Statistics from Altmetric.com
Acute coronary syndromes (ACS) are a major cause of morbidity and mortality in the developed world. The risk of future acute coronary events appears to be largely dependent on the presence of morphologically distinct, atherosclerotic plaques,1 rather than the presence of severely stenotic lesions as assessed by angiography.2 Angiography, however, cannot provide details of the vascular wall and identify these so-called “vulnerable plaques”. To identify, study and potentially treat these lesions many invasive and non-invasive imaging modalities are undergoing evaluation. While techniques such as intravascular ultrasound (IVUS), especially combined with radiofrequency analysis, have been significant developments providing valuable information in this quest, all have the fundamental drawback of inadequate spatial resolution to image important lesion characteristics. With the development of optical coherence tomography (OCT) and the demonstration of its feasibility for intravascular imaging, a technique with micron-scale resolution became a reality, potentially enabling the ability to capture in vivo what was previously seen only through a pathologist’s microscope. Further, with the widespread adaptation of percutaneous coronary intervention (PCI) for the treatment of coronary disease, OCT is also emerging as a promising technology both to image the acute results of PCI and to monitor the response of the vessel wall to stent deployment. With ongoing development of this modality, OCT has the potential to be an important clinical imaging modality, complementary to angiography and IVUS.
OCT TECHNOLOGY AND TECHNIQUE
OCT is the optical analogue of pulse-echo IVUS, where electromagnetic waves using a light source as opposed to acoustic (sound) waves are used to create the image.3 A light source is emitted towards the sample. Information from the echo time delay (time for the light to be reflected back) and the intensity of backscatter of light from internal microstructures with varying optical properties within the sample is used to create the image. The speed of light is much faster than sound and the echo time delay is extremely fast and cannot be measured directly by electronic detection. To overcome this, OCT uses a technique called low-coherence interferometry, where light reflected or backscattered from inside the sample specimen is measured by correlating it with light that has travelled a known reference path length. All OCT platforms are built on this basic principle. So far, the majority of intravascular OCT imaging in patients has been performed using time-domain OCT (TD-OCT) systems.
A typical TD-OCT interferometer system is shown schematically in fig 1. A low-coherence light source (near infrared, broad bandwidth light with high irradiance) is divided by a fibre-optic coupler (comparable to an optical beam splitter) into two separate arms, one directed down the sample or measurement arm and via the imaging catheter at the sample, and the other transmitted down the reference arm to a moving retro-reflecting mirror. The movements of this mirror are calibrated such that, for any given position of the mirror, the path length that the reflected light travels is known. When light reflected from the sample and the reference arm (having travelled a known reference length) is combined, interference occurs only when the light in both arms arrives at the same time, having travelled the same optical distance. The intensity of interference is detected by a photodetector, and is translated after processing into an intensity map and encoded using grey or false colour scale. Axial scans are acquired while the beam is scanned across the surface to acquire a complete two-dimensional dataset.
For TD-OCT, light in the near infrared range with a wavelength centred around 1300 nm is used because light with a shorter wavelength (such as visible light) is prone to a higher degree of scattering and absorption, while light with a longer wavelength results in unacceptable absorption attenuation. At ∼1300 nm a spatial resolution of 4 to 16 μm is achieved with a penetration depth of 2 to 3 mm. With image acquisition times of approximately 200 ms, high-resolution imaging can be obtained without significant motion artifacts.
Intravascular time-domain OCT systems
Two TD-OCT systems have been used for intravascular imaging in the clinical setting: the Massachusetts General Hospital (MGH) OCT system4 and the LightLab OCT system (LightLab Imaging, Westford, MA). A brief description of the systems accompanies the supplemental figures (supplementary figs 1 and 2).
Red blood cells scatter light, and for intravascular imaging with OCT a bloodless imaging field is required. With the MGH system this was achieved by 8–10 ml of saline flushed through the guiding catheter, allowing an image acquisition for approximately 2–3 seconds. This allowed sampling of only discrete transverse segments, precluding continuous imaging of longer sections of arterial wall. The LightLab system uses low pressure (∼0.5 atm) proximal balloon occlusion combined with distal flush from the tip of the catheter which is maintained while pull-back of the imaging fibre is undertaken. With balloon occlusion for 35 seconds and a pull-back rate of 1 mm/second, continuous imaging of a 35 mm coronary segment is possible.
