A Translational View of Biomarkers in Preterm Labor
Nazeeh Hanna, Division of Neonatology, Winthrop University Hospital, 259 First street, Mineola, NY 11501, USA.
Preterm labor and birth continue to pose a significant challenge to physicians in the obstetrics and neonatal fields. Until specific and effective therapeutic treatments are developed to prevent preterm labor, the best means of reducing preterm birth rate is early detection and diagnosis. However, current approaches to predict preterm labor have had variable success in the clinical setting. In this review, we discuss several limitations of using biomarkers from biological samples to predict preterm labor. In addition, we propose strategies for improving our ability to predict preterm labor, as well as directing therapies that are best suited to the underlying cause of preterm labor.
Preterm labor and birth are responsible for the majority of neonatal morbidity and mortality including cerebral palsy, blindness, and deafness, resulting in an annual cost of over 26 billion dollars in 2005. Not surprisingly, a tremendous amount of effort has been expended to counter the rising trend in preterm births.
Clinicians are under increasing pressure to practice ‘evidence-based medicine,’ which is often mistakenly interpreted as ‘randomized controlled trials’. Using that criterion, there is a paucity of effective interventions or predictive tools to stop preterm labor. For example, the lack of evidenced-based data suggests we abandon interventions such as IV hydration and reduced activity, which many clinicians believe (at least anecdotally) are effective in some patients. Moreover, the data from ‘the evidence’ appear inconsistent, at least on the surface. For example, midtrimester short cervix (<25 mm) has been shown to be a risk factor for spontaneous preterm birth. Some randomized trials have shown cervical cerclage to be an effective intervention for this condition, while others have not.[3, 4] One of the largest trials addressing cerclage failed to show a reduction in preterm birth prior to 35 weeks in patients with a cervical length <25 mm, but a sub-analysis showed efficacy for those patients with a cervical length <15 mm. These data highlight the primary limitation of randomized trials' generalizability. By design, in order to assess the efficacy of the intervention, the patients need to be the same, which is unrealistic in the clinical setting. Therefore, there needs to be some way of designing trials that will allow us to assess interventions, while at the same time produce information that are applicable to patients in the everyday heterogeneous clinical setting.
Because the pathogenesis of preterm labor is multi-factorial, biomarkers may prove to be useful in following the progression of pregnancy-associated diseases and direct evaluations and therapeutic options toward a particular cause of preterm labor such as inflammation-mediated preterm labor. In addition to their diagnostic value, identifying specific biomarkers may provide clues to developing novel-targeted therapeutics and predict the response and efficacy of such treatment(s).
Immunopathology and current clinical approach
Preterm labor and birth have been proposed to be the end result of a cascade of events, which may begin with infection, inflammation, ischemia, premature activation of the fetal hypothalamic-pituitary axis, maternal-fetal hemorrhage, or uterine over-distension.[6-8] Each of these mechanisms can lead to cervical shortening and preterm labor. Therefore, it is unlikely that the mediators involved in the cascade are identical regardless of the underlying etiology. Differentiating the inciting event may allow for pathway-specific therapy directed at the underlying cause of preterm labor, rather than at the end result (i.e., cervical shortening or preterm labor). An ideal biomarker(s) needs to have several characteristics including good specificity and sensitivity, ability to differentiate between diseases that might have similar clinical presentation, and be accessible by non-invasive means such as blood, saliva, urine, or vaginal/cervical secretions. In addition, in order to be useful in allowing for timely intervention, the ideal biomarker(s) would be detectable during early in-utero events that can predict preterm labor later in pregnancy. Finally, cost-effectiveness and reproducibility in a low-risk population (where most preterm births originate) would allow incorporation into routine practice.
There are several questions that, if adequately addressed, can help identify those ideal biomarkers for preterm labor. Those questions include: Would a multi-marker strategy be a better predictor of preterm labor compared to single marker strategy? Which non-invasive body fluid can predict intrauterine environment? Can those biomarkers reflect different intrauterine compartments equally? Can a physical finding be a marker for inflammation and preterm labor? We will try to address some of those questions.
Would Multi-Marker Strategy be Better Predictor of Preterm Labor Compared to Single Marker Strategy?
