Understanding and predicting dose dependent and idiosyncratic events

® 

Modeling Hepatoxicity: 

Understanding and predicting dose 

dependent and idiosyncratic events. 

1

Disclosure 

I chair the scientific advisory committee for the 

DILIsim Initiative and own 

equity in the commercial spin off company 

DILIsym Services Inc. 

2

Outline of Talk 

1). Problem of liver safety 

2). The DILIsim Initiative 

3). Progress/challenges 

4). Conclusions

Reasons for Termination of Programs due to Safety by 

Organ System 

4

Posted: 5:17 p.m. Friday, Nov. 4, 2016 

The Associated Press 

WASHINGTON — 

Cempra is one of a handful of drugmakers developing new 

antibiotics amid growing bacterial resistance to decades-old 

drugs like penicillin. 

On Wednesday Cempra shares plunged more than 60 

percent after the FDA posted an online review highlighting 

irregular liver enzyme measurements reported with the drug, 

called solithromycin…….. 

Copyright The Associated Press


1). Problem of drug safety 


3). Progress 

4). Conclusions

UNC Institute for Drug Safety Sciences 

In Silico Modeling 

DILIsym ® 

Patients 

In Vitro 

Cutting Edge 

Pre-clinical Models 

7

DILIsym ® : 'Middle Out' and Multi-Scale 

Mitochondrial dysfunction 

Cellular life-cycle 

Drug distribution 

& metabolism 

Patient variability 

(SimPops) 

Kuepfer 2010, Molecular Systems Biology 

8

Approach 

1). Build mechanistic “modules” using differential equations 

– perform experiments to fill in knowledge gaps. 

2). Integrate the modules with the outcome of hepatocyte 

death and release of traditional and novel serum 

biomarkers. 

3). Vary model parameters to create simulated patient 

populations (Simpops®) 

4). Refine the aggregate model through incorporating data 

obtained from successive “exemplar” drugs

DILIsym ® Modules 

10





4). Conclusions

DILIsym ® Modules 

12

Drugs Can Inhibit Bile Acid Transporters 

Hepatotoxicity 

MRP3/4 

BSEP 

Bile 

Acids 

Drug 

OĀ 

NTCP 

(OATP) 

Bile 

Acids 

Drug

Drug PBPK model 

BA Transport Inhibition Model 

Bile Acids and DILI 

Drug inhibits BA 

transporters 

Bile Acid Homeostasis Model 

Cellular ATP Model 

Bile acid 

accumulation 

disrupts cellular 

energy balance 

BA H+ 

Gradient 

Uncoupling 

Hepatocyte Life-Cycle 

Disrupted cellular 

energy balance can 

cause hepatocyte 

necrosis 

14

AMG 009 

No evidence of liver injury in multiple species 

• Rats, mice, non-human primates, hamsters or rabbits 

– During Phase I clinical trials in healthy volunteers, 5/8 patients 

showed significant and reversible transaminase elevations 

– Development of AMG 009 was halted 

– Bile acid transporter inhibition was the only mechanism 

identified as likely contributors to AMG 009 hepatotoxicity 

• No reactive metabolites, covalent binding, or mitochondrial toxicities 

were detected 

15

Some AMG 009 Transporter Inhibition Data 

X 

Bile acids 

AMG009 

Ryan Morgan - Amgen

Modeling in vitro transporter 

inhibition 

Bile acids 

Drug 

Transporter 

Inhibition 

constants 

Model average 

human 

Model simulated 

patient 

population 

(Simpops TM )

AMG 009 Modeling Approach 

PBPK Modeling 

• In vitro PK data 

• In vivo PK profile 

• Parameter optimization 

Mechanistic 

Toxicity Data 

•Transporter inhibition 

for AMG 009 (Ki, type of 

inhibition) 

PBPK model and toxicity data 

inputs were not modified after 

hepatotoxicity simulations 

Stop criteria 

used* 


Simulation 

• Baseline 

• Population (SimPops) 

Combine PK and in vitro Tox 

Further optimization using 

AMG 009/853 clinical 

toxicity data was not 

necessary 

* 

ALT screening before AM dosing on days 2,3,4,5,8,11, and 14 

If ALT>3X ULN, dosing discontinued after 24 h 

Employed to recapitulate the clinical protocol of AMG 009 

Compare to clinical data 

Clinical Data and 

Simulation Results 

18

DILIsym ® Predicts Dose-Dependent AMG 009 Hepatotoxicity in 

Human SimPops 

HUMANS 

100 mg BID 14 d Dose-Response 

(25-100 mg BID 14 d) 

