4

Developing Endpoints for Gene Therapy Clinical Trials

The third workshop panel explored the successes with and challenges to accurately measuring clinical endpoints and outcomes for gene-based therapies and moving products through the translational pathway. Larissa Lapteva, the associate director in the Division of Clinical Evaluation, Pharmacology, and Toxicology, Office of Tissues and Advanced Therapies, Center for Biologics Evaluation and Research at FDA, moderated the session. She noted in her introductory remarks that for any clinical development program with a novel therapeutic product, the choice of the primary endpoint for a clinical trial intended to demonstrate substantial evidence of that product or that agent’s effectiveness can be the most vulnerable part of the entire development program. This can either unite all of the elements of that development or make the product nonviable, she said. Lapteva also described some of the important concepts regarding endpoints and outcomes in clinical trials. Dwight Koeberl, a professor of pediatrics and molecular genetics and microbiology in the Department of Pediatrics at Duke University and the medical director of the Duke University Health System Biochemical Genetics Laboratory, discussed endpoints for clinical trials in Pompe disease; Albert Maguire, a professor of ophthalmology at the Hospital of the University of Pennsylvania and the Presbyterian Medical Center of Philadelphia, discussed clinical endpoints for a Phase 3 inherited retinal dystrophy gene therapy trial; and Julie Kanter, an associate professor of hematology and oncology at the University of Alabama at Birmingham School of Medicine, discussed her work determining optimal endpoints for gene therapy trials in sickle cell disease.

DEVELOPING ENDPOINTS FOR CLINICAL TRIALS

The concept of “substantial evidence of effectiveness” for human drugs and biological products has been defined and codified in the Code of Federal Regulations and is described and discussed in many FDA guidance documents, Lapteva said. Briefly, she explained that FDA requires two adequate and well-controlled clinical trials for most diseases, though in some cases—for example, in rare diseases where a second trial might not be feasible or ethical—FDA will accept one adequate and well-controlled trial with supportive confirmatory evidence.

In the traditional regulatory approval pathway, the endpoints used in the trials that are intended to demonstrate the product’s evidence of effectiveness would be clinical endpoints that directly measure clinical benefit or surrogate endpoints that have been validated to predict clinical benefit.¹ A second pathway, accelerated approval, has been around since the 1990s and is typically reserved for serious and, often, rare diseases for which there are no available treatments. Accelerated approval has been used in cases where the disease course may be prolonged and an extended period would be needed to observe clinical benefit.² In order to make development feasible and also improve access to care for those who need it, accelerated approval allows for the use of surrogate endpoints that predict clinical benefit with reasonable likelihood, Lapteva said.

Clinical outcomes, when used as endpoints, directly measure clinical benefit, which FDA views as how a patient feels, functions, or survives, Lapteva said. Surrogate endpoints (which may be laboratory parameters) can be measured earlier than clinical outcomes, but their use in trials should be supported by an ability to predict clinical benefit. For years the term “surrogate endpoint” was largely misunderstood and used interchangeably with other similar terms, confusing its meaning, Lapteva said, so FDA and NIH collaborated on developing the Biomarkers, EndpointS, and other Tools (BEST) resource, which was published in 2015.³ According to BEST, surrogate endpoints can be divided into the following summary categories:

___________________

¹ For FDA guidance on clinical trial endpoints, see https://www.fda.gov/regulatory-information/search-fda-guidance-documents/clinical-trial-endpoints-approval-cancer-drugs-and-biologics (accessed January 13, 2020).

² For FDA guidance on expedited programs for serious conditions, see https://www.fda.gov/regulatory-information/search-fda-guidance-documents/expedited-programs-serious-conditionsdrugs-and-biologics (accessed January 13, 2020).

³ Additional information on the Biomarkers, EndpointS, and other Tools resource is available at https://www.ncbi.nlm.nih.gov/books/NBK338448 (accessed January 13, 2020).

• Validated surrogate endpoints are those supported by a clear mechanistic rational as well as by clinical data providing evidence that the surrogate endpoint predicts a specific clinical benefit.
• Reasonably likely surrogate endpoints are supported by a strong mechanistic or epidemiologic rationale but lack adequate clinical data showing that the surrogate endpoint will predict a specific clinical benefit.
• Candidate surrogate endpoints are still under evaluation as to how they may predict clinical benefit.

