A workshop was held to discuss the uses of data on DNA adduct measurement in humans and in experimental systems in vitro and in vivo. The discussions focused principally on the understanding of the toxicological significance of DNA adducts as provided by information from animal models. An Expert Panel concluded that human DNA adduct data have utility in several aspects of risk assessment. The presence and amount of specific adducts that can be correlated with a chemical exposure are relevant for hazard identification and risk evaluation. Data from experimental systems have established dose-response relationships between the level of adducts and exposure, but these remain complex and depend on metabolic fate. Although structure-activity relationships have been useful retrospectively to explain the DNA-reactive nature of some chemicals or classes of chemicals, there are currently no means outside the laboratory to specifically predict the adduct-producing potency of a compound. Analysis of DNA adducts in tissues of laboratory animals and humans has revealed sensitive subpopulations, a finding that has important relevance for human risk assessment. Adduct analysis may be one of the best tools available to characterize exposures to DNA from complex mixtures for purposes of epidemiological investigation. Consensus statements were developed based on presentations by R. Gupta, W. Lutz, R. Nath, and B. Singer [see Regul. Toxicol. Pharmacol. 23(1), 1996] and subsequent discussions. First, rigorous scientific criteria should be met for the detection and characterization of specific DNA adducts in vitro and in target tissues in vivo. Second, the use of adduct data in risk extrapolation has the greatest value when there is characterization of adduct structure, an understanding of the role of repair in DNA adduct removal, and demonstration of biological relevance for each adduct. Third, the detection of DNA adducts in a tissue does not necessarily indicate a specific tumorigenic risk for that tissue. Fourth, the mutagenic potency for specific adducts varies by several orders of magnitude. Fifth, the role of DNA adducts induced by exogenous agents must be placed in perspective of endogenously produced adducts. The biological significance of a type of DNA adduct is related to several factors, including the efficiency of conversion to mutation, the amounts of similar endogenous adducts, and the variety of exogenous DNA adducts found in DNA from humans. The biological relevance of DNA adducts may be deduced from the dose-response relationships for adducts and tumors at physiologically relevant doses as well as from data showing mutations in targets such as oncogenes or tumor suppressor genes. There is convincing evidence in the literature for an association between some specific DNA adducts, mutation, and the carcinogenic process. As a general conclusion, the Panel suggested that the current technological capabilities for detection of DNA adducts exceed our ability to define the biological significance of adducts as it relates to toxicity or health outcome. DNA adducts are likely to play an important role in human risk for cancer induction and progression, but the quantitative aspects of this relationship remain to be determined.