A Meta-Analysis of Children's Hand-to-Mouth Frequency Data for Estimating Nondietary Ingestion Exposure
Abstract
Because of their mouthing behaviors, children have a higher potential for exposure to available chemicals through the nondietary ingestion route; thus, frequency of hand-to-mouth activity is an important variable for exposure assessments. Such data are limited and difficult to collect. Few published studies report such information, and the studies that have been conducted used different data collection approaches (e.g., videography versus real-time observation), data analysis and reporting methods, ages of children, locations, and even definitions of “mouthing.” For this article, hand-to-mouth frequency data were gathered from 9 available studies representing 429 subjects and more than 2,000 hours of behavior observation. A meta-analysis was conducted to study differences in hand-to-mouth frequency based on study, age group, gender, and location (indoor vs. outdoor), to fit variability and uncertainty distributions that can be used in probabilistic exposure assessments, and to identify any data gaps. Results of this analysis indicate that age and location are important for hand-tomouth frequency, but study and gender are not. As age increases, both indoor and outdoor hand-to-mouth frequencies decrease. Hand-to-mouth behavior is significantly greater indoors than outdoors. For both indoor and outdoor hand-to-mouth frequencies, interpersonal, and intra-personal variability are ∼60% and ∼30%, respectively. The variance difference among different studies is much bigger than its mean, indicating that different studies with different methodologies have similar central values. Weibull distributions best fit the observed data for the different variables considered and are presented in this article by study, age group, and location. Average indoor hand-to-mouth behavior ranged from 6.7 to 28.0 contacts/hour, with the lowest value corresponding to the 6 to <11 year olds and the highest value corresponding to the 3 to <6 month olds. Average outdoor hand-to-mouth frequency ranged from 2.9 to 14.5 contacts/hour, with the lowest value corresponding to the 6 to <11 year olds and the highest value corresponding to the 6 to <12 month olds. The analysis highlights the need for additional hand-to-mouth data for the <3 months, 3 to <6 months, and 3 to <6 year age groups using standardized collection and analysis because of lack of data or high uncertainty in available data. This is the first publication to report Weibull distributions as the best fitting distribution for hand-to-mouth frequency; using the best fitting exposure factor distribution will help improve estimates of exposure. The analyses also represent a first comprehensive effort to fit hand-to-mouth frequency variability and uncertainty distributions by indoor/outdoor location and by age groups, using the new standard set of age groups recommended by the U.S. Environmental Protection Agency for assessing childhood exposures. Thus, the data presented in this article can be used to update the U.S. EPA's Child-Specific Exposure Factors Handbook and to improve estimates of nondietary ingestion in probabilistic exposure modeling.
1. INTRODUCTION
Individuals have the potential for exposure to toxic chemicals through the nondietary ingestion route. Toxic chemicals can be transferred from contaminated surfaces or soil to the hand and then ingested via hand-to-mouth activity. Thus, frequency of hand-to-mouth behaviors is an important variable for estimating mass of chemical ingested by humans via mouthing. This pathway is particularly important for children because as part of their natural development, children mouth their fingers and other objects (Hubal et al., 2000). Data to assess this pathway are limited and difficult to collect. Few published studies report such information, and the studies that have been conducted used different data collection approaches, data analysis, and reporting methods, ages of children, locations, and even definitions of “mouthing” (i.e., contact with lips, inside of mouth, tongue). Because this difference in reporting of data makes it challenging to compare results among individual studies, the available data have not been analyzed collectively for use in probabilistic exposure models.
Generally, children's mouthing behavior is studied using both direct observation and videotaping methodologies (Zartarian et al., 1997; Reed et al., 1999; Freeman et al., 2001; Ferguson et al., 2005). Observations may be conducted by a parent, after being given special instructions, or by a trained observer. Videotaping to record the child's behavior is done by a trained technician. Videotape footage is then translated by a person who watches the videotapes and records information by hand (e.g., number of hand-to-mouth contacts) or uses video translation software (Zartarian et al., 1997; Ferguson et al., 2005). Data analyses from these studies are reported as either a frequency of contact (i.e., contacts/time) or as an exposure period (i.e., minutes). This article focuses on frequency of hand-to-mouth contact; future research could analyze of available data for hand-to-mouth duration data.
