- Email: [email protected]
Risk of hip fracture according to the World Health Organization criteria for osteopenia and osteoporosis
Bone Vol. 27, No. 5 November 2000:585–590
Risk of Hip Fracture According to the World Health Organization Criteria for Osteopenia and Osteoporosis J. A. KANIS,1 O. JOHNELL,2 A. ODEN,3 B. JONSSON,4 C. DE LAET,5 and A. DAWSON6 1
WHO Collaborating Centre for Metabolic Bone Diseases, University of Sheffield Medical School, Sheffield, UK Department of Orthopaedics, Malmo General Hospital, Malmo, Sweden 3 Consulting Statistician, Gothenberg, Sweden 4 Department of Economics, Stockholm School of Economics, Stockholm, Sweden 5 Institute for Medical Technology Assessment, Rotterdam, The Netherlands 6 Lilly Research Centre, Ltd., Windelsham, Surrey, UK 2
Key Words: Hip fracture; Relative risk (RR); Population risk; Bone mineral density (BMD); Osteoporosis; Osteopenia.
The risk of hip fracture is commonly expressed as a relative risk. The aim of this study was to examine the utility of relative risks of hip fracture in men and women using World Health Organization (WHO) diagnostic criteria for low bone mass and osteoporosis. Reference data for bone mineral density (BMD) at the femoral neck, from the third National Health and Nutrition Examination Survey (NHANES III), were applied to the population of Sweden. Relative risks (RRs) were calculated from the known relationship between BMD at the femoral neck and hip fracture risk. The apparent prevalence of low bone mass and osteoporosis depended on the segment of the young population chosen for reference ranges. Using a reference derived from women aged 20 –29 years, the prevalence of osteoporosis was 21.2% in women between the ages of 50 and 84 years and 6.3% in men. The RRs associated with osteoporosis depended markedly on the risk comparison. For example, in men or women aged 50 years, the RR of hip fracture in those with osteoporosis compared to those without osteoporosis was 7.4 and 6.1, respectively. The RR of those at the threshold value for osteoporosis compared to those with an average value for BMD at that age was 6.6 and 4.6 in men and women, respectively. RRs were lower comparing those at the threshold value compared to the risk of the general population at that age (4.2 and 2.9, respectively). When RR was expressed in relation to the population risk rather than to the risk at the average value for BMD, RR decreased at all ages by 37%. Such adjustments are required for risk assessment in individuals and for the combined use of different risk factors. Because the average T score at each age decreased with age, the RR of hip fracture at any age decreased with advancing age in the presence of osteoporosis. The decrease in relative risk with age is, however, associated with an increase in absolute risk. Thus, for clinical use, the expression of absolute risks may be preferred to relative risks. (Bone 27: 585–590; 2000) © 2000 by Elsevier Science Inc. All rights reserved.
Introduction Age and bone mineral density (BMD) are the strongest risk factors for hip fracture thus far discovered. These risk factors are in part independent, so that absolute values for BMD have a different significance at different ages.7 Diagnostic criteria for osteoporosis (and low bone mass) have been proposed by an expert panel of the World Health Organization (WHO),15,24 and are widely used in epidemiology by regulatory agencies for practice guidelines and in clinical research.3,10,20,23,25 The diagnostic criteria stratify individuals according to the BMD of the young healthy population, so that osteoporosis in women is defined as a BMD value that lies 2.5 standard deviations (SD) or more below the average value in young healthy women (T score ⫽ ⫺2.5 SD). An advantage of using young normal reference ranges is that the prevalence of osteoporosis defined in this way increases exponentially with age in much the same way as does the frequency of many fractures due to osteoporosis. Some have advocated the use of the Z score rather than the T score. The Z score represents the number of SDs that a given value for BMD deviates from the average value for age and gender. Thus, an individual aged 60 years with a Z score of ⫺1 SD would lie in the lowest 16% of the distribution of BMD for that age, assuming a normal distribution of BMD. Similarly, an individual aged 80 years with a Z score of ⫺1 SD would lie at the same point within the normal distribution. For this reason, the use of the Z score to denote osteoporosis would mean that the prevalence of osteoporosis would not increase with age and is, therefore, inappropriate in a clinical setting. On the other hand, the use of the T score undervalues the importance of age in contributing to risk. For example, the significance of osteoporosis in a woman aged 50 years differs from that of a woman with the same BMD (and T score) who is aged 70 years. The relative risk is higher in the younger woman, although the absolute risk is lower, at least in the short term. The risk of hip fracture is commonly expressed as a relative risk (RR). RRs may be used by clinicians to make treatment decisions. The expression of RR is, however, not standardized. For example, many observational studies describe the risk of hip fracture as the increase in fracture risk per SD decrease in BMD. Epidemiological studies commonly report the RR in individuals
Address for correspondence and reprints: Dr. John A. Kanis, Centre for Metabolic Bone Diseases, WHO Collaborating Centre, University of Sheffield Medical School, Beech Hill Road, Sheffield S10 2RX, UK. E-mail: [email protected] © 2000 by Elsevier Science Inc. All rights reserved.
585
8756-3282/00/$20.00 PII S8756-3282(00)00381-1
586
J. A. Kanis et al. Risk of hip fracture according to WHO
Bone Vol. 27, No. 5 November 2000:585–590
Table 1. Prevalence (%) of low bone mass and osteoporosis at the ages specified utilizing different ages to derive cutoff values Mena
Low bone mass 50 60 70 80 Osteoporosis 50 60 70 80
Menb
Women
20–29 yr
20–39 yr
50 yr
20–29 yr
20–39 yr
20–49
50 yr
20–29 yr
20–49 yr
50 yr
38.5 47.1 56.6 69.3
32.1 40.6 50.2 63.4
16.0 23.0 30.9 43.7
40.9 61.5 80.1 92.8
36.7 57.0 74.5 90.8
31.5 51.1 71.3 87.7
16.0 30.3 49.6 71.0
24.2 32.4 41.4 54.8
17.5 24.9 33.0 46.0
7.9 12.9 18.6 29.2
2.8 5.5 8.6 15.4
1.9 3.4 6.4 12.6
0.6 1.6 2.7 5.8
5.4 12.3 24.5 43.3
3.9 9.2 19.3 36.1
2.5 6.1 13.7 27.5
0.6 1.8 4.6 11.1
2.4 4.8 7.6 14.0
1.0 2.4 4.0 8.1
0.3 0.7 1.3 3.1
a Prevalences derived from NHANES male reference ranges for bone mineral density. Thresholds: 20 –29 years, 0.792 g/cm2; 20 –39, 0.770; 50, 0.702 for osteoporosis and 0.585, 0.564, and 0.510, respectively, for low bone mass. b Prevalences derived from NHANES female reference ranges. Thresholds: age 20 –29 years, 0.740 g/cm2; 20 –39, 0.729; 20 – 49, 0.710; 50, 0.649 for osteoporosis and 0.577, 0.556, 0.533, and 0.407 for low bone mass, respectively.
with a risk factor compared to those without the factor; for example, the RR of hip fracture in women with or without osteoporosis. Others describe the risk of fracture compared with an average; for example, the risk of hip fracture based on BMD compared to an individual with an average BMD. All these approaches give differing RRs. For clinical purposes it is appropriate to describe the risk of hip fracture in an individual (or a group of individuals) compared with the average risk of hip fracture (not average BMD). The aims of this study were to explore the relationships between BMD, age, and RR of hip fracture using the WHO thresholds for osteoporosis and low bone mass (osteopenia). A further aim was to derive RRs appropriate for clinical rather than epidemiological use.
bone mass was 0.740 g/cm2 and 0.577 for osteoporosis. This accords with the proposals recommended by the International Committee for Standards in Bone Densitometry and those used by the NHANES investigators.17 Unlike Looker et al.,17 we chose the same thresholds for men because the risk of hip fracture in men is similar to that in women for the same BMD.3,4 The use of the term “below a threshold” for osteoporosis denotes a BMD value that is less than the threshold value (i.e., a more negative T score). The risk of hip fracture was assumed to increase 2.6-fold (95% confidence interval 2.0 –3.5) for each SD decrease in BMD at the femoral neck, in line with a recent metaanalysis.18 The risk of hip fracture was taken from the National Register Source for Sweden 1994.21 Relative risk by age was expressed in several ways (Figure 1):
Materials and Methods
A. The risk of an individual at the threshold value compared with the hip fracture risk of the total population of that age. This is the RR most appropriate for clinical use.
