From January 2010 to July 2018, 15,918 consecutive patients underwent PCI at a single high-volume center. Inclusion criteria were: (1) PCI of a bifurcation lesion with significant SB; (2) initial stenting of the main branch with a provisional approach to the SB; (3) performance of coronary CTA within 30 days before attempted PCI. Coronary bifurcation lesion was defined as a coronary artery narrowing occurring adjacent to or involving the origin of a significant SB [15]. Significant SB was defined according to the 11th consensus document from the European Bifurcation Club as a branch that the operator would not want to lose in the global context of an individual patient [16]. The primary endpoint was defined as SB occlusion represented by either any decrease in thrombolysis in myocardial infarction (TIMI) flow grade (including TIMI 1–2 flow grade) or the absence of flow in the SB (TIMI 0 flow grade) after MV stenting [15].

Our study was a post hoc analysis of the participants in the CT-PRECISION registry (NCT03709836), and was approved by the Institutional Review Board and complied with the declaration of Helsinki. The registry was described previously [4]. Out of 400 bifurcation lesions among 363 patients, we excluded 23 lesions with in-stent restenosis in 23 patients, due to the limitations of quantitative plaque analysis in stented segments. Thus, the final study cohort included 377 bifurcation lesions in 340 patients (Fig. 1).

During the study period, three generations of dual-source CT scanners were used (SOMATOM Definition, SOMATOM Definition Flash and SOMATOM Force, Siemens, Erlangen, Germany). Patients without contraindications received intravenous metoprolol (2.5 mg boluses titrated up to 10 mg) to limit heart rate below 65 beats/min, and sublingual nitroglycerin was administered prior to image acquisition on a regular basis. The contrast transit time was estimated by injection of a test bolus. For acquisition of the volume dataset, 60–120 ml iodinated contrast material (Iomeron 400, Bracco Imaging SpA, Milan, Italy) was injected intravenously at a rate of 6 ml/s (Definition, Flash) or 4.5 ml/s (Force). Injection rate could be decreased as a part of contrast material reduction strategy with lower tube voltages generated by the third-generation dual-source CT scanners (Force). Data were acquired using a retrospectively gated or prospectively electrocardiogram-triggered protocol with a beam collimation of 64 × 0.6 mm (Definition), 128 × 0.6 mm (Flash) or 192 × 0.6 mm (Force) and tube voltage of 70–120 kV adjusted to body mass index. Radiation dose reduction strategies including electrocardiogram-gated tube current modulation and prospectively electrocardiogram-triggered sequential acquisition were used whenever feasible. Image data were reconstructed in mid-to-end systole and diastole (35–45% and 65–75% of the R–R interval) with 0.6-mm slice thickness and 0.4 mm increment.

Coronary bifurcation was divided into: the proximal MV, the distal MV, and the SB. The beginning of the distal MV segment was defined at the carinal level as previously described [17]. Segments within 5 mm proximal and distal to the carinal point for MV, and 5 mm distal from the ostium of the SB were analyzed [17]. Coronary arteries were assessed with multiplanar, curved multiplanar, and transverse reformations in the diastolic phase of the cardiac cycle. First, each segment of the bifurcation was visually evaluated for the presence of coronary plaque defined as any discernible structure greater than 1 mm² that could be assigned to the coronary artery wall [18]. Segments with evidence of coronary plaque were further analyzed using semi-automated software (AutoPlaque version 2.5, Cedars-Sinai Medical Center, Los Angeles, California) allowing for quantitative 3-dimensional assessments of plaque morphology and stenosis (Fig. 2). As described and validated previously, scan-specific thresholds for calcified and noncalcified plaques were automatically generated. Briefly, the attenuation threshold levels were calculated by fitting a Gaussian curve to the image histogram from the aortic normal blood pool (defined by placing a circular region of interest within the ascending aorta), after adjusting with the proximal-to-distal luminal contrast enhancement distribution in the analyzed vessel [19]. Low-density plaque was defined as noncalcified plaque subset using a fixed attenuation of < 30 Hounsfield units [20]. Coronary plaque measurements included absolute volumes (in mm³) and corresponding burdens (respective plaque volumes × 100%/vessel volume). Furthermore, maximal diameter stenosis, maximal area stenosis, minimal lumen diameter, minimal lumen area, vessel remodeling index, and contrast density difference were measured. Maximal diameter stenosis was calculated as the ratio of the narrowest luminal diameter and the mean of two non-diseased reference points within the same bifurcation segment. Maximal area stenosis was calculated as the ratio of the narrowest luminal area and the mean of two non-diseased reference points within the same bifurcation segment [21] Minimal luminal diameter and area were determined semi-quantitatively at the region of maximal stenosis degree [20]. The remodeling index was determined as the ratio between maximum vessel area and the vessel area at the beginning of the bifurcation segment [22]. The luminal contrast density, defined as attenuation per unit area (similar to area gradient), was computed over 1-mm cross-sections of the arterial segment within bifurcation lesion. The contrast density difference was defined as the maximum percent difference in contrast densities across the bifurcation segment with respect to the proximal reference cross-section [23]. True bifurcation lesions were defined as a Medina classification of 0.1.1, 1.0.1, or 1.1.1.

