Blood Plasma Hormone–Level Influence on Vocal Function

Abstract

Purpose

This preliminary study examined the influence of menstrual cycle phase and hormone levels on acoustic measurements of vocal function in reproductive and postmenopausal females. Mean fundamental frequency (f0), speaking fundamental frequency (Sf0), and cepstral peak prominence (CPP) were evaluated. It was hypothesized that Sf0 and CPP would be lower during the luteal and ischemic phases of the menstrual cycle. Group differences with lower values in postmenopausal females and greater variability in the reproductive females were also hypothesized.

Method

A mixed factorial analysis of variance was used to examine differences between reproductive and postmenopausal females and the four phases of the menstrual cycle. Separate analyses of variances were implemented for each of the dependent measures. Twenty-eight female participants (15 reproductive cycling, 13 postmenopausal) completed the study. Participants were recorded reading the Rainbow Passage and sustaining the vowel /a/. Mean vocal f0, Sf0, and CPP were determined from the acoustic samples. Blood assays were used to determine estrogen, progesterone, testosterone, and neuropeptide Y levels at four data collection time points.

Results

Group differences in hormone levels and Sf0 values were established with the postmenopausal group having significantly lower hormone levels and significantly lower Sf0 than the reproductive cycling group across the phases. Analysis of the reproductive group by hormone levels and cycle phase revealed no significant differences for CPP or Sf0 across phases. Higher estrogen was identified in the ovulation phase, and higher progesterone was identified in the luteal phase.

Conclusions

Significant differences in hormone levels and Sf0 were identified between groups. Within the reproductive cycling group, the lack of significant difference in acoustic measures relative to hormone levels indicated that the measures taken may not have been sensitive enough to identify hormonally mediated vocal function changes. The participant selection may have biased the findings in that health conditions and medications that are known to influence voice function were used as exclusion criteria.

The National Institute on Deafness and Other Communication Disorders estimates that 7.5 million individuals experience difficulty with their voice. Women have a significantly higher prevalence of voice disorders than men, at 46.3% and 36.9%, respectively (Roy et al., 2005, 2004). Women have received clinical advice to avoid extensive voice use prior to menstruation for avoidance of vocal injury, indicating a possible correlation between cyclical hormonal fluctuations and voice function (Butler et al., 2001; Ward et al., 2002). A review of the literature linking ovarian hormonal fluctuations and female voice production described the larynx as a hormonal target organ in that hormonal changes with the menstrual cycle affect laryngeal morphology, histology, and function (Abitbol et al., 1999). This relationship between the genital tract and the larynx have been established through comparison of cytological smears of the vocal folds in conjunction with cervical smears as well as through documentation of hormone receptors in the mucosa and epithelium of the vocal folds (Abitbol et al., 1999).

Hormonally driven voice changes are evident across the life span. Compared to a child, the fundamental frequency (f0) of an adult female decreases by approximately a third and the f0 of an adult male decreases by approximately an octave (Abitbol et al., 1999). Voice acoustic changes (Raj et al., 2010), edema, vascular changes, superficial viscosity differences (Abitbol et al., 1999), and limb joint laxity differences (Shultz et al., 2004) associated with the menstrual cycle have been identified. Abitbol et al. (1999) indicated that reproductive females can experience premenstrual vocal syndrome prior to menstruation. Characteristics of premenstrual vocal syndrome include edema of the posterior third of the vocal folds, microvarices, and loss of vibratory amplitude.

The menstrual cycle is characterized by a synergistic interplay among the following hormones, each of which could influence vocal function: estrogen, progesterone, androgens, luteinizing hormone (LH), and neuropeptide Y (NPY). Estrogen increases capillary permeability and has a proliferative and hypertrophic effect on the mucosa, making mucus thinner and more watery (Abitbol et al., 1999). Progesterone decreases or inhibits capillary permeability, thus increasing edema, and has an antiproliferative and dehydrating effect on mucosa, making mucus thicker, opaque, and more viscous (Abitbol et al., 1999). Progesterone is secreted by the corpus luteum, which is an endocrine structure that is formed temporarily from an ovarian follicle. Progesterone markedly declines at menopause with the degeneration of the corpus luteum. Androgens in women can cause the f0 of the voice to decrease and can have a thickening and drying effect on the mucosa (Abitbol et al., 1999). While the influence of estrogen and progesterone on the menstrual cycle is understood, the role of NPY on the menstrual cycle is less clear. Evidence indicates that NPY acts as a vasoconstrictor and regulates immune and inflammatory responses (Dimitrijević et al., 2008), suggesting that it may result in decreased blood flow to the mucosa (Lyon, 2000). NPY has also been observed to have stimulatory effects on reproduction by increasing LH release that triggers ovulation (Coiro et al., 2006).

The reproductive menstrual cycle is divided into four primary phases: follicular, ovulatory, luteal, and ischemic. Day 1 of the menstrual cycle is typically considered the first day of the follicular phase, and the average menstrual cycle lasts from between 28 and 30 days. The follicular and luteal phases can be easily distinguished from one another and are separated by ovulation. During the follicular phase, progesterone levels are lower, and follicle-stimulating hormone will activate the secretion of estrogen that progressively increases until it reaches a peak around Day 13. The ovulatory phase is marked by an increase in LH, which peaks around Day 14 and is followed by ovulation. The luteal phase is marked by the creation of the corpus luteum and secretion of greater progesterone levels than estrogen. The ischemic phase is marked by a rapid drop in estrogen and progesterone levels, which triggers menstruation. Menopause results when estrogen and progesterone levels remain consistently low and ovarian follicular activity ceases. During menopause, the influence of androgens becomes more prominent.

