Chee Lee, MBBS (Hons), MMedSci (Clin Epid), MBiostat, PhD, FRACP Biomarker-Based Clinical Trials: Practical and Design Considerations
Biomarker “Any characteristic that can be objectively measured as an indicator of normal or pathological biological processes or the response to a therapy” Biomarkers Definitions Working Group. Clin Pharmacol Ther. 69(3):89-95, 2001
No Therapy Factor 1 Neg Factor 1 Pos 100% cure PROGNOSISPROGNOSIS 50% cure 10% cure Pure Prognostic Factor Classifies an individual’s baseline risk of having a clinical event Hayes et al. Breast Cancer Res Treat. 1998;52:
Factor 1 Neg Factor 1 Pos Classifies the magnitude of an individual’s response to treatment Pure Predictive Factor No Therapy 100% cure PROGNOSISPROGNOSIS 50% cure 10% cure Hayes et al. Breast Cancer Res Treat. 1998;52:
Factor 1 Neg Factor 1 Pos Marker Can Be Prognostic and Predictive No Therapy 100% cure PROGNOSISPROGNOSIS 50% cure 10% cure Hayes et al. Breast Cancer Res Treat. 1998;52:
Predictive Biomarkers Best evaluated in a prospective study with concurrent control arm Different study designs – enriched vs unselected designs Can be “prospective” and “retrospective”
Enrichment Design: Enrol Only Those Thought Likely to Respond Romond et al. N Engl J Med. 2005;353: Lee et al. Med J Aust. 2009;190:
100% 50% 25% Pegram et al. J Clin Oncol. 2005;23: Simulated Phase III Trial in Which 100% of Patients Show a Treatment Effect: 200 Active Patients With Median = 27 Months, 200 Placebo Patients With Median = 22 Months Biomarker+ population Improves mOS from 22 to 27 months (≈25% improvement) Biomarker- population No benefit with treatment
Enrichment Design: Improves Efficiency Prevalence Biomarker+ Relative Efficacy Efficiency Gain 25%100%16× 25%50%2.5× 50%100%4× 50% 1.8× 75%100%1.8× 75%50%1.3× Gains in efficiency depend on marker prevalence and relative efficacy in biomarker+ and biomarker- patients Simon and Maitournam. Clin Cancer Res. 2004;10:
Using Markers to Restrict Trial Eligibility: Beware No difference in benefit based on HER2+ expression 1 After 10 years, new study of Herceptin in HER2- patients? 1. Paik et al. N Engl J Med. 2008;358: Paik. ASCO Abstract
NSABP B-47 POPULATION Female Invasive BC Node positive or high-risk node negative HER2 normal STRATIFICATION Age <50 y, ≥50 y Pos nodes (0-3, 4-9, 10+) Hormone receptor status ER+ and/or PgR+, ER- and PgR- Chemotherapy regimen RANDOMISATIONRANDOMISATION ARM 1 TC or AC WP ARM 2 TC or AC WP + Trastuzumab × 1 year TC: Docetaxel + cyclophosphamide q3w × 6 AC WP: Doxorubicin + cyclophosphamide q2-3w × 4, paclitaxel q1w × 12 Fehrenbacher et al. J Clin Oncol. 2013;31(suppl). Abstract TPS1139.
Enrichment Design Consideration Enrol only those thought likely to respond Exclude those unlikely to benefit –Need to understand the scientific basis of the targeted agent and thus the population likely to benefit for treatment –Statistical simulations demonstrate improved efficacy through enrichment with the population likely to benefit for treatment –Danger of excluding “biomarker negative” patient
Gefitinib vs Carboplatin/Paclitaxel for Clinically Selected Chemotherapy-Naive Patients With Advanced NSCLC in Asia (IPASS) *Limited to maximum of 6 cycles. Fukuoka et al. J Clin Oncol. 2009;27(suppl). Abstract Fukuoka et al. J Clin Oncol. 2011;29: Mok et al. N Engl J Med. 2009;361: R Eligibility: Chemotherapy naive Age ≥18 years Adenocarcinoma Never or light ex-smokers Life expectancy >12 weeks Measurable stage IIIB/IV disease Gefitinib 250 mg/day Carboplatin AUC 5 or 6 paclitaxel 200 mg/m 2 3 weeks* Primary endpoint: PFS (non-inferiority) Secondary endpoints: ORR, OS, QOL, disease-related symptoms, safety Exploratory biomarkers: EGFR mutation, EGFR gene copy number, EGFR protein expression N=1217 Cross-over allowed upon disease progression
Probability of PFS At risk: Gefitinib Carboplatin/ paclitaxel Months Gefitinib Carboplatin/ paclitaxel N Events453 (74.4%) 497 (81.7%) HR (95% CI) = (0.651, 0.845) P< Median PFS (months) months progression free61% 74% 6 months progression free48% 48% 12 months progression free25% 7% PFS in ITT Population Primary Cox analysis with covariates. HR <1 implies a lower risk of progression on gefitinib. Mok et al. ESMO Abstract LBA2.
