Developing Intermediate Endpoints in Immunotherapy

Get Permission

James P. Allison, PhD

Axel Hoos, MD, PhD

The immune system holds tremendous potential for long-term sustained antitumor activity.

—James P. Allison, PhD
We need to develop immune-related response criteria that incorporate both early and delayed response, as well as tumor volume that increases before it recedes.

—Axel Hoos, MD, PhD

“The immune system holds tremendous potential for long-term sustained antitumor activity,” said James P. Allison, PhD, Immunology Chair, The University of Texas MD Anderson Cancer Center, Houston, as he opened a panel discussion at a meeting cosponsored by the Friends of Cancer Research and the Brookings Institution, in Washington, DC.1

He added that the approach may be amenable to treating resistant tumors, and it is probably less toxic than conventional treatment. Moreover, there is transformational clinical potential. That said, “some patients do well on immunotherapy—durable responses for 10 years or more—and some don’t,” he added.

Existing Agents and Challenges

Immunotherapeutic agents work successfully to varying degrees. But there are challenges. They require a variety of developmental approaches and strategies to produce antitumor responses because immunotherapy is not all the same.

  • Cytokines interact with immune cells to modulate a particular type of response. Interleukin-2 (IL-2 [Proleukin]) is the best-known example of these agents and was approved for the treatment of metastatic renal cell carcinoma in 1992.
  • Therapeutic vaccines induce or amplify a host immune response to a specific tumor antigen.
  • Cellular therapies include active transfer of live immune cells (tumor-infiltrating lymphocytes or peripheral blood T cells) engineered to recognize tumor-specific targets—for instance, sipuleucel-T (Provenge), approved in 2010 for the treatment of metastatic castration-resistant prostate cancer.
  • Immune-modulating antibodies enhance natural immune responses or overcome tumor-induced immune tolerance.

“Some agents bind to and inhibit negative regulators of T-cell activity that normally serve as “checkpoints” to maintain self-tolerance, while others stimulate positive T-cell co-receptors,” said Ramy Ibrahim, MD, Senior Medical Director, Oncology Clinical Development, MedImmune, Washington, DC.

He added that because some tumors co-opt these checkpoints to induce immune tolerance to their own cells, “checkpoint blockade” enables an antitumor response. The first of these agents to be tested was ipilimumab (Yervoy), which inhibits cytotoxic T-lymphocyte antigen-4 (CTLA-4) and was approved for treatment of metastatic melanoma in 2011.

A similar class of checkpoint inhibitors that inhibit the T-cell co-receptor programmed cell death protein 1 (PD-1) or its ligand (PD-L1) has raised significant interest.

Using the immune system to fight cancer has been of interest for a long time, said Dr. Ibrahim, “but we do not sufficiently understand its response to cancer, and we lack a methodology of research that recognizes the unique features of immunotherapy.” There are many challenges in such research.

For example, conventional pharmacokinetic approaches cannot guide dosing decisions because pharmacodynamic effects may be prolonged. Dose escalation until toxicity does not work. “We need new phase I designs for dosing,” said Dr. Ibrahim.

Dose-ranging comparative studies may not better inform dose selection. Sometimes, if there is an early dramatic signal of activity, a trial might progress too quickly from phase I to phase III. Furthermore, animal models may not adequately mimic the human immune response to cancer, and there are too few immune-based biomarkers that correlate with an effective response.

Because the immune system is dynamic and multifactorial, a successful approach must facilitate recognition of tumor cells by immune effector cells, as well as overcome the immunosuppressive tumor microenvironment. In other words, the target is the immune system itself and not the cancer.

Finally, different immunotherapies require different criteria to assess response rate, progression-free survival, and other parameters.

Surrogate Endpoints

Clinical trials have long relied on intermediate or surrogate endpoints to predict long-term survival. But immunotherapy doesn’t “try out” the same way as cytotoxic drugs. “Unconventional responses [for example, tumor shrinkage sooner than expected] to immunotherapy are relatively frequent and shouldn’t be ignored,” said Axel Hoos, MD, PhD, Vice President of Oncology, GlaxoSmithKline, Philadelphia. “We need to develop immune-related response criteria that incorporate both early and delayed response, as well as tumor volume that increases before it recedes.”

Drug companies use surrogate endpoints to decide whether they will continue testing a particular drug. Perhaps even more important, without an accurate surrogate, a trial that might have ended successfully could be prematurely terminated. This is what happened with a trial of tremelimumab, an anti-CTLA-4 antibody that was discontinued in 2008.

