Clinical Trial Designs For Advanced Therapies

Volume 5, Issue 11
CELL & GENE

THERAPY INSIGHTS
SPOTLIGHT ON:
Clinical Trial Designs for Advanced Therapies
Guest Edited by Dr Timothy Miller
CELL & GENE THERAPY INSIGHTS
Volume 5, Issue 11
ISSUE SPOTLIGHT:
Clinical Trial Designs for Advanced
Therapies
Guest Edited by Dr Timothy Miller, President & CEO, Abeona
Therapeutics
EXPERT INSIGHT EXPERT INSIGHT EXPERT INSIGHT

Towards individualized, low toxic Post-marketing safety and CRISPR surgery for inherited
conditioning in autologous efficacy surveillance of cell and retinal diseases: landmarks in
gene-transduced hematopoietic gene therapies in the EU: A the 21st century
cell transplantation critical review Alexander H Chai & Stephen H
Rick Admiraal, Susan Prockop & Enrico Fritsche, Magdi Tsang
Jaap Jan Boelens Elsallab, Michaela Schaden,
Spencer Phillips Hey, Mohamed
Abou-El-Enein
1495–1503 1505–1521 1451–1456
EXPERT INSIGHT EXPERT INSIGHT EXPERT INSIGHT
Bayesian phase 1/2 trial designs Considerations for patient The evolution of
and cellular immunotherapies: a selection for cell and gene adeno-associated virus capsids
practical primer therapy trials using tumor for CNS gene therapy
Jordan Gauthier, Ying Yuan & associated antigens as target in Steven J Gray
Peter Thall early phase development
Stephanie Traub & David Edwards
1483–1494 1213–1224 1361–1368
REGULATORY PERSPECTIVE INTERVIEW INTERVIEW
Clinical trials of advanced therapy Keys to success for foundation: Breaking new ground: bringing
investigational medicinal products industry clinical development an iPS cell therapy to the clinic
in Spain: preparing for the collaborations Kapil Bharti
European clinical trials regulation Brian Fiske
Juan Estévez Álamo, Marcos
Timón, Cristina González
Gómez-Platero, Carmen Doadrio
Abad, Marta Velasco González,
María Yolanda de Mingo
Ballesteros, María Ángeles Martín
de la Sierra San Agustín & María
Antonia Serrano Castro
1347–1359 1267–1274 1369–1375
IN FOCUS:
Highlights from our Vector Channel
Adherent Culture Methods
EXPERT INSIGHT INTERVIEW

Evaluation of AAV vector Assessing the future prospects
production from the iCELLis of upstream bioprocessing
fixed bed bioreactor vessel systems for commercial AAV
Shelley Nass, Bindu Nambiar, production
Maryellen Mattingly, Scott A Jeffers
Denise Woodcock & Catherine
O’Riordan
1461–1471 1275–1279
CLINICAL TRIAL DESIGNS FOR

ADVANCED THERAPIES
EXPERT INSIGHT
Towards individualized, low toxic

conditioning in autologous
gene-transduced hematopoietic cell
transplantation
Rick Admiraal, Susan Prockop & Jaap Jan Boelens
Allogeneic hematopoietic cell transplantation (Allo-HCT) has become

much safer over the last couple of years. This, together with the rapidly
evolving autologous HCT-gene therapy options, is expected to result in
an increase in the number of patients receiving HCT. Autologous HCT-
gene therapy is a more advanced, safer and precise option for monoge-
netic life threatening disorders. The efficacy of gene therapy, however,
does not only rely on the gene construct itself, but also on the condi-
tioning applied before the gene therapy. In this review we describe how
the conditioning can impact the outcomes of the allo-HCT and gene
therapy and we will provide a future perspective on how to further
improve the efficacy and reduce the short- and long-term toxicity of the
conditioning.
Cell & Gene Therapy Insights 2019; 5(11), 1495–1503
DOI: 10.18609/cgti.2019.155
INTRODUCTION there is a long-standing history of Most of the non-malignant dis-

In pediatric allogeneic hematopoi- performing transplants for malig- eases are mono-genetic diseases
etic cell transplantation (allo-HCT) nant and non-malignant disorders. of the immune-system, red cells,
www.insights.bio 1495
lysosomal enzyme deficiencies and not be the same, due to variabili-

bone marrow failure syndromes. ty in pharmacokinetics (PK) and
As transplant has become safer over pharmacodynamics (PD) of agents
the last decade, the number of paused in the conditioning [2–4]. PK
tients receiving an allogeneic-he- includes all processes that influence
matopoietic cell transplantation concentration over time, i.e., clear-
(allo-HCT) for these disorders has ance, distribution, absorption, etc.
increased to approximately 50% of PD on the other hand describes the
the pediatric patients transplanted relationship between concentration
annually worldwide (CIBMTR. over time and drug effects (efficacy
org). With safer allo-HCT strate- and toxicity). Variables such age,
gies and the rapid evolving autol- body size, organ function and con-
ogous HCT-gene-therapy options, comitant medications can influence
the expectation is that this number the PK profile resulting in variable
will further grow in upcoming de- PD outcomes. Busulfan, an alkylat-
cade. Also, with the advent of new ing agent, is the best studied agent in
technologies, it is expected that the conditioning regimen and seems
standard allo-HCT transplants to also crucial in the conditioning be-
be gradually replaced by autologous fore GT [5–11]. Therefore we will
gene-transduced hematopoietic cell mainly focus on this agent, because
transplantation (gene-therapy: GT) immune suppression with agents
for patients with a mono-genetic like fludarabine and ATG (anti-thy-
life-threatening disorder: a more mocyte globuline) are less/not rel-
advanced, safer, precision strategy. evant in the conditioning for GT.
The effect of the GT does however Busulfan PK has been studied by
not only rely on the gene-construct several groups (adult and pediatric),
itself but also on the conditioning which has led to several PK-models,
applied before the GT. In other mainly developed in cohorts of in-
words, the efficacy of a GT treat- fants and children but also in some
ment depends on the package of adult cohorts [4,5,7]. The optimal
conditioning and GT product com- therapeutic window of busulfan
bined. In this review we describe exposure has been established in
how the conditioning can impact multiple reports [5,8]. This optimal
the outcomes of the allo-HCT and exposure appears to be independent
GT and we will provide a future on cell source, match grade, indica-
perspective on who to further im- tion and concomitant conditioning
prove the efficacy and reduce the agents. Although the optimal ex-
short- and long-term toxicity of the posure was similar when receiving
conditioning. 1 (Busulfan as single alkylator), 2
(Busulfan combined with cyclo-
phosphamide; Cy) or 3 alkylators
(Busulfan, Cy and Melphalan), pa-
BACKGROUND tients receiving only busulfan com-
It is well recognized that differences bined with fludarabine had lowest
in conditioning regimen may con- toxicity and superior overall surviv-
tribute to differences in outcome al chances (due to lower toxicity:
[1]. Even within patients receiving e.g., veno-occlusive disease [VOD],
the same conditioning regimen with graft versus host disease [GvHD]
comparable doses, outcomes may and idiopatic pneumonia syndrome
1496 DOI: 10.18609/cgti.2019.155

expert insight
[IPS]). The optimal cumulative tar- is made that the pharmacokinetics

get exposure for Bu AUC0–4 days (PK; e.g., clearance, volume of dis-
was found to be 90 mgxh/L (range tribution) also increase linearly with
80–100 mg*h/L, over 4 days), for body weight in order to reach com-
all cell sources, including cord parable concentrations. In addition,
blood [5]. Optimal myeloablation the assumption is made that the
also seems of great importance for concentration–effect relationship is
GT. Sessa et al. showed in the len- comparable between children and
tiviral GT trial in early-onset meta- adults. However, since developmen-
chromatic leukodystrophy that pa- tal changes are mostly non-linear
tients with sub-ablative exposure [14], empirical dosing can lead to
of busulfan had lower engraftment underdosing or overdosing. This is
of gene-transduced cells, resulting especially true in the very young
in lower (not supra-normal; which children and adolescents, thereby
was the goal of this intervention) introducing toxicity or reduced effi-
enzyme levels [12]. Patients with cacy [15,16]. In order to reach opti-
ablative (AUC0-4 days of 80– mal exposure in all patients, the PK
100 mg*h/L) busulfan exposure and pharmacodynamics (PD) need
achieved supra-normal levels (5–10 to be described, including the influ-
times normal) [12]. Also, in other ence of predictors such as body size
autologous gene-transduced HCTs, on PK and PD. With these models,
aka GT (e.g., Wiskot Aldrich, Fabry the optimal dose for any individu-
disease, Beta-thalassemia) sufficient al patient can be predicted to reach
ablation seems important for opti- optimal exposure. This approach
mal effect. has been demonstrated in pediatric
HCT [10]. While most cytostatic
agents used in HCT are dosed us-
ing a fixed mg/kg or mg/m2 dose
POPULATION for all patients, busulfan dose is
PHARMACOKINETICS & fully individualized and controlled
PHARMACODYNAMICS: using therapeutic drug monitor-
TOWARDS PRECISION ing (TDM) [10]. Recent work has
DOSING IN TRANSPLANT shown that actual exposure to bu-
Because the GT solutions developed sulfan impacts outcome in terms of
are for congenital diseases, most GT toxicity, graft failure and relapse as
transplanted patients will be pedi- described above [4,5].
atric patients, although some con- The population approach, using
genital disease may occur in adult- advanced non-linear mixed effects
hood. Knowing this we need to modeling and high computing
understand that many drugs (used power, is the preferred method for
in transplant) are not evaluated in PK analyses according to both the
children, contributing to off-la- FDA and EMEA guidelines [17,18].
bel or unlicensed use in as high as In the population approach, data
49–87% of drugs used in tertiary from all patients is pooled to es-
care hospitals [13]. Pediatric dosing timate a population mean for all
regimens are often empirical, linear- PK-parameters [19]. Next, based on
ly extrapolated from adult dosing individual concentrations inter-in-
based on body weight. When using dividual variability and residual er-
a per kilogram dose, the assumption ror are calculated for each patient.
Cell & Gene Therapy Insights - ISSN: 2059-7800 1497

Main advantage of the population of incorporating the individualized

approach is the ability to use sparse- dosing in clinical practice. As such,
ly sampled and unbalanced (dif- most centers use individualized bu-
ferences in number of samples and sulfan dosing with therapeutic drug
sample times between patients, as monitoring [5,10]. Less of interest
often the case) data [20]. This makes for conditioning in GT (more for
the population approach particu- allo-HCT), individualized dosing
larly attractive in pediatrics, where for ATG and fludarabine has been
few samples are available, and the designed and is currently evaluated
absolute dose varies significant- in clinical trials based PKPD analy-
ly between children. Additionally, ses [22–24]. PKPD analyses is also
the estimation of PK-parameters is of importance for novel, chemo-free
more robust compared to non-com- regimens in the future, such as de-
partmental analyses such as the scribed in next section.
two-stage approach as the software Finally, the currently available
is able to differentiate between real models may be further sophisticat-
inter-individual variability and re- ed, describing not only PK or PD,
sidual error (a combination of in- but rather the complete spectrum of
correct sample times, measurement drug treatment, including dose, PK,
errors and model misspecification) biomarker response, clinical efficacy
[21]. Altogether, from an ethical, and toxicity in one comprehensive
practical and methodological point model. We expect development and
of view, the population approach implementation of individualized
is the preferred method for PK dosing to take place in the next
analyses. 10 years, thereby improving the
After describing the population knowledge and efficacy of clinical
pharmacokinetics, the relationship drug therapy, and improving clini-
between concentrations or exposure cal outcome following HCT. With
and effects or toxicity (PD) needs individualized dosing, unwanted
to be determined. The PD-analy- variability in drug exposure will be
sis will give further insight into the reduced, leading to predictable, ad-
therapeutic window and will set an justable and improved outcome of
optimal target exposure. Next, an allo-HCT and GT.
individualized dosing regimen can
be designed using the population
PK model, aiming for optimal ex-
posure. The proposed individualized ANTIBODY-BASED
dosing regimen should be evaluated CONDITIONING AS
in a prospective trial, both for exter- FUTURE PERSPECTIVE
nal validation of the PK-model and In addition to achieving better effi-
the clinical safety and efficacy [2]. cacy and less toxicity by improving
Individualized drug dosing is in- the way agents are dosed in condi-
creasingly incorporated, especially tioning regimens for allo-HCT and
in pediatrics where differences in GT, there is an emerging approach
PK between children of different age exploring the use of antibody-based
groups are major. While individual- conditioning. This because chemo-
ized dosing regimens are designed therapy-based regimens, although
according to the above in many due to PKPD considered less toxic
fields, we feel HCT is at the front in the short term, they come with
1498 DOI: 10.18609/cgti.2019.155

expert insight
significant late effects, including grafts, engraftment may have been

infertility. Antibody based condi- mediated by donor immune popu-
tioning relies on using specifically lations not only antibody mediated
targeted antibodies to transiently or immune ablation. While there is
permanently deplete components of preclinical data suggesting it might
the recipient hematopoietic system. be possible [28] it remains contro-
There is the potential to use naked versial whether anti-CD45 target-
antibodies or antibodies conjugat- ing antibodies alone can effectively
ed to drug or radiolabeled. In ad- achieve conditioning for transplant
dition, this approach could be used [29]. Methods to enhance the effica-
in isolation or in combination with cy of anti-CD45 are being explored
chemotherapy. including radio-conjugated An-
Early trials (NCT00590460, ti-CD45 (I131) Apamistamab [30]
NCT00056979 and or drug conjugated anti-CD45 with
NCT00579137) run by investiga- saponin or Amanitin [31,32].
tors at Baylor College of Medicine Depletion of recipient hemato-
and Texas Children’s used a com- poietic stem cells (HSC) alone may
bination of reduced intensity che- be sufficient to achieve clinically
motherapy (fludarabine and low meaningful responses in the setting
dose irradiation or cyclophospha- of autologous GT for diseases where
mide) with monoclonal antibodies partial engraftment of genetically
targeting most hematopoietic cells modified HSCs is sufficient to cor-
(anti-CD45 antibodies YTH.24 rect the phenotype or in allo-HCT
and YTH.54) in combination with for disorders of immunity where
alemtuzumab (anti-CD52). As re- the donor cells cannot be rejected
ported in ClinicalTrials.gov a total and partial donor chimerism is suf-
of 15 patients have received this an- ficient to correct the phenotype. The
tibody combination including three potential for antibody mediated de-
patients with SCID and five with pletion of hematopoietic stem cells
Fanconi anemia. Published results has been demonstrated using the
are reported in two patients with monoclonal antibody ACK2 in an
SCID [25] one of whom engrafted, immune deficient mouse model [33]
and in seven patients transplanted and in a Fanconi anemia mouse [34].
for malignant disease who had co- ACK2 recognizes and antagonizes
morbidities limiting the use of more c-kit and interferes with the inter-
aggressive cytoreduction [26]. These action between c-kit and its ligand
seven were transplanted from un- stem cell factor [35]. Administration
related donors after cytoreduction of the monoclonal antibody resulted
with fludarabine, 450cGy of TBI, in transient decreases in phenotypic
alemtuzumab and CD45 mono- and functional HSCs in the recipi-
clonal antibodies [26]. Six of these ent mice. In addition, although the
seven recipients engrafted includ- decrease was transient, infusion of
ing 3 of 4 recipients of mismatched purified donor HSCs during a win-
unrelated grafts. In this study the dow after clearance of the antibody
dosing of the anti-CD45 monoclo- from the serum, but prior to recov-
nal antibodies was based on kinet- ery of endogenous HSC function al-
ics previously established in a rat lowed for durable donor chimerism.
model [27]. In addition, as these A humanized version of this anti-
patients received T-replete stem cell body, AMG 191, is now in clinical

trial (NCT02963064) with promis- immuno-ablation are required, the

ing early results presented in abstract combination of antibody-based
at TCT 2019 [36]. This trial employs conditioning with conventional
antibody depletion to achieve stem reduced intensity chemotherapy
cell engraftment for individuals with may be necessary. These approaches
SCID who despite prior transplant will all likely require individualized
do not have donor myeloid engraft- (PK guided) dosing as for example,
ment and have poor immunity. As combining conventional chemo-
demonstrated in the-preclinical therapy with antibody based thera-
studies [37], the timing of infusion py can change the kinetics of clear-
of donor HSCs relative to antibody ance and the biodistribution of the
level is critical to engraftment. The antibody [42].
ongoing clinical trial is a dose esca- Understanding the kinetics of
lation trial that involves real time monoclonal antibody clearance and
monitoring of serum antibody level individualized dosing will be critical
to determine individualized timing to the success of these trials. Effec-
for infusion of the stem cell graft. tive use of antibody-based depletion
The HSC clearance mediated by may require ongoing TDM as the
AMG 191 is in part dependent on kinetics of effect and of clearance
Fc-mediated effector functions [38] will likely depend on non-linear
which may restrict its applicability factors such as the size of the tar-
in patients with defective effector get population of cells. Thus, as in
function. In addition, to extend this the example provided by Busulfan,
approach to settings where high lev- careful monitoring of dosing ki-
el chimerism is required, the use of netics and individualized timing of
a combination of anti-c-kit and anti dosing if not dosing itself, will likely
CD47 [38] to enhance the activity be essential to the broader applica-
of anti-c-kit depletion by blocking tion of these approaches that have
the CD47-SIRPa interaction and the potential to decrease the toxicity
enhancing phagocytosis is being de- of our therapies.
veloped. In addition, as in the case
for anti-CD45, pre-clinical data is
emerging for the use of saponin and
amanitin conjugated CD117 anti- IN CONCLUSION
bodies [39–41]. Allo-HCT has become much saf-
Additionally, there are examples er over the last couple of years. An
where antibody mediated lymphoid important factor in this is applying
depletion alone may be sufficient knowledge gathered from PK and
such as lymphodepletion prior to PD analyses of agents used in the
adoptive therapy with CAR T cells. conditioning regimens. Individual-
Similarly, to the timing of infusion izing agents used in the condition-
of hematopoietic stem cells after ing regimens prior to allo-HCT is
stem cell targeting antibody, the also important for GT to achieve
timing of infusion of CAR T cells sufficiently high engraftment of
would depend on clearance of T gene-transduced cells to prevent
cell targeting antibody(ies) used for disease. Currently busulfan (with
depletion. an optimal myelo-ablative tar-
In some instances, where com- get of 90; 80–100 mg*h/L) seems
plete donor HSC engraftment and the best studied and easiest to
1500 DOI: 10.18609/cgti.2019.155

expert insight
target myeloablative agent used in late effects. In the coming years,

allo-HCT and GT. Although short- we expect to see more individual-
term toxicities if targeted appropri- ized, chemo-free, dosing regimens
ately are significantly less nowadays, emerging in the field of HCT and
it may come with late complica- GT. This way, outcome will be pre-
tions, such as infertility. Replac- dictable and adjustable based on in-
ing busulfan for antibody-based dividual patients’ needs and associ-
conditioning is an interesting de- ated with very minimal short-term
velopment, that doesn’t have these and long-term toxicity.
AUTHORSHIP & CONFLICT OF INTEREST

Contributions: All named authors take responsibility for the integrity of the work as a whole, and have given their approval
for this version to be published.
Acknowledgements: None.
Disclosure and potential conflicts of interest: SP states that MSKCC has been selected as a third site for the ongoing clinical
trial NCT02963064.
Funding declaration: SP reports support for sponsored clinical trials from Atara Biotherapeutics, and Advisory board and
support for sponsored clinical trials from Mesoblast, outside the submitted work. JJB reports personal fees from Advanced
Clinical, personal fees from Bluebird Bio, personal fees from Avrobio, personal fees from Omeros, personal fees from Magen-
ta, personal fees from Bluerock, grants from Sanofi, personal fees from Takeda, outside the submitted work.
ARTICLE & COPYRIGHT INFORMATION

Copyright: Published by Cell and Gene Therapy Insights under Creative Commons License Deed CC BY NC ND 4.0 which
allows anyone to copy, distribute, and transmit the article provided it is properly attributed in the manner specified below.
No commercial use without permission.
Attribution: Copyright © 2019 Rick Admiraal, Susan Prockop & Jaap Jan Boelens. Published by Cell and Gene Therapy
Insights under Creative Commons License Deed CC BY NC ND 4.0.
Article source: Invited; externally peer reviewed.
Submitted for peer review: Oct 2 2019; Revised manuscript received: Nov 12 2019; Publication date: Nov 18 2019.
REFERENCES
1. Bacigalupo A, Ballen K, Rizzo D et 3. Admiraal R, van Kesteren C, Boelens paediatric haematopoietic stem cell
al. Defining the intensity of condi- JJ, Bredius RGM, Tibboel D, Knibbe transplantation patients: towards
tioning regimens. Biol. Blood Marrow C a J. Towards evidence-based dosing individualized dosing. Clin. Pharma-
Transpl. 2009; 15: 1628–33. regimens in children on the basis of cokinet. 2012; 51: 331–45.
population pharmacokinetic phar-
2. Admiraal R, Boelens JJ. Individu- 5. Bartelink IH, Lalmohamed A,
macodynamic modelling. Arch. Dis.
alized conditioning regimes in cord van Reij EML et al. Association of
Child. 2014; 99: 267–72.
blood transplantation: Towards busulfan exposure with survival and
improved and predictable safety 4. Bartelink IH, Boelens JJ, Bredius toxicity after haemopoietic cell trans-
and efficacy. Expert Opin. Biol. Ther. RGM et al. Body weight-dependent plantation in children and young
2016; 2598: 1–13. pharmacokinetics of busulfan in adults: a multicentre, retrospective

cohort analysis. Lancet Haematol. malignancies. Biol. Blood Marrow Data Analysis Methods. Drug Metab.
2016; 3: e526–36. Transplant. 2011; 17: 341–50. Rev. 1984; 1.
6. ten Brink MH, Zwaveling J, Swen JJ, 12. Sessa M, Lorioli L, Fumagalli F et 21. Sheiner L, Beal S. Evaluation of
Bredius RG, Lankester AC, Gu- al. Lentiviral haemopoietic stem- methods for estimating population
chelaar HJ. Personalized busulfan and cell gene therapy in early-onset pharmacokinetics parameters. I.
treosulfan conditioning for pediatric metachromatic leukodystrophy: an Michaelis-Menten model: routine
stem cell transplantation: The role of ad-hoc analysis of a non-randomised, clinical pharmacokinetic data. J.
pharmacogenetics and pharmacoki- open-label, phase 1/2 trial. Lancet Pharmacokinet. Biopharm. 1980; 8:
netics. Drug Discov. Today 2014; 19: 2016; 388: 476–87. 553–71.
1572–86.
13. Krekels EHJ, van Hasselt JGC, 22. Langenhorst J, van Kesteren C, van
7. Savic RM, Cowan MJ, Dvorak CC Tibboel D, Danhof M, Knibbe C a Maarseveen E et al. Individualized
et al. Effect of weight and maturation J. Systematic evaluation of the de- Fludarabine Dosing for Predictable
on busulfan clearance in infants and scriptive and predictive performance Immune Reconstitution and In-
small children undergoing hema- of paediatric morphine population creased Survival Chances after Alloge-
topoietic cell transplantation. Biol. models. Pharm. Res. 2011; 28: neic Hematopoietic Cell Transplanta-
Blood Marrow Transplant. 2013; 19: 797–811. tion. Biol. Blood Marrow Transplant.
1608–14. 2018; 24: S306–7.
14. Kearns GL, Abdel-Rahman SM,
8. McCune JS, Bemer MJ, Barrett JS, Alander SW, Blowey DW, Leeder JS, 23. Admiraal R, van Kesteren C, Jol-van
Scott Baker K, Gamis AS, Holford Kauffman RE. Developmental phar- Der Zijde CM et al. Association
NHG. Busulfan in Infant to Adult macology—drug disposition, action between anti-thymocyte globulin
Hematopoietic Cell Transplant and therapy in infants and children. exposure and CD4+ immune recon-
Recipients: A Population Pharmaco- N. Engl. J. Med. Drug Ther. 2003; stitution in paediatric haematopoietic
kinetic Model for Initial and Bayesian 349: 1157–67. cell transplantation: a multicentre ,
Dose Personalization. Clin. Cancer retrospective pharmacodynamic co-
15. Knibbe CAJ, Krekels EHJ, Danhof
Res. 2014; 20: 754–63. hort analysis. Lancet Haematol. 2015;
M. Advances in paediatric pharma-
2: e194–e203.
9. Long-Boyle J, Savic R, Yan S et al. cokinetics. Expert Opin. Drug Metab.
Population Pharmacokinetics of Bu- Toxicol. 2011; 7: 1–8. 24. Admiraal R, Nierkens S, de Witte
sulfan in Pediatric and Young Adult MA et al. Association between an-
16. Knibbe CAJ, Danhof M. Individ-
Patients Undergoing Hematopoietic ti-thymocyte globulin exposure and
ualized dosing regimens in chil-
Cell Transplant: A Model-Based survival outcomes in adult unrelated
dren based on population PKPD
Dosing Algorithm for Personalized haemopoietic cell transplantation: a
modelling: are we ready for it? Int. J.
Therapy and Implementation into multicentre, retrospective, pharma-
Pharm. 2011; 415: 9–14.
Routine Clinical Use. Ther. Drug codynamic cohort analysis. Lancet
Monit. 2015; 37: 236–45. 17. EMEA. Guideline on reporting the Haematol. 2017.
results of population pharmacokinet-
10. Bartelink IH, van Kesteren C, Boel- 25. Patel NC, Chinen J, Rosenblatt HM
ic analyses. 2007.
ens JJ et al. Predictive performance et al. Outcomes of patients with
of a busulfan pharmacokinetic model 18. FDA. Guidance for Industry, Popula- severe combined immunodeficiency
in children and young adults. Ther. tion Pharmacokinetics. 1999. treated with hematopoietic stem cell
Drug Monit. 2012; 34: 574–83. transplantation with and without
19. Bauer RJ. NONMEM users guide: preconditioning. J. Allergy Clin. Im-
11. Nemecek ER, Guthrie K a, Sorror introduction to NONMEM 7.2.0. munol. 2009; 124: 1062–1069.e4.
ML et al. Conditioning with treo- 2011.
sulfan and fludarabine followed by 26. Popat U, Carrum G, May R et
allogeneic hematopoietic cell trans- 20. Sheinerm L. The Population Ap- al. CD52 and CD45 monoclonal
plantation for high-risk hematologic proach to Pharmacokinetic Data antibodies for reduced intensity
Analysis: Rationale and Standard hemopoietic stem cell transplantation
1502 DOI: 10.18609/cgti.2019.155

expert insight
from HLA matched and one antigen Targeting CD45 Permits Engraft- and effective platelet gene therapy of
mismatched unrelated donors. ment across Immunologic Barriers. hemophilia A mice. Blood Adv. 2019;
Bone Marrow Transplant. 2005; 35: Blood 2018; 132. 3: 2700–11.
1127–32.
33. Czechowicz A, Kraft D, Weiss- 40. Li Z, Czechowicz A, Scheck A, Rossi
27. Dahlke MH, Lauth OS, Jäger MD et man IL, Bhattacharya D. Efficient DJ, Murphy PM. Hematopoietic
al. In vivo depletion of hematopoietic Transplantation via Antibody-based chimerism and donor-specific skin
stem cells in the rat by an anti-CD45 Clearance of Hematopoietic Stem allograft tolerance after non-genotox-
(RT7) antibody. Blood 2002; 99: Cell Niches. Science (80- ) 2007; ic CD117 antibody-drug-conjugate
3566–72. 318: 1296–9. conditioning in MHC-mismatched
allotransplantation. Nat. Commun.
28. Jäger MD, Vondran FWR, Ramack- 34. Chandrakasan S, Jayavaradhan R,
2019; 10: 1–7.
ers W et al. A depleting anti-CD45 Ernst J et al. RhIg-prophylaxis is not
monoclonal antibody as isolated influenced by FCGR2/3 polymor- 41. Czechowicz A, Palchaudhuri R,
conditioning for bone marrow phisms involved in red blood cell Scheck A et al. Selective hemato-
transplantation in the rat. PLoS One clearance. Blood 2017; 129: 1045–8. poietic stem cell ablation using
2016; 11: 1–17. CD117-antibody-drug-conjugates
35. Ogawa M, Matsuzaki Y, Nishikawa
enables safe and effective transplan-
29. Wulf GG, Luo KL, Goodell MA, S,et al. Expression and Function of
tation with immunity preservation.
Brenner MK. Anti-CD45-mediated c-kit in Hemopoietic Progenitor
Nat. Commun. 2019; 10: 1–12.
cytoreduction to facilitate allogene- Cells. J. Expir. Med. 1991; 174: 2–5.
ic stem cell transplantation. Blood 42. Jang JK, Khawli LA, Park R et al. Cy-
36. Agarwal R, Dvorak C, Prohaska S.
2003; 101: 2434–9. toreductive chemotherapy improves
Toxicity-Free Hematopoietic Stem
the biodistribution of antibodies
30. Agura E, Gyurkocza B, Nath R, Cell Engraftment Achieved with
directed against tumor necrosis in
Litzow M. Targeted Conditioning Anti-CD117 Monoclonal Antibody
murine solid tumor models. Mol.
of Iomab-B (131I-anti-CD45) Prior Conditioning. Biol. Blood Marrow
Cancer Ther. 2013; 12: 2827–36.
to Allogeneic Hematopoietic Cell Transpl. 2019; 25.
Transplantation Versus Conventional
37. Kwon H-S, Logan AC, Chhabra A,
Care in Relapsed or Refractory Acute
Pang WW, Czechowicz A, Keri T.
Myeloid Leukemia (AML): Prelim- AFFILIATIONS
Anti-human CD117 antibody-medi-
inary Feasibility and Safety Results
ated bone marrow niche clearance in Rick Admiraal
from the Prospective, Rand. Blood
nonhuman primates and humanized Department of Pediatrics, UMC
2018; 132.
NSG mice. Blood 2018; 132. Utrecht, Utrecht, the Netherlands
31. Palchaudhuri R, Hyzy S, Proc-
38. Chhabra A, Ring AM, Weiskopf K Susan Prockop
tor J. Antibody Drug Conjugates
et al. Hematopoietic stem cell trans- Stem cell transplantation and cel-
Targeted to CD45 or CD117 Enable
plantation in immunocompetent lular therapy, MSK Kids, Memorial
Allogeneic Hematopoietic Stem Cell
hosts without radiation or chemo- Sloan Kettering Cancer Center,
Transplantation in Animal Models.
therapy. Sci. Transl. Med. 2016; 8: New York, USA
Blood 2018; 132.
1–22. Jaap Jan Boelens
32. Persaud S, Cooper M, Rettig M, Stem cell transplantation and cel-
39. Gao C, Schroeder JA, Xue F et
DePersio F. Conditioning for Hema- lular therapy, MSK Kids, Memorial
al. Nongenotoxic antibody-drug
topoietic Stem Cell Transplantation Sloan Kettering Cancer Center,
conjugate conditioning enables safe
Using Antibody-Drug Conjugate New York, USA


ADVANCED THERAPIES
EXPERT INSIGHT
Post-marketing safety and efficacy

surveillance of cell and gene therapies
in the EU: A critical review
Enrico Fritsche, Magdi Elsallab, Michaela Schaden,
Spencer Phillips Hey, Mohamed Abou-El-Enein
The implementation of new regulatory tools, such as the PRIority

MEdicine (PRIME) scheme, by regulatory authorities in Europe enabled
faster patient access to innovative therapies. This early access tool goes
along with a clear need for a thorough assessment of safety and efficacy
upon marketing authorization. Due to the higher degree of uncertain-
ty when evaluating novel therapies such as advanced therapy medicinal
products (ATMPs), post-marketing surveillance studies for these prod-
ucts should be designed to make up the evidential shortfall and provide
additional evidence to inform clinical practice. Here, we describe the
status and regulatory requirements of post-marketing surveillance for
ATMPs, which we found often resembling traditional, pre-market trials,
focusing on biological mechanisms and efficacy in narrowly defined pa-
tient populations. We close by proposing the pragmatic trial concept as
a potential solution to improve data quality and evidence generation in
settings closer to real-world.
DOI: 10.18609/cgti.2019.156
FROM CLINICAL TRIALS evidence [5,6]. Developers of ad-

TO POST-MARKET: THE vanced therapy medicinal prod-
CURIOUS CASE OF ATMPS ucts (ATMPs) have proactively
integrated PRIME and other such
Evaluating a marketing-authori-
regulatory tools into their product
zation application under the con-
development strategies [7–9]. For
ventional centralized procedure of
example, by mid 2019, three AT-
the European Medicine Agency
MPs (Kymriah®, Yescarta®, and
(EMA) can be a lengthy process,
Zynteglo®) were approved under
taking up to 210 days (and that
the PRIME scheme (although
is excluding any additional time
only Zynteglo benefited from the
required for applicants to respond
accelerated assessment; assessment
to EMA requests for additional in-
of Kymriah and Yescarta were re-
formation). In an effort to support
verted to the standard timetable
drug development, the EMA has
since major objections were raised
devised several early access routes
during the regulatory evaluation
for drug developers in the EU [1]
and could not be resolved within
where some of them can be uti-
the accelerated timetable).
lized via the PRIority MEdicine
However, the uncertainty about
(PRIME) scheme [2]. The PRIME
the benefit-risk profile of new-
scheme was launched in March
ly-approved ATMPs is not merely
2016, specifically to support the
due to regulatory flexibility. There
more rapid translation of products
are several other features of clini-
targeting an unmet medical need
cal trials for ATMPs that can leave
by enhancing the early interaction
critical gaps in the evidence base
and dialogue with regulators before
concerning product safety and ef-
submission of marketing authori-
ficacy [10–15]. For example, since
zation application (MAA), as well
ATMPs often target rare diseases
as accelerating the regulatory as-
[16], the pre-market clinical trials
sessment procedure of MAA [3,4].
are mostly small, single-arm trials
While such regulatory tools can
that face an increased risk of bias
help to more expeditiously satisfy
and other translational challenges
unmet medical needs, this comes
[17]. Surveys among ATMP-devel-
at the cost of having a less com-
opment companies in Europe have
prehensive data set, and therefore,
shown that for many rare diseases
greater uncertainty about the prod-
of interest, little is known about
uct’s benefit-risk balance at the
disease progression or the chal-
time of marketing authorization.
lenges associated with creation and
However, to offset this initial lack
interpretation of reliable endpoints
of data, EMA obligates product
for new indications [18]. Selection
developers to perform extensive
of endpoints is of particular impor-
post-marketing studies in order
tance since ATMP trials mainly rely
to generate more robust evidence
on surrogate endpoints due to the
supporting the overall safety and
lack of clinically meaningful ones
efficacy profile of these products.
for many indications, such as vari-
Similar policies are adopted by the
ous cardiac cell therapy approaches
US Food & Drug Administration
[19]. However, relying on surrogate
(FDA) in cases when new drugs are
endpoints in pre-market trials only
approved on the basis of limited
amplifies the uncertainty for how
1506 DOI: 10.18609/cgti.2019.156

