MEMORANDUM

To:	Tim Leighton, David Miller, Philip Villaneuva, EPA 

From:	Jonathan Cohen, ICF International

Date:	6 March, 2008

Re:	EP-W-06-091, WA 0-02, TAF CM 19: Computations of human triclosan
dose based on NHANES urine concentrations.



	

SUMMARY

The National Health and Nutrition Surveys (NHANES) are a series of US
national surveys of the health and nutrition status of the
non-institutionalized civilian population conducted by the Centers for
Disease Control and Prevention. As part of the 2003-2004 NHANES, urinary
concentrations (μg/L) of triclosan
(2,4,4’-trichloro-2’-hydroxydiphenyl ether) were measured on a
random sample of 2,517 participants of ages 6 and over.  These
measurements represent concentrations in spot urine samples, and the
corresponding human dose (μg/kg-day) was not measured or estimated by
NHANES or CDC. Various methods have been proposed to estimate the dose
from the measured concentration, but they can be categorized into two
main groups:  one that uses measured pesticide concentrations in urine
directly and the other that standardizes urinary concentrations on the
basis of creatinine, a by-product of metabolism which is felt by some to
be less variable than urinary output.  A total of seven dose conversion
variations are presented in this document (and are designated Dose1 to
Dose7). These are summarized below: 

 

Mage Method: Mage et al. (2004, 2007) use the estimated daily creatinine
excretion for a demographic group; the triclosan concentration is
divided by the creatinine concentration, multiplied by the daily
creatinine excretion in μg/day, and divided by the body weight. There
are three variations on this, designated as Dose1, Dose2, and Dose3.

PANNA Method: Schafer et al (2004) use the estimated daily urine
excretion in L/day and the average body weight for a demographic group;
the triclosan concentration is multiplied by the daily urine excretion
in L/day, and divided by the average body weight.  There are two
variations on this method, designated as Dose4 and Dose5.

Geigy Method: For specific situations, some researchers at the EPA
Office of Research and Development (ORD) use the estimated daily urine
excretion in L/kg-day for a demographic group; the triclosan
concentration is multiplied by the estimated daily urine excretion in
L/kg-day. However, at this time, no ORD-wide approach has been adopted.
The two variations on this method, using urine excretion data from Geigy
(1981), are designated as Dose6 and Dose7.  

In this memorandum we describe the various conversion methods and
compare the results by demographic group for the estimated triclosan
dose.  An important caveat to all the reported results should be noted:
In this review, the pharmacokinetics of triclosan have not been
accounted for in the dose estimate because the use of human data would
require prior approval by the Human Studies Review Board (HSRB). The
calculations presented here assume 100 % of triclosan is excreted in
urine. In fact, a pharmacokinetic study has shown that only about 50 %
of triclosan is excreted in urine. On this basis, all the dose estimates
presented in this memorandum ought to be doubled (approximately). This
constant adjustment factor does not affect the comparative results.

 

 case, was 138 μg/kg-day.

NHANES

0 μL collected from a random one-third sample of 2,517 subjects of ages
6 and older. The dose conversion calculations also used the NHANES
measurements of creatinine concentrations, body weight, body height, as
well as the age, gender, and race of each subject. The NHANES 2003-2004
data were obtained from the NHANES website:   HYPERLINK
"http://www.cdc.gov/nchs/nhanes.htm"  www.cdc.gov/nchs/nhanes.htm 

The NHANES use a complex multi-stage, stratified, clustered sampling
design. Certain demographic groups were deliberately over-sampled,
including Mexican-Americans and Blacks. The publicly released
environmental phenols file L24_eph_c.xpt with the triclosan data
includes survey weights to adjust for the one-third random sampling as
well as for the over-sampling of certain demographic groups,
non-response, and non-coverage. The statistical analyses used the
laboratory survey weights (WTSCYR) to re-adjust the urinary triclosan
data to represent the national population. NHANES also includes the
strata and primary sampling unit (PSU) survey design parameters to
account for the stratification and clustering, but (for confidentiality
reasons) the public release version provides approximate pseudo-values
for these parameters instead of the actual values. The statistical
analyses used the pseudo-strata and pseudo-PSU values to properly
account for the uncertainty of the survey estimates and thus to compute
confidence intervals for the estimated dose.

NHANES urinary metabolite concentration data are not collected in a way
to directly determine the dose, and CDC has not reported dose estimates
for triclosan based on NHANES measurement data.

TRICLOSAN

asured urinary concentrations of total triclosan (free plus conjugated
species) on 2,517 randomly selected subjects of ages 6 years and over.
Total triclosan was measured on 100 μL samples using on-line
solid-phase extraction coupled to high performance liquid
chromatography-isotope dilution-tandem mass spectrometry. The lower
detection limit was 2.27 μg/L. 75% of the samples (survey-weighted)
were above the detection limit. For these analyses, we followed the CDC
recommended approach that substitutes the limit of detection divided by
the square root of 2 for all non-detects (values below the detection
limit). Measured concentrations ranged from 2.3 to 3,790 μg/L. The
geometric mean concentration (survey-weighted) was 13 μg/L.

More details about the NHANES triclosan data as well as summary
statistics and regression analyses for triclosan (μg/L) and
triclosan/creatinine (μg/g), are provided in Calafat et al, 2007. To
confirm our calculations, we calculated the same summary statistics
(geometric mean, 10th, 25th, 50th, 75th, 90th, and 95th percentiles,
with 95 percent confidence intervals) for the same demographic groups,
both for triclosan and triclosan/creatinine. Our tabulations are shown
in the attached Excel file “dose.conversions.030608.xls.” We used
SAS and SUDAAN software for these calculations. The geometric means and
confidence intervals either agreed exactly (most cases) or differed by
0.1. The differences for the percentiles were larger. For example, for
all subjects, Calafat et al (2007) report the 50th percentile triclosan
(μg/L) as 9.2 (7.9-10.9) but our calculations gave 9.5 (8.7-11.0).
Similarly, Calafat et al (2007) report the 95th percentile triclosan
(μg/L) as 459.0 (386.0-522.0) but our calculations gave 463.0
(393.0-527.0). These differences are likely due to the percentile
calculation algorithm. We tried various alternative approaches but were
not able to get exact agreement for any of the methods that we tried.

