Estimating pesticide dose from urinary pesticide concentration data by creatinine correction in the Third National Health and Nutrition Examination Survey (NHANES-III)

Abstract

The Third National Health and Nutrition Examination Survey (NHANES-III) of the Centers for Disease Control and Prevention (CDC) recorded data on the urinary concentrations of 12 chemicals (analytes), which were either pesticides or their metabolites, that represent exposure to certain pesticides, in urine samples collected from 1988 to 1994 from a cohort of 978 volunteer subjects, aged 20–59 years. We have used each subject's urinary creatinine concentration and their individual daily creatinine excretion rate (g/day) computed from their age, gender, height and weight, to estimate their daily excretion rate in μg analyte/kg/day. We discuss the mechanisms of excretion of the analytes and certain assumptions needed to compute the equivalent daily dietary intake (μg/kg/day) of the most likely parent pesticide compounds for each excreted analyte. We used literature data on the average amount of parent compound ingested per unit amount of the analyte excreted in the urine, and compared these estimated daily intakes to the US EPA's reference dose (RfD) values for each of those parent pesticides. A Johnson S_B distribution (four-parameter lognormal) was fit to these data to estimate the national distribution of exclusive exposures to these 12 parent compounds. Only three such pesticides had a few predicted values above their RfD (lindane 1.6%; 2,4-dichlorophenol 1.3%; chlorpyrifos 0.02%). Given the possibility of a subject's dietary intake of a pesticide's metabolites incorporated into treated food, our results show that few, if any, individuals in the general US population aged 20–59 years and not employed in pesticide application were likely to have exceeded the USEPA RfD for these parent compounds during the years studied.

Introduction

The National Health and Nutrition Examination Study (NHANES) is an ongoing program administered by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC). In each phase, it represents a stratified random sample of the entire US population, with certain groups oversampled in order to increase the precision of estimates for that group. All persons chosen were assigned a sampling weight related to their inverse probability of selection. All subjects chosen who agreed to participate completed a questionnaire, and were invited to undergo a complete physical examination. In the Third NHANES (NHANES-III), conducted from 1988 to 1994, a subset of the examined subjects, aged 20–59 years, volunteered to provide a urine sample for analysis of selected pesticide residues or their metabolites as an indication of exposure to these pesticides (US Department of Health and Human Services (USDHHS), 1999). This subset was a nonrepresentative convenience sampling of the survey participants; however, the demographic representation of the subset was diverse.

The usage of a creatinine correction for urinary concentration data has long been controversial in the pesticide exposure literature. This paper reviews the data generated from these analyses performed by the National Center for Environmental Health (NCEH) at CDC, and suggests an innovative means by which these urine data can be creatinine corrected, interpreted and compared to reference doses established by the US Environmental Protection Agency (USEPA) to better characterize intake doses of concern for human health. We also discuss the factors that can cause these exposure-related values to be either an overestimation or underestimation of actual dose received.

Method of creatinine correction

As a part of the NHANES-III study, CDC collected and analyzed a spot sample of urine from a group of 978 adult volunteer subjects for 12 analytes that represent exposure to pesticides. A “spot-sample” is a void without a record of the volume (ΔV) or the elapsed time in hours from the previous void (Δt). Consequently, these pesticide analyses do not allow extrapolation of the total mass of analyte measured (Δm) to a daily excretion rate as either 24Δm/Δt or V Δm/ΔV, where V is the expected daily urine volume excreted by the subject with known age, gender and body mass, at that time of the year. However, a correction using the creatinine (Cn) concentration measured in the same samples can be used for this estimation, as described in detail in a following section:

The CDC reported concentration data for 12 pesticide-related analytes in units of micrograms of analyte per liter of urine (μg/l) (Hill et al., 1995). Table 1 lists the analytes (pesticide residues and metabolites) reported, the possible parent pesticides of the metabolites, and their respective reference doses (RfD). The RfD is an average daily intake value over a lifetime thought to be acceptable, based upon currently available animal toxicity studies. For some pesticides, the analyte is the pesticide itself; for example, 2,4-dichlorophenoxyacetic acid (2,4-D). However, in most cases, the analyte is a metabolite of the parent pesticide, generally the primary metabolite.

