osure

Lisa Jo Melnyk1, Margie Byron2, Gordon Brown2, Andy Clayton2, anciLarry Michael

1 - USEPA, National Exposure Research Laboratory,Cincinnati,OH
2 - RTI International, Research Triangle Park, NC

Introduction

Assessment of total aggregate exposure necessitates consideration of the important role
that the dietary route contributes to overall exposure. For young children, by virtue of
their unique activities, the dietary exposure contribution may actually be the dominant
contributor. However, unlike dietary exposures of adults, where the contribution to the
total intake predominantly arises directly from the food, children's dietary exposure in-
cludes indirect pathways. Contaminated hands and surfaces are also significant
sources of exposure during eating, particularly for younger children where hand-to-
food and surface-to-hand-to-food frequencies are relatively high.

A linear deterministic model, termed the Children's Dietary Intake Model (CDIM), was
developed to identify the various input parameters needed to assess total dietary expo-
sure. The parameters included not only the pesticide residue concentration of the
food or surface, but also the frequency and duration of transfer (e.g., surface-to-food),
including the transfer efficiencies between the surfaces, and transfers between con-
taminated hands and foods (e.g., surface-to-hand-to-food). The deterministic model
evaluations concluded that indirect ingestion activities were significant to overall ex-
posure.

Important research is still ongoing by EPA/NERL to model pesticide residues 011 foods.
Residue data, along with food consumption information, have already been compiled
into national databases and have been modeled with the Stochastic Human Exposure
and Dose Simulation (SHEDS) modeling system. CDIM is now being evaluated as a
probabilistic model to facilitate prediction of dietary intake of young (i.e., < 6 yrs.)
children. CDIM evaluations focus on the increased dietary exposure due to surface-to-
food and surface-to-hand-to-food contacts in connection with SHEDS dietary module.

/=R .FT + [es • FS • TSF • ASF] + [(CS • SHR • TSH* ASH)(THF • AHF • /•//)]

Where

/ = total dietary intake (pg) of pesticide
R = pesticide residue (fxg/g food) in food item
FT = total amount (g) of the contaminated single piece of food consumed
CS = pesticide residue on the contacted surface (jig/cm2 of hardwood floor)
FS = total food surface area in contact with surface (cm2)

= product of the portion of food surface area in contact with the contaminated
portion of the surface (PS) and the total food surface area (Q)
TSF = surface-to-food transfer efficiency (dimensionless)

ASF = surface-to-food contact frequency (dimensionless)

SHR = factor to account for surface area contacted by hand (=1 if no wiping effect; >1
otherwise)

TSH = surface-to-hand transfer efficiency (dimensionless)

ASH = surface-to-hand contact frequency (dimensionless)

THF = hand-to-food transfer efficiency (dimensionless)

AHF ~ hand-to-food transfer frequency (dimensionless)

FH = total food surface area in contact with contaminated portion of hand (cm )
= product of the portion of food surface area in contact with the contaminated
portion of the hand (PH) and (Q)

Objectives

The overall objective was to stochastically evaluate the deterministic CDIM.

Figure 1: Variation in Partial F-Statistic with Model Parameter by Pesticide,
Surface Type, and Selected Food Type

Specifically,

1 To carefully select and use input variables to determine the impacts on

each parameter within the model
1 To determine the range of dietary values using national database infor-
mation, in-house dietary data, and published dietary data with varied
input parameters
1 To determine the impact of reasonable ranges for activity factors
1 To estimate total dietary intake for selected pesticides in selected foods

Approach

Identify relevant data sources

->Data distributions were created

->Data distributions were synthesized or otherwise derived from other data
Incorporate SHEDS data

-^Several datasets received from EPA to support creation of the distributions for Term 1
->USDA-sponsored pesticide analysis results for specific food items
->Food consumption data by age and gender
Create Bridges

-^Between food items in the pesticide residue and food consumption databases
-^Between food items in the food consumption database and food diary data for selected studies
Construct SAS Datasets

->Each data source was compiled in Excel at the sample matrix level
-^Relevant variables were extracted and standardized to create a single SAS dataset for each
parameter type

Where needed, measurements were converted to ensure compatibility between all parameters and
terms in the model
Parametric Distributions of Raw Data

