Public Stakeholder Workshop
to Inform EPA's Upcoming IRIS

Toxicological Review
of Inorganic Arsenic

SESSION 3:

Dose-Response

Tuesday, January 8 &
Wednesday, January 9
RTP, North Carolina

Hosted by EPA's National Center
for Environmental Assessment


-------
DO NOT CITE OR QUOTE

Dose-Response

Weihsueh Chiu

Dose-Response Assessment =
Smoothly Meshii^Three Components

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

1


-------
DO NOT CITE OR QUOTE

Dose-Response Challenges for iAs

Modeling & i
Extrapolation
Approaches

8, nntrwitc

Multiple risk management
objectives.

Multiple methods used for
risk characterization.
Multiple exposure
pathways & scenarios.
Need to address
susceptibility

Dose-Response Challenges for iAs

Both experimental and
epidemiologic studies.
Multiple different
measures of exposure.
Multiple health effects and
endpoints.

Multiple covariates that
may affect susceptibility.

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

2


-------
DO NOT CITE OR QUOTE

Dose-Response Challenges for iAs

Modeling & i
Extrapolation
Approaches
& Outputs -

•	Multiple analyses of
diverse datasets.

•	Multiple extrapolations
necessary to meet risk
management needs.

•	Schedule may limit
development of new

pproaches.

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

3


-------
DO NOT CITE OR QUOTE

IS ARSENIC AN
APHRODISIAC?

The sociochemistry of an element
William R Cullen
Royal Society of Chemistry

arsenous acid
As(lll)

arsenic acid
As(V)

OH

i

HO—As—OH

HO—As—OH

in

monomethylarsonic acid O
MM A	II

CH,-As-OH

I

dimethylarsinIc acid
DMA

arsenobetaine
AsB

OH

0

ii

CH,—As—OH

1

CH,

CH.—.

tetramethylarsonium ion
TETRA

ch3 9"

1 . A>

As-—' u

I

CH3
CH,

CH —

As—CH,

in,

"SIMPLE" ARSENICALS IN THE
ENVIRONMENT

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

4


-------
DO NOT CITE OR QUOTE

Arsenosugars AsS	Thioarseriosugars thioAsS

0	As=S replaces As=0

R

cL



H OH

R= V Y^OH	AsS-OH	thioAsS-OH

jH

ASS-P04	thioAsS-P04

OH OH OH

SO,H	AsS-S03

AsS-S04

thioAsS-S03

thioAsS-S04

OH 0



OH

1 CH3® 11

2e

1

HO-As: 	CH3— As — OH

	> ch3

— As:

OH OH



OH

0

OH



CH3® II 2e

1

CH®

	^ CH3—As—OH 	>

CH3—As: -

	-v

ch3

ch3



0





II 2e

* •



CH3-As-CH3 	> ch3-

As — CH3



X

-o

ch3



Challenger





Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicolbgicail Review of Inorganic Arsenic

5


-------
DO NOT CITE OR QUOTE

^-SH

SH

t

MeAs V
As V

As III

S- adenosylhomocysteine

S- adenosylmethionine

Lipoic acid

Reductive Elimination

,SR

Me3As(V)0 + 2RSH -> Me3As(V) -> Me3As(lll) + RS-SR

^SR

Reglinski 1984

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

6


-------
DO NOT CITE OR QUOTE

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

7


-------
DO NOT CITE OR QUOTE

Identifying Factors Relevant to
Dose Response

3.1 What types of exposure could contribute to the
aggregate dose, and in what ways might this
impact how an iAs dose-response
characterization is used/applied? How can we
estimate impact of drinking water exposure
alone vs. aggregate exposure on possible effects
of iAs exposure?

Lead Discussants: Karen Bradham, Bill Mendez	

3.1. What types of exposure could contribute to
the aggregate dose, and in what ways might this
impact how an iAs dose-response
characterization is used/applied? How can we
estimate impact of drinking water exposure
alone vs. aggregate exposure on possible effects
of iAs exposure?

Bill Mendez

Inorganic Arsenic Public Stakeholder
Workshop

January 8-9, 2013

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

8


-------
DO NOT CITE OR QUOTE

What types of exposure could contribute to
the aggregate dose...

•	Dietary

-	Food contaminated by arsenic from soil/water

•	Contaminated Soil/Dust

-	Inhalation

-	Ingestion

•	Severity of As contamination varies widely

-	Many natural and man-made sources

...and in what ways might this impact how an iAs
dose-response characterization is used/applied?

•	Adverse effects in epidemiological studies usually
reported as function of single exposure medium
concentration

•	Failure to account for other exposures can:

-	Bias magnitude/form of dose-response

-	Effect statistical significance of relationship

-	Serious problem if "background" dose is a
large fraction of dose from primary exposure
medium

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

9


-------
DO NOT CITE OR QUOTE

How can we estimate impact of drinking water
exposure alone vs. aggregate exposure on
possible effects of iAs exposure?

