Randomised controlled trial of ketamine augmentation of electroconvulsive therapy to improve neuropsychological and clinical outcomes in depression (Ketamine-ECT study)

Clinical objectives

• To determine whether or not patients receiving ketamine, compared with saline, would show better/less impaired:

  • VF, autobiographical memory, visual memory and attention/WM at mid-ECT and at the end of treatment with ECT

  • anterograde memory at the end of treatment

  • autobiographical memory at 1 and 4 months after the last ECT.

• To determine whether or not patients receiving ketamine, compared with saline, would show:

  • greater improvement in depression, anxiety and quality of life at mid-ECT

  • more rapid improvement in depression as shown by earlier time to remission and response and fewer ECT treatments needed to achieve remission.

• To assess the safety and tolerability of ketamine augmentation of ECT as shown by:

  • adverse events (AEs) and adverse reactions (ARs)

  • reorientation after ECT treatments

  • psychotic and manic symptoms.

Mechanistic objectives

• To determine whether or not patients receiving ketamine, compared with saline, would show, at mid-ECT and at the end of treatment, better/less impaired:

  • haemodynamic prefrontal cortex responses to a VF task using fNIRS, related to changes in VF performance

  • prefrontal cortex metabolism (TOI) using fNIRS, related to changes in VF performance.

• In addition, although not in the original protocol, we wished to take the opportunity to compare patients and controls on:

  • measures of neuropsychological function

  • haemodynamic prefrontal cortex responses to VF and WM tasks using fNIRS

  • prefrontal cortex metabolism (TOI) using fNIRS.

(Note: in the original protocol mechanistic studies involving fMRI and MRS had been planned, but they were not able to be pursued because of insufficient recruitment.)

Inclusion criteria

• Consent given for ECT as part of standard clinical care.

• Male or female aged ≥ 18 years.

• Current Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (DSM-IV)116 diagnosis of a major depressive episode, moderate or severe as part of a unipolar or bipolar mood disorder, diagnosed by the Mini International Neuropsychiatric Interview (MINI). 117

• American Society of Anesthesiologists score (excluding mental health considerations in the scoring) of 1, 2 or stable 3 and judged as suitable to receive ketamine by an anaesthetist.

• Verbal intelligence quotient (IQ) equivalent to ≥ 85 and sufficiently fluent in English to validly complete neuropsychological testing.

• Capacity to give informed consent.

• Willingness to undertake neuropsychological testing as part of the study.

Exclusion criteria

• DSM-IV diagnosis of a primary psychotic or schizoaffective disorder, current primary obsessive–compulsive disorder or anorexia nervosa.

• History of drug or alcohol dependence (DSM-IV criteria116) within the past year.

• Electroconvulsive therapy in the past 3 months (to avoid confounding the assessment of cognitive outcomes – this was 6 months in the initial protocol but was shortened to 3 months after the commencement of the trial; see Chapter 3, Challenges to recruitment) or previously received ECT in the current trial.

• Known hypersensitivity or contraindication to ketamine or excipients in the injection, including significant cardiovascular disease, uncontrolled hypertension, glaucoma, cirrhosis or abnormal liver function or liver disease.

• Known hypersensitivity or contraindication to concomitant medications used for ECT [propofol, thiopental (thiopental sodium) and suxamethonium (suxamethonium chloride) or excipients in the injections].

• Evidence of organic brain disease including dementia, neurological illness or injury or medical illness that may significantly affect neuropsychological function.

• Detained under the Mental Health Act (MHA) 1983 (as amended in 2007)118 or unable to give informed consent.

• Pregnancy, or at risk of pregnancy and not taking adequate contraception, or breastfeeding.

• Score of < 24 on the MMSE. 119

Inclusion criteria

• Male or female aged ≥ 18 years.

• Psychiatrically well, confirmed through the MINI,117 and on no current psychotropic medication.

• Good physical health.

Exclusion criteria

• Personal history of psychiatric disorder, as revealed by the MINI. 117

• First-degree family history of major psychiatric illness requiring treatment.

• Significant physical illness including organic brain disease, neurological illness or injury that could interfere with the interpretation of the results.

• Psychotropic medication or other medication that could interfere with the interpretation of the results.

• Score of < 24 on the MMSE. 119

Screening visit

This occurred once a potentially eligible patient had consented to ECT and had agreed to see a member of the research team. Written informed consent was obtained after full explanation of the study. Eligibility was determined through a mixture of case note information and a semistructured interview to obtain demographic and background details and by using the MINI to determine diagnosis. Particular care was taken to determine potential physical reasons for exclusion from the study through the results of the standard clinical work-up for ECT including blood tests (full blood count, urea and electrolytes, liver function tests), an electrocardiogram (ECG) and physical examination. IQ and handedness were determined (see Assessments).

Baseline visit

At baseline the neuropsychological tests and efficacy assessments were carried out and, in those who consented to the mechanistic study, this was followed by near-infrared spectroscopy (NIRS) testing. This could be carried out over one or two visits according to patient wishes and practicality.

Randomisation

Randomisation occurred as close to the first ECT as possible to try to minimise dropout, as the modified intention-to-treat (mITT) population was based on patients starting ECT. In practice, this usually meant the day before ECT to allow for the study medication to be prepared and delivered to the ECT suite. Patients were randomised in a 1 : 1 ratio to ketamine or saline using permuted block randomisation (varying randomly from four to eight), stratified by NHS trust by the Christie Hospital Clinical Trials Unit (CTU). The relevant randomisation code was held by the CTU and also provided to the local site pharmacies. When a patient required randomisation the site research assistant (RA) contacted the CTU by telephone with the participant details; the CTU then contacted the local site pharmacy to prepare the drug in blinded, tamper-proof, secondary packaging according to the randomisation code.

Timing of neuropsychological and near-infrared spectroscopy assessments

Neuropsychological assessments were yoked to ECT sessions rather than to time from randomisation because of not-infrequent delays or missed ECT sessions as a result of patient (e.g. infections, failing to fast on the morning of ECT) or organisational (bank holidays, unavailability of staff) factors, as well as the need to carry out the tests within a specified time after the ECT session to capture cognitive impairment. Assessments were carried out at least 24 hours after an ECT session to avoid acute post-seizure and post-anaesthetic effects. The primary outcome assessment was carried out after four ECT sessions (usually just over 2 weeks after the first ECT session), when the effects of ECT should be apparent on cognition anterograde memory. 120 Given the individual variation seen in the number of ECT treatments administered, this time point was chosen to capture as many patients as possible before they stopped ECT, either because of dropout or improvement. Further assessments were carried out after the last ECT session (end of treatment) and at follow-up at 1 and 4 months after the last ECT session.

For those patients who had consented to the NIRS substudy, the imaging was carried out after the neuropsychological assessments.

Timing of efficacy assessments

Efficacy assessments were carried out weekly until the end of treatment, with the week-2 assessment mostly coinciding with the mid-ECT neuropsychological assessment. The primary efficacy measure was improvement in the Montgomery–Åsberg Depression Rating Scale (MADRS),121 an observer-rated measure of depression severity.

