Volume 97, Issue 7 p. 838-844
Original Research Article
Free Access

Postnatal depression in a community-based study of women with polycystic ovary syndrome

Wendy A. March

Corresponding Author

Wendy A. March

Adelaide Medical School, The University of Adelaide, Adelaide, Australia

Robinson Research Institute, The University of Adelaide, Adelaide, Australia

Correspondence

Wendy A. March, Obstetrics & Gynecology, The University of Adelaide, Adelaide, South Australia 5005, Australia.

E-mail: [email protected]

Search for more papers by this author
Melissa J. Whitrow

Melissa J. Whitrow

Robinson Research Institute, The University of Adelaide, Adelaide, Australia

School of Public Health, The University of Adelaide, Adelaide, Australia

Search for more papers by this author
Michael J. Davies

Michael J. Davies

Adelaide Medical School, The University of Adelaide, Adelaide, Australia

Robinson Research Institute, The University of Adelaide, Adelaide, Australia

Search for more papers by this author
Renae C. Fernandez

Renae C. Fernandez

Adelaide Medical School, The University of Adelaide, Adelaide, Australia

Robinson Research Institute, The University of Adelaide, Adelaide, Australia

School of Public Health, The University of Adelaide, Adelaide, Australia

Search for more papers by this author
Vivienne M. Moore

Vivienne M. Moore

Robinson Research Institute, The University of Adelaide, Adelaide, Australia

School of Public Health, The University of Adelaide, Adelaide, Australia

Fay Gale Center for Research on Gender, The University of Adelaide, Adelaide, Australia

Search for more papers by this author
First published: 20 February 2018
Citations: 13
The authors have stated explicitly that there are no conflicts of interest in connection with this article.

Abstract

Introduction

Women with polycystic ovary syndrome are susceptible to depression and anxiety and so may also be at risk for postnatal depression. This study investigates whether women with polycystic ovary syndrome have an elevated risk of postnatal depression.

Material and methods

Cross-sectional data for parous women (= 566) were available from a birth cohort. Polycystic ovary syndrome was diagnosed using the Rotterdam criteria. Details of reproductive history, pregnancy, birth, and postnatal depression were obtained through structured interview. Comparisons were made between women with and without polycystic ovary syndrome using logistic regression analysis, including the investigation of interactions.

Results

A positive but statistically non-significant association was found between polycystic ovary syndrome and postnatal depression (odds ratio 1.6, 95% confidence interval 0.9–2.9). Compared with their counterparts, women with polycystic ovary syndrome were substantially more likely: to have difficulty conceiving (odds ratio 5.2, 95% confidence interval 2.9–9.4), to have conceived with medical assistance (odds ratio 11.6, 95% confidence interval 5.5–24.4), and to have pregnancy complications (gestational diabetes, pregnancy-induced hypertension, or preeclampsia; odds ratio 2.0, 95% confidence interval 1.1–3.5). Where women with polycystic ovary syndrome had a history of miscarriage or conceived with medical assistance, the combination interacted (= 0.06 and < 0.05, respectively), with over half of such women having postnatal depression.

Conclusions

Although women with polycystic ovary syndrome may not have an excess risk of postnatal depression overall, those who had suffered a miscarriage or required medical assistance to conceive were at substantially elevated risk. Findings point to vulnerability inherent in polycystic ovary syndrome being amplified, either by stressful experiences on the pathway to pregnancy/childbirth or by specific fertility treatment regimens.

Abbreviations

  • BMI
  • body mass index
  • CES-D
  • Center for Epidemiological Studies Depression Scale
  • PND
  • postnatal depression
  • PCOS
  • polycystic ovary syndrome
  • Key Message

    As the first study to investigate postnatal depression (PND) risk for women with polycystic ovary syndrome, we revealed a substantially greater risk of PND if miscarriages were experienced or medical assistance to conceive sought.

    Introduction

    Maternal depression associated with pregnancy and childbirth, known as postnatal depression (PND), is a debilitating condition that can have a significant impact on the health and wellbeing of the woman, her child and her wider family 1, 2. PND is likely to be under-diagnosed but is estimated to affect 10–20% of mothers in western countries 2-4.

