Clinical Trial System Badly in Need of Overhaul, Say Panelists at Friends-Brookings Conference

Get Permission

Richard Pazdur, MD

Lalitha K. Shankar, MD, PhD

Geoffrey R. Oxnard, MD

Lawrence H. Schwartz, MD

Tatiana Prowell, MD

I wanted to bring an academic flavor to the agency. I was interested in developing drugs, in working with pharmaceutical companies—not just saying yes or no to something someone else invented. I wanted the FDA to be part of the development process.

—Richard Pazdur, MD
The purpose of RECIST was to create a reproducible metric that could be compared among studies. The method worked well, … but it has limitations because tumors are sometimes hard to access, they are irregularly shaped, and their boundaries are diffuse.

—Lalitha K. Shankar, MD, PhD
In a clinical trial, the higher the response rate, the greater the likelihood of approval, but RECIST doesn’t consider the depth of response, so progression may or may not have much clinical meaning.

—Geoffrey R. Oxnard, MD

Cancer clinical trials in three distinct phases, as they have been conducted for decades, are probably no longer the best way to bring a drug or biologic agent to market. This was the consensus of three panels at the 8th Annual Conference on Clinical Cancer Research convened by Friends of Cancer Research (FOCR) and the Center for Health Policy at the Brookings Institution in Washington, DC.

Improved understanding of the way cancers behave is eroding the long-established system of beliefs about treatment. Moreover, new tools and mechanisms of discovery have opened the door to new types of evidence and ways of thinking. Adaptive trials, Bayesian statistics, biomarkers, surrogate endpoints, data from postmarketing studies, and other evidence beyond randomized controlled trials are the future of oncology drug research.

Role of the FDA

Before the panels convened, Ellen V. Sigal, PhD, Chair and Founder, FOCR, interviewed Richard
Pazdur, MD
, Director, U.S. Food and Drug Administration (FDA) Office of Hematology and Oncology Products and one of Fortune’s 50 most influential leaders for 2015. Dr. Sigal asked Dr. Pazdur why he had come to the FDA in 1999 from The University of Texas MD Anderson Cancer Center.

“I wanted to bring an academic flavor to the agency,” he replied. “I was interested in developing drugs, in working with pharmaceutical companies—not just saying yes or no to something someone else invented. I wanted the FDA to be part of the development process.”

Dr. Pazdur has long expressed interest in toxicity and how it affects patients’ lives. He witnessed it firsthand with his wife, Mary, who died recently following several years of treatment for ovarian cancer. Now he sees it anew. “I underestimated it,” he said. “Toxicity is far more than a symptom or a lab test or an unpleasant reaction. It can be an all-consuming phenomenon that takes over a patient’s life.”

He also has a new appreciation for how difficult it is to obtain expanded access to as-yet-unapproved drugs. “I understand pharmaceutical companies’ position on protecting their proprietary interest, but we need more flexible criteria for who can get these drugs that can have a profound effect on remaining days.”

Dr. Pazdur emphasized that the FDA’s function is to ensure safety and efficacy. “We want patients to survive, of course we do. But survival—or other endpoints—is not necessarily our main goal. We are here to bring safe and effective drugs to patients who need them.”

What other issues does he think about for the agency? asked Dr. Sigal. “Looking at new ways to do clinical trials, developing informed consent instruments that people can understand, and listening more closely to patients: how they feel and what they want. It is patients who must drive the show.”

RECIST Becoming Outmoded

Changes in tumor size can and do indicate the activity of treatment but are based primarily on historical precedent and were developed as supportive tools in clinical trials. They may not be the best way to measure response.

Objective criteria for evaluating tumor change were developed by the World Health Organization (WHO) in the 1970s and consist of complete response, partial response, stable disease, and progressive disease. There have been refinements of measurement, but in general, thresholds are based on palpation.

In the early 1990s, WHO created the Response Evaluation Criteria in Solid Tumors (RECIST), which specified the minimum size of a lesion considered measurable, as well as the maximum number of lesions per organ. RECIST was updated in 2009, with a reduction in the number of lesions to be assessed, further clarification of disease progression, and additional imaging recommendations.

“RECIST is a morphologic assessment,” said Lalitha K. Shankar, MD, PhD, Chief, Diagnostic Imaging Branch, National Cancer Institute (NCI) Cancer Imaging Program. “Changes in tumor size can be slow or static and not reflective of biological status. Moreover, [RECIST] has limited utility in malignancies such as mesothelioma and neuroendocrine tumors.”

