Repeated-dose 90-day Oral Toxicity Studies on Whole ... - Europabio

<strong>Repeated</strong>-<strong>dose</strong> <strong>90</strong>-<strong>day</strong> <strong>Oral</strong>
<strong>Toxicity</strong> <strong>Studies</strong> on Whole
Food/Feed
Presenters:
Kevin Leiner, Margaret Nemeth, and
Keith Ward
1

Introduction
• EFSA has stated that its guidance on repeated-<strong>dose</strong> <strong>90</strong>-<strong>day</strong> oral toxicity
studies in rodents on whole food/feed is “not intended to provide
prescriptive experimental test protocols.”
• EuropaBio Members agree there should be flexibility in the design of future
studies with whole food/feed and that justification for the approach to the
study is important.
• The new EFSA guideline will:
– Impact design, conduct, analysis and interpretation both toxicologically and
statistically.
– Create challenges that require additional discussion with EFSA and other
stakeholders in the EU.
• EuropaBio Members appreciate the opportunity to discuss these new
challenges, and their proposed solutions to them, with key stakeholders in
the EU.
2

Existing safety data and the
weight of evidence
• Prior to the conduct of the <strong>90</strong>-<strong>day</strong> study several safety
assessments have been conducted including:
– Compositional analysis, agronomic performance, and molecular analysis of the
crop.
– Evaluations of the inserted trait and its protein product:
• Bioinformatic evaluations, in vitro digestibility and heat stability assays as well as
protein toxicity studies on a case-by-case basis.
• Establishing Substantial Equivalence between the GM and
conventional crop, and an absence of adverse findings in the
indicated tests, provides a weight of evidence for the safety of
the whole food (Codex, 2009).
• EuropaBio Members believe the 15 + year history of safe use of
GM crops evaluated and approved under this approach
attests to its strength.
3

Hypothesis of safety and
study justification
• In such cases, EuropaBio Members believe there is no
specific hypothesis to test in the <strong>90</strong>-<strong>day</strong> study and
thus no justification for them.
– EFSA published opinion appears to agree with this position.
• EFSA Panel on Genetically Modified Organisms (GMO). Scientific opinion-
Guidance for risk assessment of food and feed from genetically modified
plants. EFSA Journal 2011; 9(5):2150
• However, if potential indications of toxicity are
present in the existing data, then EuropaBio Members
believe the principles of Hazard Identification should
be followed to maximise the possibility of detecting
toxic effects of the test diet.
4

Hazard Identification and
multiple <strong>dose</strong> levels
• For Hazard Identification purposes, administering a <strong>dose</strong> lower than the
highest possible (which maintains nutritional balance) is of disputable
value.
• Testing at the highest possible concentration maximises exposure and thus
the potential to detect unanticipated adverse effects.
– Consistent with the limit <strong>dose</strong> approach detailed in OECD TG 408 (1998).
– Requirement of nutritional balance limits concentration in diet.
– EuropaBio Members estimate the highest possible concentration of soybean
and maize will typically exceed human intake in the EU 10X to 100X (WHO,
2003; EFSA 2011c).
• It is difficult to scientifically justify the inclusion of the low <strong>dose</strong> group if
regulating on the basis of Hazard Identification.
– A “clean” low <strong>dose</strong> level would then likely be irrelevant if the high <strong>dose</strong> is
adverse.
5

The regulatory paradigm for
GM crops
• Based on practical experience, EuropaBio
Members have presumed the current
regulatory paradigm for GM crops is based
on Hazard Identification (or rather an
absence of Hazard).
• EFSA’s inclusion of a low <strong>dose</strong> level could be
interpreted as a willingness to regulate GM
crops on the principle of Risk.
– Risk = Hazard × exposure
6

Section summary
• EuropaBio Members believe:
– When there is no evidence of unintended changes, the
<strong>90</strong>-<strong>day</strong> study would not normally be expected to add
value to the safety assessment – it is largely
confirmatory.
– The current regulatory paradigm for GM crops is based
on Hazard Identification.
– The purpose of testing a <strong>dose</strong> below the highest dietary
incorporation level that maintains nutritional balance is
unclear and needs to be resolved.
7

