(73/223) worked within an interdisciplinary team, and 85% (192/ 225) used fee-for-service billings. Patients were older in practices of men FPs (66.4% versus 56.6% of patients over the age of 65 years) as compared with women FPs. The odds of appropriate X-ray ordering were lower if the physician was a woman (OR: 0.76; 95% CI: 0.63, 0.92), provide house calls (OR: 0.71; 95% CI: 0.59, 0.85), worked within an interdisciplinary team (OR: 0.70; 95% CI: 0.56, 0.88), or used fee-for-service billings (OR: 0.66; 95% CI: 0.51 0.85). The odds of appropriate X-ray ordering were higher if the physician provided afterhour call coverage (OR: 1.87; 95% CI: 1.46, 2.38) or worked in a teaching practice (OR: 1.29; 95% CI: 1.06,1.58). Conclusions: Physician characteristics may influence X-ray order- ing and the appropriate management of the disease. Conflict of interest: GI: none; AP: honoraria, grants received, or consultancies-Eli Lilly and Company, Merck Frosst, Amgen Inc, The Alliance for Better Bone Health, Novartis Pharmaceuticals Corpora- tion; LT: none; AG: none; BK: honoraria or consultancies — The Alliance for Better Bone Health; AH: honoraria or consultancies — Eli Lilly and Company, Merck Frosst, NPS-Allelix, Zelos Therapeutics, Servier, Pfizer Pharmaceuticals USA, Novartis Pharmaceuticals Cor- poration, The Alliance for Better Bone Health, GlaxoSmithKline Consumer Healthcare; FJ: none; AW: employee — Procter and Gamble Pharmaceuticals; JDA: honoraria, grants received, or consultancies: — Eli Lilly and Company, Merck Frosst, Amgen Inc, The Alliance for Better Bone Health, Novartis Pharmaceuticals Corporation, GlaxoS- mithKline Consumer Healthcare, Servier, Roche, Wyeth. doi:10.1016/j.bone.2009.03.213 P302 Comparison of quantitative ultrasound and dual-energy X-ray absorptiometry for prediction of 10-year absolute risk of fracture among older men and women A. Moayyeri a, *, S. Kaptoge a , R.N. Luben a , S. Bingham b , N.J. Wareham c , J. Reeve d , K. Khaw a a Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK b MRC Dunn Human Nutrition Unit, University of Cambridge, Cambridge, UK c MRC Epidemiology Unit, University of Cambridge, Cambridge, UK d Department of Medicine, University of Cambridge, Cambridge, UK The role of quantitative ultrasound (QUS) in the management of osteoporosis is debated. Although QUS of the heel is known to be correlated with bone mineral density (BMD), the performance of QUS for prediction of fractures in comparison to dual-energy X-ray absorptiometry (DXA), the current reference standard for diagnosis of osteoporosis, is unclear. We examined this in a sample of men and women in the European Prospective Investigation into Cancer (EPIC)-Norfolk who had both heel QUS and hip DXA between 1995 and 1997 and were followed for incident fractures up to March 2007. From 1,454 participants (701 men) aged 65–76 years at baseline, 79 suffered a fracture over 15,567 person-years of follow- up (mean 10.3 ± 1.4 years). In a sex-stratified multivariate Cox proportional-hazard model including age, height, weight, past history of fracture, smoking, alcohol intake and DXA total hip BMD, 1 SD decrease in BMD was associated with a hazard ratio for fracture of 2.28 (95% CI 1.75–2.98). In a multivariate model with heel broadband ultrasound attenuation (BUA) in place of BMD, hazard ratio for 1 SD decrease in BUA was 2.02 (95% CI 1.54–2.65). While the global measures of model fit (including Bayesian and Akaike's information criteria, LR chi-square, D-statistic, Nagelkerke's and Cox- Snell R-square) slightly favored the model with DXA measure, measures of discrimination (area under ROC curve 0.685 vs. 0.679) and calibration (Hosmer–Lemeshow statistic 0.82 vs. 0.45) showed relative superiority of the model with BUA. Using the alternate Cox models with DXA and BUA, we calculated exact 10-year absolute risk (probability) of fracture for all participants and categorized them in three groups of <5%, 5% to <15%, and >=15%. Comparison of groupings based on DXA vs. BUA models showed a total reclassifica- tion of 24.9% of participants. The predicted risks for individual participants using BUA model were closer to the observed risks (especially among the 52.2% reclassified highest-risk group). This study shows that the power of QUS for prediction of prospective fractures and calculation of 10-year probability of fractures among the elderly is comparable to that of DXA. Given the ease and low cost of ultrasound, risk assessment tools (such as newly-developed FRAX tool) may include QUS measures in their algorithms to attain more accurate estimates of fracture risk. Further studies or meta-analyses are required to clarify whether QUS with clinical risk factors would be cost-effective for primary care case-finding when DXA is not available. Conflict of interest: None declared. doi:10.1016/j.bone.2009.03.214 P303 Length of follow-up required to achieve an accurate estimate of 10-year absolute risk of fracture among the elderly A. Moayyeri a, *, S. Kaptoge a , S. Bingham b , N.J. Wareham c , R.N. Luben a , J. Reeve d , K. Khaw a a Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK b MRC Dunn Human Nutrition Unit, University of Cambridge, Cambridge, UK c MRC Epidemiology Unit, University of Cambridge, Cambridge, UK d Department of Medicine, University of Cambridge, Cambridge, UK Estimation of 10-year absolute fracture risk for older populations is now recommended by the World Health Organization (WHO). These estimates should ideally come from country-specific prospec- tive studies that follow representative members of the community for a long time. Given practical and resource constraints, cohort studies usually follow participants for a shorter interval (typically 4– 7 years) and extrapolate their results to generate 10-year predic- tions. The most widely used statistical methods are based on exponential distribution of fracture risk and using Poisson regres- sion (as used in FRAX study by the WHO Collaborating Group). However, the length of follow-up necessary to achieve an accurate estimate of 10-year risk is not clear. The original cohort in the European Prospective Investigation into Cancer-Norfolk study comprised 25,312 men and women aged 39–79 years living in the general community in Norfolk, United Kingdom, recruited in 1993– 1997 and followed-up for incident fractures to 2007 (mean follow- up 11.4±1.3 years). Sex-specific Poisson regression models adjust- ing for age, history of fracture, height, body mass index, smoking and alcohol consumption were employed to obtain 10-year absolute fracture risks in 10 different sub-cohorts with one year added interval of follow-up (the original cohort was re-arranged to produce 10 cohorts with follow-up period of 1, 2, 3, …, and 10 years; incident fractures after the follow-up period were censored for each cohort). While 758 fractures were observed in the first 10 years of follow-up among EPIC-Norfolk participants, models with 5, 6, 7, 8, 9, and 10 years of follow-up, respectively, predicted this number to be 423, 491, 569, 638, 685, and 761 fractures. Choosing arbitrary cutoffs for definition of high-risk (10- year risk of >6% in <65-years-old and >12% in ≥65-years-old participants to reflect the fracture incidence in the original cohort) and using the model with a 10-year follow-up as the gold standard, the sensitivity of the model with 6 years follow-up was only 49.2%. This number increased for models with 7, 8, and 9 years of follow- up, respectively, to 55.8%, 79.6%, and 91.8%. This study suggests that Abstracts / Bone 44 (2009) S339–S450 S367