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.
7 Examples of Planning for Groups SUMMARY Several applications of group planning are presented in this chap- ter. Two examples focus on normal group fouling situations where the distribution of intakes is shifted but the shape of the clistribu- tion is not explicitly changed. Two examples focus on planning for heterogeneous groups using a simple and a complex (but theoreti- cally more correct) nutrient density approach. The final two examples discuss the problem of planning interventions designed to change the shape of the usual intake distribution of one or more nutrients in a targeted population group. It is often difficult to plan cliets that will achieve exactly the clesireci effect. Therefore, when planning normal cliets or clietary interven- tions it is critically important to assess the likely effects not only on the target group, but also on other groups that would be affected by the intervention. Important unpredictable factors such as food preferences, partici- pation rates in food assistance programs, or population-based edu- cational programs make the job of an intervention planner very difficult. Typically, forecasting the effect of an intervention is not straightforward, and several cycles of planning followed by assess- ment may be neecleci. The applications clevelopeci in this chapter are hypothetical. 107
108 DIETARY REFERENCE INTAKES INTRODUCTION Planning diets for population subgroups is carried out in many diverse settings and thus has multiple and varied applications. Some of the more visible group-planning applications include planning diets for institutionalized groups, food and nutrition assistance pro- grams, food fortification, nutrition education for groups, and mili- tary food and nutrition planning. The discussion below provides an in-depth analysis of six specific planning applications. Examples (1) an assisted living facility for seniors and (2) school nutrition programs, present the principles described in Chapter 3 for shifting the distribution of usual intakes. Examples (3) a group of teen boys, adult men, and adult women using the simple nutrient density approach and (4) a group of teen boys, adult men, and adult women using the nutrient density distri- bution approach, present the approaches described in Chapter 4. Finally, examples (~) nutrient supplementation and (6) food fortifi- cation, illustrate how interventions intended to shift the distribu- tion of usual intakes may also change the shape of the usual intake distribution. This discussion is not intended to prescribe how these planning activities should be conducted. Rather, based on the prin- ciples for group planning developed in Chapters 3 and 4, the dis- cussion of these examples is intended to present the issues involved in these planning applications. The group-planning framework should be applied in pilot situa- tions before it is adopted for large-scale programs. PLANNING DIETS IN AN ASSISTED-LIVING FACILITY FOR SENIOR CITIZENS An example of planning diets for institutionalized groups is menu planning for senior citizens who reside in an assisted-living facility. Menus planned for these institutions usually assume that the resi- dents have no other sources of foods or nutrients, and thus the menus are designed to meet all nutrient needs of the residents. Based on the framework developed in Chapter 3, the goal of menu planning is to provide meals that supply adequate nutrients for a high proportion of the residents, or conversely, to ensure that the prevalence of inadequate intakes are acceptably low among the res- idents. An important note, and caveat perhaps, is that to fully im- plement the planning approaches described in this report, data on usual intakes must be available. Unfortunately, such data are sel- dom available; planners for these and other institutionalized groups
EXAMPLES OF PLANNING FOR GROUPS 109 (e.g., prisons, boarding schools) frequently do not collect clietary intake ciata in order to evaluate their menu planning. It is possible to generate usual intake ciata on the target population through ciai- ly food intake records or intake recalls on each incliviclual. Howev- er, if the facility is large (e.g., more than 100 resiclents), intakes could be measured on a representative subsample of residents. Us- ing this technique, two nonconsecutive clays or three consecutive clays of food intake records or recalls are necessary. Alternatively, records of amounts served and plate waste ciata for inclivicluals mon- itoreci, again for a minimum of two nonconsecutive or three con- secutive clays, can be used. In both cases, ciata should be acljusteci to remove within-person variability and to obtain the usual nutrient intake distribution by using procedures such as those clevelopeci by Nusser and colleagues (1996) or the National Research Council (NRC, 1986~. Another possibility is to use usual nutrient intake distributions from another group in which the members are of similar age to the target group. Ideally, such ciata would also be for a similar (e.g., gentler mix, ethnicity) institutionalized population, since the varia- tion in the distribution of usual intakes is likely to differ among inclivicluals who live in institutionalized settings and those who do not. If such comparable usual intake ciata are not available, then the only option may be to use usual intake distributions from national surveys such as the Continuing Survey of Food Intakes by Inclivicluals (CSFII) or the Third National Health and Nutrition Examination Survey (NHANES III). From the most appropriate ciata set available as clescribeci above, the Planner examines the proportion of the group with usual intakes less than the Estimated Average Requirement (EAR) (for each of the nutrients for which EARs have been established as an estimate of the prevalence of inacloquate intakes. If the prevalence is unacceptably high for one or more nutrients, then intakes need to be increased. As clescribeci in Chapter 3, to estimate the amount of the increase for a given nutrient, the difference between the EAR for that nutrient and the usual intake level corresponding to the selected percentile of the current usual intake distribution (which is the chosen acceptable prevalence of inacloquacy) is cletermineci. The meclian usual intake should be increased by this amount, assuming the shape of the distribution is not expected to change. It is crucial to reassess intakes after the change is macle, especially if the change is large, because it is possible (even likely) that the shape of the distribution will change. As an example, consider a planner who is developing a menu for
110 DIETARY REFERENCE INTAKES an assisteci-living facility in which the residents are retired nuns age ci 70 years and above. For this age group, the EAR for vitamin B6 is 1.3 mg/ciay (IOM, 1998a). Assume that no ciata can be located on the distribution of usual intakes of this group or a similar group, and that resources are not available to conduct a clietary survey in the institution. How could the planner proceed to determine the target intake distribution of vitamin B6 neecleci to attain an acceptable prevalence of inacloquacy? Step I. Determine an acceptably Stow prevalence of inadequacy. For vitamin B6, the EAR was set at a level acloquate to maintain plasma pyricloxal phosphate levels at 20 nmol/L (IOM, 1998a). This plasma level is not accompanied by observable health risks, and thus allows a moderate safety margin to protect against the clevelop- ment of signs or symptoms of deficiency. This cutoff level was select- eci recognizing that "its use may overestimate the B6 requirement for health maintenance of more than half the group" (IOM, 1998a). For this reason, assume that the planner has cletermineci that a 10 percent prevalence of inadequacy (i.e., 10 percent with intakes below the EAR) would be an acceptable planning goal. Step 2. Determine the target usual nutrient intake distribution. Next, the planner needs to position the intake distribution so the nutrient intake goals are met. In this example, the planner clecicles that the prevalence of inacloquacy in the group will be set at 10 percent, and as a result the usual intake distribution of the group should be positioned such that only 10 percent of the group has usual intakes less than the EAR. Using the EAR as a cut point for estimating the prevalence of inadequate intakes builds directly on the approaches previously clescribeci for assessing intakes (IOM, 2000a). Because ciata on the usual nutrient intake distributions of the residents are not available, other sources must be used to estimate the target usual nutrient intake distribution. Data on the clistribu- tion of usual clietary intakes of vitamin B6 from CSFII (conclucteci in 1995), NHANES III (conducted between 1988 and 1994), and the Boston Nutritional Status Survey (conclucteci between 1981 and 1984) are available (IOM, 1998a).i The adjusted percentiles for ~ Caution should be used when selecting data sets. If more recent data sets were used in this example, it would provide a better reflection of changes in fortifica- tion levels.
