Document 7r6aEMj2RBLMeZVyBQ8DYVEE
Corporate Occupational Medicine
3M Center, Building 220-3W-05 St. Paul, MN 55144-1000 651 737 4230 Telephone 651 733 9066 Fax
6. An Epidemiologic Investigation of Reproductive Hormones in Men with Occupational Exposure to Perfluorooctanoic Acid
This published paper by Olsen et al (J Occup Env Med 1998;40:614-622) further examined the observation initially reported by Gilliland in his doctoral thesis that serum PFOA levels (as measured by total organic fluorine) may be associated with reproductive hormone changes. In 1993 and 1995, the fluorochemical medical surveillance program had 111 and 80 male production workers (Cottage Grove, MN) voluntarily participate. Unlike the 1990 program which analyzed for serum total organic fluorine, serum PFOA was measured in 1993 and 1995 by mass spectrometry methods. Serum PFOA levels were then compared to several reproductive hormones. Serum PFOA was not significantly associated with estradiol or testosterone. A 10% increase in mean estradiol levels was observed among employees who had the highest levels of serum PFOA although this association was confounded by body mass index. Neither was PFOA consistently associated with any of the other measured hormones.
003467
614 Epidemiologic Study of Perfluorooctanoic Acid Olsen et al
An Epidemiologic Investigation o f Reproductive Hormones in Men with Occupational Exposure to Perfluorooctanoic Acid
Geary W. Olsen, DVM, PhD Frank D. Gilliland, MD, PhD Michele M. Burlew, MS
A lthough fluoride (inorganic ionic flu oride) was identified in human blood 140 years ago,1'2 the presence of fluorine in a covalently bound or
Jean M. Burris, MPH
ganic state was first reported in
Jack S. Mandel, PhD Jeffrey H. Mandel, MD
1968.3-4 Guy subsequently identi fied perfluorooctanoic acid (PFOA, C7F |5C 0 2H) as a major component
Perfluorooctanoic acid (PFOA), a potent synthetic surfactant used in industrial applications, is a peroxisome proliferator that has resulted in dose-related increases in hepatic, pancreatic acinar, and Leydig cell adenomas in laboratory animals. In addition, PFOA increased serum estradiol levels through the induction of hepatic aromatase activity. In
of the serum organic fluorine frac tion.5 Ammonium perfluorooctanoate, a potent synthetic surfactant used in industrial applications, rapidly dis sociates in aqueous solution to PFOA.
1993 and 1 995, we conducted two cross-sectional studies o f 111 and 80
In laboratory animals, PFOA acid,
production workers, respectively, an d specifically measured their serum or its salts, is absorbed by ingestion,
PFOA in relation to several reproductive hormones to determine whether inhalation, or dermal exposure6-8
such an effect occurs in humans. PFOA was not significantly associated and is not metabolized.9-12 PFOA is
with estradiol or testosterone in either year's study. A 1 0 % increase in mean estradiol levels was observed among employees who had the highest levels o f serum PFOA, although this association was confounded by body mass index. Neither was PFOA consistently associated with the other measured hormones. O ur results provide reasonable assurance that, in this production setting, there were no significant hormonal changes associated with PFOA a t the serum levels measured. Limita tions o f this investigation include its cross-sectional design, the feiv subjects exposed at the highest levels, and the lower levels o f serum PFOA measured, compared with those levels reported to cause effects in
distributed primarily in the plasma and liver of male rats and the liver, plasma, and kidney in female rats." The major route of elimination is via urine and feces. In the female rat, there is a tenfold-greater renal ex cretion rate.111314 Castrated male rats treated with estradiol have PFOA urinary excretion rates similar to those of female rats.10 "
Peroxisome proliferators, like
laboratory anim al studies.
PFOA, are a diverse class of chemi
cals that cause hepatic peroxisome
proliferation and enzyme induction,
liver hyperplasia, and, in some in
From (he Medical Department. 3M Company. St. Paul. Minn. (Dr Olsen. Ms Burlew. Ms Burris. Dr J.S. Mandel): the Epidemiology and Cancer Control Program. School of Medicine. University of New Mexico. Albuquerque. NM (Dr Gilliland); and the Division of Environmental and Occupational Health.
stances, hepatocarcinogenesis in rats and mice.15-21 Two-year feeding studies in CrlrCD BR (CD) rats at a
School of Public Health. University of Minnesota. Minneapolis. Minn. (Dr J.H. Mandel). Address correspondence to. Geary Olsen, DVM. PhD. Medical Department. 3M. 220-3W-05. St.
