Intercorrelation of physiological seizure parameters and hormonal changes in electroconvulsive therapy

Abstract Objective Physiological parameters that predict electroconvulsive therapy (ECT) effectiveness may reflect propagation of the induced epileptic seizure. As an indication of seizure propagation to the diencephalon, we here examined the correlation between prolactin increase after ECT and clinical seizure evaluation parameters, focusing on peak heart rate. As a proxy for peripheral endocrine stress response, we examined the correlation to postictal cortisol increase. Methods Participants were consecutively recruited from clinical ECT patients (n = 131, age 18–85 years). The first ECT session in a series was examined. For each participant, blood serum concentrations of prolactin and cortisol were measured immediately before and within 30 min after the seizure. Physiological parameters were extracted from clinical records: peak heart rate (HR) during seizure, electroencephalography (EEG) seizure duration, and motor seizure duration. Correlations were calculated using non-parametric tests. Results Serum prolactin increased after ECT and correlated with peak HR, EEG seizure duration, and motor seizure duration. Peak HR during seizure also correlated positively with both EEG seizure duration and motor seizure duration. Correlations were unaffected by age, sex, baseline prolactin levels, antipsychotics, or beta-blocking agents. Serum cortisol increased after ECT but did not correlate with the seizure evaluation parameters, nor with prolactin concentrations. Conclusions Our findings of a positive correlation between peak HR and prolactin that was independent from the peripheral endocrine stress response might be in line with the idea that tachycardia during ECT seizures reflects seizure propagation to the diencephalon. This supports the practice of monitoring cardiovascular response for ECT seizure evaluation.


Introduction
Electroconvulsive therapy (ECT) response usually takes several treatments repeated over weeks, making it difficult to ascertain directly how each treatment session affects the targeted illness [1]. However, seizure propagation beyond the stimulated area seems necessary to achieve treatment effect [2]. Thus, the potential efficacy of each seizure is clinically evaluated by parameters believed to be markers of seizure propagation, such as cardiovascular response, epileptic activity on an electroencephalogram (EEG), and observations of motor seizure manifestations [3][4][5].
Among commonly used parameters to evaluate seizures, peak heart rate (HR) during seizure [6,7] and postictal EEG suppression have been found to correlate with treatment effectiveness [8][9][10]. While the duration of epileptic activity on EEG has not shown a linear association with treatment outcome [9,11], it can be assumed that some minimal duration is required, which is grounds for the widespread use of EEG seizure duration as a treatment evaluation parameter [5]. Regarding duration of motor seizures, professional consensus deem very short motor seizure duration to be a negative predictor of treatment effect [5].
Among the seizure evaluation parameters mentioned above, peak HR relates well to ECT treatment effect, and may thus offer an opportunity to understand the underlying therapeutic mechanism. ECT-induced tachycardia could reflect a central nervous sympathetic reaction due to propagation of epileptic activity to insular cortex, hypothalamus, or brain stem nuclei [12][13][14][15]. However, tachycardia might also be elicited as an indirect effect of the ECT procedure, with peripheral stress response including adrenal gland activation causing catecholamine release [16]. Common EEG monitoring in ECT only reflects superficial cortical activity [17], and motor seizures seems to reflect an even smaller aspect of elicited seizures [18].
A marker of seizure propagation to the diencephalon is postictal increase of serum prolactin, which has been observed in both epilepsy [19,20] and ECT [21,22]. A prolactin increase has been suggested to reflect postictal increase of gamma-aminobutyric acid (GABA) signaling in the diencephalon, which in turn suppresses tonic dopaminergic inhibition of prolactin release from hypothalamic-pituitary neurons [23]. If tachycardia in ECT reflects propagation of the epileptic seizure to subcortical brain areas [12][13][14][15], peak HR in the ECT seizure should correlate with elevation of serum prolactin after ECT. Also, postictal EEG suppression, which at least partially seems to be GABA-dependent [24], should correlate with elevation of serum prolactin after ECT.
The alternative model is that ECT tachycardia is caused by a peripheral stress response from the adrenal gland as an indirect seizure effect. A cortisol increase could be a possible marker with similar pattern as described above. A postictal increase of serum cortisol has been observed in ECT [25,26]. To our knowledge, no study has hitherto examined the correlations between cortisol and peak HR or EEG suppression.
The aim of this study was thus to evaluate clinical seizure evaluation parameters focusing on peak HR as a proxy marker for subcortical seizure propagation in ECT. To this end, we explored how peak HR during ECT seizures correlates to other clinical seizure evaluation parameters as well as to prolactin increase as a marker of postictal inhibition in the diencephalon, and to cortisol increase as a possible marker of peripheral endocrine stress activation.

