How Medicines May Help Prevent Depression Returning After Electroconvulsive Therapy: A Nationwide Study

This is a nationwide observational study using data from national Swedish registers. Data regarding ECT procedures and outcomes were gathered from the Swedish National Quality register for ECT (Q-ECT). Data for the Q-ECT are gathered prospectively. During 2013 the Q-ECT had an inclusion rate of 85% and during 2019 93%. [20, 21] Data from the Q-ECT have previously been validated and shown up to 95% accordance when compared to patient records. [22] Data on comorbidities and pharmacotherapy were gathered from the National Patient Register and the National Prescribed Drug Register, respectively. A study protocol with included variables, outcomes and definitions was pre-registered. [23] This study was approved by the regional ethical review board in Uppsala, Sweden, and the Swedish Ethical Review Authority. No informed consent was required for this register-based study using anonymized data. The report of this study adhered to the STROBE guidelines for cohort studies. [24]

All patients 18 years or older who were treated with ECT for MDD and included in the Q-ECT 2013-2019 were considered for inclusion. The following diagnoses, according to the International Classification of Diseases, Swedish version 10 (ICD-10), were included: F32.1 Moderate depressive episode, F32.2 Severe depressive episode without psychotic symptoms, F32.3 Severe depressive episode with psychotic symptoms, F33.1 Recurrent depressive disorder, current episode moderate, F33.2 Recurrent depressive disorder, current episode severe without psychotic symptoms and F33.3 Recurrent depressive disorder, current episode severe with psychotic symptoms. Only patients receiving an index series, defined as two or more treatments per week, of ECT were included, patients with continuation or maintenance treatments were not considered for inclusion. For our primary analysis, only patients with distinct response to treatment were included. Distinct response was defined as a score of 1 (1=very much improved) on the Clinical Global Impressions – Improvement scale (CGI-I). [25] In a secondary analysis, we included all patients that responded to treatment. This was defined as a CGI-I score of 1 (very much improved) or 2 (much improved). The rationale for using CGI-I = 1 in the primary analysis was to define a population most closely approximating clinical remission. This decision was made a priori, based on the assumption that relapse prevention strategies are most relevant and interpretable among patients with the most distinct treatment response. If a patient received several ECT series during the study period, the first series meeting the criteria of response was used.

Patients with at least one visit listing an ICD-coded diagnosis in the National Patient Register of schizophrenia, schizoaffective disorder, psychosis, bipolar disorder, or epilepsy within 5 years before ECT were excluded (see Supplement for ICD-10 codes).

Exposure was defined as prescriptions dispensed during the last week of the ECT-series, and up to 4 weeks after the ECT series end, of the following drugs or drug categories: selective serotonin reuptake inhibitors (SSRI), serotonin–norepinephrine reuptake inhibitors (SNRI), mirtazapine, tricyclic antidepressants (TCA), other antidepressants (agomelatine, bupropion, mianserin, and vortioxetine), antipsychotics, lamotrigine, and lithium. Patients with no filled prescriptions during the 4-week period post-ECT but who were dispensed at least one of the drugs within 4 weeks before the start of ECT were considered exposed to the same drugs post-ECT. Data on dispensed prescriptions was gathered from the National Prescribed Drug Register and is recorded using Anatomical Therapeutic Chemical (ATC) codes (seefor specific ATC-codes). To prevent immortal time bias, patients were considered exposed only from the date of their first dispensation within the exposure window. If an outcome occurred before any such dispensation, the patient was classified as unexposed for that episode. supplementary material

The primary outcome was relapse defined as a composite outcome of suicide, intentional self-harm, psychiatric hospitalisation or renewed ECT within a year after response to ECT. Suicide was defined using ICD-coded data from the National Cause of Death Register including unknown unnatural causes of death. Intentional self-harm was defined as any hospital admission with a primary or secondary discharge diagnosis of intentional self-harm. The following ICD-10 codes were used to define completed and attempted suicide: X60-X84 and Y10-Y34. In this study, the term intentional self-harm refers to hospital admissions for intentional self-harm or self-inflicted injuries of undetermined intent, as defined by ICD-10 codes X60–X84 and Y10–Y34. We acknowledge that these codes may include events with varying levels of suicidal intent. Psychiatric rehospitalisation was specified as any psychiatric rehospitalisation for any cause for a minimum of one calendar day, which was gathered from the National Patient Register. Renewed ECT was defined as a minimum of one ECT session administered as an additional series, based on data from the Q-ECT.

The Swedish self-rated version of the Montgomery-Åsberg Depression Rating Scale (MADRS-S) was available in the Q-ECT register for a subset of patients. [26] The scale includes nine items rated from 0 to 6, covering core symptoms of depression. Although not used as an outcome measure in the present study, MADRS-S data were used for descriptive purposes in a subsample.

