Methods

Participants

A population-based cohort, the Spenshult cohort, was recruited to explore the natural course of knee OA. Inclusion criteria were age 35–54 years, chronic knee pain (>3 months), and no history of previous knee injury or inflammatory joint disease.2 For this study 143 subjects, 63 women and 80 men, median age at baseline (range) 45.0 years (35–54), median body mass index (BMI) at baseline (range) 25.6 kg/m² (18.3–37.5) who participated in the 12-year follow-up were included (fig 1). Ethics approval was obtained (LU 312–90).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

The flow from the sample of population at baseline to study cohort at follow-up.

Radiographic examination

At baseline posteroanterior radiographs of both tibiofemoral (TF) joints were obtained, with straight legs in weight-bearing position and with the weight equally distributed between both legs.2 At follow-up radiographs in a skyline view of patellofemoral (PF) joints and posteroanterior radiographs of both TF joints were obtained in weight-bearing position using a fluoroscopy unit. The patients stood with almost their entire weight on the leg being examined, with the knee flexed 30–50°, and with the patella and the big toe touching the table of the fluoroscopy unit.3 For this study, data on TF joints from baseline and the 12-year follow-up were used. PF joints were examined at 5- and 12-years’ follow-up only.

All radiographs were classified according to Kellgren and Lawrence (K/L)4 by an experienced radiologist. Radiographs from baseline and follow-up were evaluated on different occasions and the reader was blinded to other information about the participant. The inter- and intraobserver agreement for reading the radiographs has been evaluated previously, showing high agreement (κ = 0.72–0.98).2 5 6

Radiographic TF OA was defined as K/L grade ⩾1 in at least one knee.7 The definitions used for incidence and progression were determined a priori. Incidence was defined as a change from K/L grade 0 to grade ⩾1 or more from baseline to follow-up. Progress was defined as a change from K/L 1 to any higher grade.8 9 To be able to compare yearly incidence rate with existing data the accumulated incidence was divided by 12 years.8 The 5-year incidence rate has been reported previously.10 In addition, a secondary cut-off point was used, defining TF OA as K/L grade 2. Incidence using this threshold was defined as a change from K/L grade 0 or 1 to K/L grade ⩾2, and progress as a change from K/L grade 2 to any higher grade. Both knees were examined, but only one knee per participant was used in the analysis. To avoid a false-negative result the knee with the most radiographic change from baseline to follow-up was used. If both knees showed no change or equal change, a right or left knee was chosen alternately. The joint space width (JSW) of the PF joint was measured with a ruler and OA was defined as JSW <5 mm in at least one knee.11

Clinical classification

The participants were examined by a rheumatologist at baseline, including patient history, palpation and range of motion. Clinical OA was classified using the proposed clinical criteria described by Altman: knee pain; crepitus; morning stiffness ⩽30 min; bony enlargement; age ⩾38 years12 (fig 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Clinical classification scheme of knee osteoarthritis (OA) according to Altman et al.12

Data analysis

Incidence in this paper refers to development of new radiographic OA from baseline to the 12-year follow-up. The denominator equals 12 years in all cases. An estimate of overall annual incidence rate was determined using the overall proportion of incident OA divided by 12 years.8 The relative risk for knee OA development was determined for the group with clinical OA at baseline. The Mann–Whitney U test was used for between-group comparisons. A p level <0.05 was regarded as statistically significant. A 95% binominal confidence interval for radiographic changes in the TF compartment was calculated.

Results

Baseline prevalence of OA

Seventy-six (53%) of the 143 subjects had no radiographic TF OA (K/L 0) at baseline, 30 (21%) had TF OA in one knee (K/L ⩾1) and 37 (26%) had TF OA in both knees. According to the clinical classification of OA 100 (70%) people presented with knee OA. Fifty-one (36%) subjects had both clinical and radiographic OA (tables 1–3).

The baseline characteristics (age, sex, BMI or radiographic OA) of the 61 who were lost to follow-up were not statistically different from the population studied (p>0.2, data not shown). Twenty-three (38%) of the 61 missing subjects had TF OA (K/L ⩾1) in at least one knee.

