New Approaches in the Derivation of Acceptable Daily Intake (ADI)

INTRODUCTION Current methods for estimating human health risks from exposure to threshold-acting toxicants in water or food. such as those established by the U.S. Environmental Protection Agency (U.S. EPA. 1~1; Stara tt ai., 19812,the Food and Drug Administration (FDA) (Kokoski. 1976),3the National Academy of Sciences(NAS. 1977).4 the World Health Organization (WHO. 1972)' and the Food and Agricultural Organization (FAO) (Bigwood. 19736; Venorazzi. 1976.'1980'; Lu. 19839).consider only chronic or lifetime exposure to individual chemicals. These methods generally estimate a single. constant daily intake rate which is low enough to be considered safe or acxeptable. This intake rate is termed the acceptable daily intake (ADI). Two problems with this approach have been recognized(Krewski ~tal., 1984).1°The first problem is that this method does not readily account for the number of animals used to detennine the appropriate "no-observed-effect level." (NOEL). For example. if a chemical has a NOEL based on 10 animals and a similar NOEL basedon 100animals. the risk assessorwill often choose the NOEL based on the larger study because it may appear to yield greater confidence in the resulting AD I. However. if these NOELs were for different chemicals. similar ADls might be derived even though one would be associated with less confidence. It would be useful c~ TD.ricDlofY 1_. Vol. 1. No.1. IIp. J.5-q

~.~

§

~

38

.ac ~

8

0' ~

.. .~

e> 4

~ ~

~

~ z~ ~ ~

cz ~c~, ~2.' o~o' ~:o!

c c , ,

I

f:>

r

,,

I , I I

,

~01

.,

I

,

~

'

I.

, ,~ ,

I,

I , I , I

,I ,I

. ~



~

0 ~

~

0 ...

0 ...

~

0:

,.. ~ 0

.

; . . -

Z C &.

.

~

-... ~

-~(oJ..

u

\J .:

J

~

>..-

Q,--

~

.~£;~= ~\J

",

'>'

.~-

(oJ

c =

~

= - -'- -"'---

>.;":~u= "'CJVc..,c

~"

->,.=>~ C 0

".

. ~

~ .,.. ;.;

=

O~CGOU'" ~ =?;~ "'\J~;':-': >. '" '" ~ '... GO -,.~==

O':;.:~X;,;

e

u

-:="",,;V ~-~~u

~A~8-u -O&.~

~,c=~-~

..~-=~:j ->.~

- u

c :.:'

~ c..>~GO.>.~:->~ ~

.~ ; _uc..

--- -uc "

_fJ ~.c~!Gc",~

-u-"

!G

~

GO=U ~

C"'~-

~>.c:=-

c \J~ >

~.,

u

- ~ '" . -----,c., -.~~J-S~.c-"- =e " --- -

~ C -(oJ-=~vu ~

'

~~-GO...~

=uu~GO~~ .~~~~~= Q~\J~-

=~~>f-~c .-=~

;II

~

e

Q

\I

~

u

CoI_~ ~

=

.

u . ~

",,=

, ~~~!GU - 10/

-

e~~~~"'~ = e -- e\J .-N'- U - -u

.£\J-->-_V U ---"'~ ,c,-.~

~,.-->.u~ =f.~,c'=1o/

o-~>.__!G

~~ 10/\1"Co. c

-."c~"~~> =

~

~

",,~ \I~'

",'

~

~

,c

--c.. - --

~~cr-=\J: .=-'" ~;: e

-

0 u .c--.-c~--~...;.- .~ ~

u

=

u

-

.c

..:

\J.\J

"'-

>-.:

'"

=~

~'" t

= ~--u.- C J(oJ -~---V ~O ==-e -='~.~ ~e

.~

~~_CoI o-?:c==V .

'G>'.c~~'G> ~~-

-

-

"

-

,~

--

':-

~

-

,,~ --

.-

~

c

~~ -

..

u

~ e _..~.~-

-~u.cV)~=

-", -\J --.-U . ~ U . w~~c~~~ ~._c,, ~ -

~.£

~ ~ c.. c..~ ou~=-=_J ->_":c~~

~

>-"

e~Q-e"

CI>~~y"

~~"CI~~

-

.::

U

..

Ct"-

>.

CI

CI

C .-

U::-

>. -"'0-0 .-

CI >.~

...

~C "' ! "'~Q -

.~

U

CI e

~W~O~~ ..' >

-

..~

..U

~

-

..

:

-

- "' c ,c C > .. '" = CI-~..~"'O~"'O'"

-r,...

u..C

U ~.--~,--

~C"Q..

