risk stratified analysis improves prediction of treatment benefit over subgroup analysis: findings...

12
Risk Stratified Analysis Risk Stratified Analysis Improves Prediction of Improves Prediction of Treatment Benefit Over Subgroup Treatment Benefit Over Subgroup Analysis: Findings from Analysis: Findings from Intergroup N9741 Intergroup N9741 HK Sanoff, ME Campbell, HC HK Sanoff, ME Campbell, HC Pitot, RM Goldberg, DJ Sargent Pitot, RM Goldberg, DJ Sargent University of North Carolina at Chapel University of North Carolina at Chapel Hill, and Mayo Clinic for the NCCTG, CALGB, Hill, and Mayo Clinic for the NCCTG, CALGB, ECOG, SWOG, NCIC ECOG, SWOG, NCIC ABSTRACT # 4018

Upload: winfred-mitchell

Post on 05-Jan-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

Risk Stratified Analysis Improves Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Prediction of Treatment Benefit Over

Subgroup Analysis: Findings from Subgroup Analysis: Findings from Intergroup N9741Intergroup N9741

HK Sanoff, ME Campbell, HC Pitot, HK Sanoff, ME Campbell, HC Pitot, RM Goldberg, DJ SargentRM Goldberg, DJ Sargent

University of North Carolina at Chapel Hill, and University of North Carolina at Chapel Hill, and Mayo Clinic for the NCCTG, CALGB, ECOG, SWOG, Mayo Clinic for the NCCTG, CALGB, ECOG, SWOG,

NCICNCIC

ABSTRACT # 4018

Page 2: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

Despite recent advances in therapy for metastatic Despite recent advances in therapy for metastatic colorectal cancer (MCRC), individuals’ responses to colorectal cancer (MCRC), individuals’ responses to treatment and clinical courses remain treatment and clinical courses remain heterogeneous. heterogeneous.

A substantial minority of patients progress rapidly, A substantial minority of patients progress rapidly, dying of disease within a year of diagnosis.dying of disease within a year of diagnosis.

Several baseline factors indicate a poor prognosis, Several baseline factors indicate a poor prognosis, i.e.:i.e.:– Performance status (PS) Performance status (PS) >> 2 2– Alkaline phosphatase (ALK) > twice upper limit of normal. Alkaline phosphatase (ALK) > twice upper limit of normal.

No baseline factors currently are used in treatment No baseline factors currently are used in treatment selectionselection

BACKGROUNDBACKGROUND

Page 3: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

Risk stratified analysis (RSA) assesses Risk stratified analysis (RSA) assesses treatment effect according to baseline risktreatment effect according to baseline risk

–Uses pre-existing risk models to divide patients Uses pre-existing risk models to divide patients into risk groups.into risk groups.–Outcomes stratified based on these risk groupsOutcomes stratified based on these risk groups

RSA advocated as a better method for RSA advocated as a better method for predicting patient specific treatment predicting patient specific treatment benefit over subgroup analysis. benefit over subgroup analysis.

–RSA does not artificially divide according to RSA does not artificially divide according to one factor when many may coexist within one one factor when many may coexist within one patient.patient.

BACKGROUNDBACKGROUNDRisk Stratified AnalysisRisk Stratified Analysis

Kent & Hayward, JAMA 2007; 298: Kent & Hayward, JAMA 2007; 298: 12091209

Page 4: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

Perform RSA of a large, phase III trial Perform RSA of a large, phase III trial of first-line chemotherapy for MCRCof first-line chemotherapy for MCRC

Compare RSA to subgroup analysis Compare RSA to subgroup analysis by PSby PS

Assess for differences in treatment Assess for differences in treatment benefit by risk group benefit by risk group

OBJECTIVESOBJECTIVES

Page 5: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

Individual patient data from N9741, Phase III Individual patient data from N9741, Phase III trial of FOLFOX vs. IROX vs. IFLtrial of FOLFOX vs. IROX vs. IFL

N=1682N=1682 RSA based on risk model of Köhne et al.RSA based on risk model of Köhne et al. Köhne model developed in patients with Köhne model developed in patients with MCRC treated with 5FU. 3 risk groups based on:MCRC treated with 5FU. 3 risk groups based on:

– ECOG PSECOG PS– WBCWBC– Alkaline phosphataseAlkaline phosphatase– Number of sites of metastatic disease Number of sites of metastatic disease

METHODSMETHODS

KöhneKöhne et al. Ann Oncol 2002; 13: 308.

Page 6: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

METHODSMETHODS

RISK GROUPS:RISK GROUPS:

OS and TTP were compared by risk group, PS. OS and TTP were compared by risk group, PS. Cox models assessed the relative predictive Cox models assessed the relative predictive utility of PS and risk group.utility of PS and risk group.

