Comparison of the Prognostic Value of Global and Regional Myocardial Flow Capacity Radius, Myocardial Flow Reserve, and Stress Myocardial Blood Flow Using Rubidium-82 with SiPM PET/CT

Matthieu DIETZ (  Matthieu.dietz@chuv.ch ) CHUV: Centre Hospitalier Universitaire Vaudois https://orcid.org/0000-0001-6041-2531 Christel H Kamani CHUV: Centre Hospitalier Universitaire Vaudois Gilles Allenbach CHUV: Centre Hospitalier Universitaire Vaudois Vladimir Rubimbura CHUV: Centre Hospitalier Universitaire Vaudois Stephane Fournier CHUV: Centre Hospitalier Universitaire Vaudois Marie Nicod Lalonde CHUV: Centre Hospitalier Universitaire Vaudois Niklaus Schaefer CHUV: Centre Hospitalier Universitaire Vaudois Eric Eeckhout CHUV: Centre Hospitalier Universitaire Vaudois Olivier Muller CHUV: Centre Hospitalier Universitaire Vaudois John O Prior CHUV: Centre Hospitalier Universitaire Vaudois https://orcid.org/0000-0003-1429-1374


Introduction
Myocardial perfusion imaging is a powerful non-invasive functional tool for risk strati cation, recommended by clinical practice guidelines (1)(2)(3)(4)(5). Compared with relative perfusion images, absolute quanti cation of myocardial blood ow (MBF) and myocardial ow reserve (MFR) by positron emission tomography (PET) could improve risk strati cation (6). Global and regional perfusion provide information on different aspects of myocardial perfusion. Impairment in global perfusion may be caused by either multivessel epicardial disease or microcirculatory dysfunction. Regional absolute perfusion measurements may enable the additive detection of small regional defects caused by epicardial coronary artery disease (CAD), which could not be detected with average global perfusion measurements (6).
The myocardial ow capacity (MFC) concept is an alternative approach integrating both MFR and stress MBF to depict MBF (6, 7). MFC may overcome some of the limitations of using MFR or stress MBF alone and represents a promising tool to improve clinical decision-making (7)(8)(9). However, despite a robust conceptual validation and recently promising clinical data especially after revascularization, studies demonstrating added clinical value of the MFC concept over stress MBF and MFR are still limited (8-10).
Silicon photomultipliers with digital readout (SiPM) PET is a new PET technology with improved spatial and timing resolution and a relatively high sensitivity and count-rate capability as compared to PET scanners using conventional photomultiplier tubes (11). Rubidium-82 ( 82 Rb) is a widely clinically used PET perfusion tracer produced from a strontium-82 ( 82 Sr)/ 82 Rb generator.
The most reliable quantitative variable on 82 Rb cardiac PET/CT for predicting cardiovascular events is not fully established and is unknown using SiPM technology.
The aim of this study was to prospectively compare on a SiPM PET camera, with 82 Rb, the prognostic value for cardiovascular events of global and regional MFR, stress MBF, and MFC radius.

Study Population
We prospectively enrolled participants with suspected myocardial ischemia to undergo 82 Rb cardiac SiPM PET/CT between June 2018 and June 2019 at the Lausanne University Hospital. Participants' cardiovascular risk factors and medication use were ascertained at time of PET imaging. The Local Ethics Committee approved this study protocol (#PB_2017-00634), and all participants gave written informed consent prior to inclusion.
Imaging protocol with SiPM 82 Rb PET/CT For each participant, a rest and adenosine or regadenoson stress SiPM PET/CT scan was performed, using a single dedicated camera (Biograph Vision 600, Siemens Medical Solutions, Knoxville, USA). Participants were instructed to fast for 6 h and avoid caffeine-containing food or beverages 24 h prior to the test. At rest, a 15-25 s i.v. infusion of 5 MBq/Kg of 82 Rb (Ruby-Fill® generator and Rb-82 elution system [v3], Jubilant DraxImage, Kirkland, QC, Canada) was administered with an automatic infusion system and 3D dynamic PET images were acquired starting at the beginning of the infusion over 6 min 19 s (12 × 8, 5 × 12, 1 × 30, 1 × 60, and 1 × 120 s). A second acquisition was then started following the same protocol with similar activity 2 min into an adenosine infusion (140 mg/kg/min over 6 min) or following a regadenoson administration (400 µg over 10 s). A low-dose CT (100 keV, 16 mAs) transmission scan was used for attenuation correction. Images were reconstructed by ordered subsets expectation maximization algorithms (4 iterations, 5 subsets, 4.0 mm FWHM gaussian post-lter, 220 × 220-pixel matrix size). Blood pressure, heart rate, and a 12-lead ECG were recorded throughout the procedure. The radiation dose for a 70 kg participant was estimated to be 2 × 0.39 mSv for rest and stress 82 Rb, and 1 × 0.17 mSv for the low-dose attenuation correction CT plus CT scout, resulting in a total dose of 0.95 mSv.

