Women and men in Oslo, the capital of Norway have the highest incidence rates of hip and forearm fractures and among the highest prevalence of vertebral fractures ever reported internationally [1–4]. In accordance with this, an European multi-centre study found that bone mineral density (BMD) in men and women in Oslo were among the lowest in Europe [5].

Within this high risk city of osteoporosis and osteoporotic fractures regional differences in hip fracture incidence have been reported [6]. The highest fracture rates are found in Eastern parts of the city, which is less affluent than the Western part of the city, with RR ranging from 1.23 to 1.67. Equivalent regional differences have been found regarding mortality and other health indicators [7]. These health inequalities are connected to differences between the regions regarding standard of living and socio-economic status of the habitants, with the highest annual income and highest level of education in West [8, 9].

Socioeconomic status is well established as a determinant for several health aspects and disease outcomes. Regarding fracture risk, a study including white Americans 50 years and older, reported a linear decrease in hip fracture rates with increasing income after controlling for age and sex [10]. Likewise, an ecological study from Wales found higher fracture incidence among socio-economically deprived adults [11]. This effect diminished with age and was not observed in older age groups (above 75 years). In a Norwegian population based case-control study including men and women, low educational level was associated with increased risk of hip fracture [12]. This was not confirmed in a large Swedish case-control study among women [13]. They found however that being unemployed, being unmarried and living in an apartment (vs. one-family house) were associated with increased risk of hip fracture.

Studies addressing the association between BMD and socioeconomic status are likewise heterogeneous and not conclusive. Elliot et al found highest BMD in men from lower socio-economic groups [14], but several other studies have found an association of low BMD with low-income, low education and living in socially deprived areas [15–21]. These studies had limitations which could make it difficult to conclude and generalize – i.e. they included mainly postmenopausal women, they studied partly small samples and the participation was based on self-selection or referral from general practitioners/consultants.

Using data from a population-based health survey, we wanted to examine whether BMD was lower in the socioeconomic less affluent Oslo East compared to the more privileged Oslo West. Further, we wanted to assess whether the association could be attributed to socio-demographic factors, level of education or lifestyle factors.

We found lower BMD in women in the socioeconomic less affluent Oslo East compared to the more privileged Oslo West. In the age specific analysis the difference was only significant in women aged 60 years. In men the difference was less pronounced and not statistically significant. The difference in BMD in women between the socioeconomic regions persisted after controlling for known risk factors. Individual differences in socio-demographic and lifestyle factors did not explain the regional differences we found.

Socioeconomic region assign all persons living in the same geographical area to the same socioeconomic position independent of individual socioeconomic characteristics. Since the effect of socioeconomic region on BMD in women sustain after adjusting for years of education and other covariates, there are probably other factors associated or influenced by socioeconomic region affecting BMD than those we have included in our model. Nutritional and neighbourhood factors are not controlled for, and common environmental factors (e.g. lead exposure) and water supply source are unexplored possible explanations.

From an individual perspective the socioeconomic differences we found in BMD do not seem substantial – at least not compared to other health outcomes and risk factors such as cardiovascular disease and smoking. However, small differences in BMD between regions may have large impact on fracture incidence in the population. A difference in BMD of e.g. 0.014 g/cm² in women, which is the difference in age-adjusted mean BMD between East and West, amount to about 1/4 SD. In a meta-analysis Marshall et. al showed that the relative risk for hip fracture in women was 1.8 per SD decrease in distal forearm BMD [29]. Roughly estimated a decrease of 1/4 SD in BMD may increase the risk of hip fracture by about 20%. In women it seems that the difference in BMD in our study concurs with the risk estimates from the hip fracture study in Oslo – 23% increased risk of hip fracture in women living in inner East compared with inner West [6]. In men the corresponding fracture difference was 67%, which might suggest that other factors than BMD is of importance in explaining fracture risk differences between socioeconomic regions. However, these comparisons should be interpreted with caution since we cannot fully compare the regions in our study with those in the hip fracture study in Oslo.

Few published studies have focused on socioeconomic differences in BMD and some of them have obvious limitations as for example over-representation of well-educated individuals [19, 20] or comparison of groups of unequal ethnic composition [15]. An American study reported no association between BMD in the forearm and education in women above 65 years [30]. In another American study they found a positive education-BMD association in black, but not in white and Hispanic postmenopausal women. However, they found an income-BMD association in white, but after adjustments for behavioural factors, associations with education and income were eliminated [21]. Our results corresponds, however, to other studies which reported a relation between educational level and BMD in postmenopausal women [16, 17, 20], although a Chinese study reported education-BMD association in weight bearing sites only and not in the arms [17].

Socioeconomic differences in BMD in men are particularly scarcely explored, possibly reflecting their lower incidence of osteoporotic fractures. Elliot et al found highest BMD in men from lower socio-economic groups [14]. In that study a larger percentage of those in the lower socioeconomic groups were employed in manual labour, but that did not explain the differences they found. A negative association between education and hip BMD in old men was also found in an American study, but the effect of education was eliminated after adjustment for weight [31].

It should be noted that the socioeconomic distribution of risk factors for BMD might differ between different populations. In the Italian study smoking were more common in those with high educational level [20] and in the Turkish study those in the lower socioeconomic group had lower alcohol intake and smoked less than the others [18]. The distribution of risk factors between socioeconomic groups may contribute to blur the effect of socioeconomic position on BMD. A risk factor for many health conditions such as high body mass index, which is more prevalent in lower socioeconomic groups, is a protective factor for BMD. In our study, neither body mass index nor any of the other confounding factors added in the full multivariate model could explain the difference in BMD between the socioeconomic regions (table 4). Neither could educational level, which was one of the criteria for constructing the two regions. Further research is needed to investigate other factors influencing people's bone health in the different socioeconomic regions in Oslo.

The strength of this study is the population-based setting and the bone measurements in both men and women in the age-range 40–75 years. In addition we had information on the most important socio-demographic and lifestyle factors that is known to be associated with BMD. Further, the grouping of socioeconomic region was based on an index of social- and health-related indicators, which has previously been used to show large differences in mortality between Oslo West and Oslo East [7].

On the other hand, although the overall population size was large enough to detect differences in BMD of about 0.015 g/cm² with power 80% and 5% significance level, we did not have enough power to detect small differences in the age groups with the fewest participants (i.e. men 75 years). Furthermore, the attendance rate was generally low in both the main survey and the BMD sub-study and selection bias cannot be excluded. Attendance was positively associated with age, income, education, living in Outer West and Outer East. However, the impact of self-selection in the Oslo Health Study has been evaluated and the prevalence estimates of factors associated with BMD such as body mass index, smoking and self perceived health was found to be quite robust [22]. The associations between disability pension (as a proxy of unhealthy persons) and several background variables, including level of education and region in Oslo, were compared in attendees and non-attendees and were found to be unbiased in Norwegian born individuals [22].

Another limitation of our study was that BMD was measured at the forearm. The NORA study has, however, shown that peripheral measurements, including the forearm, have a strong relationship to later osteoporotic fracture [32]. Forearm SXA is also one of the most precise bone densitometric methods [33].

In conclusion, we found socioeconomic differences in BMD in women, but not in men. The differences between socioeconomic regions in women were not explained by education, lifestyle or the other recorded covariates.