Background
For many women, the period surrounding childbirth is accompanied by distress. Indeed, the prevalence of postpartum maternal distress symptomatology ranges from 8 to 40% for depression [
1‐
4] and 13–40% for anxiety [
5,
6]. In turn, these types of distress have been related to problems in children’s emotional, behavioural, and cognitive development (e.g. [
7‐
9]). Preventing maternal distress will thus enhance both maternal and child well-being and health. The aim of the current review was to systematically review the evidence on the effectiveness of preventive interventions on distress offered to pregnant women.
The focus in most prevention studies of postpartum distress has been on
indicated (or secondary) and
selective (or primary) prevention. Indicated prevention means that an intervention is focused on pregnant women who already display symptoms of a psychological disorder without fulfilling the criteria for a full-blown disorder (e.g. [
10,
11]). Selective prevention is aimed at pregnant women at risk for developing a disorder, for example women with a history of psychopathology, pregnancy complications, adverse life events, or low social support (e.g. [
12‐
16]). Previous reviews and meta-analyses have suggested that both indicated [
17,
18], as well as selective prevention [
19‐
21] during the perinatal period are effective for the prevention of depression symptomatology. Even though research indicated that anxiety disorders might be more prevalent than depressive disorders during the perinatal period [
22], much less is known about the effects of indicated and selective prevention on other disorders or symptomatology beyond depression, such as anxiety and stress. Recent reviews indicated that the few studies that have been done were either not effective [
23], or that the number of available studies was too low to be able to assess their effectiveness properly [
21].
In contrast to indicated and selective prevention,
universal prevention is aimed at all women regardless of their risk status or symptoms. Given the relatively high level of maternal distress symptomatology after birth that prenatal symptoms likely continue in postpartum symptoms [
9], and that postpartum distress might affect sensitive parenting important for a whole range of child outcomes [
24‐
27], a preventive approach aimed at all pregnant women might be valuable, for both mother and child. Moreover, it is important to intervene as early as possible, preferably before birth, since parental distress symptomatology can impact child development from birth onwards [
28,
29]. However, little is known about the effectiveness of universal prevention of symptoms of depression, anxiety and stress during pregnancy [
21,
23].
Therefore, the aim of the current study was to systematically review and meta-analyze the available evidence on the effectiveness of preventive interventions on symptoms of depression, anxiety, and stress offered to
universal populations of pregnant women compared to routine care. Previous meta-analyses included, but did not systematically investigate and differentiate, universal preventive psychological interventions [
20,
21,
30,
31]. Moreover, this review will be the first to also include partner and infant outcomes. The prevalence of fathers’ symptomatology is estimated to be about 10% for mild to moderate depression [
32,
33] and/or anxiety disorders [
34]. As both maternal and paternal distress symptoms can impact infant development [
9,
35‐
37], it is important to investigate whether effects of universally applied psychological interventions extend from the parents to the infant.
Method
Protocol and registration
A comprehensive literature search was performed in the bibliographic databases PubMed; Embase; Ebsco/PsycINFO; Ebsco/CINAHL; and Wiley/Cochrane Library in collaboration with a medical librarian. Databases were searched from inception up to 15 November 2018. The following terms were used (including synonyms and closely related words) as index terms or free-text words: “Parents”, “Pregnancy”, “Prevention”, “Education”, “Cognitive therapy”, “Stress”, “Anxiety”, “Depression”, “Well-being”, “RCTs”. The search was performed without date, language or publication status restriction. Duplicate articles were excluded. The full search strategies for all databases and the number of identified items per database can be found in Additional File
1.
