The FDA issued a guidance on the safety testing of metabolites in February 2008, in which they stated that metabolites of concern are those that are detected at levels greater than 10% of the systemic exposure of the parent at steady state. This has presented many challenges in determining the circulating human metabolites at an early stage of development. The intention of this perspective is to address the question of how effective in vitro metabolism and early exploratory clinical data are in predicting the circulating metabolites from both a qualitative and a quantitative perspective. To this end, data were reviewed from 17 molecules in the Lilly portfolio for which there were in vitro data and a radiolabeled study in humans. Twelve example cases are presented in detail to demonstrate trends for when in vitro data adequately predicted in vivo (41%), when in vitro data underpredicted the circulating metabolites (35%), and when in vitro data overpredicted the circulating metabolites (24%). In addition, cases that present special challenges due to very low levels of the circulating parent or long half-lives of the parent and/or metabolites are presented. The trends indicate that the more complex the metabolism, the less likely the in vitro data were to predict the circulating metabolites. The in vitro data were also less predictive for N-glucuronidations and non-P450-mediated cleavage reactions. Although the in vitro data were better at predicting clearance pathways, the data set often failed to predict the quantity of metabolites, which is needed in consideration of whether or not a "disproportionate" metabolite may be circulating in human plasma.