Correlates of preconception and pregnancy hair cortisol concentrations

Abstract Assessing factors that influence chronic stress biomarkers like hair cortisol concentrations (HCCs) in pregnancy is critical to prevent adverse pregnancy outcomes. Thus, we aimed to identify correlates of HCC preconception and during pregnancy. 2,581 pregnant women participated in the study. HCC was available at four time periods: pre-pregnancy (0–3 months preconception, n = 1,023), and in the first (1–12 weeks, n = 1,734), second (13–24 weeks, n = 1,534), and third (25–36 weeks, n = 835) trimesters. HCC was assessed using liquid chromatography tandem mass spectrometry (LC-MS/MS). Sociodemographic, pregnancy- and hair-related characteristics, and measures of psychosocial stress, were interrogated as potential correlates of HCC. Spearman correlations, paired t-tests, and ANOVA were used to assess differences in log-transformed values of HCC (logHCC) across maternal characteristics. Multivariable linear regressions were used to identify the correlates of HCCs after adjusting for confounders. Mean logHCC values increased across the four prenatal periods ( P  < 0.001). In multivariable analyses, pre-pregnancy BMI was consistently associated with all HCCs, while gestational age, economic hardship, hair dyeing, and depression, showed time-specific associations with HCC. In conclusion, this study showed evidence of factors influencing HCC levels before and during pregnancy. The most consistent association was seen with pre-pregnancy BMI. Depression was also associated with HCC concentrations.


INTRODUCTION
Psychological distress is one of the most pressing issues of public health worldwide.Globally, more than 250 million people are affected by depression 1 and more than 45 million people suffer from anxiety, making both disorders the leading contributors to global burden of disease and global disability [1][2][3][4] .Women are disproportionally more exposed to stressors and factors that contribute to depression and anxiety than men 4,5 , particularly in women of reproductive age and from low-and middle-income countries (LMICs) 5,6 .In LMICs, symptoms of antenatal psychological distress, including depression and anxiety, are often unrecognized and untreated 7 , and are associated with an elevated risk for preterm birth [8][9][10][11][12] , low birth weight 13 , postpartum depression [14][15][16] , and impaired infant neurodevelopment 17,18 among others, implicating adverse health outcomes across generations 8 .
The biological mechanisms linking antenatal distress to adverse health outcomes remain incompletely understood, partly because the speci c etiology of antenatal distress, such as depression and anxiety itself is highly multifaceted [19][20][21] .One facet that contributes to anxiety and depression etiology is the experience of severe or chronic stress 22 .A well-understood biological stress response system is the hypothalamic-pituitary-adrenal (HPA) axis, which becomes increasingly activated during stressful experiences, leading to the secretion of glucocorticoids and catecholamines [23][24][25] .The effector glucocorticoid of the HPA axis is cortisol, which can be used as an objective biomarker of stress [23][24][25] .Due to situational and diurnal changes in cortisol secretion, cortisol measurement is challenging 26 .Responding to these challenges, cortisol measurement in hair emerged as a promising biomarker, because it provides retrospective, aggregate measurement of cortisol secretion over long time periods (e.g. up to 9 months) 27,28 .
Interestingly, HPA axis glucocorticoids also undergo massive alterations during the pregnancy course [29][30][31][32] , indicating that HPA axis activity may be a decisive correlate to the elevated mental disorder prevalence of pregnant women.However, previous studies investigating the association between hair cortisol concentration (HCC) and antenatal psychological distress yielded inconclusive results showing both, positive associations 33,34 and null ndings [35][36][37] .Some of these studies have only poorly or highly heterogeneously addressed the in uence of pregnancy-related and hair-related characteristics on HCC 26 , despite the importance of characterizing HCC correlates to understand its in uence on maternal mental health and pregnancy outcomes.Lastly, sample sizes in previous studies have been rather small (N range 23 to 768, median N = 108), and may not have had adequate statistical power for detecting the effects investigated [37][38][39][40][41] .Given (i), the intergenerational impact on health outcomes associated with antenatal mental health disorders, particularly in LMICs, and (ii) inconclusive prior results about pregnancy-related, characteristics, and hair-related HCC correlates, we investigated (a) the in uence of pregnancy and hairrelated characteristics on HCC levels before and throughout pregnancy, and (b) the association between prenatal HCC and symptoms of antenatal depression, anxiety, and stress, using a large sample of pregnant Peruvian women (N = 2,581).

