Can stress biomarkers predict preterm birth in women with threatened preterm labor?

Abstrak

Background: Preterm birth is a major paediatric challenge difficult to prevent and with major

adverse outcomes. Prenatal stress plays an important role on preterm birth; however, there are few

stress-related models to predict preterm birth in women with Threatened Preterm Labor

(TPL).
 Objective: The aim of this work is to study the influence of stress biomarkers on time

until birth in TPL women. Methods: Eligible participants were pregnant women between 24 and

31 gestational weeks admitted to the hospital with TPL diagnosis (n = 166). Stress-related

biomarkers (α-amylase and cortisol) were determined in saliva samples after TPL diagnosis.

Participants were followed-up until labor. A parametric survival model was constructed based on

α-amylase, cortisol), TPL gestational week, age, parity, and multiple pregnancy. The model was

adjusted using a logistic distribution and it was implemented as a nomogram to predict the labor

probability at 7- and 14-day term.

Results: The time until labor was associated with cortisol (p = 0.001), gestational week at TPL

diagnosis (p = 0.004), and age (p = 0.02). Importantly, high cortisol levels at TPL diagnosis were

predictive of latency to labor. Validation of the model yielded an optimum corrected AUC value

of 0.63.
 Conclusions: High cortisol levels at TPL diagnosis may have an important role in the

preterm birth prediction. Our statistical model implemented as a nomogram provided accurate

predictions of individual prognosis of pregnant women.

1. Intraduction
Preterm birth is a global health problem and a major paediatric challenge. The incidence has been

growing and at present, preterm birth represent in developed countries up to 12.5% of all deliveries,

and it is associated with a significant increase in mortality and morbidity (Blencowe et al., 2012).

Reliable tools to predict preterm birth in women with threatened preterm labor (TPL) are required

in order to prevent neonatal complications. TPL is the onset of possible labor between 24 and 37

weeks of pregnancy. TPL women show regular and painful uterine contractions and their cervix

starts to open (Romero et al., 2006). One-third of hospitalized pregnant women are TPL

(Gazmararian et al., 2002). Tocolytic drugs and glucocorticoids are administered to stop uterine

contractions and to induce maturation of preterm fetal lungs, respectively (Honest et al., 2009).

However, more han 50% of TPL women do not progress to active labor (Gazmararian et al., 2002).

The lack of prediction of preterm birth in TPL women is of concern due to the side effects of

tocolytic drugs in cases with low risk of preterm birth, the unnecessary hospitalization, and the

elevated cost to the public health system (Lucovnik et al., 2013). Therefore, reliable methods to

identify TPL women with high risk of preterm birth are required.
Fig. 1. Flow diagram describing the recruitment process, the exclusion determinants and the

patients who completed the study.

Traditionally, preterm birth risk factors included adverse obstetric conditions, such as,

infections, previous preterm birth, short cervical length, as well as multiple pregnancy, in-vitro

fertilization, obesity, and high parity (Goldenberg et al., 2008). Nevertheless, the same adverse

obstetric conditions that may lead to preterm birth are present in a significant number of term

births. Among the new factors with influence over TPL, the role of maternal stress has become

increasingly relevant (Goldenberg et al., 2008), especially in healthy women with-out obstetric or

medical adverse conditions. Studies addressing mater- nal stress have employed self-reported

questionnaires on anxiety and depression as well as stress biomarkers (García-Blanco et al., 2017).

Most research has focused on the relationship between self-reported anxiety and depression along

pregnancy and preterm birth outcome but findings remain unclear. In fact, some studies found that

women who had a preterm birth showed higher levels of anxiety and depressive symptomatology

during pregnancy (Field et al., 2008; Liu et al., 2016; Lilliecreutz et al., 2016), and other studies

reported no differences (Ding et al., 2014; Pearce et al., 2010).

Several confounding factors such as the time of prenatal testing as well as the subjective perception

of these symptoms may explain these inconsistences (Kramer et al., 2013). As regards the time of

prenatal testing, identifying levels of stress biomarkers after TPL diagnosis would be useful for

making a clinical decision depending on the risk of preterm birth. However, many women admitted

to hospital with TPL have the subjective perception of feeling very anxious and with low mood.
In this sense, the objective measure of stress biomarkers after TPL diagnoses can help to

understand the relationship between maternal stress and preterm birth outcomes (Mercer et al.,

2006).

