Atherothrombosis is a major cause of morbidity and mortality in the Western world and is now widely accepted as a chronic systemic vascular inflammatory disorder. Inflammation is a recognized risk factor for the vulnerable atherosclerotic plaque and is fundamental to lesion progression and destabilization. The macrophage is one of the key cellular mediators of this process (1), and histological studies have shown that the degree of macrophage infiltration increases the risk for plaque rupture and subsequent thromboembolism (2). Detection of macrophage activity, and hence inflammation within atheroma, could potentially distinguish between vulnerable and more stable states and remains one of the key goals of atheroma imaging.

It has recently been possible to identify carotid plaque inflammation noninvasively with ultrasmall superparamagnetic iron oxide (USPIO)-enhanced magnetic resonance imaging (MRI) in both animal (3) and in vivo human studies (4). The concept that the USPIO contrast medium Sinerem (Guerbet, Roissy, France) may be useful in the assessment of inflammatory activity in atherosclerotic plaque is strongly supported by the endothelial dysfunction theory (5). Dysfunctional endothelium initiates and sustains an inflammatory reaction within the arterial wall and allows the accumulation of plasma components, such as low-density lipoproteins (LDLs), into the subendothelial space. Oxidized LDL particles are then phagocytosed by macrophages, ultimately forming foam cells. The USPIO particles, of a diameter similar to that of LDL (15 to 25 nm), enter atherosclerotic plaques with an increased endothelial permeability and accumulate in atheroma with high macrophage content. This has been shown by experimental investigations in an atherosclerotic animal model (6) using hyperlipidemic rabbits. Histology and immunohistochemistry colocalized USPIO to activated macrophages in the plaque. These changes were not observed in animals that did not receive USPIO. It was concluded that USPIO was phagocytosed by macrophages in atherosclerotic plaques in a quantity sufficient to cause susceptibility artifact detectable by MRI.

Since the first statin was approved by the Food and Drug Administration (FDA) in 1987, these lipid-lowering agents have quickly become the gold standard for treatment of hypercholesterolemia and the cornerstone of therapy for primary and secondary prevention of atherosclerotic disease (7). Statins exert their benefits primarily by inhibiting de novo cholesterol synthesis and increasing the expression of low-density lipoprotein cholesterol (LDL-C) receptors, which result in significant reductions of total cholesterol and LDL-C plasma levels. Furthermore, a recent meta-analysis demonstrated a dose-dependent relationship between statin therapy and cardiovascular outcome (8), and high-dose statin therapy may even decrease the size of atheromatous plaques over time (9). It remains controversial whether LDL-C lowering is the only mechanism underlying the clinical benefit of statins, and it has been postulated that statins may also exert their positive clinical effect by LDL-C independent or pleiotropic mechanisms (10- 11). One of these is attenuation of inflammation, because statins have been shown to decrease systemic inflammatory markers (12- 13). To date, the only noninvasive evidence that statins reduce plaque inflammation in clinical practice has been with ¹⁸fluorodeoxyglucose positron emission tomography, but the associated radiation exposure limits the number of positron emission tomography studies that patients can undergo, reducing its potential contribution in assessing therapeutic interventions.

The aim of this study was to evaluate the effects of low-dose (10 mg) and high-dose (80 mg) atorvastatin (Lipitor, Pfizer, New York, New York) on macrophage activity and carotid plaque inflammation as determined by USPIO-enhanced carotid MRI. The hypothesis was that 12 weeks of treatment with atorvastatin would reduce USPIO-enhanced MRI signal loss (marker of plaque inflammation) as compared with baseline and that the reduction in signal loss would be greater in the high-dose group.

The evaluation using USPIO-enhanced high-resolution MRI in carotid disease was done as a prospective, randomized, single-center, double-blind clinical trial (ISRCTN64894118) (14). The Institutional Review Board at the study center approved the protocol, and all participating patients provided written informed consent.

