Background

Depression during the perinatal period is common, occurring in 10–15 % of pregnancies [1, 2], with the prevalence of depression peaking at 3 months postpartum. The adverse effects of perinatal depression on maternal and child health are profound [3–6] with estimated costs to the UK of £8.1 billion per year [7]. The Saving Mothers' Lives Report (2011) identified 29 women who died as a result of suicide between 2006 and 2008 [8]. Of these 29 women, 6 of them (21 %) had an underlying diagnosis of severe depressive illness. Despite this, few women seek professional advice due to physical and attitudinal barriers, long-waiting times for face-to-face psychotherapies, social stigma [9–13] and uncertainties about the potential risks of pharmacological therapies during pregnancy [14–16]. There is therefore an urgent need for a validated, effective, easily accessible, low-cost, non-pharmacological psychosocial evidence-based intervention to prevent or treat perinatal depression.

Computerised Cognitive Behavioural Therapy (CCBT) has proved an acceptable and effective treatment for depression in non-pregnant individuals [17–26] with a comprehensive meta-analysis demonstrating significant superiority compared to comparator therapies in the general population [18]. Based on this evidence, CCBT is one of the low-intensity psychosocial interventions that the National Institute for Health and Care Excellence (NICE) (which provides national guidelines for delivery of healthcare in England and Wales in the UK) recommends for treating persistent subthreshold depressive symptoms or mild to moderate depression in non-pregnant individuals [27].

NICE also suggests that CCBT should be offered to treat perinatal depression [28]. This is despite the scarcity of psychosocial interventions specifically designed to treat perinatal depression (particularly that originating antenatally) and the changing and specific psychosocial needs of women during pregnancy and post-partum [29, 30]. Further, women are increasingly using the Internet as a source of health information during pregnancy. For example, in a study of 613 women, almost 50 % used the Internet to enhance their involvement in pregnancy-related decision-making [31] and 90 % felt that health professionals should actively suggest suitable Internet sites to guide their further reading during pregnancy [31]. A CCBT resource for the self-management of perinatal depression could therefore represent an efficacious, accessible and economically sound resource [32]. CCBT also provides a level of anonymity not possible in clinical care, can be accessed at a time and place suitable for the participant and can be adapted to different cultures, languages and literacy levels to ensure the inclusion of vulnerable groups [33]. However, before CCBT is embedded in clinical practice for treatment of perinatal depression, there is a need for a systematic review of the existing evidence to determine whether web-interventions for pregnant and post-partum women are efficacious for prevention and treatment of perinatal mood disorders and potentially to inform the design of future studies if gaps in the current evidence base are identified.

Methods

Data sources

The Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines were followed for the conduct [34], and the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines for the reporting [35] of this systematic review. The literature search was conducted using MEDLINE, PsycINFO, Embase and Cumulative Index to Nursing and Allied Health Literature (CINAHL) from inception until March 26th 2015. The searches were limited to human studies and used terms as both keywords and indexing terms (Medical Subject Headings, MeSH) to identify studies related to depression AND pregnancy AND therapy AND online: ‘depression’, ‘depressive’, ‘depressed’, ‘mental-health’, ‘well-being’, ‘post-natal depression’, ‘pregnancy’, ‘perinatal care’, ‘pre-natal/ante-natal care’, ‘post-partum/post-natal care’, ‘pregnancy complication’, ‘self-help’, ‘cognitive behavioural therapy’, ‘behavioural therapy’, ‘psychotherapy’, ‘internet’, ‘online’, ‘online therapy’, ‘computer program’, ‘mobile application’, ‘E-health’ and ‘telemedicine’ (see Additional file 1 for full search strategy). Finally forward and backward citation tracking was conducted through Web of Science for all articles identified for full article review and one heavily cited review in this area [26]. One identified record was a ‘letter to the editor’ [36]. The corresponding author was contacted and confirmed that the data within the letter were also included in a further publication identified in the literature search [37]. In addition, the corresponding author of three articles written by the same author [38–40] was contacted who confirmed that all three publications included data collected from one single study. The data from only the largest study [40] were included in the final data synthesis to avoid data duplication. No supplementary data or further studies were identified by author contact.

Study selection

Two researchers (EL, KH) screened articles for eligibility. Any discrepancies that could not be resolved were discussed with the other researchers (FD, RR). Studies were included if they a) administered an internet/web/online/computer intervention during the perinatal period, defined as the start of pregnancy to 1 year post-partum, b) had maternal mental health as an outcome measure and c) an assessment of mental health pre-and post-intervention. There were no exclusions related to study design or participants’ age, ethnicity, socioeconomic status, parity, depressive status or depression therapy.

