Exploring longitudinal pathways from intelligence to morbidity and mortality risk
Calvin, Catherine Mary
Human population-based studies of longitudinal design observe that higher intelligence in youth confers protection from premature mortality in adulthood. This field of study (“cognitive epidemiology”; Deary & Batty, 2007) has firmly established associations between intelligence and health outcomes, and has begun to address the likely mechanisms involved. The present thesis assessed some social, educational, and lifestyle factors that potentially confound and/or mediate the intelligence-mortality link. First, I carried out a systematic review of longitudinal cohort studies reporting intelligence differences in youth in relation to adult mortality risk, and in meta-analysis I aggregated the effect sizes from 16. A one SD advantage in intelligence scores was associated with 24% (95% CI 23% to 25%) lower risk of death, during 17- to 69-year follow-up; this magnitude showed no sex differential. Socioeconomic status in early life did not explain the effect. Rather, the person’s own occupational status in adulthood and educational attainment explained a third and a half of the association, respectively. One issue in controlling for education, in such models, is its strong correlation with intelligence test performance, which could lead to statistical overadjustment. A second aspect of this thesis, therefore, addressed the nature of the intelligence-education covariance in two behaviour-genetic studies of large general population-based samples of schoolchildren from England and The Netherlands. Previous studies that reported intelligence—education genetic covariances were potentially biased in their use of twin self-selection or pre-selection sampling. Moreover, the analysis in this thesis used a novel statistical approach, and included non-twin data to represent fully the variance in performance scores of a population. Analysis of the English cohort confirmed the top end of estimates from previous studies: 76% to 88% of the phenotypic correlation was due to heritability. The Dutch cohort showed greater variance for equivalent estimates (33% to 100%). The results indicate a limit to the extent to which education and intelligence might be causative of one another suggesting caution in interpreting some of the substantive attenuation effects by education reported in the literature. Third, I investigated pathways from intelligence to cardiovascular disease risk factors, given the consistent and robust finding that an advantage in intelligence relates to lower cardiovascular disease-outcomes. I used data from the 1958 National Child Development Study to investigate age-11 intelligence in association with inflammatory and haemostatic biomarker status at age 46 years. The results replicated inverse associations previously reported in an older age sample, and a one SD advantage in intelligence related to a 1.1mg/L decrease in C-reactive protein. The effect was largely mediated by lifestyle factors, including smoking, occupational status and abdominal obesity. In two further studies I used the west of Scotland Twenty-07 cohort, to investigate processing speeds among 16, 36 and 56 year-olds in relation to: (1) Inflammation, and (2) metabolic-risk, after 20 years. The advantage of experimental rather than psychometric measures of cognitive ability is their reduced cultural and social bias. Faster reaction time predicted lower systemic inflammation in the youngest male cohort, which appeared to be partially confounded by baseline smoking and socioeconomic status. Furthermore, advantage in reaction time performance in the young and middle-aged cohorts significantly predicted reduced metabolic risk. This was partially explained by occupational status, but retained statistical significance in some fully-adjusted models. A one SD advantage in age 16 simple reaction time variability, related to the 21% (95% CI 12% to 30%) reduced odds of metabolic syndrome by age 36 in the basic model, and this effect remained unchanged after controlling for all covariates. The growing evidence for specific social and behavioural factors that mediate intelligence-to-mortality pathways are discussed, in respect of indirect evidence that underlying system integrity or early life confounding may contribute incrementally to the effect.