Forty-seven subjects were randomly assigned to a high- or low-dose statin group. The high-dose statin group received 80 mg atorvastatin once daily, and the low-dose group received 10 mg atorvastatin once daily for 12 weeks. The USPIO-enhanced MRI was performed at baseline (i.e., before randomization) and at 6 and 12 weeks. Blood samples including fasting lipid profile, creatine kinase, liver function tests, and serum creatinine levels were obtained at each pre-USPIO infusion visit to ensure patient safety. Blood samples for atorvastatin pharmacokinetics and cardiovascular biomarkers, including highly specific C-reactive protein, myeloperoxidase, and lipoprotein-associated phospholipase A2 (Lp-PLA2) activity were also collected at these visits and were outsourced for independent analysis (Quest Diagnostics, London, United Kingdom). Transcranial Doppler monitoring of the middle cerebral artery was performed to measure thromboembolic signals at baseline and at 6 and 12 weeks.

Patients were eligible if they had clinically documented atherosclerotic carotid disease and had demonstrated the presence of inflammation within their carotid lesions on USPIO-enhanced MRI regardless of symptomatic status and met the inclusion criteria and exclusion criteria. The patients were randomly allocated after the initial carotid ultrasound had shown at least 40% stenosis and the MRI demonstrated signal loss on USPIO-enhanced MRI consistent with inflamed lesions. Inclusion criteria were as follows: 1) male or female 18 to 80 years of age at baseline; female subjects had to be of nonchildbearing potential (post-menopausal women who have been amenorrheic >1 year or pre-menopausal women with a documented hysterectomy or bilateral oophorectomy); 2) positive USPIO-enhanced MRI of carotid plaque confirmed by an attending neuroradiologist (J.H.G.); this was defined as a signal decrease of ≥10% in 2 or more quadrants between pre- and post-USPIO MRI studies (see Image Analysis section); and 3) must either have been statin naïve or have been on a stable dose of a statin for ≥4 weeks before screening, with no evidence of statin intolerance; permitted statins and total daily dose were as follows: atorvastatin ≤10 mg, simvastatin ≤40 mg, pravastatin ≤40 mg, fluvastatin ≤80 mg, and rosuvastatin ≤10 mg.

A subject was not eligible for inclusion in this study if any of the following criteria applied: 1) required continued use of nonstatin lipid-modifying therapies (e.g., fibric acid derivatives, niacin, resins, ezetimibe, fish-oil supplement) or therapy with any other lipid-regulating medications not specified (dietary supplements or multivitamins containing a daily dose of niacin <250 mg were allowed); 2) history of statin intolerance; 3) history of chronic viral hepatitis (including presence of hepatitis B surface antigen or hepatitis C antibody) or other chronic hepatic disorders, alanine aminotransferase or aspartate aminotransferase >1.5 times the upper limit of normal, or alkaline phosphatase or total bilirubin >1.5 times the upper limit of normal of laboratory reference range at screening; 4) renal impairment with creatinine clearance <50 ml/min; 5) history of myopathy or inflammatory muscle disease, or elevated total serum creatine kinase (3 times upper limit of normal) at screening visit; 6) Doppler assessment of <40% stenosis or carotid occlusion during screening assessment; 7) history of clinically significant atopy (e.g., anaphylaxis, skin rash to medication or topical therapies, hypersensitivity to iodinated contrast media, allergies to food such as shellfish, bronchial asthma) and allergies to dextran and iron salts; 8) clinical contraindications to MRI including but not limited to intracranial aneurysm clips (except Sugita) with an appropriate operative conformation, history of intraorbital metal fragments, pacemakers and non-MRI compatible heart valves, inner ear implants, or history of claustrophobia in MRI; 9) planned carotid surgery or endovascular intervention within 10 weeks of the study period; 10) serum triglycerides >400 mg/dl (4.52 mmol/l) at screening; 11) untreated or uncontrolled hypothyroidism or thyrotoxicosis; 12) patients with poorly controlled diabetes mellitus with hemoglobin A1c ≥8.5%; 13) patients with poorly controlled hypertension (systolic blood pressure ≥160 mm Hg and/or diastolic blood pressure ≥100 mm Hg); 14) history of malignancy within the past 2 years, other than nonmelanoma skin cancer; 15) evidence of recent (<4 weeks) severe infection (e.g., pneumonia, cellulitis); 16) current life-threatening condition other than vascular disease (e.g., very severe chronic airways disease, human immunodeficiency virus positive, life-threatening arrhythmias) that may prevent subject from completing the study; 17) history of myocardial infarction or cerebrovascular accident within the previous 2 weeks of screening; 18) concomitant use of potent CYP450 3A4 inhibitors (e.g., clarithromycin, grapefruit juice [>240 ml daily], itraconazole, ketoconazole); 19) chronic use of nonsteroidal anti-inflammatory drugs and oral steroids therapy (“cardiovascular” doses of aspirin up to 325 mg daily were allowed); 20) chronic use of immunosuppressants (cyclosporine, methotrexate, and so forth); 21) use of an investigational drug within 30 days or 5 half-lives (whichever is the longer) preceding the first dose of study medication; 22) inability to give informed consent; 23) prior ipsilateral carotid endarterectomy or neck irradiation; and 24) alcohol or drug abuse within the past 6 months.