Data extraction

Relevant information on study characteristics and methodology (Table 1) and study results (Table 2) was extracted using pre-specified and standardised data extraction forms based on the template devised by the Cochrane Consumers and Communication Review Group [41]. One researcher (EL) completed the data extraction tables and a second researcher (KH) independently checked content. Any disagreements were discussed and resolved by consensus.

Table 1 Data extraction: Study design and details of intervention
Full size table
Table 2 Data extraction: Outcomes
Full size table

Quality assessment

Two researchers (EL, KH) independently assessed the risk of bias for each study as an assessment of study quality, using a modified tool [42] with objective criteria relating to sample population and recruitment, intervention, randomisation, blinding, comparator group, use of validated outcome measure, follow-up and data analysis. A paper could attain a maximum score of 8, a score of 1–3 indicating poor quality, 4–6 intermediate quality and 7–8 good quality.

Data analysis and synthesis

Due to considerable study heterogeneity a descriptive synthesis of results was conducted. This included a description of the study design and sample studied, details of the intervention including mode and duration of delivery, and the outcomes including change in mood scores post intervention and at later follow-up as well as attrition rates.

Results

Study design and participants

From 547 abstracts, 39 full text articles were assessed (Fig. 1) and four studies were included in the final data synthesis. The four eligible studies included three randomised controlled trials (RCTs) [40, 43, 44] and one feasibility trial [45] (Tables 1 and 2) with data on 1274 participants from four countries. All studies included women ≥18 years of age with no history or current symptoms of severe or life-threatening mental illness, proficiency in the written language of the intervention and no participation in psychotherapy at recruitment. Those study participants who met criteria for depression ranged from ‘clinical impairment’ to ‘major depressive episode.’

Fig. 1
figure 1

Flow (PRISMA) Diagram of Included Studies. Initial searches of bibliographic databases and reference tracking identified 547 records. Applying inclusion/exclusion criteria to the title and abstract identified 75 entries for full-article analysis. This was reduced to 39 following removal of duplicates and of records that did not have a corresponding full article. 4 studies were fully eligible for inclusion in the systematic review. Reasons for exclusion are outlined within the figure

Full size image

Intervention

The web-based intervention was administered postnatally in all four studies [40, 43–45], one of which recruited women with a recent loss of pregnancy [40]. None of the studies administered the intervention antenatally. Intervention duration ranged from 5 to 15 weeks with all four studies employing a modular structure. In addition to a web-based component, all studies included a therapist/external contact element to the intervention. In two studies [40, 43], contact with a therapist or other professional occurred through text-based contact i.e., email, instant messenger or chat-rooms whilst two studies included telephone sessions [44, 45]. Three studies provided their participants with access to a web-forum of the study cohort [43–45]. Studies differed in their reliance on an external support, with weekly feedback and monitoring proving a key element for some. In the three RCTs the comparator group were either assigned to treatment as usual or a waiting list condition.

Outcomes

Depressive symptoms immediately post-intervention were a primary outcome for three of the studies [43–45] and a secondary outcome for the other [40]. Other primary outcomes included complicated grief and stress. All studies used well-recognised and validated scales for assessing depressive symptoms with two studies using the Edinburgh Postnatal Depression Scale (EPDS) [43, 44]. Three of the studies used a combination of outcome measures with one reporting findings from a single measure [43]. The EPDS score reductions in the two RCTs were (mean ± SEM) 8.52 ± 0.22 (Range 19.46 to 10.94) and 9.19 ± 0.63 (Range 20.24 to 11.05) for treatment groups and 5.16 ± 0.25 (Range 19.44 to 14.28) and 6.81 ± 0.71 (Range 21.07 to 14.26) for comparator groups. Three of the studies included further follow-up at a period ranging from 3 to 12 months post-completion of the intervention. The most common follow-up points were at 3 and 6 months post-treatment. Overall, all four studies described a benefit in maternal mood following intervention, either at post-treatment or follow-up.

Overall rates of attrition in the trials ranged from 13 to 61 % at the completion of the intervention assessment. However, the feasibility study [45] had extremely high attrition rates, with many women struggling to complete the intervention; excluding this study the early attrition ranged from 11 to 13 %. A common theme was the gradual increase in attrition as the study progressed, ranging at follow-up from 24 to 29 %. In one attrition was higher in younger women [40].

