点击显示 收起
National Institute for Food and Nutrition Research Via Ardeatina, 546 00178 Rome Italy
Dear Sir:
We appreciate Stensel's comments on our article (1). He questions our conclusion that "BMR is depressed in anorexia nervosa ... suggesting that the metabolic activity of the active tissue mass may been reduced." Stensel points out that "a plausible explanation for the lower BMR values in subjects with anorexia nervosa after control for fat-free mass is their greatly reduced fat mass."
We agree that differences in fat mass (FM) can explain part of the interindividual variability in basal metabolic rate (BMR), but only to a very limited extent, and we maintain that the mechanism of this effect does not invalidate our conclusions. FM consists of ether-soluble lipids and, as such, is to be considered metabolically inert. The contribution to BMR that has been attributed to FM by some authors (2–4), especially in obese women, is still the subject of controversy; a plausible explanation can be found only by equating FM with adipose tissue (2). Lipids are deposited in the body as adipose tissue, which requires a cellular matrix that, in healthy subjects, represents 20% of this tissue. This cellular matrix is metabolically active and therefore can contribute to BMR. The proportion of cellular matrix in adipose tissue is inversely associated with FM, and it has been calculated that in obese subjects—whose fat cells are more tightly packed with lipids than are those of lean subjects—it would represent no more than 16%; in anorectic individuals—whose adipocytes have lost most of their lipid deposit—it might increase to 21% (5). Therefore, we can expect that the relatively larger cellular matrix and the smaller fat deposit of the anorectic patients might result in an increase in BMR after control for FM.
Stensel suggests performing an analysis of covariance (ANCOVA) with simultaneous inclusion of FM and fat-free mass (FFM). This analysis is inappropriate statistically for our data because the regression slopes of BMR versus FM of the 2 groups were significantly different (P < 0.01), whereas the regression slopes of BMR versus body weight and FFM were parallel. The key assumption of homogeneity of the regression slopes is required for performing ANCOVA and the violation of this assumption would lead to gross misinterpretation of results and introduce serious biases (6, 7). Because of this limitation, we had performed an extension of ANCOVA suitable for use with heterogeneous regression slopes (Johnson-Neyman procedure; 8). This statistical technique computes "regions of significant differences" and was applied by other authors for a similar problem (7). The results confirm that FM does not represent a significant covariate in the model (P = 0.94) and that FFM and FM are not unique estimators of BMR. When the 95% CI of nonsignificance was calculated, BMR controlled for both FFM and FM was significantly lower in subjects with FM <10 kg. For these reasons, we maintain our conclusions that BMR is reduced in anorexia nervosa and that this observation suggests a reduction in the metabolic activity of the active tissue mass in anorexia nervosa.
REFERENCES