The Food Propensity Questionnaire: Concept, Development, and Validation for Use as a Covariate in a Model to Estimate Usual Food Intake

Estimating usual food intake distributions by using the multiple source method in the EPIC-Potsdam Calibration Study

Estimating usual food intake distributions from short-term quantitative measurements is critical when occasionally or rarely eaten food groups are considered. To overcome this challenge by statistical modeling, the Multiple Source Method (MSM) was developed in 2006. The MSM provides usual food intake distributions from individual short-term estimates by combining the probability and the amount of consumption with incorporation of covariates into the modeling part. Habitual consumption frequency information may be used in 2 ways: first, to distinguish true nonconsumers from occasional nonconsumers in short-term measurements and second, as a covariate in the statistical model. The MSM is therefore able to calculate estimates for occasional nonconsumers. External information on the proportion of nonconsumers of a food can also be handled by the MSM. As a proof-of-concept, we applied the MSM to a data set from the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Calibration Study (2004) comprising 393 participants who completed two 24-h dietary recalls and one FFQ. Usual intake distributions were estimated for 38 food groups with a proportion of nonconsumers > 70% in the 24-h dietary recalls. The intake estimates derived by the MSM corresponded with the observed values such as the group mean. This study shows that the MSM is a useful and applicable statistical technique to estimate usual food intake distributions, if at least 2 repeated measurements per participant are available, even for food groups with a sizeable percentage of nonconsumers.

Feasibility of innovative dietary assessment in epidemiological studies using the approach of combining different assessment instruments

Objective

To assess the feasibility of combining short-term and long-term dietary assessment instruments as new concept for improving usual dietary intake assessment on the individual level.

Design

Feasibility study of completing three 24 h dietary recalls (24-HDR) and a self-administered food propensity questionnaire (FPQ). The 24-HDR was conducted by monthly telephone interviews, using EPIC-SOFT software. The FPQ was completely standardized across cohorts and offered either as a web-based tool or in paper format.

Setting

Random sample derived from five ongoing European cohort studies (EPIC-San Sebastian, EPIC-Florence, EPIC-Potsdam, Estonia Genome Center (EGC) and Norwegian Women and Cancer study (NOWAC)).

Subjects

A total of 400 participants.

Results

Overall, the total participation rate for the present study was 65·3 % (n 261). On average, completion of the 24-HDR was highest for the first 24-HDR (63·0 %) and decreased slightly for the second (60·3 %) and third 24-HDR (56·3 %). The proportions of selecting the web-based FPQ varied among the study centres, with the highest in EGC (92·9 %) and NOWAC (70·0 %) and the lowest in EPIC-San Sebastian (25·5 %) and EPIC-Potsdam (33·9 %). Web users rarely requested support and were younger and more highly educated than those who completed the paper format.

Conclusions

The present study supports the feasibility of a combined application of three 24-HDR and an FPQ in culturally different populations. The varying acceptance of the web-based instrument across populations requires a flexible application of assessment instruments.

Risk assessment to underpin food regulatory decisions: an example of public health nutritional epidemiology

The approach used by food regulation agencies to examine the literature and forecast the impact of possible food regulations has many similar features to the approach used in nutritional epidemiological research. We outline the Risk Analysis Framework described by FAO/WHO, in which there is formal progression from identification of the nutrient or food chemical of interest, through to describing its effect on health and then assessing whether there is a risk to the population based on dietary exposure estimates. We then discuss some important considerations for the dietary modeling component of the Framework, including several methodological issues that also exist in research nutritional epidemiology. Finally, we give several case studies that illustrate how the different methodological components are used together to inform decisions about how to manage the regulatory problem.