To calculate the percentiles and their confidence intervals for
triclosan, triclosan/creatinine, and the estimated dose we used the CDC
method given in CDC 2005, Appendix C. First, the percentile is estimated
using the survey weights and the SAS default method (in PROC
UNIVARIATE). Second, SUDAAN is used to estimate the proportion of values
below the estimated percentile, the standard error of the estimated
proportion, and hence the degrees-of-freedom-adjusted effective sample
size  (see Korn and Graubard, 1999, p. 65). Third, the Clopper-Pearson
method is used to calculate a 95% confidence interval for the proportion
in step 2. Fourth and finally, the percentiles corresponding to the
estimated proportion and its lower and upper bound are estimated using
the survey weights and the SAS default method.

Confidence intervals around the mean, geometric mean, 90th percentile,
95th percentile, and 99th percentile do not fully account for the
variability/uncertainty inherent in each of the parameters, and the
combination of variability and uncertainty across all of the combined
parameters, used in estimating the dose. For example, the procedures use
an assumption of steady state exposure. This may not be the actual
situation for many or most individuals, so the variability associated
with different exposures at different time intervals for different
durations prior to measurement may not be captured. The impact of
parameter variability and uncertainty are of particular concern for dose
estimates made for an individual for many chemicals. On the other hand,
because of the large NHANES sample size, some of the variability may
already be captured in the data set. More work in this area is needed. 
While it is difficult at this time to generate accurate confidence
intervals or uncertainty bounds, these analyses only account for the
statistical variability across the estimated dose estimates for the
population ignoring any uncertainty and variability in the dose
conversion equations themselves. By comparing the results across the
different dose estimation methods, we address only some of the
uncertainty in the dose conversion.  

DOSE CONVERSION

 data are spot urine samples for which the concentration in μg/L was
measured. To interpret these concentrations is difficult because they
cannot be directly related to a health effect. EPA evaluates health
effects in terms of a reference dose that represents the daily intake in
mg or μg per kilogram body weight that is not expected to be associated
with an adverse health effect. The conversion of measured spot urine
concentrations to daily doses is difficult because of highly variable
dilution caused by wide fluctuations in the fluid intake and excretion.
Dose calculation is also difficult because there is no way to determine
from the NHANES data how (oral, dermal, inhalation) and when (duration
and time interval prior to measurement) the actual contact with the
chemical occurred, and because of uncertainty and variability in the
absorption, distribution, metabolism, and excretion (ADME) parameters.
If NHANES collected total daily urine excretion for each participant,
then that participant’s dose could be accurately estimated by
multiplying the triclosan concentration by the total daily urine
excretion and then dividing by the body weight. However, NHANES only
collected spot urine samples so that total urine excretion was not
measured. The Schafer et al (2004) dose conversion methods and the
similar methods used by some ORD researchers estimate the daily urine
excretion using average values for a given age/gender group. The Mage et
al (2004, 2007) dose conversion methods are based on estimates of the
daily creatinine excretion (μg/day) from various studies relating the
daily creatinine excretion to age, gender, race, weight, and height. The
dose is estimated by multiplying the concentration per g of creatinine
by the estimated daily creatinine excretion and then dividing by the
body weight. These alternative methods are detailed in this section.

We also note that we have assumed throughout this discussion that
triclosan and other analytes of interest are not metabolized so that all
the ingested analyte is excreted in the urine. To apply these same
methods to parent contaminants that are metabolized, it is necessary to
make two further adjustments to estimate the dose. The first adjustment
is the ratio of parent to metabolite molecular weight. The second
adjustment is the ratio of mols of parent ingested to mols of metabolite
excreted in the urine. Human pharmacokinetic data do exist for triclosan
but would require prior approval by the HSRB before the Office of
Pesticides Program (OPP) can use the data in risk assessments. As
mentioned above, a pharmacokinetic study has shown that only about 50 %
of triclosan is excreted in urine. On this basis, all the dose estimates
presented in this memorandum ought to be doubled (approximately).

Mage et al (2004):  Dose1

Mage et al (2004) derived creatinine excretion rates for adults 18 and
older using creatinine clearance equations that were obtained by
Cockcroft and Gault (1976) based on 24-hour urine data from 249 male and
18 female adults in Canada. They obtained the following equations:

Creatinine Excretion (μg/day) =  1.93 (140 – Age) Weight1.5
Height0.5, Males 18+		(1) 	

Creatinine Excretion (μg/day) =  1.64 (140 – Age) Weight1.5
Height0.5, Females 18+	(2)	

In these equations, and throughout this memorandum, Age is age in years,
Weight is weight in kg, Height is height in cm.

The dose conversion equation is given by:

Dose (μg/kg-day) = 

{Triclosan (μg/L) / Creatinine (μg/L)} ( Creatinine Excretion
(μg/day) / Weight		(3)

To compare this method to the other dose conversion methods, we applied
the same equations to all adults and children, regardless of age.

