University of Pittsburgh Critical Care Medicine · 2012. 12. 12. · CRISMA Critical Care Medicine the University of Pittsburgh But… No single therapy specifically for sepsis Only

CRISMA Critical Care Medicine the University of Pittsburgh

University of Pittsburgh Critical Care Medicine

www.ccm.pitt.edu

the Clinical Research, Investigation, and Systems Modeling of Acute illness

CRISMA Center, Department of Critical Care Medicine

Department of Health Policy and Management McGowan Institute for Regenerative Medicine Clinical and Translational Science Institute

School of Medicine and Graduate School of Public Health University of Pittsburgh

Derek C. Angus, MD, MPH, FRCP

Making Inroads in Sepsis


Today’s talk

Introductory statement of the problem

Discussion of why sepsis has been difficult to study

Consideration of potential solutions under a ‘precision medicine’ framework


Sepsis …

... is a life-threatening condition when the body's response to an infection injures its own organs

… leads to shock, multiple organ failure and death especially if not recognized early and treated promptly

… is the primary cause of death from infection despite advances in modern medicine

Vaccines

Antibiotics

Intensive care

Millions of people die of sepsis every year worldwide #1 cause of death in US hospitals

#1 cause of disability-adjusted life-years lost worldwide


But…

No single therapy specifically for sepsis Only corticosteroids currently recommended

And there is considerable on-going controversy

Many (>50) failed trials despite Advances in understanding of biology

Energy

Money

Many clinical questions have No trials

Inconsistent trials

And, even positive trials have Methodologic challenges

Logistic challenges Angus DC JAMA 2011


ACCESS

Placebo-controlled phase 3 trial Prompted by (sort of) positive Phase 2 trial

Tidswell et al. Crit Care Med 2010

N=2000 patients with severe sepsis >300 sites worldwide

Eritoran Lipid A analog

Anti-TLR4 antagonist

28-day all-cause mortality

~ $1billion budget


PROWESS Shock

Drotrecogin alfa withdrawn from worldwide market



Today’s questions …

Are we identifying and enrolling the right patients?

Do we have the right strategies to generate novel drug targets?

Animal

T1 transitions

Do we deliver new therapies properly?

Is short-term mortality the right end-point?

Non-mortal outcomes

Do we have the right trial designs?

How can we stimulate new approaches? Are funders (government and industry) too risk averse?


And yet little changes …

The definition of insanity is to keep doing the same thing over and over, and expect different results

Albert Einstein

Currently 502 adult sepsis trials registered on clinicaltrials.gov 501 use ‘classic’ entry criteria and outcome

The one exception is based on an Italian design

Endotoxin assay to select patients for endotoxin removal


Why is sepsis so difficult?


Bone et al. Ann Intern Med 1992; Levy et al. CCM 2003

What do we think sepsis ‘is’?

Patients still die DESPITE effective antibiotics

Sepsis is a host response to infection gone awry! A case of harm by friendly fire

When organs fail, the sepsis is called ‘severe’ 1992 and 2003 International Consensus Definition

Trauma

Burns

Virus

Fungus

Bacteria

Parasite

Sepsis

Sepsis

Host

inflammatory

response

Infection


The host response to infectious challenge

6 healthy volunteers

Fong et al. J Clin Invest 1990;85:1896 Fong et al. J Immunol 1989;142:2321

TNF IL-6


The host inflammatory cascade - an inflammatory response gone awry -

Time

IL-10 RECOVERY

IL-6 TNF

IL-1

Anti-LPS

Anti-TNF

Anti-IL-1

Anti-IL-10

Blood Purif.

Immunologic Support

HMGB1


IL-6 and severe sepsis L

og

me

an

s

0

1

2

3

4

5

6

1 2 3 4 5 6 7

Hospital day

CAP, severe sepsis, died

CAP, severe sepsis, alive

CAP, no organ failure

Kellum et al. Arch Int Med 2007

Multicenter CAP cohort, n=1883


Undetectable IL-6 levels in CAP

0

0.1

0.2

0.3

0.4

0.5

1 2 3 4 5 6 7

Hospital day

pro

po

rtio

n u

nd

etec

tab

le

Died by 60d

Survived

Kellum et al. Arch Int Med 2007

CRISMA Critical Care Medicine the University of Pittsburgh Hotchkiss and Karl NEJM 2003




Evolutionary pressure …

Variance not explained by genetic drift alone

Asp299Gly/399WT haplotype

Negative effects in sepsis

Positive effects in malaria

More common in subSaharan Africa


In silico trials of anti-TNF AB for sepsis

Clermont et al. Crit Care Med 2004

Pathogen

load Pathogen

virulence

Host TNF

responsiveness

Host

anti-inflammatory

responsiveness


So what can be done?


Precision (personalized) medicine to the rescue?