Data on psychiatric comorbidities and substance use disorders was gathered from the National Patient Register. A comorbidity (from the groups: anxiety disorders, personality disorders, and substance use disorders) was present if at least one ICD-10 code was registered within 3 years before the first ECT. Considering that ICD codes are recorded in the register at the end of a hospital stay, ICD codes registered within one week after completing treatment course or at discharge after the current ECT-series were also included. See supplementary material for a full list of ICD-10 codes.

An intention-to-treat approach was used where switching or discontinuation of medications were not considered. Individuals were followed up from the end of the ECT series until the onset of the primary composite outcome (see above), death from other cause than suicide, or until a year from the end of the ECT series, whichever occurred first. Hazard ratios (HR) and 95% confidence intervals (CI) were estimated using multivariable Cox proportional hazards models associated with exposure (yes/no) to any of the drugs or drug categories defined above (see “Exposure”). An initial model was adjusted for age (continuous), sex, birth country (Nordic, non-Nordic), and education ( ≤ 9, 10–12, ≥13 years, missing). Our primary model was further adjusted for a number of different classes of psychiatric drugs dispensed within 2 years before start of ECT (continuous), number of psychiatric hospitalizations within five years before ECT (continuous), history of substance abuse (yes, no), history of psychiatric comorbidity (yes, no), psychotic symptoms during the current depressive episode (yes, no), and concomitant benzodiazepine use (yes, no), medication switch before and after the current ECT series. The proportional hazards assumption for the primary model was tested and the assumption was met (p = 0.32).

In the primary analysis, we excluded individuals (ECT-series) where CGI-I was missing. We conducted a sensitivity analysis to check the robustness of our findings in the presence of missing CGI-I using two extreme scenarios. In the first, all missing CGI-I was imputed with “1” indicating distinct response whereas in the second missing CGI-I was imputed as “3” indicating complete non-response.

Data management and analyses were conducted using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA).

Due to the mandatory nature of the inpatient portion of the National Patient Register, the National Prescribed Drug register and the National Cause of Death Register, we had complete data. However, a total of 2 726 (23.8%) patients of our initial patient selection had missing data on CGI-I. For further analysis of missing data, see sensitivity analysis.

We found 11 465 patients meeting our initial eligibility criteria. For our primary analysis, 5349(46.7%) patients were excluded due to indistinct response, 2726(23.8%) due to missing data on CGI-I; 3 390 patients remained. An additional 531(4.6%) patients were excluded due to a prior diagnosis of bipolar disorder, schizophrenia, schizoaffective disorder, or epilepsy. Our final analysis sample included 2 858 patients with distinct response (Fig. 1). Using the same inclusion and exclusion criteria for our secondary analysis, we identified 6 760 responders, defined as individuals with a CGI-I score of 1 or 2.

Table 1 shows clinical and sociodemographic characteristics for the two separate analysis, distinct responders and all responders. The average age among distinct responders was 59.9 (SD 17.4). More than half (57%) of patients among distinct responders were female and 37% had psychotic symptoms during their current depressive episode.

Prescription patterns of post-ECT maintenance medications are presented in Table 2. Among patients with distinct response, the three most common categories of psychiatric drugs dispensed were antipsychotics (39.7%), mirtazapine (38.0%) and SSRIs (35.9%). The most commonly prescribed antipsychotic drug was olanzapine (18.7%) and the most common SSRI was sertraline (14.8%).

Main findings on the association of drug exposure and relapse after ECT are presented in Table 3. Among distinct responders, 52.3% relapsed during the first year after treatment. Among patients with distinct response that relapsed, 61.8% were hospitalized, 34.0% received a new series of ECT, 3.3% presented with intentional self-harm, and 0.9% died by suicide. Among all responders, 54.4% of patients relapsed during the first year. Among distinct responders, patients who filled a prescription for antipsychotics after ECT had a 17% increased risk of relapse compared to those who did not (HR 1.17, 95% CI 1.05-1.31, p = 0.006), after adjusting for exposure to other study medications and confounders. Lithium was associated with a non-significantly reduced risk of relapse (HR 0.86, 95% CI 0.69-1.07, p = 0.17). In a secondary analysis of all responders, patients who filled a prescription of antipsychotics had an increased risk of relapse (HR 1.11, 95% CI 1.03-1.19, p = 0.005). There were no other statistically significant associations between medications prescribed after ECT and an increased, or decreased, risk of relapse.

