Changes in Visual Long-Term Potentiation Show Preserved Cyclicity in Human Females Taking Combined Oral Contraceptives Open Access

For many women, hormonal fluctuations are associated with cyclical physical and neurological symptoms, such as in premenstrual dysphoric disorder (PMDD) and catamenial epilepsy [1, 3]. The neurosteroid withdrawal hypothesis attributes menstrual cycle-related disorders to the steep withdrawal of progesterone and allopregnanolone in the late-luteal phase [3, 4] and dominance of pro-excitatory oestradiol that is concurrent with increased menstrual cycle-related symptom severity [1, 2]. Clinically, seizure frequency increases in catamenial epilepsy, and mood worsens in PMDD during the perimenstrual phase in line with the withdrawal of neurosteroids [5].

The combined oral contraceptive (COC) pill is a popular form of hormonal contraception, underpinned by its action of suppressing fluctuations of the gonadal hormones and for this reason its potential to reduce symptoms of menstrual cycle-related disorders. The combination of synthetic oestradiol and progestogen in a standard 21/7 COC regimen prevents ovulation by acting on the hypothalamic-pituitary-ovarian axis to suppress the release of the gonadotropins and the subsequent release of oestradiol and progesterone from the ovaries [6]. Gonadal neurosteroid peripheral blood concentration levels are seen to be kept relatively consistent and suppressed throughout the 21 active hormone pills [6, 9]. For this reason, COCs should be protective against the symptoms of menstrual cycle-related disorders that arise with cyclic hormonal fluctuations but often are not [10, 11]. There is a lack of insight into the effect that COC has on the central nervous system and the efficacy of suppression of endogenous neurosteroids [12]. Empirical evidence has alluded to a phenomenon of preserved cyclicity, which details recurrent menstrual-related symptoms pre-menstruation, despite COC use [11, 13, 15]. The recurrent menstrual cycle-related symptoms are indicative of preserved menstrual cycling in the brain independent of peripheral gonadal hormone suppression.

Animal and human researches have been essential to elucidating the roles gonadal steroids play in cortical excitation and inhibition balance across the healthy menstrual cycle and the role that they may play in menstrual cycle-related neurological disorders. Oestradiol is largely pro-excitatory, and it has been shown to lead to increases in synaptic plasticity via enhanced long-term potentiation (LTP), a mechanism of greater synaptic strength and connectivity [16, 17]. It also increases dendritic branching and spinogenesis [17, 21]. In rodents on the morning of proestrus when oestradiol and progesterone levels are also high (comparable to the early part of the luteal or premenstrual phase), there is decreased hippocampal LTP – which can be attributed to dominating mechanisms of GABAergic inhibition [22]. By contrast, levels of cortical excitation are greatest at the peak of oestradiol and progesterone levels on the afternoon of proestrus (approximate to the premenstrual phase in humans) [23] and LTP is briefly enhanced before decreasing again at the onset of oestrus [22].

Several non-invasive human measures of neural plasticity including repetitive transcranial magnetic stimulation [24], paired-pulse stimulation [25], and electroencephalography [26] have demonstrated sensitivity to the effect of oestradiol in humans. Using a visual LTP task to compare the mid-follicular to the mid-luteal phase, Sumner and Spriggs [26] found a trend to reduce plasticity in the mid-luteal phase which may have been equivalent to the morning proestrus finding in rats by Sabaliauskas and Shen [22]. Potentially more equivalent to the afternoon of proestrus, other studies have demonstrated a general trend of oestradiol-driven cortical excitation and enhanced neural plasticity during the premenstrual phase [24, 25, 27, 28].

One of the most widely used and validated methods of non-invasively assaying the state of LTP in the human brain has been the induction of NMDA receptor-dependent LTP in the sensory cortices, recorded via the consequential enhancement of the visually evoked potential (VEP) amplitude change using EEG [29, 32]. An overview of the visual LTP task itself can be found in Sumner et al. [32], including a breakdown of the VEPs. Briefly, studies most often report modulation of the N1 and P2 components of the VEP [32]. Evoked during early visual processing, these components represent summed synchronised activity of tens of thousands of active neurons in cortex recorded at the EEG electrodes, whereby the change in topography and valence is determined by changes in active location/s in the brain over time as information moves through cortex and alters the detected projection of electric fields. As well as sensitivity to the menstrual cycle [26], studies on visual LTP have demonstrated clinical sensitivity, such as decreased LTP in depression [33] and schizophrenia [34]. Thus, the visual LTP paradigm may demonstrate sensitivity to LTP mechanisms that may underlie preserved cyclicity in the brain.