•	Need to account for water and non-water As

-	Estimate As dose from primary medium exposures and
exposure factors

-	Add in dose from other media

• Varies by region, proximity to point sources

-	Consider speciation, bioavailability, absorption

-	Fit dose- (instead of exposure-) response model to
estimate risk

•	Need to consider As intake of target
population (U.S. general public)

•	Account for uncertainty through sensitivity
analysis, simulation

Speciation/Bioavailability Assessment for Rice

Arsenic Species in Uncooked Rice (Juhasz et al. 2006)

800

¦	Inorganic As

¦	"Organic" As

¦	"Other" As

mi

•ft

^	^	^	J?

\*>	if	^

Lull

/ * Z ~ / J? ^ f f S *

*	J?	J?

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

10


-------
DO NOT CITE OR QUOTE

1,300
1,170
1,040

Dietary and Water Arsenic Intake,
Bangladeshi Women (Kile et al. 2007)

¦ Drinking water As
~ Dietary As











cc





¦a 780





CO
c





(0

a

390
260

Ml Jill











<1 | 1-10 | 11-50 | 51-100 | 101-200 | >200
Tube well As concentration (ng/L)

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

11


-------
DO NOT CITE OR QUOTE

Bioavailability of Inorganic Arsenic

Workshop on the Integrated Risk Information System (IRIS)
Toxicological Review of Inorganic Arsenic

Presented by Dr, Karen Bradham

U.S. Environmental Protection Agency
Office of Research and Development

Potential for exposure to toxicants at contaminated sites

U.S. Population within Four-Mile Buffers of Superfund Sites

| Supedund
I with 4-mie buffet

Center for Icvsciturkmil Eirdi
Seimcc InfwmMKXi Network

2QQ0 U.S Consul
Sf 1 Population

High 4OT&8



Sourc** EPA 2008 A AT5DR 1 (AvalWM* »t
M MtpjVsedac.cwsin .colurrfiModu.Wifctpop MfW>,
U.S. Census. ESRI

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

12


-------
DO NOT CITE OR QUOTE

2007 Priority list of hazardous substances

1	ARSENIC

2	LEAD

3	MERCURY

4	VINYL CHLORIDE

5	POLYCHLORINATED BIPHENYLS

6	BENZENE

7	CADMIUM

8	POLYCYCLIC AROMATIC HYDROCARBONS

9	BENZO(A)PYRENE

10	BENZO(B)FLUORANTHENE

FREQUENTLY OCCURRING AT NFL SITES
TOXICITY POTENTIAL FOR HUMAN EXPOSURE

Arsenic exposure at contaminated sites

•	Oral ingestion of soil and dust-"risk driver" for
human exposure

•	Conventional methods or default values - do not
adequately address metal bioavailability under site
conditions

•	Bioavailability of metals in soils and dusts vary
depending on the mineralogy and physicochemical
properties

•	Default assumption for assessing risk from arsenic in
soil is that the bioavailability of arsenic in soil is the
same as the bioavailability of arsenic dissolved in
water

• Recent bioavailability studies conducted in animal
models show that the bioavailability of arsenic in soil
is typically less than that of highly water soluble
forms of arsenic (< 100%)

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicolbgicail Review of Inorganic Arsenic

13


-------
DO NOT CITE OR QUOTE

EPA bioavailability guidance

•	EPA's Technical Review Workgroup, Bioavailability Committee
recently conducted a review of all available in vivo estimates of soil
arsenic bioavailability

•	103 studies identified, most studies had values < 60% and studies
were performed to support remedial investigations and risk
assessments of specific sites

•	OSWER Directive 9200.1-113:

-	Based upon evaluation of current data sets of arsenic bioavailability, the
upper percentile of the data set results in a default value of 60%

-	The default value for arsenic in soils should only be used if site-specific
assessments for arsenic are not feasible

•	EPA's Guidance for Evaluating the Bioavailability of Metals in Soils
for Use in Human Health Risk Assessment

http://www.epa.aov/superfund/bioavailabilitv/auidance.htm

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicolbgicail Review of Inorganic Arsenic

14


-------
DO NOT CITE OR QUOTE

What kinds of dose-response characterization may

be needed (e.g., reference value, incremental
change in risk with dose, probabilistic risk at dose)
for aggregate (e.g., urine, blood) and source-
specific (e.g., food, water) dose metrics?

Weihsueh Chiu

3.3 Dose-Response Characterization:

What kind(s) of output needed (for each endpoint)?

What can the
available

data
support?

Dose-Response
Outputs

Point of departure
(e.g., Benchmark Dose)

Exposure limit (e.g.,
RfD)

Changes in continuous

parameters as a
function of exposure

Incidence of effects as
a function of exposure

Risk

Characterization
Method

Different
regulatory
programs
have
different risk
management
needs

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

20


-------
DO NOT CITE OR QUOTE

3.3 Dose-Response Characterization:

What metric(s) of dose/exposure are needed?

What can the
available data
support?

v	y

Can we
extrapolate
across dose
metrics?

Source specific

Exposure
Durations
&

Life stages

Depends on
how each
regulatory
program
estimates
exposure.

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

21


-------
DO NOT CITE OR QUOTE

Approaches to Dose-Response Analysis

3.4 What kinds of approaches are available to
analyze dose-response data (e.g., statistical
models, non-parametric approaches)?

Lead Discussants: Weihsueh Chiu, Jeff Gift

What kinds of approaches are available to
analyze dose-response data (e.g., statistical
models, non-parametric approaches)?