EXAMPLES OF PLANNING FOR GROUPS 111 women age ci 70 years and above (in the Boston survey, age ci 60 years and above) are summarized in Table 5-1. Assuming there are no changes in the shape of the distribution, the amount of the shift can be calculated as the aciclitional amount of the nutrient that must be consumed to recluce the proportion of the group that is below the EAR. This is accomplished by determining the difference between the EAR and the intake at the acceptable prevalence of inacloquacy (in this case, the 10th percentile of the usual intake distributions. Examination of the ciata from the three surveys shows that estimated usual intakes of vitamin B6 vary by as much as 30 percent among the surveys. As a result, the difference between the EAR of 1.3 mg and the intake at the 10th percentile varies, clepenci- ing on which ciata are used: for NHANES III the difference is 0.26 mg (1.3 mg - 1.04 mg = 0.26 mg); for CSFII, the difference is 0.42 mg (1.3 mg- 0.88 mg = 0.42 mg), and for the Boston survey, the difference is 0.7 mg (1.3 mg- 0.6 mg = 0.7 mg). In this exam- ple, the planner may have no reason to choose ciata from one par- ticular survey as "more applicable" to his group than another, so he may estimate target usual nutrient intake distributions using all three ciata sets. Accordingly, the target intake distributions shift up by 0.26 ma, by 0.42 ma, and by 0.7 mg using NHANES III, CSFII, and the Boston survey, respectively. In each case the target usual nutrient intake distribution would leaci to the accepted prevalence of inacloquacy. Rather than choosing one set of survey ciata over another, the planner could simply average the summary measures clescribeci in the next section. TABLE 5-1 Selected Percentiles of the Distributions of Usual Intake of Vitamin B6 from Foods in Older Women Percentile of Usual Intake Distribution of Vitamin B6 (mg/day) Studya n 5th 1 0th 25th 50th 75th 90th 95th CSFII 221 0.76 0.88 1.11 1.41 1.76 2.12 2.35 NHANES III 1,368 0.92 1.04 1.24 1.53 1.93 2.43 2.76 Boston 281 0.5 0.6 0.7 1.0 1.3 1.6 1.8 a CSFII = Continuing Survey of Food Intakes by Individuals (women > 70 y), NHANES III = Third National Health and Nutrition Examination Survey (women > 70 y), Boston = Boston Diet Study (women > 60 y). SOURCE: IOM (1998a).
112 DIETARY REFERENCE INTAKES Step 3. Select a summary measure of the target usual nutrient intake distribution to use in planning. After the planner has estimated a target usual intake distribution, this information neecis to be operationalizeci into a menu. In order to do this, the planner will first have to select a summary measure of the target usual nutrient intake distribution to use as a tool in plan- ning the menu. The meclian of the target intake distribution is the most useful; it can be calculated as the meclian of the current intake distribution, plus (or minus) the amount that the distribution neecis to shift to make it the target usual intake distribution. In the current example, although the baseline intakes at the 10th percentile and the meclian differ among the three surveys, the esti- mates of the meclians of the target usual intake distributions are quite similar, as shown in Table 5-2. Assuming that a 10 percent prevalence of intakes below the EAR was consiclereci acceptable, a meclian intake for vitamin B6 of 1.7 to 1.8 mg/ciay would be the planning goal. Accordingly, the menu would neeci to be planned so that vitamin B6 intakes would be at this level. Estimates of target nutrient intakes must be converted to esti- mates of foocis to purchase, offer, and serve that will result in the usual intake distributions meeting the intake goals. As cliscusseci previously, designing menu offerings to meet intake targets is a dif- TABLE 5-2 Identification of the Target Meclian Intakea of Vitamin B6 to Obtain a 10 Percent Prevalence of Inacloquacy in Olcler Women Difference Target Intake at (EAR- Median Median EAR 10th Percentile intake at 10th Intake Intake Studyb (mg/day) (mg/day) percentile) (mg/day) (mg/day) CSFII 1.3 0.88 0.42 1.41 1.83 NHANES III 1.3 1.04 0.26 1.53 1.79 Boston 1.3 0.6 0.7 1.0 1.70 a The target median intake is estimated by adding the difference between the Estimated Average Requirement (EAR) and the intake at the acceptable prevalence of inadequacy (in this case, 10%) to the observed median intake. b CSFII = Continuing Survey of Food Intakes by Individuals, NHANES III = Third National Health and Nutrition Examination Survey, Boston = Boston Diet Study. SOURCE: IOM (1998a).
EXAMPLES OF PLANNING FOR GROUPS 113 ficult task. Meals with an average nutrient content equal to the meclian of the target usual nutrient intake distribution may not meet the planning goals, as inclivicluals in a group tend to consume less than what is offered and served to them. Thus, the planner might aim for a menu that offers a choice of meals with a nutrient content range that includes, or even exceeds, the meclian of the target usual nutrient intake distribution. Step 4. Assess implementation of the plan. Ideally, after the menu has been planned and implemented, a survey would be conclucteci to assess intakes and determine whether the planning goal has been attained. This would then be used as the basis for further planning. PLANNING MENUS FOR A SCHOOL NUTRITION PROGRAM Probably the largest group planning application in the United States is for the nutrition assistance programs sponsored by the U.S. Department of Agriculture (USDA) . These include the Food Stamp Program; the Supplemental Nutrition Program for Women, Infants, and Children; the Child and Adult Care Feecling Program; the National School Lunch Program (NSLP); the School Breakfast Pro- gram (SBP); and the Summer Food Service Program. The NSLP and SBP are federally administered nutrition programs that operate ciaily in the nation's schools. The primary objective of these programs is "to safeguard the health and well-being of the Nation's children" (Richarci B. Russell National School Lunch Act, 42 U.S.C. § 1751 (2) t20021 ~ . The Recommencleci Dietary Allowanc- es (RDAs) have long formed the basis for fooci-baseci menu plan- ning in the school nutrition programs. USDA regulations require that NSLP lunches provide, over time, one-thirci of the RDA for key nutrients. The goal of the SBP is to provide one-fourth of the RDA. Findings from two school nutrition clietary assessment studies incli- cate that, on average, school meals meet or exceed their goals of offing one-thirci of the RDA for lunch and one-fourth of the RDA for breakfast (Burgharcit et al., 1995; Devaney et al., 1995; Fox et al., 2001~.2 2 It is important to note that program regulations are based on the former RDAs. In addition to the implications of the framework developed for group plan- ning in this report, the concepts underlying the new RDAs and differences between the new and old RDAs are important considerations in planning school meals.
