Optimal Frequency for Seizure Induction With Electroconvulsive Therapy and Magnetic Seizure Therapy in Nonhuman Primates

TL;DR

This study finds that a specific stimulation frequency minimizes electricity use for seizure induction in nonhuman primates, which could reduce side effects in therapies like ECT and MST.

Contribution

The study identifies the optimal frequency range for seizure induction in nonhuman primates using ECT and MST.

Findings

The lowest seizure thresholds were observed at 16 pps with similarly low thresholds at 10 and 25 pps.

Higher and lower frequencies required more pulses to induce seizures, increasing charge and energy.

The lowest frequencies (5 and 10 pps) produced the greatest ictal power measured by EEG.

Abstract

Electroconvulsive therapy (ECT) and magnetic seizure therapy (MST) are effective in the treatment of medication-resistant depression. Determining the stimulus frequency that results in the lowest seizure threshold could produce fewer adverse effects by reducing the overall stimulus intensity.

To determine the optimal frequency for seizure induction, 4 male rhesus macaques were titrated with an increasing number of pulses at fixed frequencies ranging from 5 to 240 pulses per second (pps) using ultrabrief pulse right-unilateral ECT and circular-coil-on-vertex MST. Bilateral electroencephalography was recorded to characterize the seizure expression.

The seizure threshold dependence on stimulus frequency was similar for ECT and MST. While higher frequencies required progressively shorter trains to induce a seizure, the middle frequency range was associated with the fewest pulses (and therefore the least charge and energy), with a minimum at 16 pps and similarly low thresholds for 10 and 25 pps. The number of pulses at seizure threshold increased markedly at lower and higher frequencies. The lowest stimulus frequencies, 5 and 10 pps, were associated with the greatest ictal power measured by electroencephalography.

While neither efficacy nor side effects were assessed in this study, the results highlight the significance of stimulus frequency for seizure induction, suggest efficient titration schedules that minimize exposure to the electrical stimulus, and can inform studies to assess the impact on clinical outcomes. These data can also support safety guidelines for interventions such as transcranial magnetic stimulation that must avoid seizure induction.

Electroconvulsive therapy (ECT) and magnetic seizure therapy (MST) are forms of noninvasive brain stimulation that act by generating a seizure and are effective in the treatment of depression. This study in nonhuman primates identified the stimulation frequency that minimizes the amount of electricity delivered to the brain to generate a seizure, which could potentially minimize adverse effects. While the study did not assess therapeutic efficacy or side effects, the results highlight the significance of stimulation frequency for seizure generation and suggest efficient strategies for clinical practice and future clinical trials that explore the impact of frequency on efficacy and side effects.

Electroconvulsive therapy (ECT) and magnetic seizure therapy (MST) are forms of noninvasive brain stimulation that act by generating a seizure and are effective in the treatment of depression. This study in nonhuman primates identified the stimulation frequency that minimizes the amount of electricity delivered to the brain to generate a seizure, which could potentially minimize adverse effects. While the study did not assess therapeutic efficacy or side effects, the results highlight the significance of stimulation frequency for seizure generation and suggest efficient strategies for clinical practice and future clinical trials that explore the impact of frequency on efficacy and side effects.

Figures

Illustration of the stimulus delivery configurations in a representative nonhuman primate subject. (A) Right unilateral electrode electroconvulsive therapy configuration. (B) Circular-coil-on-vertex magnetic seizure therapy configuration. Adapted with permission from Lee *et al.* ([32](#bib32)).

