Estimation of genotype-specific risks at a disease susceptibility locus is an important question that is best carried out in a prospective study. Nevertheless it is usually desirable to make use of data from the families that have already been collected to identify the susceptibility locus. Assuming that the families have been collected without regard to genotype at the locus in question, most of the information can be extracted by writing the likelihood in terms of the risk for a genotype relative to the standard genotype and conditional on the parental mating type. Parameters may then be estimated by explicit solution of likelihood equations. This method permits estimation of risks for heterozygotes and homozygotes for different alleles, testing of different modes of inheritance and heterogeneity of risk between alleles. It is applicable to risk alleles for any disease locus or incompletely penetrant phenotype. We have used the method to estimate risks of Crohn disease for different CARD/NOD2 15 mutations, using the families originally collected to identify this susceptibility locus. The odds ratio of Crohn disease were, respectively, 1.97+/-0.85, 3.05+/-* and 4.55+/-1.34 for the R702W, G9068R and 1007fs heterozygotes and 3.29+/-0.64, 12.13+/-* and 34.66+/-12.87 for the corresponding homozygotes. (* Signifies insufficient data to estimate these values.) These results confirm the dosage effect for CARD15/NOD2 mutations and demonstrate that the disease risks are very different in homozygotes. This last observation illustrates the power of this approach, especially for alleles with low or moderate frequency in the general population.