Osteoarthritis (OA) is a major cause of morbidity, affecting 60% of men and 70% of women over the age of 65 years 1 . In particular, the prevalence of knee OA has been shown to increase with age throughout the elderly years and affect women more than aged-matched men (11% and 7% respectively) 2 . Progressive destruction of articular cartilage and synovial inflammation are the hallmarks of OA 1 , with the severity of clinical disease being proportional to articular cartilage lost 3 . However the pathogenesis of this common condition remains poorly understood. Whether subclinical systemic inflammation precedes the clinical manifestation of OA is not known. Although the systemic inflammatory response that characterizes other forms of arthritis, such as rheumatoid arthritis (RA), is not usually manifest in OA, levels of the acute phase reactant C-reactive protein (CRP) have been found to be elevated in some individuals with established OA 4 5 6 . However, in established OA, serum levels of CRP have not been found to correlate with either radiographic joint space width in a cross-sectional study 4 or disease severity 5 .

Magnetic resonance imaging (MRI) has good tissue contrast and anatomical resolution 7 , thus allowing a non-invasive examination of the joint structure in pre-disease or early stage OA. MRI can visualize joint structure directly 8 9 and has been recognized as a valid, accurate and reproducible tool to measure articular cartilage volume 9 10 . Loss of knee cartilage has been shown to be related to knee pain 11 and joint replacement with more than 10% lost before any radiological changes can be detected 12 .

As the knee is one of the most common joints to be affected by OA 1 , and OA commonly occurs in women in late mid-life, to further explore whether low grade systemic inflammation is associated with characteristics of articular cartilage we have examined the relationships between high sensitivity (hs)CRP and tibial and patella cartilage volumes measured by MRI in asymptomatic women at mid-life. As the evaluation of the direct relationship between hsCRP and knee cartilage is potentially complicated by the concurrent relationships between CRP and increased body fat, notably intra-abdominal fat 13 14 , and body mass index and knee cartilage volume 15 , we have taken an analytical approach to establish the independent effects of hsCRP on knee cartilage.

Participants were women recruited to a cross-sectional study of the role of androgens in women using a database established from the electoral roll in the southern Australian state of Victoria between April 2002 and August 2003 16 . The database of eligible women was created using the following methodology: women were recruited by telephone using a database of individuals from household addresses selected at random on a weekly basis from Australian electoral areas. Starting addresses were selected at random from the electoral roll for each of the sampling points. Interviews were conducted in person on Saturdays and Sundays between 9.00 am and 4.00 pm. Eight interviews were conducted per sampling point. Only one eligible person was recruited per household, and people recruited to the sample tended to stay on the active database for about 2 years. Women underwent telephone screening and were excluded if they were pregnant or less than 6 weeks post-partum, or had experienced any of the following in the preceding 3 months: an acute psychiatric illness; acute renal, liver, cardiovascular disease or any other acute major illness; gynecological surgery; active malignancy or cancer treatment, excluding non-melanotic skin cancer. All participants provided a detailed medical history in response to specific questioning and those with an active medical condition were not recruited.

Women from the original cross-sectional study were eligible for this study if they were aged 40 to 67 years, had not undergone a hysterectomy and had agreed to be re-contacted for further studies. As our intent was to investigate subjects with no significant current or past knee disease, individuals were excluded if they had had any of the following: a clinical diagnosis of knee OA as defined by American College of Rheumatology criteria, knee pain lasting for > 24 h in the last 5 years, a previous knee injury requiring non-weight bearing treatment for > 24 h or surgery (including arthroscopy), or a history of any form of arthritis diagnosed by a medical practitioner. A further exclusion criterion was a contraindication to MRI including pacemaker, metal sutures, presence of shrapnel or iron filings in the eye, or claustrophobia. Of the 355 women who fulfilled these criteria, 176 remained eligible after further exclusions based on the above or the subject being unlikely to be available to complete the longitudinal protocol of re-assessment at 2 years.

Each participant attended for a single morning fasting blood test and measurement of height (cm) and weight (kg) 1 to 2 years (mean 1.53 years, standard deviation (SD) 0.24 years) prior to their knee joint MRI. At this time each woman completed a questionnaire that provided information about smoking and alcohol consumption. Fasting blood drawn at the time of recruitment to the original cross-sectional study was stored at -80°C until assayed.

The study was approved by the Southern Health Human Research and Ethics Committee and the Monash University Human Research and Ethics Committee, Clayton, Victoria, Australia and all participants gave written informed consent. All participants were identified by number, not by name, and the study was conducted in accordance with the Declaration of Helsinki principles.

The mean age of the 176 women included in the analysis was 52.3 (6.6) years. Other participant characteristics are listed in Table 1.

The univariate results are presented in Tables 2 and 3. For tibial cartilage volume, the variables of particular interest from the univariate analysis included bone area, age and ln hsCRP. For the patella cartilage volume, the variables of interest included bone volume, age, BMI and ln hsCRP.

