This was a cross-sectional observational study performed at a single-center tertiary hospital in Lisbon, Portugal. The study population consisted of all consecutive patients referred for CCTA for suspected CAD between October 2016 and 2019. The following exclusion criteria were applied sequentially: age <30 years, known CAD, suspected acute coronary syndrome, asymptomatic, symptoms other than chest pain, preoperative assessment, known left ventricular ejection fraction <50% and inadequate image quality.

Clinical data were systematically collected from electronic medical records and included patient demographics and cardiovascular risk factors. Angina characteristics were assessed at the time of CCTA and classified according to the following three criteria: (1) substernal chest pain or discomfort; (2) caused by physical or psychological stress; and (3) relieved by rest and/or nitroglycerine. Chest pain was classified as typical if all three criteria were fulfilled, as atypical when two criteria were present, and as nonanginal if one or no criteria was reported.10 Hypertension was defined as a history of blood pressure ≥140/90 mmHg and/or treatment with antihypertensive medications.11 Diabetes was defined as fasting glucose of ≥126 mg/dl and/or use of hypoglycemic agents.12 Dyslipidemia was defined as a total cholesterol level ≥200 mg/dl or treatment with lipid-lowering medication.13 Current smoking status was defined as active smoking within three months of presentation. Family history of CAD was defined as myocardial infarction or cardiac death in a first-degree relative.14 From these data, the PTP of disease was calculated according to the new 2019 ESC PTP model and by the MDF method.

The 2013 ESC guidelines on the management of stable CAD recommended direct referral for ICA if PTP was >85%, non-invasive testing for PTP 15-85%, and no further testing for PTP <15%. The new 2019 ESC guidelines modified these recommendations: no specific threshold for choosing anatomical vs. functional testing is now recommended, and the threshold to withhold additional testing was lowered to PTP <5%. Considering this, in order to gauge how the method for estimating PTP might influence the diagnostic approach according to these new guidelines, we measured the proportion of patients who would be reclassified into a different risk category (low <15%; intermediate 15-85%; high >85%) when assessed by the new 2019 ESC PTP model instead of the MDF method. Additionally, we assessed the proportion of patients who would have no indication for further testing according to the old and new guidelines (MDF PTP <15% and 2019 ESC PTP <5%, respectively).

The study was approved by the local institutional review board and all patients gave written informed consent. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki.

All scans were performed on a dual-source 64-slice scanner (SOMATOM Definition®, Siemens Healthineers, Erlangen, Germany) in accordance with the Society for Cardiovascular Computed Tomography guidelines valid at the time. A cardiologist and/or a radiologist with level III-equivalent experience analyzed all scans on an Aquarius® workstation (Terarecon Inc., USA) using axial images, multiplanar reconstructions and maximum intensity projections, as appropriate.

Obstructive CAD was defined as any ≥50% luminal stenosis on CCTA. Coronary atherosclerotic lesions were assessed by visual estimation and any disagreements were resolved by consensus. Patient records were also reviewed for downstream ICA performed during the subsequent six months. Whenever ICA was performed, patients were reclassified accordingly using the same ≥50% luminal stenosis threshold.

An Agatston score threshold of 400 was used as a general reference for withholding CCTA, but the ultimate decision was left to the discretion of the performing physician, taking into consideration the clinical context and the distribution of calcium in the coronary tree. Patients who had CCTA withheld due to high coronary artery calcium (CAC) were classified according to downstream testing if performed or, if not, were conservatively assumed to have obstructive CAD, since symptomatic patients with high CAC scores have a high likelihood of angiographically significant coronary stenosis.15

Categorical data are presented as frequencies and percentages. Continuous data are presented as mean ± standard deviation (parametric) or as median with interquartile range (IQR) (nonparametric). Baseline characteristics were compared between the two groups with the chi-square test (categorical data) and the unpaired Student's t test or Mann-Whitney U test, as appropriate (continuous data). The related-samples Wilcoxon signed rank test was used to assess the difference in PTPs calculated with the different methods. The median PTPs of patients with and without obstructive CAD were compared using the independent samples Mann-Whitney U test.

Model calibration reflects how close predicted probabilities are to the observed prevalence of obstructive CAD and was assessed by calibration-in-the-large (which compares the mean observed frequency of obstructive CAD with the mean predicted probability) and discordance ([expected percentage-observed percentage]/observed percentage).

