Positron emission tomography of the vulnerable atherosclerotic plaque in man--a contemporary review.

Atherosclerosis is the primary underlying cause of cardiovascular disease (CVD). It is the leading cause of morbidity and mortality in the Western world today and is set to become the prevailing disease and major cause of death worldwide by 2020. In the 1950s surgical intervention was introduced to treat symptomatic patients with high-grade carotid artery stenosis due to atherosclerosis--a procedure known as carotid endarterectomy (CEA). By removing the atherosclerotic plaque from the affected carotid artery of these patients, CEA is beneficial by preventing subsequent ipsilateral ischemic stroke. However, it is known that patients with low to intermediate artery stenosis may still experience ischemic events, leading clinicians to consider plaque composition as an important feature of atherosclerosis. Today molecular imaging can be used for characterization, visualization and quantification of cellular and subcellular physiological processes as they take place in vivo; using this technology we can obtain valuable information on atherosclerostic plaque composition. Applying molecular imaging clinically to atherosclerotic disease therefore has the potential to identify atherosclerotic plaques vulnerable to rupture. This could prove to be an important tool for the selection of patients for CEA surgery in a health system increasingly focused on individualized treatment. This review focuses on current advances and future developments of in vivo atherosclerosis PET imaging in man.

Introduction

Atherosclerosis is a degenerative inflammatory vascular disease (Ross, 1999). It is the primary underlying cause of CVD and by causing myocardial infarction and stroke; it has become the leading cause of morbidity and mortality in the Western world today (WHO, 2012; Go et al., 2013). The economic revolution in Southeast Asia and the accelerating urbanization of China and India, combined with the prevalence of traditional CVD risk factors in particular, is expected to increase CVD incidence dramatically by 2020 (Celermajer et al., 2012). Our knowledge of the cellular and molecular mechanisms causing and aggravating atherosclerosis has increased substantially in the last several years, however, we still face serious challenges trying to discriminate the relatively benign stable atherosclerotic plaque from the high-risk (vulnerable) plaque in a clinical setting (Fleg et al., 2012; Joshi et al., 2012). The use of ultrasound allow some qualitative evaluation of carotid plaques; however, selection criteria for surgical intervention for high-risk plaques of the internal carotid artery; carotid endarterectomy (CEA) remains largely based on the degree of carotid stenosis (Hobson et al., 2008). What is needed is a clinical tool for identification of the vulnerable plaque so that CEA is only performed on patients standing to benefit from the procedure, thereby reducing the numbers needed to treat which is currently about six to one (Chaturvedi et al., 2005). Molecular imaging has the potential as a tool for identification of the vulnerable plaque; allowing for risk stratification and individualized preventive intervention, as well as enabling clinicians to monitor the effect of medical therapy.

Molecular imaging is a technology enabling visualization of the molecular interaction and distribution of a probe (hereon referred to as a tracer) with its target or pathway in an intact biological system, i.e. in vivo. Existing molecular imaging modalities comprise contrast-enhanced ultrasound, optical imaging, magnetic resonance imaging (MRI), single photon emission computed tomography (SPECT) and positron emission tomography (PET). It is the scope of this review to present current advances and potential future applications of PET for in vivo atherosclerotic disease management in human patients.

Pathogenesis of atherosclerosis

Atherosclerosis is a systemic artery vessel wall disease characterized by inflammation (Ross, 1999; Hansson & Libby, 2006). Originally, plaque development was subdivided into four major stages, whereas the modern perception is that of a more seamless transition over time. Initiation of plaque development is a result of endothelial dysfunction with a decrease in nitric oxide (NO) bioavailability, oxidative excess and subsequent inflammation (Endemann & Schiffrin, 2004; Vita, 2011). A pro-inflammatory state results and expression of vascular cell adhesion molecule-1 (VCAM-1) and intercellular adhesion molecule-1 (ICAM-1) by endothelial cells (ECs) initiate white blood cell recruitment (Endemann & Schiffrin, 2004). Monocytes then infiltrate the artery intima where they differentiate into macrophages and start to engulf lipids such as oxidized low-density lipoprotein (oxLDL), primarily via scavenger receptors class A, B and D, resulting in macrophage transformation into foam cells and development of a fatty streak (Libby, 2002; Rahaman et al., 2006; Ashraf & Gupta, 2011). With increased complexity of the lesion, foam cells become overwhelmed and form a necrotic core. Chemokines such as C-C motif ligand-2 (CCL-2) and monocyte chemoattractant protein-1 (MCP-1) secreted from ECs and vascular smooth muscle cells (VSMCs) stimulate a continuous influx of white blood cells into the lesion (Hansson & Libby, 2006). Foam cell remnants and necrotic debris then make up a highly thrombogenic lipid rich core and VSMCs start migrating into the lesion to form the basis of a fibrous cap by modification of the extracellular matrix. A hypoxic microenvironment stimulate the formation of microvessels from the vasa vasorum forming a highly immature plaque vasculature facilitating further white blood cell recruitment and haemorrhagic episodes due to microvessel immaturity (Virmani et al., 2005). The fulminant plaque may then protrude into the arterial lumen increasing the risk of organ ischemic events distal to the plaque, see Fig.2. An acellular plaque with a large lipid core separated from the blood by a thin fibrous cap may rupture and lead to myocardial infarction or stroke and is therefore by definition a vulnerable plaque (Finn et al., 2010).

