Ann. occup. Hyg., Vol. 46, No. 2, pp. 209-217, 2002
© 2002 British Occupational Hygiene Society
Published by Oxford University Press
Article |
A Dirichlet Tessellation-based Sampling Scheme for Measuring Whole-body Exposure
Health and Safety Executive, Magdalen House, Stanley Precint, Bootle L20 3QZ; Health and Safety Laboratory, Broad Lane, Sheffield S3 7HQ, UK
Received 22 December 2000; in final form 22 October 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Dermal sampling can be conducted using small pads or patches attached to various areas of the skin or clothing, or by using a whole-body coverall. Both techniques are recognized standardized methods for collecting chemicals. Patch sampling is simple to perform and inexpensive to analyse compared with an entire overall, but may require some user intervention. Extrapolation from a small sampled area to the total body area can lead to inaccurate estimates of total body exposure because of a lack of uniformity of deposition. Whole-body overall analysis eliminates the problems associated with using patches and gives a more accurate estimate of total body exposure. Therefore, if it were possible to measure the whole-body overall accurately and quickly, we would have a better assessment of dermal exposure. In this study we develop a working protocol using a standardized approach, to measure the contamination over an entire overall. The protocol takes into account size differences and establishes a reproducible pattern of sampling in order to map the distribution of contamination over each overall. The working protocol has been applied to 10 overalls collected from companies using copper-based biocides. A portable X-ray fluorescence spectrometer (PXRF) was used to measure the copper in the biocide. The exposure estimate from the PXRF results uses an averaging scheme based on the Dirichlet tessellation of the sampling locations. This allows unbiased estimates to be obtained from a complex sampling scheme that allocates more measurements to areas of high exposure. The Dirichlet tessellation method has been compared to the patch sampling method and the conventional total digestion of the entire overall method. Using the whole-suit digestion method as the benchmark, exposures ranged from 92.0 to 5848.5 mg. Mean absolute percentage errors (from the benchmark acid digestion of the whole suit) varied from ~20% for the Dirichlet-based PXRF method to 60% for the patch methods. The patch methods underestimated the true dermal exposure (–28 to –82% for acid digestion of the patches). Analysis of this data indicates that the Dirichlet PXRF method gives a more accurate estimate of whole-body contamination than the patch method. Furthermore, the 104 measurements give a much greater spatial resolution to the exposure data than analysis of the whole overall or patches by inductively coupled plasma-atomic emission mass spectrometry (ICP-AES). This detailed knowledge of the pattern of deposition on the body is of potential importance in chemical risk assessments.
Keywords: patch sampling; dermal; overalls; exposure; Dirichlet tessellation; portable X-ray fluorescence spectrometry
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The Biocide and Pesticides Assessment Unit (BPAU) of the Health & Safety Executive (HSE) registers various classes of non-agricultural pesticide and biocide products for use in the UK. The approvals are subject to periodic review under the Control of Pesticides Regulations 1986. Risk assessments to human health form part of the review, and estimates of dermal and inhalation exposure inform the risk assessments. This enables a realistic assessment to be made of the risk involved in using these products to the operator, consumer and the environment.
The skin is an important route to systemic dose and may itself be a target organ for some chemicals, particularly pesticides and biocides. For the dermal route, just like the inhalation route, the concentration of contaminant and the duration and frequency of exposure are important elements to be considered when assessing risk. Dermal exposure results from direct and indirect contact with the product. Assessing dermal exposure is always more difficult and intrusive than inhalation exposure. In inhalation exposure, the assessments rely on collecting samples from a well-defined location, i.e. the breathing zone of the individual worker. For dermal exposure to be assessed accurately the assessments must be in terms of the whole body. The variability in dermal exposure can be very large, with some parts of the body, such as hands and knees, being more prone to contamination than other parts. Dermal exposure can be estimated using a range of techniques: direct removal of the contaminant from the skin by wiping or washing; recovery of contaminants from clothing, termed ‘whole-body dosimetry’; recovery of contaminant from patches, known as ‘patch sampling’; using gloves; visualization of contamination using a fluorescent marker; and biological monitoring (EH74/3). Each method has its relative merit and advantages and disadvantages. No one method assesses dermal exposure completely. Of the methods mentioned, patch sampling is the most commonly used.
