does ect work for depression — what the trials actually show

someone said the letters and the room went quiet. ect. you have seen the film. your prescriber says it is nothing like that now, and that it is the most effective thing they have. nobody hands you the trials, or the memory studies, or the relapse number.

here they are. what the largest comparison ever run found, what the sham-controlled trials showed and why the critics say they do not count, what the memory evidence actually measures and what it cannot, why the electrode placement matters, how often people relapse afterwards, how ect compares with ketamine, and what the uk guideline and the fda actually say. seventeen sources, funding and device-maker ties disclosed where the record states them.

what ect is, in one paragraph

electroconvulsive therapy induces a brief seizure under general anaesthesia and a muscle relaxant, usually two or three times a week over several weeks. the variables that matter in the trials are where the electrodes go (both temples, bitemporal, or one side, right unilateral), the dose relative to the person's seizure threshold, and the pulse width (brief or ultrabrief)^[9],[12]. the sine-wave machines that produced the worst cognitive outcomes in older studies are largely obsolete in high-income countries^[6].

the biggest comparison: 113 trials, 6,750 people

the 2019 bmj network meta-analysis of every randomised trial of non-surgical brain stimulation put ect at the top. bitemporal ect had a summary odds ratio for response of 8.91 against sham (95% confidence interval 2.57 to 30.91) and high-dose right unilateral ect 7.27 (1.90 to 27.78); the next best was priming tms at 6.02 and plain high-frequency tms sat at 3.17^[1]. those two forms of ect were also the only strategies that beat other active treatments head-to-head, not just sham^[1].

look at the intervals. 2.57 to 30.91 is wide because few ect trials have a sham arm, and the network borrows strength from indirect comparisons. the finding is that ect is very likely the most effective option in the set; the size of the advantage is uncertain^[1].

the sham trials, and the argument about them

the direct evidence is older. the uk ect review group's 2003 lancet meta-analysis pooled six randomised trials, 256 patients, of real versus simulated ect (anaesthesia without the seizure) and found a standardised effect of −0.91 (−1.27 to −0.54) in favour of real ect. against medication, across 18 trials and 1,144 people, the effect was −0.80 (−1.29 to −0.29)^[2]. those six sham-controlled trials date from the 1970s and 1980s, and there have been none since.

the memory question, split the way the evidence splits it

objective tests. a meta-analysis of 84 studies, 2,981 patients and 24 cognitive variables found decreases in 72% of variables in the first three days after ect, effect sizes from −1.10 to −0.21. from four to fifteen days, all but one interval included zero. after fifteen days there were no negative effects, and 57% of variables were better than before treatment: processing speed, working memory, anterograde memory and some executive function^[5]. the depression was lifting, and thinking got easier.

what that analysis could not measure. its own abstract says it: “No standardized retrograde amnesia tests were identified.”^[5] the memories people most often say they lost, of events before treatment, are the ones the objective literature was least equipped to test. the largest prospective study, 347 patients across seven facilities in the new york area, found retrograde amnesia on an autobiographical memory interview that was still detectable six months after treatment, more severe and more persistent after bilateral than right unilateral ect, and worse with sine-wave stimulation, older age and lower premorbid intellectual function; outcomes differed between facilities largely because technique did^[6]. a 2024 meta-analysis of ten studies argued the standard screening tools were too blunt and found a deficit in long-term learning on an ect-specific instrument, alongside no detriment in memory, attention or language and a gain in executive function^[7].

what patients say. across 72 studies and 5,699 people, 48.1% reported cognitive complaints after ect. whether subjective cognition improved or worsened depended on the instrument: a multi-item memory scale showed large improvement, single-item scales showed worsening, and the strongest predictor of feeling sharper was how much the depression had lifted^[8].

bitemporal, unilateral, ultrabrief

seven randomised trials, 792 people, compared moderate-dose bitemporal ect with high-dose right unilateral. no difference in efficacy: hedges's g −0.03 on depression scores, remission risk ratio 1.06, relapse at twelve months 1.42 (0.90 to 2.23). unilateral had the cognitive advantages: reorientation after each session about eight minutes faster, and less retrograde autobiographical memory loss (g −0.46)^[9]. the 2003 analysis had found bilateral more effective than unilateral; the difference is the dose, and the later trials used high-dose unilateral^[2],[9].

