Ultrasound protocols to measure carotid intima-media thickness in trials; comparison of reproducibility, rate of progression, and effect of intervention in subjects with familial hypercholesterolemia and subjects with mixed dyslipidemia

Abstract

Background. Current ultrasound protocols to measure carotid intima-media thickness (CIMT) in trials rather differ. The ideal protocol combines high reproducibility with a high precision in the measurement of the rate of change in CIMT over time and with a precise estimate of a treatment effect. To study these aspects, a post-hoc analysis was performed using data from two randomized double-blind, placebo-controlled trials: one among 872 subjects with familial hypercholesterolemia (FH) and the other among 752 subjects with mixed dyslipidemia (MD), respectively. Participants were randomized to torcetrapib or placebo on top of optimal atorvastatin therapy.

Methods. CIMT information was collected from the left and right carotid artery from two walls (the near and far wall) of three segments (common carotid, bifurcation, and internal carotid artery) at four different angles (right: 90, 120, 150, and 180 degrees on Meijer's carotid arc; left: 270, 240, 210, and 180 degrees, respectively). Based on combinations of these measurements, 60 different protocols were constructed to estimate a CIMT measure per participant (20 protocols for mean common CIMT, 40 protocols for mean maximum CIMT). For each protocol we assessed reproducibility (intra-class correlation coefficient (ICC), mean difference of duplicate base-line scans); 2-year progression rate in the atorvastatin group with its standard error (SE); and treatment effect (difference in rate of change in CIMT between torcetrapib and placebo) with its SE.

Results. Reproducibility: ICC ranged from 0.77 to 0.91 among FH patients and from 0.68 to 0.86 among MD patients. CIMT progression rates ranged from −0.0030 to 0.0020 mm/year in the FH trial and from 0.00084 to 0.01057 mm/year in the MD trial, with SE ranging from 0.00054 to 0.00162 and from 0.00083 to 0.00229, respectively. The difference in CIMT progression rate between treatment arms ranged from −0.00133 to 0.00400 mm/year in the FH trial and from −0.00231 to 0.00486 mm/year in the MD trial. The protocol with the highest reproducibility, highest CIMT progression/precision ratio, and the highest treatment effect/precision ratio were those measuring mean common CIMT with measurements of the near and far wall at multiple angles. When the interest is in the mean maximum CIMT, protocols using multiple segments and angles performed the best.

Conclusion. Our findings support the position that the number and specific combination of segments, angles, and walls interrogated are associated with differences in reproducibility, magnitude, and precision of progression of CIMT over time, and treatment effect. The best protocols were mean common CIMT protocols in which both the near and far walls are measured at multiple angles.

Trial registration: ClinicalTrials.gov identifier: NCT00136981.

Trial registration: ClinicalTrials.gov identifier: NCT00134264.

Key words::

• Carotid intima-media thickness
• epidemiology
• statins
• trial
• ultrasound

Key messages

• Our findings support the opinion that the number and specific combination of segments, angles, and walls interrogated are associated with differences in reproducibility, magnitude and precision of progression of carotid intima-media thickness (CIMT) over time.

• The choice for an ultrasound protocol to mea sure CIMT should depend on a well consi dered balance of differences in reproducibility and magnitude and precision of progression of CIMT over time before the study is carried out.

• Not all CIMT measurements are the same; neither do they reflect the same process.

Introduction

Carotid intima-media thickness (CIMT) is a non-invasive marker of atherosclerotic vascular disease that has been studied extensively since 1986 (1). CIMT is widely used in observational and intervention studies to assess determinants of atherosclerosis and its consequences for cardiovascular events (2–23). In addition, change in CIMT over time is currently used as an alternative for cardiovascular events as primary outcome in intervention studies (24–26).

Measurements can be made of the near and/or the far wall along the common carotid artery (CCA), the carotid bifurcation (BIF), and the internal carotid artery (ICA) at different angles of insonation. At present there is quite some diversity in the choice of ultrasound protocols used to measure CIMT. The most commonly used ultrasound protocols allow CIMT measurements to be taken from different combinations of segments (CCA; CCA and BIF; CCA, BIF, and ICA), walls (only far wall; both far and near wall), and angles (single angle, or a combination of angles). Moreover, with these parameters at least two different CIMT outcome measures can be estimated: the mean common CIMT and the mean maximum CIMT. The mean common CIMT is estimated as the mean value of the mean CIMT measurements that are performed over a 10 mm part of the far wall or both the far and near wall of the common carotid artery segment. The mean maximum CIMT is a summary measure that is computed as the mean of the single maximum CIMT measurements that are measured in 6 to 12 standard carotid artery walls: the far wall or both the far and near wall of the two or three distinct carotid artery segments: the common carotid segment, the carotid bifurcation, and the internal carotid artery segment at both the left and right sides.

For the choice of which ultrasound protocol (based on combination of segments, angles, and walls measured) would be most optimal, many issues are important, including the reproducibility of the method; the ability of the method to assess change over time (the CIMT rate of progression); and the ability of the method to show an effect of treatment. Up until now, formal evaluations of differences in methodological aspects between ultrasound protocols are rather limited (27).

We set out to determine, in post-hoc analyses using the RADIANCE 1 and RADIANCE 2 studies, the protocol that could provide the best balance between reproducibility, magnitude of CIMT change over time and its associated precision, and magnitude of effect of intervention on CIMT change over time and its associated precision.

Methods

General

We used data of the RADIANCE 1 (28,29) and RADIANCE 2 (29,30) studies. These studies have been described in detail before. In summary, RADIANCE 1 was a double-blind randomized placebo-controlled multi-center trial in which 850 patients with heterozygous familial hypercholesterolemia (FH) were randomly assigned to receive either atorvastatin monotherapy or atorvastatin combined with 60 mg of torcetrapib for 2 years to study the effect on CIMT progression. RADIANCE 2 was a comparable study, a double-blind, placebo-controlled multi-center trial in which 752 participants with mixed dyslipidemia (MD) were randomly assigned to atorvastatin monotherapy or atorvastatin combined with 60 mg of torcetrapib for 2 years to assess the effect of torcetrapib, a cholesteryl ester transfer protein (CETP) inhibitor, on the progression of atherosclerosis, by measuring the thickening of carotid intima-media.

