Characterization of Salivary Progesterone in Female Smokers

Abstract

Introduction

Fluctuations in ovarian hormones have been associated with changes in cigarette smoking behavior, which can be measured through both serum or less invasive salivary procedures. The primary aim of this exploratory study is to characterize the progesterone profiles of salivary progesterone measurements and to compare that with the profiles estimated from a previously measured serum sample.

Aims and Methods

Nontreatment-seeking, cigarette smoking women (n = 82; ages 18–45 years) provided daily salivary hormone samples every morning for 14 consecutive days. Time-dependent random effects functions were used to approximate daily salivary progesterone (ng/mL) levels over the course of a standardized menstrual cycle. Serum measures of progesterone from a previous study of female cigarette smokers were examined for consistency with established profiles and compared with the salivary profile using the same methodology.

Results

The salivary model fit exhibits relative stability during the follicular phase and a clear unimodal peak during the luteal phase. Parameter estimates from the non-linear function show correspondence to serum data. Although the profiles estimated from salivary and serum data agree in functional form, we observed larger between-subject heterogeneity both in the follicular level and the peak luteal level in salivary measures.

Conclusions

The pattern of salivary and serum progesterone measured across the menstrual cycle is similar in form, which is noteworthy given that the saliva and serum samples were drawn from independent sample of female smokers. Inter- and intra-individual variation in salivary measures may be greater than in serum measures.

Implications

Measuring progesterone level variation across the menstrual cycle via saliva samples has several benefits relative to serum sampling methods in that they are easily obtained, noninvasive, and low-cost. Inter- and intra-individual variation in measurements may be greater than those in serum measurements. However, the functional form of the salivary progesterone profile is isomorphic to serum progesterone.

Introduction

Despite advances in prevention and treatment, cigarette smoking remains a leading cause of preventable death in the United States. While men and women are equally likely to attempt to quit smoking cigarettes, women are less likely to achieve abstinence both during and following cessation treatment^1–4 (although this observation may not hold for varenicline⁵). Ovarian hormone levels and fluctuations during the menstrual cycle may play a role in successful abstinence attempts among female smokers. Our prior work shows that (1) within-subject daily salivary progesterone levels are associated with the number of cigarettes smoked per day (CPD) in non-treated female smokers during ad libitum smoking and (2) increased levels of serum progesterone are associated with greater abstinence rates.^6,7 Specifically, early luteal increases in progesterone levels may significantly attenuate reported CPD and promote abstinence; however, these effects diminish as progesterone plateaus and begins to decline in the middle and later portions of the luteal phase.^6,7

Given the potential importance of ovarian hormones in the smoking behavior of women, it is beneficial to assess varied hormone measurement strategies. One newer strategy that is being adopted by investigators derives hormone levels from saliva rather than serum. Saliva measures have several advantages including lower cost and invasiveness, making them more amenable to frequent collection in longitudinal studies.⁸ Saliva samples can be collected by study participants at home without the need for trained personnel, reducing patient burden, and study cost. While this less burdensome strategy is more research-friendly, there have been few studies that have compared salivary hormone profiles with those derived from serum samples. Therefore, the primary aim of this exploratory analysis is to characterize the average salivary and serum progesterone profiles during a normal menstrual cycle in cigarette smoking women.

Methods

Study Population

Saliva Data Participants

Nontreatment-seeking, cigarette smoking women (n = 115; ages 18–45 years) were recruited from the Charleston, SC area from March 2013 through May 2017 to participate in a study that examined the association between salivary ovarian hormone levels and daily cigarette craving and smoking behavior. Study eligibility was based on the following inclusion criteria: (1) smoke an average of 5 cigarettes/day for at least 6 months, (2) submit a breath carbon monoxide sample at screening of at least 5 parts per million, (3) be post menarche/pre-menopausal with regular menstrual cycles from 25 to 35 days based on self-report, and (4) if recently pregnant, participants had to be at least 3-month post-delivery/breastfeeding. Exclusion criteria included: (1) any serious or unstable medical condition (eg, cancers or untreated psychotic disorder), (2) meeting criteria for post-traumatic stress disorder, (3) any medication that may interfere with psychophysiological monitoring during a laboratory session, (4) current substance dependence other than nicotine/caffeine in the past month (measured by the Structured Clinical Interview for Diagnostic and Statistical Manual-IV Axis I Disorders), (5) use of other tobacco products, and (6) participants who were pregnant, breastfeeding, post-hysterectomy or bilateral oophorectomy, or taking birth control or hormone replacement medication that would affect the menstrual cycle. Participants could begin the study at any time during their menstrual cycle leading to full cycle coverage when data were combined across participants. Study procedures were reviewed and approved by the Institutional Review Board at the Medical University of South Carolina (MUSC).