Timing of safety assessments

Safety assessments were carried out during and after each ECT session in accordance with standard good practice in the administration of ECT and included monitoring of vital signs and pulse oximetry. Degree of reorientation 30 minutes after ECT was used to assess recovery after ECT and staff were instructed to record any problem during the recovery period as an AE or possible AR. Weekly mania and psychosis ratings were carried out as part of the efficacy assessments.

Electroconvulsive therapy

Standard ECT protocols based on the Royal College of Psychiatrists’ (RCPsych) ECT Handbook122 were agreed between centres, with ECT treatments scheduled twice weekly. For each ECT treatment, after pre-oxygenation with 100% oxygen, anaesthesia consisted of propofol (or rarely thiopental if clinically indicated) combined with the muscle relaxant suxamethonium. Etomidate was not permitted as an induction agent because of its action in elevating blood pressure. Anaesthetic drugs remained the same for all ECT treatments unless change was required for clinical reasons. ECT treatment was given after motor end-plate depolarisation was determined by cessation of muscle fasciculation in the feet. Electrode placement was standard bifrontotemporal (bilateral) or RUL (D’Elia placement) and the treatment consisted of constant-current brief-pulse (0.5-millisecond pulse width) stimuli to induce seizure. The pulse width could be increased to 1.0-millisecond if clinically indicated. Treatment dose was determined by rapid stimulus titration to determine the seizure threshold in the first session treatment, based on that given in the RCPsych ECT Handbook. 122 Target treatment doses were 1.5 times threshold for bilateral and 4–6 times threshold for RUL electrode placement and the stimulus parameters remained the same until after the fourth treatment unless change was required for clinical reasons. Standard operating procedures were agreed between treatment centres to determine criteria for an adequate seizure, and for re-stimulation and dosage adjustments later in the course, based on the RCPsych ECT Handbook. 122

Oral psychotropic medication was continued by the clinical team and remained unchanged for the first four ECT treatments and, if possible, until the end of ECT, unless changes were required for safety or clinical reasons.

The goal was to treat patients to remission (MADRS score of ≤ 10) in accordance with NICE guidance5 but the final decision to finish ECT treatment rested with the clinical team in consultation with the patient and the ECT team.

Study medication and blinding

The experimental intervention was 0.5 mg/kg of intravenous ketamine or an equal volume of normal saline (0.9% sodium chloride solution), given in accordance with the randomisation schedule by the anaesthetist as a slow bolus directly before the anaesthetic induction agent and before suxamethonium at each ECT treatment. Ketamine (at a concentration of 10 mg/ml) or saline was provided as a clear solution in ampoules retaining their original labelling, provided in secondary tamper-proof trial packaging by the local clinical trials pharmacy. The anaesthetic team broke the seal away from the psychiatric ECT team, drew up the trial medication into a syringe and disposed of the ampoule without revealing which drug was being administered. Members of the research team responsible for outcome assessment did not attend ECT sessions to minimise the risk of treatment allocation being inadvertently revealed.

Mini International Neuropsychiatric Interview

The MINI is a brief structured interview for the major axis I psychiatric disorders in the DSM-IV,116 with good concordance with more comprehensive diagnostic interviews for mood disorders. It can be used with little training by clinicians but can be used by lay interviewers after training (provided in this study by IMA) and takes approximately 22 minutes to administer. 117

Mini Mental State Examination

The MMSE is a 30-item screening tool for dementia with scores of < 24 being considered abnormal. 119

Wechsler Test of Adult Reading

The Wechsler Test of Adult Reading123 is a language test to assess premorbid intellectual functioning and is thought to be relatively insensitive to cognitive deterioration. It uses vocabulary level as a correlate of IQ with the patient being presented with irregularly spelled words and prompted to pronounce each one. It consists of 50 irregularly spelled words of progressive difficulty and is discontinued after 12 consecutive incorrect pronunciations. Each correct pronunciation is given a score of 1 and the raw score is then standardised by age and compared with the scores predicted for the patient’s demographic classification to provide an estimate of premorbid IQ.

Modified Edinburgh Handedness Questionnaire

The modified Edinburgh Handedness Questionnaire124 asks participants to indicate which hand they use for eight common tasks, with responses ranging from ‘always left’ to ‘always right’ on a five-point scale. This is scored to determine right-, left- or mixed-handedness.

Massachusetts General Hospital Scale

The Massachusetts General Hospital Scale (MGHS)125 is a method for staging pharmacological treatment of depression that generates a continuous variable representing the degree of treatment resistance, taking into account the intensity and optimisation of each previous treatment. Non-response to an adequate trial of a marketed antidepressant (≥ 6 weeks at an adequate dose) scores 1 point per trial, with 0.5 points per trial added for each optimisation strategy. ECT increases the score by 3 points. In one study a score of 3 was found to confer a < 20% chance of remission with the next treatment, which diminished as the score increased further. 126

Other assessments

Other assessments consisted of those carried out by the clinical team as work-up for ECT and included a physical examination, including height and weight, cardiovascular system and blood pressure, an ECG and blood tests, including liver function tests.

Monitoring before, during and after anaesthesia and electroconvulsive therapy treatment

Safety measures were those applied in standard practice, consisting of monitoring heart rate, blood pressure and peripheral oxygen saturation using pulse oximetry and, if indicated, capnography.

Reorientation

Reorientation was used as a measure of recovery from ECT and is a routine clinical assessment. To provide a standardised measure, reorientation 30 minutes and 60 minutes after eye opening was recorded, defined as being correct in four out of five orientation items (name, place, day of the week, age and birthday). 127 In addition, the number of correct orientation items at these time points was recorded. Reorientation usually occurs before 30 minutes49 and a delay beyond 60 minutes is extremely rare. The 30-minute time point was used to compare treatment arms.

Agitation and psychosis

This was not formally assessed during the recovery period but ECT staff were asked to record events that were unusual or more severe than expected as AEs, as well as any suspected ARs. The 18-item Brief Psychiatric Rating Scale (BPRS),128 with an additional 19th item for elevated mood, was used to assess symptoms of psychosis (sum of items 3, 4, 7, 8, 11, 12, 15 and 16) and mania (sum of items 8, 10, 17 and 19) outside the ECT sessions, at the same time as efficacy assessments.

Hopkins Verbal Learning Test – Revised

The HVLT-R115 is a word-list learning task in which the participant is asked to listen as a word list consisting of 12 words is read out to them and then to recall as many words as possible; this is carried out three times. Following an interval of 30 minutes the participant is given an unannounced recall task for the words previously learnt and a forced-choice recognition task in which the previously presented words are mixed with 12 words not heard before. Although repeated administration means that the delayed recall is no longer a surprise, this does not appear to affect performance. 131 The sum of the number of words recalled after the three presentations is used as a measure of learning and the number of words recalled after the delay is used as a measure of delayed recall (the primary outcome in this study). In addition, a retention score is calculated (the number of words on trial 4 divided by the highest score on trials 2 and 3 expressed as a percentage), as well as a recognition discrimination index (the number of true positives minus the number of false positives).