    Polycystic ovary syndrome (PCOS) is an endocrine disorder, manifesting in symptoms that include hirsutism and menstrual irregularity. Women with PCOS have increased susceptibility to depression and anxiety 5, which is an identified risk factor for PND 2, 4. This susceptibility could be due to their perturbed hormonal and metabolic profiles or to distress created by PCOS symptoms and their management 6-8.

    In addition, women with PCOS often have difficulty conceiving as well as complications during pregnancy, which may contribute to stress and dissatisfaction and, thereby, to PND 2, 4. To elaborate, women with PCOS have elevated risk of miscarriage, gestational diabetes, pregnancy-induced hypertension and preeclampsia 9. Impaired fertility means that women with PCOS are frequently clients of fertility clinics 5, 10, 11, a source of stress during treatment 12, 13. Pregnancies achieved with medical assistance are, in turn, associated with increased risk of complications 14-16. Finally, women with PCOS have relatively more neonates that are preterm or small-for-gestational-age, partly due to an excess of twin pregnancies following use of ovulation induction drugs 9.

    Thus, women with PCOS are likely to have a variety of risk factors for PND, including a history of depressive symptoms, experience of miscarriage, obstetric and perinatal complications, but there has been no investigation of PND in this group to date. As PCOS is the most common endocrine disorder in women of reproductive age 8, affecting between 9 and 18% 18, it may contribute to a substantial burden of PND.

    We aim to address this gap in knowledge by comparing the occurrence of PND in women with and without PCOS in a well characterized community-based sample. We also explore whether prior experiences of difficulty conceiving, medically assisted conception, pregnancy complications or suboptimal birth outcomes contribute to any increased risk of PND.

    Material and methods

    This work is based on the first wave of follow up undertaken for a retrospective cohort study of women, thus it has a cross-sectional design. Cohort participants were born in 1973–75 in a large maternity hospital in Adelaide, South Australia. Around 30 years later, the 2199 eligible female births were traced. The 1984 (90.2%) women who were confirmed still to be living, without any severe impairment, were invited to participate by letter and telephone [for further details of the cohort establishment, see March et al. 17].

    After acceptance, first wave data were obtained from 947 (49.1%) women at a median age of 30.2 years. Participants were broadly representative of all eligible female births, although with a slightly greater proportion of mothers from the highest socioeconomic category (8.8% vs. 5.5%, as derived from postcode of residence) 18. Among participants, 568 (60.0%) had given birth to one or more children by the time of follow up and all but two had completed questions on PND; these 566 women form the current analysis set.

    In an appointment with a trained research nurse, women completed a structured interview and a series of questionnaires. Administration was in person for over three-quarters of women and by telephone for the remainder, principally because they no longer resided in Adelaide. The structured interview covered many aspects of women's lives, symptoms of PCOS, and medical history, including pregnancies. For women seen in person, height and weight were measured following standard protocols, otherwise self-reports were obtained.

    The Rotterdam criteria were used to identify women with PCOS 19, i.e. a positive diagnosis of PCOS required the presence of at least two of the following three symptoms: oligo- or amenorrhea; clinical and/or biochemical hyperandrogenism; polycystic ovaries. The process of identifying these symptoms has been described in detail elsewhere 17. Briefly, women provided details of their menstrual cycle, and menstrual irregularity was defined as chronic amenorrhea, a cycle length of <21 or >35 days, or variation of more than four days between cycles. Women were shown pictures representing the modified Ferriman–Gallwey scale which were compared with body hair, or reports of the extent prior to removal, to classify hirsutism, with a score of more than 7 corresponding to clinical hyperandrogenism. Women were asked to provide a morning blood sample as soon as convenient after the appointment, where possible in the follicular phase; free testosterone was used to determine biochemical hyperandrogenism. Women with at least one symptom were offered a transvaginal ultrasound to ascertain the presence of polycystic ovaries.