Table 1 NHANES-III analytes, possible parent compounds and reference doses

Whereas detection of a pesticide in a urine sample above the minimum detectable level (MDL) means that the subject was exposed to that pesticide, a detectable urinary concentration of a pesticide metabolite only indicates that the subject was exposed to one or more of its parent pesticides, to the metabolite itself, or to a pharmaceutical or other chemical with the same metabolite. The numerical value of the analyte concentration in μg/l is determined by several unknown quantities which are discussed in detail below: (1) the amounts of the parent pesticide(s) and their metabolites to which the subject has been exposed; (2) the hydration status or fluid intake of the subjects as evidenced in the dilution of the urine; and (3) body chemistry.

Pesticide Intake

The presence of a detectable concentration of the analyte in the urine of an NHANES-III participant only means that the subject had been exposed through some unknown pathway to the analyte, its parent pesticide, or another chemical capable of producing the same metabolite. Furthermore, if any given metabolite is also an environmental degradation product, one cannot determine whether the exposure, in whole or in part, was to the parent pesticide itself or to the environmental degradate. Presence of the degradation product in food may arise from reaction of the parent pesticide due to thermal action during cooking or environmental factors in the field where it was applied (e.g., by plant or bacterial metabolism or through the action of moisture and sunlight). For the organophosphate and carbamate pesticides, the degradation product and the human metabolite may be the same chemical. For the purpose of this interpretation, we make the following assumptions:

(1) that the metabolite in the urine was derived solely from exposure to the parent pesticide because the RfDs are associated only with the parent pesticide, not the metabolite;

(2) that 100% of the pesticide dose was converted in vivo to the metabolite (e.g., 100% of the chlorpyrifos dose was converted to 3,5,6-trichloro-2-pyridinol) as no human experimental data showing otherwise are available at this time; and

(3) that 100% of the metabolite was excreted in urine unless experimental data show otherwise, as is only the case for 3,5,6-trichloro-2-pyridinol (Nolan et al., 1984).

A spot sample gives no information as to the previous time course of exposure, such as whether the exposure was a single occurrence at some previous point in time or a variable daily exposure. This is because each corresponding residue has a low-concentration first-order half-life in the blood after absorption that differs among individuals, depending in part on the state of health of the kidney and liver enzyme functions. For example, the subject could have been recently exposed to a quantity sufficient to rapidly attain the measured urinary value, or exposed to twice as much one half-life previously, or four times as much two half-lives previously. Alternatively, the subject could have been exposed continuously to randomly varying daily amounts that happened to result in the measured urinary concentration. From a single urine sample, it is impossible to differentiate between the myriad numbers of potential exposure profiles that could have produced that single measured value.

Fluid Intake

The concentration of the analyte in the urine (μg/l) also depends upon the quantity of fluid excreted by the kidneys. Urine is created to excrete metabolic waste products to prevent their build-up in the blood. If one has zero recent fluid intake from food and drink, a minimal amount of fluid is passed as a carrier vehicle. When food and drink are consumed, the excess fluids increase the excreted fluid volume, which decreases the concentration of the pesticide or its metabolite in the urine. For example, diuresis (e.g., excessive liquid intake or hydration) may artificially reduce the concentration of a pesticide or its metabolite in the urine, but not the amount. Traditionally, urinary creatinine has been used to correct for variable urine dilution (see discussion below). In addition, urinary specific gravity and osmolality, not recorded in the NHANES-III, have been used to help correct for hydration status in spot urine samples, but such methods have their own methodological uncertainties.

