Toxicological Risk Assessment of Chemicals: A Practical Guide - Chapter 5 potx

5
Standard Setting:
Threshold Effects
Adve rse h ealth effect s can be consi dered to be of two types (see Secti on 4.2): those consi dered to
have a threshold, k nown as ‘‘thres hold effects ’’ (effec ts such as, e.g., organ-speci fic, neurologica l,
immunol ogical, non-gen otoxic carci nogenicity , reprod ucti ve, develo pment al), and those for whi ch
there is consi dered to be some risk at any exposur e level, know n as ‘‘non-t hreshold effect s’’ (effects
such as, e.g., mut agenicity, genotoxici ty, genoto xic carcinogeni city). Tho ugh it is not possi ble to
demon strate experiment ally the presen ce or absence of a thres hold, diffe rences in the approac h
to the hazard asses smen t of threshold versus non-thresho ld effects have been adopte d widely. The
distinct ion in approac he s is based primari ly on the prem ise that simple events such as in vitro
activati on an d covale nt bindi ng may be line ar over many orders of magnitud e, i.e., that these events
occur even at very low exposur e levels. How ever, a simple pragm atic dist inction on this basis is
incre asingly probl ematic as it is likely that there is a thres hold for a numbe r of genoto xic effect s; this
is addres sed in detai l in Cha pter 6.
In the hazard asses sment proces s, descri bed in detail in Cha pter 4, all effect s observ ed are
evalua ted in terms of the type and severi ty (adver se or non-adv erse) , their dose –response relation-
ship, and the relevanc e for human s of the effects observ ed in experi mental animals. For threshold
effects, a No- or a Lowes t-Obser ved-Advers e-Effect Lev el (N=LOAE L), or alternati vely a Ben ch-
mark Dos e (BMD), is deriv ed for every single effect in all the avail able studi es provi ded that data
are suffi cient for such an evalua tion. In the last step of the hazard asses sment for thres hold effects,
all this informat ion is asses sed in tota l in order to ident ify the critical effect(s) and to de rive a
NOA EL, or LOAE L, for the critical effect(s) .
The approac h of deriv ing a tole rable inta ke by divi ding the N=LOA EL, or alternati vely a BMD
for the critical effect(s) by an asses smen t facto r has been descri bed and discu ssed extens ively in the
scien tific literat ure. It is beyond the scope of this book to revie w all these refere nces. This chapte r
presen ts an overvi ew of published extrapolat ion methods for the deriva tion of a tolerable inta ke
based on the assessmen t factor approac h, i.e., limited to addres s effects with thres hold charact er-
istics, and is not meant to be exhaust ive. The mai n focus is on the rationale for and the use of the
asses sment facto rs. Pe rtinent guidan ce docum ents and revie ws for the issues add ressed in
this chapte r include WHO=IPCS (1994, 1996, 1999), US-EPA (2002, 2004), IGHRC (2003) ,

ECETO C (2003) , KEMI (2003) , Kal berlah and Schneid er (1998) , Vermei re et al. (1999) , and
Nielse n et al. (2005) .
The approac h of standard setting for non-t hresho ld effect s is addres sed in Cha pter 6.
The development of regulatory standards derived from a standard such as, e.g., the Tolerable
Daily Intake or a Ref erence Dos e, is addres sed in Chapter 9.
5.1 INTRODUCTION
Acco rding to the OECD=IPC S de finitions listed in Annexure 1 of Cha pter 1 (OECD 2003):
Threshold is ‘‘Dose or exposure concentration of a substance below that a stated effect is not
observed or expected to occur.’’
ß 2007 by Taylor & Francis Group, LLC.
Tolerabl e Intak e is ‘‘Estimate d maxi mum amoun t of an agent, expres sed on a body mass basis,
to which each indi vidual in a (sub) p opulation may be exp osed over a speci fied perio d wi thout
appreci able risk .’’
A tolerable intake may have different units dependi ng on the route of adminis tration upon which
it is based, an d is general ly expres sed on a dail y or weekly basis . For the oral and derm al route s, a
tole rable intake is general ly expres sed on a body weight basis , e.g., mg=kg body wei ght per day.
Tho ugh not strictly an ‘‘ intake, ’’ tolerable intakes for inhalati on are generally expres sed as an
airborne concentration, e.g., mg=m
3
.
Acco rding to the OEC D=IPCS de finitions listed in Anne xure 1 of Chapter 1 (OECD 2003):
Acceptable=Tolerable Daily Intake is ‘‘Estimated maximum amount of an agent, expressed on a
body mass basis, to which an individual in a (sub) population may be exposed daily over its lifetime
without appreciable health risk.’’
Reference Dose is ‘‘An estimate of the daily exposure dose that is likely to be without
deleterious effect even if continued exposure occurs over a lifetime.’’
Related terms: Acceptable=Tolerable Daily Intake.
The term ‘‘acceptable’’ is used widely to describe ‘‘safe’’ levels of intake and is applied
for chemicals to be used in food production such as, e.g., food additives, pesticides, and veterinary
drugs. The term ‘‘tolerable’’ is applied for chemicals unavoidably present in a media such as contamin-

ants in, e.g., drinking water and food. The term ‘‘PTWI’’ (Provisional Tolerable Weekly Intake) is
generally used for contaminants that may accumulate in the body, and the weekly designation is used to
stress the importance of limiting intake over a period of time for such substances. The tolerable intake is
similar in definition and intent to terms such as ‘‘Reference Dose’’ and ‘‘Reference Concentration’’
(RfD=RfC), which are widely used by, e.g., the US-EPA. For some substances, notably pesticides, the
‘‘ARfD’’ (Acute Reference Dose), is also established, often from shorter-term studies than those that
would support the ADI. The ARfD is defined as the amount of a substance in food that can be consumed
in the course of a day or at a single meal with no adverse effects.
In inhalation studies, laboratory animals are generally exposed to an airborne chemical for a
limited period of time, e.g., 6 h a day, 5 days per week. Adjustment of such an intermittent exposure
to a continuous exposure scenario is regularly applied as a default procedure to inhalation studies
with repeated exposures but not to single-exposure inhalation toxicity studies. Operationally, this is
accomplished by a correction for both the number of hours in a daily exposure period and the
number of days per week that the exposures were performed. In an inhalation study in which
animals were exposed to an airborne concent ration of a substance at 5 mg=m
3
for 6 h a day, for
5 days per week, the adjustment of this intermittent exposure concentration to a continuous exposure
concentration would consider both hours per day and days per week : 5 mg=m
3
3 6=24 h 3 5=7
days=week ¼ 0.9 mg=m
3
, with 0.9 mg=m
3
being the concentration adjusted to continuous exposure.
For systemi c effects observed in inhalation studies, the determining factor for effects to occur at
the systemic target is generally the total dose rather than the concentration of the chemical in the air.
In such cases, a tolerable intake (expressed as mg=kg body weight per day, or mg=m
3

depending on
the standard to be derived, i.e., a tolerable intake in its strict meani ng, or a tolerable concentration) is
established from the NOAEC, or LOAEC, derived in the inhalation study and adjusted for
continuous exposure.
For local effects, in contrast, the deter mining factor for effects to occur at the site of first contact
(mucous membrane of the respiratory tract, the eyes, or the skin) is generally the concentration of
the chemical in the air rather than the total dose at the site of first contact. In such cases, a tolerable
concentration (expressed as mg=m
3
) is established from the NOAEC, or LOAEC, derived in the
inhalation study without an adjustment to a continuous exposure.
The overall principles for the derivation of a tolerable intake are equal irrespective of chemical
class (e.g., food additives, pesticides, veterinary drugs, contaminants) although it should be recog-
nized that the available database for chemicals deliberately added to, e.g., food is generally more
ß 2007 by Taylor & Francis Group, LLC.
compr ehensive than for contam inants. This is because there a re extens ive regul atory demands for
toxicit y data in relat ion to marketin g of che micals, which are intenti onally a pplied to food, etc. For
threshold effect s, a tole rable intake is generally deriv ed from the NOA EL, or LOA EL, for the
critical effect (s) by dividing the NOA EL=LOA EL, by an overall assessmen t facto r.
Acco rding to the OECD=IPC S de finitions listed in Annex ure 1 of Chapter 1 (OECD 20 03):
Assessm ent Fa ctor is ‘‘ Numeri cal adjus tment used to extrap olate from experi mentall y deter -
mined (dose –respon se) relationsh ips to e stimate the agent exposur e below whic h an advers e effect is
not like ly to occur. ’’
Relate d term s: Safety Factor , Unce rtainty Factor, Extrapo lation Factor , Adjust ment Factor ,
Conversi on Factor.
The re is an enormous variability in the extent and natur e of different databa ses for chemical
substanc es. Fo r examp le, in some cases, the evalua tion of a chemi cal must b e based on limit ed
data in experi mental animals, whereas in other cases detai led infor mation on the vario us end-
points, toxicokin etics, an d mode( s) of action may be available. In some cases, the evalua tion can be
based on data on effect s in exposed human populatio ns. Clearl y as the amount of infor mation

available incre ases, the degree of und erstanding of the hazards expres sed also incre ases, and the
uncert ainties due to lack of informat ion decreas e. However , even with complex databa ses, uncer-
tainties sti ll remain.
The asses sment facto rs general ly applied in the estab lishmen t of a tolerable inta ke from the
NOA EL, or LOA EL, for the crit ical effect (s) are appli ed in order to compe nsate for uncert ainties
inher ent to extrapolat ion of experiment al animals data to a g iven human situation, and for uncer-
tainties in the toxicolog ical databa se, i.e., in cases wher e the substanc e-speci fic knowledge requi red
for risk asses smen t is not avail able. As a co nsequen ce of the variabili ty in the extent and natur e of
different databa ses for chemi cal substances, the range of assessmen t facto rs applied in the estab -
lishmen t of a tole rable intake has been wide (1 –10,000), althoug h a value of 100 has been used most
often . An overview of diff erent approac hes in using assessmen t factors, historic ally and currently,
is provi ded in Secti on 5.2.
The key areas of uncert ainty when using data from experi mental anim als include uncertaint y
related to:
.
Extrapo lation from anim al speci es to human s (Secti on 5.3)
.
Variabi lity in the human popula tion (Secti on 5.4)
.
Route-t o-route extra polation (S ection 5.5)
.
Durati on of exposur e in experi mental studies (Section 5.6)
.
Dose –respon se curve=NOA EL n ot estab lished (Section 5.7)
.
Nature and severity of the effects (Section 5.8)
.
Gaps or other de ficienci es in the databa se (Section 5.9)
5.2 ASSESSMENT FACTORS: GENERAL ASPECTS
In the context of assessment factors, it is important to distinguish between the two terms ‘‘variabil-