Discussion
• EuropaBio Members agree with EFSA’s Scientific
Opinion (2011a) that an applicant could forego
conducting a <strong>90</strong>-<strong>day</strong> study if data indicating the GM
crop and its parental isoline:
– Are Substantially Equivalent in regard to nutritional composition.
– Demonstrate no indications of unintended adverse effects in
molecular, compositional, or phenotypic analyses.
• EuropaBio Members request clarification on whether
inclusion of two (or more) <strong>dose</strong> levels is mandatory
and, if so, how results from the low <strong>dose</strong> will be
used.
8

Part 1: Study design and conduct
9

Challenges in study design
and conduct
1. Designing for appropriate power
2. Experimental design and conduct
a) Practical aspects regarding cage location
b) Variation controls
3. Housing conditions and impact on the
Experimental Unit (ExpU)
4. Statistical power investigation
10

1. Defining toxicologically
relevant differences
• As per guidance, relevant endpoints, consistent with OECD TG
408, will be monitored.
– TG 408 specifies that each test animal will be monitored on >200 individual
endpoints.
– Ensures a thorough examination of the whole animal and that known
mechanisms of toxicity will be evaluated.
• Toxicologists will review the data for consistent trends in related
endpoints associated with toxic mechanisms.
– ↑ serum ALT, ↑ liver weight, histopathological alteraons of liver ssue.
– ↑ serum ALT, in the absence of other indicaons of liver damage, is of equivocal
importance.
• To reach scientifically valid conclusions, the weight of evidence
must be considered to determine if the difference observed is
treatment-related.
11

1. Defining toxicologically relevant
differences (continued)
• Identifying a toxicologically-relevant difference for individual
endpoints in isolation is very difficult, especially when there
is no evidence of adverse changes in related parameters.
• However, to perform power calculations, relevant effect sizes
must be defined.
• EFSA guidance acknowledges that this is no easy task, and
does not provide specific guidance in this respect.
• To determine endpoints and effect sizes with the greatest
potential to identify toxicity, EuropaBio toxicologists
examined numerous guidance documents.
• Guidance documents on determining a Maximum Tolerated
Dose (MTD) provided definitive endpoints and effect sizes.
12

1. Endpoints and effect
sizes
Endpoint Change relative to control
Body weight Decrease 10%
Liver weight Increase 25%
Kidney weight Increase 25%
Leukocyte count Decrease/increase 30%
Lymphocyte count Decrease/increase 30%
Cholesterol Increase 200%
Blood urea nitrogen Increase 50%
Creatinine Increase 50%
Alkaline phosphatase Increase 200%
(Foster, 2002; EPA, 2002; EPA, 2003; Rhomberg et al., 2007).
13

1. Toxicological endpoints
for power calculations
• The MTD has also been defined as the <strong>dose</strong>
that, “…does not produce mortality, clinical
signs of toxicity, or pathologic lesions…”
(Dorato et al., 2008; Sontag, 1976).
• Hence, it is a <strong>dose</strong> that elicits signs more
subtle than overt toxicity.
• Endpoints and effect sizes are taken from the
indicated guidance documents.
14

1. Toxicological endpoints for
power calculations (continued)
• The toxicological relevance of any differences observed will be
determined on a case-by-case basis.
– Alkaline phosphatase factors: AST levels, ALT levels, liver weight, liver
histopathology, etc.
• All effect sizes proposed have been considered on the scale of
original units.
– This is the scale that makes most sense to toxicologists as it provides
necessary context for evaluation of biological significance.
• Specifying effect sizes on the original units scale removes the
need to consider effect sizes on the standardised scale.
– To be discussed in greater detail later.
15

1. Section
summary/discussion
• EuropaBio Members believe the endpoints
and toxicologically-relevant effect sizes chosen
provide a robust platform on which to base
subsequent power calculations for the study
as a whole.
• All effect sizes proposed have been
considered on the scale of original units; the
scale which makes the most sense to
toxicologists.
16