Body Chemistry

The rate of excretion of any solute from the blood into the urine depends on several factors: passive glomerular filtration of smaller molecules; facilitated secretion of larger molecules through the renal tubules; and downstream reabsorption of substances from renal tubules. Blood pressure (BP), diabetes, renal disease, renal blood flow, urinary tract obstruction and certain drugs affect the glomerular filtration rate and tubular secretion rate differently, and these rates may vary widely among individuals based upon their unique combination of physiological parameters. It should be noted that the tubular secretion processes are not strictly linear because an increase of a tubularly-secreted pesticide analyte in the blood may not always lead to an increased elimination. Each facilitated renal transport system has a maximal rate, or transport maximum (Tm), at which it can secrete a given solute with a molecular size too large to allow passage through the glomerular filter (Berkhin and Humphreys, 2001). In this process, a specific carrier protein can bind to one of possibly several ligands, such as a pesticide analyte. The bound chemicals transport through the tubular walls, the ligand (analyte) is released into the urine and, the protein returns to pick up another ligand. Thus, the amount of a particular solute secreted by this mechanism is approximately proportionate to the amount present up to the Tm for that solute when no other competing ligands are present. However, at higher concentrations, the transport mechanism is saturated and there is no appreciable increment in the transport rate because there are no more carrier proteins available. Consequently, a spot urine sample from a pesticide applicator who has had a high accidental exposure leading to a large dosage of a mixture of pesticides may provide misleading information, because there may be competition for the limited number of transport proteins that facilitate the process of tubular secretion for a given target pesticide analyte.

For example, a subject who had a recent therapeutic dose of penicillin may preferentially tubularly secrete the higher concentration of penicillin rather than a lower concentration of a targeted pesticide analyte if they both required the same carrier protein for their transport. Thus, a subject who may have been exposed to a pesticide may be excreting a lesser quantity of a target analyte because of the presence of a competitive ligand. In NHANES-III, some subjects reported taking medication, but did not specify which pharmaceuticals they took. Also, they may have received significant exposures to other environmental chemicals concurrent with or near the time of exposure to the pesticides measured. Therefore, recent use of medication or over the counter drugs or concurrent environmental exposures may have impacted the urinary concentration of the pesticide analytes of interest in this study.

Finally, the transient factors that influence glomerular filtration may be independent of the factors that influence the rates of tubular secretion of pesticide analytes. Thus, the elimination rate of a smaller molecule by glomerular filtration does not parallel the secretion rate of those larger analyte molecules excreted by tubular secretion. These uncertainties may be partially resolved by use of the subject's excretion of creatinine (Cn), a metabolic breakdown product of cellular creatine, under a specific set of circumstances which might be related to a subject's normal day-to-day exposure activities.

Creatinine Correction and Daily Excretion

Creatinine is a byproduct of skeletal muscle metabolism of creatine, and is picked up continuously in the blood at a relatively constant rate that is proportional to the muscle mass of the individual. Additional increments of creatinine occur from dietary red-meat intake and use of creatine as a body-building supplement. Creatinine is cleared from the blood plasma in the kidney at an approximately constant rate, primarily by filtration through the glomerulus (∼80%) but a small amount is cleared by tubular secretion (∼20%). Conversely, pesticide or pesticide metabolite intake is a transient process and it is not relatively constant from day to day. For example, dietary intake may occur at different times of the day and the pesticide content of the diet changes markedly with the source and identity of the food eaten. Inhalation and dermal contact with pesticides may occur infrequently (e.g., once per month when pets are treated with flea powder) or by chance when a subject visits a location that was recently treated and touches a contaminated object. Consequently, in the blood, a pesticide analyte concentration fluctuates much more rapidly than the concentration of creatinine which is normally kept in homeostasis.

