Evaluation of a robotic transfer replicator: machine parameters that affect measurements of transfer of particulates from carpet surfaces to human skin versus human skin-like surfaces

Abstract

We have designed and built a robotic machine (robotic transfer replicator) to replicate the process by which various particulates are transferred from carpet surfaces to human skin. The particulates tested included bovine serum albumin, dust mites, paint dust, pesticides, and pollen. We have investigated the effect of time of transfer, applied pressure, skin-like receptor materials, particulate size, carpet composition and construction, and motion of the receptors on the percent transfer. The amount transferred was determined to be directly proportional to the amount of particulate applied to the carpet surface. The percent transfer values ranged from 0.05% to 28% for different combinations of particulates, carpet types, and method of transfer. The percent transfer was shown to be independent of receptor size or applied pressure. To explain the variations in the percent transfer, a three-zone model for particulate penetration into carpets is proposed. While the robotic transfer replicator was not capable of reproducing exactly the results from human finger transfer, it does provide a controlled and reproducible approach to the transfer process and the ability to use toxic or pathogenic materials that could not be used with human subjects.

Keywords

testing fabrication measurement materials performance properties structure-properties sorption chemistry

The transmission of particulates, microorganisms, pesticides, toxins, and allergens from various fabric surfaces to human skin through physical contact is a major safety concern. Precise and accurate measurements of such transfers are necessary to design and manufacture more consumer friendly fabrics. Many laboratories have performed experiments on the transfer of pesticides and particulates from a variety of donor surfaces to a variety of receptor materials, including human hands and fingers.^1–20 A few of these researchers have calculated the percent transfer (amount transferred/amount available for transfer).¹ Fewer researchers have documented transfer after multiple transfer events from the same surface. One paper has addressed the famous “five second rule”.²¹ A review of the current literature has failed to reveal any recent publications involving such transfer.

In principle, it may be possible to perform transfer measurements using a Crockmeter, which has been designed to assess the extent of color transfer from a donor fabric to a standard receptor fabric. However, the Crockmeter lacks the ability to alter the motion of rubbing, the area of contact, and the applied pressure to match the transfer process to human finger experiments.

Our goal was to design, construct, and test a robotic machine capable of replicating the transfer process of small particles from fabric surfaces to human skin, in this study, human skin was represented by the finger pads of both the middle and ring fingers. The objectives were to be able to replicate the transfer of particulates to human skin to automate the transfer process and to use toxins and pathogens, which would be too harmful for human subjects, as materials to be transferred. The following particulates were chosen to provide a cross-sectional sample of typical materials that would be expected to be present on contaminated carpet and tile surfaces. These particulates included dust mites, paint dust, pesticides, pollen, and protein. Bovine serum albumin (BSA) was chosen as the representative protein. The carpets and tiles were chosen to be typically those that are commonly found in commercial installations.

Experimental procedures

Instrumentation

A machine was designed to provide repetitive and repeatable motion between the donor surface and the receptor surface. This machine, shown in Figure 1(a), is referred to as the Robotic Transfer Replicator 1 (RTR1). To begin the transfer process, a 133 mm (5.25 inch) diameter sample of the carpet or tile donor was placed into a 140 mm (5.5 inch) diameter plastic Petri dish. Two multi-toothed armatures were activated to swing over and land on the edge of the donor surface to hold it securely in place. A known weight of particulates was placed over the donor surfaces of the tiles or carpets, using a sieve to distribute the particulates. The receptor material was held in position on a 20 mm diameter aluminum nose piece, using a rubber O-ring. To start the pre-programmed mechanical operation sequence, the RTR1 picked up the nose piece and placed the nose piece over the donor surface. The RTR1 then lowered the nose piece so that it descended onto the donor surface and pressed against the donor surface using a programmed hydraulic system to maintain a constant pressure. The RTR1 then executed a fixed number of motions of the nose piece over the donor surface creating an hourglass (figure eight) pattern. This is defined as one rub cycle. At the end of the rubbing cycle, the RTR1 raised the nose piece, delivered it to the mouth of a glass jar which contained 20 mL of a 0.10% sodium dodecyl sulfate (SDS) solution, and released it. As the final step, the RTR1 returned to the start position. The glass jar was capped and shaken vigorously. Aliquots were then removed for assay. Subsequently, a second Robotic Transfer Replicator (RTR2) was constructed with the additional capability of varying the applied static pressure and of varying the rub cycle pattern and the total number of motions within the rub cycle. This is shown in Figure 1(b). In addition, the RTR2 was physically smaller in design so that it could fit into an isolation hood. In all other aspects, the two machines operated in a similar manner.

Figure 1.

Robotic Transfer Replicator 1 (a) and Robotic Transfer Replicator 2 (b).

Materials

Particulates

The BSA powder from Sigma-Aldrich was ground in a mortar and pestle for 10 minutes. The ground powder was passed through a 100-mesh sieve (150 µm) and stored at 4℃ until use. Milled whole body Dermatophagoides pteronyssinus dust mites were obtained from Stallergenes Greer at greater than 99% purity. The dust mites were passed through a final sieve mesh of 350–357 µm prior to shipment. The dust mites were stored at 4℃ until use. Just prior to use, the dust mites were passed through a 20-mesh (850 µm) sieve to apply to the donor surface (carpet or tile). Oven-dried Giant Ragweed (Ambrosia trifida) pollen was obtained from Stallergenes Greer and stored at 4℃ until use. This source of pollen contained less than 0.1% foreign pollen, mold, rust, or smut spores. It was non-defatted. To provide a means to quantify the amount of paint dust, a fluorescent tracer, tetramethyl rhodamine, was added to flat latex paint (Evermore interior latex paint manufactured by Glidden) at a level of 0.4 g/100 mL of paint. The paint mixture was applied to the surfaces of ceramic tiles and permitted to dry. The dried paint was scraped from the tile surface, ground to a fine powder, and applied through a 40-mesh (400 µm) screen to the carpet or tile surface. To create different sizing of the particulates, BSA and paint dust were applied separately to the top of a stack of metal sieves of increasing mesh sizes. After shaking for 5 minutes for BSA and 25 minutes for the paint dust, the particulates were removed from each sieve and stored until use.