114 DIETARY REFERENCE INTAKES Thus, planning for the school nutrition programs has focused on what is offered in school meals. Since it can be assumed that the intent of the USDA programs is to protect the intakes of the target population, the following approach to planning is indicated. Multiple program objectives for school-baseci meals leaci to im- portant analytic issues in applying the group-planning framework. If the objective of the school nutrition programs were simply to provide meals that would replicate what school children would get in the absence of the programs, then application of the group- planning framework cliscusseci in Chapter 3 would not be appropri- ate. Planners would simply examine the distributions of usual nutri- ent intake at breakfast and lunch and attempt to provide school meals that would result in these same usual intake distributions. Since the school nutrition programs, however, have nutritional objectives such as safeguarding the health of the nation's children through the provision of nutritionally acloquate meals in school (as stated in the language of the federal legislation) then the group- planning framework clevelopeci in Chapter 3 is relevant and the question is how best to apply it. Actual application of the framework is difficult since school meals supply only part of chilciren's usual ciaily intake, while Dietary Reference Intakes (DRIB) are clefineci on the basis of usual ciaily intake. USDA has aciciresseci this issue in its current regulations that specify that school lunches and breakfasts must provide, on average, one-thirci and one-fourth of the RDA, respectively. However, the current practice of prorating of the RDA for meals offered does not imply that it is appropriate to prorate the DRIs for clietary planning or assessment. The DRIs are a set of clietary reference values baseci on nutrient intakes over a period of time and are not meant to be clivicleci into parts of a clay. In acicli- tion, the proportion of usual intake accounted for by breakfast and lunch varies considerably among inclivicluals. Despite these difficult conceptual issues, there are some options for applying the framework for planning school meals. The first step is to examine ciaily usual intakes of a representative group of children covered by the school nutrition programs. Table 5-3 pre- sents ciata on the usual intakes of vitamin A, vitamin C, and zinc for boys 9 to 13 years of age from the Third National Health and Nutri- tion Examination Survey and the Continuing Survey of Food Intakes by Inclivicluals (IOM, 2000b, 2001~. These ciata suggest a low preva- lence of inadequacy for the intakes of vitamin C and zinc. For vita- min A, the estimated prevalence of inacloquacy is ~ to 10 percent. Suppose planners were interested in using information on the usual intakes of school children to plan the school meals consumed
EXAMPLES OF PLANNING FOR GROUPS TABLE 5-3 Daily Usual Intake of Vitamins A and C and Zinc, Boys 9 to 13 Years of Age 115 Vitamin A Vitamin C Zinc (RAE) a (mg) b (mg) b (EAR= (EAR= (EAR= Percentile 445 ,ug RAE) 39 ma) 7.0 ma) 1 311 44.1 5.4 2 350 47.9 6.0 3 377 51.7 6.3 5 415 59.2 6.9 10 480 65.9 7.7 25 606 85.6 9.1 50 774 1 19.3 1 1.2 95 1 ,330 334.6 1 8.5 99 1,635 598.3 28.5 Approximate percent < EAR 5-10% 0% 5% Target median intake 774 + 80 a Usual intake from food only. Taken from the Continuing Survey of Food Intakes by Individuals and converted to retinal activity equivalents (RAE) using data on vitamin A and carotenoid intakes. EAR = Estimated Average Requirement. b Usual intake from food and supplements. Taken from the Third National Health and Nutrition Examination Survey and adjusted for day-to-day variation using the Iowa State University method. SOURCE: IOM (2000b, 2001). by program participants. As described in Chapter 3, determining the target usual intake distribution first involves selecting a group prevalence of inadequacy. In the case of these selected nutrients, planners are likely to conclude that the usual intakes of vitamin C and zinc are acloquate, and would therefore plan to maintain cur- rent intakes. For vitamin A, however, if the acceptable group preva- lence of inacloquacy is set at 2 to 3 percent rather than the current ~ to 10 percent, planners would aim to shift the usual intake clistri- bution by about 80 ,ug retinal activity equivalents (RAE) so only 2 to 3 percent are below the EAR, resulting in a target median intake of 854 ,ug RAE. The next step in applying the group-planning framework is to clecicle how the school nutrition programs should or could be used to achieve the targeted usual intake distribution. Two possible options are (1) to derive the target ciaily usual intake distribution
116 DIETARY REFERENCE INTAKES and prorate the target intakes across meals, or (2) to derive the target ciaily usual intake distribution, estimate the cleficit in 24-hour intakes, and plan for intakes from school meals to make up these cleficits. The first of these options is consistent with the way in which the school nutrition programs currently operate, where the amount offered in the school meals is a specified proportion of the RDAs. Implementing this option in the case of vitamin A, for example, would entail prorating the target usual intake distribution, with the target meclian intake of 854 ,ug RAE, in such a way that a certain proportion is consumed at breakfast and at lunch. The second option makes the nutritional objectives of the school nutrition programs more explicit. Implementing this option involves planning school breakfasts and lunches such that the distribution of usual ciaily intakes of participants is the target usual intake clistri- bution. In this case, the school meals are expected to make up the cleficit in usual ciaily vitamin A intake of 80 ,ug RAE. The cleficit could be macle up by planning menus that would acici 80 ,ug RAE to the meclian intake at breakfast or lunch. This amount could also be split between the two meals. Tailoring food choices or portion sizes at the point of service may be impractical. Thus, a methodology of planning for heterogeneous groups may be neecleci. In summary, application of the group-planning framework for the U.S. food and nutrition assistance programs is a complex task that involves several considerations related to program goals, nutritional considerations, and program implementation. Like any new para- cligm, it must first be tested for its feasibility and practicality. The discussion of the school nutrition programs above is intencleci to identify the main issues involved in applying the framework and options to consider in its implementation it is not intencleci to prescribe how this framework should be implemented in the con- text of school fouling. . . PLANNING DIETS FOR A HETEROGENEOUS GROUP USING A NUTRIENT DENSITY APPROACH The examples provided to this point have assumed that planning is occurring for a group that consists of a single life stage and gender group or life stage and gentler groups with similar requirements. Frequently, however, planning will occur for groups that encom- pass multiple life stage and gentler groups with very different nutri- ent and energy requirements. Two examples that incorporate the nutrient density approaches clescribeci in Chapter 4 are provicleci
