Reproducibility and Validity of a Semiquantitative Food Frequency Questionnaire in Men Assessed by Multiple Methods

Among 626 participants of the Men's Lifestyle Validation Study (2011-2013), we evaluated the validity and reproducibility of a self-administered 152-item semiquantitative food frequency questionnaire (SFFQ) using two 7-day dietary records (7DDRs), 4 Automated Self-Administered 24-hour dietary recalls (ASA24s), four 24-hour urine samples, 1 doubly labeled water measurement (repeated in 104 participants), and 2 fasting blood samples, collected over 15 months. Compared with 7DDRs, SFFQs underestimated energy intake, macronutrients, and sodium intake but overestimated some micronutrients. The mean of the Spearman correlation coefficients was 0.66 (range, 0.38-0.88) between 46 energy-adjusted nutrients estimated from 7DDRs and the final SFFQ, deattenuated for within-person variation in the 7DDRs. These deattenuated correlations were similar using ASA24s as the comparison. Relative to biomarkers, SFFQs underestimated energy, sodium, and protein intakes, as well as the sodium:potassium ratio. The energy-adjusted correlations between the final SFFQ and the biomarkers were slightly lower than the correlations between the SFFQ and 7DDRs. Using the method of triads to calculate validity coefficients, the median validity coefficient between SFFQ and true intake was 0.65 and 0.69 using 7DDRs and ASA24s, respectively, as the third method. These data indicate that this SFFQ provided reasonably valid estimates for a wide range of nutrients when evaluated by multiple comparison methods.