Introduction to IMAPP (and other topics) Alicia Carriquiry, David Osthus Outline Estimation of usual nutrient intake distributions The DRI and prevalence of inadequate intakes Brief introduction to PC-SIDE Planning intakes for groups IMAPP capabilities (Alicia) IMAPP demonstration (David) 2 Alicia Carriquiry - Iowa State University 11/18/2012 Daily intakes and usual intakes In many surveys and other studies, we use 24-hour recalls to measure daily nutrient intake for individuals. The daily intake of a nutrient varies from day-to-day within an individual. From a public health perspective, we are interested in usual intakes and not on daily intakes. Usual intake is the long-run average intake for a person. The usual intake varies from person to person in a group. 3 Alicia Carriquiry - Iowa State University 11/18/2012 Daily intakes and usual intakes (cont) We can express daily intake as a sum of usual intake plus a deviation from usual intake: Daily intake = Usual Intake + Error For example, a person may have a usual intake of folate equal to 300 DFE per day, but three daily measurements for that same person may be 100 DFE, 450 DFE and 245 DFE. The average of the three days is 265 DFE. The SD of the three daily intakes is 219 DFE - this reflects the within-person variability in folate intake. 4 Alicia Carriquiry - Iowa State University 11/18/2012 Daily intakes and usual intakes (cont) If we collect daily intakes for a group of persons, the variance of observed intakes will have two components: Variation in daily intakes = Within-person + Between-person variability To estimate usual intake for a single person, we need to average many daily intakes. Example: We observe daily intakes for two persons over 100 days. Mean intakes are 64 and 40 units. Within-person SDs are 15 and 6, so variances are 225 and 36, with average within-person variance of 130.5. 5 Alicia Carriquiry - Iowa State University 11/18/2012 Example: two persons, 100 days 6 Alicia Carriquiry - Iowa State University 11/18/2012 Relative size of SD Age group Females 14-18y Nutrient Ratio 3.54 3.13 1.13 Prog (g/k) 0.46 0.60 0.66 357.93 346.78 1.03 Vit B12 5.39 4.15 1.39 Prog (g/k) 0.42 0.53 0.79 681.32 514.06 1.33 Vit A 7 Between SD Vit B12 Vit A Females 19-30y Within SD Alicia Carriquiry - Iowa State University 11/18/2012 Distribution of usual intakes At the group level, however, we can estimate the distribution of usual intakes with one or two days of observation for each person. What data do we need? Daily intakes for a sample of persons. At least for some, we need a second observation. Second observation should be (hopefully) taken on a nonconsecutive day. There are several methods to estimate usual intake distributions. Oldest is NRC (1986). We use the ISU Method (Nusser et al. 1996; IOM, 2000) today. 8 Alicia Carriquiry - Iowa State University 11/18/2012 Removing the within-person variance The within-person variability in intakes is a nuisance (molesta!). To estimate important quantities such as the proportion of persons with inadequate intakes or with excessive intakes, we need to use the usual intakes. If we estimate usual intake distributions using a single day of intake or the mean of a few days of intake, we overestimate the prevalence of inadequacy (and the prevalence of excess). 9 Alicia Carriquiry - Iowa State University 11/18/2012 Children 4-8 NHANES 2005-2006 Mean 50th 75th 90th 95th One day 102 82 136 208 248 Adjusted 102 98 122 148 165 One day 9 8 11 14.6 16.9 Adjusted 8.9 8.7 10.2 11.7 12.8 Vit C Zinc 10 Alicia Carriquiry - Iowa State University 11/18/2012 Vitamin B6 – Women 31-50 11 Alicia Carriquiry - Iowa State University 11/18/2012 Serum 25(OH)D NHANES 2001-2006 <30 nmol/L 30-39 nmol/L 40-49 nmol/L 50-75 nmol/L >75 nmol/L Unadjusted 8.2 9.3 14.4 47.2 20.8 Adjusted 4.9 9.8 17.1 48.8 19.4 12 Alicia Carriquiry - Iowa State University 11/18/2012 Example dataset ID Number Day Sex Vitamin C Iron 126701 1 F 98.5 14.1 126701 2 F 56.1 0.5 145665 1 F 76.3 8.9 133289 1 M 124.6 7.0 133289 2 M 101.2 19.9 134678 1 M 32.7 24.0 156700 1 F 65.2 6.1 13 Alicia Carriquiry - Iowa State University 11/18/2012 The ISU Method and PC-SIDE A statistical approach to estimate usual intake distributions when we have daily intakes from a group of persons. We need a second day on at least a sub-sample. Software: PC-SIDE. PC-SIDE is distributed free of charge and can be found in Documentation can also be found on that site. 14 Alicia Carriquiry - Iowa State University 11/18/2012 What does PC-SIDE do? PC-SIDE removes the nuisance within-person variance from daily intakes. It also adjusts daily intake for the effects of: Day of week Season Any other factor that affects nutrient intakes. It produces estimates of usual intake distributions. Estimates prevalence of nutrient inadequate intakes or of excessive intakes. In complex surveys, it uses the survey weights to produce unbiased estimates of means, variances, percentiles…. 