Exogenous hormones in the pill may modulate gray matter volumes
in the brain
Merel T. de Klerk
A R T I C L E I N F O A B S T R A C T
Keywords:
Oral contraceptive use Endogenous sex hormones Exogenous sex hormones Brain volume changes
It is estimated that approximately 16% of the women at a reproductive age worldwide use oral contraceptives (OCs), which contain synthetic variants of the sex hormones estradiol (E2) and progesterone (P). Two brain areas that are marked with a high density of receptors for both E2 and P, and which are consequently affected by their fluctuations, are the hippocampus and the amygdala. Women using OCs have been reported to have reduced gray matter (GM) volumes in both brain areas. It remains unknown whether this reduction is due to the low levels of endogenous sex hormones, due to the presence of exogenous sex hormones, or due to the interaction of both. As endogenous sex hormones have been linked to mood and mental health, further information is needed for the millions of women that use OCs or want to start using them. We aimed to replicate the reduction in GM volume and highlight the contribution of endogenous and exogenous sex hormones on brain volume changes by assessing GM volumes of the hippocampus and amygdala using magnetic resonance imaging of 114 young healthy women. In addition, we measured endogenous (i.e. E2 and P) and exogenous (i.e. ethinylestradiol and progestin) hormonal blood levels, from OC users and naturally cycling (NC) women in the early follicular (low E2 and low P) and periovulatory phase (high E2 and low P). We found lower GM volumes in the hippocampus of OC users compared to NC women, but not in the amygdala. Moreover, we found a trend for larger hippocampi with longer OC intake duration, independent of androgenicity. It seems as if in the first months of OC intake a reduction in volume took place, which may become normalized with longer OC intake. For the amygdala, a trend for volume change over the course of OC intake was dependent on androgenicity, which might explain our null finding. Furthermore, we observed a number of preliminary associations between hormonal levels and brain volumes. E2 + ethinylestradiol levels were associated with greater GM volumes in the right amygdala in OC users. Additionally, P levels were associated with greater amygdala GM volumes in the early follicular phase. In conclusion, it seems as if both endogenous and exogenous, or an interaction of both, affect the brain volume changes as seen in OC users, although these exploratory and preliminary findings have to be interpreted with caution.
1. Introduction
Worldwide, it is estimated by the United Nations that 151 million women use oral contraceptives (OCs), which equals 16% of women at a reproductive age. In some countries, such as Germany and the Netherlands, the percentage is even higher, with about one third of these women using OCs (United Nations, 2019). With the advent of the pill, around 1960, women gained more control over their own body. Starting a family could now be planned, allowing women to pursue a professional career. This has been revolutionary for society and the role of women in economy (Goldin & Katz, 2002). The most commonly used OC is the combined OC, i.e. it contains synthetic variants of the
hormones estradiol (E2) and progesterone (P; Bezemer et al., 2016). These hormones are normally produced in the body, but their endogenous production is decreased when synthetic hormones are administered, as they bind readily to the same receptors (Rabe et al., 1997). As a result, the concentration of endogenous E2 and P remains low at all times in OC-taking women, while fluctuations occur across the natural cycle (Montoya & Bos, 2017; see Figure 1). The natural cycle is usually divided into two phases: the first phase is the follicular phase, which starts on the first day of menses and ends when ovulation takes place. E2 and P levels are initially low but E2 levels rise towards ovulation. Ovulation is followed by the luteal phase, the second half of the natural cycle, when E2 levels drop slightly but P levels increase (Boron & Boulpaep, 2016).
Figure 1. Endogenous hormone levels in a natural cycle (a) and during combined OC use (b). Adapted from (Montoya & Bos, 2017).
Although the synthetic estradiol in OCs usually consists of ethinylestradiol, different progestins are on the market (Sitruk-Ware, 2006). An important distinction must be made with regards to the androgenicity of the progestins: many older generation progestins exert androgenic actions by binding to the androgen receptor, in addition to the progesterone receptor (PR), with various side effects as a result. By contrast, newer progestins are anti-androgenic, because they bind more specifically to the PR. The brain contains many receptors for E2 (ER and ER) and P (PR-A and PR-B). Not surprisingly, variations of these hormones thus affect cognition and behavior (Gogos et al., 2014; Montoya & Bos, 2017). Two brain areas that are marked with a high density of receptors for both E2 and P are the hippocampus and the amygdala (Brinton et al., 2008; Österlund et al., 1998).
The hippocampus is involved in many functions related to cognition and emotion. As such, it is responsible for cognitive behaviors, such as learning as well as episodic and spatial memory, and for motivational behaviors related to stress, emotion and affect (Fanselow & Dong, 2010). The functionality of the hippocampus is impacted by endogenous sex hormones: in rats, injection of E2 increases spine density and enhances memory (Packard & Teather, 1997; Woolley & McEwen, 1993). Spine density in the hippocampus is decreased, however, by P (McEwen & Woolley, 1994). In humans, higher activity in the hippocampus is observed during both a spatial navigation and a verbal fluency task at the time of the late follicular phase, when E2 levels are high (Pletzer, Harris, Scheuringer, et al., 2019). Moreover, the ovulatory phase as well as high E2 levels in general have been associated with both increased fractional anisotropy (FA), which is related to the white matter (WM) structure, and gray matter (GM) volume in the hippocampus (Barth et al., 2016; Lisofsky et al., 2015; Pletzer et al., 2018; Protopopescu et al., 2008). With regards to OC users, smaller GM volumes in the hippocampus compared to naturally cycling (NC) women have been repeatedly reported (Hertel et al., 2017; Lisofsky et al., 2016; Pletzer, 2019). Notably, the longer an OC has been previously used, the greater the GM volume (Pletzer, Harris, & Hidalgo-Lopez, 2019). Hence, it is unclear how OC use precisely relates to hippocampal GM volume as it seems contradictory.
The amygdala consists of several subnuclei and is the so-called salience detector that is essential for the generation of emotional responses, fear conditioning, and emotional memory (LeDoux, 1993; Sander et al., 2003). Within the amygdala, sensory information is received in the lateral nuclei, which is sent to the basolateral amygdala for processing. The central nucleus serves as the output channel and is responsible for the actual changes in the body and behavior. The amygdala also responds to endogenous sex hormones: during ovulation (i.e. high E2 levels) women had reduced amygdala activity during perception of emotional images compared to the early follicular phase (i.e. low E2 levels; Goldstein, 2005). Moreover, administration of P led to increased amygdala reactivity to salient stimuli and emotional memory formation was positively correlated with P levels (Ertman et al., 2011; van Wingen et al., 2008). Furthermore, In OC users, activation of the amygdala via emotional stimuli is generally lower compared to NC women (Gingnell et al., 2013; Petersen & Cahill, 2015). Similarly, women using OCs show altered memory formation for emotional events (Nielsen et al., 2011). These findings are endorsed by longitudinal findings on structural morphology: GM volume in the left amygdala decreased when women started taking OCs (Lisofsky et al., 2016). Importantly, no significant association of GM volume with previous OC intake duration was found, as opposed to the hippocampus (cf. Pletzer, Harris, & Hidalgo-Lopez, 2019).
In conclusion, both brain areas are affected by sex hormone fluctuations. Additionally, OC users usually have smaller GM volumes in both brain areas. It remains unknown whether the effects that arise in the brain caused by OCs are due to the low levels of endogenous sex hormones, due to the presence of exogenous sex hormones, or due to the interaction of both. As endogenous sex hormones have been linked to mood and mental health (Montoya & Bos, 2017), further information is needed for the millions of women that use OCs or want to start using them. Therefore, the aim of the current study is to highlight the contribution of exogenous and endogenous sex hormones on brain volume changes, as well as to replicate these brain volume changes itself. More specifically, the following research questions were composed:
(1) Is there a significant difference in structural morphology of the hippocampus and amygdala
between OC users and NC women measured during ovulation and the early follicular phase?
