Celebrating the 10th Anniversary
A decade of discoveries in Cancer Discovery. For the community. By the Community.
Their impactful studies. In their words.
Remembering the late Waun Ki Hong and his extraordinary contribution to Cancer Discovery’s first clinical trial research article.
Comments by Senior Author Scott Lippman, MD.
The landmark Biomarker-based Approaches of Targeted Therapy for Lung Cancer Elimination trial, or BATTLE, was conceived in 2004 by Waun Ki Hong at a time before precision targeted therapy as we now know it began. This trial began even before the breakthrough discovery of EGFR activating mutations and their early characterization as a profound predictive marker for EGFR TKIs. The simplicity of the idea of assigning patients to treatments most likely to work based on all available data, including real-time outcome data from BATTLE itself, was rivaled only by the complexity and difficulty of its design, implementation, conduct, and interpretation. Although debate surrounding BATTLE focused almost entirely on the exploratory biomarker groups and drugs, Bayesian adaptive design and core needle biopsy (CNB) in the 2nd line setting—it’s important to note that even the very basic choice of clinical trial primary endpoint was a major design issue at the time. The novel Bayesian inferential adaptive framework required a valid short-term clinical efficacy endpoint, but there were no established short-term endpoints of clinical benefit in this setting until a fortunately timed SWOG trial presentation at the 2006 ASCO Annual Meeting established disease control rate (DCR) at week 8 as a robust predictor of subsequent survival—providing the final, vital piece of evidence we needed to justify the primary endpoint--and thus related statistical assumptions needed for power and sample-size calculations--just months before BATTLE was activated in late 2006.
We agreed to take a chance on a new journal, first issue, no impact factor. Ki had a unique way of motivating and inspiring people, and a keen, remarkable sense of prescience.
Virtually every aspect of this trial was new and highly innovative. Although clinically untested in this context, the novel complex Bayesian adaptive design was shown in BATTLE to be well accepted, feasible, and successfully applied. The novelty, innovation, and impact of BATTLE follow-on reports were wide-ranging and multidisciplinary, including primary statistical methods and the interventional radiology publications. BATTLE was a pinnacle team-science effort, a remarkably "Manhattan-Project"–like approach bringing together impressive teams, including molecular pathologists, imagers, and interventional radiologists, medical and surgical oncologists, statisticians, ethicists, and computational and basic scientists. Study accrual was so much better than expected, highlighting the power of this concerted team-science trial with patients, investigators, and sponsors all working together toward a common goal.
It was such a new approach, with substantial and reasonable uncertainty related to targets, biomarker groups, drugs, and adaptive design framework at the time that funding agencies and many colleagues doubted its feasibility. The ethics of this design, which mandated risky CNBs to obtain essentially exploratory, rapidly evolving biologic information, in heavily pretreated patients with recurrent and metastatic non–small cell lung cancer (NSCLC) who had received up to 9 prior therapies, also drew major attention. This concern was further accentuated by the fact that eligible patients may have had CNBs of previously radiated, densely fibrotic lesions, creating the real possibility or question as to whether these CNBs would even yield enough viable tumor tissue for analysis of the primary pre-planned molecular biomarker panel using first-generation sequencing technology from the early 2000s.
Hong remarkably secured the cooperation of several different drug companies, testing and comparing their drugs head-to-head with drugs from other companies in the same trial (an incredible feat in itself), and federal funding from the U.S. Department of Defense. The difficulty in establishing the infrastructure and multidisciplinary collaborations necessary to successfully carry out 254 CNBs with real-time genomic sequencing and multiplexed biomarker analyses cannot be overstated. The second-line NSCLC trial space was very crowded at that time in my department, with a number of ongoing trials competing for the same patient eligibility criteria, adding further prioritization pressures for BATTLE accrual. There was also great pressure to accrue rapidly while many BATTLE translational studies remained relevant in a constantly changing treatment landscape (e.g., the nearly entire ALK translocation/crizotinib precision therapy bench-to-bedside breakthrough and clinical delivery occurred during the course of BATTLE). Ki was legendary for the impeccable rigor of his clinical and translational research throughout his career, peaking with BATTLE, which had an unprecedented intense, rigorous real-time surveillance during standing weekly 5 am meetings to closely monitor accrual, potential CNB complications, adaptive design, new information from the literature and other sources, and other unknown unknowns at that time. This intense monitoring was to allow us to be as dynamic and flexible as possible within the constraints of a randomized phase II clinical trial.
Although it was a daunting, seemingly impossible task, the approach worked in spectacular fashion, randomizing 254 patients in fewer than 3 years, well ahead of the protocol-planned schedule, remarkable for any randomized trial, even more amazing given the many uncertainties and challenges (several noted above) facing BATTLE--unequivocally establishing the feasibility of an incredibly challenging precision medicine design to become a leading edge in cancer medicine. This study showed that our exceptional Interventional Radiology team’s experience, expertise, and CNB procedures were remarkably safe (<1% incidence of serious events this second-line setting) and that real-time biomarker assessment was possible (83% of patients had enough viable tumor to be characterized and categorized molecularly), then fed into the adaptive randomization algorithm for allocation of patients into the most promising predefined biomarker-treatment group (among the 4 treatments, 5 biomarker group options). However, despite the intense interest in the concept of precision targeted therapy, very little was actually known at this time for lung cancer. Our study, therefore, ended up with some important limitations. First, and probably most important, our biomarker groups (which drove the adaptive randomization) were less predictive in some cases than individual biomarkers, which diluted the impact of strong predictors in determining treatment probabilities. By far, the most striking example of this problem involved the “EGFR" biomarker group. Lumping EGFR activating mutations with EGFR copy number gain, the striking predictive difference between these two EGFR metrics became unequivocally apparent relatively early during the trial, but there was no way to fix or address this after trial activation, with the dynamic adaptive design or in any other way and still maintain trial integrity. In retrospect, the limited biomarker–drug knowledge at this time provided a rapidly closing window of opportunity and relevance for BATTLE; i.e., if designed a couple years later, it would have been extremely difficult to consider or even imagine such a randomized trial design at a time when EGFR activating mutations, ALK fusions, etc., and matched targeted drugs were well known, characterized, and available. The unfortunate decision to group the EGFR markers also created the serious ethical possibility that patients with EGFR activating mutations may be assigned to non-EGFR TKI treatment arms and also impacted the other marker groups and their interactions with other treatments, resulting in a suboptimal overall DCR. On the other hand, the KRAS mutation marker group remained valid as there were (and still are) no proven KRAS-targeting drugs or predictive markers before, during, or after BATTLE, and the adaptive design did direct KRAS mutant–positive patients away from the EGFR inhibitor arms, later proven ineffective against KRAS-mutant lung cancers. In fact, BATTLE exploratory studies and follow-on publications centered around KRAS and sorafenib –e.g., predictive mutations, markers, and signatures—were important translational discoveries still relevant today.
As strange as it sounds now, it was actually a hard sell to submit BATTLE to Cancer Discovery at that time, given the broad, pivotal importance and implications of this trial during and driving the initial stages of precision therapy. Most co-authors wanted to submit the primary report to an established, high-impact journal. I remember discussing this with Ki on several occasions, but eventually we agreed to take a chance on a new journal, first issue, no impact factor. Ki had a unique way of motivating and inspiring people, and a keen, remarkable sense of prescience—the paper has been cited >850 times and the Journal’s impact trajectory has been remarkable--and of course, an incredible sense of loyalty, a central theme of his remarkable life --in this case his intense belief, commitment, and enthusiasm in this new journal, deeply confident by the full-on backing and vision of AACR, and the inimitable Dr. Marge Foti.