A flow chart of the recruitment process is given in Figure 1. Radiologists and orthopedic surgeons at the Halmstad County Hospital, in the south-west of Sweden, and general practitioners within the catchments area of this hospital, were informed about the study and asked to list patients with radiographic knee osteoarthritis on a "patients eligible for research" list. Between October 1998 and October 2001 121 patients, referred by their general practitioner for radiographic examination because of long standing knee pain, were listed. Ninety-seven fulfilled the inclusion criteria: age 35–65, living in the defined geographic area, and diagnosis of radiographic osteoarthritis of Kellgren and Lawrence grade III or more, i.e. definite osteophytes and joint space narrowing. All listed patients received written information about the study. One week after the information was sent, patients were contacted by telephone, and invited to participate in the study. Twenty-eight patients declined participation for various reasons, the most common reason being lack of time and interest. To ensure only patients with symptoms due to knee osteoarthritis and eligible for exercise intervention were included, the following exclusion criteria were used: inflammatory joint disease, anterior cruciate ligament injury, known symptomatic injury to the menisci, hip symptoms more aggravating than the knee symptoms, about to have knee replacement surgery within 6 months, and co-morbidities not allowing exercise (Figure 1).

When eight or more patients fulfilled the inclusion criteria they were invited to baseline interview and examination for determination of exclusion criteria. Randomization was performed after the baseline examination. All patients were informed that they could be randomly allocated to either the exercise or the control group. After written informed consent, sixty-five subjects were randomized. Randomization was performed by the patient drawing a sealed envelope containing a piece of folded paper with either the word "exercise" or "control" written on it. Four persons were falsely randomized (one was too old at inclusion, one had severe hip osteoarthritis, one had fibromyalgia, and one had only joint space narrowing and no significant osteophytes), and thus 61 subjects entered the study. Thirty persons were allocated to the exercise group and 31 to the control group. Patients in the control group were offered exercise classes after the six-month follow-up period.

Post-hoc, a power analysis was performed to estimate the number of patients needed to show a clinically significant difference between groups. Estimating the least clinical significant difference to be 11 ± 15 KOOS points, a total of 30 subjects in each group were needed to detect a difference with 80% power, p = 0.05.

Data were analyzed using nonparametric tests. P-values of less than or equal to 0.05 were considered to be significant, and all tests were two-tailed. To compare groups, Mann-Whitney U-test was used. Friedman's test was used for repeated measures analysis of variance. Six weeks was considered as the time-point of primary interest, and 6 months as follow-up. Wilcoxon signed rank test was performed to compare changes from baseline to six weeks and 6 months respectively. Analyses were performed using SPSS 12.0.1 for Windows 30.

The study was approved by the Research Ethics Committee at Lund University, Sweden (LU 99–98), and is in compliance with the Helsinki Declaration.

The mean age of the 61 included subjects was 56 ± 6 years, and the mean BMI was 29.5 ± 4.8 kg/m². Patient characteristics are shown in table 1. Twenty-eight patients in each group were available for follow-up. The reasons for dropout were lack of time, reorganization at work, sudden illness, and increased knee symptoms (Figure 1). There were no clinically significant differences in baseline characteristics between the groups. Patient characteristics are shown in table 1.

The total number of performed supervised exercise sessions by the 28 patients available for follow-up in the intervention group was 302/336 (89.9%). Patients participated on average in 11 out of 12 possible exercise classes (12 classes (n = 11), 11 (n = 9), 10 (n = 6), 9 (n = 1), 2 (n = 1)). The most common reason for absence was illness not related to knee osteoarthritis, and work-related lack of time.

There was no difference between groups in pain or self-estimated function at either 6 week or 6 month follow-up. Quality of life improved significantly in the exercise group compared to the control group at 6 weeks (4.0 vs. -0.7, p = 0.05) and the results persisted at 6 months (5.1 vs. -2.3, p = 0.02, Table 2)

The individual differences ranged from clinically significant improvement of at least 10 points to clinically significant deterioration in all KOOS subscales and in both the exercise and control group (Figure 3).

