Q J Med 2000; 93: 677-680
© 2000 Association of Physicians
Prevalence of hip fracture risk factors in women aged 70 years and over
1 From the Departments of Medicine and Therapeutics 3 Medicine for the Elderly, and 4 General Practice, University of Aberdeen, Aberdeen, and 2 Department of Health Studies and Centre for Health Economics, University of York, York, UK
Received 7 April 2000 and in revised form 24 July 2000
 |
Summary |
|---|
 Top Summary IntroductionMethodsResultsDiscussionReferences |
|---|
We estimated the prevalence of common risk factors for hip fracture and the numbers needed to treat (NNT) to prevent a hip fracture in various high-risk population groups, using a postal risk factor survey of women aged 70 years and above from General Practices in Grampian and Yorkshire. Recorded risk factors included prior fracture of any type; low body weight; smoking; and family history of fracture. The prevalence rates of hip fracture risk factors were 34%, 7% and 11% for previous fracture, maternal hip fracture and smoking, respectively for the Grampian practices (low body weight being defined as falling in the lowest quartile) and 34%, 7% and 7% for a single practice in the York area. Applying previously published estimates of risk, NNT analysis produced a value of about 300 for women with no risk factors, whilst for women with three risk factors it was between 32 and 71, depending on which risk factors were present and assuming intervention reduced fracture rates by 30% or 50%. Groups of women at high risk of hip fracture can easily be identified in primary care and offered treatment, with realistic prospects of hip fracture prevention.
|
Introduction |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Osteoporosis is a major cause of morbidity, mortality and cost among the older population. Approximately 10–20% of hip fracture patients will die within 6 months;1 furthermore, in the UK it has been estimated that the annual cost of osteoporosis is in excess of £940m.2 However, while there is an increasing number of therapeutic options for fracture prevention, these require targeting towards high-risk groups if they are to be cost-effective.3 To date, there are no readily available estimates of hip fracture risk in a UK primary care population.
To identify patients at high risk of hip fracture, we need to know which risk factors are important and the magnitude of the fracture risk. The largest prospective study of hip fracture risk factors has been the Study of Osteoporotic Fractures (SOF) in the US.4 This study identified 16 independent risk factors for hip fracture for White women aged 65 and over. However, whilst this study presented the relative risk ratios of hip fracture for women with a risk factor compared with those without a given risk factor, this information cannot be used to estimate absolute risk of fracture. If relative risk ratios are multiplied by the fracture incidence rate for a given age group, this will overestimate hip fracture risk, because population fracture risk is partly a function of the very risk factors which are being used to estimate absolute risk. Therefore the greater the prevalence of the risk factor, the lower will be the relative population risk of fracture. For example, guidelines from the National Osteoporosis Foundation (NOF) estimate that the relative risk of hip fracture for a woman with a family history of hip fracture is 1.40, compared with a similar woman without this risk factor.5 However, given a prevalence of the risk factor of 7%, the population relative risk falls slightly to 1.36. For a more prevalent risk factor such as low body weight (prevalence 25%), the relative risk falls from 1.90 for individual relative risk, to 1.55 for population relative risk.5
A later re-analysis of the SOF risk factor data5,6 reduced the 16 risk factors down to a more manageable four—low body weight, personal fracture history, family history of fracture and current smoking. Importantly, this re-analysis gave ‘population relative risk’ figures as well as the usual relative risk ratios of fracture risk for the risk factors. However, population relative risk estimates, unlike ordinary relative risk ratios, are dependent upon the prevalence of the risk factor. Should the risk factor prevalence from SOF be significantly different from that in the UK population, their risk estimates will need to be adjusted for UK populations. Unfortunately, there are no published data on the prevalence of the SOF risk factors in the UK. Therefore, we have undertaken two population surveys, in Grampian and Yorkshire, to assess hip fracture risk factor prevalence in women aged 70 years and over. Furthermore, we have calculated the number needed to treat (NNT) to prevent one hip fracture for each risk factor category.
|
Methods |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
We surveyed women aged 70 years and over registered with five General Practices in the Aberdeen area (n=2672) and one General Practice in York (n=856) using a simple risk factor postal questionnaire. No reminder letter was sent. Women were asked about the following risk factors: personal fracture history; mother's fracture history; smoking status; current body weight; and whether any sibling had had a hip fracture (York only).