ity’’ and ‘‘uncertainty.’’ Variability refers to observed diff erences attributable to true heterogeneity
or diversity, i.e., inherent biological differences between species, strains, and individuals. Variability
is the result of natural random processes and is usually not reducible by further measurement or
study although it can be better characterized. Uncertainty relates to lack of knowledge about, e.g.,
models, parameters, constants, data, etc., and can sometimes be minimized, reduced, or eliminated
if additional information is obtained (US-EPA 2003).
It should be recognized that a lack of knowledge of variability is a source of uncertainty.
The terminology within this area is not standardized. Other terms include ‘‘safety factor,’’
‘‘uncertainty factor,’’ ‘‘extrapolation factor,’’ ‘‘adjustment factor,’’ and ‘‘conversion factor.’’ None
ß 2007 by Taylor & Francis Group, LLC.
of these terms are ideal . For examp le, the term safety facto r has implic ations of absol ute safety,
wher eas the term uncert ainty facto r, alth ough being broader, may be interpret ed different ly in
relation to varia bility a nd uncert ainty. For the sake of clarity in this book, the term asses sment
facto r is used and is meant as a g eneral term to cover all facto rs desig nated in the literatu re as safety
facto r, unce rtainty facto r, extrapolat ion facto r, adjus tment facto r, convers ion facto r, etc. The other
ment ioned term s are not used unless refere nce is made to a speci fic term or met hod. The asses sment
facto r can cover both varia bility and uncert ainty.
The follow ing secti on gives an overview of different approac hes in using assessmen t factors,
histori cally and curren tly, beginn ing with the introduct ion of the so-cal led ‘‘safet y factor approac h’’
in the mid-1950s and re flecting the develo pment up to the regul atory approac hes c urrently used by
inte rnational and federal b odies. The overview does not attempt to cover all publi cations in this
field, but includes the approac hes sugges ted by diff erent scien tifi c groups and inte rnational and
federal bodies , which are considered as being the most central ones in the development of the
approaches currently used regulatory. Default assessment factors used or suggested in the various
approac hes are summ arized in Table 5 .1.
5.2.1 ASSESSMENT FACTORS:VARIOUS APPROACHES
Historically, the so-called safety factor approach was introduced in the United States in the mid-
1950s in response to the legislative needs in the area of the safety of chemical food additives
(Lehman and Fitzhugh 1954). This approach proposed that a ‘‘safe level’’ of chemical food
additives could be deriv ed from a chronic NOAEL from animal studies divided by a 100-fold

safety factor. The 100-fold safety factor as proposed by Lehman and Fitzhugh was based on a
limited analysis of subchronic=chronic data on fluorine and arsenic in rats, dogs, and humans, and
also on the assumption that the human population as a whole is heterogeneous. Initially, Lehman
and Fitzhugh reasoned that the safety factor of 100 accounted for several areas of uncertainty:
.
Intraspecies (human-to-human) variability
.
Interspecies (animal-to-human) variability
.
Allowance for sensitive human populations due to illness when compared with healthy
experimental animals
.
Possible synergistic action of the many intentional and unintentional food additives or
contaminants
In 1961, the Joint FAO=WHO Expert Committee on Food Additives (JECFA) and the Joint Meeting
of Experts on Pesticides Residues (JMPR) adopted this approach in a slightly modified form: The
safe level was called the Acceptable Daily Intake (ADI) and expres sed in mg=kg body weight per
day (Vermeire et al. 1999, ECETOC 2003). Usually, a safety facto r of 100 is used by JECFA and
JMPR for establishing ADIs by this ADI approach; however, the procedures adopted by JECFA and
JMPR do not generate a clear justification for deviation from the factor of 100, but in some
individual cases, an expert explanation is given for the use of factors other than 100 (Vermeire
et al. 1999).
It is apparent that the factor of 100 has no quantitative bases, and the choice of the value 100 is
more or less arbitrary (Vermeire et al. 1999). Retrospectively, some attempts have been made to
support a 100-fold factor (Bigwood 1973, Lu 1979, Vettorazzi 1977 as reviewed in Vermeire et al.
1999 and KEMI 2003), and the 100-fold factor was found to be justified.
The 100-fold safety factor has tradi tionally been interpreted as the product of two factors with
default values of 10. For example, according to WHO=IPCS (1987), the safety factor is intended to
provide an adequate Margin of Safety (MOS) by assuming that the human being is 10 times more
sensitive than the test animal and that the difference of sensitivity within the human population is in

a 10-fold range.
ß 2007 by Taylor & Francis Group, LLC.
TABLE 5.1
Default Assessment Factors Used or Suggested for the Establishment of a Regulator y Standard or Health-Base d Guidance Value
for Threshold Effects
Factor
JECFA
JMPR
US-EPA
(2002)
Renwick
(1993) WHO
LLN
(1990)
ECETOC
(2003)
TNO
(1996)
Kalberlah and
Schneider (1998)
KEMI
(2003)
a
D-EPA
(2006)
Interspecies 10 10 10 10
Toxicokinetic A
b
44 A
b

4 (rat)
Toxicodynamic 10
0.5
2.5 2.5 1 2.5
Oral A
b
3 3A
b
3 2– 3
Inhalation 3
Interindividual 10 10 10 5 10 25 10
Toxicokinetic 4 3.16 8 3–5
Toxicodynamic 2.5 3.16 3 3.16
Occupational 33
Route-to-route
Duration exposure 10
Subacute-to-subchronic 10 3
Subacute-to-chronic 624
Subchronic-to-chronic 10 2 10 8 10
LOAEL-to-NOAEL 10 10 3 1 3–10 10
Nature and severity 11–10 up to 10
Confidence database 10 10 1 1–10 1–10
Nonscientific1
a
See also Table 5.2.
b
A is a calculated adjustment factor allowing for the differences in caloric requirement.
ß 2007 by Taylor & Francis Group, LLC.
5.2.1. 1 US-E PA Appr oach
In 1988, the US-EPA adopte d the ADI approac h in its regul atory meas ures against envir onmen tal

poll ution; with a numbe r of modi fi cations (US-E PA 1988, 1993). Inst ead of the terms ADI and
safet y facto r, the term s Reference Dose (RfD) and uncert ainty facto r (UF), respec tively, wer e
selec ted. The RfD is deriv ed from the NOA EL by divi ding by the overall UF. The o verall UF
origina lly sugges ted and recon firmed in 2002 (US-E PA 2002) general ly consi sts of a 10-fol d factor
for each of the following:
.
Huma n varia tion in sensi tivity (UF
H
)
.
Inter species extra polation (UF
A
)
.
Use of the NO AEL obtained from a less than lif etime study (UF
S
)
.
Use of a LOAE L in the absence of a NOAEL (UF
L
)
.
Adeq uacy of the total database (UF
D
)
Acco rding to US-EPA (1993), the fi rst four of the above-ment ioned factors a re adapte d from
Dour son and Stara (1983).
The exact value of the UFs chosen shoul d depend on the quali ty of the studie s av ailable, the extent
of the database, and scien tific judgment (US- EPA 2002 ). The default facto rs typic ally used cover a
single order of magni tude (i.e., 10

1
). By convent ion, in the US-EPA , a value of 3 is used in place of
one-hal f powe r (i.e., 10
0.5
) when appropriat e. The se half- power values should be facto red as whole
numbe rs when they occur singly but as powers or logs when they occur in tandem . A composit e UF
of 3 and 10 would thus be expres sed as 30 (3 3 10
1
), whereas a compo site UF of 3 and 3 would
be expressed as 10 (10
0.5
3 10
0.5
¼ 10
1
). It shoul d be n oted, in addition, that rigid applicati on of log
(i.e., 10
1
) or ½ log (i.e., 10
0.5
) un its for UFs could lead to an illogic al set of refere nce values ; therefore,
it ha s been empha sized that appli cation of scien tific judgm ent is critical to the overall proces s.
It is also noted that there is overl ap in the individua l UFs and that the appli cation of five UFs of
ten for the chronic reference value (yielding a total UF of 100,000) is inappr opria te. In fact, in cases
wher e maxi mum uncertaint y ex ists in a ll fi ve areas, it is unlikely that the databa se is suf ficient to
deriv e a refere nce value. Uncertainty in four areas may also indi cate that the database is insuf fi cient
to deriv e a reference value. In the c ase of the RfC, the maxi mum UF would be 3,000, whereas the
maxi mum would be 10,000 for the RfD. Thi s is because the deriv ation of RfCs and RfDs has
evolve d some what different ly. The RfC methodol ogy (US- EPA 1994) recommend s divi ding the
inte rspecies UF in half, one-half (10

0.5
) each for toxicokin etic and toxi codynam ic consi derations,
and it includes a Dosimetric Adjustment Factor (DAF, represents a multiplicative factor used to adjust
an observed exposure concentration in a particula r laboratory species to an exposure concentration
for humans that would be associated with the same delivered dose) to account for toxicokinetic
differences in calculating the Human Equivalent Concentration (HEC), thus reducing the interspecies
UF to 3 for toxicodynamic issues. RfDs, however, do not incorporate a DAF for deriving a Human
Equ ivalent Dos e (HED), and the interspeci es UF of 10 is typi cally applied, see also Secti on 5.3 .4. It is
recommended to limit the total UF applied for any particular chemical to no more than 3000, for both
RfDs and RfCs, and avoiding the derivation of a reference value that involves application of the full
10-fold UF in four or more areas of extrapolation.
In addition, a modifying factor (MF) could be applied (US-EPA 1993). The MF is in reality an
additional UF that is greater than 0 and less than or equal to 10; the default value is 1. The MF should
account for uncertainties of the study and database not explicitly handled by the use of the general UFs;
e.g., the completeness of the overall database and the number of species tested. In the 2002 review of the
RfD and RfC processes (US-EPA 2002), it was recommended that use of the MF be discontinued as it
was considered that the uncertainties accounted for by the MF is sufficiently accounted for by the
general UF.
The US-EPA staff paper from 2004 titled ‘‘An Examination of EPA Risk Assessment Principles
and Practices’’ (US-EPA 2004) provides comprehensive and detailed information on the
ß 2007 by Taylor & Francis Group, LLC.
pract ices empl oyed in risk asses smen t, incl uding use of UFs and use of default and extra polation
assumpti ons.
5.2.1. 2 Calabres e and Gilbert Appro ach
Calabres e and Gilber t (1993) hav e demonstra ted the lack of indepe ndence of the interspeci es and
intraspeci es UFs, as well as of the intraspeci es and the less -than-lifeti me UFs. Based on thei r
analys es, the author s conclu ded that most of the recommend ed US-EPA stand ards based on animal
models needed to have some of their UFs modi fied. They recom mended the following modi fi cations
of the intraspeci es UF, see also Secti on 5.4 .2:
.

