BACKGROUND & AIMS: Current evidence suggests that some isolated polyphenols (PP) may exert promising effects for the risk of depression in young adults, however studies among older adults remain limited. The aim of the current study was to examine the prospective association between patterns of PP intake and the risk of depressive symptomatology (DS) in older adults. METHODS: The study sample was based on the Three-City (3C) Bordeaux cohort of adults aged 65 years and over and without DS at the time of recruitment. The intakes of PP, summarised into 21 PP classes, were determined using a 24-h recall combined with the Phenol-Explorer database. In addition, the patterns of PP intake were identified by a Principal Component Analysis (PCA). DS was evaluated using the Center for Epidemiologic Studies Depression Scale (CES-D) over a period of 15 years with a reassessment every 2-3 years. The incident DS was reported for CES-D score ≥16 and/or the use of antidepressant treatment. The association between the patterns of PP intake at baseline and the risk of DS was computed using multivariate random-effect logistic regression models. RESULTS: Among the 1074 participants (mean age 75.7 y, SD 4.8 y), 423 (39.4%) developed a DS during the follow-up. Distinct patterns of PP intake were identified, explaining up to 50% of the variance. The two first patterns, mainly driven by stilbenes and dihydroflavonols and by hydroxyicnnamic acid and alkylmethoxyphenols respectively, were not associated with the odds of DS. Furthermore, a higher score on the third pattern, mainly driven by monomeric flavanols and theaflavins, was associated with a significant 27% lower risk of DS over time (Odd Ratio = 0.73, 95% Confidence Interval [0.55; 0.97]). CONCLUSION: This prospective study suggested that a pattern high in monomeric flavanols and theaflavins intakes, mainly provided by tea, was associated with a reduced risk of DS in older adults. These results provide promising evidence on combined PP intakes that would require further confirmation in other samples.< Réduire