Multitumor Analysis

Jeff Gift
(January 9, 2013)

Benchmark Dose Training California Environmental Protection Agency Training Workshop	128

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

22


-------
DO NOT CITE OR QUOTE

3ERA

Analyzing Dose-response Data

0.8

0.6

0.4

0.2

BMDL

BMD

50

100
dose

150

200

14:40 01/25 2007

Benchmark Dose Training California Environmental Protection Agency Training Workshop

Possible Topics

IMDL?

Parajfrflric or semi para metric modeling?
Prob^^fetic risk?

Mod^^n certainty?

OtheQ^rametric models?

Othe^^thods for estimating BMDLs for an endpoint
Multitumor analysis

Benchmark Dose Training California Environmental Protection Agency Training Workshop

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

23


-------
DO NOT CITE OR QUOTE

3ERA

Multiple Tumor Analysis

•	Basing unit risk on one tumor type may
underestimate cancer risk of a chemical that
induces neoplasia at multiple sites (NRC, 1994).

•	Tumor selection and derivation of confidence

limits are key aspects of a multitumor analysis.

•	Tumors must be independent of one another.

MOA for arsenic related tumor formation not known
No reason to assume tumors not formed independently

•	Low risk tumors combined with high risk tumors

For example, though skin cancer is rarely fatal and less
potent than internal tumors, its risk can theoretically be
accounted for in a combined risk analysis.

Selection of Tumors

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

24


-------
DO NOT CITE OR QUOTE

Confidence Limits for Combined

Risk

LED (BMDL) from BEIR IV model is not 95% LB on dose.

"Normally distributed uncertainties" assumption used to get
LB for combined risk has been criticized in peer review.

Two alternative approaches are used in the EPR draft of
the EPA 1,4-Dioxane assessment:

Markov Chain Monte Caro (MCMC)/Bayesian computational
approach (Kopylev et al., 2009)

BMDS multitumor (MS_combo) profile likelihood approach
(epa.gov/ncea/bmds)

These approaches were well received by peer reviewers,
but do not account for time/age dependency.

Benchmark Dose Training California Environmental Protection Agency Training Workshop

uose-response modeling results
for male rat tumors that inhaled
1,4-dioxane for 2 years

Rat Exposure Human Equivalent

Tumor Type

Nasal squamous cell
carcinoma

Hepatocellular adenoma
or carcinoma
Renal cell carcinoma
Peritoneal mesothelioma
Mammary gland
fibroadenoma
Zymbal gland adenoma
Subcutis fibroma
Bayesian Total Tumor Analysis
BMDS Multitumor (MS_Combo)

Multistage
Model
Degree

1

Inhalation

(ppm)

(mg/m3)

Unit









Risk

BMC10

BMCL10

BMC10

BMCL10

(M9/m3)-1

1107

629.9

712.3

405.3

2.5 x 10"7

252.8

182.3

162.7

117.3

o

X
LO
CO

1355

1016

872

653.7

o

X
LO

82.21

64.38

52.89

41.42

2.4 x 10"6

1635

703.0

1052

452.4

2.2 x 10"7

1355

1016

872

653.7

o

X
LO

141.8

81.91

91.21

52.70

(D

X

O

CT>

39.2

31.4

25.2

20.2

5.0 x 10"6

40.5

32.3

26.1

20.8

4.8 x 10"6

Benchmark Dose Training California Environmental Protection Agency Training Workshop

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

25


-------
DO NOT CITE OR QUOTE

3ERA

Questions

•	Should all "confirmed" As-related tumors (lung, bladder,
kidney and skin) be included in a multitumor analysis?

•	If so, how should lower risk cancers (e.g., skin) be
"weighted" relative to the other cancer risks?

•	Should a time dependent multitumor modeling approach
be developed that can calculate more defensible
confidence limits?

• Most recent approach proposed for As:

"Upper confidence limits on the combined cancer risks
can be calculated based in the assumption that the
uncertainties in the two CSFs are both normally
distributed. If this is the case, the 95% upper bound, U,
for the combined cancer potency can be calculated as:

U = (m1 + m2) + (u 1 - m 1) + (u2 ~ m2) (Equation 5-5)

where mi and ui, i = 1,2, are respectively mean and 95%
upper bound cancer potency for the two tumor types."

Confidence Limits for Combined

Risk

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

26


-------
DO NOT CITE OR QUOTE

Approaches to Dose-Response Analysis

3.5 What are factors (e.g., toxicokinetics,
bioavailability, water consumption rates,
background exposure, susceptibility) that can
impact the dose response analysis, and how
could these factors be transparently accounted
for?

Lead Discussants: William Cullen, Hisham El-Masri

* 3*

What are factors (e.g., toxicokinetics,
bioavailability, water consumption rates,
background exposure, susceptibility) that can
impact the dose-response analysis, and how
could these factors be transparently
accounted for?

Hisham El-Masri

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

27


-------
DO NOT CITE OR QUOTE

Exposure-Dose-Response Paradigm

Exposure

Nonavailability

Interna! Dose

\

Biologically Effective Dose
\

Early Biological Effects
\

Susceptibility

Altered Function/Structure

Susceptibility

\

Clinical Disease

Modified from Schuite, 1989	Prognostic Significance

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

28


-------
DO NOT CITE OR QUOTE

IAsv

a)

iAs

~

Induce chromosomal aberrati
(4), genetic instability (5).

