Serum Dioxin Concentrations and Risk of Uterine Leiomyoma in the Seveso Women's Health Study

Abstract

Uterine leiomyomata (fibroids), benign neoplasms of the smooth muscle, are a major cause of hysterectomy. Exposure to hormonally active chemicals may play an etiologic role. The authors investigated the risk of uterine leiomyoma associated with exposure to 2,3,7,8,-tetrachlorodibenzo-p-dioxin (TCDD) for women who resided near Seveso, Italy, in 1976 at the time of a chemical explosion. Twenty years later, women enrolled in the Seveso Women's Health Study were asked about history of fibroids, medical records were obtained, and vaginal ultrasonography was performed for a subset. Serum collected soon after the explosion was analyzed for TCDD. A likelihood-based method that combines both historical and current status (ultrasound) data was adapted to estimate the hazard ratio. Of 956 eligible women, 251 (26.3%) had fibroids. Compared with that for women with TCDD levels of ≤20 parts per trillion, the age-adjusted hazard ratios were 0.58 (95% confidence interval: 0.41, 0.81) for women with levels of 20.1–75.0 parts per trillion and 0.62 (95% confidence interval: 0.44, 0.89) for women with levels of >75.0 parts per trillion. This finding suggests that TCDD may have antiestrogenic effects in the uterine myometrium, in contrast to apparently estrogenic effects previously found in the breast of Seveso Women's Health Study women.

endocrine disruptors, leiomyoma, tetrachlorodibenzodioxin, uterus

Uterine leiomyomata, or fibroids, are benign neoplasms of uterine smooth muscle. Fibroids are associated with menorrhagia, pelvic pain, spontaneous abortion, and infertility, and they account for one third of all hysterectomies in the United States (1, 2). Although their frequency can be more than 70 percent in reproductive-age women (3, 4), the etiology of fibroids is largely unknown (5). Development and growth of fibroids appear to be influenced by ovarian hormones since they appear during the reproductive years, can regress after menopause, and are responsive to gonadotropin-releasing hormone agonists, which induce a hypoestrogenic state (5).

Fibroids have been associated with exposure to exogenous estrogens, such as diethylstilbestrol, in animals (6–8). About 65 percent of Eker rats spontaneously develop uterine leiomyoma (5). In the Eker rat model, xenoestrogens stimulate proliferation of uterine myometrial cells. The effect is dampened by administration of the antiestrogen ICI-182 780 (9). These xenoestrogens also stimulate transcription via the estrogen receptor and up-regulate expression of an estrogen-responsive gene, the progesterone receptor, in uterine leiomyoma cells. Selective estrogen receptor modulators (e.g., tamoxifen) that bind the estrogen receptor reduce the incidence of leiomyomas in Eker rats (10), suggesting that uterine myometrium and mammary tissue may have similar profiles of estrogen responsiveness.

In humans, one cohort study of randomly selected women in a prepaid health plan found a positive association between diethylstilbestrol exposure in utero and fibroids (11), but another study (12), following women exposed and not exposed to diethylstilbestrol, found no association. Concentrations of a different group of exogenous endocrine active chemicals, 2,2-bis-(p-chlorophenyl)-1,1,1-trichloroethane and its metabolites, were higher in human leiomyomatous tissue than in normal uterine tissue (13) and also in serum of women with fibroids compared with women without (14).

Another chemical known to disrupt multiple endocrine pathways is 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) (15). This compound has been shown to induce antiestrogenic responses in rat uterine and breast cells (16, 17) but also to induce an estrogen-like gene expression profile in the uteri of immature ovariectomized mice (18). The opposing actions of TCDD, antiestrogenic in the presence of estrogen and estrogenic in its absence, suggest that the effects of TCDD may vary depending on developmental stage at exposure (18).

On July 10, 1976, as a result of a chemical explosion, residents of Seveso, Italy, experienced the highest levels of TCDD exposure in a human population. Twenty years later, we initiated the Seveso Women's Health Study, a retrospective cohort study, to investigate the impact of TCDD exposure on women's health. We have previously shown that high TCDD exposure was associated with a doubling in risk of breast cancer (19). In the present analysis, we examined the association between serum TCDD levels and occurrence of uterine fibroids. We hypothesized that if TCDD acts as an antiestrogen in uterine myometrium, higher levels of TCDD would be associated with lower risks of fibroids. If, however, uterine myometrium responds similarly to mammary tissue, higher TCDD levels would be associated with higher risks of uterine fibroids.