We used the published data from the third National Health and Nutrition Examination Survey (NHANES III, 1988 –1994) as the reference values for BMD at the femoral neck.16 Threshold values to categorize individuals as having low bone mass or osteoporosis utilized the WHO criteria so that osteoporosis was defined as a T score of less than ⫺2.5 SD. Low bone mass denoted a T score of ⫺1 to ⫺2.5 SD.24 Threshold values were chosen in young women based on the age ranges 20 –29 years, 20 –39 years, 20 – 49 years, and at the age of 50 years, and applied to the male and the female population of Sweden, 1994. In addition, a threshold derived for men at the age of 50 years was utilized. The prevalence of osteoporosis and low bone mass was markedly dependent on the age range chosen to determine threshold values. At the age of 80 years, prevalences varied fourfold in women and nearly threefold in men (Table 1). The use of reference ranges that included successively more elderly women at or below the age of 50 years decreased the apparent prevalence of osteoporosis. For example, where reference ranges and diagnostic thresholds were calculated at the age of 50 years, the prevalence of osteoporosis in the Swedish female population was only 4.3%, and increased with the inclusion of younger women. The use of the age range 20 –29 years as a reference range gave a prevalence of osteoporosis of 21% in women aged 50 – 84 years. This corresponds broadly with the original WHO estimates applied to the UK and USA.15,24 The value of 21% approximates the lifetime risk of hip fracture in Sweden.21 For the purposes of this study we used the age range 20 –29 years for deriving diagnostic thresholds. The threshold for low
Figure 1. Different methods of expressing RR in individuals or populations. The diagram shows the normal distribution of BMD in elderly women (aged approximately 65 years). The average BMD is 1 SD below the mean value for young healthy women and approximately 16% have osteoporosis (T score ⱕ ⫺2.5 SD). RR of hip fracture may be expressed as: (A.) that of an individual at the threshold value for osteoporosis compared with that of the general population; (B.) that of all women with osteoporosis compared with that of the general population; (C.) that of a woman at the threshold value for osteoporosis compared with a woman with an average BMD for that age; and (D.) that of osteoporotic women compared with those without osteoporosis.
Bone Vol. 27, No. 5 November 2000:585–590
J. A. Kanis et al. Risk of hip fracture according to WHO
B. The risk of individuals below the threshold value compared with the risk of hip fracture of the total population of that age. This RR is appropriate for considering a group of individuals (e.g., all those with osteoporosis). C. The risk of an individual at the threshold value compared to an individual with a BMD at the mean value at the same age. This is the RR usually used to describe the performance characteristics of densitometry. D. The risk of those below the threshold value vs. those above the threshold value. This is the most common expression of risk in epidemiological studies. The risk of fracture in the short perspective (i.e., not accounting for deaths) of those with BMD below the threshold for BMD (g) at a certain age is: Yearly incidence of the age group ⫻ ⌽(( g ⫺ )/ ⫹ log(RR))/⌽(( g ⫺ )/)
Table 2. Prevalence of osteoporosis at the age intervals shown in Sweden using female-derived threshold values from the young population, aged 20 –29 years (bone mineral density threshold ⫽ 0.577 g/cm2) Men
Women
Age range (years)
Percent of population
Number affected (000)
Percent of population
Number affected (000)
50–54 55–59 60–64 65–69 70–74 75–79 80–84 50–84
2.5 3.5 5.8 7.4 7.8 10.3 16.6 6.3
7.0 7.6 11.4 14.2 14.6 13.7 14.7 83.2
6.3 9.6 14.3 20.2 27.9 37.5 47.2 21.2
17.0 21.1 30.0 43.7 63.0 68.3 67.8 310.9
(1)
where is the mean and is the SD of BMD at the current age and log(RR) is the e-log of the risk ratio of an individual with 1-SD-lower BMD as compared with another (RR estimated to be 1.4 –2.6, depending on type of fracture and type of BMD measurement). ⌽ is the normal distribution function with mean ⫽ 0 and SD ⫽ 1. For individuals with BMD ⫽ x, the risk is: Yearly incidence of the age group ⫻ exp(⫺log(RR)( ⫺ )/ ⫺ (log(RR)) 2/ 2)
587
(2)
If the mean, , of BMD at a certain age and the standard deviation, , are known, we can calculate the risk ratios vs. the normal population for: (a) those below a threshold g of BMD; and (b) those at a value x of BMD. The confidence estimate of the gradient of risk/SD change in BMD can be used for sensitivity analyses to test differing assumptions concerning the gradient of risk. All relative risks shown can be multiplied by 0.77 or 1.35 to accomodate the range of the confidence intervals. For brevity, this is not done here. Results The prevalence of osteoporosis in the Swedish population increased as expected with age. Approximately 15% of the male and 50% of the female population aged 80 – 84 years had osteoporosis (Table 2). Approximately 6% of men and 21% of women aged 50 – 84 years were classified as having osteoporosis. The prevalence of low bone mass was, as expected, higher than that of osteoporosis at all ages, but did not increase markedly with age (Table 3). Thus, the ratio of individuals with low bone mass to those with osteoporosis varied with age. For example, in women aged 50 –54 years, the number of individuals with low bone mass was sixfold higher than the number with osteoporosis. In the age range 80 – 84 years, the number with either diagnosis was nearly equal. The risk of hip fracture by age, gender, and diagnostic category is shown in Tables 4 and 5. In women aged 25 years with a T score of ⫺1 SD, the relative risk of hip fracture was 2.6 compared to women with average BMD, as expected from the choice of reference material (i.e., age 20 –29 years) (Table 4, column C). The relative risk decreased with age because the mean BMD decreased progressively with age. Indeed, at the age of 55 years, greater than half the female population had osteopenia or osteoporosis; therefore, at this age and older, a T score of ⫺1 was protective (RR ⬍ 1.0).
The risk in women with a T score of ⫺1 (osteopenia) compared with the whole population risk was lower (Table 4, column A) by 37% at all ages. Thus, at the age of 50 years, women with a T score of ⫺1 had a lower risk than the general population of the same age. In all women who fell below the ⫺1 SD threshold, the risk at this age compared with the risk of the whole population of the same age (column B) was 1.9 and remained above unity at all ages. RRs were higher when those below the threshold were compared with those above the threshold (column D) and did not fall with age. This was because the proportion of women with osteopenia at any given age did not decrease with age (see Table 3). A similar phenomenon was observed for osteoporosis in women (Table 5). Thus, at the age of 25 years, a woman at the threshold for osteoporosis (T score ⫽ ⫺2.5) had a 10.9-foldhigher risk of hip fracture than women with an average BMD (column C). The RR decreased with age and was close to unity at the age of 80 years, when approximately 50% of the female population has osteoporosis. In men, expressions of RR were qualitatively, although not quantitatively, similar to findings in women. The RRs of hip fracture compared with the population risk of fracture is shown for women with osteoporosis in Figure 2. As expected, the RR was higher in all osteoporotic women than women at the threshold value and, in both, the RR decreased with advancing age. Table 3. Prevalence of low bone mass at the age intervals shown using female-derived threshold values of BMD from the young population, aged 20 –29 years (bone mineral density between the thresholds of 0.740 g/cm2) Men
Women
Age range (years)
Percent of population
Number affected (000)
Percent of population
Number affected (000)
50–54 55–59 60–64 65–69 70–74 75–79 80–84 50–84
23.0 (25.5) 26.0 (29.5) 28.4 (34.2) 31.0 (38.4) 35.7 (43.5) 40.1 (50.4) 40.9 (57.5) 30.4 (36.7)
66.4 (71.6) 57.0 (64.7) 55.8 (67.2) 59.4 (73.6) 66.6 (81.2) 53.4 (67.1) 36.2 (50.5) 394.8 (475.9)
39.1 (45.4) 46.8 (56.4) 50.5 (64.8) 53.6 (73.8) 56.1 (84.0) 53.2 (90.7) 46.7 (93.9) 49.1 (70.3)
105.7 (122.7) 103.1 (124.2) 106.0 (136.0) 115.9 (159.6) 126.7 (189.7) 96.9 (165.2) 67.1 (134.9) 721.3 (1032.2)
Values in parentheses include all individuals below the threshold for low bone mass and therefore include those with osteoporosis.
588
J. A. Kanis et al. Risk of hip fracture according to WHO
Bone Vol. 27, No. 5 November 2000:585–590
Table 4. Proportion of the male and female population at the ages shown to have low bone mass (osteopenia) and the risk of hip fracture. (Relative risk is expressed for each age, as shown in the footnote below.) Relative riska
Table 5. Proportion of the male and female population at the ages shown identified as having osteoporosis and the risk of hip fracture. (Relative risk shown for each age, as shown in the footnote below.)
