Archives of Osteoporosis

, Volume 6, Issue 1, pp 179–188

Epidemiological burden of postmenopausal osteoporosis in the UK from 2010 to 2021: estimations from a disease model

Authors

    • Amaris
    • University of Glasgow
  • John A. Kanis
    • University of Sheffield
  • Yling Jiang
    • Amaris
  • Monique Martin
    • i3 Innovus
  • Juliet E. Compston
    • University of Cambridge School of Clinical Medicine
  • Fredrik Borgström
    • LIME/MMC, Karolinska Institutet
  • Cyrus Cooper
    • University of Southampton
    • University of Oxford
  • Eugene V. McCloskey
    • University of Sheffield
Original Article

DOI: 10.1007/s11657-011-0063-y

Cite this article as:
Gauthier, A., Kanis, J.A., Jiang, Y. et al. Arch Osteoporos (2011) 6: 179. doi:10.1007/s11657-011-0063-y

Abstract

Summary

This article describes the adaptation of a model estimating the burden of postmenopausal osteoporosis (PMO) to the UK.

Purpose

The purpose of this study was to estimate the present and future epidemiology of PMO in the UK.

Methods

For each year of the study, the ‘incident cohort’ (women experiencing a first osteoporotic fracture) was identified and run through a Markov model using 1-year cycles until 2020. Health states were based on the number of fractures and death. Fracture by site was tracked for each health state. Transition probabilities reflected fracture site-specific risk of death and subsequent fractures.

Results

Assuming that the rate of incident fractures by age is constant over time, the model estimated the total number of PMO fractures at 224,219 in 2010, including 51,927 hip and 30,994 clinical vertebral fractures. These estimates were predicted to increase by 17.2%, 16.6% and 17.9%, respectively, by 2020. The number of postmenopausal women living with osteoporosis was predicted to increase from 1.8 million in 2010 to 2.1 million in 2020 (+16.5%). A sensitivity analysis demonstrated that the estimated number of fractures is most sensitive to the assumption made on the trends in the rate of incidence.

Conclusion

The PMO disease model, first developed for Sweden, was adapted to the UK. Due to demographic changes, the burden of osteoporosis is expected to increase by almost a fifth by 2020. Due to the lack of country-specific data, these results rely on several assumptions regarding the incidence of non-hip fractures, trends in BMD and rate of incidence over time.

Keywords

Bone mineral densityEpidemiologyFractureOsteoporosisT scoreUK

Introduction

The internationally agreed description of osteoporosis is ‘a systemic skeletal disease characterised by low bone density and microarchitectural deterioration of bone tissue, leading to increased risk of fragility fractures’ [1]. The diagnosis of osteoporosis is made when the bone mineral density (BMD) is lower than 2.5 standard deviations below the average BMD in healthy young Caucasian women (a T score of –2.5 SD or less) [2]. More recently, the T score derived from the measurement of BMD at the femoral neck by dual-energy X-ray absorptiometry (DXA) has become the WHO international reference standard for the diagnosis of osteoporosis [3].

BMD decreases with age in women, with a sharp decline at menopause, and it is estimated that one in two women aged 50 years from the UK will sustain a fracture during her remaining lifetime [4]. In an environment in which the population is ageing, the burden of osteoporosis is expected to increase in the coming years [5, 6].

The morbidity of osteoporosis lies in fragility fractures, which are associated with pain, poor functional outcomes and increased mortality [2]. Among 1,000 patients studied prospectively in two UK centres, only 43.5% of patients living at home at the time of the proximal femoral fracture were able to return home after the injury. One year after hip fracture, 57% of patients who had been able to walk unaided prior to the injury required walking sticks or a walking frame [7]. Similarly, the mortality observed in women in the age range 50–55 years who sustain a hip or a clinical vertebral fracture necessitating hospital admission is up to tenfold that of the general population within the first year of fracture [8, 9]. As a result, fractures are associated with significant costs, and in the UK, the direct medical cost of osteoporotic fractures is predicted to exceed £2.1 billion per year in 2020 [10].

Despite the increasing awareness of the consequences of osteoporosis [2], the burden of the disease is not well documented. In the UK, the most recent estimate dates back to 2001, when Burge et al. [10] presented the results of a Markov model simulating the progression of osteoporosis. The authors estimated the number of osteoporotic fractures at 190,000 in 2000 and predicted a 21% increase by 2020, to reach a total of 230,000 fractures. This study was based on fracture incidence collected in 1994 from a single A&E department from Cardiff [11].

