• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

When is an Individual an Individual Versus a Member of a Group? An Issue in the Application of the Dietary Reference Intakes

Posted on: Thursday, 15 June 2006, 09:00 CDT

By Beaton, George H

The Institute of Medicine (IOM) reports on Dietary Reference Intakes (DRIs) present a new paradigm for the description of nutrient requirements. Tradition and the DRI reports themselves have created an erroneous impression that the Recommended Dietary Allowance (RDA) is intended for use with individuals and the Estimated Average Requirement (EAR) is to be used with populations. Two recent IOM committee reports (one on nutrition labeling and one on Women, Infants and Children [WIC] food packages) challenge that interpretation, revealing the critical issues surrounding individuals versus groups or populations. The present paper examines the issues and concludes that the median requirement and Tolerable Upper Limit (UL) are the critical reference values for both individuals and populations. The RDA is both unneeded and lacking a sound scientific basis.

Key words: daily value, Dietary Reference Intakes (DRI), food guides, nutrition labeling, Recommended Dietary Allowance (RDA), WIC food packages

2006 International Life Sciences Institute

doi: 10.1301/nr.2006.may.211-225

BACKGROUND

Prior to the development of the Dietary Reference Intakes (DRIs) for use in Canada and the United States, numerous concerns had been expressed about the longstanding series of Recommended Dietary Allowances (RDAs) reports. The scientific basis of the RDAs as published in those reports was uncertain (and often not specified). The committees focused upon traditional deficiency diseases, and inadequate attention was directed to possible adverse effects of high nutrient intakes. There was a growing volume of documented criticism about the inappropriate application of the RDA, until recently the only published number. In 1986, the "probability theory" as a basis of interpretation was given formal recognition.1 Total distributions of requirements, and of usual intakes, must be considered in nutrition assessment and planning. In 1994, a Food and Nutrition Board report2 presented a plan for a major review of human nutrient needs and a restructuring of the manner of their description. This led to the initiation of the series of DRI reports on human nutritional needs, the last of which was published in 2005, ending a 10-year process almost 20 years after formal recognition of the need to consider distributions of both requirements and intakes. A new framework for presentation and application of nutrient needs as distributions had been evolving in Food and Agriculture Organization of the United Nations/World Health Organization (FAO/ WHO) reports and most specifically in a UK report on nutrient requirements.3,4 The UK report estimated three points on the requirement distribution: a midpoint and upper and lower limits of the distribution.

The US Institute of Medicine (IOM) adopted the UK model with modification. The DRIs included an estimate of the Estimated Average Requirement (EAR) and of the old RDA; an estimate of a lower limit of the requirement distribution was not presented. When requirement could not be estimated, the DRIs presented the Acceptable Level of Intake (AI).a The DRIs also estimated a Tolerable Upper Limit (UL) for many but not all nutrients. For the macronutrients, the DRIs presented Acceptable Macronutrient Distribution Ranges (AMDRs). This multiplicity of DRIs is very different from the single numbers (RDA or RNI) that had been presented and used over the preceding 60 years in both Canada and the United States. Given only the single number, the RDA was used for almost all purposes. The new DRI paradigm both allowed and demanded a review, conceptual and scientific, and revision of approaches to application of nutrient requirement estimates.

In 2000, the IOM established a committee to consider appropriate uses of the new DRIs. Two reports were published, one dealing with application in assessment of intake5 and the second with application in planning.6 In keeping with a model developed by Beaton,7 each of these reports divided the topic between application to individuals and application to populations or large groups. There was limited discussion of the integration of application and advice, and no specific guidance was offered as to when an individual was to be seen as an isolated individual and when he/she was to be seen as a member of a large group. The Beaton model7 as interpreted and applied was inadequate and served to encourage this very unfortunate separation of discussions and recommendations. The net result is substantial confusion, inconsistency, and controversy in application at the national level, in industry, and in direct counseling. A key issue hinges on when to base applications on the RDA (suggested as applying to individuals) or the EAR (suggested as applying to groups). This dichotomy and apparent declaration of function is perpetuated in the DRI summary tables being published by the IOM (for examples, see the summary tables in the report on electrolytes8 and in the report on macronutrient intakes9). The use of the RDA was intuitively accepted as a continuation of traditional approaches to application (when only an RDA was available). The two reports on general applications of the DRIs5,6 spent almost as much effort in demonstrating the fallacy and outright error of many past applications of the RDA as they expended in demonstrating the validity and necessity of using the new EAR in certain specific applications. Much of the argument was statistical, relating to the issues surrounding distributions of intakes and requirements. This is not the sort of argument most often found in the nutrition literature and was not intuitively accepted. Important conceptual and philosophical arguments concerning application of the DRIs were not addressed in the reports.5,6

Subsequently, a report on the application of the DRIs in nutrition labeling was published.10 Because of voiced criticisms of that report11,12 and an earlier criticism13 of the underlying principles adopted by the IOM committee, the critical conceptual issues relating to application of the DRIs to individuals and groups have surfaced. On the heels of the labeling report, still another IOM committee reporting on reexamination of Women, Infants, and Children (WIC) food packages14 presented a parallel interpretation of the DRIs, which some are questioning. Failure of the IOM committee on nutrition labeling and, by implication, the committee on WIC food packages, to recommend continuation of past practice (base the reference value for labeling on the highest RDA) was criticized as bad science.12 It is now imperative that the underlying arguments, scientific, conceptual, and philosophical, be exposed in the open literature for objective, informed debate and resolution. That is the objective of the present paper. Inescapably, the paper examines the scientific validity and practical utility of the RDA as developed in the DRI reports, as well as probing the conceptual issues concerning individuals and groups.

IS THE RDA NECESSARY? SCIENTIFIC VALIDITY

In the early dietary standards published by the League of Nations, Canada, and the United States in the 1930s and 1940s, the RDA began by adding a "margin of safety," or as some called it, the "margin of ignorance," above the best available estimate of nutrient requirement.15 That was in an era when nutrient requirements were estimated from extremely limited data relating to the prevention or cure of classical deficiency diseases. The margin was a reflection of uncertainty relating to other functions of the nutrients (would higher intakes confer a still-unrecognized benefit?) and uncertainty about reliability of the estimates themselves. The magnitude of the margin differed between nutrients depending on perceived uncertainty. Gradually variation in nutrient requirement among individuals was recognized and became an added reason for setting an RDA higher than the best estimate of actual requirement (the mean or median of multiple requirement estimates). In more recent years, and particularly in the DRIs, panels have included assessment of potential roles in preventing chronic disease through comprehensive reviews of the scientific literature. The reliability of actual estimates of the EAR has been improved. Former major sources of uncertainty about requirements have all but disappeared. In the current DRI reports, estimated or assumed variability of requirement among similar individuals has become the sole argument for setting an RDA above the current best estimate of actual requirement (the EAR). Probability theory and distribution of requirements have become a major theoretical consideration in the DRIs.

