Repeat Bone Mineral Density Screening Measurement and Fracture Prediction in Older Men: A Prospective Cohort Study (2024)

Associated Data

Data Availability Statement

Abstract

Keywords: bone mineral density, fracture risk, older men

The majority of osteoporosis-related fractures in men occur in those men aged 65 years and older (1). Low bone mineral density (BMD) is a strong independent risk factor for fractures in older men (2, 3). Thus, several professional societies (4-6) have recommended osteoporosis screening with initial BMD testing in men aged 70 years or older. However, these guidelines have not addressed the timing of re-screening with repeat BMD measurement as it is uncertain whether repeated BMD testing improves fracture prediction in older men above and beyond that provided by an initial BMD. Previous studies in older (7) and early and late postmenopausal women (8) reported little additional value of repeat BMD 3 to 8 years after the initial BMD in the prediction of incident fractures. Similarly, a second BMD 4 years after the initial BMD did not meaningfully improve the prediction of hip or major osteoporotic fracture (MOF) in a study of 802 older adults that included 310 men (9).

In contrast, a previous analysis of 4470 older community-dwelling men enrolled in the Osteoporotic Fractures in Men (MrOS) study found that men with accelerated hip bone loss compared with men who maintained hip BMD during a time period of 4.6 years had increased risks of subsequent hip and any nonvertebral fracture (10). However, BMD change in this analysis was estimated with random effects regression models using BMD measurements at 2 to 3 time points and quadratic terms for age. Thus, results may not be applicable to the clinical practice setting where simple linear BMD change is the measure that is readily accessible for clinical decision making. In addition, this investigation did not evaluate the value of adding BMD change to a fracture prediction model based on initial BMD alone in the discrimination of incident fracture outcomes.

To evaluate whether a second BMD 7 years after an initial BMD improves fracture prediction in older community-dwelling men, the present study used data from 3561 participants in the MrOS study with hip BMD measurements at baseline and Year 7 examinations and subsequent follow-up for any clinical fractures, including MOF and hip fracture.

Methods

Results

A total of 3651 men with initial hip BMD measurement at the baseline examination and repeat hip BMD measurement at the Y7 examination were included in the analytical cohort. The mean (SD) age of participants was 72.3 (5.1) years at the initial BMD and 79.1 (5.1) years at the repeat BMD. The mean (SD) initial total hip BMD was 0.97 (0.14) g/cm² and 683 men (18.7%) had a BMD T-score at the total hip or femoral neck ≤ −1.5. The mean (SD) total hip BMD change was −0.38% (0.75%) or −0.004 (0.007) g/cm² per year. Of the 3586 men with an initial BMD T-score at the femoral neck and total hip > −2.5, only 91 (2.5%) transitioned to a BMD T-score of ≤ −2.5 at either skeletal site at the Y7 examination. The correlation coefficient between initial and Y7 total hip BMD was 0.94.

During a mean (SD) follow-up of 8.2 (4.2) years after the Y7 repeat BMD, 793 men (21.7%) experienced 1 or more clinical fractures including 426 men (11.7%) with 1 or more MOF and 193 (5.3%) men with 1 or more hip fractures. Men with vs men without incident MOF were more likely to be White, older, and less active, have experienced a prior confirmed fracture, have reported falling in the past year, and have lower BMI and greater weight loss (Table 1). Mean value of initial hip BMD was lower and annualized rates of hip bone loss were slightly greater among men with vs those without an incident MOF. On average, men who experienced a MOF lost 0.52% or 0.005 g/cm² per year at the total hip compared with 0.36% or 0.003 g/cm² per year in men without MOF (P value < 0.001 for both comparisons).

In unadjusted and adjusted models based on initial total hip BMD, each SD decrement in initial BMD was associated with an increased risk of fracture during the follow-up period after the repeat BMD measurement (multivariable HR [95% CI] 1.77 [1.57-1.98] for MOF, 1.51 [1.39-1.64] for any clinical fracture, and 2.01 [1.69-2.38] for hip fracture) (Table 2). In similar models based on annualized percent total hip BMD change, each SD decrement in BMD change was also associated with risk of fracture (multivariable HR [95% CI] 1.22 [1.10-1.35] for MOF, 1.21 [1.12-1.31] for any clinical fracture, and 1.53 [1.33-1.75] for hip fracture). In models based on the combination of initial BMD and BMD change, both initial BMD and BMD change were associated with fracture risk independent of each other; but the association was stronger for initial BMD (multivariable HR [95% CI] per 1 SD decrement for initial BMD vs. BMD change 1.76 [1.57-1.98] vs. 1.19 [1.08-1.32] for MOF, 1.51 [1.39-1.64] vs. 1.20 [1.11-1.29] for any clinical fracture and 1.99 [1.67-2.36] vs.1.46 [1.28-1.67] for hip fracture).

