Efficacy and safety of intravenously applied caffeine augmentation in electroconvulsive therapy

Abstract

Electroconvulsive therapy (ECT) is a neuromodulation procedure that uses short electrical pulses causing a brief generalized seizure to treat depression, mania, schizophrenia, and catatonia. Achieving sufficient seizure length is crucial for maintaining therapeutic efficacy, as it influences treatment response. However, over successive sessions, seizure length becomes shorter requiring an increased energy set to generate a sufficient seizure, potentially causing more side effects. Caffeine augmentation offers a promising tool by enhancing neuronal excitability via adenosine receptor modulation. This retrospective study assessed the effects of intravenous caffeine citrate (200 mg) on seizure quality during ECT. Data from 276 ECT sessions across 69 patients at the University Psychiatric Clinics Basel were analyzed. We examined seizure indicators—EEG seizure length, peak heart rate, and energy set— by comparing the last two ECT sessions before caffeine administration with the first two sessions after, controlling for anesthetic effects and their interactions. Caffeine augmentation significantly increased EEG seizure length (t (67) = 8.20, p < 0.001, d = 1.23) and significantly reduced the required energy (t (67) = 4.42, p < 0.001, d = 0.65). There was no significant change in peak heart rate (t (45) = 1.66, p = 0.17, d = 0.17) or other seizure quality metrics or side effects. Type of anesthesia did not affect outcomes, except for energy set where we observed a significant interaction—etomidate combined with caffeine resulted in a smaller energy increase compared to propofol (F (198) = 0.04, η² = 0.04). Our findings suggest that intravenous caffeine augmentation safely enhances seizure length and slows energy set increases in ECT without affecting cardiovascular markers, supporting its use as an effective augmentation strategy to improve ECT efficacy.

Highlights

• •Intravenous caffeine augmentation is a safe & effective strategy to increase seizure quality in electroconvulsive therapy.
• •IV caffeine significantly increased EEG seizure duration (p < 0.001, d = 1.23) & reduced required energy (p < 0.001, d = 0.65).
• •Caffeine had no effects on peak heart rate (p = 0.17, d = 0.17) or other potential side effects.
• •Etomidate combined with caffeine resulted in a smaller increase of required energy than propofol (η² = 0.04).

1. Introduction

Electroconvulsive therapy (ECT) is a well-established treatment modality for serious mental illnesses (SMI), i.e. major depressive disorder, bipolar depression, mania, schizophrenia/psychosis spectrum disorders, catatonia, and suicidal ideations (Odermatt et al., 2025; Tor et al., 2021). Although primarily used in mental disorders, ECT has also been applied in some non-psychiatric conditions such as Parkinson's disease, refractory status epilepticus, neuroleptic malignant syndrome, and certain chronic pain syndromes (Payne and Prudic, 2009). It involves the application of short low-energy, high-voltage electrical pulses between two electrodes placed on the scalp inducing a generalized seizure, which is measured via electroencephalography (EEG) and often additionally via electromyography (EMG) on a skeletal muscle (Kayser et al., 2013). Since its debut almost 90 years ago, when Ugo Cerletti and Lucio Bini (Cerletti and Bini, 1938) introduced ECT, it has been continuously improved to enhance its efficacy and safety (Metastasio and Dodwell, 2013). Meta-analyses and data from large cohorts have repeatedly demonstrated ECT's strong therapeutic effect (Song et al., 2015), with results being 20 % more effective than tricyclic antidepressants and 40 % more effective than monoamine oxidase inhibitors (Janicak et al., 1985; Martin et al., 2025; UK ECT Review Group, 2003).

ECT's efficacy is inter alia associated with seizure characteristics such as seizure length, amplitude, and postictal suppression. A seizure length between 30 and 120 s is considered optimal for therapeutic effect and remission (Gillving et al., 2024; Sackeim et al., 1993), while longer seizures are linked to increased side effects. However, over successive sessions, the seizure length decreases despite applying the same energy set. This is thought to be due to a rising seizure threshold (Sackeim et al., 1987). To counteract the typically decreasing EEG seizure length, clinicians often increase the energy set during stimulation, but this practice is thought to be associated with a rise in side effects and their severity (Abrams, 2002a). Side effects are most common during the index ECT series, which typically consists of two to three ECT sessions per week over four to six weeks. Mild side effects such as acute nausea, headache, and myalgia are common after a session (Andrade et al., 2016). Aside from mild side effects, cognitive impairments such as short-term deficits especially in episodic memory and executive functioning during the index ECT series are among the most prominent concerns, even though they are generally transient (Semkovska and McLoughlin, 2010). These impairments often lead to patient apprehension about starting or continuing ECT treatment and, in some cases, early termination. These side effects are of particular concern when there is an increase in energy set, as it is associated with greater cognitive side effects, including retrograde and anterograde amnesia, and heightened acute physical symptoms including headaches, muscle pain, and cardiovascular changes (Andrade et al., 2016). Cardiovascular complications, historically a leading cause of morbidity and mortality in ECT, have immensely decreased due to improved monitoring and anesthesiology (Rasmussen et al., 2002). Currently, ECT carries the same mortality risk as general anesthesia (Blumberger et al., 2017; Tørring et al., 2017).

