Prerpints.org logo
Preprint
Article

A Novel Predictive Machine Learning Model Integrating Cytokines in Cervical-Vaginal Mucus Increases the Prediction Rate for Preterm Birth

Altmetrics

Downloads

212

Views

41

Comments

0

A peer-reviewed article of this preprint also exists.

Submitted:

24 August 2023

Posted:

29 August 2023

Read the latest preprint version here

Alerts
Abstract
Preterm birth (PB) is a leading cause of perinatal morbidity and mortality. PB prediction is per-formed by measuring cervical length, with a detection rate of around 70%. Although it is known that a cytokine-mediated inflammatory process is involved in the pathophysiology of PB, no screening method implemented in clinical practice includes cytokine levels as a predictor varia-ble. Here, we quantified cytokines in cervical-vaginal mucus of pregnant women (18-23.6 weeks of gestation) with high- or low-risk for PB determined by cervical length, collecting relevant obstetric information. IL-2, IL-6, IFN-γ, IL-4, and IL-10 were significantly higher in the high-risk group, while IL-1ra was lower. Two different models for PB prediction were created using the Random Forest machine-learning algorithm: the full model with 12 clinical variables and cyto-kine values, and the adjusted model, including the most significant variables -maternal age, IL-2, and cervical length- (detection rate 66 vs 87%, false positive rate 12 vs 3.33%, false negative rate 28 vs 6.66%, and area under the curve 0.722 vs 0.875, respectively). The adjusted model integrat-ing cytokines showed a detection rate 8 points higher than the gold standard calculator, which may allow us to identify PB risk more accurately and implement strategies for preventive in-terventions
Keywords: 
Subject: Medicine and Pharmacology  -   Obstetrics and Gynaecology

1. Introduction

Preterm birth (PB) is one of the primary clinical issues in obstetric practice as the leading cause of perinatal morbidity and mortality, and the significant economic burden to the health care systems and families, with an estimated worldwide prevalence within 5–18% and up to 90% of cases occurring in developing countries [1,2]. Although evidence suggests that preterm labor is a heterogeneous condition triggered by multiple factors [3,4], the inflammatory response elicited at the maternal-fetal interface, it is considered a hallmark of the pathology. The increase in the local cytokine release that orchestrate the inflammatory response is observed in the presence of intrauterine infection as well in spontaneous preterm labor that is not associated with infection, thereby described as the intrauterine inflammatory response syndrome [5].
In vivo, cytokines are known to be part of the molecular network that mediate the innate immune and inflammatory responses [6]. All cytokines expressed locally in the cervix participate in complex interactions with prostaglandins and nitric oxide, which regulate the production of extracellular matrix proteases and other factors associated with the cervical shortening, rupture of membranes, and uterine contractions that lead to labor either at preterm or term [5,7,8]. An increase in the pro-inflammatory cytokines IL-6 and IL-8 in the cervical-vaginal fluid has been associated with cervical effacement [9,10], while an increased TNF-α is associated with cervical ripening [11]. On the other hand, IL-10 concentration has been identified as the main anti-inflammatory cytokine involved in the PB pathogenesis [12]. So far, the results of studies analyzing the relationship between cytokines and PB are inconsistent, and none have identified a biomarker that can accurately predict preterm delivery [6]. These inconsistencies can be explained by the complexity of the local inflammatory processes and the heterogeneity of the pathways that trigger PB.
Among the screening techniques to detect women at high-risk for PB, the most used measurement in clinical practice is cervical length (CL), obtained by vaginal ultrasound in the second trimester of gestation, with a detection rate of 50 to 70%. Also, there are some clinical calculators incorporating CL, gestational age, and obstetric history to provide a patient-specific risk for developing PB. The calculator that is considered the gold standard in clinical practice is the Fetal Medicine Foundation (FMF) calculator, having a detection rate of approximately 70% [13].
In recent years, the use of artificial intelligence has been extended in different fields of healthcare to predict, prevent, diagnose, and monitor different pathologies, even in obstetrics [14]. Models using machine learning are more accurate than risk calculators as they analyze massive data and, through algorithms, identify patterns to make predictions [15]; furthermore, it has been proposed that machine learning models can be useful in personalized pregnancy management, especially in low- and middle-income countries [14]. The goal of identifying women at high-risk for developing PB is to individualize the clinical follow-up and offering medical preventive strategies (e.g., progesterone) that are up to 90% effective [16].
Although the role of cytokines as key mediators in the inflammatory process that triggers labor has been demonstrated [17], very few predictive models consider their measurement into the PB screening process [18,19,20], but so far, none have been implemented in clinical practice. Therefore, this study aimed to characterize the cytokine profile in cervical-vaginal mucus in pregnant women with high- and low-risk for PB, and then integrate them in a PB screening model, using a predictive machine learning analysis.

2. Results

A total of 60 pregnant women were recruited, including 40 participants that were considered at low-risk and 20 at high-risk for PB. The low-risk group had a prevalence of preterm birth of 7.5% versus 40% in the high-risk group. The maternal characteristics of both groups are shown in Table 1. The study groups were similar in terms of age, maternal BMI, socioeconomic level, smoking, and gestational age at which CL was measured. As expected, CL was significantly lower in the group of women at high risk for PD (p = 0.02).

