X/99/$03.00/0 Vol. 84, No. 11 The Journal of Clinical Endocrinology & Metabolism Copyright 1999 by The Endocrine Society

Transcription

1 X/99/$03.00/0 Vol. 84, No. 11 The Journal of Clinical Endocrinology & Metabolism Printed in U.S.A. Copyright 1999 by The Endocrine Society Prospectively Measured Levels of Serum Follicle- Stimulating Hormone, Estradiol, and the Dimeric Inhibins during the Menopausal Transition in a Population-Based Cohort of Women* HENRY G. BURGER, EMMA C. DUDLEY, JOHN L. HOPPER, NIGEL GROOME, JANET R. GUTHRIE, ADELE GREEN, AND LORRAINE DENNERSTEIN Prince Henry s Institute of Medical Research (H.G.B.), Monash Medical Centre, Victoria 3168, Australia; The Office for Gender and Health (E.C.D., J.R.G., L.D.) and Department of General Practice and Public Health (J.L.H.), The University of Melbourne, Melbourne 3050, Australia; Oxford Brookes University (N.G.), Oxford OX3 OBP, United Kingdom; and Queensland Institute of Medical Research (A.G.), Brisbane 4029, Queensland ABSTRACT The aims of this study were: 1) to describe, in relation to the date of final menses, the average hormone levels of women in the years before and after this date and to determine the extent to which these average levels were dependent on age and body mass index (BMI); and 2) to determine the degree of tracking in residual hormone levels [i.e., the extent to which individuals above (below) the mean for their age or time relative to final menstrual period (FMP) and BMI remain above (below) the mean as time progresses]. Serial levels of serum FSH, circulating estradiol (E 2 ), and the dimeric inhibins (INH) A and B were measured annually in 150 women who experienced a natural menopause during 6 years of follow-up. Means of the log-transformed hormonal levels were analyzed as a double-logistic function of time relative to FMP, as well as age and BMI and correlations between repeated hormonal levels, were measured. Mean FSH levels started to increase from about 2 years before the FMP, increased most rapidly about 10 months before the FMP, and had virtually plateaued by 2 years after the FMP. FSH levels were, on average, 3% greater for each year of age and 2% lower for each kg/m 2 of BMI. After adjusting for IT IS WELL established that major changes occur in the pituitary-gonadal axis in relation to the menopausal transition and the menopause (1). Underlying these changes is a marked decline in primordial follicle numbers in the ovaries (2). Cross-sectional studies have demonstrated that a large increase occurs in levels of serum FSH and a large decrease in levels of circulating estradiol (E 2 ) (3). Postmenopausal levels of FSH are 10 to 15 times higher than those seen in the early follicular phase of women of reproductive age, whereas E 2 levels are about 90% lower. A small number of previous Received March 8, Revision received July 22, Accepted July 30, Address correspondence and requests for reprints to: Prof. H.G. Burger, Prince Henry s Institute of Medical Research, Level 4, Block E, Monash Medical Centre, 246 Clayton Road, Clayton, Victoria 3168, Australia. henry.burger@med.monash.edu.au. * The Melbourne Women s Mid-Life Health Project is supported by Grants from the Victorian Health Promotion Foundation and the Public Health Research and Development Committee of the Australian National Health and Medical Research Council and the Australasian Menopause Society. Support for the Hormone Assays has also been provided by Organon Australia Pty. Ltd. time relative to the FMP, logfsh showed modest tracking. Ageadjusted values of logfsh were moderately correlated across time, and much of this tracking was explained by the actual timing of a woman s FMP. Mean E 2 levels started to decrease about 2 years before the FMP, decreased most rapidly around the time of the FMP, and had virtually plateaued by 2 years after the FMP. E 2 levels were lower, on average, by about 9% per year of age, and residual values showed weak tracking. Levels of both INHA and INHB decreased, on average, in the years before the FMP and were undetectable (INHA, 10 pg/ml; INHB, 25 pg/ml) in the majority of women by the time of the FMP and in almost all women by 4 years post-fmp. Significant negative correlations between log serum FSH and log E 2 (r 0.73) and log INHA (r 0.41) and log INHB (r 