It is available in six alternative forms, which were used at the different time points in the study, with two different orders of presentation assigned according to even or odd participant numbers.

Controlled Oral Word Association Test

The Controlled Oral Word Association Test (COWAT)132 tests letter and category fluency, with participants asked to generate as many words as they can in 1 minute starting with the letters F, A or S and in a given category (animals or fruit and vegetables), with rules excluding proper names or variations of a word. The same letters were presented on subsequent tests and the two categories were alternated with a different order of presentation assigned according to even or odd participant numbers.

Autobiographical Memory Interview – Short Form

The Autobiographical Memory Interview – Short Form (AMI-SF)133 tests memory for personal events in the past with reference to an initial interview eliciting personal memories. This covers six areas with a score given to any events remembered with sufficient detail to be questioned about them again on subsequent occasions. On retesting, only items that had been mentioned the first time can be scored, with points deducted for missing items, so it is possible for scores to decrease even if a richer memory overall is recalled. The AMI-SF has been criticised as it does not allow for improvements to be detected, for its lack of normative data for the normal loss of consistency over time and for a lack of information about the independent effect of mood,22 although it has been defended against these criticisms. 19 An alternative extended scoring method has been recently proposed that allows more details to be scored at baseline, with points given for all items recalled. 134 A decrease in the consistency of memory of about 20% over 6 months in healthy volunteers has been found with the extended scoring method. 134 In this study we used both the standard and the extended scoring method to assess autobiographical memory. We did not administer the AMI-SF to HCs as we were not retesting them, and the baseline scores have little comparative value, although depressed patients do produce fewer episode-specific memories than healthy volunteers. 134

Medical College of Georgia Complex Figure Test

The Medical College of Georgia Complex Figure Test (MCGCFT)135 consists of first copying, and then reproducing from memory (immediately and after a delay), a complex line drawing with multiple elements, which tests not only visuospatial abilities but also memory, attention and planning. It is available in four alternative forms and different forms were used for the first four time points in the study, with two different orders of presentation assigned according to even or odd participant numbers.

Digit span

For the digit span task, strings of numbers of increasing length are read out slowly and in an even rhythm and participants are asked to repeat them back. 136 In the forward condition they are repeated in the same order as presented and in the backward condition they are repeated in reverse order. This task is a test of attention and being able to hold the numbers in WM to reproduce them, particularly for the backward condition. The number of digits remembered is reported (the ‘clinical’ digit span).

Self-reported Global Self-Evaluation of Memory

One of the problems in research into memory problems following ECT is the lack of agreement between the impact on subjective memory and objective memory measures. The self-rated Global Self-Evaluation of Memory (GSE-My)137 has been reported to correlate with AMI-SF scores after ECT. Patients are asked to rate their global memory and the effect of ECT on their memory on a 7-point scale.

Montgomery–Åsberg Depression Rating Scale

The MADRS is a 10-item observer-rated depression scale, which is sensitive to change and has internal consistency. 138 Each item is scored from 0 to 6 on the basis of defined anchor points; a score of ≤ 10 was taken as the cut-off for remission. 139 In addition to the standard MADRS, an alternative bespoke form was scored that included atypical depressive symptoms of increased appetite and sleep (MADRS atypical).

Clinical Anxiety Scale

The Clinical Anxiety Scale (CAS)140 is a 7-item observer-rated scale derived from the Hamilton Anxiety Scale for measuring the severity of anxiety. The seventh item is meant to be scored separately. Each item is scored from 0 to 4 on the basis of defined anchor points, and we calculated the total score and the total score of the first six items for this study.

Clinical Global Impression-Severity and Clinical Global Impression-Improvement

The Clinical Global Impression scale is a widely used 7-point observer-rated scale that is used to measure global severity [Clinical Global Impression-Severity (CGI-S)] judged in comparison to the overall range of depression seen in patients in clinical practice (1 = ‘normal, not at all ill’ to 7 = ‘among the most extremely ill patients’) and overall global improvement [Clinical Global Impression-Improvement (CGI-I)] compared with baseline (1 = ‘very much improved’ to 7 = ‘very much worse’). 141

Brief Psychiatric Rating Scale

The BPRS is an observer-rated instrument used to measure the overall range of psychiatric symptomatology. We used the 18-item version with an additional 19th item for elevated mood. Each item is scored from 1 to 7 on the basis of defined anchor points. Subscales were used to assess psychosis and mania as described under Safety assessments.

Quick Inventory of Depressive Symptomatology – Self Report

The Quick Inventory of Depressive Symptomatology – Self Report (QIDS-SR)142 is a 16-item self-rated depression scale that was used in the STAR*D study4 and rates the nine DSM-IV depression items, each on a scale of 0–3.

European Quality of Life-5 Dimensions three-level version

The European Quality of Life-5 Dimensions three-level version (EQ-5D-3L)143 is a self-rated quality-of-life instrument, with participants rating their degree of impairment or experience on five dimensions (mobility, self-care, usual activities, pain/discomfort and anxiety/depression) on a scale from 1 to 3. Based on standardised population studies, each item is weighted and an overall quality-of-life ‘index’ score is calculated to give a utility score ranging from 0 (death) to 1 (perfect quality of life); the lowest score is –0.59, that is, worse than death. In addition, a visual analogue scale (VAS) is scored from 0 (worst imaginable health state) to 100 (best imaginable health state).

Continuous outcomes

The weekly efficacy data were analysed using a random-effects (random intercepts and slopes) ANCOVA model with time (in weeks) since the first ECT as a quantitative explanatory variable. To prevent bias from efficacy assessments being delayed, time from first ECT to assessment was included as a continuous variable in a longitudinal mixed-effect model rather than the notional assessment week recorded by the RA (RA week). Only data until the end-of-treatment assessment were included (i.e. excluding follow-up data). The fixed covariates included in each model were treatment, the time × treatment interaction and the same prespecified baseline participant characteristics as used for the neuropsychology repeated measures analyses. For each model, two correlated random-effects terms were included to capture between-subject variation in the intercept and slopes, assuming that people might have different rates of improvement with treatment. All models involved the use of robust standard errors and associated CIs and p-values.

A sensitivity analysis was performed on the primary efficacy outcome, the MADRS score, by fitting a longitudinal mixed-effect model using time since randomisation instead of time from first ECT.

Binary outcomes

To assess efficacy data up to the end of acute treatment for the remission and response binary variables, the frequency and cumulative frequency of first occurrence according to the week recorded by the RA (RA week) up to the end of treatment were tabulated. When the end of treatment occurred < 13 days after the previous week’s assessment it was designated as being 1 week after the last assessment; when it was ≥ 13 days then a gap of 2 weeks was assigned instead.

