Atherosclerosis 390 (2024) 117449

Contents lists available at ScienceDirect

Atherosclerosis
journal homepage: www.elsevier.com/locate/atherosclerosis

Comprehensive biomechanical and anatomical atherosclerotic plaque

metrics predict major adverse cardiovascular events: A new tool for clinical
decision making
Sophie Z. Gu a, #, Mona E. Ahmed b, c, #, Yuan Huang d, e, Diaa Hakim b, Charles Maynard f,
Nicholas V. Cefalo b, Ahmet U. Coskun g, Charis Costopoulos h, Akiko Maehara i,
Gregg W. Stone j, Peter H. Stone b, **, 1, Martin R. Bennett a, *, 1
a
Section of Cardiorespiratory Medicine, Department of Medicine, University of Cambridge, Cambridge, UK
b
Cardiovascular Division, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA
c
Department of Molecular Medicine and Surgery, Karolinska Institutet Karolinska University Hospital Solna, 171 76, Stockholm, Sweden
d
Centre for Mathematical and Statistical Analysis of Multimodal Imaging, University of Cambridge, Cambridge, UK
e
Department of Radiology, University of Cambridge, Cambridge, UK
f
Department of Health Services, University of Washington, Seattle, WA, USA
g
Mechanical and Industrial Engineering, Northeastern University, Boston, MA, USA
h
Department of Cardiology, Royal Papworth Hospital, Cambridge, UK
i
Cardiovascular Research Foundation, New York City, New York, USA
j
The Zena and Michael A. Wiener Cardiovascular Institute, Icahn School of Medicine at Mount Sinai, New York, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Background and aims: Anatomical imaging alone of coronary atherosclerotic plaques is insufficient to identify risk
Atherosclerotic plaques of future adverse events and guide management of non-culprit lesions. Low endothelial shear stress (ESS) and
Acute coronary syndrome high plaque structural stress (PSS) are associated with events, but individually their predictive value is insuffi­
Patient outcomes
cient for risk prediction. We determined whether combining multiple complementary, biomechanical and
Risk stratification
Intravascular ultrasound
anatomical plaque characteristics improves outcome prediction sufficiently to inform clinical decision-making.
Methods: We examined baseline ESS, ESS gradient (ESSG), PSS, and PSS heterogeneity index (HI), and plaque
burden in 22 lesions that developed subsequent events and 64 control lesions that remained quiescent from the
PROSPECT study.
Results: 86 fibroatheromas were analysed from 67 patients. Lesions with events showed higher PSS HI (0.32 vs.
0.24, p<0.001), lower local ESS (0.56Pa vs. 0.91Pa, p = 0.007), and higher ESSG (3.82 Pa/mm vs. 1.96 Pa/mm, p
= 0.007), while high PSS HI (hazard ratio [HR] 3.9, p = 0.006), high ESSG (HR 3.4, p = 0.007) and plaque
burden>70 % (HR 2.6, p = 0.02) were independent outcome predictors in multivariate analysis. Combining low
ESS, high ESSG, and high PSS HI gave both high positive predictive value (80 %), which increased further
combined with plaque burden>70 %, and negative predictive value (81.6 %). Low ESS, high ESSG, and high PSS
HI co-localised spatially within 1 mm in lesions with events, and importantly, this cluster was distant from the
minimum lumen area site.
Conclusions: Combining complementary biomechanical and anatomical metrics significantly improves risk-
stratification of individual coronary lesions. If confirmed from larger prospective studies, our results may
inform targeted revascularisation vs. conservative management strategies.

* Corresponding author. Section of Cardiorespiratory Medicine, University of Cambridge, Heart & Lung Research Institute, Papworth Road, Cambridge Biomedical
Campus, Cambridge, CB2 0BB, UK.
** Corresponding author. Harvard Medical School, Director, Vascular Profiling Research Group, Division of Cardiovascular Medicine, Brigham & Women’s Hospital,
75 Francis Street, Boston, MA, 02115, USA.
E-mail addresses: [email protected] (P.H. Stone), [email protected] (M.R. Bennett).
#
These authors contributed equally as joint first authors.
1
These authors contributed equally as joint senior authors.

https://doi.org/10.1016/j.atherosclerosis.2024.117449
Received 30 October 2023; Received in revised form 18 December 2023; Accepted 9 January 2024
Available online 11 January 2024
0021-9150/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
S.Z. Gu et al. Atherosclerosis 390 (2024) 117449

