Daily, Weekly, Seasonal, Within- and Between-individual Variation in Nutrient Intake According to Four Season Consecutive 7 Day Weighed Diet Records in Japanese Female Dietitians

Objective: To study daily, weekly, seasonal, within- and between-individual variance in intake of selected nutrients and minimal days necessary for assessing true intake with a specified degree of error based on four season consecutive 7 day weighed diet records (WDRs). Subjects and Methods: We evaluated consumption of energy and 30 nutrients based on four season consecutive 7 day WDRs from 80 Japanese female dietitians in 1996-1997. We examined daily, weekly, seasonal, within- and between-individual variation in nutrient intake, relative contributions of their variances to total variance, and minimal days required to estimate a person’s nutrient intake within 10% and 20% of their true mean with 95% confidence intervals. Results: The relative contributions of variation for all nutrients by person were larger than those by day, week and season. Within-individual variances were greater than the between-individual variances. The ratios of within- vs. between-individual variances thus ranged from 1.3 - 26.9. Minimal days necessary for estimating nutrient consumption per person within 10% (20%) of the true mean with 95% confidence intervals ranged from 10-35 (3-9) days for energy and major nutrients and 15-640 (4-160) days for micro-nutrients. Conclusions: The relative contributions of variability by person were largest for all nutrients, followed by those due to sequence of days, season and day of week. Within-individual variation was greater than between-individual variation. Minimal days necessary for ascertaining major nutrients were in general fewer than micro-nutrients.


INTRODUCTION
There are several approaches for estimating consumption of nutrients, which include diet records/weighed diet records (abbreviated DRs/WDRs hereafter), 24-hour recall, food frequency questionnaires, diet history, replicate food methods and biochemical analysis 1,2).Each procedure has its own strengths and weaknesses in terms of elements of diet, time frame and the scope of the study.Some are adequate for assessing nutrient intake in the population and can be employed for ecological studies, while others are more appropriate for ranking/categorizing subjects for epidemiological research, including cohort and case-control studies.
Among available approaches, DRs/WDRs appear the most precise and are often used for quantifying consumption of foods/nutrients in defined populations and ascertaining popula-tion means.Accordingly, there are many articles reporting intake of foods and nutrients using DRs/WDRs 1, 2), Multiple/long-term data are accepted as gold standards/references for dietary surveys including in Japan 3).We have also carried out a relative validation study and observed satisfactorily higher relative validity of a semi-quantitative food frequency questionnaire (abbreviated SQFFQ hereafter) versus four season consecutive 7 day WDRs (28 day WDRs) in Japanese dietitians, as reported previously 4,5) Although information from DRs/WDRs seems accurate on the actual day of food intake, as is reported 1,2), there are variances in dietary consumption according to day, week, season and person, and many days are needed to estimate habitual intake of foods and nutrients.Here we examined intake of energy and 30 nutrients and relative magnitude of daily, weekly, seasonal, within/intra-and between/inter-individual variance, and calculated the number of days needed to evaluate an individual's nutrient intake within 10% and 20% of the true mean with a specified degree of the error on the basis of four season consecutive 7 day WDRs provided by 80 Japanese female dietitians.Few studies on daily, weekly and seasonal fluctuations on intake of foods and nutrients have been conducted in Japanese population.Furthermore, dietitians are professionals of diet and health, and information on consumption of selected nutrients may naturally be assumed to be accurate/reliable.Little research, however, has been carried out employing dietitians as study subjects, not only in Japan but also in other areas of the world.