As a part of a post-hoc analysis we saw that among distinct responders, 55.4% of patients with psychotic symptoms were dispensed antipsychotics and 30.4% of patients without psychotic symptoms were dispensed antipsychotics after ECT. Among patients with distinct response, 62.4% had available data on MADRS-S post treatment. The mean MADRS-S score among distinct responders’ post-index treatment was 8 (SD 7), where remission is defined a score of 10 or less. [26] Both those prescribed an antipsychotic drug and those who were not had the same average score. There was no statistically significant difference between these groups (p-value: 0.83). In an additional post-hoc analysis, we examined the proportion of patients who refilled a prescription for the same drug within three months of the initial dispensation. Among patients with lithium 86.9% of patients filled a new prescription within three months and among patients with antipsychotics 77.7% of patients re-filled their prescription. Data on all categories can be found in the supplementary material.

Patients with missing data on CGI-I were compared to our primary group of patients, distinct responders. Patients with missing data on CGI-I were on average younger than distinct responders (56.5 vs. 59.9 years, P < 0.001). No significant differences were observed in gender distribution. The group with missing data on CGI-I had a higher proportion of patients with psychiatric comorbidities (22.7 vs. 12.9%, P < 0.001) and substance use disorders (13.0 vs. 9.3%, P < 0.001) compared to distinct responders. Patients with missing data on CGI-I had a lower frequency of psychotic symptoms during their depressive episode (19.4 vs 37.0%, p < 0.001).

Due to the missing data regarding CGI-I, a sensitivity analysis for distinct responders with two scenarios was carried out. In the first, all patients with missing CGI-I score were assumed to be distinct responders, while in the second they were considered as non-responders. In these scenarios, we observed the following statistically significant results; in the first scenario, lithium was associated with a lower risk of relapse (HR 0.84, 95% CI 0.71-0.99) and antipsychotics were associated with a higher risk of relapse (HR 1.34, 95% CI 1.23-1.46). In the second scenario, lithium was associated with a non-significant reduced risk of relapse(HR 0.91, 95% CI 0.78-1.06) and antipsychotics were associated with a higher risk of relapse (HR 1.28, 95% CI 1.17-1.39). Full results of these two models can be found in the. In an additional sensitivity analysis, we considered all medications dispensed two months before ECT and one month after ECT as the exposure, thus widening the definition. Findings were similar in this model. Full results from this model can be found in the. supplementary material supplementary material

The aim of this study was to examine real-world prescription patterns following ECT for major depressive disorder and to explore how these patterns relate to the risk of relapse. Although there was a non-significant association between lithium and a reduced risk of relapse, we were unable to corroborate our hypothesis that lithium would be associated with a lower risk of relapse following ECT for MDD. By contrast, we found an association between dispensation of antipsychotics and an increased risk of relapse. For other medications, we observed no statistically significant associations with risk of relapse.

Among distinct responders, 52.3% relapsed within the first year after receiving treatment with ECT for MDD. The high rate of relapse post ECT is in line with previous studies, where 51% of patients relapsed within one year. [4] A previous study showed an association between lithium and lower risk of relapse in a similar patient population. [12] However, we could not replicate this finding in our study. In our study there was a trend for association between lithium and a lower risk of relapse, although this was not statistically significant. The study by Brus et al. looked at all patients that received ECT, not just responders, and that study included a total of 7 350 patients, a larger study population than ours. Thus, the non-significance of the reduced risk of relapse with lithium dispensation after ECT in our study could possibly be attributed to a lack of power. Our primary patient population, distinct responders, had a higher frequency of psychotic symptoms and a lower frequency of psychiatric comorbidity compared to the previous study’s population. In addition, the outcome was defined as suicide or hospital readmission, while our study also included intentional self-harm or renewed ECT. When considering patients who did not respond to ECT and their subsequently prescribed medications, the research question shifts from relapse prevention to treatment of ECT-resistant depression, and the latter question was not considered for our study.

In our material we found that a large portion of patients with MDD were prescribed antipsychotics after treatment. There are previous studies based on Swedish registry data examining drugs prescribed post ECT, [12] but large-scale prescription patterns following ECT in other countries are largely unknown. A previous study suggested that olanzapine may be effective for relapse prevention in patients with psychotic depression; however, the evidence is not definitive, and this potential benefit was not specifically examined in patients who received ECT as their primary treatment. [17] In our material, we observed that responders who were dispensed an antipsychotic post ECT were more likely to relapse. Increased risk of relapse after ECT in patients dispensed antipsychotics has been reported in previous studies but have then been attributed to confounding by indication and it has been suggested that patients who receive antipsychotics post ECT have a more severe disease and residual symptoms. [12, 27] Altogether, there is currently no support for the use of antipsychotics after ECT for MDD. In addition, there is a risk of side-effects of certain second-generation antipsychotics, such as weight gain and diabetes mellitus. [28]

