ABC | Volume 110, Nº2, February 2018

Original Article Gripp et al Global Longitudinal Strain Accuracy for Cardiotoxicity Arq Bras Cardiol. 2018; 110(2):140-150 Tests at different follow-up times were randomly drawn, defining a sample of approximately 10% of all calculations of the GLS analyzed during the study. Data were reassessed by the same observer, blind to the treatment instituted and specialized in the method, so that intraobserver agreement could be assessed. The interobserver analysis was performed by another professional, also specialized in the method with experience in GLS assessment, the major variable of this study. The second observer used the same clip selected by the first observer, with predefined configurations, such as depth, gain, value of pulse repetition frequency (PRF); however, the new regions of interest for myocardial markers were freely chosen during the reanalysis. If the observer agreed on the region of interest marked, the next step would be the approval of the six segments according to the walls assessed. Upon approval with a command on the working station screen, the values of GLS and segment strains were calculated and demonstrated by use of bull’s eye. Therefore, the calculations of the LV GLS, RV strain and RV free wall strain were repeated at the working station by the second observer, who was blind to the time the Doppler echocardiography was performed, the treatment and the patient’s outcome. Statistical analysis Data were prospectively recorded in the program SPSS 15.0 for Windows, also used for statistical analysis. The categorical variables were expressed as frequency, being compared by use of chi-square test. The continuous variables were expressed as mean and standard deviation or median and interquartile range, according to their distribution, and compared by use of paired Student t test or MannWhitney U test. The baseline values and those at 3, 6, 9 and 12 months from Doppler echocardiography were compared by use of one-way analysis of variance (ANOVA). Cox regression analysis was used to identify independent echocardiographic variables predictive of cardiotoxicity. Receiver operating characteristic (ROC) curves were created to define the most accurate cutoff points for the continuous variables independently associated with the event assessed. The intra- and interobserver variabilities were analyzed with intraclass correlation coefficients, and Bland-Altman plots were created to show the results of the interobserver analyses. For all tests, the statistical significance level adopted was p ≤ 0.05. Results Of the 58 female patients consecutively referred to the Oncology Outpatient Clinic of the HUCFF, 49 were included in this study. Nine patients were excluded because of their high body mass index (BMI), which generates an inappropriate acoustic window to the LV GLS acquisition and EF calculation with the Simpson’s method. Themean ageof thepopulation studiedwas 49.7±12.2 years, and the follow-up duration, 381 ± 29,8 days. Table 1 shows the baseline characteristics of the patients included in this study and of those excluded from it. Regarding the oncological data, the most common histological type of tumor was invasive ductal carcinoma, observed in 70% of the patients. In 51% of the patients, the tumor was located in the left breast, 40.8% of the patients underwent surgery before chemotherapy, and 53.1%, radiotherapy (all of them after chemotherapy). The patients underwent serial Doppler echocardiography, the first test being performed prior to treatment, and the following tests, on the third, sixth, ninth and twelfth months, in accordance to the study protocol. The LV GLS and the EF (Simpson’s method) were obtained at all tests of the 49 patients. The mean time between undergoing the first Doppler echocardiography and initiating the antineoplastic treatment was 9 days. The population studied and that excluded from the study were compared, and a similarity between the groups was observed. Intra- and interobserver analysis of global longitudinal strain The intraobserver intraclass correlation coefficients for LV GLS, RV strain, and RV free wall strain were 0.97 (95%CI: 0.91‑0.99), 0.98 (95%CI: 0.93-0.99) and 0.98 (95%CI: 0.95-0.99), respectively. The interobserver intraclass correlation coefficients were 0.97 (95%CI: 0.92-0.99), 0.97 (95%CI: 0.92-0.99) and 0.98 (95%CI: 0.93-0.99), respectively. The results showed excellent inter- and intraobserver agreements. The excellent result of the interobserver analysis of the LV GLS, RV strain, and RV free wall strain can also be observed in Figures 1A, 1B and 1C (Bland-Altman plots). Characteristics of the population that developed cardiotoxicity All patients in our study received anthracyclines, and 80% of them underwent radiotherapy after chemotherapy. During the follow-up, five patients (10%) developed cardiotoxicity, two on the third month and three on the sixth month. Despite the lack of a statistically significant association, the mean age of the patients with cardiotoxicity was higher than that of the 44 patients without it. In addition, 80% of those patients underwent radiotherapy, which is clinically relevant. All patients used anthracyclines. For two patients (40%) who developed cardiotoxicity, trastuzumab was associated to the antineoplastic regimen. The baseline characteristics of the patients who developed cardiotoxicity are shown in Table 2. Description of the echocardiographic parameters The means of the echocardiographic variables of the patients with and without cardiotoxicity are shown in Table 3. On the thirdmonth, the mean LV GLS, as well as its difference regarding the baseline value, were significantly higher in the group with cardiotoxicity. Although the EF value on the third month differed between the groups, its difference from the baseline value did not behave like that. On the sixth month, there was a significant drop in the EF and LV GLS, in addition to changes in the S wave of the left ventricle and E/E’. Table 4 shows the five cases of cardiotoxicity. 142