To address the challenge of more severe cognitive side effects, and potentially more intense general side effects, due to a required higher energy set, several strategies have been explored. These strategies range from reducing or discontinuing medications with anticonvulsant properties —especially benzodiazepines—or counteracting their effects with flumazenil (Zolezzi, 2016), modifying anesthetic agents such as propofol due to its anticonvulsant properties, to employing augmentation techniques like hyperventilation or caffeine administration (Loo et al., 2010). Among these, intravenously applied (i.v.) caffeine has gained particular interest because of its unique combination of efficacy, ease of administration and high i.v. bioavailability (100 %) (Abrams, 2002a; Bozymski et al., 2018; Sackeim et al., 1987).

Caffeine primarily blocks A1 and A2A receptors, which regulate neuronal excitability via metabotropic signaling, influencing the cyclic adenosine monophosphate (cAMP) pathway (Gahr, 2020). A1 receptors are found in seizure-related regions like the hippocampus, cortex and thalamus (Wall et al., 2022). A2A receptors are located in dopamine-rich areas such as the striatum where they interact with D2 receptors, reducing GABAergic tone and, thus, increasing neural excitability (Alsabri et al., 2018). It also blocks presynaptic A1 receptors in the striatum, influencing glutamate and dopamine neurotransmission, further lowering seizure threshold (Boison, 2012; Gahr, 2019). Caffeine's primary metabolites—paraxanthine, theobromine, and theophylline—also act antagonistic at adenosine receptors (Agritelley and Goldberger, 2021). Moreover, caffeine can enhance the activity of various monoamine transmitters, including dopamine, noradrenaline, and 5-hydroxytryptamine, leading to stimulatory effects on the central nervous system (CNS) (Malviya et al., 2023). Animal studies have shown that very high doses of i.v. caffeine (400 mg/kg) increase cortical excitability and can even directly trigger seizures (Chu, 1981; van Koert et al., 2018). Caffeine increased the neural excitability when combined with other convulsing agents and reduced the efficacy of several antiepileptic drugs. The increased neural excitability is likely due to pharmacological interactions (e.g., receptor modulation) rather than a clear direct proconvulsant effect, which remains uncertain in humans (Coffey et al., 1990; Loo et al., 2010). It also acted antagonistically to diazepam (van Koert et al., 2018). Mechanistically, caffeine interferes with seizure termination processes by increasing intracellular calcium levels and altering potassium currents, leading to heightened neuronal excitability and synchrony (Albertson et al., 1983).

Intravenously applied caffeine has emerged as a particularly promising augmentative agent in ECT due to its rapid onset, precise timing, and predictable pharmacokinetics. Only i.v. caffeine achieves immediate therapeutic levels, reaching the brain in seconds and peaking in plasma concentration almost instantaneously — within approximately three heartbeats (Nehlig, 2018). In contrast, oral caffeine peaks much later (15–120 min), making precise timing in clinical settings challenging (Malviya et al., 2023), shows variable bioavailability (80 ± 16 %) and delayed absorption (Jaffe and Dubin, 1992). The rapid, reliable pharmacodynamics of i.v. caffeine makes it especially well-suited for ECT, where control over seizure threshold and timing is critical.

The use of caffeine as an augmentation strategy in ECT began in the 1980s (Shapira et al., 1985). Studies found that intravenous administration of 250–500 mg caffeine sodium benzoate 5 min before stimulation significantly increased seizure length in a dose dependent way —by up to 200 % with 500 mg—without requiring a higher electrical dose (Coffey et al., 1987; Shapira et al., 1987). Despite these promising findings, interest declined in the 1990s due to safety concerns, particularly regarding cardiovascular risk (e.g., arrhythmias, takotsubo cardiomyopathy) and presumed neurotoxicity based on very high-dose animal studies (Enns et al., 1996). In recent years, however, smaller clinical studies have renewed interest in caffeine, demonstrating both efficacy and tolerability. For example, Bozymski et al. (2018) reported that 250–500 mg of i.v. caffeine extended seizure length by an average of 24 s, with only modest heart rate increases and mild agitation. Similarly, Pinkhasov et al. (2016) found a 48 % increase in seizure length in two-thirds of patients using caffeine citrate. However, existing studies remain limited by small sample sizes, heterogeneity in ECT and dosing protocols, and a lack of systematic comparison across different anesthetic agents—factors known to interact with seizure threshold and quality.

