tsk 1032 due thysday 6pm PST

Risk Analysis, Vol. 34, No. 7, 2014 DOI: 10.1111/risa.12166

Dermal Versus Total Uptake of Benzene from Mineral Spirits Solvent During Parts Washing

Kenneth T. Bogen∗ and Patrick J. Sheehan

Quantitative approaches to assessing exposure to, and associated risk from, benzene in min- eral spirits solvent (MSS), used widely in parts washing and degreasing operations, have fo- cused primarily on the respiratory pathway. The dermal contribution to total benzene uptake from such operations remains uncertain because measuring in vivo experimental dermal up- take of this volatile human carcinogen is difficult. Unprotected dermal uptake involves simul- taneous sustained immersion events and transient splash/wipe events, each yielding residues subject to evaporation as well as dermal uptake. A two-process dermal exposure framework to assess dermal uptake to normal and damaged skin was applied to estimate potential daily dermal benzene dose (Dskin) to workers who used historical or current formulations of recy- cled MSS in manual parts washers. Measures of evaporation and absorption of MSS dermally applied to human subjects were modeled to estimate in vivo dermal uptake of benzene in MSS. Uncertainty and interindividual variability in Dskin was characterized by Monte Carlo simulation, conditioned on uncertainty and/or variability estimated for each model input. Dermal exposures are estimated to average 33% of total (inhalation + dermal) benzene parts washing dose, with approximately equal predicted portions of dermal dose due to splash/wipe and to continuous contact with MSS. The estimated median (95th percentile) dermal and to- tal daily benzene doses from parts washing are: 0.0069 (0.024) and 0.025 (0.18) mg/day using current, and 0.027 (0.085) and 0.098 (0.69) mg/day using historical, MSS solvents, respectively.

KEY WORDS: Chemical; exposure; occupational; percutaneous; permeability; solvents; skin

1. INTRODUCTION

Benzene is classified as a known human carcinogen,(1–3) and workplace exposures to benzene have been a concern in the United States for more than 60 years. Although the primary focus of ben- zene exposure assessments has historically been on respiratory exposure among workers in industries that produced or used substantial quantities of benzene, more recently, questions have been raised regarding multiroute worker exposures to products with lower benzene levels where there is potential

Exponent, Inc., Oakland, CA 94596, USA *Address correspondence to Kenneth T. Bogen, kbogen@ exponent.com.

skin contact, as well as inhalation of volatile emis- sions during use. Mineral spirits solvent (MSS), also called white spirits solvents, are petroleum distillates composed primarily of aliphatic and alicyclic hydro- carbons with relatively small fraction of aromatic hydrocarbons (approximately 1–8 wt%), including trace levels of benzene (<0.1 to <0.002 wt%).(4)

MSS is used industrially to clean and degrease metal machine tools and parts; in painting, printing, and textile work as an alternative cleaning agent to turpentine; and together with cutting oil as a thread-cutting and reaming lubricant. Widespread MSS use by mechanics and other workers who engage in cleaning and degreasing applications, such as parts washing with a mechanical washer, presents an opportunity to evaluate both respiratory and

1336 0272-4332/14/0100-1336$22.00/1 C© 2014 Society for Risk Analysis

Dermal vs. Total Uptake of Benzene 1337

dermal exposures to benzene, as well as the relative contribution of these exposure routes to a worker’s total benzene exposure from cleaning applications.

Estimating dermal exposures to benzene from MSS among workers using parts washers is compli- cated by the need to consider the two different up- take processes that occur with activities that involve both sustained immersion and splash/wipe events, and in some cases by potential uptake enhancement due to skin damage. Advancement of occupational dermal exposure assessment and of corresponding protective workplace criteria for this pathway has long been hampered by critical deficits of relevant data and modeling frameworks.(5) Integration of accurate dermal permeability data with straight- forward, biologically-based mathematical models currently provides a basis for reasonably accurate predictions of dermal uptake for certain chemicals and simple immersion exposure scenarios.(6) How- ever, the feasibility of this approach alone is limited in realistic occupational contexts involving complex patterns of exposure to chemical mixtures as those involved in parts washing where there may be both continuous contact and occasional splash or wipe exposure to the solvent. Although separate dermal uptake studies have evaluated continuous immer- sion and thin-layer uptake of benzene, they have generally not evaluated simultaneous uptake from both processes, and/or relied on dubious estimates of in vitro benzene permeability from dermally contacted MSS. A recent study by Petty et al.(7) pro- posed a two-process dermal framework; however, the absence in vivo solvent-specific uptake and evap- oration data substantially limit application of their approach, particularly in addressing uptake of mis- cible benzene from the thin-layer solvent exposure. In addition, previous studies addressing benzene exposure from cleaning and degreasing operations have focused on intake via inhalation and offer no detailed consideration of the potential significance of the dermal pathway to total worker exposure.(8–10)

Based on mechanistic dermatotoxicological con- siderations and a detailed review of related dermal uptake (including human in vivo) data, an alterna- tive two-process dermal exposure framework was used to address this problem, and then applied to estimate potential daily and lifetime cumulative der- mal benzene exposures of workers who used histor- ical or current formulations of a recycled mineral spirit solvent in manual parts washers. Experimen- tal measures of MSS evaporation and uptake from

skin of human volunteers were obtained and mod- eled to improve the basis for estimated dermal up- take of benzene from dermally contacted MSS. Ex- perimental data were also obtained to estimate rates of splash-related MSS deposition on upper hand and lower forearm regions of skin typically experienced by parts washers. Uncertainty and interindividual variability in estimated exposures was explicitly char- acterized by Monte Carlo analysis conditioned on un- certainty and/or variability estimated for each model input.(11–13) Results were compared to estimates of potential dose from corresponding inhalation expo- sures for workers using MSS reported previously by Sheehan et al.(10) to evaluate the significance of the dermal pathway.

2. DATA AND METHODS

2.1. Benzene Concentration in Bulk MSS (Cbulk)

The benzene contents of bulk samples of Safety- Kleen 105 solvents (Safety-Kleen, Elgin, IL, USA) from “historical” (1992–1993) and “current” (1995– 1999) formulations were characterized, as described previously.(10) The current solvent is a low-aromatic- content hydrogenated MSS, and the historical sol- vent is the higher-aromatic-content version of the same MSS produced without hydrogenated distilla- tion. The arithmetic mean value, ±1 standard de- viation (SD), of benzene concentration by weight (Cbulk) for historical solvent was approximately 35 ± 15 ppmw, with corresponding 2.5th, 50th, and 97.5th percentiles of 14, 35, and 69 ppmw, respectively; the arithmetic mean (±1 SD) value of Cbulk for the cur- rent solvent was approximately 9.4 ± 5.0 ppmw, with corresponding 2.5th, 50th, and 97.5th percentiles of 3.5, 8.5, and 21 ppmw, respectively.(10)

2.2. Experimental Rates of In Vivo Dermal MSS Loss and Washing-Splash Droplet Deposition

2.2.1. Materials

As there are no published in vivo studies of simultaneous dermal uptake and evaporation loss of benzene in MSS from a thin splash or wipe layer on skin, an experimental study was undertaken. All measures of MSS loss from skin used Klean–Strip Odorless Mineral Spirits brand of MSS (100% distillates (petroleum) hydrotreated light, CAS No. 64742–47–8 (W.M. Barr, Memphis, TN, USA);

1338 Bogen and Sheehan

produced from a distillate with the same CAS number as the current formulation of Safety-Kleen 105. The MSS was applied using half a preweighed, double-tipped Q-tips R© Precision TipsTM (PTQ) applicator (Unilever, Englewood Cliffs, NJ, USA). Sections of 0.7-mm-thick adhesive-backed aluminum Scotch R© Foil Tape 3311 (3M, St. Paul, MN, USA), with one or more 1-cm2 square exposure-window (EW) areas each with >0.5-cm borders, were cut using a sharp blade and a 0.01-mm-resolution digital caliper (Ultratech Tool System No. 1433, General Tools & Instruments, Montreal, Canada). Each dermally exposed area on the tip of a central finger (F), on the back of a hand (BH), or on the ventral mid to lower forearm (FA) studied as described below was defined and bounded by a 1-cm2 square cut from 0.7-mm-thick adhesive-backed aluminum Scotch Foil Tape 3311 (3M). To measure MSS evap- oration rate per se, experiments each used a control apparatus (CA) consisting of an Al-tape section with one EW very firmly affixed either (1) to a smooth circular Al surface of a small inverted Al weighing cup, or (2) to a larger underlying stippled surface of the same Al foil tape after: (a) the adhesive side of the underlying tape was first firmly affixed to a glass microscope slide, and (b) its upper surface was finely and densely stippled (but not perforated) using a file brush in order to simulate irregular skin-surface morphology. All weighings were performed using a certified calibrated digital analytical balance (Model AE163, Mettler-Toledo, Inc., Columbus, OH, USA).

Splash kinetics during simulated parts washing were measured using a kitchen sink as a surrogate for a parts washer sink (∼60 cm × 40 cm × 22 cm deep), a hand-held synthetic-bristle scrub brush, and a 10-inch iron skillet held under and/or just near a continuous stream of household water flowing at ∼75 mL/sec. Each splash rate (ksplash, expressing a fraction of total area covered per hour) for a left or right combined lower arm and upper hand area was measured using “splash dosimeter” (SPD), consisting of a precut rectangular section of 0.12-mm thickness graph paper 17.8 cm wide × 23.0 cm long (total area Atot = 40,940 mm2). On each SPD, each splash-residue pattern (SRP) was traced with a pen less than five minutes postdeposition, which pattern was later measured using a digital caliper (see above) as described below. Preliminary experiments showed that all SRPs deposited on each SPD over a three-minute simulated washing period were clearly retained and could be traced by pen during the following two minutes.

2.2.2. Experimental Procedures

An IRB-approved human-subjects protocol was applied using materials described above as follows to measure in vivo dermal uptake of MSS from 1-cm2 areas of F, P, BH, and FA skin of four volun- teers, accounting for MSS evaporation as measured in control experiments using 1-cm2 area of stippled aluminum (Al) tape (to mimic dermal surface irregu- larity) instead of skin. Preliminary experiments using MSS applied to marked square 1-cm2 FA or BH der- mal areas showed that MSS migrated beyond the in- tended application boundaries at those sites. In con- trast, very little MSS migration was observed from marked 1-cm2 application areas on F or P sites, so MSS was applied directly to these sites without a sur- rounding EW, whereas an Al-tape strip containing three EWs was affixed firmly to each FA and BH area tested. In all EW experiments, each EW contained an applied, thin 1-cm2 layer of MSS, loss of which was measured either over time by repeat weighing (con- trol experiments), or (in all dermal experiments) by measuring after MSS application at time t = 0 the fi- nal elapsed time (t = T0) at which the optical sheen clearly visible for MSS on skin was no longer visi- ble, using a digital stopwatch evaluated to ±5 sec- onds. Applied initial MSS load (L0, in mg/cm2) was in each case measured to ±0.1 mg accuracy (by sub- tracting pre- from postloaded PTQ weights, using a mini-clip to elevate the PTQ tip during weighing, or by periodically reweighing each CA) using a certified calibrated digital analytical balance (Model AE163, Mettler-Toledo, Inc.). Between each repeated weigh- ing, each CA was returned to the surface of a glass- plate-covered hot plate maintained at an approxi- mate average target temperature of ∼32 ◦C typical for human skin,(14) measured using a conventional thermometer in a beaker filled with 100 mL of wa- ter centered on the warmed surface (experimental range: 30–36 ◦C).

In each simulated parts washing experiment, a subject continuously brushed varying parts of an iron skillet manipulated in a (conservatively) dynamic manner in a flowing kitchen sink water stream for a period of three minutes (Tsim = 0.05 h), brush- ing either at vigorous rate of ∼4.0 cycles/sec (Sub- ject 1) or a moderately vigorous rate of ∼2.4 cy- cles/sec (Subject 2), with each cycle consisting of a down/away and a return stroke. Splash onto the com- bined upper hand and lower forearm (FA) area of each arm of each subject was measured on a separate SPD wrapped cylindrically around that entire area,

Dermal vs. Total Uptake of Benzene 1339

meeting along the FA and wrist opposite the thumb, held in place using two rubber bands (one around the wrist, the other at mid-FA). Virtually no SPD over- lap or lack of skin coverage occurred, except for a triangular ∼4.5-cm2 area diverging from the crease at each mid-FA, at its widest along the FA edge of the SPD. Two subjects each performed six such experi- ments, immediately after each one of which the adja- cent countertop and the subject’s lower hands were carefully dried, the left and right SPDs were identi- fied with the subject ID, and each SRP on each SPD was rapidly but carefully traced within two minutes after the end of that splash period.