The conduct of the MRI studies did not cause a delay in surgical intervention of any of the enrolled subjects because planned carotid surgery or endovascular intervention earlier than 10 weeks within the study period was an exclusion criterion. Patients who withdrew from the study for safety reasons or otherwise were replaced with new patients with a corresponding treatment unit number so that the final number in each group would be balanced (n = 20).

Multicontrast imaging of both the left and right internal carotid artery was acquired at 1.5-T before USPIO infusion and 36 h after infusion, using a whole-body MRI system (Signa HDx, GE Diagnostic Imaging, Waukesha, Wisconsin) and a custom-designed 4-channel phased-array neck coil (PACC, Machnet BV, Elde, the Netherlands) to improve signal-to-noise ratio. Movement artifact was minimized using a dedicated vacuum-based head restraint system (VAC-LOK Cushion, Oncology Systems Limited, Crawley, United Kingdom) to fix the head and neck in a comfortable position and allow close apposition of the surface coils.

After an initial coronal localizer sequence, axial 2-dimensional time-of-flight MR angiography was performed to identify the location of the carotid bifurcation and the region of maximal stenosis on each side. Axial images were acquired through the common carotid artery 12 mm (4 slices) below the carotid bifurcation to a point 12 mm (4 slices) distal to the extent of the stenosis identified on the time-of-flight sequence. This method ensured that the entire carotid plaque was imaged and also facilitated image coregistration during repeat imaging.

The following 2-dimensional electrocardiography-triggered, fat-suppressed fast spin echo pulse sequences employing double-inversion recovery blood suppression with a voxel size of 0.4 × 0.4 × 3mm were used in each case: T₁-weighted (repetition time [TR]/effective echo time [Teeff]/echo train length [ETL]: 1 R-R/7.8 ms/12), short T₁inversion recovery (STIR) (TR/TEeff/TI/ETL: 2 R-R/40 ms/150 ms/24), intermediate T₂-weighted (TR/TEeff/ETL: 2 R-R/40 ms/24). T₂*-weighted imaging was performed using a 2-dimensional multishot spiral acquisition employing a spectral-spatial excitation pulse (TR/TE:2 R-R/5.6 ms). The multishot spiral sequence involved the acquisition of 22 spiral interleafs each comprising 4,096 data points, resulting in an effective in-plane pixel size of 0.42 × 0.42 mm, 2 signal averages were performed, and a quadruple inversion recovery (QIR) preparation was utilized to null the signal from blood before and after USPIO. Slices were acquired sequentially with a 3-mm thickness and no interslice gap.

The use of a small field of view means that there are no pure noise regions with which to adjust the measurements. The apparent tissue-free regions in the images actually contain structured noise, probably due to subtle patient motion such as swallowing. Because we are using spiral imaging sequences, these artifacts are diffusely spread around the image rather than being coherent ghosting artifacts as would be expected in Cartesian acquisitions. The usual method of obtaining noise estimates is to take a region well away from any tissue to ensure that ghosting artifacts do not contribute to the measurement. The images in the figures are windowed such that these artifacts are not apparent, but that does not mean that the tissue-free regions provide valid estimates of background noise.