Patch sampling techniques have been around since the late 1950s for monitoring pesticides. A varying number of patches, of fixed size (10 x 10 cm), are positioned at representative points on the body. The amount of contaminant on the patch is measured and then scaled up to provide representative estimates of exposure for each body part. Patches have several advantages, including their convenient size, portability, low cost and passive sampling, which eliminates the need for pumps that require calibration and maintenance. There are many opinions as to the best place and way to use patches. A comprehensive treatment of the use of patches for sampling has been published (Popendorf and Ness, 1994). The main method used by HSE’s Health Directorate in assessing dermal exposure to both pesticides and biocides is a modified seven-patch method originally developed by the World Health Organization (WHO, 1982). This method is described in the HSE’s Method for the Determination of Hazardous Substances MDHS 94.
The primary limitation of using absorbent patches is the assumption that exposure is uniform over the various body parts. This has been shown (Fenske, 1990) not to be the case. Therefore, a potential source of error is the extrapolation from concentrations measured on relatively small surface areas compared to the entire body surface area.
The whole-body dosimetry method can be used to determine if there is significant skin exposure and to what areas of the body it is occurring. Disposable whole-body overalls have been recommended as a collection medium for contaminants during spraying operations and a procedure has been developed (WHO, 1982). In theory, all contaminants that contact the skin are collected. The method differs from the patches in that it does not require a surface area extrapolation. The problem with the method is that it is both time consuming and expensive, as each overall has to be cut down into manageable pieces for chemical disposition.
Work has shown that there is a correlation between the whole-body overall and patch sampling method for the determination of potential dermal exposure for spraying activities. The WHO patch method represents ~8% of the body surface and the modified HSE’s method 3%. Patch sampling gives an overestimate or underestimate of total deposit; but the result according to Tannahill (1997) is a true representation of total body exposure.
This study sought to investigate the potential of using a statistical approach to measure the total contamination on workers’ overalls.
The use of the portable X-ray fluorescence spectrometer (PXRF) to map the distribution of a copper-based biocide had previously been validated in an internal feasibility study entitled ‘Mapping of biocide contamination on overalls’ (Foster et al., 1999). In this study the PXRF had been successfully demonstrated to give a good estimate (within 25% of acid digestion of the whole suit) of contamination on a single overall, when compared to conventional acid digestion and analysis by inductively coupled plasma-atomic emission mass spectrometry (ICP-AES). However, the method used to demonstrate the distribution was complex, time consuming and non-reproducible. By applying the Dirichlet tessellation (Stoyan et al., 1992; Moller, 1994) and establishing clear and precise protocols, we were able to map the distribution of biocide contamination over the entire overall, within 3 h, compared to a day and a half for conventional acid-digestion and ICP-AES analysis.
There were three main objectives to the study. The first was to establish a working standardized protocol that could be used to measure biocide contamination over the entire overall. The protocol took into account the different sizes of the overalls and established a set pattern for cutting and sampling the distribution of contaminant over the overall with the minimum number of measuring points. The second objective was to validate the working protocol on 10 contaminated overalls: nine obtained from treating salmon fishing nets with biocide and one from boat painting. The method was compared with the patch sampling method and whole-suit acid digestion method. The third objective was to produce an easy to understand colour contour map of the biocide contamination over the overalls as a simple method of presenting the data.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Sampling scheme
A working protocol has been devised to provide a standard PXRF sampling regime to measure the contamination over the entire overall. The protocol took into account different sizes of overall to establish a reproducible pattern of sampling, and specified the number of sample points (104) to map the distribution of contamination over each overall. The total number of sample points took into account the practicality and time needed actually to make the measurements. The overalls were all cut exactly the same way into six sections: head, left arm/upper body, right arm/ upper body, midriff, left leg/lower torso and right leg/ lower torso. For each section and for each size of overall, templates were made with measuring portals cut into them. The PXRF could be located over these measuring portals and a reading could be taken in exactly the same position on each section and overall, independent of size. Briefly: 26 measurements were taken on each leg, six on the midriff (between the waist and bottom of the rib cage), 21 on each arm and corresponding half of the upper torso and four on the hood. This split of sampling positions was decided a priori based on previous experience and the spatial findings of a meta-analysis of previously published HSE dermal exposure data (Llewellyn et al., 1996; Garrod et al., 1998, 1999, 2000; HSE, 1999). Within each section of the overall, sampling positions were not regularly spaced. Instead, a greater density of measurements was decided upon in areas perceived to be at greater risk of dermal exposure, such as the front of the thighs and forearms. Correspondingly, fewer measurements were taken in areas perceived to be low risk, e.g. back of legs and upper back. The devised protocol was applied to 10 overalls contaminated with a copper-based biocide.
