Secondary Logo

Journal Logo

Ovarian Hormones in Premenopausal Women: Variation by Demographic, Reproductive and Menstrual Cycle Characteristics

Windham, Gayle C.1; Elkin, Eric1; Fenster, Laura1; Waller, Kirsten1; Anderson, Meredith1; Mitchell, Patrick R.1; Lasley, Bill2; Swan, Shanna H.1

ORIGINAL ARTICLES
Free

Background.  Ovarian function influences many areas of concern in women’s health, including breast cancer and other chronic diseases. However, ovarian function has been little studied in healthy, premenopausal women, partly because of cyclical variation.

Methods.  We measured biomarkers of ovarian function (daily urinary metabolites of estrogen and progesterone) among 411 women age 18–39 years, who were Kaiser Permanente members in Northern California in 1990–1991. We have summarized the hormone metabolite levels of about 1,500 cycles and examined their associations with demographic and menstrual cycle characteristics.

Results.  Cycles with a short follicular phase showed elevations of 10–13% in both baseline (days 1–5) and average follicular-phase estrogen metabolite levels, as well as some elevations in progesterone metabolites. Progesterone metabolite levels were directly related to the length of the luteal phase. Compared with whites, Hispanics had estrogen metabolite levels that were 7–13% higher in the follicular and luteal phases, whereas nonwhite, non-Hispanic women (primarily Asians) had slightly lower levels. Generally, women with a prior pregnancy or those with a later age at menarche had lower estrogen metabolite levels, whereas women with prior induced abortions had higher levels. Luteal-phase progesterone metabolite levels tended to be lower among women who were overweight, were less educated, were older at their first livebirth, or had an induced abortion.

Conclusions.  Some menstrual cycle characteristics provide a crude surrogate of the hormonal milieu, particularly luteal-phase length and progesterone levels. Hormone levels varied by reproductive characteristics, potentially explaining their relevance to breast cancer risk.

From the 1Department of Health Services, Division of Environmental and Occupational Disease Control, Oakland, and

2Institute of Toxicology and Environmental Health, School of Medicine, University of California, Davis, CA.

Address correspondence to: Gayle C. Windham, Department of Health Services, Division of Environmental and Occupational Disease Control, 1515 Clay Street, 17th Floor, Oakland, CA 94612; gwindham@dhs.ca.gov

This study was supported in part by funds from the California Legislature, as well as the California Tobacco-Related Disease Research Program (Grant 3RT-0093). The hormone laboratory work was in part supported by NIH Grant ESO-6198.

Submitted 21 August 2001; final version accepted 23 July 2002.

Ovarian hormone excretion is critically important to several aspects of women’s health. However, these hormones are difficult to study on a population basis because of their complex cyclical patterns in premenopausal women. The reproductive risk factors identified for breast cancer, such as young age at menarche and nulliparity, indicate a key role for hormones, with the estrogens particularly implicated. 1–4 Hormonal mechanisms for other reproductive cancers, 5,6 as well as other chronic conditions such as osteoporosis or fractures 7,8 and cardiovascular disease, 9 have also been suggested. Proper functioning of the hypothalamic-pituitary-ovarian axis is critical to ovulation and conception, as well as to maintenance of a pregnancy. 10–12 Thus, it is important to understand factors that may affect hormone production or metabolism.

Most prior investigations of hormone function in general populations are based on one, or at most a few, serum measurements. This limitation cannot reflect the large day-to-day changes that occur among premenopausal women. Comparison across studies is complicated because of the difficulty of timing the samples in relation to ovulation. Recently developed assays for urine metabolites of sex steroid hormones allow examination of integrated hormone production throughout an entire cycle in a noninvasive manner useful for epidemiologic studies. 13,14 Although a few studies have measured urinary metabolites, 15,16 none to our knowledge have examined daily measures. The Women’s Reproductive Health Study was designed to investigate factors related to menstrual function and early pregnancy loss in a relatively large, population-based sample of healthy reproductive-aged women who collected urine daily. 17,18 The purpose of the current analysis is to summarize the steroid metabolite levels during various phases of the cycle and to describe their interrelations. We also examine the relation of these hormone patterns to more easily obtained characteristics of the menstrual cycle. The study further provides the opportunity to identify predictors of ovarian steroid levels that may be related to adverse health outcomes.