The creatinine conversion Equation 3 uses the concentration of
creatinine to adjust for varying urine volumes, due to variation among
the individuals in a demographic group and due to variation within the
same individual, e.g., the higher levels of fluid intake and excretion
in the hotter summer season. Creatinine is a natural metabolic byproduct
of muscle tissue creatine. Creatinine is picked up by the blood at a
rate roughly proportional to the amount of muscle tissue. Additional
amounts of creatinine are obtained from dietary red meat intake or from
creatine body-building supplements.  Creatinine, triclosan, and other
blood solutes are excreted from the blood into the urine by passive
glomerular filtration and/or by secretion through the renal tubules. 

Assumptions:  

The subjects in the creatinine studies are representative of the
demographic groups.

The spot sample triclosan/creatinine ratio is representative of the
24-hour average.

Subjects have reached steady state, equilibrium conditions with respect
to creatinine and triclosan urinary concentrations (in particular, this
assumes an approximately constant daily triclosan exposure). 

For each demographic group, the absorption, distribution, metabolism,
and excretion (ADME) parameters are the same across all individuals and
are constant within individuals over time.

Subjects are omnivorous, neither eating excessive red meat nor being
strictly vegetarian (since creatinine increases with the amount of red
meat eaten).

Subjects are moderately active (neither sedentary nor body builders). 

Subjects are healthy (kidney disease or muscle wasting diseases such as
Duchenne muscular dystrophy or Cushing’s syndrome reduce the
creatinine excretion rate).

Subjects do not exceed the triclosan’s tubular secretion transport
maximum, so that the tubular secretion mechanism is not saturated and
the process is (approximately) linear.

Subjects are not on medicines such as penicillin that reduce creatinine
tubular secretion.

Subjects have not been significantly exposed to other contaminants,
since if not, those contaminants might compete for the same carrier
protein used to facilitate tubular secretion.

Subjects are not taking creatinine dietary supplements.

Subjects have moderate muscle mass.

Subjects do not have febrile disorders.

Subjects do not have decreased renal function (e.g., caused by GFR or
tubular disorders).

Female subjects are not pregnant (since pregnancy increases creatinine
excretion).

Mage et al (2007) without obesity correction: Dose2

Mage et al (2007) extended the analyses of Mage et al (2004) to include
creatinine excretion rates for children and to make adjustments for
race, obesity, and continuity. In this section we present the equations
without the extra adjustment for obesity. The list of assumptions needed
for these analyses is the same as the list in the previous subsection
“Mage et al (2004).” Since the demographic groups are smaller
(including separate groups for children and by race), the assumption
that the creatinine subjects are representative of their demographic
groups is more valid.   

The Mage et al (2007) includes creatinine excretion equations for
children up to 36 months (from Viteri and Alverado, 1970), children ages
3 to 18 years (from Remer et al, 2002), and adults 18 and older (from
Cockcroft and Gault, 1976). In this memorandum we do not report or use
the equations for children up to 36 months since the urinary triclosan
concentration data, and, indeed, most of the NHANES concentration data,
are only for persons ages 6 and older.

The following creatinine excretion equations were used, together with
the dose conversion Equation 3.

Creatinine Excretion (μg/day), Males 3-14, Height 90-168 cm, 

=  1.085 Height {6.265 + 0.0564 (Height – 168)} M ( 1000					(4) 

Creatinine Excretion (μg/day), Males 3-14, Height 168-186 cm, 

=  1.085 Height {6.265 + 0.2550 (Height – 168)} M ( 1000					(5)

Creatinine Excretion (μg/day), Males 15-17, Height 90-168 cm, 

=  1.085 Height {6.265 + 0.0564 (Height – 168)} {1 + 0.045 (Age –
14) B} M ( 1000	(6)

Creatinine Excretion (μg/day), Males 15-17, Height 168-186 cm,

=  1.085 Height {6.265 + 0.2550 (Height – 168)} {1 + 0.045 (Age –
14) B} M ( 1000	(7)

Creatinine Excretion (μg/day), Males 18+,

 =  0.926 ( 1.93 (140 – Age) Weight1.5 Height0.5 (1 + 0.18 B)				(8) 	

Creatinine Excretion (μg/day), Females 3-14, 

=  1.008 Height {2.045 exp(0.01552 (Height – 90)} F ( 1000				(9)

Creatinine Excretion (μg/day), Females 15-17, 

=  1.008 Height {2.045 exp(0.01552 (Height – 90)} {1 + 0.045 (Age –
14) B} F ( 1000	(10)

Creatinine Excretion (μg/day), Females 18+,

 =  0.993 ( 1.64 (140 – Age) Weight1.5 Height0.5 (1 + 0.18 B)				(11)

In these equations, B is the indicator for Black, Non-Hispanic, based on
the NHANES variable RIDRETH2 that re-assigns the race to the main race
category if a subject reports multiple races and also reports a primary
race. Note that the NHANES public release data does not separate race
and ethnicity. The available categories are: White, Non-Hispanic; Black,
Non-Hispanic; Mexican- American; and Other. M and F are adjustments for
children’s body surface area when their body weight does not equal the
median weight for their age:

B = 1, if Black, Non-Hispanic

B = 0, if not Black, Non-Hispanic								(12)

M = {Weight / [14 + 1.433 (Age – 3) + 0.22 (Age – 3)2 – 0.00533 
(Age – 3)3] }0.5		(13)

F = {Weight / [14 + 0.4 (Age – 3) + 0.52 (Age – 3)2 – 0.024  (Age
– 3)3] }0.5			(14)

