When Liver Signals Stall Late-Stage Pipelines: The Critical Role of Pre-Market DILI (Drug-Induced Liver Injury) Assessment

A recent report by BioSpace “BMS reports liver injury events in lung disease program, with key readout near” detailing liver injury events in Bristol Myers Squibb’s lung disease program serves as a stark reminder of an enduring reality in drug development: Drug-Induced Liver Injury (DILI) remains one of the primary drivers of clinical holds, regulatory rejections, and post-marketing withdrawals.

For clinical development teams and biostatisticians alike, evaluating hepatic safety is not merely a routine check of routine laboratory panels. It is an intricate, multi-dimensional assessment where subtle shifts in baseline aminotransferases and bilirubin can mean the difference between FDA NDA/BLA approval and a Complete Response Letter (CRL).

The Regulatory Framework: FDA Premarketing Guidance on DILI

The cornerstone of hepatic safety review is the FDA’s guidance, Drug-Induced Liver Injury: Premarketing Clinical Evaluation. The agency emphasizes that the liver possesses immense functional reserve; hepatocellular damage can occur extensively before overt organ dysfunction appears. Consequently, pre-market evaluation centers on distinguishing benign, transient enzyme elevations (adaptation) from signals of impending acute liver failure.

The Gold Standard: Hy’s Zimmerman’s Rule (Hy’s Law)

The FDA considers Hy's Law the single most reliable predictor that a drug carries severe hepatotoxic liability. Historically, finding even a single true Hy’s Law case in a clinical trial database implies a mortality risk of roughly 10% for severe DILI, signaling that broader population exposure could result in fatal liver failure.

To classify a subject as a Potential Hy’s Law Case, three core criteria must be met concurrently or in close temporal proximity:

1. Hepatocellular Injury: Aminotransferase elevation (ALT or AST)  Upper Limit of Normal (ULN).
2. Impaired Clearance/Excretion: Total Bilirubin (TBL)  ULN, without initial signs of cholestasis.
3. Absence of Cholestasis: Alkaline Phosphatase (ALP)  ULN (or an -ratio , pointing to hepatocellular rather than cholestatic injury).
4. Alternative Etiology Ruled Out: Comprehensive workups negative for viral hepatitis (A, B, C, E), autoimmune hepatitis, biliary obstruction, alcoholic liver disease, or ischemic hepatopathy.

FDA Standard Safety Tables & Visualizations for DILI

Under the FDA CDER Standard Safety Tables and Figures (ST&F) integrated review guidelines, hepatic safety requires standardized tabulations and visual analytics to systematically parse DILI risk across treatment arms.

1. Categorical Shift Tables (Threshold Counts)

Sponsors are expected to summarize the incidence of peak post-baseline laboratory elevations relative to ULN:

Parameter	Standard Evaluation Thresholds	Clinical & Regulatory Interpretation
ALT / AST	, , ,  ULN	Detects hepatocellular necrosis; helps differentiate mild adaptation () from severe necrosis ().
Total Bilirubin	,  ULN	Evaluates clearance capacity;  ULN triggers Hy's Law screening.
Alkaline Phosphatase (ALP)	, ,  ULN	Differentiates cholestatic injury patterns from hepatocellular damage.
INR		Assesses synthetic function failure (coagulopathy).

2. Standardized Figures: eDISH and mDISH

Visual exploration of DILI datasets relies primarily on the eDISH (evaluation of Drug-Induced Serious Hepatotoxicity) plot, a standard log-log scatter plot of individual patient peak post-baseline lab values:

• Horizontal Axis (X-axis): Peak ALT or AST ( fold of ULN), split by the  ULN threshold.
• Vertical Axis (Y-axis): Peak Total Bilirubin ( fold of ULN), split by the  ULN threshold.

     

• Upper-Right Quadrant (Hy’s Law Zone): Patients with ALT  ULN and TBL  ULN. Points falling here immediately trigger multidisciplinary adjudication.
• Lower-Right Quadrant (Temple’s Zone / Hepatocellular Necrosis): Patients with ALT  ULN but TBL  ULN. Indicates enzyme leakage without functional clearance compromise (often adaptive, but flags liability).
• Upper-Left Quadrant (Cholestatic / Gilbert’s Syndrome): Elevated bilirubin without transaminase surge.
• mDISH (Modified eDISH): Incorporates the -ratio () or color-codes points by ALP

          elevation to weed out obstructive cholestasis from true hepatocellular Hy’s Law events.