2.2.3. Modeling/Analysis of Experimental Data

MSS-evaporation data from control experiments were fit to the biexponential loss model described in Appendix A (see Section 3.1). Data on dermal MSS loss by simultaneous evaporation and dermal uptake (consisting of L0 and T0 measures for each subject, dermal site, and replicate) were used to calculate a corresponding value of dermal flux (J) by numerical optimization with respect to the nonlinear differential equations corresponding to biexponential evaporation, conditional on L0, T0, and assumed simultaneous dermal flux Ji(t) = J MSSi(t)/�MSSi(t) from each MSSi component (i = 1,2) modeled as being subject to first-order evapora- tion, as described in Appendix 1A. After confirming approximate (p > 0.05) normality by Shapiro-Wilk tests(15) p-values from which were adjusted for mul- tiple comparisons,(16) and approximate (p > 0.05) variance homogeneity by Bartlett’s tests(17) for data grouped by subject or by dermal region, two-way analysis of variance (ANOVA) was performed(18)

on the resulting estimates of ln(J) to assess the sig- nificance of these two categorical variables, where ln denotes natural logarithm. The geometric standard deviation (GSD) of interindividual J-variability was assumed to be 1.5, calculated from human in vivo data on dermal permeability for each of six dilute contaminants of JP8 jet fuel estimated for each of 10 volunteers dermally exposed to JP8.(19) (The assumed GSD is nearly identical to that estimated from MSS-uptake data on four participants in this study—see Section 3.) Because only four participants were used to study MSS loss kinetics, uncertainty in the mean value of (ln(GMJ)) of ln(J) was modeled, in terms of the corresponding sample mean ln(GMJ,est), as ln(GMJ) = ln(GMJ,est) + T3 ln(GSD)/

√ n, where

T3 is a Student t-distributed random variable with three degrees of freedom (df), GSD = 1.5, and

n = 4. Corresponding values of dermally absorbed benzene mass were then calculated as described below (Section 2.3.2.4, Appendix A).

From each simulated parts washing splash- deposition experiment, each SRP on each SPD af- ter experiment s (s = 1, . . . 6) was classified as be- ing approximately either elliptical (E), rectangular (R), or triangular (T), the corresponding length or height or major axis (L) and perpendicular width or minor axis (W) were each measured by digital caliper. Corresponding SRP-specific areas (Asj, j = E, R, or T) were each calculated accordingly. Each complex SRP pattern was divided into components of type E, R, or T to estimate Asj. Total area (As) for each sth left or right SPD for each subject was calculated by summing all SRP-specific areas over all types j, and a corresponding rate of fractional SPD coverage was calculated as ksplash = As/(Atot Tsim). Normality of ksplash estimates grouped by sub- ject and/or by side (left vs. right) was assessed by Shapiro-Wilk tests; homogeneity of approximately normal data sets was assessed by standard or Welch’s t-test(20) as appropriate, otherwise or additionally by Wilcoxon(21) or Kolmogorov two-sample test(22)

as indicated.

2.3. Model of Dermal Exposure to Benzene During Parts Washing

Although no human or animal studies have di- rectly measured dermal uptake of benzene arising from occupational contact with MSS, dermal absorp- tion of benzene is expected as a contaminant of MSS or of other petroleum-based solvents that contact skin in occupational settings and benzene absorp- tion from MSS has been measured in vitro with hu- man skin samples.(23) Dermal contact with MSS is ex- pected to result in percutaneous uptake of benzene or other MSS constituents that are absorbed into the outermost dead layer of skin, the stratum corneum (SC), during dermal contact. Uptake of benzene from MSS during parts washing with unprotected hands/forearms may occur both during sustained im- mersion events and during splash and/or wipe events that are subject to rapid evaporation, and in some cases uptake by these two processes may be en- hanced by any damage present on the contacted skin.

2.3.1. Two-Process Framework for Estimating Percutaneous Uptake

Once absorbed into SC, fat-soluble (lipophilic) compounds such as benzene and other solvent

1340 Bogen and Sheehan

constituents are available for further partition into adjacent dermal tissues, for systemic circulation, for metabolism, and/or for subsequent excretion. Only large organic molecules trapped in the outermost lay- ers of skin for extended periods are subject to sub- stantial exfoliative SC loss over a period of many days due to ongoing SC shedding and renewal.(24)

Daily dermal benzene dose Dskin (in mg) arising from dermal uptake during parts washing with MSS was thus modeled as the sum of uptakes via two mass- transfer processes: Process 1 involving continuous dermal immersion in liquid MSS (continuously re- leased by a parts washer hose) from which evapo- rative benzene loss is assumed to be negligible, and Process 2 involving transient splash and/or wipe de- position of MSS onto skin from which benzene is ex- pected to evaporate. This study applied a detailed pa- rameterization of this general modeling approach for workers cleaning without protective gloves, assuming (as discussed below) that dermal uptake of benzene in MSS is proportional to dermal uptake of MSS.

Process 1, absorption during continuous contact exposure, involves direct uptake into skin of the MSS constituents, including a well-dissolved benzene concentration (Cbulk). For Process 1 to occur, the skin must be in continuous contact with liquid MSS. Therefore, a simplifying assumption is used that sol- vent adjacent to skin is mixed and/or replenished at a rate sufficient to ensure that Cbulk remains constant during the entire duration (T) of liquid contact with skin. This assumption is valid for parts washing by workers whose hands are partly drenched with sol- vent that is continuously supplied from a dispenser hose during the washing process, consistent with im- mersing the hands in the solvent.

Via the more complex Process 2, benzene in MSS can enter the skin after a small amount of solvent is splashed or wiped onto the skin, forming a thin layer subject to evaporation as well as dermal uptake. The fraction of volatile solvent mass deposited as a thin layer on skin at any point in time after initial contact is approximately linearly proportional to the total duration during which that solvent remains in contact with exposed skin, and has traditionally been esti- mated as a percentage of dermally applied dose that is dermally absorbed.(25) However, in the absence of in vivo measurements of benzene uptake into human skin from MSS that contacts skin during tran- sient Process 2 events, a more accurate estimate of benzene uptake in vivo must address the kinetics of simultaneous evaporation and dermal uptake of MSS and any benzene it contains. By definition, Process

2 (a thin-layer process) cannot operate in an area of skin being affected by Process 1 (which involves sus- tained contact with a well-mixed liquid volume), but could occur multiple times with splashes in areas out- side that immersed in solvent during parts washing activities.

Total daily dose (Dskin), in milligrams (mg), of dermally absorbed benzene arising from the oper- ation of these two processes can be approximated in terms of corresponding process-specific measures of effective dermal load (EL1[continuous liquid con- tact] and EL2[transient splash and/or wipe contact], respectively). Each of these two measures of der- mal load incorporates a corresponding effective ef- ficiency of benzene transfer from MSS into skin that may be affected by skin damage, if applicable. Equations (1a–d), involving other variables defined below, express these relationships:

Dskin = CbulkU SA(EL1 + EL2), (1a) in which

EL1 = fTwet T (f Aliquid fEliquid J ), (1b)

EL2 = [fTwet T(fTsplash f Asplash fEsplash ksplash L0,splash ) + (f Awipe fEwipe L0,wipe)], (1c)

fEj = fBZ j [(1 − f ADj ) + NDf ADj ], for j = liquid, splash, wipe, (1d)

where Cbulk is the concentration of benzene in bulk MSS (ppmw), U is the unit conversion factor = 10−6 mg benzene/(mg MSS × ppmw), SA is the to- tal skin surface potentially contacted by MSS dur- ing parts washing (cm2), ELj is the effective der- mal load of MSS; defined by Equations (1b–c) (mg MSS/cm2, for i = 1,2), T is the total daily duration of parts washing (hour), fTwet is the fraction of the total parts washing duration T that is “active” in the sense that dermal liquid immersion-, splash-, or wipe- mediated dermal contact with MSS actually occurs (unitless), fTsplash is the fraction of the wet contact duration that involves splash-generating brushing ac- tivity (unitless), fAj is the fraction of SA that is po- tentially contacted by MSS during the parts washing contact process i, for j = liquid, splash, or wipe, where fAsplash = (1 − fAliquid) (unitless), fEj is the effective benzene exposure efficiency of contact process j, for j = liquid, splash, or wipe; defined by Equation (1d) (unitless), J is the dermal MSS flux (mg MSS/cm2/h),

Dermal vs. Total Uptake of Benzene 1341

ksplash is the fraction of area fAsplash SA covered by splashed MSS per unit time (h−1), L0,j is the ini- tial dermal MSS load from contact process j, for j = splash or wipe (mg MSS/cm2), fADj is the fraction of area fAj SA assumed to contain severe damage (e.g., cracking or removal) down to at least half the SC depth, for j = liquid, splash, or wipe (unitless), fBZj is the MSS-to-dermal-transfer efficiency for benzene (i.e., fraction of initial benzene mass contained in dermally contacted MSS that together with nonevap- orated MSS becomes dermally absorbed) for MSS- contact process j, for j = liquid, splash, or wipe (unit- less), and ND is the factor by which estimated ben- zene uptake through severely damaged skin is as- sumed to exceed that through normal skin (unitless).

Equations (1a–d) assume contact between a con- stant skin surface area (SA) with MSS containing a constant concentration (Cbulk) of benzene over a daily time interval (T). If C and SA vary over inter- val 0 ≤ t ≤ T as functions C(t) and SA(t) of time t, then Equation (1a) can be reexpressed as the cor- responding integral involving the product C(t) SA(t) over time t.(26) Parameters involved in the processes described above are discussed individually below. All of these parameters are summarized in Table I, to- gether with assumptions and estimates used, as ex- plained below, to model their respective values for the purpose of quantitative exposure assessment.

2.3.2. Model Parameters

2.3.2.1. Daily Parts Washing Duration (T). In- terindividual variability in lifetime time-weighted (TWA) average daily duration of time (T) during which parts washing activity occurs was assumed to have an average (±1 SD) value of approximately 28 (±24) minutes, and to have a compound lognormal distribution (Table I), based on an analysis of ac- tivity data obtained for parts washers as described previously.(10) Observations of workers performing parts washing show that portions of both hands and both lower forearms are contacted by solvent during parts washing.

2.3.2.2. Fraction of T Involving Active Wet and Splash Contact (fTwet, fTsplash). Methods used to es- timate the extent and duration of dermal contact, such as patches, body suits, or fluorescent tracers, are typically designed to detect accumulated dry- mass particles of material,(30,31) rather than poten- tial uptakes associated with contacted liquid. Der-

mal contact during occupational activities involving hand manipulations during liquid applications—such as application of biocides, or work with semisyn- thetic metal working fluids (i.e., products used as lu- bricants and coolants during the machining or treat- ment of metal components)—tends to be measured in units of cumulative mass contacted or potential flux (as mass per cm2/h), rather than in terms of ar- eas and/or durations of liquid contact per se, with re- sults that exhibit substantial magnitudes of interindi- vidual variability.(32–36) Only a percentage equal to 100(fTwet)% of total parts washing duration T was assumed to generate MSS exposure by either Pro- cess 1 or 2, where fTwet was estimated to be 100% for T < 10 minutes and 50% for T ≥ 10 minutes, based on field observations indicating that durations of wet-hand contact for manual work involving the use of liquids do not typically exceed 50%.(28,29) Of the total of the wet duration (fTwet T), only an uncer- tain fraction (fTsplash) was assumed to involve brush- ing activity that generates splash droplets contacting worker skin. It was assumed that splash exposure re- sults exclusively from this more aggressive cleaning activity. A limited set of observations of onsite pro- fessional parts washing activity made for this study indicated that daily activity may include little or no brushing, and were consistent with an assumption that the lifetime TWA value of fTsplash is unlikely to exceed 30%. Accordingly, joint variability and uncer- tainty in fTsplash was assumed to be symmetrically tri- angularly distributed between 15% and 30%.