FDA, Lapteva said, has posted lists of endpoints used in traditional and accelerated approvals of drugs and biological products.⁴ In the case of gene therapies under development, FDA recognizes that for many diseases for which gene therapies may be beneficial (i.e., rare genetic diseases), there are no reliable clinical or surrogate endpoints, particularly endpoints reflective of early disease manifestations. In those cases, she said, investigators need to develop novel endpoints.

In closing, Lapteva listed some points to consider in choosing endpoints for clinical trials with gene therapies:

• The possibility of long-term or potentially irreversible effects of gene therapy treatments leaves little room for uncertainty about endpoint performance at the stage of study design and requires increased vigilance concerning the validity and accuracy of endpoint measurements during the study.
• Endpoints reflective of common pathogenic pathways, but mechanistically agnostic to the target disease or condition, may not be sufficiently sensitive in gene therapy clinical trials. Indeed, the increased availability of genetic screening, early diagnosis, and advanced laboratory testing has shifted the demand toward surrogate and clinical endpoints reflective of early disease manifestations, while the identification of genetic defects associated with poorly characterized phenotypes has increased the need for novel clinical endpoints.
• In addition to finding disease-specific surrogate endpoints, there may be an opportunity to identify and validate surrogate endpoints along the universal pathway of gene transcription, transgene protein synthesis and levels, functional activity, and clearance. The

___________________

⁴ Additional information is available at https://www.fda.gov/drugs/development-resources/table-surrogate-endpoints-were-basis-drug-approval-or-licensure (accessed December 11, 2019) and https://www.fda.gov/drugs/development-resources/clinical-outcome-assessment-compendium (accessed December 11, 2019).

• principles of such endpoint identification and validation may be applicable to multiple diseases and different types of gene therapy products.

ENDPOINTS FOR GENE THERAPY CLINICAL TRIALS IN POMPE DISEASE

Focusing in on clinical endpoints for a specific condition, Koeberl began his presentation by giving a brief background on Pompe disease. Pompe disease, or glycogen storage disease type II, is caused by a deficiency of the enzyme acid alpha-glucosidase (GAA) in skeletal muscle and heart. It can be treated successfully by giving patients GAA, which is taken up by a receptor in muscle and heart cells. Enzyme replacement therapy is the standard of care for the disease; by contrast, the approach that Koeberl and his colleagues have taken is to create a recombinant AAV8 vector to deliver the GAA gene to the liver, where it can produce high levels of GAA that enters into blood circulation and eventually travels to the heart and muscle cells, correcting the GAA deficiency.

The primary advantage of this liver-specific expression, Koeberl explained, is that it suppresses the production of antibodies against GAA, which can interfere with the current GAA enzyme replacement therapy.⁵ Other potential advantages of a gene therapy approach versus enzyme replacement therapy that have been identified from preclinical studies include sustained levels of GAA in blood, an increased uptake of GAA by muscle, a more complete correction of the enzymatic deficit, potentially decreased mortality, and a one-time dose of the gene therapy vector versus required injections of GAA every 1 to 2 weeks (Bond et al., 2019). Experiments in mice showed that GAA levels in the liver, heart, diaphragm, and quadriceps increased more in the mice that received one dose of gene therapy than in mice that received four injections of the enzyme (Han et al., 2017, 2019). Glycogen content, however, is a more sensitive measure of biochemical correction, and both treatments significantly decreased the amount of glycogen in the heart and diaphragm.

Koeberl discussed how his group chose the endpoints for its clinical trial from the multiple endpoints that have already been developed for Pompe disease. One of the challenges in developing a standalone gene therapy for the disease, he said, is having to manage the interactions between the standard-of-care treatment and the gene therapy because withholding standard of care in the early phases of disease would be unethical. “We

___________________

⁵ The liver-targeted gene therapy delivery and expression of GAA induces immune tolerance by suppressing regulatory T cells.

have to consider that when we are designing endpoints and when we are collecting data,” Koeberl said.