1. Introduction methods from coronary angiography and IVUS to reconstruct coronary

artery lumens and perform computational fluid dynamics (CFD) calcu­
Despite advances in medical and interventional therapies, future lations (Supplementary Methods) [10]. Coronary angiographic and
major adverse cardiovascular events (MACE) such as cardiovascular IVUS centrelines were merged for 3D coronary artery reconstruction
death or re-infarction following intervention in acute coronary syn­ [11], and entire reconstructed arteries divided into serial 3-mm seg­
dromes (ACS) are 8.1–15.8 % and 10.0–14.5 %, respectively, up to 5 ments to calculate local ESS, ESSG, plaque area, external elastic mem­
years [1,2]. Many events are due to destabilisation of non-culprit lesions brane (EEM) area, lumen area, PB, and arterial remodelling index. The
(NCLs), but the prognostic value of anatomy-based plaque risk assess­ lowest and highest local ESS site was determined along the whole lesion
ment alone is disappointing. For example, combining high-risk in moving 90◦ arcs around the vessel circumference and related spatially
anatomical features such as large plaque burden (PB), small minimum to the MLA. Since developed flow is established after an entrance length
lumen area (MLA), thin-cap fibroatheroma (TCFA), and lipid core of 3 mm, any lesion with a length less than 9 mm is considered too short
burden index predicts only 18.2–23.1 % of NCL MACE over 3.4 years by for any meaningful ESS evaluation [11]. We excluded bifurcations and
intravascular ultrasound [3–5], 18.9 % by optical coherence tomogra­ side branches from our CFD analyses, as do most investigators, because
phy [6], 19.3 % by near-infrared spectroscopy [7], and 4.1 % by coro­ of technical limitations of their segmentation, quantification of blood
nary computed tomography angiography [8]. flow, and 3D reconstruction necessary for CFD [18–20]. Our methods
Endothelial shear stress (ESS) describes the frictional force of blood have been previously validated and have excellent reproducibility [21].
acting on arterial wall endothelial cells [9], and low ESS is associated
with plaque development, progression, and destabilisation [10,11]. 2.2.2. Local plaque structural stress computation
Similarly, high ESS, particularly high ESS gradient (ESSG) due to abrupt Vessel geometry and plaque composition were extracted from VH-
changes in plaque-lumen interface topography, can also promote plaque IVUS data and imported into proprietary software (MATLAB R2022b,
disruption [12]. High plaque structural stress (PSS), which is determined MathWorks, Inc, Massachusetts), allowing construction of 2D solid
by material properties and spatial relationships of individual plaque models. Frames containing side branches or immediately adjacent to
components [9,13], and its longitudinal heterogeneity (PSS–HI) also bifurcations were excluded as they violated the plane strain assumption
promote plaque progression and vulnerability [14–17]. However, while for 2D solids. PSS was calculated using dynamic finite element analysis
single biomechanical plaque metrics enhance prognostication of indi­ (FEA) as described previously (Supplementary Methods) [13,22], with
vidual coronary plaques, their predictive accuracy is not sufficient to PSS heterogeneity index (HI) calculated as standard deviation/mean PSS
guide clinical management. across the whole plaque. Maximum peri-luminal region principal stress
We examined a unique and comprehensive combination of comple­ was used to indicate critical mechanical conditions.
mentary biomechanical and anatomical plaque metrics to determine Biomechanical analysis focused on lesion sites, but we also calcu­
their prognostic synergy. Specific combinations of local ESS, ESSG, PSS, lated ESSmin, ESSG, and PSS HI10mm at regions with PB<40 % adjacent to
and PSS-HI with high-risk anatomical plaque features significantly significant plaque sites. MACE during follow-up were analysed against
improve NCL risk-stratification, potentially informing clinical decisions baseline ESS, PSS and PROSPECT anatomical characteristics (PB, MLA,
such as targeted revascularisation vs. conservative management strate­ and morphology (TCFA and ThCFA)) [3]. Low ESS was defined as <1.3
gies for individual lesions. Pa, a well-accepted definition for low/physiological ESS associated with
NCL-MACE [11].
2. Patients and methods
2.3. Statistical analysis
The Providing Regional Observations to Study Predictors of Events in
the Coronary Tree (PROSPECT; NCT00180466) trial is a prospective Data variables are presented as mean ± standard deviation (SD) or
natural history study of coronary atherosclerosis [3]. 697 ACS patients median (interquartile range, IQR). Numerical variables were compared
underwent successful culprit lesion(s) intervention, followed by baseline using Mann-Whitney U tests and categorical variables using chi-square
angiography and 3-vessel virtual histology intravascular ultrasound tests. Outliers were removed using the median absolute deviation
(VH-IVUS) [3]. Vessel analysis was performed offline by an independent method with threshold 3.5. As each plaque had multiple VH-IVUS
core laboratory, and patients followed for a median of 3.4 years. MACE frames, linear mixed-effects (LME) models were used to account for
comprised the composite of cardiac death, myocardial infarction or hierarchical data structure and clustering in individual patients with
rehospitalization due to progressive or unstable angina. A clinical events results presented as mean ± standard error (SE). Model diagnostics were
committee adjudicated each suspected MACE with no knowledge of performed by inspecting residual and Q-Q plots to test model assump­
other patient data. Events were attributed to culprit lesions or NCLs tions. Association between continuous variables was assessed by Pear­
based on follow-up angiography. The study protocol conforms to the son’s correlation coefficient and linear regression. Receiver-operating
1975 Declaration of Helsinki ethical guidelines, was approved by each characteristics (ROC) curve analysis was performed for pre-defined
institutional review board, and all participants provided written biomechanical variables to identify the best cut-off that predicted
informed consent. MACE, which was subsequently used to categorize low and high ESS,
ESSG, and PSS heterogeneity. Time-to-event curves are presented as
2.1. VH-IVUS image acquisition and analysis Kaplan-Meier estimates of cumulative hazard and compared using the
log-rank method. Individual ESS and PSS predictors of MACE were
VH-IVUS image acquisition, analysis and plaque classification were identified using Cox regression analysis and statistically significant
as described previously [3] (Supplementary Methods). In brief, plaques variables entered in a multivariable model to identify independent
were defined as ≥3 consecutive frames with PB ≥ 40 %. As thin and predictors. Proportional hazards assumption was tested by checking
thick cap fibroatheromas (TFCA and ThCFA) caused 88 % of all MACE, interaction between the time measure and covariates. Chi-square test
we focused on these lesions. was used for differences in frequency distribution of biomechanical
metrics. Statistical significance was indicated by two-tailed p<0.05.
2.2. Biomechanical analysis Analyses were performed in SPSS (version 27, IBM, New York) and R
(version 4.2.2, R Foundation for Statistical Computing, Vienna).
2.2.1. Local endothelial shear stress computation
MACE and randomly-selected non-MACE NCLs were studied at ~1:3
ratio. Local plaque ESS was assessed using validated vascular profiling

2
S.Z. Gu et al. Atherosclerosis 390 (2024) 117449

Fig. 1. Study design workflow.

(A) Lesion selection, (B) plaque structural stress (PSS) and endothelial shear stress (ESS) calculation on co-registered coronary lesions. CFD = computational fluid
dynamics; ESSG = ESS gradient; FA = fibroatheroma; FEA = finite element analysis; MACE = major adverse cardiovascular events; nc = non-culprit; NCL = nc lesion;
VH-IVUS = virtual histology intravascular ultrasound.

