In vivo patellar kinematics during total knee arthroplasty

Abstract

Patellar maltracking after total knee arthroplasty often results in complications, including anterior knee pain, instability and impingement, and is therefore better resolved intraoperatively. Many factors can affect patellar kinematics during knee replacement, including component position, implant design, joint alignment, and soft tissue tensions. However, to our knowledge, the impact of arthroplasty on patellar kinematics has not been previously reported in vivo. A computer-assisted surgery (CAS) system was developed to measure the pre-arthroplasty patellar kinematics, display the distance between this path and the surface of the planned femoral component, and compare the post-arthroplasty path to the pre-arthroplasty path. Three surgeons from three centers used this CAS system to measure the in vivo pre- and post-arthroplasty kinematics of 18 patients. There was a small, but consistent, proximal shift in the tibial joint lines (mean: 4.2 mm), resulting in pseudo patella-baja, i.e., relatively more distal contact of the patella on the femoral component. This led to significant changes in proximodistal and anteroposterior patellar positioning as well as patellar flexion following arthroplasty (p < 0.008). Mediolateral shift, tilt and internal/external spin had the greatest magnitudes of change (mean: 4.1 mm, 4.6° and 4.6°, respectively) relative to their mean pre-arthroplasty ranges (averaging 2.1 mm, 5.8° and 5.8°, respectively); however, these changes were distributed almost equally medially and laterally, indicating no surgical bias in any one direction. Female patients had more lateral tilt on average than male patients throughout flexion (p < 0.004 post-arthroplasty, p < 0.03 pre-arthroplasty, in later flexion), as well as other kinematic differences; there may therefore be potential for improving overall kinematic results by focusing on gender differences during research, design and surgery. This study demonstrated the feasibility of using a CAS system to measure patellofemoral kinematics. Intraoperative awareness of patellar tracking, including knowledge of the tibiofemoral joint line, could have an impact on the surgical plan and thereby improve the postoperative outcome.

• Computer-assisted surgery
• total knee arthroplasty
• patella
• kinematics
• patellar tracking
• maltracking

Introduction

Patellar maltracking after total knee arthroplasty (TKA) can lead to anterior knee pain, increased component wear, and a higher risk of component loosening, patellar fracture and instability [1–6]. Anterior knee pain alone affects 12–26% of patients following TKA [7–10]. Maltracking may result from improper component design, poor component positioning, excessively tight soft tissues, or poor surgical technique [11–15]. While lateral retinacular release is sometimes used intraoperatively to improve maltracking, this may mask other problems such as component malrotation, and is associated with several avoidable postoperative problems including pain and morbidity, avascular necrosis, patellar fracture and longer rehabilitation [16–18].

Patellar complications are the consequence of multiple factors and usually indicate other surgical errors [14], [15]. Every step of the surgery is important for patellar tracking: tibiofemoral alignment, the tibial cut (coronal orientation, slope, tibial component rotation, and joint line level), the femoral cuts (distal cut coronal orientation, joint line level, anteroposterior position, component size and, most importantly, femoral component rotation), the patellar cut (mediolateral and superoinferior angles as well as resection level) and patellar component positioning [19–21]. Finally, soft tissue handling may influence patellar performance.

Postoperative maltracking can be difficult to resolve: conservative techniques rarely help, and a second operation is not only undesirable but may not rectify the problem [7]. Complex revision surgery may be required to fix the underlying causes. Identifying and correcting problems intraoperatively may not only decrease the rate of anterior knee pain and other patellar complications, but could avoid the need for a revision operation.

Computer-assisted surgery (CAS) can potentially benefit the patella by improving the proper positioning of the femoral and tibial components and by allowing changes to the surgical plan to improve patellar tracking. To our knowledge, however, there is no patellar CAS system in widespread clinical use, and only limited clinical or ex vivo results are available [22–24].

A number of researchers have explored the effects of arthroplasty on patellar kinematics ex vivo, comparing the intact joint to the replaced joint [24–31]. While these studies revealed important factors relating to implant design and surgical technique, none of them detailed all six degrees of freedom. In most cases, only mediolateral shift and tilt were reported, showing no significant changes with arthroplasty, mainly due to the large variability in the results. No ex vivo study can recreate all the physiological conditions; therefore, the absolute patellar path remains uncertain.

