Fear acquisition across the menstrual cycle: The moderating role of vagally mediated heart rate variability

¹University of Potsdam, Department of Biological Psychology and Affective Science, Faculty of Human Sciences, Karl-Liebknecht Strasse 24/25, 14476 Potsdam, Germany

²Faculty of Health Sciences Brandenburg, University of Potsdam, Potsdam, Germany

 

Corresponding author:

Berenike L. Blaser

University of Potsdam

Karl-Liebknecht Strasse 24/25

14476 Potsdam, Germany

Email: berenike.blaser@uni-potsdam.de

 

Running head: Fear acquisition and menstrual cycle

Keywords: fear potentiation, human fear conditioning, menstrual cycle, vagal tone, heart rate variability, startle, skin conductance, fear discrimination

  

During the luteal phase of their menstrual cycle, most menstruating individuals experience aversive physical and affective symptoms (Tschudin et al., 2010). These symptoms might be linked to cycle phase-dependent neurophysiological and autonomic processing, such as changes in neurotransmitter systems (Nappi et al., 2022) and a decrease in parasympathetic activity (Schmalenberger et al., 2019). Heightened anxiety is a frequently reported symptom during this phase (Allen et al., 1991). Notably, anxiety is consistently associated with reduced vagally mediated heart rate variability (vmHRV) (Chalmers et al., 2014), such that vmHRV fluctuations likely contribute to the mood and affect fluctuations observed during the menstrual cycle (Matsumoto et al., 2007).

VmHRV is an indicator of autonomic processing and is associated with the inhibitory functional connectivity between the medial prefrontal cortex (PFC) and the amygdala (Sakaki et al., 2016). In their neurovisceral integration model of fear, Battaglia and Thayer (2022) propose that vmHRV serves as a relevant biomarker for studying inter- and intraindividual differences in fear learning. This is attributed to the fact that vmHRV is considered a peripheral indicator of the interplay among crucial brain structures involved in fear acquisition and processing, with one of the primary regions being the amygdala. Notably, vmHRV is consistently found to be reduced during the luteal phase of the menstrual cycle (Schmalenberger et al., 2019). There is also evidence that suggests that stronger reductions in vmHRV are accompanied by a higher degree of aversive symptoms, including anxiety-related symptomatology, during the late luteal phase, known as premenstrual symptoms (Matsumoto et al., 2007). This may indicate that the heightened aversive symptoms during the luteal phase are linked to reduced amygdala inhibition, as reflected by lower vmHRV.

A research paradigm that serves as a laboratory model for anxiety (Beckers et al., 2023) and in which the amygdala has consistently been found to be involved (Kuhn et al., 2019; Sun et al., 2020) is fear conditioning. Fear conditioning paradigms, therefore, may provide valuable insights into the relationship between anxiety-related symptomatology and cycle-related fluctuations. One of the most commonly used fear conditioning paradigms is the differential conditioning procedure in which neutral stimuli are either paired (CS+) with an aversive unconditioned stimulus (UCS) or not (CS-) (Lonsdorf et al., 2017). The CS+/UCS association (fear acquisition takes place) leads to conditioned responses to the CS+. Additionally, an association between CS- and the absence of UCS is formed, giving the CS- the quality of a safety signal. The difference in reactions to the sole presentation of CS+ and CS- is then termed fear discrimination, representing how effectively the individual distinguishes between the "dangerous" and "safe" conditions. This process of differential fear and safety acquisition can be enhanced by providing verbal instructions about the pairings (Mertens et al., 2018). Two of the most commonly used physiological outcome measures of fear acquisition are skin conductance response (SCR) and the startle blink reflex (i.e., fear-potentiated startle), characterized by larger responses to the CS+ compared to the CS-. SCR is typically interpreted as an indicator of sympathetic activity, while the startle blink reflex is suggested as an indicator of subcortical fear processing in fear-inducing contexts because it is directly mediated by the amygdala (Lonsdorf et al., 2017; Wendt et al., 2023).

In a recent review, Merz et al. (2018) highlighted the potential relevance of menstrual cycle-related fluctuations in fear conditioning paradigms. However, the evidence presented by the authors is inconclusive, and there is little further research on this topic. Armbruster et al. (2018) observed a trend for overall higher startle magnitude in the luteal phase. This effect is especially pronounced in persons experiencing strong affective premenstrual symptoms (Epperson et al., 2007). Critically, Glover et al. (2013) showed higher startle magnitudes for CS+ compared to CS- (i.e., fear discrimination), during the luteal than the follicular phase. In skin conductance measures, however, previous studies showed no overall effects between menstrual phases  in SCR (Lonsdorf et al., 2015; Milad et al., 2006), as well as no differential fear discrimination effects (Carpenter et al. (2022); Milad et al., (2006). In a small study of 31 naturally cycling individuals, van der Molen et al. (1988) compared those who were currently in the last week of the luteal phase (premenstrual phase) to those in all other phases (including early and mid-luteal) and found higher fear discrimination in SCR  in premenstrual individuals (n = 8).

