Abstract
Introduction
Preterm birth remains one of the most important yet intractable issues confronting obstetricians. While it only accounts for 10% of births, it is the second leading cause of perinatal mortality and serious perinatal morbidity [1]. A number of factors have been implicated in the etiology of preterm birth, the most commonly cited being hemorrhage, inflammation, over-distention, and stress. Among those, stress is perhaps the least rigorously assessed. There are several reasons why stress has not been as frequently studied, including the difficulty of measuring stress, and the obvious bias when measures of stress are obtained in the wake of the birth of a preterm infant. Cortisol is a biomarker that has been used as a measure of stress in many studies since it plays crucial roles in the stress response, including recruiting energy from adipose and muscle, and suppressing the immune system. Hypercortisolism has been associated with medical conditions such as major depression, end-stage renal disease diabetes, cardiovascular disease, and obesity [2]. However, an emerging psychoneuroendocrinology literature has also shown evidence of the relationship of hypocortisolism to stress-related conditions such as chronic fatigue syndrome, fibromyalgia, post-traumatic stress disorder, history of childhood trauma, and most recently, has been reported among children who are victims of bullying [3–7]. The underlying etiology of hypocortisolism has not yet been elucidated, but is thought to be related to hypothalamic-pituitary-adrenal (HPA) axis dysregulation which may make an individual more vulnerable to stressors, and which can manifest as lower than expected cortisol levels. The relationship of maternal cortisol levels to preterm birth has been studied, and most reports have shown an increased cortisol level in women who deliver preterm compared to those who deliver at full term [8–11]. However, these studies used spot serum or salivary cortisol measurements to quantify acute changes in cortisol. Such measurements are limited by physiological cortisol fluctuations due to normal circadian rhythm or acute stress events (such as the preterm birth itself). Thus the elevated cortisol level may well have been the consequence, rather than the cause, of the preterm birth. A single measurement cannot reflect systemic, long term cortisol exposure. Further, while other studies have evaluated cortisol levels using other techniques in women giving birth to term and preterm infants (vide infra), those studies did not control for gestational age (i.e. term patients had levels determined at term), and cortisol levels normally increase throughout pregnancy. Hair cortisol is a validated assay that enables investigators to assess long term cortisol levels over the several months it takes hair to grow [12–14]. It represents the cumulative activity of the HPA axis as an average measure over a period of time. Although it may seem counterintuitive to hypothesize that low cortisol levels are related to preterm birth, the psychoneuroendocrinology literature suggests that impaired cortisol production could be a contributor to some disease states. Therefore, we designed a pilot study that sought to determine whether hypocortisolism is correlated with preterm birth, using hair cortisol as a marker of long term HPA axis activity. In particular we were interested in determining if either stress per se (e.g., loss of employment or loss of a home while pregnant) is associated with harms, or if the inability of an individual to mount an adequate stress response is the trigger for adverse consequences. Thus, we sought to determine if women who have similar measures of stress on validated scales of life events fare differentially in biologic outcomes if they differ in their ability to mount a normal response to “stressful” events as evidenced by their hair cortisol levels.
Materials and methods
After approval by the IRB, we performed a prospective, matched, case-control study in which 29 women who had a preterm birth (PTB) at 24–36 weeks 5 days gestation were compared to 29 women who delivered at term, matched for maternal age (+/– five years), gestational age at time of hair sampling (+/– seven days gestation), and race. Cases’ hair samples were collected within 72 hours of their preterm birth and controls (who delivered at term) had their hair samples collected at the same gestational age (GA) as the corresponding case. Blinded analysis of the hair was performed at the Department of Psychology, Dresden University of Technology. Two hair cortisol measurements were taken; one each from the proximal and distal ends of the strands. The proximal 0.5 cm of the hair strand was discarded to avoid measuring stress events occurring in direct relation to the preterm birth itself and/or the administration of the corticosteroids for fetal lung maturity. The next more proximal segment and the distal segment were measured.
Hair sample collection, preparation, and cortisol assay
Hair strands were carefully cut with fine scissors as close as possible to the scalp from a posterior vertex position. The number of strands obtained differed in accordance with the subject’s permission to cut more or less hair. However, a minimum of 10 mg of hair for a 3 cm segment was obtained from each participant. In the laboratory, the strands were lined up and cut into 3 cm segments. Depending on the individual hair length, between three and nine segments were obtained from the women. For washing of hair and steroid extraction, the protocol of Davenport et al. [14] was employed. In brief, each hair segment was put into a 15 ml Falcon tube, then 2.5 ml isopropanol was added, and the tube gently mixed on an overhead rotator for three minutes. After decanting, the wash cycle was repeated two times. Then the hair samples were allowed to dry for at least 12 hours. Next, the hair segments were weighed out and 7.5 mg were transferred into a 2 ml glass vial. 1.5 ml of pure methanol was added and the steroid extraction was performed for 18 hours. Samples were then spun in a microcentrifuge at 10.000 rpm for 2 min, and 1 ml of the clear supernatant was transferred into a new 2 ml glass vial. The alcohol was evaporated at 60 degrees Celsius under a constant stream of nitrogen until the samples were completely dried. Finally, 0.4 ml of water was added and the tube vortexed for 15 sec. Twenty microliters were removed from the vial and used for cortisol determination with a commercially available immunoassay with chemiluminescence detection (CLIA, IBL-Hamburg, Germany). The intraassay and interassay coefficient of variance of this assay is below 8%.