ultrabrief pulse, marketed as the gentlest form, is the one to ask about. across 30 studies and 1,478 people, ultrabrief high-dose right unilateral ect had a pooled remission rate of 32.2% and a response rate of 45.3%; 28.1% of patients were switched to standard brief-pulse ect for inadequate response. the authors' conclusion: it “cannot be regarded as ‘state-of-the-art’ ECT”^[12]. the senior author of that review declares speaker honoraria from mecta, an ect device manufacturer, and from otsuka and janssen^[12]. that is disclosed in the record and it is on this page for the same reason.

the relapse number

ect is an acute treatment. across 32 studies, in the modern-era ones where people took continuation medication afterwards, 51.1% of those who had responded to ect relapsed by twelve months (44.7% to 57.4%), most of them, 37.7%, within six months. continuation ect gave a similar six-month figure, 37.2%. in the randomised trials, antidepressants halved the six-month relapse risk against placebo (risk ratio 0.49, number needed to treat 3.3), and the strongest evidence was for tricyclics; for the newer antidepressants most people are actually given, the evidence was “limited or non-existent”^[10]. four small randomised trials, 254 people, later found that continuation ect plus medication reduced relapse a further 43% compared with medication alone (risk ratio 0.57, number needed to treat 7), with no difference in dropout^[11].

one in two people relapsing within a year is not a footnote. it means the question to ask is not only whether to have ect but what the plan is for the twelve months after it.

safety, and the death rate

the pooled mortality analysis covers 15 studies, 32 countries and 766,180 treatments, with 16 deaths attributed to ect: 2.1 per 100,000 treatments (1.2 to 3.4). in the nine studies published after 2001, covering 414,747 treatments, there was one. the authors compare that with a reported 3.4 per 100,000 for general anaesthesia in surgery^[13]. the risks that matter more in practice are the cognitive ones above and the anaesthetic, which is why both nice documents require a documented assessment of “the risks associated with the anaesthetic; current comorbidities; anticipated adverse events, particularly cognitive impairment; and the risks of not having treatment”^[16].

ect versus ketamine

six trials, five of them randomised, 340 inpatients who were all eligible for ect: ect had the efficacy advantage, standardised mean difference −0.69 (−0.89 to −0.48), with no significant difference on cognition, memory or serious adverse events, and different side-effect profiles (ketamine less headache and muscle pain; ect less blurred vision, vertigo and double vision)^[14]. a larger non-inferiority trial in outpatients without psychosis was published after that search and reached a different conclusion; it is not in the pooled estimate and this page does not quote its numbers because it was not fetched for this page^[14]. the honest summary is that in the sickest inpatients ect wins the comparison that has been run, and in less severe outpatients the question is open. the ketamine and esketamine evidence has its own page in the pipeline.

what nice and the fda actually say

the current uk guideline, ng222 (2022), recommendation 1.13.1: “Consider electroconvulsive therapy (ECT) for the treatment of severe depression if: the person chooses ECT in preference to other treatments based on their past experience of ECT and what has previously worked for them or a rapid response is needed (for example, if the depression is life-threatening because the person is not eating or drinking) or other treatments have been unsuccessful”^[15]. 1.13.2 requires that people be “fully informed of the risks, and of the risks and benefits specific to them”, including “potential adverse events, in particular cognitive impairment”, the risks for older people, and “the risks associated with not having ECT”, and 1.13.3 requires a shared decision with informed consent^[15]. the older technology appraisal still stands for catatonia and mania and for its process rules: “treatment should be stopped when a response has been achieved, or sooner if there is evidence of adverse effects”^[16].

in the united states, the fda issued a final order on 26 december 2018 reclassifying ect devices from class iii to class ii with special controls “for use in treating catatonia or a severe major depressive episode (MDE) associated with major depressive disorder (MDD) or bipolar disorder (BPD) in patients age 13 years and older who are treatment-resistant or who require a rapid response due to the severity of their psychiatric or medical condition”, and requiring premarket approval for all other uses^[17]. a classification is a statement about regulatory controls, not a trial result; what it tells you is which uses the regulator considers established enough to need only special controls, and that everything else does not.

what to ask before you decide

these are questions, not advice. they are the ones the trials make it reasonable to ask.

• which placement and pulse width: bitemporal, high-dose right unilateral, or ultrabrief? the efficacy and the memory findings differ by form^[9],[12]
• how will my cognition be tested, before, during and at the end of the course, and who reads the result?^[6],[16]
• what is the plan for the twelve months afterwards, given that half of responders relapse?^[10],[11]
• if i have had ect before, what happened, and does that change the recommendation?^[15]
• was ketamine or esketamine considered, and why was ect preferred for me?^[14]
• what happens if i say no, and what is the risk of not having it?^[15],[16]