Study design
Observational cohort study.

Setting
This study is a subsample of the Predictors for ECT (PREFECT) multicentre study [27,28], including participants from one study site, Uppsala University Hospital.
The study was approved by the Regional Ethical Review Board in Stockholm, Sweden (registration numbers: 2012/ 1969-31/1 and 2018/869-32). All procedures were performed according to the ethical standards of relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.

Participants
Inclusion for the PREFECT study ran from 2014 through 2016. All clinical patients referred for ECT were invited to participate with inclusion criteria: age ! 18 years, the capacity to give informed consent, and a planned ECT series of at least six treatments.
At the Uppsala University Hospital study site, 133 participants were included.

ECT procedure
ECT was performed with a Thymatron V R System IV (Somatics LLC, Lake Bluff, IL, USA). Initial stimulus dose was decided by the local ECT practitioner based on age and sex in accordance with the manufacturer's instruction manual [29]. Only right unilateral (RUL) electrode placement was used. Pulse width was almost exclusively 0.5 ms. Median (Q1, Q3) stimulus charge was 252 (152, 350) milliCoulomb. Information on anesthetic agent and dose was available in 106 of 131 participants. In these, thiopental was used in all treatments except one, where propofol (2.46 mg/kg) was used instead. Median (Q1, Q3) thiopental dose was 4.43 (3.91, 4.90) mg/kg. Suxamethonium was routinely used for muscle relaxation. No adjunctive medications apart from anesthetics and muscle relaxants were used as part of the routine anesthetic procedure for the first ECT treatment.

Variables
The planned two main outcomes were peak HR during seizure (beats per minute, bpm) and occurrence of postictal EEG suppression (yes or no). Correlations were planned between these main outcomes and the following four parameters: (1) EEG seizure duration (s), (2) duration of motor seizures (s), and changes in serum concentrations of (3) prolactin (measured in Normalized Protein Expression (NPX), further described under 'measurements' below), and (4) serum cortisol, (ng/ml). Information on type and dose of anesthetic agents and concurrent use of any antipsychotic or oral betablocking agents was collected from the medical charts.

Measurements
Seizure evaluation parameters (peak HR, postictal EEG suppression, EEG seizure duration, and motor seizure duration) were extracted by a retrospective chart review. All measurements had been recorded by clinical staff (physician, nurses, nurse's assistant) as per the clinical routine at the site during the study period. Baseline HR and Peak HR during seizure was observed on continuous electrocardiography. Postictal EEG suppression was assessed visually and recorded as confirmation or negation of occurrence of suppression (yes or no). EEG seizure duration was measured by a two-channel-EEG recording (one channel over each temporal lobe). The endpoint of epileptic activity was determined visually. Motor seizure duration was assessed by visual inspection, with muscle relaxant effect (neither any cuffing method nor electromyography were used).
Blood samples were collected before the first ECT treatment and within 30 min after seizure termination. Blood was drawn in 10 ml serum tubes, left 30-60 min at room temperature, then centrifuged for 15 min at 2000xg. After centrifugation, blood serum aliquots were stored at À20 C for a maximum of 30 days, transferred to biobank, and then stored at À70 C until analysis.
Serum concentrations of prolactin were analysed as part of a larger protein assay in the PREFECT study, using the commercially available Proximity Extension Assay (PEA) technique (Olink Proseek V R Multiplex ONC Lv2 panel, OLink Bioscience, Uppsala, Sweden) [30]. PEA results are given as NPX, a relative unit which can be used to compare changes in protein concentrations within one assay, but not for comparison of separate analyses [30].
From the PREFECT study biobank, 0.25 ml of serum blood was analysed for cortisol using ultra-performance supercritical fluid chromatography-tandem mass spectrometry (UPSFC-MS/MS) [31].

