All patients underwent the routine institutional protocol of adenosine-stress SPECT-MP. This protocol includes the injection of 10mCi of TC-99m-tetrofosmin during stress and 30mCi at rest. Adenosine was infused at a rate of 140μg/kg/min for 6min, and the radiopharmaceutical was administered at 3min of adenosine perfusion. SPECT imaging was acquired in a Gamma Camera Infinia Hawkeye, GE Healthcare, 15–30min after radiopharmaceutical administration. Attenuation correction was applied on a case-by-case basis, depending on patient body mass index and breast volume, using an integrated computed tomography scanner.

QGS/QPS software (Cedars-Sinai Medical Center, United States) was used to quantify the extension and severity of perfusion defects. Myocardial perfusion was assessed through expert visual analysis in a 20 LV segmentation model. Semi-quantitative and quantitative analyses were used to assess the severity and extent of perfusion defects. Myocardial perfusion was considered abnormal if experts reported perfusion defects with a summed stress score ≥4.

All patients underwent the routine institution protocol, implying an incremental dobutamine infusion at 3-min intervals (10, 30 and 40mcg/kg/min) followed by a final 0.5mg atropine bolus if the heart rate was less than 85% of the age-predicted maximum value at the final stage. Echocardiographic imaging was acquired in a Vivid E9 ultrasound machine (GE Healthcare, United States). Echocardiographic contrast (SonoVue, Bracco, Italy) was used on a case-by-case basis if the endocardial borders were poorly visualized. EchoPAC (GE Healthcare, United States) software was used for imaging post-processing and measurements. Wall motion abnormalities were assessed through expert visual analysis in a 16 LV segmentation model. A test was considered abnormal if ischemia was reported in at least one LV segment.

All patients underwent the routine institution protocol for CTCA using a multidetector scanner (SOMATOM Sensation 64 Scanner from 2006 to 2017 and a SOMATOM Force Scanner from 2017 to 2020, Siemens Medical Solutions, Germany). After scout images, two low-dose ECG-triggered acquisitions were performed, one prospective without contrast to assess coronary calcium, and the other during the first pass of iodinated contrast for luminal evaluation of coronary arteries (prospective triggered scan for most patients). Intravenous beta-blockade was used as needed (2.5–20mg metoprolol, targeting a heart rate of 60beats/min) and all patients received 0.5mg of sublingual nitroglycerin before the scan. The resulting acquisitions were sent to a post-processing workstation (Aquarius; TeraRecon Inc. or SyngoVia; Siemens Medical Solutions, Germany) and analyzed by expert visual analysis. The test was classified as abnormal if it revealed stenosis with maximal lumen diameter narrowing ≥50%.

Coronary angiography was performed with the standard Judkins approach and coronary angiograms were interpreted visually by experienced physicians. Each vascular territory was assessed individually regarding ischemia localization on the non-invasive exam (SPECT-MPI and DSE).

For abnormal non-invasive tests, significant CAD was defined as ≥50% maximal lumen narrowing on a vascular territory consistent with ischemia localization on the non-invasive exam (or consistent diseased vessel in the case of CTCA). The non-invasive test was classified as a false positive if these criteria were not met.

In case of stenosis ≥50% in a vascular territory where ischemia was not reported, significant CAD was defined by either a functionally significant stenosis (fractional flow reserve ≤80% or instantaneous wave-free ratio ≤89%), a maximal lumen narrowing ≥90%, or multivessel CAD that triggered referral for coronary artery bypass grafting. In this case, the non-invasive test result was classified as a false negative. A normal CTCA was classified as false negative if a stenosis ≥50% was found on coronary angiography.

Categorical variables were presented as frequencies and percentages, and continuous variables as means and standard deviations, or medians and interquartile ranges for variables with skewed distributions. Normal distribution was checked using skewness and kurtosis. The diagnostic accuracy tests used were Sensitivity (%)=(100*true positives)/(true positives+false negatives); Specificity (%)=(100*true negatives)/(true negatives+false positives); Negative predictive value (%)=(100*true negatives)/(true negatives+false negatives); and Positive predictive value (%)=(100*true positives)/(true positives+false positives). The predicted accuracy for each test was also determined, Predictive accuracy (%)=(100*True positive results+True negative results)/Total number of patients. Means were compared with independent samples T-test (two groups) or one way ANOVA (three groups) and medians were compared with a Wilcoxon test (two groups) or a Kruskall–Wallis test (three groups). Categorical variables were compared with the chi-square test. A P<.05 was considered significant. In those cases where the diagnostic accuracy tests showed a significant P, a direct comparison was made between two non-invasive tests. In these cases, the P presented were adjusted with the Bonferroni correction to reduce the probability of error.