We have recently developed a burden of disease model that aims to estimate the burden of osteoporosis at a national level [12]. This model has been validated using Swedish data and is readily adaptable to other countries. In this paper, we have adapted this model to a UK setting. The aim of the study was to estimate the current burden of osteoporosis in the UK, as judged by the incidence and prevalence of osteoporotic fractures, prevalence of osteoporosis and osteopenia, and the number of deaths attributable to osteoporosis. We additionally estimated the likely burden over the next 10 years.

Methods

Modelling approach

The details of the modelling approach have been described in a previous publication [12]. In summary, the model consisted of tracking the progression of women experiencing a first fragility fracture. For each calendar year of the study period and for each year of age between 50 and 100, the number of women sustaining a first fragility fracture was estimated. Each of these ‘incident cohorts’ was then run through a model with four different health states: first fracture, second fracture, third fracture or death. The cycle length was 1 year. However, as most deaths due to fragility fractures occur within 3–6 months of fracture [8, 9], the model assumed that deaths due to fracture occurred 6 months after the incident fracture. Over subsequent years, excess mortality due to a high prevalence of comorbidities in the fracture population was accounted for by applying the relative risks of deaths observed at 1 year in a fracture cohort compared with the general population [8, 9].

The model structure is depicted in Fig. 1. The incident cohort entered the model in the ‘first fracture’ health state. Following fracture, women may die due to the fracture, in which case death occurs within the same year. If they survive 1 year, they may stay in the same health state, sustain a second fracture, or die.
https://static-content.springer.com/image/art%3A10.1007%2Fs11657-011-0063-y/MediaObjects/11657_2011_63_Fig1_HTML.gif
Fig. 1

Model structure

Although the different health states do not record the fracture sites, the split by site (hip, vertebral and non-hip non-vertebral) is tracked in the model and the model parameters (risk of a first osteoporotic fracture, risk of subsequent fractures and excess mortality) are specific to each site.

Due to the challenges of identifying whether a fracture is directly attributable to osteoporosis [2, 13], the model identifies fracture sites as osteoporotic when the corresponding incidence increases with age and with decreasing BMD. The definition used by Kanis et al. [14] included the following sites: hip, vertebral, rib, pelvis, humeral shaft, proximal humerus, clavicle, scapula, sternum, other femoral fractures, tibia and fibula (excluding ankle), and distal forearm.

One of the objectives of the model was to obtain information on the prevalence of osteoporosis and osteopenia, and the number of women below a specific threshold by age and history of fracture. As the risk of fracture increases with decreasing BMD, women having experienced a fracture tend to have lower BMD. Based on a meta-analysis, Kanis et al. [15] reported that the mean difference in BMD between women who have experienced a fracture versus those who have not was 0.11 standard deviation. In order to estimate the distribution of BMD for each year of the study period, the following approach was used. For a given T score measured at the femoral neck (calculated using the Third National Health and Nutrition Examination Survey (NHANES III) ‘non-Hispanic white’ women aged 20–29 years as a reference group [16]), the corresponding Z score was estimated. For each calendar year and age, Z scores were assumed to be normally distributed with a mean of zero and a standard deviation of 1 (by definition of the Z score). The mean difference in BMD between women with and without a history of fracture (0.11 SD) was applied to estimate the mean Z score in women with and without a history of fracture. In order to derive the proportion of women below a specific T score threshold with or without a prior fracture, it was assumed that the variance in BMD was homogeneous between women with and without a history of fracture.

Based on the publications by Kanis et al. [8, 9], excess mortality observed within 6 months of a hip or vertebral fracture compared with the general population was assumed to be caused by a combination of excess mortality due to the fracture (causally related death) and excess mortality due to a higher prevalence of comorbidities in patients sustaining a fracture. One year after the occurrence of fracture, the risk of death is still higher than in the general population and does not decrease further with time. This increased mortality was identified as due to comorbidities and was applied over the years following fracture, over the cohort’s lifetime. The number of deaths attributable to fracture was calculated using the incremental excess mortality observed at 6 months compared with 1 year after the occurrence of fracture to ensure that the mortality due to comorbidities was not included in the calculations.

