Amniotic immune biomarkers as risk factors in women with different symptoms of threatened late miscarriage

To investigate risk factors that can help identify the possibility of pregnancy loss in threatened late miscarriage (TLM) patients with and without spontaneous uterine contractions.


| INTRODUC TI ON
Late miscarriage, also called second-trimester or mid-trimester loss, usually refers to a miscarriage that happens when a baby dies after 14 weeks and before 22 or 24 weeks of pregnancy (depending on the country). [1][2][3] However, the highest gestational age of late miscarriage is defined as 28 weeks in China, and so "Threatened late miscarriage (TLM)" is defined by the shortening or opening of the cervix with or without uterine contractions at or after 14-week gestation and before 28 weeks. Although there is no consensus on the definition and diagnosis of "cervical insufficiency (CI)" in published studies, 4-6 "TLM without uterine contractions" is commonly considered as CI. So the CI patients in this study are TLM patients without uterine contractions. If not treated properly, TLM may progressively lead to late miscarriage or preterm birth, presenting a very difficult clinical situation for both patients and doctors. A better understanding of risk factors for adverse pregnancy outcomes in these two groups of patients can contribute to the understanding of the mechanisms leading to miscarriage and ultimately the identification and management of high-risk pregnancies.
It is widely accepted that the act of parturition is the final step of a proinflammatory cascade that is coordinated by an intrauterine environment connected with hormonal signals 7 and maternal peripheral or amniotic immune biomarkers have for several decades been the focus of many published papers investigating various adverse pregnancy outcomes. As early as 2008, a study of the association of IL-6, IL-8, and TNF-alpha with PROM 8 found that only one of IL-6 or IL-8 was necessary. An investigation of only IL-6 in an African American population in 2011 9 found this cytokine to be strongly associated with preterm birth, and more recently, a study of three cytokines (IL-6, IL-2R, TNF-alpha) found moderate associations with preterm birth. 10 Since then, the cytokines IL-6 and TNF-alpha have been found to be strongly associated with fetal inflammatory response 11 and to have significantly higher levels in women with preterm delivery. 12 To our knowledge, there are no studies examining the joint effect of all common cytokines for their independent contribution to risk. The presence of sludge in amniotic fluid has also been shown to be associated with the risk of preterm delivery, 13,14 but these studies did not adjust for the potential contribution of amniotic cytokines.
Given the accumulation of evidence, we hypothesized that some of the amniotic biomarkers that are routinely available from intrauterine inflammation screening are independently associated with TLM, with possibly different effects for patients with and without uterine contractions. Separate analysis of these groups is rare in the scientific literature, and furthermore, most studies present results from univariate analyses and many do not explicitly state if they include/exclude pregnancies with intrauterine infection (itself a risk factor for late miscarriage). We restricted our study population to TLM patients without intrauterine infection. The aim of this study is to compare the amniotic fluid in TLM pregnancies with and without uterine contractions for the levels of inflammatory markers and presence of sludge, and to investigate the effect of the biomarkers on risk of late miscarriage in the two groups.