In order to permit more detailed inference on the EEG evoked potential results and the mechanism by which LTP might change, a computational model of thalamo-cortical connectivity and receptor changes can be fitted to evoked potential data from the LTP paradigm [35]. The model is an extension of more classic canonical microcircuit models typically implemented using dynamic causal modelling (DCM) [36] that provide simplified yet realistic representations of laminar, pooled population neural dynamics in the visual cortex that can contribute to evoked activity. In addition to the four interconnected populations (cortical layer IV spiny stellates (ss), layer II/III and V pyramidal neurons, and a single inhibitory interneuron population), the thalamocortical model includes separate inhibitory interneuronal input into deep and superficial layers. In the model, thalamus contains relay (rl) and reticular cells, as well as cortico-thalamic projection pyramidal neurons into layer VI. AMPA, NMDA, GABA-A, GABA-B, M and H channels are also parameterised. This model has previously demonstrated sensitivity to (invasively demonstrated [37, 38]) changes to the visual cortex evoked by LTP [35].

The primary aim of this study was to investigate the changes in neural plasticity across the menstrual-like COC cycle of healthy females, through a modified visual LTP paradigm from Teyler, Hamm [39], and Sumner, McMillan [40]. LTP across the menstrual-like COC cycle was studied at two time points during the active hormone pills (days 3 and 21) and one time point during the 7-day hormone-free interval (HFI, day 24).

We hypothesised that there would be evidence of preserved cyclicity via greater visual LTP in day 21 than day 3. This would provide evidence of preservation of an imbalance in excitation/inhibition occurring in the perimenstrual phase. DCM is expected to localise the biological underpinning of this – either as a decrease in inhibitory input into the cortical microcircuitry, an increase in excitatory connectivity, or a change in NMDA/AMPA channel parameters. Secondarily, this study hypothesised that if the synthetic hormones are additionally mediating changes to excitation and inhibition then there may be greater LTP when on active pills (days 3 and 21) than when in the HFI (day 24), driven by ethinylestradiol. Blood samples were taken to confirm there were no significant changes in endogenous progesterone or oestradiol.

26 females volunteered to participate in the study, and basic demographics are provided in Table 1. Participants were eligible to participate if aged between 18 and 35 years old and had been taking the COC for at least 3 months. Participants were required to be on a monophasic 28-day COC regimen, with 21 days of active hormone pills and then experiencing a withdrawal bleed during the 7-day HFI. COCs that were excluded were those that contained drospirenone, due to research indicating efficacy in alleviating premenstrual symptoms [41, 42]. Participants were excluded if they had been diagnosed with psychiatric disorders such as depression and PMDD but included if they experienced premenstrual symptoms. Participants were also required to not be taking any psychotropic medication or forms of hormonal medication other than the COC at the time of participation.

The study used a three-session counterbalanced, within-subject design. The three sessions occurred on days 3 and 21 which were during the active hormone pills, and day 24 which was the third day of the inactive pill HFI and roughly when the withdrawal bleed begins. Each session was approximately 2 h in length and started between 2 p.m. and 4 p.m. to control for the diurnal fluctuations in progesterone and oestradiol that are greatest during the morning [43]. Session order was counterbalanced to account for order effects.

For at least one pill cycle, participants were required to fill out a Daily Record of the Severity of Problems (DRSP) questionnaire as an indicator of preserved cyclicity symptoms. The DRSP consisted of 26 questions on a 6-point scale (1 = “not at all,” 6 = “extreme”) regarding physical and mental symptoms and feelings experienced throughout the menstrual cycle. Participants received a unique link by either text or email every evening to fill out the DRSP questionnaire. 25 participants provided a full COC cycle of DRSP results, as one participant withdrew their consent to use the data. Compliance was good with 96.3% of questionnaires returned completed. Six participants had datasets with more than one missing value (in these 6, the mean missing days = 4, range 2–6 days). A mixed-effect model was used to compare the DRSP scores for each week (days 1–7, days 8–14, days 15–21, and the HFI days 22–28) in R [44]. Bonferroni correction for multiple comparisons was applied to individual contrasts.

