Improved Assessment of Coronary Artery Disease in Obese Subjects with Flurpiridaz-¹⁸F PET Myocardial Perfusion Imaging: A Prespecified Subgroup Analysis of the AURORA Phase 3 Study

Abstract

Background:

Non-invasive assessment of coronary artery disease (CAD) in obese patients is challenging due to suboptimal image quality. We performed a pre-specified secondary analysis of obese subjects from the second phase III trial of flurpiridaz-¹⁸F PET.

Methods:

In total, 604 patients with suspected CAD underwent flurpiridaz-¹⁸F PET and ^99mTc-SPECT MPI before invasive coronary angiography (ICA) across 48 sites. MPI images were interpreted by three blinded experts. The primary endpoint was sensitivity and specificity for diagnosing CAD (≥50% stenosis on quantitative coronary angiography), requiring the lower 95% confidence interval to exceed 60% by 2 same blinded expert core lab readers, assessed using one-sided z-test (α=0.025). Sensitivity and specificity were compared between PET and SPECT, and between obese (BMI ≥30 kg/m²) and non-obese subjects.

Results:

Of 578 evaluable subjects, 298 (51.6%) were obese (mean age 62.0 years, 64.4% male, mean BMI 35.6 kg/m²) and 117 (39.3%) had CAD by ICA. In obese subjects, flurpiridaz-¹⁸F PET MPI met the primary endpoint, with lower 95% confidence limits >60% and one-sided p <0.025 by the same 2 readers (Readers 1 and 2). Sensitivity across 3 readers ranged from 70.1% to 88.0% and specificity ranged from 53.6% to 74.0%. In obese subjects, flurpiridaz-¹⁸F PET MPI showed significantly higher sensitivity compared to ^99mTc-SPECT MPI by Readers 1 and 3 (Reader 1: 72.6% vs. 60.7%, p=0.01; Reader 3: 88.0% vs. 74.4%, p=0.002), with non-inferior specificity by the same two readers (Reader 1: 68.0% vs. 61.3%; Reader 3: 53.6% vs. 50.8%; p<0.01 for non-inferiority, for both), noting that the primary endpoint was met by a partially overlapping reader pair (Readers 1 and 2). Diagnostic performance of flurpiridaz-¹⁸F PET was higher than SPECT across BMI categories. No statistically significant difference in diagnostic performance was observed between obese and non-obese subjects (Sensitivity p=0.26, specificity p=0.25 and accuracy p=1.0).

Conclusion:

Flurpiridaz-¹⁸F PET MPI demonstrates high diagnostic efficacy for detecting CAD across BMI categories in obese patients compared to ICA and ^99mTc-SPECT, with no loss of diagnostic performance in obese relative to non-obese patients.

Clinical trial registration:

NCT03354273

Introduction

Rates of obesity are high and rising across the world.¹ Obesity is associated with a higher prevalence of cardiovascular risk factors such as hypertension, diabetes and dyslipidemia, resulting in an increased incidence of coronary artery disease (CAD).² Obese patients with cardiovascular disease are also at a higher risk of long term morbidity and mortality.^3,4 Obese patients are often referred for testing for possible CAD, due to frequent symptoms such as dyspnea or reduced effort tolerance. Unfortunately, the accuracy of most non-invasive testing strategies is lower in the obese, especially the morbidly obese.^5,6

Evaluation of CAD in obese patients is limited by poor diagnostic quality of non-invasive tests due to poor acoustic windows in the case of stress echocardiography, activity limitations with respect to exercise treadmill testing, and soft-tissue attenuation artifacts with SPECT myocardial perfusion imaging (MPI).⁷ They are often at an increased risk of complications with invasive angiography, suffer reduced image quality and have a necessity for higher radiation exposures with CT coronary angiography.⁸ Despite its limitations, SPECT MPI has been a preferred modality for evaluation of CAD among obese patients due to its widespread availability and presence of extensive literature to support its role in diagnosis and prognosis of CAD.⁹ However, excess soft tissue attenuation and tracer activity from adjacent organs can produce artifacts, reduce signal to noise ratio and affect the diagnostic quality and interpretive certainty of image reads. Positron Emission Tomography (PET) MPI has inherent attenuation correction and offers better spatial resolution, improving image quality, and providing higher diagnostic accuracy.^10,11 However, current PET radiotracers require either the presence of an on-site cyclotron or the purchase of expensive generators, limiting its widespread use.

Flurpiridaz-¹⁸F is a new FDA-approved cardiac PET perfusion radiotracer that binds to the mitochondrial complex 1. Unlike other PET radiotracers, its longer half-life makes it amenable to be procured as a unit dose from a regional cyclotron, without the need of an on-site cyclotron or generator. Its shorter positron range and high myocardial extraction results in improved spatial resolution and superior diagnostic image quality, possibly mitigating the shortcomings of SPECT in obese patients. It has been shown to offer higher diagnostic accuracy and better image quality than conventional SPECT.^12,13 The current study is a pre-specified subgroup analysis of obese patients with suspected CAD enrolled in the second Phase III study (Clinical trials.gov: NCT03354273) assessing the diagnostic efficacy of flurpiridaz-¹⁸F PET against invasive coronary angiography as the primary end-point and a comparison with ^99mTc SPECT MPI as secondary endpoint.

