Committee on Mutagenicity of Chemicals in Food,

Consumer Products and the Environment (COM)

Statement 2018/S1

Statement on the quantitative approaches to the

assessment of genotoxicity data

[Link]
chemicals-in-food-consumer-products-and-the-environment

COM Secretariat
c/o Public Health England
Centre for Radiation, Chemical and Environmental Hazards
Chilton, Didcot, Oxfordshire OX11 0RQ
© Crown copyright 2015

You may re-use this information (excluding logos) free of charge in any format or medium, under the
terms of the Open Government Licence v3.0. To view this licence, visit OGL or email
psi@[Link]. Where we have identified any third party copyright information you
will need to obtain permission from the copyright holders concerned. Any enquiries regarding this
publication should be sent to COM@[Link]

1
 (COM/2018/S1)

COMMITTEE ON MUTAGENICITY OF CHEMICALS IN FOOD, CONSUMER

PRODUCTS AND THE ENVIRONMENT

Statement on the quantitative approaches to the assessment of genotoxicity

data

Introduction and background to the current review

1) Genetic toxicology has traditionally been based on the development and

implementation of in vitro and in vivo assays designed to identify substances which
cause damage to DNA and/or other cellular components which regulate the fidelity of
the genome. The information derived from these testing strategies is used in a
qualitative manner, to establish whether or not the chemical is a genotoxic or
mutagenic hazard. Accordingly, risk management approaches are based on this
dichotomous (yes/no) decision, which helps protect against public exposure to
potentially genotoxic [and therefore potentially carcinogenic] agents (COM 2011;
EFSA 2011a). These assays are also useful during product development to ‘design
out’ genotoxic liability. However, this is a conservative approach that can result in
potentially valuable chemicals being screened out and discarded unnecessarily, or
the implementation of strategies to remove agents from the environment or food,
despite the fact that exposure, and risk, may be very low (Kirkland et al 2007;
Pottenger and Gollapudi 2010).

2) Conventional approaches to assessing the risk of chemicals which are toxic

/non-genotoxic are generally based on establishing a non-toxic level in in vivo
studies (the reference dose (RfD), derived from the point of departure (POD) and
applying uncertainty factors to estimate an exposure which represents a Health-
based Guidance Value (HBGV) such as a maximum acceptable daily intake (ADI))
(IPCS 2009). In general, for genotoxic carcinogens, the view is that there is no
threshold. A margin of exposure (MOE) approach based on a POD derived from a
carcinogenicity study can be utilised for carcinogens that are genotoxic and for which
there is unavoidable exposure (EFSA 2005; Barlow et al 2006; Benford et al 2010;
COC 2012). Currently, there is considerable interest in the development and
evaluation of methodologies which would enable the analysis of genotoxicity dose-
response data to be carried out in a quantitative manner.

3) Modification of the traditional [yes/no] approach to genotoxicity is a substantial

departure from current practices. Development of a strategy based more on
quantitative analyses would require extensive evaluations of the dose response
methodologies employed and a more detailed understanding of the relationship of
the genotoxicity endpoint to a human health effect, before it would be possible to
2
establish the appropriateness and/or usefulness of quantitative assessments of
genotoxicity data. Reports from the International Workshops on Genotoxicity Testing
(IWGT) working group in quantitative approaches to genetic toxicology risk
assessment (the QWG) (MacGregor et al 2015a,b) and publications arising from a
workshop organised by the Health and Environmental Sciences Institute (HESI)
Genetic Toxicology Technical Committee (GGTC) (summarised in White and
Johnson 2016) provide insight into how international groups are addressing this
changing risk assessment environment.

4) It is suggested that refining approaches for the assessment of genotoxicity

data could contribute to reductions and improvements in the use of animals in
toxicity testing (reduction, refinement, replacement; 3R’s) (Johnson et al 2014;
Soeterman-Hernández et al 2016).

5) The Committee on Mutagenicity (COM) first considered quantitative

approaches for assessing genotoxicity data, and how they may be used in chemical
risk assessment, at its Horizon Scanning exercise in June 2013. Members were
aware of the work being conducted by IWGT and HESI on quantifying genotoxic
responses and assessing non-linear dose-response relationships, and agreed that
the implications of this work should be considered. The possibility of developing
quantitative (or semi-quantitative) methods for the analysis of dose-response data
from in vivo genotoxicity studies for chemicals present in the environment, which
have not been tested for carcinogenicity, similar to that utilised for an MOE approach
using carcinogenicity data, was raised.

6) The COM were given a presentation by Dr George Johnson (Swansea

University), a member of key working groups, and considered papers summarizing
the key research in the field (MUT/2016/07; MUT/2017/02; MUT/2017/03).The
following key themes and questions were considered pivotal to the evaluation of this
topic:

 What dose response modelling methods are available, and which are most
appropriate for evaluating genotoxicity data?

 Which POD metric is best for assessing genotoxicity data and how can
appropriate benchmark responses (BMR) be established?

 How do factors such as endpoint, tissue, sampling time and study design
impact on assessing data quantitatively?

 Can quantitative information from genotoxicity data be used in risk

assessment, and if so, how?

 Is it possible to characterise carcinogenic risk from genotoxicity data alone?

This statement is a summary of the information considered by the COM and the
resultant discussions and opinions.
3
Current hazard and risk assessment approaches

7) The genotoxicity testing strategy currently recommended by COM (COM

2011) for the detection of mutagenic hazard is based upon a core set of in vitro tests,
chosen to provide information on three types of genomic damage; gene mutation,
clastogenicity and aneuploidy. These are followed, if necessary, by appropriate in
vivo tests designed to investigate whether in vitro genotoxic activity, including the
specific mutagenic effect identified, also occurs in vivo (i.e. in the whole animal). The
testing strategy may also include assays for specific target organs (e.g. site of
contact tissues or site of rodent tumours detected in carcinogenicity bioassays) or
germ cells. If a chemical is considered to be genotoxic it is generally assumed that
there is no exposure level below which there is no effect. For chemicals for which
potential exposure cannot be eliminated, the ALARA (as-low-as-reasonably-
achievable) or ALARP (as-low-as-reasonably-practicable) approach is advised. This
suggests that levels of the chemical must be controlled to ensure that intake is
minimised to be as low as reasonably, or technically, possible (Barlow et al 2006)
and is a widely adopted principle used by regulatory authorities in Europe and many
other regions. It is a purely qualitative (hazard-based) risk management approach,
and there is no consideration of the genotoxicity or carcinogenicity data in a
quantitative manner.

8) A few exceptions to the ‘no safe level’ assumption have previously been
established. These are based on the demonstration of a non-linear dose response
and a mode of action that is biologically relevant and exhibits a threshold. COM
generated a Guidance Statement on thresholds for in vivo mutagens in April 2010
(COM 2010). A number of different threshold terms were defined in this document
(i.e. true threshold, threshold dose, practical threshold, biologically meaningful
threshold, threshold mode of action).

