Feasibility of myocardial perfusion assessment with contrast echocardiography: can it improve recognition of significant coronary artery disease in the ICU?

We performed a prospective, observational study in the ICU at Nepean Hospital, Sydney, Australia, between May 2014 and January 2017 on non-consecutive patients (S.O. is the sole MCPE operator in the unit, hence dependent on availability). All patients or authorized representatives (next of kin) gave written consent to be involved in the study which was approved by the Nepean Blue Mountains Local Health District human research and ethics committee (study 15/17-LNR/15/NEPEAN/37). Inclusion criteria were as follows: adult (> 18 years), raised (high sensitivity) troponin I levels (greater than 50 ng/L), acute coronary artery occlusion being considered, and a request made for consideration of coronary angiography. Patients were excluded if they had urgent angiography already performed due to STEMI criteria being met, were unable to have apical echo imaging performed (2 patients), past medical history of ischaemic heart disease, contraindications to echo contrast (allergies, known significant intracardiac shunts, severe pulmonary hypertension), significant valvular abnormalities, pregnant, and study refusal.

Six intensive care specialists with a high level of echo experience (i.e. DDU qualification or equivalent) as well as seven cardiologists were invited to review all relevant patient data and echo imaging to provide an estimate, based on clinical acumen, of the likelihood of the presence of significant CAD needing inpatient intervention on a Likert scale. Data included admission history, ECG, serum troponin levels, and past medical history (particularly including history of hypertension, diabetes, smoking, family history). In addition, time of admission, imaging and troponin, APACHE III, SOFA score, and haemodynamic data were recorded. The presence of significant CAD was assessed by coronary angiography, nuclear imaging, MRI, CTCA, or normal repeat echocardiography shortly after initial imaging in patients with stress-induced cardiomyopathy diagnosis.

A full comprehensive echo was performed initially by S.O. or trained sonographers with either a Vivid 9 or Vivid I echo machine (General Electric, Boston, MA, USA). The studies and analysis were performed in accordance with the leading echo organization guidelines [7, 8] to obtain LV size, ejection fraction, and regional wall motion abnormalities to gain a wall motion severity index score: average of 16 segment model score based on normal thickening = 1, hypokinesis = 2, akinesis = 3, and dyskinesis = 4. In addition, speckle tracking echocardiography (STE) analysis was also performed to determine global longitudinal strain. STE analysis was completed in manner as previously described [9, 10] by S.O. (who has performed over 1000 analyses) in accordance with a consensus document from leading organizations [11]. S.O. then performed the MCPE examination with a Vivid 9 echo machine with a M3S matrix array transducer, at the earliest time frame possible from the inclusion criteria being met.

The contrast agent Definity® was used for MCPE analysis. The microspheres have a mean bubble size of 1–10 μm enabling passage through the pulmonary vasculature [4]. Definity was drawn up into a 20-ml syringe with normal saline and injected in one to two ml increments to enable a homogenous contrast enhancement in the myocardium with no attenuation. Once imaging was optimized, a flash of higher-intensity ultrasound (MI ~ 1.0) for 30 frames (to cover at least one systolic period) timed to coincide with systole on the ECG, was manually triggered. Images were recorded for two to five beats prior to the flash and 8–15 beats after the flash to adequately assess for replenishment.

Images were transferred to an EchoPACS reporting station (General Electric, Boston, MA, USA) for off-line quantitative analysis for each segment that was able to be visualized. Analysis was attempted to be performed in a blinded fashion to achieve outcome results in a manner as previously described in other studies [12, 13]. Subjective (visual) analysis was performed with a simple scoring system: 0 normal, 1 contrast perfusion deficit. Quantitative analysis was performed by measuring ultrasound signal intensity in the ‘region of interest’ (ROI) at each myocardial segment following a standard 16-cardiac segment model (see Fig. 2). The ROI size was optimized to include as much of the myocardial segment as possible while avoiding the low-intensity signals from the pericardium or high-intensity signals from the LV cavity. The first end-systolic frame after the ‘flash’ was considered t₀, and signal intensity (SI) was calculated for each ROI at each end-systolic frame and plotted against time and fitted to the exponential function: y(t) = A(1 − e^{−β(t − t0)}) + C. y is the SI in the ROI at the end-systolic frame, A is the plateau SI corresponding to myocardial blood volume, β is the exponential decay function (decay constant) representing the rate of SI rise, and C is the intercept at the origin reflecting the background intensity level. A × β provides an estimate of the initial rate of contrast replenishment, and this provides a surrogate of myocardial blood flow [14].

Feasibility of both subjective and quantitative analysis was assessed at segmental and coronary artery territory distribution (see Fig. 2) as well as the apex vs mid vs basal levels. Segments were excluded from analysis if the myocardium vasculature lacked opacification on visual and quantitative analysis. Coronary artery territories were considered ischaemic if a third (or more) of the segments in that territory had impaired contrast filling. The left anterior descending (LAD) and right coronary artery (RCA) territories were considered unable to be assessed if two segments were unable to be examined. The left circumflex coronary artery (LCx) territory was considered unable to be assessed if one segment could not be examined.