B](#fig2)), and therefore the seizures that were recorded were markedly longer and perhaps more robust than the average. Nonetheless, this was not true for the 10 pps condition, which produced the highest ictal power for MST; moreover, the individual seizures with highest power were recorded for the 5 pps and 10 pps conditions. Also, the differences in seizure expression between ECT and MST may be confounded by the anesthesia differences, reflected in the baseline power, although prior studies of nonhuman primates have reported differences in seizure EEG expression under the same anesthesia as

Average electroencephalography (EEG) spectral power of ictal (A, B) and postictal (C, D) period relative to baseline for electroconvulsive therapy (ECT) (left) and magnetic seizure therapy (MST) (right) for various stimulation frequencies. The bottom row within each subplot corresponds to global EEG power, and rows above it correspond to discrete EEG bands (delta through beta). Individual raw and normalized data points are shown in Figures S9–S12. pps, pulses per second.

Full text

Electroconvulsive therapy (ECT) is highly effective for medication-resistant depression (1). Manipulations of stimulus parameters, such as the pulse width, electrode placement, and number of pulses (or, equivalently, total charge), can significantly reduce side effects while maintaining efficacy (2, 3, 4, 5). Magnetic seizure therapy (MST) (6) can achieve similar antidepressant effects (7, 8, 9, 10) but with a potentially more favorable side-effect profile than ECT (10, 11, 12, 13, 14), ostensibly due to its ability to induce seizures with less electric field exposure (15). Many of the parameters associated with seizure therapies have yet to be examined for their clinical impact.

Pulse train frequencya is commonly manipulated to individualize dose (1), but its effects on seizure threshold (ST) and clinical outcomes are poorly understood. Studies suggest that low frequencies are superior for seizure induction (5). Stimuli in the range of 3 to 20 pulses per second (pps) are optimal for photic seizure induction (16), and stimuli in the range of 25 to 60 pps are optimal for the induction of after-discharges, which are considered the precursors of seizure (17). Lower frequencies (20–60 pps) appear to be more efficient than higher frequencies for seizure induction in clinical ECT and MST (18, 19, 20, 21), and titrating dose by increasing train duration results in lower ST than increasing frequency (22). However, the most efficient stimulus frequencies for seizure induction have not been mapped.

Here, we report the dependence of ST on stimulus frequency over a wide frequency range in nonhuman primates and show that there is a narrow frequency band that minimizes the ST in both ECT and MST.

Methods and Materials

This study was approved by the Institutional Animal Care and Use Committees of New York State Psychiatric Institute, Columbia University, and Duke University. We have previously reported detailed methods for nonhuman primate models of ECT and MST (23,24).

Animals

Subjects were 4 male pathogen-free rhesus macaques (Macaca mulatta). Ages and weights are given in Table 1.Table 1. Age and Weight Range and Individualized Stimulation Current Amplitude Used for All Procedures in the 4 Nonhuman PrimatesAnimalAge Range, YearsWeight Range, kgStimulation AmplitudeECT, mAMST, % MSOCH11.1–13.68.5–9.956586%DY9.9–12.410.7–12.838076%MA10.1–12.38.3–9.823066%RZ17.2–18.88.1–12.128084%ECT, electroconvulsive therapy; MSO, maximum stimulator output; MST, magnetic seizure therapy.

Study Design

Two modalities were studied: ECT with right unilateral (RUL) electrode placement and MST with circular-coil-on-vertex configuration, illustrated in Figure 1. ST was determined for a range of stimulus pulse train frequencies. One modality × frequency condition was tested per session. Each condition was repeated 3 times per animal. Sessions were separated by at least 5 days to minimize carryover effects.Figure 1. Illustration of the stimulus delivery configurations in a representative nonhuman primate subject. (A) Right unilateral electrode electroconvulsive therapy configuration. (B) Circular-coil-on-vertex magnetic seizure therapy configuration. Adapted with permission from Lee et al. (32).

