Statistical Analysis
This study was designed to detect a 5% absolute difference between groups with 90% power and a 5% type I error rate. We assume that the incidence of PPD was about 30% in the PE group and 15% in the control group. Therefore, a sample size of 380 (88 in the PE group and 292 in the control group) was needed. The MedSci Sample Size Tools (MSST, version 5.7.15, copyright 2020 MedSci.cn) were applied for calculating. We recruited 130 PE patients and 295 healthy women.
Variables in accordance with normal distribution were compared via an independent t-test and presented as mean ± standard deviation. Otherwise, variables were described as mean ± quartile and examined by the Kruskal–Wallis test. Differences in the classified variables were evaluated by the Chi-squared test. P < 0.05 was considered as significant. A Multivariate logistic regression model was used to evaluate adjusted odds ratios. Confounding factors including age, BMI, gestational days, baby weight, delivery model, Leakage Index Score, milk-feeding ways, and Pain Scale, which were previously reported to be connected with PPD or unmatched between PE and normal groups. All statistical analyses were conducted on SPSS 23.0 (SPSS Inc, Chicago, USA).
Results
Clinical characteristics between normal and PE mothers
A total of 130 PE patients met the inclusion and exclusion criteria. We randomly selected 295 normal women who met the inclusion criteria during the same period. In the PE group, 74 patients were diagnosed with mild PE, the others with severe PE. Clinical characteristics were compared between the normal and PE groups in Table S1. The layer distribution of pre-pregnancy body mass index (BMI) was significantly different, and PE mothers had a higher BMI than the normal ones (P=0.005). PE patients suffered a much higher rate of caesarean section (93.08% vs. 33.22%) and less gestational age (260 vs. 277 days) (both P<0.001). As a result, the birthweight of the fetuses in the PE group was inferior to the normal one (2960 vs. 3255 g, P<0.001). In terms of feeding, infant formula was more frequently used in PE group, no wonder the exclusive breastfeeding rate was lower (33.08% vs. 57.63%,P<0.001). However, other clinical characteristics were not significantly different between groups, such as BMI increase, gravidity, parity, percentage of primipara, sex of the fetus, educational background, annual family income, and scores of the Leakage Index Questionnaire and Pain Scale.
Clinical characteristics between PPD and non-PPD groups
All participants were asked to finish EPDS, and the scores were compared between the normal and PE groups. No differences for clinical characteristics were found between PPD and non-PPD mothers in the normal group (Table S2). However, in PE group, mothers who had babies with FGR and low neonatal weight tended to develop PPD (P=0.024 andP=0.007; Table 1). Postpartum pain was another high-risk factor for PPD in PE group (P=0.012). Unlike PE group, the scores of pain scale showed no difference between PPD and non-PPD women in the normal group (P=0.209).
Severe PE and FGR womenwere inclined to develop higher EPDS scores
We tried to explore the associations between PPD and PE. The average EPDS score in the normal group was significantly lower than the mild PE subgroup (7.09±4.41 vs. 8.62±4.35, P=0.008; Table S3 and Figure 1). In the severe PE subgroup, the average EPDS score was even worse (10.58±5.41), indicating most of the severe PE patients developed PPD. Furthermore, 38 PE mothers were complicated with FGR and got the highest EPDS scores among the subgroups (11.61±5.29, P<0.001).
Rather than caesarean section, PE showed direct tendency on PPD development
There was a higher caesarean section rate among PE patients than normal women (93.08% vs.33.22%). To determine the effect of caesarean section on PPD development, we compared the EPDS scores and PPD incidence between C-section and vaginal delivery in the normal group. There was not any difference in the EPDS score (6.74±4.42 vs. 7.27±4.41, P=0.337) and PPD incidence (13.27% vs.15.32%, P=0.728) between the two delivery modes (Table S4). Interestingly, when compared the normal and PE groups suffered from cesarean section, it was found that both the EPDS (9.54±4.80 vs. 6.74±4.42,P<0.001) score and PPD incidence (32.23% vs.13.27%,P=0.001) were much higher among PE group than the normal group (Table S5). It could be inferred that it was not cesarean section but PE directly increased the risk of PPD.
