Economic and healthcare resource utilization assessments of PET imaging in Coronary Artery Disease diagnosis: a systematic review and discussion of opportunities for future economic evaluations

Abstract

Aims

This systematic literature review (SLR) consolidated economic and healthcare resource utilization (HCRU) evidence for positron emission tomography (PET) and single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) to inform future economic evaluations.

Materials and methods

An electronic search was conducted in MEDLINE, Embase, and Cochrane databases from 2012–2022. Economic and HCRU studies in adults who underwent PET- or SPECT-MPI for coronary artery disease (CAD) diagnosis were eligible. A qualitative methodological assessment of existing economic evaluations, HCRU, and downstream cardiac outcomes was completed. Exploratory meta-analyses of clinical outcomes were performed.

Results

The search yielded 13,439 results, with 71 records included. Economic evaluations and comparative clinical trials were limited in number and outcome types (HCRU, downstream cardiac outcomes, and diagnostic performance) assessed. No studies included all outcome types and only one economic evaluation linked diagnostic performance to HCRU. The meta-analyses of comparative studies demonstrated significantly higher rates of early- and late-invasive coronary angiography and revascularization for PET- compared to SPECT-MPI; however, the rate of repeat testing was lower with PET-MPI. The rate of acute myocardial infarction was lower, albeit non-significant with PET- vs. SPECT-MPI.

Limitations and conclusions

This SLR identified economic and HCRU evaluations following PET- and SPECT-MPI for CAD diagnosis and determined that existing studies do not capture all pertinent outcome parameters or link diagnostic performance to downstream HCRU and cardiac outcomes, thus, resulting in simplified assessments of CAD burden. A limitation of this work relates to heterogeneity in study designs, patient populations, and follow-up times of existing studies. Resultingly, it was challenging to pool data in meta-analyses. Overall, this work provides a foundation for the development of comprehensive economic models for PET- and SPECT-MPI in CAD diagnosis, which should link diagnostic outcomes to HCRU and downstream cardiac events to capture the full CAD scope.

Keywords:

• Coronary artery disease
• economic evaluations
• healthcare resource utilization
• meta-analysis
• myocardial perfusion imaging
• positron emission tomography
• single photon emission computed tomography
• systematic literature review

JEL Classification Codes:

• I00
• I
• I19
• I1
• I

Introduction

Coronary artery disease (CAD) represents a significant burden on healthcare systems globally, leading to substantial morbidity and mortality^1–3. CAD is associated with downstream cardiovascular events, extensive testing, and revascularization procedures, resulting in a substantial economic strain on healthcare systems^1,4,5. The incidence of CAD is expected to rise further due to the increasing prevalence of risk factors, including diabetes and obesity³.

The significant burden of CAD emphasizes the need for accurate and timely CAD diagnosis; however, there are several challenges in the diagnostic pathway such as suboptimal test accuracy and lack of access to novel technologies⁶. False negative (FN) and inconclusive results from cardiac diagnostic testing are common and result in missed diagnoses, subsequent cardiovascular events, hospitalization, and mortality. False positive (FP) results often require additional and sometimes unnecessary diagnostic follow-up testing and therapeutic invasive imaging, thereby escalating healthcare expenses^7–11.

An additional confounding challenge in diagnosing CAD is associated with the fact that current CAD screening and diagnosis professional guidelines are inconsistent, especially in difficult-to-image populations, including women and patients with obesity or diabetes^12–15. Accurately diagnosing CAD in these specific patient subgroups is often challenging due to patients’ anatomy and the mechanisms of these underlying diseases^2,16,17. For example, coronary microvascular disease commonly occurs in female, diabetic, and obese patients^18–23; however, traditional semi-quantitative perfusion diagnostic tests may not accurately capture the condition as a result of poor image quality^24–26. More specifically, women with large breasts and obese individuals experience diminished diagnostic sensitivity due to anterior wall attenuation artifacts, consequently increasing the risk of a FP diagnosis¹⁹. A large proportion (52%) of all diagnostic scans occur in the aforementioned subpopulations (women and patients with obesity or diabetes). Therefore, ensuring accurate diagnosis of these high-risk subgroups is imperative in facilitating timely and effective healthcare²⁷.

While various diagnostic testing options are available, nuclear-based myocardial perfusion imaging (MPI) modalities, such as single-photon emission computed tomography (SPECT) and positron emission tomography (PET), currently serve as gatekeepers for invasive coronary angiography (ICA), the standard of care for diagnosing CAD^28,29. SPECT is the most commonly used MPI modality³⁰; however, it has limitations^6,30. PET-MPI has emerged as an alternative due to its superior image quality^31,32, improved attenuation correction (AC)³³, lower radiation exposure³⁴, better myocardial extraction³¹, and the ability to measure myocardial blood flow and coronary reserve flow^31,35,36. Using PET-MPI to diagnose CAD leads to more accurate prediction of cardiac events, and better risk stratification in patients with suspected CAD^{21,25,37–39}.

Several clinical trials and large-scale meta-analyses have demonstrated improved diagnostic (i.e. sensitivity and specificity) and prognostic performance of PET- compared to SPECT-MPI in the general population^40–47. Improved sensitivity and specificity have also been observed in a head-to-head clinical trial of PET- compared to SPECT-MPI in subpopulations including women and patients with obesity and diabetes⁴⁷. Despite this evidence, PET-MPI use is currently limited due to low availability of PET machines, incongruent guidelines, reimbursement challenges, and a lack of evidence on the economic benefits compared to SPECT-MPI^38,48,49.

