Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
The relationship between diet during adolescence and disease later in life is of considerable research interest, because adolescence may be a more etiologically relevant time period than adulthood for some chronic diseases. The retrospective recall of adolescent diet by adults, although not ideal, may provide a practical way to assess diet–disease relationships, but few studies have assessed the validity of this recall.
We designed the 124-item High School Food Frequency Questionnaire (HS-FFQ) to ask women, ages 33 to 52 years, in the Nurses’ Health Study II (NHSII),1 about their diet during high school. In a previous study, we found good 4-year reproducibility (r = 0.65) and adequate validity (r = 0.40) for nutrient intakes when comparing participants’ diet data with their mother's reports.2 In this investigation, we used dietary measures actually collected during adolescence from NHSII offspring to further examine the validity of the HS-FFQ.
This protocol was approved by the Partners Institutional Review Board at Brigham and Women's Hospital. Voluntary return of the self-administered questionnaire was considered implied consent.
The young adults in this investigation were participants in a previous validation study, conducted 10 years earlier, comparing the Youth/Adolescent Questionnaire (YAQ) with 3 24-hour recalls.3,4 In this current study of the HS-FFQ, we selected all eligible participants (n = 153) who had been between the ages of 13 to 18 years when completing the first YAQ. Of the 108 NHSII offspring who we could contact, 90 responded to the questionnaire. Ten offspring were subsequently excluded based on established criteria (caloric intake <600 or >5000 kcal/d; more than 70 food items left blank; or more than one food section, other than dairy or meat sections, left completely blank), leaving 80 participants for this analysis. The mean age of these participants was 25 years (range = 23–29 years) when the HS-FFQ was administered, with most being female (69%) and white. These responders were largely similar demographically to the original study sample from which they were selected (data not shown).
Briefly, the HS-FFQ5 is a 124-item food-frequency questionnaire (FFQ) that was self-administered in 2003. It asked how often, on average, participants ate a specified food or supplement over the 4 years of high school. The nutrient database was constructed for the years when these participants were in high school.6–9 The 3 24-hour diet recalls had been telephone-administered 10 years earlier (between 1993 and 1994) and spaced 5 months apart to account for seasonal variation. Three recalls were considered practical and feasible for this population of adolescents and are generally accepted as the minimum needed to characterize usual individual intake.10 Lastly, a 131-item self-administered FFQ (YAQ), asking about diet in the past year, had been administered twice, in 1993 and 1994. It is similar in content to the HS-FFQ. More detailed information about the content, development, and nutrient databases of these assessments are provided in other reports.2–4
Nutrients were adjusted for energy intake using the residual method to account for variation in nutrient intakes resulting from total energy intake11 and log-transformed to improve their normality for the correlation analyses.10
We calculated Pearson correlation coefficients to compare nutrient intakes from: 1) the HS-FFQ and the average of 3 24-hour recalls (our gold standard); 2) the HS-FFQ and the average of the 2 YAQs, an alternative standard, to analyze additional nutrients that were not available from the 24-hour recalls, and 3) the average of YAQs and the 3 24-hour recalls to assess the use of the YAQ as a standard for adolescent diet (we used the average of the YAQ to better represent diet over all years in high school).
Random fluctuations above and below a person's true long-term average such as those resulting from day-to-day variability in nutrient intake can attenuate correlations.10 Thus, we corrected for the effects of this within-person variation to obtain a value similar to that obtained from a large number of replicates (24-hour recalls or YAQs depending on the analysis).12,13 We also calculated the 95% confidence intervals for these corrected correlations.14
The mean nutrient intakes, unadjusted for energy, from the 24-hour recalls and HS-FFQ were within 20% of each other, with the exception of total fat, monounsaturated fat, cholesterol, total protein, dietary fiber, vitamin A, and beta-carotene (Table 1). Notably, mean calorie intake was higher for the HS-FFQ than either the 24-hour recalls or YAQ. When adjusted for total calories (percent of energy), total fat, protein, and carbohydrates were similar across measures.
