Making the most of multiple measurements in estimating carrier probability in Duchenne muscular dystrophy: the Bayesian incorporation of repeated measurements using logistic discrimination.

In carrier detection studies, females-at-risk are usually tested several times if the results are ambiguous, whereas subjects in the control and obligate carrier reference groups may not be tested as often. The question is how to incorporate the multiple measurements most effectively with information in the family pedigree into combined carrier risks. Sets of measurements on individuals are not independent, but are related with the correlation coefficient 0 less than rho less than 1. We have developed a procedure for incorporating repeated measurements on individuals and their a priori chance of having the disease into logistic models. This procedure utilizes the set of measurements and an estimate of rho. We describe application of this procedure to carrier detection in Duchenne muscular dystrophy (DMD) using serum creatine kinase (CK) measurements as the biochemical indicator of carrier status. Estimates of rho for controls and obligate DMD carriers did not differ significantly from 0.5. Repeated testing with use of rho = 0.5 significantly decreased the median logistic carrier probability for controls and increased it for carriers. In some cases four to six rather than the three CK tests conventionally used in genetic counseling were necessary to obtain a stable logistic carrier probability for a subject.