To date, no published systematic literature reviews (SLR) or meta-analyses have collated evidence on healthcare resource utilization (HCRU) and downstream clinical outcomes following PET- and SPECT-MPI use or captured economic evaluations for PET- compared to SPECT-MPI for diagnosing CAD. This SLR aimed to consolidate and analyze existing economic evaluations and reports of HCRU and downstream cardiac outcomes following PET- and SPECT-MPI in CAD. Specific objectives included a qualitative assessment of parameters and limitations of existing economic evaluations, as well as a comprehensive appraisal of comparative and single-arm trials reporting on HCRU and downstream cardiac outcomes to inform the development of future health economic evaluations of MPI in CAD.

Methods

Search strategy

A multiphase search strategy was conducted on 2 November 2022, and included searches of both electronic databases and manual references. The database search was performed using the Ovid search interface and included: MEDLINE, Embase, Cochrane CENTRAL, and Cochrane Database of Systematic Reviews. The search strategy was designed by an experienced information specialist and independently peer-reviewed by a second information specialist using the Peer Review for Electronic Search Strategies tool⁵⁰. The search was restricted to studies conducted in adults, humans, and written in the English language. See Supplementary Appendix 1 for a full description of the search strategy.

To ensure that all relevant studies were included, a manual search of the bibliographies of relevant SLRs, meta-analyses, and network meta-analyses identified through the database search was performed. To further complement bibliography searches, an additional targeted search was conducted in PubMed in March 2023 to confirm all relevant articles had been captured. For economic evaluations, any additional relevant studies identified during the bibliographic search of on-topic SLRs were included irrespective of their publication date to ensure that all available PET-MPI economic studies were captured and appraised.

Selection criteria and data extraction process

Studies were eligible for inclusion based on criteria established a priori (Table 1)⁵¹. Articles meeting the following criteria were included: (a) Studies conducted in adults with suspected and/or confirmed CAD; (b) Single-arm and/or comparative studies or economic evaluations involving the utilization of PET- and/or SPECT-MPI; (c) Studies comprising a minimum of 30 patients each; (d) Studies conducted in countries with comparable availability and utilization patterns of PET- and SPECT MPI technology; and (e) Articles published from 2012 to the present. Exclusion criteria included: (a) Narrative review articles, case study reports, opinion pieces, or conference abstracts; (b) Animal or phantom studies; and (c) Studies focusing on SPECT- or PET-MPI without addressing HCRU, downstream CAD events, or economics.

Table 1. Description of eligibility criteria.

Study Parameter	Description of Eligibility Criteria
Population	Patients with suspected or confirmed CAD
Intervention	PET-MPI used for the detection of CAD with CA or FFR as the reference/gold standard
Comparator	SPECT-MPI used for the detection of CAD with CA or FFR as the reference/gold standard
Outcomes	Economic (e.g. cost-effectiveness, utilization), healthcare resource use (e.g. repeated testing, coronary angiography, revascularization [CABG or PCI], hospitalization visits), cardiac events (e.g. MI, CD)
Study Design	All individual clinical studies (single arm, comparative, observational) and economic evaluations
Region	Studies conducted in United States, Canada, United Kingdom, France, Finland, Germany, Switzerland, Spain, Italy, Denmark, and Japan.
Date	2012 to Present

Abbreviations: CA, coronary angiography; CABG, coronary artery bypass graft; CAD, coronary artery disease; CD, cardiac death; ECHO, echocardiography; FFR, fractional flow reserve; MPI, myocardial perfusion imaging; PCI, percutaneous coronary intervention; PET, positron emission tomography; SPECT, single-photon emission computed tomography.

The study selection process was conducted using DistillerSR (Ottawa, Canada) SLR software in two levels⁵². Level one included screening the titles and abstracts of records retrieved from the initial database search for relevance by two independent reviewers to combat individual inclusion bias. Level two consisted of reviewing the full texts of studies deemed potentially relevant in duplicate. Any discrepancies or conflicts between the reviewers regarding record inclusion were resolved by a third impartial reviewer.

Data extraction was performed using a standardized Microsoft Excel template by a single reviewer and validated by a second reviewer. General study information (e.g. reference identification, country/region of study), key patient baseline characteristics (e.g. age, sex, prevalence of patients with suspected or confirmed CAD, body mass index [BMI], obesity, hypertension, dyslipidemia, diabetes, current or history of smoking), study design, diagnostic modality, and study protocol details were extracted from the included studies. In addition, outcomes of interest: (a) Any economic outcomes (e.g. analysis type [cost-effectiveness analyses [CEA], budget impact models, observational utilization], duration/time horizon, costs, quality adjusted life years) from economic evaluations of PET-MPI; (b) HCRU (coronary angiography [CA], coronary revascularization, repeat testing, and hospitalization rates); and (c) Downstream cardiac outcomes (myocardial infarction [MI] events, major adverse cardiovascular events [MACE], and cardiac death [CD]) were also extracted. Early CA referrals and revascularization were deemed to occur within 3 months post-MPI, while late CA and revascularization occurred after 3 months and within the study follow-up period. Revascularization procedures included percutaneous coronary intervention (PCI) and coronary artery bypass graft (CABG). Acute MI referred to reported nonfatal heart attacks occurring post-MPI. MACE encompassed any reported composite endpoints of MI, and CD or all-cause mortality, stroke, unstable anginal or hospitalization due to heart failure. Repeat testing referred to additional non-invasive imaging procedures following an initial diagnostic test, often conducted during subsequent evaluations or follow-up visits.