Eligibility criteria
The following eligibility criteria were applied during the data collection process: (a) randomized controlled trials; (b) testing psychological interventions for pregnant women (with or without inclusion of their partner); (c) starting prenatally; (d) aimed at preventing maternal depression, anxiety and/or stress (e) comparing the active condition with care-as-usual, placebo or waitlist and (f) published in English in (g) international peer-reviewed journals. Care-as-usual could consist of regular consults with professionals in (prenatal) health care, such as midwives, general practitioners, or obstetric nurses. These consults are typically focused on monitoring the health of the mother and the fetus and on providing information about pregnancy and the delivery. Except for the psychological character of the interventions, there were no specific criteria for eligibility. Examples of (elements of) interventions that could be included are: psychoeducation, relaxation techniques, mindfulness, and social support. The interventions could be implemented through education materials (booklets, websites, or videos), individual meetings, group meetings, home visits, or combinations of these. Trials were excluded if they were aimed at indicated prevention (pregnant women with pre-existing psychopathology following DSM-IV or scoring above cut-off on validated clinical measures such as the Edinburgh Postnatal Depression Scale (EPDS, [
38]) or at selective prevention (aimed at pregnant women with a high risk to develop psychopathology such as low-income pregnant women, teenage pregnancies, or HIV positive pregnant women). Furthermore, studies reporting insufficient outcome data to calculate effect sizes were excluded (such as the non-reporting of standard deviations, or the reporting of plotted data only).
Data collection process
After our literature search, we removed duplicates. Two independent assessors (MM and TD) examined the titles and abstracts. The full-text of all remaining potentially eligible papers was retrieved after which the selection of studies, based on the above eligibility criteria, was done by two researchers (MM for all studies and MC or CM for half of the studies). Differences between the two raters were solved by discussion. In case of disagreement, the paper was discussed with the other members of the review team (AvS and/or TD) until consensus was reached. For data extraction, a piloted standardized form was used. This form included the following categories: study characteristics, risk of bias assessment, and data to calculate effect sizes. Study characteristics that were coded are: 1) year of publication; (2) country (high/low income); 3) participant characteristics (N, age, SES), 4) inclusion of the partner (yes/no); 5) type of intervention (psychoeducation, cognitive-behavioural therapy (CBT), mindfulness, or another intervention); 6) timing of the intervention (prenatal or a mix of prenatal and postnatal implementation); 7) delivery method of the intervention (individual, group, or mixed format); 8) materials used (e.g. booklet or video); 9) number of sessions; 10) training and supervision of the providers of the intervention (type and frequency of training); 11) method of recruitment (ads, hospital, midwives, other); 12) type of control group and/ or characteristics of the alternative treatment (wait-list, care-as-usual, alternative intervention); 13) type of randomization and number of arms; and 14) primary and secondary outcomes of the study.
Risk of bias in individual studies
Risk of bias was assessed with The Cochrane Risk of Bias Assessment Tool [
39]. This tool consists of the following criteria: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data and selective reporting. Again, two researchers (MM, and MC or CM) independently assessed risk of bias for each study. Discrepancies in ratings between the two researchers were resolved by discussion, led by a third researcher (AvS or TD).
Statistical analysis
We performed a random-effects meta-analysis, using the ‘Comprehensive Meta-analysis’ software package for Windows (CMA; version 3; available from
www.metaanalysis.com). To calculate the pooled effect size of the intervention, we used the post-test measures of different measures of distress and expressed them in Cohen’s
d [
40]. This value refers to the number of standard deviations the intervention group scores better (or worse) than the control group. An effect size of 0.20 can be considered as small, of 0.50 as moderate, and 0.80 as large [
40]. For studies using different instruments measuring the same outcome (i.e. two different depression scales), the outcomes were combined in one effect size per outcome (the mean of the two separate effect sizes). When multiple interventions were compared with a non-treated control group [
41,
42], the effect of the intervention was compared to both the active intervention as well as to the control condition. Thus, in this case, we included both comparisons (intervention A – vs. control group and intervention B vs. control group) in our analysis.
First, we calculated the pooled effect size for studies measuring maternal distress, thereby combining depression, anxiety, stress, and/or parenting stress. We checked for outliers (defined as a case in which the 95% confidence interval of an individual study did not overlap with the 95% confidence interval of the overall pooled effect size). After removal of two outliers, we repeated the main analysis. We then repeated the analysis (without the two outliers) on the measures of distress separately, namely depression, anxiety, and (general or parenting) stress.
