Study Finds Biobank Participation Bias Impacts Genetic Associations, Downstream Analyses

The individuals who participate in volunteer-based biobanks rarely represent the broader population, which can skew the results of research that relies on biobank samples, according to a study appearing in Nature Human Behaviour. The findings suggest that investigators should take steps to account for such potential biases to ensure the accuracy of their results. Biobanks have become the cornerstone of genetic epidemiology in recent years, but studies using biobank data typically do not account for the disconnect between their target populations and the individuals from whom the data is collected. UK Biobank participants, for example, are known to be better educated, healthier, and older than the general UK population. To evaluate the impact of such selective participation, a team led by University of Lausanne researchers derived probability weights for the UK Biobank based on 14 variables harmonized with data from a representative sample, then conducted inverse-probability-weighted genome-wide association analyses on 19 traits. Conducting genome-wide analyses in a more representative sample of the UK Biobank, the researchers found that selective participation can distort genome-wide findings and downstream analyses, particularly as they relate to socio-behavioral traits. "Moving forward, more efforts ensuring either sample representativeness or methods correcting for participation bias are paramount, especially when studying the genetic underpinnings of behavior, lifestyles, and educational outcomes," the study's authors conclude.