Improved knee biomechanics among patients reporting a good outcome in knee-related quality of life one year after total knee arthroplasty

Between the years 2010 and 2014, 40 patients were recruited from two orthopedic departments in Stockholm County, Sweden (Ortho Center, Löweströmska Hospital and the Department of Orthopedics at Karolinska University Hospital). Inclusion criteria were: being scheduled for TKA within one month due to primary knee osteoarthritis, the ability to walk 10 m repeatedly without assistance of a walking aid, and the ability to understand verbal and written information in Swedish. Exclusion criteria were: other severe joint pain or previous major orthopedic surgery in the lower extremities (including traumatic knee injury), rheumatoid arthritis, neurological disorder, diabetes mellitus, body mass index (BMI) > 40, and/or other medical condition affecting walking ability. All patients that met inclusion and exclusion criteria and accepted participation were included in this prospective cohort study [11]. Twenty-five age and gender matched, healthy controls without any known musculoskeletal disease or neurological disorder were recruited through acquaintances between the years 2013–2015. The control group was matched to the osteoarthritis group by age strata across five age groups (40–49, 50–59, 60–69, 70–79, 80–89 years of age). The mean age of the control group was 66 years (SD 10), and they had a mean BMI of 24.9 (SD 2.9). The present study is a secondary analysis and therefore, participants and methods are described more in detail elsewhere [11]. The study was approved by Stockholm’s regional ethical review board (DNR: 2010/1014-31/1), and all study participants provided written informed consent in accordance with the Declaration of Helsinki.

Out of the 40 patients included at baseline, 28 patients (18 females), with a mean age of 66 (SD 7) years, completed the one year follow-up. Reasons for not completing the one year follow up were: not going through with the planned surgery (n = 2), post-operative infection causing re-operation (n = 2), TKA in the contralateral limb within the following year (n = 5), death (n = 1), pelvic fracture during the following year (n = 1), and cancer diagnosis (n = 1). Patients who did not complete the one year follow up (n = 12) did not differ statistically from the studied group with regards to distribution of age, gender, weight, height, BMI or duration of years with symptomatic knee osteoarthritis.

Baseline evaluations were performed within one month prior to surgery (mean 20 (SD 13) days) and postoperative evaluations one year after surgery (mean 12 (SD 0.9) months). Three-dimensional gait data and patient-reported outcomes were collected at the Motion Analysis Laboratory at Karolinska University Hospital, Solna, Sweden. Each test session started with a physical examination. Passive range of motion of the lower extremity joints were recorded using a goniometer with the patient in a supine position for all measures except hip extension which was recorded with the patient in a prone position. Anthropometric measures were recorded using calibrated scales. After the initial examination, 35 retro-reflective markers were placed on anatomical landmarks (head, trunk, pelvis, lower and upper extremities), according to the conventional biomechanical model Plug-In-Gait (Vicon Motion Systems Ltd, Oxford, UK) [14].

Three-dimensional gait data were collected using an eight camera system (Vicon Motion Systems Ltd, Oxford, UK) (sampling rate 100 Hz), and two force plates embedded in the floor (Kistler, Winterthur, Switzerland) (1000 Hz). Kinematic, kinetic, and time and distance parameters were collected simultaneously. Kinetics were expressed as internal moments (forces from muscles, ligaments and tendons acting on the specific joint). The studied kinematic variables included knee flexion-extension range during an entire gait cycle, peak knee flexion angle in swing phase, and peak varus angle during stance phase. The studied kinetic variables during stance phase included peak knee flexion moment, peak knee extension moment, and peak knee valgus moment. Patients were instructed to walk barefoot at self-selected speed along a defined 10 m walkway. Recordings were made in two directions (back and forth). Approximately five gait trials, with clean force plate strikes, were analyzed for each patient at each test session. Gait variables from these strides were averaged to obtain one value for each variable of interest, for each patient, at each test session. Raw motion capture data were filtered in a Woltring Filter [15] with a mean squared error setting of 15, and 3D gait kinematics and kinetics were computed according to the Plug-in-Gait model [14]. The gait kinematics and kinetics data were then exported to the software program MATLAB®, R2014a (The MathWorks, Inc, Natick, MA, USA) where discrete values (maxima, minima) were computed for the participants.