In addition, RECIST is not useful for immunotherapy or targeted agents, treatment modalities that have potential for long-term, sustained antitumor activity—but on a different time scale from cytotoxic drugs. Delayed responses, even transient progression prior to regression, are common; therefore, metrics that take these phenomena into account need to be developed for both clinical and regulatory decision-making.

“The purpose of RECIST was to create a reproducible metric that could be compared among studies. The method worked well, both for ongoing studies and for historical comparisons. But it has limitations because tumors are sometimes hard to access, they are irregularly shaped, and their boundaries are diffuse,” said Dr. Shankar.

Some limitations have been overcome by molecular or functional imaging and measurement of tumor metabolism. However, she said, “Tumors are progressively variable, and thresholds that define discrete categories of response are arbitrary. That is, there may be only a two or three percentage point difference in tumor size, but to RECIST that might mean the difference between stable and progressive disease, thus affecting future treatment. Moreover, response rates do not necessarily reflect depth or duration.”

Other reasons why RECIST may not be appropriate:

  • Some targeted therapies may stabilize disease with only minimal tumor shrinkage.
  • Antiangiogenics can improve survival with little tumor shrinkage.
  • Objective progression may not necessarily indicate treatment failure. For instance, some tyrosine kinase inhibitors (sunitinib in renal cell carcinoma and EGFR inhibitors in non–small cell lung cancer [NSCLC]) can provide patient benefit, even after RECIST-defined progression. Continued treatment is often appropriate, and tumor growth rate is a more accurate measure of drug activity than is tumor size.
  • Immune checkpoint inhibitors, such as those targeting programmed cell death protein 1 (PD-1) and cytotoxic T-lymphocyte–associated protein 4 (CTLA-4), have had unusual response and progression kinetics. Sometimes tumor progression appears before regression, and lesions appear enlarged because of immune cell infiltrates and inflammation at tumor sites.

Geoffrey R. Oxnard, MD, a thoracic oncologist at Dana-Farber Cancer Institute, talked about response vs regression. “In a clinical trial, the higher the response rate, the greater the likelihood of approval, but RECIST doesn’t consider the depth of response, so progression may or may not have much clinical meaning.”

Lawrence H. Schwartz, MD, James Picker Professor and Chairman of the Department of Radiology, Columbia University Medical Center, agreed. “Oncology drug development is inefficient. We need to use volumetrics to assess tumor burden, and RECIST doesn’t do that.”

Immune-Related Approaches

One possible remedy is immune-related response criteria (irRC), based on data from phase II trials of ipilimumab (Yervoy). It describes four potential patterns of response: immediate tumor regression with no new lesions, durable stable disease, tumor regression following an initial increase in tumor volume, and regression of the index lesion in the presence of new lesions. A variation, immune-RECIST, is more in line with the original RECIST guidelines.

Both irRC and immune-RECIST more accurately predict survival in melanoma, but neither has been widely used for new melanoma immunotherapies such as ipilimumab, nivolumab ­(Opdivo), and pembrolizumab (Keytruda).

Improved Tumor Growth Metrics: VolPACT

There had been no systematic way to test or validate alternate metrics until the Biomarkers Consortium of the Foundation for the National Institutes of Health created collaborations among industry, academia, and others. The result was Volumetric CT for Precision Analysis of Clinical Trial Results (VolPACT), which collects imaging data and associated metadata from completed large phase III trials in several solid tumors to study a variety of potential quantitative response metrics. Using existing data retrospectively precludes waiting for ongoing trial outcomes.

Computed tomography images are transferred to an academic laboratory and analyzed to determine unidimensional, bidimensional, and volumetric measurements for each lesion at various times. Phase II trials will be simulated to determine how well various metrics correlate with actual survival results. The goal is to identify metrics applicable across various therapeutic modalities and tumor types.

One stumbling block so far has been establishing a way that pharmaceutical companies can share (de-identified) patient data. In addition, images must be centrally stored, which also requires data-sharing.

Imaging for Immunotherapy

Because of its unique progression characteristics, immunotherapy presents a challenge for retrospective analysis, said Dr. Oxnard. “We need improved response and progression metrics. Documentation typically results in a change in clinical management, and imaging usually stops when the disease progresses. If a patient goes off study at RECIST-defined progression, but before true clinical progression, it may be difficult to use that patient’s data to develop alternative metrics.”