EXPERT INSIGHT
to appropriately use these products of study/ status”, “categorization

in the clinical setting [20]. 1-3”, “objectives”, “safety concerns
To make up for this epistemic addressed”, “date for submission of
shortfall, post-marketing stud- interim or final reports (planned or
ies are therefore a critical tool for actual)” and optional “milestones”
gathering the much-needed fol- or separate information on “sta-
low-up data, as well as allowing tus”. This sub-chapter represents
for additional evidence synthesis ‘additional pharmacovigilance ac-
efforts to inform appropriate use tivities’, a regulatory term that en-
of such products [21,22]. In order compasses all pharmacovigilance
to better understand the character- activities not considered as routine,
istics of the post-marketing studies and can include clinical studies or
associated with ATMPs approval, non-interventional post-authori-
we examined the regulatory land- zation safety studies (more details
scape of post-marketing studies provided in Box 1). These activities
and performed a systematic review can be assigned at the time of mar-
of the European Public Assessment keting authorization to one of three
Reports (EPAR) that is describing categories that need to be followed
the evaluation of ATMPs autho- when implementing post-market-
rized via the centralized procedure. ing authorization studies on AT-
MPs. Category 1 is mandatory and
comprises post-marketing studies
that are imposed as conditions to
SEARCH STRATEGY AND the marketing authorization. These
SELECTION CRITERIA studies should provide key infor-
Data on post-marketing clinical mation to the benefit-risk profile
studies (planned and ongoing) of of the product. Category 2 is also
authorized ATMPs were extracted mandatory and entails specific ob-
from the respective specific Eu- ligations only in case of a condi-
ropean public assessment report tional marketing authorization or
(EPAR) and the corresponding a marketing authorization under
page in clinicaltrials.gov registry exceptional circumstances. Final-
for each product. The EPAR search ly, any other studies for investi-
was based on the “find medicine” gating a specific safety concern or
search function on the EMA web- evaluating the effectiveness of risk
site www.ema.europa.eu/en/med- minimization activities fall under
icines. The cut-off-date for data category 3. Category 3 comprises
entry is November 1, 2019. Rele- activities which are conducted or
vant data were extracted from the financed by the MAH for investi-
pharmacovigilance plan that, upon gating specific safety concerns, but
marketing authorization, each mar- ‘do not include studies which are
keting authorization holder MAH imposed or which are specific ob-
has to provide within a detailed ligations’ (i.e. excluding categories
Risk management plan. Detailed 1 or 2) [23].
information on this can be found Categorization of each ATMP
within the EPAR, chapter “Risk to “Gene-“, “Somatic cell-“, or
Management Plan”, sub-chapter “Tissue-engineered” was based on
“Pharmacovigilance Plan”, which information extracted from the
includes a table describing “type EPAR of the respective authorized

ff BOX 1 intended to be used for the same

The regulatory framework for post-marketing studies. essential functions in the body. Tis-
Traditionally, upon marketing authorization of a new ATMP, the identified sue-engineered products contain
risks and mitigation strategies have to be outlined in a separate risk man- cells or tissues that have been mod-
agement plan (RMP), as defined in the general ‘Guideline on Good Pharma- ified so they can be used to repair,
covigilance Practices (GVP), Module V – Risk Management Systems (Rev2)’
and an additional ATMP specific guidance on follow-up and risk manage-
regenerate or replace human tissue
ment [57,58]. According to these guidelines, the marketing authorization [34,35].
holder should outline routine pharmacovigilance duties representing the The last published information
primary/minimum set of activities required to fulfil the legal requirements
on the post-marketing studies
contained in Directive 2001/83/EC [35]. However, due to the complex
nature of ATMPs, the assessment of potential long-term safety concerns was collected by searching Clin-
or lack of durable efficacy is of importance. These risks, such as germline icaltrials.gov database using the
transformation and vector transmission events or other genotoxic sce- tradename, international non-pro-
narios for genetically modified cell products, such as CAR-T cell products
challenge the traditional pharmacovigilance systems. Therefore, a special prietary name or, if available, clin-
focus is given on additional pharmacovigilance activities, as defined by the icalTrials.gov identifiers (Data-cut-
GVP guideline [57]. These activities should be performed in addition to off: November 1, 2019).
the routine pharmacovigilance activities and Regulation (EC) No 726/2004
[59], and thus are expected to lead to a more informed benefit-risk balance
[36]. If new safety data, indicating a substantial or potential risk, or new
efficacy data under real-life conditions become available during conducted
studies, follow-up measures have to be implemented accordingly and are CHARACTERISTICS
legally binding. OF POST-MARKETING
Article 14(2) of Regulation (EC) No 1394/2007 provides a specific frame-
work for RMPs related to ATMPs, which specifies how the MAH plans to
STUDIES FOR ATMPS
further characterize the safety and efficacy concerns. Usually, the MAH The results of our data extraction
addresses such additional activities by performing post-authorization safe- for the 10 EMA-aproved AT-
ty studies (PASS) and/or post-authorization efficacy studies (PAES). Both MPs are presented in Table 1. The
types of studies have to be in line with Directive 2001/83/EC [35] and
Regulation (EC) 726/2004 [59]. PASS usually aim to obtain further infor- composition of the post-authori-
mation on specifically identified safety concerns or to measure the effec- zation studies of ATMPs was an
tiveness of the designed risk-management measures. PASS can either be equal split between interventional
an interventional or observational study, with the latter typically relying on
“real-world data (RWD)” collected from registries, to aggregate and dis- studies (50%) and observational
seminate the long-term safety and efficacy data for ATMPs [25,26]. For studies (50%) (Figure 1). 35% of
instance, the cell therapy registries of the ‘European Society for Blood and the interventional studies includ-
Marrow Transplantation (EBMT)’ (EU) is one such registry that was modi-
ed were already ongoing at the
fied to satisfy the requirement to collect more long-term outcome data on
cellular therapies. The corresponding cellular therapy module of the EBMT time of marketing authorization
registry recently got a qualification opinion by the EMA Committee for Me- (MA), and the applicant would
dicinal Products for Human use (CHMP), describing the contexts in which be required to provide an update
EMA considers the use of registry data suitable [60]. PAES aims principally
to further evaluate the efficacy of the approved products in order to gain on the results of the studies, while
more evidence on long-term product efficacy. 15% were newly designed studies.
The newly planned interventional
trials have generally adopted de-
ATMP [24–33]. According to EMA signs that resemble pre-market tri-
classification, gene therapy prod- als—eg., using single-arm designs,
ucts function by inserting ‘recom- small sample sizes, or short-term
binant’ genes into the body, usu- follow-up periods with primary
ally to treat a variety of diseases, outcomes often focused on answer-
including genetic disorders, cancer ing hypothesis from pre-marketing
or long-term diseases. Somatic-cell scenarios, focusing on a narrow
therapy products contain cells or study population rather than test-
tissues that have been manipulat- ing real-world scenarios in a broad-
ed to change their biological char- en population (Table 1). Observa-
acteristics or cells or tissues not tional trials were either long term
1508 DOI: 10.18609/cgti.2019.156

CELL & GENE THERAPY INSIGHTS EXPERT INSIGHT
ff TABLE 1
Overview of additional pharmacovigilance activities in post-marketing settings for all currently marketed ATMPs within Europe.
Name of ATMP Pivotal Sample Follow-up Status at

Class of ATMP Category Study design Identifier Study objectives to fulfil PMAS requirement
(# of PMAS) (main study) size length time of MA
3 PASS n.a. X n.a. n.a. Safety (Long-term) Planned
Alofisel
(2) Phase III, randomized, double-blind, parallel-group, placebo-controlled, international, multi-
1 NCT03279081 X 600 52 weeks Safety (Long-term) Ongoing
Somatic center study (ADMIRE-CD-II)
cell therapy 3 Interventional, Paediatric Investigation Plan (PIP) trial (TK009) n.a. X n.a. n.a. Safety Planned
medicinal 3 Interventional, PIP trial (TK010) n.a. X n.a. n.a. Safety Planned
products Zalmoxis
(4) 2 Phase III, randomized, interventional, open-label, clinical trial (TK008) NCT00914628 X 170 12 months Safety Ongoing
Safety & effectiveness in real clinical practice
1 Non-Interventional PASS & PAES (TK011) n.a. X n.a. n.a. Planned
Safety & efficacy (long-term)
Safety (long-term)
3 Registry study, observational, prospective cohort [Patient registry] NCT02173171 X 340 3 years Ongoing
Efficacy (long term)
Post-marketing, prospective cohort study among patients treated in daily routine clinical Safety (long-term)
3 NCT02910557 X 920 5 years Planned
Imlygic practice Efficacy (long term)
(5) Biodistribution and Shedding of talimogene
3 Phase II, Interventional, open-label, multicenter, single-arm trial (single group assignment) NCT02014441 X 61 60 months Ongoing
laherparepvec deoxyribonucleic acid (DNA)
3 Phase I, multicenter, open-label, dose de-escalation study (single group assignment) NCT02756845 X 18 24 months Safety in pediatric subjects Planned
3 Randomized, controlled study n.a. X n.a. n.a. Safety in pediatric subjects Planned
Non-interventional, study (CCTL019B2401) with secondary use of data from two registries
1 n.a. X n.a. 15 years Safety (Long-term) Planned
conducted by EBMT and CIBMTR
Efficacy and Safety (Real-world evidence
1 Post-authorization efficacy study based on CCTL019B2401 observational registry study n.a. X n.a. n.a. data) in paediatric patients (< 3 years with Planned
B-ALL) treated in a commercial setting
Kymriah
Efficacy follow-up in r/r diffuse large B-cell
(5)
1 Prospective, observational PAES study in DLBCL(C2201) n.a. 🗸 n.a. n.a. lymphoma (DLBCL), patients evaluated in Ongoing
study C2201
1 Study CCTL019H2301, a randomized open-label parallel-group multicenter Phase III trial NCT03570892 X 318 5 years Efficacy Planned
Study CCTL019A2205B, Long-term follow-up, observational, registry, non-randomized,
Gene 3 NCT02445222 X 1250 15 years Safety (Long-term) Ongoing
open-label, single group assignment
therapy Post-authorization, multicenter, multinational, longitudinal, single-group, prospective, obser-
medicinal 1 n.a. X n.a. n.a. Safety (Long-term) Planned
Luxterna vational, safety registry study (SPKRPE-EUPASS)
products
(2) Observational, multi-site, non-randomized, prospective cohort, long-term safety and efficacy
1 NCT03602820 X 41 15 years Safety (Long-term), Efficacy (Long-term) Ongoing
follow-up study (LTFU-01)
Long-Term, Prospective, Non-Interventional Follow-up of Safety and Efficacy, patient registry
1 NCT03478670 X 50 15 years Safety (Long-term) Ongoing
study (200195)
3 Long-term follow up of patients from study AD1115611 NCT00598481 🗸 18 4 to 8 years Safety (Long-term) Ongoing
Strimvelis
Effectiveness of additional risk minimization
(4) 3 Surveys to HCPs/PIDs and parents/carers of pediatric patients n.a. X n.a. n.a. Planned
measures (e.g. educational materials)
RIS analysis to predict malignancy due to
3 Post-marketing approval methodology study n.a. X n.a. n.a. Planned
insertional oncogenesis
1 Non-Interventional Registry Study (PASS) n.a. X n.a. n.a. Safety Planned
3 Prescriber survey n.a. X n.a. n.a. Safety Planned
12 to 24 Safety,
3 Interventional, Phase I/II, multicenter, open-label study (single group assignment) (ZUMA-1) NCT02348216 🗸 290 Ongoing
months Efficacy
Yescarta
(9) Safety (Long-term),
3 Interventional, Phase II, multicenter, open-label study (single group assignment) (ZUMA-2) NCT02601313 X 105 15 years New indication: r/r Mantle Cell Lymphoma Ongoing
(MCL)
Safety
3 Interventional, Phase I/II, multicenter, open-label study (single group assignment) (ZUMA-3) NCT02614066 X 125 24 months Ongoing
(r/r Acute Lymphoblastic Leukemia (ALL))
The identified studies were colour coded to highlight their type as follows: Interventional studies planned at time of marketing authorization, Interventional studies ongoing at time of marketing authorization, Observational studies planned at time of marketing authorization, Observational studies ongoing at
ADMIRE-CD, Adipose-derived mesenchymal stem cells for induction of remission in perianal fistulizing Crohn’s disease; ALL, Acute lymphoblastic leukemia; ATMP, Advanced therapy medicinal product; CIBMTR, Center for International Blood & Marrow Transplant Research; CNS, Central nervous system; DLBCL,
Diffuse large B-cell lymphoma; DNA, Deoxyribonucleic acid; EBMT, European Society for Blood and Marrow Transplantation; HCP, Healthcare provider, LTFU, Long-term follow-up; MA, Marketing authorization; MCL, mantle cell lymphoma; MF, Microfracture; n.a., not applicable; NHL, Non-Hodgkin lymphoma; PAES,
Post-authorization efficacy study; PASS, Post-authorization safety study; PIP, Paediatric Investigation Plan; PMAS, Post-marketing authorization studies; RIS, Retroviral Insertion Site; r/r, relapsed/refractory; TDT, Transfusion-dependent β-thalassaemia.
1509 DOI: DOI: 10.18609/cgti.2019.156 Cell & Gene Therapy Insights - ISSN: 2059-7800 1509
CELL & GENE THERAPY INSIGHTS EXPERT INSIGHT
ff TABLE 1 (CONT.)
Overview of additional pharmacovigilance activities in post-marketing settings for all currently marketed ATMPs within Europe.
Name of ATMP Pivotal Sample Follow-up Status at

Class of ATMP Category Study design Identifier Study objectives to fulfil PMAS requirement
(# of PMAS) (main study) size length time of MA
Interventional, Phase I/II, multicenter, open-label study (single group assignment) Safety
3 NCT02625480 X 100 24 months Ongoing
(ZUMA-4) (pediatric r/r ALL patients)
Interventional, Phase II, multicenter, open-label study (single group assignment) Safety (Long-term)
Yescarta 3 NCT03105336 X 160 15 years Ongoing
(ZUMA-5) (r/r indolent NHL)
(9)
(CONT.) Interventional, Phase I/II, multicenter, open-label, supportive study Safety in combination with Atezolizumab (Long-term)
3 NCT02926833 X 37 5 years Ongoing
(single group assignment) (ZUMA-6) (refractory DLBCL)
Gene Interventional, Phase III, randomized, open-label, multicenter study
3 NCT03391466 X 359 5 years Safety (Long-term) Ongoing
therapy (parallel assignment) (ZUMA-7)
medicinal Safety (Long-term)
products 1 Long-term observational registry study (including product registry REG-501) n.a. X n.a. n.a. Planned
Efficacy (Long-term)
(CONT.)
Safety, Efficacy
Zynteglo 2 Single-arm interventional, phase III, open-label study (HGB-207) NCT02906202 🗸 23 24 months (transfusion-dependent ß-Thalassemia (TDT) Ongoing
(4) patients without ß0/ß0 genotype)
Safety, Efficacy
2 Single-arm interventional, phase III, open-label study (HGB-212) NCT03207009 🗸 15 24 months Ongoing
(TDT patients with ß0/ß0 genotype)
2 Prospective, observational case-only, long-term follow-up study (LTF-303) NCT02633943 🗸 94 15 years Safety (Long-term), Efficacy (Long-term) Ongoing
2 Multinational, multicentre, prospective, open-label, uncontrolled study (HLSTM03) n.a. X n.a. n.a. Safety (Long-term), Efficacy (Long-term) Planned
Long-term safety and efficacy follow-up study connected to HLSTM03 Safety (Long-term), Efficacy (Long-term)
Holoclar 3 n.a. X n.a. n.a. Planned
(HLSTM03FU) Success after keratoplasty
(3)
Tissue- Post-authorization observational, patient registry study during routine clinical
3 n.a. X n.a. 5 years Safety (Long-term) under routine clinical conditions Planned
engineered practice
products (TEP) Safety (Long-term), Efficacy (Long-term)
n.a. Prospective, randomized, open-label, multicentre Phase III clinical trial (cod 16HS13) NCT01222559 🗸 102 60 months Ongoing
Spherox (Compare to active comparator ‘microfracture’ (MF))
(2) Safety (Long-term), Efficacy (Long-term)
n.a. Prospective, randomized, open-label, multicentre Phase II clinical trial (cod 16HS14) NCT01225575 🗸 75 60 months Ongoing
(Compare three different doses)
The identified studies were colour coded to highlight their type as follows: Interventional studies planned at time of marketing authorization, Interventional studies ongoing at time of marketing authorization, Observational studies planned at time of marketing authorization, Observational studies ongoing at
ADMIRE-CD, Adipose-derived mesenchymal stem cells for induction of remission in perianal fistulizing Crohn’s disease; ALL, Acute lymphoblastic leukemia; ATMP, Advanced therapy medicinal product; CIBMTR, Center for International Blood & Marrow Transplant Research; CNS, Central nervous system; DLBCL,
Diffuse large B-cell lymphoma; DNA, Deoxyribonucleic acid; EBMT, European Society for Blood and Marrow Transplantation; HCP, Healthcare provider, LTFU, Long-term follow-up; MA, Marketing authorization; MCL, mantle cell lymphoma; MF, Microfracture; n.a., not applicable; NHL, Non-Hodgkin lymphoma;
PAES, Post-authorization efficacy study; PASS, Post-authorization safety study; PIP, Paediatric Investigation Plan; PMAS, Post-marketing authorization studies; RIS, Retroviral Insertion Site; r/r, relapsed/refractory; TDT, Transfusion-dependent β-thalassaemia.
1510 DOI: DOI: 10.18609/cgti.2019.156 Cell & Gene Therapy Insights - ISSN: 2059-7800 1510
EXPERT INSIGHT
follow up of the patients treated ffFIGURE 1

in premarketing studies or mostly Percentage of post-marketing authorization study types at MA for all
a registry study to collect data on marketed ATMP.
particular safety and efficacy pa-
rameters in real-world settings.
The number of total post-mar-
keting studies ranged from two to
nine across the 10 ATMPs (Fig-
ure 2). The sample size of the trials
ranged from 15 to 1250 patients
(Table 2), with most of the tri-
als intending to enrol fewer than
200 patients (Figure 3). Duration
of follow-up in trials ranged from
12 to 180 months (Table 2), which
is mostly dependent on the use
of viral vectors in the treatment
since these products require lon-
ger follow-up for safety-related
ffFIGURE 2
Number and types of post-marketing authorization studies submitted by each applicant.
ATMPs were categorized according to the regulatory class as follows:

(1) Cell therapy medicinal product (CTMP)
(2) Gene therapy medicinal product (GTMP),
(3) Tissue-engineered product (TEP).

ff TABLE 2. parameters. Half of applicants

Summary statistics of post-marketing authorization design pa- (50%) included Phase III interven-
rameters among ATMPs. tional trial designs that tradition-
Sample size Follow-up (months) ally focus on product efficacy as a
Valid 23 25 primary endpoint and usually form
N
Missing 2 0 the basis of the regulatory submis-
Mean 229,17 80,44 sions and authorization (Figure 4).
Median 102,00 60,00 Most of these trials (86%) were
Std. Deviation 310,228 65,775 initiated before the MA but only a
Minimum 15 12 few of them represent pivotal trials
Maximum 1250 180 upon which the MA was acquired
and the rest continued to perform
ffFIGURE 3 these trials as part of the post-mar-
Frequencies of sample sizes specified in the post-marketing studies of keting surveillance phase which
ATMPs. would have traditionally been in-
cluded in a MA submission pack-
age (Table 1).
CURRENT POST-
MARKETING TRIALS FOR
ATMPS RATHER ATTAIN
PRE-MARKETING DESIGNS
Our analysis suggests that a high
degree of variability between trial
designs of ATMPs in post-market-
ing settings cthat can be explained
by the wide range in evidence gen-
erated from clinical studies at the
time of MAA, the rarity of the in-
dication and the specific character-
istics of the product itself (whether
ffFIGURE 4
Frequency of study phases submitted as post-marketing authorization studies.
1512 DOI: 10.18609/cgti.2019.156

EXPERT INSIGHT
it is genetically modified or not). ffFIGURE 5

This inherited variability requires Frequencies of follow-up periods specified in the post-marketing studies of
developers to devise a post-mar- ATMPs.
keting strategy based on a case-by-
case scenario. Moreover, many of
the post-marketing trials, intended
to address the critical benefit-risk
knowledge gaps for ATMPs, are
phase III interventional trials that
may not be well suited to inform
routine clinical use of the products.
The sample sizes of post-marketing
trials are relatively small for studies
that should aim to reflect real-life
situations. Further analysis of the
data also showed high variability in
the follow-up periods specified for
each study (Figure 5). These obser-
vations add the challenge of iden-
tifying a clear cut-off between pre-
and post-marketing studies and the
exact added value of PMAS.
The dominant trial paradigm
for decades has been explanatory wider population?” Addressing this
in its orientation—which is to say question requires a more pragmat-
that trials have been designed to ic orientation to trial design, which
test experimental interventions un- often involves more flexible proto-
der “idealized” or “laboratory-like” cols, broader and more heteroge-
conditions that are optimized to neous patient populations, and a
detect a treatment effect. Explana- mix of different healthcare settings.
tory trial designs thus generally in- For instance, early evidence of the
volve strict intervention protocols, real-world efficacy performance of
many patient inclusion/exclusion Yescarta, a CAR T cell-based prod-
criteria, and high-resourced settings uct used in treatment of non-Hod-
in an effort to control for system- gkin lymphoma (NHL), showed
ic errors (e.g. confounding, bias) that the efficacy signals generated
and deliver statistically credible re- once the product was used in clin-
sults of high internal validity [36]. ical practice, applied to more het-
In the pre-market setting, this ap- erogeneous population are slightly
proach to trial design makes good inferior to that generated from the
sense, since the primary question clinical trial [37].
from a regulatory perspective may This raises the question of
be of the form: How will this new whether it is more sensible to
intervention work for this particular continue reproducing data in
patient in a controlled setting? By post-marketing settings within the
contrast, in the post-market setting, framework of phase III-type trials
the primary question is a different or rather introduce more flexible
form: “Can this new intervention designs that would account for re-
be beneficial once available to the al-world heterogeneity.

PRAGMATIC CONCEPTS, pregnant women) and patients with

A PROMISING OPTION comorbidities (e.g., neurological or
TO ENHANCE POST- hematological disorder, autoim-
MARKETING TRIALS mune disease or infections) [40].
More heterogenic outcome data
Pragmatic trials are meant to in-
would be of particular value for
form a clinical or policy decision
these products, due to commonly
by evaluating the effectiveness of
underlying heterogenic baseline
interventions in real-world clini-
parameters immanent to the types
cal practice [38]. We believe that
of rare diseases that are frequent-
the distinction between explana-
ly of interest. Thus, it seems that
tory and pragmatic trial concepts
more pragmatic trials would be
may be valuable here when de-
suitable for ATMP post-market-
signing post-marketing trials that
ing safety and efficacy surveillance.
are needed to ensure safe and re-
Although the data from pragmatic
liable use of ATMPs and fulfil the
trials is noisier, it can nevertheless
requirements of multiple stake-
provide a more representative pic-
holders (regulators, HTA bodies,
ture of whether an intervention ac-
payers etc.) [39].
tually has utility in clinical practice
For ATMPs, the major draw-
(or how its utility may vary from
backs of an explanatory orientation
one set of conditions to the next).
are the small sample size due to low
However, when characterizing
incidence levels, a short time frame
explanatory and pragmatic trial
of observation, limited or com-
concepts, it is crucial to observe
plete exclusion of distinct patient
that the explanatory/pragmatic
population such as vulnerable pop-
distinction is not a dichotomy, but
ulations (e.g., children, elderly and
a multi-dimensional continuum.
In recognition of this point, Thor-
ff BOX 2 pe et al. implemented the PRag-
The Pragmatic-Explanatory Continuum Indicator Summary 2 matic-Explanatory Continuum In-
(PRECIS-2) wheel (adapted from Loudon et al., 2015 [51]). dicator Summary (PRECIS) tool,
which breaks down a trials prag-
1. “Eligibility” domain should describe who is selected to participate in matism (or lack thereof ) along 10
the trial.
different dimensions as described
2. “Recruitment” domain describes how are participants recruited into in their published report [41].
the trial.
As a result of subsequent ex-
3. “Setting” domain includes information on where is the trial being done. tensive discussion on the concept
4. “Organisation” domain provides information on what kind of expertise of pragmatism within clinical re-
and resources are needed to deliver the intervention.
search methodology [40,42–50],
5. “Flexibility: delivery” domain describes how the intervention should the requirements for characteriz-
be delivered.
ing a study as pragmatic trial were
6. “Flexibility: adherence” domain comprises information on what optimized and validated, leading
measures are in place to make sure participants adhere to the
intervention. to the implementation of PRE-
CIS-2 [51]. This improved tool
7. “Follow up” domain describes how closely are participants followed-up.
represents a nine-spoked ‘wheel’
8. “Primary outcome” domain provides information on how relevant
with nine domains based on trial
findings are to participants
design decisions. The features of
9. “Primary analysis” domain should summarize to what extent all data
are included.
PRECIS-2 were summarized in
Box 2.
1514 DOI: 10.18609/cgti.2019.156

EXPERT INSIGHT
AN EXEMPLARY further exploration of the effects of

APPLICATION OF different treatment schemes on the
PRAGMATIC TRIAL product outcome, and thereby fill-
DESIGN TO CAR-T CELL ing some of the critical information
THERAPEUTICS gaps for distinct patient popula-
tions, e.g. co-morbidities, variable
To concretely illustrate how we
pre-treatment or age groups. The
believe pragmatic, post-market-
EBMT registry could be useful here
ing trials for ATMPs should be
as well since it enables entering stan-
designed, we applied a PRECIS-2
dardized data that can be analyzed
analysis to hypothetical trials for
in larger post-marketing studies.
the two marketed CAR-T cell ther-
apies, tisagenlecleucel (Kymriah®,
Novartis) and axicabtagene cilo-
leucel (Yescarta®, Kite Pharma/Gil- Organization
ead). In what follows, we discuss
To deliver clear information on
particular PRECIS-2’s dimensions
what kind of expertise and resources
that are relevant to our case study
are needed to deliver the interven-
and how we believe trials of Kymri-
tion of interest, current approach-
ah and Yescarta should be oriented
es to post-marketing trials tend to
along the pragmatic spectrum.
focus on interventions being only
performed by physicians/hospitals
Eligibility criteria that are specially trained within the
control distribution program as a
Current post-marketing studies risk minimization measure. Given
for Yescarta and Kymriah relied the risks associated with Kymriah
on restricted enrollment according and Yescarta, we believe that this is
to the authorized indication with the right approach and that a more
deliberate consideration of contra- pragmatic design along this dimen-
indications, special warnings, and sion is not necessary.
precautions for use (see Annex I,
Summary of product characteris-
tics) [52,53]. An exception to this Flexibility of delivery
was given in case of enrolling spe-
cial patient populations not covered Relating to mechanisms on how
at marketing authorization, there- to deliver the intervention, cur-
by addressing missing information rent post-marketing concepts for
according to the RMP (e.g. use in Kymriah and Yescarta CAR-T cell
HBV/HCV/HIV infection, use in products focus on administration
patients with active Central Ner- only within the authorized dosing
vous System (CNS) involvement in regimen. However, clinicians may
malignancy). However, to be more often have to adjust dosing in prac-
pragmatic and better inform clini- tice, for example, for patients with
cal use, we believe that there should low baseline T-cell concentrations
be greater flexibility in the eligibil- in leukapheresis starting materials.
ity criteria, for example, allow for A more pragmatic approach would
patients with different schemes of thus include administration of
pre-conditioning lymphodepleting products that do not meet the com-
therapies. This flexibility will enable mercial specifications. However, an

important pre-requisite to allow engine’ in the long-run. This has the

for this kind of approach would be potential to enhance patient compli-
that no overwhelming safety con- ance and commitment for a longer
cerns have been identified during follow up period and promote pa-
manufacture and release of such tient consent to register data in the
out-of-specification (OSS) product. EBMT registry and use it, for in-
This would enable a deeper inves- stance, as a source of external con-
tigation of the dose-response rela- trol data for comparative purposes.
tionship for OSS concentration lev-
els of the CAR-T cell product. For
instance, the authorized dosage of Primary outcome
Kymriah is 0.2-5x106 CD19+ CAR
T-cells (body weight-based), with The current post-marketing trials
a maximum dose of 2.5x108 CAR for Kymriah and Yescarta aim at
T-cells (non-body weight-based). further characterization of safety
However, even lower dose ranges profiles, specifically related to Cy-
(e.g. ≤0.03 x 106) in acute lympho- tokine Release Storm (CRS), neu-
blastic lymphoma trials showed a rotoxicity, infections, prolonged cy-
clinical response [54]. topenias, growth and development,
reproductive status and pregnancy
outcomes. Some trials aim to char-
Flexibility of adherence acterize further the efficacy profile
related to Overall Response Rate
A pragmatic approach here would (ORR), CD19 CAR T-cell level,
ideally deviate as little as possible incidence/exacerbation of pre-exist-
from standard practice by avoid- ing comorbidities, relapse/progress
ing highly stringent time frames for disease, incidence death and moni-
study visits, thus increasing flexi- toring of replication-competent len-
bility and patient adherence. Less tivirus. A pragmatic approach along
stringent follow-up visits may be this dimension could apply a more
also motivating for practitioners, in concise set of outcomes, particularly
reducing the associated monitoring the ones that interfere with patients’
and workload without jeopardizing daily productivity and quality of life
patients safety. while keeping additional tests or vis-
its to a minimum. In order for the
pragmatic approach to provide add-
Follow up ed value outside of the controlled
environment of an explanatory trial,
The procedure for follow up under it should also incorporate relevant
current post-marketing settings re- patient decision-making criteria to
lies on an evaluation period of up to provide meaningful evidence and
15 years for safety and efficacy sur- build upon the evidence generated at
veillance after CAR-T cell adminis- the time of MAA.
tration. When applying pragmatic
trial concepts in the post-marketing
settings, post-market studies need to Primary analysis
interface with the patient registries
and rely on them, which in turn will There was no access to detailed sta-
be the most pragmatic ‘information tistical analysis plans to evaluate
1516 DOI: 10.18609/cgti.2019.156

EXPERT INSIGHT
primary analysis mechanisms under hypotheses that a more suitable to

the current conventional post-mar- pre-market trials. Thus, we believe
keting setting for Kymriah and Yes- that there is room for improvement
carta. For a pragmatic trial approach in terms of designing (or mandating)
here, including a heterogeneous post-market trials of ATMPs that
patient population and planning will better meet the information-
for sub-group analyses would al- al needs of patients, clinicians, and
low for detecting clinically-relevant payers.
safety and efficacy signals. Indeed, We have suggested that applying
given the limited data that is often the tools of pragmatic trial design
available for ATMPs, we would rec- (as made explicit by the PRECIS-2
ommend a maximally pragmatic framework) may help to fill this
approach to the primary analysis of gap. In particular, RWD generated
post-marketing trials to try and in- from registry-based pragmatic trials
clude as much patient-relevant data would have great potential to gen-
as possible. eralize findings and better inform
the use of ATMPs across diverse
patient populations. However, there
are certainly challenges regarding
CONCLUDING data quality, as patient populations
REMARKS AND FUTURE will become more divergent, thus
CONSIDERATIONS increasing the risk for confounding
While the new regulatory tools de- and a higher degree of diffuse data
veloped by EMA to facilitate rapid generated. On the other hand, the
marketing authorization in cases aggregation of clinical data collected
of major public health interests are from RWD can increase the robust-
warranted, these tools must be ac- ness of meta-analyses derived from
companied by sound post-autho- post-marketing studies. In a recent
rization strategies to generate long report, regulators from the Swedish
term evidence for safety and efficacy. Medical Products Agency called for
The current regulatory landscape more attention to methodological
for conducting post-authorization basics of post-marketing studies
studies is very complex and demands that can help generate reliable re-
an enormous effort from the MAH sults and affirmed the regulatory
to navigate. As we have shown, value of the pragmatic trial concept
there is a wide variety in post-mar- regardless of labelling the studies as
ket study designs for ATMPs, both ‘pragmatic’ or ‘real-world evidence’
for studies that are required by the [55]. Moreover, the field of ATMP
regulatory authorities and for those may learn from medical device eval-
studies conducted voluntarily by the uations, where regulatory agencies
MAH to investigate a specific safety already implemented guidance on
concern or to evaluate the efficacy/ how to apply real-world data for
effectiveness of risk minimization regulatory decision-making [56].
activities (classified in category 1, 2 While all aspects of the approach
and 3). However, within that variety, of a pragmatic trial do not need to
we observed that many post-mar- be implemented at once, working on
keting surveillance trials for ATMPs improving the current post-market-
adopt explanatory trial design fea- ing study methodology and supple-
tures and are focusing on answering menting it with new ideas can help

to enhance the current practice (Box health insurance providers, to obtain

3). We believe that such an approach real-world data from clinical routine
would not only be beneficial for faster after marketing authorization
MAHs but also for regulators and for ATMPs.
ff BOX 3
Considerations for using pragmatic trial concepts for
post-authorization safety and efficacy studies (both
observatonal and interventional).
1. Closing the evidence gap between Early Access Program (EAP) mediated marketing authorization for innovative
medicines and the lack of data on their long-term safety and efficacy at the time of marketing authorization
2. Gaining data on a more heterogeneous study population and none or only partially addressed risks at an earlier
stage after marketing authorization in a setting more closer to real-time and real-life scenarios
3. Implementation of patient broad consent to share data among registry studies
4. Reducing enormous timely and regulatory effort by submitting more broaden study protocols for pragmatic trial
instead of numerous single protocols for individual conventional explanatory studies
5. Pooling of study objectives in combination with implementing patient registry to enter standardized data sets and
share them for statistical analysis in sub-groups
6. Addressing pharmacovigilance requirements of GVP module V, referred to in the post-authorization development
plan as part of the risk management plan
7. Conferring economic benefits for MAH, regulators, and health insurance providers by accelarating the process
for obtaining real-world data from clinical routine upon marketing access
8. Allowing for taking into account more/all variabilities on patient and product level (e.g. patient baseline status,
product design) to potentially enabling optimization of patient access

for this version to be published. Fritsche E, Elsallab M and Schaden M contributed equally as authors on this article. Hey SP
and Abou-El-Enein M contributed equally as senior authors.
Disclosure and potential conflicts of interest: The authors declare that they have no conflicts of interest.
Funding declaration: MS received funding from the Erasmus SMT Mobility Program during her work on this project.
ME received funding from the Arab-German Young Academy of Sciences and Humanities (AGYA), a project of the Ber-
lin-Brandenburg Academy of Sciences and Humanities, funded by the Federal Ministry of Education and Research (BMBF).
The project was partially funded from the European Union’s Horizon 2020 research and innovation programme under grant
agreement No. 825392 (Reshape) and No. 820292 (Restore)

Attribution: Copyright © 2019 Fritsche E, Elsallab M, Schaden M, Hey SP, Abou-El-Enein M. Published by Cell and Gene
Therapy Insights under Creative Commons License Deed CC BY NC ND 4.0.
Submitted for peer review: Oct 2 2019; Revised manuscript received: Nov 2 2019; Publication date: Nov 18 2019.
1518 DOI: 10.18609/cgti.2019.156

EXPERT INSIGHT
REFERENCES 9. European Medicines Agency (EMA). Disease Innovation, Discovery, and

Development support and regulato- Drug Development. J. Gen. Intern.
1. Elsanhoury A, Sanzenbacher R, Rein-
ry tools for early access to medicines. Med. 2014; 29(S3), 804–807.
ke P, Abou-El-Enein M. Accelerating
2016; 44, 1–4.
Patients’ Access to Advanced Ther- 18. ten Ham RMT, Hoekman J, Hövels
apies in the EU. Mol. Ther. Methods 10. M. Abou-El-Enein, Elsanhoury A, AM, Broekmans AW, Leufkens
Clin. Dev. 2017; 7, 15–1. Reinke P. Overcoming Challenges HGM, Klungel OH. Challenges in
Facing Advanced Therapies in the EU Advanced Therapy Medicinal Product
2. Neez E, Hwang T, Sahoo SA, Naci H,
Market. Cell Stem Cell. 2019; 9(3), Development: A Survey among Com-
European Medicines Agency’s Prior-
293–297. panies in Europe. Mol. Ther. - Methods
ity Medicines (PRIME) scheme at 2
Clin. Dev. 2018; 11, 121–130.
years: An evaluation of clinical studies 11. M. Abou-El-Enein and Hey SP. Cell
supporting eligible drugs, Clin. Phar- and Gene Therapy Trials: Are We 19. K. Malliaras and Marbán E. Moving
macol. Ther. 2019; 1669. Facing an ‘Evidence Crisis’?. EClini- Beyond Surrogate Endpoints in Cell
calMedicine. 2019; 7, 13–14 . Therapy Trials for Heart Disease. Stem
3. Mullard A. PRIME time at the EMA,
Cells Transl. Med. 2014; 3(1), 2–6.
Nat. Rev. Drug Discov. 2017; 16(4) 12. Heneghan C, Goldacre B, Mahtani
226–228. KR. Why clinical trial outcomes fail 20. Kesselheim AS. Characteristics of
to translate into benefits for patients. Clinical Trials to Support Approval
4. European Medicines Agency (EMA),
Trials 2017; 18(1) 122. of Orphan vs Nonorphan Drugs for
Enhanced early dialogue to facili-
Cancer. JAMA 2011; 305(22), 2320.
tate accelerated assessment of PRI- 13. Institute of Medicine (US) Forum on
ority Medicines ( PRIME ), EMA/ Drug Discovery; Development and 21. S. de Wilde et al. EU decision-making
CHMP/57760/2015, Rev. 1. 2018; Translation. Chapter 3: Challenges for marketing authorization of ad-
44, 1–12. in Clinical Research. in Transforming vanced therapy medicinal products: a
Clinical Research in the United States: case study. Drug Discov. Today. 2018;
5. Naci H, Smalley KR, and Kesselheim
Challenges and Opportunities: Work- 23(7), 1328–1333.
AS. Characteristics of Preapproval
shop Summary., Washington, D.C.:
and Postapproval Studies for Drugs 22. G. Detela and Lodge A. EU Regula-
National Academies Press, 2010.
Granted Accelerated Approval by the tory Pathways for ATMPs: Standard,
US Food and Drug Administration, 14. Khozin S, Blumenthal GM, and Accelerated and Adaptive Pathways
JAMA. 2017; 318(7), 626. Pazdur R. Real-world Data for Clin- to Marketing Authorisation. Mol.
ical Evidence Generation in Oncol- Ther. - Methods Clin. Dev.. 2019; 13,
6. Pease AM, Krumholz HM, Downing
ogy. JNCI J. Natl Cancer Inst. 2017; 205–232.
NS, Aminawung JA, Shah ND, Ross
109(11).
JS. Postapproval studies of drugs ini- 23. European Medicines Agency (EMA).
tially approved by the FDA on the 15. Booth CM, Karim S, and Mackillop Guidance on format of the risk man-
basis of limited evidence: systematic WJ. Real-world data: towards achiev- agement plan (RMP) in the EU part
review, BMJ 2017; 1680. ing the achievable in cancer care. III: Pharmacovigilance Plan. no.
Nat. Rev. Clin. Oncol. 2019; 16(5), EMA/714913/2012, pp. 1–6, 2012.
7. European Medicines Agen-
312–325.
cy (EMA). PRIME: a two- 24. Committee for Medicinal Prod-
year overview,. 2018. [Online]. 16. M. Abou-El-Enein et al. Good Man- ucts for Human Use (CHMP).
Available: https://www.ema.eu- ufacturing Practices (GMP) manufac- European Public Assessment Re-
ro p a . e u / e n / d o c u m e n t s / re p o r t / turing of advanced therapy medicinal port (EPAR): Alofisel. EMA/
prime-two-year-overview_en.pdf. products: a novel tailored model for CHMP/64055/2018. 2017. [Online].
optimizing performance and estimat- Available: https://www.ema.europa.
8. European Medicines Agency (EMA).
ing costs. Cytotherapy 2013; 15(3), eu/en/documents/assessment-report/
Early Access to Medicines Scheme
362–383. alofisel-epar-public-assessment-re-
guidance. Hum. Med. Res. Dev. Sup-
port_en.pdf.
port Div. 17. Pariser AR, Gahl WA. Important
Role of Translational Science in Rare