METHODS

Study design and population

The study design and protocol has been published previously.¹⁴ In brief, this was a phase 3, prospective, open-label study to determine the diagnostic efficacy of flurpiridaz-¹⁸F PET. A total of 730 patients with suspected but no known CAD referred for a clinically indicated invasive coronary angiogram (ICA) were enrolled across 48 clinical sites in the United States, Canada and Europe from June 5, 2018, to May 9, 2022. The major exclusion criteria included unstable cardiovascular status (myocardial infarction [MI]/unstable angina pectoris within 6 months; transient ischemic attack/stroke within 3 months; symptomatic valvular disease; significant congenital heart disease; NYHA functional class III/IV heart failure), history of known CAD (prior coronary artery bypass grafting, percutaneous coronary intervention, MI), nonischemic cardiomyopathy, and history of heart transplantation. All patients provided written informed consent and the study was approved by Institutional Review Boards or ethics committees at all study sites.

The study design is shown in Supplement Figure 1. All patients were scheduled to undergo both rest and stress flurpiridaz-¹⁸F PET and ^99mTc SPECT MPI, with same type of stress (pharmacologic or exercise) for both studies, within 60 days before ICA.

Obese subgroup

This was a pre-specified subgroup analysis of the second phase III study. Obesity was defined based on a body-mass index (BMI) threshold of ≥ 30 kg/m².

Imaging protocol

Protocols for the study have been described in detail in a previous publication.¹⁴ All SPECT and PET cameras used for this study were qualified for technical capability and quality. SPECT protocols were based on standard ASNC guidelines. Prespecified Tc-99m SPECT tracer dosing was 8.0 to 12.0 mCi for rest and 24.0 to 36.0 mCi for stress with a minimum stress/rest ratio of 3.0. Choice of one or two day protocol for obese subjects was site-dependent, no specific recommendations were provided from the sponsor. For the PET protocol, a rest dose of 2.5–3 mCi of flurpiridaz-¹⁸F and a stress dose of 9–9.5 mCi (for exercise stress) or 6–6.5 mCi (for pharmacologic stress) was used with an inter-dose interval of 30 min (pharmacologic stress) to 1 hour (exercise stress). The CT attenuation correction component is not included in the reported PET effective dose. To prevent comparisons between different stress agents and modalities between the PET and the SPECT study, it was required that the same stressor be used for both MPI studies. Raw data images were submitted to a central laboratory (BioClinica, Inc., Newtown, PA) for quality control, reconstruction and preparation and coordination of the blinded reads.

Image interpretation

All PET and SPECT studies were read randomly in a blinded fashion by three independent experts board certified in nuclear cardiology. No clinical patient or site information was provided to the readers, and SPECT images were read at a separate time from PET images. All rotating projection images, gated images, and perfusion images from the SPECT study, and overlay of transmission/emission data, gated images, and perfusion images from the PET study were made available to the blinded readers. Prior to study reads, readers underwent study-specific calibration and training sessions with representative cases to standardize interpretation criteria across modalities. These sessions included review of the study protocol, interpretation guidelines and recognition of modality-specific artifacts, including physiologic apical thinning on flurpiridaz PET and attenuation-related artifacts and count-density variations on conventional and CZT SPECT systems. All SPECT studies were read from non-attenuation corrected perfusion images to maintain consistency between subjects scanned from different sites on different SPECT scanners. The use of non-AC SPECT was a pre-specified harmonization decision across different sites; this choice may have disadvantaged SPECT in obese patients where inferior-wall attenuation is most pronounced. Based on the perfusion images and gated data, readers provided an assessment of presence or absence of CAD, which was used to determine the primary end-point. Image quality was also assessed and scored as excellent, good, fair, poor or unevaluable. Diagnostic confidence was scored as definitely normal, probably normal, equivocal, probably abnormal, or definitely abnormal. Diagnostic certainty was defined as definitely normal or definitely abnormal. Readers also provided a semi-quantitative segmental perfusion and wall motion assessment per standard reporting guidelines.^15,16

Invasive Coronary Angiography (ICA)

ICA was performed per local institutional practice, after which data were submitted to an angiography core laboratory (Sponsors International) for quality control and a centralized blinded read by a single reviewer. Quantitative invasive angiography (QCA) was considered the gold standard for CAD determination. QCA was performed using validated software (CAAS Workstation version 8.1, Pie Medical Imaging, Maastricht, The Netherlands). Measurements were performed according to standardized core laboratory procedures, including calibrated vessel analysis using end-diastolic frames and assessment of percent diameter stenosis in at least two orthogonal projections whenever technically feasible. When image acquisition was insufficient to permit reliable quantitative assessment, lesions could be classified as non-interpretable at the discretion of the core laboratory reader. Patients were considered disease positive if there was a stenosis of ≥50% in ≥1 major coronary artery or ≥1 major branch vessel on QCA. An independent Data and Safety Monitoring Board monitored patient safety during the trial.

Study Endpoints

The primary study endpoints were sensitivity and specificity of flurpiridaz-¹⁸F PET for diagnosis of CAD (by QCA as defined above). The prespecified criteria of the overall study and for this subgroup analysis were statistical superiority of sensitivity and specificity of flurpiridaz-¹⁸F PET MPI in the detection of CAD over the threshold of 60% by the same two out of 3 expert readers, in agreement with the U.S. Food and Drug Administration. Thus, the lower bound of the 2-sided 95% CI for both sensitivity and specificity must exceed 60%. the primary study results have been published.¹⁷ In the main study, Flurpiridaz-¹⁸F PET met the pre-specified thresholds for sensitivity and specificity.

Comparison of the sensitivity and specificity of flurpiridaz-¹⁸F PET and SPECT was a key secondary endpoint. Additional secondary efficacy endpoints included radiation exposure, defect extent/severity, image quality (rated as excellent, good, fair, poor, uninterpretable) and diagnostic certainty (rated as definitely abnormal or definitely normal). Primary endpoint success required the same two readers exceeding the projected sensitivity and specificity thresholds. “Majority rule” (based on agreement of 2 of 3 readers) was also considered and reported.

