Data was collected using a pre-post ECT cohort design with longitudinal follow-up assessment points of 2 to 4 weeks post-ECT, 6 months, and 12 months post-ECT. The study received ethics approval from the Research Ethics Board of St. Joseph’s Healthcare Hamilton (SJHH) and the Hamilton Integrated Research Ethics Board (ID 11052). As approved, the pre-ECT and 2-to 4-week post-ECT assessments were delivered as part of standard clinical care. The 6-month and 12-month follow-up assessments were not part of standard clinical care and participants received compensation of $\$$ 40 per assessment and paid hospital parking.

Participants were recruited from a naturalistic sample of adults receiving an acute course of ECT between September 2010 and November 2020 at the ambulatory ECT clinic at SJHH an urban academic healthcare centre affiliated with McMaster University and the second largest provider of mental health services in Ontario Canada with a large urban, suburban, and rural catchment area. Clinical criteria for referral for ECT for depression included: inadequate response to at least three prior adequate trials of antidepressant medication, or acutely suicidal; and for bipolar depression in addition to three failed antidepressant medication trials at least one adequate trial of a mood stabilizer; or acutely suicidal. Inclusion criteria: referred to SJHH ambulatory ECT clinic for treatment of unipolar or bipolar depressive episode, age 18 to 75 years, any gender, normal or corrected to normal vision, ability to comprehend and communicate in English, medically cleared for ECT treatment. Exclusion criteria: diagnosis of Major Neurocognitive Disorder (dementia), primary psychotic disorder, Intellectual Disability, Acquired Brain Injury with loss of consciousness, visually impaired or unable to communicate adequately in English for testing purposes. Patients were evaluated by an ECT psychiatrist and anesthesiologist to determine suitability for ECT before being approached to participate in this study. Eligible patients were informed about the study by the ECT nurse, and those who expressed interest were contacted by telephone by the neuropsychology administrative assistant and scheduled for their baseline assessment where they provided written informed consent administered by a research team member. They were contacted by neuropsychology staff by telephone to arrange all assessment visits. Assessments were performed pre-ECT, mid-ECT, within 2 to 4 weeks post-ECT, at 6-months and 12-months post-ECT in an ambulatory hospital setting in the Clinical Neuropsychology Services or ECT Clinic of SJHH by a licensed clinical neuropsychologist and/or clinical or research staff under the direct supervision of a licensed clinical neuropsychologist. The administrative staff randomized patients on recruitment to receive test order A or B at baseline and alternate versions of selected neuropsychological tests were administered at subsequent assessment visits (see Fig. 1 for additional information on assessments across timepoints).

Fig. 1.

Study flow diagram. Illustration of number patients approached, number and percent recruited and retained in the study protocol or lost to follow up (LTFU), timing of testing sessions over 12 months and tests administered at each time point. Abbreviations: ECT, electroconvulsive therapy; TOPF, Test of Premorbid Function; MINI, Mini International Neuropsychiatric Interview; BDI-II, Beck Depression Inventory-II; PAI, Personality Inventory; RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; SSMQ, Squire Subjective Memory Questionnaire; WHODAS, World Health Organization Disability Assessment Schedule.

ECT was administered twice weekly with a Thymatron^TM System IV ECT Instrument (40182 and 42922, Somatics, LLC, Venice, FL, USA) using bilateral temporal/bitemporal (BIL) or right unilateral (RUL) electrode placement [28] determined based on discussion between the patient and ECT physician. Most participants (82%) received BIL temporal electrode placement, 6% received RUL, 6% started BIL and switched to RUL, and another 6% started RUL and switched to BIL based on clinical decision making. Patients received either brief pulse or ultra-brief pulse ECT. The ECT psychiatrist clinically monitored response to ECT. Typical number of sessions of ECT delivered in a course of naturalistic ECT in this setting is 12. Mean number of treatments administered in this cohort was 11.1 (range = 2 to 19) with those receiving fewer sessions either showing significant response resulting in completion termination of the series or electively withdrawing from ECT before completing treatment. The length of the ECT series and adjustments to the treatment plan (e.g., electrode placement, stimulus energy, frequency), were determined by the ECT team in consultation with the ECT recipient and family. The patient’s regular outpatient treatment team were informed of the patient’s progress through the course of ECT and advised of any recommended medication changes via the weekly ECT team progress note (see [27] for more details).

