www.thelancet.com Vol 368 October 14, 2006 1387
Establishing risk of human experimentation with drugs:
lessons from TGN1412
M J H Kenter, A F Cohen
Administration of a chemical or biological compound to
a human being is never without risk. Although knowledge
about risks increases during the development process,
risks are still present even when a substance is marketed.1
Particular care is necessary when a new drug is given to
healthy volunteers without previous human testing.
General principles for such research have been laid down
in guidelines as early as 1983, and these were the basis
for many current regulations.2 Most drugs at that time
were small molecules with fairly well characterised,
classic, pharmacological mechanisms. Proposed primary
objectives for studies in healthy people were therefore to
show pharmacological action in man and the dose (or
concentration) response curve. This approach was judged
safe and was lent support by fi ndings of available
surveys.3,4 Over time, the main objectives for these trials
changed—perhaps owing to the perceived safety of new
traditional (small-molecule) medicines—to general
tolerance and safety.
The advent of increasingly potent and selective
compounds for human-receptor systems led to situations
in which predictability from animal data was diminishing.
The fi rst substances in this category were small molecules
with fairly foreseeable pharmacokinetics, and any
unexpected adverse events were mostly fully reversible.
Biotechnology provided compounds with unique
specifi city for human targets, potentially further reducing
the predictability of animal work. However, the deaths of
two volunteers in clinical studies5–7 led to the realisation
that they could have been prevented by proper
examination of existing data.
The serious adverse events that arose during the very
fi rst administration of
CD28 superagonistic antibody, have led to immediate
reactions from diff erent regulators,8,9 ranging from a
moratorium on CD28 research to rules about how many
individuals should receive a new compound at the same
time.10 A common theme was that special care should be
given to ill-defi ned high-risk drugs. In this Viewpoint,
we propose a set of factors facilitating rational risk
analysis of all new substances to be administered to
human beings (fi gure 1). We use TGN1412 as an
illustration because it represents a new compound with
a complex and novel mechanism.
TGN1412, the so-called
Risk analysis of TGN1412
What is known about the mechanism of action of
TGN1412? This molecule is a humanised version of the
mouse antibody 5.11A1, which is an agonist of the CD28
antigen that activates T cells without specifi c engagement
of the T-cell receptor with the antigen-presenting cell.
Although the overall biology of this immunological
interaction is fairly well under stood,11,12 the precise
mechanism by which mitogenic anti-CD28 activate
T cells is unknown.13 Because the TGN1412 compound is
novel, little published data exist for its specifi c mech-
anism and, therefore, risk for unexpected occurrences is
The TGN1412 study was the fi rst trial of this type of
compound that was undertaken in man, so only a small
amount of human data were available for risk analysis.
Nevertheless, much can be learned from fi ndings of
similar clinical trials of antibodies, such as interference
with the cytotoxic T lymphocyte-associated antigen
(CTLA)-4 receptor. This somewhat distinct but related
biological mechanism is at present being tested in cancer
patients using the MDX-010 antibody.14 This molecule is
not a mitogenic antibody but it inhibits CTLA-4-mediated
signals that turn off T-cell responses. It causes severe
side-eff ects,15,16 probably owing to activation of autoreactive
T-cell clones. Additionally, several clinical studies of
antibodies against the T-cell CD3 antigen have been
done, in which massive systemic release of several
cytokines and an array of toxic eff ects have been
recorded.17 The results of this trial led to modifi cation of
the Fc-receptor binding domain of anti-CD3.18,19 In the
TGN1412 clinical trial protocol, a cytokine burst was
judged theoretical without any scientifi c consideration.20
These two analogous mechanisms suggest that T cells
can be triggered either by an agonist at an activating site
(CD3) or by an antagonist on an inhibitory site (CTLA 4)
and that such activation could produce serious toxic
eff ects. Since the CTLA-4 receptor-mediated mechanism
is closely related to that of CD28, these facts augment the
risk profi le of TGN1412 even further.
