Further considerations of difficulties of estimating familial risks from pedigree data

In elucidating the etiology of complex diseases through various epidemiologic and genetic analyses, it is necessary to define the expected risks for each member of a pedigree collected through an affected individual. It is shown that several factors are to be considered without which such expected risk estimates may be biased. The presence of secular trends in disease incidence introduces a bias in familial relative risk estimates, the magnitude and direction of which may be judged from the nature of the secular trends and the age and cohort composition of the individuals included in the study. Nonrandom follow-up time of the relatives, on the contrary, produces a bias in relative risk estimates, the magnitude and direction of which cannot be ascertained without any prior assumption about the distribution of follow-up time. In addition, competing causes of diseases and risk heterogeneity among different relatives and among families are also sources of indeterminate biases, which must be removed in order to interprete the results of relative risk studies and segregation analysis of discrete disease traits.