A few comments are needed to explain how the Mage et al (2007) equations
were interpreted. In Mage et al (2007), Equations 5 and 6 are assumed to
apply to males 15-18. The adjustment factors 1.085 and 0.926 in
Equations 4 to 8 were chosen by Mage et al (2007) to make the equations
continuous at age 18 for a male of the median height and weight. Thus
Equation 8 is the continuity- and race-adjusted version of Equation 1.
However, these adjustments do not make the equations consistent at age
18 for a male of a different height or weight. Therefore we assume here
that Equations 5 and 6 only apply to males 15-17 and not to males 18 or
over (where Equation 8 applies). Similarly, for females, the factors
1.008 and 0.993 were chosen by Mage et al (2007) to make the Equations 9
to 11 continuous at age 18 for a female of the median height and weight,
and we decided here to make Equations 9 and 10 apply to females 15-17
only. Another issue is that Mage et al (2007) do not state how the
category Black, Non-Hispanic is to be defined for NHANES data, although
it is apparent from the text that the intended adjustment is for all
Black persons but it has to be applied to the slightly smaller category
Black, Non-Hispanic since Black is not reported as a separate category
in the public release version. We used the RIDRETH2 definition since it
appears from the tabulated results that is the definition used in the
triclosan paper by Calafat et al, 2007. The impacts of these choices are
expected to be small since the number of NHANES subjects affected is
small.

Mage et al (2007) with obesity correction: Dose3

Mage et al (2007) noted that the creatinine excretion Equations 4 to 14
do not apply for the obese. Subjects with excessive weight for their
height caused by ascites (abdominal fluid retention) or an increase in
adipose tissue (non-lean body mass) do not have excess creatinine since
serous fluids and adipose tissue do not produce creatinine. Thus the
original equations will tend to overestimate creatinine excretion in the
obese. For this version of the dose conversion equations we applied the
following equations from Mage et al (2007) for obese subjects:

(μg/day), Obese Males,

 =  (137 – Age) (0.0805 + 3.42 /  BMI)							(15)

Creatinine Excretion (μg/day), Obese Females,

 =  (146 – Age) (0.081 + 2.75 /  BMI)								(16)

BMI = Body Mass Index = Weight (kg) / [Height (m)]2					(17)

Although a more accurate definition of obesity could be obtained based
on the waist circumference,  and skinfold thicknesses (to separate
muscularity from obesity), we used the simple definition of obesity as
BMI > 30. Thus for the obesity-adjusted Mage et al (2007) equations we
apply Equations 4 to 14 if BMI <= 30 and apply Equations 15 and 16 if
BMI > 30. Mage et al (2007) also offer an alternative obesity adjustment
method based on estimating lean body mass; we did not use this
alternative method.

Shafer et al (2004): Dose4

Shafer et al (2004), for the Pesticide Action Network of North America
(PANNA), used a simpler approach based on scaling up the spot urine
volumes to the estimated daily urine volume excreted (L/day). They used
the following equation:

Dose (μg/kg-day) = 

Triclosan (μg/L) ( Urine Volume (L/day) / Average Weight					(18)

They used average urine excretion and weight values quoted from Snyder
et al (1975), although those average values are not precisely reported
in the tables and graphs in Snyder et al (1975).

Table 1. Urine volumes and body weights used in PANNA (2004), based on
Snyder et al (1975) 

Gender	Age	Body Weight (kg)	Urine volume (L/day)

All	6-11	30	0.8

All	12-19	50	1.0

Female	20-59	55	1.1

Male	20-59	70	1.5



For these analyses we modified the Schafer et al (2004) equations so
that the averages for ages 20-59 were assumed to apply to all adults 20
and over.

Assumptions:  

The average urine volumes (L/day) and body weights (kg) are
representative of the demographic groups. More precisely, since only the
ratios are used, the assumption is that the ratio of average urine
volume to average body weight is representative of the urine
volume/weight ratios for the demographic groups.

Subjects have reached steady state, equilibrium conditions with respect
to triclosan urinary concentrations. 

For each demographic group, the absorption, distribution, metabolism,
and excretion (ADME) parameters are the same across all individuals and
are constant within individuals over time.

The spot sample triclosan concentration is representative of the 24-hour
average.

Although the number of assumptions made is only four, these assumptions
are very approximate. The first assumption is that body weight and daily
urine volume are approximately constant for all subjects in each
demographic group, but these values vary appreciably. Mage et al (2007)
pointed out that Snyder et al (1975) say that “within an age group,
the daily volume of urine probably increases with body size.” They
also point out that the variability in daily urine volume is much
greater than the variability in daily urine excretion. For example, for
males 9-13, they quote that the 5th to 95th percentile daily urine
volume is from 12 to 43.2 mL/kg-day (Mattson and Lindstrom, 1995) but
the daily creatinine excretion is from 11.3 to 27.7 mg/kg-day (Remer et
al, 2002). Furthermore, urine volumes are not consistent from day to day
and tend to be higher in hot weather; the Mage et al (2004, 2007) method
uses the creatinine concentration to normalize this effect.

The Schafer et al (2004) approach can be expected to be more accurate if
a finer demographic classification could be made available; this assumes
that more detailed daily urine volume data can be made available.

Schafer et al (2004) with actual body weights: Dose5

The Schafer et al (2004) approach can be slightly improved if the actual
body weight is used instead of the demographic group average body
weight:

Dose (μg/kg-day) = 

Triclosan (μg/L) ( Urine Volume (L/day) / Weight						(18)

We applied this revised approach using the same demographic groups as in
Table 1. The assumptions are similar:

The average urine volumes in L/day are representative of the demographic
groups.

Subjects have reached steady state, equilibrium conditions with respect
to triclosan urinary concentrations.

For each demographic group, the absorption, distribution, metabolism,
and excretion (ADME) parameters are the same across all individuals and
are constant within individuals over time.

The spot sample triclosan concentration is representative of the 24-hour
average.

