Estimating Sibling Recurrence Risk in Population Sample Surveys

Background/Aims: Sibling recurrence risk (SRR) is a measure of familial aggregation of a disease and is often used in family-based studies in genetic epidemiology to indicate the existence of possible genes conferring susceptibility of disease. Estimating SRR requires information about the disease status of sibships of families with two or more siblings where at least one is affected. Since family-based studies are not usually random samples, estimates of SRR derived from these studies may be biased. Network sampling used in survey research offers a way to ascertain the disease status of sibships from interviewed individuals in household survey samples, in order to obtain (approximately) unbiased estimators of SRR and its related SRR ratio (SRR divided by the prevalence of disease). Methods: Two methods of ascertaining sibships of affected families are considered: in one method the siblings' affected status is reported by an individual, regardless of the individual's affected status, and in the other method only affected individuals can report their siblings' affected status. Network estimators of SRR and SRR ratio and estimators of their standard errors are provided. Results: Reported diabetes for siblings from the 1976 National Health Interview is used to illustrate the estimation methods. The SRR ratio for diabetes among living siblings was 5.79% [relative standard error (RSE) = 5.12%], and for living or deceased siblings, it was 7.66% (RSE = 3.76%). Conclusions: Network sampling estimators can provide population estimates of SRR and SRR ratio for diseases such as diabetes.