Use of T-score to visualize the absolute differences in BMD makes them more applicable to the clinical situation, as a decrease of 1 SD at the distal radius by SPA has been predicted to give a 50–60% increase in the risk of hip fracture (Cummings et al. 1990
, Stone et al. 2003
). The absolute differences in T-score between the cohorts measured in 1988/89 and 1998/99 were small and were not statistically significant in both the rural cohort and the urban cohort. Thus, if one only takes changes in BMD into account, there ought to have been a relatively unchanged incidence of hip fracture during the study period.
The proportion of hip fractures attributable to osteoporosis has been calculated to be only 20–30% (Stone et al. 2003
), and a recent report has even proposed that the burden of low trauma fractures is not related to the prevalence of osteoporosis but to a more comprehensive assessment of fracture risk (Langsetmo et al. 2008
). These observations highlight the importance of evaluating other risk factors apart from BMD. Low weight and low BMI are well-known risk factors (Cummings et al. 1995
, De Laet et al. 2005
), but we did not find any changes in these factors over time. Another important risk factor is the risk of falling. Recent publications have actually suggested that we should shift our attention in fracture prevention from osteoporosis to reduction of the number of falls (Jarvinen et al. 2008
)—or at least to consider both (Langsetmo et al. 2008
). In our secondary analysis, we did not find any differences in reported fall incidence even though both balance and gait velocity were better at the second measurement in 1998/99, indicating a possible influence on fall risk.
As we anticipated, in 1998/99 there was an increased proportion of women who had ever used estrogen therapy and/or oral contraceptives compared to 1988/89. Interestingly, balance was better and gait velocity faster in both urban and rural women at the second measurement—as was the proportion of women consuming alcohol, all 3 of which could possibly have affected the number of fractures, but in different directions. In addition, rural women generally had lower grip strength in 1998/99 than in 1988/89. We did not evaluate changes in other risk factors for hip fracture such as bone structure, vision, prevalence of hip arthroplasties, and prophylactic fracture treatment (Cummings et al. 1995
, Kannus et al. 1996
), and we only evaluated some medications and co-morbidities.
The strengths of this study are the use of a population-based cohort for both BMD and fracture evaluation and the fact that changes in BMD and hip fracture incidence over time were estimated in the same populations. The BMD measurements were all done using the same bone scanner operated by the same technician throughout the study. The prospective evaluation of scanner drift is also another strength of the study. Furthermore, incidence of hip fracture was not presented in 5- or 10-year age classes, a study design that could be biased by changes in demographics within the age classes. Instead, we used an age-adjustment by direct standardization, thus compensating for changes within the age classes. Fracture data were collected from a complete national database in which every inpatient record is linked back to the home municipality, thus enabling us to find virtually all patients with a hip fracture, even if the fracture was sustained and treated away from the home municipality. Also, by using both a diagnosis code and an operation code for the inclusion, we improved our ability to catch only acute hip fractures within the cohorts by excluding patients with complications that were secondary to an earlier hip fracture.
One weaknesses of the study was the BMD evaluation in the forearm by the SPA technique, a clinical scanning technique that is no longer used today. Fracture prediction from BMD has been investigated more extensively by dual-energy X-ray absorptiometry (DXA), where a decrease of 1 SD in the hip has been calculated to give a 3-fold increase in the risk of hip fracture (Marshall et al. 1996
) compared to SPA. With SPA, a 1-SD reduction in forearm BMD has been calculated to give a 50–60% increase in hip fracture risk (Cummings et al. 1990
, Stone et al. 2003
). A high correlation has been shown between BMD values measured with SPA in the distal radius and those measured with DXA in the hip (Karlsson et al. 1993
), and both have been shown to be predictors of hip fracture (Cummings et al. 1993
, Marshall et al. 1996
, Stone et al. 2003
). Ideally, the second measurement should have been done after a separate randomization, but no differences in age-adjusted BMD were evident between those attending both measurements and those attending only the initial measurement. Mean age in the 4 cohorts was similar, and therefore no age adjustment for BMD was done. Post-hoc testing with age adjustment did not alter the results (data not shown).
It would also have been preferable to collect data for hip fracture incidence and for BMD/prevalence of osteoporosis/BMI and the secondary investigation over exactly the same years and in the same cohorts. However, the longer fracture ascertainment period and age span were chosen in order to increase our ability to detect even small changes in the incidence of hip fracture. Ideally, the rural area for fracture evaluation should have included only the municipality of Sjöbo were the BMD data were collected, but no differences in hip fracture incidence or hip fracture progression were evident between Sjöbo and the 8 surrounding municipalities that were included. We acknowledge that in spite of the fact that we increased the sample size by including 8 regions around Sjöbo when evaluating the rural hip fracture incidence, the discrepancy in changes in hip fracture incidence over time compared to a variety of other reports in western societies (Melton et al. 1996
, Rogmark et al. 1999
, Lofman et al. 2002
, Jaglal et al. 2005
, Kannus et al. 2006
, Nymark et al. 2006
, Chevalley et al. 2007
, Icks et al. 2008
) may have been due to a type II error.
To summarize, we did not find any changes in age-adjusted hip fracture incidence from 1987 to 2002 and we found no secular differences in BMD or in the prevalence of osteoporosis between 1988/89 and 1998/99. This indicates that changes in other risk factors for hip fracture such as balance and gait velocity, as found in our secondary investigation, are either of minor importance or are counteracted by changes in other risk factors.