Back to Top | Article Outline

Methods

The methods of the prospective Women’s Reproductive Health Study have been described previously 17,18 and are summarized below. The institutional review boards of both Kaiser Permanente and the California Department of Health Services approved the protocol and participants provided written informed consent.

Almost 6,500 women age 18–39 years, enrolled in the Kaiser Permanente Medical Care Program in Northern California during 1990–1991, were screened by a short telephone interview to identify those who could potentially become pregnant and were willing to collect and freeze first-morning urine samples daily for up to 6 months. Of 1,092 eligible women, 561 agreed to participate. Of these, 89 dropped out during urine collection and 61 changed eligibility status, leaving 411 women who completed urine collection. On average, women collected urine during 5.6 menstrual cycles, but because urine collection was not timed to the cycle start and end dates, a mean of 3.6 complete cycles were collected per woman. On average, women collected urine on 92% of appropriate days. Participants completed a detailed baseline telephone interview that asked about demographics, reproductive history, lifestyle factors and various exposures.

Back to Top | Article Outline

Definition of Endpoints

We analyzed urine samples for the estradiol metabolites, estrone sulfate and estrone glucuronide (estrone conjugates [E1C]), and the progesterone metabolite, pregnanediol-3-glucuronide (PdG), by enzyme immunoassays. 13 These were indexed for creatinine concentration, as reported previously. 18 An entire menstrual cycle was analyzed on a single microtiter plate that included standards and internal control samples. Precision of the assays was determined from pooled samples containing high, medium and low concentrations of analyte. The intraplate coefficient of variation (CV) for the E1C assay was 1.6%, whereas the interplate variations for the high, medium and low concentrations were 4.9%, 6.6% and 11.2%, respectively. The intraplate CV for PdG was 1.8%, whereas the interplate variation was 5.2%, 6.9% and 11.0%, respectively. We determined ovulatory status for each cycle on the basis of a relative rise in progesterone above baseline levels. 18,19 The day of ovulation was estimated using a previously validated algorithm. 18,20 We reassayed all cycles that were not ovulatory or in which the day of ovulation was day ≥20 of the cycle (N = 533); in 46% of these the “abnormality” was confirmed, whereas in 54% we used the more normal reassay results. 18 We also reassigned the day of ovulation in a small proportion of samples (5.6%) to correspond better to individual graphical plots of steroid metabolite levels. 18

To examine menstrual characteristics, we divided the cycle into the follicular phase (calculated from the first day of menses through the estimated day of ovulation) and the subsequent luteal phase. Cycle and phase lengths were categorized as short and long on the basis of the 5th and 95th percentiles of their distributions. Thus, “average” cycle, follicular-phase and luteal-phase lengths were 25–35, 12–23 and 11–14 days, respectively.

We examined hormone parameters in the cycles for which a day of ovulation was assigned; 1,451 cycles had complete follicular phases and 1,459 had complete luteal phases. The primary hormone parameters calculated for each cycle are defined below.

Back to Top | Article Outline

1. Baseline

The E1C baseline value was calculated as the mean over the first 5 days of the cycle. To avoid including elevated “spillover” progesterone values from the previous luteal phase, the PdG baseline is the mean over days 6–10 of the cycle. 19 Because this baseline period might overlap with rising PdG levels in short cycles, we examined alternative baselines, including the minimum 5-day moving average of PdG that occurred before the greatest 5-day average. To calculate this variable, 3 of the 5 days had to have nonmissing values.

Back to Top | Article Outline

2. Daily Average

Mean E1C and PdG levels were calculated over the entire cycle, the follicular phase and the luteal phase.

Back to Top | Article Outline

3. Area Under the Curve or “Total”

The sum of the daily E1C values during the follicular phase and that of daily PdG values during the luteal phase were calculated.