Statistical heterogeneity was assessed with the I2-statistic (fixed effects model), which refers to the variance between studies as a proportion of the total variance. High percentages indicate substantial heterogeneity. Numbers-needed-to-be-treated (NNT) were calculated from the effect sizes. Publication bias was examined by a visual inspection of the funnel plot and by Egger’s test of the intercept. An estimation of the effect size while taking publication bias into account was performed by means of the Duval and Tweedie trim and fill procedure.
Sub-group analyses on the combined outcome of depression, anxiety and stress, were performed for the following variables: timing of the intervention (prenatal only or a combination of prenatal and postpartum elements); intervention type (psychoeducation; CBT; mindfulness or other interventions); intervention delivery mode (delivered in a group or on an individual basis); whether the partner was included in the intervention; timing of post-test (during pregnancy or in the first 6 months after birth); and methodological quality (based on the risk of bias assessment performed through the Cochrane tool). We used the mixed effect model, which pools studies within subgroups with the random effects model but tests for significant differences between subgroups with the fixed effect model.
Discussion
This meta-analysis focused on the effectiveness of preventive psychological interventions offered to universal populations of pregnant women on symptoms of depression, anxiety, and general stress. Paternal and infant outcomes were also included. The meta-analysis suggested that psychological interventions among pregnant women without a specific risk of psychopathology and implemented during pregnancy, are effective in the prevention of maternal distress symptomatology. The meta-analysis showed that these interventions have a moderate effect on the combined measure of distress (d = .52) as well as on depressive symptoms (d = .50), and stress (d = .52). The effect on anxiety (d = .30) was somewhat smaller. These results indicate that, next to indicated and selected prevention, universal prevention has value in its own right. Since the results with regard to anxiety and stress are based on a considerably lower number of studies, the effectiveness of universal prevention on the prevention of distress beyond depression should be interpreted with caution.
Two studies were outliers and thus excluded [
48,
50]. These studies differed from the other studies in several respects. The Haga et al. [
48] study was the only study in which a multimodal intervention was tested, consisting of elements of meta-cognitive therapy, mindfulness, acceptance and commitment therapy, positive psychology, cognitive-behavioral therapy, and psychoeducation. The online intervention consisted of 44 sessions. While these sessions took not much time to complete (about 10 min), the number of sessions was much higher than in the other studies. It is possible that the high number of sessions, but also the multimodal nature of the intervention, explain why this study found a much larger effect size than the other studies. It might be that by offering multiple elements of different interventions and therapies, women can choose those elements that work best for them to alleviate distress symptoms. The intervention in the Khorsandi et al. [
50] study differed from the other interventions as it was exclusively focused on stress, namely psychoeducation about stress and how to handle signals of stress. The content seemed to be not exclusively geared to pregnant women, while this was the case for the other included interventions. Moreover, it was difficult to assess the methodological quality of this study. For example, no flow chart and no info on timing of measurements was reported. Also, lack of active and regular participation in the training stages of the intervention was described as an exclusion criterion. It is not clear if this criterion actually resulted in exclusion of participants, but if so, the sample could be biased towards more highly motivated women (potentially leading to a higher effect size). It is important to emphasize that exclusion of these two studies lead to more conservative effect estimates. As a result, the true effects might even be higher than the ones we reported.
The studies we included were aimed at universal prevention. The majority of the included studies (n = 9) excluded women with a diagnosis of anxiety or depression, or women who scored above a cut-off on questionnaires. The remaining three studies included all pregnant women (and their partners) regardless of pre-existing symptomatology or risk status, but did not exclude women with a diagnosis or (severe) symptoms. The overall mean baseline scores of all studies showed that the women had some symptoms, but that the depression, anxiety, and stress scores were well below clinical cut-off. While indicated and selected prevention efforts are exclusively aimed at women who are screened on their (considerable) level of distress symptomatology (using validated cut-off scores), or their relative risk on developing psychopathology based on the presence of one or more known risk factors (e.g. pregnancy complications, low social support), universal prevention targets those women who have no known risk factors and experience low to moderate levels of distress. The current analysis suggested a considerable effect on symptoms of depression, anxiety, or general stress for these women, indicating the added value of universal prevention of distress in pregnancy.