In this retrospective study, we investigated the proconvulsive effects of i.v. caffeine augmentation as an add-on treatment during ECT, examining whether i.v. caffeine administration increased seizure length while reducing the increase of energy set. Our focus was to increase (or rather maintain) good-quality seizures with sufficient length throughout the treatment series in cases where seizure quality and length were not sufficient. Additionally, we looked at caffeine’s potential side effects, such as cardiac arrhythmia and headaches. We identified potential avenues for future research and discussed the implications for clinical practice.

2. Material and methods

This retrospective analysis was conducted at the Center for Affective, Stress and Sleep Disorders (ZASS) at the Psychiatric University Hospital of Basel. At our center, the use of caffeine as an augmentation strategy during ECT is part of routine clinical practice. Approval from the local ethics committee, Ethikkommission Nordwest-und Zentralschweiz (EKNZ) (2024–00410), has been obtained for the retrospective data collection, in accordance with the Declaration of Helsinki. All included participants signed a written informed consent for retrospective data analysis. Patient and treatment characteristics were thoroughly documented during ECT using the GENET-GPD software (Freundlieb et al., 2023); specifically, patient characteristics, anesthetics, caffeine dosing and time point of administration, stimulation parameters, manually assessed seizure length and qualitative evaluations of seizure expression.

We included patients treated with caffeine augmentation during their ECT index series between March 01, 2021 and September 30, 2024, as we started using the GPD software on March 1, 2021. Inclusion criteria were that patients had at least four sessions in the index series, of which at least two successive sessions were augmented with caffeine. Exclusion criteria were patients with only treatments during the maintenance series or less than four ECT index sessions, caffeine augmentation in the first two sessions of the ECT index series, missing data for the specified four time points, changes in anesthetics within the specified four sessions or patients who had no written consent for data analysis.

For each participant, four ECT sessions were analyzed: the last two sessions before caffeine administration and the first two sessions after caffeine augmentation. Choosing two sessions before and two after caffeine administration enhances data stability by averaging values across the two sessions. The temporal close proximity to caffeine intake allows to isolate caffeine's acute effects as it reduces the influence of natural variability in the measured parameters that occur independently of caffeine augmentation during the course of ECT (e.g. energy set will most likely continue to increase after several sessions). This mirrored design further allows for within-subject comparisons of each outcome variable across four successive time points.

2.1. Patients

After applying the inclusion and exclusion criteria described in the Material and methods section to our dataset of 117 patients, we included 69 patients with a mean age of 52 years and 40 females (see Table 1).

Table 1.

Sociodemographic characteristics of patients at baseline.

Baseline characteristic
Gender
Female (n (%))	40 (58.8)
Male (n (%))	29 (41.2)
Age in years (mean (SD))	51.9 (15.6)
Body weight in kg (mean (SD))	75.0 (20.7)
Diagnosis (n (%))∗
Schizophrenia (F20, F29)	5 (7.4)
Schizoaffective disorder (F25)	3 (4.3)
Bipolar affective disorder (F31)	7 (10.1)
Major depressive disorder (F32, F33)	53 (79.4)
Anesthetic (n (%))
Propofol	46 (67.6)
Etomidate	22 (32.4)
Remifentanil	69 (100.0)
Mean Treatment number of first caffeine augmentation	7.5
Number of total index treatments (mean (SD))	13.2 (6.7)

Note. (SD) = standard deviation. ∗ according to ICD-10 diagnoses.

2.2. ECT

We administered ECT using the Thymatron System IV (Somatics, LLC, Lake Bluff, IL, USA) with the double dose setting, which delivers a brief pulse stimulation at a constant 900 mA for up to 8 s. We placed the electrodes in the standard bifrontal position. These settings were not changed throughout the treatment. Seizures were induced under anesthesia using either propofol (1–1.5 mg/kg) or etomidate (0.15–0.3 mg/kg) in combination with Remifentanil (typically target-controlled infusion (TCI) level of 6 μg kg⁻¹), and Succinylcholine (1.5–2 mg/kg) and in rare cases Rocuronium (0.6 mg/kg) as muscle relaxant. The initial energy set was selected based on the predictive ‘age = dose %' stimulation strategy, according to Abrams (2002b). Two channel EEG was recorded using bilateral frontomastoidal electrode placement.

