https://orcid.org/0000-0003-2348-0484
Abstract
The magnitude of alterations in which coronary arteries remodel and narrow over time is not well understood. We aimed to examine changes in coronary arterial remodelling and luminal narrowing by three-dimensional (3D) metrics from serial coronary computed tomography angiography (CCTA).
From a multicentre registry of patients with suspected coronary artery disease who underwent clinically indicated serial CCTA (median interscan interval = 3.3 years), we quantitatively measured coronary plaque, vessel, and lumen volumes on both scans. Primary outcome was the per-segment change in coronary vessel and lumen volume from a change in plaque volume, focusing on arterial remodelling. Multivariate generalized estimating equations including statins were calculated comparing associations between groups of baseline percent atheroma volume (PAV) and location within the coronary artery tree. From 1245 patients (mean age 61 ± 9 years, 39% women), a total of 5721 segments were analysed. For each 1.00 mm3 increase in plaque volume, the vessel volume increased by 0.71 mm3 [95% confidence interval (CI) 0.63 to 0.79 mm3, P < 0.001] with a corresponding reduction in lumen volume by 0.29 mm3 (95% CI −0.37 to −0.21 mm3, P < 0.001). Serial 3D arterial remodelling and luminal narrowing was similar in segments with low and high baseline PAV (P ≥ 0.496). No differences were observed between left main and non-left main segments, proximal and distal segments and side branch and non-side branch segments (P ≥ 0.281).
Over time, atherosclerotic coronary plaque reveals prominent outward arterial remodelling that co-occurs with modest luminal narrowing. These findings provide additional insight into the compensatory mechanisms involved in the progression of coronary atherosclerosis.
Introduction
From landmark histopathological observations in left main arteries by Glagov et al.,1 it is presupposed that the early formation of atherosclerosis is associated with dilatation of coronary arteries in order to preserve blood flow into the myocardium. These initial findings at single timepoints were later confirmed for the other major epicardial arteries by subsequent two-dimensional (2D) cross-sectional studies using invasive coronary angiography.2,3 Recent technological advancements now allow for non-invasive three-dimensional (3D) volumetric quantification of coronary plaque, vessel, and lumen using coronary computed tomography angiography (CCTA).4 We previously published findings from the Progression of AtheRosclerotic PlAque DetermIned by Computed TomoGraphic Angiography Imaging (PARADIGM) registry revealing changes in coronary plaque composition and high-risk plaque features following intercurrent, non-randomized treatment.5 The current secondary analysis aimed to examine changes in coronary arterial remodelling and luminal narrowing over time by applying volumetric quantification metrics from serial CCTA.
Methods
Study design and population
PARADIGM is a prospective, dynamic, multicenter, observational registry, enrolling patients at 13 sites in 7 countries across Asia, Europe, North- and South America between 2003 and 2015. The study design has been published in detail.6 Of those enrolled in the PARADIGM registry, a total of 1478 patients with suspected coronary artery disease (CAD) underwent clinically indicated serial CCTA with an interscan interval of ≥2 years. The study protocol was approved by each site’s institutional review board or ethics committee, and all patients provided written informed consent. For the present analysis, patients with absence of coronary plaque on both scans (n = 233) were excluded. Patients who experienced an adverse event between baseline and follow-up CCTA were not omitted. Thus, 1245 patients were included comprising a total of 5721 segments and 3294 lesions (Figure 1).
CONSORT study diagram. 2D, two-dimensional; 3D, three-dimensional; CAD, coronary artery disease; CCTA, coronary computed tomography angiography. aNo coronary plaque on both scans. bAbsent values on a per-segment level were due to no coronary plaque, revascularization, small size <2 mm, artefacts and chronic total occlusions, amongst others. cAbsent values on a per-lesion level were due to revascularization and newly developed lesions with therefore no available measurements of coronary plaque, vessel, and lumen at baseline CCTA, amongst others.
