This study investigated pre-initiation and post-initiation effects of dietary ellagic acid (EA) on 7,12-dimethylbenz[a]anthracene (DMBA) multi-organ carcinogenesis in rainbow trout (Oncorhynchus mykiss). EA at 100, 250 (study 2), 1000 and 2000 (study 1) p.p.m. suppressed stomach adenopapilloma incidence by 33, 60, 70 and 78% (P < or = 0.001), respectively, as well as tumor multiplicity (P < 0.01) and size (P < 0.001) when fed continuously following DMBA initiation. However, continuous EA feeding also produced modest (250 p.p.m.) to extensive (1000, 2000 p.p.m.) growth rate suppression in these studies. Retrospective logistic regression modeling of the data allowed separation of growth-related from non-growth-related inhibitory effects. By this approach: (i) tumor development showed a similarly strong dependence (same regression slope) on animal growth rate in all treatment groups; (ii) EA-mediated reduction in mean population growth contributed to suppressed stomach tumor response above 250 p.p.m. EA; and (iii) even at high, toxic doses EA displayed inhibitory mechanisms additional to, and distinct from, growth suppression effect. The effects of post-initiation EA were organ specific. Chronic EA treatment significantly suppressed swim-bladder as well as stomach tumor incidence at doses > or = 1000 p.p.m., but increased liver tumor incidence at doses > or = 250 p.p.m. Three protocols examined EA effects on the initiation process. EA fed at 1000 p.p.m. concurrently with 750 p.p.m. dietary DMBA for 7 weeks modestly reduced stomach tumor incidence (from 85 to 78%, P < 0.05) and multiplicity (from 6.3 +/- 4.3 to 4.9 +/- 2.9, P < 0.01), but did not alter swim-bladder or liver response. The effect of EA pretreatment prior to DMBA single-dose initiation by gill uptake was also examined. When fed for 1 week prior to initiation, 2000 p.p.m. EA again imposed a small reduction in stomach adenoma incidence (from 88 to 78%; P < 0.05) and multiplicity (from 5.5 +/- 3.2 to 4.4 +/- 3.2; P < 0.01). However, when EA was pre-fed for 3 weeks instead of 1 week, protection in the stomach was lost and response in liver and swim-bladder significantly increased. In sum, these studies demonstrate that EA influence on DMBA tumorigenesis in this multi-organ model is highly protocol dependent and organ specific. Post-initiation dietary EA consistently suppressed stomach tumor development in trout, at EA doses far lower than those required for protection in rodents. At higher doses, however, EA also displayed toxicity and a potential in some protocols to enhance tumor response in other organs.