Copper-biocide contaminated overalls
Tyvec Protech Chemical Protective Clothing, the classic model suits were used. The suits were full-body disposable overalls with hoods in white Tyvek 131N material, manufactured by Dupont Nonwovens, Luxembourg. These suits were given to volunteer workers in two companies that treated salmon fishing nets with a copper-biocide. In company A, three volunteers used a water-based biocide called Netrex AF (Mobile Ltd, UK). In company B, six volunteers used a xylene solvent-based biocide called Hempanet 1750 (Hempal Paints Ltd, UK). A further volunteer was recruited from a company that painted the underside of a boat. This volunteer used a xylene solvent-based biocide called Tiger Cruising anti-foul, colour souvenirs (Blake Marine Paints, UK). A total of 10 copper-biocide contaminated overalls were used to validate the technique.
Patch sampling
Normally, in the assessment of dermal exposure to biocides the modified seven-patch method is used. This technique was originally developed by WHO (WHO, 1982) and is described in the HSE’s Method for the Determination of Hazardous Substances MDHS 94: Pesticides in Air and/or on Surfaces. The method uses 10 x 10 cm gauze sampling patches which are attached to standard positions on the operator’s body: head, sternum, right forearm, left thigh, left ankle and back. Instead of using gauze patches, areas were marked on each of the overalls to represent where the patch samples would have been placed. These patch locations enabled the exact same position on each overall to be used in the study. The patches were analysed in two ways: first, they were analysed using the PXRF and, second, they were cut from the overalls and, in the conventional way, acid digested and analysed by ICP-AES.
The patch locations were measured using the PXRF by dividing the patch location into nine segments. The reason for dividing the patch location into nine was to accommodate the size of the PXRF’s analysis window, which is 2.5 cm in diameter. The biocide concentration of the patch location was calculated by averaging the readings of the nine segments.
Measurement with PXRF
The PXRF was used in the study to measure copper in both the patches and the whole suits. The PXRF has the advantage of being able to measure copper in situ and requires no sample preparation. The PXRF uses gamma rays in three radioactive sources to excite the copper on the surface of the overalls. To give the maximum accuracy and the minimum analysis time, the three sources were set to the following times: Cd109 40 s, Fe55 10 s and Am241 10 s. This makes the analysis of the whole suit very quick. It is necessary to make readings with all three sources, as the instrument compensates for interference effects of other elements present and matrix effects are compensated and eliminated by the use of a thin film application. The detection window of the PXRF is 2.5 cm in diameter (area 4.91 cm2). The lower limit of detection for copper using the PXRF was 0.65 µg/cm2. Values below this limit of detection were set to zero.
Measurement with ICP-AES
Samples to be analysed by ICP-AES must first be acid-digested to release the copper from the suit. The suits used are acid resistant and consequently are not significantly attacked by the treatment with hot nitric acid, permitting the biocide to be acid-leached from the surface. The lower limit of detection for copper using the ICP-AES was ~0.1 µg/ml.
Acid digestion of 100 cm2 patches for ICP-AES analysis
The procedure for the 100 cm2 patches consisted of cutting each patch into smaller squares and placing all these squares into a 150 ml beaker. To each beaker, 10 ml of concentrated nitric acid was added. The beakers were covered and placed on a hot plate set at 120°C for 15 min, swirling or tamping periodically to ensure continuous immersion of the fabric. In the case of copper it was easy to see the leaching occurring as the copper biocide is a red colour and as leaching occurred the acid took on a blue coloration. Once the colour change had occurred, 50 ml of de-ionized water was added to the beaker and each beaker was heated for a further 15 min. When cooled, the leached fabric was removed and the remaining liquid made up to 100 ml with de-ionized water. The mixture was allowed to stand for 1 h to allow the suspended matter to settle out. Once settled, 40 ml was removed and centrifuged to ensure no fibres were present. The solution was then analysis by ICP-AES.