After completing the gait trials, patients rated their perceived pain experienced during the gait trials using a visual analogue scale (VAS) where 0 represents “no pain” and 100 mm represents “severe pain” [16].

The KOOS was used to evaluate patient-reported pain, symptoms, function in activities of daily living (ADL), function in sport and recreation, and knee-related QoL [17]. KOOS generates a final score for each separate subscale, ranging from 0-100, where 0 represent “severe difficulties” and 100 “no problems at all”. Adequate test-retest reliability (intra class correlation range 0.85 – 0.9) has been reported for all subscales [18]. The questionnaire is widely used, and is valid and responsive to change in patients with knee osteoarthritis receiving both conservative [19] and surgical treatment [20].

According to standard practice at each hospital, preoperative standing anterior-posterior radiographs were taken. Two experienced orthopedic surgeons together performed the classification of radiographic severity of osteoarthritis (for the knee joint as a whole) according to the modified Kellgren and Lawrence’s (KL) classification [21] (Table 1).

The surgical procedures were performed by seven different senior orthopedic surgeons from two hospitals, all using a posterior cruciate ligament retaining cemented TKA (PFC-Sigma, DePuy, Johnson & Johnson, Warsaw, Poland). Surgeons were equally distributed across the two groups. After surgery, patients were allowed full weight bearing (together with use of an appropriate walking aid), and unrestricted range of motion. The postoperative rehabilitation was performed according to the standard practice at each hospital (in-patient physiotherapy <1 week). Thereafter, patients received rehabilitation in a primary care setting of their choice, which lasted for a median duration of 3 months (range 1-6 months) in the good outcome group and 5 months (range 1-6 months) in the poor outcome group.

The minimally detectable change (MDC) of KOOS knee-related QoL is reported to be 21.1 points one year after TKA [8, 22], and this was used as a cut-off to classify individuals postoperative outcome as either a “good outcome” (change ≥ MDC), or a “poor outcome” (change < MDC). The MDC is a statistical estimate that provides a threshold for interpretation of a measurement [23]. When a change score exceeds this threshold, there is reasonable certainty that it represents a true change, and not a measurement error [23]. The MDC is not an absolute value, and should be considered a guideline [23]. In this sample, 19 (68%) out of 28 patients reported change greater than the MDC in KOOS knee-related QoL at the one year follow-up, and were classified as having a good outcome (Fig. 1). Nine patients reported change smaller than 21.1 points in knee-related QoL and were classified as having a poor outcome. The good and poor outcome groups were analyzed separately and compared (Table 1).

A significance level was set at p <0.05 and data analyses were performed using IBM SPSS Statistics version 22 (Chicago, IL, USA). Depending on data distribution, means with standard deviation and medians with range, were used to describe the explored variables. Normal distribution of data was assessed using Shapiro-Wilk’s test and Q-Q plots. Sample size calculations were made a priori for the primary analysis with another aim [11]. In this secondary analysis, post hoc power analyses were performed with regards to change in knee gait biomechanics. The sample size of the good outcome group (n = 19) had sufficient power (range 93-99%) to detect significant differences in knee flexion-extension range, peak varus angle and peak valgus moment within the group [24]. Power for the corresponding variables within the poor outcome group (n = 9) was low (range 23-55%) [24]. Knee gait biomechanics, passive range of knee motion, and time and distance parameters were analyzed using a two-way repeated measures ANOVA with the within groups factor Time (prior to TKA and one year post TKA) and the between groups factor Group (the good outcome group and the poor outcome group) and the interaction Group*Time. The Group*Time interaction refers to the statistical test of whether the response profile for one group is the same as for the other group. In case of a significant interaction, simple effects were tested, i.e. effects of one factor holding the levels of the other factor fixed. To adjust for preoperative differences between groups an ANCOVA was performed. Receiver operating characteristic (ROC) curves were calculated to evaluate whether change in knee gait biomechanics could be used to correctly classify patients into either the good or poor outcome group. The area under the ROC curve (AUC) and 95% confidence intervals (CI) were calculated. An AUC of at least 0.70 was considered to be appropriate [25]. Differences in baseline data and change in VAS pain raw scores between the good and the poor outcome group were evaluated using independent samples t-tests and Mann Whitney U test, depending on data distribution. Fisher’s exact test was used to determine whether the proportion of patients differed between the groups with regards to the KL classification of radiographic severity of osteoarthritis.