The core model was developed and validated using Swedish data. The validation consisted of comparing the model predictions with observed epidemiological data. The model is based on the run of incident cohorts. Therefore, subsequent fractures of women having experienced a first fracture before the first year of the study period are not captured. As a consequence, the model underestimates the total number of fractures at the beginning of the study period; it needs to be run over a period of 20 years before achieving a steady state and providing reliable estimates. Therefore, the model was run from 1970 in order to provide reliable results from 1990.

The model inputs include: the number of women and annual probability of death by calendar year and single year of age; the incidence of fracture by site (hip, vertebral and non-hip non-vertebral); and by 5-year age group. It also requires the relative risk of death in women sustaining a fracture within the year of fracture (assumed to reflect the excess mortality due to comorbidities and due to fracture) and over the subsequent years (assumed to reflect the excess mortality due to comorbidities). Lastly, the relative risk of subsequent fractures by site as a function of the site of prior fracture, the mean and standard deviation of BMD measured at the femoral neck by 5-year age groups, and the mean BMD difference between women who have sustained a fracture and those who have not are used as model inputs.

The model outputs include: the overall incidence of fractures and incidence by osteoporotic status by site (hip, vertebral, non-hip non-vertebral); the prevalence of fractures, osteopenia and osteoporosis; and the number of deaths due to osteoporosis. All outputs are reported by calendar year and by single year of age.

Data source

Based on the opinion of clinical experts and on the absence of evidence of a country’s specific effect, some parameters from the core model remain unchanged. This was the case for the relative risk of subsequent fractures, which was taken from a meta-analysis of clinical trials conducted by Klotzbuecher et al. [17] (see Electronic supplementary material (ESM)). The mean BMD difference between women who have sustained a fracture and those who have not was obtained from a published meta-analysis [15, 18]. The relative risk of death following a fracture (see ESM), obtained from Kanis et al. [8, 9], was also kept unchanged. Indeed, the country effect was assumed to be captured by the annual probability of death in the general population, to which the relative risks are applied to derive the absolute risk of death in women having sustained a fracture. Lastly, due to a lack of data, the proportion of osteoporotic fractures which are first fractures was obtained from a Swedish observational study [19] (see ESM).

The total number of women aged ≥50 years, by single year of age and calendar year, was obtained from the Office for National Statistics (ONS), for the years 1982–2006 [20]. For years beyond 2006, the official projections from the Government Actuary’s Department (GAD) were used [21]. For 1970–1981, the estimates from 1982 were used as this period is only used to initialize the model.

Similarly, the annual probability of death was obtained from the ONS for years 1990–2004 [20] and the official projections from the GAD [21] were used from 2005. For years prior to 1990, death rates reported in 1990 were used as these data were only used to initialize the model.

The incidence of hip fracture was obtained from a study reported by Singer et al. [22] who conducted a prospective study of the incidence of hip fractures in 1992–1993 in the region of Edinburgh and its surrounding district. The incidence rates as a function of age were smoothed using an exponential model.

As reliable UK data on the incidence of symptomatic vertebral fractures were not available [23], the incidence of a clinical vertebral fracture was calculated by assuming that the ratio of clinical vertebral fracture to hip fracture would be similar in the UK and in Sweden by age group. Similar assumptions have been used in the health economic assessment of osteoporosis [2426]. The same approach was used to assess the risk of other osteoporotic fractures. This approach assumes that the age-specific distribution of the incidence of fracture is similar across countries and, from limited available data, appears to hold true [15].

The derived incidence of osteoporotic fractures was smoothed using an exponential function, whereas a linear model was applied to the incidence of vertebral fractures.

The available data on the incidence of fracture over time suggests that the increasing age-specific incidence of hip fracture seen in the 1970s [27] has reached a plateau [28] and was assumed, therefore, to remain constant over the study period in the base case analysis.

Few publications provide a quantitative estimate in the change of age-adjusted incidence of fracture over time. Some studies conducted in developed countries have reported an increase in the age-adjusted incidence of hip fractures until the mid-1980s to mid-1990s, followed by a levelling off [29] or a downward trend [3032]. In order to assess the potential effect of a change in the rate of incidence on the total number of fractures at the national level, a sensitivity analysis was undertaken using the incidence rates from Singer et al. for the years 1992 and 1993 and assuming an annual rate of increase of 1% in the rate of incidence by age between 1970 and 1992 and an annual rate of decrease of 1% over the 1993–2020 period.