Population at baseline
We described the characteristics of the study population in Table 1.Brie y, the mean ± SD age of participants was 28 ± 6.3 years, most of them identi ed themselves as mestizo (84%), were married (83%), had normal (18.5-24.9kg/m 2 ) pre-pregnancy BMI (48%) and were multiparous (56%).Half of the participants reported having more than 12 years of education, almost the same proportion (47%) was employed in pregnancy, 39% reported having di culty accessing basic foods, and 57% indicated that the index pregnancy was unplanned.Only 2% and 8% of participants reported tobacco smoking and using alcohol during pregnancy, respectively.Antenatal symptoms of depression and anxiety showed a prevalence of 23% and 48% in our study sample, respectively, and the mean PSS value was 19.9 7.38.

Table 1
Characteristics of participants at enrollment (N = 2,581).

Characteristics
Mean or N SD or %

Bivariate analysis
In the bivariate analysis, pre-pregnancy BMI was positively associated with logHCCs in all time periods (r range 0.11 to 0.18, P < 0.001).Gestational age at HCC collection was negatively associated with logHCC in the rst and second trimesters (r range − 0.064 to -0.065, P < 0.01), while a positive association was seen with logHCC in the third trimester (r = 0.15, P < 0.001) (Table 2).Parity was correlated with logHCC values pre-pregnancy (r = 0.10, P < 0.001) and in the third trimester (r = -0.09,P < 0.01), and the PSS was inversely correlated with logHCC in the second trimester (r = -0.02,P < 0.01).GAD-7 and the PHQ-9 scores showed a tendency towards an association with logHCC in the second (r = -0.05)and third trimester (r = 0.07) (P ≤ 0.05).Including adjustment variables (maternal age, gestational age, pre-pregnancy BMI, and parity) did not change the magnitude of these correlations substantially (Table 2).Statistically signi cant differences in the geometric mean of HCCs were also seen with respect to categories of maternal age, pre-pregnancy BMI, parity, education, employment, di culty accessing basic foods, infant sex, hair dyeing, frequency of hair cutting, and GAD-7 in at least two time periods (Table 3).As shown in Fig. 3, a trend towards increasing values of logHCCs was seen with increasing values of pre-pregnancy BMI in all time periods (P ANOVA < 0.01).LogHCCs in the rst and second trimesters decreased with increasing years of education (P ANOVA < 0.001), and a U-shape association was seen with maternal age (P ANOVA < 0.05) in the same trimesters.Having di culty accessing basic foods, dyed hair, and a female infant, were commonly associated with higher logHCC in the prepregnancy and pregnancy periods (Fig. 3).Participants with symptoms of antenatal anxiety (GAD-7 score 7) had lower logHCCs in the second and third trimesters compared to women without symptoms of anxiety (Supplementary Fig. 2).No differences in the mean of logHCCs were observed with respect to symptoms of antenatal depression and perceived stress (PSS) in the four periods evaluated (Supplementary Fig. 2).

Multivariable regressions
Parity was excluded from the multivariable analysis as it was strongly correlated with maternal age.Implementing fully adjusted models, we observed that pre-pregnancy BMI, gestational age at HCC, and hair dyeing, were associated with logHCC in different time periods, while di culty accessing basic foods and symptoms of antenatal depression were associated with logHCC at a single time in pregnancy (Table 4).For instance, a 1 kg/m 2 increase in pre-pregnancy BMI was on average associated with a 0.02 to 0.03 unit increase in logHCCs (P < 0.001) across the four prenatal periods (Table 4).Every 1-week increase in gestational age was associated with a 0.01 unit decrease in logHCC in the rst and second trimesters (Table 4), but this association was positive with third trimester logHCC ( = 0.04 logHCC units, 95%CI = 0.02, 0.06).Likewise, women who dyed their hair had 0.12 and 0.20 higher logHCC in the rst and second trimesters, respectively.Participants with symptoms of antenatal depression (PHQ-9 score ≥ 10) had 0.11 (95%CI= -0.20, -0.02) lower logHCC in the rst trimester compared to women with no/mild antenatal depression.Having di culty accessing basic foods was associated with higher logHCC in the third trimester alone ( = 0.18, 95%CI= 0.08, 0.29).Similar associations were obtained in the stepwise adjusted models (Supplementary Table 2).