Carotid ultrasound examinations

The ultrasound protocol for assessment of the carotid intima-media thickness has been described in detail elsewhere (29). In short, duplicate scans were made at base-line and at each patient's final visit and single scans at visits at 6, 12, and 18 months, to give a maximum of seven scans for each patient. At each visit sonographers acquired and recorded CIMT images of 12 artery wall segments of the near and far walls of the right and left carotid artery for the common, bifurcation, and the internal carotid artery segments. In the trials a time slot of 45 minutes for image acquisition of one participant was recommended. The segments were defined as the common carotid artery (CCA) (from 10 to 20 mm proximal to the tip of the flow divider (TFD)), carotid bifurcation (from the TFD to 10 mm proximal to the TFD), and the proximal 10 mm of the internal carotid artery (ICA). At each segment images were collected at four predefined angles of 30° steps (90° to 180° on the right side and 270° to 180° on the left side) using Meijer's carotid arc (Figure 1). This resulted in 48 possible measurements per patient. All imaging centers used the same imaging acquisition protocol and equipment (Sequoia 512 scanners equipped with 8L5 transducers; Siemens AG, Munich, Germany). A total of 48 5-second image sequences (video clips) were saved in DICOM format (Digital Imaging in Communications in Medicine; National Electrical Manufacturers Association, Rosslyn, VA, USA). Imaging data were transferred directly from the study sites to the two reading centers (Vascular Imaging Center, University Medical Center, Utrecht, the Netherlands, and Wake Forest University Medical Center, Ultrasound Reading Center, Winston-Salem, NC, USA), where standardized equipment and protocols were used to process stored images. From every image sequence, readers selected one frame in end-diastole for measurement of carotid intima-media thickness. For each image sequence, the reader selected one frame for measurement when the heart is in systole (elctrocardiogram (ECG) tracing is on QRS complex). One-centimeter lengths of leading edge (far wall) and trailing edge (near wall) of media-adventitia and blood-intima boundaries were traced within the region of interest specified by the reader using an automated edge detection program (Artery Measurement System software; Chalmers University, Gothenburg, Sweden (31)). Maximum CIMT was determined from a set of measurements perpendicular to the media-adventitia boundary. Quality assurance protocols have been described elsewhere (29). In short, the sonographers and readers received a uniform training and underwent a uniform predefined certification procedure. Furthermore, quality assurance was continued during the conduct of the trial. Also, two sonographer/reader meetings were scheduled during the study to discuss the issue at hand, to reinforce the important work the sonographers did, and to suggest potential improvements, when appropriate. The quality assurance/quality control (QA/QC) was initiated to attempt to keep the quality of imaging at the highest level.

Figure 1. Meijer's arc.

Assessment of various protocols to measure CIMT

By selecting CIMT information of different combinations of segments, angles, walls, and both carotid arteries, 20 ultrasound protocols were constructed to estimate a mean common CIMT value and 40 ultrasound protocols to estimate a mean maximum CIMT value, resulting in a total of 60 ultrasound protocols (Table I). The protocols that have been used in the analyses of the RADIANCE 1 and RADIANCE 2 studies were protocol 50 for the mean maximum CIMT and protocol 10 for the mean common CIMT.

Table I. The 60 different ultrasound protocols constructed using the study data on segments, walls, and angles measured.

									Angles
				Walls		Segments			90	120	150	180
Segment	Wall	Angles	Approach	FW	NW	CCA	BIF	ICA	270	240	210	180
Mean common CIMT protocols
CCA	FW & NW	1 angle	01	+	+	+			+
			02	+	+	+				+
			03	+	+	+					+
			04	+	+	+						+
		2 angles	05	+	+	+			+	+
			06	+	+	+				+	+
			07	+	+	+					+	+
		3 angles	08	+	+	+			+	+	+
			09	+	+	+				+	+	+
		4 angles	10	+	+	+			+	+	+	+
	FW	1 angle	11	+		+			+
			12	+		+				+
			13	+		+					+
			14	+		+						+
		2 angles	15	+		+			+	+
			16	+		+				+	+
			17	+		+					+	+
		3 angles	18	+		+			+	+	+
			19	+		+				+	+	+
		4 angles	20	+		+			+	+	+	+
Mean maximum CIMT protocols
CCA & BIF	FW & NW	1 angle	21	+	+	+			+
			22	+	+	+				+
			23	+	+	+	+				+
			24	+	+	+	+					+
		2 angles	25	+	+	+	+		+	+
			26	+	+	+	+			+	+
			27	+	+	+	+				+	+
		3 angles	28	+	+	+	+		+	+	+
			29	+	+	+	+			+	+	+
		4 angles	30	+	+	+	+		+	+	+	+
	FW	1 angle	31	+		+	+		+
			32	+		+	+			+
			33	+		+	+				+
			34	+		+	+					+
		2 angles	35	+		+	+		+	+
			36	+		+	+			+	+
			37	+		+	+				+	+
		3 angles	38	+		+	+		+	+	+
			39	+		+	+			+	+	+
		4 angles	40	+		+	+		+	+	+	+
CCA & BIF & ICA	FW & NW	1 angle	41	+	+	+	+		+
			42	+	+	+	+			+
			43	+	+	+	+				+
			44	+	+	+	+					+
		2 angles	45	+	+	+	+	+	+	+
			46	+	+	+	+	+		+	+
			47	+	+	+	+	+			+	+
		3 angles	48	+	+	+	+	+	+	+	+
			49	+	+	+	+	+		+	+	+
		4 angles	50	+	+	+	+	+	+	+	+	+
	FW	1 angle	51	+		+	+	+	+
			52	+		+	+	+		+
			53	+		+	+	+			+
			54	+		+	+	+				+
		2 angles	55	+		+	+	+	+	+
			56	+		+	+	+		+	+
			57	+		+	+	+			+	+
		3 angles	58	+		+	+	+	+	+	+
			59	+		+	+	+		+	+	+
		4 angles	60	+		+	+	+	+	+	+	+

CIMT = carotid intima-media thickness; CCA=common carotid artery segment; BIF = bifurcation segment; ICA = internal carotid artery segment; NW = near wall; FW=far wall.

Data analysis

The CIMT estimates of each ultrasound protocol were compared on reproducibility, rate of CIMT progression, and effect of intervention on change of CIMT over time with its associated precision.

Reproducibility was assessed in two ways. Intra-class correlation coefficients (ICC) and the mean difference of the duplicate CIMT scan at base-line with its corresponding standard deviation were calculated for all protocols, and the protocols were ranked by the level of the ICC (the higher, the better) and the standard deviation (SD) of the mean difference (the lower, the better). The ideal and best protocol would have a mean difference of zero with a small variation around this mean difference.

Annual CIMT progression rate in the comparison group (atorvastatin only) was calculated by subtracting the mean CIMT value of the two duplicate scans at base-line from the mean CIMT value of the two duplicate scans at end of study and dividing by 2. In this group the rate of CIMT progression was expected to be slowed down and pointing towards zero, as lipid-lowering trials have shown statins to hold progression (32). Precision of CIMT progression was defined by the standard error (SE). To obtain a parameter that weighs the magnitude of the progression rate and the precision of the measurement, an absolute progression/precision (PP) ratio was constructed by dividing the mean progression rate by its SE and transforming this value into its absolute value. The absolute value was chosen as it is unknown whether atorvastatin treatment would only slow down CIMT progression or would even cause regression (negative rate of progression). The protocols were ranked by this absolute PP ratio from small to large, with the largest value defined as the protocol that achieved the best balance between CIMT progression rate and precision, as ideally one would like to use an ultrasound protocol that shows the highest progression with the highest precision (i.e. smallest SE).

To study the effect of the intervention on CIMT change, differences in CIMT progression and their standard errors were estimated with regression models in which CIMT progression over time was the dependent variable and treatment (atorvastatin with torcetrapib versus atorvastatin alone) was the independent variable. Next an absolute treatment/precision (TP) ratio was calculated by dividing the regression coefficient (represents change in CIMT progression due to treatment with torcetrapib) by its corresponding standard error and transforming this into an absolute value. Next the ultrasound protocols were ranked by this absolute TP ratio from low to high, as ideally one would like to use an ultrasound protocol that shows the highest effect of treatment on CIMT progression (largest regression coefficient) with the highest precision (i.e. smallest SE).

