![[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
*
School of Dietetics and Human Nutrition and
Department of Animal Science, McGill University, Canada, H9X 3V9
**
Centre de recherche sur le vieillissement, Institut universitaire de gériatrie de Sherbrooke, Canada, J1H 4C4; and
Department of Epidemiology and Biostatistics, McGill University, Canada, H37 3J7.
3To whom correspondence should be addressed. E-mail: gray-donald{at}macdonald.mcgill.ca.
 |
ABSTRACT |
|---|
 TOP ABSTRACT SUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
KEY WORDS: • within-person variability • between-person variability • 24-h recall • dietary methodology
An increasing number of studies point to dietary intake as a risk factor for numerous chronic diseases (1
). Accurate measurement of usual intake, which refers to long-term average daily nutrient intake of an individual (2
), is required to make links between diet and disease. Because dietary intake of an individual is not constant from day to day (3
), an understanding of variability in food intake is required to estimate usual intake. Variability in food intake arises both because each individual differs in the types and amounts of food consumed from one day to the next (within- or intra-subject variability) (4
) and because individuals differ from each other in their food intakes (between- or inter-subject variability) (5
,6
). Variability in dietary intake influences the number of days required to estimate food and nutrient intakes accurately. The number of days required to obtain reliable estimates of food and nutrient intakes for individuals (4
,7
) varies from that required to classify individuals correctly into groups for analytical purposes (8
,9
).
A number of studies have examined within- and between-subject variability for different nutrients (7
,8
,10
,11
); however, less work has been done on variability in food intakes (12
,13
). With increasing interest in the association between foods and disease risk (14
,15
), a clearer understanding of variability in food intakes is important.
When 
2 d of intake data are available, both between- and within-subject variability can be determined by ANOVA (7
,8
,16
–19
). Mixed models that take into account both fixed and random effects are now available (20
). These models can be used to control for other factors that may influence variability.
The objectives of the present study are to compare ratios of within- to between-subject variability for foods and nutrients by the mixed model procedure, adjusting for several factors with the mixed model procedure unadjusted for other factors, and to estimate the number of days required to classify subjects correctly into groups using within- and between-subject variances and the number of days required to assess usual intakes for individuals with accuracy using within-subject variability alone in the calculation.
|
SUBJECTS AND METHODS |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). Briefly, a sample of 1543 noninstitutionalized adults aged 18–65 y was randomly selected from five regions of Canada, including the Atlantic provinces, Quebec, Ontario, the Prairie provinces and British Columbia, using a multistage random sampling strategy (21
). They were interviewed between September 1997 and August 1998. Pregnant and lactating women and those who did not speak English or French were excluded. The final sample included 572 men and 971 women. For the present study, two subjects who did not report their level of education were excluded resulting in 571 men and 970 women. A repeat interview was conducted on 29% of subjects to estimate within-subject variability in nutrient intake. Systematic sampling was used for this purpose; the second person initially interviewed and every third person thereafter were selected, providing a subsample of 446 subjects. Information on height, weight, smoking status and educational level was collected by questionnaire. Dietary intake was recorded by dietitians using the 24-h recall method in a face-to-face interview. Detailed descriptions of all foods and beverages consumed during the 24-h period before the interview, including the quantity, cooking method and brand names were recorded. Quantities were estimated using standard graduated glasses, bowls, spoons and a ruler. Supplement intakes were not considered for the present analyses. Nutrient intakes were analyzed using the Candat nutrient analysis program (Godin London Inc., London, Canada) and the 1997 Canadian Nutrient File.
Nutrients examined in this analysis included the macronutrients, fat, protein and carbohydrate, and the micronutrients, calcium, iron, folate, vitamin A and vitamin C. Carotene and vitamin E were not assessed because data are not available for these nutrients for many foods in the Canadian nutrient file.