Touching on all of the available endpoints and outcomes for Pompe disease, Koeberl briefly described a recent Phase 1/2 clinical trial of clenbuterol (Koeberl et al., 2018). The intent of this study was to gain a better understanding of how patients who received clenbuterol improved from baseline over time in terms of endpoints such as the 6-minute walk test, pulmonary function tests, and a muscle biopsy. Switching gears back to gene therapy approaches for Pompe disease, Koeberl discussed an ongoing Phase 1 trial of the AAV8 vector, in which safety is the primary endpoint. Safety in this trial is being evaluated by the incidence of adverse events and through monitoring clinical laboratory abnormalities, he said. Secondary endpoints include muscle function and pulmonary function tests, GAA activity and glycogen content in muscle biopsies, antibody formation, a urinary biomarker, and serum levels of GAA, the last of which Koeberl characterized as “very exploratory.” Except for the last endpoint, each of the secondary endpoints was validated in the earlier clinical trial with clenbuterol, with two markers of muscle and pulmonary function suitable for using in a regulatory submission, he said (Koeberl et al., 2018).

On the issue of how to tease apart the effects of standard of care and gene therapy, Koeberl explained that standard of care will stabilize muscle function, but further improvements will decline after the first couple of years of therapy (Harlaar et al., 2019). Therefore, the plan will be to enroll stably treated patients, which will make it possible to credit any improvement in muscle function to gene therapy, not enzyme replacement therapy. Koeberl noted, however, that for individuals, it is important to look at baseline levels and history before attributing improvements to gene therapy because someone who has been treated with enzyme replacement therapy for a long time may have a number of variables to consider. GAA levels also fluctuate during enzyme replacement therapy, he said, so the timing of a muscle biopsy relative to treatments with enzyme replacement therapy is important. It will also be necessary to stop standard of care at some point to demonstrate that gene therapy can serve as a standalone treatment, which will require well-designed criteria for withdrawing standard of care as well as reinstituting it if needed. Muscle glycogen content could be a good surrogate endpoint, once validated, Koeberl said, given that Pompe disease is a glycogen storage disease with glycogen accumulation being integral to disease pathogenesis.

A NOVEL OUTCOME MEASURE FOR GENE THERAPY FOR A FORM OF CONGENITAL BLINDNESS

The discussion on developing endpoints was continued by Maguire, who spoke about a novel outcome measure for Leber congenital amaurosis, a rare, autosomal recessive form of congenital retinal dystrophy that causes blindness. This condition results from lack-of-function mutations in the RPE65 gene as described in Katherine High’s earlier presentation. Humans and affected animals—there are naturally occurring dog models of this disease—with these mutations have early-onset blindness, abnormal eye movements, and flat electrical responses to light stimulation, Maguire said. Before gene therapy, there was no treatment for the disorder. As High noted in her presentation, gene therapy with an AAV vector carrying the wild-type RPE65 gene restored some vision to affected individuals.

For the Phase 1 trial of the AAV.hRPE65v2 gene therapy, Maguire and his colleagues used pupillometry, which provided objective evidence for improved function as an outcome measure. Maguire explained that pupillary light reflex is restored in retina exposed to the gene therapy construct and not in the uninjected, or contralateral, retina (Maguire et al., 2008, 2009). The issue with using this objective test for the FDA Phase 3 trial, Maguire said, was that pupillometry was not considered a clinically meaningful outcome. As one FDA reviewer put it, patients care about vision, not about their pupils, a sentiment with which Maguire agreed.

At the time, he said, there was no recognized outcome measure considered clinically meaningful, except for one surrogate endpoint: three lines of improvement on an eye chart. While some patients showed an improvement in visual acuity, which is central vision, the main improvement with the AAV.hRPE65v2 gene therapy is mediated by rod photoreceptors, which are involved in peripheral vision and night vision, he explained. The problem was that there was no test for this type of vision that satisfied the clinically meaningful, clinically significant mandate.

In the Phase 1/2 trial, Maguire’s team looked at mobility testing, which is essentially the ability to navigate an obstacle course in a certain amount of time, as an exploratory endpoint. An initial test of this endpoint showed that children who received the gene therapy were able to go through the course much more quickly when using their injected eye than when using their uninjected eye (Maguire et al., 2009). A more advanced form of this test measured the time to complete the obstacle course at seven different light levels ranging from 1 lux to 400 lux. FDA suggested that this test could be a good outcome measure and essentially assigned Maguire and his colleagues the task of creating a new outcome measure based on this framework. “FDA provided some excellent feedback on developing this into

a standardized, statistically rigorous test that would satisfy the clinically meaningful mandate,” Maguire said.

