QPPV.COM

eDISH: Evaluating Drug-Induced Serious Hepatotoxicity in Clinical Trials

Understand how eDISH plots work, what the principal regions represent, how eDISH supports Hy's Law assessment and clinical-trial liver-safety review, and why an eDISH outlier requires medical evaluation rather than automatic attribution to the investigational product.

Audio Lesson 12 min
Knowledge Assessment Test your understanding of this article. Take the assessment →

eDISH: Evaluating Drug-Induced Serious Hepatotoxicity in Clinical Trials

Introduction

Drug-induced liver injury (DILI) is one of the most important safety concerns that may emerge during clinical drug development. Mild or moderate elevations of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) are relatively common and have many possible causes. Much less commonly, a medicinal product may cause severe liver injury that progresses to hepatic failure, transplantation or death.

The central challenge in clinical development is therefore not simply to identify abnormal liver tests. It is to identify patterns that may indicate a clinically important drug-related liver-safety risk while distinguishing those patterns from background abnormalities and alternative causes.

The Evaluation of Drug-Induced Serious Hepatotoxicity (eDISH) approach was developed to help address this problem.

eDISH provides a graphical way of examining liver-test data across clinical-trial subjects. It commonly plots the maximum or relevant post-baseline ALT or AST value, expressed relative to the upper limit of normal (ULN), against total bilirubin, also expressed relative to ULN. The resulting distribution allows investigators to identify subjects whose laboratory findings fall into regions requiring particular attention.

The approach is closely associated with Hy's Law.

Hy's Law identifies a concerning pattern of hepatocellular injury accompanied by bilirubin elevation without a more plausible alternative explanation. eDISH provides a population-level visualisation that helps investigators find subjects who may warrant evaluation for such a pattern.

The distinction is important.

An eDISH plot does not diagnose DILI.

It does not establish causality.

It does not determine whether a patient has acute liver failure.

It does not replace medical review.

Instead, eDISH is best understood as a clinical-trial liver-safety screening and visualisation tool.

FDA guidance on premarketing DILI evaluation emphasises the importance of evaluating liver-test abnormalities during drug development because severe DILI is uncommon and may not be directly observed even in a clinical-development programme. Appropriate evaluation of lesser biochemical abnormalities can nevertheless provide an early signal of potential severe liver injury. [1]

For pharmacovigilance and clinical-development professionals, understanding eDISH therefore requires more than knowing how to produce a scatter plot. The reviewer needs to understand what the axes represent, how the regions should be interpreted, how baseline abnormalities affect interpretation, how treatment groups should be compared, and how subjects identified by the plot should subsequently be medically evaluated.

Learning Objectives

After reading this article, the reader should be able to:

Understanding eDISH

What Does eDISH Mean?

eDISH stands for Evaluation of Drug-Induced Serious Hepatotoxicity.

The approach was developed as a way to organise and visualise clinical-trial liver-test data so that subjects with potentially important combinations of abnormalities could be identified more readily.

The original eDISH publication described the approach as a graphical organisation of liver laboratory data, displaying peak serum ALT and total bilirubin levels for individual subjects and allowing the reviewer to connect plotted subjects with the underlying clinical and laboratory information. [2]

The approach became particularly useful because clinical trials may contain thousands of subjects and very large numbers of laboratory observations.

A conventional laboratory table can show whether individual values are abnormal.

It is much less effective at showing the relationship between two different liver-safety markers across an entire study population.

eDISH addresses this problem by displaying the relationship graphically.

Why Was eDISH Developed?

The development of eDISH reflects a basic problem in drug development.

Severe DILI is uncommon.

A clinical programme may therefore contain many subjects with elevated ALT but no subject with acute liver failure.

The absence of observed liver failure does not necessarily demonstrate the absence of a clinically important hepatotoxicity risk.

Investigators therefore need methods capable of identifying earlier biochemical signals.

