WHAT IS ALREADY KNOWN ON THIS TOPIC
-
The diagnostic performance of various non-invasive imaging techniques for detecting coronary artery disease (CAD) in patients without prior cardiac history is well documented.
-
The effectiveness of these imaging modalities in patients with a history of myocardial infarction or percutaneous coronary intervention (prior CAD) has not been thoroughly investigated, leading to potential gaps in clinical decision-making for this high-risk group.
WHAT THIS STUDY ADDS
-
This comprehensive meta-analysis compares the diagnostic accuracy of coronary CT angiography (CCTA), CCTA combined with CT perfusion (CCTA+CTP), cardiac MRI (CMR) and single-photon emission CT (SPECT) in patients with prior CAD.
-
It highlights that while CCTA, CCTA+CTP and CMR show high sensitivity for detecting obstructive CAD, SPECT demonstrates significantly lower sensitivity, indicating limitations in its clinical utility for this patient population.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
-
The findings could inform clinical guidelines and policies, promoting the use of more effective imaging techniques like CCTA, CCTA+CTP and CMR for high-risk patients, ultimately optimising resource allocation and care quality.
Introduction
Non-invasive cardiac imaging has become the diagnostic cornerstone for the assessment of symptomatic patients with suspected obstructive coronary artery disease (CAD).1 2 Recent advances in hardware technology have expanded the armamentarium of non-invasive diagnostic imaging techniques. For instance, coronary CT angiography (CCTA) as stand-alone imaging or using a hybrid approach with CT perfusion (CTP) has evolved from an anatomical technique into a comprehensive tool for evaluating CAD severity and its functional impact.3–5 Intuitively, functional imaging modalities such as single-photon emission CT (SPECT), positron emission tomography (PET), stress echography (SE) and cardiac MRI (CMR) have been preferred as first-line tests in patients with a cardiac history. Current guidelines emphasise local availability and expertise, without recommending one functional test over the other.2 Non-invasive imaging of patients with prior CAD is generally considered more challenging. Anatomical scan results are adversely impacted by stent artefacts in patient with prior percutaneous coronary intervention (PCI). Moreover, the impact of myocardial scar on tracer kinetics in patients with prior myocardial infarction (MI) remains uncertain, given the heterogeneous nature of myocardial scar, which includes islands of viable myocardium.6 In addition, patients with prior CAD have, by definition, proven atherosclerosis, a condition associated with microvascular disease, which further complicates the interpretation of non-invasive imaging.7 Therefore, the majority of studies evaluating the diagnostic value of non-invasive imaging methods have been performed in patients without a cardiac history.8–10 As such, there is a paucity of data on the performance of non-invasive imaging in patients with prior CAD, which represent the majority of patients in cardiac outpatient clinics.11 The PACIFIC 2 (Functional stress imaging to predict abnormal coronary fractional flow reserve: the PACIFIC 2 study) was the first trial to assess the diagnostic performance of non-invasive cardiac imaging, in a true head-to-head fashion, in symptomatic patients with a cardiac history. Notably, myocardial perfusion imaging was only moderately accurate for diagnosing haemodynamically significant CAD as referenced by invasive fractional flow reserve (FFR).12 The authors questioned whether non-invasive imaging aids clinical decision-making in this challenging population. Therefore, this meta-analysis aims to assess the diagnostic value of various non-invasive cardiac imaging methods in symptomatic patients with prior CAD.
