Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
I t d t. t · ~ Dletary P. annlng This report is one of a series of publications resulting from a comprehensive effort initiated by the Institute of Meclicine's Food and Nutrition Board in 1993 to expand the approach to the clevel- opment of clietary reference stanciarcis. The new categories of refer- ence values have specific uses and thus are a significant departure from the previous Recommencleci Dietary Allowances (RDAs) in the United States and Recommencleci Nutrient Intakes (RNIs) in Canada. The focus of this report is to examine the appropriate use of each of the available types of Dietary Reference Intake (DRI) values in planning nutrient intakes of groups and inclivicluals. This report should be of particular use to nutrition and public health researchers in their work, to dietitians and nutritionists responsible for the education of the next generation of practitioners, and to the government professionals involved in the development and imple- mentation of national cliet and health assessments, public ecluca- tion efforts, and food assistance programs. The report reviews the statistical underpinnings for the application of the various types of DRI values in planning, illustrates sample applications, and provides guidelines to help professionals determine when specific uses are anoronriate or inanoronriate. Planners neeci to have a good unclerstancling of the DRIB, incluci- ing how each requirement was cleriveci, and whether the Tolerable Upper Intake Levels were baseci on all sources of nutrients or just fortificants and supplements. An unclerstancling of basic statistics is also needed, especially for group planners. Planners must under- stanci the concepts of risk and probability. 19
20 DIETARY REFERENCE INTAKES BACKGROUND The term Dietary Reference Intakes (DRIB) refers to a set of nutrient- baseci reference values, each of which has special uses. The clevelop- ment of DRIs expands on the periodic reports called Recommended Dietary Allowances (RDAs), which have been published since 1941 by the U.S. National Academies, and the Canaclian Dietary Stanciarcis, called Recommended Nutrient Intakes (RNIs) published since 1938 by the Canaclian government. This comprehensive effort has been uncler- taken by the Stancling Committee on the Scientific Evaluation of Dietary Reference Intakes of the Food and Nutrition Board, Insti- tute of Medicine, National Academies, at the request of the U.S. government and Health Canada. A previous report in this series (IOM, 2000a) examined the use of DRIs in clietary assessment for inclivicluals and groups. Dietary assess- ment, whether for an incliviclual or a group, compares usual nutrient intakes with estimated nutrient requirements and examines the probability of inacloquate or excessive intake. Dietary planning, on the other hanci, aims for the consumption of cliets that have accept- ably low probabilities of inacloquate or excessive nutrient intakes. Dietary planning involves using the DRls to set goals for what intakes should be. Dietary planning may be done at several different levels. It may refer to an individual planning a meal and making relevant food purchases, a food service manager in an institution planning daily menus, or a government agency planning large nutrition or food assistance programs. For the pur- poses of this report, dietary planning applies to planning intake, rather than the amount offood purchased or served. Nutritional considerations are only one component of clietary planning. Other considerations include incorporating food prefer- ences of the incliviclual or group being planned for, and the cost and availability of foocis. However, using estimates of nutrient requirements to set intake goals should be part of the planning activity. Figure 1-1 illustrates a conceptual framework clescribecT by Beaton (1994) that can be applied to the interpretation and uses of the DRIB. As shown in the framework, knowledge about both nutrient requirements and nutrient intakes feecis into two general applica- tions: cliet planning and cliet assessment. Within each of these gen- eral categories, the applications differ according to whether they are for an incliviclual or for population groups.