2.1. Cytokine profile in low- and high-risk for PB women

The mean concentration of the pro-inflammatory cytokines IL-2, IL-6, and IFN- γ were significantly higher in the high-risk group (p = 0.001). In relation to cytokines with anti-inflammatory function, an increase in the concentration of IL-4 and IL-10 was found in the group of women at high-risk for PB (p = 0.001), while the IL-1ra concentration was significantly lower (p < 0.01) compared to women in the low-risk for PB group. Table 2, Figure 1.

2.2. Machine learning predictive model

Two predictive models were generated by machine learning analysis, the full model, which included all predictor variables, showed a detection rate of 66%, with FPR = 12%, FNR = 28%, and an AUC of 0.722. The adjusted model that included only the most statistically significant variables (Maternal age, CL, and IL-2 concentration) showed a detection rate of 87%, with FPR = 3.33%, FNR = 6.66%, and an AUC of 0.875, to compare our model with the performance of the calculator used by the FMF, we entered our data in the FMF's online calculator (https://fetalmedicine.org/research/assess/preterm/cervix) and obtained the following results a detection rate of 79%, with FPR = 6.60%, FNR = 11.66% (Table 3, Figure 2). The relevance of each of the predictors in the adjusted model was 26% for maternal age, 38% for CL, and 36% for IL-2 concentration (Figure 2).