ALT or AST elevation can indicate hepatocellular injury.

Bilirubin elevation provides additional information because significant bilirubin elevation occurring with hepatocellular injury may indicate a more serious disturbance.

Plotting these variables together makes potentially important combinations easier to identify.

FDA guidance on premarketing clinical evaluation of DILI describes the importance of evaluating such liver-test signals during drug development. [1]

eDISH Is a Visualisation Tool

The most important conceptual point is that eDISH is a visualisation tool.

It does not itself perform the complete medical assessment.

The plot can answer questions such as:

It cannot independently answer:

Those questions require medical review.

Anatomy of an eDISH Plot

The Aminotransferase Axis

A conventional eDISH plot displays the maximum or relevant post-baseline ALT or AST value on one axis, usually expressed as a multiple of ULN.

For example:

The use of multiples of ULN allows laboratory results from different reference ranges to be represented on a comparable scale.

The exact implementation can vary between analyses.

The analysis specification should therefore document which laboratory variable is used, how the maximum value is defined and which time period is considered.

The Bilirubin Axis

The other axis commonly represents total bilirubin, again expressed as a multiple of ULN.

For example:

The combination of aminotransferase and bilirubin results is central to the assessment of potential serious hepatocellular injury.

Why Logarithmic Scales Are Useful

eDISH plots commonly use logarithmic scales.

This is useful because liver-test abnormalities can span a very wide range.

A subject with ALT of 3× ULN and another with ALT of 100× ULN are both clinically relevant, but plotting them on a simple linear scale can make the lower values difficult to distinguish.

A logarithmic representation allows a broad range of values to be displayed more effectively.

The precise graphical implementation should be documented in the analysis plan or reporting specification.

Maximum Versus Time-Specific Values

An eDISH analysis commonly uses maximum post-baseline laboratory values.

However, the definition of "maximum" must be understood.

For example, an analysis may use:

Using independently occurring maxima can sometimes produce a combination that did not occur at the same time in the patient.

This is an important limitation.

Medical review should therefore examine the time course of the laboratory results, not simply the two plotted maxima.

The Major eDISH Regions

The Hy's Law Region

The upper-right region of the traditional eDISH plot contains subjects with substantial aminotransferase elevation together with substantial bilirubin elevation.

The commonly used thresholds are:

Subjects in this region require particular attention.

However, being located in the region does not automatically mean that the subject has a Hy's Law case.

The reviewer must still evaluate:

A subject in the upper-right region is therefore best described initially as a potential Hy's Law-range subject rather than automatically as a confirmed Hy's Law case.

The Temple's Corollary Region

The lower-right region contains subjects with substantial aminotransferase elevations without substantial bilirubin elevation.

This region is commonly associated with Temple's Corollary.

The concept is important because subjects with significant ALT or AST elevation but no concurrent bilirubin elevation may represent an earlier or less severe manifestation of drug-related hepatocellular injury.

Frequent monitoring in clinical trials may identify these subjects before bilirubin rises.

The presence of subjects in this region can therefore provide useful information about the liver-safety profile of a treatment.

It should not, however, be interpreted as evidence of serious hepatotoxicity by itself.

The Cholestatic or Hyperbilirubinaemia Region

The upper-left region contains subjects with bilirubin elevation without substantial aminotransferase elevation.

This pattern can be associated with cholestasis or other causes of hyperbilirubinaemia.

Potential explanations include:

The region is therefore clinically important even though it is not the classical Hy's Law region.

The Lower-Left Region

The lower-left region generally contains subjects without substantial elevations in either plotted parameter.

This region usually represents the majority of subjects in a clinical trial.

The value of eDISH comes from identifying the smaller number of subjects distributed elsewhere and understanding whether those patterns differ between treatment groups.

eDISH and Hy's Law

Hy's Law Provides the Clinical Concept

Hy's Law and eDISH are closely related but should not be treated as synonyms.

