![[logo] US EPA](../gif/logo_epaseal.gif){kind=logo}
August 23-26, 2005 Charge to the Panel
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. , 20460
OFFICE OF
PREVENTION, PESTICIDES AND TOXIC SUBSTANCES
August 8, 2005
MEMORANDUM
Subject: Transmission of the Preliminary Cumulative Risk Assessment for the N -Methyl Carbamate Pesticides for review by the FIFRA Scientific Advisory Panel.
To: Myrta Christian, Designated Federal Official
FIFRA SAP
Office of Science Coordination and Policy (7101C)
From: Anna B. Lowit and David J. Miller
Office of Pesticide Programs,
Health Effects Division (7509C)
Through: George Herndon, Associate Director
Office of Pesticide Programs,
Health Effects Division (7509C)
A meeting of the FIFRA SAP is scheduled for August 23-26, 2005 . This meeting will focus on the Preliminary Cumulative Risk Assessment for the N -Methyl Carbamate Pesticides. This memo provides the charge questions to the Panel.
Questions for the SAP
HAZARD
EPA's hazard and dose-response chapter (I.B) and associated appendices (II.B.1-6) of the Preliminary Cumulative Risk Assessment describe the application of the Relative Potency Factor (RPF) method to the N-methyl carbamate pesticides. These documents a) outline the steps in developing the dose-response relationships for each pesticide and its capacity to inhibit AChE in rats; b) describe the data used in the assessment; c) summarize the empirical dose-response modeling which provides the basis for the relative potency factors (RPFs), points of departure (PoDs), and estimates of AChE inhibition half life; and d) provide the rationale for selecting oxamyl as the index chemical.
HAZARD QUESTION #1
Empirical Dose-Response and Time Course Modeling
At the February, 2005 meeting of the FIFRA SAP, EPA proposed an empirical model for use in the cumulative risk assessment of the N-methyl carbamates. This model contains a dose-response and a time to recovery component. Based on the comments from the Panel and following experience with its application EPA made some modifications to this proposed model. EPA has applied this revised empirical model to the available RBC and brain cholinesterase data for the N-methyl carbamates. BMD and BMDL estimates provided in the preliminary assessment were derived from cholinesterase data from multiple studies and in some cases, using different cholinesterase measurement techniques.
H1a. Please comment on the mathematical/statistical approach to modeling cholinesterase data used to estimate benchmark dose values and time to half-life recovery in the preliminary cumulative risk assessment. Please address biological and mathematical/statistical considerations in your response.
H1b. Please comment on the adequacy, clarity, and transparency of the documentation provided for the empirical dose-response and time course modeling.
HAZARD QUESTION #2
Selection of the Index Chemical
EPA's cumulative risk assessment guidance indicates that the index chemical should be selected based on the availability of high quality toxicity database for the common mechanism endpoint. The selection of the index chemical is an important step in the cumulative risk assessment; the BMD for oxamyl was used to calculate RPFs and the BMDL for oxamyl was used as the PoD for extrapolating cumulative risk.
H2. Please comment on the rationale provided for the selection of the index chemical. Should any additional factors be included in the rationale for the selection of oxamyl as the index chemical?
HAZARD QUESTION #3
Selection of Brain ChE data for developing RPFs and PoDs
EPA has used data for brain ChE as the basis for the RPFs and PoDs. The rationale for this selection was provided in I.B.
H3. Please comment on the rationale provided for the selection of the brain ChE as the basis for RPFs and PoDs in the preliminary cumulative risk assessment. Should any additional factors be considered?
WATER
WATER QUESTION #1
Revised Conceptual Model for Ground Water
Based on recommendations of the February 2005 SAP, OPP revised its ground water modeling approach to estimate pesticide concentrations in the upper meter of a fixed saturated zone (ground water) that starts at 3.5 m below the surface. The Agency has included two additional adjustments to the original conceptual model since the earlier SAP. The models consider variable degradation rates with depth and account for setback distances between the well and the application area by using lateral velocity to estimate the additional travel time for a pesticide to reach the well.
W1. Please comment on the Agency's revisions to the ground water modeling approach to account for variable degradation rates with depth and varying setback distances between the well and treated fields.