15 Alicia Carriquiry - Iowa State University 11/18/2012 The DRIs to assess intakes The DRIs (Dietary Reference Intakes) defined by the IOM and Health Canada refer to nutrient requirements. There are four DRIs: 16 EAR: Estimated Average Requirement is the typical requirement of a nutrient of a person in an age-sex group. The EAR meets the needs of 50% of individuals in the group. RDA: Required Daily Allowance is a level of usual intake that exceeds the requirements of over 97% of persons in the group. UL: Upper Tolerable Level is the highest level of usual intake that is likely to be safe for almost everyone in the group. AL: Adequate Intake is established when there is not enough information about requirements to set an EAR. Alicia Carriquiry - Iowa State University 11/18/2012 17 Alicia Carriquiry - Iowa State University 11/18/2012 DRIs to assess intakes of groups The only two DRIs that are useful for groups are the EAR and the UL. The RDA should NEVER be used to assess group intakes. Beaton (1994) and Carriquiry (1999) proposed a simple method to estimate prevalence of inadequate intakes called the EAR cut-point method. The proportion of persons in a group with intakes below their requirements is approximately equal to the proportion of persons with intakes below the EAR for the nutrient. 18 Alicia Carriquiry - Iowa State University 11/18/2012 19 Alicia Carriquiry - Iowa State University 11/18/2012 20 Alicia Carriquiry - Iowa State University 11/18/2012 21 Alicia Carriquiry - Iowa State University 11/18/2012 Planning interventions From survey or other representative data we get a baseline estimate of the nutritional status of our population. We may find that the prevalence of inadequate intakes in some sub-populations is unacceptably high. Food fortification with micronutrients can improve health outcomes in populations. Guidelines on Food Fortification with Micronutrients was published in 2006 by WHO and FAO. 22 Alicia Carriquiry - Iowa State University 11/18/2012 The idea behind planning 23 Alicia Carriquiry - Iowa State University 11/18/2012 Planning intakes (cont’d) 1. 2. 3. 4. 5. 24 Establish a target prevalence of inadequacy. Decide on a set of possible food vehicles. Define the “dose” or the amount of the nutrient to be added per 100 g of the vehicle. Simulate the intakes that may be observed after fortification assuming that consumption patterns do not change due to cost, taste, etc. Evaluate whether the dose under consideration suffices to achieve the target prevalence. Alicia Carriquiry - Iowa State University 11/18/2012 Baby Example We have baseline daily intakes of vitamin A and of three possible vehicles: sugar, vegetable oil and wheat flour. The limit of fortification is: 700 mg RAE / 100 g of sugar 2500 mg RAE / 100 g of vegetable oil 240 mg RAE / 100 g of wheat flour. Retinol is added as the fortificant. 25 Alicia Carriquiry - Iowa State University 11/18/2012 Baby example (cont’d) If a person consumes 149 mg RAE of vitamin A on day 1 at baseline and also consumes 15 g of sugar, 8 g of oil and 0 g of wheat flour, the additional intake of vitamin A in RAE would be: From sugar: 700 x 15 / 100 = 105 mg RAE From oil: 2500 x 8 / 100 = 200 mg RAE From wheat flour: 240 x 0 / 100 = 0 mg RAE If we fortify the three vehicles, then the person’s intake on day 1 would be : Baseline + vit A from sugar + vit A from oil + vit A from flour: 149 + 105 + 200 + 0 = 454 mg RAE. 26 Alicia Carriquiry - Iowa State University 11/18/2012 Baby example (cont’d) For each person, we construct the “new” daily intake values. 27 Alicia Carriquiry - Iowa State University 11/18/2012 Baby Example (cont’d) With those new intake values, we re-compute the prevalence of inadequate intakes as usual. If the new prevalence is approximately equal to the target prevalence, then we are done. If the new prevalence is smaller or larger than the target, we reconsider the amount of the fortificant added to each vehicle or the mix of vehicles and try again. 28 Alicia Carriquiry - Iowa State University 11/18/2012 Planning intakes (cont’d) Additional data needs: daily intake of possible food vehicles by all persons in the sample. Given a user-chosen target prevalence, we wish to compute: 29 The gap between baseline intake levels and intake levels that would meet the target prevalence. The approximate amount of the nutrient to be added to 100 g of a vehicle to achieve the target prevalence. Alicia Carriquiry - Iowa State University 11/18/2012 Estimating the Gap The gap between baseline and target is the difference between: The EAR (the percentile at baseline proportion of usual inadequate intakes) and The quantile corresponding to the desired (or target) proportion of usual inadequate intakes. E.g. if observed prevalence is 23% and target prevalence is 10%, gap equals EAR – 10th quantile of usual intake. 