-^Variables were coded to specified categories to maintain consistency across studies

Analyses were conducted on cis-permethrin, trans-permethrin, chlorpyrifos, and diazinon
->THF and AHF were examined across food groups
->TSF and ASF were examined across surface types and food groups
-^Normal, lognormal, beta, exponential distributions were used for model parameter values
->Goodness-of-fit statistics were used to determine the distribution that best fit the data
->A11 pesticide data were used as a surrogate to obtain the distribution shape when specific
pesticide information was lacking
Monte Carlo Analysis

->Used to obtain distributions of total dietary intake of one pesticide
-^Terms 1, 2 and 3 values were randomly drawn from corresponding distributions
->N = 5000 children in a given age class was used
-^Combinations of pesticides, surfaces and foods were used
Sensitivity Analysis

-^Compare results using an F-statistic and graphic display to measure the amount of variation in the
model attributed to each parameter
Assess the sensitivity of the parameter to the intake model by the partial F-test statistic, i.e., the
ratio of the variation in pesticide intake attributed to a particular model parameter when it's fixed
at low, medium, and high values over the variation in pesticide intake that is not attributed to the
particular model parameter

Results

Table 1: Pesticide Intakes For Selected Food Diary Items



Food Amount (g)

Food Amount

No. of Pieces Mass (g)/



Pesticide

Intake/

Total Intake



Diary Food Item

(USDA)

(Diary)

of Food Piece

Food Code

ID.



(Hg)

Pesticide Name



90

1 slice

1 90.0

51000180

024

0.79

0.79

Diazinon

Bread

90

1 slice

1 90.0

51000180

160

0.39

0.39

Chlorpyrifos

Bread

90

1 slice

1 90.0

51000180

222

0.71

0.71

cis-Permelhrin

Bread

90

1 slice

1 90.0

51000180

223

0.77

0.77

trans-Pcrmethrin

Soft Taco

50



1 50.0

52215350

024

0.38

0.38

Diazinon

Soft Taco

50



1 50.0

52215350

160

0.06

0.06

Chlorpyrifos

Soft Taco

50



1 50.0

52215350

222

0.18

0.18

cis-Permethrin

Soft Taco

50



1 50.0

52215350

223

0.33

0.33

trans-Permethrin

Range of F statistic values for Pesfidde: Diazinon

Range of F statistic values far Pesticide: da-Permdhrfn

Range of F statistic values for Pesticide: Chkxpyrifoe

Range of F statistic values for Pe

je of F statistic vriluefl tor Surface: Pcfoue

Range of F statistic values tor food Type: B

Figure 2: Variation in Partial F-Statistic with Model Parameter by Pesticide,
Surface Type, and Selected Food Type; Surface Concentration Excluded

es tor Peetlckls: Dtazhon

of F statistic values tor Pestdde: Crtorpyrttos

or PesBdde: trans-Perniettrti

s tor food Type: Bread

38 tor Surface: Non-porous

Range of F statistic values tor food type; To

2









-u _

_ +



Conclusions

Surface concentration (CS) is the dominating influence when calculating total dietary exposure, for most of the combi-
nations of pesticide, food type, and surface.

Pesticide residue (R) and amount consumed (FT) are also influential for many combinations of pesticide, surface type,
or food type; in one case, more influential than surface concentration (CS).

In general, transfer efficiencies TSH and THF are modestly influential within pesticide, surface type, and food type.
Conversely, TSF does not appear to have as much effect. This may reflect a lack of data as opposed to little influence.
No other parameter, with the possible exception of the food surface area, appears to be influential.

Parameters AHF, SHR, ASH, and ASF are rarely, if ever, influential. This may be a result of the sparcity of information
for these parameters, incorrect assumptions on data interpretation, or possibly may suggest the existence of parameter
interactions or confounding terms.

Surface concentration (CS) appears to be more influential for foods with higher transfer efficiencies. This suggests that
the intra-term parameters are not truly independent in Lhe model.

Acknowledgement

The work presented could not have been possible without the help from Jim Xue, USEPA, NERL/RTP, Gerry Akland, consul-
tant, and Natalie Freeman, consultant.


-------