Osteoporosis International

, Volume 21, Issue 4, pp 561–568

Population screening for osteoporosis risk: a randomised control trial of medication use and fracture risk


    • Bone and Musculoskeletal Research ProgrammeUniversity of Aberdeen
  • A. Stewart
    • Bone and Musculoskeletal Research ProgrammeUniversity of Aberdeen
  • D. J. Torgerson
    • York Trials UnitUniversity of York
  • D. M. Reid
    • Bone and Musculoskeletal Research ProgrammeUniversity of Aberdeen
Original Article

DOI: 10.1007/s00198-009-1007-x

Cite this article as:
Barr, R.J., Stewart, A., Torgerson, D.J. et al. Osteoporos Int (2010) 21: 561. doi:10.1007/s00198-009-1007-x



Randomised control trial of osteoporosis screening in 4,800 women aged 45–54 years was carried out. Screened group observed an increase of 7.9% in hormone replacement therapy (HRT) use (p < 0.001), 15% in other osteoporosis treatments (p < 0.001) and a 25.9% reduction in fracture risk compared with control. Screening for osteoporosis significantly increases treatment use and reduces fracture incidence.


Population screening programmes can identify menopausal women with low bone mineral density (BMD) and elevated risk of future fracture but require to be proven effective by a randomised control trial.


A total of 4,800 women, 45–54 years, were randomised in equal numbers to screening or no screening (control) groups. Following screening, those in the lowest quartile of BMD were advised to consider HRT. Nine years later, the effect of screening on the uptake of treatment and the incidence of fractures were assessed by postal questionnaire. Categorical differences were assessed using χ2 test. Cox regression was used to assess hazard ratio (HR).


Of the screened and the control groups, 52.4% vs 44.5%, respectively, reported taking HRT (p < 0.001). In addition, 36.6% of the screened vs 21.6% of the control groups reported the use of vitamin D, calcium, alendronate, etidronate or raloxifene (p < 0.001). In a per protocol analysis of verified incident fractures, a 25.9% reduction in risk of fractures (of any site) in the screened group was observed (HR = 0.741, 95% CI = 0.551–0.998 adjusted age, weight and height).


Screening for osteoporosis as assessed by low bone density significantly increases the use of HRT and other treatments for osteoporosis and reduces fracture incidence.


FractureHormone replacement therapyRandomised control trial


Osteoporosis is a common condition leading to fractures, which occur in one in two women in the UK [1] and is responsible for a high degree of morbidity and mortality and resource use [2, 3]. Although a number of randomised trials of different osteoporosis treatments have demonstrated that future fracture risk can be reduced following appropriate treatment, the management of established osteoporosis is unsatisfactory because most treatment addresses the problem after a major fracture has occurred [46]. Therefore primary prevention of osteoporosis is desirable. Population screening programmes are one means to identify women with low bone density and elevated risk of future fracture. However, population-based screening for osteoporosis risk is controversial because, to date, no study has directly shown the effect of screening on the incidence of fracture [7]. The current recommendations for screening in the USA rely on indirect evidence [8].

Although osteoporosis screening studies have been performed, to date, these have been non-randomised [9], and the most robust method to test the effectiveness of screening for osteoporosis is through a randomised controlled trial. In 1997, preliminary results from the Aberdeen Prospective Osteoporosis Screening Study (APOSS) demonstrated that, in a randomised controlled trial of osteoporosis screening of 1,600 women, followed for 2 years after screening, HRT use was significantly increased in the screened population compared with controls, although no effect on fractures was noted [10]. This analysis was part of a larger trial, which was designed to have a long-term follow-up. Here, we describe the full long-term results of the APOSS randomised controlled trial of osteoporosis screening in 4,800 women followed for up to 9 years.


The APOSS is a community-based study involving women aged between 45 and 54 years of age and living within a 32-km radius of the Osteoporosis Screening Unit, Aberdeen, Scotland. Women invited to take part were chosen at random from a primary care patient register, the Community Health Index (CHI). In 1993, as part of this wider study, 4,800 of the identified women were further randomised and assigned in equal numbers to either the screening arm of the study or an inactive control arm where women were identified from the patient register but were not invited to actively participate in the assessment arm of the screening study (Fig. 1).
Fig. 1

Study flow diagram showing disposition and follow-up of screened and control women

The programme initially consisted of a letter sent directly from the screening unit inviting the active potential participants (the screened group) to take part in the programme, which consisted primarily of a dual energy X-ray absorptiometry scan of the hip and spine using a Norland XR-36. Women randomised to the inactive arm (controls) were not contacted at baseline. Screened women who attended were considered ‘at risk’ of osteoporosis if their bone mineral density (BMD) at either the femoral neck or lumbar spine (L2-4) fell within the lowest quarter of the first 1,000 women screened [11]. Scan results and appropriate advice were sent to the women's GPs. For women ‘at risk’, the GP was advised to offer HRT when the women reached the menopause provided that there were no contraindications. For women with BMD values in the top three quarters, the advice was that HRT was not specifically required for osteoporosis intervention, unless drugs such as corticosteroids were subsequently prescribed. In addition, the letters included advice on lifestyle changes the GP might recommend particularly to women unable or unwilling to take HRT.