Signi ficantly less-than -lifeti me anim al study : 5
.
When the anim al study is for a norm al experi mental lifetime (2 years in rodent s): 4
.
Occu pational epidemiolo gical study: 10
.
Environ menta l epidemiol ogical study , if study was for a no rmal human life span: 5
5.2.1. 3 Renwic k Appro ach
The approac h propos ed by Renwick (1991, 1993) is also b ased on the 100-fo ld facto r. It attempt s to
give a scien tifi c basis to the default values of 10 for the interspeci es and 10 for the intraspeci es
(interin dividual human ) diff erences. Ren wick also propos ed a divi sion of each of these UFs into
sub-f actors to allo w for separa te evalua tions of diff erences in toxicokin etics and toxicodyna mics.
The advanta ge of such a subdi vision is that compo nents of these UFs can be addres sed wher e data
are avail able; for example, if ava ilable data show simil ar toxicokin etic s of a given chemical in
experi mental anim als and humans, then only an inte rspecies e xtrapolat ion facto r would be nee ded to
account for diff erences in toxi codynam ics. Renwi ck examined the relative magnitud e of toxicoki-
netic and toxicodynam ic varia tions between and wi thin species in detail. He found that toxi cokinetic
differences were generally greater than toxicodyn amic d ifferences resul ting in the propos al that the
10-fold facto rs (for inter- and intraspeci es variation ) shoul d, by defaul t, be subdi vided into factors of
4 for toxicokin etics and 2.5 for toxi codynam ics. It should be no ted that the propos ed defaul t values
were deriv ed from limited data.
The WHO=IPCS (1994) ha s adopte d the approac h set forth by Ren wick (1993) with one
deviation, see Figure 5.1. Whil e the UF for inte rspecies (animal- to-human) extra polation shoul d be
split into defaul t values of 4 for toxi cokinetics and 2.5 for toxicodynam ics, the UF for intraspeci es
(hum an-to-hum an) extra polation shoul d be split evenly be tween both aspect s, i.e., a sub-f actor of 3.16
for both toxicokin etics and toxicodynam ics. The reason for this deviation from Renwi ck ’s initial
sugges tion was that the WHO=IPCS consi dered that the slightly greater varia bility in the kinetics in
human s compa red with dy namics was not suf ficient to warrant an unequal subdivisi on of the 10-fol d
facto r into a toxic okinetic facto r of 4 and a toxicodynam ic facto r of 2.5. Actual data shoul d be used to
repla ce the defaul t values if availab le. It was furt hermore noted that precise defaul t values for kinetics

and dy namics cannot be expect ed on the basis of a subdivisio n of the imp recise 10-fold compo site
facto r for interspeci es as wel l as for the inte rindivid ual varia tion. Acco rding to WHO=IPCS, the
defaul t v alues sugges ted above wer e consi dered as being reason able since they provide a positive
value greater than 2 for both aspects and are compatible with the species differences in physiological
parameters such as renal and hepatic blood flow. It was also noted that since the database examined
was limited, the default values suggested for subdivision of interspecies and interindividual variation
should be adopted on an interim basis.
5.2.1. 4 Lewis –Lyn ch–Nikiforov Appr oach
In 1990, Lewis et al. published a new approach introducing flexibility such that both new
information and expert judgment could be readily incorporated. The Lewis–Lynch–Nikiforov
(LLN) method, and its refinements, are extensions of established principles and procedures, and
ß 2007 by Taylor & Francis Group, LLC.
guides the data evalua tor to adjus t experi mentall y determin ed ‘‘no-eff ect ’’ (or ‘‘ minim um effect ’’)
level s from expe rimental animal studies taking the foll owing aspects into account :
.
Known diff erences between labor atory anim als and human s and between experi mental
condit ions and the real world
.
Sensit ivity of the exposed human popula tions
.
Streng th of eviden ce that the chemi cal presen ts a real haz ard to human healt h
.
Gene ral quali ty of the experi mental databa se
.
Unce rtaintie s in extrapolat ing from labor atory anim als to h umans
.
Potency of the toxic agent
.
Typ e and severi ty of the putat ive advers e effect
Acco rding to Lewis et al. (1990) , a step- by-step sequenc e is used. Initially, a quali tative deter min-

atio n is made as to the strength o f eviden ce that the putative toxi c agent presen ts an a ctual health
hazard to human s, i.e., how like ly is this agent to p roduce the suspec ted adverse effect in human s?
In contrast to the ADI and RfD method wher e no speci fic c onsideration is given to judgi ng the
like lihood that a chemical presents a real health hazard, the ‘‘stre ngth of the quali tative evidence ’’ is
scored expli citly and separa tely in the LLN approac h. The NA EL
human
is estimat ed from laboratory
resear ch results, using the follow ing a lgorithm:
NAE L
human
¼
NOA EL
animal
[S]
[I][R][ Q
1
][Q
2
][Q
3
][U][ C]
[S] is the aggrega te ‘‘scali ng facto r ’’ to account for known quanti tative diff erences betw een species
and between labor atory experi mental condition s and the real world. The default value is 1,
indi cating that animals and humans are equivalent in these dim ension s.
[I] is the adjus tme nt facto r to account for anticipat ed g reater suscep tibility among mem bers of
the test animal populatio n than was ob served in the experi ment, i.e., to account for intraspeci es
varia bility. The default value is 10, indi cating that extre mely high variability was observ ed (or
would be expect ed) a mong anim als.
Uncertainty factor 100
Interspecies differences

(differences between
humans and common
laboratory animals)
10
Toxico-
dynamics
10
0.4
(2.5)
Intraspecies differences
(differences among
humans; interindividual
differences)
10
Toxico-
kinetics
10
0.6
(4.0)
Toxico-
dynamics
10
0.5
(3.2)
Toxico-
kinetics
10
0.5
(3.2)
FIGURE 5.1 Subdivision of the 100-fold UF showing the relationship between the use of UFs (above the

dashed line), and the proposed subdivisions (below the dashed line) based on toxicokinetics and toxicody-
namics. (From Renwick, A.G., Food Addit. Contam., 10, 275, 1993; WHO=IPCS Assessing human health risks
of chemicals: Derivation of guidance values for health-based exposure limits. Environmental Health Criteria
170. Geneva, 1994. Available at http:== www.inchem.org=documents=ehc=ehc=ehc170.htm)
ß 2007 by Taylor & Francis Group, LLC.
[R] is the adjus tment facto r to account for anti cipated differences in suscep tibility be tween
human s and the laboratory animals, i.e., to account for inte rspecies varia bility. The defaul t value is
10, indicati ng that humans are much more susceptible.
[Q
1–3
] and [U] are adjustm ent factors to account for varia tions in the reliabilit y of the databa se
(data quali ty) and other source s of uncert ainty in the data e valuation p rocess.
[Q
1
]reflects the data evalua tor’ s certa inty that the agent actually causes the speci fic ‘‘critical
effect ’’ in human s. The default value is 1, indi cating that the agent causes similar toxic effect s in
animals and humans.
[Q
2
] is employed when e xtrapolat ing data from subchro nic studies to estimat e risk from lifelong
exposur es. The defaul t value 10, indi cating great uncert ainty in esti mating the NOAEL
chronic
from
the NO AEL
subchronic
.
[Q
3
] is empl oyed when e xtrapolat ing LOAE Ls to NOA ELs. The defaul t value 10, indicati ng
extre mely great uncert ainty associated with using a LOAE L

animal
to estimate a NAEL
human
.
[U] is used to account for resi dual uncertaint y in estimates of [S], [I], and [R]. The default value
is 10 indicati ng very great overal l uncertaint y, which has not alrea dy been account ed for in [Q
1–3
].
[C] is a nonsci enti fi c, judgm ental safet y factor, i.e., a socia l or poli tical value judgm ent. The
defaul t value is 1, indicating that no addit ional MOS is needed over that provi ded by the inheren tly
conser vative procedu re above.
An aggrega te adjus tment factor of about 2 50 is typical; the theor etical maximum v alue is
100,000.
By appli cation of factors [Q
1–3
] and [U], this approac h attempted to separa te scien tific judg-
ments from poli cy=value judgmen ts. Acco rding to the author s, there are three distingu ishing
featu res of the LLN approac h. The first is the empha sis on careful discr iminat ion among the
adjus tments. The second is on discr iminat ion betw een ‘‘best esti mates ’’ of the correct adjus tments
for [S], [I], and [R] and the compl etely separa te adjus tment for overal l un certainty . The thir d is on
securing scien tific consens us on the adjus tment values . It should be recognized, howe ver, that in
pract ice, it will not be possible to dist inguish all these facto rs, and that some factors may not be
indepe ndent of each other. It could also be que stioned whether a nonsci enti fic factor [C] shoul d
be discussed in a scien tific risk asses sment.
5.2.1.5 EU TGD Approach
The process of human health risk assessment has been extensively addressed within the EU
framework of Risk Assessment of New and Existing Chemical Substances. According to the
EU Technical Guidance Document (TGD) on Risk Assessment of New and Existing Chemical
Substances (EC 1996), the risk characterization is carried out by quantitatively comparing the
outcome of the effects assessment to the outcome of the exposure assessment, i.e., a comparison

of the NO AEL, or LOAE L, and the exposur e estimate, see Secti on 8.3.3. The ratio resul ting from
this comparison is called the Margin of Safety (MOS). The TGD recommends the following
parameters to be considered in assessing the MOS:
.
Uncertainty arising, among other factors, from the variability in the experimental data and
intra- and interspecies variation
.
Nature and severity of the effect
.
Human population to which the quantitative and=or qualitative information on exposure
applies
.
Differences in exposure (route, duration, frequency, and pattern)
.
Dose–response relationship observed
.
Overall con fidence in the database
These parameters are parallel to those being considered in the evaluation of the assessmen t factors to
be applied in the establishment of a tolerable intake.
ß 2007 by Taylor & Francis Group, LLC.
The TGD has been revised and the second edition was published in 2003 (EC 2003). However,
the human health risk characterization part was not included in this second edition. A final draft
version of the human health risk characterization part was released in 2005 with a d etailed guidance
on, among other s, the main issues to be included in derivation of the ‘‘reference MOS’’ (MOSref),
which is analogous to an overall assessment factor. The individual factors contributing to the
MOSref are described separately and guidance is given on how to combine these into the MOSref.
The guidance provided in this draft version has been extensively used in relation to the risk
assessment of prioritized substances carried out since the draft version was released; however,
this version is not publicly available.
In the new EU chemicals regulation REACH, which entered into force on 1 June 2007, detailed

guidance documents on different REACH elements, including risk characterization and the use of
assessment factors, are currently in preparation (spring 2007). These documents will probably be
available on the EU DG Environment REACH Web site (EU 2006) when published.
5.2.1.6 ECETOC Approach
The approach recommended by the ECETOC (1995) is to derive the best scientific estimate of a
Human No-Adverse-Effect Level, referred to in the report as the Predicted No-Adverse-Effect Level
(PNAEL). The approach distinguishes three stages:
.
Application of a scientifically derived adjustment factor to the NOAEL, or LOAEL, of the
critical effect established in the pivotal study. It is stated that if the database is inadequate,
then human PNAELs cannot be derived scientifically.
.
Application of a UF to the PNAEL to take into account the degree of scientific uncertainty
involved. The following degrees of confidence in the human PNAEL are suggested:
high ¼ 1, medium ¼ 1–2, low ¼ larger UF.
.
Application of a nonscientifically based safet y factor to take into account political aspects,
socioeconomic aspects (cost–bene fit considerations), or risk perception factors (the nature
of the effect may justify the use of an additional factor).
The scientifically derived adjustment factors include the following elements:
.
Experimental exposure in relation to the expected human exposure: a default value of 3 for
extrapolation from short-term to subchronic exposure; a default value of 2– 3 for extrapol-
ation from subchronic to chronic exposure
.
Extrapolation from LOAEL to NOAEL: a default value of 3
.
Route-to-route extrapolation: no default value
.
Interspecies extrapolation (animal-to-human): a default value of 4 for oral exposure (for the

