Prevalence of perinatal depression among Japanese women: a meta-analysis

Background Perinatal depression is one of the important mental illnesses among women. However, not enough reviews have been done, and a certain consensus has not been obtained about the prevalence of perinatal depression among Japanese women. The purpose of our study is to reveal the reliable estimates about the prevalence of perinatal depression among Japanese women. Method We searched two databases, PubMed and ICHUSHI, to identify studies published from January 1994 to December 2017 with data on the prevalence of antenatal or postnatal depression. Data were extracted from published reports. Results We reviewed 1317 abstracts, retrieved 301 articles and included 123 studies. The point prevalence of postpartum depression at 1 month was 14.3% incorporating 108,431 Japanese women. The period prevalence of depression at pregnancy was 14.0% in the second trimester and 16.3% in the third trimester. The period prevalence of postpartum depression was 15.1% within the first month, 11.6% in 1–3 months, 11.5% in 3–6 months and 11.5% in 6–12 months after birth. We also identified that compared with multiparas, primiparas was significantly associated with a higher prevalence of postpartum depression; the adjusted relative risk was 1.76. Conclusions The prevalence of postpartum depression at 1 month after childbirth was found to be 14.3% among Japanese women. During pregnancy, the prevalence of depression increases as childbirth approaches, and the prevalence of depression was found to decrease in the postpartum period over time. In addition, we found that the prevalence of postpartum depression in primiparas was higher than that in multiparas. Hence, we suggest that healthcare professionals need to pay more attention to primiparas than multiparas regarding postpartum depression.

of an equal employment policy for women in 1985, the employment rate of women has rapidly increased in Japan. However, there is insufficient social infrastructure for childcare, such as daycare, and men are not very involved in parenting. In addition, with the aging population and the increasing prevalence of nuclear families, social support in the perinatal period tends to be insufficient. In particular, the aging rate is 27.3% [8], which is the highest rate among developed countries, and support from family members, such as maternal parents, is weakening. For this reason, mental stress in women during the perinatal period is strong, and the risk of developing depression may be high. Therefore, it is problematic to apply current epidemiology data from different countries and regions to the Japanese context because of the social differences. Previous reports have suggested that perinatal depression may be affected by differences in economic status, social support, or ethnicity in the country where patients live [2,5]. For this reason, we thought it would be relevant to conduct research focused on the country and culture of Japan.
In recent years, a large, prospective nationwide cohort study (n = 82,489) called the "Japan Environment and Children's Study" (JECS) showed a 13.7% prevalence of postpartum depression among women 1 month after childbirth [9]. Although other studies on postpartum depression with various sample sizes have been carried out in Japan, with most of them written in Japanese, meta-analyses have not been conducted. For this reason, we collected articles for this study including those written in Japanese. The aim of our meta-analysis was to calculate a reliable estimate of the prevalence of postpartum depression among Japanese women. In addition, some studies have reported that the birth experience influences postpartum depression [9,10], while other studies have indicated that there is no relation between the childbirth experience and postpartum depression [11,12]. Therefore, we undertook a subanalysis of the relationship between postpartum depression and the childbirth experience.

Study selection
This systematic review was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) standards (a protocol used to evaluate systematic reviews) [13]. We searched for published studies related to perinatal depression in the PubMed electronic database. The  In addition, the ICHUSHI database (http://searc h.jamas .or.jp/) was searched for articles written in Japanese. ICHUSHI contains bibliographic citations and abstracts from biomedical journals and other serial publications published in Japan. We used comparable Japanese search terms without the terms "Japan" and "Japanese" to search ICHUSHI.
The two electronic databases, PubMed and ICHUSHI, were searched for studies published from January 1, 1994, to December 31, 2017. We excluded older literature before the release of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) [14]. Then, we examined the list of references included in the articles.