Percent of population (%)
A
B
C
D
Age (years)
Men 25 30 35 40 45 50 55 60 65 70 75 80 84
8.4 11.2 14.5 17.6 21.3 24.2 27.5 32.4 36.7 41.4 47.1 54.8 59.9
2.4 2.0 1.7 1.5 1.4 1.2 1.1 1.0 0.9 0.8 0.7 0.6 0.5
4.0 3.6 3.2 2.9 2.6 2.5 2.3 2.1 2.0 1.9 1.7 1.6 1.5
3.7 3.2 2.7 2.4 2.1 2.0 1.8 1.5 1.4 1.2 1.1 0.9 0.8
5.5 5.2 5.0 4.9 4.8 4.7 4.7 4.7 4.7 4.7 4.8 5.0 5.2
Women 25 30 35 40 45 50 55 60 65 70 75 80 84
15.9 18.8 21.8 26.4 31.0 40.9 52.9 61.5 69.5 80.1 89.1 92.8 94.9
1.7 1.5 1.3 1.2 1.0 0.79 0.59 0.48 0.39 0.28 0.20 0.16 0.13
3.0 2.8 2.6 2.4 2.2 1.9 1.6 1.5 1.3 1.2 1.1 1.1 1.1
2.6 2.3 2.1 1.8 1.6 1.3 0.93 0.76 0.61 0.45 0.31 0.25 0.21
4.9 4.8 4.8 4.7 4.7 4.7 5.0 5.3 5.7 6.7 8.4 9.8 11.1
Age (years)
a (A) for an individual at the threshold value for osteopenia (T score ⫽ ⫺1) compared with the whole population; (B) for all individuals with a value exceeding the threshold for osteopenia (T score ⱕ ⫺1) compared with the risk of the whole population of that age; (C) for an individual at the threshold value (T score ⫽ ⫺1) compared with an individual at the mean value for that age; and (D) the risk of those exceeding the threshold value (T score ⱕ ⫺1) compared with those under the diagnostic threshold (T score ⬎ ⫺1).
Relative riska
Percent of population (%)
A
B
Men 25 30 35 40 45 50 55 60 65 70 75 80 84
0.5 0.8 1.2 1.6 2.1 2.4 2.8 4.8 7.3 7.6 8.0 14.0 19.4
7.3 6.3 5.4 4.9 4.4 4.2 3.9 3.1 2.5 2.5 2.4 1.8 1.4
10.4 9.1 8.0 7.3 6.7 6.4 6.1 5.0 4.2 4.2 4.1 3.2 2.8
11.6 10.0 8.6 7.8 7.0 6.6 6.2 4.9 4.0 3.9 3.8 2.8 2.3
10.9 9.7 8.8 8.2 7.6 7.4 7.1 6.2 5.7 5.6 5.6 5.0 4.8
Women 25 30 35 40 45 50 55 60 65 70 75 80 84
0.6 1.0 1.5 2.4 3.7 5.4 8.1 12.3 17.7 24.5 33.8 43.3 51.0
6.9 5.9 5.1 4.2 3.5 2.9 2.4 1.9 1.5 1.2 0.95 0.74 0.62
9.8 8.6 7.5 6.4 5.5 4.8 4.1 3.4 2.9 2.5 2.1 1.8 1.6
10.9 9.3 8.0 6.6 5.5 4.6 3.8 3.0 2.4 1.9 1.5 1.2 0.98
10.4 9.3 8.3 7.3 6.6 6.1 5.6 5.1 4.9 4.7 4.7 4.8 4.9
C
D
a (A) for an individual at the threshold value for osteoporosis (T score ⫽ ⫺2.5) compared with the whole population at that age; (B) for all individuals with a value exceeding the threshold for osteoporosis (T score ⱕ ⫺2.5) compared with the risk of the whole population of that age; (C) for an individual at the threshold value (T score ⫽ ⫺2.5) compared with an individual at the mean value for that age; and (D) the risk of those exceeding the threshold value (T score ⱕ ⫺2.5) compared with those under the diagnostic threshold (T score ⬎ 2.5).
Discussion In this study we chose to adopt threshold values for low bone mass and osteoporosis using the age range 20 –29 years as the reference. More stringent criteria (i.e., including older premenopausal women) would have made the apparent frequency of osteoporosis much less common. The chosen threshold gave prevalence rates of osteoporosis comparable to those given by the WHO15,24 and corresponded to that chosen by the NHANES investigators.17 It is a matter of debate whether local reference ranges should be used or an international platform adopted. Ranges for BMD in Sweden and The Netherlands are comparable to the NHANES data,5 but vary in some other countries.11 Because hip fracture risk in different regions of the world varies much more than variations in BMD,1,6,8,19,22 it would seem appropriate to utilize the large and adequately sampled NHANES reference values until further research tempers this view.11 The results of our analysis would, however, differ markedly if different age ranges were used to determine threshold values for osteoporosis. With the use of the age range of 20 –29 years as a referent, the RR of hip fracture in a population of female osteoporotic patients aged 50 years was 4.8 compared with the risk of the general popula-
tion (Table 5, column B). Had we chosen the cutoff from women aged 50 years, the computed relative risk would have been 9.8 rather than 4.8 because of the difference in threshold value for BMD (data not shown). At many sites, BMD remains relatively constant from skeletal maturity up to the age of menopause. For example, lumbar BMD remains constant from the age of 20 years until menopause.2 In the NHANES data, values for BMD at the femoral neck are lower with each decade from the age of 20 years up to 50 years. The extent to which this is a cohort effect is unknown, but one of the consequences of adopting the reference age of 20 –29 years is that the prevalence of osteopenia is high even at the age of 50 years. If peak bone mass were constant up to the average age of menopause, then the expected frequency of osteopenia would be approximately 15%, whereas in this analysis it was three times higher. A consequence is that women with a T score of ⫺1 at the age of 50 years were at no greater risk than the average premenopausal population, although, in all women with low bone mass (T score ⬉ ⫺1 SD), the risk ratio at the age of 50 years was 1.9. This suggests that caution needs to be applied to this threshold if, for example, prevention were to be targeted to a
Bone Vol. 27, No. 5 November 2000:585–590
J. A. Kanis et al. Risk of hip fracture according to WHO
589
Figure 2. Population relative risks in Swedish women with osteoporosis, according to age. RR is shown for women at the threshold value for osteoporosis (T ⫽ ⫺2.5 SD) or those below the threshold (T ⱕ ⫺2.5 SD). Fracture risk is assumed to increase 2.6-fold/SD change in BMD ⫾ the 95% confidence estimate of the gradient.
minority of the postmenopausal population, rather than to the 50% of the population below the threshold for low bone mass. The present study emphasizes the importance of age as a risk factor for hip fracture. Relative risks were high in women with osteoporosis at the age of 50 years, although the absolute risk (e.g., 5 years) is low.9 If BMD were the sole determinant of hip fracture risk, then from the age-dependent decline in BMD of the female Swedish population the risk of hip fracture would be 3.8-fold higher in women aged 85 years compared with women aged 55 years. The increase in 1 year probability between these ages is, however, a 44-fold rise. Thus, over this age range, the impact of age is 11-fold greater than the impact of BMD. This illustrates problems inherent with the use of age-specific relative risks. First, age is a more important indicator of fracture probability than BMD. Thus, BMD alone can never be used to provide accurate assessments of fracture probability. A second and related problem is that relative risk decreases with age, but that absolute risk increases with age. This is likely to be confusing for clinicians who may assume that the need to treat the elderly is less, because the RR is less. A more satisfactory approach would be to express risk as an absolute risk, rather than an RR, a view endorsed by the International Osteoporosis Foundation.11 This will demand knowledge of the hazard function for fracture risk (the first fracture) at each of the common sites of fracture, and the hazard function of death over an appropriate period of time.14 There is no hard-and-fast rule concerning the most appropriate absolute risk to be used. Lifetime risks, although useful for counseling patients and assessing the burden of disease, are inappropriate for treatment thresholds, because treatments are rarely given over a lifetime. The optimal duration of treatment is not well established, but interventions of 3–5 years correspond to information available from clinical trials and models for costeffectiveness. For a number of treatments, effects appear to persist when treatment is stopped,9 so that 10 year risks are appropriate.11,14 A 10 year timeframe also corresponds to the long-term utility of BMD to predict fracture risk, which wanes with time.13 If RRs are to be used the present study draws attention to the need to choose the appropriate ones. In the field of osteoporosis, at least three different risk ratios are used.