Advanced Clinical Solutions, Test of Premorbid Functioning (TOPF) [29]. The TOPF is a single word reading test developed to predict premorbid general intellectual ability in adults. Participants read aloud a list of words that increase in complexity and familiarity until a discontinue criterion is reached. Scale scores (SS) with a mean of 100 and standard deviation of 10 are calculated.

Beck Depression Inventory-II (BDI-II) [30]. The BDI-II is a 21-item self-report measure used to assess severity of depressive symptoms. Participants are asked to rate each item reflecting on their experience over the preceding two weeks, with item ratings from 0 (none) to 3 (severe). Total score is calculated with severity ratings as follows: 0–13 = minimal, 14–19 = mild, 20–28 = moderate, 29–63 = severe.

MINI International Neuropsychiatric Interview (MINI) version 7.0.0 [31, 32]. The MINI is a clinician administered interview developed and validated as a brief structured interview for the major Axis I psychiatric disorders in the Diagnostic and Statistical Manual of Mental Disorders 4th Edition (DSM-IV) and International Classification of Diseases, 10th Revision (ICD-10).

The Personality Assessment Inventory (PAI) [33, 34]. The PAI is a 344-item self-report inventory used to measure psychological and emotional constructs and symptoms associated with DSM-IV diagnostic classifications and personality features. Each item is rated on a 4-point scale from ‘not at all true’ to ‘very true’. The PAI yields 22 non-overlapping scales, including validity scales, clinical scales, treatment scales, and interpersonal scales. The full length PAI was completed by participants at baseline and the 160-item short form PAI (PAI-SF) was administered at all post-ECT follow up visits. The PAI and PAI-SF yield 11 clinical scale scores: Somatic Complaints (SOM), Anxiety (ANX), Anxiety-Related Disorders (ARD), Depression (DEP), Mania (MAN), Paranoia (PAR), Schizophrenia (SCZ), Borderline Features (BOR), Antisocial Features (ANT), Alcohol Problems (ALC) and Drug Problems (DRG).

Repeatable Battery for Assessment of Neuropsychological Status® (RBANS) [35]. The RBANS consists of 12 subtests and was used to measure objective cognitive functioning in five domains: Immediate memory (List Learning, Story Memory), Delayed memory (List Recall, List Recognition, Story Recall, Figure Recall), Visuospatial Constructional (Figure Copy, Line Orientation), Language (Picture Naming, Semantic Fluency), and Attention (Digit Span, Coding). To minimize practice effects, the standardized equivalent alternate forms of the RBANS were used in repeat assessment (Form A and Form B). RBANS variables of interest included the Immediate Memory, Delayed Memory, and Visuospatial Constructional indices, the Coding subtest (measure of processing speed), Digit Span subtest (measure of working memory), and the Semantic Fluency subtest (measure of semantic retrieval and executive functioning). Standardized scores (SS; mean = 100, SD = 10) were calculated with higher scores reflecting better cognitive performance. The RBANS was selected for its clinical utility, the availability of normative data, alternate test versions, and ease of administration.

Squire Subjective Memory Questionnaire (SSMQ) [36]. The SSMQ is an 18-item self-report measure originally developed to assess memory before and after ECT. Participants are asked to think about how they were functioning “compared to before I began to feel bad and went to the hospital” and rate a variety of statements on a nine-point scale ranging from ‘worse than ever before’ (–4) to ‘better than ever before’ (+4). A total score (range: –72 to +72) is computed by summing the item responses, with negative scores indicating a subjective worsening of memory functioning.

The World Health Organization’s Disability Assessment Schedule 2.0 (WHODAS 2.0) [37]. The WHODAS 2.0 is a 36-item self-report measure developed to assess health and disability functioning in daily life across 6 dimensions: Cognition, Mobility, Self-care, Getting along with others, Life activities, Participation. Higher total scores reflect greater disability (score range: 0–48).

