| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Department of Medicine, Division of Endocrinology, University of British Columbia, and Vancouver Hospital and Health Sciences Centre, Vancouver, British Columbia, Canada V5Z 1C6
 |
Abstract |
|---|
 Top Abstract I. IntroductionII. Defining the PerimenopauseIII. Classic Studies of...IV. Prospective Epidemiological...V. Systematic Studies of...VI. Histological Studies of...VII. Physiological Studies of...VIII. Hypotheses to Explain...IX. Hormonal Physiology of...X. Summary and Necessary...References |
|---|
|
I. Introduction |
|---|
|
TopAbstractI. IntroductionII. Defining the PerimenopauseIII. Classic Studies of...IV. Prospective Epidemiological...V. Systematic Studies of...VI. Histological Studies of...VII. Physiological Studies of...VIII. Hypotheses to Explain...IX. Hormonal Physiology of...X. Summary and Necessary...References |
|---|
Another difficulty in the study of the perimenopause is that it has been virtually ignored. The only mention of perimenopause in a recent two-volume gynecological reference text is its definition as "the time preceding the normal menopause during which declining ovarian function causes oligomenorrhea or dysfunctional uterine bleeding, symptoms of estrogen deficiency and elevated gonadotrophins" (11). That definition is in a chapter on premature menopause (11). This same reference publication devotes 160 pages to menopause (11). Likewise, two national consensus documents relating to midlife women do not mention the perimenopause (12) or, only in passing, indicate that perimenopausal women may find oral contraceptive therapy to be useful (13).
The purpose of this review is to integrate the best available data on hormonal levels, menstrual cycles, bone changes, and experiences of perimenopause as a midlife hormonal and sociocultural transition. The primary focus is to review hormonal and spinal bone change data from prospective studies in population-based samples of perimenopausal women. A second purpose is to integrate the available hormonal data into a hypothesis explaining the (patho)physiology of erratic and often high estrogen levels, high [but inconsistent (14)] FSH levels, and prevalent nonovulation in perimenopausal cycles (15, 16). Finally, this review will synthesize the available data to arrive at an understanding of perimenopause-related symptoms, bone density changes, and to point toward research that is still needed.
Cultural (17, 18) and social status (19) differences in the reporting of midlife symptoms and the meaning of menopause in different societies (20), although very important, have been reviewed elsewhere (9, 21, 22). In addition, because Sowers and La Pietra (2) and Khaw (23) have recently reviewed the epidemiology of menopause, and there are few ways to relate the endocrinological events to the age at menopause, this review will not examine the epidemiology of menopause nor factors contributing to age at onset of menopause (23, 24, 25, 26, 27). Finally, this review will not attempt to evaluate the sparse prospective evidence that risk factors for cardiovascular disease increase during the menopausal transition (2, 28, 29).
This review used a systematic search of Medline and Premedline references from 1990 to the present under the following search titles: human, perimenopause, prospective, hormone, vasomotor symptoms (VMS), bone. In addition, earlier epidemiological, histological, and endocrinological sources were used to obtain data on studies on women 45–55 yr of age published before 1990. These sources were supplemented with additional references cited in the papers that were reviewed. All data were assessed for their prospective design and appropriate prospective statistical analysis (e.g., each woman’s perimenopausal data compared with her own premenopausal data). Cross-sectional data were excluded unless no comparable prospective data were available.
For the purposes of this review, all sources were critically evaluated for inclusion of sampling methods, physical and sociocultural description of the participants, and methods of assessment of women’s experiences, hormone levels, and bone measures. (Because few hormonal studies contained adequate descriptive data on subject selection/exclusion and sociodemographic characteristics, no study was excluded on these grounds).
Primary data on serum hormone levels were all converted into SI units (estradiol was converted from picograms/ml to picomoles/liter through multiplication by 3.671; progesterone was converted from nanograms/ml to nanomoles/liter through multiplication by 3.18). A weighted mean was generated (in which a larger study influenced the mean data proportionately more than a small one). Many data were acquired from graphs or figures; some were in primary tabular form. The variances in SI units were derived from the primary data where possible; ranges are included where no variance could be obtained. A final option was to assume similar variances as those reported in other data sets for the studies lacking them. These data were then combined and analyzed using the Fisher’s method of combining P values (30). Because hormone assay standards vary, when perimenopausal data were reported, control data from pre- and menopausal women from the same study or, lacking that, from the same laboratory were included for comparison. [For example, the largest population from Melbourne Midlife Women’s Health Study (31) did not include any premenopausal control data. Therefore, comparison data were obtained from two studies of premenopausal women from the same laboratory (32, 33).]
Estradiol levels were generally reported for cycle days 4–7 after start of menstrual flow, which is in the early or midfollicular phase (depending on the follicular phase length). This sampling point will be designated simply as the follicular phase (FP) in this review, and the specific cycle days noted if they are different than days 4–7. These days were chosen for hormonal sampling by the Melbourne investigators (31) and those in several other centers and are also in common clinical use. Estradiol and progesterone levels obtained during the week before flow were designated as premenstrual (because ovulation was inconsistent and therefore use of the term "luteal" would often have been inappropriate). For urine hormone levels, it was difficult to compare and combine data from different studies. This is true, in part, because some used overnight data corrected for creatinine (34) and others used 24 h urine collections (35, 36, 37) without creatinine correction. Following the suggestion of Metcalf and MacKenzie (38), this review has analyzed the number of days in a cycle in which the total urinary estrogen excretion exceeded the normal midcycle peak (MCP) level in that assay system. The maximum number of days of MCP estrogen levels that should occur is 4 (median, 2; range, 0–4) (38). Using criteria for diagnosis of ovulation in each assay, pregnanediol glucuronide (PDG) levels were interpreted to show whether or not a luteal phase was present; a luteal phase duration of less than 10 days was considered to be a short luteal phase (16).
In the development of an integrating hypothesis for the hormonal changes of the perimenopause, this review sought explanations in the gynecological literature including studies of ovarian hyperstimulation for the purposes of in vitro fertilization (IVF) in women over 40 yr of age. The new hypothesis that excessive endogenous ovarian follicular stimulation occurs during the perimenopausal transition (39) is based on the physiology of intra- and extra-ovarian hormonal and paracrine controls of folliculogenesis as described in clinical studies of IVF and investigations in primates, and on cross-sectional anatomical studies of ovarian follicle numbers from ovaries of women of different ages who died or had ovariectomy.