3. Discussion

In this study, we firstly focused on the characterization of inflammatory cytokine profiles in cervical-vaginal mucus in pregnant women with high- and low-risk for PB, as classified according to the CL measured in the second trimester of gestation. In agreement with previous reports, a higher concentration of IL-2, IL-6, IFN-γ was found in the cervical-vaginal samples of the high-risk group, as well as a higher concentration of IL-1ra in the low-risk group [22,23,24].
On the other hand, IL-4 and IL-10 were found to be increased in the high-risk group, in contrast to what was expected according to the inflammatory pathogenesis of PB [25]. These differences could be explained by the complexity of the inflammatory processes involved in PB and owing to the different methodologies and designs used in other studies [26]. Elevation of anti-inflammatory cytokines in the high-risk group for PB could be a physiological attempt to moderate the inflammatory process to maintain homeostasis; this hypothesis was previously proposed by Wang et al., as they found that IL-37 (an anti-inflammatory cytokine) was elevated in the fetal membranes of women with preterm labor, as an attempt to stop the inflammatory process caused by IL-6 [27].
IL-6 is one of the most important mediators in infection and inflammation and one of the most studied biomarkers associated with cervical shortening in preterm labor. In our study, IL-6 was found increased in cervical-vaginal samples from the high-risk group, which coincides with that reported by other studies [28]. Goepfert et al. found that cervical IL-6 concentrations measured at 24 weeks of gestational age (wGA) were elevated in women who had preterm delivery before 32 wGA, compared to women whose pregnancies were carried to term, and this finding was even more evident in women with a history of preterm delivery [28,29,30]. In our study, 40% of the women in the high-risk group had a history of a previous delivery before 37 wGA [29,30,31].
In our study, a higher concentration of IFN-γ was found in the group of patients at high-risk for preterm delivery. Accordingly, in a recent report by Sandoval-Colin et al, it was found a positive correlation between elevated IFN-γ, TNF-α, IL-1β, IL-6, and the onset of labor [18], in what appears to be a functional “maturation” of the immune system due to the inflammatory response. Additionally, it has been described that the activation of Th1 cells may increase secretion of TNF-α, INF-γ, and IL-1β in the fetal membranes and in the amniotic fluid in preterm labor [32].
IL-1ra belongs to the IL-1 family of cytokines with an anti-inflammatory action [33]; in our study, we found IL-1ra to be higher in the group of patients at low-risk for PB. To our knowledge, there is no literature on cervical concentration of this cytokine in the pathogenesis of preterm delivery; however our IL-1ra results could be well compared with those of the soluble IL-6 receptor (sIL-6r) that has been associated with a reduction in the risk of preterm delivery (RR 0.4 CI 95% 0.15–0.80) [34,35].
IL-2 is crucial for the maintenance of immune homeostasis and a correlation has been previously demonstrated between elevated IL-2 levels and chronic inflammation [36]. Few studies have demonstrated the relevance of IL-2 and its receptor in the pathogenesis of preterm labor; the possible explanation is that in normal pregnancy, the concentrations of IL-2 and its receptor are relatively low and difficult to measure, but in the case of preterm labor, an increase of this interleukin has been demonstrated that could imply the chronicity of an inflammatory process and not the acute response as in the case of IL-6 [37].
In most previous publications concerning cytokines in cervical-vaginal fluid, the elevated concentrations of IL-10 have been proposed as a protective factor for preterm delivery [38,39], as belongs to the group of cytokines with anti-inflammatory activity. In our study, the highest concentrations were observed in the high-risk group, which is in agreement with the study by Vogel et al. reporting that elevated IL-10 concentrations could also be associated with an increased risk of preterm delivery (RR 3.1 CI 95% 0.96–9.7) [40], as a response to an inflammatory mechanism that was previously initiated [38].
In the second phase of our study, we developed an artificial intelligence model for PB prediction using machine learning (ML), incorporating the quantification of cytokines, as they are the crucial mediators of the inflammatory process observed in preterm labor. Machine learning is the engine that is driving advances in the development of artificial intelligence. The main areas that may benefit from the application of ML techniques in the medical field are diagnosis and outcome prediction; ML has the potential to transform the way medicine works [15,41]. Using AI methods in medical care could facilitate individual pregnancy management and improve public health, especially in low- and middle-income countries. ML allows us to analyze interactions between variables different from what we are conventionally used to, overcoming limitations such as sample size and data distribution. In our study, two predictive models were performed, the full model and the adjusted model. The full model included all the predictor variables (obstetric history, CL, and concentration of all the cytokines measured). The ML analysis allowed us to choose only the variables with the highest predictive significance to build and train the adjusted model that included only maternal age, CL, and IL-2 concentration. Our results demonstrated that the adjusted model, even when it included a smaller number of variables, had a better predictive performance; this is possible because of the type of analysis used, where the variables that full the model are eliminated to improve performance [41,42,43].
In most of the studies involving cytokines in high- and low-risk groups for PB, the classification is made only considering obstetric history [44]. So, one of the main strengths of our work, is that we studied the cervical-vaginal fluid concentration of the principal cytokines involved in the inflammatory process in patients classified as high- and low-risk for PB by CL, the gold standard in current screening and a variable which itself indicates the onset of cervical shortening and therefore the phase prior to the onset of labor. Additionally, the screening was performed at the appropriate weeks of gestation according to international guidelines [45], and was not biased by interventions such as cerclage or progesterone before measurement. Besides, the presence of vaginal infection at the time of sampling for cytokine determination was ruled out.
Another contribution of our work is the comparison between the FMF calculator and the predictive model obtained by the Random Forest analysis. With this analysis we demonstrate that if we add the measurement of IL-2 to the prediction model we can increase almost 8 percentage points the detection rate reported by the gold standard in clinical practice, which would help to identify more accurately those cases at high-risk for PB allowing to implement preventive medical strategies (e.g., progesterone) that can offer an efficiency up to 90% [46,47].
The main limitation of our study is the number of patients included; however, it was possible to identify differences between the groups with significant statistical power. Considering the construction of the model, this fact is compensated by the nature of the analysis performed, since the Random Forest analysis builds the ideal model and then replicates it thousands of times to test its efficiency [48]. However, we consider that a larger number of patients is required to strengthen the model so that it can be considered as a reference for PB prediction in the clinical practice [49].
As our aim was to design an effective screening model for PB with the inclusion of clinically applicable markers based on cost, availability, efficacy, and timeliness of the test, the next step is undoubtedly to perform a cost/benefit analysis to assess whether adding the IL-2 measurement to the screening model is endorsed by the resources saved with the risk detection and prevention of PB.
Overall, our results contribute to a better understanding the cervical-vaginal inflammatory network elicited in patients classified as high- and low-risk for preterm delivery according to the current screening gold standard. The prediction model integrating cytokines in cervical-vaginal mucus showed a DR 8 points higher than the gold standard used in clinical practice, with a lower FNR. The increase in DR with the reduction in FNR may allow to identify women at risk for PB more accurately and implement strategies to prevent this outcome. Our work could represent a significant advance in both screening and reducing the prevalence of PB, a goal that has not being achieved in the last 30 years yet.

4. Materials and Methods

4.1. Ethics statement

The study was conducted at the Instituto Nacional de Perinatologia in Mexico City between January 2019 and December 2021. The protocol was approved by the Ethics, Research, and Biosafety Internal Review Boards (2017-2-69). Women who met the inclusion criteria were invited to participate, and they read and sign the informed consent form.

4.2. Study population

Recruitment was performed by considering women who attended the Maternal-Fetal Medicine Department at the Instituto Nacional de Perinatologia in Mexico City, to screen for PB with a singleton 18–23.6 weeks of gestation pregnancy. Clinical data including age, weight, pregestational weight, socioeconomic status, smoking, history of preterm delivery, gestational age by date of last period and corroborated by first trimester US, presence of urinary or vaginal infections, use of antibiotics, follow-up of pregnancy until its termination, were collected. Fetuses with structural alterations or women with a confirmed diagnosis of isthmic cervical incompetence were not included.
Participants were classified as having high- or low-risk for developing PB according to the cervical length measurement, using a cut-off point <20 mm in patients with a history of preterm delivery and <25 mm for patients without a history of PB [21]. In high-risk patients, our institutional protocol was applied by administering micronized progesterone 200 mg vaginally every 24 h, with a follow-up every 2–3 weeks.