Induce alterations in methylation
patterns (6).

Generate reactive oxygen
species (7) and 8-oxo-dG
adducts (8).

Interfere with DNA repair (9).

Induce p53 (10) and cell
proliferation (11).

Mouse carcinogen (12) and co-
carcinogen (13).

Inhibit DNA repair (9). Non
turnorigenie to mice and rats
(14).

MMAsv

0)

MMAs'"

\ ra .

DMAs1'

Induce chromosomal aberrations
and DNA breaks (15).

Generate reactive oxygen species
(16) and 8-oxo-dG adducts (17).
Induce cell proliferation (18).
Inhibit DNA repair (9).

Methylation
Reduction '

Induce DNA damage (19) and 8-oxo-
dG adducts (20).

Induce p53 (10) and cell proliferation
(11).

Rat bladder carcinogen (20) and rat
bladder tumor promoter (21).

(2)

Induces 8-oxo-dG adducts (17).
Rat liver carcinogen (24).

/

DMAs

III

Induce chromosomal aberrations and DNA
breaks (15,22)

Generate reactive oxygen species (16)

Inhibit DNA repair (9)

Induce p53 (10) and cell proliferation (23).

(2)

TMAsv

(3)

TMAsH")

Accumulation of Arsenicals Varies
Significantly Across Tissues

IAs

MMA

DMA

O 4 -

O

Blood Liver Lung Kidney Bladder Skin
Tissue

Female C57BI6 Mice -12 week drinking water exposure to As(V)

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

29


-------
DO NOT CITE OR QUOTE

What Makes Arsenic Unique?

•	Pancarcinogenic in humans, whereas rodents
are much less responsive

•	Large cross-species differences in metabolism

•	Tissue-specific differences in metabolite
accumulation

•	Toxicity most likely mediated by metabolism

•	Known variations in metabolism due to age and
ethnicity in humans

•	Polymorphisms identified in AS3MT, the
principal As metabolizing enzyme

Factors Impact on Dose-Response

•	TK

—	Linkage to a hypothesized mode of action to identify key
target tissue dosimetry

•	HUMAN VS ANIMAL

•	Epidemiological studies (human only)

—	Exposure (speciation of organic vs inorganic As, DMA,
M MA..etc.)

•	Is As dose-response a mixture problem?

—	Susceptible populations

—	Can biomarkers (e.g. levels of DMA in urine) be
misleading?

•	Need of appropriate target tissue levels (blood levels of affected
population at minimum)

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

30


-------
DO NOT CITE OR QUOTE

Approaches to Dose-Response Analysis

3.6 EPA has traditionally addressed uncertainty in
modeling dose-response data by using a
statistical lower confidence bound on the
benchmark dose. What other approaches are
available to address and transparently convey
the impact of uncertainty on the dose-
response analysis?

Lead Discussants: Bill Mendez, Warner North

3.6

"EPA has traditionally addressed
uncertainty in modeling dose-response
data by using a statistical lower
confidence bound on the benchmark dose
(BMD). What other approaches are
available to address and transparently
convey the impact of uncertainty on the
dose-response analysis?"

D. Warner North

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

31


-------
DO NOT CITE OR QUOTE

Comments for the Arsenic/IRIS Workshop Jan 8-9, 2013

D. Warner North

President and Principal Scientist, NorthWorks, Inc.
E-mail: northworks@mindsprinq.com

Web: www.northworks.net

3.6:

"EPA has traditionally addressed
uncertainty in modeling dose-response
data by using a statistical lower
confidence bound on the benchmark dose
(BMD). What other approaches are
available to address and transparently
convey the impact of uncertainty on the
dose-response analysis?"

Section 1.3 North comments

•	Inorganic arsenic is unusual as an IRIS entry.

•	Not all risk assessments should be done the
same way - for IRIS, or in other contexts.
(Reference: the NAS "Color" books)

•	Focus should be on health risk at potential
low-dose human exposure.

-	Is the health risk potentially significant?

-	If so, want quantitative estimate(s) and uncertainty
disclosure. Key aspect: dose-response relationship,
built upon evidence from epidemiology, mode of
action/toxicology investigations.

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

32


-------
DO NOT CITE OR QUOTE

Section 1.3 North comments, 2

• For important cases/IRIS entries, may want more than
a review of published papers.

-	Convene a gathering of the best experts for
discussion and debate. Publish the proceedings.

-	Frame the problem first. What is included?

-	Want a transparent process, understandable by
stakeholders.

-	Don't preclude evidence: Assemble it, then evaluate
it.

Reference: Public Participation in Environmental
Assessment and Decision Making, National Academy
Press, 2008, See esp. Chapter 6, "Practice:

Integrating Science," particularly page 141 on defaults
and guidelines. (Page 141 to be handed out). On web
at www.nap.edu/catalogue.php7record id=12434.

Dose-Response with uncertainty

Goal: Summarize the information available,

disclosing uncertainty.

Consider three levels : (response y,

given dose x)

1.	A range (e.g., zero to a plausible upper
bound)

2.	A probability distribution over the range

3.	A model that computes a probability
distribution based on inputs, permitting
sensitivity analysis to these inputs.

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

33


-------
DO NOT CITE OR QUOTE

Section 3.6: D-R with uncertainty

Goal: Summarize the information available,

disclosing uncertainty.

