Confused thinking about discrepancies in predicted fracture risk in older people
================================================================================

* Eugene McCloskey
* Juliet Compston
* John Kanis

Bolland and colleagues argue that estimates of fracture in older people should use a short time horizon of three to five years yet ignore competing mortality because it precludes effective treatment of these patients.1

They assume that adjustment for mortality risk in FRAX is based only on average mortality rates for the population, but the tool accommodates the fact that many risk factors that predict fracture risk also influence mortality (older age, previous fracture, low body mass index, smoking).2 The incorporation of competing mortality directly addresses the concerns raised over the time horizon used; a three to five year time horizon in older people is exactly what FRAX produces (table 2 of the article). If life expectancy is less than 10 years, then the fracture probability equals the remaining lifetime risk of fracture (table 2).

The authors ignore well recognised systematic differences in the output of the fracture prediction tools (fig 2), reflecting differences in calibration, input risk variables, outcome fractures, and incorporation of competing mortality. It is nonsensical to compare the tools against intervention thresholds that have been derived for only one of the tools (FRAX).

The real problem is not FRAX, but the setting of intervention thresholds and the complexities therein, which Bolland and colleagues do not address. Fracture rates alone as an outcome show an exponential rise with age so that treatment is indicated in all older people. Like Bolland and colleagues, we are keen to ensure optimal and appropriate use of osteoporosis drugs at all ages, but there are risks of both undertreatment and overtreatment. The authors also fail to acknowledge the importance of clinical judgment. An intervention threshold is a guideline not an absolute; clinical judgment is espoused within all guidelines, including that of the National Osteoporosis Guideline Group.

## Notes

**Cite this as:** *BMJ* 2013;346:f1427

## Footnotes

*   Competing interests: The authors were involved in the development of FRAX or the NOGG guideline.

## References

1.  Bolland MJ, Jackson R, Gamble GD, Grey A. Discrepancies in predicted fracture risk in elderly people. BMJ2013;346:e8669. (21 January.)
    
    [FREE Full Text](https://www.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiRlVMTCI7czoxMToiam91cm5hbENvZGUiO3M6MzoiYm1qIjtzOjU6InJlc2lkIjtzOjE3OiIzNDYvamFuMjFfMS9lODY2OSI7czo0OiJhdG9tIjtzOjIzOiIvYm1qLzM0Ni9ibWouZjE0MjcuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

2.  Kanis JA, on behalf of the WHO Scientific Group. Assessment of osteoporosis at the primary health-care level. Technical report. University of Sheffield, 2008. [www.shef.ac.uk/FRAX/pdfs/WHO\_Technical\_Report.pdf](http://www.shef.ac.uk/FRAX/pdfs/WHO_Technical_Report.pdf).