Geigy Method – Mean: Dose6 

ORD has not established a standardized or recommended procedure for
converting urinary concentrations to dose, but they routinely do not
correct for creatinine excretion (as, e.g., Mage does)  Although no
universally recommended ORD procedure exists, in some situations, some
researchers from ORD have and do use a very similar approach to
Schaf牥攠⁴污⠠〲㐰⸩䠠睯癥牥‬湩瑳慥⁤景甠楳杮琠
敨搠浥杯慲桰捩朠潲灵愠敶慲敧甠楲敮瘠汯浵⁥湩䰠搯
祡‬桴祥栠癡⁥獵摥琠敨搠浥杯慲桰捩朠潲灵愠敶慲敧
甠楲敮瘠汯浵⁥湩䰠欯ⵧ慤⹹吠楨⁳楧敶⁳桴⁥煥慵楴
湯ഺ

Dose (μg/kg-day) =  Triclosan (μg/L) ( Average Urine Volume (L/kg-day)
		(19)

Morgan et al (2008) applied this method to estimate the children’s
dose of 2,4-D. Since a default appropriate set of average urine
excretion volumes has not been formalized by ORD, and the choice of such
standard urine volumes may depend upon the study and the available urine
volume data,  we have attempted to simulate this method by using  age
group and body-weight urine excretion volume data from Geigy Scientific
Tables (Geigy, 1981).

Geigy (1981) provides two tables with urine volume data. Their Table 2
gives the mean (m) and  standard deviation (s) of the urine volume
(mL/kg-day) for the following groups of ages 6 and above. Since the
groups overlap, we re-defined the  groups to avoid overlaps: 

“5-7 years,” redefined as ages 5 and 6. m = 25, s = 7.

“7-11 years,” redefined as ages 7, 8, 9, and 10. m = 25, s = 7.

“11-14 years,” redefined as ages 11, 12, and 13. m = 19, s = 3.

‘Young males,” not used.

To obtain data for adults we used their Table 1, which includes the mean
and standard deviation of the urine volume (mL/day) for the following
groups:

“33 men,” redefined as males ages 14 or over. m = 1360, s = 443.

“39 women, no ovulation inhibitors,” redefined as females ages 14 or
over. m = 1130, s = 423.

‘30 women, ovulation inhibitors.” m = 980, s = 449. Not used.

Note that the two sets of data use different units (mL/kg-day and
mL/day). To calculate the mean and standard deviation for persons 14 and
over, we divided the urine volume (mL/day) by the body weight. We also
divided these rates by 1000 to convert from mL to L. Also note that for
simplicity we chose not to separate the females according to their use
of ovulation inhibitors, although such information can be extracted from
NHANES.

For the Geigy – Mean approach we used the values of m to get the mean
urine volumes for each age/gender group.

Assumptions:  

The average urine volumes in L/kg-day are representative of the
demographic groups. 

Subjects have reached steady state, equilibrium conditions with respect
to triclosan urinary concentrations.

For each demographic group, the absorption, distribution, metabolism,
and excretion (ADME) parameters are the same across all individuals and
are constant within individuals over time.

The spot sample triclosan concentration is representative of the 24-hour
average.

Just as for Schafer et al (2004), the first assumption is very
approximate.

Geigy Method – Upper Bound: Dose7

Finally, we used the above-described (ORD-like) approach to get an upper
bound estimate of the dose by using information on the variability
associated with urine volumes (specifically, the standard deviation
listed in the Geigy (1981) publication). Assuming the urine volumes are
normally distributed, the 95th percentile urine volume is given by

95th percentile urine volume = Mean  + Z95 ( Standard Deviation				(20) 


Z95 is the 95th percentile of a standard normal distribution,
approximately 1.68.

The values of mean and standard deviation are estimated by the values of
m and s in the subsection “Geigy Method – Mean.” To calculate the
mean and standard deviation for persons 14 and over, we divided the m
and s for urine volume (mL/day) by the body weight. The upper bound dose
estimate is then given by:

Dose (μg/kg-day) = 

Triclosan (μg/L) ( 95th Percentile Urine Volume (L/kg-day)					(21)

This estimate is not the 95 percentile estimated dose, but rather an
upper bound estimate of the mean dose which was derived by using an 95
percentile urine volume. If all the dose conversion assumptions applied,
if the triclosan concentration was exactly measured, and if the only
uncertainty and variability was the urine volume, then this estimate
would also be the 95th percentile estimate dose for that subject. 

RESULTS

We applied the seven dose conversion methods listed in the previous
section to the NHANES 2003-2004 triclosan data.

The attached Excel file “triclosan.data.030608.xls” contains the raw
data and the estimated doses for each subject. The “Variables”
worksheet lists and describes the variables in the database and the
“TRICLOSAN” worksheet contains the data. The same data is provided
in the SAS file “triclosan.sas7bdat.” The SAS code used is in the
attached SAS program “convert.030608.sas.”    

Using the same demographic groupings as in Calafat et al (2007), we
calculated various summary statistics for the triclosan concentration,
triclosan/creatinine, and the seven estimated dose values. We computed
the mean, geometric mean, and selected percentiles. We also computed 95%
confidence intervals for these summary statistics. The detailed results
are given in the attached Excel file “dose.conversions.030608.xls.”
The means, geometric means, medians, and 99th percentiles for each dose
conversion method are compared in the attached Excel file
“dose.conversions.short version.030608.xls” and in Table 2. 

