Clinical Factors May Predict Fracture Risk Better Than FRAX

MONTREAL — The clinical risk factors of age and bone mineral density at the hip appear to predict the probability of a hip fracture or a major osteoporotic fracture significantly better than the World Health Organization's more complex Fracture Risk Assessment Tool, according to an analysis of data from a prospective study of 6,252 older white women.

The results suggest that “the addition of [complex] clinical risk factor information to age and BMD alone does not enhance the prediction of these fractures in older women,” said Dr. Kristine E. Ensrud, professor of medicine at the University of Minnesota, Minneapolis.

The FRAX algorithm is designed to predict only the 10-year probability of a hip fracture or a major osteoporotic fracture. It builds a risk profile based on nine clinical risk factors (age, sex, prior history of fracture, oral glucocorticoid use, presence of rheumatoid arthritis, parental history of hip fracture, smoking status, alcohol consumption, and body mass index).

Dr. Ensrud and her colleagues used data from 6,252 participants in the Study of Osteoporotic Fractures, which enrolled women during 1986–1988. Hip BMD measurements and all of the FRAX clinical risk factors were available for these women, who had an average age of 71 years. Incident fractures were confirmed in 98% of cases reported during the study's 10-year follow-up period.

A combination of age and hip BMD had significantly greater ability to predict the 10-year risk of a hip fracture than did the FRAX algorithm alone, based on area-under-the-curve (AUC) statistics that the investigators calculated from receiver operating characteristic curves that were built with data from the study. However, the AUC statistic that was derived from the age-plus-hip-BMD model (0.76) was similar to that obtained with FRAX plus hip BMD (0.75). The AUC statistic for FRAX alone was 0.71. (An AUC of 0.50 reflects a predictive ability equal to chance.)

Similarly, age plus hip BMD had a significantly greater ability to predict both major osteoporotic fractures (hip, clinical spine, wrist, or humerus) and clinical fractures (defined as nonvertebral or clinical vertebral fractures) than did the FRAX algorithm alone, Dr. Ensrud reported at the annual meeting of the American Society for Bone and Mineral Research.

To refine their analysis further, Dr. Ensrud and her associates compared the proportion of women across the models who were classified in the highest decile of fracture risk and who actually experienced a fracture outcome. This type of assessment enhances the clinical usefulness of a risk prediction model because it contains a higher proportion of women who actually experienced the outcome in question, she said.

This simple model of age and prior fracture history (rather than hip BMD) also predicted the 10-year risk of hip, osteoporotic, and clinical fractures just as well as the FRAX algorithm alone did. This finding suggests that “in a setting where BMD is not available, the addition of [complex] clinical risk factor information to age and prior fracture history alone does not enhance the prediction of these fractures in older women,” Dr. Ensrud said.

The study was funded by the National Institute on Aging.

MONTREAL — The clinical risk factors of age and bone mineral density at the hip appear to predict the probability of a hip fracture or a major osteoporotic fracture significantly better than the World Health Organization's more complex Fracture Risk Assessment Tool, according to an analysis of data from a prospective study of 6,252 older white women.

The results suggest that “the addition of [complex] clinical risk factor information to age and BMD alone does not enhance the prediction of these fractures in older women,” said Dr. Kristine E. Ensrud, professor of medicine at the University of Minnesota, Minneapolis.

The FRAX algorithm is designed to predict only the 10-year probability of a hip fracture or a major osteoporotic fracture. It builds a risk profile based on nine clinical risk factors (age, sex, prior history of fracture, oral glucocorticoid use, presence of rheumatoid arthritis, parental history of hip fracture, smoking status, alcohol consumption, and body mass index).

Dr. Ensrud and her colleagues used data from 6,252 participants in the Study of Osteoporotic Fractures, which enrolled women during 1986–1988. Hip BMD measurements and all of the FRAX clinical risk factors were available for these women, who had an average age of 71 years. Incident fractures were confirmed in 98% of cases reported during the study's 10-year follow-up period.