| Enrollment of study cohort
The study participants were identified according to the Chinese Guideline for Diagnosis and Therapy of Preterm Birth 15 from women who attended the obstetrics and gynecology department of Sun Yat-Sen memorial hospital from January 1, 2016, to December 30, 2017. Women determined to have a high risk of late miscarriage were enrolled in the study and classified by their clinical symptoms into two groups based on whether or not they had uterine contractions ( Figure 1). These high-risk pregnant women were enrolled between 16 +0 and 27 +6 gestational weeks if they met all of the inclusion criteria and none of the exclusion criteria, resulting in 76 pregnant patients in the no-contractions group and 33 patients in the contractions group. The inclusion criterion for the no-contractions group was an intravaginal cervical length < 15 mm or cervical os open (<4 cm) without uterine contractions. The inclusion criterion for the contractions group was a normal cervix (cervical length ≥ 25 mm) with regular contractions F I G U R E 1 Flowchart of the cohort study at first visit to the emergency department, but shortening to 0-20 mm during the observation period. Women under observation in the emergency department are monitored at least every 4 hours and up to a maximum of 24-48 hours, according to the patient's condition. Women were excluded if the cervical length was ≥ 20 mm after observation in the emergency department or the cervical os opened during the observation period. Women were excluded from the study if they had intrauterine infections. In addition, the following six exclusion criteria were used: serious heart, liver, or kidney disorders; fetal death; placenta previa with uncontrolled vaginal bleeding; placenta accrete; preterm labor; acute genital tract inflammation, especially chorioamnionitis: maternal temperature of ≥38.0°C and two or more of the following: (a) uterine tenderness; (b) malodorous vaginal discharge; (c) maternal leukocytosis (WBC ≥ 15 000 cells/mm 3 , or neutrophils > 90%); (d) maternal tachycardia (>100 beats/min); (e) fetal tachycardia (≥160 beats/min). All patients signed informed consent to have an amniocentesis performed. For multiple gestation, the specimen of amniotic fluid was obtained from the lower amniotic sac (ie, closest to the cervix).

| Amniotic immune biomarker measurements and "sludge" assessment
All patients were screened for intrauterine inflammation using a standard set of inflammation biomarkers that are widely used in clinical practice and research. A 20-milliliter specimen of amniotic fluid was transported to the pediatric laboratory within two hours after amniocentesis and analyzed by chemiluminescence using the Siemens Immulite 1000 equipment, which provided levels of IL2βR, IL6, IL8, IL10, IL1β, and TNFα; The amniotic WBC was tested by Bayer AD, VIA2120 automatic blood cell instrument; The presence of amniotic microorganisms (bacteria, mycoplasma, and candida albicans) was assessed by culture. Meanwhile, a 4-milliliter sample of maternal peripheral blood was sent to the general laboratory for testing a standard set of markers (IL2βR, IL6, IL8, IL10, and IL1β) on the same day, using the same method and the same type of analyzer. For imaging, we used transvaginal ultrasound (Voluson E8; GE Healthcare, probe: RIC6, 12, D, 22Hz). Observations of cervical dynamic changes were performed for several minutes. The cervical length measurements were repeated after pressure on the pubic symphysis or cough, and the shortest length recorded after three repeated measurements. "Sludge" was defined as the presence of dense aggregates of particulate matter in the proximity of the internal cervical os. Diagnosis of "sludge" was checked by two ultrasound specialists.

| Treatment and follow-up of patients
A flowchart of the screening procedures is presented in Figure S1. Cervical length and "sludge" were assessed every 2-4 weeks until 34-weeks gestation. Once "sludge" was definitively diagnosed, the assessment interval was decreased to 1-2 weeks. If cervical length was not continuously shortened for two consecutive assessments, the screening interval was prolonged to the original 2-4 weeks. For the "no-contractions" patients, stressed McDonald cerclage was performed if the cervix was short, and emergency McDonald cerclage was performed if the cervical os was open (<4 cm). The cerclage was performed on such patients at no more than 27 +6 weeks (the highest gestational age in this group of our cohort was 27 +4 weeks). For patients with uterine contractions, tocolytic therapy was given after amniocentesis. Phloroglucin was prescribed for patients with a gestational age <20 weeks, and the highest dose was 200 mg/d. The first-line tocolytic medicine after 20 gestational weeks was ritodrine hydrochloride. Ritodrine hydrochloride was considered suitable for women with a singleton pregnancy without vaginal bleeding or other contraindications (including overactive thyroid gland, extreme loss of body water, high blood pressure, increased pressure of pulmonary circulation, abnormal heart rhythm, or diabetes). Atosiban was commonly considered for patients whose gestational age ≥24 weeks, especially with multiple gestation, diabetes or gestational diabetes, cardiovascular disease, or hyperthyroidism. Atosiban was also offered to patients at less than 24-week gestation who had contraindications of ritodrine. Tocolytic therapy of no more than seven continuous days was continued until no regular uterine contractions were detected by external tocography (AN24™: maternal-fetal holter, Beijing) or abdominal palpation. If the first-line medicine failed to control the contractions, nifedipine or indomethacin was used.
Tocolytic agents were not given after 34 +6 gestational weeks.
Prophylactic tocolytic agents were also given after cerclage for women without contractions for up to 3 days.