*WBC estimated from absolute granulocyte count (AGC) based on *WBC estimated from absolute granulocyte count (AGC) based on AGC= -0.7 + 0.8(WBC) [Benson, Cancer 1985]. AGC= -0.7 + 0.8(WBC) [Benson, Cancer 1985].

PS

# sites

# sitesWBC

High Risk

High Risk

Med

Alk 1

Low

High Risk

Med

0, 1> 1

<10 x 109>10 x 109

1>1

>300 <300

>1

Page 7: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

RESULTSRESULTSOverall survival by Risk Overall survival by Risk GroupGroup

0

20

40

60

80

100

0 12 24 36 48 60

Time (M onths)

Low Median= 20.8 Mos

Intermediate Median= 17.4 Mos

High Median= 9.4 Mos

P-Value = <0.0001

Page 8: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

RESULTSRESULTSOverall survival by Performance Overall survival by Performance StatusStatus

0

20

40

60

80

100

0 12 24 36 48 60

Time (M onths)

PS 0 Median= 20.4 Mos

PS 1 Median= 14.8 Mos

PS 2 Median= 9.1 Mos

P-Value = <0.0001

Page 9: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

RESULTS: RESULTS: OS Multivariate OS Multivariate modelmodelPrediction improved by PS and Risk GroupPrediction improved by PS and Risk Group

Risk GroupRisk Group HR (95% CI)HR (95% CI) XX22 P-Value P-Value

PS 0PS 0 11

PS 1PS 11.4 (1.4 (1.3, 1.6)1.3, 1.6) 41.041.0 <0.0001<0.0001

PS 2PS 21.5 1.5 (1.2, 2.0)(1.2, 2.0) 10.310.3 0.00140.0014

LowLow 11

IntermediateIntermediate1.4 (1.4 (1.2, 1.5)1.2, 1.5) 31.731.7 <0.0001<0.0001

High High 2.3 2.3 (1.9, 2.8)(1.9, 2.8) 71.671.6 <0.0001<0.0001

Likelihood Ratio X2 for PS = 83.8, for Kohne = 114.1, Combined Model=157.6

Page 10: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

RESULTS RESULTS OS by treatment arm, risk OS by treatment arm, risk groupgroupRisk GroupRisk Group Trt Trt

ArmArmNN Median Median

(Months)(Months)HR (95% CI)HR (95% CI) P-ValueP-Value

LowLow

FOLFOXFOLFOX 247247 27.527.5 11

IFLIFL 146146 18.218.2 1.7 1.7 (1.4, 2.1)(1.4, 2.1) <0.001<0.001

IROXIROX 135135 19.719.7 1.5 1.5 ( 12, 1.8)( 12, 1.8) <0.001<0.001

IntermediatIntermediatee

FOLFOXFOLFOX 384384 19.219.2 11

IFLIFL 227227 13.813.8 1.5 1.5 (1.3, 1.8)(1.3, 1.8) <0.001<0.001

IROXIROX 196196 17.817.8 1.3 1.3 (1.1, 1.6)(1.1, 1.6) 0.0050.005

HighHigh

FOLFOXFOLFOX 6060 10.710.7 11

IFLIFL 5555 9.49.4 1.3 1.3 (0.87, 1.9)(0.87, 1.9) 0.210.21

IROXIROX 4949 9.19.1 1.3 1.3 (0.87, 1.9)(0.87, 1.9) 0.220.22

Interaction p value risk group X treatment arm=0.08

Page 11: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

CONCLUSIONSCONCLUSIONS

RSA using the Köhne model is prognostic RSA using the Köhne model is prognostic of survival in this cohort treated with of survival in this cohort treated with combination chemotherapy.combination chemotherapy.

FOLFOX is superior to IFL in all risk FOLFOX is superior to IFL in all risk groups.groups.

Trend towards less benefit in high risk Trend towards less benefit in high risk group.group.– Small # high risk patients supports need for Small # high risk patients supports need for

pooled analyses of these patients.pooled analyses of these patients.

Page 12: Risk Stratified Analysis Improves Prediction of Treatment Benefit Over Subgroup Analysis: Findings from Intergroup N9741 HK Sanoff, ME Campbell, HC Pitot,

CONCLUSIONSCONCLUSIONS

RSA adds predictive ability to a RSA adds predictive ability to a multivariate model above PS alone.multivariate model above PS alone.

RSA should be considered as a way to RSA should be considered as a way to present clinical trial data to better present clinical trial data to better inform pts and physicians of treatment inform pts and physicians of treatment benefitbenefit

RSA is useful for designing clinical RSA is useful for designing clinical trials to ensure balanced trials to ensure balanced randomizationrandomization