25. Committee for Medicinal Products for 30. Committee for Medicinal Prod- Off. J. Eur. Union. 2007; L(324),
Human Use (CHMP). European Pub- ucts for Human Use (CHMP). 121–137.
lic Assessment Report (EPAR): Zal- European Public Assessment
35. European Commission. EU Direc-
moxis. EMA/CHMP/589978/2016. Report (EPAR): Holoclar.
tive 2001/83/EC. 2013. [Online].
2016. [Online]. EMA/25273/2015. 2014. [Online].
Available: https://ec.europa.eu/
Available: https://www.ema.europa. Available: https://www.ema.europa.
health/sites/health/files/files/eudralex/
eu/en/documents/assessment-report/ eu/en/documents/assessment-report/
vol-1/dir_2001_83_consol_2012/
zalmoxis-epar-public-assessment-re- holoclar-epar-public-assessment-re-
dir_2001_83_cons_2012_en.pdf.
port_en.pdf. port_en.pdf.
36. M. Abou-El-Enein, Duda GN,
26. Committee for Medicinal Prod- 31. Committee for Medicinal Prod-
Gruskin EA, and Grainger DW. Strat-
ucts for Human Use (CHMP). ucts for Human Use (CHMP).
egies for Derisking Translational Pro-
European Public Assess- European Public Assessment
cesses for Biomedical Technologies.
ment Report (EPAR): Imlygic. Report (EPAR): Spherox.
Trends in Biotechnology 2017; 35(2),
EMA/734400/2015. 2015. [Online]. EMA/349863/2017. 2017. [Online].
100–108.
Available: https://www.ema.europa. Available: https://www.ema.europa.
eu/en/documents/assessment-report/ eu/en/documents/assessment-report/ 37. C. A. Jacobson et al. Axicabtagene
imlygic-epar-public-assessment-re- spherox-epar-public-assessment-re- Ciloleucel in the Real World: Out-
port_en.pdf. port_en.pdf. comes and Predictors of Response,
Resistance and Toxicity. in ASH An-
27. Committee for Medicinal Products for 32. European Medicines Agen-
nual Meeting, 2018.
Human Use (CHMP). European Pub- cy (EMA). European Public As-
lic Assessment Report (EPAR): Lux- sessment Report (EPAR) Ky- 38. I. Ford and Norrie J. Pragmatic Tri-
turna. EMA/CHMP/700911/2018. mriah (EMA/485563/2018), als. N. Engl. J. Med. 2016; 375(5),
2018. [Online]. Committee for Medicinal Products 454–463.
Available: https://www.ema.europa. for Human Use (CHMP): As-
eu/en/documents/assessment-report/ sessment report. 2018. [Online]. 39. D. Schwartz and Lellouch J. Explana-
luxturna-epar-public-assessment-re- Available: https://www.ema.europa. tory and Pragmatic Attitudes in Ther-
port_en.pdf. eu/en/documents/assessment-report/ apeutical Trials. J. Clin. Epidemiol..
kymriah-epar-public-assessment-re- 2009; 62(5), 499–505.
28. Committee for Medicinal Products for
port_en.pdf.
Human Use (CHMP). European Pub- 40. Patsopoulos NA. A pragmatic view on
lic Assessment Report (EPAR): Strim- 33. European Medicines Agency pragmatic trials. Dialogues Clin. Neu-
velis. EMA/CHMP/272303/2016Rev (EMA). European Public Assess- rosci.. 2011; 13(2), 217–24.
1. 2016. [Online]. ment Report (EPAR) Yescarta
41. Thorpe KE et al. A pragmatic–explan-
Available: https://www.ema.europa. (EMA/481168/2018), Commit-
atory continuum indicator summary
eu/en/documents/assessment-report/ tee for Medicinal Products for
(PRECIS): a tool to help trial design-
strimvelis-epar-public-assessment-re- Human Use (CHMP): Assess-
ers. J. Clin. Epidemiol.. 2009; 62(5),
port_en.pdf. ment report. 2018. [Online].
464–475.
Available: https://www.ema.europa.
29. Committee for Medicinal Products for
eu/en/documents/assessment-report/ 42. Anderson ML et al. The Food and
Human Use (CHMP). European Pub-
yescarta-epar-public-assessment-re- Drug Administration and pragmat-
lic Assessment Report (EPAR): Zyn-
port_en.pdf. ic clinical trials of marketed medical
teglo. EMA/CHMP/226273/2019.
products. Clin. Trials J. Soc. Clin. Tri-
2019. [Online]. 34. The European Parliament and the
als. 2015; 12, no. 5, 511–519.
Available: https://www.ema.europa. Council of the European Union.
eu/en/documents/assessment-report/ Regulation (EC) No 1394/2007 on 43. M. Roland and Torgerson DJ. Un-
zynteglo-epar-public-assessment-re- advanced therapy medicinal products derstanding controlled trials: What
port_en.pdf. and amending Directive 2001/83/EC are pragmatic trials?. BMJ. 1998;
and Regulation (EC) No 726/2004. 316(7127), 285–285.
1520 DOI: 10.18609/cgti.2019.156

EXPERT INSIGHT
44. Godwin M et al. Pragmatic controlled product-information/yescarta-ep- AFFILIATIONS

clinical trials in primary care: the ar-product-information_en.pdf.
Enrico Fritsche
struggle between external and inter-
54. D. W. Lee et al. T cells expressing Berlin Institute of Health (BIH)
nal validity. BMC Med. Res. Methodol.
CD19 chimeric antigen receptors Center for Regenerative Therapies
2003; 3(1), 28.
for acute lymphoblastic leukaemia in (BCRT), Charité–Universitäts-
45. Ernst E. Limitations of ‘pragmatic’ tri- children and young adults: a phase medizin Berlin, Berlin, Germany
als. Postgrad. Med. J.. 2005; 81(954), 1 dose-escalation trial. Lancet. 2015; Magdi Elsallab
203–203. 385(9967), 517–528. Berlin Institute of Health (BIH)
46. S. Treweek and Zwarenstein M. Mak- 55. Gedeborg R, Cline C, Zethelius B, Center for Regenerative Therapies
ing trials matter: pragmatic and ex- and Salmonson T. Pragmatic clinical (BCRT), Charité–Universitäts-
planatory trials and the problem of trials in the context of regulation of medizin Berlin, Berlin, Germany
applicability. Trials 2009; 10(1), 37. medicines. Ups. J. Med. Sci.. 2019; Michaela Schaden
124(1), 37–41. Clinical Trials Coordination Centre,
47. D. M. Kent and Kitsios G. Against
Medical University of Vienna,
pragmatism: on efficacy, effectiveness 56. FDA, CDRH, and CBER. Use of
Vienna, Austria
and the real world. Trials 2009; 10(1), Real-World Evidence to Support Reg-
48. ulatory Decision-Making for Med- Spencer Phillips Hey
ical Devices Guidance for Industry Program on Regulation, Ther-
48. Eldridge S. Pragmatic trials in prima- apeutics, and Law (PORTAL),
and Food and Drug Administration
ry health care: what, when and how?. Brigham and Women’s Hospital,
Staff Preface Public Comment. p. 24,
Fam. Pract.. 2010; 27(6), 591–592. 1620 Tremont Street, Boston, MA
2016.
49. J. Sugarman and Califf RM. Eth- 02120, USA
57. European Medicines Agency and;
ics and Regulatory Complexities for
(EMA). Guideline on good phar- Center for Bioethics, Harvard
Pragmatic Clinical Trials. JAMA, vol.
macovigilance practices (GVP) Medical School, Boston, MA, USA
311, no. 23, p. 2381, Jun. 2014.
Module V – Risk management
systems (Rev 2). 2017. [Online]. Mohamed Abou-El-Enein
50. Ratner J, Mullins D, Buesching DP,
Available: www.ema.europa.eu/docs/ Author for correspondence,
and Cantrell RA. Pragmatic clinical
en_GB/document_library/Scientific_ Clinical Development Platform,
trials: U.S. payers’ views on their val-
guideline/2012/06/WC500129134. BIH Center for Regenerative
ue. Am. J. Manag. Care. 2013; 19(5),
pdf. Therapies (BCRT), Charité Uni-
e158–65.
versitatsmedizin Berlin, Corpo-
51. Loudon K, Treweek S, Sullivan F, 58. EMEA/CHMP. Guideline on safety rate Member of Freie Universität
Donnan P, Thorpe KE, and Zwaren- and efficacy follow-up and risk man- Berlin, Humboldt-Universität
stein M. The PRECIS-2 tool: design- agement of Advanced Therapy Medic- zu Berlin, and Berlin Institute of
ing trials that are fit for purpose. BMJ. inal Products. EMEA/149995/2008 Health (BIH).
2015; 350, h2147–h2147. (rev.1). 2018; 1–18. Augustenburger Platz 1, D-13353
Berlin, Germany
52. “Annex I: Summary of product charac- 59. Council of the European Union.
and;
teristics for Kymriah. 2018. [Online]. REGULATION (EC) No 726/2004.
Berlin Center for Advanced
Available: https://www.ema.europa. Off. J. Eur. Communities. 2004;
Therapies (BeCAT), Charité–Uni-
eu/en/documents/product-informa- 138(March 2004), 1–15.
versitätsmedizin Berlin, Berlin,
tion/kymriah-epar-product-informa- Germany
60. EMEA/CHMP. Qualification opinion
tion_en.pdf. Tel.: +49 30 450 539 594
on Cellular therapy module of the Eu-
53. “Annex I: Summary of product charac- ropean Society for Blood & Marrow mohamed.abou-el-enein@
teristics for Yescarta. 2018. [Online]. Transplantation ( EBMT ) Registry. charite.de
Available: https://www.ema. EMA/CHMP/SAWP/792574/2018.
europa.eu/en/documents/ 2019; 1–46.


ADVANCED THERAPIES
EXPERT INSIGHT
CRISPR surgery for inherited

retinal diseases: landmarks in the 21st
century
Alexander H Chai & Stephen H Tsang
Gene therapy was first conceptualized in 1972 as clarification on viral

DNA-altering mechanisms was done. Since then, the field of gene thera-
py has transformed from a biological fantasy into a valid clinical treatment
in humans, in part due to significant innovations in the field of molecular
genetics. The development of gene therapy technology and the ensu-
ing research has laid a strong foundation for the advancement of gene
therapy, which has the potential to correct dominantly inherited disor-
ders that were previously incurable. In November 2018, a drug named
Luxturna became the first in vivo CRISPR/Cas9 genome surgery treat-
ment to be FDA-approved for use in clinical trials, which are set to take
place in patients with Leber congenital amaurosis 10 in the fall of 2019
[1]. However, there remain a number of scientific and practical barriers to
resolve before genomic medicine can become a widespread treatment.
DOI: 10.18609/cgti.2019.151
PROGRESS IN CRISPR conducted on the DNA-altering for a limited number of genetic

THERAPEUTICS mechanisms of viruses [2]. Since diseases, particularly recessive loss-
Gene therapy was first concep- then, genomic medicine has slow- of-function disorders [3]. Most
tualized in 1972 as research was ly transformed into valid treatment FDA-approved gene therapy trials
have been targeted at diseases with multiple specific guide RNAs into
an autosomal-recessive mode of in- the viral delivery vectors, therefore
heritance, where researchers have guiding the cutting enzyme to mul-
found that supplementation with tiple locations in the genome [8].
the wild-type version of the mu- Ophthalmology in particular has
tant gene may restore healthy func- become the leading field for the de-
tion in cells. This method is called velopment of genomic medicine;
gene augmentation. Unfortunately, the eyes are very favorable targets
this supplementation method only for these treatments and testing.
works for recessive disorders – au- The eye’s duplicity enables research-
tosomal-dominant (gain-of-func- ers to test the effects of treatments
tion) disorders do not respond the on one eye while having a dynamic
same way. A number of treatments control to compare the natural pro-
are being researched for dominant gression of disease with, and they
disorders, including RNA interfer- don’t require invasive procedures for
ence therapeutics, but currently the treatment or observation. The eye
most developed hope for a cure is also has a special relationship with
genome surgery to repair or remove the immune system. The eye exhib-
DNA (Table 1). its a reduction in antigen-present-
Fortunately, recent advancements ing cells and immunomodulatory
in molecular genetics, particularly factors in the vitreous humor when
the progress of CRISPR (clusters compared with other cells, which
of regularly-interspaced short pal- allows it to better tolerate the ad-
indromic repeats)/Cas9 technology, ministration of gene surgery vectors
have provided this hope. CRISPR/ [9]. Immunosuppressive cytokines
Cas9 technology became popular in and surface molecules displayed
2013 when it was shown to success- on ocular parenchymal cells, which
fully edit the DNA of human cells; interact with regulatory T cells to
since then, researchers have been fo- dampen inflammatory responses,
cusing on the development and re- also contribute to the eye’s relatively
finement of this technology for clin- immune-privileged state and ability
ical purposes [4]. CRISPR’s greatest to tolerate gene therapy [10,11].
advantage over other gene-editing In December of 2017, the first in
technologies is its low cost and vivo gene therapy was approved by
high efficiency when compared to the FDA for treatment of patients
other techniques such as transcrip- with Leber congenital amaurosis
tion activator-like effector nucleases 2 [12]. The drug, Luxturna (vore-
(TALEN) or zinc-finger nucleases tigene neparvovec-rzyl), is com-
(ZFN) [5]. This is mainly due to the posed of an adeno-associated virus
fact that CRISPR’s cutting mecha- containing human RPE65 cDNA
nism is guided by a strand of RNA, and is delivered subretinally. It was
which is much simpler to engineer first sold commercially in March of
than the complex proteins which 2018, and was a groundbreaking
ZFN and TALEN technologies rely step towards the widespread use
on [6]. CRISPR is also the only one of gene therapy as a treatment in
of these technologies capable of humans.
targeting more than one genetic lo- More recently, in November
cation via multiplexed genome sur- 2018, a groundbreaking in vivo
gery [7]. This is done by packaging CRISPR/Cas9 treatment for Leber
1452 DOI: 10.18609/cgti.2019.151

expert insight
ff TABLE 1
Ongoing current ocular gene augmentation/surgery trials.
Disease Treatment Phase End date Sponsor

Achromatopsia Subretinal administration of 1/2 2022 Applied Genetics
rAAV2tYF-PR1.7-hCNGB3 Technologies Corp.
Atrophic age-related RPE transplantation 1/2 2020 Chinese Academy
macular degeneration of Sciences
Choroideremia Subretinal administration of 2 2021 Bryon Lam
AAV2-REP1
Subretinal administration of AAV2/ 2 2021 University of
REP1 Oxford
Subretinal administration of 2 2019 Spark Therapeutics
AAV2-hCHM
Subretinal administration of 3 2020 Nightstar
AAV2-REP1 Therapeutics
Leber congenital amau- Subretinal administration of 3 2029 Spark Therapeutics
rosis 2 (LCA2) AAV2-hRPE65v2
1 2024 Spark Therapeutics
1/2 2026 Spark Therapeutics
Subretinal administration of AAV2/5 1/2 2023 MeiraGTx UK II Ltd.
OPTIRPE65
Subretinal administration of 1 2026 University of
rAAV2-CBSB-hRPE65 Pennsylvania
Leber congenital amau- Subretinal administration of 1/2 2024 Editas/Allergan
rosis 10 (LCA10) AGN-151587
Leber hereditary Intravitreal administration of GS010 3 2019 GenSight Biologics
optic neuropathy (rAAV2/2-ND4) versus sham intravit-
(LHON) real administration
2019 GenSight Biologics
2021 GenSight biologics
Intravitreal administration of 1 2019 John Guy, Universi-
scAAV2-P1ND4v2 ty of Miami
Neovascular age-relat- Subretinal administration of RGX-314 1 2020 Regenxbio Inc.
ed macular degenera-
tion (AMD)
Intravitreal administration of 1 2022 Adverum
ADVM-022 Technologies
Subretinal RetinoStat 2027 Oxford Miomedica
Stargardt disease Subretinal administration of 1/2 2019 Sanofi
SAR422459
2034 Sanofi
Usher syndrome 1B Subretinal administration of UshStat 1/2 2021 Sanofi
(EIAV-CMV-MYO7A)
1/2 2036 Sanofi
X-linked retinitis pig- Subretinal administration of 1/2 2024 Applied Genetic
mentosa (XLRP) rAAV2tYF-GRK1-RPGR Technologies Corp
Subretinal administration of 1/2 2020 MeiraGTx UK II Ltd.
AAV2/50hRKp.RPGR
Subretinal administration of 1/2 2019 Nightstar
AAV-RPGR Therapeutics
X-linked retinoschisis Intravitreal AAV8-scRS/IRBPhRS 1/2 2021 National Eye
Institute
Intravitreal rAAV2tYF-CB-hRS1 1/2 2022 Applied Genetic
Technologies Corp.
Information in table sourced from clinicaltrials.gov and DiCarlo et al. [18].

congenital amaurosis 10 was ap- gene mutation. Each treatment

proved by the FDA for Phase 1 must cut and replace only one gene
trial, the first approved in vivo use out of the estimated 30,000 genes
of CRISPR technology in humans in the human genome [15], or the
[13]. Patients in the trial will be patient could suffer from extreme
treated by subretinal injection of side effects. For a disease like adRP,
AGN-151587 in one eye, which a blinding disease caused by any
should theoretically cut out the mal- one of the 150 mutations in the
functioning gain-of-function muta- RHO gene discovered so far [16], it
tion and repair the malfunctioning would take millions of dollars and
CEP290 gene. This gene codes for several decades to develop the 150
a protein vital in the development gene-specific treatments to cure
of ciliogenesis, radial microtubule patients with just this one disease.
organization, and centriolar satel- Additionally, each gene-specific
lite clustering [14]. This gene, via treatment must be individually ap-
knockout studies, has been shown proved by the FDA after years of
to be vital for the development of trials proving the safety and efficacy
healthy photoreceptor cells. This of the treatment. With such a broad
trial is a landmark development spectrum of pathogenic mutations
in the field of genetic medicine, as and a relatively small number of pa-
the safety and efficacy of CRISPR tients with each mutation, it would
technologies directly injected into be financially unsustainable to de-
humans can finally be evaluated. velop such a wide variety of highly
Pending the results of this early in- specific genome surgery treatments.
vestigation, this study could provide Due to these factors, there is cur-
framework for further development rently no effective therapeutic op-
of novel CRISPR-based genome tion for patients with adRP or any
surgery treatments of inherited ret- other patient with gain-of-function
inal diseases and propel further re- photoreceptor degenerations.
search on the safety and efficacy of Additionally, even with signifi-
CRISPR-based treatments in other cant advancements, today’s CRIS-
organ systems. PR/Cas9 technology also contains
an inherent risk of off-targeting that
cannot be fully addressed by today’s
technology. Off-target gene abla-
TRANSLATION INSIGHT tion can have significant side effects,
There are a number of questions possibly even death. These effects are
to be answered surrounding the difficult to predict and mitigate; al-
practicality of precision genome though technologies can determine
surgery. First-generation CRISPR/ most of the off-target sequences
Cas9-based genome surgery’s great- in a genome, current methods are
est strength, its extreme mutation not guaranteed to comprehensively
specificity, is also its greatest weak- identify all of these sequences [17].
ness. Clinical gene therapy trials In addition to the costs of advanced
for autosomal dominant retinitis bioinformatics and next-genera-
pigmentosa (adRP) that focus on tion sequencing technologies, the
repair of a single gene, even if suc- results are confounded by the high
cessful, would only be applicable degree of variation of genetic ma-
to patients carrying that specific terial within each individual [17].
1454 DOI: 10.18609/cgti.2019.151

expert insight
These risks must be evaluated on an type cDNA that is modified to in-

individual basis with patients con- troduce mismatches through silent
sidered for CRISPR-based genome mutations and makes it resistant to
surgery. gRNA targeting. This design ensures
There have been some efforts at that gene ablation and replacement
correcting gain-of-function diseases happen simultaneously, which is an
using mutation nonspecific meth- important safety feature. Hopeful-
ods that can fix faulty DNA at spe- ly, with these rapid developments,
cific sites, but these methods can- the next generation of retina and
not completely replace traditional genetic medicine specialists will be
CRISPR methods. One method, inspired to advance these technolo-
referred to as CRISPR2.0, deliv- gies and the treatment of inherited
ers a second viral vector with wild disorders.

Funding declaration: Jonas Children’s Vision Care is supported by the National Institute of Health 5P30CA013696, R24
EY027285, 5P30EY019007, R01EY018213, R01EY024698, R01EY026682, R21AG050437, Abeona Therapeutics, the
Schneeweiss Stem Cell Fund, New York State [C029572], the Foundation Fighting Blindness New York Regional Research
Center Grant [C-NY05-0705-0312], Nancy & Kobi Karp, the Crowley Family Funds, The Rosenbaum Family Foundation, Al-
con Research Institute, the Gebroe Family Foundation, the Research to Prevent Blindness (RPB) Physician-Scientist Award,
unrestricted funds from RPB, New York, NY, USA.

Attribution: Copyright © 2019 Alexander H Chai & Stephen H Tsang. Published by Cell and Gene Therapy Insights under
Creative Commons License Deed CC BY NC ND 4.0.
Submitted for peer review: Aug 23 2019; Revised manuscript received: Nov 1 2019; Publication date: Nov 18 2019.
REFERENCES
1. Single Ascending Dose Study 2. Friedmann T, Roblin R. Gene thera- recessive forms of inherited retinal
in Participants With LCA10: py for human genetic disease? Science dystrophies. Gene Ther. 2012; 19:
https://clinicaltrials.gov/ct2/show/ 1972; 175: 949–55. 154–161.
NCT03872479?term=CRIS-
3. Smith AJ, Bainbridge JW, Ali RR. 4. Cong L, Ran FA, Cox D et al.
PR&cond=blindness&rank=1
Gene supplementation therapy for Multiplex genome engineering using

CRISPR/Cas systems. Science 2013; 10. Caspi RR. A look at autoimmunity 16. Tsang SH, Sharma T. Atlas of Inher-
339(6121): 819–23. and inflammation in the eye. J. Clin. ited Retinal Diseases. New York, New
Invest. 2010; 120(9): 3073–83. York (2018).
5. Xu CL, Park KS, Tsang SH. CRIS-
PR/Cas9 genome surgery for retinal 11. Stein-Streilein J, Taylor AW. An eye’s 17. Vakulskas CA, Behlke MA. Eval-
diseases. Drug Discov. Today Technol. view of T regulatory cells. J. Leukoc. uation and Reduction of CRISPR
2018; 28: 23–32. Biol. 2007; 81(3): 593–8. Off-Target Cleavage Events. Nucleic
Acid Ther. 2019; 29(4): 167–74.
6. DiCarlo JE, Mahajan VB, Tsang SH. 12. FDA approves novel gene therapy
Gene therapy and genome surgery in to treat patients with a rare form of 18. DiCarlo JE, Mahajan VB, Tsang SH.
the retina. J. Clin. Investing. 2018; inherited vision loss: https://www. Gene therapy and genome surgery
128: 2177–88. fda.gov/news-events/press-announce- in the retina. J. Clin. Invest. 2018;
ments/fda-approves-novel-gene-ther- 128(6): 2177–88.
7. Cong L, Ran FA, Cox D et al.
apy-treat-patients-rare-form-inherit-
Multiplex genome engineering using
ed-vision-loss
CRISPR/Cas systems. Science 2013;
339: 819–23. 13. Go-ahead for first in-body CRISPR AFFILIATIONS
medicine testing: https://www.nature.
8. Cong L, Ran FA, Cox D et al. Alexander H Chai
com/articles/d41587-018-00003-2
Multiplex genome engineering using Yale University,
CRISPR/Cas systems. Science 2013; 14. Firat-Karalar EN. The ciliopathy New Haven, Connecticut
339(6121): 819–23. gene product Cep290 is required for
Stephen H Tsang
primary cilium formation and micro-
9. Sonoda KH, Sakamoto T, Qiao H et Department of Ophthalmology,
tubule network organization. Turk J.
al. The analysis of systemic tolerance Columbia University, New York,
Biol. 2018; 42(5): 371–81.
elicited by antigen inoculation into New York
the vitreous cavity: vitreous cavity-as- 15. Human Genome Project and;
sociated immune deviation. Immu- FAQ: https://www.genome. Department of Pathology and Cell
nology 2005; 116(3): 390–9. gov/human-genome-project/ Biology, Columbia University,
Completion-FAQ New York, New York
1456 DOI: 10.18609/cgti.2019.151


ADVANCED THERAPIES
EXPERT INSIGHT
Bayesian phase 1/2 trial designs

and cellular immunotherapies: a
practical primer
Jordan Gauthier, Ying Yuan & Peter Thall
Bayesian phase 1/2 trial designs remain underused in biomedical re-

search and are virtually absent from the field of cellular therapy. In this
review, we highlight the severe limitations of the maximum tolerated
dose (MTD) concept and the traditional phase 1/phase 2 paradigm. Next,
we introduce statistical concepts underlying most adaptive Bayesian trial
designs. We use the EffTox design [1,2], one of many adaptive Bayesian
designs, as an example to illustrate ‘state-of-the-art’ phase 1/2 designs.
We highlight how these designs can be helpful to the cellular therapy
field specifically. Furthermore, we provide the reader with practical ex-
amples, links to freely available web applications, and R packages. We
hope this will incentivize investigators to implement these designs for
chimeric antigen-receptor-engineered T cell therapy trials, as well as oth-
er T cell-based therapies.
DOI: 10.18609/cgti.2019.152
INTRODUCTION past three decades, Bayesian phase designs, remain underused in bio-
Despite the development of many 1/2 designs, which hybridize con- medical research. Reflecting a re-
novel clinical trial designs over the ventional phase 1 and phase 2 cent surge of interest in innovative
trial designs, the FDA recently dose defined as the MTD. Phase 1
launched a Complex Innovative trials often are followed by an expan-
Trial Designs Pilot Program [3], sion cohort, in which additional pa-
and issued a draft guidance docu- tients are treated at the MTD. Con-
ment entitled ‘Adaptive Designs for ventionally, most expansion cohorts
Clinical Trials of Drugs and Biolog- are devoid of any experimental de-
ics’ [4]. In this review, we first will sign, in particular without statistical
highlight severe limitations of the justification for the sample size [12];
maximum tolerated dose (MTD) they often generate confusion when
concept and the traditional phase unexpected toxicities are observed at
1/phase 2 paradigm. Next, we will the previously selected MTD [13].
introduce statistical concepts un- Expansion cohorts are still used quite
derlying the Bayesian machinery commonly, notably in trials of chi-
used by most adaptive Bayesian tri- meric antigen receptor-engineered T
al designs. Last, we use the EffTox (CAR-T) cell therapy [14,15].
design based on efficacy-toxicity Many phase 1 studies in oncolo-
trade-offs [1,2], one of many adap- gy rely on so-called ‘3+3’ algorithms
tive Bayesian designs [5–11], as an [16]. The main advantage of 3+3 al-
example to illustrate ‘state-of-the- gorithms is that they do not require
art’ phase 1/2 designs. We define a computer program or a statisti-
the key EffTox design parameters cian to implement. This apparent
and explain how to interpret trial simplicity comes with a heavy price.
simulations done using freely avail- An example of a 3+3 algorithm is
able software. While Bayesian I/II shown in Table 1. In most cases,
designs are more complex than tra- 3+3 algorithms generate unreliable
ditional 3+3 algorithms, we believe estimates of the true probability of
they are driven by concepts that can toxicity at each dose (Table 2). When
be grasped easily by clinicians and compared to alternative designs, in-
researchers. Although we focus on cluding EffTox, 3+3 algorithms are
the EffTox design as an example, far less likely to choose a truly opti-
our goal is to popularize the whole mal dose [17]. Another major draw-
family of Bayesian phase 1/2 de- back of 3+3 algorithms is that they
signs. We wish to make their un- leave many decisions to be made
derlying concepts more accessible, solely, and subjectively, using clin-
and to encourage researchers in the ical judgement.
field of cellular immunotherapy to A key assumption underlying the
use them. notion of MTD and 3+3 algorithms
is monotonicity. This says that a
higher dose is necessarily associat-
ed with both higher toxicity and
LIMITATIONS OF THE higher efficacy probabilities. This
MTD CONCEPT & THE motivates the common practice of
TRADITIONAL PHASE 1/ finding the highest dose with ‘ac-
PHASE 2 PARADIGM ceptable’ toxicity, called the MTD.
The traditional paradigm splits early We will highlight limitations of the
clinical drug development into two MTD paradigm by considering sev-
successive phases: phase 1 trials, to eral scenarios that are obvious sim-
determine the MTD, and phase 2 plifications of more complex biolog-
trials, to evaluate the efficacy of the ical processes; the dose-toxicity and
1484 DOI: 10.18609/cgti.2019.152

expert insight
ff TABLE 1
Example of a phase 1 protocol 3+3 algorithm.
Number of patients with a DLT at a Escalation decision rule

given dose level
0/3 Enter 3 patients at the next dose level
1/3 Enter at least 3 more patients at this dose level
If 0 of these 3 patients experience a DLT, proceed to the next dose level
If ≥1 of this group suffer DLT, this dose exceeds the MTD and dose escala-
tion is stopped. Three additional patients will be entered at the next lower
dose level if only 3 patients were treated previously at that dose
≥2 Dose escalation will be stopped. This dose level will be declared the max-
imally administered dose (highest dose administered). Three additional
patients will be entered at the next lower dose level if only 3 patients were
treated previously at that dose
MTD: the highest dose at which no more than 1 of 6 evaluable patients has had a DLT. Six patients should be treated before the dose
is declared as MTD.
DLT: Dose-limiting toxicity; MTD: Maximal tolerated dose.
dose-efficacy relationship shapes dose 5. In this scenario, the gain in

can be sigmoid, U-shaped, dome- efficacy clearly outweighs the risk of
shaped, and more complex shapes toxicity. In this case, choosing dose
are also possible. Many factors re- 4 over dose 5 would be detrimental
lated to cellular immunotherapies for patients, leading to underdosing
– such as patient and disease-related in terms of efficacy with comparable
variables, cell product character- toxicity. A flaw with the monotonic-
istics, in vivo kinetics – may alter ity assumption is also seen in a less
these relationship shapes. Another favorable scenario, where the exper-
advantage of most Bayesian phase imental agent is toxic but has very
1/2 designs, such as EffTox, is that low efficacy (Figure 2). In this case,
they make limited assumptions as no dose is acceptable, but a 3+3 al-
to the shapes of these relationships. gorithm still will choose a MTD.
A key point is that the MTD con- To relax these oversimplifications,
cept overlooks the fact that the rela- later in this review we will explore
tionships between dose and toxicity, the concept of risk–benefit trade-
and dose and efficacy, may differ off utilized in the EffTox design. Of
significantly, as it is often expected note, 3+3 designs are known to be
with cellular therapies. This leads outperformed by many other phase
to the concept that, among two or 1 designs, such as continuous reas-
more acceptable doses, some might sessment methods [18], and mod-
be more desirable than others, for ified toxicity probability interval
instance when the gain in efficacy is (mTPI) designs [19,20].
large while the increase in toxicity is The biological properties of im-
small. Let us consider a target toxic- mune cells are very distinct com-
ity rate of 0.3, which is a commonly pared to conventional cytotoxic
used threshold value. In the sce- agents or antibodies, setting them
nario depicted in Figure 1, toxicity apart from the conventional rules of
nearly plateaus at dose 4 with a very pharmacokinetics. For example, us-
slight increase from dose 4 to dose ing a mixed-effect model Stein et al.
5. In contrast, we observe a sharp [21] did not observe any relationship
increase in efficacy from dose 4 to between the dose of tisagenlecleucel

ff TABLE 2
Example of DLT estimates of a simulated phase 1 trial using the 3+3 algorithm showed in Table 1.
Dose Number of patients Number of DLTs 95% CI* of the true probability of DLT
1 3 0 0.00–0.71
2 6 1 0.00–0.64
3 3 2 0.09–0.99
*Confidence intervals were computed using the Clopper and Pearson method.
CI: Confidence interval; DLT: Dose-limiting toxicity.
– an FDA-approved CD19-target- future beliefs (e.g., about a treat-

ed CAR-T cell product for pediat- ment effect) are a consequence of
ric and young adults with relapsed our current beliefs (prior probabil-
or refractory acute lymphoblastic ities) and newly gathered evidence
leukemia (ALL) – and the Cmax, (likelihood function). This proba-
nor any other model parameter. In bility relationship, known as Bayes’
contrast, the Seattle group report- Law, can be simplified as follows:
ed a strong association between the
dose of JCAR014 and the in vivo Posterior probability of outcome
CAR-T cell expansion in lympho- (toxicity or efficacy) ∝
ma patients [22], and to a lower ex- Likelihood × Priors
tent in adult ALL patients [15]. The
fundamentally different nature of The posterior probability of an
cellular therapies compared to other outcome of interest, given the ob-
anti-cancer agents calls for designs served data and our prior beliefs
capable of capturing various dose–ef- (priors), is a product of the likeli-
fect relationships without relying on hood (how likely the data are for
stringent assumptions. More, e.g., a each parameter value) and our
higher dose, might not be necessar- prior beliefs. Consequently, par-
ily better, e.g., more effective, while ticular care should be taken when
being associated with significant- defining priors in collaboration
ly higher toxicity. In this situation, with statisticians, biologists, and
Bayesian phase 1/2 designs such as clinicians. Priors reflect both prior
EffTox, which do not assume lin- beliefs (prior means, or ‘most likely
earity and monotonicity, can help probabilities’ that are easily elicit-
us better understand and estimate ed from investigators) and prior
the shape of the dose-response and uncertainty (prior variance or stan-
dose-toxicity relationships, and iden- dard deviations that have no intu-
tify the dose maximizing efficacy itive meanings, often determined
while minimizing toxicity. through computer simulations).
Several approaches are possible to
construct these priors, which are
beyond the scope of this review
STATISTICAL CONCEPTS [2,13,23]. One way of eliciting ‘con-
Prior & posterior sensus’ prior means, is to simply
probabilities average the probabilities obtained
from several experts or investiga-
Bayesian statistics is based on the tors. Examples of prior means are
probabilistic principle that our shown in Table 3. In practice, priors
1486 DOI: 10.18609/cgti.2019.152

expert insight
and other design parameters are DESCRIPTION OF THE

calibrated by assessing the perfor- EFFTOX DESIGN
mance of the design across a broad Dose acceptability
range of scenarios where the true
probability of toxicity and efficacy While a number of Bayesian phase
are specified. Extreme care should 1-2 methods have been developed,
be taken running these computer we will focus here on the EffTox de-
simulations to detect and avoid design, developed by Thall et al. [1] and
signs with ‘pathological’ behaviors, refined by Thall et al. [2]. This phase
such as the failure to detect the op- 1/2 Bayesian design has already been
timal dose with high probability, or used successfully in several clinical
the dose finding procedure getting trials [24–26]. Free software for trial
stuck at a suboptimal dose. design and implementation is avail-
able at https://biostatistics.mdander-
son.org/SoftwareDownload/. EffTox
requires the specification of two cri-
Application to adaptive
teria defining a dose as ‘acceptable’.
decision-making
A dose is considered acceptable if,
Under this framework, decisions to given the current data, there are rea-
modify the treatment regimen (or sonably high posterior probabilities
simply the dose) are based on pos- that
terior probabilities computed un- 1. the efficacy probability is above a
der the Bayesian model. Let Dosen pre-defined threshold; and
denote the dose assigned to the nth
patient. This Bayesian sequential 2. the toxicity probability is below a
decision process can be described as pre-defined threshold
follows:
Dosen→Datan→
Posteriorn→Dosen+1→ ffFIGURE 1
Datan+1→Posteriorn+1→... Favorable scenario where the monotonicity assumption is broken.
The outcomes observed after n
patients (Datan) are used to com-
pute new posterior probabilities.
The decision to repeat or alter the
next dose or regimen for the (n+1)
th patient (Dosen+1) is based
on these posterior probabilities.
Thus, by repeated applying Bayes’
Law, one may sequentially adapt
decisions by using new data to ob-
tain updated posterior probabili-
ties. This often is done after a new
cohort of three patients has been
treated and evaluated. In the next
section, we will specify the process
of defining posterior probability
criteria for decision-making in the
context of the EffTox design.

