Osteoporos Int (2009) 20:363–370 DOI 10.1007/s00198-008-0688-x

ORIGINAL ARTICLE

Weight and body mass index predict bone mineral density and fractures in women aged 40 to 59 years S. Morin & J. F. Tsang & W. D. Leslie

Received: 13 March 2008 / Accepted: 29 May 2008 / Published online: 17 July 2008 # International Osteoporosis Foundation and National Osteoporosis Foundation 2008

Abstract Summary Weight and body mass index are associated with low bone mineral density and fractures in older women. This retrospective cohort study confirms a similar relationship in women aged 40 to 59 years. Introduction Risk factors for the prediction of osteoporosis and fractures have been less thoroughly studied in younger women. We evaluated the associations between weight, body mass index (BMI), the Osteoporosis Self-Assessment Tool (OST), bone mineral density (BMD) and fracture risk in women aged 40 to 59 years. Methods Using administrative health management databases, we conducted a retrospective cohort study in 8,254 women aged 40–59 years who had baseline BMD testing. Linear regression and Cox proportional multivariate models were created to examine the associations with weight, BMI, OST, BMD, and subsequent fractures throughout a 3.3-year follow-up. Results Body weight, BMI, and OST had a similar overall performance in their ability to classify women with femoral neck T-score ≤ −2.5. Throughout 27,256 person years of On behalf of the Manitoba Bone Density Program. S. Morin Department of Medicine, McGill University, Montreal, Canada J. F. Tsang : W. D. Leslie Faculty of Medicine, University of Manitoba, Winnipeg, Canada S. Morin (*) Division of General Internal Medicine, McGill University Health Center (MUHC), 1650 Cedar Avenue, Room B2–118, Montreal, QC H3G 1A4, Canada e-mail: [email protected]

observation, 225 women experienced one or more fractures. After adjustment for age, prevalent fractures, and use of corticosteroids, each standard deviation decrease in weight was associated with a 19% increase in the risk of incident fracture (95% CI: 1.01–1.35). Femoral neck BMD and the presence of prevalent fractures were also associated with the risk of incident fractures. Conclusions Low weight and BMI predict osteoporosis and are associated with increased fracture risk in younger women. The negative impact of low body weight on bone health should be more widely recognized. Keywords Body mass index . Body weight . Bone mineral density . Osteoporosis fractures . Younger women

Introduction Osteoporosis is characterized by low bone mass and an increased risk of fracture [1]. Fractures most commonly associated with osteoporosis are those of the hip, the vertebrae, and the wrist, and these are responsible for morbidity and excess mortality. Many clinical guidelines recommend risk factor assessment and measurement of bone mineral density (BMD) through dual energy X-ray absorptiometry (DXA) to identify individuals at high risk of fracture [2–4]. Risk factors have been extensively characterized in women over the age of 65 years and are used in practice, often in combination, to predict fractures [5–8]. However, the application of these algorithms in younger women has not been well studied. Body weight is positively associated with BMD, from childhood through adulthood (with correlations to the order of 0.3 to 0.6) [9]. This relationship is known to be stronger

364

in older women in whom both weight and body mass index (BMI) have been shown to explain an large proportion of the variance in BMD (8.9–19.8% of total variance) [9, 10]. There are currently insufficient data on the predictive value of body weight and BMI in younger women. Low weight and low BMI are also related to an increased fracture risk. In a large meta-analysis of 12 prospective population-based cohorts, De Laet et al. documented that the age-adjusted risk of a hip fracture was increased 2-fold in older individuals with a BMI of 20 kg/m2 compared with a BMI of 25 kg/m2 [11]. In the Study of Osteoporosis Fractures (SOF), women aged 65 years and older in the lowest quartile for weight had twice as many hip fractures as those in the other quartiles [5]. Similar data exist for the risk of vertebral and forearm fractures in older women [12–14]. The purpose of this study was to investigate the associations between body weight, BMI, BMD and subsequent fractures in women aged 40 to 59 years to guide clinicians in the evaluation of fracture risk in this population.

Osteoporos Int (2009) 20:363–370

contacts. A province-wide pharmacy database provides information on all prescription medications dispensed to out-patients. The Manitoba Bone Density Program database can be linked with these provincial computerized health databases through an anonymous personal identifier. The accuracy of these databases has been well validated [17, 18]. We identified all women aged 40 to 59 years with a first bone density measurement. Women were required to have results for the lumbar spine (L1–L4) and proximal femur (total hip, femoral neck and trochanter sites) prior to 31 October 2002 and medical coverage from Manitoba Health during the observation period ending 31 March 2004. As earlier software versions did not provide total hip measurements, records prior to May 1998 were not used in the analysis. For women with more than one eligible set of measurements, only the first record was included. The study was approved by the Research Ethics Board for the University of Manitoba and the Health Information Privacy committee of Manitoba Health. Body weight measurements