First, the performance characteristics of densitometric techniques to predict fracture are commonly compared to the risk of individuals with an average BMD. For example, the risk of hip fracture changes 1.4 –2.6-fold for each SD of BMD from the mean value depending of the technique used.18 In this study, modeled on hip fracture risk from measurements at the hip, a gradient of risk of 2.6 was chosen. Because the reference material used to compute threshold values for loss of bone mass and osteoporosis was drawn from women aged 20 –29 years, women aged 25 years at the threshold osteoporosis (T score ⫽ ⫺2.5 SD) had an RR of hip fracture of 2.6 compared to women with average BMD (Table 4, column C). Second, from an epidemiological viewpoint, relative risks are usually presented using the risk of those with the disease compared with the risk in those without the disease. In the example given earlier, the risk of those with osteoporosis is increased 4.9rather than 2.6-fold (Table 4, column D). Third, in a clinical context, and from a societal perspective, is it appropriate to express risk of groups in individuals compared with the average risk of hip fracture. The average risk of hip fracture is, however, not equal to an individual with an average BMD. BMD is normally distributed, whereas hip fracture risk increases exponentially with decreasing BMD. Thus, the 50year-old woman with a T score of ⫺2.5 SD at the hip has an RR of 2.9 rather than 4.6 (Table 4, column A). In other words, expression of RR to average BMD overestimates that of the general population by slightly more than 50%. The appropriate population RR for decisionmaking is shown in Figure 2. In reality, the long-term risk is likely to be even lower, because bone loss occurs at variable rates in individuals.13 Few individuals have a BMD that lies at the threshold for osteoporosis. It is, therefore, often appropriate to express risk as the risk of individuals that lie within the diagnostic criterion relative to the risk of fracture of the whole population risk. Risk ratios for populations with osteoporosis or osteopenia are given in Table 4 (column B) and Figure 2. Thus, a randomly drawn population of women aged 25 years with a BMD value of less than ⫺1 SD would have an RR of 1.9, and an osteoporotic population of women at the same age an RR of 4.8. It should be noted that the estimates of risk that we provided
590
J. A. Kanis et al. Risk of hip fracture according to WHO
were modeled for hip fractures from measurements made at the hip with BMD using dual-energy X-ray absorptiometry (DXA). The risks computed would differ from those derived from measurements made with other techniques or the same at other sites. In addition, the T score identifies different populations and is not appropriate for techniques other than BMD with DXA.11 In view of these difficulties, it may be appropriate to express densitometric assessments in units of risk. In this study we chose RRs, but these can now be converted, where appropriate, to absolute (e.g., lifetime) risks.12 Several important conclusions arise from this work. First, the use of the threshold for osteopenia (a T score of ⫺1.0 SD) is associated with only modest increases in hip fracture risk compared with that of the general population. Indeed, in postmenopausal women at the age of 50 years at the threshold for osteopenia, the fracture risk is less than that of the general population of the same age. This is due to the choice of age range (20 –29 years) for the reference values of BMD on which to compute the T score. If this range is to be used, then it is inappropriate to use the threshold of osteopenia for clinical decisions, even in menopausal women. A second broad conclusion is that the use of age-specific RRs is problematic. It is important to choose or adjust RRs to the general population. Where this is done, the RR at each diagnostic threshold decreases with age, although the absolute risk (probability) of hip fracture increases. This apparent paradox is confusing for clinicians. Although a T score of ⫺2.5 SD or less is a reasonable treatment threshold,10 it is not sufficiently accurate to predict absolute risk, because this varies markedly with age. It will therefore become important to derive absolute (e.g., 10 year) risks to more accurately stratify risk of individuals and to thereby direct interventions.
Acknowledgments: The authors thank the Lilly Research Centre for their support of this work.
Bone Vol. 27, No. 5 November 2000:585–590
7. 8.
9.
10.
11.
12.
13.
14.
15. 16.
17.
18.
19. 20.
References 21. 1. Bacon, W. E., Maggi, S., Looker, A., Harris, T., Navi, C. R., Giaconi, J., Honkanen, R., Ho, S. C., Peffers, K. A., Torring, O., Gass, R., and Gonzalez, N. International comparison of hip fracture rates in 1988 –1989. Osteopor Int 6:69 –75; 1996. 2. Bonjour, J. P., and Rizzoli, R. Bone acquisition in adolescence. In: Marcus, R., Feldman, D., and Kelsey, J. Eds. Osteoporosis 1996. New York: Academic; 1996; 465– 476. 3. Committee for Proprietary Medicinal Products (CPMP). Note for Guidance on Involutional Osteoporosis in Women (CPMP/EWP/552/95). London: European Agency for the Evaluation of Medicinal Products; 1997. 4. De Laet, C. E. D. H., Van Hout, B. A., Burger, H., Hofman, A., and Pols, H. A. P. Bone density and risk of hip fracture in men and women: Cross sectional analysis. Br Med J 315:221–225; 1997. 5. De Laet, C. E. D. H., Van Hout, B. A., Burger, H., Hofman, A., Weel, A. E. A. M., and Pols, H. A. P. Hip fracture prediction in elderly men and women: Validation of the Rotterdam study. J Bone Miner Res 13:1587–1593; 1998. 6. Elffors, L., Allander, E., Kanis, J. A., Gullberg, B., Johnell, O., Dequeker, J., Dilsen, G., Gennari, C., Lopez Vaz, A. A., Lyritis, G., Mazzuoli, G. F.,
22.
23. 24.
25.
Miravet, L., Passeri, M., Perez Cano, R., Rapado, A., and Ribot, C. The variable incidence of hip fracture in southern Europe: The MEDOS study. Osteopor Int 4:253–263; 1994. Hui, S. L., Slemenda, C. W., and Johnston, C. C. Age and bone mass as predictors of fracture in a prospective study. J Clin Invest 81:1804 –1809; 1988. Johnell, O., Gullberg, B., Allander, E., and Kanis, J. A. The apparent incidence of hip fracture in Europe: A study of national register sources. Osteopor Int 2:298 –302; 1992. Jonsson, B., Kanis, J. A., Dawson, A., Oden, A., and Johnell, O. Effect and offset of effect of treatments for hip fracture on health outcomes. Osteopor Int 10:193–199; 1999. Kanis, J. A., Delmas, P., Burckhardt, P., Cooper, C., Torgerson, D., on behalf of the European Foundation for Osteoporosis and Bone Disease. Guidelines for diagnosis and management of osteoporosis. Osteopor Int 7:390 – 406; 1997. Kanis, J. A. and Glu¨er, C.-C., for the Committee of Scientific Advisors of the International Osteoporosis Foundation. An update on the diagnosis and assessment of osteoporosis with densitometry. Osteopor Int 11:192–202; 2000. Kanis, J. A., Johnell, O., Oden, A., Jonsson, B., Dawson, A., and Dere, W. Risk of hip fracture derived from relative risks: An analysis applied to the population of Sweden. Osteopor Int 11:120 –127; 2000. Kanis, J. A., Johnell, O., Oden, A., Jonsson, B., De Laet, C., and Dawson, A. Prediction of fracture from low bone mineral density measurements overestimates risk. Bone 26:387–391; 2000. Kanis, J. A., Johnell, O., Oden, A., Sernbo, I., Redlund-Johnell, I., Dawson, A., De Laet, C., and Jonsson, B. Long-term risk of osteoporotic fracture in Malmo. Osteopor Int. In press. Kanis, J. A., Melton, L. J., Christiansen, C., Johnston, C. C., and Khaltaev, N. The diagnosis of osteoporosis. J Bone Miner Res 9:1137–1141; 1994. Looker, A. C., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., Heyse, S. P., Johnston, C. C., and Lindsay, R. L. Proximal femur bone mineral level of US adults. Osteopor Int 5:389 – 409; 1995. Looker, A. C., Orwoll, E. S., Johnston, C. C., Lindsay, R. L., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., and Heyse, S. P. Prevalence of low femoral bone density in older US adults from NHANES III. J Bone Miner Res 12:1761–1768; 1997. Marshall, D., Johnell, O., and Weder, H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporosis fractures. Br Med J 312:1254 –1259; 1996. Melton, L. J. The prevalence of osteoporosis. J Bone Miner Res 12:1769 –1771; 1997. National Osteoporosis Foundation. Analyses of the effectiveness and cost of screening and treatment strategies for osteoporosis: A basis for development of practice guidelines. Osteopor Int 8(Suppl. 4):1– 88; 1998. Oden, A., Dawson, A., Dere, W., Johnell, O., Jonsson, B., and Kanis, J. A. Lifetime risk of hip fracture is underestimated. Osteopor Int 8:559 – 603. Ross, P. D., He, Y., Yates, A. J., Coupland, C., Ravn, P., McClung, M., Thompson, D., and Washich, R. I. Body size accounts for most differences in low density between Asian and Caucasian women. The EPIC Study Group. Calcif Tissue Int 59:339 –343. Royal College of Physicians. Clinical Guidelines for the Prevention and Treatment of Osteoporosis. London: RCP. World Health Organisation. Assessment of Fracture Risk and Its Application to Screening for Postmenopausal Osteoporosis. WHO Technical Report Series 843. Geneva: WHO; 1994. World Health Organisation. Guidelines for Preclinical Evaluation and Clinical Trials in Osteoporosis. Geneva: WHO; 1998.