The general equation for estimating nondietary ingestion of chemical residue via hand-to-mouth contact involves the product of hand residue or soil loading (ug/cm2 or ug/g), hand-to-mouth frequency (# contacts/hr), hand surface area mouthed per mouthing event (cm2), and exposure duration (hr/day). Thus, to enhance estimates of nondietary ingestion in exposure assessments, reliable hand-to-mouth frequency data are important. Currently, data from four studies are typically used to assess exposure through this pathway (U.S. EPA, 2001a, 2002). The mean and 95th percentile values of hand-to-mouth frequencies from one of the available studies (Reed et al., 1999) are currently used by the U.S. EPA Office of Pesticides Programs in its Standard Operating Procedures (U.S. EPA, 2001a) for estimating nondietary ingestion exposures. Updated distributions reflecting all available data can be used in probabilistic models (e.g., Calendex™, developed by Exponent, Inc., http://www.exponent.com/practices/foodchemical/calendex.html; CARES®, developed by the International Life Sciences Institute, http://cares.ilsi.org/; Lifeline™, developed by The Lifeline Group, The Lifeline Group, Inc., 2006, http://www.thelifelinegroup.org/; and SHEDS, developed by U.S. EPA's Office of Research and Development, Zartarian et al., 2000) that are used to conduct children's exposure and risk assessments.
A workshop on Micro/Macro-Activity Data Needs to Improve Multi-Media, Multi-Pathway Exposure/Intake Dose conducted by the EPA in 2001 (U.S. EPA, 2001b) brought scientists together to discuss activity studies to date and different approaches used (e.g., real-time hand recording and videotaping), and to identify research needs. The analysis presented here is an attempt to respond to the workshop recommendation to assess how similar or different activity data are across studies, and determine any data gaps for additional research. Although the studies used to conduct this analysis collected data on other mouthing behaviors (e.g., object-to-mouth), the analysis presented in this article focuses on hand-to-mouth behavior only.
This article is the first attempt to compile hand-to-mouth frequency data from all available studies, and to conduct a meta-analysis with the following objectives:
- 1
Examine differences across studies by age (using the new EPA recommended age groupings (U.S. EPA, 2005)), gender, and indoor/outdoor location;
- 2
Fit variability distributions to the available hand-to-mouth frequency data for use in 1-D Monte Carlo exposure assessments;
- 3
Fit uncertainty distributions to the available hand-to-mouth frequency data for use in 2-D Monte Carlo exposure assessments; and
- 4
Assess hand-to-mouth frequency data needs using the new EPA recommended age groupings (U.S. EPA, 2005).
2. METHODS
2.1. Approach for Examining Differences Across Studies by Age, Gender, and Indoor/Outdoor Location
Few published studies containing hand-to-mouth frequency data are available, and those available have collected and reported the data in different ways. Those studies that were available in 2002 are described in the U.S. EPA Child-Specific Exposure Factors Handbook (U.S. EPA, 2002). Additional studies are being included in the updated version of the handbook, expected to be released in 2007. A summary of the 9 studies used in the meta-analysis for this article are given in Table I. For each cited study, all of the protocols and procedures related to human subjects research were reviewed and approved by an independent institutional review board (IRB) and complied with all applicable requirements of the Common Rule regarding additional protections for children. While most of these involved videotaping methods, several used parents and trained observers (Tulve et al., 2002; Greene, 2002). Some of the videography studies (Black et al., 2005; Beamer et al., in preparation; Leckie et al., 2000; Zartarian et al., 1997; Hore, 2003) used video translation software developed by Stanford University (Ferguson et al., 2005; Zartarian et al., 1997) and others (Freeman et al., 2001; Reed et al., 1999) used manual recording from video observations. As mentioned in AuYeung et al. (2004), mouthing of hands can be defined differently across studies (lips, inside of mouth, tongue).