2.3.2.3. Exposed Dermal Surface Areas (SA, fAi). The TWA value of total dermal surface area (SA) potentially contacted by MSS during parts washing for a reference male adult worker was de- fined in terms of hand (H) and forearm (FA) related components as SA = 2 SAH + SAFA = 1,810 cm2, where SAH = 535 cm2 = the area of one hand, and SAFA = 740 cm2 = half the total area (1,480 cm2) of both forearms, with interindividual variability in SA (and all regions thereof) assumed to be approx- imately normally distributed with a relative coeffi- cient of variation (CV%) equal to 10% (i.e., the same CV% as that estimated for total body surface area).(27) Based on qualitative observations made on multiple workers using parts washers that continu- ous immersion in solvent during washing typically in- volves the proximate surface of a single hand and (possibly) forearm that is typically used to grasp or manipulate the part being washed. Therefore, fAliquid

1342 Bogen and Sheehan

Table I. Summary of Two-Process Dermal-Uptake and Inhalation Model Parameters and Assumed Values

JUV Parametersb a, Variablea Typea Symbol Unita Distributiona b, . . . EV, SDa,b Basis (Section Discussed)

Benzene concentration in MSS – historical

U Cbulk ppmw Multi-LNC – 35.3, 15.1 Sheehan et al. (2010);(10)

see Section 2.1 in text Benzene concentration

in MSS – current U Cbulk ppmw Multi-LNC – 9.4, 5.0 Sheehan et al. (2010);(10)

see Section 2.1 in text

Interworkplace Cair variability

V T41 – T(41) – 0, √

3 See Section 2.4 in text

Total ventilation V Qtot m3 /h LN(a, b) ln(1.21), ln(1.28) 1.247, 0.312 U.S. EPA (2011);(27) see Section 2.4 in text

Duration of potential MSS contact

V T h LNC (a,b) 0.353, 2.09 0.47, 0.40 Sheehan et al. (2010);(10)

see Section 2.3.2.1 in text

Fraction of fTwet T involving brushing

U, V fTsplash – Tri(a,b,c) 0.15, 0.225, 0.30 0.25, 0.0306 See Section 2.3.2.2 in text

Surface area potentially contacted

V SA cm2 N(a,b) 1810, 181 1810, 181 EPA (2011);(27) see Section 2.3.2.3 in text

Immersed fraction of SA V fAliq – Tris[a/b, 1/2] 535, 1810 0.296, 0.0603 See Section 2.3.2.3 in text

Splashed fraction of SA U/V fAsplash – Tris[a/b, 1/2] 1235, 1810 0.682, 0.139 See Section 2.3.2.3 in text

Wiped fraction of SA V fAwipe – Tris[a/b, 1/2] 1070, 1810 0.591, 0.121 See Section 2.3.2.3 in text

Benzene uptake efficiency from MSS splashed on skin

V fBZsplash – LN(a, b) 0.18325, 1.34528 0.1914, 0.05794 See Section 3.1 in text

Benzene uptake efficiency from MSS wiped on both hands

V fBZwipe – X2 (a, b) 13.3405, 51.1755 0.2604, 0.1034 See Section 3.1 in text

Dermal flux of MSS V J mg/cm2 /h LN(a, b) 3.0, 1.5 3.39, 1.43 See Sections 2.2.3, 3.1 in text

Uncertainty in ln J sample mean

U T3 – T(3) – 0, (41/39)1/2 See Section 2.2.3 in text

Dermal splash coverage rate

V ksplash h−1 BiU(a,b,c,0.75) 0.2, 0.6, 1.2 0.525 0.254 Section 3.2 in text

Fraction of T involving C fTwet – CC: 1 if T < T 0.577, 0.180 Wassenius et al. (1998);(28)

wet contact 10/60, else 0.5 Cherrie and Aspley (2007);(29) see Section 2.3.2.2 in text

Dermal load for splashed MSS

C L0,splash mg/cm2 C 7.0 7.0 See Section 2.3.2.4 in text

Dermal load for wiped MSS

C L0,wipe mg/cm2 C 1.0 1.0 See Section 2.3.2.4 in text

Damage-enhanced uptake factor

C ND – C 5 5 See Section 2.3.2.6 in text

Benzene uptake efficiency from liquid MSS on skin

C fBZliquid – C 1 1 See Section 3.1 in text

Fraction of (fAliquid × SA) severely damaged

C fADliquid = fADwipe – C 40/535d 0.0748 See Section 2.3.2.3 in text

Fraction of (fAsplash × SA) severely damaged

C fADsplash – C 40/1275d 0.0221 See Section 2.3.2.3 in text

aADP = apparent dermal permeability, JUV = joint uncertainty (U) and interindividual variability (V), – = unitless, EV = expected arithmetic mean value, SD = expected standard deviation. Distribution types (parameters): LN(a, b) = lognormal with a = geometric mean (GM) and b = geometric SD (GSD); LNC(a,b) = compound lognormal (a = GM, b = GSD, log-values of which were estimated to have corresponding error-SD values of 0.051 and 0.063, respectively); Multi-LNC = multiple facility-specific LNC models were each sampled with equal likelihood; N(a,b) = normal (a = EV, b = SD); T(df) = Student t-distributed with df degrees of freedom; BiU(a, b, c, p) = biuniform as U(a,b) with likelihood p and U(b,c) with likelihood 1 – p, where U(x,y) = uniform between x and y; Tri(a,m,b) = triangular between a and b with mode at m; Tris(m,p) = Tri(p m, m, [1 + p]m) for 0 ≤ p ≤ 1; C = constant (in which case SD = 0); CC = conditional constant; ln denotes natural logarithm. All 15 of the distributed variables listed are assumed to be independent except fTwet, which is modeled as the indicated function of each simulated value of T. bParameters are listed in the order they appear in the specified distribution. Dash (–) denotes “not applicable” or that omitted values correspond to the described empirically-based distribution(s) evaluated by Monte Carlo simulation.

Dermal vs. Total Uptake of Benzene 1343

was modeled by assuming that, typically, one hand is contacted approximately 75% of the time and the other 25% of the time by MSS during period fTwet T associated active parts washing, implying fAliquid = (3/4 SAH + 1/4 SAH)/SA = SAH/SA. For splash- and wipe-related MSS exposures, corresponding frac- tions were assumed to be fAsplash = (1 – fAliquid) + (SAFA/SA) = (SAH + SAFA)/SA, and fAwipe = 2 fAliquid = 2 SAH/SA, respectively. Associated in- terindividual variability in fractions fAi were each assumed to be triangularly distributed with a corre- sponding modal value of fAi, ranging from 0.5fAi to 1.5fAi.

2.3.2.4. Dermal Flux of MSS and Benzene (J, L0,j, fBZj). Few studies provide relevant data for estimating in vivo dermal flux (JBZ) of benzene from MSS in contact with skin. Measures of steady-state dermal permeability (Kp = J/C) obtained from in vitro diffusion-cell studies have been found gen- erally to be lower than corresponding estimates of apparent “effective” permeability (Kp,eff) obtained from in vivo studies of short-term aqueous dermal exposure to dilute organic chemicals, even after increasing Kp values to account for expected levels of nonsteady-state tissue loading.(37,38) This appears also to be true to an even greater extent for organic liquid exposures, as shown by limited available data summarized in Table II. Therefore, pertinent in vivo human data are preferred as a basis for estimating JBZ. Ideally, J could be measured in human volunteers dermally exposed to an effectively harmless (e.g., parts per trillion) concentration of 14C-radiolabeled benzene in MSS, which could be detected by accelerator mass spectrometry (AMS) analysis of samples from a dermal-exposure chamber in a well-controlled disappearance-method study. Absent such direct measures, JBZ can only be estimated by indirect and/or modeling methods.

Estimates of in vivo permeability of neat ben- zene into human skin (discussed in Appendix C) exhibit considerable inconsistency and, more impor- tantly, do not directly address the dermal permeabil- ity of benzene that is contained in MSS in which skin remains immersed in the context of Process 1. The effective permeability of three different formulations of liquid white spirits (a solvent similar to MSS) into tail skin of rats exposed for three hours overlap cor- responding values for neat benzene (Appendix C), suggesting that dermal flux of neat benzene and that of MSS may be similar in humans. Although human

in vivo data are not available to verify this hypothe- sis, an approach assuming such similarity is consistent with similar estimates of human in vivo permeability observed for six JP8 (jet fuel) contaminants that are each miscible in JP8; naphthalene, two methyl naph- thalenes, and dodecane all exhibited permeabilities estimated to differ by less than threefold, while those for decane and undecane were ∼10-fold lower.(52)

In view of these considerations, for dilute ben- zene as a miscible contaminant of MSS in which skin is immersed, JBZ was defined as the following fraction of MSS flux into skin: JBZ = Cbulk (10−6 g/μg) fBZliquid J. Benzene uptake by the splash or wipe MSS-exposure pathways associated with Pro- cess 2 were modeled as described in Appendix A, ac- counting for all expected joint, independent evapo- rative losses of MSS, and of benzene from MSS, as well as concurrent dermal uptake of MSS (includ- ing its benzene content) while MSS remains in con- tact with skin. By this approach, MSS in contact with healthy skin by process j includes an effective mass (Mj) from which all of any contained benzene is der- mally absorbed, where Mj = fBZj L0,j fAj SA. Be- cause continually contacted MSS is freshly released from a parts washer without prior benzene evapo- ration, it follows that fBZliquid = 1. Values of J and fBZj (j = splash, wipe) were estimated from exper- imental data, and were modeled in a way that ac- counts explicitly for uncertainty (due to small sam- ple estimation error) and interindividual variability (see Section 2.2.3 and Appendix A). Data were con- sistent with each splashdrop residue imparting an ini- tial load of L0,splash = ∼7.0 mg/cm2 (see Section 3.2), which also is approximately that estimated postim- mersion for bath oil on human skin (see tables 7– 24 in Ref. 27). Dermal MSS-loss experimental re- sults (see Sections 2.2.2, 3.1) showed that greater MSS loads are infeasible, and lesser loads (particu- larly on BH and FA skin) tend to spread over time to a smaller value we approximated as L∞,splash = L0,wipe = 1.0 mg/cm2, which is also the approxi- mate load observed for partially wiped bath oil af- ter skin contact (see tables 7–24 in Ref. 27). Ben- zene was assumed to evaporate by first-order loss from the relatively rapidly evaporating fraction of MSS (see Section 3.1). The corresponding first-order rate constant (kBZ = 0.242/min) was estimated from experimental data of Hui et al.(23) indicating that no more than ∼0.5% of 14C-radiolabelled benzene re- mained in an “open-air” 10 μL/cm2 MSS load placed in a glass vial at room temperature (25 ◦C) for 30 min. This estimated rate accounts for a slightly

1344 Bogen and Sheehan

Table II. In Vivo Versus In Vitro Dermal Uptake of Neat Organic Liquids

In Vivo Kp,eff

Kp,eff Data In vitro Kp Kp Data Kp,eff/Kp Ratio

Chemical Species (cm/h) Source(s) (cm/h) Source(s) (Unitless)

Aniline Human 0.0032 Baranowska- Dutkiewicz(39)

0.000293 Korinth et al.(44) 11

Benzene Human 0.0097a Dutkiewicz and Tyras;(40)

Baranowska- Dutkiewicz(39)

0.00064 0.0003 0.00011

Blank and McAuliffe;(45)

Franz;(46) Loden;(47) 15 32 88

2-Butoxyethanol Human 0.00022 0.00029

Johanson et al.;(41) Jakasa et al.;(42)

0.000019 0.00005

Lockley;(48)

Korinth et al.;(49) 11–15 4–6

Carbon disulfide Human 0.0077 Baranowska- Dutkiewicz(39)

0.0022b DuPont (2005)(50) 3.0

Toluene Human 0.0202 Dutkiewicz and Tyras(43) 0.0001 Ursin et al.(51) 200

aHanke et al. (1961) reported data corresponding to a lower mean Kp,eff (0.00043 cm/h) for seven measures of the net loss of 20 μL (17.6 mg) of benzene placed in a watchglass pressed for 1.25 hours against forearms of study volunteers in a study that preceded Dutkiewicz and Tyras (1969) in the same laboratory. bApproximated from dermal absorption reported during the first 30 minutes of extended in vitro dermal exposure to 1.2 mL/cm2 of 0.4319 g/mL CS2 in isopropyl myristate.

(∼1.37-fold) shorter duration over which this rate of benzene evaporation is expected to occur at skin temperature (32 ◦C).(53)

2.3.2.5. Exposure Duration. Mathematical models of dermal uptake predict that J tends to decrease over time as increased chemical concen- trations in dermal tissue compartments reduce the corresponding concentrations gradient that drives mass transfer across a dermal membrane according to Fick’s law.(54,55) This phenomenon has been observed in experimental studies of the uptake of organic chemicals such as aniline, styrene, and xylene into skin of human volunteers; for example, uptake of neat aniline was reduced ∼10-fold in volunteers dermally exposed for 300 compared to 30 minutes.(39) Equation (1a) was not modified to reflect an expectation that Kup may decrease over time during extended periods of exposure. As noted above, this approach is conservative to the extent such decrease may occur during dermal-immersion exposures that occur during total daily parts washing activity, which typically is relatively brief (Table I).