3
S.Z. Gu et al. Atherosclerosis 390 (2024) 117449

Table 1
Cox proportional hazard regression: Factors associated with MACE.
Univariate Analysis Multivariate Analysis Adjusted Modela

HR (95 % CI) p value HR (95 % CI) p value HR (95 % CI) p value

Low ESSmin 32.5 (0.57–1843) 0.091 – - - -

High ESSmeanb 2.27 (0.95–5.42) 0.065 – - - -
High ESSG 3.96 (1.71–9.18) 0.001 3.36 (1.40–8.07) 0.007 3.34 (1.39–8.01) 0.007
High PSS 0.58 (0.20–1.70) 0.316 – – – –
High PSS HI 5.03 (1.96–12.9) <0.001 3.91 (1.47–10.4) 0.006 3.86 (1.45–10.27) 0.007
MLA≤4 mm2 2.89 (1.25–6.70) 0.013 – - - -
PB ≥ 70 % at MLA 4.35 (1.88–10.1) <0.001 2.60 (1.07–6.31) 0.035 2.59 (1.07–6.29) 0.035

CI = confidence interval; ESS = endothelial shear stress; ESSG = ESS gradient; HI = heterogeneity index; HR = hazard ratio; MACE = major adverse cardiovascular
events; MLA = minimum lumen area; PB = plaque burden; PSS = plaque structural stress.
a
Age, sex and smoking status adjusted model.
b
High ESSmean cut-off = 2.7Pa, determined by Youden index on receiver operating characteristic analysis.

3. Results ESSmin <1.3 kPa, but so did many non-MACE plaques (Supplementary
Fig. 2A). High PSS HI combined with low ESSmin best predicted MACE vs.
3.1. Baseline patient and anatomical lesion characteristics non-MACE lesions (72.7 % vs. 17.2 %, p<0.001) (Supplementary
Fig. 2A), compared with 50 % vs. 10.9 % (p<0.001) for low ESSmin/high
Fifty-one NCLs caused subsequent MACE in PROSPECT, including 8 ESSG, and 36.4 % vs. 3.1 % for high ESSG/high PSS HI (p<0.001)
non-fibroatheromas, which were excluded. Two lesions were too short (Supplementary Figs. 2B–C). Similarly, combined high ESSG/high PSS
(<9 mm) for any meaningful ESS or PSS heterogeneity evaluation. ESS HI identified fewer MACE vs. non-MACE lesions if only lesions with a
vascular profiling CFD analyses require specific biplane coronary an­ low baseline ESS (<1.3Pa) were studied (36.4 % vs. 4.5 %, p = 0.002)
giograms to reconstruct coronary 3D structure and to allow the merging (Supplementary Fig. 2D).
of artery centreline with IVUS imaging. There were 19 (37 %) lesions Of all biomechanical and anatomical parameters, only high ESSG,
with insufficient imaging to perform CFD [11] and were excluded high PSS HI, smaller MLA and PB>70 % were associated with MACE on
(Fig. 1A). The final dataset comprised 86 NCLs (22 MACE plaques from univariate analysis (Table 1). High ESSG (hazard ratio [HR] 3.36, 95 %
20 patients, 2 MIs, 4 unstable angina, 14 rehospitalization; 64 confidence interval [CI] 1.40–8.07, p = 0.007), high PSS HI (HR 3.91,
non-MACE plaques from 47 patients), and patient demographics were 95 % CI 1.47–10.4, p = 0.006), and large PB (HR 2.60, 95 % CI
well matched with the full PROSPECT cohort, apart from more current 1.07–6.31, p = 0.035) were also independent predictors of MACE in
smokers (63.6 % vs. 47.7 %, p = 0.02) (Supplementary Table 1). Patients multivariable analyses. Similar results were seen in an age-, sex- and
had a median age of 57.5 years, 83.6 % were male and patient de­ smoking-adjusted multivariate model (Table 1).
mographics were well matched between MACE (n = 20) and non-MACE
(n = 47) groups (Supplementary Table 2). 3.3. Incorporating PSS heterogeneity and ESSG improves prediction of
Lesions were distributed across all main coronary arteries, with 51.2 MACE
% in the proximal vessel (Supplementary Table 3). Overall, 64 % were
TCFAs, with similar TCFA and ThCFA proportions in MACE and non- While MACE only occurred in plaques with baseline low ESSmin
MACE groups. MACE plaques had significantly smaller MLA (3.83 vs. (<1.3Pa), 77.2 % of non-MACE plaques had focal areas with low ESSmin
5.05 mm2, p = 0.002), larger PB at MLA (70.0 % vs. 59.4 %, p<0.001), (Supplementary Fig. 2A). However, combining high ESSG or high PSS HI
and were longer (29.0 vs. 19.8 mm, p = 0.001). By VH-IVUS, MACE individually with low ESSmin increased cumulative NCL-MACE proba­
plaques had significantly larger overall PB (54.1 vs. 50.9 %, p = 0.02); bilities from 33 % to 64.6 % and 61.5 %, respectively, in all NCL
otherwise, EEM, lumen or plaque areas or plaque composition were fibroatheromas (Fig. 2A–B), and from 32.5 % to 87.3 % and 59.8 %,
similar (Supplementary Table 3). respectively, in TCFAs (Fig. 2C–D). Combining high ESSG/high PSS HI
with low ESSmin increased cumulative MACE probabilities further
3.2. Association of baseline ESS, ESS gradient and PSS heterogeneity with (Fig. 2E), with observed MACE rates of individual plaques increasing
future MACE with number of high-risk metrics present. For example, NCL MACE rates
(Fig. 3A) and plaque Hazard Ratios (HR) (Fig. 3B) were higher in pla­
Baseline ESS and PSS parameters were calculated on 55 TCFA and 31 ques with a low baseline ESS combined with a high baseline ESSG, a high
ThCFA co-registered lesions (Fig. 1B). MACE plaques had lower baseline PSS HI, or both. Indeed, lesions containing these three biomechanical
local minimum ESS (ESSmin, 0.56 vs. 0.91Pa, p = 0.007), higher local metrics showed 80 % MACE rates, and lesions containing all three
maximum ESS (ESSmax, 5.81 vs. 4.37Pa, p = 0.011) and higher ESS biomechanical metrics plus PB ≥ 70 % had 100 % MACE rates (HR 8.29)
gradient (ESSG, 3.82 vs. 1.96 Pa/mm, p = 0.007) compared to non- (Fig. 3A–B). As expected, the prevalence of these progressively higher
MACE plaques (Supplementary Table 4). MACE plaques had lower risk combinations was progressively less frequent.
maximum PSS (PSSmax, 115.5 vs. 124.4 kPa, p = 0.038) in this un­
matched cohort consistent with their higher PB and lower MLA (ac­ 3.4. Plaque risk stratification
cording to Laplace’s law), but higher PSS heterogeneity index (PSS HI,
0.321 vs. 0.241, p<0.001 (Supplementary Table 4). The progressive increase in MACE rates in plaques with more
ROC analysis was performed to identify ESS and PSS parameters biomechanical features provides a valuable platform to find the best
thresholds that best predicted future MACE, and used to classify each sequence of metrics to predict MACE in NCLs. We developed a stepwise
parameter into ‘low’ and ‘high’ indices. Cumulative MACE probabilities decision tree based on PPV and NPV to both identify and exclude high-
were higher in plaques with lower local ESSmin (≤1.3Pa, p = 0.005), risk plaques. Combining high ESSG and low ESSmin improved PPV to
higher ESS gradient (ESSG, ≥4.0 Pa/mm, p< 0.001), or higher PSS HI 61.1 % with a NPV of 83.8 % (Fig. 3C). Adding high PSS HI to low ESSmin
(≥0.29, p< 0.001)(Supplementary Fig. 1). However, the ability to increased PPV from 33.3 % to 59.3 %, maintaining a high NPV at 89.8 %,
identify MACE lesions varied across different combinations of biome­ but combining PSS HI, ESSG, and ESSmin further improved PPV to 80 %,
chanical parameters. For example, all MACE plaques had a baseline still maintaining a high NPV at 81.6 % (Fig. 3C). Similarly, different