30 Alicia Carriquiry - Iowa State University 11/18/2012 Fortifying vehicles Planner chooses the number of units of the nutrient that is added to 100 g of the vehicle. This is the dose. Then computes additional nutrient intake per person per day: Additional intake = g vehicle x dose / 100. Nutrient intake after fortification: Intake after fortification = Baseline intake + Additional intake 31 Alicia Carriquiry - Iowa State University 11/18/2012 Fortifying vehicles (cont’d) Initial “guess” of optimal dose: Dose ~ Gap x 100 / Average vehicle consumption (g). E.g., if gap = 3 mg and average consumption of vehicle per day is 10 g, then initial guess for dose is Dose = 3 x 100 / 10 = 30 mg per 100 g of vehicle. It is easier to guess the optimal dose when vehicle is consumed in approximately the same amounts by all persons in a group. 32 Alicia Carriquiry - Iowa State University 11/18/2012 Fortifying vehicles (cont’d) A food scientist determines what is feasible Technically, accounting for taste, manufacturing constraints, etc. Economically, from a cost-wise perspective. At present, we have the tools to consider one vehicle at a time, but ideally we wish to construct a possibilities region, composed of all combinations of vehicles and doses that meet the goals. 33 Alicia Carriquiry - Iowa State University 11/18/2012 IMAPP IMAPP (Intake Monitoring, Assessment and Planning Program) is software that can be used to Estimate usual nutrient intake distributions at baseline Plan fortifications. The program was developed by Alicia Carriquiry, (Iowa State U), Suzanne Murphy (U of Hawaii) and Lindsay Allen (UC Davis), in collaboration with Bruno de Benoit and Lisa Rogers (WHO). Funding was provided by WHO. The program implements the methods in the 2006 WHO/FAO report Guidelines on Food Fortification with Micronutrients . 34 Alicia Carriquiry - Iowa State University 11/18/2012 Uses of IMAPP Assesses group intake data in terms of prevalence of inadequate and excessive intakes. Estimates whether a given fortification strategy will be safe and effective for all population groups consuming a fortified food vehicle. Enables practitioners to choose between different fortification strategies on the basis of safety and efficacy. (By safety, we mean proportion of persons with usual intakes below the UL.) Screen 1 Screen 2 Program characteristics IMAPP’s engine is PC-SIDE. IMAPP does everything that PC-SIDE does, but also permits modeling fortification strategies. It produces estimates the prevalence of inadequate and excessive intakes before and after different levels of fortification. It is based on recommendations from WHO/FAO 2006 report Guidelines on Food Fortification with Micronutrients. Adjusts the nutrient intake distributions for day-to-day variability in intakes using the ISU method. Program characteristics (cont’d) User-friendly, and does not require expertise in statistical calculations. Supplies default values for many of the statistical and nutritional parameters, including nutrient requirements and external variance estimates. Clearly documents and explains all steps in a user’s manual. User provides: Daily dietary intake data (e.g., from 24-hr recalls or food records) for each person in the sample. Age, gender and reproductive status (pregnant, lactating or neither) for each person in the sample. Nutrient intakes as well as daily intakes of potential food vehicles for each person. User can select: Bioavailability of dietary iron and of iron used to fortify. The form of the nutrient (e.g., folate or folic acid, retinol or carotenoids). The DRIs for the nutrient (if different from default values). User provides (cont’d): Two or more days of dietary data for at least a representative subsample; if not, default values are suggested for day-to-day variance estimates. Persons’ body weights (in kg) to assess protein intake; if not, default weight values are suggested. Bioavailability factors for iron and zinc can be supplied by user; if not, default values are suggested. Nutrient standards used by IMAPP Software uses harmonized nutrient requirements obtained by compiling EARs from the US/Canada DRIs, RNIs from the FAO/WHO tables and other information from the literature. The user may choose to input own values or may use the default “harmonized” values in the program, which were developed for each gender/age/reproductive status group. Safe upper levels of intake are based largely on the ULs from the US/Canada DRIs. To conclude… Today is the official launch of IMAPP! IMAPP is free to anyone who wishes to use it. Next presentation: David will do several step-by-step demonstrations on how to use the program. The program may have some bugs – PLEASE let us know whether you encounter problems when you use it. When you download the program, we will keep your name and email address. We will be able to contact you to tell you about updates, corrections and other things that may be of interest. 43 Alicia Carriquiry - Iowa State University 11/18/2012 THANKS FOR YOUR ATTENTION alicia@iastate.edu 44 Alicia Carriquiry - Iowa State University 11/18/2012