Estimating the Dose of a Pesticide

When an analyte value is reported above the MDL, all we know is that the subject has been exposed to it or its parent compound, but we have absolutely no information on the magnitude or timing of that previous exposure. If, and only if, the subject has a relatively constant pesticide/metabolite intake (μg/kg/day) and a constant dietary intake of red meat, and the tubular secretion transport mechanism is not saturated for the target pesticide or its metabolite, the body will equilibrate and excrete in the urine approximately constant amounts of that analyte and creatinine per day. This set of circumstances may not exist in practice, and it restricts our analysis to a set of conditions where a linear tubular secretion process is taking place below the transport-maximum concentrations where saturation makes the clearance process zero-order non-linear. Human Tm values for these analytes are not reported in the literature so we cannot check the validity of our assumptions concerning them. However, it is the only way that the NHANES-III spot-sample data can be transformed from a urinary concentration distribution (μg/l) into an exposure distribution (μg/kg/day) consistent with the RfD. This Cn correction is not straightforward and its usage is controversial (Dell'Orto et al., 1987). We have analyzed these data using such a Cn correction as described below. It is a potentially important addition to the body of knowledge for NHANES data analysis and usage in risk assessment.

In practice, there is no consistent convention for use of Cn to correct for the influence of hydration. For example, the CDC did not report analyte data for adult urine samples in NHANES-III, for which the Cn concentrations were below 0.3 g/l on the basis that it represented a highly dilute urine sample (Hill et al., 1995; Adgate et al., 2001). The need for an arbitrary choice of Cn cut off (a value of 0.30 g/l acceptable and 0.29 g/l not acceptable) arises because extremes of hydration and dehydration may cause different changes in the rates of Cn and pesticide analyte excretion (Alessio et al., 1985; Sata et al., 1995). Creatinine clearance may also be affected by the renal functional status of the subject, BP, pregnancy, drugs, exercise and red meat intake (Neithardt et al., 2002). The ratio of pesticide or pesticide metabolite clearance to Cn clearance may also be affected by these same conditions. For example, Radha and Bessman (1983) studied the effect of a 1-h intensive exercise on young males placed on a meat-free diet. Immediately following the exercise, Cn excretion rates decreased significantly. Bennett and Wilkins (1993) found that their subjects' BP was inversely related to their urinary Cn concentration. Bingham and Cummings (1985) found that adding 260 g/day of cooked meat to volunteers on a uniform metabolic diet raised their Cn excretion rates by 23%. With fluctuations of these quantities, the average coefficients of variation of sequential daily Cn excretion in healthy adults were measured as 4% over a short interval of 5 days (Cho and Yi, 1986) and 12% over longer periods of 24–97 days (Greenblatt et al., 1976; Waterlow, 1986). Given these caveats, we make the following simplifying assumptions:

• 1) There is no pathology involved, such as kidney failure or a muscle-wasting disease;

• 2) The food intake is by typical omnivorous diet, with neither excessive red meat consumption nor a strict vegetarian diet.

• 3) The ratio of pesticide analyte to Cn in a spot urine sample represents an average value corresponding to a complete 24-h total urine collection centered on the urine sample interval.

• 4) There is no pesticide metabolite intake by dietary ingestion or any other route of exposure.

• 5) The pesticide analyte's Transport Maximum is not exceeded by any subject.

Results

Computation of Equivalent Pesticide Dose

The pesticide concentration in the blood represents an integrated net effect of the subjects' previous exposures to pesticides via ingestion (food, water, hand to mouth activity), inhalation and dermal contacts. The health effects of such exposures are usually a monotonic increasing function of the systemic dose to the exposed person, in terms of micrograms per kilogram body weight per day (μg/kg/day). Although people may be exposed to a pesticide by inhalation and dermal contact, we choose to report these data as if all the pesticide/metabolite residue had a dietary origin, because the USEPA Office of Pesticide Programs has established a set of RfDs based primarily on dietary intake of their parent pesticides. Thus, the person with the highest imputed dosage in μg/kg/day is considered to have been the person with the highest dietary exposure to the pesticide. As described previously, each person excretes Cn at an approximately constant rate on a daily basis, with homeostasis regulated by the kidney. The effect of diuresis can also be corrected using urine specific gravity (Araki, 1980; Vij and Howell, 1998) that may be more accurate than the Cn method, but there are also more physiological factors influencing specific gravity than Cn (Trevisan, 1990).