9) An example of a threshold in a mutagenic response is that demonstrated by

some low molecular weight alkylating agents, a consequence of the repair of DNA
adducts. An extensive investigation and human risk assessment were undertaken
following the discovery of ethyl methanesulfonate (EMS), a known genotoxic
carcinogen, as an impurity in tablets of Viracept (nelfinavir mesilate), an HIV
protease inhibitor (Walker et al 2009; Muller and Gocke 2009). It was estimated that
consumption of contaminated drug batches at the maximal daily dose resulted in
patients ingesting EMS at up to 0.045 mg/kg/day (daily Viracept dosage of 2.92
g/day). The responsible pharmaceutical company (Roche) went on to perform a
comprehensive quantitative risk assessment of EMS, agreed with European
regulatory agencies (Muller and Singer 2009), and determined a ‘safe level’. The
disparity between the frequency of DNA adducts and mutations suggested that a
DNA repair factor was involved in the conversion of adducts to mutations, and that
this mechanism exhibits a threshold (Jenkins et al 2005; Doak et al 2007). Therefore,
it is possible that an organism could be subjected to a low level of DNA damage
without deleterious effects because the damage is effectively and efficiently repaired,
4
and it is only when repair mechanisms are exhausted or overwhelmed that a
mutation occurs. The risk assessment was based entirely on establishing a mode of
genotoxic action which had a clear threshold from which a POD was established.

10) The COC has defined approaches for risk characterisation of carcinogens and
these are described in Guidance Statements COC/G-05 and COC/G-06 (COC 2012;
2014). These are broadly in accordance with those proposed by European Food
Safety Authority (EFSA) (EFSA 2005). These include the MOE approach and the
threshold of toxicological concern (TTC). The TTC is a de minimis approach
developed to facilitate the risk management of substances, primarily contaminants in
food, for which good (or at least conservative) exposure estimates are possible but
when chemical-specific toxicity data, including genotoxicity data, are insufficient for
normal risk characterisation (Kroes et al 2004; Dewhurst and Renwick 2013).
Exposure levels below which safety concerns are not anticipated are given for
different classes of chemicals including genotoxic carcinogens1.

11) When applied to chemicals shown to be genotoxic and carcinogenic, the MOE
approach takes into account carcinogenic potency and estimated exposure (EFSA
2005; Barlow et al 2006). The MOE is calculated using a POD derived from suitable
rodent bioassay data or human epidemiology information, which is divided by the
measured or estimated exposure. The resulting value, which is a ratio, has been
classified by the COC (based on MOEs calculated using animal carcinogenicity data)
as follows:

 may be a concern (MOE<10,000);

 unlikely to be a concern (MOE 10,000-1,000,000) or

 highly unlikely to be a concern (MOE >1,000,000)

12) This method has gained acceptance by some regulatory bodies (including
EFSA, European Medicines Agency (EMA) and World Health Organisation (WHO))
for managing genotoxic carcinogens that cannot be avoided (e.g. contaminants).
EFSA recommend using a benchmark dose (BMD) as the POD for MOE
calculations. The approach uses mathematical modelling to calculate the lower one-
sided 95% confidence limit of a dose BMD i.e. the BMDL causing a defined response
(Benchmark Response (BMR) or Critical Effect Size (CES)), typically a 10% increase
in tumours in a cancer bioassay, i.e. the BMDL10 (EFSA 2009; 2016). This is also
replacing the ‘traditional’ no observed adverse effect level (NOAEL) approach for
non-cancer endpoints. Furthermore, because the models use all the dose–response

1
A TTC approach has been developed for DNA reactive (mutagenic) impurities in pharmaceuticals
(ICH M7R1A).
([Link]
_R1_Addendum_Step_4_31Mar2017.pdf)

5
data, confidence intervals provide a quantitative estimate of the uncertainties and the
quality of the data. To date this approach is only useful when good quality
carcinogenicity studies are available.

General publications on quantitative risk assessment of genotoxicity data

13) The COM considered a number of publications which examined the

application of a range of dose response modelling methods using data from a variety
of genotoxicity studies and the quantitative analyses resulting from them (Gollapudi
et al 2013; Johnson et al 2014; MacGregor et al 2015a,b). Three principle POD
metrics were compared; these were:

i) The no observable genotoxic effect level (NOGEL). This is the highest

experimental dose level where there is no statistically significant increase in
the genotoxic effect measured in the study.

ii) The threshold effect, lower confidence limit (TdL). This was used in the
EMS/Viracept analysis and is based on the assumption of a ‘hockey stick’
dose-response (Lutz and Lutz 2009; Gocke and Wall 2009). It involves fitting
a mathematical model which assumes that the dose response is bi-linear with
a region where there is no effect (it is similar to breakpoint dose (BPD) and
Slope Transition dose (STD) models). It has been argued that the
assumptions made with the use of this model need to be supported by
mechanistic data.

iii) The BMD approach. This is determined by mathematical modelling of the

dose –response curve and has been widely used in other branches of
toxicology. The approach involves, firstly, fitting a mathematical model to
experimental dose response data and, secondly, determining the BMD which
is estimated to produce a defined increase in the response over the
control/background level (BMR or CES). For example, using data from a
carcinogenicity study a 10% increase in tumours over the control incidence is
considered the BMR and the estimated dose is termed the BMD 10. The lower
one-sided 95% confidence limit (or bound) on the dose, termed the BMDL 10 is
then used as the POD in further considerations related to risk assessment
such as the derivation of a MOE.

14) Recent interest in the development of quantitative analysis of genotoxicity

data has focused on developing a similar BMD approach to that used elsewhere in
toxicology. The QWG and HESI groups agreed that BMD modelling is the preferred
approach for deriving a POD for genotoxicity data (Gollapudi et al 2013; Johnson et
al 2014; MacGregor et al 2015a). It was also noted that the BMDL usually produces
a lower and, hence, more conservative value for the POD than the other metrics
(NOGEL, BPD, STD) considered. The BMDL takes account of the amount of
variability in the data by considering the width of the confidence interval of the BMD;
i.e. the ratio of the BMDU (the upper one-sided 95% confidence limit (or bound) of

6
the BMD) to the BMDL has been proposed as a useful metric for the assessment of
the uncertainty in the BMD estimate (EFSA 2017). The COM agreed that the BMDL:
BMDU ratio reflects the overall quality of the data and will be a useful metric for use
in risk management scenarios (e.g. choice of uncertainty factor).

