MCHB/EPI Miami Conference — December 7 - 9, 2005

Preterm Delivery: Current Challenges — Transcript

CLAUDIA HOLZMAN: I want to start by thanking you for inviting me to present this morning. Bao-Ping gave you background on the trends in preterm delivery risk and the associated infant mortality and morbidity. So I'm going to identify the challenges we face in understanding and preventing preterm delivery and I'll show you an approach we're using to study this problem.

A major challenge has been disentangling risk factors for preterm delivery. This is due in part to the association between poverty and risk of preterm delivery which has been reported in many countries including the U.S. The data in this table are from a study in Canada showing the increasing percentage of preterm delivery with decreasing neighborhood income measured at the ecological level. Because preterm delivery is associated with poverty other poverty related factors become associated with preterm delivery. The challenge is separating true mediating factors from those that are confounded.

Here's a list of some of the most frequently studied risk factors for preterm delivery. Many of these factors are interrelated and are more prevalent in disadvantaged populations. In addition many of these factors that need to be addresses or would be best addressed before pregnancy. One of the most puzzling and troubling findings is the persistently high risk of preterm delivery among African/Americans. Despite the high prevalence of poverty in many minority groups in the U.S. their preterm delivery risks are never as high as that found in African/Americans. And more troubling still the ethnic disparity increases as the gestational age of delivery gets lower.

Several studies have reported that among minorities mothers born in the U. S. have a greater risk of delivering preterm than do mothers who are foreign born. While some of this difference may reflect selection bias of health immigrants it may also reflect cumulative adverse health effects across generations of minority women living in the U. S. These affects again seem most striking for African/Americans.

Multiple studies have tried to disentangle the ethnic and socioeconomic status effects on risk of preterm delivery. Sometimes by studying military populations where employment and access to health care are similar or by restricting the population to samples of college educated women or parents. In all cases the ethnic disparities and preterm delivery remain but some argue that these study methodologies cannot overcome inherent social inequalities that persist and continue to confound ethnic comparisons of preterm delivery risk.

Moving to another challenge preterm delivery unlike a disease is the premature occurrence of a normal event (inaudible). It can be influence by medical practice through labor induction and/or C section. About 25 percent to 35 percent of preterm deliveries are initiated by clinicians and percent is rising. This not only affects trends in preterm delivery but also complicates research. First, there's no way to know if a medically indicated preterm birth would have needed as a spontaneous preterm birth sometime later and second the threshold criteria for initiating preterm delivery may vary by such factors as maternal reproductive history and medical practice settings. A challenge that's often over looked is our limited knowledge about the physiology of human (inaudible) both term and preterm. Our earliest understanding of (inaudible) came from sheep models in which endocrine based mechanisms for (inaudible) direct the transition from uterine (inaudible) to active labor stages.

The rise in fetal coritsol results in a drop off in maternal progestin that is detectible in maternal blood. But this fall in progesterone is mediated by a glycol (inaudible) inducible enzyme and this enzyme is in the sheep placenta but it is not found in the human placenta. Unlike the sheep, in humans there's no discernable change in maternal blood progesterone levels in association with (inaudible). While we know the basic physiologic changes that must be in place for regular sustained contractions in human (inaudible) we're much less clear about the precise details that trigger these changes. Is the timing of (inaudible) under the control of the mother, the placenta and membranes, the fetus or cross top between all three? Much of what goes on is confined to the gestational tissues and then through process that are autocrine and peracrine communication between cells and adjacent tissues making it difficult to study this process in vivo in humans. In addition there's ample evidence to suggest that when (inaudible) occurs prematurely the causes may vary and labor may or may not be triggered in the same way. This raises the question should we be grouping preterm deliveries as a way to gain insights into the etiology and target interventions? But what's the best way to group preterm deliveries?

