Case-control studies of genotypic relative risks using children of cases as controls

The use of a child as a case's control in a case-control study of genetic factors may be advantageous in some situations. We describe three methods of analysing such data. A method based upon the unconditional likelihood using case/child pairings apparently is the most efficient method. However, a conditional likelihood method using case/child pairs is more robust in that it allows for heterogeneity of the genetic trait among subpopulations as long as matings occur only within the same subpopulation. We argue that the child can be considered a genetic surrogate for the missing spouse and hence such designs are as valid as are those using spouses.