We have previously shown that hip bone mineral density (BMD), heel broadband ultrasound attenuation (BUA) and bone resorption markers are independent predictors of hip fracture in elderly women. We investigated whether a combination of these three parameters could improve the predictive value of a single test in a nested case-control analysis (75 hip fractures and 228 age-matched controls) of the EPIDOS prospective study comprising 7598 healthy women 75 years of age and older followed prospectively for a mean 22 months. At baseline, prior fracture, femoral neck BMD by dual-energy X-ray absorptiometry (DXA), heel BUA and urinary type I collagen C-telopeptide breakdown products (CTX) were assessed. The area under the receiver operating characteristic curve was significant for the three diagnostic tests, heel BUA being the best single predictor. The added value of urinary CTX to either BMD or BUA depends on the cutoff point chosen to define patients at risk and on the therapeutic strategy that is considered. Defining patients at risk as those with low BMD (or low BUA) or high CTX resulted in a significant increase in the sensitivity compared with BMD or BUA alone--a strategy that could be applied when a broad treatment is considered. However, this increased sensitivity was also obtained simply by increasing the BMD and BUA cutoffs, suggesting that a combination of CTX with BMD/BUA is not useful for that type of treatment strategy. Conversely, defining patients at risk as those with both low BMD and high CTX increases the specificity (88% vs 78%) with a similar number of hip fracture patients being identified (30% vs 32%)--a combination that could be useful when the strategy is to target treatment to a subset of high-risk patients. This strategy appears to be more cost-effective than bone mass measurement alone as indicated by the 37% fewer patients who need to be treated to avoid one fracture per year. If DXA or ultrasound is not available, the combination of a bone resorption marker with a history of any type of fracture after the age of 50 years gave a predictive value similar to that obtained with femoral neck BMD or heel BUA alone, for both types of treatment strategy. We conclude that the combination of urinary CTX with hip BMD could be useful for the identification of elderly women at high risk for hip fracture, resulting in higher specificity for a given sensitivity threshold than BMD measurement alone. If DXA is not available, the combination of history of fracture and urinary CTX performs as well as hip BMD to assess hip fracture risk in elderly women.