Previous studies on TPL women have focused on stress biomarkers secreted by the hypothalamic-

pituitary-adrenal (HPA) axis in TPL women (Campbell et al., 2005; Mercer et al., 2006). The HPA

axis under stress conditions is regulated by corticotrophin releasing hor- mone (CRH) and cortisol

(Cook 2002; Kramer et al., 2013; Miller et al., 2013). Previous works carried out with TPL women

found that high cortisol levels were a reliable predictor of preterm birth at 48 h after TPL diagnosis

(Campbell et al., 2005).

In addition to HPA axis, mechanisms based on sympathetic adrenal medullary (SAM) axis may be

involved in maternal stress (Glover, 2015). SAM axis shows lower habituation to stress than HPA

axis, and it is sensitive to maintained stress conditions (Allen et al., 2014; Schumachera et al.,

2013). Actually, adrenergic dysregulation reflected by high levels of α-amylase have been

observed in depressive pregnant women (Braithwaite et al., 2015), although this finding has not

yet been replicated. Thus, α-amylase could be a biomarker of maternal stress and its role as

predictor of preterm birth in TPL women is yet unknown. Therefore, it would be interesting to

determine both cortisol and α-amylase levels as potential biomarkers of preterm birth (Lilliecreutz

et al., 2016).

To our knowledge, this is the first follow-up study carried out with healthy women with TPL until

labor. The aim of this work was to study stress biomarkers (cortisol, α-amylase) as potential

predictors of preterm birth at time of TPL diagnosis in women without additional adverse obstetric
and medical conditions. We hypothesized that, higher stress biomarkers levels, especially cortisol,

would indicate a short time until labor in TPL women. Moreover, subjective maternal stress (i.e.,

self-reported anxious-depressive symptoms) would not reflect objective maternal stress (i.e.,

biomarkers levels), since many TPL women would feel very anxious and with low mood .

2. Materials and methods

2.1. Study design and participants

This is a prospective cohort study performed in the Division of Obstetrics of the University and

Polytechnic Hospital La Fe (Valencia; Spain) during a 12-month period in 2015. The ethics

committee of the hospital approved the study and informed consent was signed by all participants.

Eligible participants were healthy women between 24 and 31 weeks gestation admitted to the

hospital with the diagnosis of TPL. All eligible patients were subjected to the same tocolytic

treatment. Hence, TPL was diagnosed if regular uterine contractions associated to cervical changes

(≥80% cervical effacement or ≥2 cm cervical dilata- tion) were present. Currently, the cervical

length is measured by the cervical ultrasound (< 25 mm). The follow-up was carried out until labor.

Exclusion criteria included severe medical conditions (e.g.: diabetes mellitus), severe obstetric

complications during pregnancy (abruptio placenta, preeclampsia, intrauterine growth restriction,

cervi- cal dilatation > 4 cm, clinical signs of infection, obstructed labor), fetal anomalies, use of

teratogenic substances, body mass index < 18.5 or > 30, in-vitro fertilization, and risk of social

exclusion, which is considered a stressful condition that may act as a confusing variable. To

evaluate the risk of social exclusion, we used the multidimensional criteria defined by the Europe

2020 strategy (European Council, 2010): (i) at-risk-of poverty, (ii) severe material deprivation,

and (iii) jobless household (Savova, 2005). Fig. 1 shows the flow diagram of recruit- ment.
After TPL diagnosis, an abdominal ultrasound to check fetal cardiac activity was carried out. Then,

fetal cardiac activity and uterine contractions were monitored and blood and urine samples were

collected. If contractions ceased home resting was recommended. Contrarily, if contractions

continued, patients were admitted to the Obstetric ward to initiate antenatal steroid to and tocolytic

therapy in all cases (ACOG, 2012). Tocolytic therapy was Atosiban or Nifedipine. Atosiban was

initiated with a bolus of 6.75 mg for 1 min. Then, 300 μg min−1 as loading dose for 3 h and 100

