Intraoperative Passive Knee Kinematics During Total Knee Arthroplasty Surgery Kathryn L. Young,1 Michael J. Dunbar,1,2 Glen Richardson,2 Janie L. Astephen Wilson1 1
School of Biomedical Engineering, Dalhousie University, Dentistry Building, 5981 University Avenue, Halifax, NS, Canada B3H 3J5, Department of Surgery, Dalhousie University, Halifax, NS, Canada
2
Received 9 January 2014; accepted 13 May 2015 Published online 3 June 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jor.22945
ABSTRACT: Surgical navigation systems for total knee arthroplasty (TKA) surgery are capable of capturing passive three-dimensional (3D) angular joint movement patterns intraoperatively. Improved understanding of patient-specific knee kinematic changes between pre and post-implant states and their relationship with post-operative function may be important in optimizing TKA outcomes. However, a comprehensive characterization of the variability among patients has yet to be investigated. The objective of this study was to characterize the variability within frontal plane joint movement patterns intraoperatively during a passive knee flexion exercise. Three hundred and forty patients with severe knee osteoarthritis (OA) received a primary TKA using a navigation system. Passive kinematics were captured prior to (pre-implant), and after prosthesis insertion (post-implant). Principal component analysis (PCA) was used to capture characteristic patterns of knee angle kinematics among patients, to identify potential patient subgroups based on these patterns, and to examine the subgroup-specific changes in these patterns between pre- and post-implant states. The first four extracted patterns explained 99.9% of the diversity within the frontal plane angle patterns among the patients. Post-implant, the magnitude of the frontal plane angle shifted toward a neutral mechanical axis in all phenotypes, yet subtle pattern (shape of curvature) features of the pre-implant state persisted. © 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 33:1611–1619, 2015. Keywords: surgical navigation; computer-assisted surgery; knee kinematics; total knee arthroplasty; principal component analysis
The demand for total knee arthroplasty (TKA) surgery is increasing in the younger, more physically demanding population, and older adults receiving the surgery have higher functional expectations than previous recipients. We also know that there is considerable variability among individuals presenting for TKA surgery, in terms of demographics (gender), anatomy, and joint-specific factors such as morphology and passive and active movement and loading patterns.1,2 The general goal of implant design is to develop a durable prosthesis that leads to a well-functioning joint.3 Joint function is predominantly based on simulations of average “normal” knee joint kinematics,4 which do not necessarily reflect the movement variability within the patient population. TKA success, therefore, tends to be variable, with some patients fairing well and others poorly,5 and with little understanding of the mechanism for the variability in function, satisfaction, and longevity. Previous work has demonstrated evidence of natural frontal plane knee alignment variability in healthy populations, with statistically significant deviations from a neutral mechanical axis.6–8 If this were the case for healthy populations, it would be reasonable to expect that patient populations have equivalent or greater frontal plane alignment variability. Biomechanical studies have demonstrated a link between varus and valgus knee alignment and distinctive knee Conflicts of interest: None Grant sponsor: Canadian Institute of Health Research (CIHR); Grant sponsor: Natural Science and Engineering Research Council of Canada (NSERC); Grant sponsor: Dalhousie University Department of Surgery. Correspondence to: Janie L. Astephen Wilson (T: þ1-902-494-6950; F: þ1-902-494-6621; E-mail: [email protected]) # 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.
motion characteristics in those presenting for TKA, both passively and actively.9,10 Recent post-TKA assessment studies have also reported no long-term survival improvements in implants that deviated from a conventional axis after arthroplasty,11 and significantly better function and general outcomes in individuals who were varus pre-operatively and remained mildly varus post-operatively, compared to those surgically altered to clinically neutral.12 In knees that are not neutrally aligned, the surrounding musculature and soft tissue are likely adapted to the mechanical environment, influencing dynamic joint motion. That said, there has been minimal investigation into the variability in knee joint motion in the TKA population,9,13 and how it relates to post-surgical outcome. The utility of computer-assisted navigated TKA surgery has been debated, and as a result, studies that have compared outcomes relative to conventional surgery in terms of longevity, static joint alignment, operating time, blood loss, and post-operative length of stay are equivocal.14–18 This may be because computer-assisted TKA surgery has not yet been used to improve current surgical protocols. It has been used primarily to improve the precision of implant positioning. However, techniques have remained conventional with an overall goal of achieving neutral mechanical alignment. Whether or not this alignment profile contributes to optimized post-operative function, satisfaction and longevity for every individual is yet to be determined. It is likely that this one-size-fit all neutral alignment profile is not optimal for every patient, and that the pre-operative patient-specific joint dynamic envelope contribute to outcome.19 The full potential of computer-assisted TKA surgery, therefore, lies in the possibility of incorporating patient-specific joint dynamics into intra-operative decision-making. We JOURNAL OF ORTHOPAEDIC RESEARCH NOVEMBER 2015
1611
1612
YOUNG ET AL.
