Risk stratification and pathological mechanisms in preterm delivery

Abstract

The delivery of infants before 37 weeks gestation is a leading cause of perinatal mortality and morbidity in the United States. Traditional methods of predicting women at risk relying on obstetric history or premonitory symptoms (detected clinically or by tocodynamometry) are neither sensitive nor specific. Recent approaches to predicting preterm delivery have included sonographic measurement of cervical length and various biochemical assays. Although more sensitive than traditional methods, none of these alone exhibits sufficient accuracy to warrant widespread use. We contend that the failure of current approaches to predicting preterm delivery reflects an inadequate understanding of the underlying pathogenesis. Clinical and experimental evidence support the concept that most cases of preterm delivery reflect four pathogenic processes, which share a common final biological pathway leading to uterine contractions and cervical changes with or without premature rupture of membranes. These pathogeneses are: (1) activation of the maternal or fetal hypothalamic‐pituitary‐adrenal axis; (2) decidual‐chorioamniotic or systemic inflammation; (3) decidual haemorrhage (i.e. abruption); and (4) pathological distention of the uterus. Our research seeks to combine the most useful biophysical and biochemical markers of such processes with optimal clinical and epidemiological predictors into a composite, easily applied risk tool. This integrated approach has the potential to identify at‐risk asymptomatic patients with high sensitivity, specificity, and positive and negative predictive values, and also to ascertain underlying pathogenic processes that can lead to targeted therapy. To accomplish these goals, we employ logistic regression and artificial neural network models to assess and apply the appropriate weight to markers associated with each of the above pathogenic pathways, in addition to markers of the final common pathway leading to fetal membrane rupture, cervical extracellular matrix degradation, and myometrial activation. By combining these markers, we expect ultimately to produce a predictive model that is more robust than any existing method, and that identifies the relative contribution of each pathogenic process. Further analysis of this model using a neural network will enable us to identify asymptomatic patients destined to deliver preterm with high sensitivity, specificity, positive and negative predictive values, and to assess the relative contribution of each of the four distinct pathogeneses to this preterm delivery risk.