Our analysis focuses on the idea that the eff ects of an
untested mechanism of action in man can be adequately
predicted from work done in animal models or human
cell systems. A prerequisite for this theory is that an
analogous mechanism is in operation in the relevant
animal species in-vitro systems, and human beings. The
qualitative and quantitative response must be similar,
which requires comparable receptor structure, expression,
binding, and second-messenger eff ects.
The rhesus monkey (Macaca mulatta) tolerated large
doses of TGN1412 without any serious side-eff ects, and
the cynomolgus monkey (Macaca fascicularis) was used
for fi nal toxicology studies. According to the investigator’s
brochure,21 100% homology exists between the CD28
TGN1412 binding site in human beings and monkeys,
restricted to the so-called C”D loop. However, no sequence
comparison was included in the disclosed information.
Lancet 2006; 368: 1387–91
Central Committee on Research
Involving Human Subjects, The
(M J H Kenter PhD,
A F Cohen MD); and Centre for
Human Drug Research, Leiden,
Netherlands (A F Cohen)
Dr A F Cohen,
CHDR, Zernikedreef 10,
2333 CL Leiden, Netherlands
www.thelancet.com Vol 368 October 14, 2006
Figure 2 compares the human and rhesus monkey
CD28 aminoacid sequences, and clear diff erences can
be seen. The potential importance of the sequence
variation can be deduced from the crystal structure of
the human CD28 molecule,13 which indicates 14 contact
residues with the parental antibody 5.11A1. A non-
conservative change is noted at position 65 (Gly [G] to
Glu [E]). Epitope mapping in a previous CD28 study
recorded an identical aminoacid variation in the species
(rat, mouse, and man) specifi city of agonistic antibodies.22
The human CD28 sequence is glycosylated at position
53 but not at this site in the rhesus monkey counterpart.
A search of rhesus monkey CD28 aminoacid sequences
in the National Center for Biotechnology Information
(NCBI) database retrieved neither cynomolgus nor
additional rhesus CD28 sequences (accessed April,
The non-conservative variation at position 65 could
lead to diff erences in binding characteristics of TGN1412
to the human and monkey CD28 molecule, which
might result in varying amounts of T-cell activation in
man and rhesus monkeys.13 Unfortunately, the
investigator’s brochure only provides information on
the affi nity of TGN1412 to the human CD28 molecule
(1·88×10–9 mol/L), not for its monkey counterpart. In
our risk analysis, these factors increase the risk category
of the antibody and lead to further questions that can
only be answered by
Furthermore, rapid and fairly longlasting human T-cell
depletion was noted in a mouse model with a human
immune system after in-vivo administration of low
doses of the parental antibody 5.11A1,24 but these data
were not included in the investigator’s brochure. Also,
this document did not provide results of a comparison
of in-vitro activation of human and monkey peripheral
blood mononuclear cells (PBMC) by TGN1412. Such
fi ndings could have provided insight into the similarity
or otherwise of the activation of human and monkey
T cells. In-vitro stimulation of human PBMC by the
parental antibody 5.11A1 has been reported, showing its
potent mitogenic capacity.22
The main proposed action of TGN1412 reported in the
investigator’s brochure21 is activation of so-called
regulatory T cells. However, specifi city for a particular
T-cell subpopulation is not expected because the human
CD28 antigen is expressed on most CD4+ T cells and half
of CD8+ T cells. Data also indicate a lack of specifi city in
activation of T-cell populations since both anti-
infl ammatory and pro-infl ammatory cytokines are
generated (table). Production of interleukin 10 and
tumour growth factor β by activated regulatory T cells
was not ascertained for the investigator’s brochure. The
claim in the clinical trial protocol20 of preclinical evidence
that TGN1412 inhibits pro-infl ammatory cytokine
production and activates regulatory T cells is not
substantiated by in-vivo data. Moreover, published work
reporting that the human CD28 antigen is also expressed
on granulocytes25,26 was not included in the investigator’s
The mechanism of the antibody TGN1412 suggested
that deleterious eff ects in man could not be ruled out
conclusively from fi ndings of animal experiments. There
are two potential areas of concern. First, TGN1412
administration could lead to T-cell activation and massive
cytokine release. Second, the antibody could result in a
strong expansion of regulatory T cells and non-specifi c
immunosuppression. Therefore, either possibility—
activation or immunosuppression—could not be ruled
out with the available data and, since these eff ects would
have serious outcomes, the risk category should have
been increased accordingly.
