A Prospective Cohort Study providing Insights for Markers of Adverse Pregnancy Outcome in Women of Advanced Maternal Age


 Background Advanced maternal age (AMA; ≥35 years) is associated with increased rates of adverse pregnancy outcome. Better understanding of underlying pathophysiological processes may improve identification of AMA mothers who are at greatest risk of adverse outcome. This study aimed to investigate changes in oxidative stress and inflammation in AMA women and identify clinical and biochemical predictors of adverse pregnancy outcome in women of AMA.Methods The Manchester Advanced Maternal Age Study (MAMAS) was a multicentre, observational, prospective cohort study of 527 mothers. Participants were divided into three age groups for comparison 20-30 years (n=158), 35-39 years (n=212) and ≥40 years (n=157). Demographic and medical data were collected along with maternal blood samples at 28 and 36 weeks’ gestation. Multivariable analysis was conducted to identify variables associated with adverse outcome, defined as one or more of: small for gestational age (<10th centile), FGR (<5th centile), stillbirth, NICU admission, preterm birth <37 weeks gestation or Apgar score <7 at 5 minutes. Biomarkers of inflammation, oxidative stress and placental dysfunction were quantified in maternal serum. Univariate and multivariable statistical analyses were used to identify associations with composite adverse fetal outcome.Results: Maternal smoking was associated with adverse outcome in older mothers (Adjusted Odds Ratio (AOR) 4.34, 95% Confidence Interval (95%CI) 1.88, 9.99), whereas multiparity reduced the odds (AOR 0.56, 95% CI 0.34, 0.99). In uncomplicated AMA pregnancies, lower circulating anti-inflammatory IL-10, IL-RA and increased antioxidant capacity (TAC) were seen. In AMA with adverse outcome, TAC and oxidative stress markers were increased and levels of maternal circulating placental hormones (hPL, PlGF and sFlt-1) were reduced (p<0.05). Of these, placental growth factor had the strongest predictive accuracy (Area Under the Receiver Operator Characteristic (AUROC) = 0.74) followed by TAC (AUROC=0.69).Conclusions: This study identified alterations in circulating inflammatory and oxidative stress markers in AMA women and in AMA women with adverse pregnancy outcome providing preliminary evidence of mechanistic links. Further, larger studies are required to determine if these markers can be developed into a predictive model of an individual AMA woman’s risk of APO, enabling a reduction in stillbirth rates whilst minimising unnecessary intervention.

There is international recognition that women of AMA should undergo additional antepartum screening or intervention to address the increased risk of stillbirth (14)(15)(16)(17). The RCOG and SOGC recommend offering induction of labour (IOL) at 39 weeks and/or additional monitoring from 38 weeks' gestation (2,15,18).
Although not associated with an increase in the rate of Caesarean Sect. (19), IOL may be viewed as an unnecessary intervention as the majority of mothers will have uncomplicated pregnancies. Furthermore, induction may not be an acceptable intervention for women of AMA without further indication, with poor recruitment (only 13.6%) of eligible women to the 35-39 trial consenting to be randomised (19). Identi cation of mothers with highest risk would result in fewer interventions to prevent stillbirths.
Many pregnancy pathologies are associated with changes in oxidative stress and in ammatory status (20)(21)(22). Similar changes are reported in aging processes although these are usually researched in older populations (23,24). If these alterations were present in AMA they could adversely affect placental function (25)(26)(27). Previous work found evidence of placental dysfunction in AMA pregnancies including (but not limited to) reduced amino acid transport, aberrant cell turnover and reduced placental e ciency (28). Therefore, it was hypothesised that the increased risk of adverse pregnancy outcome results from an aging maternal environment and that a combination of biomarkers of aging, placental dysfunction, and clinical risk factors might identify women at greatest risk. This study aimed to determine whether there were changes in oxidative stress and in ammation in AMA pregnancies and whether changes in these biomarkers were evident in adverse pregnancy outcomes in this population.