2. Experimental design
• EFSA guidance advocates use of a randomised complete block
design (RCBD) in which males and females are randomised
together.
• However, current toxicology study animal husbandry practices
intentionally separate males and females (e.g., by assigning
genders to different cage racks).
• There is a concern that housing opposite genders in adjacent
cages, particularly in paired housing setups, may lead to intra-
MSK2 MSK3
cage aggression in males.
• EuropaBio Members’ strong recommendation is that the
current practice of gender separation be maintained.
17

Slide 17
MSK2 Follow up with technical folks.
Animal behavior studies.
Best practices, etc.
Michael Koch; 1/10/2012
MSK3 Will be stated in much stronger language if available information allows. Still being researched.
"Housing genders in adjacent cages is specifically avoided to prevent intra-cage aggression.."
Michael Koch; 9/10/2012

2. Experimental design
(continued)
• EFSA guidance states that use of designs other than
RCBD “may be acceptable provided that appropriate
justification for using them is given”.
• EuropaBio Members agree that, in circumstances
where a good reason for blocking exists, an RCBD is
an appropriate choice.
• However, in the absence of a known blocking
structure, a completely randomised design (CRD) is
the most obvious choice and indeed is likely to be
slightly more powerful than an RCBD.
18

2. Variation controls
• The following variation control measures are
typically utilised to minimise variation amongst the
ExpUs:
– Age-matched animals (single stock, single supplier)
– A small range in initial body weight is specified.
– Housed in a single, environmentally-controlled room.
• Temperature, humidity, photoperiod, and fresh air exchanges
closely monitored.
– In some cases, cage-rack position is periodically rotated
within the animal room to further minimise potential
environmental variability.
19

2. Section
summary/discussion
• EuropaBio Members seek agreement
from EFSA that:
– Maintaining the current practice of
physically separating males and females is
acceptable.
– The absence of obvious blocking factors is
sufficient justification for the use of a CRD.
20

3. Group or single housing
• EFSA’s Scientific Opinion document states that “In accordance with
the European Directive 2010/063, the test animals should be housed
socially”, and recommends that animals of the same sex are housed in
pairs.
• EuropaBio Members understand this requirement for studies
conducted within the EU.
• For studies performed in the U.S., housing animals singly is acceptable
from an animal welfare perspective and is commonplace.
• EuropaBio Members believe group and single housing are equally valid
and should remain options.
– Existing historical control data based on single housing remains relevant.
– It acknowledges different animal welfare requirements exist in testing
locations utilized for whole food toxicity testing.
21

3. Group or single housing
(continued)
• In cases where animals are housed two (or more)
per cage, the guidance makes clear that cage
should be regarded as the ExpU.
• The guidance then explains how it is possible to
test for cage effects and, in cases where the
between-cage variation is not significantly greater
than the between-animal variation, the statistical
analysis can be based on individual animals.
22

3. Group or single housing
(continued)
• While the analysis can be done in this way, the majority of
EuropaBio statisticians believe it is more appropriate to
conduct tests of main effects at the cage level regardless of
whether the cage effect is significant or not (in which case
testing the significance of the cage effect is unnecessary).
• EuropaBio members seek assurance that the statement
“statistical analysis can be based on the individual animals” is
not prescriptive and that an approach in which main effects are
always tested at the cage level is acceptable.
23

3. Section
summary/discussion
• EuropaBio members believe single or pair housing
arrangements are valid for toxicology studies with
whole foods and request that EFSA permit flexibility
in animal housing conditions consistent with local
jurisdiction.
• EuropaBio members seek assurance that an
approach in which main effects are always tested at
the cage level is acceptable (in which case testing
the significance of the cage effect would be
unnecessary).
24

4. Statistical power
investigation
• EuropaBio Members have carried out an extensive
power investigation based on existing <strong>90</strong>-<strong>day</strong> feeding
data.
– Mean response levels and estimates of variation based on 13 studies
for corn-based diets and 10 studies for soy-based diets.
– Some studies based on one animal per cage and others based on two
or more animals per cage.
• The endpoints and effect sizes used in this exercise
were those discussed previously, that are widely
considered by experts in the field to be
toxicologically-relevant.
25