Figure 2

Atherosclerosis and molecular imaging: The vulnerable atherosclerotic plaque protrudes into the vessel lumen as a result of progressive inflammation of the vessel wall intima. Monocytes are continuously recruited from the blood and into the intima where they differentiate to macrophages and become foam cells due to lipid ingestion. Eventually, foam cells are overcome and become apoptotic amassing to a lipid rich necrotic core which is covered by a thin fibrous cap. Expansion of the intima leads to hypoxia which drives angiogenesis whereby new blood vessels sprout from the vasa vasorum in the vessel wall media. PET-tracers are depicted in blue and arrows point to their respective molecular targets. Integrin αVβ3, integrin receptor dimer alphaVbeta3; 18F-FDG, 2-[18F]-fluoro-2-deoxy-D-glucose; 18F-FMISO, 18F-fluoromisonidazole; 18F-NaF, 18F-sodium fluoride; 68Ga-DOTATATE, 68Ga-[1,4,7,10-tetraazacyclododecane-N,N′,N″,N‴-tetraaceticacid]-ᴅ-Phe1,Tyr3-octreotate; GLUT, glucose transporter; PET, positron emission tomography; SSTR2, somatostatin receptor subtype 2.

Positron emission tomography

The PET is a medical imaging technique based on the detection of radioactive decay from positron emitting radionuclides (β+ decay). The radionuclide coupled tracer is usually administered intravenously when clinically used. The tracer is designed to target a specific biochemical or pharmacological interaction in vivo, which can then be visualized in 3D following computer reconstruction, qua the associated radioactive decay which is recorded by the PET scanner (Cherry & Dahlbom, 2004). By drawing regions of interest (ROIs) on reconstructed imaging data the quantity of tracer uptake can be calculated for that region. The first in vivo detection of annihilation radiation (see below) as we know it today was performed in two dogs in 1975 using a prototype positron emission tomograph (Phelps et al., 1975; Ter-Pogossian et al., 1975). The positron is essentially a positive electron (e+) ejected from the isotope nucleus with a kinetic energy in the range of a few mega-electron volt (MeV). As the positron travels through tissue, the kinetic energy is deposited by multiple direction changing interactions with tissue electrons prior to uniting with an electron (e−) to undergo annihilation whereby the mass of the positron and an electron is converted into electromagnetic energy. The reaction produces two photons with the energy of 511 kilo-electron volt (keV) emitted back to back.

The PET scanner detects annihilation photons that escape the body. The detection system is configured so that opposite detectors is coupled to form a coincidence circuit in which the back to back emitted photons can be detected within a specified time window. The volume between the detector pairs where the coincidence detection takes place is called the line of response (LOR) and is made up by voxels each representing a 3D volume in space. Detection of thousands and thousands of these events at different angles and radial offsets followed by extensive computer analysis allows reconstruction of 3D images (Phelps et al., 1975; Cherry & Dahlbom, 2004).

A particular challenge to clinical PET imaging is the imaging of small structures as that of the vessel wall in atherosclerotic disease. The inherent ‘blurring’ of PET images is due to the finite spatial resolution of the system of which two fundamental limitations are of paramount importance; acollinearity and positron range (Cherry & Dahlbom, 2004; Moses, 2011). Of these factors only positron range can be affected and only by either selecting an alternate parent isotope or by using a powerful magnetic field to affect positron trajectory thereby decreasing positron range. Both options present great challenges in either chemical synthesis effort, or high equipment cost and are therefore considered impractical (Moses, 2011). However, the recent emergence of the first combined clinical PET/MRI scanners may prove an exception to at least part of this consensus (Schlemmer et al., 2008). Another important consideration when imaging small structures is the partial volume effect which describes how a small lesion with high tracer uptake in a low tracer background will result in the signal being spread out. This spill-out of signal from the lesion is also true for signal spilling back in from the background making the partial volume effect very hard to predict and compensate. In a small lesion this interprets to a lesser maximal value of lesion tracer uptake than the actual maximum value tracer uptake as well as an apparent larger lesion size (Cherry & Dahlbom, 2004; Soret et al., 2007). It is important to understand that partial volume effect does not cause signal loss, but that it does cause signal displacement (Soret et al., 2007).

Annihilation photons are likely to experience either attenuation or scatter events predominantly by Compton interactions in the tissue and the result is the removal of the annihilation photon from the original LOR. Recent advances have led to the implementation of analytical models correcting for single scatter events, by single scatter scaling (SSS), which is a reasonable approximation of total scatter when compared with modelling of events of second or higher order of scatter (Polycarpou et al., 2011). With the introduction of the first clinical hybrid PET/CT scanners (Beyer et al., 2000), attenuation correction can be performed by direct measurement using the CT modality to calculate a map of attenuation coefficients, of the patient, which is subsequently used to correct PET emission data (Burger et al., 2002).