In vivo studies of patellar tracking have been conducted for the natural knee joint using bone pins, radiographs or magnetic resonance imaging (MRI) [32]; however, these studies do not provide information on either the pre-arthroplasty arthritic joint, or on the changes in kinematics during the arthroplasty operation. In vivo studies have been conducted on the replaced joint, using MRI [33], [34] and fluoroscopy [35], [36], but none of them has looked at changes due to arthroplasty, and only sagittal-plane patellar kinematics are available when using fluoroscopy. There is, therefore, a lack of in vivo data regarding intraoperative patellar kinematics and the changes caused by arthroplasty. Such data could help to prevent, diagnose and treat the clinically undesirable effects.

The objectives of this study were (1) to test the clinical feasibility of a CAS system for measuring intraoperative patellar kinematics; (2) to determine the in vivo patellar kinematics of a selected number of patients before and after component placement; and (3) to establish whether arthroplasty resulted in any consistent, significant changes in patellar kinematics that could affect the surgical outcome.

Materials and methods

Patellar computer-assisted surgery system

A patellar kinematics module was incorporated into an existing CAS system designed to guide the femoral and tibial cuts for total knee arthroplasty. Both the existing system and the patellar module were developed by Praxim (Grenoble, France). The patellar rigid body reference consisted of two parts: a base attached to the patellar bone and a marker array inserted into the base (Figure 1). The rigid body was designed such that the markers were visible whether the patella was in the upright or everted position, with the optical digitizer placed either medially or laterally. The marker array could be removed from the surgical field when not in use and reproducibly reattached when needed. The surgeon displaced the skin and underlying tissues laterally and then secured the marker base and its spikes to the anterior surface of the patellar bone using a screw, aligning the arrow-shaped marker base mediolaterally on the patella. The locations of the femoral, tibial and patellar marker arrays were detected using a Polaris optical digitizer (Northern Digital, Inc., Waterloo, Ontario, Canada) as part of the CAS system. The accuracy of locating the markers has been reported to be less than 0.35 mm RMS [37].

Figure 1. The patellar marker array was attached in two steps. First, the base was attached with a bone screw to the anterior surface of the patella, mediolaterally, facing towards the optical digitizer. Second, the marker array was screwed onto the base in one of three positions, suiting an upright or everted patella, with either a laterally or medially placed optical digitizer and a medial or lateral incision. [Color version available online.]

The surgeon recorded the pre-arthroplasty kinematics before opening the joint capsule, manipulating the leg through several range-of-motion cycles while capturing continuous kinematic data of the patella, femur and tibia simultaneously (Figure 2); kinematic data were recorded as Euler matrices representing the patellar or tibial coordinate system relative to the femoral coordinate system. After opening the joint capsule, the surgeon digitized the articular surface of the patella (for statistical shape-modeling purposes, although this information was not used in the present study), as well as two specific points on this posterior surface. The two points were the center point on the median ridge (Figure 3), which became the reference point for the trajectory, and the distal extent of the median ridge, which was used to indicate the rotation of the patella.

Figure 2. The surgeons flexed the leg through 3–7 full range-of-motion cycles, while the system recorded the tibiofemoral and patellofemoral kinematics. This was done before opening the joint capsule and again after completing the arthroplasty. [Color version available online.]

Figure 3. The trajectory point of the patella was defined by digitizing the centre of the median ridge on the posterior surface of the patella. This point also defined the origin of the patellar coordinate system. Although pre-arthroplasty kinematics were recorded before opening the joint capsule, the trajectory information was displayed only after opening the joint capsule and digitizing the posterior point. [Color version available online.]

Once the surgeon had planned the femoral cuts in the standard way, based on defined landmarks and/or ligament balancing according to surgeon preference, the CAS system presented the pre-arthroplasty patellar trajectory information to the surgeon, offering the opportunity to change the planned femoral cuts on the basis of this information, while seeing the consequences in real time. The information presented included an image of the femoral component with the trajectory overlaid; the average trajectory of the patella at particular flexion angles; a color-coding of the trajectory in 2-mm increments, showing the distance between the trajectory and the surface of the planned femoral component; the direction of the median ridge at the beginning and end of the trajectory; the thickness of the patella; and the femoral planning options available for change.