Paul. MN 55144. 107b- 2752/98/4007-0614S3.00/0 Copyright <D by American College of Occupational and Environmental Medicine
maximum amount of 300 parts per million (ppm) PFOA showed liver adenomas and an increased inci dence of pancreas acinar cell adeno-
C03468
JOEM Volume 40, Number 7, July 1998
615
mas2* 21 and Leydig cell adeno mas.17,22 PFOA is not mutagenic6 and thus the induction of these tu mors most likely occurs through nongenotoxic mechanisms such as oxidative stress.'9,24
To determine whether the Leydig cell adenomas were the result of an endocrine-related mechanism. Cook et al18 gavaged CD rats for 14 days with up to 50 mg/kg of ammonium perfluorooctanoate. A significant in crease in serum estradiol and de crease in testosterone levels were observed. The estradiol increase may be due to an induction of hepatic aromatase activity.18The decrease in serum testosterone levels might be the result of reduced conversion of 17-alpha hydroxyprogesterone (17HP) to androstenedione (via the in hibition of the C-17,20 lyase en zyme). However, Biegel et al19 were unable to replicate the negative tes tosterone association.
CD rats fed 100 ppm PFOA for a maximum of 13 weeks showed in creased estradiol but not testosterone levels.25'26 Elevated estradiol levels were found among CD rats fed 300 ppm during a 2-year bioassay, with no dose-related differences for tes tosterone, luteinizing hormone (LH), and follicle-stimulating hormone (FSH).22
The 3M Company has conducted medical surveillance of employees involved in PFOA production. Lev els of serum organic fluorine (1.0071.00 ppm) in production employees were 10- to 50-fold greater than val ues (0.01-0.13 ppm) reported from human sera.27 Since PFOA could be an endocrine modulator,18 a crosssectional study among workers with potential exposure to PFOA was conducted in 1990.28 Among 115 men engaged in PFOA production, total serum fluorine levels (sodium biphenyl extraction m ethod29) ranged from 0 to 26 ppm (mean = 3.27 ppm; standard deviation [SD] = 4.68 ppm). It had been estimated that 80%-90% of human total serum flu orine levels consisted of PFOA.27 Adjusting for potential covariates
and hormones of a priori interest resulted in a positive nonlinear (qua dratic) association with estradiol, positive linear associations with pro lactin and TSH, and negative nonlin ear (square root) associations with free or bound testosterone.28
A mortality study of employees in the chemical division, which in cluded the PFOA production build ings, found no significantly in creased cause-specific standardized mortality ratio for either male or female employees.50There were four deaths from prostate cancer, com pared with 1.97 expected (95% con fidence interval [Cl], 0.55-4.59). Only one employee had worked di rectly in the PFOA production build ings. An association between PFOA exposure and prostate cancer was considered biologically plausible based on the animal and human data, which showed associations between PFOA and reproductive hormones.
The purpose of this report is to describe the results from two crosssectional studies from the same plant that were done in 1993 and 1995.
Methods
PFOA Production
PFOA production at this plant be gan in 1947. PFOA, a white powder, is produced by an electrochemical process.31 The products of this elec trolysis cell reaction are highly fluorinated compounds, with the endproduct defined by the starting material. Production involves a fourstage process: isolating and convert ing the chemical to a salt slurry, converting the slurry to a salt cake, drying the cake, and packaging. The greatest likelihood for exposure to PFOA occurred in the drying area.
Subject Selection
General medical surveillance is performed biennially for employees at this plant. There were 111 male employees in 1993 and 80 male em ployees in 1995 who participated in medical surveillance, hormone test ing, and serum PFOA determination.
Sixty-eight employees participated in both years. The surveillance con sisted of a medical questionnaire: measurement of height, weight, and pulmonary function: standard bio chemical and urinalysis tests: PFOA determination; and several hormone assays.
PFOA Analysis
A thermospray mass spectropho tometry assay was used to determine serum PFOA levels in 1993 and 1995.32 The range of serum PFOA was 0 to 80 ppm in 1993 and 0 to 115 ppm in 1995. The upper limit of detection in 1993 was 80 ppm, whereas there was no upper limit of detection in 1995. Levels were highly correlated among the 68 em ployees studied in 1993 and 1995 (r = .91, P = 0.0001). There was also high correlation between total serum fluorine level measured with the 1990 study28 and the PFOA mea sured in 1993 (r = .72, P = 0.0001, n = 94 subjects) and in 1995 (r = .84, P = 0.0001, n - 63 subjects). These findings were not unexpected, because of the estimated 18- to 24month half-life of PFOA in hu mans.27
Hormone Assays
Serum samples were analyzed by the University of Minnesota's Endo crinology Laboratory (Minneapolis, MN) or the Endocrine Science Ref erence Laboratory (Tarzana, CA). Eleven hormones were assayed: cor tisol, dehydroepiandrosterone sulfate (DHEAS), estradiol, FSH, 17ahydroxyprogesterone (17-HP), free testosterone, total testosterone, LH. prolactin, thyroid-stimulating hor mone (TSH) and sex hormone binding globulin (SHBG). All but SHBG were analyzed at the Univer sity of Minnesota's Endocrinology Laboratory.