Use of menstrual cycle phases has been a common approach for studying the influence of hormonal fluctuations on voice function. The ischemic phase of the menstrual cycle has been reported to have increased frequency perturbation, jitter, when compared to the follicular phase (Chae et al., 2001). Raj et al. (2010) compared acoustic voice measures of reproductive age and postmenopausal women. Reproductive females were observed to have greater maximum phonation duration, higher f0 values, a greater harmonic-to-noise ratio, and a decreased frequency range (FR) during the ischemic phase of the menstrual cycle. A lower harmonic-to-noise ratio was observed during the ovulatory phase. They specified that ovulation, when estrogen is greatest, was associated with the best vocal quality, and the vocal quality of the reproductive females was better than that of the postmenopausal females. The postmenopausal participants were described to have lower f0 values, greater FRs, and greater jitter and shimmer values when compared to the reproductive females.

The prevalence of voice complaints associated with menopause ranges from 17% (Abitbol et al., 1999) to 77% (Boulet & Oddens, 1996). Literature on the effects of menopause on voice function indicates that menopause can affect laryngeal tissues and result in muscular and mucosal atrophy, fluid retention and swelling of the vocal folds, and increased viscosity of the mucosa (Abitbol et al., 1999; Caruso et al., 2000). Vocal discomfort reported with menopause has been associated with the perception of dryness, throat clearing, a lower f0 (Abitbol et al., 1999; Lindholm et al., 1997), reduced FR (Raj et al., 2010), reduced intensity (Lindholm et al., 1997), vocal fatigue, and increased roughness and hoarseness (Caruso et al., 2000; Schneider et al., 2004). Others have not supported the findings of a lower f0 or reduced FR (Mendes-Laureano et al., 2006; Meurer et al., 2004a, 2004b). These differences may be attributed to task differences during data collection (e.g., acoustic vs. perceptual) or the age of the participant pools. Frequency differences observed from perceptual differences but not acoustic may indicate that participants and clinicians are sensitive to effort changes but the larynx may be adaptable enough to make the adjustments necessary to result in insignificant acoustic changes.

A confounding factor when studying postmenopausal voice function is the influence of laryngeal senescence and aspects of aging in general. Most of the age-related changes to vocal function that negatively affect the viscoelastic properties of the vocal fold cover are described for individuals of advanced age—over the age of 65 years (Lenell et al., 2019). Neuromuscular changes with aging have been summarized as a relative preservation in the number of functioning motor units in limb skeletal muscle up to the age of 60 years, after which there is a drop to less than half of the original number of motor units by the age of 70 years (MacIntosh et al., 2006). The extent to which these changes affect vocal function in women as they age following menopause is not yet well understood. Hormone levels can mediate neuromuscular function, and sex differences attributed to morphological and hormone-mediated factors have been recently reviewed (Lenell et al., 2019). Changes in acoustic measures that are typically attributed to menopause may be due to or confounded by factors other than hormone levels (Awan, 2006). For example, pulmonary compliance change with aging could influence acoustic measures. Given the complex nature of vocal function through the life span, identification of relationships of hormone levels with vocal function measures at differing time points in aging needs to be better understood.

Laryngeal changes associated with hormone fluctuations across the menstrual cycle for reproductive females and after menopause are widely believed to influence voice function. The impact of these changes for professional voice users such as singers, teachers, clergy, and customer support personnel is often discernable to the user but is not well understood. Using a cross-sectional prospective study design, the primary aim of this study was the identification of the relationship between blood plasma–derived hormone levels and voice measures in reproductive and postmenopausal females. Use of blood plasma to objectively quantify hormone levels instead of sole reliance on inferred cycle phases provides a clearer delineation of the degree to which hormone levels may influence changes in vocal frequency (vocal f0). Describing the relationship between sex hormones and voice acoustic measures routinely gathered during clinical voice assessment in both reproductive and postmenopausal women has implications for more accurate interpretation of routine voice acoustic measures taken clinically, more targeted clinical care of occupational voice users, as well as the potential effects of hormone therapy. For this investigation, we hypothesized the following:

Hypothesis 1 (H1): There will be group differences between reproductive and postmenopausal females for the following variables: (H1a) speaking fundamental frequency (Sf0), (H1b) mean vocal f0, and (H1c) CPP amplitudes. Specifically, postmenopausal females will have lower amplitudes, lower Sf0, and lower mean f0 than reproductive females.

The next set of hypotheses pertain to the hormonal changes that occur during the menstrual cycle phases for reproductive females only as they experience hormonal changes based on cycle phase.

Hypothesis 2 (H2): Reproductive females will have the highest Sf0 values (derived from the Rainbow Passage), highest mean f0 (H2a), and larger cepstral peak prominence (CPP) amplitudes (H2c) during the ovulatory phase when estrogen levels are highest and lowest values during the luteal or ischemic phases when progesterone is highest.

Exploratory Hypotheses: Hormones

Hypothesis 3 (H3): Based on previous work, we anticipated that estrogen and progesterone levels may predict the magnitude change in Sf0, mean vocal f0, and CPP amplitudes, especially for reproductive females. We conducted a series of exploratory regressions to determine the effects of testosterone and neuropeptide on these variables. Additionally, we conducted a hierarchal regression using all four hormones to predict performance on the three dependent variables (Sf0, f0, and CPP) to determine which hormones uniquely accounted for the variance in the dependent measures. Greater NPY values would be associated with lower CPP amplitude and lower Sf0 values. NPY is anticipated to be highest during the luteal and ischemic phases of the menstrual cycle. Therefore, lower CPP amplitude and Sf0 values would be anticipated during the luteal and ischemic phases.

