In a recent meta-analysis of 37 studies and 1004 participants, we reported a significant decrease in cardiac vagal activity from the follicular to the luteal phase [
28]. However, the question remained which of the two ovarian hormones (E2, P4, or the interaction of both) is associated with these cyclical HRV changes. Here, we presented two longitudinal studies on the unique roles of E2 and P4 in menstrual cycle-related changes in HRV. The two studies share their repeated assessment of ovarian hormones and HRV in naturally-cycling participants while they differ in various design aspects (most prominently, that study one assessed the effects of hormones on HRV regardless of cycle phase, while study two assessed these effects at very specific cycle phases). As the focus of the present work is on ovarian hormones (as opposed to cycle phase, given that recent meta-analytic results are already available on this topic [
28]), these design differences allow for testing the effects of the hormones themselves and the robustness of these effects.
In line with the results of our recent meta-analysis [
28], study one found reduced HRV in the midluteal phase relative to the mid-follicular phase. Analyses in study one, which used backward-counting (a less precise method for identifying the ovulatory phase), also revealed that the ovulatory phase was associated with higher HRV than the midluteal phase. However, in study two, which scheduled participants for ovulatory visits only following a positive urine LH test, we did not find a difference in HRV between the ovulatory and midluteal phases. This may be because this more precise measure led to consistently post-ovulatory visits in study two, during which P4 is already rising and potentially exerting effects on HRV, which may therefore diminish the contrast between the ovulatory and midluteal phases in study two. Also, both studies revealed significantly lower HRV levels in the mid-luteal compared to the perimenstrual phase. Therefore, our data are consistent with our first hypothesis, that the midluteal phase is associated with a general reduction in HRV.
The goal of the present paper was to test (and conceptually replicate) the within-person associations of endogenous E2 and P4 with HRV. Consistent with our hypotheses, and with the midluteal phase reduction in HRV, within-person fluctuations in P4 (and not E2 or their interaction) were significantly associated with (lower) HRV in both studies. Of note, both studies also found significant random slopes for the influence of P4 on HRV, indicating possible individual differences in these associations that should be explored further in larger samples. Our confidence in the validity of these associations is bolstered by our ability to conceptually replicate these P4-HRV associations across two studies with differing methods.
4.1. Possible Underlying Mechanisms
Our recent meta-analysis demonstrated a significant decrease in cardiac vagal activity from the follicular to luteal phase [
28]. The results of the present work suggest that it is likely high levels of P4 (rather than P4 withdrawal) that lead to decreased midluteal HRV. Although the observational study design of the present studies (as opposed to experiments) does not allow for firm conclusions about causal underlying mechanisms, several possible pathways are discussed below.
The ANS is composed of the peripheral sympathetic and parasympathetic branches and the central autonomic network (i.e., a network of brain areas involved in autonomic control; CAN), which jointly regulate HRV. As reviewed by Thayer and colleagues [
4,
12,
13], the CAN includes the prefrontal, cingulate, and insula cortices and the amygdala. Animal research on ovarian hormones has revealed that P4 receptors are broadly expressed throughout the brain, including CAN areas important for HRV regulation like the amygdala and the frontal cortex [
46,
47,
48]. It is therefore possible that menstrual cycle-related changes in P4 may cause HRV fluctuations by acting in certain CAN areas.
In general, there are two primary central mechanisms by which P4 can act: (1) ‘classical’ P4 receptor-mediated pathways (genomic effects), and (2) alternate ‘non-genomic’ mechanisms [
46]. These non-genomic mechanisms generally refer to the metabolism of P4 to neuroactive steroids, of which allopregnanolone (ALLO) and pregnanolone are among the two most studied [
48]. In naturally-cycling individuals, plasma levels of ALLO reach approximately 0.2–0.5 nmol/L in the follicular and up to 4 nmol/L in the luteal menstrual cycle phase [
49]. ALLO modulates the GABA system by acting as a positive allosteric modulator of GABA-A receptors. The common effects of ALLO and other positive modulators of GABA-A receptor (e.g., alcohol, benzodiazepines, barbiturates) are inhibitory (e.g., anesthetic, sedative, anticonvulsant, anxiolytic) [
49,
50]. Research on the effects of alcohol, another positive allosteric modulator of the GABA-A receptor, on HRV indicates that acute alcohol ingestion quickly reduces HRV [
51]. Further, alcohol-induced reductions in HRV are thought to be mediated by the GABA system [
52], and experimental animal data suggests that alcohol modulates GABA-mediated cardiovascular control mechanisms in the brainstem [
53]. Future studies should investigate whether these two modulators of the GABA-A receptor (alcohol and ALLO) effect HRV via similar mechanisms.
