Genomic Data Changing Design of Oncology Clinical Trials
Drug discovery is known to be a long, expensive process. On average it can take decades and cost more than $1 billion to bring a drug to market. Yet, despite the large investment in time and money, less than 10 percent of drugs succeed in obtaining U.S. Food and Drug Administration (FDA) approval. Two factors are driving significant changes to the drug discovery and approval process—the sharp drop in the cost of high-throughput genomic sequencing and the growing recognition that molecularly targeting therapies to a smaller subset of patients increases effectiveness. Randomized controlled trials (RCTs) have been the gold standard for the design of clinical trials needed to obtain drug approval. But in the era of personalized medicine RCTs may not be the best fit. RCTs require enrolling hundreds or often thousands of patient participants. After the fact, post-hoc analysis identifies if there is a population subset that particularly benefited from the therapy. However, personalized medicine focuses on identifying baseline predictive markers that will reduce the trial and error effect in therapy selection. Increasingly, these baseline predictive markers are being incorporated early in the drug discovery process. So, experts say that further along the clinical trial spectrum—in phase II and […]
Drug discovery is known to be a long, expensive process. On average it can take decades and cost more than $1 billion to bring a drug to market. Yet, despite the large investment in time and money, less than 10 percent of drugs succeed in obtaining U.S. Food and Drug Administration (FDA) approval. Two factors are driving significant changes to the drug discovery and approval process—the sharp drop in the cost of high-throughput genomic sequencing and the growing recognition that molecularly targeting therapies to a smaller subset of patients increases effectiveness.
Randomized controlled trials (RCTs) have been the gold standard for the design of clinical trials needed to obtain drug approval. But in the era of personalized medicine RCTs may not be the best fit. RCTs require enrolling hundreds or often thousands of patient participants. After the fact, post-hoc analysis identifies if there is a population subset that particularly benefited from the therapy. However, personalized medicine focuses on identifying baseline predictive markers that will reduce the trial and error effect in therapy selection. Increasingly, these baseline predictive markers are being incorporated early in the drug discovery process. So, experts say that further along the clinical trial spectrum—in phase II and phase III trials—it does not make sense to test drugs in patients who are known to not have the markers associated with therapeutic benefit. Targeted drug trials will naturally require fewer participants.
There is a lot of interest in innovative trial design for targeted therapies to address the challenges created by the increasing molecular fragmentation of diagnosis, particularly for cancer.
“Due to better understanding of genetic effect in disease or disease caused by different molecular features, contemporary trial designs are used not only in oncology but also in psychiatry and neurology and might be extended to all other medical fields,” writes co-author Ja-An Lin, from University of North Carolina at Chapel Hill, in a review of innovative biomarker trial designs in the British Medical Bulletin last year. “The development in medical research motivated the invention of new clinical trial designs, and the revolutionary trials can help the medical research move forward to better understand the disease mechanism and eventually lead to better treatment to serve the patients.”
New Models for Clinical Trials
Enriched models first screen patients to identify those who have the mutation targeted by the treatment in the trial. The trial sample is “enriched” with those patients expected to have a treatment response because of a specific biomarker. The targeted population provides the advantage of more likely validating the effectiveness of a particular treatment in a smaller, cheaper trial, compared to all-comer trials, but runs the disadvantage of not knowing the effect of treatment in a general population. Some enrichment models with a hybrid design allow those patients who do not have the responsive biomarker to serve as a control arm, receiving the standard of care.
Unlike enriched models, which are still testing a single drug response against a single therapy, in adaptive trials researchers can essentially conduct several trials in parallel. Adaptive models have multiple arms and use Bayesian approaches to incorporate information to identify the cohort which is most benefited by intervention treatment gained early in the study to modify the trial implementation later on.
Last year researchers from Memorial Sloan Kettering Cancer Center (New York) published the first results showing the promise of basket studies for studying cancer drugs. Basket studies focus on a specific gene mutation, regardless of cancer type. The histology- independent, biomarker-selected, early phase 2 basket study led by Memorial Sloan Kettering enrolled 122 patients from 23 centers around the world. All patients had BRAF V600 mutations, but across a total of more than six nonmelanoma cancer types. Interestingly, the basket trial design enables mutation-positive people with rare tumors to participate in trials that traditional histology-based studies did not.
