Reliability and validity of a semi-quantitative, short food frequency questionnaire for assessing macronutrient and free and added sugars intake in New Zealand adults

Background: Global rates of non-communicable disease have risen at an unprecedented rate, leading to large investment into the research of diet-disease association. Food frequency questionnaires (FFQs) are a cost-effective nutritional assessment method with relatively low respondent burden; which makes them ideal for use in epidemiological studies. In New Zealand, there are few up-to-date multi-nutrient FFQs, and none have been validated for free and added sugar intake. Objective: To evaluate the reliability and validity of a semi-quantitative food frequency questionnaire to assess typical intake of ten macronutrients over the previous three months in New Zealand adults. Design: Healthy adults aged 18-70 years were recruited from the Dunedin population. Data from 65 participants were included in the analysis. Participants completed the FFQ twice, nine to ten weeks apart. Seven day weighed diet records (7DDR) were used as the reference method for validation, and were collected once per week for seven consecutive weeks in the period between the FFQ administrations. Validation was assessed using Spearman correlations, Bland-Altman analysis, and cross-classification. Spearman correlations and cross-classification were also used to assess test-retest reliability. Results: The FFQ demonstrated good test-retest reliability with Spearman correlation coefficients ranging from 0.51 to 0.69, and gross misclassification by tertile no greater than 9.2%. The validity of the FFQ was acceptable, with Spearman correlations ranging from 0.44 for carbohydrate to 0.58 for total fat and monounsaturated fat. The highest rate of gross-misclassification between the FFQ and 7DDR was 12.5% for carbohydrate. Bland-Altman analysis produced mean percentage agreements ranging from 79 to 96, and ultimately showed that the FFQ statistically significantly underestimated energy, fat, saturated fat, monounsaturated fat, polyunsaturated fat, and free sugar compared to the 7DDR. Conclusion: The results of this study suggest that the FFQ can produce reproducible estimates of nutrient intake for all selected nutrients. It has acceptable validity for ranking individuals by nutrient intake, but like most multi-nutrient FFQs, it is not appropriate for estimating absolute nutrient intake. Further work is required to assess the validity of the FFQ in ethnic-specific or disease-specific populations