Observer Variability of Iliac Artery Measurements in Endovascular Repair of Abdominal Aortic Aneurysms

Observer Variability of Iliac Artery Measurements in Endovascular Repair of Abdominal Aortic Aneurysms Brajesh K. Lal, MD,1,2 Joaquim J. Cerveira, MD,1,2 Craig Seidman, MD,1 Paul B. Haser, MD,1,3 Richard Kubicka, MD,1 Zafar Jamil, MD,1,3 Frank T. Padberg, MD,1,2 Robert W. Hobson, MD,1,3 and Peter J. Pappas, MD,1,2 Newark and East Orange, New Jersey Accurate measurement of iliac arteries is essential for successful delivery of aortic endografts without iliac limb endoleak. Although intravascular ultrasound measurements may be reliable, they require an invasive procedure. Therefore, helical computed tomography (hCT) has become the most commonly used modality for obtaining preprocedure arterial diameter measurements. The accuracy of hCT remains ill-defined, however, because an anatomic gold standard with which to compare the measurements is not available. We therefore assessed inter- and intraobserver variability of hCT measurements. We also applied accepted cutoff measurements to determine the clinical impact of observer variability in predicting the need for adjunctive iliac access and iliac limb seal procedures. hCT scans were analyzed in 30 patients who had undergone successful placement of a bifurcated endograft (26 Ancure, 4 Aneurex). Mean age of patients was 75 years, the male/female ratio was 27:3. Three blinded observers measured transverse diameters (maximal aortic aneurysm [Amax], narrowest infrarenal aortic neck [Amin], maximal common iliac [Imax], and narrowest iliac artery [Imin]). Inter- and intraobserver variability was calculated as standard deviation of mean pair differences according to the method of Bland and Altman. The true incidence of adjunctive procedures to facilitate delivery of the device into the aorta and ensure iliac limb seal was compared with that predicted by the observers to obtain sensitivity, specificity, and positive (PPV) and negative predictive value (NPV) for the measurements. Interobserver variability of iliac measurements was higher than intraobserver variability (p < 0.05). Interobserver variability of Amax ranged from 4.37 to 10.73% of the mean Amax. Conversely, variability of Amin was 8.91-18.89%, that of Imax was 12.11-22.23%, and that of Imin was 10.51-18.73% (p < 0.05 vs. Amax). Therefore, interobserver variability influenced aortic neck and iliac diameter twice as much as it did aneurysm measurements. To successfully place 30 endografts we performed 8 adjunctive access procedures (4 angioplasties, 4 common iliac artery conduits) and 17 adjunctive procedures in 60 limbs to ensure limb seal (9 unilateral IIA coil embolizations, 8 stents). We used 8.5 (Ancure) and 8.0 (Aneurex) mm as lower limits of acceptability for uncomplicated access, and 13.4 (Ancure) and 16 (Aneurex) mm as the upper limits of acceptability for uncomplicated iliac limb seal. These limits were applied to measurements from the three observers to predict need for adjunctive access or iliac seal procedures in this cohort. Sensitivity, specificity, PPV, and NPV of these observer measurements for a need to perform additional access procedures were 0.67, 0.80, 0.55, and 0.87; the same values for a need to perform additional seal procedures were 0.71, 0.74, 0.52, and 0.86, respectively. Interobserver variability was approximately 20% of measured iliac diameter. 1 Division of Vascular Surgery, Department of Surgery, University of Medicine and Dentistry of New Jersey–New Jersey Medical School, Newark, NJ, USA. 2 Section of Vascular Surgery, Veterans Administration, New Jersey Healthcare System, East Orange, NJ, USA. 3 Section of Vascular Surgery, Saint Michael’s Medical Center, Newark, NJ, USA. Presented at the 13th Annual Winter Meeting of the Peripheral Vascular Surgery Society, Steamboat Springs, CO, January 31-February 2, 2004. 