4. Endpoints and effect
sizes
Endpoint Change relative to control
Body weight Decrease 10%
Liver weight Increase 25%
Kidney weight Increase 25%
Leukocyte count Decrease/increase 30%
Lymphocyte count Decrease/increase 30%
Cholesterol Increase 200%
Blood urea nitrogen Increase 50%
Creatinine Increase 50%
Alkaline phosphatase Increase 200%
26

4. Measures of variation
• Given that effect sizes are expressed in percentage terms, the most
appropriate measure of variation is the coefficient of variation (CV).
• Cage was regarded as the ExpU; CV at the cage level is dependent
on the number of rats per cage.
• Evidence of differences in CV between males and females was
detected.
• For each of the endpoints considered, CV’s for a given gender and
number of rats per cage were found to be fairly consistent across
studies.
– Also, no evidence to suggest that CV differed between corn and
soy-based diets.
• So, for each endpoint, an average CV was generated for each
gender / housing policy combination, to be used in subsequent
power calculations.
27

4. Average coefficients of
variation
Endpoint
1 animal/cage
Male Female
2 animals/cage
Male Female
Body weight 9.4% 8.3% 4.9% 4.9%
Liver weight 12.4% 10.3% 7.9% 6.9%
Kidney weight 10.8% 10.5% 6.4% 6.5%
Leukocyte count 24.3% 28.5% 16.6% 20.3%
Lymphocyte count 27.5% 34.7% 18.1% 21.7%
Cholesterol 22.6% 22.2% 12.9% 16.7%
Blood urea nitrogen 14.0% 14.9% 10.0% 9.9%
Creatinine 19.5% 16.9% 7.5% 7.5%
Alkaline phosphatase 21.8% 28.0% 12.8% 20.8%
28

4. Design scenarios
considered
Three design scenarios were considered:
• 1 animal/cage; 3 treatment groups x 12 cages per treatment
per gender = 72 animals in total
– high <strong>dose</strong>, low <strong>dose</strong>, control
• 2 animals/cage; 3 treatment groups x 6 cages per treatment
per gender = 72 animals in total
– high <strong>dose</strong>, low <strong>dose</strong>, control
• 2 animals/cage; 4 treatment groups x 5 cages per treatment
per gender = 80 animals in total
– GM high incorporation rate, non-GM high incorporation rate,
GM low incorporation rate, non-GM low incorporation rate
29

4. Analysis details
• For the 3-treatment group designs, power was based on two
analysis scenarios:
– the comparison of individual treatments,
– the comparison between control and mean of high and low <strong>dose</strong>s.
• For the 4-treatment group design, power was based on the
comparison of treatments (i.e. GM vs non-GM) averaged over
incorporation rates, in line with EFSA guidance.
• In all cases, cage was regarded as the ExpU.
• Males and females were considered separately.
• All results assume a two-sided alternative hypothesis.
30

4. Results
1 animal/cage, 3 treatment groups, 12 replicate
cages/group; individual treatment comparisons
Endpoint Effect Size
Predicted Power
Males Females
Body weight 10% 72% 82%
Liver weight 25% >99% >99%
Kidney weight 25% >99% >99%
Leukocyte count 30% 84% 71%
Lymphocyte count 30% 74% 54%
Cholesterol 200% >99% >99%
Blood urea nitrogen 50% >99% >99%
Creatinine 50% >99% >99%
Alkaline phosphatase 100% >99% >99%
31

4. Results
1 animal/cage, 3 treatment groups, 12 replicate
cages/group; control vs. mean of low and high <strong>dose</strong>
Endpoint Effect Size
Predicted Power
Males Females
Body weight 10% 83% 91%
Liver weight 25% >99% >99%
Kidney weight 25% >99% >99%
Leukocyte count 30% 92% 82%
Lymphocyte count 30% 85% 66%
Cholesterol 200% >99% >99%
Blood urea nitrogen 50% >99% >99%
Creatinine 50% >99% >99%
Alkaline phosphatase 100% >99% >99%
32