CCTA acquisition and image analysis
All patients were scanned with ≥64 detector row computed tomography (CT) scanners, and protocols with regard to the acquisition and post-processing of scans were in direct accordance with the Society of Cardiovascular Computed Tomography (SCCT) guidelines.7,8 Protocols included site-specific systematic administration of betablockers and sublingual nitroglycerine before acquisition. All scans were analysed according to an 18-segment SCCT model by level III-experienced readers at a central independent core laboratory, who were blinded to clinical and other test results. Qualitative and quantitative analysis was performed, the latter by use of semi-automated software with appropriate manual correction (QAngio CT Research Edition version 2.1.9.1, Medis Medical Imaging Systems, Leiden, the Netherlands).4,9 Quantitative analysis consisted of a 2D cross-sectional evaluation at 500 µm slice thickness, with interpolation across length to compute 3D volumes. All coronary segments ≥2 mm in diameter were evaluated for coronary plaque, vessel, and lumen dimensions with manual adjustment where necessary. Coronary plaque was defined as any tissue ≥1 mm2 within or adjacent to the coronary lumen that could be discriminated in >2 planes from pericardial tissue, epicardial fat, or lumen.10 All measurements were performed on a per-segment and per-lesion level, and summation of these values generated per-patient data. Intra and interobserver intraclass correlations for measurements were excellent as previously reported.5,6
Serial arterial remodelling and luminal narrowing
The present analysis applied standardized definitions of serial arterial remodelling and luminal narrowing: changes in coronary vessel and lumen dimensions between scans performed at a median of 3.3 years apart.11,12 Changes in coronary plaque, vessel, and lumen dimensions were evaluated with both serial 3D and 2D metrics.
Serial 3D metrics
Per-segment changes in coronary plaque, vessel, and lumen volume were calculated between baseline and follow-up CCTA (Supplementary data online, Part A).5,13 Segments were co-registered at baseline and follow-up.5 Serial 3D arterial remodelling was defined as the volumetric change in vessel volume (mm3) for each 1.00 mm3 change in plaque volume. Similarly, serial 3D luminal narrowing was defined as the volumetric change in lumen volume (mm3) for each 1.00 mm3 change in plaque volume. Furthermore, these changes were evaluated according to: (i) baseline percent atheroma volume (PAV) with the median as a cut-off based upon non-normal distribution, and (ii) baseline location within the coronary artery tree including left main, proximal and side branch status. PAV was calculated as plaque volume/vessel volume * 100%, and the median baseline PAV of our study population on a per-segment level was 6.6% [interquartile range (IQR) 1.0–17.9%]. Left main segments included by definition the left main artery, whilst the other 17 segments were considered non-left main segments. Proximal segments included the proximal right coronary artery (RCA), left main artery, proximal left anterior descending artery (LAD), proximal left circumflex artery (LCx), and ramus intermedius branch, whilst the distal RCA, right or left posterior descending artery (PDA), right or left posterolateral branch (PL), distal LAD, and distal LCx were considered distal segments.8 Side branch segments included the right or left PDA, right or left PL, first or second diagonal branch, ramus intermedius branch, and first or second obtuse marginal branch, whilst the other 9 segments were considered non-side branch segments.
Serial 2D metrics
Per-lesion changes in coronary plaque, vessel, and lumen area were calculated at the cross-section of the minimal luminal area (MLA) between baseline and follow-up CCTA (Supplementary data online, Part A).5,13 Lesions were matched at baseline and follow-up using fiduciary landmarks such as distance from the ostium and side branches.5 Distance from the ostium to the cross-section of the MLA was numerically similar on both scans. Serial 2D arterial remodelling was defined as the cross-sectional change in vessel wall area (mm2) for each 1.00 mm2 change in plaque area. Likewise, serial 2D luminal narrowing was defined as the cross-sectional change in lumen area (mm2) for each 1.00 mm2 change in plaque area. Furthermore, these changes were evaluated according to baseline cross-sectional plaque burden with a cut-off of 40%, which was considered the Glagovian limit for outward arterial remodelling based on prior published reports.1,14 Cross-sectional plaque burden was calculated as plaque area/vessel wall area * 100%.