Subgroup analyses:

The following post-hoc subgroup analyses were also performed in obese subsets: (i) extreme obesity defined as subjects with BMI ≥40 kg/m2; (ii) male vs. female sex and (iii) subjects who underwent SPECT using cadmium-zinc telluride scanners.

Statistical Analysis

A modified intent-to-treat (mITT) population which included all patients with evaluable rest and stress flurpiridaz-¹⁸F PET MPI, rest and stress SPECT MPI, and ICA was used for all efficacy analyses. The statistical superiority of both sensitivity and specificity to a threshold of 60% in flurpiridaz PET was performed using a 1-sided 1-sample z-test at α = 0.025 for all readers or the same 2 out of 3 readers.

The criteria for key secondary endpoint comparison of flurpiridaz-¹⁸F PET MPI with ^99mTc-SPECT MPI was statistical superiority in sensitivity and non-inferiority in specificity (non-inferiority margin of 10%). The efficacy of the key secondary endpoint would need to be demonstrated for the same 2 of the 3 readers for whom the primary endpoint had superior efficacy. The test of superiority of sensitivity in Flurpiridaz PET over SPECT was performed with a 1-sided McNemar’s test at α = 0.025. Similarly, the test of specificity in noninferiority between flurpiridaz PET and SPECT was performed with Nam’s RMLE method for noninferiority (margin = 0.1) at a 1-sided α = 0.025.

Receiver-operating characteristic (ROC) curves were generated from plotting sensitivity versus (1-specificity) predicted by summed stress scores (median value among the 3 readers) using logistic regression. The ROC curves were compared between PET and SPECT based on the DeLong et al. method¹⁸ and the best thresholds were identified based on the Youden Index method. For flurpiridaz-¹⁸F PET MPI, the comparison of ROC curves between obese and non-obese subgroups were made using Fisher’s exact test. Since the lack of attenuation correction can negatively impact SPECT, similar analyses were also performed based on summed difference scores. Secondary endpoints of interpretive certainty, image quality, radiation exposure and defect extent/severity were compared between PET and SPECT groups using Fisher’s Exact test or Wilcoxon signed sum test (for continuous variables). Inter-reader agreement and intra-reader reproducibility were assessed in obese subjects between the 3 readers using Cohen’s and Fleiss’ kappa statistics. As there were only 39 patients in the obese subgroup who underwent exercise stress, patients undergoing both exercise and pharmacologic stress MPI were grouped together for reporting and analysis. All analyses were performed using SAS^® 9.4 software. A two-sided p-value of 0.05 or one-sided p-value of 0.025 was considered as significant. No multiplicity adjustment was performed for secondary endpoints.

Results

Of 578 enrolled subjects with evaluable studies, 298 (51.6%) were classified as obese (BMI > 30 kg/m²). The mean BMI of obese subjects was 35.6 ± 5.5 kg/m². Mean age of obese subjects was 62.0 ± 9.5 years, 64.4% were male and 117 (39.3%) had CAD on invasive angiography. The majority (n=259, 87%) of subjects in the obese subgroup underwent pharmacologic stress testing. Compared to non-obese subjects, obese subjects were younger, more likely to have hypertension and diabetes, more likely to need pharmacologic stress for SPECT, had a lower median summed stress score on PET and were less likely to have multi-vessel disease on subsequent coronary angiography (Table 1).

Table 1:

Baseline characteristics, subdivided by obese vs. non-obese participants.

Patient Characteristics	Obese (N= 298)	Non-obese (N=280)	p-value
Age, years	62.0 (9.55)	65.6 (9.01)	<.0001
Male sex	192 (64.4)	198 (70.7)	0.11
Race			0.05
Caucasian	243 (81.5)	230 (82.1)
African American	26 ( 8.7)	9 ( 3.2)
Other	6 (2.0)	7 (2.5)
Not reported	23 ( 7.7)	34 (12.1)
Ethnicity			0.01
Hispanic	43 (14.4)	36 (12.9)
Non-Hispanic	220 (73.8)	185 (66.1)
Not Reported	35 (11.7)	58 (20.7)
Weight, kg	104.66 (19.718)	76.74 (11.069)	<.0001
BMI, kg/m2	35.6 (5.47)	25.9 (2.51)	<.0001
Chest Pain			0.86
Asymptomatic	57 (19.1)	59 (21.1)
Atypical angina	120 (40.3)	106 (37.9)
Non-anginal chest pain	22 ( 7.4)	24 ( 8.6)
Typical angina	99 (33.2)	91 (32.5)
Pre-test probability			0.64
Low/Very Low	64 (21.5)	62 (22.1)
Intermediate	219 (73.5)	199 (71.1)
High	15 ( 5.0)	19 ( 6.8)
Risk Factors
Hypertension	241 (80.9)	184 (65.7)	<.0001
Hyperlipidemia	234 (78.5)	217 (77.5)	0.77
Smoking	160 (53.7)	155 (55.4)	0.69
Angina	64 (21.5)	74 (26.4)	0.16
Diabetes Mellitus	128 (43.0)	66 (23.6)	<.0001
Renal Impairment	13 ( 4.4)	12 ( 4.3)	0.96
Hepatic Impairment	54 (18.1)	35 (12.5)	0.06
MPI Variables
Stressor			0.01
Exercise
Pharmacologic	39 (13.1)	58 (20.7)
	259 (86.9)	222 (79.3)
Median Summed Stress Score
PET	2.5 [0, 10]	5.0 [0, 12]	0.03
SPECT	4.0 [1, 8]	4.0 [0.5, 9]	0.99
Median Summed Rest Score
PET	0.0 [0, 1]	0.0 [0, 1]	0.86
SPECT	2.0 [0, 3]	1.0 [0, 3]	0.36
Angiography Results
CAD	117 (39.3)	132 (47.1)	0.06
Multivessel disease	49 (16.4)	68 (24.3)	0.02

Continuous variables were described as mean (SD) or median [quartile 1, quartile 3] and compared using t-test or Wilcoxon rank-sum test for summed scores and categorical variables were described as n (%) and compared using chi-square test or Fisher Exact test. CAD defined as stenosis of ≥50% in ≥1 coronary artery. Multivessel disease defined as stenosis of ≥50% in 2 or more epicardial coronary arteries.

