点击显示 收起
Osteoporosis Research Center
Creighton University Medical Center
601 North 30th Street, Suite 4841
Omaha, NE 68131
E-mail: rheaney{at}creighton.edu
Dear Sir:
In a recent issue of the Journal, Drewnowski (1) usefully summarized the important effort to provide the general public with information about the nutritional quality of various foods. As he pointed out, today's practice often defines a food as "good" on the basis of what it does not contain. Current nutrition labeling laws, in fact, list the presumably "bad" nutrients first, usually in bold-faced type, and the "good" nutrients afterward. This emphasis on potential harm may reflect an antinutrition bias in medicine, a point noted by Goodwin and Tangum (2). It may, as well, be a residue of the disease paradigm that prevailed 100 y ago (at the birth of nutrition as a science), which held that all disease was caused by external invaders, either bacteria or toxins (3). To some extent, current nutritional policy and labeling practices continue to treat foods as potential toxins. Thus, the development of a food "goodness" score is a much-needed initiative.
For the naturally nutrient rich (NNR) score he favors, Drewnowski noted a computational difficulty with respect to foods that have a low energy density but a high content of certain micronutrients. He used red peppers as an example. For vitamin C, they yield a percentage daily value (DV) of 18 900. When incorporated into an average, such very high values distort the overall score, especially when the other 13–15 nutrients that make up the composite score for a food have relatively small values. Drewnowski noted that one approach has been to truncate high single-nutrient values by assigning to them an arbitrary maximum figure. But this is a problem common to all ratios, and preferable methods of aggregating such data exist. One such method is the transformation of the individual nutrient scores (each of which is a ratio) by using its arctan. This approach seems particularly apt, because the tangent is itself a ratio, and its range of values is essentially the same as the potential range of NNR component values. Another approach is to use the geometric mean of the raw ratios. Both stratagems minimize the distortions produced by aberrant values and should be considered in the further development of food scores. (NB: The geometric mean requires the assignment of very small, nonzero values to nutrients not present in a particular food.)
In Table 1, using the nutrient content values in ESHA FOOD PROCESSOR software (version 7.8; ESHA Research, Salem, OR), we present the NNR scores for the 14 nutrients that Drewnowski listed in Table 5 from a representative range of foods. The arithmetic means were obtained by applying equation 4 from the Drewnowski article to the raw ratios, and the arctan and geometric means were obtained by transforming the raw ratios. Neither of those transformed means is unduly perturbed by the high vitamin C and vitamin A scores of such foods as red peppers.
View this table:
TABLE 1. NNR scores: alternative calculations1
We have arranged the foods in the table in descending order of their arctan values. (The geometric mean values have nearly the same order.) Appropriately, milk and eggs rank relatively high. They provide, as is generally recognized, total nutrition for their respective developing organisms. So they ought, perhaps, to rank highest. Red peppers, which outrank them by a wide margin according to the standard NNR, are appropriately downgraded by using either of the 2 transforms. Nevertheless, red peppers, spinach, and similar vegetables (not shown in the table) still have misleadingly high scores. This discordance highlights a problem with any score that combines noncaloric with caloric nutrients. Vitamin A adds no calories, and hence extra vitamin A augments only the numerator of an energy-based ratio; protein, by contrast, contributes to both numerator and denominator. Thus, protein-rich foods can have only modest scores. No mathematical transform, alone, can fully compensate for this feature of caloric nutrients. Nevertheless, as Table 1 shows, the arctan and geometric transforms produce values that accord somewhat better with conventional nutritional wisdom than do the plain NNR values.
It must also be noted that the NNR is inevitably influenced by both the choice of nutrients used to compile a score and the current estimates of the DVs (or whatever reference value one chooses). We find it odd, for example, that niacin is not included in the NNR. Also, today one would want to include the antioxidant capacity of a food (eg, the US Department of Agriculture's oxygen-radical absorbance capacity), which would improve the current, relatively humble ranking of blueberries. And, if the DVs used for vitamin C and folate were as high as respectable segments of the nutrition community have proposed, the scores for several of the low-calorie fruits and vegetables (eg, red pepper and spinach) would drop somewhat in ranking relative to those of milk, eggs, and meat.
But these are fine-tuning issues, not substantive criticisms. We judge that this endeavor is worthy of vigorous pursuit.
ACKNOWLEDGMENTS
Neither author had a personal or financial conflict of interest.
REFERENCES