rat with a body weight of 250 g and based on caloric demands); a default value of 1 for
inhalation
.
Intraspecies extrapolation (human-to-human): a default value of 3 for the general popula-
tion; a default value of 2 for workers
This approach discriminates factors to a large extent in order to distinguish between the single
adjustments and to separate best estimates from uncertainty. It should be noted that the ECETOC
approach does not mention the establishment of an overall factor and although they mention that all
discriminated aspects introduce uncertainties, they do not give guidance on how to account for this.
It could also be questioned here whether a nonscientific factor should be discussed in a scientific risk
assessment.
In a more recent report, ECETOC (2003) has further developed many of the principles
established in the previous report (ECETOC 1995) and replaced the guidance provided there in on
ß 2007 by Taylor & Francis Group, LLC.
the use of asses sment facto rs in human health risk assessmen t. The report provi des a step-by-st ep
guidan ce for deriv ing an approximat ion of a safe exposur e level for human s from the appropriat e
NOA EL, or LOAE L, observ ed in anim al studies, inclu ding guidance on asses sment facto rs.
The most scienti fically supportabl e values for de fault asses sment facto rs recom mende d to be used
in the absence of subst ance-s pecifi c infor mation include:
.
LOAE L-to-NO AEL extrapolat ion: a default value of 3
.
Durati on extrapolat ion: subacute to c hronic, a default value of 6; subchr onic to chronic, a
defaul t value of 2; local effects by inhalati on, a defaul t value of 1
.
Route-t o-route extrapolat ion: oral to inhal ation, no defaul t value; oral to derm al, no default
value
.
Interspeci es extrapolat ion (animal- to-hu man): systemi c effect s (scal ing): mouse, a default
value of 7; rat, a default value of 4; monke y, a defaul t value of 2; dog, a defaul t value o f 2.

Local effects by inhalati on: a default value of 1
.
Intraspeci es extra polation (huma n-to-huma n): syst emic and local effects: a defaul t value of
5 for the general popula tion; a d efault value of 3 for workers
Similar ly to the previous ECET OC app roach, this revised approac h does not ment ion the estab lish-
ment of an overall factor.
5.2.1. 7 Dutch Approa ches
TNO (the Netherl ands Organis ation for Applied Scient i fic Research) has set up a method for
setting Health-based Occu pational Ref erence Valu es (HBOR Vs) (Hakkert et al. 1996). The
HBOR V is derived from the selec ted NOA EL by app lication of asses sment factors compe nsating for
uncert ainties inher ent to extrapolat ion of experi mental data to a given h uman situati on an d
for uncertaint ies in the toxi cological databa se. The asses smen t facto rs should be derived considerin g
the toxi city pro file of the substance; if the availab le data are insuf ficient , an overal l assessmen t facto r
is used compr ising vario us sub-factor s related to:
.
Interspecies differences (animal-to-human): mouse, a default value of 7 3 3; rat, a default
value of 4 3 3; rabbit, a default value of 2.4 3 3; dog, a default value of 1.4 3 3. The first
factor for each species is a calculated adjustment factor, allowing for differences in basal
metabolic rate (proportional to the 0.75th power of body weight). The second factor of 3 is
the assessmen t factor appli ed for remainin g uncertaint ies (Se ction 5.3.3), for which the
default value is 3. For local skin and respiratory tract effects, the assessment factor is 3, as
adjustment for differences in body size is inappropriate.
.
Intraspecies differences (human-to-human): a default value of 3 (workers), a default value
of 10 (general population)
.
Differences between experi mental conditions and exposur e patterns for workers: chronic-
to-chronic ex posure, a default value of 1; subacute-to-subchronic exposure, a default value
of 10; subchronic-to-chronic exposure, a default value of 10; other aspects, a default
value of 1

.
Type of critical effect: a default value of 1
.
Dose–response curve: a default value of 1
.
Confidence of the database: a default value of 1
.
Route-to-route: no default value
Principally, the overall assessment factor is established by multiplication of the separate factors.
The authors note that in practice it is not possible to distinguish all above-mentioned factors, and
some factors are not independent of each other. Therefore, straight forward multiplication may lead
ß 2007 by Taylor & Francis Group, LLC.
to an unreas onably high ov erall factor. Di scussion and weighin g of the individua l factors are
there fore essent ial to estab lish a reliable an d just ifiable overal l asses sment factor.
Ver meire et al. (1999) have p ublished a discu ssion paper with focus on asses smen t facto rs
for human h ealth risk asses sment. The stat us quo with regard to asses smen t facto rs is reviewed and
the paper discusses the develo pment of a formal, harm onized set of asses sment facto rs. Options
are presen ted for a set of d efault values and probabi listic distrib utions for asses smen t facto rs based on
the state of the art. Meth ods of combining defaul t values or probabilistic distributions of assessment
facto rs (Secti on 5.11) are also descri bed. In relation to asses smen t facto rs, the authors recommend ed:
.
For interspecies (animal-to-human) extrapolation, allometric scaling on the basis of caloric
demands (the 0.75th power of body weight) is considered preferable above scaling on body
weight.
.
Traditional extrapolation approach, based on more or less arbitrarily chosen factors of 10,
is considered simple to apply but obscures the relative contributions of scientific arguments
and policy judgments. The other default approac hes, including the application of a toxicity
profile-derived factor as suggested by TNO, make better use of the data available.
.

Worst-case character of the traditional default assessment factors is considered doubtful as
the 95th percentile for the proposed distributions for the interspecies (animal-to-human)
factor and the subchronic-to-chronic duration factor are considerably higher than 10. In
addition, the limited data on intraspecies (human-to-human) variation is also considered to
indicate that a default factor of 10 may not be sufficient.
.
Derivation of approximations of the distribution of assessment factors from historical data
(based on NOAEL ratios) has limitations as the use of the NOAEL instead of the True No-
Adverse-Effect Level brings along the variation (error) in the NOAELs.
.
Application of assessment factors derived from currently estimated distributions of assess-
ment factors may lead to very wide distributions of the overall assessment factor.
.
Probabilistic multiplication of distribution s of assessment factors is preferred above the
simple multiplication of percentiles to avoid extreme conservatism.
A more recent Dutch report (Vermeire et al. 2001) provides a practical guide for the application of
probabilistic distributions of default assessment factors in human health risk assessments, and it is
stated that the proposed distributions will be applied in risk assessments of new and existing
substances and biocides prepared at RI VM (the National Institute of Public Health and the
Environment) and TNO. The report concentrated on the quantification of default distributions of
the assessment factors related to interspecies extrapolation (animal-to-human), intraspecies extrapol-
ation (human-to-human), and exposure duration extrapolation.
5.2.1.8 Kalberlah and Schneider Approach
In a report on a research project ‘‘quantification of extrapolation factors’’ (Kalberlah and Schneider
1998), it is noted that extrapolation factors are intended to replace lack of knowledge by a plausible
assumption, and that institutions with responsibility for establishing the rules must decide which
level of statistical certainty, e.g., applicable for 50% or for 90% of a representative selection of
substances, is desired for the selection of a standard value. It is furthermore noted that extra polation
factors are requi red for: (1) time extrapolation, e.g., from a subchr onic to a chronic duration of
exposure; (2) extrapolation from the LOAEL to the NAEL; (3) interspecies extrapolation, i.e., from

experimental animals to humans; and (4) intraspecies extrapolation, i.e., from groups of persons
with average sensitivity to groups of persons characterized by special sensitivity. In addition to these
extrapolations, route-to-route extrapolation, e.g., oral-to-inhalation or dermal-to-oral must also be
discussed.
ß 2007 by Taylor & Francis Group, LLC.
If no substance-s peci fic know ledge at all is avail able for one of the extrapolat ion steps,
the extra polation facto r in each case is used in unalt ered form; this facto r is described as the
stand ard value.
On average, a factor of 2–3 was consid ered suf ficient for time extra polation from a subchr onic
to a ch ronic duration of exposure . A higher facto r is required in order to cover the 90th percentil e.
Extrapo lation from the LOA EL to the NAE L using standard facto rs shoul d not be und ertaken;
the ben chmark method shoul d be used instead.
If physiologi cally based pharm acokin etic (PBP K) models cannot be used, interspeci es extrapol-
ation is best undertak en by means of scali ng accordi ng to basal met abolic rate, see Secti on 5.3.2.3.
A second aspect, interspeci es varia bility, shoul d be consi dered in cases where a higher than average
level of safety (achi eved by consi deration of a higher percentil e of the substances) is desired.
In general, an intraspeci es facto r of 10 should be suffi cient to re flect the toxi cokinetic va riability
between healt hy adults; ho wever, it is not suf ficient wi th regard to toxi codynam ic varia bility and
possibly only consi ders risk groups to a lim ited extent.
If subst ance-s pecifi c know ledge regard ing indi vidual extrapolat ion steps is avail able but not
suf ficient to be able to d ispense wi th the extrap olation enti rely, the substance-s peci fic infor mation is
used on a priority basis for the purpos e of modi fying the stand ard value, reduct ion or possi bly also
an increase.
A furt her probl em lies in the combinati on o f the indi vidual extra polation facto rs to form a
total extra polation factor. The type of combi nation results from the dependen ce or indepe ndence of
the individua l sub-factor s. According to current know ledge, multiplica tive combi nation of the
individua l facto rs is assum ed. Su bstance-spe ci fic knowledge about the inte rdepend encies among
the sub-f actor s may lead to modi ficati on, i.e., a reduct ion of the total extra polation facto r.
5.2.1. 9 UK Appro ach
The Interdepa rtmental Group on Heal th Risks from Che micals (IG HRC) in the United Ki ngdom has

