tions, can be limited by methodological weaknesses, such as the absence of detailed information on the sampling frame or referral patterns that generated the study sample, the degree to which the study sample is representative of the population from which it was drawn, exposure histories of the subjects, detailed assessments of health status, and the nature of severity of possible confounding biases.

Case-control studies, in which the exposure status (or history) of individuals with a certain health outcome (case) is compared with the exposure status of individuals without the health outcome (controls), can provide a much stronger basis for drawing inferences about exposure-disease associations. Among the challenges of such studies, however, are assembling a representative group of cases and a comparable group of controls, collecting adequate information on critical aspects of exposure history (which, in the case of long-latency diseases, might mean exposures that occurred decades before), and identifying the critical potential confounding biases. A case-control design, however, might be the only efficient way to study rare health outcomes.

Cohort designs (e.g., cross-sectional, retrospective, and prospective) provide a number of advantages. Instead of being selected on the basis of outcome status, as in case-control studies, study subjects are either randomly selected from the target population or selected on the basis of particular exposure characteristics (e.g., over-sampling of extremes of exposure distribution). The former strategy might be used if the goal is to enhance the generalizability of the study inferences to the target population, and the latter might be used if the goal is to estimate, with the greatest precision, the nature of the dose-response relationship within a certain region of the dose distribution. Another advantage of a cohort design is that multiple health outcomes can be measured and related to the index of exposure. A cohort study that incorporates prospective assessments of the study sample generally provides opportunities to assemble more-comprehensive exposure histories of the study subjects and to examine the natural history of a dose-response relationship, including factors that modify risk. As with all epidemiological studies, the methodological challenges of cohort studies include accurate classification of exposure and outcome status and the assessment and control of confounding bias.