| Study | Study Location | No. of Subjects | M:F Ratio | Subjects' Age (Min, Mean, Max) Yrs. | Child Observation Period |
|---|---|---|---|---|---|
| Zartarian et al., 1998 | CA | 4 | 0.5 | 2.4, 3.3, 4.2 | 6–10 hrs |
| Reed et al., 1999 | NJ | 30 | 0.4 | 1.8, 4.2, 6.2 | 3–7 hrs |
| Leckie et al., 2000 | CA | 20 | 0.35 | 1, 3.3, 12 | 1–2 hrs |
| Freeman et al., 2001 | MN | 19 | 0.42 | 3, 7, 12 | 4 hrs |
| Greene, 2002 | TX and IL | 169 | 0.53 | 0.3, 1.5, 3.0 | 4 hrs on 2 different days |
| Tulve et al., 2002 | WA | 86 | 0.5 | 0.8, 2.8, 5 | 5 to 60 mins per day for 1–6 days |
| Hore, 2003 | NJ | 10 | 0.6 | 2, 3.4, 4.6 | 4 hrs |
| Black et al., 2005 | TX | 68 | 0.44 | 0.6, 2.3, 5 | 4 hrs |
| Beamer et al., in preparation | CA | 23 | 0.44 | 0.5, 1.3, 2.3 | 2–6 hours |
Because the published studies reported summary statistics and study results in different ways and for different age groups, the authors of individual studies were contacted to obtain hourly hand-to-mouth frequency data so that the data could be pooled and reanalyzed collectively in this meta-analysis, using the new EPA age groupings. For the 3 to <6 months age group, indoor hand-to-mouth frequency data were available for 10 children, and no outdoor hand-to-mouth frequency data were available. For the 6 to <12 months age group, there were 65 subjects with indoor hand-to-mouth frequency and 10 with outdoor. For the 1 to <2 years age group, there were 118 subjects with indoor hand-to-mouth frequency and 28 with outdoor. For the 2 to <3 years age group, there were 92 subjects with indoor hand-to-mouth frequency and 41 with outdoor. For the 3 to <6 years age group, there were 96 subjects with indoor hand-to-mouth frequency and 41 with outdoor. For the 6 to <11 years age group, there were 14 subjects with indoor hand-to-mouth frequency and 15 with outdoor. For the 11 to <16 years age group, there were 3 subjects with indoor hand-to-mouth frequency and 2 with outdoor. Tables II and III present the summary statistics of indoor and outdoor hand-to-mouth frequency, respectively, from the data in each of the individual studies.
| Study | No. of Observations | Mean | Std | P50 | P5 | P25 | P75 | P95 |
|---|---|---|---|---|---|---|---|---|
| Zartarian et al., 1998 | 4 | 3.9 | 2.1 | 3.4 | 2.1 | 2.4 | 5.4 | 6.9 |
| Reed et al., 1999 | 30 | 9.5 | 6.4 | 8.5 | 2.0 | 5.0 | 11.0 | 25.0 |
| Freeman et al., 2001 | 19 | 6.7 | 6.0 | 5.0 | 0.4 | 2.4 | 10.2 | 20.6 |
| Tulve et al., 2002 | 238 | 15.9 | 19.5 | 10.0 | 0.0 | 2.0 | 24.0 | 59.0 |
| Greene, 2002 | 342 | 18.6 | 19.1 | 12.0 | 0.0 | 5.0 | 25.0 | 57.0 |
| Hore, 2003 | 10 | 19.9 | 14.1 | 16.0 | 1.0 | 10.3 | 26.1 | 47.8 |
| Black et al., 2005 | 68 | 16.1 | 13.5 | 13.3 | 3.0 | 9.0 | 19.4 | 35.9 |
| Beamer et al., in preparation | 23 | 20.2 | 13.9 | 17.0 | 3.0 | 11.3 | 26.4 | 47.6 |
| Study | No. of Observations | Mean | Std | P50 | P5 | P25 | P75 | P95 |
|---|---|---|---|---|---|---|---|---|
| Leckie et al., 2000 | 20 | 7.2 | 13.4 | 1.8 | 0.0 | 0.3 | 7.6 | 40.1 |
| Freeman et al., 2001 | 17 | 2.0 | 3.8 | 0.0 | 0.0 | 0.0 | 2.3 | 11.9 |
| Tulve et al., 2002 | 60 | 9.3 | 10.1 | 7.0 | 0.0 | 0.0 | 14.0 | 35.5 |
| Black et al., 2005 | 46 | 9.7 | 12.7 | 4.9 | 0.0 | 1.3 | 10.1 | 40.9 |
| Beamer in preparation, 2006 | 17 | 9.2 | 8.1 | 7.1 | 0.0 | 2.1 | 14.0 | 26.0 |
No data were available in any of the studies for infants from birth to <1 month or for 1 to <3 months old. General linear models and mixed effects models were used to test for any statistical difference between studies, age category, gender, and location (indoor vs. outdoor). Summary statistics were obtained by study, age, and location. The data were merged if there was no statistical difference among studies.
