Consumption of ultra‐processed foods and cognitive decline in the ELSA‐Brasil study: a prospective study

Abstract

BackgroundConsumption of ultra‐processed foods (UPF) has progressively increased worldwide in the last 30 years. High intake of UPF is associated with cardiovascular disease and cancer and hypothesized to induce systemic inflammation and oxidative stress. Although UPF may also be a risk factor for cognitive decline through these pathways, little is lnown about the effects of UPF on cognition. We aimed to investigate the longitudinal association between UPF consumption and cognitive decline in the ELSA‐Brasil study during 9 years of follow‐up.MethodParticipants were evaluated in three study waves (2008‐10, 2012‐14 and 2017‐19).Data on diet measured by food frequency questionnaire was categorized according to degree of food processing using NOVA. Cognitive performance was evaluated using a standardized battery of tests ‐ the immediate recall, late recall, recognition, semantic and phonemic verbal fluency and the trail‐making tests. A global composite z‐score was derived from these tests. Association of consumption of UPF (% of total daily gram intake) with cognitive performance over time was evaluated using linear mixed effects models, adjusted for sociodemographic, lifestyle, and clinical variables.ResultIn 8,160 participants (mean 50.6±8.7 y, 56% women, and 55% white) during a median follow‐up of 9 years, the mean baseline calorie intake was 2,842±982 kcal, 28% (493g) of which came from UPF (Table 1). During follow‐up time, consumption of UPF in the fourth and fifth versus the first quintile were related to a decline in the executive funtion (β = ‐0.007, 95% CI = ‐0.013; ‐0.002, p = 0.004) and memory performance (β = ‐0.004, 95% CI = ‐0.009; 0.000, p = 0.048), respectively (Table 2). No association of UPF consumption and verbal fluency was observed (Table 2).ConclusionDuring nine years of follow‐up, high consumption of UPF was associated with cognitive decline, particularly in the memory and executive funtion domains.