4.3. Sample collection

A cervical-vaginal mucus sample from the posterior vaginal fornix was obtained, using a dacron polyester-tipped swab, that was rinsed in collection buffer containing 1X PBS, 0.05% Tween-20, 1% BSA, and a protease Inhibitor Cocktail (Roche). Samples in collection buffer were then centrifuged at 3200 rpm, 15 min at 4 °C, and the supernatants were stored at -80°C until further cytokine analysis. Before preservation, an aliquot was tested to rule out the presence of subclinical vaginal or uterine infection by fresh examination and negative microbiological cultures.

4.4. Cervical-vaginal cytokine quantification

Cytokine concentration in cervical-vaginal mucus was measured using the Luminex X-Map platform with the Bio-Plex Pro Human Cytokine 10-plex that includes IL-2, IL-4, IL-6, IL-8, IL-10, INF-γ, TNF-α, IL-1β, IL-12p70, and IL-1ra. The protocol was performed according to the manufacturer's instructions, and the value for each cytokine was expressed as pg/ml. Inter- and intra-assay variation was <10%.

4.5. Statistical analysis

Data were analyzed with the SPSS software, version 24. Descriptive statistics were performed to characterize groups: for qualitative variables, frequency measures expressed in percentages were used, while for quantitative variables, measures of central tendency (mean, median), as well as measures of dispersion standard deviation (SD). Cytokine concentration was the main point to be evaluated, and differences between groups were assessed using the Mann-Whitney U test.

4.6. Machine learning model

A Random Forest classifier model was created to predict the incidence of preterm infants using 12 clinical variables (women age, weight, pregestational weight, socioeconomic status, smoking history, parity, previous preterm delivery, gestational age at screening (date of last period and corroborated by first-trimester US), precedent urinary or vaginal infections, use of antibiotics, and gestational age at the time of pregnancy resolution) and 10 cervical-vaginal cytokine determinations (IL-2, IL-4, IL-6, IL-8, IL-10, INF-γ, TNF-α, IL-1β, IL-12p70, and IL-1ra), to create the first full model. To adjust the model, we performed hyperparameter selection by out-of-bag error and simple cross-validation resulting in 50 trees, 1 mtry (number of predictors considered in each split), and 1 as maximum depth, thus generating the adjusted model. The models were analyzed in a mathematical matrix to determine false positive rates (FPR) and false negative rates (FNR), their overall efficiency, as well as for comparing the area under the ROC curves (AUC).

Author Contributions

Conceptualization, H.B.-O.; methodology, M.J.R.-S., A.E.-N. and A.F.-P.; formal analysis, H.B.-O., G.E.-G. and J.M.-O.; writing—original draft, H.B.-O. and G.E.-G.; funding acquisition, H.B.-O.; resources, H.B.-O., G.E.-G., M.J.R.-S., A.E.-N., A.F.-P., R.G.-C. and J.C.E.-A.; writing—review and editing, H.B.-O., G.E.-G. and A.F.-P. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Instituto Nacional de Perinatología (Grant 2017-2-69).

Institutional Review Board Statement

The study was conducted at the Instituto Nacional de Perinatologia in Mexico City between January 2019 and December 2021. The protocol was approved by the Ethics, Research, and Biosafety Internal Review Boards (2017-2-69).

Informed Consent Statement

Women who met the inclusion criteria were invited to participate, and they read and sign the informed consent form.

Conflicts of Interest

The authors declare no conflict of interest. The funders had no role in study design, data collection, and analysis, decision to publish, or preparation of the manuscript.