In the mid 1990's (inaudible) along with other investigators raised concerns about the common practice of grouping preterm deliveries by clinical circumstances or by weeks of gestation. The main point was that these groupings may not be sufficiently discriminating for uncovering etiologic pathways to preterm delivery. Many investigators believe that we would make more progress if we could group preterm deliveries according to underlying processes. Some of the processes include inflammation, impaired vascular exchange, hemorrhage and uterine distention. In many cases these pathways may overlap. One strategy is to identify biomarkers such as cytokines and thrombin that may promote preterm labor and then develop interventions. Another strategy is to measure biomarkers of final common pathways that are farther downstream. These downstream biomarkers will likely perform better in predicting preterm delivery but at this point it may be difficult or unwise to interrupt the process. What's the evidence for an inflammation infection pathway to preterm delivery?

Positive preterm deliveries been associated with positive cultures and amniotic fluid obtained and mid trimester, (inaudible) in delivered placentas, evidence of a fetal inflammatory response, urinary tract infections in pregnancy, a condition known as bacterial vaginosis and periodontal disease. In addition there are animal models that have produced preterm delivery by introducing bacteria or bacteria related antigens into gestational tissues. Studies find that infection inflammation is most strong associated with preterm deliveries occurring at the earliest gestational ages. The microorganisms that have been cultured most frequently in association with preterm delivery include those that are typically over abundant in the vaginal tracts of women with a condition known as bacterial vaginosis or BV. This supports the hypothesis that the inflammation infection associated with preterm delivery arises from ascending infections into the upper genital tract either prior to or during pregnancy. Despite the compelling evidence for ascending infections as a cause of preterm delivery the magnitude of the association between BV and preterm delivery is modest. About a two fold increase in odds ratio. And only slightly stronger for the earlier preterm births. This has led investigators to consider more complex questions about BV and preterm delivery. In this (inaudible) shift in vaginal (inaudible) are some bacteria more key than others? Do these bacteria acquire virulence factors that affect their pathogenicity? And how does the host immune system respond and influence the vaginal ecology and the risk of delivering preterm when BV is present? To add to the complications we return to the challenge of disentangling risk factors.

In a study in the U. K. pregnant women were enrolled before their 10 th week of pregnancy, assessed for BV and followed to determine rates of miscarriage in BV positive and BV negative women. The first odds ratio is the association between BV and miscarriage. The second and more powerful is the association between BV and the decision to terminate the pregnancy. Clearly, a confounded association that shows BV is not independent from other social (inaudible) factors. Like BV periodontal disease an inflammatory condition in the mouth is more common in disadvantage populations. Microorganisms that predominate in periodontal disease such as the (inaudible) bacterium may travel through the blood to the gestational tissues or a woman's heightened immune response to periodontal disease may trigger preterm delivery. Results from studies on preterm delivery and periodontal disease have been mixed but among the positive studies the odds ratios are considerably larger than that found in studies of BV and preterm delivery. The presence of periodontal disease and/or BV may serve as indicators of unique patterns of immune response, the host, which are underlying causes of preterm delivery.

For the pathway that includes impaired vascular exchange studies are consistent showing increased risk of medically indicated preterm births among women with clinical symptoms such as hypertension and protinuria that define pre eclampsia. There's less agreement on the relationship between impaired vascular exchange and spontaneous preterm birth though again there's the problem of right censoring. There's some evidence that in (inaudible) women in pregnancy placental abnormalities related to low uteral blood flow do increase the risk of spontaneous preterm delivery. Multiple studies have shown a positive association between bleeding and pregnancy and risk of preterm delivery. Particularly if bleeding occurs in more than one trimester. Deicidual hemorrhage may present as clinically evident abruptions or associated thrombin maybe contribute to premature breakdown of fetal membranes and membrane rupture. What about maternal stress? Which I'm feeling a little now. There are biologic mechanisms suggesting that stress could play a role in any of the three pathways discussed. In addition stress may have a direct effect on the timing of (inaudible). Currently investigators are attempting to gain empirical evidence to support the proposed biologic mechanism involving stress.