Method

Study Population

Following receipt of approval by the institutional review board, women were recruited locally and informed consent was obtained from each participant prior to screening for eligibility. Participants were required to complete a health questionnaire and were excluded if they had any of the following health conditions: diabetes, allergic rhinitis, reflux, respiratory problems, active smoker at time of investigation, neurological problems, hormonal imbalance, pregnancy, had been breastfeeding over the preceding 6 months, or were taking any drying medications. Normally cycling females (24–34 days) were included if they were not using any hormonal contraceptives (i.e., oral, intrauterine device, shot) for 3 consecutive months prior to data collection. Postmenopausal females were included if they were not taking hormone replacement therapy. All participants were prescreened to rule out laryngeal pathology or voice disorder. None of the postmenopausal participants recruited exhibited any evidence of vocal fold bowing or laryngeal senescence.

Participants

The participants in this cross-sectional, repeated-measures, prospective study included 15 normally cycling young women, between the ages of 20 and 35 years (M = 23) who were in the reproductive stage of life and 13 postmenopausal women between the ages of 53 and 65 years (M = 59) who had been in menopause for more than 2 years. To establish cycle length for each of the cycling participants, they provided the date of onset for their last two menstrual cycles. If a participant was unable to provide the last two cycle onset dates, she was asked to track her next two menstrual cycles to establish whether she had a regular cycle that was predictable within 2–3 days. Once a regularly occurring cycle length for the participant was established, the window of each menstrual cycle phase was determined using an ovulation calculator that took into account cycle length and cycle start date. The participants were given a 2- to 6-day window that encompassed each phase of the menstrual cycle targeted in this study. To capture the ovulatory cycle, that is approximately 48 hr in length, ovulation test strips (Wondfo USA Co., Ltd.) designed to detect an LH surge were given to each participant before the ovulatory phase so they could determine the point of their LH surge at home. When the participant tested positive for LH (> 25 miu/ml), they were scheduled for data collection within 24 hr after LH detection. For example, a person with a 28-day cycle would be given the following time frames: follicular (F): Days 1–11, ovulatory (O): Days 13–15, luteal (L): Days 18–23, and ischemic (I): Days 26–28. For each of the menstrual cycle phases described above, participants completed all experimental tasks. Postmenopausal women were scheduled to come in once per week for 8 weeks. Each participant was scheduled at approximately the same time of day for data collection to avoid any data variability secondary to circadian rhythm. That is, postmenopausal females were scheduled for 8 consecutive weeks on the same day and time to mirror a 28-day cycle phase for a reproductive female.

Blood Collection and Assays

Plasma samples were collected from each participant by drawing blood using a standard venipuncture in the antecubital vein. Blood was centrifuged at 1,000g and 4 °C for 10 min to separate plasma, which was drawn off and stored at −80°C until analyses. Plasma was analyzed for estradiol (estradiol ELISA, Catalog Number 582251; Cayman Chemical), testosterone (testosterone ELISA, Catalog Number 582701; Cayman Chemical) and progesterone (progesterone ELISA kit, Catalog Number 2077-18; Diagnostic Automation/Cortez Diagnostics, Inc.), and NPY (Peptide Enzyme Immunoassay Protocol V, Catalog Number S-1145, Peninsula Laboratories International) using the standard protocols.

Acoustic Data Collection

The participants were recorded sustaining the vowel /a/ for a minimum of 8 s and reading the entire Rainbow Passage at their own pace using the Computerized Speech Laboratory (CSL; Pentax Medical). The recordings were collected as .wav files, with a sampling rate of 44 kHz with the headset microphone positioned at a constant distance of approximately 4 cm from the participant's mouth. The participant was instructed to speak at a comfortable pitch and loudness for both of the tasks. The CSL was used to analyze the data with a sampling rate of 44.1 kHz and 24-bit A/D conversion. The mean f0 was derived from the middle 6 s of the sustained vowel using the CSL main program. The Sf0 and CPP values were derived from the entire production of the Rainbow Passage using the Analysis of Dysphonia in Speech and Voice program. The Analysis of Dysphonia in Speech and Voice settings were as follows: CPP threshold = 0 (dB); cepstral extraction range: 60–300 Hz; spectral window size = 1,024 points; maximum frequency for regression line calculation = 10,000 Hz; frame overlap = 75%; and cepstral timing average = 7.

Statistical analyses: To test hypotheses H1a–H1c, a mixed factorial analysis of variance (ANOVA) was used to examine differences between reproductive and postmenopausal females (i.e., as the between-participants variable) and the four phases of the menstrual cycle (follicular, ovulation, luteal, and ischemic) as the within-participant variable. A separate ANOVA was implemented for each of the three dependent measures: Sf0, CPP, and vocal f0. Additionally, to test H2a–H2c, planned contrasts and one-way ANOVA were conducted per dependent measure for reproductive females only.

Results

SPSS software (Version 24) was used to analyze the data for the study. Prior to conducting analyses, the data were examined for potential outliers. Specifically, the data that were identified as potential outliers in the descriptive outlier and box plot analyses were rechecked for both groups. Some of the participants did not have data for all variables across the different time points. Because of the small sample size, all data were retained, and any missing variables were excluded from analyses; hence, variation in sample size for different tests is reported.