The study group consisted of 149 patients referred for coronary angiography, including 77 patients who previously underwent a SPECT-MPI, 39 who underwent DSE and 33 who underwent CTCA. The median time from a test result to coronary angiography was 3.5 (1.2–8.6) months. Characteristics of the study cohort stratified by the diagnostic test used are shown in Table 1. Patients undergoing SPECT-MP, DSE and CTCA were generally similar in terms of demographics and comorbidities, except for lower prevalence of CKD in CTCA and higher prevalence of previous revascularization in DSE. The CTCA group had a higher prevalence of patients with non-anginal chest pain, and fewer asymptomatic patients or those with dyspnea. Patients who underwent CTCA were less medicated with beta-blockers compared to patients who underwent the other non-invasive tests. All patients in the CTCA group received nitrate as part of the study protocol. Analysis of the baseline ECGs showed that LBBB was the most prevalent in all groups, compared to right ventricular pacing, and there were no significant differences in QRS duration between the groups.

The overall results of SPECT-MPI, DSE, CTCA and coronary angiography are shown in Table 2. The prevalence of obstructive CAD in the study group was 30%, with no significant differences in the three cohorts (30%, 33% and 27% for SPECT-MPI, DSE and CTCA, respectively, P=.852). Despite the similar prevalence of obstructive CAD, there was a higher proportion of abnormal results in the SPECT-MPI cohort compared to the other cohorts (84% vs 64% vs 61% for SPECT-MPI, DSE and CTCA respectively, P=.009). There were no statistically significant differences between the prevalence of abnormal results in DSE and CTCA. Regarding the results of the diagnostic accuracy tests, specificity was found to be significantly lower in the SPECT-MPI group compared to the DSE and CTCA cohorts (18% vs 50% vs 54%, respectively, P=.012), driven by the low rate of truly negative results. Comparing the DSE and the CTCA cohorts, there were no significant differences in specificity between the two groups. There were no significant differences in sensitivity, negative predictive value and positive predictive value between the three groups.

Predicted accuracy was significantly lower in the SPECT-MPI group (39% vs 64% vs 67%, P=.006). When directly comparing the accuracy of the non-invasive tests, SPECT-MPI performance was lower than DSE (39% vs 64%, P=.03) and lower than CTCA (39% vs 67%, P=.024). The DSE performance was non-inferior to CTCA regarding predictive accuracy (64% vs 67%, P>.999).

The individual test results in the study group and the characterization of true and false positive results are shown in Table 3. In the SPECT-MPI cohort, the mean LVEF estimated by the study was 43±14%. Most patients (83%) had a partial or completely reversible perfusion defect. The most frequent location of the perfusion defect was in the apical segments (83%), followed by the septal segments (79%). The prevalence of septal defects was similar in patients with true positive and false positive results (85% vs 89%, P=.420). The same was true for apical defects (95% vs 87%, P=.655) suggesting that defect location is not a reliable marker to differentiate false positive results in these patients. Additionally, there were no differences between true and false positive results regarding the extent of the stress defect (P=.357) and the severity of the stress defect, assessed by the summed stress score (P=.789), the summed rest score (P=.793) or the summed difference score (P=.865).

In the DSE cohort, the mean LVEF was 42±13%. The most frequent location of ischemia or scar was in the septal wall (49%), followed by the inferior wall (46%). There were no differences in the location of ischemia or scar between true positives and false positives. Non-sustained ventricular tachycardia was an unreliable marker to differentiate true positive results (P=.076).

In the CTCA cohort, the median Agatston calcium score was 171 (45–425), with no differences between true and false positive results (P=.408). The lesion severity was also an unreliable predictor of false positive results since there were no differences between true and false positive results in patients with a CAD-RADS classification of 3 (P=.127) or 4 (P=.127). Similarly, the coronary artery involved was also an unreliable predictor of false positive results, with a similar distribution between the two groups for the anterior descending artery (P=.479), circumflex artery (P>.999) and right coronary artery (P=.566).

Due to the observational nature of the study, there is a risk of referral bias. Since patients with negative test results are not commonly referred to coronary angiography, this introduces referral bias, which has been shown to result in an overestimation of sensitivity and underestimation of specificity.15 In our study, this specificity bias may be more relevant concerning SPECT-MPI since few patients with a normal test underwent coronary angiography and also because there is a high rate of false negatives, reducing specificity.

The small sample size, especially in the DSE and CTCA groups, is also a limitation of the study. Additionally, since patients with previous myocardial infarction or incomplete revascularization were excluded, it may influence the external validity of the results.

The different segmentation of the heart adopted by non-invasive functional exams could be a source of bias and constitute a limitation of the study.

Given that the time between the non-invasive test and coronary angiography was not the same for all patients, this is also a limitation of the study. Furthermore, the gold standard in this analysis was coronary angiography without the use of invasive functional tests, which would have been the true gold standard of non-invasive functional tests.

Finally, the inclusion of patients who underwent myocardial perfusion assessment by magnetic resonance imaging was not possible because of the low referral of patients meeting inclusion criteria in the study centers.