Neuroticism is associated with poor health outcomes, but its contribution to the accumulation of health deficits in old age, that is, the frailty index, is largely unknown. We aimed to explore associations between neuroticism and frailty cross-sectionally and longitudinally, and to investigate the contribution of shared genetic influences.
Data were derived from the UK Biobank (UKB; n = 274,951), the Australian Over 50’s Study (AO50; n = 2849), and the Swedish Twin Registry (Screening Across the Lifespan of Twins Study [SALT], n = 18,960; The Swedish Adoption/Twin Study of Aging [SATSA], n = 1365). Associations between neuroticism and the frailty index were investigated using regression analysis cross-sectionally in UKB, AO50, and SATSA and longitudinally in SALT (25–29 years of follow-up) and SATSA (6 and 23 years of follow-up). The co-twin control method was applied to explore the contribution of underlying shared familial factors (SALT, SATSA, AO50). Genome-wide polygenic risk scores for neuroticism were used in all samples to further assess whether common genetic variants associated with neuroticism predict frailty.
High neuroticism was consistently associated with greater frailty cross-sectionally (adjusted β [95% confidence intervals] in UKB = 0.32 [0.32–0.33]; AO50 = 0.35 [0.31–0.39]; SATSA = 0.33 [0.27–0.39]) and longitudinally up to 29 years (SALT = 0.24 [0.22–0.25]; SATSA 6 years = 0.31 [0.24–0.38]; SATSA 23 years = 0.16 [0.07–0.25]). When adjusting for underlying shared genetic and environmental factors, the neuroticism-frailty association remained significant, although decreased. Polygenic risk scores for neuroticism significantly predicted frailty in the two larger samples (meta-analyzed total β = 0.059 [0.055–0.062]).
Neuroticism in midlife predicts frailty in late life. Neuroticism may have a causal influence on frailty, whereas both environmental and genetic influences, including neuroticism-associated common genetic variants, contribute to this relationship.
From the Department of Medical Epidemiology and Biostatistics (Danielsdottir, Jylhävä, Hägg, Lu, Pedersen, Mosing, Lehto), Karolinska Institutet, Stockholm, Sweden; Department of Genetics and Computational Biology (Colodro-Conde, Martin), QIMR Berghofer Medical Research Institute, Brisbane, Australia; Department of Neuroscience (Mosing), Karolinska Institutet, Stockholm, Sweden; and Department of Chronic Diseases (Lehto), Institute for Health Development, Tallinn, Estonia.
Address correspondence to Kelli Lehto, PhD, Department of Medical Epidemiology and Biostatistics, Karolinska Instituet, Nobels Väg 12a, 171 77, Sweden. E-mail: Kelli.email@example.com
Received for publication December 30, 2018; revision received May 8, 2019.
Online date: August 30, 2019