The rationale for comparing the original patellar trajectory to the femoral component surface is that, for example, medial placement of the femoral component could result in tightening of the lateral retinaculum, while either anterior placement or a too-large femoral component could lead to overstuffing of the patellofemoral joint, resulting in reduced range of motion and potentially increased pain. Correct rotation of the femoral component was considered of paramount importance.

Post-arthroplasty kinematics were recorded after all cuts were made and the components inserted. If the patella had been resurfaced, the surgeon digitized a new trajectory point at the central apex of the patellar component. Post-arthroplasty and pre-arthroplasty patellar trajectory information were displayed side by side and stored for future reference.

Prior to clinical use, the patellar software and hardware was verified on artificial bones and cadaveric specimens.

Pilot study

Three experienced surgeons (J.L.B., C.d.L., E.S.) from three different institutions were selected to test the clinical application of the new patellar system in a pilot study, which began in February 2003. All three surgeons had routinely used the tibiofemoral portion of the CAS system (KneeLogics LCS; Praxim, Grenoble, France). The respective institutional review boards approved the study as part of a larger prospective clinical evaluation of computer-assisted surgery funded by the French government.

Pre-arthroplasty kinematic data were obtained for 20 patients (10 female, 10 male; age 55–81; 13 right knees, 7 left); however, post-arthroplasty kinematics were only available for 18 patients (10 female, 8 male; age 55–81; 11 right, 7 left). The post-arthroplasty kinematics were not recorded for one patient due to time constraints, and were only recorded for part of the flexion cycle for another, probably as a result of line-of-sight issues preventing visibility of the marker array over the remainder of the cycle. Results are only presented for the 18 patients, to allow a direct comparison between the pre-arthroplasty and post-arthroplasty results. Patients were selected sequentially for each surgeon from the CAS cohort of the CAS/non-CAS randomization of the larger study. All patients required surgery as a result of arthritis rather than trauma. Of the three surgeons, one (J.L.B.) performed 13 cases, the second (C.d.L.) performed 5 cases, and the third (E.S.) performed 2 cases. The patella was resurfaced in 7 patients (4 by J.L.B., 3 by C.d.L.). Their decision to resurface was based on notable patellar arthritis and/or patellofemoral pain at the preoperative clinical examination. For the post-arthroplasty kinematic measurements, the surgeon closed the incision using three sutures (E.S.), or three towel clips (J.L.B.) or left it open due to normal tracking (C.d.L.).

Posterior-cruciate-sacrificing LCS Complete™ rotating platform knee components (Depuy, Warsaw, IN) were used with a medial parapatellar approach. The patellar component, when used, was a metal-backed mobile-bearing design. All surgeons based the femoral cuts on both femorotibial alignment and ligament balancing, and made an effort not to overstuff the joint, especially laterally [19]. The surgeons favored external rotation of the femoral component, averaging 5.5° across all patients, relative to the posterior condyles. A lateral release was performed in three patients. In this pilot study, the surgeons made no adjustments to the femoral component planning on the basis of the patellar kinematic data provided. In one case (preoperatively valgus), the patella had moderate subluxation postoperatively, while in another there was mild subluxation. No other complications were noted, based on a thorough clinical review performed for the extended CAS study. There have been no revisions to date, with an average of 4.2 years having elapsed since the time of surgery.

Visualization of patellar tracking

In addition to the quantitative analyses, several types of visualization provided useful qualitative data during the data analysis phase. At the end of the surgery, the CAS system displayed and saved images of the pre-arthroplasty and post-arthroplasty trajectories overlaid on the femoral component: this gave an immediate impression of the mediolateral and proximodistal shift of the trajectory, including where the trajectory crossed the component notch. During data analysis, we (1) overlaid the pre-arthroplasty patellar trajectory on the bone-morphed femoral bone geometry [38] and (2) overlaid the post-arthroplasty trajectory on the manufacturer-provided femoral component geometry and original bone geometry (Figure 4). This displayed the six-degree-of-freedom (6-DOF) placement of the component with respect to the bone, as well as the 6-DOF trajectory of the patella with respect to the natural trochlea on the femur or the groove on the femoral component, and could be used particularly to examine potential overstuffing, especially laterally.

Figure 4. A sample distoproximal visualization of a pre-arthroplasty femur (left), derived from bone-morphed geometry, is shown beside the corresponding post-arthroplasty femur overlaid with the femoral component position (right). The righthand (green) line shows the trajectory of the centre of the patellar median ridge, while the lefthand (blue) line shows the trajectory of the distal end of the median ridge throughout several cycles of flexion. The custom visualization software also allowed the patellar path to be viewed dynamically. [Color version available online.]

