Variation in maternal urinary cortisol profiles across the peri-conceptional period: a longitudinal description and evaluation of potential functions

P.A. Nepomnaschy1,2,*, K.G. Salvante1,2,*, L. Zeng3, C. Pyles1,2,†, H. Ma3,‡, J.C. Blais3, L. Wen3, and C.K. Barha1,2 Maternal and Child Health Laboratory, Faculty of Health Sciences, Simon Fraser University, 8888 University Drive, Burnaby, BC, Canada V5A 1S6 Human Evolutionary Studies Program, Simon Fraser University, 8888 University Drive, Burnaby, BC, Canada V5A 1S6 Department of Statistics and Actuarial Sciences, University of Waterloo, 200 University Avenue West, Waterloo, ON, Canada N2L 3G1


Introduction
Physiological stress has been implicated in a wide range of negative reproductive outcomes in women including decreased fecundability and increased risk of early miscarriage, as well as alterations in the developmental trajectories of their offspring, in utero and post-natally (Nepomnaschy et al., 2004(Nepomnaschy et al., , 2006Jasienska et al., 2006;McGowan et al., 2009;Parker and Douglas, 2010a,b;D'Anna-Hernandez et al., 2011). Thus, stress as a covariate has become ubiquitous in women's reproduction and maternal-child health studies (Hruschka et al., 2005;Sarkar et al., 2013). Yet, our understanding of women's stress physiology is still very limited. Indeed, assessment of women's stress levels is often carried out in the absence of knowledge regarding the longitudinal changes that the stress axis often undergoes, which can lead to the misinterpretation of results.
Basal functioning and responsivity of the stress axis, or hypothalamicpituitary-adrenal axis (HPAA), changes as women transition between reproductive stages (between phases of the menstrual cycle, from ovarian cyclicity to conception, across gestation, peri-menopause and menopause) (Nepomnaschy et al., 2004(Nepomnaschy et al., , 2006(Nepomnaschy et al., , 2007(Nepomnaschy et al., , 2011Nepomnaschy, 2005). Thus, in order to properly assess women's physiological stress, it is important to first understand how normative ('average' or 'basal' as opposed to stress-induced) HPAA activity varies across reproductive transitions. For several reproductive transitions, however, changes in normative HPAA activity have not been fully described. This is the case for the transition from regular ovarian cyclicity to conception and early gestation.
The earliest available information regarding HPAA activity around the time of conception corresponds to the later weeks of the first trimester (Brien and Darlrymple, 1976;Demey-Ponsart et al., 1982;Abou-Samra et al., 1984;Allolio et al., 1990;D'Anna-Hernandez et al., 2011;Jung et al., 2011). The paucity of information on HPAA activity during the weeks immediately preceding and following conception is a consequence of the designs used in most prenatal studies. For practical reasons, women tend to be recruited after clinical confirmation of their pregnancy; that is at least 6 weeks post-conception. Such recruitment protocols do not allow for the prospective collection of biological specimens before the sixth gestational week.
Producing an accurate description of longitudinal changes in the HPAA's normative (basal) activity in the weeks immediately preceding and following conception is not only important for the correct assessment of women's stress levels, but also for understanding the role of the stress axis in conception, pregnancy maintenance and early embryonic development. Deviations from basal HPAA activity have been linked to lower fecundability (Nepomnaschy et al., 2004;Beehner and Lu, 2013;Whirledge and Cidlowski, 2013), higher risk of miscarriage (Nepomnaschy et al., 2006;Arck et al., 2008;Brunton, 2013) and changes in fetal developmental trajectories (MacLaughlin and McMillen, 2007;Nepomnaschy and Flinn, 2009;Parker and Douglas, 2010a,b). In order to detect changes in HPAA activity (i.e. detect physiological stress), it is first necessary to establish basal function.
To address this important gap in our knowledge, here we present the first longitudinal description of within-and between-women variation in basal urinary cortisol levels during the 12-week period centered on the day of conception (hereafter 'peri-conceptional period'). After describing longitudinal cortisol trajectories, we align them with concurrent reproductive events and embryonic ontogeny to develop hypotheses regarding the potential biological functions of the observed changes.
The results presented here should aid in the interpretation of women's cortisol data collected during the peri-conceptional period and, in so doing, contribute to our understanding of the role of the HPAA in regulating fecundability, pregnancy fate and embryonic development.

Ethical approval
The work performed for the present study was approved by the Research Ethics Board at Simon Fraser University (file number: 2010s0557).