2.3. Caffeine augmentation

We administered 200 mg caffeine citrate intravenously, using a caffeine citrate solution (100 mg/5 ml; equivalent to 10 mg caffeine base per ml; water as solvent for injection). We administered caffeine on average 43.7 s before stimulation (range: 15.2–168.5 s). The decision to augment with caffeine was made by the treating psychiatrist based on the quality of the most recent seizure(s). Caffeine augmentation is one option discussed when seizure quality is considered insufficient or re-stimulation is required. Other strategies commonly considered include the use of etomidate as anesthetic instead of propofol or to increase the energy setting. In weekly supervision meetings, each case is discussed and reviewed by the clinical team to promote consistency in clinical decision-making and to ensure that all relevant factors are carefully considered by the entire treatment team.

2.4. Main outcomes

We compared the following ECT seizure quality measures from the first two seizures with caffeine to the last two seizures without caffeine:

• •EEG seizure length: The duration of the recorded electroencephalogram (EEG) seizure.
• •Energy Set: The percentage of the total available energy delivered (charge applied) during the ECT procedure (up to 200 % available). At 100 % energy, the system delivers up to 99.8 J (504 mC) at a resistance of 220Ω.
• •Peak heart rate: The highest measured heart rate achieved during the ECT seizure.
• •Seizure Evaluation: quality of each seizure rated by the ECT practitioner on a Likert scale from 1 to 3, i.e. 1 = insufficient, 2 = sufficient, and 3 = optimal. The evaluation was based on multiple parameters, including seizure length (with seizures under 25 s generally rated as insufficient), fast EEG amplitude recruitment after stimulation, general high amplitude, hemispheric symmetry, rapid and complete postictal suppression, and vegetative responses such as heart rate increase. The seizure evaluation directly guided subsequent treatment: a rating of 3 usually required minimal or no changes for the next session; a rating of 2 often led to increased energy, caffeine addition, or switching to etomidate; and a rating of 1 typically resulted in direct re-stimulation unless contraindicated by anesthesia or patient factors such as high blood pressure.

2.5. Secondary outcomes

For secondary outcomes, we conducted exploratory analyses to investigate the effect of caffeine augmentation on these common seizure quality measures:

• •Postictal suppression index: A seizure quality indicator, which measures the degree of EEG flattening immediately following a seizure, with higher PSI values indicating a more abrupt cessation of ictal activity.
• •Maximum sustained coherence: quantifies the highest level of brainwave synchronization maintained over time, reflecting the stability and coordination of neural activity during the seizure
• •Midictal amplitude: a seizure quality indicator, which measures the average voltage of EEG waveforms during the middle portion of the ictal phase, with higher amplitudes reflecting greater seizure intensity and cortical engagement.
• •
  Side effects:

  • o
    Headaches: To explore the occurrence of headaches, we examined whether patients had received paracetamol at any of the four included treatments by searching for the term “paracetamol” across all medication-related variables in the GENET-GPD output.
  • o
    Heart rate: To analyze potential side effects of caffeine on heart rate as a proxy for cardiac side effects, we examined in the GENET-GPD output whether patients received a beta-blocker, specifically Metoprolol and Bisoprolol, at any point during their treatment. Additionally, we reviewed ECT protocol notes for any documented mentions of tachycardia or arrhythmias.

2.6. Design and data analysis

Descriptive statistics, including means, standard deviations, and percentages, were calculated for all relevant variables.

For each main outcome, the mean of the two pre-caffeine sessions was compared with the mean of the two post-caffeine sessions (i.e. post starting caffeine augmentation, thus the first two sessions with caffeine) using paired t-tests. Effect sizes were calculated using Cohen's d. To examine whether observed differences were influenced by the anesthetic used (propofol or etomidate), we conducted independent t-tests comparing outcomes between anesthetic groups. Given that the analysis of the secondary outcomes was exploratory, effect sizes (Cohen's d) are reported where applicable. Side effects, also considered secondary outcomes, were not directly assessed but inferred from added medication for reported or observed side effects (e.g. analgesics, antiemetics, antihypertensives, etc.) and were therefore described in absolute numbers and percentages.

