How well does ketamine work?

The most defensible summary of the evidence: among people with depression that hasn't responded to standard antidepressants, somewhere around half experience a meaningful reduction in symptoms from a course of ketamine, often within days. That is a genuinely strong result for a population that has already failed multiple treatments. It is also a long way from a cure, and the benefit fades without maintenance for many people. Both halves of that sentence matter when you're deciding whether to spend thousands of dollars.

What "success" means in the studies

Research on depression treatments uses two specific words, and clinic marketing tends to blur them. Response means symptom scores dropped by at least half. Remission means scores fell into the range considered minimal or no depression. Response rates are usually higher than remission rates, and "success rate" on a website is often the response number, and it may come from the most favorable time point. When you see a claim, ask which definition, measured when, in how many patients.

The key numbers

• Single infusion: in a well-known randomized trial against an active placebo (midazolam, chosen because it produces noticeable sedation), 64% of ketamine patients responded at 24 hours versus 28% on midazolam. Fast and real, but temporary without further treatment.
• Esketamine (Spravato®): of the three short-term trials behind its 2019 approval, one showed esketamine plus a new oral antidepressant beating placebo spray plus the same antidepressant, by 4 points on the 60-point MADRS depression scale. The other two, one in adults under 65 and one in adults 65 and older, did not reach statistical significance. In the trial behind its 2025 approval as a standalone treatment, esketamine alone beat placebo spray by 5 to 7 points at four weeks. In one manufacturer-run maintenance trial, people already in stable remission who kept taking esketamine had about half the relapse risk of those switched to placebo, though a published critique noted that one study site accounted for much of that result.
• Real-world clinics: a 2022 meta-analysis of real-world ketamine studies in treatment-resistant depression (79 studies, 2,665 patients) found about 45% responded and about 30% reached remission, with wide variation between patients. A U.S. analysis of 537 patients treated at community infusion practices found about 54% responded and 29% reached remission after an initial series, and about 8% got worse. Observational data flatters every treatment. There's no placebo arm, and people who feel nothing tend to quit and disappear from the statistics.

Blinding is a real problem in ketamine research. Patients usually know when they've received a dissociative drug, which can inflate apparent effects even in controlled trials. That does not erase the findings, since the midazolam comparison was designed to address exactly this. Still, it's part of why careful researchers describe the evidence as strong-but-imperfect rather than settled.

When response shows up, and what non-response means

There's no reliable way to predict individual response yet, which means trying it is still the only test. An international expert review found that most people who respond to IV ketamine do so within the first one or two infusions, though some take four to six. If a full induction series produces nothing, more infusions are very unlikely to change that, and a clinic that keeps selling them anyway is serving its own incentives. Non-response is a signal to look at other options, from TMS to medication changes to a fresh diagnostic look. Our guide to what comes next if ketamine didn't work goes through each one.

How to use these numbers

If you're considering treatment, three things follow. First, a real chance of fast relief exists. Second, plan for maintenance from day one, because many people relapse after stopping; our guide to treatment schedules covers what that involves. Third, insist on measurement: a program that scores your symptoms at every visit can tell you by the end of the first series whether your money is buying anything. "Trust the process" is not a measurement plan. The numbers above describe populations; a clinician who knows your history can tell you what they mean for you. More on vetting programs in choosing a clinic.