2.3.2.6. Skin Damage (ND, fADi). Percuta- neous chemical uptake is, in some cases, enhanced when this occurs through damaged skin, resulting in an ND-fold greater value of chemical flux than through nondamaged skin, where the factor ND de- pends on the chemical involved, as well as on the na- ture and extent of skin damage.(56,57) Because MSS constituents may damage skin under long continuous exposure conditions,(58) enhanced MSS-related der-

mal uptakes may occur in workers with exposed ar- eas of damaged skin on the hands. Data discussed in Appendix B indicate that ND is unlikely to exceed 5 for benzene from MSS through severely damaged ar- eas of human skin, but may be near 1 for slightly-to- moderately damaged skin. Considerable uncertainty is associated with generally applying this estimate to assess dermal uptake, insofar as it is based on neat benzene through skin denuded of SC, and smaller permeability shifts, or even permeability reductions, have been observed for uptakes involving milder lev- els of damage and/or different solvents. In view of these considerations, ND = 5 provides a reasonable upper-bound estimate of enhanced dermal uptake to assess the impact of damaged skin on dermal expo- sures of workers with small areas of the hands with severely damaged skin from cuts or cracks that con- tinued to use MSS for parts washing.

For quantitative exposure assessment, Equa- tion (1b) was evaluated by assuming that ND = 5 for severely damaged dermal surface area (SAdamaged) estimated to comprise, on average over a working lifetime, all of the area lying within a 2-mm width tracing all major crease lines together amounting to ∼7.5% of the surface of the bottom half of both hands (SAdamaged = 0.075 × SAH = ∼40 cm2). It was assumed on average that only half the sur- face of both hands is immersed during parts-washing tasks involving wet contact, that MSS splash can potentially affect all nonimmersed regions, and that the entire surface of both hands is wiped at the end of parts washing each day. The severely damaged frac- tions of dermal surfaces assumed to be subject to

Dermal vs. Total Uptake of Benzene 1345

such immersive contact, to MSS splash, and to wiping were thus assumed to be fADliquid = SAdamaged/SAH, fADsplash = SAdamaged/(SA – SAH), and fADwipe = fADliq/2, respectively (see Section 2.3.2.3). The rel- ative increased magnitude of predicted dermal expo- sure due to skin damage was examined by comparing results with and without the upper-bound ND factor incorporated (see Section 2.3.3).

2.3.3. Relative Contributions of the Two Modeled Processes

From Equation (1), the fractions of total dermal dose (Dskin) attributable to Processes 1 and 2 were denoted fD1 and fD2, respectively. The correspond- ing dermal exposure-enhancement factor FDskin/1 by which uptake due to Process 1 must be increased to equal that predicted to be due to both Process 1 and Process 2 is:

FDskin/1 = 1/(1 − f D2) (2) In terms of FDskin/1, total dermal uptake given by Equation (1a) may be reexpressed more simply as:

Dskin = FDskin/1CbulkSA EL1(10

−6g/μg), (3)

which can be used directly to approximate Equa- tion (1a), provided realistic bounds on fD2 are avail- able.

2.3.4. Relative Contribution of Uptake Through Damaged Skin

From Equation (1), the predicted fraction (fDD) of Dskin contributed by benzene uptake specifically through damaged skin is given by

f DD = 1 − [( Dskin conditional on ND = 1)/Dskin] (4)

in which Dskin and ND were defined in relation to Equations (1a–b).

2.4. Respiratory and Total Benzene Dose from Parts Washing

Daily respiratory benzene dose Dair, in mil- ligrams, arising from inhalation during work with parts-washing solvents from each period was esti- mated as described previously(10) using the following

equations:

Dair = Cair × CF × Qalv × Tevent, (5a)

Cair = 0.00215 Cbulk × 2.71T41 , (5b) where Cair is the estimated benzene concentration in air (in ppm), T41 is the a Student’s t-distributed ran- dom variable with 41 degrees of freedom, used to model interworkplace variability, CF is the air con- centration conversion factor for benzene = 3.19468 (mg/m3 per ppm), Qalv is the alveolar ventilation rate = 0.70 Qtotal (m3/h), Qtotal is the total ventilation rate (m3/h), and Tevent is the duration of occupational parts washing using MSS, conditional on occurrence of this activity on a given work day (h/day).

Individual variability in total ventilation (Qtotal) was assumed to be approximately lognormally dis- tributed with geometric mean (GM) and geomet- ric standard deviation (GSD) values of 1.21 m3/h and 1.28, respectively. These values correspond (by the method of moments) to an arithmetic mean of ∼1.25 m3/h and a coefficient of variation of 25% for Qtotal, and are within ∼10% of corresponding esti- mates for adult males between ages 20–60 assum- ing ∼60% light and ∼40% moderate working activ- ity, and for adult males doing car maintenance.(27)

Doubly-compound lognormally distributed interindi- vidual variability in the career average amount of time Tevent (h/day) (arithmetic mean ±1 SD = 0.47 ± 0.41 h/day; 50th, 92nd, and 99.5th percentiles of ap- proximately 0.42, 1, and 2.5 h/day) spent working at a parts washing station each day was based on sur- vey data for mechanics working at service facilities, as described by Sheehan et al.(10)

Because benzene concentrations in historical samples from parts washing operations were fre- quently below analytical detection limits, benzene in workplace air (Cair) was estimated from con- centrations from a broader data set of volatile constituents, including benzene, toluene, ethylben- zene, and xylenes (BTEX) in MSS and in air, in conjunction with a physical-chemical theory-based model shown to be statistically consistent with these data, as described previously.(10) Airborne benzene concentrations from use of historical MSS were based on the benzene content of MSS samples ana- lyzed 1992–1993 with benzene levels consistent with a limited number of samples from the 1980s (ear- liest from 1980). Airborne benzene concentrations from use of current MSS were based on the benzene content of MSS samples analyzed 1995–1999 and are

1346 Bogen and Sheehan

believed to be representative of the MSS used in parts washers from the mid-1990s to 2000s. By this approach, the estimated 50th (95th) percentile of Dair was determined to be 0.079 (0.77) mg for historical solvent, and 0.020 (0.20) mg for current solvent, re- spectively, and corresponding 95th percentile eight- hour TWA respiratory exposure estimates were es- timated to be <10% of the current Occupational Safety and Health Administration permissible expo- sure limit of 1.0 ppm for benzene.(10)

The total benzene dose (Dtotal) in milligrams per day for workers using parts washers was calculated for each period as the sum of inhalation and dermal doses defined previously:

Dtotal = Dair + Dskin, (6) from which the fraction (fDskin) of Dtotal attributable to the dermal pathway was defined as fDskin = Dskin/Dtotal, and the dermal exposure-enhancement factor (FDtotal/air) by which Dair must be increased to yield Dtotal was thus defined as FDtotal/air = 1/(1 – fDskin).

2.5. Probabilistic Analysis

Probabilistic risk assessment methods offer a dis- tinct advantage in characterizing human health risks, in that they explicitly provide a quantitative descrip- tion of the degree of variability and uncertainty in risk estimates.(11,12) Because probabilistic analyses in this study were intended to model a parts washing worker drawn at random from among all potentially applicable workers, a systematic distinction was not maintained between interindividual variability (i.e., person-to-person differences) and (if/as applicable) uncertainty (i.e., lack of knowledge due to estimation error, fundamental gaps in applicable data or theory, etc.) pertaining to model parameters summarized in Table I; for this type of prediction, no such distinction was required.(11,12) For example, interfacility variabil- ity per se in Cbulk was, for the purpose of this anal- ysis, treated effectively as a source of uncertainty contributing, together with regression-model estima- tion error, to aggregate uncertainty in Cair. Prelim- inary Monte Carlo calculations of estimated distri- butions for ln(GMJ), fBZsplash, and fBZsplash were carried out using 106, 20,000, and 20,000 nested ran- dom samples, respectively. Conditional on these re- sults, Monte Carlo calculations were then carried out using 20,000 systematic Latin Hypercube sam- ples of each of a total set of 14 simulated variables involved in models of dermal exposure to histori-

cal and current MSS (Table I). In each simulation, all variables were modeled as independent random variables, the target (identity) rank-correlation ma- trix being induced in each case using the method of Iman and Connover,(59) supplemented by Jennrich’s chi-square test to ensure no spurious significant dif- ference (p ≤ 0.05) occurred between each targeted and realized correlation matrix.(60) All calculations were performed using Mathematica R© 9.0 software.(61)

3. RESULTS

3.1. Experimentally Observed Rates of MSS Loss to Air and Human Skin

Data from stippled-Al control experiments done to measure rates of MSS evaporation to air are sum- marized in Fig. 1a, in which data from three repli- cates (23 open points) are shown together with a cor- responding biexponential model fit (curve), and data from an inverted Al-cup (smooth-surface) control experiment (solid square points). The fitted model is that described in Appendix A, with parameters, p = 0.376, b1 = 0.0754/min, and b2 = 0.00389/min (R2 = 0.864).

Nested F-tests performed (each yielding an Fdf1,df2 statistic with df1 and df2 degrees of freedom) showed that an outlying data point relative to the fit- ted model was significant (solid black circle, F1,20 = 57.2, p < 10−6; this point was excluded from other fits examined), that separate replicate-specific model fits did not significantly improve the overall fit to the data (F6,14 = 2.65, p = 0.062), and that a biexpo- nential model fit the data significantly better than a single-exponential model (F2,20 = 8.12, p = 0.0026). An outlier test comparing the fitted data in Fig. 1a to the measures obtained for evaporation from smooth Al surfaces (solid square points) shows the latter evaporation to be significantly slower than from ir- regularly stippled Al surfaces tested (F5,20 = 2.65, p = 0.00055).

The experimental measures of in vivo MSS (com- bined evaporative and dermal) loss from F, P, BH, and FA regions of human skin, and correspond- ing estimates of MSS dermal flux (J) calculated as described in Section 2.2.2.5 and Appendix 1A, are summarized in Table III. Two-way ANOVA ap- plied to log-transformed estimates of J indicate that subject (p = 0.018), but not dermal region (p > 0.05), was a significant categorical predictor (Ta- ble III). The sample GM and GSD of subject-specific

Dermal vs. Total Uptake of Benzene 1347

Fig. 1. (Top panel) data from three repli- cated experiments (23 open points) on MSS evaporation from an irregularly stippled, 1-cm2 Al surface at ∼32 ◦C, fit to a biexpo- nential loss model (curve) excluding one out- lier (solid circle); similar data using a smooth Al surface (solid squares, not fit). (Bottom panel) corresponding MSS dermal flux from each splash/wipe event in relation to its MSS load (L0) and assumed dermal flux (J).

estimates of flux J were estimated to be approx- imately 0.050 mg/cm2/min (i.e., 3.0 mg/cm2/h) and 1.49, respectively (Table III). The corresponding GSD estimate is virtually identical to that of 1.5 as- sumed in this study, which was calculated from values of dermal permeability for six organic JP8 contam- inants reported for 10 volunteers dermally exposed to JP8.(19) These estimates correspond to an effective dermal permeability of Kp,eff = 0.0037 cm/h, with an estimated 90% confidence interval (CI) of 0.0019– 0.0072 cm/h, assuming an MSS density of 0.8102 (W.M. Barr MSDS for Odorless Mineral Spirits).

Fig. 1b plots percentages of benzene dermally absorbed from illustrative values of splash/wipe (Pro- cess 2) MSS load L0 calculated by numerical eval-

uation of the model described in Appendix A, as functions of dermal MSS flux J. Conditional on GM = 3.0 mg/cm2/h and GSD = 1.5 for J, and an ini- tial posthand-wipe load equal to L0,wipe = 1.0 mg/cm2, this model was found to predict a cumulative prob- ability distribution function (cdf) for fBZwipe that is approximately (R2 = 0.99995) that of χ 2/51.17555, where χ 2 is chi-square distributed with df = 13.34046. Combining the same assumptions for J with assumed initial loads of L0,splash = 7.0 mg/cm2 for each area on which splashed MSS covers skin, the dynamic- spreading version of the same model (Appendix A) was found to predict a cdf for fBZsplash that is ap- proximately lognormal (R2 = 0.99995) with GM = 0.18325 and GSD = 1.34528.

