First validation of myocardial flow reserve assessed by dynamic ^99mTc-sestamibi CZT-SPECT camera: head to head comparison with ¹⁵O-water PET and fractional flow reserve in patients with suspected coronary artery disease. The WATERDAY study

From November 2014 to November 2016, 109 consecutive, stable, symptomatic patients with positive exercise stress test (downsloping ST segment depression <1 mm) or perfusion imaging (reversible perfusion defect in at least 2 out of 17 myocardial segments) were referred for diagnostic invasive coronary angiography (ICA) and screened for possible eligibility in the WATERDAY trial. Information regarding patient age, gender, weight, medical history, risk factors (family history of CAD, hypertension, diabetes, smoking or hypercholesterolemia), medication, blood pressure, heart rate and symptoms were recorded. Among these patients, 45 were enrolled in the WATERDAY study (clinicaltrials.​gov unique identifier NCT02278497) and underwent FFR measurement.

Inclusion criteria were: patients between 18 and 80 years old, with angiography-proven CAD, affiliated to a social security system, who have been informed of the study and gave informed consent. Exclusion criteria were: absence of ≥50% stenosis of at least one artery at ICA, absence of FFR measurements on the 3 main epicardial arteries, delay greater than 30 days between both PET and SPECT imaging and the ICA, patient with a recent history of acute coronary syndrome, patient with extra-cardiac disease whose prognosis can interfere with the treatment decision, pregnant or lactating women, allergy to angiographic contrast media, patient with renal failure with a modification of diet in renal disease (MDRD) clearance less than 60 ml/min, patient under guardianship, or unable to understand the purpose of the study.

Finally, only 30 out of 45 patients enrolled underwent both dynamic SPECT and PET within 30 days following ICA (Fig. 1). To avoid potential changes in MBF and MFR measurements, no revascularization procedure or medical therapy optimization was performed in any of the 30 analyzed patients between PET and SPECT imaging (Table 1). ICA/FFR identified 51 significant coronary artery lesions (Table 2).

The study protocol was approved by the Regional Ethics Committee (CPP Nord-Ouest III, France), written informed consent was obtained from all the patients and the procedures were in accordance with the Declaration of Helsinki.

Maximal hyperemia was obtained by intravenous (i.v.) infusion of adenosine [200 μg/kg/min for the left anterior descending artery (LAD) and left circumflex artery (LCx), and 100 μg/kg/min for the right coronary artery (RCA) [12]. FFR was measured using a sensor-tipped 0.014-in guide wire (PressureWire; St. Jude Medical, St. Paul, MN, USA) in all coronary arteries (LAD, LCx, and RCA) with a diameter of ≥2 mm, and for every stenosis ≥50% in a vessel segment ≥2 mm in diameter. FFR was calculated as the ratio of mean distal pressure divided by the mean proximal pressure. A pullback recording was performed and recorded for each measurement.

The D-SPECT system is comprised of nine individual detector columns, which rotate independently and are able to focus on a fixed region of interest (ROI). Throughout the dynamic list mode acquisition, the detectors perform a continuous step and shoot a scanning pattern consisting of multiple sweeps forward and backwards. Each sweep includes 10 positions over approximately 3 s. Therefore, a frame of 30 s would contain information from 10 sweeps while a frame of 3 s would contain information from only a single sweep. Rest and stress dynamic images are acquired in list mode over 6 min.

For rest imaging, an initial dose of approximately 37 MBq of ^99m Tc-sestamibi was used to position the patient’s heart within the field of view [13]. Three MBq/kg was then injected at a rate of 1–2 cm³/s using an automatic injector (Nemoto, Tokyo, Japan) and flushed by 30 mL of saline to ensure consistent delivery of a tight bolus. Regadenoson (400 μg) was then administered, followed by the injection of 9 MBq/kg of ^99mTc-sestamibi at peak hyperemia. Rest–stress dynamic acquisitions were completed within 75 min. Data were rebinned into 32 frames consisting of 21 × 3-sec, 1 × 9-sec, 1 × 15-sec, 1 × 21-sec, 1 × 27-sec and 7 × 30-sec frames. An ordered subset expectation maximization (OSEM) algorithm was used for image reconstruction with 4 iterations and 32 subsets. The procedure for dynamic CZT-SPECT imaging is summarized in Fig. 2.