2.2. Approach for Fitting Variability Distributions
To fit variability distributions, the data from the 9 studies were compiled and put into comparable units of frequency (hourly rate) as the ratio of number of hand-to-mouth contacts divided by a time interval. The denominator of that ratio was reported differently across studies: the average was ∼1 hour for outdoors and ∼3 hours for indoors. Thus, an “observation” in Tables 2 through 8 is defined as the number of reported hand-to-mouth frequency ratios, which could have different denominators. As shown in Table I, the study periods and data collection intervals also varied across studies. A unique ID was assigned to each individual, and the data were sorted into the new EPA age groupings for risk assessment (U.S. EPA, 2005) (3 to <6 months, 6 to <12 months, 1 to <2 years, 2 to <3 years, 3 to <6 years) according to the age of the child. Method of moments and maximum likelihood estimation methods were used along with visual inspection of the data, and then goodness-of-fit tests (Kolmogorov-Smirnov, Cramer-von-Mises, Anderson Darling, chi-square) were applied to verify the selection among a number of distributions, including lognormal, Weibull, and normal distributions. Analyses were also conducted on the inter- and intra-personal variability of indoor and outdoor hand-to-mouth frequency.
2.3. Approach for Fitting Uncertainty Distributions
Some exposure models (e.g., Zartarian et al., 2000, 2006; Xue et al., 2006; Burke et al., 2001) have the option of conducting single-stage Monte Carlo sampling to assess the variability in exposure or dose for a population of interest (expressed as population percentiles), or two-stage Monte Carlo sampling to evaluate uncertainty along with variability (expressed as a range of values for each percentile). Generally, an exposure model run in two-stage Monte Carlo sampling consists of generating N simulations each comprised of M iterations, which produces a family of N predicted distributions of population exposures. The entire process of a single sampling of uncertain parameters, followed by repeated sampling from the variability parameters, is referred to as a simulation (MacIntosh et al., 1995).
To generate uncertainty distributions presented in this article, the bootstrap approach described in Xue et al., 2006 was used as follows. For two-parameter distributions like the Weibull, the parameter pairs are chosen from a list generated by the following steps:
- 1
Fit a variability distribution (the “parent distribution”), estimating parameters v1 and v2 (e.g., geometric mean and geometric standard deviation), to all data from the original S studies using the method of moments.
- 2
Fit a variability distribution (using the shape of the parent distribution) to data in each of the original S studies using the method of moments, and examine the scatter plot of the S v1 and v2 values to get a sense of the scale of uncertainty.
- 3
Sample B data points from the parent distribution K different times (B is the bootstrap sample size; K is the number of samples of parameter pairs to be saved for uncertainty runs, typically 200).
- 4
For each of those K sets of B data points, fit the parent distribution and compute the parameter values of interest. This gives K (v1,v2) pairs.
- 5
Overlay the scatter plot of the K (v1,v2) pairs with the S (v1,v2) pairs obtained in Step 2.
- 6
Repeat Steps 3–5 with different values of B, until the scatter plot from Step 4 satisfactorily matches the spread seen in the scatter plot from Step 2.
Steps 5 and 6 require one to compare the input variable uncertainty with that seen in the original studies, leading to a choice for B. The size of B depends on the sample size of the original data and how well the bootstrap scatter plots fit over the data points among the various data sources (Step 5). Lower B values correspond to higher uncertainties.
Using those B values, 200 data pairs of Weibull distribution parameters were generated for each age group both for indoor and outdoor frequencies. Each pair was used for one variability run with 100 (M) variability simulations to generate 200 (N) sets of distributions (cumulative density functions) for conducting uncertainty analyses. After ranking the medians of those 200 runs, the 50th and 95th percentile variability distributions were selected (i.e., those distributions with the 50th and 95th percentile ranked medians). The 5th and 95th percentiles were calculated to get the ratio of the 95th vs. 5th for each variability run. This ratio indicates the variability (shown in the first two columns of data in Table VII). Next, the 50th and 95th percentiles of each of the 200 variability runs were selected, and the 5th and 95th percentiles were calculated for those 200 50th and 95th percentiles. The ratio of the 95th to the 5th percentile indicates the uncertainty from the selected various pairs of 200 Weibull distribution parameters (results shown in last two columns of the data in Table VII). This process for variability and uncertainty analyses was conducted for each age group and for the 2 locations (indoor/outdoor).
