Inadequacy of computed tomography for pre-operative planning of patellofemoral arthroplasty

To evaluate the accuracy of preoperative planning for patellofemoral arthroplasty (PFA) by comparing: (1) virtual implant positioning simulated on pre-operative images versus (2) real implant positioning from post-operative images. The authors prospectively studied 15 patients that received a PFJ implant (Tornier, Montbonnot France). A pre-operative planning software was established to determine the size and position of the trochlear component. Pre-operative scans were used to perform virtual implantations by two different operators, which were then compared to the post-operative scans to calculate errors (epsilon) in implant positioning and intra-class correlation coefficients (ICC) for intra- and inter-observer repeatability. Analysis was performed for 13 patients, for whom agreement between virtual and real surgery was excellent for anteroposterior (AP) position (ICC = 0.84; epsilon (max) = 3.5 mm), fair for proximodistal (PD) position (ICC = 0.50; epsilon (max) = 9.5 mm), and poor for mediolateral (ML) position (ICC = 0.07; epsilon (max) = 9.0 mm). It was fair for flexum-recurvatum (FR) alignment (ICC = 0.53; epsilon (max) = 8.2A degrees), poor for varus-valgus (VV) alignment (ICC = 0.34; epsilon (max) = 10.0A degrees), and internal-external (IE) rotation (ICC = 0.34; epsilon (max) = 10.6A degrees). Pre-operative planning was insufficiently accurate to follow intra-operatively, the greatest errors being angular alignment (VV and FR). The clinical relevance of these findings is that PFA is difficult to plan pre/operatively due to non-visibility of cartilage on CT scans and to trochlear dysplasia in most cases. Prospective evaluation of operative tools on consecutive patients, Level III.