A Method that Maintains Accuracy in the Prediction of Vitamin A Total Body Stores when Population-Based Modeling of a Limited Number of Theoretical Subjects Is Used with Retinol Isotope Dilution

Retinol isotope dilution (RID) equations are used to predict vitamin A total body stores (TBS). Including population-based ("super-subject") modeling with RID provides group-specific values for the equation coefficients. Objective was to test an approach that would accommodate a limited super-subject sample size without compromising accuracy in RID predictions of TBS. We used Simulation, Analysis and Modeling software to simulate fraction of dose in plasma (FD ) at 16 times from 3 h-56 d after tracer ingestion in 20 theoretical adults. Then we modeled geometric mean FD ("full dataset") to determine group mean TBS and the coefficients Fa (FD in stores) and S (specific activity in plasma/stores) in the RID equation TBS (μmol) = FaS/plasma retinol specific activity. Using the same FD data, we also generated four datasets with reduced subject numbers at times other than that designated for RID (d 21). Then, we adjusted individual FD using the ratio (individual FD on d 21/mean FD on d 21) ("adjusted datasets"), modeled each, and determined TBS and FaS for comparison to the full dataset values. Mean ratio of model-predicted TBS for adjusted/full dataset was 0.962 (range, 0.920-1.06) and for FaS, it was 0.945 (d 14), 0.971 (d 21), and 0.984 (d 28). For these theoretical data, adjusting individual FD values based on relationship to the group mean FD at an appropriate time (21 d) maintained the accuracy of model predictions of TBS and the RID composite coefficient FaS. If these results are confirmed using real data, values for FaS determined in a small super-subject study could be applied to confidently predict TBS by RID in that group's individuals. This approach would be especially useful when resources are limited for studies of vitamin A status in community settings.