Inclusion and exclusion criteria
Studies were eligible for inclusion if they (a) included women who were 16 years or older; (b) assessed prenatal or postpartum depression using a validated self-report instrument; (c) reported the results of peer-reviewed research based on cross-sectional or prospective studies; and (d) reported data to estimate the prevalence of prenatal or postpartum depression using the EPDS or Center for Epidemiologic Studies Depression Scale (CES-D). Studies were excluded if they (a) recruited only highrisk women; (b) reported results for only a subsample of a study population; (c) reported duplicate data from a single database; (d) reported only mean data; (e) did not report a cutoff point for depression; or (f ) had < 100 participants (these studies were excluded to avoid a smallstudy effect) [15]. For studies with duplicate data from a single database, we selected the study with the larger sample size. Case reports, comments, editorials, letters, and studies not performed on human participants were also excluded. Two researchers (KT and NS) independently searched the literature. After all papers had been assessed, any discrepancies in the responses were identified and discussed to reach a consensus on the best option. Disagreements about the inclusion of a study were resolved through discussion with the senior author (NYF). Data were extracted from each article using a standardized form including the first author, publication year and other information.

Data extraction
From each study, we extracted information about the publication year, sample size, measures used to assess depression, cutoff point used for each measure, time points for depression assessment, and percent of the prevalence of prenatal or postpartum depression.
Publication year, parity, and perinatal depression prevalence were used as continuous variables.
The time of measurement was defined as the first trimester (i.e., 0 to 3 months gestation; Time 1 [T1]), second trimester (i.e., > 3 to 6 months gestation; Time 2 [T2]), third trimester (i.e., > 6 months gestation to childbirth; Moreover, for intervention studies, only the baseline data were extracted. For longitudinal studies, only data on the rate of depression from one time point in each period (e.g., prenatal and postpartum) were included in the analyses. For most studies, the first time point was used, as the participants were least familiar with the study tool at that point and were unlikely to exhibit priming effects.
We collected papers that evaluated postpartum depression using the Japanese versions of the EPDS and CES-D.
The EPDS is a self-report instrument measuring postnatal depression with 10 items rated on a 4-point scale (from 0 to 3). The total score ranges from 0 to 30; the higher the score, the worse the symptoms of depression are. The reliability and validity of the Japanese version of the EPDS were reported by Okano, and a cutoff point above 9 was established [16]. Our meta-analysis also included a paper that evaluated depression by using the Japanese version [17] of the CES-D [18]. This tool consists of 20 questions about depression, and the total score ranges from 0 to 60 points. We collected papers that defined the presence of depression based on a CES-D score ≥ 16.

Statistical analysis
First, we assessed the pooled prevalence of postpartum depression at the time of the checkup 1 month after childbirth. Then, we assessed the pooled prevalence of perinatal or postpartum depression during each period (T1 to T7). Third, we conducted a trend analysis applied the generalized linear mixed model [19]. The t tests on the contrast vectors for regression coefficients of the time variable were conducted in order to evaluate the difference between time points in the prenatal period, and the trend of proportion in the post period. Finally, we calculated the relative risk to investigate the differences in the prevalence of postpartum depression between primiparas and multiparas.
We used the I 2 statistic and its 95% CI to estimate heterogeneity. The I 2 statistic was considered high when it was 75% or higher [20]. The significance level was set at p < 0.05. The meta-analysis and related statistical analysis were performed with meta-package version 4.9-1 in R version 3.5.0., and the GLIMMIX procedure in SAS ver. 9.4.

Search results and included participants
After excluding duplicate or irrelevant papers, we found 123 publications that met the inclusion criteria (Fig. 1). The final sample included 108,431 people assessed at the time of the checkup 1 month after childbirth. The sample sizes of the studies ranged from 100 to 82,489 people. More details on the included studies and participants are presented Tables 1 and 2.

Prevalence of perinatal depression and subgroup analysis
The point prevalence of postpartum depression at 1 month after childbirth was calculated by integrating the 108,431 people from 63 publications and was found to be 14.3%. (95% CI 13.2-15.4%). The level of heterogeneity was I 2 = 88.3%. Because of the high heterogeneity, the prevalence was calculated by a random-effects model (Fig. 2).
A visual inspection of the funnel plot at 1 month after childbirth revealed symmetry (Fig. 3), and Egger's regression test for funnel plot asymmetry was statistically nonsignificant (t = 0.5958, p = 0.5535).
Next, a subanalysis of the effect of the childbirth experience on postpartum depression was performed. The data for a total of 102,006 people described in 39 papers were integrated, and a meta-analysis was performed at the relative risk level. The result showed that primiparas had a significantly higher prevalence of postpartum depression than multiparas, with a relative risk of 1.76 (95% CI 1.59-1.96). The level of heterogeneity was I 2 = 52.2%; the meta-analysis of relative risk was performed using a random-effects model (Fig. 5).