Psychosocial questionnaire
All study participants were asked to complete a series of ten validated psychosocial scales via a questionnaire regarding stress and childhood trauma. The scales include the Brief Cope Scale [15], Childhood Trauma Questionnaire – Short Form (CTQ) [16], Multiethnic Identity Measure [17], Perceived Stress Scale (PSS) [18, 19], Patient Health Questionnaire Depression scale [20], Peri Life Events Scale [21], Silencing the Self Scale [22], the Social Provisions Scale [23], and Neighborhood Problems Scale [24]. All scales are validated indicators of emotional functioning. All scales yielded a numeric score; however, the subscales of the CTQ (Emotional Abuse, Physical Abuse, Sexual Abuse, Emotional Neglect, and Physical Neglect) were categorized such that a positive response on any item in the subscale was categorized into a “positive history” as compared to a “negative history” of the specific form of childhood trauma.
Statistical methods
The Wilcoxon signed-rank test was used to compare the distribution of hair cortisol scores between cases and their matched controls (i.e., subjects were treated as case-control dyads). Separate analyses were conducted for proximal & distal samples. Conditional logistic regression was used to predict case vs. control (conditioned on belonging to a case-control-matched pair), from hair cortisol score, which was treated as a linear predictor. Variables used to match each case with its control (fetal gestational age [GA] at hair cortisol measurement, patient age, ethnicity) were introduced as covariates, as were any prior history of premature delivery, and hair sampling locus (scalp-distal vs. -proximal). Sample size did not permit use of all of the psychometric measures in regression models; the PSS total score and the 5 dichotomized sub-measures of the CTQ were selected as mostly likely to be salient. Odds ratios (ORs) reported are based on a 10 pg mg–1 increase in hair cortisol level; 95% confidence intervals (CIs) are also reported. Area under the receiver operating characteristic curve (AROC) was used as a general measure of overall diagnostic utility of the logistic regression model; this is a metric that ranges from 0.5 (zero utility) to 1.0 (perfect predictability). Loess plots and the Akaike information criterion were used to choose between raw hair cortisol score and log-transformed hair cortisol score as the better predictor of case v. control. A test of interaction between hair cortisol score and sampling locus was conducted, to assess the validity of pooling proximal & distal scores in a single analysis. The Hosmer-Lemeshow goodness-of-fit test was applied, to assess the validity of using hair cortisol score, GA, PSS and age as linear predictors.
Results
29 case-control pairs were analyzed. Six subjects, 3 cases and 3 controls, having only proximal measurements were included in analysis. Since subjects were matched on age, gestational age, and race, those characteristics did not vary between groups (see Table 1). Nonparametric tests showed no significant difference between cases & controls in distribution of hair cortisol scores from either proximal (p = 0.099) or distal (p = 0.175) samples (Fig. 1). Hair cortisol raw scores rather than log-transformed scores were selected for the parametric analysis. A test of interaction of hair cortisol score with sampling locus was not significant (p = 0.427), so the final analysis pooled proximal with distal sample scores. Adjusting for sampling locus, age, GA, ethnicity, PSS, CTQ and history of premature delivery, hair cortisol level was a significant predictor of case v. control (p = 0.008); each 10-pg.mg–1 increase in hair cortisol level was associated with an estimated 33% decreased odds of being a case (95% confidence interval 10–50%). AROC was 0.85 for a model with all predictors included, compared with 0.80 with covariates only; so while hair cortisol was a statistically significant predictor, its predictive utility was modest. A history of emotional neglect increased the chances of PTB (63.4% vs. 31.6% p = .021). All other parameters of childhood trauma and perceived stress did not differ significantly between groups.
