Guiding early revascularization using [¹⁵O]H₂O positron emission tomography myocardial perfusion imaging: impact of atrial fibrillation

Abstract

Aims

Myocardial perfusion imaging (MPI) using [¹⁵O]H₂O positron emission tomography (PET) is used to guide the selection of patients with angina for invasive angiography and possible revascularization. Our study evaluated (i) whether atrial fibrillation (AF) reduces global hyperaemic myocardial blood flow (MBF) and (ii) whether [¹⁵O]H₂O PET MPI effectively guides revascularization procedures for patients with ongoing AF.

Methods and results

We prospectively recruited 346 patients with angina and persistent or paroxysmal AF referred for baseline/hyperaemic [¹⁵O]H₂O PET MPI. The primary outcome was revascularization within 3 months of MPI. In the analyses, patients were divided into four groups based on whether they had ongoing AF or sinus rhythm (SR) and whether they had previously documented coronary artery disease (CAD) or not. Thus, four groups were compared: SR-noCAD, AF-noCAD, SR-CAD, and AF-CAD. Hyperaemic MBF was affected by both ongoing AF and prior CAD [MBF (mL/min/g): 2.82 (SR-noCAD) vs. 2.12 (AF-noCAD) vs. 2.22 (SR-CAD) vs. 1.80 (AF-CAD), two-way analysis of variance P < 0.0001]. In multiple linear regression, ongoing AF was independently associated with reduced hyperaemic MBF. Every 0.1 mL/min/g decrease in hyperaemic MBF was associated with a 23% increase in odds of early revascularization. Receiver operating characteristic (ROC) analysis of vessel-specific hyperaemic MBF to predict early revascularization yielded the following areas under the ROC curve: SR-noCAD: 0.95 (P < 0.0001); AF-noCAD: 0.79 (P < 0.0001); SR-CAD: 0.78 (P < 0.0001); and AF-CAD: 0.88 (P < 0.0001).

Conclusion

Ongoing AF is associated with 19–25% reduced global hyperaemic MBF as measured by [¹⁵O]H₂O MPI PET. Regardless, vessel-specific hyperaemic MBF still predicts early revascularization in patients with AF.

Graphical Abstract

Open in new tabDownload slide

AF, atrial fibrillation; CABG, coronary artery bypass graft; PCI, percutaneous coronary intervention; PET, positron emission tomography; SR, sinus rhythm.

Introduction

Coronary artery disease (CAD) is characterized by atherosclerotic plaque formation in the coronary arteries¹ resulting in stress-induced ischaemia and angina. The diagnostic strategy for patients with angina depends on individual risk assessment and may include electrocardiogram, coronary computed tomography angiography, myocardial perfusion imaging (MPI), and invasive coronary angiography (CAG).^2,3 Whereas CAG allows for near-simultaneous diagnosis and treatment of flow-limiting stenoses, an obvious drawback of the technique is its invasive nature. Non-invasive MPI using positron emission tomography (PET) radiotracers rubidium-82, nitrogen-13 ammonia, or oxygen-15 water (¹⁵O-H₂O) has therefore been used for decades to determine whether patients should be referred for invasive CAG. For patients with no previously documented CAD and in sinus rhythm (SR), the diagnostic accuracy of vessel-specific myocardial blood flow (MBF) measured by ¹⁵O-H₂O PET MPI to detect fractional flow reserve (FFR)-significant coronary artery stenosis has been established at 88–92% using a receiver operating characteristic (ROC)-based cut-off of 2.3 mL/min/g.^4–6 However, a significant proportion of patients with angina have either previously been diagnosed with CAD or may suffer from persistent or paroxysmal atrial fibrillation (AF). We have recently shown that global MBF values are markedly lower in patients with prior coronary artery bypass graft (CABG) and that the cut-offs indicating subsequent revascularization are significantly lower than in patients with no previous history of CAD.⁷ Likewise, there are several studies that indicate that AF is associated with reduced global MBF.^8–11

AF is a common supraventricular tachyarrhythmia characterized by uncoordinated atrial activation and contraction, which leads to irregular and thus ineffective ventricular contraction.¹² The irregular cardiac rhythm is either persistent or paroxysmal in which case the cardiac rhythm alternates between AF and SR. Multiple studies have shown that the global hyperaemic MBF of patients in ongoing AF is reduced compared with both healthy control subjects or indeed other AF patients in SR.^8–11 However, successful radiofrequency ablation (RFA) therapy appears to be able to restore global MBF.⁸ These findings indicate that the pre-RFA low hyperaemic MBF could be secondary to the AF arrhythmia itself. Consequently, it is possible that the reduced myocardial perfusion associated with AF could cause a high rate of false positive referrals to CAG, partly due to angina caused by transient AF-related hypoperfusion, and partly due to a global MBF below the cut-off previously established to detect flow-limiting stenoses.