ffFIGURE 2 desirable when they maximize an

Unfavorable scenario where the monotonicity assumption is broken. objective function that quantifies
the trade-off between efficacy and
toxicity. To construct this function,
efficacy-toxicity trade-off contours
are constructed based on three
equally desirable pairs of Pr(effica-
cy) and P{r(toxicity). For example:
1. The probability of efficacy in the

absence of toxicity is .40;
2. The probability of toxicity is 65%

and the probability of efficacy is
100%;
3. An optimal pair of efficacy/toxicity

probabilities equally desirable to
1. and 2. E.g., Pr(efficacy) =.70
with Pr(toxicity) =.25.
Efficacy–toxicity trade-off con-

tours allow us to visualize the optimal
These acceptability thresholds are de- trade-offs (Figure 3). The desirability
fined by the clinical investigators. trade-off increases as the pair [Pr(ef-
For example, one can consider a ficacy), Pr(toxicity)] moves from the
dose unacceptable – and conversely upper left to the lower right corner of
that it is acceptable – if either of the the contour plot. The shapes of the
following conditions is satisfied: contours help quantifying the trade-
off: two dots – two pairs of toxicity/
ff There is a ≥ 90% posterior efficacy probabilities – located on the
probability that the toxicity same contour being equally desirable.
probability is > .30; While the trade-off contour can be
difficult to understand and interpret,
ff There is a ≥ 90% posterior we have shown that for the EffTox de-
probability that the efficacy sign to perform well, this contour has
probability is < .50. Only to be steep [2]. We acknowledge that
acceptable doses are given to defining these efficacy-toxicity trade-
patients. If no dose is acceptable, off values can be challenging in prac-
the trial is stopped and no dose tice; they might also evolve over time.
is selected. Utility-based designs [8,27], which
quantify risk–benefit trade-offs, can
in part address this limitation.
Dose desirability
Beyond being acceptable, some Validation through

doses may be more desirable than simulations
others. This concept of desirabili-
ty is the cornerstone of the EffTox Before one can use the design for
design. We consider doses as more trial conduct, one must carry out a
1488 DOI: 10.18609/cgti.2019.152

expert insight
ff TABLE 3
Examples of prior mean probabilities (priors) elicited from the investigators.
Elicited mean prior probabilities of toxicity Elicited mean prior probabilities of efficacy
Dose 1 0.02 0.20
Dose 2 0.04 0.40
Dose 3 0.20 0.60
Dose 4 0.50 0.70
Dose 5 0.80 0.80
large number of simulations (usu- the available data, in particular pre-

ally 1,000) under each of several clinical data from animal models.
distinct scenarios (usually 5–10) to
ensure the design performs well. A
design that performs well has high Trial conduct
probabilities of choosing doses that
are truly optimal, and high proba- During the trial, a Bayesian adap-
bilities of stopping the trial early in tive decision process – as described
the unfortunate scenarios where no above – is used to determine the
dose has both acceptable efficacy optimal dose for each successive
and toxicity. cohort, relying on the prior prob-
The scenarios should reflect a abilities of efficacy and toxicity
range of possible cases, for example:
ffFIGURE 3
ff Best scenario: high efficacy, low
Efficacy-toxicity trade-off contours.
toxicity at the best dose;
ff Worst-case scenario: low

efficacy, high toxicity at all doses;
ff Middle ground: intermediate

efficacy, intermediate toxicity at
the best dose;
ff Non-linear effects: Pr(toxicity)

plateaus, Pr(efficacy) keeps
increasing;
ff Non-linear effects: Pr(efficacy)

plateaus, Pr(toxicity) keeps
increasing;
ff Nothing happens: low toxicity,

low efficacy at all doses.
This is not by any means an ex-

haustive list and it is crucial that in- The target contour of the function δ is defined by three efficacy-toxicity probability
vestigators, pharmacists, biologists, pairs, initially elicited from the investigator, that are considered to be equally desirable.
These points are plotted on the graph with blue dots and connected by an orange
and statisticians consider all plausi- line representing the target contour. This plot was was generated using free software
ble and relevant scenarios, based on available at https://biostatistics.mdanderson.org/SoftwareDownload/.

determined by the investigators for the field, such as the impact

and the statisticians. The mean of prior therapies, tumor burden,
posterior probabilities of efficacy disease subtype, lymphodepletion,
and toxicity associated with each and variables related to the man-
dose are computed, and efficacy/ ufacturing of the cellular therapy
toxicity trade-off values (reflecting product (e.g., immune cell phe-
each dose’s desirability) are deter- notypes, in vitro T cell function-
mined. The next cohort will be ality). Moreover, cellular therapies
given a dose both satisfying the ac- such as CD19-targeted CAR-T cell
ceptability criteria and having the therapy can be associated with se-
highest desirability. If no dose is vere toxicities, namely cytokine re-
acceptable, the trial will be stopped lease syndrome and neurotoxicity
early with no dose chosen. [32,33]. This highlights an urgent
need to inco rporate novel statis-
tical tools to help design and ana-
lyze cellular immunotherapy trials.
APPLICATIONS TO Future research should also aim at
TRIALS OF CELLULAR comparing the cost-effectiveness of
IMMUNOTHERAPY phase 1/2 designs to the conven-
Despite a broad panel of design optional paradigm (phase 1 with an
tions, Bayesian phase 1/2 designs expansion cohort, or phase 1 fol-
have been rarely used in trials of lowed by a phase 2 trial).
immunotherapy. Some examples The following section highlights
of Bayesian designs can be found some of the advantages of phase 1/2
in a limited number of vaccination Bayesian trials over conventional
trials [28,29], although not concur- 3+3 algorithms:
rently evaluating toxicity and effi-
cacy. Some on-going CAR-T cell ff Higher probabilities of choosing
therapy phase 1 trials rely on the the optimal dose [17];
mTPI approach [12,30]. An on-go-
ing trial of CAR NK cell therapy ff By simultaneously assessing
[31] is the only one in this field, to efficacy and toxicity, phase
date and to our knowledge, using a 1/2 designs could dramatically
Bayesian phase 1/2 design. accelerate the development of
Trials of cellular immunother- cellular therapies;
apy pose unique challenges. First,
the enrollment potential of these ff Evaluating toxicity as well as
trials is also limited; they are in- efficacy concurrently provides
deed often logistically challenging, significantly more information
extremely costly, and these thera- than a conventional phase 1
pies are currently restricted to se- trial, which ignores efficacy, thus
lected malignancies in the relapsed/ leading to better estimates of the
refractory setting. Furthermore, dose–effect relationships;
most cellular therapies are only
available in a limited number of ff The acceptability criteria
centers. Taken together, these high reduce the number of
costs and relatively small numbers patients potentially treated at
of patients challenge our ability to inacceptable doses, e.g. if toxicity
prospectively address key questions is too high or efficacy is too low.
1490 DOI: 10.18609/cgti.2019.152

expert insight
REMAINING CHALLENGES One major impediment to adopt

novel phase 1/2 designs is that each
The efficacy and toxicity of cellular decision of dose escalation/de-es-
therapies, in particular CAR-T cell calation requires real-time, often
therapies, are not only functions of complicated, model estimation and
the dose administered, they also are computation, which can be logis-
strongly dependent on the in vivo tically challenging. This issue has
expansion and persistence of the been addressed on several fronts.
infused cells. The biological prop- Freely available, user-friendly soft-
erties of immune cells are very dis- ware and web applications (https://
tinct compared to conventional cy- biostatistics.mdanderson.org/Soft-
totoxic agents or antibodies, setting wareDownload/ and http://www.
them apart from the conventional trialdesign.org), as well as com-
rules of pharmacokinetics. Specif- mercial platforms (https://udesign.
ically, T cell functionality [34,35], laiyaconsulting.com/), can facilitate
lymphodepletion [15,22,36], tumor real-time model parameter re-esti-
burden [15,37] and expression lev- mation and decision making. For
els of the target antigen [38,39] are R users, several packages have also
known to dramatically impact the been implemented to facilitate the
in vivo kinetics of CAR-T cells, use of Bayesian phase 1/2 designs
which subsequently alter the risk (e.g., trialr, EffToxDesign, dfcomb).
of toxicity. These marked varia- Moreover, newer phase 1/2 designs,
tions in the risk of toxicity, for such as the U-BOIN (utility-based
example based on tumor burden, Bayesian optimal interval) design
should be taken into account and [27], remove the requirement of
more complex phase 1/2 designs complicated model estimation.
have been developed to account for U-BOIN’s dose escalation and
prognostic covariates [40]. The lim- de-escalation rule can be pre-tabu-
ited enrollment potential of cellu- lated and included in the protocol
lar immunotherapy trials prompts prior to the trial conduct, making
the field to maximize information its implementation as simple as the
gain from trial data; although Ef- 3+3 design, while yielding competi-
fTox was designed to assess only tive performance.
two binary outcomes, new Bayes- We anticipate combinatorial ap-
ian phase 1/2 designs have been proaches, for example CAR-T cells
developed to evaluate multiple combined with immune check-
co-primary endpoints while ac- point inhibitors [30,41] or other
counting for prognostic covariates molecularly targeted agents [42,43],
[40]. One of such designs is used to become an area of intense focus
in for a trial of CAR NK cell ther- in the near future. Many phase 1/2
apy currently on-going at the MD designs are available to optimize
Anderson Cancer Center [31]. This two-agent combinations [44,45],
design evaluates five co-primary or the dose and schedule of ad-
outcomes (time to severe toxicity, ministrations [46], accommodate
cytokine release, syndrome, disease late-onset events [47] and immune
progression or response and death) response [48], and account for pa-
across six prognostic subgroups tient genetic heterogeneity [18].
characterized by the disease type The U-BOIN design can use im-
and tumor burden. mune biological outcomes, such

as a prognostic immune biomark- advance the field of cellular thera-

er, to predict delayed outcomes py and should be considered when
[27]. These novel designs can help planning early phase trials.

Funding declaration: The authors received no financial support for the research, authorship and/or publication of this article.

Attribution: Copyright © 2019 Jordan Gauthier, Ying Yuan & Peter Thall. Published by Cell and Gene Therapy Insights under
Creative Commons License Deed CC BY NC ND 4.0.
Submitted for peer review: Sep 10 2019; Revised manuscript received: Oct 30 2019; Publication date: Nov 18 2019.
REFERENCES
1. Thall PF, Cook JD. Dose‐Find- Ratios. Biometrics 2006; 62(3): seamless phase I/II clinical trials. Stat.
ing Based on Efficacy–Toxicity 777–87. Med. 2017; 36(26): 4106–20.
Trade‐Offs. Biometrics 2004; 60(3):
6. Pan H, Huang P, Wang Z, Wang L, 10. Zang Y, Lee JJ, Yuan Y. Adaptive
684–93.
Li C, Xia J. A Novel Bayesian Seam- designs for identifying optimal bio-
2. Thall PF, Herrick RC, Nguyen HQ, less phase I/II Design. PLoS ONE logical dose for molecularly targeted
Venier JJ, Norris CJ. Effective sample 2013; 8(9). agents. Clinical Trials 2014; 11(3):
size for computing prior hyper- 319–27.
7. Liu S, Johnson VE. A robust Bayes-
parameters in Bayesian phase I–II
ian dose-finding design for phase I/ 11. Li DH, Whitmore JB, Guo W, Ji
dose-finding. Clinical Trials 2014;
II clinical trials. Biostatistics (Oxford, Y. Toxicity and Efficacy Probability
11(6): 657–66.
England). 2015; 17(2): 249–63. Interval Design for phase I Adoptive
3. FDA. Complex Innovative Trial Cell Therapy Dose-Finding Clinical
8. Lee J, Thall PF, Ji Y, Müller P. Bayes-
Designs Pilot Program; 2019. Trials. Clin. Cancer Res. 2017; 23(1):
ian Dose-Finding in Two Treatment
13–20.
4. FDA. Adaptive Design Clinical Trials Cycles Based on the Joint Utility of
for Drugs and Biologics; 2018. Efficacy and Toxicity. JASA 2015; 12. Dahlberg SE, Shapiro GI, Clark JW,
110(510): 711–22. Johnson BE. Evaluation of Statis-
5. Yin G, Li Y, Ji Y. Bayesian Dose‐
tical Designs in phase I Expansion
Finding in phase I/II Clinical Trials 9. Lin R, Yin G. STEIN: A simple tox-
Cohorts: The Dana-Farber/Harvard
Using Toxicity and Efficacy Odds icity and efficacy interval design for
1492 DOI: 10.18609/cgti.2019.152

expert insight
Cancer Center Experience. JNCI 21. Stein AM, Grupp SA, Levine JE et 29. Zohar S, Baldi I, Forni G, Merletti
2014; 106(7). al. Tisagenlecleucel Model‐Based F, Masucci G, Gregori D. Planning
Cellular Kinetic Analysis of Chime- a Bayesian early‐phase phase I/II
13. Yuan Y, Nguyen HQ, Thall PF.
ric Antigen Receptor–T Cells. CPT study for human vaccines in HER2
Bayesian Designs for phase I–II
2019; 8(5): 285–95. carcinomas. Pharmaceut. Stat. 2011;
Clinical Trials (ed 2016): Chapman
10(3): 218–26.
& Hall; 2016. 22. Turtle CJ, Hanafi L-A, Berger C et al.
Immunotherapy of non-Hodgkin’s 30. Hirayama AV, Gauthier J, Hay KA et
14. Locke FL, Ghobadi A, Jacobson CA
lymphoma with a defined ratio of al. Efficacy and Toxicity of JCAR014
et al. Long-term safety and activity of
CD8+ and CD4+ CD19-specific chi- in Combination with Durvalumab
axicabtagene ciloleucel in refractory
meric antigen receptor–modified T for the Treatment of Patients with
large B-cell lymphoma (ZUMA-1):
cells. Sci. Trans. Med. 2016; 8(355). Relapsed/Refractory Aggressive
a single-arm, multicentre, phase 1–2
B-Cell Non-Hodgkin Lymphoma.
trial. Lancet Oncol. 2018; 130: 2017. 23. Gelman AC, John B, Stern Hal S,
Blood (ASH abstract) 2018; 132(Sup-
Dunson DB, Vehtari A, Rubin DB.
15. Turtle CJ, Hanafi L-A, Berger C et pl. 1).
Bayesian Data Analysis; 2013.
al. CD19 CAR–T cells of defined
31. Lee J, Thall PF, Rezvani K. Opti-
CD4+:CD8+ composition in adult 24. de Lima M, Champlin RE, Thall PF
mizing natural killer cell doses for
B cell ALL patients. J. Clin. Invest. et al. Phase I/II study of gemtuzumab
heterogeneous cancer patients on the
2016; 126(6): 2123–38. ozogamicin added to fludarabine,
basis of multiple event times. J. Royal
melphalan and allogeneic hemato-
16. Tourneau C, Lee JJ, Siu LL. Dose Es- Stat. Soc. Series C (Appl. Stat.) 2019;
poietic stem cell transplantation for
calation Methods in phase I Cancer 68(2): 461–74.
high-risk CD33 positive myeloid
Clinical Trials. JNCI 2009; 101(10):
leukemias and myelodysplastic syn- 32. Gauthier J, Turtle CJ. Insights into
708–20.
drome. Leukemia 2008; 22(2): 258. cytokine release syndrome and
17. Yan F, Thall PF, Lu KH, Gilbert neurotoxicity after CD19-specific
25. Whelan HT, Cook JD, Amlie-Lefond
MR, Yuan Y. phase I–II clinical trial CAR-T cell therapy. Curr. Res. Trans.
CM et al. Practical Model-Based
design: a state-of-the-art paradigm Med. 2018; 66 (J. Clin. Investig. 126
Dose Finding in Early-Phase Clinical
for dose finding. Ann. Oncol. 2017; 6 2016): 50–52.
Trials. Stroke 2008; 39(9): 2627–36.
29(3): 694–99.
33. Gauthier J, Yakoub-Agha I. Chime-
26. Konopleva M, Thall PF, Yi C et al.
18. O’Quigley J, Pepe M, Fisher L. ric antigen-receptor T-cell therapy
Phase I/II study of the hypoxia-acti-
Continual reassessment method: a for hematological malignancies and
vated prodrug PR104 in refractory/
practical design for phase 1 clinical solid tumors: Clinical data to date,
relapsed acute myeloid leukemia
trials in cancer. Biometrics 1990; current limitations and perspectives.
and acute lymphoblastic leukemia.
46(1): 33–48. Curr. Res. Trans. Med. 2017; 65(3):
Haematologica 2015; 100(7).
93–102.
19. Ji Y, Wang S-J. Modified Toxicity
27. Zhou Y, Lee JJ, Yuan Y. A utili-
Probability Interval Design: A Safer 34. Fraietta JA, Beckwith KA, Patel PR
ty-based Bayesian optimal interval
and More Reliable Method Than the et al. Ibrutinib enhances chimeric
(U-BOIN) phase I/II design to
3 + 3 Design for Practical phase I antigen receptor T-cell engraftment
identify the optimal biological dose
Trials. J. Clin. Oncol. 2013; 31(14): and efficacy in leukemia. Blood 2016;
for targeted and immune therapies.
1785–91. 127(9): 1117–27.
Stat. Med. 2019.
20. Guo W, Wang S-J, Yang S, Lynn H, 35. Rossi J, Paczkowski P, Shen Y-W
28. Morita S, Oka Y, Tsuboi A et al. A
Ji Y. A Bayesian interval dose-finding et al. Preinfusion polyfunctional
phase I/II Trial of a WT1 (Wilms’
design addressing Ockham’s razor: anti-CD19 chimeric antigen receptor
Tumor Gene) Peptide Vaccine in
mTPI-2. Contemp. Clin. Trials 2017; T cells are associated with clinical
Patients with Solid Malignancy:
58: 23–33. outcomes in NHL. Blood 2018;
Safety Assessment Based on the phase
132(8): 804–14.
I Data. Japanese J. Clin. Oncol. 2006;
36(4): 231–6.

36. Hirayama AV, Gauthier J, Hay KA et (CAR)–modified T cells: refueling 47. Jin I, Liu S, Thall PF, Yuan Y. Using
al. The response to lymphodepletion the CAR. Blood 2017; 129(8): Data Augmentation to Facilitate
impacts PFS in aggressive non-Hod- 1039–41. Conduct of Phase I–II Clinical Trials
gkin lymphoma patients treated with With Delayed Outcomes. J. Am. Stat.
42. Gauthier J, Hirayama AV, Hay KA
CD19 CAR-T cells. Blood 2019. Assoc. 2014; 109(506): 525–36.
et al. Comparison of Efficacy and
37. Turtle CJ, Hay KA, Hanafi L-A et Toxicity of CD19-Specific Chimeric 48. Liu S, Guo B, Yuan Y. A Bayesian
al. Durable Molecular Remissions Antigen Receptor T-Cells Alone or in Phase I/II Trial Design for Immu-
in Chronic Lymphocytic Leukemia Combination with Ibrutinib for Re- notherapy. J. Am. Stat. Assoc. 2018;
Treated With CD19-Specific Chi- lapsed and/or Refractory CLL. Blood 113(523): 1016–27.
meric Antigen Receptor–Modified (ASH abstract) 2018; 132(Suppl. 1).
T Cells After Failure of Ibrutinib. J.
43. Mestermann K, Giavridis T, Weber
Clin. Oncol. 2017; 35(26).
J et al. The tyrosine kinase inhibitor AFFILIATIONS
38. Pont MJ, Hill T, Cole GO et al. dasatinib acts as a pharmacologic on/
Jordan Gauthier
γ-secretase inhibition increases off switch for CAR T cells. Sci. Trans.
Author for correspondence
efficacy of BCMA-specific chimeric Med. 2019; 11(499).
Clinical Research Division, Fred
antigen receptor T cells in multiple
44. Yuan Y, Yin G. Bayesian phase I/II Hutchinson Cancer Research Cen-
myeloma. Blood. 2019.
adaptively randomized oncology tri- ter, Seattle, WA , USA
39. Ramakrishna S, Highfill SL, Walsh Z als with combined drugs. Ann. Appl. and
et al. Modulation of Target Anti- Stat. 2011; 5(2A): 924–42. Department of Medical Oncology,
gen Density Improves CAR T-cell University of Washington, Seattle,
45. Takeda K, Taguri M, Morita S.
Functionality and Persistence. Clin. WA, USA
BOIN‐ET: Bayesian optimal interval
Cancer Res. 2019; 25(17): 5329–41. jgauthier@fredhutch.org
design for dose finding based on both
40. Thall PF, Nguyen HQ, Estey EH. Pa- efficacy and toxicity outcomes. Phar- Ying Yuan
tient‐Specific Dose Finding Based on maceut. Stat. 2018; 17(4): 383–95. Department of Biostatistics, M.D.
Bivariate Outcomes and Covariates. Anderson Cancer Center, Hous-
46. Guo B, Li Y, Yuan Y. A dose–sched- ton, Texas, USA
Biometrics 2008; 64(4): 1126–1136.
ule finding design for phase I–II
Peter Thall
41. Chong EA, Melenhorst JJ, Lacey clinical trials. J R Stat. Soc. Ser. C
Department of Biostatistics, M.D.
SF et al. PD-1 blockade modu- Appl. Stat. 2016;65(2):259-272.
Anderson Cancer Center, Hous-
lates chimeric antigen receptor
ton, Texas, USA
1494 DOI: 10.18609/cgti.2019.152


ADVANCED THERAPIES
EXPERT INSIGHT
Considerations for patient selection

for cell and gene therapy trials using
tumor associated antigens as target in
early phase development
Stephanie Traub & David Edwards
For cell and gene therapies, including those using tumor associated anti-
gens (TAAs) as targets, effective patient selection is critical for success.
In this paper, we discuss considerations for patient selection for cell and
gene therapy products in early phase clinical development. Surprisingly,
many obvious key factors like the TAA themselves, the major histocom-
patibility complex (MHC), as well as practical implication of patient se-
lection on the trial design and conduct are not given the consideration
that they should be given. The article focuses on ideal patient selection
for cell and gene therapies using TAAs and implications for clinical trial
design.
DOI: 10.18609/cgti.2019.149
CELL & GENE THERAPIES TAAs are attractive anti-cancer effects, it can reduce many of the
USING TUMOR targets. Where tumor specificity side effects commonly observed
ASSOCIATED ANTIGENS is optimized to reduce off-tissue with more conventional therapies
[1]. CAR T-cell therapies and TCR cell expresses it. Instead, TAA posi-
T-cell therapies utilize TAAs to at- tive cells display a patchy appearance
tack the tumor directly; and indi- and this might limit the efficacy of
rect vaccination approaches like au- any TAA targeted therapy [6]. When
tologous and allogenic dendritic cell selecting a TAA for a cell and gene
(DC) therapies use TAAs to induce therapy product, the selectivity of
antigen-specific T cells. the therapy and the heterogeneity of
the antigen in the tumor need careful
consideration to minimize off-target
toxicity and optimize efficacy.
TUMOR ASSOCIATED
ANTIGENS
Less than 100 TAAs have been
identified so far and some of them
MAJOR
tested in clinical trials [2,3]. The key
HISTOCOMPATIBILITY
factor for a good antigen target is
COMPLEX
tumor-specificity, i.e., an antigen The major histocompatibility com-
which is ideally not expressed in plex (MHC), also called HLA (hu-
other tissues. In reality, this is rarely man leukocyte antigen) complex, is
the case: most TAAs are expressed a set of genes that are co-dominantly
in tissues other than tumor. What expressed and are highly polymor-
is of importance, however, is that phic [7]. MHC class one (MHC-I)
TAA expression is higher in the tu- and two (MHC-II) proteins are ex-
mor than in the other tissues. For pressed on APCs as well as B lym-
example, prostatic acid phosphatase phocytes and MHC-I on almost all
(PAP) has been found to be ex- nucleated non-APCs [8,9].
pressed in the prostate. PAP is not The inheritance of the HLA hap-
restricted solely to prostate tissue, lotypes from each parent results in
but its expression in other tissues is a random combination of different
~1–2 orders of magnitude less than HLA loci. However, some HLA hap-
that observed in the prostate [4]. lotypes are over-represented in certain
Specifically, when targeting a new populations: HLA-A1, -B8, -DR17
unknown antigen which has been is the most common HLA haplotype
not intensively explored, it is of im- among Caucasians [8]; the subtype
portance to check for tissue expres- genotype HLA-A*02:01, is found
sion in addition to its expression in different frequencies in popula-
on tumor tissues. TAA selectivity tions, e.g., in Finland, HLA-A*02:01
is important; there have been cases would be found in 34.4% of the
where unexpected toxicity has oc- population while in Thailand, 1.8%
curred due to recognition of an epi- would have the genotype [10,11]. It
tope from an unrelated protein [5]. might be that regional differences
An additional consideration is due to HLA prevalence can occur
the degree of heterogeneity that and impact patient selection. The
the TAA displays; ideally, every tu- implications of this variability need
mor cell would express the TAA. In to be carefully considered both when
some cases, such as NY-ESO-1 (New designing MHC restricted therapies
York-esophageal squamous cell car- for a particular region and when de-
cinoma-1) expression in hepatocel- ciding where to locate clinical trial
lular carcinoma, not every tumor centers that will efficiently recruit.
1416 DOI: 10.18609/cgti.2019.149

expert insight
The predicted abundance of the A second aspect to consider is

MHC allele in the population needs the actual failure of the analytical
to be factored in when considering method or issues around the sample
eligibility screening; underestimation given by the patient. In the MAG-
can impact on time lines and budget. RIT trial 7.4% of patient screened
did sign the consent but no sample
was sent, insufficient tumor sample
was available, contaminations in the
TAA & HLA analytical method did not give a re-
Prevalence of TAA sult, and there were other issues. In
There is often discordance between addition, even after being identified
the prevalence of tumor antigens as MAGE-A3 positive, a large num-
described in the literature and that ber of patients (45%, 2312/4210)
encountered during trial screening. did not meet the eligibility criteria
Literature data sets are often rela- – e.g. patients variously not having
tively small with limited numbers signed informed consent for main
of cases of tumor samples. In addi- study, not being found to be free of
tion, protein data via immunohis- metastasis, having concurrent severe
tochemistry (IHC) or expression medial problems and many more
data via RT-PCR (reverse tran- [14] (ClinicalTrials.gov Identifier:
scriptase polymerase chain reac- NCT00480025). Logistical con-
tion) might have been taken into siderations can also impact on the
consideration, which could give efficiency of patient recruitment;
differences in results. Additionally, slot availability can be a significant
the method of the IHC retrieval issue. In the absence of a treatment
system might give higher antigen slot, clinical sites are not motivated
numbers in certain cases [12]. An- to screen patients; even if patients
other potential confounder occurs are available, they may progress be-
if the stage of disease in the trial fore they can be recruited or move
patients differs from that of the on to other treatments.
reference population. For example, To conclude, TAA prevalence
for MAGE-A3 (Melanoma Antigen impacts the efficiency of trial re-
Gene A3) 31 out of 105 (29.5%) cruitment. To ensure that a trial can
stage I non-small cell lung cancers efficiently recruit, the number of
and 49 out of 99 (49.5%) stage centers open to recruitment should
II non-small cell lung cancers ex- reflect the frequency of the TAA in
pressed MAGE-A3 [13]. In clinical the patient population. Secondly,
practice, e.g. in the MAGRIT trial the trial should be designed to avoid
stage IB, II and IIIA NSCLC pa- unnecessary delays to recruitment;
tients have been recruited and 33% for example, expedited trial escala-
had MAGE-A3 positive tumors tion decisions can help increase the
(4210/12820) [14] (ClinicalTrials. recruitment speed.
gov Identifier: NCT00480025).
Therefore, it may be prudent to
generate this data in support of a TAA presentation in HLA
trial using biobank material, espe-
cially when selecting for less well Foreign- and self-peptides are pre-
known TAAs to estimate patient sented to the cells of the immune
recruitment requirements. system through the MHC system.

This presentation takes place in the the first FDA-approved autologous

groove of the MHC-I for endog- cellular immunotherapy, Sipuleu-
enous and intracellular peptides, cel-T, which uses autologous anti-
and for exogenous and extracellu- gen presenting cells cultured with
lar peptides the presentation takes a fusion protein consisting of PAP
place in the groove of the MHC-II. (prostatic acid phosphatase) linked
Through cross-presentation, exoge- to granulocyte-macrophage colo-
nous antigens can be presented by ny-stimulating factor. In the Phase
MHC-I and endogenous antigens 3 trial, patients with positive IHC
can be presented by MHC-II when staining of PAP in at least 25% of
they have been degraded by auto- cells were eligible for trial entry [17].
phagy [15]. Due to the properties Several approaches to generate al-
of the MHCs, only specific peptide logeneic DCs have been taken – for
epitopes can be presented; length example, DCs generated from cord
and structure play a major role for blood [18] have been tested in the
the capability of a peptide epitope clinic (ClinicalTrials.gov Identifi-
to be presented. Consequently, not er: NCT01373515) [19]. Another
all of the possible antigen fragments trial used embryonic stem cell-de-
are equally presented on the cell rived DCs [20], which were tested
surface. Furthermore, of the anti- in a clinical trial in a confirmed
gens that are presented, not all are HLA-A*02:01 positive popula-
equally immunogenic. As an ex- tion to guarantee at least one HLA
ample, Table 1 illustrates some of match with each receiving patient’s
the antigens that can be obtained immune system (ClinicalTrials.gov
from NY-ESO-1 and their relative Identifier: NCT03371485). How-
immunogenicity. Clearly, not all ever, these allogeneic cells must be
TAAs induce strong immunogenic- HLA matched to the recipient to
ity and the T cell response can be ensure that the DC can interact
destructive and non-destructive for with host T cells in a productive
the tumor [16]. When developing a way and the DC is not rejected by
therapeutic using TAA, the choice the host. In the case of autologous
of appropriate epitope selection is DCs, patient selection should be
essential if the intervention is going considered for the TAA; for alloge-
to induce highly immunogenic and neic DCs, patient selection should
tumor destructive T cells. ideally consider selection for TAA
and match for HLA.
PATIENT SELECTION FOR

CELL THERAPY TCR-T cell therapy & CAR-T
Autologous & allogeneic cell therapy
dendritic cell therapy Another potentially promising
form of cell therapy is T-cell recep-
TAA match with cell therapy is of tor engineered T-cells (TCR-T).
importance, as without the tumor TCR-T rely on the interaction of
expressing the antigen no specific peptide:MHC, formed by peptide
immune response will be possible. bound to MHC. To effectively kill,
One successful example of a cell the T-cell receptor must be matched
therapy using patient selection is with at least one HLA allele from
1418 DOI: 10.18609/cgti.2019.149

expert insight
ff TABLE 1
NY-ESO-1 epitope sequence, HLA presentation and immune response.
MHC Allele Epitope sequence Immune response Ref.
class
I HLA-A2 SLLMWITQCFL157–167 NY-ESO-1-specific CD8+ T [9]
cells, some stabilization of
disease, and regression of
individual metastases in some
patients
I HLA-A2 SLLMWITQC157–165 Very efficiently recognized by [34–36]
CD8+ T cells from HLA-A*0201
melanoma patients and epitope
identified in patients with spon-
taneous immunity
I HLA-A2 QLSLLMWIT155–163 Is poorly immunogenic and [37–39]
CD8+ CTLs recognizing this
epitope are rarely detected in
cancer patients
I HLA-A2 LMWITQCFL159–167 Not naturally processed [34]
II HLA-DPB1*0401–0402 SLLMWITQCFLPVF157–170 CD4+T cell responses were [40–42]
induced in a high proportion of
patients
This list is not inclusive of every epitope sequence that can be possibly presented, and only the principle of the relationship of epitope
processing and presentation is illustrated.
Colour indicates identical amino acid sequence in peptides.
the patient [21]. One of the tech- signaling moiety for T cell activa-
nologies is Adaptimmune’s SPEAR tion. Therefore, recognition and
T cells (ClinicalTrials.gov Identifi- activation is independent of MHC
er: NCT04044859), where patient presentation, binding is defined only
selection is done and eligibility for by the antibody domain [23]. No
the trial requires either positivity of selection for MHC is necessary, but
HLA-A*02:05 or HLA-A*02 allele, selection for the presence of the TAA
and tumor must show confirmed on the tumor cells would be of prime
MAGE-A4 expression. Therefore, importance for CAR-T therapy.
for effective therapeutic effect,
TCR-T therapy trials should ideally
select for TAA and HLA match. HLA general considerations
CAR-T cell therapy has made
great progress in CD19+ hemato- HLA genotypes are involved in the
logical malignancies [22] although response to treatment; certain HLA
recent trials in solid tumors have genotypes are associated with a more
failed to replicate these initial suc- favorable response than others [24].
cess. CAR-T therapies targeting Retrospective analysis of patients
antigens like EGRFvIII, IL-13Rα2, treated with immune checkpoint
HER2, EphA2 and GD2 are in clin- inhibitor showed improved overall
ical development [22]. CAR-T cells survival in patient with HLA-B44,
do not require selection for HLA, as whereas HLA-B*15:01 might im-
they are independent of MHC pre- pair T cell recognition of neoanti-
sentation. The binding works via an gens [24]. Others have shown that
antibody-derived domain for bind- HLA variations might be associated
ing with the TAA and the intracel- with adverse events to checkpoint
lular part is that of a TCR-derived inhibitors [25], and patients with a

specific genotype are more likely to rarely generated from FFPE tissue,
develop side effects due to their gen- but from snap frozen biopsies or bi-
otype. There is a clear therapeutic ra- opsies stored in RNA stabilizing re-
tionale to test cell and gene therapies agents. The latter two require fresh
in combination with checkpoint tumor biopsy, which places further
inhibitors and under these circum- constraints on trial design and logis-
stances, consideration should be giv- tics. For IHC assessment of TAAs, a
en to the effect of the genotype on central lab might be preferred over a
the combination therapy. The ideal local lab as local differences in stain-
HLA genotype for the cell and gene ing procedures, cut-off assessment
therapy might not be favorable for impacts upon the positivity rate. If
the combination agent. the use of local labs is unavoidable,
this can be solved by circulating
positive tissues between local labs
Practical implications for and getting concordance between
TAA & HLA selection the pathologist doing the readout of
the staining. HLA selection can be
Patient selection is a powerful tool done conveniently from peripheral
to help ensure that maximum pa- blood with which no issues with
tient benefit can be achieved. How- availability of tissue sample are nor-
ever, tissue based TAA selection re- mally encountered.
quires the availability of an adequate
amount of suitable tissue making
tumor accessibility a key consider- HLA loss on tumors &
ation; inaccessible tumors may not other mechanisms of tumor
be readily biopsied. Archival tumor disguise
biopsies might be considered if sta-
bility data are available which show Tumors are adept at hiding from
the TAA is stable in FFPE (for- their host’s immune system by uti-
malin-fixed paraffin-embedded). lizing numerous mechanisms; one
Also, data should be available that such mechanism that leads to tumor
the TAA expression is not altered resistance to therapy [27] is tumor
by previous or ongoing treatment cell MHC-I loss or downregulation
during which the archival tumor bi- to avoid recognition and elimina-
opsy has been taken. Additionally, tion of T lymphocytes [28]. Reduced
the biopsy itself has an intrinsic risk HLA expression as well as HLA loss
factor associated, especially in diffi- of heterozygosity by cancer cells
cult to access tumor locations like helps cancer cells escape and avoid
the lung. Increased risk is associated cytotoxic T lymphocytes [29,30].
with the lung biopsy, e.g., pneumo- As well as reducing the abundance
thorax and bleeding [26]. of antigens expressed on the cell it
The method of choice for TAA is also possible that proteins can be
selection is IHC when the protein processed differently in cancer com-
of interest is expressed on the cell pared with healthy cells resulting in
surface, or RT-PCR when either the presentation of unique antigens
there is no specific antibody avail- in MHC-I [31,32]. These aspects
able to detect the TAA in FFPE have been underexplored regarding
tissue or the TAA is only expressed patient selection until now. It might
intracellularly. RT-PCR samples are be worth considering including in
1420 DOI: 10.18609/cgti.2019.149

expert insight
patient selection when the cell and other exclusion criteria or progress
gene therapy is MHC dependent. If during screening, and these failures
there is a requirement for HLA en- need to be factored in when consid-
gagement by a cell and gene therapy ering optimal screening rates. Pa-
drug product, selection for presence tient selection can be complex, and
of MHC-I on tumor cells might be a balance must be struck between
considered. recruiting individuals who are likely
Other mechanisms of acquired to respond and have benefit from
resistance include defects in antigen the treatment, and the clinical fea-
processing and presentation result- sibility of the trial. Screening is a
ing in the loss of peptide presenta- major but often overlooked hurdle
tion in the MHC complex. In addi- to successful translation, but when
tion, loss of immunogenic antigens considered carefully, it can lead to
in general by the tumor can occur; fruitful clinical outcomes.
peptides of the tumor are displayed
in the MHC:peptide complex, but
no immunogenic peptides remain
and immunogenic response against TRANSLATIONAL INSIGHT
the tumor is abrogated [33]. Considerations for patient selection
for cell and gene therapy trials using
tumor associated antigen as target
in early phase development:
CONSIDERATIONS FOR
CLINICAL TRIAL DESIGN ff Selection of the right TAA
FOR EARLY PHASE taking into consideration the
STUDIES immunogenicity of the TAA,
Typically, Phase 1 trials focus upon selectivity and specificity, and
the safety of the investigational drug the heterogeneity in the tumor
product. However, efficacy assess-
ment and understanding the mech- ff Prevalence of the TAA and the
anism of action to aid development resulting impact on patient
decision-making is becoming more recruitment
important. The effort required to
recruit adequate numbers of po- ff Ideal selection strategy for
tentially responsive (correct TAA
autologous DCs and CAR-T cell
expression and HLA genotype) pa-
therapy should include selection
tients even for a small trial should
for TAA, allogeneic DCs and
not be under estimated. To maxi-
TCR-T cell therapy should include
mize the chance of success, careful
selection for TAA and HLA
consideration should be given to
the accuracy of any literature data
ff Selection for loss of HLA on
on TAA prevalence; generation of
tumors might be considered
pilot data should be considered.
Flexible cohort size and flexible tri-
al design will minimize lack of slot ff Accessibility of tumor and the
availability. Consideration should resulting choice of indication
be given to how to incentivize trial
sites to continually screen patients. ff Optimization of trial design to
Patients can also fail eligibility on ensure efficient recruitment