Kaplan–Meier plots and Cox analyses (adjusting for the same baseline variables as used for the repeated measures and longitudinal mixed-effect models, including baseline MADRS score) were undertaken. For these analyses the time to ‘event’ was defined as the difference between the first ECT date and the date of assessment. For censored subjects, their last known efficacy assessment date while on acute treatment relative to their first ECT date was used.

The proportion worsening at follow-up after acute treatment (MADRS increase of ≥ 4 points + CGI-S increase of ≥ 1 point to CGI-S ≥ 3 from the end-of-treatment assessment or, if not available, the last RA week assessment) was determined provided that there was a MADRS and CGI-S score at the end of treatment and at least one follow-up time point. The proportions worsening in each group were compared using Fisher’s exact test.

Imaging tasks

The fNIRS tasks were administered after the main study neuropsychological tasks and, mindful of this, were designed to be as short and as straightforward as possible. The total time for testing was about 40–50 minutes (21 minutes performing the tasks).

Near-infrared spectroscopy analysis

Poor signal seen in the 50-mm channels meant that these data could not be used and the planned TOI analysis was not possible; the analysis method for this is therefore not described.

Functional near-infrared spectroscopy data were analysed with the Homer2 NIRS processing package146 based in MATLAB [see uk.mathworks.com (accessed 30 August 2016)]. For each participant, channels that measured a very low optical intensity (below the noise level of the device, e.g. because of poor optode–skin contact) were discarded from the analysis. Intensity data were then converted into optical attenuation data. For this preliminary analysis motion artefacts, that is, time points for which data exceeded selected changes in amplitude (AmpThresh) or standard deviation (SD) (StdThresh) over a period of 0.5 seconds, were identified in each channel. The choice of these parameters is data dependent, and a compromise between the number of motion artefacts identified in noisy channels and the number identified in less noisy channels is needed. In this study, AmpThresh = 10, StdThresh = 0.5 for all of the HC data and the patient mid-ECT data for the n-back task and AmpThresh = 8, StdThresh = 0.5 for all of the other patient data. The identified motion artefacts were corrected by applying a spline interpolation algorithm,147 which models the artefact with a cubic spline interpolation, subtracts it from the original signal and then corrects for the baseline shift caused by the subtraction. As some of the motion artefacts might still be present in the data even after this correction step, trials with remaining motion artefacts were rejected. A band-pass filter (third-order Butterworth filter) with cut-off frequencies of 0–0.5 Hz was applied to the data to reduce high-frequency noise. The filtered optical attenuation data were finally converted into concentration changes by applying the modified Beer–Lambert law,148 assuming a differential pathlength factor of 6.26. 149 Mean haemodynamic responses were calculated by block averaging all trials in an interval from 5 seconds before stimulus presentation to 60 seconds after stimulus presentation (hence containing both the 30-second task period and the 30-second rest period). Mean haemodynamic responses were baseline corrected by subtracting the mean value in the interval from –5 seconds to 0 seconds.

This produced one mean haemodynamic response for each channel in each participant, which was divided into three blocks of 20 seconds (0–20 seconds, 20–40 seconds and 40–60 seconds) (Figure 4). The integral of the mean haemodynamic response in each of these blocks was computed [area under the curve (AUC) for blocks 1, 2 and 3] for analysis.

FIGURE 4.

Example of mean haemodynamic response and block timing. The timings of the three 20-second blocks used in the analysis are shown (block 1: 0–20 seconds; block 2: 20–40 seconds; block 3: 40–60 seconds). The value of the integral in these blocks was used for fNIRS analysis. Hb, haemoglobin.

Given that the haemodynamic response typically lags behind neuronal activation by about 6 seconds, block 1 includes the onset of the haemodynamic response during the active task, block 2 includes the period of greatest sustained response during the task and block 3 includes the return of blood flow to rest level. As displayed in Figure 3, the areas of cortex in which HbO and HbR are measured by the array form a ring around the detectors. The data were collapsed to cover four regions, two on each side, an inferior and more posterior region consisting of BA 44, BA 45 and the inferior part of BA 9 [left ventrolateral prefrontal (LVF) and right ventrolateral prefrontal (RVF) regions] and a superior and more anterior region consisting of BA 46 and parts of BA 9 and BA 10 [left dorsolateral prefrontal (LDF) and right dorsolateral prefrontal (RDF) regions] (see Figure 3). This was carried out by calculating the mean of the channel values for that region [LVF: mean(D, C, L); LDF: mean(K, H, G); RVF: mean(F, I, J); RDF: mean(E, A, B); letters refer to channel sources shown in Figure 1] (see Figure 3). The long 50-mm channels had a poor signal and were excluded from the analysis, so that each region consisted of the mean of up to five channels (three 35-mm channels and two 43-mm channels).

The optimal haemoglobin factor for detecting cortical activation remains unclear: HbO is regarded as the most sensitive and has been most commonly reported in tasks of cortical activation, HbR changes less and appears minimally influenced by motion-induced changes and the difference between HbO and HbR values (Hbdiff) may provide an overall ‘oxygenation index’. 150 Analysis was carried out using all three measures.

Analysis was carried out in IBM SPSS Statistics version 22.0. To identify whether or not a haemodynamic response occurred in each group, a repeated measures ANOVA was carried out in HCs and patients at baseline separately over the three time blocks for each task. The groups were compared (using patient baseline data) by analysing the individual mean AUC values over the three time blocks by a univariate repeated measures ANCOVA with one between-group factor (group) and two within-group factors (region and time). Age and sex were used as covariates given the group differences (see Table 16), which, although non-significant, may influence fNIRS signal. 151,152 The effect of ECT and drug were analysed in patients alone using a repeated measures ANOVA with one between-group factor [drug (saline vs. ketamine)] and three within-group factors [ECT (baseline vs. mid-ECT), region and time]. The Huynh–Feldt correction was applied to correct for sphericity violation, and post hoc exploration of individual regions was carried out by repeated measures ANOVA or ANCOVA by region. For the VF task, the LVF was identified a priori as a region of interest as fNIRS haemodynamic responses in this region have been shown to relate to fMRI BOLD signal change during a VF task. 153 For correlational analysis between regional haemodynamic responses and behavioural data, block 2 AUC values, the time of maximum task-related response, were used, employing Pearson’s product–moment correlation coefficient (r). Results are presented as mean (SD) or mean ± standard error of the mean (SEM) with unadjusted degrees of freedom for clarity and statistical significance at p < 0.05. Given the small patient numbers and exploratory nature of the analysis, trend significance (p = 0.05 to p < 0.1) is also reported and no correction was made for multiple comparisons.

