(P19-003-23) Updates to the Dietary Supplement Ingredient Database (DSID)

Objectives: The DSID provides online calculators to convert labeled ingredient information into analytically predicted amounts, permitting more accurate calculation of micronutrients and other ingredients intake from DS. To further improve DSID estimates, the definition of MVM was modified and regression analyses generating predicted content of MVMs now incorporate data from market share estimates calculated from NHANES DS reports.

Methods: To harmonize with the MVM definition used in the Dietary Supplement Label Database and by nutrition researchers, the new DSID definition of MVM requires the presence of at least one mineral in a product, in contrast to the old definition where minerals were optional. Data and mixed model regression analyses were updated for the children’s MVM study. For the dataset with serving sizes for four years and older, a total of 14 of 64 products were removed. Market share estimates (weights) in regression analyses were adjusted accordingly. In the SAS code for the latest DSID release, the WEIGHT statement used with a RANDOM statement applied the weights to model variance but not fixed effects (% difference from label). The new code uses REPEATED statements to fit the supplement variance component and residuals and computes weighted estimates of the fixed effects. This standardized approach will be applied in all nationally representative DSID studies.

Results: The impact of these changes was evaluated by comparing the new weighted to new unweighted results, and to original regressions. For folic acid, vitamins A, B-12 and E, iron, and phosphorus, the application of weights changed the model type. Overall, the changes in predicted % differences from label at the most commonly reported labeled level ranged from +9.4 to -6. For 9 ingredients (niacin, riboflavin, thiamin, vitamin B-6, copper, iron, magnesium, phosphorus and zinc), the weight application did not change the model type: the % differences from label at the most common level were all between 0-10% above label.

Conclusions: Changing the definition of MVM and applying the NHANES weights had significant impacts on six ingredients in children MVMs by changing the model types. Less impact was seen for the other 9 nutrients. These intake estimates can be combined with nutrient intakes from foods to better estimate total dietary intakes.

Funding Sources: NIH/ODS and USDA