| Statistical analysis
Biomarkers were inspected for their pairwise correlation using a scatter plot matrix, and the Spearman rank correlation coefficient was used to assess the degree of associations between gestational age (days) and the biomarker levels. Box plots and violin plots were

| Amniotic IL2βR, IL8, and TNFα levels were higher in the group with contractions
Reproductive characteristics of patients with/and without uterine contractions were compared and no significant differences were found (    Table 3 (with details of immune biomarker levels in Table S1). In the contractions group, late miscarriage was significantly associated with maternal age (P = .029) and primigravida (P = .025) and there was also a borderline association with multiple gestation (P = .095), but none of these factors were significantly associated with late miscarriage in the no-contractions group. All of the amniotic immune biomarkers had significant crude associations with late miscarriage in the no-contractions group and four (IL10, IL8, IL1β, and TNFα) in the contractions group, where there was also a trend for an association with IL6 (P = .091). "Sludge" was found to be borderline significant in the no-contractions group (P = .057). Amniotic immune biomarker levels were also compared for patients with and without "sludge" stratified by uterine contractions (Table S2), with no significant differences found in the no-contractions group, but higher IL10, IL8, and IL1β in the contractions group. Comparisons using the log scale identified the same significant differences (data not shown). only IL8 was found to be an independent risk factor for late miscarriage (

| D ISCUSS I ON
In this study, we found differences in TLM patients with and without contractions not only for their amniotic immune biomarker levels and "sludge," but also for the impact of these biomarkers on the pregnancy outcomes. The contractions group had more late miscarriages and PPROM, and higher levels of amniotic IL2βR, IL8, and TNFα than the no-contractions group. We also found that amniotic