Primary Endpoint:

Sensitivity and specificity of flurpiridaz-¹⁸F PET for detection of ≥50% by ICA for each of the 3 readers for the obese subgroup are shown in Table 2. Per Reader 1 and Reader 2, sensitivity and specificity of flurpiridaz were significantly higher than the prespecified threshold value of 60% in the obese subgroup. The third reader’s sensitivity was also significantly higher than the prespecified threshold (P < 0.0001), although the specificity was not significant (P = 0.96). Therefore, the primary efficacy endpoint of the study was met by the same two of three readers exceeding the prespecified threshold for both sensitivity and specificity in the obese participant subgroup. Similar results were seen in obese subjects when assessed using majority rule (agreement between 2 of 3 readers): Sensitivity by majority rule was 76.9% (95% CI: 69.3, 84.6, p=0.0001) and specificity by majority rule was 66.9% (95% CI: 60.0, 73.7, p=0.03).

Table 2:

Primary endpoint results in the obese subgroup for Flurpiridaz-¹⁸F PET MPI.

	Sensitivity,% (95% CI)	p-value	Specificity, % (95% CI)	p-value
Reader 1	72.6 (64.6, 80.7)	0.0026	68.0 (61.2, 74.8)	0.01
Reader 2	70.1 (61.8, 78.4)	0.013	74.0 (67.6, 80.4)	0.0001
Reader 3	88.0 (82.2, 93.9)	<0.0001	53.6 (46.3, 60.9)	0.96
Majority Rule	76.9% (69.3, 84.6)	0.0001	66.9% (60.0, 73.7)	0.03

Confidence intervals are from 2-sided 95% confidence intervals (CIs) based on the normal approximation to the binomial distribution. The hypothesis tests are 1-sided 1-sample z-test tests with a significance level of 0.025 for sensitivity and specificity.

Comparison between 18F-Flurpiridaz PET vs. 99mTc SPECT in Obese Subjects

Diagnostic Performance:

When comparing PET to SPECT, flurpiridaz-¹⁸F PET was superior in sensitivity and non-inferior in specificity when compared ^99mTc-SPECT when assessed by Reader 1 and Reader 3. (Figure 1). Using majority rule, sensitivity (76.9% vs. 69.2%, p=0.064) of flurpiridaz-¹⁸F PET MPI did not reach pre-specified significance threshold for superiority over SPECT. By majority rule, flurpiridaz PET was non-inferior in specificity (66.9% vs. 61.9%, p=0.001 for non-inferiority) than ^99mTc-SPECT MPI.

Figure 1: Diagnostic Efficacy of 18F-Flurpiridaz PET vs. SPECT for obese subjects (N=298).

Area under Receiver Operating Characteristic (ROC) curves in Figure 2 were created based on median summed stress scores of three readers. ROC curve analysis showed superior diagnostic performance for flurpiridaz-¹⁸F PET MPI vs. ^99mTc-SPECT in obese subjects (0.79 vs. 0.67, p<0.001; Figure 2). The best threshold value was 8, identified based on the Youden Index method, for both PET and SPECT. With the best threshold, sensitivity of flurpiridaz-¹⁸F PET MPI (60.7% vs. 45.3%, p=0.003) was superior to ^99mTc-SPECT MPI and specificity (88.4% vs. 81.2%, p<0.001) was non-inferior. Figure 3 shows an example of a 60 year old female subject with a BMI of 46 kg/m2 who had significant mid and distal vessel stenosis in the right coronary artery on her coronary angiography, where the SPECT study was falsely negative, while the PET correctly identified a medium sized region of moderate to severe intensity reversible perfusion defect in the inferior regions supplied by the right coronary artery. Among obese subjects, area under ROC curve with median summed stress score for flurpiridaz-¹⁸F PET was higher than SPECT across different categories of BMI (<30, 30–39.9, >=40 kg/m²; Figure 4)

Figure 2: Area under Receiver Operating Characteristic (ROC) curve generated from median summed stress scores across 3 readers for PET vs. SPECT in obese vs. non-obese subjects.

ROC curves generated from median summed stress score (SSS) across 3 readers using logistic regression. AUC compared between PET vs. SPECT by DeLong et al. method, obese vs. non-obese by Fisher Exact test.

Figure 3: Example of a true positive 18F-flurpiridaz PET and a false negative SPECT in a 60 year old female subject with body mass index of 46 kg/m² with significant stenosis in the mid and distal right coronary artery on invasive coronary angiography.

Figure 4: Area under Receiver Operating Characteristic (ROC) Curve generated from median summed stress scores across 3 readers for PET vs. SPECT, across BMI categories.

ROC curves generated from median summed stress score (SSS) across 3 readers using logistic regression. AUC compared by DeLong et al. method.

The ROC curves in Figure 5 and Supplementary Figure 2 were generated based on median summed difference scores of three readers. The superiority of overall diagnostic performance was demonstrated for flurpiridaz-¹⁸F PET MPI in obese subjects (0.75 vs. 0.67, p=0.016; Figure 5) using summed difference score. The values of best threshold for median summed difference scores were 5 and 2 for PET and SPECT respectively. With the best threshold estimated from each ROC curve, sensitivity was 63.2% vs. 60.7% (p=0.311) and specificity was 81.8% vs. 68.0% (p<0.001).

