American Journal of Epidemiology
ª The Author 2007. Published by the Johns Hopkins Bloomberg School of Public Health.
All rights reserved. For permissions, please e-mail: [email protected].
Vol. 166, No. 4
DOI: 10.1093/aje/kwm104
Advance Access publication June 8, 2007
Original Contribution
Work-related Factors Associated with Age at Natural Menopause in a Generation
of French Gainfully Employed Women
B. Cassou1, L. Mandereau2, P. Aegerter3, A. Touranchet4, and F. Derriennic5
1
Université de Versailles Saint Quentin, Hôpital Sainte Périne, Paris, France.
INSERM U 170, Villejuif, France.
3
Département d’Information Hospitalière et de Santé Publique, Université de Versailles Saint Quentin, Hôpital Ambroise Paré,
Boulogne, France.
4
Inspection Médicale Régionale du Travail des Pays de Loire, Nantes, France.
5
INSERM U 88, Saint Maurice, France.
2
Received for publication August 10, 2005; accepted for publication March 2, 2007.
This study’s purpose was to identify occupational factors that may influence the age at natural menopause in
a random sample of gainfully employed French women born in 1938 (n ¼ 1,594). Occupational physicians selected
the subjects from their files and interviewed them during their annual visits in 1990 and 1995. The authors used
Kaplan-Meier survival curves to estimate median age at menopause (52 years) and multiple Cox models to
estimate associations among women’s characteristics, occupational factors, and age at menopause separately
within two strata distinguished by a self-reported history of depression. Among women without such a history,
earlier menopause was associated with smoking more than 10 cigarettes per day in 1990 (p < 0.001), a high-strain
job (p ¼ 0.01) in 1990, and difficult schedules before 1990 (p ¼ 0.03). Later menopause was associated with higher
educational status (p ¼ 0.003) and repetitive work in 1990 (p ¼ 0.005). Among women with a history of depression,
a later menopause was associated with having at least one child (p < 0.001) and menarche later than the age of
13 years (p ¼ 0.004). Earlier menopause was associated with a high job control in 1990 (p ¼ 0.03) and high
school education (p < 0.01). These results suggest that certain physical job stressors may be related to age at
menopause.
aging; depression; education; France; menopause; occupations; smoking; women
Abbreviation: ESTEV, study on health, work, and aging.
The average age at menopause in populations in North
America and Europe ranges from 48 to 52 years (1). Some
authors suggest that age at natural menopause may be
a marker of health and aging (2, 3), and a better understanding of the determinants of its onset may therefore have
important clinical and epidemiologic implications (4).
Behavioral, reproductive, social, and demographic factors
(5, 6) are known to be related to age at menopause, probably
intertwined with genetic influences (7, 8). Smoking has consistently been found to be related to this age (9–12), with
natural menopause occurring earlier among women who
smoked compared with those who never did. Other factors
observed with varying degrees of consistency to be associated with earlier natural menopause include menstrual
cycles of less than 26 days (13, 14), nulliparity (9, 10, 12–
15), and obesity (10, 12, 15). Similarly, a self-reported history of medically treated depression is more frequent among
women with early natural menopause than among those
with average or late menopause (16). On the other hand,
early menarche (17), oral contraceptive use (17, 18), and
Correspondence to Dr. Bernard Cassou, Laboratoire Santé-Vieillissement, Hôpital Sainte Périne, 49 rue Mirabeau, 75016 Paris, France
(e-mail: [email protected]).
429
Am J Epidemiol 2007;166:429–438
430 Cassou et al.
marriage (15, 18, 19) are associated with later onset. Relatively few surveys (20) have examined other factors (e.g.,
alcohol use, exercise, and diet). The magnitude of the effects
of most of these factors on age at menopause is in the range
of 1–2 years. Prenatal factors, such as exposure in utero to
diethylstilbestrol, may also influence age at menopause (21).
Although higher socioeconomic class appears to be a relatively strong predictor of later menopause (9, 15, 22–24),
the respective roles in this relation of such factors as ethnicity, education, place of residence, marital status, income,
occupation, and working conditions remain unclear. At the
same time, scientific evidence (25, 26) is accumulating
about the detrimental effects of workplace activities and
exposures (such as intense physical activity and exposure
to toxic products) on reproductive health (fertility disorders,
adverse pregnancy outcome, and irregular menstrual
cycles). Two studies (27, 28) suggest that higher body burdens of organochlorides may be associated with earlier menopause, but little is known about the effect of other working
conditions on the timing of ovarian senescence.
To further our understanding of the subject, we used data
from the French longitudinal study on health, work, and
aging (ESTEV; enquête santé, travail, et vieillissement) that
included a large representative sample of workers from various occupations to examine the determinants of age at natural menopause and especially the relation of work-related
factors other than toxic exposures with age at menopause.