Subjects and 28 day WDRs
We earlier evolved an evidence-based SQFFQ according to multiple regression and contribution analyses, and validated the SQFFQ against 28 day WDRs, as described elsewhere 4, 5).This study was based on the same 28 day WDRs or four season consecutive 7day WDRs.Briefly, in autumn 1996, we first mail-administered the SQFFQ to 106 (21 male and 85 female) middle-aged Japanese dietitians living in Aichi Prefecture, in Central Japan and self-administered consecutive 7 day WDRs approximately a week later, and at about 3 month intervals in winter, spring and summer 1997.
Eighty female dietitians completed the four season consecutive 7 day WDRs.Male dietitians were excluded because they were rather small in number.This investigation was therefore based on the 28 day WDRs provided by eighty female dietitians with a mean age (years of age) * standard deviation (minimum -maximum) of 48 * 8 (32 -66) in autumn, 1996.The values for height (cm), weight (kg) and BMI (kg/ma) were 156.9 * 5.1 (146.1 -172.1),52.2 * 5.5 (37.5 -69.4), and 21.5 * 2.2 (15.9 -29.5), respectively.Food weight was quantified before cooking if it was prepared at home, otherwise after cooking.Foods were generally weighed individually; while those like soup/stew were divided by the number of household members.For foods eaten out, reference was made to model recipes.For immeasurable foods, standard portion size was applied.Although WDRs were satisfactorily filled out, research nutritionists reviewed their accuracy as well as completeness.

Calculation of Intake of Energy and Nutrients
We computed average daily consumption of major and micro-nutrients by multiplying the food intake (in grams) or serving/portion size and the nutrient content per 100 grams of food as listed in the Standard Tables of Food Composition, Version 4, the Follow-up of Standard Tables of Food Composition 6. 7), Composition Table of Processed Foodstuff and Table of Trace Element Contents in Japanese Foodstuff 9).Since the composition tables are incomplete for certain foods, estimation was largely made according to analogies from same/similar foods and to assessment based on model recipes for relevant foods as reported 10)

Statistical Analysis
Using all data of 2,240 day WDRs or four season consecutive 7 day WDRs completed by 80 female dietitians, we firstly calculated mean and standard deviation along with minimum and maximum values, and compared differences in intakes of selected nutrients by day, week, season and person according to analysis of variance (ANOVA) 11).
We examined the relative contributions of daily , weekly, seasonal, between-individual variation and residual to total variation according to analysis of components of variance [11][12][13] The ratio of within-and between-individual variance with 95% confidence intervals was then computed [11][12][13].In this calculation, variation due to day, week and season was included in within-individual variation.Furthermore, we quantified minimal days required to estimate nutrient consumption within 10% and 20% of the true mean with 95% confidence intervals on the basis of within-individual variance 11-13).

Intake of Energy and Nutrients
Average and standard deviation (minimum-maximum) of daily intakes of energy and 30 nutrients are shown in Table 1.Mean daily intakes of selected nutrients by season are given in Table 2. Most nutrients, except for carbohydrate, phosphorus, vitamin D, oleic acid, * -linolenic acid and cholesterol, were significantly different by season.Remarkable seasonal differences were observed for carotenes, vitamin A, vitamin C, iron, zinc, AA, n-3 PUFAs, EPA, DHA and dietary fiber, including TDF, SDF and IDF (p<0.001).Among these, daily intakes of carotenes and vitamin C in autumn were about 20% larger than in summer.Table 3 shows results of ANOVA for intake of  Relative Contributions of Daily, Weekly, Seasonal and Personal Variances As shown in Table 4, the relative contributions of variance (%) by person were greatest for all nutrients.The contributions of residual variation of 52.3 -91.1 were larger than betweenindividual variation of 6.9 -45.3.The values for sequence by days, day of week, and by season ranged form 0.3 -1.6, 0 -0.3, and 0.1 -2.7, respectively.

Comparison Between Within-and Between-individual Variances
Within-individual variances were larger than between-individual variances (Table 5), and the ratios were greater than one.They ranged from 1.3 (0.7 -2.6) for potassium , TDF and IDF -26.9 (15.1 -53.0) for DHA.The ratios for micro-nutrients (1.3 -26.9) were generally larger than those (2.1 -3.6) for energy and major nutrients.

Number of Days Needed to Evaluate the True Intake of Energy and Nutrients
Table 5 also shows coefficient of variance with 95% confi- dence intervals, and minimal days needed to ascertain an individual's nutrient consumption within 10% (20%) of the true mean with 95% confidence intervals, which ranged from 15 -640 (4 -160) for micro-nutrients, longer than those for energy and major nutrients which ranged from 10-35 (3 -9).

