Lafita and Garcia question the usefulness of dual-energy x-ray absorptiometry measurements of bone mineral density (BMD) in the identification of persons at risk for fracture. In point of fact, BMD is the most powerful measurable determinant of risk for fracture available today. The relative risk for fracture with each standard deviation reduction in BMD is 1.5 to 2.6 (1). Furthermore, as far as we know, antifracture efficacy has not been demonstrated for any agent in the absence of at least modest increases in BMD. Although resorption markers such as urinary pyridinolines and N-telopeptide are predictors of fracture, they are not as powerful as BMD itself (2, 3). We measured urinary hydroxyproline excretion in this study and found the expected changes in response to HRT. In our experience, urinary hydroxyproline measurement is as good as or better than the newer markers as an index of bone resorption. Its drawback is that it is technically difficult and time-consuming.

Lafita and Garcia indicate that alternatives to HRT should be considered for elderly persons because of potential side effects. We agree. Indeed, our results suggest that low-dose HRT combined with adequate calcium and vitamin D constitutes just such a reasonable alternative to conventional doses of HRT (because it results in fewer side effects). It is also much less expensive than many other alternatives.

Elasy and Levinsky present concerns about the data analysis and presentation. The data were analyzed by repeated-measures analysis of variance (SAS PROCedure MIXED [4]). This analysis showed that the difference between the groups was highly statistically significant for all sites but the femoral neck. The fact that not all time points showed a significant difference between placebo and treatment is not germane. The sentence regarding the minimal important difference of 4.5% was inserted into the manuscript at the request of reviewers and the editor in response to their need to include the power and sample size calculations that were part of the original protocol. The 4.5% was not a criterion for clinical importance. It was the difference for which our sample size had an a priori 80% power, given assumptions about data dispersion. Even a 1% difference is clinically important for bone. Whether because of better than anticipated measurement precision or good luck, we demonstrated an improvement in BMD that was both statistically significant and clinically important.

Regarding clinical outcomes, reduction in fracture risk by increasing or maintaining BMD can take place without apparent improvement in clinical outcome (that is, improvement in the risk for fracture). Our study did not have power to detect antifracture efficacy.

We thank Barnes and Bess for pointing out the unfortunate error that occurred when the paper was prepared for printing.