644 Correspondence to: Brajesh K. Lal, MD, Department of Surgery, Division of Vascular Surgery, University of Medicine and Dentistry of New Jersey–New Jersey Medical School, 185 South Orange Avenue, MSB H578, Newark, NJ 07103, USA, E-mail: lalbk@umdnj. edu Ann Vasc Surg 2004; 18: 644-652 DOI: 10.1007/s10016-004-0102-x Ó Annals of Vascular Surgery Inc. Published online: 6 October 2004 Vol. 18, No. 6, 2004 Observer variability in iliac measurements for AAA repair 645 This explains why helical CT measurements were noted to have low PPV in predicting the need for an adjunctive access or limb seal procedure. These data establish PPV and NPV for hCT and provide objective evidence for the need to improve iliac artery imaging. Until more accurate imaging becomes available, we recommend oversizing of iliac limbs by 10-20% in patients with wide landing zones and that surgeons be prepared to resolve unexpected iliac artery access or seal problems intraoperatively. INTRODUCTION METHODS Improved imaging technology has played a central role in the development and success of endografting as a viable alternative to surgical repair in patients with infrarenal abdominal aortic aneurysm (AAA). Appropriate patient selection and endograft sizing are important determinants of procedural outcome. Aortic aneurysm measurements made on computerized tomography (CT) images are consistently less variable and more accurate than physical examination1 or duplex ultrasonography.2,3 Patient and endograft selection have therefore been made on the basis of arterial measurements of CT scans. The aortic aneurysm diameter is an important determinant for making a decision on whether to treat AAA. The aortic neck diameter is an important determinant for the type of repair to be offered (open vs. endograft) as well as for type I endoleaks and graft migration at the proximal attachment site. There is increasing awareness that iliac artery diameter is an important determinant of successful delivery of the device into the aorta, as well as for successful deployment and distal attachment of the device without a type I endoleak from an iliac limb. The variability in ultrasound determined aortic measurements has been extensively documented.2,4-6 However, few studies have evaluated CT-determined aortic measurements.2,3,7,8 Information on variability of iliac artery measurements is even more limited and is primarily based on conventional CT alone as opposed to helical CT (hCT) technology.7,8 The clinical impact of observer variability in iliac artery measurements remains difficult to ascertain and has not been reported. Incorrect sizing of aortic or iliac landing zones can result in endoleaks, limb occlusions, and conversion to open procedures. Therefore, it is critically important that preoperative diameter measurements be accurate to avoid complications and predict the need for possible adjunctive procedures. We therefore assessed intra- and interobserver variability of aortoiliac measurements made from hCT images by three independent blinded clinicians and evaluated their ability to preoperatively predict successful negotiation and delivery of endograft devices without iliac limb leak. Patients All patients undergoing aortic aneurysm endograft repair were prospectively entered into a patient database and followed postoperatively. Thirty consecutive patients undergoing endograft repair of aortoiliac aneurysms in a single program from January 2001 to December 2001 were included in this study. Patients with isolated iliac artery aneurysms were excluded. Demographic data pertaining to age, genders, and comorbidities were recorded. All preprocedure hCT images and data on type of endografts used were analyzed. Procedural details including adjunctive maneuvers used to deliver the device through the iliac arteries, distal attachment site type I endoleaks, and maneuvers required to control them were prospectively recorded. Patients were followed for 1 year to determine the incidence of postprocedural iliac limb complications. The follow-up protocol involved clinical vascular examination, ankle/brachial index, plain films of the abdomen, duplex ultrasonography for aneurysm sac flow and iliac limb patency, and a contrast-enhanced hCT scan. Helical Computerized Tomography Scans CT scans were obtained according to a standardized protocol. All scans were performed using a helical machine (Siemens Plus 4, New York, NY) with no gantry angulation, 3 mm collimation, and 2:1 pitch. Scanning was performed without and with nonionic contrast for a volume of 130 mL injected intravenously at 3 mL/sec and a delay of 25 sec. Scanning commenced at the level of the celiac axis and continued to 1 cm below the ischial tuberosity. Images were