HISTORY AND DEVELOPMENT OF OCT
The current OCT systems have their genesis in optical coherence-domain reflectometry (OCDR). OCDR, a one-dimensional optical ranging technique that uses low-coherence interferometry, was applied to optical electronic components and fibre-optics to perform micron-resolution optical measurements with uses such as finding defects in fibre-optic cables.5 6 Its use in biological tissues was soon recognised for measurement of eye length and corneal thickness.7 8 OCT was subsequently developed by a team from Massachusetts Institute of Technology and described as a modality for non-invasive detailed cross-sectional imaging in biological systems.3 The first OCT image of a human coronary artery ex vivo was provided as an example.
Early ex vivo studies on explanted human aorta and coronary arteries established the feasibility of OCT to image atherosclerotic plaque morphology at a micron scale with accurate delineation of morphology when compared with histology9–11 and also clearly documented its superior resolution and contrast when compared with IVUS imaging.12 13 Subsequently, the feasibility of in vivo intravascular imaging with OCT was assessed in the rabbit aorta.14 This work showed that high-resolution intravascular imaging could be performed in vivo using OCT. Following on from these early feasibility studies, an active research programme to validate this technology in evaluating vulnerable plaque was pursued by our group at MGH.15 The first systematic characterisation of human atherosclerotic plaque with OCT was performed at MGH, establishing OCT criteria for the primary components of these plaques,16 and the ability of OCT to detect and quantify the amount of macrophages in atherosclerotic plaque was demonstrated.17 Subsequently the in-human application of OCT was successfully undertaken.18 19 The more widespread accessibility of OCT through the LightLab OCT system has now resulted in a substantial increase in the use of this technology. Associated with this has been an increase in research both validating previous studies and adding important new information on the use of OCT in the assessment of plaque and its role in the context of PCI.
CURRENT APPLICATION OF OCT IN CORONARY DISEASE
Atherosclerotic plaque characterisation
Disruption of an atherosclerotic plaque with subsequent thrombus formation is the underlying pathophysiological process in most ACSs.1 Postmortem studies suggest that disruption occurs by fibrous cap rupture (up to 70%), plaque surface erosion (up to 25%) or disruption due to superficial calcium (5%).1 Morphological features that predispose plaques to rupture include (1) a large lipid-rich core, (2) a thin fibrous cap usually ⩽65 μm depleted of collagen, (3) accumulation of macrophages localised in the fibrous cap and (4) neovascularisation. Since OCT has a spatial resolution sufficient to image many of these morphological features, much work has been done to try and realise this potential with the ultimate hope of being able to identify and study high-risk vulnerable plaque in vivo.
Ex vivo characterisation of atherosclerotic plaque
The first step towards the utilisation of OCT as a tool for plaque assessment was the characterisation of human atherosclerotic lesions using OCT to identify markers of plaque vulnerability.
Using 357 atherosclerotic segments from 90 cadavers,16 OCT criteria for plaque components were established using 50 segments as a training set. Three types of plaque were identified: fibrous, fibro-calcific, and lipid-rich (fig 2). The 307 remaining segments were then blindly interpreted. High sensitivity and specificity were obtained for detection of lipid and fibro-calcific plaques with OCT as compared with histology (90% and 92%; 96% and 97%, respectively). Sensitivity and specificity for fibrous plaque were 79% and 97%, respectively. Interobserver and intraobserver agreements for characterisation of plaque type with OCT were high (κ = 0.88 and 0.91, respectively). Similar results were subsequently published by other studies using the same plaque criteria.20 21 Two small studies demonstrated lower sensitivities for detecting plaque type (table 1). Methodological differences and a small sample size may partly account for this.22 23
Fibrous cap thickness
At autopsy, 95% of ruptured plaques have a fibrous cap thickness near the rupture site of <65 µm.24 OCT is the only modality with the resolution required for direct quantitative assessment of cap thickness. We compared the cap thickness measured by OCT and histology in 29 lipid-rich plaques. There was close correlation between the two methods (r = 0.89, p<0.001; unpublished data). Kume et al duplicated these findings in 35 ex vivo lipid-rich human coronary plaques demonstrating a similarly high correlation (r = 0.90, p<0.001),25 confirming the ability of OCT to precisely measure cap thickness, a critical feature associated with plaque rupture.
