Postpartum depression (PPD), also known as postnatal depression, affects women typically within first year following childbirth [1]. It is the most common complication of childbearing as it occurs approximately in one out of every eight deliveries [1].  In different studies, the prevalence of PPD ranged from 7%-20% in UK, US, UK and Australia [2]. PPD negatively affects health seeking behaviors of women not only for themselves but for their infants as well [3]. Studies have reported a low level of social engagement, slow physical growth, and high level of stress reactivity among children born to mothers with untreated PPD [4-6]. Various treatment options, pharmacological and non-pharmacological, are available for treatment of PPD [7]. Non-pharmacological treatment options are preferred in the postpartum period because of adverse effects of pharmacological agents in term of breast feeding [8-10]. Among non-pharmacological options, cognitive behavior therapy (CBT) is one of the effective treatment options for postpartum depression [11, 12]. Previous systemic reviews have examined the effectiveness of overall non-pharmacological interventions including psychosocial and psychological options (not CBT specifically) for PPD [13-18]. These reviews were also not focused to any specific postpartum time period in term of effectiveness of interventions. In addition, the available trials on CBT have reported conflicting results and most trials have limited sample sizes. The aim of this systematic review was to assess the effectiveness of CBT for postpartum depression at least at 24 weeks of postpartum period as compared to the provided standard postpartum care. This specific time period for long term effectiveness of CBT was chosen as most relapses are reported to occur in this period and raised questions over effectiveness of specific intervention [11, 12]. 

Search and selection of studies: We searched three electronic databases, Medline, CINAHL and psycINFO, with search terms, cognitive behavior therapy, postpartum depression, postnatal depression, pregnancy, women, and standard postpartum care. These electronic databases contain original articles from subject of medicine, psychology and allied health sciences. The search was done from January 2000 to March 2013 to get most up to date information about chosen area. Search was carried out by using various combinations of key words (including MeSH terms) for interventions (CBT), control (routine care, postpartum care), outcome (postpartum, postnatal) and study design (randomized controlled trials).  Selection of studies for review was done in accordance with developed protocol which was approved by all authors after critical review. Along with electronic databases, reference and hand search of journals was carried out to identify studies that might have been missed from electronic search. We included all peer-reviewed randomized controlled trials (RCTs) published in English language. We limited our studies to RCTs as they are the gold standards for assessing effectiveness of an intervention. We included only those trials that had enrolled women with normal vaginal delivery and had postnatal depression. Trials in which women had twin pregnancy, or had undergone instrumental or operative delivery were excluded because of higher risk of depression linked with these conditions as compared to normal vaginal delivery.  We included trials with individual-based CBT session as well as with group-based CBT sessions. We excluded trials with pharmacological intervention, psychodynamic therapy, non-directive counseling, behavioral therapy, or action therapy. The control in all the trials needed to be standard postpartum care given by either health visitors or community support workers, though there was a not exclusion criterion. The primary outcome considered for review was PPD at least at 24 weeks of postpartum period. Trials assessing the effectiveness of CBT at less than 24 weeks postpartum were excluded. The secondary outcomes considered for review include anxiety, stress, postpartum marital relationship and cost effectiveness of postpartum intervention. Data extraction: Two independent reviewers used a structured form for data abstraction and discrepancies were solved through discussion. Assessment of methodological quality: Methodological qualities of the included studies were assessed by using “The Cochrane Risk of Bias Tool” as per guidelines [19, 20]. Data Analysis: A qualitative analysis of the results of all included studies was performed to examine the long term effect of CBT on treatment of PPD and to explore the similarities and differences across studies. In meta-analysis quantitative data of the trials were pooled to estimate the effect size of CBT benefit. Software “Review manager” was used to calculate risk ratio (effect measure) along with 95% confidence intervals to assess effectiveness of CBT sessions as compared to standard treatment by using random effects model. We assessed heterogeneity among studies by using I2 statistics and performing test for heterogeneity. I2> 50% and p-value of Chi2 ≤ 0.05 were considered as evidence for heterogeneity. Due to small number of studies, we did not examine publication bias.

