A pharmaceutical form of THC in a first trial for PTSD nightmares
STUDY: Roepke S et al, Nature Medicine 2026
STUDY TYPE: Randomized, placebo-controlled trial
FUNDING: Bionorica SE (unrestricted grant; also supplied the study medication)
Background
Dronabinol, the pharmaceutical form of THC, is a schedule-III controlled med approved for nausea from chemotherapy and anorexia from AIDS. It works on cannabinoid receptors, which are also involved fear extinction and sleep, shifting the balance toward non-REM sleep in a way that may help nightmares in PTSD.
This is the first large trial to test it there, but first keep in mind that THC comes with some risks:
- Anxiety, paranoia, hallucinations
- Addiction and withdrawal
- Cognitive impairment, driving impairment
- Heart complications
- Seizures
Design
- 171 adults with PTSD-related nightmares (79% female, mostly civilian assault survivors); 15% dropped out
- Randomized to dronabinol with placebo (both titrated 2.5-15 mg nightly) over 10 weeks
- About two-thirds also took other psychotropic medications, balanced across arms by randomization but not otherwise standardized.
Results
Dronabinol beat placebo on the primary nightmare outcome, with a moderate effect size (Cohen’s d = 0.65, p < 0.001, number needed to treat of 4).
Over half of patients on dronabinol responded (57.9% vs 29.0%), and more than a third hit full remission (36.8% vs 14.5%). Patients felt the difference too: global improvement (PGIC) was far more common with dronabinol (odds ratio 4.66). Self-reported PTSD symptoms (PCL-5) improved more as well. Clinician-rated overall PTSD severity (CAPS-5) and depression (MADRS) didn’t separate from placebo after correcting for multiple comparisons.
Side Effects
No concerning safety signals emerged in this trial. Dry mouth and COVID infections were more common in the dronabinol group. Serious adverse events occurred only with dronabinol (8.0%), though none were judged treatment-related, and discontinuation rates were similar between groups.
Limitations
The trial ran only 10 weeks, so durability and tolerance are unknown. Objective sleep measures weren’t collected. Unblinding was imperfect: 81% of dronabinol patients correctly guessed their assignment, versus 53% on placebo. The population was mostly female civilians, so findings may not extend to combat-related PTSD.
Practice Implications
- This is a first, and opens the door for more research, but we’ll need more safety data given its psychiatric and medical risks.
- Be careful with off-label use. The FDA warns against using it in patients with psychiatric disorders.
- Imagery rehearsal therapy is a simple technique that reduced nightmares in nine randomized trials.
—Chris Aiken, MD
Director, Psych Partners
Editor in Chief, Carlat Psychiatry Report