Although the limited spatial resolution of PET is a major disadvantage when imaging small structures such as the vessel wall this can be complemented by co-registration with either CT or MRI, whereby detailed anatomical information can be combined with the PET imaging modality to locate precisely the anatomical distribution of tracer uptake and ease the drawing of ROIs (Beyer et al., 2000; Von Schulthess et al., 2006; Schlemmer et al., 2008). With acollinearity and positron range being fundamental limitations, the best intrinsic spatial resolution of full body clinical PET scanners achievable is just below 3 mm at the centre field of view (Moses, 2011); however, using algorithms to modify the point spread function (PSF) of the scanner a resolution as good as 2 mm is practically achievable (Panin et al., 2006; Levin et al., 2012).

PET tracers in atherosclerosis

When considering the bulk of PET tracer use there are basically two types of tracers clinically used; 2-[18F]-fluoro-2-deoxy-D-glucose (18F-FDG) and the more recently introduced non-18F-FDG tracers. At the Department of Nuclear Medicine, Clinical Physiology & PET, Rigshospitalet in Copenhagen approximately 88% of all PET imaging in the years 2004–2012 was performed using 18F-FDG (internal communication). Being a university hospital, Rigshospitalet has a high research activity and a comparable conventional PET facility would perform approximately 95–100% of all PET imaging using 18F-FDG. Therefore, we will discuss the research of atherosclerosis according to these two tracer types. Important milestones and novel research areas of each type are summarized in Tables1 and 2 respectively.

Table 1

Milestones; In vivo 18F-FDG-uptake in human atherosclerosis using PET

Target	Modality	Study type	Notes	Reference
	Standalone PET	Retrospective study of cancer patients	First report on 18F-FDG-uptake in arteries	Yun et al. (2001)
	Sequential PET/CT with retrospective imaging alignment	Prospective study (first)	First study specifically of atherosclerosis in humans using PET/CT	Rudd et al. (2002)
	Hybrid PET/CT	Retrospective study of cancer patients	First report using true hybrid PET/CT to study atherosclerosis	Tatsumi et al. (2003)
MΦ (CD68), immunochemistry	Sequential PET/CT and PET/MRI with retrospective imaging alignment	Prospective and correlational to cell type (MΦ)	First noninvasive study to assess inflammation by MΦ infiltration quantitatively	Tawakol et al. (2006)
Drug intervention, Simvastatin	Sequential PET/CT with retrospective imaging alignment	Prospective study of patients screened for cancer	First interventional study, attenuation of 18F-FDG-uptake found	Tahara et al. (2006)
PET reproducibility in atherosclerosis	Hybrid PET/CT	Prospective study of patients with vascular disease	Very good interscan variability for the internal carotid artery: 0·90; CI: 0·68–0·97	Rudd et al. (2007)
CD68, cathepsin K, MMP-9 and IL-18 gene expression	Hybrid PET/CT	Prospective study of patients receiving surgery (CEA) for atherosclerosis	First study of the molecular pathology of atherosclerosis	Graebe et al. (2009)
Angiogenesis; gene expression of αVβ3, CD34 and VEGF	Hybrid PET/CT	Prospective study of patients receiving surgery (CEA) for atherosclerosis	First study of angiogenesis using 18F-FDG	Pedersen et al. (2012)
Hypoxia; gene expression of HIF-1α	Hybrid PET/CT	Prospective study of patients receiving surgery (CEA) for atherosclerosis	First study of hypoxia using 18F-FDG	Pedersen et al. (2013)
Comparison of PET/MRI to PET/CT	Hybrid PET/MRI	Feasibility of simulta-neous PET/MRI for 18F-FDG imaging of carotid arteries	First PET/MRI in carotid arteries	Ripa et al. (2013)

αVβ3, integrin dimer consisting of integrin αV and integrin β3; CD34, cluster of differentiation 34; CD68, cluster of differentiation 68 – a macrophage marker; CEA, carotid endarterectomy; CI, confidence interval (95%); CT, computed tomography; 18F-FDG, 2-[18F]-fluoro-2-deoxy-D-glucose; HIF-1α, hypoxia inducible factor-1α; IL-18, interleukin 18; MΦ, macrophages; MMP-9, matrix metalloproteinase-9; MRI, magnetic resonance imaging; PET, positron emission tomography; VEGF, vascular endothelial growth factor.

Table 2

Novel PET tracers for human use: Potential in atherosclerosis risk stratification?

Target	Modality	Study type	Notes	Reference
Active calcification using 18F-NaF	Hybrid PET/CT	Retrospective study of cancer patients	First report and feasibility study of atherosclerosis in different arterial vascular beds using 18F-NaF	Derlin et al. (2010)
MΦ activity (SSTR2)	Hybrid PET/CT	Retrospective study of cancer patients	First study of MΦ activity in atherosclerosis using the tracer 68Ga-DOTATATE	Rominger et al. (2010)
MΦ activity (SSTR2) and glycolysis	Hybrid PET/CT	Retrospective study of cancer patients	First comparison of 18F-FDG and 68Ga-DOTATATE in atherosclerotic disease	Li et al. (2012)
Hypoxia	Standalone PET	Prospective study of cancer patients	First study of hypoxia using 18F-FMISO uptake in cancer patients	Valk et al. (1992)
Hypoxia	Standalone PET	Prospective study of cancer patients	First study of hypoxia using 62Cu-ATSM uptake in cancer patients	Takahashi et al. (2000)
Angiogenesis (αVβ3)	Standalone PET	Biodistribution and pharmacokinetics study	First study of angiogenesis using 18F-Galacto-RGD in cancer patients	Beer et al. (2005)

