Article Text

Download PDFPDF

Long-term outcome following total knee arthroplasty: a controlled longitudinal study
  1. J Cushnaghan1,
  2. J Bennett1,
  3. I Reading1,
  4. P Croft2,
  5. P Byng1,
  6. K Cox1,
  7. P Dieppe3,4,
  8. D Coggon1,
  9. C Cooper1,4
  1. 1
    MRC Epidemiology Resource Centre, University of Southampton, Southampton, UK
  2. 2
    Primary Care Sciences Research Centre, University of Keele, Keele, UK
  3. 3
    MRC Health Services Research Collaboration, University of Bristol, Bristol, UK
  4. 4
    Institute of Musculoskeletal Sciences, University of Oxford, Oxford, UK
  1. Professor C Cooper, MRC Epidemiology Resource Centre, Southampton General Hospital, Tremona Road, Southampton, SO16 6YD, UK; cc{at}


Objectives: To assess long-term outcome and predictors of prognosis following total knee arthroplasty (TKA) for osteoarthritis.

Methods: We followed-up 325 patients from 3 English health districts approximately 6 years after TKA, along with 363 controls selected from the general population. Baseline data, collected by interview and examination, included age, sex, comorbidity, body mass index (BMI), functional status and preoperative radiographic severity of osteoarthritis. Functional status at follow-up was assessed by postal questionnaire. Predictors of change in physical function were analysed by linear regression.

Results: Between baseline and follow-up, patients reported an improvement of 6 points in median Short Form 36 Health Survey (SF-36) physical function score, whereas in controls there was a deterioration of 14 points (p<0.001). Median SF-36 vitality score declined by 10 points in patients and 5 points in controls (p = 0.005), while their median SF-36 mental health scores improved by 12 and 13 points, respectively (p = 0.2). The improvement in physical function was smaller in patients who were obese than in patients who were non-obese, but compared favourably with a substantial decline in the physical function of obese controls. Better baseline physical function and older age predicted worse changes in physical function in patients and controls. Improvement in physical function tended to be greater in patients with more severe radiological disease of the knee, and was less in those who reported pain at other joint sites at baseline.

Conclusions: Improvements in physical function following TKA for osteoarthritis are sustained beyond 5 years. The benefits are apparent in patients who are obese as well as non-obese, and there seems no justification for withholding TKA from obese patients solely on the grounds of their body mass index.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Osteoarthritis of the knee is a common cause of pain and disability at older ages.1 When severe, it is often treated by total knee arthroplasty (TKA), and it is the main reason2 for the 55 000 primary knee replacements that are performed annually in England.3 There is good evidence that in the short term, TKA is a safe and cost-effective method of alleviating pain and restoring physical function in patients who do not respond to non-surgical therapies.4 However, evidence on longer-term outcome is less extensive, most studies having followed patients for only 6–12 months after surgery.5 Moreover, there are uncertainties about which patients are likely to benefit most from the procedure. In particular, at least one National Health Service (NHS) trust in the UK has been reported to apply a policy that TKA should not be offered to patients with a body mass index (BMI) >30 kg/m2.6 Obesity is a strong risk factor for knee osteoarthritis,7 and many patients with the disease have high BMIs. However, it does not necessarily follow that patients who are obese will benefit less from TKA than others.

To obtain further information about long-term outcome following TKA and about its predictors preoperatively, we followed up participants from an earlier case–control study of knee osteoarthritis 2–8 years after surgery.


We studied 657 matched pairs of patients and controls who had participated in an earlier case–control study of knee osteoarthritis.7 8 The patient group comprised residents of three English health districts (Portsmouth, Southampton and Stoke-on-Trent) aged 45 years or older, who were placed on a waiting list for TKA because of primary knee osteoarthritis over a 2-year period from 1995 to 1997. Patients with an inflammatory cause for osteoarthritis were excluded. The controls, none of whom had previously undergone knee surgery for osteoarthritis, were selected from the general population and were matched to the patients for age, sex and general practice.