The mean and standard deviation of BMD by age group was obtained from a study conducted by Holt et al. [33] (personal communication with A Gauthier) that reported BMD measured at the femoral neck from 5,173 females randomly recruited from different health registers in the UK. The corresponding mean and standard deviation BMD by age are available from the ESM. Although a more recent publication by Noon et al. [34] was identified, the study conducted by Holt et al. was selected for the following reasons: Firstly, their data covered a wider geographical area; secondly, the study conducted by Noon et al. was based on three different samples, which may not reflect the characteristics of the general population (the first sample consisted of women referred for a DXA by their GP, young hospital personnel and women volunteering for clinical research; the second included women with a twofold risk of breast cancer; and the last involved twin volunteers).

Results

In 2010, there were 11.494 million women aged ≥50 years in the UK (Table 1). Half of these women were younger than 65 years, slightly more than one third (34.2%) were aged between 65 and 79 years, and 15.8% of them were aged ≥80 years. The total number of osteoporotic fractures was estimated at 224,219 in postmenopausal women. It was estimated that 45,556 osteoporotic fractures (20.3%) occurred in women aged 50–64 years, 83,905 (37.4%) occurred in women aged 65–79 years, and 94,757 (42.3%) fractures were sustained by women older than 80 years. Hip fractures accounted for 23.2% of fractures (51,927 hip fractures), and vertebral fractures accounted for 13.8% of fractures (30,994 vertebral fractures).
Table 1

Burden of fractures and osteoporosis in women from the UK by age in 2010

 

5054

5559

6064

6569

7074

7579

8084

≥85

≥50

Population (in thousands)

2,012

1,813

1,920

1,514

1,311

1,108

884

933

11,494

Clinical vertebral fractures

175

1,611

2,917

3,788

4,690

5,081

5,271

7,459

30,994

Hip fractures

523

1,042

1,976

2,947

4,677

7,478

10,981

22,302

51,927

Other osteoporotic fractures

10,508

11,234

15,570

16,028

18,456

20,758

22,492

26,252

141,298

All osteoporotic fractures

11,207

13,887

20,463

22,764

27,824

33,317

38,744

56,013

224,219

Deaths attributable to fracture

37

70

153

244

446

910

1,786

5,282

8,928

Women with osteoporosis

99,522

154,032

243,061

242,044

269,413

276,822

239,640

252,943

1,777,476

T score ≤ −3 SD

31,348

52,956

90,020

94,456

124,297

158,553

149,433

157,730

858,794

T score ≤ −3.5 SD

7,856

14,345

26,081

28,745

47,446

80,472

84,611

89,309

378,865

T score ≤ −4 SD

1,557

3,037

5,855

6,748

14,988

36,079

43,303

45,708

157,274

Women with prior fragility fractures

32,259

86,593

167,078

202,551

247,712

269,056

258,448

280,302

1,543,998

Women with osteoporosis or prior fragility fractures

129,658

231,519

385,121

407,099

459,480

471,240

421,294

450,100

2,955,512

The number of deaths directly attributable to fracture was estimated at 8,928, and the vast majority (7,068, 79.2%) occurred in women aged ≥80 years. The rate of death by fracture (estimated as the number or deaths attributable to fracture divided by the total number of fractures occurring in the age group) was estimated at 0.57%, 1.91% and 7.46% in women aged 50–64, 65–79 and ≥80 years, respectively.

Demography

In the UK, the population of women aged ≥50 years is expected to grow by 15.6% over the 10-year period from 2010 to 2020, the most important growth being expected in women aged 65–79 years (+20.1%) due to the ageing of the post–World War II baby boom generation (see Table 2).
Table 2

Demographics, fracture incidence and mortality in the UK: estimates from 2010 to 2020

 

2010

2015

2020

Increment 20102020 (%)

Women aged ≥50 years (in thousands)