0.36) were observed. It is concluded that substantial changes in reproductive hormone levels occur within 1 2 yr on each side of the FMP, that falling concentrations of E 2 and the INH contribute to the rising concentrations of FSH, and that there is no single reliable hormonal marker of menopausal status for an individual woman. (J Clin Endocrinol Metab 84: , 1999) studies have described the pattern of hormonal changes in relation to the final menstrual period (4, 5), but none has examined the changes in the levels of the dimeric inhibins (INH) A and B. Few of these previous studies have been longitudinal in design, measuring changes prospectively. In this study, we present hormonal measurements from an ongoing longitudinal study of a population-based cohort of women aged yr and pre- or perimenopausal at baseline, whose experience of the menopausal transition and menopause is being recorded annually (3). Most of the blood samples drawn from cycling women were taken during the early part of the follicular phase. Because the dates of final menses have been identified prospectively, this study provides the first comprehensive description of the concomitant natural history of FSH, E 2, and the dimeric INH in relation to final menses. Subjects and Methods The recruitment strategy and baseline characteristics of the Melbourne Women s Midlife Health Project have been described previously in detail (3, 6). In summary, an initial sample of 2001 Melbourne women, between 45 and 55 yr of age, was recruited by random digit telephone 4025

2 4026 BURGER ET AL. JCE&M 1999 Vol 84 No 11 dialing. Of a subset of 779 women who had menstruated within the preceding 3 months, 438 (56%) agreed to participate in a longitudinal study that entailed annual interviews and an annual blood sample, so that their menopausal experiences and hormonal changes could be documented. To be eligible to participate in the longitudinal study, a woman had to have at least one intact ovary, not be taking hormone therapy, and to have menstruated in the last 3 months. A comparison of participants vs. eligible nonparticipants in the longitudinal study has been reported (3). The longitudinal study commenced in 1991, and the content of this study represents 6 yr of data collected up until Date of menopause At each interview, women were asked to complete a menstrual calendar recording dates of menstrual bleeding between interviews. At each follow-up interview, the menstrual calendar was collected and used to verify the date of her last menstrual period. A woman was deemed to have completed the menopausal transition and to be postmenopausal once she had reported absence of menses for 12 or more months. Thus, for this group of women, the date of the final menstrual period (FMP) was defined. Hormonal assays Fasting blood samples were drawn between the 4th and 8th days of the menstrual cycle or after 3 months of amenorrhoea. This protocol was achieved in 91% of samples. Serum FSH and E 2 were measured as described previously (3). Dimeric INHA and INHB were measured by specific ELISA, as specified by the authors (7, 8). INHA was measured in terms of the First International Standard for INHA (Human Recombinant, 91/624, 30 34k; NIBSC, Potters Bar, UK), expressed in terms of its nominal vial content (5 g). INHB standard was provided by the author (NG), the sensitivities for INHA and INHB ELISA, as determined from two sd above the mean blank value, were 10 pg/ml and 25 pg/ml, respectively. The between-plate within-assay variations for INHA and INHB ELISA were 6.4% and 7.9% (n 6), and between-assay variations from eight assays were 12% and 19%, respectively. INH assays were undertaken in Melbourne using the stored serum from the first 5 yr of the study only. Samples below assay sensitivity were given the value of assay sensitivity: INHA 10 pg/ml (n 300, 63%); INHB 25 pg/ml (n 357, 74%); and E 2 20 pmol/l (n 232, 29%). Study sample The present study is based on those 150 women who had experienced a natural menopause during the 6 years of follow-up, whose date of FMP was defined, and who provided at least one blood sample during followup. Excluded were women who experienced a surgical menopause through hysterectomy, bilateral oophorectomy, endometrial or iatrogenic ablation; and women who took hormone