Neither activation nor immunosuppression was
reported in non-human primate studies, and a starting
dose of 0·1 mg/kg was selected for the clinical trial. This
amount was ascertained by a fraction of the so-called
no-adverse-eff ect dose concentration in the cynomolgus
Previous exposure of human beings to similar
• Investigate direct mechanism
• Assess related mechanisms and analogue
• Investigate primary and secondary pharmacology
Can the primary or secondary mechanism be
induced in animals or in human cell material?
• Receptor homology
• Post-receptor mechanism similar
• Measurement systems applicable
• Human ex-vivo tests available
Level of knowledge about mechanism of action
• Is there a plausible mechanism?
• Is there adequate clinical and pathophysiological
knowledge about the mechanism?
Selectivity of the mechanism to target tissue
• Receptor distribution in tissues
• General pharmacological studies
• Toxicology studies
Analysis of potential effects
• Dose or concentration relation
• Vital organ systems affected
• Half-life in relevant effect compartment
• Pharmacokinetic dynamic relations
• Active or toxic metabolites
Predictability of effect
• Biomarkers for effect in animal and man
• Precision and accuracy of measurement
• Relation of marker to clinical effect
Can effects be managed?
• Antidotes or antagonists
• Other countermeasures
Main issue of concern
Figure 1: Main of issues of concern to be included in a risk analysis of a new compound
This analysis assumes acceptable and stable pharmaceutical and chemical quality.
www.thelancet.com Vol 368 October 14, 2006 1389
monkey. However, cytokine release was already recorded
at a low dose in this species (table). Therefore, a proper
starting dose would most probably be much less than a
500th of the concentration causing eff ects in the
monkey—even assuming the sensitivity of man and
monkey to TGN1412 was equal.
Most monoclonal antibodies have long plasma
half-lives, and animal data in the investigator’s
brochure21 show that TGN1412 has a half life of about
8 days. Thus, full removal from the body would take
about a month. This factor is an additional risk because
any untoward eff ects would be equally longlasting.
The eff ect of the antibody TGN1412 could be expected
to relate to dose or plasma concentration since the
compound exerts its action by receptor binding.
Post-receptor eff ects in the immune system could,
however, be amplifi ed easily by disturbing the delicate
balance between several T-cell subpopulations, as seen
in early anti-CD3 clinical trials. This fact makes the
eff ects described above unpredictable with respect to
dose or concentration dependency.
Individuals can have quite diff erent reactions to a
drug, and results of the experiment in which TGN1412
was added to human and animal blood ex vivo could
have given information about cytokine release or T-cell
expansion, as is typically done with infl ammatory
substances such as lipopolysaccharide.27 According to
the clinical trial protocol,20 this test was done only with
PBMC from patients with B-cell lymphatic leukaemia,
and the results showed polyclonal expansion and
activation of T lymphocytes. These standard experiments
might have provided the data needed to predict eff ects
In the investigator’s brochure,21 little guidance is given
to doctors on how side-eff ects can be controlled and
treated. Potential non-specifi c longlasting immuno-
suppressive eff ects would need particular care and
instructions for the treating clinician and study
participants—eg, in case of infections. Management of
activation of autoreactive T-cell clones would require
special long-term monitoring and, if necessary, treatment
with high-dose corticosteroids. A clear strategy would
also be needed for control of cytokine-release syndromes
and rapid decline of T cells. In any case, these adverse
reactions would most probably be serious and diffi cult to
manage, again increasing the risk of administration of
the antibody TGN1412.
