Regulatory Toxicology and Pharmacology 42 (2005) 265–274

www.elsevier.com/locate/yrtph

The no-observed-adverse-eVect-level in drug safety

evaluations: Use, issues, and deWnition(s)
Michael A. Dorato a,¤, JeVery A. Engelhardt b
a
Toxicology Division, Lilly Research Laboratories, A Division of Eli Lilly and Company, GreenWeld, IN 46140, USA
b
Preclinical Safety Assessement, Amgen, Thousand Oaks, CA 91320, USA

Received 28 February 2005

Available online 23 June 2005

Abstract

The no-observed-adverse-eVect-level (NOAEL) is an important part of the non-clinical risk assessment. It is a professional opin-
ion based on the design of the study, indication of the drug, expected pharmacology, and spectrum of oV-target eVects. There is no
consistent standard deWnition of NOAEL. This is based, in part, on the varied deWnitions of what constitutes an adverse eVect. Toxi-
cologists, either investigating or reviewing, have not been consistent in deWning an eVect as either adverse or acceptable. The common
deWnition of NOAEL, “the highest experimental point that is without adverse eVect,” serves us well in general discussions. It does
not, however, address the interpretation of risk based on toxicologically relevant eVects, nor does it consider the progression of eVect
with respect to duration and/or dose. This paper will discuss the issues and application of a functional deWnition of the NOAEL in
toxicology evaluations.
 2005 Elsevier Inc. All rights reserved.

Keywords: Adverse; Toxicologically signiWcant; Biologically important; Toxicologically relevant; Margin of safety; Hormesis; Benchmark dose;
No-observed-adverse-eVect-level

1. Introduction opinions on where harmful and non-harmful eVects

The identiWcation of eVects in animals that may be
predictive of adverse events in humans is the cornerstone
of non-clinical safety testing of pharmaceuticals for
human therapeutic use. IdentiWcation of toxic eVects in
animals is generally expected to follow a dose–response
pattern relative to incidence and severity, allowing the
determination of dose levels where substantial, impor-
tant or relevant eVects occur and dose levels where these
eVects do not occur. The responsibility of the toxicolo-
gist is to deWne, without compromise, where on the dose–
response curve eVects occur that could be considered
potentially harmful to humans. This is a simple state-
ment for a complicated, and iterative, process. The initial
*
Corresponding author. Fax: +1 317 277 4436.
E-mail address: MADorato@Lilly.com (M.A. Dorato).
occur may change as additional information related to
the molecule is generated. Data from longer-term studies
may give the toxicologist new, or additional, perspective
on whether or not an observation thought to be subtle or
minor in a short-term study is an early indicator of a
harmful eVect. For example, a liver enzyme elevation in a
short-term non-clinical study, without a histopathology
correlate, will be noted but may not be considered to be
harmful. The outcome of longer-term studies, indicating
a progression of eVect and the occurrence of histological
alteration in the liver, provides a new interpretation of
the importance and/or relevance of the response and the
duration of exposure associated with the Wrst indications
of an unwanted response.
It has also been shown that adverse eVects in non-
clinical studies of potential new drugs may vary in
dose response from species to species, and even from
study to study within the same species. This may

0273-2300/$ - see front matter  2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.yrtph.2005.05.004
266 M.A. Dorato, J.A. Engelhardt / Regulatory Toxicology and Pharmacology 42 (2005) 265–274
represent normal biological variation, or study design
issues. Safety evaluation of new chemical entities,
whether they are industrial chemicals or potential new
drugs, is not simple. It requires objectivity, careful atten-
tion to detail, knowledge of experimental models, histor-
ical perspective, knowledge of government regulatory
requirements, and experience in data interpretation.
Safety evaluation of potential new drugs diVers in
approach to that used with industrial chemicals. The
quantitative details of the exposure duration and inten-
sity, or intention, with industrial chemicals may not be
well understood. In addition, there is no therapeutic
indication to aid in a risk/beneWt safety assessment. For
potential new drugs, after the approval of the Wrst clini-
cal protocol, human exposure is controlled, and carefully
measured. The non-clinical and clinical studies generally
include a full set of evaluations targeted at detecting
unwanted and/or harmful eVects. The potential new
drugs are designed to be active in biological systems and
have a spectrum of desired and undesired eVects. The
risk/beneWt analysis, therefore, must take into account
the disease for which the drug is intended, the natural
history of the disease without therapy, and the availabil-
ity of other treatments. With potential new drugs, not
every eVect seen in a non-clinical study can be consid-
ered as unwanted or harmful. For example, drugs
designed to treat severe obesity would be expected to
show signiWcant weight loss in non-clinical studies using
non-obese animals.
From the literature and regulatory guidelines, it may
be concluded that there is no consistent deWnition of the
terms used by many toxicologists to indicate presence or
absence of undesired responses, e.g., no-observed-eVect-
level (NOEL), no-observed-adverse-eVect-level
(NOAEL), adverse or non-adverse, biologically signiW-
cant, and toxicologically signiWcant/relevant. Therefore,
the nature of the toxicology proWle for pharmaceutical
agents will be evaluated as will the nature and deWnition
of the NOAEL. Finally, a working deWnition of the
NOAEL, found useful in non-clinical safety assessments
of potential new drugs, will be provided.