Anesthesia and Monitoring

Anesthesia and monitoring followed previously described methods (23, 24, 25). For transport to the procedure room, animals were sedated with intramuscular (i.m.) ketamine (5.0–10 mg/kg) and xylazine (0.3–0.5 mg/kg) for the MST procedures or i.m. ketamine (3.0 mg/kg) and dexmedetomidine (0.075–0.15 mg/kg) for the ECT procedures. Xylazine was replaced with dexmedetomidine during the course of the study because of increasing tolerance of the animals to xylazine. An additional dose of ketamine (2.5 mg/kg) was administered if needed to maintain sedation. Anesthesia and muscular paralysis were induced with intravenous methohexital (1.0 mg/kg) and succinylcholine (3.5 mg/kg), with a second bolus of each given at the initial dosages if the procedure could not be completed prior to anesthetic or paralytic emergence. Methohexital and succinylcholine are commonly used for clinical ECT (26).

Stimulation Modalities

ECT was administered using a customized Spectrum 5000Q device (MECTA Corp.) that allowed a wider range of stimulus current parameters. The electrical stimuli consisted of 0.2 ms bidirectional rectangular pulses delivered through electrodes 2.5 cm in diameter placed in the RUL position (27) (Figure 1A). RUL electrode placement was used because it has fewer cognitive side effects than bilateral placement and is often the first-line choice in clinical ECT (3,4,28).

MST was administered using a custom-made MagPro MST device (MagVenture A/S). MST stimuli consisted of 0.36 ms cosine pulses delivered through a 10-cm-diameter coil placed on the vertex of the head (27) (Figure 1B).

Seizure Threshold Titration

ST was titrated using an ascending method of limits procedure (5,29) utilizing successive stimulus trains with an increasing number of pulses and therefore increasing train duration, while the other stimulus parameters (pulse amplitude, pulse width, and train frequency) were kept fixedb. Amplitude was set to twice the individual amplitude–titrated ST determined at frequencies of 50 pps and 500 pulses in a previous study (27) (range 230–565 mA for ECT and 66%–86% of maximum stimulator output for MST) (see Table 1 for individual amplitudes). Amplitude was selected in this manner to 1) compensate for individual anatomical variability and thus normalize the electric field exposure in the brain across subjects (5,30,31) and 2) provide electric field intensity in the brain that is comparable to that in clinical ECT as estimated by computational electric field models (15,31,32).

ECT ST was titrated at 5, 10, 16, 25, 50, 100, and 240 pps. MST ST was titrated at 5, 10, 16, 25, and 50 pps for all animals and at 100 pps for 2 of the 4 animals (MA and DY). Limitations of the MST device, namely its inability to stimulate at a high enough amplitude and frequency simultaneously, prevented titration at 100 pps in the other 2 animals (CH and RZ) and at 240 pps for all animals. MST data were acquired first, followed by the ECT data. ST data were collected from a total of 150 sessions.

We aimed to collect 3 ST estimates for each modality × frequency condition in separate sessions. Within modality, the order of the sessions with different frequency conditions was randomized. The first titration for each modality × frequency condition used coarse steps (70% increments in the number of pulses) to locate a range for the individual ST. Subsequent titrations were finer (30% pulse increments) to provide a more accurate estimate of the ST. Titration steps were separated by intervals of approximately 20 seconds. Seizure was determined by visual observation of tonic-clonic motor activity and concomitant electroencephalography (EEG) signal recorded by the MECTA Spectrum device. In instances where evidence of a seizure was ambiguous, i.e., if very brief (∼1–2 seconds) motor or EEG seizure activity was noted after the end of the stimulus, the interval was lengthened to 60 seconds before the subsequent titration step. No more than 6 pulse trains were delivered in a single session. Sessions during which the animal had a seizure on the first titration step were discarded, because it is possible that the threshold was lower than what was observed. For 5 pps trains, because the seizure can take place completely during the stimulus train, strong tonic contractions during the stimulus train followed by motor suppression on the subsequent titration step were considered manifestation of a seizure. Sessions in which no clear seizure was identified were repeated.

Seizure Characteristics

The duration of the motor seizure was defined as the interval between the beginning of the stimulus and the end of the motor seizure expression. The one exception to this definition was for the 5 pps stimulation condition, because the motor expression typically started up to tens of seconds after the initiation of the stimulus due to the slow rate of pulse delivery (see Figure S6A). Therefore, for the 5 pps condition, we defined the beginning of the seizure when the motor expression started during the stimulus train. Ten sessions for which the seizure start was not noted were excluded from the seizure duration analysis.