The resulting MLMT is a novel metric that measures the speed and accuracy with which a subject can ambulate independently under different ambient light conditions. Maguire and his team built 12 obstacle courses, the choice of which was randomized for each run at a specific light level corresponding to various light conditions encountered during daily living (Russell et al., 2017). For example, a moonless summer night would be 1 lux of intensity, an outdoor train station at night would be 50 lux, and a bright office building would be 400 lux. Subjects were dark-adapted for 40 minutes and then asked to navigate a course as quickly as possible with the fewest errors possible, Maguire said, adding that FDA helped his team develop the test to be rigorous, objective, and reproducible. Each patient video was reviewed and graded by two examiners, who were blind as to whether the test was performed pre- or post-treatment, Maguire said.

A validation study of the MLMT showed that its results correlated with measures of visual acuity and visual field. Normal-sighted subjects all passed the test (with regard to time and accuracy) at all light levels (Chung et al., 2018). None of the individuals with inherited retinal disease improved from baseline to year 1, and 28.5 percent of the subjects declined in performance over 1 year.

In the Phase 3 trial that FDA approved, Maguire said, the MLMT provided clear evidence for a statistically significant improvement in visual function for patients receiving gene therapy (Chung et al., 2018). In addition, he said, the results of this test correlated well with measures of visual function such as sensitivity, which is the ability to perceive different ambient light levels, and for two different measures of visual field.

In summary, Maguire said, the MLMT is a novel test that was developed to provide a primary outcome measure for subjects receiving investigational products for inherited retinal dystrophies resulting in reduced retinal sensitivity and visual field. Its essential features include the ability to differentiate low-vision subjects from normal subjects, to detect changes in clinically meaningful visual function over time, and to identify a wide range of performance characteristics among the visually impaired.

DETERMINING OPTIMAL ENDPOINTS FOR GENE THERAPY IN SICKLE CELL DISEASE

Developing endpoints for sickle cell disease has long been a challenge for the field and was recently the focus of a workshop hosted by the

American Society of Hematology and FDA, Kanter said.⁶ One reason it has been difficult to develop clinical endpoints is because there is a large amount of phenotypic diversity associated with the disease, which is not necessarily accounted for by hemoglobin genotypes, she said. “While clinical patterns exist, each individual with sickle cell disease is unique and may have a unique clinical course,” Kanter said.

The one hallmark feature of sickle cell disease, she said, is pain, the primary reason why individuals with this disease will encounter medical specialists throughout their lives. The consequence with the most impact, however, is death. While childhood mortality has improved significantly since the 1970s, when more than 10 percent of children with the sickle cell mutation died by age 4, little progress has occurred regarding adult mortality, with sickle cell patients dying, on average, in their early 40s (Paulukonis et al., 2016; Quinn et al., 2010).

One of the challenges facing researchers working on new treatments for sickle cell disease, Kanter said, is that there is no national registry of patients even though there are more than 100,000 individuals with sickle cell disease in the United States alone. The lack of a national registry makes it hard to conduct natural history studies, she said, as well as making it difficult to determine optimal endpoints for clinical trials.

Current therapies, Kanter said, include small molecule drugs (e.g., hydroxyurea, L-glutamine), blood transfusions, and palliative pain management. Due primarily to the wide disparity in benefits and side effects, none of these therapies are broadly accepted. Stem cell transplants, as Courtney Fitzhugh described in her earlier presentation, can cure sickle cell disease and are particularly promising for children with matched related donors; the risk versus benefit for adults is improving as well, and early studies paint an optimistic picture for improved outcomes and quality of life (Aslam et al., 2018). Stem cell transplants do come with the risk of GVHD and can require immune suppressive medication over the long term. There is also the risk of a late rejection of the transplant. Gene therapy, Kanter said, would circumvent the need to find a matched donor; it can take the form of either adding a new gene that produces normally functioning hemoglobin or gene editing, which would correct the mutation in the body.