Figure 5: Area under Receiver Operating Characteristic (ROC) Curve generated from median summed difference scores across 3 readers for PET vs. SPECT in obese vs. non-obese subjects.

ROC curves generated from median summed difference score (SDS) across 3 readers using logistic regression. AUC compared between PET vs. SPECT by DeLong et al. method.

Among obese subjects, inter-reader agreement for PET (Fleiss kappa=0.61, 95% CI: 0.54–0.68) was higher than for SPECT (Fleiss kappa=0.49, 95% CI: 0.43–0.56). Intra-reader reproducibility kappa ranged 0.68–0.92 (PET) and 0.61–0.79 (SPECT) (Supplementary Tables 2 and 3).

In territory-level analysis (Supplementary Table 4), PET demonstrated significantly higher sensitivity than SPECT in the LAD territory (majority rule: 61.8% vs. 30.9%, p<0.001) and the RCA territory (53.8% vs. 35.4%, p=0.005). Performance was similar between modalities in the LCx territory. Specificity of PET was non-inferior to SPECT in all three territories.

Defect extent/severity:

Median summed difference and summed rest scores were significantly lower in obese subjects on flurpiridaz-¹⁸F PET vs. ^99mTc-SPECT MPI (<0.001; Table 3).

Table 3:

Comparison of secondary end-points between Flurpiridaz-¹⁸F PET and 99mTc SPECT MPI among 298 obese subjects.

	Flurpiridaz-18F PET	SPECT	p
Interpretive Certainty			<0.0001
Definitely normal/abnormal	83.6% (79.3%, 87.8%)	47.0% (41.3%, 52.6%)
Probably normal/abnormal/equivocal	16.4% (12.2%, 20.7%)	53.0% (47.4%, 58.7%)
Image quality
Excellent/Good at rest	90.9% (87.7%, 94.2%)	66.8% (61.4%, 72.1%)	<0.0001
Excellent/Good at stress (Pharm)	95.7% (93.2%, 98.2%)	73.4% (68.0%, 78.8%)	<0.0001
Excellent/Good at stress (Exercise)	94.9% (87.9%, 100.0%)	89.7% (80.2%, 99.3%)	0.41
Radiation Exposure (mSv; mean ± SD)	6.28 ± 0.76	11.93 ± 2.70	<.0001
Median Summed Stress Score (Q1, Q3)	2.5 (0, 10)	4 (1, 8)	0.26
Median Summed Difference Score (Q1, Q3)	0 (0, 7)	1 (0,5)	<0.001
Median Summed Rest Score (Q1, Q3)	0 (0, 1)	2 (0, 3)	<0.001

All comparisons provided for Majority Rule (agreement between 2 out of 3 readers); or median value of the 3 readers (for the summed scores). Confidence intervals are from 2-sided 95% confidence intervals (CIs) based on the normal approximation to the binomial distribution. P-value measured using Fisher’s Exact test or Wilcoxon signed rank test

Image quality:

Among subgroup of obese subjects, image quality rating of excellent or good of flurpiridaz-¹⁸F PET was higher than that of SPECT for rest and pharmacologic stress images (Table 3).

Diagnostic Certainty:

Among obese subjects, diagnostic certainty (% studies rated as definitively normal or abnormal) of flurpiridaz-¹⁸F PET was significantly higher than that of SPECT MPI (Table 3).

Radiation Exposure:

Flurpiridaz-¹⁸F PET MPI had significantly lower radiation exposure compared to ^99mTc-SPECT (6.28 ± 0.76 mSv PET vs 11.93 ± 2.70 mSv SPECT), p<0.0001) even among obese subjects.

Flurpiridaz-¹⁸F PET performance in obese vs. non-obese subjects

Diagnostic Performance:

There was no significant difference between sensitivity (p=0.26), specificity (p=0.25) and diagnostic accuracy (p=1.0) of flupriridaz-¹⁸F PET between obese and non-obese subjects (Figure 6 and Table 4). ROC curve analysis showed similar area under curve for obese and non-obese subjects (0.79 vs. 0.81, p>0.05; Figure 2).

Figure 6: Diagnostic Performance of 18F-Flurpiridaz PET between obese vs. non-obese subjects.

Error bars represent the width of the 95% confidence intervals.

Table 4:

Comparison of primary and secondary endpoints for 18F-flurpiridaz PET between obese and non-obese subjects.

Flurpiridaz-18F PET	Obese N=298	Non-obese N=280	p-value
Sensitivity	76.9% (69.3%, 84.6%)	83.3% (77.0%, 89.7%)	0.26
Specificity	66.9% (60.0%, 73.7%)	60.1% (52.2%, 68.0%)	0.25
Diagnostic Accuracy	70.8% (65.6%, 76.0%)	71.1% (65.8%, 76.4%)	1.0
Interpretive Confidence			0.35
Definitely normal/abnormal	83.6% (79.3%, 87.8%)	86.4% (82.4%, 90.4%)
Image quality
Excellent/Good at rest	90.9% (87.7%, 94.2%)	95.7% (93.3%, 98.1%)	0.03
Excellent/Good at stress	95.6% (93.3%, 98.0%)	97.1% (95.2%, 99.1%)	0.38
Radiation Exposure (mSv; mean ± SD)	6.3 ( 0.6 )	6.4 ( 0.6 )	0.41
Median Summed Stress Score (Q1, Q3)	2.5 (0, 10)	5 (0, 12)	0.03
Median Summed Difference Score (Q1,Q3)	0 (0, 7)	4 (0, 10)	0.002
Median Summed Rest Score (Q1, Q3)	0 (0, 1)	0 (0, 1)	0.86

All comparisons provided for Majority Rule (agreement between 2 out of 3 readers); or median value of the 3 readers (for the summed scores). Confidence intervals are from 2-sided 95% confidence intervals (CIs) based on the normal approximation to the binomial distribution. P-value based on Fisher's exact test or Wilcoxon signed rank test

Image Quality:

Image quality deemed excellent/good was similar at stress for flurpiridaz-¹⁸F PET between obese and non-obese subjects (p=0.38); but was lower for rest images in obese subjects vs. non-obese subjects (p=0.03, Table 4).