Hormone Levels Check

Between-Groups Comparison

Because we expected there to be differences in hormonal levels between these two groups due to how hormones are affected across the life cycle, we conducted independent t tests to evaluate hormone levels between these groups for each phase of the menstrual cycle (see Table 1 for descriptive means and t-test comparisons between the two groups). These comparisons confirmed group differences in hormone profiles between the groups and, in particular, for the estrogen and progesterone hormone profiles (see Table 1). The estrogen and progesterone values for the postmenopausal females were significantly less than the reproductive females and remained consistently low throughout the 4 points of data collection as expected.

Table 1.

Descriptive statistics and t tests between reproductive and postmenopausal participants' hormone levels across four time points.

Hormone	Phase	Reproductive			Postmenopausal			t	df	d
		N	M	SD	N	M	SD
Estrogen (pg/ml)	F	15	67.170	60.034	12	3.47	2.668	4.104***	14.069	1.499
	O	15	167.497	106.127	12	3.714	3.099	5.974***	14.030	2.182
	L	15	92.148	73.026	12	4.711	4.712	4.625***	14.146	1.690
	I	15	79.911	68.905	11	5.005	5.288	4.193***	14.242	1.533
Progesterone (ng/ml)	F	15	0.664	0.500	13	0.551	0.364	0.671	25.620	0.258
	O	15	0.983	0.702	13	0.499	0.310	2.411*	19.845	0.892
	L	15	5.180	2.463	13	0.295	0.215	7.649***	14.246	2.794
	I	15	2.963	1.895	12	0.304	0.226	5.387***	14.512	1.970
Testosterone (ng/dl)	F	15	437.745	171.99	13	128.792	71.821	6.348***	19.289	2.344
	O	15	508.947	264.692	13	178.671	188.828	3.836***	25.151	1.437
	L	13	532.175	481.210	13	154.807	106.898	2.760*	13.181	1.250
	I	15	437.904	348.457	13	112.190	92.466	3.482**	16.242	1.278
Neuropeptide (ng/ml)	F	14	0.360	0.237	13	0.505	0.125	−2.007	20.002	0.765
	O	14	0.346	0.254	13	0.505	0.140	−2.038	20.523	0.775
	L	14	0.364	0.276	13	0.563	0.166	−2.288*	21.580	0.874
	I	15	0.382	0.261	13	0.651	0.179	−3.208*	24.8	1.202

Note. Because the equal variance assumption did not hold for most t tests, the degrees of freedom (df) reported are adjusted to reflect conducting the t test without assuming equal variances for each variable between the groups. Cohen's d is represented in the last column and indicates the strength of the effect size between groups for each comparison. F = follicular; O = ovulatory; L = luteal; I = ischemic.

^* p < .05.

^** p < .01.

^*** p < .001.

Reproductive Females

Because reproductive females are expected to have hormone fluctuation differences across the cycle, we conducted a one-way repeated-measures ANOVA using only reproductive females per hormone. As expected and consistent with reproductive females who are cycling regularly, estrogen fluctuated across phases, F(3, 42) = 6.176, p < .01, η_p ² = .306. This difference was accounted for by the significant increase in estrogen at the ovulatory phase compared to follicular, t(14) = −2.875, p = .012, d = 1.163; luteal, t(14) = −2.983, p = .010, d = 0.83; and ischemic, t(14) = −3.467, p = .004, d = 0.98, phases. Furthermore, progesterone fluctuated across phases, F(3, 42) = 32.676, p > .001, η_p ² = .700. This difference was accounted for by significantly increased progesterone during the luteal phase compared to follicular, t(14) = −7.828, p = .000, d = 2.542; ovulatory, t(14) = −8.249, p = .000, d = 2.318; and ischemic, t(14) = −2.939, p = .011, d = 1.009, phases. The paired comparisons for progesterone levels also achieved significance between follicular and ovulatory, t(14) = −3.390, p = .004, d = 0.525; follicular and ischemic, t(14) = −4.941, p = .000, d = 1.660; and ovulatory and ischemic, t(14) = −4.298, p = .001, d = 1.386, for reproductive females. The group mean hormone values indicated that data were collected accurately across the four phases of the menstrual cycle with the highest estrogen values identified at the ovulatory phase and highest progesterone values identified during the luteal phase for the reproductive group. Testosterone and neuropeptide did not produce significant within-hormone differences between phases for reproductive females in this study sample.

Sf0

Table 2 presents the average Sf0 as well as vocal f0 and CPP for reproductive and postmenopausal groups at each phase.

Table 2.

Average (standard error) speaking frequency, mean vocal frequency, and cepstral peak prominence (CPP) amplitude between reproductive and postmenopausal participants.

Variable	Reproductive				Postmenopausal
	Phase	N	M	SDE	N	M	SDE	d
Speaking fundamental frequency (Sf0)	F	15	200.690	4.973	13	181.686	5.313	0.989
	O	14	202.801	6.217	13	178.658	4.771	1.180
	L	15	201.090	5.773	13	180.994	5.585	0.945
	I	14	199.838	5.731	13	181.081	4.970	0.949
Mean vocal frequency (f0)	F	15	215.767	3.555	13	195.947	5.264	1.195
	O	15	220.125	5.928	13	197.957	4.321	1.129
	L	15	216.515	3.970	13	195.926	4.410	1.316
	I	15	214.239	3.543	13	195.618	4.480	1.242
CPP amplitude (dB)	F	15	5.885	0.169	13	6.096	0.139	0.281
	O	14	5.799	0.179	13	6.182	0.160	0.462
	L	15	6.006	0.177	13	6.356	0.151	0.435
	I	14	5.920	0.168	13	6.157	0.146	0.308

Note. Cohen's d is represented in the last column and indicates the strength of the effect size between groups for each comparison. F = follicular; O = ovulatory; L = luteal; I = ischemic.

