Long-term psychodynamic therapy for depression earned a Strong recommendation and a very-low-certainty rating, in the same paper

Sixteen trials, 1,992 people, at least nine months of therapy each. Pooled against all comparators the effect was modest but high certainty; against active treatments specifically, the evidence was rated very low.

Long-term psychodynamic psychotherapy is expensive in time and money, so the question of whether it beats shorter or different treatments is not academic. Researchers led by Max Moser ran a preregistered systematic review and meta-analysis, searching MEDLINE, PsycINFO, Embase and the Cochrane register through January 2025 for randomised trials of therapy lasting at least nine months and at least 30 sessions. Sixteen trials with 1,992 participants qualified. At the end of treatment, the pooled reduction in depressive symptoms against all comparators was a standardised mean difference of 0.34, and against active treatments specifically 0.27. Both are modest effects. Gains were generally sustained at follow-up. A subgroup comparison against active control conditions produced a larger figure, 0.61, and this is where care is needed: that subgroup carried greater heterogeneity and the authors rate its certainty very low. Applying GRADE, they describe high-certainty evidence for symptom reduction across all trials and moderate certainty against active treatments, which under updated empirically-supported-treatment criteria supports a Strong recommendation. The same paper rates the active-control comparison as supporting only a Weak one. Risk of bias was a concern in most of the included studies, and there were too few of them to run moderator analyses. Evidence on functional outcomes, cost, and long-term maintenance was too thin to support a stronger endorsement. So the fair reading is neither headline on its own: the therapy reduces depressive symptoms, by a modest amount, on evidence that is solid in aggregate and weak precisely where it is being compared with the alternatives someone would actually choose between.