Investigators analyzed a retrospective registry of patients referred for noninvasive coronary imaging because clinicians suspected obstructive coronary artery disease. From 971 initially considered, exclusion criteria (prior myocardial infarction or revascularization, heart failure, cardiomyopathy, moderate or greater valvular disease, pacemaker, technically unsuccessful imaging, or rhythms other than AF or SR during PET MPI) produced a final pool of 874 patients. The cohort included 31 patients with ongoing atrial fibrillation during PET MPI and 843 with sinus rhythm. Using a propensity score approach, the team matched the 31 AF patients to 120 SR patients at a 1:4 target ratio. Matching variables were age, sex, body mass index, number of traditional CAD risk factors (hypertension, dyslipidemia, diabetes, smoking history, family history), calcium score, and total CCTA status. Two AF patients were matched to only two SR controls due to the predefined propensity-score window.
The center performed coronary computed tomography angiography (CCTA) and myocardial perfusion imaging with [15O]H2O PET as part of routine diagnostic workflows. CCTA scanners included three Siemens models; imaging incorporated unenhanced CT for Agatston calcium scoring and contrast-enhanced, ECG-gated angiography. The investigators graded each major epicardial artery on a 0–5 scale adapted from CAD-RADS, then used the highest individual-vessel grade as a patient-level total CCTA status. PET MPI used continuous intravenous adenosine at 140 µg/kg/min to induce hyperemia, followed by weight-adjusted [15O]H2O bolus and dynamic image acquisition. The clinical protocol included only hyperemic imaging; rest images were not available for most patients and therefore excluded from analysis. Hyperemic MBF quantification employed a single-tissue compartment model and the standard 17-segment scheme; global and territorial MBF values were recorded in mL/g/min. The study defined impaired hyperemic MBF as any segment (excluding basal septal segments 2 and 3) below 2.3 mL/g/min.
The matched groups showed no statistically significant differences in age, sex, body mass index, or the five traditional CAD risk factors, consistent with the matching strategy. Investigators observed that beta-blocker and diuretic use was more common among AF patients. Laboratory values differed: AF patients had lower glomerular filtration rate, higher hemoglobin, and elevated NT-proBNP compared with SR patients. Hemodynamic measures during PET MPI varied at baseline; AF patients had higher diastolic blood pressure, heart rate, and rate-pressure product at rest. Immediately after the [15O]H2O bolus, systolic blood pressure, heart rate, and rate-pressure product were similar between groups, though diastolic pressure remained higher in AF patients. The authors report that premedication with beta-blockers before CCTA and calcium score distributions were similar between groups, and no differences were found in individual epicardial artery status on CCTA.
Global hyperemic myocardial blood flow differed markedly between the groups. Median global MBF in AF patients was 1.9 mL/g/min (IQR 1.4–2.6) versus 3.0 mL/g/min (IQR 2.4–3.6) in matched SR patients; the between-group p-value was < 0.001. When assessed by standardized perfusion territories, AF patients had lower hyperemic MBF across territories compared with SR controls. The predefined threshold for impaired MBF (any segment < 2.3 mL/g/min) was met by 25 of 31 AF patients (81%) and by 50 of 120 SR patients (42%); this difference was also statistically significant (p < 0.001). Investigators emphasize that total CCTA status and calcium score did not differ between groups, indicating that the reduced MBF in AF patients was not explained by greater epicardial atherosclerotic burden as measured by the CCTA parameters reported.
To probe independence, the study team employed three multivariate general linear models with AF as a fixed factor. Model 1 adjusted for the six variables used in propensity matching (age, sex, body mass index, number of CAD risk factors, calcium score, and total CCTA status). Model 2 added hemoglobin and beta-blocker use—variables that differed between groups and related to global hyperemic MBF. Model 3 incorporated the covariates of Model 2 plus diastolic blood pressure after tracer bolus, GFR, and diuretic use—variables that differed between AF and SR but lacked a direct statistical association with global MBF. Across all three models, the presence of AF remained significantly associated with lower hyperemic MBF (p < 0.001 in each model). The authors report no violations of model assumptions: Levene’s test showed no unequal variances, residuals were normally distributed, and variance inflation factors were consistently < 2.5.
The investigators acknowledge limits inherent to the registry-based, retrospective design; causal inference cannot be established. The source did not report data on AF burden (beyond rhythm classification during PET MPI), nor did it provide rest MBF images for most patients because clinical protocol focused on hyperemic imaging. The study was single-center and included a relatively small number of patients with ongoing AF (n = 31). The data underlying the analysis cannot be publicly shared because of statutory restrictions; eligible researchers may apply for access from the Wellbeing Services County of North Savo per the contact information provided by the authors. These restrictions and the lack of rest perfusion imaging were noted as constraints on interpretation and opportunities for further investigation.