Between-Groups Comparison

The main effect of group emerged where reproductive females generated significantly higher f0 values than the postmenopausal females, F(1, 25) = 7.267, p < .05, η_p ² = .225. Four paired-samples t tests were conducted to follow-up on observed differences between the two groups within each phase. Family-wise error rate was controlled across the test by using the Bonferroni approach. Differences in Sf0 between the two groups were observed across the four phases of the menstrual: follicular, t(26) = 2.610, p = .015; ovulation, t(25) = 3.047, p = .005; luteal, t(26) = 2.483, p = .020; and ischemic, t(25) = 2.457, p = .021. The postmenopausal group had significantly lower Sf0 values than the reproductive group across phases. The main effect of phase and the interaction between group and phase were not statistically significant, F(3, 75) = 0.176, p = .913, η_p ² = .007 and F(3, 75) = 1.347, p = .266, η_p ² = .051. We did not expect postmenopausal females to demonstrate differences in f0 for each phase as they are no longer experiencing cycle phases, which accounts for lack of significant differences for the main effect of phase as well as the interaction between phase and group. As such, H1a was supported that general group differences in f0 were present.

Reproductive Females

To test H2a that reproductive females would have the higher Sf0 during the ovulatory phase and lowest amplitudes during the luteal or ischemic phases, we conducted a one-way repeated measures ANOVA. Sf0 did not differ between phases, F(3, 39) = 0.759, p ≥ .05, η_p ² = .055. As can be noted from Table 2, Sf0 remained more or less constant across phases for reproductive females.

Mean Vocal f0

Between-Groups Comparison

The main effect of group was significant, F(1, 26) = 14.230, p < .001, η_p ² = .354, whereby the cycling women had a higher mean vocal f0 than the postmenopausal participants. The main effect of phase and the interaction between group and phase were not statistically significant, F(3, 78) = 0.354, p = .787, η_p ² = .013 and F (3, 78) = 0.437, p = .727, η_p ² = .017. We did not expect postmenopausal females to demonstrate differences in vocal f0 for each phase as they are no longer experiencing cycle phases, which accounts for lack of significant differences for the main effect of phase as well as the interaction between phase and group. As such, H1b was supported that general group differences emerged in mean vocal f0.

Reproductive Females

To test H2b that reproductive females would have the higher vocal f0 during the ovulatory phase and lowest amplitudes during the luteal or ischemic phases, we conducted a one-way repeated-measures ANOVA. Vocal f0 did not show differences across phases, F(3, 42) = .768, p ≥ .05, η_p ² = .052. As can be noted from Table 2, vocal f0 remained more or less constant across phases.

CPP

Between-Groups Comparison

The main effect of group comparing reproductive and postmenopausal females was not statistically significant, F(1, 25) = 1.479, p = .225, η_p ² = .056. The main effect of phase and the interaction between group and phase were also not statistically significant, F(3, 75) = 1.213, p = .311, η_p ² = .046 and F (3, 75) = 0.454, p = .715, η_p ² = .018. We did not expect postmenopausal females to demonstrate differences in CPP for each phase as they are no longer experiencing cycle phases, which accounts for lack of significant differences for the main effect of phase as well as the interaction between phase and group. However, we did not observe the group differences as expected, and as such, H1c was not supported.

Reproductive Females

To test H2c that reproductive females would have the highest CPP amplitudes during the ovulatory phase and lowest amplitudes during the luteal or ischemic phases, we conducted a one-way repeated-measures ANOVA. CPP amplitudes did not overall show significant differences across phases, F(1, 13) = 2.025, p = .178, η_p ² = .135. As can be noted from Table 2, we did not observe large differences in CPP amplitudes between phases for the reproductive females.

Hormonal and Phasic Effects in Sf0, Vocal f0, and CPP Amplitude

Analyses

Simple linear regressions were used to determine whether each individual hormone during each phase predicted each dependent variable (Sf0, f0, and CPP) during the same phase. For example, do estrogen levels during follicular phase predict Sf0 during follicular phase? Additionally, we conducted exploratory stepwise multiple regressions where all four hormones were added to the model during a specific phase per dependent variable to examine whether the combination of hormones explained the variance in the dependent measures. The former analyses investigated each hormone's effect individually, whereas the latter exploratory analyses allow us to examine whether multiple hormones have an effect on predicting the performance levels per dependent variable.

Sf0

Table 3 presents the individual simple linear regressions for each hormone and phase including the unstandardized coefficient (B), standard error of coefficient [SE (B)], standardized beta coefficient (β), and tests of the coefficient (t and significant p value) as well as the variation in speaking frequency explained by hormone (r ²).

Table 3.

Simple linear regressions examining each hormone's effect on speaking fundamental frequency across four time points.