Blood samples were taken at the beginning of each study session (10 mL into K2EDTA BD Vacutainers^®), to measure plasma oestradiol and progesterone levels. Tubes were refrigerated immediately and centrifuged at 1,500 g for 15 min 2–2.5 h after collection. Plasma was then aliquoted and frozen at −80°C to permit batch processing. Samples were processed within 2 years of collection and showed no effects of time in storage on hormone concentration. Progesterone (endogenous) and oestradiol (specifically endogenous 17β-oestradiol) concentration in plasma was quantified and processed by the Liggins Institute (Auckland, New Zealand) by electrochemiluminescence immunoassay. Assays were performed according to Roche Oestradiol III (2016) (range: 5–3,000 pg/mL) and Progesterone III (2020) (range: 0.05–60 ng/mL) assay guidelines using a Roche Cobas 8,000 analyser (e601 module). This includes the use of PreciControl Universal (PCU) for quality control (Oestradiol III: mean PCU = 3.24%; Progesterone III: mean PCU = 1.52%). Samples were run in single and loaded undiluted as standard. Cross reactivity for synthetic progestogens and 17α-ethinylestradiol is reported as ≤0.33%. To confirm no significant change in hormones, a repeated measures ANOVA was run on days 3, 21, and 24 concentrations of progesterone and oestradiol separately.

EEG acquisition parameters can be found in the online supplementary material (for all online suppl. material, see https://doi.org/doi/10.1159/000530805). The visual LTP paradigm used in the current study was modified from [40], designed to induce LTP-like enhancements of the early VEP components. 24 participants took part in a day 24 session, and 23 completed all three sessions. Three participants did not complete due to changes in circumstances/eligibility over the time of intended participation (exacerbated by COVID-19 lockdowns).

In the protocol plan of the current study, it was indicated that the photic tetanus condition of the visual LTP paradigm was high-frequency (9 Hz) circular sine gratings. However, because of a programming error the photic tetanus was at a frequency of 6.5 Hz for all three study sessions of the first 10 participants. Within-subjects, the frequency was always the same across sessions. 9 Hz is considered an optimal frequency for inducing LTP [30]. Previous research has shown a lower 5 Hz frequency does not induce visual LTP, though a trend to increased N1b was seen [30]. We tested for a significant effect of tetanising speed on visual LTP before combining the datasets by integrating it as a grouping variable in the first ANOVA (shown in Fig. 1 and see section Visual LTP Data Analysis for more description on how this was implemented).

The stimuli used for the paradigm consisted of vertical and horizontal circular sine gratings with a spatial frequency of 1 cycle per degree, displayed upon a grey background at full contrast subtending 8 degrees of visual angle. Participants were seated at a distance of 90 cm between the screen and their eyes and were required to passively fixate on the red dot at the centre of the screen.

Conditions are shown in Figure 2. The baseline condition consisted of 4 min of randomly ordered vertical and horizontal circular sine gratings presented at low frequency (1 Hz) to the central visual field. The vertical and horizontal stimuli were presented 120 times (for a total of 240 presentations) for 34.8 ms each. The interstimulus interval (ISI) differed between 5 intervals ranging from 897 ms to 1,036 ms that occurred randomly and at equal frequencies. The baseline condition occurred at 3 different time points during the task: firstly, at the pre-tetanus condition, which was immediately before the tetanus, and then for two post-tetanus conditions: early post-tetanus and late post-tetanus which occurred 2- and 40-min post-tetanus, respectively. The baseline condition enabled comparisons between the event-related potential (ERP) amplitudes for the pre-tetanus and post-tetanus conditions. The 2-min break which followed the photic tetanus and was prior to the early post-tetanus condition allowed for any retinal images to dissipate. The purpose of the early post-tetanus condition was to identify any changes in the visual response that occurred immediately after the induction of LTP. 40 min after the photic tetanus condition, the late post-tetanus condition assessed whether the early changes in the visual response were maintained. LTP maintenance in the late post-tetanus condition is crucial in determining whether long-term changes of neural plasticity have occurred.

The photic tetanus condition occurred immediately after the pre-tetanus condition. It consisted of 1,000 presentations of moderate- (6.5 Hz) or high-frequency (9 Hz) circular sine gratings for 33 ms. The ISI had jitter that occurred randomly but of equal frequencies. The 6.5 Hz ISI occurred for either 101.6, 120.8 or 134.6 ms to cause jitter, and the mean 9 Hz ISI occurred for either 62.5, 76.3, or 90.1 ms. The photic tetanus for both 6.5 Hz and 9 Hz took 2 min, with the direction of horizontal or vertical gratings pseudo-randomised between participants, and the tetanising frequency presented was split between participants. Between the early and late post-tetanus, resting-state EEG data were collected for eyes open and closed for 5 min each.

