In this work, we evaluate whether in vitro systems are good predictors for in vivo estrogenic activity in fish. We focus on UV filters being used in sunscreens and in UV stabilization of materials. First, we determined the estrogenic activity of 23 UV filters and one UV filter metabolite employing a recombinant yeast carrying the estrogen receptor of rainbow trout (rtERalpha) and made comparisons with yeast carrying the human hERalpha for receptor specificity. Benzophenone-1 (BP1), benzophenone-2 (BP2), 4,4-dihydroxybenzophenone, 4-hydroxybenzophenone, 2,4,4-trihydroxy-benzophenone, and phenylsalicylate showed full dose-response curves with maximal responses of 81-115%, whereas 3-benzylidene camphor (3BC), octylsalicylate, benzylsalicylate, benzophenone-3, and benzophenone-4 displayed lower maximal responses of 15-74%. Whereas the activity of 17beta-estradiol was lower in the rtERalpha than the hERalpha assay, the activities of UV filters were similar or relatively higher in rtERalpha, indicating different relative binding activities of both ER. Subsequently, we analyzed whether the in vitro estrogenicity of eight UV filters is also displayed in vivo in fathead minnows by the induction potential of vitellogenin after 14 days of aqueous exposure. Of the three active compounds in vivo, 3BC induced vitellogenin at lower concentrations (435 microg/l) than BP1 (4919 microg/l) and BP2 (8783 microg/l). The study shows, for the first time, estrogenic activities of UV filters in fish both in vitro and in vivo. Thus we propose that receptor-based assays should be used for in vitro screening prior to in vivo testing, leading to environmental risk assessments based on combined, complementary, and appropriate species-related assays for hormonal activity.