Unicompartmental knee arthroplasty has received renewed interest in recent years. Placement of a unicompartmental knee implant involves less dissection of soft tissue, less removal of bone, and better preservation of knee anatomy. In comparison with total knee arthroplasty, unicompartmental knee arthroplasty has been associated with certain short-term advantages, such as shorter hospitalization and faster recovery1,2. Less morbidity in the form of less postoperative pain, less infection, less thromboembolic disease, and better range of motion has also been observed following unicompartmental knee arthroplasty1,3-5. The longevity of unicompartmental knee implants has also been found to compare well with that of total knee implants in single-center studies2,6 and in randomized clinical trials7. Indications for the use of unicompartmental knee arthroplasty are, however, debated, as the rate of revision following unicompartmental knee arthroplasty has been reported in register studies to be approximately twice as high as that following total knee arthroplasty5,8,9. The higher revision rates for unicompartmental knee implants have been shown to be mainly due to aseptic loosening, pain, and periprosthetic fracture5. Since the better short-term results of unicompartmental knee arthroplasty should be weighed against the possible higher risk of revision5, more knowledge of the patient's perception of pain and function after knee arthroplasty is needed.
The aim of this study was to compare the levels of pain and function at a minimum of two years following surgery among patients who underwent unrevised total knee arthroplasty or unicompartmental knee arthroplasty.
The Norwegian Arthroplasty Register
Practically all patients (99%) in Norway who undergo primary arthroplasty of the knee are reported to the Norwegian Arthroplasty Register10. The register was established in 1987 as a hip prosthesis register, but, in 1994, its range was extended to the registration of all artificial joints, including knee implants11,12. The Norwegian Arthroplasty Register receives information directly from orthopaedic surgeons. Information on patient-related outcome, such as pain and function, is not reported to the register and was therefore obtained in this study through a mail survey of selected patients from the register.
Participants
Possible participants were patients registered in the Norwegian Arthroplasty Register with at least one unrevised cemented primary unicompartmental knee implant or total knee implant that was inserted as a result of primary osteoarthritis. The study patients had to be eighty-five years of age or less when asked to participate in the study, and the procedure had to have been performed at least two years prior to the survey to ensure that the result of the intervention had stabilized13-15.
The operation volume was set to a minimum of at least 100 registered operations for each prosthesis brand. All eligible patients with a Genesis Uni (Smith & Nephew, Memphis, Tennessee) (n = 136) or a Miller-Galante all polyethylene tibial Uni (Zimmer, Warsaw, Indiana) (n = 129) prosthesis were included. To limit the number of patients asked to participate in the survey, 200 patients with Oxford III (Biomet, Bridgend, South Wales, United Kingdom) prostheses were selected randomly among 956 eligible patients. In total, 465 individuals with a unicompartmental knee arthroplasty were asked to participate. Total knee prosthesis brands were included if the number of registrants associated with a particular brand included at least 100 patients who underwent patellar resurfacing and 100 patients who did not undergo patellar resurfacing. The total knee prosthesis brands were AGC (Biomet), Genesis I (Smith & Nephew), LCS (DePuy, Leeds, United Kingdom) and NexGen (Zimmer). All available brands of prostheses with patellar components were included (134 AGC, 186 Genesis I, 238 LCS, and 112 NexGen) while a subset of prostheses without a patellar component (134 AGC, 180 Genesis I, 238 LCS, and sixty-two NexGen) were randomly chosen from all eligible implants without a patellar component (715 AGC, 931 Genesis I, 1766 LCS, and 152 NexGen), and were group-matched by prosthesis brand and year of operation. It was not possible to match all NexGen prostheses with a patellar component with corresponding NexGen prostheses without a patellar component because patella resurfacing was almost exclusively used early in the study period. Hence, only sixty-two patients who received a NexGen prosthesis without a patellar component were included. A detailed account of the selection procedure is given in Figure 1. We did not differentiate between total knee arthroplasties with patellar resurfacing and those without patellar resurfacing because a previous study from this same register16 did not demonstrate any difference in pain or function between the two groups.
