Intraobserver and interobserver variability of 64-row computed tomography abdominal aortic aneurysm neck measurements

Methods 

Twenty-five consecutive patients (23 men and 2 women; mean age, 75.5 ± 6.25 years) undergoing 64-row follow-up CTA after EVAR were prospectively studied. The following endografts had been used to treat AAA with a mean maximal diameter of 55.13 ± 11.84 mm (range, 42-79 mm): Vanguard (Boston Scientific, Oakland, NJ), n = 1; Excluder (WL Gore, Flagstaff, Ariz), n = 1; Ancure (Guidant, Indianapolis, Ind), n = 2; Aorfix (Lombard Medical, Oxfordshire, UK), n = 2; Anaconda (Sulzer Vascutek, Hamburg, Germany), n = 3; and Talent (Medtronic, Sunrise, Fla), n = 16. Ethics committee approval was not obtained because CTA scans were performed on a regular clinical basis.

CTA measurements 

CTA was performed on a 64-channel multislice computed tomography (CT) scanner (Somatom Sensation Cardiac 64; Siemens, Forchheim, Germany). Data acquisition parameters were 100 kVp, 320 mA quality reference, a 32 × 1.2-mm collimation, a pitch of 1.25, a reconstruction interval of 1.5 mm, and a 0.5-second gantry rotation speed. A total of 100 mL of nonionic iodinated contrast (350 mg/mL; Ioversol; Guerbet SA, Zurich, Switzerland) was injected at a flow rate of 3 mL/s by a power injector by using real-time bolus-triggering software (C.A.R.E.-Bolus; Siemens). Soft tissue window settings with a width of 600 HU and a center of 100 HU were used.

Aortic neck volume was measured from the raw CTA data set by using Syngo CT 2006A UB 20B software (Siemens). The aortic neck was defined as the nondiseased portion of the infrarenal aorta (a nondilated cylinder beginning distally to the most caudal renal artery and extending to the first CTA slice showing a ≥15% larger aortic wall diameter as compared with the diameter measured directly below the lowermost renal artery; Fig 1). To standardize measurements, the exact extent of the infrarenal aneurysm neck was predefined in consensus by two experienced readers (F.D. and N.D.) providing start and stop numbers of CTA slices for all readers assessing aortic morphometry. Aortic neck volumetry then was performed semiautomatically by each reader by manually segregating each single consecutive axial CTA slice strictly between start and stop slices and calculated by the above-mentioned software (Fig 1). Measurement of AAA neck morphology yielded neck volume (given in milliliters) and aortic neck diameter measured from outer adventitia to outer adventitia (given in millimeters) as the smallest (D2amin) and the largest diameter (D2amax) on axial CTA slices on the first slice distal to the lowermost renal artery and 10 mm below this landmark (D2bmin, D2bmax).

View Large Image Download to PowerPoint

Fig 1. a, Infrarenal aortic aneurysm neck. RRA, Right renal artery; N, extent of the aortic neck on the first transverse computed tomographic slice distal to the RRA. b, Infrarenal aortic aneurysm neck. N, Extent of the aortic neck on the last transverse computed tomographic slice of the neck segment defined as the infrarenal neck by consensus agreement of two experienced readers. c, Infrarenal aortic aneurysm neck. N, Extent of the aortic neck on two-dimensional reconstruction of the infrarenal aorta.

Results 

Detailed information on AAA neck dimensions is given in Table I.