αVβ3, integrin receptor dimer alphaVbeta3; CT, computed tomography; 62Cu-ATSM, 62Cu-diacetyl-bis(N4methyl-thiosemicarbazone); 18F-FDG, 2-[18F]-fluoro-2-deoxy-D-glucose; 18F-FMISO, 18F-fluoromisonidazole; 18F-NaF, 18F-sodium fluoride; 68Ga-DOTATATE, 68Ga-[1,4,7,10-tetraazacyclododecane-N,N′,N″,N‴-tetraaceticacid]-ᴅ-Phe1,Tyr3-octreotate; MΦ, macrophages; PET, positron emission tomography; SSTR2, somatostatin receptor subtype 2.

18F-FDG

18F-FDG is the workhorse among the PET tracers. It is a glucose analogue with the positron emitter 18F (t½ = 110 min) substituted for the OH-group in the second position of the carbon backbone of D-glucose. A reproducible synthesis method of 18F-FDG was first reported in the late 1970s (Gallagher et al., 1977; Ido et al., 1978); however, as early as 1976, the first two human volunteers at the University of Pennsylvania were administered 18F-FDG. 18F-FDG enables visualization of tissues with an elevated level of glycolysis by a process of metabolic trapping (Gallagher et al., 1978), Fig.1.

Figure 1

Metabolic trapping of 18F-FDG: The inflammatory active cells take up 18F-FDG via GLUT1/3. Phosphorylation by hexokinase in the cytosol yields 18F-FDG-6-Phosphate which cannot be further processed by the metabolic machinery of the cell, effectively trapping 18F-FDG. The slim arrow depicts the reciprocal reaction which does occur but is negligible. 18F-FDG, 2-[18F]-fluoro-2-deoxy-D-glucose.

In 1977 the first 18F-FDG scan was reported using a human subject. The images were recorded of the cerebrum, however, a non-PET scanner was used (Reivich et al., 1977). The discovery that a high metabolic demand for glucose was a key requirement of cancer cells was discovered as early as the 1920s by Otto Warburg and was aptly referred to as the Warburg effect ever since (Warburg et al., 1924). Today the heritage of Otto Warburg is carried on by 18F-FDG as the main PET tracer of choice in standalone PET and hybrid PET/CT imaging in cancer patient management (Von Schulthess et al., 2006; Czernin et al., 2010a). It has come to the attention of clinicians and researchers in turn that accumulation of 18F-FDG is linked not only to cancer, but to benign and more generalized pathology as well. Accordingly, 18F-FDG uptake is expected to be noticeable during the course of atherogenesis in which inflammatory activity and thereby glycolysis is high (Vallabhajosula & Fuster, 1997; Shreve et al., 1999).

In 2001 this awareness led to the first retrospective report of 18F-FDG uptake in the large arteries of cancer patients (Yun et al., 2001) and this was soon followed by the first prospective study designed to evaluate atherosclerotic disease using 18F-FDG PET/CT in man (Rudd et al., 2002). The latter study also used an elegant approach with tritiated deoxyglucose (an in vitro analogue of 18F-FDG) and autoradiography to demonstrate that activity accumulated in macrophage rich areas of atherosclerotic plaques. This made it plausible that macrophages were responsible for the 18F-FDG-uptake seen by PET/CT in vivo, see Fig.2, an observation very much in line with the emerging consensus that atherosclerosis is an inflammatory disease (Ross, 1999). As a natural consequence of this work, the first study to seek to quantitate lesion macrophages by immunohistochemistry (% CD68 staining) to in vivo 18F-FDG uptake; target to background ratio (TBR) found a good correlation between the two (r = 0·70; P<0·0001) when comparing sections of internal carotid lesions to corresponding imaging data obtained prior to CEA (Tawakol et al., 2006). An even better correlation was found when comparing mean 18F-FDG uptake to mean lesion (% CD68 staining) inflammation (r = 0·85; P<0·0001) (Tawakol et al., 2006).

This was later corroborated in two studies using patient subsets from the prior study to correlate% CD68 staining with TBR (Fifer et al., 2011; Figueroa et al., 2012). The earlier of those two studies made the comparison indirectly and was a cross-sectional study of retrospective nature. Although highly significant and comparable correlations were found in the three studies, the question of timing from the 18F-FDG PET scan to CEA could be raised as patients went for up to a month before having their plaque removed. That question was addressed when it was determined that 18F-FDG PET scans performed 2 weeks apart on the same cohort of patients exhibited an interscan variability (0·90, 95% confidence interval; CI: 0·68–0·97), inter-observer agreement (0·97; CI: 0·89–0·99) and finally intra-observer agreement (0·95; CI: 0·89–0·99) for the carotid arteries (Rudd et al., 2007). The latter study was followed by a set of recommendations for 18F-FDG PET/CT imaging in human subjects with atherosclerosis and a demonstration of even better performance in a study with a comparative design, but including more different vascular beds (Rudd et al., 2008). The low numbers of patients included in these two studies is a limitation (11 and 19 patients, respectively) and two retrospective cancer studies have quite clearly demonstrated that long term 18F-FDG-uptake in major arterial segments change in a large proportion of re-evaluated patients over the course of several months (Ben-Haim et al., 2006; Wasselius et al., 2009a). However, if time from the initial PET scan to CEA can be reduced to match the former studies (Rudd et al., 2007, 2008) or perhaps even shorter, the validity of the correlation between the imaging modality and the subsequent tissue analysis should, theoretically, pose no problem.