11,494

12,377

13,285

15.6

 Aged 5064 years

5,744

6,014

6,451

12.3

 Aged 6579 years

3,933

4,444

4,723

20.1

 Aged ≥80 years

1,817

1,919

2,110

16.1

All osteoporotic fractures

224,219

240,712

262,847

17.2

 Aged 5064 years

45,556

46,575

50,366

10.6

 Aged 6579 years

83,905

93,575

101,532

21.0

 Aged ≥80 years

94,757

100,562

110,949

17.1

Clinical vertebral fractures

30,994

33,315

36,556

17.9

 Aged 5064 years

4,704

4,624

5,101

8.4

 Aged 6579 years

13,560

15,148

16,443

21.3

 Aged ≥80 years

12,730

13,543

15,012

17.9

Hip fractures

51,927

55,388

60,540

16.6

 Aged 5064 years

3,541

3,538

3,856

8.9

 Aged 6579 years

15,103

16,623

18,199

20.5

 Aged ≥80 years

33,283

35,227

38,486

15.6

Deaths

8,928

8,833

9,275

3.9

 Aged 5064 years

260

238

250

3.9

 Aged 6579 years

1,600

1,525

1,590

0.6

 Aged ≥80 years

7,068

7,070

7,435

5.2

The rate of increase is projected to be similar between 2010 and 2015 (+7.7%) and 2015 and 2020 (+7.3%) over the whole population of women aged ≥50 years. However, the population of women aged 65–79 years is expected to grow markedly until 2015 (+13.0%), followed by a levelling off over the second 5-year period (+6.3%), whilst the opposite situation is expected for women aged 50–64 years (+4.7% and +7.3%, respectively) and ≥80 years (+5.6% and +10%, respectively).

Number of fractures

The incidence of all osteoporotic fractures is expected to rise by 17.2% between 2010 and 2020. In line with the demographic changes, the greatest increase is expected to occur in women aged 65–79 years (+21.0%). In women aged 50–64 years, the number of fractures is expected to increase from 45,556 in 2010 to 50,366 in 2020 (+10.6%), and in women aged ≥80 years, osteoporotic fractures are expected to increase by 17.1%.

A comparable increase is projected for the different fracture sites, with a slightly greater rise in the number of vertebral fractures: +17.9% compared with +16.6% for hip fractures and +17.3% for non-hip non-vertebral fractures.

The number of fractures is predicted to further increase over time as the annual rate of increase is estimated at 1.2–1.6% over the first 5-year period followed by an annual increase of 1.6–1.9% between 2015 and 2020 (Fig. 2). This annual rate of change is predicted to be more marked for vertebral fractures, especially from 2017 where the rate is expected to reach 1.9–2.0% until 2020.
https://static-content.springer.com/image/art%3A10.1007%2Fs11657-011-0063-y/MediaObjects/11657_2011_63_Fig2_HTML.gif
Fig. 2

Year-on-year increase in fracture incidence

Although the number of fractures associated with excess mortality (hip and vertebral) is expected to increase by 17.1%, the improvement in life expectancy (associated with a lower annual probability of death in the general population) is expected to limit the rise in the number of deaths directly attributable to fractures (+3.9%) in comparison with the increase in the number of fractures.

Prevalence

The number of postmenopausal women living with osteoporosis, based on the definition of a BMD at least 2.5 standard deviations lower than a young healthy women (T score ≤ −2.5 SD), is predicted to increase from 1.8 million in 2010 to 2.1 million in 2020 (+16.5%; Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs11657-011-0063-y/MediaObjects/11657_2011_63_Fig3_HTML.gif
Fig. 3

Number of women with BMD below a threshold

As a result, the prevalence of osteoporosis in the general population of women aged ≥50 years is assumed to remain stable over time, at approximately 15.5%.

The rates of increase predicted for lower thresholds (T score less than −3, –3.5 and −4 SD) are expected to reach approximately 13% between 2010 and 2020 (16.8%, 17.0% and 16.9%, respectively; see Table 3).
Table 3

Prevalence of fracture by number of previous fractures

 

2010

2015

2020

Increment 20102020 (%)

Number of previous fractures

≥1 fracture

1,543,998

1,693,314

1,884,289

22.0

≥2 fractures

378,929

422,313

482,713

27.4

≥3 fractures

96,018

108,294

125,577

30.8

When extending the description of postmenopausal osteoporotic women to women aged ≥50 years with a T score below −2.5 SD or with a history of fragility fracture, the prevalence is predicted to rise more markedly (+18.5%, from 2.956 million women in 2010 to 3.503 million in 2020).