Serum Data Participants

Treatment-seeking cigarette smokers (n = 147) were recruited from the Charleston, SC area. Inclusion and exclusion criteria were similar to the saliva study and are detailed elsewhere.⁷ Serum progesterone was collected at weekly visits for up to 6 weeks to examine the association between changes in weekly progesterone measures and abstinence from cigarettes.

Saliva Data Collection

Following informed consent and eligibility determination, participants began a 14-day observational study, allowing for ad libitum smoking throughout. Participants collected daily saliva samples every morning and completed a daily morning cigarette report on a mobile app for each of the 14 days. Participants completed a Menstrual History Diary to assess the timing of the menstrual cycle for 90 days prior to study entry. Daily diaries assessed the onset of menses during the study.

Ovarian Hormone Collection and Assays

Participants collected salivary samples 30 min after awakening. Samples were placed in containers supplied by study staff and stored in participants’ home freezers until they were delivered to study staff at a subsequent visit (days 7 and 15). The samples were assayed at the South Carolina Clinical & Translational Research (SCTR) Institute Research Nexus (MUSC) using competitive enzyme immunoassay kits manufactured by Salimetrics. Manufacturer data indicate that the saliva measures of progesterone are similar with respect to (1) coefficients of variation and (2) correlation with serum assay levels (rs = 0.8, ps ≤ .001 for saliva/serum progesterone). These daily home collection and storage procedures have been used in numerous previous studies with excellent participant adherence and sample quality.^9–11

Data Reduction and Cleaning

The primary focus of the analysis was to characterize/contrast salivary- and serum-derived progesterone profiles across the menstrual cycle in normally cycling female smokers. Hormone data and matched self-reported menstrual history diaries were examined for this purpose. As neither parent study was designed to determine cycle phases, urinary luteinizing hormone was not available to confirm cycle phase for centering. Of the 115 participants enrolled in the salivary study, 104 (90%) had salivary hormone samples available (11 women did not provide any hormone measurements). Prior to analysis, data were reviewed for consistency and errors. Since the focus is to establish profiles of the hormone trajectories over the entire menstrual cycle, participants providing inadequate information were excluded. Specifically, participants with extended cycle length (Oligomenorrhea, reporting >35 days since last self-reported menses during the study, n = 6), those not exhibiting progesterone peak during the luteal phase (possible oligomenorrhea based on self-report, n = 7), data that did not match menstrual history diary (n = 9) were not included in the analysis. The remaining 82 participants (79%) were included in the analysis. These participants provided 1244 total salivary samples for an average of 41 hormone samples for each day (range: 26–54 samples/day). In the previous study of serum progesterone, 108 (77%) of the 141 randomized participants had at least one hormone measure available and were included in the model. Serum progesterone was measured weekly (rather than daily) and centered around each self-reported menses for a total of 449 observations.

