Article Text

Extended report
Weight loss as treatment for knee osteoarthritis symptoms in obese patients: 1-year results from a randomised controlled trial
  1. Henning Bliddal1,
  2. Anthony R Leeds2,
  3. Lise Stigsgaard1,
  4. Arne Astrup2,
  5. Robin Christensen1
  1. 1The Parker Institute, Copenhagen University Hospital, Frederiksberg, Copenhagen, Denmark.
  2. 2Department of Human Nutrition, Faculty of Life Sciences, University of Copenhagen, Copenhagen, Denmark
  1. Correspondence to Professor J Bliddal, The Parker Institute, Copenhagen University Hospital, Frederiksberg, DK-2000F, Denmark; henning.bliddal{at}


Objective To evaluate 1-year symptomatic improvement in obese patients with knee osteoarthritis (OA) on an intensive low-energy diet (LED) maintained by frequent consultations with a dietician compared to minimal attention.

Methods The LED programme consisted of group therapy with dietary consultations and two periods of a low-calorie diet of 810 kcal/day during weeks 0–8 and weeks 32–36. The control group only received dietary instruction and attention for 2 h at baseline, and at weeks 8, 32, 36 and 52. The primary end point (total Western Ontario and McMaster Universities (WOMAC) index) was assessed as the mean group difference during and after 1 year.

Results The study population consisted of 89 patients, 89% women, average age 63 years. After 1 year, mean weight loss in the LED group was −10.9 kg (11%) versus −3.6 kg (4%) in the control group (p<0.0001). There was no difference between the groups in total WOMAC index (p=0.11), although both groups improved. However, the LED intervention resulted in less WOMAC pain (7.7 mm), with a group mean difference of 7.2 mm (95% CI 1.0 to 13.4, p=0.022). After one year 14 (32.8%) responded to LED versus 7 (15.6%) to control, with an absolute benefit of 16.3% (−1.1& to 33.6%, p=0.066).

Conclusion Continuous reinforcement of a weight loss programme can be successful over a year in obese knee OA patients. Weight loss was statistically reflected only by a reduction in pain. However, the overall clinical benefits of the intervention on health should lead to a strong recommendation of weight loss in this group of patients.

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.


Epidemiological studies have shown obesity to be an important risk factor for the development of knee osteoarthritis (OA),1 and there is a definite association between weight bearing in both knee and hip joints and the development of OA.2 In fact, obesity is probably the single most important risk factor for the development of severe OA of the knee, more so than other potential predisposing factors, including heredity.1 3 Several factors may account for this association, while the weight per se may be modified by therapy. Obesity will increase the load on the knee, and when further aggravated by varus mal-alignment, which is seen in about 50% of patients, the resulting effect on the joint could well be responsible for the degeneration of the cartilage as measured by grade of severity in a cross-sectional study of OA patients.4 5 With the increasing prevalence of obesity, it is essential to develop and assess suitable treatment strategies which will result in long-term weight reduction and maintenance of weight loss.6 7 Weight loss is difficult to achieve, and maintaining weight loss is an even greater challenge.8 The greater the initial weight loss, the better the subsequent outcome in terms of sustaining that weight loss, perhaps as a consequence of the initial weight loss reflecting better compliance with treatment.9 Typically, a weight loss diet would have a deficit of 500–600 kcal/day below the current requirement for energy balance, leading to a weight reduction of 0.5 kg a week. Very low-energy diets (LEDs) may produce better initial weight loss, which might improve motivation, but long-term weight loss (ie, maintenance) achieved in this way is rarely any greater than the loss achieved with low fat diets.10

Significant weight loss is frequently recommended for the treatment of knee OA,11 an option recently supported by a meta-regression analysis showing that disability in obese patients with knee OA is significantly improved when weight is reduced by more than 5% at a rate of >0.25% per week.12 As a consequence, it is recommended that patients with knee (and hip) OA who are overweight should be encouraged to lose weight and maintain their weight at a lower level.13 14

The objective of the present study was to evaluate sustained symptomatic improvement in a group following an intensive LED programme providing 810 kcal/day15 and supported by consultations with the same dietician throughout a year, compared to matched individuals randomly allocated to a control group.



