The impact of revascularization on myocardial blood flow as assessed by positron emission tomography

Patients were instructed to fast for 4 h and to abstain from caffeine and theophylline for 24 h prior to PET imaging. PET imaging was performed using list-mode 2D acquisition on an Attrius PET scanner (Positron, Westmont, IL) with ⁶⁸Ge rod source attenuation correction. Rest and stress emission data were obtained over 7 min beginning immediately upon intravenous injection of 1,295–1,850 MBq of generator-produced ⁸²Rb (Bracco Diagnostics, Monroe Township, NJ) at 50 ml/min. Hyperemia was achieved by intravenous infusion of dipyridamole (142 µg/kg/min). Attenuation correction, registration, image reconstruction of topographic views, arterial input selection and computation of MBF were performed as previously reported [12, 13]. For each myocardial quadrant, rMBF, sMBF, CFR and CFC were calculated using FDA-approved software (HeartSee™; University of Texas, Houston, TX) [2].

Relative PAs can exist in the presence or absence of severe reductions in CFC, and vice versa. Therefore, each left ventricular (LV) myocardial quadrant was categorized in terms of the presence or absence of a significant PA, and separately in terms of the presence or absence of a significant severe reduction in CFC.

The size, location and severity of the PA were automatically quantified as the percentage of the topographic maps with relative activity ≤60% of maximal activity. A significant PA was defined as ≥10% change in size and/or severity from the resting scan in a contiguous area of LV myocardium with activity ≤60% of maximum. These regions appeared as green and/or blue defects on relative perfusion images, as shown in Fig. 3. This threshold for declaring significance (size and relative change) identified visually obvious defects and excluded small deviations/variations in radiotracer uptake or reconstruction and stress-induced LV conformational changes [2, 12]. Defects smaller than 10% of the LV myocardium, although possibly not normal, are visually small and could be caused by any of the aforementioned confounders.

A severe reduction in CFC was defined as sMBF ≤0.91 cm³/min/g and CFR ≤1.74, on a pixel-by-pixel basis. These thresholds were determined based on clinical features of ischemia [2, 8]. We considered a reduction in CFC significant when a contiguous region spanning ≥10% of the LV myocardium met these conditions. Regions with severe reductions in CFC were color-coded blue on CFC maps, as shown in Fig. 3. These definitions led to four distinct classifications of myocardial “quadrant normalcy”, as shown in Fig. 3:

Our methodology for assessing sMBF and CFC responses to revascularization has been described previously [12]. Each epicardial vessel was assigned to a myocardial quadrant (or quadrants) based on its perfusion pattern and review of angiographic data. Appropriate designations of territories were made for left and codominant circulations as needed. Following revascularization, angiographic and/or surgical data were reviewed to determine which quadrants received revascularization. Quadrants were subsequently classified based on their “quadrant normalcy” on the baseline PET scan and whether the quadrant received revascularization. This method led to eight permutations of possible postrevascularization “quadrant types”, as shown in Fig. 3.

First, we performed a patient-level analysis to determine the overall impact of revascularization on the study population. We subsequently identified three patient categories based on whether revascularization was congruent with the PET findings. “Concordant” patients were defined as those who had revascularization only to territories with a significant PA and/or reduced CFC, “discordant” patients as those who had revascularization only to territories without a PA or reduced CFC, and “mixed-concordance” patients as those who had revascularization to both abnormal and normal regions.

Next, we performed a regional analysis based on the postrevascularization quadrant types. Using this methodology, we intended to determine whether the presence of a PA, or decreased CFC, or both, portended a favorable regional response to revascularization. A representative case example is shown in Supplementary Fig. 1.

SPSS version 20.0 (IBM, Armonk, NY) was used for statistical analysis. The results are expressed as means ± standard deviations or as medians and interquartile ranges, as appropriate. Categorical variables were compared using Fisher’s exact test. The normality of the distributions of continuous variables was assessed using histograms, Q-Q plots, and the Shapiro-Wilk test. Because the distributions of the tested continuous variables were largely nonnormal, intrapatient changes in MBF and CFC were assessed using the Wilcoxon signed-ranks test, and quadrant-type changes in MBF and CFC were assessed using the Mann-Whitney U test. For all analyses, two-sided p values <0.05 were considered statistically significant.

