A detailed description of the analytical method can be found in Panuwet et al. (in press). Briefly, we introduced 200 μL urine onto a dual online solid-phase extraction (reversed-phase phenyl–hexyl and strong cation exchange) system, which was operated by a novel switching mechanism. Analytes were preconcentrated using high-performance liquid chromatography on a guard column, and the remaining urine components were washed off, using 10% methanol in 0.1% formic acid, into a waste container. The valves were switched, and the analytes were backwashed onto a reversed-phase or strong anion exchange analytic column and separated using a linear gradient beginning with 10% methanol in 0.1% formic acid and ending with 100% methanol. Samples were analyzed using atmospheric pressure chemical ionization–tandem mass spectrometry with one precursor–product ion pair being used for quantification and two precursor–product ion pairs being used for confirmation (Table 2). Quantification was achieved using isotope dilution calibration, for which isotopically labeled standards were available. When they were not available, the most closely eluting labeled standard was used for quantification. Each analytical run consisted of a full eight-standard calibration set, three positive (i.e., fortified urine samples at three levels spanning calibration range) and two negative (i.e., blank) control samples, and up to 75 unknown samples. The limits of detection ranged from 0.1 to 1 ng/mL, with relative standard deviations typically < 12% over the calibration range.

Human samples were collected from residential turf applicators (i.e., commercial lawn care applicators), nonapplicators with low-level exposures (i.e., nonoccupationally exposed individuals with documented ATZ exposure based upon detectable levels of AM in their urine) and volunteers in Georgia with no known acute exposure to ATZ. All samples were collected as part of previous studies and were reanalyzed to determine total metabolites. All protocols were reviewed and approved by the CDC Institutional Review Board for ethical treatment of human research subjects. Samples were stored at −70°C until used and before analysis were thawed at room temperature and mixed thoroughly.

Simple statistics (mean, standard deviation, and ratios) were performed using Excel software (Microsoft, San Jose, CA). Because the number of samples tested was small, no significance testing was performed.

A mass chromatogram of a spiked urine sample is shown in Figure 2. Because the most polar analytes were not adequately retained on the reversed-phase column, a strong cation exchange column was used. DACT was partially retained on the reversed-phase column, and the remainder was retained on the strong cation exchange column resulting in two peaks for DACT. Both peaks were summed to calculate the total amount of DACT present. All peaks were resolved by either time or mass-to-charge ratio.

The metabolite profile of the higher exposure category of turf applicators is shown in Figure 3 (n = 8). The two graphs represent the exposure assessment using only AM and the exposure assessment using all metabolites. DACT was the most predominantly detected metabolite (mean = 51%), followed closely by DEA (mean = 31%). AM was detected on average in only 12% of the samples tested. The interperson variation (calculated as the relative standard deviation of the percentage of each metabolite percentage among persons) in urinary concentrations among these most detected analytes was between 33 and 51%. The interperson variability was much greater for the less frequently detected metabolites.

The metabolite profiles for the lower-level exposure category (n = 5) are shown in Figure 4. On average, DEA (33%) and DACT (28%) were detected in about equal proportions, with only 6% detection of AM. Similar to the higher-level exposures, the interperson variation in their urinary concentrations was large.

For the environmental exposure category, DACT was by far the most predominantly detected metabolite (77%). DEA was detected the next most frequently (15%) and AM was detected in only 2% of the samples. Again, the interperson variability in metabolic profile concentrations was large.

The small amount of data that we present here clearly demonstrate that exposure to ATZ-related chemicals can be misrepresented by measurement of AM alone. However, it is important to note that the measurement of ATZ or AM in urine would be the only unequivocal indication that a person was exposed to ATZ and not an environmental degradate.

Also, the metabolite profiles differ dramatically based upon the exposure scenario (Figure 5). Occupational or lower-level acute exposures, perhaps after ATZ use on lawns, are probably more likely to be direct exposures to ATZ and lesser exposures to the degradation products. However, the environmental exposure scenario is quite different. In environmental exposures, persons likely are exposed through food or water, which would mean that dealkylation and hydrolysis products might make up a larger percentage of the exposure. Of course, exposure to the dealkylation products is still important because these chemicals remain biologically active. Thus, the presence of the chlorinated dealkylation products or their glutathione-mediated mercapturic acid metabolites would indicate exposure to a biologically active component. Presence of the hydroxylated metabolites may indicate exposure to the hydroxylated products themselves or to their chlorinated counterparts.

Our data demonstrate that we likely will need to measure most or all metabolites of ATZ to accurately assess ATZ-related exposures. However, further evaluation is necessary because of our small sample size and because the possible presence of glucuronide metabolites was not considered. In the future, we will explore further the role of hydroxyl metabolites in the metabolite profiles by evaluating glucuronide-hydrolyzed urine. Also, we need to include the mercapturates of the dealkylation products in our methodology to glean the full picture of ATZ metabolism. Futhermore, we will use this method to evaluate larger populations with high-level occupational exposures (e.g., manufacturers, farmers) and background exposures in the general U.S. population.

Although we found detectable concentrations of ATZ metabolites in most of the urine samples tested, we are uncertain what, if any, health effects result from these levels of exposure. In general, animal dosing studies that have investigated health effects (Ashby et al. 2002; Cooper et al. 1996, 2000) have used doses much larger than those to which we could assume (from back calculation) participants were exposed in our study. Further studies evaluating health outcomes at typical human exposure levels are warranted.

We have clearly been underestimating exposure to ATZ-related metabolites in the U.S. population and in other selected studies. Our newer data are more in line with exposures we might expect to see based upon ATZ use and environmental persistence. It is likely that multiple metabolites must be measured to accurately classify exposure categories. Although DACT appeared to be the predominant metabolite detected in each exposure category, the interperson variations in its concentrations were consistently about 30%. Regardless, DACT and DEA appear to be the most important metabolites to measure to evaluate exposures to ATZ-related chemicals.

Clearly, exposure to ATZ or its degradates appears more pervasive than previously believed; however, more data are needed to confirm this observation. Where measures exist to mitigate or lessen exposures to biologically active ATZ degradates such as the use of high efficiency filters in municipal water systems or other mitigation strategies for pond and surface waters (Acosta et al. 2004; Agdi et al. 2000), they should be used to ensure the best protection of public health.

The percent contribution for “Low-level acute exposures” in Figure 5 was changed from that in the original manuscript published online.