Drug-induced liver injury, or DILI, is a major concern for the development and sustained sale of drugs. For the 548 NCEs that were approved in the US from the mid-seventies to the end of the century, adverse drug reactions led to 16 approved drugs being withdrawn and 56 being given black box warnings. A recent article in Nature Reviews Drug Discovery highlights a major new project in the Innovative Medicines Initiative (IMI) which will provide €32 million to a consortium of 17 industry & 9 academic partners. The article also points out a couple of well-worked examples which make the point of the current lack of understanding. The first, panadiplon, a GABA-A partial agonist being developed for non-sedating anxiety was terminated in Phase I over a decade ago due to liver injury despite no evidence for related toxicity in rat, dog or monkey pre-clinical tox. This was discussed in more depth by Abbott in a paper over 10 years ago, which discusses some additional toxicology in rabbits which was later discovered to have some connection with the human observations. This paper nicely illustrates the need for both more openness on the part of sharing toxicology data and also the challenges of interpreting the data across species.
The second example plucked from the other end of the spectrum is CAL-101, a phosphoinositide-3 kinase δ inhibitor in development for various leukaemias. This molecule also showed liver toxicity in Phase I despite no warnings pre-clinically, but this toxicity has been manageable and the agent has progressed to Phase III, leading to the acquisition of Calistoga by Gilead (and the renaming of the drug to GS-1101). The review of the tox and clinical data is covered in a very interesting set of blog articles.
Many papers have appeared over the last few years dealing with the hopelessness of ever understanding these problems as well as some more pragmatic approaches – many are referenced from a recent blog or the 2010 NRDD paper on chemically reactive metabolites. This new IMI project seeks to combine data sets from toxicology studies and to both evaluate existing models and devise new mechanistic platforms (with emphasis on human derived rather than animal in vivo) to improve future prediction of DILI. The goals are clearly aspirational and the sharing of toxicology data more widely is absolutely essential, however these large EU projects provide a challenging framework to make rapid progress.