Cortisol was assayed using a fluo rescence polarization immunoassay (Abbott TDx, North Chicago. 1L). Radioimmunoassays (RIA) were used for DHEAS (Pantex, Santa Monica, CA), estradiol (modified
003469
616
Pantex). I7-HP (modified CIS) and total testosterone (Coat-A Count; Di agnostic Product Corp.. Los Angeles, CA). Free testosterone was deter mined using equilibrium dialysis. LH. FSH and prolactin were assayed using a microparticle enzyme immu noassay (Abbott Imx). TSH was de termined using a chemiluminescence immunometric assay (Nichols, San Juan Capistrano, CA). SHBG was assessed via a radioimmunoassay af ter chromatographic sample purifica tion (Endocrine Science Reference Laboratory). Bound testosterone was calculated as total testosterone less free testosterone. The same assays were used for both 1993 and 1995 analyses.
Epidemiologie Study of Porfluoroootanoie Acid Olson st al
comparing parameter estimates, using indicator and continuous variables. Stepwise selection procedures were also used. Study results were analyzed by the SAS System."
We did not examine hormone changes between the two examina tions because of the estimated halflife of PFOA (approximately two years) and intraindividual variability in hormones. Since the results for the 68 employees who participated in both years were similar to those ob tained for the entire study, only the results for all employees are pre sented below.
Results
Serum PFOA levels were not highly correlated with either the co variates or the hormones. These cor relation coefficients (in parentheses) for 1993 and 1995 data, respectively, for PFOA and the variable of interest were the following; age (-.2 2 , .14); alcohol (.10, .18); BMI (.11, .10); cigarettes (.05, .11); cortisol (.07. -.05); DHEAS (.13, .12); estradiol (.12, .15); FSH (-.1 2 , -.13); 17-HP (.11, .30); LH (-.0 6 , .13); prolactin (.04, -.0 4 ); SHBG (-.0 7 , .03); bound testosterone (.01, .02); free
1993 n=111
Data Analysis
Simple and stratified analyses, anal ysis of variance (ANOVA), Pearson correlation coefficients, and ordinary multivariable regression were used to evaluate associations between PFOA and each hormone, with adjustment for potential confounding variables. For stratified analyses, employees were di vided into four PFOA categories: 0-1 ppm, 1-< 10 ppm, 10-<30 ppm, and >30 ppm in order to determine if an effect existed at the highest serum levels. For multivariable evaluation, PFOA, age, body mass index (BMI), alcohol use, and cigarette use were examined as both categorical and con tinuous variables. Alcohol use was an alyzed as less than 1drink per day, 1-3 drinks per day, and non-response to the questionnaire item. Cigarette use was recorded as either current smoker or nonsmoker. Regression models were fitted with PFOA entered as a contin uous variable, using linear, square, and square root transformations in order to assure that associations were not missed. The possible nonlinear associ ation of estradiol, free testosterone, and bound testosterone was evaluated. Nonlinear dose-response relationships were examined by model fit and by
1995 n=80
Fig. 1. Scatterplot of serum estradiol (pg/ml) by perfluorooctanoic acid (PFOA. in parts per million [ppm]) for employees in I993 and I995.
003470
JOEM Volume 40, Number 7, July 1998
617
1993 nrlil
PFOA 1995 n=80
Fifl. 2. Scatterplot of serum free testosterone (ng/dl) by PFOA (ppm) for employees in 1993 and 1995.
testosterone (.09, .01); and TSH (.03, .15).
Figures 1 and 2 are scatterplots for estradiol and free testosterone for each year. Simple linear regression of the natural log of each hormone with PFOA, treated as a continuous variable, resulted in no statistically significant coefficients in 1993 for any hormone and only one in 1995: 17-HP (beta coefficient = 0.006, P = 0.03, R2 = .06). This result was dependent upon one person. In 1995, this person had a level of 198 ng/dL of 17-HP and 114 ppm of serum PFOA. In 1993, this person's values
were 206 ng/dL of 17-HP and 80 ppm (upper limit of detection in 1993) for PFOA.
Table 1 provides the mean, me dian, standard deviation, and range of the covariates and several hor mones, by four levels of PFOA cat egorization (0 -< l, l-< 10, 10-< 30, and 3 0 ppm). Seventy-five percent of the employees with serum PFOA levels at 10 ppm or greater partici pated in both years. From Table 1, several observations are noteworthy. First, the mean of the PFOA ppm categories differed by two orders of magnitude between the lowest and
highest categories for both years. Second, the 30-ppm PFOA cate gory had the youngest mean em ployee age in both years. Third. BMI was the greatest among employees in the 30-ppm PFOA category in 1995. Fourth, mean estradiol levels were not significantly different be tween PFOA levels in either year, although the 30-ppm PFOA cate gories had mean estradiol levels that were 10% higher than the other PFOA levels. Fifth, there were no discernible trends between PFOA and either bound or free testosterone. Sixth, 17-HP levels were highest in the 30-ppm PFOA group in both years. No significant associations were observed for cortisol. DHEAS, FSH, LH, and SHBG (data not shown).
As expected,34 estradiol was highly correlated with BMI (1993: r = .41, P < 0.001; 1995: r = .30, P < 0.01) and free testosterone with age (1993: r = -.4 8 , P < 0.001; 1995: r = -.4 0 , P < 0.001); thus Table 2 provides mean estradiol and free testosterone values stratified by BMI and age, respectively. It should be noted that all five employees in 1995 with serum PFOA levels 30 ppm had BMIs 28.