Kinematic analysis

The anatomical coordinate system of the femur was defined as follows: the mediolateral (ML) axis (Y) was calculated from the medial to the lateral posterior condyle, the condyles being identified with a digitizing probe. The anteroposterior (AP) axis (Z) was perpendicular to both the ML axis and the line connecting the top of the femoral notch (digitized by the surgeon) to the hip center (calculated while manipulating the knee in a circular arc). The proximodistal (PD) axis (X) was mutually perpendicular to the other two axes, completing the coordinate system. The origin of the coordinate system lay at the top of the femoral notch. Once the anatomical coordinate system was defined, the translations and rotations of the arbitrarily oriented femoral marker array, tracked by the optical digitizer, were transformed to the anatomical system, which is necessary for clinical interpretation. The tibial anatomic coordinate system was based on the intercondylar eminence, the medial third of the anterior tibial tubercle, and the midpoint between the medial and lateral maleoli at the ankle.

The axes of the patellar coordinate system were coincident with those of the mediolaterally directed patellar marker array; the origin was at the digitized trajectory point (the center of the median ridge in the unresurfaced patella or the apex of the patellar component in the resurfaced patella). Patellar translations were defined relative to the femoral coordinate system, e.g., patellar shift was calculated parallel to the femoral mediolateral axis. The sequence used to define the patellar rotations was as follows: patellar flexion about the femoral flexion axis, followed by patellar spin about the patellar anteroposterior axis, followed by mediolateral tilt about the patellar superoinferior axis. Patellar thickness was reported by the CAS system based on the perpendicular distance from the base of the marker array to the digitized trajectory point on the patellar articular surface or, when replaced, to the apex of the patellar component.

We separated the resulting kinematic data into flexion and extension segments for each cycle, interpolated the data at one-degree increments of tibiofemoral flexion, and then averaged these points across the 3–7 cycles at each increment. We analyzed all six DOFs relative to the tibiofemoral flexion angle, and also projected the trajectory onto the three anatomical planes (ML, AP, PD) for easier visualization. We present the results for 15° flexion to 120° flexion, as these were common to most cycles.

Tibial joint line

As part of the standard CAS procedure, a navigated device was used to measure and record the plane of the final tibial cut surface. The distance (in the tibial coordinate system) of this plane from the original tibial joint line, as defined by the height of the less worn medial or lateral tibial plateau, was recorded in the log file. We subtracted this distance from the planned insert thickness to determine the change in the tibial joint line.

Statistical analyses

We tested the statistical hypothesis that there were no changes in patellar kinematics between the pre- and post-arthroplasty states. To do so, we extracted the patellar data at 15°, 45°, 90° and 120° tibiofemoral flexion, and then performed a repeated-measures ANOVA (α = 0.05). Given significance, we used paired Student's t-tests with a Bonferroni correction factor for the six DOFs (α = 0.05/6 = 0.008) to compare the pre-arthroplasty and post-arthroplasty results.

In addition to detecting significant changes in kinematics that occurred in a consistent direction, we were interested in determining the magnitude of change regardless of direction. We spline-fit a histogram of the absolute results to estimate a probability distribution function, and compared the mean absolute change for each degree of freedom to its range during pre-arthroplasty flexion. The goal of this analysis was to understand which degrees of freedom were most affected relative to the total range across flexion.

Finally, we performed an ANOVA on the female versus male results and resurfaced versus non-resurfaced results at each flexion angle. Given significance, we compared the results using unequal-variance t-tests. As above, we set the level of significance at α = 0.008 (0.05/6) so as to have the most conservative approach to detecting significance in multiple comparisons.

Results

Clinical use of the patellar kinematics module by three surgeons for multiple patients with both non-resurfaced and resurfaced patellae demonstrated the feasibility of the system in vivo.