| Age Group | Selected Variability Runs Ratio of 95th vs. 5th Percentile | Selected Variability Percentiles Ratio of 95th vs. 5th Percentile | ||
|---|---|---|---|---|
| 50th Percentile | 95th Percentile | 50th Percentile | 95th Percentile | |
| Indoor Hand-to-Mouth Frequency (No. of contacts/hr) | ||||
| 3 to <6 months | 16 | 26 | 2 | 2 |
| 6 to <12 months | 13 | 6 | 5 | 4 |
| 1 to <2 years | 8 | 77 | 3 | 3 |
| 2 to <3 years | 23 | 138 | 3 | 5 |
| 3 to <6 years | 73 | 211 | 5 | 5 |
| 6 to <11 years | 6 | 6 | 2 | 2 |
| Outdoor Hand-to-Mouth Frequency (No. of contacts/hr) | ||||
| 6 to <12 months | 38 | 12 | 3 | 3 |
| 1 to <2 years | 113 | 20 | 2 | 2 |
| 2 to <3 years | 13 | 73 | 8 | 7 |
| 3 to <6 years | 932 | 378 | 7 | 9 |
| 6 to <11 years | 53077 | 1287 | 17 | 17 |
2.4. Approach for Assessing Hand-to-Mouth Frequency Data Needs
Determining where additional data would help improve the probability distributions presented was based on the sample size for each age group, study, and indoor/outdoor location and also on the uncertainty analysis.
3. RESULTS
Combining all those hours of behavior observation data for the 429 children ages 0.3 to 12 years across all 9 studies, it was found that age and location (indoor vs. outdoor) were important for hand-to-mouth frequency, but study and gender were not. The overall general linear model and mixed model show that results across studies were not statistically significantly different. Hand-to-mouth frequency was also not statistically different between genders. However, location (indoor vs. outdoor) and age groups are statistically significant for hand-to-mouth frequency. Summary statistics are different among studies for both indoor and outdoor hand-to-mouth frequency, mainly from age differences among those studies. As age increases, both indoor and outdoor hand-to-mouth frequencies decrease. For both indoor and outdoor hand-to-mouth frequency, interpersonal and intra-personal variability are ∼60% and ∼30%, respectively. Also, the variance difference among different studies is much bigger than its mean (averaged mean relative error in %).
Weibull distributions were found to best fit the observed data for all analyses conducted. Fig. 1 illustrates this for two of the age groups as examples. Tables IV and V present the Weibull distribution parameters and summary statistics for indoor and outdoor hand-to-mouth frequency, respectively, by age group for each of the studies. To fit Weibull distributions, a minimum sample size of 9 was used, so some subjects were not included in Tables IV and V for certain age groups because there were fewer than 9 data points. Fitting distributions by study and age group revealed that the mean and median were constant across studies by age group for indoor hand-to-mouth frequency, except for the 3 to <6 years age group where Reed et al. (1999), Greene (2002), and Tulve et al. (2002) are lower than Hore (2003) and Black et al. (2003). The same finding was observed with the outdoor hand-to-mouth frequency data.
Fits of alternative variability distributions for the indoor hand-to-mouth frequency variable.
| Study | Age Group | Distribution | Weibull Shape Parameter | Weibull Scale Parameter | Chi-Square | No. of Observations | Mean | Std | p50 | p5 | p25 | p75 | p95 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Greene, 2002 | 3 to <6 months | Weibull | 1.28 | 30.19 | Fail | 23 | 28.0 | 21.7 | 23.0 | 3.0 | 8.0 | 48.0 | 65.0 |
| Greene, 2002 | 6 to <12 months | Weibull | 1.12 | 20.67 | Pass | 88 | 19.8 | 17.8 | 14.5 | 2.0 | 7.0 | 29.0 | 50.0 |
| Tulve et al., 2002 | 6 to <12 months | Weibull | 0.40 | 5.37 | Fail | 9 | 14.4 | 23.6 | 5.0 | 0.1 | 0.1 | 15.0 | 71.0 |
| Black et al., 2005 | 6 to <12 months | Weibull | 1.30 | 20.92 | Fail | 11 | 19.1 | 17.1 | 13.7 | 3.3 | 9.3 | 19.7 | 56.8 |