15) The COM acknowledges that developments in dose response modelling have
been made which make it possible for genotoxicity data, of acceptable quality, to be
analysed quantitatively rather than only qualitatively and that the authors of these
publications have provided essential contributions to these developments. The COM
broadly agreed with the conclusion that the BMD approach provides the best
representation of the dose response. However, it was agreed that these publications
present an overly optimistic view of the ease with which dose response modelling
can be applied. It was considered that more comprehensive discussion is required,
in particular the biological relevance of each endpoint and the choice of BMR and
CES, before the utility of the quantitative approaches can be realised. A lack of
consensus amongst users of the approach was also highlighted.

Benchmark dose approach

16) A number of areas were identified which were considered important for the
COM to address in more detail when evaluating the potential of using genotoxicity
data in a quantitative manner. In particular, there appear to be substantial
differences in the use of the dose response modelling and in the derivation of BMD
metrics. These differences include; choice of software package, the dose response
models, the statistical evaluation of model fit, the use of constraints/options, the
choice of BMR and methods for selecting or combining multiple BMDs. COM noted
that these areas are highly technical and require further clarification. It is important
that the rationales for the choices made are transparent and can be understood by
the toxicologists and risk assessors who will be working with the results or the
modelling processes.

Software, dose response modelling and BMD metrics

17) There are two principle software packages for the derivation of BMDs (Davis
et al 2011; EFSA 2016). The BenchMark Dose Software (BMDS) package was
developed by the US Environmental Protection Agency (EPA) in order to standardize
approaches to evaluating dose response assessments. The software has over 30
different mathematical models or model variants which can be used for the analysis
of quantal data, continuous data, nested developmental toxicology data, multiple
tumour analysis, and concentration-time data. The software is freely available on the
EPA website [Link] There are also extensive documentation
guides and training webinars on its use. New versions of the software are released
from time to time.

18) The PROAST software package has been developed by the Dutch National
Institute for Public Health and Environment (RIVM), and is freely available from their

7
website
[Link] A
comprehensive discussion of the software is available in the EFSA scientific opinion
(2009 – appendix p47-72). Various guides to its use are also provided with the
instructions for its installation.

19) COM noted that the PROAST software is frequently updated but that these
changes were not documented so that users could understand the impact of the
changes. COM also highlighted some uncertainty with regard to how the versions
are numbered. For example, the current version available at the RIVM website is
version 38.9. (July 2017) but the examples in EFSA (2016) use version 61.6.2

20) Both packages provide methods for fitting similar mathematical models to
dose-response data. However, there are some differences in the methodologies
used which are the subject of debate. Two major differences in the default
approaches have been described:

i) In the transformation of response data. PROAST (RIVM) uses the default

assumption of a log-normal distribution and transforms the data using logs
whereas BMDS (EPA) recommends choosing the most appropriate
transformation of the response data for the analyses (which may or may not
be a log transformation) based on an assessment of how well the models
describe the data, with the default being no transformation.

ii) Choice of BMR or CES: BMDS uses 1 standard deviation (1SD) above the
background as the default BMR for continuous data, whereas PROAST uses
a percentage increase e.g. 5%, 10% or some other percentage which may be
appropriate for a particular endpoint, above the background for the CES.
However, recent versions of BMDS can also be used in this way.

21) The COM also discussed the various dose response modelling methods used
in BMD analysis. The IWGT consider, for risk assessment, that it should be possible
to relate the POD to an acceptable exposure level by extrapolating from data which
includes mode of action (MOA) and mechanistic information if available (i.e. so that a
threshold mechanism, if demonstrated, can be taken into account). It was also noted
that BMD10 for quantal and continuous data will be substantially different. For
continuous genotoxicity data this represents a percent increase above a
spontaneous incidence as opposed to an absolute increase of a quantal parameter;
i.e. a 10% increase in micronuclei (MN) formation (from say 2 to 2.2 micronuclei
(MN)/1000) compared to a 10% increase in tumour incidence relative to the

2
In February 2018 PROAST released a new version 65.5 which allows for model averaging for
quantal data and two web applications of PROAST which avoid R and the installation of software.
These web applications do not, however, include all the options available in the R version of
PROAST,

8
unaffected control population (from, say 5% to 15% in the number of animals with
tumours in a carcinogenicity study) (MacGregor et al 2015a).

22) A direct comparison between PROAST and BMDS, based on BMDL10 and
BMD1SD values (respectively), from different in vivo and in vitro genotoxicity studies
on methylnitrosourea (MNU) was undertaken by Johnson et al (2014). From this
limited analysis, the authors concluded that the two approaches produce comparable
results and that both can be recommended for defining POD’s for continuous data. In
a study examining the correlation between genotoxicity and carcinogenicity, the
BMD05, calculated from bone marrow micronucleus (BMMN) data, was selected for
comparison of PODs with the BMDL10 values derived from carcinogenicity studies
(Soeteman-Hernández et al 2016). No rationale was given for selecting a 5%
increase as the BMR for calculating the BMMN POD, but the authors stated that the
choice of BMR was not crucial for their analyses. COM commented that the choice of
BMD05 as a BMR in this study was not transparent, which meant these results were
difficult to interpret. EFSA (2009) concluded that a default BMR value of 10% be
used for quantal and 5% for continuous toxicological data from animal studies in the
absence of specific information on what constitutes a biologically relevant change.
Both EFSA and EPA noted that, where specific information is available, the BMR
should be based on statistical or toxicological considerations. However, no specific
considerations of genetic toxicity data are given.

23) COM established that further explanations of the basic assumptions used and
the uncertainties that are applied to each model were required before they would be
able to come to any conclusions or make any recommendations on which software
model should be used. COM highlighted the current lack of concordance with regard
to choice of BMR for genotoxicity endpoints and what represents biologically relevant
responses. Furthermore, COM agreed that it was not obvious at present that the
BMD modelling could be transposed directly from its use with other toxicological
endpoints to use in genetic toxicology.

Endpoints and tissues

24) How the most accurate and/or conservative risk estimations should be derived
when using genotoxicity data has not yet been broadly addressed. For example, the
relative increase in DNA damage measured by the comet assay is likely to differ
appreciably from the relative increase in BMMN induced by the same chemical in the
same animals, since each response will be influenced by the chemical’s MOA, and
the ability to detect a response of a defined magnitude will be determined by the
dynamic range. Furthermore, it is not certain what the background levels of damage
or the induced increases of each genotoxicity biomarker represent in a risk
assessment scenario. The COM considered how results from the different types of
genotoxicity studies (and therefore endpoints) or different tissues will impact on the
derivation of POD values for use in potency estimations or risk assessment
scenarios. The importance of sampling time in the development and detection of

9
damage measured in genotoxicity assays (i.e. that sampling tissues at a single point
in time may not represent the peak response for different chemicals) was
highlighted, and may be important when PODs are being used to compare potency.
A number of publications were examined with the aim of addressing the importance
of differences. Many of these studies investigating the differences in genotoxicity
endpoints have focused on the alkylating agents ethylmethanesulphonate (EMS);
methylmethanesulphonate (MMS); 1-methyl-1-nitrosourea (MNU); and 1-ethyl-1-
nitrosourea (ENU), although some publications also examined polycyclic aromatic
hydrocarbons (PAH) as model genotoxicants.