We estimated the absolute risk of hip fracture by taking the population relative risk of fracture5 and multiplying it by the population risk of hip fracture using a recently published epidemiological study.6 We also estimated the NNT to prevent one hip fracture for each risk category for treatments which reduced hip fractures by 30% or 50%.
|
Results |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Aberdeen
The response rate for the Grampian practices was 64.5% (1724) with a further 154 (5.8%) questionnaires returned uncompleted due to inability to take part (for example, due to dementia, incorrect address, recent death or unwillingness to participate). The prevalence of the four main risk factors, compared with the SOF risk factor prevalence, are shown in Table 1
. There is a surprisingly similar risk factor prevalence between the American data and Grampian.
|
York
The response rate to the survey in York was lower than in Aberdeen, with only 46% responding (n=387 after exclusion of nine returned questionnaires due to either death or dementia). However, although the confidence intervals for the risk factor prevalence were wider, due to the smaller sample size, the estimates of risk factor prevalence were similar to the Aberdeen results, apart from smoking prevalence which was somewhat lower in Yorkshire compared with either Grampian or the US.
Number needed to treat (NNT)
Table 2 shows the various population proportions for given risk factor combinations with their absolute risk of hip fracture with the NNT7 per year for treatments with a 30% (e.g. calcium and vitamin D8) and 50% (e.g. alendronate9) efficacy. These estimates are based on a hip fracture incidence of 1.91%;7 however, excluding women aged under 75 will reduce the NNT by about 40% because of the increasing incidence of hip fracture in older women.
|
|
Discussion |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
There are an increasing number of treatments available for fracture prevention. To use the vast majority of these treatments in a cost-effective manner requires careful targeting of treatment to those patients who will derive most benefit (i.e. those at highest risk of fracture). In this paper we have described the prevalence of risk factors and the likely NNT for hip fracture prevention. Presenting the information in this way can help inform decision-making with respect to preventing hip fractures in primary care. For instance, if a practice were considering offering a treatment to all women with three or more risk factors, then this would result in approximately 50 women being offered treatment (out of an average practice population of 1000 women aged 70 years and over). Thus, the pharmaceutical cost of such a strategy could be calculated by multiplying this number by the annual cost of treatment, whilst the number of averted hip fractures can be estimated by multiplying the incidence estimates from Table 2
by estimates of effectiveness. This paper includes an NNT analysis allowing targeting of preventative therapy for hip fractures using clinical risk factors. The presence of three clinical risk factors gives NNT values of between 32 and 71, depending on the specific risk factors present and the efficacy of the proposed intervention. We have chosen to consider the likely efficacy of the relatively inexpensive intervention, calcium and vitamin D which reduces hip fracture rates by about 30%8 and an expensive intervention, which may reduce hip fracture rates by 50%.9 However, we acknowledge the limitations of using NNTs without clear experimental data. For instance, we have implicitly assumed that women who are at risk of a hip fracture because they have a family history of the disease respond equally well to therapy as do women who are at risk because they have low body weight. Such assumptions may not be valid. Similarly, we have assumed that treatments are about as effective in the ‘real world’ setting as they are in a trial setting. Further economic modelling and primary research will be required for the cost-effectiveness of such an approach to hip fracture prevention to be validated. However, until better data become available, our data should still be able to assist clinicians with decisions on whom to treat to prevent fracture.