Study outcomes
The primary outcome was serial 3D arterial remodelling and luminal narrowing, defined as the per-segment change in respectively coronary vessel and lumen volume from a change in plaque volume between scans. Secondary outcomes were to evaluate serial arterial remodelling and luminal narrowing according to: (i) baseline PAV, (ii) baseline location within the coronary artery tree, and (iii) conventional 2D metrics.14
Statistical analysis
Continuous data are reported as means ± standard deviations or medians with IQRs, where appropriate on the basis of distribution. Categorical data are reported as counts with percentages. For continuous data, comparisons between baseline and follow-up CCTA were made with the paired T-test or Wilcoxon Signed Rank test. For categorical data, these comparisons were made with the McNemar’s test. To account for within patient correlation, univariate and multivariate generalized estimating equations were calculated to assess the association between serial per-segment or per-lesion CCTA measurements. Multivariate analysis included statins as a covariate to adjust for the potential effects of this therapy on coronary plaque formation (Supplementary data online, Part B). Only for side branch segments statins remained a significant covariate, and therefore only for this group, multivariate estimates are displayed. Strength and direction of association were calculated with Pearson correlation coefficients. For visual interpretation, scatterplots and forest plots were constructed. All statistical tests were two-sided and a P-value of <0.05 indicated statistical significance. All analyses were performed with R (version 3.6.1, R Development Core Team, Vienna, Austria) and SPSS software (version 26, SPSS IBM Corp., Armonk, NY, USA).
Results
Study population
Baseline characteristics of the study population are listed in Table 1. A total of 1245 patients (mean age 61 ± 9 years, 39% women) underwent serial CCTA with a median interscan interval of 3.3 years (IQR 2.6–4.8 years). Largely symptomatic patients with prevalent cardiac risk factors were included.
Baseline characteristics of PARADIGM registry
| n = 1245 | |
|---|---|
| Mean ± SD, median (IQR) or % | |
| Age, years | 61 ± 9 |
| Female | 39% |
| BMI, kg/m2 | 25.5 ± 3.3 |
| CCTA interscan interval, years | 3.3 (2.6–4.8) |
| Symptoms | |
| Typical angina | 5% |
| Atypical angina | 70% |
| Non-cardiac | 9% |
| Asymptomatic | 15% |
| Dyspnoea | 9% |
| Cardiac risk factorsa | |
| Hypertension | 57% |
| Dyslipidaemia | 41% |
| Diabetes mellitus | 22% |
| Family history of CAD | 28% |
| Smoking history | 39% |
| Cardiac medication | |
| Aspirin | 41% |
| Beta blockers | 29% |
| Calcium channel blockers | 24% |
| Renin–angiotensin system inhibitors | 31% |
| Statins | 46% |
| Laboratory findings | |
| Total cholesterol, mg/dLb | 188.0 (163.0–215.0) |
| Low-density lipoprotein, mg/dLb | 114.0 (90.0–138.0) |
| High-density lipoprotein, mg/dLb | 48.0 (41.0–58.0) |
| Triglycerides, mg/dL | 126.0 (91.0–181.5) |
| C-reactive protein, mg/dL | 0.5 (0.1–1.5) |
| HbA1c, % | 6.0 (5.6–6.8) |
| n = 1245 | |
|---|---|
| Mean ± SD, median (IQR) or % | |
| Age, years | 61 ± 9 |
| Female | 39% |
| BMI, kg/m2 | 25.5 ± 3.3 |
| CCTA interscan interval, years | 3.3 (2.6–4.8) |
| Symptoms | |
| Typical angina | 5% |
| Atypical angina | 70% |
| Non-cardiac | 9% |
| Asymptomatic | 15% |
| Dyspnoea | 9% |
| Cardiac risk factorsa | |
| Hypertension | 57% |
| Dyslipidaemia | 41% |
| Diabetes mellitus | 22% |
| Family history of CAD | 28% |
| Smoking history | 39% |
| Cardiac medication | |
| Aspirin | 41% |
| Beta blockers | 29% |
| Calcium channel blockers | 24% |
| Renin–angiotensin system inhibitors | 31% |
| Statins | 46% |
| Laboratory findings | |
| Total cholesterol, mg/dLb | 188.0 (163.0–215.0) |
| Low-density lipoprotein, mg/dLb | 114.0 (90.0–138.0) |
| High-density lipoprotein, mg/dLb | 48.0 (41.0–58.0) |
| Triglycerides, mg/dL | 126.0 (91.0–181.5) |
| C-reactive protein, mg/dL | 0.5 (0.1–1.5) |
| HbA1c, % | 6.0 (5.6–6.8) |
BMI, body mass index; CAD, coronary artery disease; CCTA, coronary computed tomography angiography.
