We recently evaluated associations of
biomarker-calibrated
protein intake,
protein density,
carbohydrate intake, and
carbohydrate density with the incidence of
cardiovascular disease,
cancer, and diabetes among postmenopausal women in the Women's Health Initiative (1993-present, 40 US clinical centers). The
biomarkers relied on serum and urine metabolomics profiles, and
biomarker calibration used regression of
biomarkers on food frequency questionnaires. Here we develop corresponding calibration equations using food records and dietary recalls. In addition, we use calibrated intakes based on food records in disease association estimation in a cohort subset (n = 29,294) having food records. In this analysis, more
biomarker variation was explained by food records than by FFQs for absolute macronutrient intake, with 24-hour recalls being intermediate. However, the percentage of
biomarker variation explained was similar for each assessment approach for macronutrient densities. Invasive
breast cancer risk was related inversely to
carbohydrate and
protein densities using food records, in analyses that included (calibrated) total energy intake and body mass index. Corresponding analyses for absolute intakes did not differ from the null, nor did absolute or relative intakes associate significantly with
colorectal cancer or
coronary heart disease. These analyses do not suggest major advantages for food records or dietary recalls in comparison with less costly and logistically simpler food frequency questionnaires for these nutritional variables.