The USPIO compound, Sinerem, was supplied as a dry powder and initially made up to a volume of 15 ml with normal saline. The contrast medium was further diluted in 100 ml of normal saline and given as a slow infusion through an in-dwelling large-bore intravenous cannula over 30 min. The dose used was 2.6 mg/kg. The physiochemical properties and safety data for this contrast agent have been published previously (15). Documented contrast reactions, albeit rare, include headache, nausea, vomiting, backache, and hypersensitivity reactions.

All MRI scans were analyzed by 2 experienced independent readers (T.Y.T. and S.P.S.H.). The readers were blind to the patients' demographic data and statin dose. The sequence order of images on the pre-infusion images was maintained on each set of post-infusion images, but the latter was viewed in a random order.

Before and after USPIO, MRI scans were manually coregistered according to plaque morphology and distance from the carotid bifurcation at the time of imaging. After this, images were manually segmented into quadrants, using pre-defined rules, and therefore less susceptible to interobserver error, excluding the luminal blood pool (CMR Tools, London, United Kingdom). A horizontal line was constructed across each image and through the lumen center. Likewise, a perpendicular line was constructed ensuring that the “cross hairs” were centered through the lumen. These quadrants were defined as regions of interest (ROIs), and mean signal before and after USPIO infusion was reported within the CMR Tools software. For each patient, only the slices containing plaque were analyzed.

Because the neck coil was always in a slightly different position between imaging sessions, there was not only spatial but also temporal variation to consider. Hence, signal intensities in each quadrant were normalized to the adjacent sternocleidomastoid muscle before and after USPIO, which does not take up Sinerem in any appreciable amount. We have attempted to address the issue of coil positioning variability by normalizing to the adjacent muscle. Although there is still a coil inhomogeneity effect, we consider that the variation across the diameter of the vessel is not a major influence. Signal intensity change (ΔSI) was then calculated as the difference (after minus before) in these normalized quantities. Note that the greater the USPIO concentration (i.e., the greater the macrophage activity), the more negative the resultant value for ΔSI. It follows that a reduction over time in macrophage activity will reduce the post-USPIO concentration, and result in a positive change in ΔSI with time.

The Embo-Dop transcranial Doppler (TCD) system (Compumedics Germany GmbH, Singen, Germany) was used (version 2.1, Multi X-Laboratory, Compumedics Germany GmbH). This insonates simultaneously at 2.0 and 2.5 MHz, and the standard system settings designed to optimize emboli signal (ES) detection as recommended in the Consensus criteria (16) were adhered to. Mean depth was 51.1 mm (SD 4.6). A sample volume of 5 mm³ was routinely employed, although in a few cases, this had to be increased to obtain an adequate signal. Power was 64 to 100 mW. Bilateral recordings were performed for 1 h, although if there were problems with insonation, preference was given to the ipsilateral middle cerebral artery. The transducers were held in place by a proprietary headpiece supplied with the system. The Embo-Dop system uses an event detector system using a previously published algorithm (17) to identify potential candidate events. No distinction was made between whether the system categorizes ES into solid or gaseous emboli, as it was likely that all emboli were solid in nature. The spectra from all saved ES were recorded onto the hard drive of the computer, and all signals were later reviewed offline in consensus by 2 experienced observers in ES detection who were blinded to the patients' demographic and lipid profiles. These ES were used for further statistical analysis.

A sample size of 40 patients randomly allocated 1:1 to high (80 mg) or low (10 mg) doses of atorvastatin was expected to provide 90% power overall to detect a difference (with 5% type-1 error) between the groups in USPIO-enhanced MRI signal change at 12 weeks. Assumptions used in this calculation were based on the results of a study comparing a group of symptomatic to a group of asymptomatic patients (18) that estimated the difference between these groups to be 0.097, with a between-subject standard deviation of 0.093.

The trial database was transferred from the trial center to the sponsor once all end points had been analyzed. Unblinding of the treatment assignment occurred only after this had happened to avoid bias, and permitted independent confirmation of the analyses.