A total of 1749 individuals were then invited to participate in the survey. A reminder was sent out after two months to those who failed to respond to the initial questionnaire. In all, 1344 patients completed the questionnaire, 393 either declined or did not respond, and twelve patients were deceased or unable to be located by the post office.
Questionnaire
To calculate the knee score, we used the valid and reliable, self-administered and knee-specific Knee Injury and Osteoarthritis Outcome Score (KOOS) questionnaire17. A Norwegian translation from the Swedish version of the KOOS was developed for this study and was approved as the official Norwegian translation. A description of the validation process of this translation can be found on the KOOS web site (http://www.koos.nu)17.
To assess the potential effect of general health factors, the questionnaire also included questions for the purpose of calculating the Charnley category as modified to knee arthroplasty18,19 and for the purpose of calculating the preoperative and most recent EuroQol-5D (EQ-5D) index scores20. The EQ-5D questionnaire is a valid and reliable non-disease-specific instrument for measuring quality of life. In addition, questions asking about the patient's satisfaction with the surgery and the degree of pain from the involved knee were included. With the exception of the latter two questions, which were presented in the form of a visual analog scale (VAS), all questions had standardized answer options that were presented in the form of a Likert scale.
The study was approved by the Norwegian Data Inspectorate, Norwegian Social Science Data Services (date of issue: 04/25/2006; registration number: 2005/453-2), and The Regional Committee for Research Ethics in Western Norway (date of issue: 02/23/2006; registration number: 046:06). The patients received the questionnaire together with an information letter, and they returned the questionnaire to the register with a signed consent to participate in the study.
Outcome Measures
The KOOS questionnaire, the result of which was used as the primary outcome regarding the patient's perception of pain and function, consists of forty-two individual questions, comprising five subscales: pain, other symptoms (Symptoms), function in daily living (activities of daily living), function in sport and recreation, and knee-related quality of life (QOL). Patients were asked to consider only the most recent week when answering most of the questions, and each question could be scored on a range from 0 to 4. A normalized score (with 100 indicating no symptoms and 0 indicating extreme symptoms) was calculated for each subscale. Calculation of the scores and treatment of missing data were done in accordance with stated guidelines (http://www.koos.nu)17.
In addition, we used "pain from the operated knee" (Pain) and "satisfaction with the operation" (Satisfaction) as outcomes, both as measured on a VAS. In the analyses, these VAS scores were reversed, with 100 indicating the best possible state and 0 indicating the worst possible state. Improvement in health-related quality of life (?EQ-5D), calculated as the difference between the present and preoperative EQ-5D index scores multiplied by 100, was also used as an outcome.
To further investigate possible underlying mechanisms in the performances of the two groups, we also studied all forty-two KOOS questions (in which a score of 0 indicated the best possible state and a score of 4 indicated the worst possible state) as outcomes.
Statistics
The minimal perceptible clinical difference is 8 to 10 units for KOOS subscales17 and 9 to12 units for a VAS21. When assessing the detailed KOOS questions, a 10% difference between the two groups, which corresponded to 0.4 unit, was assumed to be clinically significant. To have an 80% chance of detecting as significant (at the two-sided 5% level) a ten-point difference in mean KOOS subscales, with an assumed standard deviation of twenty, sixty-four individuals in each treatment group were required. Thus, to ensure good representation for both groups of treatment, a minimum restriction on operation volume of each brand was set to at least 100 registered operations.