Overall, despite the potential impact of the menstrual cycle on anxiety-related mechanisms (Merz et al., 2018), the existing research on the relationship between fear acquisition and the menstrual cycle has been rather scarce and characterized by the use of unimodal indicators of fear acquisition, leading to heterogeneous findings. To deepen our understanding of the impact of the menstrual cycle on fear processing, we therefore investigated fear conditioning measures of startle magnitude and SCR in individuals in the follicular phase and individuals in the luteal phase of their menstrual cycle. Furthermore, extending previous findings, we will also investigate the potential interacting role of vmHRV as one of the most relevant modulators of the relationship between anxiety symptomatology and the cycle phase.

We tested healthy participants who took part in a larger biofeedback intervention study (Wendt et al., 2020, December 16). Participants were recruited through postings at the Universities of Potsdam and Greifswald, online platforms, and through posts on social networks. Exclusion criteria included having a body mass index lower than 18.5 kg/m² or higher than 30 kg/m², cardiovascular, neurological, or respiratory diseases, impaired hearing or color vision, claustrophobia, pregnancy, and the use of medications that alter the normal functions of the autonomic nervous system. A total of 128 participants were tested. Of the tested participants, 58 were included in the data analysis who reported a regular active menstrual cycle, no hormonal contraception and were currently either in the luteal or follicular phase of their cycle. All participants gave informed consent and were compensated with course credit or money. The data were collected between June 2020 and October 2022 and the project was approved by the Ethics Committee of the University Medicine Greifswald.

The testing took place in the context of a large intervention study described elsewhere (Wendt et al., 2020, December 16) which investigates the effects of heart rate variability biofeedback on extinction learning in healthy individuals. The focus of this report is on data from the acquisition phase of the fear conditioning paradigm that was used in this project.

After the introduction to the study, participants were led to a darkened experimental room where sensors were attached to measure the physiological signals. Participants went through a 6-minute electrocardiogram (ECG) measurement with closed eyes, followed by a fear acquisition protocol.

For the fear paradigm, a blue square and an orange circle served as conditioned stimuli. The stimuli were displayed on a black background in the center of the computer screen, each with a duration of 6 seconds. The intertrial intervals (ITI) varied in length between 14, 15, and 16 seconds. The stimulus presentation was realized using Presentation software (Version 20.3, Neurobehavioral Systems, Inc.).

The assignment of the CS+ and CS- stimulus was counterbalanced across participants. The CS+ was partially (50%) reinforced by an electro-tactile stimulus for a 1ms duration (UCS). The intensity was individually adjusted for each participant before training to a level judged to be "very unpleasant but not painful." (M = 6 mA, SD = 3.7, Range [1.3; 21.0]). The CS- was never accompanied by an electro-tactile stimulus. The stimulation electrode was placed on the inside of the right leg, approximately 3 cm above the ankle. Electrical stimulation was controlled using a DS7A Constant Current Stimulator (Digitimer, Hertfordshire, UK) in Potsdam and an S-48K stimulator (Grass Instruments, West Warwick, RI, USA) in Greifswald.

Bursts of white noise (duration of 50 ms duration, 95 dB volume) served as startle acoustic startle probes and were delivered binaurally through headphones (Potsdam: Audio-Technica ATH-PRO700MK2, Greifswald: AKG K 66). The startle stimuli were administered 4.5 or 5.5 seconds after each CS onset and in half of the ITI (i.e., 16 times) 7.5 seconds after CS offset.

The conditioning phase consisted of 16 trials (8 CS+, 8 CS-) of uninstructed acquisition, followed by a slide informing the participants which of the geometric figures was associated with the aversive stimulus. Then 16 trials of instructed acquisition followed. Before and after the acquisition protocol, participants rated the stimuli on valence and arousal using a 9-point visual analog scale ranging from 1 (valence: highly unpleasant, arousal: calm) to 9 (valence: very pleasant, arousal: exciting).

All physiological recordings were performed using a BIOPAC MP160 amplifier system and AcqKnowledge 5.0.2 software (BIOPAC Systems, Inc., Goleta, CA, USA). All data were sampled at a rate of 2,000 Hz and filtered at various sample rates (see below). A silicon grounding electrode (TerniMed) was attached to the participant's left upper arm.

For the ECG, two Ag/AgCl electrodes (10mm contact surface diameter; Schuler Medizintechnik GmbH) filled with electrode paste (CareFusion) were applied on the right forearm (approximately 2 cm below the elbow) and the left leg (approximately 2 cm proximal to the ankle). The ECG data was digitally sampled at a rate of 400 Hz. The processing of the ECG data was executed using Kubios HRV Software (University of Eastern Finland, Kuopio, Finland) following the recommendations of the Task Force of the European Society of Cardiology (Malik, 1996). The root mean square of successive differences (RMSSD) was used as a vmHRV measure due to its robustness to breathing rate influences, which were not controlled for in the present study (Chapleau & Sabharwal, 2011).

