K. Michaelsson et al., EFFECT OF PREFRACTURE VERSUS POSTFRACTURE DIETARY ASSESSMENT ON HIP FRACTURE RISK ESTIMATES, International journal of epidemiology, 25(2), 1996, pp. 403-410
Background. Dietary factors are presumed to have influence on bone mas
s and hence fracture susceptibility. Most information in this respect
is based on retrospective assessment of previous dietary habits. In a
population-based case-control study nested within a cohort, we collect
ed dietary information both before and after a first hip fracture. Thu
s it was possible to study reported changes in dietary habits, intenti
onal as well as unintentional, among hip fracture patients after a fir
st hip fracture and to compare postfracture with prefracture dietary i
nformation. Methods. More than 65 000 women born 1914-1948 in two coun
ties in central Sweden completed a food frequency questionnaire regard
ing their usual current dietary habits, before attending a mammographi
c screening between the years 1987 and 1990. Subsequently 123 of them
sustained a first hip fracture and were defined as cases in the presen
t study. For every case, one control, individually matched by age and
county of residence, was selected from the cohort. A second identical
food frequency questionnaire was mailed to both cases and controls on
average 2 years after the hip fracture event. In total 98 case/control
pairs could be included in the analysis. The association between diet
and hip fracture was evaluated and the results from the two dietary a
ssessments were contrasted. Women who themselves claimed that they had
not changed their diet in recent years were analysed separately. Resu
lts. The hip fracture cases, compared with the controls, had reduced t
heir reported dietary intake of dairy products after the fracture. App
arently this was not intentional since this effect was more pronounced
among those cases who claimed that their diet was unchanged. The chan
ges were most apparent among the younger cases with a more recent hip
fracture and with a body mass index above the median. Half of the case
s, more than twice the frequency in controls, who were initially class
ified as having high intake of dairy products were classified as havin
g low intake (<800 mg calcium/day) after the hip fracture. This also l
owered, in fact reversed, the relative risk estimates of hip fracture
both for intake of dairy products and calcium. Crude odds ratios of hi
ghest quartile of intake versus lowest, changed from 3.0 to 0.6 for da
iry products and from 2.6 to 0.9 for calcium. No other foods or nutrie
nts displayed such notable differences between the two surveys. Conclu
sion. We conclude that the use of current and retrospective dietary in
formation after a hip fracture can lead to a differential misclassific
ation in dietary studies and to biased estimates of hip fracture risk
as compared with prospectively collected dietary information.