![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
3 Department of Epidemiology, University of Washington, Seattle, WA 98195-7236; 4 Department of Epidemiology and Cancer Control, St. Jude Children's Research Hospital, Memphis, TN 38105-2794; and 5 Channing Lab, Harvard Medical School and Brigham and Women's Hospital, Boston, MA 02115
* To whom correspondence should be addressed. E-mail: beresfrd{at}u.washington.edu.
Studies of dietary behavior are an important, but sometimes neglected, aspect of epidemiological studies of nutrition and health. To be relevant to public health, epidemiological evidence of causal influences of dietary factors on disease risk should be followed by interventions to change those dietary factors to improve population health. For public health recommendations to be implemented successfully, studies of effective ways to encourage dietary behavior change are an essential component. Clearly, to know whether or not a behavioral change intervention works, it is necessary to assess the behavior at the beginning and at the end of the study and to detect differential change between the intervention and comparison condition. Indeed, the health promotion field needs measures that perform well for behavior change interventions targeting dietary intake.
The field of dietary assessment has developed enormously in the last 30 y. It is now recognized that methods for clinical assessment may not necessarily be suitable in large-scale studies (1). Furthermore, commonly used questionnaire methods may have more limitations than were initially appreciated (2). It is not necessarily the case that optimal assessment methods for an observational epidemiological study are also optimal for assessing dietary change. In particular, an instrument may yield a biased estimate of intake at a single point in time, but an unbiased estimate of change if the individual-level bias does not change between baseline and follow-up. This framework (assessing change) for the assessment of dietary measures has received less attention in the literature and is one place where some of the articles in this supplement can make a contribution.
It is important to assess the performance of measures with demonstrated reproducibility and strong correspondence with other established methods of dietary intake assessment in different studies of dietary behavior change using different populations. In practice, what is needed is a measure, or set of measures, that has face validity, takes a fairly short time to complete, is easy to understand, is not difficult or expensive to score, and is broadly applicable to large and diverse groups (3). Again, in a cross-sectional setting, measures need to be reproducible and congruent with other established measures; in an intervention setting, they need to be sensitive to dietary change and the resulting estimate of change should be congruent with good reference measures. Throughout this supplement, we have chosen specifically to avoid the use of the word "validity" in relation to metric properties of the short dietary screeners, in recognition of the lack of a gold standard for measuring dietary change.
In this context, several behavioral change studies funded by NIH and coordinated by the Office of Behavioral and Social Sciences Research (4) included a dietary component as part of the intervention and evaluation. Together these studies were included in a working group within a larger consortium of studies, details of which are described in this supplement (5). The Nutrition Working Group (NWG)6 within the Behavioral Change Consortium (BCC), as it is known, received additional funding from the National Cancer Institute (NCI) that provided a unique opportunity to explore measures of dietary behavior across diverse settings and demographic characteristics. Such a range of different groups would have been difficult or impractical to achieve in a single study. The development of this supplement issue of the Journal allowed investigators to explore the potential robustness of measurement properties of dietary assessment methods used in variable conditions across study sites, and with participants of differing race/ethnicity backgrounds, income levels, years of age, and health status. The properties are examined both cross-sectionally and longitudinally, where the measures are used to assess change.
The opportunity provided by the NWG has both strengths and weaknesses. Strength is that the measures used in the BCC NWG sites were those that are eminently practical in large population studies and had been demonstrated to have sound metric properties in a cross-sectional context (concordance with other measures) in at least one other study. The collection of studies in the BCC NWG allowed a reasonable test of robustness of performance in diverse conditions. A weakness of the approach, on the other hand, is that the original intervention studies were not designed to evaluate the performance of the measures as one of their primary objectives. Rather, each study was designed somewhat independently for the purpose of evaluating an intervention approach adapted to the behavior under study and to the population of interest.
The articles included in this supplement form a logical progression from cross-sectional to longitudinal designs. They deal with measures of percentage of energy from dietary fat (PFat), of fruits and vegetables, and of dietary supplements. A detailed overview of the contributing studies, their methods, and their study populations is provided by the investigators (5).
Four of the centers included fat intake as one of the behaviors of interest and incorporated the use of the NCI short screener for PFat and a traditional measure of dietary intake assessment (the multiple nonconsecutive 24-h recalls) in their baseline assessment. The article by Thompson et al. (6) describes the PFat screener, and the detailed steps taken to use it to estimate PFat and to adjust the comparison measure for within-person variability. This article compares absolute intake estimates within gender and site between the PFat screener score and PFat from the 24-h recall. It also estimates the correlation between the 2 methods using a measurement error model. The details provided about these statistical methods will be a useful resource for other investigators exploring the correspondence between 2 dietary assessment methods.
The NCI PFat screener also was assessed in the same 4 sites in terms of ability to measure change in fat intake (7). Specifically the authors evaluated the correlation between the PFat score and PFat from 24-h recalls over time. The deattenuated correlations from the measurement error model were estimated separately by gender and intervention condition for all sites combined. Consistency in estimating the intervention effect also was examined between the 2 measures.