Challenges to recruitment

Patient recruitment was slow, especially at first, and this was compounded by fewer sites than planned taking part initially and a relative low rate of ECT use within trusts; two of the six original trusts were not able to participate but this became apparent only after study initiation. Given that recruitment was from a fixed population of patients receiving ECT, the only way to increase numbers was by recruiting new sites. The initiation of new sites took considerable time because of uncertainties engendered by NHS reorganisations, the organisational complexity of ECT provision (varying at different ECT suites within each trust), practical difficulties and governance requirements in undertaking a clinical trial involving a controlled medicine and having to recruit sites with little research experience. By the end of the study seven mental health trusts had been recruited, involving 11 ECT suites; a further two ECT suites that had been initiated closed because of internal trust reorganisations during the trial (Figure 6). In spite of the predicted shortfall in recruitment it was not practically or financially feasible to recruit more sites; in discussion with the funder and the DMEC the primary outcome was revised to a single outcome (rather than three inter-related measures) to maintain as much statistical power as possible (see Sample size).

FIGURE 6.

Timeline of trusts and ECT suites starting in the trial.

At most trusts the referral process for ECT occurred with short notice, so that timely identification of patients to allow recruitment and assessment was challenging. Attempts to minimise this barrier involved working with the ECT suites so that they could accommodate the study within their individual referral pathways; however, organisational and time pressures, and clinical need, necessarily took precedence, given the severity of illness of the population. Particular emphasis was given to raising awareness of the study within participating trusts by the study research teams, supported by the Mental Health Research Network/Local Clinical Research Networks, which also provided cover for screening and assessments at a number of sites.

At an early stage it was recognised that a significant number of patients had received ECT in the last 6 months and were therefore being excluded. This criterion was reviewed in the light of the literature that cognitive function improved rapidly after the end of ECT and was altered to 3 months in a protocol amendment in May 2013.

A major hurdle to recruitment was the higher than predicted proportion of patients who were ineligible because of being detained under the MHA, even though a proportion might have had capacity to consent to ECT and therefore potentially to the study. This criterion was therefore reviewed carefully early in the study, in discussion with the service user group. The decision was made that it should remain in place given concerns that the principle of fully informed consent could be undermined if patients were detained in hospital against their will, that is, some patients might feel under an obligation to participate. In practice, this is unlikely to have greatly affected recruitment as most detained patients receiving ECT lacked capacity to consent to the study, either because of the severity of their depression or because of cognitive impairment. 9

In mid-2014 an open letter was published by a group of academics and clinicians requesting that the funder and ethics committee suspend the trial on the basis of risks to patients and the participant information being misleading (details can be found in the study team’s rebuttal on the Guardian newspaper’s headquarters site154). This request was deemed unfounded by the funder and the Health Research Authority and was therefore rejected; it did not appear to directly impact on recruitment.

Problems in ketamine supply were a major threat to the trial from early in the study when, in early 2014, Pfizer suspended the production of Ketalar, the only available ketamine preparation in the UK (supplies had not been resumed by the end of recruitment). Alternative sourcing was obtained to allow the trial to continue, first from Germany and then, when these supplies also dried up (because of clinical demand in the absence of Ketalar), by successfully importing a non-EU-licensed product from the USA. Re-sourcing ketamine was a time-consuming process, involving considerable uncertainty and substantial amendments to the MHRA Clinical Trial Authorisation. As a result, some sites temporarily ran out of ketamine (which could not be transferred between sites for governance reasons), but, fortunately, only two potentially eligible patients were missed for this reason.

End of study

The trial was stopped at the end of the study period, as agreed with the Efficacy and Mechanism Evaluation (EME) Board. Because of recruitment problems, this involved a 4-month extension and a revised recruitment target and curtailed follow-up of the last few patients to maximise the duration of recruitment for the primary outcome. Recruitment remained problematic to the end, largely because of the paucity of eligible patients; the final mITT sample of 70 gave 65% power to detect the effect (ES 0.57) hypothesised in the revised sample size determination.

Primary outcome

The primary outcome was HVLT-R delayed recall as a measure of anterograde verbal memory at mid-ECT. Figure 7 and Table 6 show the number of words recalled by visit over the course of the trial; the treatment effect at each assessment point is given in Table 6. No significant difference between treatment arms was found, with the estimate favouring saline. The sensitivity analysis excluded assessments carried out 4 and 5 days after the last ECT session at mid-ECT (n = 4) and assessments carried out 6–12 days after the last ECT session at the end of treatment (n = 11), with no substantial change in the results (mid-ECT –0.56, 95% CI –1.91 to 0.79; end of treatment –0.56, 95% CI –1.82 to 0.71; follow-up 1 –0.53, 95% CI –1.65 to 0.60; follow-up 2 –1.43, 95% CI –2.95 to 0.10).

FIGURE 7.

Hopkins Verbal Learning Test – Revised delayed recall scores by visit in patients receiving ECT randomised to saline or ketamine. Numbers refer to observed valid assessments at each time point. Values are means ± SEM.

Secondary neuropsychological outcomes

There was no benefit of ketamine for any secondary neuropsychological outcome (see Tables 6 and 7). Most effects were in the direction of favouring saline but were not significant, apart from two measures at different time points: end-of-treatment HVLT-R recognition and mid-ECT forward digit span. Self-evaluation of memory showed that the majority of patients reported a negative effect of ECT; the scores indicated that patients on average judged the effect of ECT to be mildly detrimental (Table 8).

Change in neuropsychological function during the electroconvulsive therapy course

Inspection of test scores at mid-ECT and end of treatment showed only small changes, if any, from baseline in most scores. Repeated measures ANOVA showed that the only measures with a significant deterioration over the course of ECT were AMI-SF standard and extended scoring (both p < 0.001) and MCGCFT immediate recall (p = 0.026) and delayed recall (p = 0.001).

From the end of treatment to follow-up 2 (4 months after the end of ECT) repeated measures ANOVA showed significant improvements in most measures (HVLT-R delayed recall, p = 0.01; HVLT-R retention, p = 0.012; COWAT category fluency, p = 0.018; MCGCFT immediate recall, p < 0.001; MCGCFT delayed recall, p < 0.001; AMI-SF with standard scoring, p = 0.011), although a trend was seen only with the extended scoring method (p = 0.089). No change over time during the follow-up period was seen with digit span forwards or backwards or COWAT letter fluency. The pattern of improvement was for HVLT-R measures to improve between the end of treatment and follow-up 1, for COWAT category fluency and MCGCFT measures to improve progressively and for AMI-SR with standard scoring to improve only between follow-up 1 and follow-up 2. Multivariate repeated measures ANOVA for these six measures confirmed a significant measure × time interaction (Wilks’ lambda p = 0.012), supporting a different pattern of improvement over time between measures.

In the comparison with baseline, values were numerically better at follow-up 2 for all measures except for the AMI-SF. However, significant improvements (using paired t-tests) were seen only with MCGCFT immediate (p = 0.004) and delayed (p = 0.019) recall, digit span backwards (p = 0.046) and HVLT-R recognition (p = 0.016). A significant decrease was seen with AMI-SF standard and extended scoring (p < 0.001).