Foods including meat, vegetables, fruits (including juice), green leafy vegetables (lettuce/spinach/cabbage), milk and bread were examined as were the four food groups based on Canada’s Food Guide to Healthy Eating (22
). These latter groups did not include mixed dishes from different food groups in which foods were entered as a mixed dish. For mixed foods, when specific amounts for each ingredient were described, it was possible to categorize into specific food groups. Total grams of the food grouping consumed was used as a measure of intake. The frequency of consumption of foods was examined in the subsample with 2 d of intake to provide information on variability in food intake.
The distribution of each nutrient was examined for normality, and appropriate transformations (log and square root) were performed for nutrients with skewed distributions (23
). An appropriate transformation could not be found for Vitamin A.
Within- and between-subject variability were estimated by the mixed model procedure for men and women in two ways, i.e., one that was unadjusted for the fixed effects and the other adjusted for the fixed effects of gender, age, education, smoking, season and size of family. It should be noted that in the analyses, means and variances are considered separately. Mixed model procedures also enable examining variances by gender in addition to adjusting for the fixed effect means of gender, age, smoking, education, family size and season. An analysis of heterogeneity of variances yielded variances attributed to men and women separately in both models. Thus, the results presented (which it should be noted, are variances and variance ratios) are stratified by gender. The above analyses were performed using both untransformed and transformed data but because similar ratios were obtained for all nutrients, the results for untransformed data are reported. The mixed model procedure permits the use of data for subjects with either 1 or 2 d of intake; data from both days of intake were used for estimating within-subject variability, whereas one day’s intake was used for estimating between-subject variability. The above analyses were performed using the mixed model procedure (Proc Mixed) of SAS (version 6.12, 1996, Cary,NC).
Within- and between-subject variances obtained by the mixed model procedure that adjusted for factors were used to determine the number of days required to obtain reliable estimates of food and nutrient intake by two different methods, one using both within- and between-subject variances and the other using only the within-subject variability. The first method allows estimation of the number of days required to obtain a specified level of correlation between observed and true intakes and is obtained by the formula,
 |
where d is the number of days, r represents the unobservable correlation between observed and true mean nutrient intakes of subjects, and sw2/sb2 is the within/between-subject variance ratio (8
). A higher value of r indicates a higher proportion of subjects correctly classified and a lower proportion misclassified (8
). If the ratio of variances is low, then fewer days of observation are required to classify subjects correctly (8
,9
), which may be because of low within-subject variability or high between-subject variability.
For some purposes, it may be necessary to assess actual intake of individuals with a given level of confidence (7
). The number of days of observation required for a given level of confidence (7
) can be calculated as follows: d = (Z
CV0/D0)2 where d is the number of days required per person, Z is the normal deviate, e.g., 1.96, CV0 is the within-subject variation, calculated as square root of within-subject variance (SD)/mean intake and D0 is the specified limit as a percentage of long-term intake (24
). Using this calculation, the number of days needed for the observed estimate of a person’s intake to lie within a specified percentage of the true mean, 95% of the time can be obtained (24
). All analyses were performed using SAS (version 6.12, 1996, Cary,NC).
|
RESULTS |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Within- to between-subject variability ratios (sw2/sb2) obtained for the selected nutrients using the mixed model procedure that adjusted for other factors tended to be higher than the unadjusted model (1.07 vs. 0.49 and 2.04 vs. 1.76 for energy in men and women, respectively) (Table 1
). The higher ratios obtained using the adjusted mixed model procedure occurred because adjusting tends to reduce between-subject variability.
|
2–6 d were required to estimate nutrient intakes with good accuracy (r = 0.8) (Table 2
). Using within-subject variability alone to estimate the accuracy of individual measurements, many more days were required to estimate nutrient intakes within 20 or 30% of usual intake. Comparison of both methods indicates that more repeat observations are required to obtain estimates of usual nutrient intakes for individuals than are required to place subjects in groups relative to each other.