Methods
This meta-analysis was reported in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.13 The protocol was published in the PROSPERO database (CRD42022322348). PubMed, Embase, Web of Science and the Cochrane Library were systematically searched, spanning the period from 2005 until March 2022 for articles in English. A manual reference check was performed to identify potential missed studies by our search strategy. Relevant studies between March and September 2022 were manually added. The concise search syntax is presented in table 1 and the full syntax in online supplemental table 1. Reports that employed duplicative cohorts or overlapping data were excluded and the study with the largest population was included. Screening and quality assessment was performed by two independent reviewers (RJ and JD) according to the revised version of the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2).14 Disagreements were resolved by consensus or a third reviewer (ID). For all studies identified by the search that fulfilled the inclusion criteria and comprised ≥20 patients with prior CAD, defined as a history of MI or PCI, the first and corresponding authors were approached and requested twice to contribute by retrieving 2×2 tables for patients with a history of CAD from their respective studies. After a response rate of 4% (6/162), we used the published data and included two types of studies: (1) Studies for which a 2×2 table specified for patients with prior CAD was deductible. Those included populations consist entirely of patients with prior CAD. (2) If a 2×2 table exclusively for patients with prior CAD was not deductible from the published data we included studies with populations that consisted of at least 50% patients with prior CAD. As such, in those studies, the included population was a mix of patients with (≥50%) and without (<50%) prior CAD. Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.
Concise search syntax
Study eligibility
The inclusion criteria were as follows: (1) the diagnostic performance of non-invasive imaging to detect obstructive CAD was studied; (2) invasive coronary angiography (ICA) or invasive FFR served as reference; (3) the data were prospectively collected; (4) all patients underwent ICA, irrespective of non-invasive imaging results; (5) the interpretation of the non-invasive imaging was blinded to ICA results; and (6) a 2×2 table specifically for patients with prior CAD was deductible. Alternatively, studies of which patients with prior CAD comprised ≥50% of the study population were also included. To ensure applicability, studies published before 2005 were excluded. CT studies were excluded if used hardware was <64 slices, if unevaluable segments were censored from analysis or if results reported only on stented segments leaving the remaining vasculature unanalysed. Patients with known coronary artery bypass grafting were, if separately reported, excluded to increase homogeneity.
Data collection
Data extraction was performed by two reviewers (RJ and JD). For each included study the following were collected: number of patients with prior CAD split for MI and PCI, hardware, criteria for abnormal imaging results and criteria for defining obstructive CAD on ICA. If the amount of patients grouped as prior CAD was not reported we considered the highest percentage of patients with either prior PCI or MI as percentage of patients with prior CAD, assuming maximal overlap between PCI and MI. True and false positive, and true and false negative numbers were extracted from the articles and summarised in a 2×2 contingency table. Subsequently, analyses were stratified for CCTA, CCTA+CTP hybrid imaging, CT-FFR, CMR, SPECT, PET and SE. If a study compared multiple imaging modalities to ICA, each modality was evaluated separately. If available, FFR was chosen as endpoint. If FFR was not available % diameter stenosis (DS) was chosen as reference with 70% DS preferred over 50% DS.
Statistical analysis
All analyses were performed at the per-patient level. Modalities that included a minimum of three separate studies were included for analysis. Intraobserver agreement between reviewers with regard to the quality assessment and study selection was assessed by the Cohen’s kappa test. Based on the results from the 2×2 tables, pooled prevalence, sensitivity, specificity, diagnostic OR (DOR), negative likelihood ratio (NLR) and positive likelihood ratio (PLR) were calculated. The prevalence of obstructive CAD, proportion of patients with prior CAD and proportion of negative non-invasive test results were meta-analysed using random-effects models with the inverse variance method and logit transformation. A simple Z test was used to detect differences in prevalence of obstructive CAD between modalities.15 Sensitivity and specificity were jointly pooled using a Bayesian bivariate meta-analysis.16 NLR, PLR and summary receiver operating characteristic (sROC) curves were meta-analysed using the Reitsma approach in a bivariate model.17 The area under the curve (AUC) for each modality along with its 95% CI was calculated according to Lehman.18 Diagnostic performance between modalities was compared by using the 95% CI of each modality’s diagnostic measure. Heterogeneity for NLR among studies was quantified by calculating the I2 statistic. The degree of heterogeneity was considered low if I2<50%. A two-sided p value <0.05 was considered statistically significant. Three separate sensitivity analyses were conducted, excluding (1) mixed cohorts that comprised patients with and without prior CAD, (2) cohorts that reported the inclusion of any patient with prior bypass surgery and (3) cohorts where the reference standard was not defined by DS% (50 DS% or 70 DS%). All statistical analyses were performed using RStudio software V.4.0.3 (R Foundation, Vienna, Austria) and SAS software V.9.4 (SAS Institute).