Linear and nonlinear relationships, taking into account potential confounders (especially age and BMI) as well as other covariates that may be on the biologic pathway of effect, resulted in no significant associa tions with PFOA except for 17-HP in the 1995 analyses (data not shown). Again, this association was depen dent on the one employee discussed earlier.
Because a primary hypothesis of the present study was whether PFOA increased estradiol and decreased testosterone serum levels in a nonlin ear fashion, we replicated these prior models28 with our 1993 and 1995 data. PFOA was not significantly associated with serum estradiol, free testosterone, or bound testosterone (data not shown). There was no sig nificant association (data not shown) between PFOA and prolactin among
003471
618 Epidemiologie Study of Perfluorooctanoic Acid Olsen et el
TABLE 1
Mean, Median, Standard Deviation (SD) of Mean and Range of Perfluorooctanoic Acid (PFOA), Demographic and Hormonal
Values by Serum PFOA Levels, and Year of Data Collection*
1993 Data
1995 Data
PFOA (ppm)
PFOA (ppm) 0-<1* 1-<10 10-<30 a30
Mean
Median
SO
Range
Mean
Median
SD
Range
0.48*
0.47
3.34*
2.49
16.26*
15.40
60.13*
64.45
F = 253.25, P = 0.0001
0.27 2.17 3.39 24.01
0.00-0.99 1.03-8.92 11.90-21.00 31.60-80.00
0.31 0.2
3.03 2.4
17.11*
14.3
55.96*
42.4
F = 77.57, P = 0.0001
0.32 1.84 6.90 33.29
0.00-0.90 1.10-8.20 10.30-28.20 34.20-114.10
Age (yr) 0-<1 1-<10 1 0 -0 0 230
43.6 45.0
39.2 38.0
39.9
39.5
33.3
32.5
F = 3.67, P = 0.01
9.2 27.0-61.0
42.0
41.0
7.7 27.0-60.0
41.3
40.0
4.2 34.0-45.0
45.1
46.0
7.4 25.0-43.0
38.2
35.0
F = 0.88, P = 0.46
8.3 29.0-60.0 8.6 24.0-58.0 7.4 30.0-55.0 9.2 27.0-50.0
Alcohol (drinks/day) 0-<1 1-<10 1 0 -0 0 230
BMI (kg/m*) 0-<1 1-<10 1 0 -00 >30
Cigarettes (cigarettes/day) 0-<1 1-<10 1 0-00 230
Estradiol (pg/mL) 0-<1 1-<10 1 0 -00 230
17-HP (ng/dL) 0-<1 1-<10 1 0 -00 230
Prolactin (p.g/L) 0-<1 1-<10 1 0 -00 230
Bound testosterone (ng/dL) 0-<1 1-<10 1 0 -00 230
0.4 0.3 0.7 0.5 0.9 0.7 0.9 ' 0.7 F = 3.05, P = 0.03
28.0
27.5
26.8
26.3
29.1 28.8
28.5
28.4
F = 1.60, P = 0.19
2.6 0.0 6.0 0.0 2.5 0.0 5.0 0.0 F = 1.26, P = 0.29
54.7
53.0
56.0
55.0
54.8 58.0
62.8 63.0
F = 0.54, P = 0.66
106.8
106.0
120.2
115.5
97.9 105.5
126.5
123.0
F = 1.55, P = 0.21
8.2 8.0
8.8 8.0
15.0*
9.0
7.5 7.5
F = 3.67, P = 0.01
528.7
513.7
609.7
609.2
485.2
477.5
569.6
596.5
F = 2.48, P = 0.07
0.5 0.0-1.9
0.5
0.3 0.7 0.0-2.9
0.7 0.0-3.4
0.5
0.4 0.5 0.0-1.9
0.6 0.4-2.1 0.8 0.7 0.7 0.0-2.1
0.8 0.0-2.0
0.5
0.4 0.6 0.0-1.4
F = 0.94, P = 0.43
4.2 20.9-42.0
27.6
26.8
2.5 21.6-32.5
28.6
27.9
1.8 27.1-32.0
27.8
27.7
1.6 26.9-30.2
29.8
28.9
F = 0.77, P = 0.52
4.2 21.9-45.2 3.4 22.1-38.3 4.0 21.2-34.8 1.8 28.2-32.6
7.5 0.0-30.0
3.8
0.0 9.4 0.0-40.0
10.5 0.0-40.0
2.6
0.0 6.0 0.0-20.0
7.1 0.0-20.0
9.1
0.0 15.2
0.0-40.0
10.0 0.0-20.0
6.0
0.0 8.9 0.0-20.0
F = 1.26, P = 0.30
13.5 32.0-83.0
68.1
66.0 11.7 45.0-94.0
12.0 34.0-81.0
65.2
62.0 14.9 45.0-96.0
11.6 39.0-70.0
67.1
66.5 9.1 54.0-79.0
8.4 55.0-70.0
73.2
75.0 6.7 64.0-81.0
F = 0.69, P = 0.56
34.9 44.0-203.0
91.6
94.0 32.2 39.0-190.0
41.5 45.0-249.0
110.6
105.5
35.6 54.0-179.0
28.4 54.0-134.0
110.3
85.5 77.5 46.0-297.0
66.8 54.0-206.0
123.0
102.0
54.7 72.0-198.0
F = 1.67, P = 0.18
3.5 4.6 15.2 0.6
178.0 168.2 113.9 81.6
2.0-18.0 2.0-22.0 6.0-51.0 7.0-8.0
10.9 10.0 11.8 10.0 12.9 14.0
9.4 9.0 F = 0.66, P = 0.58
220.9-1059.5 212.2-1021.6 301.0-651.6 450.9-634.4
534.8
518.5
567.7
564.9
554.4
549.7
567.8
623.0
F = 0.26, P = 0.85
5.1 4.0-23.0 6.0 5.0-28.0 5.3 3.0-21.0 2.7 7.0-14.0
150.4 152.3 185.5 155.0
278.7-1059.5 216.4-898.4 238.1-823.8 341.7-703.0
003472
JOEM Volume 40, Number 7, July 1998
619
TABLE 1
Continued
1993 Data
1995 Data
OCO Al
PFOA (ppm)
Free testosterone (ng/dL) 0-<1 1-<10 10-<30 30
TSH (mU/L) 0-<1 1-C10 10-<30
Mean
Median
15.0 14.8 16.6 16.5 14.2 13.5 17.4 18.5 F = 1.81, F = 0.15
1.4 1.3 1.4 1.2 2.1 2.2 1.2 1.1 F = 2.21, P = 0.09
SD
4.0 4.4 2.5 4.7
0.8 0.7 0.8 0.4
Range
6.1-28.1 6.8-28.4 10.0-17,4 11.1-21.6
0.2-4.3 0.5-3.1 1.2-2.9 0.8-1.8
Mean
Median
14.2 14.0 14.2 14.4 13.2 14.1 14.4 16.0 F = 0.31, P = 0.82
1.7 1.5
1.7 1.5
2.9s
2.5
1.7 1.3
F = 5.47, P = 0.002
SD
2.7 3.1 3.5 4.3
0.8 0.9 1.1 0.6
