Does menopausal transition really influence mental health? Findings from the prospective long‐term Zurich study

Abstract

In the prospective long‐term Zurich study, we re‐examined the hypothesized association between mental health problems in women and the transition through menopausal stages. One hundred sixty‐eight women from a population‐based Swiss community cohort were prospectively followed up from age 21 to 50. At age 50, the occurrence of hot flushes/night sweats and sleep disturbances was significantly more frequent in peri‐ and post‐menopausal women. Irritability/nervousness was increased only in peri‐menopausal women, but that association was accounted for by neuroticism trait scores at age 30. Transitions to peri‐ or post‐menopause were not related to changes in either the prevalence rates of DSM major depressive episode or anxiety disorders, or the course of psychopathological syndromes as assessed by the Symptom Checklist 90 ‐ Revised. The null associations held when adjusting for duration of reproductive period or age at menopause. Preceding mental health problems between ages 21 and 41, increased neuroticism trait scores at age 30, and concurrent psychosocial distress were significantly related to mental health problems occurring between ages 41 and 50. Depending upon the cut‐off point that was chosen, the arbitrary dichotomization of a continuous depression outcome produced spurious associations with the menopausal transition. We conclude that mental health problems between ages 41 and 50 are probably not directly related to the menopausal transition, and that previously reported associations could be false positives due to inadequate dichotomizations, reporting bias, undisclosed multiple adjustments or overfitting.

The menopausal transition typically begins in the mid‐forties and lasts on average up to five years. The mean age at final menstrual period (menopause) in Western countries is approximately 50 years1.

The menopausal transition is a time in a woman's life that can be marked by various physical and psychological changes2. The fluctuations and decline in levels of ovarian hormones can cause physical symptoms such as hot flushes, night sweats, urogenital atrophy with incontinence, vaginal dryness, sexual dysfunction, osteoporosis, and metabolic changes.

It has been thought that the menopausal transition is also a time of increased risk for mental health problems, especially depression3, 4. Although most prospective cohort studies failed to find statistically and practically significant associations between menopausal transitions and mental health problems5, 6, 7, 8, 9 (see also Vesco et al10), some recent investigations have suggested that a relationship does exist. These longitudinal studies, all performed in the U.S., have reported that the menopausal transition entails an increased risk not only for depressive symptoms11, 12, 13, but also for major depressive disorder14, 15.

Although the rise in depressive symptoms might seem to be associated with hormonal changes during the menopausal transition, especially the falling and fluctuating levels of estradiol and corresponding increases in levels of follicle‐stimulating hormone (FSH) and luteinizing hormone15, some well‐conducted longitudinal studies have found no correlation with female sex hormones14, 16. Thus, at present, the literature indicates no consistent relationship between circulating estradiol/FSH levels and depression3, 17.

An increase in mood symptoms has been partly attributed to the manifold psychosocial changes that often occur in women's lives during that time span, such as alterations in family structure, losses and role transitions, stressful life events, and a lack of social support6, 10, 18, 19. Furthermore, the relationship between the menopausal transition and depression seems to be strongly influenced by preceding mental disorders5, 12, 14. This implies that depressive symptoms during the menopausal transition could represent the re‐occurrence of pre‐existing disorders or reflect a general vulnerability to develop mental health problems during stressful life events1. In this respect, although the personality trait of neuroticism has been repeatedly reported to be a crucial risk factor for mental health problems and psychosocial dysfunction20, 21, 22, no study on the psychopathology‐menopause association has included that trait thus far.

Most studies conducted to date on this issue have suffered from major limitations. In particular, all the longitudinal studies which reported a significant impact of menopausal transition on depressive symptoms treated depression as a dichotomous outcome (i.e., absent versus present), even though the evidence is compelling that depression23, 24, and psychopathology in general25, are continuously distributed in nature. Moreover, many studies focused exclusively on depression, and covered only short intervals (i.e., 5‐10 years) that did not allow one to address the question of preceding mental health problems. Furthermore, to the best of our knowledge, no investigation has tested the prospective effect of neuroticism on mental health problems ascribed to the menopausal transition.