Usual quantitative myocardial perfusion analysis
Perfusion was assessed quantitatively measuring MBF in milliliter per gram per minute at rest and stress, using the highly automated FlowQuant v2.7 software (Ottawa, Ontario, Canada), with a 1-tissue compartment model with a ow-dependent extraction correction (12). MFR was calculated; MFR = stress MBF/rest MBF. Rate-pressure product adjusted rest MBF and MFR were determined to account for high resting heart rate or systolic blood pressure by multiplying rest MBF by 8500 mmHg/min and dividing by rate-pressure product (resting heart rate multiplied by resting systolic blood pressure). To reduce the potential spillover in image-derived blood activity curves, a dual spill-over correction was systematically applied (13), as well as global partial-volume recovery correction and motion correction (14).
Myocardial ow capacity MFC, developed by Johnson and Gould using 82 Rb PET imaging, is a metric that integrates combinations of resting MBF, stress MBF and MFR (7). To select an optimal threshold value for MFC and allow a straightforward and reliable comparison between MFC, MFR and stress MBF, the Euclidian distance in the Gould MFC diagram (x = stress MBF, y = MFR) was calculated to stratify MFC as per patient single continuous variables. This newly developed variable was called "MFC radius" and de ned as MFC

Quantitative analysis
In the current study we used continuous variables expressed as the average MFR and stress MBF values from (i) the entire ventricle (global MFR, global stress MBF and global MFC) and (ii) all vascular territories derived from standard segmentation: left anterior descending, left circum ex, and right coronary artery. Minimal regional values, representing the minimal average values among these 3 vascular territories, where used as representative values for regional MFR (minimum regional MFR), regional stress MBF (minimum regional stress MBF), and regional MFC radius (minimum regional MFC radius).
This methodology is different from the pixel-wise assessment of quantitative ow metrics created by Gould et al. Our methodology is more easily applicable, as pixel-wise assessments are not widely available. The used global and regional average MFR and stress MBF values in the current study are available in most used software solutions. MFC radius could be easily computed, integrating combinations of stress MBF and MFR as per patient single continuous variables. Moreover, global MFR and stress MBF used in the current study are well-studied quantitative parameters since more than two decades.

Clinical follow-up
The endpoint of the study was major adverse cardiovascular event (MACE), de ned as myocardial infarction, delayed revascularization (>6 months post-PET/CT), heart transplantation, hospitalization for congestive heart failure or de novo stable angina, and cardiac death. Early revascularizations observed within the rst 6 months of post-PET/CT were considered to have been triggered by the myocardial perfusion study and were excluded. Hospitalization for de novo stable angina was de ned as angina or chest pain of cardiac origin and requiring further investigations and hospitalization. Death from cardiac cause was de ned as death from myocardial infarction, congestive heart failure, valvular heart disease, sudden death, death without a witness or of unknown cause, and cardiac interventional/surgical procedure related. Outcome information were obtained from medical records available in the hospital information system. If unsuccessful, participant follow-up was obtained by a phone call to cardiologists or general practitioners and/or participants. In participants with multiple MACE, only the rst one was considered for survival analysis. Outcome data were collected in January-February 2021.