Categorically assigned at each participating site in accordance with accurate guidelines at the time.
For reasons of uniformity, medians with interquartile ranges (IQRs) are displayed despite normal distribution.
Baseline characteristics of PARADIGM registry
| n = 1245 | |
|---|---|
| Mean ± SD, median (IQR) or % | |
| Age, years | 61 ± 9 |
| Female | 39% |
| BMI, kg/m2 | 25.5 ± 3.3 |
| CCTA interscan interval, years | 3.3 (2.6–4.8) |
| Symptoms | |
| Typical angina | 5% |
| Atypical angina | 70% |
| Non-cardiac | 9% |
| Asymptomatic | 15% |
| Dyspnoea | 9% |
| Cardiac risk factorsa | |
| Hypertension | 57% |
| Dyslipidaemia | 41% |
| Diabetes mellitus | 22% |
| Family history of CAD | 28% |
| Smoking history | 39% |
| Cardiac medication | |
| Aspirin | 41% |
| Beta blockers | 29% |
| Calcium channel blockers | 24% |
| Renin–angiotensin system inhibitors | 31% |
| Statins | 46% |
| Laboratory findings | |
| Total cholesterol, mg/dLb | 188.0 (163.0–215.0) |
| Low-density lipoprotein, mg/dLb | 114.0 (90.0–138.0) |
| High-density lipoprotein, mg/dLb | 48.0 (41.0–58.0) |
| Triglycerides, mg/dL | 126.0 (91.0–181.5) |
| C-reactive protein, mg/dL | 0.5 (0.1–1.5) |
| HbA1c, % | 6.0 (5.6–6.8) |
| n = 1245 | |
|---|---|
| Mean ± SD, median (IQR) or % | |
| Age, years | 61 ± 9 |
| Female | 39% |
| BMI, kg/m2 | 25.5 ± 3.3 |
| CCTA interscan interval, years | 3.3 (2.6–4.8) |
| Symptoms | |
| Typical angina | 5% |
| Atypical angina | 70% |
| Non-cardiac | 9% |
| Asymptomatic | 15% |
| Dyspnoea | 9% |
| Cardiac risk factorsa | |
| Hypertension | 57% |
| Dyslipidaemia | 41% |
| Diabetes mellitus | 22% |
| Family history of CAD | 28% |
| Smoking history | 39% |
| Cardiac medication | |
| Aspirin | 41% |
| Beta blockers | 29% |
| Calcium channel blockers | 24% |
| Renin–angiotensin system inhibitors | 31% |
| Statins | 46% |
| Laboratory findings | |
| Total cholesterol, mg/dLb | 188.0 (163.0–215.0) |
| Low-density lipoprotein, mg/dLb | 114.0 (90.0–138.0) |
| High-density lipoprotein, mg/dLb | 48.0 (41.0–58.0) |
| Triglycerides, mg/dL | 126.0 (91.0–181.5) |
| C-reactive protein, mg/dL | 0.5 (0.1–1.5) |
| HbA1c, % | 6.0 (5.6–6.8) |
BMI, body mass index; CAD, coronary artery disease; CCTA, coronary computed tomography angiography.
Categorically assigned at each participating site in accordance with accurate guidelines at the time.
For reasons of uniformity, medians with interquartile ranges (IQRs) are displayed despite normal distribution.
CCTA findings
CCTA findings of the 5721 analysed segments are listed in Table 2. At baseline CCTA, median PAV was 6.6% (IQR 1.0–17.9%) and median plaque, vessel, and lumen volume were 12.4 mm3 (IQR 1.7–35.2 mm3), 206.0 mm3 (IQR 126.5–315.9 mm3), and 180.0 mm3 (IQR 108.5–281.7 mm3), respectively. At follow-up CCTA, PAV, plaque volume, and vessel volume had increased significantly, while the lumen volume was reduced (P < 0.001 for all). In general, comparable findings were observed over time for the 3294 analysed lesions at the cross-section of the MLA (Supplementary data online, Part C).
