This article is more than 5 years old. Information may no longer be current.

FDA develops new model to predict DILI severity

FDA researchers developed a novel and effective tool to determine drug-induced liver injury risk and assist in choosing drug development candidates, according to research published in Hepatology.

“A DILI score algorithm was developed that provides a scale of assessing the DILI risk in humans associated with oral medications,” Minjun Chen, PhD, of the National Center for Toxicological Research at the FDA, and colleagues wrote.

Building on the “rule-of-two” (RO2) model, Chen and colleagues collected data on 354 FDA-approved oral medications. After temporarily removing those deemed less of a DILI concern, 192 drugs remained. These were further classified as most-DILI concern (n = 124) or no DILI concern (n = 68).

Researchers studied daily doses of less than 10 mg; between 10 mg and 100 mg; and greater than or equal to 100 mg. The medications’ lipophilicity and reactive metabolism (RM) formation was recorded and each drug assigned a logP of less than one, one to three, or greater than or equal to three. Of the drugs considered, 117 generated RM but with a specificity of 79%, this feature alone was not enough to develop the model, researchers wrote.  

“Adding RM into the RO2 equation improved prediction considerably,” researchers wrote.  

Using a category daily dose greater than or equal to 100 mg and logP greater than or equal to three resulted in specificity of 100% (vs. 96%) and similar sensitivities (38% vs. 40%).  When logP was greater than or equal to one, but less than three — and category daily dose remained the same —  the model also performed well, with an increased specificity (97% vs. 90%) and a similar sensitivity (19% vs. 24%). The only instance where RM did not enhance prediction was when category daily dose was greater than or equal to 100 mg, logP was less than one, or the daily dose was less than 100 mg.

Chen and colleagues also studied drugs that produce RM but are of little or no DILI concern.

“Taking this ambiguity into account, the RM alone still predicted DILI with an extraordinarily high OR but at a slight reduction in specificity, even though it was superior compared to the RO2 alone,” Chen and colleagues wrote. “By comparing the performance of RM and the RO2, the obvious trade-off lies between sensitivity and specificity, with the RO2 improving specificity and RM increasing sensitivity.”

The DILI calculator is on FDA’s website. DILI scores greater than or equal to seven and less than three have severe adverse effects. The algorithm drew inspiration from MELD score, and was validated in 159 cases from the NIH’s LiverTox database, researchers wrote. – by Janel Miller

Disclosures: The researchers report no relevant financial disclosures.