The energy-adjusted nutrient correlations between the 3 24-hour recalls and HS-FFQ were modest (average r = 0.31; range = 0.10–0.50) (Table 2). After we corrected for within-person variation, the average corrected correlation for nutrients was 0.45 (range = 0.16 for carbohydrates to 0.68 for vitamin C). Because correlation coefficients are a function of between-person variation, they were higher for energy-adjusted nutrients with relatively high between-person variation (such as vitamin C and caffeine) compared with carbohydrates and protein.
The correlations between the 24-hour recalls and the mean of the 2 YAQs were moderate (mean energy-adjusted r = 0.49, range = 0.26–0.69; and mean corrected r = 0.69, range = 0.45–0.97) (Table 2). Overall, the correlations for this subsample of participants were higher than from our previous validation study of 261 children and adolescents, ages 9 to 18 years, from whom these 24-hour recalls and YAQs were originally collected (mean energy-adjusted r = 0.45 and mean corrected r = 0.54).4
The energy-adjusted correlations between the HS-FFQ and the average of 2 YAQs were moderate (mean r = 0.48; range = 0.31–0.69), whereas the mean of the corrected correlations was 0.58 with a range of 0.40 to 0.80 (Table 2). Because the correlations for nutrient intakes with and without supplements were similar, we report total intakes only. The corrected correlations for those nutrients not available from the 24-hour recalls were varied: animal fat = 0.40, vegetable fat = 0.64, total fructose = 0.44, vitamin E = 0.57, and vitamin D = 0.68. The correlations for women (mean corrected r = 0.60) and men (mean corrected r = 0.57) were similar to the total sample.
Our results suggest that the HS-FFQ is able to reasonably capture adolescent diet as recalled by adults. After accounting for total energy, most intakes were similar across all 3 measures, indicating that the HS-FFQ is able to adequately represent absolute intake. Moreover, the corrected correlations between the HS-FFQ and 24-hour recalls (our gold standard measure) were good overall, as were the corrected correlations between the HS-FFQ and 2 YAQs. In addition, the YAQ had good validity when we compared it with the 24-hour recalls in this group of participants, suggesting it has merit as a measure of real-time adolescent diet. Because the ability to assess diet–disease relationships depends in part on the between-person variation (which is reflected by the correlation coefficient), these results also suggest that the HS-FFQ can reasonably rank individuals.
The performance of the HS-FFQ among older individuals is perhaps of greater interest because they are typically studied in investigations of chronic disease. However, our results in this population of young adults are consistent with our earlier findings of good 4-year reproducibility and adequate validity (when comparing participants’ reports with mother's reports) among NHSII participants.2
The correlations we observed were higher than in other studies investigating the validity of recalled adolescent diet. For example, Wolk et al15 reported modest nutrient correlations (average r = 0.27) when comparing adolescent diet remembered by adult participants with recalled reports from their adult siblings about their own diet. Moreover, Dwyer et al16 found lower correlations (median r = 0.12) comparing recalled foods eaten at age 18 from a FFQ with original diet histories, which were rather crude measures. Our results, however, are consistent with the many studies addressing the validity of adult diet, recalled 11 to 24 years in the past (range of average nutrient correlations for these studies, 0.23–0.59),17–20 which have also been reviewed elsewhere.21,22
This study has some limitations. A larger sample size and additional days of 24-hour recalls would have improved precision (Supplementary Table 1, available with the online version of the article). Also, we did not have data to assess bias as a result of current diet. However, in the NHSII study, current diet was only weakly correlated with recalled adolescent diet (mean r = 0.20), suggesting that this bias is likely to be minimal.2 In addition, our findings may not be generalizable to substantially older individuals or to populations of different ethnicity.