Hormone	Phase	B	SE(B)	β	t	p	r 2
Estrogen	Follicular	0.150	0.070	.394	2.146	.042*	.156
	Ovulatory	0.117	0.041	.501	2.833	.009**	.251
	Luteal	0.182	0.056	.548	3.279	.003**	.301
	Ischemic	0.116	0.066	.344	1.759	.092 +	.119
Progesterone	Follicular	−9.454	9.297	−.196	−1.017	.319	.038
	Ovulatory	−4.462	7.852	−.113	−0.568	.575	.013
	Luteal	1.232	1.478	.161	0.834	.412	.026
	Ischemic	1.744	2.197	.160	0.794	.435	.026
Testosterone	Follicular	0.021	0.020	.207	1.080	.290	.043
	Ovulatory	0.013	0.017	.150	0.757	.456	.022
	Luteal	0.022	0.011	.375	1.983	.059 +	.141
	Ischemic	0.000	0.014	−.005	−0.025	.980	.000
Neuropeptide	Follicular	−49.095	18.309	−.473	−2.681	.013*	.223
	Ovulatory	−38.955	20.754	−.358	−1.877	.073 +	.128
	Luteal	−42.835	17.216	−.446	−2.488	.020*	.446
	Ischemic	−29.777	15.329	−.362	−1.943	.063 +	.131

⁺ p < .10.

^* p < .05.

^** p < .01.

Estrogen. The hormone estrogen significantly predicted Sf0 during the first three phases: follicular, ovulatory, and luteal phases. However, estrogen did not significantly predict Sf0 during the ischemic phase. Based on the regression slope coefficients (B), we can interpret that, for every estrogen picogram per milliliter, Sf0 increased by 0.150 Hz in the follicular phase, increased by 0.117 Hz in the ovulatory phase, and increased by 0.182 Hz in the luteal phase.

Progesterone and testosterone. Neither progesterone nor testosterone significantly predicted Sf0 during any of the phases. Thus, neither hormone contributed to significantly increasing Sf0.

Neuropeptide. The hormone neuropeptide did significantly predict Sf0 during the follicular phase and the luteal phase; however, neuropeptide did not predict Sf0 in the ovulatory and ischemic phases. Interestingly, for every neuropeptide nanogram per milliliter, Sf0 decreased by 49.095 Hz in the follicular phase and decreased by 42.835 Hz in the luteal phase.

Exploratory analyses with all hormones. To explore the effect of all hormones as multiple predictors of Sf0, we conducted a multiple regression where all four hormones per phase were entered simultaneously into the model. Table 4 presents the multiple regression analyses for each phase for mean vocal f0. For the follicular phase, adding in all the hormones strengthened the model, R ² = .426, F(4, 21) = 3.891, p < .05, such that 42.6% of variance of vocal f0 was accounted for in the model and only progesterone produced a statistically significant coefficient. For the ovulatory phase, adding in all the hormones strengthened the model, R ² = .385, F(4, 20) = 3.134, p < .05, and 38.5% of variance of vocal f0 was accounted for in the model with estrogen and progesterone resulted in statistically significant coefficients. The overall regression model was not significant for either the luteal phase, R ² = .281, F(4, 19) = 1.853, p > .05, or the ischemic phase, R ² = .215, F(4, 20) = 1.368, p > .05.

Table 4.

Multiple regression using all hormones to predict fundamental speaking frequency at each phase.

Phase	Variables	B	SE(B)	β	t	p	R 2
Follicular	Intercept	204.764	10.438
	Estrogen	0.032	0.072	.087	0.443	.662
	Progesterone	−6.669	7.918	−.143	−0.842	.409
	Testosterone	0.037	0.019	.368	1.960	.063 +
	Neuropeptide	−49.546	18.105	−.495	−2.737	.012*
							.426
Ovulatory	Intercept	200.591	11.471
	Estrogen	0.122	0.050	.523	2.418	.025*
	Progesterone	−9.246	7.538	−.231	−1.227	.234
	Testosterone	−0.002	0.018	−.019	−0.088	.931
	Neuropeptide	−30.603	19.538	−.282	−1.568	.133
							.385
Luteal	Intercept	193.158	16.281
	Estrogen	0.094	0.243	.213	0.388	.703
	Progesterone	−0.009	2.423	−.011	−0.004	.997
	Testosterone	0.008	0.022	.142	0.382	.707
	Neuropeptide	−23.328	24.871	−.246	−0.898	.381
							.281
Ischemic	Intercept	201.988	13.211
	Estrogen	0.098	0.087	.292	1.126	.273
	Progesterone	−0.640	3.369	−.058	−0.190	.851
	Testosterone	−0.009	0.020	−.124	−0.436	.668
	Neuropeptide	−25.779	19.676	−.311	−1.310	.205
							.215

⁺ p < .10.

^* p < .05.

Mean Vocal f0

Table 5 presents the individual simple linear regressions for each hormone and phase for reproducing females.

Table 5.

Simple linear regressions examining each hormone's effect on mean vocal frequency across four time points.

Hormone	Phase	B	SE(B)	β	t	p	r 2
Estrogen	Follicular	0.123	0.066	.348	1.858	.075 +	.121
	Ovulatory	0.094	0.035	.473	2.682	.013*	.223
	Luteal	0.138	0.046	.511	2.974	.006**	.261
	Ischemic	0.081	0.054	.291	1.488	.150	.084
Progesterone	Follicular	−14.747	7.981	−.341	−1.848	.076 +	.116
	Ovulatory	−10.236	7.102	−.272	−1.441	.161	.074
	Luteal	1.982	1.127	.326	1.759	.090 +	.106
	Ischemic	2.033	1.755	.226	1.159	.258	.051
Testosterone	Follicular	0.025	0.017	.272	1.439	.162	.074
	Ovulatory	0.002	0.016	.030	0.151	.881	.001
	Luteal	0.009	0.009	.201	1.005	.325	.040
	Ischemic	0.001	0.011	.018	0.91	.928	.000
Neuropeptide	Follicular	−30.351	17.791	−.323	−1.706	.100 +	.104
	Ovulatory	−35.686	19.724	−.340	−1.809	.082 +	.116
	Luteal	−19.340	14.775	−.253	−1.309	.202	.064
	Ischemic	−16.720	12.650	−.251	−1.322	.198	.063

⁺ p < .10.

