Contribution of cognitive performance and cognitive decline to associations between socioeconomic factors and dementia: A cohort study

Abstract

Background: Socioeconomic disadvantage is a risk factor for dementia, but longitudinal studies suggest that it does not affect the rate of cognitive decline. Our objective is to understand the manner in which socioeconomic disadvantage shapes dementia risk by examining its associations with midlife cognitive performance and cognitive decline from midlife to old age, including cognitive decline trajectories in those with dementia.

Methods and findings: Data are drawn from the Whitehall II study (N = 10,308 at study recruitment in 1985), with cognitive function assessed at 4 waves (1997, 2002, 2007, and 2012). Sociodemographic, behavioural, and cardiometabolic risk factors from 1985 and chronic conditions until the end of follow-up in 2015 (N dementia/total = 320/9,938) allowed the use of inverse probability weighting to take into account data missing because of loss to follow-up between the study recruitment in 1985 and the introduction of cognitive tests to the study in 1997. Generalized estimating equations and Cox regression were used to assess associations of socioeconomic markers (height, education, and midlife occupation categorized as low, intermediate, and high to represent hierarchy in the socioeconomic marker) with cognitive performance, cognitive decline, and dementia (N dementia/total = 195/7,499). In those with dementia, we examined whether retrospective trajectories of cognitive decline (backward timescale) over 18 years prior to diagnosis differed as a function of socioeconomic markers. Socioeconomic disadvantage was associated with poorer cognitive performance (all p < 0.001). Using point estimates for the effect of age, the differences between the high and low socioeconomic groups corresponded to an age effect of 4, 15, and 26 years, for height, education, and midlife occupation, respectively. There was no evidence of faster cognitive decline in socioeconomically disadvantaged groups. Low occupation, but not height or education, was associated with risk of dementia (hazard ratio [HR] = 2.03 [95% confidence interval (CI) 1.23-3.36]) in an analysis adjusted for sociodemographic factors; the excess risk was unchanged after adjustment for cognitive decline but was completely attenuated after adjustment for cognitive performance. In further analyses restricted to those with dementia, retrospective cognitive trajectories over 18 years prior to dementia diagnosis showed faster cognitive decline in the high education (p = 0.006) and occupation (p = 0.001) groups such that large differences in cognitive performance in midlife were attenuated at dementia diagnosis. A major limitation of our study is the use of electronic health records rather than comprehensive dementia ascertainment.

Conclusions: Our results support the passive or threshold cognitive reserve hypothesis, in that high cognitive reserve is associated with lower risk for dementia because of its association with cognitive performance, which provides a buffer against clinical expression of dementia.

Fig 1. Flow chart for study of cognitive decline and dementia.

Abbreviations: IPW, inverse probability weighting. ^¶Of the 329 participants with dementia recorded until 31 March 2015, 9 were in those with data missing on covariates or onset of dementia was prior to 1997, allowing 320 participants to be included in IPW. In those with cognitive data (N = 7,499), a total of 195 participants had a dementia diagnosis over the follow-up. ^± Cognitive tests were introduced to the study at the 1997–1999 assessment, repeated at subsequent assessments. ^† A total of 7,499 participants were included in the analysis; 43% had cognitive data at all 4 waves, 28% at 3 waves, 15% at 2 waves, and 14% only at 1 wave.

Fig 2. Trajectory of global cognitive score in participants with dementia and other participants in the years leading to dementia diagnosis in the total population (panel A) [29] and in those with dementia as a function of height (panel B), education (panel C), and occupation (panel D).

^{† †}Panel A depicts marginal effects of dementia on trajectories of the global cognitive score (composed of tests of memory, reasoning, and phonemic and semantic fluency), adjusted for age at the end of follow-up (dementia diagnosis, death, or 31 March 2015), 5-year birth cohort, sex, and education. Results for height, education, and occupation include age at onset of dementia, sex, ethnicity, their interactions with time and time squared, 5-year birth cohort to take cohort effects into account, and time-dependant marital status.