MATERIALS AND METHODS
Study population
Data for this analysis come from the French ESTEV
study, a prospective longitudinal epidemiologic investigation designed to identify occupational factors that might
modify the development of a variety of health characteristics. The population and methods of this study have previously been described in detail (29). The sample population
was randomly selected from exhaustive lists of workers in
the private sector who saw occupational physicians for legally required annual examinations. These physicians were
recruited through regional associations that covered seven
socioeconomically contrasting regions of France (Bretagne,
Dauphiné, Ile de France, Nord, Pays de Loire, Val de Loire,
and Ile de la Reunion). In all, 387 physicians (approximately
60 percent of those initially approached) volunteered to
participate.
The ESTEV study focused on four specific cohorts, representing four different generations born in 1938, 1943,
1948, and 1953. Each physician classified and filed patients’
records according to sex and year of birth. From each of
these eight files, they used a specific sampling coefficient to
randomly select an equal number of participants of each age
for the sample. We used available national data to define
these eight sampling coefficients.
The ESTEV survey took place in two waves, in 1990 and
1995, during annual occupational medical examinations. In
1990, 21,378 participants who volunteered for the first phase
took part, including 8,932 women born in 1938, 1943, 1948,
and 1953. Only women born in 1938 were eligible for this
analysis, however, because most of them had experienced
either natural or artificial menopause by 1995. This birth
cohort contributed 2,089 women to the study in 1990. The
rate of participation (87 percent) did not vary between regions by more than one percentage point. In 1995, 1,778 of
these women were seen again, for a follow-up rate of 85
percent. Women born or living on the island of La Reunion
(n ¼ 110) were excluded from the analysis because their
lifestyle and reproductive history differed so greatly from
those of women living in metropolitan France. Menopausal
status was unknown for 15 women, and date of menopause
was unknown for 59. After all exclusions, the analysis included 1,594 women who had experienced menopause.
Data collection and measures
Data on occupational history and working conditions
came from the participants with additional verification by
the occupational physician in 1990 and 1995. We assessed
working conditions with a 30-item questionnaire, derived
from previous ergonomic and epidemiologic studies in
France (30). We verified the intelligibility of the questions
by a validation study of approximately 200 workers. There
were three possible responses for each question: exposed at
work at the time of the first interview (in 1990), not exposed
in 1990 but exposed in the past, never exposed. For those
who reported exposure in 1990, the occupational physician
checked the consistency between each worker’s self-reports
and the workplace information in the medical files.
Women were considered exposed to physical constraints
at work if they reported at least one of the following difficulties: carrying heavy loads, exposure to vibrations, and
exertion required to operate tools or machines. They were
considered to be exposed to certain physical stressors
(standing work, high noise, which was defined as the impossibility of hearing someone at 2 m, work for 48 hours/
week or more, and shift work) if they answered the following question positively for each physical stressor: ‘‘During
your job, have you been exposed to this stressor?’’ Performance of repetitive work under time constraints was assessed by the question: ‘‘Have you been exposed to
repetitive work under time constraints?’’ They were considered exposed to difficult schedules if they reported at least
one of the following difficulties: getting up early or night
work. Duration of exposure to these working conditions was
classified as never exposed, exposed for less than 10 years,
and exposed for 10 or more years.
Job-related psychological stress, characterized in terms of
job control and job demand, was assessed with an abbreviated version of the instrument developed by Karasek (31).
Women were considered to have low job control if they
reported at least one of the following factors in 1990: lack
of means (material, information, or time) to perform highquality work, lack of opportunity to choose how the work
should be carried out, no diversification of work, or no opportunity to learn new things at work. Job demand was
considered high when they reported at least one of the following factors in 1990: need to rush, need to do several
things at the same time, or frequent interruptions while
working. We used job control and job demand to construct
Am J Epidemiol 2007;166:429–438
Work-related Factors and Age at Natural Menopause
a four-category job strain indicator based on Karasek’s
model: 1) high strain, when job demand was high and job
control low; 2) low strain, when job demand was low and
job control high; 3) passive, when both were low; and 4)
active, when both were high.
Socioeconomic category was defined by classifying occupation in 1990 into one of the four basic INSEE (Institut
National de la Statistique et des Etudes Economiques) categories: 1) white-collar workers, that is, senior and junior
managers and technical workers; 2) clerks, that is, clerical
and sales personnel; 3) blue-collar workers, that is, foremen,
skilled and unskilled workers, and service workers; and 4)
other occupations. Educational level was classified in three
levels: less than 7 years of education (primary), between 7
and 11 years (high school), and more than 11 years of education (at least some university). Marital status in 1990 was
defined as married, single, widowed, or divorced. Unmarried women cohabiting with someone (n ¼ 42) were classified with married participants. Physicians asked about
smoking habits and recorded the age at which participants
began smoking, how long they had smoked, and how much
(number of cigarettes per day). Height and weight, measured by the occupational physician, were used to calculate
body mass index (weight (kg)/height (m)2). We chose to
divide the sample into three tertiles of equal size: low
(22.5), intermediate (22.6–25.8), or high (25.9) body
mass index groups because conventional analysis would
have resulted in a very limited number of women with a body
mass index of 19 or less. The French version of the Nottingham Health Profile was used to evaluate self-perceived
health status and especially emotional reactions (32).
