We reanalyzed data from 218 two-year rodent carcinogenicity studies carried out by the National Toxicology Program (NTP). These data were originally collected for the purpose of identifying potential human carcinogens. However, the objective of our analysis was to investigate the frequency of possible anticarcinogenic effects in these data, since recurring cases of chemical-associated tumor reductions have been noted in the course of these studies over time. Our analysis reveals that most (>90%) NTP-tested chemicals show at least one statistically significant (p<0.05) decrease in site-specific tumor incidence. Because of the large number of statistical comparisons made in a long-term bioassay, random variability can account for many of these tumor decreases. However, we found that certain tumors (predominantly those with a high spontaneous incidence) show chemically related decreases far more frequently than chance expectation. Many of these decreases, particularly those for pituitary and mammary gland tumors, adrenal pheochromocytoma and uterine polyps in rats and liver and lung tumors in mice, are associated with the reduced body weights frequently observed in the dosed groups. The chemically related decreased incidences of leukemia in rats appear to be related to spleen damage, i.e., chemically related splenic toxicity is evident for most chemicals showing decreased incidences of leukemia. While random variability, associations with body weight and splenic toxicity can account for most of the decreased tumor incidences observed in NTP studies, there are other tumor decreases that could not be totally explained by these factors. Further investigations of possible mechanisms of action are underway. These data are relevant to the concept of chemoprevention as well as to the task of using long-term laboratory animal studies to predict enhanced human environmental-cancer risk for regulatory purposes.