We used data from the Zurich study, a longitudinal community study that spanned a 30‐year period and included participants beginning at age 20. The following questions were addressed: a) Does the severity of psychopathological syndromes increase during the menopausal transition? b) Is there an increase in prevalence rates of DSM‐diagnosed major depressive episode or anxiety disorders? c) If a change in mental health does occur during the menopausal transition, is it related to a pre‐existing vulnerability to psychopathology, a personality trait of neuroticism, concurrent psychosocial distress, or advancing age? d) Can treating depression as a dichotomous outcome produce false positive results?

METHODS

Sampling and procedure

The Zurich Study comprised a cohort of 4,547 subjects (2,201 males and 2,346 females) representative of the canton of Zurich in Switzerland, who were screened in 1978 by the Symptom Checklist 90 ‐ Revised (SCL‐90‐R)26 when they were 19 years old (if males) or 20 years old (if females).

Male and female participants were sampled with different approaches. In Switzerland, every male person must undertake a military screening test at the age of 19. With the consent of military authorities, but independent of their screening procedure, we randomly screened 50% of all male conscripts of the canton of Zurich in this age group. The refusal rate was 0.3%. Female participants were identified from the complete electoral register of the canton of Zurich. Again, 50% of them were randomly selected and received questionnaires by mail; 75% responded.

We selected a sub‐sample of 591 subjects for interview, with two‐thirds consisting of high scorers (defined by the 85th percentile or more of the global severity index of the SCL‐90‐R) and one‐third consisting of a random sample of those with scores below the 85th percentile. Altogether, seven interview waves were conducted: in 1979 (292 males and 299 females), 1981 (220 males and 236 females), 1986 (225 males and 232 females), 1988 (200 males and 224 females), 1993 (192 males and 215 females), 1999 (162 males and 205 females), and 2008 (144 males and 191 females).

For the present study, we included only women who had participated consistently through 2008 (at age 50) and who were still menstruating in 1999 (at age 41), to exclude cases of premature menopause. This produced a final sample size of 168 subjects.

Instruments and measures

All information about menstruation and menopause was obtained during the interviews in 1999 and 2008, when participants were 41 and 50 years old, respectively. Interviews were conducted according to the Structured Psychopathological Interview and Rating of the Social Consequences of Psychological Disturbances for Epidemiology (SPIKE)27. This semi‐structured interview, developed for epidemiological surveys in psychiatric research, assesses data about socio‐demography, psychopathology, substance use, medication, health services use, impairment, and social activity. Its reliability and validity have been reported elsewhere28.

Menopause status was comprehensively assessed in the last interview in 2008. We defined as post‐menopausal those women whose last menstruation had occurred more than 12 months before. We defined as peri‐menopausal those women whose last menstruation had occurred within the past 12 months and who had not menstruated in the preceding two months, and those whose last menstrual cycle had occurred less than one month before and who reported menstrual irregularities. We defined as pre‐menopausal those women whose most recent menstruation had occurred within the past month and who had experienced no menstrual irregularities during the past 12 months.

The following symptoms were assessed by the SPIKE during the 2008 interviews, and were included in our analysis: hot flushes and/or night sweats, sleep disturbances, depressed mood, irritability and/or nervousness, anxiety and/or panic, physical and mental tiredness, sexual problems, urinary incontinence, vaginal dryness, and joint pain. Participants rated distress related to each of these symptoms using a five‐point Likert scale ranging from 1 (“not at all”) to 5 (“extremely”).

We also assessed psychopathology through the SCL‐90‐R, in which distress from each symptom is rated according to a five‐point Likert scale ranging from 1 (“not at all”) to 5 (“extremely”). We covered the most recent four‐week period at each interview. The 90 items of the checklist were grouped into nine subscales (anxiety, depression, hostility, interpersonal sensitivity, obsessive‐compulsivity, paranoid ideation, phobic anxiety, psychoticism, and somatization), with the score on each subscale calculated as the average of the scores on the corresponding items (thus ranging from 1 to 5). Psychopathological vulnerability was evaluated using the mean SCL‐90‐R global severity index of each individual between 1979 and 1999. The SCL‐90‐R has shown good internal consistency and test‐retest reliability29, 30.