Continuous variables are expressed as mean ± SD and were compared using one-way ANOVA. Reproducibility for paired studies was examined by the Pearson correlation coefficient and Bland-Altman plot analysis. Abnormal global or regional stress MBF and MFR were defined using PET results as stress MBF < 2.5 ml/min/g and MFR < 2, respectively [19, 20]. The diagnostic value of CZT-SPECT for stress MBF impairment and abnormal MFR was assessed using receiver-operating characteristic (ROC) analysis. Concordance between categorical variables was assessed using a Kappa test. All tests were two-tailed and a P value ≤0.05 was considered statistically significant.

Figure 1 presents the flow chart of the study, and patients’ characteristics are summarized in Table 1. Regadenoson infusion was well-tolerated in all cases and hemodynamically [heart rate (HR) and blood pressure (BP)] responses did not change significantly during both modalities at rest (CZT-HR = 65.3 ± 11 vs. PET-HR: 65.71 ± 10 bpm; CZT-systolic BP = 118 ± 18 vs. PET = 120 ± 21 mmHg; diastolic CZT-BP = 66.5 ± 9 vs. PET = 63 ± 8 mmHg) and at stress (CZT-HR = 85.45 ± 13 vs. PET-HR = 84.6 ± 15 bpm; CZT systolic BP = 100 ± 20 vs. PET =102 ± 17 mmHg; diastolic CZT-BP = 65 ± 8 vs. 64.6 ± 6 mmHg; P = NS for all data). However, regadenoson induced significant increase of HR and decrease of SBP between rest and stress in both modalities (P < 0.05).

Table 3 shows all myocardial blood flow and reserve quantification for the whole LV and for vessel territories assessed by CZT-SPECT and PET in 30 patients. As expected, pharmacological stress significantly increased MBF in both modalities. CZT-SPECT yielded higher stress and rest MBF compared to PET for global, and LAD and LCx, territories but not in RCA territory. Nevertheless, MFR was similar in global, and each vessel territory, for both modalities. No significant difference in MFR quantification was observed in each vessel territory between CZT or PET.

As demonstrated in Fig. 3a and b, there was a good correlation between CZT-SPECT and PET for global MBF, with a mean difference between the two techniques of 0.33 (–1.06 to 1.71). On Bland-Altman plots, the difference between CZT-SPECT and PET varied slightly over a range of MBF. ROC curve analysis found a 2.5 (ml/min/g) cutoff for the diagnosis of abnormal stress MBF using CZT-SPECT with a sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive value (NPV) of 58.3, 94.4, 80, 97.5 and 77.3%, respectively. The area under the curve (AUC) was 0.77 (Fig. 3c).

There was also a significant correlation between CZT-SPECT and PET for global MFR, with a mean difference between the two techniques of 0.19 (–0.89 to 1.28; see Fig. 4a and b). In addition, the mean MFR did not vary between CZT-SPECT and PET. As depicted in Fig. 4c, the ROC curve found a cutoff of 2.1 for detection of abnormal MFR using CZT-SPECT, with a sensitivity, specificity, accuracy, PPV and NPV of 83.3, 100, 96.7, 100 and 96% respectively. The AUC was 0.96.