Statistical analyses were conducted using IBM® SPSS® Statistics Version 27 (IBM Corp., Armonk, NY, USA) with a 0.05 alpha level, two-tailed. Descriptive statistics (mean, standard deviation, percentages) were used to describe the sample demographics and characteristics. Due to the naturalistic design of the study, data is missing from various measures across time points as indicated. Missing data were excluded using pairwise deletion on an analysis-by-analysis basis and test assumptions were met unless specified within individual analyses. Data imputation methods were not used given the extent of missing data. Baseline associations between SSMQ and cognitive performance (RBANS SS) with other relevant clinical variables were explored using Pearson correlations (rp) or Spearman’s rho (rs) correlations for non-normally distributed variables (Shapiro-Wilk p 0.05)

In linear mixed model analyses, the selected covariance type and estimation model were selected based on a combination of study design (e.g., number of timepoints) and Akaike Information Criterion (AIC)/Bayesian Information Criterion (BIC) optimization. Normality of residuals was evaluated using Shapiro-Wilk test. Treatment response and effectiveness of naturalistic ECT at reducing depressive symptoms was evaluated using linear mixed modelling. For statistical analyses in the current study, post-ECT Remitters and Responders were collapsed into one group, named Responders, and compared to Non-Responders (See Table 1). A five (Time) $\times$ two (Response Group) linear mixed model for SSMQ scores using AR(1) Heterogenous repeated covariance type and maximum likelihood (ML) estimation was conducted to evaluate change in subjective cognition over time; however, the model residuals were not normally distributed (Shapiro-Wilk = 0.971, p 0.001). The linear mixed model was run excluding three outliers ($\pm$3 SD) on the SSMQ, which then passed the normality assumption (Shapiro-Wilk = 0.971, p = 0.052). Four (Time) $\times$ two (Group) linear mixed models using AR(1) Heterogenous repeated covariance type and maximum likelihood estimation were conducted for each of the RBANS SS variables of interest. When there was a significant interaction, simple main effects were conducted across Time within each response Group. When model residuals were not normally distributed, non-parametric Friedman’s Two-Way ANOVA by Ranks were performed. To examine whether post-ECT performance on the RBANS remained significantly lower than expected when compared to baseline TOPF estimated premorbid cognition, a series of paired-samples t-tests or Wilcoxon Signed Ranks tests for non-normally distributed variables (Delayed Memory, Digit Span, and Coding) were performed. Post-hoc analyses were Bonferroni corrected for multiple comparisons using a corrected alpha threshold of 0.05 divided by the number of comparisons in each analysis as specified in relevant results sections. For correlational analyses, Cohen’s guidelines for Pearson’s r and Spearman rho (rs) = 0.10, 0.30, and 0.50 are used to interpret observed effect sizes as small, medium, and large, respectively. Cohen’s d observed effect sizes of d = 0.2, 0.5, and 0.8 are interpreted as small, medium, and large, respectively. Bonferroni corrections were not applied to correlational analyses.

Absolute change scores were computed for variables that showed a significant effect of Time (i.e., SSMQ, RBANS Fluency SS, and Immediate Memory SS) by subtracting baseline scores from 2 to 4 week post-ECT scores. To examine whether changes in depressive symptoms predicted changes in subjective and/or cognitive functioning across time, a series of linear mixed-effects models were conducted using AR(1) Heterogenous repeated covariance type and restricted maximum likelihood (REML) estimation. We examined the effects of time (five levels) and $\mathrm{\Delta }$BDI change scores (continuous, person-centered change in BDI-II score from baseline) on scores on the SSMQ and RBANS SS variables of interest. To further explore whether changes in depressive symptoms predicted changes in subjective memory and cognitive performance, additional absolute change scores were computed for mid-ECT and 2 to 4 week-post-ECT BDI-II scores (Mid-ECT minus Baseline and Post-ECT minus Baseline, respectively) and mid-ECT SSMQ scores (Mid-ECT minus Baseline). Mid-ECT and post-ECT SSMQ change scores were non-normally distributed (Shapiro-Wilks p 0.05). Spearman rho (rs) correlations were conducted for these variables, showing similar results to the linear regressions; thus, only the linear regression findings are reported.

