Background
Screening tools for the detection of coronary artery disease (CAD) are of considerable interest in light of skyrocketing risk factors. Recent work suggests that carotid plaque has a relatively unexplored role in CAD risk prediction but has previously been limited by the difficulty in quantifying its irregular architecture using two-dimensional (2D) ultrasound. The aim of this study was to investigate the utility of a novel automated three-dimensional (3D) ultrasound-based carotid plaque volume quantification technique as a negative predictor of CAD.
Methods
In this prospective study, 70 consecutive patients referred for coronary angiography underwent same-day 2D and 3D carotid ultrasound scans for the purpose of plaque quantification in the carotid bulbs. Two-dimensional plaque thickness was measured in its maximal value perpendicular to the vessel wall. Total 3D plaque volume was quantified using a stacked-contour method. Luminal narrowing of coronary arteries was analyzed using the established 16-segment model for coronary arteries to produce an overall angiographic score. Receiver operating characteristic curves, negative predictive value, and sensitivity of 2D and 3D plaque quantification relative to coronary angiography were determined.
Results
The novel 3D carotid ultrasound method resulted in a higher negative predictive value and sensitivity relative to 2D carotid ultrasound at their optimal thresholds as determined by Youden indices of receiver operating characteristic curves. In particular, total 3D plaque volumes less than the threshold of 0.09 mL accurately predicted the absence of significant CAD in 93.3% of patients (98.0% sensitivity), whereas maximal 2D plaque thickness less than the threshold of 1.35 mm provided significantly lower negative predictability at 75% (93.9% sensitivity).
Conclusions
Using the determined threshold of 0.09 mL for plaque volumes, this feasibility study suggests that automated 3D ultrasound-based carotid plaque quantification may serve as an important clinical screening tool to help identify patients who are at low risk for significant CAD.
Coronary artery disease (CAD) is a leading cause of death in North America, and developing methods for early detection is critical to reducing the burden of this disease. Currently, coronary angiography is the “gold standard” for diagnosis, but the relatively high cost and inherent risks associated with this invasive procedure make it difficult to use in routine screening for CAD or monitoring disease progression. A safe and noninvasive imaging technique that allows for quick and effective screening during routine cardiology referrals is necessary to improve the clinical evaluation of this disease in both asymptomatic and symptomatic individuals. Numerous studies have examined the role of carotid intima-media thickness (CIMT) measurements via two-dimensional (2D) ultrasound as a predictability index for CAD risk stratification. However, there is a paucity of data examining the role of CIMT for immediate correlation to the extent of atherosclerosis by angiography. Several recent studies, including the Atherosclerosis Risk in Communities trial, have suggested that the measurement of carotid plaque is also an important predictor of cardiac risk. A recent small study conducted in a Japanese-only population demonstrated the strong association of 2D carotid plaque thickness with coronary artery lesions assessed by angiography. Furthermore, a recent meta-analysis also confirmed that carotid plaque measurement is superior to CIMT alone for predicting significant CAD. These studies highlight the potential of carotid plaque quantification to significantly enhance risk prediction and selection of patients for angiography.
Because arterial vessels are three-dimensional (3D) structures, plaque quantification in two planes can often lead to the omission of plaque that is outside of the image plane. The plaque itself can be irregular in shape and size ( Figure 1 ). These factors limit our ability to quantify plaque and investigate its role as a predictor of CAD. This is the first study to investigate the clinical utility of carotid plaque volumes quantified using a novel automated 3D ultrasound acquisition technique to rule out significant CAD as determined by immediate angiography.
Methods
In this prospective study conducted over 1 month, 70 consecutive patients were recruited and underwent both angiography and carotid ultrasound on the same day. The indications in patient consideration for coronary angiography included angina, abnormal findings on resting or exercise electrocardiography or nuclear tests, and/or positive results on stress echocardiography. At the attending physician’s discretion, these tests were performed in a number of patients presenting with chest pain or significant risk factors, such as family history of CAD, hypertension, hyperlipidemia, and long-term tobacco use ( Table 1 ). All procedures were conducted and interpreted by investigators blinded to other procedural results. All angiographic studies were conducted for clinically indicated purposes, and angiographic scores were used as gold-standard diagnostic values for the presence of significant CAD.