Date Received: December 8, 1999 Date Revised: June 20, 2000 Date Accepted: June 20, 2000
Risk of Hip Fracture According to the World Health Organization Criteria for Osteopenia and Osteoporosis J. A. KANIS,1 O. JOHNELL,2 A. ODEN,3 B. JONSSON,4 C. DE LAET,5 and A. DAWSON6 1
WHO Collaborating Centre for Metabolic Bone Diseases, University of Sheffield Medical School, Sheffield, UK Department of Orthopaedics, Malmo General Hospital, Malmo, Sweden 3 Consulting Statistician, Gothenberg, Sweden 4 Department of Economics, Stockholm School of Economics, Stockholm, Sweden 5 Institute for Medical Technology Assessment, Rotterdam, The Netherlands 6 Lilly Research Centre, Ltd., Windelsham, Surrey, UK 2
Key Words: Hip fracture; Relative risk (RR); Population risk; Bone mineral density (BMD); Osteoporosis; Osteopenia.
The risk of hip fracture is commonly expressed as a relative risk. The aim of this study was to examine the utility of relative risks of hip fracture in men and women using World Health Organization (WHO) diagnostic criteria for low bone mass and osteoporosis. Reference data for bone mineral density (BMD) at the femoral neck, from the third National Health and Nutrition Examination Survey (NHANES III), were applied to the population of Sweden. Relative risks (RRs) were calculated from the known relationship between BMD at the femoral neck and hip fracture risk. The apparent prevalence of low bone mass and osteoporosis depended on the segment of the young population chosen for reference ranges. Using a reference derived from women aged 20 –29 years, the prevalence of osteoporosis was 21.2% in women between the ages of 50 and 84 years and 6.3% in men. The RRs associated with osteoporosis depended markedly on the risk comparison. For example, in men or women aged 50 years, the RR of hip fracture in those with osteoporosis compared to those without osteoporosis was 7.4 and 6.1, respectively. The RR of those at the threshold value for osteoporosis compared to those with an average value for BMD at that age was 6.6 and 4.6 in men and women, respectively. RRs were lower comparing those at the threshold value compared to the risk of the general population at that age (4.2 and 2.9, respectively). When RR was expressed in relation to the population risk rather than to the risk at the average value for BMD, RR decreased at all ages by 37%. Such adjustments are required for risk assessment in individuals and for the combined use of different risk factors. Because the average T score at each age decreased with age, the RR of hip fracture at any age decreased with advancing age in the presence of osteoporosis. The decrease in relative risk with age is, however, associated with an increase in absolute risk. Thus, for clinical use, the expression of absolute risks may be preferred to relative risks. (Bone 27: 585–590; 2000) © 2000 by Elsevier Science Inc. All rights reserved.
Introduction Age and bone mineral density (BMD) are the strongest risk factors for hip fracture thus far discovered. These risk factors are in part independent, so that absolute values for BMD have a different significance at different ages.7 Diagnostic criteria for osteoporosis (and low bone mass) have been proposed by an expert panel of the World Health Organization (WHO),15,24 and are widely used in epidemiology by regulatory agencies for practice guidelines and in clinical research.3,10,20,23,25 The diagnostic criteria stratify individuals according to the BMD of the young healthy population, so that osteoporosis in women is defined as a BMD value that lies 2.5 standard deviations (SD) or more below the average value in young healthy women (T score ⫽ ⫺2.5 SD). An advantage of using young normal reference ranges is that the prevalence of osteoporosis defined in this way increases exponentially with age in much the same way as does the frequency of many fractures due to osteoporosis. Some have advocated the use of the Z score rather than the T score. The Z score represents the number of SDs that a given value for BMD deviates from the average value for age and gender. Thus, an individual aged 60 years with a Z score of ⫺1 SD would lie in the lowest 16% of the distribution of BMD for that age, assuming a normal distribution of BMD. Similarly, an individual aged 80 years with a Z score of ⫺1 SD would lie at the same point within the normal distribution. For this reason, the use of the Z score to denote osteoporosis would mean that the prevalence of osteoporosis would not increase with age and is, therefore, inappropriate in a clinical setting. On the other hand, the use of the T score undervalues the importance of age in contributing to risk. For example, the significance of osteoporosis in a woman aged 50 years differs from that of a woman with the same BMD (and T score) who is aged 70 years. The relative risk is higher in the younger woman, although the absolute risk is lower, at least in the short term. The risk of hip fracture is commonly expressed as a relative risk (RR). RRs may be used by clinicians to make treatment decisions. The expression of RR is, however, not standardized. For example, many observational studies describe the risk of hip fracture as the increase in fracture risk per SD decrease in BMD. Epidemiological studies commonly report the RR in individuals
Address for correspondence and reprints: Dr. John A. Kanis, Centre for Metabolic Bone Diseases, WHO Collaborating Centre, University of Sheffield Medical School, Beech Hill Road, Sheffield S10 2RX, UK. E-mail: [email protected] © 2000 by Elsevier Science Inc. All rights reserved.
585
8756-3282/00/$20.00 PII S8756-3282(00)00381-1
586
J. A. Kanis et al. Risk of hip fracture according to WHO
Bone Vol. 27, No. 5 November 2000:585–590
Table 1. Prevalence (%) of low bone mass and osteoporosis at the ages specified utilizing different ages to derive cutoff values Mena
Low bone mass 50 60 70 80 Osteoporosis 50 60 70 80
Menb
Women
20–29 yr
20–39 yr
50 yr
20–29 yr
20–39 yr
20–49
50 yr
20–29 yr
20–49 yr
50 yr
38.5 47.1 56.6 69.3
32.1 40.6 50.2 63.4
16.0 23.0 30.9 43.7
40.9 61.5 80.1 92.8
36.7 57.0 74.5 90.8
31.5 51.1 71.3 87.7
16.0 30.3 49.6 71.0
24.2 32.4 41.4 54.8
17.5 24.9 33.0 46.0
7.9 12.9 18.6 29.2
2.8 5.5 8.6 15.4
1.9 3.4 6.4 12.6
0.6 1.6 2.7 5.8
5.4 12.3 24.5 43.3
3.9 9.2 19.3 36.1
2.5 6.1 13.7 27.5
0.6 1.8 4.6 11.1
2.4 4.8 7.6 14.0
1.0 2.4 4.0 8.1
0.3 0.7 1.3 3.1
a Prevalences derived from NHANES male reference ranges for bone mineral density. Thresholds: 20 –29 years, 0.792 g/cm2; 20 –39, 0.770; 50, 0.702 for osteoporosis and 0.585, 0.564, and 0.510, respectively, for low bone mass. b Prevalences derived from NHANES female reference ranges. Thresholds: age 20 –29 years, 0.740 g/cm2; 20 –39, 0.729; 20 – 49, 0.710; 50, 0.649 for osteoporosis and 0.577, 0.556, 0.533, and 0.407 for low bone mass, respectively.
with a risk factor compared to those without the factor; for example, the RR of hip fracture in women with or without osteoporosis. Others describe the risk of fracture compared with an average; for example, the risk of hip fracture based on BMD compared to an individual with an average BMD. All these approaches give differing RRs. For clinical purposes it is appropriate to describe the risk of hip fracture in an individual (or a group of individuals) compared with the average risk of hip fracture (not average BMD). The aims of this study were to explore the relationships between BMD, age, and RR of hip fracture using the WHO thresholds for osteoporosis and low bone mass (osteopenia). A further aim was to derive RRs appropriate for clinical rather than epidemiological use.
bone mass was 0.740 g/cm2 and 0.577 for osteoporosis. This accords with the proposals recommended by the International Committee for Standards in Bone Densitometry and those used by the NHANES investigators.17 Unlike Looker et al.,17 we chose the same thresholds for men because the risk of hip fracture in men is similar to that in women for the same BMD.3,4 The use of the term “below a threshold” for osteoporosis denotes a BMD value that is less than the threshold value (i.e., a more negative T score). The risk of hip fracture was assumed to increase 2.6-fold (95% confidence interval 2.0 –3.5) for each SD decrease in BMD at the femoral neck, in line with a recent metaanalysis.18 The risk of hip fracture was taken from the National Register Source for Sweden 1994.21 Relative risk by age was expressed in several ways (Figure 1):
Materials and Methods
A. The risk of an individual at the threshold value compared with the hip fracture risk of the total population of that age. This is the RR most appropriate for clinical use.
