This was a randomized, double-blind, parallel-group, standard-controlled pilot trial with a 1-month follow-up period. This single-center RCT was conducted at the Second People’s Hospital of Huizhou (Mental Health Center of Huizhou, Huizhou, Guangdong, China) in accordance with the Declaration of Helsinki (revised edition, 2008). The Human Ethics Committee of the Second People’s Hospital of Huizhou approved the study protocol (approval number: 2019-H04-01). Patients or their legal guardians could withdraw from the trial at any time for any reason. This trial was registered in the Chinese Clinical Trial Registry (see text footnote 1), registry number: ChiCTR1900027701 and is reported according to the CONSORT statement (15).

Thirty MDD inpatients with or without psychotic features were recruited from December 1, 2019 to August 24, 2021. Eligible participants aged between 18 and 60 years or their legal guardians provided written informed consent. All patients met ICD-10 criteria for the diagnosis of major depression (16) using the Mini International Neuropsychiatric Interview (MINI) (17) and scored ≥ 21 on the HAMD-24 (18, 19). We excluded the patients if they: (1) failed to respond to earlier ECT; (2) had received ECT or repetitive transcranial magnetic stimulation (rTMS) over the previous 3 months; (3) had a diagnosis of bipolar disorder with or without psychotic features; (4) took medications incompatible with ECT treatment (such as lithium, benzodiazepines, antiepileptic drugs), with any use of these drugs stopped at least five half-lives before the start of ECT; (5) had a lifetime diagnosis of unstable, serious comorbidities (e.g., Parkinson’s disease, multiple sclerosis, stroke, alcohol use disorder) or history of epilepsy; (6) had < 9 years of education; (7) were pregnant or women without adequate contraception; and (8) were in other clinical studies or were unsuitable for participation as assessed by the investigators. All patients continued their usual antidepressive and psychotropic medications; no antidepressant pharmacotherapy changes were made during the treatment course.

We used the bitemporal electrode placement, since bitemporal electrode placement achieved a faster earlier decrease in symptom ratings in a previous RCT (11). For this placement, the two electrodes were applied 2–3 cm above the midpoint of the line connecting the outer canthus of the eye and the external auditory meatus on each side of the head (20).

ECT/LCE procedures were standardized as follows: treatments were three times per week using a spECTRUM 5000Q ECT instrument (MECTA Corporation, OR, United States) with a pulse width of 1 ms and a fixed current of 800 mA. A dose titration procedure to determine seizure threshold was conducted at the first treatment following Mankad et al. (21), and a seizure duration (monitored by electroencephalography; EEG) > 15 s was considered successful (22, 23). Subsequent treatments were administered at 1.5-times seizure threshold for ECT sessions and, for LCE sessions, the energy was set at half the seizure threshold, regardless of whether seizures were induced. Anesthesia included administration of etomidate (0.2–0.4 mg/kg), muscle relaxation with succinylcholine (1–2 mg/kg) and atropine (0–1 mg) depending on heart rate, and ventilation with 100% O₂ throughout the entire ECT or LCE procedures as described in our previous trial (14).

The decision to discontinue ECT/LCE was made by patient’s psychiatrist on the following basic principles based on the daily review: (1) relief of depressive symptoms; (2) insignificant benefits between the two recent ECT/LCEs; (3) side effects; and (4) other medical considerations.

The ECT arm was phase 1, three ECT sessions; phase 2, ECT for the remaining sessions. The HECT arm was phase 1, three ECT sessions (same as the ECT arm); phase 2, LCE for the remaining sessions (see Supplementary Figure 1).

The individualized pharmacological regimen was determined by the patients’ psychiatrists. Patients in both arms maintained their previously prescribed antidepressants and antipsychotics (usually used for patients with depression and psychotic symptoms) during the trial. Anticonvulsant drugs, lithium, or mood stabilizers were discontinued during the ECT/LCE treatment. Single dose short half-life benzodiazepines were used as necessary when patients became agitated or felt anxious, but benzodiazepines were prohibited 24 h before ECT/LCE sessions. When patients suffered from insomnia, zopiclone, eszopiclone, or zolpidem were temporarily prescribed. Medication during the follow-up period was essentially the same as that during the ECT/HECT phase but with lithium, mood stabilizers, or antiepileptic drugs allowed.

A 1:1 allocation sequence was generated using a random number generator¹ by using a simple randomization method, and the sequence list remained concealed in opaque envelopes from the other researchers. The enrolled patients were assigned to the ECT or HECT arm according to this allocation sequence. Participants, the neuropsychological measurement rater, psychiatrists, nurses, and researchers were blinded to patient treatment assignment. To prevent the allocation information from being guessed based on the energy setting parameters on the ECT device, the ECT operator reset the parameters of the ECT equipment after each ECT/LCE session. Allocation status was concealed until the end of the follow-up period.