One of the known markers for preterm birth is the ultrasonographically identified short cervix.[2, 9] As part of the randomized trials evaluating different interventions to treat the short cervix, we collected amniotic fluid samples and aliquots were frozen for subsequent analysis. These samples were analyzed for inflammatory mediators through the Bio-Plex™ Array (Bio-Rad, Hercules, CA, USA). Regression analysis from this data identified monocyte chemotactic protein-1 (MCP-1) as the mediator most predictive of preterm delivery (among patients who received no intervention in the randomized trials). The sensitivity and specificity for predicting delivery <32 weeks were 91 and 86%, respectively, with a positive predictive value of 88% and negative predictive value of 90%. Although this was an example of what looks to be a useful marker, most similar single markers failed to be reproducible in low-risk populations and in diverse clinical settings. This again highlights the heterogeneity of etiological factors responsible for preterm labor and the multifactorial cascades ending in uterine contraction and preterm labor.
Using multiple biomarkers from different and distinct biologic pathways may better predict the risk of preterm labor. In order to overcome the shortcomings of evaluating individual cytokines, we created a novel amniotic fluid inflammatory score based on a comprehensive evaluation of multiple cytokines and inflammatory mediators in asymptomatic women with short midtrimester cervix. Amniotic fluid from singleton gestations (n = 44) with a cervical length of ≤25 mm between 16 and 24 weeks was assayed for 25 inflammatory mediators. Patient data were stratified according to gestational age at delivery (<34 versus ≥34 weeks) to determine whether there was a difference in the mediator levels between these two groups. Mediators that reached statistical significance were included in the amniotic fluid inflammatory score. Patients were assigned 1 point for each significant mediator if their level was in the upper quartile. The amniotic fluid inflammatory score was determined, and its relationship to other clinical characteristics was examined. The receiver-operator characteristic (ROC) curve yielded a score ≥8 as predictive of delivery prior to 34 weeks with a sensitivity of 87.0%, specificity of 100%, positive predictive value of 100%, and negative predictive value of 87.5%. In addition, when this scoring system was applied to a different cohort of patients who were undergoing routine genetic amniocentesis, all of those patients were classified as having a low inflammatory score. None of those patient delivered prior to 35 weeks. We believe these data demonstrate that multiple markers are superior to a single marker in predicting preterm labor, at least in the context of women presenting with short cervix.
Which Non-Invasive Body Fluid Can Predict Intrauterine Environment?
The primary limitation of studies utilizing biomarkers identified in amniotic fluid is that they require an invasive and sometimes risky procedure (i.e., amniocentesis) in order to determine the in-utero environment. For a biomarker to be incorporated into routine practice, the information needs to be obtainable in a non-invasive manner. However, there are several challenges using non-invasive sampling to predict intrauterine environment including: what is the best non-invasive sampling site that can predict a specific intrauterine immune compartment? For example, is it the urine or the blood sample that has the best biomarkers predictive of placental immune environment? Is this non-invasive sample also predictive of other compartments' immune environment such as amniotic fluid? Can we combine several biomarkers from different non-invasive samples to predict for example placental immune environment?
To begin to answer these questions, we conducted a pilot study comparing inflammatory mediators from non-invasive samples (maternal blood, urine, saliva, vaginal, or cervical secretions) with traditional gold standard invasive samples (amniotic fluid and placenta samples). Term, non-laboring patients without major maternal, or fetal complications undergoing Cesarean delivery were recruited (n = 20). We obtained fluid samples from different maternal and fetal compartments and determine the inflammatory mediator expression in each. These mediators include cytokines, chemokines, and growth factors that again were measured via the Bio-Plex™ Suspension Array system.
The results indicated that different intrauterine compartments are mostly immunologically distinct with few compartments showing similar cytokine expression (Table 1). This finding provides important insight into what has been shown in other studies. For example, in the placenta, low IL-10 has been linked to preterm labor; however, high IL-10 and high pro-inflammatory mediators were observed in amniotic fluid samples associated with preterm labor. Although this finding might appear contradictory, it may indicate a primary deficiency of placental IL-10 production (the pathology) that triggers intrauterine inflammatory environment and increased production of pro-inflammatory mediators. Such inflammatory environment will initiate a feedback up-regulation of anti-inflammatory molecules such as IL-10 in amniotic fluids (the response).
