Seasonal misclassification error and magnitude of true between-person variation in dietary nutrient intake: a random coefficients analysis and implications for the Japan Public Health Center (JPHC) Cohort Study

OBJECTIVE: We examine (1) the extent to which seasonal diet assessments correctly classify individuals with respect to their usual nutrient intake, and (2) whether the magnitude of true variation in intake between individuals is seasonal. These effects could lead, respectively, to bias in estimates of relative risk for associations between usual nutrient exposure and disease, and to an increase in required sample size. SUBJECTS AND SETTING: One hundred and twenty-seven families in four regions of the Japan Public Health Center (JPHC) Cohort Study. DESIGN: On average, 48 weighed daily food records were collected per family over six seasons of 1994 and 1995. RESULTS: A random slopes regression model was used to predict the correlation between seasonal and annual average intakes, and to estimate true between-person variation in intakes by season. Mean vitamin C intake was greatest in summer and autumn, and seasonal variation was attributable to the consumption of fruit and vegetables. Predicted correlations between seasonal and annual average vitamin C intake ranged from 0.62 to 0.87, with greatest correlations in summer and autumn. True between-person variation in vitamin C intake was also strongly seasonal, ranging from 45 to 78% of total variance, and was again greatest in summer and autumn. These effects were less seasonal among energy and 13 other nutrients. CONCLUSIONS: It may be possible substantially to reduce both seasonal misclassification of individuals with respect to their usual vitamin C intake, and required sample size, by asking subjects to report high-season intake of fruit and vegetables in the JPHC Study