Pre-processing and data analysis were carried out using SPM12 (http://www.fil.ion.ucl.ac.uk/spm/software/spm12/). See the online supplementary material for details.

For all analyses, main effects were considered significant at p < 0.05 family-wise error corrected (FWE-c). A more liberal threshold is reported for detecting interaction effects (p < 0.001 uncorrected), and any post hoc contrasts are FWE-c (as in [26, 40, 45, 46]). Simple effect tests were conducted as appropriate. Results were interpreted at the peak level; where multiple significant peaks occurred for the same component, only the most significant peak within a cluster is reported.

The sequence of analyses is shown in Figure 2. An initial repeated measures ANOVA was used to investigate the effect of time (i.e., if LTP occurred during the early post-tetanus condition and late post-tetanus condition), whether LTP displayed specificity to stimulus orientation (tetanised vs. non-tetanised) [47], and as a between-subject effect whether LTP displayed specificity to tetanising speed (6.5 Hz vs. 9 Hz). The data for the early post-tetanus condition were utilised as a check for early LTP and short-term potentiation, while late post-tetanus checked for maintained early LTP. The initial ANOVA used results from day 24 trials only, due to this day being during the HFI and therefore suitable as a baseline without influence from the active pill day synthetic hormones.

This 2 × 2 × 2 ANOVA was run across a 250-ms time window (Fig. 1a). Henceforth, the early post-tetanus condition is calculated and defined as early post-tetanus minus pre-tetanus, and the late post-tetanus condition is calculated and defined as late post-tetanus minus pre-tetanus. Ten participants had the 6.5 Hz photic tetanus, and 14 participants had the 9 Hz photic tetanus.

The results of this analysis informed the subsequent parameters including the time window of significantly modulated ERP components to be considered for an effect of COC and cyclicity (Fig. 1b). To reduce the number of comparisons, a narrow time window focussing on the N1 and P2 where group effects are typically found was used. Additionally, tetanising speed and stimulus need to be considered and carried forward as factors in the design or can be combined/averaged to reduce design complexity.

These primary outcome analyses were constructed as one-way within-subject’s ANOVAs for the early post-tetanus and late post-tetanus blocks (Fig. 1b). We averaged across tetanising stimulus and combining participants with different tetanising speeds. We tested for two a priori main effects that reflect the hypotheses of the current study. The first main effect explored an effect of COC, by testing active pill days (days 3 and 21) versus inactive pill days (day 24) (two-tailed contrast). The second a priori main effect was tested for an effect of preserved cyclicity during the active pill days (day 3 vs. day 21) (two-tailed contrast).

Sources analysis was carried out using multiple sparse priors, as implemented in SPM12, for group inverse reconstruction [48]. The source analysis was run on the time window 170–190 ms as in the ERP analyses. A 2 × 3 repeated measures ANOVA was run (early post-tetanus × late post-tetanus across day 3 × 21 × 24). A peak was selected using a FWE-c F contrast that encompassed both early and late effects of the photic tetanus on day 3 and 21. The selected peak was extracted from each participant as the local field potential, with a 5-mm spherical radius around the MNI coordinate [12, −96, −2] in the right calcarine cortex.

This study implemented a thalamocortical model of interlaminar connectivity (Fig. 3) as in Sumner and Spriggs [35], parameterising the following populations: superficial pyramidal (sp) cells, superficial interneurons (si), ss, deep pyramidal (dp) cells, deep interneurons, thalamic pyramidal (tp) cells, reticular (rt) cells, and relay (rl) cells. The model also parameterises the decay constant of AMPA, NMDA, GABA-A, GABA-B, and M and H channels. Equations governing the model are explained in the online supplementary material. The parameters that were allowed to vary (during model fitting) were set according to the Douglas and Martin [49] canonical microcircuit while also allowing all of the additional parameters within layers II/III and IV to vary. In summary, rl > ss, ss > ss, ss > sp, ss > si, si > sp, si > si, sp > sp, sp > si, si > ss, sp > dp, dp > tp, tp > rl were allowed to vary and the rest of the parameters were fixed. AMPA and NMDA decay rates were also allowed to vary in the model.

