However, the relationship between obesity and OA in hand(s) and hip(s) remains controversial [7,8]. A significant relationship between obesity and radiographic hip OA has been found in some cross-sectional studies [9-13] as well as in longitudinal studies [14-16]. In large longitudinal studies of Gelber et al [4] and Reijman et al [6] high BMI was not associated with hip OA. Large cross-sectional studies have failed to show a significant association between obesity and hand OA in either males or females [17-19], whereas some prospective data have demonstrated that obesity predicted hand OA [14,20].

From 1994 to 2004, 5.8% (CI 4.3–7.3) developed OA in the hips, 7.3% (CI 5.7–9.0) in the knees, and 5.6% (CI 4.2–7.1) in the hands. The 10-year incidence for hand OA was significantly (p = 0.001) higher among women (5.6%, CI 4.2–7.1) compared to men (2.5%, CI 1.3–3.6), whereas no significant differences between the genders were found in the incidence of OA in hip (5.8%, CI 4.3–7.3 for women and 3.8%, CI 2.4–5.2 for men, p = 0.060) and knee (7.3%, CI 5.7–9.0 for women and 6.2%, CI 4.4–7.9 for men, p = 0.346) (Table ​(Table1).1). Only a minor proportion of the responders reported OA in more than one body region: 1.3% reported hand/knee OA, 1.4% hip/knee OA, 1.1% hip/hand OA, and 0.6% reported OA in all three body regions (hip, knee, and hand).

Table ​Table11 shows that there were few cases with OA in the lowest (< 20) and highest (> 30) categories of BMI. Obesity, defined as BMI above 30, was significantly associated with knee OA 10 years later with an OR of 2.8, and a dose-response relationship was found for this association (Table ​(Table2).2). Table ​Table22 also shows that obesity was significantly associated with hand OA with an OR of 2.6 in the multivariate analysis. The analyses did not reveal any indication of an association between obesity and hip OA. For hand- and knee OA, there was no statistically significant interaction effect between gender and BMI, age and BMI, or any of the other confounding variables and BMI.

The main limitation of this study is that the epidemiological case definition of clinically diagnosed OA was based on self-report through the response to a written question and not based on radiographic evidence. However, the present outcome question referred to OA diagnosed by a medical doctor, and not only to pain in the actual body regions. In a previous study, we found that this OA question to a large extent differentiated between OA and musculoskeletal pain in the actual joints as approximately 18% of those who reported pain also reported OA in the same joint [23]. Moreover, the present 10-year incidence estimates of hip and knee OA is at the same level as the radiographic based estimates in the recent published study of Reijman et al [4,6], who found that 5.5 % and 3.9% developed radiographic incident knee OA and hip OA, respectively, during a mean follow-up time of 6.6 years. However, it's obvious that with the current definition this survey will not capture OA without pain or other symptoms. Hence, the current findings relate to obesity and symptomatic OA. It can be argued that the relationship between obesity and OA is mostly of interest in symptomatic OA, and that radiographically diagnosed ("silent") OA perhaps has less practical interest for clinicians. In a research perspective there is definitely a need to explore the precision of self-reported diagnosis of OA.

Another limitation may be a response of 64% with the lowest response among males and among the youngest and oldest age groups. The oldest age group born in 1918–1920 was excluded. As expected, a low incidence of OA was observed in the youngest age group. Thus the results in the present study are not likely to be influenced by low response in this group. However, there was a higher proportion of women than men in this sample, which should be taken into account when interpreting these results. Furthermore, the number of OA cases was small among subjects with BMI < 20 and > 30, and the results for these groups should be interpreted with caution. A final limitation is that we lack data on the reliability of self-reported BMI. However, self-reported and measured BMI mostly correlate well, even though there has been reported a tendency to under report the BMI, especially among adolescents with overweight [24].

The main strength of this study is that it was carried out in the general population with people aged between 24 and 66 years at the start of the study. Furthermore, this prospective study had a relatively large number of respondents taking into consideration the long follow-up period of 10 years. Population-based studies are important as the cases are unselected for severity in comparison with hospital-based populations.

The present results confirm that obesity is a strong determinant for knee OA [1-6]. A non-significant relationship between obesity and hip OA is also in line with two previous large longitudinal studies [4,6]. However, some other large prospective studies have reported different results regarding the impact of high BMI on hip OA [14-16]. A systematic review of the influence of obesity on hip OA included five longitudinal and seven cross-sectional studies, and found moderate evidence for a positive association between obesity and hip OA with an OR of approximately 2 [25].

Similarly, it is an open question whether obesity is associated with an increased risk of hand OA [7]. Our results support that a relation is possible, but the results were less consistent and overall weaker than the association that were observed between BMI and knee OA. Furthermore, we did not find any dose-response effect so this finding should be interpreted with carefulness. Since only a minor proportion of 1.3% reported OA in both hand and knee, it's not likely that this overlap can explain the BMI and hand OA association.