Study population and participants
This article is based on data and urinary specimens collected in the context of the 'Society, Environment and Reproduction' (SER) study. SER is a naturalistic, longitudinal study on the interactions between the stress and reproductive axes in women. This study began in the year 2000, and it is based on two neighboring rural Kaqchikel Mayan communities located in the southwest highlands of Guatemala. All women in these two communities who matched the following inclusion profile were invited to participate in SER: parous, living with a co-resident male partner, not pregnant and not using any form of chemical contraceptive method (Nepomnaschy et al., 2004(Nepomnaschy et al., , 2006. A total of 107 women were enrolled and provided informed consent, of which 22 conceived and maintained their pregnancies during the study period through, at least, the first trimester of gestation. The 22 participants were not taking any medication at the time of sample collection. They ranged in age from 19 to 36 years (mean + SEM ¼ 27.5 + 1.07 years), in body mass index (BMI) from 19.13 to 31.29 kg/m 2 (mean + SEM ¼ 21.88 + 0.61) and in parity from 1 to 6 children (mean + SEM ¼ 2.9 + 0.3 children). All the analyses in this paper are based on the data from these 22 women. The collection of the urinary specimens analyzed for this article took place between November 2000 and October 2001.

Data and bio-specimen collection
We collected a FMU sample three times per week (i.e. every other day with the exception of Sundays) from each participant for the duration of the study or until a clinical pregnancy was detected (i.e. greater than Week 6 of gestation) (Nepomnaschy et al., 2004(Nepomnaschy et al., , 2006. On average participants provided 27 + 8 (mean + SD) FMU samples (range: 8 -35) during the 12-week periconceptional period (average rate of compliance with the collection protocol was 75%, i.e. 27 of 36 possible samples). This sampling design allowed us to monitor day-to-day variation in urinary free cortisol levels within-and between-individuals (Nepomnaschy et al., 2011). Upon waking, participants collected their first urinary voids in clean, dry, inert plastic containers provided by our research team the previous night. Local female research assistants visited participants at their homes to obtain each FMU sample on the day it was collected and transported the samples in coolers filled with blue-ice packs to our field station within 4 h of the urinary void. FMU samples were aliquoted into 2 ml cryo-vials and stored at 2108C at the station. Samples were transported on dry ice from the field to our laboratory, where we stored them at 2808C until analyses.

Hormone analyses
We evaluated urinary concentrations of biomarkers of stress, ovulation, conception and pregnancy maintenance. FMU levels of cortisol were measured to assess HPAA function. FMU estrone-3-glucuronide (E1G) and pregnanediol glucuronide (PdG), urinary estradiol and progesterone conjugates, were quantified to identify the day of ovulation, while human chorionic gonadotropin, beta sub-unit (hCG-b) was used to identify and confirm conception and pregnancy maintenance. All hormone assays were performed in 2012 at the Maternal and Child Health Laboratory at Simon Fraser University, BC, Canada.
We quantified FMU cortisol, hCG and E1G levels using a multiplex enzyme immunoassay (EIA) array (Quansys Biosciences, Logan, UT, USA) that was previously validated by our laboratory (Salvante et al., 2012). FMU PdG levels were quantified using a competitive solid-phase EIA (O'Connor et al., 2003) based on the Quidel anti-PdG monoclonal antibody, clone 330, kindly provided by Dr. Bill Lasley, University of California at Davis, Davis, CA, USA. We ran all assays in duplicate and re-assayed samples for which the coefficient of variation between duplicates was .13%. All intraand inter-assay coefficients of variation were within acceptable ranges (Table I). To account for variability in hydration state, we corrected concentrations for specific gravity using refractometry (Miller et al., 2004;White et al., 2010).
Treatment of outliers (i.e. data points .2 SD within individuals): (a) if all hormones quantified in a given sample were outside an individual's ranges, we inferred an error in the assay and re-ran the hormone array for that sample. If the re-run yielded similar results, we assumed the sample was unreliable and excluded it from further analyses; (b) if a single hormone value was outside an individuals' range, but the rest of the hormone levels were within their expected ranges we assumed that there were no problems with sample or assay quality and, thus, we retained the data point.
FMU cortisol as a biomarker of HPAA activity FMU cortisol levels are regularly interpreted as the reflection of adrenal cortisol secretion during the period between the last urinary void before an individual retires for the day and their first morning urinary void (Jerjes et al., 2006;Sarkar et al., 2013). This interpretation has resulted in the use of FMU cortisol, also known as nocturnal urinary cortisol, as a marker of HPAA activity (Pollard, 1995;Nepomnaschy et al., 2004Nepomnaschy et al., , 2006Nepomnaschy et al., , 2011Nepomnaschy et al., , 2012Jerjes et al., 2006;Valeggia and Ellison, 2009;o Hartaigh et al., 2012;Sarkar et al., 2013;James et al., 2014). Circulating cortisol is metabolized by cytochrome p450 enzymes into 6beta-hydroxycortisol. Therefore, urinary levels of cortisol reflect the non-metabolized portion of circulating free cortisol. Although the ratio of urinary cortisol:6beta-hydroxycortisol is often used as an indirect measure of CYP3A4 activity, for technical reasons we were unable to assay 6beta-hydroxycortisol in our urine samples. Cytochrome p450 enzymes also metabolize gonadal steroids such as estradiol and progesterone. Thus, if the affinity of this group of enzymes were higher for  progesterone and estradiol than for cortisol, it could be argued that regular changes in circulating levels of those gonadal steroids could influence the levels of cortisol measured in urine. There is no evidence, however, for the existence of differences in affinity of cytochrome p450 enzymes for any of those three steroids (Kerlan et al., 1992;Domanski et al., 2000;Bu, 2006). Furthermore, there is no evidence suggesting that changes in circulating levels of gonadal steroids in regular ovarian cycles are associated with variation in urinary free cortisol levels. Specifically, in a recent study we found that the day of the menstrual cycle did not predict urinary cortisol levels during an average 28 day cycle (i.e. between days 214 (follicular phase) through +14 (luteal phase), wherein day 0 is the day of ovulation), despite the large predictable fluctuations in both estrogen and progesterone on specific days of the cycle (Nepomnaschy et al., 2011). These results were consistent with other studies based on salivary cortisol (Kudielka and Kirschbaum, 2003;LeRoux et al., 2014), which also reflects levels of circulating free cortisol (Lewis, 2006;Grö schl, 2008). Therefore, for this study we assume that urinary cortisol levels reflect HPAA activity and that changes in gonadal steroids across the menstrual cycle do not affect levels of free cortisol measured in urine.