At the time of the case–control study (baseline), patients and controls were interviewed using a structured questionnaire, which among other things collected information about: sex; age; smoking habits; previous injury to the knees leading to consultation with a doctor; pain (on most days of the past month) in the hands, shoulders and hips; and comorbidity from medically diagnosed diabetes, hypertension and thyroid disease. It also included sections of the Short Form 36 Health Survey (SF-36)9 10 relating to physical function, vitality and mental health. In addition, the interviewer measured the subject’s height (using a portable stadiometer) and weight (using electronic scales) from which BMI was calculated, and examined his or her hands for the presence of Heberden nodes. Routine anterior-posterior (A-P) knee radiographs of the patients that were listed for surgery were reviewed by a trained research assistant, who scored them for overall severity of osteoarthritis according to the Kellgren and Lawrence grading system.11

During 2001–2004, we wrote to all of the original study group who could be contacted, asking them to complete a self-administered questionnaire, which sought confirmation of any knee replacement surgery with the year(s) in which this had been carried out, and again included sections of the SF-36 questionnaire concerning physical function, vitality and mental health. In addition, we sought permission from the patients to review their hospital records. Where permission was given and the records were available, we abstracted details of the knee surgery, including the type of prosthesis, and whether there had been any subsequent revision. Ethics committee review was undertaken in all three health districts, and approval granted.

Statistical analysis, which was carried out with STATA V.9.2 software (StataCorp, College Station, Texas, USA), was restricted to patients who underwent TKA after the baseline interview and at least 2 years before completing the follow-up questionnaire, and to controls who were not treated by TKA during the follow-up period. SF-36 scores at baseline and follow-up were compared using the Mann–Whitney U test and Wilcoxon signed rank sum test. Baseline predictors of change in SF-36 physical function score over follow-up were explored by linear regression, using data from the subset of subjects who provided complete information on all relevant variables.


From the 657 matched pairs who were recruited at baseline, 363 patients (55%) and 367 controls (56%) were successfully contacted at follow-up. Among the 584 non-responders, 236 (121 patients and 115 controls) had died, 170 (87 patients and 83 controls) could not be contacted because they had moved or were no longer registered with the same general practitioner, 37 (18 patients and 19 controls) were too ill to complete the questionnaire, 131 (65 patients and 66 controls) declined to participate and 10 (3 patients and 7 controls) had other reasons for non-response.

Of the 363 patients who were successfully followed-up, 32 were excluded from analysis because they had not undergone TKA, 2 because the baseline interview was not conducted until after their TKA and 4 because the interval between TKA and completion of the follow-up questionnaire was less than 2 years. We also excluded four controls who reported receiving a TKA during the follow-up period. This left 325 cases and 363 controls who were analysed.

Table 1 shows the completeness of follow-up and of inclusion in the analysis according to various characteristics of the study participants at baseline. Older patients and controls experienced higher mortality during the follow-up period, but other than by age, there were no major differences between those analysed and the baseline sample.

Table 1 Distribution of patients and controls and completeness of follow-up according to baseline characteristics

Of the 325 patients included in the analysis, most (n = 238; 73%) underwent TKA within 2 years of the baseline interview, the maximum interval being 5 years. Follow-up of the patient group was at a mean interval of 6.9 years (median 7.0, range 5.4–7.9 years) from the baseline interview, and at an average of approximately 6 years (range 2–8 years) after their TKA. They included 112 men and 213 women, with a mean age at follow-up of 69.8 years.

The 363 controls (139 men and 224 women) were followed-up at a mean of 6.5 years after baseline interview (median 6.5, range 5.2–7.7 years). Their mean age at follow-up was 69.9 years.

Hospital records were accessible for 53 (16%) of the patients. The prostheses used most frequently were the Genesis (33 patients; Smith & Nephew, London, UK) and Low Contact Stress (13 patients; De Puy, Warsaw, Indiana, USA). Three arthroplasties were documented as requiring subsequent revision.

Figure 1 compares the SF-36 scores of patients and controls at baseline and at follow-up for physical function, vitality and mental health. At baseline, the patient group reported markedly worse physical function than the controls (median score 20 v 89, p<0.001); vitality was significantly better in the patients (median score 60 v 55, p = 0.003); and the median scores for mental health in the two groups were the same (64). By the time of follow-up, the median physical function score of the patients had improved by 6 points (p<0.001), while that of the controls had deteriorated by 14 points (p<0.001), the net effect being to narrow the gap in scores between the two groups. The difference in change from baseline among the patients as compared with controls was highly significant statistically (p<0.001). By contrast, vitality declined to a significantly greater extent during follow-up in the patient group than the controls (changes in median scores –10 and –5 points respectively, p value for difference in change between patients and controls = 0.005), while mental health improved to a similar extent in both groups (p = 0.2).

Figure 1

Short Form 36 Health Survey (SF-36) scores in patients and controls at baseline and at follow-up.