Personality was assessed by the Freiburg Personality Inventory31 in 1988, when the women were 30 years old. We utilized an empirically derived subscale of neuroticism consisting of 16 items, which has been found to have good validity and reliability32, 33.

At each interview, the 12‐month prevalence of major depression episode and anxiety disorders was assessed on the basis of the information provided by the SPIKE. The criteria for major depressive episode, agoraphobia, social phobia, specific phobia and obsessive‐compulsive disorder were those of the DSM‐III‐R, whereas the criteria for generalized anxiety disorder and panic disorder were those of the DSM‐III (see Angst et al34 for further information).

The assessment of psychosocial distress was based on participants’ perceived discontent, expressed using a five‐point Likert scale ranging from 1 (“not at all”) to 5 (“extremely”), with six psychosocial domains: employment, financial situation, friendships, health, partnership, and family. Because the intercorrelation of these six variables was high, we used a single variable obtained by computing the mean score across the six domains.

Statistical analysis

We conducted a series of ordinal logistic regression analyses with the various five point‐scaled menopause symptoms entered separately as the dependent variable and menopause status considered the predictor variable. Estimates of variance explained were reported according to Nagelkerke's pseudo R².

All other associations were analyzed longitudinally, using variables measured in 1999 (age 41) and 2008 (age 50). For this purpose we used a series of generalized estimating equations. These analyses were introduced to fit regression models that account for within‐subject correlations, which is an inherent part of longitudinal studies that rely on repeated measures35. Psychopathology and mental disorders were entered as the dependent variable. Owing to the right‐skewed distribution of the continuous SCL‐90‐R psychopathology syndromes, a gamma distribution with a log link‐function best fitted our data. Models with dichotomous dependent variables (i.e., diagnoses of major depressive episode and anxiety disorders) were fitted with a binomial distribution and logit link‐function. To reduce the effects of influential observations, we used a robust estimator of the parameter estimate covariance matrix. Menopause status was entered as the predictor variable.

For all generalized estimating equations, in addition to adjusting for within‐subject correlations, time was included as a between‐subject effect to account for the influence of aging as participants progressed from 41 to 50 years old. This is a common procedure in longitudinal data analysis when outcomes are assumed to increase or decrease over time36. Finally, to test for the effects of intervening variables such as psychosocial distress and psychopathological vulnerability, we fitted a series of multiple predictor models in which all interesting predictors were entered simultaneously as main effects. The analyses examining the effect of dichotomization were weighted to compensate for power loss.

All statistical analyses were performed with SPSS version 20 for Macintosh.

RESULTS

Of the 168 women who were still regularly menstruating at age 41 in 1999, 54 (32.1%) were considered in 2008 (at age 50) to be pre‐menopausal, 65 (38.7%) peri‐menopausal, and 45 (26.8%) post‐menopausal. The remaining four women (2.4%) could not be assigned to any of these groups because of missing data.

At age 50, when menopause status was assessed, the 12‐month prevalence rates of major depressive episode in pre‐, peri‐, and post‐menopausal women were 18.5%, 13.8%, and 11.1%, respectively, while the 12‐month prevalence rates of anxiety disorders were 24.1%, 23.1%, and 22.2%, respectively. The lifetime prevalence rates in pre‐, peri‐, and post‐menopausal women were 53.7%, 55.4%, and 42.2%, respectively, for major depressive episode, and 63.0%, 63.1%, and 60.0%, respectively, for anxiety disorders. None of the differences in prevalence rates across the three groups reached statistical significance according to Pearson χ² tests (all p>0.10).

