Associations Between Polycystic Ovary Syndrome and Adverse Obstetric and Neonatal Outcomes

A Population Study of 9.1 Million Births

Ginevra Mills; Ahmad Badeghiesh; Eva Suarthana; Haitham Baghlaf; Michael H. Dahan


Hum Reprod. 2020;35(8):1914-1921. 

In This Article


The objective of this study was to determine the independent effect of PCOS on the likelihood of experiencing adverse delivery and neonatal outcomes using a large nationwide population-based healthcare database. Multiple studies in the literature have been published which have assessed the possible increased risk of PTD, IOL, C/S delivery and SGA infants in women with PCOS (Boomsma et al., 2006; Palomba et al., 2010, 2018; Qin et al., 2013; Yu et al., 2016; Khomami et al., 2019). Unfortunately, there are significant disagreements within the scientific literature about such attributable risks. The largest meta-analysis to date on this subject includes 63 studies and 224 136 pregnant women, including PCOS and controls and suggests an increased risk for C/S delivery and IOL in women with PCOS (Khomami et al., 2019). These results are not in agreement with the next largest meta-analysis conducted in 2013, which included 27 studies involving 4982 women with PCOS and 119 692 control women (Qin et al., 2013). This meta-analysis suggested that there was not an increased risk of either C/S or IOL. Furthermore, this publication also failed to detect an increased risk in PCOS women for PTD or SGA babies (Qin et al., 2013). These findings conflict with later studies, which suggest an increased risk for both PTD and SGA infants (Itskovitz-Eldor et al., 2000; de Wilde et al., 2017; Lonnebotn et al., 2018; Zheng et al., 2019; Foroozanfard et al., 2020). These contradictory findings between various studies and meta-analyses are likely due to the inherent difficulties in comparing different study designs, as well as the large number of confounding effects on such delivery outcomes, such as GDM, HDP and PEC. Pre-existing medical conditions may also complicate interpretations. These pre-existing and pregnancy-induced metabolic conditions are known to both occur more often in women with PCOS and to increase the risk of delivery and neonatal outcomes (Joham et al., 2016; Christ et al., 2018), including those assessed in this study. Our large population study, which included 9 096 788 pregnant women analysed concurrently and with identical methodology, provides an evaluation of the independent risk of PCOS on adverse delivery and neonatal outcomes while effectively controlling for these potential confounders. It also represents the largest study on the subject to date with more than 40 times more subjects than the next largest study (Khomami et al., 2019).

The prevalence of PCOS in our study population of pregnant women whose pregnancies resulted in a live birth was 163.6 per 100 000 pregnant women. This prevalence is lower than the previously published prevalence of PCOS in the general population of 8–10% (Azziz, 2006; Lizneva et al., 2016). A large proportion of women with a known diagnosis of PCOS have difficulty achieving (Norman et al., 2007) and maintaining a pregnancy, and would therefore be excluded from our population, thereby significantly decreasing the prevalence in our study population. The lower prevalence in our study population also suggests that PCOS likely remains an under-recognised condition by medical professionals (Azziz and Adashi, 2016). It is possible, therefore, that PCOS women who conceived without the assistance of fertility treatments were erroneously not identified as having the diagnosis, thereby remaining in the control group. However, data on the use of fertility treatments are available in the database, and a large proportion of the women in the PCOS study group did not require fertility care to conceive suggesting that such an error would remain minimal. Despite the lower prevalence of PCOS in our study population, we did observe a significant increase in prevalence of the condition from 49.1 per 100 000 in 2004 to 366.04 per 100 000 in 2014. The cause of this steep increase in prevalence is likely multifactorial. First, with the publication of the Rotterdam criteria for objective diagnosis of PCOS in 2003, it was expected that more cases of PCOS would be observed by clinicians, thereby adding more ovulatory subjects to the diagnosed group. As well, we suspect physicians are more knowledgeable about PCOS which has led to greater diagnosis. Secondly, a complex metabolic relationship exists between obesity and PCOS (Joham et al., 2016), and given the rapid increase in obesity rates in the USA, it is not surprising that the increase in prevalence of PCOS would mirror this trend. Finally, as techniques for ovulation induction and IVF treatments for women with PCOS advance, more women with PCOS are able to become pregnant (Sterling et al., 2016).

