defining response in gvhd treatment

Defining Response in GVHD TreatmentDefining Response in GVHD Treatment Trials

Paul Carpenter, MB.BS. Associate Member, FHCRCAssociate Professor, UW

Disclosures:– Contract Support for clinical trialsContract Support for clinical trials

• Novartis: Nilotinib for relapse prophylaxis after HCT

5/19/09 FDA AGVHD 1

What is Clinical Benefit in Acute GVHD lTreatment Trials?

Direct Benefit:Direct Benefit:– Prompt reversal of Grade III/IV symptoms clearly benefits the patientbenefits the patient

Survival Benefit:l l f b h– Complete resolution of GVHD may be neither

durable nor associated with long survival

P t t GVHD th i th t hi hi h CR t– Potent GVHD therapies that achieve high CR rates may increase infection mortality rates

“Minimal” residual acute GVHD activity may be– “Minimal” residual acute GVHD activity may be compatible with long‐term survival

5/19/09 FDA AGVHD 2

St d Q tiStudy Question

Can response to GVHD therapy be used

as a surrogate for survival in order

to demonstrate clinical benefit?

5/19/09 FDA AGVHD 3

“Not too hot, not too cold, this one’s just right”:Not too hot, not too cold, this one s just right : is how Goldilocks would pick GVHD therapy

What is the distribution of response rates among those destined to survive?Is strict CR the best predictor?

5/19/09 FDA AGVHD 4

Martin PJ, Best Pract Res Clin Haematol 2008;21:364

Study HypothesesStudy HypothesesGVHD responses that are more predictive of 6

h i l h ld idmonth survival should consider:1. Relaxation of the requirement for CR because minimal

residual GVHD activity is not clearly deleteriousresidual GVHD activity is not clearly deleterious.

2. Whether response durability improves prediction

3. Whether steroid dose predicts survivalp

Why choose 6‐month survival ?1. Survival analysis makes fewer assumptions than NRM1. Survival analysis makes fewer assumptions than NRM

2. Deaths < 6 months are often due to acute GVHD

3. Deaths > 6 months have multiple causes (esp, relapse)

5/19/09 FDA AGVHD 5

Study DesignStudy DesignAdditional data collection performed on a subgroup f l bli h d l i hfrom a recently published large retrospective cohort (Mielcarek et al, Blood 2009)

D t il d GVHD t i t D 7 28 d 56Detailed GVHD organ staging at Days 7, 28 and 56 after start of GVHD therapy

Assigned complete response (CR) at each time pointAssigned complete response (CR) at each time‐point

Also defined “very good partial response” (VGPR)

D t il d t id d d t il bl f thiDetailed steroid dose data were available for this cohort.

5/19/09 FDA AGVHD 6

Distinguishing VGPR from CR based on Organ Staging

Stage Skin Liver GutBSA (%) Bilirubin (mg/dL) stool (L/day)

0 0 < 2.0 ≤ 0.5

1 <25 2.0 ‐2.9 > 0.51

2 25‐50 3.0 – 5.9 > 1

3 >50 6.0 ‐14.9 >1.5

4 Bullae ≥ 15 0 >224 Bullae ≥ 15.0 >2

1 – or pesistent anorexia, nausea, vomiting2 – or severe abdominal pain ileus

VGPR included: 1.Minor rash resolving, evanescent, but not progressive2.Minor stable or resolving elevations of bilirubin

5/19/09 FDA AGVHD 7

2.Minor stable or resolving elevations of bilirubin3.Minimal but stable or resolving GI symptoms

Study CohortStudy Cohort

Began with data set of Mielcarek et al (Blood 2009;Began with data set of Mielcarek et al (Blood 2009; 113:2888‐2894)

N = 733 adults transplanted 2000‐2005p

N = 248 selected for further data collection because acute GVHD was diagnosed before Day 30 g y

Day 30 criterion balances the need to:

1. Maximize probability for capturing p y p gresponse data at Days 7, 28 and 56.

2. Minimize bias from including subjects whose g jstay in Seattle was prolonged by illness

5/19/09 FDA AGVHD 8

Number of subjects evaluable for hresponse at each time‐point

Evaluable Not EvaluableDay

Evaluable(N)

Not Evaluable (N)