Improvements in life expectancy are also expected to result in a significant increase in the prevalence of fractures and of multiple fractures in postmenopausal women. Indeed, the number of women with a history of fracture is expected to rise by 22.0% (from 1.544 to 1.884 million), and the number of women with three or more fractures is predicted to increase by 30.8% (see Table 4).
Table 4

Sensitivity analysis on the rate of incidence: fracture incidence and mortality estimates from 2010 to 2020

 

2010

2015

2020

Increment 20102020 (%)

All osteoporotic fractures

191,690

195,607

202,763

5.78

 Aged 5064 years

38,393

37,260

38,252

0.37

 Aged 6579 years

71,667

75,798

77,844

8.62

 Aged ≥80 years

81,630

82,549

86,667

6.17

Clinical vertebral fractures

26,510

27,013

28,061

5.85

 Aged 5064 years

3,963

3,702

3,881

2.08

 Aged 6579 years

11,527

12,171

12,483

8.30

 Aged ≥80 years

11,020

11,140

11,698

6.15

Hip fractures

44,238

44,782

46,489

5.09

 Aged 5064 years

2,985

2,832

2,932

1.78

 Aged 6579 years

12,836

13,414

13,945

8.64

 Aged ≥80 years

28,417

28,536

29,612

4.20

Deaths

7,694

7,258

7,253

5.73

 Aged 5064 years

219

191

190

13.43

 Aged 6579 years

1,371

1,240

1,223

10.75

 Aged ≥80 years

6,105

5,827

5,840

4.33

Sensitivity analysis

Assuming a 1% annual increase in the rate of fracture prior to 1992 and a 1% annual decrease after 1993, the model predicted lower numbers of osteoporotic fractures, varying from 191,690 fractures in 2010 and 202,763 in 2020 as compared with the base case scenario (−17% and −30%, respectively). The predictions in the number of fractures by site (hip, vertebral and non-hip non-vertebral) and deaths attributable to fractures were 14–16% lower than the estimates obtained assuming a constant incidence rate by age over time. For 2015, the differences varied between 18% and 20%, depending on the endpoint; for 2020, it reached 21–24%. The details of the estimates are reported in Table 4. Under these assumptions, the number of fractures was still predicted to increase, but to a lower extent than in the base case scenario (+5.78% versus +17.2%).

Discussion

This study aimed to estimate the present and future burden of osteoporosis in the UK by adapting an existing disease model that was developed and validated against Swedish data.

The present analysis assumed an unchanging age-specific BMD (mean and standard deviation) and incidence of fractures over time, so that our estimates are driven by demographic changes. These forecasts are relatively robust in the sense that all women who will be elderly in 2020 are already adults. Based on these assumptions, the number of osteoporotic fractures in postmenopausal women was estimated at 224,219 in 2010. Hip fractures accounted for 23.2% of fractures (51,927), and vertebral fractures accounted for 13.8% of fractures (30,994). Although non-hip non-vertebral fractures accounted for the vast majority of fractures (63.0%), the burden associated with these fractures is poorly documented. The age group bearing most of the burden was women aged ≥70 years, who incurred 68% of all osteoporotic fractures whilst accounting only for 37% of the study population. Lastly, the prevalence of osteoporosis was estimated to vary from 4.9% in women aged 50–55 years to 27.1% in women aged ≥85 years.

The number of women aged ≥50 years is expected to increase by 15.6% between 2010 and 2020, with an important contribution of the age group 65–79 (explaining 44% of the increase), followed by the 50–64 age group (explaining 39% of the increase).

The number of fractures is expected to increase by 17.2% by 2020 to reach 262,847 fractures (including 60,540 hip fractures and 36,556 vertebral fractures).

The increase in the number of vertebral fractures is expected to be slightly higher than for other sites (+17.9% compared with +16.6% for hip and +17.3% for non-hip non-vertebral fractures). This may be explained by the marked increase in the population aged 65–79 in whom the proportion of vertebral fractures is relatively high (approximately 16%) compared with younger (10% in women aged 50–64) and older women (13% in women aged ≥80 years) and by the effect of a longer life expectancy combined with the high risk of recurrent vertebral fractures (relative risk of 4.4) [17].

