Timing and Risk Factors for Clinical Fractures Among Postmenopausal Women

BMC Medicine Volume 4
An Open Access Research article
Published 9 October 2006

Abstract

Background
Many risk factors for fractures have been documented, including low bone-mineral density (BMD) and a history of fractures. However, little is known about the short-term absolute risk (AR) of fractures and the timing of clinical fractures. Therefore, we assessed the risk and timing of incident clinical fractures, expressed as 5-year AR, in postmenopausal women.

Methods
In total, 10 general practice centres participated in this population-based prospective study. Five years after a baseline assessment, which included clinical risk factor evaluation and BMD measurement, 759 postmenopausal women aged between 50 and 80 years, were re-examined, including undergoing an evaluation of clinical fractures after menopause. Risk factors for incident fractures at baseline that were significant in univariate analyses were included in a multivariate Cox survival regression analysis. The significant determinants were used to construct algorithms.

Results
In the total group, 12.5% (95% confidence interval (CI) 10.1–14.9) of the women experienced a new clinical fracture. A previous clinical fracture after menopause and a low BMD (T-score <-1.0) were retained as significant predictors with significant interaction. Women with a recent previous fracture (during the past 5 years) had an AR of 50.1% (95% CI 42.0–58.1) versus 21.2% (95% CI 20.7–21.6) if the previous fracture had occurred earlier. In women without a fracture history, the AR was 13.8% (95% CI 10.9–16.6) if BMD was low and 7.0% (95% CI 5.5–8.5) if BMD was normal.

Conclusion
In postmenopausal women, clinical fractures cluster in time. One in two women with a recent clinical fracture had a new clinical fracture within 5 years, regardless of BMD. The 5-year AR for a first clinical fracture was much lower and depended on BMD.

Background
According to the World Health Organization (WHO), osteoporosis is the second leading health problem in the developed world in terms of numbers [1,2], owing to the ageing of the western female population [1-9]. The incidence of fractures is related to the age of the population, and to skeletal and extra-skeletal factors [1-9]. The worldwide lifetime risk of osteoporotic fractures in women is 40%, and because of the ageing of the population, the overall prevalence of osteoporotic fractures is expected to rise considerably [10]. The mortality rate for hip fractures varies between 15 and 30% [10-12]. In 1990, there were 1.7 million hip fractures worldwide, and this could increase to 6.3 million by 2050 [10]. The resulting hospital costs alone amount to over US$3976 million in the European Union, US$500 million in Australia, US$5700 million in the USA, and US$9359 million in Japan [10]. In view of this high morbidity, mortality and economic burden, it is important to identify those groups of patients who are at high risk and for whom interventions to prevent osteoporotic fractures would be most effective [3,4,6-9].

In 1992–1994, a cross-sectional population-based study assessed BMD and other determinants that might influence osteoporosis among 4203 postmenopausal women aged 50 to 80 [7-9]. A random sample of women from the above study was re-examined 5–6 years later, allowing us to study the prognostic value of relevant determinants of experiencing a fracture.

Several researchers have addressed the same subject [13-22], and the WHO is currently developing a fracture risk assessment tool with which family physicians can predict 5-year and 10-year fracture risk [23]. However, data are scarce about the absolute risk (AR) of fractures and the timing of a previous clinical fracture [14-16]. Available studies suggest that fracture risk is highest immediately after a fracture, with the risk of an osteoporotic fracture being 12.0%[14] and that of any fracture being 12.2% during the first 2 years[16]. The risk of morphological vertebral fractures is 19.2% at 1 year after a previous morphological vertebral fracture[15]. The goal of our study was to assess the 5-year AR of incident clinical fractures among postmenopausal women.

Methods

Ethics
The ethics review committee of the Maastricht University and the Maastricht University Hospital approved the study (reference number MEC 94-196.1).

Participants
In total, 20 general practitioners (GPs) from 10 general practice centres participated. The participating general practices were invited based on their familiarity with the rules of 'good clinical practice' and based on their participation in earlier studies. The region can best be described as consisting of two cities surrounded by suburban villages [7].

A random sample of 1686 postmenopausal women, aged between 50 and 80 years, was drawn from the above study population [7-9]. Participating women signed an informed consent form.

Measurements
The assessments in 1992–1994 started with weight (in kilograms), height (in centimetres), and BMD (measured by a computer-guided DXA instrument; Hologic QDR-1000, Hologic Europe, Brussels, Belgium). Four experienced and specially trained research nurses performed all the measurements. After the measurements, the participants completed a questionnaire, which concentrated on variables possibly related to osteoporosis or low BMD [7-9]. In addition, all women were questioned about their medical history (including fracture history), family history and diet [7].

In the 1998–2000 assessments, one specially trained research nurse, or the principal investigator, completed a questionnaire in the presence of the participant, which recorded the history of fractures. The research assistant confirmed all fractures reported by the patients by searching the medical files in the participating GP centres. In the Netherlands, clinical fractures are treated in hospital and then reported to the patient's GP, who always includes this information in the patient's medical file. In our study, every fracture reported by a patient could thus be confirmed by her medical file.

Statistical analysis
SPSS software (version 11.0; SPSS Inc., Chicago, IL, USA) was used for the statistical analysis. The dependent variable was the first fracture suffered by a woman after baseline, taking only the first fracture after menopause into account. A power calculation for our main variable (timing of a previous fracture) was performed. This calculation compared two proportions [24].

The hazard of first fracture and the death hazard were used to compute the 5-year probability of fractures. Univariate Cox regression analysis was used to analyse possible predictors of suffering a fracture. Before being entered into the regression analysis, the non-discrete variables were dichotomised at the mean. The cutoff points for coffee intake per day and alcohol intake per week were set at five or more consumptions. The variables 'smoking' and 'sports', both past and present, were dichotomised as yes/no. The resulting significant variables were entered into a multivariate Cox regression to calculate the 5-year probability of fracture.

Further analysis investigated the influence of timing of a previous clinical fracture. A discrete variable was used, categorised as participants with a recent previous fracture (in the past 5 years; n = 68) and participants with an older previous fracture (longer than 5 years previously, but still after menopause; n = 62). The 5-year cutoff point was the optimum compromise between finding a sufficient number of fractures to allow statistical evaluation and minimising the loss to recall [7]. Thereafter, a multivariate subgroup analysis was performed, including the timing of previous fractures. These significant determinants were used to construct algorithms, which included relative risk (RR), AR and 95% confidence intervals (CI).

At baseline (1992–1994), BMD values for lumbar spine were divided into two categories: BMD T-score <-1.0 (including osteoporosis plus osteopenia) and T-score ≥ -1.0, according to WHO criteria [1].

Internal server error