The results of the patient-level analysis are summarized in Table 3 and Fig. 4.

Median sMBF significantly increased from 1.26 to 1.35 cm³/min/g (p = 0.002). Median CFR increased from 1.38 to 1.60 (p = 0.005). The median size of PAs decreased from 14.0% to 2.0% (p < 0.001). Median percentage of the myocardium with severely reduced CFC decreased from 33% to 12% (p < 0.001). Revascularization was concordant in 26 patients. In these patients, sMBF significantly increased from 1.11 to 1.28 cm³/min/g (p < 0.001), and CFR increased from 1.34 to 1.58 (p < 0.001). Revascularization was disconcordant in 12 patients. In these patients, neither sMBF nor CFR changed significantly: 1.43 vs. 1.37 cm³/min/g (p = 0.37), 1.95 vs. 1.66 cm³/min/g (p = 0.82), respectively. Revascularization was mixed-concordant in 12 patients. In these patients, sMBF increased from 1.26 to 1.60 cm³/min/g (p = 0.01).

Among the various concordance groups, there were no significant differences between risk factors and a history of prior revascularization (Supplementary Table 2). Furthermore, risk factors, medication usage at the time of PET1, and PET findings were analyzed in univariable and multivariable analyses (Supplementary Table 3). Among risk factors and medications, none independently predicted improvement in sMBF. A severe baseline PA was a predictor of improvement in sMBF in the multivariable analysis; however, when a severe reduction in CFC was added to the model, a severe baseline PA did not remain a predictor. A severe reduction in CFC was the only independent predictor of an increase in whole-heart sMBF (β = 25.338, p = 0.032). Furthermore, baseline sMBF, baseline CFR, and baseline percentage of LV with severely reduced CFC were similarly analyzed (Supplementary Tables 4 and 5). In these analyses, a severe reduction in CFC remained the best and sole predictor of (1) any improvement in sMBF and (2) improvement greater than the expected day-to-day variability (>20%) in sMBF (β = 0.948, p = 0.004 and β = 0.015, p = 0.004, respectively) [7].

There were 200 regions analyzed (four in each patient: anterior/septal/lateral/inferior). The frequencies of quadrant types are shown in Fig. 5 and Supplementary Fig. 2. The relevant changes in flow, CFR and CFC are discussed below and are summarized in Table 4 and Fig. 6.

Following revascularization rMBF in SevereAbn/+Revasc quadrants increased by 11.1% (n = 37; +0.05 cm³/min/g, p = 0.001) and in CFCAbn/+Revasc quadrants increased by 18.2% (n = 26; +0.12 cm³/min/g, p = 0.01). In quadrants that did not receive revascularization (n = 89), regardless of relative perfusion or flow capacity abnormality, there was no significant change in rMBF (−0.07 cm³/min/g, p = 0.30).

Following revascularization sMBF in SevereAbn/+Revasc quadrants increased by 59.3% (n = 37; +0.51 cm³/min/g, p = 0.001) and in CFCAbn/+Revasc quadrants increased by 40.0% (n = 26; +0.35 cm³/min/g, p = 0.01). In contrast, there was no significant change in sMBF in Norm/+Revasc quadrants (n = 46; +0.07 cm³/min/g, p = 0.56) nor in Norm/−Revasc quadrants (n = 70; −0.03 cm³/min/g, p = 0.77). Furthermore, there was no significant difference in the magnitude of the changes in sMBF between these two groups (p = 0.63). In quadrants that did not receive revascularization (n = 89), regardless of perfusion or CFC abnormality, there was no change in sMBF (median 0.0 cm³/min/g, p = 0.76).

