Clinical experience regarding the diagnostic value of segment-by-segment coronary computed tomography angiography in comparison with that of invasive coronary angiography

Objective To compare the degree of coronary stenosis (≥ 50% luminal narrowing) determined by coronary computed tomography angiography (CCTA) with that determined by invasive coronary angiography (ICA), using segment-by-segment analysis. Materials and Methods This was a retrospective study of the records of patients who underwent CCTA and ICA between January 2014 and June 2018 at a general hospital in Brazil. Receiver operating characteristic curve analysis was applied, and the areas under the curve were used in order to assess the overall accuracy of the methods. Results The degree of coronary stenosis was evaluated in a total of 844 arterial segments. The diagnostic performance of CCTA was good, with a sensitivity of 82.3%, a specificity of 96.4%, and a negative predictive value of 97.7% (95% CI: 96.5-98.5). In the segment-by-segment analysis, CCTA had excellent accuracy for the left main coronary artery and for other segments. Conclusion In clinical practice at general hospitals, CCTA appears to have diagnostic performance comparable to that of ICA.

Unitermos: Tomografia computadorizada multidetectores; Angiografia coronária/métodos; Doença da artéria coronária/diagnóstico por imagem. effectiveness of this method has been demonstrated in various studies, in which its correlation with invasive coronary angiography (ICA) has been examined. The use of CCTA is very important to exclude or detect coronary artery disease, even at subclinical levels (5)(6)(7) . This method has a negative predictive value (NPV) of 96-100%, making it reliable for the exclusion of coronary artery stenosis (6) .
To our knowledge, there have been no studies involving segment-by-segment and per-patient analyses of the correlation between CCTA and ICA findings in a hospital setting in Brazil. The objective of this study was to evaluate INTRODUCTION Cardiovascular diseases are responsible for the death of more than 17.9 million people annually, accounting for 31% of all deaths worldwide (1) . Eighty percent of those deaths are caused by acute myocardial infarction or stroke (1,2) . The diagnosis of coronary artery disease is important for the initiation of specific therapy and the prevention of ischemic events (3,4) .
Approximately two decades ago, noninvasive evaluation of the coronary arteries using coronary computed tomography angiography (CCTA) became possible. The the degree of coronary stenosis (≥ 50% luminal narrowing) determined by segment-by-segment analysis at a general hospital, comparing CCTA and ICA. We sought to determine whether CCTA and ICA are similar in terms of their ability to predict coronary artery disease in daily clinical practice at a general hospital in Brazil.

MATERIALS AND METHODS
This was a retrospective, cross-sectional, observational study of data related to patients who underwent CCTA and ICA involving cardiac catheterization, between January 2014 and June 2018, at the Complexo Hospitalar de Niterói, in the city of Niterói, Brazil. The Complexo Hospitalar de Niterói is a tertiary care hospital, operated by Universidade Federal Fluminense, with a 24-h emergency department that is a referral center for trauma cases. The Research Ethics Committee of Universidade Federal Fluminense approved the study (Reference no. 85407818.4.0000.5243). Because of the retrospective nature of the study, the requirement for written informed consent was waived.
Imaging studies in the Digital Imaging and Communications in Medicine format were identified by a search of the Picture Archiving and Communication System of the hospital. We reviewed the records of all adult patients (≥ 18 years of age) who, at the request of their physicians, had undergone ICA < 4 months after CCTA, to monitor chronic coronary artery disease. We included only imaging studies that were of diagnostic quality, with no artifacts that would render analysis unviable (e.g., pronounced arrhythmia, involuntary movements, and respiratory motion). A flow chart of the study selection process is displayed in Figure 1.

CCTA protocol
All CCTA examinations were performed in a 64-slice CT scanner (Somatom Sensation 64; Siemens, Forchheim, Germany), using a specific electrocardiogram-gating protocol, before and after intravenous injection of contrast medium. If the heart rate was above 65 bpm and there were no contraindications to the use of metoprolol tartrate (e.g., asthma or difficult-to-control heart failure), it was prescribed at a dose of 5-30 mg. Prior to injection of the contrast medium, all of the patients were given sublingual isosorbide dinitrate (5 mg) for coronary vasodilation, except for the patients with contraindications to its use. Patients at risk for adverse reactions to contrast medium were submitted to a desensitization protocol: oral prednisone (20 mg every 6 h), starting 12 h before the procedure; and oral diphenhydramine (50 mg), at 1 h before the procedure.
Non-ionic contrast (60 mL, Henetix 350; Guerbet, Villepinte, France) was injected into an antecubital vein at 5 mL/sec, after which 20 mL of an isotonic saline solution (0.9% NaCl) were administered with a dual-syringe injection pump (Stellant; Medrad, Indianola, PA, USA). The contrast bolus trigger was used in order to determine the timing of CCTA acquisition, allowing the arrival of the contrast medium in the middle ascending aorta to be noted.