Treatment 

Stop criteria 

used 

• DILIsym ® predicts dose-dependent, delayed 

presentation of AMG 009 hepatotoxicity and 

recovery after discontinuation 

• Incidence rates were fairly similar to observations 



19

No Hepatotoxicity Predicted in the 

Rat SimPops Administered AMG 009 

1500 mg/kg/day PO for 1 month 

1500 mg/kg PO 1 month † 

Direct BA Toxicity Model 

No. Deaths 0/187 

ALT > 3X 0/187 

Bili > 2X 0/187 

Treatment 

RATS 

Preclinical Data and 


20

AMG 853 

• AMG 853 was the backup to AMG 009 advanced into 

the clinic 

– No evidence of liver injury in preclinical species. 

– No evidence of human toxicity in Phase 1 or 2 

clinical trials 

– AMG 853 was a more potent BSEP inhibitor than 

AMG 009 with a Ki = 1.8 µM vs 2.4 µM 

21

DILSYM Modeling of AMG 853 

• DILIsym predicted that AMG 853 was safe in simulated 

humans (SimPops) although you would predict toxicity 

based on the BSEP Ki value….. 

Why? - 

Inhibition type was the key

Why mechanism of transport inhibition matters 


MRP3/4 

BSEP 

Bile 

Acids 

Drug 

OĀ 

NTCP 

(OATP) 

Bile 

Acids 

Drug

Conclusion 

QSP modeling was able to predict the 

hepatoxic potential of AMG 009 and the safety 

of AMG853 based on in vitro 

liability assessments




On Wednesday Cempra shares plunged more than 60 

percent after the FDA posted an online review highlighting 

irregular liver enzyme measurements reported with the drug, 

called solithromycin…….. 

What exactly did the FDA say?

What the FDA Briefing Document Said 

“serious liver safety concern” and suggested two pathways forward: 

1). “the number of study subjects treated with solithromycin should be 

increased from 924 to approximately 12,000 and carefully assessed for 

liver safety events, either in expanded Phase III randomized trials or in a 

large clinical safety study, in advance of making a regulatory decision 

regarding approval” 

2). “Alternatively, if solithromycin treatment has been found to convincingly 

offer a clinically substantial benefit over other currently approved 

treatments a second avenue might be considered….”

Two issues with solithromycin: 

– Imbalance in serum ALT elevations 

– Structural similarity with telithromycin

Structures of Macrolide Antibiotics

DILIsym ® Mechanism-Based Modeling 

Drug Properties 

• Oxidative stress 

• Mitochondrial dysfunction 

• Bile acid transporter inhibition 

Human population 

Liver Exposure 

PBPK Modeling 

• Compound Properties 

SimPops® 

• Tissue penetration studies 

• Pharmacokinetic data 

• in vitro data 

30

The rates of serum ALT elevations in clinical trials are 

reasonably predicted by DILIsym 

Compound 

Protocol 

Peak ALT > 3X ULN 

Observed 

Simulated 

Solithromycin 

Erythromycin 

Oral 

(CE01-300) 

IV-to-Oral 

(CE01-301) 

500 mg 

QID 10 days 

5.4% (3.2%) 3.9% 

9.1% (5.5%) 6.0% 

1-2% 2.8% 

Clarithromycin 500 mg BID 7 days 1-2% 2.8% 

Simulation Results and 

Clinical Data 

Telithromycin 800 mg QD 10 days 0.0-0.8% 0% 

data 

Data presented at Nov 4 2017 anti-infective Ad com 

31

Contribution to Predicted ALT elevations 

in Simulated Human Population 

DILI Mechanism Solithromycin Telithromycin Erythromycin Clarithromycin 

Mitochondrial 

Respiration 

Inhibition 

Predominant None None Predominant 

Oxidative 

Stress 

None None Minor None 

Bile Acid 

Transporter 

Inhibition 

Minor (Predominant) Predominant Minor 


32

Excerpt from FDA Briefing Document 

…. A somewhat surprising additional 

unexplained gap in the analysis submitted by 

DILIsym Services is the absence of the parallel 

testing of telithromycin hepatoxicity in a 

simulated CAP population……The use of 

telithromycin as a “positive control” in the 

model with comparative liver test data would be 

highly relevant and might support the utility of 

the model….”