The most commonly used primary endpoint has been pain severity during a vaso-occlusive crisis, but, as Kanter pointed out, pain is subjective and

___________________

⁶ Following the workshop, in December 2019, two publications that discuss the findings from the American Society of Hematology and FDA meeting were released. The first paper covers patient-reported outcomes, pain, and issues with the brain and can be found at https://ashpublications.org/bloodadvances/article/3/23/3982/429244 (accessed January 26, 2020). The second paper that explores renal and cardiopulmonary endpoints along with a measurement of the cure and a discussion of low-resource settings can be found at https://ashpublications.org/bloodadvances/article/3/23/4002/429243 (accessed January 26, 2020).

can have many causes having nothing to do with sickle cell disease. Various biologic endpoints have been proposed, but none have been validated in sickle cell disease, and the same is true for biologic predictors of disease severity. “We cannot identify, when a person is born with sickle cell disease, if it will be severe, if they will have frequent pain crises and be in the hospital frequently, or even if the disease will not manifest itself until the individual is a young adult,” Kanter said.

There was hope that the presence of fetal hemoglobin might predict individuals who will have an easier disease course, but studies have found that many adults with persistent fetal hemoglobin have all of the same complications as those without persistent fetal hemoglobin, even though the disease appears 10 to 20 years later in those with persistent fetal hemoglobin. Total hemoglobin does seem to correlate with some disease-specific mortality measures, such as renal dysfunction and stroke, Kanter said, but it is unclear if altering an individual’s total hemoglobin prior to disease manifestation will change the disease course.

The outcomes of stem cell transplants have demonstrated something important, Kanter said, and have shown that sufficient engraftment of donor stem cells leads to curative therapy. These studies have also provided evidence that stem cell transplants are successful when they result in non-sickle hemoglobin engraftment accounting for at least 50 percent of total hemoglobin production (produced from as little as 20 percent of the stem cells in patients with mixed chimerism). With gene therapy, she explained, the goal is to have pancellular expression, where every bone marrow cell expresses both non-sickle (hemoglobin A or F) and sickle cell hemoglobin. When red blood cells express both forms of hemoglobin, the normal hemoglobin can outcompete the sickled form resulting in red blood cells with a normal (or near normal) shape. Incomplete transfection or subtherapeutic doses that do not result in pancellular expression would allow for the formation of some red blood cells with only the mutant hemoglobin, which would cause sickling and clinical complications.

Work done with bluebird bio (a biotechnology company based in Massachusetts), using a vector that delivers a functional adult hemoglobin gene rather than a fetal hemoglobin gene, has shown that the vector copy number (the average number of gene therapy vectors delivered to a sample of blood stem cells), the percentage of stem cells that have been transduced or have received a gene therapy vector, and cell dose (the amount of a patient’s own blood stem cells returned to the patient after transduction was delivered) correlate well with the quantity of hemoglobin those cells produce. The vector copy number can predict how much novel therapeutic hemoglobin the patient will make, Kanter said. She noted, too, that clinical studies have shown that over time the amount of healthy hemoglobin increases because those cells outlive the sickled hemoglobin-containing cells.

While clinical results so far have been encouraging, it will be important to determine how individuals receiving gene-based therapy do over the long term regarding resolution of vaso-occlusive pain, decrease in stroke risk, and stabilization of organ dysfunction. There are safety concerns with gene therapies, Kanter noted, including insertional oncogenesis and a lack of sustainable protein production. There might also be off-target effects that will be difficult to identify, and novel mutations may occur. Research needs to determine the stopping points at which an investigation would end with poor outcomes and identify measures to define success that will be accepted by FDA and other regulatory agencies, she said. Currently, there are no stopping guidelines; however, Kanter said, as more information is gathered, there will be the ability to identify new surrogate endpoints, such as persistently low vector copy numbers, which likely means there is insufficient healthy hemoglobin production to modify the disease course. Because there are many unique ongoing trials, a stopping rule would likely have to consider outcomes other than just safety and efficacy, she said. Finally, Kanter added, investigators will need to monitor gene therapy recipients over the long term to see if these therapies provide long-term disease management or a cure and to determine how to make this type of therapy available, affordable, and universal.

