The menopause and its consequences have begun to receive much-needed attention over the recent years. However, uncertainty abounds. Surprisingly little is known about this natural change that affects around half of the world’s population. In the absence of facts, fear and speculation can mushroom. High-quality research can help narrow the problem space, providing information to individuals and clinicians.
Around 20% of women experience the menopausal transition simply as a cessation of periods (Brinton et al., 2015). The remainder are symptomatic in some way, with around 50% experiencing changes to mood, sleep, or cognition (Brinton et al., 2015). These symptoms are often problematic, inconvenient, or distressing, but only a subset will meet the threshold of a mental disorder such as a depressive episode. What proportion is unclear, with a recent review (Brown et al., 2024) identifying only two studies and 600 perimenopausal women in which this was assessed by clinicians as opposed to using symptom scales (one study found an increase in depressive episodes perimenopausally, one did not).
The impact of the menopause on severe mental illnesses such as schizophrenia and bipolar has received even less attention. In women with established bipolar disorder, prospective studies of around 100 women found that the perimenopause is a time of symptom worsening for the majority, with an increase in both depressive and manic symptoms (Marsh et al., 2015; Marsh et al., 2008). In schizophrenia, the evidence is more circumstantial. A large meta-analysis using data from England found that after the age of 40 (when the perimenopause is assumed to begin) approximately a quarter more women than men are newly diagnosed with schizophrenia, a reversal of the trend before this age (Kirkbride et al., 2012). In Finland, women have more hospital admissions and a reduction in antipsychotic effectiveness compared to men after age 45 (Sommer et al., 2023).
The study by Shitomi-Jones and colleagues at Cardiff University published recently in Nature Mental Health is a welcome addition to the existing scarce literature on the association of menopause with mental disorder. It utilized the large, prospective, well-characterized dataset in the UK Biobank to test the hypothesis that the perimenopause is a time of increased risk of new onset psychiatric disorders compared to the late premenopausal stage.
This study is a welcome addition to the existing scarce literature on the association of the menopausal transition with mental disorder.
Methods
Sample
Postmenopausal female participants (sex but not gender was assessed) in the UK Biobank were included in the primary analyses. Females were excluded if they had experienced early menopause (< 40 years) or if their age of menopause could not be determined due to surgery (post hysterectomy, oophorectomy, or uterine ablation), hormonal medication (intrauterine system or oral contraception) or inconsistent answers. People using hormone replacement therapy were not excluded. The final sample size was 128,294 participants.
To assess for sex-specific effects, age-matched males were also assessed.
Reproductive stage definitions
Perimenopause was defined as two years before or after the final menstrual period. Note that “perimenopause” is variably defined – the World Health Organisation uses it to mean the length of time from periods becoming irregular until 12 months after they stop (often in the region of 5-10 years).
The premenopause was defined as the 6-10 years prior to the final menstrual period. The postmenopause was defined as the 6-10 years after the final menstrual period. This means that data from 2-6 years pre or post the final menstrual period was not included in the analyses. The authors state this was to “increase distinction between the time periods and to minimize the likelihood of misclassification due to inaccuracies in menopausal timing”.
Mental disorder ascertainment
Psychiatric diagnoses and age at onset were obtained using a combination of interviews with research nurses at baseline and a self-report questionnaire completed by a subset (about 30%) of participants 5-10 years after recruitment. ‘Major depressive disorder’ required participants to have at least 2 cardinal symptoms of depression (as defined by DSM-5) and at least 5 in total. ‘Mania’ referred to diagnoses of mania, bipolar, or manic-depression. ‘Schizophrenia spectrum disorder’ referred to schizophrenia or any other type of psychosis. ‘Any psychiatric disorder’ included depression, mania, psychosis, anxiety, substance use, PTSD, OCD, eating disorders, and insomnia.
Results
The first onset of a psychiatric disorder during the perimenopause was reported by 0.88% of females. This is equivalent to a rate of 2.33 new cases per 1,000 person years. In other words, if we monitored 1,000 perimenopausal females for a year, there will be an average of 2.33 new cases of psychiatric disorder in that time.
Rates of first onset psychiatric disorder in the postmenopausal period (0.50%) were similar to the premenopausal period (0.59%) at 1.53 and 1.66 cases per 1,000 person years respectively.
The risk varied by nature of psychiatric disorder, with the largest increase in perimenopausal risk seen for new onset mania with a relative risk of 2.12 (95% CI 1.30 to 3.52). This means that the risk of having a first episode of mania doubles during the perimenopausal period. However, the absolute risk remains low at 0.11 per 1,000 person years during the perimenopause and 0.05 outwith that time.
New onset major depressive disorder during the perimenopause had a relative risk of 1.30 (95% CI 1.16 to 2.45).
