| Summary: | Abstract Background Nutrition interventions typically rely on self-reported intake that may be susceptible to differential reporting bias due to exposure to the intervention. Such differences may result from increased social desirability, increased attention to eating or improved recall accuracy, and may bias estimates of the intervention effect. This study investigated differential reporting bias of fruit and vegetable intake in youth with type 1 diabetes participating in a randomized controlled trial targeting increased whole plant food intake. Methods Participants (treatment n = 66, control n = 70) completed 3-day food records at baseline, 6-,12-, and 18-months, from which fruit and vegetable intake (servings/day) was calculated. Serum carotenoids were assessed at these visits using a high-performance liquid chromatography-based assay. Linear regression estimated associations of fruit and vegetable intake with serum carotenoids by treatment assignment. Multiplicative interaction terms tested the interaction of treatment assignment with fruit and vegetable intake on serum carotenoids for each visit and within each group over time. Results The association of fruit and vegetable intake with serum carotenoids was significantly lower in the control versus intervention group at baseline (β = 0.22 Vs 0.46) and 6-month visits (β = 0.37 Vs 0.54), as evidenced by significant interaction effects. However, the association of fruit and vegetable intake with serum carotenoids did not significantly differ over time for either group. Conclusions While the stronger association of fruit and vegetable with carotenoids in the treatment arm suggests greater reporting accuracy, this difference was evident at baseline, and did not change significantly over time in either group. Thus, results indicate greater subject-specific bias in the control arm compared to the treatment, and lack of evidence for reactivity to the intervention by treatment assignment. Clinical trial registry number and website NCT00999375
|
|---|
