Cooper, Glinda S.*; Klebanoff, Mark A.†; Promislow, Joanne*; Brock, John W.‡; Longnecker, Matthew P.*
Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
The persistent organic pollutants polychlorinated biphenyls (PCBs) and the major dichlorodiphenyltrichloroethane (DDT) metabolite p,p′-DDE (1,1-dichloro-2,2-bis(p-chlorophenyl)ethylene (DDE) affect the endocrine system, possibly at background-exposure levels.1 Menstrual cycle characteristics are one aspect of women's health that has received relatively little attention from researchers,2 and experimental studies in rats3,4 and monkeys5,6 suggest this may be a relevant area of study with respect to these organochlorine compounds. However, there have been few epidemiologic studies of PCBs or DDE in relation to menstrual cycle patterns.7–9
We recently measured levels of PCBs and DDE in maternal serum specimens from a cohort study that enrolled pregnant women in the United States in the early 1960s when PCB levels were higher than at present10 and when DDT use in the United States was at its peak.11 Upon enrollment in the study, women reported information about their menstrual cycle characteristics (usual length, regularity, and bleeding duration). The data from 2314 women with PCB and DDE measurements were used in this analysis.
The subjects were enrolled in the Collaborative Perinatal Project, a prospective study of neurologic disorders and other conditions in children.12 Pregnant women were recruited between 1959 and 1965 from 12 study centers (in Baltimore, Boston, Buffalo, Memphis, Minneapolis, New Orleans, New York , Philadelphia, Portland, Providence, and Richmond). Details of the eligibility criteria and recruitment have been presented previously.13,14 The characteristics at registration of women in the sample were essentially the same as those in the sampling frame.12
Nonfasting blood was collected from the women at enrollment and at approximately 8-week intervals during pregnancy, at delivery, and 6 weeks postpartum. Sera were stored in glass at –20°C, with no recorded thaws. Approximately 42,000 women enrolled in the study.
We measured serum organochlorine levels in a subset of Collaborative Perinatal Project participants. In all, there were 44,075 women who had delivered a live born singleton after registration and for whom a 3-mL aliquot of third-trimester maternal serum was available. We used 3 sampling methods to select 9 groups that constituted the analysis subset, which was designed to assess the association between organochlorine levels and birth outcomes, including neurodevelopment.13 One of the methods was to obtain a simple random sample (n = 1200). We also selected the mothers of male infants with cryptorchidism, hypospadias, or polythelia (n = 232, 213, and 185, respectively). The remaining 5 groups were selected according to the child's performance on tests at various ages: neonatal reflexes, neonatal tone, Bayley Scale of Infant Development at 8 months, IQ on the Weschler Intelligence Scale for Children at age 7 years, and audiometric examination at age 8 years; n = 993). The 9 samples were obtained independently, and the 71 pregnancies selected for more than 1 type of sample were included only once in this analysis. Use of the stored sera for these analyses was approved by the Institutional Review Board of the National Institute of Environmental Health Sciences.