The physician recorded past and present health problems
during the clinical examination in 1990 and 1995. During
the face-to-face interview in 1990, the physician also asked
whether participants ‘‘had ever been diagnosed with depression’’ and ‘‘at what age was depression first diagnosed?’’
Data about reproductive history variables (measured in
years) were collected at the clinical examination: age at first
menstruation and at first and last childbirth, number of livebirths, use of contraceptive pills (yes or no, age at first use,
age at last use, and duration of use), and use of hormone
replacement therapy. Women were asked to report their age,
in years, at their last menstrual period. They were also asked
whether the occurrence of their menopause was natural and,
if not, the circumstances under which it had occurred. A
woman was considered to be naturally postmenopausal
when, at the time of the clinical examination in either
1990 or 1995, at least 12 consecutive months had passed
without menstruation not due to surgery or medical causes
(medication, radiotherapy), but age at natural menopause
was defined as age at cessation of menses (n ¼ 1,163).
For the 48 women who had not yet experienced menopause
in 1995 (28 perimenopausal and 20 still menstruating regularly), age at menopause was censored at the year of the
second interview (1995). The natural menopause group included 53 women who began hormone replacement therapy
during the perimenopausal period and for whom age at
menopause could therefore not be accurately determined.
They were censored at the date of the beginning of the
treatment. Women whose menopause was medically inAm J Epidemiol 2007;166:429–438
431
duced (n ¼ 122) were censored at the year of artificial
menopause, and women whose menopause was induced
by hysterectomy and/or oophorectomy (n ¼ 261) were censored at the date of surgery.
Statistical analysis
Several analyses were performed to examine age at natural menopause. Social, demographic, reproductive, and occupational factors were analyzed as qualitative predictor
variables. A composite index based on smoking status and
number of cigarettes/year was calculated for smoking exposure. Another composite index assessed oral contraceptive
use, based on duration and starting age. Bivariate relations
between variables and menopause were described by
Kaplan-Meier curves. They were assessed by the log-rank
test and the likelihood ratio test of a univariate Cox model.
The proportional hazard hypothesis was verified by a graphic
method. Because we observed a strong interaction between
self-reported history of depression and job strain on time to
natural menopause (p < 0.004), we categorized women according to whether they reported a history of depression
before performing separate multivariate modeling. All variables associated with the risk function of menopause with
a p value of less than 0.2 were then introduced as candidates
in a multiple proportional hazards model. Variables were
recoded as binary when it was possible to do so without loss
of information. Starting with a full model with all candidates, we performed a descending stepwise procedure based
on the likelihood ratio test and the Akaike Information Criterion to remove uninformative variables. Some variables
were very strongly correlated, such as marital status and
parity or smoking habits and use of oral contraceptives.
Nevertheless, including these variables in multiple analyses
did not produce collinearity errors, such as high estimator
variance. Finally, we tested for interactions between the
remaining covariates. The simplest adequate coding of the
relation between the covariate and the response was chosen
after exploration of trends with both continuous and ordered
categorical covariates. Since the stratum with history of depression was relatively empty, we verified the stability of
predictors by a resampling procedure involving 200 bootstrap samples. Statistical analyses were performed with the
R software package (http://www.R-project.org/), with the
final significance level set at 5 percent.
RESULTS
The estimated median age at menopause for the entire
sample was 52 years. The mean age at menopause was
51.0 years for naturally postmenopausal women. A total
of 83.2 percent of naturally postmenopausal women (n ¼
1,007) reported menopause between 45 and 55 years of age,
7.5 percent (n ¼ 91) before 45 years, and 9.3 percent (n ¼
113) after 55 years.
Table 1 shows the relations between the social and health
variables and median age at menopause. Earlier menopause
was associated with smoking (for current smokers, smoking
at least 10 cigarettes/day) (p < 0.001) and self-reported
432 Cassou et al.
TABLE 1. Median age (years) at menopause according to social, demographic, and
medical variables in a cohort of 1,594 gainfully employed French women born in 1938
Variables
No. of
women
% of total
sample
Median age
(years)*
Log-rank
p value
16.1
51
0.30
Social category
Executives
257
Clerks
830
52.1
51
Blue-collar workers
463
29.0
52
44
2.8
51
Other
Marital status
Married
1,125
70.6
52
Single
130
8.2
52
Widowed
169
10.6
50
Divorced
169
10.6
51
Missing data
0.16
1
Education
Primary
837
52.5
51
High school
507
31.8
52
University
230
14.4
51
20
1.3
1,278
80.2
52
81
5.1
52
Missing data
0.04
Smoking status in 1990 and intensity
Never smoked
Former smokery and 10 cigarettes/day
Former smoker and >10 cigarettes/day
82
5.1
52
Current smoker and 10 cigarettes/day
70
4.4
52
Current smoker and >10 cigarettes/day
77
4.8
50
6
0.4
Missing data
<0.001
Body mass index (kg/m2) in 1990
22.5
477
29.9
51
22.6–25.8
570
35.8
52
25.9
543
34.1
52
4
0.2
Missing data
0.07
Depression
Never
In 1990 or before
1,355
85.0
52
239
15.0
50
<0.001
* Menopause-free survival curve.
y Participant had stopped smoking at least 1 month before the interview in 1990.
history of depression (p < 0.001). For body mass index,
median age at menopause was lower for the leanest women;
this difference was nearly statistically significant (p ¼ 0.07).