INTRODUCTION TO DIETARY PLANNING | Nutrient l | Requirements l l Nutrient Intakes (food plus supplements) 1 :~ Assessing ~~ L Diets | /. Group | | Individual | | Group I | Individual FIGURE 1-1 Conceptual framework uses of dietary reference standards. SOURCE: Adapted from Beaton ( 1994~ . 21 The simplicity of the above statements belies the complexity in using and interpreting DRIs to plan and assess diets. Two important factors account for this complexity. In the past, both planning and assessment applications have reliecT primarily on the former RDAs and RNIs because these were the only nutrient stanciarcis wiclely available. Often, the concepts underlying the former RDAs and RNIs were not well unclerstoocT, and thus some applications for both assessment and planning purposes were inappropriate (IOM, 1994~. Therefore, aciclitional types of reference intakes have been clevelopecT (Estimated Average Requirement, Acloquate Intake, and Tolerable Upper Intake Level). With the three additional categories of dietary reference intakes now available, applications need to be carefully consiclerecT and clearly explained so each of the categories are used appropriately. DRIs can be used in situations such as planning incli- viclual cliets; planning nutrition and food procurement for the mili- tary, prisons, nursing homes, and other institutionalized groups; food labeling and nutritional marketing; clinical dietetics; food for- tification; developing new or modified food products; and assessing food safety.
22 DIETARY REFERENCE INTAKES The approaches cliscusseci in this report for using the DRIs as a guide in planning clietary intakes for inclivicluals and for groups rely on the same basic principles that were presented in the previous report on applications of the DRIs in clietary assessment (IOM, 2000a). Those principles provide the rationale for using each of the DRIs for incliviclual and group cliet assessment, and the same ratio- nale extends to the use of the DRIs in cliet planning. WHAT ARE DIETARY REFERENCE INTAKES? As inclicateci above, the term Dietary Reference Intakes (DRIB) refers to a set of at least four nutrient-baseci reference values that can be used for planning and assessing cliets and for many other purposes. An important principle underlying both the former Recommencleci Dietary Allowances (RDAs) and Recommencleci Nutrition Intakes (RNIs) and the new DRIs is that these are standards for healthy people they are not appropriate for individuals or groups who are ill orfor repletion of deficient individuals. The concepts underlying the new DRIs differ from the former RDAs and RNIs as inclicateci in Box 1-1. Processes Used to Establish the Dietary Reference Intakes In establishing the EAR or Acloquate Intake (AI) for nutrients, a requirement is clefineci as the lowest continuing intake level of a
INTRODUCTION TO DIETARY PLANNING 23 nutrient that will maintain a clefineci level of nutriture in an inclivici- ual. The chosen criterion of nutritional acloquacy upon which this requirement is baseci is different for each nutrient and is iclentifieci in the DRI nutrient reports (IOM, 1997, 1998a, 2000b, 2001,2002a). In some cases, the criteria may differ for inclivicluals at different life stages for the same nutrient. In developing the DRIB, emphasis is placed on the reasons underlying the particular criterion of acle- quacy used to establish the requirement for each nutrient. A more cletaileci discussion of the origin and framework of the DRIs is pre- sented in Appendix A. The EARs are baseci on a thorough review of the scientific litera- ture for health outcomes associated with the nutrient. The criteria and eviclence-baseci rationale used for setting each EAR are clearly specified. An estimate of the variation in the requirement is also specified, and is used to set the RDA. When ciata are inacloquate to establish an EAR and RDA, other approaches are used to establish an intake goal, which is clesignateci an AI. The process used to estab- lish the UL involves the estimation of an uncertainty factor that is applied to a no-observeci-aciverse-effect level (NOAEL) or to a lowest-observeci-aciverse-effect level (LOAEL) baseci on human or animal ciata related to iclentifieci hazards. · . · . · . Estimated Average Requirement1 The Estimated Average Requirement (EAR) is the usual intake level that is estimated to meet the requirement of half the healthy incli- vicluals in a life stage and gentler group. At this level of intake, the other half of the healthy inclivicluals in the specified group would not have their neecis met. The EAR is baseci on a specific criterion of acloquacy, cleriveci from a careful review of the literature. When selecting the criterion, reduction of disease risk is consiclereci along with many other health parameters. For example, the EAR for vitamin C is baseci on "an amount thought to provide antioxidant protection as derived from the correlation of such protection with neutrophil ascorbate concentrations" (IOM, 2000b). For energy, 1 It is recognized that the definition of EAR implies a median as opposed to a mean or average. The median and average would be the same if the distribution of requirements followed a symmetric distribution, and would diverge as the distribu- tion became skewed. Two considerations prompted the choice of the term EAR: (1) data are rarely adequate to determine the distribution of requirements, and (2) precedent has been set by other countries that have used the term EAR for reference values similarly derived (COMA, 1991~.
