For many diseases genome-wide association studies are currently being used to identify markers associated with an increased risk of disease. The number of such markers is increasing steadily and methods for implementing their use as diagnostic tools are now being developed. Genetic information on presence or absence of a number of risk alleles is combined with population data on mortality and disease incidence to derive individual risk estimates of disease. The consequences of the population heterogeneity induced by the genetic variation in susceptibility have so far been ignored in these calculations. A frailty model is developed to investigate the impact of the population heterogeneity on the risk estimates. The approach is an example of the life-table method for heterogeneous populations introduced by Vaupel, Manton & Stallard in 1979 and further developed by Hougard in 1984. Danish population data on prostate cancer together with simulated data on the population heterogeneity introduced by 35 SNPs used in an ongoing prostate cancer trial in Denmark are used to illustrate the methodology and assess the validity of the standard approach and of a simple first order approximation.