Attribution: Copyright © 2019 Stephanie Traub & David Edwards. Published by Cell and Gene Therapy Insights under Cre-
ative Commons License Deed CC BY NC ND 4.0.
REFERENCES
1. Siu D. Cancer therapy using tu- and melanoma. Blood 2013; 122(6): 10. Middleton D, Menchaca L, Rood H,
mor-associated antigens to reduce 863–71. Komerofsky R. New allele frequency
side effects. Clin. Exp. Med. 2009; database: http://www.allelefrequen-
6. Liang J, Ding T, Guo ZW et al., Ex-
9(3): 181–98. cies.net. Tissue Antigens 2003. 61(5):
pression pattern of tumour-associated
403–7.
2. Novellino L, Castelli C, Parmiani G. antigens in hepatocellular carcinoma:
A listing of human tumor antigens association with immune infiltration 11. Allele Frequencey Net Database:
recognized by T cells: March 2004 and disease progression. Br. J. Cancer http://www.allelefrequencies.net
update. Cancer Immunol. Immunoth- 2013; 109: 1031.
12. Ezaki T. [Antigen retrieval: its signifi-
er. 2005; 54(3): 187–207.
7. Williams TM. Human leukocyte cance and drawbacks in immunohis-
3. Hollingsworth RE, Jansen K. Turn- antigen gene polymorphism and the tochemistry]. Kaibogaku Zasshi 1996;
ing the corner on therapeutic cancer histocompatibility laboratory. J. Mol. 71(6): 615–28.
vaccines. npj Vaccines 2019; 4(1): 7. Diagnost. 2001; 3(3): 98–104.
13. Sienel W, Varwerk C, Linder A et
4. Graddis TJ, McMahan CJ, Tamman 8. Choo SY. The HLA system: genetics, al. Melanoma associated antigen
J, Page KJ, Trager JB. Prostatic acid immunology, clinical testing, and (MAGE)-A3 expression in Stages I
phosphatase expression in human tis- clinical implications. Yonsei Med. J. and II non-small cell lung cancer:
sues. Int. J. Clin. Exp. Pathol. 2011; 2007; 48(1): 11–23. results of a multi-center study. Eur.
4(3): 295–306. J. Cardiothorac. Surg. 2004; 25(1):
9. Adler LN, Jiang W, Bhamidipati
131–4.
5. Linette GP, Stadtmauer EA, Maus K et al. The Other Function: Class
MV et al. Cardiovascular toxicity II-Restricted Antigen Presentation by 14. Vansteenkiste JF, Cho BC, Vanakesa
and titin cross-reactivity of affin- B Cells. Front. Immunol., 2017; 8: T et al. Efficacy of the MAGE-A3
ity-enhanced T cells in myeloma 319–319. cancer immunotherapeutic as adju-
vant therapy in patients with resected
1422 DOI: 10.18609/cgti.2019.149

expert insight
MAGE-A3-positive non-small-cell 22. Liu B, Yan L, Zhou M. Target 31. Ma W, Vigneron N, Chapiro J et
lung cancer (MAGRIT): a ran- selection of CAR T cell therapy in al. A MAGE-C2 antigenic peptide
domised, double-blind, placebo-con- accordance with the TME for solid processed by the immunoproteasome
trolled, phase 3 trial. Lancet Oncol. tumors. Am. J. Cancer Res. 2019; is recognized by cytolytic T cells
2016; 17(6): 822–835. 9(2): 228–241. isolated from a melanoma patient
after successful immunotherapy. Int.
15. Gfeller D, Bassani-Sternberg M. Pre- 23. Chmielewski M, Hombach AA, Ab-
J. Cancer 2011; 129(10): 2427–34.
dicting Antigen Presentation-What ken H. Antigen-Specific T-Cell Acti-
Could We Learn From a Million vation Independently of the MHC: 32. Woods K, Knights AJ, Anaka M et
Peptides? Front. Immunol. 2018; 9: Chimeric Antigen Receptor-Redirect- al. Mismatch in epitope specifici-
1716–1716. ed T Cells. Front. Immunol. 2013; 4: ties between IFNgamma inflamed
371–371. and uninflamed conditions leads to
16. Ilyas S, Yang JC. Landscape of Tumor
escape from T lymphocyte killing in
Antigens in T Cell Immunother- 24. Chowell D, Morris LGT, Grigg CM
melanoma. J. Immunother. Cancer
apy. J. Immunol. 2015; 195(11): et al. Patient HLA class I genotype
2016; 4: 10.
5117–5122. influences cancer response to check-
point blockade immunotherapy. 33. Beatty GL, Gladney WL. Immune
17. Olson BM, McNeel DG. Sipuleu-
Science 2018; 359(6375): 582–587. escape mechanisms as a guide for
cel-T: immunotherapy for advanced
cancer immunotherapy. Clin. Cancer
prostate cancer. Open Access J. Urol. 25. Hasan Ali O, Berner F, Bomze D
Res. 2015; 21(4): 687–692.
2011; 3: 49–60. et al. Human leukocyte antigen
variation is associated with adverse 34. Dutoit V, Taub RN, Papadopoulos
18. Kumar J, Kale V, Limaye L. Umbil-
events of checkpoint inhibitors. Eur. KP et al. Multiepitope CD8(+) T
ical cord blood-derived CD11c(+)
J. Cancer 2019; 107: 8–14. cell response to a NY-ESO-1 peptide
dendritic cells could serve as an alter-
vaccine results in imprecise tumor
native allogeneic source of dendritic 26. Wiener RS, Wiener DC, Gould MK.
targeting. J. Clin. Invest. 2002;
cells for cancer immunotherapy. Stem Risks of Transthoracic Needle Biopsy:
110(12): 1813–22.
Cell Res. Ther. 2015; 6: 184. How High? Clin. Pulmon. Med.
2013; 20(1): 29–35. 35. Bownds S, Tong-On P, Rosenberg
19. van de Loosdrecht AA, van Wetering
SA, Parkhurst M. Induction of tu-
S, Santegoets SJAM et al. A novel 27. Vinay DS, Ryan EP, Pawelec G et al.
mor-reactive cytotoxic T-lymphocytes
allogeneic off-the-shelf dendritic cell Immune evasion in cancer: Mecha-
using a peptide from NY-ESO-1
vaccine for post-remission treatment nistic basis and therapeutic strate-
modified at the carboxy-terminus to
of elderly patients with acute myeloid gies. Semin. Cancer Biol. 2015; 35:
enhance HLA-A2.1 binding affinity
leukemia. Cancer Immunol. Immuno- S185–S198.
and stability in solution. J. Immuno-
ther. 2018; 67(10): 1505–1518.
28. Marincola FM, Jaffee EM, Hicklin ther. 2001; 24(1): 1–9.
20. Case CC, Kevin Nishimoto K, DJ, Ferrone S. Escape of human
36. Valmori D, Dutoit V, Liénard D et
Whiteley E, Srivastava R, Lebkowski solid tumors from T-cell recognition:
al. Naturally occurring human lym-
J. AST-VAC2: An Embryonic Stem molecular mechanisms and function-
phocyte antigen-A2 restricted CD8+
Cell-Derived Dendritic Cell Cancer al significance. Adv. Immunol. 2000;
T-cell response to the cancer testis
Immunotherapy. Mol. Ther. 2015; 74: 181–273.
antigen NY-ESO-1 in melanoma
23(Suppl. 1): S87.
29. Lin A, Yan WH. Human Leukocyte patients. Cancer Res. 2000; 60(16):
21. Zhang J, Wang L. The Emerg- Antigen-G (HLA-G) Expression in 4499–506.
ing World of TCR-T Cell Trials Cancers: Roles in Immune Evasion,
37. Jager E, Chen YT, Drijfhout JW et al.
Against Cancer: A Systematic Metastasis and Target for Therapy.
Simultaneous humoral and cellular
Review. Technol. Cancer Res. Treat. Mol. Med. 2015; 21(1): 782–791.
immune response against cancer-tes-
2019; 18: 1533033819831068–
30. Starling S. Immune editing shapes tis antigen NY-ESO-1: definition of
1533033819831068.
the cancer landscape. Nat. Rev. Im- human histocompatibility leukocyte
munol. 2017; 17: 729. antigen (HLA)-A2-binding peptide

epitopes. J. Exp. Med. 1998; 187(2): 40. Odunsi K, Qian F, Matsuzaki J et 42. Zarour HM, Storkus WJ, Bru-
265–70. al. Vaccination with an NY-ESO-1 sic V et al. NY-ESO-1 encodes
peptide of HLA class I/II specifici- DRB1*0401-restricted epitopes rec-
38. Romero P, Dutoit V, Rubio-Godoy
ties induces integrated humoral and ognized by melanoma-reactive CD4+
V et al. CD8+ T-cell response to NY-
T cell responses in ovarian cancer. T cells. Cancer Res. 2000; 60(17):
ESO-1: relative antigenicity and in
Proc. Natl Acad. Sci. 2007; 104(31): 4946–52.
vitro immunogenicity of natural and
12837–12842.
analogue sequences. Clin. Cancer Res.
2001; 7(3 Suppl): 766s–772s. 41. Jager E, Jäger D, Karbach J et al.
Identification of NY-ESO-1 epitopes AFFILIATIONS
39. Gnjatic S, Jäger E, Chen W et al.
presented by human histocompatibil-
CD8(+) T cell responses against a Stephanie Traub
ity antigen (HLA)-DRB4*0101-0103
dominant cryptic HLA-A2 epitope Centre for Drug Development,
and recognized by CD4(+) T lym-
after NY-ESO-1 peptide immu- Cancer Research UK
phocytes of patients with NY-ESO-
nization of cancer patients. Proc. stephanie.traub@cancer.org.uk
1-expressing melanoma. J. Exp. Med.
Natl Acad. Sci. USA 2002. 99(18):
2000; 191(4): 625–30. David Edwards
11813–8.
Centre for Drug Development,
Cancer Research UK
1424 DOI: 10.18609/cgti.2019.149


ADVANCED THERAPIES
EXPERT INSIGHT
The evolution of adeno-associated

virus capsids for CNS gene therapy
Steven J Gray
Adeno-associated virus (AAV) is emerging as a dominant gene therapy

delivery vehicle, and the broad tool kit of naturally-occurring AAV capsid
variants has allowed tailoring of approaches for specific applications. For
example, Glybera® (AAV1) is targeted to muscle, Luxterna™ (AAV2) is
targeted to the retina, and Zolgensma® (AAV9) is targeted to the central
nervous system (CNS). In the context of CNS gene therapy, the discovery
of AAV9 was largely responsible for a shift from direct intraparenchymal
brain injections to approaches that more globally target the brain, such as
intravenous injection and/or injection into the cerebrospinal fluid (CSF).
In fact, one could divide CNS gene therapy into ‘pre-AAV9’ and ‘post-
AAV9’ eras, due to the dramatic leap that this vector technology enabled.
One can envision a similar future leap coming, as lab-derived improve-
ments to capsids are being made that could further increase the efficien-
cy and specificity of CNS-directed gene therapy. Recent advancements
in AAV vectors are discussed.
DOI: 10.18609/cgti.2019.141
CNS GENE THERAPY: The first CNS-directed AAV administration of AAV2 vectors for
HOW WE GOT TO gene therapy trials started in the Canavan disease, Parkinson’s disease,
WHERE WE ARE NOW early 2000s, with stereotaxic and CLN2 Batten disease [1–13].
These approaches were well-tolerat- a wealth of potentially new vector

ed, and localized treatment effects tropisms to uncover [20]. By 2009-
could be seen. However, overall 2011, a critical discovery was inde-
the efficacy of the treatments was pendently described by three different
limited and generally not deemed laboratories that one of these novel
to be highly disease-modifying. capsids, AAV9, was able to cross the
Other naturally-occurring AAV blood brain barrier (BBB) [21–23].
serotypes were identified and char- Furthermore, the ability of an intra-
acterized, such as AAV5, AAVrh8, venously-injected AAV9 vector to
and AAVrh10, which demonstrat- target the brain was shown in these
ed greater spread from intraparen- initial three reports to translate from
chymal injection sites compared to mice to adult cats and non-human
AAV2 [14–16]. As an additional ad- primates. The technology was rapidly
vancement, novel injection param- adopted by laboratories around the
eters such as convection-enhanced world, with numerous publications
delivery were optimized to provide demonstrating substantial efficacy
greater and more controlled spread in small and large animal models.
of AAV vectors from a stereotaxic, In contrast to direct intracranial ap-
intraparenchymal injection [17,18]. proaches, the translation of dose and
While these represented improve- vector administration of intravenous
ments in the general approach to AAV9 to large animals and humans
CNS gene therapy, they were incre- was relatively straightforward. As a
mental and typically did not trans- result, after 5 years the first intrave-
late into a transformative benefit nous AAV9 clinical trial was initiated
to patients. Exceptions to this are for Spinal Muscular Atrophy (SMA)
cases where localized expression of in 2014. In 2017 the initial results
the transgene could be particularly from the Phase 1 SMA trial were
disease-modifying, such as in the published, documenting unequivo-
clinical trial for aromatic L-amino cal and unprecedented benefit for a
acid decarboxylase (AADC) defi- CNS disease from gene therapy [24].
ciency [19]. The AADC clinical trial In 2019, the AAV9/SMN1 vector
in particular was groundbreaking, became an FDA-approved drug,
as the clearest demonstration at the termed Zolgensma®. In the mean-
time of substantial benefit imparted time, the success of the SMA program
by an AAV vector for a CNS dis- prompted many similar AAV9 efforts
ease. Following bilateral stereotaxic to move into clinical trials, such as
injection of an AAV2/AADC vec- for Mucopolysaccharidosis (MPS)
tor to the putamen, patients gained Type IIIA, MPS IIIB, GM1 Gangli-
new motor function with a dramatic osidosis, and Danon disease (clinical-
improvement on quality of life. On trials.gov identifiers NCT02716246,
the whole, however, the approach NCT03315182, NCT03952637,
to target an organ the size of a hu- and NCT03882437, respectively).
man brain by multiple stereotaxic In short order, intravenous AAV9
injections progressed in a very lim- became a platform approach to treat
ited and incremental fashion, with multiple CNS disorders in a similar
treatment of most CNS disorders fashion.
impractical by this approach. It should be noted that intravenous
In 2006, the discovery of dozens injection of AAV9 has several major
of new AAV capsid variants provided limitations, including the high dose
1362 DOI: 10.18609/cgti.2019.141

expert insight
of vector that needs to be adminis- against AAV9, due to natural AAV

tered (and manufactured) per kilo- infections [27–29]. These patients
gram body weight, the substantially would be excluded from trials or
higher (>100x) biodistribution of treatments utilizing intravenous ad-
the vector to peripheral organs com- ministration of AAV9. One strategy
pared to the brain, and the relatively to overcome these naturally-occur-
high prevalence of natural antibodies ring antibodies would be to utilize
against AAV9 in the human popu- lab-engineered capsid variants that
lation. It was found that injection would be serologically distinct from
of AAV9 into the CSF could over- natural AAVs.
come many of these limitations, and Another limitation is the bio-
the intra-CSF approach with AAV9 distribution pattern of the AAV9
still translated well from rodents to vector. When administered IV, less
larger animals [25,26]. Similar to the than 1% of the vector is localized
expanding use of AAV9 in intrave- to the brain [21]. When adminis-
nous clinical trials to target the CNS, tered intrathecally, a large portion
AAV9 intrathecal trials initiated in of the vector distributes to periph-
2015 for Giant Axonal Neuropa- eral organs, with vector genome
thy, followed by trials for other dis- copy numbers in the liver exceeding
eases such as CLN6 Batten disease, that of the spinal cord near the site
MPS I, MPS II, and CLN3 Batten of injection [25]. Thus, AAV9 lacks
disease (clinicaltrials.gov identifiers specificity or even preferred tropism
NCT02362438, NCT03580083, for the CNS. Related to the biodis-
NCT03566043, NCT02725580, tribution of the vector, within the
and NCT03770572). Based on the CNS AAV9 has been described to
results of the SMA clinical trial and target neurons and astrocytes, and
numerous strong preclinical study to a lesser degree oligodendrocytes
results across many disease models, and endothelial cells. Although
the expectation is that the number of there is some inconsistency in the
AAV9-based clinical trials (intrave- literature, there are multiple reports
nous or intrathecal) will continue to that in juvenile or adult primates
expand at an accelerating pace. the AAV9 vector has a preference
for astrocytes over neurons especial-
ly when administered intravenously
[21,22,30].
LIMITATIONS OF AAV9 & There are some concerns about
THE NEED FOR BETTER potential toxicities using AAV9 vec-
AAV VECTORS tors. In the SMA clinical trial using
While AAV9 has transformed the IV-injected AAV9 vectors, transient
field of CNS gene therapy and can liver toxicity was noted, but it was
mediate transformative treatments responsive to steroids and could be
for many diseases, it has substan- managed with a prophylactic ste-
tial limitations. As mentioned, if roid regimen [24]. This is consistent
it is administered intravenously it with findings seen in other clinical
is susceptible to rapid neutraliza- trials administering AAV8 vectors
tion by circulating anti-AAV9 an- IV such as for hemophilia, where
tibodies. Approximately 17–47% dose-responsive liver toxicity was
of the human population has de- observed that could be resolved with
tectable neutralizing antibodies steroids [31]. A recent publication

from Jim Wilson’s laboratory docu- transduction efficiency would be

mented dorsal root ganglia toxicity needed to maximize therapeutic
in NHPs and pigs following admin- efficacy.
istration of an AAV9-like variant
[32], but this finding has not been
corroborated in other similar NHP
studies using AAV9. Of greater con- BEYOND AAV9:
cern was the rapid death of one of STRATEGIES TOWARD THE
the NHPs following IV administra- NEXT LEAP IN VECTOR
tion of the AAV9-like variant. This TECHNOLOGY
appears to have been an isolated and AAV9 is one of over 100 natural-
unexplained incident that may or ly-occurring AAV capsid variants
may not have been directly linked that have been isolated. While it is
to the AAV capsid. Regardless, these possible that another naturally-oc-
cautionary potential toxicities could curring AAV capsid could be iden-
presumably be managed better if a tified with superior CNS-targeting
more efficient AAV capsid was avail- capabilities, most efforts to devel-
able that would allow lower vector op a better AAV-based CNS vec-
doses to be equally therapeutic. tor have focused on creating novel
In terms of overall efficiency of laboratory-derived AAV capsids.
CNS gene transfer, ideally a vector Toward this end, there are 2 main
would target a majority of affect- strategies: 1) rational design and 2)
ed cells across the CNS. However, directed evolution.
with AAV9 there has never been a Rational design takes a hypoth-
report of saturating (near 100%) esis-driven approach to generate
transduction efficiency across the AAV capsid variants, utilizing struc-
CNS in any animal model, with the ture-function knowledge about
exception of studies dosing prenatal AAV capsid biology. In two inde-
or neonatal rodents. Rather, biodis- pendent approaches, strategies were
tribution numbers are typically be- pursued to increase the spread of
low 0.5 copies per cell on average, AAV vectors by knocking out their
meaning that large numbers of cells primary proteoglycan receptor. In
across the CNS are not receiving the first instance, Albright et al.
any vector DNA, let alone stably investigated the role of sialic acid
expressing the transgene. For the binding in the ability of an AAV1
sake of discussion, an assumption variant (AAV1RX) to cross the BBB
is made that intravenous doses have and transduce neurons in the CNS
a ceiling at roughly 1 x 1014 vg/kg, [33]. AAV1 normally binds cell sur-
and intrathecal doses might have a face proteoglycans with terminal
ceiling of a 10 mL injection volume sialic acid. When the variable re-
in humans (1 x 1015 vg total); thus, gion I from the BBB-crossing rh10
higher transduction efficiency can’t capsid was swapped into AAV1 to
easily be achieved simply by increas- create AAV1RX, it allowed AAV1
ing the dose. While it is clear from to cross the BBB while also strong-
multiple studies that this level of ly reducing the dependence of
transduction efficiency is sufficient AAV1RX to utilize sialic acid [34].
to provide at least some level of After testing a variety of AAV1 and
therapeutic benefit for many diseas- AAVrh10 mutants with different
es, for most CNS disorders a higher levels of dependence on sialic acid
1364 DOI: 10.18609/cgti.2019.141

expert insight
binding, Albright et al. proposed a have lower cross-reactivity to anti-

model whereby fine-tuning of sial- bodies resulting from natural AAV
ic acid binding becomes important infections [37]. Directed evolution
for CNS transduction after intra- approaches have been utilized to
venous administration. Some sialic generate novel AAV capsids with
acid binding was necessary for CNS greatly enhanced CNS transduction
transduction; however too much following intravenous administra-
sialic acid binding led to high liver tion in mice (PHP.B, [38]), efficient
tropism and reduced BBB crossing. transduction of oligodendrocytes
In the second approach to modi- following intracranial injection
fy primary proteoglycan receptor in rats and non-human primates
binding, Sullivan et al. introduced (AAV-Olig001, [39,40]) and en-
R585A and R588A mutations into hanced retrograde axonal transport
AAV2 (AAV2HBKO) to knock out in mice (AAV2-retro, [41]). While
its ability to bind heparin sulfate the PHP.B variant of AAV9 confers
proteoglycans, the primary receptor approximately 50-fold enhanced
for AAV2 [35]. Upon direct intra- CNS transduction compared to
cranial administration into the stri- AAV9 in mice following IV admin-
atum of mice, AAV2HBKO showed istration, unfortunately this greater
considerably larger transduced areas CNS transduction does not translate
compared to AAV2, presumably to primates [42–44]. In contrast, the
due to lower heparin binding at preferred oligodendrocyte tropism
the site of injection allowing bet- and degree of spread of Olig001 af-
ter spread. In a more unorthodox ter intraparenchymal injection into
third approach to rational design, the striatum does translate between
the Vandenberghe lab has taken rodents and non-human primates
an in silico strategy to reconstruct [40]. Thus, while directed evolution
ancestors of modern AAVs. One has the potential to generate valu-
reconstructed variant, termed An- able capsid variants with novel char-
c80L65, was able to cross the BBB acteristics, they have a mixed track
after intravenous administration in record of effectively translating out
adult mice, and transduce approx- of rodent models.
imately 3–4 times the number of
neurons and astrocytes compared to
AAV9 [36]. Overall, these examples
demonstrate the potential of ratio- EXPERT INSIGHT
nal design strategies to yield better Despite considerable effort to de-
CNS vectors. rive AAV vectors in the laboratory
Directed evolution approaches with superior abilities, and the first
have also shown considerable prom- use of a non-natural AAV capsid
ise to generate novel AAV capsids (AAV2.5) in a clinical trial reported
with substantially altered proper- in 2010 [45,46], the vast majority of
ties. These approaches utilize librar- AAV clinical trials still use unmod-
ies of AAV capsid variants produced ified capsids that can be found in
by random peptide integration, ran- nature. In the context of CNS gene
dom mutagenesis, and/or shuffling therapy, AAV9 is the gold standard
of multiple AAV capsid sequences. and has provided the first sugges-
Shuffled capsid AAV variants and tion of a ‘on-size-fits-all’ gene trans-
lab-engineered AAV capsids can fer approach to treat many CNS

disorders. While this is likely the transfer in non-human primates.

case for dozens of inherited CNS The gold standard that any new
disorders, the vast majority would capsid needs to meet is to be tested
benefit greatly from (or outright directly against AAV9 in non-hu-
require) better vector technology. man primates. If a novel AAV cap-
With the tools available to the field, sid is proven to have five- or ten-
it is likely that a vector will become fold greater transduction efficiency
available that is superior to AAV9 in broadly across the CNS compared
terms of specificity and/or efficien- to AAV9, following an intravenous
cy. However, at this time there are or intra-CSF injection in non-hu-
no published reports demonstrating man primates, we can expect anoth-
an AAV capsid more efficient than er monumental leap in our ability
AAV9 for widespread CNS gene to treat CNS disorders.

Disclosure and potential conflicts of interest: Dr. Gray has received patent royalty income from Asklepios Biopharma related
to the Olig001 capsid, and patent royalty income from Abeona Therapeutics related to CNS-directed AAV capsids.
Funding declaration: Dr Gray reports personal fees from Asklepios Biopharma, grants, personal fees and non-financial sup-
port from Abeona Therapeutics, grants, personal fees and non-financial support from Neurogene, outside the submitted
work. In addition, Dr Gray has a patent AAV vectors targeted to oligodendrocytes with royalties paid to Askleopios Biophar-
ma, and a patent AAV Vectors Targeted to the Central Nervous System with royalties paid to Abeona Therapeutics.

Attribution: Copyright © 2019 SJ Gray. Published by Cell and Gene Therapy Insights under Creative Commons License
Deed CC BY NC ND 4.0.
Submitted for peer review: Sep 4 2019; Revised manuscript received: Oct 11 2019; Publication date: Oct 29 2019.
REFERENCES
1. Janson C, McPhee S, Bilaniuk L et Hum. Gene Ther. 2002; 13(11): gene therapy for canavan disease.
al. Clinical protocol. Gene therapy 1391–412. Sci. Transl. Med. 2012; 4(165):
of Canavan disease: AAV-2 vector for 165ra163.
2. Leone P, Shera D, McPhee SW
neurosurgical delivery of aspartoacy-
et al. Long-term follow-up after 3. Hackett NR, Redmond DE,
lase gene (ASPA) to the human brain.
Sondhi D et al. Safety of
1366 DOI: 10.18609/cgti.2019.141

expert insight
direct administration of AAV2(CU) an open-label, phase I trial. Lancet 18. Varenika V, Kells AP, Valles F et al.
hCLN2, a candidate treatment for Neurol. 2008; 7(5): 400–8. Controlled dissemination of AAV
the central nervous system manifesta- vectors in the primate brain. Prog.
11. Mittermeyer G, Christine CW,
tions of late infantile neuronal ceroid Brain Res. 2009; 175: 163–72.
Rosenbluth KH et al. Long-term
lipofuscinosis, to the brain of rats
evaluation of a phase 1 study of 19. Hwu WL, Muramatsu S, Tseng SH et
and nonhuman primates. Hum. Gene
AADC gene therapy for Parkinson’s al. Gene therapy for aromatic L-ami-
Ther. 2005; 16(12): 1484–503.
disease. Hum. Gene Ther. 2012; no acid decarboxylase deficiency. Sci.
4. Barker RA. Continuing trials of 23(4): 377–81. Transl. Med. 2012; 4(134): 134ra61.
GDNF in Parkinson’s disease. Lancet
12. Marks WJ Jr, Bartus RT, Siffert J et 20. Gao G, Vandenberghe LH, Wilson
Neurol. 2006; 5(4): 285–6.
al. Gene delivery of AAV2-neurturin JM. New recombinant serotypes of
5. Christine CW, Starr PA, Larson for Parkinson’s disease: a dou- AAV vectors. Curr. Gene Ther. 2005;
PS et al. Safety and tolerability of ble-blind, randomised, controlled 5(3): 285–97.
putaminal AADC gene therapy for trial. Lancet Neurol. 2010; 9(12):
21. Gray SJ, Matagne V, Bachaboina L et
Parkinson disease. Neurology 2009; 1164–72.
al. Preclinical differences of intravas-
73(20): 1662–9.
13. Muramatsu S, Fujimoto K, Kato S cular AAV9 delivery to neurons and
6. Eberling JL, Jagust WJ, Christine et al. A Phase I Study of Aromatic glia: a comparative study of adult
CW et al. Results from a phase I safe- L-Amino Acid Decarboxylase Gene mice and nonhuman primates. Mol.
ty trial of hAADC gene therapy for Therapy for Parkinson’s Disease. Mol. Ther. 2011; 19(6): 1058–69.
Parkinson disease. Neurology 2008; Ther. 2010.
22. Foust KD, Nurre E, Montgomery CL
70(21): 1980–3.
14. Cearley CN, Vandenberghe LH, Par- et al. Intravascular AAV9 preferential-
7. Gasmi M, Herzog CD, Brandon EP ente MK et al. Expanded repertoire ly targets neonatal neurons and adult
et al. Striatal delivery of neurturin by of AAV vector serotypes mediate astrocytes. Nat. Biotechnol. 2009;
CERE-120, an AAV2 vector for the unique patterns of transduction 27(1): 59–65.
treatment of dopaminergic neuron in mouse brain. Mol. Ther. 2008;
23. Duque S, Joussemet B, Riviere C
degeneration in Parkinson’s disease. 16(10): 1710–8.
et al. Intravenous administration of
Mol. Ther. 2007; 15(1): 62–8.
15. Cearley CN, Wolfe JH. Transduction self-complementary AAV9 enables
8. Kaplitt MG, Feigin A, Tang C et al. characteristics of adeno-associated vi- transgene delivery to adult motor
Safety and tolerability of gene therapy rus vectors expressing cap serotypes 7, neurons. Mol. Ther. 2009; 17(7):
with an adeno-associated virus (AAV) 8, 9, and Rh10 in the mouse brain. 1187–96.
borne GAD gene for Parkinson’s Mol. Ther. 2006; 13(3): 528–37.
24. Mendell JR, Al-Zaidy S, Shell R et
disease: an open label, phase I trial.
16. Markakis EA, Vives KP, Bober J et al. al. Single-Dose Gene-Replacement
Lancet 2007; 369(9579): 2097–105.
Comparative transduction efficiency Therapy for Spinal Muscular Atro-
9. LeWitt PA, Rezai AR, Leehey MA of AAV vector serotypes 1-6 in the phy. N. Engl. J. Med. 2017; 377(18):
et al. AAV2-GAD gene therapy for substantia nigra and striatum of 1713–22.
advanced Parkinson’s disease: a dou- the primate brain. Mol. Ther. 2010;
25. Gray SJ, Nagabhushan Kalbur-
ble-blind, sham-surgery controlled, 18(3): 588–93.
gi S, McCown TJ et al. Global
randomised trial. Lancet Neurol.
17. Bankiewicz KS, Eberling JL, Kohut- CNS gene delivery and evasion of
2011; 10(4): 309–19.
nicka M et al. Convection-enhanced anti-AAV-neutralizing antibodies
10. Marks WJ Jr, Ostrem JL, Verha- delivery of AAV vector in parkinso- by intrathecal AAV administration
gen L et al. Safety and tolerabil- nian monkeys; in vivo detection of in non-human primates. Gene Ther.
ity of intraputaminal delivery of gene expression and restoration of 2013; 20(4): 450–9.
CERE-120 (adeno-associated virus dopaminergic function using pro-
26. Meyer K, Ferraiuolo L, Schmelzer
serotype 2-neurturin) to patients drug approach. Exp. Neurol. 2000;
L et al. Improving single injection
with idiopathic Parkinson’s disease: 164(1): 2–14.
CSF delivery of AAV9-mediated

gene therapy for SMA: a dose-re- 33. Albright BH, Simon KE, Pillai M rodents and nonhuman primates.
sponse study in mice and nonhuman et al. Modulation of Sialic Acid Acta Neuropathol. Commun. 2017;
primates. Mol. Ther. 2015; 23(3): Dependence Influences the Central 5(1): 47.
477–87. Nervous System Transduction Profile
41. Tervo DG, Hwang BY, Viswanathan
of Adeno-associated Viruses. J. Virol.
27. Mimuro J, Mizukami H, Shima M S et al. A Designer AAV Variant
2019; 93(11).
et al. The prevalence of neutralizing Permits Efficient Retrograde Access
antibodies against adeno-associated 34. Albright BH, Storey CM, Murlidha- to Projection Neurons. Neuron 2016;
virus capsids is reduced in young ran G et al. Mapping the Structural 92(2): 372–82.
Japanese individuals. J. Med. Virol. Determinants Required for AAVrh.10
42. Hordeaux J, Wang Q, Katz N et al.
2014; 86(11): 1990–7. Transport across the Blood-Brain
The Neurotropic Properties of AAV-
Barrier. Mol. Ther. 2018; 26(2):
28. Boutin S, Monteilhet V, Veron P PHP.B Are Limited to C57BL/6J
510–23.
et al. Prevalence of serum IgG and Mice. Mol. Ther. 2018; 26(3): 664–8.
neutralizing factors against adeno-as- 35. Sullivan JA, Stanek LM, Lukason MJ
43. Hordeaux J, Yuan Y, Clark PM et
sociated virus (AAV) types 1, 2, 5, 6, et al. Rationally designed AAV2 and
al. The GPI-Linked Protein LY6A
8, and 9 in the healthy population: AAVrh8R capsids provide improved
Drives AAV-PHP.B Transport across
implications for gene therapy using transduction in the retina and brain.
the Blood-Brain Barrier. Mol. Ther.
AAV vectors. Hum. Gene Ther. 2010; Gene Ther. 2018; 25(3): 205–19.
2019; 27(5): 912–21.
21(6): 704–12.
36. Hudry E, Andres-Mateos E, Lerner
44. Matsuzaki Y, Konno A, Mochizuki
29. Ellsworth JL, O’Callaghan M, Rubin EP et al. Efficient Gene Transfer
R et al. Intravenous administration
H et al. Low Seroprevalence of Neu- to the Central Nervous System by
of the adeno-associated virus-PHP.B
tralizing Antibodies Targeting Two Single-Stranded Anc80L65. Mol.
capsid fails to upregulate transduc-
Clade F AAV in Humans. Hum. Gene Ther. Methods Clin. Dev. 2018; 10:
tion efficiency in the marmoset brain.
Ther. Clin. Dev. 2018; 29(1): 60–7. 197–209.
Neurosci. Lett. 2018; 665: 182–8.
30. Bevan AK, Duque S, Foust KD et al. 37. Li W, Asokan A, Wu Z et al. Engi-
45. Mendell JR, Campbell K, Rod-
Systemic gene delivery in large spe- neering and selection of shuffled AAV
ino-Klapac L et al. Dystrophin
cies for targeting spinal cord, brain, genomes: a new strategy for produc-
immunity in Duchenne’s muscular
and peripheral tissues for pediatric ing targeted biological nanoparticles.
dystrophy. N. Engl. J. Med. 2010;
disorders. Mol. Ther. 2011; 19(11): Mol. Ther. 2008; 16(7): 1252–60.
363(15): 1429–37.
1971–80.
38. Deverman BE, Pravdo PL, Simpson
46. Bowles DE, McPhee SW, Li C et al.
31. Nathwani AC, Tuddenham EGD, BP et al. Cre-dependent selection
Phase 1 gene therapy for Duchenne
Rangarajan S et al. Adenovirus-as- yields AAV variants for widespread
muscular dystrophy using a transla-
sociated virus vector-mediated gene gene transfer to the adult brain. Nat.
tional optimized AAV vector. Mol.
transfer in hemophilia B. N. Engl. J. Biotechnol. 2016; 34(2): 204–9.
Ther. 2012; 20(2): 443–55.
Med. 2011; 365(25): 2357–65.
39. Powell SK, Khan N, Parker CL et al.
32. Hinderer C, Katz N, Buza EL et Characterization of a novel adeno-as-
al. Severe Toxicity in Nonhuman sociated viral vector with preferential
Primates and Piglets Following High- oligodendrocyte tropism. Gene Ther. AFFILIATIONS
Dose Intravenous Administration of 2016. Steven J Gray
an Adeno-Associated Virus Vector Department of Pediatrics, Uni-
40. Mandel RJ, Marmion DJ, Kirik D
Expressing Human SMN. Hum. versity of Texas Southwestern
et al. Novel oligodendroglial alpha
Gene Ther. 2018. Medical Center, Dallas, TX, USA
synuclein viral vector models of
steven.gray@utsouthwestern.edu
multiple system atrophy: studies in
1368 DOI: 10.18609/cgti.2019.141


ADVANCED THERAPIES
REGULATORY PERSPECTIVE
Clinical trials of advanced therapy

investigational medicinal products
in Spain: preparing for the European
clinical trials regulation
Juan Estévez Álamo, Marcos Timón, Cristina
González Gómez-Platero, Carmen Doadrio Abad,
Marta Velasco González, María Yolanda de
Mingo Ballesteros, María Ángeles Martín de la
Sierra San Agustín & María Antonia
Serrano Castro
Clinical trials (CT) of Advanced Therapy Medicinal Products are a reality

worldwide. Although ATMPs are still very innovative therapies, it is inter-
esting to investigate what relevant information can be obtained from the
analyses of authorized CT and the investigated products. The aim of this
study was to follow the evolution of CT with Advanced Therapy investi-
gational Medicinal Products (ATiMP) authorized in Spain from May 2004
to June 2019 on the basis of information available at the Spanish Agency
for Medicinal Products and Medical Devices and their real status (also
taking into consideration their status in three different official Registries).
We will also discuss how sponsors and Authorities can prepare for the
coming new clinical trial regulation and take advantage of the opportu-
nities it may present.