The amounts of BSA, dust mites, and pollen were determined using the standard BCA (bicinchoninic acid) assay, which is available from Pierce Chemical Company, and BSA as the standard protein. By comparing the assay of unknown amounts of dust mites and pollen to the assay response to the standard protein, it was possible to convert the protein assays into an absolute amount of both dust mites and pollen. The ratio of BSA to dust mites was determined to be 0.1538, and the ratio of BSA to pollen was determined to be 0.2001. The amounts of paint dust were determined by measuring the fluorescence intensity using a 554.5 nm excitation wavelength and a 575 nm emission wavelength and comparing with standard solutions. In all experiments, the particulate concentrations of the elution solution from the nose piece material or the finger washes were measured at three different dilutions, and the three resultant values were averaged.

The pesticide used was the “Flea Spray for Homes” as supplied by Enforcer Products (product code – FSH128). The active ingredients were listed as permethrin (0.25%), related reaction products (0.02%), and Nylar® (0.01%). We focused on measuring the levels of permethrin, since it was the major component of the pesticide. The amount of permethrin in each sample was determined using gas chromatography at the Agriculture Services – Pesticides and Hazardous Waste Laboratory at the University of Georgia. The gas chromatograph was a Tracor 222 with a packed GP 1.5% SP-2250/1.95% SP-2401 pm 100/120 Supelcoport (Supelco Cat #1-1947) column. The carrier gas was an argon–methane mixture, “P5”. The sample volume was 10 µL. The column temperature was 260℃, and the detection was by electron capture. The chromatogram peaks were compared with known amounts of the standard.

Donor materials

Carpets

The carpets were chosen to be representative of commercial grade carpets, which would be used in high traffic areas. These carpets were obtained from a local building supply house. Some of the characteristics are listed in Table 1. The numerical values are averages of five to ten measurements.

Table 1.

Carpet characteristics

Carpet	construction	Areal	Tuft	Overall	Backing	Stitches/cm
ID		Density	Height	Height	thickness	(lengthwise)
		(kg/m²)	(mm)	(mm)	(mm)	× crosswise)
“A”	Cut loop	2.08	5.0	7.4	2.3	3.50 × 4.21
“B”	Multilevel loop	2.19	4.4	6.8	2.4	3.55 × 3.64
“C”	Cut loop	1.47	5.6	7.8	2.1	2.86 × 2.10
“D”	Multilevel loop	1.82	4.8	7.2	2.5	3.50 × 3.50
“E”	Level loop	2.22	5.1	7.1	2.1	3.67 × 3.94
“F”	Level loop	2.28	4.8	7.2	2.0	3.81 × 3.56
“G”	Cut loop	2.14	5.5	7.7	2.3	3.98 × 3.85

After cutting the circular sections of the various carpets with a clicker press, the carpet samples were stored for at least 24 hours in a conditioning room which maintained the temperature at 21 ± 1℃ and the relative humidity at 65 ± 2%.

Tiles

The tiles were Armstrong commercial grade tiles manufactured from vinyl composite and purchased from Lowes Hardware Store. The surface was labeled “Imperial Texture”, and the pattern code was 51899 (Lot # C016A). The circular tile samples were cut using the same clicker press die.

Receptor materials

In this study, the term “receptor materials” is used to identify either the fabrics or human skin to which the particulates were transferred from the donor surface of either carpet or tile. In the case of the fabrics, these materials were stretched over the aluminum nose piece and were securely held in place by rubber O-rings. Four different receptor materials were used in previous research: bare fingers (human skin), nylon/spandex Comfortwear® compression fabric, which was a double warp knit, 100% cotton knit fabric, which was a single jersey, and 100% cotton woven fabric, which was a twill.¹⁵ Of the three fabrics, the compression fabric matched human fingers the best in terms of selectivity of the four different microorganisms that were studied.²² That is, the ratios of transfer of the different microorganisms, as measured by transfer to human fingers, were matched by the ratio of transfer values to the compression fabric. Because of the many variables inherent in transfer studies, the compression fabric was not an absolute replacement for human fingers. This fabric was chosen to continue the experiments with particulate transfer. Five other additional fabrics also were tested as receptor fabrics, and they are described in Table 2.

Table 2.

Characteristics of receptor fabrics

Fabric identifier	Weight (g/m²)	Thickness (mm)	Composition
Compression	276	0.58 ± 0.01	Nylon/spandex
Satin	91.5	0.16 ± 0.01	Polyester
Microfiber woven	121	0.27 ± 0.01	Polyester
Nonwoven	38.7	0.34 ± 0.04	Polypropylene
Headliner – cloth face	110	0.69 ± 0.03	Nylon
Headliner – foam backing	198	6	Polyurethane
Warp knit – napped on one side	122	0.52 ± 0.01	Nylon

The compression fabric was a warp knit of 77% nylon and 23% spandex (280 denier dull Lycra, type 127) that was obtained from the Marena Group in Lawrenceville, GA. This fabric was used for most of the experimental transfer measurements. The satin fabric was a polyester satin weave, produced from polyester fibers and yarns and obtained from JoAnn Fabrics. The microfiber woven fabric was a plain weave microfiber, obtained from JoAnn Fabrics. The nonwoven fabric was a spunbound-melt blown-spunbound (SMS) point bonded produced from olefin fibers, which was marketed as a surgical gown material. The headliner fabric was an automobile headliner material composed of a nylon warp knit fabric that was bonded to polyurethane open cell foam. This was also obtained from JoAnn Fabrics. This fabric was mounted on the nose piece so that the knit fabric was in contact with the donor surface. The warp knit was also obtained from JoAnn Fabrics and mounted so that the napped surface was in contact with the donor surface.

