Abstract
Aim
To study the effect of obesity on clinical and radiographic outcomes of computer-navigated knee arthroplasty.
Materials and methods
117 patients underwent primary computer-navigated total knee arthroplasty. Eight were lost to follow-up and 8 had incomplete data.
Results
Eighty-four (83.2%) female, 17 (16.8%) male patients age 65.3 ± 6.9 years with a pre-operative BMI 27.2 ± 4.1 (18.6–40.0) kg/m2, 7.3 ± 0.98 years follow-up. Forty-two (41.6%) had a BMI>27.5 kg/m2 indicative of obesity in Singapore. Post-operative radiographic alignment, 2-year Oxford knee scores and ROM were not significantly associated with BMI.
Conclusion
BMI is not a determinant of functional scores when computer navigation is used.
Keywords: Knee arthroplasty, Computer navigation, Obesity
1. Introduction
Computer aided total knee arthroplasty has been shown to improve the outcome in outliers with consistent results with the ability to provide the surgeon with intra-operative measurements of the patient's limb alignment.1 The intra-operative measurements derived from current imageless navigation systems are then used to determine the various bone cuts. Prior to the advent of computer navigation, standard radiographs were the means of obtaining pre –operative measurements of limb alignment to plan and perform a total knee arthroplasty with post-operative radiographs to evaluate the outcome of such surgery as alignment errors may affect implant function and lead to decreased survival of the implant.2
The accuracy of computer navigation derived intra-operative measurements is dependent on the surgeon's identification of landmarks during registration while measurements from radiographs are affected by the patient's weight-bearing status, limb positioning and inter-observer variability.3 Weight-bearing has been described by Gbejuade et al. to alter the alignment of the lower limb axis comparing supine CT scanograms with weight-bearing radiographs.4
The technical difficulties in performing a total knee arthroplasty in the overweight or obese patient include anaesthetic complications, difficult exposure and identification of landmarks leading to malalignment or malpositioning of implants and post-surgical wound complications and difficulty with rehabilitation.5,6 Instability has been described too in obese patients undergoing total knee arthroplasty.7 The morbidity of obesity is not limited to osteoarthritis and includes the metabolic syndrome including hyperlipidaemia, hypertension and diabetes mellitus. This further complicates the surgical management of such patients.
However through the use of computer navigation, Shetty et al. reported that there was no significant difference in post-operative lower limb alignment between the obese and non-obese.8 Likewise, Yogeesh et al. reported that there was no significant difference in operating time with computer navigated knee arthroplasty between the obese and non-obese.9
2. Aim
The aim of this study is to study the effect of body mass index on the radiographic pre- and post-operative lower limb long-axis with intra-operative measurements derived from computer navigation and clinical outcomes including range of motion and scoring performed by a single surgeon at a single centre.
3. Materials and Methods
Institutional Review Board approval was obtained for this study. Data was prospectively collected by our Orthopaedic Diagnostic Centre for 117 consecutive patients undergoing primary computer-navigated total knee arthroplasty using Ci Brainlab system with J&J PFC posterior stabilised implants by a single surgeon. For a meaningful important difference in the Oxford Knee Score of 5 points between two groups of patients as recommended by Beard et al., with a mean 2 year Oxford knee score of 18 and a standard deviation of 5.6 based on a previously presented unpublished data from our institution, with a 2-sided significance of 0.05 and a power of 0.8, 42 patients would be the required minimum sample size.10
Clinical data including pre-operative body mass index (BMI) was recorded. Obesity in Singapore is defined as a BMI of more than 27.5 kg/m2 and overweight as a BMI of more than 23 kg/m2 as per World Health Organization (WHO) guidelines for Asians.11 Our patients were thus stratified into 3 groups: BMI more than 23 kg/m2, BMI of more than 27.5 kg/m2 and BMI more than 30 kg/m2 which is the international definition of obesity.
Pre-operative and post-operative standing long-leg films and weight-bearing films of the knees were taken and the long-axis was measured by a single observer. Intra-operative computer navigation long-axis values were stored as screenshots intra-operatively after registration and after implant was cemented. Pre and post-operative range of motion (ROM), 36-Item Short Form Health Survey (SF-36) and Oxford knee scores were recorded by an independent observer. Patient satisfaction after surgery was recorded as well. Minimum 5-year follow-up. 8 patients were lost to follow-up and 8 patients had incomplete range of motion data.