F I G U R E 3
Violin plots of amniotic immune biomarker levels in patients with and without uterine contractions, stratified by pregnancy outcomes. *Indicates log transform a biomarker of delivery onset. 21 Other immune biomarkers in our study were not found to have independent effects on late miscarriage. Although some articles have found independent effects on miscarriage/preterm birth of some biomarkers, 22,23 none of these studies mutually adjusted for all the same biomarkers that we investigated in our study: we found very high correlations between some of the biomarkers so that it is unsurprising that there were so few with independent contributions. This was also observed by Velemínský who found that only one of IL-6 or IL-8 was necessary to consider for estimating the risk of PROM. 8 A novel aspect of our work is the investigation of immune biomarker levels and "sludge" in patients stratified by different symptoms. Other authors have studied amniotic immune biomarkers as risk factors in cervical insufficiency (CI) patients, [23][24][25][26] who are comparable to the "no-contractions" patients in our study: one retrospective study 23 found IL1β, IL6, IL7, IL17α, TNFα (but not IL8) and cervical dilation to be independently associated with very early preterm birth (≤32 weeks of gestation). Another strength of our study is that, in contrast to many previous studies, we did not choose a subset of the cytokines, but examined all those available from a comprehensive laboratory package in order to determine their independent effect.
"Sludge" has also been reported by others as an independent risk factor for preterm birth in "short cervix" patients. [27][28][29] In our study, we conducted a joint analysis of cytokines and "sludge" and failed to see an independent effect of "sludge" on our primary outcome (late miscarriage) when adjusted for cytokine levels, possibly due to low statistical power. However, there was a clear pattern of earlier "pregnancy loss" in women with "sludge" in the no-contractions group ( Figure S2). Interestingly, three patients in the no-contractions group whose "sludge" disappeared during follow-up all had good pregnancy outcomes, delivering live-born infants at 35 +6 weeks, 37 +1 weeks, and 38 +3 weeks.
For the patients with contractions, we failed to build a reasonable model to analyze the independent contributions of the risk factors, partly due to the extreme distribution of the amniotic immune biomarker levels but also the small sample size. Our limited model found TNFα to be associated with pregnancy loss, but a larger study is needed to verify if IL8 is an independent risk factor for these patients. We also failed to find a significant effect of "sludge" on late miscarriage in this group, although other researchers have identified it as an independent risk factor 13,14 for preterm birth and one study reported a much higher prevalence of "sludge" in women who delivered preterm. 30 Considering we had only 6 patients with "sludge" in the contractions group, larger prospective studies are needed.
Few studies have focused on the association between amniotic immune biomarkers and "sludge." One study 31 reported that the presence of "sludge" was related to intra-amniotic inflammation with or without microorganisms. In our study, we found that amniotic immune biomarker levels were not different between patients with and without "sludge" in the no-contractions group, but IL10, IL8, and IL1β were different in the contractions group, suggesting potentially different mechanisms leading to their different pregnancy outcomes, including late miscarriage. Furthermore, the association of sludge with some amniotic immune biomarkers in the contractions group can partly explain why it was not found to be an independent risk factor.
The strengths of this study are that we analyzed two different clinical phenotypes of TLM and investigated the value of immune biomarkers in the absence of intrauterine infection. Although many immune biomarkers have been reported as risk factors for late miscarriage and preterm birth, there is still no consensus and few studies have focused on TLM patients with contractions. 12,32 Since amniocentesis was conducted prior to tocolytic therapy or cerclage to rule out intrauterine infections, we could exclude pregnancies with infection and thus add to the evidence from previous reports of worse perinatal outcomes for pregnancies with abnormal amniotic fluid markers indicative of inflammation or subclinical intra-amniotic-infection. 33 By conducting the analysis in each group separately, we ensured that patients were similar with respect to treatment: all patients in the no-contractions group had cerclage, and no patient in the contractions groups had this procedure. In addition, the exclusion of patients with intrauterine infections reduces the potential bias from differential treatment.
Despite the similarities between patients in the two groups, a potential limitation of the study is that different tocolytic drugs may be used. However, amniotic immune biomarkers are not part of this decision, so the choice of drug is unlikely to be a mediator of the biomarker's effect on pregnancy outcome. Factors associated with the tocolytic medicine include gestational age and multiple gestation, both of which have been adjusted in our analysis.
Other factors that impact on the choice of drug are cardiovascular diseases, diabetes, thyroid diseases, and other medical conditions: TA B L E 4 Risk factors related to late miscarriage in patients with and without uterine contractions, from multiple logistic regression and Cox regression analysis we did not have access to this information for our cohort, but since these conditions are rare in our study population, they are unlikely to have confounded the effect of the biomarkers. There remains the possibility that during the follow-up period, different patients may have required different treatment strategies which could mediate the effect of the biomarkers on the pregnancy outcome.
Thus, our estimates are of the total effects of amniotic immune biomarkers, and a more in-depth clinical study would be needed to answer the question regarding direct effects. Furthermore, since this is a relatively small study in a single hospital, further work is needed to validate the role of the amniotic immune biomarkers, and instead of simple categorization of high and low levels based on cohort-specific medians and quartiles, to identify clinical cutoffs that might be "transported" to other populations.
In summary, our work demonstrates that high levels of amniotic immune biomarkers present different characteristics of TLM pregnancies with and without contractions, in the absence of intrauterine infections. These findings can improve our understanding of the underlying biological mechanisms of miscarriage and help in the identification of at-risk pregnancies.

ACK N OWLED G M ENTS
The authors wish to thank Chen Wang and Yingying Yang for their assistance with analyzing the data, and Xianghua Lin for giving the researchers access to the laboratory records in this study.

CO N FLI C T O F I NTE R E S T
We have no conflicts of interest to disclose.

AUTH O R CO NTR I B UTI O N S
Hui Chen and Lili Meng conceived and designed the study. Lili Each author has confirmed compliance with the journal's requirements for authorship.