References

  1. Griggs, K.M.; Hrelic, D.A.; Williams, N.; McEwen-Campbell, M.; Cypher, R. Preterm Labor and Birth: A Clinical Review. MCN Am J Matern Child Nurs 2020, 45, 328–337. [Google Scholar] [CrossRef] [PubMed]
  2. Blencowe, H.; Cousens, S.; Oestergaard, M.Z.; Chou, D.; Moller, A.B.; Narwal, R.; Adler, A.; Vera Garcia, C.; Rohde, S.; Say, L.; et al. National, regional, and worldwide estimates of preterm birth rates in the year 2010 with time trends since 1990 for selected countries: a systematic analysis and implications. Lancet 2012, 379, 2162–2172. [Google Scholar] [CrossRef] [PubMed]
  3. Romero, R.; Dey, S.K.; Fisher, S.J. Preterm labor: one syndrome, many causes. Science 2014, 345, 760–765. [Google Scholar] [CrossRef] [PubMed]
  4. Di Renzo, G.C.; Tosto, V.; Giardina, I. The biological basis and prevention of preterm birth. Best Pract Res Clin Obstet Gynaecol 2018, 52, 13–22. [Google Scholar] [CrossRef]
  5. Wei, S.Q.; Fraser, W.; Luo, Z.C. Inflammatory cytokines and spontaneous preterm birth in asymptomatic women: a systematic review. Obstet Gynecol 2010, 116, 393–401. [Google Scholar] [CrossRef] [PubMed]
  6. Simhan, H.N.; Bodnar, L.M.; Kim, K.H. Lower genital tract inflammatory milieu and the risk of subsequent preterm birth: an exploratory factor analysis. Paediatr Perinat Epidemiol 2011, 25, 277–282. [Google Scholar] [CrossRef] [PubMed]
  7. Taylor, B.D.; Holzman, C.B.; Fichorova, R.N.; Tian, Y.; Jones, N.M.; Fu, W.; Senagore, P.K. Inflammation biomarkers in vaginal fluid and preterm delivery. Hum Reprod 2013, 28, 942–952. [Google Scholar] [CrossRef]
  8. Socha, M.W.; Flis, W.; Pietrus, M.; Wartęga, M.; Stankiewicz, M. Signaling Pathways Regulating Human Cervical Ripening in Preterm and Term Delivery. Cells 2022, 11. [Google Scholar] [CrossRef] [PubMed]
  9. Torbé, A.; Czajka, R. Proinflammatory cytokines and other indications of inflammation in cervico-vaginal secretions and preterm delivery. Int J Gynaecol Obstet 2004, 87, 125–130. [Google Scholar] [CrossRef]
  10. Dowd, J.; Laham, N.; Rice, G.; Brennecke, S.; Permezel, M. Elevated interleukin-8 concentrations in cervical secretions are associated with preterm labour. Gynecol Obstet Invest 2001, 51, 165–168. [Google Scholar] [CrossRef] [PubMed]
  11. Ashford, K.; Chavan, N.R.; Wiggins, A.T.; Sayre, M.M.; McCubbin, A.; Critchfield, A.S.; O'Brien, J. Comparison of Serum and Cervical Cytokine Levels throughout Pregnancy between Preterm and Term Births. AJP Rep 2018, 8, e113–e120. [Google Scholar] [CrossRef] [PubMed]
  12. Zaga-Clavellina, V.; Flores-Espinosa, P.; Pineda-Torres, M.; Sosa-González, I.; Vega-Sánchez, R.; Estrada-Gutierrez, G.; Espejel-Núñez, A.; Flores-Pliego, A.; Maida-Claros, R.; Estrada-Juárez, H.; et al. Tissue-specific IL-10 secretion profile from term human fetal membranes stimulated with pathogenic microorganisms associated with preterm labor in a two-compartment tissue culture system. J Matern Fetal Neonatal Med 2014, 27, 1320–1327. [Google Scholar] [CrossRef]
  13. Celik, E.; To, M.; Gajewska, K.; Smith, G.C.; Nicolaides, K.H. Cervical length and obstetric history predict spontaneous preterm birth: development and validation of a model to provide individualized risk assessment. Ultrasound Obstet Gynecol 2008, 31, 549–554. [Google Scholar] [CrossRef] [PubMed]
  14. Kim, H.Y.; Cho, G.J.; Kwon, H.S. Applications of artificial intelligence in obstetrics. Ultrasonography 2023, 42, 2–9. [Google Scholar] [CrossRef] [PubMed]
  15. Arain, Z.; Iliodromiti, S.; Slabaugh, G.; David, A.L.; Chowdhury, T.T. Machine learning and disease prediction in obstetrics. Curr Res Physiol 2023, 6, 100099. [Google Scholar] [CrossRef] [PubMed]
  16. da Fonseca, E.B.; Damião, R.; Moreira, D.A. Preterm birth prevention. Best Pract Res Clin Obstet Gynaecol 2020, 69, 40–49. [Google Scholar] [CrossRef] [PubMed]
  17. Meyyazhagan, A.; Kuchi Bhotla, H.; Pappuswamy, M.; Tsibizova, V.; Al Qasem, M.; Di Renzo, G.C. Cytokine see-saw across pregnancy, its related complexities and consequences. Int J Gynaecol Obstet 2023, 160, 516–525. [Google Scholar] [CrossRef]