4. Surgical technique
Surgery was performed either under a regional anaesthetic or general anaesthesia. After induction of anaesthesia, a tourniquet was applied round the thigh and sterile scrub performed. Tibia and femur navigation trackers were inserted percutaneously through stab incisions. Hip centre registration was performed first followed by a medial parapatellar arthrotomy to expose the knee joint. Femur, followed by tibia registration was then performed. The navigated tibia cutting guide was then used for a tibia-first technique followed by navigated gap balancing with a tensioner device to determine the femoral cuts. The patella was resurfaced in all cases. The knee was held in extension in neutral alignment with the aid of navigation trackers while waiting for cement to harden. Capsule and skin closure was performed with the knee held flexion.
Patients were allowed to stand on walk on the first post-operative day under supervision of our physiotherapist. All patients were placed on sequential compression devices for the calves for deep vein thrombosis prophylaxis. Subcutaneous enoxaparin was routinely administered to all obese and diabetic patients till discharge from the hospital ward.
Statistical analysis was performed with SPSS for Windows using the Chi-square test for categorical variables and the t-test for continuous variables.
5. Results
There were 84 (83.2%) female patients and 17 (16.8%) male patients with 49 (48.5%) left and 52 (51.5%) right knees with a mean follow-up duration of 7.3 ± 0.98 years (range 5.9–9.0). The mean age was 65.3 ± 6.9 years. Mean pre-operative BMI was 27.2 ± 4.1 (18.6–40.0) kg/m2 with 88 (87.1%) with BMI>23 kg/m2, 42 (41.6%) with BMI>27.5 kg/m2, 23 (22.8%) with BMI>30 kg/m2. At 2 years after surgery, BMI was 27.6 ± 3.9 (20.5–40.2) kg/m2 which was significantly different from the pre-operative BMI (p < 0.017). Mean operative duration was 96.0 ± 10.7 min.
Pre-operative radiographic axis was 9.3 ± 10.6° (−31.0 to 33.0) with 11 valgus knees (10.9%). Thirty (29.7%) patients had a pre-operative coronal plane deformity of more than 15°. Pre-operative radiographic axis significantly different from the initial intra-operative navigation axis after exposure and registration of landmarks (9.3 ± 10.6° vs. 6.5 ± 9.0°, p < 0.0) but the post-operative radiographic axis was not significantly different from the post-implantation navigation axis (0.22 ± 1.6° vs. −0.05 ± 3.0°, p < 0.36). 28/101 (27.7%) had a post-operative radiographic axis >3°.
There was significant improvement in knee extension, knee flexion, SF-36 and Oxford knee scores at 2 years (Table 1 and Fig. 1). No significant improvement in extensor lag and straight leg-raising at 2 years (Table 1). Thirty (42.3%) patients had a pre-operative radiographic axis deformity of more than 15°. This was not significantly associated with operative duration (95.0 ± 9.4 vs. 98 ± 12.9 min, p > 0.132).
Table 1.
Overall results at 2 years.
| Variable | Pre-op | 2 years post-op | p-value |
|---|---|---|---|
| Knee extension (°) | 8.7 ± 11.4 | 1.7 ± 4.9 | 0.0 |
| Knee flexion (°) | 119.3 ± 16.9 | 113.9 ± 15.2 | 0.0 |
| Fixed flexion deformity (°) | 3.3 ± 4.5 | 0.6 ± 2.1 | 0.0 |
| Extensor lag (°) | 0.3 ± 1.4 | 0.2 ± 1.0 | 0.48 |
| Straight-leg raising lag (°) | 0.4 ± 2.2 | 0.2 ± 1.0 | 0.35 |
| Oxford Knee score | 33.8 ± 8.1 | 18.9 ± 5.6 | 0.0 |
Fig. 1.
Pre-operative and 2-year post-operative SF-36 results.
We found that a pre-operative BMI more than 23 kg/m2 was significantly related to a longer operative duration (p < 0.021) (Table 2). A BMI more than 23 kg/m2 was not significantly associated with pre-operative Oxford knee score or pre-operative range of motion. Nor was it significantly related to 2-year Oxford knee score or range of motion at 2 years.
