The frequency of U-shaped dose responses in the toxicological literature

Hormesis has been defined as a dose-response relationship in which there is a stimulatory response at low doses, but an inhibitory response at high do ses, resulting in a U- or inverted U-shaped dose response. To assess the pr oportion of studies satisfying criteria for evidence of hormesis, a databas e was created from published toxicological literature using rigorous a prio ri entry and evaluative criteria. One percent (195 out of 20,285) of the pu blished articles contained 668 dose-response relationships that met the ent ry criteria. Subsequent application of evaluative criteria revealed that 24 5 (37% of 668) dose-response relationships from 86 articles (0.4% of 20,285 ) satisfied requirements for evidence of hormesis. Quantitative evaluation of false-positive and false-negative responses indicated that the data were not very susceptible to such influences. A complementary analysis of all d ose responses assessed by hypothesis testing or distributional analyses, wh ere the units of comparison were treatment doses below the NOAEL, revealed that of 1089 doses below the NOAEL, 213 (19.5%) satisfied statistical signi ficance or distributional data evaluative criteria for hormesis, 869 (80%) did not differ from the control, and 7 (0.6%) displayed evidence of false-p ositive values. The 32.5-fold (19.5% vs 0.6%) greater occurrence of hormeti c responses than a response of similar magnitude in the opposite (negative) direction strongly supports the nonrandom nature of hormetic responses. Th is study, which provides the first documentation of a data-derived frequenc y of hormetic responses in the toxicologically oriented literature, indicat es that when the study design satisfies a priori criteria (i.e., a well-def ined NOAEL, greater than or equal to 2 doses below the NOAEL, and the end p oint measured has the capacity to display either stimulatory or inhibitory responses), hormesis is frequently encountered and is broadly represented a ccording to agent, model, and end point. These findings have broad-based im plications for study design, risk assessment methods, and the establishment of optimal drug doses and suggest important evolutionarily adaptive strate gies for dose-response relationships.