Results
Systematic search yielded 13 747 articles. After removal of duplicates, screening of title and abstract and full-text review of 36 articles were considered relevant. Stand-alone CTP, CT-FFR, PET and SE yielded ≤3 studies per modality and were excluded for this meta-analysis. A total of 18 studies (24 modalities) were successfully included. The flow chart of the article search and selection process is depicted in figure 1. The included modalities comprised 6 CCTA studies (615 patients), 5 CCTA+CTP studies (455 patients), 9 CMR studies (1689 patients) and 4 SPECT studies (506 patients). A minimum of 91% of patients had a history of prior CAD, consisting of MI (46%) and/or PCI (76%). The reference was FFR for 6 modalities, ≥70% DS for 10 modalities, ≥50% DS for 6 modalities and a combined ≥50% DS and SPECT endpoint for 2 modalities. Further study and patient characteristics are shown in online supplemental tables 2 and 3. Per-vessel analyses could not be conducted since the majority of studies reported patient outcomes only. Obstructive CAD was found in 64% of patients. The prevalence of obstructive CAD did not differ between modalities (p=0.46). Based on FFR studies only, the prevalence of obstructive CAD was 58% (online supplemental table 4). The methodological quality of the included studies is shown in figure 2. The agreement between the independent reviewers was high (κ=0.86). The QUADAS-2 score for individual studies is depicted in online supplemental table 5 (risk of bias) and online supplemental table 6 (applicability concerns). The risk of bias split per criterion is depicted in online supplemental figure 1. In general, studies were at low risk of bias. 11 of 18 studies were considered at high risk of applicability concerns since these studies did not exclusively comprise patients with prior CAD.
Study flow chart illustrating the process of literature search and selection algorithm. In total, 18 studies comprising 24 modalities were included. CAD, coronary artery disease; CCTA, coronary CT angiography; CMR, cardiac MRI; CTP, CT perfusion; FFR, fractional flow reserve; ICA, invasive coronary angiography; PET, positron emission tomography; SE, stress echography; SPECT, single–photon emission CT.
Bias and applicability score. Assessment of methodological quality of included studies using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) criteria. Stacked bars represent the proportion (with absolute numbers) of studies with a low risk (green), unclear risk (yellow) or high risk (red) of bias or applicability concerns.
Diagnostic performance
Pooled estimates of per-patient diagnostic accuracy are shown in table 2 and figure 3. The per-patient sensitivity of CCTA (0.95; 95% CI 0.92 to 0.98), CCTA+CTP (0.93; 95% CI 0.84 to 0.98) and CMR (0.91; 95% CI 0.86 to 0.94) was similar, whereas the sensitivity of SPECT was lower (0.63; 95% CI 0.52 to 0.73). The specificity of SPECT was higher than CCTA (0.66; 95% CI 0.56 to 0.76 vs 0.37; 95% CI 0.29 to 0.46) but comparable to CCTA+CTP (0.59; 95% CI 0.49 to 0.69) and CMR (0.69; 95% CI 0.53 to 0.81). Sensitivity and specificity for individual studies categorised per modality are depicted in online supplemental figures 2 and 3. The NLR of CCTA and CMR (0.21; 95% CI 0.13 to 0.33 and 0.15; 95% CI 0.08 to 0.24) was better than SPECT (0.58; 95% CI 0.43 to 0.73), whereas no statistically significant difference was observed with CCTA+CTP (0.19; 95% CI 0.05 to 0.47). CMR demonstrated a better PLR than CCTA (3.27; 95% CI 2.21 to 4.83 vs 1.50; 95% CI 1.29 to 1.80) and similar to CCTA+CTP (2.17; 95% CI 1.68 to 2.80) and SPECT (1.87; 95% CI 1.43 to 2.51). SPECT yielded the lowest DOR (3.31; 95% CI 2.18 to 4.81) compared with CCTA (7.70; 95% CI 4.00 to 13.50), CCTA+CTP (16.73; 95% CI 3.92 to 47.56) or CMR (24.75; 95% CI 10.20 to 50.82). The sROC curves for each modality are depicted in figure 4. CCTA displayed a higher AUC than SPECT (0.91; 95% CI 0.86 to 0.98 vs 0.70; 95% CI 0.58 to 0.87). The AUCs of CCTA+CTP and CMR were comparable (0.89; 95% CI 0.73 to 1.00 and 0.91; 95% CI 0.80 to 1.00). Normal non-invasive test results were observed in 17% (95% CI 13% to 20%) of CCTA, 33% (95% CI 19% to 40%) of combined CCTA+CTP, 28% (95% CI 21% to 36%) of CMR and 49% (95% CI 33% to 66%) of SPECT scans. Heterogeneity among study results for the NLR was low for all modalities.
