[Empirical estimation of recurrence risks in multifactorial diseases with variable age-at-onset: conditional probability approach].

Conditional probability approach in estimation of recurrence risks considered in the report is naturally extended to the case of diseases with variable age-at-onset which are divided into many age-at-onset incompatible classes, the distribution of families for classes of affected and normal siblings being polynomial, modified with regard both for classes by age manifestation and for age of relatives at the moment of observation. This case is realised within the framework of the so called "life-table" approach of general use in demography. It is emphasized that the estimate of "accumulated morbidity" obtained within this framework--as a resulting estimate of "forthcoming risk" for individuals which are at premanifestational age--is the "a posteriori" probability for relatives to be affected by the disease. It is indicated that conditional probability approach, being free of any biological models of the disease origin, is not free of ascertainment models: different kinds of stratifications of the patients for the features related to their ascertainment as probands (age-at-onset, age by the time of observation, sex, birth order, etc.) can naturally generate inconsistency of the ascertainment probability pi both within and among the families, violating the independence of the ascertainment of patients as probands. In this connection, the maximum likelihood equations are given for estimation of the ascertainment probabilities of different incompatible groups of patients as probands and the necessity to account for possible differences in pi-s is pointed out.