Diagnostic Certainty:

Proportion of studies rated as definitely normal or definitely abnormal was similar between obese and non-obese subjects (p=0.35; Table 4).

Defect extent/severity:

Obese subjects had significantly lower median summed stress score and summed difference scores on PET than non-obese subjects (Table 4).

Radiation Exposure:

Effective radiation exposure was similar between obese and non-obese subjects (Table 4).

SUBGROUP ANALYSES:

Subgroup Analysis By Sex:

Among obese subjects, sensitivity of flurpiridaz PET was similar in both males and females (p=1.0); specificity was higher in females compared to males (p=0.019) (Table 5).

Table 5:

Subgroup Analysis: Comparison of Sensitivity and Specificity of Flurpiridaz-¹⁸F PET between Sex, Among Obese Subjects

Sensitivity (95% CI)			Specificity (95% CI)
Male N=192	Female N=106	P-value	Male N=192	Female N=106	P-value
76.3% (67.8%, 84.8%)	80.0% (62.5%, 97.5%)	1.00	58.9% (49.1%, 68.8%)	75.6% (66.5%, 84.7%)	0.019

Results for Majority Rule (agreement between 2 of 3 readers). Confidence intervals are from 2-sided 95% confidence intervals (CIs) based on the normal approximation to the binomial distribution. P-value based on Fisher's exact test.

Subgroup Analysis by use of CZT SPECT technology:

Overall, small number (n=67, 22.5%) of obese subjects underwent Tc-99m SPECT using newer generation cadmium-zinc-telluride scanners. PET was similar in sensitivity (ranging from 71% to 89% across readers) compared to SPECT (61% to 82% across readers) in this subgroup (Table 6). Specificity of PET (41% to 69% across readers) was non-inferior to SPECT (41% to 46% across readers) in this subgroup by 2 out of 3 readers.

Table 6:

Subgroup Analysis by use of newer generation SPECT technology (cadmium-zinc-telluride camera).

	Sensitivity (95% CI)				Specificity (95% CI)
Reader	PET MPI N=28	SPECT MPI N=28	DF: Difference	P-value	PET MPI N=39	SPECT MPI N=39	DF: Difference	P-value
Reader 1	20/28 (71.4%)	18/28 (64.3%)	7.1%		22/39 (56.4%)	18/39 (46.2%)	10.3%
	(54.7%, 88.2%)	(46.5%, 82.0%)	(−17.2%, 31.5%)	0.2398	(40.8%, 72.0%)	(30.5%, 61.8%)	(−10.3%, 30.8%)	0.0227
Reader 2	20/28 (71.4%)	17/28 (60.7%)	10.7%		27/39 (69.2%)	17/39 (43.6%)	25.6%
	(54.7%, 88.2%)	(42.6%, 78.8%)	(−13.9%, 35.4%)	0.1587	(54.7%, 83.7%)	(28.0%, 59.2%)	(4.7%, 46.6%)	0.0001
Reader 3	25/28 (89.3%)	23/28 (82.1%)	7.1%		16/39 (41.0%)	16/39 (41.0%)	0.0%
	(77.8%, 100.0%)	(68.0%, 96.3%)	(−11.1%, 25.4%)	0.2071	(25.6%, 56.5%)	(25.6%, 56.5%)	(−21.4%, 21.4%)	0.1790
Majority Rule	21/28 (75.0%)	20/28 (71.4%)	3.6%		21/39 (53.8%)	17/39 (43.6%)	10.3%
	(59.0%, 91.0%)	(54.7%, 88.2%)	(−19.6%, 26.7%)	0.3694	(38.2%, 69.5%)	(28.0%, 59.2%)	(−11.3%, 31.9%)	0.0298

SPECT= Single photon emission computed tomography, PET= Positron emission tomography, MPI= Myocardial perfusion imaging. Confidence intervals are from 2-sided 95% confidence intervals (CIs) based on the normal approximation to the binomial distribution. The hypothesis tests are 1-sided McNemar's tests with a significance level of 0.025 for sensitivity and Nam's RMLE method with a significance level of 0.025 of noninferiority for specificity (margin=0.1).

Subgroup analysis in extremely obese subjects (BMI ≥40 kg/m²):

PET sensitivity ranged from 81% to 95% across readers. SPECT sensitivity ranged from 52% to 86% across readers. Only reader 1 achieved statistical significance (Figure 7 and Supplemental Table 1). By majority rule, sensitivity of PET was not superior to SPECT in this subgroup (75% vs. 71.4%, p=0.37). Specificity of PET ranged from 47% to 88% across readers, and was non-inferior to SPECT (ranging from 44% to 47% across readers) by Reader 1, 2 and majority rule (Figure 7 and Supplemental Table 1). These findings should be interpreted cautiously given the limited sample size in the morbidly obese subgroup (n=57), which resulted in wide confidence intervals and limited statistical power.

