点击显示 收起
Cell Biology and Immunology Group
Wageningen University
PO Box 338
6700 AH Wageningen
Netherlands
E-mail: hans.verhoef{at}wur.nl
Dear Sir:
I respectfully disagree with much written in the letter by Wieringa et al. They are misguided that we equated plasma retinol concentrations with vitamin A status (1); instead, we quoted a recommendation from leading experts, which was reviewed by the International Vitamin A Consultative Group, that serum retinol concentration is the prime indicator for routine assessment of the occurrence and degree of vitamin A deficiency (2). This recommendation and the description in our article about the predictive role of serum retinol concentration agree with the earlier conclusion by Olson (3) that "serum vitamin A values are useful nutritional indicators when they are very low when they fall in the normal range [20-300 µg/g liver], they are not particularly helpful."
We examined the validity of the relation between modified-relative-dose-response (MRDR) values and serum retinol concentrations (1) because we were concerned that several articles, including one by Wieringa et al (4), based inferences on an assessment of this relation, which we viewed as incorrect. Thus, it is surprising that Wieringa et al now assert that this relation is not the issue.
The values provided by Wieringa et al appear to suggest that the mean MRDR can be measured with much more precision than can mean plasma retinol concentrations. Although they do not provide a source for these data, the MRDR values they provided contradict their own data (Figure 3 in reference 4), which suggests that the variance, and thus the SEM of the MRDR, is larger than that of the mean plasma retinol concentration, particularly in the range of plasma retinol concentrations that indicates vitamin A deficiency (<0.70 µmol/L). More importantly, we questioned the validity of the MRDR, not its precision.
We agree with Wieringa et al that the key question regarding the validity of the MRDR is how well it corresponds to vitamin A status; in addition, to judge whether the MRDR is a better indicator than is serum retinol concentration, it would be important to know whether it agrees better with vitamin A status. Unfortunately, Wieringa et al did not refer to published data from human or animal studies that would validate the MRDR on the basis of its relation with liver concentrations of vitamin A. As noted in our article, the provisional cutoffs for abnormal MRDR values to define vitamin A deficiency appear instead to have been based on studies with few subjects from populations believed to be vitamin A sufficient.
We stated that "if serum retinol concentration reflects vitamin A status....then estimators, such as the MRDR, that incorporate the reciprocal of serum retinol concentration will, by definition, be biased." It makes sense to incorporate the reciprocal of an indicator in a derived estimator if this would linearize the relation with what it is supposed to measure. For example, incorporation of the reciprocal of the serum ferritin concentration in the ratio of serum concentrations of soluble transferrin receptor to ferritin is helpful because this derived estimator is linearly related to body iron stores (5). To explore this idea further for serum retinol concentration, I redrew its "idealized" (3) relation with liver concentrations of retinol in the range of low vitamin A stores, ie, liver concentrations <50 µg/g (Figure 1, panel A). As shown in the same figure (panel B), plotting the reciprocal of serum retinol concentration is not helpful because in the range of vitamin A that indicates a deficiency (liver concentrations <20 µg/g), it does not produce a linear relation with liver concentrations of vitamin A; by contrast, this relation is even less linear than that with serum retinol concentration.
FIGURE 1.. Idealized relation between serum retinol concentrations and low (<50 µg/g) liver vitamin A concentrations (panel A), redrawn from reference 3, and between the reciprocal of serum retinol concentrations and liver vitamin A concentrations (panel B) derived from panel A.
Contrary to what was stated by Wieringa et al, we assumed in our models 1 and 2 that supplementation with a fixed dose of retinyl ester produces a fixed effect on average. Our assumption that serum retinol 5 h after dosing (R5) is independent of serum retinol at baseline (R0) was guided by evidence from rat studies that R5 values in vitamin A-depleted rats "approximated the homeostatic levels repeatedly found in our laboratory for rats receiving identical vitamin A-sufficient diets" (6). On the basis of this observation (Figure 2 in reference 6), the RDR was originally conceived as "the degree by which plasma vitamin A levels otherwise ‘normal’ to an individual are reduced by deficiencies in hepatic stores. " Contrary to what was stated (without reference to published evidence) by Wieringa et al, this would imply that R0 and R5 are independent. Wieringa et al seem to suggest that the regression-to-the-mean effect as reported in our models 1 and 2 were due to the independence between R0 and R5. This would be incorrect: such an effect cannot exist when 2 variables are independent, and the values for these 2 variables in our models were generated independently and randomly by computer. The regression-to-the-mean effect that we reported was due to the fact that R0 occurs both on the x and y axes of Figures 1 and 2 in reference 1.
Our concern with the article by Wieringa et al (4) relates mainly to their analysis of the relation between the MRDR and plasma retinol concentrations; we acknowledged that their MRDR values were less than expected for the serum retinol concentration in infants who had received iron. This does not mean that the mathematical artifact we described did not play a role in their analysis, which it must have for the reasons outlined in our article.
We referred to the article by Wieringa et al (4) as being from "our group" because it was conducted under the academic supervision of Clive West. To the extent that we may have given the erroneous impression that the work by Wieringa et al (4) was carried out exclusively by the group at Wageningen University, I acknowledge that our wording was unfortunate.
Our assertion that the MRDR is not normally distributed was indeed not based on the original data by Wieringa et al (4), but rather on mathematical considerations, and was supported by visual inspection of our Figure 3 (1) and their Figure 3 (4).
ACKNOWLEDGMENTS
HV was partially supported by a grant from the Netherlands Foundation for the Advancement of Tropical Research (W 01.65.215.00). CEW, who coauthored the article referred to in this letter, died on 27 August 2004.
REFERENCES