EXAMPLES OF PLANNING FOR GROUPS 117 below. The first illustrates the simple nutrient density approach, in which the target meclian intake for each subgroup is compared to the average energy neecis of the subgroup. The second example illustrates the nutrient density distribution approach, which includes a consideration of the variability of energy and nutrient neecis within each subgroup. To compare and contrast the two approaches, both examples con- sicler the vitamin C intakes of a group consisting of adolescent boys age ci 14 to 18 years, women age ci 19 to 50 years, and men age ci 19 to 50 years. As in most of the examples in this chapter, ciata used here are real ciata, in this case collected in the 1994-1996 Continuing Survey of Food Intakes by Inclivicluals. Intake distributions of vita- min C and of energy for the three subgroups were acljusteci using the Iowa State University method (IOM, 2000a; Nusser et al., 1996~. The estimated usual intake distributions of energy in each of the subgroups were used as estimates for the distributions of require- ments of energy. The examples were constructed using the ciata presented in Table 5-4. Simple Nutrient Density Approach Step I. Obtain the target median vitamin C intake for adolescent boys, adu;tt women, and adu;tt men. Adolescent Boys. The estimated prevalence of vitamin C inacloquacy in this particular subgroup of adolescent boys is approximately 19 percent when comparing usual intakes to their Estimated Average Requirement (EAR) of 63 mg/ciay. Thus, a target vitamin C intake distribution would be obtained by shifting the baseline usual intake distribution by an amount sufficient to move the 3rd percentile of the distribution from its current 31 mg to approximately 63 mg (assuming that a prevalence of inacloquacy of 2 to 3 percent is what is desired). By shifting the intakes of vitamin C by 32 mg/day (EAR- 3rci percentile: 63 - 31 = 32), the target vitamin C intake distribution is obtained (as was clescribeci in Chapter 3~. In this target vitamin C intake distribution, the 3rci percentile is now approximately at the EAR of 63 mg/ciay. The target meclian intake is now 139 mg/ciay. Adult Women. The prevalence of inacloquacy among the women in this example is approximately 33 percent compared to their EAR of 60 ma. To obtain the target vitamin C intake distribution, it is necessary to shift the distribution by approximately 37 mg/ciay (EAR- 3rd percentile: 60 - 23 = 37), so that the proportion of
118 DIETARY REFERENCE INTAKES TABLE 5-4 Usual Vitamin C and Energy Intakes of a Group Containing Three Discrete Subgroups Percenti Subgroup EARa n Median Mean SD b 3rd Usual Vitamin C Intake (mg/day) Boys 14-18 y 63 474 107 70 31 Women 19-50 y 60 2,498 77 48 23 Men 19-50 y 75 2,726 95 67 26 Usual Energy Intake (kcal/d ay) Boys 14-18 y 2,801 2,881 782 Women 19-50 y 1,685 1,719 430 Men 19-50 y 2,561 2,659 809 a EAR = Estimated Average Requirement. b SD = standard deviation. SOURCE: USDA/ARS (1997). target usual intakes below the EAR of 60 mg/ciay is about 3 percent. The target meclian intake is now 1 14 mg/ciay. Adult Men. The prevalence of inacloquacy among the men in this example is approximately 35 percent baseci on their EAR of 75 ma. To obtain the target vitamin C intake distribution, it is necessary to shift the distribution by approximately 49 mg/day (EAR- 3rd per- centile: 75 - 26 = 49), so that the proportion of target usual intakes below the EAR of 75 mg/ciay is now about 3 percent. The target meclian intake is now 144 mg/ciay. Step 2. Divide the target median vitamin C intake by the mean energy intake or expenditure in each subgroup to obtain the target median nutrient intake re;tative to energy. In this step, the median of the target usual intake distribution of the nutrient (vitamin C), which has been clevelopeci to exceed the requirements of most members of the group, is clivicleci by the mean energy intake. The mean energy intake, rather than the meclian, is used because for energy, assuming the group (or subgroup) is in energy balance, the mean energy intake is equal to the mean energy requirement, and there are negative effects to providing energy above or below the requirement.
EXAMPLES OF PLANNING FOR GROUPS 'OUp 119 Percentile SDb 3rd 5th 95th Prevalence of Inadequacy (%) 70 31 38 256 19 48 23 2 8 1 78 33 67 26 3 1 238 35 782 1,747 4,288 430 1,071 2,248 809 1,537 4,112 Adolescent Boys. The target meclian vitamin C intake for adolescent boys in this example is 139 mg/day. With a mean energy intake of 2,881 kcal/ciay, this leacis to a target meclian vitamin C intake of 48.2 ma/ 1,000 kcal. Adult Women. The target meclian vitamin C intake for adult women of 114 mg/ciay is clivicleci by their mean energy intake of 1,719 kcal/ clay, for a target meclian intake of 66.3 mg/1,000 kcal. Adult Men. The target meclian vitamin C intake for adult men of 144 mg/day is divided by their mean energy intake of 2,659 kcal, for a target median intake of 54.2 mg/1,000 kcal. Step 3. Compare the target median nutrient intakes relative to energy for each discrete subgroup to identify the subgroup with the reference intake (i.e., the highest nutrient requirement relative to energy intake) and set planning goals for the whole group. Ensure that intakes of the other subgroups wiR not be above the Tolerable Upper Intake [ever (UT). Among these three groups, women have the highest target median vitamin C intake relative to their mean energy intake. Thus, the target reference intake for planning purposes would be 66.3 ma/ 1,000 kcal.