Per-segment volumetric CCTA findings
| Baseline CCTA (n = 5721) | Follow-up CCTA (n = 5721) | P-valuea | |
|---|---|---|---|
| Median (IQR) | |||
| Volumes | |||
| Plaque volume, mm3 | 12.4 (1.7–35.2) | 24.0 (9.5–53.5) | <0.001 |
| Vessel volume, mm3 | 206.0 (126.5–315.9) | 210.8 (132.6–323.0) | <0.001 |
| Lumen volume, mm3 | 180.0 (108.5–281.7) | 170.9 (105.3–270.1) | <0.001 |
| PAV, % | 6.6 (1.0–17.9) | 12.3 (5.1–25.8) | <0.001 |
| Baseline CCTA (n = 5721) | Follow-up CCTA (n = 5721) | P-valuea | |
|---|---|---|---|
| Median (IQR) | |||
| Volumes | |||
| Plaque volume, mm3 | 12.4 (1.7–35.2) | 24.0 (9.5–53.5) | <0.001 |
| Vessel volume, mm3 | 206.0 (126.5–315.9) | 210.8 (132.6–323.0) | <0.001 |
| Lumen volume, mm3 | 180.0 (108.5–281.7) | 170.9 (105.3–270.1) | <0.001 |
| PAV, % | 6.6 (1.0–17.9) | 12.3 (5.1–25.8) | <0.001 |
CCTA, coronary computed tomography angiography; PAV, percent atheroma volume. a P-values are calculated with Wilcoxon Signed Rank test.
Per-segment volumetric CCTA findings
| Baseline CCTA (n = 5721) | Follow-up CCTA (n = 5721) | P-valuea | |
|---|---|---|---|
| Median (IQR) | |||
| Volumes | |||
| Plaque volume, mm3 | 12.4 (1.7–35.2) | 24.0 (9.5–53.5) | <0.001 |
| Vessel volume, mm3 | 206.0 (126.5–315.9) | 210.8 (132.6–323.0) | <0.001 |
| Lumen volume, mm3 | 180.0 (108.5–281.7) | 170.9 (105.3–270.1) | <0.001 |
| PAV, % | 6.6 (1.0–17.9) | 12.3 (5.1–25.8) | <0.001 |
| Baseline CCTA (n = 5721) | Follow-up CCTA (n = 5721) | P-valuea | |
|---|---|---|---|
| Median (IQR) | |||
| Volumes | |||
| Plaque volume, mm3 | 12.4 (1.7–35.2) | 24.0 (9.5–53.5) | <0.001 |
| Vessel volume, mm3 | 206.0 (126.5–315.9) | 210.8 (132.6–323.0) | <0.001 |
| Lumen volume, mm3 | 180.0 (108.5–281.7) | 170.9 (105.3–270.1) | <0.001 |
| PAV, % | 6.6 (1.0–17.9) | 12.3 (5.1–25.8) | <0.001 |
CCTA, coronary computed tomography angiography; PAV, percent atheroma volume. a P-values are calculated with Wilcoxon Signed Rank test.
Serial 3D arterial remodelling and luminal narrowing
For each 1.00 mm3 increase in plaque volume, the vessel volume increased by 0.71 mm3 [95% confidence interval (CI) 0.63 to 0.79 mm3, P < 0.001; r = 0.45, P < 0.001] (Figure 2). This increase was similar in segments with low and high baseline PAV (0.68 mm3 vs. 0.74 mm3, interaction P = 0.496). Conversely, a 0.29 mm3 reduction in lumen volume was observed for each 1.00 mm3 increase in plaque volume (95% CI −0.37 to −0.21 mm3, P < 0.001; r = −0.19, P < 0.001). Likewise, this reduction was similar in segments with low and high baseline PAV (−0.32 mm3 vs. −0.26 mm3, interaction P = 0.498). Furthermore, serial 3D arterial remodelling and luminal narrowing was evaluated according to baseline location within the coronary artery tree including left main, proximal and side branch status (Figure 3A and B). No differences were observed between left main and non-left main segments, proximal and distal segments and side branch and non-side branch segments (interaction P ≥ 0.281). Moreover, a strong and positive correlation was observed between the change in vessel volume and the change in lumen volume (r = 0.79, P < 0.001) (Supplementary data online, Part D). Finally, when our analysis was restricted to segments that showed exclusively progression of coronary atherosclerosis, a comparable pattern of serial 3D arterial remodelling and luminal narrowing was observed (Supplementary data online, Part E). Again, no differences were seen according to baseline PAV (interaction P ≥ 0.859) or location within the coronary artery tree (interaction P ≥ 0.206).