Inclusion criteria and the alignment of ovarian cycles
To avoid the influence of early pregnancy loss, which has been linked to increased cortisol levels (Nepomnaschy et al., 2006), on cortisol trajectories we excluded conceptive cycles that resulted in miscarriages during the first trimester. Given the importance of accurately ascertaining day of ovulation for our analyses, we also excluded pregnancies that were conceived before women enrolled in the study or while they were temporarily absent from it. Conceptive and non-conceptive cycles were aligned on the day of ovulation for an appropriate comparison of cortisol trajectories during the follicular phase (Nepomnaschy et al., 2004(Nepomnaschy et al., , 2006. Alignment of conceptive and non-conceptive cycles on the day of ovulation is more accurate than alignment on the first day of the cycle due to potential differences among women in the length of the follicular phase. Furthermore, the lack of a midluteal progesterone peak in conceptive cycles, which do not contain complete luteal phases, prevents their alignment using this hormonal milestone.

Statistical analyses
All statistical analyses were performed using R (Version 3.0.1, 2012, R Development Core Team, Vienna, Austria). To normalize the data distribution, we log transformed (base 10) all specific gravity-corrected FMU cortisol values (referred to simply as cortisol hereafter) (Keene, 1995). While the number of participants included in our analyses is modest (n ¼ 22), our repeatedmeasures design includes an average of 27 + 8 (mean + SD) data points per participant. When analyzed using mixed-effects modeling, as we have used in this study, the heteroscedasticity and dependence among the repeated measurements is incorporated through random effects, which increases the precision of estimates and results in higher statistical power (Burton et al., 1998). A P value ,0.05 was considered as being significant.

Average cortisol trajectory across the peri-conceptional period
A graphic representation of the average cortisol trajectory was developed using the locally-weighted scatterplot smoothing (LOESS) 'xyplot' function in the 'lattice' package in R. LOESS is a non-parametric regression smoothing method that fits local polynomial regressions to localized subsets of the data and joins them together to build a smooth function describing the deterministic variation in the data (Cleveland, 1979). The objective is to identify the estimated curve that best represents the population curve. LOESS curves allow for the visualization of the underlying structure of the longitudinal hormone profiles; however, no confidence intervals and tests on individual coefficients associated with the time trend can be obtained. To that aim we used linear mixed-effect models with regression splines, which can account for both intra-and inter-individual variation in cortisol through random effects (Burton et al., 1998;Edwards et al., 2006). Mixed-effects models were, thus, used to isolate changes in cortisol as predicted by day and reproductive status (e.g. pre-versus post-conception, days from conception etc.) (Burton et al., 1998;Hruschka et al., 2005). We included maternal age, BMI and parity as covariates in the spline analysis as these factors have previously been shown to influence maternal cortisol secretion during pregnancy (D'Anna- Hernandez et al., 2012;Peer et al., 2013;Conde and Figueiredo, 2014;de Vries et al., 2014).