MATERIALS AND METHODS

Study population

Women eligible for the Seveso Women's Health Study were newborn to 40 years of age in 1976, had resided in one of the most highly contaminated areas (zones A or B), and had adequate stored sera collected soon after the explosion (20). Enrollment began in March 1996 and was completed in July 1998. Of 1,271 eligible women, 17 could not be located or contacted, 21 had died, and 12 were too ill to participate. Of the 96 percent of eligible women who could be contacted, 981 (80 percent) participated. For this analysis, we excluded 25 women who had been diagnosed with fibroids before July 10, 1976. Thus, the final sample included 956 women.

Procedure

Details of the study procedure are presented elsewhere (20). Briefly, after informed consent was obtained, each woman underwent a blood draw and was interviewed by a trained nurse-interviewer who was blind to the participant's TCDD level and zone of residence. Information collected during the interview included demographic characteristics; personal habits; and occupational, menstrual, reproductive, and medical histories. Medical records were requested for all gynecologic treatments or conditions. After the interview, women who were 50 years of age or younger at interview or who were still menstruating were invited to undergo a gynecologic examination and transvaginal ultrasound. (Ultrasound was not offered to postmenopausal women over 50 years of age because the Seveso Women's Health Study was primarily designed to study endometriosis, which was expected to regress after menopause.) This study was approved by the institutional review boards of the participating institutions.

Definition of fibroid cases

Historical diagnosis of fibroids was assessed from the interview and medical records. Those who responded yes to the question, Have you ever been diagnosed with fibroids? were asked, How old were you when you were first diagnosed with fibroids? and past medical records were then requested for confirmation. If the medical record did not contradict the self-reported historical diagnosis, the woman was categorized as a “case” and was assigned an age at first diagnosis based on her self-report. Two women who reported a history of fibroids had a medical record that contradicted this self-report. These women were categorized as having no prior diagnosis of fibroids.

For 25 women who responded no to the question, Have you ever been diagnosed with fibroids?, medical records were available (requested for other diseases) that indicated a diagnosis of fibroids. These women were categorized as fibroid cases with age at first diagnosis based on that reported in the medical record.

The presence of fibroids at the time of follow-up was assessed for 634 women who underwent transvaginal ultrasound. The ultrasound examinations were performed by a gynecologist (N. G.) at the Mangiagalli Hospital of the University of Milan or, in a few instances, by a gynecologist at Desio Hospital; all had specific expertise in ultrasonography. All examinations used the Aloka color Doppler 680 and SSD 2000, 5-MHz transvaginal probe for imaging (Aloka Co., Tokyo, Japan) and 6-MHz pulsed Doppler system for blood flow analysis. Volume in cubic centimeters for each fibroid was calculated by using the formula for a prolate ellipse (21), a validated estimate of fibroid volume (22). For women with multiple fibroids, the total volume of all fibroids was computed from the sum of the volumes of the individual fibroids. All ultrasound examinations were videotaped and reviewed by an independent gynecologist (D. O.).

Laboratory analyses

Archived serum samples, collected after the explosion, had been stored at −20^oC in the Desio Hospital Laboratory. For analysis of serum for TCDD, we preferentially selected the first sample available that was of adequate volume (>0.5 ml) and that was collected between 1976 and 1981. For 872 of the women (91 percent), TCDD was measured in sera collected between 1976 and 1977. For 56 women (6 percent), TCDD was measured in sera collected between 1978 and 1981. For 28 women (3 percent), the volume of the archived serum specimen was inadequate for analysis; therefore, a sample collected in 1996 was analyzed. The samples were measured for TCDD by high-resolution gas chromatography/mass spectrometry (23). Values were reported on a lipid-weight basis in parts per trillion (ppt) by dividing TCDD on a whole-weight basis by total serum lipid content, estimated from measurements of triglycerides and total cholesterol (24). The median serum sample weight was 0.65 g, the median limit of detection was 18.8 ppt, lipid adjusted, and the analytical coefficient of variation for TCDD was 10–12 percent.