Women with a high school education had the highest median age at menopause (p ¼ 0.04). A history of depression
was more frequent among women who reported exposure to
high job strain (p < 0.04) and repetitive work (p < 0.004)
(data not shown) in 1990. No association was observed with
social category or marital status.
Table 2 summarizes the relations between reproductive
variables and median age at menopause. Earlier menopause
was associated with oral contraceptive use for less than 10
years (p ¼ 0.05). Women who had never had a child reached
menopause earlier, and this difference was nearly statisti-
cally significant (p ¼ 0.06). No association was observed
with age at menarche or at first or last birth.
Table 3 reports the relations between occupational factors
and median age at menopause among women without a history of depression. Median age at menopause was higher
among women who reported repetitive work in 1990 than
among those who did not (p ¼ 0.04). Although the difference was not statistically significant, it is interesting that the
median age at menopause was lowest for women who reported high-strain jobs in 1990 and highest for those who
reported active jobs (p ¼ 0.11). We observed no association
with the other physical work stressors.
Table 4 shows the relations between occupational factors
and the hazard ratio of menopause among women with
Am J Epidemiol 2007;166:429–438
Work-related Factors and Age at Natural Menopause
433
TABLE 2. Median age (years) at menopause according to reproductive history in
a cohort of 1,594 gainfully employed French women born in 1938
Variables
No. of
women
% of total
sample
Median age
(years)*
Log-rank
p value
0.55
Age at menarche (years)
11
188
11.8
51
12
349
21.9
52
13
343
21.5
52
14
381
23.9
51
15
318
19.9
52
15
1.0
No child
209
13.1
51
22
506
31.7
52
23–25
464
29.1
52
26
402
25.3
52
13
0.8
No child
209
13.1
51
27
513
32.2
52
28–31
435
27.3
52
32
424
26.6
52
13
0.8
Missing data
Age at first birth (years)
Missing data
0.14
Age at last birth (years)
Missing data
0.22
Parity
0
209
13.1
51
1
1,376
86.3
52
9
0.6
Missing data
0.06
Oral contraceptive use
Never
1,050
65.9
52
10 years and age at first use 30 years
153
9.6
51
10 years and age at first use >30 years
247
15.5
51
>10 years and age at first use 30 years
65
4.1
52
>10 years and age at first use >30 years
61
3.8
52
Missing data
18
1.1
0.05
* Menopause-free survival curve.
a history of depression. Later menopause was associated
with job strain (p ¼ 0.03) and was highest for those with
passive jobs in 1990. Median age was higher for women
who were exposed in 1990 to shift work (p ¼ 0.08) or repetitive work (p ¼ 0.09) than for women not so exposed.
Age at menopause was lower for women who were exposed
to 48 hours/week or more of work in 1990 than for the other
women (p ¼ 0.07), but the former group was rather small.
We found no associations with physical work constraints,
standing work, noise, or difficult schedules.
Among the women with no history of depression (table 5),
the multivariate Cox final model showed that earlier menopause was significantly and independently associated with
smoking more than 10 cigarettes per day in 1990 (p <
0.001), high-strain jobs in 1990 (vs. no such job in 1990,
p ¼ 0.01), and difficult schedules before 1990 (vs. in 1990 or
Am J Epidemiol 2007;166:429–438
never, p ¼ 0.03). Later menopause was associated with high
school educational level (p ¼ 0.003) and repetitive work in
1990 (vs. before 1990 or never, p ¼ 0.005). For women with
a history of depression (table 6), later menopause was significantly and independently associated with parity (p <
0.001) and age at menarche of 13 or more years (p ¼
0.004), and earlier menopause was associated with high
school educational level (p < 0.01) and job control (p ¼
0.03). No significant interaction between variables was observed in either stratum.