Contribution to Predicted ALT elevations 

in Simulated Human Population 

DILI Mechanism Solithromycin Telithromycin Erythromycin Clarithromycin 

Mitochondrial 

Respiration 

Inhibition 

Predominant None None Predominant 

Oxidative 

Stress 

None None Minor None 

Bile Acid 

Transporter 

Inhibition 

Minor Predominant Predominant Minor 


34

Summary 

– Imbalance in serum ALT elevations 

• Well-characterized clinically and mechanistically 

– Structural similarity with telithromycin 

• Measured solithromycin effects on the liver differ

FDA panel narrowly backs Cempra antibiotic 




A panel of federal health advisers has narrowly recommended 

approval for an experimental antibiotic from Cempra Inc., a small 

North Carolina drugmaker. 

The Food and Drug Administration's outside experts voted 

7-6 in favor of the drug, saying its effectiveness outweighed 

risks of liver toxicity seen in company studies. The vote is 

nonbinding but the FDA often follows the advice of its panelists. 

Copyright The Associated Press

DILIsym ® Sub-models 

37

Multiple Steps Involved in 

Idiosyncratic DILI 

38



39



40

FDA Submission: A New Antibiotic 

that Caused Serum ALT Elevations 

in Phase 1 Study 

X X X 

Serum 

biomarkers 

ccK18 

CK18 

Oxidized 

HMGB1 

No acetylated 

HMGB1 

interpretation apoptosis 

41

Summary 

New mechanistic biomarkers promise to 

revolutionize the interpretation of liver safety 

in clinical trials and ultimately in the clinic. 

42





4). Conclusions

DILIsym ® Mechanism-Based Modeling 

Drug Properties 

• Oxidative stress 

• Mitochondrial dysfunction 

• Bile acid transporter inhibition 

Human population 

Liver Exposure 

PBPK Modeling 

• Compound Properties 

SimPops® 

• Tissue penetration studies 

• Pharmacokinetic data 

• in vitro data 

44

Optimal requirements for the cell system 

1). Metabolically competent 

2). Measure intracellular concentrations of 

parent and metabolites 

3). Identify the three major mechanisms 

4). Provide quantitative parameters for 

direct incorporation into DILIsym 

including PB/PK parameters 

5). Validation with the ~30 drugs successfully 

modeled already in DILIsym 

45

Confidence in Predictions 

Depends on Confidence in Data Used 

in vitro 

tox 

in vivo 

tox 

in vitro 

mechanistic 

in vivo 

mechanistic 

Clinical 

mechanistic 

Sims required for confidence in predictions 

Confidence in data for predictions




3). Successes and Challenges 

4). Conclusions

Conclusion 

Quantitative Systems Pharmacology 

is already having an impact on drug 

development decisions both within 

industry and regulatory agencies. 

The impact should rapidly increase in 

the years ahead.

49 

The DILI-sim Team Sept 2015

Special thanks to the DILIsim Partners 

DILIsym ® 

And Cindy Ashfari and Ryan Morgan - Amgen 

50

Thanks 

UNC Institute for Drug Safety 

Sciences Leaders 

DILIsim Team and SAB 

Mouse Genetics: Merrie Mosedale 

Human Genetics: Tom Urban 

Organotypic liver cultures: Ed LeCluyse 

Cellular Imagine Core: Joe Trask 

UNC –Hamner Organ Toxicity 

Biomarker Core: Rachel Church 

UNC-Hamner Metabolomics Core- Jeff 

MacDonald 

UNC Chapel Hill 

Kim Brouwer 

U. of Liverpool 

Kevin Park 

Dean Naisbitt 

Daniel Antoine 

Munir Pirmohamed 

Qualyst 

Ken Brouwer 

51

Back up

DILIsym ® Performance Review – Level 2 

• Key Question: how do the frequency, dose response effect, and mechanisms predicted compare to the observed cases? 

Drug 

Injury 

Frequency 

Injury Dose- 

Response 

Injury 

Severity 

Injury Timing 

Injury 

Mechanism 

Entacapone (Clean) 

Tolcapone (DILI) 

Methapyrilene (Clean) 

Troglitazone (DILI) 

Pioglitazone (Clean) 

AMG009 (DILI) 

Compound A (DILI) 

Bosentan (DILI) 

Telmisartan (Clean) 

Compound D (DILI) 

Compound B (DILI) 

Color Key – Accuracy of DILIsym® 

Excellent 

Good 

Fair 

Poor 

Unavailable 

HUMAN 

Compound C (DILI) 

Etomoxir (DILI) 

Compound E (DILI) 



53

Understanding and predicting dose dependent and idiosyncratic events

Create successful ePaper yourself

Delete template?

Save as template?