Our data support the largest and most recent meta-analysis to date, which concludes that PCOS is an independent risk factor for C/S delivery (Khomami et al., 2019). Our study suggests an OR of 1.5 for the risk of C/S delivery (after controlling for all possible confounders), which is higher than any other published study. Earlier and smaller meta-analyses were unable to detect this increased risk, suggesting that the samples in such groups were insufficiently sized to detect a difference in the presence of significant confounding factors (Qin et al., 2013; Palomba et al., 2018). The higher rates of C/S in this population may be related to the propensity for impaired glucose tolerance independent from a diagnosis of diabetes (Gilbert et al., 2018). Unfortunately, the database does not capture women with a single pathological value on an oral glucose tolerance test as they would not have been coded as diabetic. As such, women with impaired glucose tolerance may either grow larger babies, increasing their risk for labour dystocia, or they may experience abnormal placentation (Palomba et al., 2014b), which would predispose them to abnormal foetal heart rate patterns in labour, all of which would necessitate C/S delivery. Furthermore, increased C/S rates are often related to increased rates of pregnancy complications seen in the PCOS. Interestingly, the PCOS and control groups had similar rates of labour induction in conjunction with such a different C/S rate despite the fact that the pregnancy complications seen in the PCOS group may have favoured increases in labour induction. It is important to consider that PCOS women in this study group were characterised by a higher household income, more private insurance and more treatment in urban teaching hospitals, which may suggest that there are differences in practice patterns by physicians who treat patients in such settings. These differences may contribute to the similar rates of labour induction in the PCOS and controls in the setting of higher C/S rates.

When singleton pregnancies were evaluated in our data, women with PCOS were 1.55 times more likely to experience PPROM and 1.44 times more likely to experience PTD. These finding are, again, in agreement with several previously publications (Khomami et al., 2018; Wang et al., 2019; Zheng et al., 2019), despite a 2013 meta-analysis failing to detect a difference in the risk of PTD (Qin et al., 2013). Interestingly, the risks of both PPROM (OR 1.27) and PTD (OR 1.28) are increased less in multiple gestations pregnancies with PCOS than they are in singleton pregnancies with PCOS. This finding is likely due to the fact that in groups with multiple gestations, the multiple pregnancies themselves are the greatest risk for PPROM and PTD, overshadowing the contribution of PCOS.

Our data further suggests an increased risk of placental abruption (OR 1.63) in women with PCOS, which is an outcome that has not been previously reported in the literature. It has been suggested that PCOS can influence early placental development in women with PCOS, often leading to smaller placentas which are more prone to the development of microthombosis (Palomba et al., 2013, 2014b). Therefore, it is possible that placentas in PCOS-affected pregnancies are at increased risk of early failure and detachment from the uterine wall. It can also be hypothesised that the IR of both PCOS and pregnancy in isolation (Koster et al., 2015), through metabolic modulation together may synergistically alter early placentation in the ways described above. Furthermore, such changes in early placentation may also increase the risk of pregnancy-associated hyptertensive disorders, which are also additive risk factors for placental abruption.

Our study found no statistical difference in the number of women undergoing IOL between the PCOS and the reference group overall for all pregnancies in this study. These findings are not in agreement with the most recently published meta-analysis, which suggests an OR of 2.55 for PCOS women to undergo IOL (Khomami et al., 2019). Given the high prevalence of pre-existing diabetes, GDM, HDP and PEC in PCOS pregnancies (Qin et al., 2013; Palomba et al., 2015; Yu et al., 2016), and the fact that most pregnancies complicated with these conditions will require IOL, it is not surprising that more pregnancies with PCOS could be categorised as requiring IOL. However, given the strength of our study in being able to control for all these confounders, we remain confident that PCOS itself is not the risk factor leading to this increased risk of IOL. This is a rather important finding, as previous publications have suggested that the higher rates of PTD in women with PCOS are secondary to an increased rate of IOL. However, our data suggest this is not the case and that the increased risk of PTD in women with PCOS occurs despite there being no increased risk of undergoing IOL.

The neonatal outcomes assessed in our study included SGA, IUFD and congenital anomalies. Previous literature suggests that the risk of SGA infants is increased in pregnant women with PCOS (Wang et al., 2019; Zheng et al., 2019; Foroozanfard et al., 2020). Our data, however, do not support this finding, as we report no difference in the occurrence of SGA infants between the PCOS and control groups when all pregnancies are analysed. Our data do, however, suggest a slightly increased risk of SGA babies in PCOS women with multiple gestation (OR 1.27, P = 0.04). Given the significantly smaller sample size of the multiple gestation group and the fact that the CI for this value is just above 1, however, it is highly likely that this is a clinically marginal finding.