CD28 Homo sapiens
CD28 Macaca mulatta
Figure 2: Comparison of deduced human and rhesus monkey CD28 aminoacid sequences
Human (Homo sapiens) accession NM 006139.1; rhesus monkey (Macaca mulatta) accession AF344855.1. Asterisks denote variations between the human and
monkey sequences. Putative contact residues with the TGN1412 antibody are marked by orange boxes; the C’’ D loop13 is indicated by black bars.
Infl ammation type Peak cytokine concentration (ng/L)
Low dose (5 mg/kg)
High dose (50 mg/kg)
Tumour necrosis factor α
Data are mean (range). Data taken from table 9 in the investigator’s brochure.21
Table: Cytokine production in cynomolgus monkey on administration of TGN1412
www.thelancet.com Vol 368 October 14, 2006
The above risk analysis, undertaken with data available
in the research fi le and public domain before the
TGN1412 trial started, shows that essential information
was absent and the antibody was a high-risk compound
unlikely to be suitable for administration to healthy
people without additional preclinical experiments. A
prerequisite for thorough assessment of the protocol and
preclinical data for any clinical trial is that all parties
involved have access to all necessary fi ndings. The
sponsor has main responsibility for making these results
available and should include and discuss the data in the
research fi le. Relevant new information that becomes
available after submission should be added and discussed
as soon as possible. This process is of special importance
in the early and rapid development of a new medicine.
From the information that was disclosed, we conclude
that the assessors did not receive all relevant fi ndings.
Even when all data are available, the diff erent people who
assess risk of a human study should communicate their
fi ndings in a consistent and orderly manner to boost the
chance that the right questions are asked. Our proposed
scheme will ensure that all parties cover the indicated
points in a transparent and critical manner, followed by a
synthesis. This approach can be used by investigators,
regulators, research ethics committees, and for internal
review in the clinical research unit.
In the UK, scientifi c assessment is done by the
competent authority at the Medicines and Health Care
Products Regulatory Agency (MHRA). The report of the
TGN1412 trial by the MHRA includes three distinct
subreports: a medical, a pharmaceutical, and a
pharmaco-toxicology (safety) assessment that were
fi nalised on separate dates.28 The published MHRA
document suggests that the subreports are the result of
isolated assessments from diff erent individuals without
much interdisciplinary interaction. Moreover, the safety
report contains several passages that seem to be copied
from text that was supplied by the sponsor company in
the investigator’s brochure. This work does not suggest
independent critical assessment.
The primary investigator takes scientifi c and medical
responsibility for the participants, which requires full
understanding of risks. This responsibility cannot
be devolved to the employer, a governmental agency, or
the research ethics committee. The TGN1412 trial
was undertaken by two
venture-capital-driven company (ie, sponsor) and a
clinical research organisation with a strong interest in
the actual implementation of the study. Both relied
heavily on the regulators to provide clearance for rapid
undertaking of the trial.
Administration of high-risk interventions should be
done in an institution at which adequate evaluation and
monitoring can be done by in-house experts. For
example, a university medical centre (in the case of
TGN1412, with a clinical and research immunology
companies: a small
department) with a well equipped and good clinical
practice-compliant research unit.
The interim report of the expert scientifi c group on
phase I clinical trials has now been published.29 The
document provides a thorough overview on the events
of the TGN1412 trial based on information that was
available to the expert scientifi c group, and it lists
22 recommendations to increase safety of volunteers in
such trials that test a compound for the fi rst time in man.
Later in 2006, a fi nal report is scheduled to be published,
which will take account of opinions and comments on
the interim report. We welcome most of the
recommendations of the expert scientifi c group but
regret that important data are still not in the public
domain for an independent and rational assessment by
the scientifi c community. As an example, the group
conclude that “Sequence analysis of the extracellular
domain revealed a 100% amino acid homology to the
human counterpart, thereby confi rming identical binding
characteristics of the TGN1412 to human and cynomolgus
monkey CD28 [p 13]”,29 but (again) no sequence
comparison and functional analysis is given to
substantiate this claim.
