A comprehensive literature search was conducted in the PubMed, Medline, and Cochrane Library databases from inception to April 2023. The following medical subject heading terms and keywords were used for the database searches:

We applied no language restrictions and included studies in non-English languages, and the relevant data were translated for interpretation using the Google Translate service. The references of relevant literature were carefully checked for potentially eligible studies. Disagreements were resolved through consensus and, when necessary, through arbitration by a third researcher.

Studies that met the following criteria were included: patients aged 18 years and older undergoing ECT and receiving etomidate as an intravenous induction agent and propofol in the control group.

Additionally, we conducted a manual examination of the sources cited in published meta-analyses addressing similar subjects. We included studies regardless of language restrictions and utilized the assistance of the Google Translate service.

The primary goal is to evaluate the duration of seizures, including motor seizure duration and EEG seizure duration. We included prospective and retrospective controlled trials, including crossover trials. We excluded case note analysis, non-human studies, phase I clinical trials, case reports, editorials, abstracts, reviews, comments and letters, expert opinions, studies without original data, and duplicate publications.

Two investigators (SA and SRa) independently extracted the following information from each included study: study characteristics (first author, year of publication, sample size, and study type), participant baseline characteristics, indication for ECT, type of induction agents, dosages, route of administration, and outcomes of interest.

In cases where the data were presented as the median and interquartile range (IQR), we attempted to obtain the mean and standard deviation values from the authors through communication. However, if we did not receive a response, as a final option, we calculated the mean using a validated formula: mean = 2m + a + b/4, where “m” represents the median, and a and “b” represent the 25th and 75th percentiles, respectively (16). To estimate the standard deviation (SD), we used the formula provided by the Cochrane Collaboration: IQR = 1.35 SD (17, 18). The risk of bias in the eligible RCTs and crossover trials was assessed using the “Cochrane collaboration's tools for quality assessment” (19). The following six components were assessed: (1) random sequence generation, (2) allocation concealment, (3) blinding of participants and personnel, (4) blinding of outcome assessment, (5) incomplete outcome data, and (6) selective reporting.

As for the crossover trials, we assessed the studies in four additional domains: (1) appropriate crossover design, (2) random sequence generation, (3) carryover effect, and (4) unbiased data (20). The Newcastle–Ottawa Scale (NOS) (21) (range: 0–9 stars) was used to rate the methodological excellence of each retrospective cohort study. Three categories, namely, selection, comparability, and outcome, were used to grade the studies. A total of ≥5 stars indicated that the quality of the research was relatively high. All items were independently assessed by two investigators (SA and SS). Disagreements were resolved through consensus and, when necessary, through arbitration by a third researcher (SR).

The meta-analysis was carried out based on the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statements (22). RevMan (version 5.4; Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014) was used for all statistical analyses. The standard mean difference (SMD) with 95% confidence intervals (CIs) was calculated on pooled effects for continuous variables.

To assess potential statistical heterogeneity among trials, the Higgins I² statistics and Cochrane's Q test were used. The meta-analysis was conducted using fixed-effect modeling. Subsequently, after assessing the heterogeneity with fixed modeling, the analysis was repeated using random-effects methods. As a result, all the values reported in the present analysis are derived from random-effect modeling. We measured the degree of heterogeneity among the trials using the I² statistics. I²-values below 40% were deemed insignificant, while those ranging from 30 to 60% indicated moderate heterogeneity. High heterogeneity was represented by values between 50 and 90%, and values exceeding 75% denoted substantial heterogeneity. When the heterogeneity was high, subgroup analysis or sensitivity analysis was used to identify sources of heterogeneity. The results of the meta-analysis were visually examined by a forest plot, and the potential publication bias was shown by a funnel plot and Egger's test. Egger's test was performed using the R statistical software (23) and meta package v4.17-0 (24). A p-value of < 0.05 was considered statistically significant.

