The relationship between delivery mode and postpartum depression: a systematic review and meta-analysis

Background: To assess the relationship between delivery mode and postpartum depression and to examine whether cesarean section (CS) has a higher risk of postpartum depression than vaginal delivery (VD). Methods: We searched the Cochrane Library, PubMed, and EMBASE from inception to 30 April 2019 without language limitations. Two authors independently selected studies, assessed the quality of included studies, and extracted data. Any disagreements were resolved by discussion with a third author. We used the Newcastle-Ottawa Scale and GRADE methods to assess the quality of the included studies and evidences. This study had four included cohort studies data and carried out xed-effect model meta-analysis. Results: The ndings demonstrated a signicant difference in the risk of postpartum depression between CS and VD. Compared with the control group, the CS group was associated with a higher prevalence of postpartum depression symptoms (Risk Ratio =1.29; 95% CI: 1.11-1.51). Conclusions: The ndings supported a relationship between delivery mode and postpartum depression. Particularly, we found that CS


Background
Postpartum period has a high probability for mental disorder. Postpartum depression, rst identi ed by Pitt in 1968, is one of the common complications after giving birth [1]. This symptom may relate to psychological disturbances between mothers and children [2].
From the clinical viewpoint, the delivery mode could be divided into cesarean delivery (CS) and vaginal delivery (VD). CS is a relatively complicated pregnancy outcome compared with VD.
Previous studies identi ed that CS has a higher risk of postpartum depression than VD [3]. Some other evidence-based studies also showed the same ndings about the relationship between the type of delivery and postpartum depression [4][5][6]. However, some researchers pointed out that there was no obvious association between mode of delivery and maternal postpartum psychiatric disorders [7][8].
In clinical practice, quite a few countries appeal to control the high rate of CS in recent years [3]. When mothers require CS procedures, clinicians should consider the impact of delivery mode on maternal postpartum mental disorders, especially the relationship between CS and postpartum depression. In addition, clinicians should convince pregnant women without medical indications of CS not to select caesarean, the association between delivery mode and postpartum depression is an essential clinical issue. Thus, this study is conducted to explore the impacts of delivery mode on postpartum depression and to compare the risk of postpartum depression in CS and VD.

Literature search and search strategy
In this study, we searched the Cochrane Library, PubMed, and EMBASE for relevant studies from inception to 30th April 2019. The search strategy was '((modes of delivery OR delivery mode OR delivery experience OR type of delivery OR delivery or childbirth) AND (cesarean section OR cesarean delivery OR cesarean OR elective cesarean delivery OR emergency cesarean delivery) AND (vaginal delivery OR normal delivery OR normal vaginal delivery OR natural delivery OR vacuum extraction) AND (postpartum depression OR depression OR postpartum OR depression postnatal OR postpartum depressive symptomatology OR maternal depression OR mental disorder))' with no limitations on language. Table 1 shows the details of the search strategy. The protocol of this systematic review was registered in PROSPERO under the number CRD42019148154. Studies were included if they satis ed the following inclusion criteria: (1) these studies belonged to a cohort study design, (2) the exposure group were women delivered by cesarean and the control group were vaginal delivery, and (3) the outcomes of these articles were the number of postpartum depression patients of the exposure group and that of control group, respectively. For all the publications that accord with our inclusion criteria, we scanned their titles and abstracts to determine whether they have relevant information, and then read the full text of relevant articles to evaluate whether to include them or not.
Two authors (Chen and Tung) independently selected the related studies, and then their differences were resolved by discussing with a third author (Chien). Several remarks are made to clarify our selection strategy: 1) Since our goal was to explore the relationship between delivery mode and postpartum depression, we did not include studies about prenatal or pregnant depression. 2) In the cesarean section could have elective or emergency CS, the difference between them is that elective CS is decided before the operation [9]. 3) We excluded cross-sectional or case-control studies that could not assess the causal relationship between mode of delivery and depression in the postpartum period.

Data extraction and quality assessment
We extracted the following data from the included studies by using an extraction form: rst author, publication year, country, database used, study design, study duration, study subjects, and outcomes. We also used the Newcastle-Ottawa Scale (NOS) to assess the quality of the included cohort studies. Three domains (selection of study groups, comparability, and outcome assessment) of the NOS were used to evaluate the quality of the cohort studies [4]. Each item of selection and outcome category could have a maximum of one star, except the comparability, which could have up to two stars. As a result, a highquality study was awarded with seven or more stars.

Interrater reliability for selection and data extraction
To establish a consistent selection and data extraction, the kappa statistic was used to assess the interobserver reliability between the two independent reviewers, which showed a kappa value of 0.755 for data extraction (95% con dence interval (CI): 0.671-0.839).

Statistical Analysis
Review Manager 5.3 (The Nordic Cochrane Centre, The Cochrane Collaboration, 2014) was used to conduct the meta-analysis. The risk of postpartum depression is presented in terms of the risk ratio (RR) and 95% con dence interval (CI). The I2 statistic, which evaluates the degree of variation across studies due to heterogeneity rather than chance alone [5], was used to assess the heterogeneity. Speci cally, if a study's I2 value is higher than 50%, we considered it has substantial heterogeneity and performed random-effects model meta-analysis. On the contrary, if the I2 value of a study is below 50% (I 2 = 36%), we carried out xed-effect model meta-analysis. A funnel plot was used to evaluate the publication bias of each outcome. We further used the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) method to summarize and assess the quality and certainty of the available evidence.