Acid digestion of whole-suit panels for ICP-AES analysis
The procedure for the whole suit panels consisted of further cutting the panels into smaller manageable pieces, keeping the relative sections together (head, left arm/upper body, right arm/upper body, midriff, left leg/lower torso, right leg/lower torso). Each panel was placed into a 1 l beaker and 80 ml of concentrated nitric acid was added. The beakers were covered and placed on a hot plate set at 130°C for 30 min, swirling or tamping periodically to ensure continuous immersion of the fabric. Once the red of the copper biocide had disappeared, 500 ml of de-ionized water was added to the beaker and each beaker was heated for a further 30 min. When cooled, the leached fabric was removed and the remaining liquid made up to 800 ml with de-ionized water. The mixture was left to stand for 1 h to allow the suspended matter to settle out. Once settled, 40 ml was removed and centrifuged to ensure no fibres were present. The solution was then analysis by ICP-AES.
Calculations and statistics
Comparisons of the masses of copper determined by the summation of PXRF results were made, with the masses determined by conventional acid digestion and analysis by ICP-AES of patches and of the entire overall.
Three different methods of calculating dermal exposure from the PXRF results have been used.
The first takes the mean of the 104 PXRF readings and multiplies this by the calculated surface areas of the various sized overalls [medium 27 775 cm2, large 29 380 cm2, extra large (XXL) 33 910 cm2].
The second calculation averages the PXRF measurements over six separate body part regions and multiplies by each region’s area. Whole-body exposure is then based on the summation of these results. Table 1 lists the areas of the six body parts for each of the different sized suits, along with the number of associated PXRF readings.
|
The third method takes a weighted sum of the PXRF readings, where each PXRF result is multiplied by the area of its ‘tile’ in the Dirichlet tessellation of the overall taken with respect to the positions of the PXRF measurements. The Dirichlet tessellation is a geometric construction that divides a space populated by n points into n regions, each corresponding to all locations closer to a particular point than any other. If the space is convex then these regions will be connected and commonly referred to as tiles. Comprehensive reviews of the properties and applications of Dirichlet tessellations have been published (Stoyan et al., 1992; Moller, 1994). An alternative name for the tiles is Theissen polygons after an early application (1911) in meteorology, where an estimate of rainfall over a region in Switzerland was to be obtained from spatially irregular rain gauges. This corresponds to estimating exposure at a location on an overall by taking the closest PXRF measurement to that location. Total body exposure is then calculated by integrating over the entire overall. This third estimate is the main PXRF method under consideration in this study and will be referred to as the Dirichlet PXRF method.
The tessellation has been calculated using the software package MATLAB. This is only capable of constructing the tessellation over planar regions and so the tessellation has been calculated over sections of the cut-up overall. Extreme care has been taken to account for edge effects to ensure that the tessellations on separate sections of the overall match up along their common edges. With the arms and legs having been cut open along their inside seams, there is an association between some of the edges of the arm and leg sections of the overalls. To account for this toroidal effect, ‘virtual’ PXRF sample points were added to the tessellation outside the boundaries of the overall sections to mimic this effect in two dimensions. This geometric manipulation is necessary as there are currently no software packages capable of calculating the Dirichlet tessellation over the surface of general three-dimensional objects.
Figure 1 shows the Dirichlet tessellation over the left arm (and upper left body) of a large-sized overall; it also shows the marked PXRF sampling locations. Some tiles appear to have no corresponding PXRF sampling location. This is because these tiles are along the seam of the overall that has been cut. These ‘orphaned’ tiles are part of the same tiles that appear on the opposite side of the overall panel.
|
Both Pearson correlation coefficients and Spearman rank correlation coefficients have been used to measure the agreement between the methods. Because hypothesis tests using the Pearson correlation coefficients were not performed, it is not necessary to assess assumptions about the underlying distributions of the data. Presented are the percentage errors by which the methods differ from the benchmark values obtained through whole-suit acid digestion. Additionally, for each technique (other than whole-suit acid digestion) the mean absolute percentage error is given. Wilcoxon signed-rank tests have been performed to detect differences between the (main) measurement methods. No attempt has been made to discern a statistically significant difference between the three whole-suit PXRF methods. These estimates of dermal exposure are based upon the same experimental data and their values are inherently different due to their differing methods of calculation. The assumptions made about the spatial distribution of exposure that were the basis for the non-homogeneous density of sampling locations have been assessed by calculating the Spearman rank correlation coefficients between the PXRF results and their corresponding areas in the Dirichlet tessellation. Negative correlations would indicate higher exposures in areas with higher densities of sampling locations.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Comparison between methods
Potential dermal exposures to antifouling products varied considerably between individuals. Table 2 gives the estimates of potential dermal exposure arising from the different measurement and calculation methods. Using acid digestion of the whole suit as a benchmark, exposures ranged from 92.0 to 5848.5 mg, with a mean of 1796.5 mg. All other measures of dermal exposure were highly correlated with the results of acid digestion of the whole suit, but performance varied between methods.