DISCUSSION
Pregnancy is one of the most critical periods of human life, requiring massive changes in extremely complex physiological circuits.Interferences in these circuits can have adverse effects across generations.Hence, understanding these changes is key to preventing adverse effects and promoting a healthy life for children.Our study investigated HCC, a biomarker of HPA axis activity, with several covariates, including psychosocial stress measures (stress, anxiety, and depression) during the pre-pregnancy and pregnancy periods in a large sample of pregnant Peruvian women.Overall, we observed increased levels of HCC before and throughout pregnancy, peaking in the third trimester.Furthermore, we found that HCC was associated with pre-pregnancy BMI in all four prenatal periods, but time-speci c associations were seen between HCC and gestational age at HCC collection, hair treatment, di culty accessing basic foods, and symptoms of antenatal depression, after controlling for well-known confounders.We found no statistically signi cant associations between HCC and symptoms of anxiety or perceived stress neither before nor during pregnancy.
The mean HCC values observed in our sample were, on average, lower compared to other studies that investigated HCC during pregnancy, as shown in a recent systematic review of 56 studies by Marceau et al. 51 .Nevertheless, the observed mean values in our sample ranging between 3.38 and 5.59 pg/mg, lie within most mean levels observed by Marceau et al. (2020).We are aware of only one study investigating HCC in a pregnant Peruvian sample 52 (not overlapping with samples in this study) 41,44,45,53,54 .Interestingly, Dobernecker et al. 52 observed considerably higher HCC levels in 39 participants, which may be due to the small sample size used and the characteristics of the participants, given that more than 50% reported previous experiences of trauma, implicating a higher chronic cortisol secretion.Furthermore, we observed an increase in HCC across pregnancy, peaking in the third trimester.This increase in HCC across pregnancy was also found in previous studies, although not consistently.For example, Marceau et al. 51 did not nd this increase in half of the reviewed studies, questioning this often-referenced assumption.Recently, this questioning was mitigated, as qualitatively more precise studies examining individual HCC trajectories 55 or more ne-grained time intervals 56 obtained HCC increases across pregnancy, although in a non-linear way, with interindividual differences and massive within-person variations.This non-linear cortisol increase is in accordance with established biological ndings.Beginning with gestational week seven, cortisol secretion is increasingly stimulated by the temporary growth of the pituitary gland and by the placenta [57][58][59][60][61] , due to an isolated excitatory in uence of the placental cortisol-releasing hormone on cortisol secretion, leading to cortisol elevations up to ve-fold in certain tissues 61- 64 .
For providing a methodological foundation for further investigations of HCC in pregnant samples, we investigated the bivariate in uence of twenty-one pregnancy-related, sociodemographic, and hair-related covariates on HCC.The only covariates that showed a signi cant bivariate in uence on HCC in at least two time points and in a consistent manner were maternal BMI, maternal education, di culty accessing basic foods, and infant sex, which is partly consistent with previous studies [38][39][40]45,65 . For istance, we showed a signi cant difference in prenatal HCC by infant sex, whereby women giving birth to a female infant had signi cantly higher HCC preconception and in the rst trimester, compared to women giving birth to a male infant.The vast majority of previous studies did not nd such an effect 33,38,39,44,61,[66][67][68][69][70][71][72][73] .However, Romero-Gonzalez et al. 65 recently reported signi cant differences in prenatal HCC by infant sex.Similar to our ndings, they found a higher rst-trimester HCC in women carrying a female compared to a male baby 65 .These ndings indicate that prenatal experiences of stress may in uence the survival of the fetus and the secondary sex ratio distribution in the population; however, further results from Romero-Gonzalez et al. 65 found no signi cant impact on infant sex by perceived stress.
The correlation observed between educational level and HCC is similarly inconsistent based on results from previous pregnancy studies.While some studies found a statistically signi cant in uence of educational level on HCC [38][39][40]74,75 , others found no such effect 61,68,76-78 . For eample, in a subsample of N = 62 pregnant women from Spain, HCC differed signi cantly regarding educational level across the three trimesters, with participants of higher education showing lower HCC in the rst and second, but not in the third trimester 40 .Our results were directionally and time-consistently related to the associations identi ed by Garcia-Leon et al. 40 . Given the cose relationship between educational attainment and income 79 , it is possible that our nding of the correlation between education and HCC corresponds with the association identi ed between di culty accessing basic foods and HCC in our study.This notion is supported by the fact that we found both variables, education and di culty accessing basic foods, to be correlated, although this correlation was weak.We did not nd other studies except our own 45,53 investigating di culty accessing basic foods among pregnant women, which might be explained by the uniqueness of our LMIC sample, as most studies examining HCC among pregnant women have been conducted in high-income countries in North America and Europe 51 where including di culty accessing basic foods might be less important.
In bivariate and multivariable analyses, we found that pre-pregnancy BMI was the strongest predictor of HCC before and during pregnancy.This nding is in accordance with some 39,44,55,66,74,75,80,81 but not all 38,45,67,77,82,83 previous studies.To our knowledge, only two studies have investigated the association between HCC and maternal BMI across all three trimesters 40,74 with partly con icting ndings.While Bosquet Enlow et al. 74 found a signi cant positive correlation between BMI and HCC in the second but not in the rst and third trimesters in a sample of n = 93 U.S.-American women, Garcia-Leon et al. 40 failed to nd such an association in their Spanish sample of n = 62 women.As we investigated HCC in a pregnant sample three to ten-fold larger than in previous studies, our ndings of an unbiased and consistent association between pre-pregnancy BMI and HCC in all four prenatal periods (i.e., from preconception to the third trimester), add to this existing evidence.An association between BMI and HCC is likely, as associations between measures of obesity and both, HPAaxis activity 84 and stress 85 , are well established in non-pregnant samples.
From the different psychosocial stress measures considered in our study, maternal symptoms of antenatal depression was the only measure associated exclusively with rst-trimester HCC after controlling for relevant covariates.While such ndings have been reported in previous studies, evidence is still con icting about the temporal aspect of this association.For instance, there have been studies that report a positive association of HCC with antenatal depression in every trimester of pregnancy 33,34 , except for the pre-pregnancy period 41 .However, an equal number of studies support a null association of antenatal depression with HCC in late pregnancy [35][36][37] , or throughout pregnancy 41 .Identifying an association of depression with HCC restricted to the rst trimester is a plausible nding.One explanation for this nding is the fact that levels of HCC in early pregnancy are lower than in subsequent trimesters, where they tend to be more in uenced by physiological factors, which implies that the associations between psychosocial stressors like depression and HCC are more likely to be detected in early versus in late pregnancy 86 .This observation agrees with the concept of a "ceiling effect" for stressors measured in pregnancy, which suggests that their in uence is only apparent when the physiological response measured is low to moderate 41 .The fact that in our study we assessed antenatal depression at the time of the interview in early and mid-pregnancy and HCC from preconception to the third trimester, indicates that our results are less likely to be in uenced by the time in which the stressor was measured and more related to the time in which the psychoendocrine association is interrogated.
BG, SES, and LD conceived and designed the study.DLJQ conducted the statistical analyses.CK led the laboratory analysis.DLJQ, RK, BG drafted the manuscript.DLJQ, RK, SES, LD, CK, GLT, and BG interpreted the data, critically revised the draft for important intellectual content, and gave nal approval of the manuscript to be published.