The thickness of the vessel walls of the large arteries are on the edge of the resolution of the best standalone PET systems today. Therefore, the emergence of true hybrid PET/CT scanners (Beyer et al., 2000) represented a leap in molecular imaging, which was quickly evaluated to establish its potential in human atherosclerosis research (Tatsumi et al., 2003). The first prospectively designed study to use optimal post-injection scan times (3 h) in combination with almost immediate plaque recovery came from our group in 2009. We demonstrated a correlation between molecular pathology (gene expression) and 18F-FDG-uptake using hybrid PET/CT with a focus on molecular markers of inflammation and vulnerability (Graebe et al., 2009). In that study CEA was performed the day after the 18F-FDG PET/CT scan and the correlation between 18F-FDG-uptake and gene expression of the macrophage marker CD68 (r = 0·71; P = 0·02 and n = 10) was comparable with the first immunohistochemical findings on a per-patient level (Tawakol et al., 2006). Further investigations which included more patients reproduced the findings regarding CD68 (r = 0·38; P<0·0001 and n = 17) using a slice-by-slice approach of each lesion for comparison with 18F-FDG uptake (Pedersen et al., 2010).

Atherogenesis leads to remodelling and thickening of the vessel wall intima which may precipitate hypoxia and lead to angiogenesis as a compensatory measure (Ribatti et al., 2008), Fig.2. Evidence of hypoxia was first produced in an elegant non-imaging study (Sluimer et al., 2008) and this was corroborated by an in vitro study using an 18F-FDG-analog to produce evidence that hypoxia, not inflammation, increased glucose-uptake by macrophages per se (Folco et al., 2011). The first study directly linking the marker of hypoxia; hypoxia inducible factor-1α (HIF-1α) to 18F-FDG-uptake in vivo was recently published by our group (Pedersen et al., 2013). So far a single study has explored the association between neoangiogenesis, by gene expression of the integrin dimer αVβ3 and 18F-FDG-uptake in man, however, no correlation was found (Pedersen et al., 2012). Finally, a brand new study introduced demonstrated the feasibility of simultaneous hybrid PET/MRI of the carotid arteries using 18F-FDG (Ripa et al., 2013).

The promise of 18F-FDG PET as a method to monitor medical intervention in atherosclerosis was explored in the first interventional study using the HMG-CoA reductase inhibitor simvastatin (in a dynamic dose regime; 5–20 mg day−1) in addition to dietary management compared with a purely dietary management group (Tahara et al., 2006). That study found that simvastatin alone attenuated plaque 18F-FDG-uptake in the thoracic aorta and/or carotid arteries, and that this correlated with high-density lipoprotein cholesterol (HDL-C) elevation. A more recent study using atorvastatin in low (5 mg day−1) and high (20 mg day−1) dose found that a reduction in 18F-FDG-uptake in the ascending aorta and the femoral artery was associated with low-density lipoprotein cholesterol (LDL-C) reduction in both groups, however, only the high dose of atorvastatin led to significant attenuation of the 18F-FDG signal when compared with the baseline scan (Ishii et al., 2010). Furthermore, no difference in 18F-FDG-uptake between the treatment groups was found, and therefore any dose-response relationship could not be elucidated. The latest statin-based interventional study used atorvastatin (40 mg day−1) to treat previously statin free (1 year) subjects with confirmed significant atherosclerosis of at least one vascular territory (Wu et al., 2012). Seven different arterial segments were measured to assess 18F-FDG-uptake and again a significant reduction of 18F-FDG-uptake was found with treatment. In agreement with the earlier study (Ishii et al., 2010) circulating LDL-C was lowered but in contrast to the first interventional study (Tahara et al., 2006) HDL-C was now found to be significantly reduced (Wu et al., 2012). No change, however, was seen in the coronary artery calcium (CAC) score or various adipose fat volume measurements. Earlier findings suggested that statin treatment reduce 18F-FDG active plaques and with it calcified late-stage plaque burden which support the notion that 18F-FDG accumulation represents a transient inflammatory state culminating in a calcified non-reversible endpoint (Wasselius et al., 2009b).

Very recently 18F-FDG PET has been introduced in clinical drug trials with TBR measurements as primary endpoint of an index vessel selected on the basis of the highest average maximum TBR before and after treatment with losmapimod (a p38 mitogen-activated protein kinase inhibitor) as add-on therapy to statins. Significant attenuation of 18F-FDG uptake in active sites of inflammation was found, however, not upon evaluation of the vessels using an all-segment TBR approach (Elkhawad et al., 2012). The dal-PLAQUE study used dalcetrapib (a cholesteryl ester transfer protein modulator) as add-on therapy to LDL lowering therapy. It was found that an increase in HDL-C was associated with a decrease in TBR of most-diseased-segments of artery as well as attenuation of total vessel area; however, that effect was only recorded in the carotid arteries, not the index vessels (Fayad et al., 2011). Active sites of inflammation are the most cell rich sites of atherosclerotic plaques with macrophages and vascular smooth muscle cells (VSMC) making up the bulk of the cells (Jonasson et al., 1986), we therefore suggest that any effect measured by 18F-FDG PET would most likely be associated with such characteristics.