Despite the great number of vessels with significant stenosis, FFR was abnormal in only 12 coronary arteries (13% of the total number of coronary territories assessed). This low prevalence is likely to influence the predictive values of quantitative perfusion CZT-SPECT. However, there was a high concordance between MFR and FFR for both PET (kappa = 0.85) and CZT-SPECT (kappa = 0.81), as illustrated in Table 4. Discrepancies between FFR and CZT-SPECT and PET were observed in 17/90 and 13/90 vascular territories, respectively. The five territories with low FFR (≤0.8) and normal MFR (≥2.1) by CZT-SPECT corresponded to focal coronary lesions in five different patients. Most of these patients had an FFR value of 0.79 to 0.8, except one patient had an FFR of 0.66 on distal LAD. All of the remaining 12 patients with normal FFR (>0.8) and abnormal MFR (<2.1) by CZT-SPECT had more than three cardiovascular risk factors and five patients had diabetes. Nevertheless, MFR evaluated by CZT-SPECT and PET had similar diagnostic values for the detection of abnormal FFR (see Table 4).

Figure 5a shows a patient with normal FFR and normal global and regional MBF and MFR in CZT-SPECT and PET in all three vessels.

Figure 5b showed a patient with normal FFR but abnormal global MBF and MFR in CZT-SPECT and PET in all three vessels.

The mean dose of ^99mTc-sestamibi was 294 MBq (169–443) at rest and 780 MBq (482–1229) at stress, leading to a total effective dose of 8.73 mSv (5.8–13.7) for dynamic SPECT [21]. The effective dose from the complete PET study (rest-stress H₂¹⁵O and CT attenuation scan) was 3.1 mSv (2.6–4.2) [22, 23].

Recently, Nkoulou et al. [3] reported lower MFR values with ^99mTc-tetrofosmin than typically seen with PET and ¹³N–ammonia, due to the lower MBF values observed by SPECT at high-flow conditions, while resting MBF values were highly comparable with PET. The underestimation of MBF with ^99mTc-tetrofosmin compared to ¹³N–ammonia is not only evidenced by a higher extraction fraction at resting conditions but also based on the fact that the given extraction fraction remains constant over a much larger range of MBF. Conversely, in the WATERDAY study, we observed increased global stress MBF values compared to ¹⁵O–water PET after correction of the ^99mTc-sestamibi extraction fraction [15], while resting MBF values were highly comparable with ¹⁵O–water PET. The difference between the two techniques increased for greater flow values, likely due to the absence of AC in CZT-SPECT. The plot shows that the agreement between SPECT and PET is rather good when MBF is <2 ml/min/g (resting conditions), and decreases for greater flow values (vasodilation). One of the reasons can be the difficulty to distinguish the myocardial edges because of photon attenuation, leading to a possible myocardial ROI positioning error inducing some fluctuations in MBF calculation. In addition, the absence of AC results in increasing the height of the input function curve, causing a distortion of the input function and hence a significant overestimation in the calculated MBF values [23].

Regional analysis demonstrated an increased stress and rest MBF in LAD and LCx using CZT-SPECT compared with PET. CZT-SPECT provided global and regional MFR similar to PET and yielded high diagnostic values for the detection of impaired PET flow reserve with high PPVs and NPVs (respectively, 100% and 96%) in our population.

Despite the limited sample size, the strength of these results lies in the fact that each patient served as his own control and the cut-off for abnormal perfusion reserve was a widely accepted PET MFR threshold (<2) [33‐35]. ¹⁵O–water PET has high reproducibility and agreement between software packages [7], but shares several limitations with ^99mTc-sestamibi CZT-SPECT, including potential problems with bolus delivery, patient motion and suboptimal hyperaemia. The agreement between software packages for dynamic ^99mTc-sestamibi tracer using CZT cameras should be assessed in further studies. MFR reflects CAD beyond the epicardial section of the coronary tree involving microcirculation and endothelial dysfunction. PET studies showed an increased sensitivity in detecting CAD using global MFR as an adjunct to perfusion analysis [35]. Therefore, it is of paramount importance to develop dynamic SPECT imaging that could benefit a wider cardiac patient population. Our results demonstrated high agreement of dynamic CZT-SPECT with PET for the identification of abnormal global and regional MFR. However, per-segment (17 segment model) MFR analysis was not performed as segmental counts statistics were not sufficient enough to draw high-quality TACs compatible with reliable measurements.

