点击显示 收起
Obesity Research Center, Weight Control Unit, St Luke's–Roosevelt Hospital, 1090 Amsterdam Avenue, 14th Floor, New York, NY 10025, E-mail: sbh2{at}columbia.edu
Dear Sir:
We acknowledge Watson's points concerning our application of Bland-Altman (B-A) plots (1); this cross-validation analysis method is well described in Watson's letter. As pointed out by Watson, the plot we applied in our paper (2) is not an original B-A plot but rather a modified version [ie, X - Y versus X, where X is skeletal muscle (SM) measured by use of magnetic resonance imaging (MRI) and Y is predicted SM from a developed model). We minimally modified the B-A plot because we wanted to investigate the bias pattern of the between-method difference versus the reference method (ie, MRI-measured SM) rather than the between-method difference versus the average of MRI-measured SM and predicted SM [ie, (X + Y)/2]. The method average does not have implications in our study because predicted SM is not an instrument-measured entity. Although the original form of the B-A plot may be more statistically sound, the modified version is more appropriate for the research questions we addressed.
Concerning the results from Equation 4, the statement in the Discussion is incorrect and the statement in the Results is correct. That is, there was no significant difference between MRI-measured SM and predicted SM (0.4 ± 3.0; P = 0.28). Concerning the results from Equation 6, the statement in the Discussion that "a small bias was observed when the model developed in group A was cross-validated in the nonobese group B subjects" is based on the results in Figure 2B, which shows that the difference tended to increase as MRI-measured SM increased, although the difference was not significant (-0.3 ± 2.7 kg; P = 0.17). We hope that the issues addressed by Watson are now clarified and we thank Watson for his careful reading of our paper.
REFERENCES