... couples “established indices with state-of-the-art molecular profiling to create diagnostic, prognostic, and therapeutic strategies precisely tailored to each patient's requirements”

Mirnezami et al. NEJM 2012

Examples of success … Trastuzumab for HER-2 positive breast cancer

Lebrikizumab for asthma

Vemurafenib, dabrafenib for BRAF V600E mutation melanoma


A framework for launching precision medicine …

As articulated by …. National Academies of Science

NIH Pharmacogenomics Research Network

‘Knowledge network’ Large bank of clinical, genetic, and molecular ‘omics’ data

Baby steps taking place in sepsis …

Use network for Drug target discovery

Drug-response phenotypes

‘Biomarkers’ of drug responsiveness

Novel RCT design incorporating drug-response phenotypes

Theragnostics – biomarker-guided treatment


A framework for launching precision medicine …

As articulated by …. National Academies of Science

NIH Pharmacogenomics Research Network

‘Knowledge network’ Large bank of clinical, genetic, and molecular ‘omics’ data

Use network for Drug target discovery

Drug-response phenotypes

‘Biomarkers’ of drug responsiveness

Novel RCT design incorporating drug-response phenotypes

Theragnostics – biomarker-guided treatment


Sotiriou C and Piccart MJ. Nat Rev Cancer. 2007 Jul;7(7):545-53.

Precision medicine in cancer


Managing respiratory infection based on PCT concentrations



Pneumonia Severity Index (PSI) and Procalcitonin …

0.5

0.25Mo

rta

lity

10 20 30

PCT tier I

PCT tier II

PCT tier III

PCT tier IV

Day

0

0

PSI IV

PCT tier I

PCT tier IIPCT tier III

PCT tier IV

0.5

0.25Mo

rta

lity

10 20 30

PCT tier I

PCT tier II

PCT tier III

PCT tier IV

PCT tier I

PCT tier II

PCT tier III

PCT tier IV

Day

0

0

PSI IV

PCT tier I


PCT tier IV

Among subjects with high risk,

PCT provides important additional discrimination

0.5

0.25

10 20 30

PCT tier I


PCT tier IV

Day

0

0

PSI V0.5

0.25

10 20 30

PCT tier I


PCT tier IV

Day

0

0

PSI V

Huang et al. Ann Emerg Med 2008


Low risk Gene signature and Clinical profile Discordant risks

High risk Gene signature and Clinical profile

TAILORx (n=10,500 women) and MINDACT (n=6,000 women)

Bringing molecular prognostic signatures to daily clinical practice

Node-negative B.C. population

CHEMOTHERAPY •RANDOMIZE

CHEMO YES or NO (TailorX)

•RANDOMIZE FOR the decision-making tool (Mindact)

ENDOCRINE THERAPY


But Sepsis isn’t the same as Breast Ca …

Biomarkers of drug-response are not known at the start of the trial

Biomarkers could be discovered in post-hoc analyses But, inefficient, because requires validation

Biomarkers could be ‘guessed at’, but highly risky IL-6 as a biomarker for severe sepsis in the MONARCHS trial

N=2,600

IL-6 predicted outcome but NOT drug-responsiveness

Panacek et al Crit Care Med 2004

Biomarkers may not be available ‘real-time’, requiring complicated enrollment and hypothesis testing


Drug-response phenotypes …

Definitely tricky ….

Consider agnostic, data-driven and knowledge-driven analyses

Examples Corticosteroids

aPC


Knowledge-driven analyses in steroids …

Known genetic variation in GC resistance Drug-response phenotypes in chronic inflammatory diseases


Discovery-driven analysis of aPC in PROWESS

Post-hoc analysis discovered gene with large treatment interaction Man et al. Pharmacogenomics Journal 2012


But what would a new trial look like?

Begin the trial with uncertainty about Drug effects

Drug by phenotype interaction

Either … Narrow over time to best-performing drug-response phenotype

Adaptive signature trial

Requires

Response-adaptive randomization

Biomarker result available pre-randomization

Or, divide into stages

Stage 1 – evaluate candidate biomarkers

Predictability of drug-response

Frequency rate

Stage 2 – enroll based on expected frequency of biomarker

Biomarker NOT necessary pre-enrollment


Adaptive designs

Multiple arms

In-trial adaptation based on event rates and success rates of interventions

Figure 1. Response adaptive randomization for perpetual trial with multiple arms


A global adaptive trial program for sepsis?

Enroll patients in one of several arms Randomization based on probability of success

Closer to QI – broader participation?

Thresholds for dropping arms Sample size a function of required certainty, not a pre-

determined number

Selective randomization for … Different types of patients

Different clinical settings

Different use of co-interventions


400-800 patient phase 2 RCT WITH simultaneous evaluation of 3-5

candidate biomarker signatures of response to therapy

~2000 patient phase 3 study of sepsis patients

Tests sequential hypotheses:

• Does drug improve outcomes in all subjects enrolled?

