Given the success of checkpoint inhibitors and the desire to test them in combination with other immunotherapies and targeted therapies, hundreds of clinical trials have been launched. To most efficiently study these agents, researchers and the FDA are exploring the use of novel endpoints, the use of new preclinical models, and adaptive trial designs. However, the cost and demands associated with the conduct of increasingly sophisticated early-phase clinical trials are putting smaller companies and some academic medical centers at a disadvantage.

Since the FDA approved the first checkpoint inhibitor, ipilimumab (Yervoy; Bristol-Myers Squibb), in 2011 to treat metastatic melanoma, immunotherapy research has exploded. Six checkpoint inhibitors have now been approved by the agency to treat a variety of cancers, and more than 1,000 clinical trials involving the agents are under way. In addition to changing the way that many cancers are treated, immunotherapies—and in particular checkpoint inhibitors, which are increasingly being tested in combination with one another and with other agents—are beginning to alter clinical-trial design.

Traditionally, overall survival has been the primary endpoint in oncology trials, but survival data take a long time to mature. “We don't want to wait a decade for overall survival to give us a statistically valid assessment of whether treatment a + b is better than treatment a,” says John Kirkwood, MD, of the University of Pittsburgh Cancer Institute in Pennsylvania. In addition, “as cancer patients are living longer due to improved therapies, powering a study for overall survival will become increasingly challenging,” says Marc Theoret, MD, acting associate director of Immuno-oncology Therapeutics at the FDA.

In the past, investigators tackled this problem by measuring progression-free survival (PFS). However, PFS may not always be a reliable measure of checkpoint inhibitor effectiveness, says Leisha Emens, MD, PhD, of the Johns Hopkins University School of Medicine in Baltimore, MD. Patients may take longer to respond to immunotherapy than other treatments, so the disease may advance before physicians see a response. Additionally, pseudoprogression, a temporary increase in lesion size due to the aggregation of immune cells, may be mistaken for progression.

“The best endpoints aren't clear,” says Emens. “We have to think about using alternative measures, such as the durable rate of response—some composite endpoint of response rate and disease control rate.” The FDA has already approved an immunotherapy—the oncolytic virus talimogene laherparepvec (Imlygic; Amgen), which is used to treat melanoma—on the basis of a phase III trial in which durable response rate was the primary endpoint. In addition, investigators have recently developed alternative guidelines for measuring response to immunotherapies, including irRECIST, mRECIST, and iRECIST (Cancer Discov 2017;7:446–7).

Another challenge facing the immuno-oncology community is the sheer number of possible therapeutic combinations involving checkpoint inhibitors. Drug companies are eager to combine checkpoint inhibitors with other drugs in their portfolios, including other checkpoint inhibitors, targeted agents, and traditional chemotherapies. New checkpoint inhibitors are being developed as well, further increasing the number of potential combinations. “There are a lot of immunotherapy drugs in development, so trying to figure out how to prioritize them alone and in combination is a huge challenge,” says Emens.


Treatments that are deemed effective will be “advanced” to a phase III trial; those that do not extend progression-free survival will be dropped. New treatment arms can be added throughout the study. Biomarker signatures from all patients will be monitored and correlated with response

Preclinical models play an important role in this process, says Geoffrey Shapiro, MD, PhD, of Dana-Farber Cancer Institute in Boston, MA. He emphasizes the importance of rapid and widespread development of useful and translatable preclinical models. In particular, humanized mouse models that feature both a competent immune system and the ability to accept human cells offer a promising path forward. “There's a rush to mix drugs with PD-1/PL-L1 blockades, but much more work in these types of preclinical models needs to be done to assess potential detrimental interactions that could occur, as well as favorable interactions,” he says.

Given the many checkpoint inhibitors and combinations that investigators expect to test, designing trials for maximum efficiency is key. “Using traditional clinical trials to explore multiple combinations would potentially lead to hundreds of clinical trials that would take many years to read out,” says William Grossman, MD, PhD, of Genentech. He also notes that comparing the results of these trials to identify the most beneficial combinations would be virtually impossible, which makes adaptive trials increasingly attractive.

“We need to do trials that will involve 50 to 100 patients, not 500 to 1,000 patients,” says Kirkwood, explaining that adaptive trials require fewer patients because multiple treatment arms can be evaluated against a single control arm. Moreover, if one arm is performing better than another, patients can be reassigned to the more successful treatment and the less successful treatment can be dropped.

