The American Cancer Society Cancer Prevention Study-3 FFQ Has Reasonable Validity and Reproducibility for Food Groups and a Diet Quality Score

FFQs are commonly used to assess dietary intake and it is important to evaluate their performance in the target population. We evaluated the reproducibility and relative validity of the Cancer Prevention Study-3 (CPS-3) FFQ in estimating usual intake of 63 food groups and diet quality in accordance with the American Cancer Society dietary guidelines for cancer prevention. Methods: A subset of participants from the CPS-3 (433 women, 244 men), 31-70 y of age, were included in a cross-sectional diet assessment substudy (2015-2016). Reproducibility was assessed by comparing estimates from repeat FFQs, approximately 1 y apart, using Spearman correlation coefficient (r_s) and Pearson correlation coefficient (r_p) correlations for food groups and diet quality, respectively. Validity was assessed similarly by comparing FFQ estimates with estimates from ≤6 interviewer-administered 24-h dietary recall (24HR). Analyses were stratified by sex and race/ethnicity. Results: Reproducibility correlations for repeated FFQs were > 0.50 for 83-97% of food groups analyzed across strata of sex and race. Although participants tended to overreport plant foods (e.g., fruits and legumes) and underreport refined grains and sugar-sweetened beverages, the median energy-adjusted, deattenuated Spearman correlations comparing the second FFQ to the 24HR were 0.50 and 0.52 among men and women (range: 0.05-0.82), respectively, suggesting that ranking was preserved for most food groups. Validity was highest for coffee, alcohol, and total dairy, and lowest for pasta and regular-fat yogurt. Median validity across food groups varied by race/ethnicity and was highest among whites (r_s = 0.54) followed by Hispanics (r_s = 0.49) and African Americans (r_s = 0.45). The diet quality score had good validity in all subgroups examined, but was higher among men (r_p = 0.69) than women (r_p = 0.61), and lower among whites (r_p = 0.62) than Hispanics (r_p = 0.64) or African Americans (r_p = 0.73). Conclusions: This study indicates good reproducibility and validity of the CPS-3 FFQ for most major food groups and the diet quality score in all sex and race/ethnicity groups examined.