Owing to the large interindividual variation in Cn excretion, we chose to model this variability for each subject, instead of using the same mean values for all subjects. There are established norms for excretion of Cn on a daily basis, as a function of age, gender and body surface area (BSA). The BSA determines the daily heat loss that must be recovered through cellular metabolism that generates the Cn that is excreted by the urine. The BSA can be modeled as a function of a person's age, gender, height and weight. An expression for Cn production by “normal” males and females in the NHANES-III can be developed because there is an age, gender, height and weight recorded for each subject.

For BSA, the Mosteller (1987) formula is

For creatinine clearance (CnCl), ml plasma cleared/min/1.73 m² BSA, the Cockcroft and Gault (1976) formula is

Therefore, the expected daily production of Cn is

Male: (mg Cn/dl serum) (ml plasma/min/1.73 m² BSA) =(140 − age [year])^*wt(kg)/72.

Simplifying, and substituting for BSA, the expected micrograms Cn excreted per day is:

¹

Using these formulae, an 80 kg man, 40 years old and 170 cm tall, would have an expected Cn daily production of 1.78 g/day, and a female of the same age, height and weight would have a Cn production of 1.51 g/day.

The NHANES-III did not collect urine from children for reporting pesticide analyte concentration data, so the procedure we developed was not applied for estimating a child's dose of pesticide. However, provided the appropriate equations for children are used to predict a child's daily creatinine clearance rate as a function of gender, age and body surface area, the same technique is applicable (Shull et al., 1978; Schwartz et al., 1987). Such equations for children are not expected to be as accurate as those for adults owing to rapidly changing metabolism and BSA from infancy through puberty. We note that this proposed method using urine concentration may be useful especially for infants when it is difficult to collect a urine sample.

Computation of Total Exposure

The reported urinary analyte concentration in μg/l is corrected to units of eliminated dose in μg/kg/day by the Cn concentration. This makes sense if, and only if, the subject is at steady state with respect to the analyte, and the tubular secretion Tm is not exceeded. Then, the parent pesticide intake in molar units would also be equal to the target analyte excretion on a molar basis (assuming that one molecule of parent pesticide metabolizes to one or more different molecules of analyte). To correct the analyte excretion for parent pesticide intake, that value must be multiplied by the ratio of molecular weight pesticide to the molecular weight of the analyte. However, not all analytes are excreted in the urine, as some are soluble in the bile and are eliminated in the feces. Also, some of the pesticide may be excreted unchanged, so we may be neglecting that part of the exposure. Thus, to account for the amount of pesticide absorbed that is distributed, metabolized, or excreted differently, the amount of pesticide corresponding to the analyte excreted also must be multiplied by the ratio of mol pesticide intake to the known mol analyte excreted in urine from that exposure. Nolan et al. (1984) reported an average of 1.47 mol chlorpyrifos intake per mole TCPY excreted in the urine but not all researchers make this correction (Buck et al., 2001).

Following our data analysis procedure, we can convert each pair of values for urinary analyte concentration and Cn concentration into a value of the virtual parent pesticide exposure as a daily dietary intake in units of μg/kg/day that would equilibrate to those measured values, subject to the restrictive assumption of nonsaturation of tubular secretion. This exposure can then be compared to an RfD established by the EPA for the parent compound. There are uncertainty factors involved in setting an RfD; so a value above the RfD, but below the lowest observed effect level, does not necessarily imply a health effect is likely.

Since some analytes have multiple parents, such as 3,5,6-TCPY which is a metabolite of both chlorpyrifos and chlorpyrifos-methyl, we have chosen the parent compound that people not occupationally applying pesticides would most likely be exposed to in normal daily activities. In the case cited above, chlorpyrifos would be the most likely compound the subject was exposed to, as chlorpyrifos methyl is an insecticide used primarily in fumigation of stored grain.