Table I. Morphologic characteristics of proximal aortic necks in this series (n = 25)

	Neck dimension	Data
	Neck length (mm)	31.9±20.4(range,18.3-75)
	D2amin (mm)	24.6±3.7(range,18.6-33.4)
	D2amax (mm)	28.2±5.0(range,19.8-39.7)
	D2bmin (mm)	25.1±3.5(range,17.8-34.2)
	D2bmax (mm)	28.6±4.3(range,21.3-38.4)
	NeckVol (mL)	23.4±9.9(range,10.9-50.7)
	Neck angulation 0°-15°	n=19(76%)
	Neck angulation 16°-30°	n=5(20%)
	Neck angulation 31°-45°	n=1(4%)
	Circumferential thrombus 0%-25%	n=3(12%)
	Circumferential thrombus 26%-50%	n=1(4%)
	Circumferential thrombus >50%	n=0(0%)
	Tapered neck	n=3(12%)
	Reverse tapered neck	n=1(4%)

D2amin, Smallest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2amax, largest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2bmin, smallest diameter on the axial slice 10 mm distal to the lowermost renal artery; D2bmax, largest diameter on the axial slice 10 mm distal to the lowermost renal artery; NeckVol, neck volume.

Intraobserver variability of CTA volume and diameter measurements 

Intraobserver agreement was high for dedicated neck volumetry by all readers, with individual mean differences of less than 1 mL (P > .05 by t test; Table II, Table III, Table IV; Fig 2). However, intraobserver variability for neck diameter measurement was significant for D2amin, D2bmin, and D2bmax in reader 2 as well as for D2amin in reader 3 (P < .05 by t test; Table II, Table III, Table IV). Intraobserver variability did not differ among different endografts (P > .05 by t test).

Table II. Intraobserver differences for repeated CTA measurements in 25 patients for reader 1

	Variable	Difference	SD	Limits of agreement	P values⁎
	NeckVol	−0.58	1.48	−1.19to0.03	.062
	D2amin	−0.06	0.22	−0.15to0.04	.225
	D2amax	−0.06	0.26	−0.17to0.04	.230
	D2bmin	−0.008	0.24	−0.11to0.09	.863
	D2bmax	0.006	0.28	−0.11to0.12	.917

D2amin, Smallest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2amax, largest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2bmin, smallest diameter on the axial slice 10 mm distal to the lowermost renal artery; D2bmax, largest diameter on the axial slice 10 mm distal to the lowermost renal artery; NeckVol, neck volume.

⁎ Paired two-sided t test.

Table III. Intraobserver differences for repeated CTA measurements in 25 patients for reader 2

	Variable	Difference	SD	Limits of agreement	P value⁎
	NeckVol	0.25	2.67	−0.86to1.35	.645
	D2amin	0.58	1.36	0.02to1.15	.043
	D2amax	−0.87	7.32	−3.89to2.16	.559
	D2bmin	0.77	1.84	0.01to1.53	.047
	D2bmax	0.82	1.21	0.32to1.32	.002

D2amin, Smallest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2amax, largest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2bmin, smallest diameter on the axial slice 10 mm distal to the lowermost renal artery; D2bmax, largest diameter on the axial slice 10 mm distal to the lowermost renal artery; NeckVol, neck volume.

⁎ Paired two-sided t test.

Table IV. Intraobserver differences for repeated CTA measurements in 25 patients for reader 3

	Variable	Difference	SD	Limits of agreement	P value⁎
	NeckVol	−0.19	0.58	−0.44to0.05	.110
	D2amin	−1.24	1.90	−2.03to−0.45	.003
	D2amax	0.32	2.50	−0.71to1.36	.524
	D2bmin	−0.51	1.46	1.12to0.09	.093
	D2bmax	−0.24	1.00	−0.65to0.17	.242

D2amin, Smallest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2amax, largest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2bmin, smallest diameter on the axial slice 10 mm distal to the lowermost renal artery; D2bmax, largest diameter on the axial slice 10 mm distal to the lowermost renal artery; NeckVol, neck volume.

⁎ Paired two-sided t test.

View Large Image Download to PowerPoint

Fig 2. a, Plot of intraobserver differences against the average volume of the aortic neck for reader 1 according to Bland and Altman.22 b, Plot of intraobserver differences against the average volume of the aortic neck for reader 2. c, Plot of intraobserver differences against the average volume of the aortic neck for reader 3.