Serum cholesterol and triglycerides were measured using standard enzymatic assays. Serum levels of 11 polychlorinated biphenyls, DDE, and p,p′-DDT (the parent compound and active ingredient of the pesticide) were measured at the Centers for Disease Control and Prevention in 1997–1999 using solid-phase extraction, clean-up, and dual-column gas chromatography with electron capture detection.15 The results shown are not recovery-adjusted.14 The specific PCB congeners measured were PCBs 28, 52, 74, 105, 118, 138, 153, 170, 180 194, and 203. These specific congeners were chosen because they were the congeners detected at the highest levels in human samples (PCBs 74, 118, 138, 153, 170, 180, and 194)16 or because they had been found to be neurotoxic in animal experiments and were detectable in at least some portion of human specimens (PCBs 28, 52, 105, 203).17 We included the potential neurotoxic congeners because of the use of these samples in our analysis of neurodevelopment.18
A batch consisted of 1 quality control sample, 10 unknowns, a reagent blank, an instrument blank and 1 standard. If any quality control sample was outside the 99% confidence limits, or outside the 95% confidence limits in 2 consecutive batches, the entire batch was reanalyzed. All laboratory personnel were blinded to sample status and quality control status during the analyses. The laboratory participated in international round-robin testing to verify consistent and comparable results on spiked serum samples. The between-assay coefficient of variation for total PCBs was 19% at 3.54 μg/L and for DDE was 19% at 29 μg/L (n = 291). These coefficients of variation are similar to values reported by other researchers using blinded samples.19
Of the 2823 subjects selected for inclusion in the study, a laboratory result for both DDE and PCBs was not obtained for 7% (n = 211), mainly because the measured value did not meet the quality control standards. Thus, results for DDE and PCBs were available for 2612 subjects. For DDE, all mothers had levels above the detection limit (0.61 μg/L). The proportion with values below the detection limit for specific PCBs was: PCB 118, 11.1%; PCB 138, 5.6%; and PCB 153, 9.6%. The proportion with values less than the detection limit was higher (>35%) for the remaining congeners (that is, they were present at lower levels). If the level of a given congener was less than the detection limit, imputation was not performed; rather, the measured levels below the detection limit reported by the laboratory were included in the analysis.20
A gynecologic history was taken by study personnel at the initial study visit (median approximately 20 weeks’ gestation). Scripted questions were not used, but the information was recorded using a structured form. The history included questions about age at menarche, usual duration of menses, amount of flow (heavy, medium, or light), and usual cycle length. There was also an open-ended question about unusual features of the menstrual period, and a question about dysmenorrhea (coded as “none,” if no discomfort noted; “slight,” if discomfort noted but no medication required; “moderate,” if discomfort required medication but usual activities not affected; and “severe,” if patient required bed rest or lost time from work). If the participant reported a range of values for cycle length or duration (eg, 25–30 days, or 3–5 days), both values were recorded. We categorized a woman as having irregular menstrual cycles if she met any of the following criteria: menstrual cycles reported as irregular; amenorrhea as suggested by response on the open-ended question (“skipped one or more menstrual periods regularly”); the range of values for usual length of menstrual cycles greater than 7 days. Coding of duration of menses was limited to 1 to 8 days, with 8 or more days included in the “8” response. Information on endometriosis was not collected in the Collaborative Perinatal Project.
Height was measured at the initial study visit, and information was collected on prepregnancy weight, age, smoking, and socioeconomic index. The height and weight data were used to calculate prepregnancy body mass index (kg/m2). The socioeconomic index was the mean of 3 percentile scores (for education, occupation, and family income),21 calculated for all participants in the Collaborative Perinatal Project.
Of the 2612 pregnancies for which measurements of PCBs and DDE were available, we excluded those whose menstrual cycle regularity was “unknown” (n = 83), those who reported use of oral contraceptives (n = 18), those who had missing data on body mass index (n = 176), and those who had missing data on age, race, or lipid levels (n = 2). We also excluded 19 pregnancies that were the second study pregnancy from the same mother, so that only one observation per participant was used for this analysis. The total sample size for these analyses was 2314.
We examined cycle length and bleeding duration in relation to levels of PCBs and DDE using ordinary least-squares models. There was considerable digit preference (ie, “28” for cycle length and “5” for bleeding duration) and skewness of these variables, but examination of residuals did not reveal evidence of poor model fit or influential data points. We also conducted additional analyses using polytomous logistic regression with 3-level outcomes (< 26 days, 26–30, 31+ days for cycle length; < 4, 4 to 6, and 7+ days for bleeding duration). These results were similar to those of the linear regression analyses and are not presented here. We chose cut-points for categories of PCBs and DDE that had been used in previous analyses13,18 with the highest 2 categories corresponding to approximately 20–25% of the distribution.