To examine whether patient’s primary diagnosis (MDD versus Bipolar Disorder) and/or presence versus absence of clinical comorbidities was associated with post-ECT changes in outcome variables of interest, a series of independent-samples t-tests or Mann-Whitney U tests for non-normally distributed variables (SSMQ, Delayed Memory, and Coding) were performed. Variables of interest included post-ECT change scores on SSMQ, BDI-II, or RBANS SS Immediate Memory, Delayed Memory, Visuospatial Construction, Semantic Fluency, Digit Span, and Coding.

Eligible patients (n = 224) were approached and 100 met inclusion and exclusion criteria, consented to research, and were enrolled in the cohort (M = 45.7 years old, SD = 11.0, 69% women). The sample had a mean level of education just above post-secondary (M = 14.4 years, SD = 2.5). At baseline, all of the cohort met criteria on the MINI for a current or past Major Depressive Episode and/or had baseline BDI-II score in the severe range (BDI-II; M = 41.0, SD = 9.4, n = 99). Primary MINI diagnoses (n = 97), were Major Depressive Disorder (n = 79; 81.4%) and bipolar spectrum disorders (n = 18; 18.9%). At least one MINI diagnostic comorbidity was identified for 62.9% of the sample with the majority being anxiety disorders (see 27 for further details). Mean TOPF estimated premorbid intellectual ability registered in the average range compared to standardized normative data. Baseline cognitive performance on the RBANS was also in the average range, but was significantly lower than expected compared to the mean TOPF SS [27].

The number of participants completing evaluations and percentage lost to follow up (LTFU) at various points is shown in Fig. 1. Participants were LTFU due to being admitted to hospital and discontinuing ECT (n = 6), 1 moved away, 1 died of other causes, and the remainder declined follow-up testing, did not respond to scheduling telephone calls, or could not be scheduled for 6- and/or 12-month visits due to coronavirus disease of 2019 (COVID-19) related research restrictions.

Treatment response and effectiveness of naturalistic ECT at reducing depressive symptoms in this heterogeneous cohort is reviewed thoroughly in a previous publication [27]. Briefly, we found that for the group as a whole, BDI-II score improved significantly by the 4th ECT session (p 0.001) and this was maintained at 2 to 4 weeks (p 0.001), 6-months (p 0.001), and 12-month post-ECT assessments (p = 0.003) [27]. At 2 to 4 weeks post-ECT (n = 64), the clinical remission rate was found to be 27% (Remitters: BDI-II score $\le$13 and $\ge$50% improvement from baseline, n = 17), with a treatment response rate of 42.2% ($\ge$50% improvement in BDI-II score from baseline; n = 27, including remitters). Non-Responders sample (50% improvement in BDI-II score) made up 58.7% (n = 37) of the cohort who completed the 2 to 4-week post-ECT assessment. These findings should be interpreted with caution in the context of a 36% 2 to-4-weeks post-ECT attrition rate. Differences between those who completed post-ECT testing and those LTFU were explored previously [27]. Briefly, those LTFU had significantly lower scores at baseline on the RBANS Immediate Memory index (p = 0.022) and the Delayed Memory Index (p 0.001), and were less likely to have received previous ECT (p = 0.005), compared to those who completed post-ECT testing.

As shown in Table 2, SSMQ score was not significantly correlated with RBANS SS on Immediate Memory, Delayed Memory, Visuospatial Constructional, Digit Span, Coding, or Fluency. SSMQ score was weakly correlated with RBANS Visuospatial Constructional index (rs = 0.20, p = 0.050) and trended towards correlation with Digit Span subtest score (rs = 0.1, p = 0.067).

As shown in Table 3, SSMQ score was significantly negatively correlated with PAI DEP (rs = –0.24, p = 0.021), STR (rs = –0.23, p = –0.032), and SUI (rs = –0.2, p = 0.049) scales, such that worse subjective memory appraisals were associated with higher levels of self-reported depression, stress, and suicidality symptoms. Delayed Memory scores were negatively correlated with ANX (rs = –0.2, p = 0.013) and BOR Features (r = –0.32, p = 0.002) scales, and Visuospatial Constructional scores were significantly negatively correlated with the ANX (r = –0.27, p = 0.009) and SOM (rs = –0.27, p = 0.008) scales, such that higher clinical symptomatology was associated with worse cognitive performance. There were no significant correlations between any clinical variables and RBANS Immediate Memory, Digit Span, Coding, and Fluency.