With careful consideration of sample size; what is cumulatively learned by these interventional studies is that statin treatment reduces 18F-FDG-uptake in atherosclerotic plaques whether it be reported by the mean/maximal standardized uptake value (SUVmean, SUVmax) (Tahara et al., 2006; Wasselius et al., 2009b) or the target to background (TBR) model of 18F-FDG accumulation (Ishii et al., 2010; Wu et al., 2012). Interventional studies using novel drugs demonstrate that 18F-FDG PET can be used to assess drug effects on plaques in a clinical setting (Fayad et al., 2011; Elkhawad et al., 2012). The case is made for large cohort event-driven prospective studies to further elucidate the anti-inflammatory effect of statins as well as novel drugs for treatment of atherosclerosis and the eligibility of 18F-FDG-PET as a surrogate endpoint to monitor such non-invasive therapy.

Non-FDG PET-tracers in atherosclerosis imaging

From the diversity of studies presented, it becomes apparent that for its excellent properties 18F-FDG has one major drawback; its lack of specificity. As all living cells can utilize glucose we need a tracer that specifically targets the cell-mediated key molecular processes associated with the vulnerable atherosclerotic plaque. Most prominent of these targets are macrophage infiltration, calcification, apoptosis, hypoxia and neoangiogenesis of the intima/media (Ross, 1999; Tawakol et al., 2006; Ribatti et al., 2008; Sluimer et al., 2008; Dweck et al., 2012).

18F-NaF

One feature in atherogenesis is localized ‘spotty’ calcification which is thought to promote plaque vulnerability (Motoyama et al., 2007). The tracer sodium [18F]-fluoride 18F-NaF was originally developed and introduced for bone scintigraphy in 1962 (Blau et al., 1962) but has re-emerged recently with the introduction of PET, see Fig.2. 18F-NaF is deposited by chemisorption onto hydroxyapatite (Czernin et al., 2010b) and that infers hydroxyapatite presence within the atherosclerotic lesion itself. Mineralization of the artery wall can be initiated by VSMCs responding to stress by oxidized lipids (Watson et al., 1994; Yan et al., 2011) as well as inflammation and macrophage accumulation which coincide with osteogenic activity (Aikawa et al., 2007). It is beyond the scope of this review to extensively cover the background of osteogenesis in atherosclerosis; which has excellently been discussed previously (Doherty et al., 2003; Demer & Tintut, 2011). Using hybrid PET/CT and a retrospective approach it was recently described how 18F-NaF accumulated in atheroma of the aorta, iliac, femoral and carotid arteries (Derlin et al., 2010, 2011); however, also that coincidental 18F-NaF and 18F-FDG uptake (14 of 215 lesions; 6·5%) is rare (Derlin et al., 2011). This could be due to different lesion properties, e.g. inflammation versus calcification, although it cannot be ruled out that time between 18F-NaF and 18F-FDG scans may have influenced this result (Derlin et al., 2011). A recent study scanned the patients 1 day apart and found a modest correlation (r2 = 0·171, P<0·001) between 18F-NaF and 18F-FDG uptake in calcific aortic valvular disease (Dweck et al., 2013). Although this seems to support the initial observation of poorly overlapping uptake patterns of 18F-NaF and 18F-FDG, the studies are not directly comparable and data are not identically presented. Contrary to 18F-FDG, 18F-NaF specifically targets bone and bone turnover and we have no reason to believe that uptake of these two tracers should perfectly coincide. This lack of tracer uptake co-localization earlier led investigators to hypothesize that it was a reflection of differences in disease stage (Derlin et al., 2011).

In another study 18F-NaF uptake was found to be associated with major adverse cardiovascular events as well as an increase in Framingham risk prediction score (Dweck et al., 2012). More importantly 41% of patients with CAC score >1000 had no significant 18F-NaF uptake and that adds weight to the hypothesis that only metabolically active calcific plaques are imaged by 18F-NaF (Dweck et al., 2012), i.e. late plaques developing an increasingly vulnerable phenotype. However, bona fide histological validation of this hypothesis is still lacking. With careful consideration of the research we find it justified to assume that 18F-NaF PET scan adds independent information to what is learned from an 18F-FDG PET scan alone. A very recent publication produced evidence that plaque calcification followed focal inflammation in the same arterial location as detected by 18F-FDG PET at an earlier time point (Abdelbaky et al., 2013). Based on the newest data, further assessment of 18F-NaF PET in atherosclerosis should certainly be warranted, however, the first prospective studies are still to be seen.