A final aspect of this review is an effort to provide a hormonal or physiological explanation for the changes, symptoms, and signs reported by perimenopausal women. The data support the premise that the perimenopause includes a time of erratic estrogen production (both high and low) and that the times of high estrogen levels (especially coupled with anovulation) are causally associated with significant clinical manifestations such as short or long intermenstrual intervals, heavy flow, increased dysmenorrhea, breast tenderness and nodularity, emotional stress-type experiences, weight gain, and cyclic VMS. It is hoped that a more accurate understanding of the dynamic ovarian hormonal changes of the menopausal transition will lead to needed observational and therapy research and to rational, effective approaches to the education and clinical care of perimenopausal women.
|
II. Defining the Perimenopause |
|---|
|
TopAbstractI. IntroductionII. Defining the PerimenopauseIII. Classic Studies of...IV. Prospective Epidemiological...V. Systematic Studies of...VI. Histological Studies of...VII. Physiological Studies of...VIII. Hypotheses to Explain...IX. Hormonal Physiology of...X. Summary and Necessary...References |
|---|
There has been a general consensus for some years that the perimenopause ends when 1 yr has elapsed without menstrual flow (2, 40, 41, 42, 43). That is part of the perimenopause definition proposed by the WHO (10). However, menopause is defined by WHO as the "final menstrual period" (retrospectively defined as 1 yr without flow) (10). Therefore, during the last year of the perimenopause, women are also in the "postmenopause" as defined by WHO (10) or are in their first year after menopause (although they cannot know either classification except in retrospect). Because of the 12-month overlap in these definitions, it is difficult to understand whether the year beyond the final menstrual period is or is not included in the perimenopause, even when comparing papers from the same investigators using the same dataset, that focus, respectively, on defining the perimenopause (44), and on describing the normal menopause transition (45). Some reports in the literature even refer to menopause as beginning when 3 or 6 months have elapsed without flow (46), although, in statistical terms, the permanent end of flow requires 12 months without a menstrual period (47). Therefore, although the WHO definition of the perimenopause clearly includes the year after the final menstrual period, the end of the perimenopause and the beginning of menopause are problematic both in definition and in their use in the literature.
The definition of the onset of the perimenopause, like its end, also poses problems for epidemiologists, clinical investigators, physicians, and women. The WHO definition says that it is the period "immediately" (which implies a short span of time) before menopause when the "endocrinological, biological and clinical features of approaching menopause commence" (10). That definition implies that there are a typical set of features that will make onset of the perimenopause obvious. However, variability is the hallmark of the menopausal transition, and no operational definition was given of those features by the WHO (10). A definition of "the inception of perimenopause" for use in epidemiological studies was recently constructed from prospective data of the Massachusetts Women’s Health Study (based on a population-based sample of 1,550 women followed for 5 yr) (44). Perimenopause was defined as beginning when the woman reported "at least three but less than 12 months of amenorrhea" or "a self report of increased menstrual irregularity" (44). Those definitions had a positive predictive value of 0.70 for final menstrual period within the next 3 yr (44). For epidemiological purposes, that report and the earlier one (45) found the perimenopause duration to be an average of "about four" years. That is consistent with the 4.8-yr transition documented by Treloar (48) in a prospective study of cycle intervals and symptoms. Women’s self-reports of VMS or lighter flow were not very strongly predictive of the onset of perimenopause (44); however, clusters of experiences were not tested as a new factor in the prediction equation. In addition, women who reported a change in cycle regularity "at only isolated interviews" were classified as premenopausal rather than perimenopausal by at least one of the New England Research Institute studies (45), although all participants were over age 45 and 13% of the "premenopausal women" reported VMS (which would be consistent with the "clinical features" of menopause). Finally, about 10% of women apparently changed from pre- to menopausal status between two interviews separated by 9 months (45). [That, of course, is an impossibility if the 12 months of amenorrhea beyond the final menstrual period are considered in the perimenopause as defined by WHO (10).]
It is useful to have an epidemiological definition for the onset of the perimenopause (44). That is especially true because FSH levels increase gradually (49), are often intermittently high and normal, and are not diagnostic (14). If this review’s hypothesis about perimenopause physiology is confirmed, it is possible that inhibin levels or inhibin to estradiol relationships might provide a biochemical indicator of the onset of perimenopause. At present, however, a better definition of the onset of the perimenopause is needed for appropriate clinical care as well as for research. It is important for the WHO definition of perimenopause onset to be made into something that can assist a woman to understand what lies ahead for her and that can guide physicians in their care of women in their late thirties through their fifties.
Given the current absence of a biochemical, hormonal, or symptom-cluster marker for the onset of perimenopause, it is possible that women may be able to accurately determine when they begin to be perimenopausal. Dennerstein and colleagues (50) asked women in the Melbourne Women’s Midlife Health Study’s cross-sectional baseline survey of cycle regularity and flow changes to classify themselves related to whether they believed they had or had not started into the transition, were mid or late in the perimenopause, or had completed it. In ANOVA factor scores, premenstrual complaints were significantly related to dysphoria, vasomotor, skeletal, digestive, respiratory, and general somatic symptoms. These investigators speculate that "self reported menopausal status is a more sensitive measure of endocrine status" (50). However, the subsequent study of hormones in a subset of that Melbourne population did not attempt to use women’s self-classification (31), although it found follicular phase estradiol levels that were significantly higher than in premenopausal women reported from the same laboratory (32, 33). Based on the results of this review, an increase in or new onset of high estrogen-related symptoms may signal the onset of perimenopause. If such a symptom cluster were adopted, it is likely that the perimenopause would begin sooner and the transition last longer than an average of 4 yr, as is currently asserted (44, 45).