therapy before the cessation of menses. In total, 795 blood samples were taken (average, 5.3 per woman). Two thirds of women contributed a complete set of six blood samples, whereas one third contributed, on average, 3.9 samples due to either dropout from the study or refusal to give blood. Statistical analysis The women experienced menopause at varying times in the 6 years of follow-up. By aligning each measurement from a woman according to the date of her FMP, we were able to summarize the accumulated hormone data over a 9-yr time-scale (from 4 yr before through to 5 yr after the FMP). Geometric means were calculated for the following 6-month groupings: all samples obtained within 3 months on either side of the FMP were grouped and referred to as the time 0 samples. Samples obtained between 3 and 9 months on each side of the FMP were pooled and referred to as either 6 months before or 6 months after final menses, respectively. The other samples were pooled and grouped similarly in 6-month intervals, extending to 4 yr before and 5 yr after final menses. The numbers of samples analyzed at each time point are shown in Table 1. Modeling of log(fsh) and log(e2) as a function of time to menopause The longitudinal data were analyzed as repeated measures ordered by time (t), either as chronological age or as time relative to FMP, taking into account any correlation between successive observations within individuals. Log transformation of FSH and E 2 values was performed before analysis to ensure that the distributions of the residuals (observed values-fitted values) were close to normal. Nonlinear mixed effects analyses were used to model the means of the log-transformed hormone levels,, as a function of time, t (measured in years), according to the equation t h 1 2(h 1 h 1 )/{exp[ 0 (t )] exp[ 1 (t ]} 1 (age) 2 (BMI), where is the approximate time of maximum rate of change, h 1 is the maximum mean, h is the mean at time of maximum change, and 0 and 1 are parameters representing rates of change (9 11). This function was shown to provide a good description of the changing hormonal levels. The regression coefficients 1 and 2 represent the linear effects of age and body mass index (BMI), respectively, when adjusting for time. When both age and BMI were found to be statistically significant, their interaction was tested by the addition of the term 3 (age. BMI) into the equation. The correlation between repeated measures, adjusted for time and possibly age and BMI, was estimated as a function of the absolute time between repeated measures, t j -t k, where t j and t k represent the times of measures, and is a measure of the tracking of levels in an individual about the fitted mean. Tracking refers to the extent to which individuals above (below) the mean for their age or time relative to FMP and BMI, remain above (below) the mean as time progresses. The models were fitted by restricted maximum likelihood using an iterative procedure, assuming normal error structure (12). Analyses were carried out in SAS (SAS Institute Inc., Cary, NC) (13), using the macro NLINMIX. Because a large proportion of the INHA and INHB data were below the level of sensitivity, means only are presented, with the values below the threshold presumed to take the value of the threshold. Results By 1997, the 150 women had a mean (sd) age of 55.7 (2.3) yr (range, 50 61). Fifty-seven percent were in the paid workforce (full- or part-time), and 73% were married or living with a partner. The median parity was 3 (range, 0 9). Geometric means for FSH, E 2, and the dimeric INH at 6-month intervals around the FMP are shown in Fig. 1. For INHA and INHB assays, the percentage of samples that fell at or below the level of sensitivity was calculated and shown in parentheses in Fig. 1b. Geometric mean (95% confidence interval) FSH levels 4 yr before final menses were 10.5 ( ) IU/L. Increases in mean FSH levels were observed across the 9-yr time scale, but it was during the 18 months on either side of the FMP that TABLE 1. Number of blood samples analyzed for FSH, E 2, Inhibin A, and Inhibin B by six monthly time categories relative to FMP Hormone assay Time category (yr) relative to FMP Total FSH, E Inhibin A, inhibin B a a Inhibin assays were performed on blood samples collected before September 1996.

