Discussion
Stress echocardiography has become an established and mature technique for the assessment of CAD.1 Interpretation remains challenging, even with experienced practitioners.1,2Additional parameters can be utilised to assist with this difficult interpretation, including strain and strain rate,16,17 and assessments of diastolic function.1,18 None of these have become routine practice.
Left ventricular SV is the amount of blood ejected with each heartbeat. It is the difference between the end-diastolic volume and end systolic volume. 11,19-20 Exercise results in an increase in cardiac contractility. The changes in the mechanical properties of contracting cardiac muscle fibers with changes in inotropy results in an increase in SV. 11,19-20 A decrease in cardiac contractility causes a downward shift in the Frank-Starling curve, with subsequent drop in SV (and an increase in left ventricular end diastolic pressure). 11,19-20 Myocardial ischemia results in changes in contractile state of the ventricle. These changes result in alterations to SV and CO. 11,20-21
Mechanistically, active human systems require a constant blood supply with delivery of oxygen and nutrients and removal of waste products to function normally. Exercise increases these requirements. Any alteration in supply (e.g. inducible myocardial ischemia) reduces provision of these essential roles, resulting in decreased function. At a ventricular level this results in a drop in SV (and CO). 11,19-21These alterations are due to changes in systolic (via a loss of intrinsic inotropy) and diastolic (loss of compliance) function. Left ventricular contractility has been shown to change similarly to SV.20,22 At a basic science level, nitric oxide inhibition at an endothelial level results in hemodynamic changes.24,25
Echocardiography can reliably estimate SV and CO, before and after stress testing. (see Materials and Methods).7,11,18,26,27 The changes in SV and measures of diastolic function have been shown to be reliable when compared to invasive techniques.18,28-31 Changes in Doppler echocardiography have been documented in the setting of CAD.32-34 These concepts contributed to the hypothesis that SV could be a simple and readily measurable parameter to assist with analysis for stress echocardiography.
This study was designed prospectively to measure SV before and after treadmill exercise, and then to subsequently look at adverse cardiac outcomes in follow-up, as a method of validation. Estimation of the SV before and after exertion provided additional, complimentary and confirmatory information when added to the traditional stress echocardiogram analysis. The estimated SV was used for analysis, as it was a purer and simpler value (there was significantly more variation with the CO estimations, due to marked variability between patients with respect to heart rate, at rest and after stress). Indexing SV resulted in a lower AUC and lower sensitivity and specificity. As such SV was preferred to SVi. A significant increase in SV with exertion is a normal response. An inadequate response was shown to be either a decrease or a mild increase (<10ml). This suboptimal SV increment with exertion resulted in significantly worse cardiac prognosis compared to patients who had an appropriate increase in these volumes.
It is proposed that this suboptimal SV change is a marker of ischemia and the increase in events were likely to be driven by this. Its negative predictive value is very high, adding reassurance to negative studies. An abnormal result associated with a non-ischemic SE may prompt consideration for a more careful analysis of the regional wall motion analysis, or further clinical evaluation. The SV analysis appears to provide statistically incremental analytical and prognostic information to the standard SE evaluation.
An interesting subset of patients were those with a non-ischemic SE, but an abnormal ΔSV. These patients had increased risk compared to non-ischemic patients with a normal ΔSV response to exertion, even following adjustment for age, gender, EF, exercise capacity and Framingham risk (see Figure 5). Careful re-evaluation of the SE did not reveal post exertion regional wall motion abnormalities or reduced cardiac augmentation. Looking at the individual cases, the events for this cohort appeared to be ischemic in nature. Possible reasons for an abnormal ΔSV despite a non-ischemic SE could include a false negative SE, endothelial dysfunction, subclinical ischemia, inducible diastolic dysfunction (except that E/e’ did not universally correspond with these abnormal ΔSV cases), exercise induced pulmonary hypertension, or by chance. These patients did not have increased wall thickness on echocardiography compared to those with a normal ΔSV. These changes may represent myocardial alterations that are too subtle for visual detection. As the events in this subset appear to be ischemic, this may the driver and the most likely explanation. This finding suggests an incremental value for assessing ΔSV in SE.
This technique is quick and readily performed and has been shown here to be valid when measured after the routine regional wall motion analysis. It can be performed in the vast majority of patients. It requires no additional equipment or preparation. It adds only seconds to the resting echocardiogram and to the post exercise analysis and can be done after regional wall analysis but before the estimations of diastolic function.18 The proposed cut-off appears to apply for the detection of ischemia and for the detection of adverse cardiac events.