Substitution of annualized absolute change for annualized percent change in total hip BMD and use of BMD at the femoral neck (Table 3) rather than at the total hip did not substantially alter these findings. Results were also similar in analyses excluding the 336 men taking osteoporosis drug treatment at the repeat BMD measurement or during follow-up (results not shown). In analyses that estimated associations of total hip BMD measures with risk of fracture outcomes using subdistribution hazards models accounting for death as a competing risk, findings were generally similar to analyses estimating associations using Cox proportional hazards models. However, for models based on the combination of initial BMD and BMD change that considered death as a competing risk (Table 4), multivariable associations of BMD change with fracture outcomes were further attenuated and not significant in the case of MOF.

Associations between annualized percent change in total hip BMD and risks of MOF and hip fracture (but not any clinical fracture) appeared to be somewhat more pronounced among men with higher baseline BMD T-score (ie, T-score at femoral neck or total hip > −1.5) (Table 5). For prediction of MOF, the HR (95% CI) per 1 SD decrement in percent BMD change was 1.41 (1.26-1.57) among men with initial T-score > −1.5 vs 1.12 (0.93-1.36) in men with initial T-score ≤ −1.5 (P value for interaction 0.009). For prediction of hip fracture, the HR (95% CI) per 1 SD decrement in percent BMD change was 1.82 (1.58-2.10) among men with initial T-score > −1.5 vs 1.49 (1.18-1.88) in men with initial T-score ≤ −1.5 (P value for interaction 0.006). Results were similar in analyses substituting annualized absolute for annualized percent total hip BMD change, though tests for interaction were of borderline significance (P value for interaction 0.05 for MOF and 0.09 for hip fracture). Findings were also essentially unchanged in analyses expressing initial BMD as ≤ median value at total hip vs > median value (P value for interaction between percent BMD change and initial BMD category 0.05 for MOF and 0.03 for hip fracture). In contrast, we found no evidence that the association of total hip BMD change and risk of any of the fracture outcomes differed by age group (≥80 years vs <80 years) or whether men had or had not experienced weight loss of 5% or more between initial and repeat BMD measurement.

In terms of discriminating between men who did vs those who did not experience a MOF during the follow-up period after the Y7 repeat BMD measurement, performance of models based on initial total hip BMD alone was somewhat better than that of models based on total hip BMD change alone and similar to that of models based on the combination of initial total hip BMD and total hip BMD change. The AUC (95% CI) for multivariable models was 0.68 (0.66-0.71) for the initial BMD model, 0.63 (0.61-0.66) for the BMD change model, and 0.69 (0.66-0.71) for the model based on combination of initial BMD and BMD change (Fig. 2A). With respect to the outcome of any clinical fracture, findings regarding the comparison of model discrimination were similar, though AUC values were slightly lower (AUC [95% CI] for multivariable models 0.64 [0.62-0.66] for the initial BMD model, 0.60 [0.58-0.62] for the BMD change model, and 0.64 [0.62-0.67] for the model based on combination of initial BMD and BMD change) (Fig. 2B). AUC values were highest for models predicting hip fracture. Hip fracture discrimination of the initial BMD model was somewhat better than that of the BMD change model; the AUC (95% CI) for multivariable models was 0.73 (0.69-0.76) for the initial BMD model vs 0.68 (0.64-0.72) for the BMD change model, P value for difference 0.02 (Fig. 2C). Discrimination of the combination of initial BMD and BMD change model appeared slightly better than that of the initial BMD model, but the difference in AUC values was smaller in magnitude and not significant for multivariable models (AUC [95% CI] 0.73 [0.69-0.76] for initial BMD model vs 0.74 [0.71-0.77] for the model based on combination of initial BMD and BMD change, P value for difference 0.06).

Results regarding the comparisons of model discrimination were generally similar in analyses stratified by characteristics including age group, category of initial BMD T-score, and weight loss category (Table 6).