Fibrous cap macrophage content
Identifying macrophages in vivo would provide invaluable information in assessing the inflammatory state of a plaque and its vulnerability to rupture.
With the relatively large size of macrophages (20–50 µm) and their high degree of optical contrast, it was hypothesised that plaques rich in macrophages should exhibit a high OCT signal variance. It was assumed that by determining the normalised standard deviation (NSD) of the OCT signal within the fibrous cap it would be possible to quantify the macrophage content within that region.17 OCT imaging for macrophage density was performed using raw OCT NSD signal data in 26 lipid-rich plaques and compared with macrophage density quantified histomorphometrically by immunoperoxidase staining with CD68 (fig 3). A strong correlation (r = 0.84, p<0.001) was found between the raw OCT NSD signal intensity and macrophage content determined by histology. This correlation was independent of cap thickness (controlling for cap thickness r = 0.80, p<0.001). These results demonstrated the capability of OCT in evaluating macrophage content within fibrous caps of atheroma with a high degree of accuracy. It should be noted, however, that, given the resolution of OCT, the processing steps used and the range of sizes of macrophages, individual macrophages are not resolved. The computed macrophage density (NSD) originates from the reflectivity differences between the macrophages and surrounding cap matrix.
The OCT characteristics of coronary thrombi confirmed at postmortem examination were examined in 108 coronary arterial segments.26 White thrombi were identified as signal-rich, low-backscattering protrusions in the OCT image while red thrombi were identified as high-backscattering protrusions inside the lumen of the artery, with signal-free shadowing in the OCT image. Using a measure of the attenuation of intensity of the OCT signal within the thrombus, the authors demonstrated differentiation of red and white thrombi with high sensitivity and specificity.26 In a rabbit model of atherosclerosis and acute thrombosis, OCT identified all histologically confirmed thrombi.27 In addition, the appearance of the red thrombi was consistent with that in the autopsy study.
Clinical application of OCT in plaque characterisation
Following the successful first-in-man application of OCT in 10 patients19 using the ex vivo validated OCT markers of plaque vulnerability, a number of clinical studies have systematically evaluated the associations between plaque and patient characteristics.
Plaque characterisation in patients with acute and stable coronary syndromes
OCT imaging of culprit plaque was performed in 57 patients (20 patients following ST elevation myocardial infarction (STEMI), 20 patients with non-ST elevation ACS (NSTEACS) and 17 with stable angina pectoris (SAP)) prior to PCI.28 Lipid-rich plaque (defined as lipid occupying ⩾2 quadrants of the cross-sectional area) was observed in 90% of STEMI, 75% of NSTEACS, and 59% of SAP patients (p = 0.09). The median fibrous cap thickness and the frequency of TCFA (lipid-rich plaque with a fibrous cap thickness ⩽65 µm) differed significantly between the presentation groups (table 2). An example of a patient with STEMI and a patient with SAP is shown in fig 4. The significant association between fibrous cap attenuation, TCFA, and patients who presented with unstable coronary syndromes provided, for the first time, in vivo corroboration of the findings of the autopsy studies. It reinforced the importance of fibrous cap thickness as a marker of plaque vulnerability, demonstrating the benefit of the high spatial resolution of OCT. Fibrous cap rupture detected in this study was lower than expected and was probably a result of the discrete sampling method necessitated by the short image acquisition times and low frame rates of the first-generation MGH OCT system. Imaging in 30 patients with STEMI using the LightLab OCT system29 demonstrated fibrous cap rupture in 73% of culprit plaque. The authors also describe the OCT characteristics of what they propose is plaque erosion, which occurred in 23% of culprit plaque. While OCT characteristics of erosion have never been validated against histology, the appearance of the lesions on OCT (fig 5) and their frequency appear to support this assumption. The frequencies of lipid-rich plaque, TCFA and the mean fibrous cap thickness in this study (93%, 83% and 47 µm respectively) were remarkably similar to those seen in the STEMI group from the previous study.
