Estimates of incidence of perinatal mental illness vary widely from 3 to 25% of women. The perinatal period runs from the beginning of pregnancy through to one year after the birth of the child, and there’s a substantial amount of research that points to the mental health risks that women face during this period.

A recent systematic review published in the Journal of Affective Disorders investigates the effectiveness of CBT for treating and preventing perinatal depression. The review includes a meta-analysis, which was conducted with two aims:

1. To assess how effective CBT (cognitive behavioural therapy) interventions are at preventing or treating depression in postpartum women
2. To look at the moderating impact of a range of demographic, treatment and research-related factors on treatment efficacy.

Investigating these two questions simultaneously is an almost counter-productive process, since the more ‘varied’ each of the individual papers, the less meaning can be derived from the overall meta-analytic verdict on treatment efficacy. It could be argued that Sockol is merely being pragmatic about the inevitable differences between clinical studies and that her work could help researchers to target proven moderating variables, although it is perhaps good fortune for the author that very few of the potential moderating factors appears to have any effect on treatment efficacy.

Sockol’s introduction eschews much background to perinatal depression, focusing instead on the many variables that might influence treatment outcomes, which include the content, format and delivery site of the CBT intervention, the demographic profile of the women being treated, and the involvement (or not) of the womens’ partners in the intervention.

Methods

Effectively, this meta-analysis is actually two closely related meta-analyses, both concerning therapies containing CBT for postpartum depression:

1. One looking at the effect on women already experiencing depression (treatment studies, n=26)
2. One on women without depression symptoms (prevention studies, n=14)

Clear paper searching and selection criteria are provided and a similarly thorough description is offered for the statistical analyses conducted. A broad set of search terms returned 349 papers; 216 were excluded at the abstract-reading stage and it is unclear whether any potentially important studies were lost at this stage; after this, a further 93 studies were removed for various reasons specified by Sockol.

The 40 remaining papers deemed appropriate for the analysis had a range of therapy interventions of 3 to 16 sessions, with a substantial variation in the amount of CBT administered. This makes interpretation of the overall efficacy data from this meta-analysis quite difficult in terms of positing that the effects are specifically related to CBT interventions, especially in light of the variation between control conditions between the studies analysed.

Sockol applied a ten-point quality assessment to each paper, in order to assess the moderating influence of methodological and analytic rigour on study outcomes. She details the ten points in her paper and these might be a useful guide for other researchers looking to assess the impact of research quality on data outcomes.

Results

Sockol’s results are reported separately for treatment and prevention papers.

Treatment

In terms of treatment, a corrected effect size of 0.64 was found at the p<0.001 significance level, showing greater reductions in depressive symptoms for women receiving treatment than women in ‘control conditions’.

In terms of moderating factors, the only apparent moderating influence was whether the treatment was offered antenatally or postnatally, with the latter providing greater reductions in depression symptoms.

Prevention

In terms of prevention, two analyses were conducted, one to look at depression symptom levels pre and post treatment (i.e. the same analysis as for the treatment studies) and one to look at the occurrence of depressive episodes in treatment and control groups post-intervention.

For depression symptom reduction, the data here was less impressive than in treatment studies, which seems logical given that baseline symptoms were obviously much lower than in the treatment studies. An effect size of 0.39 at the p<0.001 level was reported, but this conclusion was only based on eight studies, only four of which reported significant reductions in symptoms when comparing intervention to control.

This suggests that the evidence for effective CBT-based prevention therapies for postnatal depression is highly variable and that the evidence base is currently too small to permit robust meta-analysis. It is of note that Sockol found that those prevention studies meeting more of her quality criteria reported smaller effect sizes, although interpreting this is difficult without knowing which quality criteria were and were not met by each study.

The prevalence of depressive episodes post-intervention was significantly reduced following prevention interventions relative to control conditions at a moderate corrected effect size of 0.71 (p<0.01).

Several demographic and intervention factors which did not influence the treatment studies did act as moderators for the prevention studies. Women who were single, who were non-white and/or who already had children benefited more from intervention, and women receiving individual rather than group therapy reported fewer depressive episodes post-intervention. Of the latter, it is important to note that the individual study data is only pooled from two studies, and evidently further research is needed to see if this finding is robustly replicable. Sockol explores some of the logical arguments for using group therapies such as cost-effectiveness and social support and is clear to point out that group prevention interventions were effective when compared to control conditions.

Conclusions

In her discussion of the results, Sockol talks of how her findings indicate the value of both CBT-based prevention and treatment interventions for reducing depressive symptomatology in postnatal women. Whilst it is broadly hard to argue with this conclusion, it is more challenging to support the ‘CBT-based’ element, when it is clear that there was substantial variation in the amount and style of CBT administered.

Sockol talks about the adaptations of CBT made in several of the studies, particularly for non-white populations; it feels as though it would have been important to talk about this earlier in the paper as a possible explanation for the greater reductions in symptoms for non-white participants. A therapy that is more personally tailored is (one would hope) likely to produce more positive outcomes, but then this idea of individualisation becomes a more crucial moderating factor in itself and perhaps a more accurate one to assess than, in this case, ethnicity.

Sockol pulls on her results as encouraging signs for the clinical applicability of CBT in a variety of formats, which she appears to hope will be useful for clinicians working with more limited time and financial budgets. I am not entirely convinced that the evidence here is strong enough to support such a position. Although the data is encouraging, this feels more of a general overall snapshot that would indicate the need for more research in several more specific areas, and perhaps more research on a manualised intervention of specified duration, where the protocol can be more closely replicated and therefore meaningfully compared meta-analytically.

Link

Primary paper

Sockol LE. A systematic review of the efficacy of cognitive behavioral therapy for treating and preventing perinatal depression. J Affect Disord. 2015 May 15;177:7-21. doi: 10.1016/j.jad.2015.01.052. Epub 2015 Feb 2. [PubMed abstract]

Other references

Gavin NI, Gaynes BN, Lohr KN, Meltzer-Brody S, Gartlehner G, Swinson T. (2005). Perinatal depression: a systematic review of prevalence and incidence. Obstetrics and Gynecology, 106 (5, Pt 1):1071-1083. [PubMed abstract]