For continuous variables with a normal distribution, the mean and SD were reported. For log-normally distributed variables (e.g., laboratory parameters), the geometric mean and coefficient of variation (%CV) were calculated. For count data, the median and range were reported. Baseline balance was not tested statistically. Interobserver and intraobserver agreement for reproducibility and repeatability, respectively, were determined for quadrant analysis for USPIO signal change using the Bland-Altman method.

Statistical modeling for estimation and hypothesis testing was carried out as described in the following text. All reported p values were 2-sided and not adjusted for multiple testing. A value of p < 0.05 was considered to indicate statistical significance.

USPIO-enhanced MRI signal change (ΔSI) was analyzed using a repeated-measures mixed model with change from baseline in ΔSI as the outcome variable. Group (low dose/high dose), visit (week 6/week 12) and baseline signal change were fitted as fixed effects and patient as a random effect. Estimates of mean signal change within each group were calculated with appropriate 95% confidence interval (CI), adjusting for the correlation between measurements made on the same subject. Note that a significant change in 1 group and a nonsignificant change in another group do not necessarily imply that the groups changed differently. To address this directly, an estimate of the difference between groups with an appropriate 95% CI was also calculated, together with a p value testing the hypothesis of no difference between groups. Other end points were analyzed using similar models, but assuming a Poisson rather than a normal distribution for count data. Model assumptions regarding homogeneity of variance were verified by inspection of residual plots. Distributional assumptions regarding normality were verified by assessment of normal probability plots.

Recruitment started in July 2006, and the last subject visit was in August 2007. Sixty-four patients were screened, and 47 patients were enrolled into the study. Seven patients did not complete the study because of an adverse event (n = 2; both patients were in the high-dose group and had deranged liver function tests during the study that were outside the limits of acceptability from the protocol), withdrawn consent (n = 1), or other reasons not associated with this specific study (n = 4). All patients remained asymptomatic during the study, although 1 patient on the low-dose statin regimen had a completed ipsilateral hemispherical stroke 3 weeks after finishing the trial. All 40 patients completing the study had, by definition of trial design, matched baseline and 6- and 12-week USPIO-enhanced MRI studies for identification of the vessel boundaries and compositional analysis. These 40 patients comprised the analysis population (6). No enrolled patient had a high-grade symptomatic stenosis, as they would have been operated upon within 2 weeks of presentation according to local protocol.

Patient demographics are shown in (Table 1). A comparison of laboratory measures, image analysis, and TCD end points at baseline is presented in (Table 2). In general, differences between the high- and low-dose groups were small. The 6-year difference in age between the groups was not considered clinically relevant.

Patients had between 3 and 8 MRI slices, providing cross-sectional images of the carotid plaque. This resulted in a mean of 4.6 slices per high-dose patient (range 3 to 6), and 4.75 slices per low-dose patient (range 3 to 8). Two patients (1 high dose, 1 low dose) had bilateral carotid disease. Both sides from these patients were included for analysis.

Interobserver agreement was high for USPIO signal change calculations. The intraclass correlation coefficient (1-way random effects) was 0.90. The 95% limits of agreement were −0.146 to 0.168. This was based on 10 patients, 24 paired patient visits, 118 slices, and 906 matched quadrants. There was some evidence that for quadrants with a relatively large signal, 1 reader derived a less extreme value than the other, although the number of quadrants affected by this was <3%. Repeatability was also based on 10 patients, 58 slices, and 232 matched serial quadrants. The intraclass correlation coefficient was 0.85 between 0 and 6 weeks and 0.81 between 6 and 12 weeks on the pre-USPIO imaging scan.

Results of the statistical analysis of the change from baseline in image analysis, microemboli counts, laboratory measures, and biomarkers are presented in (Table 3). Significant differences were seen between groups at 12 weeks for ΔSI, microemboli count, LDL-C, total cholesterol, and plasma Lp-PLA2 activity. Atorvastatin concentrations in the 2 groups are also presented. The mean concentration was approximately 5-fold higher in the high-dose group, though the range of concentrations was very large.

Examples of the imaging obtained from a low- and high- dose patient are shown in (Figures 1, 2), respectively. (Figure 3) is another example of a high-dose patient at baseline and at 12-week imaging. It also depicts the quadrant analysis approach used in the image analysis. (Figures 4, 5) show the change in signal intensities and microemboli count, respectively, at 6 and 12 weeks.