Consider three levels : (response y

given dose x)

1.	A range (e.g., zero to a plausible upper
bound - old pre-BMD, EPA cancer risk
number-see page 141 handout)

2.	A probability distribution over the range

3.	A model that computes a probability
distribution based on inputs, permitting
sensitivity analysis to these inputs.

Section 3.6: D-R with Uncertainty

Goal: Summarize the information available,
disclosing uncertainty.

Consider three levels : (response y

given dose x)

1.	A range (e.g., zero to a plausible upper
bound)

2.	A probability distribution over the range

3.	A model that computes a probability
distribution based on inputs, permitting
sensitivity analysis to these inputs.

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

34


-------
DO NOT CITE OR QUOTE

Example for Level 3:
(ecological risk, introduced
species)

Reference:

D. W. North, "Limitations, definitions, principles, and
methods of risk analysis, Rev. sci. tech. Off. Int.
Epiz.14(4), 913-923, 1995

Available at: http://www.northworks.net/limitations.pdf

Microbes

Do we infect Mars with
terrestrial microbes by landing
a spacecraft on its surface?

(Decision facing NASA in 1972 for
billion-dollar Project Viking Mission,
first surface landing)

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

35


-------
DO NOT CITE OR QUOTE

Planetary Environmental Protection

(an example of risk assessment
with almost no "data")

Mars from Viking Orbiter Viking Lander, 1976

Mission Contamination Model

BIO-BURDEN



RELEASE



TRANSPORT



REPRODUCTION



External



ImpJarw





RESISTANCE







l ANDl.NO MOOe











Covered



USABLE WATCH

Reach
Usable

TO

ENVIRONMENT



RtCONI AMfcNAT IOM



REiEASt

Efotion



Water



Growth

Mated

MECHANISM



MECHANISM







•IIOEXPSHIMENT

Encap-
sulated

RELEASE

Vibration

TRANSPORT



NUTRIENTS



LETHALITY



























m

PROBABILITY \

" )
, CON T AMI NAT SON /



Bioexperlment Contamination

Mission scenarios, fate and transport of microbes on lander


-------
DO NOT CITE OR QUOTE

Bio-burden Submodel
(Microbes on Spacecraft)

Input Elements:

•	Pre-sterilization burden by location

•	Sensitivity to sterilization

•	Sterilization regime

•	Recontamination

•	Inflight mortality or proliferation

•	Contamination and subsequent
amplification in biology experiment:

(probability = 10"6)

Outputs:

•	Bio-burden estimates by Viking Project

•	Expected number of Viable Terrestrial
Organisms (VTOs), by location on
spacecraft

BIO-BURDEN

SiHStflVt

11

External

VARIABLES
• EltXlMl.

16
Covered

WKaiWjUM

Mo-buf
-------
DO NOT CITE OR QUOTE

Transport Submodel
(How Many Get to "Usable Water"?)

Input Elements:

•	Expected number of VTOs released, by
mechanism

•	Lethality of UV radiation, in normal
atmosphere and dust storm (probability that
a VTO survives transit)

•	Extent of usable water

-	Probability it exists anywhere

-	Portion of surface covered

Output

•	Expected number of VTOs that will reach
usable water

0.045
Implan-
tation

14.9
Erosion

0.11
Vibration

TRANSPORT

sensitive

VARIABLES

~ Availability ol

usable v*a\tt
» Drgr«c ol UV
shielding

i

1.2 x 10"3!
Beach
Usable !
Water
	~

Reproduction Submodel
(Can at Least One Do It?)

Input Elements:

•	Fraction of VTOs that are
facultatively anaerobic and
psychrophilic: (0.05)

•	Probability that nutrients needed for
reproduction will be present in the
water microenvironment (0.10)

Output:

•	Expected number of VTOs that
reproduce at least once, defined as
"contamination"

HE PRODUCT ION



SENSITIVE



1.2 * Iff3

VARIABLES



Reach



6 x 10"

Usable

• PllCafll

Water

ptyctirophilic lod

Growth



* AvmubiJiiv ol
mm will ncciiH'T
1 oi microtowl jrdnMh



Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

38


-------
DO NOT CITE OR QUOTE

Mission Contamination Model
Results

BIO-BURDEN	RELEASE	TRANSPORT	REPRODUCTION

Mission Contamination Model



Marginal

Sensitivity Analysis













Probability of Contamination

Contamination Model Variables





Values





Units: = 10**-6



Extreme

Intermed.



Intermed.

Extreme

Nominal

5.9



Low

Low

NOMINAL

High

High

Low

High

Bio-Burden Variables















1. bio External

2.2

5.5

11

22

55

5

10.7

2. bio Covered

3.2

8

16

32

80

3.1

20.2

3. bio Encapsulated

4,000

10,000

20,000

40,000

100,000

5

10.4

Release Variables















1. rel Hard Landing Probability

0.0004

0.001

0.002

0.004

0.01

5.2

9.6

3. rel Newly Exposed/Hard, Encaps

0.0001

0.0002

0.001

0.005

0.01

5.4

10.9

4. rel Implanted, Soft

0.0001

0.0002

0.001

0.005

0.01

5.7

8.7

6. rel VTO/Vibration

0.001

0.002

0.01

0.05

0.01

5.4

11.1

9. rel VTO/Erosion, Encaps

0.00001

0.00002

0.0001

0.0005

0.001

5.4

10.9

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

39


-------
DO NOT CITE OR QUOTE

Mission Contamination Model
Marginal Sensitivity Analysis -2

Probability of Contamination

Contamination Model Variables





Values





Units:

= 10 **-6



Extreme

Intermed.