However, if imaging continues, data can be used to inform the design of ongoing and future trials: evidence of progression, such as presence and characteristics of new lesions; brain metastasis; lesions where radiology can’t find them; and pretreatment disease trajectory.

The challenge lies in diseases where progression-free survival is a standard regulatory endpoint because the benefit of immune checkpoint inhibitors may not be accurately captured, said Dr. Oxnard. But, he added, “Several agents have been approved, and we can learn from them how to facilitate future drug development.”

Patient Perspective on Adverse Events

The second panel discussed how patients experience adverse events, how rarely oncologists pay enough attention to them, what this means to their treatment, and how their experiences can be used in designing future trials.

Adverse events can have a profound effect on adherence to treatment. Evidence shows that severity of symptoms is not taken seriously by clinicians, many of whom don’t ask about symptoms in the first place, thus potentially skewing both clinical trials and treatment.

In clinical trials, adverse events are (or should be) reported to and assessed by providers using the Common Terminology Criteria for Adverse Events (CTCAE), which contains 800 of them (laboratory based, observable/measurable, and symptomatic), graded on a scale of 1 to 5 for severity and effect. As Alicyn Campbell, MPH, Global Head of Patient Outcomes Research for Oncology, Genentech, said, “We have to change the way we assess patients’ experience in order to evaluate how treatment will affect outcome.”

It has become obvious that many side effects (eg, mucositis, pain, fatigue) are best heard about directly from patients who know what they feel and how strongly they feel it—that is, patient-reported outcomes.

Patty Spears, Research Specialist, North Carolina State University, defined a patient-reported outcome as any report of a patient’s health that comes directly from a patient, without interpretation by a clinician or anyone else. Patient-reported outcomes are critically important because they provide a more complete and accurate picture of treatment effect than just what shows up on a laboratory slide or x-ray—and can help regulate dose.

Moreover, she said, “Once a drug is approved, patients and physicians are faced with treatment decisions, which are often complex. Patient-reported outcomes can measure highly subjective benefits and harms, evaluating them by factoring in what matters to patients.”

Patient-reported outcomes are increasingly being used in clinical trials, but by no means universally. However, this is changing. Patients are more involved in research and want to have a voice. They need information not only from clinical investigators, but from other patients on a trial, said Ms. Spears. Unfortunately, there is no standardization in collecting patient-reported outcomes, and the instruments used to collect data and ways to analyze them vary across the board.

In particular, few existing patient-reported outcome instruments meet FDA guidance; trials conducted at multinational sites lead to even more variety; trial designs often do not account for inclusion and interpretation of patient-reported outcomes; data contain missing or incomplete components; patients can feel burdened by being asked too many questions too often; and there is a perception that patient-reported outcomes are biased, although they rarely are.

Ms. Spears had suggestions about how to collect patient-reported outcomes data in clinical trials. Start in phase I, and then develop precision instruments for later trials. The most important element, she said, is that patients have the opportunity to say what matters to them. The instrument should be easy to complete, and the information gathered should be made available to patients. “Asking the right question at the right time in the right way is critical,” she said.

PRO-CTCAE May Be a Solution

NCI has developed a Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE) to measure adverse events related to treatment from the patient perspective.

Take mucositis, for example. The original CTCAE divided the score from 1 (asymptomatic or mild) to 5 (life-threatening consequences)—as clinicians saw it. PRO-CTCAE asks the patient to describe the severity of mouth or throat sores and how much they interfered with usual activities. This personalized the mucositis and related it to what the patient felt, not what a clinician saw.

Patients can easily understand PRO-CTCAE items, and they demonstrate better validity, reliability, and responsiveness than the original scale. For instance, it takes into account moderate toxicity over a long period of time rather than only high toxicity measured at predetermined times. The latter may be relatively easily treated, whereas the former may not.

PRO-CTCAE is important as treatment agents change; that is, the side effects of biologics are different from those of cytotoxics, and administration is no longer the same: a daily tablet for a long time rather than an intravenous infusion every few weeks for six or eight cycles.

Items in the PRO-CTCAE library of symptoms and adverse events can be used across treatment modalities and can serve as a standard method of patient self-reporting, as well as a way for clinical investigators to keep track of significant symptoms and toxicities. Other uses for the library follow:

  • In early-phase trials, it can provide initial information about toxicity and side effects, especially severe ones, to determine future development.
  • In expansion cohorts, it can elaborate on toxicity in broader populations, refine dosing, identify chronic toxicities that affect adherence, generate data to power clinical outcome assessments, and explore approaches to reducing severity.
  • It could characterize symptoms at baseline.
  • In postmarketing studies, particularly for accelerated approval, it can contribute to safety surveillance, comparative effectiveness research, and routine clinical practice.