4
S.Z. Gu et al. Atherosclerosis 390 (2024) 117449

Fig. 2. Incorporation of high ESSG and high PSS HI improves risk stratification of low ESS lesions.
Cumulative MACE probability in lesions with low ESS and high or low (A) ESSG, (B) PSS HI, in TCFA lesions with high or low (C) ESSG, (D) PSS HI, or (E) com­
bination of ESS, ESSG and PSS HI. ESS = endothelial shear stress; ESSG = ESS gradient; HI = heterogeneity index; MACE = major adverse cardiovascular events; PSS
= plaque structural stress; TCFA = thin cap fibroatheroma.

combinations of biomechanical features were associated with lesser or coronary interventions (PCI). Low ESSmin and high ESSG sites were
even no risk, with no events with physiologic/high local ESS (≥1.3Pa). identified in each plaque, together with the highest PSS HI over a 10 mm
plaque segment (PSS HI10mm) 5 mm proximal and distal to a reference
point moving along each plaque (Fig. 4A–B). 95 % of MACE plaques
3.5. Spatial relationship of high-risk features in individual plaques and exhibited two or more higher-risk features (low ESSmin, high ESSG, high
their relationship to the MLA PSS HI10mm) clustered within 1 mm of each other, compared with only
36 % of non-MACE plaques (p<0.0001) (Fig. 4C). Importantly, the
We further examined whether the 3 prognostic biomechanical met­ mean distance between the colocalised high-risk features and the MLA
rics were clustered spatially or located distant from each other, and their was ≥13 mm in both MACE and non-MACE plaques.
relationship to the MLA, the site targeted by most percutaneous

5
S.Z. Gu et al. Atherosclerosis 390 (2024) 117449

Fig. 3. Combined ESS, ESS gradient (ESSG) and plaque structural stress (PSS) heterogeneity index (HI) improves MACE prediction.
(A) Observed MACE rates in different combinations of biomechanical metrics. (B) Hazard Ratio (HR) and feature prevalence at lesion level. (C) Stepwise decision tree
for plaque risk stratification based on predictive values. CI = confidence interval; MACE = major adverse cardiovascular events; NPV = negative predictive value;
PPV = positive predictive value.

6
S.Z. Gu et al. Atherosclerosis 390 (2024) 117449

Fig. 4. MACE lesions exhibit more spatially colocalised biomechanical metrics.

(A) Example of a MACE plaque, showing the plaque thickness, lumen radius, ESS, ESSG and the PSS HI10mm profile. (B) Representative example of co-localization of
low ESSmin and high PSS HI10mm in the same MACE plaque, with their spatial relationship indicated by the distance between the red and green dashed lines. (C)
Different combinations of biomechanical features (low ESSmin<1.3Pa, high ESSG≥4 Pa/mm, and high PSS HI10mm > 0.29) clustering together in MACE plaques vs.
no-MACE plaques. ESS = endothelial shear stress; ESSG = ESS gradient; HI = heterogeneity index; MACE = major adverse cardiovascular events; MLA = minimal
lumen area; PSS = plaque structural stress. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

4. Discussion synergistic high-risk biomechanical features cluster within a plaque,

which are typically located distant from the MLA.
Our study is the first to comprehensively examine the prognostic Early prognostic studies found that anatomical characteristics,
significance of multiple combinations of known baseline high-risk and including large PB, small MLA, TCFA morphology, and large lipid core
mechanistically complementary biomechanical and anatomical metrics were associated with future MACE [3,7,23]. We similarly found that
in non-culprit coronary plaques. We observed that: (1) each biome­ smaller MLA and PB>70 % were predictors in univariate analysis in our
chanical parameter (ESSmin, ESSG, PSS HI) alone increased MACE pre­ subgroup, and larger PB in multivariate analysis. TCFA morphology was
diction; (2) high ESSG, high PSS HI, and high PB were independent not a significant event predictor, but this may be due to our excluding
predictors of MACE; (3) combining baseline low ESS, high ESSG and lower risk lesions (PIT, fibrocalcific and fibrotic plaques) and the
high PSS HI provided the highest prediction (88.9 %) of future MACE, generally low event rate of both TCFA and non-TCFA lesions (4.9 % vs.
with even higher prediction if combined with high PB; (4) different 1.3 %) in PROSPECT. We also found that low ESS was not an indepen­
metrics combinations segregated lesions into high-, intermediate- and dent predictor of MACE, as observed in the PREDICTION study [10],
low-risk groups that could guide clinical decisions for targeted revas­ which included lower-risk plaques and patients. Our current study was
cularisation or conservative treatment, and (5) complementary and limited to non-culprit higher-risk fibroatheromas in ACS patients. In

7
S.Z. Gu et al. Atherosclerosis 390 (2024) 117449

Fig. 5. Graphical abstract.