We used the published data from the third National Health and Nutrition Examination Survey (NHANES III, 1988 –1994) as the reference values for BMD at the femoral neck.16 Threshold values to categorize individuals as having low bone mass or osteoporosis utilized the WHO criteria so that osteoporosis was defined as a T score of less than ⫺2.5 SD. Low bone mass denoted a T score of ⫺1 to ⫺2.5 SD.24 Threshold values were chosen in young women based on the age ranges 20 –29 years, 20 –39 years, 20 – 49 years, and at the age of 50 years, and applied to the male and the female population of Sweden, 1994. In addition, a threshold derived for men at the age of 50 years was utilized. The prevalence of osteoporosis and low bone mass was markedly dependent on the age range chosen to determine threshold values. At the age of 80 years, prevalences varied fourfold in women and nearly threefold in men (Table 1). The use of reference ranges that included successively more elderly women at or below the age of 50 years decreased the apparent prevalence of osteoporosis. For example, where reference ranges and diagnostic thresholds were calculated at the age of 50 years, the prevalence of osteoporosis in the Swedish female population was only 4.3%, and increased with the inclusion of younger women. The use of the age range 20 –29 years as a reference range gave a prevalence of osteoporosis of 21% in women aged 50 – 84 years. This corresponds broadly with the original WHO estimates applied to the UK and USA.15,24 The value of 21% approximates the lifetime risk of hip fracture in Sweden.21 For the purposes of this study we used the age range 20 –29 years for deriving diagnostic thresholds. The threshold for low
Figure 1. Different methods of expressing RR in individuals or populations. The diagram shows the normal distribution of BMD in elderly women (aged approximately 65 years). The average BMD is 1 SD below the mean value for young healthy women and approximately 16% have osteoporosis (T score ⱕ ⫺2.5 SD). RR of hip fracture may be expressed as: (A.) that of an individual at the threshold value for osteoporosis compared with that of the general population; (B.) that of all women with osteoporosis compared with that of the general population; (C.) that of a woman at the threshold value for osteoporosis compared with a woman with an average BMD for that age; and (D.) that of osteoporotic women compared with those without osteoporosis.
Bone Vol. 27, No. 5 November 2000:585–590
J. A. Kanis et al. Risk of hip fracture according to WHO
B. The risk of individuals below the threshold value compared with the risk of hip fracture of the total population of that age. This RR is appropriate for considering a group of individuals (e.g., all those with osteoporosis). C. The risk of an individual at the threshold value compared to an individual with a BMD at the mean value at the same age. This is the RR usually used to describe the performance characteristics of densitometry. D. The risk of those below the threshold value vs. those above the threshold value. This is the most common expression of risk in epidemiological studies. The risk of fracture in the short perspective (i.e., not accounting for deaths) of those with BMD below the threshold for BMD (g) at a certain age is: Yearly incidence of the age group ⫻ ⌽(( g ⫺ )/ ⫹ log(RR))/⌽(( g ⫺ )/)
Table 2. Prevalence of osteoporosis at the age intervals shown in Sweden using female-derived threshold values from the young population, aged 20 –29 years (bone mineral density threshold ⫽ 0.577 g/cm2) Men
Women
Age range (years)
Percent of population
Number affected (000)
Percent of population
Number affected (000)
50–54 55–59 60–64 65–69 70–74 75–79 80–84 50–84
2.5 3.5 5.8 7.4 7.8 10.3 16.6 6.3
7.0 7.6 11.4 14.2 14.6 13.7 14.7 83.2
6.3 9.6 14.3 20.2 27.9 37.5 47.2 21.2
17.0 21.1 30.0 43.7 63.0 68.3 67.8 310.9
(1)
where is the mean and is the SD of BMD at the current age and log(RR) is the e-log of the risk ratio of an individual with 1-SD-lower BMD as compared with another (RR estimated to be 1.4 –2.6, depending on type of fracture and type of BMD measurement). ⌽ is the normal distribution function with mean ⫽ 0 and SD ⫽ 1. For individuals with BMD ⫽ x, the risk is: Yearly incidence of the age group ⫻ exp(⫺log(RR)( ⫺ )/ ⫺ (log(RR)) 2/ 2)
587
(2)
If the mean, , of BMD at a certain age and the standard deviation, , are known, we can calculate the risk ratios vs. the normal population for: (a) those below a threshold g of BMD; and (b) those at a value x of BMD. The confidence estimate of the gradient of risk/SD change in BMD can be used for sensitivity analyses to test differing assumptions concerning the gradient of risk. All relative risks shown can be multiplied by 0.77 or 1.35 to accomodate the range of the confidence intervals. For brevity, this is not done here. Results The prevalence of osteoporosis in the Swedish population increased as expected with age. Approximately 15% of the male and 50% of the female population aged 80 – 84 years had osteoporosis (Table 2). Approximately 6% of men and 21% of women aged 50 – 84 years were classified as having osteoporosis. The prevalence of low bone mass was, as expected, higher than that of osteoporosis at all ages, but did not increase markedly with age (Table 3). Thus, the ratio of individuals with low bone mass to those with osteoporosis varied with age. For example, in women aged 50 –54 years, the number of individuals with low bone mass was sixfold higher than the number with osteoporosis. In the age range 80 – 84 years, the number with either diagnosis was nearly equal. The risk of hip fracture by age, gender, and diagnostic category is shown in Tables 4 and 5. In women aged 25 years with a T score of ⫺1 SD, the relative risk of hip fracture was 2.6 compared to women with average BMD, as expected from the choice of reference material (i.e., age 20 –29 years) (Table 4, column C). The relative risk decreased with age because the mean BMD decreased progressively with age. Indeed, at the age of 55 years, greater than half the female population had osteopenia or osteoporosis; therefore, at this age and older, a T score of ⫺1 was protective (RR ⬍ 1.0).
The risk in women with a T score of ⫺1 (osteopenia) compared with the whole population risk was lower (Table 4, column A) by 37% at all ages. Thus, at the age of 50 years, women with a T score of ⫺1 had a lower risk than the general population of the same age. In all women who fell below the ⫺1 SD threshold, the risk at this age compared with the risk of the whole population of the same age (column B) was 1.9 and remained above unity at all ages. RRs were higher when those below the threshold were compared with those above the threshold (column D) and did not fall with age. This was because the proportion of women with osteopenia at any given age did not decrease with age (see Table 3). A similar phenomenon was observed for osteoporosis in women (Table 5). Thus, at the age of 25 years, a woman at the threshold for osteoporosis (T score ⫽ ⫺2.5) had a 10.9-foldhigher risk of hip fracture than women with an average BMD (column C). The RR decreased with age and was close to unity at the age of 80 years, when approximately 50% of the female population has osteoporosis. In men, expressions of RR were qualitatively, although not quantitatively, similar to findings in women. The RRs of hip fracture compared with the population risk of fracture is shown for women with osteoporosis in Figure 2. As expected, the RR was higher in all osteoporotic women than women at the threshold value and, in both, the RR decreased with advancing age. Table 3. Prevalence of low bone mass at the age intervals shown using female-derived threshold values of BMD from the young population, aged 20 –29 years (bone mineral density between the thresholds of 0.740 g/cm2) Men
Women
Age range (years)
Percent of population
Number affected (000)
Percent of population
Number affected (000)
50–54 55–59 60–64 65–69 70–74 75–79 80–84 50–84
23.0 (25.5) 26.0 (29.5) 28.4 (34.2) 31.0 (38.4) 35.7 (43.5) 40.1 (50.4) 40.9 (57.5) 30.4 (36.7)
66.4 (71.6) 57.0 (64.7) 55.8 (67.2) 59.4 (73.6) 66.6 (81.2) 53.4 (67.1) 36.2 (50.5) 394.8 (475.9)
39.1 (45.4) 46.8 (56.4) 50.5 (64.8) 53.6 (73.8) 56.1 (84.0) 53.2 (90.7) 46.7 (93.9) 49.1 (70.3)
105.7 (122.7) 103.1 (124.2) 106.0 (136.0) 115.9 (159.6) 126.7 (189.7) 96.9 (165.2) 67.1 (134.9) 721.3 (1032.2)
Values in parentheses include all individuals below the threshold for low bone mass and therefore include those with osteoporosis.
588
J. A. Kanis et al. Risk of hip fracture according to WHO
Bone Vol. 27, No. 5 November 2000:585–590
Table 4. Proportion of the male and female population at the ages shown to have low bone mass (osteopenia) and the risk of hip fracture. (Relative risk is expressed for each age, as shown in the footnote below.) Relative riska
Table 5. Proportion of the male and female population at the ages shown identified as having osteoporosis and the risk of hip fracture. (Relative risk shown for each age, as shown in the footnote below.)