Cross‐sectional associations between menopause symptoms and menopause status at age 50 are presented in Table 1. The occurrence of hot flushes and/or night sweats was significantly increased in women classified as peri‐ and post‐menopausal (OR=2.79 and OR=3.04, respectively) when compared with pre‐menopausal women. Sleep disturbances were also more common in peri‐ and post‐menopausal women (OR=2.26 and OR=2.58, respectively), while irritability and/or nervousness were significantly increased only in peri‐menopausal women (OR=2.34). Vaginal dryness was significantly increased only in post‐menopausal women (OR=2.86). Estimates of explained variance in these outcomes ranged from 3.9% for vaginal dryness to 6.7% for hot flushes/night sweats (corresponding to small‐to‐medium effect sizes). Neither depressed mood nor anxiety/panic (both R²=0.005) was related to menopause status.

Table 1.

Cross‐sectional associations between menopause status and symptoms assessed by SPIKE at age 50

	Menopause status	OR (95% CI)	p	R2
Hot flushes, night sweats	Peri‐menopausal Post‐menopausal	2.79 (1.33, 5.84) 3.04 (1.35, 8.86)	0.007 0.007	0.067
Sleep disturbances	Peri‐menopausal Post‐menopausal	2.26 (1.00, 5.10) 2.58 (1.07, 6.25)	0.049 0.035	0.040
Depressed mood	Peri‐menopausal Post‐menopausal	1.40 (0.60, 3.25) 1.29 (0.50, 3.28)	0.433 0.599	0.005
Irritability, nervousness	Peri‐menopausal Post‐menopausal	2.34 (1.02, 5.35) 1.08 (0.41, 2.84)	0.044 0.871	0.044
Anxiety, panic	Peri‐menopausal Post‐menopausal	1.45 (0.47, 4.46) 1.49 (0.44, 5.09	0.518 0.523	0.005
Physical and mental tiredness	Peri‐menopausal Post‐menopausal	2.00 (0.84, 4.74) 1.61 (0.61, 4.20)	0.117 0.333	0.021
Sexual problems	Peri‐menopausal Post‐menopausal	1.25 (0.52, 3.04) 1.59 (0.61, 4.13	0.617 0.340	0.007
Urinary incontinence	Peri‐menopausal Post‐menopausal	3.14 (0.65, 15.24) 4.32 (0.85, 21.91)	0.041	0.156 0.077
Vaginal dryness	Peri‐menopausal Post‐menopausal	1.56 (0.65, 3.72) 2.86 (1.14, 7.20)	0.316 0.025	0.039
Joint pain	Peri‐menopausal Post‐menopausal	1.51 (0.57, 3.98) 0.69 (0.21, 2.25)	0.409 0.544	0.041

SPIKE ‐ Structured Psychopathological Interview and Rating of the Social Consequences of Psychological Disturbances for Epidemiology

Pre‐menopausal status served as reference category

When neuroticism was added to the analysis, the transition to peri‐menopause was no longer related to irritability and/or nervousness, while neuroticism was (for one standard deviation increase in neuroticism: OR=1.67, 95% CI: 1.16‐2.39, p=0.006). Thus, neuroticism as assessed at age 30 fully explained the association between the transition to peri‐menopause and irritability and/or nervousness at age 50. In contrast to menopause status, neuroticism at age 30 also strongly predicted the occurrence of depressed mood (OR=1.94, 95% CI: 1.33‐2.82, p=0.001) and anxiety/panic (OR=1.81, 95% CI: 1.15‐2.83, p=0.010) at age 50.

Longitudinal associations between menopause status and the prevalence of mental disorders between ages 41 and 50 are shown in Table 2. No association reached statistical significance. In other terms, increases or decreases in prevalence rates of major depressive episode or anxiety disorders between 41 and 50 did not differ between women who remained pre‐menopausal over time and those who transitioned from pre‐ to peri‐ or from pre‐ to post‐menopause. That null association held when adjusting for preceding psychopathological vulnerability, duration of reproductive period, or age at menopause. In contrast, sensitivity analyses revealed that the personality trait of neuroticism as assessed at age 30 significantly predicted increases in the prevalence of major depressive episode (OR=1.37, p=0.045) and anxiety disorders (OR=1.93, p<0.001).

