A new statistical test for age-of-onset anticipation: application to bipolar disorder

The discovery of expanding microsatellite repeat mechanisms for a variety of diseases has spurred renewed interest in testing for genetic anticipation in complex diseases. However, standard statistical methods can be shown to be inappropriate for this purpose. We derive a new statistical test for comparing mean age of onset in parents with mean age of onset in children, based on a random sample of affected parent-child pairs. This test incorporates the right-truncation of the age-of-onset distributions and utilizes age-at-interview information. We illustrate the use of the test by application to bipolar affective disorder data. In this example, the correct p-value is 0.014 whereas the results based on the biased estimates give a p-value of 0.0001 [McInnis et al., 1993]. We also briefly consider the effects of ascertainment bias when parent-child pairs are obtained from samples collected for purposes of linkage analysis.