1348 Bogen and Sheehan

Table III. Summary of Experimental Data on Dermal Uptake of Mineral Spirits Solvent (MSS)

MSS Load, L0 Uptake Duration, T0 Estimated Flux, J

Subject Dermal Region n Mean (mg/cm2) SDa (mg/cm2) Mean (min) SD (min) GMa (mg/cm2/min) GSDa (Unitless)

1 Back of hand 3 3.83 1.38 63.94 31.18 0.0383 1.41 2 Back of hand 3 2.13 0.50 22.11 4.16 0.0738 1.59 3 Back of hand 3 2.60 1.35 18.23 0.87 0.1019 1.73 1 Fingers 3 1.83 1.10 16.46 1.27 0.0811 1.90 2 Fingers 3 1.17 0.99 26.67 1.53 0.0259 2.39 3 Fingers 3 3.87 2.25 40.06 2.80 0.0588 1.91 4 Fingers 3 2.13 1.23 44.07 6.09 0.0289 2.13 1 Forearm 3 4.20 1.23 84.65 12.09 0.0274 1.15 2 Forearm 3 2.90 1.11 49.52 14.58 0.0372 1.03 3 Forearm 3 2.40 0.69 15.89 3.34 0.1219 1.07 1 Palm 3 2.20 1.42 19.98 0.58 0.0754 1.92 2 Palm 3 1.20 0.62 19.33 2.52 0.0435 2.24 3 Palm 2b 3.20 0.71 32.67 9.66 0.0704 1.80 4 Palm 3 2.47 0.74 42.66 15.8 0.0402 1.50

Allc Means 4 2.62 0.48 36.22 9.57 0.0527 1.19 Meansc All 4 2.54 0.57 36.30 9.99 0.0499 1.49 All All 41 2.57 1.34 35.51 22.11 0.0520 1.96

aMean = arithmetic mean, SD = standard deviation, GM = geometric mean, GSD = geometric SD. Values mentioned in Section 3.1 are listed in bold. bExcludes one replicate for which the corresponding calculated ln(J) value is a significant outlier for ln(J) data grouped by subject and dermal region, as determined by two-way ANOVA (F1,34 = 15.8, p = 0.00035). cSubject, but not dermal region, is a significant categorical predictor of ln(J), as determined by two-way ANOVA performed on the ln(J) data grouped by subject and dermal region (F1,34 = 3.86 and 0.67, and p = 0.018 and 0.58, respectively).

3.2. Dermal Splash Deposition from Simulated Parts Washing

Experimental dermal splash droplet deposition data obtained from simulated parts washing by two subjects are summarized in Table IV. The data indicate that for each of the two (right-handed) subjects studied, splash exposures measured on the left compared to the right-side hand/FA splash dosimeter (SPD) were significantly greater (by ∼10- fold and 22-fold for subjects A and B, respectively) (Table IV). Subject A, who had a (∼1.7-fold) greater brushing rate, had a similarly (∼2.2) greater splash exposure to the dominantly exposed left-side SPDs (Table IV). The quite vigorous brushing rate of 4.0 cycles/sec used by subject A reflects an upper bound that is plausible for only intermittent periods. There- fore, the biuniform distribution compared in Fig. 3 to the subject-specific ksplash distributions was used to model interindividual variability in the lifetime TWA value of ksplash applicable to parts washers. The biuniform ksplash distribution predicts this variability to range from 20%/h to 60%/h (the approximate range of values exhibited in all six 3-minute splash deposition experiments done by subject A and three

of six done by subject B) with 75% likelihood, and from 60%/h to 120%/h with 25% likelihood.

3.3. Two-Process Exposure Model Simulation Results

Fig. 2 shows the calculated distributions of cumulative-average daily, dermally absorbed ben- zene dose (Dskin) associated with parts washing using historical or current MSS. Again, the distri- butions shown both reflect joint uncertainty and intraindividual variability, pertaining to a worker selected at random with respect to the model- parameter distributions summarized in Table I. Corresponding statistics for the Dskin distribution are summarized in Table V, where they are compared to those for Dair and Dtotal for each time period. Av- erage values of each dose estimate listed in Table V for each time period are proportional to the average benzene concentration in MSS (Cbulk) applicable to that period (see Table I), so those applicable to each current estimate are each approximately 26.6% of the corresponding value applicable to the

Dermal vs. Total Uptake of Benzene 1349

Table IV. Summary of Experimental Simulated Parts Washing Splash Dataa

Fellipsoid Aspect Ratio (Unitless) (Unitless) Area (mm2)

Subject Side Experiment n Ave GSD Ave GSD Ave GSD

A Left 1 32 0.701 1.97 865.5 A Left 2 38 0.431 1.87 2026.8 A Left 3 45 0.668 1.93 1086.3 A Left 4 33 0.501 2.87 1197.8 A Left 5 26 0.650 2.15 1455.0 A Left 6 23 0.124 3.30 3516.7 A Left All 197 0.512 1.93 2.35 1.27 1691.4b,d 1.66

A Right 1 4 1.000 1.30 50.5 A Right 2 12 1.000 1.62 162.1 A Right 3 16 0.838 5.82 163.3 A Right 4 27 0.844 2.23 141.3 A Right 5 36 0.832 3.31 229.9 A Right 6 22 0.838 2.00 238.5 A Right All 117 0.892 1.1 2.71 2.71 164.3b 1.76

B Left 1 21 0.564 2.55 281.8 B Left 2 22 0.135 2.79 986.2 B Left 3 25 0.181 2.44 891.5 B Left 4 33 0.357 2.40 1282.0 B Left 5 73 0.405 2.81 843.8 B Left 6 15 0.811 2.22 322.1 B Left All 189 0.409 1.97 2.54 2.53 767.9c,d 1.88

B Right 1 3 1.000 1.43 29.1 B Right 2 2 0.450 1.00 6.5 B Right 3 2 1.000 1.67 22.3 B Right 4 3 0.337 1.59 64.6 B Right 5 2 1.000 1.50 2.5 B Right 6 4 1.000 3.95 81.5 B Right All 16 0.798 1.64 1.86 1.86 34.4c 3.84

aFellipsoid = area-weighted fraction of all splash droplet patterns that were classified as approximately ellipsoid (E) (rather than as triangular [T] or rectangular [R]) for the purpose of estimating splash droplet area A[type] as: A[E] = π LW, A[T] = LW/2, A[R] = LW, where L = greatest length, W = width or height perpendicular to L, aspect ratio = L/W, Ave = arithmetic mean, GSD = geometric standard deviation. Total splash-dosimeter area = 40,940 mm2. bSignificant difference by side (F1,5 = 16.0, p = 0.010), but not by experiment (F5,5 = 1.21, p = 0.42), by two-way ANOVA. cSignificant difference by side (F1,5 = 20.2, p = 0.0064), but not by experiment (F5,5 = 0.94, p = 0.53), by two-way ANOVA. dSignificant or marginally significant left measure difference by subject (p = 0.041 by Wilcoxon test, p = 0.023 by Kolmogorov two-sample test, p = 0.058 by two-tail t-test). The same comparisons were all nonsignificant (p > 0.40) when measures for subjects A and B were normalized by corresponding subject-specific brushing rates (4.0 and 2.4 cycles/sec, respectively).

historical period. As noted in Table V, an average of approximately one-third of the total benzene dose associated with parts washing (regardless of time period) is estimated to be due to the dermal route of exposure, implying that this route accounts for a substantial, although typically not dominant, fraction of total estimated benzene dose. Values of Dskin for both periods correlate substantially positively with those of Dtotal (R2 = 0.58), while values of the dermal fraction (fDskin) of Dtotal also correlate substantially,

albeit slightly weaker and negatively, with those of Dtotal (R2 = 0.40).

The fraction of the dermal dose associated with the two uptake processes as described by variables fD1, fD2, and FDskin/1 defined in relation to Equation (2) (see Section 2.3.3), and the frac- tion predicted to occur through damaged skin as described by variable fDD defined in relation to Equation (4) (see Section 2.3.4) are also shown in Table V. Based on both median and mean estimates,

1350 Bogen and Sheehan

Fig. 2. Cumulative distribution of estimated dermal benzene dose associated with expo- sure to historical and current MSS used in parts washers. The distribution for histori- cal MSS has median, arithmetic mean, and 95th percentile values of 0.027, 0.035, and 0.085 mg/day, respectively, while that for cur- rent MSS has corresponding values of 0.0069, 0.0093, 0.024 mg/day, respectively.

Subject A

Subject B

0.0 0.5 1.0 1.5 2.0 0.0

0.2

0.4

0.6

0.8

1.0

Fraction of area covered per hour

C u

m u

la tiv

e lik

e lih

o o

d

Fig. 3. Cumulative empirical distributions for the rate (ksplash) of splash droplet cov- erage measured on the left upper hand and lower forearm for two subjects during simulated, conservatively dynamic parts washing (six 3-minute experiments per subject) (solid functions). Based on these data, the plotted biuniform distribution (dashed function) was used to model in- terindividual variability in the lifetime TWA value of ksplash for all parts wash- ers, noting that subject B used a sustained, relatively vigorous brushing rate of 4.0 cycles/sec.

continuous contact and splash/wipe dermal expo- sures were approximately equal contributors to total dermal dose. The fraction of dermal dose associated with estimated uptake through damaged skin was approximately 22.5%.

4. DISCUSSION

Even the recent improvements in sensitivity in the biological monitoring of workers exposed to ben- zene, such as using urinary benzene as a marker,(62)

have had limited success in distinguishing and char- acterizing exposures at the low levels associated with

petroleum solvents such as MSS. This fact empha- sizes the importance of collecting representative air and solvent concentrations of benzene and relying on representative models to estimate and characterize worker exposures. Many solvent use scenarios in- volve both continuous contact and incidental splash and/or wipe (thin-layer) exposures and these two distinctive types of exposure require different expo- sure models. Most dermal exposure assessments to date have not recognized the need to assess and sum the dose from both types of dermal contact to ob- tain a total dermal dose for the exposed worker. This need was recognized recently by Petty et al.,(7) who

Dermal vs. Total Uptake of Benzene 1351

Table V. Summary of Simulation Results Using the Two-Process-Dermal and Inhalation Model

Estimators

Variablea Symbol Unit Median Arithmetic Average 90th Percentile 95th Percentile

Dermal dose, Historical MSS Dskin mg/day 0.027 0.035 0.065 0.085 Inhalation dose, Historical MSS Dair mg/day 0.066 0.17 0.38 0.63 Total dose, Historical MSS Dtotal mg/day 0.098 0.20 0.43 0.69

Dermal dose, Current MSS Dskin mg/day 0.0069 0.0093 0.018 0.024 Inhalation dose, Current MSS Dair mg/day 0.017 0.045 0.10 0.17 Total dose, Current MSS Dtotal mg/day 0.025 0.054 0.12 0.18

Frac dermal via Process 1b,c fD1 – 0.51 0.51 0.76 0.81 Frac dermal via Process 2b,c fD2 – 0.49 0.49 0.74 0.79 Frac dermalc fDskin – 0.29 0.33 0.65 0.75 Frac due to skin damagec fDD – 0.225 0.224 0.228 0.229 DEEF due to Process 2c FDskin/1 – 1.97 2.36 3.83 4.81 DEEF due to dermalc FDskin/air – 1.41 1.85 2.89 4.03

aHist = historical, Curr = current, Frac = fraction, DEEF = dermal exposure enhancement factor, – = not applicable. Estimates listed are rounded to 2 or 3 significant digits (as indicated), and all assume that daily parts washing has occurred on the day for which the listed estimate is applicable. bListed estimates apply to both historical and current MSS. cVariable subject to the constraint: 1 = fD1 + fD2, such that fD1 and fD2 are perfectly negatively correlated (r = –1).

developed a two-process framework; however, they applied non-MSS-specific flux value for dermal ex- posure to benzene rather than the value for MSS esti- mated in this study, and failed to provide data needed to evaluate competing volatilization and dermal up- take components that are simultaneously occurring during thin-layer MSS (or other petroleum solvent) exposures.

In this study, an alternative two-process dermal exposure framework was developed and parameter- ized for detailed application to workers washing parts with MSS. This framework is supported by data from in vivo human studies of benzene uptake from MSS by both processes and worker-derived data for most other dermal exposure parameters. Detailed, experi- mentally supported models of dermal uptake for im- mersion and splash exposures to benzene from MSS were incorporated. The approach described provides a basis to refine exposure estimates for combined continuous contact and splash/wipe exposures and to compare inhalation and dermal doses from parts washing with both historical and current MSS for- mulations. Interestingly, the results obtained suggest that benzene uptakes from combined splash and wipe exposures to MSS typically contribute approximately as much to total estimated benzene dose as those from continuous dermal immersion in MSS during parts washing. Continuous contact uptake generally has been considered to be the dominant form of dermal exposure for parts washers, and frequently

has been the only dermal uptake process evaluated. The fraction of cleaning that involves active brush- ing (governed by fTsplash, defined to determine the relative contribution of splash exposure to the der- mal dose) is thus clearly an important element of the proposed exposure model that merits improved characterization.