| Beamer et al., in preparation | 6 to <12 months | Weibull | 1.93 | 16.34 | Pass | 11 | 14.6 | 7.9 | 17.0 | 2.0 | 9.6 | 20.8 | 26.4 |
| Greene, 2002 | 1 to <2 years | Weibull | 0.90 | 19.21 | Pass | 132 | 20.1 | 21.0 | 14.0 | 0.1 | 5.5 | 26.0 | 65.0 |
| Tulve et al., 2002 | 1 to <2 years | Weibull | 0.82 | 16.98 | Fail | 84 | 18.5 | 19.2 | 14.0 | 0.1 | 5.0 | 27.0 | 54.0 |
| Black et al., 2005 | 1 to <2 years | Weibull | 1.94 | 19.35 | Pass | 20 | 17.1 | 9.6 | 14.6 | 4.7 | 9.4 | 23.6 | 35.4 |
| Greene, 2002 | 2 to <3 years | Weibull | 0.75 | 11.61 | Pass | 87 | 13.6 | 16.3 | 8.0 | 0.1 | 2.0 | 20.0 | 39.0 |
| Tulve et al., 2002 | 2 to <3 years | Weibull | 0.58 | 8.66 | Pass | 41 | 12.1 | 13.6 | 8.0 | 0.1 | 2.0 | 15.0 | 36.0 |
| Black et al., 2005 | 2 to <3 years | Weibull | 1.24 | 12.28 | Pass | 24 | 11.6 | 8.7 | 10.5 | 1.6 | 3.8 | 16.2 | 31.5 |
| Reed et al., 1999 | 3 to <6 years | Weibull | 1.64 | 11.32 | Pass | 25 | 10.1 | 6.6 | 9.0 | 2.0 | 5.0 | 13.0 | 25.0 |
| Greene, 2002 | 3 to <6 years | Weibull | 1.01 | 11.79 | Fail | 12 | 11.8 | 9.8 | 10.0 | 0.1 | 3.5 | 17.5 | 30.0 |
| Tulve et al., 2002 | 3 to <6 years | Weibull | 0.61 | 11.06 | Pass | 104 | 15.5 | 21.2 | 7.5 | 0.1 | 2.0 | 22.0 | 62.0 |
| Hore, 2003 | 3 to <6 years | Weibull | 1.90 | 24.91 | Pass | 9 | 22.0 | 13.2 | 19.4 | 8.9 | 11.0 | 26.1 | 47.8 |
| Black et al., 2005 | 3 to <6 years | Weibull | 1.29 | 22.47 | Fail | 13 | 20.5 | 20.3 | 14.8 | 4.9 | 9.8 | 22.5 | 83.9 |
| Freeman et al., 2001 | 6 to <11 years | Weibull | 1.26 | 7.25 | Pass | 12 | 6.7 | 6.0 | 4.2 | 1.7 | 2.3 | 10.2 | 20.6 |
| Study | Age Group | Distribution | Weibull Shape Parameter | Weibull Scale Parameter | Chi-Square | No. of Observations | Mean | std | p50 | p5 | p25 | p75 | p95 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Tulve et al., 2002 | 1 to <2 years | Weibull | 1.14 | 13.74 | Pass | 10 | 13.3 | 9.7 | 11.0 | 0.1 | 8.0 | 17.0 | 35.0 |
| Black et al., 2005 | 1 to <2 years | Weibull | 1.05 | 12.90 | Fail | 12 | 12.6 | 13.7 | 7.0 | 1.3 | 4.2 | 15.6 | 42.2 |
| Tulve et al., 2002 | 2 to <3 years | Weibull | 0.49 | 2.99 | Pass | 17 | 5.5 | 9.1 | 4.0 | 0.1 | 0.1 | 7.0 | 38.0 |
| Black et al., 2005 | 2 to <3 years | Weibull | 0.59 | 3.56 | Pass | 20 | 5.3 | 8.2 | 2.2 | 0.1 | 0.4 | 6.7 | 26.6 |
| Leckie et al., 2000 | 3 to <6 years | Weibull | 0.53 | 1.47 | Fail | 9 | 2.5 | 3.8 | 1.1 | 0.1 | 0.1 | 3.0 | 11.5 |
| Tulve et al., 2002 | 3 to <6 years | Weibull | 0.63 | 7.90 | Pass | 33 | 10.1 | 10.4 | 7.0 | 0.1 | 2.0 | 16.0 | 36.0 |
| Black et al., 2005 | 3 to <6 years | Weibull | 0.51 | 6.09 | Fail | 10 | 10.4 | 14.9 | 5.8 | 0.1 | 0.1 | 8.7 | 40.9 |
| Freeman et al., 2001 | 6 to <11 years | Weibull | 0.45 | 1.02 | Fail | 12 | 2.6 | 4.4 | 0.1 | 0.1 | 0.1 | 3.5 | 11.9 |
Table VI presents the Weibull distribution parameters and summary statistics for indoor and outdoor hand-to-mouth frequency by age group, with data pooled from all studies. This table illustrates that hand-to-mouth frequency generally decreases as age increases. Table VII presents the results of the uncertainty analysis. The uncertainty was large for age groups 6 to <11 years old and 3 to <6 years old, as can be seen by studying the ratio of the 95th percentile to the 5th percentile. The results in Table VII also reveal that the variability is much bigger than the parameter uncertainty. In comparison with other age groups, the 3 to <6 year old group had large uncertainty both for indoor and outdoor hand-to-mouth frequency. The 6 to <11 year age group had very large uncertainty for outdoor hand-to-mouth frequency. The age group of 6 to <12 months also had large uncertainty for indoor hand-to-mouth frequency. These high uncertainties are due to small sample sizes and/or differences across studies.