Behavioural data

Although no behavioural data were collected during the fNIRS task, participants were tested on a similar category VF task using the COWAT and, as with the whole group, the fNIRS subgroup of patients performed significantly less well on this task than HCs (see Table 15). However, the COWAT task is not paced and therefore tests both the speed with which words are retrieved and the number of words that are able to be retrieved, unlike the fNIRS task; thus, the results may not reflect the patients’ performances while being scanned.

Haemodynamic response to the verbal fluency task in each group

Haemoglobin AUC data were available for 49–51 HCs and 16 or 17 patients depending on region (one patient did not complete the VF task).

The results for HbO values are shown in Table 17 and the mean effect of time averaged over all regions is illustrated in the group comparison in Figure 11. The repeated measures ANOVA in HCs showed a significant effect of time (F_2,96 = 7.404, p = 0.002) and a trend to a region × time interaction (F_6,288 = 2.408, p = 0.076), with changes over time in all regions. In patients there was only a trend to an effect of time (F_2,30 = 2.868, p = 0.081) because of the values tending to rise over the blocks. ANOVAs of individual regions in patients showed a significant effect in the RVF only (see Table 17).

FIGURE 11.

Time course of HbO, HbR and Hbdiff AUCs in HCs and patients during the VF task. Responses are averaged over all regions. Values are adjusted means ± SEM (covaried for age and sex). (a) HbO all regions; (b) HbR all regions; and (c) Hbdiff all regions.

For HbR (Table 18 and see Figure 11), in HCs there was a trend to an effect of time (F_2,96 = 3.116, p = 0.055) and a significant region × time interaction (F_6,288 = 2.770, p = 0.037); a decrease in AUC was seen between block 1 and block 2 in all regions, with an increase in block 3 apart from in the LVF where it decreased further. All regions had significant (LDV, RVF) or trend significant (LVF, RDF) changes over time. In patients, although mean AUC values increased in all regions between blocks 2 and 3, no significant main effects or interactions were found.

For Hbdiff values (Table 19 and see Figure 11), in HCs there was a highly significant effect of time (F_2,96 = 10.501, p < 0.001) with no significant effect of region or region × time interaction. The effect of time was due to a peak in block 2 seen in all four regions, confirmed by the post hoc regional analysis. In patients, although mean values increased over the blocks, no significant effect of time was seen in the ANOVA, although when regions were examined separately there was a trend in the RVF.

Correlations between performance and mood and haemodynamic responses

No significant correlations were found in either all participants taken together or each group separately between COWAT category fluency and block 2 AUC values (chosen as the period encompassing the maximum response; see Figure 4). For MADRS scores a negative correlation was found in patients for the baseline block 2 Hbdiff AUC (r = –0.529, p = 0.035) in the LVF region (i.e. more severe depression was associated with lower responses). The corresponding correlation for the baseline block 2 HbO AUC in the LVF was in the same direction but not significant (r = –0.411, p = 0.11).

Group comparison of the time course of the haemodynamic response at baseline

Data were available in all regions for 49 HCs and 16 patients.

The ANCOVA of HbO AUC values, covaried for age and sex, showed a significant group × time interaction (F_2,122 = 3.936, p = 0.022) with no other main effects or interactions. The time course for the mean AUC over all regions for the two groups is illustrated in Figure 11; patients had lower values in blocks 1 and 2 than HCs. Post hoc ANCOVA by region showed significant group × time differences in the LVF (F_2,126 = 3.531, p = 0.041) and RVF (F_2,128 = 3.575, p = 0.031).

For the HbR AUC ANCOVA no significant main effects or interactions were found (see Figure 11).

The Hbdiff AUC ANCOVA showed similar findings to the HbO analysis (see Figure 11) with a trend to a significant time × group interaction (F_2,122 = 3.023, p = 0.053) and a significant effect of region (F_3,183 = 3.295, p = 0.032) (because of higher overall mean AUC values in the LDF and RVF than in the LVF and RDF; see Table 19). Exploratory ANCOVA by region showed a significant group × time interaction in the LVF only (F_2,126 = 3.191, p = 0.048).

The effect of electroconvulsive therapy and ketamine on the time course of the haemodynamic response in patients

Data were available for 11 patients at both baseline and mid-ECT (four of whom had received ketamine).

The ANOVA of HbO AUC values showed significant interactions for ECT × time (F_2,18 = 8.240, p = 0.003) (Figure 12) and region × time (F_6,54 = 3.111, p = 0.011), with a trend for an overall effect of time (F_2,18 = 3.518, p = 0.051). There were no significant drug main effects or interactions. Post hoc ANOVA showed that all regions, apart from the RDF, had significant ECT × time interactions (i.e. the time course of the haemodynamic response was significantly different at mid-ECT compared with baseline). At mid-ECT the pattern was for lower AUC values in blocks 1 and 2 and higher values in block 3 than at baseline.

FIGURE 12.

Time course of (a) HbO, (b) HbR and (c) Hbdiff AUCs for the VF task in patients. Responses are averaged over all regions and are shown at baseline (pre ECT) and mid-ECT. Values are means ±SEM.

For HbR AUCs (see Figure 12) there was a significant interaction for ECT × time (F_2,18 = 5.187, p = 0.018) and trend significance for ECT × drug (F_1,9 = 3.612, p = 0.09) and ECT × drug × time (F_2,18 = 3.069, p = 0.073). At mid-ECT there were smaller reductions in HbR than at baseline in blocks 2 and 3. Post hoc ANOVA showed a significant ECT × time interaction only in the LVF (F_2,18 = 4.101, p = 0.036) and a significant ECT × drug × time effect in the RDF (F_2,18 = 4.244, p = 0.031). The drug interactions were largely driven by baseline effects in patients receiving ketamine (Figure 13).

FIGURE 13.

Time course of HbR AUCs for the VF task in patients receiving (a) saline (n = 7) or (b) ketamine (n = 4). Responses are shown in the RDF cortex at baseline (pre ECT) and mid-ECT. ECT × drug × time interaction: F_2,18 = 4.244, p = 0.031. Values are means ± SEM.

For Hbdiff AUCs there was a similar pattern to that seen with HbO (see Figure 12). The repeated measures ANOVA showed trends to effects of time (F_2,18 = 3.287, p = 0.061), ECT × time interaction (F_2,18 = 3.088, p = 0.07) and region × time interaction (F_6,54 = 1.928, p = 0.096). There were no significant main or interaction effects of the drug. Exploratory ANOVA by region showed significant ECT × time interactions in the LDF (F_2,20 = 3.854, p = 0.038) and RVF (F_2,20 = 4.280, p = 0.034).

In summary, at mid-ECT, compared with baseline, there was an altered time course of the haemodynamic response, with blunting of the changes in HbO and HbR during the active task. No consistent effect of ketamine was found but the numbers in the two treatment arms were very small, meaning that no conclusions can be drawn.