DISCUSSION
This study corroborates the findings of previous studies
(11, 16, 24) that suggest that educational level, smoking, and
434 Cassou et al.
TABLE 3. Median age (years) at menopause and unadjusted relative risk according to occupational
exposures among French women without self-reported history of depression who were born in 1938
Variables
No. of women
(n ¼ 1,355)
% of total
sample
Median age
(years)*
57.9
52
Hazard ratio
(Cox model)
Likelihood ratio
test p value
Physical work constraintsy
Never exposed
785
1
0.50
Exposed before 1990
289
21.3
52
0.96
Exposed in 1990
281
20.8
52
1.08
540
39.9
52
1
Standing work
Never exposed
0.71
Exposed before 1990
209
15.4
52
1.02
Exposed in 1990
606
44.7
52
1.05
1,080
79.7
52
1
High noise
Never exposed
0.56
Exposed before 1990
135
10.0
52
1.02
Exposed in 1990
140
10.3
52
0.90
Never exposed
789
58.2
52
1
Exposed before 1990
513
37.9
52
1.04
53
3.9
52
1.20
Work 48 hours/week
Exposed in 1990
0.42
Difficult schedules
Never exposed
1,013
74.8
52
1
Exposed before 1990
210
15.5
51
1.16
0.18
Exposed in 1990
132
9.7
52
0.95
Shift work
Never exposed
1,071
79.0
52
1
Exposed before 1990
143
10.6
51
1.10
0.22
Exposed in 1990
141
10.4
52
0.87
Repetitive work
Never exposed
1,029
75.9
52
1
Exposed before 1990
194
14.3
52
1.08
0.04
Exposed in 1990
132
9.8
53
0.79
Job strain in 1990
High strain
119
8.8
51
1
Passive
265
19.6
52
0.79
0.11
Active
622
45.9
52
0.77
Low strain
349
25.7
52
0.81
* Menopause-free survival curve.
y Exposed to carrying heavy loads or vibrations or work exertion required to operate tools or machines.
self-reported history of depression are associated with age at
menopause. Among women with no history of depression,
our results suggest a relation between high job strain, difficult schedules, and repetitive work under time constraints on
the one hand and age at menopause on the other. Among
women with a history of depression, however, our results
suggest a relation with job control.
We hypothesized that adverse working conditions and
especially high-strain jobs may lead to earlier menopause.
Exposure to physical and mental job stress has several
known physiologic consequences (33–35). Fenster et al.
(33) found that the risk of short cycles (24 days) more
than doubled among women in stressful jobs compared with
women without stressful jobs. As noted above, Wheelan
et al. (13) reported that women with shorter cycles (<26
days) are likely to be younger at menopause. According to
Bromberger et al. (14), psychosocial stress is predictive of
earlier age at menopause among African-American women
and among those with irregular cycles at baseline. Highstrain jobs (combination of high work demands and low
job control) may affect the autonomic nervous system, neuroendocrine activity, and therefore reproductive function.
High job control can moderate the stress caused by the high
demands in active jobs. We hypothesized that passive and
Am J Epidemiol 2007;166:429–438
Work-related Factors and Age at Natural Menopause
435
TABLE 4. Median age (years) at menopause and unadjusted relative risk according to occupational
exposures among French women with self-reported history of depression who were born in 1938
No. of women
(n ¼ 239)
% of total
sample
Median age
(years)*
118
49.4
50
1
Exposed before 1990
55
23.0
50
0.76
Exposed in 1990
66
27.6
50
1.04
68
28.5
50
1
Variables
Hazard ratio
(Cox model)
Likelihood ratio
test p value
Physical work constraintsy
Never exposed
0.25
Standing work
Never exposed
Exposed before 1990
48
20.1
50
0.88
123
51.4
50
0.74
167
69.9
50
1
Exposed before 1990
34
14.2
50
1.06
Exposed in 1990
38
15.9
50
1.01
Never exposed
129
54.0
50
1
Exposed before 1990
101
42.3
50
0.83
9
3.7
50
2.07
Exposed in 1990
0.21
High noise
Never exposed
0.96
Work 48 hours/week
Exposed in 1990
0.07
Difficult schedules
Never exposed
172
72.0
50
1
Exposed before 1990
53
22.2
50
0.92
Exposed in 1990
14
5.8
50
0.86
0.81
Shift work
Never exposed
176
73.6
50
1
Exposed before 1990
41
17.2
50
1.11
Exposed in 1990
22
9.2
52
0.63
0.08
Repetitive work
Never exposed
152
63.6
50
1
Exposed before 1990
55
23.0
50
1.05
Exposed in 1990
32
13.4
52
0.65
High strain
36
15.1
50
1
Passive
45
18.8
51
0.85
0.09
Job strain in 1990
Active
Low strain
105
43.9
50
1.42
53
22.2
50
1.44
0.03
* Menopause-free survival curve.
y Exposed to carrying heavy loads or vibrations or work exertion required to operate tools or machines.
low-strain jobs with low demands had no effect on the reproductive function. However, among women with a history
of depression, we found that earlier menopause was associated with active and low strain, that is, high job control. This
surprising result may be explained by the fact that women
reporting a history of depression may perceive job strain,
especially high decision latitude, differently from other
women without depression. Job strain, though, was based
on the job held in 1990, which may not be representative
of a woman’s usual occupational position, and perhaps the
depression led to a job change. It is also possible that this
result could be due to some idiosyncrasy in this small sample
Am J Epidemiol 2007;166:429–438
(n ¼ 239). Indeed this factor appeared significant in only 37
percent of the bootstrap samples.