The included preventive interventions varied from (online) mindfulness-based self-help interventions to interventions consisting of multiple group sessions based on principles of established therapeutic techniques, such as cognitive-behavioural therapy (problem solving and communication skills) and interpersonal psychotherapy (underlining the importance of social relationships). Most interventions were exclusively aimed at pregnant women and included psychoeducation about postnatal distress, relaxation techniques and the acquisition of emotion regulation skills. Also, most interventions were offered in a group setting (in a local hospital) and facilitated by a mental health professional or a midwife. A minority of interventions was offered in a (internet-based) self-help format. A subset of interventions also offered postnatal sessions: these interventions were also aimed at the couple relationship and/or the transition to parenthood. We could not demonstrate that one type of intervention was more effective than the other types. However, due to the high heterogeneity among the interventions and the small number of studies per type of intervention, the analyses might have lacked sufficient power to demonstrate differences in effect.
When considering the three indicators of distress separately, we found considerable effect sizes for depression (d = .50), stress (d = .52), and a somewhat lower effect size for anxiety (d = .30). This implies that the interventions were effective on all three indicators of distress. Moreover, analyzing the indicators of distress separately resulted in less heterogeneity. It is important to keep in mind that most of the included studies focused on symptoms of depression (n = 10), and that the results for symptoms of stress and anxiety were based on a considerably lower number of studies (n = 5).
The impact of universal prevention on symptoms of
depression is in line with conclusions of earlier reviews and meta-analyses showing the effectiveness of selected and indicated prevention for depression and depressive symptomatology [
17,
19‐
21]. For example, one of the studies [
51] in our meta-analysis considered the incidence of a depressive disorder by 6 weeks postpartum as an outcome. While 10 women from the control group were diagnosed with depression (9.3%), only three women (2.7%) were diagnosed in the intervention group. Other studies (
n = 4) reported the percentage of women that scored above cut-off scores for depression, using the Edinburgh Postnatal Depression Scale (EPDS [
38];). The rates varied between 11.8 and 40.6% for the control groups, and, in contrast, between 8.7 to 17.4% for the intervention groups. However, different cut-off scores were used (ranging from 10 to 14), which makes the percentages difficult to compare. To be able to detect whether universal prevention leads to less cases of a depressive disorder (and thus to genuinely assess the effect of universal prevention during pregnancy on the development of psychopathology), future studies are strongly encouraged to report the incidence rate of depression and other mental disorders as an outcome. When cut-off scores are used for this aim, it is important to use comparable cut-off score across studies.
While earlier reviews were not able to quantify the effect of universal prevention on symptoms of
anxiety [
21,
23], results of this meta-analysis indicate that, next to depression, universal prevention has a moderately preventing effect on symptoms of anxiety. As an accumulating number of studies indicated that women can experience considerable levels of anxiety symptomatology after childbirth [
5,
8], even resulting in an anxiety disorder [
13], it is an important finding that universal prevention apparently works to alleviate anxiety symptoms. However, the number of interventions focusing on the prevention of anxiety was rather low and the effect size seemed to be smaller (
d = 0.30) than for depression (
d = 0.50) and stress (
d = 0.52). Therefore, in line with earlier meta-analyses [21, 23), we hope that future prevention trials will include anxiety as a target of intervention.
There are two other reviews, which examined effects on general
stress [31, 32]. These reviews did not indicate an effect of antenatal universal prevention. However, these reviews included a mixture of universal, selective and indicated prevention, possibly explaining the different results. Also, there were differences in the nature of the stress measures included. In Fontein-Kuipers et al. [
30] distress was broadly measured and included symptoms of depression and anxiety next to perception of stress, parenting stress, and parental worry. It is possible that different types of stress need different types of intervention, and that a potential effect of interventions on this broad index of stress would thus be more difficult to detect.