Another challenge is the limited knowledge we have on genetics and preterm delivery. There's agreement that like other complex disorders preterm delivery is likely to be polygenetic with gene-gene and gene environment interactions. The increased risk of delivering preterm among women with a previous preterm delivery is well established and there's some evidence that mothers born preterm have a higher risk of delivering preterm. While these associations have been used to argue for the importance of genetics in preterm delivery they're also consistent with a persistent adverse environment as a major influence on preterm delivery risk. Twin studies offer more convincing evidence of estimated heritability but these estimates may vary greatly with the population studied. There remain many unanswered questions regarding the best way to investigate the genetic contribution to preterm delivery risk. Whose genes matter? The mother, the fetus or the interaction of the two? If we try to do full genome scans we need very large numbers of pregnancies that will likely include a great variety of populations. This approach works best when there is one or more clear (inaudible) or subtypes, something we're lacking in our current stage of preterm delivery research. The candidate gene approach is more targeted but cannot easily account for the gene-gene interactions.

Some intriguing results have come from studies that model gene environment interactions using gene polymorphisms that affect immune response such as cytokine levels. Here data show that among women with BV those who are also a carrier of T and F alpha two have a significantly increased risk of delivering preterm. To date we have had minimal success with interventions to prevent preterm delivery. (Inaudible) used to stop preterm delivery are mainly useful in prolonging labor by one to seven days long enough to administer steroids for fetal lung maturation if needed. In a recent review of (inaudible) clinical trials the authors concluded that the odds of prolonging the pregnancy to term were in the range of 1.6 to 2.5. The trials were not methodically rigorous and (inaudible) did not show a clear a benefit for parinatal outcomes. Some argue that distinct sub groups of women with preterm labor may particularly benefit from (inaudible) but clinical trials have not been designed to adequately test this distinction. Results from clinical trials using cervical (inaudible) have been controversial.

In a recent meta analysis the overall effect estimate suggested no benefit of (inaudible) with an odds ratio of .84 and a confidence interval that included one. But in subgroup analysis significantly lower rates of preterm delivery were found in women with documented short cervical length and singleton pregnancies. Clinical trials of antibiotics for women not in labor have mainly focused on lowering the risk of preterm delivery by treating bacterial vaginosis. In women with asymptomatic bacterial vaginosis treatment with (inaudible) in pregnancy has been largely unsuccessful in both high risk indicated her by HR and lower risk LR women. In some studies treatment has been associated with an increased risk of preterm delivery. In a recent clinical trial by Andrews and colleagues women who had a previous delivery before 34 weeks were randomly assigned to receive placebo or regular antibiotic treatment in the interpregnancy interval.

Again, the group receiving antibiotics had an increased risk of preterm delivery. Clinical trials of (inaudible) use to treat BV in pregnancy are fewer and smaller and only two include oral treatment. The trial by (inaudible) in Vienna , Austria used a different approach. Pregnant women were screened for BV and results were shared or not shared with the prenatal care provider based on random assignment. The treatment protocol was standardized and the group whose screen results were passed on to the provider had a significantly lower risk of preterm delivery. In a recent clinical trial conducted by the NICHD Maternal Fetal Medicine Units Network weekly injections of 17P progesterone beginning at 16 to 20 weeks gestation lowered the risk of preterm delivery in women with a previous preterm delivery. However, several factors in the trial are worth noting. Only half of all eligible women participated. An unusually high percentage or participants had more than one previous preterm birth and by chance a larger proportion of these women were in the placebo group. The 54.9 percent preterm births in the placebo group was unusually high and the 36.3 percent preterm births in the treated group was closer to the percentage reported in most observational studies.