Continuous variables are expressed as the mean ± standard deviation or median and interquartile range, and were compared with Student’s t-test or Mann–Whitney U test as appropriate. Chi-square or Fisher’s exact tests were used for comparison of categorical variables expressed as counts and percentages. Discriminatory performance was determined by the C-statistic, and compared using the method of DeLong et al. [24] Receiver operating characteristic (ROC) curves were generated for the presence of plaque in respective bifurcation segments alone or in combination with quantitative plaque analysis. In the combined analysis, plaque presence was given hierarchical superiority. Optimal cut-offs for quantitative plaque features were identified using the Youden index. All probability values were two -tailed, and a p value of < 0.05 was considered statistically significant. Data were processed using the SPSS software, version 23 (IBM SPSS Statistics, New York, USA).

Out of 377 bifurcation lesions, 1131 bifurcation segments were screened for quantitative plaque analysis (Fig. 1). Overall, the presence of coronary plaque was noted in 331 (87.8%) proximal MV, 334 (88.6%) distal MV, and 133 (35.3%) SB segments. SB occlusion occurred in 28 (7.5%) of 377 bifurcations: the absence of flow and decrease in TIMI flow were noted in 12 (43%) lesions and 16 (57%) lesions, respectively.

The baseline clinical characteristics were similar in patients with and without SB occlusion (Table 1). Per-lesion angiographic characteristics showed higher incidence of true bifurcations in lesions with versus without SB occlusion (Table 2).

Table 3 displays quantitative anatomical plaque characteristics stratified by the occurrence of SB occlusion for respective bifurcation segments. While the SB occlusion group displayed higher contrast density difference in the proximal MV segment than the SB non-occlusion group (16.5 vs 9.3%, p = 0.034), no significant between-group differences were found in distal MV. Lesions with SB occlusion had lower minimal luminal diameter (1.5 ± 0.6 vs 2.0 ± 0.7 mm, p = 0.002) and minimal lumen area (1.9 [IQR 0.9–2.6] vs 3.1 [IQR 1.7–4.6] mm², p = 0.003) within SB segments compared to lesions without SB occlusion.

Quantitative morphological plaque characteristics stratified by the presence of SB occlusion are summarized in Figs. 3, 4. Lesions with occluded SB had significantly higher total plaque volume compared to lesions without SB occlusion in proximal MV segment (47.2 [IQR 37.9–58.7] vs 34.9 [IQR 25.5–49.9] mm³, p = 0.004) (Fig. 3). Whereas SB occlusion group demonstrated higher volumes of noncalcified plaque in proximal (41.2 [IQR 32.0–52.0] vs 31.0 [IQR 22.0–43.3] mm³, p = 0.003) and distal (35.3 [IQR 28.0–41.6] vs 27.8 [IQR 19.6–40.2] mm³, p = 0.025) MV segments, the volume of low-density plaque differed only in proximal MV with higher values in the SB group than non-SB group (5.7 [IQR 2.0–6.7] vs 3.0 [IQR 1.2–5.7] mm³, p = 0.007). Calcified plaque volume differed between the two groups exclusively in the SB segment, where lesser volume was observed in lesions with SB occlusion than without SB occlusion (0.0 [IQR 0.0–0.1] vs 0.0 [IQR 0.0–4.2] mm³, p = 0.015).

Both proximal (58.75 [IQR 50.0–64.1] vs 48.7 [IQR 38.6–62.3]%, p = 0.024) and distal (57.3 [IQR 46.3–66.5] vs 49.7 [IQR 40.1–59.9]%, p = 0.016) segments of MV but not SB (56.2 [IQR 47.3–60.0] vs 51.0 [IQR 38.6–59.3]%, p = 0.074) had higher total plaque burden in lesions with than without SB occlusion. Whereas SB occlusion group was characterized with increased noncalcified plaque burden across all three bifurcation segments, the low-density plaque burden was significantly higher in proximal MV (6.6 [IQR 2.8–10.3] vs 3.8 [IQR 1.6–7.7]%, p = 0.013) and SB segments (7.3 [IQR 4.9–19.2] vs 6.1 [IQR 2.6–9.5]%, p = 0.016) among lesions with versus without SB occlusion (Fig. 4).

The presence of visually identified plaque in the SB segment, but not in the proximal and distal MV segments, was the only qualitative parameter that predicted SB occlusion with an area under the ROC curve (AUC) of 0.792 (Table 4). Among quantitative plaque parameters calculated for the SB segment, only the addition of noncalcified plaque burden (AUC 0.840, 95% CI: 0.764–0.915, p = 0.003) and low-density plaque burden (AUC 0.836, 95% CI: 0.760–0.912, p = 0.012) yielded significant improvements in predictive accuracy of visually identified SB plaque presence (Fig. 5).

The presence of coronary plaque in the SB segment had 89% sensitivity and 69% specificity for the prediction of SB occlusion. Using ROC curve analysis, optimal cut-offs for non-calcified plaque burden and low-density plaque burden were > 33.6% (86% sensitivity and 78% specificity) and > 0.9% (89% sensitivity and 73% specificity), respectively.