Diagnostic performance per modality
Diagnostic performance forest plots. CCTA, coronary CT angiography; CMR, cardiac MRI; CTP, CT perfusion; SPECT, single–photon emission CT.
Diagnostic performance sROC curves. AUC, area under the curve; CCTA, coronary CT angiography; CMR, cardiac MRI; CTP, CT perfusion; SPECT, single–photon emission CT; sROC, summary receiver operating characteristic.
Additional sensitivity analyses are displayed in online supplemental table 7. The first sensitivity analysis which exclusively covers cohorts in which all patients had prior CAD revealed that the performance of CCTA, CCTA+CTP and SPECT was similar, whereas CMR was characterised by a lower specificity (0.40; 95% CI 0.13 to 0.66), worse NLR (0.24; 95% CI 0.06 to 0.61) and lower AUC (0.78; 95% CI 0.54 to 1.00) compared with the primary analysis. The second sensitivity analysis in which all cohorts with any reported bypass surgery patient were excluded yielded similar results as the primary analysis. The third sensitivity analysis, employing an angiographic (50% DS or 70% DS) reference standard, demonstrated diagnostic performances which were concordant with the primary findings for all modalities.
Discussion
To the best of our knowledge, this is the first meta-analysis to evaluate the diagnostic performance of non-invasive cardiac imaging for the diagnosis of obstructive CAD in patients with prior CAD. The analysis based on 18 studies with 3265 patients revealed that CCTA, CCTA+CTP and CMR imaging exhibited high sensitivity, ranging from 91% for CMR to 93% for combined CCTA+CTP, and 95% for CCTA as a stand-alone tool when compared with an invasive reference standard. In contrast, SPECT demonstrated a poor diagnostic performance with a low sensitivity (63%) leading to a high rate of false negative findings, impacting clinical decision-making adversely. Specificity was moderate for CMR (69%) and SPECT (66%), whereas CCTA alone had a low specificity (37%) which improved to 59% for the hybrid approach with CTP. Obstructive CAD was highly prevalent in our meta-analysis (64%) and as such normal scan results for CCTA (17%), CCTA+CTP (33%), CMR (28%) and SPECT (49%) represented a minority. Sensitivity analyses which focused exclusively on cohorts in which all patients had prior CAD revealed that the performance of CCTA, CCTA+CTP and SPECT was similar, whereas CMR was characterised by a lower specificity.