|
Pearson correlation coefficients for each of the whole-suit PXRF methods with the results of acid digestion are ~0.99, while acid digestion of patches and the PXRF ‘patch’ method have correlation coefficients of 0.93 and 0.96, respectively. These good linear agreements somewhat disguise the true performance of the various methods. Mean absolute percentage errors (from the benchmark acid digestion of the whole suit) varied from ~20% for the Dirichlet PXRF method to 60% for the two patch methods. The two patch methods consistently underestimated true dermal exposure (–28 to –82% for acid digestion of patches) and, whilst a scaling factor could be added to improve the fit, this has already been done in the protocol for the HSE patch-based method (a factor of one-half). This factor resulted from a previous study (Tannahill et al., 1996), which suggested that the patch-based method overestimated dermal exposure.
Wilcoxon signed-rank tests found significant differences between acid digestion of the whole suit and both acid digestion of patches and the PXRF patch method (both P-values = 0.002). No significant difference was found between the two patch methods (P = 0.770), or acid digestion of the whole suit and the Dirichlet method (P = 0.322).
Of the three PXRF suit methods, our preferred measure is that based on Dirichlet areas. This should theoretically produce more accurate estimates of potential dermal exposure than either of the other two estimates based on the PXRF measurements and this has been borne out in practice by the results. Whilst the correlation coefficients are almost identical for all three methods (
0.99), the mean absolute error of 20.4% is lower for that based on Dirichlet areas than the other two methods (24.5% where the PXRF results are averaged over each body part, 30.1% where averaged over the whole suit). Seven (out of 10) measurements using the Dirichlet weighted method are within 12% of their corresponding value obtained by acid digestion of the whole suit. The other three measurements were all consistently higher than the benchmark method. Interestingly, the seven within 12% were all xylene solvent-based copper biocide; the other three were water-based copper biocide. The difference could be with the acid digestion method not releasing the water-based copper from the suits, resulting in an underestimate of copper on the suit and an overestimate when measurements are made with the PXRF. This needs to be investigated further, but does not effect our conclusions. Figure 2 plots the estimates obtained using the Dirichlet PXRF method and acid digestion of patches against the values obtained through acid digestion of the whole suits. A line representing a one-to-one correspondence between the estimates and the ‘true’ value (acid digestion of the whole suit) has been added to this figure.
|
Spatial distribution of exposure
Table 3 gives the estimates of dermal exposure obtained by acid digestion and the whole-suit Dirichlet PXRF method by body part. The split between lower and upper body has been made at a level around the bottom of the ribcage. Because a single panel was digested covering the midriff between the waist and the chest, the ICP-AES results for these panels were divided equally between the four body parts. This clearly will have led to some error in the body part results for acid digestion, but the level of exposure on this panel was low in comparison with other parts of the body so the effect should be minimal. The hood results are not included in the two upper body exposures.
|
Agreement by body part is good: 24 (out of 40) of the PXRF estimates of exposure were within 25% of their corresponding ICP-AES result. Average absolute errors were 22.7, 45.7, 23.1 and 38.9% for the upper left (UL), upper right (UR), lower left (LL) and lower right (LR) parts, respectively. The two highest values being heavily affected by values of +216.5% for the upper right section of overall 7 and +102.3% for the lower right section of overall 8, which are both overalls contaminated with the water-based biocide. Correlation between body parts ranged from 0.96 to 1.00. This is a poorer agreement than for the whole overall, but this is to be expected—the additional ‘averaging’ of errors over the whole body provides an improved agreement.
Table 2 gives Spearman rank correlation coefficients calculated from the PXRF results (surface concentration of copper in µg Cu cm–2) and the areas of the tiles in the Dirichlet tessellation that correspond to each PXRF reading. All correlations are negative and, though only modest (–0.05 to –0.45), six are individually significant at a 5% level and collectively provide strong evidence of higher exposures in areas where the sampling density was highest (and therefore the tessellation has tiles of smaller area), reaffirming the logic of the sampling scheme.
