December 3, 2003 Questions and Charge to the Panel
FIFRA SCIENTIFIC ADVISORY PANEL (SAP)
DECEMBER 3-5, 2003
FIFRA SAP WEB SITE http://www.epa.gov/scipoly/sap/
OPP Docket Telephone: (703)305-5805
Docket Number: OPP-2003-0316
SHERATON CRYSTAL CITY HOTEL
1800 JEFFERSON DAVIS HIGHWAY
ARLINGTON, VIRGINIA 22202
Draft Preliminary Probabilistic Exposure and Risk Assessment For Children Who Contact CCA-Treated Playsets and Decks and CCA-Containing Soil Around These Structures
Questions/Charge to the Panel
I. Exposure Assessment
Background: revisions to the SHEDS-Wood model and the probabilistic CCA exposure assessment since the August 2002 assessment
The current SHEDS-Wood model incorporates a number of code changes and new analyses that address comments and suggestions received from the 2002 SAP Panel members, EPA reviewers, public and other external groups as summarized in Tables 1 and 2 (and in more detail in Appendix 1) of the exposure assessment report. For example, some of the code changes and new analyses performed, include the following:
Scenario-specific changes or analyses
comparing results for children exposed to public playsets only to those also exposed to residential CCA- treated wood
conducting separate analyses for children 1-13 years to assess sensitivity of results to changing the age group from 1-6 years
conducting special analyses, such as changing dermal and GI absorption values
conducting a separate analysis for children who exhibit pica soil ingestion behavior
Changes to model inputs or assumptions using revised body weight and surface area equations based on NHANES III to update the body weight and hand size monthly rather than annually
for residue dermal exposure calculations replacing the transfer efficiency and fraction of skin contacted per time with user-specified distributions
determining the maximum dermal loading based on surface concentrations and transfer efficiency
changing the SHEDS-Wood approach for bathing events by allowing a variable number of days between baths
Sensitivity and uncertainty analyses
conducting sensitivity analyses by varying each variable up or down by 1 standard deviation
allowing the use of Beta, Weibull, and Gamma distributions for fitting distributions to model inputs
sampling parameter pairs during the uncertainty analyses step, rather than independently sampling each parameter
Evaluation of SHEDS-Wood model results
evaluating the comparability the SHEDS-Wood model results to results from other earlier (primarily) deterministic CCA assessments.
Specific Issues and Questions for the FIFRA SAP
Issue 1: Documentation, completeness, and clarity of the model source code and the exposure assessment reportBoth the SHEDS-Wood source code and the probabilistic exposure assessment report have been significantly revised since the August 2002 SAP.
Question A: The Source Code Directory on the CD provided to the SAP includes annotated code for the exposure and dose algorithms used in the SHEDS-Wood model. Are these algorithms consistent with the descriptions in the SHEDS-Wood CCA exposure assessment report? Does the revised SHEDS-Wood version 2 code (i.e., the code submitted for the December 2003 SAP) accurately reflect changes to the version 1 methodology (i.e., the code and methodology presented to the August 2002 SAP) described in the report?
Question B: The SHEDS-Wood CCA exposure assessment report presents the model construct, selected model inputs, model results, and comparison to other CCA model estimates. Please comment on the clarity, completeness and usefulness of this document.
Issue 2. Modifications to SHEDS-Wood model code and the exposure scenarios selected
A number of modifications to the model code and scenario-specific changes have been made to the SHEDS-Wood model since the August 2002 SAP.Question A: Considering the limitations of available information and state-of-the-art modeling methods required for the assessment of children's exposures from contacting CCA treated wood residues and CCA containing soil, are the revisions made to the SHEDS-Wood code or algorithms scientifically sound and acceptable ?Question B: The SHEDS-Wood model has been modified using feedback from the August 2002 SAP. In particular, the recent assessment, includes: assessment of exposures of children contacting only CCA treated public playsets; sensitivity of results to changing the age group of exposed children to 1-13 years, and; a separate analysis for children exhibiting pica soil ingestion behavior. The Panel is requested to comment on the appropriateness of the new exposure scenarios in the revised probabilistic exposure and dose assessment.
Issue 3. Key input variables and specification of associated variability distributions
Sensitivity and uncertainty analyses of the SHEDS-Wood model results identified the following as key input variables influencing the model results: wood surface residue-to-skin transfer efficiency; wood surface residue levels; fraction of hand surface area mouthed per mouthing event; and GI absorption fraction for residues. In addition to the above variables, sensitivity and uncertainty analyses also indicated the importance of following additional variables: average number of days per year a child plays around CCA-treated playsets, frequency of hand washing, daily soil ingestion rate, and average fraction of non-residential time a child plays on/around CCA-treated playsets.
Question A. Has the Agency used the best available information for developing input distributions for these variables? If not, are there any other data that EPA should be aware of? Considering the limitations and uncertainties with available information, are the choices made in developing distributions for each of these key variables using the available information reasonable and scientifically sound?