Preconception versus post-conception cortisol levels
To compare cortisol levels between the pre-and post-conception periods, we fit the following linear mixed effects model where x ij indicates whether the sample was collected pre-(x ij ¼ 0) or postconception (x ij ¼ 1), b 0 represents the average preconception log cortisol, b 1 is the average change in log cortisol post-conception, b 0i and b 1i are individual specific random effects for preconception values and change in postconception values, respectively and e ij is random error due to intra-individual variation.
To evaluate the amount of inter-individual variation in the change in cortisol from pre-to post-conception, we conducted a likelihood ratio test (LRT) to compare the fit of two different linear mixed effects models: one including the random effect term b 1i x ij and another without it. The model that included the b 1i x ij term allowed individuals to vary from one another in the extent to which their post-conception cortisol changed from their preconception values.

Directional trends in cortisol across the peri-conceptional period
The LOESS curve ( Fig. 1) showed that cortisol's trajectory varied dynamically. Thus, we decided to evaluate the magnitude and significance of these directional trends across the whole peri-conceptional period. To that aim, we used a mixed-effects model with regression splines ('lme' function in the 'nlme' package in R). We began with a mixed model containing maternal BMI, age and parity as fixed effects and cubic regression splines with weekly knots (i.e. potential directional changes at Days 235,228,221,214,27,0,7,14,21,28 and 35) as both fixed and random effects. The Akaike information criterion, Bayesian information criterion and log-likelihood values from the fitted reduced models were used to choose the optimal spline basis and knots. The best-fitted model had fixed linear regression splines with knots at Days t* ¼ 27, 0, 28 and 35 and a random linear time trend: where (t -t*)+, known as a truncated line function, equals (t -t*) when t . t* and equals zero when t ≤ t*. The slope of the first period, b 1 , was statistically compared with zero (0), while the slope of each successive period was compared with the slope of the preceding period through coefficients g's. The random linear trend, a i1 , allows between-individual variation in cortisol slopes within each period. This model allows for a more detailed description of the cortisol pattern across the peri-conceptional period because the coefficients associated with the linear time trend variable for each period can be estimated and tested. For the ease of presentation, the final model is equivalently written as separate linear equations for each of the five periods defined by the knots.
Maternal peri-conceptional urinary cortisol profiles 1) Day 242 through Day 27: 2) Day 26 through Day 0: 3) Day +1 through Day +28: 4) Day +29 through Day +35: 5) Day +36 through Day +42: where b 0 is cortisol at the beginning of the study (Day 242), b 2 is associated with maternal BMI, b 3 is associated with maternal parity, b 4 is associated with maternal age, b 1 is the slope of FMU cortisol during the first period, b 1 + g 1 is the slope of cortisol during the second period, b 1 + g 1 + g 2 is the slope of cortisol during the third period, b 1 + g 1 + g 2 + g 3 is the slope of cortisol during the fourth period, b 1 + g 1 + g 2 + g 3 + g 4 is the slope of cortisol during the fifth period, a i0 is the individual-specific random intercept, a i1 is the individual-specific random slope to account for between-individual variation, and e ij is the random error term. The slope of the first period is compared with slope ¼ 0, while the slope of each subsequent period is compared with the slope of the preceding period.

