The SU.VI.MAX (SUpplémentation en VItamines et Minéraux AntioXydants) study was a population-based, double-blind, placebo-controlled, randomized trial assessing the efficacy of a daily antioxidant supplementation on the incidence of cardiovascular disease and cancer 
. The trial was launched in 1994–95 with a planned follow-up of 8 years (1994–2002). From the full SU.VI.MAX cohort (N
13,017), a total of 6,850 participants who had agreed to participate in a post-supplementation observational follow-up were recruited for the SU.VI.MAX 2 study (2007–2009) which included a neuropsychological evaluation 
The SU.VI.MAX and SU.VI.MAX 2 studies were conducted according to the guidelines laid down in the Declaration of Helsinki and were approved by the Ethics Committee for Studies with Human Subjects of the Paris-Cochin Hospital (CCPPRB n° 706 and n° 2364, respectively) and the Comité National Informatique et Liberté (CNIL n° 334641 and n°907094, respectively). Written informed consent was obtained from all participants.
Among the 6,850 adults included in the SU.VI.MAX 2 study (2007−2009), all cognitive tests were completed by a total of 4,447 individuals who were aged 45–60 y at the start of the SU.VI.MAX trial in 1994. Among them, 2,835 participants had available data on sedentary behavior and physical activity from both 2001 and 2007. From that subsample, a total of 2,612 participants had available data for all covariables of interest. Finally, 33 participants who had been confined to bed for more than 1 month during the period covered by the physical activity questionnaires were excluded, leaving 2,579 participants for inclusion in the present analyses.
Cognitive functioning assessment
At the SU.VI.MAX 2 phase, the participants were invited to take part in a neuropsychological evaluation using validated tests regarding three memory domains (lexical-semantic, episodic and working) and mental flexibility. Lexical-semantic memory was assessed by verbal fluency tasks, including a phonemic fluency task (citing words beginning with the letter P) and a semantic fluency task (naming as many animals as possible). The score for each task was the number of correct words produced during a 2-min period 
. Episodic memory was evaluated using the RI-48 test, which is a delayed cued recall test with a maximum score of 48 
. Working memory was assessed with the forward and backward digit span. One point was scored for each sequence repeated correctly, with a maximum score of 14 points for digit span forward as well as backward 
. Mental flexibility was assessed through the Delis-Kaplan trail-making test (TMT), connecting numbers and letters alternating between the two series. The score was the time in seconds needed to complete the task 
Sedentary behaviors and physical activity
Sedentary behaviors and physical activity were assessed in 2001 and 2007 using a self-administered French version of the Modifiable Activity Questionnaire (MAQ) 
The original MAQ was designed to assess sedentary and physical activity during both leisure time and work over the past 12 months 
and was the instrument used in the Diabetes Prevention Program 
. The MAQ has been validated against energy expenditure measurements using the doubly-labeled water technique, and the test-retest properties of the questionnaire have been established 
. Generally, participants were asked to report each leisure-time physical activity performed at least 10 times for at least 10 min per session over the past 12 months. The frequency and duration of each activity were also reported. For each reported physical activity, the number of hours per week was multiplied by its estimated metabolic cost expressed in metabolic equivalent tasks (MET) 
. Then, a summary score was obtained expressed in MET-h per week. In the same questionnaire, participants were asked to report their average daily time spent watching TV, using a computer, or reading, and each of these variables was expressed in min per day.
At the start of the SU.VI.MAX trial, information on gender, date of birth, education and social position was collected. Retirement status was assessed at the time of each MAQ administration. At the SU.VI.MAX 2 phase, depressive symptoms were assessed using the French version of the Center for Epidemiologic Studies Depression Scale (CES-D) and the score was modeled as a covariate 
. Tobacco use status (never, former, current smoker) and self-rated health status were collected through a self-administered questionnaire. Self-rated health status was assessed on a 5-point scale (excellent, good, fair, poor, very poor). During the clinical examination, weight was measured using an electronic scale, with participants wearing indoor clothing and no shoes. Height was measured under the same conditions with an electronic wall-mounted stadiometer. During the entire follow-up (1994–2009), the incidence of cardiovascular disease, hypertension and diabetes was documented 
Principal component analysis (PCA) was performed in order to yield summary measures accounting for the correlations among the cognitive scores, thereby maximizing the explained variance 
. These summary scores were converted into T scores (mean
10). Thus, a one-point difference in the test score corresponded to one-tenth of a SD difference. For each sedentary behavior, the original values (duration expressed in min/day) and change over time were categorized in three classes based on the respective distribution (tertiles or none/low/ high according to the median value). Body mass index (BMI) was calculated as the ratio of weight to squared height (kg/m2
). Time-dependent retirement status was computed as follows: retired at baseline, retired during follow-up, not retired at the end of follow-up.
Descriptive characteristics of the study sample are reported as mean (SD) or percentage. From the 13,017 participants included in SU.VI.MAX, those retained and those excluded from the present analyses were compared in order to assess potential selection bias. Reported P-values refer to the chi2 test or Wilcoxon-Mann-Whitney test, as appropriate. Covariance analyses were used to estimate the difference in mean (95% confidence interval, CI) cognitive scores across categories of sedentary behaviors as well as the 6-year change in sedentary behaviors using the lowest category as reference. P for trend was assessed, using linear contrast tests across categories. The initial model was adjusted only for the interval between sedentary behavior assessment and cognitive evaluation. The second model was adjusted for age (y), education (primary, secondary, university or equivalent), supplementation group (active or placebo), BMI, occupational category (unemployed, manual workers, employed, self-employed or farmers, managerial staff) and retirement status (yes/no). The third model was further adjusted for tobacco use status, BMI, depressive symptoms, history of hypertension (yes/no), history of diabetes (yes/no), and history of cardiovascular diseases (yes/no). The full final model was further adjusted for leisure-time physical activity (MET-h/week) and the remaining sedentary behaviors. The longitudinal models were similarly conducted except for an adjustment for the baseline score for each sedentary behavior. Retirement was considered a time-dependent covariable and the remaining sedentary behaviors (ie, those not modeled as the respective independent variable) in the full model were accounted for as change over time.
In an effort to assess the robustness of our results, we carried out sensitivity analyses after removing participants with the lowest cognitive performance scores (eg, below the education level-specific tenth percentile), as these participants may have modified their physical activity or sedentary behaviors following changes in cognitive abilities. Interaction terms with gender and retirement status regarding sedentary behaviors were tested.
All tests were two-sided and type I error was set at 5%. Statistical analyses were performed using SAS software (version 9.1, SAS Institute Inc, Cary, NC, USA).