The generative model described above was fit to the empirical EEG data using DCM. We incorporated a parameterised general linear model into the inversion protocol. LTP was modelled as a linear change from baseline that is greatest in the late post-tetanus block [−1 0 1], to reflect early phase LTP and the typical time course of P2 potentiation [32] as well as a non-linear change from baseline that peaks in the first post-tetanus block (to model general excitability and short-term potentiation) [−1 1 0] and following the time course of N1 potentiation in these data. This was combined in the contrast [−1 1 0; −1 0 1] allowing for both non-linear and linear contributions to describe the condition-specific effects and representing the model that best fit in Sumner and Spriggs [35] to describe visually induced LTP.

Parametric empirical Bayes (PEB) (as described in detail in [50]) was used to estimate the change in parameters of primary visual cortex microcircuitry that contribute to the effect of preserved cyclicity on LTP. The design matrix comprised the group effect (a column of 1 s) and the effect of cyclicity (a column of -1 s for day 3 and 1 s for day 21) as well as random between-subject effects (columns of 1 s per participant). The parameters identified above as being allowed to vary in the initial model fitting (often termed the “full” model [50]) were re-estimated at the second level using Bayesian model reduction that iteratively searches all possible parameter contributions to the effect of cyclicity and reduces free parameters until only those which meaningfully contribute to the model evidence remain. Bayesian model averaging was implemented to identify the direction and size of the effect of this subset of parameters on the effect of cyclicity on LTP. Their significance was thresholded by only reporting posterior probabilities (Pp) of the reduced model that were ≥0.99.

The global DRSP score was calculated for each participant, and when taking into account all questions, visual inspection of the graphs, some evidence for preserved cyclicity is evident (Fig. 4). The mixed-effect model showed a main effect of week (F(3,646.17) = 3.91, p = 0.008). Individual contrasts revealed that this is driven by an increase in symptoms in week 4. The difference in week 2 versus week 4 (t(646) = −3.412, p = 0.0007) survived Bonferroni correction for multiple comparisons. The difference between week 1 versus 4 was only significant uncorrected (t(646) = −1.997, p = 0.046). All other comparisons were ps > 0.05.

The literature on preserved cyclicity typically reports physical premenstrual symptoms (e.g., Sulak, Scow [15]). Incidence of a symptom as a percent of participants scoring over 1 on each of the four physical-based DRSP questions is plotted in Figure 5. Visual inspection of the graphs shows that physical evidence of preserved cyclicity is difficult to disentangle from background noise. Symptoms such as headache, joint and muscle pain do appear to become more prevalent in our sample in the HFI.

Blood samples were successfully collected from 22 participants for all sessions (shown in Table 2). Mean values are shown in Figure 6, and while small differences are seen between days 3, 21, and 24, Figure 6b provides comparison with low and high normal menstrual cycle plasma concentrations of oestradiol and progesterone demonstrating effective suppression of peripheral hormone changes in our sample. Using a repeated measures ANOVA, there was no significant effect of session, meaning no change in either progesterone (F(2,20) = 1.12, p = 0.346) or oestradiol (F(2,20) = 1.44, p = 0.261).

The initial 2 × 2 × 2 ANOVA determined a 20-ms time window, and from 170 to 190 ms post-stimulus would be appropriate for the analyses of the preserved cyclicity and effect of LTP (Fig. 7; see orange-shaded ERP section in A and B). Significant post-tetanus enhancement of the N1 occurred in the early post-tetanus block at 175 ms. Significant enhancement of the P2 occurred in the late post-tetanus block, peaking at 179 ms. The online supplementary material contains the detailed results.

A significant interaction was found between tetanising speed and the early post-tetanus block at 89 ms (F(1,88) = 17.39, p = 0.046 FWE-c, p = 0.00007 uncorrected). Post hoc t-contrasts revealed that the potentiated VEP component in the early block was the P1 and was greater for 9 Hz than 6.5 Hz tetanising speed at 89 ms (t(88) = 4.17, p = 0.023 FWE-c, p = 0.00004 uncorrected), indicating that the P1 demonstrates tetanising speed sensitivity (Fig. 7c). However, this significant interaction was found outside of the time windows that the initial ANOVA established and so participant trials for 6.5 Hz and 9 Hz tetanising speed were combined into a single factor for the subsequent one-way within-subject’s ANOVAs.