We comprehensively searched PubMed, Medline, and Cochrane Library databases from inception until April 2023 using the keywords “etomidate,” “propofol,” and “electroconvulsive therapy,” and a total of 70 articles were identified. After removing the duplicates, the articles were shortlisted first by the topics, then by reading the abstracts, and finally by way of full-text review. This left us with a total of 16 relevant articles (25–40) that were included in this meta-analysis. A summarized result of our literature search was reported in the PRISMA flowchart (Figure 1).

The baseline characteristics of the included studies are found in Table 1. Out of the included studies, two were of retrospective cohort nature (25, 31), while the rest were clinical trials (27–30, 32, 37, 40), some of which were crossover in nature (26, 33–36, 38, 39). The included articles used etomidate and propofol as interventions. In addition, some studies also included thiopental, thiopentone, ketofol, ketamine, methohexital, and methohexitone as interventions, although we disregarded their results in our analysis. Most of the studies did not mention the indication for ECT; however, those that did usually stated their demographic as depressed (35, 39) or schizophrenic (36, 38). The most reported outcomes of the studies in our analysis included hemodynamic stability (heart rate change, systolic blood pressure change, diastolic blood pressure change, and mean blood pressure), electroconvulsive seizure motor duration, and seizure EEG duration. The mean age of the patients in the included studies ranged from 30.77 to 49.6 years for etomidate and 31.34 to 62 years for propofol.

We assessed the quality of the seven RCTs and seven crossover trials using the Cochrane risk of bias tool (19), and overall, the RCTS were found to have a low risk of bias, which are comprehensively shown in Figure 2. Two out of the seven studies were deemed high quality as they had a low risk of bias in all assessing criteria, while the rest of the five studies had an unclear risk mainly in selection, attrition, or reporting bias.

The crossover studies had a low risk of bias in general, as shown in Figure 3. Two of the studies (33, 35) had a low risk of bias in all assessing criteria and were deemed high quality. The remaining five studies were graded as moderate to good quality as they had problems with only 1 to 2 items of unclear risk. The only two retrospective cohorts (25, 31) included in our study were assessed using NOS (21), and both got 6 and 7 stars, as explained comprehensively in Table 2.

All the studies included reported this outcome except these two studies (25, 32). These crossover studies by Tan and Lee (34) divided the study into two phases and performed a crossover, and this study by Avramov et al. (39) classified the patients into three groups of different doses in a double-blind manner; all these results were included in the comparison. Figure 4 shows the analysis of motor seizure duration comparing etomidate and propofol. The overall motor seizure duration with etomidate lasted longer than with propofol (SMD = 1.09, 95% CI = 0.72–1.45; P < 0.00001; Figure 4), and the P-value (P < 0.00001) for this comparison reached statistical significance. The overall heterogeneity (I²) was high (79%). A subgroup analysis was performed based on study design, i.e., RCTs, crossover, and retrospective cohort studies. In the RCTs and crossover studies subgroup, the results significantly favored etomidate over propofol (SMD = 1.48, 95% CI = 0.67 to 2.28; P = 0.0003; Figure 4) and (SMD = 0.83, 95% CI = 0.56–1.10; P < 0.00001; Figure 4), respectively. The I²-values for the comparison were 90% and 22%, respectively. However, the only cohort study showed the difference to be statistically insignificant (P = 0.55) (SMD = 0.18, 95% CI = −0.41 to 0.77; Figure 4).

Furthermore, sensitivity analysis revealed this RCT by Mir et al. (29) to be a cause of heterogeneity, and after excluding it, the RCT subgroup and overall result still showed etomidate as having significantly longer seizure motor duration than propofol (SMD = 1.06, 95% CI = 0.56–1.56; P < 0.0001; Figure 5) and (SMD = 0.90, 95% CI = 0.64–1.15; P < 0.00001; Figure 5), while the heterogeneity lowered down to 72 and 55%, respectively (Figure 5).

All the included studies reported this outcome except for these five studies (26, 28, 29, 37, 40). Among the crossover studies, Tan and Lee (34) classified the patients into two groups and conducted a crossover design. On the other hand, Avramov et al. (39) classified the patients into three groups, each receiving different doses, in a double-blind manner. The findings from all these studies were incorporated into the comparison.

