Remarks by Scott Gottlieb, M.D.
Commissioner of Food and Drugs
The National Comprehensive Cancer Network Policy Summit
(Remarks as prepared for delivery)
Good morning. It’s an honor to be here.
I’d like to talk about four topics today.
First, some of the new steps that FDA is taking to modernize the clinical trial process, and make sure that development programs are well-suited to evaluate the more targeted therapies that increasingly form the backbone of cancer care;
Second, how we’re harnessing real world data and advanced analytics to better support these regulatory approaches, throughout the full life-cycle of FDA approved products;
Third, how we’re taking new steps to enable payors and providers to leverage some of these same kinds of information and approaches to deliver more value to their patients.
And, finally, the role that payors can play in providing more open access to the significant amount of data that they have.
Real world data that’ll help patients and providers better assess benefit and risk; that’ll aid developers in bringing new therapies to market; and will inform FDA’s pre- and post-market efforts.
There’s a common theme that links these initiatives.
It’s the need to modernize drug development to harness the full medical potential of this rapidly advancing science, while ensuring that innovation remains affordable for patients.
If outdated regulations delay or derail the development of innovative, safe and effective products, patients suffer.
And if FDA-approved drugs are priced out of reach of patients, then the full benefits of innovation won’t be realized.
Both of those outcomes should be unacceptable to us, whether we’re policy makers or providers, product developers, or payors.
Cancer may be the field where the medical benefits and fiscal challenges of rapid cycle innovation are most apparent.
Cytotoxic chemotherapies were the standard of care for most metastatic cancers for nearly 50 years, from the late 1940s through the mid-1990s. Aside from drugs for testicular cancer, and some lymphomas and leukemia’s, these treatments had very modest efficacy. And they came with a lot of toxicity.
Patients often dreaded the drugs almost as much as they dreaded the disease they were intended to treat.
Over the last 20 years, our rapidly advancing understanding of tumor biology and how cancer cloaks itself from the immune system has enabled the development of highly targeted treatments for metastatic disease.
These new therapies can show significant efficacy in early stage trials, in patients who are carefully selected based on their tumor biology. And these therapies often come with a lot less toxicity compared to the prior standard of care.
In some cases, this allowed the FDA to make a positive review decision on clinical data from smaller trials. Some trials are as small as just a few dozen patients. And they’re based on results that accrue just a few years after first-in-human dosing.
These opportunities are based on a new paradigm, where the new drugs are so carefully targeted to the underlying biology of cancer, and the patients so carefully selected for their likelihood to benefit, that we’re seeing more compelling responses in much earlier stages of development.
These opportunities challenged us to consider our existing review paradigm. We had to ask ourselves some hard questions: What should we do, for example, guided both by ethics and science, when a new drug in a heavily pre-treated group of terminal cancer patients, is in a Phase 1 study where it shows a dramatic response in more than half of these patients?
Shouldn’t these kinds of findings change our approach to how we design the Phase II studies of this drug?
And how we randomize patients to such a treatment?
A recent letter in the New England Journal of Medicine noted that the overall response rate -- both complete and partial tumor responses -- in Phase 1 oncology trials was 20 percent.
This mirrors the FDA’s internal analysis, where we’ve found that Phase 1 trials that are based on an enrichment design – in other words, trials where the patients are selected for a treatment based on how their tumors express a specific biomarker – are associated with a significantly higher probability of demonstrating a clinical benefit than those that are not.
This can allow for different kinds of development programs.
One is the use of seamless drug development programs, where drug developers can plan to quickly expand enriched Phase 1 cohorts, and evaluate the drug in one continuous clinical trial.
This leapfrogs the traditional sequential three phases of clinical trials. It eliminates all of the time spent starting and stopping drug development in between each of those different stages.
The evolution in targeted therapies was the driving force behind the development of the Breakthrough Therapy Designation. This new, more efficient development program was codified into law by Congress in 2012 through the Food and Drug Administration Safety and Innovation Act (FDASIA).
Today, oncology is responsible for more Breakthrough Designations than any other therapeutic area. A full 50% of the total number of Breakthrough Designations that’ve been granted since 2013 are for products that treat cancer.
There are some critics that view these developments with skepticism. They argue that the FDA has somehow weakened its approval standards. I believe that this sort of criticism lacks historical and scientific context that undermines its validity.
Over the last decade, the FDA consulted with the Oncologic Drugs Advisory Committee, numerous external stakeholders, patient advocacy groups, and cancer researchers on appropriate endpoints in distinct tumor types. We received repeated advice to re‐evaluate overall survival endpoints.