25) A comprehensive evaluation of the dose-responses generated in vivo

following MNU and ENU exposure for a variety of endpoints including BMMN, gene
mutations in lacZ transgenic mice, or in Pig-A (in mice), was undertaken as part of a
programme of work developing POD-based evaluations of genotoxicity data
(Gollapudi et al 2013; Johnson et al 2014). The lowest BMDL value for each
chemical was derived from the in vivo gene mutation studies. These values were
conservative (lower) when compared with the values derived from the cancer
bioassay. However, COM commented that the generalisation that a value derived
from genotoxicity data will always be conservative compared to cancer bioassay,
cannot be made based only on data from this class of chemical. It is noted that whilst
potent in vivo genotoxins are likely to be potent carcinogens, dependent on MOA,
some weak genotoxins may also be potent carcinogens.

26) Zeller et al (2016) used MMS to examine the relationship of an endpoint with
the chemical MOA and to address the effect this has on the choice of CES/BMR.
The results showed that MMS acts primarily as a clastogen and its potency as a
gene mutagen is lower. Accordingly, it would not be appropriate to apply the same
CES to both chromosomal damage and gene mutation endpoints for this chemical.
The authors concluded that a ‘one-size-fits-all’ CES for genotoxicity data may be
sub-optimal because of the variability in baseline values, scoring systems and the
inherent differences in the characteristics of each end-point.

27) Detailed comparisons of endpoints and dose responses following

administration of a number of PAH’s including benzo[a]pyrene (BaP) and
dibenz(a,h)anthracene (DBahA) to MutaMouse were undertaken with a view to
improving the interpretation of genotoxicity dose response data (Wills et al 2016a).
BMMN, Pig-A and lacZ gene mutations were examined from a variety of tissues.
Covariate analyses (e.g. combining data from sexes or different tissues) were used
and the BMR was chosen as a 100% increase relative to control (i.e. doubling).
Confidence interval data indicated that tissue specific differences in BMD values
spanned an order of magnitude. Such large increases could have a significant
impact if they were used in a risk assessment or MOE evaluation to establish
acceptable human exposure limits. It is noted that sampling time was not considered
as a variable and it is known that mutation patterns vary with expression time and
show tissue specific responses.
10
28) An examination of the response of gpt-delta transgenic mice, which have a
lower spontaneous mutation frequency than in the earlier MutaTMMouse studies, to
EMS indicated substantially lower POD’s in the gpt-delta mice (Cao et al 2014)
although the NOGELs were similar. The COM commented that it was not clear
whether the results were a consequence of the lower baseline, a different strain of
mouse or a particularly sensitive endpoint, but the lower POD’s may also reflect
more heterogeneous data (larger BMDL-BMDU range) than in MutaMouse.
Accordingly, it demonstrated the importance of understanding a chemical MOA, the
appropriateness of the endpoint, the sensitivity of the genetic target and the quality
of the data in interpreting genotoxicity data quantitatively.

29) Establishing genotoxic MOA information was highlighted as vital in deciding

on the most relevant endpoints to use for POD determination (Gollapudi et al 2013;
MacGregor et al 2015b; Johnson et al 2014). Furthermore, it was suggested that the
selection of appropriate tissues for a quantitative analysis should be based on the
following: site-specific toxicity; mechanisms of toxicity; distribution and metabolism;
any chemical accumulation; cell proliferation; the ability for DNA repair capacity to be
induced by the chemical. Sensitivity of each endpoint and background mutation or
micronucleus frequency, will also affect the outcome of the analysis. COM noted
that, to date, there has been no discussion of the importance of sampling time when
deriving BMD’s despite the knowledge that time to maximum mutation frequency is
tissue specific in transgenic mice (Wang et al 2005) and time to maximum MN
frequency will depend on chemically-induced cell cycle delay.

30) The COM broadly agreed with the use of covariate analyses for combining
data from different tissues where this was appropriate. The preliminary data
available to them highlighted the importance of the selection of relevant endpoints
and tissues if quantitative data were going to be used effectively. However, it was
noted that it will be crucial for the developers of the software to provide clarity on
how these factors are incorporated into the modelling and how the data are intended
for use.

31) COM agreed that, whilst studies examining different endpoints and tissues
contribute useful information to this area of research and the development of the
quantitative analysis approaches, it was not possible to extrapolate findings from
specific chemicals or chemical classes (e.g. alkylating agents) to generate broad
assumptions. They considered that not enough is known about the quantitative
relationships of different genotoxic or mutagenic effects, pre-neoplastic lesions and
tumours to be able to interpret dose-response data accurately from a particular
endpoint/tissue for each chemical. They suggested that more robust analyses of a
larger number of more varied chemicals were required before any conclusions could
be reached. An evaluation of the use of comet assay data in quantitative analyses
has not been undertaken. COM recommends that a database which enables the
comparison of BMDs across chemicals, endpoints and tissues would provide useful
starting material for a more comprehensive evaluation of the utility of quantitative
11
assessment of genotoxicity data. The COM also pointed out that the applicability of
the quantitative approaches to germ cell mutagenesis had not been addressed.

Use in carcinogenicity risk assessment

32) Proponents of these developments have argued that using quantitative

methods for the analysis of genotoxicity data will provide the potential to move away
from a ‘hazard-only’ approach towards a risk-based approach (Johnson et al 2013;
MacGregor et al 2015a,b). To do this, a detailed evaluation of the biological
relevance of the endpoints and BMR’s is required. COM examined some
publications from groups exploring the possibility of using POD’s derived from in vivo
genotoxicity studies in place of those generated from long term carcinogenicity
studies, for example in MOE assessments (Sanner and Dybing 2005; Soeteman-
Hernández et al 2015; Soeteman-Hernández et al 2016). It is understood that, with
regard to potential exposure to chemicals that are (or could be) genotoxic
carcinogens, there are a number of risk management needs. These range from
determination of the potential level of concern for exposure to unavoidable
contaminants or constituents of the diet, to market authorisation of new products
(such as pesticides and human medicines). Hence, it is unlikely that a single
approach would be suitable for all risk management situations. Indeed, as discussed
above (para 11-12), the approach currently utilised by the COC varies depending on
the risk management context.

