Radiologic assessment of age-related knee joint space changes in women: A 4-year longitudinal study

Authors


Abstract

Objective

To analyze a population-based cohort of women in order to establish normal values of joint space width (JSW) and to evaluate the existence of age-related joint space loss (JSL).

Methods

Knee radiographs were performed 4 years apart in women from the OFELY (Os des Femmes de Lyon) Cohort. Posteroanterior radiographs of both knees were taken in semiflexion with a standardized fluoroscopically assisted protocol. Radiographs were qualitatively evaluated using a scoring system based on the Altman score that assessed joint space narrowing, osteophytes, and sclerosis for each tibiofemoral compartment and each side. For quantitative assessment, radiographs were digitized using a video camera, and specific software was used to measure JSW in every compartment.

Results

We evaluated the radiographs of 606 women (ages 39–90 years, mean 62 years) and found that in all subjects, JSW significantly decreased with age in every compartment (r = −0.12 to −0.16, P < 0.001), including in 358 subjects without any radiographic abnormality related to osteoarthritis (OA) at baseline. The longitudinal analysis confirmed a significant loss over 4 years of ∼0.30 mm (6%) for the medial compartment. Multiple regression analysis did not identify significant predictors of JSL among clinical risk factors and biochemical markers of bone and cartilage turnover.

Conclusion

In this first longitudinal study of a population-based cohort of women, we have established normal values of JSW and shown that JSW decreases with aging, especially at the medial compartment, even in subjects without any radiographic abnormalities related to OA.

INTRODUCTION

Due to an aging population and the secular trend of an increase in body weight, knee osteoarthritis (OA) is an increasing public health problem. The natural history of OA remains poorly studied for several reasons. First, many previous epidemiologic studies were conducted in elderly or osteoarthritic populations (1, 2). Second, in most studies (3–5), knee radiographs were obtained with legs in full extension, a technique adequate to analyze bony changes such as osteophytes and sclerosis, whereas joint space width (JSW) is better assessed on semiflexed radiographs with standardized procedures (6). Radiographs are usually analyzed using the Kellgren/Lawrence (K/L) scale score, which provides more emphasis on osteophytes than on JSW (7, 8).

Measurement of JSW is recommended as the best feature of the progression of tibiofemoral OA (9, 10); however, few data about normal values for JSW are currently available (11). In addition, the existence of a physiologic age-related joint space loss (JSL) independent of OA is debated. Dacre et al (12) showed that JSW decreases with aging, although Lanyon et al (13) found no significant decline in JSW with increasing age.

The aim of this study was to prospectively analyze JSW in a population-based cohort of women ages 39–90 years in order to establish normal values of JSW, as well as to establish the existence of age-related JSL using a standardized semiflexed radiographic protocol and a semiautomated measurement of tibiofemoral JSW using a dedicated software.

SUBJECTS AND METHODS

Subjects.

Subjects consisted of women from the OFELY (Os des Femmes de Lyon) Cohort. OFELY is a previously described (14, 15) prospective study of the determinants of bone loss in 1,039 female volunteers ages 31–89 years recruited between February 1992 and December 1993 with an annual followup. Written informed consent was obtained from each subject, and the study was approved by the local ethics committee.

In the present study, we analyzed 616 women ages 39–90 years (mean ± SD age 61.6 ± 10.5 years) and assessed knee radiographs performed 8 years after their inclusion into the OFELY study and subsequently median ± interquartile range 4.0 ± (3.8–4.1) years later. We recorded baseline clinical, biologic, and bone densitometric data. Major exclusion criteria were unilateral or bilateral knee replacements (n = 4) and insufficient quality of radiographs for valid analysis (n = 6). A total of 606 subjects were included and both radiographs of these subjects were qualitatively and quantitatively analyzed.

Radiographs..

Radiographic protocol.