Table 2.
Pre-operative variables and 2-year results: BMI ≥23 kg/m2.
| Variable | BMI <23 kg/m2 n = 13 | BMI ≥23 kg/m2 n = 88 | p-value |
|---|---|---|---|
| Op duration (min) | 88.8 ± 10.8 | 97.1 ± 10.3 | 0.008 |
| Pre-operative Knee extension (°) | 4.8 ± 7.8 | 9.5 ± 11.9 | 0.19 |
| Pre-operative Knee flexion (°) | 121.3 ± 19.3 | 118.7 ± 16.7 | 0.63 |
| Pre-operative Fixed flexion deformity (°) | 2.2 ± 3.1 | 3.6 ± 4.8 | 0.33 |
| Pre-operative Extensor lag (°) | 1.3 ± 3.1 | 0.17 ± 0.91 | 0.26 |
| Pre-operative Straight-leg raising lag (°) | 2.0 ± 5.8 | 0.18 ± 1.0 | 0.30 |
| Pre-operative Oxford Knee score |
32.2 ± 6.4 |
34.0 ± 8.4 |
0.45 |
| 2-year Knee extension (°) | −0.2 ± 4.0 | 2.1 ± 5.0 | 0.12 |
| 2-year Knee flexion (°) | 116.7 ± 13.9 | 113.6 ± 15.3 | 0.50 |
| 2-year Fixed flexion deformity (°) | 0.15 ± 0.55 | 0.72 ± 2.2 | 0.36 |
| 2-year Extensor lag (°) | 0.77 ± 1.9 | 0.11 ± 0.75 | 0.24 |
| 2-year Straight-leg raising lag (°) | 0.62 ± 1.6 | 0.13 ± 0.83 | 0.29 |
| 2-year Oxford Knee score | 18.0 ± 2.1 | 19.0 ± 6.0 | 0.55 |
| Post-op radiographic axis >3° | 1/13 (7.7%) | 3/88 (3.4%) | 0.42 |
A pre-operative BMI greater than 27.5 kg/m2 was statistically significantly related to pre-operative extensor lag (p < 0.033) (Table 3). It was not significantly related to 2-year Oxford knee score. However, having a BMI greater or equal to 27.5 kg/m2 was statistically significant for 2-year post-operative knee extension (p < 0.005), flexion contracture (p < 0.024) and extensor lag (p < 0.044). Although the differences between the two groups may not be clinically significant as they represent an absolute difference of less than one degree.
Table 3.
Pre-operative variables and 2-year results: BMI ≥27.5 kg/m2
| Variable | BMI <27.5 kg/m2 n = 59 | BMI ≥27.5 kg/m2 n = 42 | p-value |
|---|---|---|---|
| Op duration (min) | 95.0 ± 11.4 | 97.5 ± 9.5 | 0.25 |
| Pre-operative Knee extension (°) | 10.5 ± 13.7 | 6.7 ± 7.0 | 0.72 |
| Pre-operative Knee flexion (°) | 121.4 ± 16.4 | 115.7 ± 17.3 | 0.93 |
| Pre-operative Fixed flexion deformity (°) | 4.0 ± 5.2 | 2.5 ± 3.5 | 0.91 |
| Pre-operative Extensor lag (°) | 0.52 ± 1.8 | 0.0 ± 0.0 | 0.033 |
| Pre-operative Straight-leg raising lag (°) | 0.69 ± 2.9 | 0.0 ± 0.0 | 0.075 |
| Pre-operative Oxford Knee score |
33.8 ± 8.8 |
33.8 ± 7.2 |
1.0 |
| 2-year Knee extension (°) | 2.9 ± 5.7 | 0.33 ± 3.3 | 0.005 |
| 2-year Knee flexion (°) | 115.7 ± 15.5 | 111.6 ± 14.4 | 0.18 |
| 2-year Fixed flexion deformity (°) | 1.0 ± 2.6 | 0.17 ± 0.82 | 0.024 |
| 2-year Extensor lag (°) | 0.34 ± 1.3 | 0.0 ± 0.0 | 0.044 |
| 2-year Straight-leg raising lag (°) | 0.32 ± 1.2 | 0.0 ± 0.0 | 0.050 |
| 2-year Oxford Knee score | 18.3 ± 4.8 | 19.6 ± 6.6 | 0.25 |
| Post-op radiographic axis >3° | 4/59 (6.8%) | 0/42 (0.0%) | 0.14 |
A pre-operative BMI of more than 30 kg/m2 was not significantly related to pre-operative Oxford knee score (Table 4). It was significantly related to pre-operative knee flexion (p < 0.02) and extensor lag (p < 0.033). At 2 years, a BMI of more than 30 kg/m2 was not significantly related to 2-year Oxford knee score. It was significantly related to 2-year extensor lag (p < 0.045). Once again, the differences between the two groups may not be clinically significant as they represent an absolute difference of less than one degree except for the difference in pre-operative knee flexion.