The pre-arthroplasty patellar path, in most cases, showed minimal mediolateral deviation throughout flexion with respect to the anatomical coordinate system of the femur; in other words, the plane of movement was essentially perpendicular to the posterior condylar axis (Figure 5). On average, the trajectory was 2–3 mm lateral to the central plane, with a consistent standard deviation (averaging 4.2 mm) throughout the range of motion (Figure 6). Flexion and extension trajectories were quite similar, except for a tendency to be more proximal and extended during early flexion. Mediolateral tilt was more variable across patients (mean SD: 11.7°), but typically remained relatively steady throughout the range of motion, averaging 6.3° lateral. Patellar spin became increasingly external with greater flexion, changing from a mean of 5° external in early flexion to a mean of 15° external in late flexion (mean SD: 11°). The remaining three degrees of freedom were primarily controlled by the joint geometry, with the patella moving distally and posteriorly, and becoming more flexed with increasing tibiofemoral flexion. Even at close to full extension, the patella was located relatively distally on the femur, approximately 15–20 mm above the notch, crossing the level of the notch at a mean of 74°, although this varied widely across patients (SD: 21°).

Figure 5. The pre-arthroplasty patellar trajectory (mean and standard deviations) had minimal mediolateral deviation throughout flexion. Geometric data for the femurs were obtained intraoperatively by matching digitized portions of the femoral surface to a statistical shape model. The schematic shows a typical femoral shape. [Color version available online.]

Figure 6. The six degrees of freedom of pre-arthroplasty kinematics are shown relative to tibiofemoral flexion for both the flexion and extension phases (mean and standard deviation). On average, the patella tracked slightly laterally (upper left), with lateral tilt (upper right) and external spin (mid right). The other translations and rotation (AP, PD and flexion) were primarily controlled by the joint geometry (see Figure 5). [Color version available online.]

Post-arthroplasty patterns were similar to the pre-arthroplasty results (Figure 7), but with shifts in some of the absolute values. The direction of these changes appears to be paradoxical (e.g., more proximal post-arthroplasty despite contacting the femoral component more distally) because these values represent the patellar bone with respect to the femoral bone (to which the marker arrays were attached) rather than with respect to the femoral component. The Discussion section provides further explanation and analysis. Relative to the femoral bone, the post-arthroplasty patella was more proximal in mid-to-late flexion (mean: 2.2 mm; p < 0.008), more posterior throughout (mean: 5.5 mm; p < 0.001), and slightly more flexed in early flexion (mean: 2.9°; p < 0.008) (Figure 8). The more proximal/posterior/flexed kinematics post-arthroplasty were seen across all three surgeons. There were no consistent differences between the resurfaced and non-resurfaced knees studied. Resurfacing increased the patellar thickness by a mean of 1.3 mm (range: −1.5 mm to +2.7 mm, the maximum being for an originally thin patella of only 16 mm); this slight increase in thickness probably occurred because the patellae chosen for resurfacing were severely arthritic and more likely to be worn away.

Figure 7. The post-arthroplasty patellar trajectory (mean and standard deviation) also had minimal mediolateral deviation throughout flexion. All three surgeons resected less from the tibia than the insert size, thus raising the joint line. This resulted in a more proximal femoral component, more distal contact of the patella on the femoral component, and a longer portion of the trajectory within the notch. This is called pseudo patella baja, as the distal patellar contact is not due to shortening of the patellar tendon. Geometric data for the femoral component were obtained from the manufacturer; the schematic shows a typical placement of the component on the femur. [Color version available online.]

Figure 8. The six degrees of freedom of post-arthroplasty kinematics are shown relative to tibiofemoral flexion for both the flexion and extension phases (mean and standard deviation). The overall patterns are similar to the pre-arthroplasty patterns (Figure 6), but with shifts in the absolute values (see Figure 9). [Color version available online.]

Pre/post-arthroplasty changes for mediolateral shift, mediolateral tilt and internal/external rotation (spin) were not in a consistent direction and were therefore not statistically significant (Figure 9). Disregarding the direction of change, the absolute pre/post changes for shift, tilt and spin were 4.1 mm, 4.6° and 4.6°, respectively (Figure 10). These changes were large relative to the overall range through the pre-arthroplasty flexion cycle (see Figures 5 and 6), which averaged 2.1 mm, 5.8° and 5.8°, respectively. As a result, the largest relative absolute changes due to arthroplasty were for ML shift, ML tilt and internal/external spin (Figure 10).