We made several comparisons between quadrant types. There was a greater relative change in sMBF in SevereAbn/+Revasc quadrants (n = 37) than in Norm/+Revasc quadrants (n = 46; +59% vs. +5%, p < 0.001), indicating that quadrants with a stress-induced PA and reduced CFC showed dramatically better improvement following revascularization than quadrants with neither abnormality. CFCAbn/+Revasc quadrants (n = 26) showed a greater change in sMBF than Norm/+Revasc quadrants (n = 46; +40% vs. +5%, p = 0.002), indicating that improved sMBF could be predicted solely from an abnormal baseline CFC. Finally, there was a significant difference in the percentage improvement in sMBF between SevereAbn/+Revasc quadrants (n = 37) and CFCAbn/+Revasc quadrants (n = 26; +59% vs. +40%, p = 0.03), which demonstrates that the presence of both a PA and a low CFC predicts greater improvement in sMBF when compared with a low CFC alone.

There were significant decreases in the percentages of the myocardium with severely reduced CFC in the SevereAbn/+Revasc and CFCAbn/+Revasc quadrants: −12% (p < 0.001) and −9.5% (p < 0.001), respectively. Quadrants of all other types, specifically Norm/+Revasc quadrants, did not demonstrate significant changes.

Appropriately targeted revascularization consistently improved MBF and CFC. Following revascularization, in quadrants with a significant PA and severely reduced CFC rMBF increased by 11%, sMBF increased by 59%, and the percentage of myocardium with severely reduced CFC decreased by 12%. Furthermore, in regions with a large burden of severely reduced CFC but with minimal if any PA, rMBF increased by 18%, sMBF increased by 40%, and the extent of severely reduced CFC decreased by 9.5%. Furthermore, the regional analysis confirmed that improvement in quantitative perfusion metrics was due to revascularization and was not a result of lifestyle changes or medical therapy. Considering recent findings indicating that CFC is directly correlated with mortality and the risk of MI [3], a revascularization strategy focused primarily on improving CFC has the potential to reduce the risk of death and MI.

Based on their angiographic appearance, 72 quadrants without a PA at baseline underwent revascularization. Of these, 46 (64%) had preserved baseline CFC (Norm/+Revasc), and 26 (36%) had significantly reduced baseline CFC (CFCAbn/+Revasc). The differential improvement in sMBF following revascularization between these two quadrant types further demonstrates the value of CFC. While quadrants of these two types had normal relative perfusion images, they contained vessels with visually high-grade disease on angiography, leading to intervention. However, despite visually obstructive CAD, there was no improvement in MBF following revascularization unless CFC was reduced. The differential response in sMBF following revascularization was striking: +40% vs. + 5% (p = 0.002) in vessels with and without severely reduced CFC, respectively. Our data highlight physiologic discordances between the widely used functional measurements, relative perfusion imaging (without MBF quantification), angiographic percentage stenosis and CFC, in addition to the impact of revascularization (or lack thereof) on changes in MBF. Furthermore, these discordances were not uncommon: in the current study: about one third of all quadrants were of these two types, and these quadrant types accounted for about two thirds of all interventions. In fact, Norm/+Revasc quadrants were found in 24 of 50 patients (48%) and accounted for 41% of quadrants that received revascularization. Twelve patients had intervention only to quadrants with normal relative perfusion and normal CFC. Clearly, patients who underwent revascularization in Norm/+Revasc quadrants were exposed to procedural risk and potential long-term complications without improvement in either sMBF or CFC.

On the surface, a severe reduction in CFC without a corresponding PA seems incongruent. The physiologic explanation for this finding is also the basis for the well-documented and well-described discordance between CFR, sMBF and FFR – that is, diffuse epicardial disease [11, 15]. The appearance of reversibility on radionuclide imaging requires a wide variability in counts throughout the myocardium. The region with the lowest count, when displayed with surrounding regions with high counts, results in what is termed “reversibility.” In the setting of a discrete epicardial stenosis, an abrupt stepdown in radionuclide count density leads to an obvious visual and measurable relative stress defect. However, in the setting of diffuse CAD, there are smaller differentials between areas with higher counts and areas with lower counts. Therefore, reversibility and perfusion defects become less apparent despite concomitant reductions in sMBF and CFC. Furthermore, quantification of diffuse disease on angiography is challenging, subjective and operator-dependent if not measured by intravascular ultrasonography or FFR pullback.