The perimenopause was not associated with an increased risk of new onset schizophrenia spectrum disorders (relative risk 0.95 (95% CI 0.48 to 1.88). However, the number of new cases of schizophrenia or related disorders in the sample was very small (fewer than 50).
Similar patterns were not seen in males of the same age.
The risk of having a first onset of any psychiatric condition was significantly increased during the perimenopausal period, with the largest increase in risk seen for new onset bipolar disorder.
Conclusions
A major take-home message from this study is that 99% of females did not experience a new onset psychiatric condition during the two years either side of their final menstrual period. This is reassuring.
However, for those who are in some way vulnerable, the two years either side of the final menstrual period represent a time of increased risk for new onset bipolar and major depressive disorder.
Reassuringly, 99% of females in the relatively healthy UK Biobank cohort did not experience a new onset psychiatric disorder during the perimenopause. However, the risk of the rare outcome of new onset mania was doubled.
Strengths and limitations
The key strengths of this study are firstly the large sample size and secondly being able to assess age at menopause using participant’s self-report rather than relying on age as a proxy. Relying on age has been a limitation of much previous work, as the age at natural menopause varies widely between individuals. In this sample it ranged by 20 years from age 40-60 (reported in this paper’s supplemental text).
A partially addressed limitation of this study is selection bias. The proportion of participants who report severe mental illness in the UK Biobank is much lower than the proportion in the general population, suggesting that those who have experienced severe mental illness are less likely to volunteer. Participants in the UK Biobank are also not representative of the UK population with regard to risk factors for mental illness: they are wealthier, slimmer, drink less, and smoke less. The study performed a number of supplementary sensitivity analyses, which found largely similar effects of perimenopause on new onset psychiatric disorders in participants at extremes of these characteristics within the UK Biobank. Participants are also less likely to be of a minority ethnicity, which may be important as there are ethnic differences in both the experience of menopausal symptoms and the risk of mental disorder. This selection bias would make the study likely to underestimate an effect rather than find a false positive.
As the study relies largely on retrospective self-report data there may also have been recall bias. Participants will have been aware of when their last menstrual period was and may have falsely recalled the onset of their mental health issue as being around that time.
There may also have been some classification bias in that the assessment of psychiatric disorders relied on self-report questionnaires or participants’ description of symptoms to a trained nurse and was not by gold-standard clinical interview. This is a trade off as the large sample size would have made such assessment very expensive.
The lack of detailed psychiatric disorder classification available means that the study has used the term “mania” to describe a bipolar episode of any type and has not been able to distinguish depressive vs. manic, mixed, or hypomanic episodes in bipolar.
The study’s definition of perimenopause and the windows of time used for pre and post menopause are not standard in the field, but this is an emerging field of study and these time windows represent pragmatic choices.
Age and life events related to age are the major potential confounders. The authors’ analysis of age-matched males partially addressed this by showing no similar pattern of age on new onset psychiatric disorders in males, but could not account for an interaction between age and sex, for example, culturally; ageing men are often viewed more positively than ageing women.
Another potential confounder is hormone replacement therapy (HRT), which is a treatment often commenced in the perimenopausal period for troublesome menopausal symptoms. Just over a third of those asked had ever used HRT at baseline in the UK Biobank (publicly available data). Although systemic HRT is associated with improvements in perimenopausal depressive mood in the majority, potentially it could be associated with unhelpful mood changes in some, such as those sensitive to dysphoric side effects of progestogens (Sharma et al., 2023). Of course, those struggling with mood may also be more likely to commence HRT. A sensitivity analysis exploring whether rates of new perimenopausal psychiatric disorder were similar in those on and not on HRT would have been interesting, albeit hard to interpret.
Limitations of this study include selection bias (the UK Biobank is not representative of the UK population as a whole) and that the reliance on retrospective self-report to determine mental disorder may have introduced recall and classification bias.
Implications
As a practicing psychiatrist, this study made me reflect on my practice and training. I have cared for many women who have become abruptly mentally unwell in midlife either for the first time or after a long period of stability, but the potential role of the menopause has not been part of my assessment. This study makes me think we should more often ask patients about their menopausal status and experience of menopausal symptoms, as it may be relevant to understanding why they have become unwell. This knowledge alone could be helpful for the narrative around an often devastating episode of illness in a person’s life which otherwise is experienced as “out of the blue” or attributed to something else. In the future, knowing that someone is experiencing a perimenopausal mood episode may even inform treatment choice and prognosis.
The key research avenue opened up is to ask why some people who have reached midlife without experiencing mental disorder are vulnerable to becoming severely mentally unwell around the time of the menopause. This is a time of biopsychosocial change in many areas of a person’s life, and there are many candidate mechanisms. Understanding what matters most could improve prediction and open up new treatment modalities.
