The present study shows high feasibility of CFVR in a population with generally limited echocardiographic acoustic window. Reproducibility of CFVR was good especially when performed within a week. Agreement with PET was overall modest, primarily because of poor agreement between methods in participants with prior MI whereas agreement with PET was good in participants without prior MI. Both the reproducibility and method agreement sub-studies indicate that variability over time in CFR should be taken into account e.g. when planning studies with purpose of testing the effect of interventions on CFR.
Feasibility of CFVR
Although CFVR has low cost compared to other CFR methods and advances in echocardiographic technology have led to improvement in accessibility, CFVR has not achieved wide routine use in clinical practice, primarily due to concerns of feasibility in representative patient populations and validity of the measures achieved [
38].
We found a feasibility of 97 % in an unselected group of patients with a median BMI above 30 kg × m
−2. This feasibility is comparable to those previously obtained in patient populations not excluding overweight subjects. A large multicentre study comprising 1544 patients had a feasibility of 92 % (exclusive of 2 % for whom examination was stopped prematurely because of side-effects) with contrast used in 36 % [
7]. One study with a population mean BMI of 30 (
n = 38) reported feasibility of 92 % with 10 % requiring contrast usage and other studies not reporting on body habitus have achieved success rates up to 100 % (
n = 124) [
9,
39]. In comparison feasibility of 95 and 98 % has been reported for mitral annular velocities as measured by speckle tracking and PW tissue Doppler, respectively [
40].
Reproducibility of CFVR
LOA of CFVR repeated within a week in the present study is almost identical to LOA of 0.32 in a study of 13 patients referred for CAG in whom CFVR was repeated with a delay of one hour, and to LOA of 10 % in hypertensive patients (
n = 8) examined 3–5 days apart, as well as a CV of 6 % reported for lean healthy subjects (
n = 8) [
12,
18,
41]. In fact the reproducibility could not be expected to be better, since the present within-subject CV, corresponding to error of measurement, is on par with the previously reported observer variability of repeated off-line readings [
24]. The within-subject CV tends to be higher for CFV than CFVR, which probably is due to the fact that minor differences between examinations that may cause difference in measured CFV between examinations (e.g. placement of probe or sample volume and Doppler angle to the flow), tends to have no significant impact on CFVR as long as it is kept constant during the same examination.
CFVR seems to perform as well as several other measures applied in clinical practice. The reproducibility of CFVR repeated within a week is comparable to measures as LVEF by Simpson biplane method, M-Mode mitral annular excursion, and peak early mitral annular velocity (e’) (CV 5–12 %) in terms of CV [
40,
42,
43]. Similarly, we find the overall reproducibility of CFVR to be at least as good as for MFR by PET (CV range 17 to 26 %) [
30,
44‐
46].
The good reproducibility of CFVR has practical implications as it underpins the usefulness of serial evaluation of CFVR e.g. before and after revascularization, or as an outcome measure in clinical as well as preclinical trials of e.g. drug therapy on coronary microvascular function.
Reliability is an estimate of the proportion of all variation that is not due to measurement error [
33,
34]. For laboratory measurements a reliability above 0.90 is desirable [
34]. We report reliability for examinations repeated within a week that is significantly higher, meaning that more than 90 % of the variability in the measurements of CFVR was due to genuine differences in CFVR between the participants. Reliability is influenced by the heterogeneity (variance) of the study population with regard to the measured parameter, i.e. the greater between-subjects variance the greater reliability. As we have made no selection on CFVR values, the reported reliability is representative for a population of overweight and obese revascularized and stable CAD patients. The relatively high reliability of CFVR makes it suited for distinguishing patients on this parameter. This is also reflected in its ability to prognosticate [
6].
Method agreement: CFVR vs. MFR
There are obvious differences between methods of estimating CFVR by TTE Doppler and MFR by PET. CFVR is the ratio of peak diastolic flow velocities whereas MFR estimates the perfusion volume-velocity per myocardial tissue mass (mL × g
−1 × min
−1) for the entire heart cycle. Further, the methods can be affected be different factors. For example, a change between rest and hyperaemia in the calibre of the artery where the sample volume is placed would have effect on the flow velocity and the measured CFVR. Whereas,
82Rb extraction can be decreased by severe acidosis, hypoxia, and ischemia; thus in addition to blood flow,
82Rb uptake could be affected by metabolism and myocardial cell integrity [
47].