Discussion

In this large population-based study of community-dwelling older men, repeating the hip BMD measurement 7 years after the initial BMD measurement did not meaningfully improve subsequent prediction of fracture including MOF, any clinical fracture and hip fracture. Performance of models based on initial BMD alone in discriminating between men with and without an incident fracture outcome was somewhat better than that of models based on BMD change alone and similar to that of models based on the combination of initial BMD and BMD change. While initial BMD (per 1 SD decrement) and BMD change (per 1 SD decrement) were associated with increased risk of fracture independent of each other, associations were stronger for initial BMD. These results suggest that there is little additional value in routinely incorporating a repeat BMD measurement 7 years after the initial BMD measurement into a screening strategy to assess fracture risk in older community-dwelling men.

Our findings are consistent with those of previous studies conducted in postmenopausal and older women that evaluated the value of repeat hip BMD screening in fracture prediction. A study in 4124 older women (7) (mean age 72 years at the initial BMD) found no significant differences between a model based on initial BMD, repeat BMD 8 years later, and a model based on the combination of initial BMD and BMD change in discriminating between women who did and did not experience fracture (any non-spine, hip fracture) during an average follow-up of 5 years after the repeat BMD. Of note, the AUC statistic of the hip fracture models based on initial total hip BMD (0.73) and the combination of initial total hip BMD and BMD change at the total hip (0.74) were identical to those observed in our study of older community-dwelling men. Similarly, a study of 7219 postmenopausal women (8) (mean age 66 years at repeat BMD) with repeat total hip BMD 3 years after the initial BMD reported that discrimination of MOF and hip fracture during the 9 years after the repeat BMD was essentially indistinguishable using models based on initial BMD (AUC 0.61 for MOF and 0.71 for hip fracture) compared with models based on the combination of initial BMD and BMD change (AUC 0.61 for MOF and 0.73 for hip fracture). Furthermore, investigations in middle-aged (22), postmenopausal (8), and older women (23) also found that initial BMD (expressed as a continuous measure) remained a strong robust predictor of fracture risk after adjustment for BMD change (expressed as a continuous measure), while BMD change was weakly or not associated with fracture risk after consideration of initial BMD. Of note, the modest associations of BMD change with risk of fracture after adjustment for initial BMD that we observed in our study of older men were further attenuated in analyses that accounted for the competing risk of mortality. The ability of baseline hip BMD to predict subsequent risk of hip and nonvertebral fractures over 20 to 25 years of follow-up was previously investigated in a cohort of 8130 community-dwelling women aged 67 years and older (mean age 73.4 years) (24). This study reported that initial hip BMD was a robust predictor of long-term risk of fracture with little or no degradation over time.

Fewer studies examining repeat BMD screening and fracture prediction have included or been limited to men. Findings from our large study of 3651 men are in agreement with those of a previous study conducted in 802 older adults (mean age 75 years) including 310 men (9). In this study, a repeat femoral neck BMD 4 years after the initial BMD did not meaningfully improve discrimination of incident MOF and hip fracture during an average follow-up of 9.6 years after the repeat BMD. For example, the AUC statistic for hip fracture was 0.71 for the initial BMD model, 0.68 for the BMD change model, and 0.72 for the model based on the combination of initial BMD and BMD change. Models performed similarly in analyses stratified by sex. A Canadian study of 5502 middle-aged and older adults that included 1417 men (mean age 64 years) (25) reported that a model based on BMD change (expressed as a continuous variable) was not better than a model based on initial BMD (expressed as a continuous variable and measured within the first 5 years of the study) in predicting the odds of most self-reported fracture outcomes in men during the first 7 years of the study including main (hip, pelvis, vertebral, rib, and forearm), forearm, and rib fractures. However, the model based on total hip BMD change (but not femoral neck BMD change) was better than the model based on initial BMD in predicting any low-trauma fracture in men. Unlike our investigation, this study analyzed fracture outcomes during a time period that was concurrent with the time period of measurement of BMD change and did not address whether repeating BMD improved the prediction of subsequent fracture risk. Of importance, a prior study in our MrOS cohort (10) estimated BMD change using hip BMD measurements at 2 to 3 time points over 4.6 years with random effects linear regression models that included a quadratic term for age to account for nonlinear increases in BMD loss with advancing age. After adjustment for initial BMD, men with accelerated BMD loss (estimated BMD change ≥ 1 SD below mean BMD change) compared with men who maintained BMD (estimated BMD change ≥ 0 g/cm²) had a 7-fold increase in risk of hip fracture and a 2-fold increase in risk of any nonvertebral fracture during an average follow-up of 4.5 years after the final BMD. There was no difference in fracture risk between men with expected BMD loss (estimated BMD change between 0 and 1 SD below mean BMD change) and those who maintained BMD. The statistical approach used to estimate BMD change differed between the previous and present study. Our current study addressed the clinical utility of routinely integrating repeat BMD testing into a screening strategy to assess subsequent fracture risk in older men. Thus, we analyzed simple linear BMD change rather than rate of BMD change generated from mixed effects models, as the former (but not the latter) measure is readily available in the practice setting for shared clinical decision making such as whether or not to recommend initiation of drug treatment for fracture prevention. In addition, the previous analysis did not address the extent to which the repeated BMD measurements improved fracture risk prediction.