For the Electrodermal Activity (EDA) recording, the non-dominant hand of the participants was used. Two Ag/AgCl sintered biopotential electrodes (8mm contact surface diameter, Easycap GmbH) were filled with isotonic electrode contact gel (0.5% NaCl, GEL101, BIOPAC Systems, Inc.) and then attached palm-side using double-sided adhesive rings over the hypothenar muscles. The EDA was recorded using an EDA100C module (BIOPAC Systems, Inc.), employing a constant voltage method (0.5V). The EDA data were passed through a 10 Hz low-pass filter.

The startle response was measured using electromyography (EMG) of the Orbicularis Oculi muscle. For this purpose, two electrodes (Ag/AgCl electrodes, 5mm contact surface diameter; Schuler Medizintechnik GmbH) were filled with electrode contact paste (CareFusion) and positioned under the left eye of the participant, with the first electrode located approximately 0.5 to 1 cm below the eye vertically aligned with the pupil, and the second electrode placed laterally adjacent to it (parallel to the eyelid contour) approximately 1 cm from the outer corner of the eye. The EMG signal was recorded using the EMG100C module (BIOPAC Systems, Inc.). The EMG data was digitally sampled at a rate of 1,000 Hz and filtered using 30 Hz high pass, 400 Hz low pass and 50 Hz notch-filter.

The startle magnitude and EDA data were preprocessed using MATLAB.

SCRs were analyzed using the trough-to-peak method (TTP) in Ledalab Version 3.4.9 (Benedek & Kaernbach, 2010). In the TTP method, the SCR amplitude is defined as the difference between the skin conductance at the peak of the response and its preceding trough in a determined time window. Adhering to the guidelines (Boucsein et al., 2012), the response window was set from 1 to 4 seconds after CS onset. The startle magnitude procedure adhered to Blumenthal et al.’s (2005) recommendations. Blink response onset and peak were automatically identified within 20-120 ms after probe onset, with a peak before 150 ms, employing the algorithm of Globisch et al. (1993). In a subsequent visual inspection, trials without blinks were scored as zero, and trials with excessive background activity or artifacts were considered missing. As our emphasis was on interindividual variability in startle magnitude, we decided to use raw data instead of T-transformed data (Bradford et al., 2015).

Cycle phases were assessed through self-report using the forward-count method (Schmalenberger et al., 2021). The follicular phase was assigned when participants reported being in days 1-11 of their cycle. To determine the luteal phase window, 11 days were subtracted from the reported average cycle length. If participants were between these phases or if their cycle day or average cycle length could not be reliably assessed, they were excluded from the analysis.

All analyses were conducted in R version 4.2.2. Linear mixed models were calculated, using SCR and startle magnitude from instructed acquisition trials as dependent variables respectively. Participant intercepts were introduced as random effects to cluster the trials by participant. Predictor variables included trial condition (CS+, CS- or ITI/UCS, contrast-coded), menstrual phase (follicular or luteal, dummy-coded) and their interaction. RMSSD and its interaction with trial condition as well as participants’ age (control variable) were included if they improved the model fit as indicated by Likelihood Ratio Tests.

Out of the 58 individuals included in the analysis, 36 reported being in the follicular phase, while 22 reported being in the luteal phase. The difference in group sizes is attributed to the higher reliability of reporting the follicular phase using the forward-count method, resulting in more individuals currently in the luteal phase being excluded from the analysis. The groups did not differ significantly in terms of age (m_luteal = 22.68±2.32, m_follicular = 24±2.9).

The results of the model can be viewed in Table 1. There was no significant main effect of the cycle phase, t(63.4) = 1.54, p = .13., indicating no differential overall startle magnitude between phases. The interaction effect of Condition x Cycle Phase yielded significant terms in the model, t(1296.1) = -2.58, p < .05 (CS+ vs. CS-), t(1296.1) = -2.68, p < .01 (CS+ vs. ITI).

Table 1 Results of mixed model predicting startle magnitude

 Note. The final model had the following structure: startle magnitude ~ condition * phase + (1|participant). CS+ – stimulus paired with unconditioned stimulus; CS- – stimulus not paired with unconditioned stimulus; ITI – intertrial interval; lut – luteal phase.

 

A visualization of the interaction effect, including significance levels from post-hoc contrast testing, can be seen in Figure 1. Although CS+ trials evoked higher startle magnitudes compared to CS- and ITI trials in both the follicular and luteal phases, the disparity between CS+ and the other trials was more pronounced during the luteal phase, indicating heightened fear discrimination. Post-hoc contrast testing, however, showed no significant difference between startle responses to the CS+ in the follicular and the luteal phase group, t(63.4) = -1.54, p = 0.13. Adding the RMSSD as a main effect or interaction to the model did not change the results. Age was not included as it did not improve the model fit.

 

Figure 1 Startle magnitude by condition and menstrual cycle phase.