Table 4.
Pre-operative variables and 2-year results: BMI ≥30 kg/m2.
| Variable | BMI <30 kg/m2 n = 78 | BMI ≥30 kg/m2 n = 23 | p-value |
|---|---|---|---|
| Op duration (min) | 95.8 ± 10.8 | 97.0 ± 10.5 | 0.64 |
| Pre-operative Knee extension (°) | 9.7 ± 12.5 | 6.1 ± 6.9 | 0.19 |
| Pre-operative Knee flexion (°) | 121.2 ± 16.6 | 111.8 ± 16.7 | 0.020 |
| Pre-operative Fixed flexion deformity (°) | 3.7 ± 5.0 | 2.3 ± 3.2 | 0.099 |
| Pre-operative Extensor lag (°) | 0.39 ± 1.6 | 0.0 ± 0.0 | 0.033 |
| Pre-operative Straight-leg raising lag (°) | 0.52 ± 2.5 | 0.0 ± 0.0 | 0.075 |
| Pre-operative Oxford Knee score |
34.1 ± 8.2 |
32.6 ± 7.8 |
0.44 |
| 2-year Knee extension (°) | 2.3 ± 5.2 | 0.2 ± 3.5 | 0.069 |
| 2-year Knee flexion (°) | 114.8 ± 16.0 | 111.3 ± 11.7 | 0.33 |
| 2-year Fixed flexion deformity (°) | 0.78 ± 2.3 | 0.22 ± 1.0 | 0.10 |
| 2-year Extensor lag (°) | 0.26 ± 1.1 | 0.0 ± 0.0 | 0.045 |
| 2-year Straight-leg raising lag (°) | 0.25 ± 1.1 | 0.0 ± 0.0 | 0.051 |
| 2-year Oxford Knee score | 18.3 ± 4.7 | 20.7 ± 7.8 | 0.70 |
| Post-op radiographic axis >3° | 4/78 (5.1%) | 0/23 (0.0%) | 0.57 |
A post-operative radiographic axis deviation of more than 3° was not significantly related to 2-year Oxford knee score (Table 5). It was significantly related to extensor lag (p < 0.045) at 2 years and was not significantly related to pre-operative BMI.
Table 5.
Post-operative radiographic axis alignment.
| Variable | Post-operative Radiographic Axis |
p-value | |
|---|---|---|---|
| within 3° n = 78 | more than 3° n = 23 | ||
| 2-year Knee extension (°) | 1.5 ± 4.2 | 2.6 ± 6.6 | 0.35 |
| 2-year Knee flexion (°) | 114.7 ± 14.0 | 112.1 ± 17.8 | 0.45 |
| 2-year Fixed flexion deformity (°) | 0.50 ± 1.5 | 1.0 ± 3.1 | 0.38 |
| 2-year Extensor lag (°) | 0.28 ± 1.2 | 0.0 ± 0.0 | 0.045 |
| 2-year Straight-leg raising lag (°) | 0.26 ± 1.1 | 0.0 ± 0.0 | 0.051 |
| 2-year Oxford Knee score | 18.9 ± 5.9 | 18.8 ± 5.1 | 0.96 |
| BMI more than 23 kg/m2 | 62/73 (84.9%) | 26/28 (92.9%) | 0.51 |
| BMI more than 27.5 kg/m2 | 31/73 (42.4%) | 11/28 (39.2%) | 0.83 |
| BMI more than 30 kg/m2 | 16/73 (21.9%) | 7/28 (25.0%) | 0.79 |
No revisions were performed. There were 2 (2.0%) patients with prosthetic joint infections, one with a BMI of 30.9 kg/m2 which presented 5 years after the index surgery with cultures growing Bacteroides Fragilis requiring debridement, washout and change of liner and long term antibiotics suppression as patient did not want a revision and one other patient, BMI 22.1 kg/m2, who presented 4 years after index surgery with a knee aspirate growing Staphylococcus Lugdunensis also on long-term antibiotic suppression. Both patients did not have diabetes mellitus. One patient died 6 years after surgery.