In 2002, all of the women, who could be contacted, from the original cohorts regardless of whether the 2,400 potential active assessment participants attended for screening or not were identified. Two thousand one hundred eighty-one (90.9%) of the 2,400 control group (controls) and 2,256 (94%) of the 2,400 participants randomised to the screened group were mailed a questionnaire. The questionnaire asked about general health, medical conditions, medication pertinent to bone health, falls in the last year and fractures since the age of 50 years. Women were also asked to report their current weight and height. The remaining 219 control (9.1%) and 144 screened (6%) women were lost to follow-up through death or change of abode prior to mailing the questionnaire (Table 1). The study was approved by the Grampian Research Ethics Committee.
Table 1

Response to questionnaire


Control group% (n)

Screened group% (n)

Returned completed

56.8 (1364)

59.7 (1433)

Not returned

26.2 (628)

25.6 (614)


0.3 (8)

1.2 (28)


0.0 (1)

0.1 (2)

Returned blank

3.9 (93)

3.5 (85)

Return to sender


3.8 (91)

Asked to withdraw

0.0 (1)

0.3 (6)

Died before questionnaire

3.2 (77)

3.1 (75)

Moved before questionnaire

5.9 (142)

2.7 (66)


100 (2,400)

100 (2,400)

Self-reported fractures were verified through consultation of individual X-ray reports written by a trained radiologist and included on the local radiology reporting system. In those cases where fractures could not be verified from the X-ray report, letters were sent to the individual's primary care physician asking whether they could confirm the reported fracture from their records. Only fractures validated by one of these methods were included in the analysis.

Statistical analysis

The primary analysis of drug use was undertaken by ‘intention-to-treat’ analysis, where all subjects in the screened and control groups who responded to the follow-up questionnaire were included. In addition, a per-protocol analysis was undertaken to investigate fracture rates, and here, only screened women who attended for screening were included in the analysis. SPSS statistical software (version 13.01; SPSS Inc, Chicago) was used for statistical analyses. Chi-square (χ2) tests were used to assess categorical differences between the two groups and Student's t test to assess non-categorical differences. Cox regression analysis was performed to estimate the hazard ratio of fracture. Due to differences in age, weight and height between those women found to be at ‘high risk’ of osteoporosis and those in the highest three quartiles of BMD, data were adjusted for age, weight and height.

Stata 9.2 was used to perform treatment received analysis to estimate the hazard ratio of fracture adjusted for treatment effect.


There was a small but significant difference (p < 0.001) in the time to follow-up in the screened group (9.1 years) compared with the control group (8.8 years). A significant difference (χ2 = 16.8, p = 0.019) was observed in the response rates of the entire control group compared with the entire screened group. A completed questionnaire was returned by 56.8% of the entire control group compared with 59.7% in the entire screened group (Table 1). This difference was attributable to loss to follow-up prior to mailing out the questionnaire, as there was no significant difference in response rate found between the two groups when only those mailed the follow-up questionnaire were included in the analysis (χ2 = 0.457, p = 0.499). Of those mailed a follow-up questionnaire, 62.5% (n = 1,364) of the control group and 63.5% (n = 1433) of the screened group returned a completed questionnaire.

No significant difference between the two groups, at follow-up, was found in any of the demographic parameters (age p = 0.72, weight p = 0.65 and height p = 0.83; Table 2). However, compared with women in the highest three quartiles of BMD, women in the lowest quartile of BMD were found to be significantly older (59.2 vs 58.0 years; p < 0.001), lighter (65.1 vs 71.4 kg; p < 0.001) and shorter (1.60 vs 1.61 m; p < 0.001).
Table 2

Group demographics (mean ± SD)




p value

Age (years)

58.4 ± 3.3

58.4 ± 4.0


Weight (kg)

68.8 ± 12.9

69.1 ± 18.5


Height (m)

1.61 ± 0.06

1.61 ± 0.06


General health (%)

Very good



χ2 = 2.00




p = 0.74





Not so good








No significant difference (χ2 = 2.0, p = 0.74) was observed in the self-reported general health of the two groups, with 31.1% of the control group and 31.4% of the screened group reporting their general health over the last year to be excellent while only 2.1% of the control group and 1.5% of the screened group reporting poor general health (Table 2).