Table 2.

Longitudinal associations between menopause status and change in prevalence rates of DSM‐III‐R/DSM‐III mental disorders over time as participants aged from 41 to 50

	Menopause status
	1999	2008	OR (95% CI)	p
Major depression episode	Pre	Pre	Reference
	Pre	Peri	0.71 (0.34, 1.51)	0.375
	Pre	Post	0.57 (0.24, 1.37)	0.209
Anxiety disorders	Pre	Pre	Reference
	Pre	Peri	1.10 (0.55, 2.20)	0.788
	Pre	Post	0.92 (0.43, 1.97)	0.824

Major depressive episode, agoraphobia, social phobia, specific phobia and obsessive‐compulsive disorder were diagnosed according to DSM‐III‐R criteria, and generalized anxiety disorder and panic disorder according to DSM‐III criteria

Longitudinal associations between menopause status and psychopathological syndromes are outlined in Table 3. No syndrome score increased or decreased significantly over time in relation to menopause status. These null findings were not merely due to a lack of statistical power, because the effect sizes were also practically insignificant (all d<0.2). The null associations held after adjusting for duration of reproductive period or age at menopause.

Table 3.

Longitudinal associations between menopause status and course of SCL‐90‐R psychopathological syndromes over time as participants aged from 41 to 50

	Menopause status
	1999	2008	b (95% CI)	p
Anxiety	Pre	Pre	Reference
	Pre	Peri	0.027 (−0.157, 0.210)	0.774
	Pre	Post	−0.091 (−0.283, 0.101)	0.352
Depression	Pre	Pre	Reference
	Pre	Peri	0.090 (−0.131, 0.311)	0.426
	Pre	Post	0.000 (−0.220, 0.220)	0.998
Hostility	Pre	Pre	Reference
	Pre	Peri	0.052 (−0.086, 0.191)	0.460
	Pre	Post	−0.071 (−0.187, 0.045)	0.228
Interpersonal sensitivity	Pre	Pre	Reference
	Pre	Peri	0.051 (−0.146, 0.248)	0.612
	Pre	Post	−0.003 (−0.226, 0.221)	0.982
Obsessive‐compulsivity	Pre	Pre	Reference
	Pre	Peri	0.029 (−0.182, 0.240)	0.789
	Pre	Post	−0.045 (−0.271, 0.182)	0.700
Paranoid ideation	Pre	Pre	Reference
	Pre	Peri	−0.005 (−0.199, 0.189)	0.958
	Pre	Post	−0.063 (−0.268, 0.141)	0.543
Phobic anxiety	Pre	Pre	Reference
	Pre	Peri	0.026 (−0.085, 0.137)	0.644
	Pre	Post	−0.005 (−0.139, 0.128)	0.937
Psychoticism	Pre	Pre	Reference
	Pre	Peri	−0.009 (−0.140, 0.121)	0.887
	Pre	Post	−0.031 (−0.163, 0.102)	0.651
Somatization	Pre	Pre	Reference
	Pre	Peri	0.066 (−0.084, 0.217)	0.388
	Pre	Post	−0.066 (−0.253, 0.120)	0.486

SCL‐90‐R – Symptom Checklist 90 ‐ Revised

Longitudinal associations among menopause status, important covariates and course of psychopathology over time as participants aged from 41 to 50 years are shown in Table 4. Again, changes in menopause status were not related to the course of psychopathology. In contrast, preceding psychopathological vulnerability, i.e. the severity of preceding psychopathological syndromes between ages 21 and 41, was significantly associated with all syndromes. When neuroticism as assessed at age 30 was included in the analysis, it similarly accounted for significant increases in all psychopathological syndromes between 41 and 50 years (all p<0.001). Concurrent psychosocial distress was also related to the course of all syndromes over time, except for phobic anxiety. The progression in age was significantly associated with the course of hostility, interpersonal sensitivity, and paranoid ideation.

Table 4.