The OUTCOME-AARHUS cohort was designed to investigate the role of ¹⁵O-H₂O PET MPI as part of a standard clinical work-up for suspected obstructive CAD. The cohort consists of prospectively recruited patients with angina referred for ¹⁵O-H₂O PET MPI and broadly includes patients with various comorbidities such as previously documented CAD and AF.

The specific aim of the current sub-study was (i) to investigate whether ongoing AF is associated with reduced global hyperaemic MBF measured by ¹⁵O-H₂O PET MPI and (ii) to examine the prognostic value of vessel hyperaemic MBF measured by ¹⁵O-H₂O PET MPI to identify vessel-specific early revascularization for patients with known AF.

Methods

Subjects

All patients referred for ¹⁵O-H₂O PET at the Department of Nuclear Medicine & PET Centre, Aarhus University Hospital, Denmark, between July 2020 and July 2022, were invited prospectively to participate in the study. When consent to participate was given, patients were enrolled into the OUTCOME-AARHUS cohort. The protocol was approved by the Central Denmark Region Committees on Health Research Ethics in accordance with the Helsinki Declaration, case number 1-10-72-125-20. The storage of data was approved by the regional Data Protection Agency. The cohort is registered at ClinicalTrials.gov as NCT04451551.

We performed a sub-group analysis of patients with angina and previously documented persistent or paroxysmal AF from the OUTCOME-AARHUS cohort’s initial 1826 participants. Patients with documented atrial flutter were included as persistent or paroxysmal AF in the study. A total of 346 patients with AF were identified. A total of 237 had paroxysmal AF and 109 had persistent AF. A CONSORT diagram is presented in Figure 1.

To determine the independent effects of AF and previous CAD on hyperaemic MBF, included patients were divided into four sub-groups based on cardiac rhythm (AF or SR) at the time of ¹⁵O-H₂O PET MPI and previously documented CAD (CAD or no CAD). Thus, the following subgroups were analysed: SR-noCAD, AF-noCAD, SR-CAD, and AF-CAD. Patients were categorized as in ‘SR’ if AF or atrial flutter was not documented during the MPI procedure. Of note, few patients categorized as in SR had various pace rhythms, of which details are not readily available. Previously documented CAD was defined as previously documented ST-elevation myocardial infarction (STEMI), non-STEMI, chronic total occlusion, or previous invasive revascularization treatments [percutaneous coronary intervention (PCI) or CABG]. Early revascularization was defined as PCI or CABG within 3 months of MPI. Information about medications, co-morbidity, and previous cardiac procedures was collected from the Electronic Patient Journal system of the Central Denmark Region. The collection of clinical and follow-up data was not anonymized, nor was the data collector systematically blinded for MPI results. However, we do estimate the risk of biased data collection to be minimal.

Patients were managed according to local clinical guidelines. Consequently, the MPI was performed on clinical indications and the treating cardiologists were informed of the MPI results. The ¹⁵O-H₂O PET report contained information about (i) the absolute MBF globally and in the coronary artery territories, (ii) the global myocardial flow reserve, and (iii) the size (in %) of myocardium with a relative defect in MBF.

Consent

All participants completed consent forms before inclusion. No patients were included without informed consent.

¹⁵O-H₂O PET MPI procedure

Patients were instructed to abstain from caffeine, dipyridamole, and xanthine for 24, 48, and 72 h, respectively, before the procedure and they were also instructed to fast for 2 h before the scan. All patients were scanned during rest and pharmacologically induced hyperaemic conditions on a GE Discovery MI Digital Ready PET/CT (Aarhus, Denmark). Scans were 4-min dynamic acquisitions initiated at the same time as the infusion of an IV bolus of 400 MBq ¹⁵O-H₂O. Hyperaemia was induced by a 6-min continuous IV infusion of adenosine (0.14 mg/kg/min) initiated 2 min prior to the hyperaemic scan acquisition. For attenuation correction, a low-dose CT scan was performed. Images were reconstructed using the VuePointFX algorithm in isotropic voxels (3.27 × 3.27 × 3.27 mm³).