Materials and methods Study design and population Using comprehensive health care databases of the Province of Manitoba in Canada (population 1.2 million in 2004), we designed a retrospective historical cohort study of women aged 40 to 59 years who underwent clinical BMD testing in the province for evaluation of fracture risk. Citizens of the province of Manitoba have universal access to publicly funded medical care including BMD testing. The Manitoba Bone Density Program is a unique, integrated program that has managed all clinical DXA testing of the province since 1997 [15]. Criteria for testing are consistent with most published guidelines and, include female sex, age 65 years or older, premature ovarian failure, prior fragility fracture, X-ray evidence of osteopenia, prolonged corticosteroid use, and other pertinent clinical risk factors (www.gov.mb.ca/ health/programs/mbd). DXA testing rates for this program have been published and the program’s database has been shown to be over 99% complete and accurate [15, 16]. The hospital discharge summary database provides information on hospital admissions for the entire province. The physician claims database provides information on all in- and out-patient therapeutic procedures and physician visits. Hospital discharge abstracts provide up to 16 diagnoses and up to four procedure codes using the International Classification of Disease-9-Clinical Modification (ICD-9-CM) system and physician billing claims provide one ICD-9-CM diagnosis code and one or more procedure codes for all in-patient, out-patient, and office-based

Height and weight were by self-report prior to 2000 and were measured by a wall-mounted stadiometer and bathroom scale from 2000 onward. BMI was calculated as weight in kilograms divided by height squared in meters. Although weight is used in many osteoporosis risk assessment strategies, we also evaluated BMI as it has been selected by the World Health Organization (WHO) as one of several fracture risk assessment parameters for inclusion in an absolute fracture risk prediction tool [11]. Bone density measurements Prior to 2000, DXA measurements were performed using a pencil-beam instrument (Lunar DPX; GE Lunar, Madison, WI, USA) and after this date a fan-beam instrument was used (Lunar Prodigy; GE Lunar, Madison, WI, USA). Instruments were cross-calibrated using 59 volunteers and anthropomorphic phantoms. No clinically significant differences were identified (T-score differences < 0.2). Therefore, all analyses are based upon the unadjusted numerical results provided by the instrument. Densitometers showed stable long-term performance (coefficient of variation [CV] <0.5%) and satisfactory in vivo precision (CV 1.7% for L1–L4 and 1.1% for the total hip) [19]. Dual-energy X-ray absorptiometry scans were performed and analyzed in accordance with the manufacturer’s recommendations. Lumbar spine T-scores (number of standard deviations [SD] above or below young adult mean BMD) and Z-scores (number of SDs above or below age-matched mean BMD) were calculated using the manufacturer’s USA white female reference values based upon the revised

Osteoporos Int (2009) 20:363–370

NHANES III reference data (Prodigy version 8.8) [20]. Vertebral levels affected by artifacts were excluded by experienced physicians using conventional criteria [21]. A T-score of −2.5 or lower was considered to be osteoporotic based upon the WHO classification [2]. Osteoporosis Self-Assessment Tool Because BMD measurements are not widely available in certain communities, several risk assessment tools have been developed to target women at higher risk of osteoporosis for BMD testing [22]. The Osteoporosis Self-Assessment Tool (OST) uses self-reported data and is derived by score = 0.2 × (weight in kg − age), truncated to an integer with a lower score, identifying women at higher risk of osteoporosis at the femoral neck [23]. OST has been reported to have a sensitivity of 89.2% and a specificity of 45.0% (using a cut-off point of ≤1) for identifying osteoporosis in postmenopausal women aged 45 to 64 years [24]. Fracture ascertainment Each participant’s longitudinal health service record was assessed from the date of the BMD testing to 31 March 2004 for the presence of ICD-9-CM fracture codes that were not associated with trauma codes (ICD-9-CM E800-E879 and E890-E999). Specific fracture sites of interest were the hip (ICD-9-CM 820–821), the spine (ICD-9-CM 805), forearm (ICD-9-CM 813), and proximal humerus (ICD-9-CM 812) [25]. As vertebral fractures are frequently asymptomatic, we were only able to ascertain those that necessitated medical attention. In addition, we required that hip fractures and forearm fractures be accompanied by a site-specific fracture reduction, fixation or casting code. Hip, spine, forearm, and proximal humerus fractures defined in this way were collectively designated “osteoporotic” fractures. Using these fracture definitions, we have previously shown that BMD measurements predict fractures in our clinical cohort as well as those reported in large meta-analyses [26].