4. Results
2 animals/cage, 3 treatment groups, 6 replicate
cages/group; individual treatment comparisons
Endpoint Effect Size
Predicted Power
Males Females
Body weight 10% 91% 91%
Liver weight 25% >99% >99%
Kidney weight 25% >99% >99%
Leukocyte count 30% 83% 67%
Lymphocyte count 30% 77% 61%
Cholesterol 200% >99% >99%
Blood urea nitrogen 50% >99% >99%
Creatinine 50% >99% >99%
Alkaline phosphatase 100% >99% >99%
33

4. Results
2 animals/cage, 3 treatment groups, 6 replicate
cages/group; control vs. mean of low and high <strong>dose</strong>
Endpoint Effect Size
Predicted Power
Males Females
Body weight 10% 97% 97%
Liver weight 25% >99% >99%
Kidney weight 25% >99% >99%
Leukocyte count 30% 92% 79%
Lymphocyte count 30% 87% 73%
Cholesterol 200% >99% >99%
Blood urea nitrogen 50% >99% >99%
Creatinine 50% >99% >99%
Alkaline phosphatase 100% >99% >99%
34

4. Results
2 animals/cage, 4 treatment groups, 5 replicate cages/group;
GM vs. non-GM averaged over incorporation rates
Endpoint Effect Size
Predicted Power
Males Females
Body weight 10% 99% 99%
Liver weight 25% >99% >99%
Kidney weight 25% >99% >99%
Leukocyte count 30% 97% 87%
Lymphocyte count 30% 94% 83%
Cholesterol 200% >99% >99%
Blood urea nitrogen 50% >99% >99%
Creatinine 50% >99% >99%
Alkaline phosphatase 100% >99% >99%
35

4. Provisos for power
analyses
• Power investigation for 3-treatment group scenario
comparing control with mean of high and low <strong>dose</strong>s
assumes that, in reality, mean of low <strong>dose</strong> = mean
of high <strong>dose</strong>.
• Power investigation for the 4-treatment group
design assumes that there is no interaction between
treatment (i.e. GM vs. non-GM) and incorporation
rate.
– Analysis of previous data indicates that this
assumption should typically hold in practice.
36

4. Should males and females
be analysed together?
• Physically separating males and females does not
necessarily preclude analysing data from both genders
together but there are good reasons for doing so:
– Analysis across genders for some variables is problematic due
to well-established gender-specific differences in variance,
which experience suggests cannot be readily overcome by the
use of a data transformation.
– Large differences in gender means (e.g., for body and organ
weights) may make results of a combined analysis more
complicated to interpret (Gad, 2006).
– Analysis across genders would be more complex and could be
difficult for some CRO’s to implement correctly.
37

4. Should males and females
be analysed together?
• Reasons for analysing genders separately (continued):
– Analysis by gender is the standard approach in toxicology.
– Certain endpoints (e.g., testes or uterus weights) are limited to
one gender and thus necessitate separate analyses by default.
– Results from the power investigation show that powers based
on gender-specific analyses are generally more than adequate
for the designated purpose.
• EuropaBio Members therefore recommend analysing
genders separately.
38

4. Section
summary/discussion
• The power profile as a whole is very positive for each of the
three proposed design scenarios.
– Each scenario strikes a reasonable balance between
power and ethical considerations.
• Findings demonstrate that the power inherent in previous
studies of this size (i.e., 10-12 animals per treatment group
per gender, in line with OECD TG 408) should in general be
similarly adequate.
• While power is important, the toxicological relevance of all
observed differences are discussed in any case regardless of
statistical significance.
39

Part 2: Study analysis and
interpretation
40

Challenges in study analysis
and interpretation
1. Toxicological interpretation
2. Standardised effect sizes (SES)
3. Statistical Analysis Plan & summary
statistics and raw analysis outputs
41

