VARIABILITY AND ACCURACY OF COMPUTER-ASSISTED MEASUREMENT OF WEAR IN KNEE ARTHROPLASTY

Abstract

Introduction. Using conventional tests, manual computer-assisted methods offer a satisfactory reproducibility and accuracy in wear evaluation in knee arthroplasty. However, repeatability may be underestimated. Various calibration methods were proposed but not tested. Accuracy may decrease with fluoroscopic images.

Method. We recalculated intra- and interobserver variability based on 132 fluoroscopic images of 3 unimplanted specimens, using specific analyses like the Brand and Altman test or the repeatability coefficient. Calibration was tested in comparison to a reference method (based on a circular shape of known diameter). The influence of a radiological marker embedded into the central part of the polyethylene to control the incidence of fluoroscopic guided images was tested. Accuracy was compared between fluoroscopic and digitized plain radiographs.

Results. The coefficient of repeatability indicated a variation of 0.4 mm in repeated measurements by the same observer. Although a high interobserver reproducibility (r=0.99), the Brand-Altman test showed inconsistent values in about 30% of cases with both observers. The method of calibration influenced the results significantly. Under ideal conditions of radiographic incidence, the accuracy was better with a digitized x-ray (0.10 mm) than with fluoroscopic images (0.25 mm), and compared favourably with fully automatized methods. If small variations of radiographic incidence occured, the accuracy droped to 1 mm. A comparison of successive radiographs of the same implant showed that 1° tilting induced an average reduction of 0.36 mm in the apparent thickness of the polyethylene.

Discussion. When comparing wear measurements in TKA using manual, computer-assisted methods, examination by a single observer should be preferred. The difference between the measurements of the thickness of the polyethylene insert on two images taken with exactly the same incidence should be considered significant if it is greater than twice the precision of the method, i.e. greater than 0.20 mm on digitized X-ray images, or greater than 0.50 mm on fluoroscopic images. In the absence of a strict control of the radiographic incidence, it appears that the difference between the measurements of the thickness of the polyethylene on two images should be considered significant only if it is greater than 2 mm. Given the estimated wear rate of conventional TKR, 10 years of use of the prosthesis would be needed before such systems can detect wear with certainty.

Conclusion. A metallic marker of known dimension embedded in the central part of the polyethylene insert appears mandatory to control calibration as well as the angle of incidence of fluoroscopically guided radiographs. This would improve the performance of radiographic measurements enough to offer clinical relevance.