WATER QUESTION #2
Comparisons of the Three Models
The three models used by the Agency (PRZM, RZWQM, and LEACHP) provided predicted concentrations that were similar on average, but short-term concentration differences among the models varied considerably. Differences in peak concentration estimates ranged from a factor of 2 to 5 in Florida to as much as a factor of 20 in North Carolina ; however, there was no consistency with regard to which model gave the highest or lowest predictions. Some of these differences may due to differences in the way the models handle degradation-temperature relationships, evapotranspiration, and weather generation.
W2. Given that no model stands out as superior when compared to the monitoring data evaluated so far, can the SAP suggest criteria for further evaluation of the models?
WATER QUESTION #3
Evaluation of the Ground Water Model Estimates
The Agency compared NMC concentrations in ground water estimated with the three models (PRZM, RZWQM, LEACHP) to results of available prospective ground water monitoring studies (oxamyl in NC and MD and methomyl in GA), two well-monitoring studies along the central ridge of FL, and published literature on in-field monitoring studies. Using the FL well monitoring data, known fate characteristics of the NMC pesticides, and soil and hydrologic data, the Agency identified the conditions under which exposures similar to that estimated in the NMC CRA may occur: private wells drawing from shallow, acidic ground water with high to very high saturated hydraulic conductivities in the soil and vadose zone. This has allowed the Agency to move toward a spatially-explicit characterization of potential high exposure areas.
W3. Please comment on the performance of the models against the available monitoring data. What additional considerations should be taken when applying modeled estimates to risk assessments for areas where monitoring data are not available?
FOOD
FOOD QUESTION #1
The food portion of the N-methyl carbamate cumulative risk assessment used similar data sources and techniques to those used for the organophosphate pesticide for estimating cumulative risk from food. This included use of both the USDA's Continuing Survey of Food Intakes by Individuals (CSFII) as a data source for food consumption and Pesticide Data Program Data (PDP) as a data source for food residues.
F1. Please comment on the planned intermediate- and longer- term activities associated with sensitivity analyses identified in Section I of the document. Does the Panel have any suggestions for other or additional activities which the Agency should consider?
RESIDENTIAL
RESIDENTIAL QUESTION #1
Use of REJV Data and Professional Judgment
To generate estimates of exposure from residential use of NMC pesticides, the probabilistic models use a variety of inputs to address potential exposure from multiple use scenarios. Critical inputs include the percent of households applying the various pesticide products, and the timing of those applications. These two inputs, coupled with potential exposure from pesticide residues in drinking water and the diet, directly impact per capita estimates of cumulative exposure. In its February Case Study, the Agency presented background information on the Residential Exposure Joint Venture (REJV) survey. The Agency used this database as the primary source for data on the inputs relating to timing of applications and percent of households using NMC products. Details regarding the empirical details of the REJV survey are presented in Appendix II.E.1.
In February 2005, the SAP expressed reservations regarding the REJV data. In response to SAP concerns, EPA used other non-survey information in this preliminary CRA, in addition to estimates from REJV, to develop use/usage inputs and seasonal timelines of pesticide use which were representative of the Southern region of the U.S.
As previously mentioned, the REJV survey can be used to generate empirically-based estimates of percent of household use and the frequency of product specific applications. But, because the REJV did not collect information regarding the reason for the reported pesticide use (pest treated) or how much of the product was used, the empirical timing and frequency information (based on a national survey) may not provide a clear picture of regional use. Therefore, to establish the timing of pesticide applications for the scenarios likely to result in the highest exposure, OPP made these estimates based on a combination of REJV data, product label information, professional judgment, and pest pressure information available from the Cooperative State Extension Services. Specific examples of how these sources were used to determine timing and frequency of pesticide use for PNMC residential assessment are presented in Section E of the preliminary NMC CRA document.
R1. Please comment on the use of information sources other than REJV to establish periods of pesticide use and other use/usage information. Does the Panel suggest an alternative method to improve the use of REJV in the NMC assessment? Does the Panel know of other data sources that may be available?