Fitting a bivariate measurement error model for episodically consumed dietary components

There has been great public health interest in estimating usual, i.e., long-term average, intake of episodically consumed dietary components that are not consumed daily by everyone, e.g., fish, red meat and whole grains. Short-term measurements of episodically consumed dietary components have zero-inflated skewed distributions. So-called two-part models have been developed for such data in order to correct for measurement error due to within-person variation and to estimate the distribution of usual intake of the dietary component in the univariate case. However, there is arguably much greater public health interest in the usual intake of an episodically consumed dietary component adjusted for energy (caloric) intake, e.g., ounces of whole grains per 1000 kilo-calories, which reflects usual dietary composition and adjusts for different total amounts of caloric intake. Because of this public health interest, it is important to have models to fit such data, and it is important that the model-fitting methods can be applied to all episodically consumed dietary components.We have recently developed a nonlinear mixed effects model (Kipnis, et al., 2010), and have fit it by maximum likelihood using nonlinear mixed effects programs and methodology (the SAS NLMIXED procedure). Maximum likelihood fitting of such a nonlinear mixed model is generally slow because of 3-dimensional adaptive Gaussian quadrature, and there are times when the programs either fail to converge or converge to models with a singular covariance matrix. For these reasons, we develop a Monte-Carlo (MCMC) computation of fitting this model, which allows for both frequentist and Bayesian inference. There are technical challenges to developing this solution because one of the covariance matrices in the model is patterned. Our main application is to the National Institutes of Health (NIH)-AARP Diet and Health Study, where we illustrate our methods for modeling the energy-adjusted usual intake of fish and whole grains. We demonstrate numerically that our methods lead to increased speed of computation, converge to reasonable solutions, and have the flexibility to be used in either a frequentist or a Bayesian manner.

Dietary acrylamide exposure and hemoglobin adducts--National Health and Nutrition Examination Survey (2003-04)

The objective of this study is to evaluate the relationship between dietary AA and hemoglobin adducts using the National Health and Nutrition Examination Survey (NHANES, 2003-04). Measured acrylamide (AA-Hb) and glycidamide (Gly-Hb) hemoglobin adducts for over 7000 participants >3 years, 24-h dietary recall, food frequency questionnaire (FFQ), lifestyle and demographic data, and anthropometric measurements are available from NHANES (2003-04). The 24-h dietary recall and FFQ data were combined with AA concentration data in food from the US FDA to estimate "usual" AA dietary exposure. The associations between dietary AA and AA-Hb and Gly-Hb were evaluated using linear regression models with smoking, age, gender, energy and macronutrient intake, body surface area, and activity level as covariates. Dietary AA positively correlates with AA-Hb and Gly-Hb (p<0.05) but the correlation is small (R-Squared<3.5%). Relative to the background adduct levels, the incremental increase in AA-Hb and Gly-Hb from average dietary AA is small (7% and 9% for AA-Hb and Gly-Hb, respectively). Non-dietary sources of exposure, measurement errors associated with the use of the FFQ, and uncertainty in the data on AA levels in foods are possible explanations for the observed lack of association between dietary AA and AA-Hb and Gly-Hb.

Selecting informative food items for compiling food-frequency questionnaires: comparison of procedures

The authors automated the selection of foods in a computer system that compiles and processes tailored FFQ. For the selection of food items, several methods are available. The aim of the present study was to compare food lists made by MOM2, which identifies food items with highest between-person variance in intake of the nutrients of interest without taking other items into account, with food lists made by forward regression. The name MOM2 refers to the variance, which is the second moment of the nutrient intake distribution. Food items were selected for the nutrients of interest from 2 d of recorded intake in 3524 adults aged 25-65 years. Food lists by 80 % MOM2 were compared to those by 80 % explained variance for regression on differences between the number and type of food items, and were evaluated on (1) the percentage of explained variance and (2) percentage contribution to population intake computed for the selected items on the food list. MOM2 selected the same food items for Ca, a few more for fat and vitamin C, and a few less for carbohydrates and dietary fibre than forward regression. Food lists by MOM2 based on 80 % of variance in intake covered 75-87 % of explained variance for different nutrients by regression and contributed 53-75 % to total population intake. Concluding, for developing food lists of FFQ, it appears sufficient to select food items based on the contribution to variance in nutrient intake without taking covariance into account.