24 DIETARY REFERENCE INTAKES the situation is somewhat different. Energy requirements are esti- mateci on an incliviclual basis using a person's gentler, age, height, weight, and physical activity level to estimate total energy expencli- ture; thus the specific criterion of acloquacy is maintenance of a healthy body mass inclex with a healthy level of physical activity. Recommended Dietary Allowance The Recommended Dietary Allowance (RDA) is the clietary intake level that is sufficient to meet the nutrient requirement of nearly all healthy inclivicluals in a particular life stage and gentler group. If the distribution of requirements in the group is assumed to be normal, the RDA is computed from the EAR by Hilling two stanciarci deviations of the requirement (8DREQ) as follows: RDA = EAR + 2 SDREQ The stanciarci deviation of the requirement distribution can be observed directly if sufficient ciata are available. Often this is not the case, and the stanciarci deviation is estimated by assuming a specific coefficient of variation (CV) for the average requirement. A CV of 10 percent has been used for many nutrients (IOM, 1997, 1998a, 2000b, 2001), and for these, the RDA equals 120 percent of the EAR. Therefore, assuming a normal distribution, 97 to 98 percent of the individuals in the group will have a requirement that is below the RDA. If the distribution of requirements is known to be skewoci rather than normal (for example, iron requirements of menstruat- ing women), the RDA is obtained by fincling the usual intake level that is at the 97th to 98th percentile of the requirement clistribu- tion. In either case, the RDA developed in the DRI process differs conceptually from the former RDAs and RNIs since with the estab- lishment of an EAR, the RDA is cletermineci quantitatively rather than through the use of judgment-based safety factors. The RDA is intencleci for use primarily as a goal for intake of inclivicluals. Because the RDA is often cleriveci directly from the EAR and an estimate of variability of the requirement distribution, if ciata are insufficient to establish an EAR, no RDA can be set. Adequate Intake If sufficient scientific evidence is not available to establish an EAR, and thus determine an RDA, a reference intake called an Adequate Intake (Al) may be cleriveci instead. The AI is a value baseci on
INTRODUCTION TO DIETARY PLANNING 25 experimentally derived levels of intake or the mean nutrient intake by a group (or groups) of apparently healthy people who are main- taining a defined nutritional state or criterion of adequacy. Exam- ples of defined nutritional states include normal growth, mainte- nance of normal circulating nutrient values or biochemical indices, or other characteristics of nutritional well-being or general health related to the nutrient. For example. the AI for voting infants is based on the daily mean . , ~ , · · . . ~ . . . · . . ~ . . . ~ . . nutrient In take supplied by human milk tor healthy, tull-term infants who are exclusively breasted. For adults, the AI may be based on data from a single experiment (e.g., the AI for choline tIOM, 1998a]), on estimated dietary intakes in apparently healthy population groups (e.g., the AI for pantothenic acid tIOM, 1998a] ), or on combined data from different approaches (e.g., usual dietary intake and experimentally altered intakes of calcium in adult women tIOM, 19974~. The AI is thus expected to exceed the true EAR (and often the RDA) if it could be set for the same specified criterion of nutritional adequacy. In the absence of an EAR (and RDA) for a nutrient, the AI can be used as the intake goal. The issuance of an AI is an indication that more research is needed to determine with confidence the mean and distribution of require- ments for a specific nutrient. As this research is completed, it