Methods

Particulate transfer from carpet to fingers

Determination of finger contact area

To determine the area of the fingers of each of five human volunteers, a piece of cardboard was placed on a balance and the balance was tared to zero. Each person dipped both of their middle and ring fingers into an ink solution and pushed those fingers against the cardboard surface until a reading of 1.13 kg was achieved. The resultant ink image was traced onto paper, cut out of the paper, and weighed. The areas were determined by converting the weight to surface area using the weight of a square of known dimensions from the same paper.

Transfer protocol

To initiate the transfer process with fingers, a 133 mm diameter carpet or tile donor surface was secured with a double-sided tape to a top loading balance. This balance was an electronic balance normally used to weigh refrigerant bottles and was capable of withstanding horizontal motions. The balance was set to zero. A known weight of particulates was placed over the surface of the tiles or carpets, using a sieve to distribute the particulate as evenly as possible. The exact pattern and distribution of particulates on the surface was determined to be not critical to the transfer measurements. The fingers were washed with a detergent solution, rinsed with distilled water, dried with paper towels, and finally rinsed with 95% ethanol. The top fingers pads of the middle and ring fingers of the right hand were placed onto the surface and pressed to obtain and maintain a pressure reading of 1.13 ± 0.23 kg during the entire period of transfer. For the slide technique, the middle and ring fingers of the right hand were moved in a continuous “zig-zag” pattern over the surface for a period of 20 seconds to achieve transfer (see Figure 2). This is defined as one rub cycle. For the second transfer technique, which is hereafter referred to as “compression”, the middle and ring fingers of the right hand were pressed to the carpet or tile surface and then released to achieve transfer. The highest and lowest pressure readings were recorded.

Figure 2.

“Zig-zag” slide pattern of finger pads on the carpet.

After the transfer was completed, the exposed fingers were placed one at a time over the opening of a 13 mm × 100 mm test tube, which contained 1.00 mL of a 0.10% sodium dodecyl sulfate (SDS) solution. The test tube was inverted 20 times within a period of 20 seconds. After elution, the fingers were cleaned for the next transfer. The elutions from the two fingers were combined to give a total volume of 2.00 mL and mixed vigorously. Aliquots were removed from the combined solution and subjected to the assay protocol, which was appropriate for the specific type of particulate, as indicated above. The particulate transfer studies were repeated for a minimum of five times for each type of particulate and surface combination.

Calculations

The amount of the particulate that transferred to the finger pads or to the receptor material when using the RTR1 or RTR2 was calculated based on the appropriate assay and dilution factors. This value was divided by the weight of particulate that was deposited onto the surface of the donor material. This ratio is defined as the percent transfer. The values were not corrected for the contact area of the receptor when using the RTR1 or RTR2. However, because of the small diameter of the sampling test tube, the percent transfer values were corrected to reflect the total amount of particulate on the finger pads.

Results and discussion

Transfer to human fingers

Effect of different fingers

To compare the effect of different fingers, five different volunteers were selected to perform the transfer of particulates from carpet to human fingers. The area of each person’s finger pads was determined as described above, and the values are listed in Table 3. Each person then followed the finger protocol, as listed in the Methods section, with the exception that the target weight on the balance was corrected for the differences in finger pad area so that the pressure was 17.2 kPa for all measurements. Specifically, the larger the finger surface area was, the greater was the weight that was applied to the donor surface. The nominal amount of particulate, which was BSA for these experiments, applied to the carpet surface was 50 mg. The experiments were run in triplicate. The carpet used was the “B” loop pile carpet. The percent transfer values for the initial rub cycles with correction for the finger area values are also listed in Table 3. The errors listed are the standard deviations. The data were ranked from the lowest to the highest transfer.

Table 3.

Effect of different human fingers on the percent transfer of bovine serum albumin

Person	Finger area (cm²)	Percent transfer
A	4.97 ± 0.08	0.66 ± 0.12
C	6.47 ± 0.11	0.77 ± 0.10
B	6.50 ± 0.06	0.83 ± 0.06
D	5.05 ± 0.09	1.00 ± 0.04
E	6.17 ± 0.13	1.40 ± 0.18
Average		0.93 ± 0.29
Ratio of the standard deviation to the mean		0.31

The area of the 13 mm × 100 mm test tubes used to extract the particulate samples from the fingers was 0.95 cm². This area is smaller than the areas of the fingers that were in contact with the donor surface. Correcting for the ratio of total finger area to the area sampled (0.95 cm²) generates the transfer values that are listed in Table 3. There is no correlation between the finger area and the percent transfer. Admittedly, the range of finger areas tested was rather small on a percentage basis (1:1.3), and the smoothness of the skin was not kept constant. While the individual transfer values had low standard deviations and, therefore, rather high precision, the variations between individual values were much higher than the standard deviation values. This last conclusion illustrates the difficulty in measuring absolute percent transfers to human skin with a high degree of accuracy.

Transfer to human skin-like receptors

Effect of different receptor materials

To test the effect of using different receptor materials, several different fabrics were tested. These fabrics were chosen to be a wide range of possible surfaces both in composition and structure. An approximate 50 mg sample of BSA was weighed and applied to the surface of a loop carpet. The indicated fabric was attached to the nose piece of the RTR1 with an O-ring. The experiment was performed using the standard protocol. The values listed in Table 4 are means and the standard deviations from triplicate measurements.

Table 4.