Literature Search and Studies Characteristics
As illustrated in Fig. 1, our search recognized 3352 records after removing duplicates. Four cohort studies were included in this systematic review. Eventually, these four studies offered data for meta-analysis. Particularly, one of the four involved studies had two sets of data that acquired from the CS group and VD group, respectively [6]. We treat these two data sets as two different studies. We combined this study with the other three studies, because all of them had the same outcome indicators. The characteristics of the included studies are summarized in Table 2. As can be seen, the published time of these studies range from 2011 to 2017. All four cohort studies were population-based cohort studies from Iranian, Italy, and China. In addition, all four of the included studies had seven or more stars on the NOS scale and were of high quality.  [11]. As shown in Fig. 2, the xed effects model meta-analysis demonstrated a signi cant difference in the risk of postpartum depression between CS and VD. The total RR is 1.29 (95%CI: 1.11-1.51). The studies had a small statistical heterogeneity(I²=36%). In addition, the Funnel plot showed little publication bias (test for overall effect: = 3.25, = 0.001) (Fig. 3).

GRADE Summary of Findings
The results of the GRADE assessment are summarized in Table 3. Particularly, the quality of the evidence from the included observational studies was initially judged to be low. Meanwhile, all evidence did not have serious of risk of bias, inconsistency, indirectness, imprecision, as well as publication bias, and the quality of all evidence is still low without downgrade. The risk in the intervention group (and its 95% con dence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Con dence interval; RR: Risk ratio GRADE Working Group grades of evidence High quality: We are very con dent that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately con dent in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our con dence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have little con dence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect 4. Discussion

Clinical Implications
To the author's best knowledge, few systematic reviews and meta-analyses examined the relationship between delivery mode and postpartum depression. We analyzed the evidence about the postpartum depression rates in CS and VD, respectively. Our study not only supported the hypothesis that delivery type is relevant to the incidence depression in the postpartum period, but also concluded that CS is associated with a higher risk of postpartum depression symptoms than VD.
Postpartum depression is a multi-factorial behavior related to delivery mode [10], maternal age [11], season of delivery [12], and social support [13]. The mode of delivery has substantial impact on physical and psychological health [14] and is one of the most concerned variates in research about postpartum depression [15]. However, few studies focused on the association between the type of delivery and the incidence of postpartum depression. To examine the causal relationship between delivery mode and postpartum depression, our study included only cohort studies. Generally, ordinary meta-analyses on the e cacy of interventions obtain high quality evidence from randomized controlled trials [5]. However, randomized trials are not suitable for our study, beause the duration of randomized trials is often too short to explore long time or rare adverse consequences [16]. It is also impossible to randomize mothers into the categories 'CS' or 'VD'. Therefore, only cohort studies are included in our study and the obtained results are shown to be effective. All included cohort studies may reveal the effects of delivery mode on maternal mental disorders and further effects of the different modes of delivery on postpartum depression.

Methodological Considerations
From the methodological perspective, there are some limitations in our study. Firstly, only four cohort studies were identi ed, which may make the results of the funnel plot unreliable since the results might be in uenced by the low statistical power of insu cient studies [17]. In addition, the included studies were limited to the northern hemisphere. Whether the risk of postpartum depression differs in patients with different delivery modes elsewhere in the world is unclear [18]. Secondly, we were unable to conduct subgroup analyses in terms of maternal age, season of delivery, social support, indications for cesarean section, and neonatal admission to the intensive care unit because the selected cohort studies did not offer enough information. Thirdly, although the kappa statistics for the agreement of interobserver reliability seemed acceptable, nondifferential misclassi cation or bias data extraction still may have occurred. Fourthly, according to our obtained results from GRADE assessment, the evidence from the included observational studies was initially rated as low quality and later not upgrade [5,19]. More research with precision and abundant data in the future are likely to increase the quality of the evidence. Fifthly, only one study indicated that postpartum depression is more common in vaginal delivery than cesarean section [6]. Although the VD group showed a more decreased Edinburgh Postnatal Depression Scale (EPDS) from 2 months to 4 months after delivery, it is di cult to have enough information to estimate the 2 months postpartum depression in this study. Sixthly, this study's search keyword is "vaginal delivery OR normal delivery OR normal vaginal delivery OR natural delivery OR vacuum extraction". However, there are several types vaginal delivery, including spontaneous vaginal delivery (SVD), assisted vaginal delivery (AVD) or instrumental vaginal delivery, induced vaginal delivery, and normal vaginal delivery (NVD), which means that we might missed some of the studies related to this topic. Seventhly, from the included VD types might cause issues such as vaginal tears, retained placenta, postpartum hemorrhage (PPH), and shoulder dystocia could lead to maternal physical and psychical problems. Mothers' states are also very dependent on the baby's condition; we do not focus on the maternal population whose babies had abnormalities. Finally, Chen et al (2017) included most of the numbers, however, their study did not show the signi cance result. We also have a funnel plot (Fig. 3) to prove there is no publication bias.

Conclusions
Although the available best evidence supports an association between delivery mode and postpartum depression, more large-scale cohort studies are required to further reveal the relationship between the mode of delivery and depression in the postpartum period. Future studies also may divide the delivery modes into detailed modes such as elective cesarean delivery, emergency cesarean delivery [20], planned vaginal delivery, assisted vaginal delivery, spontaneous vaginal delivery [21,22], uncomplicated spontaneous vaginal delivery, complicated vaginal delivery [23], to study the role of delivery modes in postpartum.

Availability of data and materials
All data underlying the ndings are within the paper.  PRISMA2009checklist.doc