The study was performed in accordance with the principles of good clinical practice. Ethics committee approval and informed patient consent were obtained. The participants were informed that the control group would receive less attention but would be provided with some instructions and material to enable them to lose weight if compliant with a low-fat diet. The study was approved by the local ethics committee of The Capital Region of Denmark (KF 01-104/02) and the randomised controlled trial (RCT) was carried out according to the Helsinki criteria.

Study design

This was a prospective RCT with blinded outcome assessment.15 The 1-year trial evaluated sustained outcomes at five prespecified time-points by applying a repeated measures design. The end point was assessed as the group mean difference during and after the 52-week therapy, which consisted of a 44-week sustainment phase (open label) following a initial 8 weeks, the results of which have been published previously.15

Patient selection

Patients were recruited from the outpatient clinic of the Department of Rheumatology, Frederiksberg Hospital, Denmark. Overweight patients over 18 years of age, with primary knee OA diagnosed according to the American College of Rheumatology criteria,16 were considered eligible for the trial. Major exclusion criteria were a history of other rheumatic diseases possibly responsible for secondary OA, diabetes mellitus or other endocrine disorders, and substantial abnormalities in haematological, hepatic, renal or cardiac function. Overweight was defined as a body mass index (BMI) ≥28 kg/m2. Only patients who explicitly expressed a clear, unequivocal desire for weight loss and were fluent in Danish, were invited to participate. Before inclusion in the study, fasting blood glucose was measured as well as haemoglobin and Thyroid Stimulating Hormone. The participants were asked not to change any medications during the 52 weeks of the study.


All the subjects were randomly assigned to either 52 weeks of intensive weight loss therapy or a 52-week moderate conventional hypo-energetic, high protein diet (approximately 1200 kcal, 5 MJ/day), defined as the control group. For every 16 patients included, stratified randomisation was carried out according to (1) gender, (2) BMI and (3) age to ensure homogeneity between intervention groups. The random allocation sequence was concealed until interventions were assigned: each randomisation list was drawn up by the study statistician (RC) and given to the secretariat at the Parker Institute, who subsequently informed patients when to meet the dietician (ie, only implicitly referring to group allocation).


LED group

The intensive diet was a formula diet providing 810 kcal/day as previously described in detail.15 In brief, the formula LED consisted of nutrition powder (Speasy; Dansk Droge, Ishoj, Denmark) that was dissolved in water and taken as six daily meals.15 This met all recommendations for the daily intake of nutrients. This formula was administered for the first 8 weeks of the programme and was followed by instruction, in groups of eight participants, every week for another 24 weeks. The guidance given aimed to achieve a 1200 kcal/day diet. For another 4 weeks (weeks 32–36), maintenance of the patients' weight loss was reinforced by use of the same formula diet as in the first 8 weeks (810 kcal/day) and from weeks 36 to 52 yet again by instruction in groups every second week. This meant a total of 44 visits supervised by the same dietician. The weekly sessions were 1.5 h long, to give a total of 66 h of instruction.

Control group

The conventional diet programme consisted of a presentation by the same dietician as for the LED group, who gave nutritional advice in a 2 h session at baseline, and in weeks 8, 32, 36 and 52. The dietician recommended eating ordinary foods in amounts which would provide the patients with approximately 1200 kcal/day. The control group attended five meetings altogether with a total of 10 h of instruction. The effect of the intervention was enhanced by the instruction in weeks 32 and 36, which may have been crucial for compliance with the programme. Also, the result was supported by group dynamics and continuous encouragement from the dietician.