Exploratory analyses
We conducted two sets of exploratory analyses on relevant topics for which we had a smaller sample size: (i) a comparison of pre-ovulatory cortisol trajectories between conceptive and non-conceptive cycles and (ii) a comparison of cortisol trajectories during the early post-conception period between women carrying female and male embryos. Given the size of our sample for these analyses (see below), our goal is not to draw conclusions but to promote research and scientific discussions on these two critical topics.
Cortisol trajectories leading up to ovulation in conceptive versus non-conceptive cycles. Of the 22 women who conceived pregnancies, 12 also provided data from complete non-conceptive ovarian cycles. We used this small sample to explore potential differences in pre-ovulatory cortisol trajectories between conceptive and non-conceptive cycles. Using a mixed-effects model we compared their cortisol levels starting at Day 220 and ending at ovulation (Day 0): Figure 1 Locally weighted scatterplot smoothing curve (blue line) and fitted average (red line) for log 10 -transformed, first-morning urinary cortisol levels in women across the peri-conceptional period (+6 weeks around conception) of 22 pregnancies. Cortisol was corrected for specific gravity, then log 10transformed. Day 0 is the day of conception. The vertical dashed lines at Days 27, 0, +28 and +35 represent 'knots' (i.e. inflection points) at which the trajectory of cortisol changes.
where t ijk is the kth day in the cycle with respect to the day of ovulation (t ijk ¼ 0) for woman i in cycle j, X ij ¼ I indicates that woman i is in a conceptive cycle, where the jth cycle is the conceptive cycle, b 0 is the average log cortisol on the day of ovulation in a non-conceptive cycle, b 1 + 2b 4 t ijk represents the average rate of change in log cortisol in the non-conceptive cycles, b 2 is the average difference between log cortisol levels at ovulation when women are in a conceptive cycle versus when women are in a non-conceptive cycle, b 3 and b 5 represent the difference between the quadratic trends of conceptive and non-conceptive cycles, (b 1 + b 3 ) + 2(b 4 + b 5 )t ijk represents the average rate of change in log cortisol levels in the conceptive cycles, b i is the random woman effect, a ij is the random cycle effect for a woman and e ij is the random measurement error effect.
Cortisol trajectories during the early post-conceptional period by sex of the embryo. Sex data were available for 19 of the 22 pregnancies in this study.
We used t-tests to evaluate the relationship between the sex of the developing embryo and maternal cortisol during each of the first 6 weeks of gestation.

Results
Cortisol fluctuated significantly across the 12-week study period of the 22 pregnancies studied (Table II; Fig. 1). Overall, cortisol increased an average of 2.6 ng/ml per week across the entire peri-conceptional period, from 58.14 + 36.00 ng/ml (geometric mean + SD) 6 weeks prior to conception to 89.29 + 46.76 ng/ml 6 weeks after conception (Table II). Maternal age, parity and maternal BMI were included in the analysis of cortisol trajectory during the peri-conceptional period as potential confounders. The multivariate model showed that, on average, cortisol levels increased 7.6% per year of maternal age (P ¼ 0.0028). Parity, however, was negatively correlated with cortisol levels such that each additional offspring was associated with a 18.4% decrease in cortisol levels on average (P ¼ 0.0077). We found no association between maternal BMI and cortisol levels (P ¼ 0.9375).
In the set of conceptive cycles analyzed (n ¼ 22), cortisol was an average of 18.85% higher in the 6 weeks following conception than in the 6 weeks preceding conception (b 1 ¼ 0.075 + 0.033 (coefficient + SE), P ¼ 0.022). The magnitude of the increase between the pre-and post-conception periods, however, varied significantly between women (LRT statistic ¼ 8.0017, P ¼ 0.005 based on a mixed x 2 distribution). The linear spline mixed effects models revealed significant differences between the population slopes (Table III; Fig. 1). From Days 242 to 27, cortisol followed a mild linear increasing trajectory (1.36% per day; P ¼ 0.0007). Then, after peaking at Day 27, cortisol decreased (P ¼ 0.0013 compared with the trajectory observed during the preceding period) at a rate of 4.02% per day until the day of conception (Day 0). After conception, cortisol started to slowly increase at a rate of 1.73% per day for 4 weeks (P ¼ 0.0008 compared with the preceding period). Between Days 29 and 35 after conception, the previously increasing trajectory turned into a decline (P ¼ 0.0002) at a rate of 6.60% per day. After that 1 week decline period, cortisol's trajectory changed once again to a strong increase (P ¼ 0.0020) at a rate of 8.86% per day between days 36 and 42 postconception, the last week of the study period (Table III; Fig. 1).
Twelve women provided samples from both conceptive and nonconceptive cycles. The analysis of those samples revealed that preovulation cortisol trajectories of conceptive and non-conceptive cycles from the same women were marginally different (P ¼ 0.0633). While in their conceptive cycles cortisol followed a parabolic trajectory, increasing from Days 220 to 26, then decreasing until the day of ovulation (rate of cortisol change ¼ 20.02 -0.0035t ng/ml/day; Table IV), in their non-conceptive cycles cortisol's trajectory was rather flat and time independent. Another interesting source of variation in maternal cortisol levels was the sex of the gestating embryo. During gestational week 2, mothers carrying female embryos (N ¼ 10) had higher mean cortisol levels than those carrying male embryos (N ¼ 9) (t(17) ¼ 2.28, P ¼ 0.04).