Finally the protocols were ranked by all three parameters for the mean common CIMT protocols, the mean maximum CIMT protocols, and all 60 protocols. A summary score was calculated for each protocol, based on summation of the overall ranks of reproducibility, progression/precision ratio and treatment/precision ratio after categorization of these scores in groups of five and six. Groups of five (for the 20 mean common CIMT protocols and 40 mean maximum CIMT protocols) and six (for the total ranking: 60 protocols) were used instead of the absolute ranks, as it was assumed that differences between the protocols were not large enough to justify a difference of 60 versus 1 between the best and the worst protocol, whereas a difference of 5 or 6 versus 1 between the best and the worst protocol was acceptable.

For the present study only data of the duplicate scans before randomization and the duplicate scans at the end of 24 months of study were used. Participants that had data for all four visits were included in the analyses (RADIANCE 1: n = 748; RADIANCE 2: n = 521). Analyses were performed by the complete-case analysis principle. Imputation of missing data was not done. SPSS statistical software (version 12.0) was used for the analyses.

Results

The study populations of the RADIANCE 1 and RADIANCE 2 studies differed on several baseline characteristics (Table II). Participants of the RADIANCE 1 study (FH patients) were younger, had higher levels of low density lipoprotein (LDL) cholesterol, lower levels of triglycerides, lower waist circumference values, and lower body mass index than participants of the RADIANCE 2 study (MD patients). There were more male participants and less hypertension in the RADIANCE 1 study compared to the RADIANCE 2 study.

Table II. Base-line characteristics of the RADIANCE 1 and RADIANCE 2 study populations. Values presented as mean (standard deviation) or number (percentage).

	RADIANCE 1	RADIANCE 2
	n=872	n=752
Age (years)	45 (12.5)	57 (8.2)
Male, n (%)	431 (49.4)	482 (64)
Mean maximum CIMT (mm)	1.14 (0.30)	1.31 (0.31)
Mean common CIMT (mm)	0.72 (0.15)	0.83 (0.15)
Systolic blood pressure (mmHg)	116 (11)	120 (11)
Diastolic blood pressure (mmHg)	73 (7)	74 (7)
HDL cholesterol (mg/dL)	52 (13)	47 (11)
LDL cholesterol (mg/dL)	139 (37)	100 (20)
Triglycerides (mg/dL)	113 (64)	183 (80)
Waist circumference (cm)	89 (12)	100 (13)
BMI (kg/m^2)	26.7 (4.4)	30.1 (4.4)
Hypertension, n (%)^a	230 (26.4)	390 (52)

BMI = body mass index; CIMT = carotid intima-media thickness; HDL = high density lipoprotein; LDL = low density lipoprotein. ^aHypertension defined as: systolic blood pressure >130 mmHg or diastolic blood pressure >85 mmHg or use of blood pressure-lowering medication.

The findings in patients with familial hypercholesterolemia

Data on both pre-randomization examinations and both end-study examinations were available for 748 out of 850 participants (88%) for the RADIANCE 1 study.

Reproducibility. ICC ranged from 0.774 to 0.908 overall, from 0.803 to 0.908 in the mean common CIMT protocols, and from 0.774 to 0.902 in the mean maximum CIMT protocols (Table III). The highest values of ICC were generally found for mean maximum CIMT and mean common CIMT protocols that included measurements at both walls at multiple angles (≥3 angles). The mean difference in CIMT between the two base-line examinations was close to 0 for almost all ultrasound protocols (Table III). The standard deviation (SD) of the mean difference ranged from 0.0777 to 0.0866 for the mean common CIMT ultrasound protocols and from 0.1215 to 0.2051 for the mean maximum CIMT ultrasound protocols. Mean common CIMT protocols had smaller SD than mean maximum CIMT protocols. Mean common CIMT protocols with measurements of both walls at ≥2 angles had the smallest SD. Mean maximum CIMT protocols with measurements of both walls of all three carotid segments had the smallest SD.

Table III. Reproducibility of CIMT measurements at base-line in the RADIANCE 1 study (n=748), estimated as intra-class correlation coefficients (ICC) and the mean difference in CIMT (mm) between duplicate scans. Protocols were ranked in an order of low to high reproducibility (increasing ICC and decreasing standard deviation of mean difference): the higher the rank, the better the protocol. Ranking ICC for all protocols (reported for CCA the top 4; for MeanMax the top 8, and for all the top 10). Please see Table I for precise information on the angles used in each protocol.

					Ranking ICC					Ranking SD
Segments	Walls	Angles	Approach	ICC	Mean common CIMT	Mean maximum CIMT	Overall	Mean	SD	Mean common CIMT	Mean maximum CIMT	Overall
CCA	FW & NW	1 angle	1	0.843				0.0032	0.0866			53
			2	0.868				0.0058	0.0852			54
			3	0.864				0.0000	0.0874			52
			4	0.811				0.0029	0.1000
		2 angles	5	0.891				0.0075	0.0805	18		58
			6	0.904	18		58	0.0020	0.0777	20		60
			7	0.884				0.0014	0.0849			55
		3 angles	8	0.908	20		60	0.0048	0.0784	19		59
			9	0.904	17		57	0.0026	0.0808	17		57
		4 angles	10	0.906	19		59	0.0041	0.0814			56
	FW	1 angle	11	0.803				0.0016	0.1162
			12	0.830				0.0058	0.1094
			13	0.846				−0.0004	0.1043
			14	0.809				0.0033	0.1129
		2 angles	15	0.856				0.0070	0.1061
			16	0.879				0.0023	0.0984
			17	0.871				−0.0007	0.0993
		3 angles	18	0.886				0.0036	0.0985
			19	0.886				0.0007	0.0983
		4 angles	20	0.891				0.0016	0.0979			51
CCA & BIF	FW & NW	1 angle	21	0.839				0.0023	0.1476
			22	0.863				−0.0061	0.1391
			23	0.869				−0.0025	0.1323
			24	0.844				0.0023	0.1363
		2 angles	25	0.880				0.0000	0.1425
			26	0.890				−0.0081	0.1354
			27	0.886				−0.0034	0.1314
		3 angles	28	0.893				−0.0032	0.1425
			29	0.896		34		−0.0074	0.1368
		4 angles	30	0.899		36	52	−0.0032	0.1427
	FW	1 angle	31	0.774				−0.0022	0.2051
			32	0.815				−0.0058	0.1932
			33	0.846				−0.0073	0.1790
			34	0.817				0.0008	0.1858
		2 angles	35	0.836				−0.0023	0.1963
			36	0.869				−0.0116	0.1778
			37	0.868				−0.0115	0.1760
		3 angles	38	0.868				−0.0059	0.1871
			39	0.870				−0.0140	0.1859
		4 angles	40	0.876				−0.0086	0.1883
CCA & BIF & ICA	FW & NW	1 angle	41	0.864				0.0006	0.1295		33
			42	0.883				−0.0104	0.1239		39
			43	0.879				0.0010	0.1215		40
			44	0.851				0.0037	0.1286		36
		2 angles	45	0.893		33		−0.0045	0.1290		35
			46	0.900		37	53	−0.0081	0.1246		38
			47	0.888				0.0004	0.1253		37
		3 angles	48	0.898		35	51	−0.0040	0.1342
			49	0.902		40	56	−0.0071	0.1293		34
		4 angles	50	0.901		38	54	−0.0042	0.1373
	FW	1 angle	51	0.902		39	55	−0.0056	0.1721
			52	0.844				−0.0111	0.1696
			53	0.861				−0.0024	0.1593
			54	0.812				0.0025	0.1777
		2 angles	55	0.865				−0.0097	0.1697
			56	0.881				−0.0134	0.1616
			57	0.871				−0.0054	0.1661
		3 angles	58	0.882				−0.0105	0.1691
			59	0.884				−0.0130	0.1699
		4 angles	60	0.890				−0.0109	0.1723

CIMT = carotid intima-media thickness; CCA=common carotid artery segment; BIF = bifurcation segment; ICA = internal carotid artery segment; NW = near wall; FW=far wall; ICC = intra-class correlation coefficient; mean = mean difference; SD = standard deviation.