DISCUSSION
As reported in the literature 12.14-22), the greatest source of variation in nutrient intake was brought about by person.Within-individual variation was greater than between-individual variation, which may be largely thanks to our free-living lifestyle in the real world, far from data for basic sciences like chemistry and physics.The extent of within-individual variability would unduly yield misclassification of subjects and inconsistencies in epidemiological observations 1, 2).Every effort should therefore be exerted to evade categorization bias in studies related to dietary surveys in addition to statistical procedures for adjusting within-individual fluctuation.
When assessing within-individual variation and the ratio of within-vs.between-individual variance, the settings for DRs/WDRs, including age and sex of the subjects, number of days, and study frames (food vs. recipe based DRs/WDRs), should clearly be taken into account.Those values for most nutrients in this study were generally similar to those measured in the general population in Japan 20,21); however, in common, they were greater than those calculated from DRs/WDRs in western countries [14][15][16][17][18][19] which may be partly due to the wide variety of the Japanese diet.
The next largest variation was derived from season 23-28).For this reason, we should keep in mind to seasonal effects in dietary studies.For example, intakes of certain vitamins and Table 4. Relative contributions of variance(%) for intake of energy and 30 nutrients .minerals in autumn and winter were found to be larger than those in spring and summer , which may be associated with fluctuations in consumption of fruit , vegetables and certain seafood because they are important sources of relevant micronutrients.Contrary to the comparison based on absolute values , the relative contribution of variance by season to the total variance were not so large as expected because the consumption of fruit and vegetables may be evenly increased in the study subjects in autumn and winter.There were differences in intakes of selected nutrients by sequence of days but not by day of week as reported in the literature 12. 14, 19) The difference by day of week was greatest for carbohydrate in this study (p<0.001),rather than sequence of days, because bigger meals were usually consumed on the weekend than on weekdays.Therefore , the information of dietary intake during weekend should not be overlooked in dietary surveys.
With our WDRs, minimal days required for evaluating a person's true intake of energy and macro-nutrients were generally fewer than for micro-nutrients in line with earlier findings 12,18,21,22) At least 10 days were needed for energy to secure values within 10% of the true mean , in contrast to one year or more for vitamin A, vitamin D, EPA and DHA .Accordingly, we should always remember that short-term DRs/WDRs invariably yield invalid/inconsistent information , when assessing the association between consumption of micro-nutrients , in particular, and h ealth/disease.The accuracy of DRs/WDRs is , as mentioned, greatest when taken on the actual day of ingestion of foods/nutrients .DRs/WDRs have been adopted for nutritional surveys in the world including Japan and used for estimating the population mean.They are employed as gold standards/references for dietary surveys; however, diet varies according to day , week, season and person.In order to precisely assess average individ -Table 5. Intake of energy and 30 nutrients, within-and between-individual variation, coefficients of variation and minimal days needed to estimate nutrient intake within 10% and 20% of the true mean with 95% confidence intervals.
ual habitual consumption of foods/nutrients, especially for certain micro-nutrients, we need long-term DRs/WDRs due to within-individual variation as described above.It is obviously very laborious to keep multiple DRs/WDRs.Thus, it does not seem pertinent to administer multiple DRs/WDRs, particularly to the general populace, but rather apt for FFQs/SQFFQs instead to ascertain long-term dietary intake.Although less precise/less quantifiable than DRs/WDRs, FFQs/SQFFQs may be suitable for case-referent and prospective studies.In this sense, we developed a data-based SQFFQ and examined relative validity of the SQFFQ versus 28 day WDRs as well as its reproducibility.Using the validated/modified SQFFQ, we proposed the JADE (Japanese Dietitians' Epidemiologic) Study and self-administered SQFFQ to

Japanese dietitians in autumn
1999, 2000 and 2001, which is worldwide a new approach, to elucidate relationships between diet/lifestyle and health/disease.ACKNOWLEDGMENTS This study was partly supported by A Grant-in-Aid from the Ministry of Education, Science, Sports and Culture (06454242).The authors are grateful to the members of the Aichi Prefectural Dietetics Association for their participation in this study, and to Ms. Y. Kubo, Ms. Y. Ito, Ms. K. Nishiyama and Dr. M.A. Moore for their technical and language assistance in preparing this manuscript.

Table 2 .
Average daily intake of energy and 30 nutrients by season.

Table 3 .
Results of analysis of variance for intake of energy and 30 nutrients.