Posteroanterior radiographs of both knees were taken in semiflexion with a standardized fluoroscopically assisted protocol using the SynaFlex X-ray Positioning Frame (Synarc, San Francisco, CA) as previously described (16). Rotation of the feet was fixed at 10° by a V-shaped support. Both knees were in contact with the cassette, removing magnification effects. The great toe touched the vertical wall of the frame, and the knees were flexed until the thighs were in contact with the vertical wall (∼20° flexion). The focus-to-film distance was 110 cm.

The x-ray beam was centered on the femorotibial joint and tilted at an angle of 10° according to fixed flexion view (16). Medial tibial plateau alignment was assessed by fluoroscopy. In case of large intermargin distance, the beam angle was adjusted similarly to Lyon schuss view (17). At baseline and 4 years later, radiographs were performed in the same unit using the same equipment and the same radiograph technicians, who were adequately trained and used a written operations manual with a refresher training session prior to the followup radiographs. At the followup visit, the technicians visualized the initial radiographs to improve reproducibility.

Radiograph assessment

All readings were performed by 2 trained rheumatologists (DG, MA). Measurements were made separately at baseline and followup. The second reader was blind to previous measurements.

Qualitative study

Knee radiographs were read using a scoring system based on the Altman atlas (18) that assessed each tibiofemoral compartment and each side of the knee for joint space narrowing (JSN) and osteophytes independently on a 4-point scale (0–3), and assessed sclerosis on a 2-point scale (0–1). Radiographic OA was defined as a total score ≥2 in any compartment (e.g., JSN score of 1 and osteophyte score of 1 in a single compartment). Therefore, we defined women as either with (score ≥2) or without (score <2) radiographic OA. Among the women without OA, we defined healthy women as being without radiographic abnormalities related to OA (osteophyte, JSN, and sclerosis scores of 0).

The intraobserver reproducibility was satisfactory as assessed medially and laterally on 100 radiographs, with kappa scores (95% confidence intervals [95% CIs]) of 0.86 (0.68–1.0) and 0.73 (0.48–0.98) for JSN, 0.56 (0.22–0.90) and 0.93 (0.81–1.0) for osteophytes, and 0.86 (0.68–1.0) and 1.0 for sclerosis, respectively.

The interobserver reproducibility (kappa score [95% CI]), which was assessed on 100 radiographs, was good (κ > 0.70) for all evaluations at all sites except for sclerosis in the medial compartment (κ = 0.66).

Quantitative study

Radiographs were first digitized with a 190 μm pixel size according to a standardized procedure using a video camera (PAL 768*576; Sony, Tokyo, Japan), a 24-mm Nikon lens (Nikon, Tokyo, Japan), and VISIOLAB 2000 software (Explora Nova, La Rochelle, France). The distance between camera and film was fixed at 47 cm. On every film, a graduated ruler was placed to allow calibration. In every compartment, femorotibial JSW was assessed using specific software (Morpho-Expert, Explora Nova) according to the procedure detailed in Figure 1.

Figure 1.

Computerized tibiofemoral joint space width (JSW) measurement. Two vertical lines are in contact with the convexity of the femoral condyle margins. Two lines were automatically generated in every compartment, the first at 15 mm from the condyle line and the second at 15 mm from the preceding one. In the area delimited by these 2 lines, the operator detected tibial and femoral bone edges. Then, the area delimited by these lines and the least diameter of the circle included in this area were automatically measured. This diameter corresponds to the minimum JSW.

Intraobserver reproducibility (DG) assessed on 42 radiographs was satisfactory. The intraclass correlation coefficients (ICCs) were 0.91 and 0.87 in the medial compartment and 0.84 and 0.70 in the lateral compartment on the right and left sides, respectively, with a root mean square error (RMSE) ranging from 1.32–1.61. Coefficients of variation (CVs) were 2.5 and 3.3 in the medial compartment and 5.2 and 4.9 in the lateral compartment on the right and left sides, respectively.