Assessment of efficacy

Changes in body weight and body composition were examined as independent predictors of changes in knee OA symptoms. At baseline and after 8, 32, 36 and 52 weeks, the body weights of all patients were measured using a decimal scale (TANITA BWB-600S; Frederiksberg Vægtfabrik, Copenhagen, Denmark). Symptoms of OA, as perceived by patients prior to the assessment, were monitored using the Western Ontario and McMaster Universities (WOMAC) OA index, a validated, disease-specific questionnaire addressing the severity of joint pain (five questions), stiffness (two questions) and limitation of physical function (17 questions). The total WOMAC index was deemed the primary outcome. The patient assessed each question on a 0–100 mm visual analogue scale (VAS) and the results were presented as normalised WOMAC scores. Regarding secondary outcomes, the Health Assessment Questionnaire (HAQ) was used to investigate patients' disability,17 and patients experiencing a reduction of at least 50% in disease symptoms according to the total WOMAC index were considered to have clinically significant improvement. The OMERACT-OARSI response criteria did not apply, as we did not use an absolute lower limit for pain (change) in our inclusion criteria.

Power and sample size

We wanted to detect a significant effect size (ES) (ie, Cohen's standardised mean difference) of 0.8, indicating a large clinical effect compared to placebo.15 We set α=5% and with a power (1−β) of 90% anticipating a large clinical effect (ES of 0.8), thus calculated we needed a sample size of 34 knee OA patients in each of the two groups.18 For practical reasons, 16 participants were included at a time and we increased the sample size to 48 patients per group to allow for a dropout rate of more than a quarter.

Efficacy analyses

The primary efficacy analysis performed was assessment of the difference between groups in the change in the total WOMAC index after 52 weeks in the studied population (ie, all randomised patients seeing the dietician at baseline). The baseline observation carried forward (BOCF) approach was used for patients who did not complete the study (as illustrated in the trial profile), as this method seems the most conservative in weight loss trials,19 referred to as non-responder analysis. However, as this trial was designed as a pragmatic trial, baseline data were only imputed when the patient did not attend follow-up visits, and was not based on adherence rate or compliance considerations. For sensitivity, we report the results for the major outcomes as a complete analysis in an online supplementary appendix table presented for each individual time point. To analyse the longitudinal element of the randomised trials, a linear approach was used for repeated measurements, fitted in SAS using the procedure ‘PROC MIXED’ based on restricted maximum likelihood (REML) estimates of the parameters.20 The factor (subject) was applied as a random effects factor. Assessment of the treatment and time effects was of interest in testing for a possible interaction, and both treatment and time were included as systematic factors, using the baseline value as covariate to reduce random variation and increase power.21 Unless stated otherwise, results are expressed as the difference between the group means and 95% CI with the associated p values, based on the mixed linear model. To determine clinical efficacy, based on the binary relief of symptoms (yes/no), we calculated the number needed to treat (NNT)22 in terms of a clinically significant improvement15 based on symptoms at 52 weeks. As a secondary hypothesis, we assessed the average ES during the entire study, using Cohen's index, evaluating the overall patient experience in the dietary regimen compared to the control group, which was analysed based on the same mixed model, however with only the main effects of group and time without the interaction in the model.


Participant flow

Participants underwent a 1-year follow-up examination as previously reported.15 The enrolment, randomisation and treatment adherence details of the participants are shown in figure 1. There were no significant differences in the number of patients who withdrew for various reasons.

Patient characteristics

All our patients had radiographic severity grade 2 or 3 OA on the Kellgren and Lawrence scale.15 The demographic and clinical characteristics of the randomised patients were comparable at baseline (table 1). As expected, the majority of patients were female (n=79, 89%). The mean age was 63 (SD 11) years ranging from 36 to 90 years. On average, patients were obese, corresponding to category II obesity, with a BMI ranging from 29 to 55 kg/m2. Inflammatory joint disease was an exclusion criterion per se, but signs of modest inflammation, that is, mild joint effusion, were present in 10–15% of patients; the median level of C reactive protein was 3.5 mg/l. The mean (normalised) total WOMAC index at baseline was 39 mm (=39% of the maximum), indicating disease with low-to-moderate symptom severity.