• Does drug improve outcomes in those who are biomarker positive

Reject null for all patients

Drug approved for a broad sepsis

population

Fail to reject null for all subjects

Test drug efficacy in biomarker positive

subjects

Reject null for biomarker +ve patients

Drug approved for biomarker +ve

patients only

Fail to reject null

Drug not approved

Stage I

Stage II

2-stage ‘traditional’ RCT design incorporating potential drug-response biomarkers


Key elements of 2-stage novel RCT design

Enter stage I with 3-5 uncertain biomarkers and a drug

Narrow to best 1 (or more) biomarkers for stage II

Enroll broadly, power based on frequency of phenotype and estimate of drug-by-phenotype interaction

Sequential hypothesis testing, with adjusted alpha spend function

Do NOT surrender possibility of overall effect

‘Confirm’ biomarker during drug trial Simultaneously test drug and biomarker


Conclusions

There is a need for RCTs in sepsis that are … More plentiful

Larger

Easier

More likely to be positive

But, industry, government, and trialists are risk averse Continuing to do the same old thing

There are models from other fields They will need customization

Sepsis researchers will need to be bold and innovative


With huge thanks to …

The CRISMA Center Sachin Yende

Lisa Weissfeld

Elsewhere on campus Ferrell, Kammerer (genetics)

Don Defranco (glucocorticoids)

Elsewhere ANZICS CTG – Finfer, Venkatesh and Cohen

Melbourne, Australia - Montgomery

VA Clinical Studies Program – Meduri

Basel – Christ-Crain

SEPNet, Germany – Brunkhorst

UNC biostats – Kosorok and Zhao


New Approaches and Designs

Zelen’s design

Cluster randomization

Facilitated enrollment

Sequential design

Biomarker-guided selection

Adaptive (Bayesian) designs Ex – Pravigard Ex – ISPY-2


VA Cooperative EMR-based randomization

Fiore et al. Clinical Trials 2011


STAR*D

N=4000

Multiple settings

Wide number of simultaneous comparisons in one RCT

Rush et al. Controlled Clinical Trials 2004


The dying ‘process’ is elusive in sepsis

Well-documented changes in immune function Early hyperinflammatory response

Late humoral and cellular hypoinflammatory response

Hotchkiss Nat Rev Immunol 2006

But, organ ‘failure’ generally has bland histology Little evidence of parenchymal cell death

Hotchkiss et al Crit Care Med 1999

Notable exceptions are the lung and gut, but … The lung recovers …

Red then grey hepatization of the lung can completely resolve

Intestinal epithelial cells have very high turnover anyway …

Primary predictor of death following MOF Physician gives up and decides to withdraw support

Cook et al NEJM 2003


Don’t worry.

I had the same

thing and they

cured me!

First clue that the latest medical

breakthrough isn’t quite there yet.


Sepsis and chronic health

AJRCCM 2005;172:1440-46.


The lingering consequences of sepsis

Longitudinal analysis of the Health and Retirement Study

3-fold increase in moderate/severe cognitive impairment

3-fold increase in physical function impairment


Inflammatory markers and long-term survival

90 180 365

0.80

0.85

0.90

0.95

1.00

IL-10 < 5 pg/ml

IL-10 > 5 pg/ml

P<0.0001

90 180 365

0.80

0.85

0.90

0.95

1.00

Days post discharge

Su

rviv

al

IL-6 < 2 pg/ml

IL-6 > 2 pg/ml

P<0.0001

90 180 365

0.80

0.85

0.90

0.95

1.00

IL-10 < 5 pg/ml

IL-10 > 5 pg/ml

P<0.0001

90 180 365

0.80

0.85

0.90

0.95

1.00

Days post discharge

Su

rviv

al

IL-6 < 2 pg/ml

IL-6 > 2 pg/ml

P<0.0001

Hazard over time:

IL-6 - decays over 90 days

IL-10 - constant

Yende et al. AJRCCM 2008


Perfect

Health

Death

Time

Burd

en o

f dis

ease

Sepsis

Sepsis

A new model?

Sepsis …

Precipitated by chronic disease

Stimulates deterioration of chronic disease

Chronic health (e.g., heart disease)

E.g., Plaque

destabilization

Yende and Angus Curr Inf Dis Rev 2008


Alternative scenarios for sepsis study with therapy that has strong effect on sepsis process.

Large modifiable risk

Small modifiable risk

Control Treatment Control Treatment

100%

Components of Risk for Death

Sepsis

Underlying co-morbidity

Patient selection

Absolute outcome

Attributable outcome

Response to therapy

Wood, et al. CCM 2003


Novel pathways and drug discovery - severe H1N1

IFITM3 splice variant associated with ‘severe’ H1N1

ventilated H1N1 ‘ARDS’

Pathogenicity in knock-out mice

Novel drug target

Everitt et al Nature 2012

Documents

University of Pittsburgh Critical Care Medicine · 2012. 12. 12. · CRISMA Critical Care Medicine the University of Pittsburgh But… No single therapy specifically for sepsis Only