Exploration of threshold analysis in the relation between stressful life events and preterm delivery.

Biologic evidence suggests that the hormones activated by stress affect gestational length, but the results of epidemiologic investigations are inconsistent. The authors of this paper know of no threshold models that have been studied; these models assume that stress does not affect preterm delivery until a certain amount of stress has been experienced but that each unit of stress above the threshold adds to the risk of preterm delivery. By using standard logistic regression, the authors compared threshold and nonthreshold models of the relation between number of stressful life events and preterm delivery in 11 US states. They used data on 1990-1995 births from the Pregnancy Risk Assessment Monitoring System. The risk of preterm delivery among multiparas who gave birth in 1990-1993 increased 7% for each event over five they experienced, but no relation was found for 1994-1995 births. Among primiparas who gave birth in 1994-1995, the risk increased 5% for each event over two, but no relation was found for 1990-1993 births. These results suggest that a threshold model may fit the relation between stress and preterm delivery better than one with no threshold. However, the inconsistent results are difficult to reconcile with a biologic threshold in the relation between stress and preterm delivery.