reconstructed and printed with 3 mm axial spacing in a 12 images-per-page format, representing the most frequently used methodology in the community. Measurement Technique Three observers blinded to patient clinical history and procedural outcome measured the following structures in hCT images: widest aortic aneurysm diameter (Amax), narrowest infrarenal aortic neck 646 Lal et al. diameter (Amin), widest common iliac artery diameter (Imax), and narrowest diameter along the length of the iliac artery (Imin). Observers were required to select the axial image of their choice to perform measurements and were blinded to the measurements made by the other reviewers. One observer repeated the same measurements in a blinded fashion after a one-month interval; he had no access to his previous readings. All measurements were made from hCT images printed with the same magnification, using the same view box and illumination, with an electronic caliper. Diameters were registered to one decimal place in millimeters, the standard output of the caliper. Aortic aneurysm (Amax) diameter was determined by measuring the outer wall diameter at the widest point in any plane; all other arterial diameters (Amin, Imax, Imin) were determined by measuring the lumen. Diameter measurements were made perpendicular to the direction of tortuosity (the narrowest axis) in tortuous arteries to correct for oblique axial hCT slices. Indirect measurements off reconstructed images in three dimensions were not used for the study. Statistical Analysis To assess intraobserver variability we calculated the differences between paired measurements of each diameter made by one observer at two different occasions performed one month apart. Variability was expressed as standard deviation (SD) of mean pair differences. Graphical illustration of the mean differences of the paired observations was made according to the method described by Bland and Altman.9 Possible systematic bias between the two measurements of the observer was assessed by determining whether the mean differences were significantly different from zero. The same analyses were used to determine interobserver variability of measurements performed by 3 different observers. To evaluate the clinical importance of the observed variability in measurements, three analyses were performed. First, to determine the magnitude of the problem, the frequency with which differences between measurements exceeded 1, 2, and 3 mm between observations was calculated as a percentage of all observations. Second, to determine the relative importance of the problem, variability was calculated as a percentage of the mean diameter being measured. Finally, to determine how variability influenced the predicted outcome in patients, the incidence of actual adjunctive access procedures and iliac limb endoleaks was determined and compared with those predicted by Annals of Vascular Surgery observers, based on their measurements. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the measurements to predict these adjunctive procedures were then calculated. RESULTS Patients Thirty consecutive endovascular endograft procedures were performed to treat aortoiliac aneurysms in a single program over a 1-year period. These patients were entered into the study. Preprocedure hCT images were available for all patients. Mean age of the patients was 75 years. Twenty three (77%) patients were males, 26 (81%) had hypertension, 10 (31%) had diabetes, 24 (75%) had coronary artery disease, 18 (56%) had hypercholesterolemia, 5 (16%) had chronic obstructive airway disease, and 5 (22%) were currently smoking. Intraobserver Variability Thirty paired measurements were performed by the same observer on two separate occasions for each of the following: maximum aortic aneurysm diameter (Amax) and the least aortic neck diameter (Amin) for a total of 120 measurements (Table I). The mean Amax was 58.05 mm, and intraobserver variability for Amax measured as the SD of mean differences in the paired measurements was 1.84 mm. The mean Amin was 22.81 mm with an intraobserver variability of 0.73 mm. Sixty paired measurements were performed on two separate occasions by the same observer for each of the following: widest common iliac (Imax) and narrowest iliac artery (Imin) diameters for a total of 240 