Clinical Trial Designs for Studying Targeted Therapies

Clinical Trial Designs for Studying Targeted Therapies

As therapeutic research has shifted focus from cytotoxic agents to targeted drugs, the primary challenge is to understand drug activity in molecularly defined subsets of patients, and the current drug development paradigm may not be appropriate for handling this challenge. Strategies have been developed over the last several years to help expedite the drug development process and ensure that patients receive the most appropriate therapies. Several of these strategies-—enrichment trial, umbrella, and basket trial designs—were discussed in the Education Session “Improving Clinical Trial Efficiency: Thinking Outside the Box,” on Sunday, May 31.

Variations on Enrichment Designs

Enrichment design, or targeted design, has become the most common phase III design for the simultaneous development of new cancer drugs with companion diagnostics. It has successfully been used in the study of multiple targeted therapies in several different cancer settings, including trastuzumab and pertuzumab for HER2-overexpressed breast cancer. In the classic setting, patients must screen positive with the diagnostic test in order to be eligible for the clinical trial. Patients are then randomly assigned to receive either the test drug or control treatment.

The design is most appropriate in settings in which patients who are biomarker negative are not likely to benefit from the novel drug, as demonstrated through strong phase II data or a biologic rationale supporting the role of the biomarker in disease. In these cases, the use of enrichment design can avoid exposing patients to inappropriate drugs and adverse effects. The design can also reduce sample size requirements, as a direct result of increases in effect size, although it may still be necessary to screen a large group of patients with the diagnostic test.

As part of the enrichment design, a run-in period can be conducted if it is predicted that fewer than half of patients will benefit from the treatment.1 In this setting, patients are screened for a biomarker, but, prior to random assignment, all participants receive the drug for a short period of time. Their response, which could be pharmacodynamic, immunologic, or based on tumor biology, is used as a predictive biomarker. Patients who have a response are assigned to the test treatment or the control regimen as in a typical enrichment design.

Dr. Richard Simon
 Two variations of the enrichment design can be implemented in settings in which patients who are negative for the biomarker could benefit from the drug, or when the cut-point for at least one of the candidate biomarkers is unclear. Regarding the first scenario, Richard Simon, DSc, of the National Cancer Institute (NCI), described what he called an adaptive stratification design. In this approach, patients are screened for a biomarker, and then randomly assigned irrespective of biomarker status. An intermediate endpoint, such as progression-free survival, would be measured and used to determine the predictive probability that, if completed, the trial would allow rejection of the null hypothesis. If the predictive probability is high, the trial could continue as is; if the predictive probability is low, the trial could switch to exclusive accrual of patients who are biomarker positive.

For the second scenario, in which the biomarker cut-point is unclear, an adaptive enrichment design could be used. In this framework, the eligibility criteria are restricted over the course of the trial based on interim analyses that use a pre-determined decision algorithm and can include cut-point, biomarker status, or other factors. At the end of the trial, the primary statistical significance test is performed using all of the randomly assigned patients. “Adaptive enrichment with only one interim analysis can greatly improve the power to reject the global null hypothesis,” Dr. Simon said.

Umbrella Design Creates Infrastructure for Multiple Trials

The umbrella design focuses on a single tumor type or histology. It involves a group of two or more enrichment designs, or substudies, that are connected through a central infrastructure that oversees screening and identification of patients. The trials can be phase II and exploratory or proof of concept, or they can be phase II/III or III definitive trials.

“The reason and rationale for the umbrella trial design first and foremost is to facilitate screening and accrual,” Sumithra J. Mandrekar, PhD, of Mayo Clinic, said. Because of the central platform, a large number of patients can be screened for multiple biomarkers, which is particularly beneficial for low-prevalence markers. The design also affords sufficient hit rate so patients and physicians stay engaged.

At least six umbrella trials are currently ongoing or have been completed. Dr. Mandrekar discussed four: BATTLE, whose results have been published, as well as ALCHEMIST, FOCUS 4, and Lung-MAP, which are ongoing. In the ALCHEMIST trial, coordinated through NCI, new arms will be added in the near future for EGFR wild-type and ALK-negative cohorts. The infrastructure of the umbrella design allows the creation of new substudies to test new drugs and biomarkers.

Setting up the infrastructure for umbrella trials, including the collection and flow of data, is a complicated process. Once a patient is registered, samples are sent to centers that perform genotyping and tissue analysis, and these data are stored along with clinical data in a repository. The centralization of testing ensures standardization; however, as Dr. Mandrekar noted in the question-and-answer period, results from various local laboratories could be less reproducible, as in the context of real-world data sets. An advantage of housing large data sets in the umbrella design is that patients who do not enter into substudies can be tracked for future trials for which they could be eligible, Dr. Mandrekar said.

Basket Trials Explore Treatment Options for Multiple Tumor Types

Dr. Suzanne Dahlberg

In contrast to the umbrella design, basket trials allow the study of multiple molecular subpopulations of different tumor or histologic types all within one study. They can include highly rare cancers that would be difficult to study in randomized controlled trials. Genetic aberrations leading to changes in BRAF, HER2, and EGFR expression, for example, can play different roles in different tumor types and in the context of different drugs. Uncertainties remain about whether an aberration is actionable, and “resolving these uncertainties is the reason for conducting these types of studies,” Suzanne Dahlberg, PhD, of the Dana-Farber Cancer Institute, said. Because the basket design affords the flexibility to continually open and close arms of the study, many drugs for many different diseases can be screened.

Basket trials are discovery-based and can be phase I trials. Many of the early-phase immunotherapy studies used the basket design. Similar to the umbrella design, there is a central screening and treatment infrastructure. As with both the umbrella and enrichment designs, it is important to have evidence supporting the potential usefulness of a biomarker for the selection of drug treatment, Dr. Dahlberg said.

The NCI-MATCH: Molecular Analysis for Therapy Choice study is a basket trial led by Dr. Dahlberg’s colleagues at NCI and ECOG-ACRIN Cancer Research Group, which is scheduled to start enrollment in the next few weeks. Up to 3,000 patients with refractory solid tumors or lymphoma will be screened, with the aim of enrolling 35 patients into 20-25 biomarker subgroups. The drugs studied in NCI-MATCH have predictive biomarkers and include both U.S. Food and Drug Administration–
approved and investigational drugs.

Myriad challenges exist both in designing a basket trial and interpreting the data generated by the trial, such as writing 20 or more protocols—one for each cohort—and creating a screening and treatment infrastructure. At the end of the trial, there could more answers than questions, Dr. Dahlberg said. Because of the small patient population in these types of trials, it can be difficult to determine if outcomes are meaningful. “At the heart of the issue is, how do we define an exceptional responder?” she asked. Identifying this group of patients, who have excellent outcomes associated with a treatment, could require special statistics or endpoints built into the study.

Watch the session: Visit the ASCO Virtual Meeting website.