For subjective memory, there was a significant interaction of Time $\times$ Group, F(4,46.06) = 2.75, p = 0.039. A significant main effect of Time was found in the Responder group, F(4,24.57) = 6.97, p 0.001, and the Non-Responder group, F(4,40.02) = 3.25, p = 0.021, but post-hoc analyses revealed a different pattern of change for each treatment response group (see Fig. 2).

Fig. 2.

Change in Subjective Memory following ECT (Time $\mathrm{\times }$ Group). Change in subjective memory scores on the Squire Subjective Memory Questionnaire (SSMQ) across Time and Response Groups (Responders and Non-Responders). Baseline n = 97 Mid-ECT n = 92, Post-ECT n = 59, 6-months Post-ECT n = 30, 12-months Post-ECT n = 15. *p 0.05, **p 0.01. Post-hoc analyses for Responder group are marked with ‘_a’ and comparisons are made from Baseline to identified timepoint. Post-hoc analyses with Non-Responder group are marked ‘_b’ and comparison is made from Mid-ECT to Post-ECT. Error bars reflect standard error of the mean.

Responders’ SSMQ scores significantly improved from baseline to mid-ECT (p = 0.009), returned to levels comparable to baseline at post-ECT (p = 1.00), then significantly improved by 6-months (p = 0.034), and 12-months post-ECT (p = 0.015) compared to baseline. In contrast, the Non-Responder group mean SSMQ significantly worsened from mid-ECT to post-ECT (p = 0.020) and did not show significant differences between baseline and any subsequent time points (p $>$ 0.05). Notably, these post-hoc analyses were no longer significant after Bonferroni correction for multiple comparisons (corrected p 0.005).

Main effects and interaction effects for Immediate Memory, Digit Span, and Coding were non-significant (p $>$ 0.05). There was no significant effect of Time (p $>$ 0.05), for either Delayed Memory or Visual Constructional indices. The Semantic Fluency model showed a significant main effect of Time, F(3,36.17) = 10.25, p 0.001, and non-significant effects for Group and Time $\times$ Group interaction. Post-hoc analyses (Bonferroni corrected p 0.008) revealed that Semantic Fluency scores significantly worsened from baseline to post-ECT (p 0.001; Cohen’s d = 0.55), and improved significantly from post-ECT to 6-months post-ECT (p = 0.007; Cohen’s d= 0.43) (see Fig. 3).

Fig. 3.

Change in semantic fluency following ECT (Time $\mathrm{\times }$ Group). Change in RBANS Semantic Fluency standardized scores (SS) across Time and Response Groups (Responders and Non-Responders). Baseline n = 99, Post-ECT n = 64, 6-months Post-ECT n = 34, 12-months Post-ECT n = 19. **p 0.01, ***p 0.001 for post-hoc analysis for Time collapsed across Group. Error bars reflect standard error of the mean.

Interestingly, removing Group from the model, revealed a significant main effect of Time for RBANS Immediate Memory, F(3,42.42) = 3.76, p = 0.018. Post-hoc analyses showed that Immediate Memory scores significantly improved from baseline (n = 99) to 2-to-4-week (n = 64) post-ECT (p = 0.017), but baseline scores did not differ significantly from 6-month (n = 34) or 12-month (n = 19) follow-up. At 2 to 4 week post-ECT testing (n = 64), RBANS SS performance again fell in the average descriptor range (SS: 90–110) for all indices and subtests of interest, except for Semantic Fluency, which fell in the low average range. Post-ECT RBANS SS (n = 64) for Delayed Memory (M = 99.13, SD = 15.71), z = –3.34, p 0.001, Coding (M = 95.24, SD = 15.75), z = –4.31, p 0.001, and Semantic Fluency (M = 87.79, SD = 14.44), t(57) = 9.88, p 0.001, d = 1.297 remained significantly lower than expected compared to estimated premorbid function per baseline TOPF SS (M = 107.98, SD = 10.91). Post-ECT RBANS SS for Digit Span (Wilcoxon Signed Ranks, p = 0.032), Immediate Memory (p = 0.030), and Visuospatial Construction (p = 0.056) were not significantly different from TOPF SS after Bonferroni correction.