published a docum ent entitled ‘‘ Uncertai nty Factor s: The ir Use in Human Health Risk Ass essment
by UK Gove rnme nt’’ (IG HRC 200 3). The docum ent inte nded to lay out the principle s used in the
United Kingdom .
The uncertaint ies fall into two broad catego ries . Firstly, there are the uncertaint ies related to
the extrapo lation of the key data from exp erimenta l anim al species to the ‘‘av erage’’ human (animal-
to-hu man), and then from the ‘‘average ’’ human to other mem bers of the popula tion with different
charact eristics (hum an-to -human), i.e., those with great er sensitivity . Secondl y, there are then a
numbe r of uncertaint ies relat ed to the avail able database incl uding those aris ing from route -to route
extra polation, duration of exposure, NOA EL not estab lished or not firml y estab lished, and gaps or
other deficiencies in the database.
Cha pter 5 of the docum ent reviews the UFs used by UK Governm ent department s, agenci es,
and their advisory committees in human health risk assessment. Default values for UFs are provided
in Table 3 in the UK document with the factors separated into four classes: (1) animal-to-human
factor, (2) human variability factor, (3) quality or quantity of data factor, and (4) severity of effect
factor. The following chemical sectors are addressed: food additives and contaminants, pesticides
and biocides, air pollutants, drinking water contaminants, soil contaminants, consumer products and
cosmetics, veterinary products, human medicines, medical devices, and industrial chemicals.
5.2.1.10 Swedish National Chemicals Inspectorate’s Approach
The Swedish National Chemicals Inspectorate (KEMI) has published an extensive review on human
health risk assessment with focus on the application of assessment factors in risk assessments for
plant protection products, industrial chemicals, and biocidal products within the European Union
(KEMI 2003).
ß 2007 by Taylor & Francis Group, LLC.
One of the main conclusions drawn from the evaluation of the available data on default
assessment factors was that the conventionally used factor of 100 (10 for animal-to-human and
10 for human-to- human variations) is probably an underestimate. It is stated that it is likely that the
animal-to-human extrapolation is greatly underestimated, and in the case of human-to-human
variability, an assessment factor of 10–16 is considered as a minimum.
Attention is also drawn to the fact that there are some other elements not included in the
traditional assessment factor of 10 including adequacy of the database, nature of the effect, duration

of exposure, route-to-route extrapolation, and considerations of extra-sensitive subpopulations such
as children, the elderly, and patients under medical treatment.
The use of default assessment factors is recommended in risk assessments, when justifiable,
although the scientific background for such factors in general was considered unsatisfactory. The
default assessment factors suggested are summarized in Table 5.2. It is recommended to use
assessment factors derived from probabilistic distributions in favor of deterministic assessment
factors, see Table 5.2.
TABLE 5.2
Deterministic and Probabilistic Assessment Factors Suggested for Use
in Human Health Risk Assessment
Area to be Extrapolated
Assessment Factor
Deterministic Approach Probabilistic Approach
Adequacy of the toxicological database
relevance, validity, reliability 1–5 —
children 1–10 —
Nature of the effect
adversity, severity, potency 1–10 —
Duration of exposure
subacute (1 month) to subchronic (3 months) 3 —
subchronic (3 months) to chronic (24 months) 8 10 (90th); 16 (95th); 37 (99th)
subacute (1 month) to chronic (24 months) 24 25 (90th); 39 (95th); 92 (99th)
Route-to-route extrapolation
dermal NOAEL from oral NOAEL 100% or case-by-case —
inhalation NOAEL from oral NOAEL 100% or case-by-case —
Dose–response curve
shape of the curve Case-by-case —
LOAEL to NOAEL BMDL
5
or 3–10 —

Interspecies extrapolation
rat to human 4 (TK) 3 2.5 (TD) ¼ 10 28 (90th); 48 (95th); 132 (99th)
mouse to human — 49 (90th); 84 (95th); 231 (99th)
Intraspecies extrapolation 3–5 (TK) 3 3.16 (TD) ¼ 10–16 —
Source: Modified from KEMI, Human health risk assessment. Proposals for the use of assessment (uncertainty) factors.
Application to risk assessment for plant protection products, industrial chemicals and biocidal products within the
European Union. Report No. 1=03, Solna, Sweden, 2003.
TK: toxicokinetics.
TD: toxicodynamics.
BMDL
5
: 5% lower confidence limit of the benchmark dose.
ß 2007 by Taylor & Francis Group, LLC.
5.2.1. 11 Dani sh EPA’ s Appr oach
In Denm ark, health-bas ed quality criteria are set for chemical subst ances in soil, drink ing water,
and ambi ent air accordi ng to princ iples laid do wn in a guidan ce docum ent from the Dani sh
Environ mental Protec tion Agency (D-E PA 2006). The princ iples laid down in the guidan ce
docum ent are based on an extensive revie w addressing the hazard asses smen t o f chemicals ,
including applicati on of asses sment facto rs (Niels en et al. 2005).
For threshold effects, a Tolerabl e Daily Intake (TDI ) is calcula ted by divi ding the NOA EL
(or LOA EL) for the crit ical effect(s) with an overal l UF. The current practice according to the
D-EPA in relation to the setting of quali ty crit eria for chemical substances in soil , drink ing water,
and ambient air is to divide the overal l UF into three catego ries (D- EPA 2 006):
.
UF
I
account s for the inte rspecies variation in suscep tibili ty. The default value is 10 when
correction for diff erences in body size between human s and experi mental anim als is based
on the body weight.
.

UF
II
accou nts for the differences in interindi vidual suscep tibili ty. The defaul t value is 10.
.
UF
III
accoun ts for the quality and relev ance of the databa se, i.e., account s for the
uncert ainties in the estab lishmen t of a NO AEL for the crit ical effect . The UF
III
includes
elements such as (1) the quali ty of the databa se, e.g., data on speci fic toxic endpoi nts are
lacking or inadeq uate, default value of 1–10; (2) route-to- route extra polation, e.g., no
studies using the appropr iate exposur e route are av ailable, no defaul t value; (3) LOA EL-to-
NOA EL extra polation, e. g., a NOAEL cannot be estab lished for the critical effect, default
value of 1 0; (4) subchr onic-to-ch ronic extra polation, e.g., no chroni c studi es on whi ch to
estab lish the NOAEL are availab le, default value of 10; and (5) nature and severi ty of
toxicit y, e.g., the crit ical effect is toxicity to reprod uction, carci nogeni city or sensitiz ation,
defaul t value of up to 10. A de fault value for UF
III
has no t been recommend ed; however, a
value from 1 to 100 is generally used. The value is ev aluated case-b y-case based on expert
judgme nt.
The overal l UF is derived by mul tiplication of the single UFs. Recogniz ing that the overall UF
might be unrealist ically high, a final revie w of the overall UF is pe rformed in relat ion to the
available data. If the magni tude of the overal l UF is very high (e.g., a bove 10,000), the databa se
is con sidered as being too limit ed in order to set a healt h-base d quality criteria in soil, drink ing
water, and ambient air for the speci fi c chemi cal subst ance.
5.2.1. 12 Chem ical-Speci fic Assess ment Factors
AWHO=IPC S (2005) Har monizati on Projec t Docu ment has propos ed using chemical- speci fi c
toxicological data instead of default assessment factors, when possible. The concept of Chemical-

Specific Adjustment Factors (CSAFs) has been introduced to provide a method for the incorporation
of quantitative data on interspecies differences or human variability in either toxicokinetics or
toxicodynamics into the risk assessment procedure, by modifying the relevant default UF of 10.
Incorporation of toxicokinetic or toxicodynamic data becomes possible if each factor of 10 is
divided into appropriately weighted sub-factors as suggested by Renwick (1991, 1993) and adopted
by WHO=IPC S (1994) , see Secti on 5.2.1 .3.
When appropriate chemical-specific data are available, a CSAF can be used to replace the relevant
default sub-factor; for example, suitable data defining the difference in target organ exposure in animals
and humans could be used to derive a CSAF to replace the uncertainty sub-factor for animal to human
differences in toxicokinetics (a factor of 4). The overall UF would then be the value obtained on
multiplying the CSAF(s), used to replace default sub-factor(s), by the remaining default sub-factor(s) for
which suitable data were not available. In this way, chemical-specific data in one area could be
introduced quantitatively into the derivation of a tolerable intake, and data would replace uncertainty.
ß 2007 by Taylor & Francis Group, LLC.
The WHO=IPCS (2005) guidan ce docum ent describ es the types and quali ty of data that could be
used to deriv e a CSAF . The guidance is separa ted into four main sections coveri ng each of the four
diff erent areas wher e CSAFs can be introd uced to replace a defaul t sub-f actor:
.
Data relate d to interspeci es diff erences in toxicokin etics
.
Data relate d to interspeci es diff erences in toxicodynam ics
.
Data relate d to human varia bility in toxicokin etic s
.
Data relate d to human varia bility in toxicodynam ics
The combi nation of adjustm ent facto rs and defaul t UFs to deriv e an overal l UF is also addres sed.
In the 2002 revie w of the RfD and RfC proces ses (US- EPA 2002), the growing suppor t for the
use of CSAFs in place of DAFs was noted, and this will provide an incent ive to fill exist ing data
gaps. The US-EPA has not yet estab lished a guidan ce for the use of chemi cal-speci fic data for
deriv ing UFs, but the divi sion of UFs into toxicodynam ic an d toxi cokine tic compo nents is in the

RfC methodol ogy (US-E PA 1994). It was pointed out that, for many substances, there are relat ively
few data available to serve a s an adequat e basis to repla ce defaul ts for inte rspecies diff erences and
human varia bility with more informat ive CSAFs. Current ly, relevant data for consi deration are often
rest ricted to the component of u ncertaint y relat ed to interspeci es diff erences in toxicokin etic s.
5.2.1. 13 Chi ldren-Sp ecifi c Assessmen t Factor
Con cern has been rais ed that infan ts and children are at higher risk than adults from exposur e to
envir onmen tal c hemicals . The quest ion of an extra assessmen t facto r in the hazard and risk
asses sment for chemi cals of concern for chil dren has there fore been raised and the rationale for
such a children-s pecifi c asses smen t facto r has been discussed.
Ren wick et al. (2000) have performed an analys is of the need for an addit ional UF for infants and
chil dren. The y consi dered that the propos al to introduce an additional 10-fold facto r when exposur e
of infan ts and chil dren is anticipat ed implie s either age -related differences between speci es or
diff erences wi thin human s, which exceed those presen t in adults. Altern ativ ely, the extra factor
could be relat ed to de ficienci es of curren t testing methods or concern s over irreversibil ity in devel-
oping organ syst ems. The y conclu ded that the available data did not provide a scienti fic rationale for
an extra facto r due to inadequacy of inter- and intraspeci es UFs. Justi fication for the facto r therefore
must relate to the adequacy and sensitivity of curren t met hods or concern about irreversi ble effects in
the develo ping organi sm. They also pointed out that when adequate reprod uction, multigenerat ion, or
develo pment al studies are cond ucted, there wi ll be no need for an addit ional 10-fold facto r.
In setting pesticide tole rances , the U.S. Fo od Qual ity Protec tion Act (FQPA) adopte d in 1996
directed the US-EPA to apply an extra safety factor of 10 in assessing the risks to infants and children
(US-EPA 1996). This additional 10-fold MOS should take into account the potential for pre- and
postnatal toxicity, and the completeness of the toxicology and exposure databases recognizing that
maturing organ systems of infants and children may be susceptible to injury by chemicals. There may
be developmental periods, i.e., windows of vulnerability, when endocrine, reproductive, immune, and
nervous systems are particula rly sensiti ve to certa in chemi cals, see Secti on 5.4 .1.1. When data are
missing or inadequate for an evaluation of the age group or of a window of vulnerability during
development, the application of the extra factor, in addition to the general default factor of 10 for
intraspecies variation (human-to-human), was considered appropriate.
The FQPA authorizes the US-EPA to replace this additional 10-fold factor with a factor of a

