Dietary self-reports may or may not usefully complement blood concentrations for micronutrients

From the Nutrition and Physical Activity Assessment Group

The present article is a follow-up study from the Nutrition and Physical Activity Assessment Group in the Public Health Sciences Division. In the first publication,  Dr. Prentice and the Group evaluated and reported on novel biomarkers for the intake of four micronutrients (α-carotene, β-carotene, lutein plus zeaxanthin, and α-tocopherol) which are also analyzed in the current study. These reports build on a human feeding study that aimed at approximating a participant’s typical diet over a 2-week period for 153 Women’s Health Initiative (WHI) participants from the Seattle area. In the first follow-up article, the Nutrition and Physical Activity Assessment Group enhanced the biomarker equations to include potential confounding factors for outcomes such as cardiovascular disease (CVDs), cancers, and diabetes. These were then applied in a national WHI subcohort consisting of 5488 women for whom blood concentrations for the four micronutrients were measured routinely and demonstrated significant inverse associations of the three carotenoids with these chronic diseases, as well as a direct association of α-tocopherol with certain vascular diseases.  However, the number of participants the first study was limited, so in the second study the research group wanted to examine the same associations in larger WHI cohorts that included about 80,000 participants having self-reported diet, but not these blood concentrations. It’s possible that the larger set of participants would increase the precision of the measurements, in spite of limitations of self-reported diet.

Various cohorts were used in this second study.  In the WHI Dietary Modification (DM) trial, 48,835 participants were randomly selected; 29,294 were a part of the usual-diet comparison group (DM-C) and 93,676 participants were enrolled in the WHI observational study (OS).  In the DM trial, Food Frequency Questionnaires, FFQs, were administered at baseline, year one, and every three years during the trial intervention period. In the OS, FFQs were administered at baseline and during year three of the study.

Subsets of the WHI Clinical Trial and OS were selected for the baseline Measurement Precision Study. A total of 5,488 baseline participant serum samples were analyzed to calculate the 4 micronutrient concentrations and other analytes. The measurements were used to establish calibration equations that yield micronutrient intake estimates from the following: FFQs, dietary supplement reports, and participant characteristics only. The Nutrition and Physical Activity Assessment Study (NPASS) Nutritional Biomarker Study consisted of 450 OS participants. Fourteen participants of NPAAS enrolled in the NPAAS Feeding Study. The data from the remaining 436 OS participants are utilized for equation development for the four micronutrient concentrations. Last, the NPAAS Feeding Study (FS) enlisted 153 women in the Seattle area. FFQs were collected before the feeding period. In this study, serum nutrient concentrations were assessed for micronutrient intakes. The Measurement Precision Study (n=5,488) and the NPAAS Nutritional Biomarker Study (n=436) used serum aliquots from fasting blood samples.

After FFQs and serum concentrations were collected from various cohorts and subset of cohorts, the study resources provided the research group the possibility to address two questions.  The recent findings of the publication are located in the American Journal of Clinical Nutrition. First question: can FFQ dietary data explain supplementary variation in micronutrient intake in the NPAAS FS – variation not explained by serum nutrient concentrations and participant characteristics? Second question: Can calibrated intake procedures be established in the NPAAS Nutritional Biomarker Study (n=436) and Measurement Precision subcohort (n=5,488) utilizing FFQ data for micronutrients and participant characteristics – using equations developed in NPAAS FS? The co-primary outcomes in the study consisted of coronary heart disease, CHD, and invasive breast cancer. 

Graphical representation Photos courtesy of USDA Agricultural Research Service
Photo courtesy of USDA Agricultural Research Service Image from Dr. Prentice

There are significant associations (p< between 0.05) between the NPASS FS intake and FFQ dietary intake before the feeding period for β-carotene and lutein plus zeaxanthin (L+Z). There is no association between α- carotene and α-tocopherol.  R square values, which indicates variation in the model, were in the 25-75% range for the relationship between four micronutrients, outcomes (CHD, cancer, diabetes), and covariates. There was more variation between the models from the NPAAS Nutritional Biomarker Study than for the Measurement Precision subcohort analyses.  The biomarker analyses show reductions in various outcomes (coronary artery bypass graft, invasive breast cancer, and diabetes) with α and β carotene. The biomarker analyses also showed reductions in stroke and diabetes with (L+Z); there was an increase in percutaneous coronary intervention (CABG/PC with α-tocopherol. However, disease association analyses did not agree closely between those using the stronger intake measure in the Measurement Precision Study (n=5488), with those relying on FFQ-based intake estimates without the serum concentration measurements. The lack of agreement was thought to arise from FFQ measurement errors which were not sufficiently corrected by the calibration procedure.   

The researchers summarized its findings by reporting that carotenoid intake based on serum concentrations and participant characteristics had moderately significant associations with CVD, cancer, and diabetes; α-tocopherol was associated with an increased risk for certain CVD outcomes utilizing the WHI Measurement Precision subcohort. The research group commented that in recent years, concentrations for quantitative intake have been considered in addition to serum concentrations to measure the risk of chronic disease. This study indicates that nutritional variables should be rescaled to correspond to intake and the need for rescaling is dependent upon the participant characteristics.

By answering the first research question, the research group learned that if intake measures without self-reported data is acceptable biomarker criteria and is available throughout the study’s cohort, it may be best to use objective intake measures without FFQ data.  For the second question, the team learned that if FFQ data explained variation in the FS beyond that of serum concentrations and covariates, then FFQ data is needed in the intake assessment to avoid bias in the calibrated intake and disease association analyses. Per contra, the reliability of the calibrated intake and disease association analyses is dependent upon the quality of the biomarker assessment from the feeding study, and the quality of the calibration equation which produced the intake estimates from FFQ, supplementary data, and participant characteristics only, the latter of which may be inadequate in the larger cohort analyses.

In conclusion, the analyses in the study stresses the importance of critically evaluating research methods for intake assessment in nutritional epidemiology.  Intake biomarkers could strengthen study results and limit bias within the analyses.  It’s encouraged to evaluate intake assessment using feeding studies for biomarker identification, rather than just relying on self-reported dietary data in cohort studies.  

This research was supported by the National Heart, Lung, and Blood Institute and National Cancer Institute.

Fred Hutch/UW Cancer Consortium members: Ross Prentice, Marian Neuhouser, Ying Huang, Garnet Anderson, Johanna W Lampe

Prentice RL, Pettinger M, Neuhouser ML, Tinker LF, Huang Y, Zheng C, Manson JE, Mossavar-Rahmani Y, Anderson GL, Lampe JW. Can dietary self-reports usefully complement blood concentrations for estimation of micronutrient intake and chronic disease associations? The American Journal of Clinical Nutrition. 2020 Mar 4.