Table 1

Baseline characteristics for all randomly assigned patients

Clinical efficacy

As shown in figure 2, treatment was successful for the average patient in terms of weight loss achieved (figure 2A) and improvements in individual symptoms overall according to the total WOMAC index (figure 2B). For the WOMAC index overall, there was no group×time interaction (p=0.16) and no main effect of time (p=0.69), however, we found a significant group difference (p=0.002). Figure 2C shows the percentages of responders as regards weight loss (squares) and the total WOMAC index (triangles) favouring LED across all time points, indicating that the clinical response can be sustained after assessment at 8 weeks.

Figure 2

Longitudinal clinical efficacy for the primary outcomes. Values are mean±SE. (C) The proportions responding are shown for both total WOMAC index (>50% reduction) and clinically significant weight loss (>10%). CD, control diet; LED, low-energy diet; WL, weight loss; WOMAC, Western Ontario and McMaster Universities index.

Focusing on the endpoint of clinical efficacy at week 52, the LED group experienced a mean weight change of −10.9 kg (11%), whereas body weight in the control group reduced only by −3.6 kg (4%), resulting in a statistically significant difference between the groups of 7.3 kg (95% CI 5.0 to 10.0 kg, p<0.0001). As anticipated, the magnitude of weight loss was greater the more obese the patients were at baseline. This is illustrated in the online supplementary appendix table 1. Likewise, the number of patients who were able to achieve and sustain a weight loss of at least 10% after 1 year in the LED and control groups was 24 (54.5%) and 4 (8.9%), respectively (p<0.0001). This corresponded to an absolute benefit of 45.7%. According to the normalised changes in the total WOMAC index after 1 year, there was a statistically significant improvement in the LED group of −7.3 mm compared to −3.0 mm in the control group (table 2), giving a mean difference of 4.3 mm (−0.9 to 9.6 mm, p=0.11). The response rates of 14 (31.8%) in the LED group and 7 (15.6%) in the control group correspond to an absolute benefit of 16.3% (−1.1% to 33.6%). Although statistically not significant, this indicates an NNT of around six patients needing LED therapy compared to control. In contrast, when focusing on the pain subscale of the WOMAC index, the LED group improved significantly (7.7 mm), whereas the control group did not experience any reduction in pain (0.5 mm), resulting in a statistically significant group mean difference of 7.2 mm (1.0 to 13.4 mm, p=0.022). The proportion of losses-to-follow-up was generally high in both the LED and control groups (25% and 49%, respectively), with a statistically significantly higher attrition rate in the control group (χ2=5.44, p=0.020). It was twice as likely (49% vs 25%) that the baseline BOCF imputation technique applied in the control group as in the LED group; for the purpose of sensitivity analysis we also report test results and estimates based on the data observed using no imputation (online supplementary appendix table 2). Sensitivity analyses indicated a consistent finding for the total WOMAC index showing no statistical difference between the groups (p=0.16), whereas the pain assessment at the end point was not statistically significant due to the small sample size (33 and 23 patients compared in the two-sample t test).

Table 2

Change in outcomes from baseline after 52 weeks in knee osteoarthritis patients who were randomised to either an intensive low-energy diet (LED) or a control group

As the repeated measures analyses supported a main effect of dietary group (p=0.002), we explored – on a post hoc level – the average clinical improvement that one would expect during the entire 1-year programme (online supplementary appendix figure 1). With an average weight loss of 7.6 kg (8%, 5.3–9.9 kg, p<0.0001) compared to the control group, our data indicate that the average patient would experience a moderate-to-large clinical effect in terms of pain reduction (ES 0.57, 95% CI 0.15 to 0.99), disability (ES 0.55, 0.13 to 0.97), the total WOMAC index (ES 0.58, 0.15 to 1.00) and the HAQ (ES 0.61, 0.18 to 1.03). There was no indication of clinical improvement in joint stiffness as also assessed with the WOMAC index (ES 0.28, −0.14 to 0.70).