Question B. In some of these instances (see Table 12, page 58), because of data limitations, the Agency has made simplifying assumptions to represent them as point estimates based on professional judgement. Are the simplifying assumptions presented in the draft exposure assessment for making these decisions adequately supported by relevant scientific data? Are the choices made to quantify these variables (i.e., selected distributions or point estimates) reasonable and sound? \
Question C. Are the methods used for fitting variability distributions that are assigned to model input variables for the CCA assessment appropriate?
Question D. The Panel is requested to comment on whether any other model inputs are either key drivers of results or sources of large model uncertainty. Do these model input variables and the distributions assigned to them appropriately reflect available scientific data ? Did EPA appropriately integrate the available data to derive the distributions for these input variables?
Issue 4: Methods and results for sensitivity and uncertainty analyses
EPA's draft CCA Exposure Assessment includes a formal sensitivity and uncertainty analysis as well as discussion of various sources of uncertainty in the model analyses.Question A: The Panel is requested to comment on the utility and suitability of the statistical diagnostic tools used by SHEDS for analyzing model results (e.g., variability analyses, sensitivity analyses, uncertainty analyses).
Question B: Is the bootstrap approach that is used for fitting uncertainty distributions, which has been revised in response to prior SAP comments, implemented properly, or are there alternative approaches that are recommended?
Question C: Are the uncertainty distributions assigned to chemical and non-chemical specific model input parameters appropriate?
Question D: The Panel is requested to comment on whether the modeling approach and documentation appropriately identify and address critical sources of uncertainty in the model and the resulting exposure estimates. Does EPA's documentation adequately describe the uncertainties inherent in the data used for modeling and the influence of these uncertainties on interpretation of the modeling results?
Question E. Does the Panel recommend performing any additional uncertainty analyses to evaluate the impacts of using alternative input distributions on the modeling results (e.g., to address uncertainties in various factors determining the frequency of children's exposures to CCA-treated wood in playsets and decks)?
Issue 5: Special Model Simulations
A number of special simulations with the SHEDS-Wood model were conducted in order to examine the importance of specific exposure scenarios or the impact of certain input assumptions. For example, some of these analyses included conducting separate simulations for children exposed to public playsets only, modeling exposures of the 7-13 year old age group, and studying exposures of children exhibiting pica behavior. Additional analyses were also conducted to examine the impacts of using data or assumptions about increased GI absorption, decreased dermal absorption, lowering the transferable wood residue concentrations by sealants, and hand washing after play events. The results from these special analyses were not significantly different than the baseline model results, except for the large impact of assuming the use of sealants would greatly reduce wood residues.
Question A. The Panel is requested to comment on the appropriateness of the justifications made in characterizing the key factors or inputs for each of these special simulations. Did the Agency provide adequate technical rationale and justification for its choices for these alternative exposure scenarios or input distributions? Do the results from these special analyses reflect proper use of available information?
Question B: Do any of the findings from these special analyses necessitate the Agency to consider revising certain scenarios or inputs to the baseline assessment?
Issue 6: Evaluation of the SHEDS-Wood model results
The Agency has evaluated the probabilistic CCA exposure model results by comparing them to results from other earlier deterministic CCA assessments. In particular, the SHEDS-Wood model results were found to compare well to a deterministic CCA assessment performed by the Gradient Corporation, and SHEDS-Wood upper percentiles compare well to deterministic Consumer Product Safety Commission estimates.
Question A: Has EPA provided adequate documentation of the overall plausibility of the exposure estimates generated by the SHEDS-Wood model for CCA? Are the comparisons with the results of other selected exposure assessments appropriate and appropriately presented? Are there any other types of benchmarking approaches or data to assess the reliability of the overall exposure model or specific model elements?
Issue 7: Overall completeness and acceptability of the SHEDS-Wood probabilistic CCA exposure assessment
EPA has revised the August 2002 SHEDS-Wood exposure assessment after carefully considering numerous comments and suggestions that it has received from various parties, including those from the August 2002 FIFRA SAP members, EPA/ORD and EPA Program Office peer-reviewers of the preliminary draft September 2003 report, and from the general public and other external groups.
Question A: In addition to the comments and suggestions already offered by the Panel members under the specific issues raised previously, considering the availability of data and information, does the Panel recognize any critical gaps in information or methodologies that still need to be addressed for the CCA exposure and dose assessment?
II. Risk Assessment
FIFRA SCIENTIFIC ADVISORY PANEL (SAP)
Issue 8: In the study by Nico et al. (2003), X-ray absorption spectroscopy (XAS) was used to determine the chemical and structural state of arsenic and chromium molecules in CCA-treated wood residue samples.
Based on the results of their analysis, Nico et al. (2003) determined that arsenic and chromium formed a "chemical complex bonded to the wood structure." Based on this study, the dominant oxidation state of the two elements is As(V) and Cr(III), and the local chemical environment of the two elements is best represented as a stable Cr/As cluster consisting of a Cr dimer bridged by an As(V) oxygen ion. Nico et al. (2003) also maintained that this chemical complex was quite resistant to leaching.