Methods
Women were recruited to the Manchester Advanced Maternal Age Study (MAMAS) from March 2012-October 2014 from six UK maternity units after providing written informed consent. Ethical approval was obtained from the NRES Committee North West, (12/NW/0015). Pregnant women aged between 20-30 years (controlsoptimal reproductive age), 35-39 years and ≥ 40 years were approached at 28 weeks' gestation between April 2012-June 2014. Women with multiple pregnancy, body mass index (BMI) < 18.5 or > 30 kg/m 2 , fetal abnormalities, and pre-existing maternal medical conditions were excluded. We estimated a 20% incidence of adverse pregnancy outcome (APO), therefore approximately 600 participants were required to obtain 120 women with APO; 120 women with APO would allow multivariable analysis of six covariates.
In addition to usual antenatal care, participants attended prenatal research appointments at 28 and 36 weeks' gestation (± 1 week), at which detailed demographic, medical data and maternal blood samples for plasma and serum fractionation were collected. After delivery, outcome data were collected from medical records.
Biochemical analyses were conducted after delivery, therefore not altering participants' prenatal care.
The Index of Multiple Deprivation (IMD) -a measure of relative social deprivation (29) -was calculated from the mother's address using NPEU-IMD tool (University of Oxford, UK). A composite adverse pregnancy outcome (APO) was de ned as one or more: small for gestational age (SGA) or FGR (< 10th /<5th centile respectively using individualised birthweight centiles (IBC) (30), stillbirth, admission to the NICU, PTB without infection (< 37 weeks gestation), and 5 minute Apgar score < 7 in the absence of maternal diseases (diabetes/hypertension).
Normal pregnancy outcome was de ned as a term live birth (38-42 weeks), appropriately grown (IBC between 10-95th centile) and absence of maternal or fetal complication (not limited just to those included in our APO de nition).

Nested Case Control Studies
Two nested case-control studies were conducted i) to determine whether AMA associated with elevated circulating biomarkers of in ammation and oxidative stress and ii) to determine whether APO was associated with markers of aging and placental dysfunction in women ≥ 35 years of age. In the rst nested case-control study samples from participants ≥ 40 years were matched (1:1) to mothers aged 20-30 and 35-39 years (n = 40/group) for demographic (BMI, IMD, ethnicity (groups), marital status (married, single, partnership), smoking status (current/ex/non-smoker) and obstetric characteristics (normal vaginal delivery (NVD)). Women with adverse pregnancy outcome, who conceived via assisted reproductive technology (ART) or in whom maternal disease that developed post 28 weeks' gestation were excluded. Sample sizes were determined following power calculations based on previous studies for detecting a difference in circulating oxidative stress (31)(32)(33)(34) and in ammatory mediators (35)(36)(37). For example, to detect a difference in cytokines TNF-α and IL-6 between normal and adverse outcomes in AMA with 80% power with a 5% signi cance level required between 28 and 43 participants in each group.
For the second nested case control study, cases were 43 women ≥ 35 years of age who had an APO as de ned above. Controls were participants with maternal age ≥ 35 years without an APO. Participants that had maternal disease that developed post 28 weeks, PTB associated with infection, or incomplete outcome data were excluded. Groups were matched for demographic characteristics (maternal/paternal age, ethnicity, BMI, smoking status, marital and housing (ownership/rental) status and parity (primi/multiparous. Biomarkers of aging were measured in maternal serum or plasma samples for participants included in the nested case control studies (Supplementary Table 1). Absorbances were measured using a microplate reader (FLUOStarOmega, BMG Labtech) for all Enzyme linked immunosorbent assays (ELISAs). Pro/antiin ammatory cytokines (interleukin (IL)-1α, IL-1β, IL-1Ra, IL-6, IL-10 and tumour necrosis factor (TNF)-α) were quanti ed in maternal serum using DuoSet®ELISAs (R&D Systems, Abingdon, UK) following an optimised protocol (27).
Maternal plasma antioxidant concentration was quanti ed using OxiSelect™ Total Antioxidant Capacity Assay Kit (Cell Biolabs, Inc., San Diego, USA). Oxidative damage markers 8-Isoprostane and DNA/RNA oxidative damage were measured by EIA Kits and Protein Carbonyl Calorimetric Assay Kit (Cayman Chemical Company, Michigan, USA). Maternal serum placental hormones (hCG, PAPP-A, Progesterone and hPL were quanti ed using DRG ELISA kits (DRG Instruments, Marburg, Germany). Placental growth factor (PlGF) and soluble fmslike tyrosine kinase-1 (sFlt) were quanti ed using DuoSet® ELISAs (R&D systems). All assays were conducted according to the manufacturer's standard protocols.