Figure 3 shows a three-dimensional representation of the contamination on overall 2. This clearly indicates the high exposure experienced on the front of the thighs, inside of calves and the inside sleeves. Similar patterns of exposure were observed on the overalls of the other workers. The image has been constructed by assuming uniform exposure over each of the Dirichlet tiles (equal to the corresponding PXRF measurement) and ‘mapping’ each of these tiles onto a three-dimensional mannequin.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The whole-suit Dirichlet PXRF method samples a total of 510 cm2 (1.5–1.8% of the total suit area, depending upon the suit size), whilst acid digestion of patches samples 600 cm2 of the available surface area. With PXRF measurements taken all over the body, no two measurements are closer than 7.5 cm, while with the patch-based methods the sampled area is concentrated in a small number (six) of locations. It is not surprising therefore that the resulting estimate of total body exposure is superior using the PXRF-based methods. Consider each patch partitioned into a number of disjoint areas (each patch has the surface area of ~20 PXRF readings). Then the acid digestion patch results can be considered as the aggregate of the exposures over these (sub) areas. As these (sub) areas are adjacent to one another, their exposures will be highly correlated and consequently there is a loss of ‘information’ in comparison with a scheme that samples a similar surface area but is spread over many localities.
A comparison study of whole-suit acid digestion and patches (Tannahill et al., 1996) analysed 22 suits by the two methods. Three different work sectors were investigated: five pairs of measurements were from spray application of pesticides; eight from timber preservation; and nine from the spray application of EAF paints. The correlation between the two methods for their data (
= 0.94) was similar to the correlation reported here. Average absolute percentage error, after applying a scaling factor of one-half, was 43.2%—an improvement over the 59.9% achieved in this study. Nevertheless, this is still considerably below the accuracy achieved by the PXRF Dirichlet technique. Only four (out of 22) of the patch estimates of exposure are within 10% of the value obtained by acid digestion of the whole suit. Within each industry sector the relationship between the two estimates varied. For pest control and timber treatment, the patch-based method gave results that were on average approximately twice those of the suit, while for EAF painting, the patch method underestimated acid digestion of the suits.
Of the three methods of calculating whole body exposure from the PXRF results, the one using the areas from the Dirichlet tessellation is the most accurate. This is in agreement with what we would theoretically expect. Dermal exposure is not uniform over the entire body. Areas such as the front of the thighs, the inside of the ankles and lower legs, and the forearms have much higher levels of dermal exposure than the back of the legs and torso. Consequently, our sampling scheme devotes more PXRF readings to these high-exposure parts of the body and correspondingly fewer sampling points to those parts of the body where we suppose that dermal exposure will be low. Because of this unrepresentative sampling scheme, an estimate based on the mean of the PXRF readings (equivalent to appropriating equal areas to each measurement) will be biased to give estimates of total body exposure that are too high. This bias may be reduced, but not eliminated, by averaging over body parts instead of the whole suit. This bias will only disappear when the pattern of dermal exposure is uniform over the body. Here, uniform exposure does not mean that any individual having completed some job has even dermal exposure over their whole body. Instead, it implies that there is no systematic tendency for particular body parts to have consistently higher or lower exposure than other parts of the body. The accuracy of whole-body estimates is not the only advantage of the PXRF method. With 104 measurements taken over the exterior of an overall, a high level of spatial resolution is achieved. However, to produce images showing the distribution of exposure requires a method of extrapolation from the PXRF results to all parts of the overall. The Dirichlet tessellation provides an intuitive means of doing this, whereas the other two PXRF methods sacrifice spatial resolution by averaging the results over the entire body and six body parts, respectively.
The performance of any sampling scheme adopted for PXRF measurements will depend upon the distribution of exposure over the body. We have chosen the positions of the PXRF measurements to produce a greater sampling density over parts of the body that are generally more prone to high exposure. This should, in general, provide more accurate estimates. Estimates of exposure for tasks that have an unusual pattern of exposure, such as predominantly over the back, will be correspondingly less accurate.