In practice, variability of requirement did not receive equivalent scientific consideration. Under assumptions of normality, and most often an assumed CV, the RDA was declared to be two standard deviations (SD) of requirement above the best estimate of median requirement, the EAR, for all nutrients (except iron, for which the distribution among menstruating women is highly skewed). The EAR + 2 standard deviations was stipulated as meeting or exceeding the requirement of almost all individuals (97%-98% of individuals were believed to have true requirements at or below the RDA). It was stated repeatedly that if \usual intake was at the level of the RDA, then there was only a 2% to 3% probability that dietary intake would be below the true requirement.13 The assertion was repeated in each DRI report and was emphasized in sections of the reports addressing application, as well as in the two generic reports on application of the DRIs.5,6

Estimating the actual distribution of requirements is very difficult and requires very large data sets. Indeed, the SD is the most difficult parameter of a distribution to estimate reliably. In the current DRIs, data-based estimates of the SD of requirement were presented for only three nutrients (protein, iron, and vitamin A)16,17 and for energy.9 In the case of protein, the observed variance of requirement in data pooled across multiple studies was reduced by about one-half on the basis that the excluded variance arose from experimental error and inter-laboratory variation rather than true, between-person variation in requirements. Of necessity, important assumptions or judgments were a part of these computations.18 The adopted variance of protein requirement made no specific provision for the effect of variations in protein quality (amino acid composition and digestibility) within the US population. Further, while the primary protein requirement estimate, and its variability, was estimated as grams protein per kilogram body weight per day,9 the IOM summary tables8,9 present the RDA for protein in terms of intake per day for a reference individual (e.g., an adult man weighing 70 kg). That number is not applicable to free-living individuals in the general population, and indeed could be seriously misleading. If describing an RDA for men in the United States as grams per day, one would also need to take into account the variance in body weights. Comparative RDA estimates for adult men are: DRI summary table RDA = 56 g/d8,9; the RDA taking account of measured weights of 6630 men age 20 to 70 in NHANES III = 83 g/d (Beaton GH, unpublished calculations).c

In the case of vitamin A, the observed coefficient of variation (CV) of requirement was reduced from 21% to 20% on the argument that observed variation incorporated experimental error.16 The precise basis of this reduction is not described in the report.

In the case of iron, variation of requirements was based on a modeled variation of factorial components of requirement for absorbed iron.16 There were no actual data on observed variability of total dietary requirement. No specific provision was made for the added variance arising from variation in bioavailability of iron among individuals at intakes just meeting the requirement for absorbed iron.

Further, evidence was presented that the requirement distributions for protein, and more particularly for iron in menstruating women, departed from normality. The departure for protein was not great, but should remind all that the characteristics of requirement distributions remain unknown in almost all cases. Departure of the requirement distributions from normality poses a serious challenge to the central probability assumptions made in deriving the RDAs.

For all of the other nutrients for which an EAR could be estimated, but variance estimates were not available, the variance of requirements was assumed (normal with a CV of 10%, except for niacin, where the CV was assumed to be 15%d) and the RDA was set as the EAR +20% or +30%.16,17 The assumption of CV = 10% was said to have been based on the variation in basal metabolic rate (BMR), an approach that can be traced back as far as a 1965 FAOAVHO report on protein requirements.19 In the examination of energy expenditure, the IOM macronutrient panel collected research data from selected studies using a doubly labeled water estimation of total energy expenditure. That data set also included measured BMR, a variable used by the IOM panel to partition observed energy expenditure between maintenance and activity. Noting that these data had been vetted by a panel of energy experts, it is relevant to ask about the observed CVs of measured BMR and measured total energy expenditure in this reference data set. Among adults 19 through 50 years of age, the CV of BMR, expressed as kcal/d, was 14.0% for men and 15.0% for women; the CV of total energy expenditure was 16.3% for men and 16.4% for women (Beaton GH, unpublished analyses). These estimates are substantially higher than the adopted 10%. Black and Cole20 attempted to differentiate within-person and between-person variance of total energy expenditure in a different set of doubly labeled water studies. Their work suggests that there is a substantial random error component (within-subject variation) in observed variation in total energy expenditure. We do not know the true variability of BMR or of total energy expenditure, and the DRI series of reports suggest a major inconsistency between the estimated variability of energy expenditure and the assumed variability of nutrient requirement distributions. This leaves us with two critical questions.

First, is it reasonable to assume that the CV of BMR or total energy expenditure is a valid proxy for the unknown variability of nutrient requirement? If that is scientifically reasonable, then we must ask whether the old estimate of the variability of BMR is a valid estimate of the actual variability of BMR. Without knowledge of the true variability of a nutrient requirement, one cannot offer hard scientific evidence to answer the first question. It is noted that for the three nutrients with actual databased estimates, the assumed CV of 10% with normal distribution would be a serious misfit. The second question, the reliability of the CV of BMR, has been answered in the negative. The "scientific basis" and hence validity of RDAs calculated with uncertain or assumed variation and unknown distribution characteristics must be questioned.

The message seems clear. The probability theory that underlies the RDA construct remains valid. However, at present, we do not have data to support the application of this theory with the assumed precision that is now common. Indeed, we have no justification for continued presentation of the RDA as a science-based estimate of the level of intake conferring a specified low probability (0.025) of intake below actual requirement. That simply is not true.

It is important to recognize that these criticisms of the scientific basis of the RDA do not impact on the EAR or its utility in application. It has been demonstrated1'5 in theory and in empirical examples that for assessment the EAR cutoff approach is essentially independent of the variance of requirement as long as the requirement distribution approximates symmetry (normality not required). The same independence would seem to hold for the use of the EAR in the applications discussed in the present paper.

IS THE RDA NECESSARY? CONCEPTUAL UNDERSTANDINGS

Is it necessary or advisable to urge people to try to achieve the RDA for nutrients through diet and/or supplements if the goal is a low likelihood of inadequate intake? The immediate answer of most in the nutrition field would be "yes, of course it is necessary." That answer would be wrong. Therein lies the root issue.