We adjusted for study center in all multivariate models because of differences in exposure levels and because of potential differences in recruitment or data collection practices among centers. For the evaluation of confounding, PCBs and DDE were modeled as continuous linear variables. Serum PCB and DDE levels are affected by the concentration of serum lipids and, therefore, we included serum triglycerides and cholesterol as linear variables in all multivariate models. Potential confounding by additional factors was evaluated first by comparing the coefficients for PCBs and DDE from models that included only center and lipid variables to the coefficients in a model with each additional factor. If the coefficient per 1 μg/L of PCBs or DDE changed by 15% or more after the addition, the factor was further considered as a confounder. The factors considered as potentially confounding were age (years), race (black, white, other), prepregnancy body mass index (continuous), socioeconomic index (mean percentile score), smoking (smoked fewer than 5 packs ever, smoked at least 5 packs ever), age at menarche, parity, and pregnancy interval. Age, race, and prepregnancy body mass index met the criteria for confounding and, therefore, we included them in the fully adjusted model. Including DDE and PCBs in the same model had little effect on the results for the individual organochlorine measures and so were not included as adjustment variables.
We used logistic regression to estimate the odds of menstrual cycle irregularity in relation to level of PCBs and DDE, using the fully adjusted model described previously. We also examined the odds of having heavy menstrual flow (compared with the combined group of medium or light flow) in relation to the organochlorine levels. For the analysis of dysmenorrhea, we compared the odds of reporting severe or moderate dysmenorrhea with the combined categories of slight and none. Results were similar when we excluded the moderate category to make a clearer distinction between the categories; only the results of the analyses with all groups are shown.
To evaluate potential interaction by center (12 sites), by sampling group (9 groups), and by race (comparing whites and blacks), we included interaction terms between these variables and the 5-level trend test variable (median value within each group) for PCBs and DDE. There was evidence of interaction by sampling group in the analysis of PCBs and bleeding duration (P = 0.02). Because of this interaction, we limited the sample for the PCB-bleeding duration analysis to the 967 women who we had selected as the simple random sample group. There was not evidence of interaction by sampling group for the other analyses (all P > 0.20), and so results for the full sample are presented for these other analyses.
To assess the potential effect of various forms of misclassification, we repeated the analyses, excluding women who had given birth in the 2 years preceding enrollment because they may have had more difficulty reporting usual menstrual cycle characteristics. We also repeated the analyses limiting the sample to primiparous women to reduce misclassification that may result from the effect of previous pregnancies and lactation on organochlorine levels. We also examined analyses using the organochlorine measures as a per-serum lipid measure.
Seven of the PCB congeners (PCBs 74, 105, 118, 138, 153, 170, and 180) were highly correlated, with a Spearman's correlation coefficient greater than 0.60. To assess the independent effects of individual PCB congeners, we created 1 variable based on the sum of the concentration of these 7 congeners and included it in a model with the remaining 4 congeners represented as individual terms (PCBs 28, 52, 194, and 203).18
An overview of participants’ background characteristics, organochlorine exposures, and menstrual cycle parameters is represented in Table 1. The mean age of participants was 24 years (±SD 6.2 years), and ranged from 13 to 45 years. The mean serum DDE level was 30.0 μg/L (±19.6 μg/L), with a median value of 24.7 μg/L. The mean PCB concentration was 3.1 μg/L (±1.9 μg/L), with a median of 2.7 μg/L. The mean cholesterol level was 239 μg/L (±67 μg/L) with a median of 233 μg/L, and the triglycerides mean was 210 μg/L (±81 μg/L) with a median of 196 μg/L. The mean usual cycle length among the 2150 women classified as having regular cycles was 28.4 days (±2.6 days), with a median of 28.0 days. Usual bleeding duration ranged from 1 to 8 days (mean = 4.8 ± 1.3; median = 5.0 days). Seven percent of participants reported irregular cycles, 19% reported heavy bleeding, and 9% reported severe dysmenorrhea.
Mean cycle length decreased slightly with increasing age, but little difference in cycle length, bleeding duration, or irregular cycles was observed in relation to body mass index, smoking, or parity (Table 2). Cycle length and bleeding duration were lower among black patients compared with white patients, and the prevalence of irregular cycles was higher among white patients compared with black patients.
Bleeding duration, heavy bleeding, and dysmenorrhea were not associated with levels of DDE or PCB (Table 3). These results were essentially unchanged in the subgroup analyses that excluded women who had a live birth within 2 years of enrollment and in the analyses limited to primiparous women (data not shown). Results were similar regardless of whether organochlorines were assessed on a wet-weight basis or per serum lipid.