The relation between age at menopause and repetitive
work may have been biased by selection of the sample population at the first phase of the survey (e.g., the healthy
worker effect), and this may explain the inverse relation
between repetitive work under time constraints and age at
menopause. Women who reported repetitive work in 1990
may have been in better health, given that they had already
put up with this working condition for a long time (36). The
effect on age at menopause of difficult schedules could be
due to conflict between night work schedules and circadian
436 Cassou et al.
TABLE 5. Multivariate Cox model among French women without self-reported history of depression who were born in 1938 (n ¼
1,330*)
Full model
Variables
Hazard
ratio
Final model
Likelihood ratio
test p value
Hazard
ratio
95% confidence
interval
Likelihood ratio
test p value
0.8
0.7, 0.9
0.003
<0.001
Education: high school
0.8
0.004
Smoking: current and >10 cigarettes/day
1.9
<0.001
1.9
1.4, 2.5
Body mass index (kg/m2): 22.5–<25.9
0.9
0.3
—y
—
—
Body mass index (kg/m2): 25.9
0.9
0.07
—
—
—
High strain job in 1990
1.3
0.01
1.3
1.1, 1.6
0.01
Repetitive work in 1990
0.8
0.02
0.7
0.6, 0.9
0.005
Shift work before 1990
1.0
0.90
—
—
Shift work in 1990
0.9
0.35
—
—
—
Difficult schedules before 1990
1.2
0.06
1.2
1.0, 1.4
0.03
Difficult schedules in 1990
1.0
0.69
—
—
—
—
* Twenty-five participants with missing data in at least one variable were not included in the analysis.
y —, variable not included in the final model.
rhythms (37). Although multiple comparisons may account
for these results, the data show specific relations between
psychosocial stressors and age at menopause and no relations with physical constraints.
Our study has several strengths. Because the women
included in the analysis were all born in 1938, our results
are not affected by a generational bias due to the variability
of reproductive histories across different generations (38).
Our sample was relatively homogeneous in terms of the
timing of reproductive events and culturally determined variables. Data were, in part, prospectively collected, and the
follow-up rate was high. The population was relatively large
and not clinically based. A standardized protocol was used
for the interview. Many potential confounders were examined. However, various factors that may influence ovarian
senescence were not taken into account, including dietary
factors, menstrual cycle regularity, and alcohol consumption. In addition, women were not asked if they had ever
undergone induction of ovulation (ovarian hyperstimulation), although it is unlikely that many women born in
1938 did. Because of the unreliability of information about
age at menopause of the participants’ mothers, we did not
TABLE 6. Multivariate Cox model among French women with self-reported history of depression who were born in 1938 (n ¼ 226*)
Full model
Variables
Final model
Hazard
ratio
Likelihood ratio
test p value
Hazard
ratio
95% confidence
interval
Likelihood ratio
test p value
Social category: executive
1.1
0.8
—y
—
—
Education: high school
1.6
0.12
1.9
1.2, 3.1
<0.01
Former smoker and 10 cigarettes/day
0.8
0.67
—
—
—
Former smoker and >10 cigarettes/day
1.0
0.99
—
—
—
Current smoker and 10 cigarettes/day
0.4
0.02
—
—
—
Current smoker and >10 cigarettes/day
1.1
0.66
—
—
—
Smoking
Age at menarche: 13 years
0.6
0.004
0.6
0.5, 0.9
0.004
Parity: 1
0.4
<0.001
0.4
0.3, 0.6
<0.001
Oral contraceptive use 10 years and age
at first use 30 years
1.2
0.55
—
—
—
High job control in 1990
1.3
0.15
1.4
1.0, 2.0
0.03
Repetitive work in 1990
0.9
0.83
—
—
—
Work 48 hours/week in 1990
1.5
0.33
—
—
—
Shift work in 1990
0.6
0.90
—
—
—
* Thirteen participants with missing data in at least one variable were not included in the analysis.
y —, variable not included in the final model.
Am J Epidemiol 2007;166:429–438
Work-related Factors and Age at Natural Menopause
collect this information. Murabito et al. (39) estimate that
genetic factors may explain at least 50 percent of interindividual variability.
These results must be interpreted in light of the specificities of our sample. The ESTEV population included workers in the second half of their careers, followed by an
occupational physician and working mainly in private industry and trade. Although the geographic regions surveyed
cover the socioeconomic diversity of France, we cannot
guarantee that the sample population is representative of
the entire French working population. Moreover, the physician participation rate was quite low at 60 percent. The high
return rate by participants in 1995, stable in all six regions,
limits the potential bias from nonparticipation. The reasons
for nonparticipation in 1990 were as follows: 1) refusal to
participate, by 50 percent of those selected but not interviewed; 2) nonemployment (resignation, layoff) for 16.6
percent; 3) failure to come to the occupational clinical examination for 8.3 percent; 4) sick leave for 8.3 percent; and
5) unknown reasons for 16.8 percent.
Menopause was, as is standard, diagnosed retrospectively.
The section of the questionnaire concerning menopause was
completed by a physician aware that women must not have
menstruated for at least a year to be considered postmenopausal. The reliability of reported age at natural menopause
declines with time since menopause (40). In our study,
women had been naturally postmenopausal for an average
of 2 years when they reported age at natural menopause. A
total of 66.6 percent (n ¼ 807) of the women reported natural menopause before 1990, 29.4 percent (n ¼ 356) between 1990 and 1994, and 4 percent (n ¼ 48) had not yet
reached menopause in 1995. Of those who had experienced
natural menopause before 1990 and were interviewed twice,
41.4 percent reported exactly the same age at menopause
both times, and another 41.7 percent reported an age at
menopause within 1 year at the first and second interviews.
