Validation and reproducibility of an FFQ for use among adults in Botswana

Objective: To evaluate the validity and reproducibility of a 122-item interviewer-administered quantitative FFQ developed to determine food and nutrient intakes of adults in Botswana. Design: Relative validity of the FFQ was evaluated by comparing nutrient and food group intakes against four non-consecutive 24 h recalls administered over 12 months. The FFQ was repeated after 1 year to assess reproducibility. Setting: Kanye, Botswana. Subjects: Seventy-nine adults aged 18-75 years. Results: Spearman correlation coefficients for the validity of energy-adjusted nutrients ranged from 0.42 (carbohydrate) to 0.49 (protein) for macronutrients and from 0.23 (Fe) to 0.44 (PUFA) for micronutrients. Exact agreement of quartile distribution for nutrients between the FFQ and recalls ranged from 27% to 72%. Weighted kappa values were lowest for retinol (0.13), Fe (0.22) and beta-carotene (0.25) and ranged from 0.33 (SFA) to 0.59 (folate) for other nutrients (energy, carbohydrate, protein, fat, Ca and vitamin E). Spearman correlation coefficients between the recalls and FFQ for food groups ranged from 0.18 (dark green leafy and yellow vegetables) to 0.58 (poultry). Reproducibility correlation coefficients (energy-adjusted) varied between 0.39 for retinol and 0.66 for vitamin E, with most values falling between 0.50 and 0.60. Conclusions: The FFQ had good relative validity for estimating habitual food group and nutrient intakes, but was poor for some micronutrients (Fe, retinol and beta-carotene) and foods (fruits and dark green leafy vegetables).