There was a growing recognition that these traditional endpoints were based on a previous era of highly toxic chemotherapies. They were standards designed for an era that no longer exists. It was an era where the cytotoxic chemotherapies offered little advance in efficacy. And what little treatment advantage they offered, often came with serious side effects. So measuring survival was the key.
To justify the significant toxicity of these drugs, the ability to extend survival was considered the only appropriate tradeoff.
It’s not that the other endpoints weren’t clinically relevant -- endpoints like objective response rate, where you can show that you can substantially shrink the tumor burden for an extended period of time; or progression‐free survival, where you show that you can delay disease progression.
It’s just that the drugs were so toxic, that these benefits generally were not seen as sufficient to justify the risks.
This was especially true when we knew that it was only a small percentage of patients who were going to derive any benefit at all from that toxic chemotherapy.
That’s changed now that the drugs are less toxic, and the benefits that they deliver can be more certain to accrue to patients who are carefully selected based on their tumor types.
We needed to adjust our approach for the era we’re in today.
And that’s what we set out to accomplish.
After careful consideration of the changes in the kinds of drugs that are being developed, and the way that they’re being used, FDA agreed that these other endpoints like progression and objective response were acceptable for either accelerated or regular approval, depending on a variety of factors, including the tumor type and whether other therapies are available.
We didn’t evolve our approach overnight.
And we didn’t evolve on a whim.
We followed the science.
And we followed the growing refrain from the broader clinical community, and especially from patients, who themselves compelled us to chart a new course because of the way that this science is bringing with it new opportunities to extend life.
Let me take a moment to reflect on those opportunities that we’re seeing, and take measure of what’s been realized on behalf of patients. Enriched trials can more easily demonstrate a robust impact on intermediate clinical endpoints or surrogate endpoints that are likely to correlate with long-term clinical outcomes. Effects like tumor shrinkage or progression free survival can translate into the ability of patients to stay off more toxic regimens for a longer stretch of time, therefore improving their quality of life. These observed effects may challenge our ability to achieve clinical trial equipoise, which is the doubt about whether new therapy may be more effective than the existing standard of care. This is especially true when there are few or no effective treatment alternatives.
Under these circumstances, where we know a drug can produce a robust response in a tumor that’s otherwise hard to treat, we may not be able to ethically randomize patients.
Pivotal trials testing these products are also likely to have crossover designs, meaning that if a patient’s cancer grows on the comparator treatment, then the patient is switched over to the experimental arm. Both arms receive the experimental drug, although in different sequence. Statistically, this may make it difficult to show an overall survival advantage for the experimental arm. Patients aren’t left on the comparator arm for long enough to show that they would have died sooner if left off the new drug. There’s a simple bottom line here. When new drug shows a robust effect on things like tumor size or tumor progression, patients are understandably less willing to forgo a new therapy in the name of a good “p” value.
Take for instance imatinib, for the treatment of CML. It was approved based on cytogenic response rate, with subsequent trials showing both efficacy and safety advantages compared to chemo. But it took more than a decade to prove an overall survival advantage, because every patient eventually crossed over to the new drug. A lot of patients gained a lot of extra years of life as we waited a little longer to show p equaled .05
Later trials randomizing patients with EGFR‐mutant or ALK rearranged advanced non-small cell lung cancer to either targeted therapy or chemotherapy, with high cross‐over to the experimental arm, also confounded interpretation of overall survival. That doesn’t mean these drugs don’t work.
Standard of care for a number cancers now includes FDA products approved using earlier endpoints like progression free survival, cytogenic response, or pathologic complete response.
This includes merkel cell carcinoma, medullary thyroid cancer, gastrointestinal stromal tumor, metastatic basal cell carcinoma, pancreatic neuroendocrine tumor, multiple myeloma, CML, chronic lymphocytic leukemia, and EGFR‐ and ALK‐driven non-small-cell lung cancer, to name just a few deadly diseases.
In some cases, access to additional therapies that were approved based on early endpoints in targeted drug classes mean patients whose tumors develop resistance to first line therapies can switch into second and third line treatments.
For instance, patients with metastatic lung cancer with ALK or EGFR mutations whose tumors progress on first line treatments go on to second and third line therapies. In some cases, these patients have remained stable for years on targeted therapies, or checkpoint inhibitors approved based on these endpoints.
I know not every patient benefits from these therapies.
And not every result is durable.
But the overall trends are unmistakable.
Recently released SEER data shows that between 2011 and 2015, overall cancer death rates decreased 1.8 percent per year for men, and 1.4 percent for women.