Six weeks of intensive exercise had no effect on self-reported pain or function in middle-aged patients with symptomatic and moderate-severe radiographic knee osteoarthritis.

Quite opposite to previously published studies on exercise in knee osteoarthritis we found no improvement in pain or function. Possible reasons for this include our study group having moderate to severe osteoarthritis compared with mild to moderate in most previous studies, being younger than previously studied groups and the intervention being of comparably high intensity.

It has been suggested that the responsiveness to exercise is modified by the loss of joint space width 31. The homogeneity of this study population, with regard to radiographic changes, provided us the possibility to study the effects of exercise on patients with moderate to severe radiographic knee osteoarthritis. Can significant improvements of pain or self-reported function be expected in patients with radiographic knee osteoarthritis corresponding to Kellgren & Lawrence grade 3 or more? In this study, no improvements were seen on group level in pain or function. However, regular exercise in general is important to prevent diseases caused by inactivity 6, and thus patients with knee osteoarthritis should be encouraged to exercise. In clinical practice, patients with severe knee osteoarthritis should have treatments based on individual preferences and different stages of motivation 32.

It can be argued that the exercise intensity was too high for this group with moderate to severe knee osteoarthritis. Even though the intensity of each exercise was individually adapted, all individuals exercised at a minimum of 60% of HR max. It has been suggested that pain during exercise might be a protective mechanism in knee osteoarthritis, i.e. an increase in pain from too intensive exercises may restrain patients from further joint loading, which otherwise could cause further cartilage damage 33. Patients in the current study were told to reduce the exercise intensity if pain during exercise was perceived as worse than 'acceptable', or persisted more than 24 hours.

It is suggested that the different degrees of varus-valgus laxity should be taken into account in exercise interventions, to enhance the functional outcome 3435. Severe knee osteoarthritis is associated with a hip-knee-ankle malalignment and an increase in varus-valgus laxity compared to healthy knees 36. It is possible that varus-valgus laxity mediated the effect of exercise on pain since all patients had radiographic changes corresponding to Kellgren and Lawrence grade III or more. Malalignment may cause increased joint loads, and greater quadriceps strength might further increase joint load by the muscles compressing the articular surfaces 37.

Younger patients are usually more physically active than elderly 38, and have higher demands on level of physical function and physical performance at work or leisure time. Thus, moderate to severe knee osteoarthritis might be perceived as more disabling by younger individuals compared to elderly. Our study population was younger (<65 years) and comprised more men (49%) than most other populations with knee osteoarthritis described in randomized controlled trials of exercise 1314153940, which might have reduced the effect on self-reported function in the present study.

This study showed no significant differences on self-reported pain and function either between or within groups. A post-hoc analysis was performed to study the possibility that the benefit from exercise was larger in subjects with worse pain at baseline. Fifteen patients in the exercise group were compared to 13 from the control group who had worse than total group median pain score (KOOS Pain 58 on a 0–100, worst to best scale) at baseline. The groups had comparable patient characteristics. The changes seen in these subgroups were however not different from the changes seen in the total groups.

A possible limitation could be lack of power. A post-hoc analysis was performed to estimate the number of patients needed to show a clinically significant difference of 11 ± 15 KOOS-points 20. The standard deviation of 15 is supported by results from randomized controlled trials of glucosamine supplementation 41 and a nutritional supplement 42 for knee osteoarthritis, where significant group differences were found in KOOS pain and ADL subscales. The number of subjects in each treatment arm in these RCT:s ranged from 15 to 27. The standard deviations in KOOS subscales have not previously been determined in exercise interventions.

Only one of the five KOOS subscales showed a statistical significant improvement, and it can not be excluded that this result could be due to chance. The improvement of the KOOS subscale Quality Of Life in the exercise group was of doubtful clinical significance, however the improvement persisted over time, and is in accordance with previous findings of impact from exercise on mental health aspects in patients with knee osteoarthritis 314344. Group dynamics, support, or attention received may possibly have influenced the quality of life more than the exercise itself in the present study. Psychosocial factors are important determinants of physical function 45, and our results suggest that supervised exercises and follow-up are important, and that quality of life should be evaluated in osteoarthritis interventions.