Osteoarthritis (OA) and osteoporosis are both common conditions in elderly women [1, 2]. It would be anticipated that the conditions frequently coexist due to their high prevalence, but some studies have suggested that there is an inverse association between the occurrence of OA and osteoporosis [3–12]. Increased bone mass has been proposed as a risk factor for OA independent of age and other factors, because higher bone mass may increase bone stiffness and thereby increase loading of articular cartilage and lead to cartilage damage [13, 14]. Alternatively, OA and osteoporosis may only be indirectly related through weight, mechanical stress, genetic factors, or other factors that affect both bone mass [15, 16] and the risk of OA [17, 18].

Recent reviews reported that bone mass or bone mineral density (BMD) (measured with various techniques at different skeletal sites) was significantly increased in the OA cases compared with age- and sex-matched controls [19–22]. A recent large epidemiologic study showed higher bone density at both central (hip and lumbar spine) and peripheral (radius and calcaneus) sites in elderly women with moderate-severe radiographic grades of hip OA after adjusting for age, weight and other anthropometric characteristics [11]. Population-based epidemiologic studies have shown that women with radiographic changes of knee [5, 9, 12, 23] or hand OA [5, 10, 12] have increased bone mass after adjustment for age and other covariates, compared to those without OA.

Most previous studies reported the association of OA with bone mass among Caucasian populations, and there are few data among Asian populations. One study reported that Japanese patients with knee OA had greater bone mass at the hip and lumbar spine compared to age and weight matched controls [24]. We reported elsewhere that, compared to Caucasians, Japanese women appear to have a substantially higher prevalence of knee OA, but much lower prevalence of hand OA except distal interphalangeal (DIP) joint [25]. This suggests that the prevalence and/or types of risk factors for OA may differ between Japanese and Caucasian populations.

Although BMD declines throughout the skeleton with age, there is significant variation between skeletal sites, even within an individual. Therefore, we hypothesized that bone measurements at the heel might be more strongly related to OA at the knee, since both sites are weight-bearing, whereas hand BMD measurements might be more strongly related to hand OA. Furthermore, it has been suggested that certain features of OA (such as osteophytes) may be reparative in nature, whereas other features (such as joint cartilage damage) are destructive, and may therefore have a different relationship with bone mass at nearby (or distant) skeletal sites. To investigate these issues, as well as the general issue of whether bone density is associated with OA in Japanese, we examined the cross-sectional association between radiographic knee or hand OA and appendicular bone mass at weight-bearing and non-weight-bearing sites among a community-based sample of Japanese women. More specifically, we examined the association between bone mass and different severity levels of radiographically defined knee and hand OA.

The Hizen-Oshima Study is a prospective population-based cohort study of musculoskeletal conditions (i.e., osteoporosis and osteoarthritis). We recruited community-dwelling women aged 40 years or over in Oshima town, Nagasaki prefecture, Japan. The women were identified by the municipal electoral list and contacted through mailings. As of March 31, 1999, the town of Oshima had a population of 5817 (2840 male and 2977 female), in which the population of women aged 40 or over was 2006. The proportion of women aged 40 or over in Oshima (2006/5817 = 34%) was somewhat higher than the national average of 27%, based on the 1995 Census for all of Japan. All women aged 40 or over were invited to participate. The baseline examination was performed at the Oshima Health Center between 1998 and 1999. A total of 586 women (approximately 30% of eligible women) participated in the study. The response rates in the age groups 40–49, 50–59, 60–69, 70–79, and 80 years or over were 10, 28, 45, 39, and 15%, respectively. The low response among younger women may reflect time constraints of work or family, or they may have not been interested in their musculoskeletal health. Low response among the oldest women may reflect difficulty attend the study site because of poor health or lack of transportation. We compared the age distribution of participants vs. non-participants. In total, the mean age of participants (63.9 years) was significantly higher than that of non-participants (61.1 years). Despite having a shipyard in the town, Oshima is a mainly rural (farming/fishery) district. Approximately half of the women who participated in the study continue to grow rice and vegetables by manual labor, sometimes using machinery. All participants were noninstitutionalized, living independently at baseline, and were able to ambulate independently (with or without a cane), and all women except one were able to go up and down stairs. All subjects gave written informed consent before examination. This study was approved by the local Ethics Committee.