* BMI, body mass index; 17-HP, 17-alpha hydroxyprogesterone; TSH, thyroid-stimulating hormone. T Samples sizes; 0 -< 1 ppm: 1993, n 53; 1995, n = 39. 1--<10 ppm: 1993, n = 46; 1995, n = 26. 10-<30 ppm: 1993, n = 8; 1995, n = 10. 30 ppm: 1993, n = 4; 1995; n = 5. * Mean significantly different (Bonferroni t-test, p < .05) than the three other PFOA ppm levels. * Mean level significantly different (Bonferroni t-test, p < .05) than the 0 -< 1 ppm and 1--<10 ppm PFOA categories.
Range
9.3-20.5 5.6-19.4 7.0-18.0 7.3-18.0
0.6-4.0 0.5-3.7 1.9-5.8 1.1-2.5
TABLE 2
Mean, Median, Standard Error (SE) of Mean and Range of Estradiol by Body Mass Index and Free Testosterone by Age,
Stratified by Serum PFOA Level and Year of Data Collection
BMI (kg/m*) by PFOA (ppm)
n Mean
1993 Data Median SE
Range
n Mean
1995 Data Median SE
Range
Estradiol (pg/mL) BMI <28 0--<1 ppm 1--<10 10-<30 30 BMI 28 0-1 ppm 1-<10 10-<30 30
Free testosterone (ng/dL) Age <40 0--<1 ppm 1-<10 1Q-<30 30 Age 40 0--<1 ppm 1-<10 10-<30 30
30 48.4 30 55.0
3 54.3 2 62.5
23 63.0 16 57.8
5 55.0 2 63.0
47.0 55.0 56.0 62.5
66.0 55.5 60.0 63.0
1.7 32.0-68.0 23 66.0
2.2 34.0-81.0 13 62.0
5.5 44.0-63.0
5 64.6
7.5 55.0-70.0
0--
2.9 32.0-83.0 16 71.1
3.1 34.0-79.0 13 68.3
6.1 39.0-70.0
5 69.6
7.0 56.0-70.0
5 73.2
20 17.3 28 16.8
4 15.2 3 19.5
33 13.6 18 16.2
4 13.2 1 11.1
16.8 17.4 14.9 20.1
13.4 15.8 13.1 11.1
0.9 10.5-28.1 0.7 10.1-24.4 1.0 13.4-17.4 1.4 16.9-21.6
18 15.3 13 14.7 2 15.7 3 15.9
0.6 6.1-21.2 21 13.2
1.3
6.8-28.4
13 13.7
1.4 10.0-16.6
8 12.6
----
2 12.2
66.0 62.0 64.0 --
72.0 65.0 72.0 75.0
2.2 48.0-87.0 3.7 48.0-91.0 4.1 54.0-79.0 ----
3.2 45.0-94.0 4.5 45.0-96.0 4.1 55.0-78.0 3.0 64.0-81.0
15.2 0.6 11.3-20.5 14.6 0.6 10.8-17.8 15.7 2.4 13.3-18.0 16.0 1.2 13.8-18.0
13.4 0.6 9.3-18.3 14.3 1.0 5.6-19.4 14.1 1.2 7.0-16.2 12.2 4.9 7.3-17.0
moderate drinkers, as was previously reported.28
Discussion We conducted two cross-sectional
studies of PFOA production workers
to investigate the relation between serum PFOA levels and several re productive hormones: in particular, estradiol and testosterone. Although we did not observe a significantly positive association between PFOA
exposure and estradiol, mean estra diol levels were 10% greater among employees with the highest serum PFOA levels (S30 ppm); however, this was confounded by BMI, and any interpretation is limited by the
003473
620 Epidemiologic Study of Perfluorooctanoic Acid Olsen et al
few subjects at this PFOA level. Gilliland also observed an approxi mate 10% increase in mean estradiol levels from his lowest (0-1 ppm) to highest (15-26 ppm) total serum or ganic fluorine levels among these production employees.28 Unlike the present study, this previous report also observed a significant nonlinear positive association between estra diol and total serum organic fluo rine.2i* Possible reasons for the dif ferent results include the following: (1) use of different measurements of exposure (total serum organic fluo rine in 1990 and serum PFOA in 1993 and 1995); (2) the possibility that the multivariate model used in 1990 may have transgressed the homoscedasticity assumption of regres sion analysis35; (3) possible misclassification of confounding variables (eg, the expected relationship be tween BMI and estradiol was not observed in 1990: correlation coef ficient = --.01); (4) different sub jects analyzed (94 employees partic ipated in both the original 1990 and the 1993 surveys, compared with 61 employees who participated in 1990 and 1995); and (5) differences in the estradiol assays.
Dose, threshold effect, and species sensitivity may account for the ap parent differences between the ani mal and human studies. We did not observe a significant association be tween estradiol and PFOA but did observe a 10% increase at the highest serum levels of PFOA. Serum PFOA in these workers was likely below the observable effect levels in animal studies: the observable effect level in the CD rat is somewhere above a mean serum level of 55 ppm PFOA.1825 All but three PFOA mea surements from employees in our study were below 55 ppm PFOA. The 10% increase in mean estradiol levels observed among employees with the highest levels of serum PFOA (^30 ppm) could suggest a threshold response. The discovery of the convergence of peroxisomal proliferators and estradiol at the level of their nuclear hormone receptors pro