Statistical analysis
Demographic data were compared using Fisher's Exact test for categorical data and Kruskal-Wallis tests with Dunn's multiple comparisons or Mann Whitney U tests for continuous data. Univariate logistic regression was conducted on the whole dataset to identify demographic or clinical variables associated with APO. Multivariable logistic regression was used to quantify the effect of maternal age on APO; in this analysis maternal ethnicity, use of ART, smoking status, parity, IMD and home ownership were included as categorical variables and in an second model, paternal age was included as an additional categorical variable (< 30, 30-34, 35-39 and 40 + years). The logistic regression analyses were conducted using STATA (Version 14, StataCorp, Texas, USA). Biomarker analysis was performed on GraphPad Prism (Version 6.04, GraphPad Software Inc., La Jolla, USA) using Kruskal-Wallis with Dunn's multiple comparisons tests on untransformed data or Mann-Whitney U where appropriate. Due to wide dispersion in the cytokine data, including values at the lower limit of detection of the assay, these data were transformed as log(y + smallest detectable value) for analysis. Gestational age effects were assessed using Spearman's correlation. Markers that showed statistical signi cance at the p < 0.01 level were analysed to test their predictive potential as markers of APO in women aged ≥ 35 years by calculation of the area under the receiver operator characteristic curve (AUROC).

Results
Overall, 1,134 women were approached to participate in MAMAS. 585 of these mothers (51%) either did not meet the inclusion criteria or declined to participate (Fig. 1). 549 mothers consented before 28 weeks gestation.
A further 40 participants were recruited between 28 weeks' gestation and delivery. 62 mothers (11%) were either lost to follow up or withdrew from the study between 28 weeks' gestation and delivery. Therefore, demographic, medical and pregnancy data was collected on a nal cohort of 527 participants (n = 158 20-30, n = 212 35-39 and n = 157 ≥ 40 year olds).
There were no differences in circulating cytokines at 28 weeks' or 36 weeks' gestation between AMA women with normal and adverse pregnancy outcome ( Supplementary Fig. 2). TAC was higher in AMA mothers with APO compared to normal outcomes at both 28 and 36 weeks' gestation (p = 0.002 and 0.006 respectively, Concentrations of hPL, PlGF and PlGF:sFlt ratio were unchanged at 28 weeks' gestation ( Fig. 5A, C, G). hPL was lower at 36 weeks' gestation in women with APO (p = 0.007; Fig. 5B). Similarly, PlGF concentrations were lower at 36 weeks' when measured alone (p < 0.001, Fig. 5D) or adjusted for sFlt-1 (human VEGF R1/Flt-1) concentrations (p = 0.03, Fig. 5H). sFlt-1 was lower in women with APO at 28 weeks' with a similar trend at 36 weeks' gestation (p = 0.05 and 0.07, Fig. 5E,F). No differences were detected in circulating hCG, PAPP-A or progesterone with APO ( Supplementary Fig. 3). ROC curves were created for all biomarkers that reached a statistical signi cance of p < 0.01 between NPO and APO. TAC and 8-isoprostane had predictive area under the curve values of 0.69 and 0.66 respectively (ranked as a poor prognostic markers, Fig. 6A-B), whilst hPL and PlGF had predictive values of 0.68 (poor) and 0.74 (fair), respectively (Fig. 6C-D).

Discussion
This prospective study found that increased maternal age was associated with increased total antioxidant capacity and a reduction in anti-in ammatory IL-10 and IL-RA indicating changes in oxidative stress and in ammation over the timeframe of reproductive life span. In mothers ≥ 35 years of age, adverse pregnancy outcome was associated with a further increase in total antioxidant capacity and a reduction in placentallyderived biomarker, hPL and PlGF. In common with large population studies, adverse outcome (most frequently the birth of a small for gestational age infant) was more common in women ≥ 35 years of age. In this population, maternal smoking and primiparity were independent risk factors for adverse outcome, consistent with retrospective study ndings (7,10,38).

Strengths and Limitations
This prospective study was designed to explore potential mechanisms underpinning the association between AMA and adverse pregnancy outcome, to identify potential biomarkers and generate further hypotheses which can be explored. The multi-centre approach used in this study offered diversity, making it more representative of the UK maternity population. However, this study would have bene tted from larger overall sample size to increase the statistical power to con rm associations between APO and biomarkers and assess the predictive ability of combinations of clinical and biochemical markers. Furthermore, a larger cohort would have enabled larger nested case control studies, with greater statistical power, particularly for cytokine biomarkers which had wide variation in levels between individuals. Despite these limitations, MAMAS is a large prospective study investigating associated factors for adverse pregnancy outcome in AMA and our detailed data collection provided the ability to adjust for multiple confounding variables, and delineate the effects of paternal age, ethnicity and parity and socioeconomic status, all of which may be associated with stillbirth (39)(40)(41)(42).