Hy's Law describes a safety concept involving:

eDISH displays selected laboratory findings that can identify subjects who may require evaluation for this pattern.

The relationship can therefore be represented as:

eDISH

↓

Identify subjects with concerning laboratory combinations

↓

Medical review

↓

Determine injury pattern

↓

Evaluate competing causes

↓

Assess causality

↓

Determine whether a convincing Hy's Law case exists

The graphical plot is therefore only one part of the overall assessment.

Why an eDISH Outlier Is Not a Hy's Law Case

Consider a subject in the upper-right region.

The laboratory results might show:

ALT = 8× ULN

and:

total bilirubin = 3× ULN.

The subject could appear prominently in the Hy's Law region.

However, further evaluation might demonstrate:

The subject would still be an important liver-safety case.

But the drug might not be responsible.

This is why eDISH should identify subjects for investigation rather than assign causality automatically.

Why eDISH and Hy's Law Are Complementary

The two approaches solve different parts of the same problem.

Hy's Law provides a clinical and regulatory safety concept.

eDISH provides a population-level graphical screening method.

The combination allows investigators to move from:

large dataset

to:

potentially important subjects

to:

individual medical review

to:

drug-level safety assessment.

Baseline Liver Tests

Why Baseline Matters

One of the most important limitations of simple eDISH interpretation is that many patients entering modern clinical trials do not have completely normal liver tests.

Patients may have:

A subject may therefore appear in a concerning region because the baseline liver tests were already abnormal.

The investigator must distinguish treatment-emergent injury from pre-existing abnormality.

Baseline Is Not Necessarily One Value

A single screening laboratory result may not adequately represent baseline status.

Some subjects have fluctuating liver tests.

For such subjects, the safety team may need to examine:

This becomes particularly important when interpreting subjects with chronic liver disease.

Multiples of ULN Can Be Misleading

Consider two subjects.

Subject A:

Baseline ALT = 1.0× ULN

Peak ALT = 6× ULN

Subject B:

Baseline ALT = 4× ULN

Peak ALT = 6× ULN

The two subjects have the same peak ALT relative to ULN.

Their clinical interpretation may nevertheless be very different.

Subject A developed a substantial treatment-emergent abnormality.

Subject B may have returned to a previously observed baseline range.

This is why eDISH should be interpreted together with baseline-adjusted analyses and longitudinal laboratory data.

Treatment and Control Groups

Why Comparison Matters

An eDISH plot is more informative when treatment groups can be compared.

Background ALT and bilirubin abnormalities occur in clinical-trial populations.

A pattern that occurs equally in placebo and active treatment groups may have a different interpretation from a pattern concentrated in the investigational treatment group.

The analysis should therefore consider:

Dose Response

If potentially concerning subjects become more frequent with increasing dose, the finding may increase concern.

However, dose response is not mandatory for idiosyncratic DILI.

Therefore:

absence of a dose relationship does not exclude causality.

Similarly, an apparent dose relationship does not prove causality.

It is one component of the overall evidence.

Exposure Duration

The timing of laboratory abnormalities can also matter.

A cluster appearing early after treatment initiation may suggest a different mechanism from a pattern emerging only after prolonged exposure.

The safety team should therefore examine:

Individual-Subject Medical Review

From Plot to Patient

The eDISH plot identifies subjects.

The next step is to return to the individual clinical data.

A medical reviewer should reconstruct the subject's liver-safety timeline.

This may include:

The objective is to understand the clinical story behind the point on the graph.

The Importance of ALP

ALT and bilirubin are the primary variables displayed in a traditional eDISH plot.

ALP is nevertheless essential to the subsequent medical assessment.

A subject with ALT and bilirubin elevations may have:

The R-value can help characterise the pattern.

The classical Hy's Law concept is primarily concerned with hepatocellular injury.

Therefore, an eDISH point alone cannot establish whether the subject has the relevant injury pattern.