Combined biomechanical parameters including endothelial shear stress (ESS), ESS gradient (ESSG) and plaque structural stress (PSS) heterogeneity index (HI)
improves prediction of adverse events in patients.

contrast to some studies [24,25], high ESS (i.e. high ESSmean, cut-off = biomechanical and anatomical features (Fig. 3D). No lesion with a
2.7Pa determined by Youden index on ROC analysis) was also not an physiological/high ESS (≥1.3Pa) alone led to MACE during a median
independent predictor of future MACE. A luminal stenosis creates both follow-up of 3.4 years, suggesting that baseline ESS could act as the first
high shear stress at the MLA and low shear stress up and down stream to screen. High ESS gradient and high PSS HI were additive in
the MLA. The metric of “average” plaque ESS used in those studies may risk-stratifying lesions, leading to a stepwise decision tree to aid plaque
miss highly focal areas of low ESS (<~1.0Pa), located adjacent to high risk-stratification. Plaques with low ESS, high ESSG and high PSS HI had
ESS areas (>5Pa), which becomes subsumed by averaging. In contrast, a PPV of ≥80 %, whereas different combinations of parameters could
the metric “ESSG” reflects both the combined pathobiology of high ESS identify intermediate-risk lesions. Heterogeneity is a consequence of the
and the adjacent low ESS, and thereby provides a powerful independent dynamic and variable plaque natural history and vascular remodelling
predictor. along the course of each lesion [30], resulting in changes in plaque and
Although previous studies have shown that individual biomechan­ lumen morphology [31], ESS, ESSG, and PSS within individual plaques.
ical metrics (low ESS or high PSS HI) improve MACE prediction when A comprehensive plaque risk assessment thus requires careful evalua­
combined with high-risk anatomical variables [10,11,17], their predic­ tion of multiple plaque and arterial wall features along the entire lesion,
tive accuracy was still low. Our current results show that incorporating and the co-localised highest-risk portion of the plaque may be distant
mechanistically complementary biomechanical metrics together with from the MLA.
anatomical large PB dramatically improves MACE prediction, identi­
fying almost all future MACE plaques over 3.4 years of follow-up. The
additive effect of these three biomechanical variables is mechanistically 4.1. Limitations
reinforced by their each being independent predictors of MACE. These
destabilizing features are spatially located near each other in MACE This study has some important limitations: (1) Although we studied
plaques compared to similar no-MACE plaques, and likely represent a one of the largest prospectively enrolled patient cohorts of the natural
local “perfect storm” of high-risk features of both plaque anatomical history of atherosclerosis, only 54 % of MACE NCLs were technically
vulnerability (PSS, PSS HI, PB) and local haemodynamic destabilizing suitable for local ESS calculation, potentially introducing some selection
influences (ESS, ESSG). Other biomechanical stresses, such as axial bias. However, a large number of randomly selected, well-matched
plaque stress [26] or other local flow disturbances [27,28], may further control (non-MACE) plaques were evaluated, and the additive value of
increase the prognostic ability of these combined biomechanical and combined biomechanical variables was consistently demonstrated in
anatomical plaque risk features. Higher-risk biomechanical features multiple analyses. In addition, patients in our study are well-matched
co-localised spatially, and our finding that these features are relatively with the main PROSPECT cohort (Supplementary Table 2). (2) This is
distant from the MLA in most future MACE plaques is important clini­ a retrospective, post-hoc analysis of a specific patient population from
cally. Standard clinical practice deploying coronary stents around the the PROSPECT trial, limiting the generalisability of our results.
MLA may not cover areas of adverse biomechanical and anatomical Although it provides an invaluable proof-of-concept incorporating all of
plaque features located substantially proximal or distal to the MLA [29]. the currently available biomechanical and anatomical prognostic met­
Coronary atherosclerotic plaques are complex, lengthy, and hetero­ rics, large-scale prospective studies involving a wider patient population
geneous, and accumulating evidence suggests that MACE are related less with these or other biomechanical/anatomical metrics will be required,
to flow-limiting stenosis than overall plaque burden, be it obstructive or assessing both clinical and cost-effectiveness outcomes. (3) Side
non-obstructive [29]. Our findings suggest the potential value of branches were not considered in our CFD analysis, which could alter the
screening all portions of atherosclerotic plaques for higher-risk ESS calculation. However, such single-conduit models have reasonable
test accuracy to rule out pathological ESS (sensitivity 67.7 %, specificity