The comparison of EEG seizure duration between etomidate and propofol is reported in Figure 6. The overall EEG seizure duration with etomidate lasted statistically significantly longer than with propofol (SMD = 0.98, 95% CI = 0.64–1.33; P < 0.00001; Figure 6). The heterogeneity turned out to be 69%. A subgroup analysis was performed based on study design, i.e., RCTs, crossover, and retrospective cohort studies. In the RCTs and crossover studies subgroup, the results significantly favored etomidate over propofol (SMD = 1.29, 95% CI = 0.38–2.21; P = 0.02; Figure 6) and (SMD = 1.01, 95% CI = 0.61–1.41; P < 0.00001; Figure 6), respectively. The I²-values for the comparison were 82 and 53%, respectively. While the cohort subgroup showed the difference to be statistically insignificant (P = 0.08) (SMD = 0.35, 95% CI = −0.04 to 0.74; P = 0.05; Figure 6). The I²-value for the comparison was 69%.

Furthermore, sensitivity analysis revealed this crossover trial by Tan and Lee (34) to be a cause of heterogeneity, and after excluding it, the pooled result for crossover studies and the overall result turned out to be (SMD = 0.73, 95% CI = 0.46–0.99; P < 0.00001; Figure 7) and (SMD = 0.86, 95% CI = 0.53–1.19; P < 0.00001; Figure 7) significantly longer with etomidate, while the overall heterogeneity reduced to 64% and between crossover studies completely resolved to 0% (Figure 7).

Two funnel plots for the outcomes of motor seizure duration (Figure 8) and EEG seizure duration (Figure 9) revealed that almost all of the studies shown by scattered points in the funnel plot were distributed in the middle and top of the baseline and were located in the range of the inverted funnel, which implies that there is no significant publication bias and the results from this study are reliable.

Furthermore, to confirm our assessment of publication bias, we performed Egger's test as the number of articles exceeded 10. Egger's test for regression gave an intercept value of 3.227 and a p-value of 0.08, indicating no evidence of publication bias for the studies reporting motor seizure duration; however, for EEG seizure duration, Egger's test gave an intercept value of 3.754 and a p-value of 0.015, indicating the presence of funnel plot asymmetry and the likelihood of publication bias.

Although the findings from our study significantly favor the use of etomidate as the primary drug for ECT, it is important to acknowledge the limitations of our study and exercise caution when interpreting the results.

First, a meta-regression could have been performed as the number of studies exceeded (10) the recommended threshold of 10, as advised by Cochrane (52). This would have allowed for a more in-depth analysis of the variability between studies. Even after performing subgroup and sensitivity analyses, we still detected significant heterogeneity among the studies in the comparison, which we could not eliminate.

Second, the seizure duration can be significantly affected by the dosage of the agent used. Unfortunately, a wide range of doses were employed across all the trials, potentially contributing to the high heterogeneity. Subgroup analysis could have been conducted, but there was a lack of sufficient data. Third, crossover trials, such as the one conducted by Tan and Lee (34), might have influenced the seizure duration outcomes of the agent used post-crossover. However, there is no statistical means of evaluating this hypothesis. Additionally, the site of stimulation for ECT may have affected the seizure duration, but due to a deficiency in available data, we were unable to perform a subgroup analysis. Fourth, due to a lack of sufficient data, this meta-analysis did not take into account side effects or potential complications related to the narcotics, whether they were laboratory-related, clinical (e.g., nausea and vomiting), metabolic (effect on adrenal function), or cerebral (myoclonus). Considering these factors generally influences the strength of recommendations for any intervention. Finally, a risk of publication bias was detected when conducting Egger's test for studies pooled together for the EEG seizure duration outcome. Therefore, additional original studies with larger sample sizes and clear indications of variables such as dosage and method of ECT are required to further confirm the findings of our study.