Discussion
To our knowledge this article presents the first description and analysis of women's basal peri-conceptional cortisol trajectories. Our longitudinal evaluation of this period uncovered that: (a) FMU cortisol excretion varies significantly across the peri-conceptional period; (b) in conceptive cycles ovulation is preceded and immediately followed by an overall  Maternal peri-conceptional urinary cortisol profiles increase in cortisol excretion, which appears not to be the case for nonconceptive cycles and (c) this overall increasing trajectory is interrupted by two transitory drops in cortisol levels: the first one during the week preceding conception and the second one during the fifth gestational week. An exploratory analysis suggests that basal post-conception cortisol trajectories may differ between women carrying male and female embryos. This novel information highlights the need for further research on HPAA activity during the peri-conceptional period. This knowledge is paramount to accurately assess women's stress and, importantly, to develop interventions to prevent or moderate the undesired effects of physiological stress on women's health, reproductive outcomes and embryonic development.

Preconceptional period
Our observation that conceptive cycles were preceded by at least 5 weeks of a sustained increase in basal FMU cortisol is in line with earlier reports Valeggia and Ellison, 2009). Their analyses of monthly urine specimens from a group of women transitioning from post-partum amenorrhoea to the resumption of ovarian cyclicity showed an increase in c-peptide and basal cortisol in the months prior to the first ovarian cycle Valeggia and Ellison, 2009). Rises in c-peptide production indicate an increase in insulin secretion, a critical mediator of energy intake and storage for specific tissues (Polonsky et al., 1988;Emery Thompson and Knott, 2008). Cortisol increases trigger gluconeogenesis, leading to higher levels of circulating glucose. Ellison and Valeggia argue that this increase in circulating levels of metabolic energy is important for the resumption of ovarian function after post-partum amenorrhea. Consistent with their hypothesis, the increase in basal cortisol we observed preceding conceptive cycles but not non-conceptive ones in our population supports the notion that fecundability may be linked to levels of available energy. Our findings are also compatible with an important body of literature linking stress challenges to reproductive suppression. Psychosocial, energetic and health stress challenges trigger the activation of the HPAA leading to changes in the allocation of available metabolic energy to tissues involved in the stress response (Selye, 1976;McEwen and Wingfield, 2003;McEwen, 2004;Nepomnaschy et al., 2004Nepomnaschy et al., , 2006Landys et al., 2006;Uhart et al., 2006). As metabolic energy is finite, its re-allocation is achieved at the expense of other metabolic tasks that can be momentarily postponed such as reproduction (Ellison and Lager, 1986;Bentley et al., 1999;Jasienska, 2001Jasienska, , 2003Ellison, 2003, 2009;Jasienska and Ellison, 2004;Nepomnaschy et al., 2004Nepomnaschy et al., , 2006Nepomnaschy et al., , 2007Vitzthum et al., 2009).
Importantly, in our sample, the increasing basal cortisol trajectory was interrupted by a period of significant decline during the week preceding conception. This decline may be critical for fecundability. The week preceding conception corresponds to the follicular phase of the ovarian cycle, a critical period for follicle maturation, ovulation and conception. Indeed, elevated levels of cortisol and salivary alpha-amylase, another stress biomarker, during this period have both been associated with lower chances of conception (Louis et al., 2011;Lynch et al., 2014). Stress-induced HPAA activation during this week has been linked to deleterious changes in the secretion of gonadotrophins and gonadal

Table III
Linear spline mixed effect model for log 10 -transformed first-morning urinary cortisol trajectory in women across the peri-conceptional period, estimated regression coefficients (fixed effects).
steroids which can affect follicle maturation (Rivier and Vale, 1990;Chrousos et al., 1998;Nepomnaschy et al., 2007;Zhang et al., 2011). Thus, it is plausible that the observed decrease in basal cortisol levels during the week preceding ovulation of conceptive cycles plays a protective role in the processes involved in follicle maturation, ovulation and conception.

