Nationally Representative Within-Person to Between-Person Variance Ratios of Nutrient Intakes and Their Application for Estimating Usual Nutrient Intake Distributions Using 1-Day Dietary Data in Japan

BACKGROUND: Collecting multiple-day dietary data is generally recommended to estimate usual intake distributions; however, in the National Health and Nutrition Survey (NHNS) in Japan, only 1-d dietary data are available. OBJECTIVES: This study aimed to calculate within-person to between-person variance ratios (WBVRs) of nutrient intake using multiple-day dietary data and to estimate usual nutrient intake distributions by applying the WBVRs to a 1-d dietary record (DR) of the National Health and Nutrition Survey (NHNS). Estimation robustness was also assessed. METHODS: The WBVRs were calculated from nonconsecutive 8-d paper-based DRs including 1375 males and 1422 females nationwide. Distributions of individual usual nutrient intake (food only; excluding supplements) were estimated from 1-d household-based DRs (paper-based) of 5053 males and 5652 females in the NHNS (2018-2019) by adapting the Best-Power method. We examined the influence of sex, age group, and sample size on WBVR values and the influence of WBVR values on estimated distributions. RESULTS: The WBVRs did not vary substantially by sex or age group but did vary by nutrient, ranging from 0.71 for potassium to 3.08 for n-3 (ω-3) fatty acids in males and from 0.79 for potassium to 3.77 for vitamin D in females. The estimated usual intake distributions were narrower than 1-d intake; for example, the 90th percentiles of protein intake were 103 g/d for usual and 114 g/d for crude 1-d, whereas those for vitamin A were 738 and 875 μg retinol activity equivalents/d in males. Considerable variation in the calculated WBVR values was observed when a small number of participants and survey days were randomly selected from the multiple-day DRs. WBVR variation affected estimated distributions. CONCLUSIONS: This study estimated nationally representative usual nutrient intake distributions in Japan. The WBVRs estimated in this study, derived from a large and nationwide dataset, may be the best presently available for use in future 1-d dietary surveys aimed at estimating usual nutrient intake distributions in Japan.