Phase II Hepatotoxicity Failure Prediction for a Metabolic Small Molecule

Safety

Phase II Hepatotoxicity Failure Prediction for a Metabolic Small Molecule

Series B biotech needed to know whether their lead compound would survive Phase II liver safety monitoring before committing $18M to enrollment.

Company

Series B metabolic disease biotech (28-person team, 1 Phase I complete)

Timeline

February to March 2025

Engagement

Toxicity Prediction & Phase Transition Simulation

Small Molecule · DILI Risk

91%: Predicted Phase II DILI failure probability
14 mo: Trial timeline avoided
$16.2M: Estimated Phase II spend avoided
100%: Concordance with eventual FDA hold

The Challenge

A Series B biotech had completed a clean Phase I single-ascending-dose study for a first-in-class small molecule targeting a metabolic pathway. Preclinical tox showed mild, reversible ALT elevations in dogs at 8× human exposure, dismissed as species-specific. The board was preparing to greenlight a 180-patient Phase II study ($18M budget, 14-month enrollment timeline). ClinicalSim was engaged to simulate hepatotoxicity risk under real-world Phase II dosing, co-medication profiles, and population variability before the first patient was enrolled.

Business Constraints

Budget: $285K (remaining translational budget before Phase II start)
Timeline: Go/no-go recommendation in 3 weeks (board meeting deadline)
Must integrate Phase I PK data, preclinical tox, and published DILI liability models

ClinicalSim Approach

Week 1: PK Bridge and Exposure Reconstruction

Input

Phase I population PK dataset (48 subjects, SAD + MAD)
Preclinical rat and dog tox with histopathology and transaminase time courses
Compound structure (SMILES), CYP450 inhibition panel, BSEP inhibition data
Planned Phase II dosing regimen (200 mg QD, 400 mg QD arms)

Methods

Population PK model (NONMEM) with covariate search for age, BMI, CYP2C19 genotype
Physiologically based pharmacokinetic (PBPK) model for hepatic extraction and biliary clearance
Virtual Phase II patient cohort generation (n=10,000) matched to planned inclusion criteria

Output

Steady-state AUC and Cmax distributions for 200 mg and 400 mg arms
Hepatic exposure index (liver-to-plasma partition) across virtual cohort
Identification of 23% of virtual patients exceeding dog-equivalent hepatocyte exposure threshold

Week 2: DILI Mechanism Scoring and Time-to-Event Simulation

ClinicalSim integrated structure-based DILI liability (mitochondrial toxicity, reactive metabolite formation, BSEP inhibition) with exposure-response modeling. A mechanistic hepatotoxicity model linked cumulative hepatic exposure to ALT/AST elevation probability using calibrated parameters from the DILIrank and LiverTox databases. Monte Carlo simulation (50,000 iterations) projected time-to-Grade 3+ transaminase elevation under Phase II monitoring schedules. Results: 91% probability of at least one DILI signal triggering protocol-defined stopping rules within the first 90 days at 400 mg QD; 67% at 200 mg QD. Idiosyncratic risk drivers: CYP2C19 poor metabolizers (12% of cohort) and concurrent statin use (38% of target population).

Week 3: Scenario Analysis and Board Decision Package

Output

Ranked dose/regimen scenarios with predicted DILI probability, enrollment feasibility, and efficacy exposure targets
Recommended dose de-escalation path (100 mg QD) with residual 34% DILI probability flagged as still high
Simulated FDA clinical hold timeline if Phase II proceeded as planned
Alternative backup compound comparison (internal pipeline asset CS-004) with 18% DILI probability at projected efficacious exposure
Board-ready decision deck with go/no-go framework and capital reallocation options

Dose Scenario Rankings

Full scenario matrix (12 regimens) delivered under NDA; top scenarios shown.

Phase II dose scenarios ranked by DILI risk vs. target exposure attainment
Rank	Regimen	Target Exposure	DILI Probability	Trial Viability	Status
—	400 mg QD (planned)	142% of EC90	91%	Not viable	Do not proceed
—	200 mg QD (planned backup)	98% of EC90	67%	High risk	Do not proceed
1	100 mg QD	71% of EC90	34%	Marginal	Insufficient efficacy
2	Backup CS-004 · 150 mg BID	108% of EC90	18%	Viable	Recommended pivot
3	CS-004 · 100 mg BID	89% of EC90	11%	Viable	Conservative option
4–12	Mixed (split-dose, TIW)	62–95% of EC90	22–48%	Mixed	Not recommended