.ClCI_!"M>-_~MO

U "CI>=C

.Po '~ e] "OuC"'~

's

.s6u~;u::£

..

-

~-

-~, >

V

-~-~

c

...

'A

"'

M

..

...

~

U

-

~z-=~~=u ..~ u -=' ~ -->...- ,-e~- e > ~-~C..>.O --~;':"'C CI ~= ..> ..",u=-

.-

-

C

:u C

."

-

...

-

-

~

-

~

CI

c'"

(of)

~ . >..~ Z

CiO

~~-"~;~

e .-

~

-

1--C" >~ "'C"'.OCl~ Z ~C-"'-W ~ .~ "= 'U C ..

]

...

CI~

.~

U-

."

= ;~~Q&.~.. C-CCI -~ ..C"-C~:U> .. e r,- '"CI ~'-.. ~C ::~~ ~ - ~. = . = --

.~

u

>=

-U..-Q

---V

::>

.-

>

..

"-"

U

...

..

'C'

~

CI ~

u

.. ..

..

.~

0

.:

>. .-

-.. Q,.

...

=

._M

.-

~~~

-..J

~ --~

CI

"'00'" ~"'.."'O

-==~~>~ - 0 ~ - g,.- ;.= >. ~ ~ > CI .. = ~~"'O CI" -U"~~~~ =~>'CI-=~C

_.:

-

--,c-0 ~ ~

..'-'

..r-C~~

-..'"

C"C~=,,Q~~"->"

0> -~~Q~ ~_U"

.-

.-

S .-

CI "=-~ ~

" >. ,c,cC

=C"-->~ ~..

~- U

~

~~~ - OC.."=e..,- ~ '" CI

Q .-

e ~"'>:':Q..~- ~

-..~~~~= -.. W.._Cl"~U..

>.

e< .=OU~CI e r- 'C

,cCC~""=

N~ -

~ ~

:I

O~";'~?U~ ->. 0 =""" ,,",c-~~~~

39

and its usefulness for risk assessment(with larger size denoting better quality or usefulness). . After graphic representation.of all available toxicity data. a smooth boundary line is estimated (in Figs. 1 and 2 the line has been fitted by eye) which represents for any given time the highest no-observed-advene-effect level (NOAEL) for which no lower adverseeffect level (AEL) is observed. Interpolation along this NOAEL curve can be performed to estimate the NOAEL for any desired partial-lifetime exposure. To obtain a corresponding acceptable intake, the estimated NOAEL is divided by an uncenainty factor. In Figs. 1 and 2 an uncenainty factor of 10 is used and accounts for the expected interhuman variability \0 the toxicity of a chemical (in lieu of chemicai.specific data). Both the choice of the highest TABLE I Various effect levels aDd tbeir definiciOllSused in filS. 1 and 2

AEL

.

NOAEL

0

NOEL

0

Frank-Effea Level. ThaI exposure level which produces unmistakable advenc effects, such as ineveniblc f\IDctjonaJimpairmeol or monalilY. al a stalistlcally or biololically siloificaol increase in frequency or severity betWeenan exposed popula. tion and its appropriale control. Advenc-EffeCl Level. ThaI exposure level al wbjch theR are stalisticaUy or biologically ~caal In. creases in frequency or severilY of adverse effects betWeenthe exposed populatioo and its appropriate conlrol. No-Obsc1'Ved-Advene-Effea level. ThaI exposure level al which theR are no statislically or biologically silJliflCanl increasesio frequency or severilY of ad. verse effeCtSbetWeen the exposed population and its appropNle conlrol. Effeas ~ produced at this level. bul they arc not considered to be adverse. No-Observed-Effea Level. That exposure level at w~ there ~ DOstaustically or biologically SIIm~nt increasesin frequency or seventy of effects betWeenthe exposed populauon and its appropriate control.

.Listed in order of decreasing severity. b

Adverv effectsare consideredas functionalimpainnentor pathologicallesIOns

wbich may affect the performance of the whole organism, or which reducc an organisms ability to resfM)ndto aD additioaal challenge (U.S. EPA. 1~).