Statistical analysis

Meta-analyses were conducted using R (Version 4.0.5, “meta” package). For comparative studies, meta-analyses were performed by pooling hazards ratios of PET- and SPECT-MPI across studies and calculating the 95% confidence interval with the inverse variance method to evaluate HCRU and downstream cardiac event outcomes. Considering the limited number of head-to-head studies, a fixed-effects model was utilized to reduce the uncertainty that could be introduced with a random-effects model by inaccurately estimating between-study variance under a small sample size.

Additionally, single-arm meta-analyses were conducted to estimate the pooled rates of early and late CA and revascularization rates, MI, and CD for overall and categorized suspected CAD (<25%, 25–75%, >75%). Univariable meta-regression with random effects was performed to assess the impact of suspected CAD status on study outcomes. To mitigate the high variability in follow-up time across studies, logistic regression was used, utilizing the complementary log-log link function. By applying logistic regression, hazard estimates were obtained and utilized in the meta-analysis and meta-regression analyses. This standardized approach enabled reliable comparisons and combined analysis of data from different studies, regardless of follow-up durations. Bubble plots were generated to show the association between suspected CAD status and HCRU and event rates after adjustment for follow-up time. Studies with data points lying far from the regression line were considered outliers. When considering MI, late revascularization and late ICA, outliers were determined based on visual inspection, whereas for early revascularization and ICA, outliers were considered if the value was above quartile 3 + 1.5 interquartile range (IQR) or below quartile 1–1.5 IQR.

Results

Search results

The search yielded 13,439 results. Of this, 791 full texts were assessed for eligibility, resulting in 727 excluded records and 64 included records. Additionally, seven records were included through citation searching and targeted searches for a total of 71 included records. The results of the systematic search and study selection process are summarized in Figure 1.

Figure 1. Study identification, screening, and inclusion PRISMA diagram.

Economic evaluations

Eight PET-MPI economic evaluations were identified. Parameters and limitations of these evaluations were qualitatively assessed with significant variability observed across studies. Six records included both PET- and SPECT-MPI^53–58 and two records included PET-MPI only^59,60. Four economic evaluations were published within the past decade, while the remaining four were published between 17 and 28 years ago. Five of the evaluations were CEA^55–58,60 and the remaining three were observational or registry evaluations encompassing HCRU, cardiac outcomes, and associated costs^53,54,59.

The economic evaluations mainly focus on a general CAD patient population, with only two publications^56,57 providing women subgroup data. Time horizons ranged from 4-weeks to 30-years. There was also considerable variation in the types of parameters included in the evaluations. Few evaluations included both short-term (e.g. diagnostic characteristics, early HCRU) and long-term (e.g. MI events) outcomes. Two reports include HCRU only^59,60, two reports include both HCRU and downstream cardiac outcomes^53,54, and three records include both HCRU and diagnostic performance^55,57,58. Only one record linked diagnostic performance to HCRU, but it did not establish a connection between diagnostics and downstream cardiac outcomes⁵⁶. This heterogeneity in model structure and included parameters may drive the differences in costing results, where some studies demonstrate lower costs with PET-MPI vs. SPECT-MPI use^54,57,58,60 and others report higher costs^53,55,56,59. Detailed characteristics and outcomes of the economic evaluations are presented in Table 2.

Table 2. Common characteristics and variability of included economic studies.