measurements (Table I). The mean Imax was 15.7 mm and the intraobserver variability was 1.77 mm. The mean Imin was 9.2 mm with an intraobserver variability of 0.77 mm. The mean differences of all observation pairs were not significantly different from zero (p > 0.05 across measurements of Amax, Amin, Imax, and Imin), indicating that there was no observer bias. Figure 1 demonstrates plots of the intraobserver variability of measurements for Amax, Amin, Imax, and Imin according to the method described by Bland and Altman.9 Interobserver Variability Thirty measurements each were performed by three separate observers for both the Amax and Amin for a total of 180 measurements (Table II). The interobserver variability of measurements for Vol. 18, No. 6, 2004 Observer variability in iliac measurements for AAA repair 647 Table I. Intraobserver differences and variability with hCT measurements Object measured Paired observations (n) Mean difference (mm) p Standard deviation (mm) Aortic aneurysm Aortic neck Widest common iliac artery Narrowest iliac artery 30 30 60 60 0.44 0.17 0.04 –0.25 0.2 0.21 0.86 0.07 1.84 0.73 1.77 0.77 The mean difference of paired observations was not significantly different from zero (p > 0.05) for all four measurements, indicating that there was no observer bias. Fig. 1. Intraobserver variability of hCT scan measurements for widest aortic aneurysm (A), narrowest infrarenal aortic neck (B), widest common iliac artery (C), and narrowest iliac artery (D) diameters. Solid line de- notes the mean difference of observations; dotted line denotes 2 standard deviations (SD) above and below that measurement. The majority of values fall within 2 SD of the mean difference. Amax by the three observers, measured as the SD of mean differences in the paired measurements, ranged from 2.49 to 6.11 mm. Interobserver variability for Amin ranged from 1.45 to 2.33 mm. Sixty measurements each were performed by three separate observers for both the Imax and Imin artery diameters for a total of 360 measurements (Table II). The interobserver variability for Imax was 2.06-3.91 mm, and for Imin it was 0.96-1.71 mm. Absolute variability for Amax was significantly higher than for Amin, Imax or Imin (p < 0.05). There was no observer bias across the measurements (mean differences between observations for all measurements were not significantly differ- ent from zero, p > 0.05). Figure 2 demonstrates plots of the interobserver variability in the measurements for Amax, Amin, Imax, and Imin according to the method described by Bland and Altman.9 Interobserver Variability as a Percent of Observed Diameter To determine the relative magnitude of the problem, interobserver variability was calculated as a percent of mean diameter of each structure measured. It was noted that interobserver variability of Amax ranged from 4.4 to 10.7% of the mean 648 Lal et al. Annals of Vascular Surgery Table II. Interobserver differences and variability with hCT measurements Observer 1 vs. 2 Object measured Aortic aneurysm Aortic neck Widest common iliac artery Narrowest iliac artery Observer 1 vs. 3 p Standard deviation (mm) Mean difference (mm) p Standard deviation (mm) 3.01 –1.49 –0.16 0.06 0.06 0.76 3.76 2.07 3.91 1.99 –1.63 –0.21 0.09 0.06 0.67 6.11 2.33 3.78 0.04 0.86 1.71 0.36 0.07 1.50 p Standard deviation (mm) Mean difference (mm) 1.02 0.14 0.05 0.07 0.60 0.84 2.49 1.45 2.06 –0.32 0.06 0.96 Mean difference (mm) Observer 2 vs. 3 The mean difference of paired observations was not significantly different from zero (p > 0.05) for all measurements, indicating that there was no observer bias. Amax. However, variability of Amin was 8.918.9% of the mean Amin, that for Imax was 12.122.2%, and for Imin was 10.5-18.7%. Therefore, even though absolute interobserver variability was highest for Amax (2.49-6.11 mm, Table II), variability influenced Amin, Imax, and Imin twice as much as the Amax when calculated as a percentage of the mean diameter of the structure measured (p < 0.05 vs. Amax). Stratification of Intraobserver and Interobserver Variability The differences between paired observations for each measurement were then stratified according to their magnitude. Table III demonstrates stratified absolute intraobserver variability. Amax measurements varied by >2 mm in 46.7% of paired