Clinical Context

Laboratory results should be interpreted alongside:

Clinical context can materially alter the interpretation of the same laboratory pattern.

eDISH and Causality Assessment

eDISH Does Not Determine Causality

Causality requires a broader assessment.

The reviewer should consider:

An eDISH plot provides evidence that a laboratory pattern exists.

It does not identify its cause.

Competing Causes

Potential competing causes may include:

The extent of investigation should be proportionate to the clinical circumstances.

Concomitant Medicines

A particularly important issue in clinical trials is polypharmacy.

A subject may receive several medicines during the trial.

If liver injury occurs, the investigational product is not automatically the most likely cause.

The reviewer should construct a medication timeline and assess each potentially hepatotoxic exposure.

Dechallenge and Rechallenge

Improvement after treatment discontinuation may support causality.

Recurrence following re-exposure can provide stronger evidence but may carry significant clinical risk.

Deliberate rechallenge should not ordinarily be undertaken merely to establish a pharmacovigilance diagnosis.

Where inadvertent or clinically necessary rechallenge occurs, the information should be carefully evaluated.

eDISH at the Study Level

Counting Subjects in Each Region

One useful application of eDISH is to compare the number and proportion of subjects in areas of interest.

The analysis may examine:

The absolute number alone is not sufficient.

The denominator matters.

Exposure-Adjusted Interpretation

Two studies may identify the same number of potential cases but have very different exposure.

For example, three cases among 300 treated subjects represent a very different signal from three cases among 30,000 treated subjects.

The safety team should therefore consider:

Cluster Versus Isolated Outlier

A single subject can be clinically important.

However, multiple subjects showing similar patterns may provide stronger evidence of a drug-level signal.

The safety team should therefore examine whether subjects:

eDISH and Population-Level Signal Detection

From Individual Cases to a Drug Signal

eDISH can contribute to the progression from individual observations to population-level safety assessment.

A possible sequence is:

  1. eDISH identifies subjects of interest.
  2. Medical review evaluates each subject.
  3. Potential DILI cases are adjudicated.
  4. Cases are compared with controls and background rates.
  5. Patterns across subjects are assessed.
  6. The overall drug-related liver-safety signal is evaluated.

This approach helps prevent two opposite errors:

Signal Detection Is Not Case Counting

Counting subjects in the Hy's Law region is not equivalent to counting confirmed Hy's Law cases.

Each subject requires appropriate medical evaluation.

A case may be excluded from a drug-related signal because of:

Conversely, clinically important drug-induced liver injury may occur outside the classical Hy's Law region.

Signal detection therefore requires broader clinical judgement.

eDISH and Cholestatic Liver Injury

The Limitation of the Traditional Plot

Traditional eDISH focuses heavily on ALT or AST and total bilirubin.

This makes it particularly useful for identifying hepatocellular patterns associated with classical Hy's Law.

However, not all serious DILI is hepatocellular.

Drug-induced cholestatic injury can also result in significant morbidity.

Modern DILI experience has therefore raised questions about whether liver-safety visualisation should also incorporate measures capable of identifying cholestatic patterns more directly.

Recent FDA regulatory reviews have used eDISH-AP, a complementary plot examining maximum alkaline phosphatase and total bilirubin values, to screen for potential cholestatic DILI. [3]

Why Cholestatic DILI Matters

A drug may cause:

A traditional ALT-bilirubin eDISH plot is strongest for the first category.

It should not be interpreted as a comprehensive visualisation of every possible form of drug-related liver injury.

Modern Applications

Recent literature has highlighted several factors that challenge the universal application of traditional Hy's Law and eDISH concepts, including:

These developments do not make eDISH obsolete.

They reinforce the need to interpret it as one component of a broader liver-safety framework. [4]

Practical eDISH Review Workflow

Step 1: Define the Dataset

Before producing the plot, specify:

Step 2: Calculate Relative Values

Convert the relevant laboratory values into multiples of ULN.