3 FSH, E 2, AND THE INHIBINS DURING MENOPAUSE 4027 FIG. 1. Geometric means (a, FSH and E 2 ; b, INHA and INHB) in relation to the FMP. The horizontal axis represents time (yr) with respect to the FMP (0); negative (positive) numbers indicated time before (after) the FMP. The parentheses above the time scale (b) indicate the percentage of measured INHA and INHB at or below the assay sensitivity. most of the change was observed. Mean FSH levels rose 5-fold over 3 yr, from 17.5 ( ) at yr, to 48.4 ( ) at the time of FMP, and to ( ) at yr. Mean E 2 levels were characterized by large variability in the premenopausal years, as shown in Fig. 2. Mean E 2 levels were 287( ) pmol/l at 4 yr and fell by 60% to 113 (83 155) pmol/l at the time of the FMP and to 35 (29 41) at yr. Between 2 and 5 years mean E 2 levels approached the sensitivity level for the assay (20 pmol/l). Mean levels of INHA averaged 25.0 ng/ml ( ) at yr before FMP and fell by 50% to 13.5 ( ) ng/l at the time of the FMP, with a minor fall thereafter. The proportion of values undetectable 6 months after FMP was 68%, and this rose to 94% by 4 years after FMP. Mean levels of INHB were 48.9 ( ) ng/l at yr before FMP, and fell by 40% to 28.5 ( ) at 6 months after FMP. INHB averaged 35.2 ( ) ng/l at the time of the FMP. The majority of INHB samples (83%) were undetectable at 6 months after the FMP, and this rose to 100% by yr. Log FSH was negatively correlated with the other hormones: loge2 (r 0.73), loginha (r 0.41), and log- INHB (r 0.36). Log(E2) was positively correlated with loginha (r 0.45) and loginhb (r 0.39); all P Modeling of FSH and E 2 When log(fsh) was modeled as a function of age, the fitted curve was increasing almost linearly from age FIG. 2. Observed and fitted levels [a, mean log(fsh) by age (yr), stratified by BMI groups; b, mean log(fsh) across the menopausal transition, stratified by BMI groups; c, mean log(fsh) across the menopausal transition, stratified by age categories]. The horizontal axis represents time (yr) with respect to the FMP (0); negative (positive) numbers indicated time before (after) the FMP.

4 4028 BURGER ET AL. JCE&M 1999 Vol 84 No 11 and then flattened abruptly to be a constant from about age 56 onward. The effect of BMI was significant; BMI 0.015, s.e 0.006, P The model for mean log (FSH) vs. age is shown in Fig. 2a, with separate curves drawn for varying levels of BMI (20, 25, and 30 kg/m 2 ). The correlation between the residuals were about 0.6 for measures one year apart, 0.4 for those two years apart, and plateaued at 0.3 for those 3 or more years apart. Log(FSH) was similarly modeled as a function of time relative to FMP. The nonlinear equations provided a good fit to the means for years before and after the FMP. In the best fitting model for log(fsh), BMI had a significant effect ( BMI 0.022, s.e 0.004, P ), such that the greater the BMI, the lower the level of serum FSH; age also had a significant effect ( age 0.034, s.e 0.011, P 0.01). The rate of increase in FSH reached a maximum at approximately 0.85 (s.e 0.45) yr, or 10 months, before the FMP. Mean values stabilized by about 2 yr after the FMP. The best fitting model for mean log(fsh) is shown in Fig. 2, b and c, with separate curves drawn for varying levels of BMI (20, 25, and 30 kg/m 2 ) and varying ages (45, 50, and 55 yr); the corresponding parameter estimates are shown in Table 2. The vertical displacement of these lines illustrates the small (although statistically significant) effects of BMI and age on mean FSH when compared with the change across the period under observation and the overall variation. The correlations between the residuals were modest and significantly greater than zero; they decreased from 0.4 for measures 1 yr apart, to 0.3 for those 2 yr apart, to 0.2 for those 3 yr apart, and plateaued at 0.2 for those 4 or 5 yr apart. When log(e 2 ) was modeled as a function of age, the fitted curve decreased almost linearly between the ages of 48 and 56 yr, then gradually approached the level of sensitivity for the assay (3.0 on the log scale); as shown in Fig. 3a. The effect of BMI was not significant ( BMI 0.004, s.e 0.008, P 0.6). The correlation between the residuals were about 0.4 for measures 1 yr apart, 0.2 for those 2 yr apart, and plateaued at 0.1 for those 3 or more years apart. Log(E 2 ) was similarly modeled as a function of time