Although the following was not identified as a primary or secondary end point before trial initiation, we evaluated our results for possible associations between changes in serum cholesterol levels and changes in USPIO-enhanced MRI-defined inflammation and other measurements. We observed a number of moderate correlations (Spearman |r| = 0.4 to 0.6), including between relative changes in ΔSI and LDL-C (Spearman r = −0.48, p = 0.0036), between changes in ΔSI and change in microemboli count (Spearman r = −0.58, p = 0.0004), and between changes in LDL-C and change in microemboli count (Spearman r = 0.55, p = 0.0018).

Macrophage infiltration has been identified as one of the leading risk factors for plaque rupture. USPIO-enhanced MRI has been shown to be potentially able to identify macrophage-rich plaques in vivo in human carotid arteries (4), and once a vulnerable plaque is found, sequential follow-up of this is important to ensure that effective medical or surgical therapy is instituted to prevent cerebral events. Monitoring lipid therapy alone is insufficient for determining the therapeutic effect of drugs because although statins can reduce plasma cholesterol, they do not always decrease macrophage infiltration (19). To study change in plaque morphology after a therapeutic intervention, an imaging method is required that can provide accurate assessment of plaque tissue components and activity. Several imaging methods including X-ray angiography, ultrasonography, and computed tomography have been used to study the arterial wall and assess atherothrombosis, but none of these modalities can adequately characterize plaque macrophage infiltration. Recent developments in MRI have made it possible to robustly identify and measure plaque tissue composition in vivo (20- 21).

The findings presented herein represent the first prospective molecular MRI study on the in vivo effects of statin therapy on carotid plaque inflammation. We found a significant reduction from baseline in USPIO-enhanced MRI-defined plaque inflammation in the high-dose atorvastatin group at both 6 and 12 weeks. No difference was observed with the low-dose regimen. At 12 weeks, there was a significant mean signal difference between the 2 groups. Atorvastatin-induced LDL-C lowering was moderately correlated with both attenuation of plaque inflammation and a clinical correlate—a reduction in microemboli count on TCD. These findings provide in vivo evidence that high-dose statin therapy might have a beneficial effect on plaque stability.

High-resolution MRI has also previously been used for plaque burden regression assessment. In the case of statin-induced atherosclerotic plaque modulation, these effects have been consistently reported only after at least 1 year of therapy: Corti et al. (22) showed a decrease in carotid and aortic plaque size and increase in luminal area after 2 years of simvastatin monotherapy. In a follow-up study, they extended their observations to note that aggressive dosing of simvastatin compared with conventional therapy did not result in a greater change in plaque size, and concluded that changes in vessel wall parameters are more related to the LDL-C reduction rather than to the dose of statin given (23). The recently reported ORION (Outcome of Rosuvastatin Treatment on Carotid Artery Atheroma: a Magnetic Resonance Imaging Observation) trial found a significant reduction in the proportion of carotid plaque composed of lipid and a significant increase in fibrous tissue over the course of 2 years treatment with rosuvastatin (24), but again, did not observe a measurable difference as a consequence of using a higher dose of rosuvastatin. Lima et al. (25) demonstrated statin-induced plaque regression in the thoracic aorta 6 months after the onset of MRI-monitored statin therapy, although the temporal difference in the onset of documented plaque modulation could be explained by the difference in vascular bed under scrutiny.