8
S.Z. Gu et al. Atherosclerosis 390 (2024) 117449

90.7 %) [19], and the effect of side branches on ESS can be very local­ References
ised, and generally <5 % if outflow difference is small [32]. (4) Both ESS
and PSS can be calculated from fully-coupled fluid structure interaction [1] T. Jernberg, P. Hasvold, M. Henriksson, H. Hjelm, M. Thuresson, M. Janzon,
Cardiovascular risk in post-myocardial infarction patients: nationwide real world
(FSI) models, but this requires significantly higher computational power data demonstrate the importance of a long-term perspective, Eur. Heart J. 36
and may provide little incremental prognostic value. (5) Finally, ESS and (2015) 1163–1170.
PSS features were examined by co-registering plaque/artery segments [2] D.L. Steen, I. Khan, K. Andrade, A. Koumas, R.P. Giugliano, Event rates and risk
Factors for recurrent cardiovascular events and mortality in a contemporary post
and frames, potentially introducing co-registration errors. However, acute coronary syndrome population representing 239 234 patients during 2005 to
plaque outcome prediction analysis was performed on whole-plaque 2018 in the United States, J. Am. Heart Assoc. 11 (2022) e022198.
level data (i.e. minimum ESS, maximum ESSG, and PSS HI across the [3] G.W. Stone, A. Maehara, A.J. Lansky, B. De Bruyne, E. Cristea, G.S. Mintz,
R. Mehran, J. McPherson, N. Farhat, S.P. Marso, H. Parise, B. Templin, R. White,
whole plaque) to minimize the effects of individual frame misalignment. Z. Zhang, P.W. Serruys, A prospective natural-history study of coronary
atherosclerosis, N, Engl. J. Med. 364 (2011) 226–235.
[4] P.A. Calvert, D.R. Obaid, M. O’Sullivan, L.M. Shapiro, D. McNab, C.G. Densem, P.
4.2. Conclusions M. Schofield, D. Braganza, S.C. Clarke, K.K. Ray, N.E.J. West, M.R. Bennett,
Association between IVUS findings and adverse outcomes in patients with coronary
Comprehensive combinations of mechanistically synergistic ESS and artery disease, J Am Coll Cardiol Img 4 (2011) 894–901.
[5] J.M. Cheng, H.M. Garcia-Garcia, S.P.M. De Boer, I. Kardys, J.H. Heo, K.
PSS and anatomical parameters at baseline were associated with most M. Akkerhuis, R.M. Oemrawsingh, R.T. Van Domburg, J. Ligthart, K.T. Witberg,
future MACE from NCLs in PROSPECT (Fig. 5). Identification of the site E. Regar, P.W. Serruys, R.J. Van Geuns, E. Boersma, In vivo detection of high-risk
and magnitude of these biomechanical and anatomical metrics may coronary plaques by radiofrequency intravascular ultrasound and cardiovascular
outcome: results of the ATHEROREMO-IVUS study, Eur. Heart J. 35 (2014)
represent a new tool to guide revascularisation strategies to high-risk 639–647.
plaques, while also providing reassurance for more conservative man­ [6] F. Prati, E. Romagnoli, L. Gatto, A. La Manna, F. Burzotta, Y. Ozaki, V. Marco,
agement for low-risk lesions. A. Boi, M. Fineschi, F. Fabbiocchi, N. Taglieri, G. Niccoli, C. Trani, F. Versaci,
G. Calligaris, G. Ruscica, A. Di Giorgio, R. Vergallo, M. Albertucci, G. Biondi-
Zoccai, C. Tamburino, F. Crea, F. Alfonso, E. Arbustini, On behalf of C.
Trial registration Investigators, Relationship between coronary plaque morphology of the left
anterior descending artery and 12 months clinical outcome: the CLIMA study, Eur.
Heart J. 41 (2019) 383–391.
ClinicalTrials.gov Identifier: NCT00180466. [7] R. Waksman, C. Di Mario, R. Torguson, Z.A. Ali, V. Singh, W.H. Skinner, A.K. Artis,
T. Ten Cate, E. Powers, C. Kim, E. Regar, S.C. Wong, S. Lewis, J. Wykrzykowska,
S. Dube, S. Kazziha, M. van der Ent, P. Shah, P.E. Craig, Q. Zou, P. Kolm, H.
Declaration of competing interest
B. Brewer, H.M. Garcia-Garcia, H. Samady, J. Tobis, M. Zainea, W. Leimbach,
D. Lee, T. Lalonde, W. Skinner, A. Villa, H. Liberman, G. Younis, R. de Silva,
The authors declare that they have no known competing financial M. Diaz, L. Tami, J. Hodgson, G. Raveendran, N. Goswami, J. Arias, L. Lovitz,
interests or personal relationships that could have appeared to influence R. Carida II, S. Potluri, F. Prati, A. Erglis, A. Pop, M. McEntegart, M. Hudec,
U. Rangasetty, D. Newby, Identification of patients and plaques vulnerable to
the work reported in this paper. future coronary events with near-infrared spectroscopy intravascular ultrasound
imaging: a prospective, cohort study, Lancet 394 (2019) 1629–1637.
[8] M.C. Williams, A.J. Moss, M. Dweck, P.D. Adamson, S. Alam, A. Hunter, A.S.
Financial support
V. Shah, T. Pawade, J.R. Weir-McCall, G. Roditi, E.J.R. van Beek, D.E. Newby, E.
D. Nicol, Coronary artery plaque characteristics associated with adverse outcomes