vides a plausible mechanism for a possible threshold relationship be tween PFOA and estradiol.16 17 While responses to peroxisome pro liferators. like PFOA, are readily ob served in rats and mice, other spe cies-- including humans-- have shown no such responses to many types of peroxisome proliferators at equivalent dose levels.38-41
We did not observe any significant associations between PFOA and free or bound testosterone. However, we did observe a significant positive as sociation between 17-HP and PFOA in the 1995 but not 1993 analyses. We examined 17-HP, a precursor of testosterone, because Cook et al18 suggested that PFOA may affect the conversion of 17-HP to testosterone via inhibition of 17,20-lyase. If this enzyme was inhibited, the expected result would be an increase in 17-HP levels, which was observed in both years' studies, although significantly in only the 1995 study. Recent labo ratory work suggests that there may be an accommodation by the CD rat to the initial testosterone-lowering effect of PFOA.19 A previous re port28 observed a significant nega tive nonlinear association between total serum fluorine and free or bound testosterone. This observation was dependent upon one influential data point, that of an employee who had no detectable total serum organic fluorine level but had the highest free testosterone level measured.
Several methodological issues should be considered in evaluating the results from this study. First, the cross-sectional design does not allow for a direct analysis of the temporal ity of an association. Given the longhalf life of PFOA, it is conceivable that there may be some biological accommodation to the effects of PFOA, as suggested by Biegel et al.19 Second, the two cross-sectional analyses cannot be viewed as inde pendent populations because 68 em ployees were studied in both years. Fewer employees participated in se rum measurements in the 1995 as sessment, although the majority of
those with the highest serum PFOA exposure levels in 1993 also partici pated in 1995. This reduced sample size resulted in lower statistical power. Third, we specifically mea sured serum PFOA levels. Use of total serum organic fluorine may rep resent other perfluorocarbons, which could be peroxisome proliferators, although data suggest that PFOA would represent the greatest fraction of total serum organic fluorine levels in this employee population.27,29 Fourth, there could be measurement error in important confounding vari ables. Analysis of the 68 subjects who participated in both years showed good correlation for the con founding factors of BMI (r = .93, P = 0.0001) and the self-reported aspects of alcohol consumption (r = .67, P = 0.0001) and cigarette smok ing (r = .84, P = 0.0001). Fifth, the quality of medical surveillance data can be evaluated by whether known associations are observed.42 In this regard, we observed various ex pected associations (eg, estradiol and BMI, free testosterone and age). Fi nally, the pulsatile nature of some of the hormones studied (eg, FSH, LH, testosterone) has resulted in prior recommendations that mean hor mone measurements should be the result of pooled blood from multiple samples taken at short intervals.43 In our study, multiple samples were not feasible because of the low probabil ity of employees voluntarily giving three samples over a 45- to 60minute time period. In summary, we conducted two cross-sectional studies in 1993 and 1995 and did not observe a signifi cantly positive association between PFOA exposure and estradiol or a significantly negative association with testosterone. Our study may not have been sensitive enough to detect whether an association between PFOA and estradiol could exist in humans because measured serum PFOA levels were likely below the observable effect levels suggested in the animal studies. Our results pro vide reasonable assurance that sig-
003474
JOEM Volume 40, Number 7, July 1998
621
nificant hormonal changes among these male production employees were not apparent in relation to their measured serum PFOA levels.