  18. Sandoval-Colin, D.E.; Godines-Enriquez, M.S.; Espejel-Núñez, A.; Beltrán-Montoya, J.J.; Picazo-Mendoza, D.A.; de la Cerda-Ángeles, J.C.; Bello-Chavolla, O.Y.; Meraz-Cruz, N.; Chavira-Suárez, E.; Vadillo-Ortega, F. Cervicovaginal Cytokines to Predict the Onset of Normal and Preterm Labor: a Pseudo-longitudinal Study. Reprod Sci 2023, 30, 221–232. [Google Scholar] [CrossRef] [PubMed]
  19. Park, J.W.; Park, K.H.; Lee, S.Y. Noninvasive prediction of intra-amniotic infection and/or inflammation in women with preterm labor: various cytokines in cervicovaginal fluid. Reprod Sci 2013, 20, 262–268. [Google Scholar] [CrossRef] [PubMed]
  20. Cetin, O.; Karaman, E.; Alisik, M.; Erel, O.; Kolusari, A.; Sahin, H.G. The evaluation of maternal systemic thiol/disulphide homeostasis for the short-term prediction of preterm birth in women with threatened preterm labour: a pilot study. J Obstet Gynaecol 2022, 42, 1972–1977. [Google Scholar] [CrossRef]
  21. Berghella, V.; Palacio, M.; Ness, A.; Alfirevic, Z.; Nicolaides, K.H.; Saccone, G. Cervical length screening for prevention of preterm birth in singleton pregnancy with threatened preterm labor: systematic review and meta-analysis of randomized controlled trials using individual patient-level data. Ultrasound Obstet Gynecol 2017, 49, 322–329. [Google Scholar] [CrossRef] [PubMed]
  22. Yavari Kia, P.; Baradaran, B.; Shahnazi, M.; Asghari Jafarabadi, M.; Khaze, V.; Pourasad Shahrak, S. Maternal Serum and Cervicovaginal IL-6 in Patients with Symptoms of Preterm Labor. Iran J Immunol 2016, 13, 229–236. [Google Scholar]
  23. Hanna, N.; Bonifacio, L.; Reddy, P.; Hanna, I.; Weinberger, B.; Murphy, S.; Laskin, D.; Sharma, S. IFN-gamma-mediated inhibition of COX-2 expression in the placenta from term and preterm labor pregnancies. Am J Reprod Immunol 2004, 51, 311–318. [Google Scholar] [CrossRef]
  24. Buxton, M.A.; Meraz-Cruz, N.; Sanchez, B.N.; Foxman, B.; Castillo-Castrejon, M.; O'Neill, M.S.; Vadillo-Ortega, F. Timing of Cervico-Vaginal Cytokine Collection during Pregnancy and Preterm Birth: A Comparative Analysis in the PRINCESA Cohort. Int J Environ Res Public Health 2021, 18. [Google Scholar] [CrossRef]
  25. Chatterjee, P.; Chiasson, V.L.; Bounds, K.R.; Mitchell, B.M. Regulation of the Anti-Inflammatory Cytokines Interleukin-4 and Interleukin-10 during Pregnancy. Front Immunol 2014, 5, 253. [Google Scholar] [CrossRef]
  26. Park, J.S.; Park, C.W.; Lockwood, C.J.; Norwitz, E.R. Role of cytokines in preterm labor and birth. Minerva Ginecol 2005, 57, 349–366. [Google Scholar] [PubMed]
  27. Wang, L.; Liu, Z.; Huang, D.; Ran, Y.; Zhang, H.; He, J.; Yin, N.; Qi, H. IL-37 Exerts Anti-Inflammatory Effects in Fetal Membranes of Spontaneous Preterm Birth via the NF-κB and IL-6/STAT3 Signaling Pathway. Mediators Inflamm 2020, 2020, 1069563. [Google Scholar] [CrossRef]
  28. Gilman-Sachs, A.; Dambaeva, S.; Salazar Garcia, M.D.; Hussein, Y.; Kwak-Kim, J.; Beaman, K. Inflammation induced preterm labor and birth. J Reprod Immunol 2018, 129, 53–58. [Google Scholar] [CrossRef] [PubMed]
  29. Goepfert, A.R.; Goldenberg, R.L.; Andrews, W.W.; Hauth, J.C.; Mercer, B.; Iams, J.; Meis, P.; Moawad, A.; Thom, E.; VanDorsten, J.P.; et al. The Preterm Prediction Study: association between cervical interleukin 6 concentration and spontaneous preterm birth. National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. Am J Obstet Gynecol 2001, 184, 483–488. [Google Scholar] [CrossRef]
  30. Challis, J.R.; Lockwood, C.J.; Myatt, L.; Norman, J.E.; Strauss, J.F., 3rd; Petraglia, F. Inflammation and pregnancy. Reprod Sci 2009, 16, 206–215. [Google Scholar] [CrossRef] [PubMed]
  31. Paternoster, D.M.; Stella, A.; Gerace, P.; Manganelli, F.; Plebani, M.; Snijders, D.; Nicolini, U. Biochemical markers for the prediction of spontaneous pre-term birth. Int J Gynaecol Obstet 2002, 79, 123–129. [Google Scholar] [CrossRef] [PubMed]
  32. Helmo, F.R.; Alves, E.A.R.; Moreira, R.A.A.; Severino, V.O.; Rocha, L.P.; Monteiro, M.; Reis, M.A.D.; Etchebehere, R.M.; Machado, J.R.; Corrêa, R.R.M. Intrauterine infection, immune system and premature birth. J Matern Fetal Neonatal Med 2018, 31, 1227–1233. [Google Scholar] [CrossRef]
  33. Smith, A.J.; Humphries, S.E. Cytokine and cytokine receptor gene polymorphisms and their functionality. Cytokine Growth Factor Rev 2009, 20, 43–59. [Google Scholar] [CrossRef] [PubMed]