6. Discussion
In our study, we have found that there was a statistically significant difference between pre-operative radiographic long-axis measurements and intra-operative navigation measurements while the post-operative long-axis measurements correlate with the intra-operative, navigation-derived, long-axis after implant insertion. It is accepted that radiographic measurements are accurate in normal patients, however, standard positioning may be difficult in patients with joint deformity such as patient with osteoarthritis.12
We have found that pre-operative radiographs exaggerate deformity compared with navigation measurements which is in agreement with the findings reported by Yaffe et al. who reported a discrepancy of 4.66° up to a 12° difference using the Aesculap Orthopilot system.3 The influence of simulated weight-bearing of half body-weight on cadaveric limbs was reported by Kendoff et al. to cause a deviation of 0.4°.13 This deviation in mechanical axis was reported by Specogna to be greater during single-leg standing compared with double-leg standing and supine.12 Likewise, although we had only 4 patients with a computer navigated axis deviation of more than 3° post-implantation, this was exaggerated on weight-bearing with 28 patients with a radiographic mechanical axis deviation. We recognise that a post-operative mechanical axis deviation has been reported by Huang et al. to affect functional outcome though we did not find this at 2 years in our patient population.2
We found that a high pre-operative BMI was not associated with clinically significant differences in functional scores and range of motion after a computer navigated total knee replacement. Few studies compare functional outcomes of obese versus non-obese undergoing computer navigated total knee arthroplasty. Our results are consistent with Shetty et al. who found that obesity did not affect post-operative lower limb alignment after a computer navigated total knee arthroplasty and Yogeesh et al. who had better alignment and a shorter operating time in obese patients using computer navigation compared to conventional technique.8,14 This contrasts with Jarvepaa et al.'s series of 48 patients who underwent a conventional total knee arthroplasty where his obese patients had a poorer outcome.5 Millar et al. reported that there was less blood loss with computer navigation compared to conventional technique in the obsess patient.15 Despite computer navigation, the obese patient presents with technical challenges on table as illustrated by an additional 8.3 min of operating time in our patients with a BMI greater than 23 kg/m2. We found too that extensor lag and weakness in straight-leg raising which was present pre-operatively persisted post-operatively.
Conventional total knee arthroplasty is a reliable and predictable procedure. We find that computer navigation is useful when exposure and landmarks to assess alignment are difficult such as in the obese patient where determining the transepicondylar axis of the femur is difficult in a deep wound or the conventional tibia cutting guide won't sit against the bone due to thick subcutaneous tissue as we do not have to display all bony landmarks simultaneously to determine alignment. Electricwala et al. reported more obese patients required a revision at 5 years than non-obese patients.6 We had 2 patients, one obese and one non-obese who needed a revision for infection but were on long term antibiotic suppression.
7. Conclusion
Computer navigated total knee arthroplasty can be a useful tool to overcome technical challenges in the obese patient to improve functional and radiographic outcomes.
Compliance with ethical standards
Funding
There is no funding source.
Ethical approval
Institutional Review Board approval was obtained for this study.
Declaration of competing interest
The authors declare that they have no conflict of interest.