|
Variability ratios (sw2/sb2) for food groupings were computed by the mixed model procedures described previously for nutrients (Table 3
). The within/between-subject ratios for most food/food groups tended to be slightly higher by the mixed model procedure that adjusted for other variables (1.15 vs. 0.96 and 2.07 vs. 1.87 for grain products among men and women, respectively), indicating that, as for nutrients, adjustment tended to reduce the between-subject variability, thereby increasing the ratio. The variance ratios were generally higher for food groupings than for nutrients, with the exception of fruits (including juices) and milk food groups. These higher variance ratios mean that more days of food intake would be required than those estimated for nutrients.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Adjusting for several factors when estimating variance ratios results in a reduction in between-subject variance. This may be due to differences in total intake because of age, sex, smoking status or physical activity. The resulting higher variance ratio indicates that more days are needed to obtain reliable estimates of nutrient intakes. Not adjusting for these factors and thus estimating a lower number of days required could result in the study having insufficient power to detect differences in intakes when these variables are controlled in multivariate analyses.
In the present study, gender, age, smoking, education, season and size of family contributed to variability in the intakes of most foods and nutrients. Similar to other studies, gender, age and smoking contributed to variability in nutrient intakes (7
,25
,26
). It has been reported that there are differences in consumption of certain foods by level of education and family size (25
,27
–29
). As reported in other studies, season did not contribute to variability in the present study (11
,30
).
The within- to between-subject variability ratios were generally > 1 as reported in other studies (7
,8
,10
–12
). The ratios for energy, protein, carbohydrate, calcium, vitamin C, iron, (7
,8
) grains, vegetable and fruit food groups (12
) were similar to those reported for similar populations. The variance ratio for fat was similar to that reported in literature among men (7
); however, the ratio was higher for women. For women, the within-subject variability was higher than for men and the between-subject variability lower, possibly reflecting inconsistent use of low fat products or less regular consumption of fat-containing foods. The number of days required to estimate usual intakes for carbohydrate and calcium was similar to that reported in other studies for similar populations using within/between variance methodology (8
).
Nutrients had lower within- to between-subject variability ratios compared with foods possibly because nutrients come from many food sources. Among foods, there was greater variability in the intake of specific foods compared with whole food groups. It is possible that 2 d of measured intake for each individual is not sufficient to obtain a true picture of variability in some less routinely eaten foods.
A question often asked at the design stage pertains to the number of days of observation required to assess usual intakes of individuals and groups (7
,9
). In our analyses, considerably more days were required to obtain reliable estimates of intakes for individuals compared with relative ranking of subjects into groups. If the objective is to obtain accurate estimates of individuals for counseling purposes (7
), then the method involving the use of within-subject variability must be considered due to the large day-to-day variation in dietary intakes of each individual. Studies have indicated that most nutrients have high within-subject variability, resulting in a greater number of days to estimate intakes reliably for individuals (7
,24
,9
). The food frequency method may be an option for specific foods (7
); however, food-frequency questionnaires have been estimated to measure nutrient intakes only as accurately as 2–3 repeat 24-h recalls (31
).
A possible limitation of the present study was that the day of the week effect (3
,24
) was not considered. Attention was given to avoid conducting a repeat interview on the same day of the week for each subject; however, the choice of days was not necessarily a weekday and a weekend day. Interviews for the same subject were not done on consecutive days to avoid misleading correlations associated with consecutive days of dietary assessment (3
).
In conclusion, within/between-subject ratios for foods and nutrients tended to be higher with adjustment compared with the unadjusted model, indicating the need to adjust for confounding variables when calculating the number of days to obtain reliable estimates of nutrient intakes.
|
FOOTNOTES |
|---|
2 Supported by Fonds de la recherche en santé du Québec and the Beef Information Center with funds obtained from the Beef Industry Development Fund.
Manuscript received 8 June 2002. Initial review completed 10 July 2002. Revision accepted 18 October 2002.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
1. Shikany, J. M. & White, G. L., Jr (2000) Dietary guidelines for chronic disease prevention. South. Med. J. 93:1138-1151.[Medline]
2. Nusser, S. M., Carriquiry, A. L., Dodd, K. W. & Fuller, W. A. (1996) A semiparametric transformation approach to estimating usual daily intake distributions. J. Am. Stat. Assoc. 91:1440-1449.
