Participants
Women (N = 21) between the ages 18 and 45 reporting regular menstrual cycles lasting 21 to 35 days, reportedly smoking 10 or more cigarettes per day, were recruited. Exclusion criteria included the following: elevated depressive symptoms, defined as a score of 16 or higher on the Center for Epidemiologic Studies Depression Scale (CES-D) [15], use of medications affecting mood or ovarian hormones (e.g., anti-depressants, anxiolytics, hormonal birth control), current use of smoking cessation aids or programs, self-reported psychiatric disorders (e.g., bipolar disorder, psychotic disorder), and pregnancy and lactation (either currently or in the last 12 months). The University of Regina Ethics Board approved this study. Though 22 participants were initially recruited for the study, one participant was excluded due to an excessive amount of missing data (i.e., missing diary submissions and failure to return urine samples) and loss of contact.
Procedures
Enrollment session
During the in-person enrollment session, participants’ eligibility was confirmed and written informed consent was obtained. Participants were provided with ovulation predictor tests, urine collection kits and instructions for both. Relevant demographic information was collected and the Fagerstrom Test for Nicotine Dependence (FTND) [16] was administered. The FTND consists of 6 questions and is a well-established instrument for assessing the intensity of the physical addiction to nicotine. Each answer has a corresponding score which is summed to provide an overall dependency score ranging from low dependence (score of 1–2) to high dependence (score of 8 or higher). The assessment starting point was counter-balanced such that half of the participants began their participation in the study with the follicular phase, while the other half began with ovulation predictor testing and luteal phase testing prior to completing follicular phase testing.
Ovulation predictor tests
Starting on cycle day 8, participants began taking daily ovulation predictor tests (Easy@Home Ovulation Test Sticks; Easy@Home, Chicago, IL) until a positive test was obtained, indicating that ovulation would occur in approximately 12–24 h. Therefore, the day following a positive test was considered “ovulation day” and post-ovulation day 1 was considered the first day of the luteal phase. Participants were instructed to send a picture of each test result to the researcher, allowing the researcher to make a final determination on whether the test was positive or not. In the event of an anovulatory cycle, the participant was instructed to continue with ovulation testing until a positive test was obtained. Any luteal phase data already collected was replaced with the data collected after the positive ovulation predictor test.
Smoking diary
Smoking behavior was assessed over the course of a menstrual cycle on days 1, 3, 5, and 7 of the follicular phase, in addition to days 1, 3, 5, 7, 9, 11, and 13 post-ovulation (Fig. 1). Specifically, on those days, participants were asked to log into the Expimetrics phone application (Expimetrics Inc., Lafayette, IN) and create a new diary entry each time they smoked a cigarette. The number of total entries served as the number of cigarettes smoked per day. For each entry, participants were also asked to indicate their perceived need (5-point Likert scale ranging from 1—could have done without it to 5—desperate) and their enjoyment (5-point Likert scale ranging from 1—hated it to 5—loved it) of each cigarette. Text message reminders were sent by the researcher on the morning of each tracking day.
Urine collection and hormone assays
On the morning following each smoking diary day, participants collected a first-morning voided urine sample, for a total of 11 samples over the course of one menstrual cycle. Samples were stored in the participant’s home freezer until the last urine sample was collected, at which time they were shipped to the lab and stored at −40°C until assayed.
Hormone assays
Urine samples were assayed for estrone-3-glucuronide (E1G) and pregnanediol glucuronide (PdG), which are the urinary metabolites of estradiol and progesterone, respectively. These metabolites have been shown to correlate very highly (rs = 0.93–0.97 [17]; with serum levels of estradiol and progesterone measured 1 day prior to urine collection. For this reason, urine collection occurred the morning following the smoking diary assessment questionnaire completion.
E1G was assayed using an enzyme immunoassay (Arbor Assays, Ann Arbor, MI), with sensitivity at < 22.5 pg/ml-1. The intra-assay and inter-assay variabilities for E1G were 5.1% and 9.1%, respectively. PdG was also assayed using an enzyme immunoassay (Arbor Assays, Ann Arbor, MI), with sensitivity at < 0.180 ng/ml-1. The intra-assay and inter-assay variabilities for PDG were 6.5% and 16.0%, respectively.
Analytic approach
Given the nested data structure of days within participants, Linear Mixed-Effect Modeling (LMEM) was used to analyze the data using a model building approach. The primary outcome variables were (1) the number of cigarettes smoked per day, (2) the perceived need, and (3) perceived enjoyment of each cigarette. Since hormones are well known to follow complex cyclical change patterns [18,19,20], two different analytic approaches were used to explore the relationship between E1G and PdG with the outcome variables. First, the curvilinear (i.e., quadratic) relationship between current hormone levels and outcome variable was modeled. In this set of models, the hormone levels were person-centered according to each person’s average levels (i.e., each participants’ average hormone level subtracted from each of their daily levels). The second modeling approach sought to capture how daily fluctuations in hormones were related to outcome variables. Change scores were calculated between hormone levels across two successive days, capturing day-to-day variability. For example, positive values reflect an increase that day relative to the previous, while negative values reflect a drop in hormone levels from the previous day. Importantly, both sets of models included an interaction between E1G and PdG, to examine if the influence of one hormone was dependent upon levels of the other. All analyses were conducted using the LMER package in R.
Power calculations were conducted using the methods recommended for repeated measures data [21]. Statistical power in the current study was therefore determined by the sample size, the number of repeated observations, and the intraclass correlation observed between repeated outcome measures. Given 21 participants, 11 repeated measures per participant, and an intraclass correlation coefficient of 0.43 for the daily number of cigarettes smoked, a sample size of 21 allowed for the detection of a medium effect of f2 = 0.17 where an f2 of .02 = small effect, .15 = medium effect, and .35 = large effect, as outlined by [22].