Background It is difficult to determine the nature of the association between social relationships(SR) and cognitive capability, as different aspects of the former likely have different effects on the latter. Moreover, whether SRs are associated solely with cognitive performance or also with rate of decline remains unclear. We aim to examine associations between structural (marital/cohabitation status, network size) and interactional (frequency of contact, social participation, positive and negative social support) aspects of SR with average and change in measures of memory, executive function and processing speed over time.

Methods We used data from the MRC National Survey for Health and Development (NSHD,n=2,109) and the English Longitudinal Study of Ageing(ELSA,n=8460). Participants from NSHD aged 53y at baseline were followed up three times over 16y. Participants from ELSA aged 64.3y(9.8SD) at baseline were followed up 9 times over 17y. We applied multilevel models to examine each SR variable at baseline with each standardised measure of cognitive performance. We first examined average performance adjusting for time, then rate of decline by including an interaction term between SR and time. Models were adjusted for age at baseline, sex, social class, education, limitations in activities of daily living, and health conditions. We adjusted for depressive symptoms in sensitivity analyses.

Results In NSHD, having more positive support or less negative support was associated with higher memory scores(β:0.05[95%CI:0.02–0.09]; β0.06[95%CI:0.02–0.09]), but not processing speed(executive function not measured).We did not observe associations between the other measures of SR and cognitive performance in NSHD. In ELSA, being married (β:0.04[95%CI:0.01–0.08]), having a larger network size (β:0.12[95%CI:0.02–0.21] >6 vs. zero), more frequent contact (β:0.25[95%CI:0.08–0.24] 1–2/week vs <1/year), greater social participation (β:0.31[95%CI:0.28–0.35] high vs low), more positive support (β:0.02[95%CI:0.01–0.03]), or less negative support (β:0.04[95%CI:0.02–0.05]) was associated with higher memory scores. Similar associations were observed for executive function but not processing speed. We did not observe any associations between SRs and decline in memory or processing speed in NSHD. In ELSA, being married and greater social participation were associated with less decline in memory and executive function but not processing speed.

Discussion Preliminary results suggest a relationship between both structural and interactional aspects of SRs with memory, and executive function but not processing speed. However, only marital status and social participation were associated with decline in memory and executive function over time in one cohort. We will also present replication of these analyses in the population-based SNAC-K and Rotterdam Study cohorts.