Figure 9. The changes between pre-arthroplasty and post-arthroplasty kinematics (mean and standard deviation), for all six degrees of freedom, flexion and extension phases, are shown with respect to the femoral bone reference. Arrows indicate significant differences.The patella was more posterior, more proximal and more flexed after arthroplasty (p < 0.008) due to “rounding the femoral corner” sooner; see the text and Figure 12 for an explanation of these paradoxical results given more distal patellar contact on the femoral component. None of the primary tracking characteristics (ML shift, ML tilt, int/ext spin) showed a significant bias in one direction or the other, although individual changes were potentially relevant (see Figure 10) [Color version available online.].

Figure 10. The absolute change in each degree of freedom shows the effect of arthroplasty regardless of the direction of change (i.e., the absolute value of the data in Figure 9). The histograms sum the results for all four flexion angles studied (15°, 45°, 90° and 120°). The curves show a spline fit to a non-parametric probability density function estimate of the histogram data. The arrow indicates the mean change. At the top of the graph, this mean absolute change is compared to the pre-arthroplasty range for that degree of freedom (see Figures 5 and 6). This shows that mediolateral tilt, shift and internal/external spin had the highest relative change due to arthroplasty. [Color version available online.]

The tibial joint line shifted proximally in all cases (mean: 4.2 mm; range: 0.7 to 8.9 mm), resulting in pseudo patella baja, i.e., more distal contact of the patella on the femoral component, relative to the notch or distal condyles (Figure 7), and consequent changes in the proximodistal and anteroposterior patellar position as well as patellar flexion (p < 0.008; Figure 9). In contrast to true patella baja [39], there was no evidence of an abnormally short patellar tendon.

Female patients in this study had noticeably more lateral tilt than males, averaging about 10° more lateral consistently throughout flexion for both pre- and post-arthroplasty, reaching significance in deep flexion (>90°) post-arthroplasty (p < 0.004) (Figure 11). This result was not expected. Females also had significantly more flexed patellae in deep flexion (p < 0.008) and significantly more distal patellae in mid flexion (p < 0.001). Neither males nor females showed significant changes in mediolateral tilt with arthroplasty. There was no significant difference between females and males regarding the change in the joint line. There were also no significant differences in the remaining degrees of freedom, notably mediolateral shift.

Figure 11. Female and male mediolateral tilt are compared for pre-arthroplasty, post-arthroplasty, and change due to arthroplasty. Female patients in this study had, on average, approximately 10° more lateral tilt than male patients, both pre-arthroplasty and post-arthroplasty; the difference was significant in later flexion (p < 0.004). Arthroplasty had a minimal effect on tilt. [Color version available online.]

Discussion

We investigated the intraoperative impact of arthroplasty on in vivo patellar kinematics using a novel computer-assisted surgery system. For the patients in this study, there were statistically significant, and potentially clinically significant, changes in kinematics. To our knowledge, this is the first report of in vivo patellar kinematics before and after arthroplasty.

In all cases, there was a rise in the tibial joint line. The surgeons intentionally under-resected the tibia to counteract the loss of flexion stability after PCL excision (described below). To achieve equal flexion-extension gaps, the surgeons then placed the femoral component more proximally. This rise in joint line also appears to be a philosophy of the LCS system, based on the instrumentation available, in order to conserve tibial bone. The insert sizes used were relatively small, normally 10.0 mm or 12.5 mm (only two were 15.0 mm), indicating that the change in joint line was not due to an extra-large insert; also, anteroposterior dimensions were visually similar between the anatomical femur and the femoral component, indicating that the change in joint line was not due to an extra-small femoral component. The more proximal femoral component, for the same length of patellar tendon, caused the patella to contact the femoral component relatively distally (pseudo patella baja) and therefore to “round the corner” towards the distal condyles sooner than in the natural joint (Figures 7 and 12). This can lead to patellotibial impingement in deep flexion, decreased range of motion (ROM), and pain [40]. The CAS system did not specifically provide joint line information to the surgeons during our study.

Figure 12. When the joint line is raised due to under-resection of the tibia (whether intentional or unintentional) and the femoral component is consequently positioned more proximally, the patella contacts the femoral component more distally (relative to the notch or distal condyles) due to the constant length of the patellar tendon. However, relative to the coordinate frame of the femoral bone (to which the femoral marker array is attached), the position of the patella remains relatively unchanged in extension (blue dot, anteriorly). In flexion, the patella “rounds the corner” sooner than in the intact knee (red dots, distally). Relative to the femoral coordinate frame, this paradoxically causes the patella to be more flexed in early flexion and more proximal and posterior in later flexion relative to the pre-arthroplasty kinematics (Figure 9). [Color version available online.]