Back to Top | Article Outline

4. Peak

We calculated a 3-day average around the maximum E1C and PdG values, respectively. For the estrogen metabolite, the maximum daily value was selected within a 6-day window around the day of ovulation to capture the periovulatory peak. For progesterone, the maximum value during the luteal phase was selected. If values were missing for any of the 3 days the peak variable was not calculated.

In the regression models, we weighted each hormone parameter by the proportion of nonmissing values within the appropriate time frames defined above. Thus, if there were no missing values, the weight was 1; mean weights for each parameter ranged from 0.77 to 0.91. The mean values of the weighted hormone parameters were within about 1.5% of the nonweighted means.

Back to Top | Article Outline

Covariates and Statistical Analysis

We examined the correlations of pairs of hormone parameters using Pearson correlation coefficients (r). As correlations may be particularly affected by non-normal distributions, we also transformed the hormone parameters using the square root and recalculated the correlation coefficients. These correlations were even stronger, so we present the slightly more conservative, untransformed results. We examined measures of between- and within-woman variation for each parameter by calculating the mean and standard deviation for each woman across her own cycles (if she had at least two). We compared the standard deviation of the distribution of woman-level means (between-woman variation) with the mean of these woman-level standard deviations (within-woman variation) as components of the overall cycle level variation.

We analyzed associations with a number of covariates and menstrual cycle characteristics at the cycle level. 21,22 Because of the expected within-woman correlation, mixed models that account for repeated measures were used, 23,24 effectively increasing the standard error of the estimates. The compound symmetry covariance structure, which assumes that all repeated units (eg, cycles) within a woman are equally correlated, fit the data and was used in all models.

This analysis focuses on demographic and reproductive history variables. From a pregnancy history we calculated gravidity, parity, number of pregnancy losses, number of induced abortions, and age at first livebirth. Body mass index (BMI) was calculated from weight and height (kg/m2, classifying <19.1 as “underweight” and >27.3 as “overweight” according to the pre-1998 National Center for Health Statistics recommendation). All independent variables were categorized using cutpoints indicated in the s. We repeated the covariate analyses restricted to cycles of average length so as to limit influences of extreme cycle lengths. We also repeated the analysis restricted to parous women to examine age at first livebirth and other pregnancy variables.

Back to Top | Article Outline

Results

Participants were predominantly white, educated and parous (Table 1), with a mean age of 31 years. Figure 1 shows the daily values of E1C and PdG averaged over all participants’ cycles (centered on day of ovulation). This illustrates the characteristic periovulatory estrogen peak followed by a smaller estrogen rise during the luteal phase, and the rise and decline of progesterone during the luteal phase.

TABLE 1

TABLE 1

FIGURE 1

FIGURE 1

Back to Top | Article Outline

Properties of Hormone Parameters

The estrogen metabolite parameters during the follicular phase were generally highly correlated with each other (r = 0.71 to 0.84) and with the luteal-phase average (r = 0.66 to 0.71), but showed more moderate correlations with the baseline value (r = 0.42 to 0.71). The progesterone parameters during the luteal phase were even more highly correlated (r = 0.91 to 0.94), but less so with the baseline and the follicular-phase average (r = 0.38 to 0.54). The daily average of the estrogen metabolite during the entire cycle was highly correlated with its average during the follicular phase (r = 0.89), as well as with its peak (r = 0.83); similarly, the daily average of the progesterone metabolite throughout the cycle was highly correlated with its average during the luteal phase (r = 0.95) and with its peak (r = 0.90). The estrogen and progesterone metabolite parameters were only weakly correlated to each other (r = −0.003 to 0.13).

The variability of the hormone parameters across all cycles includes components of between- and within-woman variation (Table 2). Assay variation may also play a role, as well as time elapsed between laboratory assays. As expected, the mean variability within a woman’s cycles (Table 2, last column) is less than that between women by about one-third. Had we examined hormone parameters on the woman level, the variation would be about 20–25% less than for the overall cycle-based analysis.