We can examine Table 2 in detail to compare the dose conversion methods
overall and for different demographic groups. Since the Dose7 method is
intended to be an upper bound estimate, we will discuss the other six
dose conversion methods. For all subjects, the means vary from 1.4 (Mage
et al methods Dose2 and Dose3) to 1.9 (Schafer et al method Dose4)
μg/kg-day, so that the maximum difference is 35%. However, the
differences in the means when population subgroups are considered range
from 22% for females to 67% for children ages 6-11 and 72% for black,
non-hispanics. For the age group 6-11, the minimum is 0.6 for Dose1
(Mage et al (2004)) and the maximum is 1.1 for Dose5 (Shafer et al
(2004) with actual body weight). This difference is mainly because the
Mage et al (2004) approach was based on data from adults and was
intended to be applied to adults 18 and older. The Mage et al (2007)
paper extends the analysis to children and these revised equations give
a mean dose of 0.9 for children ages 6-11. For the black, non-hispanic
group, the minimum estimated dose is 1.07 ug/kg-day derived from  Mage
et al (2004) while the maximum estimated dose  is 1.83 ug/kg-day from
the Schafer et al (2004) Dose 4 approach. Similar findings apply to the
geometric means and medians. For all subjects the geometric means range
from 0.22 to 0.27 (25% difference) and the medians range from 0.15 to
0.20 (29% difference). The biggest differences are for the black,
non-hispanic subpopulation where the geometric means differ by 57% and
the medians differ by 70%. In general the lowest estimated doses are for
the three Mage et al. methods (Dose1-Dose3), and the highest estimated
dose is for the Schafer et al (2004) Dose4 method, with the ORD-type
method (Dose6) providing  intermediate estimates.    

The 99th percentile dose estimates show even larger percentage
differences. For all subjects, the 99th percentiles range from 16 (Mage
et al, 2007) to 27 (Schafer et al, 2004), a 75% difference. The biggest
demographic group differences are 243% (7 to 25) for the 6-11 age group
and 200% (12 to 36) for black, non-hispanics.

The same comparisons are displayed graphically in Figures 1 to 13.
Figures 1 to 3 display all the mean, geometric mean, and 99th percentile
dose estimates, respectively, for each demographic group. The dose is
plotted on a linear scale. Figures 4 to 13 display the mean, geometric
mean, 90th, 95th, and 99th percentile dose estimates for three methods
on a logarithmic scale. Each of Figures 4 to 13 shows all the summary
statistics for a different demographic group. The three methods shown
are Mage et al (2007) with obesity correction (Dose3), Schafer et al
(2004) with actual bodyweights (Dose 5), and the Geigy Method – Mean
(Dose6).

As shown in Table 2 and Figures 1-3, the Geigy Method  – Upper Bound
dose estimate is the highest of the seven estimated doses. This is, of
course, expected, since this method was intended as an upper bound
method. The maximum 99th percentile estimated dose was 62.4 μg/kg-day
for the Mexican-American group. The maximum estimated dose for all
subjects was 138 μg/kg-day.   Table 2. Selected summary statistics for
estimated triclosan dose (ug/kg-day), by calculation method*, age,
gender, and race.







































	Mean	Geometric Mean	Median	99th Percentile

Group	1	2	3	4	5	6	7	1	2	3	4	5	6	7	1	2	3	4	5	6	7	1	2	3	4	5	6	7

All	1.5	1.4	1.4	1.9	1.6	1.6	2.4	0.2	0.2	0.2	0.3	0.2	0.2	0.3	0.2	0.2	0.2
0.2	0.2	0.2	0.3	16.8	15.5	15.5	27.2	23.6	23.6	38.1

6-11	0.6	0.9	0.9	1.0	1.1	0.9	1.3	0.2	0.2	0.2	0.2	0.2	0.2	0.3	0.1	0.1	0.1
0.2	0.1	0.1	0.2	7.2	10.9	10.9	10.3	24.6	9.7	14.2

12-19	1.5	1.4	1.4	2.3	1.7	2.2	3.4	0.2	0.2	0.2	0.3	0.2	0.3	0.4	0.2	0.1
0.1	0.2	0.2	0.2	0.3	20.1	16.6	16.6	42.0	25.5	28.8	46.5

20-59	1.7	1.6	1.5	2.0	1.7	1.6	2.6	0.3	0.3	0.3	0.3	0.2	0.2	0.4	0.2	0.2
0.2	0.2	0.2	0.2	0.3	20.6	19.1	19.1	28.1	29.1	29.9	48.3

>= 60	1.1	1.1	1.0	1.5	1.2	1.2	1.8	0.2	0.2	0.2	0.2	0.2	0.2	0.3	0.1	0.1
0.1	0.1	0.1	0.1	0.1	14.4	14.1	14.4	21.4	17.1	14.8	22.7

Male	1.9	1.8	1.7	2.4	2.0	2.0	3.0	0.3	0.3	0.3	0.4	0.3	0.3	0.4	0.2	0.2	0.2
0.3	0.2	0.2	0.3	20.6	19.0	19.0	36.0	35.1	35.2	54.1

Female	1.1	1.1	1.1	1.3	1.1	1.2	1.9	0.2	0.2	0.2	0.2	0.2	0.2	0.3	0.1	0.1
0.1	0.2	0.1	0.1	0.2	14.2	14.9	14.7	23.4	17.8	17.6	28.5

Mexican-American	2.0	2.0	1.9	2.6	2.3	2.2	3.5	0.3	0.3	0.2	0.3	0.3	0.3	0.4
0.2	0.2	0.1	0.2	0.2	0.2	0.3	26.0	24.0	20.6	44.4	42.4	40.6	62.4

White, Non-Hispanic	1.5	1.4	1.4	1.8	1.5	1.5	2.3	0.3	0.2	0.2	0.3	0.2	0.2
0.3	0.2	0.2	0.2	0.2	0.2	0.2	0.2	15.0	15.0	15.0	23.4	16.3	19.0	29.1