Rate of CIMT progression over time in the atorvastatin group. The rate of CIMT progression in the atorvastatin alone group ranged from −0.00302 to 0.00029 mm/year for the mean common CIMT protocols (Table IV). All but one protocol showed negative slopes. The CIMT progression for the mean maximum protocols showed positive and negative directions depending on the angles that were used. The absolute progression/precision ratio (PP ratio) ranged from 0.44 to 3.73 for the mean common CIMT protocols and from 0.06 to 1.50 for the mean maximum CIMT protocols. The highest ratios were obtained with mean common CIMT protocols that included measurements of the near and far wall at multiple angles. In the mean maximum CIMT protocols the highest ratios were found in two-segment protocols that included measurements of the far wall and in three-segment protocols with measurements of the near and far wall.

Table IV. Left side: Annual rate of progression of CIMT (mm), its precision (standard errors), absolute ratio of progression/precision and ranking of the ratio for all 60 ultrasound protocols in the RADIANCE 1 study for the atorvastatin only group (n=364). Right side: Effect of treatment on CIMT rate of progression over time in RADIANCE 1 (n=748). Regression coefficients represent the increase in CIMT per year due to treatment with torcetrapib. Protocols were ranked in an order from low to high: the higher the rank, the better the protocol. Ranking reported for CCA the top 4; for MeanMax the top 8, and for all the top 10. Please see Table I for precise information on the angles used in each protocol.

			Progression rates						Treatment effects
						Ranking ratio						Ranking ratio
Segments; Walls	Angles	Approach	Progression	SE	Ratio	Mean common CIMT	Mean Max CIMT	Overall	Regression coefficient	SE	Ratio	Mean common CIMT	Mean Max CIMT	Overall
CCA; FW & NW	1 angle	1	0.00029	0.00066	0.44				0.00107	0.00089	1.21
		2	−0.00056	0.00054	1.05				0.00196	0.00078	2.50
		3	−0.00220	0.00063	3.52	18		58	0.00320	0.00091	3.52	17		57
		4	−0.00302	0.00081	3.73	20		60	0.00315	0.00104	3.02			51
	2 angles	5	−0.00040	0.00059	0.67				0.00156	0.00081	1.93
		6	−0.00172	0.00055	3.15			56	0.00310	0.00083	3.74	19		59
		7	−0.00259	0.00071	3.65	19		59	0.00366	0.00094	3.91	20		60
	3 angles	8	−0.00145	0.00061	2.36				0.00252	0.00086	2.92
		9	−0.00210	0.00065	3.25	17		57	0.00334	0.00089	3.73	18		58
	4 angles	10	−0.00167	0.00067	2.48				0.00281	0.00091	2.73
CCA; FW	1 angle	11	−0.00092	0.00085	1.09				0.00087	0.00116	0.75
		12	−0.00069	0.00074	0.93				0.00192	0.00104	1.86
		13	−0.00220	0.00081	2.71			54	0.00374	0.00112	3.35			56
		14	−0.00193	0.00087	2.21				0.00180	0.00116	1.55
	2 angles	15	−0.00150	0.00073	2.07				0.00207	0.00100	2.07
		16	−0.00186	0.00072	2.60			52	0.00328	0.00102	3.23			55
		17	−0.00212	0.00079	2.68			53	0.00342	0.00108	3.16			54
	3 angles	18	−0.00215	0.00074	2.90			55	0.00313	0.00102	3.07			52
		19	−0.00188	0.00075	2.52				0.00307	0.00104	2.96
	4 angles	20	−0.00197	0.00078	2.53			51	0.00287	0.00105	3.09			53
CCA & BIF; FW&NW	1 angle	21	0.00133	0.00115	1.16		36		0.00305	0.00155	1.97		36
		22	0.00045	0.00096	0.47				0.00043	0.00129	0.34
		23	−0.00036	0.00091	0.40				0.00343	0.00131	2.63		40
		24	−0.00018	0.00113	0.16				0.00208	0.00150	1.39
	2 angles	25	0.00069	0.00105	0.65				0.00159	0.00140	1.14
		26	−0.00037	0.00092	0.40				0.00265	0.00129	2.06		37
		27	−0.00020	0.00097	0.20				0.00328	0.00140	2.35		39
	3 angles	28	0.00020	0.00098	0.20				0.00270	0.00138	1.96		35
		29	−0.00017	0.00097	0.18				0.00208	0.00136	1.53		34
	4 angles	30	0.00037	0.00102	0.36				0.00207	0.00142	1.46		33
CCA & BIF; FW	1 angle	31	−0.00022	0.00157	0.14				0.00213	0.00209	1.02
		32	0.00022	0.00130	0.17				−0.00098	0.00174	0.56
		33	−0.00084	0.00133	0.63				0.00400	0.00186	2.15		38
		34	0.00196	0.00162	1.21		38		−0.00048	0.00214	0.22
	2 angles	35	−0.00052	0.00136	0.38				0.00046	0.00182	0.25
		36	−0.00101	0.00127	0.80				0.00243	0.00174	1.40
		37	0.00031	0.00136	0.23				0.00248	0.00196	1.27
	3 angles	38	−0.00099	0.00132	0.75				0.00233	0.00178	1.31
		39	−0.00008	0.00133	0.06				0.00082	0.00186	0.44
	4 angles	40	−0.00012	0.00139	0.09				0.00074	0.00188	0.39
CCA & BIF & ICA; FW & NW	1 angle	41	0.00163	0.00108	1.50		40		0.00173	0.00144	1.20
		42	0.00035	0.00086	0.41				−0.00022	0.00117	0.19
		43	0.00090	0.00089	1.01		34		0.00141	0.00126	1.12
		44	0.00006	0.00110	0.06				0.00124	0.00140	0.88
	2 angles	45	0.00109	0.00098	1.11		35		0.00055	0.00130	0.42
		46	0.00049	0.00090	0.54				0.00077	0.00124	0.62
		47	0.00054	0.00092	0.59				0.00171	0.00131	1.30
	3 angles	48	0.00126	0.00096	1.31		39		0.00095	0.00131	0.72
		49	0.00053	0.00093	0.57				0.00058	0.00129	0.45
	4 angles	50	0.00119	0.00099	1.19		37		0.00058	0.00136	0.43
CCA & BIF & ICA; FW	1 angle	51	0.00043	0.00140	0.31				0.00126	0.00189	0.67
		52	−0.00030	0.00111	0.27				−0.00133	0.00154	0.86
		53	0.00032	0.00116	0.28				0.00200	0.00167	1.20
		54	0.00134	0.00149	0.90		33		−0.00012	0.00192	0.06
	2 angles	55	−0.00042	0.00116	0.36				−0.00006	0.00162	0.03
		56	−0.00040	0.00113	0.35				0.00057	0.00157	0.37
		57	0.00026	0.00122	0.21				0.00178	0.00180	0.99
	3 angles	58	−0.00012	0.00116	0.10				0.00094	0.00161	0.59
		59	−0.00023	0.00120	0.19				0.00014	0.00171	0.08
	4 angles	60	−0.00010	0.00122	0.08				0.00043	0.00173	0.25