Intake of added sugars and selected nutrients in the United States, National Health and Nutrition Examination Survey (NHANES) 2003-2006

In the Institute of Medicine (IOM) macronutrient report the Committee recommended a maximal intake of < or = 25% of energy from added sugars. The primary objectives of this study were to utilize National Health and Nutrition Examination Survey (NHANES) to update the reference table data on intake of added sugars from the IOM report and compute food sources of added sugars. We combined data from NHANES with the United States Department of Agriculture (USDA) MyPyramid Equivalents Database (MPED) and calculated individual added sugars intake as percent of total energy then classified individuals into 8 added sugars percent energy categories, calculated usual intake with the National Cancer Institute (NCI) method, and compared intakes to the Dietary Reference Intakes (DRIs). Nutrients at most risk for inadequacy based on the Estimated Average Requirements (EARs) were vitamins E, A, C, and magnesium. Nutrient intake was less with each 5% increase in added sugars intake above 5-10%. Thirteen percent of the population had added sugars intake > 25%. The mean g-eq added sugars intake of 83.1 g-eq/day and added sugars food sources were comparable to the mid-1990s. Higher added sugars intakes were associated with higher proportions of individuals with nutrient intakes below the EAR, but the overall high calorie and the low quality of the U.S. diet remained the predominant issue. With over 80% of the population at risk for select nutrient inadequacy, guidance may need to focus on targeted healthful diet communication to reach the highest risk demographic groups for specific life stage nutrient inadequacies.

Modeling data with excess zeros and measurement error: application to evaluating relationships between episodically consumed foods and health outcomes

Dietary assessment of episodically consumed foods gives rise to nonnegative data that have excess zeros and measurement error. Tooze et al. (2006, Journal of the American Dietetic Association 106, 1575-1587) describe a general statistical approach (National Cancer Institute method) for modeling such food intakes reported on two or more 24-hour recalls (24HRs) and demonstrate its use to estimate the distribution of the food's usual intake in the general population. In this article, we propose an extension of this method to predict individual usual intake of such foods and to evaluate the relationships of usual intakes with health outcomes. Following the regression calibration approach for measurement error correction, individual usual intake is generally predicted as the conditional mean intake given 24HR-reported intake and other covariates in the health model. One feature of the proposed method is that additional covariates potentially related to usual intake may be used to increase the precision of estimates of usual intake and of diet-health outcome associations. Applying the method to data from the Eating at America's Table Study, we quantify the increased precision obtained from including reported frequency of intake on a food frequency questionnaire (FFQ) as a covariate in the calibration model. We then demonstrate the method in evaluating the linear relationship between log blood mercury levels and fish intake in women by using data from the National Health and Nutrition Examination Survey, and show increased precision when including the FFQ information. Finally, we present simulation results evaluating the performance of the proposed method in this context.

Reproducibility of ad libitum energy intake with the use of a computerized vending machine system

BACKGROUND

Accurate assessment of energy intake is difficult but critical for the evaluation of eating behavior and intervention effects. Consequently, methods to assess ad libitum energy intake under controlled conditions have been developed.

OBJECTIVE

Our objective was to evaluate the reproducibility of ad libitum energy intake with the use of a computerized vending machine system.

DESIGN

Twelve individuals (mean + or - SD: 36 + or - 8 y old; 41 + or - 8% body fat) consumed a weight-maintaining diet for 3 d; subsequently, they self-selected all food with the use of a computerized vending machine system for an additional 3 d. Mean daily energy intake was calculated from the actual weight of foods consumed and expressed as a percentage of weight-maintenance energy needs (%WMEN). Subjects repeated the study multiple times during 2 y. The within-person reproducibility of energy intake was determined through the calculation of the intraclass correlation coefficients (ICCs) between visits.