The most frequent adverse events reported after following the LED were constipation in five patients (11%), increased flatulence in four (9%), dizziness in two (5%) and heightened sensitivity to cold in two (5%). According to the design, we did not systematically collect data on adverse events for the control group as they were accorded a minimal amount of attention. None of these episodes resulted in patient withdrawal. However, during the first 8 weeks three patients reported that they were not able to comply with the prescribed LED intervention, which may be regarded as an adverse effect per se.


At 1-year follow-up, the knee OA patients with concomitant obesity allocated to a low-fat, high carbohydrate and protein diet showed superior results compared to a control group receiving a minimum amount of dietary attention. This resulted in a clinically important reduction in pain due to knee OA. The popular clinical opinion that weight loss achieved at a slow rate is better was not confirmed in this randomised, controlled setting with participating patients with severe functional limitations due to OA. These results are in accordance with previous reports in other obese subjects23 and demonstrate that bad knees are no excuse for not losing weight. In agreement with previous indications,12 15 24,,27 there was a significant clinical improvement according to the total WOMAC index, which in the intervention group showed a decrease approximately twice that of controls. The results indicate that some 30% of patients with knee OA and concomitant obesity can achieve significant long term weight loss and that the intervention used in this study had twice the efficacy of the control group programme. The NNT for the intervention was six patients, and the ESs for most self-reported outcomes of pain and function were just above 0.5, a moderate effect which is still better than achieved with most medications for OA28 as well as light exercises.29 The only exception to this was joint stiffness as assessed by questionnaire, which was not affected by the treatment.

It may be deducted from the follow-up results of this trial (illustrated in figure 2) that a better outcome for knee OA depends on maintaining weight loss. Although the number of responders decreased during follow-up, the response rate was still twice as high among those on continuous dietary support as compared to control participants. Thus, the results suggest continuous reinforcement of the weight loss programme is mandatory for treatment of these patients. Our programme8 encouraged adherence of the LED treated patients through a weekly meeting with a dietician. The overall response to weight loss in the patients is excellent in the short term, that is, within 3 months of implementation, this therapy with its NNT of 4 and ES of 0.8,15 resulted in unparalleled improvements in patients with combined obesity/knee OA. Even though the 1-year follow-up does not indicate a similar response rate but only an NNT of 6 and ES of about 0.5–0.6, the face validity of the intervention is still high. The quality of the evidence is moderate,30 while at a personal and political level it may be strengthened by other factors such as, for example, an associated response in life-style related diseases. This notion is substantiated by the fact than more than half of the patients in the intervention group maintained a weight loss of 10% or more after 1 year. The reduction in pain was the only measure to show statistical significance. The lack of difference in function and disability is disappointing and most likely due to an underpowered sample size. This can be attributed to the high drop-out rate and an overly optimistic power calculation.

In conclusion, in knee OA patients with concomitant obesity, we found a sustained, although moderate, response to a weight loss intervention during and after 1 year, as compared to the initial short-term response. We would encourage clinicians to apply these data in clinical practice as most patients should receive instruction on how to lose weight. In our opinion these data suggest that policy makers should adopt, support and/or reimburse this form of treatment in most situations.


The authors are indebted to laboratory assistants Inger Wätjen, Jette Nielsen, Salomea Hirschorn and Tove Riis Johannessen, and the database consultant Christian Cato Holm.


Supplementary materials


  • Funding This study was supported by grants from The Oak Foundation, Frederiksberg Hospital, Copenhagen Hospital, The Danish Rheumatism Association, Hørslev Fonden and Bjarne Jensen Fond. The dietary products and the dietician were supported by Dansk Droge A/S; none of the authors have received grants or personal fees from this company.

  • Competing interests AR Leeds is employed as medical director by the Cambridge Manufacturing Company (Cambridge Diet). Henning Bliddal, Arne Astrup and Robin Christensen have received travel grants from the Cambridge Manufacturing Company to attend scientific meetings.

  • Ethics approval This study was conducted with the approval of the local ethics committee of The Capital Region of Denmark.

  • Provenance and peer review Not commissioned; externally peer reviewed.