  34. Gonçalves, L.F.; Chaiworapongsa, T.; Romero, R. Intrauterine infection and prematurity. Ment Retard Dev Disabil Res Rev 2002, 8, 3–13. [Google Scholar] [CrossRef] [PubMed]
  35. Mitsuyama, K.; Matsumoto, S.; Rose-John, S.; Suzuki, A.; Hara, T.; Tomiyasu, N.; Handa, K.; Tsuruta, O.; Funabashi, H.; Scheller, J.; et al. STAT3 activation via interleukin 6 trans-signalling contributes to ileitis in SAMP1/Yit mice. Gut 2006, 55, 1263–1269. [Google Scholar] [CrossRef]
  36. Gaffen, S.L.; Liu, K.D. Overview of interleukin-2 function, production and clinical applications. Cytokine 2004, 28, 109–123. [Google Scholar] [CrossRef] [PubMed]
  37. Alvarez-de-la-Rosa, M.; Rebollo, F.J.; Codoceo, R.; Gonzalez Gonzalez, A. Maternal serum interleukin 1, 2, 6, 8 and interleukin-2 receptor levels in preterm labor and delivery. Eur J Obstet Gynecol Reprod Biol 2000, 88, 57–60. [Google Scholar] [CrossRef] [PubMed]
  38. Cao, X.L.; Zhou, X.Y.; Xu, N.X.; Chen, S.C.; Xu, C.M. Association of IL-4 and IL-10 Polymorphisms With Preterm Birth Susceptibility: A Systematic Review and Meta-Analysis. Front Immunol 2022, 13, 917383. [Google Scholar] [CrossRef]
  39. Busse, M.; Zenclussen, A.C. IL-10 Producing B Cells Protect against LPS-Induced Murine Preterm Birth by Promoting PD1- and ICOS-Expressing T Cells. Cells 2022, 11. [Google Scholar] [CrossRef] [PubMed]
  40. Vogel, I.; Goepfert, A.R.; Thorsen, P.; Skogstrand, K.; Hougaard, D.M.; Curry, A.H.; Cliver, S.; Andrews, W.W. Early second-trimester inflammatory markers and short cervical length and the risk of recurrent preterm birth. J Reprod Immunol 2007, 75, 133–140. [Google Scholar] [CrossRef] [PubMed]
  41. Bertini, A.; Salas, R.; Chabert, S.; Sobrevia, L.; Pardo, F. Using Machine Learning to Predict Complications in Pregnancy: A Systematic Review. Front Bioeng Biotechnol 2021, 9, 780389. [Google Scholar] [CrossRef] [PubMed]
  42. Mohammad, N.S.; Nazli, R.; Zafar, H.; Fatima, S. Effects of lipid based Multiple Micronutrients Supplement on the birth outcome of underweight pre-eclamptic women: A randomized clinical trial. Pak J Med Sci 2022, 38, 219–226. [Google Scholar] [CrossRef] [PubMed]
  43. Saberi-Karimian, M.; Khorasanchi, Z.; Ghazizadeh, H.; Tayefi, M.; Saffar, S.; Ferns, G.A.; Ghayour-Mobarhan, M. Potential value and impact of data mining and machine learning in clinical diagnostics. Crit Rev Clin Lab Sci 2021, 58, 275–296. [Google Scholar] [CrossRef] [PubMed]
  44. Oskovi Kaplan, Z.A.; Ozgu-Erdinc, A.S. Prediction of Preterm Birth: Maternal Characteristics, Ultrasound Markers, and Biomarkers: An Updated Overview. J Pregnancy 2018, 2018, 8367571. [Google Scholar] [CrossRef]
  45. Glover, A.V.; Manuck, T.A. Screening for spontaneous preterm birth and resultant therapies to reduce neonatal morbidity and mortality: A review. Semin Fetal Neonatal Med 2018, 23, 126–132. [Google Scholar] [CrossRef]
  46. Jung, E.Y.; Park, J.W.; Ryu, A.; Lee, S.Y.; Cho, S.H.; Park, K.H. Prediction of impending preterm delivery based on sonographic cervical length and different cytokine levels in cervicovaginal fluid in preterm labor. J Obstet Gynaecol Res 2016, 42, 158–165. [Google Scholar] [CrossRef]
  47. Norman, J.E. Progesterone and preterm birth. Int J Gynaecol Obstet 2020, 150, 24–30. [Google Scholar] [CrossRef]
  48. Lee, K.S.; Ahn, K.H. Artificial Neural Network Analysis of Spontaneous Preterm Labor and Birth and Its Major Determinants. J Korean Med Sci 2019, 34, e128. [Google Scholar] [CrossRef]
  49. Giouleka, S.; Tsakiridis, I.; Kostakis, N.; Koutsouki, G.; Kalogiannidis, I.; Mamopoulos, A.; Athanasiadis, A.; Dagklis, T. Preterm Labor: A Comprehensive Review of Guidelines on Diagnosis, Management, Prediction and Prevention. Obstet Gynecol Surv 2022, 77, 302–317. [Google Scholar] [CrossRef] [PubMed]
Preprints 83228 g001
Figure 1. Differences in cytokines measured in cervical mucus: (A) Boxplot with proinflammatory cytokines with significant differences between groups: IL-2, IL-6 and IFN- γ. (B) Anti-inflammatory cytokines with significant differences between groups: IL-4, IL1ra and IL-10. *p < 0.05, **p < 0.01, ***p < 0.001.