| Age Group | Distribution | Weibull Shape Parameter | Weibull Scale Parameter | Chi-Square | No. of Observations | Mean | Std | p50 | p5 | p25 | p75 | p95 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| INDOOR | ||||||||||||
| 3 to <6 months | Weibull | 1.28 | 30.19 | fail | 23 | 28.0 | 21.7 | 23.0 | 3.0 | 8.0 | 48.0 | 65.0 |
| 6 to <12 months | Weibull | 1.02 | 19.01 | pass | 119 | 18.9 | 17.4 | 14.0 | 1.0 | 6.6 | 26.4 | 52.0 |
| 1 to <2 years | Weibull | 0.91 | 18.79 | fail | 245 | 19.6 | 19.6 | 14.0 | 0.1 | 6.0 | 27.0 | 63.0 |
| 2 to <3 years | Weibull | 0.76 | 11.04 | fail | 161 | 12.7 | 14.2 | 9.0 | 0.1 | 2.9 | 17.0 | 37.0 |
| 3 to <6 years | Weibull | 0.75 | 12.59 | pass | 169 | 14.7 | 18.4 | 9.0 | 0.1 | 3.7 | 20.0 | 54.0 |
| 6 to <11 years | Weibull | 1.36 | 7.34 | pass | 14 | 6.7 | 5.5 | 5.7 | 1.7 | 2.4 | 10.2 | 20.6 |
| OUTDOOR | ||||||||||||
| 6 to <12 months | Weibull | 1.39 | 15.98 | pass | 10 | 14.5 | 12.3 | 11.6 | 2.4 | 7.6 | 16.0 | 46.7 |
| 1 to <2 years | Weibull | 0.98 | 13.76 | pass | 32 | 13.9 | 13.6 | 8.0 | 1.1 | 4.2 | 19.2 | 42.2 |
| 2 to <3 years | Weibull | 0.56 | 3.41 | fail | 46 | 5.3 | 8.1 | 2.6 | 0.1 | 0.1 | 7.0 | 20.0 |
| 3 to <6 years | Weibull | 0.55 | 5.53 | fail | 55 | 8.5 | 10.7 | 5.6 | 0.1 | 0.1 | 11.0 | 36.0 |
| 6 to <11 years | Weibull | 0.49 | 1.47 | fail | 15 | 2.9 | 4.3 | 0.5 | 0.1 | 0.1 | 4.7 | 11.9 |
Fig. 2 illustrates graphically the uncertainty distributions for the 1 to <2 year and 3 to <6 year age groups examined, corresponding to the variability distributions in Fig. 1. The uncertainty clouds covered all the studies well by age group. The B value ranged from 10 to 15, which indicates that we have reasonable data to define uncertainty. As discussed for Table VII, the uncertainty was large for age groups 6 to <11 years old and 3 to <6 years old.
Uncertainty bootstrap distributions for indoor hand-to-mouth frequency (contacts per hour).