observations; however, Amin measurements did so in only 3.3%; Imax differed by >2 mm in 15%, and Imin in 8.3% of paired measurements. The impact of interobserver variability was also assessed by similar stratification (Table IV). There was a >2 mm difference between the three observers in 40-70% of Amax observations; 19.4-25.5% of Amin observations; 16.7-25% of Imax observations (p < 0.05 vs. Amax); and 6.7-23.3% of Imin observations (p < 0.01 vs. Amax). Sensitivity, Specificity, and Negative and Positive Predictive Values of Iliac Measurements to Predict Complicated Access and Seal To further assess the clinical impact of interobserver variability on iliac artery diameter measurements, we compared the predicted versus observed incidence of complex iliac artery access procedures (artery too narrow for passage of the device) and complex iliac limb seal maneuvers (artery too wide for the selected device limb diameter). To limit observer variability as the only variable to be studied, we imposed the following criteria on measurements from all observers (90 aortic, 180 iliac) according to the actual endograft used. The following cutoffs were used to predict the need for additional procedures to facilitate passage of the device: <8.5 mm (for patients receiving the Ancure device, Guidant Corp., Menlo Park, CA) and <8.0 mm (for those receiving the AneuRx device, Medtronic Corp., Santa Rosa, CA). Similarly, the following cutoffs were used to predict the need for additional procedures to facilitate distal limb seal: >13.4 mm (Ancure) and 16 mm (AneuRx). These cutoffs were applied according to the manufacturer’s recommendations at the time the endografts were placed. The limits were applied to measurements from all three observers to obtain predicted complicated access and seal events in these patients. During the placement of 30 bifurcated endografts (21 Ancure, and 9 Aneurex) we encountered access difficulties in 8 instances (4 required angioplasty and 4 required common iliac artery prosthetic conduits to deliver the device). We also performed additional maneuvers to obtain adequate iliac limb seal in 17 of 60 limbs (9 unilateral internal iliac artery coil embolizations, 8 extender grafts/stents). The predicted incidence of complex iliac artery access and seal maneuvers based on the measurements obtained from the three observers during this study was compared to the true incidence. This yielded the sensitivity, specificity, PPV, and NPV of these measurements to predict such events (Table V). DISCUSSION Our study noted lower intraobserver than interobserver variability. This is consistent with results from published literature, for any measurement. It Vol. 18, No. 6, 2004 Observer variability in iliac measurements for AAA repair 649 Fig. 2. Pooled results of interobserver variability of hCT scan measurements for widest aortic aneurysm (A), widest infrarenal aortic neck (B), widest common iliac artery (C), and narrowest iliac artery (D) diameters. The majority of values fall within 2 SD of the mean differences between all observer pairs. also reflects the increased consistency of image selection and measurement technique that occurs when the same observer repeats the measurement compared with the consistency of different observers. Most published studies have limited their analysis of observer variability to aortic aneurysm measurements. Lederle et al.3 noted that interobserver variability in aneurysm measurements could be as high as 5 mm or more. In another study of 19 patients, Jaakkola et al.2 noted interobserver variability of aneurysm measurements to be 6.9 mm. Using hCT technology, our study showed a lower interobserver variability, ranging from 2.5 to 6.1 mm. Two studies with 107 and 298 patients have measured variability in aortic neck measurements ranging from 3.9 to 7.8 mm. In one of these studies the axial slice to be measured was predetermined. In our study, we required that physicians select the image for measurement, thereby eliminating any risk of selection bias. However, we still obtained lower interobserver variability, 1.5-2.3 mm, than previously reported. Table III. Impact of intraobserver variability Object measured Aortic aneurysm Aortic neck Widest common iliac artery Narrowest iliac artery Difference between paired observations (mm) n (%) >1 >2 >3 >1 >2 >3 >1 19 14 2 6 1 0 22 (63.3) (46.7) (6.7) (20) (3.3) (0) (36.7) >2 >3 >1 >2 >3 9 4 14 5 1 (15) (6.7) (23.3) (8.3) (1.7) Only two previous studies have attempted to measure variability in iliac artery measurements. They reported interobserver variabilities ranging from 3.9 to 4.6 mm.7,8 The small diameters of iliac 650 Lal et al. Annals of Vascular Surgery Table IV. Impact of interobserver variability Object measured Aortic aneurysm Aortic neck Widest common iliac artery Narrowest iliac artery Observer 1 vs. 3 Observer 2 vs. 3 n (%) n (%) n (%) >1 >2 >3 >1 >2 >3 >1 >2 >3 >1 >2 >3 16 12 4 17 6 2 25 10 7 16 4 2 (53.3) (40) (13.3) (54.8) (19.4) (6.5) (41.7) (16.7) (11.7) (26.7) (6.7) (3.3) 24 21 18 22 11 8 43 25 13 36 14 5 (80) (70) (60) (70.9) (25.5) (25.8) (71.7) (21.7) (21.7) (60) (23.3) (8.3) 26 21 16 25 16 9 40 27 18 34 15 2 (86.7) (70) (53.3) (83.3) (23.3) (30) (66.7) (25) (30) (56.7) (15) (3.3) Table V. Sensitivity, specificity, PPV, and NPV of iliac artery diameter measurements to predict need for complex iliac limb procedures Complicated access Iliac limb endoleak Observer 1 vs. 2 Difference between paired observations (mm) Sensitivity Specificity PPV NPV 0.67 0.71 0.8 0.74 0.55 0.52 0.87 0.86 arteries and their increased tortuosity in multiple planes have the potential to cause increased variability. However, the two studies7,8 did not differentiate between measurements made on ectatic or stenotic iliac artery segments, and one study preselected the axial image in which to take measurements. Our investigation tested measurements that have improved clinical utility. We measured the widest and narrowest iliac artery segments, thereby testing variability to different-sized structures separately. Each observer was required to select the image slice to be used for the measurement, which is more representative of clinical practice and eliminates selection bias. All images were obtained by current hCT technology. We noted iliac interobserver variability to be lower than previously reported. In addition, the size of the target artery influenced variability, as variability for Imax was higher (2.06-3.91 mm) than that for Imin (0.961.71 mm). The lower interobserver variability for aortic and iliac measurements noted in our study may be the result of routine use of electronic calipers, a standardized method of measurement, and a standardized definition of luminal diameter. It may also relate to the use of hCT scanning technology with 3-mm slices. Previous studies have only used conventional CT,2,3,7,8 with one using 10-mm slice thickness.8 Their results are not applicable to the new-generation hCT scanners. The most recent study recognized this as a shortcoming and encouraged future studies with hCT images.8 Previous variability assessments have not been accompanied by evaluations of their clinical impact. We used three measures to make this assessment. First, we determined the magnitude of the problem by calculating the percentage of paired readings with differences exceeding 1, 2, and 3 mm. Second, we determined the relative importance of the problem by calculating variability as a percentage of the mean diameter being measured. Finally, we measured sensitivity, specificity, NPV, and PPV of the iliac measurements to predict the need for additional procedures to ensure device delivery and distal seal. Forty to 70% of paired measurements for Amax (Table II) varied by >2 mm (Table IV). This could adversely influence decisions on whether to treat patients with aneurysms. Similar stratification of iliac variability demonstrated that only 16.7-25% of Imax observations (p < 0.05 vs. Amax) and 6.725% of Imin observations (p < 0.01 vs. Amax) varied by >2 mm. Therefore, unlike aneurysm measurements, large variations in iliac measurements occur less frequently. However, when variability was calculated as a percentage of the target vessel diameter, Amax measurements varied by only 4.4-10.7% of the aneurysm diameter, compared to 8.9-18.9% of the aortic neck diameter (p < 0.05). Therefore, the clinical impact of percent variability in Amin is more important than that in Amax. Most clinicians and U.S. Food and Drug Vol. 18, No. 6, 2004 Administration (FDA)-approved device manufacturers recommend oversizing the endograft by 2-4 mm (10-20%) in relation to the measured neck