Lines Between Phases No Longer Clear

The traditional stepwise approach to clinical trials is increasingly difficult to maintain, as illustrated by the development of pembrolizumab.

The first human trial of this drug was begun in 2011 to determine the dose in patients with advanced melanoma. The striking initial response prompted an increase in sample size. The potential for activity in NSCLC added a cohort to the study. Promising results meant expanding the sample size, and even when the number exceeded 1,200, more patients were added—almost unheard of for phase II. One cohort of 173 patients with unresectable or metastatic melanoma was randomly assigned to different doses of pembrolizumab.

The agent was granted accelerated approval in September 2014—just 3 years after clinical trials began—demonstrating the potential to make transformative therapies available years before they would be at the standard pace of development. Now there are about 40 such trials nominally called phase I. But they have far more patients than usual, and the data collected are more complex. Sometimes the ultimate goal is way beyond phase I—all the way to approval.

What has driven these changes? Tatiana Prowell, MD, Medical Officer and Breast Cancer Scientific Lead, FDA Office of Hematology and Oncology Products, said that scientific advances have resulted in more effective drugs—or rather, “the right drug for the right patient.” The focus is on disease pathway instead of tissue of origin, and the agency could be more efficient by avoiding delays inherent in discrete trial phases.

Expansion cohorts involve a variety of tumor types and molecular subsets. But the paperwork can get a little fuzzy. For some trials, the unusual sample size has not been justified, nor have descriptions of objectives for additional cohorts. Informed consent may not have been revised to reflect the nature of the trial, and there are concerns about patient safety.

Dr. Prowell noted that the FDA wants to provide guidance for expansion cohorts. “We now have more than three dozen commercial INDs [Investigational New Drugs] with active phase I trials enrolling more than 100 patients each. Many, however, have up to 1,200 patients and up to 14 expansion cohorts of 10 to 180 patients per cohort. This new paradigm can result in quick and nimble drug development, but we still have to ensure that appropriate patient protections are in place.”

Regulations must keep pace with the new science, so the FDA has several questions and issues under discussion:

  • Cancer drug development could conceivably occur with a single first-in-human protocol, which would create a need for more meetings between the agency and the sponsor and would necessitate oversight by various disease experts and divisions.
  • The size and quality of safety databases would come under scrutiny, as well as the adequacy of data required for approval.
  • Should these types of protocols be reserved for Breakthrough Therapy designation, or can they be applied to any drug?
  • What type and level of independent oversight is needed: scheduled pauses in a trial to review data observed thus far, improvement in transparency to reduce bias, or assurance of statistical rigor?

Blurring Phases, Polishing Trials Lens

Mostly because of expansion cohorts, but also because of fuzzy objectives and endpoints, it is no longer clear when phase I bleeds into phase II, or even phase III. Lines need to be drawn—perhaps not as sharply as previously, but new thresholds for phase I trials could include:

  • When a cohort exceeds a typical phase I size (about 20 patients). At that point, the investigator would write a formal statement of hypothesis and a plan to justify the expansion.
  • When randomization is introduced. Then the study becomes comparative and can serve as a basis for statistical inference.
  • When the sponsor intends to use the study for registration.

In addition, all such studies should have clear written objectives, endpoints, a statistical plan, and a statement of trial governance, including purpose, sample size, study design, minimal bar for futility or activity, early stopping rules, informed consent, institutional review board and regulatory authority notification of cohort expansion, and a patient follow-up plan. ■

Disclosure: The FOCR forum and corresponding full report were supported by a grant from the Burroughs Wellcome Fund. Drs. Sigal, Pazdur, Shankar, Oxnard, Schwartz, and Prowell, and Ms. Spears reported no potential conflicts of interest. Ms. Campbell is an employee of Genentech and a Member of the Roche Group. [The opinions and thoughts expressed by Ms. Campbell are her own and do not reflect nor represent those of F. Hoffman-La Roche AG, nor of Genentech, a Member of the Roche Group.]


Related Articles

Friends of Cancer Research Releases White Paper on Patient-Reported Outcomes

On November 17, Friends of Cancer Research (FOCR) released a white paper report, Enhancing Use of Patient-Centered Data in Regulatory Decision Making.1 The contents of that paper are summarized below.

Improving Patient Input

Many stakeholders agree that to ensure truly transformational therapies, ...