should be possible to replace estimates of AIs with EARs and RDAs. To;terab;te Upper Intake [event The Tolerable Upper Intake Level (UL) is the highest level of con- tinuing daily nutrient intake that is likely to pose no risk of adverse health effects to almost all individuals in a specified life stage and gender group. As intake increases above the UL, the potential risk of adverse health effects increases. The term tolerable intake was chosen to avoid implying a possible beneficial effect from levels of intakes above the RDA. Instead, the term is intended to connote a level of intake that can, with high probability, be tolerated biologi- cally. The UL is not a recommended level of intake, and there is no currently established benefit to healthy individuals associated with ingestion of nutrients in amounts exceeding the RDA or AI. The UL is baseci on an evaluation conclucteci using the methoclol- ogy for risk assessment of the adverse effects of nutrients (IOM, 1999~. (A cletaileci explanation of this methodology is also inclucleci in all of the DRI nutrient reports.) The neeci to establish ULs grew in part out of the increased fortification of foocis with nutrients and the increased use of dietary supplements. Details are given for each
26 DIETARY REFERENCE INTAKES nutrient on how the UL was established (IOM, 1997, 1998a, 2000b, 2001, 2002a). For some nutrients there may be insufficient ciata on which to develop a UL. The lack of a UL cannot be interpreted as meaning that high intake poses no risk of adverse effects. Unless otherwise stated in the DRI nutrient reports, values given for EARs, RDAs, AIs, and ULs are baseci on the total intake of the nutrient naturally occurring in food, acicleci to food as a fortificant, and from supplements. IMPLEMENTATION OF DIETARY PLANNING FOR INDIVIDUALS AND GROUPS Planning cliets refers to determining what usual nutrient intake should be. Regardless of whether one is planning cliets for inclivicluals or groups, the goal is to have cliets that are nutritionally acloquate, or conversely, to ensure that the probability of nutrient inacloquacy or excess is acceptably low. As will be clescribeci in depth in this report, how this goal is implemented differs when planning for inclivicluals compared to planning for groups. Nevertheless, the underlying considerations are similar. At the incliviclual level, usual intake is clefineci as the incliviclual's average intake over a long period of time. As cliscusseci in greater detail in the Dietary Reference Intake (DRI) report on clietary assess- ment (IOM, 2000a), because of the large ciay-to-ciay variation in incliviclual intake, intake on one or even several clays may provide inaccurate estimates of an incliviclual's usual intake. Similarly, for groups, the focus for cliet planning is the distribution of usual intake, which is the distribution of the long-term average intakes of incli- viduals in the group. Usual intake distributions can be estimated by adjusting the observed intake distributions using statistical tech- niques (NRC, 1986; Nusser et al., 1996) . By removing the ciay-to-ciay variation in intakes (within-person variation), the resulting adjusted distribution better reflects the incliviclual-to-incliviclual variation of intakes within the group. Another consideration in the implementation of clietary planning is the concept of an acceptably low probability of nutrient inacle- quacy (probability that intake does not meet requirement) or, con- versely, a high probability of nutrient acloquacy. For inclivicluals, an acceptably low probability of nutrient inacloquacy has been tracli- tionally accomplished by planning for the incliviclual's usual intake to be at the Recommencleci Dietary Allowance for the nutrient, such that the probability of inacloquacy floes not exceed 2 to 3 percent.