The large amount of data produced by the Dirichlet PXRF method suggests further analysis that should be carried out. Of particular merit would be a more detailed investigation of the spatial distribution of exposure in order to determine which areas of the body are at high risk of exposure and to determine the correlation structure between the PXRF measurements. The data should help to illuminate the effect that the number of sample points has on the accuracy of the resulting estimates. Sub-samples of varying sizes could be drawn from the original 104 measurements and estimates recalculated for each. This approach is limited somewhat in that for all possible sub-samples of size 50 (say), the restriction of having to use locations that were designed for 104 measurements may result in a sampling design that is quite different to what is optimal.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The standardized and statistical approach to the analysis of the entire overall used in this study was a reliable and reproducible method. An entire overall could be mapped in 3 h using the Dirichlet PXRF method, instead of the one and a half days required by the ICP-AES method. Analysis of the data indicates that the PXRF method gives a more accurate estimate of whole-body contamination than the patch method. Furthermore, the 104 measurements give a much greater spatial resolution to the exposure data than analysis of the whole overall or patches by ICP-AES. This detailed knowledge of the pattern of deposition on the body is of potential importance in chemical risk assessments.
Acknowledgements—The authors would especially like to thank Paul Roberts of the Exposure Control Section and Rob Foster, Barry Smith and Monica Martinez of the Inorganic Exposure Assessment Section of HSL for the analysis of samples. The Pesticide Registration Section of the Health & Safety Executive (UK) sponsored this project.
|
FOOTNOTES |
|---|
* Author to whom correspondence should be addressed.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Fenske RA. (1990) Non-uniform dermal deposition patterns during occupational exposure to pesticides. Arch Environ Contam Toxicol; 19: 332–7.
Foster R, Guiver R, Wheeler JP. (1999) PXRF mapping of biocide contamination on overalls: feasibility study, IEAS/99/06, internal HSL report.
Garrod ANI, Rimmer DA, Robertshaw L, Jones T. (1998) Occupational exposure through spraying remedial pesticides. Ann Occup Hyg; 42: 159–65.
Garrod ANI, Martinex M, Pearson J, Proud A, Rimmer DA. (1999) Exposure to preservatives used in the industrial pre-treatment of timber. Ann Occup Hyg; 43: 543–55.
Garrod ANI, Guiver R, Rimmer DA. (2000) Potential exposure of amateurs (consumers) through painting wood preservative and antifouling preparations. Ann Occup Hyg; 44: 421–6.
HSE. (1999) Dermal exposure to non-agricultural pesticides—exposure assessment document EH74/3. Health & Safety Executive.
Llewellyn DM, Brazier A, Cocker J, Evans J, Hampton J, Nutley BP, White J. (1996) Occupational exposure to permethrin during its use as a public hygiene insecticide. Ann Occup Hyg; 40: 499–509.
Moller J. (1994) Lecture notes on random Voronoi tessellations. Lecture Notes in Statistics (87). Springer-Verlag.
Popendorf WJ, Ness SA. (1994) Pad dosimetry methods. In Ness SA, editor. Surface and dermal monitoring for toxic exposures, chapter 14. New York: Van Nostrand Reinhold.
Tannahill SN. (1997) Assessment of exposure to non-agricultural pesticides and effectiveness of protective clothing, Institute of Occupational Medicine phase 1 report, HSE research project R3228/R46.60.
Tannahill SN, Robertson A, Cherrie B, Donnan P, MacConnell ELA, MacLeod GJ. (1996) A comparison of two different methods for assessment of dermal exposure to non-agricultural pesticides in three sectors, Institute of Occupational Medicine report TM/96/07. Edinburgh: Institute of Occupational Medicine.
Stoyan D, Kendall WS, Mecke J. (1992) Stochastic geometry and its applications, Wiley Series in Probability and Mathematical Statistics. Wiley.
WHO. (1982) Field surveys of exposure to pesticides. Standard protocol: VBC/8282.1. Geneva: Division of Vector Biology and Control/World Health Organization.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. ROFF, D. A. BAGON, H. CHAMBERS, E. M. DILWORTH, and N. WARREN Dermal Exposure to Electroplating Fluids and Metalworking Fluids in the UK Ann. Hyg., April 1, 2004; 48(3): 209 - 217. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. W. HUGHSON and R. J. AITKEN Determination of Dermal Exposures During Mixing, Spraying and Wiping Activities Ann. Hyg., April 1, 2004; 48(3): 245 - 255. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. ROFF, D. A. BAGON, H. CHAMBERS, E. M. DILWORTH, and N. WARREN Dermal Exposure to Dry Powder Spray Paints Using PXRF and the Method of Dirichlet Tesselations Ann. Hyg., April 1, 2004; 48(3): 257 - 265. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||