Much higher screening of PPD in PE mothers than the normal ones
We compared the rate of positive screening of PPD in each subgroup (Table 2). Totally 83 people were screened positive for PPD, while the remaining 342 were negative. The incidence of PPD was 14.58% in the normal group, whereas the rate was much higher among PE mothers. About 30.77% of women in the PE group met the diagnostic criterion for PPD. Furthermore, the incidences of PPD dramatically increased with the severity of PE and its complications. For instance, the incidences of PPD in the mild PE and severe PE subgroups were 27.03% and 36.96% respectively, which were significantly higher than the normal mothers (14.58%, P=0.014 & P=0.002). We also tried to explore the associations between PE complications and PPD development. In the PE+ FGR subgroup, the incidence of PPD was the highest among all the subgroups (44.74%). Thirty new-borns were extremely weak and had to be sent to the neonatal intensive care unit (NICU). Obviously, when the babies were sent to NICU, their mothers tended to develop PPD. PPD incidence among these mothers increased dramatically (36.66%), which was extremely high. Preterm, one of the common complications in PE, occurred in almost half of PE mothers (61 of 130). PPD occurrence was 32.79% in the PE + preterm subgroup.
Independent risk factors for PPD
Then multiple logistic regression was performed to evaluate the independent risk factors for PPD. With PPD as the dependent variable, PE, severe PE, FGR, NICU admission were regarded as independent variables individually, while age, BMI, gestational days, baby weight, delivery model, Leakage Index Score, milk-feeding ways and Pain Scale analysed as confounding factors. Women with mild PE demonstrated 2-fold higher odds of PPD (AOR=2.117, 95% CI: 1.001-4.479; Table 3). Furthermore, severe PE, FGR, and NICU admission all increased nearly 3-fold risk for PPD positive screening. These findings indicate that the severity and complications of PE will increase the risk of PPD (as shown on Figure S2). Besides PE, postpartum pain was another independent risk factor for PPD (AOR=1.509, 95%CI: 1.078-2.114). The effect of breastfeeding on PPD has not been clearly indicated before, but in our study, exclusive breastfeeding seemed not to positively affect the mood of the mothers (AOR=0.752, 95%CI: 0.445-1.270). Pelvic floor muscle recovery has always been a concern among new mothers and can dramatically influence their moods. After evaluating pelvic floor function, we found the dysfunction of pelvic floor muscles had no negative effect on PPD (AOR=1.137, 95%CI: 0.952-1.358). Moreover, we observed there were no correlations between the caesarean section and PPD (AOR=1.177, 95%CI: 0.620-2.232).
Discussion
Among the general population, hypertension has already been proved to be an independent risk factor for depressive disorder.24Hypertension increased 1.12-fold of developing depression among 6,237 old Chinese adults.25For pregnant women, few studies were exploring the connection between PE and PPD. To our knowledge, this is the first retrospective cohort study to clarify the associations of the severity and complications of PE with PPD in Chinese population. The number of cases (425) in our trial is the largest among the existing relevant studies.
As reported, PPD occurred in 20.5% PE patients in Tanzanian and in about 21% PE mothers in Greek.1,26In our study, the percentage of a positive screening for PPD in PE group was even higher (30.77%). Besides PE, its complications could also increase the risk of PPD. Similar to our findings, Hoedjes M. et al. discovered that the prevalence of PPD was 23% in mild PE patients and 44% in severe PE.27These studies suggest PE affects PPD strongly.
For PE mothers, besides unfavourable experience of hypertension, other conditions such as additional costs, concerns of the new-borns with complications also increase mothers’ psychological burden.28,29The outcomes of infants play an important role in PPD development among severe PE patients. 27This conclusion was confirmed in our study, especially for growth-restricted babies. A study reported the prevalence of PPD among FGR family was 48.2%, which was similar to our result (44.74%).28 38% of mothers experienced significant depressive symptoms when their babies were sent to NICU.30These studies mentioned that baby conditions and financial problems maybe two of the most risk factors of PPD. In our clinical trial, mothers were asked ‘what most upsets you?’ The majority of mothers told us that they were bothered most by the poor outcomes of their babies and NICU admission. In our study, 30 neonates were admitted to NICU. Notably, the incidence of PPD among these mothers was very high (36.6%). Many randomized controlled trials have implicated that insufficient contact with babies will increase the odds of PPD.31-33 Therefore, clinical healthcare workers should provide psychological supplies to mothers with NICU babies. The financial problem was the second problem: seven of them received lower annual income (less than 80 k RMB per year), they felt huge burden on children hospital expenses.
