Training of nonsonographer physicians or staff members is needed to implement carotid intima-media thickness (CIMT) and plaque screening by ultrasound for the assessment of subclinical atherosclerosis. The purpose of this study was to determine the effect of formal training on CIMT assessment and plaque detection by medical residents.
A medical resident (R1) was trained using an abbreviated American Society of Echocardiography CIMT protocol. CIMT and plaque assessment by R1 were compared against an expert scanner on 60 subjects using a portable US system. A second medical resident (R2) was then trained on the CIMT protocol focusing on plaque visualization after the results of the first phase of the study were analyzed, and the results were compared against an expert on an additional 10 subjects.
In the first phase of the study, a total of 106 images (94% interpretable) were available for CIMT and plaque assessment by both R1 and the expert. CIMT measurements were bioequivalent within the limits of ultrasound resolution, with 88% agreement. Variability on plaque presence was high, with only 53% agreement. R2 and the expert each scanned 10 new subjects twice, from whom 40 images were available for interpretation. R2 demonstrated CIMT agreement (93%) comparable with that observed in phase 1 but with greatly improved plaque agreement (100%). Intraobserver variability during phase 2 for both R2 and the expert was extremely low.
Medical residents can undergo rapid training for CIMT measurement and plaque visualization to detect subclinical atherosclerosis compared with an expert.
Increased carotid intima-media thickness (CIMT) is a measure of subclinical atherosclerosis and has been associated with an increased risk for the development of coronary artery disease and the occurrence of cerebrovascular accidents in asymptomatic patients of differing ages, gender, and ethnicities. The assessment of carotid plaque and CIMT with high-resolution brightness-mode (B-mode) ultrasound requires great precision. Consequently, this technique has largely been reserved for the research setting, where it has been shown to be highly sensitive and reproducible when performed in the hands of highly trained sonographers using comprehensive, time-intensive scanning protocols with offline measurements made by core laboratories. On the other hand, recent guidelines have given CIMT assessment a class IIa indication in subjects with intermediate Framingham risk scores. Although labs using expert sonographers should remain the gold standard for CIMT measurement, transition of CIMT assessment from the research lab to the office setting is required for more widespread clinical availability of this test, particularly screening for atherosclerosis in a large population.
The development of an American Society of Echocardiography (ASE) consensus document with a simplification of the CIMT protocol, the improvement in ultrasound technology and resolution, dedicated laptop-sized, portable, and user-friendly ultrasound systems with touch-screen function familiar to the current “digital generation,” and the automation of CIMT measurement in current ultrasound systems suggest that it may be feasible to train persons inexperienced in ultrasound to perform CIMT measurement. A recent study, largely focused on image interpretation, demonstrated low interobserver variability in the offline measurement of CIMT and carotid plaque identification between a core lab and newly trained clinicians. That study served as a first attempt at formulating an expert-guided CIMT training protocol comprising didactic as well as hands-on training with mock subjects. CIMT assessment technology now allows online automated CIMT measurement, suggesting that the demonstration of decreased variability in image acquisition and plaque assessment between expert and inexpert scanners needs to be demonstrated for clinical integration of this test.
The purpose of this prospective study was to develop a robust training protocol in intima-media thickness (IMT) and plaque image acquisition for a medical resident (R1) and then to test this protocol prospectively by comparing IMT values and plaque presence from images obtained by R1 with those of an expert scanner and to develop refinement in training protocol prospectively after reviewing initial results.
The institutional review board of the University of Southern California approved the study protocol. All study participants provided written informed consent. A single portable ultrasound system dedicated to CIMT and plaque analysis, CardioHealth Station (CHS; CardioNexus Corporation, Houston, TX) was used. Training was based on the scanning protocol for imaging acquisition recommended by the ASE.
The study consisted of two prospective phases. The first phase included 60 subjects without history of cardiovascular disease (CVD) who were admitted for noncardiac reasons from August 2011 to November 2011 at a single tertiary medical center. All subjects underwent carotid imaging using CHS for CIMT measurement and carotid plaque assessment at a single encounter by both an expert (T.Z.N.) and R1 (K.D.). R1 initially underwent a training protocol developed by the expert for CIMT and plaque image acquisition and interpretation ( Table 1 ). Interobserver variability was determined and used to modify the original training protocol for the second medical resident (R2). A new modified training protocol ( Table 1 ), with an increased focus on plaque assessment, was implemented during a second phase in which both the expert and R2 (T.O.) scanned 10 new subjects (meeting the same clinical inclusion criteria as above). Each of these 10 subjects from phase 2 was scanned twice by both R2 and expert (1–24 hours apart). Interobserver and intraobserver CIMT and plaque variability using the modified training protocol was evaluated. Of note, both R1 and R2 used the same five mock subjects during training (phases 1 and 2) to establish baseline post-training interobserver variability between trainees. The expert scanner was a cardiologist and a core lab–trained IMT scanner with >8 years of experience in performing CIMT measurement and plaque assessment for randomized multicenter CIMT studies as well for clinical practice–implemented CIMT studies. Both R1 and R2 were third-year internal medicine residents without any prior experience in ultrasound imaging, CIMT imaging, or reading. The order of scanning between expert and R1 and R2 scanners was changed after every subject.