DOI: 10.18609/cgti.2019.147
INTRODUCTION System), which is currently under

Clinical trials (CT) are essential to development and will enormously
support the authorization of me- simplify communications between
dicinal products and are the basis sponsors and Member States (MS).
for their appropriate use in normal In the meantime, a Voluntary Har-
clinical practice. The knowledge of monisation Procedure (VHP), set
ongoing or finished CT is essen- up by the Clinical Trials Facilita-
tial in order to favor better designs tion Group, serves as a pilot for the
for future clinical investigations. coordinated EU assessment of CT
There is a CT European legislation applications foreseen in Regulation
in force since 1st May 2004 (Di- 536/2014(2). The VHP was intro-
rective 2001/20/CE) [1] that has duced in order to achieve harmo-
been reviewed in CT Regulation nized assessments and decisions on
536/2014 [2]. Under both legisla- clinical trials in the EU, and spon-
tions, the conduct of a clinical tri- sors are encouraged to use it [6,7].
al with a medicinal product in any Advanced Therapy medicinal
European Union (EU) Member products (ATMP) are a particular-
State requires prior national autho- ly innovative medicinal class that
rization. In the case of Spain, such includes gene therapy medicinal
authorization is given by the Span- products (GTMP), somatic cell
ish Agency for Medicinal Products therapy medicinal products (sCT-
and Medical Devices (AEMPS) MP), tissue engineered products
after internal CT review, provided (TEP), and combined products
that one Ethics Committee (Com- (tissue or cell associated with a
mittee of Ethics of the Investigation device). The legal and regulatory
with medicinal products – known framework for ATMPs in the EU
in Spanish as CEIm) has also given (ATMP Regulation 1394/2007) [8]
a favorable opinion. The informa- came into force on 31st December
tion that sponsors currently need to 2008 and defined common rules
provide to the competent authori- for this very innovative group of
ties (AEMPS in the case of Spain) medicinal products that have to
to be published in either EU CT comply with specific quality re-
Register (EU CTR) [3] or Spanish quirements [9].
Register on Clinical Studies (REec) Clinical investigation of ATMP
[4] is shown in Table 1. has additional difficulties due to
The above-mentioned reviewed the nature of some of the products.
Regulation came into force in June For instance, many cell-based AT-
2014 and introduced important MPs are autologous (i.e. prepared
changes; among them, a European from material taken from the pa-
coordinated assessment of CT and tient) which makes standardization
additional transparency require- a real challenge for manufacturers.
ments with respect to terms cur- In addition, Advanced Therapy
rently in force and shown in Table investigational Medicinal Prod-
1, related to CT information and ucts (ATiMP) have to comply not
documents that will be available to only with the general legislations
the public [5]. However, its whole for clinical trials and ATMP, but
applicability is still pending the also with legislation from different
availability of the new EU CT Por- frameworks, such as the tissues and
tal and Database (CT Information cells Directive (Directive 2004/23)
1432 DOI: 10.18609/cgti.2019.147

regulatory perspective
ff TABLE 1
Information to be provided by the sponsors to the AEMPS to be published in the EU CTR and REec
according to EU and national legislation [2,14].
CT information to be provided by Publication in EU CTR Rules for publication of CT in REec

the sponsor to NCA to be public
Summary of CT design (since All CT authorized since 1st May All CT authorized since 1st January
initial CT application for 2004. However, Phase 1 CT not 2013. Phase 1 CT not including pe-
authorization) including pediatric population are diatric population may only include
not published abbreviated information, if this is the
sponsor choice
Date of CT start (within following Yes Yes
15 days)
Date of end of recruitment in No Yes
Spain (within following 15 days)
Dates of end of CT in Spain and Yes Yes and in case of premature end,
of global CT end, clarifying if the reasons are also published after
end is premature or not (within assessment
following 15 days)
Temporary halts affecting Spain Yes, reasons are not published Yes, and reasons are also published
clarifying if global or not and after assessment
reasons (within the following 15
days)
Summary of CT results (within Results to be loaded in EudraCT and Results of Phase 1 CT not including
one year of the date of global CT also submitted to the AEMPS pediatric population are currently
end) not public
for the donation, procurement and be able to comply with the new EU
testing of the starting materials to be CT legislation.
converted into cell-based medicinal
products [8], or that for genetically
modified organisms (GMO) (Di-
rectives 2001/18 and/or 2009/41) METHODOLOGY
[8] when the product belongs to this All valid clinical trial applications
category. In this latter case, lack of on ATiMP received at AEMPS since
harmonization between different 1st May 2004 until 30th June 2019
GMO authorities across the EU has have been considered for the anal-
prompted the development of com- ysis. Description of the characteris-
mon voluntary procedures for some tics of the clinical trials authorized
categories of products [10]. by AEMPS takes into account the
In spite of the above difficul- information available on the inter-
ties, Spain has been identified as nal CT database of this Agency re-
the Member State with the highest gardless of substantial amendments.
number of CT on ATMPs [6,11]; ATiMPs were classified according
taking advantage of this, we set out to the definitions set out in Regu-
to analyze characteristics of these lation 1394/2007 and Directive
CT. This article is focused on the 120/2009, and following the prin-
analysis of the characteristics of the ciples highlighted in the reflection
clinical trials on ATiMP authorized paper on classification of ATMPs
by AEMPS from 1st May 2004 to published by the Committee for
30th June 2019, also paying atten- Advanced Therapies (CAT) [12].
tion to the IMP being investigated. Products used in clinical trials be-
The purpose is to identify possible fore these definitions were published
areas of improvement in order to have been reclassified according to

these criteria in order to have a har- developed by AEMPS, defines not

monized approach. Products con- only the cell type but a number of
taining or consisting on genetically additional attributes (tissue of or-
modified cells (e.g., CAR-T cells) igin, expansion in culture, other
are generally considered GTMPs manipulations, etc.) as a pre-requi-
in the EU, except when the genetic site to the final identification. The
modification is not directly linked to analysis of the products’ characteris-
the therapeutic activity of the cells. tics showed in this paper took into
In Spain, all medicinal products account our register of PEI ATiMP.
without a marketing authorization Number and characteristics of
in any country of the European the ATiMP in the authorized CT,
Economic Area (EEA) that contain owners of such products (commer-
an active substance or combination cial, i.e., pharmaceutical companies,
of substances not included in any or non-commercial, i.e., facilities
of the medicinal products marketed within the National Health System)
in Spain need to obtain a number and number of CT per ATiMP have
of Product under clinical investiga- been analyzed.
tion (known in Spanish as PEI) and The following aspects have been
sponsors need to cross-reference analyzed and verified for all autho-
this number for every new CT ap- rised ATiMP CT during this period
plication. A PEI covers all pharma- on the basis of information available
ceutical forms and strengths of an on CT Applications and electronic
investigational product. ATiMPs CT Dossier Documents:
and especially, cell-based ATiMPs,
are very complex and sometimes it ff Type of sponsor (commercial or
is difficult to determine whether a non-commercial [14];
particular product should be con-
sidered the same or a different PEI. ff Distribution of the CT according
For instance, a different PEI number to type of ATiMP (sCTMP, GTMP,
is required when the same cell prod- TEP) and GMO character;
uct changes from an autologous to
an allogeneic use. Normally, when ff Phase of CT as indicated by the
changes were introduced – e.g., in sponsor;
final formulation – the new prod-
uct was considered as being differ-
ff Therapeutic area of investigation
ent. When substantial changes were
taking into account MeSH terms
introduced in manufacturing with-
used by EudraCT [15] to define
out a proper comparability study,
the Therapeutic Area;
the final products were also con-
sidered as being different. Different
manufacturers require different PEI ff Population (i.e., adults (18–64
numbers unless equivalence of the years), elderly (>65 years) and/or
products is shown through strong pediatrics (less than 18 years);
comparability studies. To clearly de-
fine and identify the different drug ff National or International
substances used in clinical trials in character taking into
Spain, a guideline on nomenclature consideration geographical
of cell-based medicinal products distribution of the participant
was followed [13]. This guideline, sites;
1434 DOI: 10.18609/cgti.2019.147

ff According to the number of sites ff End of Recruitment: date of end of

in Spain, single-site or multi-site recruitment has been provided;
CT;
ff Temporarily Halted: temporary
ff CT status and availability of halt date has been received;
results.
ff Prematurely Ended (According
In addition, publication has been
to the Regulation No 536/2014
checked in REec [4], EU CTR [3]
[2], early termination of a clinical
and ClinicalTrials.gov [16]. For CT
trial means the premature end of
found in these three registers, consis-
a clinical trial due to any reason
tency in terms of the status displayed
before the conditions specified in
taking into consideration Table 2 and
the protocol are complied with)/
the availability of results with respect
Completed: end of trial date has
to the information received in AE-
been received. CT having included
MPS has been also reviewed. Taking
a significantly lower than planned
into account that the status in REec
number of subjects or those not
reflects the situation of the CT in
having completed all parts defined
Spain, possible differences in the re-
in the protocol have also been
cruitment status among Registers for
considered as prematurely ended
international CT have not been con-
for this analysis, even if the end
sidered incoherent. Consistency of
was not notified as premature;
the end of trial status for internation-
al trials has also taken into account,
if the global end of trial had been ff Unknown: in cases where there
notified to the AEMPS. The latest have not been notifications by the
available status from the following sponsor within the last 2 years.
is shown, provided that at least one
Results have been considered as:
notification has been received for the
trial in the last 2 years: ff Yes: available results
ff Not Initiated: CT authorized,
without reception of date of start; ff No: no available results
ff Recruiting: date of CT start ff NA (not applicable): when the

received; CT has not finished yet or when
ff TABLE 2
Equivalence of CT status among the different CT Registries checked on this research.
REec EUCTR Clinicaltrials.gov

– – Unknown
Not initiated Not yet recruiting
Recruiting (or restarted) Recruiting
Ongoing (or restarted)
Enrolling by invitation
End of recruitment Active, not recruiting
Temporarily halted Temporarily halted Suspended
Prematurely ended Prematurely ended Terminated
Withdrawn (no patients)
Completed Completed Completed

the global end date of the CT sponsors (in fact, from 2005 to
has taken place within the 2010 ATiMP CT sponsors are only
last year and the deadline to Spanish). From 2011 onwards,
submit official results has not sponsors from other countries
been reached yet, according started to sponsor ATiMP CT in
to National Law and European Spain. The proportion of interna-
Regulation tional sponsors increased up until
2018, when approximately half of
Verification of all CT status and the sponsors were from other coun-
results has taken place during Sep- tries (see Figure 2). In this sense, the
tember 2019. figures from 2018 are especially re-
vealing, since 29 out of the CT run
by a sponsor not based in Spain
RESULTS were authorized. CT on GTMP are
mainly run by commercial sponsors
During the period from 1st May and have a greater relevance since
2004 to 30th June 2019, AEMPS 2016, showing a great peak in 2018
received 331 valid CT applications coinciding with international CT
on ATiMP that represent 2.9% of increase, as shown in Table 3. This is
the total number of valid CT ap- consistent with the evolution in the
plications in that period. Status for type of ATMP being investigated
these CT applications on 20th July (see Figure 1).
2019 (data analysis starting date) Most ATiMP CT are early
was: 290 authorized, 14 rejected, 19 phases: Phase 1, Phase 1/2 and
withdrawn and 8 under assessment. Phase 2 represent 80.3% of all au-
Spain takes part in approximately thorized CT during the study pe-
23% of ATiMP CT authorized in riod. Non-commercial sponsors are
Europe [3]. more focused on early phases clini-
cal trials, as opposed to commercial
sponsors who conduct the majority
CT according to type of Phase 2/3, 3 and 4 trials. There is
of ATiMP, sponsor & no significant relationship between
international character the type of therapy and phases of
CT. Most national clinical trials
The distribution of authorized CT have non-commercial sponsors
according to the type of ATiMP (88%). International trials are
and sponsor along the analyzed mostly Phase 2 or 3, while national
period is shown in Figure 1. Total trials are Phase 1 and 2. Non-com-
numbers of ATiMP CT and distri- mercial sponsors mostly conduct
bution according to type of prod- single-site trials while multi-site
uct, sponsor, country of the spon- trials are conducted by commercial
sor, international character, number sponsor (see Table 4).
of sites in Spain and phase is shown Regarding the Voluntary Har-
in Table 3. monisation Procedure (VHP),
Clinical investigation of ATiMP available for CT planned to be con-
shows an important increase since ducted in two or more EU Mem-
2010. Until 2013, it was mostly ber States, Spain has participated
focused on sCTMP and TEP and in the evaluation of seven CT with
driven by Spanish non-commercial ATiMP by this procedure, five of
1436 DOI: 10.18609/cgti.2019.147

ffFIGURE 1
Cumulative data on authorized ATiMP CT in Spain (2005–2018).
which were with tissue engineered inflammatory bowel disease with 15

products. Most of these CT were CTs, gastrointestinal system cancer
Phase 2 or Phase 3, and both com- with 13 CTs, and heart failure, isch-
mercial and non-commercial spon- emic and non-ischemic/cardiomy-
sors used this procedure. Prevalence opathy with 13 CTs were the most
of these phases is also observed in all frequent (Table 5).
CT evaluated by VHP to date [6,7]. 17.9% of ATiMP CT include
pediatric population (together
with adults and/or elderly people
CT according to targeted [10.7%]; exclusively pediatric pop-
disease & CT population ulation [7.2%]). Most of these tri-
als investigated GTMPs (48.1%)
Globally, the most predominant and have a commercial sponsor
therapeutic area was cancer (31.7%) (61.5%). Regarding indication on
followed by cardiovascular (14.8%) exclusively pediatric CT, cancer re-
and musculoskeletal (10.0%) dis- mained the most prevalent (47.6%)
eases. 63% of cancer CT investi- followed by congenital, hereditary,
gated GTMP, while 35.9% of them and neonatal diseases and abnor-
investigated sCTMP. However, malities (e.g., spinal muscular atro-
95.5% of CT on the cardiovascular phy, Fanconi anemia, osteogenesis
area and 96.4% of those on mus- imperfecta, inborn errors of urea
culoskeletal diseases investigated cycle, etc.; 38.1%).
TEP. The indications of leukemia/ CT were equally performed in
lymphoma/myeloma with 32 CTs, both women and men.

ff TABLE 3
Number of CT on ATiMP according to type of product, sponsor, country of the sponsor, international
character, number of sites in Spain and phase.
No. OF CT ON No. OF CT ON No. OF CT ON TEP No. OF CT ON

ATiMP (N = 290) sCTMP (N = 99) (N = 107) GTMP (N = 84)
Sponsor
Commercial 124 (42.8%) 32(32.3%) 21 (19.6%) 71 (84.5%)
Non-commercial 166 (57.2%) 67(67.7%) 86 (80.4%) 13 (15.5%)
Sponsor country
Spain 209 (72.1%) 80 (81.0%) 94 (87.8%) 35 (41.7%)
USA 49 (16.9%) 9 (9.0%) 2 (1.9%) 38 (45.2%)
Rest of EU 29 (10.0%) 7(7.0%) 11 (10.3%) 11 (13.1%)
Israel 3 (1.0%) 3(3.0%) 0 (0%) 0 (0%)
International
Yes 109 (37.6%) 27 (27.3%) 22 (20.6%) 60 (71.4%)
No 181 (62.4%) 72 (72.7%) 85 (79.4%) 24 (18.6%)
Sites in Spain
Multi-site 156 (54.1%) 52 (52.5%) 49 (45.8%) 55 (65.5%)
Single-site 133 (45.9%) 46 (46.5%) 58 (54.2%) 29 (34.5%)
Phase
Phase 1 71 (24.5%) 19 (19.2%) 28 (26.2%) 24 (28.6%)
Phase 1/2 33 (11.4%) 14 (14.1%) 1 (0.9%) 18 (21.4%)
Phase 2 128 (44.2%) 46 (46.5%) 61 (57.0%) 21 (25.0%)
Phase 2/3 3 (1.0%) 1 (1.0%) 1 (0.9%) 1 (1.2%)
Phase 3 52 (17.9%) 19 (19.2%) 14 (13.1%) 19 (22.6%)
Phase 4 3 (1.0%) 0 (0%) 2 (1.9%) 1 (1.2%)
ATiMP Yescarta® and Zalmoxis®), USA

(Zolgensma®) or Spain (NC1).
Regarding investigated products, NC1 is a product prepared on a
168 different ATiMP products are non-routine basis according to spe-
being investigated in the authorized cific quality standards, and used
CT, being 54 (32.1%) TEP, 51 within Spain in a hospital under
(30.3%) sTCMP and 49 (29.2%) the exclusive professional respon-
GTMP, while 14 (8.3%) products sibility of a medical practitioner,
are being investigated as both sTC- in order to comply with an indi-
MP and TEP. They include ATiMP vidual medical prescription for a
currently having a marketing au- custom-made product for an indi-
thorization in the EU (Alofisel®, vidual patient, authorized by the
Holoclar®, Imlygic®, Kymriah®, AEMPS, as defined in Regulation
ff TABLE 4
National or International character and number of sites in Spain
for CT on ATiMP according to type of sponsor.
Non-commercial Commercial
National (N = 181) 159 (87.8%) 22 (12.2%)
Single-site 103 5
Multi-site 56 17
International (N = 109) 7 (6.4%) 102 (93.6%)
Single-site 2 23
Multi-site 5 79
1438 DOI: 10.18609/cgti.2019.147

ffFIGURE 2
Authorized ATiMP CT per Sponsor country.
[EC] No 1394/2007 [8] and Royal system, while the other nine pertain
Decree 477/2014 [17]). to a pharmaceutical company.
40 of the GTMP and two of It is remarkable that 91 out of 168
the sCTMP are GMO and they ATiMP belong to non-commercial
are being investigated in 75 CT. owners; most of them are sTCMP
Their Product Owners are mainly and TEP, in consistency with the
commercial (85.7%). Sponsors for type of CT run by non-commercial
GMO CT are from Spain and USA sponsors. On the other hand, most
(43% each), while sponsors for the of the products that belong to com-
other 14% are from other European mercial owners are GTMP.
countries.
The number of CT per PEI has
ranged from 1 (for 112 ATiMP) to CT status
9 (for 2 ATiMP). 21 products have
been investigated on at least four CT, According to EU legislation, spon-
including Alofisel®, Imlygic® and sors have the obligation to report
Kymriah®, which have a marketing National Competent Authorities
authorization in the EU, and NC1 relevant dates and information for
(authorized in Spain according to the CT in order to make its status
the national legislation for ‘hospital transparent. Certain information,
exemption’). Twelve out of these 21 such as the annual safety report,
products are manufactured in a facil- should be provided yearly along
ity pertaining to the national health the CT duration. In addition, the

ff TABLE 5
Number of trials in the four most investigated disease areas: cancer, cardiovascular diseases, musculo-
skeletal diseases and digestive system diseases.
Disease area Diseases Number of trials

Cancer Leukemia/lymphoma/myeloma 32 (34.8%)
Gastrointestinal system cancer 13 (14.1%)
Brain cancer 9 (9.8%)
Skin cancer 8 (8.7%)
Bladder or renal cancer 5 (5.4%)
Respiratory system cancer 4 (4.3%)
Prostate cancer 3 (3.3%)
Breast cancer 3 (3.3%)
Others 15 (16.3%)
TOTAL 92
Cardiovascular diseases Heart failure, ischemic and non-ischemic/ 13 (30.2%)
cardiomyopathy
Myocardial infarction/coronary 12 (27.9%)
Limb ischemia and peripheral arterial disease 12 (27.9%)
Stroke 6 (14.0%)
TOTAL 43
Musculoskeletal diseases Joint or bone arthrosis 11 (37.9%)
Bone defects 9 (31.1%)
Spinal defects or pathology 6 (20.7%)
Tendinopathy/ligament defects 3 (10.3%)
TOTAL 29
Digestive system diseases Inflammatory bowel diseases (perianal fistules) 15 (65.3%)
Hepatic failure/cirrhosis 7 (30.4%)
Fecal incontinence 1 (4.3%)
TOTAL 23
CT should be published in the August 2011, are not published in

EU CTR [3] and in the REec [4] any of these registers, and only 117
as is indicated in Table 1. Howev- out of 290 CT are published in all
er, considering the international of them.
character of part of the CT and Considering the information
the relevance of ClinicalTrials. available at AEMPS, the status of
gov [16] also for EU investigators ATiMP CT is reflected in Table 6.
and sponsors, the registration sta- The public status displayed for
tus of all ATiMP CT authorized the CT published in REec, EU
in Spain has also been checked in CTR [3] and ClinicalTrials.gov [16]
that register. was consistent in most cases (98 out
All authorized ATiMP CT since of 117 cases). For 12 CT the status
1st January 2013 (n=165) are reg- in ClinicalTrials.gov [16] was not
istered and published in REec [4]. updated according to the informa-
In 230 out of 290 CT, Spain is a tion available in the AEMPS and in
participating country in a record in 6 cases the information seemed to
the EU CTR [3]. The 60 not pub- be more updated in ClinicalTrials.
lished CT are phase I and do not in- gov [16] than in the AEMPS. Co-
clude pediatric population. 235 CT herence of the CT status in REec [4]
are registered in ClinicalTrials.gov and the status for Spain in EU CTR
[16]. Only 8 CT, authorized before [3] is seen but this is expected since
1440 DOI: 10.18609/cgti.2019.147

AEMPS is responsible for keeping it These results are published in the

updated. EU CTR except if the CT is only
During the analyzed period, Phase 1 and does not include pedi-
142 CT were ended by the spon- atric population. Results should also
sor, 47 of which were terminated be submitted to EU Member States
earlier than expected. Table 7 shows National Competent Authorities.
the number of prematurely ended In both cases, the deadline for this
clinical trials in relation to the rea- submission is within a year of the
sons for it. The main reasons for end of trial (usually the last visit of
early termination included lack of the last patient). For CT authorized
recruitment (53.2%) and business in Spain since 1st January 2013, the
reasons (21.3%). It is remarkable summary of the results is published
that lack of efficacy and safety at REec, except for Phase 1 CT not
were the reason for the early ter- including pediatric population, for
mination in just five and two CT, which there is limited information
respectively. published.
Regarding the time elapsed from According to these criteria, AE-
the authorization until the early ter- MPS should have received the sum-
mination, less than 1 year passed mary of results for 73 completed
in 31.9%, between 1 and 3 years CT for which the due date has ex-
in 27.6% and more than 3 years in pired. Twenty-three of these trials
40.4%. Finally, the predominant have a commercial sponsor while
therapeutic areas on these CT were 50 of them have a non-commer-
cancer (40%), and cardiovascular cial one. Results for only 45 CT
diseases, coinciding with the two (61.6%) have been received and
most investigated therapeutic areas only 14 CT have results publicly
for ATiMP CT. available either in REec (n= 4) [4],
in EU CTR (n=7 plus intermediate
results for 1 CT, authorized since
ATiMP CT results 2007) [3] and/or in ClinicalTrials.
gov (n=5 CT authorized since April
According to the EU legislation, 2012) [16].
sponsors should upload a summa- With respect to prematurely
ry of CT results in EudraCT [15]. ended CT, results are expected to
ff TABLE 6
Status of ATiMP CT according to information available in AEMPS.
CT status Number of CT per status according to

information available in AEMPS
Not initiated 16
Recruiting 74
End of recruitment 30
Temporary halted 3
Prematurely ended 46
Completed 79
Unknown 42
Unknown status for 17 CT authorized before 2013 might be due to the fact that for
these trials part of the information could be in a paper File on CT, not checked for this
review, which was previous to the current AEMPS database that contains all documents
in the CT dossier presented in an electronic format.

ff TABLE 7
Number of prematurely ended clinical trials distributed by reasons
for early termination (CT authorized before and after 2013).
Reasons No. of CT (auth. < 2013) No. of CT (auth. ≥ 2013)

Lack of 20 5
recruitment
Business reasons 4 6
Enough data 1 1
Lack of efficacy 2 3
Safety 0 2
Other reasons 3 0
TOTAL 30 17
be made public as soon as possible a more systematic way of describ-

within the year following the end ing the ATiMP under development
of the CT, especially information could be of great interest as AEMPS
related to safety or lack of efficacy, has previously highlighted [13].
unless the CT ends with no rele- Regulation 1394/2007 [8] set a
vant subject participation. Follow- clear and common framework for
ing these criteria, results have been ATMPs in the EU. This regulation,
accessible for the AEMPS in most amongst other things, added tissue
of the cases (72.8%). Results are engineered products as a new class
published in a CT register in seven of ATMPs to the previously defined
cases, including all CT stopped due gene therapy and somatic cell ther-
to lack of efficacy. apy medicinal products. This meant
that many cells and/or tissue-based
treatments that were already in clin-
ical use outside the pharmaceutical
DISCUSSION legislation, became regulated as me-
This article analyses all CT on dicinal products when the ATMP
ATiMP authorized by AEMPS regulation came into force (Decem-
from 2004 to 2019. This group of ber 2008).
290 studies represents around 22% Publication of the ATMP regu-
of the entire CT with ATiMP con- lation clearly had a positive effect
ducted in the EU. As already stated, on the number of clinical trials in
Spain is one of the countries in the Spain, as observed by the increase
world with more significant activity from 2010 in Figure 1. At that time,
in this area [6,11]. most of the trials had an academ-
The interest in identifying spe- ic sponsor and research was mainly
cific numbers for IMP investigated focused on TEP and sCTMP (Fig-
is highlighted, since these numbers ure 1). The number of trials stayed
are difficult to find due to the natu- relatively high up to 2014, when a
ral evolution of the names in prod- clear drop is observed, presumably
ucts under clinical development. due to the restrictions in public in-
However, in the field of cell and vestments in clinical research associ-
tissue research, where the nature ated with the worst years of the eco-
and origin of the cells as well as the nomic crisis. Recovery in number of
autologous or allogeneic character clinical trials started from 2016, but
could greatly influence the efficacy this time driven mainly by commer-
and safety of the products, having cial research (Figure 1).
1442 DOI: 10.18609/cgti.2019.147

The huge increase in the last in organizing late-phase CT that

2 years is a clear reflection of the normally involve hundred or even
success in gene therapy clinical re- thousands of patients, and require
search directly related to industry the involvement of many investiga-
(including big pharma) becoming tors and sites in different countries.
increasingly interested in the de- An example of these difficulties is
velopment of advanced therapies highlighted in the article by inves-
(Figure 1) and expressed in the tigators of the study MESEMS [22]
availability of several GTMP mar- which due to financial constrains
keted both in the EU and USA has been designed to merge par-
since 2015 [18–20]. This increase tially independent clinical trials. In
has occurred despite the additional fact, 91 out of 168 ATiMP investi-
difficulties imposed on most gene gated are produced within the Na-
therapy medicinal products because tional Health System in non-com-
of their consideration as Genetical- mercial GMP-compliant facilities.
ly Modified Organisms (GMO). It is remarkable that sometimes the
Application of the GMO regula- results of early academic studies are
tion [21] in the EU to clinical trials the basis for the further develop-
with most gene therapy products ment of a marketed product as was
means the involvement of a differ- the case for Alofisel [23]. The fact
ent competent authority to assess that only 6 of the non-commercial
the potential environmental effects products were GTMP could be due
of such products, complicating the to the more complex manufactur-
authorization procedure. This has ing process of these products.
a greater impact on multinational On the other hand, 82.8% of
trials, because each MS has its own the CT on Phase 2/3, 3 and 4 and
GMO competent authority and the 84.5% of the CT on GTMP are
procedures are far from harmonized run by commercial sponsors. Addi-
across the EU. In an effort to unify tionally, commercial CT stand out
criteria and streamline the process in their international (82.3%) and
several activities have been initiat- multicenter characteristics, as can
ed, which have already yielded a be seen in Tables 3 & 4. This is in
number of consensus documents line with the characteristics neces-
[10]. Although these documents are sary for confirmatory CT required
not obligatory, a good number of to support the application for the
MS (including Spain) have adopted marketing authorization of any me-
them on a voluntary basis. This is dicinal product.
expected to ease the administrative Our results show that 290
burden for clinical trials authoriza- ATiMPs CT were conducted in
tions of medicinal products con- different therapeutic areas. Can-
taining or consisting of GMOs. cer, with almost a third of the trials
The main characteristics of ac- (31.7%), cardiovascular (14.8%)
ademic studies (n=166), as can be and musculoskeletal (10.0%) dis-
seen in Tables 3 & 4, are: early Phase eases were the most prevalent ones.
1, 1/2 and 2 CT (94%), national Cancer diseases were also preva-
(95.8%), unicentric (62%) and fo- lent for pediatric patients (47.6%)
cused on the investigation of sCT- due to their severity and scarce
MP or TEP (92.2%). This is consis- therapeutic alternatives, as well
tent with the logistical difficulties as congenital diseases (38.1%).

Our results regarding indications Only two CT on a GTMP have

are in line with the search in the been stopped due to safety reasons,
main international CT databases one of them with no patient partic-
performed by Hanna et al. [24] or ipation because the safety problem
the review of ATiMP CT between was detected prior to enrolment.
2004 and 2010 by Maciulatitis et al. The time elapsed between the date
[11] in the EU, or even in Europe- of authorization of the CT and the
an-country publications such as by premature end seems to be related
the Czech Republic [25]. Although to the reason for stopping the tri-
there were multiple indications, it al: the CT ended because of a lack
is important to highlight refractory of recruitment tended to last lon-
and recurrent characteristics, and ger while those CT ended due to
the scarce and poor therapeutic al- business or safety reasons usually
ternatives for them (e.g., refracto- stopped within the first year.
ry and metastatic tumours, critical Currently, all CT should be up-
limb ischemia, non-revascularisable loaded to the European CT data-
myocardium, complex perianal base EudraCT and be published
fistulas, osteogenesis imperfecta, in the EU CTR, except for those
spinal muscular atrophy, etc.). It is Phase 1 studies not including pe-
remarkable that 30.8% (n=16) of diatric population. It is remarkable
pediatric ATiMP CT were autho- that the commercial confidentiali-
rized in 2018. Indeed, 36.4% of ty principle that supported hiding
authorized ATiMP CT in 2018 in- these Phase 1 CT not in the EU
cluded pediatric population, while CTR for many years is not appli-
only 14.9% of total authorized CT cable in CT.gov, where 48 out of
(on any kind of medicinal product) 60 non-pediatric Phase 1 CT are
in 2018 included pediatric popula- published.
tion. This seems consistent with the Under-reporting of CT results is
increase in the GTMP CT. a serious problem which has been
Regarding the reasons that mo- frequently highlighted [26–28].
tivated a premature end for a CT, Publication of results is not only an
shown in Table 7, the main reason ethical issue but a legal requirement
was lack of recruitment. The impor- [1,2,14]. Our analysis shows that
tance of a correct design that takes AEMPS has received an on-time
into account all actual population report on the results for 65 out of
characteristics should be pointed 105 expected. 36 of these were with
out, in order to avoid lack of recruit- a commercial sponsor and 69 from
ment after all the efforts deployed to a non-commercial sponsor. This
set up the trial. It is important to means a rate of proper reporting of
indicate that 30 of the prematurely 72.2% for commercial and 53.6%
ended CT were authorized before for non-commercial sponsors, con-
2013, and for 20 of them, reasons firming the lower rate of reporting
for stopping the CT were related results for academic sponsors previ-
to a lack of feasibility in recruiting ously shown [29].
the necessary patients. This seems When looking into the structured
to indicate that nowadays, protocols format required to provide the results
are better adapted to true patient for the EU CTR [3] and Clinical-
characteristics, which marks an im- Trials.gov [16] registers, it is remark-
provement in their quality. able the fact that only 15 CT from
1444 DOI: 10.18609/cgti.2019.147

commercial sponsors have results academic research of medicinal prod-

uploaded to EU CTR (out of 103 ucts. It is remarkable that the vast
registered CT for which such results majority of CT for which the status
could be expected). In addition, only is unknown are old and non-com-
nine CT (five from non-commercial mercial CT, showing that there has
sponsors and four from commercial been an increasing interest in com-
sponsor) out of 104 registered and for plying with regulations thanks to
which results could be expected have efforts from several networks such as
results uploaded to ClinicalTrials.gov STARS Project (Strengthening Train-
[16] with a similar structured format. ing of Academia in Regulatory Sci-
However, for 27 CT (12 from aca- ence) [30] – an initiative funded by
demic sponsors and 13 from com- the EU with the aim of analyzing and
mercial ones) not having loaded the improving training of non-commer-
structured results as required, there is cial sponsors on regulatory science in
at least one paper in a medical jour- order to have better and faster access
nal focused on the results referenced to innovative therapies. In the case of
within the record of the CT in Clin- Spain, AEMPS has created the Office
icalTrials.gov [16] as ‘Publications for the Support of Innovation and
automatically indexed to this study Knowledge with Medicinal Products
by ClinicalTrials.gov [16] Identifier [31], responsible for giving technical
(NCT Number)’. The lapsed time and administrative advice to every in-
between the end of CT date and the novative project that is going to take
publication has been 1 year for only place in Spain or EU. Within this
one CT, longer than 1 and less than 2 Office we can find a specific Office
years for nine CT, longer than 2 and for non-commercial research, where
up to 3 years for five CT, and longer special support from the beginning
than 3 years for the other ten CT. of projects is usually needed.
These data could point to diffi- Regulation 536/2014 [2] is still
culties in in completing the current not fully applicable in Europe.
structured summary of results, es- In the meantime, all stakeholders
pecially for academic sponsors. In should get prepared to work ac-
addition, this shows the need to cording to the new CT Regulation
increase awareness around the legal rules. This implies (among oth-
need for sponsors to organize CT er things) having a single national
activity in such a way that a sum- contact in the EU to organize access
mary of results could be available for the sponsor’s users to the future
within 1 year of the end of CT date EU CT database and portal on the
(usually the date of last visit of the basis of the who does what principle
last patient). It would be important (viewer, preparer or submitter roles)
that editors of Medical Journals do and taking into account the future
not reject the publication of CT re- transparency rules [5] and the prin-
sults due to the public availability of ciple of having single consolidated
the aforementioned legally required documents for all MS [32] when
summary in official CT registers. preparing the CT dossier.
As this sample includes a big por- Transparency should be seen as
tion of non-commercial trials, this an opportunity to identify serious
concern may not only be specific health problems not yet investigat-
to ATiMP research but can also re- ed, to identify known risks to be
flect the general difficulties related to avoided/minimized in future CT, to

facilitate recruitment and to coop- between different MS, as described

erate with other sponsors (as well as above. These issues will be taken
many other positive things). How- into account for the implementa-
ever, to see this benefit, all stake- tion of the new Regulation.
holders should commit to com- Problems highlighted here es-
ply with this principle that is very pecially for non-commercial trials
much emphasized in the new EU may not be specific to ATiMP re-
legislation. search, but can also reflect the gen-
In the EU, the EudraCT num- eral difficulties related to academic
ber is a unique identifier necessary research on medicinal products due
for all CT on medicinal products. to the large number of this type of
It would be very helpful if medical sponsor represented in this sample.
Journals always required the inclu- In Europe, there are several ini-
sion of the EudraCT number, to- tiatives ongoing trying to facilitate
gether with any other relevant iden- CT under the scope of the new reg-
tifier, in any publications related to ulation. Discussions on possible im-
this type of CT with participation provement of VHP, simplification
of EU sites. Currently, this number in the Environmental assessment of
is only present in about 56% of the GMOs, the ‘Strengthening training
CT records identified in Clinical- of academia in regulatory sciences
Trials.gov [16], but it would be very and supporting regulatory scientif-
helpful if sponsors could reference ic advice’ (STARS) project, and an
this number every time the CT is update of the guidance related to
identified for a CT Register. the CT Regulation in volume 10
VHP [7] has been the basis for Eudralex [33] are among them.
the coordinated evaluation proce-
dure established in EU Regulation
536/2014 [2]. For this reason, it is
the perfect place for active adap- CONCLUSION
tation to the changes that will be Clinical research on ATMP has
implemented by the new Europe- seen a clear increase, especially on
an Clinical Trials Regulation. By GTMP, during the last few years.
using VHP, sponsors will not only This increase has been in parallel
get experience on the European co- with an improvement in the quality
ordinated assessment but could also of CT, highlighted with the rising
influence possible improvements on number of multi-site and interna-
the application of the future legis- tional CT (also a consequence of
lation itself with real cases that are the increasingly commercial spon-
presented to us on a day-to-day soring of CT).
basis. However, the VHP was only Our analysis also shows some
used by a minority of the interna- difficulties in complying with reg-
tional CT with ATiMP conducted ulatory requirements, especially for
in Spain. This may reflect a percep- non-commercial sponsors. In this re-
tion of a higher complexity for this gard, it is notable that there are sever-
procedure, especially by academic al initiatives at a European and Span-
researches. In the case of GTMP, ish level, such as the STARS Project
application of the GMO regula- [30] and European Commission ini-
tion may have also interfered with tiatives to unify GMO requirements
a harmonized assessment process [10] in the EU, or the Office for the
1446 DOI: 10.18609/cgti.2019.147

Support of Innovation and Knowl- sponsors in order to ease the regu-

edge with Medicinal Products [31] latory framework burden and pro-
in Spain, that are trying to facilitate mote clinical research in the EU.
clinical research. Since this regulation is not yet fully
Last but not least, it should be applicable, all stakeholders still have
noted that Regulation 536/2014 [2] time to adapt their workflows and
is intended to be an instrument for national legislations to the new way
cooperation between EU MS and that lies ahead.