(2) How are endogenous and exogenous E2 and P related to the structural morphology of the hippocampus and amygdala? And in OC-users, is the contribution of exogenous P dependent on androgenicity?
To address these questions, structural morphology of the two regions of interest (ROIs) was assessed using magnetic resonance imaging (MRI), in addition to endogenous and exogenous hormonal blood samples, from OC users and NC women in the early follicular (low E2 and low P) and periovulatory phase (high E2 and low P). Based on the aforementioned research, it can be hypothesized that the structural morphology of the hippocampus and amygdala will be affected by OC use. More specifically, we expect reduced GM volumes for OC users in the hippocampus and amygdala compared to NC women. Moreover, we expect to find the largest differences in the hippocampus between the OC group and oNC group, because high E2 levels (i.e. during ovulation) were previously related to larger hippocampal volumes. With regards to the second research question, it is difficult to formulate a directed hypothesis, as a significant association between E2 or P levels and GM volume of the two ROIs has only been reported once. Notably, this is to our knowledge the first study to assess the influence of exogenous hormones on brain structure.
2. Material & Methods
2.1 Participants
This thesis is part of a 3-year study on the impact of endogenous and exogenous sex hormones on social behavior and the underlying brain responses. The study was approved by the Ethics committee of the Medical Faculty of the University Tübingen (331/2016BO). A total number of 119 women was recruited by advertisements in the newspaper and on social media, by emails and by posting information around the University Tübingen and the University Hospital Tübingen. All participants were aged between 18 and 33 years to reduce confounding factors such as age on brain volume and to ensure measurements took place before menopause. All participants were Caucasian and fluent in German, reported to be heterosexual and gave written informed consent. Exclusion criteria were any mental disorder (including premenstrual dysphoric disorder; as assessed by the DSM 5, SCID interview and medical history), complicating medical problems (e.g. severe hypertension, diabetes, congestive heart failure), any hormonal treatment (except thyroxin) or medication interfering with brain activation, and lifetime pregnancy or breastfeeding. Furthermore, participants with contradictions for MRI measurements were excluded, as well as women not willing to be informed about incidental MRI findings. Suitability for participation was assessed by trained psychologists or physicians by
means of semi-structured interviews. Participants received €50 after participation.
The aim of the current study was to explore differences in brain volume. Therefore, both women using OCs and NC women were included. Because hormonal levels fluctuate during the menstrual cycle, NC women were tested in their early follicular phase or in their periovulatory phase. Participants thus corresponded to one of three groups: OC users (OC), NC women in their early follicular phase (fNC), and NC women in their periovulatory phase (oNC). The OC group was measured in their active pill phase, the fNC group was measured during day 2–5 of their menstrual cycle based on self-report, and the oNC group was measured from 3 days before up to 2 days after a positive ovulation test. Importantly, the OC users had been taking their current combined monophasic OCs for at least 6 months and previous contraceptive methods were noted down, although not used for further analyses (29.5% had used another type of hormonal contraceptive before). All NC women had a regular menstrual cycle with a duration between 21 to 35 days and no OC intake in the past 4 months. Four participants of the OC group were excluded due to measurements in their pill-free week and one woman of the oNC group was excluded because she was accidentally measured during the luteal phase. Hence, the total sample amounted to 114 women, consisting of 66 NC women and 48 OC users.
In order to further characterize the three groups and to explore whether the groups had comparable characteristics, corresponding information was gathered (see Table 1). Analyses of variance (ANOVAs) were applied with group as the between-subjects factor to test for group differences, unless when differently stated. Age was homogenous across groups (F(2) = 0.03, p = 0.97). Educational attainment was based on the highest level of education achieved: if low secondary education was completed, it was rated as low (1), if high secondary education was completed, it was rated as medium (2), and if an (applied) university degree was completed, it was rated as high (3). Educational attainment did also not differ significantly between groups (F(2) = 0.4, p = 0.67). Executive functioning was estimated with the Trail Making Test (TMT; Corrigan & Hinkeldey, 1987). In the first part (TMT-A), numbers have to be connected accurately in sequential order as quickly as possible, whereas in the second part (TMT-B), numbers and letters have to be alternated. The score is calculated by subtracting the time to finish TMT-A from the time to finish TMT-B. There were no group differences for TMT score (F(2) = 1.06, p = 0.35). Verbal IQ was estimated using a German Vocabulary Test (Wortschatztest (WST); Schmidt & Metzler, 1992). Each question of the test contains a target word and five pseudo-words. The total number of correctly identified target words is used as the score, which can be converted to verbal IQ score. Verbal IQ did not differ significantly between groups (F(2) = 0.54, p = 0.59). Handedness was also homogenous across groups, as assessed with the Kruskal-Wallis test (handedness: H(2) = 0.03, p = 0.98)
Table 1. Demographic characteristics of the participants. Where relevant, the mean SD (range) is depicted. OC fNC oNC N 48 36 30 Age (years) 23.85 3.42 (18 - 31) 23.69 3.30 (19 - 33) 23.87 2.73 (19 - 30) Education (levels) 2.24 0.61 (1 - 3) 2.35 0.49 (2 - 3) 2.52 0.60 (1 - 3) TMT B-A (sec) 18.91 10.44 (3.23 - 43.16) 17.43 10.62 (1.97 - 41.32) 15.59 7.61 (3.79 - 35.95) Verbal IQ 106.04 7.70 (92 - 122) 107.64 9.13 (86 - 125) 105.83 7.36 (95 - 122) Right-handed 93.33% 94.12% 93.10%
TMT = Trail Making Test, SD = standard deviation 2.2 Hormonal assessment
Hormonal assessments took place on the day of the MRI measurement. EDTA blood samples of 18 ml were drawn and subsequently sent to the institute of Pharmaceutical Sciences Pharmaceutical (Bio-)Analysis of the University of Tübingen led by Prof. Lämmerhofer for analysis with mass spectrometry LC-MC. The following hormones were measured:
(1) Endogenous estradiol (E2)
(2) Exogenous estrogens (i.e. ethinylestradiol, exoE) (3) Endogenous progesterone (P)
(4) Exogenous progestins (exoP), both androgenic (i.e. levonorgestrel, n = 23, exoPa) and anti-androgenic (i.e. dienogest and chlormadinone acetate, n = 23, exoPaa
2.3 MRI data acquisition and analysis
High-resolution anatomical images were obtained using a 3T PRISMA scanner (Siemens Medical Solutions, Erlangen, Germany), at the University of Tübingen. MP2RAGE (3-D Magnetization Prepared 2 Rapid Gradient Echo) sequences were acquired consisting of 167 sagittal slices (TR / TE = 5000 / 2.98ms, TI = 700 / 2500 ms, 1 x 1 x 1 mm resolution, field of view (FOV) read 256 mm, FOV phase 93.8%, slice oversampling = 0%, flip angle [FA] = 4 / 5°). The MP2RAGE images were calculated by combining the three images with different inversion times, in order to get rid of the salt-and-pepper background noise. This denoising was done using a script of Jose Marques in MatLab (version 9.9.0; The Mathworks Inc., Sherborn, MA, USA) as reported in (O’Brien et al., 2014). See https://github.com /JosePMarques/MP2RAGE-related-scripts.