The visually observed strength of the motor seizure expression in the 4 limbs (separate for tonic and clonic phase) and face were rated on a qualitative scale of none, weak, medium, and strong. Electromyographic (EMG) recordings were also made for exploratory analyses (see the Supplement).

Electroencephalography

Brain activity during seizures was recorded with EEG using 2 bilateral fronto-mastoid channels through the MECTA Spectrum ECT device and software (gain = 5000, sampling rate = 140 Hz). Data processing was performed using MATLAB (version R2023; The MathWorks, Inc.). The EEG was bandpass filtered from 0.5 to 50 Hz, manually artifacted, and resampled to 128 Hz. Prestimulus baseline activity was analyzed for 30 seconds before stimulus onset. Ictal activity was analyzed between the end of the stimulus train and the end of the EEG seizure or, if that was ambiguous, the end of the motor seizure. Artifact-free postictal activity was analyzed after the end of the seizure. Discrete wavelet transform was performed using the fourth-order Daubechies wavelet (33). The EEG power was computed from the detail functions reconstructed from wavelet coefficients corresponding to the following frequency bands: delta (0–4 Hz), theta (4–8 Hz), alpha (8–16 Hz), and beta (16–32 Hz). EEG power (in units of μV^2^/Hz) was log-transformed (base 10) and averaged over artifact-free 1-second epochs that overlapped by 0.5 seconds.

EEG power analysis was carried out for 85 sessions; the rest were excluded due to malfunction of the recording computer (7 sessions), seizures shorter than stimulus duration (8 sessions, all at 5 pps), or insufficient duration of artifact-free seizure recording (50 sessions).

Statistical Analysis

The ST, seizure duration, and EEG power data were analyzed with linear mixed-effects models in JMP Pro 17 (SAS Institute Inc.) and plotted with JMP or MATLAB. Modality (ECT or MST), stimulus train frequency, and session were treated as fixed effects, and subject was treated as a random effect. Modality and subject were treated as nominal variables, and session number was treated as a continuous variable. Stimulus frequency was treated as a nominal variable to accommodate its strongly nonlinear effect on ST. The ST data, expressed as either number of pulses or train duration, were log-transformed to reduce heteroscedasticity across frequency conditions and improve normality. Statistical analysis of the EEG power also included EEG channel laterality (left, right), phase (baseline, ictal, postictal), and band (delta, theta, alpha, beta) as fixed effects. All fixed effects interactions were modeled; to allow this, the 240 pps ECT condition was excluded because it does not have an MST counterpart. False discovery rate correction was used to account for multiple comparisons. The mixed effects analysis was followed up with t tests to compare ECT and MST as well as t tests with Dunnett or Dunnett-Hsu multiple comparison correction to identify means significantly different from the minimum across stimulus frequency. Wald test was used to assess the significance of subject as a random variable. Results were considered significant at p < .05, unless otherwise noted.

Results

Seizure Threshold

Seizure induction was generally reliable at all frequencies except for 5 pps. At 5 pps, seizures could not always be induced, and the long duration and large number of the pulses led to premature termination of some titration sessions due to the animal’s emergence from anesthesia, MST coil heating, or ECT device errors.