Five sites participated in the evaluation of the NCI fruit and vegetable screener (FVS) (8). Mean estimates were compared, and correlations with the 24-h recall were estimated using a measurement error model. All analyses were conducted separately by site and gender. Two different scoring methods were explored, and an additional comparison measure, namely the sum of the 5 major serum carotenoids, was included. A single-item question on fruit and vegetable intake was also evaluated.
Change in fruit and vegetable intake (9) was assessed in the same 5 sites that administered the FVS. Correlations between the screener and the 1-item question, and 24-h recall estimates were compared at both baseline and follow-up. Analyses were conducted for all sites combined but stratified by gender. Mean treatment effect estimates were compared at follow-up with the 24-h recall values.
Two other articles included in this supplement have explored influences on dietary behaviors. The article by Davis et al. (10) used a novel statistical method, signal detection methodology, to examine factors associated with use of dietary supplements. The factors are identified in a random half of the dataset and confirmed in the other half. Social desirability trait (11) has long been considered to be a potential influence on responses to dietary intake questionnaires, particularly in the context of behavior change evaluation. This article examines the influence of social desirability bias on responses to the screeners evaluated in other articles in the supplement (PFat, FVS, and the 1-item question on fruit and vegetable intake) relative to corresponding 24-h recall estimates. By evaluating social desirability effects at both baseline and follow-up, the article is able to contrast biases over time and by intervention status of BCC NWG participants.
Together the articles in this supplement highlight many of the issues associated with dietary intake assessment in epidemiological and health promotion research, present quality examples of the use of state-of-the-art statistical methods to address measurement and behavioral biases, and illustrate many of the challenges of collaborating among many independent projects. They should offer a rich resource for other investigators.
 |
FOOTNOTES |
|---|
2 Author disclosures: S. A. A. Beresford, L. M. Klesges, and H. R. H. Rockett, no additional disclosures.
6 Abbreviations used: BCC, Behavioral Change Consortium; FVS, Fruit and Vegetable Screener; NCI, National Cancer Institute; NWG, Nutrition Working Group; PFat, percentage of energy from fat.
|
LITERATURE CITED |
|---|
 TOP LITERATURE CITED |
|---|
1. Fraser GE. A search for truth in dietary epidemiology. Am J Clin Nutr. 2003;78(3 Suppl):521S–5S.
2. Kipnis V, Subar AF, Midthune D, Freedman LS, Ballard-Barbash R, Troiano RP, Bingham S, Schoeller DA, Schatzkin A, Carroll RJ. Structure of dietary measurement error: results of the OPEN biomarker study. Am J Epidemiol. 2003;158:14–21.
3. Glasgow RE, Ory MG, Klesges LM, Cifuentes M, Fernald DH, Green LA. Practical and relevant self-report measures of patient health behaviors for primary care research. Ann Fam Med. 2005;3:73–81.
4. Solomon S, Kington R. National efforts to promote behavior-change research: views from the Office of Behavior and Social Sciences Research. Health Educ Res. 2002;17:495–9.
5. Yaroch AL, Nebeling L, Thompson FE, Hurley TG, Hebert JR, Toobert DJ, Resnicow K, Greene GW, Williams GC, et al. Baseline design elements and sample characteristics for seven sites participating in the Nutrition Working Group of the Behavioral Change Consortium. J Nutr. 2008;138:185S–192S.
6. Thompson FE, Midthune D, Williams GC, Yaroch AL, Hurley TG, Resnicow K, Hebert JR, Toobert DJ, Greene GW, et al. Evaluation of a short dietary assessment instrument for percentage energy from fat in an intervention study. J Nutr. 2008;138:193S–199S.
7. Williams GC, Hurley TG, Thompson FE, Midthune D, Yaroch AL, Resnicow K, Toobert DJ, Greene GW, Peterson KE, et al. Performance of a short percentage energy from fat tool in measuring change in dietary intervention studies. J Nutr. 2008;138:212S–217S.
8. Greene GW, Resnicow K, Thompson FE, Peterson KE, Hurley TG, Hebert JR, Toobert DJ, Williams GC, Elliot DL, et al. Correspondence of the NCI fruit and vegetable screener to repeat 24-H recalls and serum carotenoids in behavioral intervention trials. J Nutr. 2008;138:200S–204S.
9. Peterson KE, Hebert JR, Hurley TG, Resnicow K, Thompson FE, Greene GW, Shaikh AR, Yaroch AL, Williams GC, et al. Accuracy and precision of two short screeners to assess change in fruit and vegetable consumption among diverse populations participating in health promotion intervention trials. J Nutr. 2008;138:218S–225S.
10. Davis RE, Resnicow K, Atienza AA, Peterson KE, Domas A, Hunt A, Hurley TG, Yaroch AL, Greene GW, et al. Use of signal detection methodology to identify subgroups of dietary supplement use in diverse populations. J Nutr. 2008;138:205S–211S.
11. Hebert JR, Hurley TG, Peterson KE, Resnicow K, Thompson FE, Yaroch AL, Ehlers M, Midthune D, Williams GC, et al. Social desirability trait influences on self-reported dietary measures among diverse participants in a multicenter multiple risk factor trial. J Nutr. 2008;138:226S–234S.
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