Key results are shown in Figure 8. The results of the early post-tetanus one-way ANOVA revealed that there were no significant main effects detected for both a priori hypotheses of the COC and preserved cyclicity.

The late post-tetanus one-way ANOVA also revealed that there were no significant main effects for the effect of COC. There was a main effect of preserved cyclicity that occurred in the late post-tetanus block, indicating a left-lateralised increase in visual LTP at 174 ms (F(1,144) = 14.13, p = 0.021 FWE-c; p = 0.0005 uncorrected). Post hoc t-contrasts indicated that the potentiated VEP component during the late post-tetanus condition was the P2, which was significantly more potentiated during day 21 in comparison with day 3 (t(44) = 3.76, p = 0.011 FWE-c; p = 0.0002 uncorrected).

All individual model fits explained >94% of the variance in the data indicating that the combination of linear and non-linear dynamics described the visually induced LTP task effects well. PEB revealed a reduced model of lower ss > si (posterior estimates/effect size; Ep = −0.13, Pp = 1.00) and si > ss (Ep = −0.15, Pp = 1.00) connectivity in day 21 compared to day 3 for the linear model (Fig. 9). Connectivity from tp > rl was also slightly reduced (Ep = −0.03, Pp = 1.00) in day 21 compared to day 3 for the linear model (Fig. 9). Overall, there appears to be a reduction in the layer IV to layer III inhibitory interneuronal gating mechanism on day 21 compared to day 3.

The current study investigated the changes in neural plasticity that occurred across the menstrual-like cycle of healthy female COC users. Successfully replicating previous visual LTP studies, we showed significant potentiation of the P1 and N1 VEP components in the early post-tetanus condition, and significant potentiation of the N90OP and P2 in the late post-tetanus condition. A time window encompassing the N1 and P2 was selected to study the effect of the pill (N1 and P2 are key VEP components where effects are typically found [32], and a narrower time window reduces multiple comparisons in SPM12). We showed visually induced LTP does not remain consistent across the active pill days despite peripheral hormonal suppression, and the P2 was significantly more potentiated in day 21 than day 3. This provides emerging evidence towards an effect of preserved cyclicity that is driven by an underlying neural mechanism that enhances LTP in the absence of hormonal fluctuations. Computational modelling indicates that this may be driven by a change in GABA-mediated inhibitory interneuron gating on LTP. There was no significant difference in LTP between active and inactive pill days (days 3 and 21 vs. day 24).

The global score of the DRSP did not provide evidence of cyclicity. ANOVA results indicated that there is a significant effect of the HFI where symptoms worsen compared to the active pill days. This was largely driven by week 2 compared to the HFI. Week 1 was only significantly different from the HFI in post hoc testing uncorrected. Week 3 and the HFI were not significantly different. Nor was week 1 compared to 2 or 3 or week 2 compared to 3.

The DRSP appears to provide some evidence for increases in physical symptoms in the days leading up to the HFI (Fig. 5); however, the results are very noisy. Subjective measure in a such a small sample (compared to that of the Sulak, Scow [15] study with 262 participants and Sveindottir and Backstrom [54] with 83) is prone to noise. Sulak and Scow [15] show ∼30% of participants demonstrated presence of at least one physical premenstrual symptom in current COC users prior to the HFI (particularly breast tenderness, bloating, and swelling). Sveindottir and Backstrom [54] reported that 75% demonstrate cyclicity; however, their premenstrual period counted −9 to −1 days to the onset of bleeding and will have included a few HFI days for many participants. The current study’s EEG results provided the opportunity for more objective and potentially more reliable marker of preserved cyclicity to compare with the literature. In the current study, 68% (15/22) show evidence that in the field peak voxel their visually evoked response was more positively modulated on day 21 than day 3 (Fig. 8c).

The modulation in the field peak represents significantly greater potentiation of the P2 component for day 21 compared to day 3 during the active pill days (Fig. 8b). These findings suggest there is a cyclic neural change despite suppressed peripheral blood oestradiol concentrations and despite days 3 and 21 being across active pill days and therefore involving the addition of the same amount of synthetic hormones [6]. The difference in LTP between active pill days closely aligns with human and rodent studies that have identified a general trend of oestradiol-driven enhanced neural plasticity and increased synaptic density during the premenstrual phase in the unmedicated menstrual cycle, which is approximately day 21 of the active pill days [19, 23, 24, 28, 55].