^* p < .05.

^** p < .01.

Estrogen. The hormone estrogen significantly predicted vocal f0 during the ovulatory and luteal phases and not during the follicular and ischemic phases. Thus, per picogram per milliliter of estrogen, predicted increases in vocal f0 by 0.094 Hz in the ovulatory phase and an increase by 0.138 Hz for luteal phase were identified.

Progesterone, testosterone, and neuropeptide. None of these hormones significantly predicted vocal f0 during any of the phases; thus, these hormones did not have a significant effect on affecting mean vocal f0.

Exploratory analyses with all hormones. To explore the effect of all hormones as multiple predictors of mean vocal f0, we conducted a multiple regression where all four hormones per phase were entered simultaneously into the model. Table 6 presents the multiple regression analyses for each phase for mean vocal f0. For the follicular phase, adding in all the hormones strengthened the model such that 36.8% of variance of vocal f0 was accounted for in the model and only progesterone produced a statistically significant coefficient. For the ovulatory phase, adding in all the hormones strengthened the model and 49.3% of variance of vocal f0 was accounted for in the model, with estrogen and progesterone resulting in statistically significant coefficients. The overall regression model was not significant for either the luteal phase or the ischemic phase.

Table 6.

Multiple regression using all hormones to predict mean vocal frequency at each phase.

Phase	Variables	B	SE(B)	β	t	p	R 2
Follicular	Intercept	216.153	10.257
	Estrogen	0.038	0.071	.109	0.533	.599
	Progesterone	−16.524	7.780	−.379	−2.214	.046*
	Testosterone	0.032	0.018	.345	1.753	.094 +
	Neuropeptide	−26.827	17.791	−.286	−1.508	.146
							.368
Ovulatory	Intercept	225.043	9.844
	Estrogen	0.128	0.040	.644	3.245	.004**
	Progesterone	−16.563	6.473	−.434	−2.559	.018*
	Testosterone	−0.012	0.016	−.143	−0.730	.473
	Neuropeptide	−25.082	16.792	−.240	−1.494	.150
							.493
Luteal	Intercept	195.292	13.594
	Estrogen	0.215	0.203	.596	1.058	.303
	Progesterone	0.543	2.024	.080	0.268	.791
	Testosterone	−0.009	0.018	−.195	−0.511	.615
	Neuropeptide	5.869	20.767	.080	0.283	.781
							.240
Ischemic	Intercept	207.967	10.817
	Estrogen	0.057	0.073	.204	0.777	.446
	Progesterone	1.619	2.826	.178	0.0573	.573
	Testosterone	−0.011	0.016	−.186	−0.0650	.523
	Neuropeptide	−11.233	16.242	−.165	−0.691	.497
							.140

⁺ p < .10.

^* p < .05.

^** p < .01.

CPP

We did not find any effect of hormone (i.e., estrogen, progesterone, testosterone, and neuropeptide) individually in the simple regressions to affect CPP (p > .05, r ² ranging from .000 to .144 for all models) nor in the multiple regressions (p > .05, r ² ranging from .109 to .167 for all models) when all hormones were combined per phases. Thus, we did not find evidence that CPP amplitude was affected by the hormones examined in this study.

Discussion

The combination of serum assay identification of hormone levels with routine clinical acoustic measures of vocal function represents a novel multidisciplinary approach for understanding the influence of hormone fluctuations on vocal function in reproductive and postmenopausal women. Given that voice disorders are more prevalent in women than men and hormone fluctuations are a key physiological difference between men and women, this research effort sought to identify the extent to which plasma hormone levels may influence voice function. Further plasma assay distinctions between reproductive and postmenopausal women were also identified as a means to understand acoustic differences between these two groups. Given that the larynx is a hormone target organ and voice changes have been described empirically and anecdotally to differ by phase of the menstrual cycle (Abitbol et al., 1999), it was anticipated that acoustic measures would differ between cycle phases and correspond to changes in hormone levels.

The differences between groups for Sf0 were anticipated and are supported by the literature, with postmenopausal females having average lower Sf0 values than reproductive cycling women (Berg et al., 2017). The differences identified are also supported by literature describing a significant difference between young women (aged 20–30 years) and middle-age women (aged 40–60 years), the latter group including both peri- and postmenopausal women (Eichhorn et al., 2018). While aging may play a part in the acoustic differences identified between the two groups, the degree to which hormones versus aging mechanisms contribute to this difference have not been well delineated in the basic or applied literature. To date, the literature available describes greater influence of laryngeal senescence occurring over the age of 65 years (Lenell et al., 2019), and the postmenopausal group consented for this research all fell within the middle age range except for one 65-year-old participant.