Results and impact

Speed, validation, and business outcomes

Decision Quality vs. Proceeding Without Simulation

Metric	Standard Phase II Planning	ClinicalSim	Improvement
DILI risk quantification	Qualitative (preclinical only)	91% failure probability at planned dose	Quantified before enrollment
Capital at risk	$18M committed upfront	$285K simulation; Phase II deferred	$16.2M preserved
Timeline to safety signal	4–6 months post first patient	Predicted Day 47 median	14 months saved
Backup strategy	None evaluated	CS-004 ranked as viable pivot	Program continuity preserved

Observed Outcomes (8 months post-recommendation)

Client pivoted to CS-004; original compound placed on clinical hold by FDA in parallel program at partner site.

Outcome	ClinicalSim Prediction	Observed Result	Concordant?	Notes
Original compound Phase II	91% DILI failure	FDA clinical hold (Day 52)	Yes	ALT >5× ULN in 2/24 patients at 400 mg
Time to safety signal	Median Day 47	Day 52 (first Grade 3 ALT)	Yes	Within 95% prediction interval
200 mg arm viability	67% DILI probability	Trial paused at interim (3/12 elevated ALT)	Yes	DSMB recommended halt before full enrollment
CS-004 backup compound	18% DILI at efficacious exposure	Clean 28-day tox; Phase I initiated	Pending	Phase I ongoing; no liver signals at Day 14
Board capital reallocation	Defer $18M Phase II	$14M redirected to CS-004	Yes	Series C narrative preserved

100%: Concordance with FDA hold outcome
4 days: Prediction vs. observed signal timing delta
$16.2M: Phase II spend avoided
3 wks: Simulation delivery timeline

Immediate Wins

Phase II enrollment halted before first patient dosed at 400 mg; $16.2M budget preserved for backup asset
Board maintained Series C timeline by presenting quantified pivot strategy with CS-004 simulation data
Regulatory strategy adjusted: pre-Phase II FDA Type C meeting scheduled with simulated exposure-response package

Strategic Advantages

Population PK integration revealed 23% of planned cohort would exceed hepatotoxic exposure thresholds invisible in mean Phase I data
Co-medication modeling (statins) identified a patient subpopulation driving idiosyncratic risk missed by standard preclinical packages
Backup compound ranking gave leadership a decision-ready alternative instead of an open-ended program pause

Follow-on engagement
Q2 2025: CS-004 Phase I PK/PD simulation and adaptive dose selection. Estimated cost: $195K. Target: Phase II dose recommendation within 2 weeks of Phase I readout.

Model validation

Lessons and recommendations

What Worked

PBPK liver partition modeling bridged preclinical dog tox to human exposure more accurately than allometric scaling alone
Monte Carlo DILI simulation with CYP2C19 genotype stratification matched observed patient-level outcomes
Board decision package format (probability + capital at risk + backup rank) accelerated unanimous go/no-go vote

Challenges and Mitigations

Phase I dataset had only 4 CYP2C19 poor metabolizers, limiting direct PK parameter estimation for that subgroup.

Mitigation: Used published CYP2C19 covariate priors from CPIC guidelines; validated against external popPK literature.

Backup compound CS-004 had limited preclinical tox (21-day rat only). Root cause: accelerated internal timeline.

Mitigation: Applied read-across DILI liability from structural analogs; flagged wider confidence intervals in board deck.

When to use ClinicalSim for hepatotoxicity prediction

Phase I clean but preclinical tox shows reversible liver signals at elevated exposure
Planned Phase II doses approaching or exceeding preclinical NOAEL margins
Target population includes high co-medication burden (statins, antifungals, CYP inhibitors)
Board or investor milestone requires quantified safety risk before capital commitment

ROI: approximately 57:1 ($16.2M avoided spend vs. $285K simulation cost).

Next steps: run ClinicalSim toxicity simulation on backup compounds before Phase I; integrate adaptive liver monitoring rules into protocol design.

About This Engagement

Client profile: Series B metabolic biotech, 28 employees, 1 Phase I asset + 1 backup
Project duration: 3 weeks (simulation delivery) + 8 months (outcome tracking)
Total cost: $285K
Date: February to March 2025

This case study is anonymized at client request. Compound identifiers, target pathway, and institutional affiliations have been redacted. Full simulation protocols available under NDA.

Run a similar program with ClinicalSim.

Tell us about your target. We will scope a 2 to 4 week pilot.

Schedule a Consultation [email protected]