Discussion
Our pilot study suggests that women who deliver prematurely may have lower hair cortisol levels than women who deliver at term. Hypocortisolism, as a factor in the pathophysiology of several disease states, has been reported in patients with stress-related disorders such as chronic fatigue syndrome, chronic pelvic pain, fibromyalgia, posttraumatic stress disorder, irritable bowel syndrome, burnout, and atypical depression [3], and most recently has been found in victims of childhood bullying [7]. However, until now, this phenomenon has not been reported in women delivering preterm. Many studies have correlated perceived stress with preterm birth [25, 26]. Cortisol, utilized as a biomarker for stress, has also been correlated with preterm birth. Campbell et al. examined random serum cortisol in 218 women between 22–36 weeks gestation with threatened preterm birth and demonstrated that higher cortisol levels were predictive of preterm birth in less than 48 hours for the groups between 22–27 w and 32–36 weeks [9]. Erickson et al. tested random serum cortisol levels on 59 women with a preterm birth and 300 women who delivered at term (gestational aged-matched) and showed that women with a preterm birth had higher cortisol levels [10]. Field et al. looked at urine cortisol at 20 weeks gestation in 300 women with depression, and found a positive correlation with cortisol and preterm birth [11]. Mercer et al. measured random serum cortisol in 3 groups of women, 46 women who had a recurrent spontaneous preterm birth, 92 women with an isolated spontaneous preterm birth, and 92 women with recurrent term birth, and found that women with recurrent spontaneous preterm births had higher cortisol levels compared to women with recurrent term birth and isolated spontaneous preterm birth [8]. However, all of these studies are subject to the limitation that spot cortisol measurements fluctuate with normal circadian rhythm and with acute stress. Thus preterm labor itself, or its prodrome, may be the cause of, rather than the consequence of, elevated spot cortisol. Hair cortisol measurements, which reflect cortisol levels over a several month time frame, can overcome that limitation. However, there is limited published information about hair cortisol in pregnancy, and even less about hair cortisol and its relationship with preterm birth. Kalra et al. looked at the relationship between stress and hair cortisol in healthy pregnant women and found that maternal hair cortisol levels correlated positively with measures of perceived stress [27]. One of the authors of this paper previously showed that hair segments can provide a retrospective calendar of cortisol production for an individual, and that there is a physiological increase in hair cortisol incorporation in the third trimester (i.e. higher hair cortisol measurements) [28]. This is an important fact that makes it crucial to control for gestational age when measuring hair cortisol in pregnancy. Kramer et al. found that concentrations of hair cortisol were higher in the hair of women who delivered near term than those who delivered at less than 34 weeks of gestation [25]. However, they did not control for gestational age, i.e., all women had samples obtained around the time of birth so that preterm women’s levels might differ either because of a disease state or because of the different gestational age at which they were obtained. In a prior study one of our authors (C.K) found that premature birth was negatively associated with hair cortisol measurements [29]. Though this study also did not control for gestational age, there was a negative correlation with preterm birth, a finding that our study confirms. To date, there has been no other study that examined hair cortisol as a predictor of preterm birth and that controlled for gestational age at the time of hair collection. Although our study was a pilot study with a small sample size, by obtaining samples from both the preterm and term group at the same gestational age we were able to show significantly decreased hair cortisol in women who delivered preterm when compared to their controls who delivered at full term. The mechanism for the development of hypocortisolism, and how this low cortisol state contributes to pathology is not clear. Because we assessed psychosocial functioning we were able to demonstrate that the differences were not due to lower stress in the preterm birth cohort. In fact the only significant difference was a significantly increased measure of emotional neglect in the preterm group. Some theories suggest that a blunted chronic cortisol level, such as was seen in our preterm group, may be a response to long-standing chronic stress that results in dysfunction of the normal HPA axis. The HPA axis, and cortisol specifically, has an effect on the immune system. In particular, glucocorticoids suppress the production of cytokines and therefore restrain the inflammatory reaction during stress. It’s possible that alterations in the HPA axis function effects the glucocorticoid role in the immune response in pregnancy, which is physiologically an immunosuppressive state. We need to note a few limitations and strengths. Since this was a pilot study to see if there was a sufficient difference in levels to warrant further work, we did not have adequate sample size to look at relationships between particular psychosocial makers of stress and hair cortisol levels. However, the sample size was adequate to demonstrate significant differences between cases and controls in factors linked to preterm birth. The fact that prior preterm birth was a risk-factor was reassuring since it comports with an extensive literature. The finding of a link between low cortisol and preterm birth in the same cohort we found of sufficient interest to warrant further efforts in this field. The strengths of the work include the very careful match on demographics, the sampling at the same gestational age, and the use of a biomarker that reflects chronic, rather than acute events and the blinding of lab personnel to clinical events. In sum, our pilot study is the first to demonstrate lower hair cortisol levels in women who deliver preterm compared to matched women who deliver at term. More work is needed to examine the role of HPA axis dysregulation, its link with chronic hypocortisolism, and its association with preterm birth.
Funding source
Maimonides Research Foundation.
Disclosure statements
None of the authors have any potential or actual financial interests relevant to the topics discussed.
Footnotes
Acknowledgments
Financial Support for this research: Maimonides Research Foundation; Fund #49792. The funding source had no involvement in the study design, collection, analysis or interpretation of data, writing the report, or the decision to submit the manuscript for publication.
Sharon Kim: Research Assistant, Department of Obstetrics and Gynecology, Maimonides Medical Center.