The traditional approach in epidemiology of relating exposure to an environmental agent such as a drug or infective agent has been to measure an overall risk (i.e., average and then "adjust risk for demographic variables and other confounders"). An attempt is sometimes made to define a "susceptible" subgroup. The analyses are usually based on good statistical methodology rather than an understanding of the interaction of body of host and agent. A twofold risk for 1000 exposed versus nonexposed people could be an average twofold risk for all 1000 exposed or a 20-fold risk for 100 exposed individuals (i.e., a drug-host interaction). Clearly, finding the 100 individuals with a 20-fold risk has much greater clinical importance than a twofold risk for 1000 people. The world of epidemiology may be changing-we may soon be able to define risk based on genetic susceptibility, at least sometimes.