Interpretation
Understanding the mechanisms underlying the susceptibility to adverse outcomes is essential to improve identi cation of AMA women highest risk of adverse outcomes. This study focussed on processes implicated in maternal aging and placental dysfunction. A pro-in ammatory bias and elevated oxidative stress are established hallmarks of aging (43)(44)(45) and both are strongly associated with pregnancy pathologies, particularly those characterised by placental dysfunction (46)(47)(48). The nested case control studies revealed features of biological aging in women of AMA in the absence of adverse outcome including: elevations in TAC, coincident with reduced oxidative damage (reduced lipid peroxidation) and a reduction in anti-in ammatory cytokines (IL-10 and IL-1Ra) possibly indicating a shift towards a pro-in ammatory state. The former ndings are consistent with these women delivering healthy infants and suggest that adaptive antioxidant responses are effective in maintaining the oxidant:antioxidant balance protecting against oxidative stress (49). The reduction in IL-10 levels has also been seen in serum and placenta of women perceiving reduced fetal movements (27) and in the placenta of infants with FGR (50). As both FGR and RFM are associated with placental dysfunction, the reduction in IL-10 could be consistent with the increased placental dysfunction seen in AMA mothers (28). Critically, an isolated reduction in anti-in ammatory status is not detrimental, but studies of the IL-10 knockout mouse demonstrate increased susceptibility to in ammatory stimuli resulting in PTB and fetal loss (51), and exacerbation of the vascular symptoms of preeclampsia (52) and effects of hypoxia (53).
Therefore, an age-related decline in anti-in ammatory cytokines may increase vulnerability of women of AMA to in ammation and the associated detrimental effects observed on placental function and should therefore be further explored (54). We speculate that maternal aging creates a suboptimal environment for placental and fetal development that contributes to the vulnerability to adverse outcomes.
Consistent with other larger retrospective studies the clinical factors associated with adverse outcome in women of AMA were maternal smoking and primiparity (8,10,38). However, in our study population conception by ART had no signi cant association with stillbirth, although the effect-size of other studies was within the 95% con dence intervals of our population (55,56). These ndings emphasise the need to promote smoking cessation services and suggest that women who have their rst pregnancy over the age of 35 or those who smoke should be prioritised for intervention.
In this population APO was associated with higher circulating levels of 8-isoprostane at 28 weeks' gestation, indicating elevated oxidative damage, despite higher antioxidant capacity. Inadequate antioxidant compensatory responses resulting in oxidative stress has been detected in placentas from APO, where it has been related to altered placental function (22). Future studies are required to con rm whether placental oxidative damage is implicated in the placental dysfunction observed in AMA pregnancies.
Consistent with other reports reduced maternal circulating concentrations of placental hormones (hPL, sFlt and PlGF) were detected in AMA pregnancies with APO compared to normal outcomes. The differences in placental hormones provide further evidence for placental dysfunction as an underpinning mechanism for susceptibility to adverse outcomes in women of AMA. There is a growing body of evidence that these represent biomarkers of placental dysfunction, and are lower in pregnancies with FGR, PE and stillbirth (57)(58)(59). PlGF and sFlt concentrations in the maternal circulation are correlated with fetal size as early as the rst trimester (60) and have strong potential as biomarkers for APO in a clinical setting (59,61). In common with prognostic accuracy studies from other contexts in pregnancy, PlGF had the strongest predictive value for APO in AMA mothers, (62) although the AUROC is insu cient for clinical utility at present. Further studies with larger sample size are required to determine if a prognostic model incorporating the clinical and biochemical predictors with su cient predictive power to identify women of AMA at greatest risk of APO can be derived.

Conclusion
This study has identi ed alterations in circulating biomarkers of in ammation and oxidative stress markers, in pregnant women of AMA, suggesting that biological processes seen in aging may contribute to susceptibility to adverse outcomes in this population. Furthermore, we identi ed serum biomarkers with fair predictive accuracy for adverse pregnancy outcome in AMA. With larger sample sizes and data sets, and by combining identi ed demographic and clinical variables that alter risk of adverse outcome and the measurement of aging and placental biomarkers, it may be possible to create a predictive model with su cient discrimination to delineate an individual woman's risk of adverse pregnancy outcome relating to AMA. This would allow mothers of AMA to be offered more individualised care that considers both maternal and fetal wellbeing and reduces stillbirth rates whilst minimising unnecessary intervention. Availability of data and materials The datasets generated and/or analysed during the current study are not publicly available as ethnical approval was not sought for their dissemination but are available from the corresponding author on reasonable request.