|IMT: anatomy, pathophysiology, literature review||1 h||Lecture with slides, including 15 min Q-and-A session|
|IMT acquisition protocol (ASE)||1 h||Lecture with slides, including 15 min Q-and-A session|
|Ultrasound principles, image interpretation, and measurement||1 h||Lecture with slides, including 15 min Q-and-A session|
|Practical training: equipment demonstration|
|Ultrasound system features||1 h||Touch-screen functions, data entry, selection of database for reference values, image save function, image transfer, and report generation|
|Ultrasound system knobology||0.5 h||Ultrasound transducer and frequencies, 2D, color and power Doppler, spectral Doppler, depth and gain settings, image optimization|
|Day 2||Practical training: image acquisition|
|Demonstration of IMT imaging protocol by expert: 1 mock subject||1 h||Subject positioning and neck angle, machine positioning for right and left carotid scans, sonographer positioning, transducer orientation, holding the transducer, transducer manipulation, scanner arm positioning|
|Supervised practice sessions of resident with the expert: 1 mock subject||1 h||Right and left sides, short-axis sweep, identification of plaques in short axis, confirmation of plaque in the long-axis views, CIMT assessment in anterior, lateral, and posterior angles; transducer maneuvers: heel-toe, transducer rotation, gliding, medial and lateral tilt|
|Day 3||Supervised practice sessions of resident with the expert: 2 mock subjects||1 h||Right and left sides, short-axis sweep, identification of plaques in short axis, confirmation of plaque in the long-axis views, CIMT assessment in anterior, lateral, or posterior angles; transducer maneuvers: heel-toe, transducer rotation, transducer pressure, use of extra gel for anterior angles, gliding, medial and lateral tilt, swallowing by patient, neck hyperextension|
|Resident practice scans|
|Day 4||Unsupervised practice scans: 2 mock subjects||1–2 h||Bilateral plaque assessment and CIMT measurement|
|Review of images with the expert and feedback||0.5 h||Demonstration of CCA, bulb, and ICA in the short-axis sweep, speed of sweep (phase 2), demonstration of plaque or lack thereof, demonstration of plaque or lack thereof in 3 long-axis views (phase 2), image depth, image optimization, focus, horizontal alignment, visualization of near and far wall IMT|
|Day 5||Unsupervised practice scans: 3 mock subjects||2.5 h||Bilateral plaque assessment and CIMT measurement|
|Review of images with the expert and feedback||1 h||Demonstration of CCA, bulb, and ICA in the short-axis sweep, speed of sweep (phase 2), demonstration of plaque or lack thereof, demonstration of plaque or lack thereof in 3 long-axis views (phase 2), image depth, image optimization, focus, horizontal alignment, visualization of near and far wall IMT|
|Total training time||13 h|
The training protocol is shown in Table 1 . Resident training included didactic and practical training sessions. Didactic sessions included online CIMT lectures, distribution and review of the ASE consensus article on CIMT, ultrasound principles, and image interpretation. Practical training sessions started with training residents on the features of the CHS system, followed by supervised and then unsupervised practical training sessions with the expert, as shown in Table 1 . Phase 2 training included the addition of dedicated time (5–15 min per subject) toward practicing short-axis and longitudinal circumferential sweep techniques for plaque assessment. This included a more focused and slower speed of carotid short-axis sweep, taking time to demonstrate carotid bulb and proximal internal carotid artery as well as the teaching of techniques to better differentiate plaque and artifacts (including the use of power Doppler), transducer manipulation including adjustment of probe pressure, probe angle, patient maneuvers, and demonstration of plaques in both short and one or more long axes. The distal 1 to 2 cm of common carotid artery (CCA) alone was imaged for CIMT because this appears to be representative of all segments, and all CCA, bulb, and internal carotid artery segments were imaged for plaque.
Inclusion of subjects into each phase of the study began after the expert scanner confirmed that each medical resident was capable of acquiring frozen images with adequate visual quality for CIMT measurement and plaque assessment.