As outlined above, although extensive work as been done to achieve a consensus about the definition of the perimenopause (9, 10, 44, 45, 50), two difficulties may persist with the current definition: 1) the perimenopause ends at the end of the year after the final menstrual period, which means it can only be defined in retrospect, and it overlaps with the definition of menopause, and 2) the onset of oligomenorrhea or "irregular flow" may occur after several years of observable changes (that can be hormonally characterized as different from those of premenopausal women). The definition of menopause in retrospect is of little use to the clinician or to the perimenopausal woman prospectively experiencing the transition. A more appropriate definition of the start of menopause is to define it as beginning when a year has passed without flow. The likelihood of a subsequent normal period (e.g., not caused by endometrial hyperplasia, a polyp, or some other pathology) is less than 5% (47). This review includes the last year of no flow in the perimenopause as defined by WHO (10) (and also because hormonal data from many studies to be subsequently reviewed show variable and not consistently low estrogen levels during that time). For the purposes of clarity, this review will term menopause as beginning when a year has elapsed without flow and not as the "final menstrual period" (10).
The WHO definition of the perimenopause does not indicate particular hormonal characteristics. It is commonly understood, however, that this period of time is characterized by declining estrogen levels (6, 11). This review, in contrast, will demonstrate from published data that estrogen levels are highly variable and, in the early part of the perimenopause, estrogen levels average higher than during a woman’s reproductive life. Given the lack of clarity about the onset of the perimenopause, all studies were included in this review that characterized women between the ages of 45–55, that used the term "perimenopause," or that indicated that a woman’s periods had changed, or that the last menstrual flow occurred within the subsequent 4 yr. In the latter part of this review, an attempt will be made to integrate the clinical and hormonal features of the perimenopause studies into a proposed series of five phases based on clinical and hormonal characteristics.
|
III. Classic Studies of the Perimenopause |
|---|
|
TopAbstractI. IntroductionII. Defining the PerimenopauseIII. Classic Studies of...IV. Prospective Epidemiological...V. Systematic Studies of...VI. Histological Studies of...VII. Physiological Studies of...VIII. Hypotheses to Explain...IX. Hormonal Physiology of...X. Summary and Necessary...References |
|---|
B. Early reports of women’s experiences in the perimenopause
Between the time of Tilt’s study in the mid-1800s until the
1930s, menorrhagia disappeared from the perimenopause. [Wilbush, an
anthropological historian, believes this change occurred because
hysterectomy had become prevalent (51).] A new concept that the
"complaints" and experiences of midlife women were largely
psychological was prevalent in Britain and increased in the early years
of this century. The first therapy for menopause in the "modern
era" (e.g., when medicine had evolved beyond blood letting
and leeches) was apparently the early antiepileptic sedative,
phenobarbital. Subsequently, and apparently about the time estrogens
were isolated and synthesized in the late 1930s, the idea of menopause
changed again. Menopause began to be understood as a time of estrogen
deficiency. It was low estrogen levels that explained the psychological
symptoms of menopausal women.
By 1953, some "objective" way of characterizing the response of "patients exhibiting endocrine insufficiencies" was needed to monitor responses to estrogenic preparations (53). Blatt, Kupperman, and colleagues developed a systematic description of the timing and quantity of the flow response of a menopausal woman to a given dose and estrogen preparation. The estrogen preparation (and by 1953 there were 17 choices!) was given for 3 weeks followed by a single 100-mg injection of progesterone. The onset, duration, and amount of flow were used to derive an "Amenorrhea index," which predicted the potency of the estrogen (53). A "Menopausal Index" was also created from women’s "11 most common menopausal complaints," which has become the standard assessment tool for climacteric experiences (53, 54). The Kupperman and Blatt Index, which divides symptoms into psychological and physical, has been used in most subsequent studies. It included VMS, paresthesia, insomnia, nervousness, melancholia, vertigo, weakness (fatigue), arthralgia and myalgia, headaches, palpitations, and formications. It specifically did not include questions about menstrual flow, breast, weight gain, and fluid symptoms, perhaps because symptoms were a priori defined as occurring after menstrual flow had ceased, and because they were understood to be the consequence of estrogen deficiency.
Neugarten and Kraines (55), behavioral scientists with the Committee on Human Development in Chicago, used the previously developed Kupperman and Blatt Menopausal Index (53, 54), combined with symptoms reported by patients and their physicians to occur in midlife, to derive a 22-item menopausal symptom instrument which they published in 1964. Their new instrument included weight gain, flooding, irregular cycles, cold hands, trouble concentrating, and breast pain as well as feelings of suffocation. This new instrument was reproducible to retest within women (r = 0.8) and correlated well with the Kupperman and Blatt Menopause Index (r = 0.7).
The Neugarten Menopause Symptoms instrument was validated through its completion by women of different ages. The responses of a large sample of adolescent girls, midreproductive life women, women over 40 reporting some change in cycles (who were perimenopausal but whom the authors called "menopausal"), and women more than a year after their last menstrual flow were all compared (55). Perimenopausal women shared with adolescents (and not with normally menstruating women or those after menopause) the experiences of weight gain, cold hands and feet, feeling excitable and blue, and having headaches. Significantly more than any other group of women, the perimenopausal women experienced "weight gain, cold hands and feet, skin crawls, headaches, feeling blue, cold sweats, tired feelings, excitable, can’t concentrate, crying spells." The group of women who were more than 1 yr beyond their last flow were the least symptomatic of any of the groups (55). Neugarten and Kraines’ interpretion of their data stated: "the increased production of sex hormones in adolescence (signaled by the first menses) and the decreased production of estrogen (signaled by menopause) ... are primary in producing heightened sensitivity to and the increased frequency of reported symptoms" (55). That the menopausal women 1 yr after their last period were the least symptomatic was ascribed to an improved ability to cope (55).
C. Early prospective menstrual cycle interval and basal temperature
documentation
The first physiological, longitudinal study of the perimenopausal
transition was an anecdotal study of two women, reported in 1949 by a
physiologist, Mary Collett (56). This study documented the transition
into the menopause by yearly cycle interval averages and ranges in one
woman (recorded from 34–48 yr of age), and primary cycle interval data
from a second subject (from 41–57 yr of age). Basal metabolic rate
(BMR) was periodically documented in subject 1. The BMR showed a marked
decrease of 14.5–18%, which occurred at the age of 45, coincident
with the onset of her cycle interval irregularity. The rate continued
to be suppressed until age 48, when she had her last menstrual flow.
