Self-Reported Dietary Supplement Use Is Reproducible and Relatively Valid in the Cancer Prevention Study-3 Diet Assessment Substudy

• Published in: Journal of the Academy of Nutrition and Dietetics 2022-09, Vol.122 (9), p.1665-1676.e2
• Main Authors: Hartman, Terryl J., Wang, Ying, Hodge, Rebecca A., Mitchell, Diane C., Flanders, W. Dana, Li, Chengchen, Sampson, Laura, Troeschel, Alyssa N., Patel, Alpa V., McCullough, Marjorie L.
• Format: Article
• Language: English
• Subjects: Dietary supplements; Food frequency questionnaire; Reproducibility; Validity
• ISSN: 2212-2672; 2212-2680
• DOI: 10.1016/j.jand.2021.07.006

Dietary supplement use is common, particularly among cancer survivors and those at increased risk for cancer. The objectives of this study were to assess 1-year test–retest reproducibility of dietary supplement use reported via food frequency questionnaire (FFQ-1 vs FFQ-2) and relative validity in comparison to repeated 24-hour dietary recalls (FFQ-2 vs DRs). This ancillary study was conducted within a large prospective cohort, the American Cancer Society’s Cancer Prevention Study-3. Between 2015 and 2016, 684 participants in the United States (64% women; 62% non-Hispanic White, 23% non-Hispanic Black, and 15% Hispanic) completed two FFQs and up to six unannounced telephone interviewer-administered DRs over 1 year as part of the Cancer Prevention Study-3 Diet Assessment Substudy. FFQs queried current multivitamin-mineral supplement (≥10 components) use, frequency and dose (range) for seven supplements taken individually or as part of a complex (individual/complex) including calcium, vitamins D, C, and E, folic acid, fish oil, and glucosamine. DRs allowed exact reporting of supplement frequency and dose. Weighted κ statistics were used to evaluate reproducibility between FFQ-1 and FFQ-2 and Spearman correlation coefficients assessed agreement between supplemental nutrient amounts assessed by FFQ-2 and the average of DRs. Just more than half of the participants reported taking multivitamin-mineral supplements on the baseline FFQ. Kappa statistics for the comparison of categorical responses between FFQ-1 and FFQ-2 were 0.67 for multivitamin-mineral supplements. Kappas for individual/complex supplements ranged from 0.47 for folic acid to 0.74 for vitamin D, with a mean of 0.64. Results were similar between men and women. Spearman correlation coefficients comparing FFQ-2 with the average of DRs (validity) for nutrient intakes from all sources ranged from 0.65 (fish oil for women) to 0.77 (vitamin D for men and calcium for women); results were similar among men and women. These findings suggest the FFQ used in Cancer Prevention Study-3 has good reproducibility over 1 year and yields estimates comparable to a more detailed assessment for commonly consumed dietary supplements.