Figure 1 presents a portrayal of the usual protein intakes of adult men examined in the NHANES III survey, together with the estimated requirement distribution presented as the probability of inadequacy (dotted line). Finally, the figure portrays in black the expected distribution of men with usual intakes below their true requirement. The analyses and plotting are done by 0.1 g/kg/d intervals using midpoints. While facilitating display, the approach overestimates the prevalence of inadequate intakes. The estimated prevalence of apparently inadequate intakes, estimated by IOM- recommended methods5 applied to each observation, was 2.36%. For an individual selected at random from that adult male population of 19- to 50-year-olds, the likelihood that his usual intake was below his true, but unknown, requirement was 0.024. The probability of inadequacy assumed to be associated with published RDAs is 0.023. That is, the probabilities are identical. However, in the NHANES data set, 5% of men had usual intakes below the RDA of 0.8 g/kg/d. It was not necessary to achieve the RDA to achieve a low likelihood of inadequacy. Interestingly, in that same population, it is predictable that 1.4% of those with intakes at or above the RDA have intakes below their own requirement. Among those with expected inadequacy, 37% would have had intakes above the RDA. The apparent anomalies arise from the fact that prevalence of inadequacy, and hence the likelihood of inadequate intake for the randomly selected individual, is a function of both the distribution of requirements and distribution of usual intakes. We may "eat as individuals" but we exist as members of large groups.

We are faced with a major conceptual and philosophical issue. The only thing that science can say with certainty is that to achieve a low likelihood of inadequate intake, it is not necessary to have a usual intake equal to or above the RDA. The nutrition community (scientists and practitioners) has long had a philosophy that "if a little is good, more must be better" or "if a low level is adequate a higher level must be better." The DRI reports repeatedly warn against such a philosophy. Even though incompletely understood, it is absolutely clear that high intakes of nutrients can be associated with detrimental effects. It should always have been recognized that ingesting higher levels of nutrients than actually needed can be economically wasteful. It is certainly time for consideration of benefits and costs (including potential harm) when recommending intakes of nutrients for North Americans.

Finally, Figure 1 serves to illustrate a major criticism of past usage of the RDA. Historically, and even currently, some have seen the RDA as \a target for group mean intake on the assumption that if the group mean intake approximated the RDA, only 2% to 3% of the group members would have inadequate intakes. This assumption was refuted in the DRI reports on application.5,6 Figure 1 portrays a distribution of usual intakes in which the expected prevalence of inadequacy (intake below true requirement) is 2% to 3%. The group mean intake was 1.34 g/kg/d, which is 67% higher than the RDA of 0.8 g/kg/d. Clearly, use of the RDA as a target for the group mean intake is scientifically wrong.6

Figure 1. Protein intakes of US adult men, 2001 to 2002: mean intake = 1.34 g/kg/d; median intake = 1.29 g/kg/d; Estimated Average Requirement (EAR) = 0.66 g/kg/d; Recommended Dietary Allowance (RDA) = 0.8 g/kg/d. The prevalence of inadequate intakes estimated as the sum of affected individuals within each 0.1 g/kg/d interval of intake was 3.5% (area in black); the estimate by the EAR cutoff approach was 2.4%; and by the full probability approach using log- normal requirement distribution was 2.7%. The likelihood that a randomly selected individual from this population has an inadequate intake is 2.4 to 3.5 per 100, or 0.024 to 0.035, with the best estimate being about 0.025, which is the level of "risk" that is claimed for achievement of the RDA.

Given the foregoing considerations, it is reasonable to conclude that the scientific basis of the RDA is absent or seriously wanting for most nutrients and that the utility of the RDA, even if it could be estimated reliably, is questionable.

VALIDITY OF SOME RECENT APPLICATIONS OF THE DRIs IN THE UNITED STATES AND CANADA

Throughout any discussion of past and proposed practice, it must be recognized that before the publication of the DRIs starting in 1997, only one numeric estimate of the distribution of nutrient requirements was published: the RDA in the United States and the Recommended Nutrient Intake (RNI) in Canada. It follows that past use of the RDA cannot in itself be taken as justification for its use in the future. The DRIs have opened a new window and users must be prepared to change past thinking. Four recent national applications of the DRIs in food planning in the United States and Canada give the opportunity to critique current concepts and approaches.

Two of these examples use the EAR as the basis of developing dietary guidance or actual food packages for use by individuals. Two use the EAR for part of the process, and then appear to switch to the RDA. The first two applications represent works in progress; the other two have been implemented. The IOM report on labeling is excluded from this discussion since it represents a conceptually different type of application.

Example 1 is drawn from the report of an IOM committee on revision of WIC food packages.14 That report made the clear assertion that the first nutritional goal for WIC food packages was that "the package would reduce the prevalence of inadequate and excessive nutrient intakes of participants." Nutrient adequacy was defined in relation to the EAR. That is, the first nutrient criterion of appropriate WIC food packages would be that it helped position usual intakes of as many individuals as possible between the best estimate of their requirement (the EAR for their age-sex stratum) and the estimated UL for that stratum. The RDA was not used in the identification of problem nutrients or in the committee's recommendations for the design of new food packages. The approach taken was consistent with the new DRIs and their meanings and with approaches advocated in the two IOM reports on application,5,6 It was argued6 that planning intakes such that an acceptably small proportion of individuals have usual intakes below the EAR for their age-sex group was the appropriate approach.

The second example of use of the EAR is even clearer. In the development of a food guide for application in Canada,21 the explicit nutrient intake goal was that the guide should meet the nutritional needs of almost all Canadians. In development and testing, this was interpreted as achieving a predicted low prevalence of intakes below the EAR for each target age-sex group when requirement had been estimated, or group mean intakes approximating the AIs when no estimate of the distribution of requirements was available. The guide should also minimize the proportion of expected usual intakes above the UL. To implement these goals, large series of diets complying with the test food pattern were generated and nutrient contents computed. This allowed examination of expected distributions of intakes among hypothetical individuals following that model of the food guide. The proportion of intakes below the EAR and the proportion above the UL were then estimated. For macronutrients, the derived distributions of intakes were tested against the AMDR. When neither a requirement nor an AMDR was available, group mean intakes were compared with the AI.5 The diets generated comprised random selections of individual foods within each food group category, incorporating a weighting system based on national food purchase data for items falling within the definition of the food group. Through an iterative process involving revision of food group specifications and quantities, a composite food guide that met the originally specified nutrient intake goal was developed. This is an appropriate interpretation and application of the DRIs and is consistent with the principles underlying the two DRI application reports.5,6

The RDA was not applied in either of these recent examples even though the target was consumption of nutrients by individuals. In each case the goal was to help achieve low prevalence of both inadequate and excessive intakes among individuals in the population. The examples are conceptually equivalent and mark the opening of the new path of interpretation and application of the DRIs.