Statistics
We assessed the distribution of data with the Shapiro-Wilk test. Continuous parametric variables were expressed as mean ± SD and compared using Student's t-tests. Nonparametric data are presented as median [interquartile range] and compared using the Mann-Whitney U test. The chi-square test or Fisher exact test was used for analysis of categorical variables. Receiver-operating characteristic (ROC) analysis and pairwise comparisons were used to compare areas under the curves.
Because dichotomous variables represent a straightforward way to display in survival curves and odds ratios, the Youden index was used to select an optimal threshold value from the ROC for global and minimum regional MFR, stress MBF, and MFC radius measurements. Kaplan-Meier curves were used to elucidate the survival distributions regarding MACE. Differences in the outcomes of participants were assessed using the log-rank test. A Cox proportional hazard regression with adjustment for potential confounders was performed to determine the predictors of worse outcome. To prevent over tting of the multivariable Cox proportional hazards models, only cardiovascular risk factors with p values < 0.05 in univariate Cox proportional regression models were considered in the multivariable models. To show the potential incremental value of a quantitative variable, likelihood ratio chi-square test was used. Cox regression analysis and likelihood ratio chi-square test were also performed with global or minimum regional MFR, stress MBF, and MFC radius considered as continuous variables. Only signi cant results using both binary and continuous variables were considered statistically signi cant.
Kaplan-Meier curves were also used to elucidate the survival distributions regarding MACE with participants classi ed into tertiles based on global and regional MFR, stress MBF and MFC radius.
The statistical analysis was performed using R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). A p value of < 0.05 was considered statistically signi cant.

Participant's characteristics
From June 2018 to June 2019, 82 Rb SiPM PET/CT was performed in 279 participants. Two studies were excluded because of technical issues (delayed start at infusion). Follow-up was successful in 274 of 277 remaining participants (99%). 40 participants were censored due to early revascularization (5 CABG and 35 percutaneous coronary intervention, < 6 months after PET/CT) Baseline characteristics of the remaining study population of 234 participants are given in Table 1. Participants had a high prevalence of known CAD (54%), with a high burden of cardiovascular risk factors (hypertension: 73%; current or former tobacco use 45%; dyslipidemia: 68%). Preventive therapies were highly prescribed in the overall population: 58% with aspirin, 62% with beta-blockers, 58% with angiotensin-converting enzyme inhibitors, and 66% with lipid-lowering agents.

Comparative analysis
Participants with MACE had signi cantly worse global or minimum regional MFR, stress MBF, and MFC radius when compared with participants without (Fig. 1). In contrast, global or minimum regional rest MBF were similar among both groups (Table 2). There was a signi cantly higher prevalence of known CAD or history of myocardial infarction in patients with MACE as compared to patients without MACE (Table 2).

ROC analysis
On ROC analysis, using the Youden index, we derived the maximum potential effectiveness of global or minimal regional MFR, stress MBF, and MFC radius cutoffs for MACE prediction. For global absolute myocardial perfusion measurements, a threshold of 1.98 for MFR achieved a speci city and sensitivity of 73% and 64%, a threshold of 1.94 mL/g/min for stress MBF achieved a speci city and sensitivity of 59% and 83%, and a threshold of 3.12 for MFC radius achieved a speci city and sensitivity of 60% and 77%. For minimum regional absolute myocardial perfusion measurements, a threshold of 1.75 for MFR achieved a speci city and sensitivity of 72% and 66%, a threshold of 1.7 mL/g/min for stress MBF achieved a speci city and sensitivity of 61% and 79%, and a threshold of 2.7 for MFC radius achieved a speci city and sensitivity of 65% and 77%. Overall, the areas under the ROC curve analyses were numerically similar among all such parameters (p > 0.1 for all pairwise comparisons), with greater values for global and regional MFC radius (0.73 (0.65-0.80) for both; Fig. 2).

