Objectives: To assess the relative validity of an interactive 24-h recall for estimating mineral intakes of rural Malawian women.
Design: Repeated interactive 24-h recalls were compared with weighed records collected for the same 2 days of food intake, and for 2 days 1-2 weeks prior and subsequent to the weighed record data collection period.
Setting: Three villages in traditional authority Jalasi, Mangochi District, Malawi.
Subjects: 60 rural pregnant women.
Results: Median daily intakes of most minerals (Ca, Fe, Zn, Mn) were comparable for the two methods, but slightly overestimated for recalled (R) intakes expressed per MJ (mg/MJ) compared to weighed (W) (R vs W = Ca, 48 vs 38; Fe, 2.1 vs 1.9; Zn, 0.9 vs 0.8; Mn, 0.40 vs 0.38; P < or = 0.05). By contrast, recalled median daily intakes of energy (kJ), protein (g) fat (g) and Cu (mg) were slightly underestimated (R vs W = 6588 vs 7824; 51 vs 57; 14 vs 15; 1.3 vs 1.6, respectively; P < or = 0.05). Discrepancies were attributed primarily to inaccurate estimates of main meal food portions [R vs W = nsima (the main meal cereal style) 475 vs 557; and legume relish 171 vs 118 P < or = 0.001]. For classifying intakes into tertiles, agreement between the two methods was poor for daily intakes (Cohen's kappa < 0.40), but fair when expressed per MJ, and as a percentage of energy from food groups (Cohen's kappa > or = 0.40). Variance ratios for recall data were higher than corresponding ratios for the weighed intakes (R vs W = for energy, 4.87 vs 0.87), indicating poorer recall measurement precision.
Conclusion: Results emphasise the importance of selecting the dietary method according to the study objectives, and the nutrients required.