The four visits were set at (1) baseline; (2) post-phase 1 (i.e., after three ECTs); (3) post-phase 2 (i.e., end of the treatment, within 24–48 h after the last ECT/LCE session); and (4) end of the 1-month follow-up period.

The primary outcome was the change in HAMD-24 between baseline and end of the treatment (i.e., between visits 1 and 3). Response was defined as a decrease in total HAMD-24 score > 50% between baseline and end of treatment, and remission was defined as HAMD-24 total score < 8 at the end of the treatment.

Changes between baseline and the end of the treatment in (1) Hamilton anxiety scale (HAMA) (24) and (2) positive and negative syndrome scale (PANSS) (25) were used to assess anxiety symptoms and psychiatric symptoms (only for patients with psychotic symptoms); (3) the everyday memory questionnaire (EMQ) (26) was used to measure subjective memory function and its impact on daily life at baseline and at the end of the follow-up period.

The following cognitive function tests were used: (1) the twelve subtests of repeatable battery for the assessment of neuropsychological status (RBANS) (27) (list learning, story memory, figure copying, line orientation, picture naming, semantic fluency, digit span, coding, list recall, list recognition, story recall, and figure recall) were used to identify and characterize different aspects of changes in cognitive function; (2) the Stroop color word test was used as a measure of cognitive flexibility and control or executive function (28–30).

Orientation recovery tests (ORTs) after each ECT or LCE procedure were used to measure recovery in orientation after the patients finished the ECT/LCE session and were transferred to the recovery room. A trained nurse asked the patient the following five questions: “what’s your name?” “where are you?” “what’s the date today?” “how old are you?” and “when’s your birthday.” The ORT was recorded as the time point at which all five items were correctly answered.

Any AE or patient who dropped out for any reason was recorded to analyze the safety of ECT/LCE.

The rater and patients were asked to which arm they thought the participant had been allocated at the end of treatment and the end of follow-up period. The ratio of right/wrong guesses was used to evaluate the effectiveness of the blinding procedure.

As this is the first HECT pilot study for MDD patients, we considered ethical, security, and statistical issues and set the sample number to 30 (15 participants per arm).

For the main analysis, Welch’s two-sample t-tests were used for comparisons of the primary and secondary outcomes in the intention-to-treat (ITT) sample. Hedges’ g and its 95% confidence intervals (CI) were calculated to measure the effect size. For other ECT/LCE-related data such as anesthesia doses, seizure thresholds, number of sessions, and ECT/LCE parameters, Welch’s two-sample t-tests were used for comparisons of continuous outcome measures. For the safety results, Fisher’s exact tests were used to analysis differences in dropout and AE rates of different phases between the two arms. False discovery rate (FDR) correction (31) was conducted to accommodate false positive results. For the supplementary analysis, the generalized estimating equation (GEE) model was used to estimate the between- and within-group longitudinal trends in the primary, secondary, and cognitive outcome measures at four visits. The auto-regressive first order (AR1) was set as the working correlation matrix due to the characteristics of repeat-measure clinical data (32). All p-values less than 0.05 (two-sided) were considered significant. All statistical analyses were performed using R version 4.1.0 (R Project for Statistical Computing) within RStudio version 1.4.1106 (RStudio).

Fifty-eight inpatients were entered into the study; 28 screening failures were excluded after entry, and 30 patients were randomized into the ECT or HECT arms at a 1:1 ratio. The patient demographics at baseline are shown in Table 1. Due to symptoms, the baseline and post-phase 1 cognitive tests were not completed for one patient. One patient in each arm withdrew after phase 1 treatment and the results of their post-phase 1 visit were treated as the end of treatment visit for the primary and secondary outcomes according to the ITT principle; they also participated in the follow-up visits. Two patients from the ECT arm were lost to follow-up; all patients in the HECT arm completed follow-up visits. Due to the distance and the local COVID-19 prevention policy, face-to-face interviews at the follow-up visit were impossible with three patients, so they were interviewed online by video chats and, as a result, the cognitive tests were not performed. The details are presented in Supplementary Figure 1 and Supplementary Table 1.

In the ITT sample, the percentage decreases in HAMD-24 in the ECT and HECT arms were –68.3 and –59.14%, respectively. Thirteen patients in the ECT arm and 11 patients in the HECT arm reached response criteria (Fisher’s exact test, p = 0.651), and five patients in the ECT arm and four patients in the HECT reached remission criteria (Fisher’s exact test, p = 1.000).