For post-1977 TCDD values that were detectable but ≤10 ppt (n = 8), the measured value was retained for analysis. For post-1977 TCDD levels of >10 ppt, the TCDD exposure level was back-extrapolated to 1976, according to the Filser model (25) for women aged 16 years or younger in 1976 (n = 30) and according to the first-order kinetic model for older women (n = 37) (26). For nondetectable values (n = 94), a serum TCDD level equal to one half the detection limit was assigned (27).

Statistical analyses

For data analysis, we used a method modified from that proposed by Dunson and Baird (28). Our fibroids data were of two types: 1) those fibroids detected prior to the study and reported by the participants or medical records (“historical diagnosis”); and 2) those fibroids detected for the first time by the study ultrasound examination. The historical diagnosis data consisted of age at diagnosis for those with diagnoses and age at interview for those without diagnoses. From historical diagnosis data, it is possible to construct survival models of age at clinical detection (29). The ultrasound data provide “current status” data and consist of age at screening for each woman and whether she had fibroids detected at screening. From current status data, it is possible to construct models of age at subclinical onset of fibroids (30).

Dunson and Baird (28) proposed a likelihood-based method for combining historical and current status data, when both are collected. Dunson and Baird presented a proportional odds analysis, in which cumulative odds of onset are modeled. Following a suggestion in their paper, we modified their method and present the results of a proportional hazards analysis in this paper. In a proportional hazards model, effects of covariates are expressed in terms of relative age-specific rates (“hazards”) of onset. For all models, the distributions of age at onset and of age at diagnosis were assumed to be piecewise linear, with nine change points selected to ensure an adequate number of cases between points (28). All models were fit by maximum likelihood, with confidence intervals for the hazard ratios generated by profile likelihoods. We report here only the results on fibroid onset because this outcome is the biologically relevant one.

We considered continuous and grouped versions of TCDD. In a linear model with log₁₀TCDD as the covariate, the impact on the age-adjusted hazard rate of onset is expressed as the hazard ratio associated with a 10-fold increase in TCDD. For the grouped model, we categorized TCDD into three groups: ≤20.0 ppt, 20.1–75.0 ppt, and >75.0 ppt. The lowest cutpoint was set at 20 ppt (body burden ∼4 ng/kg) because, in 11 pooled samples from unexposed women, the medians ranged between 15 ppt and 20 ppt (31). The middle cutpoint was set at 75 ppt, approximately the median TCDD level among those whose levels were higher than 20 ppt.

To check the form of the dose-response curve, we fitted the log-hazard rate to a smoothing spline in log₁₀ serum TCDD. That is, we fitted the following model for h(t | z), the hazard rate of onset of fibroids at age t for a woman with baseline hazard h₀(t) and serum level TCDD: log(h(t | TCDD)) = log(h₀(t)) + S^*(TCDD). Here, S^*(TCDD) = S(log₁₀ TCDD), and S is a natural smoothing spline. We generated the spline basis functions with 3 degrees of freedom from the S-Plus function ns (Insightful Corporation, Seattle, Washington).

We considered other covariates associated with fibroids in the literature (32–34) including parity, family history of fibroids, age at menarche, current body mass index (weight (kg)/height (m)²), smoking, alcohol consumption, and education. These variables were entered one at a time into the age-adjusted models with dichotomized TCDD (≤20 ppt or >20 ppt) and were designated potential confounders if they changed the TCDD coefficient by 10 percent or more. None of the covariates considered confounded the results when these criteria were used; therefore, only age-adjusted models are presented here. To evaluate confounding or effect modification by developmental stage, we repeated the analyses separately for women who had (n = 672) and had not (n = 284) started menstruating in the year of the explosion.

The pure current status approach uses less information than the Dunson-Baird approach (28), but it has the advantage of fewer assumptions. Therefore, using the semiparametric approach of Shiboski (35), we analyzed the current status data for 377 of 634 women for whom current status data were available (those who underwent transvaginal ultrasound). The analysis excluded 207 women who were less than age 32 years at the time of the ultrasound (age of the youngest case) and 50 women who had a history of fibroids but had no fibroid detected by ultrasound.