Whether the cesarean section will increase the risk of PPD is still controversial. In China, some healthy pregnant women would like to choose cesarean section due to social-psychological factors. In this research, mothers with PE preferred to have a cesarean section to avoid possible adverse outcomes. This can explain why the rate of operative delivery in China among PE patients is so high.
First of all, to figure out the effect of cesarean section for PPD, we compared delivery models among normal women. Patel et al. demonstrated that operative delivery would not increase the incidence of PPD in 14,633 women.34 A meta-analysis in 2017 also reported that elective cesarean section would not significantly exacerbate the odds of PPD (AOR: 1.15, 95%CI: 0.92-1.43).35 In our study, there was not any difference in the EPDS score and PPD incidence between the two delivery models among normal pregnant populations.
However, in PE group, we found that both EPDS score and PPD incidence were much higher in mothers suffered from operation. It could be inferred that PE directly increased the risk of PPD rather than cesarean section. Then we applied subgroup analysis to find the reason. In PE+ FGR subgroup, the incidence of PPD was the highest among all the subgroups. Obviously, mothers tended to showing anxiety when babies were sent to NICU. Another common complication is preterm. Almost half of PE mothers occurred preterm. As expected, mothers in the PE+preterm group experienced higher psychological distress than others. Weigl et al. pointed new mothers of preterm infants exhibited higher scores of depression, anxiety and stress than parents of term infants. Preterm mothers showed lower levels of estradiol, progesterone, and prolactin, as well as a heightened post-awakening cortisol response compared to term mothers36. These results are consistent with our finding.
Postpartum pain, urinary incontinence, and feeding methods were also evaluated in the regression model. Postpartum pain was an independent risk factor for PPD, increasing the odds by 1.5-fold. A few trials showed that untreated pain is associated with a risk of PPD.37,38The usage of painkillers can help decrease the incidence of PPD in some cases.39,40Our study implied postpartum pain as another risk factor of PPD in the PE group. This was probably because PE mothers suffered more postpartum pain from the operation. Therefore, it is reasonable to use painkillers for PE mothers during the postnatal period.
Hullfish Kl. et al has demonstrated a correlation between urinary incontinence and PPD41, but Doering AD. et al showed no such connection42. In our study, there was no significant result about urinary incontinence in PPD development. Nonetheless, more authoritative urinary incontinence scales need to be tested in the future. Non-breast-feeding was regarded as a risk factor for PPD in many cases.43,44But in our study, exclusive breastfeeding did not help decrease the incidence of PPD.
Although the connection between PE and PPD is still unclear, some mechanisms, such as clinical symptoms, inflammation, and genetic changes, have been used as hypotheses for the reason between PE and PPD. The pathogenesis for PE, a placenta disease, can be explained by the “two-stage theory”.45At the first stage, vascular remodeling disorders of uterine spiral arterioles caused by multiple factors result in “superficial placental implantation” and ultimately cause insufficient placental perfusion and impairment of placental function. In the second stage, the ischemic placenta will experience oxidative stress and release inflammatory factors, such as IL-6, leading to systemic endothelial dysfunction. Therefore, PE patients often have excessive inflammatory factors in blood circulation.46-49For example, abnormally elevated C-reactive protein (CRP) and tumour necrosis factor (TNF)-α are detected in the serum of PE mothers, resulting in vascular remodelling dysfunction of the placenta.50Consistently, like PE, inflammatory biomarkers also take part in PPD development.51 Studies confirm that increased IL-6 and TNF-α levels during the perinatal period can intensify the risk of PPD.52-54 Based on these studies, we speculate that inflammatory cytokines are released by the dysfunctional placenta in PE mothers, finally leading to the development of PPD.50,55,56Our further research will pay more attention to these inflammatory cytokines.
We must admit that there are some limitations in our study. As a retrospective study, it suffered from bias and case limitations. Firstly, patients were recruited from 2 hospitals and the local bias may be relatively reduced, but there is still a need for a study involving multiple centres. Secondly, it was hard to control operation rate in PE group, although this delivery mode was not found to be a risk factor in our regression model. In future, we would like to initiate larger randomized controlled trials and more in-depth mechanistic studies.