The study hypotheses linking cycle phase and hormone levels to acoustic measures were evidence supported for Sf0 and vocal f0 measures. No effect was observed for CPP. More specifically, increases in estrogen resulted in incremental increases in Sf0 for the follicular, ovulatory, and luteal phases and increases in vocal f0 for ovulatory and luteal phases. Null findings for CPP may indicate that, in general, laryngeal function is highly adaptable in healthy young participants with no history of voice disorder. Because the reproductive participants were several years out of puberty, they likely had many years of experience stabilizing their vocal function despite the identified changes in serum hormone levels between cycle phases. Given the claims that are made for CPP as a sensitive parameter for subtle vocal function changes, the lack of evidence was surprising. It was hoped that CPP would be more sensitive than the other acoustic parameters used to identify cycle and hormone differences.

To the best of the authors' knowledge, use of NPY as a blood plasma parameter was novel in this study of voice function. Surprisingly, NPY appeared to be the biggest predictor of Sf0 change in this preliminary investigation with increases in NPY resulting in decreases in Sf0. Future studies should incorporate measures of NPY in addition to the typical hormones that are routinely considered in voice research to better understand the role it may play in influencing vocal function.

In the reproductive group, the lack of acoustic differences between cycles for the acoustic measures may be due to the participant pool being highly constrained via the exclusion criteria used. It may be that the addition of one or more of the health criteria used for exclusion would result in greater influences on vocal function with changes on blood plasma hormone levels. The role of epithelial function in the maintenance of airway surface liquid, for example, may be impaired to a functional and measurable level when lowered estrogen and progesterone levels are combined with drying medication or reflux, both of which have been shown to impair epithelial function (Erickson & Sivasankar, 2010; Erickson-Levendoski & Sivasankar, 2011). It may be that women can adapt to one of these perturbations physiologically, but voice function declines with the addition of other health conditions. These relationships require further inquiry.

As evidenced by the large standard deviations reported in Table 1, it is also important to acknowledge the within-participant and between-participants variability in hormone levels empirically determined in this study. Large standard deviations observed in aggregate data indicated interparticipant variability in hormone levels within each given phase. That variability makes it difficult to establish normative hormone levels across individuals. A focus on hormone levels as a predictor for acoustic voice variability unrelated to phase of the menstrual cycle allows for greater predictability of more accurate voice outcomes. For example, some females have greater levels of testosterone than other females, which can result in the diagnosis of polycystic ovarian syndrome, the effects of which on voice function are not currently known. These standard deviations indicate that cycle phase alone may not be predictive of hormonally mediated voice change. Hormone-level analyses may be necessary to ascertain discrete differences in voice function.

These findings have implications for predicting how different forms of hormone therapy can impact the voice, particularly for occupational voice users who rely on reliable, durable, and excellent voice quality to maintain employment. The extent to which the voice characteristics of quality and durability may be managed with hormone therapy are yet to be described. Because of the intra- and interparticipant variability in plasma hormone levels identified in the reproductive women, it would not be in the best interest of females describing vocal impairment to treat them all in the same fashion. At the present time, it is not routine practice for many speech-language pathologists to inquire about the menstrual cycle or hormones in general during routine clinical voice assessment (Plexico & Sandage, 2018). Identification of the extent to which hormone levels incrementally influence voice function in health and in company with many health conditions and medications will further our ability to tailor voice care to the individual needs of the client. Furthermore, identification of these relationships to hormone levels will assist in clinical hypothesis generation regarding likely tipping points from vocal health to disorder. For example, while women typically experience menopause and the subsequent change in hormones in their late 40s or early 50s, some women go through menopause much earlier. Identification of hormone-level changes at an earlier age that mirror those that accompany menopause may provide a clinical rationale for the development of a presenting voice disorder. Consideration of hormone-level changes may also account for the development of a voice disorder in a middle-age woman occupational voice user who had no prior history of voice disorder.

Discovery of laryngeal mechanical differences through the menstrual cycle may help identify time points of vocal vulnerability that may predispose women to develop voice disorders, leading to better habilitation of occupational voice users (e.g., teachers) and more efficient/knowledgeable clinical voice rehabilitation. Identification of hormonal influences on voice function that occur with menopause could further delineate the risks or benefits associated with hormonal replacement therapy on voice function in future research.

A strength of this study was the identification of a preliminary evidence to understand the relationship between hormone levels and acoustic measures of voice function. Identification of how changes in estrogen and NPY can influence Sf0 provides a piece of the larger story of how voice function may change over the course of the life span, a complex interaction between hormone-level changes and aging changes unrelated to hormones. The preliminary outcome described above may better account for individual differences in voice that can be attributed to differences in hormone level versus menstrual cycle phases. Further evidence to validate the proposed predictive model is needed.

It is acknowledged that the present research effort was limited by the number of participants; however, in a complex, repeated-measures design that requires a total of nine laboratory visits, retention of high numbers of participants presented as a realistic barrier.

Conclusion

This study aspired to identify relationships between hormone levels and vocal function that could be useful to determine vocally vulnerable time points for occupational voice users within the menstrual cycle. To that end, the differences proposed between reproductive, cycling women and postmenopausal women were evidence supported. Differences in estrogen and NPY relative to cycle phase were also identified. Future efforts to refine this basic research should consider the inclusion of participants who are on birth control and have one or more health conditions that are known to negatively influence voice function. Analyses of perturbation measures could also yield a more refined understanding of laryngeal function that is hormonally mediated. Additional investigations that target male participant hormone levels would further our knowledge regarding sex differences. Inclusion of women who take birth control pills or who are identified as having high testosterone would also further this line if there is further inquiry.

Funding Statement

This study was funded by National Institute on Deafness and Other Communication Disorders Grant NIH-1R03DC013664-01A1, awarded to Laura W. Plexico.