Current guidelines for the management of patients with chronic coronary syndrome recommend non-invasive testing in patients suspected of obstructive CAD.1 2 However, there is a paucity of data on the diagnostic performance of non-invasive imaging in high-risk populations and the present guidelines have predominantly been substantiated in patients without a cardiac history. Non-invasive imaging in these patients is assumed to be more challenging due to the presence of stent artefacts and myocardial scar reducing image quality and impeding accurate interpretation. Nevertheless, even in this population, CCTA demonstrated high sensitivity and preserved its high negative predictive value for excluding obstructive CAD. The majority of CCTA studies were performed in patients with prior stenting. Studies have shown that the coronary lumen can be evaluated using 64-slice CCTA in approximately 90% of stents.19 With the advent of new CT hardware providing a spatial resolution up to 0.2 mm, in-stent lumen visibility has improved and will allow assessment of stent patency for smaller stent diameters with satisfactory diagnostic confidence.20 On the other hand, specificity of CCTA was poor, namely 39%, which is comparable to the findings of a previous meta-analysis on the diagnostic accuracy of non-invasive imaging for the detection of FFR-determined functional significant disease.9 We sought to reflect clinical care and excluded CT studies which censored unevaluable segments from analysis or reported only on the stented segments, leaving the remaining vasculature unanalysed. Notably, one may use CCTA for the mere exclusion of significant CAD in these high-risk patients and for localisation of coronary disease. The ISCHEMIA (International Study of Comparative Health Effectiveness With Medical and Invasive Approaches) trial taught us that revascularisation for moderate to severe ischaemia conveys no prognostic benefit in comparison to medical treatment.21 However, left main (LM) disease was an exclusion criterion. As such, direct referral to the catheterisation laboratory might only be judicious in the presence of significant LM disease, while for non-LM disease a ‘stepped care’ approach would be conservative treatment, followed by functional testing and eventual revascularisation in patients refractory to medical therapy by antianginal drugs. Importantly, coronary revascularisation in patients with prior CAD was found to improve perfusion to a similar postrevascularisation perfusion as patients without prior CAD and, as such, symptom reduction could be achieved.22
A high sensitivity (ie, >90%) was also observed for the hybrid CCTA+CTP approach and CMR perfusion imaging, while sensitivity of SPECT was only moderate. This might be attributable to the low spatial resolution of SPECT. Typically, SPECT has a spatial resolution of 12–15 mm, which is substantially lower compared with CT and CMR, allowing for detection of subendocardial ischaemia which might go undetected with SPECT.23–26 However, studies have shown that exploiting the full potential of SPECT by including new-generation SPECT scanners and ancillary findings such as transient ischaemic dilatation and left ventricular volumes could reduce the rate of false positive findings, while sensitivity seems unaffected by the addition of these parameters.27 In line with prior studies in patients with low to intermediate-risk chronic coronary syndrome, SPECT exhibits a low diagnostic accuracy mainly due to its high rate of false negative findings. These findings were further substantiated by a sensitivity analysis using an anatomical (DS%) reference standard. This analysis showed a lower sensitivity of SPECT (0.71; 95% CI 0.59 to 0.80) compared with a previous meta-analysis that also used a DS% reference standard but had an under-representation of patients with prior CAD (0.88; 95% CI 0.88 to 0.89).10 Results of SPECT should be interpreted with caution, and in this context direct referral for ICA and subsequent FFR measurements may guide clinical decision-making in a more salutary fashion. Of note, a sensitivity analysis focusing on cohorts with exclusively patients with prior CAD revealed a significantly lower diagnostic performance of CMR, although only three studies were eligible for inclusion in this sensitivity analysis. There were insufficient studies performed in patients with prior CAD to include CT-FFR, PET or SE. Future studies might shed light on the potential of CT-FFR to improve specificity of CCTA, while PET showed a superior sensitivity with a similar specificity in the only study that directly compared PET, qualitative CMR and SPECT in patients with a prior CAD.12 Patients with a history of CAD in the current cohort were categorised based on either prior MI or a prior coronary revascularisation. This common classification, while somewhat arbitrary, aligns with standard practices in the literature and enhances the generalisability of our meta-analysis results. Although distinguishing between these subgroups would offer deeper insights, our adherence to these conventional definitions ensures consistency with existing studies. As per standard practice, we maintain these distinctions to accurately summarise the available literature. In line with current practice, prior studies typically classified patients simply as having prior CAD or no CAD, without distinguishing between MI and PCI. Consequently, a detailed analysis focusing on these specific subgroups was not feasible. After all, the selection of a diagnostic test should be individually decided and individual contraindications should be leading. In the light of increasing healthcare costs, the European Society of Cardiology 2019 guideline advocated appropriate use of non-invasive imaging and recommends non-invasive ischaemia testing or direct referral to ICA in patients with a high pretest likelihood of CAD.2 28 Our study is the first to reliably meta-analyse the prevalence of obstructive CAD in symptomatic patients with a cardiac history against an invasive external reference standard. Obstructive CAD was highly prevalent (64%) in the present meta-analysis, and despite the satisfactory diagnostic performance of CCTA, hybrid CCTA and CTP, and CMR, the high pretest likelihood of obstructive CAD and as a consequence low rate of normal scans may justify a direct referral for ICA in patients prior CAD and refractory symptoms.
