Social Activity, Cognitive Decline and Dementia Risk: A 20-year Prospective Cohort Study

Riccardo E. Marioni; Cecile Proust-Lima; Helene Amieva; Carol Brayne; Fiona E. Matthews; Jean-Francois Dartigues; Helene Jacqmin-Gadda


BMC Public Health. 2015;15(1089) 

In This Article


The analysis cohort (n = 2854) is described in Table 1. The gender distribution was 41 % male and the majority of the cohort (57 %) had a mid-level education with 11 % having a validated secondary degree or higher qualification. The mean age of the sample at baseline was 77.0 (standard deviation (SD) 6.8) years and 783 were given a dementia diagnosis and 2200 died without dementia over the 20-year follow-up (mean time-to-diagnosed-dementia of 9.4 (SD 5.1) years). A low level of engagement was reported in 31 % of individuals, compared to 46 % and 23 % with medium and high levels, respectively. According to our classifications, around half of the cohort (48 %) had a large social network, under one fifth were satisfied with their social network, and 78 % felt that they were well understood by others.

Longitudinal Cognitive Change and Dementia Risk

In the whole population there was weak evidence for quadratic cognitive decline over time (Table 2). Having more education and being engaged in social, physical, and intellectual pursuits were associated with higher baseline cognitive ability. There was also an association between a higher educational level and slower linear cognitive decline but faster quadratic change (a compression of cognitive morbidity). Individuals who felt well understood had a slightly lower initial cognitive score but declined less over time than those who did not feel well understood.

In the time-to-dementia model (Fig. 1 and Additional file 1: Table S1 there was a decreased dementia risk for those with a medium or high level of engagement and a borderline-significant association for those who felt well understood. In the fully adjusted model (Additional file 1: Table S2, only the association for high level of engagement in social, physical, and intellectual pursuits was retained - Hazard Ratio (HR) 0.79 95 % confidence interval: 0.63, 0.99. Of the additional covariates, being widowed, having had a stroke, or having impairment with activities of daily living were all associated with an increased risk of dementia.

Figure 1.

Hazard ratios for dementia for demographic and engagement/perception variables. All estimates are from multivariate models adjusting for age, sex, education, social/physical/intellectual engagement, network size, satisfaction with relationships within the social network, and a sense of feeling well understood by social contacts

Three Homogeneous Sub-populations of Longitudinal Cognitive Change and Dementia Risk

The sex- and education-adjusted cognitive trajectories of the three latent classes are presented in Fig. 2 (left panel). The largest class contained 1997 individuals who did not show many signs of cognitive decline (non-decliners). The middle class (n = 611) showed moderate levels of cognitive decline (moderate decliners) while the smallest class (n = 246) had an accelerated rate of decline relative to the other two groups (fast decliners). The percentage of subjects diagnosed with dementia during the follow-up differed across the three groups (n = 157 (8 %), 423 (69 %), and 203 (83 %), respectively – P <0.001). The class-specific hazard functions displayed in Fig. 2 (right panel) highlight the large differences in the risk of dementia between the three sub-populations. Table 1 shows that there were few differences between the baseline characteristics of the groups, except for cognitive ability, which decreased from the non-decliners to the moderate decliners to the fast decliners (P < 0.001). The fast decliners were slightly younger than the other groups (mean age 74.3 (SD 4.8) years, P <0.001). None of the social-environment variables explained the heterogeneity in cognitive decline and dementia risk that was seen across the three sub-populations.

Figure 2.

Latent general cognitive factor trajectories and risk of dementia in the three latent class model