Diffuse disease was seen angiographically in the vast majority of patients, but was not quantified by any particular metric. However, Gould et al. have reported ranges of flow capacity when coronary disease is present and flow remains above ischemic thresholds. These regions appear yellow and green on flow capacity maps and are indicative of nonobstructive diffuse CAD. In our patient population, 100% of the patients had yellow regions on flow capacity maps. The median percentage of the LV myocardium that was yellow on the flow capacity maps was 36.5% (interquartile range 20.3–46.5%). In other words, all patients had “diffuse disease” and the majority of patients had diffuse disease in vessels perfusing about 20–45% of their LV myocardium. These concepts are illustrated in Fig. 7. This observation could explain why, mechanistically, regions with the combination of a PA and a severe reduction in CFC had a better response to revascularization than regions with just a severe reduction in CFC.

FFR-guided revascularization was performed in nine lesions using an FFR threshold of <0.80. Each of these vessels supplied territories without PA and three vessels perfused territories with severe reductions in CFC. All quadrants with normal relative perfusion images and reduced CFC demonstrated improvement in sMBF following FFR-guided revascularization. In contrast, only one of the six quadrants with normal relative perfusion images and preserved CFC demonstrated improvement in sMBF following FFR-guided revascularization. Although the sample size of patients with available FFR data in the current study was small, our results showing discordance between FFR and CFC are consistent with findings reported by others and offer an insight into the impact of revascularization based on these criteria [11, 14].

In a recent study, Driessen et al. [16] enrolled 53 patients without known CAD who underwent serial ¹⁵O-H₂O PET before and after FFR-guided revascularization. FFR was routinely measured before and after PCI. sMBF improved by 58% (1.57 ± 0.59 to 2.48 ± 0.91 cm³/min/g) after revascularization. In concordance with our data, the percentage increase in sMBF was similar. However, the absolute values of sMBF in the study by Driessen et al. were markedly higher. The reasons for this discrepancy were differences in the patient populations and also possibly in the tracers used. The patient population in the study by Driessen et al. was significantly “healthier” than our population. Exclusion criteria in their study included known CAD, whereas 80% of our population had known CAD, including 26% with prior CABG. Their population was also younger with fewer comorbidities. In addition, Driessen et al. found a baseline sMBF in normal territories of 2.45 ± 0.73 cm³/min/g, confirming a relatively healthy patient population with minimal diffuse atherosclerotic disease [2, 9]. Furthermore, plots of the relationships between FFR, sMBF and CFR confirm that the vast majority of lesions with FFR <0.80 had sMBF >0.91 cm³/min/g and CFR >1.74 [16]. In other words, CFC was not severe despite a reduction in FFR, a common finding also noted by van de Hoef et al. [11]. Hence, the majority of lesions that received revascularization in the study by Driessen et al. would be of the PerfAbn/+Revasc quadrant type. As noted, in the current study, there were only two quadrants of this type, which is consistent with our population having more advanced CAD.

Although this was a single-center study, we employed validated methods and present the results in a way that provides a generalizable methodology for comparison of quantitative PET data. While quantification of absolute sMBF is commercially available, CFC quantification is currently limited to a few centers. Percentage diameter stenosis was visually estimated by the treating cardiologist, which may have introduced interoperator variability. Symptomatic improvement was not assessed in a blinded and unbiased manner. Due to the study design, objective assessment of symptoms before revascularization was not obtained as enrollment occurred after revascularization. Therefore, conclusions regarding symptomatic improvement with improved MBF cannot be drawn. Furthermore, improvement in global and/or regional LV function could have added more relevance to the findings. However, the sample size of 50 patients in addition to the short duration of follow-up limited such an analysis. Finally, because this study was designed to assess the short-term impact of revascularization, long-term outcome data are not available.

Revascularization targeted to regions with severely reduced CFC on baseline PET yields improvement in quantitative perfusion metrics particularly if a significant relative PA is also present. Regions without reduced CFC demonstrated no improvement in quantitative perfusion metrics after revascularization.