Effect of different receptor fabrics on percent transfer of bovine serum albumin

Receptor material	Percent transfer
Compression fabric	1.43 ± 1.54
Headliner	1.19 ± 0.09
Nonwoven	0.72 ± 0.26
Microfiber woven	1.30 ± 0.16
Satin	2.44 ± 1.51
Warp knit	1.60 ± 0.24

While the mean values of percent transfer range from 0.72% to 2.44% and that difference is significant, the range for the 95% confidence of the compression fabric, which is the standard receptor fabric used in these measurements, was from 0.54% to 2.32%. This range is inclusive of all of the other receptor fabrics that were tested. Studies of microbial transfer from a wider range of possible receptor materials, including animal skin, indicated that the compression fabric was the best material that replicated the transfer to finger pads.¹⁵ For that reason, the compression fabric was used as the primary receptor material in this study.

Effect of particulate loading amounts

Preliminary transfer measurements for a wide range of weights of applied particulates and types of particulates were inconsistent. The source of this inconsistency became clear when the effect of the particulate loading amounts were compared with the amount of particulate transferred. A typical dose response experiment is shown in Figure 3. At low amounts of applied particulate, the amount transferred was directly proportional to the amount applied, which would result in a constant percent transfer. However, at a critical value of applied amount of particulate, this relationship ceased, and the amount transferred became constant. As seen by comparing Figures 3 and 4, different particulates behave differently in terms of their respective dose response curve. While paint dust shows a linearity through 200 mg of applied amount, the BSA dose response curve has a definite break at approximately 50 mg of applied amount. So, at some point, the receptor material became saturated with the particulate. As a consequence, all measurements of percent transfer were subsequently determined at applied amounts of particulates that were below the saturation amount. Therefore, it was necessary to perform a dose response curve for each combination of carpet and particulate to determine that saturation point, as the value would vary for each combination, before performing other experiments with that combination.

Figure 3.

Dose response for paint dust transfer from loop pile carpet using Robotic Transfer Replicator 1.

Figure 4.

Dose response for bovine serum albumin (BSA) transfer from loop pile carpet using finger pads.

Effect of pressure

To test the effects of the receptor pressure, variable amounts of BSA from 20 to 60 mg were distributed over loop pile carpet donor surfaces. Different weights were added to the 25 mm nose piece to give a range of total applied pressures of 10–22 kPa for the receptor surface on the donor surface. The range of pressures was limited by the constraints of the RTR2 design. The receptor was moved in a spiral motion. There were nine measurements for each pressure tested. The results of these measurements are listed in Table 5 as the mean value ± the standard deviation.

Table 5.

Effect of receptor pressure on percent transfer of bovine serum albumin from loop carpet to compression fabric

Pressure (kPa)	Percent transfer
10.3	1.72 ± 0.57
14.9	1.49 ± 0.54
17.1	1.54 ± 0.84
21.6	1.54 ± 0.57

Based on these error values, there are no significant differences between any of the values at the 95% confidence limit. Consequently, there was no effect of the applied pressure on the percent transfer over the narrow range that could be tested. Because it has been reported that the typical pressure applied by baby hands as a baby crawls across a floor is 17.5 kPa, this value was used throughout all of the measurements.¹⁵ Therefore, control of applied pressure is not critical for transfer measurements.

Effect of the area of the receptor on percent transfer

Three different size nose pieces were machined and used to transfer BSA from loop pile carpet to the standard receptor material. The amount of BSA distributed over the carpet surfaces ranged from 15 to 50 mg, which is well within the range of constant percent transfer. The applied pressure was calculated based on the total weight of the nose piece and the head weight divided by the area of the nose piece. The nose piece was rubbed over the carpet for one spiral pattern. The measured percent transfers for various combinations of nose pieces and static weights are summarized in Table 6. There is no difference between the percent transfer values at the 95% confidence level. In other words, the total amount of transferred BSA was the same no matter what was the size of the receptor surface area. It is critical to note that the percent transfer values were not corrected for the differences in the nose piece area. If such a correction were made, the percent transfer values for the 20 mm nose piece would be decreased by half, and the 25 mm nose piece values by a third relative to the 14 mm values. From the previous section, it was concluded that pressure does not affect the percent transfer. Eliminating that variable, the conclusion is that there is no effect of the size (area) of the receptor on the total amount of particulate that is transferred.

Table 6.

The effect of the nose piece size and pressure on the percent transfer of bovine serum albumin

Nose piece		Applied pressure
Diameter (mm)	Area (cm²)	kPa	Percent transfer
14	154	32.3	1.60 ± 0.51
20	314	16.0	1.38 ± 0.39
25	491	10.3	1.41 ± 0.58

As the nose piece area increases three times from the smallest to the largest nose piece, the percent transfer shows no significant change. Since the amount of BSA added to the carpet surfaces is very much below the level where the receptor becomes saturated, the conclusion would be that the amount of BSA that is transferred to the receptor is only a function of the applied amount of BSA, the particulate. The size of the donor material was kept constant in these experiments. This conclusion presumes that the nature of the particulate, its size distribution, and the nature of the donor surface remain constant.

Effect of transfer time

To determine the minimum time of rubbing that was necessary to reach a constant value for the percent transfer, two experiments were performed. In the first experiment, approximately 50 mg of BSA was weighed and placed in the center of a level loop carpet surface. The standard rub cycle was initiated using a figure eight motion with the RTR2. However, at the end of the cycle, neither the recipient compression fabric nor the carpet sample were removed. The rub cycle was repeated for the stated number of times. This results in net increased length of time of transfer so that the x-axis in Figure 5 represents the total time of transfer. The results of these experiments using the RTR1 are shown in Figure 5. For comparison, a similar set of measurements were made using paint dust as the particulate and the RTR2 as the transfer machine.

Figure 5.