Author, Year	Study Type Duration PET Radiotracer	Population	Outcome Parameters			Cost parameters	Limitations	Economic Results
			HCRU (CA, revasc.)	Cardiac Events (MI, CD)	Diagnostic performance Linkage to HCRU and cardiac events
Comparative Studies (PET-MPI vs. SPECT-MPI)
Mimouni, 2023^58	CEA 1-year ^82Rubidium	Suspected CAD Female: 46.1% Diabetes: 77% Grouped by pCAD: PTP > 15%	CA^a	Not included	TP, FP, TN, FN No link to HCRU and cardiac events	Human resources Supplies, drugs Intervention room ICA admission Re-admissions Radiology technician Senior physician	Limited follow-up time to document clinical events No subgroup analysis	Costs were lower with PET-MPI (€973) vs. SPECT-MPI (€1,192) ICER (PET-MPI vs. SPECT-MPI): −€2,730 per accurate diagnosis Driven by shorter protocol time and more scans per week with PET-MPI
Lorenzoni, 2019^55	CEA 90-days NR	Suspected CAD Female: 33–58% Diabetes: 14–28% pCAD: 34–49%	Not included^b	Not included	TP, FP, TN, FN No link to HCRU and cardiac events	Modality ICA admission Cost of a FN assumed twice the cost of ICA and FP cost of unnecessary ICA	Very limited follow-up No subgroup analysis Limited economic inputs	Costs (100 patients) were higher with PET-MPI (€117,722) vs. SPECT-MPI (€90,125) ICER (PET-MPI vs. SPECT-MPI): NR Driven by increased procedure cost of PET-MPI
Hlatky, 2014^53	Registry with HCRU and costs 2-years NR	Suspected CAD Female: 48–59% Diabetes: 16–41% pCAD: 47–52%	CA, revasc.^a	MI	No diagnostic outcomes	Modality CA, revasc. MI, death	No subgroup analysis	Costs were higher with PET-MPI ($6,647) vs. SPECT-MPI ($3,965) Driven by higher rates of subsequent revascularization procedures with PET-MPI
Merhige, 2007^54	Utilization, costs and outcomes study 1-year ^82Rubidium	Known CAD: 44-49% 46% female pCAD in known CAD: 46–47% pCAD in suspected CAD: 29–33%	CA, revasc.	MI, CD	FP No link to HCRU and cardiac events	Modality CA, revasc.	Limited follow-up time to document clinical events No subgroup analysis	Costs were lower with PET-MPI ($4,110) vs. SPECT-MPI ($5,937) Driven by reduced CABG rates (>50% lower) with PET-MPI
Patterson, 1995^57	CEA 10-years NR	Men and women, various pCAD Women subgroup analysis provided	CA, revasc.	Rate of complications and mortality	Sensitivity, specificity No link to HCRU and cardiac events	Modality CA, revasc Mortality Bayes’ theorem was used to calculate the rate of complication as non-fatal MI in 20% of FN patients	Medically unnecessary CA was not considered for FP patients	Costs were lower with PET-MPI vs. SPECT-MPI for all pCAD^c (10%: $2,710 vs. $2,940; 50%: $5,020 vs. $5,400; 100%: $7,850 vs. $8,450) Cost/QALY^c (PET-MPI): $10,000 (pCAD 10%); $3,550 (pCAD 50%); $2,790 (pCAD 100%) Cost/QALY^c (SPECT-MPI): $12,700 (pCAD 10%); $4,600 (pCAD 50%); $3,460 (pCAD 100%) Driven by higher sensitivity and specificity with PET-MPI
Garber, 1999^56	CEA 30-years ^82Rubidium	Base: men, 55 years, pCAD: 50% Sensitivity: men 45 and 65 years, women 45, 55, and 65 years, pCAD: 25–75% Women subgroup analysis provided	CA, revasc.	MI	FP, FN, TP, TN TP: cost of CA TN: appropriate avoidance of CA FN: conservative management, no revasc. FP: unnecessary CA	Modality CA, revasc MI	Downstream cardiac events were not considered for FN patients	Costs were higher with PET-MPI vs. SPECT-MPI in men ($35,093 vs. $34,047) and women ($41,551 vs. $40,510) ICER (PET-MPI vs. SPECT-MPI): $523,000/QALY (men); $520,500/QALY (women) Driven by higher modality cost and increased angiography with PET-MPI
Single Arm PET-MPI Studies
Ma, 2021^59	Retrospective observational study 1-year NR	Suspected CAD Female: 58–59% Diabetes: 32–33%	CA, revasc.	Not included	No diagnostic outcomes	Modality CA, revasc Hospitalization costs (e.g. all-cause, cardiac, outpatient, emergency)	Limited follow-up time to document clinical events No subgroup analysis PET-MPI had more high-risk patients (could explain the higher utilization rates and costs)	Costs were higher with PET-MPI ($2,358.20) vs. conventional imaging ($1,774.30) Driven by higher utilization (hospitalizations, PCI) with PET-MPI
Siegrist, 2008^60	CEA 4-weeks ^13N-ammonia	Known CAD: 79% Female: 28%	CA, revasc.	Not included	No diagnostic outcomes	Modality CA, revasc Medical management	Limited follow-up time to document clinical events No subgroup analysis	Costs (total) were lower with PET-MPI (€152,450) vs. standard of care (€170,360) ICER: NR Driven by reduced diagnostic angiography with PET-MPI

^a Refer to Table 3 for specific reported rates.

^b Coronary revascularization derived from ENVINCI data set as an alternate measure of effectiveness.

^c Data digitized from publication.

Abbreviations: CA, coronary angiography; CABG, coronary artery bypass grafting; CAD, coronary artery disease; CD, cardiac death; CEA, cost-effectiveness analysis; FN, false negative; FP, false positive; HCRU, healthcare resource utilization; ICA, invasive coronary angiography; ICER, incremental cost-effectiveness ratio; MI, myocardial infarction; MPI, myocardial perfusion imaging; pCAD, pretest likelihood of coronary artery disease; PCI, percutaneous coronary intervention; PET, positron emission tomography; PTP, pretest probability; QALY, quality-adjusted life-year; revasc., revascularization; SPECT, single-photon emission computed tomography; TN, true negative; TP, true positive.

Comparative studies

Six comparative PET- and SPECT-MPI studies were included, with studies demonstrating variability in reported outcomes (Figure 2). Two studies reported HCRU and cardiac outcomes in PET- compared to SPECT-MPI^53,61 and two studies evaluated downstream HCRU only between the two MPI modalities^62,63. Only one study reported diagnostic performance and downstream HCRU for PET- vs. SPECT-MPI⁵⁸. No comparative studies reported all three outcomes. Mimouni et al.⁵⁸ was the only comparative study reporting on the diagnostic performance of PET- and SPECT-MPI. The results demonstrated that PET-MPI had a significantly higher diagnostic accuracy compared to SPECT-MPI (.87 vs. .78; p = .048), which correlated with increased utilization of ICA procedures.