Interobserver reproducibility assessed on 44 radiographs was also good. ICCs were 0.80 and 0.79 in the medial compartment and 0.72 and 0.65 in the lateral compartment on the right and left sides, respectively, with a RMSE ranging from 1.45–1.60. CVs were 0.8 and 0.7 in the medial compartment and 1.7 and 1.2 in the lateral compartment on the right and left sides, respectively.

Clinical evaluation.

All subjects completed a standardized medical questionnaire at baseline. The following data were recorded: age, height at age 25 years (before consequences of possible vertebral fractures), weight, height, and body mass index (BMI), calculated as weight (kg)/height (m2). Knee pain, function, and stiffness were assessed using the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (19) with a maximal theoretical total score of 96.

Biologic variables.

For each woman, fasting blood samples and first void of the morning urine samples were collected on the day of clinical assessment and kept frozen at −80°C until assayed.

Serum osteocalcin was measured using an automatic test (Elecsys N-MID Osteocalcin; Roche Diagnostics, Manheim, Germany). Serum intact N-terminal propeptide of type I collagen (PINP) was measured by radioimmunoassay (Intact PINP; Farmos Diagnostica, Upsalla, Finland) (20).

Serum β isomerized C-terminal crosslinking of type I collagen (CTX-I) was measured by an automatic system (β-Crosslaps/Serum; Roche Diagnostic). Urinary C-terminal crosslinking telopeptide of type II collagen (CTX-II), a specific marker of type II collagen found almost exclusively in cartilage, was measured by enzyme-linked immunosorbent assay based on a monoclonal antibody raised against a linear 6 amino acid epitope of CTX-II (21). Urinary glucosyl-galactosyl-pyridinoline (Glu-Gal-Pyr), a specific marker of degradation of type I and type III collagens found in the synovium tissue, was measured on nonhydrolyzed samples by high-performance liquid chromatography (22). Intra- and interassay variations were <8% for all markers.

Bone mineral density.

Bone mineral density (BMD) was measured by dual x-ray absorptiometry (DXA) with a QDR 4500 device (Hologic, Waltham, MA) at total hip at baseline. The in vivo precision error of DXA, expressed as the CV, was 1%. A control phantom was scanned every day, and all DXA measurements were performed by the same experienced operator.

Statistical analysis.

The principal outcome measure was the 4-year change in JSW. This value is the JSL, expressed in mm and as the standardized response mean, which was calculated as the mean change between 4 years and baseline divided by the SD of the change (23). Secondary outcome measures were possible predictive factors of JSL. The description of baseline characteristics in age groups was performed to compare continuous variables with an analysis of variance (ANOVA), a post hoc test when the ANOVA was significant, and chi-square tests to compare categorical variables. Since osteocalcin, serum CTX-I, PINP, urinary Glu-Gal-Pyr, and urinary CTX-II values were not normally distributed, data were logarithmically transformed before analyses.

The intra- and interobserver reproducibilities of the qualitative study were assessed by kappa statistics for each compartment, and the intra- and interobserver reproducibilities of the quantitative assessment were assessed by ICCs and RMSEs. A Student's paired t-test was used to determine the 4-year JSL, which was entered as a positive variable. Stepwise regression analysis was performed to explore possible factors correlated with the JSL, including age, BMI, WOMAC, lumbar spine and hip BMD, biologic markers, and baseline JSW. All statistical analyses were performed using Statview software, version 5 (SAS Institute, Cary, NC).

RESULTS

Population at baseline.

Baseline characteristics of the 606 women with paired radiographs according to age group are shown in Table 1. There was no significant difference between age groups for weight, height at age 25 years, and BMI. Total WOMAC scores increased with age, especially in the oldest group. Total hip BMD T scores significantly decreased with age, especially after age 60 years. Except for urinary Glu-Gal-Pyr, all biochemical markers significantly increased with age.