Figure 2 shows ST versus stimulus frequency for ECT and MST. There was no significant interaction between modality and frequency (F5,121 = 0.808, p = .546), indicating that ST as a function of frequency behaved similarly for ECT and MST. The effect of session number was not significant either (F2,121 = 1.93, p = .200); this was expected because there was 1 session per week to minimize carryover effects, the frequency conditions were interleaved and randomized, and the seizure induction sessions were carried out regularly over the course of several years as part of a series of studies, so any lasting effects would have reached a steady state. Therefore, the interaction term and session number were removed from the final model. ST, quantified as the number of pulses required to induce a seizure, had a significant U-shaped dependence on stimulus frequency (F6,139 = 68.9, p < .0001) (Figure 2A). Of the tested stimulus frequencies, the mixed-effects model identified 16 pps to minimize the number of pulses for seizure induction. This minimum ST did not differ significantly from the STs at 10 and 25 pps (ts < 1.40, ps > .564), but it was significantly lower than the STs for 5, 50, 100, and 240 pps (ts > 3.79, ps < .0014). ST was affected by modality (F1,139 = 15.9, p = .0001), with MST having 18.8% lower ST than ECT, on average (t = −3.98, p = .0001). The least squares estimates of the mean ST at 16 pps were 87.7 and 71.3 pulses for ECT and MST, respectively. There was no significant ST variance contributed by the individual animals (p = .267), indicating that STs were consistent across animals.Figure 2. Seizure threshold (ST) and seizure duration for electroconvulsive therapy (ECT) (orange) and magnetic seizure therapy (MST) (blue) as a function of stimulus pulse train frequency (pulses per second [pps]). (A) ST expressed as the number of pulses needed to induce a seizure at each frequency (stimulus total charge and energy are directly proportional to the number of pulses). (B) Duration of the stimulus pulse train at ST (filled markers) and respective observed motor seizure duration (open markers). Markers and whiskers indicate the mean and SD of the animal averages. The average for each animal was computed across all titration sessions at each frequency. Asterisks (∗) mark levels significantly different (p < .01) from the respective minimum across frequencies. All axes are logarithmically spaced. Individual data points are shown in Figures S1–S3.

Another way to quantify ST is by the duration of the stimulus train, which is equal to the number of pulses divided by the train frequency. As illustrated in Figure 2B, the train duration at ST decreased monotonically with increasing frequency (F6,139 = 378, p < .0001). The range of least squares means spanned 63.4 seconds for 5 pps to 1.50 seconds for 240 pps for ECT. The mean train duration at 240 pps did not differ significantly from the one at 100 pps but was significantly shorter than those for lower frequencies (ts > 4.43, ps < .0001).

Seizure Motor Characteristics

Observed motor seizure expression strength in the unparalyzed arm was rated as medium for 88% to 89% of the seizures (Figures S4 and S5). Figure 2B shows the observed motor seizure duration for each stimulus frequency. There was no interaction between modality and stimulus frequency (F5,111 = 1.97, p = .118), indicating that stimulus frequency affected seizure duration similarly for ECT and MST. The effect of session number was not significant either (F2,111 = 1.76, p = .176). Therefore, the interaction term and session number were removed from the final model. Stimulus frequency significantly affected the motor seizure duration (F6,129 = 7.85, p < .0001). For ECT, the least squares means of the motor seizure duration ranged from 42.1 seconds at 5 pps to 21.7 seconds at 240 pps. At 5 pps, the mean duration of the motor seizure was shorter than the duration of the respective stimulus trains, which exceeded 1 minute on average; motor seizure expression often started notably later than the train initiation and ended before or shortly after the end of the train. For higher stimulus frequencies (≥10 pps), seizure duration was longer than the stimulus train. Compared with the frequency with the shortest seizure duration (240 pps), only the 5 pps and 10 pps conditions had a significantly longer seizure duration (ts > 3.76, ps < .0013).

Modality had a significant effect on the motor seizure duration (F1,129 = 22.2, p < .0001). Motor seizure duration for MST was 26.9% shorter on average than that for ECT (t = −4.71, p < .0001).

Exploratory EMG recordings indicated that the ratio between tonic and clonic phase duration was comparable across stimulus frequencies (Figure S8).