Treatment of the Minimum Detectable Levels (MDL)

The MDL established by the CDC for the chlorpyrifos metabolite 3,5,6-TCPY is 1 μg/l. Consequently, if a subject has a <MDL residue of TCPY and a Cn concentration in the urine of 0.3 g/dl, then the Cn corrected value would be (<1 μg/l)/[(0.3 g/dl)(10 dl/l)] or <0.33 μg TCPY residue/g Cn. Owing to the variable denominator Cn concentration, each <1 μg/l MDL value translates into a different maximum possible μg TCPY residue/g Cn that could not have been observed. To analyze these MDL data, with the observed data, we use the method of maximum likelihood estimation (MLE) to fit the data set with the S_B probability model (Johnson, 1949), also known as the four-parameter lognormal distribution.

Fitting the Censored S _B Distribution

The distribution of pesticide/metabolite residues in urine has two asymptotic properties quite different from those of a classic two-parameter lognormal distribution, where the variable X is bounded as 0<X<∞. These differences are that subjects who were not recently exposed to a given pesticide may have that pesticide's metabolite concentration in the urine equal to zero, and the maximum possible urine concentration cannot approach infinity and must be finite.

The S_B can be expressed by a lognormal distribution bounded by a lower concentration (X_min) and a maximum concentration (X_max) (Mage, 1980). The transformation is that the variable Y, where Y=(X−X_min)/(X_max−X) is lognormally distributed. The variable X_min can be negative, but the values predicted equal to or less than zero are censored and all are treated as equal to zero. Note that the S_B model can fit a continuum of distributions that connect the normal to the two-parameter lognormal distribution, and must fit any data set equal to or better than either of them.

The MLE method determines the distribution that has the maximum probability of drawing a given observed data set as an independent random sample. Let the probability of drawing a sample with value Y be p(Y) and the probability of drawing a sample with a value equal or less than Y be F(Y), where F(Y) is the integral of p(Y) dY between 0 and Y. The likelihood function Λ to be maximized is described by the relationship,

where ΠF(Y) represents the product function over all the transformed MDL values, and Πp(Y) represents the product function over all values above the MDL. The MLE method was applied to the Cn corrected NHANES-III data, and the optimal parameters are provided in Table 2.

Table 2 Modeled four-parameter lognormal distributions fit to pesticide analyte data as virtual dietary exposure of its parent pesticide (μg/kg/day)

The results are shown for the 11 of 12 analytes where the MLE procedure converged to a stable solution. The analyte carbofuranphenol could not be modeled because there were only 16 observations above the MDL, so our MLE procedure would not converge. Table 3 provides a typical example of the fit of the S_B model using the chlorpyrifos dosage predicted from the TCPY data. Owing to the MDL values, an exact comparison of observations and predictions is not possible, except for the portion of the table where data are all above the MDL of TCPY of 1 μg/l, which here occurs at a frequency of 85%.

Table 3 Example of S_B model fit with the chlorpyrifos estimated dosage corresponding to urinary analyte TCPY, N=973

Discussion

Interpretation of Data and Weighting Functions

The pesticide sample cohort does not represent a probability sample because the subjects were volunteers from the original random sample. Consequently, they do not constitute a true random sample from the NHANES-III random sample, and their original sampling weights strictly do not apply to these data. However, there are conditions which might allow a nonweighted analysis of their data to have some validity. In mathematical terms, let each ith subject in the sample of size n have a weighting function wi, and a measured parameter yi. The mean value (Y) would be computed as Y=Σwi yi/Σwi. If the correlation between wi and yi is zero, then the expected value of the weighted mean of {yi} would be equal to the nonweighted mean

We have analyzed the correlations of sample weights (wi) with urinary concentrations (yi) for all 12 analytes, in sets of residue data that were above their respective MDL. Only two of the 12 analytes had correlations significantly different from zero, 1-naphthol and 2,5-dichlorophenol, which had values r=0.13 and r=−0.13, respectively. Since the corrections for these low correlations are of order r², or 2% to their calculated means and distributional values, we chose to treat all 12 analyte values without making the weighting corrections, as that would provide negligible error in comparison to all other uncertainties in these data and in our analysis.