There were no significant correlations between baseline-to-post-ECT change in SSMQ and change in any of the RBANS change variables (p $>$ 0.05) (see Table 4).

All patients reported subjective memory concerns prior to ECT and continued to perceive their memory as worse than it was before the onset of their depressive disorder even among those whose depression responded to ECT, possibly due to some degree of recall bias and/or a persistent general negative cognitive bias associated with low mood. Importantly, subjective memory concerns did change significantly over the course of treatment. Although results of post-hoc analyses did not remain significant after statistical correction for multiple comparisons, there was a trend for subjective memory complaints to vary with treatment response. For Responders, subjective memory appraisals improved rapidly from baseline to mid-ECT in association with a rapid improvement in mood, returned to baseline at post-ECT, then improved by the 6-month and 12-month assessments. In contrast, Non-Responders’ subjective memory appraisals worsened from mid-ECT to post-ECT, with no significant improvements from baseline to any subsequent timepoints. In the sample as a whole, although rapid reduction in depressive symptoms at mid-ECT significantly predicted early improvements in subjective memory appraisals, changes in depressive symptoms from baseline to 2 to 4-weeks post-ECT did not predict changes in subjective memory during the same period. These findings are generally consistent with literature showing improvements in subjective cognition from pre- to post-ECT that is moderated by improvements in depressive symptoms consistent with the effects of mood congruent negative cognitive bias [22].

As mentioned above, it is also important to consider the nature of the subjective memory measure, the SSMQ, and its implications for interpreting the post-ECT subjective memory findings. The SSMQ requires a respondent to consider their memory functioning “compared to before I began to feel bad and went to the hospital” with options ranging from ‘worse than ever before’ (–4) to ‘better than ever before’ (+4). This requires accessing remote autobiographical memories, which are often impaired in depressed individuals who tend to experience an overgeneralization or lack of specificity when recalling autobiographical events [43, 44]. In the case of chronic depression lasting many years, this may also be confounded by cognitive decline associated with aging. Further, appraisals of current memory functioning in depressed individuals are thought to be influenced by a negative self-evaluation style. Thus, it is plausible that we are seeing an over-estimation of past abilities and under-estimation of current abilities in this clinical sample, and that this bias persists following ECT treatment. Another limitation when interpreting these findings is the high proportion of our sample that was lost to follow-up.

Despite the caveats to interpreting subjective memory scores, our findings suggest that substantial significant subjective memory complaints were present before and after ECT, even for those who saw a significant reduction or remission of depressive symptoms. Notably, these subjective memory complaints occurred in the context of stable or in some areas improved objective cognitive performance indicating that patients believed their memory was impaired even when their measured cognition had improved or registered at normative levels for their age. Beliefs about memory may continue to influence self-perception of functional capacity and mood, which then further impacts on daily memory abilities. Indeed, persistent subjective memory complaints are not uncommon following ECT [9, 22, 23]. A recent meta-analysis reported a weighted mean prevalence rate of 48.1% of patients reporting subjective cognitive complaints following ECT [22] and a systematic review on patients’ perspectives of ECT noted that rates of persistent subjective memory loss post-ECT ranged from 26% to 55% [9, 23]. Importantly, the finding that patients who responded to ECT still experienced persistent subjective memory complaints highlights the importance of evaluating subjective cognition as relevant outcomes of ECT in tandem to evaluating reduction in depressive symptoms and objective cognition.

Encouragingly, objective cognitive performance generally did not worsen following ECT, except for transient worsening in semantic fluency that returned to baseline levels by 6-months post-ECT. Cognitive performance on the RBANS remained largely unchanged following naturalistic ECT, and immediate memory showed significant improvement from baseline to the 2 to 4 weeks post-ECT. In line with these findings, a meta-analysis of ECT practices suggested that verbal fluency, along with verbal memory and autobiographical memory, tend to show a transient worsening in the short-term (i.e., 1 to 28 days) with return-to-baseline or improvement in the long-term (i.e., $>$1 month) post-ECT [10]. Another recent meta-analysis found long-term adverse effects of ECT on learning capacity, with long-term improvements in executive function and processing speed, and stability in most other cognitive domains [16].