Macrophage concentration and distribution
Macrophage densities in lipid-rich plaques varied significantly among the clinical groups (p<0.001 for trend in culprit and remote lesions) with a higher macrophage concentration in the ACS patients than in the patients with SAP (fig 6). Macrophage densities at both culprit and remote sites within each clinical group were similar, reflecting the concept of pan-arterial inflammation in patients presenting with an acute event. In patients with ACS with clear evidence of plaque rupture, a significantly higher macrophage density was noted at the rupture site than at the adjacent non-ruptured cap within the same image (fig 7), as demonstrated in histopathological studies. Macrophage density was also higher at the superficial endothelial surface of culprit plaque compared with subsurface macrophage density and was a better predictor of ACS; a parameter for assessing individual plaque vulnerability that warrants further investigation.
Systemic inflammation, local plaque inflammation, and plaque characteristics
To evaluate the concept that inflammation, both locally within each plaque and systemically, plays a critical role in the development of TCFA, using OCT imaging, the relationships between the peripheral white cell (WBC) count, local plaque fibrous cap macrophage density and the characteristics and presence of TCFA were assessed in subjects with ACS and SAP.31 Baseline WBC count was found to correlate with macrophage density (r = 0.483, p<0.001). Both parameters were associated with lipid-rich plaque and correlated inversely with plaque fibrous cap thickness (r = −0.547 for macrophage density and −0.412 for WBC count, p<0.015). Plaques classified as TCFA had a higher median macrophage density than non-TCFA plaques and patients with TCFA had a higher WBC count than those without TCFA (fig 8). Receiver operator curves were computed for WBC count, macrophage density and these combined parameters for prediction of TCFA and showed the area under the curves to be 0.88, 0.91 and 0.97 (p<0.001), respectively. This study demonstrates that, in the plaque of subjects with acute and stable presentations, fibrous cap macrophage density (a marker of a plaque’s local inflammatory state) correlates with a patient’s WBC count (a marker of systemic inflammation) and both parameters, independently and particularly in combination, predicted the presence of TCFA.
Remodelling and plaque characteristics
Positive remodelling is associated with unstable coronary syndromes and ex vivo histological characteristics of plaque vulnerability. The association between remodelling assessed by IVUS and underlying plaque characteristics identified by OCT was performed in 54 lesions.32 Positive remodelling compared with negative remodelling was more commonly associated with lipid-rich plaque, a thin fibrous cap, the presence of TCFA and a larger inflammatory cell infiltrate. This study supported the associations demonstrated in the autopsy studies and provided further in vivo evidence to explain the link between positive remodelling and unstable coronary syndromes.
Comparison with other imaging modalities
Several invasive and non-invasive imaging modalities are undergoing evaluation for the assessment of coronary plaque33 34 (table 3). Non-invasive techniques do not at present have the spatial or temporal resolution to image coronary plaque adequately. Due to its ease of use, ability to image through blood, good axial resolution and established role in interventional cardiology, IVUS is the most widely employed invasive imaging modality. Its major advantage over OCT is, due to its larger imaging depth, the ability to image the entire vessel allowing measurement of parameters such as vessel dimensions, remodelling, plaque area and volume and, combined with radiofrequency (RF) analysis, a quantitative assessment of the composition of plaque.35 However, due to a spatial resolution of 100–150 µm, IVUS is unable to measure fibrous cap thickness, identify true TCFA or image inflammatory cells.29 36 In addition, for the detection of lipid plaque, OCT is superior to grey-scale IVUS (table 1).19 20 22 29 37 The addition of RF analysis to IVUS improves the detection of lipid plaque ex vivo, with comparable specificity and slightly reduced sensitivity compared with OCT (table 1).37
Other invasive modalities such as angioscopy, thermography, spectroscopy and intravascular MRI have also been advocated (table 3). Of these, only angioscopy and thermography have had significant in vivo evaluation. Angioscopy is not in widespread use and, like the current clinical TD-OCT system, requires a blood-free field, achieved using a proximal occluding balloon. It is a qualitative imaging tool and direct quantitative assessment of plaque morphology is not possible.38 In the context of culprit plaque assessment in patients with ACS, OCT appears to be clearly superior.29 Temperature heterogeneity using thermography has been described as an important correlate of inflammation within the atherosclerotic plaque and in this context is a surrogate for its macrophage content.39 No direct comparisons have been made with OCT. Thermography may be a useful complementary modality for OCT systems not validated for macrophage assessment and such studies are underway.