33) The quantitative use of dose response data in MOE approaches for genotoxic
chemicals in food was considered by Benford (2016). Attention was drawn to the
importance of considering factors such as study design and quality, strain and
species, and chemical MOA when using carcinogenicity data, and that these factors
would also be critical if genotoxicity data are used. It is noted that a comparison of
potency in carcinogenicity and genotoxicity assays is necessary using a broad range
of carcinogen classes and MOAs. EFSA (2009) recommend the MOE approach for
substances that are both genotoxic and carcinogenic, when risk assessment is
necessary. They proposed the use of the BMDL10 as the POD based upon tumour
data from carcinogenicity studies. To date, EFSA has not expressed a view on the
use of a POD derived from genotoxicity data in place of a carcinogenicity value.

34) The COM was provided with a number of publications detailing comparisons
of mutagenic and carcinogenic potency using BMD dose response modeling. A
preliminary evaluation was undertaken by Sanner and Dybing (2005) who concluded
that there was a correlation between carcinogenic and mutagenic potencies. A
framework, using the lowest effect dose (equivalent to the LOGEL) in a micronucleus
study, was proposed as having the potential to be used in regulatory settings when a
chemical was considered to be mutagenic, but for which carcinogenicity studies are
either not available or of poor quality.

12
35) A comprehensive evaluation of potency estimates was undertaken by
Hernández et al (2011) using 18 chemicals listed as either IARC class 1 or 2A
carcinogens. BMD10 values for carcinogenicity and genotoxicity were derived using
the PROAST dose-response modelling current at the time. Those from genotoxicity
data were based on a range of endpoints (BMMN, comet, mutations in transgenic
mice) from various tissues and from multiple studies. Some of the carcinogenicity
studies, however, used only two treatment dose levels and different exposure routes
were used in some cases. The authors concluded that there was some degree of
association and a correlation between the BMDs for mutagenicity and
carcinogenicity, despite the differences in study designs and routes of exposure.

36) An extension of this study, using similar methodologies, evaluated 48

chemicals, (Soeteman-Hernández et al 2016) and calculated BMD05 values from MN
studies. The log10 of these BMD05 values were plotted against the log10 of the
cancer BMD10values. The BMD05 values were calculated for individual datasets (i.e.
if there were several studies available, the data were not pooled) and a version of
PROAST current at that time was used for dose-response modelling of both sets of
data. The plot of the data showed a wide scatter but the authors concluded that
there was a positive correlation between the BMMN and carcinogenic potencies (as
measured by the BMDs), although prediction of carcinogenic potency from the
genotoxicity data had an uncertainty of two orders of magnitude (i.e. factor of 100).
As stated previously (para 22), it is understood that choice of BMD05 as the BMR
was based on data comparisons rather than a value that represents a biologically
meaningful effect and is not proposed for use in risk assessments.

37) The COM considered that the causal relationship implied by the association of
the mutagenicity and carcinogenicity potency was problematic and were
unconvinced by the 1:1 associations that seemed to be inferred in the publications.
They felt that the complex relationships between adducts, mutations, pre-neoplastic
lesions and tumours make it unlikely that evidence for a simple ratio/association is
robust and that pursuit of a simple correlation is overly ambitious. It was considered
possible that the dose inducing a biologically relevant genotoxic effect in an
appropriate tissue would be lower and therefore be a more conservative POD for
protecting health than a BMDL10 for cancer. However, before this could be
substantiated, COM felt that analysis of more datasets using a much broader range
of chemicals and chemical classes is essential before any assumptions could be
made. In particular, they pointed out that much less is known about the pattern of
responses for weak genotoxins: for example, styrene, which causes tumours in nasal
turbinates, induces relatively weak responses in genotoxicity assays.

Study designs, data quality and use of uncertainty factors

38) The COM considered it important to evaluate the impact of study design, and
to consider the quality of the available data before conducting or interpreting
quantitative analysis of genotoxicity data in order to generate PODs. It was noted

13
that for an optimal statistical design for BMD modelling it is preferable to distribute a
fixed number of animals in a study into more dose groups with fewer animals per
group. However, this may not accord with current OECD guidelines for in vivo
genotoxicity tests. Nevertheless, it was agreed that there is some flexibility within
these study designs and that the two designs (i.e. for OECD and for BMD estimation)
were not necessarily incompatible Current OECD guideline designs of genotoxicity
studies were suitable for quantitative analysis for chemicals for which there are
sufficient data to determine a dose-response relationship. For chemicals where a
dose response has not been established it will be difficult to determine a POD and
carryout a BMD assessment from OECD guideline study designs which typically use
no more than three dose levels.

39) Data quality is partly reflected in the width of confidence intervals, which is
also dependent upon the number of dose groups and animals per group. The COM
commented that guidance should be provided on what level of uncertainty in the data
and what ratio of BMDU: BMDL would be considered unacceptable.

40) One important aspect of fitting mathematical models to dose-response data is

testing whether the model is a good fit to the data. Models which are not a good fit
should not be used. The COM noted that the choice of model based solely upon the
results of ‘goodness of fit’ tests is a contentious area when a number of models
provide a satisfactory fit. EFSA (2016) recommend that a model averaging approach
is used, rather than a single default model. However, it is not clear if there is
currently suitable software, readily available, to use the method with quantitative
data. COM note that there are ongoing discussions with regards to optimising study
designs, especially with regard to what is considered to be generation of the most
suitable dose response. These factors require clarification before any definitive
conclusions can be drawn.

41) Whilst the use of uncertainty factors was introduced by some authors, it was
noted that more precise attempts at quantification has not been undertaken
(Johnson et al 2014; MacGregor 2015b). COM commented that the uncertainty
factors should be a reflection of the data quality, species differences, the endpoint
measured, and presence or absence of a threshold mechanism, but until further
examples are available, no conclusion can be reached.

Use of in vitro genotoxicity data for deriving POD’s

42) There are a number of recent publications that have examined the use of
BMD assessments of in vitro genotoxicity studies for potency comparisons, or for
comparisons of in vitro with in vivo BMD’s (Soeterman-Hernández et al 2015; Bemis
et al 2016; Wills et al 2016b). COM commented that, whilst an interesting innovation,
these approaches are at an early stage of development and currently cannot be
considered for risk assessment scenarios or for potency ranking. COM decided not
to consider this in vitro use of quantitative models further at this time.

14
Overall discussion and conclusions

43) COM considered the current literature on quantitative analyses of dose-

response data from genotoxicity studies, including the reports from IWGT and
ILSI/HESI, and discussed the recent developments of the approaches in this area.

44) It is noted that a move towards quantitative assessment of data is a significant

departure from the current practices which are based principally on establishing only
whether a chemical represents a mutagenic hazard. Discussions such as this should
enable exploration of concepts which underpin the use of genotoxicity data in risk
assessment; for example, would such an approach imply that all genotoxic
chemicals have an exposure level below which the risk is considered tolerable and
how might this level be identified?