INTRODUCTION TO DIETARY PLANNING 27 To ciate, planning for groups has generally not incorporated plan- ning for a low prevalence of nutrient inacloquacy, in large part because the tools required Knowledge of the Estimated Average Requirement and the usual intake distributions have not been wicle- ly available. Thus, there is no convention about what prevalence of inacloquacy is acceptably low. It is in the professional judgment of the nutritionist or planner to determine what is an acceptably low probability of nutrient inacloquacy for an incliviclual or prevalence of inacloquacy for groups. The level selected should be clearly stated. Similarly, in applying the DRIs for planning, professional judgment is required to determine the likelihood of any recognized benefit of increasing intakes beyond their current level. CAVEATS REGARDING THE USE OF DIETARY REFERENCE INTAKES IN DIETARY PLANNING AND ASSESSMENT Dietary planning and assessment are inextricably linked. Assess- ment is used as a basis for planning and to evaluate whether the planning goals have been met. Those assessing and planning cliets should be aware of limitations in the ciata that underpin the Dietary Reference Intakes (DRIB) and their application: there is uncertainty associated with the estimates of the Estimated Average Require- ments (EARs) themselves, and clietary intake and food composition ciata are subject to inaccuracy. [imitations in the Data on Nutrient Requirements Detailed consideration of the DRI reports for specific nutrients (IOM, 1997, 1998a, 2000b, 2001, 2002a) can provide insight into both what is known and what information is still needed to further define intakes that support health. In interpreting the DRIs for use in dietary planning, planners should be aware that often the EARs are baseci on ciata from a limited number of inclivicluals; that for most nutrients the precise variation in requirements is not known and has been approximated from the variation in related physiological parameters; that, in the absence of evidence to the contrary, the variation in individual requirements has been assumed to follow a normal distribution; that the EAR has often been extrapolated from one population group to others that differ in life stage and gentler; and that the degree of uncertainty associated with the EAR has not been specified. By definition, EARs are estimates they are not clefineci with 100 percent accuracy. Thus, although the best avail- able evidence was used, gaps in the knowledge base remain.
28 DIETARY REFERENCE INTAKES Choice of Requirement Criterion Knowledge of the criterion used by the DRI panels to determine the EAR and Recommencleci Dietary Allowance (RDA) can help in assessing the potential impact of not meeting these guidelines. This may affect setting goals for nutrient intake, including selection of an acceptable group prevalence of clietary inacloquacy (e.g., the pro- portion of a group with intakes below the EAR). In establishing the DRIB, the requirements for most nutrients have been presented as a single endpoint for various life stage and gentler groups, rather than as multiple endpoints. To the extent that for most nutrients a single endpoint has been established for an EAR and RDA, this approach differs from that originally recommencleci by NRC (1986) and adopted by the Joint Food and Agriculture Organization and World Health Organization Expert Consultation on the requirements of vitamin A, iron, folate, and BE (FAD/WHO. 1988~. These groups recommencleci both a basal requirement level (the amount of nutrient neecleci to prevent a clinically detectable impairment of function) and a normative storage requirement level (the amount of nutrient neecleci to maintain a desirable level in tissues). However, the DRI process floes allow for multiple enci- points to be used where the ciata exist, and to ciate this has been clone for vitamin A. An EAR has been set for the reversal of night blindness, and an EAR and RDA have also been set for the mainte- nance of liver stores. A planner might want to ensure that intakes would result in a minimal (near zero) prevalence of inacloquacy with regard to night blindness, but might be willing to accept, and thus plan for, a somewhat higher prevalence of inadequacy with regard to maintenance of normal liver stores. Inadequate Dietary Intake Versus Inadequate Nutritional Status Planning cliets for groups involves choosing an acceptable group prevalence of clietary inacloquacy (see Chapter 3~. Theoretically, this would correspond to the prevalence of inacloquate nutritional status with regard to the criterion used to establish the EAR. For example, if planners chose to maintain the current distribution of vitamin B6 intake in the United States in women age ci 31 to 50 (see appendixes to the DRI publications for tables describing the popu- lation distributions of nutrient intakes tIOM 1997, 1998a, 2000b, 2001, 20021), they would be accepting an apparent group preva- lence of dietary inadequacy between 10 and 15 percent, according to ciata from the Third National Health and Nutrition Examination