Figure 1. Differences in cytokines measured in cervical mucus: (A) Boxplot with proinflammatory cytokines with significant differences between groups: IL-2, IL-6 and IFN- γ. (B) Anti-inflammatory cytokines with significant differences between groups: IL-4, IL1ra and IL-10. *p < 0.05, **p < 0.01, ***p < 0.001.
Preprints 83228 g001
Preprints 83228 g002
Figure 2. Performance of the proposed models generated by machine learning: (A) Full model includes all the variables included as predictors. (B) Adjusted model includes the variables with the highest statistical relevance CL, IL2, and maternal age. (C) ROC curves of the two proposed models, red "Full model" with an AUC = 0.722, green "Adjusted model" with an AUC = 0.879.
Figure 2. Performance of the proposed models generated by machine learning: (A) Full model includes all the variables included as predictors. (B) Adjusted model includes the variables with the highest statistical relevance CL, IL2, and maternal age. (C) ROC curves of the two proposed models, red "Full model" with an AUC = 0.722, green "Adjusted model" with an AUC = 0.879.
Preprints 83228 g002
Table 1. Baseline characteristics and clinical data of pregnant women included in the study.
Table 1. Baseline characteristics and clinical data of pregnant women included in the study.
Low Risk for Preterm
Delivery (n = 40)		 High Risk for Preterm
Delivery (n = 20)		 p-value
Age (years) 29 (±7.1) 31 (±5.8) 0.25
Pregestational weight (Kg) 63.7 (±13.7) 67.8 (±13.5) 0.08
Pregestational BMI (Kg/m2) 25.2 (±5.4) 27.5 (±5.3) 0.12
Socio-economic level median
(Minimum and maximum value) 		2 (1–4) 2 (1–5) 0.12
Smoking n (%) 1 (2.5) 0 (0) 0.45
History of preterm
delivery n (%) 		0 (0) 8 (40) 0.01
Gestational age at time of cervical length measurement, (weeks of gestation) 21.0 (±1.5) 21.2 (±2.0) 0.25
Cervical length (mm) 33.8 (±5.8) 13.1 (±7.7) 0.02
Mean, median, standard deviation, minimum and maximum value, comparations with Mann Whitney U Test, p-value.
Table 2. Pro-inflammatory and anti-inflammatory cytokine profile in cervical-vaginal fluid at 18.0–23.6 weeks of gestation in high (n = 20) and low risk (n = 40) groups for preterm birth.
Table 2. Pro-inflammatory and anti-inflammatory cytokine profile in cervical-vaginal fluid at 18.0–23.6 weeks of gestation in high (n = 20) and low risk (n = 40) groups for preterm birth.
Cytokine Risk group
for preterm birth		 Mean ± SD
pg/ml 		p-value
Pro-inflammatory cytokines
IL-1β High-Risk 763.87 (±1505.99) 0.814
Low-Risk 587.94 (±1432.56)
IL-2 High-Risk 5.63 (±6.07) 0.001
Low-Risk 3.60 (±14.80)
IL-6 High-Risk 856.29 (±1.98) 0.001
Low-Risk 118.32 (±0.48)
IL-8 High-Risk 5882.35 (±5638.79) 0.381
Low-Risk 9695.78 (±11,070.29)
IL-12 High-Risk 0.49 (±0.49) 0.304
Low-Risk 0.34 (0.29)
TNF-α High-Risk 104.17 (±74.62) 0.115
Low-Risk 78.63 (±50.32)
IFN-γ High-Risk 117.49 (±53.42) 0.001
Low-Risk 54.17 (±26.37)
Anti-inflammatory cytokines
IL-4 High-Risk 20.98 (±10.78) 0.001
Low-Risk 10.83 (±8.92)
IL-10 High-Risk 40.44 (±41.23) 0.001
Low-Risk 3.56 (±5.22)
IL-1ra High-Risk 29768 (±17,596) 0.002
Low-Risk
Mean, standard deviation, comparations with Mann Whitney U Test, p-value.
Table 3. Comparison between the two proposed classifiers models: “Full model”, “Adjust model” and Fetal Medicine Foundation (FMF) calculator.
Table 3. Comparison between the two proposed classifiers models: “Full model”, “Adjust model” and Fetal Medicine Foundation (FMF) calculator.
Random Forest “Full model” Random Forest “Adjust model” Fetal Medicine Foundation Calculator
Predicted Real Predicted Real Predicted Real
Term Preterm Term Preterm Term Preterm
Term 14 6 Term 20 1 Term 36 4
Preterm 2 1 Preterm 2 7 Preterm 7 13
Detection rate 65% Detection rate 87.7% Detection rate 79%
False positive rate 12% False positives rate 3.33% False positives rate 6.6%
False negative rate 28% False negatives rate 6.66% False negatives rate 11.66%
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.
Prerpints.org logo

Preprints.org is a free preprint server supported by MDPI in Basel, Switzerland.

facebook logotwitter logolinkedin logo
MDPI Initiatives
Important Links
Subscribe

Disclaimer

Terms of Use

Privacy Policy

© 2024 MDPI (Basel, Switzerland) unless otherwise stated