We also assessed the relation of serum TCDD to the natural logarithms of total fibroid volume and of the volume of the largest single fibroid by using linear regression. For these analyses, we included only the 95 fibroid cases discovered at ultrasound among the subset of 377 women with current status ultrasound data described above.

Analyses based on the combined historical and current status data were programmed in S-Plus statistical software (revision of code provided by D. Dunson). We used the R package, BAM (provided by S. Shiboski) for the pure current status analyses. The volume analyses were programmed in Stata, version 9.2 software (Stata Corporation, College Station, Texas).

RESULTS

Characteristics of the 956 women by TCDD category are presented in table 1. The mean ages at explosion and follow-up were 19.8 (standard deviation, 1.2) years and 40.4 (standard deviation, 11.5) years, respectively. All women were Caucasian, 72.5 percent were parous, 26.7 percent had a family history of fibroids (mother, sisters, grandmothers), 15.5 percent currently used oral contraceptives, 30.0 percent were overweight (body mass index 25–29.9 kg/m²) or obese (body mass index ≥30.0 kg/m²), 20.9 percent were current smokers, and 30.0 percent consumed alcohol. As previously reported (31), TCDD levels were higher in those exposed at younger ages. Women in the highest exposure category were also more likely to be never married, nulliparous, and of lower body mass index, but these factors are also related to age.

TABLE 1.

Select characteristics of women by serum TCDD* level, Seveso Women's Health Study, Italy, 1996–1998

Characteristic	TCDD (parts per trillion)
	≤20.0 (n = 151)		20.1–75.0 (n = 410)		>75.0 (n = 395)		Total (n = 956)
	No.	%	No.	%	No.	%	No.	%
Age at chemical explosion (years)
0–10	15	9.9	54	13.2	163	41.3	232	24.3
11–20	58	38.4	127	31.0	94	23.8	279	29.2
21–30	47	31.1	113	27.6	75	19.0	235	24.6
>30	31	20.5	116	28.3	63	16.0	210	22.0
Age at interview (years)
20–29	10	6.6	40	9.8	141	35.7	191	20.0
30–39	54	35.8	124	30.2	106	26.8	284	29.7
40–49	49	32.5	117	28.5	73	18.5	239	25.0
≥50	38	25.2	129	31.5	75	19.0	242	25.3
Age at menarche (years)†
<12	34	22.5	92	22.4	97	24.6	223	23.4
12–13	77	51.0	190	46.3	187	47.5	454	47.5
≥14	40	26.5	128	31.2	110	27.9	278	29.1
Education‡
<Required	54	35.8	170	41.5	102	25.8	326	34.1
≥Required	97	64.2	240	58.5	293	74.2	630	65.9
Marital status
Never married	12	8.0	44	10.7	124	31.4	180	18.8
Married/living together	129	85.4	338	82.4	257	65.1	724	75.7
No longer married	10	6.6	28	6.8	14	3.5	52	5.5
Current body mass index (kg/m2)†
<18.5	4	2.7	15	3.7	36	9.1	55	5.7
18.5–24.9	87	57.6	256	62.6	271	68.6	614	64.3
25.0–29.9	45	29.8	104	25.4	63	16.0	212	22.2
≥30	15	9.9	34	8.3	25	6.3	74	7.8
Parity
0	25	16.5	71	17.3	167	42.3	263	27.5
1	35	23.2	85	20.7	74	18.7	194	20.3
≥2	91	60.3	254	62.0	154	39.0	499	52.2
Oral contraceptive use
Never	66	43.7	205	50.0	164	41.5	435	45.5
Former	59	39.1	155	37.8	159	40.3	373	39.0
Current	26	17.2	50	12.2	72	18.2	148	15.5
Family history of fibroids
No	100	66.2	307	74.9	294	74.4	701	73.3
Yes	51	33.8	103	25.1	101	25.6	255	26.7
Current coffee consumption (cups/day)
<1	12	8.0	45	11.0	45	11.4	102	10.7
1–3	102	67.5	277	67.6	249	63.0	628	65.7
>3	37	24.5	88	21.4	101	25.6	226	23.6
Smoking
Never	98	64.9	270	65.8	251	63.5	619	64.8
Former	22	14.6	54	13.2	61	15.5	137	14.3
Current	31	20.5	86	21.0	83	21.0	200	20.9
Alcohol consumption
Never	93	61.6	257	62.7	260	65.8	610	63.8
Former	14	9.3	27	6.6	18	4.6	59	6.2
Current	44	29.1	126	30.7	117	29.6	287	30.0

*

TCDD, 2,3,7,8-tetrachlorodibenzo-p-dioxin.

