Rare genetic variants influence participation in cognitive components of the UK Biobank

Population-level biobanks are transforming health and genomic research however they are all subject to recruitment and participation biases, the impacts of which are not fully understood. The UK Biobank (UKBB) is one such cohort; it is widely used for health-related research, but its participants are not representative of the UK general population. In addition, those who complete optional components of UKBB differ from those who do not in terms of health-related exposures and outcomes. Participation in UKBB has a partly genetic basis, but the extent to which rare variants influence participation in optional study components, and how this might impact findings of rare variant association studies, is unknown. We analysed exome sequencing and array data from 398,149 UKBB participants to examine rare variant contributions to UKBB participation bias, focusing on cognitive phenotypes. We found that classes of rare variation previously shown to contribute to cognition were enriched among participants who completed fewer cognitive tests. We then compared the effects of rare variants on cognition itself between groups of participants defined by whether they completed all online cognitive tests. Rare variants had significantly weaker effects on cognition in individuals who completed all online cognitive tests than in those who did not. These findings demonstrate that participation is a non-random phenotype that is influenced by rare variants, and that individuals who engage in optional UKBB components are not representative of the broader cohort. We show how participation bias can reduce the power of rare variant association studies, potentially masking true associations. This bias should therefore be carefully considered when interpreting results derived from phenotypes with substantial missingness.