A linear model best represented the association between segment day and FEV1 (Table 3). An indicator variable for follicular or luteal phase had a significant interaction (p = 0.003) with menstrual segment day, thus separate models are presented for the follicular (Figure 1a) and luteal phases (Figure 1b). The models indicate that the number of days relative to the estimated date of ovulation is associated with percent predicted FEV1 during the follicular phase. This result indicates an overall decrease in FEV1 as the day of ovulation approaches (p < 0.001). No statistically significant association was found between menstrual segment day and percent predicted FEV1 in the five days prior to the onset of menses or the luteal phase (Table 3); however, the overall pattern suggests FEV1 decreases in asthmatic women and increases in non-asthmatic women in the luteal phase (Figure 1b).

	Beta (s.e.)	p-value
Follicular phase
Intercept	91.8 (2.6)	<0.001
Number of days relative to estimated day of ovulation	−0.49 (0.11)	<0.001
Luteal phase
Intercept	90.6 (2.9)	<0.001
Number of days relative to estimated day of ovulation	0.02 (0.10)	0.84
5 Days prior to Onset of Menses
Intercept	92.7 (3.6)	<0.001
Number of days relative to onset of Menses	−0.02 (0.45)	0.97

The GEE models that examined hormone levels were best fit with a quadratic term to accommodate the nonlinear relationships between sex hormones and FEV1. Direct model interpretation is complicated due to the quadratic term. Overall, statistically significant associations were found between the E1C/PDG ratio (Figure 2) and PDG (adjusted for E1C) (Figure 3) and FEV1 in the follicular phase (Table 4). However, different patterns emerged when the models were stratified by asthma status. Among non-asthmatic women, statistically significant associations were found between the E1C/PDG ratio (Figure 4) as well as E1C (Figure 5) and PDG (Figure 6) and FEV1 in the follicular phase (Table 4). Although not statistically significant, only the association with E1C persisted in the luteal phase. Additionally, E1C was associated with FEV1 in the five days prior to bleeding only among non-asthmatic women (Table 5). Among asthmatic women, there were no statistically significant associations between the sex hormones and FEV1 in either phase (Table 4) or in the 5 days prior to bleeding (Table 5).

Results were unchanged after excluding samples from three women without complete menstrual segments.

In the open-ended comments made at the end of the study, women reported overall that the study protocol was “easy” or “simple”. Participants commented that urine collection was complicated when travel occurred. Several commented that because urine collection was not “clean catch” it was not a burden. Some women commented that having the pink rubber duck in the bathroom was a good reminder while one woman commented that only having the urine collection cup in the bathroom worked as a reminder.

Detailed evidence of a direct relationship between sex hormones and lung function within-woman over time has been missing from the discussion of the role sex hormones play in asthma incidence and exacerbation. This research establishes methods, including statistical methods and daily hormone measurements, to be applied to future research of this important, yet poorly understood, relationship.

Our results suggest an association between sex hormones and FEV1 that varies by asthma status and menstrual phase (follicular or luteal). Interestingly, there were no statistically significant associations between sex hormones and FEV1 in women who reported a prior doctor diagnosis of asthma; associations were found among women without a prior asthma diagnosis. The differences in associations between sex hormones and FEV1 by asthma status suggests a worthy avenue to pursue in explaining why there may be worsening of asthma around times of hormonal change such as menses and pregnancy. Why might the associations between sex hormones and lung function vary by asthma status? Perhaps our asthmatic women were well controlled and thus might have different relationships between their sex hormones and their FEV1 than uncontrolled asthmatics would have. Further, a larger sample would allow detection of weaker associations and a study with at least 2 consecutive menstrual cycles would be more informative.

Mandhane et al. hypothesized there would be changes in asthma characteristics over the course of a menstrual cycle and that those changes would be “blunted” in women taking oral contraceptives (OCs – all were combined estrogen and progesterone formulations) [13]. They studied 17 asthmatic women (9 took OCs) over the course of a menstrual cycle. Daily measurements of exhaled nitric oxide (eNO) and spirometry (FEV1/FVC ratio) were performed and daily sex hormone measurements were determined from saliva. Among OC users, 17β-estradiol and progesterone levels were not associated with eNO; however, among women not using OCs, an increase in progesterone was associated with an increase in eNO and an increase in 17β-estradiol was associated with a decrease in eNO. No associations were reported between either of the sex hormones and the FEV1/FVC ratio in either OC users or non-users. The authors did not estimate ovulation among those not using OCs and did not consider menstrual “phase” as a potential effect modifier.

Farha et al. collected lung function measurements (spirometry, gas transfer, FeNO) once weekly for four to five weeks from 13 asthmatic and 10 non-asthmatic women – some who were taking hormonal contraceptives [14]. They reported that women with asthma experienced cyclic changes in airflow and gas transfer and suggested that the data supported hormonal effects on lung function. They suggested that the cyclic mechanisms differ between asthmatic and non-asthmatic women. The lack of daily measurements limits the strength of their conclusions.