Estimation of the Progesterone Profiles

Preliminary visual examination of the salivary progesterone levels using Spaghetti plots suggested a mathematical model similar to the well-established serum profile previously described.^12,13 This was accomplished using a negative exponential function shown in Equation (1), where progesterone level at time t is denoted $P_t$.

$${\displaystyle \begin{array}{l}{\displaystyle P_t=\alpha +\beta {\mathrm{e}}^{-{(t-\gamma )}^2/\delta }}\end{array}}$$ (1)

where α describes the mean salivary progesterone level during the follicular phase, β describes the increase to peak progesterone during the early luteal phase, γ describes the day when the level reaches the peak, and δ describes the variability of the day of the peak. These time-dependent functions were used to approximate daily progesterone (ng/mL) levels over the course of a standardized menstrual cycle of 30 days. Established serum models have a single peak occurring around day 22, a peak progesterone level of around 16.7 ng/mL serum, and a follicular phase level of 0.25 ng/mL serum. The salivary data demonstrated similar characteristics; the magnitude of the peaks appeared less pronounced. The goal was to provide corresponding estimates of the four parameters of the model that might be expected from typical cycles when salivary data are used. In addition, because Spaghetti plots also suggest inherent heterogeneity in salivary progesterone levels, the goal includes the estimation of variability expected in some of the characteristics of the model. The corresponding random effects non-linear model used to estimate these characteristics is described in Equation (2).

$${\displaystyle \begin{array}{l}{\displaystyle P_{it}={\alpha }_i+{\beta }_i\ast {\mathrm{e}}^{-{(t_i-\gamma )}^2/\delta }+{\epsilon }_{it}}\end{array}}$$ (2)

$P_{it}$ represents the progesterone level on the tth cycle day for the ith participant and ${\epsilon }_{it}$ represents the residual at cycle day t. The residuals, ${\epsilon }_{it}$, were assumed to be Gaussian, namely $N(0,{\sigma }^2)$, where ${\sigma }^2$ represents the sampling variance. To incorporate and examine the heterogeneity in the follicular phase progesterone level and the rate of change during the luteal phase, ${\alpha }_i$ and ${\beta }_i$ were assumed random effects, distributed Normal with variances ${\sigma }_{\alpha }^2$ and ${\sigma }_{\beta }^2$, respectively. Following completion of the salivary data analysis, serum measures of progesterone from female cigarette smokers were examined. Serum models were developed using a similar routine and descriptively compared with the salivary results noted in this study.

Results

The analysis cohort (n = 82) included in the estimation of the progesterone function was on average 30.5 (SD = 7.1) years old. The women provided on average 15.0 (SD = 3.2) progesterone samples. Excluded participants were of similar age (33.9; SD = 7.1). The serum progesterone study cohort (n = 108) was on average 32.7 (SD = 7.6) years old.

Progesterone Parameters

Study participants contributed a total of 1244 salivary progesterone measurements for the analysis. The model fit exhibited relative stability during the follicular phase and a clear unimodal peak during the luteal phase (Figure 1a). Estimates from the non-linear function shown in Equation (3) suggested a mean peak progesterone level near day 24 with a mean peak progesterone level (α + β) of 226 pg/mL, a level nearly twice the follicular phase average (α) of 122 pg/mL and began increasing between cycle days 10 and 15.

Figure 1.

Raw (a) salivary and (b) serum progesterone data showing mean with 95% CI and range of data. Cycle day is noted as the number of days since the start of the most recent self-reported menses. If menses occurred during the 14-day measurement period, measurements were centered around that menses.

$${\displaystyle \begin{array}{l}{\displaystyle P_t(\mathrm{s}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{v}\mathrm{a}\mathrm{r}\mathrm{y},\mathrm{p}\mathrm{g}/\mathrm{m}\mathrm{l})=122.1+103.7\times {\mathrm{e}}^{-{(t-23.6)}^2/68.3}}\end{array}}$$ (3)

Large observed variations between subjects in both the follicular level as well as the peak luteal level in salivary measures may be caused by random heterogeneity at the subject level. The variance of the random effect for α, ${\sigma }_{\alpha }^2$ = 60 (SE = 6.1) and for β, ${\sigma }_{\beta }^2$ = 101 (SE = 13) were both significant (p < .01).