Serial 3D arterial remodelling and luminal narrowing. (Upper panel) Arterial remodelling is displayed for all segments (n = 5721) at the left panel and a comparison between segments with low (n = 2861) and high baseline PAV (n = 2860) at the middle and right panel, respectively. The cut-off for baseline PAV was set at the median, which was 6.6% (IQR 1.0–17.9%). (Lower panel) Luminal narrowing is displayed with similar comparisons according to baseline PAV. 3D, three-dimensional; PAV, percent atheroma volume.
Forest plots according to baseline location within the coronary artery tree. (A) Arterial remodelling ‘outward’ estimates of coronary plaque formation according to left main, proximal, and side branch status. No significant differences were observed. (B) Luminal narrowing ‘inward’ estimates of coronary plaque formation according to left main, proximal and side branch status. No significant differences were observed.
Serial 2D arterial remodelling and luminal narrowing
For each 1.00 mm2 increase in plaque area at the cross-section of the MLA, the vessel wall area increased by 0.97 mm2 (95% CI 0.92 to 1.01 mm2, P < 0.001; r = 0.80, P < 0.001) (Figure 4). This increase was similar in lesions with <40% and ≥40% baseline cross-sectional plaque burden (0.97 mm2 vs. 1.00 mm2, interaction P = 0.528). Inversely, a 0.03 mm2 reduction in lumen area was observed for each 1.00 mm2 increase in plaque area, which did not reach statistical significance (95% CI −0.08 to 0.01 mm2, P = 0.154; r = −0.04, P = 0.011). Also, when lesions were stratified into <40% and ≥40% baseline cross-sectional plaque burden, no association between the change in plaque area and the change in lumen area was observed (P = 0.479 and P = 0.868, respectively). Last, when our analysis was restricted to lesions with only progression of coronary atherosclerosis, a comparable pattern of serial 2D arterial remodelling and luminal narrowing was seen (Supplementary data online, Part F).
Serial 2D arterial remodelling and luminal narrowing. (Upper panel) Arterial remodelling at the cross-section of the MLA is displayed for all lesions (n = 3294) at the left panel and a comparison between lesions with <40% (n = 1675) and ≥40% baseline plaque burden (n = 1619) at the middle and right panel, respectively. (Lower panel) Luminal narrowing is displayed with similar comparisons according to baseline cross-sectional plaque burden. 2D, two-dimensional; MLA, minimal luminal area.
Discussion
We report on serial changes in atherosclerotic coronary plaque, revealing prominent outward arterial remodelling (70%) that co-occurs with modest luminal narrowing (30%) that was not affected by either baseline PAV or location within the coronary tree (Figure 5). Our findings remain the first in a contemporary series to provide additional insight into the compensatory mechanisms involved in the progression of coronary atherosclerosis, as initially put forth by Glagov et al.
Progression of coronary atherosclerosis in the proximal RCA. Progression of a mixed coronary plaque in the proximal RCA over time as assessed with serial CCTA and 3D metrics. Coronary plaque, vessel (orange border), and lumen dimensions (yellow border) are marked on both scans. Serial 3D changes in atherosclerotic coronary plaque reveal prominent positive or outward remodelling of approximately 70% that co-occurs with luminal narrowing of 30%. 3D, three-dimensional; CCTA, coronary computed tomography angiography; RCA, right coronary artery.
