Epidemiology involves studying the distribution and determining the causes of disease. Typically, epidemiologic studies result in an identification of associations or correlations, although correlation does not equal causation. The Bradford-Hill criteria represent one approach for determining causation, and include strength of association, temporal relationship, consistency, biological plausibility, specificity in the causes, dose-response relationships, experimental evidence, analogy, and coherence. Epidemiological approaches are also applied to interpreting the results of medical testing. Such interpretation requires an understanding of the positive predictive value of a test - the likelihood that a positive result indicates the disease actually to be present. The positive predictive value is calculated using specificity and sensitivity, which are intrinsic properties of the test, along with the population prevalence or pre-test probability. For cancer screening tests, it is also important to understand cancer biology. Only some of the tumors detected by screening are of a type destined to become fatal, but if detected they are curable. Additionally, studies of cancer screening are affected by lead-time bias, length bias, overdiagnosis, and stage-migration bias. Criteria for appropriate screening require that the test is valid, reliable, reproducible, inexpensive, easy to administer, imposes minimal discomfort, detects disease in the preclinical stage, and affects outcomes.