Adjusted mean cycle length was not associated with DDE levels, although the prevalence of irregular cycles increased slightly with exposure (Table 4). Limiting the analysis to women who had a live birth within 2 years preceding enrollment and limiting the sample to primiparous women resulted in little change in the mean cycle length association with DDE, but in both of the subgroups the prevalence of irregular cycles appeared to be somewhat more strongly related to DDE than in the analysis of the full sample. The analyses of DDE on a wet-weight basis and per serum lipid produced similar results.
Increasing PCB concentration was associated with increasing cycle length, with a difference of 0.7 days in the highest compared with the lowest exposure group (Table 4). Irregularity was slightly more frequent among those in the higher categories of PCBs. Similar (or stronger) associations were seen in analyses limited to subgroups designed to decrease potential misclassification of outcome or exposure. In the analyses of PCBs expressed on a serum-lipid basis the associations with mean cycle length and irregularity were somewhat attenuated compared with the wet-weight analysis. However, in the subgroup analyses (excluding the women who had given birth within the 2 years preceding enrollment and limiting the sample to primiparous women), the association between PCB level and mean cycle length was seen in the PCB per serum-lipid analyses, with a difference of 1.0 day across categories (P = 0.01).
The analyses of the congener-specific PCB data indicated little difference between congeners in the association with bleeding duration. With respect to menstrual cycle length, however, the models using the wet-weight and the serum-lipid measures indicated a possible difference with respect to congener 52: an inverse association (decreased cycle length with increasing exposure) was observed with this congener, with an improved goodness of fit (ie, reduction in error sum of squares) of the model compared with the model with a single variable summing the 11 congeners.
We observed an association between increasing levels of serum PCBs and longer menstrual cycles. There was weaker evidence of an association with irregular cycles, but the effects on cycle length and variability were stronger in the subgroup analyses conducted to explore the effect of outcome or exposure misclassification. DDE was not associated with menstrual cycle length, but there was some evidence of an association with irregular cycles, particularly in the subgroup analyses. There was no relationship of either PCBs or DDE with bleeding characteristics (duration, volume) or dysmenorrhea. The difference we observed with respect to mean cycle length was relatively small (approximately 1.0 day increase from lowest to highest PCB category). Nonetheless, demonstration of the effect of organochlorines on menstrual cycle characteristics provides support for the use of these outcomes in studies of the potential health effects of endocrine disruptors.
Available data from experimental studies in animals support the plausibility that exposure to PCBs could cause menstrual disturbances in humans. Among monkeys exposed to low doses of PCBs (Aroclor 1254),6 which produced tissue levels comparable with those found in humans with background-level exposure, menses duration increased fairly consistently with dose, but a summary measure of this trend with precision estimates was not provided. In another study of monkeys, slightly higher doses of PCBs (Aroclor 1248) were administered and caused longer menstrual cycles, increased menses duration, and increased menstrual bleeding.5 Prolonged estrous cycles in rats were also seen in studies of PCBs (Aroclor 1254) in rats.3,4
Studies of PCBs and menstrual function in humans are somewhat limited and contradictory.7–9 In one study, shorter menstrual cycles were found in relation to consumption of PCB-contaminated fish.7 The types of PCBs and their exposure levels in this study are likely comparable with our subjects. In the more extreme situation of the poisoning from oil contaminated with PCBs and related compounds in Yucheng, Taiwan, the prevalence of “abnormally light” bleeding was somewhat higher in Yucheng women (8.8%) compared with controls (3.4%).8
PCBs alter endocrine systems by a variety of mechanisms in experimental settings. PCBs are known to decrease thyroxine in animals,22 and some data suggest background level exposure can have hypothyroid effects on humans.23 Thyroid hormone data are not available for the women in the Collaborative Perinatal Project. Hypothyroidism in humans can cause prolonged duration of menses, and hyperthyroidism may result in oligomenorrhea and decreased menstrual flow.24 PCBs in vitro have a direct effect on hypothalamic cells, resulting in increased levels of gonadotropin releasing hormone levels.25 Direct toxicity to the oocyte also has been reported.26 Which of these potential mechanisms, if any, is relevant to humans is unclear. As with several of the specific PCB congeners we examined, PCB 52 is a weak inducer of P450s that metabolize phenobarbital. With respect to the other potential mechanisms of action for PCB 52, however, this congener is somewhat different from the others we measured: it is more neurotoxic (eg, dopamine depleting), and perhaps more estrogenic.27 Whether these properties account for the difference among congeners in associations with cycle lengths is unclear.