Assuming a constant BMD by age over time, the number of women with osteoporosis is predicted to increase by 16.5% to reach 2.1 million in 2020. Using local data, the number of women with osteoporosis was predicted to increase by 16.5%, whilst the number of women with a T score below −2.5 SD or a history of fragility fracture was forecast to increase by 18.5%.

Lastly, the number of women with multiple fractures is expected to increase more markedly (+30.8%) than the number of women with single fractures (+19.4%) because of the longer life expectancy of women having experienced a first fracture.

Our prediction of the total number of fractures was 15% higher than the estimate reported by Burge et al. [10] who predicted a total number of 230,000 fractures in 2020. This may be explained by the sites considered for their estimation as they considered only forearm/wrist, upper arm, humerus, leg and ankle as non-hip non-vertebral osteoporotic sites.

Our study is subject to several limitations due to the lack of reliable epidemiological UK data. First, the incidence of fracture by age was assumed to remain constant over time, which may not be realistic due to changes in the prevalence of different risk factors (e.g. BMD, an increase in the prevention and treatment of osteoporosis, and falls). A sensitivity analysis assuming a 1% decrease per year in the incidence rate by age from 1993 was therefore implemented. Under these assumptions, the total number of osteoporotic fractures is estimated at 191,690 in 2010 and is expected to increase by 5.78% between 2010 and 2020. This demonstrates that the estimates are relatively sensitive to the incidence of fracture by age, but also that prevention and treatment campaigns have the potential to limit the incremental burden of osteoporosis in the future. It also highlights the need for further studies on the incidence of fracture in order to select the appropriate assumptions.

Secondly, due to a scarcity of data, some model parameters were based on population studies conducted in Sweden. This was the case for the proportion of fractures which are first, the proportion of vertebral and non-hip non-vertebral fractures, and the relative risk of death following an osteoporotic fracture.

Thirdly, the BMD data, obtained from a UK population [33] (personal communication with A Gauthier), provided some unexpected results as the mean BMD reported in the oldest population (women aged ≥80 years) was higher than that observed in women aged 70–79 years (0.663 and 0.657, respectively). As the standard deviation related to these estimates is high (the study sample included only 41 women aged ≥80 years), the difference is not significant, but the point estimates may lead to some discrepancies.

Based on our analysis, only 24% of women with a history of fracture have a T score lower than −2.5 SD, which implies that the guidance of NICE in the UK recommending treatment in women with a T score below −2.5 SD when they are under 75 years is conservative [24, 35]. In contrast, the guidance of the National Osteoporosis Guideline Group provides for intervention in postmenopausal women with a prior fragility fracture and in others where the 10-year probability of fracture exceeds a given intervention threshold, irrespective of any T score [36, 37].

Our results have to be interpreted with caution as they rely on several assumptions on the epidemiology of osteoporosis due to a scarcity of data. In particular, further research would be required to document the incidence of non-hip fracture, the trends in incidence of fractures and the trends in BMD over time in the UK.

Osteoporosis has recently been recognized as a major public health burden in developed countries [38]. This study confirms the importance of prevention in this area as the demographic changes in the coming 10 years are expected to lead to a marked increase in the incidence of osteoporotic fractures (+17.2%) and in the number of women living with a history of multiple fractures (+27.4%).

Acknowledgements

This study was supported by the International Osteoporosis Foundation and funded by Amgen (Europe) GmbH. The authors thank Dr. Matthew Gitlin and Sean Robbins from Amgen (Europe) GmbH for their involvement in this project, and Professor Alistair McGuire from the London School of Economics for his advice on the model development. Funds were provided by Amgen (Europe) GmbH and GlaxoSmithKline to Bioscript Stirling for minor editing and styling support.

Conflicts of interest

J.A. Kanis, E. McCloskey, and C. Cooper have no conflicts of interest. J. Compston has received grant funding from Osteotronix and Nycomed; received speaking and/or advisory fees from Novartis, Amgen, Servier, GSK, Gilead, Procter & Gamble/Sanofi Aventis, Eli Lilly, Merck Sharp & Dohme, Medtronic and Warner-Chilcott; and has provided consultancy to Novartis and Amgen. The work undertaken by A. Gauthier, Y. Jiang, M. Martin and F. Borgstrom on the analysis and model development was done under contract to Amgen.

Supplementary material

11657_2011_63_MOESM1_ESM.doc (496 kb)
ESM 1(DOC 496 kb)

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2011