This work was funded by British Heart Foundation (BHF) grants FS/ in the SCOT-HEART study, J. Am. Coll. Cardiol. 73 (2019) 291–301.
[9] A.J. Brown, Z. Teng, P.C. Evans, J.H. Gillard, H. Samady, M.R. Bennett, Role of
19/66/34658, PG/16/24/32090, RG71070, RG84554, the BHF Centre
biomechanical forces in the natural history of coronary atherosclerosis, Nat. Rev.
for Research Excellence (London, UK), the National Institute of Health Cardiol. 13 (2016) 210–220.
Research Cambridge Biomedical Research Centre (London, UK), Na­ [10] P.H. Stone, S. Saito, S. Takahashi, Y. Makita, S. Nakamura, T. Kawasaki,
tional Institute of Health (US) grants R01 HL146144-01A1, R01 A. Takahashi, T. Katsuki, S. Nakamura, A. Namiki, A. Hirohata, T. Matsumura,
S. Yamazaki, H. Yokoi, S. Tanaka, S. Otsuji, F. Yoshimachi, J. Honye, D. Harwood,
HL140498, the Swedish Heart-Lung Foundation (20200165), the M. Reitman, A.U. Coskun, M.I. Papafaklis, C.L. Feldman, Prediction of progression
Swedish Research Council (2021-00456), the Swedish Society of Medi­ of coronary artery disease and clinical outcomes using vascular profiling of
cine (SLS-961835), Erik and Edith Fernströms Foundation (FS- endothelial shear stress and arterial plaque characteristics: the PREDICTION study,
Circulation 126 (2012) 172–181.
2021:0004), and Karolinska Institutet (FS-2020:0007), the Sweden- [11] P.H. Stone, A. Maehara, A.U. Coskun, C.C. Maynard, M. Zaromytidou, G. Siasos,
America Foundation, the Swedish Heart Foundation. We are grateful for I. Andreou, D. Fotiadis, K. Stefanou, M. Papafaklis, L. Michalis, A.J. Lansky, G.
the generous support of the Schaubert Family. S. Mintz, P.W. Serruys, C.L. Feldman, G.W. Stone, Role of low endothelial shear
stress and plaque characteristics in the prediction of nonculprit major adverse
cardiac events, J Am Coll Cardiol Img 11 (2018) 462–471.
CRediT authorship contribution statement [12] V. Thondapu, C. Mamon, E.K.W. Poon, O. Kurihara, H.O. Kim, M. Russo, M. Araki,
H. Shinohara, E. Yamamoto, J. Dijkstra, M. Tacey, H. Lee, A. Ooi, P. Barlis, I.-
K. Jang, On behalf of the M.G.H.O.C.T.R. Investigators, High spatial endothelial
Sophie Z. Gu: Formal analysis, Software, development, Writing – shear stress gradient independently predicts site of acute coronary plaque rupture
original draft, Writing – review & editing. Mona E. Ahmed: Formal and erosion, Cardiovasc. Res. 117 (2021) 1974–1985.
analysis, Writing – original draft, Writing – review & editing. Yuan [13] S.Z. Gu, M.R. Bennett, Plaque structural stress: detection, determinants and role in
atherosclerotic plaque rupture and progression, Front. Cardiovasc. Med. 9 (2022)
Huang: Supervision, Software, development, Writing – review & edit­ 875413.
ing. Diaa Hakim: Supervision, Writing – review & editing. Charles [14] Z. Teng, A.J. Brown, P.A. Calvert, R.A. Parker, D.R. Obaid, Y. Huang, S.P. Hoole, N.
Maynard: Supervision, Writing – review & editing. Nicholas V. Cefalo: E.J. West, J.H. Gillard, M.R. Bennett, Coronary plaque structural stress is
associated with plaque composition and subtype and higher in acute coronary
Writing – review & editing. Ahmet U. Coskun: Software, development, syndrome, Circ. Cardiovasc. Imaging 7 (2014) 461–470.
Writing – review & editing. Charis Costopoulos: Supervision, Writing – [15] A.J. Brown, Z. Teng, P.A. Calvert, N.K. Rajani, O. Hennessy, N. Nerlekar, D.
review & editing. Akiko Maehara: Data collection, Writing – review & R. Obaid, C. Costopoulos, Y. Huang, S.P. Hoole, M. Goddard, N.E.J. West, J.
H. Gillard, M.R. Bennett, Plaque structural stress estimations improve prediction of
editing. Gregg W. Stone: Data collection, Writing – review & editing.
future major adverse cardiovascular events after intracoronary imaging, Circ.
Peter H. Stone: Supervision, Conceptualization, Writing – review & Cardiovasc. Imaging. 9 (2016) 1–9.
editing. Martin R. Bennett: Supervision, Conceptualization, Writing – [16] C. Costopoulos, Y. Huang, A.J. Brown, P.A. Calvert, S.P. Hoole, N.E.J. West, J.
review & editing. H. Gillard, Z. Teng, M.R. Bennett, Plaque rupture in coronary atherosclerosis is
associated with increased plaque structural stress, J Am Coll Cardiol Img 10 (2017)
1472–1483.
Appendix A. Supplementary data [17] C. Costopoulos, A. Maehara, Y. Huang, A.J. Brown, J.H. Gillard, Z. Teng, G.
W. Stone, M.R. Bennett, Heterogeneity of plaque structural stress is increased in
plaques leading to MACE: insights from the PROSPECT study, J Am Coll Cardiol
Supplementary data to this article can be found online at https://doi. Img 13 (2020) 1206–1218.
org/10.1016/j.atherosclerosis.2024.117449.