Interobserver variability of CTA volume and diameter measurements 

The distribution of absolute volumes measured by all readers is shown in Fig 3. Neck volumetry was associated with significant interobserver variability in both readings (P < .005 by t test; Table V, Table VI) except for measurements comparing reader 1 with reader 2. In almost the same manner, aortic neck diameter measurement was associated with statistically relevant interobserver variability (P < .015 by t test; Table V, Table VI) except for single measurements comparing results from readers 1 and 3. Interobserver variability did not differ among different endografts (P > .05 by t test).

View Large Image Download to PowerPoint

Fig 3. Distribution of aortic neck volume measurement results as measured by three blinded and independent readers.

Table V. Interobserver differences for CTA measurements in 25 patients: first reading

	Variable	Difference	SD	Limits of agreement	P value⁎
	Reader 1 vs reader 2
	NeckVol	0.84	3.39	−0.55to2.24	.225
	D2amin	1.42	2.62	0.34to2.50	.012
	D2amax	2.09	3.21	0.77to3.42	.003
	D2bmin	1.57	3.19	0.25to2.89	.022
	D2bmax	2.50	2.33	1.53to3.50	.000
	Reader 1 vs reader 3
	NeckVol	−1.53	2.49	−2.56to−0.50	.005
	D2amin	0.34	2.93	−0.87to1.55	.567
	D2amax	−0.17	2.48	−1.19to0.86	.738
	D2bmin	0.020	2.52	−1.02to1.06	.969
	D2bmax	1.22	2.32	0.26to2.18	.015
	Reader 2 vs reader 3
	NeckVol	−2.38	2.83	−3.55to−1.21	.000
	D2amin	−1.08	1.98	−1.90to−0.26	.012
	D2amax	−2.26	2.76	−3.40to−1.12	.000
	D2bmin	−1.55	1.75	−2.27to−0.82	.000
	D2bmax	−1.28	1.51	−1.90to−0.65	.000

D2amin, Smallest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2amax, largest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2bmin, smallest diameter on the axial slice 10 mm distal to the lowermost renal artery; D2bmax, largest diameter on the axial slice 10 mm distal to the lowermost renal artery; NeckVol, neck volume.

⁎ Paired two-sided t test.

Table VI. Interobserver differences for CTA measurements in 25 patients: second reading

	Variable	Difference	SD	Limits of agreement	P value⁎
	Reader 1 vs reader 2
	NeckVol	1.67	2.20	0.76to2.58	.001
	D2amin	2.56	2.74	1.43to3.69	<.001
	D2amax	3.31	3.89	1.70to4.91	<.001
	D2bmin	2.42	2.55	1.37to3.47	<.001
	D2bmax	3.26	2.88	2.07to4.45	<.001
	Reader 1 vs reader 3
	NeckVol	−1.15	2.10	−1.99to−0.30	.010
	D2amin	−0.35	2.74	−1.48to0.78	.531
	D2amax	0.80	3.91	−0.81to2.41	.317
	D2bmin	−0.41	2.61	−1.48to0.67	.441
	D2bmax	0.92	2.81	−0.24to2.08	.115
	Reader 2 vs reader 3
	NeckVol	−2.82	1.60	−3.48to−2.16	<.001
	D2amin	−2.90	1.79	−3.65to−2.16	<.001
	D2amax	−2.51	1.68	−3.20to−1.81	<.001
	D2bmin	−2.83	2.00	−3.65to−2.00	<.001
	D2bmax	−2.34	1.31	−2.88to−1.80	<.001

D2amin, Smallest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2amax, largest diameter on the axial slice on the first slice distal to the lowermost renal artery; D2bmin, smallest diameter on the axial slice 10 mm distal to the lowermost renal artery; D2bmax, largest diameter on the axial slice 10 mm distal to the lowermost renal artery; NeckVol, neck volume.

⁎ Paired two-sided t test.

Discussion 

The introduction of multidetector row CTA technology has greatly advanced its role in diagnostic imaging in patients with vascular diseases. Thus, CTA has become a key imaging modality in patients with AAA, and 64-row CT technology promises to further improve image quality, because high precision and reliability of volumetric measurements are important prerequisites for the detection of AAA neck growth after EVAR.