Seizure EEG Characteristics

EEG seizure duration was largely consistent with the observed motor seizure duration (Figure S7). Baseline global power was significantly greater in the MST than in the ECT condition (t = 9.10, p < .0001) (Figure S9), likely reflecting differences in anesthetics used across these conditions. We subsequently normalized the ictal and postictal power relative to the baseline for each condition by subtracting the log-transformed baseline power from the log-transformed ictal and postictal power. Session number and EEG channel did not significantly affect the normalized global EEG power (ps > .55). Consequently, we averaged the 2 channels within animal, condition, band, and session and did not model session number.

Figure 3 shows the average normalized EEG power data. There were significant main effects of seizure phase (F1,130 = 175, p < .0001) and stimulus frequency (F5,133 = 9.80, p < .0001). Furthermore, significant interactions were frequency × phase (F5,130 = 8.34, p < .0001), frequency × modality (F5,132 = 4.96, p = .0006), and frequency × modality × phase (F5,130 = 4.17, p = .0021). Nonsignificant effects were modality (F1,131 = 1.67, p = .231) and modality × phase (F1,130 = 0.508, p = .477). There were also no significant differences between animals (Wald test, p = .742). As expected, postictal power was lower than ictal power (t = −13.2, p < .0001). Comparing ictal power across stimulus frequencies, the 5 pps condition had significantly higher power than the minimum at 16 pps for ECT (t = 7.75, p < .0001), and the 5 pps and 10 pps conditions had significantly higher power than the minimum at 25 pps for MST (t = 3.40, p = .0042 and t = 5.97, p < .0001, respectively). In the postictal phase, there were no significant differences in power across the stimulus frequencies relative to the condition with the least negative power (ps > .112).Figure 3. Average electroencephalography (EEG) spectral power of ictal (A, B) and postictal (C, D) period relative to baseline for electroconvulsive therapy (ECT) (left) and magnetic seizure therapy (MST) (right) for various stimulation frequencies. The bottom row within each subplot corresponds to global EEG power, and rows above it correspond to discrete EEG bands (delta through beta). Individual raw and normalized data points are shown in Figures S9–S12. pps, pulses per second.

We also examined the effect of stimulation frequency on normalized EEG power within the delta, theta, alpha, and beta bands. In addition to the significant effects for global power, there were significant main effects of band (F3,533 = 14.8, p < .0001) and modality (F1,533 = 8.83, p = .0031) and a significant modality × band interaction (F3,533 = 4.88, p = .0050). Notably, all interactions involving frequency × band were not significant (ps > .591), indicating that the EEG power varied similarly with frequency across bands. Therefore, the significantly higher ictal power from low-frequency than high-frequency stimulation and the lack of significant frequency dependence in the postictal phase observed for the global power also held across bands.

Discussion

We showed that the number of pulses (and therefore charge and energy) at ST varies as a function of the stimulus train frequency in a similar way for ECT and MST. The range of frequencies that results in the lowest ST is centered around 16 pps (equivalent to 8 Hz in the conventional ECT terms of pulse-pair frequency). This value is below the lowest frequencies, 10 Hz or 20 Hz, available in standard commercial ECT devices (34, 35, 36). Our findings indicate that increasing stimulation frequency above 25 pps (50 Hz), as is typically done in conventional ECT titration, reduces seizure induction efficiency, even though higher frequencies require shorter stimulus trains. This is consistent with several clinical studies that have found a significant difference between ST at different stimulus train frequencies (18, 19, 20, 21,37).

Our analysis indicated that modality had a significant effect on ST. The higher ST for ECT than MST might be explained by the fact that the ECT sessions used a modified sedation. Additionally, the individual pulse amplitudes for ECT were determined with a unidirectional pulse train (27), whereas the trains used for frequency titration were bidirectional, and we have previously shown that bidirectional ECT trains can have higher ST than unidirectional trains (25). Nonetheless, the lack of interaction between frequency and modality supports the understanding that the fundamental mechanism of seizure induction by ECT and MST is the same, specifically tetanic stimulation of neural populations with an induced electric field (15,38).