Automated ROI segmentation was performed with FreeSurfer (version 7.1.1) using the default processing stream (recon-all -all), and the flag -qcache was added to allow for rapid group analyses (http://surfer.nmr.mgh.harvard.edu/). The stages of this procedure have been described before in other work, but in brief it includes motion correction, removal of non-brain tissue, Talairach transformation, intensity normalization, definition of the GM and WM boundary, topology correction, and identification of GM, WM, and cerebrospinal fluid (Fischl et al., 2002). Manual quality control was performed with visual inspection in FreeView and consequently four brains were edited using gcut, a flag within FreeSurfer, and/or manual edits. ROI volumes were derived from the Desikan-Killiany atlas (Desikan et al., 2006).
2.4 Statistical analysis
All statistical analyses were executed in R (version 4.0.3). All parameters were tested for normality with a Shapiro-Wilk test and, when relevant, for equal variance between samples with a Bartlett’s test. These assumptions were not violated unless when differently stated. In order to address the first research question, i.e. differences in structural morphology between OC users and NC women, first the OC group was compared with the NC group by means of a mixed analysis of covariance (ANCOVA), with absolute GM volume as the dependent variable, group as a between-subjects independent variable, laterality as a within-subjects independent variable, and total intracranial volume (TIV) as a covariate to correct for brain size variation (Buckner et al., 2004). The ANCOVA was one-tailed for the hippocampus, because we had a directed hypothesis based on previous findings (Hertel et al., 2017; Lisofsky et al., 2016; Pletzer, 2019). Subsequently, group distinctions were made in the NC group, and the OC group, fNC group, and oNC group were compared in a cross-sectional design. A mixed ANCOVA was again used to establish these differences. Significant effects of the ANCOVAs were followed up by Tukey’s HSD post-hoc tests. Importantly, there were no GM volume outliers. Because all hormonal values were within the normal human range, none of the participants were removed.
In order to address the second research question, i.e. how endogenous and exogenous E2 and P are related to the structural morphology of the hippocampus and amygdala, first a separate dataset was created. Although the hormonal levels were within the normal human range, extreme hormonal values could pose a risk and influence the model fit. Now we were interested in the actual hormone levels and therefore all participants with hormone levels that exceeded 3 standard deviations (SD) above or below the group mean were excluded. Consequently, data from six women had to be excluded (four of the OC group, one of the fNC group, and one of the oNC group). Additionally, two exoP values were missing in the OC group, because their specific progestins were not tested for in the lab. Because hormonal levels were not normally distributed, a Kruskall-Wallis test was performed to test whether hormone levels differed between groups. Dunn’s test was used as a post-hoc comparison with Bonferroni correction. In the case of significant group differences, analyses would be conducted per group. Subsequently, partial correlations were executed to find which hormones correlated with GM volumes while controlling
for TIV. When the correlation coefficient was larger than 0.05 or smaller than –0.05, the particular hormone was included as a predictor in linear regression analyses, to assess whether there is a linear relationship between GM volumes and sex hormones, and to quantify the strength of this relationship. As such, various hormones were added into one single model, as sex hormones are known to interact (Edwards et al., 1968). TIV was always included as covariate and OC intake duration was an exploratory variable of interest. Results on OC intake duration are only reported when it makes a significant difference in the models. Intake duration was defined as the total number of months OC users used their current pill, and if their previous pill consisted of the same hormones, that duration was added. Note that age, IQ and educational attainment were not included in any of the analyses, because we tried to keep the models and analyses simple and straightforward in this exploratory analysis. The threshold for statistical significance was set at p < 0.05.
3. Results
3.1 Hormonal differences between groups
None of the data on hormones, except for exoPa, was normally distributed; the median and interquartile range (IQR) for each group is reported in Table 2. The serum levels of the endogenous hormones were within the normal human range (Stricker et al., 2006). Endogenous hormone levels differed significantly between groups, as assessed with a Kruskall-Wallis test
(E2: H(2) = 75.81, p = 2.2e-16; P: H(2) = 54.86, p = 1.23e-12; see Figure 2a and b). The oNC group had significantly higher E2 levels compared to both the OC group and fNC group (OC: p = 1.13e-17; fNC: p = 5.26e-05). The fNC group had significantly higher E2 levels than the OC group (p = 1.56e-05). In terms of P, fNC and oNC did not differ significantly (p = 1), but the OC group had significantly lower values compared to both groups (oNC: p = 1.33e-09; fNC: p = 1.90e-09).
When adding the endogenous and exogenous hormone levels in the OC groups, the three groups still differed significantly in their hormonal profiles (E2 + exoE: H(2) = 54.36, p = 1.57e-12; P + exoP: H(2) = 66.79, p = 3.14e-15; see Figure 2c and d). The oNC group still had significantly higher E2 levels compared to the E2 + exoE levels of the OC group (p = 2.17e-11), but the OC group had no longer lower E2 + exoE levels compared to the E2 levels in the fNC group (p = 1). Interestingly, now the OC group had significantly higher P + exoP levels compared to the P levels of both other groups (oNC: p = 3.54e-09; fNC: p = 2.00e-13).
3.1.1 Effect of androgenicity
While E2 and E2 + exoE levels did not differ significantly between OC users taking androgenic vs. anti-androgenic OCs, as assessed with the Wilcoxon rank-sum test (E2: W = 191.5, p = 0.28; E2 + exoE: W = 122, p = 0.27), all other hormone levels differed significantly (exoE: W = 137, p = 0.005; P: W = 127, p = 0.002; exoP: W = 61, p 1.58e-06; P + exoP: W = 61, p = 1.58e-06). In all cases, the hormonal levels were lower in the androgenic pill group. In Figure 3, these results are shown for E2 + exoE (a) and P + exoP (b).
Table 2. Serum hormonal levels (pg/mL). Median (IQR) is depicted, except for exoPa, for which mean (SD) is depicted.
OC fNC oNC
Estradiol (E2) 4.71 (2.16) 2.68e+01 (1.60e+01) 1.21e+02 (1.20e+02)
Ethinylestradiol (exoE) 1.83e+01 (1.52e+01)
Progesterone (P) 3.09e+01 (1.50e+01) 9.36e+01 (1.39e+02) 1.90e+02 (7.18e+02) Androgenic progestin (exoPa) 3.16e+03 (1.88e+03)
Anti-androgenic progestin (exoPaa) 1.58e+04 (1.31e+04)
Figure 3. Estradiol + ethinylestradiol (a) and progesterone + progestin (b) levels are displayed, separated by androgenicity of the progestin. Median and IQR are shown.
3.2 Differences in GM volumes between OC users and NC women
In order to assess GM volumes between OC users and NC women, first the OC group was compared with the NC group. Subsequently, group distinctions were made in the NC group based on the cycle phase, and the OC group, fNC group, and oNC group were compared in a cross-sectional design.
3.2.1 OC users and NC women
Hippocampus. As assessed with a one-sided
mixed ANCOVA with hippocampal GM volume as the dependent variable, group as the between-subjects variable, laterality as a within variable, and TIV as a covariate, the hippocampal GM volume of OC users was significantly smaller than that of NC women (F(1, 111) = 3.48, p = 0.03; see Figure 4a). This significant difference was independent of laterality (F(1, 111) = 0.44, p = 0.51).
Amygdala. In the amygdala, no significant
differences in GM volume between OC users and NC women were found, which was assessed with a two-sided mixed ANCOVA (F(1, 111) = 1.08, p = 0.30; see Figure 4b). See Table 3 for an overview of GM volumes per group.