Intermed.

Extreme

Nominal: 5.9



Low

Low

NOMINAL

High

High

Low

High

Transport Variables















1 tra Survive Transit

0.001

0.002

0.01

0.05

0.1

2.2

45.2

2tra Find Water

0.0005

0.001

0.005

0.025

0.05

1.5

49.9

4 tra Water Deposition

0.00005

0.0001

0.0005

0.0025

0.005

5

15.2

5 tra Stay Lodged

0.1

0.2

0.5

0.8

0.9

5.5

10

Reproduction Variables















1 rep Psychrophilic, Anaerobic

0.005

0.01

0.05

0.1

0.25

0.6

29.6

2 rep Availability of Nutrients

0.01

0.02

0.1

0.2

0.5

0.6

29.6

Results

•	Contamination probability 6 x10"6, well below
mission limit of 10 4, insensitive to model
assumptions and input data

•	Simple explanation for why number is low: UV
flux through thin atmosphere kills microbes

•	NASA Scientific Advisory Committee (Carl
Sagan, Joshua Lederberg, et al.) persuaded of
acceptable safety

•	Mission flew, Viking Lander successful

•	Mid-course correction eliminated, Orbiter life
extended to one year

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

40


-------
DO NOT CITE OR QUOTE

National Academy of Sciences,
Viewpoint -1992

"... it is the unanimous opinion of the task group that
terrestrial organisms have almost no chance of
multiplying on the surface of Mars and in fact have little
chance of surviving for long periods of time, especially if
they are exposed to wind and to UV radiation."

— Space Studies Board, National Research
Council, Biological Contamination of Mars, 1992,
page 49: http://www.naD.edu/cataloci.DhD7recorcl id=12305

Retrospective on this Case Example

•	Planetary Quarantine since Viking Landing (1976):
- Not a major concern, perhaps excepting Mars

sample return. (Possibly new observations might
change the concern level.)

•	Example of Quantitative Risk Analysis built on highly
judgmental information

•	Accepted by scientific leaders and Agency managers; no
public concern or controversy regarding risk assessment

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

41


-------
DO NOT CITE OR QUOTE

A Key Conceptual Aspect for Levels 2 and 3

• Probability assessments are summaries of
information and subject to change:

-	"Necessarist" viewpoint; Ref: E. T. Jaynes, Probability Theory:
The Logic of Science, Cambridge Univ. Press, 2003.
(Endorsement in Nassim Taleb, The Black Swan, 2007)

-	Usual statistics viewpoint: probabilities are frequencies in data
obtained in independent, identical experimental trials.

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

42


-------
DO NOT CITE OR QUOTE

Extrapolation approaches

3.7 What kinds of extrapolations are needed (e.g.,
interspecies, exposure route, human variability,
low-dose/effect)?

Lead Discussants: Bill Mendez, Warner North

3.7. What kinds of extrapolations (e.g.,
high-low dose/effect, exposure route,
human variability, interspecies) are
needed?

Bill Mendez

Inorganic Arsenic Public
Stakeholder Workshop

January 8-9, 2013

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

43


-------
DO NOT CITE OR QUOTE

High to low exposure/dose

•	Context: High-quality epidemiological studies are
available for many endpoints

•	Often address primarily or exclusively high
exposures

•	Tradition: identify point of departure (POD),
extrapolate to low doses (linear, UFs)

•	Feasibility of model-based extrapolation

-	Need to account for covariates

-	Uncertainty about model form, confidence
bounds, at low exposures

-	Use in vivo, in vitro toxicity, metabolism data
to support extrapolation?

exposure route

•	Epidemiology available for inhalation and
ingestion exposures

•	Dosimetry for cross-pathway (ingestion/
inhalation) extrapolation is not well-established

•	Portal-of-entry effects can be significant

•	Consensus(?) is that cross-pathway estimation of
PODs for risk assessment is not advisable

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

44


-------
DO NOT CITE OR QUOTE

human variability

•	Exposure/dose-response

-	Important covariates (diet, smoking,
exposure to other stressors) not always
available

-	Sensitive developmental windows

-	Hard to generalize relationships between
genetic/metabolic variations and risks

•	Risk extrapolation to U.S. population

-	Differences in mortality, background
disease rates, background As exposures,
smoking, etc.

interspecies

•	In vivo tests could, in theory, be used to estimate
endpoints for risk assessment

-	Derive POD (NOAEL, BMDL) in range of data (curve
fitting)

-	Then ??

•	BMD models generally do not have a mechanistic basis

•	Apply CSAF (based on PBPK modeling) to account for
pharmacokinetic differences

•	Pharmacodynamic differences are highly species- and
endpoint-dependent, and generally not well-characterized

•	Use animal tests as back-stop for PODs derived from
epidemiology?