relative to FMP, the best fitting model of which is depicted in Fig. 3b, and corresponding parameter estimates are shown in Table 2. There was an independent age effect ( age 0.085, s.e 0.016, P ), but no independent BMI effect ( BMI 0.007, s.e 0.007, P 0.3). The maximum rate of change was estimated to occur at about the time of FMP. Again the three separate fitted curves are drawn at varying ages (45, 50, TABLE 2. Parameter estimates (SE in parentheses) for the best fitting models for mean log(fsh) and mean log(e 2 ) Parameter Log (FSH) Log (E 2 ) h 1 (maximum mean) 3.38 (0.61) 9.68 (0.80) (time at maximum growth 0.85 (0.45) 0.05 (0.17) rate) h (mean at maximum growth 2.31 (0.63) 8.92 (0.83) time) 0 (rate of change parameter) (0.0958) 2.15 (0.64) 1 (rate of change parameter) (0.315) (0.037) 1 (coefficient for BMI) (0.004) 2 (coefficient for age) (0.011) (0.016) FIG. 3. Observed and fitted levels (a, mean log(e 2 ) by age (yr); b, mean log(e 2 ) across the menopausal transition), stratified by age categories. The horizontal axis represents time (yr) with respect to FMP (0); negative (positive) numbers indicated time before (after) the FMP. and 55 yr), illustrating a strong age effect. The correlations between the residuals were weak, being 0.2 for measures 1 yr apart, 0.1 for those 2 yr apart, and negligible for those 3 or more years apart. Calculations were made of the specificity and sensitivity of FSH and E 2 measurements as markers of menopausal status. This was done by calculating for the overall data set, the level of serum FSH and of serum E 2 at the intersection of their fitted longitudinal curves with the date of final menses (107.9 IU/L and 88.5 pmol/l, respectively. The specificity was the percentage of premenopausal women whose values were below this level for FSH, above for E 2, and the sensitivity was the percentage of postmenopausal women above this level for FSH, below for E 2. The sensitivity of FSH and E 2 was 85% and 84%, respectively, and the specificity was 76% and 67%, respectively.

5 FSH, E 2, AND THE INHIBINS DURING MENOPAUSE 4029 Discussion This study has provided a comprehensive picture of the changes in FSH, E2, and INHA and INHB in relation to the time of final menses, or menopause, in healthy women. It has demonstrated that the effects of BMI and age on hormone levels are relatively small when compared with the overall variation between women during the menopausal transition. We demonstrated that analysis as a function of age gives a misleading impression. For example, although the average FSH level increases exponentially from age 46 to 54, this is not the experience of a typical woman. As shown in Fig. 2, the FSH profile will depend strongly on the timing of the woman s last menstrual period and, to a lesser extent, on her age and BMI. Similar comments apply to E 2. Our analysis has shown that the level of tracking with age is moderately strong for FSH and less for E 2. The correlation between log (FSH) levels 1 yr apart was about 0.6. Even log(fsh) measures taken 5 yr apart were correlated, after adjusting for age. These correlations are reduced somewhat by adjusting instead for time since FMP, indicating that much of the tracking of hormone levels with age as a woman passes through the menopausal transition, is explained by the actual timing of her FMP. Similarly for E 2, the age-adjusted and time since FMP-adjusted values were correlated over time. Also, the estimated effect of BMI on mean FSH levels was greater, and measured with more precision, once the time scale was adjusted for the FMP. The variability in serum E 2 levels is likely to be due to a substantial extent to the variability both within and between subjects, in the day of the follicular phase on which the single blood samples were obtained (between days 4 and 8). Whereas in young women of reproductive age levels of serum E 2 are generally stable during this part of the follicular phase, the latter is known to shorten as menopause approaches. It is, thus, likely that samples particularly those taken after days 5 or 6, may reflect the rising levels of E 2 as midcycle approaches in such older women. The logistics of the study made it impossible to confine sampling to days 3 or 4 of the cycle which might have been expected to lessen the degree of variability. Although changes at the level of the hypothalamo-pituitary axis