2. Toxicology proWle for pharmaceutical development
At a very high level, toxicology is a process used to
identify, assess, and manage risk. The toxicologist must
determine if there is a real diVerence between control
and treated groups; determine if the diVerence is an eVect
of the experimental treatment; and determine if the eVect
of treatment is adverse. This is a simple statement for a
very complicated process. The following are the basic
questions asked of a toxicology proWle:
1. What dose/exposure does/does not produce toxic
eVects in animals?
2. Were the animal models relevant to assessment of
human toxicity?
3. What were the signs/durations of toxic responses?
4. Did the eVects diVer after single or multiple dos-
ing?
5. Did the eVects progressive with duration of expo-
sure?
6. What were the target organs/systems?
7. Were the toxic eVects reversible?
8. How was the substance metabolized/cleared by the
body?
9. Were toxic metabolites produced?
10. Was accommodation to the toxic eVects observed?
11. Was the toxicity consistent across experimental
models?
Lewis et al. (2002) have presented a number of factors
useful in determination of treatment and non-treatment-
related eVects, and to clarify the use of the term adverse.
A typical toxicology proWle (Fig. 1) includes the con-
duct of studies of suYcient duration and dose to charac-
terize toxic eVects, dose dependence, exposure, target
organs, and biomarkers in support of the safe use of an
investigational drug. The identiWcation of target organs
and relevant biomarkers is extremely important. Even
though ‘biomarkers’ is the current ‘buzz-word’ for drug
safety evaluations, toxicology studies, from the very
beginning, have always addressed and identiWed bio-
markers of potential clinical eVects. The importance of
biomarkers is embodied in the quote attributed to Para-
celsus, “All things are poison, and nothing is without
poison: the dose alone makes a thing not a poisonƒ
Thus is it with medicine: it becomes what you make of it.
If it is possible to make evil out of good, it is also possi-
ble to make good out of evil” (Temkin, 1941). As with
any area of scientiWc endeavor, the technology has
improved, and new methods are available to determine
potential biomarkers that were not previously monitor-
able in non-clinical studies.
Fig. 1. Typical toxicology proWle.
 M.A. Dorato, J.A. Engelhardt / Regulatory Toxicology and Pharmacology 42 (2005) 265–274
The phasing of the various studies in a toxicology
proWle (Fig. 2) may vary. The relationship of the toxicol-
ogy studies to the phases of drug development is, to
some extent, a case-by-case decision. Individual compa-
nies may initiate the studies at diVerent times based on
patient populations, inclusion of women of child bearing
potential, interactions with regulatory agencies, etc.
The International Conference on Harmonization of
Technical Requirements for Registration of Pharmaceu-
ticals for Human Use (ICH) has provided harmoniza-
tion of new drug registration requirements for Europe,
Japan, and the United States. The reader is referred to
the ICH website for information on non-clinical safety
requirements for potential new drugs (www.ich.org). In
particular, the ICH Topic M3(M) (2000) and the ICH
Topic S6 (1997) provide guidance on study design and
duration for small molecules and biotechnology prod-
ucts, respectively. In general, a two-week repeated dose
toxicity study is the minimum study duration required in
most of the world to support an initial clinical dose. The
U.S. Food and Drug Administration (FDA, 1996) has
proposed a single dose toxicology strategy to support
single dose clinical studies. The Committee for Medici-
nal Products for Human Use (CHMP, 2004) has also
issued a position paper on the toxicology study duration
required to support a single low dose of a compound.
The major advantage of these regional diVerences to the
ICH guidelines is a reduction in the quantity of drug
necessary to conduct a non-clinical toxicology study
prior to a single dose clinical trial.
Fig. 2. High-level relationship of toxicology proWle to phases of drug development.
267
The international regulatory requirements for non-
clinical safety studies with new drugs, and existing
regional diVerences, are provided by Dorato and Buck-
ley (1998) and Dorato and Vodicnik (2001). In general,
the international guidelines for safety testing of biotech-
nology products (ICH Topic S6, 1997) involve much
more discussion with reviewing regulatory authorities
than is required for toxicology support of small molecule
development.
The principles of non-clinical safety assessment are
similar between small and large molecules. The diVer-
ence lies in how these principles are put into practice.
For the biotechnology products, species speciWcity is an
area of major concern in selection of appropriate animal
models, as is the immune response or presence of rele-
vant biological receptors. New biotechnology products
may often be developed in one, rather than two, non-
clinical animal model. Whether small or biotechnology
molecules are being evaluated, the toxicology proWle is
progressive in duration; doses usually decrease with
increasing duration of exposure; the risk/beneWt decision
is an iterative process. As with small molecules, the ICH
Topic S6 (1997) states “Dosage levels should be selected
to provide information on a dose–response relationship,
including a toxic dose and a no-observed-adverse-eVect-
level (NOAEL).”