3. National Research Council (1986) Nutrient Adequacy.Assessment Using Food Consumption Surveys. Subcommittee on Criteria for Dietary Evaluation. National Academy of Sciences 1986 National Academy Press Washington, DC.
4. Liu, K., Stamler, J., Dyer, A., McKeever, J. & McKeever, P. (1978) Statistical methods to assess and minimize the role of intra-individual variability in obscuring the relationship between dietary lipids and serum cholesterol. J. Chron. Dis. 31:399-418.[Medline]
5. Sempos, C. T., Looker, A. C., Johnson, C. L. & Woteki, C. E. (1991) The importance of within-person variability in estimating prevalence. Macdonald, I. eds. Monitoring Dietary Intakes 1991:99-109 Springer Verlag New York, NY .
6. Gibson, R. S. (1990) Precision in dietary assessment. Principles of Nutritional Assessment 1990:97-116 Oxford University Press New York, NY.
7. Beaton, G. H., Milner, J., Corey, P., McGuire, V., Cousins, M., Stewart, E., de Ramos, M., Hewitt, D., Grambsch, P. V., Kassim, N. & Little, J. A. (1979) Sources of variance in 24-h dietary recall data: implications for nutrition study design and interpretation. Am. J. Clin. Nutr. 32:2546-2559.
8. Nelson, M., Black, A. E., Morris, J. A. & Cole, T. J. (1989) Between- and within-subject variation in nutrient intake from infancy to old age: estimating the number of days required to rank dietary intakes with desired precision. Am. J. Clin. Nutr. 50:155-167.
9. Marr, J. W. & Heady, J. A. (1986) Within- and between-person variation in dietary surveys: number of days needed to classify individuals. Hum. Nutr. Appl. Nutr. 40A:347-364.[Medline]
10. Beaton, G. H., Milner, J., McGuire, V., Feather, T. E. & Little, J. A. (1983) Source of variance in 24-h dietary recall data: implications for nutrition study design and interpretation. Carbohydrate sources, vitamins and minerals. Am. J. Clin. Nutr. 37:986-995.
11. Sempos, C. T., Johnson, N. E., Smith, E. L. & Gilligan, C. (1985) Effects of intraindividual and interindividual variation in repeated dietary records. Am. J. Epidemiol. 121:120-130.
12. Sempos, C. T, Johnson, N. E., Gilligan, C. & Smith, E. L. (1986) Estimated ratios of within-person to between-person variation in selected food groups. Nutr. Rep. Int. 34:1121-1127.
13. Borrelli, R., Simonetti, M. S. & Fidanza, F. (1992) Inter- and intra-individual variability in food intake of elderly people in Perugia (Italy). Br. J. Nutr. 68:3-10.[Medline]
14. Slavin, J. L., Jacobs, D., Marquart, L. & Wiemer, K. (2001) The role of whole grains in disease prevention. J. Am. Diet. Assoc. 101:780-785.[Medline]
15. Van Duyn, M. A. & Pivonka, E. (2000) Overview of the health benefits of fruit and vegetables consumption for the dietetics professional: selected literature. J. Am. Diet. Assoc. 100:1511-1512.[Medline]
16. Black, A. E., Cole, T. J., Wiles, S. J. & White, F. (1983) Daily variation in food intake of infants from 2 to 18 months. Hum. Nutr. Appl. Nutr. 37A:448-458.[Medline]
17. Berti, P. R. & Leonard, W. R. (1998) Demographic and socioeconomic determinants of variation in food and nutrient intake in an Andean community. Am. J. Phys. Anthropol. 105:407-417.[Medline]
18. Oh, S. Y. & Hong, M. H. (1999) Within- and between-person variation of nutrient intakes of older people in Korea. Eur. J. Clin. Nutr. 53:625-629.[Medline]
19. Launer, L. J., Dardjati, S., Kusin, J. A. & Reed, G. F. (1991) Patterns of variability in the nutrient intake of nutritionally vulnerable pregnant women. Eur. J. Clin. Nutr. 45:131-138.[Medline]
20. Littell, R. C., Milliken, G. A., Stroup, W. W. & Wofinger, R. D. (1996) SAS System for Mixed Models 1996 SAS Institute Inc Cary, NC, USA.