Figure 7: Diagnostic Efficacy of Flurpiridaz-¹⁸F PET vs. ^99mTc-SPECT in extremely obese subjects (body mass index ≥40 kg/m2; N=57).

Error bars represent the width of the 95% confidence intervals.

DISCUSSION

In this pre-specified secondary analysis of the second Phase 3 flurpiridaz-¹⁸F multicenter trial, flurpiridaz-¹⁸F PET successfully met the primary study endpoint and demonstrated high diagnostic efficacy for the diagnosis of significant coronary artery disease in obese patients, a particularly challenging population to image. Sensitivity ranged from 70.1% to 88.0% and specificity from 53.6% to 74.0% across 3 readers, with both exceeding the pre-specified 60% lower confidence limit threshold (one-sided p<0.025) by the same 2 readers (Readers 1 and 2). Flurpiridaz-¹⁸F PET showed superior sensitivity than ^99mTc-SPECT MPI in obese subjects by 2 out of 3 readers (Reader 1: 72.6% vs. 60.7%, p=0.01; Reader 3: 88.0% vs. 74.4%, p=0.002), with non-inferior specificity by the same two readers (Reader 1: 68.0% vs. 61.3%; Reader 3: 53.6% vs. 50.8%; p<0.01 for non-inferiority, for both), noting that the primary endpoint was met by a partially overlapping reader pair (Readers 1 and 2). Diagnostic efficacy of flurpiridaz-¹⁸F PET did not decrease in obese subjects compared to non-obese subjects, underscoring the value of this PET tracer in a high-risk group where image quality is often compromised by soft-tissue attenuation. Overall diagnostic performance of flurpiridaz-¹⁸F was significantly better than ^99mTc SPECT MPI across all BMI categories.

Obesity has long been recognized as a technical hurdle for SPECT imaging. Attenuation artifacts due to increased soft tissue can lead to false-positive or false-negative findings, often necessitating additional imaging, 2-day protocols, higher radiation dose, all of which also complicate workflow and interpretation. PET has several technical advantages including superior spatial resolution and robust attenuation correction which help counteract soft tissue attenuation artifact, a common source of suboptimal image quality with SPECT in obese individuals. Previous studies on comparing diagnostic accuracy of PET vs. SPECT using older radiotracers have been small in size and retrospective.^10,19 In a study of 224 matched patients who underwent Rubidium-82 (⁸²Rb) PET and ^99mTc SPECT MPI by Bateman et. al,¹⁰ PET had greater diagnostic accuracy compared to SPECT in non-obese patients (87% vs. 70%). Among obese patients, diagnostic accuracy of PET was maintained at 85%, while that of SPECT decreased to 67%. Diagnostic confidence was also greater with PET compared to SPECT in both obese and non-obese subgroups in that study. In another study of 108 patients with extreme obesity, defined as BMI ≥40 kg/m²,¹⁹ Harnett and Hazra et. al. demonstrated a higher diagnostic accuracy at 86% with ⁸²Rb PET compared to 65% with ^99mTc SPECT. Hyafil et al. demonstrated that ⁸²Rb-PET provided significantly higher sensitivity than ^99mTc-SPECT with CZT cameras for detection of myocardial ischemia in women and overweight individuals (85% vs. 57%), despite equivalent specificity, suggesting that the sensitivity advantage of PET over SPECT in higher BMI patients persists even when the more technically advanced CZT platform is used as the comparator.²⁰ The overall diagnostic performance of flurpiridaz-¹⁸ F PET in obese subjects observed in the current study should be interpreted in context of the anatomic QCA gold standard at 50% threshold, high proportion of single-vessel disease, and absence of absolute flow quantification. These factors likely explain the lower observed accuracy compared to retrospective Rb-82 PET studies. PET imaging with flurpiridaz-¹⁸F offers all the intrinsic benefits of PET, including inherent attenuation correction, superior spatial resolution, along with other tracer specific characteristics such as lower positron range, high myocardial extraction fraction and higher count statistics, which likely account for the improved sensitivity observed in our obese subgroup, though the effective radiation dose at currently studied administered activities is modestly higher than that of other PET perfusion tracers. These technical advantages translate clinically into more reliable detection of perfusion defects, which is crucial for timely and accurate diagnosis in patients whose body habitus might otherwise obscure critical findings. In addition, the availability as a unit dose has the potential to significantly improve access to cardiac PET beyond high volume centers.

The specificity of ^99mTc SPECT was higher among obese subjects in our study than previously reported even in the extremely obese patients.¹⁹ This may, in part, reflect the use of cadmium zinc telluride SPECT scanners in a subset of patients (22.5% of obese subjects), which offers improved sensitivity and spatial resolution that may partially compensate for soft tissue attenuation in obese patients, even without dedicated attenuation correction. Additionally, the readers’ greater familiarity with reading SPECT images compared to PET images acquired with a new radiotracer could have contributed. Importantly, the use of a less stringent cut-off of 50% stenosis by QCA as the reference standard may have also inflated specificity estimates, since many lesions in the 50–70% range are not associated with true myocardial ischemia. A key methodological consideration is that all SPECT studies were read from non-attenuation corrected images. While this ensures inter-site consistency, it may have disadvantaged SPECT performance in obese patients where attenuation artifacts are most pronounced.The comparable diagnostic efficacy between obese and non-obese subgroups, and the superior diagnostic performance of PET compared to SPECT across BMI categories suggests that flurpiridaz-¹⁸F PET MPI can overcome the limitations imposed by increased body mass. In the per-vascular territory analyses, flurpiridaz-¹⁸F PET demonstrated significantly higher sensitivity than SPECT in both the LAD and RCA territories, findings that are consistent with the known susceptibility of non–attenuation-corrected SPECT to inferior and diaphragmatic attenuation artifacts during pharmacologic stress. Although certain radiology benefit managers currently limit cardiac PET MPI coverage to patients with a BMI ≥40 kg/m², our findings demonstrate that PET may represent a preferred strategy for suspected CAD evaluation irrespective of BMI, rather than being reserved exclusively for morbidly obese patients. Enhanced diagnostic confidence in this group may facilitate earlier intervention and more tailored therapeutic strategies, potentially leading to improved long-term outcomes, or enable avoidance of unnecessary catheterization. Future research should focus on the prognostic implications of PET MPI findings in obese patients to better define the role of this imaging modality in routine clinical practice.