Black, Non-Hispanic	1.1	1.2	1.1	1.8	1.6	1.5	2.3	0.2	0.2	0.2	0.3	0.2	0.2
0.4	0.2	0.2	0.1	0.2	0.2	0.2	0.3	12.1	14.0	13.7	36.3	26.1	28.3	45.6































* Calculation Method:





























1. Mage et al (2004).





























2. Mage et al (2007), without obesity correction.



























	3. Mage et al (2007), with obesity correction.



























	4. Schafer et al (2004).





























5. Schafer et al (2004) with actual bodyweight.





























6. Geigy Method - Mean.





























7. Geigy Method – Upper Bound.































 tc "Gplot " \f C \l 1  tc "Plot of y * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of y * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of y * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of z * x = dosenum " \f C \l 2 

 tc "Gplot " \f C \l 1  tc "Plot of y * x = dosenum " \f C \l 2 

CONCLUSION

NHANES urinary metabolite concentration data are not collected in a way
to directly determine the dose, and CDC has not reported dose estimates
for triclosan based on NHANES measurement data. The conversion of
measured spot urine concentrations to daily doses is difficult because
of highly variable dilution caused by wide fluctuations in the fluid
intake and excretion. Dose calculation is also difficult because there
is no way to determine from the NHANES data how (oral, dermal,
inhalation) and when (duration and time interval prior to measurement)
the actual contact with the chemical occurred, and because of
uncertainty and variability in the absorption, distribution, metabolism,
and excretion (ADME) parameters. We have presented several alternative
methods of estimating human dose (μg/kg-day) from a spot urine
concentration (μg/L) of an analyte and applied these methods to NHANES
2003-2004 triclosan concentrations. The analyses presented in this
document assumes that total triclosan is fully excreted in the urine.
Therefore, the use of these analyses for risk assessment purposes will
require correction for human pharmacokinetics.

There are essentially two main approaches for this conversion. The
creatinine adjustment approach of Mage et al (2004, 2007) uses an
estimate of the daily creatinine excretion (L/day) for each demographic
group and applies this to the analyte/creatinine concentration ratio.
The urine volume adjustment approach using the methods of Schafer et al
(2004) or the ORD-type methods (as used by some ORD researchers in some
situations) applies the estimated daily urine volume in L/day or
L/kg-day, respectively, to the analyte concentration. Within these two
main approaches, different versions use different sets of equations or
tables for the estimated daily creatinine or urine excretion by
demographic group. In general the two main approaches give similar
results for the mean, median, and geometric mean. The differences
between the two main approaches are more pronounced at the 90th and
higher percentiles. There are smaller differences between the different
versions of these approaches. The ORD-type method is preferable to the
Schafer et al (2004) method since urine volumes can be expected to
increase with increased body size. For triclosan, all subjects, the
means are up to 35% higher for the urine volume adjustment methods
compared to the Mage methods. The 99th percentile is up to 75% higher.

The Mage et al approach requires numerous assumptions for its validity: 

The subjects in the creatinine studies are representative of the
demographic groups.

The spot sample analyte/creatinine ratio is representative of the
24-hour average.

Subjects have reached steady state, equilibrium conditions with respect
to creatinine and analyte urinary concentrations. 

最왾氀჆

最왾氀჆

最왾氀჆

最왾氀჆

最왾氀჆

hÝ%

hÝ%

h

h

h demographic group, the absorption, distribution, metabolism, and
excretion (ADME) parameters are the same across all individuals and are
constant within individuals over time.

Subjects are omnivorous, neither eating excessive red meat nor being
strictly vegetarian.

Subjects are moderately active. 

Subjects are healthy. 

Subjects do not exceed the analyte’s tubular secretion transport
maximum. 

Subjects are not on medicines such as penicillin that reduce creatinine
tubular secretion.

Subjects have not been significantly exposed to other contaminants.

Subjects are not taking creatinine dietary supplements.

Subjects have moderate muscle mass.

Subjects do not have febrile disorders.

Subjects do not have decreased renal function (e.g., caused by GFR or
tubular disorders).

Female subjects are not pregnant.

Of these assumptions the steady state assumption, the related assumption
that the spot sample analyte/creatinine concentration ratio is
representative of the concentration ratios over the 24-hour period,  and
the assumption about not reaching the tubular secretion transport
maximum are difficult, if not impossible, to verify since there are no
NHANES data available to check these assumptions. In particular, the
recent paper by Scher et al (2007) throws some doubt on the
representativeness of spot urine samples. Using NHANES data, some of the
other assumptions can be evaluated, and, if creatinine excretion data
are available, further adjustments can be made for those subjects. We
note that a recent presentation by Dr. Allen discusses an additional 20%
adjustment for pregnancy. The Mage et al approach is also relatively
complex.

The urine volume adjustment approach is simpler. The main assumptions
are that the daily urine excretion is constant across each demographic
group, and that absorption, distribution, metabolism, and excretion
(ADME) parameters are the same across all individuals in a demographic
group and are constant within individuals over time. The method also
makes a steady state assumption and the related assumption that the spot
sample analyte concentration is representative of the concentration over
the 24-hour period. As applied by Schafer et al (2004) or by our
analyses of the ORD-type method using the Geigy (1981) tables, the
demographic groups appear to be too broad and would benefit from a large
set of urine excretion data by demographic group so that the urine
volume adjustment approach could be refined and more accurately applied.
Ideally, this data would contain information on both intra- and
inter-person variability in 24 hour urinary volumes so the issue of
whether the urinary volume or creatinine excretion is more stable could
be addressed.   Mage et al (2007) point out some disadvantages of the
urine volume adjustment approach. They note that the creatinine
excretion rates are less variable than urine excretion rates for the
same demographic groups, and also that the urine volume adjustment
approach does not adjust for the within-person variability of urine
excretion such as due to increased fluid intake in hot weather. Since
Mage et al. normalize the analyte concentration by the creatinine
concentration, the Mage et al approach better adjusts for the
within-person variability.  This assumes that (i) the Mage creatiine
excretion equation more accurately predicts creatine excretion; and (ii)
that intra-individual (e.g, day-to-day)  percent variability in creatine
 excretion is smaller than the corresponding figure for urine volume.
This could be addressed by looking at the intra-class correlation
coefficient, if data are available.