¹⁵O-H₂O PET image analysis

Semi-automatic analysis of all PET scans was performed with aQuant Research (an in-house software package developed at Aarhus University Hospital and Uppsala University).¹³ MBF was calculated separately for 17 myocardial segments of the left ventricle using a one-tissue compartment model correcting for luminal activity spill-in. Whole vessel MBF was used for vessel-specific analysis. Myocardial segments were categorized as either left anterior descending artery, right coronary artery, or circumflex artery based on expected coronary supply and in accordance with present guidelines.¹⁴ The relative flow reserve (RFR) was calculated as the ratio of hyperaemic MBF of the myocardial region with the lowest MBF (defect) to the region with highest MBF (reference) (see Supplementary data online): RFR = defect hyperaemic MBF/reference hyperaemic MBF.

Statistical analyses

Continuous variables are presented as means with standard deviations in parentheses. Categorical variables are presented as number of affected patients with percentages in parenthesis. The significance of the impact of previously documented CAD and ongoing AF on global hyperaemic MBF was calculated using two-way analysis of variance (ANOVA) and presented as P-values. The independent associations between global hyperaemic MBF and ongoing AF, previously documented CAD, and potential confounders were calculated by multiple linear regression and presented as coefficients (β) with confidence intervals (CIs). The probability of early revascularization was calculated by logistic regression analysis and presented as odds ratios (ORs) with CIs. The performance of vessel-specific hyperaemic MBF to identify subsequent early revascularization was estimated by ROC analyses and presented as ROC curves. CIs of optimal cut-off values were estimated by bootstrap with 1000 repetitions. P-values of <0.05 were considered significant. Statistical analyses were performed in the software package Stata/SE 17.0 (StataCorp. 2021. Stata Statistical Software: Release 17. College Station, TX, USA: StataCorp LLC).

Results

A total of 194 (56%) patients had documented CAD, while 152 (44%) had no previous history of CAD. Similarly, 137 (39%) AF patients had ongoing AF during the MPI procedure while 209 (61%) were in SR. This generated the following distribution of patients in the pre-specified groups: SR-noCAD (n = 78), AF-noCAD (n = 74), SR-CAD (n = 131), or AF-CAD (n = 63). Baseline characteristics of pre-specified groups are presented in Column A of Table 1.

As expected, heart failure was more prevalent in patients with known CAD. More patients with known CAD were treated with statins compared with patients with no known CAD, which is explained by a higher prevalence of hyperlipidaemia as well. Neither symptoms nor medical treatments were substantially different between the groups. Due to a few cases of pace rhythms and successful RFA treatments, 8/109 patients diagnosed with persistent AF were categorized as in SR during the MPI procedure. Of the 237 patients diagnosed with paroxysmal AF, only 36 had AF during the MPI procedure.

Global hyperaemic MBF (mL/min/g) was affected among those with ongoing AF regardless of whether patients had known CAD. Thus, in patients with no previously documented CAD, the presence of ongoing AF was associated with a markedly lower mean global MBF (2.82 vs. 2.12 mL/min/g, 25%) and similarly in patients with known CAD (2.22 vs. 1.80 mL/min/g, 19%)—see Table 2. Two-way ANOVA on previously documented CAD and ongoing AF was both P < 0.0001, while the test of interaction between documented CAD and ongoing AF was P = 0.17.

In multiple linear regression, female sex [β: 0.59 (0.38, 0.81)] was independently associated with increased global hyperaemic MBF, while ongoing AF [β: −0.41 (−0.69, −0.13)], heart failure [β: −0.21 (−0.41, −0.01)], and hyperlipidaemia [β: −0.28 (−0.49, −0.08)] were all factors independently associated with reduced global hyperaemic MBF. AF type (persistent AF) was not associated with global hyperaemic MBF [β: −0.03 (−0.31, 0.26)], and neither was heart rate nor any symptom during the scan. Results of multiple linear regression are presented in Table 3.

In total, during early follow-up (within 3 months of the MPI), 133/346 patients underwent CAG and 39 patients (51/1038 vessels) underwent revascularization. Follow-up CAG and revascularization numbers are presented in Supplementary data online, Table S2. Baseline characteristics of follow-up early revascularized vs. non-revascularized patients are presented in Column B of Table 1. In general, no substantial differences in age, body mass index, previous heart disease, or predisposing factors were observed, but patients who underwent early revascularization were more frequently male and had heart failure. More early revascularized patients were treated with beta-blockers and angiotensin-converting enzyme/angiotensin receptor blocker ACE/ARB compared with non-revascularized patients.

Per vessel-based logistic regression identified vessel-specific hyperaemic MBF [OR: 1.23 (1.15, 1.30) (95% CI) per 0.1 mL/min/g decrease] as the only factor independently associated with early revascularization. Logistic regression results are presented in Table 4. Per-patient logistic regression using the lowest vessel MBF yielded similar results and is presented in Supplementary data online, Table S3. Logistic regressions of all four subgroups are individually presented in Supplementary data online, Table S1.