different value (higher or lower, including 1) only if, on the basis of reliable data, the resulting level
of exposure would be safe for infants and children. In practice, factors of 3 and 10 have been used,
and the factor has also been used in cases when data have been sufficient but there were reasons for
concern (US-EPA 2002).
In addition to considering the FQPA-relevant areas of uncertainty, assessments of pesticide risk
to child ren also consider applying part or all of the FQPA factors in certain situations to account for
ß 2007 by Taylor & Francis Group, LLC.
areas of resi dual uncertaint y that the traditional UFs do not addres s or for which they are believed to
be insuf ficient. These areas of residual uncert ainty incl ude exposur e unc ertainties and high concern
for an observ ed suscep tibility (US-EPA 2002).
The US-EPA has conclu ded that in many cases, concern s regard ing pre- and postnatal toxicit y
can be addres sed by calculating an Rf D by using pre- or postn atal develo pmental endpoi nts
and applying the UFs (interspe cies (Secti on 5.3), intr aspecies (Section 5.4), LOAE L-to-NO AEL
(Section 5.7), subchronic- to-ch ronic (Secti on 5.6), and databa se-de ficiency (Section 5.9)) to account
for de ficienci es in the toxi city data when there are gaps consi dered essential for setting a refere nce
value, incl uding lack of data on children (US-E PA 2002).
The overlap of areas covered by the FQPA factor and those a ddressed by the traditional UFs
was recogni zed, a nd it was conclu ded that the curren t UFs, if appropr iately applied using the
approac hes recom mende d in the review (i.e., US-EPA 2002), will be adequat e in most cases to
cover concern s and uncert ainties regard ing the potential for pre- and postn atal toxi city and the
completen ess o f the toxi cology databa se. In other words, an addit ional UF is not needed in the
RfC=RfD met hodolo gy because the curren tly available facto rs are considered suf ficient to account
for uncert ainties in the databa se from whi ch the refere nce values are deriv ed (and it does not exclud e
the possi bility that these UFs may be decreased or increased from the default value of 10).
In a report prepared for the Danish Environ mental Protec tion Agency (Niels en et al. 2001) with
the purpose of revie wing the knowledge on the ex posure and vulner ability of human s to chemical
substanc es durin g the embryonic, fetal, and postn atal perio ds, it was stro ngly recommend ed to
perfor m child-spec ifi c risk a ssessments for che mical substances in product s and foods inte nded for
children (e.g., in cosm etics, toys, child care products, food addit ives in prefer red foods, and
pesticide resid ues in proces sed baby foods and infan t formulae) . In addition, in the risk asses sment

of chemical subst ances in other use catego ries than the above-m entioned, it was recom mende d
speci fically to focus on children, including the unborn child, if a potential e xposure to a
given subst ance may occur to these a ge groups . Fu rthermor e, it was recom mende d that the risk
asses sment should be perfor med by expert s on a case-b y-case basis for each substa nce and for each
exposur e scenar io. In cases where the available data are insuf ficient to evalua te the suscep tibility of
children, incl uding the unborn chil d, it was stro ngly recom mende d that addit ional safety meas ures
(choi ce of safety facto rs) should be considered when tolerable inta kes are establis hed for chemical
substanc es in products and foods inte nded for children.
In conclu sion, the traditi onal a ssessment facto rs (interspe cies, intr aspecies, subchr onic-
to-ch ronic, LOAE L-to-NO AEL, and database-de ficiency ) are considered to cover the concern s
and uncertaint ies for children adequat ely, i.e., no children-s pecifi c asses smen t facto r is needed
when setting tole rable intakes. However , it is recom mended to perfor m children-s peci fic risk
asses sments for chemical subst ances in products and foods intended for chil dren, based on speci fic
exposur e assessmen ts for children.
5.3 INTERSPECIES EXTRAPOLATION (ANIMAL-TO-HUMAN)
Data from studies in expe rimental animals are the typical starting point s for hazard and risk
assessments of chemical substances and thus differences in sensitivity between experimental animals
and humans need to be addressed, with the default assumption that humans are more sensitive
than experimental animals. The rationale for extrapolation of toxicity data across species is founded
in the commonality of anatomic characteristics and the universality of physiological functions and
biochemical reactions, despite the great diversity of sizes, shapes, and forms of mammalian species.
This section gives a short introduction regarding the biological variation between mammalian
speci es (Secti on 5.3.1 ) as a basis for the subseq uent section on allo metric scaling (Section 5.3.2).
Then a number of analyses performed regarding the validity of the default assessment factor of 10
are revie wed (Secti ons 5.3.3 and 5.3.4). Finally, the key issues are summari zed and our recom -
menda tions are presen ted (Section 5.3.5).
ß 2007 by Taylor & Francis Group, LLC.
5.3.1 BIOLOGICAL VARIATION
It is often postulated that extrapolation of biological data from animals to man must be viewed with
extreme caution because of man’s biological uniqueness.

Olson et al. (2000) examined the strengths and weaknesses of animal studies to predict human
toxicity. The examination was based on the results of a multinational pharmaceutical company
survey, which covered compounds where human toxicity was identified during clinical development
of new pharmaceuticals, determining whether animal toxicity studies identified concordant target
organ toxicities in humans. Data were compiled for 150 compounds with 221 human toxicity events
reported; multiple human toxicity was reported in 47 cases. The results showed a positive human
toxicity concordance rate of 71% for rodent and non-rodent species, with non-rodents alone being
predictive for 63% of human toxicity and rodents alone for 43%. The highest incidence of overall
concordance was seen in hematological, gastrointestinal, and cardiovascular effects, and the least
was seen in cutaneous effects.
Although testing of chemicals in experimental animals to a great extent is predictive for human
toxicity, humans might be more or less susceptible to the effect(s) exerted by a toxic chemical
compared with other mammals, as also is the case between animal species. Absorption, distribution
and storage, excretion, metabolism, site or target organ, and mechanism(s) of action are all involved
in the toxicological response to a chemical. Thus, interspecies differences result from variation in
the sensitivity of species due to differences in toxicokinetics as well as in toxicodynamics. Some of
the toxicokinetic differences can be explained by differences in body size and related differences in
basal metabolic rate (caloric requirement).
In general, absorption of chemicals is comparable among vertebrate species for the oral and
inhalation route, whereas differences in dermal absorption are much more pronounced because of
differences in skin morphology between vertebrates. The distribution and storage of chemicals, once
they have been absorbed, also tend to be comparable across vertebrates, although there are differences
related to, e.g., protein binding. In terms of renal and pulmonary excretion, the differences between
the common laboratory animals and humans are minimal. The metabolism of chemicals is, however,
generally far from comparable from species to species. Not only are different metabolites some-
times formed, but the rate of formation of identical metabolites may also be species specific. The
mechanism(s) and sites of action may also differ across species, both qualitatively and quantitatively.
A classical exa mple is the thalidomide-induced teratogenicity in humans where the variability
between species in susceptibility is thought to be largely explained by different metabolic pathways
in humans and laboratory animals, since the ultimate human teratogen is a metabolite of thalidomide.

This highlights the importance of uncertainties in interspecies extrapolation in cases where the
expression of chemical toxicity is related to its metabolism, and it is widely believed that interspecies
differences in metabolism of xenobi otics is usually the most significant explanatory factor for
observed interspecies differences (Davidson et al. 1986, Voisin et al. 1990, Calabres e et al. 1992).
RIVM, the Dutch National Institute for Public Health and the Environment, has launched a Web
site in January 2006 with information of physiological and anatomical parameter values in various
species frequently used in toxicity testing. The parameters are focused on organs and tissues
relevant for pharmacokinetics following oral exposure. The aim of the Web site is to gain insight
into the impact of anatom ical and physiological differences between species on the pharmacoki-
netics. This insight may lead to improved species selection and subsequently to improved animal-
to-human extrapolation (RIVM 2007).
In addition to the toxicokinetic and toxicodynamic differences mentioned above, other aspects
of differences between experimental animals and humans include different types of organs and
tissues, differences in digestion, and differences in the structure of the upper respiratory tract.
Furthermore, animal studies are performed in homogenous groups of animals, but the results have to
be applied for the protection of all individuals in a heterogeneous population of humans. In
consequence of this, interspecies variation must also be expected.
ß 2007 by Taylor & Francis Group, LLC.
Extrapolation of data from studies in experimental animals to the human situation involves two
steps: a first step is to adjust the dose levels applied in the experimental animal studies to human
equivalent dose levels, i.e., a correction for differences in body size between laboratory animals and
humans. A second step involves the application of an assessment factor to compensate for
uncertainties inherent in toxicity data as well as the interspecies variation in biological susceptibil-
ity. These two steps are addressed in the following sections.
5.3.2 ADJUSTMENT FOR DIFFERENCES IN BODY SIZE:ALLOMETRY=SCALING
One aspect in the extrapolation of data from studies in experimental animals to the human situation
is, as mentioned above, a correction of the dose levels in experimental animal studies to equivalent
human dose levels, e.g., a NOAEL derived from an animal study to the equivalent human NOAEL.
Adolph (1949, as cited in Davidson et al. 1986, Voisin et al. 1990, ECETOC 2003) compiled a
list of 34 morphological, physiological, and biochemical parameters, which correlated with inter-

species body weight in accordance with the following general allometric equation:
Y ¼ aW
n
where
Y is a biological function
W is body weight
a, n are species-independent constants for the biological function Y
Values obtained for the exponent n ranged from 0.08 to 1.31. The geometric mean of all
n values was 0.82 and a frequency distribution indicated that values from about 0.67 to 0.75 were
most prominent (Adolph 1949, as cited in ECETOC 2003).
Today, well over 100 biological parameters of mammals are known to be linearly related to
body weight and highly predictable on an interspeci es basis (Davidson et al. 1986, Voisin et al.
1990, Calabrese et al. 1992). The allometric equation has traditionally been used for extra polation of
experimental data concerning physiologic al and biochemical functions from one mammalian
species to another. In addition, the allometric equation has also been used extensively as the basis
for extrapolation, or scaling, of e.g., a NOAEL derived for a chemical from studies in experimental
animals to an equivalent human NOAEL, i.e., a correction for differences in body size between
humans and experimental animals.
Where an interspecies correlation is assumed to exist between Y (biological effect) and body
weight, such that if Y ¼ aW
n
and the dose (mg) associated with Y in an experimental animal equals
X, then
Y ¼ f (X) ¼ f (aW
n
)
For the observed dose in an experimental animal study, X
animal
, and the equivalent dose in man,
X

human
, the scaling factor for man from the experimental animal is
Scaling ¼ X
human
=X
animal
¼ a(W
human
)
n
=a(W
animal
)
n
and
X
human
¼ X
animal
 [W
human
=W
animal
]
n
To correct for differences in body size between humans and experimental animals, three measures of
body size are used in practice as the basis for the extrapolation: body weight, body surface area, and
caloric requirement (Feron et al. 1990, Vermeire et al. 1999, KEMI 2003).
ß 2007 by Taylor & Francis Group, LLC.
The reasons for using these three measures and the advantages and disadvantages of their

use have been described by Davidson et al. (1986) and Vocci and Farber (1988). In these
papers, it is also explained why the body weight can be used in all three cases. However, the
body weight should be taken to the power of 1, 0.67, and 0.75 for the body weight approach,
the body surface area approach, and the caloric requirement approach, respectively. These figures
indicate that the approach used to correct for differences in body size will clearly affect the value of
the NOAEL adjusted to the body size of humans.
5.3.2.1 Adjustment for Differences in Body Size: Body Weight Approach
Body weight is considered as being the most easily and accurately measurable of the three measures
of body size used in practice as the basis for the extrapolation, and most often provides the
quantitative basis for the correction of doses for differences in body size between experimental
animals and humans.
When correction for differences in body size is based on body weight, the exponent n in the
allometric equation is 1 and the human dose X
human
(expressed in mg) can be calculated as follows:
X
human
¼ X
animal
 [W
human
=W
animal
]
1
The scaling factor [W
human
=W
animal
]