Eleven of the 18 most common cancers in men showed decreases in mortality. And fourteen of the 20 most common cancers in women showed decreases in mortality.
These included leukemia, melanoma, myeloma, non-Hodgkin lymphoma, and cancers of the colon and rectum, breast, cervix, ovaries, esophagus, kidney, larynx, lung and bronchus, prostate, and stomach. Lung and bronchial cancer – among the most feared and rapidly fatal cancers - had the greatest decrease in mortality in men; and non-Hodgkin lymphoma had the greatest decrease in women. Early detection was one key for many of these diseases. But so were advances in care.
FDA needed the ability to work more closely with the scientific community and sponsors to ensure that trial designs and endpoints were appropriate for the type of therapy and the disease state under consideration. Congress signaled the importance of this with the Breakthrough designation. But it didn’t change any of FDA’s standards for drug approval.
That designation just codifies the FDA’s early and intensive engagement with sponsors for treatments that demonstrate significant clinical benefits over the current standard of care.
The FDA has also used the Accelerated Approval pathway for 25 years to speed the approval of therapies that demonstrate a meaningful advantage over available therapies for serious or life-threatening diseases. Of the 93 accelerated approvals in oncology, a majority have verified clinical benefit.
Of the 51 oncology accelerated approvals that have post-marketing requirements and verified benefit, the median time from accelerated approval to verification was 3.4 years.
Most the remaining accelerated approvals were approved in the past three years. So, confirmatory trials are still underway.
Only five cancer drugs approved through accelerated approval -- just five percent of the total -- have been withdrawn from the market for failure to confirm their clinical benefit.
Some critics of these pathways have pointed out that making approvals based on surrogate endpoints may introduce uncertainty about the final clinical benefit.
That’s, in fact, the whole point.
In some cases, for instance in rare cancers, there are simply not enough patients to power a trial to measure overall survival.
In other cases, where the disease can take several years to run its course after multiple rounds of standard treatments – like leukemia’s or lymphomas - trials could take years, or even decades, to show a statistically valid benefit on overall survival.
The treatments are getting so good patients are living too long to measure survival advantages in a reasonably sized trial.
In these cases, other endpoints like minimal residual disease or complete response, can give us the practical knowledge we need to predict survival, and are “fit for purpose” based on how drugs are designed to disrupt disease promoting pathways.
If we were only approving drugs for serious and life threatening diseases based on an overall survival benefit, and then denying patients the ability to crossover to the experimental arms in confirmatory trials, we would literally be asking patients to die so that we could achieve a lower “p” value.
So we’re trying to balance competing needs. The need for high degrees of statistical certainty, and the need for access to care.
Accelerated Approval balances disease risk with the likelihood of benefit. It accounts for uncertainty, and allows for regulatory flexibility to make safe and effective drugs available to patients sooner than permitted through the regular approval pathway.
All of this said, we know we owe it to patients and providers to be even more transparent about how we’re making these decisions, and how the science is changing our regulation.
And so, in the coming weeks, for the first time, CDER will publish on the web, a list of the surrogate endpoints that were the primary basis of approval or licensure of a drug or biological for both accelerated and traditional approvals.
We believe this list will help to address some of the questions that’ve been raised about how we apply surrogates to make a given regulatory decision. We’ll also be taking steps to provide more guidance to drug developers on these novel approaches.
This includes how to design trials based on a biomarker as a novel surrogate endpoint. As one part of these efforts, we’ll be accepting requests for meetings with sponsors -- called Type C meetings -- earlier in the drug development process.
The aim is to give sponsors explicit advice on how to design rigorous development programs under these approaches.
Innovators can use these meetings to discuss the use of a biomarker as a surrogate endpoint in clinical contexts where that marker was never used as the primary basis for approval.
These approaches require careful consideration and thoughtful approaches to clinical trial design and analysis.
Both of these efforts will help give the research community and innovators greater clarity about how the FDA utilizes endpoints, and how we expect product developers to carefully evaluate biomarkers as surrogate endpoints, in a given disease state.
But I said at the outset that we need to do more to harness real world data and real world evidence and advanced analytics to support these changes in the clinical trial ecosystem. We need to make sure that we’re informing our regulatory decision-making throughout the life cycle of FDA approved products.
Leveraging reliable post market data, including real world evidence, is a key to these new regulatory approaches.
New sources of real world evidence can help us understand the efficacy of new products in the real world clinical setting, and allow patients to make more informed treatment decisions.
Ultimately, nuanced safety and efficacy information from a variety of both pre- and post-market settings can help accelerate innovation and competition among a wider range of safe and effective oncology treatments.