The performance of vessel-specific hyperaemic MBF to identify subsequent early revascularization was evaluated by ROC analyses that yielded the following area under the ROC curve (AUC): SR-noCAD: 0.95 (P < 0.0001); AF-noCAD: 0.79 (P < 0.0001); SR-CAD: 0.78 (P < 0.0001); and AF-CAD: 0.88 (P < 0.0001). Results of ROC analyses are presented in Table 5 and ROC curves are presented in Figure 2. Vessel-specific hyperaemic MBF measurements are visually presented in Figure 3. To evaluate the importance of microvascular dysfunction, the performance of RFR < 0.65 is presented in Table 6. The RFR of patients with vessel-specific hyperaemic MBF < 2.00 mL/min/g is illustrated in Figure 4.

Discussion

In this comprehensive analysis of 346 patients with either persistent or paroxysmal AF undergoing ¹⁵O-H₂O PET MPI due to angina, we demonstrate that ongoing AF during an MPI procedure is associated with 19–25% lower global MBF compared with SR. Despite this, hyperaemic MBF—and to a lesser extent, RFR—derived from the ¹⁵O-H₂O PET MPI was strongly associated with subsequent early revascularization procedures and therefore guides clinical management of AF patients with or without previously documented CAD.

Global MBF is lower in patients with ongoing AF compared with patients in SR

Our results broadly fall in line with some smaller but comparable studies showing that hyperaemic MBF of AF patients with ongoing AF is reduced compared with AF patients in SR.^8–10 These studies indicate that the lower MBF observed during AF may be caused by the arrhythmia itself. This was recently substantiated by Takafuji et al.,⁸ who demonstrated that hyperaemic MBF measured by CT perfusion in 49 AF patients increased to levels no different than controls in SR after conversion from AF to SR by RFA therapy. Range et al.⁹ have previously documented similar results using ¹⁵O-H₂O PET MPI to measure MBF in 10 AF patients, where MBF also increased after successful RFA therapy, albeit not quite to the level of healthy controls. In contrast, Wijesurendra et al.¹¹ were unable to detect any differences in global hyperaemic MBF measured by first-pass perfusion cardiac magnetic resonance imaging in 19 patients in AF vs. 23 patients in SR.

As expected, most patients in AF had persistent AF. However, persistent AF was not an independent factor associated with reduced global hyperaemic MBF as demonstrated by multiple linear regression (see Table 3). Our findings therefore support the notion that the hyperaemic MBF of patients with paroxysmal AF could be varying dependent on the present cardiac rhythm and that AF does not necessarily alter the microvasculature akin to what is observed in type 2 diabetes-related small vessel disease.^15,16 It is, in our opinion, more likely that impaired diastolic filling is a contributing mechanism responsible for the reduced myocardial perfusion during AF. However, while hyperaemic heart rates were higher for patients in ongoing AF, the heart rate was not associated with reduced MBF in multiple linear regression (see Table 3) analysis. Nevertheless, the irregular cardiac rhythm of patients with ongoing AF may still hamper diastolic filling. All in all, our study was not designed to investigate the underlying pathological process or to determine whether a common mechanism may cause both frequent AF and reduced global hyperaemic MBF. To further improve our understanding of this problem, future studies should be designed to measure MBF in AF patients both in periods of SR and in AF.

Association between obstructive CAD and AF

It has previously been argued that AF can be caused by obstructive CAD, and multiple previous cross-sectional studies have therefore been designed to explore the association between AF and CAD. These studies are not entirely in agreement about the association between AF and CAD. However, the largest of these cross-sectional studies tend to conclude that either AF is not associated with significant coronary lesions or AF is even an independent factor associated with the ‘absence’ of significant coronary lesions as studied by CAG.^17–19 It is possible that this lack of association between AF and CAD is in fact indicative of the circulatory perturbations caused by AF. Thus, a large fraction of false positive referrals to CAG caused by the overall reduced perfusion in AF could lead to fewer findings of significant coronary stenoses compared with patients with no AF. Due to a lack of statistical power, we could not demonstrate a difference in ‘unnecessary’ CAG referrals between patients with ongoing AF compared with patients in SR (revascularization rates of 23% vs. 32%, P = 0.21) (see Supplementary data online, Table S2). Nevertheless, the notion that ongoing AF causes unnecessary CAGs is supported by a previous study in which positive single-photon emission computerized tomography (SPECT) MPI was less predictive for findings of stenosis during CAG in AF patients compared with non-AF patients.²⁰ In this paper, we demonstrate that vessel-specific hyperaemic MBF as measured by ¹⁵O-H₂O PET identified subsequent early revascularization of said coronary vessel even in AF patients scanned during ongoing AF. Regional hypoperfusion defined as RFR < 0.65 did identify early revascularizations for all groups except patients with previous CAD and ongoing AF (see Table 6). Anyhow, RFR did not outperform vessel-specific hyperaemic MBF in this regard and was likely more used as an adjunct indicator of the need to revascularize (see Table 6).

