Predicting potential liver toxicity from phase 2 data: a case study with ximelagatran.

Ximelagatran was denied marketing approval in the USA and was withdrawn from those markets in which it had been approved, because of concerns over potential liver toxicity. A retrospective analysis of phase 2 data relating to liver toxicity is performed using the methods of extreme value modelling. The analysis reveals that the phase 2 data were predictive of the phase 3 results and, had the methods been available at the time, such analysis would have provided valuable information relating to the decision to proceed with further development of the compound.