CIMT=carotid intima-media thickness; CCA=common carotid artery segment; BIF=bifurcation segment; ICA=internal carotid artery segment; SE=standard error; PP ratio=progression/precision ratio, calculated as progression/SE. Protocols were ranked in an order of lowest to highest absolute value of progression/SE ratio. Regression coefficient of CIMT represents the amount of increase in CIMT per year due to treatment with torcetrapib; treatment/precision ratio calculated as absolute value of beta divided by the SE; the regression model used: ‘Progression Rate=a+beta (treatment)’.

Effect of treatment effect on CIMT change over time. The effect of torcetrapib as compared to placebo on CIMT progression ranged from 0.00087 to 0.000374 mm/year for the mean common CIMT protocols and from −0.00133 to 0.00400 mm/year for the mean maximum CIMT protocols (Table IV). The smallest treatment effects were found in the mean maximum protocols, notably in protocols that included measurements of three segments.

The absolute treatment/precision ratios (TP ratios) ranged from 0.75 to 3.91 for the mean common CIMT protocols and from 0.03 to 2.63 for the mean maximum CIMT protocols. The highest TP ratios were found predominantly in the mean common CIMT protocols. Protocols that included measurements of both walls at multiple angles had higher ratios than protocols that were based on measurements of only the far wall at one angle. When the interest was in mean maximum CIMT protocols, the highest TP ratios were found in two-segment protocols with measurements of the near and far wall.

Overall ranking based on reproducibility, rate of CIMT progression, and effect of treatment on CIMT change over time. Mean common CIMT protocols provided the best balance between high reproducibility, large and precise rates of CIMT progression, and effect of treatment (Table VII). The highest mean common CIMT ranking protocol was the one based on measurements from both near and far wall, at angles 120, 150, and 180 of the right carotid artery and angles 180, 210, and 240 of the left carotid artery. Mean common CIMT protocols that measured both walls were ranked higher than single wall protocols. In the mean maximum CIMT protocol, the protocols with highest total rank scores were those that included measurements of the near and far wall. The highest mean maximum CIMT ranking protocol was the one based on measurements from three segments (CCA, BIF, ICA) at near and far wall, at angles 90, 120, and 150 or 120, 150, and 180 of the right carotid artery and angles 180, 210, and 240 or 210, 240, and 270 of the left carotid artery (approach #45, 46, 47, 48). Overall three-segment mean maximum CIMT protocols that included measurements of both walls at multiple angles had the highest total rank scores.

The RADIANCE 2 study findings

Data on both pre-randomization examinations and end-study examinations were available for 521 out of 752 participants (69%) in the RADIANCE 2 study.

Reproducibility. ICC ranged from 0.678 to 0.847 for the mean common CIMT protocols and from 0.684 to 0.861 for the mean maximum CIMT protocols (Table V). The highest values of ICC were found for mean common CIMT protocols that included measurements at multiple angles (≥2 angles) and for mean maximum CIMT protocols that included measurements of both the near and the far wall at ≥2 angles. The mean difference in CIMT between the two base-line examinations was close to 0 for almost all ultrasound protocols (Table V). The standard deviation (SD) of the mean difference ranged from 0.0974 to 0.1572 for the mean common CIMT ultrasound protocols and from 0.1584 to 0.2561 for the mean maximum CIMT ultrasound protocols. Mean common CIMT protocols had smaller SD than mean maximum CIMT protocols. Mean common CIMT protocols with measurements of both walls at ≥2 angles had the smallest SD. Of the mean maximum CIMT protocols, three-segment protocols using measurements of both walls at multiple angles were ranked highest.

Table V. Reproducibility of CIMT measurements at base-line in the RADIANCE 2 study (n=521), estimated as intra-class correlation coefficients (ICC) and the mean difference in CIMT (mm) between duplicate scans and ranking for all protocols. Protocols were ranked in an order of low to high reproducibility (increasing ICC and decreasing standard deviation of mean difference): the higher the rank, the better the protocol. Ranking ICC for all protocols reported for CCA the top 4; for MeanMax the top 8 and for all the top 10. Please see Table I for precise information on the angles used in each protocol.

					Ranking ICC					Ranking SD
Segment	Wall	Angles	Approach	ICC	Mean common CIMT	Mean maximum CIMT	Overall	Mean	SD	Mean common CIMT	Mean maximum CIMT	Overall
CCA	FW & NW	1 angle	1	0.771				−0.0051	0.1114			53
			2	0.805				0.0000	0.1032			55
			3	0.743				−0.0009	0.1243
			4	0.705				−0.0020	0.1295
		2 angles	5	0.836				0.0007	0.0974	20		60
			6	0.834				−0.0019	0.0987	19		59
			7	0.797				0.0010	0.1107			54
		3 angles	8	0.843	19		56	−0.0009	0.0988	18		58
			9	0.830				0.0003	0.1029			56
		4 angles	10	0.847	20		58	0.0004	0.1000	17		57
	FW	1 angle	11	0.720				−0.0100	0.1504
			12	0.757				−0.0056	0.1365
			13	0.782				−0.0049	0.1269
			14	0.678				−0.0074	0.1572
		2 angles	15	0.821				−0.0012	0.1209
			16	0.827				−0.0066	0.1163			51
			17	0.780				−0.0041	0.1330
		3 angles	18	0.841	18		52	−0.0042	0.1159			52
			19	0.828			51	−0.0061	0.1209
		4 angles	20	0.841	17			−0.0040	0.1197
CCA & BIF	FW & NW	1 angle	21	0.773				−0.0050	0.1909
			22	0.775				0.0152	0.1803
			23	0.749				0.0013	0.1880
			24	0.732				0.0030	0.1854
		2 angles	25	0.811				0.0098	0.1809
			26	0.821		32		0.0115	0.1728
			27	0.809				0.0041	0.1706		34
		3 angles	28	0.842		36	54	0.0067	0.1738
			29	0.842		35	53	0.0095	0.1706		33
		4 angles	30	0.861		40	60	0.0057	0.1696		35
	FW	1 angle	31	0.684				−0.0074	0.2556
			32	0.717				0.0200	0.2374
			33	0.695				0.0047	0.2561
			34	0.721				0.0073	0.2420
		2 angles	35	0.743				0.0150	0.2451
			36	0.769				0.0150	0.2315
			37	0.773				0.0154	0.2319
		3 angles	38	0.785				0.0134	0.2346
			39	0.814				0.0181	0.2211
		4 angles	40	0.814				0.0162	0.2280
CCA & BIF & ICA	FW & NW	1 angle	41	0.798				−0.0053	0.1743
			42	0.773				0.0116	0.1738
			43	0.772				0.0024	0.1694		36
			44	0.760				−0.0013	0.1709
		2 angles	45	0.812				0.0100	0.1769
			46	0.816				0.0098	0.1686		38
			47	0.821				0.0061	0.1584		40
		3 angles	48	0.845		38	57	0.0093	0.1692		37
			49	0.843		37	55	0.0106	0.1643		39
		4 angles	50	0.855		39	59	0.0116	0.1708
	FW	1 angle	51	0.712				−0.0121	0.2328
			52	0.732				0.0113	0.2196
			53	0.753				−0.0006	0.2175
			54	0.745				0.0073	0.2202
		2 angles	55	0.768				0.0114	0.2227
			56	0.783				0.0102	0.2163
			57	0.803				0.0128	0.2076
		3 angles	58	0.808				0.0096	0.2149
			59	0.827		33		0.0183	0.2049
		4 angles	60	0.833		34		0.0174	0.2112