In our 2019 budget, we’ve requested $23 million to establish a new medical data enterprise for real world data and advanced analytics, giving the FDA the ability to conduct near-real-time evidence evaluation down to the level of individual electronic health records for at least ten million individuals.
This data would come from a broad range of health care settings in the U.S. for a wide range of medical products.
The health care settings would be carefully selected to cover data gaps in current FDA systems. ln addition to accessing more detailed clinical data, we’d also develop and apply modern computational techniques, such as natural language processing and machine learning, to more efficiently cull real time trends in the safety and effectiveness of marketed medical products.
We’re already pursuing these sorts of efforts.
The FDA is collaborating with the National Cancer Institute to develop digital biomarkers for measuring cancer pain, quality of life, functional status, and cognitive function using biometric sensors, computer vision and voice recognition technologies.
Digital biomarkers and algorithms can replace highly variable surveys or “snapshot” clinician assessments done during clinical trials. These approaches often don’t capture the day to day burden of cancer and the treatment-related toxicities.
We’re also testing new diagnostic imaging artificial intelligence frameworks to support more objective and less volatile assessments of tumor response to treatment.
A recent meta-analysis conducted by FDA data scientists suggested a discordance of about 30% in classification of tumor dynamics based on RECIST criteria between the primary investigator and the independent radiology review assessment.
We may be able to use annotated radiology images to train artificial intelligence algorithms to deliver a more objective, holistic estimate of tumor response, by scanning very large numbers of radiology images. It could be based on more images than any single radiologist would be able to see in a lifetime. Using these approaches, we could generate a standardized measure of tumor dynamics based on a patient’s entire tumor burden.
If successful, investigators would have a much more nuanced measurement of patient response to treatment over time.
And the FDA would gain a more objective tool for following the natural history of cancer’s spread and the clinical response to different treatment strategies.
These approaches are at hand.
The technology exists to stand them up.
And the prerogative is there to seize them.
We need to make sure our approach to developing drugs is as sophisticated as the medicines that are being evaluated.
And we need to make sure that the data tools exist to fully support the kinds of clinical trial approaches that we’re embracing. That means building out more capacity to use real world data, especially in the post market setting, to continue to evaluate the long-term safety and benefits of new treatments.
Tools like these could also help the agency develop “virtual control arms” based on standard of care in different tumor types, potentially decreasing the need to expose patients to toxic treatments in control arms when the disease history is already well known – like platinum doublet arms in small cell lung cancer. It could also reduce the cost and time required to run oncology trials, while still generating high quality data.
But we know it’s not enough to develop this new science, and more efficiently advance the resulting treatments to patients.
We also must make sure we are taking steps to make these breakthrough drugs accessible to all patients who need them.
And so I’d like to spend my last few minutes talking about how we can collectively advance more innovative, value-based contracting that promotes competition and encourages payors and industry to focus on delivering the best outcomes for patients with serious and life-threatening ailments, as efficiently and as safely as possible.
In one of two recently released FDA guidance documents, we recognized that payors seek a range of information on the effectiveness, safety and cost-effectiveness of approved or cleared medical products. This includes information from drug makers to help support product selection, formulary management, and coverage and reimbursement decisions.
This information doesn’t always align with the information that the FDA reviews in making regulatory decisions, or that’s included on FDA approved product labels.
For example, a payor might want to contract for a new treatment based on its ability to decrease the number of days a patient is admitted to the hospital. This might be an objective measure of benefit that the payor can easily track to measure whether a drug is benefiting its members. But shortening hospital stay might not be an outcome for which the drug was approved. Yet it’s exactly the kind of truthful, non-misleading, and appropriate communications that companies and payors might need to exchange to develop value-based contracts.
Our aim is to help facilitate contracting for new medical products that are based on the value that these products are delivering to health systems, providers, and especially patients.
In the second recently released guidance document, we provide our views on manufacturers’ communications that are not contained in FDA-required labeling, but that are consistent with that labeling. Together, these new guidance documents are designed to clarify the FDA’s policy for these sorts of communications, and to make sure the FDA regulations on communications don’t get in the way of value-based contract negotiations. The FDA is doing our part to ensure that our regulations aren’t a barrier to these types of contracts.
But we can’t do it alone.
We also need help from industry, especially the payors, who sit on mountains of data that can inform the sort of real world evidence that we’re increasingly dependent on as part of our post-market work. So, I’d hope payors would consider doing more to share this data, through a public-private partnership perhaps. To take one example, we have a real interest in seeing payors share more post-market data on long-term patient outcomes in oncology patients, and especially adverse effects associated with targeted drugs, including checkpoint inhibitors.