μg −min−1 as main- tenance dose for 48 h were administered. Alternatively, Nifedipine 20 mg,

followed by 10 mg each 20 min until 40 mg for 1 h (Conde- Agudelo et al., 2011; Kashanian et

al., 2005). In cases of false TPL diagnosis (one contraction or less in 10 min and no cervical

change, n = 11) women were sent home. Finally, in cases of imminent labor (cervix between 4 cm

and 10 cm dilated, rate of cervical dilatation at least 1 cm/hour, effacement is usually complete,

and fetal descent through birth canal begins.), magnesium sulfate was administered to reduce the

risk of cerebral palsy. At least 12 h before saliva sample collection, all women received one

corticosteroids dose. Since corticos- teroid average lifetime is 12 h, saliva sample was collected

when antenatal steroid levels in blood have decreased considerably. The second dose of antenatal

steroid was administered after saliva sample collection. Furthermore, Atosiban or Nifedipine was

administered for < 24 h to all participants before saliva sample collection; and none of them

received magnesium sulfate.

Saliva samples from TPL women were collected between 10 and 12 a.m. (minimum 1 h

after breakfast since they should not eat or drink for 30 min before collecting the sample) on the

morning after admission. Saliva samples were collected by means of a passive drool method and
stored at −80 °C until analysis. Furthermore, anxiety and depressive symptoms were assessed by

self-reported questionnaires in order to examine their association with stress biomarkers.

2.2. Analytical determinations

Standard of cortisol was purchased from Sigma-Aldrich Química SA (Madrid, Spain). Saliva

samples were thawed on ice and homogenized. The sample treatment to determine cortisol was

based on a previous work (García-Blanco et al., 2016). Briefly, 25μL of sample were subjected to

liquid–liquid extraction (LLE) to extract cortisol, then the organic layer was evaporated to dryness

and the residues were reconstituted in water (pH 3): methanol (85:15 v/v) solution. Finally, they

were injected in the chromatographic system (ultra-performance liquid chromatography coupled

to tandem mass spectrometry) as previously described (García-Blanco et al., 2016).

Salivary α-amylase assay kit was from Salimetrics (Suffolk, United Kingdom). For the α-amylase

determination, samples were vortexed and centrifuged. Then, they were diluted with the α-amylase

diluent at 1:200 as final dilution. Finally, they were subjected to the kinetic enzyme assay as

described by the manufacturer.

2.3. Psychological assessment

Depressive and anxiety symptoms were determined, using the State form from the State-

Trait Anxiety Inventory (STAI) (Spielberger et al., 1983) and Beck Depression Inventory Short

Form (BDI/SF) (Beck et al., 1974), respectively.

The STAI − State (Spielberger et al., 1983) was used to measure the current anxiety state.

All of its 20 items for assessing current tension, nervousness, and worry were assessed on a 4-
point Likert scale. Scores range from 0 to 60. Higher scores indicate greater anxiety symptoms.

The threshold for clinical anxiety is a total score of 19. Internal consistency coefficient for the

scale was 0.89 in our sample population.

The BDI/SF (Beck et al., 1974) was used to measure the intensity of attitudes and

depressive symptoms and detecting depression in the general population. It contains 13 items,

which are rated on a 4-point scale. Scores ranged from 0 to 39. Higher scores indicate higher

depressive symptoms. The threshold for clinical depression is a total score of 4. In our sample

population we had an internal consistency for the BDI of 0.73.

2.4. Statistical analysis

Statistical analyses were performed using R software (version 3.2.3) and the rms R-

package (version 4.4-1). Overall survival, defined as the time from TPL diagnosis to date of labor,

was defined as the clinical endpoint. The overall survival probabilities were estimated using a

parametric survival model. Based on previous studies, the selected predictor variables were

gestational week at TPL diagnosis, maternal age, parity, cortisol level, α-amylase level and

single/multiple preg- nancy. Cortisol values under the limit of detection were imputed by

simulating from a gamma distribution prior to modelling (Arunajadai and Rauh, 2012). To assess

the uncertainty caused by this imputation method we performed a sensitivity analysis by fitting

and alternative model with the cortisol levels split in three categories (lower than 2.5, between 2.5

and 10 and higher than 10) to check the robustness of our results. The model was adjusted using a

log logistic distribution and it was implemented as a nomogram to predict the probability of labor

at 7- and 14-day term for each women depending on her predictor variables and total points. The

fitted survival model was validated estimating its optimism corrected area under ROC curve
(AUC) value using boostrapping. Finally, the Spearman’s correlation coefficient was used to

express the association between self-reported questionnaires and biomarkers measurements.