In a comparison of different disease states, the screened group self-reported a significantly (p = 0.05) higher prevalence of hyperparathyroidism (1.1%) than the control group (0.5%). No significant differences were observed in any of the other disease states, including vitamin D deficiency, celiac disease, malabsorption, breast cancer, multiple sclerosis and Parkinson's disease.

A significantly greater number of the screened subjects reported current or past use (for more than 3 months) of HRT (screened = 52.4% vs controls = 44.5%, p < 0.001). In HRT users, the mean duration of HRT use was not, however, significantly different between the screened and control groups (screened = 84.0 months vs controls = 86.0 months p = 0.58). Women were asked if at any time since commencing HRT, they had stopped and re-started therapy. Continuous use of HRT was reported in 78.0% of the control group and 78.9% of the screened group who had used HRT at any time (p = 0.74).

A significantly greater number of the screened group, 36.6% vs 21.6% in the control group reported having taken some form of osteoporosis medication, excluding HRT (p < 0.001). This finding was also observed when the individual treatment types were analysed with a significantly greater proportion of the screened group reporting taking vitamin D, calcium supplements, alendronate, etidronate and raloxifene than in the control group (Table 3).
Table 3

Self-reported use of medication


Control, % (n)

Screened, % (n)

p value


44.5 (600)

52.4 (747)


Vitamin D

12.5 (163)

24.2 (345)



14.1 (187)

20.0 (287)



1.3 (17)

4.7 (66)



1.6 (21)

5.1 (72)



0.4 (5)

0.6 (8)



0.3 (4)

0.9 (13)


Any osteoporosis medication excluding HRT

21.6 (280)

36.6 (517)


Any osteoporosis medication including HRT

59.4 (768)

69.0 (975)


Oral Steroids

6.8 (90)

7.1 (101)



11.1 (146)

8.5 (120)



15.3 (188)

15.3 (213)



1.3 (17)

1.4 (19)



17.6 (232)

17.1 (244)


No significant difference was observed in the use of oral steroids, diuretics, anti-epileptics or anti-depressants (Table 3), although a significantly greater number of the control group reported taking thyroxine than the screened group (control, 11.1% vs screened, 8.5%, p = 0.021).

Of the screened women in the lowest quartile of BMD at either the hip or the spine, 62.6% reported taking HRT compared with 48.4% of those in the upper three quartiles of hip and spine BMD (χ2 = 23.97, p < 0.001). There was no significant difference between HRT use in the upper three quartiles of BMD and the control group (44.5%, χ2 = 2.89, p = 0.089; Table 4). However, there was a significant difference between controls and those women in the upper three quartile of BMD in their use of vitamin D preparations (high BMD group, 26.8%; controls, 12.5%, P < 0.001), while there was a trend to increased calcium use (high BMD group, 17.0%, controls 14.1%; Table 4).
Table 4

Comparison of the mean percentage of subjects reporting current or past use of osteoporosis treatments


Control (%)


High BMD

p all

p Control vs. high BMD







Vitamin D




































Controls vs screened in lowest 25% BMD (low BMD) vs screened in the highest 75% BMD (high BMD)

No significant difference was found between the numbers of women who reported falling in the previous year in the screened group (15.6%) compared with those who reported falling in the previous year in the control group (13.9%; χ2 = 1.62, p = 0.20). Similarly, there was no significant difference between the two groups in the number of falls reported per year (screened = 0.3 falls/year vs control 0.28 falls/year, p = 0.503).

Using intention-to-treat analysis, including all women randomised to the assessment group and controls who subsequently responded to the follow-up questionnaire, 8.8% (n = 109) of the screened group of women had suffered a validated incident fracture (any type) compared with 9.4% (n = 111) of the control group. Women who suffered an incident fracture were slightly older that those women who did not (58.2 vs 58.8 years; p < 0.001). The risk of fracture was reduced by 20.9%, although this was not significant (HR = 0.791; 95% CI, 0.600–1.042; p = 0.096 adjusted age, weight and height).