Longitudinal associations among menopause status, important covariates, and course of SCL‐90‐R psychopathological syndromes over time as participants aged from 41 to 50

	Predictors	Wald χ2 (df)	p
Anxiety	Menopause status (41‐50)	0.185 (2)	0.912
	Psychopathological vulnerability (21‐41)	71.854 (1)	<0.001
	Psychosocial distress (41‐50)	8.936 (1)	0.003
	Effects of aging (41‐50)	0.023 (1)	0.880
Depression	Menopause status (41‐50)	1.648 (2)	0.439
	Psychopathological vulnerability (21‐41)	52.741 (1)	<0.001
	Psychosocial distress (41‐50)	46.843 (1)	<0.001
	Effects of aging (41‐50)	1.784 (1)	0.182
Hostility	Menopause status (41‐50)	1.137 (2)	0.566
	Psychopathological vulnerability (21‐41)	30.656 (1)	<0.001
	Psychosocial distress (41‐50)	12.277 (1)	<0.001
	Effects of aging (41‐50)	6.120 (1)	0.013
Interpersonal sensitivity	Menopause status (41‐50)	0.351 (2)	0.839
	Psychopathological vulnerability (21‐41)	72.042 (1)	<0.001
	Psychosocial distress (41‐50)	24.688 (1)	<0.001
	Effects of aging (41‐50)	13.010 (1)	<0.001
Obsessive‐compulsivity	Menopause status (41‐50)	0.064 (2)	0.968
	Psychopathological vulnerability (21‐41)	90.216 (1)	<0.001
	Psychosocial distress (41‐50)	12.735 (1)	<0.001
	Effects of aging (41‐50)	0.055 (1)	0.814
Paranoid ideation	Menopause status (41‐50)	1.072 (2)	0.585
	Psychopathological vulnerability (21‐41)	61.703 (1)	<0.001
	Psychosocial distress (41‐50)	10.777 (1)	0.001
	Effects of aging (41‐50)	4.300 (1)	0.038
Phobic anxiety	Menopause status (41‐50)	0.017 (2)	0.992
	Psychopathological vulnerability (21‐41)	77.636 (1)	<0.001
	Psychosocial distress (41‐50)	3.205 (1)	0.073
	Effects of aging (41‐50)	0.027 (1)	0.869
Psychoticism	Menopause status (41‐50)	1.299 (2)	0.522
	Psychopathological vulnerability (21‐41)	40.104 (1)	<0.001
	Psychosocial distress (41‐50)	14.828 (1)	<0.001
	Effects of aging (41‐50)	1.113 (1)	0.292
Somatization	Menopause status (41‐50)	2.044 (2)	0.360
	Psychopathological vulnerability (21‐41)	22.973 (1)	<0.001
	Psychosocial distress (41‐50)	14.773 (1)	<0.001
	Effects of aging (41‐50)	2.704 (1)	0.100

SCL‐90‐R – Symptom Checklist 90 ‐ Revised

Table 5 shows how selection of particular cut‐off scores for dichotomous depression could influence the odds ratios. Relative to pre‐menopause, we found a significant effect of transition to peri‐menopause (OR=1.57, p=0.033) only when the SCL‐90‐R depression cut‐off was set at ≥2.0. For all other cut‐off points (i.e., ≥1.5, ≥2.5, and ≥3.0), no statistically significant positive association with transition to peri‐menopause was detected. Actually, when the cut‐off was set at ≥3.0, the odds ratio was negative for the transition to post‐menopause when compared with the pre‐menopause phase (OR=0.29, p= 0.032), indicating that those who became post‐menopausal had a lower risk for depression. As indicated above, when depression was modelled as a continuous variable, no differences were found (p=0.255). Moreover, all mean differences among menopause phases were practically insignificant (all d<0.1), indicating that a lack of statistical significance was not merely a result of insufficient power, but rather a clear null result of no practical significance37.

Table 5.