Another notable finding of our study was the differences in defect extent and severity between obese and non-obese subjects and by modality. Among obese subjects, median summed stress scores were lower on PET compared to non-obese subjects, whereas no such difference was found on SPECT. This may be due to the lower prevalence of multivessel CAD among the obese patients in our study and the ability of PET to avoid attenuation-related artifacts that can exaggerate perfusion defect severity on SPECT. The lower median summed difference and rest scores observed with flurpiridaz-¹⁸F vs. SPECT in obese patients also likely reflect improved image resolution and attenuation correction, which reduce artifact-related false positives that can artificially elevate defect extent and severity on SPECT. The higher specificity of flurpiridaz-¹⁸F PET in obese women compared to men may reflect the improved image quality, attenuation correction and superior spatial resolution with PET that reduces false positive findings from breast and soft-tissue artifacts, while similar sensitivity suggests equivalent detection of true disease across sexes. However, this finding warrants cautious interpretation given the post-hoc nature and limited sample size (N=106); prospective evaluation is needed.

All readers underwent standardized training prior to blinded reads. Despite this, inter-reader variability was moderate for both modalities (PET Fleiss kappa=0.61; SPECT kappa=0.49). The higher agreement for PET compared to SPECT may reflect the superior image quality offered by this modality facilitating more consistent interpretation, even with a novel radiotracer. Intra-reader reproducibility was high for PET compared to SPECT, supporting robust interpretive reliability with PET even when using a novel tracer.There are several strengths of our study. The prospective nature of the study with blinded physician read of the scans and use of quantitative coronary angiography as the diagnostic gold-standard make the results less likely to be affected by interpretive bias. The relatively longer half-life of flurpiridaz¹⁸F compared to other commonly used PET radiotracers makes it feasible to be routinely combined with exercise treadmill stress, and 13% of all obese subjects in the study underwent exercise stress with PET, which can provide additional useful prognostic information. Contemporary SPECT technology which help with improving image quality especially among obese patients (such as cadmium-zinc-telluride scanners) was allowed. SPECT is commonly performed with higher Tc-99m doses in a 2-day protocol in obese individuals, leading to greater radiation exposure. In comparison, flurpiridaz-¹⁸F PET was associated with a significantly lower radiation dose in all patients, especially for obese patients who are more likely to undergo procedures with higher radiation doses due to image quality challenges associated with their body habitus.

The study findings should be interpreted in context of the following limitations. As all scans were interpreted visually by blinded readers per protocol, no automated perfusion quantitation was performed. PET also offers the ability to quantitate absolute myocardial flow and myocardial flow reserve, which could further improve its diagnostic accuracy,²¹ however, flow data were not available in our study. Quantitative coronary angiography with an anatomic gold standard of 50% stenosis was chosen as the end-point per FDA guidance, and invasive physiologic data such as fractional flow reserve and instantaneous wave-free ratio, which more closely correlate to physiologically significant stenoses were not available. Also, spatially relative perfusion defect assessment as done in this study has been shown to underestimate epicardial disease severity and extent in presence of balanced flow reduction such as in multivessel disease or coronary microvascular dysfunction.^22,23 However, given that both PET and SPECT were assessed using relative perfusion metrics, it is less likely to significantly affect the results of this study. All SPECT studies were read from non-attenuation corrected images; AC SPECT images were not available for analysis, precluding an AC-corrected comparison. The morbidly obese subgroup (BMI≥40; n=57) was underpowered for definitive conclusions. While the mixed pre-test probability range reflects a real-world referred population; results are most generalizable to intermediate pretest probability which accounted for >70% of the enrolled obese subjects.

Conclusions:

In summary, this study reinforces the value of flurpiridaz-¹⁸F PET MPI as a reliable and accurate diagnostic tool in obese patients with suspected CAD. By providing high image quality and maintaining diagnostic performance with increasing body habitus, flurpiridaz-¹⁸F PET MPI can help ensure that patients with obesity receive optimal diagnostic evaluation and management.

Supplementary Material

1

Supplemental Figure 1: Study Design.

Supplemental Figure 2: Area under Receiver Operating Characteristic (ROC) Curve generated from median summed difference scores across 3 readers for PET vs. SPECT, across BMI categories. ROC curves generated from median summed difference score (SDS) across 3 readers using logistic regression. AUC compared by DeLong et al. method.

Supplement table 1: Comparison of Diagnostic Efficacy (Sensitivity and Specificity) between Flurpiridaz- ¹⁸F Injection PET MPI and SPECT MPI by Reader and Body Mass Index ≥40 kg/m2 (QCA as Standard of Truth >=50% Stenosis Threshold)

Supplement Table 2: Inter-reader agreement in obese subjects

Supplement Table 3: Intra-reader agreement in obese subjects

Supplemental Table 4: Comparison of Diagnostic Efficacy (Sensitivity and Specificity) between Flurpiridaz (18F) PET MPI and SPECT MPI by Reader and Vascular Territory within obese subjects: (4.1) LAD (4.2) LCx (4.3) RCA