Methods
Consecutive patients referred to GenesisCare Cardiology clinical testing facilities in Brisbane, Australia were studied prospectively between April 2014 and November 2015. Standard Bruce protocol treadmill testing12 with digital gated echocardiography before and after exercise was performed. Patients requiring dobutamine stress echocardiography, patients with resting left bundle branch block, those with a paced rhythm or atrial fibrillation, those with reduced ejection fractions (EF < 50%), those with significant valvular regurgitation or stenosis and patients with aortic or mitral valve replacements or mitral valve repairs were excluded. Indications for the test included chest pain and dyspnea for investigation.
The echocardiographic images were acquired in the parasternal long axis, short axis, apical four, two, and three chambers. Ejection fraction was measured by Simpson’s method. The SV was estimated by the formula SV = cross sectional area (CSA) x velocity time integral (VTI). The CSA of the aortic annulus was calculated using the formula 0.785 x d2, where d is the left ventricular outflow tract (LVOT) diameter measured in the parasternal long axis plane. The VTI was obtained by measuring the area under the curve of the LVOT pulsed wave Doppler tracing in the apical five chamber view, using the leading edge of the velocity spectrum.8-10 (see Figure 1 & 2) The CO was calculated by multiplying the VTI by the heart rate at the time of the tracing.8-11 The resting SV was estimated during the pre-exercise echocardiography scan. The post-exercise value was captured after the peak regional wall assessment,7between 60 and 120 seconds after completion of exercise (see Figure 1). The LVOT diameter was not expected to change post exercise.10 The change in stroke volume (ΔSV) in millilitres (ml) was defined as the SV estimated post exertion minus the resting SV.
Exercise was replicated and quantitated using General Electric medical grade treadmills using Case systems (Milwaukee, USA). Standard Bruce protocols were used to produce exercise stress in a controlled and reproducible environment. Imaging was acquired utilising General Electric Vivid e9 (Horton, USA) and Vivid 7 (Horton, USA), Siemens SC2000 and SC2000 Prime (Mountain View, USA) and the Phillips ie33 and Epic (Best, The Netherlands) echocardiography scanners. Metabolic equivalents (METs) were used to represent exercise capacity, as per standard protocols.1,2,12
The echocardiogram was performed by cardiac sonographers with subspecialty training in stress echocardiography. All tests were supervised and read by cardiologists with subspecialty training in stress echocardiography, and an exercise physiologist. Results were then over-read, standardized and recorded by a single stress echocardiography specialised cardiologist, blinded to the results and the outcomes.
Ischemic stress tests were defined as those with new regional wall motion abnormalities in two contiguous segments, or cavity dilatation with a lack of cardiac augmentation (direct comparison of the pre- and post-exercise echocardiographic images). Non-ischemic stress echocardiograms were defined as having no evidence of myocardial ischemia on the stress echocardiogram (no new regional wall motion abnormalities, no cavity dilatation, and appropriate augmentation of cardiac contractility post exercise), with an appropriate augmentation of left ventricular function, as assessed by echocardiography. A wall motion score index (WMSI)1 was calculated for all studies by a single stress echocardiography specialised cardiologist, blinded to the results and the outcomes. A subset of patients underwent anatomic testing for coronary obstruction (invasive or CT coronary angiography), as determined by the treating physician, independent of this study.
In order to test the hypothesis that SV could be a marker of abnormal cardiac outcomes, a two-phase study design was devised. An initial cohort of patients was selected and examined (the derivation cohort), to test the hypothesis and if valid to then determine the SV cut-off value. To confirm the concept that the LVOT diameter did not change significantly post exercise, it was measured before and after peak exertion in this derivation cohort. The validation cohort, a subsequent larger group of patients, was then studied to examine the predictive value of an abnormal SV response based on prognostic outcomes.
Patients’ medical records were reviewed for up to five (5) years after the baseline SE data collection. Predetermined adverse cardiac endpoints were angina as determined by the patient’s clinical cardiologist, acute coronary syndrome, cardiac revascularization (percutaneous intervention or coronary artery bypass grafting, worsening New York Heart Association (NYHA) class, a reduction in EF of greater than 10%, and cardiovascular death. A diagnosis of angina, or worsening NYHA class was determined by the treating cardiologist at the time of follow-up. This assessment was made independent to the study, and the treating specialist was blinded to the results. The time interval to a patient’s first adverse cardiac event (a composite of the above) was analysed using Cox proportional hazards regression, adjusted for patient’s age, sex, ejection fraction, exercise capacity and pre-test Framingham risk,13 at the time of the stress echocardiogram.
The study design and methodology were reviewed and approved by the Brisbane GenesisCare Cardiology echocardiographic working group ethics subcommittee. The research protocol was carried out in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki). All patients provided written consent.
A multivariable model incorporating the resting and peak change in stroke volume for the prediction of subsequent composite adverse cardiac events in adult patients referred for stress echocardiography was developed.14 Data were exported in Microsoft Excel format for subsequent statistical analysis using MedCalc Statistical Software (MedCalc Software, Ostend, Belgium). Demographic and baseline data were tabulated and summarised using descriptive statistics. No imputation for missing data was performed and the number of analysed observations was reported for each summary proportion. Unpaired t-tests were used for cohort analysis of continuous data for separate populations. Paired t-tests were utilized to compare continuous data for dependent variables. The Mann-Whitney test for unpaired samples was used to analyse the median wall motion score index. Data expression was presented as the mean plus or minus the standard deviation.
The outcome for time-to-event analyses within each patient was the first observed element of the study composite outcome of pre-determined endpoints. Results were presented graphically using Kaplan-Meier curves assessed by log-rank tests, as well as tabulated unadjusted and adjusted hazard ratios with 95% confidence interval (CI) returned by Cox proportional hazard regression models. Variables included in adjusted Cox models were the gender, age in years at the time of baseline assessment, ejection fraction percentage, exercise capacity in METs and pre-test Framingham risk score. The selection of the variables was based on the baseline characteristic differences between groups. The proportional hazards assumption was evaluated using scaled Schoenfeld residuals and visual assessment of log-log plots. The cut-point for SV was determined using receiver operator curves (ROC).14,15