Given the extremely rare occurrence of IUFD with today's modern obstetrical care, there are no published studies directly assessing the risk of PCOS on IUFD. Not surprisingly, despite the size of our database, we were unable to detect a difference in the risk of IUFD in women with PCOS. Interestingly, babies born to women with PCOS were almost two times more likely to be born with congenital anomalies than those without PCOS. Although this finding has also not previously been reported in the literature, it is not unexpected. Hyperglycaemia during early pregnancy and embryo development is a known teratogen and is widely cited as causing cardiac, limb and neural tube defects (Barbour, 2014). In pregnancy, women with PCOS compared with non-PCOS controls demonstrate impaired glucose tolerance and hyperinsulinaemia in early pregnancy even without an overt diagnosis of diabetes (Kelley et al., 2019). Therefore, given the high concurrent presence of pre-existing diabetes, GDM and hypergylaecima in women with PCOS (Dahan et al., 2017; Christ et al., 2018; Dahan and Reaven, 2019), it follows that there would be an increased risk of such anomalies in infants born to women with PCOS, as these share a similar pathophysiological process. This finding persists even after controlling for pre-gestational diabetes and GDM, which suggests that PCOS itself may be a risk factor for congenital anomalies and the underlying mechanism should be studied in the future.

The relationship between PCOS and the development of chorioamnionitis and post-partum maternal infections has not previously been reported in the literature. A study by Koster et al. (2015), which analysed placental features of PCOS suggested an increased occurrence of chorioamnionitis and funisitis in placental gross pathology specimens. Although this would likely correlate to a higher number of clinical cases of chorioamnionitis in PCOS women, this has never been studied. Our study, therefore, is the first to demonstrate that women with PCOS are 1.6 times more likely than non-PCOS women to develop chorioamnionitis in labour. Furthermore, these women are at a similarly increased risk of developing post-partum maternal infections (such as post-partum endometritis or caesarean and perineal wound infections). The underlying aetiology of this increased risk of infection remains elusive. Women with PCOS may experience more prolonged or dysfunctional labour, secondary to higher rates of obesity and nulliparity, which may be a contributing factor to the higher rate of infectious morbidity. Ideally, length of labour, birthweight and nulliparity would be controlled for; however, such an analysis in this database is unavailable. However, it can be postulated that pre-existing hyperglycaemia (Palm et al., 2018) and low grade chronic inflammatory processes secondary to PCOS (Palomba et al., 2014a) could predispose these women to the development of such infections. The increased propensity for PCOS women to develop subclinical infections might also serve as an underlying cause for the increased risk of these women to also experience PPROM.

This study is not without its limitations. This is a retrospective analysis utilising an administrative database which relies on the accuracy and consistency of the individuals coding the data. Therefore it is possible that a remote diagnosis of PCOS or a milder phenotype or PCOS may not have been reported. In such cases, we may not have captured the entirety of the pregnant PCOS population. Despite the probability that some undiagnosed PCOS women remain in the control group, the results of this study are consistent with most smaller studies in the literature and rates of most complications were higher in our data than in the previously published studies. There are also known limitations in how accurately hospital coding is able to capture perinatal conditions and complications (Moore et al., 2006). Furthermore, the database does not provide information on parity, birthweight and duration of labour, which are potential confounding variables that cannot be accounted for in our analysis. Despite these limitations, having access to such a large data repository allows for a comprehensive assessment of the effects of PCOS on delivery and neonatal outcomes, despite its close association with multiple confounding variables. Our study, with over 9 million pregnant women and almost 15 000 of them with a known diagnosis of PCOS allows the ability to control for the known confounding effects of pre-existing conditions such as pre-gestational diabetes, pre-existing HTN and obesity, as well as the development of GDM, HDP and PEC, which are known to occur more frequently in PCOS. Our study is also the first to our knowledge to report on the increased risks for placental abruption, chorioamnionitis and maternal infections in women with PCOS and congenital anomalies in their offspring, compared to those without. Our findings, therefore, are important in clarifying conflicting data in the literature regarding increased risks of perinatal outcomes in relation to a diagnosis of PCOS. Furthermore, these findings represent novel data that can be helpful in managing and ameliorating the risk of maternal and foetal infection in the puerperium.