Right now, a lot of this data remains in silos controlled by the payors. It’s sold at high prices. This data can help inform our view on the long-term outcomes that some in the payor community say FDA isn’t doing enough to demand from drug makers. Yet increasingly, for all the reasons I outlined at the beginning of my remarks, we’re going to be dependent on real world data to answer some of these questions. And so, I hope we can find new ways to partner with the payor community, and share some of their data, to answer these questions.
It’s not enough to point fingers. We all need to work together. That includes our desire to work with payers to help answer these questions in real world settings that matter to patients.
As I mentioned earlier, the severity of these diseases -- and the pace of innovation -- demands inventive trial designs and endpoints so patients can benefit from early access to products demonstrating exceptional efficacy in early stage trials.
But efficacy seen in these trials may not represent outcomes seen in real world patients. And patients who achieve stable disease may be on these treatments for years, or even decades, requiring long-term follow-up to measure long term outcomes.
For instance, a recent article in Lancet Haematology noted that “several hematological malignancies are now chronic diseases that are treated with continuously administered therapies that have unique side-effects over time.”
The study found that “population based data [shows] that five-year survival for patients with chronic myelogenous and chronic lymphocytic leukemia, indolent B-cell lymphomas, and multiple myeloma has improved markedly.”
Because of these sorts of realities, we need to dramatically improve the speed with which we use learning from clinical trials to inform clinical practice, and use clinical practice to inform future drug development programs.
The speed of advance, particularly in immunotherapies, across a wide range of tumor types and combination treatment approaches, “each with a different profile of adverse events” has created a need for reliable databases for optimizing patient outcomes and minimizing safety risks. These are systems that no single provider, manufacturer, or insurer can create alone.
More public-private collaboration to share outcomes data from electronic health records and other real world sources, such as claims data for tracking adverse events, could provide all stakeholders with information needed to advance the standard of care. It could allow payers to better relate prices to value, including from reduced hospitalizations, physician visits, or other impacts on healthcare costs and patients’ quality of life.
Right now, FDA pays enormous sums of money to payers for access to just a fraction of this kind of post-market information.
As development costs and timelines fall, we should expect more products, even in oncology, to compete based on price and value, and without reducing market incentives to innovate.
Better post-market assessment is a key to this competition.
And so, I’d ask the broader community, don’t just criticize.
Put your data where your arguments are. Let’s look for new ways to collaborate around these post-market efforts.
Beginning with the HIV/AIDS crisis, and continuing today, highly informed and active patients and caregivers have made it very clear that they’re often willing to endure more uncertainty about the benefits of new treatment options when the alternative is an inevitably fatal or crippling medical condition.
The FDA’s role is to evaluate those benefits and risks in dialogue with the patient and providers, and ensure that the evidence generated from “adequate and well controlled trials” demonstrates safety and “substantial evidence” of efficacy.
For most indications, and for most applications that reach the agency, that “gold standard” still requires two randomized controlled trials, preferably placebo controlled; or, when that’s neither ethical or feasible, a trial compared to the standard of care or another appropriate comparator.
But statistical designs like that are most appropriate when we have the greatest uncertainty about the relative safety or efficacy of an experimental treatment compared to a comparator. Or when we’re dealing with a condition where the natural history of the disease is itself highly variable.
Highly controlled trial designs reduce other sources of uncertainty. This gives us greater certainty that any change in outcome across the treatment and control populations is due to the intervention and not a result of chance.
That’s the benefit of these approaches.
But we know that these approaches come with a cost.
They can exclude certain minority populations, patients with poor performance status, or rural populations that can’t access, or afford to access, an academic medical center hosting a trial.
In other cases, these traditional designs may not be appropriate or ethical. Or they can take many, many years to complete.
And other approaches can give us a high level of confidence that’s appropriate to the disease type or product under consideration, especially when evidence from non-randomized trials is supplemented by biology, historical controls, or pragmatic trials conducted in the post-market.
No endpoint or trial design is perfect. And science itself is best understood as the gradual reduction of uncertainty.
We need to have humility about the best tool to answer a clinical question. This is the best way to have a rational discussion about the strengths and weaknesses of various types of evidence; and how to balance the tradeoffs of various pre-market and post-market studies, and the needs of patients.
Our work demands that we continue to reflect on how we can make the science of drug development more modern and more patient-centered, so that approved products impact the features of disease that patients and families value most.
Done well, medical product development will continue to benefit from a better understanding of the patient experience, as FDA reviewers gain the critical context of a disease, including how patients live, and at times die, with these conditions.