In addition to the anesthetic, sedative, anticonvulsant, and anxiolytic effects of P4 metabolites such as ALLO, numerous studies have shown that the effects of the positive modulators of GABA-A receptor can also be exactly the opposite (i.e., paradoxical). These paradoxical effects (which can take the form of negative mood, anxiety, irritability or aggression) severely affect 3 to 6% of human subjects and moderately affect 25% [
49,
50]. Research on the emotional effects of the menstrual cycle suggests that affective symptoms experienced in the luteal phase of the menstrual cycle by individuals with premenstrual dysphoric disorder (PMDD) may be caused by the paradoxical effects of ALLO mediated by the GABA-A receptor [
49]. Thus, individuals seem to differ in the extent to which paradoxical effects of ALLO cause affective symptoms in the luteal phase. Future research should investigate whether, (and if so, for whom) paradoxical effects of ALLO might also serve as an underlying mechanism for P4-related HRV reduction in the luteal phase.
Given that the amygdala (which is involved in HRV regulation as part of the CAN [
13]) is rich in GABA-A receptors [
54], possible P4-induced changes in the activity of the amygdala (mediated by ALLO as a positive modulator of GABA-A receptors) are of interest. However, the literature is conflicting regarding a potential effect of P4 on HRV via the amygdala. Van Wingen and colleagues combined a single progesterone administration to naturally-cycling individuals with functional magnetic resonance imaging (fMRI) in two double-blind, placebo-controlled, crossover design studies [
55,
56]. Participants received the P4 administration in their follicular phase which yielded plasma concentrations of P4 and ALLO comparable to levels observed during pregnancy in one study [
55] and comparable to luteal phase levels in the other study [
56]. Interestingly, moderate (i.e., luteal phase) ALLO plasma concentrations increased amygdala reactivity (to angry and fearful faces) [
56], while higher (i.e., pregnancy) ALLO plasma concentrations decreased the amygdala reactivity [
55], indicating that ALLO may modulate amygdala activity in a biphasic dose-dependent manner. A (somewhat conflicting) study using positron emission tomography (PET) revealed a positive correlation between amygdala activity and HRV in females [
57]. Since the amygdala is a common neural basis for both P4 and HRV functioning, future studies should continue to examine effects of P4 in the amygdala as a possible mechanism for the P4-HRV correlation.
In sum, much more research is needed to understand the underlying mechanisms of the effects of cycle-related fluctuations of P4 on the regulation of HRV. This is in line with the conclusion of a recent review of research on cardiovascular disease in women, which stated that the role of P4 in the regulation of cardiovascular physiology remains largely unknown and requires further research [
47].
4.2. Clinical Implications
Regarding clinical implications, the results of the present work might be relevant for the field of cardiology: Given that HRV has established itself as a biomarker for cardiac health in research [
9], recent studies suggest its use as a tool for cardiac risk assessment in clinical practice [
8,
58]. If clinicians were to use HRV assessments for clinical reasons, they should keep the within-person fluctuations of HRV in naturally-cycling females in mind. We recommend mid-follicular assessments with low and stable P4 levels (i.e., between day +5 and +10, where the onset of menses is day +1 and there is no day 0).
Given HRV’s association with cognitive and emotional self-regulatory capacity [
1,
2,
3,
4,
5], the results of the present work could also have implications for everyday functioning. However, since no clinical outcomes (e.g., affective and cognitive symptoms) were reported in the present work, it cannot be safely concluded that the observed fluctuations in HRV across the menstrual cycle are associated with fluctuations in clinical symptoms. As will be stated below in the section on future research (see
Section 4.3), this should be investigated in future studies.
4.3. Implications for Future Research
The association of cyclical P4 with vagally-mediated HRV has several important implications for future research: First, future studies on cardiac vagal activity (indicated for example by HRV) as a predictor or outcome should control the cycle phase of the naturally-cycling participants. Cycle phase should either be included in the analyses as a covariate or, preferably, assessments should take place in the same cycle phase for all cycling individuals. A failure to take the menstrual cycle of female participants into account in HRV studies could greatly reduce the significance and validity of findings. The same implication was already found in the meta-analysis showing significant differences in cardiac vagal activity across menstrual cycle phases [
28]. However, the present work can empirically specify this implication based on its result that it is high levels (as opposed to withdrawing levels) of the ovarian hormone P4 (and not E2), which are related to the menstrual fluctuations of cardiac vagal activity (indicated by HRV). We therefore specifically recommend future studies on cardiac vagal activity to schedule assessments in the mid-follicular phase of their naturally-cycling participants since this phase is characterized by low and stable P4 levels. Assessments in this cycle phase are practicable to schedule, as it is possible to count forwards from the more clearly reportable onset of menses and set assessments between days +5 to +10.
A second implication for future research is based on the above-mentioned empirical evidence for the links between HRV and cognitive functioning [
4], emotional functioning [
1,
3,
5] and behavioral outcomes [
2]. These results mainly stem from cross-sectional studies and therefore reveal inter-individual differences in cognitive, affective and behavioral outcomes as a function of trait-level HRV. The results of the present work imply the need for studies investigating whether the intra-individual fluctuations in HRV across the menstrual cycle are paralleled by intra-individual cycle-related variations in cognitive, affective and behavioral outcomes across the cycle. There may also be differences as to which outcome (affective, cognitive or behavioral) is more closely associated to the HRV fluctuations across the menstrual cycle. Given the present work’s result that menstrual HRV fluctuations are related to the ovarian hormone P4 (and not E2), future studies can plan their repeated assessments of these outcomes in different cycle phases accordingly: Assessments in the mid-luteal phase with its high P4 levels and control assessments, for example, in the mid-follicular phase with its stable low P4 levels, should be performed.