40

NOAEL line (without lower AEu).and the suggesteduncenainty factor of 10 are consistentwith (and a logicaJextensionof) previously establishedprinciples of the U.S. EPA (1980), the FDA (Kokoski, 1976),and the NAS (1977, 19801')in the use of effect levels and uncenainty factors to estimateADIs. The basisof this methodis empiricalobservations.As examples, data in Figs. 1 and 2 summarizethe estimatedhuman equivalent doserates, exposuredurations,effect levels,studysuitability, and target organsfrom the toxicity data for methoxychlorand mirex, respectively.In theseexamples,the human equivalent dose rate (mgiday)is plotted versusthe equivalentexposureduration (years) on a log-log scaleusing dose per body surfacearea and fraction of lifespan. Upon closeinspectionof Figs. 1 and 2, cenain generalpatterns are evident. The most prominent is that frank-effectlevels(FELs) (indicated by solid triangles,seealsoTable I) declinein doserate as exposureduration (i.e., fractions of life span) increases.This pattern is expected.A dosewhich is an LD!Oin rats (and therefore a FEL) over a shon duration shouldlead to at least50% monality after longer durationssincethe longerperiod includesthe former. Note also the pattern of the solid circles (AEu) for a given duration. Solidcirclestend to lie betweenFEL andNOEL or NOAEL values(indicated by diamondsor open circles, respectively).This pattern should also be expectedsinceby definition AELs involve dosesassociatedwith adverseeffectsthat are not assevereasthose representedby FELs; whereasNOEls or NOAELs involve doses not associatedwith adverseeffects.111ispattern confinns the typical dose response,dose-effect,or dose-severityrelationship expected at any given duration. The pattern of NOAEL or NOEL valuesversusfraction of lifespanis rather well defined with methoxychlor and somewhatlessdefined for mirex. Another pattern that can be noted is that the maximum dose differencesfrom studyto study for any panicular effect level (i.e., all of the NOEls) at a given generalduration are often an order of 10or less.This smalldifferenceappearsremarkableconsidering that thesedata, althoughadjustedto representhumans,are from severaldifferent studiesand include a variety of experimentalanimals as well as humans.

41

The slopefor a toxicantis influencedby its mechanismof toxicity or its pharmacokineticsand. thus, the specifictrends indicatedby the methoxychloror mirex data may not be representativeof other chemicals.Zero or very SItIalJslopessuggestthat the NOAEL line dependsonly on the dose rate, not the duration. Negativeslopes may indicate a complex mechanismof toxicity. Someof the possible reasonsfor negativeslopesare: 1) bioaccumulation of the chemical or itS toxic metabolites; 2) accumulation of damage; 3) decrease in resistance to the toxic effects of the chemical as eXJX>SUfe continues (and the observed individuals grow older); or 4) multistage phenomena (whereby several organs or tiss'~es must be compromised before overt toxicity ensues).

MA THEMA nCAL MODEL APPROACH Traditionally, NOAELs have been defined for quantal endpoints which have non-zero backgroundincidencesby choosingan experimental dose level which does.not contribute to a statistically significantincreasein incidenceof adverseeffectswhen compared to a control group. In parallel, NOAELs have been defined for continuousdata by choosingan experimentaldoselevel which does not constitutea significantly different meanvalue for a parameter indicating an adverseeffect when comparedto a mean value for a control group. There are tWo limitations inherent in this approach.The first problem relates to the insensitivity of the current method for detennining NOELs that use different numbersof animals.0110vs. 0/100.For example.a dose-relatedtrend in a parametermay suggest a deviation from the control incidenceor mean value at an intermediatedoselevel(s)which is not statisticallysignificant.This dosewould be treated as a NOAEL. A statisticallynonsignificant responsecould have biological significanceespeciallywhen experimentalgroupsare limited to smallsamplesizesand conclusions are extrapolatedto larger populations. The secondlimitation is related to the generallack of useof the slope of the dose-responsecurve. For example. the responsein-