For example:

ALT relative value = ALT / ALT ULN

Bilirubin relative value = total bilirubin / bilirubin ULN

Step 3: Identify Subjects of Interest

Identify subjects crossing predefined thresholds.

The traditional Hy's Law region uses:

Step 4: Review Baseline

Determine whether the abnormalities were:

Step 5: Review ALP

Determine whether the pattern is:

Step 6: Evaluate Competing Causes

Review:

Step 7: Assess Clinical Course

Review:

Step 8: Assess the Study-Level Pattern

Finally, evaluate:

Worked Example: Interpreting an eDISH Plot

A Hypothetical Clinical Trial

Consider a placebo-controlled clinical trial with:

Suppose the eDISH analysis identifies:

The upper-right quadrant therefore contains several subjects in both groups.

This finding requires further evaluation.

The correct conclusion is not:

"The drug caused four Hy's Law cases."

Instead, the safety team should review each subject.

Case-Level Review

Suppose review establishes:

The final interpretation would be substantially different from the raw eDISH counts.

The eDISH plot successfully identified the subjects.

Medical review determined what the points actually represented.

Study-Level Interpretation

The safety team should then determine whether the remaining convincing case, together with other ALT elevations and the overall study data, represents a meaningful drug-related liver-safety signal.

This illustrates the proper relationship between graphical analysis and clinical judgement.

Common Mistakes in eDISH Assessment

Common errors include:

What an Experienced Safety Physician Looks For

An experienced safety physician does not look at an eDISH plot and immediately ask:

"How many Hy's Law cases are there?"

The first questions are broader:

"What is the distribution of liver-test abnormalities?"

"Is there an imbalance between treatment groups?"

"Which subjects require individual review?"

"What happened to those subjects clinically?"

The reviewer then examines:

Only after this review can the eDISH findings contribute meaningfully to a drug-level liver-safety assessment.

Pharmacovigilance and Inspection Considerations

Inspection Perspective

An organisation should be able to demonstrate how clinical-trial liver-safety signals were identified and evaluated.

Relevant evidence may include:

The important issue is not merely whether an eDISH plot exists.

The organisation should be able to demonstrate what decisions were made from the plot and why.

Documentation Expectations

For subjects identified as potentially important, documentation should allow an independent reviewer to reconstruct:

A plot without corresponding medical reasoning is insufficient to demonstrate a robust safety assessment.

Integration With Pharmacovigilance

Although eDISH is primarily associated with clinical development, the principles behind it can inform broader pharmacovigilance activities.

After approval, safety teams may encounter DILI through:

The data structure is different from a controlled clinical trial.

Nevertheless, the same principles remain relevant:

Limitations of eDISH

eDISH Is Not a Diagnostic Tool

The most important limitation is that eDISH is not a diagnostic test.

It identifies laboratory patterns.

Clinical diagnosis requires additional evidence.

eDISH Does Not Establish Causality

A subject may fall into a concerning region for reasons unrelated to the investigational product.

Therefore, eDISH should never be used as a substitute for causality assessment.

eDISH Can Miss Important DILI

Not every serious DILI case produces the classical ALT-bilirubin pattern.

Examples include:

Maximum Values Can Lose Temporal Information

A plot based on independent maximum values may suggest that two abnormalities occurred together when they did not.

For example:

ALT may peak at week 4.

Bilirubin may peak at week 12.

The eDISH point may nevertheless plot the maximum of both.

The clinical reviewer must therefore return to the longitudinal laboratory data.

Underlying Disease Can Complicate Interpretation

Modern clinical trials increasingly include patients with chronic diseases and abnormal baseline liver tests.

This can reduce the specificity of simple threshold-based interpretations.

The more abnormal the baseline population, the more important longitudinal and comparator-based interpretation becomes.

eDISH in Modern Drug Development

Beyond the Traditional Hy's Law Plot

The traditional eDISH framework remains valuable.