The predictivity of animal models for human toxicity
has been investigated by the International Life Sciences
Institute (ILSI) in a 1999 workshop. Olson et al. (2000)
have published the workshop Wndings. The workshop
268 M.A. Dorato, J.A. Engelhardt / Regulatory Toxicology and Pharmacology 42 (2005) 265–274
examined how well toxicity seen in non-clinical animal
studies predicted actual human toxicities and the dosing
duration required in animals to reveal potential toxici-
ties in human clinical trials. Toxicology studies con-
ducted in two animal species were approximately 71%
predictive of toxicities seen in man. Toxicology proWles
conducted in non-rodent species, alone, were 63% pre-
dictive of human toxicities, while those conducted in
rodents, alone, were 43% predictive of human toxicities.
Also, 94% of potential human toxicities were predicted
in studies of 61 month in duration. This survey pro-
vides a valuable perspective on the use and appropriate-
ness of non-clinical safety data.
While the purpose of non-clinical toxicology studies
is to provide an assessment of eVect and no eVect levels,
and maximum tolerated dose (MTD), the ultimate appli-
cation of the results of non-clinical toxicology studies is
in the determination of the margin of safety (MOS). The
concept of the MOS is an important part of the toxicol-
ogy proWle. Therefore, the determination of toxicity in a
toxicology study is to be expected, and desirable. The
rational use of the information is not in the application
of the absolute toxicity determination, but in relation to
the clinical dose/exposure, the clinical indication, avail-
able therapy, and overall risk/beneWt analysis of the use
of a potential new drug. The relationship of the MOS to
the eVective dose and the NOAEL is shown in Fig. 3.
The NOAEL is the dose on the toxicology dose–
response curve that is compared to the pharmacody-
namic eVective dose to establish the MOS. The eVective
dose is often the clinically eYcacious dose, but in early
drug discovery, the eVective dose may be deWned as the
dose eVective in a relevant eYcacy model.
In toxicology studies conducted with a variety of
agents it is sometimes impossible to establish a dose that
does not produce any pharmacologic eVects in the
Fig. 3. Relationship of margin of safety to toxicity proWle.
animal model. For example, compounds for the class of
drugs called selective estrogen receptor modulators
(SERMs) generally produce pharmacologic eVects at
very low levels of exposure. For members of this drug
class that are developed for indications in post-meno-
pausal women, the estrogenic or antiestrogenic eVects
seen in normal mature animals are noted, but are not
necessarily considered adverse for the post-menopausal
clinical population. This is an important consideration
in the overall interpretation of potential toxicity.
The proceedings of the First ICH conference (1991)
attempted to address the signiWcance of using NOEL
and NOAEL to aid in the description of observations in
a toxicity study. “Furthermore, one should keep in mind
that the eVect to be determined is the toxicologically rel-
evant eVect, i.e., the eVect which may endanger human
health. Therefore the term ‘no-observed-adverse-eVect-
level’ is preferred to distinguish NOEL from signs of
speciWc pharmacodynamic ‘class’ eVects” (Hess, 1991).
In developing a risk/beneWt position on a drug, it is
important to understand and discuss whether or not an
observed pharmacodynamic eVect, i.e., the hypoglycemia
produced by insulin and its analogues, should be consid-
ered as adverse, and to what patient population. An
assessment of the ability to monitor, and reverse the
eVect is also important. Evaluation of an idealized dose–
response relationship for eYcacy and toxicity provides
additional insight into the nature of the NOAEL (Fig. 4).
The distance between the curves at any chosen level, e.g.,
1, 50, or 99%, represents the chosen parameters for the
MOS. The philosophical question of whether or not
pharmacology and toxicology, in Paracelsus’ terms
(Temkin, 1941) are intrinsically ‘good’ and ‘evil,’ will not
be discussed here. However, the occurrence of what has
been termed ‘high-dose’ pharmacology complicates the
assessment of the NOAEL from non-clinical toxicology
 M.A. Dorato, J.A. Engelhardt / Regulatory Toxicology and Pharmacology 42 (2005) 265–274
Fig. 4. Idealized dose–response curves for eYcacy and toxicity, show-
ing the relationship of MOS and the occurrence of high-dose no-target
eVects that may be considered undesirable.
studies. It is expected that low doses of an experimental
compound usually produce desired eVects, whereas high
doses usually produce undesired eVects. This is repre-
sented in the eYcacy and toxicity curves in Fig. 4. The
upper section of the eYcacy curve may produce exces-
sive pharmacodynamic responses (on-target) that may
be considered undesirable. These responses should not a
priori be used to deWne the NOAEL. Rather, the nature
of these eVects as toxicologically relevant should be the
subject of discussions between the toxicologist, physician
and regulatory reviewers. These discussions should pro-
vide an appropriate perspective for risk perception,
assessment, and management. Understanding the rele-
vance of the animal model to the patient population is
also important to the discussion. For example, the study
of hypoglycemic agents in non-diabetic animal models
may produce adverse eVects because the animal model is
euglycemic. Insulin administration at pharmacodynami-
cally eVective doses in pregnant diabetic women has
been shown to decrease birth defects, the opposite may
be true in normal healthy animals (CPMP, 2002).