Previous studies have looked at using LV volumes as a marker of coronary artery disease. 35,36 These studies used planimetry and the Simpson’s biplane method. This volume measurement essentially documents a change in EF. This method is more time consuming than the Doppler technique and has limitations including the geometrical assumptions used to assess LV function, alignment issues (particularly a problem during exercise stress echocardiography), frequent foreshortening of the ventricle, and the influence of load-dependent factors during functional assessment.10,37 Suboptimal endocardial definition further reduces the accuracy of this technique. Regional wall motion abnormalities make dynamic assessment even more challenging.10,37 The Doppler estimation of SV measures a change in ventricular volume utilising a different technique which has been shown here to be independent of EF. This Doppler technique has previously been shown to be feasible in dobutamine stress echocardiography to estimate CO 11 but has not been used in the manner described in this study.
There are limitations with respect to this observational cohort. Ideally these measurements would have been compared to invasive measurements to confirm the diagnosis and measure SV invasively, blindly. Practically, ethically and from the clinical perspective, this was not feasible and would have negated the non-invasive attribute of stress echocardiography. The ethical concerns refer to universally invasively testing low to intermediate risk patients. 19,38-40The accuracy of SE for detection of myocardial ischemia has already been previously documented and validated. 1,2,39,40 In general, stress echocardiography is used to attempt to avoid an invasive test. Clinically, a select group of patients (as determined by the treating physician, and independent and blinded to the study) had an anatomical evaluation, which confirmed the efficacy and accuracy of SE. The median WMSI values were significantly different for non-ischemic patients in both the derivation and validation cohorts.
Baseline clinical characteristics were documented but blood tests were not performed. This was done to simplify the study and minimise the intrusion on the volunteers. Baseline bloods potentially could have added further data to the interpretation of these results.
Inconsistency in the measurement of SV may lead to a lack of reproducibility. Significant variability in the SV measurement could compromise the cut-off values. However, the difference between the mean ΔSV and the cut-off values were marked (see Table 2) and unlikely to fall within the margin of error of acquisition. The estimation of SV requires an accurate measurement of the left ventricular outflow tract. A small error in this measurement magnifies the error.10,37 Previous studies have suggested that the echocardiographic measurement of SV is an accurate and reproducible technique.3-5, 44 In order to minimise inter-reader variability, a single experienced, echocardiography subspecialty cardiologist measured all the stroke volume data, blinded to the outcomes. This may reduce the applicability of these data to the wider community but minimised errors.
Whilst patients achieved a mean 98±19% of maximum predicted heart rate at peak exercise, by the time the SV was acquired, this value had reduced to 68±25%, well below the value recommended to maximise sensitivity for detection of ischemia. 38,45,46Despite this, abnormal ΔSV values predicted prognostic differences, suggesting that delaying this measurement until after the regional wall motion analysis still provides valid and valuable additional information to the standard test method.
It is possible that the results were simply due to the presence or absence of ischemia, or reflective of the changes in ejection fraction post exertion. However, patients with a normal ΔSV were seen in the ischemic group, and an abnormal ΔSV was seen in patients with a non-ischemic SE. All patients with a non-ischemic SE had a normal increase in EF post exertion. The data presented here suggest that ΔSV is a more sensitive marker of events than EF or a normal SE response.
The estimation of SV cannot be performed in all patients. Heart rate did not produce any issues, but peak exercise image quality occasionally limited the completeness of the Doppler envelope, resulting in inadequate measurements. In most cases, image quality had less of an impact on the ability to estimate Doppler signals. Irregular electrical rhythms, especially bigeminy resulted in Doppler signals that were difficult to accurately measure. These incomplete measurements only occurred in a very small percentage of patients. Stroke volume could be successfully compared pre and post exercise in approximately 94% of these stress echocardiograms.
In this non-randomized, single center cohort study, there were some apparent differences between the groups at baseline (see Table 1). Despite presentation of the adjusted estimates, these baseline differences may have influenced the outcomes presented here.
The evaluation of the medical records was a potential source of ascertainment bias. It is possible for events to occur at other centres, and potentially not be recorded in these patient reviews. The ischemic patients may have been assessed more closely due to the results of the stress test, resulting in a higher reporting of events. However, given these patients had significantly more abnormal ΔSV measurements, an under-detection of events in this group would not be expected to influence these results. The SV analysis was conducted after events had occurred, and the results were not communicated to the clinicians, as the implications were not apparent at the time of data collection. While failure to appropriately account for missing data in analyses may lead to bias and loss of precision, imputation of missing results also requires additional assumptions. To have completed this analysis with complete data rather than imputation of missing values may not have been associated in an epidemiological context with substantial bias in reported regression estimates.15,41
The multivariable prediction model described in the present study was derived from echocardiographic observations at a single center. There were significantly less women than men (a common problem in cardiac research). This does reflect real world experiences and this particular study population. 18,42,43 The investigators were not blinded to the results of the stress test, making it possible for biases (including ascertainment bias) to occur. Overall event rates were low, especially in the non-ischemic patients. The present model would benefit from validation within an independently collected data set from a separate population.
The broad referral pattern of these patients does suggest that the utility of measuring SV could be extended to the general stress test population.