We measured an effect after only 6 weeks with high-dose atorvastatin, and rather than targeting plaque area/volume changes, which has been shown to take months to years to detect a difference, we have looked at change in macrophage infiltration and inflammation. Our results are in keeping with previous data that intensive treatment with atorvastatin decreases the inflammatory activity of carotid plaques at 1 month (26). Furthermore, Crisby et al. (27) demonstrated that patients after 12 weeks of statin treatment had less lipid and oxidized LDL-C and fewer macrophages and T cells in their endarterectomized carotid samples compared with patients without statin therapy. In the MIRACL (Myocardial Ischemia Reduction With Aggressive Cholesterol Lowering) trial (28) and the PROVE-IT (Pravastatin or Atorvastatin Evaluation and Infection Therapy) trial (29), significant benefits, namely, decreased incidence of unstable angina, occurred within the first month of treatment. Hence, our study may support the hypothesis that dampening of plaque inflammation rather than morphological regression plays a role in the mechanism underlying the early beneficial effects of statins seen in clinical practice. The fact that we did not observe a measurable signal difference with the low-dose group may have been because most of the patients were statin-primed before enrollment, and therefore did not show a strong serological response to low-dose atorvastatin. Many of the patients in the other trials described were either statin naïve or had their lipid-lowering drugs discontinued during the lead-in period of the study and were therefore more likely to be more sensitive to low-dose statin administration than were participants in the current trial. Placebo was not considered ethical in our study.

The effects of high- or low-dose atorvastatin on the systemic inflammatory markers, except for Lp-PLA2, were not significant, which may have been due to the small sample size used. We were also surprised to find that reduction in plaque inflammation was not correlated with the increase in high-density lipoprotein cholesterol (HDL-C) in the high-dose group. Previous studies have found that increasing HDL-C with statin therapy may have a positive impact not only on the morphology (24), namely, reduction of lipid core, but also on attenuation of plaque inflammation (30). However, it is also possible that pleiotropic effects other than an HDL-C–dependent mechanism mediate the anti-inflammatory effect of atorvastatin.

Methods must be reproducible as well as accurate if they are to be useful in clinical practice. We have shown that interobserver reproducibility of USPIO signal change on high-resolution MRI as a marker of plaque inflammation was high. That stems from the method of segmenting the plaque into quadrants using pre-defined objective rules, as described in the methodology. In fact, in the sample of plaques investigated in this study, agreement was better than for the different measurements of luminal stenosis used and for identification and quantification of carotid atherosclerotic plaque components on MRI (31). An intraclass correlation coefficient of >0.8 between signal data for matched quadrant data before USPIO at 0, 6, and 12 weeks suggests that the technique is repeatable. No USPIO uptake was seen in the pre-imaging at either 6 or 12 weeks. Therefore, it is likely that USPIO particles have been cycled out of the plaque. This finding clearly has implications for future prospective sequential studies to assess inflammatory activity because evidence of residual USPIO signal at 6 or 12 weeks before infusion would limit the usefulness of the technique.

Atorvastatin therapy itself is unlikely to affect USPIO-based macrophage detection by MRI, a prerequisite for the use of this contrast material to examine lesion macrophage burden during lipid-lowering therapy. A recent study showed that nontoxic concentrations of atorvastatin did not affect the amount of USPIO taken up by macrophages in vitro (32). Furthermore, the intracellular distribution of iron oxide nanoparticles and the resulting MRI signal intensities remained unchanged by statin treatment.

Our study should be interpreted in light of certain limitations. The sample size is relatively small but was adequately powered for the primary end point. Although this and other earlier studies (4,33- 34) have demonstrated the technical feasibility of USPIO-enhanced MRI, the diagnostic accuracy and potential effect on clinical management are yet to be defined in larger-scale studies. Moreover, the disadvantages with USPIO-MRI include the fact that at least 2 MRI studies are required before and after infusion of contrast medium. In times of low budget and a demand for fast patient throughput, a test with immediate results would be desirable. Recently, ultrashort echo time subtraction imaging has shown promise, which may remove the need to perform the pre-infusion MRI study (35).

The relatively modest mean percentage reduction in a signal loss can sometimes only be appreciated with the use of dedicated quantitative algorithms and can be masked, for instance, in plaques with heavy calcification (4). New image-acquisition methods, for example, gradient echo acquisition for superparamagnetic particles with positive contrast, that allow USPIO to be detected as a strong positive contrast enhancement have been developed in vitro with promising results but are yet to be implemented in human carotid atheroma imaging (36).