45) Broadly, COM is in agreement with the principle of evaluating genetic

toxicology data quantitatively. It is hoped that such approaches have the potential to
improve the interpretation of genotoxicity data, potentially reducing the need for long-
term carcinogenicity studies, and, hence, reduce the number of animals used in
chemical risk assessment (with 3R’s benefit). As these approaches are developed
and their utility demonstrated, there may be scope for them to be incorporated into
regulatory frameworks.

46) COM recognised the importance of the developments in the software and use
of BMD methodologies to evaluate genotoxicity quantitatively. However, it was noted
that, to date, much of the analyses have been performed by a small number of
specialists and that the continual modifications to versions of the software made it
difficult for those less well acquainted with the models and approaches to understand
the significance of the changes. Many of the analyses are complex, and will require
explanation and clarification before they can be considered by a broader audience.
Some aspects of the dose-response modelling continue to evolve whilst other
aspects vary between the developers of the methods. Therefore COM could not
conclude on the appropriateness of the different models for use with genotoxicity
data. It was concluded that changes to software should be documented and if
software comparisons are undertaken, that it is made clear which aspects of the
modelling are being compared.

47) With regards to the usefulness of POD’s from genotoxicity data in risk
assessment, COM recommend that a detailed evaluation of the different software
methodologies is undertaken before any conclusions can be reached. Furthermore,
the COM felt that a clarification of the outstanding issues in the use of the
methodology was needed (e.g. choice of dose-response models, use of constraints)
so that non-experts in the field were aware of the implications (if any) of the use of

15
the different software packages and options that have been proposed. COM
suggested that precise descriptions of the methodologies and underlying
assumptions (explicit and implicit) are developed so that a detailed and informed
evaluation can be undertaken by potential users of the methods.

48) Guidance is needed on how to assess data quality and goodness of fit of the
models to help decide on the suitability of a dataset for modelling. Clarification is
needed on the level of uncertainty in the estimates in terms of the upper to lower
confidence limit ratios which are considered acceptable and the factors which drive
these uncertainties.

49) COM noted that there is a lack of consensus with regards to the selection of
an appropriate CES/BMR for specific genotoxicity endpoints and that this was a
complex area which requires more extensive discussion and evaluation. COM felt
that it was unlikely that a similar size response (e.g. 10% increase over the negative
control value) would be suitable for different genotoxicity endpoints such as, for
instance, micronucleus induction and gene mutations. Selecting a BMR will require
an understanding of the biological relevance of each endpoint and characterisation
of the relative magnitude of response over background. Further investigations of
what constitutes an appropriate BMR/CES for determining BMDs using a variety of
genotoxicity study types is needed, with emphasis placed on the biological relevance
of the choice of BMR/CES. Overall, it was difficult to conclude on selection of
BMR/CES given the limited datasets available.

50) COM concluded that it was not possible to make any broad assumptions
based on data from limited chemical classes. COM remain to be convinced of the
close associations in comparisons of genotoxicity and carcinogenicity data reported
by some investigators and highlighted the need for a more extensive evaluation of
suitable datasets including a broader assessment of different chemicals classes,
genotoxicity endpoints, tissues and timepoints. It was considered that BMD’s from
genotoxicity studies would, generally, be lower than those from carcinogenicity
studies. However, at present, there are insufficient examples and a lack of
understanding of the appropriate BMRs for the various endpoints for COM to draw
any definitive conclusions. Consequently, the COM, at present, was unable to make
any recommendations for the inclusion of quantitative genotoxicity data in MOE
calculations.

COM

March 2018

16
References

Barlow S, Renwick AG, Kleiner J, Bridges JW, et al (2006) Risk assessment of

substances that are both genotoxic and carcinogenic report of an International
Conference organized by EFSA and WHO with support of ILSI Europe. Food Chem
Toxicol. 44(10) 1636-50.

Bemis, J, Wills, J, Bryce, S, Torous, D, Dertinger, S, Slob, W (2016) Comparison of

in vitro and in vivo clastogenic potency based on benchmark dose analysis of flow
cytometric micronucleus data. Mutagenesis 31(3) 277– 285.

Benford D, Bolger PM, Carthew P, Coulet M, et al (2010) Application of the Margin of

Exposure (MoE) approach to substances in food that are genotoxic and
carcinogenic. Food Chem Toxicol. 48 Suppl 1:S2-24.

Benford DJ (2016) The use of dose-response data in a margin of exposure approach

to carcinogenic risk assessment for genotoxic chemicals in food. Mutagenesis 31(3)
329–331.

Cao X, Mittelstaedt RA, Pearce MG, et al (2014) Quantitative dose-response

analysis of ethyl methanesulfonate genotoxicity in adult gpt-delta transgenic mice.
Environ Mol Mutagen. 55(5) 385-99.

COC (2012) Cancer risk characterisation methods. Committee on Carcinogenicity.

Available online:
[Link]

COC (2014) Defining a Point of Departure and Potency Estimates in Carcinogenic

Dose Response. Committee on Carcinogenicity. Available online:
[Link]
4/Defining_a_point_of_departure_and_potency_estimates_in_carcinogenic_dose_re
[Link].

COM (2010) Guidance: assessment of thresholds for in vivo mutagens. Committee

on Mutagenicity. Available online:
[Link]
mutagens.

COM (2011) A strategy for testing of chemicals for genotoxicity. Committee on

Mutagenicity. Available online: [Link]
strategy-for-testing-of-chemicals-for-genotoxicity.

Davis JA, Gift JS, Zhao QJ (2011) Introduction to benchmark dose methods and
U.S. EPA's benchmark dose software (BMDS) version [Link] App Pharmacol.
254(2) 181-91.

Dewhurst I, Renwick AG (2013) Evaluation of the Threshold of Toxicological

Concern (TTC)--challenges and approaches. Regul Toxicol Pharmacol. 65(1) 168-
77.

17
Doak SH, Jenkins GJ, Johnson GE, Quick E, Parry EM, Parry JM (2007)
Mechanistic influences for mutation induction curves after exposure to DNA-reactive
carcinogens. Cancer Res. 67(8) 3904-11.

EFSA (2005) Opinion of the Scientific Committee on a request from EFSA related to
a harmonized approach for risk assessment of substances which are both genotoxic
and carcinogenic. The EFSA Journal 282 1-31.

EFSA (2009) Guidance of the Scientific Committee on a request from EFSA on the
use of the benchmark dose approach in risk assessment. The EFSA Journal 1150 1-
72.

EFSA (2011a) Scientific opinion on genotoxicity testing strategies applicable to food

and feed safety assessment. The EFSA Journal 9(9) 2379-448.

EFSA (2011b) Use of BMDS and PROAST software packages by EFSA Scientific
Panels and Units for applying the Benchmark Dose (BMD) approach in risk
assessment. EN-113. Available online: [Link].

EFSA (2016) Review of the Threshold of Toxicological Concern (TTC) approach and
development of new TTC decision tree. EFSA Supporting Publication EN-1006.
Available online: [Link].