For the variable “energy set,” we calculated differences between the two sessions before caffeine augmentation and the two after (e.g., Session 2 – Session 1, Session 4 – Session 3) and compared these using paired t-tests. Additionally, to explore whether anesthetic type moderated these changes, we applied a linear mixed-effects model with anesthetic (propofol vs. etomidate) as a between-subjects factor and time as a within-subject factor (4 sessions), including the interaction (anestheticXtime), i.e., “energy_set ∼ anesthetic ∗ timepoint + (1 | patient)”. Fixed effects from this model were evaluated using Type III ANOVA. This approach was chosen specifically for the variable “energy set,” as we were interested in the change between sessions making slope-based comparisons more meaningful. To account for individual variability and explore anesthetic-time point interactions, we conducted the corresponding post hoc analyses.

Distributional shifts in paired categorical outcomes of seizure evaluation were tested using the Stuart–Maxwell test for marginal homogeneity, which generalizes McNemar's test to outcomes with more than two categories.

All model assumptions—including normality, independence, and absence of influential outliers—were assessed and confirmed via visual inspection (e.g., Q–Q plots, residual diagnostics). A significance threshold of p < 0.05 was applied to all tests, and false discovery rate (FDR) correction was used to adjust for multiple comparisons.

We performed data analysis on patient and treatment characteristics with the R statistics environment (v4.3.3; R Core Team, 2024). Data cleaning, manipulation, and summary statistics were conducted primarily using the tidyverse and dplyr packages (Wickham et al., 2019). For statistical modeling and hypothesis testing, we employed lme4 for mixed-effects models (Bates et al., 2015), car for regression diagnostics (Fox et al., 2024), and lsr (Navarro, 2021) and effsize (Torchiano, 2020) for effect size estimation.

3. Results

3.1. Main outcomes

3.1.1. EEG seizure length

We found a significant increase in mean EEG seizure length between the last two seizures without caffeine and the first two seizures with caffeine augmentation (t (67) = 8.20, p < 0.001, d = 1.23). The mean difference in EEG seizure length was 22.29 s (95 % CI: 16.86 s–27.71 s), indicating a substantial increase in seizure length following the intervention (see Fig. 1). The administered anesthetics had no significant effect (t (66) = 1.21, p = 0.23, d = 0.31). Analyzing the occurrence of exceedingly long seizures (>120 s), we identified 2 seizures (1.5 %) prior to caffeine administration and 5 seizures (3.8 %) afterward, one of which exceeded 150 s.

Fig. 1.

EEG seizure length, difference in energy set and peak heart rate before and after caffeine augmentation. Note. The difference in energy set represents the change in the energy set, expressed as a percentage (%) of the maximum applicable energy set. EEG seizure length is measured in seconds (s), and peak heart rate is recorded in beats per minute (bpm). ∗∗∗p < 0.001.

A further noteworthy result of the EEG seizure length was that 89.7 % of seizures following caffeine administration lasted longer than 30 s, compared to 51.1 % of seizures before caffeine administration.

3.1.2. Change in energy set

We observed a significantly smaller energy set increase between the two sessions with caffeine augmentation (session 3 to session 4) (t (67) = 4.42, p < 0.001, d = 0.65) compared to the two sessions without caffeine augmentation (session 1 to session 2). The mean difference in energy set increase between the two sessions without caffeine augmentation was 12.6 %, compared to 7.06 % between the two sessions with caffeine augmentation. A linear mixed-effects model with anesthetic (between-subject) and the four time points (within-subject) as factors revealed a significant interaction between anesthetic and time point regarding energy set (F (198,6) = 0.04, η² = 0.04), implicating that the energy set increased differently depending on the type of anesthetic used. However, after FDR correction, the effect was not significant anymore, but a statistical trend (p = 0.06, see Fig. 2). Post hoc comparisons revealed a significantly weaker energy set increase after caffeine augmentation when combined with etomidate than when combined with propofol.

Fig. 2.

Energy Set at each Time point under propofol or etomidate as administered anesthesia. Note. Time point 1 & 2: without i.v. caffeine augmentation. Time point 3 & 4: with i.v. caffeine augmentation. ∗p < 0.05.

3.1.3. Peak heart rate

There were no significant differences between peak heart rate means (t (45) = 1.66, p = 0.17, d = 0.17) (see Fig. 1). Anesthetic effects were not significant (t (44) = 1.23, p = 0.23, d = 0.41).

