|dc.description.abstract||Adult health problems such as obesity and diabetes have been associated with dietary intake during the pre-school years. However, due to the difficulties involved in measuring the diet of toddlers, limited data is available on their dietary intake. A food frequency questionnaire is a cost effective and easily administered dietary assessment method and is the preferred choice in many epidemiological studies. As there is no such food frequency questionnaire available for use in New Zealand toddlers, the aim of this study was to determine the ability of a 97-item food frequency questionnaire to assess the usual dietary intake over the past four weeks of New Zealand toddlers aged 12 to 24 months.
Participants were the primary caregivers and their child. Primary caregivers completed two administrations of the food frequency questionnaire for their child, approximately four weeks apart. A five-day non-consecutive weighed diet record was completed over the four weeks between questionnaires. Relative validity was assessed in 153 participants by randomly selecting either their first or second food frequency questionnaire, and comparing nutrient intakes obtained from the food frequency questionnaire with those from the diet record. Spearman correlation coefficients, cross-classification and Bland-Altman statistics were calculated. Reproducibility was assessed in 152 participants, by comparing the two administrations of the food frequency questionnaire using intraclass correlations. Validity and reproducibility were assessed for both unadjusted data, and data that was adjusted for overall fruit and vegetable intake, using a single cross-check question, as used in the Prevention of Overweight in Infancy Study (POI-adjustment) or two cross-check questions as used in the Eating Assessment in Toddlers Study (EAT-adjustment), as recommended by expert groups.
Spearman’s correlation coefficients for nutrients ranged from 0.37 for total fat to 0.66 for calcium, with a mean coefficient of 0.50. Coefficients increased minimally when the fruit and vegetable sections were adjusted for total intake. On average, 40% of participants were correctly classified into the same quartile, and only 4% were grossly misclassified. Bland-Altman statistics demonstrated that the food frequency questionnaire overestimated intakes, with mean unadjusted nutrient data being 28-78% higher than the corresponding diet record data, and mean POI adjusted intakes 12-60% higher. The single exception was vitamin C which was not significantly different from the diet record. The food frequency questionnaire exhibited good reproducibility between administrations, with POI adjusted intraclass correlations ranging from 0.65 for vitamin C to 0.75 for calcium.
The food frequency questionnaire showed adequate to good relative validity for ranking participants according to nutrient intake using POI adjusted data, and high reproducibility. Our results compare favourably with previous literature, with validity and reproducibility similar, if not higher, than previous toddler multi-nutrient food frequency questionnaires. However, consistent with previous studies, the food frequency questionnaire was not suitable to assess absolute energy or nutrient intakes in individuals, overestimating the intake of virtually all nutrients.||