Of interest, we found that the ROC-based MBF cut-off indicating early revascularization for patients with ongoing AF and/or previously documented CAD at our centre was markedly lower than the MBF threshold indicating FFR-significant stenosis previously established for patients in SR with no known CAD.⁶ Due to the limited number of revascularizations performed in each subgroup, these cut-off values should be interpreted cautiously since they obviously reflect local practice both in terms of MPI reporting and the decision to revascularize difficult stenoses. Nevertheless, our results stress the need to interpret the results of ¹⁵O-H₂O PET MPI with additional care for patients with known CAD and/or ongoing AF.

Rate of revascularization was low even for patients with severe hypoperfusion

It has recently been demonstrated by the randomized multicentre trials ISCHEMIA and REVIVE that an initial revascularization strategy over optimal medical therapy for patients with moderate to severe obstructive CAD does not improve overall survival.^21,22 In our experience, this has resulted in a decreasing rate of revascularizations of patients with obstructive CAD and angina. In line with this, the per-patient overall revascularization rate was only 11.3% in the present study, even though a much larger fraction of patients had severely reduced hyperaemic blood flow in one or more territories (see Figure 3). As such, the positive predictive value of vessel-specific hyperaemic MBF was low (9–21%). We suspect other factors must have influenced the decision to revascularize—including CAG findings and patient-related factors that were not included in logistic regression analysis. However, as illustrated in Figure 4 depicting coronary arteries with MBF < 2.00 mL/min/g, the presence of ‘regional’ hypoperfusion as measured by low RFR values was indicative of follow-up revascularization since 40/323 (12.4%) coronary arteries of patients with RFR < 0.65 (‘filled symbols’) were revascularized vs. only 4/126 (3.2%) coronary arteries of patients with RFR ≥ 0.65. Global hypoperfusion combined with no regional hypoperfusion indicating small vessel disease may therefore be useful to exclude some patients from CAG with the purpose of revascularization. Surprisingly, for patients with global MBF < 2.00 mL/min/g, regional hypoperfusion measured as RFR < 0.65 was not a major determinant for referral to CAG compared with no regional hypoperfusion (58% vs. 56%), even though this metric most closely resembles the reversible ischaemia observed using retention tracers (82Rubidium or SPECT) for MPI. It is likely that this will change once cardiologists become more familiar with the use of this ¹⁵O-H₂O MPI metric.

Limitations

Several limitations to the study must be acknowledged. First, this was a clinical study designed to investigate how ¹⁵O-H₂O PET MPI is used to guide patient management and the main outcome parameter, early revascularization, and therefore reflects clinical practice in a single site more than an established reference test such as invasive FFR would have done. Second, this study does not evaluate the appropriateness or benefit of early revascularization, since follow-up major cardiac events or symptoms were not recorded. Third, estimates of optimal MBF cut-off values are associated with significant uncertainty due to the few events (revascularizations) that were recorded. Fourth, the decision to revascularize was entirely taken by the interventional cardiologist and may therefore have been influenced both by patient-related factors such as accessibility of the stenosis and physician-related factors such as experience. Fifth, the globally reduced hyperaemic MBF of patients with ongoing AF may have resulted in inappropriate ‘referrals’ for CAG. However, we believe that the risk of inappropriate ‘revascularization’ procedures caused by AF-related perfusion defects on the MPI is likely negligible, since the decision to revascularize overwhelmingly depends on angiographic findings—not on the preceding MPI.

Conclusion

Ongoing AF is associated with a 19–25% reduced global hyperaemic MBF as measured by ¹⁵O-H₂O MPI PET. Nevertheless, vessel-specific hyperaemic MBF was a strong indicator of early revascularization with an increasing likelihood of 23% per 0.1 mL/min/g decrease in MBF. The use of ¹⁵O-H₂O MPI PET to guide clinical management is therefore feasible even in patients with ongoing AF.

Supplementary data

Supplementary data are available at European Heart Journal - Cardiovascular Imaging online.

Funding

The study was supported by an unrestricted grant from MedTrace, Hoershold, Denmark.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

References

Author notes