In the discussion of food guides in chapter 2 of the IOM planning report,6 there is a strong promotion of the concept that compliance with the guide should ensure a low probability of inadequacy for each individual. That appears to have been interpreted as achievement of the RDA by each individual. Such an interpretation is inconsistent with the later discussion in the same report6 concerning how planners should approach groups. Here, achieving a low likelihood of inadequacy for the individual equated with promoting a distribution of intakes such that the prevalence of inadequate intakes is low (low proportion below the EAR); the RDA does not come into play. The authors of the planning report6 failed to recognize or to address this major conceptual inconsistency, perhaps because of the sharp separation of individuals and groups in both the assessment5 and planning6 reports. Perhaps, too, the omitted discussion and debate in the planning report would have reflected philosophy and conceptualization more than "simple" science. The Beaton model7 as used in the DRI reports erred by not going far enough. It failed to formally recognize that the likelihood of inadequacy of intake for a randomly selected individual reflects both the distribution of requirements and the distribution of usual intakes-i.e. the likelihood of inadequacy for a member of the population is equivalent to the population prevalence of apparently inadequate intakes.

In developing dietary advice, the emphasis must be on foods and patterns of food selection associated with long-term health. Nutrient contribution is a secondary consideration, albeit important. It seems clear that as much, or more, attention should be directed to strong epidemiologic studies addressing the links between patterns of food use and health as to laboratory and clinical studies designed to elaborate the functions of specifie nutrients or testing the associations between specific nutrients and health. Basing recommendations on foods rather than on nutrients is not an abandonment of science. That does not argue against testing and adjusting food groupings to ensure that nutrient intakes are likely adequate to meet the needs of as many people as possible as long as this is seen as the secondary goal.

There are two other examples of recent applications of the DRIs at the national level, both in the United States. Unfortunately, while both correctly incorporate part of the construct of the DRIs, both have serious conceptual flaws in how they interpret and apply the RDA.

In the development of the Dietary Guidelines for Americans 2005 by the Department of Health and Human Services and Department of Agriculture,22,23 problem nutrients requiring special attention were defined on the basis of the prevalence of apparently inadequate intakes estimated as the proportion of individuals with usual intakes falling below the EAR.5 Other factors, such as the prevalence of biochemical or hematological markers suggestive of inadequacy of intake, could influence designation of problem nutrients. There was no reference to intakes below the RDA as being problematic.22 This step was consistent with the recommended application of the DRIs.5 The second group, who actually developed the final report,23 presented two examples to demonstrate that diets complying with the guidelines would meet "recommended intakes." One example was a Dietary Approaches to Stop Hypertension (DASH) meal pattern, represented by a single hypothetical 1-week dietary record based on "selected foods." The second example was based on the then- current USDA Food Guide. In this second model, the authors attempted to provide representation of expected food choices in the United States (missing from the DASH example). For each food group in the pattern, nutrient intake was computed as the composition of a weighted mixture of foods of that type based on estimated food consumed in the United States, multiplied by the recommended number of servings of that food group. In the DASH example, one is lookin\g at the possible usual intake of a single individual, and other individuals making different food selections might have higher or lower nutrient intakes. In the food guide-based example, at best the computed intake estimate would represent the expected group mean intake of a population of individuals complying with the Food Guide. The computed values for total dietary intake (both models) were compared with the RDA or AI for adult women. The comparison of a group mean intake with the RDA was a traditional but erroneous application of the RDA. This application was explicitly rejected as invalid in the first IOM report on application of the DRIs.5 A specific example of the potential nutritional shortfall was presented in the IOM report on application in planning.6 The examination of protein intakes in the United States (Figure 1) makes it clear that the group mean intake consistent with assurance of adequacy of intake for nearly all individuals was considerably higher than the RDA. Without estimation of distributions of usual intakes, comparisons of modeled intakes with any of the published DRIs except the AI (intended to represent a group mean intake of a healthy population) is without valid interpretation.5,6

The second flawed example comes from the development of the new US "My Pyramid" food guide.24,25 The approach apparently adopted for developing this guide and testing expected nutrient contribution and the inherent flaw in the approach appear to parallel those in the development of the Dietary Guidelines for Americans as described above. When establishing the numbers of food group servings to specify, the approach effectively targeted a group mean intake to meet or exceed the RDA or AI. It appears24,25 that the authors failed to examine expected distributions of intakes. That makes it impossible to estimate the proportions of people who would be expected to meet their requirement (proportion above the EAR) and avoid excessive intakes (proportion below the UL) or fall within the AMDR for macronutrients. It cannot be inferred that individuals following the guide would necessarily achieve the RDA or even have a very low likelihood of inadequacy, as many in the nutrition community have assumed. Two important messages arise from the review of these examples:

1. The RDA is not needed in the development of food guides or general dietary advice to be used by individuals.

2. It appears that, in both the United States and Canada, national bodies are beginning to interpret and implement the conceptual framework of the new DRIs. At present, errors and confusion continue, but there is definite progress.

Recent committees offering national-level recommendations are promoting intakes of individuals that would be greater than the EAR without comparing individual intakes to the RDA. Since many nutrition policies and programs then build on the guidelines being revised, it can be predicted that an EAR-based approach such as adopted in the IOM report on WIC food packages14 and in the development of a food guide for Canada21 will spread in both countries. One implication of this evolution is that the RDA really has very limited, if any, purpose in applied nutrition.

THE SPECIFIC CASE OF NUTRITION LABELING

Nutrition labeling is a tool for the description of foods such that the consumer can make more informed choices among foods, it is not a description of desirable food choices or meal patterns. In that, food labeling, and, more specifically, the choice of reference values for the nutrition facts table on food labels, is distinctly different from the work described above. It follows that the issues here are different from above, although they are conceptually related.

The US Food and Drug Administration (FDA) has stated that expressing nutrient content of foods as a percent reference dietary value (%DV) per serving "is intended to help consumers understand the role of individual foods in the context of the total daily diet."26 A Health Canada publication27 suggests the following potential uses of the nutrition facts table: to evaluate a food's nutritional value, to compare nutritional quality of foods, and to control intakes of specific nutrients, such as fat or sodium in special diets. The clearest statement of purpose is found in the charge to the IOM committee on use of the DRIs in labeling. The committee was to recognize and take into account "the purpose of reference values on food labels, specifically that consumers are expected to use the reference values to compare different food products and to determine the relative contributions of a food product to an overall healthpromoting diet."10 Since the charge to the committee was developed in consultation with the sponsors, including the FDA and Health Canada, and was explained to the committee in its opening meetings with sponsors and at workshops, it is reasonable to accept that this statement reflects the current intended purpose of the nutrition facts table in both countries.