The discriminative ability of each model was expressed by the C-statistic, which, for binary outcomes, is identical to the area under the receiver operating characteristic (ROC) curve.16 The areas under the ROC curves were compared using the nonparametric method described by De Long et al.17

Net reclassification improvement, which compares the proportions of patients moving up or down in clinically meaningful PTP categories in cases versus controls, was used to ascertain the ability of the new 2019 ESC PTP model to reclassify risk.18

Statistical analyses were conducted using IBM SPSS version 25.0 (IBM SPSS® Inc., Chicago, IL, USA) and MedCalc® version 9.3.8.0 (Mariakerke, Belgium). A two-tailed p-value <0.05 was considered statistically significant.

Overall, the MDF method overestimated the likelihood of obstructive CAD by 130%, whereas the new 2019 ESC PTP model underestimated the probability of disease by 6.5%. The observed versus expected prevalence of obstructive CAD by quartiles of PTP can be observed in Figure 2. The MDF method overestimated the prevalence of obstructive CAD in all quartiles of PTP, with the calibration-in-the-large statistic indicating significant miscalibration (p for miscalibration 0.005). By contrast, the updated 2019 method showed good calibration for predicting the likelihood of obstructive CAD, with the calibration-in-the-large statistic supporting the validity of this model (p for miscalibration 0.712). Stratification by gender produced similar results. We also performed a sensitivity analysis excluding patients who were conservatively assumed to have obstructive CAD (n=10) and the results remained similar.

Figure 2.

Comparison of observed vs. expected prevalence according to the two pre-test probability (PTP) prediction models. The 2013 European Society of Cardiology (ESC) prediction model significantly overestimated the observed prevalence of CAD across all quartiles of PTP, whereas the new 2019 ESC prediction model showed good calibration for predicting the likelihood of coronary artery disease. Abbreviations as in Figure 1.

(0.19MB).

Regarding the ability of each model to differentiate patients with and without obstructive CAD, the two approaches showed similar discriminative power, with C-statistics of 0.730 (95% confidence interval [CI]: 0.658-0.802) and 0.735 (95% CI: 0.663-0.808) for the 2013 and 2019 methods, respectively (p for comparison 0.933) (Figure 3). These results were similar in men and women.

Figure 3.

Receiver operating characteristic curves of the two models for the prediction of obstructive coronary artery disease on coronary computed tomography angiography. The two models showed similar discriminative power, with C-statistics of 0.730 (95% confidence interval [CI]: 0.658-0.802) and 0.735 (95% CI: 0.663-0.808) for the 2013 and 2019 methods, respectively (p for comparison 0.933). ESC: European Society of Cardiology; MDF: modified Diamond-Forrester model; PTP: pre-test probability.

(0.09MB).

As the calculation of PTP in the 2019 guidelines downgraded the tabular method presented in the previous document, the likelihood of disease assessed by the new method was significantly lower than the MDF estimate. On average, individual PTP was 22% lower in the new guidelines. According to the MDF method, most patients (84.4%, n=270) were classified as intermediate risk, while the 2019 ESC PTP model classified the majority of patients as low risk (58.4%, n=187). The reclassification across risk categories among patients with and without obstructive disease and the resulting net reclassification improvement are depicted in Table 2.

Overall, and assuming the same thresholds, the adoption of the new guidelines would reclassify 48.4% of patients (45.6% from intermediate-risk to low-risk [n=146], and 2.8% from high-risk to intermediate-risk [n=9]), possibly leading to a modification in the diagnostic pathway. As for the lower end of the spectrum, an MDF score-guided risk stratification would classify 12.8% of patients (n=41) as low-risk (PTP <15%), warranting no further testing. Of these, 2.4% (n=1) would have obstructive CAD. Using the new 2019 ESC PTP method and the new no-testing threshold (PTP <5%), 16.3% of our population (n=52) would not undergo further testing, 3.8% of whom (n=2) would have obstructive CAD.

Some limitations of this study should be acknowledged. First, we studied a relatively low-risk population referred for CCTA, and so extrapolation of these findings to other subsets of patients should be performed with caution. Second, our definition of obstructive CAD was based on CCTA, which has a tendency to overestimate stenosis severity; nonetheless, in order to reduce the false positive rate, whenever ICA was subsequently performed patients were reclassified accordingly. Third, patients with dyspnea were excluded from the analysis, since the method proposed by the 2013 guidelines only applies to patients with some form of chest pain. Also, the presence of cardiovascular risk factors was determined based on patient self-reporting, CCTA requests by referring physicians, and electronic medical records, but not on specific blood tests or detailed medical history. Finally, clinical outcomes were not assessed or correlated with the PTP methods, as this was beyond the scope of this study.