†

Numbers do not add to total because of missing data.

‡

Education: <required, <8 years; ≥required, ≥8 years.

Table 2 presents the numbers and proportions of women with fibroids by TCDD category and method of case ascertainment. Of the 956 women, 251 (26.3 percent) had fibroids. Of these, 37.8 percent were diagnosed by the study ultrasound, 61.0 percent by medical record only (they did not have a study ultrasound, but the medical record requested recorded a fibroid), and 1.2 percent by self-report only (they had no ultrasound or medical record). The method of ascertainment differed between those in the lowest exposure group (≤20.0 ppt) and those with levels of 20.1–75 ppt (p < 0.01) but not with those in the highest exposure group (>75 ppt) (p = 0.33).

TABLE 2.

Number and percentage of women with fibroids by serum TCDD* category and by method of ascertainment, Seveso Women's Health Study, Italy, 1996–1998

	TCDD (parts per trillion)
	≤20.0 (n = 151)		20.1–75.0 (n = 410)		>75.0 (n = 395)		Total (n = 956)
	No.	%	No.	%	No.	%	No.	%
No. with fibroids	62	41.1	110	26.8	79	20.0	251	26.3
Method of ascertainment
Ultrasound only	19	30.6	19	17.3	20	25.3	58	23.1
Ultrasound and medical record	12	19.4	10	9.1	15	19.0	37	14.7
Medical record only	29	46.8	80	72.7	44	55.7	153	61.0
Self-report only	2	3.2	1	0.9	0	0.0	3	1.2

*

TCDD, 2,3,7,8-tetrachlorodibenzo-p-dioxin.

As presented in table 3, crude prevalence of fibroids increased with age and with factors associated with age, including less education, former or current marriage, overweight or obesity, and previous pregnancy. In addition, fibroids were more common among women who had a family history of fibroids, drank at least one cup of coffee per day, or who ever drank alcoholic beverages and were less common among current oral contraceptive users. The median TCDD levels for cases and noncases were 43.9 ppt lipid adjusted and 64.1 ppt lipid adjusted, respectively.

TABLE 3.

Fibroid case status* by select characteristics of women, Seveso Women's Health Study, Italy, 1996–1998

Characteristic	No. of cases	Total	%
All women	251	956	26.3
Age in 1996 (years)
20–30	2	232	0.9
31–40	47	279	16.8
41–50	117	235	49.8
≥51	85	210	40.5
Age at menarche (years)†
<12	61	223	27.4
12–13	116	454	25.6
≥14	74	278	26.6
Education‡
<Required	147	326	45.1
≥Required	104	630	16.5
Marital status
Never married	5	180	2.8
Married/living together	219	724	30.2
No longer married	27	52	51.9
Current body mass index (kg/m2)†
<18.5	5	55	9.1
18.5–24.9	134	614	21.8
25.0–29.9	79	212	37.3
≥30	33	74	44.6
Parity
0	25	263	9.5
1	52	194	26.8
≥2	174	499	34.9
Oral contraceptive use
Never	139	435	32.0
Former	99	373	26.5
Current	13	148	8.8
Family history of fibroids
No	175	701	25.0
Yes	76	255	29.8
Current coffee consumption (cups/day)
<1	17	102	16.7
1–3	173	628	27.5
>3	61	226	27.0
Smoking
Never	168	619	27.1
Former	40	137	29.2
Current	43	200	21.5
Alcohol consumption
Never	135	610	22.1
Former	26	59	44.1
Current	90	287	31.1

*

Fibroid cases were defined by historical diagnosis or current ultrasound.

†

Numbers do not add to total because of missing data.

‡

Education: <required, <8 years; ≥required, ≥ 8 years.