1. Toxicological
interpretation
• EuropaBio Members appreciate that EFSA has commented on the
imperfect relationship between statistical significance and
toxicological relevance in the final version of the guidance.
• However, without feedback from EFSA regarding the flexibility that
will be permitted in overall design, some concerns about the
prominence of the statistical analysis persist.
• EFSA (2011b) recently stated, “Many researchers incorrectly conclude
that any statistically significant effect is biologically relevant...”
– This gives EuropaBio Members confidence that EFSA considers statistical analysis
one tool of many to be used to identify toxicologically-relevant differences.
• The scientific training of toxicologists, and their experience in
interpreting data, must ultimately determine the relationship
between the finding and treatment with the test substance.
42

2. Presentation of results on
the standardised scale
• Given that all endpoints are to be analysed and presented on
the scale of original units, EuropaBio Members question the
need to also present results on the standardised scale.
– “If the original units of measurement are meaningful, the presentation
of unstandardised effect statistics is preferable over those of
standardised effect statistics”. (Nakagawa and Cuthill, 2007)
• Original units are meaningful to trained toxicologists.
– “If researchers are familiar with their study systems, or abundant
previous research on a topic of interest exists, effect sizes in original
units are more readily interpretable than standardised effect statistics”.
(Nakagawa and Cuthill, 2007)
• The study system is familiar to trained toxicologists and the basis of
decades of previous research.
43

2. Presentation of results on
the standardised scale
• When results are presented on the standardised
scale all studies of the same design will appear to
be equally sensitive regardless of any differences in
underlying variability.
• The guidance states that “Changes in endpoints
…must also be evaluated with respect to normal
biological variation of the endpoints”; this can be
done on the scale of original units but not on the
standardised scale.
44

Original Scale Standardised Scale
Lab 1 Lab 2 Lab 1 Lab 2
Mean Control 9.77 9.77 -1.45 -0.59
Mean GM low 12.44 12.44 0.84 0.34
Mean GM high 12.17 12.17 0.61 0.25
SD 1.17 2.85 1.00 1.00
CV 10% 25% NA NA
p-value GM high vs Control 0.003 0.165 0.003 0.165
GM high - Control 2.40 2.40 2.06 0.84
95% CI
GM high - Control
0.97 - 3.83 -1.11 - 5.92 0.83 - 3.29 -0.39 - 2.07
Width of 95% CI 2.86 7.03 2.46 2.46
Comparison with historical
ranges possible?
2. Example of original vs.
standardised scaling
yes yes no no
45

2. Section
summary/discussion
• EuropaBio Members see the presentation of
results on the standardised scale as
providing no additional benefit, and
respectfully request that this aspect of the
guidance be relaxed from mandatory to
optional.
46

3. Statistical Analysis
Plan
• From the Scientific Opinion document it is unclear
whether it is mandatory to provide a separate SAP
in addition to the details concerning the proposed
statistical analysis that are included in the study
protocol.
• EuropaBio Members wish to avoid unnecessary
duplication, and believe that the level of detail that
will be provided in the study protocol should
suffice.
47

3. Summary statistics and
raw analysis outputs
• EuropaBio Members’ understanding is that EFSA wishes to see
results presented both in terms of means and p-values, and
differences and 95% confidence intervals.
• However, some requirements are less clear, and others (e.g., the
median) seem to be of questionable value in most cases.
• It is the opinion of EuropaBio Members that the following
summary statistics would generally be appropriate:
– Mean, min, max and SD for each treatment group.
– P-values for all fixed effects and specific contrasts.
– Point estimate of the difference between GM and non-GM together with
95% confidence interval for the difference.
• EuropaBio Members seek confirmation from EFSA as to whether
the above list is acceptable.
48

3. Summary statistics and
raw analysis outputs
• With regard to the provision of raw data in electronic
format, the guidance document is inconsistent
regarding whether this is a default requirement or “on
request”.
– Please clarify.
• The guidance document states that “The statistical
analysis programs, logs and outputs should be
provided for the purposes of review”.
– To provide all this information by default seems excessive.
– Please clarify whether this is a default requirement or “on
request”.
49