Trend analysis of the prevalence of perinatal depression
We performed a trend analysis by applying the generalized linear mixed model where outcome was existence of depression; link function was logit function; fixed effects were time (6 time points from T2 to T7; nominal variable) and scale (CES-D or EPDS); random effect was trial. As a result of the F test, there was no statistically significant difference between CES-D and EPDS in the prevalence of perinatal depression (F = 0.46, p = 0.501). The t tests on the contrast vectors for regression coefficients of the time variable were conducted in order to evaluate the difference between two time points in the prenatal period, and the trend of proportion in the postpartum period. The contrast vector for prenatal period was set as (− 1, 1, 0, 0, 0, 0), and postpartum period was set as (0, 0, 3, 1, − 1, − 3). The contrast vector for prenatal-postpartum comparison was set as (2, 2, − 1, − 1, − 1, − 1). As a result of trend analysis, a prevalence of prenatal depression increased statistically significantly over time (t = 3.78, p = 0.001), and a prevalence of postpartum depression decreased statistically significantly over time (t = 6.00, p < 0.001).
Comparing the prevalence of prenatal and postpartum depression, prevalence of prenatal depression was statistically significantly higher than that of postpartum depression (t = 4.11, p < 0.001).

Sensitivity analysis of the prevalence of perinatal depression
Additionally, a sensitivity analysis was performed to examine the robustness of the data. In particular, the analysis focused on heterogeneity. We found that when the study with the largest sample size (n = 82,489), i.e., the JECS [9], was excluded, the prevalence of depression was 14.1% at 1 month postpartum (95% CI 12.8-15.5%, I 2 = 88.1%, n = 25,942). There was no statistically significant difference in the prevalence of depression with or without the JECS data, and the heterogeneity was the same with or without JECS data.
The EPDS is the most frequently used measure to evaluate perinatal depression in women worldwide [21], so we examined the prevalence of perinatal depression only with statistical data from the EPDS. The prevalence of perinatal depression after the sensitivity analysis is presented below.

