Exercise as a Treatment for Depression? Forensic analysis of seven problematic studies commonly included in meta-analyses on the subject

The Medical Evidence Project uses forensic metascience to determine where problematic data or claims have entered important areas of the medical literature. Our ultimate goal is to make medical care safer for patients.

The medical literature and popular health media show considerable interest in the use of exercise in treating depression. That level of interest is not surprising; the World Health Organization estimates that, globally, over one in 20 adults suffer from depression, and exercise would seem to represent a low-risk, low-cost treatment modality. A review of influential medical guidelines and national health policies shows substantial support for the prescription of exercise to depressed patients.

Yet Medical Evidence Project analysts have been increasingly concerned about the quality of the evidence supporting the idea that exercise is an effective treatment for depression. In fact, when the latest Cochrane review on the subject became public in January 2026, our analysts were already investigating several trials in this research domain due to earlier concerns raised regarding the quality and integrity of published research in the field.

This report digs into this area, presenting forensic analysis of seven studies commonly included in systematic reviews and meta-analyses on this subject. Our analysis of these studies reveals critical issues ranging from internal inconsistencies, unlikely baseline data, mathematically impossible formulae, physiological contradictions, and, in one case, a study that reports to have administered a product described with the name of a toothpaste instead of an anti-depressant.

These findings have real-world implications because some clinical guidelines explicitly recommend exercise as a treatment for depression based on the assumption the evidence in favor of it is strong. In reality, the actual impact of exercise on depression remains largely unknown.