Hospital with patients selected from the Houston Methodist Oncologic Pharmacy Registry, which includes cancer-related therapy information from the Houston Methodist Cancer Center and 7 hospitals and affiliated cancer centers. The institutional review board of Houston Methodist approved the protocol.
Study Population and Pre-specified Risk Factors
Patients who had a baseline STE within 3 months before the initiation of anthracycline therapy and at least one follow-up echocardiogram 3 months after start of treatment from July 1, 2013, and July 1, 2019 were identified. Patients with a suboptimal strain study (impaired regional tracking in >2 myocardial segments), an abnormal left ventricular ejection fraction (LVEF) <50%, left bundle branch block, history of coronary artery bypass grafting (CABG), premature ventricular contractions, or a systolic blood pressure >180 mmHg or diastolic blood pressure >100 mmHg before treatment were excluded (Figure 1). Each chart was accessed for baseline demographics (including age, vitals, cardiovascular (CV) risk factors, medical comorbidities), echocardiography studies, and clinical outcomes (symptomatic heart failure, CV mortality, and non-CV death).
Pre-existing cardiovascular risk factors and medical comorbidities extracted from the electronic medical records, included: body mass index (BMI), tobacco use, diabetes, hypertension, hyperlipidemia, coronary artery disease, hypothyroidism, and a family history of atherosclerotic cardiovascular disease. Cumulative anthracycline dose was calculated based on the following rapid infusion doxorubicin cardiotoxicity equivalence: epirubicin, 0.66; daunorubicin, 0.75; and idarubicin, 0.53 (15,16).
Echocardiography Strain Protocol and Endpoints
The primary endpoint of the study was cardiotoxicity, defined as an LVEF <50% and absolute decline >10% in LVEF from baseline echocardiography. LVEF was calculated using quantitative (modified Simpson’s biplane method) and/or visual analysis of LVEF according to the American Society of Echocardiography standards (17).
The vendors utilized in the strain analysis included the most recent upgraded versions of GE EchoPAC and TomTec Image-Arena. The endocardial borders were traced from 3 apical viewers. Global longitudinal strain was calculated by measuring the entire endocardial line length at the end-diastole and end-systole in each view and averaging the results from the 3 views. Each measurement was taken from the average of 3 consecutive cardiac cycles. The strain quality was assessed objectively by a second observer as bad if more than two segments were un-visualized.
Secondary end points of all-cause mortality and CV mortality were assessed between the patients with normal and impaired baseline GLS. Impaired baseline GLS was defined as ≥-18% (18).
Statistical Analysis
Baseline characteristics were summarized according to cardiotoxicity status and baseline GLS status (abnormal vs normal). All data were presented as mean ± SD for continuous variables and number and % for categorical variables. Chi-square or Fisher’s exact test for categorical variables and t-test or Mann-Whitney test for continuous variables were used to determine association of pre-treatment variables with impaired baseline GLS and cardiotoxicity. The normality assumption was tested by the Shapiro –Wilk tests.
Kaplan-Meier survival curves for overall mortality was calculated in strata defined by impaired vs. normal GLS. Time-on-study was used as time scale in all survival analyses: Time begins at first anthracycline dose and ends in date of death or date of last follow-up. Two-sided log rank tests defined significance. The overall survival at 1-year, 3-year, and 5-year intervals between impaired and normal GLS were compared with the pseudo‐value approach (19). Cumulative incidence of cardiovascular related mortality was calculated after adjusted for the competing risk of non-cardiovascular death. Univariable and multivariable Cox proportional hazards regression models were used to examine associations between demographics and clinical variables with cardiotoxicity. The multivariable model included baseline GLS and variables selected from backward elimination with the significance level for removal from the model of 0.05. The full model included all statistically significant variables with p-value<0.05 in the univariable models. Subsequently, the same multivariable analysis was done when GLS modeled as a categorical variable (impaired vs. normal). The Cox proportionality assumption was verified by including time-dependent interactions of covariates with survival time in the models. There were no violations of this assumption in all models.
All analyses were performed with STATA version 16 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC). Statistical significance was defined as two-tailed p<0.05 for all tests.