120 DIETARY REFERENCE INTAKES Whether the target reference intake would leaci to intakes above the UL cannot be accurately cletermineci using the simple density approach. However, an indication of the likelihood of excessive intakes can be obtained by calculating the anticipated intake at the 95th percentiles of the energy intake distribution, using the refer- ence density. For adolescent boys, the 95th percentile of energy intake is 4,288 kcal/ciay, which would be associated with a vitamin C intake of 284 mg/day (4,288 kcal x 66.3 mg/1,000 kcal). This intake remains considerably below the UL of 1,800 mg/ciay for acloles- cents. Similarly, for adult men the 95th percentile of energy intake is 4,112 kcal/ciay, which would be associated with a vitamin C intake of 273 mg/ciay using the reference density. This too is well below the UL of 2,000 mg/day for adult men. Step 4. Assess whether the plan was successfully implemented. Ideally, after the plan has been implemented, assessment of in- takes would be conclucteci to confirm whether the acceptable preva- lence of inacloquacy has been attained and whether the prevalence of intakes above the UL is low. Nutrient Density Distribution Approach Step I. Obtain the target usual vitamin C intake distribution. The first step in the nutrient density distribution approach is sim- ilar to the first step in the simple nutrient density approach. How- ever, instead of focusing on one point of the target usual intake distribution (the meclian), in this case the entire distribution is of interest. Adolescent Boys. As clescribeci in the simple nutrient density approach, the target usual vitamin C intake distribution for adolescent boys would be shifted up by 32 mg/ciay. This would leaci to a distribution with a median intake of 139 mg/day, and Sth and 95th percentiles of 70 and 288 mg/ciay, respectively. Adult Women. For adult women, the usual vitamin C intake clistri- bution would be repositioned by 37 mg/ciay to obtain the target intake distribution. It would have a median of 114 mg/day and Sth and 95th percentiles of 65 and 215 mg/day, respectively. Adult Men. The usual intake distribution for adult men would be shifted up by 49 mg/day to obtain a target intake distribution with a
EXAMPLES OF PLANNING FOR GROUPS 121 meclian of 144 mg/ciay, and Sth and 95th percentiles of 80 and 287 mg/day, respectively. Step 2. Define the target usual vitamin C density intake distribution for each definable subgroup. Given a target nutrient intake distribution and a usual energy intake distribution. it is now possible to derive the target nutrient density intake distribution tor each subgroup. this is clone by using one of the two equations presented in Chapter 4 to compute the average nutrient density intake for each incliviclual in each subgroup (or for a sample of inclivicluals in each subgroup). The average nutrient density intake for each incliviclual is then combined to form the target nutrient density intake distribution for each subgroup. , 1 In this example, an average (over a number of possible energy intake values) vitamin C density intake was computed for a random sample of 400 inclivicluals from each of the subgroups (boys, women, men). For each incliviclual in each subgroup sample, a random sam- ple of 400 energy intakes was drawn from the usual energy intake distribution for that subgroup. The target vitamin C density intake was constructed using equation (2) from Chapter 4: Average nutrient density intake = (l/m) Im= ~ (usual nutrient intake/energy intakes.) x 1,000 Equation (2) was used rather than equation (1) because the cal- culation was performed on a random sample of each subgroup (Monte Carlo approach) rather than the entire distribution of all possible nutrient and energy intake combinations. This procedure was accomplished as follows: · A random sample of 400 intakes was drawn from the target usu- al vitamin C intake distribution for each subgroup. · Next, for each of those 400 vitamin C intakes in each subgroup, a random sample of 400 energy intakes was drawn from the usual energy intake distribution in the corresponding subgroup. Thus, a given vitamin C intake (e.g., 46 ma) was associated with 400 dif- ferent energy intakes (e.g., 46 mg/1,750 kcal, 46 mg/3,002 kcal, 46 mg/2,222 kcal, and so on). From those 400 different densities for each nutrient intake, the average nutrient density intake was calculated using the second equation (nutrient density intake = ~ 1/ m] Imj = ~ tusual nutrient intake/energy intakej] x 1,000) where m is equal to 400.
122 DIETARY REFERENCE INTAKES · This process was repeated a total of 400 times in each subgroup (for each of the 400 vitamin C intakes in each subgroup). · Then, for each subgroup, the 400 average nutrient density intakes were used to construct the target vitamin C density intake distribution. Adolescent Boys. In the case of boys age ci 14 to 18 years, the target nutrient density intake distribution has a meclian of 52 mg of vita- min C/l,OOO kcal, and Sth and 95th percentiles of 26 and 112 ma/ 1,000 kcal, respectively. Adult Women. In this example, the target vitamin C density intake distribution for women age ci 19 to 50 years has a meclian of 71 ma/ 1,000 kcal, a Sth percentile of 42 mg/1,000 kcal, and a 95th percen- tile of 135 mg/1,000 kcal. Adult Men. For the subgroup of men age ci 19 to 50 years, the resulting target vitamin C density intake distribution has a meclian of 57 mg/1,000 kcal, and Sth and 95th percentiles of 33 and 115 mg / 1,000 kcal, respectively. Step 3. Compare the target median vitamin C density for each dis- crete subgroup to set planning goals for the group as a whole. In this example, the target vitamin C density distribution for women haci the highest meclian (71 mg/1,000 kcal compared to 57 ma/ 1,000 kcal for adult men and 52 mg/1,000 kcal for adolescent boys). This amount would normally be chosen as the reference nutrient density intake distribution for the group as a whole, and intakes would be planned on this basis. The planned menus resulting from this activity should be checkoci for both total milligrams of vitamin C and milligrams of vitamin C/l,OOO kcal. Comparison of the Simple Nutrient Density Approach and the Nutrient Density Distribution Approach It is useful to compare the planning results that would be achieved when using the two nutrient density methods clescribeci above (anci in Chapter 4~. Recall that for the same group of boys, women, and men, the meclian of the target nutrient density intake distribution that would be obtained by simply dividing the target median vita- min C intake by the mean energy requirement in each of the groups was 48, 66, and 54 mg/1,000 kcal, respectively. Baseci on these values, the planner would aim for a target nutrient density intake distribution in each of the subgroups with a median equal to the
