Biomarker-driven Adaptive Trial Design Proving Informative in Non–Small Cell Lung Cancer

Get Permission

Biomarker-driven adaptive trial design is an accelerated strategy for targeted drug development that is proving informative in non–small cell lung cancer (NSCLC), according to Roy S. Herbst, MD, PhD, Chief of Medical Oncology and Associate Director of Translational Research at Yale Cancer Center, New Haven.1

At the Best of ASCO® ’12 meeting in Boston, Dr. Herbst described recent advances in drug development, drawing largely from his experience with the BATTLE (Biomarker-based Approaches of Targeted Therapy for Lung Cancer Elimination) drug development program at The University of Texas MD Anderson Cancer Center (prior to joining Yale). The new BATTLE-2 study is now open at Yale.

‘Tissue Is the Issue’

As background, Dr. Herbst noted that the traditional approach to drug development required that many thousands of patients be treated before subset analyses revealed those most likely to respond. Gene sequencing and the discovery of mutations that drive tumor growth—yielding targets for inhibitors—changed the field and has led to good treatment outcomes in properly selected patients, but “the problem remains,” he said. “Still, no one is cured of metastatic lung cancer. Resistance to targeted agents occurs. The challenge continues.”

“Tissue is the issue,” Dr. Herbst emphasized. Tumor biopsies now not only guide treatment but are pivotal to drug development. While tissue is frequently obtained from stage III resected patients, it remains rare in the stage IV chemorefractory setting.

Nevertheless, he said, biomarker-driven drug development is accelerating and has changed the landscape of NSCLC from a disease with three major histologic types to, within adenocarcinoma alone, more than a dozen key recognized mutations (EGFR, KRAS, MAPK, P13K, ALK, INSR, FGFR, PDGFR, EPHA/B, HER2, VEGFR, AKT, BRAF) as well as amplifications and fusions in a number of genes. Targeted agents for most of these abnormalities are available or in development.

“With tissue, we can understand the biology and can develop agents to target it. And this brings hope, but there is much work to be done,” Dr. Herbst said.

Challenges in Drug Development

The discovery of druggable targets is one challenge. Equally daunting are the clinical trial process, long timeline, and huge expense in bringing drugs to market. To this end, there has been much discussion about “adaptive design.”

Under an adaptive randomized design, a complex statistical process allows researchers an early glance at how patients with certain molecular signatures in their tumors are responding to various investigational agents. Researchers can then use these data to assign new participants to the treatments that appear likely to be most effective, according to tumor profile.

This is the approach taken in the BATTLE program at MD Anderson and Yale, a platform for integrated translational research that involves a clinical trial program, a novel trial design, and biomarker discovery. The hypotheses are that real-time biopsies are possible and they are accurate reflections of aberrant signaling pathways in NSCLC, and that targeted agents matched to abnormal pathways will improve disease control. A single control arm is used for multiple experimental drugs, and each drug is evaluated for efficacy in multiple biomarker-defined subgroups.

“With Bayesian adaptive randomization, more patients are assigned to more effective therapies,” Dr. Herbst said. The study moves forward based on accumulating patient data. “We learn and update as we go,” he said. “Success is dependent on good biomarkers guiding assignments to good treatment options.”

Lessons of BATTLE-1

BATTLE-1 began 7 years ago and has since enrolled almost 350 patients with refractory, biopsy-amenable NSCLC. Patients undergo two biopsies and biomarker analysis, and are then randomly assigned to a targeted agent (erlotinib [Tarceva], vandetanib [Caprelsa], erlotinib plus bexarotene [Targretin], or sorafenib [Nexavar]), depending on their tumor profile. The endpoint is disease control at 8 weeks.

Some of the lessons learned from the BATTLE-1 experience are:

  • Biomarker-based adaptive design is doable and is well received by clinicians and patients.
  • Real-time biopsies are possible, and results can be obtained in 2 weeks.
  • Adaptive randomization only works when effective drugs and predictive biomarkers are available.
  • Matching targeted agents with abnormal pathways can improve disease control in NSCLC; for example, vandetanib improves disease control in patients with high VEGFR-2 expression.
  • Eight-week disease control is an acceptable surrogate for efficacy in patients with pretreated lung cancer, as it predicts for overall survival; median overall survival was 11.3 months when patients achieved 8-week disease control vs 7.5 months when they did not (P = .002).
  • Gene-expression signatures associated with response to EGFR tyrosine kinase inhibitors do exist in patients with wild-type EGFR.

Now investigators are moving into the BATTLE-2 program, which builds on the experience of BATTLE-1, employs more specifically targeted drugs and drug combinations, avoids biomarker grouping, involves selection and validation of novel predictive biomarkers in real time, and unites multiple drug companies in a collaborative effort. ■

Disclosure: Dr. Herbst has served as a consultant or in an advisory role for Biothera, Diatech, N-of-One, Roche/Genentech, and has research funding from Genentech and GlaxoSmithKline.


1. Herbst R: Developing a targeted therapy: Issues in the age of personalized medicine. Best of ASCO Boston. Education Session. Presented August 4, 2012.

2. Kim ES, Herbst RS, Wistuba II, et al: The BATTLE trial: Personalizing therapy for lung cancer. Cancer Discov 1(1):44-53, 2011.