Question A: The Panel is requested to comment on the Nico et. al. (2003) study and particularly on the arsenic and chromium chemical complex from CCA treated wood surface residue, and whether the Panel believes that the chemical complex is formed during the fixation process. What is the meaning of this complex cluster formation to the current risk assessment
Issue 9. Casteel et al. (2003), reported that the relative bioavailability (RBA) of dislodgeable wood residue is 27%.
This value is significantly lower than the default value of 100% that is usually employed when reliable site-specific data are lacking and also lower than the RBA value recommended by the SAP 2001. The result of this study indicates that the arsenic in the dislodgeable arsenic material is not as well absorbed as soluble arsenic.Question A: Does the Panel agree that, in light of the Casteel study and the Nico study discussed in issue 8, the Agency should use 27% for the RBA to estimate the bioavailable dose.
Issue 10: In the 2001 SAP meeting, the Panel cited the research of Wester et al. (1993) as a source of the dermal absorption rate of soluble arsenic in water and soil.
The Panel recommended using a 2-3 % dermal absorption rate for arsenic residue on the surface of wood. Recently, a preliminary study by Wester et al. (2003) has been submitted by the same laboratory compares the dermal absorption of arsenic in CCA-treated wood surface residues with arsenic in water solution. Although the Agency has not receive the complete results of this study (e.g., the recovery of the arsenic in the urine of the animal given IV dose of arsenic), the preliminary results of this study indicate that the dermal absorption of 0.01% from wood surface residue was approximately two order of magnitude lower than the results in water. The dermal absorption from this study was based on urinary arsenic data following application of arsenic in CCA-treated wood residue that had been weathered by the environment.
Question A: Taking into consideration the Nico et al. study mentioned in issue 8, the Panel is requested to comment on whether this new study conducted by Wester et al. provides a more appropriate estimate of dermal absorption from contact with CCA-treated wood surfaces than the earlier 1993 Wester et al. study.
Issue 11: In the 2001 SAP meeting, the Panel recommended that a biomonitoring study be performed on children who are normally exposed to CCA-treated playground equipment and decks.
The Panel recommended that the study should be designed according to well-accepted epidemiological principles, including adequate sample size, to resolve the issue of whether there are substantial exposures to children from arsenic residues after playing on decks and playsets. The Panel indicated data from such a biomonitoring study could be directly used in the risk assessment and could be used to validate the exposure assessment model. Recently, a proposed protocol for a pilot study was submitted to OPP for peer review; this proposed protocol is an attempt to determine if changes in exposure to arsenic can be assessed by examining changes in the urinary excretion of arsenic. EPA has provided the Panel with a copy of the proposed protocol for the pilot study. In summary, the proposed pilot study will determine whether a significant difference in urinary arsenic can be discerned when a population of children are switched from arsenic-containing tap water to an essentially arsenic-free source of drinking water.
Question A. The Panel is requested to comment on the strengths and limitations of the approach to be employed in the proposed pilot study to help resolve the issue of whether there are substantial exposures to children from arsenic residues after playing on decks and playsets. In particular, please comment on the feasibility, the potential confounding background sources from the statistical analysis, the sensitivity and accuracy of analytical method for quantitation of arsenic in urine to detect changes, the determination of intraindividual variation and interindividual variation based on the current knowledge of exposure; and any other aspects of the proposed pilot study that might affect its utility.
Question B. The Panel is asked to describe approaches for gathering additional data e.g., data on the efficiency of transfer of surface residues to the skin surface (which has been identified as one of most critical model inputs based on the uncertainty analysis) to improve the estimates of exposure and / or the level of confidence in such estimates, and with respect to these approaches, as well as the proposed pilot study, to comment on the cost of data generation, the amount of time to generate the data, and the degree to which the data will reduce uncertainty about the accuracy of the model estimates.
Issue 12. Prior to the availability of probabilistic models, such as SHEDS, OPP estimated the lifetime average daily dose (LADD) and corresponding cancer risk to pesticides via a deterministic approach using central tendency input parameters (median or mean values).
Probabilistic models now allow OPP to express input parameters as distributions and subsequently generate a distribution of LADDs and corresponding pesticide cancer risks. In other words, the deterministic approach results in a single cancer risk value and the probabilistic approach results in a distribution of cancer risk values.Question A. The Panel is requested to comment on whether in this probabilistic approach of using the upper bound arsenic cancer slope factor combined with using high-end LADDs would result in a significant overestimation of the risk for the more highly exposed percentiles of the population ? If this is an overestimate, what other values would the panel recommended using as replacements, or in addition to the values that were used that would minimize the overestimation of risk without substantially underestimating the risk for such percentiles.
In this assessment, the estimated risks are considered approximations because inaccuracies may occur when exposures are summed across routes at the quartile level especially in the upper percentile. This is due to the way the Monte Carlo simulations were conducted and the outputs summarized.
Question B. The Panel is requested to comment on the range of percentiles, if any, at which there is a significant decrease in the reliability of the estimates of risk.