CIMT = carotid intima-media thickness; CCA = common carotid artery segment; BIF = bifurcation segment; ICA=internal carotid artery segment; NW = near wall, FW=far wall; ICC = intra-class correlation coefficient; mean = mean difference; SD = standard deviation.

Rate of CIMT progression over time in the atorvastatin group. The rate of progression of mean common CIMT protocols ranged from 0.00084 to 0.00539 mm/year and of mean maximum CIMT protocols from 0.00398 to 0.01057 mm/year. The progression/precision ratio (PP ratio) ranged from 0.84 to 5.61 for the mean common CIMT protocols and from 1.85 to 5.44 for the mean maximum CIMT protocols (Table VI). Overall mean common CIMT protocols that included measurements of both walls were ranked highest. Mean maximum CIMT protocols that included measurements at all three segments of both walls at multiple angles showed the highest ratios between rate of CIMT progression over time and its corresponding precision.

Table VI. Left side: Annual rate of progression of CIMT (mm), its precision (standard errors), absolute ratio of progression/precision and ranking of the ratio for all 60 ultrasound protocols in the RADIANCE 2 study for the atorvastatin only group (n = 268). Right side: Effect of treatment on CIMT rate of progression over time in RADIANCE 2 (n = 521). Regression coefficients represent the increase in CIMT per year due to treatment with torcetrapib. Protocols were ranked in an order from low to high: the higher the rank, the better the protocol. Please see Table I for precise information on the angles used in each protocol. Ranking for all protocols (for CCA top 4; for MeanMax top 8 and for all the top 10).

			Progression rates						Treatment effects
						Ranking ratio						Ranking ratio
Segments; Walls	Angles	Approach	Progression	SE	Ratio	Mean common CIMT	Mean maximum CIMT	Overall	Regression coefficient	SE	Ratio	Mean common CIMT	Mean maximum CIMT	Overall
CCA; FW & NW	1 angle	1	0.00289	0.00099	2.91				0.00199	0.00151	1.32
		2	0.00225	0.00136	4.75				0.00019	0.00132	0.14			55
		3	0.00405	0.00085	2.43				0.00156	0.00137	1.14
		4	0.00521	0.00116	0.84				0.00327	0.00152	2.14	19		54
	2 angles	5	0.00247	0.00102	5.44	19		58	0.00014	0.00131	0.11			59
		6	0.00440	0.00100	5.17	17		55	0.00098	0.00122	0.8
		7	0.00084	0.00101	3.47				0.00190	0.00134	1.41	18
	3 angles	8	0.00158	0.00119	5.61	20		60	0.00111	0.00126	0.88			57
		9	0.00474	0.00087	4.72				0.00098	0.00129	0.76			51
	4 angles	10	0.00472	0.00115	4.44				0.00071	0.00133	0.53
CCA; FW	1 angle	11	0.00429	0.00083	1.65				0.00274	0.00197	1.39	17
		12	0.00539	0.00101	4.51				−0.00088	0.00169	0.52			56
		13	0.00326	0.00094	4.39				−0.00101	0.00149	0.68
		14	0.00455	0.00108	1.33				0.00400	0.00185	2.16	20
	2 angles	15	0.00481	0.00086	4.09				0.00037	0.00164	0.22			60
		16	0.00529	0.00107	5.35	18		57	−0.00065	0.00144	0.45
		17	0.00414	0.00088	4.21				0.00089	0.00157	0.56
	3 angles	18	0.00480	0.00111	4.93			51	0.00002	0.00154	0.01
		19	0.00471	0.00092	4.34				0.00040	0.00152	0.26
	4 angles	20	0.00505	0.00114	5.13			54	0.00046	0.00160	0.29
CCA & BIF; FW&NW	1 angle	21	0.00456	0.00180	2.54				0.00283	0.00279	1.02		39	53
		22	0.00551	0.00226	3.73				0.00023	0.00229	0.10
		23	0.00560	0.00150	2.43				0.00154	0.00227	0.68
		24	0.00713	0.00206	2.26				0.00232	0.00248	0.94		38	52
	2 angles	25	0.00411	0.00169	4.27				0.00108	0.00241	0.45
		26	0.00515	0.00223	4.18				0.00025	0.00222	0.11
		27	0.00398	0.00176	3.74				0.00185	0.00240	0.77
	3 angles	28	0.00424	0.00229	4.51				0.00063	0.00234	0.27
		29	0.00675	0.00158	4.40				0.00033	0.00241	0.14
	4 angles	30	0.00745	0.00218	4.62				0.00027	0.00252	0.11
CCA & BIF; FW	1 angle	31	0.00627	0.00150	2.44				0.00303	0.00362	0.84		36
		32	0.00769	0.00203	3.47				−0.00027	0.00313	0.09
		33	0.00620	0.00166	2.30				0.00048	0.00306	0.16
		34	0.00650	0.00221	1.85				0.00486	0.00333	1.46		40	58
		35	0.00697	0.00155	3.42				0.00083	0.00318	0.26
	2 angles	36	0.00780	0.00216	3.78				−0.00178	0.00297	0.60
		37	0.00703	0.00160	2.93				0.00265	0.00323	0.82		35
		38	0.00763	0.00214	3.62				−0.00092	0.00303	0.30
	3 angles	39	0.00766	0.00166	3.57				−0.00014	0.00316	0.05
	4 angles	40	0.00796	0.00223	3.57				0.00006	0.00318	0.02
		41	0.00462	0.00171	2.70				0.00143	0.00253	0.57
		42	0.00625	0.00218	3.32				0.00172	0.00217	0.80		34
	1 angles	43	0.00463	0.00139	3.29				−0.00071	0.00211	0.34
		44	0.00631	0.00188	2.81				0.00031	0.00227	0.14
CCA & BIF & ICA; FW & NW		45	0.00509	0.00155	4.14				0.00101	0.00236	0.43
	2 angles	46	0.00707	0.00205	4.80		35		−0.00100	0.00210	0.48
		47	0.00470	0.00167	4.69		33		−0.00052	0.00218	0.24
	3 angles	48	0.00594	0.00213	4.88		36		−0.00098	0.00228	0.43
		49	0.00645	0.00156	5.44		40	59	−0.00144	0.00226	0.63
	4 angles	50	0.00838	0.00212	5.29		39	56	−0.00132	0.00242	0.54
		51	0.00667	0.00139	2.87				0.00144	0.00315	0.46
	1 angle	52	0.00874	0.00185	3.36				0.00218	0.00289	0.75
		53	0.00727	0.00155	3.44				−0.00217	0.00281	0.77		33
CCA & BIF & ICA; FW		54	0.00870	0.00210	2.79				0.00218	0.00300	0.73
		55	0.00737	0.00151	3.95				0.00103	0.00301	0.34
	2 angles	56	0.00949	0.00205	4.73		34		−0.00231	0.00275	0.84		37
		57	0.00829	0.00152	4.14				−0.00025	0.00300	0.09
		58	0.00978	0.00197	4.63				−0.00222	0.00291	0.76
	3 angles	59	0.00858	0.00162	4.95		38	53	−0.00113	0.00298	0.38
	4 angles	60	0.01057	0.00214	4.94		37	52	−0.00154	0.00311	0.50