DISCUSSION

A moderated panel discussion and question period with the workshop audience followed the presentations. Topics explored during this panel included patient perspectives on clinical endpoints, endpoint validation, and the costs associated with endpoint research.

Leveraging Patient Perspectives and Data

“How do patients feel about the endpoints used in clinical trials for Pompe disease?” a workshop participant asked, noting that tests such as the 6-minute walk test are not very popular among patients with neuromuscular diseases. There are perhaps more clinically relevant endpoints available, Koeberl said, such as the gait, stairs, gower, chair assessment, which has been validated and also correlates to the 6-minute walk test, so it will likely be used in the future. Developing more disease-specific endpoints will also be important, he added.

When asked to comment on why she thinks there is no national registry for sickle cell disease patients, Kanter replied that it comes down to decentralized patient advocacy for this condition and a lack of funding. She said that there are several groups starting to put together different forms of a registry, so collaboration is going to be key. “We want to make sure that

too many people are not in the kitchen cooking up different registries,” she said. In recent years the American Society of Hematology has undertaken sickle cell disease as its first disease-specific initiative, which is making a difference, she said, as has NIH and NHLBI’s embrace of this disease, although the institutes have indicated they are not interested in being the long-term funder of a registry.

Developing and Validating Endpoints for Gene Therapies

Will there be a time, a workshop participant asked, when the resolution of anemia in sickle cell disease will become an accepted endpoint, just as the resolution of hypertension is the endpoint for drugs designed to lower blood pressure instead of a reduction in the incidence of stroke? It will be important, Lapteva answered, to demonstrate how improving anemia correlates with reduced disease burden, such as by reducing hospitalization, improving respiratory function, and reducing fatigue. “There are all of these ways to look at the reduction of the disease burden from the perspective of how you improve the symptoms,” she added. “If you start thinking about it from that perspective, then there are ways to potentially look and validate the endpoint that you are talking about in terms of how it improves the outcomes in a patient.” Kanter said that it can take years for those downstream patient outcomes to improve, which raises concerns in the sickle cell disease community that the validation of a surrogate endpoint, such as normal hemoglobin production, will take a very long time.

One participant asked how companies can get regulatory feedback on novel outcome measures outside of a development program. Lapteva answered that FDA has programs called the Drug Development Tool Qualification Programs,⁷ which are meant to assess measures in a pre-competitive environment when there is no worry about disclosing proprietary information.

A participant asked whether microperimetry, which has the ability to interrogate the same point on the retina over time and has high sensitivity to variable light levels, might have the potential to serve as a sensitive measure of retinal disease over time. Many groups are working to validate measures like pupillometry as a surrogate endpoint, Maguire answered, but doing so will require correlating measurements with patient functional activity, such as an improved ability to walk around town.

Panelists were asked for their opinion of the statement that it is not possible to validate a surrogate endpoint in a rare disease. With many rare diseases, Maguire said, the timeline is so long that it is difficult for patients

___________________

⁷ Additional information is available at https://www.fda.gov/drugs/development-approvalprocess-drugs/drug-development-tool-qualification-programs (accessed December 12, 2019).

and researchers alike to stay engaged and to find funding sources that are willing to commit to long-term studies that might last 5 or 10 years. The larger number of sickle cell disease patients makes it possible to validate surrogate endpoints for that disease, Kanter said, but the biggest obstacle is the lack of a national registry. “We would have a much better understanding of what in a 2-year-old would predict mortality in a 25-year-old if we had a longitudinal registry,” she said. Koeberl added that there may be a couple of good surrogate endpoint candidates for retinal diseases, and it is just a matter of time to collect the data and confirm validation.

Endpoints are critical, a workshop participant said, but the costs of endpoint research will be significant. How is the field thinking about this? Maguire said that young investigators will likely not choose the topic because of the issues with funding. Maintaining a long-term focus and encouraging researchers and funders to consider endpoint development and research is challenging, he added. Sickle cell disease is somewhat different in that endpoints are a popular topic, Kanter said. In academia and in various organizations, the importance of endpoints needs to be realized in order to continue this type of research, she said. And High added, “I think we should really not underestimate the challenges involved in coming up with new endpoints that are really accepted by the clinical community, by regulators, and so forth.” This is a crucial piece of developing gene therapies, she said.

This page intentionally left blank.