Statistical methods
Statistical analyses were performed with SPSS Statistics 25 (IBM Corporation and others).
Peak HR and baseline HR were the only parameters with a normal distribution, as confirmed with Shapiro-Wilks normality test, and are presented as mean value with standard deviation. The other continuous variables, which were not normally distributed, are presented as median values with the first and third quartile presented.
Significance in changes of prolactin and cortisol concentrations following the first ECT seizure was assessed using the sign test.
Data on postictal suppression were excluded from further analysis due to few available data (n ¼ 62).
Correlations between peak HR and EEG seizure duration, motor seizure duration, and changes in prolactin and cortisol concentrations were calculated with Spearman's rank test. Mann-Whitney U was used for comparing subjects when grouped by use of antipsychotics or not.
To enable controlling for potential confounders, regression analyses using generalized linear models were applied. Peak HR was entered as dependent variable, and each variable with a significant correlation (p < 0.05) in the Spearman test above was entered as independent variable along with potential confounders (age, sex, baseline HR, baseline prolactin, antipsychotic medication, and beta-blocking medication) in separate multiple regression analyses. Three regression models (linear, gamma distribution, and negative binominal log link distribution) were tested for best fit with the data, and the model with the lowest AIC (Aikaike's Information Criterion) was selected, which yielded a linear model for all three regressions.

Participants
Of the initial 133 participants with samples collected from the PREFECT biobank, 131 participants had available medical chart data at the Uppsala site. Two participants had blocked access to their medical charts after study inclusion and were thus excluded from this study as no retrospective chart review was possible.

Descriptive data
The study population consisted of more women (n ¼ 80, 61%) than men. Median (Q1, Q3) age in the population was 45 (30, 59) years. There was no significant age difference between sexes. Sixty-six participants were on antipsychotics.
Baseline prolactin concentrations were higher in subjects using antipsychotics compared with those that did not (p < 0.01). There were no significant differences between subjects with and without antipsychotic medications regarding age (mean 46 years with antipsychotics, 45 years without, p ¼ 0.77) or sex (number of women vs. men 40/26 with antipsychotics, 40/25 without, p ¼ 0.91). Further, there was no difference between the groups regarding prolactin change, baseline HR, or peak HR (all p ! 0.16).

Main results
Seizure evaluation parameters and changes in prolactin and cortisol are shown in Table 1. Heart rate increased from mean (SD) 78 (15) bpm at baseline, to 120 (23) bpm peak HR during the ECT seizure. Both prolactin and cortisol concentrations increased after ECT, as listed in Table 1.
Individual postictal changes in prolactin and cortisol are shown in Figures 1 and 2, respectively.
When grouped by use of antipsychotics, a significant prolactin increase was seen in both groups (p < 0.001 both groups).
There was a positive correlation between peak HR on the one hand, and EEG seizure duration (r ¼ 0.56, p < 0.001), motor seizure duration (r ¼ 0.63, p < 0.001), and prolactin increase (r ¼ 0.24, p ¼ 0.03) on the other. Peak HR did not correlate with cortisol increase (r ¼ 0.03, p ¼ 0.78). A scatter plot of postictal hormone concentration changes vs. peak HR is shown in Figure 3.
The correlations between peak HR on the one hand, and EEG seizure duration, motor seizure duration, and postictal prolactin increase on the other hand, remained significant (p < 0.001) when each parameter with a statistically significant (p < 0.05) correlation to peak HR was entered into regression analyses along with potential confounders ( Table 2).