Changes in HAMD-24, HAMA, PANSS, and EMQ scores are shown in Figure 1, and detailed within- and between-group comparisons and effect sizes are shown in Table 2. Patients in both arms showed significant within-group improvements in HAMD-24, HAMA, and PANSS after treatment, but there were no significant between-group differences in these measures. For EMQ, the within- and between-group differences were both non-significant. In supplementary analyses, there were significant decreases in HAMD-24, HAMA, and PANSS in both arms, but there were no significant between-group nor group*time effects. The results of the supplementary analysis using the GEE model support the results of the main analysis and shown in Supplementary Table 2.

As shown in Table 3, ECT/LCE energy and seizure durations were significantly lower in phase 2 in the HECT arm than in the ECT group. Two participants withdrew during treatment after three ECTs: one patient from the HECT arm due to post-ECT headache, the other from the ECT arm for personal reasons. Two patients in the ECT arm were lost to follow-up for unspecified reasons. The drop-out rates were not significantly different between the two arms. As hypothesized, the differences in AE rates were not significantly different in phase 1 but were significantly lower in the HECT arm in phase 2.

As shown in Table 3, at visit 3, 15 of 28 allocation guesses made by patients and 15 of 28 made by raters were correct, while at visit 4, 15 of 28 allocation guesses made by patients and 14 of 28 made by raters were correct, which were not significantly different from the probability of random guessing. Thus, masking was successful for both patients and raters.

As shown in Figure 2 and Supplementary Table 2, there were improvements in several cognitive measurements after treatments. At visit 3 (end of treatment) and 4 (follow-up), significant group*time effects were detected in eight RBANS subtests and Stroop time interference scores. With respect to RBANS subtests, there were significantly higher list learning, story memory, semantic fluency, digit spanning, coding, list recall, figure recall, and story recall scores in the HECT arm than in the ECT arm at visit 4. Conversely, only one RBANS subtest of figure copying was significantly lower in the HECT arm than in the ECT arm at visit 3. In terms of executive function, a significant lower Stroop time interference score was observed in the HECT arm than the ECT arm at visit 3 (end of treatment). Finally, a shorter ORT was observed in phase 2 in the HECT arm (Table 3).

In summary, participants in the HECT arm generally outperformed those in the ECT arm at the end of treatment for most cognitive tests.

All individual participant data (IPD) are presented in Supplementary Table 1.

HECT has two phases, the first consisting of three standard ECT sessions to achieve a rapid and immediate decline in depressive symptoms; the second containing several LCE sessions to reduce side-effects whilst maintaining efficacy. After phase 1, there was an ∼40% decrease in HAMD-24 from baseline, consistent with a previous study (11), with rapid remission induced by the first three ECTs demonstrated as expected. In phase 2, the decrease in HAMD-24 by LCE in the HECT arm was ∼20%, similar to the ECT arm. The moderate decline in symptom relief of LCEs was similar to that seen for ECT in the later stages of treatment, as predicted. Regarding whole treatment efficacy, the within-group mean effect sizes for the HAMD-24 in the ECT and HECT arms were approximately 2.60 and 2.23, respectively, consistent with previous RCTs of ECT (11, 33, 34). The between-group effect size of HAMD-24 was 0.31, which was non-significant, and the between-group difference remained non-significant at the end of follow-up.

In current ECT protocols, after the seizure threshold is obtained by titration or other empirical formulae such as the “half-age” strategy (35), the parameters of ECT generally remain the same throughout the course of treatment. This assumes that this high energy and its induced seizures are necessary for every ECT session throughout the course of treatment. However, this trial raises the prospect that high-energy ECT and its induced seizures may not be necessary in the later stages of ECT treatment.

In ECT, it is as important to reduce side-effects as maintain anti-depressive efficacy. In this trial, HECT showed similar anti-depressive effects to ECT but a significantly better safety profile. There were no significant differences in AE rates in phase 1 treatment between the two arms, but the HECT arm showed significantly fewer AEs in phase 2 than the ECT arm, similar to our previous HECT trial in schizophrenia (14). The LCE energy in the present trial was set to half of seizure threshold, approximately 1/3 of the energy used in ECT. Although several seizures were induced during LCE sessions, patients in the HECT arm received lower energy, experienced significantly shorter seizure durations, and reported significantly fewer AEs in phase 2. Furthermore, other electrical treatments, such as rTMS and transcranial direct current stimulation, can also exert anti-depressive effects with fewer charge and side-effects (36, 37). Furthermore, the ORT was significantly shorter in the HECT group, as reported in our previous HECT trial (14). A study of magnetic seizure therapy (MST) found that the recovery times for consciousness were shorter in the MST group than ECT (38). MST also uses electrical currents to induce seizures; however, the strength of the electric field in the brain is smaller than ECT according to simulations in a realistic human head model (39). As the only difference between the HECT and ECT treatment is the much lower energy use after the third session, our trial indicates that the high energy use and its induced longer seizure durations in the later stages of ECT sequencing may further increase side-effects.