During her mid-forties she experienced an initial weight gain of 14 kg
(56). This was followed by "Evidences of endocrine imbalance ...
weight loss along with falling BMR, softening of the teeth at 46 to 48,
low hemoglobin... , hot flushes, great fatigue and extreme
nervousness." The other subject’s prospective data showed marked
menstrual cycle variability (from 19–89 days in length) beginning 4 yr
before the last flow (56). This initial study is still one of the few
that has followed women prospectively from the premenopause through to
the menopause.
Several large descriptive, prospective studies were begun in the late 1940s, at a time before hormonal measurements from urine or blood were widely available. Women were asked to record the day on which flow began on calendar cards, or calendars were coupled with basal temperature records (as a bioassay for progesterone action). These studies have documented menstrual cycle characteristics in women of different ages both cross-sectionally and prospectively (16, 47, 48, 57, 58). Collett performed one of the earliest published studies in which an unknown number of women recorded 302 cycles (57). The longest continuous sample was 176 cycles in a woman in her forties. Collett noted that with increasing age, cycle intervals shortened (from a mean of 30 days in the youngest to a mean of 25 in the oldest), that the follicular phase also shortened, and that anovulation increased to 15% of the cycles in the oldest women (57).
Two large prospective cycle interval studies, by Chiazze et al. (59) and Treloar et al. (47), showed shortening cycle intervals from a mean of 29 days for women in their twenties to a mean of 26 days for women in their forties. Treloar and colleagues prospectively followed more than 2,700 women over an average of 5 yr. A number of women recorded from their twenties until they were menopausal (47, 48). Although initially the sample only included women who were undergraduates at the University of Minnesota, it later included some of their daughters and friends. Annual questionnaires documented surgeries, pregnancies, and hormone use, which, along with age and postsecondary education, have been used in data analysis (47, 48).
The volunteers for Treloar et al. recorded the onset of flow on a menstrual card. No data about ovulation were available. The 120 women providing data through menopause (47) showed "heterogeneity," with an increase in both short and long cycle intervals (which was noted to be similar to that after puberty) (47). In a subsequent questionnaire, 14% of women reported hot flushes in the years preceding the cessation of flow (48). The cycle interval changes noted for women in their late forties and early fifties are similar to those documented by Doring [n = 67 (58)], by Collett et al. [n = 43 (57)], and by Vollman [n = 41 (16)], who sampled women from infertility clinics, from general advertising, or from a general medical practice, respectively.
The largest prospective basal temperature study was by a Swiss
physician named Vollman (16). More than 1 yr of prospective,
quantitatively analyzed rectal basal temperature data were documented
in 332 women, and the perimenopausal transition was prospectively
recorded in 41 women (16). [It is of interest that Vollman developed
the first quantitative method for evaluation of luteal phase onset from
basal temperature data—the day the temperature curve crosses the mean
temperature line correlates with the day of the midcycle LH peak day
(r = 0.891, P < 0.001)(60).] As shown
in Fig. 1
, which depicts the last 100
cycles of one woman (16), 9% of the cycles were less than 21 days in
length, and 10% were greater than 36 days long. In contrast to
Vollman’s documented 4% mean rate of nonovulation for women in their
mid-twenties to early forties, 16% of this woman’s perimenopausal
cycles were nonovulatory (16). Stratifying the data by gynecological
age (years from menarche), Vollman noted that 34% of cycles were
anovulatory in women of gynecological age 40–45. In addition, Vollman
(16) observed that, even within cycles that were ovulatory, an
increasing percentage of cycles of older women had short luteal phase
lengths of less than 10 days.
|
|
IV. Prospective Epidemiological Studies of the Perimenopause |
|---|
|
TopAbstractI. IntroductionII. Defining the PerimenopauseIII. Classic Studies of...IV. Prospective Epidemiological...V. Systematic Studies of...VI. Histological Studies of...VII. Physiological Studies of...VIII. Hypotheses to Explain...IX. Hormonal Physiology of...X. Summary and Necessary...References |
|---|
A. Manitoba Project on Women and Their Health in the Middle Years
The Manitoba Project, which began in the early 1980s, has provided
a wealth of information about the perimenopausal period of women’s
lives (7, 21, 40, 41, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77). Although the original sample was not
strictly population based, the design and subsequent conduct of the
study met the best of epidemiological principles. The design included
interview of women over the telephone at 6-monthly intervals for 3 yr.
The initial cohort of 477 women, ages 45–55, represented 87% of the
sample of an earlier cross-sectional study (40). An analysis of the
longitudinal reports of menstrual periods documented that more than 100
different patterns of menstrual flow (e.g., regular followed
by a skipped period, or no flow for 3 months followed by every other
month) occurred in just over 300 women (40). Menstrual flooding, in
addition to "menstrual problems, lack of energy, and nervous
tension," were the best predictors of a change from regular to
irregular cycles (40). In a subsequent analysis, depression scores over
16 on the Centre for Epidemiological Studies Depression Scale occurred
in 26% of all of the women at one time or another in the 3 yr of the
study. In a cross-sectional analysis of the prospective data, neither
change in menstrual cycle status nor children leaving home were
associated with depression, but having a hysterectomy was (76).
However, the potential for a causal relationship between the endocrine
changes that led to flooding and the mood symptoms appears not to have
been adequately considered in the Manitoba (76) study (or in the
Massachusetts study described below) (42).
B. Massachusetts Women’s Health Study
This important study, which was initiated in 1981, observed 2,570
women, aged 45–55, who were identified using census lists. This study
achieved a 77% initial response rate, and 93% of those participated
during the entire 4.5-yr prospective study (61), which has provided key
information about midlife women (21, 26, 42, 44, 45, 78, 79, 80, 81, 82). Women
were contacted by telephone every 9 months and were asked standardized
questions concerning their menstrual cycle status as well as other
experiences (similar to the questions and instruments used in the
Manitoba Project). Those who had menstruated in the preceding 3 months
and reported no change, or only intermittent changes in cycle
regularity, on one of two consecutive interviews were called
premenopausal; those with irregularity or 3–11 months of no flow were
called perimenopausal; and those who were more than 12 months beyond
their last flow were classified as menopausal. By these criteria, 1,178
of the sample were premenopausal at the initial telephone interview
(45). Excluding women with surgical menopause, the average age of onset
of the perimenopause was 47.5 yr, and menopause occurred at an average
age of 51.3 yr (26, 45). The average duration from onset of irregular
periods until 12 months past the last flow was approximately 4 yr.
