Skip to main content
NUTRITION 2024 Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back

Session: Nutritional Assessment and Methods (Oral 22)

(OR22-03-24) Generalizability and External Validation of Feeding-Study-Derived Macronutrient Biomarkers in Other Feeding Studies

Monday, July 1, 2024
10:24 AM – 10:36 AM CT
Location:  W375D
Objectives: Epidemiologic studies rely on dietary self-report, which is prone to measurement error. Previously, we used metabolomics to identify intake biomarkers and calibrate food frequency questionnaires in Women’s Health Initiative (WHI). Here we evaluate their transferability to other populations.

Methods: WHI Nutrition and Physical Activity Assessment Study (NPAAS) was a controlled 2 week feeding study of habitual diet in 153 women to develop intake biomarkers. Fasting blood and urine at the end of study were analyzed by LC-MS/MS and lipidomics in serum, plus GC-MS and NMR in urine, to develop metabolite signatures for macronutrient intakes (protein, carbohydrate, fiber, fat, and fatty acids). We validated these metabolites in two controlled feeding studies using the same lab and platforms: 1) low vs high glycemic load diet for 28 days each in 80 men/women (Carbohydrate And Related Biomarkers; CARB) and 2) traditional Mexican vs standard US diet for 24 days each in 53 women of Mexican descent (Comparison Of Mexican And Standard US diets; COMIDAS). In NPAAS, for each macronutrient we ran LASSO regression for variable selection; the final model was then developed through least squares on selected metabolites. Cross-validated R2 (CV-R2) was calculated by splitting data into subsets with models developed in the training set and R2 computed in the test set with CV-R2 ≥36% used as a biomarker criterion. To validate the biomarker model from NPAAS, we computed the CV-R2 of the predicted biomarker and macronutrient intake in CARB and COMIDAS. To validate selected metabolites from NPAAS, we computed the CV-R2 by re-estimating the biomarker in CARB and COMIDAS using the selected metabolites.

Results: For NPAAS, biomarkers for protein, MFA, and PFA had CV-R2 ≥0.36. CV-R2 values using coefficients developed in NPAAS and applied to CARB and COMIDAS were all < 0.36. However, the selected metabolites from NPAAS had moderate prediction power ( >0.30) for carbohydrate, fiber, fat, and SFA in the other studies.

Conclusions: We were unable to validate the macronutrient biomarkers from NPAAS in external feeding studies, likely influenced by the  standardized diet composition resulting in low intake variation in these studies. Re-estimation of the selected metabolites in the external studies had moderate predictive power and may be transferrable.

Funding Sources: US NIH: NHLBI, NCI, NIDDK, ORIP