EXAMPLES OF PLANNING FOR GROUPS 123 highest of the three values, or 66 mg/1,000 kcal. Using this method, which floes not take into account the distribution of energy require- ments in the group, results in a prevalence of vitamin C inacloquacy of approximately 8 to 9 percent for the women in the group (for adolescent boys and men the resulting intakes would be acloquate for all inclivicluals). In contrast, using the nutrient density clistribu- tion approach results in a projected prevalence of inacloquacy of approximately 2 to 3 percent for the women, and essentially zero for the men and adolescent boys. Because the nutrient density clis- tribution approach accounts for variability in energy intakes, it is more likely to achieve planning goals. INTERVENTIONS THAT MAY CHANGE THE SHAPE OF THE INTAKE DISTRIBUTION: NUTRIENT SUPPLEMENTATION Some planning applications involve interventions that aim to mollify food or nutrient intakes. One way to mollify nutrient intakes when a fooci-baseci approach is not possible is to incorporate use of a nutrient supplement within a group. If every incliviclual in the group consumed the identical supplement every clay, the clistribu- tion of usual intakes would simply shift up, with no change in shape, by the close of the supplement. In practice, however, all inclivicluals in a group may not take the supplement on a regular basis, and, among those who do take it, the close may not be constant. As a result, misleacling conclusions and practices may result if uniform supplement usage is assumed. As an example, suppose a planner wished to recluce the preclicteci prevalence of zinc inacloquacy among a group of free-living teenage girls through the use of a supplement. The first step would be to examine the current intake distribution. Let us assume that the group of teenage girls being targeted is similar to the sample of girls age ci 14 to 18 years surveyoci by the Third National Health and Nutrition Examination Survey (NHANES III), so that ciata from NHANES III can be used to estimate the current intake clistribu- tion. Participants in NHANES III are free-living and have not been the target of any national public health intervention regarding the use of zinc supplements. Table 5-5 presents information on the clis- tribution of usual intake of zinc from foocis (acljusteci for within- person variation) and from supplements. The EAR for zinc in girls aged 14 to 18 years has been set at 7.3 mg/ciay. As shown in Table 5-5, more than 25 percent of teen girls haci inacloquate usual intake of zinc from food alone. If the acceptable group risk of inaci-
124 DIETARY REFERENCE INTAKES TABLE 5-5 Estimated Usual Zinc Intake Distribution for Girls, 14 to 18 Years of Age (mg/day) Percentile of Zinc from Usual Intake Foods Zinc from Supplements Total Zinca 1 4.0 0.83 3.9 3 4.7 0.9 4.8 5 5.1 1.0 5.2 10 5.8 1.0 5.8 25 7.1 2.5 7.2 50 8.8 8.0 9.0 75 10.9 15.0 11.6 90 13.2 15.0 13.8 95 16.4 37.5 16.0 99 18.6 45.5 26.6 Sample size 949 48 949 Mean 9.27 9.75 9.82 a Because only 48 of the 949 girls used supplements containing zinc, total zinc intake does not equal the sum of the zinc intakes from food and supplements. SOURCE: IOM (2001). equacy were set at 3 percent, then the 3rci percentile of usual intake should be increased to the level of the Estimated Average Require- ment (EAR). That is, the 3rci percentile value of 4.7 in Table 5-5 should increase to 7.3, an increase of 2.6 ma. Assuming that the usual intake distribution floes not change its shape, the meclian intake would be the existing meclian intake + 2.6 mg (8.8 mg + 2.6 mg = 11.4 mg). This new usual intake distribution could be achieved if everyone took a supplement containing 2.6 mg of zinc. Before recommencling consumption of a supplement containing 2.6 mg of zinc, however, it is important to determine current sup- plement use. Accordingly, the next step is to examine the reported use of zinc supplements and the computed distribution of intakes from both sources, which are shown in Table 5-~. Note that only 48 of the 949 teen girls in the survey reported taking a zinc supple- ment (approximately ~ percent), so including supplements floes not affect the total intake for most participants. Indeed, the distri- bution of total zinc intake differs primarily in the upper percentiles, with very little change in the lower percentiles. The third percentile increases only 0.1 mg/day, from 4.7 to 4.8 mg/day. Thus, there is almost no effect of current use of zinc supplements on the preclicteci prevalence of inacloquacy. The increase that is neecleci to recluce
EXAMPLES OF PLANNING FOR GROUPS 125 the prevalence to 3 percent is now 2.5 mg/ciay (7.3 - 4.8) versus 2.6 mg/ciay when food alone is consiclereci. In theory, planners could develop an education intervention that recommencleci that teen girls consume a supplement that provides 2.5 mg of zinc/day. Special supplements providing this level of intake could even be marketed. However, several observations regarding supplement usage patterns in free-living populations are important to highlight: · Although the average supplement provicleci 9.75 mg of zinc, the change in the meclian intake of zinc, when acicling in supplement use, was only 0.2 mg (9.0 mg- 8.8 mg). · Although the meclian intake of zinc increased by 0.2 mg when supplements were inclucleci, the magnitude of the change at the 3rci percentile was only 0.1 ma. · The prevalence of inacloquate intake of zinc still exceeds 25 percent, even when intake from currently consumed supplements is acicleci to the intake from food. · As is usually the case, supplement usage was not uniform across this group of inclivicluals. Teen girls with higher intakes of zinc from food were more likely to take a supplement and perhaps more likely to take a higher-close supplement. Thus, supplement use by a free-living population may not achieve the planner's goals, and the challenge is to determine how to either shift the whole distribution by 2.5 mg/ciay or to increase the use of supplements or zinc-rich foods by individuals in the lower percen- tiles. If an aciclitional supplement of 2.5 mg/ciay of zinc was clistrib- uteci and consumed by the entire population, then the distribution would shift as clesireci. As the ciata in Table 5-5 illustrate, it may take an intensive intervention to achieve this goal. An alternative approach is to ensure supplement use by those in the lower percentiles. This might be possible if there are character- istics that would identify inclivicluals with low intakes (such as income level or age). Such interventions to increase supplement use are likely to be more successful in a confined population (where supplement use could be monitored than in a free-living one. The important conclusion from this example of planning is that an intervention to change usual intakes through supplementation can be difficult to design and implement. In a free-living popula- tion, not every person can be expected to consistently take a supple- ment (or a given food or food group rich in a specific nutrient), and interventions in such a group may be expected to change both
