Gene Therapy for Hemophilia: Are Expectations Matching Reality?

Main Text

BioMarin Pharmaceuticals’ ongoing phase 3 trial (ClinicalTrials.gov: NCT03370913) is designed to test a single dose (6E13 vg/kg) of valoctocogene roxaparvovec (an adeno-associated viral serotype 5 vector expressing B-domain deleted coagulation factor VIII [AAV5-FVIII]) in men with severe hemophilia A. In 2019, BioMarin reached an agreement with the US Food and Drug Administration (FDA), as part of the breakthrough therapy designation, that they could submit data for provisional approval on 16/134 participants at 6 months, rather than the full study results at the 1 year mark. On August 18, 2020, the FDA issued a complete response letter (CRL) to BioMarin. This turn of events came as a surprise to everyone, including BioMarin. A CRL indicates the FDA will not approve the application.¹

According to BioMarin, the FDA recommended that the company complete the phase 3 study and submit 2-year follow-up safety and efficacy data on all study participants. The FDA concluded that the differences between study 270-201 (phase 1/2) and the phase 3 study limited its ability to rely on the phase 1/2 study to support the durability of effect.² The additional year of requested follow-up will extend the study completion date into the fourth quarter of 2021. This may be the first time in over 50 years of development of cutting-edge hemophilia A and B pharmaceutical products that a CRL has been received from the FDA, hence the reason for surprise.

Although the contents of the CRL have not been shared, we can conclude a number of points. The FVIII levels and curve from 16 participants in phase 3 did not match the phase 1/2 results for the same dose and time point published earlier.³ In addition, the steep drop in expression between the end of years 1 and 2 (∼56%)⁴ is not consistent with two canine studies with AAV-FVIII and several human AAV-FIX studies, where durability of response has been observed. Finally, the absence of a response from a few participants in the phase 3 study in seronegative individuals suggests there is a non-response rate in the phase 3 trial. High neutralizing antibody titers following infusion likely preclude repeat dosing with any AAV for the foreseeable future. Certainly, 2-year data would provide longer term information on variability and durability.

AAV is being studied for many genetic and acquired indications. When given systemically, several AAV serotypes exhibit hepatic tropism owing to the unique physiology of liver sinusoids and permissive binding and internalization within hepatocytes. However, we know little about subsequent events, all of which contribute to the final outcome of protein expression. Hepatocytes, the target, are not homogeneous cells. Rather, they are zonally distinct and poorly characterized phenotypically, and they represent distinct subpopulations of metabolic activity whose differential interactions with AAV are poorly understood. Little is known also about AAV activity in the 30% of hepatocytes that are polyploid. Hepatocytes divide at baseline rates; yet, in some human trials (e.g., AAV-FIX), transgene expression is stable when it would be expected to decline. For all we do not know about how AAV vectors travel from peripheral vein infusion to systemic protein production, we also know little about the target cell for the latter. Thus, human studies will encounter issues not found in their animal model counterparts, which have proven to be poorly predictive of human gene therapy outcomes. Answers to questions such as these are not required if the data support a curative state. It is when the data do not measure up that some answers are needed.

What are the issues? Variability, predictability, and durability for starters. Small molecule drugs, protein therapeutics, and monoclonal antibodies show interindividual, and sometimes intraindividual, differences. However, the coefficient of variation around the means of these pharmacokinetic (PK) and pharmacodynamic (PD) differences is small compared to the differences one sees in gene therapy. Ethical drugs are prepared by good manufacturing practices (GMPs) using reproducible and quality controlled methods. The individual human body and how it responds to the drug is the major variable. However, with gene therapy, we have a compounded problem. First, although vector is produced by GMP, we cannot be sure each vector lot is identical to the last because we do not yet understand all the parameters to measure that are required for establishing identity between lots. Second, each individual is a bioreactor, required to make transgenic protein subject to unique PK and PD variations. We are definitely not making the protein by GMP standards within our bodies. The fundamental difference between interindividual variability between drugs and gene therapy is that the drug dose and regimen is easily changed to accommodate these differences. Gene therapy treats a population, which, by definition, will include high and low responders, with no recourse for correction of low or high responses.

Are variability, lack of predictability, and durability intrinsic functions of AAV gene therapy that must be accepted? Possibly. Considering the contributors to variability we are aware of, plus those we are not, reducing the coefficient of variation around each seems an insurmountable task. We understand so little about the entire in vivo bioproduction process that we cannot yet identify major contributors to the variability that may be manipulated. Local versus systemic delivery is an obvious example of an adjustable variable: local delivery obviates the role of antibodies and other inhibitory substances from blocking systemic access of the vector to target cells. Another example may be drugs that promote endosomal escape.⁵ The in vivo multiplicity of infection (MOI) from systemic AAV is ∼5,000–60,000 vg/hepatocyte, not unlike in vitro MOIs. The productive vector gene dose relative to the input dose is very, very small. What if the MOI could be cut 100-fold by concomitant drug therapy to influence the path to stable transgene episomal formation in the nucleus? Given the dose-dependent toxicity manifest by transaminase elevations, this is also a reasonable hypothesis to test. It is clear that the therapeutic window is presently quite small.

Disorders such as hemophilia can forgive a lot of variability, but even this disease has limits. For an individual with severe hemophilia (<1% factor levels), reaching 5% or 100% results in a huge change in the quality of life. But these two outcomes are not the same. One is still dependent on exogenous treatment for occasional bleeding episodes, while the other is effectively cured. Several questions arise. The first question is whether the benefit/risk is favorable for a suboptimal result. Second, what is the risk of transgene-related toxicity when delivering >100% levels due to individual lack of predictability? Finally, what is the risk of achieving no result due to the variability? The answers depend upon the disorder, availability of alternative therapies, risks of over- or under-treatment, and importance of being able to treat with AAV again if a suboptimal result is achieved. For hemophilia, where there are effective protein replacement therapies, assessing the benefit/risk in undergoing AAV gene therapy in its present form requires thoughtful discussion with all the stakeholders, starting with the patient.⁶

Conflicts of Interest

G.F.P. reports the following potential conflicts of interest: Consultant for Ambys Medicines, BioMarin, CRISPR Therapeutics, Decibel Therapeutics, Geneception, Generation Bio, Pfizer, and Takeda. He is on the board of directors for Voyager Therapeutics and the World Federation of Hemophilia (WFH), vice president medical for WFH, and a member of the (US) National Hemophilia Foundation Medical and Scientific Advisory Council.