A recent op-ed in the New York Times by journalist and cancer advocate Clifton Leaf asked the question, do clinical trials work? The answer is a straightforward yes. A clinical trial is a tool that when correctly employed can identify more-effective treatments for patients with cancer and other diseases.
It is important to acknowledge that some clinical trials have not been as informative as we would like. Some have shown “improvements” that might not be considered groundbreaking. But others have truly changed the standard of care for a given disease in a single day.
The questions today are, how do we increase the number of informative trials that can change practice and save lives? How do we evaluate promising therapies in the shortest possible time? And when do we stop and move on when the results are not as positive as expected? A well-designed clinical trial can accomplish these goals.
My colleagues and I are focusing on streamlining our clinical research program, which will allow more patients to participate in more trials, and also bring novel compounds and therapies to those who can benefit from them in a timely manner. At Memorial Sloan Kettering, patients and researchers are typically involved in more than 900 total clinical trials at any given time.
The Traditional Paradigm
The standard approaches for developing most cancer treatments have been built on a rigid sequence of clinical trials. Historically these studies advanced the field slowly over time with incremental improvements. Each step reset the benchmark to which subsequent treatments were compared. This approach served us for many years as standard chemotherapy agents with broad applicability across many patients and diseases were being developed.
The development of agents such as cisplatin, carboplatin, and paclitaxel all occurred in this way and these agents still form the backbone of many cancer treatments.
The traditional paradigm of evaluating agents required them to move through phase I, II, and III trials, often over a protracted period of time. The phase I trial was intended to establish a dose — the “maximum tolerated dose” — and to establish safety. These phase I trials were generally small, and an assessment of the effectiveness of the agent was usually not the goal.
The next step involved a phase II trial classically with 35 or so patients enrolled to see if a certain group of patients would have tumor shrinkage and for what period of time. We often accepted a rather low number of patients showing improvement as evidence of activity.
If results of the phase II trial were positive, the investigation would move to randomized phase III trials with large numbers of patients, known in clinical trials as cohorts. Large numbers are required to detect a difference between a new and old treatment, particularly if the differences are expected to be small. Phase III trials often are done at many medical centers around the nation, and sometimes around the world, making orchestration complex, and the time required to get results can be long.
Rethinking the Large Trial
Just as the computing tools we used five years ago would seem completely inadequate to support us now, the rigid clinical trial paradigm described above has also become outmoded in many situations. Clinical trials remain the best way to improve cancer treatments, but what we need to rethink and avoid is the large clinical trial with long follow-up looking for very small improvements. (Several examples were discussed in the op-ed by Mr. Leaf.)
In particular, this approach will not serve us well in the era of mechanism-based or targeted therapies in which the therapy being tested may only work in a small group of patients but may be highly effective in that group.
The idea of aligning patients who carry a specific therapeutic target with a specific drug has resulted in higher response rates than have been previously seen, and now combination therapy is being considered to overcome the resistance that tumors often develop to anticancer drugs. New cancer drugs are initially tested in phase I trials, but we have become more nimble in confirming these responses early on rather than moving down the traditional paradigm.
The Basket Trial
The development of the “basket” trial is one example. Instead of starting with multiple clinical trials in different diseases (which requires duplication of regulatory and infrastructure efforts), we start with one trial — the basket — and one or more targets, and allow patients with multiple diseases to enroll in cohorts or groups.
If one group shows good response, we expand this group to immediately assess whether others could benefit from the new therapy. If another group is unfortunately not showing evidence of effectiveness, this group may be closed and the patients can move on to consider other therapy. In this way, the exploration of the effectiveness of a treatment occurs early, quickly, and in one trial.
The optimal phase I trial today often explores drugs with innovative mechanisms. When a robust response rate is seen and the group is expanded sufficiently (called an expansion cohort), this can sometimes bring enough confidence of effectiveness in the phase I trial that smaller randomized trials can be immediately done to confirm the findings.
For example, a recent Memorial Sloan Kettering study of nivolumab and ipilimumab in patients with advanced melanoma, conducted by medical oncologist Jedd Wolchok and colleagues, was expanded so that 53 patients received the combination therapy in the initial trial. In this study, 53 percent of patients had significant reduction in tumor size by 80 percent or more. This trial rapidly confirmed the activity of this combination without having to go to a traditional phase II trial.
Smaller, Smarter Trials
There are two ways to increase the likelihood that a new treatment shows benefit in contemporary clinical trials. The first is to align the right patient with the right target in trials of therapeutic agents that target a specific mutation or pathway. The second is to identify biomarkers for response — for example, an abnormality measurable in the blood or detectable in the tumor specimen that when present can predict response to a particular agent — as a critical part of new drug development. In this way, we can give a new treatment to those people who are most likely to benefit.
We also need to continue to develop new statistical designs. One of these is called the Bayesian approach, where treatment arms in a given trial are frequently reviewed. The ones that are better performing will enroll more patients, and those doing less well get closed as soon as possible.
We are essentially moving toward smaller and smarter trials looking for clearly meaningful improvements. This allows us to evaluate strategies faster and, most importantly, increase the chance of benefit for individual patients.
When a clinical trial is well designed — whether the results are positive or negative — we learn important next steps. In addition, if we can match the right patient to the right trial, the number of successful approaches will continue to rise.
The clinical trial remains our best tool to identify new therapies, but as with all tools, innovation is required if trials are to remain relevant. We have more novel agents and approaches to consider than ever before, and well-designed clinical trials remain the best way forward.