Quality assessment

Quality assessment revealed one study to be of good quality, scoring 7 out of 8. Two studies were of intermediate quality (scoring 6 out of 8) where limitations included studying a selected sample (e.g., previous pregnancy loss [40], having child <12 months old [43] thus limiting the generalisability of findings to the general antenatal population. One study was categorised as poor. (Quality Assessment shown in Additional file 2).

Discussion

Although CCBT has been endorsed by NICE for management of depressive symptoms in the general population, this systematic review demonstrates that there has been much less work in this field in relation to the perinatal period. We identified only four studies meeting the inclusion criteria for the review, three of which were RCTs. Importantly none reported an intervention delivered in the antenatal period, highlighting an important research gap. Though there were small numbers of studies with heterogeneity of study design, participant characteristics, duration, timing and nature of the intervention, and intermediate quality of studies, the overall findings suggest web-based interventions implemented in the post-partum period improve maternal mood outcomes.

In accord with other CCBT interventions, the majority of the interventions followed a modular structure. Most also included interaction with a therapist or other health professional as part of the intervention. This complicates the interpretation of the impact of the web-based component alone. Further, web-based interventions placing great importance on therapist input do not address the financial burden of treating perinatal depression nor the waiting list delay. In the case where web-based materials are used solely to supplement therapist mediated interventions, the many practical barriers to face-to-face contact may remain.

All studies utilised a well-validated outcome measure to assess depression symptoms. There was reduction in symptom scores in all studies, with the reported reduction in the two RCTs using the EPDS as an outcome measure, considered a clinically significant improvement [46, 47]. This review was not able to determine whether the severity of symptoms impacts on the effectiveness of CCBT. Three of the studies included longer-term follow-up assessments with the maximum follow-up at 12 months. Due to the relapsing and remitting course of depression, longer-term outcomes are important to evaluate in future studies.

Attrition rates are often high and problematic in other studies of online therapies [48, 49]. Of note the one feasibility study also had high attrition rates (less than 40 % of women completing the follow-up EPDS questionnaire). However, early attrition rates in the RCTs was much lower, and comparable between studies, though notably did increase with longer-term follow-up. There was insufficient data to determine whether attrition rates were related to length or mode of intervention, or whether this differed according to the level of support given during the intervention. Further work is needed to understand the reason for this in order to gain a comprehensive understanding of the true acceptability of CCBT [50]. It is also unknown whether use of a general intervention, rather than an intervention tailored to specific concerns associated with the post-partum period or event triggering the onset of depressive symptoms impacted attrition rates. Future studies should consider use of tailored interventions specific to pregnancy, the post-partum period and pregnancy specific triggering events (e.g., antenatal stillbirth). To understand and limit attrition, future studies should also consider detailed qualitative assessment of reasons for completing the intervention or not, as well as acceptability of the intervention [51], and consideration of co-morbid anxiety as well as depression symptoms.

The strengths of this review include the systematic and comprehensive review process which was followed in line with PRISMA guidelines. Two researchers independently assessed eligibility of the titles, abstracts and full-text studies, extracted the data and assessed the articles for bias. There are also a number of potential limitations. The overall number of studies identified was small and none were in the antenatal period. Whilst this highlights the need for more research in this field, it does limit the interpretation of the findings. Indeed the heterogeneity of study design and lack of consistency in outcome measures meant we were unable to quantitate outcomes. Two of the studies included selected women, and one study recruited both women and their partners, limiting generalisability of the findings. Some of the studies included women with pre-existing depression or taking antidepressant therapy which may have confounded the results with treatment effects. Finally lack of information about acceptability of the interventions was a limitation of these studies.

This review highlights a number of directions for future research. More work is both warranted and necessary within the neglected area of E-health for antenatal depression. Antenatal depression has received less attention than postpartum depression, despite a similar prevalence. This is important, as antenatal depression is a major risk factor for depression in the postpartum period. Thus an intervention administered during pregnancy may also prevent the development and/or progression of antenatal depression to postpartum depression following childbirth. Whilst there is preliminary evidence that CCBT interventions are acceptable during pregnancy [51], further studies are needed to investigate the impact of such an intervention during pregnancy [33, 52, 53]. Our review would also suggest there is a need for more rigor in this field, particularly harmonisation of symptom scores used to allow future meta-analyses of outcomes. Whilst our review focused on a maternal intervention, consideration of an intervention including the partner and/or family may be an important approach for future work, particularly when the triggering event (e.g., fetal death, significant life event) is likely to affect both the woman and her partner.

Conclusions

In summary, this systematic review suggests web-based interventions may be effective for the management of maternal health in the post-partum period and highlights the need for robust, high quality studies conducted during the antenatal period.