Table 1. Baseline characteristics of the population (n = 606) according to age group*
Characteristics<50 years (n = 106)50–59 years (n = 87)60–69 years (n = 295)≥70 years (n = 118)All (n = 606)P
  • *

    Values are the mean ± SD unless otherwise indicated. For statistical analyses, osteocalcin, CTX-I, PINP, urinary Glu-Gal-Pyr, and urinary CTX-II values were logarithmically transformed. The chi-square test was used to compare categorical variables and an analysis of variance (ANOVA) was used to compare continuous variables. When the ANOVA was significant, post hoc tests were performed. P values indicate differences across treatment groups. WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; BMD = bone mineral density; PINP = serum intact N-terminal propeptide of type I collagen; CTX-1 = β isomerized C-terminal crosslinking of type 1 collagen; CTX-II = C-terminal crosslinking telopeptide of type II collagen; Glu-Gal-Pyr = glucosyl-galactosyl-pyridinoline.

  • Significantly different from <50 years and 50–59 years.

  • Significantly different from <50 years, 50–59 years, and 60–69 years.

  • §

    Significantly different from <50 years.

  • Qualitative assessment was based on Altman score; osteoarthritis is defined by a score ≥2 in ≥1 compartment.

Age, years44.7 ± 2.755.3 ± 3.163.9 ± 2.776.1 ± 4.961.6 ± 10.5 
Weight, kg61.2 ± 10.662.1 ± 11.463.0 ± 9.860.9 ± 9.462.2 ± 10.10.24
Height at age 25 years, cm162.8 ± 6.1161.5 ± 5.5161.3 ± 5.6161.2 ± 5.4161.6 ± 5.60.10
Body mass index, kg/m223.5 ± 4.024.3 ± 4.524.9 ± 3.824.8 ± 3.824.6 ± 4.00.48
WOMAC total score3.9 ± 7.78.0 ± 14.911.0 ± 12.913.2 ± 14.99.8 ± 13.2< 0.001
Total hip BMD T score0.24 ± 0.99−0.03 ± 1.01−0.57 ± 1.06−1.37 ± 1.07−0.50 ± 1.16< 0.001
Osteocalcin, ng/ml20.7 ± 5.723.3 ± 9.627.8 ± 10.928.9 ± 12.126.1 ± 10.7< 0.001
PINP, ng/ml30.3 ± 11.835.5 ± 19.140.2 ± 19.3§37.6 ± 16.5§37.3 ± 18.0< 0.001
Serum CTX-I, ng/ml0.28 ± 0.130.36 ± 0.190.43 ± 0.23§0.46 ± 0.250.40 ± 0.22< 0.001
Urinary CTX-II, ng/mmole of urinary creatinine120.5 ± 60.1154.3 ± 124.9197.5 ± 107.3247.1 ± 126.1187.7 ± 114.9< 0.001
Urinary Glu-Gal-Pyr, nmoles/nmole of urinary creatinine5.75 ± 1.705.25 ± 1.815.40 ± 1.865.26 ± 1.415.4 ± 1.70.14
Qualitative assessment, no. (%) score ≥23 (2.8)6 (6.8)63 (21.4)53 (44.9)125 (20.6)< 0.001

Baseline JSW was weakly correlated with height at age 25 years (r = 0.10–0.19, P = 0.001–0.03) and total hip BMD T score (r = 0.10–0.26, P = 0.001–0.01), and negatively related to urinary CTX-II (r = −0.18 to −0.09, P = 0.001–0.04) according to the knee compartment. These correlations were not more significant after age adjustment.

The characteristics of the population according to the presence or absence of knee OA at baseline are shown in Table 2. Radiographic OA, defined as a value of the modified Altman score ≥2 in any compartment, was present in 125 subjects (20.6%), with a prevalence that increased with age (P < 0.001). Of the 125 subjects with OA, 81 had medial OA.