We believe that thorough analysis of human, rhesus
monkey, and cynomolgus monkey complete and
functional CD28 molecules may be important for our
understanding of the adverse events that severely
aff ected six healthy volunteers. Our comparison with
publicly available CD28 sequence data shows that
variation in aminoacids between man and rhesus
monkey might account (in part) for the diff erent
outcome of administration of TGN1412 in these species.
Furthermore, we suggest that all data from the TeGenero
and Paraxel research fi le should be made publicly
available for discussion by the international scientifi c
community so that lessons from the TGN1412 trial can
be learned and better risk assessment can be developed
that will protect future healthy volunteers in clinical
studies to develop useful new medicines.
On July 3 and 13, 2006, roughly 4 months after the
TGN1412 clinical trial and during the process of
bankruptcy, TeGenero submitted
cynomolgus CD28 nucleotide sequences to the NCBI
database (accession numbers ABG77997 and ABG77998)
potentially coding for a CD28 molecule with an
extracellular domain identical to the human counterpart.
Further analysis is needed to show whether this sequence
encodes the target of antibody TGN1412 and the
functional CD28 molecule in this species. Even if further
work should indicate that the CD28 molecules of man
and cynomolgus are identical in structure and function,
we believe that in-vitro tests and a comparison of human,
rhesus monkey, and cynomolgus monkey CD28
sequences should have been included in the research fi le
and, if not, should have led to questions from all involved
in the clinical trial (investigators, clinical research
organisation, and assessors).
www.thelancet.com Vol 368 October 14, 2006 1391
The complexity of current documentation about new
compounds has the inherent risk that important fi ndings
and a scarcity of data can be hidden. Our approach to risk
assessment should be seen as an initiative for an
internationally accepted format.
Confl ict of interest statement
MJHK is executive director and AFC is vice-chairman of the Central
Committee on Research Involving Human Subjects, The Hague,
Netherlands, which is also the competent authority for drug trials in
Netherlands. AFC is director of the Centre for Human Drug Research,
which is involved in phase I trials with healthy volunteers.
1 Bresalier RS, Sandler RS, Quan H, et al. Cardiovascular events
associated with rofecoxib in a colorectal adenoma chemoprevention
trial. N Engl J Med 2005; 352: 1092–102.
2 Council for International Organizations of Medical Sciences. Safety
requirements for the fi rst use of new drugs and diagnostic agents in
man: a review of safety issues in early clinical trials of drugs.
Geneva: Council for International Organizations of Medical
3 Orme M, Harry J, Routledge P, Hobson S. Healthy volunteer
studies in Great Britain: the results of a survey into 12 months
activity in this fi eld. Br J Clin Pharmacol 1989; 27: 125–33.
4 Zarafonetis CJ, Riley PA Jr, Willis PW 3rd, et al. Clinically
signifi cant adverse eff ects in a phase 1 testing program.
Clin Pharmacol Ther 1978; 24: 127–32.
5 Bouchie A. Gene-therapy death prompts suit. Nat Biotechnol 2002;
6 Carmen IH. A death in the laboratory: the politics of the Gelsinger
aftermath. Mol Ther 2001; 3: 425–28.
7 Steinbrook R. Protecting research subjects: the crisis at Johns
Hopkins. N Engl J Med 2002; 346: 716–20.
8 Schneider CK, Kalinke U, Lower J. TGN1412: a regulator’s
perspective. Nat Biotechnol 2006; 24: 493–96.
9 Bhogal N, Combes R. TGN1412: time to change the paradigm for
the testing of new pharmaceuticals. Altern Lab Anim 2006; 34:
10 Wadman M. London’s disastrous drug trial has serious side eff ects
for research. Nature 2006; 440: 388–89.
11 Jiang H, Chess L. Regulation of immune responses by T cells.
N Engl J Med 2006; 354: 1166–76.
12 Barr TA, Carlring J, Heath AW. Co-stimulatory agonists as
immunological adjuvants. Vaccine 2006; 24: 3399–407.
13 Evans EJ, Esnouf RM, Manso-Sancho R, et al. Crystal structure of a
soluble CD28-Fab complex. Nat Immunol 2005; 6: 271–79.
14 Phan GQ, Yang JC, Sherry RM, et al. Cancer regression and
autoimmunity induced by cytotoxic T lymphocyte-associated
antigen 4 blockade in patients with metastatic melanoma.
Proc Natl Acad Sci USA 2003; 100: 8372–77.
15 Beck KE, Blansfi eld JA, Tran KQ, et al. Enterocolitis in patients with
cancer after antibody blockade of cytotoxic T-lymphocyte-associated
antigen 4. J Clin Oncol 2006; 24: 2283–89.
16 Blansfi eld JA, Beck KE, Tran K, et al. Cytotoxic
T-lymphocyte-associated antigen-4 blockage can induce
autoimmune hypophysitis in patients with metastatic melanoma
and renal cancer. J Immunother 2005; 28: 593–98.
17 Smith JA, Bluestone JA. T cell inactivation and cytokine deviation
promoted by anti-CD3 mAbs. Curr Opin Immunol 1997; 9: 648–54.
18 Carpenter PA, Pavlovic S, Tso JY, et al. Non-Fc receptor-binding
humanized anti-CD3 antibodies induce apoptosis of activated
human T cells. J Immunol 2000; 165: 6205–13.
19 Carpenter PA, Tso JY, Press OW, Yu X, Anasetti C.
Non-FcR-binding, humanized anti-CD3 antibody Hu291 induces
apoptosis of human T cells more eff ectively than OKT3 and is
immunosuppressive in vivo. Transplant Proc 2000; 32: 1545–46.
20 Parexel. Parexel clinical trial protocol: a phase-I single-centre,
double-blind, randomised, placebo controlled, single
escalating-dose study to assess the safety, pharmacokinetics,
pharmacodynamics and immunogenicity of TGN1412 administered
intravenously to healthy volunteers. London: Parexel International,
2006. Available at: http://www.circare.org/foia5/tgn1412protocol.pdf
(accessed Aug 23, 2006).
21 TeGenero Immunotherapeutics. Investigator’s brochure:
TGN1412—humanized agonistic anti-CD28 monoclonal antibody,
edn 1.1 2005-12-19. Würzburg: TeGenero AG, 2005. Available at:
(accessed Aug 23, 2006).
22 Lühder F, Huang Y, Dennehy KM, et al. Topological requirements
and signaling properties of T cell-activating, anti-CD28 antibody
superagonists. J Exp Med 2003; 197: 955–66.
23 Altschul SF, Madden TL, Schaff er AA, et al. Gapped BLAST and
PSI-BLAST: a new generation of protein database search programs.
Nucleic Acids Res 1997; 25: 3389–402.
24 Legrand N, Cupedo T, van Lent AU, et al. Transient accumulation of
human mature thymocytes and regulatory T cells with CD28
superagonist in “human immune system” Rag2–/–γc–/–mice.
Blood 2006; 108: 238–45.
25 Venuprasad K, Chattopadhyay S, Saha B. CD28 signaling in
neutrophil induces T-cell chemotactic factor(s) modulating T-cell
response. Hum Immunol 2003; 64: 38–43.
26 Venuprasad K, Parab P, Prasad DV, et al. Immunobiology of CD28
expression on human neutrophils: I—CD28 regulates neutrophil
migration by modulating CXCR-1 expression. Eur J Immunol 2001;
27 Lewis CE, McCarthy SP, Lorenzen J, McGee JO. Diff erential eff ects
of LPS, IFN-gamma and TNF alpha on the secretion of lysozyme by
individual human mononuclear phagocytes: relationship to cell
maturity. Immunology 1990; 69: 402–08.
28 Medicines and Health Care Products Regulatory Agency.
Pharmaco-toxicological (safety) assessment of TGN1412. London:
Medicines and Health Care Products Regulatory Agency, 2005.
29 Department of Health. Expert scientifi c group on phase one clinical
trials: a consultation. http://www.dh.gov.uk/Consultations/
ID=4137501&chk=x%2BoJ/%2B (accessed Aug 23, 2006).