RESIDENTIAL QUESTION #2
Uncertainties Associated with the Hand-To-Mouth Assessment
To assess non-dietary ingestion (mg/day), the following four key factors are used in the models:
Residue Concentration (turf residues, pet fur residues, and residues from hard indoor surfaces)
Hand to mouth frequency (number of events per hour)
Surface area of the inserted hand parts (cm2)
Exposure time (hours/day)
Other factors include both saliva extraction efficiency and wet hand adjustment factor. This exposure estimate is then used along with the Relative Potency Factor (RPF) and Benchmark Dose to estimate risk. In the Preliminary N-methyl carbamate assessment, risk estimates for non-dietary oral exposure result in the lowest Margins of Exposure (MOEs), and would therefore be of greatest concern to the Agency; however, these low MOEs appear to be due in part to the incorporation of micro-activity data into our macro activity models. As a result, the non-dietary ingestion scenarios in the Preliminary N-methyl carbamate cumulative risk assessment are the least refined.
The residue concentration values are derived from individual residue dissipation or deposition studies which are discussed in the Residential Chapter (Section E) of the Cumulative Risk Assessment document. The exposure durations are taken from the Agency's Exposure Factors Handbook. The hand to mouth frequencies and hand surface areas come from behavior studies relying either on observational data of young children using video tape analysis, trained observers, or parental observers. However, study data that evaluated hand-to-mouth frequency and surface area mouthed is difficult to interpret. Specifically, comparison of study results can be difficult due to differences in study practices and methodologies. For example, there are no standard definitions of mouthing (superficial contact, licking, biting, fraction of hand inserted) and thus the data for these behaviors likely differs among studies as a result of the investigators definitions. In addition, the degree to which ancillary data (such as surface area of hand contacted or inserted, the duration of contact, and the length of videotaping) are collected and reported differ among studies. This makes broad-based and generally-applicable interpretation difficult. Nevertheless, Drs. Zartarian and Xue allowed us the use their preliminary distributional analyses of these children's video data in this assessment. The studies used in the hand to mouth frequency analysis performed by Zartarian and Xue are briefly summarized in a table provided in a memorandum dated August 8, 2005 and provided to the Panel under separate cover.
The distributions of hand-to-mouth frequencies and surface area mouthed used in the Preliminary NMC CRA were based on the analysis performed by Zartarian and Xue (as detailed above). In the aggregate models used in the NMC cumulative assessment, each separate iteration selects a single value for the hand to mouth events variable from a distribution of hand to mouth frequency values. Also, each separate iteration of the model selects a single surface area from a distribution of the fraction of hand mouthed. These values are multiplied by the residues and exposure durations which are similarly selected from a distribution of residue and exposure durations as described above. This relatively simple selection process, however, ignores the numerous complexities and interrelationships involved in this critical behavior pattern. (For example, the fraction of a hand which is mouthed during each mouthing event may be inversely correlated with the frequency with which the hand is mouthed. Specifically, a high frequency of hand-to-mouth events may be associated with a smaller fraction of the hand which is mouthed. The algorithms used in the NMC CRA however, (as established by the OPP Residential Standard Operating Procedures (SOP's)) assume independence between these two parameters. This assumption likely leads to overestimates of exposures when upper percentiles of the hand-to-mouth frequency and area of hand mouthed distributions are combined. In addition, the macroactivity approach used in the NMC CRA aggregate models is based on the following assumptions:
The mouthing frequency (events per hour), as recorded during the course of observational studies, continue at the same rate for the entire exposure duration selected; in reality, a high-end mouthing frequency recorded over a short time interval (e.g., one hour) may not be likely to continue at the same intensity over a longer time period (e.g., 6 or 8 hours)
The hand is fully replenished with residues from a contaminated surface (e.g., the lawn, pet or hard flooring) between each hand to mouth event
The contact frequency and surface area data used in this assessment are taken from observational studies in which all hand contacts were recorded as hand-to-mouth events, regardless of the fraction of hand mouthed. Additionally, no adjustment was made for the duration of time the hand remained in the mouth.
R2a. The methodology used in the NMC CRA in which micro-activity data are used in macro-activity approach likely leads to systematic overestimates of exposure when upper percentiles of mouthing frequency and surface area of hand mouthed are combined. Does the Panel agree that this methodology does indeed overestimate exposure? Can the Panel suggest improvements to this methodology to further refine exposure estimates?
R2b. Does the Panel have suggestions for an alternative approach than the one used to estimate the non-dietary oral exposure pathway in the Preliminary NMC CRA? For example, would the use of a time weighted frequency value based on random hourly draws of hand frequency distributions more accurately estimate hand-to-mouth exposures?