3.2.2 OC users and NC women in their early follicular and periovulatory phase
Subsequently, we looked at whether OC users differ from NC women in different menstrual phases (i.e. fNC and oNC, see Figure 4c and d).
Hippocampus. When running a two-sided
mixed ANCOVA, as described in the statistical analysis section, no group differences in hippocampal GM volume were found (F(2, 110) = 1.73, p = 0.18).
Amygdala. Similarly, running the analyses for
the amygdala revealed no significant difference between any of the groups for amygdala GM volume (F(2, 110) = 0.65, p = 0.53).
Table 3. Brain volumes for each group. Mean SD is depicted.
OC fNC oNC Hippocampus Left 3767.88 346.07 3913.78 320.95 3884.99 334.37 Right 3771.14 305.30 3889.98 347.20 3960.99 334.03 Amygdala Left 1574.20 202.98 1614.64 190.99 1659.13 174.80 Right 1524.31 161.70 1553.98 203.63 1580.68 191.59
Figure 4. Group differences between OC users and NC women in hippocampal (a) and amygdala (b) GM volume are displayed, as well as group differences between the OC group, fNC group, and oNC group in hippocampal (c) and amygdala (d) GM volume. Means and standard error (SE) are shown.
3.3 Relationship of endogenous and exogenous hormones to GM volumes
Due to the significant group differences in hormonal levels (see section 3.1 Hormonal differences between groups), regression analyses were conducted separately for each group. As mentioned in the methods section, first a separate dataset was created where 6 participants had to be excluded due to their extreme hormonal levels. See Table 4 for the new hormonal levels.
GM volumes of the left and right hemispheres were then combined because they correlated highly, both with regards to the hippocampus (r(106) = 0.81, p = 2.2e-16) and with regards to the amygdala (r(106) = 0.69, p = 2.32e-16). Results are described for the combined ROIs as well as for left and right separately.
Partial correlations were executed to analyze whether hormones correlated with GM volumes while controlling for TIV. Supplementary table 1 (see Appendix) displays these correlations for all hormones. None of the partial correlations were significant. Subsequently, hormone levels with a correlation coefficient larger than 0.05 or smaller than -0.05 (from Supplementary table 1) were included as predictors in linear regression models, to test whether hormone levels can predict GM volumes in the hippocampus and amygdala. Supplementary tables 2-5 (see Appendix) depict the results for the linear regression models. All hormones with sufficiently large correlation coefficients were included into a single regression model due to the known hormonal interactions (Edwards et al., 1968).
Hippocampus. None of the endogenous or
exogenous hormones were associated with hippocampal GM volumes, in any of the groups.
Amygdala. E2 + exoE was positively related to
the right amygdala GM volume in the OC group (see Figure 5a). This was only the case when the duration of OC intake was taken into account, however. When one leverage point (i.e. E2 + exoE value) was removed, this association was no longer significant (see Figure 5b). Moreover, P was positively related to both total and left amygdala GM volume in the fNC group (see Figure 6a). E2 was also included in this model, but the significant relationship remained when E2 was removed from the model. When one or two leverage points (i.e. P values) were removed from the model however, because they exceeded 3 SD above the mean, this relationship no longer existed (see Figure 6b and c).
3.3.1 Effect of intake duration
Due to the significant effect of E + exoE in combination with OC intake duration on the right amygdala GM volume, here we disentangle this variable of interest further. In Figure 7, the GM volumes for both ROIs can be seen over the course of intake duration, as well as the influence of androgenicity on these changes. Androgenicity was added as an additional variable of interest, as we expected that particularly women with a longer OC intake duration would be taking OCs with androgenic progestin, because those have been on the market for longer.
For the hippocampus, both androgenic and anti-androgenic progestins were related to larger hippocampal volumes, albeit not significantly (r(42) = 0.22, p = 0.15). For the amygdala, diverging effects can be seen: androgenic progestins are associated with a decrease in amygdala GM volume, whereas anti-androgenic progestins are associated with a very small increase. Both associations were not significant (androgenic: r(20) = -0.19, p = 0.38; anti-androgenic: r(18) = 0.03, p = 0.89). A Meng’s test, using an online tool ((http://vassarstats.net/rdiff.html) revealed that these correlations did not differ significantly from each other (z = -0.69, p = 0.49).
The fact that hippocampal GM volumes were found to be smaller in OC users compared to NC women but apparently increase with longer OC use seem contradictory. Therefore, we split intake duration into two groups around the median (57.5 months; see Figure 9). As such, we could compare NC women (intake duration of 0 months) with OC users with a short intake duration (6 – 56 months) and OC users with a longer intake duration (more than 57 months). Although no group difference was found for the amygdala when controlling for TIV (F(2) = 0.17, p = 0.84), the GM volume difference of the hippocampus approached significance (F(2) = 22.47, p = 0.09). A Tukey-HSD post hoc comparison revealed that this significance would emerge between OC users with short intake duration and NC women, with OC having smaller GM volume (p = 0.08).
Table 4. Serum hormonal levels (pg/mL) after removal of 6 participants with extreme hormone levels. Median (IQR) is depicted for most hormones. For E2 (fNC), exoPa, and exoPaa (OC) mean (SD) is depicted, because those were normally distributed.
OC fNC oNC
Estradiol (E2) 4.81 (1.87) 2.83e+1 (1.06e+01) 1.26e+02 (1.22e+02)
Ethinylestradiol (exoE) 1.83e+1 (1.52e+01)
Progesterone (P) 2.98e+01 (1.38e+01) 9.15e+1 (1.25e+02) 1.85e+02 (6.88e+02) Androgenic progestin (exoPa) 3.25e+03 (1.88e+03)
Anti-androgenic progestin (exoPaa) 1.82e+05 (1.31e+04)
Figure 5. (a) Linear regression model examining the effect of E2 + exoE on amygdala GM volume. (b) Linear regression model after removal of one leverage point.
Figure 6. (a) Linear regression model examining the effect of P on amygdala GM volume. (b-c) Linear regression model after removal of one (b) and two (c) leverage points. (d) Hippocampal GM volume for NC women (0 months of intake duration), OC women with short intake duration (6 – 56 months) and OC women with longer intake duration (more than 57 months).
Figure 7. The change in GM volumes for the hippocampus (a) and amygdala (b) over the time of OC intake duration. Androgenic and anti-androgenic progestins are presented separately.
4. Discussion
In this study, we investigated whether brain volumes differ between women taking oral contraceptives (OCs) vs. naturally cycling women (NC), who were either measured during their early follicular phase (fNC) or during their periovulatory phase (oNC). Moreover, we were particularly interested in the contribution of endogenous and exogenous hormones on brain volumes. This is to our knowledge the first
study to examine these effects of exogenous hormones. We focused on two brain regions that are marked with a high density of receptors for both estradiol (E2) and progesterone (P) and have been shown previously to be affected by OC intake, i.e. the hippocampus and the amygdala.