Antero-posterior weight-bearing knee radiographs were obtained with the subjects' knees in full extension, according to the protocol by Framingham Osteoarthritis Study [25]. Postero-anterior hand radiographs were obtained with the central ray focused on the 3rd metacarpophalangeal (MCP) joint; each hand film was taken separately. Radiographs were graded according to the criteria described by Kellgren and Lawrence (K/L) [26]. In addition, since the K/L system of radiographic grading for OA does not differentiate between osteophytic and joint space narrowing changes of OA [23, 27], we separately evaluated the presence of osteophytes and of joint space narrowing for each subject, without regard to the K/L grade; first, we evaluated the osteophytes for all subjects, after that the joint space narrowing. Each joint was scored for the severity of both osteophytes and joint space narrowing, using categories of none (0), questionable (1), or definite (2). Each subject was classified according to the most severe grade within a given joint group for each of the radiographic findings (K/L grade, osteophytes or joint space narrowing). We examined the following joint groups: the knees, the second and third distal interphalangeal (DIP), the proximal interphalangeal (PIP), and the metacarpophalangeal /carpometacarpal (MCP/CMC) joints of the hands. We reported earlier the prevalence of radiographic OA of the knee and hand between Japan and the United States [25]. To standardize readings between the studies, the Japan reader (S.Y.) trained with the primary reader of the Framingham Osteoarthritis Study prior to reading the Japanese films, and used the same atlas and grading scheme. To evaluate interrater reliability for OA grade, we randomly selected 40 subjects from each population and had the Framingham and Japan readers read each set of films. In testing whether OA was present or absent, the kappa statistic was 0.77 (p < 0.001) for knee and 0.68 (p < 0.001) for hand (combining all joints), indicating good agreement between readers. Intrarater reliability (Kappa) for OA grade was 0.77 for the knee (p < 0.001), but was not evaluated separately for the hands. Spearman correlation coefficients between osteophytes and joint space narrowing were 0.59 (p < 0.001) for knee, 0.63 (p < 0.001) for DIP, 0.38 (p < 0.001) for PIP and 0.59 (p < 0.001) for MCP/CMC. Our radiograph reader (S.Y.) was blinded to the subject's age, bone mass and anthropometric characteristics.

Quantitative ultrasound of the right calcaneus (heel) bone was determined by measurement of broadband ultrasound attenuation (BUA) and the speed of sound (SOS) using the Achilles ultrasound bone densitometer (Lunar Corp., Madison, WI). Stiffness index, a function of BUA and SOS measurements, was calculated according to the manufacturer's equation. Measurements of the right second metacarpal bone mineral density (mBMD) were obtained from hand radiographs using an optical densitometer, the Bonalyzer II (Teijin Ltd., Tokyo). The reproducibility expressed as coefficient of variation (CV), measured on 5 times consecutively within 1 h with repositioning between each measurement, was 1.71% for stiffness index calculated in 17 healthy volunteers, and 0.56% for mBMD measurements calculated in a random sample of 20 of the subjects.

Height (m) and weight (kg) were measured with the subject in light clothing and without shoes, and body mass index (BMI) was calculated as weight (kg)/height (m²).

We did not detect a consistent association of peripheral bone mass with radiographic features of OA in the knee, DIP and PIP joints. However, we found that OA at the CMC/MCP was associated with higher bone mass: stiffness index was significantly higher in women with grade 3, definite osteophyte or definite joint space narrowing, compared with those with no OA, no osteophyte or no joint space narrowing, respectively, and mBMD was higher in women with definite osteophytes, compared with those with no osteophytes. Since it is well known that age and obesity are related both to OA and bone mass, and are potential confounders [15–18], our analyses were adjusted for age and BMI.

Several studies reported that radiographic knee OA was associated with increased femoral neck or spine BMD independently of age and BMI [5, 9, 23], whereas knee OA was not associated with radius or metacarpal BMD [23, 28]. Furthermore, several studies reported that radiographic hand OA was associated with increased lumbar spine or total body BMD [5, 10, 12], but not with radius or metacarpal BMD [29, 30]. It has been reported that in cases with generalized OA, the width of the radius and the metacarpal is significantly larger than in controls, and bone width can influence bone mass results expressed as BMD or % cortical area [22]. In the current study, metacarpal width in women with OA (grade 1–4) at CMC/MCP joint was greater than that in women without OA (grade 0) at the CMC/MCP joint, but this difference was not significant (p = 0.28). These findings may in part explain why we did not find an association of mBMD with OA grade and joint space narrowing status at the CMC/MCP joint despite finding an association of stiffness index with these measures of OA.

