Results
From July 2016 to March 2019, 72 patients were evaluated, of which 3 patients were excluded due to significant GLS reduction at T2, leaving 69 patients for analysis, with a mean age of 52±13 years and female predominance (97%). Cardiac comorbidities included 2 (3%) patients with ischemic heart disease of whom 1 (1%) was post myocardial infarction and cardiovascular risk factors (hypertension, hyperlipidemia and diabetes mellitus) were relatively common ranging from 12 to 30 percent of the patients. Due to comorbidities 21 patients (30%) were treated with either angiotensin converting enzyme inhibitor (ACEI), angiotensin II receptor blocker (ARB) or beta blockers (BB) and 11 (16%) with statins. All baseline characteristics are summarized on Table-1.
All patients were diagnosed with breast cancer, of whom 9 (13%) were metastatic. Cancer therapy, other than ANT (100%), included paclitaxel (88%), and the recombinant humanized monoclonal antibodies against HER2 Trastuzumab (21%) and Pertuzumab (19%) (Table 1).
All patients performed speckle strain at T1 and T2 echocardiography assessment and 50 patients performed speckle strain also at T3. All patients had normal baseline LVEF (mean 60±1%) and normal GLS (mean -21±2%). Clinically significant reduction of GLS was observed in 10 (20%) patients at T3, however only 2 (3%) patients showed a reduction of 10% in LVEF (Table 1), which considered to be clinically significant (7).
When comparing the association between Dst to e’ average, significant inverse correlations were noted in anteroseptal, apical and middle segments (B between -20.8 to -15.2, p<0.008) and in the average of all LV segments (B=-14.9, p=0.045). When comparing the association of Dst and E/e’ average, significant positive correlations were seen in the lateral, posterior and basal segments (B between 12.1 to 20.2, p< 0.04) as well as the average of all segments (B=19.2, p=0.024). These associations show that poorer diastolic function by common measurements, is associated with longer diastolic time. (Table 2)
Using logistic models to assess the predictive ability of the relative change between the various diastolic strain parameters in T2 and T1, the only Dst measurement that was able to show significant prediction capabilities for significant GLS reduction, was the one for the basal segment time (OR 1.09 for every 1% increase in basal diastolic time, p=0.03, supplementary Table-S1). Therefore, we continued investigation of this predictor only, in a multivariate fashion.
After construction of a multivariate logistic regression model of significant GLS reduction between T3 and T1, with covariates of relative basal Dst change between T2 and T1, relative GLS reduction between T2 and T1, baseline cardiac risk factors, cardiotoxic cancer therapy used and cardioprotective medication used, and application of a model selection algorithm as described in the methods section we ended up with a final multivariate model that included relative change in basal Dst, relative change in GLS, hypertension, hyperlipidemia and Pertuzumab therapy. Of those, the only significant predictors were the relative change in the basal Dst (OR 1.3 per 1% change, p=0.022, and Pertuzumab treatment (OR 159.1, p=0.035). (Table-S2)
The predictive ability of basal Dst for significant GLS reduction between T3 and T1 was moderate with a Youden index of 0.38 and an AUC of 0.732 (95% CI 0.523-0.940). When building a ROC curve for the prediction of the multivariate logistic model, predictive ability was better with a Youden index of 0.8 and an AUC of 0.950 (95% CI 0.888- 1) (p for AUC comparison = 0.05, Fig. 2). The net reclassification index (NRI) for basal Dst added to the multivariate logistic model was 0.48 (95% CI of 0.10 to 0.0.86) composed of a positive NRI of 0.50 (95% CI 0.12 to 0.87) and a negative NRI of -0.02 (95% CI -0.08 to 0.00) showing that adding the basal Dst to the multivariate model was overall significantly beneficial to its predictive ability.