Because of the lipophilic properties of PCBs and DDE, differences in serum levels of lipids will affect the observed levels of organochlorines. How best to account for the influence of serum lipids on serum organochlorines levels remains an unresolved issue.14 The standard formula used to estimate total serum lipids on the basis of levels of triglycerides and cholesterol was based on data from a small number of primarily middle-aged men and nonpregnant women of unspecified race.28 Throughout pregnancy, the composition of serum lipids changes,29 and thus the formula may introduce imprecision in accounting for the effect of lipids, compared with separate adjustment for cholesterol and triglycerides by including them as terms in a regression model. For these reasons we believe the latter approach is more appropriate for this analysis.
The mean menstrual cycle length for Collaborative Perinatal Project participants (28.4 days) is similar to previously published data for U.S. women.30 The decrease in mean cycle length with increasing age30 and slight increase with increasing body mass index are also consistent with previous reports.31,32 Racial differences (shorter bleeding duration and decreased mean cycle length among blacks, and higher prevalence of irregular cycles among whites) in this study were similar to what has been reported in some previous studies.32–34 However, previously noted associations between irregular menstrual cycles and either low or high body mass index2 and smoking35 were not observed in the Collaborative Perinatal Project participants.
This is the first large epidemiologic study to examine menstrual cycle patterns in relation to low-level exposure to the persistent organochlorines PCBs and DDE. In addition to the sample size, a strength of the study is that exposure levels were higher than would generally be seen today. Also, few women had used oral contraceptives at the time of this study (1959 to 1965). We conducted several analyses to examine the potential for bias introduced by the sampling technique or quality of the menses data (analyses detailed in the appendix, available with the electronic version of the article). We found little evidence of problems with the data from the selection of subjects, exclusion of subjects with missing values, or potential indicators of recall accuracy.
An important limitation of the data is that menstrual cycle characteristics were based on self-report, rather than on diary data or hormonal measurements. Women were recalling their prepregnancy menstrual cycles, but a specific time period was not delineated in the questions. Long-term (eg, 30-year) recall of menstrual cycle patterns may not be very accurate36,37; the validity of shorter recall periods is not established. Another issue is whether our measure of organochlorines is relevant to the time period for which menses patterns were reported. Other studies have found that short-term serum PCB and DDE levels are consistent over time among healthy adult women,38 or during pregnancy.39 Therefore, our PCB and DDE levels should represent prepregnancy exposure reasonably well. The proportion of samples not meeting quality control standards was relatively high; this problem was likely due to deterioration of the matrix in some samples interfering with the extraction and separation.40 PCBs and DDE, however, appear to be quite stable during long-term frozen storage,13 and we have no reason to suspect that the measurements in samples that met our quality control standards were unreliable.
Other issues with respect to outcome assessment arise from the data collection instrument. Our definition of irregularity was based on responses to an open-ended question about unusual features of menstrual cycles, and on ranges for usual cycle length that were given without specific prompting. Specific definitions of amount of flow (heavy, medium, or light) were not provided. The dysmenorrhea categories were based on subjective reporting of the impact of pain. Additional research among populations with relatively high exposure using more precise measures of menstrual cycle characteristics would be a useful follow-up to this study.
This study extends the experimental evidence from studies in rats and monkeys pertaining to endocrine-related (specifically, menstrual or estrous-related) effects of PCBs. The associations we observed in women with background levels of PCBs indicate that the menstrual cycle may be a sensitive marker of the effects of environmental endocrine disruptors.
We thank Donna Baird and Germaine Buck for their thoughtful reviews of the manuscript.
1.Longnecker MP, Rogan WJ, Lucier G. The human health effects of DDT (dichlorodiphenyltrichloroethane) and PCBS (polychlorinated biphenyls) and an overview of organochlorines in public health. Annu Rev Public Health. 1997;18:211–244.