The r² values for factors potentially contributing to total tibial cartilage and patella volumes are shown in Tables 4 and 5. For total tibial cartilage, total bone area explained 14% of the variation in volume. The addition of age increased the proportion of the variation in cartilage volume explained to 21%. The further addition of smoking and alcohol to this model did not significantly increase the proportion of variation in total tibial cartilage volume explained, however we retained these variables in the analysis to form our standard model against which to test the addition of both BMI and ln hsCRP. There was no significant additional benefit with the addition of BMI. However, the inclusion of the CRP term increased the proportion of cartilage volume variation explained to 28%: this was a highly statistically significant increase in the r² value, and this was seen whether BMI was included in the model or not (Table 4).

For the patella, 27% of the variation in patella cartilage volume was associated with patella bone volume. This was increased significantly to 30% with the addition of age but was essentially unchanged by the addition of the smoking and alcohol terms, although again we retained these terms in our standard model against which to test the effect of the addition of the BMI and ln hsCRP terms. Addition of either BMI or ln hsCRP separately increased the proportion of variation in patella cartilage volume explained to 32% (the change did not reach statistical significance at the 5% level in either case). Although the addition of both the ln hsCRP and BMI terms increased the r² value by 2%, again the change did not reach statistical significance at the 5% level (Table 5).

The principal finding of this study was that in women at mid-life, the level of hsCRP in serum made a significant independent contribution to the variation in the volume of total tibial cartilage measured approximately 12 to 18 months later. This observation was true whether or not the analysis was controlled for BMI. This observation did not hold true for the patella compartment. Although BMI is a major determinant of CRP 21 , our analysis indicates that the impact of hsCRP is not purely mediated through BMI.

While increased hsCRP has been reported in the early phases of OA 22 , this study provides the first evidence of an independent relationship between hsCRP and tibial cartilage volume in asymptomatic individuals. Although OA is not considered a classical inflammatory arthropathy, it is characterized by intra-articular inflammation manifest as synovitis. In this study we were not able to assess synovial effusion or thickness from the MRI. However, recent studies have shown a significant relationship between hsCRP levels and the degree of synovial inflammatory infiltration in OA 23 and RA 24 .

CRP is an acute-phase protein that is produced in large amounts by hepatocytes, upon stimulation by the cytokines interleukin (IL)6, tumor necrosis factor α (TNFα) and IL1, during an acute-phase response. Recent studies consistently indicate that both TNFα and IL1 play a key role in cartilage destruction in OA 25 26 . Although OA disease activity appears to be limited to involved joints, release of cytokines into the systemic circulation during the subclinical phase of OA may explain the association we have seen between hsCRP and tibial cartilage volume. The role of CRP in OA may be complex, possibly involving specific haplotypes of the CRP gene 27 . The relationship between CRP levels and BMI may also be complex and involve modulation of CRP gene expression within adipose tissue 28 .

Why the relationship with hsCRP was seen in the tibial compartment but not the patella compartment is unclear. However, we have previously reported a lack of association between patella and tibial cartilage loss over 2 years, suggesting that pathogenetic mechanisms for OA in the patellofemoral and tibiofemoral joint may differ 19 . The assessment of the presence/absence of cartilage defects provides another method of assessing cartilage health. In our study, the independent contribution seen by hsCRP to variation in tibial cartilage volume was not seen between hsCRP and the presence of cartilage defects in the tibiofemoral compartment (data not shown). Our findings that age and bone size influence knee cartilage volume corroborate previous findings 29 .

Strengths of our study include the community-based recruitment of asymptomatic women, the relatively large number of study participants, and measurement of CRP by a highly sensitive assay. A potential limitation of this study is that hsCRP was only measured on one occasion. Although serial sampling of CRP would have been optimal, it would not have been feasible as many of the women in rural areas had to travel considerable distances for blood sampling. To limit the possibility that CRP levels may have been affected by intercurrent illness, women were asked not to attend for their blood test on a day they were not well. Another limitation to our study is that the cross-sectional design does not provide information regarding the usefulness of hsCRP in predicting cartilage loss. However, re-assessment of our study population after 2 years will provide data for the predictive value of hsCRP for tibial cartilage loss.

That hsCRP measured several months prior to MRI scanning of the knee was significantly associated with less tibial cartilage volume suggests the possibility that the presence of subclinical systemic inflammation has a role in the early stages of the disease process that precedes symptomatic OA. However, the present report was a cross-sectional study and none of the women in this study had clinical evidence of OA. Nonetheless, our findings indicate that assessing whether inflammation precedes clinical OA merits specific investigation. If activation of inflammation is found to have a pathogenic role in OA development this could have implications as a target for preventative therapy.

BMI = body mass index; CRP = C-reactive protein; CV = coefficient of variation; hsCRP = high sensitivity CRP; IL = interleukin; ln = natural log; OA = osteoarthritis; RA = rheumatoid arthritis; SD = standard deviation; TNFα = tumor necrosis factor α.

The authors declare that they have no competing interests.

FH was responsible for data collection and analysis and interpretation of the data. RB was responsible for analysis and interpretation of the data. AW performed data analysis. FC and SDavis were responsible for study design and data analysis and interpretation and SDavison reviewed the final draft. All authors prepared, read and approved the final manuscript.