Drinking water and leukemia.

that this combination of three different tests gives the best overall concordance with rodent carcinogenicity. It is well known that the Ames test gives only just over 50% concordance with rodent carcinogenicity, but it is still extensively used. This is because it has a well-defined endpoint that one can readily understand in biological terms, i.e., genotoxicity. However, there is overwhelming evidence to show that enzymes of the cytochrome P450 super-family are involved in the metabolism and toxicity of most (-90%) chemicals. P450s play a pivotal role in toxicity and carcino-genicity, and the COMPACT system is designed to identify P450-mediated metabolism and metabolic activation. Although this system (3) was originally based solely on the structures of known P450 sub-strates, we have now generated full three-dimensional structures (4) of the mam-malian enzymes themselves (including human isoforms) which agree closely with experimental findings. However, we are also aware that there are other mechanisms of carcinogenicity that do not require P450 activation, and structure alert systems can be useful in identifying direct-acting carcinogens , for example (1,5). Nongenotoxic carcinogens are not so easy to predict but we are elaborating models to identify chemicals involved in perox-isome proliferation and other activation pathways such as 9-lyase cleavage. Eventually there will be a battery of tests in place, which we hope will adequately assess the likely risk to man from exposure to foreign compounds; so models of human enzymes and receptors which may mediate potentially carcinogenic events will be important. What is crucial is the determination of how readily a chemical is metab-olized and whether any reactive intermedi-ates (ROS or metabolites) are sufficiently long-lived to cause irreparable DNA damage. There may well be short-term test procedures developed which can assess these factors in vitro, but computer-based systems can be just as accurately predictive; however, these computer systems do not require synthesis of the chemical, are extremely rapid, and, consequently, relatively inexpensive. We appreciate that traditional toxicologists may have been suspicious of replacing biological tests with computer predictions, but there is evidence that attitudes are changing. One reason for publishing our retrospective study of the 40 NTP chemicals was to show that it is possible for a combination of tests to give reasonable concordances with rodent carcinogenicity, and we did not anticipate that moderating the equivocal results of the rodent study in the light of the pathology report presented at the NTP conference would be controversial. Our EHP …