21. Gray-Donald, K., Johnson-Down, L. & Starkey, L. J. (2000) Food habits of Canadians: reduction in dietary fat intake in a generation. Can. J. Public Health 91:381-385.[Medline]
22. Health and Welfare Canada (1992) Canada’s Food Guide to Healthy Eating 1992 Minister of Supply and Services Ottawa, Canada.
23. Armitage, P. & Berry, G. (1994) Data editing. Statistical Methods in Medical Research 1994:386-401 Blackwell Scientific Publications Oxford, UK.
24. Willett, W. (1998) Nature of variation in diet. Nutritional Epidemiology 1998:33-49 Oxford University Press New York, NY.
25. Fraser, G. E., Welch, A., Bingham, S. A. & Day, N. E. (2000) The effect of age, sex and education on food consumption of a middle-aged English cohort-EPIC in East Anglia. Prev. Med. 30:26-34.[Medline]
26. Dallongville, J., Marécaux, N., Fruchart, J. C. & Amouye, P. (1998) Cigarette smoking is associated with unhealthy patterns of nutrient intake: a meta-analysis. J. Nutr. 128:1450-1457.
27. Roos, E., Prattala, R., Lahelma, E., Kleemola, P. & Pietinen, P. (1996) Modern and healthy? Socioeconomic differences in the quality of diet. Eur. J. Clin. Nutr. 50:753-760.[Medline]
28. Jacobsen, B. K. & Thelle, D. S. (1988) Risk factors for coronary heart disease and level of education The Tromso Heart Study. Am. J. Epidemiol. 127:923-932.
29. Billson, H., Pryer, J. A. & Nichols, R. (1999) Variation in fruit and vegetable consumption among adults in Britain. An analysis from the dietary and nutritional survey of British adults. Eur. J. Clin. Nutr. 53:946-952.[Medline]
30. Van Staveren, W. A., Deurenberg, P., Burema, J., De Groot, L. C. & Hautvast, J. G. (1986) (1986) Seasonal variation in food intake, pattern of physical activity and change in body weight in a group of young adult Dutch women consuming self-selected diets. Int. J. Obes. 10:133-145.[Medline]
31. Sempos, C. T., Liu, K. & Ernst, N. D. (1999) Food and nutrient exposures: what to consider when evaluating epidemiologic evidence. Am. J. Clin. Nutr. 69:1330S-1308S.
This article has been cited by other articles:
|
|
 |
|
  I. Huybrechts, D. De Bacquer, B. Cox, E. H. Temme, H. Van Oyen, G. De Backer, and S. De Henauw Variation in energy and nutrient intakes among pre-school children: implications for study design Eur J Public Health, October 1, 2008; 18(5): 509 - 516. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. J McCabe-Sellers, S. Bowman, J. E Stuff, C. M Champagne, P. M Simpson, and M. L Bogle Assessment of the diet quality of US adults in the Lower Mississippi Delta Am. J. Clinical Nutrition, September 1, 2007; 86(3): 697 - 706. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. H. Ledikwe, H. M. Blanck, L. K. Khan, M. K. Serdula, J. D. Seymour, B. C. Tohill, and B. J. Rolls Dietary Energy Density Determined by Eight Calculation Methods in a Nationally Representative United States Population J. Nutr., February 1, 2005; 135(2): 273 - 278. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Jahns, A. Carriquiry, L. Arab, T. A. Mroz, and B. M. Popkin Within- and Between-Person Variation in Nutrient Intakes of Russian and U.S. Children Differs by Sex and Age J. Nutr., November 1, 2004; 134(11): 3114 - 3120. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