The majority (67%) of included studies defined obstructive CAD by an anatomical reference. The angiographic appearance of a coronary stenosis does not always commensurate with its functional significance, which is illustrated by a moderate sensitivity and specificity of ICA to detect FFR significant CAD.9 29 Anatomical features of CAD severity and FFR are predominantly measures of coronary atherosclerosis, whereas myocardial perfusion reflects the net impact of the coronary and microvascular bed on myocardial blood flow.30 As such, the apparent mismatch is arising from coronary physiology reflecting the different aspects of the atherosclerotic spectrum rather than the failure of either technique. Diagnostic imaging in patients with chronic coronary syndrome should not only be focused on determining lesion-specific ischaemia. Oversimplifying the complex nature of CAD will likely not improve our clinical decision making. An integrative comprehensive approach (eg, combining myocardial blood flow imaging, anatomical features of CAD and invasive coronary pressure measurements) will provide an impetus to personalised medicine by identifying patients amenable to coronary revascularisation and those who might only benefit from optimal medical therapy.
Limitations
This meta-analysis provides valuable insights based on current evidence, but there are some limitations that should be acknowledged. First, the scarcity of studies exclusively focusing on patients with prior CAD led to the inclusion of populations of which ≥50% had prior CAD. Although the proportion of patients with prior CAD is ≥91% in our meta-analysis, these cohorts comprising patients without prior CAD inevitably result in heterogeneity between studies. Second, inconsistent terminology of known and prior CAD could lead to erroneous interpretation. The potential misinterpretation of these terms may impact the reported prevalence of obstructive CAD. Third, the majority of studies did not report per-vessel data specified for patients with prior CAD, precluding per-vessel analyses. Per-vessel analyses could be informative for our understanding of test results but are of less importance if the goal is to study a non-invasive test’s potential to act as a gatekeeper for ICA. Moreover, to provide meaningful per-vessel insights in specific situations, such as territories with infarction (scar) or diffuse CAD, uniformity in study protocols and non-invasive imaging methodology are essential, and well-designed multimodality imaging studies will provide valuable insights in this matter. Fourth, it is unknown if patients underwent prior non-invasive imaging before referral for ICA. Prior abnormal non-invasive imaging could lead to selection bias and a potential overestimation of the proportion of patients with obstructive CAD on ICA. Fifth, the reference standard was a mixture of functional and anatomical parameters, which negatively impacts the generalisability of our findings. Sixth, a number of studies did not exclude patients with prior bypass surgery. Although a sensitivity analysis excluding these cohorts did not reveal different results, the exact impact of the inclusion of these patients is unknown. Last, the original studies often lacked detailed baseline characteristics beyond MI and/or PCI in patients with prior CAD, hindering the analysis of traditional risk factors’ influence in our meta-analysis despite their potential nullification in patients with prior CAD.
Conclusion
This meta-analysis focused on prospective blinded studies involving patients with prior MI or PCI. The results revealed that non-invasive imaging by CCTA, CCTA+CTP hybrid imaging or CMR exhibited a comparable high diagnostic value. SPECT was characterised by a poor sensitivity and NLR, raising concerns about its reliability to guide clinical decision-making. The prevalence of obstructive CAD was found to be high, and the added value of non-invasive imaging as gatekeeper for invasive angiography might be questioned in this high-risk cohort.
Data availability statement
Data are available upon reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
Not applicable.
