点击显示 收起
Institute of Food Research Norwich Research Park Colney, Norwich NR4 7UA United Kingdom E-mail: jack.dainty{at}bbsrc.ac.uk
Dear Sir:
In the article by Lowe et al (1), several experimental methods for assessing fractional zinc absorption (FZA) were compared in 6 women. A rigorous evaluation of the validity of the different experimental techniques is long overdue and we congratulate the authors for their attempt to clarify the issues concerning the various methods being used in different laboratories. The performance of each method was analyzed in relation to the results of a compartmental model, developed and reported on previously (2), that used the same experimental data. Although we read Lowe et al's article with interest, several issues need clarification.
The incorporation of 3 types of data (fecal, urinary, and plasma) was used as the justification for choosing the compartmental model as the gold standard to compare against other methods that used only one set of data (fecal, urinary, or plasma). We suggest that both the quantity and the quality of the data used should be the main criteria, but there is no information on quality, other than the fact that a constant fractional SD of 0.1 was used by the CONSAM program (3) when the tracer data sets were fit in the compartmental model. Are we to assume that the fecal, urinary, and plasma data all had the same uncertainty associated with them? This seems unlikely given that the sample preparations were all different and that the quantity of zinc in each sample varied widely. The precision of the parameter estimates from an earlier report by Lowe et al (2) was generally good, reflecting the excellent structure and design of the model. However, in 5 of the 6 subjects, the CV for the parameter associated with urinary excretion was >60%. On the basis of these results, we estimated that the removal of the urinary data from the model would not weaken it.
A criticism of any model is that it is just that: a model. The modeling process makes gross simplifications of the way the body works and any results from it should be scrutinized for false assumptions, unjustified complexity, and unsubstantiated claims of parameter precision. In Lowe et al's (1) discussion, there was plenty of excellent, well-argued criticism of the other methods used to calculate FZA but no criticism of the compartmental model against which these other methods were compared. Attention should have been drawn to the shortcomings of using modeling in nutritional studies so that other investigators would not be left with the impression that the results from a compartmental model are beyond contradiction.
Another weakness of Lowe et al's study (1) was the small number of data sets used. Detailed metabolic studies are often constrained by the resources available, thus limiting the number of subjects studied, the procedures that can be undertaken, or both. Although the results obtained from the different methods reviewed was interesting, the method of comparison used was not appropriate. The FZA calculated from the compartmental model is based on an equation containing 2 of the rate constants (k1,5 and k6,5), which are simultaneously fitted with the other rate constants to the data set provided. There are uncertainties associated with these parameters that were not stated in Lowe et al's (1) article, although these uncertainties were addressed in their previous study (2) in which the compartmental model was developed. In their more recent article, Lowe et al (1) used the mean and SD of the FZA calculated from the compartmental model, generated from the 6 subjects, as their reference point. Calculation of the SD of the 6 results could give a misleading picture of how good the estimate of the reference FZA is. For instance, if the uncertainties concerning the rate constants k1,5 and k6,5 are large for each individual subject's data, the corresponding uncertainty concerning each calculated FZA will be large. If, however, the difference between each of the 6 calculated FZAs is, by chance, small, the SD of the mean FZA will be small. This is the drawback to having only 6 data sets and it applies equally to other methods used to calculate FZA. The conclusion that "We therefore recommend the DITR technique with use of a spot urine sample collected 2 d after tracer administration..." cannot, therefore, be justified from the data provided in Lowe et al's (1) article.
Because of the limited number of subjects in Lowe et al's (1) study, it would have been more worthwhile to analyze each subject's data separately and to examine the random and systematic errors associated with both the collection of that data and the calculation of the FZA. The different methods could then have been assessed genuinely within each individual. Although this type of analysis would not make clear which absorption value is the most accurate, it would enable investigators to gauge which method produces a value for FZA with the lowest associated uncertainty.
REFERENCES