Percent of population (%)
A
B
C
D
Age (years)
Men 25 30 35 40 45 50 55 60 65 70 75 80 84
8.4 11.2 14.5 17.6 21.3 24.2 27.5 32.4 36.7 41.4 47.1 54.8 59.9
2.4 2.0 1.7 1.5 1.4 1.2 1.1 1.0 0.9 0.8 0.7 0.6 0.5
4.0 3.6 3.2 2.9 2.6 2.5 2.3 2.1 2.0 1.9 1.7 1.6 1.5
3.7 3.2 2.7 2.4 2.1 2.0 1.8 1.5 1.4 1.2 1.1 0.9 0.8
5.5 5.2 5.0 4.9 4.8 4.7 4.7 4.7 4.7 4.7 4.8 5.0 5.2
Women 25 30 35 40 45 50 55 60 65 70 75 80 84
15.9 18.8 21.8 26.4 31.0 40.9 52.9 61.5 69.5 80.1 89.1 92.8 94.9
1.7 1.5 1.3 1.2 1.0 0.79 0.59 0.48 0.39 0.28 0.20 0.16 0.13
3.0 2.8 2.6 2.4 2.2 1.9 1.6 1.5 1.3 1.2 1.1 1.1 1.1
2.6 2.3 2.1 1.8 1.6 1.3 0.93 0.76 0.61 0.45 0.31 0.25 0.21
4.9 4.8 4.8 4.7 4.7 4.7 5.0 5.3 5.7 6.7 8.4 9.8 11.1
Age (years)
a (A) for an individual at the threshold value for osteopenia (T score ⫽ ⫺1) compared with the whole population; (B) for all individuals with a value exceeding the threshold for osteopenia (T score ⱕ ⫺1) compared with the risk of the whole population of that age; (C) for an individual at the threshold value (T score ⫽ ⫺1) compared with an individual at the mean value for that age; and (D) the risk of those exceeding the threshold value (T score ⱕ ⫺1) compared with those under the diagnostic threshold (T score ⬎ ⫺1).
Relative riska
Percent of population (%)
A
B
Men 25 30 35 40 45 50 55 60 65 70 75 80 84
0.5 0.8 1.2 1.6 2.1 2.4 2.8 4.8 7.3 7.6 8.0 14.0 19.4
7.3 6.3 5.4 4.9 4.4 4.2 3.9 3.1 2.5 2.5 2.4 1.8 1.4
10.4 9.1 8.0 7.3 6.7 6.4 6.1 5.0 4.2 4.2 4.1 3.2 2.8
11.6 10.0 8.6 7.8 7.0 6.6 6.2 4.9 4.0 3.9 3.8 2.8 2.3
10.9 9.7 8.8 8.2 7.6 7.4 7.1 6.2 5.7 5.6 5.6 5.0 4.8
Women 25 30 35 40 45 50 55 60 65 70 75 80 84
0.6 1.0 1.5 2.4 3.7 5.4 8.1 12.3 17.7 24.5 33.8 43.3 51.0
6.9 5.9 5.1 4.2 3.5 2.9 2.4 1.9 1.5 1.2 0.95 0.74 0.62
9.8 8.6 7.5 6.4 5.5 4.8 4.1 3.4 2.9 2.5 2.1 1.8 1.6
10.9 9.3 8.0 6.6 5.5 4.6 3.8 3.0 2.4 1.9 1.5 1.2 0.98
10.4 9.3 8.3 7.3 6.6 6.1 5.6 5.1 4.9 4.7 4.7 4.8 4.9
C
D
a (A) for an individual at the threshold value for osteoporosis (T score ⫽ ⫺2.5) compared with the whole population at that age; (B) for all individuals with a value exceeding the threshold for osteoporosis (T score ⱕ ⫺2.5) compared with the risk of the whole population of that age; (C) for an individual at the threshold value (T score ⫽ ⫺2.5) compared with an individual at the mean value for that age; and (D) the risk of those exceeding the threshold value (T score ⱕ ⫺2.5) compared with those under the diagnostic threshold (T score ⬎ 2.5).
Discussion In this study we chose to adopt threshold values for low bone mass and osteoporosis using the age range 20 –29 years as the reference. More stringent criteria (i.e., including older premenopausal women) would have made the apparent frequency of osteoporosis much less common. The chosen threshold gave prevalence rates of osteoporosis comparable to those given by the WHO15,24 and corresponded to that chosen by the NHANES investigators.17 It is a matter of debate whether local reference ranges should be used or an international platform adopted. Ranges for BMD in Sweden and The Netherlands are comparable to the NHANES data,5 but vary in some other countries.11 Because hip fracture risk in different regions of the world varies much more than variations in BMD,1,6,8,19,22 it would seem appropriate to utilize the large and adequately sampled NHANES reference values until further research tempers this view.11 The results of our analysis would, however, differ markedly if different age ranges were used to determine threshold values for osteoporosis. With the use of the age range of 20 –29 years as a referent, the RR of hip fracture in a population of female osteoporotic patients aged 50 years was 4.8 compared with the risk of the general popula-
tion (Table 5, column B). Had we chosen the cutoff from women aged 50 years, the computed relative risk would have been 9.8 rather than 4.8 because of the difference in threshold value for BMD (data not shown). At many sites, BMD remains relatively constant from skeletal maturity up to the age of menopause. For example, lumbar BMD remains constant from the age of 20 years until menopause.2 In the NHANES data, values for BMD at the femoral neck are lower with each decade from the age of 20 years up to 50 years. The extent to which this is a cohort effect is unknown, but one of the consequences of adopting the reference age of 20 –29 years is that the prevalence of osteopenia is high even at the age of 50 years. If peak bone mass were constant up to the average age of menopause, then the expected frequency of osteopenia would be approximately 15%, whereas in this analysis it was three times higher. A consequence is that women with a T score of ⫺1 at the age of 50 years were at no greater risk than the average premenopausal population, although, in all women with low bone mass (T score ⬉ ⫺1 SD), the risk ratio at the age of 50 years was 1.9. This suggests that caution needs to be applied to this threshold if, for example, prevention were to be targeted to a
Bone Vol. 27, No. 5 November 2000:585–590
J. A. Kanis et al. Risk of hip fracture according to WHO
589
Figure 2. Population relative risks in Swedish women with osteoporosis, according to age. RR is shown for women at the threshold value for osteoporosis (T ⫽ ⫺2.5 SD) or those below the threshold (T ⱕ ⫺2.5 SD). Fracture risk is assumed to increase 2.6-fold/SD change in BMD ⫾ the 95% confidence estimate of the gradient.
minority of the postmenopausal population, rather than to the 50% of the population below the threshold for low bone mass. The present study emphasizes the importance of age as a risk factor for hip fracture. Relative risks were high in women with osteoporosis at the age of 50 years, although the absolute risk (e.g., 5 years) is low.9 If BMD were the sole determinant of hip fracture risk, then from the age-dependent decline in BMD of the female Swedish population the risk of hip fracture would be 3.8-fold higher in women aged 85 years compared with women aged 55 years. The increase in 1 year probability between these ages is, however, a 44-fold rise. Thus, over this age range, the impact of age is 11-fold greater than the impact of BMD. This illustrates problems inherent with the use of age-specific relative risks. First, age is a more important indicator of fracture probability than BMD. Thus, BMD alone can never be used to provide accurate assessments of fracture probability. A second and related problem is that relative risk decreases with age, but that absolute risk increases with age. This is likely to be confusing for clinicians who may assume that the need to treat the elderly is less, because the RR is less. A more satisfactory approach would be to express risk as an absolute risk, rather than an RR, a view endorsed by the International Osteoporosis Foundation.11 This will demand knowledge of the hazard function for fracture risk (the first fracture) at each of the common sites of fracture, and the hazard function of death over an appropriate period of time.14 There is no hard-and-fast rule concerning the most appropriate absolute risk to be used. Lifetime risks, although useful for counseling patients and assessing the burden of disease, are inappropriate for treatment thresholds, because treatments are rarely given over a lifetime. The optimal duration of treatment is not well established, but interventions of 3–5 years correspond to information available from clinical trials and models for costeffectiveness. For a number of treatments, effects appear to persist when treatment is stopped,9 so that 10 year risks are appropriate.11,14 A 10 year timeframe also corresponds to the long-term utility of BMD to predict fracture risk, which wanes with time.13 If RRs are to be used the present study draws attention to the need to choose the appropriate ones. In the field of osteoporosis, at least three different risk ratios are used.