The absence of AC may have an impact on flow quantification with the CZT camera. To quantify MBF, we have adjusted the input function ROI inside the LA in order to avoid crosstalk between myocardial counts and bolus ROI that is mainly due to motion (respiration, heartbeat, etc.). However, the LA is often deeper inside the body and will be more affected by the attenuation (~25% attenuation for 2 cm at 140 keV).

FFR was initially introduced as an invasive surrogate marker of relative severity of myocardial ischemia, and validated by comparison with PET-derived relative coronary flow reserve [36]. PCI has shown improved prognosis in patients with abnormal FFR, using a cut-off value between 0.75 [37] and 0.80 [10, 11]. Between these two limits, there is a substantial gray zone with a certainty of 99% that a given lesion is responsible for ischemia. Ironically, FFR was recently proposed as a reference to validate novel non-invasive flow-based functional assessment by PET [2, 20]. A key finding of our study is that a 2.1 cut-off value for CZT-SPECT MFR predicted an abnormal FFR (<0.80) with an accuracy of 81.1% and an NPV of 93%, comparable to ¹⁵O–water PET.

In our study, some of the false positives (12 patients) were likely due to the presence of microvascular dysfunction in patients with diabetes or with more than 3 cardio-vascular risks, highlighted as “slow flow” during ICA with a normal FFR. FFR is primarily focused on identifying a pressure gradient across a stenosis, but cannot directly assess microvascular resistance or flow. However, high specificity (84.6% vs. FFR) was an important finding using CZT-SPECT, thus supporting the use of this new technique to guide referrals for ICA or as an optional “add-on” investigation after ICA/FFR when the functional severity of stenosis is uncertain or when FFR is contraindicated (suboccluded artery, asthma). Some of the false negatives (five patients) were due to focal or unique lesions on the main coronary arteries identified by ICA, with a FFR less than 0.8. However, in most of these cases (4/5 patients) the FFR was within the gray zone, between 0.79 and 0.8 [19, 20]. In contrast, global and even regional MBF and MFR were in the normal range.

Due to a poor distinction between epicardial stenosis and microvascular disease, it is quite challenging to directly correlate MBF with a specific epicardial lesion. Further, the widely standardized segment model [38] is frequently inaccurate in the prediction of a diseased coronary branch; an area at risk can be mixed with adjacent normal areas or truncated by the crude segmentation.

For LAD and LCx vessels, when FFR was less than 0.8, MFR (< 2) by PET had better accuracy, specificity, PPV and NPV than MFR (2.1) by CZT. However, for RCA when FFR was less than 0.8, the accuracy was 86.7%, the sensitivity was 66.7%, the specificity was 88.9%, the PPV was 40% and the NPV was 96% for both MFR modalities. The concordance between FFR and MFR-PET was slightly better than MFR-CZT in all 90 territories with a kappa coefficient as 0.85 and 0.81, respectively. The difference in the regional MBF and MFR values could be attributed to the use of attenuation correction (AC) in PET imaging.

It would be informative to perform myocardial perfusion imaging (MPI) to obtain semiquantitative assessment of perfusion analysis (expressed as total perfusion deficit or summed scores), to show potential additional diagnostic value of flow over uptake, or of a combination of both; but such conventional acquisitions have not been finally done in this trial. Although we demonstrated the capability of CZT-SPECT MBF quantification to be close to PET quantification, the number of investigational subjects and, in particular, those with severe stenosis was still small. Future studies should enroll a larger number of subjects, and ideally utilize both attenuation and scatter correction, neither of which was performed in this study. Variability due to methodology arises from the myocardial radioactive tracer distribution (^99mTc-sestamibi vs. ¹⁵O–water), reconstruction algorithms, and flow model (^99mTc-sestamibi vs. ¹⁵O–water). However, the flow model for ^99mTc-sestamibi used in this study has been previously validated in animals [15].