Treatment effect on CIMT change over time. The effect of torcetrapib as compared to placebo on CIMT progression ranged from −0.00101 to 0.00400 mm/year for the mean common CIMT protocols and from −0.00231 to 0.00486 mm/year for the mean maximum CIMT protocols (Table VI). The absolute treatment/precision ratios (TP ratios) ranged from 0.01 to 2.16 for the mean common CIMT protocols and from 0.02 to 1.46 for the mean maximum CIMT protocols. The highest TP ratios were found mainly in the mean common CIMT protocols that used measurements of both walls. Mean maximum protocols that included measurements of only the common carotid and the bifurcation segment showed higher TP ratios than three-segment protocols. Protocols with measurements of the near and far wall had higher ratios in both the mean common and the mean maximum protocols.

Overall ranking based on reproducibility, rate of CIMT progression, and effect of treatment on CIMT change over time. Mean common CIMT protocols provided the best balance between high reproducibility, large and precise rates of CIMT progression, and effect of treatment (Table VII). The highest mean common CIMT ranking protocol was the one based on measurements from both near and far wall, at angles 90, 120, and 150 of the right carotid artery and angles 210, 240, and 270 of the left carotid artery (approach #8). Mean common CIMT protocols that measured both walls had higher total rank scores than single-wall protocols. In the mean maximum CIMT protocols, the highest total rank scores were found in protocols that used measurements of the near and far wall. The highest mean maximum CIMT ranking protocol was the one based on measurements from both near and far wall, at angles 120, 150, and 180 of the right carotid artery and angles 180, 210, and 240 of the left carotid artery (approach #49). Mean maximum CIMT protocols that included measurements of all three segments at multiple angles had the highest total rank scores.

Table VII. Overall ranking of the ultrasound protocols based on summation of the transformed rank scores (mean common CIMT and mean maximum CIMT into quintile scores; overall rank scores into sixtile scores) for the RADIANCE 1 and RADIANCE 2 study. Ranking for all protocols (for CCA top 4; for MeanMax top 8, and for all the top 10).

				RADIANCE 1			RADIANCE 2
Segments	Walls	Angles	Approach	Mean common CIMT	Mean maximum CIMT	Overall	Mean common CIMT	Mean maximum CIMT	Overall
CCA	FW & NW	1 angle	1
			2						21
			3	15
			4
			5				15		23
		2 angles	6	19		24	18
			7	17		22
			8	15		22	19		24
		3 angles	9	20		24	15		22
		4 angles	10			23	16		21
	FW	1 angle	11
			12
			13
			15						20
		2 angles	16			21
			17
			18						22
		3 angles	19
		4 angles	20			22
CCA & BIF	FW & NW	1 angle	21
			22
			23
			24
			25
		2 angles	26		16
			27					16
			28
		3 angles	29		16
		4 angles	30		16
	FW	1 angle	31
			33
			34
			35
		2 angles	36
			37
			38
		3 angles	39
		4 angles	40
CCA & BIF & ICA	FW & NW	1 angle	41		16
			42
			43		16
			44
			45		17
		2 angles	46		17			17
			47		17			16
			48		17			18
		3 angles	49		16			19	20
		4 angles	50		16			17	20
	FW	1 angle	51
			52
			53
			54
			55
		2 angles	56
			57
		3 angles	59					16
		4 angles	60					16

CIMT = carotid intima-media thickness; CCA=common carotid artery segment; BIF = bifurcation segment; ICA = internal carotid artery segment; NW = near wall, FW = far wall.

Discussion

Mean common CIMT that was assessed through measurements of both the near and far wall of the common carotid artery segment at multiple (≥2) angles provided the best balance between high reproducibility, large estimates of CIMT progression over time, large change in progression over time induced by the intervention, and its associated precision. With regard to mean maximum CIMT, protocols that included measurements of both near and far wall of all three carotid segments performed best.

Some aspects of this study need to be addressed. Firstly, all CIMT measurements in our study were fixed as regards angle through use of Meijer's carotid arc. This guaranteed that consecutive scans could be performed at approximately the same angle. Our results, however, might differ from CIMT ultrasound protocols that are dedicated to finding the previous image (insight/memory) or those that interrogate the maximum measurements (independent of angles), or interrogate from a single angle that visually measures the best image from which CIMT should be measured. However, there are no data available to indicate how protocols with absence of angle approaches would rank in our tables. In the RADIANCE studies, CIMT measurements were performed independent of whether or not a plaque or a lesion was present. In that sense our approach differed from that of the Mannheim consensus (CIMT measurement in an area free of plaque) (33), yet resembles the approach taken in many other studies on the risk of events and other intervention studies. Also, in our studies the same ultrasound equipment was used in all centers. Having different ultrasound equipment in a randomized trial, one would expect that the variability in the measurements goes up to some extent.

Secondly, reproducibility was assessed by the ICC and the standard deviation of the mean difference. The ICC is a commonly used and well known measure to evaluate reproducibility. It quantifies measurement reliability by comparing the variability of different measurements of the same subject to the total variation across all measurements and all subjects. One characteristic of the ICC is that it depends on the magnitude of the variation between individuals. If the variation between individuals is large, relative to measurement error, the ICC will be high. In contrast, if the variation between individuals is small and CIMT is measured with the same measurement error, the ICC will be small. Multiple-segment protocols could therefore produce higher ICCs than single-segments protocols, based on this difference in between-individuals variation. Therefore, protocols were also ranked according to the standard deviation of the mean difference of the duplicate scan. This measure is a suitable test to assess reproducibility and is less affected by between-individuals variations. In the assessment of reproducibility the mean absolute difference (MAD) of the duplicate base-line scan was also evaluated (results not shown). Ranking of the protocols on the absolute mean difference provided similar results as the ranking based on SD of the mean difference. The correlation between these two parameters was close to 1, indicating that these two parameters assess the same outcome.