Figure 2. Variability in reported outcomes of interest in comparative PET- vs. SPECT-MPI studies.

Table 3 summarizes key characteristics in each comparative study, and the reported rates of clinical HCRU and cardiac outcomes. The total population sample across studies consisted of 88,371 patients who underwent PET-MPI and 1,945,625 patients who underwent SPECT-MPI. Patel et al.⁶¹ was the only study that included an 100% confirmed CAD population. Conversely, two studies limited their inclusion criteria to patients with suspected CAD^53,58. The remaining studies had a mixed population of known or suspected CAD. Female patients comprised 37–69% of the total PET-MPI population and 33–69% of the total SPECT-MPI population. The mean age of patients ranged between 59–66 years old. Diabetes was prevalent in 14–77% of patients. Most patients also exhibited moderate to high mean BMI (28.8–31.8 kg/m²) or were classified with obesity.

Table 3. Characteristics of PET- vs. SPECT-MPI studies and outcomes.

Author, Year	Article type	Modality	N	Key Patient Characteristics	Suspected (or Known) CAD	FU Time (months)	Coronary Angiography (%)		Coronary Revascularization (%)		Repeat Testing (%)	MACE^a (%)	Acute MI^a (%)
							Early ≤90 d	Late^a	Early ≤90 d	Late^a
Mimouni, 2023^58	Retrospective cost‑effectiveness using prospectively collected data during the RUBIS trial	PET	308	Age: 61 Female: 46.1% DM: 76.9% Mean BMI: 31.8	100% (0 %)	12	–	15	–	–	–	–	–
		CZT-SPECT	308	Age: 61 Female: 46.1% DM: 76.9% Mean BMI: 31.8		12	–	10	–	–	–	–	–
Ananthasubramaniam, 2022^62	Retrospective, analysis of IBM MarketScan commercial and Medicare claims data	PET	224	Age: 59 Female: 58.5% DM: 50.7%	–	–	56.7	–	19.9	–	10.0	–	–
		SPECT	12,905	Age: 59 Female: 52.8% DM: 38%	–	–	70.7	–	27.3	–	14.1	–	–
Patel, 2019^61	Prospective single-center randomized comparison of clinical effectiveness	PET	161	Age: 66.5 Female: 37.3% DM: 26.7% Mean BMI: 28.8	0% (100%)	12	20.5	28.6	11.2	15.5	–	–	0.0
		CZT-SPECT	161	Age: 66.1 Female: 32.9% DM: 28.6% Mean BMI: 29.2		12	15.5	28	9.9	14.9	–	–	1.2
Knight, 2018^123	Retrospective analysis of prospectively collected data	PET	7,130	Age: 64.5 Female: 44.9% DM: 29.8% Mean BMI: 31.9	59.5% (40.5%)	12	13.9^b*	–	7.9^b*	–	–	5.7	–
		SPECT	6,686	Age: 63.8 Female: 47% DM: 29.2% Mean BMI: 31.2	63.3% (36.7%)	12	9.3^b*	–	4.3^b*	1.6	–	5.2	–
Hlatky, 2014^53	Retrospective economic evaluation using prospectively collected data during the SPARC trial	PET	548	Age: 63 Female: 59% DM: 41% Mean PTP: 47%	100% (0%)	24	11.0	15	2–5^d	2–6^d	–	–	1.1
		SPECT	565	Age: 60 Female: 51% DM: 30% Mean PTP: 51%		24	4.0	7	0.4–1^d	0.4–2^d	–	–	1.2
Pelletier-Galarneau, 2023^63	Real-world, retrospective study of de-identified claims and electronic health records	PET	83,100^e	Age: - (matched based on age) Female: 41–69% Obesity: 42–61% DM: 14–54%	45.1% (54.9%)	24	12.8	–	–	–	2.2	–	–
		SPECT	1,951,500^e		58.4% (41.6%)	24	12.4	–	–	–	3.4	–	–

^a Follow-up time as specified in previous column.

^b ≤60 days.

^c n = 6,777.

^d Includes CABG and PCI rates.

^e Calculated based on 2.5 million patients who underwent diagnostic imaging, 77% who underwent SPECT and 3.2% who underwent PET.

* p <.0001.

Abbreviations: -, not reported; ACM, all-cause mortality; BMI, body-mass index (kg/m²); CAD, coronary artery disease; CZT, cadmium zinc telluride; d, days; DM, diabetes mellitus; FU, follow-up; MACE, major adverse cardiovascular events; MI, myocardial infarction; PET, positron emission tomography; PTP, pre-test probability; SPECT, single-photon emission computed tomography.

The meta-analysis of comparative studies demonstrated that the rate of early and late ICA, as well as revascularization, was significantly higher for PET- compared to SPECT-MPI; however, the rate of repeat testing was significantly lower with PET-MPI. The rate of acute MI was lower, albeit non-significant, with PET- vs. SPECT-MPI (Table 4).

Table 4. Meta-analysis of PET- vs. SPECT-MPI studies.