The effect of the number of rub cycles on the transfer of bovine serum albumin (BSA) (closed circles) using the Robotic Transfer Replicator 1 (RTR1) or paint dust (open circles) from level loop carpet to compression fabric using RTR1.

The results for the transfer of either BSA or paint dust show a similar trend with time of transfer. By the time that the first rub cycle is completed, the receptor material contains close to the maximum amount of particulates. By the time that two rub cycles are completed, the receptor contains the maximum amount that can be transferred. The apparent decrease in amount transferred with rub times greater than two rub cycles is attributed to a redistribution of the particulate over the surface of the carpet as well as some movement of the particulate into the carpet and/or off the carpet sample itself.

Effect of rubbing patterns and particulate distribution

To test the effect of the pattern of the rub cycle, the RTR2 was utilized as the pattern could be altered through its software program. Four different programs were chosen: a figure eight pattern described a number eight or hourglass pattern, which was repeated. This pattern covered about 60% of the donor surface. A spiral pattern described a circular motion, which started in the center of the donor surface and moved out with an increasing diameter of a circle. Once the receptor was moving at the edge of the donor, the diameter of the circle was decreased until the receptor was in the center of the donor. This motion was repeated and covered about 100% of the donor surface. The steps pattern started with the receptor at the edge of the donor. The receptor was stepped down the distance of the receptor diameter and along the edge. At that point, the receptor was moved to the opposite edge of the donor until it touched that edge. The receptor was then moved along the edge for another step and was then moved across to the opposite edge. This was repeated until the receptor was opposite its starting position. Then it crossed all of the lines to return to the start position. These series of motions were repeated and covered about 95% of the donor surface. As a final variation of motion, the receptor was dropped onto the center of the surface of the donor, and the receptor nose piece was continuously moved in a very tight circle to give a spot rotation. In all of the measurements, the time of contact between the receptor and the donor surface was kept constant at 20 seconds.

As another variable, the mode of application of the particulate, which was BSA for these measurements, was investigated. Two modes of application were used. In the “center” mode, the known amount of BSA was placed directly in the center of the donor surface with no attempts to spread it on the surface. In the “spread” mode, the BSA was distributed over the surface using a sieve to assist in the process. However, the distribution was definitely not even, and no attempt was made to further distribute the BSA on the donor surface. Table 7 summarizes the results. The values given in the table are the percent transfer values ± the standard deviation. The values enclosed in parentheses are the number of measurements.

Table 7.

Effect of receptor motion and mode of particulate application on percent transfer

	Motion of receptor
	Figure 8	Spiral	Steps	Spot rotation
Application of donor
Surface area covered	∼60%	∼100%	∼95%	<20%
Center spot application	3.1 ± 1.2 (60)	1.4 ± 0.5 (29)	2.2 ± 0.4 (10)	9.5 ± 4.4 (29)
Spread unevenly	1.6 ± 0.4 (10)	1.7 ± 0.3 (18)	1.2 ± 0.2 (10)	N/D

The percent transfer values for the experiments in which the BSA sample applications were spread over the surface of the carpet are essentially independent of the motion of the receptor head. When the BSA samples were applied at a single spot in the center of the carpet, the variation in transfer values is much greater. This difference in transfer is interpreted as a lack of uniform distribution of the BSA sample when it is applied to the center of the carpet. This distribution is critical, as demonstrated by the “no motion” measurements, where the receptor surface is placed in contact with the mound of BSA. This is effectively a higher surface concentration (saturation) than when the BSA was distributed over the carpet surface. When the receptor surface is placed into motion, the result is a lower surface concentration, an opportunity to redistribute particulates over the carpet surface, and hence a lower percent transfer. Based on these results, the particulate samples were subsequently distributed as evenly as possible over the carpet surface using a sieve.

Effect of particulate size

A series of different sizes of particulates were separated using the procedure outlined in the Experimental procedures section. The percent transfer was measured using a level loop carpet designated as “F”, the compression fabric as the receptor surface, and the RTR2 as the machine. The values shown in Figure 6 are an average of two determinations for each size and particulate. The percent transfer values are plotted on a logarithmic scale because of the wide range of values.

Figure 6.

The effect of average particle size on the transfer of bovine serum albumin (BSA) (closed circles) or paint dust (open circles) from level loop carpet “F” to compression fabric using the Robotic Transfer Replicator 2.

As expected, the absolute values differed between the BSA transfer and the paint dust transfer. The percent transfer decreased orders of magnitude from the smallest particle size to the largest. This result demonstrates how sensitive the absolute percent transfer is on particle size. However, the transfer efficiency of both particulates became fairly constant for particulate sizes greater than 200 µm.

Effect of donor surface

To determine the effect of the donor surface on the percent transfer, five different carpets were used. The characteristics of these carpets are given in the Methods section (see Table 1). To initiate the transfer measurements, approximately 100 mg of BSA was accurately weighed and distributed onto the carpet surface. The standard protocol for the RTR1 was employed for the transfer process and analysis. Table 8 summarizes the values for the first transfer rub cycle for each of the carpet surfaces. The value of “n” in parentheses is the number of experiments. The carpets are ranked from low to high percent transfer.

Table 8.

Comparison of different carpets as donor surfaces for transfer of bovine serum albumin

Carpet	% transfer (mean ± S.D.)	Pile type	Tuft height (mm)
B	0.81 ± 0.18(7)	Unlevel loop	4.4
E	1.08 ± 0.22(6)	Level loop	5.1
A	1.30 ± 0.13(5)	Cut loop	5.0
G	1.83 ± 0.31(4)	Cut loop	5.5
D	2.39 ± 0.46(5)	Unlevel loop	4.8
C	3.04 ± 1.23(10)	Cut loop	5.7

Carpet “D” actually had some bare spots where the backing was accessible to the receptor surface. Carpet “C” was a shag carpet with the lowest density and therefore not as comparable with the other carpets. Its high percent transfer is likely due to motion of the piles as they are rubbed by the receptor surface. As the piles are flattened and then lifted as the receptor changes direction; this motion would unbury and redistribute the particulates and effectively increase the surface available for transfer.