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

45


-------
DO NOT CITE OR QUOTE

in vitro, "omics" data

•	Role of in vitro data? (We have lots)
-Quantitative risk assessment

-	Safety assessment
-Inform low-dose extrapolation

•	Questions

-	How to simulate transport, metabolism

-	Identification of key pathways/events

-	Modeling of complex causal networks
over time

-Assessment of uncertainty

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

46


-------
DO NOT CITE OR QUOTE

Extrapolation approaches

3.8 What approaches are available for such

extrapolations (e.g., PBPK modeling, uncertainty
factors, probabilistic factors, linear/non-linear
dose-response)?

Lead Discussants: Hisham El-Masri, Jeff Gift



What approaches are available for
such extrapolations (e.g., PBPK
modeling, uncertainty factors,
probabilistic factors, linear/non-
linear dose-response)?

Hisham El-Masri

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

47


-------
DO NOT CITE OR QUOTE

Gastrointestinal
T ract

Skin arid Lungs

Liver

ABSORPTION

Bile

Blood and Lymph
Circulation

Metabolism

y

Metabolites

	i	 	

Storage



Kidney



Lung

Organs and Bones
Fatty Tissues

Feces

Urine

Extracellular
Fluids

DISTRIBUTION/
METABOLISM

Expired Air

ELIMINATION

PBPK Models: Internal Dosimetry

•	Defines the relationship between external concentration and
an internal measure of (biologically effective) exposure in both
experimental animals and humans

*	Use of PBPK Models can account for:

—	Interspecies differences in ADME

—	Nonlinear uptake, metabolism, clearance

—	Toxicity associated with products of metabolism rather than
parent chemical only

—	Tissue interactions (e.g. GSH depletion, induction of
clearance/repair, receptor occupancy)

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

48


-------
DO NOT CITE OR QUOTE

Previous As PBPK Models

Yu (1999) model:

•	Partition coefficients were solely determined using a child
poisoning case. This study provided total arsenic levels only.
There was no information in poisoning study that would help
the researchers to determine the partition coefficients for
arsenic and its metabolites (MMA and DMA) as was
published and referenced in the Yu (1999) publication.

•	Yu (1999) stated in their publication that they used the child
poisoning study to determine metabolic parameters such as
Vmax and Km. The child poisoning study did not have any
information that can lead to these estimates.

•	Yu (1999) model simulations were not tested against data.

Previous As PBPK Models

Mann et al. (1996) model:

•	The modeling effort for the humans was based on modification
of an earlier one that was established for rabbits and hamsters.
Both models did not include descriptions of current knowledge
about metabolism of arsenic (such as the inhibition effects of
Arsenic and MMA).

•	The model calibration relied heavily on "global" optimization
of parameters such as partition coefficients, first order oral
absorption constant, methylation rate constants, oxidation and
reduction constants. All of these parameters were optimized
using urine data. "Global" optimization would yield a set of
unidentifiable parameters.

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

49


-------
DO NOT CITE OR QUOTE

As Human PBPK Model

•A physiologically-based pharmacokinetic (PBPK) model was
developed to estimate levels of arsenic and its metabolites in human
tissues and urine after oral exposure to arsenate (AsV), arsenite (Aslll)
or organoarsenical pesticides.

•	The overall model consists of interconnected individual PBPK models
forAsv, Aslll, monomethylarsenic acid (MMAv), and, dimethylarsenic
acid (DMAv).

•	Metabolism of inorganic arsenic in liver was described as a series of
reduction and oxidative methylation steps incorporating the inhibitory
influence of metabolites on methylation.

•	Unique aspects of this model development effort are that it addresses
parameter sensitivity and identifiably, utilizes human data whenever
possible and incorporates new data on arsenic methylation.







Model Evaluation





Q9

. @



TctelAS

Q9

(9



TdciAB



~Q8





/

^Q8





/



<

|Q7





0 0 o 0 0 J 0 0
0 r

<

|Q7

0

o
o

V 0 _	



rine (un
o


-------
DO NOT CITE OR QUOTE

3ERA

What approaches are available for such

extrapolations (e.g., PBPK modeling,
uncertainty factors, probabilistic factors,
linear/non-linear dose-response)?

Low Dose Linear/Nonlinear Approaches

Jeff Gift
(January 9, 2013)

Benchmark Dose Training California Environmental Protection Agency Training Workshop

EPA 2005 Cancer Guidelines

Low dose linear - "slope is greater than zero at a dose of zero"
(curvature is possible near observed data).

Low dose nonlinear - "slope is zero at (and perhaps above) a
dose of zero."

"It is the Agency's long-standing science policy position that use
of the linear low-dose extrapolation approach provides adequate
public health conservatism in the absence of chemical-specific
data indicating differential early-life sensitivity or when the mode
of action is not mutagenic."

Benchmark Dose Training California Environmental Protection Agency Training Workshop

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

51


-------
DO NOT CITE OR QUOTE

3ERA

EPA 2005 Cancer Guidelines

Linear Approach

•	MOA known, evidence for direct mutagenic activity

•	MOA known, background doses of As (or other agents
with a common MOA) near levels associated with key
precursor events.