365

province or the end of the observation period (31 March 2004). Statistical analysis Descriptive statistics were tabulated for the study cohort. Linear regression was performed to evaluate the relationship between weight or BMI and BMD after adjustment for age. We divided the cohort into quartiles of weight, BMI, and OST, and computed the proportions of individuals with low BMD scores and incident fractures in each stratum. A Chi-squared test with Cochrane–Armitage linear trend analysis was performed to test for a difference across quartiles. Receiver operating characteristic (ROC) curves were plotted for weight, BMI, and OST using femoral neck Tscore ≤ −2.5, T-score ≤ −2.5 at any site, or incident osteoporotic fractures as the reference variables. We compared the areas under the ROC curves for the three variables for all women and within predefined subgroups: age group (40–49 years and 50–59 years), and in women with no prevalent fractures or use of systemic corticosteroids. Diagnostic accuracy measures were considered using different thresholds of weight, BMI, and OST score. Using multivariate Cox proportional hazard models, the association between weight, BMI and OST and incident osteoporotic fractures was evaluated, in turn, with and without including baseline femoral neck BMD. The models each included age (in years), presence of a prevalent fracture (as a binary variable) and use of systemic corticosteroids (as a binary variable) as covariates, which were all entered into the models simultaneously. Receiver operating characteristic curves were compared using the non-parametric method of DeLong et al. [27], which allows for efficient comparison of the highly correlated curves originating from a common population (AccuROC 2.5; Accumetric, Montreal, QC, Canada). All other analyses were performed using Statistica (Version 6.1, StatSoft, Tulsa, OK, USA). A p value of less than 0.05 was considered statistically significant for all analyses.

Covariates Results The presence of a prevalent non-craniofacial fracture (any ICD-9-CM 805–829 code between 1987 and the date of BMD testing) and use of systemic corticosteroids in the year prior to BMD testing (3 months or more at a mean daily equivalent dose of at least 7.5 mg of prednisone) were documented. Clinical endpoints Women were followed until the first one of these events: occurrence of a fracture, death, relocation out of the

We identified 8,254 women aged 40 to 59 years who fulfilled the study criteria. The mean age and weight of the study cohort were 52.7 (SD 4.9) years and 69.2 (15.0) kg respectively; 589 (7.1%) had a prevalent fracture and 1,226 (14.9%) had a BMD measurement consistent with osteoporosis (Table 1). The mean OST score was 2.9 (SD 3.0, range –10 to 18); 3,149 (38.2%) had an OST of ≤ 1. Women aged 40–49 years had higher T-scores and more frequent use of systemic corticosteroids than women aged 50–59 years.

366

Osteoporos Int (2009) 20:363–370

Table 1 Baseline characteristics of women in the entire cohort and by age groups

Age (SD), years Weight (SD), kg Height (SD), kg BMI (SD), kg/m2 OST score (SD) OST score ≤1, n (%) Prevalent fracture, n (%) Systemic corticosteroid use, n (%) Lumbar spine T-score Femoral neck T-score Total hip T-score Femoral neck T-score ≤ −2.5, n (%) T-score ≤-2.5 at any site, n (%)

All women (40–59 years), N=8,254

Women 40–49 years, n=2,262

Women 50–59 years, n=5,992

P younger vs older

52.7 (4.9) 69.2 (15.0) 162.1 (6.5) 26.3 (5.6) 2.9 (3.0) 3,149 (38.2) 589 (7.1) 434 (5.3) −0.8 (1.4) −1.1 (0.9) −0.6 (1.1) 377 (4.6) 1,226 (14.9)

46.1 (2.7) 68.6 (16.2) 162.5 (6.8) 26.0 (6.0) 4.0 (3.3) 516 (22.8) 150 (6.6) 184 (8.1) −0.5 (1.3) −0.9 (0.9) −0.5 (1.1) 58 (2.6) 234 (10.3)

55.1 (2.8) 69.4 (14.5) 161.9 (6.4) 26.5 (5.4) 2.5 (2.8) 2,633 (43.9) 439 (7.3) 250 (4.2) −0.9 (1.3) −1.1 (0.9) −0.6 (1.1) 319 (5.3) 992 (16.6)

< 0.0001 0.0307 < 0.0001 < 0.0001 < 0.0001 < 0.0001 0.2738 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001

Osteoporosis Self-Assessment Tool; lower score indicates a greater number of osteoporotic factors (OST=0.2 × (weight in kg − age))

Over 27,256 person years of observation (mean 3.3 years per individual), 225 (2.7%) experienced one or more incident fractures (19 hip, 56 spine, 59 humerus, and 97 wrist). The proportion with incident fractures was similar in women aged 50–59 years (158, [2.6%]) and women aged 40–49 years (67 [3.0%]). The prevalence of osteoporosis and of incident fractures increased with decreasing weight, BMI, and OST quartile (Table 2). The proportion of incident fractures increased from 2.1% in the highest BMI quartile (BMI > 29 kg/m2) to 3.4% in the lowest quartile (BMI < 22 kg/m2; p=0.0094). The area under the ROC curves revealed that weight, BMI, and OST predicted low BMD in the whole cohort, in both age subgroups (40 to 49 years and 50 to 59 years) and in women without prior fracture or corticosteroid therapy