However, it appears unlikely that small shifts in oestradiol concentrations lead to functionally relevant neurobiological shifts in neural plasticity during active pill days in this dataset. Because if neural plasticity followed peripheral blood hormone shifts then during the active pill days a trend for enhanced neural plasticity would have been expected for days 3 and 21 when peripheral blood oestradiol levels are at the highest due to the additional synthetic ethinylestradiol, in comparison with day 24 at its lowest [56]. Instead, it was revealed that there was enhanced LTP seen on day 21, giving evidence towards preserved functionally relevant neural shifts within the brain despite hormonal suppression.

In the current study, the finding of enhanced P2 potentiation on day 21 may plausibly represent a mechanism of greater excitation or reduced inhibition underlying preserved cyclicity of LTP increases in the perimenstrual-like phase. Previous studies have shown that generative, neurobiologically informed models of cortical activation during EEG can be used to disentangle these possibilities after administration of drugs [57, 60] and in disorder [61, 62], as they have demonstrated sensitivity to the change in visual cortex microcircuitry that underlie LTP [35] and change over the menstrual cycle [63].

The modelling was implemented as in Sumner and Spriggs [35] and provided a good fit for the data. Using PEB, it was shown that on day 21 (where there was also greater P2 potentiation) there was a reduction in the reciprocal connectivity between interneuronal inhibitions in layer IV for the linear model (Fig. 9). Layer IV is an established key site of LTP in visual cortex [64]. Further, inhibition has been shown to provide a gating mechanism preventing the induction of LTP from layer IV to III [64]. Typically, in invasive LTP induction studies, this inhibition is reduced using a GABA-A receptor antagonist such as bicuculline or picrotoxin [64]. The modelling supports a change in inhibitory gating of LTP, involving the reduction of both excitatory input signalling and inhibitory output of superficial inhibitory interneurons.

The neurosteroid withdrawal hypothesis during the perimenstrual phase is thought to explain the neurobiology of PMDD and catamenial epilepsy. However, the hypothesis is contingent on large changes in hormone concentrations. Thus, the COC ought to be an effective treatment. However, COC has generally not demonstrated efficacy in alleviating symptoms of menstrual cycle-related disorders. Instead, total suppression using gonadotropin-releasing hormone agonists is the most effective treatment [65]. This suggests that a key mechanism of cyclicity has always played a causal role in menstrual cycle-related disorders and occurs somewhat independently of large hormone concentration shifts and neurosteroid withdrawal, requiring total gonadotropin-releasing hormone-medicated cyclic suppression. Alternatively, it indicates an adaptation to COC-based hormone suppression that explains why, despite a decades-old hypothesis, effective hormone treatments for menstrual cycle-related disorders have largely remained elusive. The current result implicates an inhibitory and GABA-sensitive mechanism – supporting the progress being made now with GABA steroid therapeutics such as sepranolone in treating PMDD [66, 68].

Interestingly, previous research has shown that drospirenone-containing COC is effective in reducing negative physical and psychiatric premenstrual symptoms [41, 42, 69, 70]. The efficacy of drospirenone has been largely established from subjective participant questionnaires like the DRSP used in the current study, and the neurobiological mechanisms by which drospirenone elicits symptom reduction remain unknown. Drospirenone has anti-adrenergic and anti-mineralocorticoid activity, with its efficacy suggested to arise from its pharmacological similarities to progesterone and spironolactone [69, 71, 72]. Drospirenone’s analogue spironolactone has been shown to decrease LTP in vitro [73] which may suggest that anti-mineralocorticoids function to provide more effective suppression of preserved cycling. As noted in Hampson [12], the pharmacological heterogeneity of COCs is likely to generate variation in their effect on the central nervous system and suggests the problem of generalising across COCs. Visually induced LTP, as used in the current study, may be useful as a potential objective method to check for sufficient suppression and reveal the mechanism for drospirenone’s efficacy and other hormone therapies used in PMDD and catamenial epilepsy.

It is interesting that there was an effect of cyclicity (day 3 vs. 21) but no apparent difference between active pill days (3 and 21) and inactive pill day (day 24) in the ANOVA testing for an effect of COC. There is a visual difference in potentiation of the P2 component between day 21 of the active pills and day 24 of the inactive pills (Fig. 8a). The analysis for the effect of COC used essentially averaged (half weighted) data for days 3 and 21 to compare against day 24, which means while there may have been a difference between day 24 and day 21, this is attenuated by the lack of a difference between visual LTP on days 3 and 24. Either way, it appears the effect of cyclicity is more powerful than any effect of synthetic hormone levels.

