Surrogate Endpoints in Oncology Clinical Trials: Are We There Yet? from a Practical Perspective

Surrogate Endpoints in Oncology Clinical Trials: Are We There Yet? From a Practical Perspective

Qian Shi, PhD., Professor of Biostatistics and Oncology

ASA Biopharmaceutical Section Regulatory-Industry Statistics Workshop September 24, 2020

Make drug development Un-validated surrogate endpoints can produce erroneous faster, more efficient, safer, conclusions regarding the efficacy or safety of therapies and more feasible

Shi and Sargent, Int J Clin Oncol. 2009

Prognostic association Treatment effect correlation

R2 = 0.85

Prognostic does NOT guarantee surrogacy

Munshi N et al., JAMA Oncol, 2017

Treatment effect correlation

R2 = 0.85 Context of Use: A substitute for “true” clinical endpoint in future trials

* 2 ** 2 R WLS and R Copula

©2020 MFMER | slide-6 R2 and R2 Critical trial-level factors that impact the WLS Copula estimation accuracy and variability: most commonly used estimates • Number of trials • Range of the treatment effect sizes

Variability Bias

Shi et al, JCO 2017

Pre-biologics Era Post-biologics Era

Buyse et al, JCO 2007 Shi et al, JCO 2014

Sargent et al, Shi et al, ASCO JCO 2007 2019 Oral •3-year DFS •Stage III •Stage II & III •2-year DFS •Oxa /IRi •Stage III •5FU •Stage II & III •Oxa/IRI +/- •5FU biologics Sargent et al, Sargent, et al, JCO 2005 EJC 2011

5FU+LVProlonged SAR FOLFOX Prolonged SAR, TTR, OS 1978 1995 2007 1998 2004 2009 Shi et al, JCO 2013 Salem et al, ASCO GI poster 2018

Endpoint Events Measures Surveillance

True clinical Overall Survival Deaths Death certificates Patient contact endpoint

Deaths Imaging, blood test, Routine examinations Traditional surrogate Disease-free BM biopsy, etc. endpoint Survival (SOC, standard Disease Status (RECIST, IMWG) processes)

Study specific PSA, MRD, ctDNA, Biomarker Biomarker Status Treatment specific cognitive functions Novel Outside of routine examination surrogate endpoint Composite Biomarker status + Disease status Endpoint involving Disease Status MRD requiring CR dependent Biomarker Deaths

Endpoint Events Measures Surveillance Source of heterogeneities

True clinical Overall Survival• Patient selectionDeaths for testingDeath certificates Patient contact endpoint • Testing methods

Deaths Imaging, blood test, Routine examinations Traditional surrogate Disease-free • Testing time points BM biopsy, etc. endpoint Survival (SOC, standard Disease Status (RECIST, IMWG) processes)

Minimal Residual Disease Clinical Utility Lower Higher • More accurately defining Sensitivity Sensitivity cure

Non- MRD+ MRD+ responders Statistical Challenges: Conversion rate differ per • Testing • Assay Conversion • Laboratory • Specimen

Responders MRD- MRD- • Disease characteristics Response Rate ↓ • Regimen

RECIST Responses Response Reversing?

Directing Treatment? MRD Responses

Response Lasts?

Number of Testing • Single vs. many • Varying across patients

Timing of Testing • Treatment dependence • Response dependence

Patient Selection • Response dependence • Randomization

• Clinically meaningful: • Capture the targeted treatment effect • Correlates the treatment effect on true endpoint • Statistically meaningful: • Programmable derivations based on necessary data elements • Consistency across all trials • Consistency from evaluation to future use • Practically useful in further trials: • Provides utility advantages compared to the true endpoints • Data elements needed can be routinely collected in the future trials

• Individual patient data (IPD) from RCTs • Unbiased estimates of treatment effects within trials • Consistent derivations across trials • Evaluate (and develop) surrogate endpoints with no published treatment effect data • RCT selections: Inclusive • Prespecified inclusion/exclusion criteria • More is better (specificity) • Population heterogeneities (universality) • Treatment class and treatment sequencing (universality)

• Harmonization of testing methods • Common testing time points • Common data collection •Share Data!!