A combination of age and hip BMD had significantly greater ability to predict the 10-year risk of a hip fracture than did the FRAX algorithm alone, based on area-under-the-curve (AUC) statistics that the investigators calculated from receiver operating characteristic curves that were built with data from the study. However, the AUC statistic that was derived from the age-plus-hip-BMD model (0.76) was similar to that obtained with FRAX plus hip BMD (0.75). The AUC statistic for FRAX alone was 0.71. (An AUC of 0.50 reflects a predictive ability equal to chance.)

Similarly, age plus hip BMD had a significantly greater ability to predict both major osteoporotic fractures (hip, clinical spine, wrist, or humerus) and clinical fractures (defined as nonvertebral or clinical vertebral fractures) than did the FRAX algorithm alone, Dr. Ensrud reported at the annual meeting of the American Society for Bone and Mineral Research.

To refine their analysis further, Dr. Ensrud and her associates compared the proportion of women across the models who were classified in the highest decile of fracture risk and who actually experienced a fracture outcome. This type of assessment enhances the clinical usefulness of a risk prediction model because it contains a higher proportion of women who actually experienced the outcome in question, she said.

This simple model of age and prior fracture history (rather than hip BMD) also predicted the 10-year risk of hip, osteoporotic, and clinical fractures just as well as the FRAX algorithm alone did. This finding suggests that “in a setting where BMD is not available, the addition of [complex] clinical risk factor information to age and prior fracture history alone does not enhance the prediction of these fractures in older women,” Dr. Ensrud said.

The study was funded by the National Institute on Aging.

MONTREAL — The clinical risk factors of age and bone mineral density at the hip appear to predict the probability of a hip fracture or a major osteoporotic fracture significantly better than the World Health Organization's more complex Fracture Risk Assessment Tool, according to an analysis of data from a prospective study of 6,252 older white women.

The results suggest that “the addition of [complex] clinical risk factor information to age and BMD alone does not enhance the prediction of these fractures in older women,” said Dr. Kristine E. Ensrud, professor of medicine at the University of Minnesota, Minneapolis.

The FRAX algorithm is designed to predict only the 10-year probability of a hip fracture or a major osteoporotic fracture. It builds a risk profile based on nine clinical risk factors (age, sex, prior history of fracture, oral glucocorticoid use, presence of rheumatoid arthritis, parental history of hip fracture, smoking status, alcohol consumption, and body mass index).

Dr. Ensrud and her colleagues used data from 6,252 participants in the Study of Osteoporotic Fractures, which enrolled women during 1986–1988. Hip BMD measurements and all of the FRAX clinical risk factors were available for these women, who had an average age of 71 years. Incident fractures were confirmed in 98% of cases reported during the study's 10-year follow-up period.

A combination of age and hip BMD had significantly greater ability to predict the 10-year risk of a hip fracture than did the FRAX algorithm alone, based on area-under-the-curve (AUC) statistics that the investigators calculated from receiver operating characteristic curves that were built with data from the study. However, the AUC statistic that was derived from the age-plus-hip-BMD model (0.76) was similar to that obtained with FRAX plus hip BMD (0.75). The AUC statistic for FRAX alone was 0.71. (An AUC of 0.50 reflects a predictive ability equal to chance.)

Similarly, age plus hip BMD had a significantly greater ability to predict both major osteoporotic fractures (hip, clinical spine, wrist, or humerus) and clinical fractures (defined as nonvertebral or clinical vertebral fractures) than did the FRAX algorithm alone, Dr. Ensrud reported at the annual meeting of the American Society for Bone and Mineral Research.

To refine their analysis further, Dr. Ensrud and her associates compared the proportion of women across the models who were classified in the highest decile of fracture risk and who actually experienced a fracture outcome. This type of assessment enhances the clinical usefulness of a risk prediction model because it contains a higher proportion of women who actually experienced the outcome in question, she said.

This simple model of age and prior fracture history (rather than hip BMD) also predicted the 10-year risk of hip, osteoporotic, and clinical fractures just as well as the FRAX algorithm alone did. This finding suggests that “in a setting where BMD is not available, the addition of [complex] clinical risk factor information to age and prior fracture history alone does not enhance the prediction of these fractures in older women,” Dr. Ensrud said.

The study was funded by the National Institute on Aging.