Selection Bias in Samples of Older Twins?

Abstract

Twin studies are a powerful approach for estimating genetic and environmental influences in later life, but the usual requirement that both twins are alive may introduce a selection bias in gerontological studies relative to representative samples of nontwins. In the present study, samples of older twins and nontwins in Sweden were compared across the domains of vitality, well-being, physical and cognitive functioning, and health utilization to evaluate possible selection bias. One member of each twin dyad in the OCTO-Twin Study of intact twin pairs older than age 80 was randomly selected (N = 128) and compared with a population-based sample of nontwins (N = 324) from the OCTO Study. Multiple regressions adjusting for differences in demographic variables showed significant effects for twin status in only 3 of 20 comparisons. The results suggest that twin pairs surviving into very late life are similar to a representative sample of nontwins of the same age in health status and biobehavioral functioning. These findings support the generalizability of twin studies for understanding genetic and environmental influences on aging, health, and behavior.