Post-conceptional period
Until the present study, maternal cortisol trajectories post-conception had only been described for later stages of gestation. The lack of information regarding trajectories earlier during gestation is linked to logistical limitations of study designs as most prenatal studies recruit women once they learn they are pregnant, 6-8 weeks postconception (Table V). Later gestational stages (less than gestational week 8) are characterized by a constant increase in basal cortisol levels (Brien and Darlrymple, 1976;Demey-Ponsart et al., 1982;Bustamante and Crabbé, 1984;Allolio et al., 1990;Brunton et al., 2008;Entringer et al., 2010;D'Anna-Hernandez et al., 2011;Jung et al., 2011). This hypercortisolemic state has been argued to constitute a response to the energetic demands imposed by fetal growth (Entringer et al., 2010;Jung et al., 2011),  Yes (Nepomnaschy et al., 2006) Maternal peri-conceptional urinary cortisol profiles weight and volume associated with pregnancy (Butte and King, 2005;Forsum and Lö f, 2007) and the development of energetic reserves in preparation for nursing (Butte and King, 2005). Our results suggest that the increase in basal cortisol levels actually begins earlier, at the moment of conception. The average rate of basal cortisol increase observed right after conception is significantly higher than the one characterizing the 6 weeks preceding it. Future studies should explore the function(s) of this early post-conceptional increase in basal cortisol. Increases in basal cortisol levels during this period may provide the extra metabolic energy required for the processes of implantation and placentation (Clapp et al., 1988;Butte and King, 2005;Forsum and Lö f, 2007).
The post-conceptional increase in cortisol may also contribute to the same processes that lead to hypercortisolemia later during gestation; the allocation and storage of energy to specific tissues involved in sustaining pregnancy later on, the delivery and nursing of the baby and post-natal maternal care. Nonetheless, the basal cortisol levels we observed during the first 6-week post-conception were clearly lower than those reported for later gestational stages, and the rate of increase was more gradual (Brien and Darlrymple, 1976;Demey-Ponsart et al., 1982;Bustamante and Crabbé, 1984;Allolio et al., 1990;Brunton, et al., 2008;Entringer et al., 2010;D'Anna-Hernandez et al., 2011;Jung et al., 2011). This slow post-conceptional increase may be explained by a trade-off between the emerging energy demands of early gestation (Clapp et al., 1988;Butte and King, 2005) and the risks associated with elevated glucocorticoid exposure. Elevated basal cortisol after conception may increase the risk of early miscarriage and could affect the processes of implantation, placentation and the earliest stages of embryonic development (Barnes and Adcock, 1993;Estelles et al., 1994;Barnes, 1998;Jauniaux et al., 2003;Michael and Papageorghiou, 2008). Implantation involves inflammatory processes that mediate structural changes in the endometrium preceding implantation (Granot et al., 2012;Van Sinderen et al., 2012). Thus, as cortisol is anti-inflammatory, maintaining relatively low basal cortisol levels during this period may be paramount to the successful remodeling of the endometrium and implantation (Barnes and Adcock, 1993;Barnes, 1998). Cortisol also plays an important role in placentation. It acts as a modulator of regulatory pathways involved in the invasion of the placental trophoblast (Jones et al., 1989;Ma et al., 2002;Nicholson et al., 2004;Kalantaridou et al., 2007;Michael and Papageorghiou, 2008). Excess cortisol exposure during the formation of the placenta may result in shallow trophoblast invasion. In turn, shallow trophoblast invasion has been linked to a variety of complications such as premature onset of maternal-placental circulation, leading to elevated oxidative stress and increased risk of miscarriage, intrauterine growth restriction and pre-eclampsia (Estelles et al., 1994;Jauniaux et al., 2003;Michael and Papageorghiou, 2008).
The arguments above may also help explain the significant decline in basal cortisol we observed during Week 5 post-conception. Later during gestation the fetus is partially protected from exposure to high maternal cortisol levels (both basal and stress-induced) by the placental expression of 11 beta-hydroxysteroid dehydrogenase 2 (11b-HSD2), an enzyme that converts cortisol into inactive cortisone (Benediktsson et al., 1997). If the amount of placental 11b-HSD2 produced is not sufficient to counter an increase in maternal cortisol, the excess glucocorticoid reaching the fetus can have serious effects on its development. Depending on the timing of the exposure and its magnitude, high maternal cortisol could lead to a broad spectrum of negative outcomes ranging from the alteration of fetal growth rates to an increase in the risk of cardiovascular disease in the adult (Barker, 1990;Gitau et al., 1998;Barker and Bagby, 2005;Reynolds, 2010Reynolds, , 2013Duthie and Reynolds, 2013). Importantly, 11b-HSD2 may not be present to protect the embryo during the weeks immediately following conception. Indeed, the earliest that 11b-HSD2 has been detected is 6 weeks post-conception (Shams et al., 1998;McTernan et al., 2001). Thus, as basal cortisol increases after conception, the days immediately preceding the onset of 11b-HSD2 production may be particularly sensitive to high glucocorticoid levels. Therefore, it is possible that the decline in basal maternal cortisol levels during gestational week 5 may be part of a mechanism protecting the embryo from high glucocorticoid levels at this critical developmental juncture.