The in vivo measure of Kp,eff for MSS and its constituent benzene (0.0037 cm/h) obtained in this study is slightly (approximately two- to threefold) below the range reported for in vivo percutaneous penetration of white spirits solvent into the tail of a rat (0.0075–0.0093 cm/h)(63) and for in vivo absorp- tion of neat benzene into the skin of human subjects (0.0092 cm/h).(39) However, the in vivo value of Kp,eff for MSS measured in this study is substantially (∼30- fold) greater than a Kp estimate for benzene from re- cycled Safety-Kleen 105 (0.00013 cm/h) based on an in vitro study using occluded excised human skin.(23)

This difference between in vivo and in vitro estimates is consistent with the recent comparison of aqueous chemical uptake by Bogen(38) and the comparisons of neat versus in vivo chemical uptake summarized in Table II.

A major source of uncertainty/variability in this assessment is associated with the assumption of enhanced dermal uptake through damaged skin, concerning which somewhat contradictory evidence exists among relevant published studies.(57,58,64–69)

Estimates of rates at which parts washers continue

1352 Bogen and Sheehan

daily dermal contacts with MSS involving areas of severely damaged skin on their unprotected hands throughout their parts washing careers are not available and merit future study. We assume that our inclusion of a damaged skin component makes the resulting assessment at least somewhat conservative, and by definition to overestimate dose to parts wash- ing workers with less damaged or undamaged skin. Doses to specific parts washing worker subpopula- tions could be estimated by applying the two-process framework used without its damaged-skin com- ponent, or using a damaged-skin enhancement factor estimated for a specific set of workers, if/as applicable. A recent review of benzene absorption by dermatologists suggested that a doubling of uptake rate through compromised or lesioned skin was reasonable, as opposed to the fivefold increase for seriously damaged skin that was used in this evaluation.(70) Workers with undamaged skin are ex- pected to have received dermal doses approximately 20% smaller than reported in this assessment (see Fig. 3).

The mean dermal contribution to total ben- zene dose from the use of MSS in parts washing was estimated to be nonnegligible but minor, ap- proximately 33%. Similar exposure calculations for parts washing using a solvent with a much higher benzene content, such as gasoline (>15,000 ppm by weight), indicate that dermal uptake can be a much greater contributor to total occupational benzene dose, likely overwhelming the respiratory contribution.(71)

4.1. Comparison to Standards and Criteria

The estimated daily exposures to benzene from MSS used in parts washing solvent, accounting for joint respiratory and dermal exposure pathways, are low. Daily total benzene doses estimated in this study for mechanics using parts washers (e.g., median doses of 0.025 and 0.098 mg/day for current and histori- cal solvent formulations, respectively) are less com- parable to workplace standards than to environmen- tal benzene exposures (e.g., they are both less than the median daily dose of 0.15 mg/day estimated for a nonoccupationally exposed, nonsmoking adult).(72)

Neither the current OSHA permissible expo- sure limit (PEL) of 1.0 ppmv as an eight-hour TWA exposure,(73) nor the American Conference of Gov- ernmental Industrial Hygienists eight-hour TWA threshold limit value of 0.5 ppmv (ACGIH 2012),(74)

directly addresses expected dermal uptake of ben-

zene via occupational activities that also involve res- piratory exposure. Results from this study indicate that, while dermal benzene uptake may not typically be a dominant pathway of benzene exposure during parts washing with MSS, it is nevertheless estimated to be a nonnegligible pathway, on average account- ing for about one-third of total predicted dose.

APPENDIX A. MODEL OF MSS LOSS OBSERVED BY EVAPORATION AND DERMAL UPTAKE

Control data on MSS evaporation from irregu- larly stippled Al surface at 32 ◦C were fit to the bi- exponential loss model: % MSS remaining = 100% { p exp(–b1t) + (1 – p)exp(–b2t)} for 0 < p < 1/2 with time t measured in minutes, corresponding to rates of change in two corresponding components of MSS mass/cm2 defined by dLi(t)/dt = –biLi(t) for i = 1,2, with L(t) = �iLi(t) and Li(0) = piL0, where {p1, p2} = {p, 1 – p}, and L0 = L(0) = initial MSS load (mg/cm2). Accounting also for loss due to dermal flux J (here measured in mg/cm2/min), the rate model was modified as follows: dLi(t)/dt = –biLi(t) – J (1 + g(t)) fi(t) conditional on L(t) ≥ 0, where here g(t) = 0, fi(t) = Li(t)/L(t). To adhere to the nonnegativity con- straint, loss conditional on L0 > 0 and J > 0 is as- sumed to occur only until time T0, defined as the time at which MSS(t) has decreased to 0. Values of J con- ditional on experimentally observed {L0, T0}, and of T0 conditional on { L0,j, J} for MSS-source processes j = splash or j = wipe were solved numerically us- ing the Mathematica 9.0 NDSolve function.(61) Ben- zene evaporation from MSS deposited from these two sources was modeled as first-order loss, governed by a rate constant kBZ estimated from published ex- perimental data (see Section 2.3.2.4). The TWA frac- tion (fBZj) of benzene remaining in MSS during the interval 0 ≤ t ≤ T0 was thus modeled as fBZj = [1 – exp(–kBZT0)]/(kBZT0), which was solved numeri- cally as a function of {L0,j, J}, where J was mod- eled stochastically as described in Section 2.2.3. Note that j = liquid corresponds to an immersion scenario for which kBZ = 0, so fBZliquid = 1. A final wipe at the end of parts washing was assumed to leave both hands with initial MSS load L0 = L0,wipe = 1.0 mg/cm2 (see Section 2.3.2.4).

To estimate benzene uptake from splashed MSS, the approach above was modified as follows based on experimental splash simulation results (see Sec- tions 2.2.2 and 3.2). All MSS loads were assumed to spread at rate s = 0.1/min such that (absent MSS

Dermal vs. Total Uptake of Benzene 1353

evaporation and flux) L(t) = L(∞) + [L(0) – L(∞)] exp(–s t) with time t measured in min, and L(t) = 7.0 and 1.0 mg/cm2 at times t = 0 and t = ∞, respec- tively, where by definition, L0,splash = L(0). The cor- responding MSS-covered area A(t) at time t under those conditions was similarly modeled as A(t) = A0 [1 + g(t)], where A0 = A(0), and here g(t) = 6 [1 – exp(–s t)]. The fraction fBZsplash was thus modeled as described in the preceding paragraph, conditional on the latter definition of g(t) and on L(t) defined as mg of splash-deposited MSS at time t divided by effective area A0 initially covered by splashed MSS, where A0 = ksplash fAsplash SA (see Sections 2.2.3 and 2.3.2.3).

APPENDIX B. ENHANCED MSS-RELATED CHEMICAL UPTAKE DUE TO SKIN DAMAGE

In vivo application of MSS constituents nonane and dodecane (but not tetradecane) to hairless rat skin for one hour induced mild/moderate skin irritation, as measured by elevated transepidermal water loss and visual erythema scores.(58) Rhesus monkey skin damaged by repeated tape-stripping to remove the entire SC was approximately five times more permeable to neat benzene than intact healthy skin, implying that ND = 5 for neat benzene through very severely (i.e., maximally) damaged skin.(64)

Nielsen(65) tested human skin that was only slightly damaged by three-hour in vitro pretreatment with SC-disrupting 0.3% sodium lauryl sulfate. While permeability values measured for several hydrophilic chemicals were between two- and threefold greater than corresponding values measured using intact hu- man skin for the two most hydrophobic compounds (log Kow > 3) tested, this damage-related increase in permeability was found to be less than twofold.(65)

Likewise, 123I-radiolabled protein permeability of patient skin exhibiting psoriasis was observed to be 10-fold greater than that of uninvolved skin in which dermal uptake was measured clinically by gamma counting and imaging.(66) In contrast, measures of in vivo dermal absorption of hydrophobic solvents toluene and 1,1,1-trichloroethane into shaved back and neck skin of outbred guinea pigs, normally rapid over the first 50 minutes of dermal exposure, were all observed to decrease significantly (i.e., implying ND < 1) in skin that was damaged by pretreatment with any of four types of subacute injury (tape strip- ping to remove the SC, sandpaper abrasion, needle abrasion, or delipidization by five-minute expo- sure to 2:1 chloroform:methanol 30 minutes prior to

test-chemical exposure), or that was damaged by pre- treatment inflicting either of two types of subacute contact-dermatitis injuries (treatment with a white spirit irritant 48 and 24 hours prior to test-chemical exposure, or formaldehyde-induced allergic contact dermatitis; a second contact-allergy treatment— colophony—resulted in no or minimal reduction in dermal uptake).(67) These observations are at odds with the hypothesis that significant correlations between (albeit involving only slight elevations in) urinary toluene and xylene metabolites (hippuric and methylhippuric acid) in a subset of 72 male Japanese car-body painters (who all wore Tyvek dust-proof overalls, gas masks, and nylon static gloves), and cor- responding visual scores of hand eczema (assessed by dermatologist-scored dryness, erythema, scaling, dyshidrosis, fissuring, and lichenification), indicate significantly enhanced uptake of toluene and/or xy- lene in workers with moderate to severe eczema.(68)

Although prolonged or repeated contact with gaso- line in an occupational setting is observed to deplete SC lipids and increase the prevalence of hyperker- atosis, dryness, onychosis, and dermatitis involving fissuring of the skin and nail disorders,(69) the effect of chronic gasoline-induced dermatitis on percu- taneous uptake of organic solvents remains to be quantified.

In guinea pigs continuously dermally exposed to neat 2-butoxy ethanol, dermal uptake was substan- tially enhanced by prior treatment of the absorp- tive region by severe mechanical abrasion.(67) How- ever, the opposite was true for the more lipophilic neat solvents toluene and 1,1,1-trichloroethane. Prior treatment of the absorptive dermal region by ei- ther acute injury (using methanol:chloroform mix- ture), mechanical injury (stripping with tape or scratching with a needle), or chemical injury (using methanol:chloroform mixture) were all observed to significantly reduce dermal uptake of toluene and 1,1,1-trichloroethane as measured by periodic mea- surement of blood concentrations of the tested chem- icals, implying that ND < 1 for these endpoints and chemicals.(67)

APPENDIX C. PREVIOUS ESTIMATES OF IN VIVO DERMAL PERMEABILITY OF BENZENE

Two in vivo studies of human volunteers der- mally exposed to neat benzene(40,75) each estimate an “effective” (time-weighted average, nonsteady-state) value of the dermal permeability (“Kp,eff”) of neat

1354 Bogen and Sheehan

benzene. An arithmetic mean (±1 SD) of Kp,eff = 0.0092 (±0.0031) cm/h can be calculated from apparent dermal permeability measures from the in vivo study reported by Dutkiewicz and Tyras(40)

(originally in Polish with an English abstract; a summary of benzene data from this study was repub- lished in English by Baranowska-Dutkiewicz),(39)

who used the difference method to measure dermal uptake of neat benzene in seven volunteers who each were exposed on a small area of the forearm to an initial volume of 0.15 mL of neat benzene for 10–15 minutes. Methods information from an earlier study of human dermal uptake of ethylbenzene, done by the same group and cited by those authors,(76)

indicates that neat test chemical was placed in a 4.7-cm diameter/1.2 cm-depth watch glass that “was tightly affixed to the skin of the forearm” of each subject. Normal skin has sufficient plasticity to form an exposure-chamber compression seal in vivo that under neutral airflow conditions is nearly or virtually completely resistant to vapor exchange or loss over brief periods less than one hour by volatile chem- icals, including chloroform and trichloroethylene, which have vapor pressures that bracket and are within 50% of that of benzene.(77) Such virtually air-tight seals are analogous to those produced for extended periods around human eyes by swim goggles.

The mean Kp,eff value of 0.0092 cm/h obtained by Dutkiewicz and Tyras(40) is 20-fold greater than that (0.00046 cm/h) implied by data reported in the earlier Hanke et al.(75) study from the same research group, which involved exposures that similarly involved neat benzene contained in a 8.0-cm watch glass containing 17.6 mg benzene pressed against forearm skin. The protocol used in the earlier study differed in two key respects that indicate its reported perme- ability estimate for benzene is likely to underestimate the value of Kp,eff most appropriate for the present analysis. First, the earlier study involved exposures each lasting 1.25 hours, an exposure period that was five times greater than that used by Dutkiewicz and Tyras, and was also substantially greater than typical total daily durations of parts washing activity (Table I). After exposing guinea pig skin to various durations to neat benzene in vivo, a range of de- generative changes and junctional separation were observed, including changes of slight to moderate severity after just 15 minutes of exposure, which by one hour progressed to “rather severe,” “more pro- nounced,” or “almost continuous” levels of severity for several endpoints, and generally increased in

severity thereafter.(78) Dermal uptake of each of two nonalcoholic organic solvents (toluene and 1,1,1- tetrachloroethane) was in each case reduced after four acute types of injury and four types of subacute injury investigated, with the most pronounced (∼5- to 10-fold) reductions observed during the first hour or two after just five minutes of delipidizing exposure to a mixture of chloroform and methanol.(67) Because benzene itself is an effective delipidizing agent,(79) it is likely that measures of benzene permeability re- ported by Hanke et al.(75) involving prolonged dermal exposures to neat benzene underestimated effective benzene permeability, as applicable to exposure periods typically substantially less than one hour for each affected area of skin on any one work day (Table I).