EFSA (2017) Update: use of the benchmark dose approach in risk assessment.
EFSA Journal 15(1) 4658.
Gocke E, Wall M (2009) In vivo genotoxicity of EMS: statistical assessment of the
dose response curves. Toxicol Lett. 190(3) 298-302.

Gollapudi BB, Johnson GE, Hernández LG, Pottenger LH et al (2013) Quantitative

approaches for assessing dose-response relationships in genetic toxicology studies.
Environ Mol Mutagen. 54(1) 8-18.

Hernández LG, Slob W, van Steeg H, van Benthem J (2011) Can carcinogenic
potency be predicted from in vivo genotoxicity data?: a meta-analysis of historical
data. Environ Mol Mutagen. 52(7) 518-28.

IPCS (2009) Environmental Health Criteria (EHC) 240: Principles and methods for
the risk assessment of chemicals in food. International Programme on Chemical
Safety. World Health Organization. Available online:
[Link]
Jenkins GJ, Doak SH, Johnson GE, Quick E, Waters EM, Parry JM (2005) Do dose
response thresholds exist for genotoxic alkylating agents? Mutagenesis 20(6) 389.

Ji, Z; LeBaron, MJ, Schisler, MR, Zhang, F, Bartlels, MJ, Gollapudi, BB, Pottenger
LG (2016) Dose–Response for Multiple Biomarkers of Exposure and Genotoxic
Effect Following Repeated Treatment of Rats with the Alkylating Agents, MMS and
MNU Mutagenesis 31(3) 297–308.

18
Johnson GE, Soeteman-Hernández LG, Gollapudi BB, Bodeger OG, et al (2014)
Derivation of point of departure (PoD) estimates in genetic toxicology studies and
their potential applications in risk assessment. Environ Mol Mutagen. 55(8) 609-23.

Kirkland D, Pfuhler S, Tweats D, Aardema M, et al (2007) How to reduce false

positive results when undertaking in vitro genotoxicity testing and thus avoid
unnecessary follow-up animal tests: Report of an ECVAM Workshop. Mutat Res.
628(1) 31-55.
Kroes R, Renwick AG, Cheeseman M, Kliener, J et al (2004) Structure-based
thresholds of toxicological concern (TTC): guidance for application to substances
present at low levels in the diet. Food Chem Toxicol. 42(1) 65-83.
Lutz WK, Lutz RW (2009) Statistical model to estimate a threshold dose and its
confidence limits for the analysis of sublinear dose-response relationships,
exemplified for mutagenicity data. Mutat Res. 678(2) 118-22.
MacGregor JT, Frötschl R, White PA, Crump KS, et al (2015a) IWGT report on
quantitative approaches to genotoxicity risk assessment I. Methods and metrics for
defining exposure-response relationships and points of departure (PoDs). Mutat Res
Genet Toxicol Environ Mutagen. 783 55-65.

MacGregor JT, Frötschl R, White PA, Crump KS, et al (2015b) IWGT report on
quantitative approaches to genotoxicity risk assessment I. Methods and metrics for
defining exposure-response relationships and points of departure (PoDs). Mutat Res
Genet Toxicol Environ Mutagen. 783 66-78.

Muller L, Gocke E (2009) Considerations regarding a permitted daily exposure

calculation for ethyl methanesulfonate. Toxicol Lett. 190(3) 330-3.

Muller L, Singer T (2009) EMS in Viracept--the course of events in 2007 and 2008
from the non-clinical safety point of view. Toxicol Lett. 190(3) 243-7.

Pottenger LH, Gollapudi BB (2010) Genotoxicity testing: moving beyond qualitative

"screen and bin" approach towards characterization of dose-response and
thresholds. Environ Mol Mutagen. 51(8-9) 792-9.

Sanner T, Dybing E (2005) Comparison of carcinogenic and in vivo genotoxic

potency estimates. Basic Clin Pharmacol Toxicol. 96(2) 131-9.

Soeteman-Hernández LG, Fellows MD, Johnson GE, Slob W (2015) Correlation of In

Vivo Versus In Vitro Benchmark Doses (BMDs) Derived From Micronucleus Test
Data: A Proof of Concept Study. Toxicol Sci. 148(2) 355-67.

Soeteman-Hernández LG, Johnson GE, Slob W (2016) Estimating the carcinogenic

potency of chemicals from the in vivo micronucleus test. Mutagenesis 31(3) 347-358.

Walker VE, Casciano DA, Tweats DJ (2009) The Viracept-EMS case: impact and
outlook. Toxicol Lett. 190(3) 333-9.

19
Wang J, Liu X, Heflich RH, Chen T (2004) Time course of cII gene mutant
manifestation in the liver, spleen, and bone marrow of N-ethyl-N-nitrosourea-treated
Big Blue transgenic mice. Toxicol Sci. 82(1)124-8.

White PA, Johnson GE (2016) Genetic toxicology at the crossroads—from qualitative

hazard evaluation to quantitative risk assessment Mutagenesis 31(3) 233–37.
Wills JW, Long AS, Johnson GE, Bemis JC, Dertinger SD, Slob W, White PA
(2016a) Empirical analysis of BMD metrics in genetic toxicology part II: in vivo
potency comparisons to promote reductions in the use of experimental animals for
genetic toxicity assessment. Mutagenesis 31(3) 265-75.

Wills JW, Johnson GE, Doak, SH, Soetman-Hernández LG, Slob W, White PA
(2016b) Empirical analysis of BMD metrics in genetic toxicology part I: in vitro
analyses to provide robust potency rankings and support MOA determinations
Mutagenesis 31(3) 255-63.

Zeller A, Tang L, Dertinger S, Funk, J, Duran-Pacheco, G Guérard, M (2016) A

proposal for a novel rationale for critical effect size in dose–response analysis based
on a multi-endpoint in vivo study with methyl methanesulfonate Mutagenesis 31(3)
239– 253.

20
Glossary

Acceptable daily intake (ADI): The estimate of the amount of a chemical in food or
drinking-water, expressed on a body weight basis that can be ingested daily over a
lifetime without appreciable health risk to the consumer. It is derived on the basis of
all the known facts at the time of the evaluation. The ADI is expressed in milligrams
of the chemical per kilogram of body weight (a standard adult person weighs 60 kg).

Adduct: A chemical grouping which is covalently bound to a large molecule such as

DNA or protein.

ALARA/ALARP: As Low As is Reasonably Achievable/ As Low As is Reasonably

Practicable: A risk management approach under which exposure to a substances or
mixture is reduced to the lowest level that it is deemed to be reasonably practicable
to achieve in particular circumstances.

Alkylating agents: Chemicals which leave an alkyl group covalently bound to

biologically important molecules such as proteins and nucleic acids (see adduct).
Many alkylating agents are mutagenic, carcinogenic and immunosuppressive.