Acknowledgments
The authors gratefully acknowledge the contributions of the following individuals to this study: Frances Curtis. Mary Fowler. Mary Hansen, and Drs James Johnson. Roger Perkins. James Wolter. and Larry Zobel.
References
1. Bryce H. Industrial and utilitarian aspects of fluorine chemistry. In: Simons J, ed. Fluorine Chemistry. New York: Aca demic Press; 1964:297-492.
2. Nickles M. Presence du fluor dans la sang. Compt Rend. 1856;43:885.
3. Taves D. Evidence that there are two forms of fluoride in human serum. N a ture. 1968;217:1050-1051.
4. Taves D. Electrophoretic mobility of se rum fluoride. Nature. 1968;220:582-583.
5. Guy W. Fluorocompounds o f Human Plasma: Analysis, Prevalence, Purifica tion. and Characterization. [Doctoral dis sertation.] Rochester, NY: University of Rochester; 1972.
6. Griffith FD, Long JE. Animal toxicity studies with ammonium perfluorooctanoate. Am Ind Hyg Assoc J. 1980;41:576583.
7. Kennedy G. Dermal toxicity of ammo nium perfluorooctanoate. Toxicol Appl Pharmacol. 1985;81:348-355.
8. Kennedy G, Hall G, Brittelli J, Chen H.. Inhalation toxicity of ammonium perfluo rooctanoate. Food Chem Toxicol. 1986; 24:1325-1329.
9. Ophaug R, Singer L. Metabolic handling of perfluorooctanoic acid in rats. Proc Soc Exp Biol Med. 1980;163:19-23.
10. Ylinen M. Koho A, Hanhijarvi H. Peura P Disposition of perfluorooctanoic acid in the rat after single and subchronic administration. Bull Environ Contam Toxicol. 1990;44:46-53.
11. Vanden Heuvel J. Kuslikis B. Van Refelghem M, Peterson R. Tissue distribution, metabolism and elimination o f perfluo rooctanoic acid. J Biochem Toxicol. 1991;6:83-92.
12. Kuslikis Bl. Vanden Heuvel JP. Peterson RE. Lack of evidence for perfluorodecanoyl- or perfluorooctanoyl-coenzyme A formation in male and female rats. J Biochem Toxicol. 1992:7:25-29.
13. Hanhijarvi H. Phaug R. Singer L. The sex-related difference in perfluorooclunoate excretion in the rat. Proc Soc Exp Bio! Med. 1982:171:51-55.
14. Hanhijarvi H. Ylinen M. Kojo A. Kosma VM. Elimination and toxicity of perfluo rooctanoic acid during subchronic admin istration in the Wistar rat. Pharmacol Toxicol. 1987:61:66-68.
15. Ikeda T. Aiba K. Fukuda K. Tanaka M. The induction of peroxisome prolifera tion in rat liver by perfluorinated fatty acids, metabolically inert derivatives of fatty acids. J Biochem. I985;98:475482.
16. Pastoor TP. Lee KP. Perri MA, Gillies PJ. Biochemical and morphological stud ies of ammonium perfluorooctanoateinduced hepatomegaly and peroxisome proliferation. Exp Mol Pathol. 1987;47: 98-109.
17. Sibinski LJ. Two-Year Oral fDiet) Toxiciry/Carcinogenicity Study o f Fluorochemical FC-143 in Rats. St. Paul, MN: Riker Laboratories; 1987.
18. Cook JC, Murray SM. Frame SR, Hunt ME. Induction of Leydig cell adenomas by ammonium perfluorooctanoate: a pos sible endocrine-related mechanism. Toxi col Appl Pharmacol. 192:113:209-217.
19. Biegel LB, Liu RCM, Hunt ME, Cook JC. Effects of ammonium perfluo rooctanoate on Leydig cell function: in vitro, in vivo, and ex vivo studies. Toxicol Appl Pharmacol. 1995;134:18-25.
20. Liu RCM, Hunt ME, Cook JC, Biegel LB. Effect of the peroxisome proliferator, ammonium perfluorooctanoate (C8), on hepatic aromatase activity in adult male CR1:CD BR (CD) rats. Fundam Appl Toxicol. 1996;30:220-228.
