Prediction of Acute Liver Injury Trajectory in Patients Following Acetaminophen Overdose: A Multibiomarker Machine...

Area of Science:

• Biomarker discovery and validation
• Clinical trial design optimization
• Drug development for liver injury

Background:

• Clinical translation of novel therapies is often hindered by heterogeneity in late-stage trials.
• Acetaminophen-induced liver injury (APAP DILI) trials face diluted efficacy signals due to spontaneous patient recovery.
• Efficient trial designs are crucial for advancing new treatments.

Purpose of the Study:

• To develop a prognostic enrichment tool for identifying patients with worsening APAP DILI trajectories.
• To improve the efficiency of clinical trial designs for novel APAP DILI therapies.
• To reduce sample size requirements and de-risk drug development.

Main Methods:

• Utilized serum samples from three UK cohorts (MAPP2, MAIL trial, healthy controls).
• Measured 63 biomarkers and evaluated 321,682 combinations using kernel naïve Bayes classification.
• Developed and validated a four-biomarker model (MCSFR, WBC, Sodium, K18) for predicting liver injury trajectory.

Main Results:

• The four-biomarker model achieved an AUC of 0.868 (derivation) and 0.854 (evaluation).
• Optimized model yielded a Positive Likelihood Ratio of 14.4, increasing Positive Predictive Value from 29.4% to 85.7%.
• Cost-minimization modeling suggested an application threshold reducing illustrative trial costs from $39.0M to $8.3M.

Conclusions:

• Multidimensional biomarker models can effectively resolve signal dilution in APAP DILI trials.
• Identifying patients with injury progression significantly reduces required sample sizes.
• This approach has the potential to de-risk novel therapy development for APAP DILI.