Letter to the Editor

We read with interest the article by Gerger and colleagues describing the results of a meta-analysis entitled ‘Integrating fragmented evidence by network meta-analysis: relative effectiveness of psychological interventions for adults with post-traumatic stress disorder’ (Gerger et al. 2014). This article attempted to summarize the available evidence on the effectiveness of psychological interventions for patients with post-traumatic stress disorder (PTSD). The authors included randomized trials in adults with full or subclinical PTSD that compared specific treatments head-to-head to wait-list or other control interventions.

The value of a meta-analysis depends heavily on the scope and quality of the included studies and the methodical, systematic and consistent way the search has been conducted, the studies selected, the analyses carried out and the way the results are interpreted. To this end, the study of Gerger et al. suffers from some limitations that hamper a reliable interpretation of their findings. For example, because of the considerable between-trial heterogeneity, the authors could not identify any intervention superior to other specific psychological interventions. This is not surprising given that the authors used broad eligibility criteria and included experimental studies ranging from non-clinical student samples lacking a formal PTSD diagnosis (but reporting a ‘past stressful experience’; Lytle et al. 2002) to studies with refugees suffering from complex conditions due to exposure to multiple and severe traumatic events (Paunovic & Öst, 2001). This, and the inclusion of older studies that applied preliminary versions of the therapeutic procedures, such as eye movement desensitization (EMD) rather than eye movement desensitization and reprocessing (EMDR) therapy (Lytle et al. 2002), made it likely that patients’ response to the therapies varied significantly, and therefore obfuscated the interpretation of the analytic results.

It is unclear why the authors made a division into ‘small’ and ‘large-sized trials’ and why they used 60 or more patients per trial arm as a criterion for ‘large-sized trials’. Accordingly, five cognitive behavioral therapy (CBT) trials, three exposure therapy trials and one traumatic incident reduction therapy trial were considered ‘large-sized trials’ and subsequently included in the analyses, but for instance the Power et al. (2002) study (comparing prolonged exposure and EMDR therapy with wait-list) with 105 patients was not. It is unclear how this arbitrary criterion for trial size and the selection of studies influenced the results, showing that none of the three specific psychological interventions were superior to supportive therapies. However, the statement that there is ‘most robust evidence for cognitive behavioral and exposure therapies’ (p. 1) and labeling EMDR therapy ‘promising’ (p. 11), although the authors found a consistent (non-significant) trend of higher effect sizes for EMDR therapy after evaluating more than 20 trials (Table 1 and Fig. 2), is difficult to understand and lacks scientific merit. This is underscored by the fact that prior meta-analyses cited by the authors (e.g. Bisson & Andrew, 2007; Powers et al. 2010) have reported that CBT and EMDR therapy are equally effective and empirically supported. Nothing in these data seems to support the current authors’ puzzling interpretation.

Another important reason limiting the informative value of Gerger et al.'s meta-analysis is that, although just recently published, the literature search was carried out in 2010. Accordingly, the authors missed a number of studies that were conducted since then, such as the ‘large-sized’ comparison study of brief eclectic psychotherapy with EMDR therapy (n = 140; Nijdam et al. 2012). Therefore, the contribution of Gerger et al.'s meta-analysis to decision making in clinical practice about what intervention to use to date for patients with PTSD is marginal at best.