The posterior translation of the patella with respect to the femur (mean: 5.3 mm, relatively consistently throughout flexion) appeared to be due to traveling “around the corner” sooner rather than either posterior placement of the component or undersizing of the component, based on a visualization of the patient's femoral geometry and the planned size and position of the femoral component. When the ideal component size lay between two sizes, the surgeons chose the larger size or put the femoral implant in slightly more flexion.

It is known that excision of the posterior cruciate ligament (PCL) results in increased distraction (laxity) at 90° tibiofemoral flexion (averaging 4.8 mm medially and 4.5 mm laterally in one study [41]), with only a minimal increase in extension (averaging 0.9 mm and 0.8 mm, respectively [41]). Choosing the femoral component size based on the AP dimensions therefore necessitates placing the component more proximally to create equal flexion and extension gaps. An alternative approach, used by some surgeons, is simply to allow the greater flexion gap in the belief that a PCL-sacrificed knee can withstand a larger flexion gap without loss of stability; we are unaware of patient surveys or scientific evidence supporting either opinion. Since the original femoral joint line is rarely perpendicular to the mechanical axis of the leg, it is necessary to resect more from one side distally. Likewise, axial rotation and positioning of the femoral component are important since they affect the position and orientation of the trochlea. Particular attention must be paid to avoid overstuffing the lateral patellofemoral compartment.

Many factors could influence patellar superoinferior flexion, including the contact location on the femoral component, the contact location (pivot point) on the patella, the position of the quadriceps tendon with respect to the femoral component, the tension in the patellar and quadriceps tendons, the position of the femur on the tibia, or the thickness of the patella. The patellar contact position usually traverses from slightly inferior when the knee is in extension to about 7–9 mm superior on the patella by 90° flexion [36], [42], [43]. In ex vivo contact studies, the flexion angle at which the patellar contact area split in two in the replaced joint (as it entered the notch) varied from a mean of 60° [44] to a mean of approximately 105° [28] and depended on component design [45]. The 74° found in our study falls into this range, tending towards earlier contact with the notch.

A proximal shift of the tibial joint line has been reported in other recent series as well, whereas reported femoral joint line changes are more variable. One study [46] reported an average proximal shift in the tibial joint line of 4.2 mm (primarily related to the medial condyle), whereas the femoral joint line shifted distally by an average of 3.4 mm; another study [40] described an average joint line elevation of 3.1 mm; a third study [47] reported an average elevation of 1.1 mm (ranging from −3 to +12 mm).

Pseudo patella baja (due to a change in the joint line as opposed to a change in the length of the patellar tendon) has been described by several authors [39], [40], [46], [48]. Although the consequences of true patella baja are considered to be greater, pseudo patella baja can still lead to patellotibial impingement in deep flexion, thereby damaging the tendon or the patella, a decreased lever arm (complicated by femoral rollback) and hence greater energy expenditure, a decreased range of motion, and possibly anterior knee pain [40]. Nevertheless, a change in the joint line may be necessary due to varus or valgus deformity [49]. The consequences of an elevated joint line need to be weighed against the wear benefits of a thicker polyethylene tibial insert and the loosening benefits of maintaining the stronger proximal bone in the tibial plateau. These in turn require a more proximal femoral component to maintain ligament balancing. Notwithstanding these considerations, it is generally agreed that maintaining the joint line should be a surgical goal [49].

Changes in mediolateral shift and tilt due to arthroplasty were variable between patients, but were often substantial (averaging 4.1 mm in shift and 4.6° in tilt). Maximum changes were 15.6 mm and 17.5°. Tilt is especially influenced by the lateral prosthetic condyle, as compared to the anatomical one, as well as component size, particularly in the AP dimension [19]. Since we only have data on the planned positions of the femoral and tibial components, we are unable to report the position of the patella with respect to the femoral component. Visually, based on overlaying the planned femoral component position on the individualized geometry of the patient's femur, the patella normally followed the groove of the femur pre-arthroplasty and the groove of the femoral component post-arthroplasty. Exceptions to this would indicate a greater potential for problems. Mediolateral changes may result from a shift in the groove location due to one of several reasons: translation or axial rotation of the femoral component; changes in the soft-tissue tensions; or medialization or lateralization of the patellar component. Two recent studies have reported routine (i.e., unintended) medialization of the femoral groove due to arthroplasty [50], [51]. Changes in spin were probably due to either the realignment of the leg as a result of the arthroplasty, or a change in the Q-angle based on the alignment of the components.

