Integrating mol Integrating mol ecular Profiling ecular Profiling Into Patient Se election for the Treatm ment of N Non-Small Cel S ll C l ll L ll Lung Cancer C Frances A. Sheph herd, MD FRCPC Scott Taylor Chair in L Scott Taylor Chair in L Lung Cancer Research Lung Cancer Research Princess Mar rgaret Hospital Professor of Medicine P f f M di i e, University of Toronto U i i f T
Predictive and Pr rognostic Factors • A prognostic facto or is a baseline patient or tumour c p characteristic that identifies a better o outcome regardless of treatment • A predictive factor is a baseline patient or tumour c patient or tumour c characteristic that characteristic that identifies a better o outcome from treatment (respons treatment (respons se or survival) se or survival)
In Other In Other r Words, r Words That is to o Say……. • A prognostic ma arker determines the effect of the the effect of the tumour on the tumour on the patient • A predictive mar A di ti rker determines k d t i the effect of the treatment on the tumour
Exam mple • Untreated patients with PS S 0/1 survive 6 months • Untreated patients with PS S 2/3 survive 3 months • PS is prognostic ti PS i • Treated patients with PS 0 0/1 survive 8 months compared to 6 months for compared to 6 months for control control • Treated patients with PS 2 2/3 survive 4 months compared to 3 months for control • PS is NOT predictive of a a better outcome from treatment as the relative benefit (i.e. the HR) is the same in both PS 0/1 and P same in both PS 0/1 and P PS 2/3 patients although the PS 2/3 patients although the absolute benefit is less
Predictive Factors • It is possible to ide ntify predictive factors for respons p se from a single arm g trial • However, predictive , p e factors for survival benefit can only be determined with an untreated control g roup • This is because sup perior survival in a single arm trial may y be due to inherent prognostic factors ti f t and not due to the d t d t th effect of therapy
When Using Su When Using Su bgroup Results bgroup Results Bewa are!!!! • Implication is that the results in the subgroup g give a better estimate of whether the patient p will benefit than the ov verall result • Two ways this could h Two ways this could h happen happen – Subgroup does so wel l without treatment that treatment may not be n treatment may not be n needed prognostic factor needed - prognostic factor – Subgroup is biological lly distinct and responds d differently to treatment e e t y to t eat e t t – predictive factor t p ed ct e acto
Differential Treat tment Response p • Saying that a treatmen nt “works” differently in a particular subgroup ti l b i is, in statistical terms, i t ti ti l t saying there is a treatm ment by subgroup “interaction” interaction • A patient or tumour ch haracteristic that identifies a subgroup w identifies a subgroup w within which a treatment within which a treatment effect is different from m that in other patients is a “predictive” factor a predictive factor • In statistical terms, this s is the same as saying there is a treatment by baseline covariate interaction
Intera Intera action action • Quantitative interaction Quantitative interaction n: the relative treatment n: the relative treatment effect is in the same dir rection, but of different magnitude (scale used is important) in the two groups groups – Considered less importa ant since the treatment effect is beneficial in both sub groups – EGFR mutation status • Qualitative interaction: the relative treatment effect is in opposite dir effect is in opposite dir rections in the two rections in the two groups – Obviously very importan nt when present – ERCC1 and p53 protein
Testing for th Testing for th he Statistical he Statistical Significance of f an Interaction • When we are testing fo r the significance of an interaction we are askin ng: “are the two hazard rates different from eac ch other?” • This test is less powerf ful (less likely to be positive) because: iti ) b • There are two sources of variation • The numbers in the group The numbers in the group ps being compared are smaller ps being compared are smaller • Always look for the interaction p value value
Testing for th Testing for th he Statistical he Statistical Significance of f an Interaction • Saying the test for intera action is not powerful is another way of saying th hat apparent variations in treatment effects within subgroups can easily arise by chance • This problem is not help Thi bl i t h l ed (in fact it is made d (i f t it i d worse) by doing multiple e tests for significance within subgroups within subgroups • Always look for interacti on p values in studies evaluating predictive ma evaluating predictive ma arkers of survival!!! arkers of survival!!!
Treatment Factors in n Advanced NSCLC Biologic ma ke s Biologic markers Clinical selection parameters � ERCC-1 � RRM 1 � RRM-1 � Sex S � Ras mutation � Age � Beta-tubulin eta tubu � PS � PS � EGFR � p53 Future investigations Future investigations � BRCA1 � TS � Histology � Folate pathway markers � Micro-array � Micro array profiling
ERCC1 Ex xpression Excision Re epair Cross C Complemen l nting 1 Gene ti 1 G
A Adduct formation B LKP5 Damage recognition recognition Mechanism Assembly of the nucleotide excision repair complex of Repair of Repair C of Platinum Dual incision Damage excision Adducts Adducts D Repair synthesis DNA ligation E Repaired DNA Gazdar AF. NEJM 2 0 0 7 ;3 5 6 ( 8 ) :7 7 1 -3 .