In our dosing paradigm, the stimulus amplitude was individually titrated (27) and then doubled before ST was titrated by increasing the number of pulses. The absence of significant ST differences across animals suggests that the pulse amplitude individualization compensated appropriately for individual differences and that additional titration of the number of pulses was unnecessary. Instead, a fixed number of pulses could be used for all animals for a given modality and stimulus frequency.

The EEG data indicate that while the seizure strength was comparable between ECT and MST overall, stimulus frequency significantly affected seizure expression and did so in a different way for ECT and MST. The key finding is that the lowest stimulus frequencies (5 pps and 10 pps) produced seizures with higher ictal power than higher stimulus frequencies. One limitation of this observation is that at 5 pps the duration of the stimulus train typically exceeded the duration of the seizure (see Figure 2B), and therefore the seizures that were recorded were markedly longer and perhaps more robust than the average. Nonetheless, this was not true for the 10 pps condition, which produced the highest ictal power for MST; moreover, the individual seizures with highest power were recorded for the 5 pps and 10 pps conditions. Also, the differences in seizure expression between ECT and MST may be confounded by the anesthesia differences, reflected in the baseline power, although prior studies of nonhuman primates have reported differences in seizure EEG expression under the same anesthesia as well (see Figure S13) (39). Other factors that distinguish ECT and MST and could contribute to seizure expression differences are the induced electric field distribution and the pulse shape and duration (40). For example, compared with the symmetric bilateral electric field of circular-coil-on-vertex MST, RUL ECT stimulates stronger in the right hemisphere and in depth (32).

The current study suggests that stimulus trains in the 10 to 25 pps range are more effective at generating an excitatory state that overcomes innate cerebral inhibition than trains at higher or lower frequencies. While we did not probe the underlying mechanisms, this optimal stimulation frequency range can induce prolonged depolarization and epileptiform activity in hippocampal slices (41,42) and coincides with a 10 Hz recruiting rhythm that has been observed with EEG during endogenous seizure onset (43). In contrast, higher stimulus frequencies can suppress ongoing seizure activity by extending the neural refractory period, inducing intermittent axonal block and desynchronizing firing patterns (44,45). Lower frequencies (∼1 pps) can suppress seizures as well by inducing long-lasting hyperpolarization (46), suggesting possible explanations of the U-shaped curve that we observed in Figure 2A. Our EEG analysis also revealed that frequency affected ictal expression, suggesting shifts in the dynamics of neural inhibitory and excitatory processes with stimulation frequency. An extended discussion of putative mechanisms is provided in the Supplement.

The study had some limitations. The subjects were nonhuman primates that did not model depression. Modeling depression in nonhuman primates is challenging and resource intensive (47, 48, 49, 50). Moreover, the within-subjects design, which allowed for a relatively small number of animals, would not be feasible due to potential condition sequence–dependent cumulative effects of the stimulation sessions on depression symptoms. Furthermore, the study consisted of only titration sessions, whereas repetitive stimulation above the seizure threshold is needed for therapeutic efficacy (3). Few studies have investigated the impact of frequency on clinical efficacy and side effects (21,51, 52, 53), and some of these have had confounds (e.g., mismatch in the number of pulses across frequency conditions). Given this paucity of information, we cannot make conclusions regarding stimulus frequency and clinical outcomes at this time, but we review the evidence in the Supplement. Nonetheless, our findings are consistent with the limited clinical literature on the effect of stimulus frequency on seizure induction efficiency in human patients with depression (18, 19, 20, 21, 22,37).

Conclusions

Our results, coupled with previous reports, suggest that when individualizing the ECT or MST dose, it is most efficient to maximize the stimulus train duration first and only then increase the train frequency. This dosing/titration approach may be especially advantageous for patients with a high ST, which may otherwise exceed the Food and Drug Administration–regulated device limit of 100 J (34, 35, 36). Trials will be needed to confirm whether 10 to 25 pps results in more efficient seizure induction in patients than higher frequencies and to examine the impact on clinical outcomes.