2.Harlow SD, Ephross SA. Epidemiology of menstruation and its relevance to women's health. Epidemiol Rev. 1995;17:265–286.
3.Carruthers, Christina M. “The multigenerational effects of polychlorinated biphenyl on the reproduction, development, and sex hormone levels of Sprague-Dawley rats.” Dissertation. Bowling Green State University, 2003.
4.Brezner E, Terkel J, Perry AS. The effects of Aroclor 1254 (PCB) on the physiology of reproduction in the female rats—I. Comp Biochem Physiol C. 1984;77:65–70.
5.Barsotti DA, Marlar RJ, Allen JR. Reproductive dysfunction in rhesus monkeys exposed to low levels of polychlorinated biphenyls (Aroclor 1248). Food Cosmet Toxicol. 1976;14:99–103.
6.Bryce F, Hayward S, Stapley R, Arnold DL. Consequences of Aroclor 1254 ingestion on the menstrual cycle of rhesus (Macaca mulatta) monkeys. Food Chem Toxicol. 2000;38:1053–1064.
7.Mendola P, Buck GM, Sever LE, Zielezny M, Vena JE. Consumption of PCB-contaminated freshwater fish and shortened menstrual cycle length. Am J Epidemiol. 1997;146:955–960.
8.Yu ML, Guo YL, Hsu CC, Rogan WJ. Menstruation and reproduction in women with polychlorinated biphenyl (PCB) poisoning: long-term follow-up interviews of the women from the Taiwan Yucheng cohort. Int J Epidemiol. 2000;29:672–677.
9.Windham G, Mitchell P, Petreas M, Lee D, Lasley B. Exposure to DDT and metabolites in relation to menstrual cycle length among Laotian immigrants. Epidemiology. 2002;13:S198.
10.Longnecker MP, Wolff MS, Gladen BC, et al. Comparison of polychlorinated biphenyl levels across studies of human neurodevelopment. Environ Health Perspect. 2003;111:65–70.
11.ATSDR. Toxicological profile for DDT, DDE, and DDD: update (Syracuse Research Corporation, Ed.). Atlanta, GA: U.S. Dept. of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry; 2000.
12.Niswander KR, Gordon M. The Women and Their Pregnancies. Pub no. (NIH) 73–379. Washington, DD: U.S. Government Printing Office, U.S. Department of Health, Education, and Welfare; 1972.
13.Longnecker MP, Klebanoff MA, Zhou H, Brock JW. Association between maternal serum concentration of the DDT metabolite DDE and preterm and small-for-gestational-age babies at birth. Lancet. 2001;358:110–114.
14.Longnecker MP, Klebanoff MA, Brock JW, et al. Maternal serum level of 1,1-dichloro-2,2-bis(p-chlorophenyl)ethylene and risk of cryptorchidism, hypospadias, and polythelia among male offspring. Am J Epidemiol. 2002;155:313–322.
15.Brock JW, Burse VW, Ashley DL, et al. An improved analysis for chlorinated pesticides and polychlorinated biphenyls (PCBs) in human and bovine sera using solid-phase extraction. J Anal Toxicol. 1996;20:528–536.
16.Hansen LG. Identification of steady state and episodic PCB congeners from multiple pathway exposures. In: Robertson LW, Hansen LG, eds. PCBs: Recent Advances in Environmental Toxicology and Health Effects. Lexington, KY: University Press of Kentucky; 2001;47–56.
17.Safe SH. Polychlorinated biphenyls (PCBs): environmental impact, biochemical and toxic responses, and implications for risk assessment. Crit Rev Toxicol. 1994;24:87–149.
18.Daniels JL, Longnecker MP, Klebanoff MA, et al. Prenatal exposure to low-level polychlorinated biphenyls in relation to mental and motor development at 8 months. Am J Epidemiol. 2003;157:485–492.
19.Laden F, Collman G, Iwamoto K, et al. 1,1-Dichloro-2,2-bis(p-chlorophenyl)ethylene and polychlorinated biphenyls and breast cancer: combined analysis of five U. S. studies. J Natl Cancer Inst. 2001;93:768–776.
