Basket and umbrella studies as well as adaptive enrichment design strategies represent novel approaches to testing targeted therapeutics in oncology, and a session at ASCO's annual meeting discussed the nuts and bolts of these design strategies.
Basket and umbrella studies as well as adaptive enrichment design strategies represent novel approaches to testing targeted therapeutics in oncology. These approaches have evolved rapidly in the last 2 to 3 years, with the objective of accelerating the drug development process so that appropriate therapies can be delivered quickly to suitable patients. At an early morning session on the third day of the annual meeting of the American Society of Clinical Oncology in Chicago, speakers discussed the nuts and bolts of these design strategies, the underlying statistical challenges, the logistical barriers with trial implementation, and the interpretation of results.
Richard Simon, PhD, DSc, who heads the biometric research branch in the Division of Cancer Treatment and Diagnosis at the National Cancer Institute, discussed tools to enrich clinical trial (CT) design.
Cancer is a heterogeneous group of diseases at its primary site with respect to sensitivity to treatments, he explained. Many of these treatments are expensive and work only in a subset of patients. Standard CT design can generate a high number of false negative results, according to Simon, while trials that do yield positive results may have only a small proportion of eligible patients. Therefore, he explained, it’s essential to have an elevated “number needed to treat” when designing oncology trials.
“How do we generate reliable evidence that a specific treatment will work in a particular subset of the population?” Simon asked, before describing adaptive enrichment and enrichment stratification (see Figure 1).
With the enrichment design, patients are first evaluated for biomarker expression; those who test positive are then deemed eligible for inclusion in phase 3 of the study while those who do not express the biomarker are removed. Simon indicated that this design is appropriate for phase 2 trials; if these generate biological evidence which indicates the drug is ineffective, that will point to the need to enrich the cohort being tested, and probably the need to develop a companion diagnostic as well.
Enrichment design, Simon said, has successfully been implemented in trials evaluating drugs for HER2-overexpressing breast cancers (trastuzumab), BRaf-mutated melanoma (vemurafenib), and ALK-positive lung cancer (crizotinib). Adaptive enrichment involves introducing restricted eligibility criteria at fixed interim analysis points. At the end of the trial period, a statistical significance test is performed. This design has a fixed sample size regardless of changes in eligibility, except if the trial is terminated.
The advantages of this kind of trial design, he explained, are the clarity of interpretation, and the fact that it spares the patient unnecessary exposure to the drug, particularly in cases in which the drug may not be effective. “This design helps develop a predictive biomarker, not a surrogate end point,” Simon specified.
In the case of a single binary biomarker, Simon said, where we do not want to assume that biomarker-negative patients will not benefit, an adaptive stratification design would be suitable. In this case, he explained, patients are randomized to receive either the new treatment or the control treatment and an intermediate end point is introduced during the trial to analyze results, such as progression-free survival.
Umbrella studies are designed to test the impact of different drugs on different mutations in a single cancer type, and the BATTLE trial is an example of such a trial design. Sumithra J. Mandrekar, PhD, professor at the Mayo Clinic, Rochester, Minnesota, presented the rationale behind the umbrella trial design.
Mandrekar showed the umbrella trial design scheme (see Figure 2), which she said allows for a central infrastructure with multiple subtrials to test different regimens within molecularly defined patient subsets; the various subsets can share a control arm.
The trial design assumes that the biomarker and its effects on the tumor are well understood, said Mandrekar, adding that while this design has minimal or no prognostic impact, it has predictive potential. “The goals of this trial design are to facilitate patient screening and accrual, and it is quite suitable for trials evaluating low-prevalence disease,” said Mandrekar, adding that the design can accelerate the speed of development and may prove useful for the rapid approval of new drugs. She then provided a list of umbrella trials for lung cancer (see Table) at various stages of drug development. A majority of these, she said, are biomarker-driven.
Table. Ongoing Umbrella Trials
Adjuvant non-squamous NSCLC
Phase 2 followed
by phase 3
Discovery and confirmatory
Previously treated squamous lung cancer
National Lung MATRIX trial
Single-arm phase 2
NSCLC indicates non-small cell lung cancer.
The primary features of umbrella trials, according to Mandrekar, are:
She did point out several logistical challenges associated with implementing a large-scale endeavor like an umbrella trial. These include acquiring patient consent and patient enrollment and tracking—“It takes a huge team effort.”
Basket studies are designed to test the effect of a single drug on a single mutation in a variety of cancer types. They provide a unique way of merging the traditional CT design with rapidly evolving genomic data that facilitate the molecular classification of tumors. During her talk, “Basket Trial Designs: Identifying the Exceptional Responders,” Suzanne Eleanor Dahlberg, PhD, research scientist at the Dana-Farber Cancer Institute, introduced this other novel trial design.
Basket trials aim to assess targeted therapeutics that have a dramatic clinical impact, with a focus on biological drivers of response. “So patients who harbor a specific mutation or over-express a particular protein, targeting that particular abnormal signaling pathway could yield a dramatic improvement in patient response,” Dahlberg said.
She noted that although a basket trial is an efficient way to screen multiple drugs across many patient populations, it is not a formal statistical design. Rather, it is designed based on a genetic abnormality in the patient’s tumor.
Emphasizing that the trial design can greatly improve trial efficiency, Dahlberg said that basket trials can screen multiple drugs across many cancer types. While genomic variability exists across multiple tumor types, not every mutation is necessarily actionable across all of them. She believes the basket design provides a strong rationale to pair a drug with a validated biomarker in a specific tumor.
Noting that these are discovery-phase trials, Dahlberg explained that they can be used for drug development in rare cancers. “The trials can be conducted across multiple institutions, rely on sample availability, and need a sufficient number of drugs that can target multiple tumor types.” Citing NCI-MATCH (Molecular Analysis for Therapy Choice Program) as an example of a basket trial design,1 Dahlberg said that it is a collaboration between the ECOG-ACRIN Cancer Research Group and the National Cancer Institute. The trial, which at press time was scheduled to initiate enrollment in July, will assign treatment based on “actionable mutations” in the tumor. Each of the 10 arms in the trial will enroll adults with advanced solid tumors and lymphomas who are refractory to standard therapy.1
Several criteria influence the drug selection process, she said. The molecules could be FDA-approved for a predictive indication, or have a biomarker; investigational drugs can be included if they have predictive molecular value.
Dahlberg concluded: “Basket design can accelerate the delivery of the right treatment to a patient, but it requires that strong biomarkers be associated with the drug. Additionally, heterogeneity in response across disease types is a primary consideration, and clonal variation has to be adapted during trial design.”Reference
1. NCI-Molecular Analysis for Therapy Choice Program (NCI-MATCH). National Cancer Institute website. . Accessed June 12, 2015.