Slide 13 LKP5 Some words on figure look blurry. Can this be fixed, or does it project fine? Lori K. Pender, 7/12/2007
Low ERCC 1 Express sion Correlates with Prolonged Survival afte Prolonged Survival afte er Chemotherapy with er Chemotherapy with Gemcitabine & Cisplat in in Advanced NSCLC 1.0 P=0.009 Log rank test .8 al ive Surviva .6 E ERCC1 mRNA < median value (6.7) Cum ulati .4 .2 ERCC1 mRNA > median value (6.7) E 0.0 0 0 20 20 40 40 60 60 80 80 0 0 100 100 120 120 Overall Survival ( w ee eks) Lord et al. Clin Cancer Res 2 0 0 2 ;8 :2 2 8 6 .
ERCC1 in NSCLC: A Double-Edg ged Sword? ERCC1 expression is clearly ERCC1 i i l l Survival Distribution Function linked to platinum-resistance 1 At the same time, it is a At the same time it is a ERCCI > 50 ERCCI > 50 0.75 (94.6 months) favorable prognostic factor in untreated patients with 0.50 early stage NSCLC early stage NSCLC ERCCI < 50 (35.5 months) � High expression associated 0.25 with better survival with better survival P=0.01 P=0.01 0 � Possibly due to its role 0 20 40 60 80 100 120 140 0 in cancer susceptibility Survival ( m onths) Simon et al. Chest 2005; 127: 978-83.
f ERCC1 Status in the Prognostic Effect of Prognostic Effect of f International Adjuvant Lung Cancer Trial (IALT) In the control group, ER RCC1 positive patients have a favourab h f ble prognosis!! bl b i !! Hazard Ratio 95% CI P Value Control Group ERCC1 +ve 1 0.009 ERCC1 -ve ERCC1 -ve 0.66 0 66 0 40-0 90 0.40-0.90 Chemotherapy ERCC1 +ve 1 0.34 ERCC1 -ve 1.16 0.86-1.56 All Patients ERCC1 ve ERCC1 +ve 1 1 0.26 0.26 ERCC1 -ve 0.88 0.71-0.10 Olaussen et al. NEJM 2 0 0 6 ;3 5 5 :9 8 3 -9 1 .
Overall Survival by E ERCC1 Status in the International Adjuva International Adjuva ant Lung Cancer Trial ant Lung Cancer Trial (IAL LT) Patients with ERCC1-Negative Tumors Patients with ERCC1-Positive Tumors 100 100 Chemotherapy Control (80 deaths) vival ( % ) vival ( % ) (105 deaths) 80 80 60 60 60 60 Overall Surv Overall Surv 40 40 Chemotherapy Control (113 deaths) (92 deaths) 20 20 0 0 0 0 0 1 2 3 4 5 0 1 2 3 4 5 Years Years Adjusted HR= 1 .1 4 Adjusted HR= 0 .6 5 P= 0 .4 0 P= 0 .0 0 2 P= 0 0 0 2 Patients with ERCC1- -negative, completely resected NSCLC benefit from m adjuvant cisplatin-based CT but those with ERCC1- positive tumors do not Interactio n p=0.009 Olaussen e et al. NEJM 2 0 0 6 ;3 5 5 :9 8 3 -9 1 .
ITACA Trial in Ea ITACA Trial in Ea ITACA Trial in Ea ITACA Trial in Ea arly Stage NSCLC arly Stage NSCLC arly Stage NSCLC arly Stage NSCLC Taxane Taxane Taxane Taxane High High Profile 4 Profile 4 Control* Control* TS TS TS TS Pemetrexed Pemetrexed Low Low Profile 3 Profile 3 High High Control* Control* Cis/Gem Cis/Gem ERCC 1 ERCC 1 High High Profile 2 Profile 2 Control* Control* Low Low Low Low TS TS Cis/Pem Cis/Pem Low Low Profile 1 Profile 1 Control* Control Control* Control High/Low ERCC1 & TS selected accordin High/Low ERCC1 & TS selected accordin ng to median level of mRNA expression in ng to median level of mRNA expression in historical series ; * Control arm – Investig historical series ; * Control arm – Investig gator choice of a cisplatin-based doublet gator choice of a cisplatin-based doublet
RRM1 Ex xpression Ribonucleotid de Reductase S b Subun nit M1 it M1
Recommend
More recommend