First, the performance characteristics of densitometric techniques to predict fracture are commonly compared to the risk of individuals with an average BMD. For example, the risk of hip fracture changes 1.4 –2.6-fold for each SD of BMD from the mean value depending of the technique used.18 In this study, modeled on hip fracture risk from measurements at the hip, a gradient of risk of 2.6 was chosen. Because the reference material used to compute threshold values for loss of bone mass and osteoporosis was drawn from women aged 20 –29 years, women aged 25 years at the threshold osteoporosis (T score ⫽ ⫺2.5 SD) had an RR of hip fracture of 2.6 compared to women with average BMD (Table 4, column C). Second, from an epidemiological viewpoint, relative risks are usually presented using the risk of those with the disease compared with the risk in those without the disease. In the example given earlier, the risk of those with osteoporosis is increased 4.9rather than 2.6-fold (Table 4, column D). Third, in a clinical context, and from a societal perspective, is it appropriate to express risk of groups in individuals compared with the average risk of hip fracture. The average risk of hip fracture is, however, not equal to an individual with an average BMD. BMD is normally distributed, whereas hip fracture risk increases exponentially with decreasing BMD. Thus, the 50year-old woman with a T score of ⫺2.5 SD at the hip has an RR of 2.9 rather than 4.6 (Table 4, column A). In other words, expression of RR to average BMD overestimates that of the general population by slightly more than 50%. The appropriate population RR for decisionmaking is shown in Figure 2. In reality, the long-term risk is likely to be even lower, because bone loss occurs at variable rates in individuals.13 Few individuals have a BMD that lies at the threshold for osteoporosis. It is, therefore, often appropriate to express risk as the risk of individuals that lie within the diagnostic criterion relative to the risk of fracture of the whole population risk. Risk ratios for populations with osteoporosis or osteopenia are given in Table 4 (column B) and Figure 2. Thus, a randomly drawn population of women aged 25 years with a BMD value of less than ⫺1 SD would have an RR of 1.9, and an osteoporotic population of women at the same age an RR of 4.8. It should be noted that the estimates of risk that we provided
590
J. A. Kanis et al. Risk of hip fracture according to WHO
were modeled for hip fractures from measurements made at the hip with BMD using dual-energy X-ray absorptiometry (DXA). The risks computed would differ from those derived from measurements made with other techniques or the same at other sites. In addition, the T score identifies different populations and is not appropriate for techniques other than BMD with DXA.11 In view of these difficulties, it may be appropriate to express densitometric assessments in units of risk. In this study we chose RRs, but these can now be converted, where appropriate, to absolute (e.g., lifetime) risks.12 Several important conclusions arise from this work. First, the use of the threshold for osteopenia (a T score of ⫺1.0 SD) is associated with only modest increases in hip fracture risk compared with that of the general population. Indeed, in postmenopausal women at the age of 50 years at the threshold for osteopenia, the fracture risk is less than that of the general population of the same age. This is due to the choice of age range (20 –29 years) for the reference values of BMD on which to compute the T score. If this range is to be used, then it is inappropriate to use the threshold of osteopenia for clinical decisions, even in menopausal women. A second broad conclusion is that the use of age-specific RRs is problematic. It is important to choose or adjust RRs to the general population. Where this is done, the RR at each diagnostic threshold decreases with age, although the absolute risk (probability) of hip fracture increases. This apparent paradox is confusing for clinicians. Although a T score of ⫺2.5 SD or less is a reasonable treatment threshold,10 it is not sufficiently accurate to predict absolute risk, because this varies markedly with age. It will therefore become important to derive absolute (e.g., 10 year) risks to more accurately stratify risk of individuals and to thereby direct interventions.
Acknowledgments: The authors thank the Lilly Research Centre for their support of this work.
Bone Vol. 27, No. 5 November 2000:585–590
7. 8.
9.
10.
11.
12.
13.
14.
15. 16.
17.
18.
19. 20.
References 21. 1. Bacon, W. E., Maggi, S., Looker, A., Harris, T., Navi, C. R., Giaconi, J., Honkanen, R., Ho, S. C., Peffers, K. A., Torring, O., Gass, R., and Gonzalez, N. International comparison of hip fracture rates in 1988 –1989. Osteopor Int 6:69 –75; 1996. 2. Bonjour, J. P., and Rizzoli, R. Bone acquisition in adolescence. In: Marcus, R., Feldman, D., and Kelsey, J. Eds. Osteoporosis 1996. New York: Academic; 1996; 465– 476. 3. Committee for Proprietary Medicinal Products (CPMP). Note for Guidance on Involutional Osteoporosis in Women (CPMP/EWP/552/95). London: European Agency for the Evaluation of Medicinal Products; 1997. 4. De Laet, C. E. D. H., Van Hout, B. A., Burger, H., Hofman, A., and Pols, H. A. P. Bone density and risk of hip fracture in men and women: Cross sectional analysis. Br Med J 315:221–225; 1997. 5. De Laet, C. E. D. H., Van Hout, B. A., Burger, H., Hofman, A., Weel, A. E. A. M., and Pols, H. A. P. Hip fracture prediction in elderly men and women: Validation of the Rotterdam study. J Bone Miner Res 13:1587–1593; 1998. 6. Elffors, L., Allander, E., Kanis, J. A., Gullberg, B., Johnell, O., Dequeker, J., Dilsen, G., Gennari, C., Lopez Vaz, A. A., Lyritis, G., Mazzuoli, G. F.,
22.
23. 24.
25.
Miravet, L., Passeri, M., Perez Cano, R., Rapado, A., and Ribot, C. The variable incidence of hip fracture in southern Europe: The MEDOS study. Osteopor Int 4:253–263; 1994. Hui, S. L., Slemenda, C. W., and Johnston, C. C. Age and bone mass as predictors of fracture in a prospective study. J Clin Invest 81:1804 –1809; 1988. Johnell, O., Gullberg, B., Allander, E., and Kanis, J. A. The apparent incidence of hip fracture in Europe: A study of national register sources. Osteopor Int 2:298 –302; 1992. Jonsson, B., Kanis, J. A., Dawson, A., Oden, A., and Johnell, O. Effect and offset of effect of treatments for hip fracture on health outcomes. Osteopor Int 10:193–199; 1999. Kanis, J. A., Delmas, P., Burckhardt, P., Cooper, C., Torgerson, D., on behalf of the European Foundation for Osteoporosis and Bone Disease. Guidelines for diagnosis and management of osteoporosis. Osteopor Int 7:390 – 406; 1997. Kanis, J. A. and Glu¨er, C.-C., for the Committee of Scientific Advisors of the International Osteoporosis Foundation. An update on the diagnosis and assessment of osteoporosis with densitometry. Osteopor Int 11:192–202; 2000. Kanis, J. A., Johnell, O., Oden, A., Jonsson, B., Dawson, A., and Dere, W. Risk of hip fracture derived from relative risks: An analysis applied to the population of Sweden. Osteopor Int 11:120 –127; 2000. Kanis, J. A., Johnell, O., Oden, A., Jonsson, B., De Laet, C., and Dawson, A. Prediction of fracture from low bone mineral density measurements overestimates risk. Bone 26:387–391; 2000. Kanis, J. A., Johnell, O., Oden, A., Sernbo, I., Redlund-Johnell, I., Dawson, A., De Laet, C., and Jonsson, B. Long-term risk of osteoporotic fracture in Malmo. Osteopor Int. In press. Kanis, J. A., Melton, L. J., Christiansen, C., Johnston, C. C., and Khaltaev, N. The diagnosis of osteoporosis. J Bone Miner Res 9:1137–1141; 1994. Looker, A. C., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., Heyse, S. P., Johnston, C. C., and Lindsay, R. L. Proximal femur bone mineral level of US adults. Osteopor Int 5:389 – 409; 1995. Looker, A. C., Orwoll, E. S., Johnston, C. C., Lindsay, R. L., Wahner, H. W., Dunn, W. L., Calvo, M. S., Harris, T. B., and Heyse, S. P. Prevalence of low femoral bone density in older US adults from NHANES III. J Bone Miner Res 12:1761–1768; 1997. Marshall, D., Johnell, O., and Weder, H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporosis fractures. Br Med J 312:1254 –1259; 1996. Melton, L. J. The prevalence of osteoporosis. J Bone Miner Res 12:1769 –1771; 1997. National Osteoporosis Foundation. Analyses of the effectiveness and cost of screening and treatment strategies for osteoporosis: A basis for development of practice guidelines. Osteopor Int 8(Suppl. 4):1– 88; 1998. Oden, A., Dawson, A., Dere, W., Johnell, O., Jonsson, B., and Kanis, J. A. Lifetime risk of hip fracture is underestimated. Osteopor Int 8:559 – 603. Ross, P. D., He, Y., Yates, A. J., Coupland, C., Ravn, P., McClung, M., Thompson, D., and Washich, R. I. Body size accounts for most differences in low density between Asian and Caucasian women. The EPIC Study Group. Calcif Tissue Int 59:339 –343. Royal College of Physicians. Clinical Guidelines for the Prevention and Treatment of Osteoporosis. London: RCP. World Health Organisation. Assessment of Fracture Risk and Its Application to Screening for Postmenopausal Osteoporosis. WHO Technical Report Series 843. Geneva: WHO; 1994. World Health Organisation. Guidelines for Preclinical Evaluation and Clinical Trials in Osteoporosis. Geneva: WHO; 1998.
Date Received: December 8, 1999 Date Revised: June 20, 2000 Date Accepted: June 20, 2000