3. Results

3.1. Participants

A final sample of 166 healthy TPL women were followed from the TPL diagnosis

(recruitment time) until labor. Fig. 1 shows flow diagram of the recruitment process. The

demographic and clinical characteristics are presented in Table 1. Note that 77% of them have

finally a preterm birth (< 37 gestational weeks), and the mean latency to preterm birth is about 4

weeks. The differences between both groups of participants (preterm and term labor) are indicated

in Table 2.

3.2. Construction and validation of the prediction model

A parametric survival model was constructed including the pre- dictor variables

(gestational week at TPL diagnosis, cortisol level, α- amylase level, maternal age, parity, and

single/multiple pregnancy), and adjusted using a log logistic distribution. Gestational week at TPL

diagnosis was included in the model since it is a confounding variable with influence over other

variables, such as biomarkers levels. The cortisol concentrations lower than the limit of detection

represent approximately 38%. In Table 3, the coefficients and confidence inter- vals (95%) for

each variable from the model are summarized. As we can see, all the variables show a statistically

significant association with survival time, except α-amylase, parity, and multiple pregnancies.
 Particularly important is the finding that high cortisol levels at TPL diagnosis was

predictive of shorter latency to labor. Validation of the model yielded an optimum corrected AUC

value of 0.63. The sensitivity analysis regarding the imputed cortisol values showed similar results

for this predictor: a positive association between higher cortisol levels at TPL diagnosis and shorter

latency to labor (p = 0.005).

From the previous multivariable regression model, we derived a nomogram assigning

points for each predictor and the event was defined as “labor” (Fig. 2). Its use is as follows: first

the patient’s variable is located on the corresponding axis and then a line is drawn to the points

axis. Summing the points and drawing a line from the total score axis to the 7- and 14-day event

free probability axes the final predictive result is obtained. Although stress biomarkers are

important predictor variables of 7 and 14-day event free probability, the other predictors used in

the model add useful information to discriminate patients’ prognosis preterm birth.
3.3. Association between biomarkers levels and self-reported anxious- depressive symptoms

Table 4 shows the correlation between STAI and BDI scores and biomarkers levels in TPL women.

No significant associations were found.

4. Discussion

The main findings of this study reveal that labor probability in healthy TPL women is

influenced by a variety of factors (stress biomarkers levels, gestational week at TPL diagnosis,

age, parity, single/multiple pregnancy). Of note, high levels of stress biomarkers (cortisol, α-

amylase) at the time of TPL diagnosis correspond to higher labor proximity. Therefore, maternal

stress seems to have an important role on preterm birth in healthy women. In addition, although

TPL diagnosis at later gestational age and multiple pregnancies were no significant independent

predictors, they are also variables associated with labor proximity in the developed model. Finally,

the maternal age influence consists of an increase of labor probability for younger women,

probably due to serious medical causes associated to TPL. In the survival model proposed, the

nomogram allows the prediction of a participant-specific probability of labor free at 7 and 14-days

for those women who were diagnosed with TPL.

 Our study addresses for the first time a follow-up study in women from TPL diagnosis to

labor by measuring stress biomarkers. Results have shown that high levels of salivary cortisol were

associated with short latency from TLP to labor. This is in accordance with previous results that

have shown a significant association between preterm birth and stress biomarkers during

pregnancy (Kramer et al., 2013). Of note, stress biomarkers during pregnancy have compared

mothers who finally had a term birth with mothers who finally had a preterm birth. However, there

is still a lack of reliable methods to predict preterm birth when a pregnant woman is diagnosed of

TPL.

Most of previous studies found that the determination of HPA-axis- related stress

biomarkers could be useful for an accurate prediction of preterm birth in TPL women (Campbell

et al., 2005). Nevertheless, the predictive power of stress biomarkers at long-term follow-up is

unknown because in most of these studies follow-up study was carried out just for 48 h (Campbell

et al., 2005). Our study has demonstrated that cortisol levels at the time of TPL diagnosis is a

relevant stress biomarker to predict time until labor. Similarly, in our sample α- amylase levels

improve the predictive power of the developed model. That is, high α-amylase levels are associated

with higher labor

probability; however, this trend has not achieved statistical signifi- cance. Of note, SAM axis has

been scarcely studied as predictor of preterm birth (Lilliecreutz et al., 2016). Indeed, there is some

disagree- ment on α-amylase as stress biomarker. Although previous studies have been associated

α-amylase levels and anxious and depressive symptoms (Braithwaite et al., 2015), others have

found no association (Bosch et al., 1996), or different levels depending on the different stressors

(Bosch et al., 2003; Nater et al., 2005). Importantly, in our sample, perceived anxiety and
depressive symptoms were not associated with stress biomarkers. Of note, many pregnant women

can have the subjective perception of feeling very anxious and low mood at TPL diagnosis.

