Biomarkers may be very useful to assess and control the risk of long-term outcomes associated with exposure to industrial chemicals. Even when dose-effect and dose-response relationships have been demonstrated, the process of standard setting and the definition of 'safe' exposure levels is hampered by: (i) the arbitrary nature of attempts to distinguish acceptable from unacceptable effects; (ii) the difficulties in assessing the prognostic value of observed changes; (iii) the influence of the study power on statistically defined thresholds. Integration of epidemiological findings with mechanistic studies is necessary to assure confidence in risk assessment and to undertake sound risk management.