Our research establishes methods that provide the best data to investigate hormone related asthma effects such as premenstrual worsening of asthma. Most prior research in this area has been limited by poor (or lack of) confirmation of menstrual phase, the lack of actual measurement of sex hormone levels throughout a cycle, use of a measurement of lung function other than percent predicted FEV1, statistical methods that do not take into account repeated measures and the failure to include non-asthmatic women [15–19]. Some of these studies have also focused on comparison of symptoms between asthmatic OC users and non-users generally finding fewer symptoms and better asthma outcomes in OC users [20].

Numerous epidemiologic studies have measured daily hormones to either study fertility or ovarian function [7, 8] while prior research on menstrual cycle phase and asthma has used the LMP date to approximate a woman’s menstrual cycle phase. LMP date is used to estimate the date of expected ovulation and thus the phase is established. Although this method is very inexpensive and it is easy to collect data on LMP, there are numerous problems with this approach. Recent data demonstrate that cycle length is not predictive of day of ovulation. In the Early Pregnancy Study, daily urine was collected to measure hormones to study subclinical and early pregnancy loss [7]. In this study, Wilcox et al found that among the 69 cycles that were 28 days long, only 10% of the cycles were associated with ovulation 14 days before the next menses. They also reported that the time from ovulation to the next menses ranged from menstrual cycle days 10 to 22 in these 28 day cycles. Using the same data, Harlow et al. examined the hormonal patterns of 28 cycles with follicular phases of 24 days or more [21]. There were five different hormonal patterns observed among these cycles. Four of the cycles >38 days appeared to be “double cycles” with no bleeding. They further report that the daily estrogen profiles in long cycles are heterogeneous.

These data provide evidence that using LMP date is not a valid method to estimate either hormone levels or menstrual phase in women. The follicular phase can be highly variable – even in ovulatory cycles. Thus, LMP is not a valid measure to estimate menstrual phase and infer hormone patterns. Separately, fertility monitors only indicate the timing of probable ovulation. They also require the use of a daily urine specimen, but do not provide measurements of all important sex hormones or their ratios which would be critical for their study. Fertility monitor accuracy, especially in research studies, has not been well established [22]. Daily hormone measurements are needed if an accurate relationship between actual levels of hormones (and their ratios) and lung function/asthma symptoms are to be determined.

In our study, daily spirometry was not performed by a certified provider due to financial constraints. Women were asked to complete only one useable reading. While these are limitations, women did use the same device for the entire protocol and all received the same instructions to use the device. Also, as part of the study protocol, women met with the research assistant three times per week and the research assistant was able to discuss protocol compliance, as well as confirm proper use of the device. While the device meets American Thoracic Society (ATS) recommendations for accuracy and precision in measuring peak flow, it is not clear if validation testing has been conducted for FEV1. However, we examined within-woman patterns based on the same device used day after day. It would not have been logistically or economically feasible to have women perform spirometry on an ATS approved spirometer daily under the supervision of a NIOSH certified technician. We propose that since we are conducting research and not providing clinical care or medical advice based on results from the device, use of this device is appropriate.

While including only 13 women in our analyses is a limitation, use of daily hormone measurements and repeated FEV1 measurements over a single menstrual segment for each woman are innovative. The use of longitudinal analyses to examine within woman associations provides strong evidence of patterns. We also examined whether absolute levels or ratios of hormone levels were associated with FEV1 levels which is novel.

In summary, our analyses of daily hormone and lung function measurements provide early evidence of differences between asthmatic and non-asthmatic women over menstrual segments. Most importantly, the methods demonstrated here provide optimal evidence for investigating the role of hormones in lung function in asthmatic and non-asthmatic women. These data provide preliminary evidence that the associations between sex hormones and percent predicted FEV1, a measure of lung function, vary by asthma status and phase of the menstrual cycle. The study of contiguous cycles from a larger group of women in their 20s and 30s (most likely to ovulate) would provide the strongest evidence for determining the relationships between sex hormones and lung function, which would lead to insight into the phenomena of so-called hormone-affected asthma.

We learned a series of lessons from this pilot study. First, women should wait until just before their period to start the protocol; however, we should not allow women to start the protocol if they are bleeding as a full menstrual cycle may not be captured during the study. Also, it would be ideal to select times when women do not plan to be out of town during participation. If they must travel, plans should be developed with the participant to collect and store their urine during travel. Necessary supplies, such as ice packs and coolers, should be provided for travel collection and storage. The study protocol should be extended to at least 2 full menstrual cycles (capture 3 bleeds) to allow analyses of within-woman cycle-to-cycle variability of associations.

Sample size calculations, for which our data should prove helpful, should take into consideration that some cycles will be anovulatory or will not permit estimation of ovulation day (18.8% in our study). For the FEV1 measurements, women should perform the exhale function 3 times to allow the investigator to select the best measurement from the electronically stored data. Finally, baseline spirometry with Albuterol challenge would also be useful in describing the lung function of the population and the severity of asthmatics included in the study.