A digit preference for ages 45 and 50 was not observed in
our study.
Depression was not diagnosed by validated scales but by
women’s self-reports. To avoid making the questionnaire
too long, since depression was not our principal predictor
variable, we did not use a standardized instrument, and our
method is susceptible to underreporting. However, similar
questions have been used in numerous studies of middleaged women (41), and we did find a strong correlation between the Nottingham Health Profile score of emotional
reactions and self-reported history of depression (p <
0.0001).
The validity and reliability of exposure must be discussed. The occupational physicians monitored the validity
of the information provided on working conditions and verified its consistency with information on the job specification sheet included in the medical file. They corrected
roughly 2 percent of workers’ statements. Reliability of reported toxic exposures was not good, and we did not consider these exposures in the analysis. Use of self-reports to
assess work stress, especially job strain, may be a source of
bias. Yet, previous studies in similar conditions have produced high correlations between subjective assessments and
expert ratings of job conditions (42). Recall of past work
Am J Epidemiol 2007;166:429–438
437
load is still a serious problem in this type of study, but
differential classification bias is unlikely since all participants faced this problem to the same extent. The timing of
exposures must also be considered in designing studies of
occupational exposures that may affect ovarian senescence.
For example, women exposed for less than 10 years to physical working conditions might have experienced these conditions only after menopause had already begun. This
nonetheless seems very unlikely, since examination of occupational histories showed that these exposures were probably present at the beginning of employment for nearly all
women. Moreover, only 5 percent of naturally postmenopausal women reported that their first and last jobs were
different, and depression occurred at least 5 years before
menopause for 63 percent of the naturally postmenopausal
women who had been depressed. Although it is impossible
to know the precise temporal relations between depression,
working conditions, and especially job strain and age at
menopause, we can consider that the most likely sequence
is awkward working conditions, depression, and menopause.
The most important factor determining a woman’s age at
onset of menopause is probably the number of ovarian follicles (43). Menopause occurs when the number of primordial follicles has fallen to a critical number. Certain
environmental factors, such as smoking, may damage the
follicular pool directly or indirectly. Our results suggest that
some job stressors, including high job strain, difficult schedules, and perhaps repetitive work, could change the onset of
natural menopause. Clarifying the basis for these associations requires further study.
ACKNOWLEDGMENTS
This research was supported by grants from the Institut
National de la Santé et de la Recherche Médicale
(INSERM), the Ministry of Research, the Ministry of
Labor and Social Affairs, and the Fédération Nationale de
la Mutualité Francxaise.
The authors thank the 387 occupational physicians who
participated in the ESTEV study. They thank Louise
Lafortune and Jo Ann Cahn for helpful comments on the
manuscript.
Conflict of interest: none declared.
REFERENCES
1. Sowers MF. Menopause: its epidemiology. In: Goldman MB,
Hatch MC, eds. Women and health. San Diego, CA: Academic
Press, 2000:1155–68.
2. Snowdon DA, Kane RL, Beeson WL, et al. Is early natural
menopause a biologic marker of health and aging? Am J Public
Health 1989;79:709–14.
3. te Velde ER, Dorland M, Broekmans FJ. Age at menopause as
a marker of reproductive ageing. Maturitas 1998;30:119–25.
4. Jacobsen BK, Heuch I, Kvale G. Age at natural menopause and
all-cause mortality: a 37-year follow-up of 19,731 Norwegian
women. Am J Epidemiol 2003;157:923–9.
438 Cassou et al.
5. Harlow BL, Signorello LB. Factors associated with early
menopause. Maturitas 2002;42(suppl 1):S87–93.
6. Cramer DW, Xu H. Predicting age at menopause. Maturitas
1996;23:319–26.
7. Treloar SA, Do KA, Martin NG. Genetic influences on the
age at menopause. Lancet 1998;352:1084–5.
8. De Bruin JP, Bovenhuis H, van Noord PAH, et al. The role
of genetic factors in age at natural menopause. Hum Reprod
2001;16:2014–18.
9. Parazzini F, Negri E, La Vecchia C. Reproductive and general lifestyle determinants of age at menopause. Maturitas
1992;15:141–9.
10. Kato I, Toniolo P, Akhmedkhanov A, et al. Prospective
study of factors influencing the onset of natural menopause.
J Clin Epidemiol 1998;51:1271–6.
11. Cooper GS, Sandler DP, Bohlig M. Active and passive
smoking and the occurrence of natural menopause. Epidemiology 1999;10:771–3.
12. Reynolds RF, Obermeyer CM. Age at natural menopause in
Spain and the United States: results from the DAMES project.
Am J Hum Biol 2005;17:331–40.