INTRODUCTION TO DIETARY PLANNING 29 Survey (NHANES III). If the assumptions involved in establishing the EAR were correct and applied to all population groups, one would expect to observe similar proportions consuming vitamin B6 below the EAR and with low plasma pyricloxal phosphate levels (i.e., inacloquate nutritional status with regard to the indicator used to set the EAR). In practice, however, the apparent prevalence of dietary inadequacy of a nutrient may not be equivalent to the preva- lence of inadequate nutritional status for the same nutrient. Sources of error contributing to any observed discrepancies between estimates of the prevalence of inacloquate intake and inaci- equate nutritional status include those involved in estimating clietary intakes. These have been reviewoci in the DRI report on clietary assessment (IOM, 2000a), and include an incomplete knowledge of (1) the nutrient composition offoocis, (2) thenutrientbioavailabil- ity from different food and supplemental sources, (3) the usual intakes as compared with short-term intakes, and (4) the uncler- reporting of self-reporteci clietary intakes. The uncertainties involved in estimating nutrient requirements can also contribute to observed discrepancies, as can the lack of population data on the biochemical indicators of nutrient acloquacy used to establish the requirement estimates. Sources of Error in Planning for Dietary Intake Uncertainty of Requirement Estimates For some nutrients, the sources of error in estimating intakes and requirements are not extreme, and the apparent prevalence of clietary inacloquacy (e.g., the proportion below the EAR) corre- sponcis reasonably well to the prevalence of inacloquate nutritional status with regard to the criterion used to establish the EAR. For example, the EAR for iron was established as the amount of iron needed to meet body functions with minimal storage, and this was cletermineci to be reflected by a serum ferritin concentration of about 15 ,ug/L (IOM, 2001~. When the prevalence of inacloquate iron intakes was compared to the prevalence of apparent biochemi- cal deficiency (low serum ferritin concentrations), the agreement was reasonable for most life stage and gentler groups (IOM, 2001~. If planners chose to reduce the prevalence of dietary inadequacy (anci, by inference, the prevalence of inacloquate nutritional sta- tus), this could be clone using the methods clescribeci in Chapter 3 of this report.
30 DIETARY REFERENCE INTAKES In other cases, however, errors in estimating clietary intake make it difficult to use clietary intake ciata to plan cliets with acceptable levels of inacloquacy. This is especially true for vitamin E. Food com- position ciata neeci to be upciateci for this nutrient, and clietary intakes are frequently unclerestimateci clue to underreporting (which may be particularly problematic for fat, a major carrier of vitamin E) (Mertz et al., 1991~. Data from NHANES III suggest that the majority of adults age ci 31 to 50 haci apparently inacloquate clietary intakes (IOM, 2000b), leaving the impression that cliets must be planned with aciclitional vitamin E to meet the requirements for the population. However, examination of the serum oc-tocopherol distributions in NHANES III reveals that fewer than ~ percent haci plasma concentrations below the 12 ,umol/L (516 ,ug/ciL) used to set the EAR. Thus, for vitamin E, it is clear that the apparent preva- lence of clietary inacloquacy floes not correspond to the prevalence of inacloquate nutritional status as assessed biochemically. Thus, when choosing a planning goal, especially when planning for groups, planners neeci to consider the limitations of the clietary intake ciata, the consequences of not meeting the criterion used to determine the EAR, the results of available biochemical ciata, and the goals of clietary planning for specific situations. As inclicateci earlier, a nutrient will usually have a Tolerable Upper Intake Level and either an EAR and RDA or an Acloquate Intake (AI). However, for energy and the macronutrients, this is not always the case. For example, no DRIs have been set for total fat for incli- vicluals over 1 year of age. Instead, an Acceptable Macronutrient Distribution Range of 20 to 35 percent of energy from dietary fat is recommencleci for adults to minimize risk of adverse health out- comes. For energy, no DRIs have been set an estimate of the total energy expenditure associated with an individual's gender, age, height, weight, and physical activity level is used. Uncertainty of Dietary Intake Estimates Another source of error that has potentially profound implica- tions for dietary assessment and planning is the accuracy of self- reported dietary intakes. A variety of study designs has been employed to examine the accuracy of clietary assessment techniques to measure individuals' true energy intakes over defined time periods. The weight of evidence from this extensive literature indicates that a sizeable proportion of inclivicluals systematically misreport their intakes, with the tendency toward underreporting. In a now classic study by Mertz and colleagues (1991), the usual energy intake of