1
between a man weighing 70 kg and a rat weighing 250 g is 280
when correct ion for differences in body size is based on body weight. Similarly, the scaling factor
between a man weighing 70 kg and a mouse weighing 35 g is 2000. It should be recognized that the
scaling factor and thus the uncertainty in extrapolating doses from experimental animals to
equivalent human doses is heavily dependent on the choice of body weight for man as well as for
experimental animals.
When the observed dose in an experimental animal study is expressed in mg=kg body weight,
then the equivalent human dose (in mg=kg body weight) is equal to the dose in the experimental
animal study as the scaling factor is 1. This is illustrated by the following example: a NOAEL of
1 mg has been derived from an experimental study with rats. By assuming a body weight of 250 g
for the rat, the NOAEL is 4 mg=kg body weight. The equivalent human NOAEL can be calculated
to 280 mg based on a human body weight of 70 kg (1 mg 3 [70 kg=0.25 kg]), or 4 mg=kg body
weight (280 mg=70 kg), see also Table 5.3.
TABLE 5.3
Adjustment of Dose Levels for Differences in Body Size between Humans
and Experimental Animals. Examples of Deriving a Human NOAEL by the Various
Approaches for a Chemical with a NOAEL of 1 mg in a Rat Study
Rat NOAEL (mg)
Rat NOAEL
(mg=kg bw) Human NOAEL (mg)
Human NOAEL
(mg=kg bw)
Body weight
approach
a
1mg 4mg=kg bw
(1 mg=0.25 kg)
280 mg
(1 mg 3 [70 kg=0.25 kg]

1
)
4mg=kg bw
(280 mg=70 kg)
Body surface area
approach
b
1mg 4mg=kg bw
(1 mg=0.25 kg)
43.6 mg
(1 mg 3 [70 kg=0.25 kg]
0.67
)
0.62 mg=kg bw
(43.6 mg=70 kg)
Caloric requirement
approach
c
1mg 4mg=kg bw
(1 mg=0.25 kg)
68.4 mg
(1 mg 3 [70 kg=0.25 kg]
0.75
)
0.98 mg=kg bw
(68.4 mg=70 kg)
Note: Assumed body weight (bw) of rat 0.25 kg, of human 70 kg.
a
X
human

¼ X
animal
3
[W
human
=W
animal
]
1
.
b
X
human
¼ X
animal
3
[W
human
=W
animal
]
0.67
.
c
X
human
¼ X
animal
3
[W

human
=W
animal
]
0.75
.
ß 2007 by Taylor & Francis Group, LLC.
According to Voisin et al. (1990), physiological and metabolic processes such as renal function,
metabolic rate, and cardiac function are not directly proportional to body weight and thus the toxic
effects influenced by these physiological processes are not proportional to body weight, especially
when extrapolating from smaller to larger animals. Consequently, interspecies comparisons based
directly on body weight are likely to be very inaccurate in predicting chemical-induced toxicity
across species.
5.3.2.2 Adjustment for Differences in Body Size: Body Surface Area Approach
The surface area approach has been proposed as an alternative to correction for differences in body
size based on body weight. This approach is founded on the notion that the basal metabolic rate of
vertebrates is a fundamental biological parameter, i.e., a final common expression of physiological
and biochemical functions, which is remarkably well related to the body surface area across species
and within species (Davidson et al. 1986).
The most comprehensive attempt to assess interspecies differences in susceptibility to toxic
responses, based on two diff erent dose correction approaches (body weight versus body surface
area), was published in the classic paper by Freireich et al. (1966, as reviewed in Davidson et al.
1986, Calabrese et al. 1992, Grönlund 1992). The authors attempted to standardize various
toxicological studies for 18 anticancer drugs performed in adult mice, rats, hamsters, dogs, mon-
keys, and humans. The findings of this study led to the conclusion that the toxic effects of an agent
were simil ar across species when the dose was measured on the basis of the body surface area.
Dourson and Stara (1983) have noted that dose conversions based on body surface in general
more accurately reflect differences among species in several biological parameters when compared
to conversions based on the body weight.
Calabrese et al. (1992) have also noted that the use of the body surface area approach for correction

for differences in body size between experimental animals and humans is a more conservative approach
than use of the body weight approach. The authors also noted that as the animal model approaches
human dimensions of weight and surface area, the differential in dose correction between body weight
and body surface area is minimized. Furthermore, the body surface area approach was considered likely
to account for certain toxicokinetic differences, especially those associated with interspecies differences
in blood flow to organs or enzymatic parameters of importance for the metabolism of substances, which
typically scale according to surface area. However, the body surface area approach did not appear to
address issues of interspecies variation due to, e.g., differences in absorption efficiencies, thickness of
epidermal tissue, number of hairs per square centimeter of skin, the presence and quantity of the gut
microflora, the relative dominance of oxidative and conjugative metabolic pathways, and the rate of
biliary excretion. It was also stated that the relative importance of these factors will differ from
compound to compound and from species to species, ranging from unimportant to critical and thus
standard dose correction practice does not eliminate the reality of variability with respect to how the
different species handle and respond to agents over a wide range of doses.
Renwick (1999) has noted that most physiological and many biochemical proces ses correlate
better with body surface area than with body weight. For compou nds, which are metabolized by
processes of intermediary metabolism, or for which the clearance is determined largely by blood
flow to the organ(s) of elimination, there is a significa nt discrepancy between doses expressed on the
basis of body surface area and those based on body weight, when comparing rodents with humans.
Interspecies factors of about 3–4 and 8–10 would be necessary for rats and mice, respectively, to
convert the external dose expressed in mg=kg body weight into a dose based on body surface area,
and therefore, more closely related to the species differences in basal metabolic rate and organ blood
flows between humans and these species.
The above-mentioned references thus indicate that the body surface area approach apparently is
a more feasible approach than the body weight approach in terms of dose correction for differences
in body size between experiment al animals and human s.
ß 2007 by Taylor & Francis Group, LLC.
The allomet ric equation relating the body surfa ce area (BSA) to the body weight ( W )isas
follow s:
BSA ¼ a W

2 =3
wher e the surfa ce area is a funct ion of body weight to the powe r 2=3, and a speci es-sp ecifi c
const ant, a.
When correct ion for diff erences in body size is based on body surfa ce area, the powe r n in the
allo metric equation is 2=3, or 0.67, and the human dose X
human
(expre ssed in mg) can be calculated
as follow s:
X
human
¼ X
animal
 [ W
human
=W
animal
]
0 :67
( 5:1)
The scali ng factor [ W
human
=W
animal
]
0.67
betw een a man weighin g 70 kg and a rat weighin g 250 g is
43.6 when correct ion for differences in body size is based on body surfa ce area. Similar ly, the
scaling facto r between a man weighin g 70 kg and a mous e weighin g 35 g is 163. The correspondi ng
scaling facto rs obtained based on the body weight approac h are 280 and 20 00, respec tively (Section
5.3.2.1). Thu s, the diff erence between the two extra polation bases, W

1
and W
0.67
, for ‘‘to man from
rat ’’ and for ‘‘to man from mous e ’’ is 6.4-fold an d 12.3-fo ld, respec tively, great er for W
1
compa red
to W
0.67
. The refore, scaling by the body weight approac h provides equiva lent human doses of
roughl y an order of magnitud e great er than scali ng by the bo dy surface area approac h. Henc e,
scaling by the body weight approac h without a biol ogically justi fiable reason may overes timate the
equiva lent human dose.
When the observ ed d ose in an experiment al anim al study is expres sed in mg=kg body weight,
then the equiva lent human dose (in mg=kg body wei ght) is equal to the dose in the experi mental
anim al study divi ded by a scalin g facto r a ccording to the follow ing equati on:
X
human
¼ X
animal
=[W
human
=W
animal
]
0 :33
( 5:2)
Acco rding to this e quation, the scaling facto r [ W
human
=W

animal
]
0.33
is 6.4 ‘‘ to man (70 kg) from rat
(250 g) ’’ a nd 12.3 ‘‘to man (70 kg) from mous e (35 g). ’’ This is illust rated by the follow ing examp le:
a NO AEL of 1 mg has been deriv ed from an experi mental study wi th rats. By assum ing a body
weight of 250 g for the rat, the NO AEL is 4 mg=kg b ody weight. According to Equ ation 5.1, the
equiva lent human NOA EL can be calcul ated to 43.6 mg based on a human body weight of 70 kg
(1 mg 3 [70 kg=0.25 kg]
0.67
), or 0.62 mg=kg body weight (43.6 mg=70 kg). Acco rding to
Equ ation 5.2, the eq uivalent human NOAEL can be calculated to 0.62 mg=kg body weight
(4 mg=kg 3 [70 kg=0.25 kg]
0.33
), i.e., the anim al d ose (in mg=kg bo dy weight) divided by the
scaling facto r, in this case 6.4. Se e also Table 5.3.
The surfa ce area for different speci es can be calcul ated by empi rically derived equations using a
speci es-speci fic ‘‘shape factor, ’’ which depends on the ration of weight to height (Vo isin et al.
1990). Acco rding to Voi sin et al. (1990) , there are severa l limitati ons in the accuracy of conversions
based on body surface area: (1) the surface area appears to be dif ficult to esti mate; (2) some analyses
have indicated that the exponent of 2=3 may be inaccu rate; (3) some physiologi cal parameter s are
not so well related to body surface area; (4) body surface area convers ions are inaccu rate when the
mode of ad ministration is different across speci es; and (5) n ot all types of toxi city correl ate with
body surfa ce area, e.g., skin toxi city.
5.3.2.3 Adjustment for Differences in Body Size: Caloric Requirement Approach
The caloric requirement, or metabolic rate approach, has also been proposed as an alternative to
correction for differences in body size based on body weight.
ß 2007 by Taylor & Francis Group, LLC.
A number of parameters such as, e.g., renal clearance, basal oxygen consumption (metabolic
rate), area under the curve (AUC), maximum metabolic velocity, or cardiac output correlate to the