The nutrition facts table is intended to apply to the total population 4 years of age and older. This necessitates some form of blending of different age-sex specific reference intakes. At present, both countries establish reference %DVs for selected lists of nutrients considered to be of current importance in public health, and express the nutrient content of a serving of the food as a %DV, with or without absolute contents also stated. A number of studies in both the United States and Canada suggest that nutrition label information is being used by many as an aid to purchase decisions (relative nutritional values of alternative food choices), but is not being used to plan meals or structure diets. Use of the %DV for meal planning was notably absent from stated US and Canadian goals and from the mandate of the IOM committee.10 seemingly, that is not even possible unless all foods, including restaurant and fast- food meals and supplements are labeled with the same reference %DV.

A very important trait of an appropriate %DV, however developed, is that it should yield unbiased descriptors of the nutritional quality of foods. A key principle to be borne in mind throughout the discussion is that the best available estimate of an individual's true, but unknown, requirement for a nutrient is the EAR for his/ her age-sex group. It follows that the RDA (set well above the EAR) must be seen as a biased representation (overestimation) of likely nutrient need. This principle was central to the recommendation of the IOM report10 and to Tarasuk's28 defense of the use of a derivative of the EAR.

Given a desire by the FDA and Health Canada to retain the %DV approach in labeling, the critical question becomes how should the %DV be derived to best satisfy the stated goals of nutrition labeling? A series of seven possible approaches are identified for comparison with this question in mind.

1. The median requirement of mixed populations (IOM report10) or an estimated median requirement (committee that developed dietary reference values for the United Kingdom3);

2. A population-weighted mean of age-sex group EARs;

3. A population-weighted mean of age-sex group RDAs (Murphy and Barr11);

4. The highest EAR;

5. The highest RDA (existing approach, continuation advocated by Hathcock12);

6. The median of a simulated population distribution (an alternative approach to estimating the median requirement of the total mixed population/;

7. The 97.5th centile of a simulated population distribution.

To allow comparison of the suggested derivations, specific examples of applications were examined. Alternative %DV estimates were derived for three nutrients chosen to represent specific features. Iron was included because of the heavy skew in the distribution of its requirement for menstruating women. Vitamin A was selected because of the very large estimate of the CV of requirement (20%). Folate was included to illustrate the more typical assumption of a CV of 10%. The derived %DV estimates are shown in Table 1.

Estimated nutrient needs of pregnant and lactating women were omitted from all of these suggested approaches (the population estimates include pregnant and lactating women, but they were assigned the requirement estimates of comparable non-pregnant, non- lactating women). Estimates of the EAR and RDA for IOM age-sex groupings were taken from the relevant IOM reports.16,17 Column 1 shows the %DVs presented in the IOM report.10 For approaches 2 and 3, US population projections for the year 2006(29) by IOM age-sex groups were used as weighting factors applied to the EAR or RDA for each DRI age sex grouping 4 years and older. In approaches 4 and 5, the highest specified EAR or RDA among those applying to the DRI age- sex grouping was selected (no weighting) For approaches 6 and 7, a simulated 1 in 1000 sampling of the expected 2006 US population above age 4 was generated (a total of 281,515 mock individuals). To each individual a requirement was assigned as a random member of the DRI-stipulated requirement distribution for his or her age-sex stratum.8 This simulation of the requirements of individuals in the United States (pooled simulations for age-sex strata) was then subjected to simple univariate distribution analysis, and the 50th and 97.5th centiles, conceptually equivalent to the EAR and RDA as set in the IOM reports, were estimated. Note that the estimates presented in the IOM report and the estimates of the 50th centile of the simulated population differ slightly, in part because the population estimates were for different years and drew on structurally different population projections.

Table 1. Comparison of Suggested Reference Dietary Values, Age 4 Years and Above (and Ratio to the Population-Weighted Estimated Average Requirement, or EAR)

Also shown in Table 1 are the ratios of each estimate to the baseline population EAR of the IOM committee. If we consider a single age-sex grou\p, and the CV of requirement was assumed to be 10%, the ratio of the RDA to the EAR would be 1.2 (RDA set 20% higher than the best estimate of actual requirement (the EAR). The ratios in Table 2 reflect a similar concept but across the pool of 13 age-sex groups. The ratios indicate the bias or "inflation factor" of the proposed DV compared with the best estimate of the true median requirement.

To examine the potential impact of choice of approach, and therefore derived %DVs, a series of foods was selected. The foods and their content of the target nutrients are shown in Table 2. Composition data are from a 2005 USDA compilation of food composition estimates.30 Table 3 shows the %DV values developed by applying the proposed DVs (Table 1) to current composition estimates (Table 2).

The data in Table 3 suggest that the first role of the %DV, to facilitate comparison of the same nutrient between foods, can be satisfied by any reference %DV (indeed, the absolute levels without a %DV would suffice for this purpose); this was also noted by Hathcock.12 However, if comparisons are to be made across nutrients in the same or different foods, it becomes important that the reference value should have equivalence of meaning across nutrients. From examination of the ratios shown in Table 1, it is clear that while %DVs based on weighted EARs or population-derived median requirement have equivalent meaning across nutrients (ratios the same for all three nutrients), this is not true for RDA-based %DVs, the 97.5th centile of the population distribution, the highest EAR, or the highest RDA.

One consequence of this bias of overestimation of need in the reference %DV is a bias of underestimation of nutritional value of foods. The most objective and least-biased %DV for labeling would be based on the best available estimate of the nutrient requirement of a randomly selected member of the target class, i.e. the median requirement in the target population.10 The pooling of age-specific information represents a compromise from the scientifically preferred use of age- and sexspecific EARs. The compromise is forced by space limitations in nutrition facts tables. Provision of that much detail (reference values for each age-sex group) would also impose consumer overload and might well defeat the prime purposes of nutrition labeling. The space limitation also precludes the rational proposal that %DVs should be developed for two age ranges: 4 through 12 years and 13 years and older.13 Such a step would help to avoid the substantial bias of underestimation of nutritional value of foods for children in the under-13 age group associated with any of the suggested %DV approaches.