RESULTS

Daily energy intake for all subjects was 5020 + or - 1753 kcal during visit 1 and 4855 + or - 1615 kcal during visit 2. There were no significant associations between energy intake and body weight, body mass index, or percentage body fat while subjects used the vending machines, which indicates that intake was not driven by body size or need. Despite overconsumption (%WMEN = 181 + or - 57%), the reproducibility of intake between visits, whether expressed as daily energy intake (ICC = 0.90), %WMEN (ICC = 0.86), weight of food consumed (ICC = 0.87), or fat intake (g/d; ICC = 0.87), was highly significant (P < 0.0001).

CONCLUSION

Although ad libitum energy intake exceeded %WMEN, the within-person reliability of this intake across multiple visits was high, which makes this a reproducible method for the measurement of ad libitum intake in subjects who reside in a research unit. This trial was registered at clinicaltrials.gov as NCT00342732.

Total folate and folic acid intake from foods and dietary supplements in the United States: 2003-2006

BACKGROUND

The term total folate intake is used to represent folate that occurs naturally in food as well as folic acid from fortified foods and dietary supplements. Folic acid has been referred to as a double-edged sword because of its beneficial role in the prevention of neural tube defects and yet possible deleterious effects on certain cancers and cognitive function. Previous monitoring efforts did not include folic acid from dietary supplements and are therefore not complete.

OBJECTIVE

Our objective was to combine data on dietary folate (as measured by two 24-h recalls) and folic acid from dietary supplements (collected with a 30-d frequency questionnaire) with the use of the bias-corrected best power method to adjust for within-person variability.

DESIGN

The National Health and Nutrition Examination Survey (NHANES) is a nationally representative, cross-sectional survey. Linear contrasts were constructed to determine differences in dietary and total folate intake for age and racial-ethnic groups by sex; prevalence of inadequate and excessive intakes is presented.

RESULTS

In 2003-2006, 53% of the US population used dietary supplements; 34.5% used dietary supplements that contained folic acid. Total folate intake (in dietary folate equivalents) was higher for men (813 +/- 14) than for women (724 +/- 16) and higher for non-Hispanic whites (827 +/- 19) than for Mexican Americans (615 +/- 11) and non-Hispanic blacks (597 +/- 12); 29% of non-Hispanic black women had inadequate intakes. Total folate and folic acid intakes are highest for those aged > or =50 y, and 5% exceed the Tolerable Upper Intake Level.

CONCLUSIONS

Improved total folate intake is warranted in targeted subgroups, which include women of childbearing age and non-Hispanic black women, whereas other population groups are at risk of excessive intake.

Reproducibility and comparative validity of a food frequency questionnaire for Australian children and adolescents

BACKGROUND

Dietary intake during childhood and adolescence is of increasing interest due to its influence on adult health, particularly obesity, cardiovascular disease and diabetes. There is a need to develop and validate dietary assessment methods suitable for large epidemiologic studies of children and adolescents. Limited large scale dietary studies of youth have been undertaken in Australia, due partly to the lack of a suitable dietary intake tool. A self-administered, semi-quantitative food-frequency questionnaire (FFQ), the 'Australian Child and Adolescent Eating Survey' (ACAES), was developed for youth aged 9-16 years. This study evaluated reproducibility and comparative validity of the ACAES FFQ using assisted food records (FRs) as the reference method.

METHODS

The ACAES FFQ was completed twice (FFQ1 and FFQ2) at an interval of 5 months, along with four one-day assisted FRs. Validity was evaluated by comparing the average of the FRs with FFQ2 (n = 113) as well as with the average of FFQ1 and FFQ2 (n = 101). Reproducibility was evaluated by comparing FFQ1 and FFQ2 (n = 101). The two methods were compared using correlations, Kappa statistics and Bland-Altman plots.

