Cancer Immunotherapies Have FDA, Industry Looking For New Endpoints

The unique effects of immunotherapy in cancer have FDA, industry and academia looking for new ways to measure efficacy in clinical trials.

However, discussions at a recent two-day meeting co-sponsored by FDA and the American Association for Cancer Research suggest stakeholders have a long way to go before settling on the most appropriate assessment tool in immuno-oncology drug development.

The traditional oncology endpoints of overall response rate, progression-free survival (PFS) and overall survival may not be the best measures of immunotherapy efficacy because they may underestimate treatment effects.

“What would have traditionally been called progression we recognized by the mechanisms of action of immunotherapy may actually not represent progression at all.” – Columbia University’s Schwartz

FDA staff, along with clinical trial researchers and industry representatives, discussed various alternatives to traditional endpoints, with suggestions ranging from modifications of existing criteria for tumor response to development of novel intermediate endpoints.

While the need for new endpoints has, to date, not proven to be a hurdle to regulatory approval of existing immunotherapies, such as the PD-1/PD-L1 checkpoint inhibitors, meeting participants said they will be needed as more immunotherapy combinations make their way through the clinic and are tested against increasingly more efficacious approved treatments.

Challenges With Pseudo-Progression

Unlike chemotherapy or targeted agents, immunotherapies do not have direct anti-tumor effects, and their clinical effects, which result from immune system activation, are delayed. This delayed response has proven problematic to capture with traditional clinical trial endpoints.

Also problematic are patterns of response that are not seen with traditional cancer agents. For example, in radiological scans a patient’s tumor may initially appear to have grown after immunotherapy but will appear smaller in subsequent scans. This phenomenon, often referred to as pseudo-progression or flare, can result in progressive disease being declared at an early assessment point before the immunotherapy’s clinical benefit has appeared.

The development of new lesions, or growth of non-target lesions, following immunotherapy also may lead to a premature declaration of progressive disease, even though such lesions may ultimately shrink or disappear.

“What would have traditionally been called progression we recognized by the mechanisms of action of immunotherapy may actually not represent progression at all,” said radiologist Lawrence Schwartz, Columbia University.

The need to consider alternative endpoints for immunotherapies first became apparent in the melanoma setting, said Maitreyee Hazarika, a medical officer in FDA’s Division of Oncology Products 2.

Bristol-Myers Squibb Co.’s Yervoy (ipilimumab), a CTLA-4-blocking antibody, was the first of the checkpoint inhibitors to reach the US market. It was approved in March 2011 for monotherapy treatment of unresectable or metastatic melanoma on the basis of a 34% improvement in overall survival, with a median four-month benefit. However, there was a more modest PFS benefit and the best overall response rate in the ipilimumab arm was only 10.9%. (Also see "Yervoy Approval Has Shifted Development Landscape For Melanoma" - Pink Sheet, 4 Apr, 2011.)

Tumor response-based endpoints did not adequately capture ipilimumab’s anti-tumor activity and would not have led to demonstration of effectiveness without the overall survival results, Hazarika said.

Two PD-1 inhibitors, Bristol's Opdivo (nivolumab) and Merck & Co. Inc.’s Keytruda (pembrolizumab), subsequently received accelerated approval for melanoma based on overall response rates of 24%-32%, but with response durations lasting more than 10 months, Hazarika noted. Low overall response rates with prolonged duration of response may reflect immune system dynamics, and small differences in median PFS may not represent the true treatment effect if benefits are delayed, she said.

Studies Need 'More Than Just A Guideline'

Standard RECIST (Response Evaluation Criteria In Solid Tumors) criteria used for determining when patients progress in clinical trials do not account for the unusual responses seen with immunotherapy. Although RECIST 1.1 has language that describes how to manage equivocal new lesions, this was not based on immunotherapy and may not be appropriate, Schwartz said.

He noted that several different approaches have been proposed for measuring response in immunotherapy trials. These include development of immune-related response criteria (irRC) based on World Health Organization criteria, and revised irRC that incorporate RECIST 1.1 recommendations (irRECIST).

Hazarika said that although the irRC were initially evaluated with ipilimumab in advanced melanoma patients, “the criteria has not been sufficiently evaluated for immunotherapies with other mechanisms of action or for other tumor types. Hence it is currently considered an exploratory endpoint for regulatory purposes.”

Although irRECIST was developed as a workaround for pseudo-progression and better reflects benefit with many immuno-oncology therapies, it has been adopted in different ways by different sponsors, Schwartz said.

New endpoints may require enhanced data collection.

“We have to come together on this and really have more than just a guideline but actually a criteria of very specific rules that we can apply to patients on these trials,” he said, citing the need for a uniform definition of irRECIST with specific rules of response and progression.