TABLE 2

TABLE 2

Compared with average length cycles, cycles with a short follicular phase had elevations on the order of 10–13% in the estrogen baseline and follicular-phase average parameters, and a decrease in the estrogen total (Table 3). The baseline progesterone value was also elevated by 29%. Examining the minimum PdG from the 5-day moving average (intercept 0.53 ± 0.02), short follicular phases showed elevations but on the order of 16% (β = 0.09; 95% CI = 0.02 to 0.15). There was also some elevation of the luteal-phase progesterone parameters, particularly the peak. Long follicular phase was characterized by higher estrogen values, particularly the total (as might be expected given the greater number of days summed) and the luteal-phase average (Table 3). However, the daily average follicular-phase estrogen values were somewhat reduced.

TABLE 3

TABLE 3

Both long and short follicular phases were associated with lower progesterone parameters in the luteal phase of the previous cycle (Table 3). The hormone associations found by follicular-phase length were generally observed with short and long total cycle length as well (data not shown). This is expected because more than 84% of the variance in cycle length is attributable to follicular-phase length compared with 3% explained by luteal-phase length. 18

Cycles with short luteal phases generally had higher estrogen metabolite levels by 8–19% (Table 4). Progesterone parameters were generally decreased among cycles with short luteal phases and increased among those with long luteal phases (Table 4). The minimum PdG (alternative for the baseline) was similarly elevated in cycles with long luteal phases.

TABLE 4

TABLE 4

Back to Top | Article Outline

Demographic and Reproductive Covariates

Table 5 shows the adjusted mean differences for the E1C parameters by covariates; we did not include the results for the follicular-phase average in the because it was so highly correlated with the other estrogen parameters. The variables most consistently related in univariate analyses were gravidity, prior induced abortion, and older age at menarche, but the latter two were dampened after adjustment. In the adjusted model, estrogen parameters declined in a dose-response manner for number of prior pregnancies, decreasing by about 20% for three or more. In contrast, the estrogen parameters were greater among women who had prior induced (but not spontaneous) abortions. Most estrogen parameters were somewhat decreased among cycles of women with an older age at menarche. The adjusted estrogen parameters showed little association with age at first livebirth. Among women with livebirths, the estimates for history of induced abortion were greatly attenuated.

TABLE 5

TABLE 5

In the univariate models, Asians had somewhat lower estrogen metabolite values than whites, but less so in the multivariate models. Accordingly, they were combined with the small number of women of “other” races, who shared similar hormone values (Table 5). Except for the baseline measurement, Hispanics had estrogen parameter values 7–13% higher than whites, with a greater difference after adjustment.

The progesterone parameters were strongly related to high BMI, which had lower values of all the luteal-phase parameters (Table 6). Prior induced abortion and less education also were associated with lower progesterone values. Nonwhite, non-Hispanic women had a reduced progesterone baseline level (on the order of 15%), whereas Hispanic women did not differ much from white women by any PdG parameter. Cycles of women with an earlier or later age at menarche also had a lower baseline progesterone value. Older age at first birth (≥30 years) was strongly associated with lower adjusted luteal-phase progesterone levels (by 10–15%). Among women with livebirths, associations observed with reduced progesterone parameters were strengthened, such as those for history of induced abortion, higher BMI and lower education, and also for older age at menarche.

TABLE 6

TABLE 6

Back to Top | Article Outline

Discussion

A strength of this study is that it examines urinary steroid levels on a daily basis, through multiple cycles, in relation to demographic and menstrual cycle characteristics among a relatively large sample of premenopausal women. As the sample was not clinic-based and women had to have had a recent menstrual period to qualify, little selection by adverse health status should have occurred. Rather, as hormone parameters were examined only among ovulatory cycles, this sample probably had a lower-than-average likelihood of menstrual problems. The sample is not representative of the general population, in that subjects were willing to adhere to a labor-intensive system of urine collection and had to speak English; three-quarters of the women were white, and they tended to be highly educated and parous. This may have produced less variation in the hormone parameters within our sample than would be found in the wider population.