42

cidenceor meanparametermeasurementis expressedasthe presence or absenceof a statistically significant effect at discrete intervals (i.e., the experimentaldoses).The probability of response at a dose level betWeena lowest-observed-adverse-effect level (LOAEL) and a NOAEL is not addressed.Especiallyif dosesare widely spaced.this could lead to considerableunderestimationof the thresholddose. The approachsuggestedhereis not assubjectto theselimitations becauseit usesmore of the dose-responseor dose-effectcurve. For example. an AD I might be calculatedfrom a dose-response curve by defining an adverseeffect as a risk level of more than a certain percentageabove background.such as 10%. In this presentation 10% is chosenbecausemanyof the mathematicalmodels in current useagreewell at estimatedrisksin this rangeandbecause the better studies have sufficient numbersof dosesand animals per doseto measurethis level directly. The lower 95% confidence limit (CL) on the doseassociatedwith this risk is then calculated. To obtain an ADI. the dose associatedwith this lower 95% CL might be reducedby a chemical-specific.speciesadjustmentfactor. or as in the caseof Fig. 3 the cube root of the animal to human body weightratio. Uncertaintyfactorsmight then be usedto divide this adjustedvalue to yield the ADI. In this review. uncenainty factors range betWeen10 and 100. The first uncert!inty factor of 10 is interpreted as accountingfor the expectedvariability in the general human population to the toxicity of the chemical.This uncertaintYfactor is consistentwith previous U.S. EPA guidelines(U.S. EPA. 1980)as well as other guidelines[e.g.. FDA (Kokoski. 1976);WHO (Vettorazzi. 1980); NAS, 1977. 1980].The seconduncertainty factor betWeen1 and 10 is thought to be necessarybecausethe adjusted 95% CL correspondingto 10% responserepresentsa LOAEL rather than a NOAEL. The use of this variable uncertainty factor is also consistentwith previousguidelines(U .5. EPA. 1980).In this example. the choice for the value of this variable factor should dependon both the severity of the adverseeffect (i.e.. more severeeffects yield a larger factor; U.S. EPA. 1980)and the slope of the doseresponse.or dose-effectcurve (i.e.. shallower slopesalso yield a larger factor). For example.a choicefor this variable uncertainty factor of 1.0 should be associatedwith both a minimal adverse effect and a steepdose-responseor dose-effectcurve.

43

44

An exampleof this procedureis given in Fig. 3 which is a hypothetical plot of the percentageof fats respondingwith a slight body weight decreaseof 5% versusdoserate or the percentageof dogs with liver necrosisvenus dose rate. Hypothetical responses are indicated by solid lines; lower 95% ConfidenceLimits (Cls) on the doserate are shown as dashedlines. The lower 95% Cls of the dose rates at a 10% responseare adjusted by division by th: cube root of the ratio of body weight betweenhumansand rats or dogs. For rats of 400 g weight, this value is 5.6; for dogs of 10 kg weight, it is 1.9; both calculationsassumea 70 kg body weight. To estimatean ADI from the rat data (shownin Fig. 3 as ADIR). the adjusted lower 95% CL is divided by a 10-fold uncertainty factor to accountfor the expectedvariability in the general human population to the toxicity of a chemical in lieu of specific data. and an additional 1.O-foldfaCtorbecausethe effect is both minimally severeand hasa steepdose-response slope.Thus, the total uncertainty factor is 10. To estimate an ADI from the dog data (shownin Fig. 3 as ADIo), the adjustedlower 95% CL is divided by a 10-fold uncertainty factor to accountfor the expected humanvariability, as before, and an additional10-foid uncertainty factor~cause the effeCtis both more severethan a slight body weight decreaseand the slope of the dose-responseis shallower. Thus, the total uncertainty factor is 100. DISCUSSION The primary advantageof the graphic method is that it provides a mechanismfor viewing all of the data simultaneouslyresulting in an integratedprofile of the toxicity of a compound.In addition. exposureduration-responsetrends, if present,are clearly delineated providing a possiblestrategyfor estimatingacceptableintakes for partial lifetime exposures. The graphicalmethodrelieson a simpl~severityranking system for data presentation (i.e.. NOEL, NOAEL, AEL, and FEL). Obviously with sucha simple system,effects within a given category (e.g., all AELs) may not be identical nor is it assumedthat they are. Indeed, the critical toxic effect is often a function of exposureduration. In thesecasesthe effectswithin a givencategory

4S

46

with this"approach, have been discussed(Dourson eotaI., 198618; Crump, 198619).For example, with this new approach both the slope of the dose-response curve andthe number of animalsused in an experimentcanaffectthe estimationof the ADI whenquantal or continuous toxicity d3ta are available. Another advantageof this method is that it can also estimatethe health risk for suprathreshold exposurelevels which might be useful for cost-benefit analysis.Severaldifficulties include finding appropriatedata setS to model, choosingamong equally good data setSthat may yield different ADIs and, for cost-benefitanalysis,assumingthat a certain percentagereSy'QDse in an animal stUdy is equivalent to a similar percentageresponsein humans. In summary,the methodsdescribedfor estimatingADls utilize more of the availabletoxicity datathan the current methodologies, and offer a consistentapproachfor possiblyestimatinghealth risks for less than lifetime toxicant exposureand perhapshealth risks abovethe ADIs. They also addressseveralof the criticismsof the current approachsuchasuseof dose-response slopesand the number of animals tested in defining NOELs. More work is needed, however, before either or both of thesemethodsare acceptedas the StQlusquo. Act...

...11"--R8I

The author wishesto acknowledgethe work of R. C. HertZbera.R. Hartung. K. Blackburnand K. S. Crwnp that precededthis text (Doursonet Gl., 1985;Crwnp. 1984),and alsothe manyhelpful discussions of J. F. Stan duringthe prepantion of this tat.