No significant difference was seen in the type of incident fracture (χ2 = 15.77, p = 0.33; Table 5).
Table 5

Percentage of incident fracture type


Control, % (n)

Screened, % (n)

Per-protocola screened, % (n)

No fracture

90.6 (1,073)

91.2 (1,132)

92.2 (989)


0.2 (3)

0.2 (3)

0.1 (1)


2.9 (34)

2.7 (34)

2.4 (26)


0.2 (2)

0.2 (2)

0.0 (0)


0.7 (8)

0.9 (11)

0.8 (9)


0.1 (1)

0.1 (1)

0.1 (1)


2.9 (35)

1.3 (16)

1.0 (11)


0.5 (6)

0.6 (7)

0.6 (7)


0.3 (4)

0.6 (7)

0.6 (6)


1.0 (12)

1.4 (17)

1.2 (13)


0.1 (1)

0.2 (3)

0.3 (3)


0.2 (2)

0.3 (4)

0.4 (4)


0.2 (2)

0.0 (0)

0.0 (0)


0.1 (1)

0.3 (4)

0.3 (3)

aThe per-protocol group were those women from the original 2,400 women randomised to the screened population who attended the initial screening assessment and were therefore made aware of their BMD results via their primary care physician. This group were compared in the analysis to the entire control group

Using treatment-received analysis, the adjusted treatment received estimate showed a reduction in risk of fracture of 24.1%; this was not significant (HR = 0.759; 95% CI, 0.548–1.050, p = 0.096 adjusted age, weight and height)

A per-protocol analysis was undertaken, including all women from the original 2,400 women randomised to the screened population who attended the initial screening assessment and were therefore made aware of their BMD results. In this analysis, when compared to the control women, a 22.9% reduction in risk of fractures of any site was observed in the screened group (HR = 0.771, 95% CI = 0.575–1.035, p = 0.083). However, after adjustment for age, weight and height, there was a significant 26.6% reduction in risk of incident validated fractures (HR = 0.734, 95% CI = 0.546–0.988, p = 0.041). When the per-protocol analysis was restricted to the major osteoporotic fractures (hip, wrist, vertebral and humeral), a 27.6% reduction in fracture risk was noted, although this was not statistically significant (HR = 0.724, 95% CI = 0.465-1.128, p = 0.153) even when adjusted for age, weight and height.


In the present randomised controlled trial, we have assessed the long-term effectiveness of osteoporosis screening on the use of medications pertinent to the treatment of osteoporosis and fracture risk. We have demonstrated significantly increased use of treatments in subjects randomised to screening and in a per-protocol analysis a subsequent reduction in fracture rates after adjustment for major confounders.

A limitation of our data was the response rate, which was just over 60% in both the screened and control groups mailed a follow-up questionnaire. However, given the 9-year follow-up time, this was a reasonable response rate and is similar to that achieved in a middle aged local population who took part in the Aberdeen children of the 1950s study [12].

The screened and control groups were found to be demographically similar with no significant difference neither in age, weight or height nor in self-reported general health. This supports the finding in 1997 that there was no bias, in observable variables, introduced by differential responses from the screened and unscreened groups [10].

A comparison of common diseases found no difference between the screened and the control groups, with the exception of hyperparathyroidism. In the general population, the incidence of hyperparathyroidism is one in 500 to one in 1,000 [13]. These diseases are, however, all self-reported, and it is possible that some of the participants confused hyperparathyroidism with hyperthyroidism. However, it is also possible that the report of low bone mass in the actively screened women prompt the primary care physicians to screen for secondary causes of low bone mass, although this was not specifically advised. It is also possible that due to the number of diseases tested that this observation occurred through chance.

The effects of screening on treatment uptake mirrors our short-term findings published in 1997 when we demonstrated that the use of HRT 2 years post-screening was higher in the screened group described than in the control group [10]. This current finding is in this much larger APOSS cohort (2,400 compared with the earlier 799 women) with a much longer follow-up (mean, 9.2 vs 2.0 years). The questionnaire sent in 2002 was more detailed than the 2-year post-screening questionnaire and included a question about the use of other medications pertinent to osteoporosis treatment and prevention. In this larger trial of screening, we found that the number of screened women taking each of the different osteoporosis treatment types (vitamin D, calcium, alendronate, risedronate, eitidronate and raloxifene) was significantly greater than their control counterparts. This demonstrates that, in addition to increasing the HRT usage, the use of other bone treatments was also increased.

A sub-analysis was performed to investigate whether this increased use in treatments pertinent to osteoporosis was confined to those women advised to take HRT at the time of the menopause (those in the lowest quartile of BMD, hip or spine). This revealed that, although the percentage of women above the lowest quartile of BMD taking treatments was lower than that of the advised group, the number of women in the highest 75% of the screened group who had taken some form of osteoporosis treatment was significantly greater than in the control group. As these women were not advised to take treatment for osteoporosis, it is perhaps an increased awareness of bone health acquired through taking part in the screening process that has led to this significant increase in treatment, some of which may be through self-medication with vitamin D and, to a lesser extent, calcium.