Unanswered is the argument put forth by Murphy and Barr11 and by Hathcock12 to the effect that failure to base the %DV on the RDA would leave the most susceptible individuals unprotected. That argument seems to assume a use of the nutrition facts table as an aid to meal planning and hence a direct connection between summed %DVs and adequacy of intake of individuals in the United States or Canada. Four points can be made:

Table 2. Nutrient Content per Serving of Some Common Food Items*

* As previously documented, use of the nutrition facts table in meal planning (summing %DVs to 100%) is not an intended purpose,10 even though such a usage was suggested in the IOM report on planning.6 The argument11,12 seems to be irrelevant.

* The objective of national nutrition programs is, and should be, to safeguard the health and well-being of the total target population of individuals to as great a degree as possible. In practice, this means trying to ensure that the prevalence of apparently inadequate intakes is as low as possible, while also maintaining the prevalence of potentially detrimental excessive intakes as low as possible. Given the DRI paradigm, the nutritional goal becomes to attempt to position the distribution of usual intakes of individuals between the EAR and the UL or within the AMDR. The RDA is not needed in this. The argument11,12 is unfounded.

* Both studies11,12 use the iron needs of women as a specific example of why the weighted RDA11 or highest RDA12 should be used as a %DV. Use of an RDA-based approach and particularly the highest RDA would underestimate the nutritional value of foods for about 81 % of the population to hypothetically protect menstruating women with iron needs above the population EAR (about 19% of the target population). The impact of underestimation of nutritional value would be particularly notable in the 7% of the target population 4 to 8 years of age. The argument11,12 implies an unjustifiably biased perspective.

* The final point, already documented, is that the RDAs, as presented in the DRI reports, lack a sound scientific basis. In contrast to the assertion of Hathcock,12 it is use of the RDA, not use of the EAR, that is no longer in step with science.

* The key point is, if a low prevalence of inadequacy is achieved in the target population, it follows that the likelihood of inadequate intake by a randomly selected individual in this population would be low. Even if the RDA could be reliably estimated, the rationale used to defend the RDA as the intake needed to achieve low likelihood of inadequacy for the individual does not hold.

* Table 4 offers some summary comparisons of potential impact and import of implementation of three of these approaches to derivation of the DV. The recently published evaluation of data from NHANES 2001-2002 (What We Eat in America31) adds important perspective. Those dietary analyses do not include nutrient intakes from supplements or other non-food sources. Among the 16 nutrients for which requirements had actually been estimated and an EAR published, prevalence of apparently inadequate intakes of 5% or less was seen for nine. If the DRI estimates of requirements are correct, concerns might hold for seven of the nutrients, but even here the public health concern might well depend on the expected health impact of these "inadequate" intakes.h Notable among the possible problem nutrients, with prevalence of apparently inadequate intakes exceeding 50%, were vitamin E and magnesium. In general, the intakes of all 16 nutrients from food alone appeared adequate for children (0-8 years of age).

* Among the seven nutrients for which a requirement could not be defined by the DRI panels and no AMDR was developed, only an AI was available. Here, the only valid assessment is a comparison of group mean intake with the AI.5 While this is recognized in the text of the report,31 the summary tables present the percentage of individuals with intakes above the AI31 and may be misleading. Group median intakes are presented and can be compared with AIs shown in the same tables. Making this comparison, it appears that vitamin K, potassium, calcium (for some age-sex groups), and dietary fiber exhibit intake levels that might be associated with some occurrence of inadequacy, but one cannot offer any assessment of the proportion of individuals, if any, with inadequate intakes. The other three nutrients exhibit group mean intakes above the AI, so existing intakes might be judged likely to be adequate for almost all individuals.5 Again, children appeared to have adequate intakes of all of these nutrients.

Table 3. Possible Nutrition Label Descriptions of Nutrient Content of Foods

* For 11 nutrients, ULs were established in the DRI reports. Intakes exceeding the UL by more than 5% were seen for vitamin A (retinol), folic acid, zinc, copper, selenium, and sodium. Excessive intakes of sodium were ubiquitous across all age-sex groups, with an overall estimate that 86% of the population had excessive intakes. For vitamin A, folate, and zinc, intakes above the UL were seen only in children less than 8 years of age. For copper and selenium, excess intakes were reported only for infants and children less than 4 years of age.

* The anomaly of these dietary findings and the linkage to labeling issues lies in the fact that it is among children that potential nutrient inadequacies are least likely and excessive intakes are most likely. It is for this same group that the bias of underevaluation of nutrient contribution of foods to the general diet is most severe.

* The dietary analyses also show that for the 23 nutrients examined, the US diet appears to be nutritionally quite good. That is not the impression that consumers might gain if the %DV is set higher than as a function of the best estimate of actual requirement, the EAR. (The "nutritional" problem of greatest concern, obesity, is not referenced by the dietary data.31)

* The nutrient intakes described above do not include nutrients ingested from supplements. If these were also considered, there might be amelioration of some of the apparent inadequacies but exacerbation of some of the possible excesses.

Table 4. Comparison of Impacts of Application of Reference Values (DV) Using Different Derivation Approaches*

* Given this picture of the current US dietary situation, there is added importance to using the most appropriate %DV construct. What is needed is a value that would neither lead to unwarranted addition of nutrients to foods nor encourage wasteful purchase and consumption of supplements, but would offer the consumer a realistic picture of the nutritional value of foods purchased and assist informed choices and offer objective guidance to the food and pharmaceutical industries with reference to possible merit of fortification or supplements. The recommendations of the labeling committee10 pertaining to application of the DRIs are logical (except the suggestion that in the absence of an EAR, the AI as published should be used), valid, and important. It is to be hoped that they will be implemented, perhaps with some modification, in both Canada and the United States.

CONCLUSIONS AND RECOMMENDATIONS

When it was decided to develop DRIs, the nu\trition community was accustomed to using the RDA (previously the only descriptor of requirements) for all purposes, and had paid little more than lip service to the recognition that the RDA at best was only a single point in a distribution of requirements among similar individuals. There is how a major inertia resisting changes from traditional approaches, many of which are now embedded in regulations. The first two IOM reports on application of the DRIs went far along the path of documenting erroneous interpretations and applications of the past, and developed well-documented descriptions of how the DRIs could and should be applied in specific areas of future assessment and planning. However, the reports did not go far enough. Those reports divided their focus into individuals and groups/ populations, but failed to address the critical issue of "when is an individual an individual versus a member of a group?" As a result, there are major internal contradictions and major confusion among users. The IOM report on application of the DRIs in nutrition labeling and fortification10 did address this issue in its recommendations, albeit without specific identification of the issue itself. The subsequent IOM report on WIC food packages14 also addressed the issue without declaration. Both reports presented scientifically logical interpretations of the application of the DRIs to individuals and populations. Both reports severely challenge traditional thinking.