Our findings did not vary by clinical characteristics, including age group or weight change category. However, we found some evidence to suggest that the association of BMD change (expressed as annualized percent or absolute change) with risks of MOF and hip fracture (but not any clinical fracture) was slightly higher in magnitude among men with higher vs lower initial BMD T-score. This unexpected finding may in part be due to the phenomenon of regression to the mean, although it was present in analyses stratified by the median value of initial BMD as well as in analyses stratified at an initial T-score of −1.5. While previous studies (7, 9) have not reported evidence of interactions between initial BMD and BMD change for the prediction of fracture risk, a prior investigation in 4498 women aged 40 years and older (22) found that higher baseline BMD was the factor most strongly associated with a subsequent decrease in BMD.

Our results have implications for screening strategies to assess fracture risk in older men. Several organizations (4-6) have endorsed osteoporosis screening with initial BMD testing in older men but have not addressed the timing of re-screening with repeat BMD measurement. In our population-based cohort of older relatively healthy, community-dwelling men who would be candidates for osteoporosis screening, repeat BMD testing 7 years after the initial BMD measurement provided little additional value beyond the initial BMD in predicting risk of future fracture. Our results confirm and extend those of a previous analysis of our cohort (26) that examined the utility of repeat BMD testing in identifying older men who transition to osteoporosis (BMD T-score ≤ −2.5). In the previous study, only 0.2% of the men with initial BMD T-score > −1.5 (78% of the cohort) developed osteoporosis during an average follow-up of 8.7 years and among men with initial BMD T-score between −1.50 and −2.00 (19% of the cohort), the estimated time for 10% to transition to osteoporosis was 8.5 years.

Strengths of our study include the large cohort of well-characterized community-dwelling men, rigorous quality control of hip BMD measurements performed 7 years apart, nearly complete long-term follow-up of participants for vital status and fractures, and confirmation of incident fractures with radiographic reports. However, there are some limitations. Among our cohort of relatively healthy older men, the prevalence of osteoporosis was low at baseline, in agreement with that reported in community-dwelling men 65 years and older enrolled in the nationally representative National Health and Nutrition Examination Survey 2005-2010 (27). Very few men without osteoporosis at baseline transitioned to osteoporosis 7 years later at the repeat BMD measurement. Thus, our findings are relevant to a screening population of older men. They do not apply to older men with intervening conditions that markedly increase risk for accelerated bone loss or to individuals with secondary causes of osteoporosis for whom repeat BMD testing may be indicated. Our study population included predominantly White men and results may not be generalizable to older men of other racial/ethnic groups. Our investigation addressed the value of incorporating a repeat BMD measurement into a screening strategy to assess fracture risk in older men but did not evaluate the utility of BMD measurements in monitoring response to osteoporosis drug treatment in this population.

In conclusion, a repeat BMD measurement 7 years after the initial BMD measurement did not result in a meaningful improvement in subsequent fracture prediction at the population level in community-dwelling older men. These findings suggest that repeat BMD testing within this time interval should not be routinely incorporated into screening strategies to evaluate fracture risk in this population. Future research is needed to inform recommended BMD testing intervals in targeted subgroups of older men at high risk of transitioning to osteoporosis, accelerated bone loss or high fracture risk.