Table 2. Characteristics of women at baseline according to the presence or not of radiographic knee OA based on Altman score ≥2*
 OA (score ≥2)Non-OA (score <2)OA versus non-OA
All (n = 125)Medial OA (n = 81)All (n = 481)No abnormalities (score 0) (n = 358)PP after age adjustment
  • *

    Values are the mean ± SD. OA = osteoarthritis; JSW = joint space width; see Table 1 for additional definitions.

Age, years68.9 ± 8.569.6 ± 8.459.8 ± 10.259.1 ± 9.8< 0.0001 
Body mass index, kg/m225.8 ± 4.426.3 ± 4.224.2 ± 4.423.8 ± 3.6< 0.00010.24
WOMAC total score14.3 ± 14.114.3 ± 3.68.6 ± 12.88.1 ± 12.4< 0.0010.11
Total hip BMD T score−0.57 ± 1.17−0.55 ± 1.17−0.48 ± 1.16−0.50 ± 1.140.430.67
Osteocalcin, ng/ml27.8 ± 10.228.4 ± 9.825.8 ± 10.825.7 ± 10.90.020.03
PINP, ng/ml38.3 ± 15.339.0 ± 14.837.1 ± 18.636.79 ± 18.360.160.03
Serum CTX-I, ng/ml0.43 ± 0.200.44 ± 0.210.39 ± 0.220.39 ± 0.230.030.22
Urinary CTX-II, ng/mmole of urinary creatinine253.1 ± 143.2263.4 ± 149.5170.0 ± 99.8161.1 ± 92.7< 0.0010.39
Urinary Glu-Gal-Pyr, nmoles/nmole of urinary creatinine5.38 ± 1.705.20 ± 1.415.42 ± 1.765.42 ± 1.760.810.07
JSW, mm      
 Medial right4.55 ± 1.384.26 ± 1.355.10 ± 0.765.14 ± 0.74< 0.0001< 0.0001
 Medial left4.59 ± 1.284.32 ± 1.435.14 ± 0.815.18 ± 0.78< 0.0001< 0.0001
 Lateral right5.25 ± 1.575.48 ± 1.536.02 ± 1.196.07 ± 1.16< 0.00010.09
 Lateral left5.41 ± 1.595.62 ± 1.595.98 ± 1.116.04 ± 1.09< 0.00010.01

The 125 subjects with radiographic OA were older and had higher BMI, WOMAC total score, serum CTX-I, urinary CTX-II, and osteocalcin than the 481 subjects without OA. After age adjustment, BMI, WOMAC total score, and urinary CTX-II were not more significantly associated with the presence of OA. Baseline JSW was significantly lower for subjects with OA than for subjects without OA (P < 0.0001) in every compartment, but more markedly in the medial compartments after age adjustment.

In the cross-sectional analysis, baseline JSW significantly decreased with age for all subjects in every compartment (r = −0.12 to −0.16, P < 0.001) (Table 1 and Figure 2). JSW was 15–18% lower in the oldest group of women (age >70 years, n = 118) than in the youngest group of women (age <50 years, n = 106) according to the knee compartment.

Figure 2.

Minimum joint space widths ± SEM in all subjects (n = 606) for medial and lateral tibiofemoral compartments of both knees. * Age 70 years versus age <50 years; P < 0.0001 for all compartments. equation image Age 70 years versus age 50–59 years; P ≤ 0.0004 for all compartments. ‡ Age 70 years versus age 60–69 years; P < 0.0001 for all compartments. □ Age 60–69 years versus age <50 years; P < 0.05 for all compartments. ANOVA = analysis of variance.

In the 481 women without radiographic OA (score <2), the minimum JSW significantly decreased with age (r = −0.13 to −0.24, P < 0.001). JSW was 7–16% lower in women age >70 years (n = 65) than in women age <50 years (n = 103) according to the knee compartment.

In the 358 healthy subjects (score 0), JSW significantly decreased with age (r = −0.16 to −0.20, P = 0.0002–0.03). JSW was 7–14% lower in women age >70 years (n = 38) than in women age <50 years (n = 78).