3. Section
summary/discussion
• Please clarify if a separate Statistical Analysis Plan is
mandatory or if the level of detail that can be
presented in the study protocol is adequate.
• Does EFSA regard the summary statistics proposed by
EuropaBio Members as suitable?
• Please clarify whether raw data in electronic format is
a default requirement or “on request”.
• Please clarify whether the statistical analysis
programs, logs and outputs are a default requirement
or “on request”.
50

References
• Codex Alimentarius. 2009. Foods derived from modern biotechnology. Codex
Alimentarius Commission, Joint FAO/WHO Food Standards Programme, Food and
Agriculture Organization of the United Nations: Rome, Italy.
• Dorato, M.A., McMillian, C.L., Vodicnik, M.J. “The Toxicologic Assessment of
Pharmaceutical and Biotechnology Products”. Principles and Methods of
Toxicology. Ed. A.W. Hayes. New York: Informa Healthcare USA, Inc., 2008. pp 325-
368.
• EFSA. 2008. Safety and nutritional assessment of GM plants and derived food and
feed: The role of animal feeding trials. Food and Chemical Toxicology 46:S2-S70.
• EFSA. 2011a. Guidance for risk assessment of food and feed from genetically
modified plants. EFSA Journal 9:2150.
• EFSA. 2011b. Scientific opinion: Statistical significance and biological relevance.
EFSA Journal 9:2372.
51

References
• EFSA. 2011c. Guidance of EFSA. Use of the EFSA Comprehensive European Food
Consumption Database in Exposure Assessment. EFSA Journal 2011;9:2097.
Supplemental information:
http://www.efsa.europa.eu/en/datexfooddb/datexfooddbspecificdata.htm.
• EPA. 2002. Health effects Division (HED) Guidance Document. # G0201.
Hepatocellular Hypertrophy. October 21, 2002.
• EPA. 2003. Health effects Division (HED) Interim Guidance Document. # G2003.02.
Rodent Carcinogenicity <strong>Studies</strong>: Dose selection and evaluation. July 1, 2003.
• Foster, P.M.D. 2002. EPA. 23 July 2002.
• Gad, S. 2006. Statistics and Experimental Design for Toxicologists and
Pharmacologists. Boca Raton: Taylor & Francis Group.
• Nakagawa, S., and I.C. Cuthill. 2007. Effect size, confidence interval and statistical
significance: a practical guide for biologists. Biol Rev Camb Philos Soc 82:591-605.
52

References
• OECD. 1998. Guideline for the Testing of Chemicals – <strong>Repeated</strong> Dose <strong>90</strong>-<strong>day</strong> <strong>Oral</strong>
<strong>Toxicity</strong> Study in Rodents, 408.
• OECD. 2012. Guidance Document 116 on the Conduct and Design of Chronic
<strong>Toxicity</strong> and Carcinogenicity <strong>Studies</strong>, Supporting Test Guidelines 451, 452 and453
2nd Edition.
• Rhomberg, L.R., K. Baetcke, J. Blancato, J. Bus, S. Cohen, R. Conolly, R. Dixit, J.
Doe, K. Ekelman, P. Fenner-Crisp, P. Harvey, D. Hattis, A. Jacobs, D. Jacobson-Kram,
T. Lewandowski, R. Liteplo, O. Pelkonen, J. Rice, D. Somers, A. Turturro, W. West,
and S. Olin. 2007. Issues in the design and interpretation of chronic toxicity and
carcinogenicity studies in rodents: approaches to <strong>dose</strong> selection. Crit Rev Toxicol
37:729-837.
• Sontag, J.M., Page, N.P., and Safiotti, V. 1976. Guidelines for Carcinogen Bioassays
in Small Rodents, DHHS Publ. (NIH) 76-801. National Cancer Institute, Bethesda,
MD.
53

References
• WHO. 2003. GEMS/FOOD Regional Diets: Regional per Capita Consumption Of
Raw and Semi-processed Agricultural Commodities. Revision September 2003.
54

Contributing EuropaBio
• Bayer CropScience
• BASF
• Dow AgroSciences
• DuPont Pioneer
• Monsanto
• Syngenta
Members
55