Approximate confidence intervals for risk prediction in genetic counseling.

In current genetic counseling practice, a single risk estimate is often quoted to a family rather than a range of risks. Such point estimates are predicated on knowing basic parameters like recombination fractions exactly when, in fact, there may be considerable uncertainty about them. Using the large sample theory of statistics, it is possible to derive approximate risk intervals that incorporate known statistical imprecision. The necessary theory will be briefly discussed and illustrated by an application to family counseling for Duchenne muscular dystrophy in the presence of two flanking markers. Some of the problems of the theory will be mentioned. These include lack of adequate sample size to justify the conclusions of large sample theory, pronounced nonlinearity in the risk function, and failure to take into proper account genetic interference. Except in trivial cases, sophisticated computer software is needed to carry out the computations of risk intervals.