Treatment by subgroup interaction test, P< EGFR Mutation PositiveEGFR Mutation Negative At risk: Gefitinib C/P Gefitinib (n=132) Carboplatin/paclitaxel (n=129) HR (95% CI) = 0.48 (0.36, 0.64) No. events gefitinib, 97 (73.5%) No. events C/P, 111 (86.0%) Gefitinib (n=91) Carboplatin/paclitaxel (n=85) HR (95% CI) = 2.85 (2.05, 3.98) P< No. events gefitinib, 88 (96.7%) No. events C/P, 70 (82.4%) Probability of PFS PFS in EGFR Mutation-Positive and -Negative Patients ITT population. Cox analysis with covariates. Mok et al. ESMO. 2008, Abstract LBA2. Months
Meta-analysis of EGFR-TKI trials Lee et al. JNCI. 2013, 105(9):
Biomarker Analysis Using Existing RCT Data Correlative translational science exploratory endpoint Original study not necessarily powered for correlative science endpoint Tissue not necessarily obtained in all randomised patients Outcomes considered hypothesis- generating and need to be confirmed in future prospective studies Lee et al. Med J Aust. 2009;190:
Biomarker testing in all patients Power trial separately within marker groups Upfrontstratification Biomarker by Treatment Interaction Design Sargent et al. J Clin Oncol. 2005;23: Unselected Design: Enrol All Comers Evaluate the utility of a new Rx in the subsets determined by the pre-specified classifier Useful trial design if biological evidence to limit Rx to biomarker+ patients is less clear
Unselected Design Consideration Enrol all patients (biomarker+ and biomarker-) Upfront molecular characterisation Benefit from those with biomarker+ unclear Studies are usually large, where biomarker+ and biomarker- are powered differently Hybrid designs to address particular subpopulation without resorting to a large trial
Unselected Design: Hybrids EORTC BIG 3-04 MINDACT trial design 6000 ER+ Node negative women.
Unselected Design: Hybrids
Prognostic Biomarker Best evaluated in control (no treatment or treated with non-targeted therapy) group Protocol of assay method used (specific reagents or kits used) Quality control procedures Reproducibility assessments Quantitation/scoring methods Show the relation of the biomarker to other established prognostic factors
Pepe et al. J Natl Cancer Inst. 2001;93: Phases of Biomarker Development
Incorporating Biomarkers in Early Phase Trials Surrogate biomarker endpoint potentially allows assessment of treatment effects earlier than the ultimate clinical endpoint of interest (PFS/OS) Surrogate biomarker could also demonstrate mechanisms of molecular action Valid surrogate biomarkers have both prognostic and predictive properties The challenge: Does treatment effect on biomarker reliably predict treatment effect on clinical endpoint?
STrtT Randomised treatment Potential surrogate (« intermediate ») True endpoint Buyse and Molenberghs. Biometrics. 1998;54: Validation of Surrogate Endpoints
ST This requires several trials Effects of treatment on surrogate and on true endpoint must be correlated Trt This can be shown in a single trial Validation of Surrogate Endpoints Surrogate and true endpoint must be correlated
Disease-Free Survival (DFS) vs Overall Survival (OS) Hazard Ratios (HR) by Trial Sargent et al. J Clin Oncol. 2005;23:
Conclusions Biomarkers could improve efficiency of clinical trials Multiple innovative approaches are available for assessment of efficacy of novel therapies and the accompanying biomarkers (“companion diagnostics”) Limitations of different biomarker-directed trial designs discussed