A second key way in which the Hanke et al.(75)

study differed from that of Dutkiewicz and Tyras(40)

was that the former study protocol included sur- rounding the benzene-containing watch glass pressed against skin, or against a glass surrogate for skin, by an enclosing funnel routed to a vapor trap that was all maintained at constant air flow to trap benzene volatilized both during and for 15 minutes after the dermal exposure period ended by (within the enclos- ing funnel) releasing the watch glass from its position pressed against exposed skin. The fraction of de- posited benzene mass absorbed was calculated as one minus the ratio of total benzene trapped under condi- tions of dermal exposure, to that trapped using glass instead of skin as the exposed surface. However, no specific data involving this calculation were reported. In particular, because separate measures of benzene volatilized during versus after exposure were not re- ported (nor apparently made), the absorption calcu- lation assumed without proof that a substantial frac- tion or even majority of the trapped benzene that was measured did not volatilize during rather than after each dermal exposure. To the extent such volatiliza- tion did occur, but would not have occurred during skin exposure absent sustained air flow within the apparatus used, benzene that might have remained available for dermal uptake was instead lost to the vapor trap during the exposure period, resulting in systematic underestimation of the dermally absorbed mass of benzene. During the extended postexposure period of negative air flow used, it is also possible that a substantial amount of benzene that had absorbed into and remained within top layers of exposed skin, which normally would tend to remain there and continue to be absorbed further into dermal tissue, instead volatilized back out of the skin surface under

Dermal vs. Total Uptake of Benzene 1355

the influence of artificially supplied air flow, and was therefore incorrectly excluded from each calculation of absorbed benzene mass. It is likely that these problems with data interpretation led the same laboratory to abandon the method used by Hanke et al.(75) and adopt the much simpler method used by Dutkiewicz and Tyras(40) to measure rates of human dermal uptake for a series of neat organic solvents.(39)

To estimate dermal permeability of benzene, Petty et al.(7) did not cite or discuss the Dutkiewicz and Tyras(40) study, and instead relied primarily on the Hanke et al.(75) data, which they showed to be reasonably consistent with measures implied by percent-uptake\data reported in in vitro studies (dis- cussed below), in which a small mass (e.g., 50 μL) of benzene (or benzene in another solvent) was applied to skin as a thin layer that in some cases was allowed to evaporate.

Estimates of steady-state permeability (Kp) for neat benzene from in vitro diffusion-cell studies us- ing human skin(45–47) were lower than the in vivo Dutkiewicz and Tyras(40) Kp,eff estimate by factors of approximately 15–88 (see Table II). Kp estimates from the Hui et al.(23) in vitro study of uptake of benzene from MSS through human skin samples, the only study specific to MSS, also were lower than the in vivo Dutkiewicz and Tyras(40) Kp,eff estimate by a factor of ∼90, for recycled and used solvent so- lutions. The Kp,eff value of apparent in vivo perme- ability for benzene identified above(40) also is also ∼10-fold or more greater than Kp values measured through human skin in vitro for benzene in several petroleum chemical mixtures(45) and for benzene in gasolines.(80) Although there are clear differences in species uptake from in vitro exposures, there are few relevant in vivo studies to support a compari- son of interspecies differences in dermal uptake of organic chemicals. A recent comparison of in vivo studies of dermal uptake of dilute aqueous organic solvents indicated that Kp,eff estimates obtained using guinea pig data were consistent with estimates ob- tained using data from human studies.(38) Recogniz- ing that some uncertainty remains as to species differ- ences in uptake in in vivo studies, the Dutkiewicz and Tyras(40) Kp,eff estimate does not differ significantly (p = 0.52, by t-test) from that of Kp,eff = 0.0082 cm/h implied by data reported by Morgan et al.(81) for ap- parent permeability of neat benzene into clipped skin on the backs of 10 rats exposed for 24 hours. Inter- estingly, the effective permeability of three different formulations of liquid white spirits into tail skin of

rats exposed for three hours (0.0075–0.0093 cm/h), calculated from data reported by Verkkala et al.,(63)

assuming a density for white spirits equal to that of MSS, also fall within the range of Kp,eff values for neat benzene exposures.

Measures of the uptake percentage (U%) of neat benzene dermally absorbed in vivo after application to human forearm (n = 4) and palm (n = 4)(82) do not differ significantly (two-tail p = 0.075, by t-test), and these combined measures likewise do not to dif- fer significantly from earlier measures reported by Franz(83) of U% (n = 4) made on human forearms in vivo (two-tail p = 0.17, by t-test). Franz(83) observed that after application of 5 μL/cm2 of neat benzene to the ventral surface of forearms of human volun- teers, “liquid benzene was undetectable on the skin surface within 30 seconds,” with ∼99.9% of that loss due to evaporation. There are no reported measures of the rate at which a thin layer of MSS volatilizes from skin, although Hui et al. (2010)(23) reported sim- ilarly rapid evaporation of benzene from an open-air glass vial of MSS (>99.5% in 30 minutes at 25 ◦C; see Section 2.3.2.5). Related data involving in vivo exposures to skin of primates and hairless mice to benzene and/or to benzene in rubber solvent(64,83,84)

generally provide similar or slightly greater values of U%. In this regard, it is noteworthy that sets of in vivo U%-measures made using neat benzene applied to monkey back (n = 6)(83) and to monkey forearm (n = 3)(64), and additional U-measures made using dilute benzene in rubber solvent (n = 3),(45) are all statistically homogeneous (p = 0.47, by analysis of variance), and these pooled monkey data (n = 12) do not differ significantly from the pooled human data (n = 12) mentioned above (p = 0.068, by t-test). The key drawback to U% data summarized here is that available in vivo human data do not address di- lute benzene in far less volatile MSS, in which ben- zene is miscible, so direct application of the available U% data to this specific type of exposure remains speculative.

DECLARATION OF INTEREST AND ACKNOWLEDGMENTS

Some of the original research on parts washer exposure was funded by Safety-Kleen Corporation, which is involved in litigation regarding solvent ex- posures. Preparation of this article was funded en- tirely by compensated time from Exponent, Inc., with no direct or indirect funding by any other party. One of the authors (Patrick Sheehan) served during

1356 Bogen and Sheehan

2007–2009 as an expert witness for Safety-Kleen in cases involving benzene exposure from use of parts washing solvents; neither author is currently engaged in any expert work concerning such cases. We thank Rick Nelson at Exponent for editorial assistance, and Safety-Kleen Corporation for providing data on the benzene content in Safety-Kleen 105 solvent.

REFERENCES

1. International Agency for Research on Cancer (IARC). IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans. Vol. 29: Some Industrial Chemical and Dyestuffs. Benzene, Lyon, France: World Health Organiza- tion, IARC, 1982.

2. International Agency for Research on Cancer (IARC). IARC Monographs on the Evaluation of Carcinogenic Risks to Hu- mans. Overall Evaluations of Carcinogenicity: An Updating of IARC Monographs Vols. 1 to 42. Suppl 7. Lyons, France: World Health Organization, IARC, 1987.

3. U.S. Environmental Protection Agency (EPA). Carcinogenic Effects of Benzene: An Update. Washington, DC: EPA/600/P- 97/001F, 1998.

4. International Agency for Research on Cancer (IARC). IARC Monographs on the Evaluation of Carcinogenic Risks to Hu- mans. Some Organic Solvents, Resin Monomers and Related Compounds, Pigments and Occupational Exposures in Paint Manufacture and Painting. Vol. 47. Lyons, France: World Health Organization, IARC, 1988.

5. Boeniger MF, Lushniak BD. Exposure and absorption of haz- ardous materials through the skin. International Journal of Occupational and Environmental Health, 2000; 6(2):148–150.

6. McDougal JN, Boeniger MF. Methods for assessing the risk of dermal exposure in the workplace. Critical Reviews in Tox- icology, 2002; 32:291–327.

7. Petty SE, Nicas M, Boiaeski, AA. A quantitative method for estimating dermal benzene absorption from benzene- containing hydrocarbon liquids. International Journal of Oc- cupational and Environmental Health, 2011; 17:287–300.

8. Fedoruk MJ, Bronstein R, Kerger KD. Benzene exposure as- sessment for use of a mineral spirits-based degreaser. Applied Occupational and Environmental Hygiene, 2003; 18:764–771.

9. Plisko MJ, Spencer JW. Evaluation of a mathematical model for estimating solvent exposures in the workplace. Journal of Chemical Health and Safety, 2008; 15(2):14–21.

10. Sheehan P, Bogen KT, Hicks J, Goswami E, Brorby G, Lau EC, Ott B. Benzene inhalation by parts washers: New esti- mates based on measures of occupational exposure to solvent coaromatics. Risk Analysis, 2010; 30(8):1249–1267.

11. Bogen KT, Spear RC. Integrating uncertainty and inter- individual variability in environmental risk assessment. Risk Analysis, 1987; 7:427–436.

12. National Research Council (NRC). Science and Judgment in Risk Assessment. Washington, DC: National Academies Press, 1994.

13. Bogen KT. Methods to approximate joint uncertainty and variability in risk. Risk Analysis, 1995; 15(3):411–419.

14. Song CW, Chelstrom LM, Levitt SH, Haumschild DJ. Effects of temperature on blood circulation measured with the laser Doppler method. International Journal of Radiation Oncol- ogy, Biology, Physics, 1989; 17:1041–1047.

15. Royston P. Approximating the Shapiro-Wilk W-test for non- normality. Statistics and Computing, 1992; 2:117–119.

16. Wright SP. Adjusted p-values for simultaneous inference. Bio- metrics, 1992; 48:1005–1013.

17. Snedecor GW, Cochran WG. Statistical Methods, 8th ed. Ames, IA: Iowa State University Press, 1989.

18. Wolfram Research. Wolfram Mathematica R© Documentation Center. Champaign, IL: Wolfham Research, Inc., 2013.

19. Kim D, Andersen ME, Nylander-French LA. Dermal absorp- tion and penetration of jet fuel components in humans. Toxi- col Lett, 2006b; 165(1):11–21.

20. Kendall M, Stuart A. The Advanced Theory of Statistics, Vol. 2: Inference and Relationship, 4th ed. New York: MacMillan Publishing Co., 1979.

21. Lehman EL, D’Abrera HJM. Nonparametrics: Statistical Methods Based on Ranks. San Francisco, CA: Holden-Day Inc., 1975.

22. Wilcox RR. Some practical reasons for reconsidering the Kolmolgorov-Smirnov test. British Journal of Mathematical and Statistical Psychology, 1997; 50:9–20.

23. Hui X, Wester RC, Barbadillo S, Cashmore, A, Maibach HI. In vitro percutaneous absorption of benzene in human skin. Cutaneous and Ocular Toxicology, 2009; 28(2):65–70.

24. Reddy MB, Guy RH, Bunge AL. Does epidermal turnover reduce percutaneous penetration? Pharmaceutical Research, 2000; 17:1414–1419.

25. Wester RC, Maibach HI, Gruenke L, Craig J. Bioavailability of benzene through inhalation and skin absorption. Pp. 339– 351 in Winegar ED, Keith LH (eds). Sampling and Analysis of Airborne Pollutants. Boca Raton, FL: Lewis Publishers, 1993.

26. Cherrie JW, Robertson A. Biologically relevant assessment of dermal exposure. Annals of Occupational Hygiene, 1995; 39(3):387–392.

27. U.S. Environmental Protection Agency (EPA). Exposure Fac- tors Handbook, 11th ed. EPA/600/R-090/052F, September 2011. Washington, DC: U.S. EPA Office of Research and De- velopment, 2011.

28. Wassenius O, Järvholm B, Engström T, Lillienberg L, Med- ing B. Variability in the skin exposure of machine operators exposed to cutting fluids. Scandinavian Journal of Work, En- vironment & Health, 1998; 24(2):125–129.

29. Cherrie JW, Apsley A. Semple S. A new sampler to assess dermal exposure during wet working. Annals of Occupational Hygiene, 2007; 51(1):13–18.

30. Cherrie JW, Brouwer DH, Roff M, Vermeulen R, Kromhaut H. Use of qualitative and quantitative fluorescence techniques to assess dermal exposure. Annals of Occupational Hygiene, 2000; 44(7):519–522

31. Soutar A, Semple S, Aitken RJ, Robertson A. Use of patches and whole body sampling for the assessment of dermal expo- sure. Annals of Occupational Hygiene, 2000; 44(7):511–518.