The subjects' choice whether to volunteer or not was made without any knowledge of their exposure to the 12 target parent compounds under study. If their choice to participate was not influenced by a demographic characteristic that was correlated with exposure to any of those parent pesticides, then the sample might constitute a ‘pseudo random sample’. For example, if males were more exposed more than females and only males volunteered, then the sample would be biased to a greater percentage of higher values than would exist in the entire population of males and females. In our analyses, there was no significant difference between the demographic characteristics of the subjects voluntarily providing urine samples and the subjects in the complete NHANES-III cohort. We therefore make the assumption that the choice to volunteer a urine sample for pesticide analysis was made independently of any consideration related to their likelihood of being exposed to the parent pesticides, or any other demographic characteristics correlated with their pesticide exposures.

With such correlations not withstanding, we can interpret the distributions as approximating the distribution of pesticide exposures of people in the US, aged 20–59 years during the years of sampling. Table 2 shows that two pesticide residues (2,4,5-TCP and 2,4,6-TCP) have values at the 99.9th percentile that are greater (transformed) than the RfD for lindane, their most likely parent compound. 24DCP is higher at the 99th percentile than its own dietary RfD of 3 μg/kg/day. It is not a human metabolite of 2,4-D, but it can arise from bacterial degradation of 2,4-D (USEPA, 1980). Chlorpyrifos has a predicted frequency of 0.02% above the RfD. When interpreting these data, relative to their RfD for the various parent compounds or on their own, the distribution of values we hypothesize may be equal to or less in frequency for the exposures shown, due to the possibility of a person's exposure to the pesticide's metabolite rather than to the parent pesticide, as is the case for TCPY and chlorpyrifos (Wilson et al., 2003) and the possibility that the metabolite is also derived from other parent compounds. However, for those analytes for which there are no data available on fecal excretion, these values could underestimate the intake dosage if urinary excretion was less than 100% for the analyte. Of the analytes shown in Table 1, we have found human data showing less than 100% urinary excretion only for 3,5,6-TCPY. Given that the RfD is a calculated daily exposure value for which no human health effect is expected, we conclude that, based on this one biomonitoring sample, most subjects were not exposed to the target pesticides at a rate leading to a public health concern.

Recommendation for Future Work

We recommend that in future biological monitoring of urinary pesticide residues and metabolites, when possible, it is best to collect a total 24-h void and to record knowledge of the subject's body weight so that the μg equivalents/kg/day of a subject's average daily intake of parent pesticide can be estimated directly. This would eliminate the uncertainty inherent in the creatinine correction we have used. If a total 24-h urine collection is not possible because of a high rate of subject refusal or logistical problems, then the total void volume of the spot sample should be recorded along with the time from the previous void. We propose that the technique we have developed be utilized for comparing such exposures developed from urinary concentration data to regulatory estimates of exposures based on exposure models.

Summary and conclusion

We have reviewed the literature on the usage of urinary Cn excretion to normalize measurements of urinary pesticide residues, and have shown how an individual's physiologic data (age, gender, height, weight) can be used to estimate their daily Cn excretion rate to transform their pesticide/metabolite data into individual estimates of daily pesticide urinary excretion rate. Assuming equilibrium (molar intake rate of parent=molar excretion rate of analyte), our results show that the vast majority of the sample from the adult general public, during the 1988–1994 NHANES-III survey, may not have been exposed to the 12 target pesticides at a rate above the RfD. Our analysis assumes that the 978 adult subjects sampled represent the general public that does not apply pesticides occupationally and does not receive sufficiently high pesticide exposures that reach the transport maximum for the corresponding analytes which would call into question our Cn correction procedure. Only two of the subjects whose estimated dose exceeded an RfD had current occupations in the area of “Agricultural services, forestry and fishing” and their previous occupations were as ‘construction laborer’ and ‘mover of freight, stock and materials by hand’. We have used the Johnson S_B model to predict the distribution of ingestion dose rates of the imputed parent compound for each metabolite or residue and compared them to their respective RfDs. We recommend that such a procedure be considered in future studies of urinary pesticide analyte concentrations to provide an individualistic estimation procedure to be applied to each subject in the study.