This meta-analysis showed that a minority of interventions focused on both partners. Only in three of the 12 included interventions, the partner was also involved. These interventions focused mainly on the functioning of the couple relationship during the transition to parenthood. Our meta-analysis showed that these interventions were less effective in preventing distress, than interventions that only included the mother. It might be that mothers with a partner willing to participate in a preventive intervention experience higher levels of social support at baseline, and therefore lower levels of distress. The intervention might thus be less effective for this group of women. However, given that both the content of the intervention (focus on the couple relationship), and the target group (couples) varied, no firm conclusion about the effectiveness of including the partner in preventive interventions can be drawn yet. Furthermore, preventing distress in partners might request a different approach, and it is thus worthwhile to also investigate interventions exclusively geared on the partner. Given the paucity of trials that focused on distress of the partner, and abundant research indicating that fathers also experience considerable postpartum distress [
32,
34] which might also affect the child [
36,
37], future trials should focus on the prevention of both maternal and paternal distress.
Likewise, we were not able to measure the effectiveness of the interventions on infant outcomes, as only three of the 12 included studies assessed (a variety of) infant outcomes. Beattie et al. [
43] reported no effect on various birth outcomes of their mindfulness intervention. In Feinberg and Kan [
49], parents (especially fathers) participating in a cognitive-behavioural based psychosocial prevention program reported lower levels of dysfunctional interaction and distress in the relationship with their child around 6 months postpartum. Also, infants from the intervention group showed a longer duration of orienting and greater soothability. Abkarzadeh et al. [
41] reported that in both their intervention groups (psychoeducational attachment and relaxation), the number of counted fetal movements increased compared to the control group.
Given the well-established impact of parental distress on children’s well-being and development [
9,
35,
52], future trials are encouraged to investigate whether the positive effects of the universally applied psychological interventions extend from mother to infant. Since parenting quality is a factor that can be modified by intervention [
53], focusing on the inclusion of quality-related outcomes, such as soothability and parent-infant interaction [
49] could be a promising pathway. For example, including observational measures of parental sensitivity for and responsiveness to stress signals of the infant could be included.
Limitations
The current study has several limitations. First, most of the included studies focused on depressive symptomatology as an outcome. Therefore, we were unable to draw firm conclusions regarding the other indicators of distress, namely symptoms of anxiety and general stress. Second, because none of the included studies focused on child outcomes, no conclusions about the effectiveness of the interventions on infant well-being could be drawn. Third, only a limited number of studies included the partner, which means that the effectiveness of interventions during pregnancy on preventing distress of the partner could not be analyzed. Fourth, the risk of bias assessment indicated that a large part of studies was not sufficiently transparent in reporting all information necessary to give a quality judgement based on the Cochrane Risk of Bias Assessment tool. This was mainly a problem when judging the random sequence generation and the allocation procedure, in which respectively two-third and half of the studies did not report how they handled this. Also, judgement of the incomplete outcome data criterion revealed that almost half of the studies had to deal with relatively high drop-out rates and/or did not specify the reasons for drop-out adequately. However, subgroup analyses showed no association between overall methodological quality and the size of the effect. Fifth, to be able to detect whether universal prevention would make a difference in preventing distress (i.e. if universal prevention is worthwhile from a cost-effectiveness perspective) we compared the effect of universal prevention to routine care. While routine care can be provided by midwives, there were differences between studies as to which type of routine care women have access to during pregnancy. Also, not all studies provided sufficient details about what constituted regular care in their study. This means that the regular care condition might have varied between studies. To be able to detect if additional support during pregnancy could contribute to stress reduction among pregnant women compared to different types of routine care, future trials are recommended to provide details about regular care in their specific study setting.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.