Binding of diethylstilbestrol and four different triphenylethylene derivatives: tamoxifen, toremifene, clomiphene and triparanol to DNA in rat liver, was studied using the 32P-postlabelling method with HPLC-radioactivity detection. Three different modifications of the 32P-postlabelling technique (a) a bisphosphate method with adduct enrichment by nuclease P1 (NP1)-treatment or (b) by butanol extraction and (c) a monophosphate method, were applied in order to provide an unbiased analysis of adduct formation. When tamoxifen was administered by daily gavage for 4 weeks (80 mumol/kg for 2 weeks and 40 mumol/kg for a further 2 weeks) two major adducts and about six minor adducts were produced in the liver of female Sprague-Dawley rats. Equimolar doses of toremifene produced one apparent adduct. The adduct levels in the tamoxifen and toremifene treated rats were 600 and 2/10(8) nucleotides, respectively. Under conditions used, clomiphene, triparanol and diethylstilbestrol did not produce DNA adducts. The present and previous data suggest that modification (a) is the 32P-postlabelling method of choice for risk assessment in human subjects. Modification (c) with butanol extraction after labelling has the advantage of low background radioactivity and may be preferable if large amounts of DNA are available. The main tamoxifen adducts were suggested to be alpha-(N2-deoxyguanosinyl)tamoxifen and alpha-(N2-deoxyguanosinyl)-N-desmethyltamoxifen.