Outcome	HR	95% CI	p-value
Early Coronary Angiography (<3 months)	1.05	1.03–1.07	<.0001
Late Coronary Angiography	1.58	1.25–2.01	.0002
Early Revascularization (<3 months)	1.58	1.39–1.79	<.0001
Late Revascularization	1.78	1.17–2.70	.0067
Acute MI	0.76	0.28–2.05	.5896
Repeat Testing (e.g. repeat SPECT-MPI, ECHO, etc.)	0.64	0.61–.67	<.0001

Abbreviations: ECHO, echocardiography; HR, hazard ratio; MI, myocardial infarction; MPI, myocardial perfusion imaging; PET, positron emission tomography; SPECT, single-photon emission computed tomography.

Single-arm studies

A total of 61 single-arm studies were included. Of these, 62% were SPECT-MPI^64–100 and 38% were PET-MPI studies^{21,25,59,61,101–119}. Population baseline characteristics were highly heterogeneous across the included studies. Twenty studies specifically limited their inclusion criteria to suspected CAD patients^{25,59,67,74–77,83,87,92,94–96,100,101,107,109–111,117}. Conversely, the remaining studies included a mixed patient population with confirmed or suspected CAD. The proportion of female patients ranged from 3% to 100%. The patients also exhibited various CAD comorbidities, including diabetes and obesity. Follow-up time varied significantly, ranging from 2–102 months. Supplementary Table S1 summarizes the key patient characteristics and outcomes of each single-arm study.

Figure 3 depicts rates of HCRU and CAD events in each study by the proportion of patients with suspected CAD, where ‘0′ indicates 0% suspected CAD and ‘1′ indicates 100% suspected CAD. Significant outliers have been excluded from the analysis. HCRU increases with increasing prevalence of confirmed CAD in the population in patients imaged with PET- or SPECT-MPI. See Supplementary Figure S1 for plots with included outlier studies.

Figure 3. Annualized rates of HCRU and CAD events adjusted for time in single-arm MPI studies (excluding outlier studies).

Pooled rates which controlled for variability in follow-up time and suspected CAD status (i.e. 0–100%) are presented in Table 5 excluding outlier studies and Supplementary Table S2, including outlier studies. Findings demonstrated that studies comprising over a 75% suspected CAD population exhibited significantly lower annualized rates of HCRU and cardiac events compared to studies with a population less than 25% suspected CAD. These findings are independent of the exclusion of outliers. Annualized rates were 7.7% and 7.3% for early and late ICA, 2.4% and 1.6% for early and late revascularization, and 0.5% for cardiac death for the >75% suspected CAD subgroup compared to annualized rates of 17.3% and 15.2% for early and late ICA, and 10.6% and 15.2% for early and late revascularization in the <25% suspected CAD subgroup. In all patients, regardless of suspected CAD status, the pooled HCRU and CAD event annualized rates were 9.8% and 9.6% for early and late ICA, 3.7% and 2% for early and late revascularization, 0.6% for MI, and 0.8% for cardiac death.

Table 5. Pooled annualized rates (%) of HCRU and CAD events across MPI studies adjusted for time (excluding outlier studies).

	Early ICA % (n)	Early Revasc. % (n)	Late ICA % (n)	Late Revasc. % (n)	MI % (n)	Cardiac Death % (n)
<25% suspected CAD	17.3 (3)	10.6 (2)	28.3 (2)	15.2 (2)	1.2 (2)	NA
25–75% suspected CAD	9.7 (7)	5.1 (10)	NA	1.9 (21)	0.8 (20)	1.2 (13)
>75% suspected CAD	7.7 (6)	2.4 (7)	7.3 (8)	1.6 (9)	0.5 (12)	0.5 (11)
General Population	9.8 (16)	3.7 (19)	9.6 (10)	2.0 (32)	0.6 (34)	0.8 (24)

Note: “n” indicates the number of studies included in pooled analysis, including all studies with available outcome and suspected CAD data. Bolded percentage indicates that the annualized rate of HCRU is lower than that of <25% suspected CAD subgroup (25–75% suspected CAD subgroup for cardiac mortality), and the results were statistically significant in meta-regression. Early is defined as <30 days.

Abbreviations: CAD, coronary artery disease; HCRU, health care resource utilization; ICA, invasive coronary angiography; MI, myocardial infarction; MPI, myocardial perfusion imaging; revasc., revascularization.

Discussion

This SLR sought to identify and consolidate existing economic evaluations and investigations of HCRU and downstream cardiac outcomes following PET- and SPECT-MPI for CAD diagnosis, addressing a key evidence gap. The economic and clinical evaluations reviewed in this SLR do not capture all pertinent outcome parameters (i.e. diagnostic performance, HCRU, and cardiac events) or link diagnostic performance to downstream HCRU and cardiac outcomes, thus, resulting in simplified assessments of CAD burden. This work builds a framework to inform the development of future economic evaluations of PET- and SPECT-MPI and yields HCRU and cardiac event estimates that can be used in future economic models.

Economic evaluations

The existing economic evaluations for PET-MPI are limited in number, are heterogeneous, and possess several design limitations that hinder consensus for comparative cost effectiveness and affordability of PET- relative to SPECT-MPI. Firstly, the economic evaluation time horizons are highly variable, and often the follow-up time is short (up to 1-year). Several of the economic evaluations did not include downstream cardiac outcomes, which may be due to the short time horizons that are insufficient to capture the occurrence of downstream health outcomes. Additionally, PET-MPI has been shown to have a higher rate of early referrals to CA and revascularizations which could be attributed to the higher sensitivity and specificity of PET- compared to SPECT-MPI⁵⁷. As a result, evaluations that are short and only capture early HCRU may yield higher incurred costs for PET- than SPECT-MPI; however, they are not representative of the entire resource utilization landscape.