ICA protocol
All ICA examinations were performed through transradial access with a 6F sheath. The angiography system used (Artis zee; Siemens Healthineers, Erlangen, Germany) had a 17-in. intensifier. A minimum of eight X-ray projections were acquired for the study of the coronary arteries.

CCTA imaging analysis
Two radiologists, with 5 and 17 years of experience in cardiac imaging, respectively, evaluated the images and accompanying reports. The radiologists interpreted the images by consensus, using axial source images, thin-slab maximum intensity projections, and multiplanar reconstruction on an image processing workstation (Leonardo; Siemens Healthineers). Coronary segments were identified by using the segmentation protocol devised by Raff et al. (8) . For each segment, significant stenosis was defined as luminal narrowing ≥ 50%.

ICA imaging analysis
Images from ICA examinations were stored digitally in multiple views and subsequently analyzed by a cardiologist who was blinded to the CCTA results. Coronary segmentation followed the same protocol used in the CCTA analysis.
On the basis of the segmentation protocol devised by Raff et al.

Statistical analysis
Continuous variables are expressed as means ± standard deviations, whereas categorical variables are expressed as absolute and relative frequencies. Considering ICA as the gold-standard method of imaging, we calculated the accuracy, sensitivity, specificity, positive predictive value (PPV), and NPV of CCTA, with 95% confidence intervals (95% CIs).
We assessed the performance of CCTA in the identification of ≥ 50% luminal narrowing, relative to that of ICA, by using receiver operating characteristic (ROC) curve analysis. Findings from both examinations were analyzed segment by segment and per patient. Segments with ≥ 50% and < 50% luminal narrowing, as determined by anatomical evaluation via ICA, served as true-positive and true-negative markers, respectively.
The ROC analysis was applied to the categorical responses, and the overall accuracy of each analysis was assessed by calculating the area under the curve (AUC). We considered AUCs ≥ 0.5 to < 0.7 to be indicative of poor agreement between the performance of CCTA and that of ICA in evaluating stenosis grading, whereas we considered AUCs ≥ 0.7 to < 0.9 to be indicative of good agreement and AUCs ≥ 0.9 to 1.0 to be indicative of excellent agreement.
Multiple comparisons with kappa tests were performed to assess the level of agreement between the ICA and CCTA analyses in a segment-by-segment mode and in a per-patient mode. The MedCalc statistical software package, version 14.8.1.0 for Windows (MedCalc Software, Ostend, Belgium) was used for the statistical analyses. Values of p < 0.05 on two-tailed tests were considered to be significant.

RESULTS
We reviewed data from 50 patients who underwent ICA and CCTA during the study period. Two patients were excluded due to excessive arrhythmia-related artifacts, and one patient was excluded for being under 18 years of age. Therefore, the final sample comprised 47 patients. The mean age was 69.1 ± 12.1 years (range, 18-95 years), and 36 (76.6%) of the patients were male.

DISCUSSION
This study revealed good agreement between CCTA and ICA in the identification of ≥ 50% luminal narrowing at a general hospital in Brazil. We found that CCTA showed accuracy exceeding that obtained in randomized studies (5-7) .
The NPV of CCTA for all 844 segments examined in the present study was 97.7%, which is comparable to values obtained in other studies, such as that conducted by Mahdavi et al. (9) , who reported an NPV of 97.2% for 628 segments in 47 patients. In a CCTA validation study, Budoff et al. (7) obtained an NPV of 99.0% for the identification of ≥ 50% luminal narrowing in 910 vessels. We found that the PPV of CCTA was lower than was its NPV, as was also found by Scheffel et al. (10) . In our sample, the NPV was less than 100% because some patients did not have class I recommendations for CCTA, reflecting the fact that our assessment was performed in an everyday clinical setting, without strict application of exclusion criteria, which may have led to an overestimation in the quantification of obstructive calcified plaques. Taken together, however, these findings demonstrate that CCTA performs well for the exclusion of coronary artery disease, thus minimizing the risk of unnecessary invasive procedures (11)(12)(13) .
The AUC of 0.91 obtained for CCTA in our per-patient analysis is similar to the 0.96 obtained by Budoff et al. (7) , who applied more exclusion criteria. As in the present study, Chow et al. (14) showed that CCTA had excellent sensitivity and a high NPV in comparison with ICA.

Segment-by-segment analysis
Results of the segment-by-segment analysis are presented in Table 1. The diagonal branch was omitted from the analysis because it was not present in any of the patients in our sample. The Dg3-LAD, Mg1-RCA, and Mg3-LCx corresponded to 46 segments each, none of which showed ≥ 50% luminal narrowing on CCTA or ICA. The PDA-LCx and PLV-LCx corresponded to three segments each. None of the PLV-LCx showed ≥ 50% luminal narrowing on either examination, and one PDA-LCx segment showed ≥ 50% luminal narrowing only on CCTA. For all of these segments, it was not possible, with the software used for the statistical analyses, to calculate the ROC curve. Therefore, none of these 144 segments appear in Table 1.