Behavioural data

Behavioural data from the n-back task were missing from one site for technical reasons and were available for only 20 HCs and 13 patients. Participants did not receive an n-back task as part of neuropsychological testing, but the digit span backwards provided an assessment of WM; its correlation with n-back accuracy was r = 0.39 (p = 0.025) in those receiving both. Excluding two outliers from the n-back data (one HC, one patient) increased the correlation to r = 0.57 (p = 0.001). As with the whole group, patients performed significantly less well on the tasks than HCs. The digit span backwards results are shown in Table 16; the fNIRS n-back accuracy was 87% (SD 19%) in HCs and 69% (SD 21%) in patients (p = 0.021).

Effect of the n-back task on the time course of the haemodynamic response in each group

Haemoglobin AUC data were available for 51 HCs and 18 patients.

In HCs some regions showed an increase in HbO AUC values over time, but the ANOVA was not significant (time: F_2,100 = 1.199, p = 0.31) and no significant regional effects or interactions were seen (Table 20). The ANOVA in patients showed a trend towards an effect of time (F_2,34 = 2.914, p = 0.068), reflecting decreases in mean AUC values in block 2 in all regions (Figure 14); no regions showed a significant effect of time on exploratory testing, although trends were seen in three regions (see Table 20).

FIGURE 14.

Time course of (a) HbO, (b) HbR and (c) Hbdiff AUCs for the n-back task in HCs and patients. Responses are averaged over all regions. Values are adjusted means ± SEM.

In HCs the ANOVA of HbR AUC values showed a highly significant effect of time (F_2,100 = 10.151, p < 0.001) with no main effect of region or time × region interaction. Individual ANOVAs showed significant effects of time in all regions (Table 21) because of a dip in block 2 (see Figure 14). In patients there was also a significant effect of time (F_2,34 = 5.946, p = 0.006) and significant effects of time in RVF and RDF considered individually (see Table 21).

For Hbdiff the repeated measures ANOVA in HCs showed a significant effect of time (F_2,100 = 5.388, p = 0.006) because of a peak in block 2 seen in three of the four regions (Table 22), illustrated in Figure 14, with no significant time × region interaction. Exploratory ANOVAs by region found significant effects of time bilaterally in both ventrolateral prefrontal regions (see Table 22). In patients, although mean values dipped in most regions in block 2 and increased in block 3 (see Figure 14), the ANOVA showed no significant effects, although a trend to an effect of time was seen in the RDF in the ANOVA of individual regions (see Table 22).

Correlations between performance and mood and haemodynamic responses

No significant correlations were seen with the n-back behavioural data, but the number of participants with results was relatively low. For digit span backwards, in all participants taken together there were significant correlations with block 2 AUC haemodynamic responses in LDF (HbO r = 0.39, p = 0.001; Hbdiff r = 0.37, p = 0.002) and RDF (HbO r = 0.29, p = 0.018; Hbdiff r = 0.30, p = 0.011). In HCs the digit span backwards score correlated with block 2 haemodynamic responses in the LDF only (HbO r = 0.35, p = 0.012, Hbdiff r = 0.36, p = 0.011); there were no significant correlations in patients analysed alone. In HCs the MADRS score correlated negatively with the LVF block 2 HbO AUC (r = –0.29, p = 0.037) and the LDF block 2 HbO AUC (r = –0.31, p = 0.028); negative correlations of a similar magnitude were seen in patients but were not significant.

Group comparison of the time course of the haemodynamic response at baseline

Haemodynamic AUC data were available for 51 HCs and 18 patients.

The ANCOVA of HbO AUC data, covaried for age and sex, showed a significant effect of time (F_2,130 = 6.213, p = 0.007), a trend to a significant effect of group (F_1,65 = 3.304, p = 0.074) and a strong trend for a time × group interaction (F_2,132 = 3.057, p = 0.050). In addition, there were significant effects of sex (F_1,65 = 5.887, p = 0.018), time × age interaction (F_2,130 = 3.270, p = 0.041) and time × sex interaction (F_2,130 = 4.360, p = 0.015). The trend for a time × group interaction was due to a dip in block 2 in patients, whereas in HCs the values tended to rise over time (see Figure 14). Exploratory analysis by region showed a significant group × time effect in the RVF region (F_2,130 = 3.959, p = 0.021) and a group difference in the RDF (F_1,65 = 4.202, p = 0.044).

Deoxyhaemoglobin AUC values dipped in block 2 in both groups (see Figure 14) and there was a significant effect of time in the ANCOVA (F_2,130 = 6.274, p = 0.004), with no other significant main effects or interactions, although trends were seen for both time × age (F_2,130 = 3.226, p = 0.050) and time × sex (F_2,130 = 2.513, p = 0.093) interactions. Exploratory analysis by region showed no group × time interactions.

For Hbdiff AUC values, although mean AUC values were lower in patients than in HCs, and increased in block 2 in HCs but decreased in patients, the repeated measures ANCOVA showed no significant effect of group (F_1,65 = 2.714, p = 0.104) or group × time interaction (F_2,130 = 2.034, p = 0.135) (see Figure 14). There were significant effects of time (F_2,130 = 6.208, p = 0.003) and region (F_3,195 = 3.272, p = 0.022) and also significant time × age (F_2,130 = 5.789, p = 0.004) and region × age (F_3,195 = 2.944, p = 0.034) interactions. Exploratory repeated measures ANOVA in individual regions showed a trend to a group effect in the LDF (F_1,65 = 3.411, p = 0.069) and trends to both a group effect (F_1,65 = 3.627, p = 0.061) and a group × time interaction (F_2,130 = 2.436, p = 0.094) in the RDF.

The effect of electroconvulsive therapy and ketamine on the time course of the haemodynamic response in patients

Data were available from 12 patients at both baseline and mid-ECT (five of whom had received ketamine).

For HbO data the ANOVA showed a significant effect of time (F_2,20 = 3.879, p = 0.039) and a drug × ECT × region interaction (F_3,30 = 3.117, p = 0.041), with a trend to a drug × region effect (F_3,30 = 2.485, p = 0.08). No other main effects or interactions were significant. The effect of time was due to a decrease in HbO in block 2 at both time points (Figure 15) and the drug × ECT × region interaction was driven by baseline differences in HbO between the two groups of patients, which were not present at mid-ECT (data not shown).

FIGURE 15.

Time course of (a) HbO, (b) HbR and (c) Hbdiff AUCs for the n-back task in patients. Responses are averaged over all regions and are shown at baseline (pre ECT) and mid-ECT. Values are means ± SEM.

The analysis of HbR AUC values showed a significant effect of time (F_2,20 = 4.719, p = 0.021) and a trend to a region × drug interaction (F_3,30 = 4.607, p = 0.08) but no other significant main effects or interactions.

The ANOVA for Hbdiff AUC data showed a significant ECT × time interaction (F_2,20 = 4.898, p = 0.019) with no other main effects or interactions. The time course for the mean AUC at baseline and at mid-ECT is illustrated in Figure 15. Post hoc analysis by region showed significant ECT × time interactions in the LDF (F_2,20 = 3.703, p = 0.043) and RVF (F_2,20 = 6.373, p = 0.007) with trends seen in the other two regions (p < 0.09).