For the hippocampus, we found a significant group difference with OC users having smaller hippocampal GM volumes than NC women. This is in accordance with previous findings; hippocampal GM volumes have repeatedly been reported to be smaller in
OC users (Hertel et al., 2017; Lisofsky et al., 2016; Pletzer, 2019). Smaller hippocampi have been related to depression (Sheline et al., 2002), which is also an adverse effect of OC use, particularly among adolescent users (Skovlund et al., 2016). The change in hippocampal volume with OC use could therefore be related to those depressive effects. When disentangling this difference of OC users and NC women further by considering the different cycle phases of the NC women, the significant group difference diminished. This is in contrast with our expectations, as the hippocampal volume of women in their ovulatory phase (i.e. high E2 levels) had previously been reported to be larger than that of women in low E2 phases (Lisofsky et al., 2015; Pletzer et al., 2018; Protopopescu et al., 2008). We did not replicate this, as we did not find higher GM volumes in the oNC group compared to the other groups. This lack of replication might be due to the fact that our participants in their periovulatory phase had relatively low E2 levels (Stricker et al., 2006). Nevertheless, we used a relatively new technique to assess hormone levels, which might simply estimate lower hormone levels compared to more conventional techniques (Salvagno et al., 2017). Another possibility is that the total intracranial volume (TIV), which was added as a covariate in our mixed ANCOVAs, accounted for the shared variance between the hippocampal volume and the total intracranial volume. Although TIV did not differ between the groups, it was significantly correlated to the hippocampal volume. When removing TIV from the analyses, significant group differences indeed emerged as expected.
Moreover, a trend for larger hippocampi with longer intake duration was observed, independent of androgenicity, similar to Pletzer et al. (2019). Although this seems contradictory with the aforementioned results, it seemed as if in the first months of OC intake a reduction in volume took place, which may become normalized with longer OC intake. This reduction might be due to the low levels of E2, as an E2-dependent increase in GM volume has been reported (Pletzer et al., 2018). Similarly, animal studies indicate that a decrease in GM volume might arise due to a decrease in synaptic spine density, which could potentially arise because of the low E2 levels, as E2 induces the formation of new synapses (Woolley & McEwen, 1993). Alternatively, because P is known to downregulate synapses in the hippocampus, the high progestin levels might be at stake (McEwen, 2002). For the subsequent increase in GM volume, no evidence could be gathered. Perhaps the brain adapts to the new (endogenous + exogenous) hormonal levels.
No group differences were found for amygdala volume. As such, we did not replicate previous findings of smaller amygdala GM volumes with OC use (Lisofsky et al., 2016). However, Lisofsky et al. (2016) observed volume changes of OC users within the first months of OC use, which may not be comparable to the cross-sectional design of this study. Actually, none of the other studies that performed whole-brain cross-sectional group comparisons reported amygdala volume
differences between OC users and NC women (De Bondt et al., 2013; Petersen et al., 2015). However, androgenicity did influence the brain volumes over the course of OC intake. This might have been the reason for the lack of overall group differences; there was a trend for smaller amygdala volumes with androgenic OC use and a very small trend of larger GM volumes with anti-androgenic OC use. As such, it might have cancelled each other out when put together into one OC group.
Furthermore, we observed a number of associations between hormonal levels and brain volumes. However, they did not survive leverage point removal and should therefore be interpreted with caution. E2 + ethinylestradiol (exoE) levels were associated with greater GM volumes in the right amygdala in OC users. An association between amygdala volume and E2 has not been reported before; to our knowledge the only association between E2 levels and the amygdala in general is reduced activity during perception of emotional images at the time of ovulation, when E2 levels are high, although a direct association has not been established (Goldstein, 2005). Given the high density of receptors for E2 in the amygdala, however, the finding is not surprising.
Additionally, P levels were associated with greater amygdala volumes in the early follicular phase. It has previously been reported that women in the late luteal phase had larger amygdala volumes, and our findings point in the direction of a role for P in this brain volume difference (Ossewaarde et al., 2013). However, because P levels are generally low in the follicular phase, this association could have been too weak in the present study. It might be worth replicating these findings in a design including women in their luteal phase, when P levels are generally higher.
We did not find an association between E2 or exoE levels and hippocampal GM volume, which had previously been reported for E2 (Pletzer et al., 2018). As mentioned before, this could be due to the relatively low E2 and exoE levels in our sample (Stadel et al., 1980; Stricker et al., 2006).
The present study has some limitations. To start with, our sample consisted predominantly of highly educated women in a limited age range, resulting in a relatively young sample with high intelligence (Giovagnoli et al., 1996). Even though these demographic characteristics did not differ between groups, intelligence and age have been related to GM volumes (Amat et al., 2008; Ge et al., 2002). As such, the sample might not be fully generalizable to the general population. Additionally, in the linear regression analyses (i.e. the second research question), intelligence scores and age were not included as covariates. Because the analyses were conducted per group, it could have been unbalanced within the groups.
Moreover, we propose some suggestions for future research. In addition to overcoming the limitations described above, mood could be an additional variable of interest. Particularly because smaller hippocampus and amygdala volumes have been related to negative mood and depression (Sheline et al.,
2002), and because these negative symptoms are also frequently observed among OC users (Skovlund et al., 2016), women who display greater GM volume changes, or who fail to normalize their brain volumes over the course of OC intake use may be more prone to these symptoms. Secondly, a longitudinal design could provide more insight into the effects of OC use on brain volume changes, and more specifically into GM volume changes driven by endogenous and exogenous hormones within participants. As the hormone levels varied greatly between participants, this would reveal the effects of actual hormone fluctuations. Thirdly, this study only included the hippocampus and amygdala as ROIs. Future research could attempt to explore the contribution of (exogenous) sex hormones to volume changes in other brain areas, such as the parahippocampal gyrus and prefrontal cortex, as OC-dependent differences have been reported in these brain areas as well (Lisofsky et al., 2016; Pletzer et al., 2010; van Wingen et al., 2011).
In conclusion, it seems as if both endogenous and exogenous hormones affect the brain volume changes as seen in women using OCs, although these exploratory and preliminary findings have to be interpreted with caution. When a relationship between sex hormones and brain volume changes would come to be replicated, preferably with the suggested improvements taken into account, this could be of importance to the mental health of the 151 million women worldwide who currently use OCs.
Acknowledgements
The author wishes to thank Prof. Dr. B. Derntl and A.C.S. Kimmig of the University of Tübingen for supervising me during this whole process and for reviewing this manuscript.
References
Amat, J. A., Bansal, R., Whiteman, R., Haggerty, R., Royal, J., & Peterson, B. S. (2008). Correlates of intellectual ability with morphology of the hippocampus and amygdala in healthy adults. Brain and Cognition, 66(2), 105–114.
https://doi.org/10.1016/j.bandc.2007.05.009
Barth, C., Steele, C. J., Mueller, K., Rekkas, V. P., Arélin, K., Pampel, A., Burmann, I., Kratzsch, J., Villringer, A., & Sacher, J. (2016). In-vivo Dynamics of the Human Hippocampus across the Menstrual Cycle. Scientific Reports, 6(1), 32833. https://doi.org/10.1038/srep32833
Bezemer, I. D., Verhamme, K. M. C., Gini, R., Mosseveld, M., Rijnbeek, P. R., Trifirò, G., Sturkenboom, M. C. J. M., Beest, F. J. A. P., & Herings, R. M. C. (2016). Use of oral contraceptives in three European countries: A population-based multi-database study. The European Journal of Contraception & Reproductive
Health Care, 21(1), 81–87.
https://doi.org/10.3109/13625187.2015.1102220 Boron, W. F., & Boulpaep, E. L. (2016). Medical Physiology. Elsevier Health Sciences.