RESIDENTIAL QUESTION #3
Distributional Analysis
Assessing residential exposure to pesticides is a complex process that must consider exposure from a variety of sources via multiple routes. To account for exposure from different sources, the PNMC residential exposure assessment identifies scenarios where significant exposure may occur. Each of these scenarios is defined by a specific type of activity or set of activities that may result in exposure. Generally the relationships between these activities and the resulting exposures are well-defined in that algorithms, equations, and standard operating procedures exist for calculating exposure based on the activity being performed. However the supporting data sets used to estimate exposure for various residential scenarios range from robust (e.g. unit exposure values) to limited or sparse (e.g. lawn sizes, area treated, duration of exposure, and saliva extraction factors). Additionally, information characterizing the extent to which each activity contributes to exposure for a particular scenario does not always exists (e.g. the amount of time spent in home gardens performing activities such as hand weeding versus staking tomatoes or harvesting sweet corn).
In general, the Agency has attempted to fit distributions (as described in Appendix II.E.2 of the NMC CRA) to the exposure measurements for residential activities when supporting information exists to characterize the extent to which the activity contributes to exposure for the residential scenario of interest. However, the Agency has employed uniform distributions to the data sets for which such supporting information does not exist, (e.g. lawn sizes, area treated, duration of exposure, and saliva extraction factors). The Agency has elected to create such distributions when the available data are limited to such an extent that it is uncertain how well they represent national variability. The Agency believes use of uniform distributions to be conservative in estimating potential exposure since uniform distributions tend to overestimate exposure.
R3a. Please comment specifically on the Agency's use of lognormal distributions to estimate residential exposure and the statistical methods and procedures by which the Agency has selected particular distributions (e.g. probability plots and goodness-of-fit statistics).
R3b. Does the Panel agree that the Agency's approach to creating and using of uniform distributions (i.e. ranges of values) for residential scenarios lacking adequate supporting information tends to overestimate exposure? Is the Panel aware of other data sources that may be better suited for assessing residential exposure scenarios of interest? Does the Panel have any suggestions regarding alternative distributions to use for scenarios where supporting exposure information is inadequate? To what extent should sensitivity analyses be used to assess the appropriateness of alternative distributions?
R3c. When the Agency fits distributions to various exposure values, the maximum value entered into the probabilistic models for a particular distribution is usually defined to be an upper percentile value such as the 99th percentile in order to ensure realistic input parameters. Recognizing that the Agency intends to perform sensitivity analyses to evaluate the effects of this truncation, please comment on the Agency's approach of truncating distributions that are input to the probabilistic models. Please comment on any other approaches that the Agency might use to evaluate uncertainties associated with choices about whether and where to truncate distributions.
INTEGRATION
INTEGRATION QUESTION #1
The cumulative risk assessment guidance describes key principles for conducting these risk assessments. One such principle is the need to consider the time frame of both the exposure (e.g., When does exposure occur? What is the exposure duration?) and of the toxic effect (e.g., What are the time to peak effects and the time to recovery? How quickly is the effect reversed?). EPA's Preliminary Cumulative Risk Assessment for the N-methyl carbamates describes the current limitations in data and software to fully characterize the dynamic nature of exposure, effect, and recovery for this common mechanism group. In order to address these limitations, OPP performed an examination of the exposure patterns for records from the high end of exposure distribution and found that that a large fraction (~70%) of daily records contributing to the upper tail of the food exposure distribution represent single eating occasions. Regarding drinking water and residential/nonoccupational exposure, EPA's preliminary assessment provided a characterization of the current availability regarding datasets and models and a description of the impact of these limitations on the risk estimates from specific exposure pathways (i.e., drinking water, residential).
I1a. Please comment on clarity and adequacy of the risk characterization provided in the preliminary cumulative risk assessment. Are there important aspects with respect to the strengths and weaknesses of the risk characterization other than the ones we identified?
I1b. Is the Panel aware of additional data which would aid the Agency in its cumulative risk characterization for the N-methyl carbamate pesticides? For example, is the Panel aware of any available data on the timing of water consumption events or can the Panel make any recommendations regarding reasonable assumptions that could be made to help characterize the estimated risk? Are there other sensitivity analyses and further investigations that would be equally or more important than the ones we identified?