have previously correlated the intraoperative frontal plane angle magnitude during an intraoperative passive knee flexion exercise to the knee adduction moment during gait,20 a parameter that has been linked to tibial component migration post-TKA.19,20 The objective of this study was to characterize and summarize the variability within pre-implant passive knee adduction angle patterns over a full extensionflexion range in a large cohort of TKA patients, and characterize the effect of TKA surgical intervention on each pattern.
METHODS This was a retrospective study of patients with severe knee osteoarthritis (OA) undergoing a primary TKA using a surgical navigation system (n ¼ 530). Surgeries were performed between 2007 and 2011 by one of two participating 1 surgeons (M.D., G.R.), with each patient receiving Stryker Triathlon femoral and tibial components. Ethics approval for this study has been received from the Capital District Health Authority (CDHA) Research Ethics Board. 1 The Stryker Precision Knee navigation system (Stryker Corporation, Kalamazoo, MI) was used for all cases. A standard medial parapatellar approach was used in all cases. Following capsulotomy, but prior to ligamentous or bony dissection, two infrared segment trackers were used as fixed reference points by attaching them to the femur and tibia using bicortical anchoring pins. Three markers affixed to a digitization tool were used to record the three-dimensional (3D) positions of lateral and medial epicondyles, the femoral center, medial and lateral malleoli, the tibial center, the anterior-posterior axis of the femur (Whiteside’s line), and the tibial anterior/posterior axis.20 A hip circumduction exercise with no rocking of the pelvis was used to define the hip joint center.21 After anatomical point digitization, the surgeon brought the leg through a standardized series of passive flexion and extension cycles at two time points during surgery (i) prior to osteophyte removal or any bone resection (pre-implant) and (ii) after bone resection and the definitive insertion of the prosthesis (post-implant).20 Surgical resections were performed using a standardized protocol: the distal femoral cut was set at 5˚ of valgus, and the tibial cut was set at neutral (0˚). Anterior and posterior cruciate ligaments were resected in all cases and patellae were resurfaced using an inset patellar button. The measured resection technique was used to obtain a balanced flexion and extension gap. Post-operatively, raw data from the navigation system 1 were extracted and a custom Matlab program (The Mathworks, Natick, MA) was used to calculate the 3D angular movement of the knee according to the joint coordinate system.22 A single knee flexion segment (full extension to full flexion) and its corresponding adduction angles during the passive range of motion were extracted for each patient for the pre- and post-implant states. Curve smoothing and sampling techniques were used to standardize adduction angle curves to a passive range of motion (PROM) between 10 and 110˚ of knee flexion, defining one data point for every degree of flexion. Only cases that had a complete matchedpair, containing pre- and post-implant adduction angle data through a full (10–110˚) passive range of flexion were included in the analysis. All curves were individually assigned a curve fitting condition of either piecewise cubic
JOURNAL OF ORTHOPAEDIC RESEARCH NOVEMBER 2015
hermite, piecewise cubic spline, or quintic spline interpolation, and plotted relative to the original data to visually ensure reasonable approximations. Curves were plotted to illustrate frontal plane angles on the x-axis, and the range of flexion motion on the y-axis, to align with the clinical representation using the surgical navigation system in the operating room. Principal Component Analysis (PCA) There are a number of multivariate statistical analysis techniques that have been used to characterize patterns of joint kinematic data, particularly in the gait analysis literature.23–25 We chose to use the method of principal component analysis (PCA) to extract dominant patterns within the passive adduction angle waveforms over the flexion range because it is a technique that objectively defines the patterns based on the correlation structure within the data. 1 Using a custom program in Matlab , the original data was structured into matrix X of size n p. Each row of n represented a single patient’s adduction angle waveform, and each column was a frame of data (p ¼ 101 data points from 10 to 110˚ flexion), as described by Deluzio and Astephen applied to gait analysis waveforms.26 The first four PCs (PC1–PC4) were retained for analysis, and all pre- and postimplant original waveforms were projected onto each of these four patterns to calculate PC scores, which quantify how closely the original waveform matches the pattern described by each PC. Paired two-tailed Students t-tests were used to examine statistical changes between the pre- and postimplant PC scores, and a one-way analysis of variance (ANOVA) was used to examine if the PC scores were significantly different from zero for the pre- and post-implant states. PC scores that were not significantly different from zero indicated that the group did not significantly exhibit the pattern described by that PC. Pre-implant knees were divided into varus (>2˚), valgus (
School of Biomedical Engineering, Dalhousie University, Dentistry Building, 5981 University Avenue, Halifax, NS, Canada B3H 3J5, Department of Surgery, Dalhousie University, Halifax, NS, Canada
2
Received 9 January 2014; accepted 13 May 2015 Published online 3 June 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jor.22945