(Table 3). These variables were less discriminatory, however, when it came to predicting incident fractures, particularly in the younger age group. The performance characteristics of various thresholds of each of the clinical variables are shown in Table 4. Body weight of 57 kg or less was found to have a sensitivity of 50.9% and a specificity of 80.9% for predicting a T-score of ≤ −2.5 at the femoral neck, and of 24.9% and 79.6% respectively, for prediction of incident fractures. At a cut-off point of ≤ 70 kg, the sensitivity was 89.4% and the specificity 42.4% to identify women with osteoporosis at the femoral neck. BMI below 20 kg/m2 had specificity of 93.5%, but very low sensitivity (24.9%) in detecting femoral neck osteoporosis; at a cut-off of 26 kg/m2, the sensitivity rose to 82.5%, but the specificity decreased to 45.6%. OST, at a threshold of

Table 2 Prevalence (%) of osteoporosis by BMD criteria and incident fractures in relation to weight, BMI, and OST quartiles Criteria/fractures Weight (kg) Femoral neck T-score ≤ −2.5 T-score ≤ −2.5 at any site Any incident fracture BMI (kg/m2) Femoral neck T-score ≤ −2.5 T-score ≤ −2.5 at any site Any incident fracture OST Femoral neck T-score ≤ −2.5 T-score ≤ −2.5 at any site Any incident fracture

P for trend Q1 (n=2,032) weight < 59 221 (10.9) 559 (27.5) 66 (3.2) Q1(n=2,062) BMI < 22 199 (9.7) 532 (25.8) 71 (3.4) Q1 (n=1,900) Score < 0 227 (11.9) 574 (30.2) 66 (3.5)

Q2 (n=2,135) weight 59 to 67 94 (4.4) 337 (15.8) 64 (3.0) Q2 (n=2,065) BMI 22 to 25 101 (4.9) 310 (15.0) 54 (2.6) Q2 (n=2,456) Score 0–2 98 (4.0) 367 (14.9) 74 (3.0)

Q3 (n=1,982) weight 68 to 77 48 (2.4) 216 (10.9) 48 (2.4) Q3 (n=2,064) BMI 26 to 29 46 (2.2) 251 (12.2) 57 (2.8) Q3 (n=1,919) Score 3–4 42 (2.2) 193 (10.1) 40 (2.1)

Q4 (n=2,105) weight > 77 14 (0.7) 114 (5.4) 47 (2.2) Q4 (n=2,063) BMI > 29 31 (1.5) 133 (6.4) 43 (2.1) Q4 (n=1,979) Score > 4 10 (0.5) 92 (4.6) 45 (2.3)

< 0.0001 < 0.0001 0.0215

< 0.0001 < 0.0001 0.0094

< 0.0001 < 0.0001 < 0.0001

Osteoporos Int (2009) 20:363–370

367

Table 3 Area under the receiver operating characteristic curves (95% CI) for osteoporotic BMD or incident osteoporotic fracture Femoral neck T-score ≤ −2.5 All (N=8,254) Weight 0.76 (0.73–0.78) BMI 0.72 (0.70–0.75)* 0.77 (0.75–0.79) **,*** OSTa Without prior fracture or corticosteroids (n=7,270) Weight 0.77 (0.74–0.79) BMI 0.73 (0.71–0.76)* 0.77 (0.75–0.80)*** OSTa Age 40–49 years (n=2,262) Weight 0.73 (0.67–0.79) BMI 0.69 (0.620–0.76) 0.73 (0.66–0.79) OSTa Age 50–59 years (n=5,992) Weight 0.77 (0.74–0.79) BMI 0.73 (0.71–0.76)* 0.77 (0.74–0.79)*** OSTa

T-score ≤ −2.5 at any site

Incident fracture

0.69 (0.68–0.71) 0.67 (0.65–0.69)* 0.71 (0.69–0.72)*,***

0.55 (0.51–0.59) 0.55 (0.51–0.59) 0.56 (0.52–0.60)

0.70 (0.68–0.72) 0.68 (0.66–0.70)* 0.72 (0.70–0.73)*,***

0.55 (0.51–0.60) 0.56 (0.51–0.60) 0.56 (0.52–0.60)

0.71 (0.68–0.75) 0.69 (0.66–0.73) 0.71 (0.68–0.75)

0.50 (0.43–0.57) 0.48 (0.41–0.55) 0.50 (0.43–0.58)

0.69 (0.68–0.71) 0.67 (0.65–0.79)* 0.70 (0.68–0.72)***

0.57 (0.53–0.61) 0.58 (0.53–0.62) 0.58 (0.54–0.63)

*p<0.001 or ** p<0.05 for BMI versus weight, ***p<0.001 for OST versus BMI; OST was transformed to (−OST) to construct ROC curves

a

≤ 1, had a sensitivity of 46.8% and a specificity of 81.8% for identifying women with a low T-score at the femoral neck. In univariate analysis, each standard deviation decrease in weight (15 kg) was associated with a 17% increase in the risk of incident fracture (95% CI; 1.01–1.35), and each standard deviation decrease in BMI (5.6 kg/m2) and OST [3] with a similar risk increase (Table 5). These remained constant after adjustment for age, prevalent fracture, and use of systemic corticosteroids. The addition of femoral neck BMD to the model caused the association between weight, BMI and OST and fractures to lessen. Femoral neck BMD and the presence of osteoporotic fractures at baseline were independently associated with the risk of incident

fracture in all three multivariate models (HR 1.6, 95% CI 1.4–1.9 and HR 2.7, 95% CI 1.9–3.8 respectively).