Several conclusions can be drawn from these data. The percent transfers ranged from 0.8% to 3%, or almost a fourfold change. Examining the listing of carpets from the lowest to the highest percent transfer, all pairs of carpets that are close in percent transfer are not significantly different at the 95% confidence level, with the exception of carpets A and G. All comparisons with non-adjacent carpets reveal significant differences. Also, there is no correlation between the pile type of the carpet and the percent transfer, as can been seen from Table 8. This is attributed to the variations in the carpets even within a sub-category of pile type. However, there is a small positive correlation between the tuft height and the percent transfer, as seen in Figure 7. The R-squared value is only 0.4 for a linear trend line.

Figure 7.

Effect of tuft height on percent transfer of bovine serum albumin from different carpets.

The effect of tuft height on percent transfer could be explained by the greater height of carpet tufts being more restrictive to the movement of particulates into the carpet, where the particulates would be less accessible to transfer to the receptor material. This would leave more particulates on the surface of the carpet to be in contact with the receptor material. The experiments with multiple rub cycles demonstrate that the particulates penetrate deeper into the carpet (moving further from the carpet surface) as the rubbing continues.

Effect of the number of rub cycles

All of the measurements reported so far involve the transfer of a particulate for the first time that the donor material rubs over the donor surface, which contains the particulate. A series of measurements were made that involved multiple rubs. Unlike the experiments reported in Figure 5, the receptor surface was replaced with a fresh surface after each rub. A typical result is shown in Figure 8. Other combinations of particulates and methods gave similar results, although the shapes of the curves were not identical (data not shown).

Figure 8.

Multiple rub measurement of the transfer of bovine serum albumin from loop pile carpet using the Robotic Transfer Replicator 1.

The amount transferred decreased dramatically after the first rub cycle and then decreased slowly with subsequent rub cycles. This occurred with all of the particulates. The final values varied based on the particulate and method used, with some combinations giving final percent transfer values indistinguishable from zero. It is not possible to fit the data points with any common mathematical function, including variations of exponential functions.

Effects of different particulates and donor surfaces

Table 9 summarizes the percent transfer for the five different particulates that were tested in this study. Using these data, several questions can be addressed.

Table 9.

Summary of the percent transfer values for different combinations of particulate, receptor surface, method, and donor surface. C. fabric refers to the compression fabric

			Percent transfer
Particulate	Receptor/method	Donor surface	Mean ± SD	SD/Mean
BSA	Finger/manual	Loop carpet	13.0 ± 2.0	14%
		Cut carpet	0.91 ± 0.13	14%
		Tile	22.8 ± 2.0	8%
	C. fabric/RTR1	Loop carpet	1.4 ± 0.8	54%
		Cut carpet	3.2 ± 0.4	13%
		Tile	50 ± 12	23%
	C. fabric/RTR2	Loop carpet	0.8 ± 0.2	25%
		Cut carpet	1.3 ± 0.1	10%
Dust mites	Finger/manual	Loop carpet	2.8 ± 1.0	27%
		Cut carpet	3.1 ± 0.3	8%
		Tile	98 ± 20	20%
	C. fabric/RTR1	Loop carpet	12 ± 3	22%
		Cut carpet	28 ± 2	6%
		Tile	75 ± 24	32%
Paint dust	Finger/manual	Loop carpet	0.59 ± 0.06	11%
		Cut carpet	0.33 ± 0.06	12%
		Tile	123 ± 32	24%
	C. fabric/RTR1	Loop carpet	0.21 ± 0.02	10%
		Cut carpet	0.89 ± 0.4	48%
		Tile	31 ± 22	71%
Pollen	Finger/manual	Loop carpet	0.98 ± 0.52	50%
		Cut carpet	0.78 ± 0.13	14%
		Tile	99 ± 1.3	1%
	C. fabric/RTR1	Loop carpet	5.2 ± 0.4	7%
		Cut carpet	16 ± 4	22%
		Tile	133 ± 5	4%
Pesticide	Finger/manual	Loop carpet	3.0 ± 0.1	3%
		Cut carpet	3.4 ± 0.5	14%
	C. fabric/RTR1	Loop carpet	2.4 ± 0.2	8%
		Cut carpet	1.2 ± 0.4	33%

BSA: bovine serum albumin; RTR1: Robotic Transfer Replicator 1; RTR2: Robotic Transfer Replicator 2.

Note: Please check table Alignment in pdf and process.

To determine whether the finger method or the RTR method is more precise, the standard deviations of each value were divided by the mean value to give the relative error (%CV). The results of these calculations are given in the last column of Table 9. The average relative error (%CV) for the fingers was 16% ± 12%, for the RTR1 was 25% ± 20%, and for the RTR2 was 17% ± 10%. Therefore, the precision of the finger transfer method was about the same as for the RTR2 machine and somewhat but not significantly better than for the RTR1 machine measurements. However, the differences are not significant. The relative error varied over a wide range for all three methods, as can been seen by the standard deviations of the average relative error. This variation reflects the inability to control all of the variables involved with particulate transfer and, therefore, the inherent error in these measurements.