In our study, the associations with OA were inconsistent, suggesting they may simply be due to chance. The cross-sectional nature of our study and the use of peripheral bone mass measurements may have affected our ability to detect an association with OA. A recent review [31] reported that higher BMD as measured by DXA at the spine and hip was associated with OA at hip, spine or knee, but weaker associations were observed with hand OA. Measuring BMD at skeletal sites distant from the site evaluated for OA might provide a general indication of skeletal bone mass, whereas measurements closer to the sites evaluated for OA might provide an indication of local bone effects. If local bone factors are more important, this might partly explain why heel BMD was not consistently associated with OA at other sites. On the other hand, being a weight-bearing site, one would expect that heel bone mass would reflect the physical loads borne by the knees, but not necessarily loading of bones in the hand. If so, heel bone mass might be more strongly associated with knee OA than hand OA, but this was not seen. Likewise, hand BMD was not strongly associated with hand OA in our study. The composition of measured bone might also play a role. The heel and spine have high proportions of trabecular bone than other sites such as the hip and metacarpal. However, there is no convincing evidence that the associations with OA might be stronger (or weaker) for BMD measurements at sites with high proportions of trabecular bone.

Several environmental or cultural (lifestyle) characteristics may contribute to site-specific OA occurrence [18]. Knee bending and squatting were strongly associated with OA of the knee [32, 33]. However, these were retrospective case control studies, and as such provide weaker evidence of a causal relationship than prospective studies. Life-long knee bending and squatting behaviors, a component of the traditional Japanese lifestyle, may have had a long-term effect on occurrence of knee OA. Since Japanese elderly spent most of their lives using tatami mats and Eastern-style toilets (all of which are within inches of ground level and require squatting), these daily activities of the traditional Japanese lifestyle may have contributed to the development of knee OA even among those with low bone mass (osteoporosis). Prospective studies are required to confirm a causal association between knee bending and knee OA. Furthermore, Half of the women we studied were engaged in farming, and these women may have a high risk of joint injury and of repetitive hand use, which may be risk factors for OA [34, 35]. There was no association of higher bone mass with OA at the knee, DIP and PIP joints in the current study. This suggests that environmental factors may have contributed more to the occurrence of OA at knee, DIP and PIP joints in our Japanese sample, and this might have masked an association between bone mass and OA.

Heel bone stiffness index was lower in women with grade 4 at the knee than those with no OA in the current study. Although the explanation is unclear, one possibility is that women with severe knee OA may have less physical (walking) activity, which could lead to lower bone mass than women of comparable age.

The measurements of bone mass at 2 appendicular sites (calcaneus and metacarpal) may offer insights regarding possible differences in associations of cancellous and cortical bone with OA [36]. The calcaneus is of primarily cancellous bone, and the metacarpal of primarily cortical bone. Furthermore, the calcaneus is weight-bearing, and may reflect to some degree the usual level of physical activity, whereas the metacarpal is not a weight-bearing bone. Previous studies tended to find associations between OA (at the hip, knee and hand) and BMD measured at weight-bearing sites [5, 9, 10, 12, 23], but not when BMD was measured at the hand [23, 28–30]. However, there was no apparent relationship between the type of bone measured and the presence of OA in the present study.

Our study has several limitations. Although we attempted to obtain a representative sample of the population, the subjects had to be mobile enough to attend the examination center. Women with the most severe symptoms and disability (especially in the oldest age groups) may have chosen not to participate. The response rate in the Hizen-Oshima Study was approximately 30% of eligible women, so there is a possibility of selection bias. Our mail survey did not yield a high response rate but there was only one mailing to the general community and no attempted follow-up. The subjects in this analysis were all Japanese women; thus, our results may not be generalizable to men, or men and women of different racial groups. These analyses are cross-sectional by design, and causality cannot be inferred from the differences in adjusted bone mass between women with and without knee or hand OA. For example, women with higher bone mass at an earlier point in life may be more likely to develop OA, but subsequent declines in physical activity due to OA symptoms could contribute to accelerated bone loss which would confound the cross-sectional analyses reported here. Considering the exploratory nature of these analyses, and the use of multiple endpoints and predictors, it may be appropriate to adjust the p-values for multiple comparisons. Taking the dozens of comparisons into account, none of the associations would be considered statistically significant.

Our results showed that age and BMI adjusted appendicular bone mass is increased with radiographic changes of OA at CMC/MCP joint, especially with osteophyte status, among Japanese women. Our findings also failed to detect an association of bone mass with OA for knee, DIP or PIP joint in Japanese women, in contrast to other populations. Factors associated with OA may differ by site of joint involvement [18]. However, the cross-sectional nature of our study makes it difficult to make etiologic inferences, and prospective studies are needed to investigate this issue further.