| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Letters to the Editor |
Maine Center for Osteoporosis Research & Education Bangor, Maine 04401
In response to the letter submitted above, we take issue with several comments about the article by Rosen et al. (1).
First, we have published two manuscripts that report information from this randomized longitudinal trial examining the role of biochemical markers in predicting skeletal response to calcium supplementation vs. hormone replacement therapy (HRT) in early postmenopausal women. In the first paper, Chestnut and my colleagues report primarily on study design and the a priori primary outcomes measures: urinary N-telopeptide (NTx) and bone mineral density (BMD). The overwhelming majority of data presented in that paper centered on NTx, with virtually no additional information about other markers; in fact there was only one table (Table II) containing correlation coefficients and ROCs, and a short paragraph in the Results section in respect to the other turnover indices. Complete data sets on free deoxypyridinoline (fDpd) were not available when the first manuscript was prepared and therefore could not be included in the original paper. Indeed, there was an overwhelming amount of data beyond the primary outcome measures, which necessitated, as customary in large multicenter trials, a second manuscript (2), comparing performance of other available biochemical markers, and their utility in predicting skeletal response to calcium or HRT.
Second, concern was raised by the authors of this letter as to the comparison between changes in NTx and other markers after the first month of HRT, especially in respect to inherent variability in measures. The use of the term "earliest and most significant" is a qualitative descriptor of the change in NTx and osteocalcin. These indices had the most rapid and significant response to HRT (P < 0.0001) after 1 month. Discussion of variability on the ability to measure a change is found in the Discussion section as noted:
"... however, when subjects’ coefficients of variation were compared to the percent change in the markers after the initiation of HRT, the mean change in NTx due to therapy was always greater than the biological variability, even at the early time points. This was not the case for fDpd. As an example, after 1 month of HRT, the mean percent change in NTx was -28% compared with the mean change in fDpd of -10%. After 6 months, the mean change in NTx was -42% compared with that in fDpd, which was -22%".(1)
Third, issues were raised about comparing 1st and 4th quartiles for fDpd. We could not find the reason why HRT subjects in the 3rd quartile of fDpd at baseline had a greater percent change in BMD after 1 year than those with higher fDpd in the 4th quartile. The data stand as analyzed several different times and as reported in the manuscript. We chose to use quartiles rather than medians because the goal of the analysis was to provide relevant statistics for an individual patient rather than an entire group. Indeed, this is a standard method of representing data from epidemiologic studies. Moreover, odds ratio analysis provides further evidence of the marker’s ability to predict gain in BMD if on HRT, or loss if not on HRT.
Fourth, concern was raised about ROC analysis. Providing an ROC curve at all time points would not provide additive information and would introduce problems inherent in multiple testing. The 6th month time point was chosen because it was the time when the resorption markers were closest to their nadir, the latest time point before the end of the study, and formation markers had decreased by that time point.
Fifth, there was some question about the "control" or calcium only group. The control group was included in the study as a control. The study was not designed to compare two intervention groups, calcium supplementation vs. HRT. As Chesnut et al. (2) noted, markers of bone remodeling did not change significantly in the control group throughout the study (average: -3.0% P = 0.20). Therefore we chose to represent marker values as a mean over the entire study.
Sixth, the authors inquire about use of other skeletal sites. The purpose of this paper was not to compare changes at the spine vs. those at the hip, but rather to compare markers of bone turnover. Chesnut provided sufficient femoral neck BMD to demonstrate therapeutic skeletal responsiveness. As reported previously, HRT has a much greater and more rapid effect on spine than hip BMD, so we chose a priori to examine lumbar BMD responsiveness to HRT (2). This choice of site also fits with ongoing clinical concerns related to changes in spine BMD after prolonged HRT.
Seventh, the issue of percent C.V. is raised. The letter to the editor points out that the published within-patient C.V. for fDpd is 10–16%. However, this does not affect the conclusions that the within-patient variability in fDpd is too great in this study to see a significant change before 6 months, at which time the average percent change due to therapy is only slightly greater than the C.V.
Footnotes
Address correspondence to: Clifford J. Rosen, M.D., Maine Center for Osteoporosis Research & Education, St. Joseph’s Hospital, 360 Broadway, Bangor, Maine 04401.
Received November 26, 1997.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