The good outcome group presented with significantly less knee flexion-extension range (5°) during gait preoperatively (Fig. 2), compared to the poor outcome group (p = 0.004) (Table 2). The proportion of patients with less radiographic changes (a KL score of 3b) were larger in the poor outcome group, although not statistically different (p = 0.08) (Table 1). Preoperatively, the poor outcome group reported significantly lower scores in the KOOS ADL subscale (Table 1). In both the good and the poor outcome groups several patients had gone through previous knee arthroscopy. These arthroscopic surgeries were done for diagnostic purposes for some, and in some cases it was due to degenerative meniscal tears (the tears were not repaired, only resected). The number of participants with a previous history of knee arthroscopy were equally distributed in the two groups (11 out of 19 in the good outcome group (57%), and 5 out of 9 in the poor outcome group (56%).

The good outcome group displayed significant improvements in the majority of knee gait biomechanics variables (Fig. 2). During gait, peak knee flexion angle increased by 5°, knee flexion-extension range increased by 8°, peak varus angle was reduced by 4.8°, peak flexion moment increased by 0.08 Nm/kg, and peak valgus moment was reduced by 0.16 Nm/kg (Table 2). The good outcome group also displayed a significant increase in walking speed, a reduction (normalization) of stance phase duration (% of gait cycle), and increased passive range of knee extension by 5° (Table 2).

The poor outcome group displayed a significant reduction in peak varus angle during stance phase by 2.9° (p = 0.042) (Table 2). Aside from that, no other knee gait biomechanics outcomes, or passive knee joint range of motion showed any significant change one year after surgery (Table 2). Baseline data and postoperative results, for each individual classified as having a poor outcome, are presented separately (Table 3).

At the postoperative evaluation, there were no differences between the good and the poor outcome groups, other than a tendency for the good outcome group to demonstrate larger increases in peak flexion moment during stance phase (p = 0.054) (Table 2).

No differences were found between the groups with regards to change in perceived pain during gait trials assessed with VAS, or in perceived pain during gait trials at the postoperative assessment (Table 4).

Receiver operating characteristic (ROC) curves showed that smaller change in flexion-extension range, and change in peak flexion moment had a good ability to predict a poor outcome post TKA (Fig. 3). The AUC was 0.83 for change in flexion-extension range (CI 0.67 – 0.98), and 0.77 for change in peak flexion moment (CI 0.59 – 0.94). The ability to predict a poor outcome in knee related QoL for the other evaluated knee gait biomechanics outcomes were low (Fig. 3).

At the one year follow-up, both the good and the poor outcome groups presented with significantly lower passive range of knee flexion, lower peak flexion angle during gait, lower peak extension moment, and lower peak valgus moment compared to controls (Table 2). The good outcome group was comparable to the control group with regards to walking speed and stance phase duration, while the poor outcome group walked with a reduced walking speed and with a longer stance phase duration as compared to controls (Table 2). The poor outcome group had a significantly lower flexion-extension range, and lower peak flexion moment compared to the control group (Table 2).