Static versus serial measurement
Preceding studies have calculated remodelling of coronary arteries using different methods.5,10,13–21 Though, static definitions have primarily been employed. Static arterial remodelling is calculated as a remodelling index: the ratio of the peak vessel wall area or diameter at the lesion site, as compared to an arbitrary proximal (or in some cases distal) reference site with minimal disease.11 Its threshold is set at 1.1 as has been validated against intravascular ultrasound (IVUS): <1.1 for intermediate or negative remodelling and ≥1.1 for positive remodelling.11,22 So far, the static evaluation of arterial remodelling in prior studies has revealed several limitations. First, the selection of a reference site with minimal disease remains challenging, particularly in cases with diffuse coronary atherosclerosis. Second, coronary arteries naturally taper, which by definition means that the vessel wall dimensions at the proximal reference site are larger than those at the more distal lesion site. Third, static arterial remodelling is considered an indirect measure of remodelling, since also vessel wall dimensions at the reference site can change during the dynamic atherosclerotic process.11 We surmise that direct evidence of arterial remodelling can only be derived from serial changes in vessel dimensions.12 Therefore, we applied standardized definitions of serial arterial remodelling and luminal narrowing. Hence, our analysis was the first to explore both 2D cross-sectional and 3D volumetric measurements of coronary plaque formation. The latter approach, by using all 3 dimensions of coronary plaque, vessel, and lumen, showed predominantly outward arterial remodelling with only modest luminal reduction over time.
Studies with serial 3D metrics
To date, limited IVUS studies have examined arterial remodelling and luminal narrowing over time with serial 3D measurements of coronary plaque, vessel, and lumen.23,24 Shiran et al.23 selected 31 patients, with a matching number of untreated moderate left main lesions, that underwent serial IVUS with a mean interval of 7.7 months. This study demonstrated that the percent change in atheroma volume correlated well with the percent change in external elastic membrane (EEM) volume (r = 0.448, P = 0.0115). Also, the percent change in lumen volume showed a strong and positive correlation with the percent change in EEM volume (r = 0.897, P < 0.0001). Moreover, Puri et al.24 studied 340 left main and proximal epicardial segments in patients with known CAD (i.e. ≥20% angiographic diameter stenosis in ≥1 epicardial artery) that underwent serial IVUS with a mean interval of 21 months. Likewise, this study showed a strong and positive correlation between the change in EEM volume and the change in lumen volume in both left main (r2 = 0.83, P < 0.001) and epicardial segments (r2 = 0.71, P < 0.001). Similarly, the present analysis demonstrated very similar strength and direction of correlations with serial CCTA: (i) a moderate and positive correlation amongst change in plaque and vessel volume, and (ii) a strong and positive correlation amongst change in vessel and lumen volume.
Studies with serial 2D metrics
Only a few CCTA studies have examined arterial remodelling and luminal narrowing over time with serial 2D measurements of coronary plaque, vessel, and lumen.25,26 From the Providing Regional Observations to Study Predictors of Events in the Coronary Tree-MSCT (PROSPECT-MSCT) study of 32 acute coronary syndrome (ACS) patients that after percutaneous coronary intervention (PCI) underwent serial CCTA with a mean interval of 38 months, Papadopoulou et al.25 analysed 129 segments without implanted stents or extensive calcification. In this specific post-ACS population, the majority of segments showed positive arterial remodelling over time. However, multiple IVUS studies investigated arterial remodelling and luminal reduction over time with serial cross-sectional metrics.14,27,28 From the Reversal of Atherosclerosis with Aggressive Lipid Lowering (REVERSAL) trial of 502 patients that underwent serial IVUS with an interval of 18 months, Sipahi et al.14 selected 128 focal lesions with angiographic diameter stenosis ≤50%, an increase in atheroma area at follow-up and no prior PCI, diffuse CAD or extensive calcifications. By comparing the lesion site at baseline and follow-up, it was demonstrated that each 1 mm2 increase in atheroma area produced a 1.28–1.62 mm2 increase in EEM area. Besides, as in contrast to earlier static histopathological observations of Glagov et al.,1 this increase in EEM area was not significantly different in lesions with <40% and ≥40% atheroma burden at baseline. Also, these serial observations remained valid after adjustment for low-density lipoprotein and C-reactive protein levels at follow-up, in an attempt to control for the potential effects of statins on arterial remodelling. In addition to this, Hartmann et al.27 studied 46 non-ostial left main lesions with angiographic diameter stenosis <30% in a corresponding number of patients that underwent serial IVUS with a mean interval of 18 months. Comparable results were observed by matching the lesion site at baseline and follow-up: for each 1 mm2 increase in atheroma area, the EEM area increased by 1.06 mm2. Furthermore, changes in EEM area were similar in lesions with <40% and ≥40% atheroma burden. The present analysis, which (i) was not limited to non-obstructive and/or left main lesions, and (ii) employed serial quantitative CCTA instead of IVUS measurements, was overall very consistent with the aforementioned findings. To this end, it should also be noted that CCTA is associated with low risk of complications, can provide assessment of the entire coronary artery tree and is able to analyse more severe stenoses that cannot be crossed by the IVUS catheter.