Children born to treated mothers are being followed for any unanticipated long term effects. Several studies are underway to examine the affects of treating pregnant women for periodontal disease. The one published randomized clinical trial is from Chile that enrolled women with gingivitis. Non treated women had an almost three fold increase in odds of delivering a preterm or low birth weight infant. In face of the limited success in interventions a major thrust of preterm delivery research continues to be the search for relevant biomarkers. The hope is that biomarkers will improve our understanding of (inaudible) and preterm delivery or aid clinicians in predicting preterm delivery particularly in women with preterm labor who may or may not go on to deliver preterm and may be useful in identifying subgroups of women for targeted interventions but there are many challenges in measuring relevant biomarkers and I'm going to show a few of them today to you. First, where and when should we measure the biomarkers? Though we may get important clues by examining compartments such as placenta, amniotic fluid and cord blood these sources are not easily sampled during pregnancy. Ultimately findings in these compartments need to be linked to upstream factors that can be measured through minimally invasive means and early enough in pregnancy to intervene successfully.

Here's a list of some of the most widely studied biomarkers considered as possible predictors of preterm delivery and reviewed her by Vogel and colleagues. Typically a positive likelihood ratio of greater than ten applies the biomarker could be useful clinically. In asymptomatic women no single biomarker obtained through minimally invasive modes came close to an LR of plus ten. Investigators trying out combinations of these and other biomarkers to see if they can improve their clinical applicability.

In women with preterm labor one of the most widely used biomarkers is fetal fibronectin measured in cervical vaginal fluid. Fetal fibronectin's a protein that is part of the intracellular matrix at the border between the maternal deciduas and the fetal membranes. Women who have a negative fetal fibronectin test have a very low probability of delivering before 35 weeks and therefore do not need antenatal steroids or close surveillance in the hospital. However, a positive fibronectin is less useful. Only one out of three women with a positive test deliver before 35 weeks. The fetal fibronectin test has not been useful in identifying a subgroup of women who would benefit from antibiotic intervention. Sorting out the meaning of elevated biomarker levels presents additional challenges.

Often we measure a biomarker in a compartment other than where it is produced. For example, consider elevated levels of AFP or CRH in mid trimester maternal blood, both of which have been linked to an increased risk of preterm delivery. Is a stress fetus producing more AFP or is a stressed placenta producing more CRH? Or is production unchanged but placental abnormalities allow these biomarkers to enter into the maternal seculation in excessive amounts? For some biomarkers there's considerable (inaudible) variability (inaudible) then there's the challenge of interpreting results form biomarkers with levels that normally change during the course of pregnancy. The figure on the left depicts a biomarker that increases as pregnancy progresses. Comparing levels in preterm and term at the time of labor could be misleading. It would be best to measure this biomarker at the same week of pregnancy in both the term and preterm indicated here by the arrow at 22 weeks. The figure on the right shows a biomarker that rises in the weeks preceding labor as an (inaudible) or (inaudible). Measuring this biomarker at 22 weeks gestation would result in different levels in preterm compared with term but only because the increased biomarker was assessed closer to the time of labor in the preterm. It may also be misleading to consider one biomarker at a time because of the complex interactions between sub straights, receptors, agonists and antagonists.

In this study the authors argued for models that reflect bioactivity, bioactivity of given cytokine incorporated the absolute level of that cytokine as well as levels of other biomarkers that reduce or enhance the impact of that cytokine. So in light of all these challenges researchers continue to explore new approaches to study pathways to preterm delivery and as an example I'll show you an approach that we've been using in our study the pregnancy outcomes and community health study or POCH study that was designed to assess psychosocial and biologic factors related to preterm delivery.

In an effort to better understand the different pathways to preterm delivery we use the approach of grouping pregnancies by placental histopathology findings. In the most simplified model we group pregnancies based on whether they've had infection inflammation or vascular abnormalities predominate or there are less common abnormalities or little to no abnormal findings. We then compare levels of biomarkers measured prospectively in term and preterm both within and across the histopathology based subtypes. And finally we evaluate personal and social antecedents to identify upstream factors in these pathways. We designed the POCH study to test the usefulness of this approach. Our goal was to build the capacity to assess both simplified and more complex placental histopathology subtypes to include in these models.