Covariates associated with cortisol levels
In our sample, FMU basal cortisol levels were associated with maternal age, parity and sex of the embryo. Counterintuitively, age and parity presented opposite effects. Basal cortisol levels were positively associated with age but negatively associated with parity. Previous studies based on plasma and salivary cortisol have reported modest increases in its basal levels with age (salivary cortisol: (Heaney et al., 2012); plasma cortisol: (Seeman et al., 1995;Kern et al., 1996;Lupien et al., 1996)), but others have reported declines (plasma cortisol: (Lupien et al., 1996)) or no change (plasma cortisol: (Leblhuber et al., 1993;Born et al., 1995;Thomas et al., 1999;de Bruijn et al., 2002)) with increasing age. At this point, it is unclear why these studies yield opposite results even when using the same matrices (for a discussion of this issue, see Lupien et al., 1996). Our results suggest that these inconsistencies may be explained, at least in part, by a potential buffering effect of parity on basal cortisol. Consistent with our findings, a previous study reported that first-time mothers presented more prominent increases in basal cortisol between the second and third trimesters of pregnancy than second-time mothers (Conde and Figueiredo, 2014). Nonetheless, further studies will be needed to confirm these findings and test the existence of the proposed buffering effect of parity on the increase in basal cortisol levels with age.
Sex of the embryo was another factor associated with maternal cortisol levels. To our knowledge, this is the first study to evaluate this relationship. Women carrying female embryos presented higher basal cortisol levels during post-conceptional week 2 than those carrying males. This observation is based on a small sub-sample of pregnancies and, thus, needs to be confirmed by larger studies. Furthermore, this result was not in line with our a priori predictions. Males are, on average, more energetically expensive to produce than females (Clutton-Brock et al., 1984;Byrne and Warburton, 1987;Kline and Stein, 1987;Damjanovic et al., 2009;Catalano et al., 2013). Therefore, women carrying males would be expected to present higher metabolic rates and, thus, higher basal cortisol levels than those carrying females (Damjanovic et al., 2009). Our results, however, may be explained by a sex bias in early miscarriages. The risk of miscarriage is highest immediately after conception (Kline and Stein, 1987;Catalano et al., 2013), and males are more likely to be miscarried than females (Byrne and Warburton, 1987;Wells, 2000). This is particularly so in response to stress (Chason et al., 2012;Davis et al., 1998Davis et al., , 2007Song, 2012;Catalano et al., 2013), which is associated with increases in cortisol. Thus, if mothers who had high overall (i.e. both basal and stress-induced) cortisol levels were more likely to lose males than females, then our sample of male-carrying mothers would be biased toward those who had lower overall cortisol levels.

Summary and future directions
In summary, our results suggest that basal HPAA activity varies dynamically across the peri-conceptional period. In our sample, conceptive cycles were characterized by an overall increase in FMU cortisol excretion spanning the full 12 weeks of the study period. This overall increasing trajectory was interrupted twice by 1-week declines: once during the week preceding ovulation and the other during the fifth week of gestation. FMU cortisol levels were associated with women's age, parity and, post-conception, the sex of their embryo.
We propose that the overall increase in basal cortisol secretion, suggested by its increase in FMU excretion, may mediate a shift in energy allocation from regular metabolic maintenance toward reproductive function. Preconception, the increase in circulating metabolic energy would be allocated to reproductive function, increasing the chance of conception. Post-conception, that energy would be allocated toward meeting the energetic demands of early gestation and storage for use later in pregnancy and post-natal care. Since high overall cortisol levels may elevate the risk of miscarriage and affect embryonic development, the gradual nature of basal cortisol's largely increasing trajectory and the 2 week-long declines observed during the study period may reflect a trade-off between meeting pregnancy's energetic needs and minimizing overall cortisol exposure during specific windows of vulnerability.
Future studies conducted in other populations and including larger sample sizes will be required to confirm our findings, assess their generalizability and evaluate the biological functions of the observed changes in HPAA activity across the peri-conceptional period. Fully understanding normative longitudinal variation in HPAA activity across reproductive transitions is vital to comprehend how deviations from normative patterns can influence maternal health and wellbeing, fecundability, risk of miscarriage and fetal programing as well as to develop interventions aimed at preventing these deviations or moderating their effects.