This study was conducted at six USAF bases, as a part of a broader project (Institute of Environmental and Human Health 2001), to investigate the relative contributions of JP-8 exposures through different routes (i.e., inhalation and dermal) to the total body dose of fuel-cell maintenance workers. Workers were recruited with informed consent from active-duty USAF personnel who routinely worked with, or were exposed to, JP-8. Although 339 USAF personnel were enrolled in the overall project, a total of 85 fuel-cell maintenance workers were included in our particular study. Approval for human subject use was obtained from the institutional review board for each of the participating investigators and for the USAF, and the study complied with all applicable U.S. requirements and regulations.

Questionnaires were collected after the work shift to obtain information on demographic factors, including job tasks, the use of personal protective equipment, smoking status, and other work-related characteristics. Work diaries for each individual were also recorded during the survey, including detailed information on work tasks, task durations, and the use of personal protective equipment. Workers included in this study entered fuel cells during the day sampling was conducted. They performed maintenance work inside the fuel cell and thus were expected to have the highest potential for both dermal and inhalation exposure to JP-8. To reduce inhalation exposure, all workers wore air-supplied respirators when working inside the fuel cell. Although USAF personnel had been assigned a priori into exposure groups (Egeghy et al. 2003; Serdar et al. 2004), we note that these categories did not fully correspond with the work scenarios given in the work diaries on the day of investigation. Therefore, in the present investigation we relied upon the work diaries to define job tasks and work scenarios.

All exposure data (dermal, breathing-zone air, breath, and urine) were natural log-transformed to help satisfy assumptions regarding normality and homogeneous variance. Paired analyses were performed to investigate the differences between pre- and postexposure measurements of breath naphthalene and urinary 1- and 2-naphthol levels, as well as between postexposure urinary 1- and 2-naphthol levels. Multiple linear regression analysis (Proc REG procedure in SAS, version 8.2; SAS Institute, Cary, NC) was used to investigate the contributions of JP-8 exposure (dermal and inhalation), smoking, and other covariates obtained from questionnaires to urinary 1- and 2-naphthol concentrations. Stepwise variable selection was used to determine final regression models, with inclusion and elimination decisions about predictors conducted at the α = 0.10 level. Possible collinearity problems were investigated using eigenvalue analyses and variance inflation factors. Possible outliers were examined using studentized residuals. All statistical analyses were performed using SAS software.

The multiple linear regression model structure adopted was of the general form:

Here, the outcome variable ln(urinary naphthol_i) [ln(ng/L)] is the natural logarithm of either the ith worker’s urinary 1-naphthol or urinary 2-naphthol level; X_ij represents the ith worker’s jth exposure level to JP-8 (dermal, breathing-zone, or breath naphthalene measurement); and C_ik represents the kth covariate value for the ith worker based on questionnaires providing information on smoking status (38 smokers, 47 nonsmokers), race (74 white, 11 nonwhite), sex (81 males, 4 females), job tasks [handle foams (n = 73), hold ventilation (n = 58), remove bolts (n = 55), remove foams (n = 76), remove tank door (n = 63)], and so forth.

The predictor variable effects consisted of α, the intercept; β_j, the regression coefficient for the natural logarithm of JP-8 exposure {e.g., the natural logarithm of dermal naphthalene [ln(ng/m²)], breathing-zone or breath naphthalene [ln(ng/m³)]}; and γ_k , the regression coefficient for covariate k. Two models were fitted using different inhalation markers. We used the breathing-zone naphthalene level in model 1 and the end-exhaled breath naphthalene level in model 2 as inhalation markers.

The relative contributions of predictor variables in the final regression models were determined by the proportionate contribution that each predictor made to the regression model multiple R² using the Pratt index (Pratt 1987). For each predictor in a final regression model, the Pratt index for that predictor is the product of its estimated standardized regression coefficient and the simple correlation between that predictor and the outcome variable. One particularly nice property of the Pratt index is that the sum of the Pratt indices for all predictors equals R². The Pratt index for each predictor can be rescaled by dividing it by the model R² and multiplying by 100, so that the resulting number can be interpreted as the percentage of the model R² accounted for by that predictor.

The measured dermal, breathing-zone, and breath naphthalene levels, as well as urinary 1- and 2-naphthol levels, for the 85 USAF fuel-cell maintenance workers are described in Table 1. The geometric mean (GM) [geometric standard deviation (GSD)] of dermal naphthalene level were 4,180 (9.35 ng/m²) with a range of 100 ng/m² to 5,090 μg/m². The GM (GSD) of breathing-zone naphthalene were 614,000 and (2.12 ng/m³) with a range of 670 ng/m³ to 3,910 μg/m³. The postexposure levels of breath naphthalene and urinary 1- and 2-naphthol were significantly higher than the preexposure levels (all p-values < 0.0001). In addition, postexposure urinary 2-naphthol levels were greater than postexposure urinary 1-naphthol levels (p < 0.0001).

Model 1 explained 26.6% and 26.3% of total variance in the urinary 1- and 2-naphthol levels in entrants, respectively (Table 2). In the model for urinary 1-naphthol, breathing-zone naphthalene and smoking were the only significant predictors, explaining 88.2% and 11.8% of total variance, respectively, using the Pratt index of relative importance (Pratt 1987). For urinary 2-naphthol, dermal and breathing-zone naphthalene and smoking were significant, explaining 51.1%, 35.8%, and 13.1% of total variance, respectively. These results indicate that dermal exposure to naphthalene contributed significantly to urinary 2-naphthol levels but not to urinary 1-naphthol levels among the fuel-cell maintenance workers.

Because the fuel-cell maintenance workers wore respiratory protection when entering the fuel cell, breathing-zone naphthalene could be regarded as an unreliable measure of personal inhalation exposure, because it most likely represents an overestimation of the personal inhalation exposure under these conditions. Thus, end-exhaled breath naphthalene measured immediately after the end of work was investigated as a potential inhalation marker in model 2. This model explained 31.8% and 30.9% of total variance in urinary 1- and 2-naphthol levels, respectively (Table 2). For urinary 1-naphthol, breath naphthalene and smoking were the only significant predictors, explaining 87.2% and 12.8% of total variance, respectively. For urinary 2-naphthol, dermal and breath naphthalene and smoking were significant predictors, explaining 32.3%, 52.9%, and 14.8% of total variance, respectively. These results also suggest that dermal exposure to naphthalene contributes significantly to urinary 2-naphthol levels but not to urinary 1-naphthol levels. Although the relative contribution of dermal naphthalene to urinary 2-naphthol levels decreased in model 2 relative to model 1, this may be attributed to the fact that breath naphthalene may actually reflect both dermal and inhalation exposure to JP-8.