Brinton, R. D., Thompson, R. F., Foy, M. R., Baudry, M., Wang,
J., Finch, C. E., Morgan, T. E., Pike, C. J., Mack, W. J., Stanczyk, F. Z., & Nilsen, J. (2008). Progesterone receptors: Form and function in brain. Frontiers in Neuroendocrinology, 29(2), 313– 339. https://doi.org/10.1016/j.yfrne.2008.02.001
Buckner, R. L., Head, D., Parker, J., Fotenos, A. F., Marcus, D., Morris, J. C., & Snyder, A. Z. (2004). A unified approach for morphometric and functional data analysis in young, old, and demented adults using automated atlas-based head size normalization: Reliability and validation against manual measurement of total intracranial volume. NeuroImage, 23(2), 724–738. https://doi.org/10.1016/j.neuroimage.2004.06.018 Corrigan, J. D., & Hinkeldey, N. S. (1987). Relationships between Parts A and B of the Trail Making Test. Journal of
Clinical Psychology, 43(4), 402–409.
https://doi.org/10.1002/1097-4679(198707)43:4<402::AID-JCLP2270430411>3.0.CO;2-E
De Bondt, T., Jacquemyn, Y., Van Hecke, W., Sijbers, J., Sunaert, S., & Parizel, P. M. (2013). Regional gray matter volume differences and sex-hormone correlations as a function of menstrual cycle phase and hormonal contraceptives use. Brain Research, 1530, 22–31. https://doi.org/10.1016/j.brainres.2013.07.034
Desikan, R. S., Ségonne, F., Fischl, B., Quinn, B. T., Dickerson, B. C., Blacker, D., Buckner, R. L., Dale, A. M., Maguire, R. P., Hyman, B. T., Albert, M. S., & Killiany, R. J. (2006). An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. NeuroImage, 31(3), 968–980.
https://doi.org/10.1016/j.neuroimage.2006.01.021
Edwards, D. A., Whalen, R. E., & Nadler, R. D. (1968). Induction of estrus: Estrogen-progesterone interactions. Physiology &
Behavior, 3(1), 29–33.
https://doi.org/10.1016/0031-9384(68)90027-9
Ertman, N., Andreano, J. M., & Cahill, L. (2011). Progesterone at encoding predicts subsequent emotional memory. Learning &
Memory, 18(12), 759–763.
https://doi.org/10.1101/lm.023267.111
Fanselow, M. S., & Dong, H.-W. (2010). Are the Dorsal and Ventral Hippocampus Functionally Distinct Structures?
Neuron, 65(1), 7–19.
https://doi.org/10.1016/j.neuron.2009.11.031 Fischl, B., Salat, D. H., Busa, E., Albert, M., Dieterich, M., Haselgrove, C., van der Kouwe, A., Killiany, R., Kennedy, D., Klaveness, S., Montillo, A., Makris, N., Rosen, B., & Dale, A. M. (2002). Whole Brain Segmentation: Automated Labeling of Neuroanatomical Structures in the Human Brain. Neuron,
33(3), 341–355.
https://doi.org/10.1016/S0896-6273(02)00569-X
Ge, Y., Grossman, R. I., Babb, J. S., Rabin, M. L., Mannon, L. J., & Kolson, D. L. (2002). Age-Related Total Gray Matter and White Matter Changes in Normal Adult Brain. Part I: Volumetric MR Imaging Analysis. American Journal of Neuroradiology, 23(8), 1327–1333.
Gingnell, M., Engman, J., Frick, A., Moby, L., Wikström, J., Fredrikson, M., & Sundström-Poromaa, I. (2013). Oral contraceptive use changes brain activity and mood in women with previous negative affect on the pill—A double-blinded, placebo-controlled randomized trial of a levonorgestrel-containing combined oral contraceptive.
Psychoneuroendocrinology, 38(7), 1133–1144.
https://doi.org/10.1016/j.psyneuen.2012.11.006
Giovagnoli, A. R., Mascheroni, S., Simoncelli, M., Laiacona, M., & Capitani, E. (1996). Trail Making Test: Normative values from287 normal adult controls. Italian Journal of Neurological
Sciences, 17, 305–309. https://doi.org/10.1007/BF01997792
Gogos, A., Wu, Y. C., Williams, A. S., & Byrne, L. K. (2014). The Effects of Ethinylestradiol and Progestins (“the pill”) on Cognitive Function in Pre-menopausal Women. Neurochemical
https://doi.org/10.1007/s11064-014-1444-6
Goldin, C., & Katz, L. F. (2002). The Power of the Pill: Oral Contraceptives and Women’s Career and Marriage Decisions.
Journal of Political Economy, 110(4), 730–770.
https://doi.org/10.1086/340778
Goldstein, J. M. (2005). Hormonal Cycle Modulates Arousal Circuitry in Women Using Functional Magnetic Resonance Imaging. Journal of Neuroscience, 25(40), 9309–9316. https://doi.org/10.1523/JNEUROSCI.2239-05.2005
Hertel, J., König, J., Homuth, G., Van der Auwera, S., Wittfeld, K., Pietzner, M., Kacprowski, T., Pfeiffer, L., Kretschmer, A., Waldenberger, M., Kastenmüller, G., Artati, A., Suhre, K., Adamski, J., Langner, S., Völker, U., Völzke, H., Nauck, M., Friedrich, N., & Grabe, H. J. (2017). Evidence for Stress-like Alterations in the HPA-Axis in Women Taking Oral Contraceptives. Scientific Reports, 7(1), 14111. https://doi.org/10.1038/s41598-017-13927-7
LeDoux, J. E. (1993). Emotional memory systems in the brain.
Behavioural Brain Research, 58(1), 69–79.
https://doi.org/10.1016/0166-4328(93)90091-4 Lisofsky, N., Mårtensson, J., Eckert, A., Lindenberger, U., Gallinat, J., & Kühn, S. (2015). Hippocampal volume and functional connectivity changes during the female menstrual cycle. NeuroImage, 118, 154–162.
https://doi.org/10.1016/j.neuroimage.2015.06.012
Lisofsky, N., Riediger, M., Gallinat, J., Lindenberger, U., & Kühn, S. (2016). Hormonal contraceptive use is associated with neural and affective changes in healthy young women.
NeuroImage, 134, 597–606.
https://doi.org/10.1016/j.neuroimage.2016.04.042
McEwen, B. (2002). Estrogen actions throughout the brain. Recent progress in hormone research, 57, 357-384. Montoya, E. R., & Bos, P. A. (2017). How Oral Contraceptives Impact Social-Emotional Behavior and Brain Function. Trends
in Cognitive Sciences, 21(2), 125–136.
https://doi.org/10.1016/j.tics.2016.11.005
Nielsen, S. E., Ertman, N., Lakhani, Y. S., & Cahill, L. (2011). Hormonal contraception usage is associated with altered memory for an emotional story. Neurobiology of Learning and
Memory, 96(2), 378–384.
https://doi.org/10.1016/j.nlm.2011.06.013
O’Brien, K. R., Kober, T., Hagmann, P., Maeder, P., Marques, J., Lazeyras, F., Krueger, G., & Roche, A. (2014). Robust T1-Weighted Structural Brain Imaging and Morphometry at 7T Using MP2RAGE. PLOS ONE, 9(6), e99676.
https://doi.org/10.1371/journal.pone.0099676
Ossewaarde, L., Wingen, G. A. van, Rijpkema, M., Bäckström, T., Hermans, E. J., & Fernández, G. (2013). Menstrual cycle-related changes in amygdala morphology are associated with changes in stress sensitivity. Human Brain Mapping, 34(5), 1187–1193. https://doi.org/10.1002/hbm.21502
Österlund, M., G.J.M. Kuiper, G., Gustafsson, J.-Å., & Hurd, Y. L. (1998). Differential distribution and regulation of estrogen receptor-α and -β mRNA within the female rat brain1First published on the World Wide Web on 10 December 1997.1.