Administration of insulin or insulin analogues to normal
healthy animals may induce teratogenicity not relevant
to human risk. The doses in a toxicity evaluation should
be chosen carefully, and distinguish the teratogenic
eVects of the test compound from those caused by pro-
nounced hypoglycemia (CPMP, 2002). The appropriate
use of animal models of disease in toxicology evalua-
tions is a matter of debate and is beyond the scope of
this discussion. The data from a non-clinical toxicology
study should be interpreted relative to the impact on
safety for the target population.
Histopathology has historically been the most consis-
tent criterion to establish NOEL/NOAEL. In order to
assure the best determination of eVect and no eVect lev-
els, consistency in the pathology evaluation is essential.
As such, a pathology peer review should be conducted
on all repeated dose toxicity studies. Pathology peer
review is a quality control process where an independent
269
pathologist conducts a second evaluation of a subset of
animals and tissues. This peer review evaluation ensures
that all compound-related Wndings were identiWed in the
study, that the Wndings were recorded accurately, and
that the terminology and severity grades were used con-
sistently by the primary pathologist. Additionally, the
peer review pathologist reviews the correctness of the
interpretation written by the primary pathologist. Utiliz-
ing these processes will facilitate consistent identiWcation
of adverse Wndings and determination of the NOAEL
for a pharmaceutical.
As mentioned previously, non-clinical toxicology
assessments are an iterative process. This creates some
diYculty in the interpretation of subtle pathology Wnd-
ings in early short-term studies. Without question, the
purpose of a non-clinical toxicology evaluation is to
assess the safety of a molecule, without compromise,
prior to moving into clinical trials. On the other hand,
toxicity data without the perspective necessary to deter-
mine MOS present an equally important issue. As part
of the risk/beneWt discussion between toxicologists, clini-
cians, and regulators, development of a new drug may be
approved in stages, while a complete picture of MOS is
developed. It is important to record and report all Wnd-
ings but the decision to stop development on a poten-
tially useful or critical new medicine needs equal care
and consideration.
3. DeWnitions of the no-observed-adverse-eVect-level
It is generally accepted that there is a lack of agree-
ment on the deWnition of common terms used in risk
assessment, as well as inconsistency in the application of
the terms such as NOAEL. Table 1 presents selected
deWnitions of the NOAEL from the published literature.
Lewis et al. (2002) and Filipsson et al. (2003) have
pointed out that the NOAEL is a measured/estimated
value and may be diVerent from the true no adverse
eVect level, which lies somewhere between the measured
NOAEL and the measured lowest-observed-adverse-
eVect-level (LOAEL).
Filipsson et al. (2003) have presented an alternate to
the traditional NOAEL approach. The benchmark dose
method (BMD) is based on a dose corresponding to a
speciWc change in the adverse response, compared to
controls. This dose–response modeling approach has not
been widely adopted for the evaluation of pharmaceuti-
cals. The method claims to address some of the limita-
tions of the traditional NOAEL assessment, though
cases in which the traditional approach should be used
are acknowledged. Beck et al. (1993) have provided a
comparison of the advantages and disadvantages of the
NOAEL and BMD approaches.
Calabrese et al. (1999) and Calabrese and Baldwin
(2003a) have introduced a discussion of hormesis in the
270 M.A. Dorato, J.A. Engelhardt / Regulatory Toxicology and Pharmacology 42 (2005) 265–274

Table 1
Selected deWnitions of NOAEL
Source
Hayes (2001)
ECETOC T R 85 (2002),
EPA (1995), Faustman and
Omenn (2001), Beck et al. (1993)
Leisenring and Ryan (1992)
Calabrese and Baldwin (1994)
FDA Guidance (2002)
IPCS (1999)
DeWnition
The highest dose that is without observed eVects in properly designed and executed toxicology studies
The highest exposure level at which there are no statistically or biologically signiWcant increases in the
frequency or severity of adverse eVect between exposed and control groups. Some eVects may be produced
but they are not considered adverse or precursors to adverse eVects
Experimental dose level immediately below the lowest dose that produces a statistically or biologically
signiWcant increase in rate of adverse eVects over control
Highest dose not statistically diVerent from control yet signiWcantly diVerent from lowest-observed-
adverse-eVect-level (LOAEL)
The highest dose level that does not produce a signiWcant increase in adverse eVectsƒ adverse eVects that
are statistically signiWcant and adverse eVects that may be clinically signiWcant (even if they are not
statistically signiWcant) should be considered. The deWnition of the NOAEL in contrast to that of the
NOEL reXects the view that some eVects observed in the animal may be acceptable pharmacodynamic
actions of the therapeutic and may not raise a safety concern
Simple estimate of the highest dose in which the incidence of toxic eVect was not signiWcantly diVerent from
untreated group (statistically and biologically)

evaluation of the NOAEL. In a hormetic response, a
stimulation of biologic eVects occurs below the toxicity
threshold while an inhibition occurs above the threshold.
