Texas Commission on Environmental Quality

Roberta L. Grant, Susan L. Santos, Mike L. Dourson, Stephanie Shirley, Neeraja K. Erraguntla, R. Jeffrey Lewis, and Nancy B. Beck

Society of Toxicology, March 22-26, 2015

San Diego, CA

• Focused on US Environmental Protection Agency’s Integrated Risk Information System (IRIS) as an example

• Considered experiences in other programs with varied mandates

• Used specific case studies to explore diverse approaches (including some very basic ideas)

• Work is continuing to develop implementable approaches showing how information can be presented differently to improve the consideration and use of uncertainty information by risk managers

2

*American Chemistry Council’s Center for the Advancement of Risk Assessment Science and Policy

• The Workshop examined four approaches:

1. Comparing Values to Other Relevant Peer-Reviewed Numbers

2. Unpacking Toxicity Assessments to Understand and Improve Confidence – Focus of this talk

3. Presenting Toxicological Information Visually in the Context of Alternative Values, Exposure Levels, and Biomonitoring Equivalents

4. Improving Transparency in Dose-Response Decision Making

*American Chemistry Council’s Center for the Advancement of Risk Assessment Science and Policy

3

Completeness of Database

Study Quality

Reference Concentration (RfC)/ Reference Dose (RfD)

* USEPA (1994) Methods for Derivation of Inhalation Reference Concentration and Application of Inhalation Dosimetry

4

• Many steps in toxicity assessment – the major steps are referred to as ‘elements’

• 8 major ‘elements’ of a toxicity assessment and 3 ‘confirmatory elements’ assessed

• for the purposes of our example, each element is treated as being equally important

• how confident are we that the value is likely to be accurate/precise/predictive

Accuracy: degree of closeness of the measurements to that quantities actual (true) value

Precision: the degree to which repeated measurements under unchanged conditions show the same value

Predictive: is there a balance between being protective and reasonably accurate.

6

• Used clear criteria to judge confidence and implement scaling (numerical or descriptive values)

• 1 = low confidence; 5= high confidence

• Audience = risk assessor

• help explain approach and confidence to a risk manager/decision maker

7

Element Description for High Confidence

Database

Completeness

Database included investigations of a comprehensive array of non-cancer

toxicity endpoints, established from studies of chronic duration in various

mammalian species (refer to EPA 1994)

Systematic Review A systematic approach was used to identify studies, evaluate their quality

and integrate them.

Key Study Quality The key study(ies) are well-conducted and can be used without restrictions.

Critical Effect The database is sufficient to identify the effect occurring at early time points

(i.e. critical effect). This should protect against all other adverse effects. MOA

information, if available, helps to determine if the earliest critical effect has

been identified.

Relevance of Critical

Effect

The critical effect is known to be related to human findings. If only animal

studies are available, MOA information, if available, helps to determine if the

critical effect is relevant to humans.

Point of Departure

(POD)

Dose response is well understood and NOAEL and LOAEL are identified.

Ideally, BMC/BMD modeling was performed with small differences between

BMD and BMDL.

Human Equivalent

POD

Human data are available or human equivalent dose/concentration is known

from PBPK or similar model.

Sensitive Populations Human data on sensitive subpopulations are available or PBPK or similar

model is available to account for TK/TD differences between general and

sensitive populations.

Non-Cancer Toxicity Assessment Elements

8

Cancer Toxicity Assessment Elements

Element Description for High Confidence

Carcinogenic Potential Using a weight of evidence approach, adequate data exists to classify the

chemical into EPA/IARC categories (e.g., not carcinogenic, possibly carcinogenic,

known carcinogen, etc.).

Systematic Review A systematic approach was used to identify studies, evaluate their quality and

integrate them.

Key Study Quality The key study(s) are well-conducted and can be used without restrictions.

Relevance of Critical

Effect

The tumor type/site is known to be related (or may be related) to human

findings. If only animal studies are available, MOA information, if available,

helps to determine if the critical effect is relevant to humans.

Point of Departure Dose response is well understood. Ideally, BMC/BMD modeling was performed

with small differences between BMD and BMDL.

Human Equivalent POD Human data are available or human equivalent dose/concentration is known

from PBPK or similar model.