Discussion Women at high risk of osteoporosis-related fractures can be targeted in clinical practice with specific interventions. Younger women with risk factors such as low body weight and BMI may also be at higher risk of fractures. The results from our large cohort study support the previously documented associations between weight, BMI, and BMD. Body weight has been identified in previous reports as a

Table 4 Discriminatory performance of weight, BMI and OST for the prediction of T-score ≤ −2.5 or any incident fracture Femoral neck T-score ≤ −2.5

T-score ≤ −2.5 at any site

Incident fracture

Sensitivity % (95% CI)

Specificity % (95% CI)

Sensitivity % (95% CI)

Specificity % (95% CI)

Sensitivity % (95% CI)

Specificity % (95% CI)

(49.8–52.0) (57.6–59.7) (88.7–90.1)

80.9 (80.1–81.8) 77.0 (76.1–77.9) 42.4 (41.3–43.4)

39.6 (38.5–40.6) 45.6 (44.5–46.7) 80.0 (79.2–80.9)

82.8 (82.0–83.6) 79.0 (78.2–79.9) 44.6 (43.5–45.7)

24.9 (24.0–25.8) 29.3 (28.4–30.3) 65.3 (64.3–66.4)

79.6 (78.7–80.4) 75.5 (74.6–76.4) 41.1 (40.0–42.2)

(24.0–25.9) (46.1–48.3) (81.7–83.3)

93.5 (93.0–94.0) 80.3 (79.4–81.1) 45.6 (44.5–46.7)

19.8 (19.0–20.7) 38.2 (37.1–39.2) 74.1 (73.2–75.1)

94.8 (94.3–95.3) 82.0 (81.2–82.8) 47.6 (46.5–48.6)

11.1 (10.4–11.8) 27.6 (26.6–28.5) 60.4 (59.4–61.5)

92.8 (92.2–93.3) 79.2 (78.3–80.1) 44.5 (43.4–45.5)

(59.2–61.3) (75.5–77.3) (85.5–87.0)

78.8 (77.9–79.6) 63.7 (62.6–64.7) 48.8 (47.7–49.9)

46.8 (45.7–47.9) 63.8 (62.7–64.8) 76.8 (75.8–77.7)

81.1 (80.3–82.0) 66.3 (65.3–67.3) 51.4 (50.3–52.5)

29.3 (28.4–30.3) 45.8 (44.7–46.9) 62.2 (61.2–63.3)

77.2 (76.3–78.1) 62.1 (61.0–63.1) 47.5 (46.4–48.6)

Weight (kg) ≤ 57 50.9 ≤ 59 58.6 ≤ 70 89.4 BMI (kg/m2) ≤ 20 24.9 ≤ 22 47.2 ≤ 26 82.5 OST ≤1 60.2 ≤2 76.4 ≤3 86.2

368

Osteoporos Int (2009) 20:363–370

Table 5 Risk of developing any incident osteoporotic fracture without and with adjustment for age, prior fracture history, systemic corticosteroid use and femoral neck BMD

Model 1 Weight per SD decrease Age per decade increase Prevalent osteoporotic fracture Systemic corticosteroid use Femoral neck BMD per SD decrease Model 2 BMI per SD decrease Age per decade increase Prevalent osteoporotic fracture Systemic corticosteroid use Femoral neck BMD per SD decrease Model 3 OST per SD decrease Prevalent osteoporotic fracture Systemic corticosteroid use Femoral neck BMD per SD decrease

Univariate HR (95% CI)

Multivariate HR (95% CI) without BMD

Multivariate HR (95% CI) with BMD

1.17 (1.01–1.35) – – – –

1.19 1.07 3.28 1.24 –

(1.03–1.37) (0.94–1.22) (2.36–4.57) (0.75–2.04)

0.97 0.98 2.68 0.97 1.61

(0.84–1.14) (0.85–1.12) (1.91–3.76) (0.59–1.61) (1.37–1.89)

1.17 (1.01–1.35) – – – –

1.18 1.07 3.26 1.25 –

(1.02–1.37) (0.94–1.22) (2.35–4.54) (0.76–2.06)

1.00 0.98 2.70 0.98 1.59

(0.86–1.17) (0.86–1.12) (1.93–3.78) (0.59–1.63) (1.36–1.86)