~-I.=-=r

Altboup the research(or other work) describedin this anicle hasbeen funded wboUyor in pan by the United StatesEnvironmentalProtectionAgency.it has not beensubjectedto the Agency'srequiredpeer and administrativereviewand. therefore. does not be necessarily and no official endorsement should inferred.reflect the view of. tbe AgeJICY

MICHAEL L. DOURSON U.S. EnviroNMnt4l

Prot«tion A~. 26 W. St. C/Qi, S~. CiIIdMGO". 011;0 45268

47

Ref'ennca 1. U.S. EPA. Guidelines aDd methodology used in the preparation of health effects assessment chapters of tbe.coasent decree water quality critena. Federal Register. 45. 79347- 79357 (1980). 2. J. F. Stara. M. L. Dounon aDd C. T. DeRosa. Water quality criteria: Methodology and a~tions. in Con! Proc.: EnviroliMf'n,al Risk Ass~ssm~n': How New R~,uIQ/ioIU Will Aff~cl tJv UtiJiry /lUiwrry (EJecu-ic Power Research Institute. Palo Alto. CA. 1981).

3. C. J. Kokosk.i. Written testimony of Charles J. Kokoski. Docket No. 76N. 0070. DHEW. Food aDd Drug Administration (1976). "'4. NAS (N~tionaJ Academy of Sciences). D"IIkin, W~,. and H~al,h (Washington. DC. 1m). "5. WHO (World HeaJth Organization). EvaJuation of Certain Food Additives and the Contaminants Mercury. Lead. and Cadmium. WHO Technical Repon Series No. 50S. Geneva. Switzerland. pp. 9-11 (1972). --6. E. J. Bigwood. CRC Crit. Rev. Toxicol.. pp. 41-93 June (1973). - 7. G. Vettorazzi. Safety factors and their application in the toxicological evalu-

.8. 9. 10. vII.

ation. in Th~ EvalUQ/iOllof Toxicological Data fo,. ,h~ Pro'~cnon of Public H~aldr (Perpmon. Oxford. 1976). pp. 207-223. G. Vettorazzi. Handbook of /num4IionaJ Food Rerul4lory Toxicology. Vol. I: EvaJuations(Spec:trum. New York. 1980). pp. 66-68. F. C. Lu. Regulatory ToKicoiogy and PharmKOlogy 3.121-132 (1983). D. Krewski. Charles Brown and DWKan Murdoch. Fundam. Appj. Toxicol. 4. 5383-5394 (1986). U.S. EPA. Appro8Cbcs to risk 8essmeDt for multiple chemical exposure.

EaviroamentaJ Criteriaaad...~t

Office.('".uw;i~ti.OH,EPA~

84-(XM (1984&) . "12. U.S. EP A, Selected approaches to risk asseSImeDt (or multiple cbemicaJ ex~,EDYiroameDtai Criteria aDd A~t Office, r-~~-!'~ati, OH. EPA~14a (1984b).

-{'3. J. F. Stare. R. C. Henzbera. R. J. F. BruiDS. M. L. Dourson, P. R. Durkin. L. S. Erdreid1 and W. E. Pepelko. Approaches to risk assessmentofchemicaJ mixtures. in CMmical Saf~ty ReruJ4lion and CompljQnc~.eds. F. Homburger and J. K. ~arquis. Cambridge. MA (S. Karger Ag. Basil. Switzerland. 1985a). "14. J. F. Stare. R. J. F. Bruins. M. L. Dounon. L. S. Erdreicb and R. C. Hertzberg. Risk assessmentis a developing science: Approaches to improve evaluation of sinpe chemicals and chemical mixtures. in Wo,ks#topOil M~I#todsfo, Assnsing 1M Effects of MizlUm of Ch~ic4/ (GuiJdford. Surrey England. 1985b). ]5. K. S. Crump. Fundam. Appl. Toxicol. 4.854-871 (1984). 16. M. L. Doursoo. R. C. Hertzberg. R. Hartung aDd K. Blackburn. Toxicol. Ind. HeaJth 4. 23-33 (1985). "17. NAS (Naoonal Academy of Sciences). Drink.in, Watv and H~aldr. Volume 3 (National Academy Press. Washington. DC. 1980). pp. 29-37. "18. M. L. Dourson. R. C. Hertzberg and J. F. Stara. Fundam. Appi. Toxicol. 6. 182-183 (1986). '19. K. S. Crump. Fundam. Appt. Toxicol. ,. 183-184 (1986).

48