Low Dose Extrapolation A biologically based model or PBPK model is available and MOA understanding

leads to extrapolation to lower doses with confidence.

Sensitive Populations Human data on sensitive subpopulations are available or PBPK or similar model

available to account for TK/TD differences between general and sensitive

populations. If the MOA is mutagenic, then age-dependent adjustment factors

were applied.

9

Elements for a Confirmation of Toxicity Assessment

Element Description for High Confidence

Peer Review An external independent peer review was conducted

including opportunities for public comment, written peer

review report, and the Agency has responded

appropriately to peer review and public comments.

Validation The Agency has evaluated whether the final toxicity

values are realistic and plausible based on available

information.

Risk Value

Comparison

Relevant values from high quality, peer reviewed sources,

are within three-fold of each other.

10

Validation or Reality Check

• In some cases, the aggregate impact of all risk assessment decisions involved in a toxicity assessment results in a toxicity value that is overly conservative and unrealistic.

• Example: applying UFs in multiple areas, and then

multiplying them together, may compound the conservatism as the upper bounds on each of the factors is used in the calculation

11

Toxicity Assessment Elements that Were Not Considered

• Mode-of Action • RfC/RfD Confidence • Uncertainty Factors

12

Confidence Scoring: Example – Hazard Identification: Quality of Key Study(ies)

Confidence scale and basis for scoring

1 = Low: chosen study may have deficiencies, but is still considered useful.

2-3 = Medium: chosen study is reasonably well done and can be used with some restrictions; extrapolation seems reasonable based on findings with other chemicals

4-5 = High: chosen study is well done and can be used without restriction

Implication Klimisch scores for evaluating quality of toxicology studies have international support; other scales have also been widely used. A scoring method consistent with Klimisch scores for assessing quality of human studies also exists, along with other scales. Studies with Klimish scores associated with high confidence can be used with little or no restriction.

References Potential tools:

H.J. Klimisch, M. Andreae and U. Tillmann. 1997. A Systematic Approach for Evaluating the Quality of Experimental Toxicological and Ecotoxicological Data. Regulatory Toxicology and Pharmacology Vol 25 pp 1-5

EPA’s OPP Core grades (guideline, minimum, supplementary, invalid) EPA IRIS principal study confidence scoring (high, medium or low) Bevan and Strother, 2012 Money et al., 2013 Systematic Review tool quality assessment components (e.g. NTP OHAT

approach, Rooney, 2014)

Similar tables describing basis for confidence scoring available for other Major Elements described earlier

e.g., database completeness, systematic review, critical effect, etc.

13

Confidence Scoring: Example - Dose-Response Assessment: Point of Departure (POD)a

Confidence scale and basis for scoring

1 = Low: many uncertainties exist in the POD; only a free-standing NOAEL or LOAEL identified; few dose groups; BMD modeling not possible

2-3 = Medium: some uncertainty exists in identified POD, NOAEL or LOAEL identified, but few dose groups; BMD modeling was conducted; difference between BMD and BMDL is large

4-5 = High: dose response and basis for POD are well understood: NOAEL and LOAEL identified; multiple dose groups, BMD modeling conducted; difference between BMD and BMDL is small (approximately 2-fold or less)

Implication When BMC/BMD modeling can be performed, the entire dose-response curve is used to determine the POD. Such determination improves the basis of risk management decisions.

References Many references here (e.g., IPCS, 2005; EPA, 2002; EPA, 2012)

a slope of the dose-response curve is critical information if the RfC or RfD is exceeded. A steep slope

increases the importance of the exposure assessment.

Similar tables describing basis for confidence scoring available for other Major Elements described

earlier e.g., database completeness, systematic review, critical effect, etc.