The ultimate fallacy that some in the nutrition field and USDA25 have voiced is, "since we eat as individuals, advice directed to individuals should be based on the RDAs." It seems clear that, for essentially all nutrients, there is no scientific justification for advocating nutrient intake equal to or greater than the specific level identified as the RDA. There are also very serious questions about the validity and relevance of the assertion that RDAs are necessary to ensure a low risk of inadequacy and maintain a healthy population. We can offer assurance that if the prevalence of apparently inadequate intakes in a population or large group is held at very low levels, perhaps 2% to 3%, the probability of inadequate intake in a randomly selected individual in that population is also 2% to 3%, the espoused goal of the RDA. One must question whether an RDA, even if based on sound data, has any real purpose. We are currently in a state of confusion and contradiction, with a major discontinuity between science and application.

Further consideration should be directed to nutrients for which neither requirements (EAR) nor an AMDR was estimated. Any approach to application should recognize that the AI was intended in design to refer to a group mean intake, not individual intakes. It was an error of the DRI process to suggest that it could be used as the equivalent of an RDA, adding still more confusion.

At a future time, there would be considerable merit if the IOM or another body acted to fill in an important gap in requirement estimation and guidance. Serious consideration should be given to the development of sets of clinical reference values that would find application in the unique situation of planning long-term feeding by the intravenous or oral routes, usually with fixed composition formulations. In this situation, it might be argued that the level chosen as a suitable reference intake should be close to the highest requirement among such patients unless there is contraindication to high intakes. Such reference values must take into account the route of feeding and the impact this can have on both absorption (and hence requirement) and potential for detrimental effects (e.g., iron). Clinical reference values require the input of clinical experience and judgment. By definition the DRIs were not designed for application to people who are clinically ill. Even if the scientific limitations of the existing RDAs were resolved, they would still be the wrong estimates for this special situation.

Finally, there should be substantial effort directed at enhancing our knowledge of the detrimental effects of high intakes and estimation of the UL. At present, the UL incorporates a substantial margin of safety based on perceptions of the magnitude of ignorance about both the best markers of detrimental effects and about the distributions of susceptibilities. In that regard, there is a conceptual parallelism between the current UL32 and the earliest RDA.15

The DRIs have presented a totally new (to North American users) nutritional paradigm. That is truly an important step forward in the science of nutrition and of its application. However, all users must now recognize, understand, and internalize this new paradigm and its implications and applications.

The most important conclusions of the present paper are as follows:

* The RDA lacks an acceptable basis in science. It continues to be seriously overinterpreted. This must change if the science of application is to continue to grow.

* Dietary advice to the general public should attempt to promote distributions of intakes by individuals in which as many individuals as possible have usual intakes above the EAR and below the UL, within the AMDR, or for nutrients for which neither a requirement nor a desirable range of intake has been established, a group mean intake approximating the AI.

* As to the question "when is an individual an individual versus member of a group?," the answer seems clear: "It doesn't matter, the advice that should be given is the same - usual intakes of individuals should be above the EAR and below the UL."

* The principles developed by the IOM labeling committee,10 including a %DV based on a population EAR, should be adopted in both Canada and the United States, perhaps with some refinements.

* It would seem both logical and wise that the term "Recommended Dietary Allowance" and the acronym "RDA" be removed from the vocabulary of nutrition as rapidly as possible. It designates an estimate that has limited scientific validity and has virtually no utility once the DRI concepts are understood and applied. It should not be promoted as the "recommended" intake for individuals.

ACKNOWLEDGMENT

While accepting full responsibility for the final paper, the author wishes to acknowledge Mary Bush, Margaret Cheney, George McCabe, Linda Meyers, and Valerie Tarasuk, who critiqued successive versions of this paper and/or made other conceptual contributions. He is also indebted to Suzanne Murphy and Valerie Tarasuk, who made available prepublication copies of their papers.

a In this situation, the DRI panels were to attempt to identify the mean intake of healthy population groups just as very early investigators had done. The assumption was that such a group mean intake identified a distribution of intakes adequate to maintain health in all or almost all individuals in the group. Some DRI panels instead assembled fragmentary information and presented a number that was also called an Al. The relevance of these latter estimates to the EAR, to the RDA, to the planned Al, or to the true distribution of requirements is unclear. This severely complicates attempts to develop consistent interpretations and applications of the published AIs.

b Mean 2 SD was a statistical convention that was widely used to define "normal" in clinical chemistry. It was conveniently transferred to nutrition when variability of requirement was initially recognized.

c It is likely that the best predictor of protein need is lean body mass, but that was not the basis on which requirements were estimated. It follows that neither of these estimated RDAs is scientifically correct.

d The CV for niacin requirement was set at 15% because there were some data that suggested a CV of 10% would be insufficient to cover observed variance.17

e Under the assumptions of normality and independence of intake and requirement distributions, the Required Group Mean Intake = EAR + 2 SD (intake) and RDA = EAR + 2 SD (requirement).6 Given that the variance of intake is almost always greater than the estimated or assumed variance of requirement, it follows that the Required Group Mean Intake is greater than RDA, sometimes much greater.6

f The IOM report10 presented a statistically precise estimation. The simulated population is an empirical approach that may be easier to understand. When large numbers are used in the empirical model, the estimates should be almost identical with the statistical model. The simple demographically weighted DRIs do not give full recognition to the skewing of the requirement distribution as do both the IOM estimate and the simulated population approach now presented.

g The SAS models of requirement distributions used to estimate EARs and RDAs for the DRI report on iron requirements16 (models written by GHB) were used to generate these population distributions. For other nutrients normality was assumed and mock in- . dividuals were assigned random values from the model distribution: EAR + 2 SD^sub requirement^ as specified in the DRI reports (CV = 10% for folate; 20% for vitamin A).

h "Inadequacy" of intake must not be assumed to imply a functional impairment affecting health; probability of inadequacy and risk of ill health are very different. Nutrient intake is only one of many factors that may affect health, particularly for chronic diseases.

REFERENCES

1. National Research Council. Nutrient Adequacy: Assessment Using Food Consumption Surveys. Washington, DC: National Academies Press; 1986. Available at: http://www.nap.edu/books/0309036348/html/ index.html. Accessed April 4, 2006.