Thirdly, protocols were ranked based on the absolute progression/precision ratio in the atorvastatin monotherapy group. Most ideally, one would like to rank protocols on progression/precision in a placebo-group as it is then clear in which direction the CIMT progression rate is assumed to go: progression of CIMT rates over time. However in the group that received atorvastatin, the direction was not known. As previous studies showed that statin treatment produces at most a slowing down of the CIMT progression rate that is close to zero (34,35), the absolute PP ratio was used when ranking on magnitude of progression rate (either in a negative direction or in a positive direction) and precision. Yet, in our study negative progression rates in both the mean common and the mean maximum CIMT protocols were observed. We also ranked the protocols based on direction of CIMT progression, i.e. ranking protocols with negative progression rates and high precision more highly (results not shown); however, this only had a minor effect on the ranking. Finally, we observed the same difficulty of direction of progression in the ranking of treatment effect on progression of CIMT. Although the event trial with atorvastatin and torcetrapib showed increased risk in the torcetrapib arm as compared to atorvastatin monotherapy (36), we decided to rank the ultrasound protocols as if treatment with torcetrapib was to slow down CIMT progression. This was done for two reasons: 1) CIMT trials are designed to show benefit of treatment, and if treatment is expected to cause harm a CIMT trial would not be conducted beforehand; and 2) the event trial showed that torcetrapib therapy resulted in an increased risk of mortality and morbidity of an unknown mechanism, with evidence of an off-target effect of torcetrapib (37). Currently trials with other CETP inhibitors are being conducted, and these will provide information whether this class of drugs yields an effect on CIMT progression. To overcome this difficult problem, we ranked based on the absolute TP ratio, which corresponds only to magnitude of effect with precision, whilst discarding the direction of effect of treatment (positive or negative).

The generalizability of the results of this study may depend on the study populations since reproducibility, progression and its precision, and effect of treatment and its precision may be affected by the level of atherosclerosis. However, the implications on choice of ultrasound protocol seem to be equal for both study populations. For the FH populations the best balance between the three parameters was found in mean common CIMT protocols, with increasing ratios if both walls were measured at more angles. This also applies to the MD populations; yet in the MD population, some mean maximum CIMT protocols also rank favorably.

For decades a lively discussion has been on-going regarding the desirability of performing near-wall CIMT measurements (38). The reluctance is based on the physics of ultrasound indicating that the near-wall CIMT is at best an approximation of the true value, whereas the advocates indicate that the addition of the near-wall measurements increase precision and hence increase precision in risk prediction and evaluation of drug effects. Our study indicated that the best protocols with regard to reproducibility, PP ratio, and TP ratio were all predominantly based on data collected from both the far and the near wall. The results of our study support the view that ultrasound protocols that measure both walls should be preferred above protocols that measure only the far wall.

In our study we used an automated edge detection program. These approaches have been suggested as a very useful approach to reduce considerably variability in the measurement of CIMT. Most of these automated edge detection programs have been applied for reading of the far wall of the common carotid artery, whereas the near wall is not considered, nor are measurements at the bifurcation and in the internal segments. The main potential advantage of automated edge detection programs is that they may reduce variability in CIMT readings as a result of reduction in the differences between readers, and by the elimination of change in reading behavior over time (reader drift). However, the beneficial effect of automated edge detection in reduction of measurement variability depends on the contribution of sonographers and readers on the total variability in the CIMT measurement. A number of studies have indicated that differences between sonographers have a much larger effect on the CIMT measurements than differences between readers (39,40). Furthermore, a formal quantification of the extent of reduction in measurement error using either approach has not been done. In general, when ultrasound images show clear interfaces automated edge detection programs work very well, whereas when the interfaces on the ultrasound images are less clear the automated edge detection program needs to be manually overridden, thereby eliminating the advantages of the use of automated edge detection.

In every study using CIMT measurements, whether a cohort study or a randomized controlled trial, a choice needs to be made based on the costs and the benefits. Benefits relate to high reproducibility, precise measurements, and when appropriate a high progression/precision balance, and thus a high likelihood of success. Costs relate to the time needed to acquire the CIMT information (ultrasound scanning and reading) and the financial expenses related to that process. Acquisition times may vary from 15 minutes for protocols that assess CIMT in one segment and one angle (e.g. protocols #1–4 or #11–14) to 45 minutes for protocols that assess CIMT in all three segments and at multiple angles (e.g. protocols #45–50). Similarly, off-line reading time may vary from 15 minutes for the single-segment, one-angle protocols to 1.5 to 2 hours for the multi-segment, multi-angle protocols. The balance between quality and costs should be evaluated and considered in the decision-making process. We feel that the present analyses provide sufficient information to estimate that balance in an evidence-based way.

To the best of our knowledge published data of this type are limited, most likely since no other study has such a wealth of detailed information on CIMT measurements as was available in the RADIANCE 1 and RADIANCE 2 studies. Data from other studies with similar extensive ultrasound protocols were analyzed and provided information on the best ultrasound protocol for trials in other population groups. In healthy middle-aged subjects, in the OPAL study (27), mean common CIMT ultrasound protocols that included measurements of both the near and far wall at multiple angles (24) performed best on reproducibility and providing a large rate of CIMT progression with a high precision.

In conclusion, our findings support the opinion that the number and specific combination of segments, angles, and walls interrogated are associated with differences in reproducibility, magnitude, and precision of progression of CIMT over time. Given these findings, the choice for an ultrasound protocol should depend on a well considered balance of these parameters before the study is carried out.

Trial registration

RADIANCE 1 ClinicalTrials.gov number: NCT00 136981; RADIANCE 2 ClinicalTrials.gov number: NCT00134264.

Acknowledgements

The authors thank Rudy Meijer in Schoorl, the Netherlands, for the design of the image-acquisition protocol, training, and quality control; and P. Barter, T. Roberts, D. Ambrose, A. Chin, W. Davidson, R. Burnside, A. Caffrey, M. Li, A. O'Reilly, Tu T. Nguyen, and T. Thuren. We thank the following staff at CIMT Laboratory Europe: C. van Everdingen, A. Geerts, M. Geurtsen, M. Djuanda, A. Kuin, F. Leus, R. Meijer, D. Mooiweer-Bogaerdt, K. Nijssen, H. Noordzij, L. Romkes, B. Sies, E. Stooker, F. Verhey, B. van der Vlist, L. van der Vlist, E. Wineke, and H. Wisse; and at the CIMT Core Laboratory United States: M. Barr, K. Bettermann, S. Burton, A. Conner-Day, B. Ettenger, J. Griffin, C. Halverson, B. Holley, L. Hoots, J. Fleshman, M. Lauffer, L. Passmore, C. Sharpe, M. Wilder, P. Miller, T. Vitek, and G. Wolgast.

Declaration of interest: Dr Dogan and Dr Duivenvoorden report no conflict of interest. Dr Kastelein reports receiving consulting fees and lecture fees from Pfizer, Astra-Zeneca, Merck, and Schering-Plough and grant support from Pfizer and Astra-Zeneca. Dr Shear was a former employee of Pfizer. Dr Grobbee reports receiving consulting fees, lecture fees, and grant support from Pfizer, Astra-Zeneca, Servier, Organon, Merck, and Unilever. Dr Evans reports receiving honoraria, consulting fees, and grant support for professional input on CIMT issues from Astra-Zeneca, Organon, and Pfizer. Dr Visseren reports receiving research grants from Merck and the Netherlands Organisation for Health Research and Development. Dr Bots reports receiving consulting fees, lecture fees, and grant support from Pfizer, Astra-Zeneca, Servier, Organon, Merck, and Unilever.

No other potential conflict of interest relevant to this article was reported.