Reason Not Evaluable

7 236 12 7 Missing data1 Relapse4 Deaths

28 225 23 4 Mi i d t28 225 23 4 Missing data1 Relapse18 Deaths

56 204 44 3 Missing data3 Relapse1 Second transplant1 Second transplant37 Deaths

5/19/09 FDA AGVHD 9

Causes of Death at each time‐pointpDay Deaths, N Cause

7 43 Conditioning toxicity ± Infection

d7 4

1 GVHD + conditioning toxicity

61

Conditioning toxicity ± InfectionGVHD

28 18 1181

GVHD + conditioning toxicityRelapseInfectionOth (C b l h h )1 Other (Cerebral hemorrhage)

56 37

762

Conditioning toxicity ± InfectionGVHDGVHD di i i i i56 37 2

3134

GVHD + conditioning toxicityGVHD + infectionInfectionRelapse4

2RelapseOther (Cerebral hemorrhage, Idiopathic pneumonitis)

10

We compared subjects with either CR or CR+VGPR to subjects who were non‐respondersCR VGPR to subjects who were non responders

or less than very good partial responders

CRSeattle Response

GroupsPRVGPR

+PR NRCR VGPR NR

+

paris

ons

+CR

+

NR

Com

p

VGPR+

PR

A 2 x 2 type analysis was performed

Classic 2 x 2 GVHD Response“Test” 2 x 2With

DiseaseWithout Disease

st Disease & No Disease &

Dead Alive

Dead & Alive &

Test 2 x 2

R

POS

tes Disease &

POS testNo Disease &

POS test

a = true +ve b = false +ve

Dead & not in CR

Alive & not in CR

a = true +ve b = false +ve

Not

in C

R

test

Disease & NEG test

No Disease & NEG test

Dead & in CR

Alive & in CR

NR

NEG

t

c = false -ve d = true -ve c = false -ve d = true -veC

A positive “not in CR” test predicts for death at 6 months

GVHD Response “ROC” plotGVHD Response ROC plot

Sensitivity/

ity)

= a / a+c= proportion of dead without CR= “Correctly Predicted Dead”

(Sen

sitiv

i

1 minus Specificity= b / b+d= proportion of survivors without CR

“F l l P di t d D d”= “Falsely Predicted Dead”

Dead & not in CR

Alive & not in CR

(1 - Specificity)

not in CR

Dead & in CR

not in CR

Alive & in CR

a b

5/19/09 FDA AGVHD 13

in CR in CRc d

Day 7 CR or CR/VGPR “Tests” Lack Sensitivity d S ifi i f P di i D h 6 M hand Specificity for Predicting Death at 6 Months

SurvivorsAll patients SurvivorsAll patients(includes deaths before endpoint)

CR CRCR CR

CR + VGPR CR + VGPR


CR Test at Day 28 or 56 Has Sensitivity b t L k S ifi itbut Lacks Specificity

All patients SurvivorsAll patients Survivors

5/19/09 FDA AGVHD 15Day 56 – red Day 28 – blue Day 7 ‐ black

VGPR/CR Test at Day 28 or 56 Improve Specificity with Loss of Sensitivity


CRVGPR/CR

CRVGPR/CR

VGPR/CR


Durable Day 28/56 VGPR/CR Tests Improve f fSensitivity with Little Loss of Specificity


Durable CR

Durable CR


Steroid Dose or % Dose Reduction at Day 28 or 56 t P di ti f D th t 6 M th56 was not Predictive of Death at 6 Months

Day 56y

Day 28

Day 7

Day 28steroid dose% reduction


ConclusionsConclusions

Poor surrogates for 6 month survival include:Poor surrogates for 6 month survival include:– Day 7 CR or VGPR response rates– Day 28 or 56 CR or durable CR ratesy

Durable VGPR/CR response has marginal potential as a surrogate for survivalp gA critical unknown is how a treatment that changes response will change 6 month g p gmortality via other downstream effects


Difference between Minnesota and Seattle results might relate to definition of PRmight relate to definition of PR. Where do PRs best fit in the analysis?

CRSeattle Response Groups PRVGPR

+PR NRCR VGPR

NR

CR NR

Comparisons

VGPR

++

PR

Minnesota Response GroupsCR NR


CR PR NR ComparisonsPR NR

Future DirectionsFuture Directions

Prospective evaluate durable VGPR/CR as anProspective evaluate durable VGPR/CR as an endpoint in clinical trials because it may be more useful than strict CR.Need to better understand the relationship between beneficial effects (response induction) of GVHD treatments and harmful effects that can occur downstream of CR/VGPR ( i i i f i lCR/VGPR (opportunistic infection, relapse, chronic GVHD).


Acknowledgementsg

Barry Storer, PhDy ,Marco Mielcarek, MDPaul Martin, MDPaul Martin, MD


defining response in gvhd treatment

Documents