13. Whelan EA, Sandler DP, McConnaughey DR, et al. Menstrual
and reproductive characteristics and age at natural menopause.
Am J Epidemiol 1990;131:625–32.
14. Bromberger JT, Matthews KA, Kuller LH, et al. Prospective
study of the determinants of age at menopause. Am J Epidemiol 1997;145:124–33.
15. Beser E, Aydemir V, Bozkaya H. Body mass index and
age at natural menopause. Gynecol Obstet Invest 1994;37:
40–2.
16. Harlow BL, Cramer DW, Annis KM. Association of medically
treated depression and age at natural menopause. Am J Epidemiol 1995;141:1170–6.
17. van Noord PAH, Boersma H, Dubas JS, et al. Age at natural
menopause in a population-based screening cohort: the role
of menarche, fecundity and lifestyle factors. Fertil Steril
1997;68:95–102.
18. Gold EB, Bromberger J, Crawford S, et al. Factors associated
with age at natural menopause in a multiethnic sample of
midlife women. Am J Epidemiol 2001;153:865–74.
19. Sievert LL, Waddle D, Canali K. Marital status and age at
natural menopause: considering pheromonal influence. Am
J Hum Biol 2001;13:479–85.
20. Cooper GS, Baird DD, Darden FR. Measures of menopausal
status in relation to demographic, reproductive, and behavioral
characteristics in a population-based study of women aged
35–49 years. Am J Epidemiol 2001;153:1159–65.
21. Hatch EE, Troisi R, Wise LA, et al. Age at natural menopause
in women exposed to diethystilbestrol in utero. Am J Epidemiol 2006;164:682–8.
22. Brambilla DJ, McKinley SM. A prospective study of factors affecting age at menopause. J Clin Epidemiol 1989;42:
1031–9.
23. Luoto R, Kaprio J, Uutela A. Age at natural menopause and
sociodemographic status in Finland. Am J Epidemiol 1994;
139:64–76.
24. Lawlor DA, Ebrahim S, Davey Smith G. The association of
socio-economic position across the life course and age at
natural menopause: the British Women’s Heart and Health
Study. BJOG 2003;110:1078–87.
25. Paul M. Occupational reproductive hazards. Lancet 1997;
349:1385–8.
26. Florack EM, Zielhuis GA, Rolland R. The influence of
occupational physical activity on the menstrual cycle and
fecundability. Epidemiology 1994;5:14–18.
27. Cooper GS, Savitz DA, Millikan R, et al. Organochlorine
exposure and age at natural menopause. Epidemiology
2002;13:729–33.
28. Akkina J, Reif J, Keefe T, et al. Age at natural menopause
and exposure to organochloride pesticides in Hispanic
women. J Toxicol Environ Health 2004;67:1407–22.
29. Derriennic F, Touranchet A, Volkoff S. Age, travail, santé:
études sur les salariés âgés de 37 à 52 ans, enquête ESTEV
1990. Paris, France: Les éditions INSERM, 1996.
30. Marquié JC, Paumès D, Volkoff S. Le travail au fil du
temps. Toulouse, France: Octarès, 1995.
31. Karasek RA. Job demands, job decision latitude and mental
strain: implications for job redesign. Adm Sci Q 1979;24:
285–8.
32. Hunt SM, McEwan T. The development of a subjective
health indicator. Sociol Health Illn 1980;2:231–46.
33. Fenster L, Waller K, Chen J, et al. Psychological stress in
the workplace and menstrual function. Am J Epidemiol
1999;149:127–34.
34. Garde AH, Hansen AM, Jensen BR. Physiological responses
to four hours of low-level repetitive work. Scand J Work
Environ Health 2003;29:452–60.
35. Berga SL. How stress can affect ovarian function. Contemp
Obstet Gynecol 1993;38:87–94.
36. Choi BC. Definition, sources, magnitude, effect modifiers
and strategies of reduction of the healthy worker effect.
J Occup Med 1992;34:979–88.
37. Costa G. The impact of shift and night work on health.
Appl Ergon 1996;27:9–16.
38. Morabia A, Costanza MC. International variability in ages
at menarche, first livebirth, and menopause. World Health
Organization Collaborative Study of Neoplasia and Steroid
Contraceptives. Am J Epidemiol 1998;148:1195–205.
39. Murabito JM, Yang Q, Fox C, et al. Heritability of age at
natural menopause in the Framingham Heart Study. J Clin
Endocrinol Metab 2005;90:3427–30.
40. den Tonkelaar I. Validity and reproducibility of self-reported
age at menopause in women participating in the DOM-project.
Maturitas 1997;27:117–23.
41. Bromberger JT, Meyer PM, Kravitz HM, et al. Psychologic
distress and natural menopause: a multiethnic community
study. Am J Public Health 2001;91:1435–42.
42. Theorell T, Karasek RA. Current issues relating to
psychosocial job strain and cardiovascular disease research.
J Occup Health Psychol 1996;1:9–26.
43. Ginsburg J. What determines the age at the menopause?
BMJ 1991;302:1288–9.
Am J Epidemiol 2007;166:429–438