RECOVER: a negative primary endpoint, a positive trial elsewhere

Design

12-month triple-blind randomised sham-controlled trial, with open-label second-year follow-up

Population

493 adults with markedly treatment-resistant major depression, at least four failed trials in the current episode

Primary outcome

percent time in response on the Montgomery-Asberg Depression Rating Scale over 12 months

Effect

primary outcome not different; clinician-rated response 39.6% vs 30.7% and global improvement 53.8% vs 39.8% favoured active stimulation

RECOVER randomised 493 adults with markedly treatment-resistant major depression to active or sham adjunctive vagus nerve stimulation for 12 months, triple-blind. This was a severely ill group: a mean of 13.3 lifetime antidepressant treatments and a current episode lasting 17.8 years, with at least four failed trials in the episode.

The prespecified primary outcome - percent of time in response on the Montgomery-Asberg scale - did not differentiate the groups. Several other outcomes did: clinician-rated symptom response 39.6% versus 30.7%, global improvement 53.8% versus 39.8%, work and activity impairment 43.3% versus 37.5%, quality of life 45.4% versus 37.1%, and a combined tripartite metric. Durability was the striking part: over 80% of those with meaningful benefit at 12 months still had it at 18 and 24 months, in a group expected to relapse, and about 30% of those without benefit at 12 months improved meaningfully during the second year.

How you read this depends on how strictly you hold a prespecified endpoint. A trial that misses its primary outcome has, formally, not shown its effect - and this one then reports advantages across three separate domains and two years of follow-up. The practical consequence is narrow and real: for the patient who has failed four or more adequate trials, and for whom the alternative is another switch with a low expected yield, this is grounds to discuss a device referral rather than to dismiss it. It does not move stimulation up the algorithm.

• Reserve the conversation for genuinely resistant illness - four or more failed adequate trials, not four prescriptions
• Document the trials, doses and durations before a device referral; centres will ask
• Set expectations at partial benefit: roughly 40% clinician-rated response, not remission
• Warn that benefit accrues over months and that a second year can add response
• In India, availability is limited to a few centres and cost is out of pocket - say so early rather than late

Why it matters

It forces a judgement most clinicians avoid: what to do with a trial that failed where it promised and succeeded everywhere else.

Don't overread it

The primary outcome was negative; the supportive outcomes cannot substitute for it, and the second-year durability data are open-label with no sham comparison.

The statistics, in plain English

When the prespecified primary outcome fails, every other positive result becomes supportive rather than confirmatory - the statistical protection against chance findings applies to the endpoint chosen in advance. Here the secondary outcomes point the same way across symptoms, function and quality of life, which is more persuasive than one lucky measure, but it is not the same as having met the target. Differences of 8 to 14 percentage points against sham are modest in absolute terms in a group with almost no other options.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.