3.1.4. Seizure evaluation

Overall, the number of seizures rated “optimal” (seizure evaluation score of 3/3) clearly increased after caffeine administration from 66/134 seizures rated “optimal” to 88/134 (34 %) (see Fig. 3). For etomidate, the number of optimal seizures increased from 15 to 21 (40 %), while the number of insufficient rated seizures decreased from 16 to 4 (75 %). For propofol, the number of optimal rated seizures increased from 51 to 67 (31.4 %), and the number of insufficient rated seizures decreased from 25 to 4 (84 %). The Stuart–Maxwell test confirms that this upward shift in seizure evaluations after caffeine augmentation is statistically significant (χ² (2) = 22.1, p < 0.001). Re-stimulations also declined from 64/134 before to 14/134 re-stimulations after caffeine augmentation, further reflecting this result, with reductions of approximately 80 % for etomidate (from 15 to 3 cases) and 78 % for propofol (from 49 to 11 cases), indicating a comparable effect across anesthetic agents.

Fig. 3.

Seizure Evaluation before and after Caffeine administration. Note. Seizure quality assessed by the ECT practitioner on a Likert scale from 1 to 3, i.e. 1 = insufficient, 2 = sufficient, and 3 = optimal.

3.2. Secondary outcomes

There is a small but statistically significant decrease in midictal amplitude after caffeine augmentation (d = 0.21). Analyzing any other seizure quality measures, caffeine augmentation did not have any significant effect. Maximum sustained coherence (d = 0.09) and Postictal suppression index (d = 0.06) did not show statistically significant differences when comparing the ECT sessions with and without caffeine augmentation and anesthetic effects were ruled out for each outcome.

3.2.1. Side effects

For headaches, we found that overall, 65.4 % treatments had a record of paracetamol: 61.0 % without caffeine and 69.9 % with caffeine augmentation indicating a slight increase of the occurrence of headaches under caffeine augmentation.

Analyzing potential side effects of caffeine on the heart rate, we found that 2.9 % of patients received Metoprolol and 0 % of patients received Bisoprolol during at least one out of four treatment session. When analyzed at the level of treatment sessions, Metoprolol was administered in only 1.1 % of all included treatments. Notably, none of these doses were given in the treatment sessions with caffeine administration, indicating that the caffeine augmentation had no relevant effects on the heart rate. Additionally, in the ECT protocol notes, no tachycardias or arrhythmias were reported.

4. Discussion

This study aimed to investigate the effects of intravenously applied caffeine augmentation during ECT on several seizure quality measures including seizure length, increases in energy set (difference between treatments) and peak heart rate. Our findings showed that i.v. caffeine augmentation administered shortly before stimulation effectively (1) reduced the increase in energy set during the index series, (2) prolonged the ECT seizure length, (3) did not affect the peak heart rate, and (4) increased the overall seizure quality rated by the ECT practitioner and (5) caused no additional side effects such increased heart rate, tachycardia, arrhythmias or the occurrence of headaches. Additionally, the lack of negative effects on ECT quality measures indicates that caffeine augmentation is well tolerated and does not compromise the overall quality of the procedure. Although these results were generally consistent across anesthetics, there is an interaction between anesthetic type and time point in the energy set suggesting that caffeine may facilitate seizure induction when etomidate is used compared to propofol.

Our data suggest that intravenously administered caffeine enhances seizure quality during ECT. Nearly 90 % of caffeine-augmented treatments resulted in EEG seizures meeting the recommended minimum length of 30 s, even with lower or unchanged energy doses. This is also apparent in the significant improvement of seizure quality rated by ECT practitioners. In addition, the need for re-stimulation was almost eliminated in the first two sessions with caffeine augmentation. These findings clearly support the notion that caffeine augmentation increases the seizure quality in already good seizures with insufficient length. Importantly, caffeine augmentation was associated with a smaller energy set increase, particularly when combined with etomidate. This is clinically desirable, as it may reduce the risk of dose-related side effects such as cognitive impairment, cardiovascular stress, and prolonged recovery time, while preserving therapeutic efficacy. Seizure quality as assessed by clinicians also improved: the number of optimally rated seizures (evaluation score of 3) increased from 66 to 88 after caffeine administration. These findings further underscore the potential of i.v. caffeine to improve both the quality and reliability of seizure induction, supporting its role as a simple, low-risk augmentation strategy in clinical ECT practice.