In the per-protocol analysis in this current study, we found a significant reduction in fracture rate for any site in the screened group compared with the control group. To date recommendations about screening for osteoporosis have been based on indirect evidence of fracture prevention [8]. We report for the first time a direct evidence of a significant reduction in fractures (any site) following screening for osteoporosis. We also found a reduction in osteoporotic fractures (hip, wrist, vertebral and humeral fractures), although this reduction failed to reach significance. Buist et al. [14] reported an increased uptake of anti-resorptive therapy and a reduction in hip fractures following participation in a non-randomised osteoporosis screening programme. The participants in the Buist et al. study were, however, older than the APOSS participants (mean age, 68.9 compared with 58.4 in APOSS). Women enrolled in the APOSS study were aged 45–54 and at follow-up were aged 54–63, and these women are beginning to arrive at an age when the number of osteoporotic fractures will increase dramatically. A future follow-up of women who have taken part in this RCT will determine if the reduced fracture rate is sustained.

One of the concerns of regulatory authorities and reimbursement agencies is the inappropriate use of medications following screening. The results of this study would suggest that screening for osteoporosis risk does lead to appropriate treatment albeit most subjects receiving HRT, which would not necessarily be the treatment of choice, were the study to be re-run in 2009.

The significant reduction in fracture risk, associated with participation in this study, could not be explained by a reduction in falls, since the number of fallers and the number of falls in the screened and control groups were not significantly different, again implying that increased use of osteoporosis treatments was responsible.

This RCT of osteoporosis screening was started in the early 1990s, and because of this, there are two inherent issues surrounding this study. The role of HRT in fracture prevention is well documented, and meta-analyses of HRT trials have demonstrated its efficacy at reducing both vertebral and non-vertebral fractures in postmenopausal women [15, 16]. Thus, in 1990, when the APOSS study was started, the treatment of choice for postmenopausal prevention of osteoporosis was HRT. However, in 2003, following the results of the Million women study [17], the Medicines and Healthcare Product Agency and the Committee on Safety of Medicines in the UK, in line with other US and European agencies, advised that for long-term use in preventing osteoporosis in women greater than 50 years of age, the risk/benefit ratio of HRT was unfavourable. They advised that HRT should only be used for preventing osteoporosis by those women who are unable to take other osteoporosis prevention treatments or for whom other treatments have been unsuccessful. If this RCT of osteoporosis screening were to be started today, the recommendation to primary care physicians might not be to prescribe HRT at the menopause for osteoporosis prevention. However, we also demonstrated a significant increase in at risk women taking other forms of osteoporosis treatment; therefore, treatment as a whole has increased in the screened group.

The second point relates to the identification of those with low bone density and the threshold used to advise intervention. This randomised controlled trial of osteoporosis screening began prior to the introduction of the WHO classification for osteoporosis in 1997 [18]. In the absence of these criteria, women were advised to take HRT at the time of their menopause if their bone density was found to be in the lowest quartile of bone density at either spine or femoral neck as measured in this population. This is likely to have led to a greater number of women being identified as having low bone density than if the WHO cut-off of a T-score of <−2.5 had been used. However, the women identified as at risk in this study would have an approximate one standard deviation reduction in BMD relative to an age-matched control population hence at least doubling their inherent fracture risk.

One of the limitations with this study was that the use of HRT was self-reported. Most women will know whether or not they have taken some form of HRT. However as the average length of time to follow-up was approximately 9 years, recalling some of the finer details of their HRT history, for example exact start and stop dates, may be difficult. However, it has been previously shown in a prospective cohort that recall of menopausal status is good [19].

In conclusion, this randomised controlled trial of osteoporosis screening demonstrated an increased uptake of osteoporosis treatment and a significant reduction in fracture risk in those women who participated in the screening programme. This study strongly suggests that a population screening programme to target treatment of those at risk of osteoporosis will reduce fractures. It does however remain unclear which specific aspects of screening followed by therapy gives rise to this reduction in fracture risk.

Further work is required to determine the optimum age to undertake screening to ensure a cost-effective programme.


The authors would like to thank Graeme Maclennan for his statistical advice and help calculating the adjusted treatment received estimate. DMR is grateful to the arthritis research campaign for continued infrastructure support. Grant funding pertinent to the data collection for this study was gratefully received initially from the Wolfson Foundation and subsequently from GlaxoSmithKline. Finally, we express our gratitude to all the women who gave so freely of their time to take part in this study.

Conflicts of interest


Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2009