The POCH study recruited pregnant women at the time of (inaudible) prenatal screening from 52 clinics in five Michigan communities. Women who are less than 15 years of age or who had preexisting diabetes or multiple fetuses or fetus with a known congenital anomaly were excluded. Participants were enrolled at the 16 th through 20 th week of pregnancy and we enrolled 3,038 women, 19 were lost to follow-up leaving 3,019 women. At enrollment nurses conducted interviews collected, biologic samples and women were offered the opportunity to participate in a three day at home data collection protocol that was primarily designed to assess maternal distress both through self report and stress biomarkers such as salivary cortisol and urinary (inaudible).

Approximately 72 percent of the women in the cohort were 30 years or younger, 24.6 percent were African/American but the percentage of women from other minorities was considerably less and 51.8 percent of women were insured by Medicaid and 42.8 percent were (inaudible). Within the POCH cohort we used a case cohort design to assemble a sub cohort of 1,361 women. The sub cohort consisted of all women who delivered preterm, all women with high MSAFP levels and a random sample of women with normal MSAFP levels who delivered at a term with an over sampling of African/Americans. In this sub cohort 25 percent are preterm, 16 percent have high MSAFP and 40 percent are African/American. The sampling strategy was designed to maximize statistical power for assessing high risk sub groups.

In analysis that include the entire sub cohort sampling fractions are used to reflect the composition of the original cohort. Additional data on the sub cohort included detailed gross and microscopic assessments of the placenta, information from prenatal and labor and delivery records, assays of biomarkers and stored maternal samples and contextual factors from linked census data. As an example of our approach we present data for the biomarker CRH. During pregnancy CRH is produced in the placenta, extra placental membranes and deciduas. Maternal blood levels of CRH increased throughout the second and third trimester of a normal pregnancy. Higher levels have been reported in women who later deliver preterm but the reasons remain unclear. Increases in CRH may be initiated in response to maternal factors such as inflammation, vascular problems or stress or they may be increased in relation to fetal responses. In this case the fetus experiencing inflammation or decreased blood flow.

In the POCH study we measured maternal CRH levels in blood in mid pregnancy at enrollment. We also examined nine tissue samples from the placenta indicated here by the black lines. After a detailed microscopic evaluation of the nine tissue samples the pathologist completed a summary assessment indicating whether the placenta had none, mild, moderate or severe evidence of abnormalities in each of the five broad pathology groupings listed here and the following results are from the first 848 sub cohorts placentas for which we have the data claimed. We first divided the sub cohort into three groups. Group one with no moderate to severe inflammation or vascular pathology, group two with moderate to severe vascular pathology utraplancental or intraplacental and group three with moderate to severe inflammation, acute or chronic but no moderate or severe vascular pathology. In analysis weighted by the sampling fractions we found a significantly higher percentage of African/Americans in the inflammation group and a significantly greater percentage of deliveries before 35 weeks in the vascular and inflammation groups. BV assessment at enrollment approximately mid pregnancy was significantly more prevalent in the inflammation group.

Adjusted mean CRH measured in maternal blood mid pregnancy was significantly elevated in the vascular group only. Within each of the three pathology defined groups adjusted mean CRH levels in the preterm categories were compared to that of term. Only in the vascular group were the CRH levels significantly higher in the preterm. This approach of using placental histopathology to examine pathways to preterm delivery appeared to have face validity as evidenced by the evidence of higher BV prevalence and higher percentage of African/Americans in the inflammation group, findings consistent with much of the literature. In addition, our results highlighted one of the potential benefits of this approach, the opportunity to gain new insights into biomarkers of preterm delivery. In this case we learned that maternal levels of CRH in mid pregnancy were strongly linked to vascular pathways but did not mark other pathways to preterm delivery. Once we've linked biomarkers to placental pathology sub types we'll go back and assess how these biologic findings fit into the larger picture of the personal and social antecedents that are related to preterm delivery.

Finally, I use this cartoon to remind myself that solving the problem of preterm requires not only biomarkers and tissues but consideration of the women and their environments. Finally, I'd like to acknowledge the team of members in the POCH research team that have contributed to this effort and thank you very much.