In the 125 women with radiographic OA (score ≥2), JSW significantly decreased with age (r = −0.39 to −0.29, P < 0.001). JSW was 18–31% lower in women age >70 years (n = 53) than in women age <50 years (n = 3) in the medial compartments.

JSL over 4 years.

The 4-year JSL according to the compartments is shown in Table 3. In all subjects (n = 606), JSW significantly decreased over 4 years in every compartment (P < 0.02) at a rate of 2–3 times higher in the medial than in the lateral compartments. The mean ± SD 4-year JSL was 0.31 ± 0.68 mm and 0.30 ± 0.77 mm in the medial compartment and 0.11 ± 1.11 and 0.16 ± 1.09 mm in the lateral compartment on the right and left sides, respectively. The mean ± SD annual rate of JSL was 0.08 ± 0.17 mm and 0.08 ± 0.19 mm in the medial compartment and 0.03 ± 0.28 mm and 0.04 ± 0.27 mm in the lateral compartment on the right and left sides, respectively. When this was calculated as a percentage of the initial JSW, this represented a 4-year loss of ∼6% and ∼2% for the medial and lateral compartments, respectively.

Table 3. Joint space loss over 4 years in all subjects and in women without any radiographic abnormalities related to OA (score 0)*
 Joint space loss, mean ± SD mmSRMP
  • *

    OA = osteoarthritis; SRM = standardized response mean.

  • By paired t-test.

All subjects (n = 606)   
 Medial right0.31 ± 0.680.46< 0.0001
 Medial left0.30 ± 0.770.39< 0.0001
 Lateral right0.11 ± 1.110.100.017
 Lateral left0.16 ± 1.090.150.0005
Non-OA subjects   
 Score <2 (n = 481)   
  Medial right0.31 ± 0.680.46< 0.0001
  Medial left0.27 ± 0.790.34< 0.0001
  Lateral right0.06 ± 1.090.060.10
  Lateral left0.10 ± 1.060.090.04
 Score 0 (n = 358)   
  Medial right0.35 ± 0.650.54< 0.0001
  Medial left0.29 ± 0.750.39< 0.0001
  Lateral right0.10 ± 1.030.100.06
  Lateral left0.15 ± 1.00.150.004
Subjects with radiographic OA   
 All (n = 125)   
  Medial right0.28 ± 0.860.330.0005
  Medial left0.39 ± 0.850.46< 0.0001
  Lateral right0.21 ± 1.380.150.09
  Lateral left0.37 ± 1.350.270.003
 Medial OA (n = 81)   
  Right and left0.32 ± 0.760.420.0003

In the 481 subjects without radiographic OA (score <2) and the 358 women without any radiographic abnormalities related to OA at baseline (score 0), the pattern of JSL was similar to the entire population, although the loss tended to be lower in the lateral compartments (Table 3). In the latter population, the mean ± SD annual rate of physiologic JSL was 0.09 ± 0.16 mm and 0.07 ± 0.19 mm in the medial compartment and 0.03 ± 0.26 mm and 0.04 ± 0.25 mm in the lateral compartment on the right and left sides, respectively. When this was calculated as a percentage of the initial JSW, it represented a 4-year loss of ∼6% and ∼2% for the medial and lateral compartments, respectively.

In the 125 subjects with radiographic OA at baseline, JSW of the involved compartment significantly decreased over 4 years by mean ± SD 0.28 ± 0.86 mm and 0.39 ± 0.85 mm on the right and left sides, respectively (P < 0.0001). The mean ± SD annual rate of physiologic medial JSL was 0.07 ± 0.22 mm and 0.10 ± 0.21 mm on the right and left sides, respectively. When this was calculated as a percentage of the initial JSW, this represented a 4-year loss of ∼7% and ∼5% for the medial and lateral compartments, respectively.

In the 81 subjects with medial radiographic OA, there was a mean ± SD medial JSL of 0.32 ± 0.76 mm (P = 0.0003) over 4 years, with a mean ± SD annual rate of medial JSL of 0.08 ± 0.19 mm. This represented a 4-year loss of 8% compared with initial JSW.