The well-documented short-term side effects of ECT include retrograde amnesia and acute disorientation, which typically resolve within 3 days of treatment, as well as deficits in memory and executive function that resolve in days to less than 2 weeks after finishing an ECT course (8, 40, 41). In the present study, we used the EMQ to measure daily memory and found no significant change at baseline and follow-up in the ECT group, similar to previous studies on subjective memory (42, 43). Although there was a possible trend toward a better EMQ (lower scores) in the HECT group, the difference between groups was not significant, which may be due to the relatively small sample size and the high heterogeneity of the subjective memory test. We expect to increase the sample size in future studies.

In terms of objective cognition, participants in the HECT group outperformed those in the ECT group for most cognitive tests at the end of treatment. In both the RBANS and the Stroop tests, significant group*time interactions were mainly detected in executive function and objective memory, especially in short-term and working memory at visits 3 and 4. Previous studies have shown that executive and cognitive function recover with relief of depressive symptoms (40, 44, 45). In this study, both groups showed mild average improvements in cognitive and executive function after treatment, but cognitive improvements were more pronounced in the HECT group at visits 3 and 4. The current used in ECT is widely distributed in the hippocampus, amygdala, temporal, and frontal lobes (46) which are mainly associated with memory and executive function (47, 48). Conventional ECT with high currents or epilepsy tends to impact the prefrontal lobe related to executive function, whereas LCE with its lower current may reduce the associated executive impairment. Unfortunately, we were unable to determine whether the corresponding change was caused by a decrease in current or seizure duration. Distinguishing the separate roles of electric current and seizures in ECT will require new experimental paradigms to answer this question, which we are currently developing.

The hippocampus is an important brain region related to memory function (49), and hippocampal atrophy is one of the most reliable biomarkers of depressive disorders (50, 51). After ECT, the hippocampus—especially the neuronal cells of the dentate gyrus—grow significantly (neurogenesis) (52). In animal models, there is a dose-response effect between the increase in the number of new cells in the hippocampal dentate gyrus and the number of electroconvulsive seizures (ECS) (53). In reviewing physiological studies on epilepsy, we also found that neurogenesis is enhanced in the dentate gyrus after epileptic seizures but that the morphology of proliferating neurons is abnormal, such as the formation of basal dendrites in the dentate gyrus hilum and heterotrophic migration to the hilum (54). Seizures also lead to mossy budding of neurons present prior to the seizure. Epilepsy induces the proliferation of new neurons that can be integrated into the hippocampal circuit, but their fibrous connections are abnormal (54). The current study cannot establish whether the accumulated abnormal neurogenesis and immature nerve cells triggered by repeat seizures cause cognitive impairment, but the current study does suggest that lowering the energy during the later stages of ECT is beneficial to cognition, especially in terms of memory function.

This trial has several limitations. First, under ideal circumstances, a sham arm should be used as a placebo control group, but this is unethical in practice, so routine ECT was chosen as a standard control arm. Second, the between-group difference of antidepressive effects was non-significant, it may be just due to the relatively small sample size, but as a pilot RCT, the results demonstrated that HECT, as a new energy setting protocol, deserves further research. We expect to increase the sample size in future studies. Third, while several cognitive measurements improved after treatment, these may not necessarily have been due to the therapeutic effects of ECT or LCE, since the improvement in affective state may have secondarily improved the cognitive deficits associated with MDD (55–57). Instead, the between-group effects and the group*time effects may be more robust and relevant. Fourth, another major limitation is the short 1-month follow up period. Although there was no significant difference in response and remission rates between the two arms at the 1-month follow-up visit, HECT—as a “lower-energy” treatment—may be associated with a higher risk of relapse. Unfortunately, we were unable to answer this question, now. We expect to extend the follow-up period in the next trial to observe the long-term performance of HECT. Fifth, although the retrograde and autobiographical amnesia are the main cognitive impairments of ECT, considering the complexity and longtime cost of autobiographical memory test, we did not evaluate the autobiographical memory test in trial. We hope to rearrange memory-related tests to include the retrograde and autobiographical amnesia examinations in future studies. Finally, due to the coupling of electric current and seizures in ECT, it is very difficult to analyze the separate antidepressant (or other) effects of current and seizures, but we hope that our new research paradigm will overcome this problem.