As shown in Table 3, the percent transfer can vary by as much as a factor of two with different human experimenters performing the same transfer. As shown in Table 9 and in Figure 9, the values of percent transfer as measured by human finger pad transfer are close but do not exactly match the values as measured with either the RTR1 or RTR2. This is true for most of the comparisons when the standard deviation values are taken into account. It is also noted that trends in the percent transfer values are not correlated. When comparing the loop pile with the cut pile carpets for the same particulate, the human finger pad transfer shows a change in one direction, whereas the RTR1 shows a change in the opposite direction. The discrepancy is most evident when comparing the values for transfers from tile surfaces. However, this is to be expected due to the transfer process. Unlike the carpet surfaces, the particulates remain on the tile surface with no penetration. The receptor simply pushes the particulates around the surface like a snow plow pushing the particulates out of the path of the receptor. This gives rise to very high percent transfer values that approach 100%. The values would be very sensitive to the precise motion as a consequence. This is reflected in the range of relative errors for the tile measurements of 1–71%.

Figure 9.

Comparison of the percent transfer values for different particulates, carpet surfaces, and transfer methods with the error bars representing the standard deviation values.

As shown in Table 9, the percent transfer values for the five different particulates range over a ratio of 1:20. The precise order of values depends on the carpet surface. For the loop pile carpet the ranking was paint dust < pollen < pesticide ∼ dust mites < BSA, based on the finger pad transfer method. For cut pile carpet, the ranking was paint dust < pollen ∼ BSA < dust mites ∼pesticide, based on the same method. The rankings based on the RTR1 method are paint dust < BSA< pesticide < pollen < dust mites for loop carpets and paint dust < pesticide < BSA < pollen < dust mites for cut pile carpets.

Unlike the data in Table 8, where six different carpets were tested with the same particulate, Table 9 presents data to compare two different carpets with five different particulates. The variation of the percent transfer values as measured by the finger pad method is not significantly different, with the exception of BSA as the particulate. The RTR1 method gave larger differences of the order of two to fourfold.

To explain the above conclusions, we propose a model for the transfer process. In this model, we divide the carpet into three zones: The top zone is the surface of the carpet, which is in contact with the receptor material. The bottom zone lies within the base of the tufts of the carpet. Any particulates found in that bottom zone are considered “trapped” and inaccessible to the receptor material. The inner zone lies between the other two zones. Furthermore, we assume that for low dosages of particulates, the particulates are in equilibrium between the top zone and the receptor and the particulates are in equilibrium between the top and inner zones. This latter assumption is supported by the linear dose response curves at low dosage amounts.

At the start of each rubbing experiment, the particulates are applied to the carpet surface. The receptor material is pressed against the carpet surface, and rubbing motion commences. Initially, the particulates are more or less evenly distributed over the carpet surface, and some of the particulates start to move into the inner zone. The receptor surface collects an unknown percentage of the particulate, which lies on the surface in the top zone. The amount collected is independent of the size of the receptor as the receptor moves over the entire surface of the donor surface. This transfer occurs very quickly in the rubbing cycle. If the rubbing cycle is extended, more of the particulates leave the top zone and enter the inner zone. As the rubbing cycles are repeated, a large number of particulates move from the top zone to the inner zone, as shown in Figure 8. The amount of particulate that leaves the top zone and enters the inner zone during the initial rub cycle is expected to be dependent on the size and the shape of the particles. For some combinations of particulates and donor surfaces, there is no particulate in the top zone and, consequently, no transfer occurs.

Since the percent transfer is dependent on the rate of movement of the different particulates, the reproducibility would be predicted to be greater than if it were a simple equilibrium between the particulates in the carpet and the receptor material. Because of the subtle differences between human rubbing finger pads and robotic machine rubbing, it is inappropriate to expect the two different methods to give identical results. This is especially true when different human operators are involved with the same particulate transfer from the same donor surface, giving rise to different results. In addition, it is extremely difficult to maintain a steady pressure while performing the zigzag motion with fingers.

Because tile surfaces do not have these zones available, the percent transfer from tile surfaces is much greater than that from carpet surfaces. The transfer values are very dependent upon the precise motion so that correlations between human fingers and robotic machines cannot be achieved.

Conclusions

As stated in the introduction, our goal was to develop and test a robotic machine that was capable of replicating the transfer process of various particulates from fabric surfaces to human skin. In this study, human skin was represented by the finger pads of the middle and ring fingers. Based on the data in this study, the conclusions are summarized as follows.

At low dosage amounts of the particulates, the amount transferred is directly proportional to the amount deposited. This means that the percent transfer is constant at low dosage levels. Hence, all of the data are reported as percent transfer rather than total transfer. At a particular amount of dosage and above, the amount transferred becomes constant. The implication is that the receptor material has become saturated and can no longer accept any more of the particulate. The saturation value is dependent upon the nature of the particulate.

The transfer process is rapid and essentially complete by the end of the first rub cycle. Continuation of the rubbing process does not increase the amount transferred. In fact, the opposite occurs with extended rubbing.

The values of percent transfer have a reproducibility of 15–25%. The robotic transfer process has the same precision, as does the finger pad transfer method. Examination of the literature shows similar difficulties in reproducibility.

The variation between different human subject transfer measurements was about twofold. This is similar to the comparison between the recent transfer values for the RTR1 and RTR2. The conclusion is that the absolute values of the percent transfer are very sensitive to small differences in technique. The values reported here should be considered to be approximate values.

The percent transfer values determined by the finger pad transfer method are similar to but do not match the values obtained using the robotic transfer machines. The closeness of matching values is determined by the nature of the particulate.

The percent transfer is highly dependent upon and very sensitive to the nature of the particulate and the size of the particulate. There is also a large difference between carpet and tile surfaces. The pattern of depositing the particulate on the donor surface, the motion of the nose piece, and the nature of the carpet donor surface had a smaller but significant effect on the percent transfer. For the different carpets, there was a small effect of tuft height on the percent transfer.

Surprisingly, the percent transfer is independent of the receptor area and of the applied pressure. This is true for both the transfer to finger pads and transfer to compression fabric with the robotic transfer replicator.