In the current study, changes in subjective memory from baseline to 2 to 4 weeks post-ECT were not correlated with changes in performance on immediate memory or semantic fluency measures. Similarly, changes in depressive symptoms from baseline to 2 to 4 weeks post-ECT did not predict changes in any of the cognitive domains assessed during the same period. Although overall cognitive performance did not appear to worsen, it is important to recognize that before beginning ECT our patient cohort as a whole was functioning at a level slightly lower than expected compared to their estimated premorbid intellectual functioning. At post-ECT, although not significantly worse and within normative limits, participants were still performing below expectation on measures of delayed memory, working memory/executive function (coding) and semantic fluency compared to estimated premorbid score suggesting subtle persistent declines in cognition compared to premorbid functioning. Indeed, cognitive impairment is not uncommon in depressed samples [41], and has been found to persist into euthymic periods [45, 46], particularly in patients with a greater illness burden (e.g., higher number of previous episodes, longer illness duration) [47] and may be mediated by reduction in hippocampal volume [48, 49]. These volume changes are reportedly correlated with working memory/executive functioning [49]. Although ECT has been associated with increase in hippocampal volume [50], the persistent relative impairment observed in our sample could be a consequence of the pathophysiology of depression.

Importantly, another consideration is that the RBANS may not be sufficiently sensitive to capture nuanced changes in specific domains of cognitive functioning. The memory tests employed in this study looked primarily at anterograde memory, namely the ability to form new memories before and after ECT. However, studies looking at the types of memory loss reported by ECT patients indicates that impairments to retrograde memory, particularly that of autobiographical memory, may be of greater concern [9]. Although the current study did not include a measure of autobiographical memory, the test battery did include a semantic fluency task that involved searching one’s semantic memory stores for information that matches a category cue and quickly retrieving that information, assessing elements of semantic memory and executive functioning. Whereas immediate and delayed memory tasks assess the ability to encode and retrieve new information, semantic memory relies on the ability to retrieve previously learned decontextualized knowledge and this process did transiently decline during ECT.

Interestingly, both semantic memory and executive control have been implicated in autobiographical memory functioning in healthy and depressed samples [51, 52, 53, 54]. As noted, post-ECT reductions in verbal fluency (an element of semantic memory) and autobiographical memory following ECT are fairly common [10, 38]. Fortunately, impairments in verbal fluency and autobiographical memory appear to resolve fairly quickly post-ECT [10, 38]. Thus, our finding of a transient worsening of semantic fluency post-ECT that resolved by 6-months follow-up appears in line with this literature. However, patients’ subjective reports of amnesia seem to be more persistent, lasting beyond six months post-ECT [38]. It will be important for future research to examine factors contributing to persistent subjective memory loss and the potential tertiary outcomes associated with these difficulties, such as functional impairment, perceptions of ECT and risk of depression relapse. In the interim, clinicians are encouraged to be proactive in discussing and normalizing subjective memory concerns with this patient population. Psychoeducation about the various factors that influence subjective memory appraisals within a depression population, including negative cognitive bias and recall bias, should be discussed. Clinicians are also encouraged to utilize the research findings to support patients to re-appraise subjective concerns using a Cognitive Behaviour Therapy (CBT) approach. For example, using a CBT approach, clinicians may guide clients to challenge their assumptions or appraisals of poor memory (i.e., subjective impairment) against the evidence (i.e., lack of objective memory impairment in research samples, or possibly in their own cognitive test results if available) which may help patients reframe their causal attributions and gain a more balanced and less biased evaluation of their current status, which anecdotally often leads to improved mood and subjective memory outcomes. Providing patients with more extensive psychoeducation related to cognitive biases and subjective memory challenges associated with depression that are distinguished from objective challenges may also aid in shifting beliefs about the negative impact of ECT on memory and may eventually contribute to reduced colloquial stereotypes about ECT.