“We have no fantasies that blocking one pathway will do the trick in most cases,” explains Memorial Sloan Kettering physician-in-chief José Baselga, M.D., Ph.D., the study’s senior author, in a statement. “But the repertoire of pathways that these tumors rely on is not endless—it’s finite. The second wave of these [basket] trials will be appropriate combinations, and this trial is a pioneer for that as well. It’s the way forward.”
By contrast, an umbrella study enrolls patients with a single tumor type or histology but contains multiple sub-trials, each testing a targeted therapy within a molecularly defined subset. This approach can test multiple therapies and multiple markers simultaneously. Also using Bayesian approaches, umbrella studies can add, modify, or drop subtrials based on data generated from the ongoing trial or evolving cancer research field. Umbrella trials usually are performed nationally at multiple clinical sites using a common genomic screening platform.
Biomarkers Increase Trial Success
The biotech trade association Biotechnology Innovation Organization (BIO; Washington, D.C.) partnered with the contract research organization Amplion (Bend, Oregon) to analyze the effects of biomarkers in clinical trial success. To inform the newly released report, Clinical Development Success Rates, 2006-2015, the organizations used data from their respective subscription-based databases, Biomedtracker, which tracks the clinical development and regulatory history of investigational drugs, and BiomarkerBase that tracks biomarker usage in clinical trials, drug labels, and tests (including laboratory-developed, FDA-cleared, and FDA-approved tests).
From 2006-2015, the study identified a total of 9,985 clinical and regulatory phase transitions from 7,455 development programs, across 1,103 companies. The success rate at each of the four phases of development—Phase I, II, III, and regulatory filing—was assessed for 14 disease areas.
Analysis found that the overall likelihood of approval (LOA) across all stages of development was 9.6 percent. Phase II clinical programs continued to have the lowest success rate, with only 30.7 percent of developmental candidates advancing from Phase II to Phase III.
The use of “selection” biomarkers to establish inclusion or exclusion criteria for enrolling patients into clinical studies has increased “dramatically,” the report says, but still remains small. Out of all 9,985-phase transitions, 512, or five percent, of transitions incorporated a selection biomarker for patient stratification. However, programs that utilized selection biomarkers had higher success rates at each phase of development, compared to the overall dataset.
“The large differences in Phase II and III transition success rates are quite convincing, quantitatively, of what many drug developers have long argued anecdotally—enrichment of patient enrollment at the molecular level is a more successful strategy than heterogeneous enrollment,” the report says.
Use of a selection biomarker raised the LOA from Phase I to one in four versus less than one in 10 when no selection biomarker was used. The largest percentage difference among the four phases of development by selection biomarker use was seen in Phase III, where transition success rates for selection biomarker programs were 76.5 percent (n=132), compared to 55.0 percent (n=1,254) for programs not using a selection biomarker.
“The higher success rates for trials run with biomarker-selected patients suggests that the broader industry is already on the right path,” the report says. “Experiencing one in four failures (with selection biomarkers) versus two in four Phase III failures (without biomarkers) could have significant cost implications.”
Launch of New Trial Models
With positive proof-of-principle results from these innovative trials demonstrating early successes, the number of active innovative trials is expected to continue to grow. Below is a sampling of current trials.
This fall, the European trial AcSé eSMART will launch in collaboration with the Innovative Therapies for Children with Cancer consortium. eSMART is a phase I/II basket-and-umbrella trial for children with relapsed or refractory disease that will explore the effectiveness of 10 investigational oncology drugs from at least three different pharmaceutical companies— as single agents and in combination—based on the results of pangenomic tumor profiling.
The French National Cancer Institute originally launched AcSé (Secured Access to Innovative Therapies) in 2013 to promote safe access to targeted therapies outside of their approved indications for patients lacking treatment options. Through this program, patients undergo molecular testing and, if appropriate, receive targeted therapy within the oversight of a phase II trial, if there are no other trials for which they are eligible. The 2013 proof-of-concept basket study investigated single-agent crizotinib in adults and children with an advanced-stage, incurable malignancy harboring ALK, MET, or ROS1 alterations. In France, crizotinib is only indicated for adults with ALK-positive nonsmall cell lung cancer. However, research evidence shows that more than 15 different malignancies in adults and children feature molecular alterations responsive to crizotinib treatment.