have been described in relation to and after the menopause (14, 15), the marked decline in follicle number (2) suggests that the major explanation for the hormonal changes lies within the ovary. Because both the sex steroids and the INH are products of the granulosa cells lining ovarian follicles, the disappearance of follicles from the ovary would be expected to be manifest in declining abilities to secrete the steroids and the INH, as, indeed, is evident in the present study. As had been found in previous studies (4, 5), serum FSH levels rise progressively before final menses and continue to rise for the first 2 to 4 years after the menopause itself. Average levels of estradiol in the early follicular phase are maintained (albeit rather erratically) as women approach the final menses, and it is only in the year or so before final menses that levels fall substantially. Recent studies have suggested that estrogen levels may be, if anything, higher in women during the start of the menopausal transition than in their earlier reproductive years (16, 17). No previous studies have reported changes in the dimeric INH in relation to the menopause in a longitudinal study design. Two recent cross-sectional studies (18, 19) have described INHA, INHB, and FSH levels in younger and older cycling women and shown inverse relationships, particularly between INHB and FSH. Nevertheless, neither study examined the menopausal transition. Our study has shown that mean levels of both INHA and INHB fall substantially before the final menses, the major falls in INHA being observed in the last 18 months before final menses and INHB over a somewhat longer period. Current interpretations of the sources of INHA and INHB are in accord with the concept that INHA is primarily a secretory product of the dominant follicle and the corpus luteum and INHB of small antral follicles in the recruited cohort (8). INHB levels may be regarded as indicating the size of that recruited cohort. Whether it is falling levels of INHB, which account for the progressive though small rise in FSH with increasing age in regularly cycling women, is the subject of a current study in our laboratories. Preliminary evidence for this possibility has been obtained in a study where serum INHB levels were significantly lower in a group of older regularly cycling women selected to have raised early follicular phase serum FSH levels, in comparison with a control group of young women with levels of serum FSH in the normal young range (20), and more recent evidence was cited above (18, 19). Previous investigations of levels of mrna for the INH subunits are consistent with the above interpretations (21). Thus, subunit is found in both the dominant follicle and in other members of the recruited cohort. A subunit levels are confined largely to the granulosa cells of the dominant follicle and are subsequently demonstrable in the corpus luteum, whereas B levels are characteristically found in antral follicles, but not in the granulosa cells of the dominant follicle. On the basis of these findings and interpretations, we postulate that the falling levels of INHA as final menses approaches are due to the fact that a progressively larger proportion of menstrual cycles at this time are anovulatory, with failure of development of the dominant follicle. Hence, it is hypothesized that it is only those women who continue to ovulate close to the time of final menses who would be contributing significant INHA levels to the population mean. In contrast, it is hypothesized that the steeply declining follicle numbers would be contributing to the falling INHB levels, and the observation that a higher percentage of INHB measurements are undetectable compared with INHA would be consistent with the probability that a substantial number of women have few, if any, follicles remaining at the time of final menses. It should be emphasized that the present study provides no data about the prevalence of anovulatory cycles in the population studied. For the reasons of wishing to keep at a minimum the invasive aspects of our prospective longitudinal study of the menopausal transition, and also for consideration of feasibility and cost, data to substantiate the presence or otherwise of ovulation were not collected. In summary, this study confirms previous observations