INTRODUCTION TO DIETARY PLANNING 31 266 adults (estimated from 7 to 35 days worth of food records) was determined to be insufficient to maintain body weight in 81 per- cent of subjects. The average discrepancy between self-reported en- er~v intake and the intake required for weight maintenance was Jo 1 ~ 17 ~ ~ 1 1 ~ ,r , 1 1 rim 1 · , · , 1 1 /()() kcal. More recently, seli-reports ot dietary In take have been compared to energy expenditure measured by doubly labeled wa- ter, on the assumption that energy expenditure is equivalent to intake in situations of energy balance. Such comparisons have typi- cally revealed substantial underreporting of intakes, even when changes in body stores during the study period are taken into ac- count (Bandini et al., 1990; Black et al., 1993; Johnson et al., 1998; Kaczkowski et al., 2000; Martin et al., 1996; Prentice et al., 1986; Tomoyasu et al., 1999~. Furthermore, although the nature and sources of measurement error are known to vary across dietary as- sessment methods, the problem of underreporting appears to be pervasive irrespective of whether food records, dietary recalls, diet histories, or food frequency questionnaires are used to assess intake (Black et al., 1991; Sawaya et al., 1996~. Self-reports of dietary intake have also been compared to esti- mates of energy expenditure based on factorial methods, although at the individual level, this method yields a less precise estimate of energy expenditure than the doubly labeled water technique. Typi- cally, reported energy intake (EI) is expressed as a ratio of estimated basal metabolic rate (BMReS~), based on age, sex, self-reported or measured body weight, and possibly height. A variety of approaches to evaluating the adequacy of EI/BMReS~ can be found in the litera- ture. Goldberg and colleagues (1991) have proposed a method to estimate a minimum plausible EI/BMReS~ by applying a series of assumptions that take into account within-person variation in energy intake, random error in the estimation of an individual's basal meta- bolic rate based on the predictive equation used, and variation in an individual's physical activity level. When these methods have been applied to population-based dietary survey data, comparisons indicate that 10 to 50 percent of respondents may be underreport- ing their food (energy) intakes (Black et al., 1991; Briefel et al., 1997; Johansson et al., 1998; Stallone et al., 1997) . While the underreporting of energy intakes appears well docu- mented, it is unclear how this affects the accuracy of self-reported nutrient intakes. Research into this question is limited by the absence of reliable reference biomarkers for intakes of many nutrients. Studies in which the assessment of self-reported energy intake using the doubly labeled water method has been combined with the measure- ment of urinary nitrogen excretion to assess self-reported protein J ' 1 I
32 DIETARY REFERENCE INTAKES intake suggest that energy intake may be more prone to under- estimation than protein intake (Larsson et al., 2002~. Importantly, these findings imply that all nutrients are not proportionally under- reported; rather, particular foods or classes of foods must be selec- tively underreported. When the reported intakes by individuals classed as energy underreporters have been compared to those whose energy intakes appear more plausible, underreporters have often been found to report a lower percentage of energy from fat (Becker and Welten, 2001; Becker et al., 1999; Briefel et al., 1997; Goris et al., 2000~. Such comparisons have also indicated lower reported consumption of particular classes of foods among under- reporters (Becker and Welten, 2001; Krebs-Smith et al., 2000~. How much one can infer about the nature of underreporting from these studies hinges on the validity of the assumption that underreporters' dietary patterns are the same as those not deemed to be under- reporting. Nonetheless, it would appear overly simplistic to assume that the nutrient intakes of individuals who systematically under- report their energy intakes are underreported to the same degree. The implications of underreporting for dietary assessment and planning are profound given the need to rely on self-reported dietary intakes for information about usual intake patterns. Because individuals' intakes of energy and nutrients are intertwined, the systematic underestimation of true usual energy intakes for some proportion of the population is likely to mean an underestimation ~ , . , . , . 