Whilst our survey has produced similar prevalence estimates to those of the SOF study, there are some important limitations which require comment. First, we have had to assume that the risk factors found in older Caucasian American women can be applied to British women. Ideally, a large prospective cohort study needs to confirm the validity of these risk predictors in the UK; however, until this has been achieved, it does not seem unreasonable to extrapolate from the US data. For instance, a recent prospective Dutch study has produced similar relative risk estimates, for body weight, among women aged 70 years and above living in sheltered accommodation.8 Second, these figures can only be applied to White women. For men and for other ethnic groups, it is possible that other risk factors may be more important. Thirdly, although only four relatively common risk factors are included in our study, this should not preclude treating women who have other, possibly more powerful, but rarer risk factors, such as oral corticosteroid use. Finally, our response rates, being less than 100%, could introduce some selection bias. However, our surveyed populations were from the community, unlike the SOF data, which were taken from US voter registration lists. Furthermore, comparing York data, where there was a lower response rate, with the Aberdeen data, did not show any apparent bias due to lower response rates. Despite this, the data still needs to be treated with some caution as response rates were less than 100%.
In summary, we have described hip fracture risk factor prevalence within a UK primary care population. The estimates of risk and the proportion of the population at risk should assist cost-effective planning of prevention strategies.
|
Acknowledgments |
|---|
The Aberdeen survey was undertaken as part of a pilot screening programme funded by Research into Ageing whilst the York survey was funded by the University of York. AS is an Arthritis Research Campaign Postdoctoral Research Fellow.
|
Notes |
|---|
Address correspondence to Dr A. Stewart, Department of Medicine and Therapeutics, University of Aberdeen, Osteoporosis Research Unit, Woolmanhill Hospital, Aberdeen AB25 1LD. e-mail: a.stewart{at}abdn.ac.uk
|
References |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
1. Schurch MA, Rizzoli R, Mermillod B, Vasey H, Michel JP, Bonjour JP. A prospective study on the socio economic aspects of fracture of the proximal femur. J Bone Miner Res1996; 11:1935–42.[Web of Science][Medline]
2. Dolan P, Torgerson DJ. The cost of treating osteoporotic fractures in the United Kingdom female population. Osteoporosis Int1998; 8:611–17.[Web of Science][Medline]
3. Torgerson DJ, Kanis JA. Cost effectiveness of preventing hip fractures in the elderly population using vitamin D and calcium. Q J Med1995; 88:135–9.
4.
Cummings SR, Nevitt MC, Browner WS, Stone K, Fox KM, Ensrud KE, Cauley J, Black D, Vogt TM for the Study of Osteoporotic Fractures Research Group. Risk factors for hip fracture in white women. N Engl J Med1995; 332:767–73.
5. Osteoporosis: Review of the evidence for prevention, diagnosis and treatment and cost effectiveness analysis. Osteoporosis Int1998; 8(Suppl. 4):S1–88.
6. Fox KM, Cummings SR, Powell-Threets K, Stone K for the Study of Osteoporotic Fractures Research Group. Family history and risk of osteoporotic fracture. Osteoporosis Int1998; 8:557–62.[Web of Science][Medline]
7. Johansen A, Evans RJ, Stone MD, Richmond PW, Lo SV, Woodhouse KW. Fracture incidence in England and Wales: a study based on the population of Cardiff. Injury1997; 28:655–60.[Web of Science][Medline]
8. Chapuy MC, Arlott ME, Duboeuff F, Brun J, Crouzet B, Arnaud S, Delmas PD, Meunier PJ. Vitamin D3 and calcium to prevent hip fractures in elderly women. N Engl J Med1992; 327:1637–42.[Abstract]
9. Black DM, Cummings SR, Karpf DB, Cauley JA, Thomson ED. Randomised trial of the effect of alendronate on risk of fracture in women with existing vertebral fractures. Lancet1996; 348:1535–41.[Web of Science][Medline]
10. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical Epidemiology: A Basic Science for Clinical Medicine, 2nd edn. London, Little, Brown, 1991
11. Pluijm SMF, Graafmans WC, Bouter LM, Lips P. Ultrasound measurements for the prediction of osteoporotic fractures in elderly people. Osteoporosis Int1999; 9:550–6.[Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||