The paragraph discusses two approaches that have been used to identify teratogens, or exposures during pregnancy that harm the fetus or embryo. The first approach involves identifying a rare exposure linked to a rare defect, as was done for rubella virus and fetal alcohol syndrome. The second approach uses epidemiological data to confirm an association, as was done for valproic acid and spina bifida. It concludes that no single piece of evidence confirms causal relationships, and instead they emerge from multiple lines of evidence that each suggest but do not prove the relationship.