diameter.10-12 Oversizing has been advocated on the basis of observation of neck dilation over time,13-17 as well as the suspected inaccuracy of neck measurements. Our results indicate that oversizing by 2-4 mm (10-20%) is strongly indicated because of variability in aortic neck measurements. In our study, 16.7-25% of Imax measurements varied by >2 mm. Additionally, percent variability of Imax was 12.1-22.2% of mean Imax, which was significantly higher than that for Amax (p < 0.05). Our results for interobserver variability suggest a need to routinely oversize wide distal landing zone iliac limbs by 2-3 mm or 10-20%. One previous study has noted a progressive decrease in variability as the diameter of the structure measured decreased.8 We noted a different trend in our study. Variability was higher for Imax (2.1-3.9 mm) than for the aortic neck (1.5-2.3 mm), even though the mean diameter for Imax (13.89 mm) was less than that for Amin (23.24 mm). This is probably due to the consistency with which an axial image for Amin can be selected by observers. The iliac artery is usually more tortuous and calcified, and the aortic bifurcation may be less distinct than the aortic neck. Of additional concern was our finding that measurements for the narrowest iliac artery diameter varied by >2 mm in 6.7-25% of instances. Diameters of current or previously FDA-approved endograft devices10-12 in the United States range from 18 to 23.5 Fr. They are best deployed through a guide sheath, which could range from 22 to 25 Fr. A variation of 2 mm in diameter measurement of Imin could alter the device of choice for the patient or the need for an adjunct angioplasty or prosthetic conduit to deliver the device. It could also lead to an inappropriate decision to abandon endografting in an otherwise appropriate candidate or vice versa. The influence of observer variability on the ability to predict the need for adjunctive procedures to deliver the device through the iliac artery or to achieve adequate iliac limb seal has been ill defined. We determined the true incidence of the above-defined adjunctive procedures in our cohort of 30 patients. We then analyzed the measurements made by our observers by applying cutoff measurements that were accepted at the time of implantation of the endograft and based on individual device manufacturer recommendations.10,11 This was used to predict the number of adjunctive procedures that may be required on the basis of Observer variability in iliac measurements for AAA repair 651 observer readings. When the actual and predicted incidences were compared, we noted acceptable specificity (0.80) and NPV (0.87) to predict a narrowed iliac artery requiring additional dilation or conduit procedure. We also noted acceptable NPV (0.86) for predicting a dilated iliac artery necessitating additional procedures or up-sizing to attain a seal. The sensitivity and PPV were uniformly low for predicting an artery that was too narrow (0.67, 0.55, respectively) or too wide (0.71, 0.52, respectively) for a device. These findings underline the need for improvement in imaging technology to assess the adequacy of iliac arteries for endografting. There are several potential options to improve the accuracy of measurements. It is possible that smaller collimation widths and thinner reconstruction slices of 1 or 2 mm would have made a difference in the findings. However, this would generate several-fold more images requiring multiple measurements, which may not be a practical solution unless the process could be automated by new technology. Measurements made at the workstation on further magnified images with digital calipers may also decrease variability; however, the advantages of magnification would be limited by image resolution that could result in large but hazy anatomical landmarks. Additionally, measurements would still need to be performed manually. Currently available three-dimensional displays based on maximum intensity projections (MIP) and surface-shaded display (SSD) or other signal processing maneuvers segment and reformat axial CT images. This process must also be done manually, thereby introducing additional variables. There is also potential loss of data and resolution in this process, and the measurements still need to be made manually. The ideal solution