20.Fitzgerald EF, Hwang SA, Langguth K, et al. Fish consumption and other environmental exposures and their associations with serum PCB concentrations among Mohawk women at Akwesasne. Environ Res. 2004;94:160–170.
21.Myrianthopoulos NC, French KS. An application of the U.S. Bureau of Census socioeconomic index to a large, diversified patient population. Soc Sci Med. 1968;2:283–299.
22.Brouwer A, Longnecker MP, Birnbaum LS, et al. Characterization of potential endocrine-related health effects at low-dose levels of exposure to PCBs. Environ Health Perspect. 1999;107(suppl 4):639–649.
23.Persky V, Turyk M, Anderson HA, et al. The effects of PCB exposure and fish consumption on endogenous hormones. Environ Health Perspect. 2001;109:1275–1283.
24.Koustras DA. Disturbances of menstruation in thyroid disease. Ann New York Acad Sci. 1997;17:280–284.
25.Gore AC, Wu TJ, Oung T, Lee JB, Woller MJ. A novel mechanism for endocrine-disrupting effects of polychlorinated biphenyls: direct effects on gonadotropin-releasing hormone neurones. J Neuroendocrinol. 2002;14:814–823.
26.Muller WF, Hobson W, Fuller GB, Knauf W, Coulston F, Korte F. Endocrine effects of chlorinated hydrocarbons in rhesus monkeys. Ecotoxicol Environ Saf. 1978;2:161–172.
27.Hansen LG. Stepping backward to improve assessment of PCB congener toxicities. Environ Health Perspect. 1998;106(suppl 1):171–189.
28.Phillips DL, Pirkle JL, Burse VW, Bernert JT Jr., Henderson LO, Needham LL. Chlorinated hydrocarbon levels in human serum: effects of fasting and feeding. Arch Environ Contam Toxicol. 1989;18:495–500.
29.De Vriese SR, Dhont M, Christophe AB. FA composition of cholesteryl esters and phospholipids in maternal plasma during pregnancy and at delivery and in cord plasma at birth. Lipids. 2003;38:1–7.
30.Treloar AE, Boynton RE, Behn BG, Brown BW. Variation of the human menstrual cycle through reproductive life. Int J Fertil. 1967;12:77–126.
31.Harlow SD, Matanoski GM. The association between weight, physical activity, and stress and variation in the length of the menstrual cycle. Am J Epidemiol. 1991;133:38–49.
32.Harlow SD, Campbell B, Lin X, Raz J. Ethnic differences in the length of the menstrual cycle during the postmenarcheal period. Am J Epidemiol. 1997;146:572–580.
33.Harlow SD, Campbell B. Ethnic differences in the duration and amount of menstrual bleeding during the postmenarcheal period. Am J Epidemiol. 1996;44:980–988.
34.Gordley LB, Lemasters G, Simpson SR, Yiin JH. Menstrual disorders and occupational, stress, and racial factors among military personnel. J Occup Environ Med. 2000;42:871–881.
35.Brown S, Vessey M, Stratton I. The influence of method of contraception and cigarette smoking on menstrual patterns. Br J Obstet Gynaecol. 1988;95:905–910.
36.Bean JA, Leeper JD, Wallace RB, Sherman BM, Jagger H. Variations in the reporting of menstrual histories. Am J Epidemiol. 1979;109:181–185.
37.Must A, Phillips SM, Naumova EN, et al. Recall of early menstrual history and menarcheal body size: after 30 years, how well do women remember? Am J Epidemiol. 2002;155:672–679.
38.Gammon MD, Wolff MS, Neugut AI, et al. Temporal variation in chlorinated hydrocarbons in healthy women. Cancer Epidemiol Biomarkers Prev. 1997;6:327–332.
39.Longnecker MP, Klebanoff MA, Gladen BC, Berendes HW. Serial levels of serum organochlorines during pregnancy and postpartum. Arch Environ Health. 1999;54:110–114.
40.Longnecker MP, Zhou H, Klebanoff MA, Brock JW. An unexpected distribution of sodium concentration in serum specimens stored for more than 30 years. Ann Epidemiol. 2003;13:178–181.
© 2005 Lippincott Williams & Wilkins, Inc.