Although T₂*-weighted imaging does allow the visualization of the effect of USPIO particles, there has been much debate in the literature as how best to quantify this and whether the degree of signal loss is in any way proportional to the inflammatory load within the plaque. Trivedi et al. (37) used manually delineated ROIs and calculated the normalized signal change between before and after USPIO images. The signal was normalized to the signal in the adjacent sternocleidomastoid muscle and any ROI drawn that showed an actual normalized signal drop were taken to indicate USPIO uptake; this showed only a moderate correlation with macrophages staining positively for USPIO on Perls staining (Pearson's product moment 0.6). Although useful, this technique has an inherent problem of bias and observer error, leading to some question as to its usefulness in the quantification of inflammatory burden.

A more recent approach has been to arbitrarily divide the vessel wall in each slice into quadrants by constructing perpendiculars to the horizontal axis across the image. This technique, which has been adopted in this study, has the advantage that data points come from the whole vessel and every section that has plaque, rather than the rather biased population of ROI. Thus, a quadrant showing signal loss, once normalized to adjacent muscle, is taken to mean USPIO uptake in that quadrant.

While eliminating operator bias, this technique has a number of problems associated with it. First, small focal areas of signal loss may be lost in a quadrant when regions of signal enhancement are found around it, limiting the spatial resolution of the analysis. Second, any statistical analysis needs to consider the quadrants as populations and compares 1 population with another (e.g., quadrants from symptomatic patients against quadrants from asymptomatic patients). For parametric statistics to be useful, a number of assumptions are made. A simple 2-tailed t test assumes that the quadrants are normally distributed and that each quadrant represents a totally independent observation. However, the assumption that each quadrant represents a completely independent observation is obviously overly simplistic.

One quadrant is likely to depend on another to some extent, not only within the slice but also between the slices. Data will therefore be clustered, and simple parametric statistics will therefore falter because the number of degrees of freedom for a simple t test will be related to the number of quadrants rather than to the number of patients, and therefore the overall test statistic will be calculated to be artificially high. Thus, any p value will be artificially low. In an attempt to tackle these problems, a complex repeated-measures mixed model was used to model true variances in quadrant data.

Another study limitation is that, by design, we included patients who were not symptomatic from their carotid disease, and it might therefore be argued that our population did not cover the whole spectrum of disease severity, bringing into question the generalizability of our results. However, patients included in this study are those for whom the benefit of carotid endarterectomy is still controversial and for whom much is to be expected from noninvasive imaging techniques and intervention.

There is an increasing desire to formulate surrogate end points for clinical trials, with biomarkers and imaging being possible candidates for assessing underlying disease. Such surrogates might allow more timely assessment of the efficacy or futility of a proposed intervention. The problem that pharmaceutical companies and regulatory authorities face is that, to demonstrate a clinical benefit using accepted outcome measures such as stroke or death, clinical studies need to recruit thousands of patients and rely on them undergoing such an event (38). Imaging of arterial disease that can detect changes in vessel structure or function could hasten the development process on smaller number of patients and allow intervention at an earlier stage in the disease process. There is increasing evidence from the coronary literature that imaging can be used to monitor the effects of interventions such as statin therapy. Measures of carotid intima-media thickening using ultrasonography have been validated and have been approved by the FDA to be used in clinical trials. The literature concerning the assessment of interventions using MRI is somewhat less mature.

This is, to our knowledge, the first study of a targeted MRI contrast agent used in humans to assess therapeutic response in an interventional drug trial. Simultaneously, it facilitated the enrichment of a trial population so that only subjects with plaque inflammation were selected for enrollment into the study. Aggressive lipid-lowering therapy over a 12-week period is associated with a significant reduction in USPIO-defined inflammation. As validation continues, USPIO-enhanced MRI methodology may be a useful imaging biomarker for the screening and assessment of therapeutic response to anti-inflammatory interventions in patients with carotid atherosclerotic lesions. Serial studies involving a larger population of patients are needed to determine whether attenuation of plaque inflammation as defined by USPIO-enhanced MRI can be correlated with a reduced risk of developing clinical symptoms.

For a complete list of acknowledgments and a diagram of the atheroma trial consort, please see the online version of this article.

Atorvastatin Therapy: Effects on Reduction of Macrophage Activity (ATHEROMA): Evaluation Using Ultrasmall Superparamagnetic Iron Oxide-Enhanced Magnetic Resonance Imaging in Carotid Disease