Regulatory Casualties: The Solithromycin Precedent

The pharmaceutical graveyard contains numerous programs halted by hepatic signals, but few provide a clearer case study in pre-market DILI review than Cempra Pharmaceuticals’ Solithromycin.

In 2016, Cempra submitted an NDA for solithromycin, a fourth-generation macrolide/ketolide intended to treat Community-Acquired Bacterial Pneumonia (CABP). Ketolides carried a notorious precedent: Aventis’ Ketek (telithromycin) was approved in 2004 but subsequently linked to severe liver failure and death post-marketing, leading to strict boxed warnings and restricted indications.

When the FDA evaluated Solithromycin’s NDA:

1. Premarket Signal Detection: In Phase 3 trials involving approximately 920 patients exposed to oral solithromycin, transaminase elevations (ALT  ULN) occurred at a rate significantly higher than in the active control arm (moxifloxacin). While no definitive Hy's Law mortality occurred in the ~1,000-patient registration cohort, the frequency of marked ALT elevations (including patients exceeding  and  ULN) was troubling.
2. Statistical Modeling and Rare Event Risk: Given the telithromycin precedent, FDA clinical reviewers and biostatisticians used the Rule of Three and binomial probability models to estimate that the true incidence of severe liver injury could be as high as 1 in 300 to 1 in 500 patients.
3. The Advisory Committee and CRL: The Antimicrobial Drugs Advisory Committee concluded that the pre-market safety database was too small to characterize the risk of catastrophic liver failure in a common indication like CABP. In December 2016, the FDA issued a Complete Response Letter (CRL), requiring a post-baseline safety study of roughly 9,000 to 10,000 patients to rule out severe DILI prior to approval—a requirement that effectively halted the drug's commercial path.

Other notable regulatory actions driven by DILI include premarket 'not approval'/complete response letter or post-marketing market withdrawal due to the hepatotoxicity:

• Exanta (ximelagatran): An oral direct thrombin inhibitor denied FDA approval in 2004 after approximately 8% of treated patients developed ALT  ULN, accompanied by cases of fatal acute liver injury.
• Lumiracoxib (Prexige): Withdrawn globally and denied US approval due to high rates of serious hepatotoxicity. 

• Troglitazone (Rezulin): Fast-tracked for Type 2 diabetes in 1997, only to be withdrawn in 2000 after accumulating at least 90 cases of liver failure and 63 deaths.

Strategic Takeaways for Late-Stage Programs

When an investigational agent—such as an IPF or chronic lung disease candidate—shows late-stage hepatic signals, sponsors must treat it as an existential program milestone:

1. Protocol-Defined Stopping Rules: Ensure trials implement the strict FDA-recommended cessation triggers: ALT/AST  ULN; ALT/AST  ULN persisting  weeks; or ALT/AST  ULN accompanied by TBL  ULN or INR .
2. Independent Hepatic Adjudication Committees (EAC): Blinded, prospective evaluation by independent hepatologists is mandatory to differentiate background comorbidities (e.g., right heart failure, passive hepatic congestion in respiratory patients) from intrinsic or idiosyncratic DILI.
3. Comprehensive Re-Challenge/De-Challenge Tracking: Monitor the kinetic resolution of ALT and bilirubin upon drug withholding. True DILI should show rapid recovery kinetics unless fulminant failure is underway.
4. Indication Context: Regulators calibrate risk-benefit tolerance based on clinical unmet need. In fatal progressive disorders like idiopathic pulmonary fibrosis (IPF), some hepatic lab shifts may be acceptable if accompanied by robust efficacy and clear mitigation strategies (e.g., intensive liver function monitoring regimens similar to pirfenidone or nintedanib). However, any confirmed Hy's Law signal without a defined monitoring pathway remains an insurmountable hurdle.

Immediate Implementation Checklist for Biostatistics & Clinical Safety

1. Audit SDTM / ADaM Datasets: Ensure ADLBC contains derived flags for Hy's Law criteria (R2ANRHI for peak transaminases and bilirubin) and time-to-onset flags within 30 days of initial elevation.
2. Generate Specific Summary Tables or Shift Tables for assessing DILI according to FDA's Standard Safety Tables and Figures (ST&F) integrated review guideline
3. Automate eDISH Generation: Run dynamic eDISH displays with -ratio stratification across unblinded interim cuts to detect emergent clustering in the upper-right quadrant.
4. Standardize Workup Case Report Forms (CRFs): Verify clinical sites automatically execute acute hepatitis panels, abdominal ultrasounds, and concomitant medication reconciliation for any subject crossing the  ALT +  TBL threshold.