Differences in age, the amount of time that had passed from the time of the operation until the time that the patient completed the form, and the preoperative EQ-5D index score were analyzed with the independent-samples Student t test, whereas differences according to sex and Charnley category were analyzed with the Pearson chi-square test. Multiple logistic regression was used to analyze response rates for sex (with male sex used as the reference), age, time since the operation, and treatment option (with total knee arthroplasty used as the reference). To estimate differences in mean outcome scores for unicompartmental knee arthroplasty and total knee arthroplasty, we used multiple linear regression, with adjustment for possible confounding by sex, age (younger than sixty-five years, sixty-five to sixty-nine years, seventy to seventy-nine years, and eighty years or older), time since operation, and Charnley category. Crude and adjusted differences in mean scores are presented with 95% confidence intervals and p values.
Multiple linear regression was also used to investigate a possible association between mean outcome scores and the different unicompartmental brands, with adjustment for sex, age, Charnley category, and time since operation.
In the analyses, a p value of <0.05 was considered significant. To account for chance findings, the critical p value was reduced to 0.00119, in accordance with the Bonferroni correction method, when exploring the forty-two detailed KOOS questions.
Source of Funding
No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. The study was funded by the Norwegian Rheumatism Association, which has been financed with the aid of extra funds from the Norwegian Foundation for Health and Rehabilitation. This was an external funding source that provided funding for salary for the study.
We compared the quality of unicompartmental knee arthroplasty and total knee arthroplasty on the basis of data retrieved from the Norwegian Arthroplasty Register and with use of patient-reported degrees of pain and function as the outcome. We observed only small or no differences in pain and function between unicompartmental knee arthroplasty and total knee arthroplasty at least two years following surgery (mean, 6.5 years). Some of the differences were significant in favor of unicompartmental knee arthroplasty but did not reach the level of a minimal perceptible clinical difference. Analysis of the detailed questions from the KOOS indicated, however, that patients in the two groups experienced pain and function differently, with better range of motion as the main advantage of unicompartmental knee arthroplasty.
Better performance after unicompartmental knee arthroplasty in activities that involve bending of the knee is probably related to better stability, as this procedure preserves more of the anatomy of the knee, such as the cruciate ligaments. More prostheses could be expected to be counted as nonfunctional if the end point in survival analysis were expanded to include not only revisions but also painful joints14. A higher proportion of untreated loose prosthesis components after unicompartmental knee arthroplasty may explain why our findings indicated no advantage of unicompartmental knee arthroplasty with regard to pain. Progression of arthritis has also been suggested as an important reason for revision of unicompartmental knee arthroplasty22,23 and may affect our findings in the same way when not treated.
An alternative to evidence from randomized clinical trials is evidence from large observational studies. The patient selection in our study was based on information reported to the Norwegian Arthroplasty Register. This made it possible to include large volumes of patients with several prosthetic designs implanted by many surgeons and from hospitals performing various amounts of surgery. Thus, the external validity of the results might be higher than the validity of those from randomized clinical trials. Observational studies may be affected by different forms of bias. We have treated the most common confounding factors by making adjustments in the statistical model, but there may still be variables that have not been taken into account. Preoperative differences between the treatment groups may have biased our results. Better preoperative range of motion for unicompartmental knee prostheses may have caused the results of unicompartmental knee arthroplasty to be too positive and thereby may have increased the differences in favor of unicompartmental knee arthroplasty. Thus, any advantages of unicompartmental knee arthroplasty may be smaller than indicated by the observed differences but would not have altered our conclusion. A bias in the opposite direction may have occurred if patients with bi-compartmental or tri-compartmental arthritis were treated with unicompartmental knee arthroplasty. In the study period, however, use of criteria defined by the Oxford group2 was recommended to identify patients who were suitable for treatment with unicompartmental knee arthroplasty. Hence, such a bias was not likely, since treatment with total knee arthroplasty was the only recommended option for patients with a more severe stage of the disease as determined on the basis of the Ahlbäck radiographic classification24,25 (classification 1 to 5 for total knee arthroplasty, and classification 1 to 3 for consideration of unicompartmental knee arthroplasty). Total knee arthroplasty is the main available treatment of a failed unicompartmental knee arthroplasty, and bias may have been introduced if knees that were treated with a unicompartmental knee arthroplasty were more likely to be revised for identical indications, for example pain, than were knees that were treated with a total knee arthroplasty. Then a falsely low proportion of individuals with poor results after unicompartmental knee arthroplasty would have been eligible for inclusion. If so, the results of unicompartmental knee arthroplasty would be too positive. However, this would most likely not have altered our findings of no or only small differences between total knee arthroplasty and unicompartmental knee arthroplasty. When comparing unicompartmental knee arthroplasty and total knee arthroplasty, the selection of prosthesis brands might play an important role. An adequate range of commonly used implants is represented in our material and is expected to offer a good representation of hospitals, surgical techniques, and surgeon experience for both treatment options. Since the outcomes in our study, except for the improvement in health-related quality of life (?EQ-5D), were based on the patient's perception of pain and function experienced in the last week, the risk of recall bias is assumed to be negligible. Despite not reaching the level of a minimal perceptible clinical difference, it may be that observed differences in outcomes used in this study could have clinical relevance when assessed in combination with physical-examination outcomes.