Although a few of the economic evaluations used diagnostic performance parameters to inform cost^55,57,58, only one evaluation published over two decades ago included assumptions directly linking diagnostic performance to HCRU and cardiac events⁵⁶. Garber and Solomon⁵⁶ linked true positive (TP) results to necessary CA, true negative (TN) results to the appropriate avoidance of CA, FN to conservative management, and FP to unnecessary CA. However, they did not consider the downstream cardiac events that may occur from a FN test result and associated costs. Garber and Solomon⁵⁶ also relied on sensitivity and specificity values derived from published literature, which may be influenced by publication bias. Similar challenges have been observed in other economic evaluations for diagnostic tests because of the indirect nature of the diagnostic testing impact on patient outcomes¹²⁰. Similarly, simplified methods to estimate diagnostic test accuracy are often used to inform economic models, which may present challenges for decision makers^121,122. The variable nature of parameters included in existing analyses emphasizes the need for a comprehensive economic model that links diagnostic performance to downstream health outcomes.

Another crucial aspect to consider in economic evaluations is the selection of appropriate cost parameters. Economic analyses that only included modality costs tended to favor SPECT- over PET-MPI due to the omission of pivotal high-cost components (i.e. HCRU and cardiac events)^53,55,56. However, in the most recent economic model developed by Mimouni et al.⁵⁸, several cost parameters, including human resources, materials (supplies and drugs), intervention rooms, and technician and physician time were considered. The inclusion of personnel time and intervention room costing increases the cost of 2-day SPECT-MPI procedures, which is not considered in most economic evaluations. This variability poses a substantial challenge when attempting to conduct a comprehensive cost-effectiveness comparison between PET- and SPECT-MPI, comparing economic results of existing evaluations, and trying to apply results to other economic landscapes.

Of note, the existing economic evaluations focus only on a general CAD population and lack investigations into subpopulations with diagnostic imaging challenges and a high CAD burden such as female, obese, and diabetic patients^19,24,25,37. It is suspected that there will be differences in HCRU, clinical event occurrence, and costs among these subpopulations. Additionally, as the prevalence of diabetes and obesity continue to rise, consideration of these high-risk subpopulations will be imperative to accurately realize these diagnoses, otherwise an increase in adverse clinical outcomes will further contribute to an increase in HCRU.

Furthermore, the lack of recently published economic analyses for PET-MPI underscores the need for new comprehensive economic models. PET- and SPECT-MPI technology has changed overtime; therefore, economic models published between 17–28 years ago may not accurately reflect the costs of current technology. Utilizing publications that are over two decades old to inform current health care decisions is suboptimal and reinforces the need to critically evaluate data and create a holistic and up-to-date model.

Future economic models for imaging diagnostics should address the abovementioned limitations to enhance the validity and applicability of their results. Specifically, they should consider the link between diagnostic performance and downstream outcomes namely HCRU and cardiac events, both of which are costly. In addition, future economic models comparing PET- and SPECT-MPI should consider subpopulations where underlying disease mechanisms and anatomical imaging challenges limit the performance of current imaging diagnostics. The time horizon should also be of sufficient length to capture early and downstream medical costs including referrals to CA, revascularization, hospitalizations, intensive care, and follow-up visits due to MI or MACE events. Lastly, future models should include comparisons within imaging modality tracers and address novel PET-MPI radiotracers.

Comparative studies

There are limited comparative reports of HCRU and downstream cardiac outcomes following PET- and SPECT-MPI. The existing studies have several notable limitations in their design and sources of heterogeneity. Firstly, most studies were not designed to assess the long-term impact of diagnostic testing on clinical outcomes. Only one comparative study⁶¹ was designed to assess whether diagnostic accuracy impacts downstream management; however, the follow-up time was likely too short to fully capture the impact of clinical event occurrence. All patients in this study also had a history of CAD and were medically managed; therefore, the full benefit of PET as a diagnostic modality may not be shown. Follow-up time also varied among the included studies, ranging from 12 to 24 months. Differences in follow-up duration can impact the observed outcomes and limit the comparability across studies.

A significant source of heterogeneity is the lack of randomized study designs among the comparative studies. Patel et al.⁶¹ was the only study with randomization, while others were retrospective observational designs or real word evidence claims studies. This lack of randomization introduces challenges when comparing and generalizing results. Additionally, it is often difficult to match study arms on important patient characteristics. Patel et al.⁶¹ was randomized and the reported baseline characteristics were well-balanced between the PET- and SPECT-MPI arms; however, several of the non-randomized comparative studies noted differences in baseline characteristics. Knight et al.¹²³, Hlatky et al.⁵³, and Ananthasubramaniam et al.⁶² showed that the PET-MPI population was older, and had a higher prevalence of prognostic CAD factors, including diabetes, higher BMI, chronic kidney disease, hypertension, dyslipidemia, gastroesophageal reflux disease, cardiac arrhythmias, family history of CAD, and medication history than the SPECT-MPI population. Knight et al.¹²³ also found that the PET-MPI population had a higher grade of obstructive CAD and multivessel disease, which reflected in the higher early referrals to CA and revascularization within the 60-day follow-up duration of the study. The differences in baseline characteristics among these studies demonstrate that the PET-MPI population is intrinsically sicker compared to the SPECT-MPI population, presenting challenges in the direct comparison of outcomes between PET- and SPECT-MPI.