Discussion
Our study is the first to use a meta-analysis to investigate the reliable prevalence of perinatal depression among Japanese women. The most important finding is that the point prevalence of postpartum depression was 14.3% 1 month after childbirth. The JECS [9] is a large-scale study compared with other studies, so we tried to reanalyze the data with the JECS data excluded. The prevalence of postpartum depression and heterogeneity 1 month after childbirth were almost the same with or without the JECS data. While the JECS already identified the reliable prevalence of postpartum depression, our research confirms the extent of the heterogeneity in postpartum depression among Japanese women.
According to the DSM-IV-TR [22], maternity blues are defined as depressive episodes that develop by the fifth day after childbirth and then disappear within 2 weeks. It is recommended that maternity blues and postpartum depression be clearly distinguished [22]. Thus, it might be important to establish a sampling time to investigate the condition of postpartum depression 1 month after childbirth to exclude the possibility of maternity blues.
In Japan, the rate of infant health checkups 1 month after childbirth is high at 83.6% [23], and infants' mothers are also checked for health problems at that time. Since Okano created the Japanese version of the EPDS [16], this screening tool has been used for the early detection of a high risk of depression in mothers. Epidemiological studies of perinatal depression are mainly conducted by public health nurses and midwives in Japan. Although they often report research results in Japanese, sampling bias is less likely in these studies.
In addition, every year, approximately 100 women commit suicide in Japan because of worry about childcare, and the number has remained high [24]. Recently, Takeda [25]. These women were suffering from mental illnesses, such as depression, and this figure was more than double the maternal mortality rate due to obstetric abnormalities. Therefore, it is important to estimate the prevalence of postpartum depression in Japan. In addition, postpartum depression may lead to child abuse [26]. Therefore, to protect the health of children, more substantial measures against perinatal depression are needed. Furthermore, the prevalence of postpartum depression in primiparas is higher than that in multiparas. This is a fundamentally important finding that has major implications for the national health care plan in Japan. There may be several reasons for this result. First, multiparas  114  7  275  59  31  17  33  62  32  131  222  68  31  29  16  15  40  33  66  33  91  18  20  40  17  7  66  4  30  17  43  37  50  57  8  39  24  79  19  43  36  89  233  7  49  24  18  41  127  10  38  19  19  169  36  23  8  25  11  14  191  437  11341   Total   108431   627  161  1864  402  664  356  204  252  189  691  1869  820  230  152  111  103  503  151  361  149  986  143  160  443  103  147  276  109  590  113  279  248  178  299  135  152  179  405  201  271  143  595  1744  117  314  121  101  142  690  100  179  215  244  1327  114  257  113  110  118 Fig. 2 Prevalence of postpartum depression 1 month after childbirth are expected to have some experience adapting to the stress of childbirth and childcare through the pregnancy experience. Second, a woman with a history of postpartum depression is known to have a high risk of depression during the birth of her second child [27]. For this reason, a high-risk multipara has already received psychological education for perinatal depression and may take preventive measures. Third, if a woman suffered from perinatal depression in her first childbirth and did not receive adequate care, her motivation to give birth to a second child may be reduced. Further research is needed to provide details on the relationship between postpartum depression and family planning.
According to the DSM-5 [28], 50% of cases of postpartum depression are known to have developed during pregnancy. Therefore, mood disorders not only postpartum, but also during pregnancy have also been attracting attention. Interestingly, the prevalence of depression increases as childbirth approaches during pregnancy and the prevalence decreases over time in the postpartum period. In particular, the prevalence of depression was the highest in the third trimester of pregnancy; however, a previous report suggested using different cutoff values for the EPDS for the periods before and after pregnancy [29]. A similar trend has been observed in the United States, and large-scale cohort studies have reported that the prevalence of perinatal depression reaches its peak just before childbirth [30]. During pregnancy, the prevalence of depression increases as childbirth approaches.
Sleep disorders, such as restless leg syndrome and frequent awakening at night, are known to occur most often in the third trimester of pregnancy [31,32]. On the other hand, sleep quality improves over time after childbirth [33]. In addition, urinary incontinence may also raise the risk of perinatal depression [34]. During pregnancy, frequent urination is common [35], and the degree of urinary incontinence is reported to increase as childbirth approaches [36]. The worsening of frequent urination may affect the prevalence of depression during pregnancy. These studies attributed the increase in prevalence to organic problems of an epidemiological nature, but it is not possible to claim direct causal links between depression and biological factors.
The cessation of the use of antidepressants during pregnancy may also affect the increase in maternal depression prevalence. Pearlstein reported that although antidepressants are the most common treatment for postpartum depression, women tend to prefer psychotherapy [37]. Certainly, there is strong evidence for the effectiveness of structured psychotherapy, such as cognitive behavioral therapy (CBT) [38], interpersonal psychotherapy (IPT) [39] and psychological education [30,40], for treating and preventing perinatal depression. Therefore, psychotherapy should be considered the first choice, depending on the patient's condition. However, the need for drug therapy should also be considered. Suzuki reported that depression declined in Japanese women who had been treated for depression after they had stopped antidepressants after pregnancy [41]. The JECS also showed that Japanese women tend to refrain from taking drugs when pregnant. These women again increase their rate of medication after birth [42]. Interestingly, the incidence of postpartum depression is reported to be very low among women with no history of mental illness [27]. In other words, patients with postpartum depression may have had a predisposition for depression before onset. It was also reported that women who discontinued antidepressant medication experienced a relapse of major depression during pregnancy significantly more frequently than women who maintained their medication (hazard ratio, 5.0; 95% confidence interval, 2.8-9.1; p < 0.001) [43]. Therefore, drug treatment strategies should be carefully

Conclusion
Our meta-analysis provided reliable estimates of the prevalence of perinatal depression among Japanese women. The point prevalence of postpartum depression 1 month after childbirth was found to be 14.3%, and the data had high heterogeneity. Our results indicated that during pregnancy, the prevalence of depression increased as childbirth approached, and the prevalence decreases over time in the postpartum period. In addition, we found that the prevalence of postpartum depression in primiparas was higher than that in multiparas. Hence, we suggest that healthcare professionals need to pay more attention to primiparas than multiparas regarding postpartum depression.