RESULTS

Correlation coefficients for comparative validity ranged from 0.03 for retinol to 0.56 for magnesium for transformed, energy-adjusted, deattenuated nutrient data, with correlation coefficients greater than 0.40 for total fat, saturated fat, monounsaturated fat, carbohydrate, sugars, riboflavin, vitamin C, folate, beta-carotene, magnesium, calcium and iron. Correlation coefficients for reproducibility ranged from 0.18 for vitamin A to 0.50 for calcium for transformed, energy-adjusted, deattenuated nutrient data. The ACAES FFQ ranked individuals reasonably accurately, with the comparative validity analysis showing that over 50% of participants were classified within one quintile for all nutrients, with only a small percentage grossly misclassified (0-7%).

CONCLUSION

The ACAES FFQ is the first child and adolescent specific FFQ available for ranking the dietary intakes of Australian children and adolescents for a range of nutrients in epidemiologic research and public health interventions.

Validation of a semi-quantitative food frequency questionnaire measuring dietary vitamin K intake in elderly people

The study objective was to validate a semi-quantitative food frequency questionnaire (FFQ) specifically designed to measure dietary vitamin K intake. A 50-item FFQ was interviewer-administered and compared with data previously obtained from 5-day food records. Thirty-nine community-dwelling healthy men and women aged 65 to 85 years were recruited from the Montréal metropolitan area. Absolute and relative agreements between methods were assessed. Vitamin K intake measured by the vitamin K FFQ (mean+/-standard deviation; 222+/-186 microg/day) was significantly higher than that obtained by food records (135+/-153 microg/day; P<0.001). Bland-Altman analysis on log(10)-transformed data indicated that vitamin K intake from vitamin K FFQ was 2.26 times (95% confidence interval: 1.90 to 2.67) higher than food records, limits of agreement ranging from 0.80 to 6.35. However, correlation between methods was strong and highly significant (r=0.83; P<0.001). Cross-classification also showed that 72% of participants were correctly classified into thirds and only 8% were grossly miscategorized. Weighted kappa value (kappa=0.60) also indicated a good relative agreement. In light of these results, the vitamin K FFQ is a valid tool for ranking individuals according to their vitamin K intake. The poor absolute agreement likely results from the inability for food records to adequately measure the usual intake of episodically consumed foods, particularly those high in vitamin K. The vitamin K FFQ will be useful in large-scale, population-based research on vitamin K and disease as well as in clinical practice, especially that focusing on anticoagulant therapy.

Estimated intakes of meat and fish by children and adolescents in Australia and comparison with recommendations

Long-chain n-3 PUFA are considered important for cardiovascular health and brain development. Meat other than fish contributes significantly to total intakes of long-chain n-3 PUFA in adults; however, there are limited published data examining the intake of individual meat sources in children and adolescents in the Australian population. A review of literature was conducted using PubMed, Agricola and CAB Abstracts using the terms 'intake', 'beef', 'lamb', 'pork', 'poultry', 'fish', 'children' and 'adolescents' and using reference lists in published articles. Studies and surveys were identified that contained published values for intakes of meat or fish. Two national dietary surveys of children and adolescents were conducted in Australia in 1985 and 1995 and two regional surveys were conducted in Western Sydney and Western Australia in 1994 and 2003, respectively. Comprehensive data for the intake of individual meat sources were not reported from the 1995 survey, but estimations of intake were calculated from published values. Reported intakes of meat and fish are generally lower in females than males and tend to increase with age. Weighted mean intakes of red meat (beef plus lamb) across the three most recent studies were 67.3 and 52.2 g/d, respectively, for males and females aged between 7 and 12 years and 87.7 and 54.2 g/d, respectively, for males and females aged 12-18 years. These weighted intakes are within Australian guidelines and are likely to contribute significantly to total long-chain n-3 PUFA intake in children and adolescents in the Australian population.