Molecular Brain Research, 54(1), 175–180.
https://doi.org/10.1016/S0169-328X(97)00351-3 Packard, M. G., & Teather, L. A. (1997). Intra-hippocampal estradiol infusion enhances memory in ovariectomized rats.
NeuroReport, 8(14), 3009–3013.
Petersen, N., & Cahill, L. (2015). Amygdala reactivity to negative stimuli is influenced by oral contraceptive use. Social
Cognitive and Affective Neuroscience, 10(9), 1266–1272.
https://doi.org/10.1093/scan/nsv010
Petersen, N., Touroutoglou, A., Andreano, J. M., & Cahill, L. (2015). Oral contraceptive pill use is associated with localized decreases in cortical thickness. Human Brain Mapping, 36(7), 2644–2654. https://doi.org/10.1002/hbm.22797
Pletzer, B. (2019). Sex Hormones and Gender Role Relate to
Gray Matter Volumes in Sexually Dimorphic Brain Areas.
Frontiers in Neuroscience, 13, 592.
https://doi.org/10.3389/fnins.2019.00592
Pletzer, B., Harris, T., & Hidalgo-Lopez, E. (2018). Subcortical structural changes along the menstrual cycle: Beyond the hippocampus. Scientific Reports, 8(1), 16042.
https://doi.org/10.1038/s41598-018-34247-4
Pletzer, B., Harris, T., & Hidalgo-Lopez, E. (2019). Previous contraceptive treatment relates to grey matter volumes in the hippocampus and basal ganglia. Scientific Reports, 9(1), 11003. https://doi.org/10.1038/s41598-019-47446-4
Pletzer, B., Harris, T.-A., Scheuringer, A., & Hidalgo-Lopez, E. (2019). The cycling brain: Menstrual cycle related fluctuations in hippocampal and fronto-striatal activation and connectivity during cognitive tasks. Neuropsychopharmacology, 44(11), 1867–1875. https://doi.org/10.1038/s41386-019-0435-3 Pletzer, B., Kronbichler, M., Aichhorn, M., Bergmann, J., Ladurner, G., & Kerschbaum, H. H. (2010). Menstrual cycle and hormonal contraceptive use modulate human brain structure.
Brain Research, 1348, 55–62.
https://doi.org/10.1016/j.brainres.2010.06.019
Protopopescu, X., Butler, T., Pan, H., Root, J., Altemus, M., Polanecsky, M., McEwen, B., Silbersweig, D., & Stern, E. (2008). Hippocampal structural changes across the menstrual cycle.
Hippocampus, 18(10), 985–988.
https://doi.org/10.1002/hipo.20468
Rabe, T., Nitsche, D. C., & Runnebaum, B. (1997). The effects of monophasic and triphasic oral contraceptives on ovarian function and endometrial thickness. The European Journal of
Contraception & Reproductive Health Care, 2(1), 39–51.
https://doi.org/10.1080/13625189709049933
Salvagno, G. L., Danese, E., & Lippi, G. (2017). Preanalytical variables for liquid chromatography-mass spectrometry (LC-MS) analysis of human blood specimens. Clinical Biochemistry,
50(10), 582–586.
https://doi.org/10.1016/j.clinbiochem.2017.04.012
Sander, D., Grafman, J., & Zalla, T. (2003). The human amygdala: an evolved system for relevance detection. Reviews in the Neurosciences, 14(4), 303-316.
Schmidt, K. H., and Metzler, P. (1992). Wortschatztest: WST. Weinheim: Beltz.
Sheline, Y. I., Mittler, B. L., & Mintun, M. A. (2002). The hippocampus and depression. European Psychiatry, 17, 300– 305. https://doi.org/10.1016/S0924-9338(02)00655-7 Sitruk-Ware, R. (2006). New progestagens for contraceptive use. Human Reproduction Update, 12(2), 169–178. https://doi.org/10.1093/humupd/dmi046
Skovlund, C. W., Mørch, L. S., Kessing, L. V., & Lidegaard, Ø. (2016). Association of Hormonal Contraception With Depression. JAMA Psychiatry, 73(11), 1154–1162. https://doi.org/10.1001/jamapsychiatry.2016.2387
Stadel, B. V., Sternthal, P. M., Schlesselman, J. J., Douglas, M. B., Hall, W. D., Kaul, L., & Ahluwalia, B. (1980). Variation of ethinylestradiol blood levels among healthy women using oral contraceptives. Fertility and Sterility, 33(3), 257–260.
https://doi.org/10.1016/s0015-0282(16)44589-9
Stricker, R., Eberhart, R., Chevailler, M.-C., Quinn, F. A., Bischof, P., & Stricker, R. (2006). Establishment of detailed reference values for luteinizing hormone, follicle stimulating hormone, estradiol, and progesterone during different phases of the menstrual cycle on the Abbott ARCHITECT® analyzer. Clinical
Chemistry and Laboratory Medicine (CCLM), 44(7), 883–887.
https://doi.org/10.1515/CCLM.2006.160
United Nations. (2019). Contraceptive Use by Method 2019:
Data Booklet. UN. https://doi.org/10.18356/1bd58a10-en
van Wingen, G. A., Ossewaarde, L., Bäckström, T., Hermans, E. J., & Fernández, G. (2011). Gonadal hormone regulation of the emotion circuitry in humans. Neuroscience, 191, 38–45. https://doi.org/10.1016/j.neuroscience.2011.04.042
van Wingen, G. A., van Broekhoven, F., Verkes, R. J., Petersson, K. M., Bäckström, T., Buitelaar, J. K., & Fernández, G. (2008). Progesterone selectively increases amygdala reactivity in women. Molecular Psychiatry, 13(3), 325–333.
https://doi.org/10.1038/sj.mp.4002030
Woolley, C. S., & McEwen, B. S. (1993). Roles of estradiol and progesterone in regulation of hippocampal dendritic spine density during the estrous cycle in the rat. Journal of
Comparative Neurology, 336(2), 293–306.
Appendix
Supplementary table 1. Partial correlations between hormones and GM volumes for the hippocampus and amygdala, while controlling for TIV. Correlation coefficient (p-value) is depicted.
Hippocampus Hippocampus left Hippocampus right Amygdala Amygdala left Amygdala right
OC group E2 0.072 (0.55) 0.068 (0.57) 0.047 (0.69) 0.078 (0.51) -0.029 (0.81) 0.107 (0.37) exoE 0.047 (0.66) 0.139 (0.17) 0.024 (0.82) -0.073 (0.49) -0.112 (0.29) 0.005 (0.96) E2 + exoE 0.118 (0.33) 0.157 (0.18) 0.073 (0.55) -0.007 (0.95) -0.109 (0.36) 0.130 (0.28) P -0.065 (0.54) -0.044 (0.67) -0.101 (0.34) 0.065 (0.54) 0.095 (0.37) -0.018 (0.86) exoP -0.025 (0.82) -0.056 (0.59) -0.008 (0.94) -0.093 (0.39) -0.100 (0.36) -0.004 (0.97) exoPa 0.205 (0.37) 0.248 (0.28) 0.138 (0.55) -0.099 (0.67) -0.121 (0.60) -0.049 (0.83) exoPaa -0.102 (0.90) -0.163 (0.51) -0.023 (0.93) -0.035 (0.89) -0.057 (0.82) 0.002 (0.99) P + exoP -0.027 (0.80) -0.059 (0.58) -0.010 (0.93) -0.095 (0.38) -0.102 (0.35) -0.006 (0.96) fNC group E2 -0.063 (0.72) -0.134 (0.45) 0.012 (0.95) 0.093 (0.60) 0.108 (0.54) 0.064 (0.72) P 0.164 (0.17) 0.130 (0.28) 0.163 (0.18) 0.162 (0.18) 0.143 (0.23) 0.146 (0.23) oNC group E 0.054 (0.67) 0.024 (0.86) 0.111 (0.41) 0.261 (0.17) 0.240 (0.22) 0.239 (0.22) P 0.057 (0.67) 0.157 (0.24) -0.041 (0.76) -0.045 (0.74) -0.133 (0.32) 0.083 (0.54)
E2 = estradiol, exoE = ethinylestradiol, P = progesterone, exoP = progestin, exoPa = androgenic progestin, exoPaa = anti-androgenic progestin
Supplementary table 2. Linear regression models examining the effects various hormones on hippocampal GM volume for the OC group. The regression formula always included TIV.