| Variable | Men ( n = 52) | Women ( n = 18) | Total ( n = 70) |
|---|---|---|---|
| Age (y) | 65.5 ± 11.6 | 68.4 ± 10.2 | 66.2 ± 11.4 |
| Race (% Caucasian) | 100.0% | 100.0% | 100.0% |
| Body mass index (kg/m 2 ) | 29.4 ± 5.3 | 27.6 ± 5.2 | 29 ± 5.3 |
| Mean angiographic score (range, 0–4) | 2.2 ± 1 | 1.5 ± 1.2 | 2 ± 1.1 |
| CAD risk factors | |||
| Diabetes | 23.5% | 27.8% | 24.3% |
| Hypertension | 64.7% | 77.8% | 67.1% |
| Hyperlipidemia | 68.6% | 61.1% | 65.7% |
| Current tobacco use | 25.5% | 11.1% | 21.4% |
| Former tobacco use | 60.8% | 22.2% | 50.0% |
| Family history of CAD | 88.2% | 88.9% | 87.1% |
| CAD indications | |||
| Angina | 58.8% | 72.2% | 61.4% |
| Abnormal resting ECG results | 53.8% | 38.9% | 50.0% |
| Abnormal exercise ECG results | 36.5% | 27.8% | 34.3% |
| Positive stress echocardiographic results | 3.9% | 22.2% | 8.6% |
| Positive nuclear test results | 5.8% | 16.7% | 8.6% |
| History | |||
| Previously diagnosed CAD | 35.3% | 16.7% | 30.0% |
| Past intervention | 17.6% | 11.1% | 15.7% |
Patients meeting the following inclusion criteria were eligible for study: (1) men or women aged > 18 years, (2) outpatients referred for clinically indicated angiography for assessment of CAD, (3) absence of clinical contraindications to angiography or carotid ultrasound, and (4) ability and willingness to give informed written consent.
Coronary angiography was performed according to the standard Judkins method using a GE System 2000 (GE Healthcare, Milwaukee, WI) by one of four experienced interventional cardiologists. Luminal narrowing of coronary arteries was analyzed using a 16-segment model for the coronary arteries to produce an overall angiographic score, as defined previously. After qualitative scoring of each segment, the entire angiogram was graded as follows: 0 = no or minimal disease (0%–19% narrowing in any segment), 1 = mild disease (20%–49% narrowing in any segment), 2 = moderate disease (luminal narrowing of at least one segment of 50%–69%), and 3 = severe (≥70% narrowing within any segment of the main branches of the coronary artery or ≥50% in the left main coronary artery). Patients were stratified as having insignificant CAD (angiographic score of 0 or 1) or significant CAD (angiographic score of 2 or 3). Quantitative coronary angiographic software was available with this system and left for use at the discretion of the interpreting interventional cardiologist.
Two and three-dimensional carotid ultrasound scans were conducted using a dedicated vascular ultrasonography device (iU22; Philips Medical Systems, Markham, Ontario, Canada) equipped with a VL13-5 mechanical volume transducer for 3D imaging and an L9-3 transducer for 2D imaging. The 3D full-volume acquisition was automated and did not require a manual sweep by the sonographers, eliminating the variability in time associated with manual acquisition methods. All images were stored in Digital Imaging and Communications in Medicine format. Carotid scans were conducted by a single registered diagnostic cardiac sonographer using standardized practice trained using a Meijer’s arc. Imaging angles between 30° and 60° (right carotid) or 300° and 330° (left carotid) were generally used during 2D acquisition, but optimal image acquisition angles varied among patients. Both 2D long-axis and short-axis views and 3D full volumes of the distal common carotid artery, bifurcation, internal carotid artery, and external carotid artery were acquired for each patient. Scans were conducted immediately after angiography during normal postprocedural recovery. A full carotid ultrasound protocol as described by the American Society of Echocardiography consensus statement was carried out along with additional 3D acquisition.
During postprocessing, maximal plaque thickness was measured in the long-axis or short-axis views of the carotid bulb using a manual caliper placed along the outer border of the intima and extending at right angles up to the maximal thickness of the intima or protuberance of plaque into the vessel at a 90° angle from the intima. Embedded plaque was defined as an inward-directed or outward-directed area of thickening (along the vertical plane of long-axis views) within the intima of the vessel wall, which at the least had clearly demarcated proximal and distal ends (along the horizontal plane of long-axis views). To standardize plaque quantification in both two and three dimensions, the portions of intima above and below an area of plaque were included in all measurements because in most cases, it was difficult to differentiate the vertical borders of the intima from that of embedded or protuberant plaque. Therefore, in addition to protuberant and echo-dense structures along the vessel wall, our definition of plaque included embedded plaque and abnormal thickening of the intima where there was a clear start and finish to the thickening in the long-axis horizontal plane. Moreover, rather than using a discrete cutoff measurement (such as >1.5 mm, which has been used in other studies ), we considered 2D intima thickening >50% of the mean intima thickness in the distal common carotid artery as plaque. This approach allowed us to consider the Glagov phenomenon and include the measurement of plaque growing into the artery wall. Plaque thicknesses were determined in both the left and right common carotid bifurcations (bulb) for each patient, and only the maximum value was considered. The maximal plaque thickness was categorized with respect to coronary angiographic scores and the absence or presence of significant CAD.