Another source of heterogeneity was the variability in MPI protocol among the included studies, including differences and uncertainties in the type of stress used (exercise or pharmacologic), perfusion tracers, type of imaging cameras (cadmium-zinc telluride [CZT] SPECT camera or conventional SPECT camera), and the incorporation of AC techniques. For instance, Knight et al.¹²³ and Patel et al.⁶¹ also utilized AC in all SPECT-MPI scans. Conversely, it is not clear whether AC was implemented in the other included studies. Moreover, Patel et al.⁶¹ and Mimouni et al.⁵⁸ utilized CZT cameras for imaging assessments, whereas other studies used conventional SPECT cameras. CZT cameras have been shown to improve image quality and diagnostic performance for SPECT-MPI, which may result in reduced downstream outcomes compared to studies utilizing traditional SPECT cameras¹²⁴. These technical disparities can affect the diagnostic performance of the imaging tests and thus impact the observed outcomes.

Single-arm studies

The single-arm analyses utilized a large dataset to inform pooled estimates of HCRU with various CAD populations. The results suggest that there are lower rates of HCRU and cardiac events in studies with 100% suspected CAD patients compared to studies with 100% confirmed CAD patients. The pooled estimates can be used to inform future economic modeling of PET- and SPECT-MPI for CAD diagnosis. The results presented herein highlight that economic models should differentiate between suspected and confirmed CAD patients as the rates of HCRU and cardiac events differ and consider both early- and late-HCRU.

The high heterogeneity among the included single-arm studies creates challenges when pooling data to determine if HCRU differs across studies in various subgroups. There are several sources of heterogeneity in the single-arm studies including variable population baseline characteristics, study designs, and follow-up times. There are also MPI protocol and technology differences among studies, including type of stress test used, camera type, perfusion tracer type, and the use of AC techniques. A future comparative study that assesses more closely how HCRU and cardiac event rates differ by higher risk subpopulations (e.g. diabetes or obesity) is required.

Strengths and limitations

This is the first SLR to comprehensively synthesize economic, HCRU, and cardiac outcomes associated with PET- and SPECT-MPI imaging for CAD diagnosis, which advances the scientific literature for this disease landscape. Additionally, a thorough examination of existing studies was conducted to understand the extent of heterogeneity and limitations in study design. This work will inform the development of future economic evaluations of PET- and SPECT-MPI and add to the value framework of appropriate and timely utilization of CAD diagnosis. A limitation to note is that PCI and CABG rates were treated as mutually exclusive events and were added together into an aggregate value for early- and late-revascularization when pooled. Additionally, given the heterogeneity in study designs, patient characteristics, and follow-up times, pooling the data in meta-analyses was challenging. Caution is warranted when interpreting the findings.

Conclusion

This SLR collated economic and HCRU evidence following PET- and SPECT-MPI for CAD diagnosis and highlights important limitations and challenges of pooling HCRU and clinical event data. The economic evaluations comparing PET-MPI and SPECT-MPI in populations with known and/or suspected CAD are also limited and highly variable in nature. These evaluations typically have short time horizons, resulting in an incomplete assessment of the costs associated with downstream HCRU outcomes. Despite its higher procedure cost, PET-MPI has the potential to be cost-effective based on its improved diagnostic performance compared to SPECT-MPI; however, this economic potential is underestimated in current economic model frameworks. PET-MPI can facilitate more targeted and effective interventions by accurately identifying and stratifying patients with CAD, improving patient outcomes, and potentially reducing long-term healthcare costs. Future economic models of PET- and SPECT-MPI should consider the inclusion of diagnostic performance and link to HCRU and downstream cardiac events to capture the full resource utilization landscape.

Transparency

Declaration of funding

This work was funded by GE HealthCare.

Declaration of financial/other relationships

This work received funding from GE HealthCare. GE HealthCare contracted EVERSANA to complete this work. NF, SP, LA, AWM, DW, and AOR are employees of EVERSANA. ES and AC are employees of GE HealthCare.

Author contributions

Concept and design: NF, SP, AC; Acquisition of data: LA, AWM, DW, AOR; Analysis and interpretation of data: NF, SP, LA, AWM, DW, AOR, ES; Drafting of the manuscript: NF, SP, LA, AWM, DW, AOR; Critical revision of the paper for important intellectual content: NF, SP, ES, AC; Obtaining funding: ES, AC; Administrative, technical, or logistic support: SP, ES, AC; Supervision: NF, AC.

Role of the funder/sponsor

ES and AC supported the development of this work and are employees of GE HealthCare (the Funder). The publication of study results was not contingent on the sponsor’s approval or censorship of the manuscript.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Previous presentations

Ferko N, Priest S, Ramirez AO, et al. Challenges and Opportunities with Quantifying Comprehensive Health Economics of Diagnostic Imaging Modalities Using Radiopharmaceuticals: A Case Study in Coronary Artery Disease. Poster session presented at: The Professional Society for Health Economics and Outcomes Research (ISPOR); 2023 May 7–10; Boston, MA.

Acknowledgement

Data and statistical analysis support was provided by Tim Disher of EVERSANA (Burlington, ON, Canada). Scientific advice was provided by Francois Tranquart of GE HealthCare (Chalfont St Giles, UK).