Drug safety is the main reason for the termination of drug development programs in clinical stages and the liver is amongst the organ systems most relevant for safety failures . 2D in vitro systems and animal models that are currently used to assess the hepatotoxic potential of drugs have important limitations. Hepatocytes in 2D monolayer cultures have significantly reduced functionality and metabolic capacity, which impairs their utility for prediction of drug behavior in vivo. Due to important species differences in drug metabolizing enzymes, animal models do not accurately reflect drug pharmacokinetics in man resulting in poor correlation with human toxicity. This leads to flawed selection of new candidate drugs, resulting in considerable risks for trial participants and increased costs.
Fialuridine (FIAU) provides a key example for unreliable preclinical safety assessments. FIAU, a nucleoside analogue for therapy of chronic hepatitis B infections did not show toxicity, neither in primary human hepatocyte 2D cultures nor in any animal systems tested, including mouse, rat, dog and cynomolgus monkey . Yet, in clinical trials, 7/15 patients developed severe hepatotoxicity, several weeks after beginning FIAU treatment, five of which died. The 3D spheroid model successfully predicted FIAU toxicity in vitro with an EC50 of 100nM (Fig. 1), thus indicating that the spheroid system can provide a powerful tool to predict compound toxicity during pre-clinical drug development stages, which are not detected in commonly used in vitro or animal systems.
Other in vitro systems require elevated concentrations of hepatotoxins (100x human therapeutic serum levels = cmax) and even then show only relatively sensitivity: 50% sensitivity and 5% false-positives  or 66% sensitivity with a false-positive rate of 10% . Choosing concentrations that approximate physiological exposure values is important, as mechanisms of toxicity can considerably differ at higher concentrations. As a result, unphysiological drug effects are detected that are irrelevant for treatment in man.
In contrast, the 3D spheroid model predicts hepatotoxicity with an unmatched sensitivity. In recent experiments, the system successfully flagged 70% of hepatotoxins that caused fatal liver injuries resulting in withdrawal or black box warnings as significantly toxic already at 5x human cmax. At higher concentrations, toxicity of 20/23 (87% sensitivity) hepatotoxic drugs was detected with 93% specificity (1/14 false-positives).
- Cook et al., Nat Rev Drug Disc, 2014
- McKenzie et al., NEJM, 1995
- Xu et al., Toxicol Sci, 2008
- Khetani et al., Toxicol Sci, 2013