Nulliparous women and smokers started the perimenopause at a slightly
younger age (45). Smokers and those starting perimeopause at a later
age experienced a shorter perimenopausal transition. Smokers were
approximately 2 yr younger at menopause (45).
Perimenopausal women tended to be more symptomatic (VMS, musculoskeletal symptoms, anxiety, fatigue, sleep disturbances, gastrointestinal concerns) than women menstruating regularly and those a year beyond their last flow. Approximately 20% of perimenopausal women visited their physician early in the transition with concerns about their menstrual cycles, while 12% sought help late in the transition because of symptoms they defined as menopausal (45, 61). It appeared that women with a longer perimenopausal transition were more likely to report VMS. Very few of the perimenopausal women were treated with ovarian hormone therapy or oral contraceptive agents (45, 82).
C. Kuopio Osteoporosis Risk Factor and Prevention (OSTPRE) Study
Tuppurainen and colleagues (46) reported on the design of the
OSTPRE study in which all 14,220 women ages 47–56 in Kuopio Province
were surveyed by a postal enquiry sent in 1989. This is the largest
population-based study of the menopausal transition to date but
includes no published hormonal nor menstrual cycle data. There was a
92.8% response rate to the questionnaire, which included
anthropomorphic and gynecological topics, use of medication, disease
diagnosis, and fractures. In addition to the lack of hormonal data, the
study has several drawbacks: menopause is defined as beginning after 6
months without flow; no category is provided for the perimenopause
(although the mean age of all women was 52); and 21.7% of women
aged 56 were still menstruating, probably because of hormone therapy.
As evidence of its careful epidemiological design, however, the Kuopio
study undertook a reproducibility analysis (repeat questionnaire
completion by a sample) that showed moderate correlations for milk and
cheese consumption (r = 0.548 and 0.469), high
correlations for weight, height, and oral contraceptive use
(r = 0.81 to 0.93), and low correlations for physical
activity in hours per day (r = 0.212, P
= 0.03) (46). In addition, they proved the validity of their
questionnaire for fractures against computerized medical records and
showed only three false-positive and three false-negative reports out
of a total of 2,935 fractures.
The epidemiological data from Kupio include age at hysterectomy. A total of 7,826 gynecological operations were performed on the 13,100 women who were followed longitudinally; 0.9% of these operations were because of cancer. Hysterectomy was performed in 15.8% of the population [which is lower than is reported from studies in the United States and Canada (67) but greater than in Japan]. The peak incidence of hysterectomy is 13.5/1,000 person years between the ages of 46–48 yr. The incidence was at or above 10/1000 person years from ages 40 to 52, which effectively spans the ages in which the perimenopause occurs in a North American population. These data are of importance because high estradiol levels and disturbed ovulation, which this review documents to occur during the perimenopause, are likely to be causal in abnormal menstrual bleeding problems (83), which are a common reason for hysterectomy.
The average age for the 13,100 women at the start of prospective observation in the Kuopio study was 52.4 yr; they had a body mass index (BMI) of 26.3 and consumed 817.2 mg/day of calcium from dairy products. Twelve percent of the population smoked regularly, and 8.7% had never been pregnant. At baseline, menopausal symptoms were reported by 55% of the "premenopausal" women (i.e., women who had experienced flow in the last 6 months) and by 69% of the postmenopausal women (who were 6 months or more beyond their last menstrual period). The bone and fracture data from this study will be reported in Section IX.C. (46).
The above three prospective studies of the epidemiology of the perimenopause have provided important information about women’s symptoms, the timing of events, and the process of the perimenopause. However, none of these studies has included detailed (more frequent than every 6- to 9-month telephone questionnaire) self-reports of experiences, primary menstrual cycle information, or hormone levels. To provide the needed direct hormonal information about the perimenopause, detailed physiological studies (the majority of which lack population-based validity) must be reviewed.
|
V. Systematic Studies of the Endocrinology of the Perimenopause |
|---|
|
TopAbstractI. IntroductionII. Defining the PerimenopauseIII. Classic Studies of...IV. Prospective Epidemiological...V. Systematic Studies of...VI. Histological Studies of...VII. Physiological Studies of...VIII. Hypotheses to Explain...IX. Hormonal Physiology of...X. Summary and Necessary...References |
|---|
. The designs of the hormonal studies in
perimenopausal women performed to date, except for the population-based
Melbourne Women’s Midlife Health Study (31) and the Malmo, Sweden,
study (84), have reported data from selected groups of women (often
gynecology patients or hospital staff paid for their participation) and
have not described the subjects adequately for comparison with other
studies. The Melbourne population-based study appropriately reported
the response rate (56%) and women’s descriptive characteristics (31)
but separated the hormonal data (31) and symptom information (50)
except for one paper cross-sectionally comparing follicular phase
hormone levels in women with and without VMS (85). Also, no primary
menstrual cycle data are available, although the hormonal analysis is
in five different groups based on reported menstrual cycle
characteristics (i.e., Group I includes women who reported
no change in cycle interval or flow in the last year, and Group V
designates women reporting no period for 3 months but who had
experienced flow within the last 12 months). Despite this
menstrual cycle classification (whose validity has been questioned)
(86), the Melbourne Women’s Midlife Health Study, as now reported,
does not allow integration of data on hormonal levels with
perimenopausal flow characteristics, vasomotor, and other experiences
of perimenopausal women.
|
, and the three with urine measurements are
listed subsequently. All are in chronological order by date of
publication. FSH and LH changes through the perimenopause, although
documented in many of the cited studies, will not be reviewed here.
These data confirm what has been known for many years, that with
increasing gynecological age, FSH levels increase first followed by LH
levels (32, 91, 93, 94), and that an elevated FSH level is not
diagnostic of perimenopause (14). These data also confirm that the
early follicular phase is the time in the cycle during which FSH is
most likely to first be elevated (91).