Our findings corroborate a number of previous studies reporting that intravenous caffeine reliably prolongs seizure length, facilitates completion of ECT with good clinical response, and lacks major safety concerns (Datto et al., 2002; Loo et al., 2010; Pinkhasov et al., 2016). Previous studies demonstrated that caffeine allowed for the maintenance of lower stimulus intensities, with some evidence suggesting faster and higher clinical response rates, while showing no or only minimal effects on EEG-based seizure quality measures such as seizure regularity or postictal suppression (Datto et al., 2002; Loo et al., 2010; Shapira et al., 1987). However, the underlying mechanisms remain unclear. While generally considered as low-risk, minor cardiovascular effects and occasional anxiety or agitation have been observed, though these are typically mild and inconsistent (Coffey et al., 1990; Kelsey and Grossberg, 1995; Pinkhasov et al., 2016). No significant adverse effects have been reported in any of these previous studies using appropriate caffeine doses. A recent case report by Volpe et al. (2025) even reported that a patient tolerated an unusually high dose—4000 mg of caffeine sodium benzoate (consisting of an equal mix of caffeine base and buffer-sodium benzoate at a 1:1 ratio) —without apparent adverse cardiac or cognitive effects.

Although previous studies reported slight increases in heart rate following caffeine administration during ECT (Cua et al., 2009; Jaffe et al., 1990; Loo et al., 2010), our findings could not replicate this effect when examining peak heart rate. This could also be due to methodological differences. For example, Jaffe et al. (1990) included only patients with a prior history of cardiac illness, whereas other studies employed augmentation strategies beyond caffeine alone and implemented them at varying time points—for instance, administering caffeine only after the electrical dosage had reached its maximum setting (Cua et al., 2009; Loo et al., 2010). In our sample, caffeine was not associated with a statistically significant change in peak heart rate, nor did we see a rise in the use of beta-blockers or paracetamol to manage such symptoms. However, subtle cardiovascular effects cannot be entirely ruled out. This is particularly relevant given that ECT itself typically induces transient tachycardia and hypertension (Rozet et al., 2018) as part of the sympathetic activation due to the seizure, which may on the one hand mask or interact with any minor effects of caffeine, on the other hand even provide an increased risk for such side effects. Still, our results suggest that the use of caffeine during ECT appears to be safe from a cardiovascular standpoint, particularly when looking at peak heart rate.

While some have speculated that longer seizures were associated with a higher peak heart rate (Nagler, 2013), our results suggest that seizure prolongation can occur independently of cardiovascular changes. This aligns with previous findings indicating that caffeine augmentation does not substantially affect cardiovascular parameters during ECT (Kelsey and Grossberg, 1995). Even though caffeine can enhance sympathetic activity by blocking adenosine receptors, we did not find a measurable increase in the peak heart rate. A possible explanation might be that caffeine's effect in combination with anesthesia has simply not been strong enough to affect peak heart rate or it might reflect a ceiling effect for maximum heart rate. Still, further research is needed to better understand whether and how autonomic activation relates to seizure length.

We observed no increased side effects such as anxiety, agitation, or cardiovascular problems in our sample. While earlier studies have raised concerns about these effects when using caffeine augmentation, they appear to be dose-dependent and heavily influenced by the route and timing of caffeine administration (Loo et al., 2010). Moreover, previous worries about a prolonged seizure, status epilepticus or central nervous system toxicity stem from animal studies employing extremely high doses of caffeine (Yasuhara and Levy, 1988), which do not reflect typical clinical use of caffeine augmentation. Accordingly, these side effects have not been reported in human populations receiving standard therapeutic or consumption doses. In our study, the caffeine dose (200 mg) was within the range of normal daily human consumption—approximately equivalent to 1½ double espressos—and was administered only after the patient was fully anesthetized. This timing likely reduces subjective side effects such as nervousness, stress, or anxiety, supporting intravenous caffeine application as a practical and well-tolerated method to enhance seizure length during ECT. The safety of the typical clinical caffeine dose regarding prolonged seizures is supported by our data, where we found only five out of 136 seizures that were overly prolonged (>120 s) after i.v. caffeine augmentation. Interestingly, four of them occurred when etomidate was used as anesthetic.

While we could only trace headaches retrospectively by reviewing use of paracetamol or other analgesics, Weber et al. (1997) offers valuable insights on this question: Patients who received intravenous caffeine and did not consume additional caffeinated beverages had fewer postoperative headaches (10 %) compared to those given a placebo (23 %) (p < 0.05). Conversely, patients who received both intravenous caffeine and an additional caffeinated beverage experienced more headaches (20 %) than those who received a placebo (11 %). These findings suggest that while intravenous caffeine alone may reduce postoperative headaches, combining it with additional oral caffeine can increase headache incidence.

Another important aspect to consider is how caffeine mediates its proconvulsant effect during ECT. Caffeine might affect two properties of the seizure: (a) seizure threshold, and (b) seizure termination. As we did not assess seizure threshold nor any markers for seizure termination, we cannot identify the effect of caffeine augmentation on these two properties. Nevertheless, we observed in our study that seizures were prolonged under caffeine while maintaining good electrophysiological quality, which supports the notion of a pharmacological increase of neuronal excitability due to caffeine augmentation. If so, both seizure threshold and seizure termination, were affected. Future studies incorporating threshold titration and mechanistic markers for seizure termination will be required to address this question directly.

