[Saving psychoanalysis. At any cost?].

Although long-term psychoanalysis has been criticized due to lacking scientific efficacy, a meta-analysis by Leichsenring and Rabung concludes it is even more effective than other psychological interventions and follows a dose:response relationship (i.e. the longer the better). We reassessed this meta-analysis and find that its conclusions are not warranted. First, major parts of them are not based on the included randomised clinical trials (11) but on observational studies (12), which are known to be subject to investigator bias. The definition of long-term psychoanalysis included mainly randomised clinical trials with fewer than 50 treatment sessions, which is untypically short for this kind of intervention. Comparison groups were misclassified (especially the classification "cognitive interventions"). Important disorders were not clearly defined, hindering any interpretation of which treatment is to be indicated. Moreover the authors did not consider the lack of acceptance and attrition rates, nor did they include intention-to-treat data sets. For some of the studies only 45% of samples were included in the meta-analysis. We conclude that a few of the included randomised clinical trials using psychodynamic short-term interventions truly are promising and of high quality; however these do not justify generalizations on long-term psychoanalysis, and the presented meta-analysis is clearly biased.