Although we found differences between active pill days, further research into preserved cyclicity is required to strengthen our interpretation that there are menstrual cycle-related neural changes that drive differences in visual LTP across active pill days. Additional measure days between days 3 and 21 could be utilised to illustrate whether a gradual trend for enhanced neural plasticity occurs, as the comparison between two active pill days that are 18 days apart leaves large room for possible unobserved neural fluctuations. The use of more measure days may also enable attribution of neural changes to certain gonadal hormones or receptor dynamics.

Furthermore, it would be valuable to complete this study in a group of individuals with PMDD who are on COC. A limitation of the current study is that we are unable to relate the increase in visual LTP to any symptom on the DRSP because there was a low incidence of symptoms reported in the DRSP overall. In addition to collecting higher participant numbers to overcome the noise, it would also be easier to determine if there is a relationship in a sample that is experiencing more intense and disordered premenstrual symptoms such as in PMDD.

In the current study, no VEP component demonstrated specificity towards tetanising stimulus orientation. Specificity is typically reported in the N1 component [32, 47]. Interestingly, the lack of specificity in the current study is the second time that this has occurred in an all-female study [26]. Sumner and Spriggs [26] did not show significant N1 potentiation. This warrants future specific investigation into sex effects on the visual LTP paradigm.

The blood sample analysis was completed using a single aliquot per person (rather than duplicate or more). This means we are unable to be sure there is no unreliability in an individual’s reported hormone concentrations. However, our purpose was to show no change in endogenous hormones over time while taking COC (as has been reported previously), and we were able to show this. Therefore, our approach is considered fit for the purposes of this paper.

The programming error mentioned in the methods that resulted in the use of two different photic tetanising frequencies must be acknowledged as a limitation within the current study. There was a significant interaction between tetanising speed of 9 Hz and the early post-tetanus block at 90 ms. However, this possible confound was mitigated by ensuring the time windows used in the analyses of the effect of cyclicity and COC was well clear (170–190 ms). Kirk and McNair [30] identified that potentiation of the N1 was trending at 5 Hz, but not significantly as with 8.6 Hz. We did not find a significant difference in visual LTP of the N1 between 6.5 Hz and 9 Hz. This hints at a graded response of potentiation from the upper limits of theta-band frequency towards mid-range alpha at 9 Hz. Further research is still necessary to fine-tune the critical range to induce visual LTP, given that frequencies outside of the alpha-rage can effectively induce LTP. Our findings demonstrate the need for future research replicating and investigating the mechanisms driving tetanising frequency specificity in P1 potentiation.

The current study showed increased LTP in the perimenstrual-like phase of the COC by showing greater VEP potentiation on day 21 compared to day 3. Both days are during the active hormone days of a 28-day COC regimen. Computational modelling revealed changes to GABA-mediated interneuronal gating may be implicated. This study provides objective evidence of preserved cyclicity in current COC users. Preserved cyclicity may underlie the challenges with developing effective hormonal treatments for menstrual cycle-linked disorders such as PMDD and catamenial epilepsy. Modelling suggests involvement of a GABA-sensitive inhibitory mechanism is consistent with current progress in use of GABA-A receptor-acting steroid drugs for PMDD.

The authors would like to thank Eric Thorstensen and Christine Keven at the Liggins Institute for analysis of the plasma samples.

The study was approved by the University of Auckland Human Participants Ethics Committee (Reference Number 023544). Informed written consent was given by all participants prior to participating in the study.

The authors have no conflicts of interest to declare.

This study was funded by a Neurological Foundation grant awarded to RLS.

Conceptualisation, funding acquisition, and supervision: R.L.S. methodology: R.L.S., S.M., and A.S. Investigation: R.L.S., E.S., and M.A. Formal analysis: R.L.S., E.S., and A.S. Writing – original draft: R.L.S. and E.S. Writing – review and editing: R.L.S., E.S., M.A., S.M., and A.S.

The data that support the findings of this study are not publicly available due to this having not been requested in the ethics and written consent process but are available from RLS upon reasonable request.

Elizabeth Stone and Rachael L. Sumner contributed equally to this work.