The association of an increased risk of relapse for patients prescribed antipsychotics was persistent in both our primary and secondary analysis, in initial and final models, and three scenarios in our sensitivity analysis. In addition, our model was adjusted for several factors that account for disease severity, such as previous number of hospital admissions for psychiatric reasons, previous number of prescribed antidepressant drugs (indicating treatment resistance) and psychotic features during the current episode. Despite these adjustments, residual confounding and confounding by indication still cannot be ruled out. Prescribing decisions may also have been influenced by clinical impressions not captured in structured data, such as personality traits or perceived relapse risk. Additionally, considering that 37% of distinct responders were prescribed an antipsychotic post ECT it would be difficult to argue that these patients are exceptional, considering that they make up a large portion of the study population. Furthermore, patients who were prescribed an antipsychotic post ECT did not have a higher MADRS-S score post treatment. However, MADRS-S data were only available for a subset of patients, and the scale may not fully capture differences in symptom profiles. If confounding by indication was a particular issue with our model, one could argue that similar patterns would have been seen for patients who were prescribed lithium or TCA, as these are typically reserved for MDD patients where other treatments have failed. [29] However, lithium and TCAs are typically prescribed as continuous treatments, whereas antipsychotics may often be used for intermittent use. As such, intermittent use of antipsychotics may serve as a proxy for acute severity, and this could contribute to the observed association with relapse through confounding by indication.

Our study is the first to investigate the effectiveness of a wide range of psychiatric medications in preventing relapse among patients who have responded to ECT in a real-world setting. The real-world setting provides an opportunity to evaluate treatment strategies in a less controlled environment, enabling the inclusion of patients who would not qualify for a controlled study, in a large number of patients. Our models were adjusted for a wide range of highly detailed data from validated Swedish registries. Another strength of the study is the large, nationwide sample of patients, which increases the generalizability of the findings.

This was an observational study using real-world data, and any associations observed should be interpreted with caution. Even though our models are adjusted for several variables, there is still a risk of confounding by indication. A large portion of patients had missing data on response (CGI-I) and the group with missing data on CGI-I were different than patients with available data regarding age, psychiatric comorbidities and substance use disorder. As these patients were more frequently excluded due to missing data, this could cause selection bias. However, a sensitivity analysis was conducted for two extreme scenarios regarding patients with missing data on CGI-I. In these two scenarios, our main findings were not altered. In addition, the CGI-I scale, is a relatively coarse measure, that is not specific for change in depressive symptoms. An additional limitation is that there is no data on the indication for the medications dispensed, as an example, some patients might have been prescribed mirtazapine or quetiapine for sleeping disorders and not as a relapse prevention strategy post ECT. Furthermore, we were unable to determine whether prescribed medications, particularly antipsychotics, were intended for regular or intermittent use, which may have influenced the observed associations. Another concern is the high proportion of included patients who were prescribed SSRIs or mirtazapine after ECT. These treatments have limited evidence as maintenance strategies post-ECT and are generally considered less effective than tricyclic antidepressants in patients hospitalized for MDD. [30] This prescribing pattern may have impacted the observed relapse rates in our study. Finally, our definition of relapse, which is based on hospitalization, new ECT series, suicide or intentional self-harm, may miss cases of clinical deterioration managed in outpatient settings, such as patients experiencing increased depressive symptoms without hospitalization or further ECT. A potential limitation of our statistical model is that, due to the four-week exposure period, the exposure could in some cases occur after the outcome. To address this limitation, we conducted an additional sensitivity analysis, which yielded similar results (See Table 4 in supplementary). An additional potential limitation is that patients with lithium prescriptions prior to ECT were not excluded, which could introduce prevalent user bias through depletion of susceptibles.

The risk of relapse is high even after distinct response to ECT for MDD, despite pharmacological maintenance treatment. Even though there was a trend for association between lithium and lower risk of relapse, we could not corroborate our primary hypothesis. Antipsychotics were commonly prescribed after ECT for MDD and dispensation of antipsychotics was associated with a higher risk of relapse. Currently, there is no evidence supporting the use of antipsychotics as a relapse prevention strategy among patients who have received ECT for MDD. Thus, initiation of antipsychotic drugs after ECT for MDD should be evaluated in further studies before routine implementation in clinical health care.

The data that support the findings of this study are not publicly available due to privacy concerns and the sensitive nature of patient information. Access to the data may be granted upon request to the Swedish National Quality Register for ECT, in accordance with applicable regulations.