    Women who reported a pregnancy of more than 20 weeks were asked whether they had experienced depression when pregnant or postnatally. This was cross-checked against another section of the structured interview concerning depression, in which women provided details of episodes, whether these were clinically diagnosed, and any treatment. The Center for Epidemiological Studies Depression Scale (CES-D) was also completed, with a score of 16 or more indicating current symptoms consistent with clinical depression 20, 21.

    Women provided a pregnancy history that included miscarriages, premature birth (<37 weeks, gestation) and multiples. Women were asked whether they had difficulty conceiving and, if so, for details. These included whether a practitioner of any kind was consulted (including alternative medicine) and whether conception occurred with medical assistance (including ovulation induction drugs, in vitro fertilization or other procedures). Women reported whether they had gestational diabetes, pregnancy-induced hypertension or preeclampsia.

    At follow up, women reported the highest level of education they had reached and whether they were living with a partner at the time of follow up (married or de facto). Age at first birth was determined from the dates of birth of the woman and her first born child. Body mass index (BMI), using measured or self-reported height and weight, was calculated as weight (kg) divided by the square of height; women were classified as overweight or obese using cut-offs of 25 and 30 kg/m2, respectively.

    Statistical analyses

    Almost all variables required for this analysis were categorical so, for simplicity, age was classified in five-year bands. There were some missing data items (= 25 for BMI, = 41 for CES-D), as women did not have to provide information if they were uncomfortable about any question(s). There were two missing values for PND (both in women who did not have PCOS), resulting in analyses being based on = 566 complete cases.

    Variables denoting aspects of reproductive history and pregnancy complications were considered in the analysis. It is debatable whether these should be treated as confounders or as mediators of possible associations between PCOS and PND. From a theoretical perspective, we decided to treat them as mediators with the potential to amplify any susceptibility to PND inherent in the women with PCOS. Socioeconomic status, indicated by educational attainment, is an established independent risk factor for PND 4, so it was considered as a potential confounder, although we did not anticipate it would be influential, as the distributions for women with and without PCOS had previously been shown to be similar 17.

    Descriptive statistics with proportions represented as percentages were generated. Logistic regression was undertaken to produce odds ratios (OR) and 95% confidence intervals (CI) for the association between PCOS and PND, as well as between possible confounders/mediators and PND. Logistic regression was also used to assess interactions of PCOS status and variables characterizing the course through conception-pregnancy-birth in relation to the outcome of PND.

    All data were analyzed in STATA version 14.1 (StataCorp LP, College Station, TX, USA).

    Ethical approval

    Study procedures conformed to the principles of the Declaration of Helsinki and approval was gained from the institutional review board of the hospital ethics committee and the University of Adelaide (H/36/99, March 2000). All participants gave informed written consent.

    Results

    In the analysis set of parous women, 52 (9.2%) met the Rotterdam criteria for PCOS. Three women had menstrual irregularities, hyperandrogenism and polycystic ovaries; seven had only menstrual irregularities and polycystic ovaries; eight had only hyperandrogenism and polycystic ovaries; 27 had menstrual irregularities and hyperandrogenism; some of these women may also have had polycystic ovaries as only half of those offered an ultrasound, accepted.

    Table 1 reports the characteristics of parous women with and without PCOS. Compared with other women, those with PCOS were substantially heavier (three in five with a BMI meeting the obese classification) and more likely to have depressive symptoms (half exceeding the clinical cut-off on the CES-D scale) at the time of follow up. Women with PCOS were less likely to have completed high school and although the majority were currently living with a partner, the proportion was not as high as for their counterparts.