When the donor surface is rubbed again with a fresh receptor material, the value of the second percent transfer is dramatically less than the initial transfer value. The percent transfer continues to decrease with additional rubbing cycles, reaching zero transfer for some but not all combinations of donor surfaces and particulates. The decrease with the increasing number of rub cycles is a complex mathematical function.

Despite the lack of precise agreement between the transfer values determined with finger pad transfer and those determined with the RTR1 and RTR2, the robotic transfer replicator does provide a reproducible process of determining transfer values. The advantages of using this machine to measure transfers are that reproducible relative measurements are easily made, that the values are similar to that obtained by a human subject, and that the transfer of toxic or pathogenic materials can be made, which would not be possible with human subjects. The machine can be used to explore the effect of different carpet compositions and construction as well as other donor surfaces, such as upholstery or drapery fabrics. This information can lead to the better design of fabrics that are in contact with the human population to minimize transfers of undesirable particulates and microorganisms.

Footnotes

Acknowledgement

The authors would like to recognize Roger Fortner from the University of Georgia Instrument Shop, who was responsible for the construction and programming of both models of the Robotic Transfer Replicator.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publications of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by a grant from the Carpet and Rug Institute and the Agricultural Experimental Research project number 1009239.

References

Van Dyke

Martyny

Serrano

. Methamphetamine residue dermal transfer efficiencies from household surfaces. J Occup Environ Hyg 2014; 11: 249–258.

Ferguson

Bursac

Biddle

et al.

Soil-skin adherence from carpet: use of a mechanical chamber to control contact parameters. J Environ Sci Health A 2008; 43: 1451–1458.

Ferguson

Bursac

Johnson

et al.

Computer controlled chamber measurements for clay adherence relevant for potential dioxin exposure through skin. J Environ Sci Health A 2012; 47: 382–388.

Ferguson

Bursac

Wayne

. Combined analysis of results on single and multiple contacts: potting soil, sand and clay transfers to cloth for the estimate of soil adherence to human skin. Environ Nat Resources Res 2013; 3: 33–44.

Ferguson

Bursac

Coleman

et al.

Comparisons of computer-controlled chamber measurements for soil-skin adherence from aluminum and carpet surfaces. Environ Res 2009; 109: 207–214.

Hubal

Nishioka

Ivancic

et al.

Comparing surface residue transfer efficiencies to hands using polar and nonpolar fluorescent tracers. Environ Sci Technol 2008; 42: 934–939.

Williams

Bernard

Dyk

et al.

Measurement of transferable chemical residue from nylon carpet using the California roller and a new mega-California roller. J Environ Sci Health B 2008; 43: 675–679.

Bernard

Melnyk

Berry

. Measurements of transferable residue from ceramic tile, vinyl tile, hardwood flooring, and carpets using a press sampler and C-18, PUF, and cotton sampling disks. Epidemiology 2002; 13: S194–S194.

Fenske

. Dermal transfer of chlorpyrifos residues from residential surfaces: Comparison of hand press, hand drag, wipe, and polyurethane foam roller measurements after broadcast and aerosol pesticide applications. Environ Health Persp 1999; 107: 463–467.

10.

Camann DE and Clothier JM. Dermal transfer efficiency of pesticides from new and used cut-pile carpet to dry and wetted palms. Report for U.S. Environmental Protection Agency, Research Triangle Park, NC, 1998.

11.

Rodes

Newsome

Vanderpool

et al.

Experimental methodologies and preliminary transfer factor data for estimation of dermal exposures to particles. J Exp Anal Environ Epid 2001; 11: 123–139.

12.

Ivancic

Nishioka

Barnes

et al.

Development and evaluation of a quantitative video-fluorescence imaging system and fluorescent tracer for measuring transfer of pesticide residues from surfaces to hands with repeated contacts. Ann Occup Hyg 2004; 48: 519–532.

13.

Hubal

EAC

Suggs

Nishioka

et al.

Characterizing residue transfer efficiencies using a fluorescent imaging technique. J Exp Anal Environ Epid 2005; 15: 261–270.

14.

Clothier JM. Dermal transfer efficiency of pesticides from new, vinyl sheet flooring to dry and wetted palms. Report for U.S. Environmental Protection Agency, Research Triangle Park, NC, Publication No. EPA/600/R-00/029, 2000.

15.

Annis PA, Leonas KK and Knopp JA. Microbial and particulate transfer from carpet to skin-like materials. Report, 2007.

16.

Hubal

Suggs

Nishioka

et al.

Characterizing residue transfer efficiencies using a fluorescent imaging technique. J Exp Anal Environ Epid 2005; 15: 261–270.

17.

Causer

Shorter

Sercombe

. Effect of floor covering construction on content and vertical distribution of house dust mite allergen, Der p I. J Occup Environ Hyg 2006; 3: 161–168.

18.

Sercombe

Liu-Brennan

Causer

et al.

The vertical distribution of house dust mite allergen in carpet and the effect of dry vacuum cleaning. Int J Hyg Environ Heal 2007; 210: 43–50.

19.

Suganuma

Kuno

. The effect of hydrated ferric oxide particle size on the rubbing fastness. Sen'i Gakkai shi 1987; 43: 230–235.

20.

Ferguson

Biddle

Coleman

et al.

In-vitro soil adherence for dermal exposure using a controlled mechanical chamber. J Appl Sci Res 2009; 5: 232–243.

21.

Dawson

Han

Cox

et al.

Residence time and food contact time effects on transfer of Salmonella Typhimurium from tile, wood, and carpet: test the five-second rule. J Appl Bact 2007; 102: 945–953.

22.

Yu H. The effect of chemical finishing on the microbial transfer from carpets to human skin and selected fabrics. PhD Thesis, The University of Georgia, USA, 2007.