ABSTRACT: Surgical navigation systems for total knee arthroplasty (TKA) surgery are capable of capturing passive three-dimensional (3D) angular joint movement patterns intraoperatively. Improved understanding of patient-specific knee kinematic changes between pre and post-implant states and their relationship with post-operative function may be important in optimizing TKA outcomes. However, a comprehensive characterization of the variability among patients has yet to be investigated. The objective of this study was to characterize the variability within frontal plane joint movement patterns intraoperatively during a passive knee flexion exercise. Three hundred and forty patients with severe knee osteoarthritis (OA) received a primary TKA using a navigation system. Passive kinematics were captured prior to (pre-implant), and after prosthesis insertion (post-implant). Principal component analysis (PCA) was used to capture characteristic patterns of knee angle kinematics among patients, to identify potential patient subgroups based on these patterns, and to examine the subgroup-specific changes in these patterns between pre- and post-implant states. The first four extracted patterns explained 99.9% of the diversity within the frontal plane angle patterns among the patients. Post-implant, the magnitude of the frontal plane angle shifted toward a neutral mechanical axis in all phenotypes, yet subtle pattern (shape of curvature) features of the pre-implant state persisted. © 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 33:1611–1619, 2015. Keywords: surgical navigation; computer-assisted surgery; knee kinematics; total knee arthroplasty; principal component analysis
The demand for total knee arthroplasty (TKA) surgery is increasing in the younger, more physically demanding population, and older adults receiving the surgery have higher functional expectations than previous recipients. We also know that there is considerable variability among individuals presenting for TKA surgery, in terms of demographics (gender), anatomy, and joint-specific factors such as morphology and passive and active movement and loading patterns.1,2 The general goal of implant design is to develop a durable prosthesis that leads to a well-functioning joint.3 Joint function is predominantly based on simulations of average “normal” knee joint kinematics,4 which do not necessarily reflect the movement variability within the patient population. TKA success, therefore, tends to be variable, with some patients fairing well and others poorly,5 and with little understanding of the mechanism for the variability in function, satisfaction, and longevity. Previous work has demonstrated evidence of natural frontal plane knee alignment variability in healthy populations, with statistically significant deviations from a neutral mechanical axis.6–8 If this were the case for healthy populations, it would be reasonable to expect that patient populations have equivalent or greater frontal plane alignment variability. Biomechanical studies have demonstrated a link between varus and valgus knee alignment and distinctive knee Conflicts of interest: None Grant sponsor: Canadian Institute of Health Research (CIHR); Grant sponsor: Natural Science and Engineering Research Council of Canada (NSERC); Grant sponsor: Dalhousie University Department of Surgery. Correspondence to: Janie L. Astephen Wilson (T: þ1-902-494-6950; F: þ1-902-494-6621; E-mail: [email protected]) # 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.
motion characteristics in those presenting for TKA, both passively and actively.9,10 Recent post-TKA assessment studies have also reported no long-term survival improvements in implants that deviated from a conventional axis after arthroplasty,11 and significantly better function and general outcomes in individuals who were varus pre-operatively and remained mildly varus post-operatively, compared to those surgically altered to clinically neutral.12 In knees that are not neutrally aligned, the surrounding musculature and soft tissue are likely adapted to the mechanical environment, influencing dynamic joint motion. That said, there has been minimal investigation into the variability in knee joint motion in the TKA population,9,13 and how it relates to post-surgical outcome. The utility of computer-assisted navigated TKA surgery has been debated, and as a result, studies that have compared outcomes relative to conventional surgery in terms of longevity, static joint alignment, operating time, blood loss, and post-operative length of stay are equivocal.14–18 This may be because computer-assisted TKA surgery has not yet been used to improve current surgical protocols. It has been used primarily to improve the precision of implant positioning. However, techniques have remained conventional with an overall goal of achieving neutral mechanical alignment. Whether or not this alignment profile contributes to optimized post-operative function, satisfaction and longevity for every individual is yet to be determined. It is likely that this one-size-fit all neutral alignment profile is not optimal for every patient, and that the pre-operative patient-specific joint dynamic envelope contribute to outcome.19 The full potential of computer-assisted TKA surgery, therefore, lies in the possibility of incorporating patient-specific joint dynamics into intra-operative decision-making. We JOURNAL OF ORTHOPAEDIC RESEARCH NOVEMBER 2015
1611
1612
YOUNG ET AL.
have previously correlated the intraoperative frontal plane angle magnitude during an intraoperative passive knee flexion exercise to the knee adduction moment during gait,20 a parameter that has been linked to tibial component migration post-TKA.19,20 The objective of this study was to characterize and summarize the variability within pre-implant passive knee adduction angle patterns over a full extensionflexion range in a large cohort of TKA patients, and characterize the effect of TKA surgical intervention on each pattern.