•	MOA Unknown

Nonlinear Approach

•	MOA known, evidence for nonlinearity at low doses (e.g.,
key precursor events with well defined nonlinear dose-
response relationships) and no evidence for mutagenic or
other activity consistent with linearity at low doses.

Linear vs NonLinear Approach

BMDL BMD

0

50

100

150

200

dose

14:40 01/25 2007

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

52


-------
DO NOT CITE OR QUOTE

SEPA

EGBE MOA,

hemangiosarcomas
in mice (Gift, 2005)

Benchmark Dose Training California Environmental Protection Agency Training Workshop

Arsenic History

Following the commendations of NRC (2001), EPA
(2010) used the linear approach because of "remaining
uncertainties regarding the ultimate carcinogenic
metabolites and whether mixtures of toxic metabolites
interact at the site(s) of action."

SAB (2007) concurred, indicating that:

As has the potential for a highly complex mode of action.
What is known about PK/PD properties of As not sufficient
to support a specific nonlinear dose-response relationship.
Agreed with NRC (2001) recommendation for linear dose-
response analysis of southwestern Taiwan population.

Benchmark Dose Training California Environmental Protection Agency Training Workshop

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

53


-------
DO NOT CITE OR QUOTE

Questions/Considerations

Population Data

•	Is population data useful forjudging low dose shape
given what is know about individual predispositions (e.g.,
from smoking, diet, genetic variants, bimethylation)?

MOA

•	Does MOA evidence exists for nonlinearity of Arsenic
cancer dose-responses (e.g., key nonlinear events)?

Arsenic's Low Dose Anticancer Activity

•	Is arsenic's therapeutic effect on certain cancers (e.g.,
leukemia) relevant to its ability to initiate the subject (e.g.,
lung and bladder) cancers at low doses?

Benchmark Dose Training California Environmental Protection Agency Training Workshop

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

54


-------
DO NOT CITE OR QUOTE

Extrapolation approaches

3.9 EPA has traditionally addressed uncertainty via
the application of uncertainty factors. What
other approaches may be available to address
and transparently convey the impact
of uncertainty on these extrapolations?

Lead Discussants: Ken Cantor, Weihsueh Chiu









3.9 Uncertainties in Extrapolation

Weihsueh Chiu

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

55


-------
DO NOT CITE OR QUOTE

•	Each extrapolation has associated uncertainty.

•	EPA is moving towards separating out the "adjustment"
component from the "uncertainty" component.

•	For example, for interspecies extrapolation of oral
doses:

—	BWS/4-scaling is considered a "central estimate" cross-
species adjustment.

—	A remaining 10,/2-fold "Uncertainty Factor" addresses
residual uncertainty.

—	Could also combine these probabilistically based on
analyses of multiple chemicals & endpoints.

What options are available for the
extrapolations needed for iAs?

•	Uncertainties

—	Across dose metrics

—	In variability in susceptibility across population (including
lifestages)

—	From precursor markers to disease endpoints

—	Below detectable effect levels

•	Data (or lack thereof) to support different approaches

—	Fixed factors

—	Sensitivity analyses

—	Probabilistic factors

—	Other approaches?

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

56


-------
DO NOT CITE OR QUOTE

Extrapolations to derive a RfD are
performed at a fixed level of response

Are there empirical data for iAs to support
these extrapolations?

Human

RfD

Dose

(log scale)



Disaggregating TK & TD requires an



internal dose metric



Is such data available for iAs endpoints?



TOXICO DYNAMICS

TOXICOKINETICS



Human ..

Human
(sensitive) , .. »
* ' (median)

Human (sensitive) Human (median)



/ / —

/ /

LU

CO

/ /
/ / ¦¦¦ "aT

/ /

O

O.

/ / -=
/ i p w

/ / ^ s

/ / o

/ /

i/>
LU
0£

/ / to ~
/ / °

(M	m





/ /



DOSE METRIC

EXTERNAL DOSE



(log scale)

(log scale)

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

57


-------
DO NOT CITE OR QUOTE

"Low dose extrapolation" is a bit of a misnomer,
and is also related to uncertainty

•	In the Benchmark Dose approach, we are fixing
the response level (usually in the observable
range), and estimating the associated dose.

—	In the range of observation, sensitivity to the model
form generally less.

•	Extrapolation is to lower response levels.

—	As the response level decreases below the observable
range, the uncertainty increases dramatically,
especially due to the assumed model form.

—	Linear extrapolation from the POD serves to define a
bound of the uncertainty range.

Alternative approaches to "low-
response extrapolation"

•	Both avoid the "linear/non-linear" (false) dichotomy.

•	Quantitatively characterize the extrapolation uncertainty

—	Acknowledges multiple dose-response shapes consistent with
the data.

—	Should incorporate both parameter and model uncertainty.

•	Use a dose-response characterization approach that does
not require "low-response extrapolation."

—	Even for cancer, could fix the response at "X% increased risk"
and derive an exposure limit that "protects" sensitive individuals
from that level of risk.

—	But not useful for addressing some risk management needs
(e.g., estimating overall population incidence).

—	For precursor endpoints, may shift debate to (false) dichotomy
of "adaptive" versus "adverse."

Public Stakeholder Workshop to Inform EPA's Upcoming IRIS Toxicological Review of Inorganic Arsenic

58


-------