When grouping the 4-year JSL values of right and left sides (no significant difference between sides) we found no significant difference in medial JSL between OA and non-OA subjects.

Predictive factors of JSL.

Several possible risk factors were investigated to predict JSL (Table 1). We found that clinical risk factors, BMD, and biologic markers were not predictive of JSL. Multiple regression analysis demonstrated that initial JSW in all subjects was the only significant predictor of JSL (r = 0.33 and 0.25 for the right and left medial compartments, respectively; P < 0.0001), suggesting that women with the most important baseline JSW will experience the most severe JSL over 4 years. For healthy women (score 0), we found similar results (r = 0.31 and 0.35 for the right and left medial compartments, respectively; P < 0.0001).

For subjects with medial OA, none of the studied factors, including baseline JSW of the involved compartment, were predictive of JSL.

DISCUSSION

In this cross-sectional analysis of a cohort of 606 healthy female volunteers, we found a significant decrease in JSW with age, which was 15–18% lower in the oldest group (age ≥70 years) than in the youngest group (age <50 years). These results were confirmed by the longitudinal analysis with a significant 4-year JSL of ∼0.30 mm (6%) for the medial compartment, which is greater than in the lateral compartment. Similar results were observed in healthy subjects without any radiographic abnormalities related to OA, suggesting a physiologic JSL with age.

It is well-established that age is a strong risk factor for knee OA (24). Its incidence and prevalence increased 2–10-fold from age 30–65 years and increased further thereafter (25). However, the natural progression of age-related JSW has previously been poorly studied, and there are conflicting results in relation to the existence of age-related JSL in the absence of OA.

Our results are consistent with a study by Dacre et al (12), who demonstrated a significant decrease in JSW with age in a cross-sectional analysis of 685 normal knee radiographs performed with an automated measure. It differs from the present study in that the radiographs were taken after minor knee trauma or in subjects with knee pain who presented to an emergency department. Furthermore, it was performed in both women and men with a median age of 35 years. Conversely, another cross-sectional study was performed by Lanyon et al (13), who found no significant reduction in JSW with age in 125 asymptomatic subjects (no osteophytes and without knee pain). These authors provided evidence that OA is a specific process and not an inevitable part of aging. In that study, JSW was measured by metered calipers, a manual method of measurement that may not be sensitive enough to detect changes. Cross-sectional analyses may be biased by secular changes, and JSL with age is better assessed in a longitudinal study. To our knowledge, the present study is the first to longitudinally assess JSL in a population-based cohort of women.

Most of the previous studies were performed in patients with knee OA defined by a K/L scale score ≥2 (26). One longitudinal study assessed JSL in knees without radiographic OA in elderly men and women (ages 70–79 years) with knee pain (27). Among this population, 111 subjects with a normal JSW (K/L scale score 0) had a mean ± SD medial tibiofemoral 3-year JSL of approximately 0.14 ± 0.53 mm. The authors showed that JSL increased with the severity of baseline JSW. In our study, we reported a range of 0.29–0.35 mm over 4 years in healthy subjects (score 0). We did not detect any difference in JSL between subjects with and without radiographic OA, but our definition of OA is based only on radiographic features, excluding knee pain and using the Altman score. This scale is interesting for longitudinal evaluation but may be not ideal to identify OA. Mean values of WOMAC and JSW at baseline in women with and without OA were statistically different, but the severity of OA might be insufficient to show a difference in the evolution of JSW compared with healthy subjects.

In women with medial radiographic OA, we found a medial 4-year JSL of 0.32 mm, similar to that reported in the control groups of several randomized controlled trials, including 0.31 mm (28) and 0.29 mm (29) over 3 years in 2 studies examining the effect of glucosamine, and 0.17 mm over 2 years in a recent study of patients treated with glucosamine and/or chondroitin sulfate, celecoxib, or placebo (30). A trial of risedronate demonstrated a similar JSL of 0.09 mm per year in a European cohort. Conversely, a higher rate of 0.29 mm over 1 year was reported in a randomized controlled study of chondroitin sulfate (31).

