Methods
Cohort 1 used the IDEAL collaborators registry, which involves 301 patients undergoing elective coronary angiography for suspected symptomatic coronary artery disease in four European academic hospitals.1 The exclusion criteria for IDEAL were significant valvular disease, previous coronary artery bypass surgery, acute heart failure, acute myocardial infarction within 48 hours of the procedure and prior anterior wall myocardial infarction. Measurements were not acquired in vessels with angiographically identifiable myocardial bridging or collateral arteries. Cohort 2 represents a separate cohort of patients with chest pain and positive functional test scheduled for elective coronary angiography, enrolled in a previous study.13 The exclusion criteria for this study were known ischaemic heart disease, valvular pathology, evidence of regional wall motion abnormalities and renal impairment (creatinine >120 μmol/L). In cohort 2, both invasive and non-invasive Doppler measurements of coronary flow velocity were obtained. For both cohorts, only measurements in the LAD were used since the LAD is the only coronary artery in which echocardiographic DSVR can be reliably assessed.10 In cohort 1, 228 patients were included since measurements were taken in the LAD in 228 of 301 patients (76%) in the IDEAL registry.
Invasive flow velocity measurements
Coronary angiography was performed according to standard procedures. After angiography, a 0.014-inch guidewire equipped with both a distal pressure sensor and Doppler crystal (ComboWire XT, Philips Volcano, San Diego, USA) was inserted into the LAD. In the coronary ostium, the pressure sensor was equalised with the pressure of the aortic guiding catheter. Then, the wire was advanced beyond the stenosis, or beyond the proximal segment of the LAD if there was no stenosis. Doppler flow velocity, distal coronary pressure and aortic pressure were measured under true resting conditions. In cohort 1, 200–300 μg of intracoronary nitrates were administered prior to the resting measurements and the measurements were repeated during pharmacological hyperaemia, induced either by intracoronary injection of adenosine (60–150 μg) or intravenous adenosine administration (140 μg/kg/min). In cohort 2 vasodilator drugs were not administered.13 Pressure drift was assessed at the end of the procedure, and if pressure drift was identified (>2 mm Hg) measurements were repeated or corrected for during offline analysis.
Non-invasive echocardiography measurements
In cohort 2, transthoracic echocardiography was performed as described in an earlier study.13 In brief, a Philips ie33 (Amsterdam, The Netherlands) or Esaote MyLab Twice (Genova, Italy) device was used for echocardiography. Starting in the parasternal long axis view, the probe was rotated clockwise and moved laterally across the chest wall until the LAD was clearly in view with an angulation of <20° to the probe. Pulse-wave Doppler was applied with a sampling width of 7.5–10 mm to record coronary flow velocity signals. Data were exported as high-resolution images and digitalised for data analysis using a MATLAB algorithm (MathWorks, Natick, Massachusetts, USA) with smoothing by a Savitzky-Golay filter.
Data analysis
For cohort 1, quantitative coronary angiography (using either CAAS II, Pie Medical, Maastricht, The Netherlands; or McKesson, San Francisco, USA) was performed in angiographic stenoses to quantify diameter stenosis percentage, minimal and reference lumen diameter, area stenosis, minimal and reference lumen area, and stenosis length. For both cohorts, data were analysed using an automated MATLAB script (MathWorks) as previously described.1 14 Phasic analysis yielded average values of aortic pressure, distal coronary pressure and average peak Doppler flow velocity for the whole cardiac cycle, systole specifically and mid-to-late diastole specifically. Systole was identified starting at the R peak on the ECG and ending at the dicrotic notch on the aortic pressure trace. Mid-to-late diastole corresponded with the wave-free period, which starts at 25% of diastole, as marked by the aortic dicrotic notch, and ends 5 ms before systole.14 15 Both invasively and non-invasively measured DSVRs were calculated as the ratio between time-averaged mid-to-late diastolic and systolic peak Doppler flow velocity. Figure 1 provides two exemplary cases along with the angiogram and hyperaemic pressure tracings. Formulas used for the calculations of all parameters used in this study are shown in box 1. DSVR, stenosis resistance, microvascular resistance, total vascular resistance and diastolic-systolic resistance ratio were calculated during the resting state. FFR16 was calculated during the hyperaemic state. Stenosis resistance, microvascular resistance and total vascular resistance were examined during both diastole and systole specifically. FFR was used as reference standards to test the diagnostic accuracy of DSVR, using the FFR threshold of 0.75, which corresponds best with myocardial ischaemia.16 17 For cohort 1 both pressure and Doppler flow velocity data were analysed, thereby also yielding resistance data. For cohort 2, only Doppler flow velocity data were available, precluding calculation of resistance values.
Figure 1Examples of two patients included in the study. Shown are the Doppler flow velocity measurements during resting conditions (Panel A), the coronary angiogram (Panel B) and the pressure tracing during hyperaemic conditions after injection of 150 μg of adenosine (Panel C). DSVR, diastolic-systolic velocity ratio; FFR, fractional flow reserve; LAD, left anterior descending artery.
Statistical analysis
Categorical data are presented as numbers and percentages. Continuous data are presented as mean±SD or median with IQR according to normality of the data. To analyse the diagnostic performance of DSVR, receiver operating characteristic analysis was performed. The optimal DSVR cut-off value was defined as the value with the highest combined specificity and sensitivity. Analysis of variance was used to test for differences across multiple groups. Bonferroni correction was applied for comparison across multiple groups. Linear regression was performed to test the association between two continuous variables. Log transformation was applied to independent and dependent variables in order to achieve linearity of the relationship as required for linear regression analysis. Intraclass correlation coefficients were calculated using a two-way mixed model. A two-sided p value of <0.05 was considered statistically significant. Statistical analyses were performed using SPSS V.22.0.