Although it is much more feasible to collect urine than serum for measurement of daily hormone levels, the urine levels reflect metabolism, which varies by woman depending on several intrinsic and extrinsic factors that we may not have been able to account for. Another potential limitation of the study includes the measurement of only one estrogen metabolite that may vary among women in how well it represents serum concentrations. In addition, we did not have duplicate samples for hormone assays. However, all cycles that were non-ovulatory or had an unusual day of ovulation were reassayed (24%), and the more normal and thus more conservative pattern was used for analyses. 18 Furthermore, the hormone parameters we defined were based on combining values from several days, which minimized the effect of any single extreme value.

Back to Top | Article Outline

Hormone Parameter Characteristics

Analyses were conducted with cycles as the unit of observation, accounting for repeated measures. Calculating mean hormone levels by woman would have masked the presence of within-woman variation (between cycles). Other than menstrual parameters and the amount smoked, our independent variables were primarily woman-level characteristics, but varying exposures may affect cycle level differences and should be examined in future research. As expected, the multiple parameters for each steroid metabolite were correlated and thus would not all need to be examined. The total area under the curve may be an important parameter clinically, as it represents a measure of the full “dose” or ovarian secretion of the hormone during the specified phase. This is influenced by phase length to some extent, whereas the daily average is less so. The estrogen peak does not add information over other E1C parameters, but if only limited sampling or assay is possible, measuring E1C on several days around ovulation may provide a marker of the other estrogen parameters. As the baseline levels were less correlated with the other hormone parameters and these levels are more likely to reflect non-ovarian function (eg, adrenal production), it appears useful to examine them separately. The generalized progesterone baseline is the lowest of several other time frames examined (days 3–7, 4–8 and 5–9) for typical cycles, but is influenced by short cycle length (or early ovulation). The minimum PdG determined individually for each cycle limits this problem, but requires a full cycle of data for calculation. Nevertheless, results by covariates were very similar for these two variables. In general, urinary estrogen and progesterone parameters were very weakly correlated.

Menstrual cycle characteristics presumably reflect underlying hormone patterns. However, the nature of this relation may be difficult to discern because of the variety of ways in which the hormones can be parameterized. We observed that cycles with short follicular phases had high baseline estrogen and mean estrogen during the follicular phase. Short follicular phases were also associated with higher progesterone parameters, particularly at the baseline and peak. Perhaps the hormonal feedback loop is triggered earlier in cycles with elevated baseline steroid levels, compressing time for follicle development. Women with short cycles will experience more cycles during their reproductive lifetime, thus receiving more frequent pulses of ovarian hormones and a greater daily average exposure. Cycles with a long follicular phase (which may indicate delayed follicle recruitment) had slightly lower daily average estrogen metabolite levels, but an elevation in the total follicular-phase E1C and the luteal-phase average. Another study that measured estrogen metabolites 25 similarly found higher average metabolite levels in short follicular phases and did not find appreciably lower levels in cycles with long follicular phases. In a study with daily plasma samples, 26 the follicular-phase length and cycle length were inversely related to various estradiol parameters (mean, peak and baseline), supporting our findings regarding short cycles.

We, as well as others, 26 found that luteal-phase progesterone parameters were directly associated with luteal-phase length. Therefore, measurement of luteal-phase length may be an adequate surrogate for relative PdG levels. In contrast, greater estrogen levels were observed in cycles with short luteal phases. Interestingly, a reduced progesterone response in the luteal phase was associated with alterations in the follicular-phase length of the subsequent cycle in our data. This observation is consistent with findings 27 of a relation between the follicle-stimulating hormone profile during the luteal-follicular transition and the length of the ensuing follicular phase. Thus the reduced progesterone parameters may reflect compromised corpus luteum function that has an associated perturbation of gonadotropin secretion.