In the AcSé trial, 107 pediatric patients were tested for structural genomic alterations in ALK, MET, or ROS1 using either biomarker tests or pangenomic tumor profiling. Seventeen patients were identified that harbored positive tumors. Five of these patients met the eligibility criteria to enroll in a phase I trial of ceritinib, while 11 patients entered into the AcSé trial. The objective response rate reached 45 percent.
The MyPathway study is a basket-and-umbrella trial conducted at 39 U.S. treatment centers. It is designed to evaluate agents targeting the HER2, BRAF, Hedgehog, and EGFR pathways in tumor types outside of the approved indications.
As of December 2015, 129 patients with tumors from 25 different primary sites received targeted treatment based on their tumor’s molecular profile. Patients had to have exhausted all standard treatment options and potential clinical trials. Across all tumors, 64 percent had HER2 alterations, 26 percent had BRAF mutations, 6 percent had Hedgehog pathway mutations, and 5 percent EGFR mutations. The overall clinical benefit rate was 34 percent.
Sixteen U.S. cancer centers partnered to form the Lung Cancer Mutation Consortium to prospectively examine lung adenocarcinomas for genetic and molecular abnormalities and use that information to match patients to the best possible therapies. Tumor tissue from 875 of 1,020 eligible individuals with confirmed stage IV lung adenocarcinomas was probed for 14 different genetic and molecular abnormalities. More than half (54 percent) had a genetic abnormality. The most common mutations were KRAS (25 percent) and EGFR (12 percent) genes, followed distantly by ALK rearrangements (4 percent), MET amplification (3 percent), and BRAF mutations (2.5 percent). Twenty- eight percent of the total cohort (n=242) had a driver alteration. Of these patients, more than one-half (n=131) received targeted therapy. Patients with a driver mutation who received targeted treatment achieved a median overall survival of 2.7 years—an improvement of 1.2 years over patients with driver mutations who did not receive targeted therapy and 1.0 years over patients without targetable driver mutations.
Takeaway: The increasing ease of access to genomic data is reshaping the drug development and clinical trial processes. Innovative new models will increasingly test the effects of more markers and treatment candidates simultaneously.
Efforts to Improve Clinical Trial Recruitment
As the medical field gains a greater understanding of genomics, diseases are increasingly stratified into genetically defined subtypes, in addition to other inclusion criteria such as age, treatment history, or tumor stage. Finding patients for trials involving each subtype becomes more complex, causing enrollment to possibly drag out. As it already stands, experts say that less than five percent of cancer patients participate in clinical trials. Data companies are partnering with pharmaceutical companies and medical centers to try to improve trial recruitment through the use of advanced data-mining and analytical software.
One example of this is IBM’s Watson. In late 2014, IBM initiated its Patient Matching proof of concept pilot with the Mayo Clinic. More recently IBM has announced two additional partnerships aimed at expanding matching cancer patients with clinical trials. In June,
Froedtert & the Medical College of Wisconsin Cancer Network (Milwaukee) said it is adopting Watson cognitive computing software to help match a patient’s unique health profile with the best-suited cancer clinical trial in an attempt to further personalize approaches to cancer treatment. The matching program is slated to begin this fall.
In the fall of 2015, clinical research organization ICON (Ireland) began applying Watson Clinical Trial Matching to 25 breast, lung, colon and rectal cancer trials. “Recruiting the required number of patients for clinical trials is a constant challenge for our customers and can represent more than 30% of total study costs,” said ICON’s chief operating officer, Steve Cutler, Ph.D. at the time of the announcement. “By applying IBM Watson to our clinical trials, we have the potential to revolutionize clinical trial feasibility, patient recruitment and study start-up timelines which will help our customers take significant time and cost from their development programs.”
Subscribe to Clinical Diagnostics Insider to view
Start a Free Trial for immediate access to this article