6 4030 BURGER ET AL. JCE&M 1999 Vol 84 No 11 that there is no clear-cut biological marker of the menopause or FMP. Levels of FSH rise progressively at the time when final menses are experienced, whereas levels of E 2 and the dimeric INH fall. The falling concentrations both of E 2 and of the INH are likely to contribute to the rising levels of FSH. The substantial individual variation, as evidenced by Figs. 2 and 3, and the modest to weak tracking of individual levels, provides further strong evidence to support the previous proposals that the interpretation of isolated hormonal measurements in women during the menopausal transition cannot be used reliably to define their reproductive status on an individual basis. Acknowledgments David Robertson and his staff at Prince Henry s Institute of Medical Research provided the INH-A and INH-B assays, for which Nigel Groome (Oxford Brookes University, Oxford, UK) provided the reagents. Mr. N. Balazs and his staff in the Department of Chemical Pathology (Monash Medical Centre) provided the FSH and E 2 measurements. References 1. Burger HG The endocrinology of the menopause. Maturitas. 23: Richardson SJ, Senikas V, Nelson JF Follicular depletion during the menopausal transition: evidence for accelerated loss and ultimate exhaustion. J Clin Endocrinol Metab. 65: Burger HG, Dudley EC, Hopper JL, et al The endocrinology of the menopausal transition: a cross-sectional study of a population-based sample. J Clin Endocrinol Metab. 80: Rannevik G, Jeppsson S, Johnell O, Bjerre B, Laurell-Boruli Y, Svanberg L A longitudinal study of the perimenopausal transition: altered profiles of steroid and pituitary hormones, SHBG and bone mineral density. Maturitas. 21: Longcope C, Franz C, Morello C, Baker R, Johnston CC Jr Steroid and gonadotropin levels in women during the peri-menopausal years. Maturitas. 8: Dennerstein L, Smith AMA, Morse C, et al Menopausal symptoms in Australian women. Med J Austr. 159: Groome NP, O Brien M Two-site immunoassays for inhibin, and its subunits. Further applications of the synthetic peptide approach. J Immunol Methods. 165: Groome NP, Illingworth PJ, O Brien M, et al Measurement of dimeric inhibin B throughout the human menstrual cycle. J Clin Endocrinol Metab. 81: Preece MA, Baines MJ A new family of mathematical models describing the human growth curve. Ann Hum Biol. 5: Hopper JL, Hibbert M, Macaskill G, et al Longitudinal analysis of lung function growth in healthy children and adolescents. J Appl Physiol. 70: Hopper JL, Green RM, Nowson CA, et al Genetic, common environment, and individual specific components of variance for bone mineral density in 10- to 26- year-old females: a twin study. Am J Epidemiol. 147: Lindstrom MJ, Bates DM Nonlinear mixed effects models for repeated measures data. Biometrics. 46: SAS Institute Inc SAS/STAT User s Guide, Version 6, 4th Ed, Vol 1 and 2. Cary, NC: SAS Institute Inc. 14. Reame NE, Kelch RP, Beitins IZ, Yu M-Y, Zawacki CM, Padmanabhan V Age effects on follicle-stimulating hormone and pulsatile luteinizing hormone secretion across the menstrual cycle of premenopausal women. J Clin Endocrinol Metab. 81: Santoro M, Banwell T, Tortoriello D, Lieman H, Adel T, Skurnick J Effects of aging and gonadal failure on the hypothalamic-pituitary axis in women. Am J Obstet Gynecol. 178: Shideler SE, DeVane GW, Kalra PS, Benirschke K, Lasley BL. Ovarianpituitary hormone interactions during the perimenopause. Maturitas. 11: Santoro N, Brown JR, Adel T, Skurnick JH. Characterization of reproductive hormonal dynamics in the perimenopause. J Clin Endocrinol Metab. 81: Danforth DR, Arbogast LK, Mroueh J, et al Dimeric inhibin: a direct marker of ovarian aging. Fertil Steril. 70: Welt CK, McNicholl DJ, Taylor AE, Hall JE Female reproductive aging is marked by decreased secretion of dimeric inhibins. J Clin Endocrinol Metab. 84: Klein NA, Illingworth PJ, Groome NP, McNeilly AS, Battaglia DE, Soules MR Decreased inhibin B secretion is associated with the monotropic rise of FSH in older, ovulatory women: a study of serum and follicular fluid levels of dimeric inhibin A and B in spontaneous menstrual cycles. J Clin Endocrinol Metab. 81: Roberts VJ, Barth S, El-Roeiy A, Yen SSC Expression of inhibin/activin subunits and follistatin messenger ribonucleic acids and proteins in ovarian follicles and the corpus luteum during the human menstrual cycle. J Clin Endocrinol Metab. 77: