Abstract
Food frequency questionnaires (FFQs) are a relatively simple, cost-effective method of assessing diet in population-based studies. However, there are no up-to-date, validated, multi-nutrient FFQs available for use in New Zealand adults. To address this gap, a 143-item semi-quantitative FFQ was developed to assess habitual intake of twenty-two nutrients of New Zealand adults over the past year. The aim of this thesis was to evaluate the test-retest reliability and the relative validity of this FFQ for energy, and vitamins A, C and E.
To assess reliability, the FFQ was self-administered by New Zealand adults aged 30-59 years on two occasions nine months apart. The validity of the FFQ was assessed by comparing data from the second administration of the FFQ to two validation references: (a) data derived from an 8-day weighed diet record (8DDR) collected over a 12-month period, and (b) blood concentrations of ascorbic acid, β-carotene and α-tocopherol. Intraclass correlation coefficients were used to assess the reliability of the FFQ. Crude and energy-adjusted Spearman correlations were used for comparing the performance of the FFQ against the two validation references. Absolute agreement between the FFQ and the 8DDR was assessed using Bland-Altman method.
Dietary data were available from 135 participants, and biological data from 133 participants. Mean energy intakes derived from the first and the second administration of the FFQ were 9.8MJ and 9.5MJ, respectively, and from the 8DDR was 9.2MJ. Intraclass correlation coefficients between the FFQ replicates ranged from 0.45 for retinol to 0.72 for vitamin C. The Spearman correlation coefficient for energy between the FFQ and the 8DDR was 0.25. After energy adjustment, correlation coefficients for nutrients of interest ranged from 0.39 for retinol to 0.59 for β-carotene. Using the Bland-Altman method to assess the strength of agreement between absolute nutrient levels from the 8DDR and the FFQ, the FFQ over-estimated absolute mean intakes for most nutrients but provided ideal mean estimates for energy and retinol. The lower “limit of agreement” for energy was 52% and the higher “limit of agreement” was 194%, indicating that the FFQ could provide a single measure for energy ranging from 48% lower to 94% higher than that from the 8DDR. Correlations between the FFQ and biological references were 0.36 for vitamin C and 0.30 for β-carotene after energy adjustment and the inclusion of nutrient intakes from dietary supplements. An inverse relationship was observed between the FFQ and the biomarker for vitamin E. The relative validity of the FFQ against the biomarkers for all selected nutrients was comparable to that of the 8DDR.
These results indicate that the FFQ produces reproducible estimates of intakes for selected nutrients. As is the case for most multi-nutrient FFQs, this FFQ does not provided accurate absolute measures of intakes for the nutrients of interest. However, it fulfills the main objective of the FFQ in ranking individuals according to levels of intake for vitamins A, C and E. Further work is required to assess the ability of this FFQ in ranking individual by intake of the other sixteen nutrients that the FFQ is designed to measure.