Researchers at the National Cancer Institute (NCI) and elsewhere have developed a method to model particular aspects of usual dietary intakes of foods and nutrients using 24-hour recalls. This method can be used to:
- estimate the distribution of usual intake for a population or subpopulation;
- assess the effects of non-dietary covariates on usual consumption; and
- correct (at least partially) bias caused by measurement error in estimated associations between usual dietary intakes and health outcomes using the statistical technique of regression calibration. (Note: This modeling technique does not accurately estimate usual intake for individuals.)
The NCI method addresses some, but not all, of the measurement error issues inherent in the analysis of 24-hour recall data. A detailed explanation of measurement error concepts and implications for the use of the NCI method has been written by NCI’s Biometry Research Group and is available for additional information.
The premise of the NCI method is that usual intake is equal to the probability of consumption on a given day times the average amount consumed on a "consumption day." The exact methods used for dietary components that are consumed nearly every day by nearly everyone differ slightly from those used for dietary components that are episodically consumed. In general, the former category (ubiquitously consumed or consumed daily by almost everyone) includes most nutrients whereas the latter category (episodically consumed) includes most foods, though there are exceptions.
For episodically-consumed components, a two-part model is used. The first part estimates the probability of consumption using logistic regression with a person-specific random effect. The second part specifies the consumption-day amount using linear regression on a transformed scale, also with a person-specific effect. Parts I and II are linked by allowing the two person- specific effects to be correlated and by including common covariates in both parts of the model. Intake data from 24-hour recalls provide the values for the dependent variable, while average daily intake estimated from a food frequency questionnaire (FFQ) may be incorporated as one of the covariates. Other covariates (e.g., gender, age, race) may be included, particularly if there is interest in subpopulations. The resulting estimated model parameters can then be used to estimate distributions or as input to further analyses, depending on the application of interest (see bullets above).
For dietary components that are consumed daily by most persons, the process is the same, except that the probability part of the model is not needed because the probability of consumption is assumed to be 1.
Learn more about the:
Frequently Asked Questions
Has the NCI method been validated?
Evidence for the validity of the method as it relates to estimating the distribution of usual intakes of episodically-consumed components has been published in a series of papers in the Journal of the American Dietetic Association. The use of the method to estimate the distribution of usual intakes of components consumed nearly daily by nearly all persons has been described in Statistics in Medicine (Tooze JA et al., 2010) and the Journal of Nutrition (Freedman LS et al., 2010).
Analyses establishing the method’s validity for predicting individual usual intake for use in regression analyses (e.g., to examine relationships between diet and health) have also been published. A 2009 paper published in Biometrics describes the use of the NCI method to evaluate relationships between episodically-consumed foods and health outcomes.
What data are needed to use the NCI method?
The NCI method requires the availability of data from one or more non-consecutive 24-hour recalls or food records for a representative sample of individuals from the population of interest. Data from two or more non-consecutive recalls or records are needed for at least a subset of these individuals.
What part does the frequency instrument play in the NCI method? Under what circumstances is it helpful?
The NCI method involves using data from two or more 24-hour recalls as well as covariates, which may include data from an FFQ such as the NHANES 2003-2006 Food Frequency Questionnaire (formerly called the Food Propensity Questionnaire). A frequency instrument can substantially improve the power to detect relationships between dietary intakes as predictor variables and other variables. The magnitude of improvement depends on the proportion of zeros in the report of the dietary component, with the FFQ having a great impact for those with a large number of zero intakes.
When applying the method to estimate usual intake distributions, satisfactory results can generally be obtained without the FFQ as a covariate. However, there are conditions under which the FFQ may help, particularly for estimating the tails of the distributions.
How does the NCI method compare to other methods?
For dietary components consumed nearly daily by nearly all persons, various methods have been proposed either to estimate the distribution of usual intakes or to predict individual-level usual intake for use in regression analyses. To date, no other unified methods that estimate usual intake of episodically-consumed foods from 24-hour recall data have been available that are appropriate for use in both estimating distributions and in analyses of diet-health associations.
Two other methods have been developed to estimate the distribution of usual intake of episodically-consumed components with a few days of 24-hour recalls. The simplest uses the within-person mean and usually leads to biased estimates of the prevalence of either inadequate or excess intake. This is because this method does not:
- account for reported days without consumption or for consumption-day amounts that are positively skewed;
- distinguish within-person from between-person variation;
- allow for the correlation between the probability of consumption and the consumption-day amount; or
- relate covariate information to usual intake.
The Iowa State University (ISU) method for estimating the distribution of episodically-consumed components uses modeling and meets most of the challenges noted above. However, it does not allow for correlation between probability and amount and cannot incorporate covariate information.
The NCI method was designed to meet all of these challenges and as such is an improvement over previous methods. It allows efficient estimation of usual intake distributions of daily and episodically consumed components for populations and subpopulations and prediction of individual intake for use in regression analyses. In the absence of such methods, analyses relating intakes to health have often used reported rather than predicted usual intakes, leading to biased estimation.
What are the assumptions of the NCI method?
The method assumes that the 24-hour recall is an unbiased instrument for measuring usual food intake -- in other words, that it does not misclassify the respondent's food intake and that it provides an unbiased measure of the amount of food consumed on a consumption day. Further, the macros provided to implement the NCI method required an assumption that there are no true never-consumers of a given nutrient or food. This is because the logistic regression used to model the probability of consumption does not predict a zero value.
What important caveats are associated with the NCI method?
- Many studies have found misreporting of energy intake on both 24-hour recalls and food frequency instruments, almost always in the direction of underreporting; this suggests that some foods are underreported.
- If only a limited number of repeated 24-hour recalls are available, reliable separation between non-consumers, irregular consumers, and always-consumers is not possible. Therefore, in the absence of extra information about ever- vs. never-consumption, the NCI method does not estimate the proportion of non-consumers/always-consumers of a given food.