21. Lemberger T, Desvergne B, Wahli W. Peroxisome proliferator-activated recep tors: a nuclear receptor signaling pathway in lipid physiology. Ann Rev Cell Dev Biol. 1996;12:335-363.
22. Cook JC, Hunt ME, Frame SR, Biegel LB. Mechanisms of extrahepatic tumor induction by peroxisome proliferators in Crl:CD BR (CD) rats. Toxicologist. 1994; 14:301 (abstract).
23. Ogboum JD, Frame SR. Elliot GS, Cook JC. Pancreatic oncogenic effects of Wyeth 14,643. Toxicologist. 1997:36:232 (abstract).
24. Rao MS. Reddy JK. Hepatocarcinogenesis of peroxisome proliferators. Ann NY AcadSci. 1996;804:573-587.
25. Palazzolo MJ. 13 -Week Dietary Toxicity Study with T-5180. Ammonium Perfluo rooctanoate <CAS No. 3825-26-1), in Male Rats. Madison. Wl: Hazelton Lab oratories: 1991.
26. Perkins RG. Investigation of ammonium perfluorooctanoate effect on hormone levels and peroxisomal proliferation in the rat. Toxicologist. 1992:12:38 (ab stract ).
27. Ubel FA. Sorenson SD. Roach DE. Health status of plant workers exposed to fluorochemicals. a preliminary report. Am Ind Hvg Assoc J. 1980:41:584-589.
28. Gilliland FD. Fluorocarbons and Human Health: Studies in an Occupational Co hort. [Doctoral dissertation.] Minneapo lis. MN: University of Minnesota; 1992.
29. Pothapragada V. Sodium biphenyl method for determination of covalently bound fluorine in organic compounds and biological materials. Anal Chem. 1982: 54:1132-1137.
30. Gilliland FD. Mandel JS. Mortality among employees of a perfluorooctanoic acid production plant. J Occup Med. 1993;35:950-954.
31. Simons J. Bryce T. Electrochemical fluorination. In: Simons J, ed. Fluorine Chemistry. New York: Academic Press; 1954:340-377.
32. Johnson JD, Wolter JT, Colatzy GE, Rethwill PA, Nelson RM. Quantification of perflurooctanoate and perfluorooctanesulfonate in human serum using ion-pair extraction and high performance liquid chrom atography-therm ospray mass spectrometry with automated sam ple preparation. [3M Environmental Lab oratory Report.] St. Paul, MN: 3M Com pany; 1996.
33. SAS Institute, Inc. SAS Users Guide: Statistics. Version 6. Cary, NC: SAS Institute, Inc.; 1990.
34. Griffin JE, Wilson JD. Disorders of the testes. In: Isselbacker KJ, Braunwald E, Wilson JD, Martini JB, Fauci AS, Karper DL, eds. Harrison's Principles o f Inter nal Medicine. New York: McGraw-Hill; 1994:2006-2017.
35. Montgomery DC, Peck EA. Introduction to Linear Regression Analysis. New York: John Wiley and Sons. Inc.; 1992: 67-117.
36. Keller H. Dreyer C. Medin J. Mahfoudi A. Ozato K, Wahli W. Fatty acids and retinoids control lipid metabolism through activation of peroxisome prolif erator-activated receptor-retinoid X re ceptor heterodimers. Proc Natl Acad Sci. 1993;90:2160-2164.
37. Hunter J. Kassam A. Winrow CJ, Rachubinski RA, Capone JP. Crosstalk between the thyroid hormone and peroxisome pro liferator-activated receptors in regulating peroxisome proliferator-responsive genes. Mtd Cel Endocrinol. 1996:116: 2 13 -2 2 1.
38. Stringer DA. Hepatic Peroxisome Prolif eration. | Monograph No. 17. | Brussels: European Centre for Ecotoxicology and Toxicology of Chemicals (ECETOC):
1992.
003475
622 Epidemiologic Study of Porfluorooctanoic Acid Olson st al
39. De la Iglesia FA. McGuire ED, Haskins JR. Lalwani ND. Structural diversity of peroxisome proliferators and their effects on mammalian liver cells in vivo. Ann NY AcadSci. 1996;804:310-327
40. Tugwood JD. Aldridge TC, Lambe KG. MacDonald N, Woodyatt NJ. Peroxisome proliferator-activated receptors: struc tures and function. Ann NY Acad Sci.
1997;804:252-265. 41. Tugwood JD, Lambe KG. Woodyatt
NJ. Inter-individual differences in human peroxisome proliferator-activated recep tor alpha structure and function. Taxicologixt. 1997:36:236 (abstract). 42. Olsen GW. Kusch GD. Stafford BA, Gudmundsen SL. Currier MF. The posi tive known association design: a quality
assurance method for occupational health surveillance data. J Occup Med. 1991 ;33: 998-1000.
43. Goldzieher JW, Dozier TS, Smith KD. Steinberger E. Improving the diagnostic reliability of rapidly fluctuating plasma hormone levels by optimized multiple sampling techniques. J Clin Endocrinol Metab. 1976;43:824-830.
003476