Similarly, unique serum progesterone samples (449 serum progesterone measurements) were assessed (Figure 1b). Model component estimates from the non-linear function (Equation 4) indicated an estimated peak progesterone level near study day 22 with a peak progesterone level of (α + β~) 10.6 ng/mL and the follicular phase average of (α~) 1.1 ng/mL and began increasing between cycle days 10 and 15.

$${\displaystyle \begin{array}{l}{\displaystyle P_t(\mathrm{s}\mathrm{e}\mathrm{r}\mathrm{u}\mathrm{m},\mathrm{n}\mathrm{g}/\mathrm{m}\mathrm{l})=1.06+9.52\times {\mathrm{e}}^{-{(t-22.3)}^2/46.0}}\end{array}}$$ (4)

The variance of the random effect for serum α was not significantly different from zero and for β, ${\sigma }_{\beta }^2=5.2(\mathrm{S}\mathrm{E}=0.5)$ was significant (p < .01). The values of the peak progesterone values from our serum data are lower than the values obtained from the Harris study,¹³ likely due to the weekly measurements and timing of the subject-level peak progesterone.

Discussion

Salivary progesterone measured daily during the menstrual cycle followed a similar functional form when compared with weekly serum measurements, although the peak salivary level occurred slightly later on average (~1.5 days). Results also indicated that variation in salivary progesterone across individuals may be larger than in serum studies with a less distinct peak.¹⁴ Inter- and intra-individual variation in salivary data may be greater than, and do not mirror, known serum measures derived from independent samples. This may be due to the variation in salivary progesterone throughout the day that is not present in serum progesterone levels, indicating that verification of early morning sampling may be required.¹⁵ However, it is clear that frequent monitoring of salivary sex hormones can produce response profiles similar to hormone levels derived from serum samples.

The limitations of this study design were the use of menstrual history diary data to determine cycle timing. Even with data validation and comparisons between subject’s hormone assay and menstrual history data, it is possible that individual cycle anchor points be off by one or more days. Furthermore, participants were paid for salivary sample collection at weekly visits and there was no way to verify that samples were taken on the days and times reported, whereas serum samples require in-person procedures and therefore greater precision in the temporal parameters of data collection. Moreover, since we collected an average of 14 contiguous days of hormone data for each participant, we did not capture the full menstrual cycle hormone profile of any one participant but rather averaged over a cycle across all participants.

The functional form of salivary progesterone in female smokers has not been well characterized in prior studies and the results presented here should bolster confidence in extant and future findings based on salivary progesterone collection methods. Additionally, characterization may aid in the development of sex-specific intervention in women smokers; specifically, by strategically timing medication-assisted tobacco quit attempts with greater precision. However, serum levels of progesterone may not always correlate well, and individual differences in salivary secretion may affect this correlation.¹⁴ Furthermore, the ratio of plasma to salivary progesterone varies during the cycle and is much higher during the luteal phase when compared with the follicular.¹⁶ In addition, the diurnal pattern in salivary progesterone noted by Konishi et al. may be distinct from serum levels and should be included in study design and analytic modeling. The principal advantage remains that serial salivary ovarian hormone sampling can be done with greater ease and lower participant burden than serial serum sampling while rigorous controls of sample collection must be implemented (consistent collection timing, proper freezing). A more complete understanding of the relationship between salivary- and serum-derived hormone levels could be achieved in future comparative studies where both types of samples are collected daily and across all phases of the menstrual cycle.

Supplementary Material

A Contributorship Form detailing each author’s specific involvement with this content, as well as any supplementary data, are available online at https://academic.oup.com/ntr.

Author Contributions

NLB and VR had full access to all the data in the study and take full responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: NLB, VR, KMG, and MES. Acquisition, analysis, or interpretation: all authors. Drafting of the manuscript: NLB, VR, KMG, and MES. Critical revision of the manuscript: all authors. Statistical analysis: VR and NLB.

Funding

This study was supported by National Institutes of Health grants from the National Institute on Drug Abuse (R03DA048227, U54DA016511, P50DA016511, and K12HD055885). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Declaration of Interests

KMG has provided consultation to Jazz Pharmaceuticals and Pfizer, Inc. MJC has provided consultation to Pfizer, Inc. RLT has provided consultation to the American Society of Addiction Medicine. No other authors have any financial, intellectual, or scientific conflicts of interest to disclose.

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.