68 Ga-DOTATATE

The inflammatory active atherosclerotic lesion is continuously infiltrated by blood monocytes that differentiate into macrophages which eventually become lipid loaded foam cells (Ross, 1999). The association between 18F-FDG uptake and macrophages has been substantiated earlier (Rudd et al., 2002; Tawakol et al., 2006), however, by targeting the somatostatin receptor subtype 2 (SSTR2) expressed by macrophages (Dalm et al., 2003; Armani et al., 2007), it should be possible to obtain higher specificity than what 18F-FDG PET offers (Fig.2) and potentially reveal all states in plaque development where macrophages are abundant. The somatostatin receptor ligand 68 Ga-[1,4,7,10-tetraazacyclododecane-N,N′,N″,N‴-tetraaceticacid]-ᴅ-Phe1,Tyr3-octreotate (DOTATATE) has high affinity for SSTR2 (Breeman et al., 2005) and can be used in combination with 18F-FDG for imaging and evaluation of neuroendocrine tumors by clinical PET (Kayani et al., 2008). The potential for imaging macrophages in atherosclerosis was first explored in a study comparing 68 Ga-DOTATATE uptake in the coronary arteries with presence of calcified plaques and risk factors for cardiovascular disease (Rominger et al., 2010). Male sex and prior vascular events correlated with 68 Ga-DOTATATE uptake (TBR) and a correlation with calcified plaques was also found (R = 0·34; P<0·01). However, in 14 of 25 (56%) of cases no colocalization with calcified plaques and 68 Ga-DOTATATE uptake was found, suggesting not all calcified plaques had a significant macrophage population (Rominger et al., 2010). The fact that less than half of the calcified plaques exhibited 68 Ga-DOTATATE accumulation could suggest that end-stage or ‘stable’ calcified plaques without active inflammation may account for the bulk of the 68 Ga-DOTATATE negative calcified plaques. It would be interesting to see what could be learned from comparison between 68 Ga-DOTATATE and 18F-NaF scans in this context. The second of the two existing studies using 68 Ga-DOTATATE so far, compared 68 Ga-DOTATATE to 18F-FDG uptake in a retrospective study of cancer patients (Li et al., 2012). The earlier findings was corroborated when 68 Ga-DOTATATE was found to correlate with calcified plaques (R = 0·52; P<0·05), however, less than half (43%) of vascular sites with the highest focal 68 Ga-DOTATATE uptake was also 18F-FDG positive and conversely only 28% of the sites with the highest focal 18F-FDG uptake sites co-localized with 68 Ga-DOTATATE (Li et al., 2012). Clearly 68 Ga-DOTATATE and 18F-FDG are not equally distributed. As pointed out by the researchers themselves the tracer kinetics of 68 Ga-DOTATATE and 18F-FDG could be an issue as saturation kinetics could be feature of SSTR2 receptor imaging which was not the case when using 18F-FDG (Li et al., 2012). Another study found that proliferating human umbilical vein endothelial cells expressed SSTR subtypes 2 and 5 in vitro suggesting a possible role in angiogenesis (Adams et al., 2005). We know angiogenesis is an important feature of the advanced plaque and therefore we cannot rule out that 68 Ga-DOTATATE may detect more than macrophages in atherosclerosis imaging. Considering these facts, the number of included patients (n = 70 and n = 16, respectively) and the retrospective nature of these two studies, care must be taken when interpreting the results. However, the stronger correlation of 68 Ga-DOTATATE than 18F-FDG with cardiovascular risk factors encourages further investigations of the distribution differences between these two tracers. Histological validation of 68 Ga-DOTATATE uptake in the atherosclerotic plaque is a logical next step as is the elucidation of the biological significance of tracer distribution for the vulnerable plaque.

Future PET-tracer targets

As mentioned earlier perhaps the three most obvious molecular processes to target in plaque imaging is hypoxia, neoangiogenesis and apoptosis. First, hypoxia of the atherosclerotic arterial wall has been found to colocalize with accumulation of foam cells and macrophages in vivo in an experimental rabbit model as well as in humans (Bjornheden et al., 1999; Sluimer et al., 2008). Evidence that hypoxia increase macrophage 18F-FDG uptake has been observed (Folco et al., 2011) suggests that imaging hypoxia itself may offer an alternate approach, one that is already pursued in cancer imaging (Valk et al., 1992; Takahashi et al., 2000). Second, the vitronectin receptor αVβ3 is associated with neoangiogenesis and known to be expressed in atherosclerotic arteries (Hoshiga et al., 1995). It is recognized by and binds to the arginine-glycine-aspartic acid (RGD) motif. So far human imaging studies performed with a tracer based on RGD-ligands are scarce (Beer et al., 2005, 2006) and preclinical testing has shown promise as well (Oxboel et al., 2012; Pohle et al., 2012). Third, phosphatidylserine (PS) a cell membrane phospholipid, is externalized to the outer cell membrane during apoptosis, a key feature of the vulnerable atherosclerotic plaque (Koopman et al., 1994; Kolodgie et al., 2000). Thus targeting PS is a definite possibility and has been performed as early as 2003, however, not using PET (Van de Wiele et al., 2003). With these data in mind the first clinical PET study of atherosclerosis focusing on these key molecular processes (Fig.2) is expected in the not too distant future.