126 DIETARY REFERENCE INTAKES the location and shape of the usual intake distribution. It is impor- tant to unclerstanci the patterns and predictors of supplement use in order to model and plan such interventions. Simply assuming uniform use of a supplement in free-living populations would likely result in a failure to achieve the planning goals. FOOD FORTIFICATION Fortification is often seen as a potentially desirable public health measure that could achieve an increased intake of specified nutri- ents without changes in food consumption practices or compliance with specific nutrient supplement usage. Historically, mandatory fortification programs have been applied in many countries as a means to aciciress particular public health concerns. In these pro- grams, public health authorities determine both the food vehicles and levels of fortification, and only fortified versions of the selected foocis are permitted on the market. One such example is the man- ciatory fortification of table salt with iodine in Canada, a measure undertaken to recluce iodine deficiency in the population. Alterna- tively, food fortification programs may be voluntary, with food man- ufacturers having the option of Hilling particular nutrients (some- times within prescribed limits) to foocis, but not being required to do so. One example of this approach is the fortification of orange juice with calcium; because the program is voluntary, it is possible to purchase orange juice with or without calcium acicleci. Regula- tions on food fortification differ between Canada and the United States, with voluntary fortification permitted in the United States. Regardless of whether fortification is mandatory or voluntary, if it is intencleci to achieve public health goals, then it is often necessary to "target" the fortification. Such targeting could be accomplished by selecting only foocis for fortification that are used exclusively or in substantially greater amounts by the group targeted by a fortifica- tion program, or by mounting an educational program to promote the use of specific fortified foocis by the target group. Fortification, however, also carries the potential for detrimental effects. Fortification of foocis might increase nutrient intakes to excessive levels among those persons who have high intakes of the fortified food or those who already have high intakes of the nutri- ent and then consume the newly fortified food. Minimally con- trolleci fortification of foocis, even at low levels in incliviclual foocis, can have unexpected effects, ranging from negligible benefits to public health concerns about potentially detrimental high intakes. Further, unless fortified foods reach only the target group (unusual
EXAMPLES OF PLANNING FOR GROUPS 127 in practice, except for infant foocis), it is possible that the risk of detrimental effects will appear in other sectors of the population (i.e., nontarget groups). Because of the range of potential effects that can accompany fortification programs, both beneficial and detrimental, the potential impact of proposed fortification is usually examined before implementation. In general, no simple method can be used to predict the effects of fortification. Fortifying foocis with nutrients will have impacts on the nutrient intakes of those who consume the fortified foocis and will not have impacts on those who do not consume them. Further, the degree of impact clepencis not only on the level of the nutrient acicleci, but also on the distribution of usual intakes of the food. In recent years, predicting the effect of fortification has been compli- cateci in the United States by introduction of food products forti- fieci with a nutrient while the evaluation of the neeci for fortifica- tion is still in progress. Thus, it is difficult to anticipate changes in the usual intake distribution of the nutrient when even changes in the amount of the nutrient in the food supply are almost impossible to predict. A more extencleci discussion on the issue of voluntary fortification is presented in Appendix D. The approach presented below involves mocleling and estimating the effects of a mock fortification effort by using ciata on foocis and nutrients consumed and then calculating the change in nutrient intake after the foocis are fortified. The preclicteci benefits and risks associated with the fortification can be assessed through application of assessment methods baseci on the Estimated Average Require- ment (EAR) and Tolerable Upper Intake Level (UL) (IOM, 2000a). Such an approach was utilized by Lewis and colleagues (1999) to examine the impact of folate fortification of cereal-grain products in the United States if increased fortification of foocis was manciateci. A similar approach is illustrated below for the hypothetical aciclition of vitamin A to fluici milk. For simplicity, this example assumes that only one food will be fortified with vitamin A. As was cliscusseci earlier, this assumption is unlikely to hold when voluntary fortifica- tion of foocis with vitamin A is permitted. Addition of Vitamin A to Fluid Milk Two levels of requirements for vitamin A have been established with different functional endpoints in mind (IOM, 2001~. For adult women, the EAR for prevention of functional deficiency of vita- min A is 300 ,ug retinal activity equivalents (RAE)/ciay while the EAR to establish and maintain desirable levels of liver vitamin A
128 DIETARY REFERENCE INTAKES stores has been set at 500 ,ug RAE/ciay. For adult women 19 to 50 years of age, examination of the 1994-1996 CSFII (USDA/ARS, 1997) ciata suggests that about 15 percent have intakes below 300 ,ug RAE/ciay and hence have intakes apparently inacloquate to meet their own functional requirements. The same ciata suggest that about 44 percent may have intakes inacloquate to provide minimal stores of vitamin A. These descriptors of a potential problem may moti- vate planning interventions to raise vitamin A intakes in this target group, although planners would also obtain other types of ciata (e.g., biochemical or clinical outcome information such as incidence of night blinciness) before proceeding with an intervention. Suppose that in order to increase vitamin A intake by adult women, a fortification program is consiclereci that acicis vitamin A to all fluici milk. In the United States milk is frequently fortified with vitamin A, but it is not required. This example assumes that no fortification is currently taking place. Baseci on ciata from the CSFII (USDA/ARS, 1997), Table 5-6 illus- trates the preclicteci impact of this fortification on the distribution of total vitamin A intake of adult women. Total intake equals reported TABLE 5-6 Impact of the Aciclition of Vitamin A to Milk on the Expected Distribution of Total Vitamin A Intake in Women 19-50 Years of Age Level of Addition of Vitamin A (as Retinyl Ester) to Fluid Milk (Pg./ 100 ml) Percentile of Intake 0 50 100 150 200 250 300 1 135 138 140 143 145 147 149 5 225 238 247 253 259 268 276 10 272 287 298 308 319 327 337 25 368 398 421 445 465 484 505 50 542 592 635 670 711 747 787 75 785 872 964 1,083 1,151 1,245 1,333 90 1,150 1,259 1,389 1,549 1,679 1,811 1,954 95 1,390 1,560 1,715 1,915 2,084 2,234 2,411 99 2,026 2,154 2,372 2,573 2,777 3,067 3,325 NOTE: n = 2,325 women. In this example, the amount by which vitamin A increases reflects the initial fluid milk consumption of those in the various percentile groups. For example, those in the 1st percentile drink little milk, so their vitamin A intake increases only slightly as the level of addition of vitamin A to milk increases. In contrast, those in the 99th percentile, who drink much more milk, have a much greater increase. SOURCE: USDA/ARS (1997) as reported in IOM (2001).