We found 2SD ranging from 0.49 to 1.67. For comparison, we have knowledge of only one study; Saraste et al. examined CFVR and MFR
LAD in 10 young, healthy participants with a mean delay of 13 days, and reported LOA corresponding to mean difference (2SD) of −0.20 (1.03) [
18]. Thus, agreement for obese CAD patients performed within a week, −0.14(1.17), is of comparable magnitude to that of the healthy, lean and young subjects.
Agreement between CFVR and MFR
global was as good as for CFVR and MFR
LAD. Possible explanations are that microvascular dysfunction is a process that affects the global myocardium and the reproducibility of regional estimates tend to be poorer than for global estimates in the majority of studies [
30,
44‐
46,
48].
In concordance with our findings for agreement between CFVR and MFR, reproducibility of MFR is in general not improved by correction for RPP, nor is it the tradition to correct CFVR for RPP [
30,
44‐
46,
48].
Agreement between any two methods depends on and is limited by the repeatability or reproducibility of both methods [
31]. We did not evaluate the reproducibility of MFR but it has been assessed previously by others. Reproducibility seems to be similar for
15O-water,
13N-ammonia or
82Rb, and without any obvious difference between studies of delayed or immediately repeated exams [
30,
44‐
46,
48,
49]. Sdringola et al. evaluated the reproducibility of MFR using
82Rb in 107 healthy subjects with a median delay of 22 days between repeated exams. For global estimates (regional being of comparable magnitude) they found reproducibility corresponding to LOA +/−38 % and +/−51 % of mean MFR, equivalent to absolute differences of 1.6 and 2.0 in what they termed “true” and “not-true normals”, respectively [
44]. Manabe et al. applied
82Rb in repeated exams separated by an hour and found reproducibility corresponding to LOA of +/−1.6 (37 %) in15 healthy participants [
46].
In this context, our overall findings of agreement between CFVR and MFRLAD and MFRglobal must be considered to be acceptable.
The acceptable agreement between CFVR and MFR can have practical implications in the clinical setting as CFVR as a non-invasive, non-ionising method with prompt availability and lower cost would be preferable.
Our data of both CFVR reproducibility and method agreement indicate that variability over time in CFR should be taken into account e.g. when planning studies with purpose of testing the effect of interventions on CFR. That CFR and microvascular function is a dynamic parameter is also illustrated by previous findings of improvement in CFR early (days) after stenting both in patients with and without MI, with further improvement in some but not all patients after 3–6 months [
50‐
52], which can relate to myocardial recovery [
52], and in-stent restenosis [
9,
51]. Likewise, improvement in CFR has been illustrated from 1 to 6 months after CABG [
53]. The post interventional improvement could also be influenced by rehabilitation efforts such as exercise and weight-loss [
8]. In attempt to reduce the potential effect of recovery after MI, PCI, in-stent restenosis and CABG we included only participants in stable phase at least six months after PCI, MI or CABG.
Limitations
We instructed participants and they agreed to abstain from caffeine before examinations, but we did not asses the validity of this by means of laboratory testing of caffeine levels as it would normally not be done in a clinical setting. If any participant should not have complied with the instructions on caffeine, it must be considered a random effect and as so it would tend to decrease the estimated reproducibility and agreement. It is a limitation of the method agreement substudy that only patients with revascularized stenosis were included, however the focus of the present study was on CFR as a measure of microvascular function and not on estimation of stenosis severity. We did not repeat assessment of coronary anatomy immediately before trial start. However, exercise ecg and stress echocardiography were performed at inclusion and these were without limiting angina or signs of ischemia in all patients included in the study. This was also illustrated by the fact that CFVR of LAD had as good agreement with MFRglobal as with MFRLAD.
CFVR was measured as distal as possible, in order to measure distal to grafts to the LAD. Accordingly, we consider CFVR a measure of microvascular function also in the participants with CABG.
Patients with previous CABG often have complicated coronary anatomy and thus, it can be difficult to define “matched” myocardial region perfused by the LAD vs. competitive flow from grafts, LCX or RCA for comparisons with CFVR [
54]. We did not take into account the exact coronary anatomy from CAG of the individual participant, and the regional MFR was estimated by our clinical routine practice from the standard region applied by the software. However, our agreement obtained in non-MI participants within a week is not far from agreement between repeated readings (the same examination) of MFR performed with different software-packages (LOA +/−0.3,
n = 90 patients) or between a novice and an expert reader using identical software (LOA +/−0.3,
n = 30) [
55,
56].
The sample size was limited and it was not the purpose of the study to estimate differences in agreement between subgroups (ie MI vs no-MI, and PCI vs CABG).