In an analysis restricted to the 108 patients with BMI at baseline ⩾30 kg/m2, the median physical function score increased from 17 to 20 points over the course of follow-up (p = 0.002), whereas among the 36 controls with baseline BMI ⩾30 kg/m2, the median score decreased from 61 to 25 points (p<0.001). In the subset of 82 patients who were older than 75 years at baseline, the median physical function score was unchanged over follow-up at 17 points, while in the 87 controls aged 75 years or older at baseline, median physical function score deteriorated from 83 to 43 points (p<0.001).

Table 2 shows characteristics at baseline that predicted change in the SF-36 physical function score of controls over follow-up. This analysis was based on the 335 controls who provided complete information on all of the characteristics examined. The effect estimates were derived from a single linear regression model, which incorporated baseline SF-36 physical function score in addition to all of the other risk factors. Better baseline physical function, female sex, older age, BMI ⩾30 kg/m2, current smoking and hypertension were all associated with a significantly greater decline in physical function during follow-up.

Table 2 Predictors of change over follow-up in Short Form 36 Health Survey (SF-36) physical function score of controls

Table 3 summarises the results of a similar analysis for the patient group, using a regression model that also incorporated various features of the subject’s osteoarthritis as assessed at baseline. A total of 259 patients provided information on all of the relevant variables. As in the controls, better baseline physical function, female sex and older age were associated with a significantly worse change in physical function over follow-up. However, the adverse impacts of current smoking, high BMI and hypertension were smaller than in the control group, and not statistically significant. Improvement in physical function score tended to be greater in patients with a higher Kellgren and Lawrence grade at baseline, although differences were not statistically significant. By contrast, changes in physical function were worse in those who at baseline reported pain at greater number of joint sites other than the knee (p value for trend = 0.03).

Table 3 Predictors of change over follow-up in Short Form 36 Health Survey (SF-36) physical function score of patients

When the interval between completion of the baseline interview and follow-up questionnaire was added to the regression models, it did not significantly predict outcome in either patients or controls.


Our findings are consistent with a sustained beneficial impact on physical function following TKA for osteoarthritis, but we found no evidence for a parallel improvement in vitality or mental health relative to the control group. In patients who were obese, the improvement in physical function was less than in patients who were non-obese, but compared favourably with a marked decline in the physical function of obese controls. Similarly, patients older than 75 years did not experience the marked decline in physical function that was observed in controls of the same age. Better baseline physical function and older age predicted worse changes in physical function in patients and controls. Improvement in physical function tended to be greater in patients with more severe radiological disease of the knee, and was less in those who reported pain at other joint sites at baseline.

The study design that we employed had several limitations. It was an observational investigation rather than a randomised controlled trial, and as such was liable to possible confounding effects. In particular, it cannot necessarily be assumed that the changes in patients’ health-related quality of life, had they not undergone surgery, would have been similar to those in the control group. In addition, the long-term follow-up of these two groups was made possible by the initial case–control study that had been conducted. Although the present study was not designed to answer a specific question it was the perfect design to study the long-term outcome of advanced osteoarthritis.

Information about quality of life was collected by interview at baseline, but by self-administered questionnaire at follow-up, a difference that may have distorted comparisons between the two time points. However, Picavet has shown that perceived health as reported in a postal health survey and at home interview was similar.12 Moreover, even if the difference in method of ascertainment biased estimates of absolute changes in health measures, we would not expect it to impact importantly on comparisons between patients and controls, or on the identification of prognostic indicators within the patient group and the assessment of their relative importance.

Another weakness was the incomplete follow-up of patients and controls. This was principally because of death and migration. Death rates may have been higher in subjects with worse physical function, which might falsely elevate the apparent improvement in patients if this differentially happens in patients. However, similar proportions of patients and controls died in our study, making survival bias unlikely. Attrition through death is inevitable in this age group. By contrast, moving to a new address could be associated with either better or worse function (poorer function might mitigate against moving in some people while leading others to enter care homes). As would be expected, mortality was higher at older ages in patients and controls, but differences in rates of follow-up by sex, BMI and Kellgren and Lawrence grade were small.

It was not possible to exclude the presence of osteoarthritis of the knee in the controls, hence possibly reducing the difference between the two groups. However, any osteoarthritis in the controls was not sufficiently severe to have them placed on a waiting list for TKA. The patients were considered to have advanced osteoarthritis, hence the need for TKA.

A further limitation was the frequent absence of information about surgical procedures, which meant that we were not able to take account of the surgeon or type of prosthesis when examining indicators of prognosis. As there was more than one centre involved in the study there was inevitable variation in surgical procedure that limits the generaliseability of the findings. However, this does probably reflect broad patterns of practice in the UK. We feel our data is a welcome addition to our knowledge of the outcome of TKA, which could inform policy decisions as it is a well conducted, controlled observational study.