Three-dimensional plaque volume was quantified using a stacked-contour method modified from a 3D cardiac protocol as previously reported for other cardiac structures ( Figure 2 ). Full volumes of both the left and right carotid bulbs were obtained. Adjustments along the x, y, and z axes allowed visualization of the entire volume of the carotid bulb in all possible planes. This allowed for the acquisition of carotid bulb plaque volumes regardless of the plaque size and degree of irregularity. Quantification of plaque embedded within the intima was also performed using the 3D stacked-contour method, allowing the reader to outline and quantify areas of irregular echodensities or areas exhibiting thickening with well-defined borders. Similar to 2D quantification, it was often challenging to isolate embedded plaque from the overlying or underlying intima. Thus, these portions of the intima were included in all embedded plaque volume measurements to standardize the various types of plaque under a single quantification protocol. In contrast to the 2D images, 3D carotid bulb full volumes more clearly permitted visual distinction between intima thickening and plaque formation within or on the surface of the intima. Three-dimensional plaque volumes were quantified using the stacked-contour method (QLAB GI 3DQ Plug-in; Philips Medical Systems). Briefly, a reference axis was first placed along the maximal length of the identified plaque. Along this reference axis, contours were manually drawn up to 1 mm apart, capturing the irregularity, shape, and size of the plaque along the length of the entire reference axis, generating a “true” plaque volume ( Figure 1 ). Algorithm-generated total plaque volumes were quantifiable to the lower limit of 10 μL (per the software limitations of the device; iU22 user reference manual, Philips Medical Systems). The sum of individual plaque in each carotid bulb was combined to give total plaque volume and compared with the coronary angiographic score in each patient.
In collaboration with our Medical Informatics Laboratory, a small substudy was conducted to examine the measurement accuracy and precision of the 3D stacked-contour method by comparing the measurements from three blinded readers each taking multiple diameter (four each) and volume (three each) measurements from a phantom of a carotid plaque. The phantom was specially designed for this study and composed of a bead with known volume and diameter placed within a carotid artery mimic and imaged by the 3D transducer. This plaque mimic was composed of a high-strength neoprene multipurpose rubber ball (1241T1; McMaster-Carr, Robbinsville, NJ) glued to the inside of a latex tube. Adhesion was ensured using a small amount of silicone glue. The ball and silicone glue densities were different so that the plaque mimic contour was clearly visible with ultrasound. The latex tubing diameter of 6.35 mm and wall thickness of 0.79 mm were considered suitable for a carotid artery mimic given that average carotid artery luminal diameters range from 4.3 to 7.7 mm and normal carotid wall thickness on average may range from 0.4 to 0.8 mm. The true diameter of the rubber ball used as the plaque mimic was known to be 0.50 cm, with a calculated volume of 0.065 mL. Plastic tissue-mimicking material, mixed with Sigmacell Cellulose (Sigma Aldrich, St. Louis, MO), was poured to allow approximately 0.5 cm of tissue between the probe and the “carotid artery” to mimic neck anatomy. The carotid artery phantom was filled with water but was not meant for flow physiology.
Statistical Analysis
Two investigators independently measured plaque volumes to obtain interrater reliability from a subset of patients. Interrater reliability was determined using the Shrout-Fleiss reliability test for a random set of 10 patients and the two independent raters. Additionally, a small reliability study for the angiographic score was performed with two independent, blinded raters each rating the same 10 cases to assess the interrater reliability of 3D carotid plaque quantification. Agreement was summarized by the weighted Fleiss-Cohen κ statistic.