9
S.Z. Gu et al. Atherosclerosis 390 (2024) 117449

[18] Y. Li, J.L. Gutiérrez-Chico, N.R. Holm, W. Yang, L. Hebsgaard, E.H. Christiansen, [24] C. V Bourantas, T. Zanchin, R. Torii, P.W. Serruys, A. Karagiannis, A. Ramasamy,
M. Mæng, J.F. Lassen, F. Yan, J.H.C. Reiber, S. Tu, Impact of side branch modeling H. Safi, A.U. Coskun, G. Koning, Y. Onuma, C. Zanchin, R. Krams, A. Mathur,
on computation of endothelial shear stress in coronary artery disease, J. Am. Coll. A. Baumbach, G. Mintz, S. Windecker, A. Lansky, A. Maehara, P.H. Stone, L. Raber,
Cardiol. 66 (2015) 125–135, https://doi.org/10.1016/j.jacc.2015.05.008. G.W. Stone, Shear stress estimated by quantitative coronary angiography predicts
[19] A.A. Giannopoulos, Y.S. Chatzizisis, P. Maurovich-Horvat, A.P. Antoniadis, plaques prone to progress and cause events, J Am Coll Cardiol Img 13 (2020)
U. Hoffmann, M.L. Steigner, F.J. Rybicki, D. Mitsouras, Quantifying the effect of 2206–2219.
side branches in endothelial shear stress estimates, Atherosclerosis 251 (2016) [25] V. Tufaro, H. Safi, R. Torii, B.-K. Koo, P. Kitslaar, A. Ramasamy, A. Mathur, D.
213–218, https://doi.org/10.1016/j.atherosclerosis.2016.06.038. A. Jones, R. Bajaj, E. Erdoğan, A. Lansky, J. Zhang, K. Konstantinou, C.D. Little,
[20] S.P. H, C.A. Umit, P. Francesco, Ongoing methodological approaches to improve R. Rakhit, G. V Karamasis, A. Baumbach, C. V Bourantas, Wall shear stress
the in vivo assessment of local coronary blood flow and endothelial shear stress, estimated by 3D-QCA can predict cardiovascular events in lesions with borderline
J. Am. Coll. Cardiol. 66 (2015) 136–138, https://doi.org/10.1016/j. negative fractional flow reserve, Atherosclerosis 322 (2021) 24–30.
jacc.2015.05.010. [26] J.M. Lee, G. Choi, B.-K. Koo, D. Hwang, J. Park, J. Zhang, K.-J. Kim, Y. Tong, H.
[21] A.U. Coskun, Y. Yeghiazarians, S. Kinlay, M.E. Clark, O.J. Ilegbusi, A. Wahle, J. Kim, L. Grady, J.-H. Doh, C.-W. Nam, E.-S. Shin, Y.-S. Cho, S.-Y. Choi, E.J. Chun,
M. Sonka, J.J. Popma, R.E. Kuntz, C.L. Feldman, P.H. Stone, Reproducibility of J.-H. Choi, B.L. Nørgaard, E.H. Christiansen, K. Niemen, H. Otake, M. Penicka,
coronary lumen, plaque, and vessel wall reconstruction and of endothelial shear B. de Bruyne, T. Kubo, T. Akasaka, J. Narula, P.S. Douglas, C.A. Taylor, H.-S. Kim,
stress measurements in vivo in humans, Catheter, Cardiovasc. Interv. 60 (2003) Identification of high-risk plaques destined to cause acute coronary syndrome
67–78, https://doi.org/10.1002/ccd.10594. using coronary computed tomographic angiography and computational fluid
[22] S.Z. Gu, C. Costopoulos, Y. Huang, C. Bourantas, A. Woolf, C. Sun, Z. Teng, dynamics, J Am Coll Cardiol Img 12 (2019) 1032–1043.
S. Losdat, L. Räber, H. Samady, M.R. Bennett, High-intensity statin treatment is [27] L.H. Timmins, D.S. Molony, P. Eshtehardi, M.C. McDaniel, J.N. Oshinski, D.
associated with reduced plaque structural stress and remodelling of artery P. Giddens, H. Samady, Oscillatory wall shear stress is a dominant flow
geometry and plaque architecture, Eur. Hear. J. Open. 1 (2021) oeab039. characteristic affecting lesion progression patterns and plaque vulnerability in
[23] D. Erlinge, A. Maehara, O. Ben-Yehuda, H.E. Bøtker, M. Maeng, L. Kjøller-Hansen, patients with coronary artery disease, J. R. Soc. Interface. 14 (2017).
T. Engstrøm, M. Matsumura, A. Crowley, O. Dressler, G.S. Mintz, O. Fröbert, [28] A. Hoogendoorn, A.M. Kok, E.M.J. Hartman, G. de Nisco, L. Casadonte, C. Chiastra,
J. Persson, R. Wiseth, A.I. Larsen, L. Okkels Jensen, J.E. Nordrehaug, Ø. Bleie, A. Coenen, S.-A. Korteland, K. Van der Heiden, F.J.H. Gijsen, D.J. Duncker, A.F.
E. Omerovic, C. Held, S.K. James, Z.A. Ali, J.E. Muller, G.W. Stone, O. Ahlehoff, W. van der Steen, J.J. Wentzel, Multidirectional wall shear stress promotes
A. Amin, O. Angerås, P. Appikonda, S. Balachandran, S. Barvik, K. Bendix, advanced coronary plaque development: comparing five shear stress metrics,
M. Bertilsson, U. Boden, N. Bogale, V. Bonarjee, F. Calais, J. Carlsson, Cardiovasc. Res. 116 (2020) 1136–1146.
S. Carstensen, C. Christersson, E.H. Christiansen, M. Corral, O. De Backer, [29] P.H. Stone, P. Libby, W.E. Boden, Fundamental pathobiology of coronary
U. Dhaha, C. Dworeck, K. Eggers, C. Elfström, J. Ellert, E. Eriksen, C. Fallesen, atherosclerosis and clinical implications for chronic ischemic Heart disease
M. Forsman, H. Fransson, M. Gaballa, M. Gacki, M. Götberg, L. Hagström, management—the plaque hypothesis: a narrative review, JAMA Cardiol 8 (2023)
T. Hallberg, K. Hambraeus, I. Haraldsson, J. Harnek, O. Havndrup, K. Hegbom, 192–201.
M. Heigert, S. Helqvist, J. Herstad, Z. Hijazi, L. Holmvang, D. Ioanes, A. Iqbal, [30] A.P. Antoniadis, M.I. Papafaklis, S. Takahashi, K. Shishido, I. Andreou, Y.
A. Iversen, J. Jacobson, L. Jakobsen, I. Jankovic, U. Jensen, K. Jensevik, S. Chatzizisis, M. Tsuda, S. Mizuno, Y. Makita, T. Domei, T. Ikemoto, A.U. Coskun,
N. Johnston, T.F. Jonasson, E. Jørgensen, F. Joshi, U. Kajermo, F. Kåver, H. Kelbæk, J. Honye, S. Nakamura, S. Saito, E.R. Edelman, C.L. Feldman, P.H. Stone, Arterial
T. Kellerth, M. Kish, W. Koenig, S. Koul, B. Lagerqvist, B. Larsson, J.F. Lassen, remodeling and endothelial shear stress exhibit significant longitudinal
O. Leiren, Z. Li, C. Lidell, R. Linder, M. Lindstaedt, G. Lindström, S. Liu, K. heterogeneity along the length of coronary plaques, J Am Coll Cardiol Img 9
H. Løland, J. Lønborg, L. Márton, H. Mir-Akbari, S. Mohamed, J. Odenstedt, (2016) 1007–1009.
C. Ogne, J. Oldgren, G. Olivecrona, N. Östlund-Papadogeorgos, M. Ottesen, [31] S.Z. Gu, Y. Huang, C. Costopoulos, B. Jessney, C. Bourantas, Z. Teng, S. Losdat,
E. Packer, Å.M. Palmquist, Q. Paracha, F. Pedersen, P. Petursson, T. Råmunddal, A. Maehara, L. Räber, G.W. Stone, M.R. Bennett, Heterogeneous plaque–lumen
S. Rotevatn, R. Sanchez, G. Sarno, K.I. Saunamäki, F. Scherstén, P.W. Serruys, geometry is associated with major adverse cardiovascular events, Eur. Hear. J.
I. Sjögren, R. Sørensen, I. Srdanovic, Z. Subhani, E. Svensson, A. Thuesen, Open. 3 (2023) oead038.
J. Tijssen, H.-H. Tilsted, T. Tödt, T. Trovik, B.I. Våga, C. Varenhorst, K. Veien, [32] Y. Shi, J. Zheng, N. Yang, Y. Chen, J. Sun, Y. Zhang, X. Zhou, Y. Gao, S. Li, H. Zhu,
E. Vestman, S. Völz, L. Wallentin, J. Wykrzykowska, L. Zagozdzon, M. Zamfir, J. Acosta-Cabronero, P. Xia, Z. Teng, The effect of subbranch for the quantification
C. Zedigh, H. Zhong, Z. Zhou, Identification of vulnerable plaques and patients by of local hemodynamic environment in the coronary artery: a computed
intracoronary near-infrared spectroscopy and ultrasound (PROSPECT II): a tomography angiography–based computational fluid dynamic analysis, Emerg.
prospective natural history study, Lancet 397 (2021) 985–995. Crit. Care Med. 2 (2022). https://journals.lww.com/eccm/fulltext/2022/12000/th
e_effect_of_subbranch_for_the_quantification_of.2.aspx.

10