He also highlighted the need for enhanced data collection, with more robust and complete case report forms. Currently, the limited data collected are “not designed to capture progression phenotypes or kinetics. We don’t even capture true tumor burden,” Schwartz said. “We’re capturing a tiny subset of the actual lesions and how that correlates with the total tumor burden that the patient has is really not known, and limited studies that have been done show that there’s relatively little correlation.”

There also is a need for more complete evaluation of non-target and new lesions. “Once we generally see one reason for progression, unless we’re specifically asked to complete the form differently we may not report the other reasons for progression,” Schwartz said. “If we have target lesion progression, we may not look at all the other sites of disease.”

FDA’s Intermediate Response Endpoint

Speakers from FDA, academia and industry discussed a host of potential endpoints. FDA statistical reviewer Xin Gao described the agency’s exploration of an intermediate response endpoint that is based on radiology images but also incorporates information missing from RECIST-based response rate.

In developing a candidate intermediate endpoint, FDA used patient-level data from nine PD-1 inhibitor trials that had overall survival as the sole primary endpoint or a co-primary endpoint. The proposed intermediate endpoint is a composite with three parts:

• Target lesion response (based upon baseline tumor burden, post-baseline reduction depth and post-baseline tumor change dynamics);
• Non-target lesion response; and
• Occurrence of new lesions

The overall endpoint is binary, with patients deemed as having a response or no response.

In an individual patient-level analysis, the intermediate endpoint was associated with overall survival, and individuals who obtained a response had a longer survival outcome. In a trial-level analysis, there was a moderate association between the treatment effect on the intermediate endpoint and the treatment effect on overall survival, Gao said.

“Although this is not a perfect result yet, the result is better than” the correlations seen using traditional RESIST criteria, Gao said. FDA intends to explore the endpoint further, including refining the target lesion model.

Antonio Ribas, a melanoma researcher at the University of California Los Angeles, presented the case for using response durability as an endpoint.

“PFS and OS usually favor a small benefit in a lot of patients because that’s the easiest to shift the curve on improvement in hazard ratio, as opposed to having life-changing effect in some of the patients, which is what we’re trying to achieve” with immunotherapy, he said.

However, durable response rate does capture this benefit of immunotherapies. It is independent of the performance of the control arm and the effect of post-progression therapies, which can impact the overall survival picture, Ribas said.

Industry Proposals

Daniel Chen, vice president and global head of cancer immunotherapy development at Genentech Inc./Roche, discussed the company’s use of another endpoint, immune-modified RECIST (imRECIST), which he described as “our own translation of irRC onto a RESIST-type of approach.”

The imRECIST endpoint is pre-specified in all randomized trials of the PD-L1 inhibitor Tecentriq (atezolizumab), which recently added a lung cancer indication on top of its original accelerated approval in bladder cancer. (Also see "Genentech’s Tecentriq Stakes New Claim With FDA Lung Cancer Approval" - Scrip, 19 Oct, 2016.)

Genentech also has worked to describe an endpoint known as immune-modified PFS, which is based on imRECIST and may be a better predicter of overall survival, Chen said.

Tai-Tsang Chen, executive director in the Department of Global Biometric Sciences at Bristol-Myers Squibb, made the case for an alternative survival endpoint. Milestone survival would reflect the survival probability at a predefined time point, such as two years. This approach is agnostic to the uncertainty of delayed duration and provides a predictable timing of analysis, he said.

Axel Hoos, senior vice president and head of immuno-oncology at GlaxoSmithKline PLC, noted that although the community has taken steps beyond conventional chemotherapy efficacy criteria to describe the biology of immunotherapy and how to measure response, “there’s still more work to be done in terms of landing on the right system.”

Any new efficacy measure “has to be practical for routine clinical use.” – GSK’s Hoos

However, he said any new efficacy criteria “has to be practical for routine clinical use. If it isn’t, if it’s too esoteric, it’s very hard to implement.”

Genentech’s Chen said the intent of developing new criteria should not be to come up with something that is “three or four steps behind what’s already happening in the clinic. That is, we don’t need more ways to show that PD-1, PD-L1 inhibitors work. It’s helpful. But we already have traditional methods of showing those kinds of benefits.”

Rather, “I think the biggest need for us in the future again comes back to the fact that we have 800 combination cancer immunotherapy trials already in the clinic today, and what we don’t want to be is always three steps behind where we’re redeveloping new tools for something that’s already shown clinical benefit,” Chen said. “My hope is … that we get to a place where we can show the patterns now that are important, that we can use existing databases to link them to survival and … create something that’s practical enough and flexible enough that that can then be applied to the next generation of therapies that are quickly coming.”