Table 1. Correlation Among Cytokines in Maternal-Fetal Compartments
|Amniotic fluid||(2) IL-5, RANTES||(1) IL-7||(0)||(0)||(0)||(1) MIP-1β|
|Maternal blood||(0)||(5) Eotaxin, IL-13, IL-15, MIP-1α, TNF-α||(4) G-CSF, IP-10, PDGF, TNF-α||(3) IP-10, TNF-α, VEGF||(2) IL-13, PDGF|| |
|Urine||(6) IL-1ra, IL-2, IL-4, IL-10, IL-13, VEGF||(0)||(0)||(1) MCP1|| || |
|Cervical secretions||(3) MIP-1β, RANTES, VEGF||(4) IFN-γ, IL-4, Il-5, IL-17||(25) G-CSF, Eotaxin, GM-CSF, IFN-γ, IL-(1-β,1ra,2, 4, 5, 6, 7,8,9, 10,12,13, 15, 17), IP10, MCP-1, MIP1-β, PDGF, TNF-α, VEGF, FGF|| || || |
|Vaginal secretions||(4) IL-2, IL-8, MCP-1, MIP-1β||(2) IL-17, FGF|| || || || |
|Saliva||(2) IP-10, RANTES|| || || || || |
Not surprisingly, in our study, there was significant correlation between vaginal and cervical samples. The data indicated there are several potential cytokines in non-invasive samples that can be targeted as a biomarker reflecting their expression in the intrauterine environment. Significantly, the study demonstrates that a specific correlation of an intrauterine cytokine may be reflected in one non-invasive site but not another, depending upon the type of cytokine, and the compartment from which it is secreted.
Can a Physical Finding be a Marker for Inflammation and Preterm Labor?
A physical or ultrasound finding may enable the clinician to use a non-invasive tool to predict patients at high risk for specific pathology. Such a finding may enhance the negative and/or the positive predictive value of a chemical biomarker. Previous study demonstrated that sonographic measurement of fetal membrane thickness could be helpful in the prediction of preterm delivery.
Using the amniotic fluid and cervical length data from the randomized trials noted above, we examined the relationship between cervical length and levels of inflammatory mediators in amniotic fluid. Spearman correlations were used to determine which cytokines correlate with cervical length. Stepwise regression identified the most significant cytokine predictive of early delivery, and a ROC curve determined the cervical length cutoff predictive of intra-amniotic inflammation. Our results indicate that cervical length ≤5 mm is associated with significant increases in amniotic fluid inflammatory cytokines, even in the absence of infection or labor. A cervical length of ≤5 mm was associated with significant increases in inflammatory mediators (Interleukin (IL)-1β, IL-2, IL-6, IL-8, and MCP-1), which have been previously shown to be associated with preterm labor.[18, 19] Unfortunately, the dataset was too small to allow a multivariable analysis including both cervical length and mediator levels in predicting outcome. While a very short cervical length is a good indicator of intrauterine inflammation, it represents the final common pathway of multiple inciting events that can result in preterm labor. As such, it is not an ideal biomarker when utilized alone. Many of these patients will go on to delivery prematurely despite intervention. It is likely that markers which identify earlier in-utero events will allow more effective therapies to be designed to stop the preterm labor cascade before the cervix becomes shortened.
Summary and potential clinical approach
It appears that the intrauterine compartments are mostly immunologically distinct, and the expression of inflammatory markers in various maternal-fetal compartments will be differentially expressed in non-invasive sampling sites. Because the etiology of preterm labor is multifactorial, using multiple biomarkers from distinct biologic pathways will better predict the risk of preterm labor. Furthermore, combining non-invasive tools such as a physical or ultrasound finding physical finding may improve the ability of specific biomarker in predicting outcome.
Platforms to measure for example the levels of inflammatory mediator are commercially available and can easily be incorporated into ongoing trials looking at interventions to treat preterm labor. Initially, data can be collected in an observational manner and correlated with outcomes. Taking the short cervix as an example, it is possible that the variable success of therapies (such as cerclage or progesterone therapy) in the clinical setting is a result of treating all patients presented with midtrimester cervix in a similar manner in spite of possibly different etiological factors. Biomarkers may allow us to differentiate these etiological factors. Ultimately, the purpose of any biomarker(s) is not only for its predictive value, but rather in the possibility of directing therapies best suited for an individual patient. As pathway-specific therapies to treat cervical shortening and preterm labor evolve, these data may aid in choosing the most appropriate therapy directed at the underlying cause of cervical shortening and preterm labor.