11 By.. . .~1 , , .. , ~ . ~ ot nutrient Intakes as well. lhIS IS Illustrated In a recent analysts ot data from a Swedish population survey in which the proportion of individuals with nutrient intakes below the average requirement decreased substantially when individuals reporting "implausibly or dubiously low energy intakes" (defined as EI/BMReS~ < 1.10 and 1.10 to 1.34, respectively, with EI estimated from a 7-day food record) were excluded from the analysis (Becker and Welten, 2001) . Planners are currently limited as to what they can do to correct problems of underreporting. The application of EI/BMReS~ thresh- olds to identify underreporters can be problematic, given the need to make assumptions about individuals' usual physical activity levels (often in the absence of good measures of physical activity) and the error inherent in estimates of BMR (an error that is compounded when BMR is calculated using self-reported weight and height). Further, it cannot be assumed that all those with reported energy intakes above the chosen EI/BMReS~ threshold have accurately reported their intakes. Even if underreporters are somehow identi- fied, the exclusion of their data from population-level assessments of nutrient adequacy clearly threatens the ability to generalize assess-
INTRODUCTION TO DIETARY PLANNING 33 ment results to the population as a whole. This is because it cannot be assumed that the cliets of inclivicluals iclentifieci as unclerreporters are identical to those not so iclentifieci. Well-accepteci, valiciateci methods to statistically correct for the effects of underreporting on the estimated distribution of usual intakes are presently lacking. The statistical procedures proposed to adjust intake distributions for within-person variation in intake (e.g., NRC, 1986; Nusser et al., 1996) do not correct for systematic errors in reporting. Application of the residual method of energy adjustment (Willett and Stampfer, 1986) to nutrient distributions has been proposed as one means to recluce the bias associated with energy underreporting without excluding the ciata of unclerreporters in some kinds of epiclemiological analyses (Stallone et al., 1997~. This adjustment method, however, floes not provide an appropriate correction of underreporting for clietary intake ciata to be used in assessment and planning applications of the DRIB. Energy acljust- ment methods cannot eliminate bias clue to selective underreporting of foocis; instead these methods effectively "assume" that nutrients have been unclerreporteci in direct proportion to energy. Further, energy adjustment floes not provide corrected estimates of absolute intake. Thus, energy-acljusteci ciata are not useful in assessments of nutrient acloquacy. In summary, energy underreporting is clearly a serious problem in clietary surveys; it limits the accuracy with which planners can estimate usual energy and nutrient intakes in population groups of interest. Given the current absence of inexpensive, valiciateci meth- ods to readily identify underreporting in dietary intake surveys and statistical methods to correct for underreporting in self-reported energy and nutrient intakes, planners are severely limited in their ability to aciciress this problem. This problem not only highlights the importance of employing thorough, standardized procedures to collect dietary data, but it also flags the urgent need for more research into statistical methods to analyze and adjust for underreporting in self-reported intake data. In interpreting the results of dietary assessments prior to determining planning goals, planners should look to other sources of data on nutritional status (e.g., anthropometric, clinical, or bio- chemical assessments) for corroborating evidence. In interpreting dietary assessment results, planners may also find it useful to esti- mate the extent of energy underreporting in their data by applying factorial methods to compare reported energy intakes with estimates of energy expenditure. However, the crudeness of these estimates should be recognized. Until better methods of identification and
34 DIETARY REFERENCE INTAKES adjustment are clevelopeci, it is not recommencleci that ciata acljust- ments be undertaken. Planners can use clietary intake ciata from national surveys, but should remain aware of the inaccuracies of the ciata when setting intake goals baseci on the DRIs and assessing achievement of those goals.