Back to Top | Article Outline

Demographic and Reproductive Characteristics

Our strongest findings were with age at menarche, pregnancy history variables, BMI and (to a lesser extent) race. We found generally lower estrogen parameter values among cycles of women with older age at menarche, supporting an inverse relation noted by some serum studies, 28–30 but not all. 31,32 Progesterone values were also lower with older age at menarche among parous women. Consistent with our findings of reduced estrogen metabolite parameters with higher gravidity, Bernstein et al.33 found that parous women had a 20% lower follicular-phase estrogen level (day 11) in serum and urine than nulliparous women. In our study, age at first livebirth was not related to any estrogen parameter, but was inversely associated with progesterone parameters. Two studies 16,33 support these findings with estrogen, whereas another study 32 found an inverse association. The only study 16 that examined progesterone did not find an association. We found history of induced abortion associated with reduced progesterone, but with elevated estrogen. These associations with reproductive history were also observed in models limited to average-length cycles. Thus, these patterns do not appear to reflect variation in cycle characteristics, as confirmed by our previous analysis 18 in which we did not find differences in cycle or follicular-phase length with these variables.

Our observation that women with a high BMI had cycles with lower mean luteal-phase progesterone metabolite levels is consistent with other recent findings. 16,34 Gold et al.15 reported a higher peak PdG in women with low BMI; although we did not observe this, their finding supports an inverse association. BMI was not strongly related to estrogen parameters in our study, nor in several other studies of similar-aged, primarily parous women. 16,30,35

By ethnic group, we found that nonwhite, non-Hispanic women (primarily Asians) tended to have lower mean estrogen and progesterone metabolite levels than white women, but the associations were generally reduced after adjustment. Other studies 36,37 have found that Asians had lower estrogen levels than European women; one of these studies suggested different routes of estradiol metabolism. 37 In our study, Hispanics had higher estrogen metabolite levels than whites. However, the magnitude of this difference was reduced by about half among average length cycles, indicating that some of the hormonal differences may reflect the longer mean cycle length we found in Hispanics. 18 A study of parous women 16 did not report strong associations with race, but it was based on small numbers. In that study, Hispanics also tended to have higher urinary estrogen (12%) in the mid-luteal phase than whites. We are not aware of any study of follicular-phase estrogen levels in Hispanics. In addition to biological differences, variation by ethnicity may reflect life-style differences. Although we have information on a number of covariates, there are some, such as diet, for which we have no data.

We found little association between age (within 18–39 years old) and the estrogen parameters. Other study results are inconsistent with respect to age and luteal-phase estrogen values, 16,30,32 but several report elevated follicular-phase serum estradiol levels as age increases, within a similar reproductive age range as our study. 30,32,33,35

It is commonly accepted that ovarian hormones play a critical role in breast cancer risk, and there have been several hypotheses proposed for the mechanism. 2,4 Therefore, we examined various risk factors associated with breast cancer, by their steroid metabolite levels. Factors considered protective for breast cancer such as parity, older age at menarche, and Asian ethnicity were associated with lower estrogen parameter values. This supports the hypothesis that these factors affect breast cancer risk through an estrogen mechanism. Similarly, we observed higher estrogen values with history of induced abortion, a suspected risk factor for breast cancer. Increasing weight in premenopausal women is not associated with an increased risk of breast cancer, 2 and so the lack of an association of BMI with urinary estrogen is consistent. However, these findings do not preclude a possible relation in older women. Examining progesterone with respect to breast cancer risk factors, we found slightly lower progesterone parameter values with the protective factors of older age at menarche and nonwhite, non-Hispanic race/ethnicity. However, progesterone levels were also reduced with purported risk factors such as older age at first birth and induced abortion. 1 Short cycle length has been associated with an increased risk of breast cancer 1,38; we found that with cycles short follicular phases had higher baseline levels of both steroids, higher follicular-phase daily estrogen levels, and generally higher luteal-phase progesterone levels.

In conclusion, specific cycle characteristics appear to be representative of consistent alterations in steroid metabolite patterns, providing a crude surrogate of hormonal milieu. In particular, luteal-phase length is directly related to progesterone metabolite levels. Also, short follicular phases (and thus short cycles) had elevated baseline steroid and average follicular-phase estrogen metabolite levels, preceded by reduced progesterone levels in the previous luteal phase. Several reproductive characteristics were associated with variation in urinary steroid levels. It appears that these would not be revealed by a study of menstrual cycle characteristics alone, 18 as these variations occurred in average-length cycles as well.