Due to the exploratory nature of the study, a sample size was not formally calculated and a number of 40 patients were felt suitable to make an initial assessment and guide future research. Statistical analysis was performed with JMP Pro version 13 (SAS Institute Inc., Cary, NC, USA). Continuous variables are expressed as mean ± standard deviation (SD) if normally distributed, and as median with interquartile range (IQR) if not normally distributed. Normality was assessed using the Shapiro-Wilk test. Feasibility was defined as the number of segments where analysis was possible compared to total segments from all included patients. A between-group comparison for continuous data was performed by the Student t test and non-parametric or non-normally distributed data by the Wilcoxon signed-rank/Kruskal-Wallis (rank sum) test. Categorical data was compared by Pearson’s chi-squared analysis or Fisher’s exact test. All probability values are two-sided, and p values < 0.05 are considered statistically significant. Logistic regression was used to assess the association between the presence of acute coronary artery occlusion and clinical acumen and/or subjective or visual MCPE analysis. Receiver operating curves were analysed to determine optimal MCPE values for the presence or absence of acute coronary artery occlusion. Inter-rater variability was performed on a random 15% of patients for myocardial blood flow estimation by MCPE and assessed by absolute difference and expressed as a percentage of their mean.

Feasibility and values for each myocardial segment analysis technique are shown in Table 2. Quantitative analysis was estimated to be performed 24 h–3 weeks from time of echo. 2D segmental thickening analysis showed the greatest feasibility (in 90–100% of patients) and subjective MCPE analysis the least (20–53% of patients). Despite both 2D segmental thickening assessment and longitudinal strain analysis by STE being feasible in the majority of segments assessed, there were no significant differences seen in wall motion score index or longitudinal strain analysis values between groups with significant CAD and those with no CAD diagnosed. However, in both subjective and quantitative MCPE assessment, significant differences were seen (except in the right coronary artery territory in quantitative assessment). Quantitative MCPE analysis had better feasibility than subjective MCPE assessment. The LAD territory was the most feasible to be assessed (65% by quantitative MCPE and 53% by subjective MCPE analysis) and the LCx the least (20% by quantitative MCPE and 33% by subjective MCPE analysis). The feasibility of MCPE subjective and qualitative analysis was greatest at the apical level (89% and 90% respectively). The mid-level was easier to analyse (66% and 68% respectively) than the basal level (39% and 48% respectively).

Based on ROC curve analysis, the optimal MCPE cut-off value for the presence vs absence of significant CAD is 2.9 dB/s (see Fig. 3) which had 67% sensitivity and 88% specificity. We found a positive predictive value of 50% and a negative predictive value of 91%. Of the 3 patients who were found to have perfusion deficits on MCPE, 2 had angiography and the other an MRI. Association between clinical acumen and quantitative MCPE or subjective MCPE analysis in predicting significant CAD is shown in Table 3. Clinical acumen showed no significant association in predicting presence of significant CAD (OR 0.64, p = 0.091); however, if quantitative or subjective MCPE analysis was included in the model, incremental improvement and significant association was seen (OR 17.15, p = 0.013; OR 23.05, p = 0.010 respectively). The best association was seen with subjective MCPE analysis alone (OR 33.0, p = 0.003).

Inter-rater variability was assessed by M.S. vs S.O. in random 6 patients (15%). Inter-rater variability was reasonable with mean absolute difference in estimated blood flow (± SD) of 0.31 dB/s (± 1.5) and expressed as percentage of the mean of 9% (± 36).

There are several limitations to our small, single-centre, observational study. The primary issue is the lack of a recognized reference standard to exclude acute coronary artery occlusion (angiography or cardiac MRI) in a significant portion of our patients (e.g. 28% of our patients had a simple normal follow-up echo). However, often, once an acute event had settled, the risk of angiography may outweigh the benefit (e.g. likely stress-induced cardiomyopathy in a patient with subarachnoid haemorrhage). Those with an initial abnormal echo were not excluded and then normalized findings at a later date still had coronary artery disease similar to those having exercise stress tests; however, we did find a high specificity and negative predictive value in those with significant CAD found. In addition, the analysis was performed by a single operator (S.O.) who was at times the treating clinician. The blinding of the analysis was therefore, at times, not possible. Patients were not consecutive as S.O. was the sole MCPE operator in the unit and patient inclusion was based on other clinicians highlighting potential subjects, during work hours. Several patients are likely to have been missed and this suggests selection bias. However, this was primarily a feasibility study and information may be useful as pilot data to guide future studies. The analysis of our data into coronary artery territories may also be inaccurate as individual patient coronary artery blood flow does not always follow the standard anatomical boundaries.

The use of MCPE in determining large defects in myocardial blood flow may be useful in the ICU environment. Further research should be done in this area to try and help improve our early recognition of significant CAD. Studies should focus on true blinded assessment, multi-operator analysis, and objective MCPE analysis being performed immediately after the study has taken place. Additional studies may assess the significant variation seen in clinicians estimating the likelihood of significant CAD being present and if MCPE analysis could help reduce some of this variation.