BIC Intercept (SE) TIV (SE) E2 (SE) E2 + exoE (SE) P (SE) exoP (SE) P + exoP (SE)
Total hippocampus
E2 + P + exoPa 325.77 4.542e+03 (3.318e+03) 2.194e-03 (2.326e-03) 2.120 (4.981) -5.487 (1.553e+01) 3.288e-02 (1.007e-01) E2 + P + exoPaa 204.24 7.953e+03 (2.301e+03) ** -1.696e-03 (1.921e-03) 6.442e+01 (4.010e+01) 2.527e+01 (1.137e+01) 3.073e-02 (1.366e-02) E2 + exoE + P + exoPa 325.67 4.503e+03 (3.315e+03) 2.217e-03 (2.320e-03) 2.387 (4.852) -5.527 (1.543e+01) 2.113e-02 (9.306e-02) E2 + exoE + P + exoPaa 204.77 6.943e+03 (2.314e+03) * -6.870e-04 (1.804e-03) 2.040e+01 (1.389e+01) 1.935e+01 (1.012e+01) 9.352e-03 (1.342e-02)
Left hippocampus
E2 + exoE + exoP 483.68 2.330e+03 (1.034e+03) * 9.748e-04 (7.282e-04) 1.975 (2.268) 3.330e-03 (5.331e-03)
E2 + exoE + P + exoP 483.68 2.330e+03 (1.034e+03) * 9.745e-04 (7.282e-04) 1.975 (2.268) 3.332e-03 (5.330e-03)
Right hippocampus
E2 + exoE + P + exoPa 297.62 2.881e+03 (1.643e+03) 7.270e-04 (1.150e-03) 9.509e-01 (2.405) -3.436 (7.649) -4.236e-03 (4.613e-02)
BIC = Bayesian Information Criterion, SE = standard error, TIV = total intracranial volume, E2 = estradiol, exoE = ethinylestradiol, P = progesterone, exoP = progestin, exoPa = androgenic progestin, exoPaa = anti-androgenic progestin; * p < 0.05
Supplementary table 3. Linear regression models examining the effects various hormones on amygdala GM volume for the OC group. The regression formula always included TIV.
BIC Intercept (SE) TIV (SE) E2 (SE) exoE (SE) E2 + exoE (SE) P (SE) exoP (SE) P + exoP (SE)
Total amygdala
E2 + exoE + P + exoP 497.00 2.174e+03 (1.228e+03) 6.174e-04 (8.933e-04) 1.740 (2.642) -1.621 (7.319) 3.209 (4.458) 2.972e-03 (7.947e-03) E2 + exoE + P + exoPa 296.09 2.869e+03 (1.532e+03) 2.805e-04 (1.081e-03) 1.427 (2.293) -2.875e+01 (1.695e+01) -1.409 (7.250) 1.339e-01 (1.106e-01)
E2 + exoE + P + exoP 497.00 2.174e+03 (1.228e+03) 6.174e-04 (8.933e-04) 1.740 (2.642) -1.621 (7.319) 3.209 (4.458) 2.972e-03 (7.947e-03)
Left amygdala
exoE + P + exoP 576.07 8.380e+02 (4.940e+02) 5.720e-04 (3.479e-04) -1.069 (3.657) -2.127e-03 (3.427e-03) E2 + exoE + P + exoP 461.45 9.960e+02 (7.386e+02) 4.609e-04 (5.200e-04) -6.457e-01 (1.620) -1.162e-03 (3.806e-03)
Right amygdala
E2 + exoE + intake duration 464.10 4.858e+02 (4.583e+02) 7.338e-04 (3.243e-04) * 2.630 (1.193) * E2 463.04 5.439e+02 (4.717e+02) 7.042e-04 (3.303e-04) * 1.825 (1.134)
BIC = Bayesian Information Criterion, SE = standard error, TIV = total intracranial volume, E2 = estradiol, exoE = ethinylestradiol, P = progesterone, exoP = progestin, exoPa = androgenic progestin, exoPaa = anti-androgenic progestin; * p < 0.05
Supplementary table 4. Linear regression models examining the effects various hormones on hippocampal GM volumes for the NC groups. The regression formula always included TIV.
BIC Intercept (SE) TIV (SE) E2 (SE) P (SE)
fNC group
Total E2 + P 562.09 5.766e+03 (1.431e+03) *** 1.494e-03 (9.811e-04) -4.653 (1.053e+01) 6.495e-02 (2.016e-01) Left E2 + P 515.57 3.065e+03 (7.362e+02) *** 6.603e-04 (5.048e-04) -3.514 (5.420) -1.755e-02 (1.037e-01) Right P 517.20 2.668e+03 (7.620e+02) ** 8.359e-04 (5.349e-04) 7.490e-02 (1.031e-01)
oNC group
Total E2 + P 469.12 4.520e+03 (1.918e+03) * 2.131e-03 (1.273e-03) 1.000 (1.598) 1.377e-01 (1.130e-01) Left P 429.31 2.615e+03 (9.459e+02) * 8.375e-04 (6.506e-04) 7.633e-02 (5.683e-02) Right E2 429.32 2.227e+03 (9.819e+02) * 1.153e-03 (6.599e-04) 4.657e-01 (8.036e-01)
BIC = Bayesian Information Criterion, SE = standard error, TIV = total intracranial volume, E2 = estradiol, P = progesterone; * p < 0.05, ** p < 0.01, *** p < 0.001
Supplementary table 5. Linear regression models examining the effects various hormones on amygdala GM volumes for the NC groups. The regression formula always included TIV.
BIC Intercept (SE) TIV (SE) E2 (SE) P (SE)
fNC group
Total E2 + P 520.53 1.992e+03 (7.903e+02) * 8.092e-04 (5.418e-04) -1.883 (5.818) 2.745e-01 (1.114e-01) * Left E2 + P 472.58 1.035e+03 (3.984e+02) * 3.977e-04 (2.731e-04) -1.247 (2.933) 1.732e-01 (5.614e-02) ** Right E2 + P 484.61 9.575e+02 (4.731e+02) 4.114e-04 (3.244e-04) -6.358e-01 (3.483) 1.013e-01 (6.666e-02)
oNC group
Total E2 426.37 6.458e+02 (9.331e+02) 1.693e-03 (6.271e-04) * 1.046 (7.638e-01)
Left E2 + P 392.22 5.463e+02 (5.094e+02) 7.362e-04 (3.382e-04) * 4.423e-01 (4.242e-01) -1.668e-02 (3.001e-02) Right E2 + P 394.52 2.178e+01 (5.300e+02) 9.914e-04 (3.519e-04) ** 6.894e-01 (4.414e-01) 3.878e-02 (3.122e-02)