Effect of dichotomization and choice of arbitrary cut‐offs for SCL‐90‐R depression, adjusted for preceding psychopathological vulnerability

Cut‐off on a scale from 1.0 to 5.0	Menopausal status	OR	95% CI	p
Depression ≥ 1.5	Post‐menopausal Peri‐menopausal Pre‐menopausal	0.85 1.05 Reference	0.60, 1.22 0.74, 1.50	0.382 0.770
Depression ≥ 2.0	Post‐menopausal Peri‐menopausal Pre‐menopausal	1.54 1.57 Reference	0.94, 2.54 1.04, 2.38	0.089 0.033
Depression ≥ 2.5	Post‐menopausal Peri‐menopausal Pre‐menopausal	1.11 1.29 Reference	0.58, 2.10 0.78, 2.12	0.754 0.319
Depression ≥ 3.0	Post‐menopausal Peri‐menopausal Pre‐menopausal	0.29 0.79 Reference	0.10, 0.90 0.43, 1.46	0.032 0.452
Continuously	Menopausal status	Mean score	95% CI	p
Depression scale from 1.0 to 5.0	Post‐menopausal Peri‐menopausal Pre‐menopausal	1.78 1.81 1.74	1.67, 1.89 1.70, 1.91 1.64, 1.86	0.255

DISCUSSION

This is the first prospective community study spanning 30 years and focusing on a broad range of psychopathological, psychosocial and physical problems putatively associated with the menopausal transition.

At age 50, hot flushes/night sweats and sleep disturbances were more common in peri‐ and post‐menopausal women than in pre‐menopausal ones, whereas irritability and/or nervousness were heightened only in peri‐menopausal women. However, the association between irritability/nervousness and peri‐menopause was fully explained by neuroticism at age 30, suggesting that the symptom was triggered by that personality vulnerability.

Our longitudinal analyses further revealed that changes in menopause status were not related to either the course of psychopathological syndromes or the prevalence of major depressive episode or anxiety disorders between ages 41 and 50. However, the course of psychopathological syndromes was related to psychopathological vulnerability prior to age 41 as well as to concurrent psychosocial problems. This finding emphasizes the importance of adjusting for these covariates when studying the effect of menopause on mental health. Notably, when neuroticism as assessed at age 30 was included in the analysis, it also significantly predicted increased psychopathology between ages 41 and 50. This result suggests that neuroticism is a reliable marker of persistent vulnerability to psychopathology22, 38 and is in line with emergent evidence that neuroticism has a substantial genetic overlap with depression, internalizing disorders, and even general psychopathology39, 40, 41.

Some recent longitudinal studies have reported that the occurrence of depression is associated with the transition from pre‐ to peri‐menopause11, 13, 15, 42. Two studies have also identified more symptoms in the post‐ versus pre‐menopausal phase11, 14. On the other hand, several other longitudinal studies have found no statistically and practically significant increase in mental health symptoms in relation to the menopausal transition5, 6, 7, 8, 9, 43, 44. Methodological factors probably best explain these discrepancies across studies.

First, and most importantly, all recent longitudinal studies that have described positive associations between depression and menopause status used dichotomous or dichotomized outcomes, that is, depressed versus not depressed12, 13, 14, 15, 42. When psychopathology is reduced to present versus absent, one cannot account for increases or decreases in symptom severity over time. Furthermore, because psychopathology is dimensional by nature25, treating it categorically produces a severe bias both on methodological16 and conceptual45 grounds. As demonstrated by our data, dichotomizing continuous variables can produce severe flaws and should therefore be avoided46, 47, 48. Most researchers and clinicians are probably not aware that, when dichotomization occurs at the low end of an underlying continuous construct (which is typically the case when continuous screening instruments are dichotomized), the resulting odds ratios can be severely inflated49.

Another important limitation of some previous studies is that they inferred the first onset of binary depression from a single retrospective assessment of history of lifetime depression13, 15. Since it is now well established that retrospective lifetime assessments grossly underestimate the true lifetime prevalence of mental disorders34, 50, 51, we suggest that this estimation of “first onset” may be biased, and that many women classified as having no history of depression at the outset of the above studies had already in fact experienced unrecalled or denied depressive episodes (see also Andrews et al52).