1.19 (1.03–1.38) – – –

1.20 (1.04–1.39) 3.29 (2.36–4.57) 1.24 (0.75–2.03) –

0.98 2.68 0.97 1.60

(0.83–1.14) (1.91–3.75) (0.59–1.62) (1.37–1.88)

predictor of peak bone mass in pre-menopausal women and a predictor of BMD in women aged 45 to 64 years and in older women [5, 22, 28]. We have shown that osteoporosis, as defined by a T-score of ≤ −2.5 at the femoral neck or a Tscore ≤ −2.5 at any site is associated with weight, BMI, and OST in the entire cohort and also in the subgroups of women aged 40–49 years and women aged 50–59 years. The area under the curve for all clinical parameters was consistent with better identification of osteoporosis at the femoral neck than at any other site. This may be important as prospective studies have demonstrated that proximal femur BMD has a gradient for fracture risk as high as or higher than that of other sites [29, 30]. To evaluate the prognostic performance of weight, BMI, and OST, we tested thresholds found to be predictive in previous studies, mostly performed in post-menopausal women. In a cohort of Canadian women aged 45 years and over, weight below 70 kg had a sensitivity of 87% and a specificity of 48% in identifying osteoporosis at the femoral neck (40% of the cohort was below the age of 65 years) [22]. We documented a sensitivity of 89.4% and a specificity of 42.4% at a weight threshold of 70 kg for predicting a femoral neck T-score of −2.5 or less. Body weight below 58 kg has found to be predictive of hip fractures in post-menopausal women in the Study of Osteoporotic Fractures [5]. A BMI of 25 kg/m2 has been identified as the reference point below which the risk of hip and any osteoporotic fracture starts to increase [11]. Similar conclusions were reached in the Study of Osteoporotic Fractures using a BMI of 26.2 kg/m2 [5]. With a BMI cut-off point of 26 kg/m2,

we found that the diagnostic accuracy for classifying women as osteoporotic or as having an incident fracture was similar to using a weight criterion of 70 kg. OST of >1 has traditionally been used to determine low risk designation in post-menopausal women [31]; in a cohort of post-menopausal women aged 45–65 years, the accuracy of OST for femoral neck T-score ≤ −2.5 at a cutoff value ≤ 1 was 89.2% and 45.0% for sensitivity and specificity respectively [24]. With our data, the cut-off value that produced a sensitivity of 86% and a specificity of 49% was ≤ 3. Our findings are in close agreement with a recent report from the Study of Osteoporotic Fractures that compared several fracture risk assessment tools (including OST) to identify low hip and lumbar spine BMD in 7,779 US women aged 67 years and older. OST had the greatest area under the ROC curve (AUC 0.76, 95% CI 0.74–0.77). Using published cut-off points, the risk tools had sensitivities of ≥ 85% and specificities of ≤ 48% [32]. Recently, the World Health Organization has proposed a 10-year fracture prediction tool, validated in multiple cohorts of men and women over 40 years of age, which incorporates a number of clinical parameters (including BMI) and BMD at the femoral neck [33]. Prediction tables have been created to estimate risk, based on BMI in the absence of BMD measurement, which could be of use in regions where DXA is unavailable. The positive predictive values of these anthropometric markers at cut-off points of high sensitivity were low (data not shown); this is in keeping with the low prevalence of osteoporosis and fractures in this young population and the low specificity of the tests.

Osteoporos Int (2009) 20:363–370

An important finding of our study is the association between lower weight or BMI and an increased risk of osteoporosis-related fractures in this younger population. With each standard deviation decrease in weight or in BMI, the age-adjusted risk for fractures increased by approximately 18%. Results were comparable with OST. We were unable to show an independent effect of weight or BMI on the risk of osteoporotic fracture after integration of femoral neck BMD into the models. De Laet came to similar conclusions in a meta-analysis in which the age-adjusted risk for any fracture increased with lower BMI; after adjusting for BMD, BMI was not found to be predictive of fracture risk except for hip fracture [11]. This suggests that BMD may be an important intermediate in the causal pathway. The presence of fractures was associated with an almost three-fold increase in the risk of fractures, as documented in other populations [33, 34]. This association was independent of BMD and underlines the importance of identifying women with this strong clinical marker for recurrent fractures. The determination of thresholds for high or low risk of outcomes depends on the clinical context and varies with expected benefits and costs associated with further testing, high risk designation and availabilities of effective interventions [35]. Pharmacological interventions have proven successful in reducing the risk of fractures in women selected mostly on the basis of low BMD and/or the presence of fragility fractures [1]. Interventions specifically aimed at clinical risk factors have been not been extensively studied and have not been shown to reduce fracture risk. Behavioral changes and multi-faceted approaches are necessary for long-term beneficial impact on body weight [36]. Preservation of proper weight should be encouraged as weight loss has also been related to further reduction in bone strength and fractures [37]. Our study has multiple strengths; this is the largest study to evaluate the association between weight, BMD, and fractures in women aged 40 to 59 years. We have used two statistical approaches to assessing the strength of association between the variables and the outcomes of osteoporosis and fracture. First, the discriminatory performance of weight, BMI, was evaluated with measures of sensitivity and specificity and receiver operating characteristic curves, and second, the risk of fractures as function of weight and BMI and other baseline predictors was modeled in a multivariate regression. Both techniques yielded comparable results that are consistent with data documented in observational studies and clinical trials performed in older women. Because of the observational study design and the lack of information on clinical variables in the database, we