Information about the patients’ disease state was limited by the use of generic rather than disease-specific measures of function. However, the measures used were applicable to patients and controls and thus enabled comparison between the two groups.

Against these weaknesses must be set the strengths of a long follow-up interval and the size of our study sample, which despite attritional losses gave good statistical power. The follow-up interval from baseline interview ranged from 5.2–7.9 years, and was not itself found to be a predictor of outcome in either patients or controls. This study provides evidence of the course of a group of patients following TKA as compared with expected trends in the general population. Such data provide the basis for understanding the role of joint replacement in a population context of ageing and functional decline. It was also an advantage that we collected parallel data on controls, which enabled us to assess whether prognostic indicators were specific to patients undergoing TKA or applied more generally. The original study had a matched-pair case–control design, but in the analysis presented, the patient–control matching was not taken into account. A repeat analysis based on matched pairs with complete follow-up data (135 pairs) gave similar results, though because of the lower statistical power there were fewer significant predictors of change in physical function.

At baseline, the patient group had markedly worse physical function than the controls, but they did not differ materially in their mental health. This is similar to the pattern that we have observed previously in patients awaiting hip replacement for osteoarthritis,13 and accords with observations in patients undergoing knee and hip arthroplasty in Australia.14 Although it has not been a universal finding (in Finland, Hirvonen and colleagues reported significantly lower mood in patients awaiting total hip or knee arthroplasty than in controls from the general population15), it seems clear that severe osteoarthritis is not associated with poor mental health in the same way as less specific musculoskeletal illnesses such as low back pain.16

The long-term improvement in the physical function that we observed in patients who have undergone TKA is striking when set against the decline that occurred in controls. The observation is consistent with findings from other studies,4 5 and suggests that the benefits from TKA are sustained. By contrast, vitality declined over the study period and, while mental health scores increased, the improvement was no better than in the controls. Some earlier investigations have suggested a benefit to psychological as well as physical health from TKA, although to a lesser degree.5

The improvement in physical function following TKA was apparent in patients who were obese as well as non-obese, whereas obese controls experienced a marked decline in physical function over the follow-up period. This differential from controls may in part reflect judicious selection for surgery, with a tendency to operate preferentially on patients who do not suffer from other concurrent disease. However, it suggests that if appropriate selection criteria are applied, TKA can be worthwhile in patients who are obese. Several previous studies have also found that improvements in quality of life after TKA were similar in patients who were obese and non-obese,1719 although another has suggested that a lower BMI predicts better physical function and greater satisfaction with surgery.20 Our findings give further support to the conclusion of a recent systematic review that BMI is not a strong predictor of functional outcome following TKA,21 and suggest that higher BMI should not be a contraindication to TKA, provided that the patient is sufficiently fit to undergo the short-term rigours of surgery.

Another possible contraindication to TKA is older age. In our study, there was no change in median physical function score among patients older than 75 years. However, this must be viewed in the context of a substantial decline in the physical function of controls of the same age. Again, the way in which patients are selected for surgery may have contributed to this difference. However, other studies have also indicated that patients who are older often derive worthwhile benefit from TKA.5 22 23 TKA halts the decline and maintains physical function up to the oldest age groups.

In the patient and control groups, change in physical function over follow-up was inversely related to baseline physical function. This may simply reflect smaller scope for improvement and greater opportunity for decline in those who started with a higher score, perhaps combined with an element of regression to the mean. A similar relationship has been found in other studies.5

After adjustment for baseline physical function and also for sex and age, improvement in physical function over follow-up tended to be greater in patients with more severe radiological changes of osteoarthritis in the knee preoperatively, and worse in those who reported pain at other joint sites. This could be because overall physical function is determined by the severity of disease in the knee and also by other concomitant illness. In patients with severe radiological changes in the knee, a relatively high proportion of the total deficit in physical function may be attributable to the diseased joint and therefore amenable to alleviation by surgery. Conversely, among patients with pain at other joint sites, the proportional contribution of the knee to the loss of physical function, and therefore the potential impact of surgery, may be smaller.

In summary, our results build on those of earlier investigations in indicating that improvements in physical function following TKA for osteoarthritis are sustained. These benefits extend to patients with BMI >30 kg/m3 and, provided appropriate selection criteria are applied with regard to fitness for surgery, there seems no justification for withholding TKA from patients who are obese.


The authors thank Mrs Gill Strange, who prepared the manuscript.



  • Competing interests: None declared.

  • Funding: This study was funded by a project grant from the Arthritis Research Campaign.

  • Ethics approval: Ethics approval was obtained.