If the triclosan results hold more generally, the urine volume
adjustment approach results in estimates of dose that are higher than
the Mage et al creatinine adjustment approach.  Therefore,  the urine
volume adjustment approach is more conservative (i.e., more protective
of human health).  A more conservative approach may be better given the
uncertainties in the dose conversion from spot urine samples. However,
it is not clear whether the same finding would apply to other analytes.

Different readers will reach different conclusions about which method to
recommend. My own feeling at this time is that the ORD-type method based
on urine volume in L/kg-day is preferred for routine analyses, since it
is simpler to apply and explain, requires fewer assumptions for its
validity, and may well be more conservative.  The crucial issue is to
decide which is a more accurate estimator for an given individual :  a
L/kg-day estimate based on the age group or of the individual of
interest or a gram creatinine excretion rate as estimated for that
individual using the Mage et al (2007) equation. Also of value is to
decide how consistent is within individual (day-to-day) variability of
L/kg urine volume vs. day-to-day variability in creatinine excretion.  

However, for routine application of the ORD-type method it is necessary
to find a compendium of urine excretion data for several demographic
groups (by age, gender, race), since the Snyder et al (1975) and Geigy
(1981) urine excretion data are more generalized. For less routine
analyses, the Mage et al (2007) method can be used, but the results
should be compared to the ORD-type method. 

ACKNOWLEDGEMENT

This note acknowledges the comments and editorial suggestions from the
EPA OPP and ORD reviewers of the first draft. In several cases I have
used their suggested additional text in this revised draft. 

REFERENCES

Calafat A.M., Xiaoyun, Y., Wong, L.-Y., Reidy, J.A., Needham, L.L. 2007.
Urinary concentrations of triclosan in the US population: 2003-2004.
Environ Health Perspect.   HYPERLINK
"http://dx.doi.org/doi:10.1289/ehp.10768" 
http://dx.doi.org/doi:10.1289/ehp.10768 . Online December 2007.  

CDC Third National Report on Human Exposure to Environmental Chemicals.
2005. 

Cockcroft D.W., Gault M.H. 1976. Prediction of creatinine clearance from
serum creatinine, Nephron 16:31-41.

Geigy. 1981. Geigy Scientific Tables, Volume 1. Units of measurement,
body fluids, composition of the body, nutrition. Eighth edition. (Edited
by C. Lentner). CIBA-GEIGY.

Korn E.L., Graubard B.I. 1999. Analysis of Health Surveys. Wiley.

Mage D.T., Allen R., Gondy G., Smith W., Barr D.B., Needham L.L. 2004.
Estimating Pesticide Dose from Pesticide Exposure Data by Creatinine
Correction in the Third National Health and Nutrition Examination Survey
(NHANES-III). J Exposure Anal Environ Epidemiol 14:457-465.

Mage D.T., Allen, R.H., Kodali, A. 2007. Creatinine corrections for
estimating children’s and adult’s pesticide intake doses in
equilibrium with urinary pesticide and creatinine concentrations. J
Exposure Scil Environ Epidemiol 1-9.  

Mattsson S., Lindstrom S. 1995. Diuresis and voiding pattern in healthy
schoolchildren. Br J Urol. 76:783-789.

Morgan, Marsha K, Sheldon, L.S. ,Thomas, K.W. , Egeghy, P.P. ,Croghan,
C.W. , Jones, P.A., Chuang, J.C.  and Wilson N.K. 2008. Adult and
children's exposure to 2,4-D from multiple sources and pathways, J. Exp.
Sci. Environ. Epidemiol.   Advanced online publication 2 January 2008;
doi: 10.1038/sj.jes.7500641 .  Available at: 
http://www.nature.com/jes/journal/vaop/ncurrent/pdf/7500641a.pdf

Remer T., Neubert A., Maser-Gluth C. 2002. Anthropometry-based reference
values for 24-h urinary creatinine excretion during growth and their use
in endocrine and nutritional research. Am J Clin Nutr. 75:561-569.

Scher D.P., Alexander B.H., Adgate J.L., Eberly L.E., Mandel J.S., and

Acquavella J.F., et al. 2007. Agreement of pesticide biomarkers between
morning

void and 24-h urine samples from farmers and their children. J Expo Sci

Environ Epidemiol. 17(4): 350–357.

 

Viteri F.E., Alvarado J. 1970. The creatinine height index: its use in
the estimation of the degree of protein depletion and repletion in
protein calorie malnourished children. Pediatrics 46:696-706.

Schafer, K.S., Reeves, M., Spitzer, S., Kegley, S. E. 2004. Chemical
Trespass: Pesticides in Our Bodies and Corporate Accountability. 
Pesticide Action Network North America. May 2004.

Snyder, W.S., Cook M.J., Nasset E.S., Karhausen, L.R., Howells, G.P.,
Tipton, I.H. 1975. International Commission on Radiological Protection,
No. 23. Report of the Task Group on Reference Man. Oxford, Pergamon
Press.        

	

		Page   PAGE  2  of   NUMPAGES  30 

		Dose Conversion

ICF International

  www. icfi.com

	

		Page   PAGE  28  of   NUMPAGES  30 

		Dose Conversion

ICF International