FUNDING DISCLOSURE
The PrOMIS study was supported by awards from the National Institutes of Health (NIH) (R01HD102342; R21HD102822; R21MH28985, R01HD059835).The STEM-GDM study was supported by awards from Roche Diagnostic Operations Inc (project number 208617-5074547).The funders had no further involvement in the study design, analysis, and interpretation of evidence, not in the manuscript preparation and decision to submit the paper for publication.

Figures
Figure 1 Diagram showing the segmental analysis of HCC from scalp hair samples collected across pregnancy from two pre-birth cohorts in Lima, Perú (N=2,581).
Each hair sample collected in early, mid-, and late pregnancy, was segmented into two hair segments (3 cm each) to assess current HCC measures from segment 1 (0-3 cm from the scalp), and HCC in the 3 months prior to the hair sampling from segment 2 (3-6 cm from the scalp).
Geometric means of HCC across four prenatal time periods (pre-pregnancy, rst trimester, second trimester, and third trimester) in pregnant women in Lima, Perú (N= 2,581).HCC is represented using the geometric mean (pg/mg).P-values comparing the statistical difference in the mean of logHCCs across consecutive time periods, were calculated using paired t-tests.

Table 2
Pairwise correlations between log-transformed values of HCCs (logHCCs) and continuous covariates in pregnant women in Lima, Perú (N = 2,581).Correlations were adjusted for maternal age, gestational age at HCC collection (GA), pre-pregnancy BMI and parity. ≥

Table 3
Geometric means of HCC (pg/mg) at four time periods (pre-pregnancy, rst trimester, second trimester, and third trimester) across categories of maternal ba characteristics (N = 2,581).
Mean (SD) corresponds to geometric mean values and standard deviation of geometric means of HCC in pg/mg.PHQ-9, 9-item Patient Health Questionnaire.7-item Generalized Anxiety Disorder.PSS, Perceived Stress Scale.Mean (SD) corresponds to geometric mean values and standard deviation of geometric means of HCC in pg/mg.PHQ-9, 9-item Patient Health Questionnaire.7-item Generalized Anxiety Disorder.PSS, Perceived Stress Scale.Mean (SD) corresponds to geometric mean values and standard deviation of geometric means of HCC in pg/mg.PHQ-9, 9-item Patient Health Questionnaire.7-item Generalized Anxiety Disorder.PSS, Perceived Stress Scale.

Table 4
Multivariable linear regressions assessing the association of logHCC with maternal predictors.HCC was measured in four time periods (pre-pregnancy, rst trimester, second trimester, and third trimester) (N range 766-1,636).Regression estimates for depression, anxiety and perceived stress were obtained from independent fully adjusted models. a