Last, the pattern of energy set adjustments suggests a gradual learning effect among clinicians. The increase in energy set between treatment 2 to treatment 3 was already smaller than the increase between treatment 1 and 2, reflecting an account for caffeine's anticipated effect by clinicians. However, with a confirmed caffeine effect in the individual patient in treatment 3, subsequent increases became even smaller (treatments 3 to 4). Importantly, clinical ECT quality measures—postictal suppression index, average seizure energy, and maximum sustained coherence—as well as cardiovascular stability were maintained even as seizure length increased, suggesting room for improvement in energy set adjustments. The slight reduction in midictal amplitude was small and unlikely to impact seizure quality and may instead reflect a shift in cortical dynamics than reduced efficacy. Given that average seizure length nearly doubled during the first caffeine-augmented ECT session, clinicians may consider re-evaluating energy settings when starting caffeine.

4.1. Limitations and future outlook

It is important to acknowledge the limitations of this study, such as the retrospective design. Future research should focus on well-designed prospective clinical trials, such as randomized controlled studies, to further investigate the relationship between caffeine augmentation, increased seizure length, reduced increase in energy set requirements, and treatment improvement. Additionally, varying doses of caffeine should be systematically assessed, as even lower doses may be effective. 200 mg of intravenous caffeine augmentation already produces a more-than-necessary increase in seizure length. To minimize variability introduced by individual clinician preferences, future protocols should either standardize criteria for caffeine administration or delegate the decision to a single designated clinician per site.

In the context of caffeine administration during anesthesia, potential post-anesthetic side effects such as headaches warrant further attention, particularly as we did not assess them explicitly in our study. Future studies should systematically monitor post-treatment symptoms and control for individual caffeine intake to clarify whether such effects are attributable to caffeine augmentation, anesthesia, or their interaction. In particular, a thorough evaluation of the effects and potential interactions between caffeine and etomidate on ECT parameters is needed to help develop clearer clinical guidelines, as 4 out of 5 prolonged seizures occurred under the combination of etomidate and caffeine. However, our study was not designed to assess this specific interaction. Furthermore, the effect of i.v. caffeine on midictal amplitude should be analyzed in greater detail in future studies, as we found a small effect in our exploratory analysis.

A further limitation of this study is that we did not investigate treatment response or compare outcomes between patients with and without caffeine. In our sample, caffeine was mainly given when the seizures were not rated as optimal (rating of 3), and since adequate seizures are necessary for ECT to be effective, it is likely that caffeine has beneficial effects on clinical outcomes in these cases. Future studies should examine in more detail whether caffeine can also improve clinical outcomes. Moreover, future studies should examine the potential benefits of caffeine augmentation on cognitive side effects. ECT is known to negatively impact declarative memory—especially anterograde and retrograde amnesia— with most deficits being transient, but persistent autobiographical memory loss remaining a concern for some patients (Fink, 1979; Andrade et al., 2016). The effects of caffeine on cognition in the context of ECT have not been thoroughly documented. Given caffeine's known stimulant properties (adenosine receptor antagonism and potential effects on plasticity (Hanajima et al., 2019)); it is plausible that caffeine could modulate cognitive outcomes. Preliminary evidence from a nonrandomized study suggested slight cognitive benefits, particularly in memory domains, in patients receiving caffeine during ECT (Calev et al., 1993), but concerns remain that caffeine-induced seizure prolongation might worsen cognitive side effects. Overall, these complex interactions warrant detailed investigation to clarify whether caffeine could help mitigate or inadvertently worsen cognitive outcomes following ECT.

4.2. Conclusion

In conclusion, our study strongly supports the safety and clinical utility of low-dose intravenous caffeine augmentation, which effectively prolongs seizure length during ECT without affecting peak heart rate or additional ECT seizure quality indicators, while reducing or eliminating the need for increased electrical stimulus intensity. Future studies should clarify whether caffeine's primary effect is to enhance neuronal excitability, which may facilitate the occurrence of seizures, or to prolong ongoing seizure activity by modulating inhibitory networks.

Funding

This study was supported by the Gertrud Thalmann Fonds of the University Psychiatric Clinics (UPK) Basel.

Declaration of interest statement

All authors declare that they have no conflicts of interest.

Handling editor: Prof. A. Meyer-Lindenberg