References
- 1.Todesca A., Garro L., Penna M., Bejui-Hugues J. Conventional versus computer-navigated TKA: a prospective randomized study. Knee Surg Sport Traumatol Arthrosc. 2017;25:1778–1783. doi: 10.1007/s00167-016-4196-9. [PMID: 27306985 DOI: [DOI] [PubMed] [Google Scholar]
- 2.Huang N.F.R., Dowsey M.M., Ee E., Stoney J.D., Babazadeh S., Choong P.F. Coronal alignment correlates with outcome after total knee arthroplasty. Five-year follow-up of a randomized controlled trial. J Arthroplast. 2012;27:1737–1741. doi: 10.1016/j.arth.2012.03.058. [PMID: 22868073 DOI: [DOI] [PubMed] [Google Scholar]
- 3.Yaffe M.A., Koo S.S., Stulberg S.D. Radiographic and navigation measurements of TKA limb alignment do not correlate. Clin Orthop Relat Res. 2008;466:2736–2744. doi: 10.1007/s11999-008-0427-9. [PMID: 18752032 DOI: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Gbejuade H.O., White P., Hassaballa M., Porteous A.J., Robinson J.R., Murray J.R. Do long leg supine CT scanograms correlate with weight-bearing full-length radiographs to measure lower limb coronal alignment? The Knee. 2014;21:549–552. doi: 10.1016/j.knee.2013.05.009. [PMID: 23849161 DOI: [DOI] [PubMed] [Google Scholar]
- 5.Jarvenpaa J., Kettunen J., Soininvaara T., Miettinen H., Kroger H. Obesity has a negative impact on clinical outcome after total knee arthroplasty. Scand J Surg. 2012;101:198–203. doi: 10.1177/145749691210100310. [PMID: 22968244. [DOI] [PubMed] [Google Scholar]
- 6.Electricwala A.J., Jethanandani R.G., Narkbunnam R. Elevated body mass index is associated with early total knee revision for infection. J Arthroplast. 2017;32:252–255. doi: 10.1016/j.arth.2016.05.071. [PMID: 27421585 DOI: [DOI] [PubMed] [Google Scholar]
- 7.Tripathi M.S., Heinle C.C., Manaqibwala M.I., Tria A.J. The utility of increased constraint in primary total knee arthroplasty for obese patients. Orthop Clin N Am. 2016;47:51–55. doi: 10.1016/j.ocl.2015.08.007. [PMID: 26614920 DOI: [DOI] [PubMed] [Google Scholar]
- 8.Shetty G.M., Mullaji A.B., Bhayde S., Lingaraju A.P. No effect of obesity on limb and component alignment after computer-assisted total knee arthroplasty. The Knee. 2014;21:862–865. doi: 10.1016/j.knee.2014.04.004. [PMID: 24799079 DOI: [DOI] [PubMed] [Google Scholar]
- 9.Kamat Y.D., Aurakzai K.M., Adhikari A.R., Matthews D., Kalairajah Y., Field R.E. Does computer navigation in total knee arthroplasty improve patient outcome at midterm follow-up? Int Orthop. 2009;33:1567–1570. doi: 10.1007/s00264-008-0690-0. [PMID: 19034445. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Beard D.J., Harris K., Dawson J. Meaningful changes for the Oxford hip and knee scores after joint replacement surgery. J Clin Epidemiol. 2015;68:73–79. doi: 10.1016/j.jclinepi.2014.08.009. [PMID: 25441700 DOI: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lee Y.S., Biddle S., Chan M.F. Health promotion board–ministry of health clinical practice guidelines: Obesity. Singap Med J. 2016;57:292–300. doi: 10.11622/smedj.2016103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Specogna A.V., Birmingham T.B., Hunt M.A. Radiographic measures of knee alignment in patients with varus gonarthrosis. Am J Sports Med. 2007;35:65–70. doi: 10.1177/0363546506293024. [DOI] [PubMed] [Google Scholar]
- 13.Kendoff D., Citak M., Pearle A. Influence of lower limb rotation in navigated alignment analysis: implications for high tibial osteotomies. Knee Surg Sport Traumatol Arthrosc. 2007;15:1003–1008. doi: 10.1007/s00167-007-0308-x. [PMID: 17364202 DOI: [DOI] [PubMed] [Google Scholar]
- 14.Kamat Y.D., Aurakzai K.M., Adhikari A.R. Total knee replacement in the obese patient: comparing computer assisted and conventional technique. Sci World J. 2014;2014 doi: 10.1155/2014/272838. [PMID: 24523634. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Millar N.L., Deakin A.H., Millar L.L., Kinnimonth A.W.G., Picard F. Blood loss following total knee replacement in the morbidly obese: effects of computer navigation. The Knee. 2011;18:108–112. doi: 10.1016/j.knee.2010.03.002. [PMID: 20591675 DOI: [DOI] [PubMed] [Google Scholar]