369

were unable to control for menopausal status, family history of fracture, alcohol and tobacco use, and history of rheumatoid arthritis, which have all been shown in metaanalyses to predict fractures in older women [8]. Nonetheless, we were able to adjust for age, BMD, systemic corticosteroid use and the presence of prior fractures; these variables have been selected for being highly predictive of incident fractures in clinical prediction tools and have proven to reflect a higher risk of fracture in our population as well [38]. Our cohort consisted of women referred for BMD testing, and perhaps at higher risk of fractures than the general population of similar age, and this limits the generalizibility of our study. We have documented incident fractures in 2.7% of all women during the study period. In women with no prevalent fracture or use of corticosteroids, this proportion was slightly lower at 2.3%. Referral bias had little effect on mean Z-scores for our cohort, as documented by the fact that those scores were in close agreement with manufacturer reference data (within 0.2 SD of the expected value). In summary, we have shown that low weight and BMI at specific thresholds of 70 kg and 26 kg/m 2 predict osteoporosis as determined by BMD with good sensitivity, but low specificity in women aged 40–59 years. We have also demonstrated the importance of prevalent fractures in predicting future fractures in that age group. In multivariate analysis, BMI and weight were found to be associated with fracture risk independent of prevalent fracture and use of corticosteroids, but were dependent on BMD. We conclude that the measurement of simple clinical variables such as age, fracture history, and weight (or BMI) in areas where BMD testing is not readily accessible or limited facilitates the identification of middle-aged women at risk of osteoporosis and related fractures. The negative impact of low body weight on bone health should be more widely recognized. Acknowledgements We are indebted to Manitoba Health for providing data. The authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. The results and conclusions are those of the authors, and no official endorsement by Manitoba Health is intended or should be inferred. This article has been reviewed and approved by the members of the Manitoba Bone Density Program Committee. This study was supported in part by an unrestricted educational grant from the CHAR/GE Healthcare Development Awards Programme. Conflicts of interest S. Morin: honoraria for lecture and advisory board membership from Merck Frosst, Procter & Gamble, sanofiaventis, Novartis, E. Lilly, Servier, and Amgen. W.D. Leslie: honoraria for lectures in the past year: Merck Frosst Canada. Unrestricted educational and research grants: Merck Frosst Canada, The Alliance for Better Bone Health - sanofi-aventis and Procter & Gamble Pharmaceuticals Canada, Novartis Pharmaceuticals Canada. James T. Tsang: No conflicts.

370

Osteoporos Int (2009) 20:363–370

References 1. Sambrook P, Cooper C (2006) Osteoporosis. Lancet 367:2010–2018 2. Brown JP, Josse RG (2002) 2002 clinical practice guidelines for the diagnosis and management of osteoporosis in Canada. CMAJ 167:S1–S34 3. National Osteoporosis Foundation (2003) Physician's guide to the prevention and treatment of osteoporosis. National Osteoporosis Foundation, Washington, DC 4. Kanis JA (2002) Diagnosis of osteoporosis and assessment of fracture risk. Lancet 359:1929–1936 5. Cummings SR, Nevitt MC, Browner WS, Stone K, Fox KM, Ensrud KE et al (1995) Risk factors for hip fracture in white women. Study of Osteoporotic Fractures Research Group. N Engl J Med 332:767–73 6. Van Staa TP, Leufkens HG, Abenhaim L, Zhang B, Cooper C (2000) Use of oral corticosteroids and risk of fractures. J Bone Miner Res 15:993–1000 7. Kanis JA, Borgstrom F, De Laet C, Johansson H, Johnell O, Jonsson B et al (2005) Assessment of fracture risk. Osteoporos Int 16:581–89 8. Kanis JA, Oden A, Johnell O, Johansson H, De Laet C, Brown J et al (2007) The use of clinical risk factors enhances the performance of BMD in the prediction of hip and osteoporotic fractures in men and women. Osteoporos Int 18:1033–1046 9. Reid IR (2002) Relationships among body mass, its components and bone. Bone 31:547–555 10. Felson DT, Zhang Y, Hannan MT, Anderson JJ (1993) Effects of weight and body mass index on bone mineral density in men and women: the Framingham study. J Bone Miner Res 8:567–573 11. De Laet C, Kanis JA, Oden A, Johanson H, Johnell O, Delmas P et al (2005) Body mass index as a predictor of fracture risk: a meta-analysis. Osteoporos Int 16:1330–1338 12. Johnell O, O'Neill T, Felsenberg D, Kanis J, Cooper C, Silman AJ (1997) Anthropometric measurements and vertebral deformities. European Vertebral Osteoporosis Study (EVOS) Group. Am J Epidemiol 146:287–293 13. Williams AR, Weiss NS, Ure CL, Ballard J, Daling JR (1982) Effect of weight, smoking, and estrogen use on the risk of hip and forearm fractures in postmenopausal women. Obstet Gynecol 60:695–699 14. Roy DK, O'Neill TW, Finn JD, Lunt M, Silman AJ, Felsenberg D et al (2003) Determinants of incident vertebral fracture in men and women: results from the European Prospective Osteoporosis Study (EPOS). Osteoporos Int 14:19–26 15. Leslie WD, Metge C (2003) Establishing a regional bone density program: lessons from the Manitoba experience. J Clin Densitom 6:275–282 16. Leslie WD, Caetano PA, Macwilliam LR, Finlayson GS (2005) Construction and validation of a population-based bone densitometry database. J Clin Densitom 8:25–30 17. Roos LL, Walld RK, Romano PS, Roberecki S (1996) Short-term mortality after repair of hip fracture. Do Manitoba elderly do worse? Med Care 34:310–326 18. Roos LL, Mustard CA, Nicol JP, McLerran DF, Malenka DJ, Young TK et al (1993) Registries and administrative data: organization and accuracy. Med Care 31:201–212 19. Leslie WD (2006) The importance of spectrum bias on bone density monitoring in clinical practice. Bone 39:361–368 20. Binkley N, Kiebzak GM, Lewiecki EM, Krueger D, Gangnon RE, Miller PD et al (2005) Recalculation of the NHANES database SD