A Review of the Relationship Between 100% Fruit Juice Consumption and Weight in Children and Adolescents

Studies assessing a relationship between consumption of 100% fruit juice by children and adolescents and weight are contradictory. The purpose of this review was to assess the 9 cross-sectional and 12 longitudinal studies that have looked at this association. Of the 9 cross-sectional studies reviewed, only 3 reported any association. Those studies used small, local convenience samples of preschool children; furthermore, selection criteria were not well defined. One found that preschool children consuming ≥ 12 fluid oz/day of 100% fruit juice had a higher prevalence of overweight than those who consumed less than 12 oz/day (32% vs 9%). Later, a separate study showed that this relationship held only for apple juice. Another study demonstrated an association of overweight and energy from juice. These 3 studies were not nationally representative. Only 3 of the longitudinal studies showed an association between 100% fruit juice consumption and weight; 1 found an association only in adolescent girls, and 2 reported an association in children who were already overweight. None of the longitudinal studies was nationally representative, but 5 had sample sizes of at least 1000, 2 were ethnically diverse, and 3 had geographically separate sites, suggesting that the findings could be applicable to wider populations. Based on the currently available evidence, it can be concluded that there is no systematic association between consumption of 100% fruit juice and overweight in children or adolescents. Data do support consumption of 100% fruit juice in moderate amounts and suggest that consumption of 100% fruit juice may be an important strategy to help children meet the current recommendations for fruit.

Relative reliability and validity of the Block Kids Questionnaire among youth aged 10 to 17 years

This cross-sectional study tested the reliability and validity of the Block Kids Questionnaire to assess diet during the past 7 days. Within a 7-day period, 10- to 17-year-old children and adolescents completed two 24-hour dietary recalls by telephone, followed by the Block Kids Questionnaire at the end of the week. Test-retest reliability was assessed using intraclass correlations for 18 participants who completed a second Block Kids Questionnaire 1 month later. Validity of the Block Kids Questionnaire compared to the 24-hour dietary recall was assessed for the whole sample and by age group using paired t tests and Pearson correlation coefficients adjusted for attenuation and energy intake. Participants were 83 children and adolescents (57% Hispanic, 21% African-American, and 23% white; 53% were female subjects, mean age 13 years). The Block Kids Questionnaire mean daily consumption values were higher for percent energy from carbohydrate, and servings of fruit, 100% fruit juice, and vegetables, and lower for all other categories compared to the 24-hour dietary recall. All reliability intraclass correlations were >0.30, except percent energy from protein and fruit/vegetable servings. Significant differences in the means between the two dietary assessment methods were noted for most nutrients/food groups. The adjusted correlation coefficients ranged from 0.69 for percent energy from carbohydrate to -0.03 for grain servings, with 60% of the food group servings <0.18. Overall, the majority of the correlation coefficients for children aged >12 years were higher than those aged < or =12 years. These results suggest that the Block Kids Questionnaire has validity for some nutrients, but not most food groups assessed, and appears more useful for adolescents.

[Food frequency questionnaire for adults from a population-based study]

The objective of the study was to develop a food frequency questionnaire for each of the following groups: female adults, male adults and male and female adults, based on dietary data obtained in a population-based study comprising individuals from different income levels. Dietary data was obtained from a 24-hour Dietary Recall applied to a probabilistic sample of 1,477 subjects in the city of São Paulo (Southeastern Brazil) in 2003. Food items accounting for at least 90% of total daily calorie intake and nutrients were selected. The reference time frame was the year preceding the interview and subjects had the choice of four serving sizes.