Further biases were apparently involved in previous studies. For example, Freeman et al12 reported that the transition to peri‐menopause was positively related to dichotomized scores on the Center for Epidemiologic Studies Depression Scale (CES‐D) (OR=2.89, p=0.01), but by tendency negatively related to the diagnosis of major depression (OR=0.24, p=0.21). This is surprising, because high CES‐D scores should indicate probable major depression diagnoses53. Moreover, in their bivariate analysis, early (OR=1.33, p=0.10) and late (OR=1.79, p=0.10) peri‐menopausal transitions were not significantly related to depression. Instead, after multiple adjustments and probable overfitting of their regression analyses, they achieved the significance level of p<0.05 for both early (OR=1.55, p=0.03) and late transitions (OR=2.89, p=0.01). On the other hand, in their re‐analysis of data restricted to women with no history of depression15, these authors referred to the unadjusted bivariate analysis (OR= 2.50, p=0.01), omitting to discuss that the adjusted multivariate analysis would have produced no significant result for a diagnosis of major depression (OR=1.60, p=0.34). Similarly, Cohen et al13 stated that the menopausal transition significantly increased the risk for dichotomized depression only after they adjusted for age at study entry and adverse life events, but did not show unadjusted associations in their report.

Moreover, studies reporting positive findings should also provide compelling evidence for their practical significance37. With large samples (e.g., N>400), statistical significance may easily be achieved even for trivial effects. One example of a small effect size that yielded statistical significance in a very large sample (N=67,434) can be found in a recent meta‐analysis54 which reported a protective effect of later age at menopause with regard to depression, attributed to longer exposure to endogenous estrogens. In this meta‐analysis, the odds ratio for a two‐year increment was only 0.98 (95% CI: 0.96‐0.99), which is a difference in the odds of only 2%. This effect size is practically irrelevant55, and its statistical significance is likely to be merely a result of the huge sample size56.

Several authors have postulated that a lifetime history of depression is the strongest predictor of depression during the menopausal transition5, 12, 14. Our results suggest that, while a preceding psychopathological vulnerability as expressed through high scores on neuroticism is an important predictor of psychopathology between ages 41 and 50, menopausal stages per se are not. Persons who score high on neuroticism are more susceptible to the negative effects of daily stress and critical life events57, 58, 59. Consequently, if negative affective symptoms occur during the menopausal transition, they might reflect the difficulties that vulnerable women have in coping with and adapting to the developmental changes that accompany this transition. Accordingly, it has been argued that psychopathological symptoms during menopause might be triggered by psychosocial strains such as stressful life events rather than by hormonal changes1, 43. Our data support this notion, revealing that concurrent psychosocial distress from ages 41 to 50 is strongly related to higher psychopathology during this time period, regardless of menopausal stage or preceding psychopathological vulnerability.

We recognize some limitations of our study. First, our sample size was initially moderate and was then further reduced through attrition. Second, the Zurich study was not designed specifically to examine the menopausal transition in women. Thus, we could not provide a fine‐grained assessment of that transition. Furthermore, the information on psychopathological or psychosocial outcomes that was applied in the present study relied on measurements made in 1999 and 2008. The time gap between those interviews was wide and menopause status was evaluated only twice. Third, no hormonal assessments were made.

Nevertheless, in line with three comprehensive reviews1, 2, 60, our data indicate that mental health problems between ages 41 and 50 are not directly related to the menopausal transition. All the longitudinal studies that did find associations between psychopathology and the menopausal transition used binary or dichotomized outcomes, which is problematic from a methodological46 and conceptual45 point of view. As demonstrated by our data, a dichotomization of continuous variables may produce spurious positive or negative associations.

We suggest that the relationship between psychopathology and menopause should be reconsidered carefully. Future research should incorporate not only previous history of mental disorders and current stressors, but also the personality trait of neuroticism, because this trait is likely to influence the risk of psychopathological symptoms (re‐)occurring during the age of the menopausal transition.

ACKNOWLEDGEMENT

This work was supported by the Swiss National Science Foundation (grant no. 32‐50881.97).