The third implication for future research is based on previous empirical work on the emergence of clinically significant emotional, cognitive and physical symptoms in the premenstrual phase and subsequent clearance of symptoms in the follicular phase affecting a minority of cycling individuals (i.e., the new DSM-5 diagnosis PMDD [
17]). Experimental studies demonstrate that PMDD is not characterized by abnormal fluctuations or levels of E2 and P4; rather, it is caused by abnormal responses to normal fluctuations and levels that cause these symptoms [
23]. Individuals affected by PMDD are therefore ascribed
hormone sensitivity [
59]. Previous research traced inter-individual differences in hormone sensitivity (indicated by the emergence of PMDD symptoms) back to, for example, traumatic experiences in the woman’s biography [
60,
61,
62,
63]. In the light of the present work, the question is whether intra-individual fluctuations of HRV across the cycle may be a physiological marker of hormone sensitivity (comparable to traumatic experiences as a biographical marker). Cyclical fluctuations of HRV as a biomarker for hormone sensitivity would take research on female health an important step further.
4.4. Strengths and Limitations
When considering the work as a whole, the primary methodological strength lies in the repeated measurement design investigating intra-individual variations in HRV across the menstrual cycle as opposed to cross-sectional study designs which do not do justice to the inter-individual differences in cycle effects. However, even though analyses in both studies revealed random slopes for the effect of P4 on HRV (i.e., inter-individual differences regarding menstrual cycle-related variations in HRV), there was not enough power in either study to investigate the between-person predictors of these differences in hormone sensitivity further, and this is a central limitation of the studies.
As mentioned before, the two studies included in the present work show various design differences. However, given that the focus of the present work is the association between cyclical ovarian hormones and HRV, the most central variables of interest (i.e., E2, P4, and HRV) were measured with a high degree of consistency between studies: In both studies, saliva samples were used to analyze hormone levels, the same ECG lead pattern was used, participants were seated during their ECG, high frequency RSA was employed as an index for vagally-mediated HRV, statistical analytic methods were identical, and participants were naturally-cycling females of similar age. The two studies are therefore compatible with each other with regard to the question of how the ovarian hormones are associated with HRV. Also the same results (high levels of P4 are associated with low levels of HRV) are yielded in both studies, indicating that the effects of P4 are robust to changes in study design. Given the replication crisis in psychological research, the methodological differences of the two studies could therefore be perceived as an advantage and not a disadvantage [
38].
One central design difference between the studies was the disguise of the menstrual cycle focus in study one but not in study two. On the one hand, this blinding allowed the assessment of the association between ovarian hormones and HRV in study one to be less affected by cycle-related stereotypes and expectations than in study two. On the other hand, the blinding reduced the precision of cycle phase estimations as data points in study one had to be phased in retrospect (via the cycle day count-based method). As a result, 55 of the 160 assessments had to be excluded in the cycle phase analyses of study one since they did not fall within any cycle phase time window and the ovulatory E2 concentrations did not show the typical exponential increase. Most likely, the cycle day count-based method of determining the ovulatory cycle phase included both peak E2 days as well as lower-E2 peri-ovulatory days. However, this study design allowed for testing the primary question of the present work as it allowed for clear examination of the association of hormones (and less cycle phase) and HRV within a given individual across the cycle. Another difference between the two studies refers to the ECG data lengths (5 min in study one vs. 10 min in study two). However, as the present work investigates within-person changes in HRV, and since the ECG data lengths were consistent within each participant, different ECG data lengths in the two studies are defensible. Also, separate statistical models were analyzed for each study resulting in consistent ECG data lengths within each statistical model.
A limitation shared by both studies refers to the time of day of HRV assessments. Given that HRV shows circadian variations [
64,
65] and since the present work focuses on within-person changes in HRV, we have kept the time of the testing consistent within each participant. Nonetheless, it is possible that the time between waking and testing showed within-person differences which, in turn, could distort the investigated effects of cyclical hormones on HRV. Unfortunately, we failed to collect information on the time that has passed between waking up and testing, which is a limitation of both studies.
Another limitation specific to study two refers to the reduced reliability of the ovulation tests used: Since the tests exhibited a high sensitivity to low LH levels, it is possible that they generated false-positive results in participants with a relatively high baseline LH level. In addition, the comparison of the test and control lines was often ambiguous, making the read-out susceptible to errors. For this reason, the cycle phase reclassification process included the backward-count day approach. This approach, however, does less justice to inter-individual differences in cycle lengths than does ovulation testing.
Finally, since the two studies were observational studies investigating concurrent fluctuations in ovarian hormones and HRV across the menstrual cycle, it is not possible to refer to causal cycle effects which is another central limitation of this work.