Therefore, stress biomarkers would quantify better maternal stress in TPL women than self-

reported questionnaires.

The novelty of this study is represented by the fact that we followed healthy TPL women

until labor and that we considered the simulta- neous determination of stress biomarkers from

SAM and HPA axis (α −amylase and cortisol, respectively) together with other relevant variables

(maternal age, gestational week at TPL diagnosis, parity).

In this study, the influence of maternal stress biomarkers over preterm birth in healthy women has

been shown. As a result, a survival model based on several stress biomarkers and demographic

variables (gestational week at TPL diagnosis, maternal age, cortisol, α-amylase, parity, multiple

pregnancy) has been developed, and a nomogram based on this model to predict 7- and 14-day

labor free probability in TPL women was constructed. The nomogram can be used to predict

patients’ prognosis individually and more accurately than only measur- ing stress biomarkers. The

model was derived from a prospective cohort collected during a 12-month period in a regional

referral hospital.

Although nomograms are useful predictive tools employed in several types of studies with

pregnant women (An et al., 2015), the nomogram constructed in this paper is the first to predict

the probability of labor in pregnant women with TPL diagnosis.

 The strength of this study is the successful follow-up of participants from TPL diagnosis

to labor, while its major limitation is the low sample size. However, it can be justified by the

rigorous inclusion criteria, so that participants were restricted to healthy TPL women in order to

evaluate the stress associated with TPL without the influence of other stressful social or medical

variables. In addition, the tocolytic agents and corticosteroids administered to participants could

affect stress biomarkers levels. However, it should be noted that all the participants received

similar Atosiban, and Nifedipine therapies and the same antenatal steroid dose, and samples were

collected after the corticos- teroid average lifetime, when levels in blood have decreased consider-

ably, as well as, magnesium sulfate was not administrated to any participants since they were not

subjects with imminent labor risk. Moreover, an intrinsic limitation of the study design is that

cortisol and α-amylase levels depend on the different gestational age for each participant at saliva

collection. Specifically, cortisol levels increase exponentially as pregnancy goes on, since maternal

cortisol release promotes stimulation of CRH in the placenta achieving between 60 and 700 times

higher cortisol maternal plasma levels than prior to pregnancy (Campbell et al., 1987). Also, the

chaotic mother’s sleep after birth may affect cortisol level since it depends on the circadian rhythm.

5. Conclusions

In this study, an important role of stress biomarkers over preterm birth has been observed. So, a

reliable model to predict the time until labor for healthy women with TPL diagnosis is presented.

This model can be easily applied by means of simple demographic, obstetrical, and analytical

variables as predictors, and a nomogram to determine the 7- and 14-days not labor probability.

Among the predictors, we highlight salivary cortisol levels in TPL women may have an important
influence over preterm birth, and subsequently over adverse perinatal outcomes. Of note, to

conclude that stress biomarkers are strong predictors of preterm birth in TPL women, further

studies should validate the developed model by means of a large external TPL cohort with follow-

up until labor, since it is a crucial prerequisite to be applied to clinical routine use.

Conflict of interests


The authors report no conflict of interests.

Acknowledgments

We are greatly indebted to all the pregnant women, their families, nursing and medical staff who

voluntarily participated in the present study. Without their collaboration and enthusiasm this study

could not have been completed.

MV acknowledges FIS PI14/0443 grant from the Instituto Carlos III (Spanish Ministry of

Economy and Competitiveness) and RETICS funded by the PN 2018-2011 (Spain), ISCIII- Sub-

Directorate General for Research Assessment and Promotion and the European Regional

Development Fund (FEDER), reference RD12/0026; AG-B acknowl- edges a “Rio Hortega” Grant

(CM14/00012) from the Instituto Carlos III (Spanish Ministery of Economy and Competitiveness).

CC-P acknowl- edges a “Miguel Servet I” Grant (CP16/00082) from the Instituto Carlos III

(Spanish Ministry of Economy and Competitiveness).