Acknowledgments

The Osteoporotic Fractures in Men (MrOS) Research Group: Coordinating Center (California Pacific Medical Center Research Institute and University of California, San Francisco): P.M. Cawthon (Principal Investigator), S.R. Cummings (co-Investigator), D.C. Bauer (co-Investigator), D.M. Black (co-Investigator), K.L. Stone (co-Investigator), R. Collins (Project Director), T. Blackwell, A. Burghardt, A. Chau, L. Concepcion, S. Ewing, B. Fan, S.L. Harrison, L.Y. Lui, C. Navy, S. Patel, K. Peters, C. Schambach, A. Schwartz; University of Alabama, Birmingham: J. Shikany (Principal Investigator), P. Johnson (Project Director), S. Felder, D. Lee; University of Minnesota: K. Ensrud (Principal Investigator), H. Fink (co-Investigator), S. Diem (co-Investigator), J. Schousboe (co-Investigator), B. Taylor (co-Investigator), L. Langsetmo (co-Investigator), L. Rathburn (Project Director), K. Moen, A. Kats; Stanford University: M. Stefanick (Principal Investigator), A. Hoffman (co-Investigator), D. Kado (co-Investigator), S. Downey, K. Kent; University of Pittsburgh: J. Cauley (Principal Investigator), J. Zmuda (co-Investigator), E. Strotmeyer (co-Investigator), D. Cusick (Project Director), S. Happe; Oregon Health & Sciences University: E. Orwoll (Principal Investigator), J. Lapidus (co-Investigator), L. Marshall (co-Investigator), M. Abrahamson (Project Director), J. Wiedrick; University of California, San Diego: C. Ginsberg (Principal Investigator), L. Claravall (Study Coordinator), A. Huynh.

Glossary

Contributor Information

Kristine E Ensrud, Department of Medicine, University of Minnesota, Minneapolis, MN 55455, USA. Division of Epidemiology & Community Health, School of Public Health, University of Minnesota, Minneapolis, MN 55455, USA. Center for Care Delivery and Outcomes Research, VA Health Care System, Minneapolis, MN 55417, USA.

Li-Yung Lui, California Pacific Medical Center Research Institute, San Francisco, CA 94107, USA.

Carolyn J Crandall, Department of Medicine, University of California, Los Angeles, CA 90095, USA.

Eric S Orwoll, Bone and Mineral Unit, Oregon Health & Science University, Portland, OR 97239, USA.

Lisa Langsetmo, Center for Care Delivery and Outcomes Research, VA Health Care System, Minneapolis, MN 55417, USA.

John T Schousboe, HealthPartners Institute, Bloomington, MN 55425, USA. Division of Health Policy & Management, School of Public Health, University of Minnesota, Minneapolis, MN 55455, USA.

Howard A Fink, Department of Medicine, University of Minnesota, Minneapolis, MN 55455, USA. Division of Epidemiology & Community Health, School of Public Health, University of Minnesota, Minneapolis, MN 55455, USA. Center for Care Delivery and Outcomes Research, VA Health Care System, Minneapolis, MN 55417, USA. Geriatric Research Education and Clinical Center, VA Health Care System, Minneapolis, MN 55417, USA.

Nancy E Lane, Department of Internal Medicine, University of California – Davis, Sacramento, CA 95817, USA.

Deborah M Kado, Department of Medicine, Stanford University, Stanford, CA 94304, USA. Geriatric Research Education and Clinical Center, VA Health Care System, Palo Alto, CA 94304, USA.

Jane A Cauley, Department of Epidemiology, University of Pittsburgh, Pittsburgh, PA 15261, USA.

Marcia L Stefanick, Department of Medicine, Stanford University, Stanford, CA 94304, USA.

Peggy M Cawthon, California Pacific Medical Center Research Institute, San Francisco, CA 94107, USA.

Funding Support

The Osteoporotic Fractures in Men (MrOS) Study is supported by National Institutes of Health funding. The following institutes provide support: the National Institute on Aging (NIA), the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS), the National Center for Advancing Translational Sciences (NCATS), and NIH Roadmap for Medical Research under the following grant numbers: R01 AG066671, U01 AG027810, U01 AG042124, U01 AG042139, U01 AG042140, U01 AG042143, U01 AG042145, U01 AG042168, U01 AR066160, and UL1 TR000128. This manuscript is also the result of work supported with resources and use of facilities of the Minneapolis VA Health Care System. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the U.S. Department of Veterans Affairs or the United States government.

Disclosures

Dr. Orwoll reports consulting with Amgen, Biocon, Radius, Sanofi, Ultragenyx, Bayer, NextCure; and research support from Mereo. Dr. Kado reports royalties from UpToDate. The other authors have nothing to disclose.

Data Availability

MrOS data is available to the public via the “MrOS Online” website (https://mrosonline.ucsf.edu/).

References