However, modern drug development increasingly requires more nuanced analysis.

Important considerations include:

Emerging Approaches

Recent research has proposed modifications to traditional Hy's Law and eDISH approaches to account for modern DILI phenotypes and clinical-development populations. [4]

These proposals should be understood as evolving scientific approaches rather than automatically replacing established regulatory frameworks.

For regulatory submissions, the applicable regulatory guidance and agreed analysis strategy remain fundamental.

Key Takeaways

eDISH stands for Evaluation of Drug-Induced Serious Hepatotoxicity.

It is a graphical clinical-trial liver-safety tool designed to help investigators visualise the relationship between aminotransferase and bilirubin abnormalities across subjects.

The traditional eDISH plot commonly displays:

The upper-right region corresponds to the traditional laboratory range associated with potential Hy's Law concern.

The lower-right region can identify subjects with substantial aminotransferase elevations without substantial bilirubin elevation and is commonly associated with Temple's Corollary.

The upper-left region may identify subjects with bilirubin elevation without substantial aminotransferase elevation and can include cholestatic or non-hepatocellular processes.

The most important principle is:

An eDISH plot identifies subjects for medical review; it does not diagnose DILI or establish causality.

A robust eDISH assessment therefore requires:

eDISH remains highly useful because it converts a large volume of clinical-trial laboratory data into a form that allows potentially important patterns to be recognised rapidly.

Its greatest value is therefore not the plot itself.

Its value lies in connecting population-level laboratory surveillance with structured individual medical review.

References

  1. U.S. Food and Drug Administration. Drug-Induced Liver Injury: Premarketing Clinical Evaluation. Guidance for Industry. July 2009.

  2. Watkins PB, Desai M, Berkowitz SD, Peters G, Horsmans Y, Larrey D, Maddrey W. Evaluation of drug-induced serious hepatotoxicity (eDISH): application of this data organization approach to phase III clinical trials of rivaroxaban after total hip or knee replacement surgery. Drug Safety. 2011;34(3):243-252. doi:10.2165/11586600-000000000-00000.

  3. U.S. Food and Drug Administration. FDA clinical review materials describing eDISH-AP, a cholestatic drug-induced liver injury screening plot based on alkaline phosphatase and total bilirubin.

  4. Cohen EB, Lewis JH. Modifying Hy's Law and eDISH for the modern idiosyncratic DILI era. Expert Opinion on Drug Metabolism & Toxicology. 2025;21(10):1159-1168. doi:10.1080/17425255.2025.2577677.

  5. Senior JR. Evolution of the Food and Drug Administration approach to liver safety assessment for new drugs: current status and challenges. Drug Safety. 2014;37 Suppl 1:S9-S17. doi:10.1007/s40264-014-0182-7.

  6. Temple R. Hy's law: predicting serious hepatotoxicity. Pharmacoepidemiology and Drug Safety. 2006;15(4):241-243. doi:10.1002/pds.1211.

Revision History

Related Articles

Hy's Law: Identifying Potential Serious Drug-Induced Liver Injury

Understand Hy's Law, how potential cases are identified and medically evaluated, how eDISH supports population-level liver-safety assessment, and why the concept is a signal of potential serious hepatotoxicity rather than an individual diagnostic test.

Drug-Induced Liver Injury (DILI): From Liver Test Abnormalities to Causality and Regulatory Safety Assessment

Understand how drug-induced liver injury is recognised and characterised, how biochemical patterns and clinical context inform causality assessment, and how DILI evidence is translated into pharmacovigilance and regulatory safety decisions.

Drug-Induced Cardiac Toxicity: Cardiovascular Safety Assessment in Clinical Development and Pharmacovigilance

Understand how potential drug-induced cardiac toxicity is identified and medically evaluated, how ICH E14 and S7B support assessment of QT/QTc and proarrhythmic risk, and how cardiovascular safety signals are assessed across clinical development and pharmacovigilance.