may involve the development of an automated, observer-independent, analytic method utilizing some form of edge identification that could analyze reformatted highresolution hCT scans in three-dimensional projections. CONCLUSIONS Interobserver variability of aortic and iliac measurements from hCT was consistently higher than intraobserver variability. Standardized measurement techniques in which hCT images were used reduced variability of measurements, compared with previously published results with conventional CT. Variability was influenced by the size of the artery measured and also by the anatomic location. Absolute interobserver variability was 652 Lal et al. highest for aneurysm measurements. However, variability was approximately 20% of measured aortic neck and iliac diameters, and only half as much for aneurysm diameter. This explains why hCT measurements were noted to have low PPV in predicting the need for an adjunctive access or limb seal procedure. These data establish variability statistics, PPV, and NPV for hCT and provide objective evidence for the need to improve iliac artery imaging. Until more accurate imaging becomes available, we recommend oversizing of iliac limbs by 10-20% in patients with wide landing zones and that surgeons be prepared to resolve unexpected iliac artery access or seal problems intraoperatively. Annals of Vascular Surgery 6. 7. 8. 9. 10. 11. REFERENCES 12. 1. Fink HA, Lederle FA, Roth CS, Bowles CA, Nelson DB, Haas MA. The accuracy of physical examination to detect abdominal aortic aneurysm. Arch. Intern. Med. 2000;160. 833-836. 2. Jaakkola P, Hippelainen M, Farin P, Rytkonen H, Kainulainen S, Partanen K. Interobserver variability in measuring the dimensions of the abdominal aorta: comparison of ultrasound and computed tomography. Eur. J. Vasc. Endovasc. Surg. 1996;12:230-237. 3. Lederle FA, Wilson SE, Johnson GR, et al. Variability in measurement of abdominal aortic aneurysms. Abdominal Aortic Aneurysm Detection and Management Veterans Administration Cooperative Study Group. J. Vasc. Surg. 1995;21:945-952. 4. Grimshaw GM, Docker MF. Accurate screening for abdominal aortic aneurysm. Clin. Phys. Physiol. Meas. 1992;13:135-138. 5. Singh K, Bonaa KH, Solberg S, Sorlie DG, Bjork L. Intra- and interobserver variability in ultrasound measurements of 13. 14. 15. 16. 17. abdominal aortic diameter. The Tromso Study. Eur. J. Vasc. Endovasc. Surg. 1998;15:497-504. Yucel EK, Fillmore DJ, Knox TA, Waltman AC. Sonographic measurement of abdominal aortic diameter: interobserver variability. J. Ultrasound Med. 1991;10:681-683. Aarts NJ, Schurink GW, Schultze Kool LJ. Abdominal aortic aneurysm measurements for endovascular repair: intra- and interobserver variability of CT measurements. Eur. J. Vasc. Endovasc. Surg. 1999;18:475-480. Singh K, Jacobsen BK, Solberg S, et al. Intra- and interobserver variability in the measurements of abdominal aortic and common iliac artery diameter with computed tomography. The Tromso study. Eur. J. Vasc. Endovasc. Surg. 2003;25:399407. Bland M, Altman D. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-310. Medtronic AVE. Instructions for use. AneuRx Stent Graft: Physician Training Manual. 2003, 7.1. Guidant Corporation. Patient selection and graft sizing. Ancure Endograft System: Physician Training Manual. 2001, II.9. WL Gore & Associates. Patient selection and treatment. Excluder Bifurcated Endoprosthesis. Gore, 2003, 20–24. Lederle FA, Johnson GR, Wilson SE, et al. Relationship of age, gender, race, and body size to infrarenal aortic diameter. The Aneurysm Detection and Management (ADAM) Veterans Affairs Cooperative Study Investigators. J. Vasc. Surg. 1997;26:595-601. Sonesson B, Hansen F, Stale H, Lanne T. Compliance and diameter in the human abdominal aorta—the influence of age and sex. Eur. J. Vasc. Surg. 1993;7:690-697. Sonesson B, Resch T, Lanne T, Ivancev K. The fate of the infrarenal aortic neck after open aneurysm surgery. J. Vasc. Surg. 1998;28:889-894. Sonesson B, Sandgren T, Lanne T. Abdominal aortic aneurysm wall mechanics and their relation to risk of rupture. Eur. J. Vasc. Endovasc. Surg. 1999;18:487-493. Wilson KA, Woodburn KR, Ruckley CV, Fowkes FG. Expansion rates of abdominal aortic aneurysm: current limitations in evaluation. Eur. J. Vasc. Endovasc. Surg. 1997;13:521-526.