One such trial is Genentech's just-launched MORPHEUS, which will begin by testing 17 first-in-disease immunotherapy combinations in four tumor types: HR-positive breast, pancreatic, gastric, and non–small cell lung. To carry out such a complex trial, Genentech is collaborating with academic research sites worldwide—and with other drug companies whose agents will be included. Biomarkers will be a significant focus of the trial, says Grossman, “to better understand why certain patients respond to cancer immunotherapy, and just as importantly, why certain patients do not respond to cancer immunotherapy.”

In addition, a major adaptive trial in melanoma, MICAT, is expected to launch within the next year. Developed and supported by a diverse set of partners, including AIM at Melanoma, the International Melanoma Working Group, and drugmakers, it showcases an organizational model for trials in which a single large pharmaceutical company is not taking the lead. “Whereas the largest pharmaceutical companies can execute adaptive trials on their own, smaller firms cannot,” explains Kirkwood, a key figure in organizing the trial. “They either don't have multiple agents in their own pipeline, or they don't have the resources.” Therefore, partnering with nonprofits and other drug companies offers small and mid-size companies an opportunity to test their agents in an adaptive framework (see table).

Of course, all changes in trial design ultimately have to be acceptable to the FDA, and the agency is working with industry, academia, and nonprofits to develop workable trial designs that establish drug safety and efficacy. As part of that effort, the FDA and the American Association for Cancer Research have held an annual Oncology Dose-Finding Workshop for the past few years to discuss concerns and lay out plans for trials such as MICAT.

These conversations “help companies make go/no-go decisions and prioritize combinatorial approaches,” says Theoret. The FDA is working with groups such as Project Data Sphere and the Foundation for the National Institutes of Health to encourage investigators to share data from completed clinical trials, he notes, in order “to develop better metrics to predict response and survival with immuno-oncology products.”

“I think the FDA is very much embracing the challenges of drug development in this field,” says Emens, pointing to pembrolizumab (Keytruda; Merck) as proof. “The initial approval for pembrolizumab was in the context of a phase I adaptive trial that grew.”

The novel endpoints, new preclinical models, and adaptive designs being used to study checkpoint inhibitors are likely to spread beyond the field of immuno-oncology. At Genentech, for example, “the approaches and lessons learned from MORPHEUS are likely to impact how we think about approaching traditional drug development both in and outside of oncology,” says Grossman. –Kristin Harper

Footprint of Early-phase Trials Grows

Traditionally, phase I trials have enrolled a small number of participants to assess drug safety and optimal dosing. Nowadays, however, adaptive phase I trials may involve thousands of patients, huge amounts of data, and even multiple diseases, says Geoffrey Shapiro, MD, PhD, of Dana-Farber Cancer Institute in Boston, MA. In addition, “there's a real push not to have too many dose levels, to start as close as possible to where you think you need to be based on preclinical modeling.”

Researchers are also now collecting larger amounts of sophisticated biological data in phase I trials, to better understand how treatments work—or fail to work. “We're gathering a lot more pre- and on-treatment research materials from patients, whether that be blood or tumor tissue, to understand what happens during the course of an intervention,” says Stanley Riddell, MD, of the Fred Hutchinson Cancer Research Center in Seattle, WA. He notes that techniques such as single-cell RNA sequencing offer an unprecedented opportunity to learn how immunotherapy changes the tumor microenvironment.

Although squeezing more information from early-phase trials minimizes the time needed for drugs to reach patients, investigators caution that once a drug is approved, it may become more difficult to study it and optimize its use. “An approval may be based on a dosing schedule that is not optimal, and exploring shorter or cheaper schedules may be seen as unethical, given the approval already in place,” explains Emile Voest, MD, PhD, of the Netherlands Cancer Institute in Amsterdam.

In addition, the cost of these sophisticated early-phase trials may put academic researchers at a disadvantage. Riddell says that although the NIH played a vital role in the initial discoveries that led to the development of immunotherapies, federal funding has not kept pace. “Biotech or pharmaceutical companies have more resources to put into trial research,” notes Riddell. “The amount that we can get from standard NIH grants really doesn't cover the sophisticated analytics we could do—and all of the information that we could gain.”

Early-phase trial investigators at academic medical centers are scrambling to accommodate the demands of adaptive phase I trials. “These studies have definitely put stress on early-phase groups,” says Shapiro. “They must be able to do phase I dose-escalation work, with very careful toxicity assessment, while also being able to transition to enrolling large numbers of cancer patients as quickly as possible.”

Early-phase data teams may need to expand to handle the workload, or centers may need to recruit partners to rapidly enroll patients during a trial's expansion phase. “The best trial centers will respond by becoming more facile and nimble,” Shapiro predicts. –KH

For more news on cancer research, visit Cancer Discovery online at