Univariate analysis
The Kaplan-Meier survival curves indicated that participants with global MFR < 1.98 or stress MBF < 1.94 mL/g/min or MFC radius < 3.12 had signi cantly higher rates of MACE (all p < 0.0001) as compared with those with normal perfusion. Similarly, patients with minimum regional MFR < 1.75 or stress MBF < 1.7 mL/g/min or MFC radius < 2.7 had signi cantly higher rates of MACE (all p < 0.0001) as compared with those with normal perfusion (Fig. E1 and Fig. E2 in the supplementary data online). Tertiles of global or minimal regional MFR, stress MBF, and MFC radius provided incremental information regarding MACE occurrence (Fig. 3).
On univariable Cox proportional regression, global MFR < 1.98 or stress MBF < 1.94 mL/g/min or MFC radius < 3.12, and minimum regional MFR < 1.75 or stress MBF < 1.7 mL/g/min or MFC radius < 2.7 emerged all as signi cant predictors of MACE (Table 3). Male gender, as well as known history of CAD and history of myocardial infarction were also found to be signi cantly predictive of MACE (Table 3).
When considered as continuous variables, global or minimum regional MFR, stress MBF, and MFC radius also emerged as signi cant predictors of MACE (Table E1 in the supplementary data online). Abbreviations as in Table 2 Multivariate Cox regression analysis On multivariable analysis after adjustments for sex, known CAD and history of myocardial infarction, (i) Global: MFR < 1.98, stress MBF < 1.94 mL/g/min, and MFC radius < 3.12, as well as (ii) Minimum regional: MFR < 1.75, stress MBF < 1.7 mL/g/min, and MFC radius < 2.7, emerged as all as independent predictors of MACE (Fig. 4). In contrast, male gender, known CAD, and history of myocardial infarction were predictors of events on univariable but not on multivariable analysis (Fig. 4) Fig. 5 summarizes the main ndings of multivariable analysis. Similar results were observed when global or minimum regional MFR, stress MBF, and MFC radius were considered as continuous variables (Fig. E3 in the supplementary data online).
Likelihood ratio chi-square test (Table E2 in the  supplementary data online) As binary variables, considering minimum regional values, we demonstrated an added value of MFC radius over MFR and stress MBF alone (p = 0.023 and p = 0.016 respectively), as well as an added value of MFR over stress MBF alone and of stress MBF over MFR alone (p = 0.026 and p = 0.019 respectively). As binary variables, considering global values, we demonstrated a trend for an added value of MFC radius over MFR and stress MBF alone (p = 0.061 and p = 0.087 respectively), and an added value of MFR over stress MBF alone and of stress MBF over MFR alone (p = 0.04 and p = 0.004 respectively). However, as continuous variables, no test was signi cant, except for a trend for an added value of minimum regional MFC radius over MFR alone (p = 0.083). Finally, integrating regional perfusion with minimum regional MFC, MFR or stress MBF did not increase prognostic performance compared to global perfusion alone (Table E2 in the supplementary data online).