Fig. 3 illustrates graphically the uncertainty cumulative density functions (cdfs) for 3 selected variability percentiles (5th, 50th, and 95th) of indoor hand-to-mouth frequency for 1 to <2 year olds and 3 to <6 year olds, respectively. For those 200 uncertainty runs, 200 5th percentiles form the bottom CDF, 200 50th percentiles form the middle CDF, and 200 95th percentiles form the top CDF. The 5th percentiles are not stable since the values are usually very small and sometimes even zero. For the 1 to <2 year old age group, the 95th and 5th percentiles for the middle CDF are 22 and 7 contacts per hour with ratio 3; the 95th and 5th percentiles for the top CDF are 101 and 30 with ratio 3. For the 3 to <6 year old age group, the 95th and 5th percentiles for the middle CDF are 16 and 3 with ratio 5; the 95th and 5th percentiles for the middle CDF are 106 and 21 with ratio 5. These results indicate that the uncertainty is larger for the 3 to <6 year age group than for the 1 to <2 year age group.
CDFs for 3 selected variability percentiles.
4. DISCUSSION
Children's hand-to-mouth behavior is difficult to measure for several reasons. Some of these reasons include the following: children's contacts with surfaces and objects are frequent and intermittent; observational studies are labor-intensive for data collection and data analysis; and data analysis can be subjective. Interpretation of the results is also difficult. Some researchers express mouthing behavior in terms of frequency of occurrence (e.g., contacts per hour or contacts per minute). Others express mouthing behavior as an exposure period (i.e, minutes). This discrepancy makes it more difficult to compare results among studies. To conduct the meta-analysis presented in this article, investigators of individual studies were contacted to provide, and in some cases reanalyze, their original study data by age, gender, and location. The uncertainty analyses in this article only focus on parameter uncertainty, and do not account for other uncertainties from differences in study approaches and their associated reliability and validity issues.
The meta-analysis revealed that there was not a statistically significant difference in hand-to-mouth frequency between genders or across studies. However, there was a statistically significant difference in hand-to-mouth behavior with regard to location (indoor versus outdoor) and age groups. Hand-to-mouth frequency indoors was consistently higher for all age groups than outdoor hand-to-mouth frequency. The highest difference was observed for children 2 years old with an indoor hand-to-mouth frequency of 2.4 times higher than outdoors. As age increases, both indoor and outdoor hand-to-mouth frequencies decrease. These findings of differences between mouthing behaviors indoors versus outdoors and across ages are consistent with those discussed in some of the individual studies (Freeman et al., 2001; AuYeung et al., 2004; Tulve et al., 2002). More reliable estimates of daily mouthing frequency should take into account these differences due to location and age, as well as longitudinal estimates of mouthing behavior.
Because children can be practically observed or videotaped for only several hours in a day, available hand-to-mouth frequency data are typically cross-sectional and short term. However, exposure assessors are often interested in estimating chronic exposures. Exposure assessors typically extrapolate mouthing behaviors from several hours of observations to daily estimates using data or assumptions on the amount of time children are awake during the day and not otherwise eating. Thus, future research on children's mouthing behavior would also ideally examine longitudinal estimates within a day and over time.
The analyses in this article represent a first effort to fit hand-to-mouth frequency distributions by indoor/outdoor location, and by age using the new U.S. EPA Guidance on Selecting Age Groups for Monitoring and Assessing Childhood Exposures to Environmental Contaminants (U.S. EPA, 2005). Thus, the results presented here can be used to enhance future exposure and risk assessments. Despite differences in the available studies for hand-to-mouth frequency, the results were generally consistent across studies. This is the first comprehensive analysis to report Weibull distributions as the best-fitting distribution for hand-to-mouth frequency.
This article also highlights the need for more research on mouthing behavior for children at various stages of development. For example, no data were identified for infants between birth to <1 month old or for 1 to < 3 months old. More data are also needed for the 3 to <6 year age group because the uncertainty in the available data was large. Because collection of micro-activity data is intrusive, expensive, and labor-intensive, any new data collection efforts should use a standardized protocol for micro-activity data gathering and analysis and be carefully designed so that they can be combined more readily across studies.
This article has focused on frequency of hand-to-mouth contact. Future research could also include collection and analyses of available data for hand- and object-to-mouth duration; frequency of hand-to-object, hand-to-surface, other-body-part-to-surface, and object-to-mouth contact; and surface area of objects mouthed, to assist with estimation of dermal and nondietary ingestion exposures.
ACKNOWLEDGMENTS
We gratefully acknowledge the following individuals for providing assistance on this article: James Leckie of Stanford University; Celestine Kiss of the Consumer Product Safety Commission; and Paul Lioy of the Environmental and Occupational Health Sciences Institute.
DISCLAIMER
The U.S. Environmental Protection Agency through its Office of Research and Development funded and managed the research described here. It has been subjected to Agency's administrative review and approved for publication as an EPA document.