In contrast to the expectation that perimenopausal estradiol levels are
on the decline or are low, many studies have shown at least one cycle
or group of subjects in whom follicular phase estrogen levels were high
(31, 87, 88, 89, 90, 91, 95). This anecdotal observation of erratic high estradiol
levels, however, has been confirmed by mean data and by this
meta-analysis. Mean FP estradiol levels taken from 415 perimenopausal
women were compared with mean FP levels in 292 premenopausal controls
reported in 12 different papers (32, 33, 65, 89, 90, 91, 92, 93, 96, 97, 98, 99, 100). As shown
in Table 1, the mean FP estradiol level in perimenopausal women was
higher than in controls (224.9 ± 98 vs. 174.7 ±
57 pmol/liter). A Fisher’s combined P test of the four
studies that provided serum samples of both pre- and perimenopausal
women (65, 89, 91, 99) showed significantly higher estradiol levels in
the FP in perimenopausal women (Fisher’s F = 16.13, df =
2 k = 2 x 4 = 8; P = 0.041).
The most convincing evidence that FP estradiol levels are
inappropriately high during the perimenopause comes from the Melbourne
Women’s Midlife Health Study, which provides the largest number of
women cross-sectionally sampled in a well designed population-based
study. FP estradiol data, by cycle and flow-based categories, are
illustrated in (Fig. 2) (31). Average
estradiol levels did significantly decrease across the five
menstrual cycle groups with a lower level in the group who reported
skipping three or more cycles. Group I, who reported no change in cycle
interval or flow in the preceding year and were considered
premenopausal, had a FP estradiol geometric mean of 273 pmol/liter,
while that mean was 113 pmol/liter in Group V (31). However, excluding
data from Group I (who may have been either pre- or perimenopausal)
still leaves a weighted mean estradiol level of 226
pmol/liter, which is high compared with mean EF phase estradiol of 173
pmol/liter in normal young women controls in two studies published by
the same laboratory (184 ± 30 and 162 ± 73 pmol/liter,
respectively) (32, 33). The average estradiol levels are higher than
normal premenopausal FP phase levels for all except the last group
(86). In addition to the higher mean estradiol levels, more than 10%
of the individual values in every group exceeded the mean laboratory
normal midcycle estradiol peak level of 750 pmol/liter (32, 33) (Fig. 2).
|
is more striking because in
young women FP estradiol levels are low and range from 50–200
pmol/liter, with a mean of 175 ± 57 pmol/liter in the combined
premenopausal controls (Table 1). The high perimenopausal FP levels
published by Burger and colleagues (31) are, in fact, not different
from cycle mean estradiol levels (obtained by quantitative combined FP
and midluteal phase samples) from two cycles in each of 66 ovulatory
women studied in this laboratory (86, 101).
Estradiol levels during the premenstrual (or luteal phase) portion of
the cycle are also significantly elevated in perimenopausal women
(Table 1). The mean premenstrual estradiol level exceeded the mean in
premenopausal controls (371 ± 97 vs. 304 ± 84
pmol/liter). Although only a subset of studies provided systematically
studied data from both pre- and perimenopausal women (65, 89, 99),
Fisher’s combined P test shows perimenopausal levels to be
significantly elevated (F = 15.55, df = 2 k = 2 x
3 = 6; P = 0.0164) compared with normal
premenopausal values.
The three cross-sectional studies in which urinary hormonal data were used also showed elevated estrogen levels, this time as estrogen excretions. Before lending too much weight to these data, it is worth remembering that all of these studies sampled women with abnormal flow and ovulatory disturbances (63, 87, 88). Seven of the 12 cycles showed individual cycle estrogen excretions to be high (greater than MCP levels) for durations of 8, 12, 14, 6, 15, 5, and 5 days, respectively (63, 87, 88). In addition, dysfunctional uterine bleeding was observed to be associated with high estrogen excretions that occurred during flow (87). Thus, whether by urine or serum testing, one of the characteristics of the perimenopause is high mean estrogen excretions or estradiol levels, both in the FP and premenstrually. These data, however, do not indicate that every perimenopausal woman experiences high estradiol levels consistently, nor can we yet understand what percentage of women and for what periods of time will be exposed to high estrogen levels.
Studies of ovulatory characteristics or progesterone levels are few in
perimenopausal women in contrast to the many studies that have
documented estradiol levels or estrogen excretions. Of the 12 studies
of perimenopausal endocrinology shown in Table 1, all of those
measuring urine excretions but only two of those measuring serum
hormones tabulated the percentage of the cycles that were ovulatory. In
paid midwifery students studied by Abe et al. (90), five of
eight cycles were nonovulatory. Two studies excluded otherwise eligible
women because they were nonovulatory (89, 100). Despite excluding 24%
of the sample, the midluteal progesterone levels averaged 15.75
nmol/liter in women ages 45–50 compared with 39.65 nmol/liter in women
ages 20–29 (P < 0.05) (89). That suggests ovulatory
but insufficient luteal phase cycles were occurring. Although Ballinger
et al. (65) do not categorize cycles as nonovulatory or
ovulatory, they showed significantly lower mean progesterone values
(16.3 vs. 25.3 nmol/liter) in women in the later
perimenopause compared with controls. Furthermore, Ballinger et
al. showed that mean progesterone levels over 16 nmol/liter
(e.g., indicating ovulation) occurred on only one rather
than two of the weekly samples in women 
45 yr old compared with those
less than 45 (65). These data suggest ovulatory but short luteal phase
cycles, as previously shown by Vollman (16), occur more
frequently in the menopausal transition. Studies by Lee et
al. (91) and Fitzgerald and co-workers (98) did not document any
age-related differences in progesterone levels. Unfortunately, The
Melbourne Midlife Women’s Health Study provides no information about
ovulation (31).
The three cross-sectional studies in which 24-h urinary PDG levels were
reported (Table 1), as previously mentioned, all sampled women with
abnormal bleeding or from gynecological patient populations (63, 87, 88). In these three studies, anovulation occurred 77% of the time.
This appears to exceed the frequency of anovulation from other studies
of the perimenopause (15, 16). This lends support to a probable causal
relationship between dysfunctional uterine bleeding and lower levels of
progesterone (83). Therefore, these cross-sectional data confirm the
earlier basal temperature studies (16, 57, 58) showing a significantly
increased frequency of cycles having ovulation disturbances
(anovulation, short or insufficient luteal phases) during the
menopausal transition.