Acknowledgements: The authors thank all current and past administrative staff of the Cllinical Trial Division on human
medicinal products in AEMPS for their excellent work on updating the clinical trials database. Comments of Dr. Cesar Her-
nandez García on the manuscript were also greatly appreciated.
Disclosure and potential conflicts of interest: All authors currently work in the AEMPS and have no conflict of interest to
declare. The views expressed in this article are the personal views of the authors and may not be understood or quoted as
being made on behalf of or reflecting the position of the AEMPS or working parties.

Attribution: Copyright © 2019 Juan Estévez Álamo, Marcos Timón Jiménez, Cristina González Gómez-Platero, Carmen
Doadrio Abad, Marta Velasco González, Yolanda de Mingo Ballesteros, María Ángeles Martín de la Sierra San Agustín &
María Antonia Serrano Castro. Published by Cell and Gene Therapy Insights under Creative Commons License Deed CC BY
NC ND 4.0.
Article source: Invited.
Revised manuscript received: Oct 18 2019; Publication date: Nov 8 2019.
REFERENCES
1. European Parliament, Council of the medicinal products for human use. repealing Directive 2001/20/EC. Of-
European Union. Directive 2001/20/ Official Journal L 121, 01/05/2001 P. ficial Journal L 2014;158:1-76
EC of the European Parliament and 0034 - 0044.
3. EU Clinical Trials Register. European
of the Council of 4 April 2001 on
2. European Parliament, Council of Medicines Agency: http://clinicaltri-
the approximation of the laws, regu-
the European Union. Regulation alsregister.eu
lations and administrative provisions
(EU) No 536/2014 of the European
of the Member States relating to the 4. Spanish registry of clini-
parliament and of the council of 16
implementation of good clinical prac- cal studies (REec), 2013:
April 2014 on clinical trials on me-
tice in the conduct of clinical trials on h t t p s : / / w w w. a e m p s . g o b . e s /
dicinal products for human use, and

investigacionClinica/medicamentos/ Products, European Commission 16. ClinicalTrials.gov [Internet].

docs/directrices-NSA-invest-tera- 2017: https://ec.europa.eu/health/ Bethesda (MD): National Library
pia-celular-en.pdf?x92596 documents/eudralex/vol-4_en of Medicine (US). 2000 Feb 29:
http://clinicaltrials.gov/ct/show/
5. Appendix, on disclosure rules, to 10. Medicinal products: Advanced
NCT00287391?order=1
the “Functional specifications for therapies, European Commission,
the EU portal and EU database to 2019: https://ec.europa.eu/health/ 17. Agencia Estatal Boletín Oficial del
be audited - EMA/42176/2014” human-use/advanced-therapies_en. Estado. [Royal Decree 477/2014,
(EMA/228383/2015, 2 October of 13 June, regulating advanced
11. Maciulaitis R, D’Apote L, Buchanan
2015): https://www.ema.europa.eu/ therapy medicinal product non-in-
A, Pioppo L, Schneider CK. Clinical
en/documents/other/appendix-dis- dustrial manufacturing]: https://
development of advanced therapy me-
closure-rules-functional-specifica- w w w. b o e . e s / d i a r i o _ b o e / t x t .
dicinal products in Europe: evidence
tions-eu-portal-eu-database-be-audit- php?id=BOE-A-2014-6277 Spanish
that regulators must be proactive.
ed_en.pdf
Mol. Ther. 2012; 20(3): 479–82. 18. Ginn SL, Amaya AK, Alexander IE,
6. Boran T, Menezes-Ferreira M, Reis- Edelstein M, Abedi MR. Gene ther-
12. Reflection paper on classification of
chl I et al. Clinical Development apy clinical trials worldwide to 2017:
advanced therapy medicinal products
and Commercialization of Advanced An update. J. Gene Med. 2018; 20(5):
(EMA/CAT/600280/2010 rev.1, 21
Therapy Medicinal Products in the e3015.
May 2015): https://www.ema.europa.
European Union: How Are the Prod-
eu/en/documents/scientific-guideline/ 19. Hartmann J, Schüßler-Lenz M, Bon-
uct Pipeline and Regulatory Frame-
reflection-paper-classification-ad- danza A, Buchholz C. Clinical devel-
work Evolving? Hum. Gene Ther. Clin.
vanced-therapy-medicinal-products_ opment of CAR T cells-challenges
Dev. 2019; 28(3): 126–35.
en-0.pdf and opportunities in translating in-
7. Clinical Trials Facilitation Groups novative treatment concepts. EMBO
13. Agency of Medicines and Medical
Guidance document for sponsors Mol. Med. 2017; 9(9): 1183–97.
Devices (AEMPS). Guideline of the
for a Voluntary Harmonisation Pro-
Spanish Agency of Medicines and 20. Nappi F, Capone F, Galli M. Gene
cedure (VHP) for the assessment of
Medical Devices for the nomenclature therapy success: how to keep its prom-
multinational Clinical Trial Appli-
of active substances in advanced ther- ises and improve. A regulatory per-
cations. Heads of Medicines Agen-
apies investigational medicinal prod- spective. Cell Gene Ther. Insights 2019;
cies: Version 4 (CTFG//VHP/2016/
ucts containing cells. Madrid, 2013 5(6): 719–25.
Rev6, June 2016): https://www.hma.
September: https://www.aemps.gob.
eu/fileadmin/dateien/Human_Med- 21. Commission Directive 2001/18/EC
es/investigacionClinica/medicamen-
icines/01-About_HMA/Working_ on the deliberate release into the envi-
tos/docs/directrices-NSA-invest-tera-
Groups/CTFG/2016_06_CTFG_ ronment of genetically modified organ-
pia-celular.pdf?x15721
VHP_guidance_for_sponsor_v4.pdf isms and repealing Council Directive
14. Royal Decree 1090/2015, of 4 Decem- 90/220/EEC. OJ L 2001; 106: 1–38.
8. European Parliament, Council of the
ber, regulating clinical trials with medic-
European Union. Regulation (EC) No. 22. Uccelli A, Laroni A, Brundin L et al.
inal products, Ethics Committees for
1394/2007 of the European Parliament MEsenchymal StEm cells for Multiple
Investigation with medicinal products
and of the Council on Advanced Ther- Sclerosis (MESEMS): a randomized,
and the Spanish Clinical Studies Reg-
apy Medicinal Products and amending double blind, cross-over phase I/II clin-
istry: https://www.aemps.gob.es/leg-
Directive 2001/83/EC and Regulation ical trial with autologous mesenchymal
islacion/espana/investigacionClinica/
(EC) No. 726/2004. Official Journal L stem cells for the therapy of multiple
docs/Royal-Decree-1090-2015_4-De-
2007; 324: 121–137. sclerosis. Trials 2019; 20(1): 263.
cember.pdf?x45974
9. EudraLex, The Rules Governing 23. Alofisel European Public Assessment
15. European clinical trials database
Medicinal Products in the Europe- Report. European Medicines Agency,
(EudraCT): https://eudract.ema.eu-
an Union, Volume 4: Guidelines on 2019: https://www.ema.europa.eu/
ropa.eu/
Good Manufacturing Practice spe- en/medicines/human/EPAR/alofisel
cific to Advanced Therapy Medicinal
1448 DOI: 10.18609/cgti.2019.147

24. Hanna E, Remuzat C, Auquier P, Tou- 32. Transition of Clinical Trials to Reg- Use, Agencia Española de Medica-
mi M. Advanced therapy medicinal ulation (EU) No. 536/2014: CTFG mentos y Productos Sanitarios,
products: current and future perspec- Best Practice Guide for sponsors of Calle Campezo, 1, Edificio 8,
tives. J. Mark Access Health Policy 2016; multinational clinical trials with dif- 28022 Madrid, Spain
4. ferent protocol versions approved
Carmen Doadrio Abad
in different Member States under
25. Koci Z, Boran T, Krupa P, Kubinova Clinical Trials Division, Depart-
Directive 2001/20/EC that will
S. The current state of advanced ther- ment of Medicines for Human
transition to Regulation (EU) No.
apy medicinal products in the Czech Use, Agencia Española de Medica-
536/2014. [internet] London: Head
Republic. Hum. Gene Ther. Clin. Dev. mentos y Productos Sanitarios,
of Medicines Agencies, 2018: https://
2018; 29(3): 132–47. Calle Campezo, 1, Edificio 8,
www.hma.eu/fileadmin/dateien/Hu-
28022 Madrid, Spain
26. De Angelis C, Drazen JM, Frizelle FA man_Medicines/01-About_HMA/
et al. Clinical trial registration: a state- Working_Groups/CTFG/2018_05_ Marta Velasco González
ment from the International Com- CTFG_Best_Practice_Guide_for_ Pharmacoepidemiology and Phar-
mittee of Medical Journal Editors. sponsors_of_transition_multination- macovigilance Division, Depart-
Lancet 2004; 364(9438): 911–2. al_clinical_trials.pdf. ment of Medicines for Human
Use, Agencia Española de Medica-
27. Moorthy VS, Karam G, Vannice KS, 33. EudraLex, The Rules Governing mentos y Productos Sanitarios,
Kieny MP. Rationale for WHO’s new Medicinal Products in the Europe- Calle Campezo, 1, Edificio 8,
position calling for prompt reporting an Union, Volume 10, Clinical trials 28022 Madrid, Spain
and public disclosure of interven- guidelines, European Commission
2006: https://ec.europa.eu/health/ María Yolanda de Mingo
tional clinical trial results. PLoS Med.
documents/eudralex/vol-10_es Ballesteros
2015; 12(4): e1001819.
Clinical Trials Division, Depart-
28. Ross JS, Mulvey GK, Hines EM, ment of Medicines for Human
Nissen SE, Krumholz HM. Trial pub- Use, Agencia Española de Medica-
lication after registration in Clinical- AFFILIATIONS mentos y Productos Sanitarios,
Trials.Gov: a cross-sectional analysis. Calle Campezo, 1, Edificio 8,
Juan Estévez Álamo
PLoS Med. 2009; 6(9): e1000144. 28022 Madrid, Spain
Author for correspondence
29. Goldacre B, DeVito NJ, Heneghan C Clinical Trials Division, Depart- María Ángeles Martín de la Sierra
et al. Compliance with requirement ment of Medicines for Human San Agustín
to report results on the EU Clinical Use, Agencia Española de Medica- Clinical Trials Division, Depart-
Trials Register: cohort study and web mentos y Productos Sanitarios, ment of Medicines for Human
resource. BMJ 2018; 362: k3218. Calle Campezo, 1, Edificio 8, Use, Agencia Española de Medica-
28022 Madrid, Spain mentos y Productos Sanitarios,
30. Strengthening Regulatory Science jesteveza@aemps.es Calle Campezo, 1, Edificio 8,
(Stars): https://www.csa-stars.eu/ 28022 Madrid (Spain)
Marcos Timón
Work-Packages-1695.html
Biological Products, Advanced María Antonia Serrano Castro
31. [Office for the Support of Innova- Therapies and Biotechnology Di- Clinical Trials Division, Depart-
tion and Knowledge with Medic- vision, Department of Medicines ment of Medicines for Human
inal Products].Spanish Agency for for Human Use, Agencia Española Use, Agencia Española de Medica-
Medicines and Medical Devices, de Medicamentos y Productos mentos y Productos Sanitarios,
2017: https://www.aemps.gob.es/me- Sanitarios, Calle Campezo, 1, Edifi- Calle Campezo, 1, Edificio 8,
dicamentos-de-uso-humano/ofici- cio 8, 28022 Madrid, Spain 28022 Madrid Spain
na-de-apoyo-a-la-innovacion-y-con-
Cristina González Gómez-Platero
ocimiento-sobre-medicamentos/
Clinical Trials Division, Depart-
[Spanish]
ment of Medicines for Human


ADVANCED THERAPIES
INTERVIEW
Keys to success for foundation:

industry clinical development
collaborations
BRIAN FISKE joined The Michael J. Fox Foundation for Parkinson’s
Research (MJFF) in 2004. As Senior Vice President, Research Programs,
Brian co-manages a team of professionals who stay closely linked to the
Parkinson’s community in order to develop an aggressive and innovative
agenda for accelerating research and drug development for Parkinson’s
disease. This ensures that MJFF priorities reflect and best serve the ulti-
mate needs of patients. Brian regularly meets with academic and industry
scientists around the world to identify promising ideas to support, providing
troubleshooting and ongoing management of projects as they go forward.
He currently oversees the teams focused on MJFF’s strategies for develop-
ing disease-modifying and symptomatic therapies for Parkinson’s patients.
Brian earned an undergraduate degree in biology from Texas A&M University
and a PhD in Neuroscience from the University of Virginia. After complet-
ing postdoctoral research at Columbia University, Brian spent several years
as an editor for the prestigious scientific journal, Nature Neuroscience. He
brings this broad experience and knowledge to the Foundation to help bring
new treatments to people with Parkinson’s.
DOI: 10.18609/cgti.2019.145
QQ Can you firstly give us some background on The

Michael J. Fox Foundation for Parkinson’s Research
(MJFF)’s involvement to date in the clinical
development of gene therapy product candidates?
BF: The Foundation has been around since the year 2000 and
from day one we have been focused on accelerating and enabling
therapeutic development for the more than one million people liv-
ing with Parkinson’s disease (PD) in the USA today, and the many
more patients living with the disease globally.
I’ve personally been with the Foundation for 15 years and it’s been amaz-
ing to watch the evolution of the therapeutic pipeline for PD over that time
and how robust and compelling it has become. There are a lot of diverse
approaches being tested for PD, including traditional pharmacological
treatments but also innovative approaches using gene therapy as well as
some cell-based therapies.
In the early years of the Foundation, some of the leading therapeutic
approaches included brain tissue transplantation therapy and the delivery
of certain types of growth factors and other proteins in the brain. One of
our first experiences in helping with the clinical development of a gene
therapy came around this time when we began to work with a compa-
ny called Ceregene. They were developing a gene therapy approach for a
growth factor called neurturin – their hope was to deliver this potentially
protective growth factor into the brains of people with PD to help keep
dopamine-producing cells alive. Unfortunately, the trials testing neurturin
did not show benefits so the program was halted. But it represented some
of our first experiences working with a gene therapy program.
Since that time, we’ve continued to support lots of different approaches.
Not all have been gene therapies, but we have certainly seen that field con-
tinue to grow and evolve, and when you look at the pipeline today, there
are at least a handful of gene therapy approaches in clinical development. A
couple of these (Voyager Therapeutics/Neurocrine and Oxford BioMedica/
Axovant) are trying to deliver some of the synthetic machinery for making
more dopamine in the brain, while other companies are continuing down
the path of delivering protective factors – one recent example is Prevail
Therapeutics who have a gene therapy program for PD with GBA1 muta-
tion (PD-GBA).
So we’re certainly starting to see some interesting movement and explo-
ration of gene therapy in PD.
QQ What does the Foundation seek to bring to its clinical

development collaborations in terms of capabilities
and expertise?
BF: One of the ways we can provide support is financially. Since

our early days, we’ve had mechanisms in place to provide grant funding for
1392 DOI: 10.18609/cgti.2019.145

Interview
individuals and groups to help them develop novel therapeutics, and make
the case for them being relevant and promising for PD.
However, over the years, we’ve developed a sophisticated approach that
now includes multiple ways in which we can support drug developers as
they move forward, beyond simply
funding them. For example, we
“...the biggest ‘do’ when thinking have teams here that are expert in
about collaborating with an understanding the challenges of re-
organization such as ours is to treat us cruiting patients into PD trials, and

the different ways we can engage
as a partner.” patient groups to educate them on
the value of their participation.
The Foundation is at the nexus of R&D activity in PD, so we have
connections to a lot of different external expertise that can also be lever-
aged. When a company approaches us with a therapeutic idea, they may be
struggling to find a good PD expert who can participate on their scientific
advisory board, or maybe they’re looking for input directly from patients
– they may want to hear from a certain patient group about the particular
type of treatment they’re developing. We have ways to engage such indi-
viduals and groups – to connect those dots and help companies access the
expertise they need.
So we have really expanded the menu of opportunities to help support
therapeutic development for PD.
QQ Can you summarize the key learnings from your

years of experience collaborating with the gene
therapy industry and academia in coordinating clinical
development projects?
BF: It’s been interesting over the years to figure out the best
way to work with companies, in particular. It’s one thing to work with
academic groups that are more used to the idea of a foundation or a funder
helping support R&D, but when you’re dealing with companies, it’s a dif-
ferent ballgame. There are different incentives involved, and there are many
more concerns and considerations relating to proprietary information, for
example.
In general, we’ve found that the biggest ‘do’ when thinking about col-
laborating with an organization such as ours is to treat us as a partner. A
company can come to us and be open and honest about what they’re trying
to do in PD – let us know why they think it’s important, what challenges
they’re facing. Again, the assistance we can provide is not limited to the
purely financial. There are many different ways we can assist a company

that’s developing a treatment for PD, but we can only help if we know
what those challenges are – it’s vitally important to have open and honest
communication to make a collaboration work well.
We’ve found in a number of cas-
es that a particular problem raised
“...over the last 20 years or so, there’s by one company is in fact shared
been a growing appreciation that by multiple other companies in
Parkinson’s disease isn’t just about loss the same therapeutic space. There
are often opportunities in these in-
of one certain subset of brain cells...” stances to look at pre-competitive
initiatives as a means of solving the
problems, and the Foundation can act in many ways as a neutral convener.
We can help both to identify potential solutions and then ultimately, if
everyone agrees it’s valuable, actually take the lead role in supporting a par-
ticular project or study to address that challenge. It’s obviously useful for
every company involved to be a part of such collaborations because they all
benefit from the outcome.
We do find that companies that are able to appreciate the value of that
kind of pre-competitive communication and collaboration are the ones
that often benefit the most from the work we can do, and the type of part-
nership we can build with them. We have lots of examples where this type
of multiple stakeholder collaboration has worked, both in the context of
broad programs – for example, general biomarker development in PD– as
well as some very specific therapeutic challenges that have arisen, where the
Foundation was able to step in and clarify through funded studies the issues
that were causing concern to the benefit of all involved.
QQ What are some of the key specific challenges involved

in designing and conducting clinical trials for gene
therapies against PD?
BF: Probably the biggest challenge is delivery, especially if the

presumed mechanism of action through which the therapeutic
gene is supposed to work is in the brain. So when you’re thinking
about designing trials for a gene therapy product in PD, at least with tech-
nology today, that right there is your first big hurdle: you’re most likely
going to be doing brain surgery in individuals with the disease. You need to
ask what are some of the implications that come with that fact: how are you
going to deliver the product? Is it a validated delivery approach and device?
What are the other considerations for someone with PD who has move-
ment and other potential disabling symptoms – how could that impact the
surgical procedures you might put in place? PD is also generally a disease
1394 DOI: 10.18609/cgti.2019.145

Interview
of older age, which is another thing to factor in when considering recovery

from something as significant as brain surgery.
I think the other challenge is that over the last 20 years or so, there’s been
a growing appreciation that PD isn’t just about loss of one certain subset of
brain cells – in this case, the idea that it’s all about loss of certain dopamine
producing cells in the brain – but that it’s really more of a ‘whole body’
disease. There are a lot of parts of the brain, even of the peripheral nervous
system, that might be impacted by the disease. The idea of targeting a gene
therapy product to just one region of the brain may help restore some
function or protect cells in that specific region, but there’s the whole rest of
the body to consider, and the rest of the disease process that might still be
happening around it. I think that with gene therapy, you have to appreciate
this is a reality with PD: when you develop these types of treatment you
have to understand that wherever you target it, you might only be address-
ing one component and not the entire disease process.
This is all based upon gene therapy in its current form and existing tar-
geted delivery methodologies, of course. If over time we are able to advance
gene therapy to a point where you can truly deliver it in the same way you
might deliver a small molecule – systemically to the whole body – and if we
know we can target the right cells and produce the appropriate therapeutic
response, then that equation might change. You might then see gene thera-
py becoming more of a ‘full disease’ type of approach for PD.
A third component would be deciding what gene you want to deliver.
Looking at the current gene therapy development pipeline, two of the cur-
rent leading groups mentioned above are essentially delivering the enzymes
for making more dopamine in the brain, which is certainly a valid approach
and an interesting way of targeting the disease given loss of dopamine un-
derlies much of the movement challenges seen in PD. You also have groups
trying to deliver genes that target specific mechanisms believed to under-
lie disease cause – approaches which could potentially be restorative. But
again, if these approaches only target a specific region of the brain, it may
not necessarily affect the entire body. So it’s important to be clear about
what you think your gene therapy product is doing and where.
QQ Are there any particular emerging cell and gene

therapy approaches on your radar which may hold
promise?
BF: We’re certainly aware of and excited by the development

of various approaches in the wider cell and gene therapy space
and the potential opportunities they might bring to PD – we’re
keeping a close watch on them. We’ve seen genetically engineered cell

therapies have such a strong impact in oncology, for example – it will be

interesting to see if certain transplanted cell populations can be genetically
modified to provide factors that could be protective in PD.
Obviously, gene therapy in itself is not a treatment – it’s really a platform
technology for delivering potentially therapeutic genes. We tend to look at
it that way: we don’t see ‘gene ther-
apy’ as the cure for PD, because it
“It’s core to our mission that we really depends on what it is you’re
constantly seek to push and accelerate delivering. But it’s always exciting
the pipeline for new treatments for for us to see different novel plat-
forms emerging. For me, the real
Parkinson’s disease...” promise for something like cell and
gene therapy is that it is a compar-
atively very targeted, exquisite way of addressing a specific mechanism. It’s
not a traditional small molecule that might hit a bunch of different biolo-
gies – it’s not always easy to chemically dial those off-target effects out. The
enormous value that cell and gene can bring to developing treatments for a
disease like PD certainly isn’t lost on us – it’s why we continue to monitor
the field to see where the next opportunity may arise.
QQ Moving further forward, what are The Michael J. Fox

Foundation’s chief R&D priorities and goals for the
future?
BF: I’ll come back to the fact that the current R&D pipeline is
probably the most exciting, healthy and robust that it’s been in
years. Lots of different approaches being tested, some addressing disease
mechanisms, some aimed at providing better ways to handle the most se-
rious disabling symptoms, and others seeking to address different stages of
the disease. So we see this nice mix in the clinical pipeline for PD that is
giving a lot of us hope and excitement about the opportunities in the years
ahead. (In addition, over the last couple of years we have seen a number
of new products actually getting approved for PD, which is always very
important to see).
It’s core to our mission that we constantly seek to push and accelerate the
pipeline for new treatments for PD – we’ll continue to develop both our
strategic funding and our non-funding mechanisms for how best to enable
that progress. One challenge we’re seeing in clinical trial patient recruit-
ment is that increasingly, treatments in development for PD are looking
to target genetic forms of the disease. It obviously further complicates the
patient recruitment picture when you have to think about how to identify,
screen, recruit and enroll people with certain mutations linked to PD. So
1396 DOI: 10.18609/cgti.2019.145

Interview
we and others in the community continue to develop strategies around

how do we find these people and educate them about the valuable role they
could play in developing drugs for PD so they can actually participate in
some of these trials.
There’s also been a lot of conversation lately around the regulatory paths
for developing drugs for diseases like PD, so we have a whole separate
effort ongoing that focuses on how we engage with regulators – and with
payers, too. How do we develop lines of communication with these critical
stakeholders and bring that insight to bear in identifying a clearer path for
developing drugs for PD moving forward?
In short, there’s a whole algorithm here in terms of how we accelerate the
pipeline, which is a combination of several different types of strategies we
can put in place. That is what we will continue to push in the coming years.
AFFILIATION
Brian Fiske
The Michael J. Fox Foundation for Parkinson’s Research

Disclosure and potential conflicts of interest: The author declares that they have no conflicts of interest.
Funding declaration: Dr Fiske is an employee of the The Michael J. Fox Foundation for Parkinson’s Research, which has finan-
cial partnership and sponsorship relationships with entities as listed at: https://www.michaeljfox.org/current-partnerships.

Attribution: Copyright © 2019 Brian Fiske. Published by Cell and Gene Therapy Insights under Creative Commons License
Interview conducted: Sep 20 2019; Publication date: Nov 7 2019.


ADVANCED THERAPIES
INTERVIEW
Breaking new ground: bringing an

iPS cell therapy to the clinic
KAPIL BHARTI holds a bachelor’s degree in biophysics from the Panjab
University in Chandigarh, India, where he graduated with highest honors. This
was followed by a Masters degree in biotechnology at the Maharaja Sayaji
Rao University in Baroda, India and a diploma in molecular cell biology at the
Johann Wolfgang Goethe University at Frankfurt in Germany. Supported by
an international PhD student fellowship, he obtained his PhD from the same
institution, graduating summa cum laude. His PhD work involved basic bi-
ology in the areas of heat stress, cellular chaperones, and epigenetics. From
Germany, Dr Bharti came to the National Institute of Neurological Disorders
and Stroke to work with Dr Heinz Arnheiter as a postdoctoral fellow. While
there, he published numerous papers in the areas of transcription factor reg-
ulation, pigment cell biology, and the developmental biology of the eye. It is
perhaps this combination of diverse backgrounds that led him to develop an
interest in the emerging field of stem cell biology, particularly of the retinal
pigment epithelium, as he moved into the role of staff scientist. Dr Bharti
has authored numerous publications and has won several awards, including,
most recently, being a finalist in the prestigious trans-NIH Earl Stadtman
Symposium.
DOI: 10.18609/cgti.2019.142
QQ What are you working on right now?
KB: The main focus of my lab at the National Eye Institute at

NIH is to better understand the mechanisms of various forms of
retinal degenerative diseases, and to try to develop new therapies

for them.
All the work in my lab is based around the use of induced pluripo-
tent stem cells (iPSC). One of the main focuses for us is to develop an
autologous iPS cell-based therapy for a disease called age-related macular
degeneration (AMD), which causes patients to go blind later in life. AMD
is thought to be caused by degeneration of an eye tissue. We have actually
developed this tissue in the lab from patient-derived iPS cells and we’re now
trying to develop a Phase 1 clinical trial for those patients.
QQ How do you go about approaching clinical translation

in a novel and ground-breaking technology field such
as iPSCs? What are the key considerations/lessons
that might be generally applicable for others in the
cell and gene therapy area?
KB: For me, one of the key things in the lab is to not give up
the science – to really focus on the biology and science of the
cells. When we say we’re trying to do translational research, that’s not to
say we’re not doing basic science research, too. For me, the two go hand-
in-hand. It’s really fundamental to understand everything basic about your
cells, your gene therapies, whatever technology you are working on, before
you introduce them into clinical application. The more you understand,
the more likely it is they will be safe and efficacious in patients.
To give you one example, regarding the eye tissue we’re making from
iPS cells, we spent years using developmental biology of various organisms
that had been studied, including mice and humans, and using that devel-
opmental biology to make sure that the cells we are making in a dish are as
close as possible to native eye cells. Simultaneously, we spent a lot of time
characterizing them to really gain an understanding of how they function. I
think all those fundamental discoveries are really key to ensuring that your
translation is going to be fruitful and will go forward properly.
I think the same thing is applicable to any technology: not to give up
the basic biology.
QQ More specifically, what are/were the main challenges in

bringing an iPSC-derived cell therapy to first in human
trials and how have you sought to address them?
KB: There’s not a lot of work that’s been done to date on bring-
ing iPS cells to patients – there’s just one example from Japan,
1370 DOI: 10.18609/cgti.2019.142

Interview
essentially. There’s not much background information on how to do it

and that means a lot of work in both ensuring the safety of the cell thera-
py product, and also ensuring that
“One of the main focuses for us is to you’re making it correctly. As we’re
develop an autologous iPS cell-based moving towards the trial, we’re

having to invent all of those things
therapy for a disease called age-related and to make sure that we’re doing
macular degeneration...” it right – it’s a very steep learning
curve in that sense.
The field knows that a key concern with iPS cells is that any cells left in
the final product can be tumorigenic – you need to ensure that the final
product cells don’t carry any oncogenic mutations. We spent quite some
time on this aspect to ensure our final product does not carry any iPS
cells. In fact, we demonstrated that even if you forced iPS cells into the
final product, they won’t grow under the conditions we used to grow our
test article – the one we want to transplant into patients. Even so, we went
to great lengths to demonstrate in preclinical animal models that the cells
don’t form tumors or teratomas or migrate into any other tissue.
As part of this work to ensure the safety of the cells, we developed a pro-
cess of making iPS cells from patients’ progenitor blood cells – CD34-pos-
itive cells – that helped ensure that the cells maintained their proliferation
early on (so that they could make the iPS cells) but would not accumulate
mutation during the culture process. And in fact, we then checked for the
oncogene and found out that in most cases, we don’t see any potential on-
cogenic mutations in these cells.
In terms of delivery, we then had to develop a tool that really fits the
back of the eye, that is safe for delivery, and at that is biocompatible whilst
also capable of helping maneuver the transplant.
With all of this work done, we are right now in the process of working
with the FDA towards our clinical trial approval.
QQ Tell us about the trial design you have chosen for the
first in human study in AMD patients – what have
you selected in this regard and why?
KB: Since this is a Phase 1 trial, it is a safety study by design –

we have discussed our trial design with the data monitoring board
and the Institutional Review Board (IRB) and the main goal is to
ensure that the transplant will stay safe.
What that means for us is that at least in the first cohort, we’re trans-
planting into patients who already have significant vision loss, meaning
that if the transplant were to prove to be unsafe, it wouldn’t cause further

vision loss in those patients. Essentially, what we’re looking for in the first
cohort is safety of implant and whether it integrates in the back of the eye.
As we continue and we keep demonstrating safety of the transplant in
the first cohort, we might be able to test it in a second cohort of patients
who have slightly better vision. But again, the primary outcome is going
to be safety throughout the trial.
QQ There has been some discussion around the optimal

method of delivery used for cell therapies against
AMD – can you summarize your observations and
opinions in this regard?
KB: There are multiple ways to deliver retinal pigment epithe-

lium cells into the back of the eye. One of the first approaches was
injecting a suspension of cells, and then people have also tried monolayer
patch on a plastic scaffold. We took a slightly different approach: we’re try-
ing to transplant monolayer patch on a biodegradable scaffold.
I’ve spent a lot of time studying how RPE cells function and it is clear
that these cells need to be fully polarized before they can perform any func-
tions. What I mean by ‘fully polarized’ is you need polarized tissue that
contains several thousand cells – a single RPE cell won’t be able to perform
all the functions that normal, native RPE tissue performs.
Because of this, we knew that to get optimal efficacy we would need to
be able to deliver the patch as a polarized tissue in the back of the eye. Cell
suspension may work under some conditions, but in many cases the inte-
gration of those cells is extremely challenging. And it’s a large assumption
that one would have to make to say that all the cells injected would form
a perfectly polarized monolayer in the back of the eye. There’s no real data
that supports that.
In our experiments we did see some integration of suspension injections,
but not at all at a comparable level to the integration of the monolayer
patch. That’s why we think that if the integration is so dramatically dif-
ferent between the two approach-
es, the efficacy will be dramatically
“...what we’re looking for in the first different, too. But again, there’s not
cohort is safety of implant and whether enough patient data at this point to
say for certain that one approach is
it integrates in the back of the eye..” definitely better than the other –
this is all based on preclinical work.
Plastic versus biodegradable: in our hands, both worked well. For me,
though, biodegradable intuitively makes more sense. As we allow the cells
to form a monolayer, they secrete their own exocellular matrix and then the
1372 DOI: 10.18609/cgti.2019.142

Interview
scaffold, which is no longer needed, simply degrades away. The scaffold is

only designed for proper delivery in the right place – after that it is no lon-
ger required, so why keep it there?
“We hope that we can transplant 12 But we’ll find out in due course if
our approach really enables the lon-
patients in the Phase 1 trial over the ger-term survival of the patch as op-
course of the next 2 or 3 years.” posed to the plastic scaffold.
QQ Changing tack for a moment, how do you think the

‘auto vs allo’ debate will play out in the iPS cell field?
KB: When iPS cells were first discovered in 2007, everyone

was excited that for the first time we had the possibility of devel-
oping autologous cell therapies, because at that time, everyone
accepted that allogeneic cell therapies would not work in the long-
term due to immune rejection.
But nobody did the actual experiment of comparing an autologous iPS
cell therapy to an allogeneic one, and to this day, still nobody has done that
experiment! So I went with the approach of let’s at least prove that autol-
ogous iPS cell therapy does work in a Phase 1 trial, and hopefully beyond
that. There will then be a way to compare autologous with allogeneic.
The appeal of allogeneic cell therapy is of course that the manufacturing
process is that much simpler. You would only have to make one cell bank
whereas currently we have to make cells for every single patient, which is a
lot more time-consuming and challenging in terms of financial resources.
Clearly, if allogeneic does work (by which I mean no immune rejection of
alloantigens or allotransplants) then that’s the way to go. But as of right
now, we don’t know for sure if allo will work – that’s why at this stage, I
think we should be comparing the two.
Moving forward, if allo turns out to work at least as well as auto in our
application area, and if there’s a way to switch, I think it would be perfect
for us to do so. Perhaps this will be made possible by using universal donor
cells, which is where you take an iPS or embryonic stem (ES) cell line and
knock down or knock out the actual antigens so that they are not seen by
the host’s immune system. However, while that approach is very intriguing,
it has its own challenges: if the immune system doesn’t see the allo cells at
all and those cells then make a tumor, the immune system won’t see that
either. So we’ll see how far that goes. But again, if it works, that is definitely
the most appealing idea.
An intermediate approach is making iPS cell banks from individuals
homozygous for certain MHC-haplotypes. The thinking there is that these
cell banks could be applicable to a relatively large percentage of the patient

population, especially in societies that are not particularly ethnically di-

verse. It wouldn’t work so well in the USA, but in Japan, for instance, one
estimate is you would only need around 100 banks to cover more than half
of the total population.
But at this stage, we really don’t know which approach will be successful
in the end. And I think that in the long-term, if autologous transplants do
integrate and become efficacious, healthcare or health insurance companies
will figure out a way to pay for them. And the cost could potentially be
reduced by automation, of course – we all know the challenges in manu-
facturing brought about by the amount of manual labor required, which
increases cost and therefore price. We and many other groups are working
hard on trying to figure that one out.
QQ Finally, can you share your chief priorities and goals

for the next 12–24 months?
KB: The main priority is to work towards getting the IND ap-
proval for the Phase 1 trial and then getting that started. We’re very
close – we hope if things go well, we might get it in the next several months.
Once that approval is achieved, we’ll then work towards transplanting a
few patients next year. We hope that we can transplant 12 patients in the
Phase 1 trial over the course of the next 2 or 3 years. In the meantime, we
need to work out if the Phase 1 is successful, where we want to go from
there for the Phase 2. Do we want to work on automation, or a hybrid ap-
proach for an allo cell therapy? We are discussing a lot of these possibilities
at this moment but at the end of the day, the early Phase 1 data will dictate
the direction we subsequently take.
AFFILIATION
Kapil Bharti
National Institute of Neurological Disorders and Stroke, NIH, MD, USA
1374 DOI: 10.18609/cgti.2019.142

Interview

Disclosure and potential conflicts of interest: The author declares that they have no conflicts of interest.
Funding declaration: The author received no financial support for the research, authorship and/or publication of this article.

Attribution: Copyright © 2019 Kapil Bharti. Published by Cell and Gene Therapy Insights under Creative Commons License
Interview conducted: Aug 27 2019; Publication date: Oct 29 2019.