Discussion
In this study of the first seizure in 133 patients receiving ECT, we explored how peak HR correlates to other clinical seizure evaluation parameters and changes in serum concentrations of prolactin and cortisol. We can for the first time report that 49 (À20, 93) b a Missing data: baseline heart rate n ¼ 14, peak HR n ¼ 7, motor seizure n ¼ 1, prolactin before n ¼ 40, prolactin after n ¼ 41, prolactin change n ¼ 43. b Change before/after seizure is significant at the 0.01 level with the sign test. bpm: beats per minute; SD: standard deviation; EEG: electroencephalogram; Q1: first quartile; Q3: third quartile; NPX: normalized protein expression. tachycardia during ECT seizures correlate with postictal prolactin increase (albeit the correlation coefficient of 0.24 warrants some caution in interpretation) but not to postictal cortisol increase. The positive correlation between peak HR and prolactin increase, in conjunction with the lack of correlation between peak HR and cortisol increase, strengthens the suggestion that peak HR reflects propagation of the epileptic activity to the diencephalon. Both tachycardia during ECT seizure [7,32] and postictal prolactin increase [33] have previously been reported to correlate with ECT treatment effect, but to our knowledge a possible correlation between these markers has not previously been examined.
The exclusive use of RUL electrode placement in this material should be considered, as in threshold stimulations, bitemporal electrode placement has been found to require a higher stimulus charge to reach the EEG duration threshold, but also produce a larger prolactin release [34]. Possibly reflecting similar phenomena with other measures, a more recent study on differences in seizure propagation as measured with single photon emission computed tomography reported that generalized seizures elicited with bitemporal electrode placement resulted in a more bilateral engagement of subcortical structures than RUL placement-however, suprathreshold stimulus charges were used and serum changes in prolactin were not measured [35]. With these previous findings taken into account, an investigation on peak HR relating to postictal prolactin increase in a setting with bitemporal electrode placement would be valuable, and might show a stronger correlation between the two parameters.
We found no correlation between cortisol increase and peak HR or postictal EEG suppression, which to our knowledge has not been examined in previous studies.
There are previous reports on prolactin elevation [21,22,33] and on cardiovascular activation in effective ECT seizures [6,7]. Our findings of a correlation between the two strengthens the notion of peak HR as a feasible marker for seizure propagation in clinical practice, especially as both peak HR [6,7] and postictal prolactin increase [33] have previously been proposed as markers of clinical ECT effectiveness.
Due to a lack of available data, postictal suppression was excluded as outcome variable in this study, but considering the previous connection to treatment outcome, the relation of postictal EEG suppression and other clinical seizure parameters and hormonal effects in ECT remains an interesting topic for future explorations.
There are several limitations to this study. Data from clinical routines may suffer from inherent quality issues, as they are recorded in a less standardized and reliable manner than in a controlled study setting. Several of the issues raised below stem from this fact. On the other hand, this increases the generalizability of results, as the situation for data collection is highly similar to daily clinical work. Exact time points of blood sampling with relation to the ECT treatment were not recorded, which lower the temporal precision. As only one sample was taken after seizure termination, it is also possible that the peak in prolactin and cortisol concentrations was not captured, particularly in regards to prolactin  which seems to have a faster decline of postictal concentration increase than cortisol, with peak levels of postictal prolactin repeatedly recorded at $10-15 min after an ECT seizure [25,[36][37][38]. Peak concentrations of postictal cortisol have been reported to remain 15-30 min after an ECT seizure [36,37], although one study reported this duration of the postictal cortisol peak only with a stimulus charge three times seizure threshold, as compared with threshold stimulus intensity (threshold defined as the lowest stimulus intensity producing a motor seizure !25 s) [25]. Assuming postictal changes of prolactin and cortisol from these previous studies apply to our study setting, this might result in an underestimation of the postictal hormone concentration change in the material, and in consequence increase the risk of type 2 errors.
Intravenous anticholinergic or beta-blocking drugs, which may influence peak HR, are sometimes used as part of the ECT anesthesia routine and were not controlled for in this study. However, since such drugs were not used routinely for the anesthetic procedure at the study site, especially in the first ECT treatment, we believe it unlikely that this would introduce any systematic error into the data. Further, in this study we could not control for all medications, indications for ECT, body mass index, menstrual cycle, substance use, or concomitant medical conditions which might influence seizure parameters, prolactin release, or autonomic nervous regulation. Antipsychotics increase baseline prolactin, and although subjects with and without antipsychotics did not differ in median prolactin concentration change in this study there is a possibility that patients taking antipsychotic drugs may have decreased headroom above baseline, which may have influenced the range of prolactin elevation. However, we consider it unlikely that any such factor would introduce a systematic error in a way to produce the observed correlations. Rather, the relatively small data set increases the risk for type-II errors. The calculations were not corrected for multiple comparisons and we have therefore not corrected for the risk of random significance in the findings.
Another aspect of this study data is that only the first seizure of the full ECT series was observed. Previous studies have reported a decrease in EEG seizure duration [39] and postictal prolactin increase [26,37], with an increasing number of treatments in a series of ECT. A possible explanation for these observations is the anticonvulsant effect of ECT seizures [40], and this aspect of study design can be important to consider when comparing different studies of ECT seizure quality, which might include different parts of a treatment series.
We conclude that peak HR during the ECT seizure, as well as other ECT evaluation parameters, correlate with prolactin increase after ECT, but not with cortisol increase. This supports the notion of tachycardia in ECT seizures as centrally elicited, rather than primarily caused by an indirect stress response from secondary treatment effects. As peak HR is  easy to measure, reflects epileptic activity in diencephalon, and predicts ECT effect, it seems to be a feasible parameter for seizure evaluation in ECT. Going forwards, such measurements should be further tested in prospective studies including bilateral and bifrontal electrode placements.