Table 4 presents hazard ratios and 95 percent confidence intervals for earlier onset of fibroids associated with TCDD exposure. The estimated age-adjusted hazard ratio associated with a 10-fold increase in serum TCDD was 0.83 (95 percent confidence interval (CI): 0.65, 1.07), suggestive of a weak inverse association between the hazard of fibroid onset and increasing exposure. Compared with that for women with TCDD levels of ≤20 ppt, the age-adjusted hazard ratio for women with levels of 20.1–75.0 ppt was 0.58 (95 percent CI: 0.41, 0.81) and with levels of >75.0 ppt was 0.62 (95 percent CI: 0.44, 0.89). When we stratified the analysis by menarche status at the time of the explosion, for women who were postmenarche, the results were similar to those in the entire sample. There were only four cases among women who were premenarche; therefore, we were unable to model TCDD in this group.

TABLE 4.

Association between serum TCDD* and age-adjusted hazard of fibroid onset among women, Seveso Women's Health Study, Italy, 1996–1998

TCDD (parts per trillion)	No. of cases	Total	%	Age-adjusted HR*	95% CI*
log10TCDD	251	956	26.3	0.83	0.65, 1.07
≤20.0	62	151	41.1	1.00
20.1–75.0	110	410	26.8	0.58	0.41, 0.81
>75.0	79	395	20.0	0.62	0.44, 0.89

*

TCDD, 2,3,7,8-tetrachlorodibenzo-p-dioxin; HR, hazard ratio; CI, confidence interval.

Figure 1 shows the smoothed function of TCDD in the proportional hazards model by TCDD. The log hazard decreased with exposure, up to about 47 ppt, but no change occurred thereafter. The figure suggests that the straight-line model with log₁₀TCDD is not a good fit to the data and that the categorical model provides a better fit. This conclusion is reinforced by a likelihood ratio test comparing the 3 degrees of freedom categorical model with the simpler 2 degrees of freedom trend model with fibroid categories scored as 1, 2, and 3 (p < 0.02).

FIGURE 1.

Smoothed estimate of the relation between log 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and log hazard of fibroids onset, adjusted for age, Seveso Women's Health Study, Italy, 1996–1998. Vertical lines indicate TCDD categorical cutpoints (20 parts per trillion (ppt), 75 ppt).

Fifty women who were analyzed had a history of fibroids, by medical record or self-report, but no fibroid at the ultrasound screen. This discrepancy could have happened because the fibroids regressed or were removed, but potentially these women were misclassified. To check robustness of our conclusions to such misclassification, we repeated the analyses after reclassifying these women as noncases. The results were similar. Specifically, the age-adjusted hazard ratio for women with levels of 20.1–75.0 ppt was 0.58 (95 percent CI: 0.40, 0.85) and with levels of >75.0 ppt was 0.74 (95 percent CI: 0.50, 1.09) compared with women with levels of ≤20 ppt.

Current-status-only analyses of the ultrasound data (n = 377) led to similar conclusions. Compared with that for women with levels of ≤20 ppt, the age-adjusted hazard ratio for women with TCDD levels of 20.1–75.0 ppt was 0.43 (95 percent CI: 0.26, 0.73) and for women with TCDD levels of >75.0 ppt was 0.71 (95 percent CI: 0.43, 1.18).

Figure 2 shows the estimated probability of onset by age for the three-category model. This figure is based on a light smoothing of the three piecewise linear curves, each with nine pieces. The prevalence of fibroids in women aged 50 years is estimated to be about 70 percent if serum TCDD is ≤20 ppt and about 50 percent in women aged 50 years with higher exposure.

FIGURE 2.

Estimated age-specific cumulative onset of fibroids by serum 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) level (parts per trillion (ppt), lipid adjusted), Seveso Women's Health Study, Italy, 1996–1998. Solid line represents a serum TCDD level of ≤20.0 ppt. Dashed line indicates a serum TCDD level of 20.1–75 ppt. Dotted line represents a serum TCDD level of >75.0 ppt.

In regression analyses, there was no evidence of an association of fibroid volume (largest single fibroid or total fibroid volume) with TCDD. All p values were over 0.40, and therefore individual data are not shown.

