The accuracy of computed tomography central luminal line measurements in quantifying stent graft migration

This study evaluated the accuracy of central luminal line (CLL) measurements in quantifying stent graft migration. The bias of the CLL technique together with observer variability were assessed.

Stent grafts were deployed in plastic aortic phantoms at fixed locations from two side branches. Each phantom was filled with iodinated contrast, and a 2-mm multislice computed tomography (CT) scan was performed. The stent graft was then displaced caudally, its new location determined, and again, a CT scan performed. This created a series of 15 cases with known stent graft migration. CLLs were used to measure stent graft position on the CT scans and calculate migration (3 observers). In vivo stent graft migration was then evaluated in a similar manner using a series of follow-up CT scans from nine patients (2 observers). All CLL measurements were performed independently and were repeated on a separate occasion.

The mean difference in CLL migration between the actual and observed measurements (bias) in the aortic phantoms was <1 mm. The 95% confidence intervals for the bias were within the interval (-1 and 1 mm), and the 95% limits of agreement were within -3 mm and +3 mm. The 95% limits of agreement for measurements within and between observers were -4 to 2 mm and -2 to 2 mm, respectively. The phantom study generated a coefficient of repeatability (RC) of 1 mm for within-observer measurements. Clinically, CLLs generated 95% limits of agreement within and between observers of -3 to 4 mm (RC, 2 mm) and -3 to +3 mm, respectively.

Bias from CLL-determined migration is small and insignificant from a practical point of view. A small amount of measurement variability within and between observers does exist; it should be feasible to detect changes in stent graft position that are ≥4 mm.