Estimation of the risk of a qualitative phenotype: Dependence on population risk

Individual disease risk estimated based on the data from single or multiple genetic loci is generally calculated using the genotypes of a subject, frequencies of alleles of interest, odds ratios and the average population risk. However, it is often difficult to estimate accurately the average population risk, and therefore it is often expressed as an interval. To better estimate the risk of a subject with given genotypes, we built R scripts using the R environment and constructed graphs to examine the change in the estimated risk as well as the relative risk according to the change of the average population risk. In most cases, the graph of the relative risk did not cross the line of y=1, thereby indicating that the order of the relative risk for given genotypes and the population average risk does not change when the average risk increases or decreases. In rare cases, however, the graph of the relative risk crossed the line of y=1, thereby indicating that the order of the relative risk for given genotypes and the population average risk does change owing to the change in the population risk. We propose that the relative risk should be estimated for not only the point average population risk but also for an interval of the average population risk. Moreover, when the graph crosses the line of y=1 within the interval, this information should be reported to the consumer.