POST-OPERATIVE ALIGNMENT IN TOTAL KNEE ARTHROPLASTY: LONG LEG VERSUS SHORT LEG RADIOGRAPHS

Abstract

Short leg radiographs remain the standard radiographs available in many UK hospitals. The aim of this study was to see if these radiographs are reliable when assessing the post-operative alignment of total knee arthroplasty in comparison to a Hip-Knee-Ankle (long leg) radiograph.

Twenty consecutive 6 week post-operative long leg radiographs, taken with a standardised protocol, and a short leg radiograph derived from the same digital image were each examined on two separate occasions by two observers. On the long leg radiograph the anatomical and mechanical axis were calculated and on the short leg radiograph the anatomical and surrogate mechanical axis were calculated. These data were used to investigate intra- and inter-observer error. A single observer also collected the same measurements on an additional 30 radiographs (total of 50) to further investigate any patterns of error.

On long leg radiographs, intra-observer agreement was good for both anatomical and mechanical axis for both observers (Intraclass Correlation Coefficients [ICC] of 0.95 to 0.98). The anatomical axis on short leg radiographs was also good (ICC = 0.92 and 0.76). Intra-observer agreement for the short leg radiograph derived mechanical axis was not as consistent (ICC = 0.73 and 0.56). Inter-observer variability was good for long leg radiographs for both anatomical (ICC = 0.89) and mechanical (ICC = 0.95) axis. On short leg radiographs, however, agreement was not as good, in particular for the mechanical axis (ICC = 0.51), but also the anatomical (ICC = 0.73). Taking the long leg radiograph values as the “gold standard” there was a difference in the magnitude of errors seen on short leg radiographs dependant on the knee alignment. Varus aligned knees (n=24) had an average error of 1.2° (0° to 3°) for the anatomical axis and 1.6° (0° to 4°) for the mechanical axis. Perfectly aligned knees (n=8) had an average error of 3.0° (1° to 6°) for the anatomical axis and 2.9° (1° to 5°) for the mechanical axis. Valgus aligned knees (n=18) had an average error of 3.4° (0° to 8°) for the anatomical axis and 5.8° (2° to11°) for the mechanical axis. Using a Mann-Whitney test the magnitude of error was greater for valgus knees for both anatomical (p< 0.0001) and mechanical (p< 0.00001) axes when compare to varus knees. Interestingly all except one knee measured on the long leg radiograph as valgus aligned appeared to be in varus on the short leg radiograph.

In conclusion, short leg radiographs are inadequate to make any comment on leg alignment in total knee arthroplasty. This is most pronounced in a valgus aligned knee.