METHODS This was a retrospective study of patients with severe knee osteoarthritis (OA) undergoing a primary TKA using a surgical navigation system (n ¼ 530). Surgeries were performed between 2007 and 2011 by one of two participating 1 surgeons (M.D., G.R.), with each patient receiving Stryker Triathlon femoral and tibial components. Ethics approval for this study has been received from the Capital District Health Authority (CDHA) Research Ethics Board. 1 The Stryker Precision Knee navigation system (Stryker Corporation, Kalamazoo, MI) was used for all cases. A standard medial parapatellar approach was used in all cases. Following capsulotomy, but prior to ligamentous or bony dissection, two infrared segment trackers were used as fixed reference points by attaching them to the femur and tibia using bicortical anchoring pins. Three markers affixed to a digitization tool were used to record the three-dimensional (3D) positions of lateral and medial epicondyles, the femoral center, medial and lateral malleoli, the tibial center, the anterior-posterior axis of the femur (Whiteside’s line), and the tibial anterior/posterior axis.20 A hip circumduction exercise with no rocking of the pelvis was used to define the hip joint center.21 After anatomical point digitization, the surgeon brought the leg through a standardized series of passive flexion and extension cycles at two time points during surgery (i) prior to osteophyte removal or any bone resection (pre-implant) and (ii) after bone resection and the definitive insertion of the prosthesis (post-implant).20 Surgical resections were performed using a standardized protocol: the distal femoral cut was set at 5˚ of valgus, and the tibial cut was set at neutral (0˚). Anterior and posterior cruciate ligaments were resected in all cases and patellae were resurfaced using an inset patellar button. The measured resection technique was used to obtain a balanced flexion and extension gap. Post-operatively, raw data from the navigation system 1 were extracted and a custom Matlab program (The Mathworks, Natick, MA) was used to calculate the 3D angular movement of the knee according to the joint coordinate system.22 A single knee flexion segment (full extension to full flexion) and its corresponding adduction angles during the passive range of motion were extracted for each patient for the pre- and post-implant states. Curve smoothing and sampling techniques were used to standardize adduction angle curves to a passive range of motion (PROM) between 10 and 110˚ of knee flexion, defining one data point for every degree of flexion. Only cases that had a complete matchedpair, containing pre- and post-implant adduction angle data through a full (10–110˚) passive range of flexion were included in the analysis. All curves were individually assigned a curve fitting condition of either piecewise cubic
JOURNAL OF ORTHOPAEDIC RESEARCH NOVEMBER 2015
hermite, piecewise cubic spline, or quintic spline interpolation, and plotted relative to the original data to visually ensure reasonable approximations. Curves were plotted to illustrate frontal plane angles on the x-axis, and the range of flexion motion on the y-axis, to align with the clinical representation using the surgical navigation system in the operating room. Principal Component Analysis (PCA) There are a number of multivariate statistical analysis techniques that have been used to characterize patterns of joint kinematic data, particularly in the gait analysis literature.23–25 We chose to use the method of principal component analysis (PCA) to extract dominant patterns within the passive adduction angle waveforms over the flexion range because it is a technique that objectively defines the patterns based on the correlation structure within the data. 1 Using a custom program in Matlab , the original data was structured into matrix X of size n p. Each row of n represented a single patient’s adduction angle waveform, and each column was a frame of data (p ¼ 101 data points from 10 to 110˚ flexion), as described by Deluzio and Astephen applied to gait analysis waveforms.26 The first four PCs (PC1–PC4) were retained for analysis, and all pre- and postimplant original waveforms were projected onto each of these four patterns to calculate PC scores, which quantify how closely the original waveform matches the pattern described by each PC. Paired two-tailed Students t-tests were used to examine statistical changes between the pre- and postimplant PC scores, and a one-way analysis of variance (ANOVA) was used to examine if the PC scores were significantly different from zero for the pre- and post-implant states. PC scores that were not significantly different from zero indicated that the group did not significantly exhibit the pattern described by that PC. Pre-implant knees were divided into varus (>2˚), valgus (
