Intra- and interobserver variability of target vessel measurement for fenestrated endovascular aneurysm repair

Purpose: To evaluate intra- and interobserver agreement of target vessel measured from computed tomography (CT) scans with 2 measuring techniques used in planning fenestrated endovascular aneurysm repairs (FEVAR): multiplanar reconstruction (MPR) and semi-automated central lumen line (CLL). Methods: CT datasets from 25 FEVAR patients were independently analyzed by 2 experienced observers according to a standardized protocol using the MPR (Leonardo workstation) and CLL (Aquarius workstation) techniques for each patient. Longitudinal vessel separation and clock-face position of the visceral aortic branches were measured twice. The repeatability coefficient (RC) was calculated using the Bland and Altman method to measure intra- and interobserver variability. Differences between groups were examined by paired t test (continuous data) or chi-squared analysis (categorical). Clock-face discrepancy >30 minutes was considered significant. Results: Intraobserver mean difference was insignificant regardless of the measurement technique: the observer and workstation-specific RCs varied between 3.9 and 4.9 mm. Paired measurements differed by >3 mm in 8%. Interobserver variability was greater: observer and workstation-specific RC varied between 5.6 and 7.4 mm, with a tendency toward consistency using MPR, although the mean difference was insignificant. Paired measurements differed by >3 mm in 18%. There was no significant intraobserver variation in clock-face measurement, while interobserver variation was significant in 12% of measurements using the Aquarius workstation and 6% using the Leonardo workstation (p=0.19). Conclusion: Subjective interpretation of anatomical landmarks is more important than measurement techniques or workstations used in the generation of measurement inconsistencies. Introduction of consensus regarding interpretation of anatomical detail and development of fenestrated stent-grafts tolerant of measurement errors might ameliorate some of the problems encountered in FEVAR.