We identified 293 studies from the literature search. Figure 1 shows overall study selection stages. Table 1 shows reasons of studies excluded from the review [21-39].  Five of the trials [40-44] fulfilled our eligibility criteria (Table 2). Studies were published between 2001 and 2011. Three trials were conducted in UK [41-43], one in USA [40] and one in Australia [44]. In four trials, randomization was at individual level [40, 42-44] whereas in one trial, it was at cluster (geographical areas in which people registered with specific general practitioners) level [41].  Total number of participants in included studies was 1087. Sample size ranged from 37 [44] to 595 [41].   Three trials [40-42] reported participants’ age range (17 to 42 years) while two trials reported mean age of participants 27.89 [43] and 32.2 years [44]. Intervention: In the included trials [40-44] effectiveness of CBT was compared with standard postpartum care. However, the duration of CBT sessions varied across trials. In three trials [44, 45] sessions were of one hour duration. There were differences in total length of intervention across trials. The total length of intervention in four trials was eight weeks [40, 41, 43, 44]. Of these four trials one [40] had three booster sessions at 6, 16 and 52 weeks of postpartum period. In one trial, length of intervention was up to ten weeks [42]. Studies varied in the personnel who delivered CBT and included health visitors [41, 43, 44], therapist [42], and research staff [40]. The comparator group in all included trials [40-44] got standard usual postpartum care which included routine health visits by community nurses, health visitors and community support workers. In four trials [41-44], PPD was diagnosed if participants scored ≥12 on Edinburgh Postpartum Depression Scale (EPDS) whereas in one trial [40] a score of ≥20 on Beck Depression Inventory (BDI) was used as a diagnostic criterion. The last assessment for outcome was performed at 24 weeks in two trials [43, 44] whereas in the remaining three trials it was done at 52 weeks [40], 72 weeks [41] and 60 months [42]. Secondary outcome including anxiety, postpartum marital relationship and cost-effectiveness of the intervention for postpartum depression were reported by two trials [41, 42]. Quality of studies: Quality of the included studies was assessed by using “Cochrane Risk of Bias” tool.  Table 3 below summarized the methodological quality of included trials. Blinding of participants and personnel for intervention was maintained in one trial [40]. Similarly, there was no clear information about blinding of outcome assessment in trials [40, 42-44] except one [41]. Study power calculation was provided in only two trials [41, 42]. Qualitative Data Synthesis: Two trials [41, 43] showed a significant positive effect of CBT sessions on postpartum depression at 24 weeks of postpartum period. In one trial [41], the proportion of participants with postpartum depression at 24 weeks of postpartum period was 32.9% in intervention group and 45.6% in control group. In the second trial [43] the proportion of participants with postpartum depression at 24 weeks of postpartum was 36.4% versus 65.2% in control group. Three trials [40, 42, 44] found no significant beneficial effect of CBT sessions over routine standard postpartum care provided by health visitors for prevention of postpartum depression at 24 weeks postpartum. In one of the three trials [40], the proportion of participants with postpartum depression at 24 weeks of postpartum period was 25% in intervention group and 31.3% in control group and this trial revealed no beneficial effect of CBT sessions on anxiety and postpartum marital relationship. It is important to note that this trial was conducted among high risk women and 58% to 72% of women in control group sought additional available services. Similarly, half of the women in intervention group could attend fewer than four sessions. This would have mitigated the effect of intervention. In the remaining two trials [42, 44] there was no effect of CBT session but trials had small sample sizes resulting in limited power. Meta-analysis: Effectiveness of CBT sessions was shown by statistically significant differences between the intervention and control groups in term of proportion of participants with postpartum depression at 24 weeks postpartum. Quantitative data was pooled from all included trials [40-44]. Meta-analysis revealed a statistically significant beneficial effect of CBT sessions as compared to standard postpartum care on long term management of postpartum depression. [RR: 0.70 (95% C.I: 0.55 to 0.90)]. Findings suggested 30% reduction in prevalence of depression among women at least 24 weeks of postpartum period in intervention group as compared to control. There was no statistically significant heterogeneity across studies.

We found that RCTs have conflicting results on the beneficial effect of CBT for the prevention of PPD.  Pooling the results using meta-analysis found a statistically significant 30% reduction in the risk of relapse of PPD with CBT over a 24 weeks or longer follow-up. Although four of the five studies did not find a statistically significant benefit, the direction of overall effect size was in favor of CBT. Being small, these studies were not powered to discover a beneficial effect of CBT. Lack of statistical heterogeneity is another indication that study results were generally pointing out the beneficial effect of CBT. There are certain limitations of this review, which need to be considered while interpreting findings. One trial [41] with large sample size had large weight (65.6%) and may have skewed the results in one direction. Inclusion of studies published in English language only might have induced the reporting bias. Similarly, exploration of grey literature and contact with experts in field for unpublished data would have broadened the context of review. Further high qualities RCTs especially with large sample size are warranted to assess the effectiveness of CBT in PPD. It may also be worthwhile to examine the effectiveness of CBT delivered by peers versus health visitors or support workers for postpartum depression. Similarly, importance of marital partner should be considered while designing intervention trials for postpartum depression as lack of social support and marital conflict contribute to the development of PPD.