Back to Top | Article Outline

Acknowledgments

We acknowledge the contributions of Robert Hiatt and Catherine Schaeffer of Kaiser Permanente’s Division of Research in the original design and conduct of the study. We thank David Epstein for managing the fieldwork, Ceciley Wilder and David Paniagua for manuscript preparation, and Sioban Harlow and Peggy Reynolds for valuable comments on earlier drafts.

Back to Top | Article Outline

References

1. Kelsey JL, Gammon MD, John EM. Reproductive factors and breast cancer. Epidemiol Rev 1993; 15: 36–47.
2. Pike MC, Spicer DV, Dahmoush L, Press MF. Estrogens, progesterones, normal breast cell proliferation and breast cancer risk. Epidemiol Rev 1993 17–35.
3. Rosenberg C, Pasternack BS, Shore RE, Koenig KL, Toniolo PG. Premenopausal estradiol levels and the risk of breast cancer: a new method of controlling for day of the menstrual cycle. Am J Epidemiol 1994; 140: 581–525.
4. Bernstein L, Ross RK. Endogenous hormones and breast cancer risk. Epidemiol Rev 1993; 1: 48–65.
5. Faerstein E, Szklo M, Rosenshein N. Risk factors for uterine leiomyoma: a practice-based case-control study. I. African-American heritage, reproductive history, body size and smoking. Am J Epidemiol 2001; 153: 1–10.
6. Risch HA, Marrett LD, Jain M, Howe GR. Difference in risk factors for epithelial ovarian cancer by histologic type: results of a case-control study. Am J Epidemiol 1996; 144: 363–372.
7. Sowers MR, Galuska DA. Epidemiology of bone mass in premenopausal women. Epidemiol Rev 1993; 15: 374–398.
8. Cooper GS, Sandler DP. Long-term effects of reproductive-age menstrual cycle patterns on peri- and postmenopausal fracture risk. Am J Epidemiol 1997; 145: 804–809.
9. Collins P. Vascular aspects of oestrogen. Maturitas 1996; 23: 217–226.
10. Baird DD, Weinberg CR, Zhou H, et al. Preimplantation urinary hormone profiles and the probability of conception in healthy women. Fertil Steril 1999; 71: 40–47.
11. Kolstad HA, Bonde JP, Hjollund NH, et al. Menstrual cycle pattern and fertility: a prospective follow-up study of pregnancy and early embryonal loss in 295 couples who were planning their first pregnancy. Fertil Steril 1999; 71: 490–496.
12. Harlow SD, Esphross SA. Epidemiology of menstruation and its relevance to women’s health. Epidemiol Rev 1995; 17: 265–286.
13. Munro CJ, Stabenfeldt GH, Cragun JR, Addiego LA, Overstreet JW, Lasley BL. Relationship of serum estradiol and progesterone concentrations to the excretion profiles of their major urinary metabolites as measured by enzyme immunoassay and radioimmunoassay. Clin Chem 1991; 37: 838–844.
14. Lasley BL, Shideler SE. Methods for evaluating reproductive health of women. In: Gold EB, Lasley BL, Schenker M, eds. Occupational Medicine: State of the Art Reviews: Reproductive Hazards. Philadelphia: Hanley and Belfus, 1994; 423–433.
15. Gold EB, Eskenazi B, Lasley BL, et al. Correlates of progesterone metabolite levels in women (Abstract). Am J Epidemiol 1998; 147: S56.
16. Westhoff C, Gentile G, Lee J, Zacur H, Helbig D. Predictors of ovarian steroid secretion in reproductive-age women. Am J Epidemiol 1996; 144: 381–388.
17. Windham GC, Elkin EP, Swan SH, Waller KO, Fenster L. Cigarette smoking and effects on menstrual function. Obstet Gynecol 1999; 93: 59–65.
18. Waller K, Swan SH, Windham GC, Fenster L, Elkin EP, Lasley BL. Use of urine biomarkers to evaluate menstrual function in healthy premenopausal women. Am J Epidemiol 1998; 147: 1071–1080.