2. Food and Nutrition Board, Institute of Medicine. How Should the Recommended Dietary Allowances Be Revised? Washington, DC: National Academies Press; 1994. Available at: http://darwin.nap.edu/ books/NX004037/html/index.html. Accessed April 4, 2006.

3. Department of Health. Dietary Reference Values: A Guide. Her Majesty's Statio\nery Office; London; 1991.

4. Committee on Medical Aspects of Food Policy. Dietary Reference Values for Food Energy and Nutrients for the United Kingdom. Department of Health. Report on Health and Social Subjects No. 41. Her Majesty's Stationery Office; London; 1991.

5. Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes: Applications in Dietary Planning. Washington, DC: National Academies Press; 2003. Available at: http://www.nap.edu/ books/0309088534/html/. Accessed April 5, 2006.

6. Beaton GH. Criteria of an adequate diet. In: Shils ME, Olson JA, Shike M, eds. Modem Nutrition in Health and Disease. 8th ed. Malvern, PA: Lea and Febiger; 1994; 1491-1505.

8. Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes for Water, Potassium, Sodium, Chloride, and Sulfate. Washington, DC: National Academies Press; 2004. Available at: http://www.nap.edu/books/0309091691 /html. Accessed April 5, 2006.

9. Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes for Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol, Protein, and Amino Acids (Macronutrients). Washington, DC: National Academies Press; 2005. Available at: http:// darwin.nap.edu/books/0309085373/html/. Accessed April 5, 2006.

10. Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes: Guiding Principles for Nutrition Labeling and Fortification. Washington, DC: National Academies Press; 2003. Available at: http://www.nap.edU/catalog/10872.html. Accessed April 5, 2006.

11. Murphy S, Barr Sl. Recommended Dietary Allowances should be used to set Daily Values for nutrition labeling. Am J CNn Nutr. 2006;83:1215S.

12. Hathcock J. Guiding Principles for Nutrition Labeling: New IOM Report Misses the Mark. Council for Responsible Nutrition Backgrounder. December 2003. Available at: http://www.crnusa.org/ pdfs/CRN_Bckgrndr_IOM_DVs1203.pdf Accessed April 5, 2006.

13. Council for Responsible Nutrition. Comment in Response to CL 2004/13-NFSDU, Nutrient Reference Values (NRV) for Food Labelling Purposes. April 2004. Available at: http://crnusa.org/pdfs/ CRNcommentonNRV061404.pdf. Accessed April 5, 2006.

14. Food and Nutrition Board, Institute of Medicine. WIC Food Packages: Time for a Change. Washington, DC: National Academies Press; 2004. Available at: http://www.nap.edU/cat.alog/11280.html. Accessed April 5, 2006.

15. Beaton GH. Recommended dietary intakes: individuals and populations. In: Shils ME, Olson JA, Shine M, Ross AC, eds. Modern Nutrition in Health and Disease. 9th ed. Baltimore, MD: Williams & Wilkins; 1998; 1765-1725.

16. Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes for Vitamin A, Vitamin K, Arsenic, Boron, Chromium, Copper, Iodine, Iron, Manganese, Molybdenum, Nickel, Silicon, Vanadium, and Zinc. Washington, DC: National Academies Press; 2001. Available at: http://www.nap.edu/books/0309072794/html/ . Accessed April 5, 2006.

17. Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes for Thiamin, Riboflavin, Niacin, Vitamin B6, Folate, Vitamin B12, Pantothenic Add, Biotin, andCholine. Washington, DC: National Academies Press; 1998. Available at: http:/ /www.nap.edu/openbook/0309065542/htmI/. Accessed April 5, 2006.

18. Rand WM, Pellet PL and Young VR. Meta-analysis of nitrogen balance studies for estimating protein requirements in healthy adults. Am J Clin Nutr. 2003;77:109-127.

19. Food and Agriculture Organization of the United Nations/ World Health Organization. Protein Requirements. WHO Technical Report Series No. 301. Geneva; 1965.

20. Black AE, Cole TJ. Within- and between-subject variation in energy expenditure measured by the doubly-labeled water technique: implications for validating reported dietary energy intake. Eur J Clin Nutr. 2000:54:386-394.

21. Office of Nutrition Policy and Promotion, Health Canada. Revision of Canada's Food Guide to Healthy Eating: Modeling of Food Intake Patterns. Ottawa, ON: Health Canada; 2005.

22. 2005 Dietary Guidelines Advisory Committee. Nutrition and Your Health: Dietary Guidelines for Americans. Available at: http:// www.health.gov/dietaryguidelines/dga2005/report/Accessed April 5, 2006.

23. US Department of Health and Human Services, US Department of Agriculture. Dietary Guidelines for Americans 2005. Available at: http://www. health.gov/dietaryguidelines/dga2005/document/pdf/ DGA2005.pdf. Accessed April 5, 2006.

24. U.S. Department of Agriculture. My Pyramid: Steps to a Healthier You. Available at: tittp://www.rnypyramid. gov/. Accessed April 5, 2006.

25. USDA Center for Nutrition Policy and Promotion. The Food Guide Pyramid Update, (see Federal Register 66 number 175, Sept 11, 2003 and Table 5) Available at: http://www.usda.gov/cnpp/ pyramidupdate/index.html. Accessed April 5, 2006.

26. US Food and Drug Administration, Good Reading for Good Eating. Available at: http://www.fda.gov/fdac/special/foodlabel/ goodread.html. Accessed April 5, 2006.

27. Health Canada Nutrition Labelling "Georgia"].. .Get the Facts! Available at: httpy/www.hc-sc.gc.ca/fn-an/att_ foirnats/hpfb- dgpsa4xJf/label-etiquet/te_backgrc)undle_ poirtt_e.pdf. Accessed April 5, 2006.

28. Tarasuk V. Using population-weighted Estimated Average Requirements as a basis for Daily Values in food labeling. Am J Clin Nutr. 2006;83:1228S.

29. U.S. Census Bureau, Population Division. Population Projection Branch State Interim Projections by Age and Sex. 2004 to 2030. Available at: httpy/www. cerisus.gov/populatiori/projections/ DownldFile3.xls Accessed August 10, 2005.

30. US Department of Agriculture Nutrient Data Laboratory. USUA National Nutrient Database for Standard Reference. Available at: http://www.nal.usda. gov/fnic/foodcomp/search/. Accessed April 5, 2006.

31. Moshfegh A, Goldman J, Cleveland L. What We Eat in America, NHANES 2001-2002: Usual Nutrient Intakes from Food Compared to Dietary Reference Intakes. Washington, D

Source: Nutrition Reviews

More News in this Category