32. Hughson GW, Aitken RJ. Determination of dermal exposures during mixing, spraying and wiping activities. Annals of Occu- pational Hygiene, 2004; 48(3):245–255.

33. Kromhout H, Fransman W, Vermeulen R, Roff M, van Hem- men JJ. Variability of task-based dermal exposure measure- ments from a variety of workplaces. Annals of Occupational Hygiene, 2004; 48(3):187–196.

34. Rajan-Sithamparanadarajah R, Roff M, Delgado P, Eriksson K, Fransman W, Gijsbers JH, Hughson G, Mäkinen M, van Hemmen JJ. Patterns of dermal exposure to hazardous sub- stances in European Union workplaces. Annals of Occupa- tional Hygiene, 2004; 48(3):285–297.

35. van Ravenzwaay B, Leibold E. The significance of in vitro rat skin absorption studies to human risk assessment. Toxicology in Vitro, 2004; 18:219–225.

36. Marquart H, Warren ND, Laitinen J, van Hemmen JJ. De- fault values for assessment of potential dermal exposure of the hands to industrial chemicals in the scope of regulatory risk as- sessments. Annals of Occupational Hygiene, 2006; 50(5):469– 489.

37. Bogen KT. Models based on steady state in vitro dermal per- meability data underestimate short-term in vivo exposures to

Dermal vs. Total Uptake of Benzene 1357

organic chemicals in water. Journal of Exposure Analysis and Environmental Epidemiology, 1994; 4(4):457–475.

38. Bogen KT. Dermal uptake of 18 dilute aqueous chemicals: In vivo disappearance-method measures greatly exceed in vitro- based predictions. Risk Analysis, 2013; 33(7):1334–1352.

39. Baranowska-Dutkiewicz B. Skin absorption of aniline from aqueous solutions in man. Toxicology Letters, 1982; 10:367– 372.

40. Dutkiewicz T, Tyras H. Wchlanianie benzenu przez skòre u ludzi [Absorption of benzene through skin in man ]. Zeszyty Naukowe Bromatologii i Chemie Toksykologicznej, 1969; 2(1):3–9 (in Polish, with English abstract).

41. Johanson G, Boman A, Dynésius B. Percutaneous absorption of 2-butoxyethanol in man. Scandanavian Journal of Work, Environment & Health, 1988; 14(2):101–109.

42. Jakasa I, Mohammadi N, Krüse J, Kezic S. Percutaneous ab- sorption of neat and aqueous solutions of 2-butoxyethanol in volunteers. International Archives of Occupational and Envi- ronmental Health, 2004; 77(2):79–84.

43. Dutkiewicz T, Tyras H. Skin absorption of toluene, styrene and xylene by man. British Journal of Industrial Medicine, 1968; 25:243.

44. Korinth G, Lüersen L, Schaller KH, Angerer J, Drexler H. Enhancement of percutaneous penetration of aniline and o- toluidine in vitro using skin barrier creams. Toxicology in Vitro, 2008; 22(3):812–818.

45. Blank IH, McAuliffe DJ. Penetration of benzene through hu- man skin. Journal of Investigative Dermatology, 1985; 85:522– 526.

46. Franz TJ. Percutaneous absorption of benzene. PP. 61–70 in Mcfarland HN (ed). Advances in Modern Environmental Toxicology, Volume 6. Princeton, NJ: Scientific Publishers, 1984.

47. Loden M. The in vitro permeability of human skin to benzene, ethylene glycol, formaldehyde and n-hexane. Acta Pharma- col et Toxicol, 1986; 58:382–389.

48. Lockley DJ, Howes D, Williams FM. Percutaneous penetra- tion and metabolism of 2-butoxyethanol. Archives of Toxicol- ogy, 2004; 78:617–628.

49. Korinth G, Schaller KH, Drexler H. Is the permeability coef- ficient Kp a reliable tool in percutaneous absorption studies? Archives of Toxicology, 2005; 79:155–159.

50. DuPont. 2005. Carbon Disulfide: In Vitro Dermal Absorption Rate Testing. Sponsored by the Carbon Disulfide Panel of the American Chemistry Council. DuPont – 16061. U.S. Environ- mental Protection Agency (EPA) Docket Document ID No. EPA-HQ-2003–0006–0289. (Author: Fasano Sr WJ.) Haskell Laboratory for Health and Environmental Sciences, Newark, DE. In Letter to OPPT Document Control Office, USEPA, from Susan Anderson Lewis, Managing Director, CHEM- STAR, American Chemistry Council re: In Vitro Dermal Absorption Rate Testing of Carbon Disulfide (CAS Number 75-15-0) Under TSCA Section 4 Test Rule. Posted August 25, 2005, U.S. Environmental Protection Agency (EPA) Docket Item No. EPA-HQ-OPPT-2003-0006. Available at: http://www.regulations.gov/#!documentDetail;D=EPA-HQ- OPPT-2003-0006-0289, Accessed December 22, 2013.

51. Ursin C, Hansen CM, Van Dyk JW, Jensen PO, Christensen IJ, Ebbehoej J. Permeability of commercial solvents through living human skin. American Industrial Hygiene Association Journal, 1995; 56:651–660.

52. Kim D, Andersen ME, Nylander-French LA. A dermatotoxi- cokinetic model of human exposures to jet fuel. Toxicological Sciences, 2006; 93(1):22–33.

53. CHERIC. 2013. Pure Component Properties (Queriable database). Chemical Engineering Research Information Cen- ter. Available at: http://www.cheric.org/research/kdb/hcprop/ showcoef.php?cmpid=651&prop=PVP, Accessed June 12, 2013.

54. U.S. Environmental Protection Agency (EPA). Dermal Ex- posure Assessment: Principles and Applications. EPA/600/8– 91/011B, May 1992. Washington, DC: U.S. EPA Risk Assess- ment Form, 1992.

55. U.S. Environmental Protection Agency (EPA). Risk Assess- ment Guidance for Superfund Volume I: Human Health Eval- uation Manual (Part E, Supplemental Guidance for Dermal Risk Assessment). EPA/540/R/99/005, July 2004. Washington, DC: U.S. EPA Office of Superfund Remediation and Tech- nology Innovation, 2004.

56. Babu RJ, Chatterjee A, Ahaghotu E, Singh M. Percutaneous absorption and skin irritation upon low-level prolonged der- mal exposure to nonane, dodecane and tetradecane in hairless rats. Toxicology and Industrial Health, 2004; 20:109–118.

57. Kezic S, Nielsen JB. Absorption of chemicals through compro- mised skin. International Archives of Occupational and Envi- ronmental Health, 2009; 82(6):677–688.

58. Babu RJ, Chatterjee A, Singh M. Assessment of skin irritation and molecular responses in rat skin exposed to nonane, dode- cane and tetradecane. Toxicology Letters, 2004; 153:255–266.

59. Iman RL, Connover WJ. A distribution-free approach to in- ducing rank correlation among input variables. Communica- tions in Statistics-Simulation, 1982; 11(3):311–334.

60. Jennrich RI. An asymptotic chi-2 test for the equality of two correlation matrices. Journal of the American Statistical As- sociation, 1970; 65(330):904–912.

61. Wolfram Research. 2013. Wolfram Mathematica 9 Docu- mentation Center. Champaign, IL: Wolfram Research, Inc., (www.wolfram.com), 2013. Available at: http://reference. wolfram.com/mathematica/guide/Mathematica.html, Ac- cessed January 5, 2013.

62. Waidyanatha S, Rothman N, Fusioni S, Smith M, Hayes R, Bechtold W, Dosemeci M, Guilan L, Yin S, Rappaport S. Uri- nary benzene as a biomarker of exposure among occupation- ally exposed and unexposed subjects. Carcinogenesis, 2001; 22(2):279–286.

63. Verkkala E, Pfläffi P, Savolainen H. Comparison of local neu- rotoxicity of three white spirit formulations by percutaneous exposure of rat tail nerve. Toxicology Letters, 1984; 21:293– 299.

64. Maibach HI, Anjo DM. Percutaneous penetration of ben- zene and benzene contained in solvents used in the rubber industry. Archives of Environmental Health, 1981; 36(5):256– 260.

65. Nielsen JB. Percutaneous penetration through slightly dam- aged skin. Archives of Dermatological Research, 2005; 296:560–567.

66. Gould AR, Sharp PJ, Smith DR, Stegink AJ, Chase CJ, Ko- vacs JC, Penglis S, Chatterton BE, Bunn CL. Increased per- meability of psoriatic skin to the protein, plasminogen acti- vator inhibitor 2. Archives of Dermatological Research, 2003; 295:249–254.

67. Boman A, Wahlberg JE. Percutaneous absorption of 3 or- ganic solvents in the guinea pig: (1) Effect of physical and chemical injuries to the skin. Contact Dermatitis, 1989; 21: 36–45.

68. Hino R, Nishio D, Kabashima K, Tokura Y. Percutaneous penetration via hand eczema is the major accelerating factor for systemic absorption of toluene and xylene during car spray painting. Contact Dermatitis, 2008; 58:76–79.

69. Jia X, Xiao P, Jin X, Shen G, Wang X, Jin T, Nordberg G. Adverse effects of gasoline on the skin of exposed workers. Contact Dermatitis, 2002; 46:44–47.

70. Hostynek JJ, Lamel SA, Maibach HI. Benzene absorption in animals and man: An overview. Reviews on Environmental Health, 2012; 27(2–3):85–101.

71. Laitinen J, Kangas J, Pekari K, Liesivuori J. Short time expo- sure to benzene and gasoline at garages. Chemosphere, 1994; 28(1): 197–205.

1358 Bogen and Sheehan

72. Wallace L. Environmental exposure to benzene: An update. Environmental Health Perspectives, 1996; 104(Suppl 6):1129– 1136.

73. Occupational Safety and Health Administration (OSHA). Occupational safety and health standards, toxic and hazardous substances, benzene; 29 CFR §1910.1028(c)(1), 2010. Available at: http://www.osha.gov/pls/oshaweb/owadisp. show document?p id=10042&p table=STANDARDS, Ac- cessed May 29, 2010.

74. American Conference of Governmental Industrial Hygienists (ACGIH). Threshold limit values for chemical substances and physical agents & biological exposure indices. Cincinnati, OH: ACGIH, 2012.

75. Hanke J, Dutkiewicz T, Piotrowski J. The absorption of benzene through human skin. Reprinted in translation from Medycyna Pracy, 1961; 12(5):413–426. Int. J Occup Environ Health. 2000; 6(2):104–111.

76. Dutkiewicz T, Tyrus H. A study of the skin absorption of ethylbenzene in man. British Journal of Industrial Medicine, 1967; 24:330–332.

77. Bogen KT, Colston BW, Machicao LK. Dermal ab- sorption of dilute aqueous chloroform, trichloroethy- lene and tetrachloroethylene in hairless guinea pigs. Fundamental and Applied Toxicology, 1992; 18: 30–39.

78. Kronevi T, Wahlberg JE, Holmberg B. Skin pathology fol- lowing epicutaneous exposure to seven organic solvents.

International Journal of Tissue Reactions, 1982; II(1): 21–30.

79. Brener Z. Recent developments in the field of Chagas’ disease. Bulletin of the World Health Organization, 1982; 60(4):463– 473.

80. Adami G, Larse F, Venier M, Barbieri P, Lo Coco F, Reisen- hofer E. Penetration of benzene, toluene and xylenes con- tained in gasoline through human abdominal skin in vitro. Toxicology in Vitro, 2006; 20:1321–1330.

81. Morgan DL, Cooper SW, Carlock DL, Sykora JJ, Sutton B, Mattie DR, McDougal JN. Dermal absorption of neat and aqueous volatile organic chemicals in the Fischer 344 rat. En- vironmental Research, 1991; 55:51–63.

82. Modjtahedi BS, Maibach HI. In vivo percutaneous absorption of benzene in man: Forearm and palm. Food and Chemical Toxicology, 2008; 46(3):1171–1174.

83. Franz TJ. Percutaneous absorption of benzene. Pp. 108–114 in MacFarland HN, Holdsworth CE, MacGregor JA, Call RW, Kane ML (eds). Applied Toxicology of Petroleum Hydrocar- bons. Proceedings of a symposium sponsored by the Ameri- can Petroleum Institute and the Society of Toxicology, held May 1982, Washington, DC. Princeton, NJ: Princeton Scien- tific Publishing.

84. Susten AS, Dames BL, Burg JR, Niemeier RW. Percutaneous penetration of benzene in hairless mice: An estimate of dermal absorption during tire-building operations. American Journal of Industrial Medicine, 1985; 7:323–335.

Copyright of Risk Analysis: An International Journal is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.