EXAMPLES OF PLANNING FOR GROUPS 129 intake of vitamin A plus the increase that would come from con- suming fortified milk. It is possible to determine the theoretical increase because the CSFII database can be clisaggregateci to cleter- mine the amount of milk consumed by each incliviclual. Thus, the amount of the increase in vitamin A intake will reflect the amount of milk consumed: those women who consume large amounts of fluici milk will increase their intake substantially, while those who consume little or no fluici milk will not increase their intake. Table 5-7 provides some information on the likely benefits and potential risks of this fortification. Baseci on the results for adult women, Billing vitamin A to fluici milk could be expected to have beneficial impacts by raising intakes without a major concern about possible detrimental effects. That is, as the level of fortification increases, the prevalence of usual intake of vitamin A less than the EAR to prevent night blindness (300 ,ug RAE) declines from approximately 15 percent at no fortification to approximately 7 per- cent at a fortification level of 300 ,ug of retinol/100 mL of milk. The prevalence of usual intake less than the EAR for maintaining stores (500 ,ug RAE) declines from 44 percent at no fortification to 24 TABLE 5-7 Apparent Benefits and Potential Risks Associated with the Aciclition of Vitamin A to all Fluid Milk as a Function of Level of Aciclition, Women 19-50 Years of Age Prevalence of Inadequate Intakesb (below the EAR) Prevalence of Potentially Excessive IntakesC Level of Additiona % < EAR % < EAR % > UL (Pg/100 ml) (300 ,ug RAE) (500 fig RAE) (3,000 ,ug) 0 (baseline) 14.6 44.3 0.0 50 12.1 38.9 0.0 100 10.2 35.6 0.1 150 8.8 33.3 0.1 200 8.0 29.9 0.2 250 7.6 28.8 0.3 300 6.9 24.3 0.7 NOTE: n = 2,325 women. a Added as a retinyl ester. b Based on total vitamin A intake as ,ug of retinal activity equivalents (RAE). EAR Estimated Average Requirement. c Based on preformed vitamin A only. UL = Tolerable Upper Intake Level. SOURCE: USDA/ARS (1997). =
130 DIETARY REFERENCE INTAKES percent at a fortification level of 300 ,ug of retinol/100 ml of milk. In contrast, as the level of fortification increases, the prevalence of usual intake above the UL increases only slightly from 0 to 0.7 per- cent. On the basis of this evidence only, the decision to fortify milk with vitamin A would seem a worthwhile endeavor. Other subgroups, however, may not have the same benefits or risks at that level of vitamin A fortification. Table 5-8 shows the impact of this fortification of fluici milk for boys 9 to 13 years of age. In this case, the prevalence of inacloquate vitamin A intake without fortification (at baseline) is lower than for adult women. With forti- fication, the prevalence of inacloquate intakes baseci on maintain- ing stores (EAR = 445 ,ug RAE for this age group) cleclines from about 11 percent to 3.5 percent. Since there is very little prevalence of inacloquate intake of vitamin A baseci on preventing night blinci- ness (EAR = 230 ,ug RAE for this age group) without fortification, the aciclition of more vitamin A to milk would have a negligible effect on prevalence of this criterion of inacloquate intake. On the other hanci, the potential detrimental effect with fortification is TABLE 5-8 Apparent Benefits and Potential Risks Associated with the Aciclition of Vitamin A to all Fluid Milk as a Function of Level of Aciclition, Boys 9-13 Years of Age Prevalence of Inadequate Intakesb (below the EAR) Prevalence of Potentially Excessive IntakesC Level of Additiona % < EAR % < EAR % > UL ,ug/100 ml (230 ,ug RAE) (445 ,ug RAE) (1,700 ,ug) 0 (baseline) 0.5 11.1 0.9 50 0.3 8.2 2.6 100 0~3 7.0 5~9 150 0.3 5.6 12.2 200 0.3 4.5 19.0 250 0.3 4.2 30.0 300 0.3 3.5 37.8 NOTE: n= 574 boys. a Added as a retinyl ester. b Based on total vitamin A intake as ,ug of retinal activity equivalents (RAE). EAR = Estimated Average Requirement. c Based on preformed vitamin A only. UL = Tolerable Upper Intake Level. SOURCE: USDA/ARS (1997).
EXAMPLES OF PLANNING FOR GROUPS 131 high, as shown by increasing percentages with usual intake above the UL as the level of fortification increases. Specifically, with no fortification, the prevalence of usual intakes above the UL for this age group is approximately 1 percent, while at a fortification level of 300 ,ug of retinol/100 mL of milk, the prevalence of usual intakes above the UL would increase to 38 percent. The reason for these differential impacts for adult women and boys 9 to 13 years of age is that the latter group has a higher initial intake of vitamin A, and an overall higher consumption of the vehicle chosen for fortification- milk. By combining the analyses for adult women and boys 9 to 13 years of age, the relationship between the potential benefits to women and the potential risks to adolescent boys of fortifying milk at the various levels is clemonstrateci. Figure 5-1 summarizes the benefits to adult women by the declining percentage with inacloquate intake and the increasing potential risk to boys 9 to 13 years of age by the increasing percentage over the UL. Baseci on these results, planners would have to consider the preclicteci potential risk to boys 9 to 13 years of age and the preclicteci benefits to the target group of adult women before reaching a decision on whether to fortify and at what 50 - 40 - 30- ct - ct ~ 20- Cal 10- O- Benefit to women: preserve liver stores - Benefit to women: preserve function - ma_ _ ~ l by__ .~ - Risk to boys of excess - - - - ~_ _ 0 50 100 150 200 250 300 Level of Addition (,ug/100 ml) - 50 - 40 o -30 ~ s - 20 ,c Cal - - 10 - o FIGURE 5-1 Projected benefits and potential risk associated with the addition of vitamin A to fluid milk. UL = Tolerable Upper Intake Level.
132 DIETARY REFERENCE INTAKES amount. Of course, this exercise should be repeated for other sub- groups of the population before final decisions are macle. When only a few foocis are involved in compulsory fortification, regulatory agencies run mock fortification studies (like the vitamin A example above) and weigh the expected benefits and potential risks associated with different levels of fortification. However, with voluntary fortification such as what is currently the practice in the United States, as the number of fortified foocis increases, it becomes extremely difficult to run meaningful mock fortification scenarios. In aciclition, it has not been possible to keep food composition ciata- bases current with regard to branci-specific fortified foocis, and not all nutrient composition databases in the United States are clesigneci to do so. Food composition databases in the United States used in national surveys usually reflect the average composition of foocis that are available in the market, with varieties or brancis weighted by general market share. Thus, it is difficult to investigate the effect of voluntary fortification of specific brancis of foocis unless all brancis within a category are fortified. More cletaileci survey ciata, as well as more specific food composition tables, are neecleci for investigation of branci-specific fortification. Planning Fortification: General Conclusion and Recommendation The principal conclusion drawn from this fortification applica- tion is the importance of examining the potential impacts on all groups notjust on the targeted subgroups that have a higher than clesireci prevalence of inacloquate intakes without fortification. It is recommended that a modeling approach, such as that presented here, be conclucteci prior to any major introduction of fortification.