Future role of OCT in plaque characterisation
Plaque characterisation by OCT is at present the domain of research with its clinical application yet to be determined. Our current understanding of the pathophysiology of ACS is primarily based on histomorphological and in vitro biological studies. Substantiating this paradigm in the clinical setting is crucial. While the quest to identify vulnerable plaque in vivo has been the primary focus of current research, it is vital that the question of the clinical relevance of being able to identify a vulnerable lesion is not overlooked. To date all published studies using OCT have been single time point studies. Well-designed, prospective follow-up studies using appropriate imaging modalities are needed to answer these questions. Due to its unique abilities, OCT will need to be incorporated into any such study, most likely complemented by RF-analysed IVUS. In addition, OCT may be useful to monitor plaque response to genetic, pharmacological, or other forms of intervention. For more widespread clinical application to occur, however, further work on OCT technology is needed. The current TD-OCT systems impose too many practical limitations to be applicable in comprehensive plaque assessment of patients in the clinical setting. Second-generation systems promise to overcome this hurdle. There is a definite learning curve associated with OCT image interpretation, as reflected in the results of some of the ex vivo studies. Improved image display in the form of colour-coded tissue maps similar to RF IVUS systems would be advantageous. Work is underway at present to try to achieve this using OCT signal backscatter and attenuation data. Assessment of macrophage concentration was only validated with the first-generation MGH TD-OCT system. To make this a robust OCT marker, further work is needed to validate and incorporate this parameter in a clinically useful manner with other OCT systems.
OCT in percutaneous coronary intervention
IVUS is the standard modality to assess the outcome of PCI. While IVUS is able to accurately define post-PCI lesion cross-sectional dimensions and assess stent symmetry and stent expansion, detailed information related to the stented site is restricted because the resolution of IVUS is limited and metal stent struts impair image quality. Therefore the true incidence and clinical relevance of such features as edge dissections, tissue prolapse and residual thrombus are not known. Further, in the drug-eluting stent era, new entities such as late stent malapposition and the concern about inadequate neointimal strut coverage as a risk for late stent thrombosis (LST) have emerged. OCT could thus provide potential advantages over IVUS both in imaging the acute results of PCI40–42 (fig 5) and in monitoring the response of the vessel to stent deployment43 44 (fig 9). In a study of 43 imaged stents, OCT consistently detected more incidences of dissection, tissue prolapse, and incomplete stent deployment than IVUS.40 In a smaller study of 10 patients similar advantages were demonstrated.41 Future prospective studies will need to address the consequences of these findings on the outcome after PCI. The role of OCT in complex interventions such as bifurcation PCI and chronic total occlusions is also being explored.45
With its high resolution and ability to accurately differentiate stent struts from neointimal tissue, OCT may be particularly advantageous to assess the response to stent implantation. This is particularly relevant with DES (fig 9). Prospective studies with follow-up to 12 months have demonstrated, using OCT, the delayed neointimal growth associated with DES compared with BMS, and the absence of late catch-up neointimal proliferation, at least within the period of follow-up.43 44 46 The capability of OCT to accurately measure neointimal thickness compared with histology was demonstrated (correlation r = 0.85, p<0.001) in stented rabbit carotids.47 However, this was a small study (eight stents) using only BMS and did not look at the ability of OCT to differentiate fibrin from neointima or assess the presence of true endothelialisation, factors implicated in LST.48 Thus further work is needed to validate OCT with regards to these parameters in an appropriate model before its clinical application can be fulfilled.
OCT requires a blood-free zone for imaging, accomplished by crystalloid or contrast flush. With frame rates of 4–15 frames/second, this imposes practical limitations on the system. A single bolus flush permits an acquisition time of 2–3 seconds, and imaging is only possible at discrete locations. With proximal balloon occlusion and continuous flushing, longer segments can be imaged. The disadvantage is that the configuration is relatively cumbersome compared with current IVUS systems; there is transient ischaemia in the territory of the artery under study and some concern about the local consequences of balloon inflation. In addition, inadequate displacement of blood can be a problem in vessels >3.5 mm in diameter, where large bifurcations are present and in the presence of competitive flow from collaterals or bypass grafts. Index matching constitutes an alternative approach based on the premise that the predominant source of scattering in blood is the difference in refractive index between the cytoplasm of erythrocytes and serum. Increasing the refractive index of serum to a value similar to that of cytoplasm could substantially reduce scattering. Another approach is isovolumic replacement of blood with an optically transparent haemoglobin-based blood substitute.