This study also has several strengths. It is one of the few studies that have examined recall of diet during adolescence. One advantage of asking respondents about high school is that it may help participants anchor their memories during a distinct time of life and facilitate memory of foods they had eaten during that time.23 In this study, errors from the HS-FFQ and 24-hour recalls (ie, resulting from memory, perception of portion sizes, and food lists) are largely independent, thus minimizing correlated error.10 In addition, we were able to compare intakes from the HS-FFQ with reports from a similar questionnaire (YAQ) administered when these participants were of high school age. Taken together with our earlier results, these findings suggest that a food-frequency questionnaire used to retrospectively assess diet during high school can be useful in the study of diet and disease relationships.
We thank Gary Chase, Mike Atkinson, and Karen Corsano for their technical and logistic assistance.
1. Colditz GA, Hankinson SE. The Nurses’ Health Study: lifestyle and health among women. Nat Rev Cancer
2. Maruti SS, Feskanich D, Colditz GA, et al. Adult recall of adolescent diet: reproducibility and comparison with maternal reporting. Am J Epidemiol
3. Rockett HR, Wolf AM, Colditz GA. Development and reproducibility of a food frequency questionnaire to assess diets of older children and adolescents. J Am Diet Assoc
4. Rockett HR, Breitenbach M, Frazier AL, et al. Validation of a youth/adolescent food frequency questionnaire. Prev Med
5. Nurses’ Health Study web site. Available at: http://www.channing.harvard.edu/nhs/questionnaires/index.shtml#n2hsffq
. Accessed November 8, 2005.
6. US Department of Agriculture ARS. USDA Nutrient Database for Standard Reference, Release 14, SR14. Nutrient Data Laboratory Home Page. Available at: http://www.nal.usda.gov/fnic/
7. US Department of Agriculture ARS. USDA Nutrient Database for Standard Reference, Release 11. Nutrient Lab Home Page. Available at: http://www.nal.usda.gov/fnic/
8. Holden J, Eldridge A, Beecher G. Carotenoid content of US foods: an update of the database. Journal of Food Composition and Analysis
9. Matthews RH, Pehrsson PR, Farhat-Sabet M. Sugar content of selected foods: individual and total sugars. USDA Home Economics Research Report #48.
10. Willett W. Nutritional Epidemiology
, 2nd ed. New York: Oxford University Press; 1998.
11. Willett W, Stampfer MJ. Total energy intake: implications for epidemiologic analyses. Am J Epidemiol
12. Beaton GH, Milner J, Corey P, et al. Sources of variance in 24-hour dietary recall data: implications for nutrition study design and interpretation. Am J Clin Nutr
13. Liu K, Stamler J, Dyer A, et al. Statistical methods to assess and minimize the role of intra-individual variability in obscuring the relationship between dietary lipids and serum cholesterol. J Chronic Dis
14. Rosner B, Willett WC. Interval estimates for correlation coefficients corrected for within-person variation: implications for study design and hypothesis testing. Am J Epidemiol
15. Feskanich D, Rimm EB, Giovannucci EL, et al. Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. J Am Diet Assoc
16. Wolk A, Bergstrom R, Hansson LE, et al. Reliability of retrospective information on diet 20 years ago and consistency of independent measurements of remote adolescent diet. Nutr Cancer
17. Dwyer JT, Gardner J, Halvorsen K, et al. Memory of food intake in the distant past. Am J Epidemiol
18. Wu ML, Whittemore AS, Jung DL. Errors in reported dietary intakes. II. Long-term recall. Am J Epidemiol
19. Bakkum A, Bloemberg B, van Staveren WA, et al. The relative validity of a retrospective estimate of food consumption based on a current dietary history and a food frequency list. Nutr Cancer
20. Jensen OM, Wahrendorf J, Rosenqvist A, et al. The reliability of questionnaire-derived historical dietary information and temporal stability of food habits in individuals. Am J Epidemiol
21. Sobell J, Block G, Koslowe P, et al. Validation of a retrospective questionnaire assessing diet 10–15 years ago. Am J Epidemiol
22. Friedenreich CM, Slimani N, Riboli E. Measurement of past diet: review of previous and proposed methods. Epidemiol Rev
23. Frazier AL, Willett WC, Colditz GA. Reproducibility of recall of adolescent diet: Nurses’ Health Study (United States). Cancer Causes Control