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 dysfunction occurring on the sixth month, with sensitivity of 78%, specificity of 79% and negative predictive value of 93%. The sample calculation of that study was based on the hypothesis that a 14% GLS drop could predict cardiotoxicity, exactly the same value found in our study. It is worth noting that all prognostic models, in addition to predictive accuracy, should have the variables easily obtained. Doppler echocardiography is widely available and easily accessible, involves no radiation, being performed at the bedside. Its use in the follow-up of patients with breast cancer is a criterion of quality in healthcare services, mainly when using GLS, capable of predicting cardiotoxicity in those patients. However, it should be performed by echocardiography professionals trained in the method, with excellent image acquisition, to minimize the intra- and interobserver variabilities, using the same device and software, creating an individualized set for image acquisition and subsequent assessment. In our study, those values were found using the GE software. The different brands of devices have different normality range values. An agreement regarding those values has not been achieved between the manufacturers. Most studies and guidelines use a percentage variation of strain to define the presence of cardiotoxicity. Using the patient’s baseline measures as control, and guaranteeing that all measures are taken with the same equipment and technique, the variations seem more reliable. Study limitations The sequential echocardiographies were performed by the same examiner. Although the examiner was blind to the treatment instituted, an influence of the previous assessment on the subsequent tests could exist. However, the interobserver analysis showed an excellent correlation between data, and the second observer was blind not only to treatment, but also to the echocardiography times and previous results. Thus, although an assessment bias might have existed, it would not be strong enough to alter the results found. The strain calculation requires an appropriate acoustic window. The patients excluded were those with the highest BMI, who would be at higher risk for cardiotoxicity according to the literature. In addition, in patients undergoing left breast surgery before the antineoplastic treatment, the presence of the expander or the surgical wound itself could interfere with the analysis. Limitations regarding the method also apply, and could be related to the test being performed by an untrained professional, or might be related to the devices available, taking into account that the strain values vary according to the brand of the device used. Our study showed a low incidence of cardiotoxicity, which could limit the multivariate analysis. Currently, the literature is reviewing the assumption that a robust multivariate analysis should involve at least ten outcomes for each variable analyzed. There are three well-known simulation studies that assess that criterion for regression models, and they do not agree. Currently, in addition to the number of events per variable, the regression model depends on several other factors, such as the association of variables and outcomes, and some statistical studies report on the use of a smaller number of outcomes for each variable analyzed. 23,24 Conclusions The incidence of cardiotoxicity associated with the antineoplastic treatment for breast cancer was 10% in our institution. Our population with a low cardiovascular morbidity profile showed no association between cardiotoxicity and the risk factors classically described, such as clinical and anthropometric variables and treatment. A significant LV GLS drop was observed from the third month onward, characterizing that variable as an independent predictor of cardiotoxicity, with a cutoff point of an absolute LV GLS value of -16.6% or a percentage LV GLS variation of -14%. Acknowledgements The authors thank all the staff of the Oncology Service of the Clementino Fraga Filho University-affiliated Hospital for making this study possible. Author contributions Conception and design of the research, Writing of the manuscript and Critical revision of the manuscript for intellectual content: Gripp EA, Oliveira GE, Feijó LA, Garcia MI, Xavier SS, Sousa AS; Acquisition of data: Gripp EA, Oliveira GE; Analysis and interpretation of the data: Gripp EA, Feijó LA, Garcia MI, Xavier SS, Sousa AS; Statistical analysis: Feijó LA, Garcia MI, Xavier SS, Sousa AS. Potential Conflict of Interest No potential conflict of interest relevant to this article was reported. Sources of Funding There were no external funding sources for this study. Study Association This article is part of the thesis of Doctoral submitted by Eliza de Almeida Gripp, from Universidade Federal do Rio de Janeiro. Ethics approval and consent to participate This study was approved by the Ethics Committee of the Hospital Universitário Clementino Fraga Filho – UFRJ under the protocol number 926775. All the procedures in this study were in accordance with the 1975 Helsinki Declaration, updated in 2013. Informed consent was obtained from all participants included in the study. 148