Pearson’s correlation coefficient was determined to evaluate the relationship between maximum plaque thickness, total plaque volume, and the amount of disease quantified by angiography. T -test analysis was used to compare the mean maximal plaque thickness (2D) and the mean total plaque volume (3D) between patient groups negative for significant CAD and groups positive for significant CAD. Threshold values for plaque thickness and plaque volume were determined from receiver operating characteristic (ROC) curves and weighted Youden indices. Contingency tables (2 × 2) were used to determine the positive predictability, negative predictability, sensitivity, and specificity of each carotid bulb plaque quantification method (2D maximum thickness and 3D total volumes) relative to coronary angiography in assessing patient risk for CAD.
To assess the accuracy of the 3D carotid plaque volume quantification software, the root mean square error of the diameter and volume of the carotid plaque mimic were calculated respectively by three independent readers. To assess measurement precision, the average value for each reader was determined and compared using analysis of variance to compare mean values among readers.
Results
In the 70 patients recruited for this study, the indications for angiography were chest pain (61.4%), abnormal results on resting electrocardiography (50.0%), abnormal results on exercise electrocardiography (34.3%), and positive results on stress tests or nuclear tests (8.6%). Approximately 24.3% of these patients had diabetes, 67.1% had hypertension, and 65.7% had hyperlipidemia ( Table 1 ). No patients had previously known carotid disease or stroke. The majority of patients (52 of 70) were men. The mean age for the total patient population was 66.2 ± 11.4 years. Two-dimensional, 3D, and coronary angiographic data were available for all 70 patients. The mean combined angiographic score for men and women after catheterization was 2.0 ± 1.1, indicating that the majority of patients presented with moderate to severe CAD. In particular, 49 patients were determined to have significant CAD using coronary angiography as the gold standard.
Maximal Plaque Thickness by 2D Ultrasound
Two-dimensional ultrasound was performed on both the left and right carotid bulbs, from which the maximal plaque thickness value between the two arteries was used in the contingency table analysis ( Table 2 ). Patients with coronary angiographic scores of 0, 1, 2, and 3 had mean plaque (maximal) thicknesses of 1.53 ± 0.66, 2.18 ± 1.07, 2.29 ± 1.13, and 2.80 ± 0.87 mm, respectively ( Table 3 , Figure 3 ). The mean plaque thickness for patients in the absence of significant CAD (angiographic scores of 0 and 1) was 1.87 ± 0.94 mm. In contrast, in patients with significant CAD (angiographic scores of 2 and 3), the mean plaque thickness was determined to be significantly higher at 2.64 ± 0.98 mm ( P < .005).
| 2D maximal plaque thickness (mm) | Presence of significant CAD (angiographic scores of 2 and 3) | Absence of significant CAD (angiographic scores of 0 and 1) | Total |
|---|---|---|---|
| ≥1.35 | 46 | 12 | 58 |
| <1.35 | 3 | 9 | 12 |
| Total | 49 | 21 | 70 |
| Variable | Coronary angiographic score | |||
|---|---|---|---|---|
| 0 ( n = 10) | 1 ( n = 11) | 2 ( n = 15) | 3 ( n = 34) | |
| Mean plaque thickness (mm) | 1.53 | 2.18 | 2.29 | 2.80 |
| Range (mm) | 0.74–2.55 | 0.77–4.35 | 1.18–4.85 | 1.04–4.67 |
| Variability (SD) | 0.66 | 1.07 | 1.13 | 0.87 |
| Mean plaque volume (mL) | 0.07 | 0.13 | 0.17 | 0.25 |
| Range (mL) | 0.02–0.16 | 0.02–0.43 | 0.08–0.41 | 0.1–0.51 |
| Variability (SD) | 0.05 | 0.13 | 0.10 | 0.11 |
Carotid Bulb Plaque Volume by 3D Ultrasound
Similar to 2D carotid ultrasound, 3D ultrasound was performed on both the left and right carotid bulbs, from which the total (cumulative) plaque volume was calculated and used in contingency table analysis ( Table 4 ). The 3D full-volume acquisition was automated and did not require a “sweep” of the carotid bulb by the technologist. Patients with coronary angiographic scores of 0, 1, 2, and 3 had mean plaque volumes in the carotid bulb of 0.07 ± 0.05, 0.13 ± 0.13, 0.17 ± 0.10, and 0.25 ± 0.11 mL, respectively ( Table 3 , Figure 4 ). The mean plaque volume for patients without significant CAD (angiographic scores of 0 and 1) was 0.10 ± 0.10 mL, whereas in patients with significant CAD (angiographic scores of 2 and 3), the mean plaque volume was significantly greater at 0.23 ± 0.11 mL ( P < .001).