body weight to the power of 0.75 (W
0.75
). Further support for the power of 0.75 comes from a more
theoretical approach based on fractal geome tric and energy conservation rules for mammalian
species (West et al. 1997, 1999, as cited in ECETOC 2003).
It is important to note that extrapolation using allometric scaling based on metabolic rate
assumes that the parent compound is the toxic agent and that the detoxification is related to
the metabolic rate and thus controls the tissue level. This is relevant for oral exposure only
(ECETOC 2003).
Feron et al. (1990) have concluded that, in general , adjustment for differences in body size
between experimental animals and humans should be based on caloric requirement (energy metabo-
lism) as this was considered to be both scientifically sound and of practical significance.
In 1992, the US-EPA has adopted the caloric requirement approach for oral exposures (US-EPA
1992, as cited in US-EPA 2005), and it is stated that doses should be scaled from animals to humans
on the basis of equivalence of milligrams of the agent normalized by the 3=4 power of body weight
(W
0.75
) per day (US-EPA 2005). The 3=4 power is considered as being consistent with current
science, including empirical data that allow comparison of potencies in humans and animals, and it
is also supported by analysis of the allometric variation of key physiological parameters across
mammalian species. It is generally more appropriate at low doses, where sources of nonlinearity
such as saturation of enzyme activity are less likely to occur. This scaling is intended as an unbiased
estimate rather than a conservative one. It is furthermore noted that equating exposure concentra-
tions in food or water is an alternative version of the same approach, because daily intakes of food or
water are approximately proportional to W
0.75
.
Vermeire et al. (1999) have noted that scaling on the basis of surface area or caloric demand can
be considered more appropriate compared to extrapolation based on body weight; however, they
also noted that experimental work did not answer the question regarding which of these two

methods is the most correct. Based on theoretical grounds, and supported by their own analyses,
Vermeire et al. (1999) concluded that scaling on the basis of caloric demand to adjust oral NOAELs
for metabolic size can be considered more appropriate compared with extrapolation based on body
weight. It was also noted that an allometric exponent of 0.67, i.e., the body surface area approach,
seems to better describe intraspecies relations.
Based on theoretical grounds, the TNO (Hakkert et al. 1996) and Kalberlah and Schneider
(1998) consider the interspecies extrapolation based on caloric demands (W
0.75
) as preferable above
scaling on body weight.
The allometric equation relating the caloric requirement (CR) to the body wei ght ( W)isas
follows:
CR ¼ aW
3=4
where the caloric requirement is a function of body weight to the 3=4 power, and a species-specific
constant, a.
When correction for differences in body size is based on caloric requirement, the exponent n in
the allometric equation is thus 3=4, or 0.75, and the human dose X
human
(expressed in mg) can be
calculated as follows:
X
human
¼ X
animal
 [W
human
=W
animal
]

0:75
(5:3)
The scaling factor [W
human
=W
animal
]
0.75
between a man weighing 70 kg and a rat weighing 250 g is
68.4 when correction for differences in body size is based on caloric requirement. Similarly, the
scaling factor between a man weighing 70 kg and a mouse weighing 35 g is 299. The corresponding
scaling factors obtained based on the body weight approach are 280 and 2000, respec tively.
ß 2007 by Taylor & Francis Group, LLC.
Thu s, the diff erence be tween the two extrapolat ion bases, W
1
and W
0.75
, for ‘‘ to man from rat ’’ and
for ‘‘to man from mous e ’’ is 4.1-fol d and 6.7-fold, respec tive ly, g reater for W
1
compa red to W
0.75
.
The refore, scaling by the body wei ght approac h p rovides equivalent human doses great er than
scaling by the caloric requi rement approac h. Henc e, scali ng by the body weight approac h wi thout a
biol ogically just ifi able reason may overes timate the equ ivalent human dose. As can be seen from the
two scaling facto rs d erived for rat and mous e, respec tively, the great er the diff erence in body weight
betwee n anim al and man, the great er the scaling factor; this is illust rated in Tab le 5.4.
When the observ ed d ose in an experiment al anim al study is expres sed in mg=kg body weight,
then the equiva lent human dose (in mg=kg body wei ght) is equal to the dose in the experi mental

anim al study divi ded by the scaling facto r accordi ng to the following equation:
X
human
¼ X
animal
=[W
human
=W
animal
]
0 :25
( 5:4)
Acco rding to this e quation, the scaling facto r [ W
human
=W
animal
]
0.25
is 4.1 ‘‘ to man (70 kg) from rat
(250 g) ’’ and 6.7 ‘‘ to man (70 kg) from mous e (35 g). ’’ This is illust rated by the foll owing examp le:
a NO AEL of 1 mg has been deriv ed from an experi mental study wi th rats. By assum ing a body
weight of 250 g for the rat, the NO AEL is 4 mg=kg b ody weight. According to Equ ation 5.3, the
equiva lent human NOA EL can be calcul ated to 68.4 mg based on a human body weight of 70 kg
(1 mg 3 [70 kg=0.25 kg]
0.75
), or 0.98 mg=kg body weight (68.4 mg=70 kg). Acco rding to
Equ ation 5.4, the eq uivalent human NOAEL can be calculated to 0.98 mg=kg body weight
(4 mg=kg 3 [70 kg=0.25 kg]
0.25
), i.e., the anim al d ose (in mg=kg bo dy weight) divided by the

scaling facto r, in this case 4.1. Se e also Table 5.3.
5.3.2. 4 Adju stment for Differences in Body Size: Exp osure Route
Acco rding to Feron et al. (1990) , simpli city is p robably the main reason for applying the caloric
requi rement approac h in extra polating inhalati on toxicit y data from anim als to human s. This method
is based on the assumpti on that (small) animals and human s breat he at a rate related to thei r need for
oxygen, thus automati cally at a rate dep ending on thei r caloric requi rement (energ y metaboli sm),
and thus, most imp ortantly, they are automati cally being exp osed to chemicals occurring in the
breathing atmosphere at a rate similar to that of the caloric requirement.
TABLE 5.4
Scalin g Facto rs for Adjusti ng the Dose Express ed Per Unit Body
Weight to the Dose Express ed Per Unit Caloric Requirement Taking
70 kg as the Body Weight for an Adul t Human
Animal Species Body Weight (kg) Scaling Factor
Rat 0.200 4.3
Rat 0.250 4.1
Rat 0.300 3.9
Mouse 0.025 7.3
Mouse 0.035 6.7
Mouse 0.050 6.1
Guinea pig 0.500 3.4
Dog 10 1.6
Dog 15 1.5
Note: When the dose for a given species is expressed in mg=kg body weight, the equivalent
human dose (in mg=kg body weight) is obtained by dividing the animal dose by the
scaling factor.
ß 2007 by Taylor & Francis Group, LLC.
In p ractice, this means that no adjustm ent for diff erence in body size is needed for a NO AEC
obtained for systemi c effect s in an inhalation toxicit y study (van Gend eren 19 88, Feron et al. 1990,
Vermei re et al. 1999, KEMI 2003). For example, a NOA EC of 50 mg=m
3

observ ed for laboratory
animals is also the equiva lent human NOA EC (note that so far species-sp eci fic sensi tivity has not
been taken into a ccount).
As ment ioned in Secti on 5.3 .2.3, extra polation using allomet ric scaling based on metaboli c rate
assumes that the parent c ompound is the toxi c agent and that the detoxi fi cation is relat ed to the
metaboli c rate and thus contr ols the tissue level . This is relev ant for oral exposur e only. With regard
to inhalation of subst ances, whi ch act syst emicall y, the lower detoxi fication (metabo lic) rate in
large r anim als is balanc ed b y a low er uptake (lower respirato ry rate) and thus no scaling facto r is
needed (ECETO C 2003).
For subst ances with local effect s on the respi ratory tract, no g eneral approac h for interspeci es
scaling c an be given. Anatom ical and physi ological differences in the airways be tween experi mental
animals and humans contr ibute to inte rspecies diff erences in local effects observ ed between animals
and hu mans, see Secti on 4.7.8. It shoul d be noted, howe ver, that for local effect s the deter mining
facto r for effect s to occur in the respirato ry tract is general ly the c oncentrati on of the chemi cal in the
air rathe r than the tota l dose and thus allomet ric scalin g is not relev ant.
5.3.2. 5 Adjustm ent for Differences in Body Size: PBPK Models
Extrapo lation betw een species shoul d ideal ly take into account metaboli c routes, i.e., the absence or
presen ce of metaboli tes, as wel l as the relative rate of formati on of the indi vidual metaboli tes. In
PBPK model s (Sec tion 4.3.6), both aspect s (nonlinear ity, form ation of active met abolites) are
incor porated. This modeling techni que uses compa rtments that correspond to actual tis sues or tissue
groups of the body. Size, blood flow, air flow, etc. are taken into ac count, in addition to speci fic
compo und-re lated parameter s such as partition coef ficients and met abolic rate data. Based o n such
studies, target-or gan concent rations of active metaboli tes can be predicted in experiment al animals
and human s, thus providing the best p ossible basis for extra polation (Feron et al. 1990).
Acco rding to Cl ewell et al. (2002a) , PBPK model ing provi des imp ortant capabi lities for
improvi ng the reliabilit y of the extrapolat ions across dose, speci es, and exposur e route that are
general ly required in che mical risk assessmen t regard less of the toxic endpoi nt being consi dered.
The author s have described an approac h, which provides a comm on tem plate for incorporat ing
pharm acokinetic modeling to estimat e tissue dosim etry (e.g., tis sue concentrati ons, body burdens ,
area under the curve, see Sec tion 4.3.5) into chemical risk assessmen t. They noted that chemi cal risk

asses sments typically depe nd up on co mparisons a cross speci es that often simpli fy to ratios re flect-
ing the diff erences , and have descri bed the uses of this ratio concept and discussed the advant ages of
a pharm acokine tic-based approac h as compa red to the use of default dosime try. Based on their
analys es, they con cluded that the correct relat ionship for cross-speci es dosim etry depends o n
whether the toxicity is due to the parent chemi cal or a met abolite, and in the case of toxi city from
a metaboli te, whet her the metab olite is highly react ive o r suf ficient ly stabl e to enter the circu lation.
Moreo ver, the nature of the cross -speci es relationsh ip for each of these possi bilities is different for
oral e xposure than for inhalati on. Therefor e, PBPK model ing is required to improve the reliabil ity of
cross -species extra polation that considers the natur e of the toxic entity. Thus, the avail ability
of infor mation on the parent compo und and its met abolism may allo w modi ficati on of the de fault
assessment factor.
5.3.3 REMAINING SPECIES-SPECIFIC DIFFERENCES
As mentioned earlier, the interspecies differences can be divided into differences in metabolic size
(Section 5 .3.2) an d remainin g species-sp eci fic differences . The average sensitivity of human s to the
adverse effects of chemicals (after scaling for caloric requirement) is comparable to that of other
species (KEMI 2003). However, an extra assessment factor is needed to account for the remaining
ß 2007 by Taylor & Francis Group, LLC.