Carotid Artery Scanning Technique
Scans were performed using the CHS system with a broadband 13-5V1/13-5 MHz linear-array transducer with a center frequency of 9 MHz, using B-mode display. This system displays raw radiofrequency data rather than digitally processed data displayed by the conventional ultrasound systems and has a theoretical resolution of approximately 30 μm. In phase 1 of the study, only a single-angle CIMT measurement was available on the CHS, and averaged three-angle measurement was available in phase 2. We used the single-angle mean CIMT measurement that displayed the thickest CIMT throughout phases 1 and 2 of the study. The CHS system determines systole and diastole on the basis of vessel luminal diameter allowing CIMT measurement at end-diastole without requiring external electrocardiography. The transducer was moved to align the flow divider with an on-screen vertical marker to ensure that the region of interest was automatically located at the far wall of the CCA (1 cm from the flow divider; Figure 1 ). Within the 1-cm region of interest, the system continuously tracks 24 spatial measurements at 200 frames/sec. Only the IMT data that are present at end-diastole are kept, and all nondiastolic frame IMT data are discarded, and a frozen image of tracked CIMT with measurements is then displayed. The CHS system incorporates a sensor in the transducer that tracks the scanning angle in real time and records the transducer angle on the screen along with the IMT measurement. Online manual CIMT border correction by manual tracing was also available in the setting of poor automatic border detection and was used when needed. Current ASE guidelines were followed. Head positioning was approximated to 45° to the right or left. The bilateral CCA was imaged in each patient both by the residents and the expert. For R1, plaque assessment was performed only in the live short-axis views acquired from the base of the neck to the highest segment that could be imaged above the carotid bifurcation, and long-axis assessment of plaque was performed only if plaque was seen in the short-axis view. For R2, plaque assessment was performed both in the short-axis view as well as in the long-axis sweep from the anterior to the posterior portion of the neck in all patients. CCA far wall mean CIMT was obtained by a circumferential (anterior to posterior) longitudinal at a single angle sweep along the CCA until the best horizontal single image with clear visualization of far wall lumen-intima and media-adventitia borders was obtained. A carotid plaque was defined as a focal intraluminal protrusion >50% of the surrounding IMT and with total thickness of ≥1.5 mm. A final report indicating mean right and left CCA far wall CIMT values and the presence and location of plaque, along with still images, was immediately available at the conclusion of each scan. The total time needed for manual scan completion was recorded for both the expert and novice scanners for all subjects.
Statistical analysis was performed by a single blinded observer using Excel (Microsoft Corporation, Redmond, WA). Descriptive data are presented as mean ± SD, along with the coefficient of variation (CV). CIMT variability (intraobserver and interobserver) was analyzed using the “two one-sided t test” (TOST; 0.110-mm limits) and Bland-Altman analysis to determine both equivalency and level of agreement, respectively. Logistical regression analysis was performed to assess correlation (Pearson’s correlation coefficient). For noncontinuous data, specifically plaque comparisons, χ 2 values were derived using SAS (SAS Institute Inc., Cary, NC). P values < .05 were considered statistically significant for all analyses. Although the CHS system has a theoretical axial resolution of approximately 30 μm (0.030 mm), we used the technological limits of standard ultrasound system resolution, approximating 0.1 mm when setting P TOST limits (0.110 mm) as well as expected “good” CVs (average cohort CIMT, 0.550–0.650 mm; standard deviation, 0.500–0.100 mm by ultrasound limits; expected CV ≤ 15%–20%).
Among the 60 subjects originally enrolled in phase 1, seven were excluded from final analysis because of the presence of initially unreported end-stage renal disease ( n = 3) or coronary artery disease ( n = 1), accidental capture of the jugular vein by R1 ( n = 1), an improperly saved image by R1 ( n = 1), and a single study with overall difficult image acquisition for both R1 and the expert ( n = 1), for a final phase 1 cohort of 53 subjects (106 vessels). From these 53 subjects, four more carotid vessels (right and left) were found to have nondiagnostic image quality by R1 for mean CIMT measurement analysis. The corresponding four diagnostic contra lateral vessels were also excluded for analysis purposes. One patient was noted to have concentrically thickened CIMT (>2.5 mm) and was included only for plaque analysis. The final phase 1 cohort included 96 vessels from 53 subjects for CIMT comparison and 98 vessels for plaque assessment. The frequency of nondiagnostic image acquisition was found to be <1% (one of 120) for the expert and <6% (seven of 120) for R1. The final phase 2 cohort included 10 separate subjects (each scanned twice by both the expert and R2) with 40 vessels for CIMT comparison and 20 vessels for plaque assessment. There were no subject or vessels excluded from phase 2. The expert and R1 or R2 each performed their scans first approximately 50% of the time within each respective phase cohort.
Clinical characteristics are summarized in Table 2 . Our subject cohort was at low CVD risk per Framingham risk score (mean, 6.5%). The average age of our final full cohort (total n = 63) was 45 ± 13 years (range, 25–65 years). There were 42 men, 45 Hispanics (71%), five African-Americans, six Asians, and seven Caucasians. Approximately one third each had histories of hypertension (29%) and diabetes mellitus (35%). A small number had significant or active histories of smoking (14%). The mean Framingham risk score in those with plaque versus without plaque (by the expert scans) was 9.04 ± 7.22 versus 5.11 ± 5.24 ( P < .01).
|Total population||63 (100%)|
|Total population age (y)||45 ± 13|
|Male age (y)||45 ± 11|
|Female age (y)||45 ± 15|
|African American||5 (8%)|
|Diabetes mellitus||22 (35%)|
|Smoking history||9 (14%)|
|Total cholesterol (mg/dL)||159 ± 60|
|Low-density lipoprotein cholesterol (mg/dL)||96 ± 36|
|High-density lipoprotein cholesterol (mg/dL)||43 ± 20|
|Glycosylated hemoglobin (%)||7.0 ± 2.7|
|Framingham risk score (%)||6.5 ± 5.9|