In summary, these cross-sectional data describing the serum hormonal
characteristics of one cycle each from 415 perimenopausal women show
that mean estradiol levels, both in the follicular and premenstrual
phases, are significantly higher than in younger women. Although this
observation may be caused by sampling relatively later in a shortened
FP, or in some cycles in which midcycle bleeding rather than normal
menstruation was occurring (86), approximately 10% of levels are
extremely high and exceed the normal MCP estradiol level (Fig. 2) (31).
Also, when cycles are aligned by their MCPs as is done by Santoro and
colleagues (34), FP values remain higher in perimenopausal than in
premenopausal women. In addition to higher FP estradiol values,
premenstrual estradiol levels are also higher than in premenopausal
controls. In general, these papers do not comment on the high estrogen
levels they document. This omission may occur because high estrogen
levels were contrary to their anticipated results (86). Finally,
perimenopausal women have a higher (by almost 50%) frequency of
anovulation compared with women in their twenties and thirties (16).
Despite the many studies and data from large numbers of women that
strengthen these conclusions, prospective data are needed to confirm
them and to understand the physiological changes leading to higher
estrogen productions and disturbed ovulation that appear to
characterize the perimenopause.
B. Prospective ovarian hormonal levels in the perimenopause
Eight studies, spanning the years from 1976 to 1996, have recorded
hormone levels over several menstrual cycles in each of a total of 248
women. These data are reported in detail in Table 2. Although some studies report only
summary data, 193 cycles are shown in detail in the respective
publications (34, 38, 61, 84, 93, 102, 103, 104). These data differ from the
single-cycle cross-sectional data in that they allow some appreciation
of the cycle-by-cycle hormonal variability of the perimenopause. All
except three of the prospective hormonal data sets have been obtained
from 24-h or overnight urine samples rather than serum measurements
because it is less invasive. As shown in Table 2, high and prolonged or
inappropriately timed (such as the start of flow) and high estrogen
levels (as high as the MCP) are documented in several studies (34, 38, 87, 102). Although it is counterintuitive, both Longcope et
al. (103) and Metcalf and MacKenzie (38) documented high estrogen
levels during long cycles, or, in one instance, in a perimenopausal
woman who subsequently did not have further vaginal bleeding.
|
Based on urinary PDG excretion in the prospective perimenopausal data from two women, only 56% of the 38 cycles were ovulatory. The patterns of urinary estrogen, mucus, and bleeding show anovulatory cycles with high premenstrual estrogen levels followed by cycles with short FP that are normally ovulatory. Typically, that ovulatory cycle might be followed by a very long FP cycle with a short luteal phase length. In a few instances, the well-trained subjects did not record stretchy mucus despite high estrogen and low PDG excretions. That observation suggests some decrease in cervical gland responsiveness to estrogen may develop during the perimenopause (102).
Three prospective studies reported in Table 2 used serum sampling (84, 93, 103). Only the Sherman study (93) sampled daily across several
cycles in three women while the others measured serum levels at 3- to
6-month intervals in 39 and 152 women, respectively (84, 103). Sherman
and colleagues summarized their data to say that FSH levels were rising
and estradiol levels were declining in women ages 46–56 with regular
cycles. However, when the hormonal data are examined more closely, they
include several characteristics suggesting intermittent high estrogen
levels. For example, in 4 of 13 cycles shown in detail, estradiol
levels over 260 pmol/liter occurred at the time of flow, a time when
levels have normally dropped. In addition, 3 of 13 cycles showed
clearly short FP lengths (93). Normative data for menstrual cycle
hormone levels show a significant inverse relationship between the FP
length and the mean FP estradiol level (r = -0.551)
(96). Therefore, the data of Sherman et al. (93), although
not documenting high mean estradiol levels, show intermittent high
values. Also, two of the three women whose cycles were shown in detail
had at least one anovulatory cycle. Longcope’s study, which sampled 39
peri-menopausal women at 3- or 4-month intervals for 2.5 yr, found
that the weighted mean estradiol level was 294 ± 327 pmol/liter,
which is clearly not low (103) compared with a mean cycle estradiol
level of 275 ± 112 pmol/liter in 66 ovulatory premenopausal women
(101).
The Malmo Perimenopause Project, which includes 152 women who
contributed serum hormone samples every 6 months during the interval
from 80 months before up to more than 48 months after the last flow
(84), has recently published the largest and best designed set of
prospective data. Unfortunately, as in the Longcope study (103),
serum samples were not timed within cycles. Data are presented
according to how many months the women were before or beyond their
final menstrual period. This presentation indicates that the mean
estradiol level was 379 pmol/liter during the peri-menopause (as
defined by the 48 months before and the 12 months after the last flow).
The mean serum estradiol level, however, was 459 pmol/liter if only the
4 yr before the last flow were included (84). Even in the
year after the last flow, during which the mean estradiol levels were
182 ± 163 and 171 ± 151 pmol/liter (for the first and the
final 6 months, respectively), these estradiol levels were normal for
the premenopausal FP (Table 1).
The Malmo study also shows a high estradiol variance (as SD of 427 for a mean of 513 pmol/liter) when women were 4 yr before the last flow (84). This variance may be due to the collection of serum with no relationship to flow or may be a vivid expression of the variability as well as the elevation of estradiol levels during the perimenopausal transition. In both the Malmo and the Massachusetts study by Longcope, the estradiol variance did not become less than 50% of the mean until approximately 24–48 months had passed since the last flow (84, 103). Thus, the three prospective studies of the perimenopause in which serum hormonal sampling was used, like the cross-sectional data, show high estradiol levels.
The first of several prospective studies in which urinary hormone measurements were used focused on the incidence of ovulation disturbances during three consecutive cycles in women 40 to 55 yr old (61). Metcalf documented that anovulation occurred in more than 50% of women reporting a recent history of oligomenorrhea (who were classified as perimenopausal). Anovulation, however, was present in only 5% of cycles in those who reported regular cycles, even if they were more than 50 yr old (61). Ovulation was inconsistent in 66% of women who noted recent changes in cycle length. In two other reports from New Zealand with similar design, 17 of 31 perimenopausal women ovulated within 16 weeks of their final menstrual period (15) or 52% of peri-menopausal women were ovulatory (66). Midcycle spotting occurred for 12 of 139 women; 5 of these had regular and consistently ovulatory cycles, and 4 of the 7 with irregular cycles also had inconsistent ovulation (61).