Aneugen: A chemical which induces aneuploidy (qv).

Aneuploidy: The occurrence of an abnormal number of chromosomes in a cell,

such that the total number of chromosomes within the cell is not an exact multiple of
the normal (haploid) number. Chromosomes may be lost (monosomy) or gained
(trisomy) during cell division. An extra or missing chromosome is a common cause of
genetic disorders (birth defects or spontaneous abortions). Some cancer cells also
have abnormal numbers of chromosomes. (Chemical induction of aneuploidy is
aneugenicity).

Benchmark dose (BMD): A dose of a substance associated with a specified low

incidence of risk, generally in the range of 1–10%, of a health effect; the dose
associated with a specified measure or change of a biological effect.

Benchmark dose (BMD) modelling: A mathematical modelling approach to dose-

response assessment that aims to be more quantitative than the NOAEL process.
An estimate of the dose that corresponds to a particular level of response (a
benchmark response), often 10% is derived and a measure of uncertainty is also
calculated. The lower one-sided 95% confidence limit (or bound) on the benchmark
dose is called the BMDL and the upper one-sided 95% confidence limit (or bound) is
the BMDU. The BMDL can be used as the point of departure for derivation of a
health-based guidance value or a margin of exposure.

BMDL: BMDU The ratio of the lower and upper one-sided 95% confidence limits of
the benchmark dose which is a measure of the precision of the BMD. It accounts for
the uncertainty in the estimate of the dose-response due to characteristics of the
experimental design, such as sample size.

21
Benchmark response (BMR): An adverse effect, used to define a benchmark dose
from which a reference dose can be developed. The change in response rate over
background of the BMR is usually in the range of 5-10%, which is the limit of
responses typically observed in well-conducted animal experiments. This term is
often used synonymously with Critical Effect Size (CES).

Clastogen: An agent that produces chromosome breaks and other structural

aberrations in chromosomes such as translocations. Clastogens may be viruses or
physical agents as well as chemicals. Clastogenic events play an important part in
the development of some tumours (clastogenicity).

Critical effect size (CES): The magnitude of the adverse effect seen at a lowest
dose when a vulnerable population is exposed to a chemical. This term is often used
synonymously with Benchmark Response (BMR).

Continuous Data: Continuous data is quantitative data that can be measured and
has an infinite number of possible values within a selected range.

Gene mutation: A mutation resulting from a change in a single base pair in the DNA
molecule (also called point mutation).

Genotoxicity: Genotoxicity refers to interaction with, or damage to, DNA and/or

other cellular components which regulate the fidelity of the genome. It is a broad
term that, as well as mutation includes damage to DNA such as the production of
DNA adducts, by the chemical itself or its metabolites. Cells have the capacity to
protect themselves from such potentially lethal or mutagenic genotoxic effects by
many repair processes and therefore many genotoxic events do not become evident
as mutations. However, the capacity to damage the genome (genotoxicity) is an
indicator of potential mutagenicity. Thus, some methods that measure genotoxicity
may not provide direct evidence of heritable mutation.

Genotoxic carcinogen: Carcinogen whose primary mode of action involves

deoxyribonucleic acid or chromosomal alterations.

Health-based guidance value (HBGV): A numerical value derived by dividing a

point of departure (a no-observed- adverse-effect level, benchmark dose or
benchmark dose lower confidence limit) by a composite uncertainty factor to
determine a level that can be ingested over a defined time period (e.g. lifetime or 24
h) without appreciable health risk.

LacZ gene mutations: See transgenic gene mutation models.

Margin of exposure (MOE) approach: A methodology that allows the comparison

of the risks posed by different genotoxic and carcinogenic substances. The MOE
approach uses a reference point, often taken from an animal study and
corresponding to a dose that causes a low but measurable response in animals. This

22
reference point is then compared with various dietary intake estimates in humans,
taking into account differences in consumption patterns.

Micronuclei (MN) (including bone marrow micronuclei (BMMN)): Whole or

broken chromosomes that fail to segregate normally during cell division and may be
lost from the main nuclei, but remain in the body of the cell forming micronuclei.
Centromere positive micronuclei contain DNA and/or protein material derived from
the centromere. The presence of centromere positive micronuclei following exposure
to chemicals in vitro or in vivo can be used to evaluate the aneugenic (qv) potential
of chemicals.

Mode of action (MOA): A biologically plausible sequence of key events leading to

an observed effect supported by robust experimental observations and mechanistic
data. A mode of action describes key cytological and biochemical events—that is,
those that are both measurable and necessary to the observed effect—in a logical
framework.

No Observed Adverse Effect Level (NOAEL): Greatest concentration or amount of

a substance, found by experiment or observation, that causes no adverse alteration
of morphology, functional capacity, growth, development or lifespan of the target
organism distinguishable from those observed in normal (control) organisms of the
same species and strain under the same defined conditions of exposure.

No Observed Genotoxic Effect Level (NOGEL): This is the highest experimental

dose level where there is no statistically significant increase in the genotoxic effect
measured in the study.

Pig-A gene mutation assay: An assay which utilises the Pig-A gene which codes
for one subunit of a glycosylphosphatidyl inositol anchor protein. Loss of function
arising from Pig-A mutations can readily be assessed using straightforward
immunochemistry and flow cytometric methods, thus making it a useful to measure
gene mutations induced by chemicals or radiation. The development of in vivo and in
vitro models are ongoing but are not yet recognised as fully evaluated and there are
no OECD guidelines.

Point of departure (POD): A reference point on a toxicological dose-response curve

established from experimental data which corresponds to an estimated low or no
effect level. Used for hazard characterisation (see also RfD).

Quantal Data: A quantal dose response is one in which the effect is designated to
be an all or nothing response (i.e. an animal has a tumour or it does not).

Reference dose (RfD): An estimate of the daily exposure dose that is likely to be
without deleterious effect even if continued exposure occurs over a lifetime.

Software: PROAST and BMDS – two software packages for benchmark dose
modelling. The PROAST software package developed by the Dutch National Institute

23
for Public Health and Environment (RIVM). BenchMark Dose Software (BMDS)
developed by the US Environmental Protection Agency (EPA).

Threshold: Dose or exposure concentration below which a biological effect is not

expected.

Transgenic animal models: Animals which have extra (exogenous) fragments of

DNA incorporated into their genomes. This may include reporter genes to assess in-
vivo effects such as mutagenicity in transgenic mice containing a recoverable
bacterial gene (lacZ or lacI). Other transgenic animals may have alterations of
specific genes believed to be involved in disease processes (e.g. cancer). For
example strains of mice have been bred which carry an inactivated copy of the p53
tumour suppressor gene, or an activated form of the ras oncogene which may
enhance their susceptibility of the mice to certain types of carcinogenic chemicals.

24