Clinical implications
We propose one hypothesis for the differences in the outward/inward ratios of coronary plaque formation between serial 3D and 2D metrics. Potentially, arterial remodelling at the cross-section of the MLA is more severe compared to a lesion or segment in its totality. Hence, cross-sectional vs. volumetric analysis possibly neglects the third dimension of coronary plaque—i.e. length—and therefore shows more exaggerated observations.
For consideration, in the current study, we performed a multivariate analysis including statins as a covariate to adjust for the potential effects of this therapy on serial arterial remodelling and luminal narrowing. By doing so, it was demonstrated that only in side branch segments statins remained a significant covariate. Prior research, comparing static remodelling indexes at baseline and follow-up IVUS interrogations, did observe constrictive arterial remodelling during therapy with statins.29 To this end, further research is warranted to establish how statins precisely alter serial arterial remodelling. Also, the effect of cardiac risk factors on serial arterial remodelling and luminal reduction deserves to be explored in future analyses.
Limitations
The present findings were part of a large observational cohort study with all its fundamental limitations including unmeasured confounding factors and selection bias. Patients with more rapid progression were excluded from our analysis due to the prespecified study design with an interscan interval of ≥2 years. Consequently, the current study population was at low-intermediate risk for CAD and demonstrated a relatively low prevalence of obstructive disease. Further research is necessary to examine if our findings, especially with regard to baseline plaque burden, are generalizable to a more high-risk or ‘sicker’ study population. To this end, the threshold of baseline plaque burden for predominant luminal narrowing vs. modest arterial remodelling was potentially not reached in our population. Last, quantitative analysis of coronary plaque, vessel, and lumen was performed with semi-automated software, which is currently investigational and not yet ready for routine clinical practice. Due to limitations in spatial resolution and image quality of contemporary CT scanners, it is still not feasible to quantitatively analyse coronary plaques <1 mm2. Our observations are therefore not generalizable to those plaques.
Conclusion
Although the Glagov phenomenon is well-accepted for detailing the natural history of coronary plaque progression, the evidence to date is incomplete and lacks contemporary validation in a cohort with suspected CAD. The current analysis from the PARADIGM registry demonstrates that serial CCTA measurements can be applied to define a pattern of coronary plaque formation over time. Particularly, serial 3D changes in atherosclerotic coronary plaque reveal prominent outward arterial remodelling that co-occurs with modest luminal narrowing. These findings support early work and provide insight into the magnitude of alterations in which coronary arteries remodel and narrow over time.
Supplementary data
Supplementary data are available at European Heart Journal - Cardiovascular Imaging online.
Data Availability
The data underlying this article cannot be shared publicly due to the patients' privacy. Data may be available upon reasonable request to the corresponding author, pending investigator approval of the study hypothesis and presence of an institutional data use agreement.
Funding
This work was supported by the Leading Foreign Research Institute Recruitment Program through the National Research Foundation (NRF) of Korea funded by the Ministry of Science and ICT (MSIT) (2012027176). The study was also funded in part by a generous gift from the Dalio Institute of Cardiovascular Imaging (New York, NY) and the Michael Wolk Foundation (New York, NY).
References
Author notes
Conflict of interest: J.K.M. has an equity interest in Cleerly. K.C. is a non-compensated medical advisory board member of Heartflow Inc. H.S. serves on the scientific advisory board of Philips, has equity interest in Covanos Inc., and has a research grant from Medtronic, Abbott Vascular and Philips. L.J.S. serves on the scientific advisory board for Covanos, Inc. The remaining authors have no relevant disclosures.