    Table 1. Characteristics of parous women at first wave of cohort follow up (= 566)
    Characteristic PCOS (= 52) Without PCOS (= 514) p-value
    n % n %
    Age, years
    <30 12 23.1 111 21.6 0.81
    ≥30 40 76.9 403 78.4
    BMI, kg/m2, = 541
    Lean or normal, BMI < 25 12 24.5 227 46.1 <0.01
    Overweight, 25 ≤ BMI < 30) 8 16.3 128 26.0
    Obese, BMI ≥ 30 29 59.2 137 27.9
    Current depression symptoms, CES-D > 16, = 525 22 47.8 157 32.8 0.04
    Living with a partner 33 63.5 389 75.7 0.05
    Age at first birth, years
    <25 28 53.9 251 48.8 0.21
    25–29 23 44.2 217 42.2
    ≥30 1 1.9 46 9.0
    Children
    1 25 48.1 186 36.2 0.09
    2+ 27 51.9 328 63.8
    Educational attainment
    Some high school 26 50.0 192 37.4 0.05
    Completed high school 15 28.9 238 46.3
    Tertiary 11 21.2 84 16.3
    • BMI, body mass index; CES-D, Center for Epidemiological Studies Depression Scale; PCOS, polycystic ovary syndrome.

    All women in this analysis set had given birth. The association between PND and PCOS was not statistically significant, although the prevalence of PND among women with PCOS was observed to be greater than among other women (37% vs. 27%), as shown in Table 2.

    Table 2. The occurrence of reproductive health problems and pregnancy or birth complications among parous women with and without PCOS (= 566)
    Characteristic PCOS (= 52) Without PCOS (= 514) Odds ratio 95% CI
    n % n %
    Postnatal depression 19 36.5 137 26.7 1.6 0.9–2.9
    Had difficulty conceiving 27 51.9 88 17.1 5.2 2.9–9.4
    Consulted a practitioner re fertility 21 40.4 47 9.1 6.7 3.6–12.6
    Medical assistance to conceive 16 30.8 19 3.7 11.6 5.5–24.4
    History of miscarriage 16 30.8 108 21.0 1.7 0.9–3.1
    Pregnancy complications 20 38.5 125 24.3 2.0 1.1–3.5
    Preterm birth, <37 weeks 11 21.2 74 14.4 1.6 0.8–3.2
    Multiple birth 3 5.8 19 3.7 1.6 0.5–5.6
    • CI, confidence interval; PCOS, polycystic ovary syndrome.

    Also as shown in Table 2, half of women with PCOS reported difficulty conceiving and a third had conceived with medical assistance; this was markedly different from the reproductive profile of other women, as the elevated odds ratios indicate. During pregnancy, women with PCOS were twice as likely as their counterparts to have had complications of high blood pressure, preeclampsia and/or gestational diabetes.

    Figure 1 presents subgroups of women with specific risk factors for PND to depict the joint influence of PCOS and each risk factor. Among women who had experienced miscarriage, those with PCOS were more likely to have PND than those without (56% vs. 26%, < 0.05). Among women who conceived with medical assistance, those with PCOS were more likely to have PND than those without (56% vs. 11%, < 0.01). Of note, overlap in the subgroup who had experienced miscarriage and the subgroup who had conceived with medical assistance was limited (around 25%). Logistic regression analysis formally confirmed a PCOS × miscarriage interaction (OR = 3.5, 95% CI 1.0–13.1) and a PCOS × assisted conception interaction (OR = 10.7, 95% CI 1.6–72.8).

    Details are in the caption following the image
    Prevalence of postnatal depression in parous women with and without polycystic ovary syndrome (PCOS), overall, and within subgroups based on risk factors for postnatal depression. *< 0.05.

    Discussion

    In a simple analysis, the association between PCOS and PND was not statistically significant, although the data may suggest a positive trend: over one in three parous women with PCOS had PND, whereas in women without PCOS the prevalence was around one in four. Reproductive profiles of women with PCOS contrasted with those of other women in terms of difficulty conceiving, mode of conception, pregnancy complications and neonatal outcomes, although not all increases in risk were statistically significant, particularly where events were less common. Two factors were formally shown to interact with PCOS and elevate PND – a history of miscarriage and requiring medical assistance to conceive. In both instances, the combination of an adverse reproductive exposure and PCOS was associated with occurrence of PND in more than one in two women.