Change in radiographic TF OA over 12 years

Overall, 65/76 (86%, (95% CI 75.6% to 92.6%)) developed incident TF OA over 12 years (K/L ⩾1). The incidence of radiographic TF OA was studied in two different subsets: (a) subjects with neither clinical nor radiographic OA at baseline (n = 27) and (b) subjects with clinical OA but no radiographic changes at baseline (n = 49). The proportions of subjects developing radiographic OA in the two groups were 78% (95% CI 57.7% to 91.4%) and 90% (95% CI 77.8% to 96.6%) respectively (table 1). The relative risk of OA in subjects presenting with clinical OA was 1.15 (95% CI 1.08 to 1.15). A comparison of the two subgroups showed that subjects with clinical OA were somewhat older (median 46 years (range 35–54) vs 40 years (range 35–54), p = 0.001). No significant differences in BMI (p = 0.48) or sex (p = 0.36) were seen between the two groups.

Twelve-year follow-up showed radiographic progression in 50/51 cases (98%, (95% CI 89.6% to 100%)) (table 2).

Incidence of radiographic PF OA over 7 years

The PF joint was first examined at the 5-year follow-up in 125/143 subjects, showing radiographic PF OA in 41/125 (33%) subjects. The group with PF OA consisted of more women (p<0.001) and had significantly higher BMI (median (range) 26.6 (20.4–37.5) kg/m² vs 24.7 (18.3–34.0) kg/m², p = 0.04). Twenty-three of the 41 had TF OA at baseline (data not shown). Seven years later 26/84 (31% (95% CI 21.3% to 42.0%)) subjects with JSW >5 mm at first examination had developed PF OA (table 3). At follow-up 58/143 (41%) had OA in both TF and PF compartments, and six (4%) had no radiographic signs of knee OA (data not shown).

Discussion

Our hypothesis that knee pain is an early feature of OA and that all patients with idiopathic chronic knee pain therefore would deteriorate radiographically over 12 years was not fully supported. However, in this population 86% of the patients with normal radiographs at baseline developed incident OA in the TF compartment according to the K/L 1 threshold for radiographic OA. The rationale for choosing K/L 1 (minute osteophytes, doubtful significance) as cut-off point was based on previous studies suggesting that K/L 1 is related to progression of radiographic findings, and therefore should be considered as an early feature of OA.7 Using hypothetical data from the subjects lost to follow-up and assuming that none of the 38 subjects without radiographic changes at baseline developed incident OA over 12 years, would reduce the overall proportion of incident TF OA to 57%, and the estimated yearly incidence rate to 5%. Using the K/L 2 threshold in the population studied resulted in an estimated annual incidence of 4%, which is higher than previously reported rates ranging from 2% to 3.1%.8 13 14 Female sex,14 older age13 and previous knee injury15 are related to higher incidence rates. The higher incidence rate in this population of younger men and women without known knee injuries supports our hypothesis that knee pain may be an early feature of OA.

The possibility that knee pain in subjects without TF OA at baseline was related to already existing PF OA cannot be excluded. OA in the PF compartment is related to painful knees, and it has been suggested that PF OA occurs more frequently than TF OA.16 A recent study of populations with early knee OA revealed that PF OA was present in 41% of cases with K/L grade 0–1 in the TF compartment, and that the presence of osteophytes in the PF compartment was a risk factor for developing new osteophytes in the ipsilateral TF compartment after 30 months.17 Forty-one subjects had PF OA at the 5-year follow-up, of whom 23 had no TF OA at baseline. Assuming that they had PF OA at baseline would indicate that PF OA could possibly explain the knee pain for 23/76 (30%) subjects. Eleven of the 18 subjects with missing data on PF OA had no TF OA at baseline. Assuming that they all had PF OA at baseline would give a maximum prevalence of baseline PF OA without coexisting TF OA of 34/76 (45%), suggesting that PF OA at the most could explain the knee pain in 45% of the cases with no TF OA at baseline.

Strengths of this study are the population-based longitudinal design, and the high response rate from baseline to follow-up (78%).

A possible concern is the difference in radiographic technique at baseline and follow-up. At baseline radiographs were taken with straight knees, and at follow-up a semiflexed position was considered a more reliable radiographic technique, which has later been corroborated.6 18 Methodological differences may have affected the occurrence and disappearance of radiographic findings, such as osteophytes. Such misclassification at baseline would underestimate the number of patients with radiographic knee OA. The development of radiographic findings from baseline to follow-up could thus be a progression rather than incident OA.

Conclusion

A majority of middle-aged patients with chronic idiopathic knee pain in this study developed knee OA over 12 years. We conclude that knee pain is often the first sign of knee OA.