It is agreed upon that volumetry has a higher predictive accuracy than maximum diameter measurement in assessment of total AAA size changes after stent-graft insertion.18, 19, 20 However, dedicated aortic neck volumetry involves substantially smaller values than total AAA volumetry, which is usually measured from the renal arteries to the orifice of the hypogastric artery. Therefore, it might bear higher susceptibility to observer variability. However, it has recently been shown that small lung nodules with volumes as low as 7.3 mm3 can be measured with high precision by using 4-row CT technology.22

To our knowledge, observer variability of aortic neck measurements with 64-row CTA has not been assessed. Our hypothesis was that aortic neck volumes can be reliably measured in vivo by using state-of-the-art three-dimensional CTA imaging allowing for maximum spatial resolution. Indeed, we have shown that, even though comparatively small volumes were measured, intraobserver agreement was excellent for all readers, thus indicating that relative volume changes could be reliably reproduced when assessed by a single reader under study conditions. This observation suggests that serial AAA neck measurements as performed on 64-row CTA could be implemented as a valid surrogate end point for assessment of the effect of differentiated medical treatment strategies in the prevention of aortic neck enlargement after EVAR.

Measurement methods were optimized by clear predefinition of aortic neck dimensions before individual analysis, thereby limiting potential observer variability. Despite this effort, interobserver variability was considerable in our series. This finding might be explained by the fact that varying interpretation of vascular morphology in the axial section and surrounding structures by each individual observer may lead to relevant measurement differences. Therefore, we recommend that serial volume measurements for assessment of aortic neck size changes should be performed by a single reader.

According to what has been shown for AAA diameter measurements in other studies applying less elaborate imaging methods,16, 23, 24, 25 neck diameter measurement on axial CTA slices resulted in statistically significant intraobserver variability. The use of 64-row CTA plus dedicated software did not improve the precision of axial AAA diameter measurements.

Using a two-dimensional tool such as diameter measurement to evaluate a three-dimensional object such as an aortic neck is a potential source for observer variation, especially if slices are not analyzed perpendicular to the aortic centerline. Furthermore, in contrast to volumetry, in which vascular structures are marked by the reader and semiautomatically calculated by the CT software, diameter measurement offers various possibilities for measurement variation around the vessel circumference. In the presence of more sophisticated three-dimensional imaging options, we therefore regard axial diameter measurements as not a first-line method to assess for changes in neck morphology after EVAR, especially in studies aiming at the prevention of aortic neck enlargement.

Two shortcomings of this study have to be addressed. The first limitation is the fact that the true volume of aortic necks measured in this study is not known. However, the study was focused on in vivo assessment of precision because this is most important in assessing the reproducibility of this imaging method before it can be recommended as a tool within prospective trials. Therefore, in vitro evaluation of the accuracy of 64-row CTA AAA neck measurements using calibrated phantoms is warranted to confirm its ability to validly quantify volumes of varying anatomy.

Second, because shortcomings of the software interfered with our standardized measurement protocol, we did not perform diameter measurements perpendicular to the centerline of the infrarenal aortic segment. It has been shown by others that the interobserver variability of diameter measurements can be lessened when measurements are performed perpendicular to the aortic centerline.23 However, even with standardized measurements, variability was still considerable, thus indicating that the precision required to reliably assess changes in neck dimensions over time would not be granted by using this measurement approach. Because our series mainly contained patients with favorable aortic neck morphology, we assume that diameter measurements perpendicular to the centerline would not have added substantial improvements.

In conclusion, our prospective study revealed that AAA neck volumetry can be performed with good intraobserver agreement by using software-enhanced 64-row CTA. Studies assessing differentiated therapies aiming at the prevention of aortic neck degeneration should adopt CTA volumetric measurements performed by a single experienced reader and refrain from serial AAA neck diameter measurements.

Author contributions