Food composition data: the foundation of dietetic practice and research

Food composition databases and dietary assessment systems are important tools for food and nutrition professionals. The availability and accessibility of data have improved over time along with the technology to convert the information into useful formats for planning diets, writing educational materials, counseling patients, and conducting research. Primary sources of food composition data include government, academic, and other institutional databases; the food industry; and scientific literature. Changes in the marketplace affect food availability and composition and affect the accuracy and adequacy of food composition databases. Improvements in both food composition data and in dietary assessment methods have worked synergistically to improve estimates of dietary intake. The development of databases for food frequency assessment systems requires special considerations for data aggregation for each food or food grouping in the questionnaires. Considerations for selecting a dietary assessment system include appropriateness of the data for the intended audience or purpose, efficiency of the search strategy for retrieving data, content and format of summary information, and cost. Needs for food composition data vary depending on dietetic practice area; however, most food and nutrition professionals will benefit from becoming more informed about food composition data, exploring new ways to educate themselves about databases and database systems, and advocating for what is most needed in dietetic practice.

Biomarkers for diet and cancer prevention research: potentials and challenges

As cancer incidence is projected to increase for decades there is a need for effective preventive strategies. Fortunately, evidence continues to mount that altering dietary habits is an effective and cost-efficient approach for reducing cancer risk and for modifying the biological behavior of tumors. Predictive, validated and sensitive biomarkers, including those that reliably evaluate "intake" or exposure to a specific food or bioactive component, that assess one or more specific biological "effects" that are linked to cancer, and that effectively predict individual "susceptibility" as a function of nutrient-nutrient interactions and genetics, are fundamental to evaluating who will benefit most from dietary interventions. These biomarkers must be readily accessible, easily and reliably assayed, and predictive of a key process(es) involved in cancer. The response to a food is determined not only by the effective concentration of the bioactive food component(s) reaching the target tissue, but also by the amount of the target requiring modification. Thus, this threshold response to foods and their components will vary from individual to individual. The key to understanding a personalized response is a greater knowledge of nutrigenomics, proteomics and metabolomics.

A new statistical method for estimating the usual intake of episodically consumed foods with application to their distribution

OBJECTIVE

We propose a new statistical method that uses information from two 24-hour recalls to estimate usual intake of episodically consumed foods.

STATISTICAL ANALYSES PERFORMED

The method developed at the National Cancer Institute (NCI) accommodates the large number of nonconsumption days that occur with foods by separating the probability of consumption from the consumption-day amount, using a two-part model. Covariates, such as sex, age, race, or information from a food frequency questionnaire, may supplement the information from two or more 24-hour recalls using correlated mixed model regression. The model allows for correlation between the probability of consuming a food on a single day and the consumption-day amount. Percentiles of the distribution of usual intake are computed from the estimated model parameters.

RESULTS

The Eating at America's Table Study data are used to illustrate the method to estimate the distribution of usual intake for whole grains and dark-green vegetables for men and women and the distribution of usual intakes of whole grains by educational level among men. A simulation study indicates that the NCI method leads to substantial improvement over existing methods for estimating the distribution of usual intake of foods.

CONCLUSIONS

The NCI method provides distinct advantages over previously proposed methods by accounting for the correlation between probability of consumption and amount consumed and by incorporating covariate information. Researchers interested in estimating the distribution of usual intakes of foods for a population or subpopulation are advised to work with a statistician and incorporate the NCI method in analyses.

Statistical methods for estimating usual intake of nutrients and foods: a review of the theory

Although 24-hour recalls are frequently used in dietary assessment, intake on a single day is a poor estimator of long-term usual intake. Statistical modeling mitigates this limitation more effectively than averaging multiple 24-hour recalls per respondent. In this article, we describe the statistical theory that underlies the four major modeling methods developed to date, then review the strengths and limitations of each method. We focus on the problem of estimating the distribution of usual intake for a population from 24-hour recall data, giving special attention to the problems inherent in modeling usual intake for foods or food groups that a proportion of the population does not consume every day (ie, episodically consumed foods). All four statistical methods share a common framework. Differences between the methods arise from different assumptions about the measurement characteristics of 24-hour recalls and from the fact that more recently developed methods build upon their predecessor(s). These differences can result in estimated usual intake distributions that differ from one another. We also demonstrate the need for an improved method for estimating usual intake distributions for episodically consumed foods.