10 Hz binaural beats in opioid rehabilitation: a small sham-controlled trial with very large reported effects

Anxiety and aggression are common in opioid rehabilitation and both make relapse more likely, so a cheap non-drug add-on is worth testing. Soleimanifar and Mansouriyeh randomised 30 men in a residential treatment centre in Tabriz, Iran, to either ten 30-minute sessions of alpha-band binaural beats over three to four weeks, or the same ten sessions of an acoustically comparable audio track with no beats in it. The beat was 10 Hz, produced by a 440 Hz tone in the left ear and 450 Hz in the right, generated with the open-source Gnaural software.

The trial is better built than most in this field. The sham was matched for equipment, room, session length, schedule and researcher contact, so participants could not tell which arm they were in. Allocation was concealed, the outcome assessor was blinded, the primary outcome (the Beck Anxiety Inventory) was named in advance, and nobody dropped out.

The reported results are very large. Anxiety in the beat group fell from a mean of 29.1 to 12.0 while the sham group barely moved, 23.7 to 22.9. Total aggression on the Buss-Perry questionnaire fell from 107.1 to 79.5 against 100.3 to 98.6 in the sham group. After adjusting for baseline scores the group effect was significant for anxiety and for all four aggression subscales, with partial eta-squared values between 0.44 and 0.81. No adverse events were reported.

Those effect sizes are the problem. A partial eta-squared of 0.75 means three quarters of the remaining variance in post-treatment anxiety is attributed to which audio track a participant heard, which is larger than psychotherapy achieves and far larger than the pooled estimates in binaural beat meta-analyses. With 15 people per arm, an effect that extreme is more easily explained by chance, by the sham group’s floor (its anxiety started about five points lower and had less room to fall), or by the single unblinded person in the room, than by auditory entrainment.

Limitations. Thirty participants, one centre, men only, no follow-up after the last session, and both outcomes self-reported by people who knew they were in a study about sound. The trial was registered retrospectively, in January 2026, which removes the main protection against selective reporting. It is a preprint and has not been through peer review, which is why the rubric scores the venue low even though the design scores high. We grade it moderate on method and would treat the size of the effect as unconfirmed until a larger, prospectively registered trial reproduces it.