Objective: FFQ are often used to estimate food and nutrient intakes to rank individuals by their level of intake. We evaluated the relative validity of a semi-quantitative FFQ created for use in Tanzania by comparing it with two 24 h diet recalls.
Design: We measured relative validity of the FFQ with deattenuated energy-adjusted rank correlations for nutrients, deattenuated rank correlations for food groups, and performed a cross-classification analysis of energy-adjusted nutrient quartiles using percentage of agreement and Bland-Altman analysis.
Setting: Interviews were conducted in 2014 in participants' homes in Ukonga, Dar es Salaam, Tanzania.
Subjects: We surveyed 317 adults aged 40 years or older from the general public.
Results: Deattenuated energy-adjusted rank correlation coefficients of nutrients ranged from -0·03 for riboflavin to 0·41 for percentage of energy from carbohydrates, with a median correlation of 0·21. Coefficients for food groups ranged from 0·00 for root vegetables to 0·51 for alcohol, with a median of 0·35. Relative to the average of the two 24 h diet recalls, the FFQ overestimated energy intake and intakes of all nutrients and food groups, other than tea, with ratios among nutrients ranging from 1·34 for SFA to 7·08 for vitamin A; and among food groups from 0·92 for tea to 9·00 for fruit. The percentage of participants classified into the same nutrient intake quartile ranged from 23 % for SFA to 32 % for both niacin and pantothenic acid, with a median of 28 %.
Conclusions: The FFQ performed moderately well in urban Tanzanian adults.
Keywords: FFQ; Nutritional epidemiology; Sub-Saharan Africa; Tanzania; Validation.