14

Confidence Scoring Results for Inhalation RfC

for Carbon Tetrachloride (CCl4)

Peer Review +++

Validation -

Toxicity Value Comparison -

15

Major Elements

Confirmation Elements

Confidence Scoring Results for Inhalation

Reference Value (ReV) for 4-vinylcyclohexene

(4-VCH)

0 1 2 3 4 5

Database Completeness

Systematic Review

Key Study Quality

Critical Effect

Relevance of Critical Effect

Point of Departure (POD)

Human Equivalent POD

Sensitive Populations

Peer Review ++

Validation -

Toxicity Value Comparison -

16

Major Elements

Confirmation Elements

Confidence Scoring for RfC for CCl4

Confidence Scoring for ReV for 4-VCH

17

Element Score Basis Database Completeness 3

Medium

Developmental study in different species and multigeneration study lacking

(confidence from EPA (2010) was medium)

Systematic Review 1

Low

At the time of this assessment, IRIS did not employ a systematic procedure

for data gathering, analysis and internal review

Key Study Quality 5

High

The chosen study is well done and can be used without restriction

(confidence from EPA (2010) was high)

Critical Effect 4

High

Studies are sufficient to determine the critical effect with confidence; fatty

change in liver is moderate severity

Relevance of Critical Effect 5

High

The critical effect of liver toxicity is appropriate to humans. Extrapolation

seems reasonable based on findings with humans and other experimental

animal species. Critical effect matches human experience

Point of Departure (POD) 5

High

A lower limit on the BMD was used as the POD. Multiple dose groups

Human Equivalent POD 4

High

HEC and duration adjustments were derived using a PBPK model

Sensitive Populations 3

Medium

Available life stage information does not suggest increased childhood

susceptibility

Peer Review +++

High

The external peer review seemed adequate and EPA appeared to take

comments into consideration

Validation -

Not conducted

Risk Value Comparison -

Not conducted

Confidence Scoring for RfC for CCl4

18

Confidence Scoring for ReV for 4-VCH

Element Score Basis Database Completeness 3

Medium

A subchronic inhalation study was available in two species. Inhalation

developmental study and multigeneration inhalation study lacking. An oral two

generation reproductive/developmental study in mice showed no effects on

reproductive function (confidence from TCEQ (2011) was medium).

Systematic Review 1

Low

At the time of this assessment, TCEQ did not employ a systematic procedure for

data gathering, analysis and internal review.

Key Study Quality 3

Medium

The chosen study was conducted using GLP in rats and mice, although only10

animals/sex were evaluated. (confidence from TCEQ (2011) was medium).

Critical Effect

2

Medium

Studies are sufficient to determine the critical effect with confidence. Three

concentrations were tested and multiple endpoints evaluated. The following critical

effects occurred at the highest concentration: ovarian atrophy and mortality (severe

effects) and lethargy/ tremor (moderate effects).

Relevance of Critical Effect 1

Low

Mice are sensitive for ovarian atrophy because they produce more reactive

metabolite than humans. However, since it is possible that humans produce the

reactive metabolite, a default assumption was made that ovarian atrophy may occur

in humans. The MOA for tremor/lethargy is not known, so it was assumed these

effects were relevant to humans.

Point of Departure (POD) 2

Medium

BMC modeling was not conducted because adverse effects only occurred at the

highest concentration. A NOAEL and a LOAEL were identified.

Human Equivalent (POD) 3

Medium

Default duration adjustments and animal-to-human adjustments were conducted.

Sensitive Populations 1

Low

Available life stage information was not available to indicate sensitive populations

exist.

Peer Review ++

Medium

Peer input, a 90-day public comment period, and comments were addressed.

Validation - Data are not available to conduct a reality check.

Risk Value Comparison -

Chronic inhalation values from other sources were not available.

19

• Presentation of risk assessment results in clear and concise manner is challenging

• Risk Managers must apply this information to make decision

• No ‘right’ way to communicate this type of information

• Different approaches may appeal to different people

• Consideration should be given to testing (e.g. focus groups) before adopting or adapting approaches

20

• Method presented treats each ‘element’ as equally important

• But ‘element’ weighting could be done

• Many ‘elements’ interrelated (e.g., mode of action information important to identification of critical effect and human relevance)

• The working groups welcome input and feedback on the approaches

21

Roberta L. Grant, Neeraja K. Erranguntla, and Stephanie Shirley (Texas Commission on Environmental Quality (TCEQ))

Susan L. Santos (Focus Group Consulting & Rutgers University)

Mike L. Dourson (Toxicology Excellence for Risk Assessment (TERA))

R. Jeffery Lewis (Exxon Mobile Biomedical Services)

Nancy B. Beck (American Chemistry Council)

22

Thank you! roberta.grant@tceq.texas.gov