The prospective urinary hormone studies from New Zealand described
above included no estrogen data (15, 61, 66). However, some years later
the urinary estrogen excretion data from the same set of subjects were
reported indirectly as a ratio of estrogen excretion divided by PDG
excretion (107). In 22% of the cycles, the ratio of estrogen to PDG
excretion at or above 100 (which is typical of the midcycle estrogen
peak) lasted for a prolonged time of 14 or more days. In their 16
normal premenopausal women, the urinary midcycle ratio of hormones over
100 averaged 2 days and had a range of 0–5 days (107). Extremely high
ratios (>100) occurred on only 18% of cycle days in younger women
while ratios over 100 were present on 40% of the measurement days in
the perimenopausal women (P < 0.01) (107).
Furthermore, as mentioned above, the longer the cycle the more likely
it was that a prolonged high ratio would occur. Ratios over 100
for 14 days occurred in 6.9% of cycles that were 18–35 days
in length, and in 46.9% of cycles that were 50–260 days long (107).
Although low PDG levels, as expected in long anovulatory cycles might
be postulated to explain this, a ratio over 100 required high estrogen
excretion (which was not anticipated in oligomenorrhea).
The two further prospective studies using urinary steroid excretion
levels corrected for creatinine were authored by Shideler and
colleagues (104) and by Santoro and colleagues (34). These two studies
show from three to six cycles of daily urinary data in detail for each
of several women (Table 2). Furthermore, Santoro’s group
systematically compared urinary hormonal data from 11 regularly cycling
women over 47 yr old (whom she called perimenopausal) with data from
younger women also studied across one cycle. These data, which are
illustrated in Fig. 3, show significantly
increased estrogen excretions in both the follicular and premenstrual
phases of the cycle in the older women (34). The mean estrone level of
76.9 ng/mg Cr is clearly greater than the mean level of 40.7 ng/mg Cr
in controls (34). The final important observation from these data is
that PDG excretion was significantly less in regularly cycling women
over 47 yr old compared with younger women (P < 0.015)
(34) (Fig. 3). These data confirm lower progesterone excretion levels
during the early perimenopause and suggest, in contrast to the Metcalf
data, that short or insufficient luteal phases may begin before
anovulation develops and that these subtle ovulation disturbances are
present before menstrual cycles become irregular or oligomenorrhea
develops. Data from the groups of Shidelar and Santoro show that high
estrone conjugate excretion (at greater than midcycle levels) exists
for prolonged periods that last from 3–13 days (34, 104) as
illustrated in Fig. 4. Unfortunately,
Santoro et al. do not record times of vaginal bleeding on
this graph or allow a calculation of the FP lengths which Shideler
et al. noted to be short during several cycles (104).
|
|
Both the cross-sectional (Table 1) and prospective (Table 2) published
studies on the endocrinology of the perimenopause just reviewed show
consistent evidence of significantly higher estrogen levels than are
normal in younger women. This observation contrasts sharply with our
current understanding. Is it possible that, until recently, estrogen
levels did gradually decrease during the perimenopausal transition?
That is highly unlikely given that the time of acquisition of these
data spans more than 20 yr, and that evidence of high estrogen levels
such as menorrhagia and breast tenderness have been reported to occur
during the perimenopausal transition since Tilt’s time, and short FPs
since the 1950s (57). The endocrinology of the perimenopause has not
recently changed. Rather, the construct that menopause is a time of low
estrogen production likely led to an inability of scientists and
clinicians to see the (intermittently) high estrogen levels that were
present (86).
|
VI. Histological Studies of Ovarian Changes Across the Lifespan |
|---|
|
TopAbstractI. IntroductionII. Defining the PerimenopauseIII. Classic Studies of...IV. Prospective Epidemiological...V. Systematic Studies of...VI. Histological Studies of...VII. Physiological Studies of...VIII. Hypotheses to Explain...IX. Hormonal Physiology of...X. Summary and Necessary...References |
|---|
It is well established that, from birth onward, primordial follicles
are continuously being activated, maturing partially, and then
regressing. This follicle activation continues in a constant pattern
that is independent of pituitary stimulation. Evidence suggests that
this regular follicular activation changes during late reproductive
life. Richardson and colleagues (111) performed a careful quantitative
histological study of the ovaries and endometrium coupled with a single
hormonal measurement and a reproductive history on a randomly selected
ovary from each of 17 women ages 44 to 55 who underwent ovariectomy and
hysterectomy for uterine fibroids or menorrhagia. (Although these are
important data, the women were a selected and abnormal sample because
they apparently required hysterectomy.) The six women reporting regular
cycles had an average of 1700 follicles in the ovary compared with an
average of 180 follicles in the ovaries of those who reported irregular
cycles (111). The endometria from the perimenopausal women showed no
evidence of ovulation although three of seven, by timing, should have
been in the luteal phase. Furthermore, in contrast with this review’s
reported endocrinology of the perimenopause (Tables 1 and 2), estrogen
levels were equal in the premenopausal and perimenopausal women and
ranged from 146 to over 450 pmol/liter, which is well within normal
ranges for the menstrual cycle (96).
The steady number of follicles undergoing partial stimulation and atresia without any hormonal intervention does not appear to increase at puberty (112), although few peripubertal ovaries have been examined thus far. The marked increase in pituitary stimulation of the ovary at puberty does not appear to alter that follicular "depletion rate" (112, 113). A number of pathologists have performed quantitative analysis of the remaining ovarian follicles in ovaries of women from random autopsy series or women having ovarian surgery (111, 112, 113, 114). Only recently, however, has the remaining number of follicles been related to the age of the woman to construct an age-specific rate of apparent follicle disappearance (113). When the cross-sectional ovarian follicle count data are ordered by the age of the woman examined, observations show a marked increase in the slope of follicle loss that appears to indicate an accelerated rate of follicle depletion beginning in the late thirties or early forties (111, 112, 113, 114).
What are the implications of an accelerated rate of follicle depletion? It seems obvious that more follicles would disappear if more were being stimulated consequent to gonadotrophin levels that begin to rise in women in their thirties (49). Another possibility is that the cohort of follic