Low-dose groups could show an improvement over con-
trols. This may be related to an adaptive or exaggerated
homeostatic response, and its biological limits. This
response also may be in the range of the expected phar-
macodynamic eVects that would occur in the dose range
below the NOAEL, such as that seen with hormones and
hormone-like drugs. Recognition that non-adverse
eVects may occur at doses below the NOAEL does not
mean they are not due to the test article (e.g., “normal
variation”), but are truly part of the dose–response con-
tinuum of the test material. Due to regulatory require-
ments/expectation most toxicology studies of potential
new drugs are carried out well above the low doses
where hormesis is likely to occur. The use of data on hor-
mesis is unlikely to replace the traditional method of
establishing a NOAEL until its place in risk assessment
is better established (Rodericks, 2003). Further, as there
is a recognized overlap in the low-dose stimulatory zone
between the threshold and hormetic response predic-
tions, the ability to predict the hormetic dose–response
curve will be inXuenced by the deWnition used to estab-
lish the NOAEL (Calabrese and Baldwin, 2003a).
Indeed, as we state in our deWnition, lack of statistical
signiWcance alone does not constitute a NOAEL. Rather,
the NOAEL is determined by the combined analysis of
the biological and statistical eVects.
In general use, the NOAEL is an important part of the
MOS, it is a professional opinion about toxicologically
relevant eVects and interpretation of expected pharma-
cology, and it is the subject of iterative interpretation as
the toxicology proWle develops. While we will not discuss
the application of safety or uncertainty factors, the
NOAEL is often combined with a safety factor to address
issues of variability in extrapolation across species (Bar-
nes and Dourson, 1988; Beck et al., 1993; Calabrese et al.,
1992; Dourson and Stara, 1983; Lehman and Fitzhugh,
1954). As previously mentioned, early in the development
of the toxicology proWle we may see changes in parame-
ters, i.e., liver enzymes that have no histopathology corre-
late and are not interpreted to be a safety issue. Later,
within the context of longer duration studies, these same
eVects may be seen as precursors to severe hepatic dam-
age. The discussion is then related as to whether or not
the early biomarker response should be judged as
adverse. As a response predictive of progressive toxicity,
it should be considered adverse. However, the indication,
monitorability, reversibility, and risk/beneWt analysis for
the patient population will have an eVect on how the
observation aVects continued development of the drug.
The nature of the NOAEL may best be understood,
perhaps, by its criticisms (Calabrese et al., 1999; Cala-
brese and Baldwin, 1994, 2003a; Filipsson et al., 2003;
IPCS, 1999; Leisenring and Ryan, 1992; Woutersen et al.,
2001). A compiled list of concerns about the traditional
method of establishing the NOAEL is presented in Table
2. Suggestions on improved study design as a way to
improve the NOAEL estimate have been made over the
years. These suggestions generally included the addition
of more dose groups, the addition of more animals to the
Table 2
Issues with the traditional method of NOAEL determination
• Must be one of the experimental doses tested
• Dependent on study design
• Observed value which does not consider the slope of the dose–
response curve
• It is an estimated/derived value; the accuracy of the estimate depends
on the spread of the doses
• Dose selection aVects accuracy of NOAEL estimate
• Values are dictated by experimental design not by inherent
biological properties of the test system
• Sensitive to sample size
• NOAEL tends to be higher with fewer animals/dose
• Varies from experiment to experiment
• DeWnition of adverse varies between reviewers
• Risk levels may be higher than estimated by the derived NOAEL
 M.A. Dorato, J.A. Engelhardt / Regulatory Toxicology and Pharmacology 42 (2005) 265–274
lower-dose groups to improve the detection of subtle
changes, etc. While these suggestions are deserving of
consideration, one must ask if the concerns about the
potential diVerence in the measured and true NOAEL
really impact clinical safety. Careful clinical dose escala-
tion and evaluation of relevant biomarkers, based on
information from non-clinical toxicology studies in rele-
vant animal models, has long been shown to be a valid
approach to the safe testing of potential new drugs.
The observation that there is no consistent deWnition
of NOAEL may be explained by the subjective nature
of what constitutes the deWnition of an adverse eVect. In
very broad terms, an adverse eVect may be considered
to be a change (biochemical, functional, or structural)
that may impair performance and generally have a det-
rimental eVect on growth, development or life span of a
non-clinical toxicology model. More speciWcally, an
adverse eVect in a non-clinical toxicology study should
be an eVect that would be unacceptable if it occurred in
a human clinical trial (FDA Guidance, 2002). Toxicolo-
gists have not been consistent in applying judgment to
determine if an observed eVect in a non-clinical toxicol-
ogy study is either adverse or acceptable. In practice,
the judgment on the adverse nature of an observation in
a non-clinical toxicology study is subject to discussion,
challenge, and reinterpretation. The decision that an
observation is adverse may be aVected as much by pub-
lic policy as by strict scientiWc interpretation. Lewis et
al. (2002) have presented a series of steps that may be
used to determine if a response in a non-clinical toxicol-
ogy study is either adverse or non-adverse. This
Fig. 5. Approach to classifying toxicology study results as adverse or non-adverse (modiWed from Lewis et al., 2002).
271
approach includes an evaluation of the adaptive, tran-
sient/persistent, and progressive nature of the response,
its association with other eVects, the occurrence of func-
tional impairment, and the primary or secondary nature
of the response. Though the approach is weighted
toward dealing with ‘industrial’ chemicals, it is also
valuable in assessing safety of pharmaceuticals. Any
eVect seen in a non-clinical toxicology study, whether it
is broadly deWned as pharmacology (on-target) or toxi-
cology (oV-target), may be considered undesirable and
therefore adverse. In interpreting these Wndings relative
to potential new drugs, the nature of the animal model
(normal versus disease), the target patient population,
and the therapeutic indication must be carefully
evaluated.
A simpliWed diagram of the process used to determine
if the results of a non-clinical toxicology study of a
potential new drug are adverse or non-adverse is pre-
sented in Fig. 5. The Wrst two questions are pretty
straightforward, though they are presented in very sim-
ple form. The large ‘gray area’ relates to the importance
of statistical and biological signiWcance. The expert opin-
ion of the toxicologist is the key in determining an
observed event as adverse or non-adverse. As previously
stated, consideration is given to the indication, the clini-
cal opinion, the regulatory opinion, and the relevance of
the animal model. In many cases, a dose-related response
that shows statistical signiWcance from the control group
will be treated as an adverse event. However, in the eval-
uation of potential new drugs, the therapeutic target
must also be considered. Pharmacologic eVects integral
272 M.A. Dorato, J.A. Engelhardt / Regulatory Toxicology and Pharmacology 42 (2005) 265–274
to the therapeutic indication are to be expected in non-
clinical toxicology studies with active pharmaceuticals.
These may be shown as changes, relative to control, in
monitored parameters, i.e., clinical pathology, behavior,
food consumption, weight gain, etc. Pharmacodynamic
eVects related to expected pharmacology must be distin-
guished from adverse, oV-target, eVects (Williams, 1990).
An example of an eVective regulatory decision strategy
involving no MOS relative to a non-clinical study
NOAEL has been presented by Black et al. (1999).
Toxicologists from diVerent regions have diVerent
interpretations of what constitutes an adverse event.
The impact of non-adverse statistical diVerences from
control, leading to diVerent assessments of toxicity, and
supporting the use of NOAEL rather than NOEL in
non-clinical toxicology studies of pharmaceuticals, was
addressed at the First ICH (Hayashi, 1991). Two very
simple examples of expected pharmacology that might
be considered adverse in normal subjects are the severe
body weight loss observed in non-obese animals during
the study of a drug for obesity, and the severe hypogly-
cemia observed in non-diabetic animal models during
the study of a drug for diabetes. Both of these eVects
are related to the on-target pharmacology. They should
certainly be noted as issues/warnings for clinical inves-
tigators, to provide some perspective on use. They
should not, though, prevent the clinical study of the
potential therapeutic agent or a priori be a major cause
of concern for use in the appropriate target popula-
tions. Statistical signiWcance, by itself, does not make
an event adverse. This concept is embedded in the
diVerent deWnitions of NOEL (any eVect, not just
adverse eVects, Table 1) and NOAEL (FDA Guidance,
2002). If an event does not Xag as statistically signiW-
cant, it may still be considered adverse based on the
biological signiWcance. For example, it has been our
experience that drug-induced immune-mediated cyto-
penia will not necessarily be Xagged as statistically sig-
niWcant in most shorter-term non-rodent studies, but
would certainly be considered adverse. Similarly, non-
statistically signiWcant increases in organ weights may
still be biologically signiWcant. Even the dose respon-
siveness is not a simple prerequisite for determining an
adverse eVect. Undesired receptor-mediated eVects may
show a Xat dose response, but would be considered
adverse.
Further, a NOAEL should be determined for each
toxicity study as a whole, rather than for individual Wnd-
ings. Additionally, the NOAEL should be established for
each study type and duration and be allowed to change
or mature as one gathers additional data for a particular
compound. This allows for a holistic monitoring of tem-
poral relationships of compound-related eVects identi-
Wed in the drug development process. This will help in
establishing the relevance of Wndings in animals to
human risk assessment (Olson et al., 2000).
4. Discussion
In the development of potential new drugs, a balance
must be created between the drive to move forward with
potential new and important medicines, and the drive to
make conservative decisions relative to patient safety.
This can be facilitated by active/open discussions
between regulatory agency toxicologists, sponsor toxi-
cologists, and clinicians. To further this discussion, a
clear understanding of what constitutes an adverse eVect
in a non-clinical toxicology study is important. The iden-
tiWcation of the NOAEL provides a basis for moving
forward with clinical investigations of new drugs. It must
be clearly understood that the NOAEL is not risk free.
Variability in clinical and non-clinical response and exis-
tence of sensitive sub-populations increases the proba-
bility of serious adverse eVects. We propose a functional
deWnition of NOAEL that is most appropriate for the
safety evaluation of potential new drugs to be:
The highest dose/exposure that does not cause toxico-
logically relevant increases in the frequency or severity
of eVects between exposed and control groups based on
careful biological and statistical analysis. While mini-
mum toxic eVects or pharmacodynamic responses may
be observed at this dose, they are not considered to
endanger human health or as precursors to serious
events with continued duration of exposure.
The determination of precursor eVects and additional
insights into the toxicity proWle from longer duration
studies are very important to decisions on continuing
development of a new drug. Statistical diVerences from
control groups deserve attention, but by themselves
make variable contributions to safety evaluations from
non-rodent and rodent toxicology studies. Statistical sig-
niWcance, alone, should not be used to judge the observa-
tion as adverse. In the deWnition of NOAEL oVered
above, a careful consideration of the biological and sta-
tistical outcomes of a non-clinical toxicology study is
required. In addition, the spectrum of histopathological
alterations, identiWcation of target organs, and potential
reversibility of lesions aids not only in identiWcation of
the NOAEL, but also in the overall risk assessment.
Using all of the available toxicity data rather than only
those Wndings that are statistically signiWcant will allow
for the development of a rational interpretation of the
comprehensive toxicity proWle of a potential new drug.
This comprehensive proWle should provide the greatest
assistance to regulatory assessors and clinicians in
understanding the potential safety concerns of a new
drug candidate.
We have taken what might be termed a traditional
approach in the evaluation of the dose–response in toxi-
cology studies, leading to a determination of NOAEL.
Alternative approaches have been discussed. The bench-
mark dose (Filipsson et al., 2003) claims to be a more
 M.A. Dorato, J.A. Engelhardt / Regulatory Toxicology and Pharmacology 42 (2005) 265–274
powerful statistical tool than traditional NOAEL
approaches, and claims to represent a more accurate risk
assessment. Calabrese (2004, 2005) and Calabrese and
Baldwin (1998, 2003b) have for several years represented
hormesis as a method to improve toxicology risk assess-
ment. Axelrod et al. (2004) have presented an argument
that the existing toxicology data does not support a uni-
versal extension of the hormesis concept to regulatory
policy. Toxicologists should acquaint themselves with
the concept of hormesis and use their professional judg-
ment and observations in deciding the validity of the
hormetic approach. The last 40 years of drug safety eval-
uation support the more traditional approach to
NOAEL as an appropriate tool for evaluating risk/ben-
eWt of a therapeutic agent. The toxicologist needs to be
open, however, to new approaches and challenges to
established practices.
The application of the pathology peer review pro-
cess, consistent use of the deWnition of adverse, and
implementation of an accepted deWnition of NOAEL
will allow pharmaceutical companies and drug regula-
tory authorities to come to a common understanding
with respect to pharmacological class eVects. Further, as
consistent deWnitions are recognized and used, the shar-
ing and acceptance of evaluation recommendations
among global drug regulatory authorities for the non-
clinical portion of the drug dossier may also be facili-
tated. To achieve this, though, it is necessary that a
functional deWnition of NOAEL for human pharma-
ceutical development, such as the one proposed here, be
accepted.
References
Axelrod, D., Burns, K., Davis, D., vonLarebeke, N., 2004. “Horme-
sis”—an inappropriate extrapolation from the speciWc to the uni-
versal. Int. J. Occup. Environ. Health 10, 335–339.
Barnes, D.G., Dourson, M., 1988. Reference dose (RfD): description
and use in health risk assessment. Regul. Toxicol. Pharmacol. 8,
471–486.
Beck, B.D., Conolly, R.B., Dourson, M.L., Guth, D., Hattis, D., Kim-
mel, C., Lewis, S.C., 1993. Symposium overview: improvements in
quantitative noncancer risk assessment. Fundam. Appl. Toxicol. 20,
1–14.
Black, L.E., Bendele, A.M., Bendele, R.A., Zack, P.M., Hamilton, M.,
1999. Regulatory decision strategy for entry of a novel biologic
therapeutic with a clinically unmonitorable toxicity into clinical tri-
als: pre-IND meetings and a case example. Toxicol. Pathol. 27 (1),
22–26.
Calabrese, E.J., Beck, B.D., Chappell, W.R., 1992. Does the animal-to-
human uncertainty factor incorporate interspecies diVerences in
surface area? Regul. Toxicol. Pharmacol. 15, 172–179.
Calabrese, E.J., Baldwin, L.A., 1994. Improved method for selection of
the NOAEL. Regul. Toxicol. Pharmacol. 19, 48–50.
Calabrese, E.J., Baldwin, L.A., 1998. Hormesis as a biological hypothe-
sis. Environ. Health Perspect. 106 (1), 357–362.
Calabrese, E.J., Baldwin, L.A., Holland, C.D., 1999. Hormesis: a highly
generalizable and reproducible phenomenon with important impli-
cations for risk assessment. Risk Anal. 19 (2), 261–281.
273
Calabrese, E.J., Baldwin, L.A., 2003a. The hormetic dose–response
model is more common than the threshold model in toxicology.
Toxicol. Sci. 71, 246–250.
Calabrese, E.J., Baldwin, L.A., 2003b. Toxicology rethinks its central
belief. Nature 42, 691–692.
Calabrese, E.J., 2004. Hormesis—basic, generalizable, central to toxi-
cology and a method to improve the risk-assessment process. Int. J.
Occup. Environ. Health 10, 466–467.
Calabrese, E.J., 2005. Paradigm lost, paradigm found: the re-emergence
of hormesis as a fundamental dose response model in the toxicolog-
ical sciences. Environ. Pollut., accepted.
CHMP, 2004. Position paper on non-clinical safety studies to sup-
port clinical trials with a single microdose. CPMP/SWP/2599/02/
Rev 1.
CPMP, 2002. Points to consider on the need for assessment of repro-
ductive toxicity of human insulin analogues. CPMP/SWP/2600/
01Final.
Dorato, M.A., Buckley, L.A., 1998. Toxicology in the drug discovery
and development process. In: Enna, S.J., Williams, M., Ferlanny,
J.W., Kemakin, T., Pursolt, R.D., Sullivan, J.P. (Eds.), Current Pro-
tocols in Pharmacology. Wiley, New York, pp. 10.3.1–10.3.30.
Dorato, M.A., Vodicnik, M.J., 2001. The toxicologic assessment of
pharmaceutical and biotechnology products. In: Hayes, A.W. (Ed.),
Principles and Methods of Toxicology, fourth ed. Taylor and Fran-
cis, New York.
Dourson, M.L., Stara, J.F., 1983. Regulatory history and experimental
support of uncertainty (safety) factors. Regul. Toxicol. Pharmacol.
3, 224–238.
ECETOC, 2002. Recognition of, and diVerentiation between, adverse
and non-adverse eVects in toxicology studies. Technical Report 85.
EPA, 1995. The use of the benchmark dose approach in health risk
assessment. EPA/630/R-94/007.
Faustman, E.M., Omenn, G.S., 2001. Risk assessment. In: Klaassen,
C.D. (Ed.), Cassarett and Doull’s Toxicology the Basic Science of
Poisons, sixth ed. McGraw-Hill, New York, p. 92.
FDA, 1996. Guidance for industry. Single dose acute toxicity testing
for pharmaceuticals. 61FR43934.
FDA, Guidance for industry and reviewers (draft), 2002. Estimating the
safe starting dose in clinical trials for therapeutics in adult healthy
volunteers. Draft Guidance. EECDS029ECderGuidE3814dft.doc.
Filipsson, A.F., Sand, S., Nilsson, J., Victorin, K., 2003. The benchmark
dose method—review of available models, and recommendations
for application in health risk assessment. Crit. Rev. Toxicol. 33 (5),
505–542.
Hayashi, Y., 1991. The current situation with regard to no eVect dose
vs no toxic eVect dose. In: D’Arcy, P.F., Harron, D.W.G. (Eds.),
First International Conference on Harmonization. Brussels, pp.
191–192.
Hayes, A.W., 2001. Principles and methods of toxicology, fourth ed.
Taylor and Francis, Philadelphia p. 1824.
Hess, R., 1991. Repeated dose toxicity: industry perspective. In:
D’Arcy, P.F., Harron, D.W.G. (Eds.), First International Confer-
ence on Harmonization. Brussels, p. 197.
ICH Topic S6, 1997. Preclinical safety evaluation of biotechnology-
derived pharmaceuticals. ICH Harmonized Tripartite Guide-
line.
ICH Topic M3(M), 2000. Maintenance of the ICH guideline on non-
clinical safety studies for the conduct of human clinical trials for
pharmaceuticals. ICH Harmonized Tripartite Guideline.
IPCS, 1999. Environmental Health Criteria 210. Principles of the
Assessment of Risks to Human Health from Exposure to Chemi-
cals. World Health Organization, International Programme on
Chemical Safety, Geneva, p. 23.
Lehman, A.J., Fitzhugh, D.G., 1954. 100-Fold margin of safety. Assoc.
Food Drug OV. USQ Bull. 18, 33–35.
Leisenring, W., Ryan, L., 1992. Statistical properties of the NOAEL.
Regul. Toxicol. Pharmacol. 15, 161–171.
274 M.A. Dorato, J.A. Engelhardt / Regulatory Toxicology and Pharmacology 42 (2005) 265–274
Lewis, R.W., Billington, R., Debryune, E., Gamer, A., Lang, B., Carpa-
nini, F., 2002. Recognition of adverse and nonadverse eVects in tox-
icity studies. Toxicol. Pathol. 30 (1), 66–74.
Olson, H., Betton, G., Robinson, D., Thomas, K., Monro, A., Kolaja,
G., Lilly, P., Sanders, J., Sipes, G., Bracken, W., Dorato, M., Van
Deun, K., Smith, P., Berger, B., Heller, A., 2000. Concordance of the
toxicity of pharmaceuticals in humans and animals. Regul. Toxicol.
Pharmacol. 32, 56–67.
Rodericks, J., 2003. Hormesis and toxicological risk assessment. Toxi-
col. Sci. 71, 134–136.
Temkin, C., Lillian, 1941. Seven defences, the reply to certain calumnia-
tions of his enemies. In: Sigerist, H.E. (Ed.), Paracelsus, four trea-
tises. The Johns Hopkins Press, Baltimore, MD, p. 22.
Williams, P.D., 1990. The role of pharmacological proWling in safety
assessment. Regul. Toxicol. Pharmacol. 12, 238–252.
Woutersen, R.A., Jonker, D., Stevenson, H., te Biesebeek, J.D., Slob,
W., 2001. The benchmark approach applied to a 28-day
toxicity study with rhodorsil silane in rats: the impact of
increasing the number of dose groups. Food Chem. Toxicol. 39,
697–707.