A backbone for FRAX?


  • Nelson B Watts,

    Corresponding author
    1. Bone Health and Osteoporosis Center, University of Cincinnati College of Medicine, Cincinnati, OH, USA
    • Bone Health and Osteoporosis Center, University of Cincinnati College of Medicine, 222 Piedmont Avenue, Suite 6300, Cincinnati, OH 45237, USA.
    Search for more papers by this author
  • Dima L Diab

    1. Bone Health and Osteoporosis Center, University of Cincinnati College of Medicine, Cincinnati, OH, USA
    Search for more papers by this author

  • This is a Commentary on Leslie and Lix (J Bone Miner Res. 2011;460–467. DOI: 10.1002/jbmr.248).

Whenever I go on a ride, I'm always thinking of what's wrong with the thing and how it can be improved.

—Walt Disney

Every day you may make progress. Every step may be fruitful. Yet there will stretch out before you an ever-lengthening, ever-ascending, ever-improving path. You know you will never get to the end of the journey.

—Winston Churchill

Time is a dressmaker specializing in alterations.

—Faith Baldwin

FRAX will be 3 years old on February 21, 2011. It has grown and matured through several iterations, currently Version 3.2, with most recent changes (at the time of this writing) made on December 8, 2010. It has been adapted for use in 30 countries and is accessible in 14 languages (www.shef.ac.uk/FRAX/).

FRAX uses clinical risk factors (age, sex, body mass index [BMI], prior fragility fracture, a parental history of hip fracture, smoking, excess alcohol intake, use of systemic corticosteroids, rheumatoid arthritis, and secondary osteoporosis) and expected mortality to estimate an individual's 10-year risk of hip fracture and major osteoporosis-related fracture. If bone mineral density (BMD) is not available, FRAX can be used with BMI to estimate fracture risk.1

Since FRAX was introduced, critics have claimed that it does not include enough risk factors. For example, low calcium intake, vitamin D deficiency, fall risk, and high bone turnover markers are not considered.2 Countering those critics are studies showing that more parsimonious models (ie, using only BMD and age or BMD, age, and fracture) appear to perform as well as FRAX.3 A simplified approach may work efficiently in populations where other risk factors occur uncommonly. However, circumstances may dictate the need for the more comprehensive information contained in FRAX (eg, rheumatoid arthritis and glucocorticoids in a rheumatology practice), and some specialties may find that important risk factors are not included in FRAX (eg, oncologists may want to include aromatase inhibitors or androgen-deprivation therapy, and endocrinologists may want to add type 2 diabetes and thiazolidinediones). It is possible to use FRAX and then add the additional risk for these items to the risk determined by FRAX, but such an approach is limited by not knowing if the risk applies similarly across ages, or affects men and women to the same degree, and how the additional factors might interact with other risk factors.

Risk prediction is improved (and “secondary osteoporosis” becomes a dummy button) if BMD is included in the FRAX calculation. BMD at most sites is predictive for any fracture, but site-specific fracture prediction is best with BMD from the site in question (ie, spine BMD for spine fracture and hip BMD for hip fracture).4 BMD of the femoral neck was selected for use in FRAX based on large epidemiologic studies that documented its performance for fracture prediction, especially for assessment of hip fracture risk. Femoral neck BMD was the only site reliably available in the datasets used in the construction of FRAX.

Spine/hip discordance (ie, lower BMD at the spine than at the hip) creates a dilemma with FRAX, which, being “spineless,” considers only femoral neck BMD. What does one do with a patient who is low in the spine and not low in the hip? Melton found that spine T-scores were lower than hip T-scores in about half of women, more often in young than in old women.5 The National Osteoporosis Foundation recommends pharmacologic treatment for patients whose lowest T-score is −2.5 or below in the spine or hip but suggests the use of FRAX when the lowest T-score is −2.4 or better.6 However, a younger woman with a T-score of −2.5 (and without other risk factors) is not at high risk, and an older woman with a spine T-score of −2.4 might be. In Canada, one is allowed to substitute lowest T-score (the spine T-score can be used if its lower than the hip),7 but this probably overestimates fracture risk.

In the current issue of JBMR, Leslie and Lix describe a fracture-prediction model in which spine BMD can be included.8 They used robust data on over 37,000 women from Manitoba, with a “hybrid” output being the sum of hip and other nonvertebral fractures estimated using femoral neck BMD and vertebral fractures using spine BMD. Overall, there was not much difference between the hybrid model (with spine BMD for spine fracture prediction and femoral neck BMD for nonvertebral fracture risk prediction) and FRAX (with femoral neck BMD only), which is not surprising because spine and hip BMD are similar in most women. However, in a subgroup of women with discordance at hip and spine (35% of the validation cohort who had more than 1 SD difference between spine and femoral neck T-scores), the performance of the hybrid model was better than FRAX using femoral neck BMD only: 6.1% moved from a low-risk to a moderate-risk category and 7.5% moved from a moderate-risk to a high-risk category; fewer than 1% moved to a lower-risk category.

Some women, particularly younger women, have lower T-scores in the spine than in the hip (spine/hip discordance).5 For such women, improvement in fracture risk estimation using the hybrid model is significant in those at moderate risk under the femoral neck model. Those at low risk under the femoral neck model are unlikely to move to high risk by adding lumbar spine measurements, and those already at high risk do not need further risk stratification for clinical purposes.

The rise in vertebral fracture rates is linear,9 so the effects should be consistent across ages. The hybrid model has been shown to work for women, but what about men? It works in Canada, but what about elsewhere? And the model now has been published but is not available at the FRAX Web site or elsewhere on the Internet. Leslie and Lix have demonstrated that it is possible to incorporate spine BMD into a facture-prediction model, but additional steps are needed to bring this to point of clinical utility.


NBW is co-founder, stockholder, and director of OsteoDynamics. He has received honoraria for lectures from Amgen, Novartis, and Warner Chilcott; consulting fees from Amgen, Arena, Baxter, InteKrin, Johnson & Johnson, Lilly, Medpace, Merck, NPS, Orexigen, Pfizer/Wyeth, Takeda, Vivus, and Warner Chilcott; and through the University of Cincinnati, research support from Amgen, Merck, and NPS. Both authors state that they have no conflicts of interest.