Comparison of Polyphenol Intakes in Relation to Dietary Patterns and Food Sources in the Adventist Health Study-2 Cohort

Evidence suggests a relationship between polyphenol intake and health benefits. Polyphenol intake among a large United States cohort with diverse dietary practices ranging from meatless to omnivorous diets has not been previously evaluated. The primary aim of this study was to compare polyphenol intakes of several vegetarian and non‐vegetarian dietary patterns and to assess phenolic intake by food source. To characterize dietary intake, a food frequency questionnaire was obtained from 77,441 participants (mean age 57.1 years, 65% female) of the Adventist Health Study‐2 (AHS‐2). Dietary patterns were defined based on the absence of animal food consumption as vegan, lacto‐ovo‐, pesco‐, semi‐ and non‐vegetarian. Polyphenol intakes were calculated using the AHS‐2 Polyphenol Food Composition Database, which was developed using chromatography‐derived data from Phenol‐Explorer, US Department of Agriculture databases, and relevant literature. The mean total polyphenol intake in all subjects was 801 ± 356 mg/d and the main foods contributing to polyphenol intakes were coffee, fruits and fruit juices. Total polyphenol intake differed significantly between dietary patterns, with phenolic acids from coffee contributing the greatest variation. The dominant classes and sources of dietary polyphenols differed between vegetarian and non‐vegetarian diets. Among all subjects, mean total polyphenol intake was highest among non‐vegetarians (946 ± 924 mg/d), followed by pesco‐vegetarians (754 ± 629 mg/d) and was lowest in vegans (593 ± 415 mg/d). In the non‐coffee consuming subgroup, pesco‐vegetarians had the highest total polyphenol intake (613 ± 398 mg/d), followed by vegans (568 ± 343 mg/d), non‐vegetarians (532 ± 398 mg/d), lacto‐ovo‐vegetarians (528 ± 329 mg/d) and semi‐vegetarians (488 ± 334 mg/d). Flavonoid intake was highest in pesco‐vegetarians (336 ± 243 mg/d) and phenolic acid intake was highest in non‐vegetarians (548 ± 887 mg/d). In conclusion, coffee consumers appear to have a different dietary profile than non‐coffee consumers, including a reduced proportionate contribution of fruits, vegetables and legumes to total phenolic intake. Further evaluating these differences may be important in identifying relationships between plant‐based diets and health outcomes. Support or Funding Information This work was supported by National Cancer Institute grant 1U01CA152939, the McLean Endowment, and the Hulda Crooks Grant. The funding agencies had no role in the design and analysis of the study or in the writing of this abstract. Percentiles of unadjusted total polyphenol intake stratified by dietary pattern Percentiles of total polyphenol intake (mg/d) n 5 th percentile 50 th percentile 95 lh percentile Dietary pattern All 77441 176 556 2339 Vegan 5870 202 497 1288 Lacto‐ovo‐vegetarian 22372 189 498 1575 Pesco‐vegetarian 7604 202 580 1920 Semi‐vegetarian 4298 159 514 2066 Non‐vegetarian 37297 162 635 2847 Estimated mean total dietary polyphenol intake (mg/d) in coffee consumers and non‐coffee consumers in the Adventist Health Study‐2 cohort stratified by diet pattern (mean values with 95% confidence intervals) Total polyphenol intake (mg/d), mean (95% Cl) Entire cohort Coffee consumers Non‐coffee consumers Dietary pattern n Mean 95% Cl n Mean 95% Cl n Mean 95% Cl All 74,668 562 559, 566 23,563 984 963, 1006 51,105 456 453, 458 Vegan 5,677 498 488, 507 218 954 870, 104 5,459 492 483, 501 Lacto‐ovo‐vegetarian 21,666 504 499, 509 3,720 929 907, 951 17,946 463 458, 467 Pesco‐vegetarian 7,318 589 580, 599 1,713 949 917, 981 5,605 484 476, 492 Semi‐vegetarian 4,154 573 560, 586 1,495 980 946, 1016 2,659 432 423, 442 Non‐vegetarian 35,853 662 657, 668 16,417 1119 1106, 1132 19,436 412 409, 416 ANCOVA was used to determine differences in energy‐adjusted total dietary polyphenol intake between diet groups with covariates age, gender, race, BMI level, and physical activity level. Bonferroni adjustment for multiple comparisons. Mean values within a column with unlike superscript letters indicate that pairwise comparisons were significantly different (p < 0.01).