To our knowledge, there has been only one randomized clinical trial comparing unicompartmental knee arthroplasty and total knee arthroplasty1,7. This study included 102 knees with fifteen-year follow-up and showed a tendency for better range of motion after unicompartmental knee arthroplasty as compared with total knee arthroplasty. However, a similar number of knees in the two groups were classified as either good or excellent. These findings compare well with the results in our study. No significant differences in survival between the groups were found when revision or pain assessment (Bristol knee score < 60) was used as the end point. However, as only a single brand of arthroplasty implant was used in each treatment group, the external validity of the results may be limited. Observational studies from arthroplasty registers that have focused on pain and function after knee arthroplasty are few. A study from the Swedish Knee Arthroplasty Register did not find any difference in the proportions of satisfied patients in the two treatment groups26. Using risk of reoperation as the outcome, several register-based studies have reported significant higher revision rates for unicompartmental knee arthroplasty as compared with total knee arthroplasty5,8,9. A study from a specialized center2 did not report these differences. The latter results may be affected by strict exclusion criteria, through which some of the most challenging unicompartmental knee arthroplasty interventions were left out. The limited number of surgeons and implant designs involved may also explain why no differences were found.
Use of unicompartmental knee arthroplasty is disputed, and comparative studies of the midterm and long-term quality of the two treatment options are few and inconsistent. It has been suggested that unicompartmental knee arthroplasty may be preferable to total knee arthroplasty in most patients with anteromedial osteoarthritis and an intact anterior cruciate ligament2, thus indicating that unicompartmental knee arthroplasty is suitable in as many as 20% to 30% of the patients considered for knee arthroplasty. Others suggest additional restrictions in the selection procedure, resulting in a reduced percentage of suitable patients27. Data from randomized clinical trials are currently unavailable, and further research is needed.
In conclusion, with similar levels of pain and function but with a higher risk for revision for unicompartmental knee arthroplasty, total knee arthroplasty is still an excellent option for some patients with isolated disease. A better ability to bend the knee, however, may suggest a preference for unicompartmental knee arthroplasty in patients with special need for greater range of motion.
Note: The authors thank Ms. Eva Dybvik, Ms. Kaia Furnes, Ms. Ronja Furnes, and Mr. Ørjan Sørnes for their registration of questionnaires. Furthermore, we thank Mrs. Inger Skar, Secretary of the Norwegian Arthroplasty Register, for her accurate registration of each knee replacement reported. We also extend our thanks to the Norwegian orthopaedic surgeons for reporting their cases to the register, and to all of the patients who filled in the forms for this study. The study was supported by the Norwegian Rheumatism Association and has been financed with the aid of EXTRA funds from the Norwegian Foundation for Health and Rehabilitation.