Statistical resolution of genetic heterogeneity in familial disease.

If a disease can be split into two or more groups on any criterion (clinical, biochemical, physiological or statistical) then the grouping can be tested to establish if genetically independent forms of the disease have been identified. The data required are simply the frequencies of the two disease groups in relatives of probands for each of the disease groups. A systematic search for such distinct groups is proposed in searches for genetic heterogeneity in familial diseases. In disease forms with overlapping, correlated genetic liabilities, the method of Falconer (1967) can be used to estimate the genetic correlation. However, when the groupings of the disease are confounded (such as one form precluding the other as in early and late onset diabetes) Falconer's method will be biased. Special methods of analysis to estimate the genetic parameters have been developed and are presented here. However, even when the groupings are confounded the Falconer method still gives reasonable estimates of the genetic correlation, in that they are unlikely to seriously mislead the investigator in the analysis and interpretation of observed data. In practice Falconer's simple method may be preferred to the more complex methods developed here because it involves fewer assumptions and can be applied over a wider range of circumstances.