Viral Vector Channel
ADHERENT CULTURE
METHODS EDITION
IN FOCUS:
Highlights from our Vector Channel
Adherent Culture Methods
EXPERT INSIGHT INTERVIEW

Evaluation of AAV vector Assessing the future prospects
production from the iCELLis of upstream bioprocessing
fixed bed bioreactor vessel systems for commercial AAV
Shelley Nass, Bindu Nambiar, production
Maryellen Mattingly, Scott A Jeffers
Denise Woodcock & Catherine
O’Riordan
1461–1471 1275–1279
VECTOR CHANNEL: ADHERENT

CULTURE METHODS
EXPERT INSIGHT
Evaluation of AAV vector

production from the iCELLis
fixed bed bioreactor vessel
Shelley Nass, Bindu Nambiar,
Maryellen Mattingly, Denise Woodcock &
Catherine O’Riordan
AAV gene therapy vectors have demonstrated efficacy in numerous clin-

ical trials, and gene therapy products are now a reality. In support of the
commercialization of AAV gene therapy biologics, scalable, high capacity
AAV production methods are necessary, and here we describe the use of
the iCELLis® Fixed Bed Bioreactor Vessel (Pall Corporation), a versatile
AAV production system that can be used for the production of both re-
search grade and GMP AAV vectors. The iCELLis® system is ideally suit-
ed for use with the triple transfection AAV production method utilizing
adherent HEK 293 cells. For routine AAV research vector production in
the iCELLis® Nano high compaction 4 m2 vessel is used, with the option
to combine up to four vessels in tandem, if vector yields greater than
1 × 1014 VGs are required. The use of the iCELLis® system provides a con-
tinuum in the vector production platform for pre-clinical AAV vector pro-
duction to GMP AAV production for clinical trials and commercialization.
DOI: 10.18609/cgti.2019.154
MATERIAL & METHODS as previously described [1]. Briefly, a 1:1:1 ratio of the three plasmids
AAV vectors were produced via tran- HEK293 cells were transfected us- (inverted terminal repeat [ITR]
sient transfection on planar vessels ing polyethyleneimine, (PEI), and vector, AAV rep/cap, and Ad helper
plasmid). The pAd helper used was (Beckman Coulter, Indianapolis,

pHelper (Stratagene/Agilent Tech- IN, USA). A 400 µl volume of
nologies). Cell pellets were har- sample was loaded into the sam-
vested following centrifugation ple sector of a two-sector velocity
(1,500 rpm for 15 min) and resus- cell, and 410 µl of PBS was loaded
pended in lysis buffer (20 mM Tris into the corresponding reference
[pH 7.5], 150 mM NaCl, 10 mM sector. The sample was placed in
MgCl2) prior to freeze/thawing the four-hole rotor and allowed to
[1]. Following the addition of Ben- equilibrate in the instrument un-
zonase® and 0.1%Triton™ X-100, til a temperature of 20oC and full
the lysate was incubated at 37oC vacuum were maintained for 1 h.
for 90 min and then centrifuged Sedimentation velocity centrifuga-
at 3,500 rpm before sequential fil- tion was performed at 20,000 rpm
trations using 0.8 µm and 0.45 µm and 20oC. Absorbance (260 nm)
filters. Purification of AAV was optics was used to record the ra-
achieved using a column purifica- dial concentration as a function of
tion method as described previous- time until the lightest sedimenting
ly [1]. component had cleared the optical
window (1.2 h). AUC data were
analyzed as previously described
[2].
SAMPLE PREPARATION
FOR AUC ANALYSIS
The purified vector, at a concentra-
tion of 2 × 1012 to 5 × 1012 VGs/
QUANTITATIVE PCR
ml, was buffer exchanged into PBS
ANALYSES
(pH 7.2) using a 10K MWCO The AAV vector was quantified us-
Slide-a-Lyzer™ (Thermo Scientific). ing a real-time qPCR assay (7500
The AAV vector absorbance signal Real-Time PCR System; Applied
was determined by optical density Biosystems, Foster City, CA, USA)
measurement at 260 nm (OD260) with primers specific for the polya-
using spectrophotometric meth- denylation signal. Vector levels are
ods. For consistency, the samples expressed as vector genomes per
were adjusted to a target concen- milliliter, VGs/ml.
tration (OD260 of between 0.2
and 0.8) either by direct dilution
with PBS or further concentrated
using an Amicon® Ultra-0.5/30K ANALYZING RAAV
MWCO Centrifugal Filter Device VECTOR PURITY USING
(Millipore). SYPRO® RUBY PROTEIN
GEL STAIN
Samples from purified vector were
loaded onto a NuPage™ 4-12 %
SEDIMENTATION Bis-Tris gel (Invitrogen). Typically,
VELOCITY AUC DATA 5 × 1010 VGs of purified vector was
ACQUISITION analyzed. The gel was stained with
Sedimentation velocity AUC (SV- SYPRO® Ruby Protein Gel Stain
AUC) analysis was performed (Life Technologies) and observed
using a Proteome Lab™ XL-I under a UV light source.
1462 DOI: 10.18609/cgti.2019.154

expert insight
IN VITRO TRANSDUCTION The iCELLis® Nano vessels f FIGURE 1

ASSESSMENT OF AAV were prepared according to the The vessels are controlled by a
manufacturer’s recommendations stand-alone mPath bioreactor
HEK293 cells were seeded at control tower and Pall Link; a su-
with the appropriate Pall branded pervisory control and data acquisi-
2 × 105 cells/well and infected 24h
consumable parts, DO (Dissolved tion (SCADA) software package.
later, in triplicate, with AAV at a
Oxygen Probe used to measure ox-
MOI of 1 × 106 VGs/cell in a 500 µl
ygen content in the vessel during
volume. The media were replaced
the duration of run), and pH
24 h post-infection with 1 ml of
probes. Following autoclaving,
complete DMEM containing 10%
the vessels were filled with 700ml
fetal bovine serum (FBS), penicil-
DMEM supplemented with 5%
lin/streptomycin (pen/strep), and
FBS and allowed to condition
L-glutamine. After 72 h, cells were
over night at 37°C, with stirring
lysed and assayed for vector genome
at a linear speed of 2 cm/s. Prior
copy number by qPCR assay (BGH
to inoculation with cells, the DO
target) and eGFP protein levels us-
and pH probes were recalibrated,
ing an eGFP ELISA® kit from Ab-
activated and allowed to stabilize
cam (ab 171581).
for 60 minutes. The following set A center column within the vessel
points were used throughout the provides a fixed bed of polyester
microfibers for cell attachment and
duration of the run, DO: 45%, expansion. The bottom left image shows
a microscopic view of cell attachment to
OPERATION OF THE pH: 7.2 and temperature: 37°C.
the microfibers. (Pall Corporation).
ICELLIS® FIXED BED Air flow was also activated at
BIOREACTOR this time, at a rate of 30 ml/min.
Pre-cultured HEK293 cells (Agi-
The iCELLis® Fixed Bed Bioreactor
lent) were introduced to the Nano
system is a disposable bioreactor
vessels in a concentrated volume of
vessel that is available with multi-
100 ml for a total of 4 × 108 cells
ple surface growth areas (Figure 1).
per 4 m2 iCELLis® Nano vessel
The smaller scale unit, the iCELLis®
(10,000 cells/cm2). To maximize
Nano, ranges in cell growth surface
cell attachment and initial cell
area from 0.53–4.0 m2, while a larg-
growth in the concentrated media
er manufacturing unit, the iCELLis®
environment, the speed of the stir-
500+ provides up to 500 m2 of cell
rer was adjusted to maintain the
growth surface area. Potential advan-
2 cm/s linear speed, while the cir-
tages to the iCELLis® Nano bioreac-
culation of medium was initiated
tor include pH and temperature con-
6–8 h post cell inoculation. Typi-
trol along with the replenishment of
cally, 4 l of DMEM supplemented
O2 during the AAV vector produc-
with 5% FBS was pumped into the
tion process, promoting optimal cell
iCELLis® Nano vessel at a rate of
viability. The iCELLis® bioreactor
24 ml/min, and subsequently cy-
consists of a fixed bed surrounded
cled out of the vessel at a rate of
by culture medium. The medium
28 ml/min. The increased outlet
is pumped from the bottom of the
circulation rate prevents the head-
bioreactor through the bed and then
space from inadvertently becom-
falls as a thin-film down the outer
ing too small which could reduce
wall of the fixed-bed. The O2 is de-
overall gas exchange. The working
pleted from the media but is replen-
vessel volume remained at approx-
ished within the headspace above, as
imately 668 ml throughout the
O2 is fed into the bioreactor.
duration of the perfusion event. A

7 cm piece of tubing attached to the OPTIMIZATION OF THE

‘Media Out Port’ on the underside MEDIA RECIRCULATION
of the Nano vessel lid, gave a 6 cm VOLUME TO SUPPORT
falling film height. The falling film CELL GROWTH
height can be adjusted by altering PRE-TRANSFECTION
the length of tubing allowing for During the 4 m2 iCELLis® Nano
increased gas exchange if necessary. runs, prior to cell transfection,
The vessel will not inadvertently approximately 8 l of DMEM
empty overtime because media is supplemented with 5% FBS was
only removed when the media level recirculated throughout the ves-
rises to the level of the tubing. Cell sel during a 3-day growth period.
growth proceeded for 72 h with The glucose levels (measured of-
daily sampling of media from the fline by Vi-Cell) remained >2 g/l
vessels using a ViCell MetaFlex™ throughout the duration of cell
system (Beckman Coulter). This growth. With the aim of reducing
provided an offline measurement media use and costs, the recircu-
of the levels of pH, DO, and media lation volume was reduced to 4 L
nutrients and metabolites. Cells for the 3-day growth period prior
were transfected using the triple to transfection. With the reduced
transfection method, as described volume of media, glucose levels
[1]. Prior to transfection, a cell remained >1 g/l and lactate levels
count was performed by aseptically remained below 1.5 g/l, compara-
removing carriers from the Nano ble to levels measured with an 8 l
vessel and performing a cell nuclei media recirculation volume (Figure
count. The top GL45 cap of the 2). Importantly, there was no mea-
vessel was removed in the hood and sured adverse effect on the rate of
sterile tweezers were used to manu- cell growth with reduced media
ally remove two carrier strips from volumes, cell counts of 4 × 109 to
different areas of the fixed bed. The 8 × 109 were consistently achieved.
cell strips were placed in a snap top
tube containing 300 µl of PBS. The
cells were then lysed from the strips,
by adding 300 µl of Lysis Buffer TRANSFECTION OF
Reagent A100 as described by the HEK293 CELLS IN THE
vendor (Chemoetec) and vortexing ICELLIS® NANO
for 1–2 min. An additional 300 µl HEK293 cells were transfected us-
of Reagent B was then added and ing polyethyleneimine (PEI-HCL
vortexed to stabilize the sample. A Max 40,000MW, Poly Sciences
nuclei count was performed on a Inc.), and a 1:1:1 ratio of the three
sample loaded into a Via1-Cassette plasmids (inverted terminal repeat
on the NucleoCounter® NC-200 [ITR] vector, AAV rep/cap, and Ad
System (Chemometec); since the helper plasmid); a ratio of 3:1 PEI:
strips have a known surface area of DNA was used. A total of 2.4 mgs
13.9 cm2 a cell count per vessel was of each pDNA was added to
determined. An even cell distribu- 333 ml of serum-free DMEM, an
tion was applied in the calculation, additional 21.6 mls of PEI (1 mg/
as previous studies have confirmed ml) combined with 333 mls of
that cells distribute uniformly serum-free DMEM was also add-
throughout the fixed bed [3]. ed. The PEI plasmid complex was
1464 DOI: 10.18609/cgti.2019.154

expert insight
f FIGURE 2
Comparison of glucose and lactate levels in the iCELLis® Nano over a 3-day
cell growth period using either 4L or 8L media recirculation volumes.
incubated at room temperature to clarification and further down-

for 15 minutes before addition to stream processing.
the bioreactor. Prior to adding the
transfection complex to the cells,
media circulation was halted and
the serum containing media was OPTIMIZATION OF
drained from the iCELLis® Nano PEI: DNA COMPLEX
vessel. Additionally, during trans- FORMATION
fection of the cells, the pH control The effect of incubation time on
was paused to prevent CO2 addi- PEI: DNA complex formation and
tion to the iCELLis® Nano vessel vector yield was evaluated in the
during the transfection period. 4 m2 iCELLis® Nano vessels. After
An additional 133 mls of fresh addition of the PEI to the DNA
serum-free DMEM was added to mixture, the complex was incubat-
flush out any residual complex ed for 15 min or 30 min prior to
from the sample vessel and lines adding to the vessel (two vessels
leading to the iCELLis® Nano per condition). There was an aver-
vessel. Recirculation with 6 l se- age yield of 1.6 × 1014 VGs/Vessel
rum-free DMEM was reinitiated pre-purification for both complex
2 h post transfection, and media formation times, suggesting that
samples were taken daily to assess 15 min was enough time to allow
media nutrients and metabolites, for complex formation prior to ad-
culture pH and DO levels. The dition to the cells (Figure 3).
transfected cells were harvested
96 h post-transfection, both the
vessel media and recirculation bulk
volume, along with the cell lysate ESTABLISHING OPTIMAL
and vessel rinses were pooled prior CONTACT TIME OF THE

f FIGURE 3 compaction’ bed for increased cell

AAV production (VGs / iCELLis Nano vessel) following a PEI:DNA complex
®
densities so it is reasonable to as-
formation time of 15 min or 30 min. sume that the cells may require
longer exposure time for optimal
transfection. Cells in the iCELLis®
Nano bioreactor were exposed to
the PEI:DNA complex for either a
2 or a 3-h exposure time to assess
the effects on AAV vector yield.
Following exposure of the cells to
the PEI:DNA complex, recircula-
tion of fresh media was initiated.
A titer analysis of harvest samples
was performed to determine vector
yields for both conditions; it was
determined that 2- and 3-h incu-
bation times resulted in 1.5 × 1014
total VGs/vessel, suggesting that
the 2-h exposure time of cells to
the PEI:DNA complexes was suf-
ficient to optimally transfect cells
in the iCELLis® Nano vessel.
PEI:DNA COMPLEX TO
THE CELL SURFACE
DURING TRANSFECTION
The optimal time for PEI:DNA
OPTIMIZING MEDIA
complex contact time with the cell
RECIRCULATION
surface was first determined in six-
VOLUME TO SUPPORT
well dishes, before further evalua-
CELL CULTURE
tion in the iCELLis Nano. Six-well
POST-TRANSFECTION
dishes were seeded with HEK293 For our initial runs, media recir-
cells and transfected with the 3:1 culation was initiated post trans-
PEI:DNA complex, the complex fection with 8 l of serum-free
was allowed to incubate with the DMEM. With the goal of min-
cells for 1, 2, 3 or 4 h, before re- imizing downstream processing
placing the complex with fresh se- volumes, reduced recirculation
rum-free media. Transfected cells volumes were evaluated. There was
were harvested 72 h post transfec- a direct correlation in vector yield
tion and vector yields were deter- with media volume, and it was de-
mined by qPCR. A 5-fold increase termined that 6 l of media was the
in vector yield was measured with lowest working volume that could
the 2 h complex incubation time, be used without affecting vector
compared to incubating the cells yield. In cases where recirculation
for an hour. No significant increase volume was reduced to 5 l, vector
in vector yield was measured when yields were reduced by as much
the complex incubation time was as three-fold, compared to vector
extended to 3 or 4 h. The 4 m2 yields achieved using 6-8 l of recir-
iCELLis® Nano vessels have a ‘high culation media (Figure 4).
1466 DOI: 10.18609/cgti.2019.154

expert insight
f FIGURE 4
The effect of media recirculation volume on vector yield.
Vector yield (VGs/vessel) for different vectors including AAV9, AAVrh10 and AAV PHP.B is represented as a function of recirculating
volume.
CHARACTERIZING remainder is retained intracellular-

AAV VECTOR ly. In contrast, for AAVrh10, close
RELEASE FROM CELLS to 70% of the vector is released
POST-TRANSFECTION into the media with the remain-
AAV serotype plays a critical role ing 30% retained intracellularly
in how vector fractionates between (Figure 5). Determining where a
the intracellular fraction and me- given serotype fractionates during
dia during AAV vector production. production is critical for designing
Vandenberghe et al., showed that strategies to harvest vector. Figure 5
in the context of either serum con- shows how various AAV serotypes
taining or serum free AAV produc- fractionate between the media and
tion, serotypes including AAV1, cells, following production in the
AAV8 and AAV9 can be harvested iCELLis® Nano. For all serotypes
from the medium of production evaluated, 50% or greater of the
cultures [4]. In our studies, and AAV vector fractionated to the me-
in agreement with Vandenberghe dia, suggesting that with use of the
et al., approximately half of the iCELLis® Nano system, under con-
AAV9 serotype vector is collected ditions described here, harvesting
in the harvested media, while the both the cellular lysate and media

f FIGURE 5
AAV vector distribution following production in the iCELLis® Nano.
The proportion of AAV vector retained intracellularly (red) or released into the media
(blue), is represented as a percentage of the total amount of AAV vector produced.
n ≥ 2 for each serotype.
will be necessary to maximize vec- CELLSTACK-10 &

tor yields. THE ICELLIS® NANO
BIOREACTOR
The production of AAVeGFP in the
EVALUATION OF AAV iCELLis® Nano 4 m2 vessel was eval-
VECTOR PRODUCTION uated against our standard Corning
FROM THE CORNING CellSTACK®-10 AAV production
f FIGURE 6
AUC sedimentation distribution plots for an AAV vector produced in either the 10 stack production vessels (A) or
the Nano iCELLis bioreactor (B).
The 99S species represents AAV capsids harboring the full vector genome of ~4,000 nucleotides, and the fractional content of this
capsid species is similar in vector preparations generated from both production systems; 94% for vector generated in the CellSTACK-10
production vessel (A) and 90% for vector generated in the Nano iCELLis bioreactor (B). The 78S and 82S capsid species represent
capsids harboring fragmented vector genomes [2].
1468 DOI: 10.18609/cgti.2019.154

expert insight
method. For this comparison, 15 harboring a full vector genome, f FIGURE 7

CellSTACK®-10 vessels were setup moreover, there was no evidence of SDS-PAGE analysis of AAVeGFP
for a total growth surface of 28.5 m2 empty particles or capsids harboring vectors followed by SYPRO RUBY
staining.
and four iCELLis® Nano vessels were fragmented genomes, in either of
used for a total surface area of 16 m2. the AAVeGFP vector preparations
Both vessels were inoculated with [2]. Additionally, SDS-PAGE analy-
10,000 cells/cm2 of the same stock sis revealed similar AAV capsid pro-
of HEK293 cells and cell growth tein ratios for both AAVeGFP vector
proceeded for three days. PEI: DNA preparations Figure 7. The infectiv-
complexes were used for transfec- ity of the AAVeGFP vector prepa-
tion as described above. At the time rations was compared by infecting
of transfection, the media in the HEK293 cells (1 × 106 VGs per cell)
iCELLis® Nano vessel was replaced and measuring eGFP expression
with serum-free DMEM, in contrast and vector genome copy number,
the cell stack remained in DMEM 72 h post infection. Figure 8 reveals
supplemented with 5% FBS. Ad- that both AAVeGFP vector prepara-
ditionally, at harvest the media and tions yielded similar VGs/cell and
lysed cells were processed from the levels of eGFP protein, following in-
iCELLis® Nano vessel at 96 h post fection in HEK293 cells, suggesting
transfection, while only cell pellets that both the CellSTACK-10 and
were processed from the planar ves- iCELLis® Nano bioreactor yielded
sel at 72 h post transfection. The AAVeGFP vector preparations with
AAVeGFP vector yields at harvest comparable potency.
were compared; the cell stack yield-
ed 1.88 × 1013 VGs/m2 compared to Lane 1: Mark12 Marker (Invitrogen),
4.25 × 1013 VGs/m2 from the iCEL- Lane 2 AAVeGFP vector produced in
Lis® Nano vessel, with the caveat that CONCLUSIONS the CellSTACK-10: 5 x 1010 VGs and
Lane 3 AAVeGFP vector produced in the
only the intracellular fraction was We have demonstrated that the iCELLis® Nano: 5 x 1010 VGs AAV capsid
proteins VP1, VP2, and VP3 are present
harvested from the planar vessel. The iCELLis® Nano bioreactor is an in the correct 1:1:10 ratio.
harvested material from the iCELLis® ideal option for AAV vector pro-
Nano was clarified and further production at research scale with the
cessed using TFF; AAVeGFP vector potential for scale up to support
from both production platforms was commercial demand. We show that
purified using affinity chromatogra- the AAV production in the iCELLis®
phy followed by CsCl density gra- Nano generates vector at high yield
dient purification [1]. The purified and comparable potency to vector
AAVeGFP vector preparations were generated using a more traditional
then compared by analytical ultra- planar vessel. The key advantages
centrifugation (AUC), SDS-PAGE of the iCELLis® Nano bioreactor,
for capsid protein ratio, and poten- over CellStack®-10 production, in-
cy using an in vitro infectivity assay. cludes constant pH, DO, and tem-
Figure 6 shows the AUC profiles for perature control, and improved gas
both AAVeGFP vector preparations, handling and monitoring, ensuring
post CsCl purification, revealing optimal cell viability during the
production of similar capsid spe- AAV production process. Notably,
cies from both production systems. others have shown, in the context
The predominant capsid species, in of retroviral vector production, a
both AAVeGFP preparations sedi- direct correlation between the ox-
mented at 99S representing capsids ygen level, the cell growth rate, and

f FIGURE 8
Assessment of vector potency.
HEK293 cells were infected with 1x106 VGs/cell of AAVeGFP vector generated from either the iCELLis® Nano or Corning CellStack-10
vessels. Approximately 72 hours post infection, cell lysates were assayed for vector genome copy number (VGs) by qPCR (blue) and
for eGFP protein levels by ELISA (green).
the vector titers in the iCELLis® for a systemic liver directed gene
fixed bed [3]. Additionally, we have therapy, such as hemophilia Factor
shown that the iCELLis® produc- IX, dosing of 1,000 patients would
tion system is compatible with a be supported, assuming a dose
range of AAV serotypes, including of 5 × 1011 VGs/kg [6]. A caveat
AAV5 and DJ8; all serotypes have to these calculations is that losses
consistently produced vector yields during purification of vector from
in the range of 1-2 × 1014 VGs in the iCELLis® 500+ are not consid-
the iCELLis® Nano 4 m2, with the ered, which will vary depending
potential of producing 2.5 × 1016 on the process used. Importantly,
VGs at the 500 m2 scale. In the the optimization experiments de-
context of clinical dosing this vec- scribed here and by others [7], with
tor yield would support dosing the iCELLis® Nano, provides a ba-
approximately 100,000 patients sis for further development of this
for an ocular indication, assuming system to the larger iCELLis® 500+,
a dose of 1.5 × 1011 VGs/eye, the a scale that is more compatible with
recommended dose for treating the demands of commercial AAV
LCA2 patients [5]. Alternatively, production.
REFERENCES
1. Nass S, Mattingly M, Woodcock D Recombinant AAV Vectors. Hum. 4. Vandenberghe LH, Xiao R, Lock M,
et al. Universal Method for the Puri- Gene Ther. Meth. 2015; 26(6): Lin J, Korn M, Wilson JM. Efficient
fication of Recombinant AAV Vectors 228–42. Serotype-Dependent Release of
of Differing Serotypes. Mol. Ther. Functional Vector into the Culture
3. Wang X, Olszewska M, Qu J et al.
Methods Clin. Dev. 2017; 9: 33–46. Medium During Adeno-Associated
Large-scale Clinical-grade Retroviral
Virus Manufacturing. Hum. Gene
2. Burnham B, Nass S, Kong E et Vector Production in a Fixed-Bed
Ther. 2010; 21: 1251–7.
al. Analytical Ultracentrifugation Bioreactor. J. Immunother. 2015;
as an Approach to Characterize 38(3): 127–35.
1470 DOI: 10.18609/cgti.2019.154

expert insight
5. Russell S, Bennett J, Wellman JA et Bioreactor. Hum. Gen. Ther. Methods Maryellen Mattingly
al. Efficacy and safety of voretigene 2016; 27(3): 112–21. Gene Therapy Research, Rare and
neparvovec (AAV2-hRPE65v2) in Neurologic Diseases Therapeutic
patients with RPE65-mediated inher- Area, Sanofi, 49 New York Avenue,
ited retinal dystrophy: a randomised, Framingham, MA 01701, USA
controlled, open-label, phase 3 trial. AFFILIATIONS
Denise Woodcock
The Lancet 2017; 390: 849–60. Alexander H Chai Gene Therapy Research, Rare and
Gene Therapy Research, Rare and Neurologic Diseases Therapeutic
6. Doshi BS, Arruda VR. Gene therapy
Neurologic Diseases Therapeutic Area, Sanofi, 49 New York Avenue,
for hemophilia: what does the future
Area, Sanofi, 49 New York Avenue, Framingham, MA 01701, USA
hold? Ther. Adv. Hematol. 2018; 9(9):
Framingham, MA 01701, USA
273–93. Catherine O’Riordan
Bindu Nambiar Gene Therapy Research, Rare and
7. Powers AD, Piras BA, Clark RK et
Gene Therapy Research, Rare and Neurologic Diseases Therapeutic
al. Development and Optimization
Neurologic Diseases Therapeutic Area, Sanofi, 49 New York Avenue,
of AAV hFIX Particles by Transient
Area, Sanofi, 49 New York Avenue, Framingham, MA 01701, USA
Transfection in an iCELLis Fixed-Bed
Framingham, MA 01701, USA

Acknowledgements: We would like to acknowledge Nicholas Kohlstrom (Pall Corporation) for assistance with the iCELLis Nano
setup, scientific discussions, and guidance. We would also like to thank Pall Corporation for the use of images in Figure 1.

Attribution: Copyright © 2019 Shelley Nass, Bindu Nambiar, Maryellen Mattingly, Denise Woodcock & Catherine O’Rior-
dan. Published by Cell and Gene Therapy Insights under Creative Commons License Deed CC BY NC ND 4.0.

VECTOR CHANNEL: ADHERENT

CULTURE METHODS
INTERVIEW
Assessing the future prospects of

upstream bioprocessing systems for
commercial AAV production
SCOTT A JEFFERS is a Director of Process Development at uniQure

LLC. He has been in and out of gene therapy since 1997 when as a graduate
student at Purdue University in Dr David A Sanders’ lab where he worked
on pseudotyping lentiviral and retroviral gene therapy vectors with Ebola
virus glycoproteins. He moved out of gene therapy and became a virologist
studying SARS virus with Dr Kathryn V Holmes and then made the jump to
France were he worked at the Institute Pasteur with Dr Felix Ray elucidating
the x-ray crystal structures of the glycoproteins of Rift Valley fever virus and
other deadly viruses. He finally broke into industry and back into gene ther-
apy when he worked at Brammer Bio in Florida.

DOI: 10.18609/cgti.2019.134
QQ Tell us what you are working on right now.
SAJ: uniQure is a gene therapy company looking for function-

al cures for liver and CNS diseases. We are currently working on a
late-phase hemophilia B project and we also have projects in earlier stages
for hemophilia A, Fabry disease, Huntington’s disease and spinocerebellar
ataxia type 3 (SCA-3). We’re a leader in late-phase process development.
In my role, I’m personally responsible for technology transfer and pro-

cess scaling-up, determining process robustness, and validating the process
for commercial manufacturing of our hemophilia B product.
QQ Can you outline any particular challenges you

encounter in upstream bioprocess development for
uniQure’s gene therapies?
SAJ: I started working in gene therapy in 1997 as a graduate

student at Purdue, and uniQure has been around for 20 years, but
gene therapy is still in its very early days. There are still only four ap-
proved gene therapy products out there on the market – two ex vivo thera-
pies in Kymriah and Yescarta, and two in vivo in Luxturna and Zolgensma.
All of those products required developments to be made in recombinant
virus production, but the issue is that since the early days of gene therapy,
there’s been no one expression system to use. I like to compare it to the
VHS versus Betamax scenario in the early days of home video recorders
when I was a kid: one of these systems is going to be the one that wins out
over the other.
The difficulty is that it is still early days and that scenario has yet to play
out. Today, there’s the adherent HEK system, and there’s the suspension
HEK system. They’re both useful for testing many products at small scale
and that testing can be done very quickly. The adherent HEK-293 system,
for example, can be used to produce virus very quickly; you can get it into
animals very quickly, and you can do rapid, prototype proof of concept
experiments.
But the problem is that HEK systems are not going to scale. Scaling is
difficult because, with the adherent system, for example, it’s a scale-out
instead of a scale-up. And that’s where I think the biggest difficulty of all
has been. For example, in my previ-
“There are still only four approved ous work, I’ve used adherent HEK-
293 in 10 Layer Cell Factories from
gene therapy products out there on Corning, and it would take many
the market – two ex vivo ... and two in hundreds – up to a thousand – of
these cell factories to be able to
vivo...” dose a single patient in a systemic
application.
uniQure has pioneered the use of baculovirus-induced insect cell expres-
sion system for the production of our AAV vectors. We think that it is ro-
bust and scalable, and we can perform commercial scale manufacturing in
our state-of-the-art facility in Lexington, Massachusetts. We currently have
a 500-liter system and we’re expanding to 2000 liters in the near future.
1276 DOI: 10.18609/cgti.2019.134

Interview
We are opening up bigger process development and assay development labs

and a pilot plant right now. We use the baculovirus system because it works
for our products.
QQ Can you dive a bit deeper into the technology and

process tools you use – have there been any particular
innovations that have stood out for you over recent
times, and what future improvements would you like
to see?
SAJ: In terms of adherent systems, the iCELLis from Pall is

something that’s going to boost production and yield consider-
ably. It’s a scale-up process of a scale-out process, in essence, which means
I can produce more virus in a much smaller space: one iCELLis bioreactor
is equal to about 786 CF10 flasks, each of which is a cube of approxi-
mately 33 centimeters on a side. If I have to have 786 of those flasks, plus
room for them to be manipulated and moved around, that requires a lot
of space. The iCELLis, on the other hand, has a footprint of around 1.5 m
x 1.5m x 2.5m. So it’s a great system, but it is still going to be a scale-out
kind of system. Moving forward, I would love to see some of the other
fixed bed bioreactors being built out and scaled up to increase our scale of
production.
One place where we’re going to need to see improvement moving
forward is in transfection efficiency. Transfecting all these cells is in-
credibly expensive at the moment, just in DNA costs alone. If I just
had a requirement of one milligram of DNA that I was going to use
for each CF10, and I had triple transfection (so I had three plasmids in
order to do that), I would have to have more than 2100 milligrams of
DNA to transfect the iCellis 500 system. And DNA is really expensive,
especially at the GMP level. That is certainly one of the major expenses
with mammalian adherent systems, but I would have the same issue in
a suspension system. For example, if I need to use 1 ug of DNA per mL
of suspension culture and I have three plasmids, I may need between 1.5
to 3 ug of plasmid per mL of suspension culture. This means I may have
to use up to 6000 milligrams of DNA with HEK-293 in a 2000-litre
reaction. I don’t think anyone is planning on doing this large of a scale,
and this seems to be unobtainable based on current cost of goods. To me,
this is a major advantage of the baculovirus system: I don’t have to rely
on DNA, and therefore I don’t need to rely on outsourced manufacture
of a very expensive, critical raw material in order to produce a large-scale
batch of vector. Instead, I can bring that in-house, and I can control that
critical raw material.

Transfection efficiency is a key area for future improvement. Lipofection

is very expensive and the relatively cheaper options, such as Polyethyleni-
mine (PEI), are not as efficient.
If we could get to using microcarriers, like they do with CHO cells, that
would be a super advance in the field of adherent mammalian expression
for gene therapy.
QQ Can you paint us a picture of what you expect the

future of commercial AAV-driven gene therapy
manufacture will look like in upstream bioprocessing
terms? And how will the balance between adherent
and suspensions systems develop moving forward?
SAJ: There are multiple systems, of course: for suspension,

you’ve got baculovirus, herpes, transfection of HEK-293 cells,
you’ve got producer cell lines, etc. I’m biased, but I do really think
that the baculovirus expression system is going to be the one that goes the
furthest. In my honest opinion, suspension will win out over adherent for
anything that is systemic, but again, it will require more technology and
more drive towards that eventuality.
For smaller-scale production – for delivery to the eye, for instance – I
can imagine that you could still use adherent cell culture for the very long
term. But I do think the real future is in suspension systems.
QQ Finally, what do you and uniQure have coming up

through the remainder of 2019 and through 2020?
What will be your key goals and milestones over this
period?
SAJ: We’re continuing our Phase 3 trial for hemophilia B (The

HOPE-B trial) – in fact, we just reached our target enrolment,
which is great.
We’ll dose our first patient in our
Huntington’s disease program this
“...I do really think that the year. And we’ll continue to work
baculovirus expression system is on our pipeline and growing our
knowledge on how best to produce
going to be the one that goes the these gene therapy vectors. That
furthest.” means that, though I now prefer
baculovirus, going forward we are
1278 DOI: 10.18609/cgti.2019.134

Interview
going to use the best system that’s available – the one that’s going to get us
to our targets.
I’m excited to work with uniQure and to see the growth in the gene
therapy industry in general. I think we are at a time where in medicine
where we’re not just going to be just treating patients; we’re going to be
providing functional cures for them for some pretty horrible diseases.
Luckily, I don’t have a child with a rare disease, but I do think often
about the children that have rare diseases and how we’re changing their
lives. That truly is something that brings me into work every day – having
the ability and the opportunity to make a difference.
AFFILIATION
Scott A Jeffers
uniQure LLC

Disclosure and potential conflicts of interest: Dr Jeffers is an employee of uniQure and discusses the system that uniQure
uses to manufacture its gene therapy.
Funding declaration: The author received no financial support for the research, authorship and/or publication of this article.

Copyright: Published by Cell and Gene Therapy Insights under Creative Commons License Deed CC BY NC ND
4.0 which allows anyone to copy, distribute, and transmit the article provided it is properly attributed in the manner
specified below. No commercial use without permission.
Attribution: Copyright © 2019 Scott A Jeffers. Published by Cell and Gene Therapy Insights under Creative Com-
mons License Deed CC BY NC ND 4.0.
Interview conducted: Sep 12 2019; Publication date: Nov 8 2019.

Clinical Trial Designs For Advanced Therapies

Uploaded by

Copyright:

Available Formats

Clinical Trial Designs For Advanced Therapies

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Clinical Trial Designs For Advanced Therapies

Uploaded by

Copyright:

Available Formats

What are some of the clinical trial designs discussed in the document?

What are some of the clinical trial designs discussed in the document?

What are some of the production systems discussed for AAV vectors?

What are some of the production systems discussed for AAV vectors?

Volume 5, Issue 11

CELL & GENE

EXPERT INSIGHT EXPERT INSIGHT EXPERT INSIGHT

REGULATORY PERSPECTIVE INTERVIEW INTERVIEW

Adherent Culture Methods

EXPERT INSIGHT INTERVIEW

CLINICAL TRIAL DESIGNS FOR

Towards individualized, low toxic

Allogeneic hematopoietic cell transplantation (Allo-HCT) has become

Cell & Gene Therapy Insights 2019; 5(11), 1495–1503

INTRODUCTION there is a long-standing history of Most of the non-malignant dis-

lysosomal enzyme deficiencies and not be the same, due to variabili-

1496 DOI: 10.18609/cgti.2019.155

[IPS]). The optimal cumulative tar- is made that the pharmacokinetics

Cell & Gene Therapy Insights - ISSN: 2059-7800 1497

Main advantage of the population of incorporating the individualized

1498 DOI: 10.18609/cgti.2019.155

significant late effects, including grafts, engraftment may have been

Cell & Gene Therapy Insights - ISSN: 2059-7800 1499

trial (NCT02963064) with promis- immuno-ablation are required, the

1500 DOI: 10.18609/cgti.2019.155

target myeloablative agent used in late effects. In the coming years,

AUTHORSHIP & CONFLICT OF INTEREST

ARTICLE & COPYRIGHT INFORMATION

Article source: Invited; externally peer reviewed.

Cell & Gene Therapy Insights - ISSN: 2059-7800 1501

1502 DOI: 10.18609/cgti.2019.155

Cell & Gene Therapy Insights - ISSN: 2059-7800 1503

CLINICAL TRIAL DESIGNS FOR

Post-marketing safety and efficacy

The implementation of new regulatory tools, such as the PRIority

Cell & Gene Therapy Insights 2019; 5(11), 1505–1521

FROM CLINICAL TRIALS evidence [5,6]. Developers of ad-

1506 DOI: 10.18609/cgti.2019.156

to appropriately use these products of study/ status”, “categorization

Cell & Gene Therapy Insights - ISSN: 2059-7800 1507

ff BOX 1 intended to be used for the same

1508 DOI: 10.18609/cgti.2019.156

Name of ATMP Pivotal Sample Follow-up Status at

Name of ATMP Pivotal Sample Follow-up Status at

follow up of the patients treated ffFIGURE 1

ATMPs were categorized according to the regulatory class as follows:

Cell & Gene Therapy Insights - ISSN: 2059-7800 1511

ff TABLE 2. parameters. Half of applicants

1512 DOI: 10.18609/cgti.2019.156

it is genetically modified or not). ffFIGURE 5

Cell & Gene Therapy Insights - ISSN: 2059-7800 1513

PRAGMATIC CONCEPTS, pregnant women) and patients with

1514 DOI: 10.18609/cgti.2019.156

AN EXEMPLARY further exploration of the effects of

Cell & Gene Therapy Insights - ISSN: 2059-7800 1515

important pre-requisite to allow engine’ in the long-run. This has the

1516 DOI: 10.18609/cgti.2019.156

primary analysis mechanisms under hypotheses that a more suitable to

Cell & Gene Therapy Insights - ISSN: 2059-7800 1517