19. Kassam A, Overstreet JW, Snow-Harter C, De Souza MJ, Gold EB, Lasley BL. Identification of anovulation and transient luteal function using a urinary pregnanediol-3-glucuronide ratio algorithm. Environ Health Perspect 1996; 104: 408–413.
20. Baird DD, Weinberg CR, Wilcox AJ, McConnaughey DR, Musey PI. Using the ratio of urinary estrogen and progesterone metabolites to estimate day of ovulation. Stat Med 1990; 10: 255–266.
21. Harlow SD, Zeger SL. An application of longitudinal methods to the analysis of menstrual diary data. J Clin Epidemiol 1991; 44: 1015–1025.
22. Martinez-Schenell B, Wilcox LS, Peterson HB, Jamison PM, Hughes JM. Evaluating the effects of tubal sterilization on menstrual function: selected issues in data analysis. Stat Med 1993; 12: 355–363.
23. Zeger S, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986; 42: 121–130.
24. Laird N, Ware J. Random effects models for longitudinal data. Biometrics 1982; 38: 963–974.
25. Harlow SD, Baird DD, Weinberg CR, Wilcox AJ. Urinary oestrogen patterns in long follicular phases. Hum Reprod 2000; 1: 11–16.
26. Landgren B-M, Unden A-L, Diczfalusy E. Hormonal profile of the cycle in 68 normally menstruating women. Acta Endocrinol (Copenh) 1980; 94: 89–98.
27. DeSouza MJ, Miller BE, Loucks AB, et al. High frequency of luteal phase deficiency and anovulation in recreational women runners: blunted elevation in follicle-stimulating hormone observed during luteal-follicular transition. J Clin Endocrinol Metab 1998; 83: 4220–4232.
28. Moore JW, Key TJA, Wang DY, Bulbrook RD, Hayward JL, Takatani O. Blood concentrations of estradiol and sex hormone-binding globulin in relation to age at menarche in premenopausal British and Japanese women. Breast Cancer Res Treat 1991; 18 (suppl 1): S47–S50.
29. Apter D, Reinila M, Vihko R. Some endocrine characteristics of early menarche, a risk factor for breast cancer, are preserved into adulthood. Int J Cancer 1989; 44: 783–787.
30. MacMahon B, Trichopoulos D, Brown J, et al. Age at menarche, urine estrogens and breast cancer risk. Int J Cancer 1982; 30: 427–431.
31. Bernstein L, Pike MC, Ross RK, Henderson BE. Age at menarche and estrogen concentrations of adult women. Cancer Causes Control 1991; 2: 221–225.
32. Dorgan JF, Reichman ME, Judd JT, et al. Relationships of age and reproductive characteristics with plasma estrogens and androgens in premenopausal women. Cancer Epidemiol Biomarkers Prev 1995; 4: 381–386.
33. Bernstein L, Pike MC, Ross RK, Judd HL, Brown JB, Henderson BE. Estrogen and sex hormone-binding globulin levels in nulliparous and parous women. J Natl Cancer Inst 1985; 74: 741–745.
34. Ecochard R, Marret H, Barbato M, Bodhringer H. Gonadotropin and body mass index: high FSH levels in lean, normally cycling women. Obstet Gynecol 2000; 96: 8–12.
35. Nagata C, Kaneda N, Kabuto M, Shimizu H. Factors associated with serum levels of estradiol and sex hormone-binding globulin among premenopausal Japanese women. Environ Health Perspect 1997; 105: 994–997.
36. Key TJA, Chen J, Wang DY, Pike MC, Boreham J. Sex hormones in women in rural China and in Britain. Br J Cancer 1990; 62: 631–636.
37. Aldercreutz H, Gorbach SL, Goldin BR, Woods MN, Dwyer JT, Hamalainen E. Estrogen metabolism and excretion in Oriental and Caucasian women. J Natl Cancer Inst 1994; 86: 1076–1082.
38. Whelan EA, Sandler DP, Root DL, Smith KR, Weinberg CR. Menstrual cycle patterns and risk of breast cancer. Am J Epidemiol 1994; 140: 1081–1090.
Keywords:

hormones; estrogen; menstrual cycle; progesterone; breast cancer; women’s health

© 2002 Lippincott Williams & Wilkins, Inc.