Another limitation of OCT is the limited depth range to a vessel within a radius of approximately 3–4 mm and the penetration depth in tissue of 1–3 mm, which limits detailed imaging beyond the internal elastic lamina. This precludes visualisation and measurement of the entire plaque and vessel wall. Parameters such as plaque and lipid core area and remodelling cannot be measured and remain the domain of IVUS.
While macrophage concentration is a surrogate for the biological activity of the plaque, OCT at present cannot provide physiological and functional information.
FUTURE DEVELOPMENTS IN OCT TECHNOLOGY
Improving speed, resolution and ease of use
Second-generation OCT systems have now been developed that potentially overcome many of the practical limitations of the TD-OCT scanners. This technology utilises a detection method termed Fourier or frequency-domain detection, which allows high-speed data acquisition and real-time imaging (supplementary fig 3). Systems have been developed by the Wellman Laboratories at MGH49 and LightLab Imaging, and application of this system in humans has now begun. Using guide catheter flush without balloon occlusion, imaging at ⩾100 frames per second (with 500+ A-lines/frame) with a pull-back rate of up to 20 mm/second has been achieved, contrasting with 15.6 frames/second (at 200 A-lines/frame) and pull-back rate of 1 mm/second for the TD-OCT system. This translates into an impressive ability to image an entire coronary artery in a matter of a few seconds (supplementary movie). The higher A-line density of the system results in improved image quality and provides the ability to perform three-dimensional reconstruction of long segments of vessel using the volumetric data set (supplementary fig 4). The future of clinically applicable OCT imaging, competitive with the ease of use of IVUS platforms, rests directly with these new scanners.
Novel OCT techniques
Much work is in progress to find novel OCT techniques to provide additional markers of plaque vulnerability and extend the capabilities of OCT beyond imaging of structural morphology.
Polarisation-sensitive OCT (PS-OCT), in addition to measuring the intensity of back-reflected light, analyses its polarisation state, providing information on tissue birefringence. Collagen and smooth muscle cells exhibit elevated birefringence. Fibrous caps of unstable plaque are associated with low collagen content, thinner collagen fibres, and reduced numbers of smooth muscle cells. PS-OCT birefringence in ex vivo human atherosclerotic specimens has demonstrated a positive correlation with collagen fibre content50 51 and SMC density.52
The structural and biological characteristics of the fibrous cap dictate its biomechanical properties, which are what ultimately, under the influence of external stress, result in disruption. Elastography is a means of estimating the elastic biomechanical properties of tissue. IVUS elastography has already been applied to coronary arteries in vivo to provide “strain maps” as means of evaluating plaque tensile strength. OCT with superior spatial resolution could potentially allow characterisation of biomechanical properties at precise locations. The feasibility of OCT elastography has been demonstrated in phantom models and ex vivo atherosclerotic samples.52
Doppler OCT has been able to identify sub-mm/second blood flow in capillaries within tissue. Potential applications include the accurate study of blood flow adjacent to plaque surfaces and in the identification and study of neovascularisation related to vulnerable lesions.
Cellular and molecular imaging with OCT
Cellular and molecular imaging provides the opportunity to image beyond structural anatomy and is the focus of intense research.53 While OCT is insensitive to inelastically scattered light such as that emitted by fluorescent probes, the use of microsphere contrast agents and molecular contrast OCT is being explored for this purpose.53
OCT is a rapidly evolving optical imaging technique with tremendous potential, having already contributed to our knowledge of the atherosclerotic process in living patients. With its unique high-resolution capability for intravascular imaging, it has provided, for the first time, in vivo corroboration of important morphological features of plaque and their relation to clinical coronary syndromes. As a complementary imaging modality to other established techniques such as IVUS, it is an important tool in vulnerable plaque research. Whether it will have an established clinical role in vulnerable plaque detection and treatment, however, must depend on the outcomes of future prospective natural history studies. OCT has also demonstrated its effectiveness in imaging the short-term and long-term results of PCI, and if the next generation of high-speed FD-OCT systems live up to their expectations the role of IVUS as the preferred imaging modality for interventions will be seriously challenged. The development of novel OCT techniques promises to add another dimension to the utility of this modality and provides for an exciting future.
( Additional figures and a video are published online only at http://heart.bmj.com/content/ vol94/issue9
Competing interests: None declared.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.