It is difficult to compare our absolute tracking results to results of previous studies because of the large variations between and even within studies [24–32]. Nevertheless, the patterns and directions that we found for mediolateral shift and tilt, and for patellar spin, fall within the range of those previous studies. Importantly, since the values for tilt and shift were close to zero, the definitions of “medial” and “lateral” depend on the definition of the coordinate system. The large changes in AP and PD translations and the linear changes in patellar flexion (proportional to tibiofemoral flexion) are consistent with all studies reporting those values.

The significant kinematic differences found between female and male patients in this study, particularly the greater lateral tilt in females, suggest the potential for improving overall kinematic results by focusing on gender differences during research, design and surgery. Possible causes for such differences include differences in the bony geometry of the femur [52–55] or patella [56], [57]; more severe joint degeneration at the time of surgery [58–60]; a larger quadriceps (Q) angle resulting from shorter average stature [61]; or possible differences in soft tissue tension or orientation [62]. Potential areas of improvement include implant design, implant positioning, or increased awareness of intraoperative tracking through computer- assisted surgery.

One limitation of this study is that the number of patients studied was relatively small; despite this, we were able to identify significant effects due to arthroplasty. The surgeons were reluctant to do more cases with the patellar module because it made no change to their surgical practice (in this first phase) and was time-consuming to use. The diversity of conditions, including the participation of different surgeons and the inclusion of both resurfaced and non-resurfaced patellae, is both a strength and a weakness of the study. The weakness is that the data were sub-divided into groups that were too small to compare reliably; the strength is that the feasibility of the system was shown across a variety of users, subjects and implant conditions, although for only one implant design. While the pre-arthroplasty kinematics were taken with the capsule closed, post-arthroplasty kinematics did depend the tension imposed by the medial sutures or towel clips. It would have been informative to measure the kinematics between femoral resurfacing and patellar resurfacing in order to determine their independent effects; however, this option was not available at the time. The CAS system described would be most useful in the case of patellar resurfacing, when the surgeon could adjust the placement of the patellar component and the thickness of the patella in response to a display of the kinematics after femoral resurfacing. Nevertheless, simply measuring patellar kinematics encourages the surgeon to give femoral, tibial and patellar component placement greater consideration with regard to patellar tracking than would otherwise be the case.

The hardware and software for the patellar CAS module are currently being updated based on this pilot study, mainly to improve the presentation of the information to the surgeon and reduce the time required. The revised module should be used with a larger number of patients, including an investigation of patella-specific outcomes for patients with and without the use of the patellar computer-assisted surgery module.

In summary, patellofemoral issues, especially anterior knee pain, are common after total knee arthroplasty. Component positioning and surgical technique have been shown to play an important role in these complications. Patellar tracking may be improved intraoperatively by selecting different component sizes or changing the position and orientation of the components. We have demonstrated the feasibility of a computer-assisted surgery system in vivo, and have shown that arthroplasty does affect patellar kinematics. In particular, we have gained a better understanding of the changes in patellofemoral mechanics as a result of TKA. This should inspire new designs to address these issues more effectively in TKA. Since it is much easier to resolve patellar problems intraoperatively than postoperatively, use of a patellar CAS system–indeed, simply increasing intraoperative awareness of patellar tracking, joint line level and potential gender-related differences–could reduce the rate and severity of pain and complications after total knee arthroplasty.

Acknowledgements

The authors wish to thank Carinne Granchi of Praxim for creating the visualization program showing the patellar trajectory overlaid on the anatomical femur and the femoral component. We would also like to thank Dr. Christian Roux, Gwenael Guillard, Nathalie Perrin and Joel Savean of the Laboratoire de Traitement de l’Information Médicale INSERM of Brest, France, for their work on the analysis of the initial subjects.

Conflict of interest statement: One author (E.S.) has a consulting agreement with Praxim; another (C.A.) had a consulting agreement with Praxim during the first phase of data analysis; and a third (C.P.) is an employee of Praxim.