In summary, at mid-ECT, compared with baseline, a significant altered time course was seen only for Hbdiff during the active task, which may be a measure of oxygenation; no consistent effect of ketamine was found but the numbers in the two treatment arms were very small, meaning that no conclusions can be drawn.

Electroconvulsive therapy experience

Those surveyed reported a wide range of responses to ECT, with some people reporting that they felt it had changed, or saved, their life and others reporting that they did not feel any benefit; some respondents had felt initially better but had since required further ECT treatment. Only a few people reported that the experience of the ECT itself was unpleasant, with most people feeling that they would have preferred not to have had it but that they had not been particularly upset by it. Interestingly, almost all respondents focused on the practical arrangements of the treatment, with much discussion about travel arrangements, and, in particular, there was considerable strength of feeling expressed around not wanting to be in hospital. This was due to the disruption that it would cause to people, both emotionally and practically. It was also noted that support was still needed from a friend or a relative as an outpatient:

Because I didn’t want to go into hospital, we tried to do it as an outpatient and I just found . . . I found because I had to include people I didn’t want to . . . because I live on my own, I just found the whole thing too . . . I know it sounds silly, but too stressful and I just didn’t go again.

About half of the respondents reported memory problems after ECT treatment, with most saying that the problems were temporary and that their memory returned quite quickly after treatment. However, two people said that they felt that their memory was still poorer than before ECT treatment. Some people commented that their memory was poor anyway, which they attributed to their condition at the time, and so they struggled to assess whether or not their memory had been affected by ECT. One respondent felt frightened by this, saying:

Again, this is a little bit difficult, because my memory’s quite difficult from around that time anyway . . . so it’s very difficult thinking back to a lot of that time now, because there are so many blanks in it . . . I think a lot of what I was frightened of was the blanks in my memory, not being able to remember things. I think I found that quite difficult.

Electroconvulsive therapy information

There was a range of experiences and opinions when asking people to talk about the information they received about ECT, its timing and their decision to undergo the treatment. However, none of the respondents reported feeling rushed into having ECT or felt that they had not received sufficient information. Those that could remember the time before ECT generally recalled receiving a lot of information and a doctor being available to answer questions.

Opinions about ECT, before it had been suggested to them as a treatment, varied considerably and included the opinions of family and friends; some reported feeling that it was an old and outdated treatment, ‘from the Dark Ages’, whereas others knew about ECT either from their own profession or through the professions of people around them and felt it to be an effective treatment. Some people had received ECT previously and wanted the treatment again as they had found it useful in the past. However, the most striking responses came from people who had not previously had ECT, with the most common response when it had been suggested being fear:

 . . . but it can impair permanently the function of the brain and it does harm the brain.

I think the fear about the treatment really was that it might affect my memory.

Fear, I think.

I think it’s always come across as being quite scary, sort of having an electric shock.

Well I was frightened.

I kind of felt it was a bit of a scary last resort type treatment.

So I was a bit scared of it, but when I thought it’s kind of reached the point where it’s the last resort almost.

As demonstrated by these last two quotations, several people felt that they were so ill that they had reached a point where they were willing to try anything that might help.

The following respondent felt that she wanted to try something because her illness was so severe:

Because I was so ill and no drugs seemed to be working, I was keen to try anything . . . but when you’re so ill and you think nothing is ever going to be right, and you want to die, and you think anything will help, so yes, I was quite happy about it [ECT].

Study information

Most of the respondents could remember being given information about the study but a small number reported not remembering because they were unwell at the time. One person said, ‘when I got this letter about . . . the survey, I thought I didn’t know I did a research programme, but I found my leaflets and realised, oh I did!’

All those who did remember the information provided recalled it as being set out well and feeling that they understood what the study was about. The most reported reaction to being asked to participate in the research was altruism; over half of respondents recognised the possible benefit for future patients undergoing ECT, with many feeling that participating in research is important for that reason.

None of the respondents felt concerned about the possibility of receiving ketamine, with the majority either having a knowledge of the clinical use of ketamine – ‘my son’s an anaesthetist, so I sort of valued his opinion on taking part’ – or feeling that they trusted the design of the study and the professionals involved – ‘well, I assumed it wouldn’t be harmful, so I was quite happy to go along with it, yeah’.

A number of respondents felt hopeful that they would receive the ketamine and that it might help their experience of ECT; the only people who reported any concern about not knowing which drug they would receive were those who wanted to be sure that they were getting the study drug, as exemplified by the statement: ‘I would have felt happier knowing that I was getting the ketamine, obviously, rather than a placebo’.

Experience of the research study

None of the respondents reported negative feelings about being approached about the research, with the majority having no particular feelings either way. In terms of taking part in the research, respondents reported feelings ranging from neutral (‘I think I was relatively OK with it’; ‘Neutral I guess. I wasn’t bothered’) to very positive (‘I was pleased to be able to help, so it was a positive and good feeling taking part’; ‘I felt good to think I was doing something useful’).

None of the respondents reported feeling concerned about the research activities and the most frequently reported feelings about the neuropsychological assessments were around wanting to know what the assessments showed; for example, this person looked to the assessments to monitor any improvement:

Fine really. I didn’t mind giving them. I think in a way it was good to like see whether you were improving or . . . .

Another respondent found them interesting in a different way:

They were fine. I found them interesting, because it was obvious from mine, that I had a mind that sort of had a number slant. I could remember numbers very easily, but not so much words . . . . The drawings from memory, I was absolutely atrocious on.

Nearly half of the respondents, without responding to a specific prompt, highlighted the research staff who carried out the assessments; all reported very positive experiences using terminology such as ‘brilliant’, ‘very nice and friendly’ and ‘very helpful’.

Life since electroconvulsive therapy

This theme included a broad range of experiences. Some respondents had had a further course of ECT since participating in the study and others felt that the treatment had not been successful at all, saying things like, ‘I’m still battling depression . . . I don’t think it’s been as effective as I thought it would have been’. Others felt that ECT had worked for them, with one person saying, ‘I’ve been very lucky’. Other people did not feel an immediate effect from the treatment; for example, one person stated that:

It was strange for me, because I had 13 of them and, at the end, I didn’t actually feel that [I] was much better. It was 2 days later; it was like a light switch went on . . . I thought nothing’s changed. But then, 2 days later, I remember waking up – it was 11 April.

The majority of respondents were still accessing mental health services following their ECT treatment, with most mentioning that they see their psychiatrist, some as often as once per month but others as little as twice a year. Most people felt that they had some support in the form of community teams or psychiatric nurses. A number of people reported accessing psychological therapy and feeling supported, with the most positive outcome being from a woman who was due to end therapy shortly: ‘I’m delighted. They’re thinking of discharging me in a couple of weeks’.

In contrast, a number of people also felt that they had only had medication offered to them and did not know about other services. One person said:

I’ve got a mental health worker who contacted me, but she didn’t offer me anything . . . I’d like to access any help that’s there, but I don’t know what’s there, because no one has ever told me.