Dr. Hoos said that immune-related responses must be incorporated into surrogate endpoints, such as:

  • Calculating the overall clinical benefit rate (for example, response and stable disease for 6 months or more).
  • Taking a baseline measurement several months after the initiation of therapy rather than at the beginning. This allows for the time it takes for a tumor to respond.
  • “Milestone survival”: the proportion of trial participants alive at a designated point that is late enough to surpass late-onset immune response but early enough for accelerated approval.

Tai-Tsang Chen, PhD, Lead Statistician, Bristol-Myers Squibb, New York, described immune checkpoint modulators as offering the possibility of long-term survival and delayed clinical effect.

Log-rank analysis is not sufficient as the sole characterization of overall survival because it does not capture long-term survival effect, he said. The time to final analysis may be lengthened based on the kinetics of survival effect. But milestone survival (for example, 2 years) can be used as an intermediate endpoint to quantify long-term survival in the assessment of survival benefit.

There are advantages and disadvantages, said Dr. Chen. The former include earlier assessment of benefit and risk, greater statistical power when the delayed clinical effect is present, predictable timing of analysis, and the ability to use overall survival as both intermediate and final endpoints.

Disadvantages include not maintaining study integrity after the milestone analysis (unblinding prior to final overall survival analysis), as seen in all interim analyses not accounting for the totality of overall survival data, and the danger of making false-positive or false-negative errors in a registration trial when prior data do not lend themselves to appropriate milestone selection.

Other Endpoints

Other surrogate endpoints take into account unusual response patterns and delayed treatment effect. They can be used to characterize the activity of checkpoint modulators or to accelerate their development.

Clinical benefit rate is the rate of partial responses plus complete responses plus durable stable disease in patients whose disease had been progressing. Because stability sometimes reflects the natural course of a disease, a clinical trial should require evidence of measurable progressive disease before patients begin the study drug.

Gated progression-free survival is a modified approach to traditional progression-free survival and is assessed 2 or 3 months after treatment begins rather than at the time of treatment initiation. This minimizes the effect of early and/or pseudo progressions, thus allowing time for a full immunologic response. However, it requires a protocol in which patients are kept on the investigational treatment past initial radiographic progression. This may be appropriate only for patients with refractory disease and for whom there are no other treatment options.

Tumor growth rate constants, which are now exploratory, can be used as a potential measure of efficacy. Analysis of changes in tumor growth rate may translate into survival improvement. Growth rates can be calculated from the tumor burden data usually collected for progression-free survival assessment. They are particularly relevant for treatments that induce low levels of response, as well as for describing effects beyond response.

The Future

Much of the data demonstrating delayed treatment effects and unusually long-lasting responses come from a single agent: ipilimumab. Trials with therapeutic vaccines and early clinical results from PD-1 and PD-L1 inhibitors show similar effects, suggesting that the phenomenon might occur in 5% to 10% of patients. Such positive intermediate results could reduce the chance of prematurely ending a trial and increase the likelihood of accurately characterizing the activity profile of new immunotherapies. It also can reduce the time to marketing by serving as a basis for accelerated approval.

A major question for the future concerns randomized trials. Are they at all ethical when early immunotherapy trials have shown significant and reproducible evidence of efficacy? Randomized trials with survival endpoints may not be feasible when multiple drugs in the same class are developed simultaneously by several companies (eg, PD-1/PD-L1).

Single-arm registration trials might be appropriate for adoptive cell transfer approaches or therapies with high response rates. It is unclear, though, what endpoints would be appropriate if a therapy induced a long-term benefit but a low response rate. Validated endpoints that reflect mechanism of action are needed. ■

Disclosure: Dr. Allison is an inventor of intellectual property owned by the University of California, Berkeley, and licensed to Bristol-Myers Squibb; he is also a recipient of royalties from Bristol-Myers Squibb and founder of Jounce Therapeutics. Dr. Ibrahim works for and owns stock in AstraZeneca. Dr. Hoos is Vice President, Oncology, for GlaxoSmithKline. Dr. Chen reported no potential conflicts of interest.


1. 2013 Friends-Brookings Conference on Clinical Cancer Research. Session One: Facilitating Development of Immunotherapies: Intermediate Endpoints for Checkpoint Blockade. Presented November 7, 2013.