Using a computer model developed by HTPC partner, the Hamner Institutes for Health Sciences, researchers at UNC Eshelman School of Pharmacy were able to predict human liver failure induced by the diabetes drug, troglitazone.

Troglitazone was introduced in 1997, and withdrawn from the market three years later after 63 patients died from liver failure. Though the drug had been shown to be safe in rat studies, and had only mildly adverse effects in human trials, introduction to a larger human population revealed that some patients could not metabolize the drug.

The UNC research team used DILIsym, a computer model developed by scientists at the Hamner Institutes, to combine chemical information about troglitazone with clinical human liver data, to see if the model could predict which patients in a simulated population would develop liver failure.

The model correctly predicted those at-risk patients, and also predicted that rats would respond differently than humans. From the UNC press release:

“Before DILIsym, no one had been able to completely explain troglitazone liver injury or suggest improved approaches so drug companies could avoid similar problems in the future,” (study author, Dr. Kim) Brouwer says. “It turns out that animals do a poor job predicting human drug-induced liver injury. There are lots of explanations, but one important reason is that bile acids are different in each species. Recent data suggest that the use of humanized systems has greater predictive power for adverse events like DILI.”


“Rare liver toxicity is now the major safety concern with new drugs and can often be detected only after many thousands of patients have received treatment,” Watkins says. “We believe that the application of DILIsym will greatly improve drug safety while minimizing animal testing and reducing the costs of new medicines.”

Read more about the development of DILIsym software here.