Indications for introduction of atherosclerosis PET imaging

A significant number of patients are known to suffer acute ischemic events in the absence of significant (≤50%) artery stenosis (Aldrovandi et al., 2012; Kovacic & Fuster, 2012). Indications for CEA are based on ipsilateral symptomatic ischemic events distal to a carotid plaque of ≥70% degree stenosis as determined by ultrasound. Although some qualitative information can be evaluated using modern ultrasound systems, selection of patients for CEA remains largely based on the degree of stenosis (Hobson et al., 2008). In our eyes this is an inadequate approach considering the patient cohort with non-significant carotid artery stenosis who experience acute ischemic stroke. What is needed is plaque characterization whereby plaque vulnerability can be evaluated. PET could potentially be the tool that, together with ultrasound, improves patient selection criteria thereby reducing the need to treat ratio. Recent work has indicated that CEA of symptomatic low grade carotid stenosis is safe and thus better selection criteria sensitivity are warranted and should be pursued (Ballotta et al., 2013). PET is a powerful technique with unsurpassed sensitivity and should be considered as a tool for achieving this goal.

Challenges in atherosclerosis PET imaging

Recently, the concept of the ‘vulnerable patient’ has been introduced (Naghavi et al., 2003a,b). First, to find a regional vulnerable internal carotid artery plaque using PET does not take into account the global status of the patient being evaluated for CEA and therefore co-morbidities characterizing the vulnerable patient could be missed in such a setup. Second, although arterial 18F-FDG uptake is highly correlated between different large artery vascular territories (Rudd et al., 2009), evaluation of 18F-FDG uptake in large versus small arteries such as the coronary and cerebral arteries is largely hampered by the significant myocardial and cerebral 18F-FDG uptake. On the other hand, CT assessment of plaque burden (calcium score) in the aorta and coronary arteries revealed a strong correlation between the two (Kim et al., 2011). This could be an indication of what should be further investigated using PET in a similar setup if technically feasible. Bearing this in mind, atherosclerosis is a systemic disease and vascular territories should be evaluated separately when clinically indicated to ensure optimal treatment. A third challenge is cost: PET scans are not inexpensive and therefore PET has been subjected to economic evaluations from early clinical implementation and proved cost effective in the management of an array of cancers (Buck et al., 2010). Accordingly, if effective in identifying vulnerable carotid plaques, the cost of an unnecessary CEA surgery clearly outperforms the cost of a PET scan several-fold not to forget the unnecessary discomfort and added risk inflicted upon the patient. Finally, a clinical 18F-FDG PET/CT scan subjects the patient to ionizing radiation. In a typical whole body scanning procedure the absorbed dose from the tracer is less than half the dose received by that from the CT modality alone (Brix et al., 2005). The unmatched sensitivity, the fact that the technique is non-invasive and quantifiable, is considered to outweigh the drawbacks when imaging human atherosclerosis (Owen et al., 2011). The focus of this review has been the potential role of PET in atherosclerotic disease management. Hybrid PET/CT and PET/MRI provide complementary information to the PET modality; however, it is beyond the scope of this review to venture further into the specific advantages of each alternative imaging modality. However, some key characteristics of different important imaging modalities in plaque characterization have been summarized in Table3.

Table 3

Comparison of non-invasive imaging modalities for atherosclerotic plaque characterization

	PET	MRI	CT	US
Plaque ‘activity’ (metabolism)	+++	+	−	+
Plaque composition	−	+++	+	++
IMT – vessel wall characteristics	−	++	−	+++
Calcification	−	+	+++	++
Sensitivity	+++	++	−	−

+++, very good performance; ++, intermediate performance: +, limited performance; −, not applicable; CT, computed tomography; IMT, intima-media thickness; MRI, magnetic resonance imaging; PET, positron emission tomography; US, ultrasound.

Concluding remarks

It is perhaps not immediately intuitive, but many molecular targets are shared by cancer and atherosclerotic disease. No finer example of this was the way atherosclerosis imaging research started; with the observation that some cancer patients had an increased 18F-FDG uptake in the large arteries (Yun et al., 2001). From that, as well as the other examples reviewed, it is likely that future tracer ‘firsts’ in atherosclerosis imaging will continue to be of retrospective nature with a foundation in cancer research. It follows that by rationally and systematically re-evaluating ‘old’ tracers in retrospective studies before putting them to use in a new context, the achievements in PET-based atherosclerosis research brings hope for a better understanding of this disease on a molecular level. From what is seen in this review it is obvious that there is a continuous drive in the imaging community towards testing of existing tracers in new contexts. As important as this, there is focus on the development of new tracers for molecular imaging of atherosclerosis using PET. What knowledge is gained from the existing tracers should be applied for further development of new tracers followed by comprehensive prospective preclinical testing in animal models. Finally, promising new tracers should be introduced in small populations paving the way for full clinical implementation. In atherosclerosis research, the aim is the development of a safe clinical tool for non-invasive in vivo risk stratification of patients with symptomatic atherosclerotic disease. With apoptosis and new angiogenesis tracers close to clinical introduction the future for molecular imaging is rich in perspectives. The goal should be to learn as much as possible whereas improving and ideally substituting current implementations to benefit future patients in need of stratification for vascular surgery.

On a final note the first true event-driven prospective studies in atherosclerosis using molecular imaging were initiated in 2008/2009; the high-risk plaque initiative (Falk et al., 2011) and the associated BioImage study (Muntendam et al., 2010). As we are waiting for these results we should commit ourselves to continuously apply new as well as established tracers in new contexts in clinical research working towards an even brighter future for patients with atherosclerosis as well as for molecular imaging itself.

Conflict of interest

The authors have no conflicts of interest.