DISCUSSION

We have shown that women from the Seveso Women's Health Study with serum TCDD levels of >20 ppt in serum collected soon after the explosion had a lower age-adjusted risk of uterine leiomyomas than women with lower exposures. We found no further decrease in risk with TCDD levels of >20 ppt based on the categorical data or >47 ppt based on the smoothed function of the continuous data. These results are consistent with earlier findings in rodents that TCDD acts as an antiestrogen in uterine cells in vitro (17) but not with recent findings that TCDD induces an estrogen-like transcriptional response in uteri of ovariectomized immature animals (18). Our study also confirms previously published reports of high prevalence of fibroids at the end of the reproductive years in Caucasian women with, presumably, background-only exposure (36).

Walker (37) suggests that a signaling defect in uterine leiomyoma may lead to tumor growth under conditions such as following the withdrawal of steroid hormones, in which normal myometrial cells would undergo apoptosis. Although uterine leiomyoma growth is influenced primarily by estrogen through the modulation of the estrogen receptor, progesterone and other hormones also play a critical role (38). Thus, uterine leiomyomas may be sensitive to the influences of chemicals that act as exogenous hormones (5). Chemicals, which are estrogen-receptor ligands, can be highly tissue, cell, and promoter specific, and the ability of the estrogen receptor to activate gene expression may depend on the ligand and the context (5). For example, tamoxifen decreases the risk of recurrence of estrogen-responsive breast cancer but increases the risks of endometrial cancer (39, 40) and of leiomyomas in postmenopausal women (21). In contrast, tamoxifen reduced the incidence and volume of leiomyomas in the Eker rat model (10), suggesting that its effects may also be species specific.

TCDD does not act by directly binding to the estrogen receptor (41) and, to our knowledge, has never been tested in the Eker rat model. However, like other hormonally active chemicals (9), it may differ in its effect on the breast and uterus, and even in its effect on the uterine endometrium and myometrium. We report here a decrease in risk of uterine fibroids, although no change in fibroid volume, and no clear dose response. This result is in contrast to the significant dose-response increase for breast cancer (19) and the nonmonotonic increase in earlier onset of menopause up to 100 ppt TCDD but not above (42), previously reported in the Seveso Women's Health Study.

This study has several advantages for studying the onset of fibroids. The study population is a true cohort, assembled on the basis of residence at the time of the accident. Also, most individual serum TCDD concentrations were from blood samples taken within 2 years of exposure. The statistical approach of Dunson and Baird (28), which we adapted, made it possible to include information from the postmenopausal women over 50 years of age who, by protocol, were not offered the study ultrasound. One limitation of the study is that differences in risk by stage of development at exposure (e.g., in utero, lactational, premenarcheal) could not be determined because all women in the age range in which fibroids would appear were exposed postnatally, and only four cases were observed among women who were premenarcheal at exposure. Additional follow-up of the cohort will mean that younger women will reach the ages at which fibroids occur and would therefore shed light on this question.

In summary, we found that higher serum levels of TCDD following the explosion in Seveso, Italy, are associated with a lower risk of fibroids. This finding supports other evidence that TCDD acts as an antiestrogen in the uterine myometrium (17) and may differ in its effects on mammary tissue and on uterine endometrium. The mechanism for the observed inverse association is unclear.

Abbreviations

Abbreviations
 
• CI

  confidence interval

 
• ppt

  parts per trillion

 
• TCDD

  2,3,7,8-tetrachlorodibenzo-p-dioxin

This study was supported by grants R01 ES07171 and F06 TW02075-01 from the National Institutes of Health, R82471 from the US Environmental Protection Agency, EA-M1977 from the Endometriosis Association, 2P30-ESO01896-17 from the National Institute of Environmental Health Sciences, and 2896 from Regione Lombardia and Fondazione Lombardia Ambiente, Milan, Italy.

The authors gratefully acknowledge Dr. Paolo Brambilla and Dr. Stefania Casalini for coordinating data collection at the Hospital of Desio, Dr. Luigi Bonsignore for examining women at the Hospital of Desio, and Wayman Turner of the Centers for Disease Control and Prevention for serum TCDD measurements.

Conflict of interest: Dr. Olive is a member of the Speaker's Bureau of US HealthConnect, Inc. (Fort Washington, Pennsylvania) and of the Advisory Boards of Takeda Pharmaceuticals North America, Inc. (Deerfield, Illinois), TAP Pharmaceutical Products Inc. (Lake Forest, Illinois), Wyeth (Madison, New Jersey), and The Proctor & Gamble Company (Cincinnati, Ohio).