The rate of JSL differs according to the radiographic protocol and the quality of alignment of the medial tibial plateau (32). A recent study compared the fixed flexion (16) and Lyon schuss (17) techniques in a 12-month longitudinal study (11). They reported a JSL of 0.22 mm with the Lyon schuss view and −0.01 mm with fixed flexion view in knees with OA, suggesting that the Lyon schuss view was more sensitive because of fluoroscopic alignment of the medial tibial plateau. These authors have assessed JSW in 99 normal knees (K/L scale score 0) and did not find any variation over 12 months with both views. The duration of the study was probably too short to detect a JSL.

We found no predictive factors of JSL. In particular, BMI is usually considered a risk factor of incident OA. However, the association between mean BMI and JSL was not found in a recent review (26). Our results are in contrast with studies showing a positive association between increased urinary CTX-II levels and progression of knee (33) or hip (34) OA, but are consistent with a study by Mazzuca et al, who found no relationship between urinary CTX-II and JSL (35). Type II collagen constitutes the most abundant protein of intervertebral discs. Lumbar spine disc degeneration could contribute to increase urinary CTX-II independently of radiologic knee OA (36).

We found that the initial JSW was predictive of JSL in medial compartments, i.e., the higher the baseline JSW was, the higher the 4-year JSL. This result is consistent with a subanalysis of a 3-year study of 212 patients with knee OA evaluating the effects of glucosamine sulfate with a strong correlation between baseline JSW and JSL, suggesting that subjects with less severe radiographic OA will experience the most severe radiographic progression (37). These data were also suggested in a study showing a decrease of tibial cartilage volume at a rate of 5% per year assessed by magnetic resonance imaging in 123 subjects followed for 2 years (38). The main predictive factor was the initial cartilage volume, with a more rapid loss early in the disease. Alternatively, the fact that baseline JSW was a predictor of JSL could be a bias related to the regression to the mean due to random error or variability in the measurement (39, 40).

Our study has several strengths and some limitations. The radiographs were performed using a standardized radiographic protocol and semiautomated measurements in a population-based cohort of women. To our knowledge, it is the first longitudinal study involving a large number of women without OA over a 4-year period. Radiographs were obtained in a single radiographic unit using the same equipment by the same trained technicians. In contrast to most studies, we also assessed JSW in the lateral compartment. We found that baseline JSW was greater in lateral than in medial compartments, and that the 4-year JSL was lower in lateral than in medial compartments. Our study is limited by the small number of subjects with radiographic OA and by the definition of OA based only on radiographic features. In addition, we did not report data on varus–valgus alignment, ligamentous laxity, meniscal damages, knee trauma, or concurrent OA in other joints, usually considered risk factors for progressive knee OA.

In conclusion, our results strongly suggest, both in the cross-sectional and longitudinal studies, an age-related JSL, even after exclusion of women with radiographic OA. We did not find a significant difference in JSL between subjects with and without radiographic OA, probably because of the small number of OA subjects and the definition of OA using only radiographic features. Further studies comparing these 2 populations, including a higher number of well-established OA subjects, are required to confirm these findings.

AUTHOR CONTRIBUTIONS

Dr. Gensburger had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Gensburger, Arlot, Sornay-Rendu, Roux, Delmas.

Acquisition of data. Gensburger, Arlot, Sornay-Rendu, Roux, Delmas.

Analysis and interpretation of data. Gensburger, Arlot, Sornay-Rendu, Roux, Delmas.

Manuscript preparation. Gensburger, Arlot, Sornay-Rendu, Roux, Delmas.

Statistical analysis. Gensburger, Arlot, Sornay-Rendu, Roux, Delmas.

Ancillary