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

improves T-score agreement and reduces osteoporosis prevalence. J Bone Miner Res 20:195–201 Hansen KE, Binkley N, Christian R, Vallarta-Ast N, Krueger D, Drezner MK et al (2005) Interobserver reproducibility of criteria for vertebral body exclusion. J Bone Miner Res 20:501–508 Cadarette SM, Jaglal SB, Murray TM, McIsaac WJ, Joseph L, Brown JP (2001) Evaluation of decision rules for referring women for bone densitometry by dual-energy x-ray absorptiometry. JAMA 286:57–63 Koh LK, Sedrine WB, Torralba TP, Kung A, Fujiwara S, Chan SP et al (2001) A simple tool to identify Asian women at increased risk of osteoporosis. Osteoporos Int 12:699–705 Gourlay ML, Miller WC, Richy F, Garrett JM, Hanson LC, Reginster JY (2005) Performance of osteoporosis risk assessment tools in postmenopausal women aged 45–64 years. Osteoporos Int 16:921–927 Kanis JA, Johnell O, Oden A, Dawson A, De Laet C, Jonsson B (2001) Ten year probabilities of osteoporotic fractures according to BMD and diagnostic thresholds. Osteoporos Int 12:989–995 Leslie WD, Tsang JF, Caetano PA, Lix LM (2007) Effectiveness of bone density measurement for predicting osteoporotic fractures in clinical practice. J Clin Endocrinol Metab 92:77–81 DeLong ER, DeLong DM, Clarke-Pearson DL (1988) Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 44: 837–845 Hawker GA, Jamal SA, Ridout R, Chase C (2002) A clinical prediction rule to identify premenopausal women with low bone mass. Osteoporos Int 13:400–406 Kanis J, Johnell O, Oden A, Johansson H, Eisman J, Fujiwara S et al. (2006) The use of multiple sites for the diagnosis of osteoporosis. Osteoporos Int 17:527–534 Leslie WD, Lix LM, Tsang JF, Caetano PA (2007) Single-site vs multisite bone density measurement for fracture prediction. Arch Intern Med 167:1641–1647 Rud B, Hilden J, Hyldstrup L, Hrobjartsson A (2007) Performance of the Osteoporosis Self-Assessment Tool in ruling out low bone mineral density in postmenopausal women: a systematic review. Osteoporos Int 18:1177–1187 Gourlay ML, Powers JM, Lui LY, Ensrud KE (2008) Clinical performance of osteoporosis risk assessment tools in women aged 67 years and older. Osteoporos Int doi:10.1007/s00198-007-0555-1 Kanis JA, Johnell O, Oden A, Johansson H, McCloskey E (2008) FRAX and the assessment of fracture probability in men and women from the UK. Osteoporos Int 19:385–397 Kanis JA, Johnell O, De Laet C, Johansson H, Oden A, Delmas P et al (2004) A meta-analysis of previous fracture and subsequent fracture risk. Bone 35:375–382 Pepe MS, Feng Z, Huang Y, Longton G, Prentice R, Thompson IM et al (2008) Integrating the predictiveness of a marker with its performance as a classifier. Am J Epidemiol 167:362–368 Vescovi JD, Jamal SA, De Souza MJ (2008) Strategies to reverse bone loss in women with functional hypothalamic amenorrhea: a systematic review of the literature. Osteoporos Int 19:465–478 Ensrud KE, Cauley J, Lipschutz R, Cummings SR (1997) Weight change and fractures in older women. Study of Osteoporotic Fractures Research Group. Arch Intern Med 157:857–863 Siminoski K, Leslie WD, Frame H, Hodsman A, Josse RG, Khan A et al (2005) Recommendations for bone mineral density reporting in Canada. Can Assoc Radiol J 56:178–188