    In agreement with our findings on depressive symptoms at interview, other studies of women with PCOS have reported that they have an elevated risk of depression and anxiety 5, 6, 22. Although distressing symptoms associated with PCOS may contribute to relatively poor mental health profiles 6, 7, a large meta-analysis showed that physical symptoms including hirsutism, infertility and obesity did not fully account for associations between PCOS and depression, anxiety or emotional distress 22. Biochemical characteristics of PCOS were not considered in the meta-analysis and may also contribute 23.

    Our findings concerning the reproductive profiles of women with PCOS are in agreement with those reported previously, including difficulties conceiving 10, 11 and requiring medical assistance to conceive 5, 11. Existing studies have demonstrated that women with PCOS have an elevated prevalence of miscarriage 9, suggesting that the lack of a statistical significance for some differences observed in our study reflects the relatively modest sample size.

    The interactions between PCOS and both history of miscarriage and medically assisted conception in relation to PND demonstrate the role of the path to pregnancy/birth in increased susceptibility to PND. Patterns in our data suggest that a wider range of fertility-related exposures could potentially exacerbate susceptibility, but a larger sample would be needed to explore this.

    There are several possible explanations for these findings. One is that the underlying endocrine profile of PCOS predisposes some women to PND in addition to subfertility. In women with PCOS, elevated androgens have been shown to be associated with negative affect and depressive symptoms 24. In other women, elevated androgens pre- and postpartum have been associated with PND 25, but results are inconsistent 26, 27 and the relation between androgens and mood disorders in women remains controversial 28, 29.

    It is also possible that a heightened stress response in women with PCOS culminates in PND when difficulties with conception/pregnancy are encountered. Hyper-responsivity of the hypothalamic-pituitary-adrenal axis has been proposed to contribute to poor mental health, including depression, in women with PCOS 30, 31. Studies have also shown that, compared with controls matched for age and BMI, women with PCOS (without diagnoses of mental disorders) have greater plasma adrenocorticotropic hormone and serum cortisol levels following an experimental stressor, despite similar levels of emotional distress (reported state anxiety) 30. Such hyper-responsivity may mean women with PCOS are less resilient to stress caused by infertility, treatment and pregnancy complications. The possibility that interactions between stressful exposures and the endocrine system in PCOS could contribute to a range of symptoms and disorders has been raised by several authors 6, 22, 32.

    Thirdly, it is possible that specific aspects of assisted reproductive treatment contribute to PND. Exposure to ovulation induction during fertility treatment has been related to both miscarriage and mood disorders 33, 34. Women with PCOS are sensitive to ovulation induction with regard to the adverse outcome of ovarian hyperstimulation syndrome, during which mood disorders are common 34.

    This is the first study to investigate specifically the relation between PCOS and PND. A strength of this study is the use of a community-based sample, thus considering a more representative group of women with PCOS than possible with clinic-based samples.

    This study has several limitations. A proportion of women with either anovulation or hyperandrogenism declined the invitation to have an ultrasound to assess ovaries for the presence of cysts. Consequently, there is likely to be a degree of under-ascertainment of PCOS, with some women in the group without PCOS probably misclassified. This would have a conservative influence on results (i.e. reducing observed differences between the groups with and without PCOS). The analysis sample had to comprise parous women, which meant the sample size was relatively modest, resulting in some established differences between women with and without PCOS (for example, history of miscarriage) trending in the expected direction but not achieving statistical significance. Alternatively, these may truly be null associations.

    Our findings suggest that women with PCOS may be susceptible to PND, especially if they have reproductive health problems. We endorse the need for research on ovulation stimulation protocols for this patient population to reduce the risk of adverse outcomes, as highlighted by others 35. Given that many women with PCOS are in contact with health services before conception, there is an opportunity for greater awareness, monitoring and support to prevent PND.

    Acknowledgments

    We thank the clinical nurses for their role in the cohort establishment. We are also grateful to Kendal Smith for the study coordination and Nanette Kretschmer for management. Also, thank you to the many other staff members involved in interviews, database construction and data entry, and to Chris Davies for assistance with additional data queries.

      Funding

      This work by W.M. was supported by the Australian National Health and Medical Research Council funded Center for Research Excellence in Polycystic Ovary Syndrome (Grant ID APP1078444).