Discussion
Our results support that global and regional MFR, stress MBF, and MFC radius, as assessed by using the latest advances in 82 Rb PET imaging, are powerful predictors of cardiovascular event, outperforming traditional cardiovascular risk factors such as the presence of known CAD or history of myocardial infarction. In a comprehensive analysis, we show similar results for global and regional MFR, stress MBF, and MFC radius as prognostic factors.
Our study has notable strengths. We utilized state-of-the-art 82 Rb PET imaging, using the latest advances in SiPM PET technology, including a Time of Flight of 214 ps and smaller crystals with higher PET resolution, which greatly ameliorate the image quality as compared to conventional PET with photomultiplier tubes (11). We used standardized automated post processing protocols of the dynamic data including dual spillover, global partial-volume recovery and motion corrections, and resting ratepressure-product (RPP) adjustments (13,14). PET perfusion was quanti ed by automated, purely objective measurements. All scans were performed using the same machine and the same acquisitions protocols. We had a 99% successful follow-up rate. We also used a novel quanti cation technique, MFC radius, that allows per patient single continuous variables to stratify MFC, and, therefore, represents a straightforward tool to select an optimal threshold value for MFC.
There is a major interest in using accurate non-invasive parameters to improve patient assessment and risk strati cation. In this line, absolute quanti cation of the myocardial perfusion as assess by PET/CT offers a powerful opportunity to tackle this challenge (15,16). However, the most reliable quantitative variable on 82 Rb cardiac PET/CT for predicting MACE is not fully established and is unknown using SiPM technology. Our similar results for global and regional MFC radius, MFR, and stress MBF as prognostic factors were unexpected and show some con icting results with previous studies.  (20,21).
Regional quantitative parameters as expressed by MFC have shown promising clinical data. Gould et al. showed in recent observational studies with large cohorts over long-term follow-up that the extent of severe regional impairment of MFC, expressed as percentage of left ventricular per-pixel regional MFC, provide optimal risk strati cation and is associated with a survival bene t gained by revascularization (8,9). This risk strati cation was not entirely assessed by global MFR and global stress MBF (9). Even limited extended area of abnormal regional MFC (< 0.5% of the left ventricle) were shown to be predictive of MACE, when integrating the relative stress perfusion (9). Using severe regional MFC impairment as target for revascularization, Bober  prognostic value of regional MFR < 2, being superior to global stress MBF and MFR (24). In this study the authors emphasized several methodological differences for regional measurements between them: (i) a decrease in at least two adjacent segments, (ii) Bom

Limitations
This study must be interpreted in the context of its single-center design, with a relatively modest sample size, which limits extensive subgroup analysis for outcomes. The period of follow-up was middle range, with a low incidence of hard cardiac events such as cardiac death. MFC, MFR, and stress MBF have, by de nition, interactions. Therefore, these parameters could not be assessed as independent predictors between them without collinearity. Our study emphasized myocardial blood ow measurements, and left ventricular ejection fraction, left ventricular volumes, regional wall motion, coronary artery calcium score and semi-quantitative evaluation of relative perfusion defects were not assessed in this study despite representing important information that could be part of routine PET/CT imaging.

Conclusions
Using the latest SiPM PET technology with 82 Rb, global and regional MFR, stress MBF, and MFC radius, are similar powerful predictors of cardiovascular event.

Declarations
Ethics approval All procedures performed in this study were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its last amendments or comparable ethical standards. The Local Ethics Committee approved this study protocol (#PB_2017-00634), and all participants gave written informed consent prior to inclusion.

Consent for publication
Informed consent was obtained from all individual participants included in the study.

Con icts of interest/Competing interests
The authors declare that they have no con ict of interest.

Funding
Dr. Dietz is supported by Research Fellowship Awards from the Société Française de Radiologie, Paris, France, and from the Agence Régionale de Santé Auvergne-Rhone-Alpes, Lyon, France.

Availability of data and material
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Receiver Operator Curve Analysis for the prediction of MACE in the study population. Receiver operator curves for MACE prediction by global MFR (A), global stress MBF (B), global MFC radius (C), minimum regional MFR (D), minimum regional stress MBF (E), and minimum regional MFC radius (F) Figure 3 MACE-free Survival Curves (n = 234) According to tertiles based on Global and Regional MFR, stress MBF, and MFC radius. (A) Global MFR, (B) Minimum regional MFR, (C) Global stress MBF, (D) Minimum regional stress MBF, (E) Global MFC radius, (F) Minimum regional MFC radius Figure 4 Predictors of MACE on Cox Proportional Hazards Modeling. Forest plots of hazard ratios derived from multivariable modeling with 95% con dence intervals for global MFR (A), global stress MBF (B), global MFC radius (C), minimum regional MFR (D), minimum regional stress MBF (E), and minimum regional MFC radius (F) values, along with covariates: male gender, known CAD and history of myocardial infarction Figure 5 Scatterplots summarizing hazard ratios derived from multivariable analysis for global (A) and minimum regional (B) MFR, stress MBF and MFC radius measurements

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. OnlineTablesandOnlineFigures.docx