De-implementing treatments from the healthcare system: trauma-focused versus non-trauma-focused therapy of post-traumatic stress disorder as an example.

Recently, a proposal was made to de-implement ineffective or harmful psychotherapeutic treatments from the healthcare system. Given the potential wide-ranging implications of this proposal, a critical review is warranted. As an example of an 'implementation failure', the authors of the proposal discussed non-trauma-focused treatments of post-traumatic stress disorder (PTSD). For this reason, we reviewed the evidence for non-trauma-focused psychotherapies compared with trauma-focused or exposure-based psychotherapies of PTSD in adults, published in the recent 7 years. We focused on efficacy and safety as the central criteria for potential (de-)implementation. The majority of meta-analytical results showed non-trauma-focused psychotherapies to be non-inferior to trauma-focused psychotherapies. In only a few analyses, non-inferiority was not found or exposure therapy achieved statistically significantly better effect sizes, which, however, were small and not clinically significant. Furthermore, non-trauma-focused psychotherapies were found to be safe and well-tolerated with low dropout. Hence, key criteria for de-implementation are not met. The results question the claim that non-trauma-focused therapies constitute an implementation failure. Thus, the example of trauma-focused versus non-trauma-focused psychotherapy of PTSD highlights fundamental problems associated with the proposal to de-implement specific treatments from the healthcare system. The reviewed results suggest that researcher allegiance may have an important impact on (de-)implementation decisions and needs to be rigidly controlled for. Several open questions remain, including consented criteria on which de-implementation decisions would be based, who decides about the actors responsible for making these decisions and, crucially, the mechanisms by which those in positions of authority are held accountable.