Use of time to event analysis to estimate the normal duration of human pregnancy.

Abstract

Current estimates of the average duration of human pregnancy are flawed by inaccurate estimation of the time of conception and by failure to account adequately for the effect of routine elective delivery post-term. In this study, 1514 healthy pregnant women were studied in whom the discrepancy between the menstrual history and first trimester crown-rump length estimated gestational age was within -1 to +1 day difference. The duration of gestation was estimated using time to event analysis: non-elective delivery was taken to be the event, and elective delivery was taken to be censoring. The median time to non-elective delivery using the Kaplan-Meier product limit estimate was 283 days after last menstrual period (LMP) and there was no difference comparing male and female fetuses. The median was significantly greater for nulliparous women compared with multiparous women (284 versus 282 days, P < 0.0001). Multivariate analysis using Cox's proportional hazards model confirmed the independent effect of nulliparity on duration of pregnancy [hazard ratio, 0.75; 95% confidence interval (CI) 0.67-0.85] and demonstrated no effect of maternal age, previous abortions, fetal sex, high parity, or bleeding before 24 completed weeks of gestation. Bleeding in the third trimester of pregnancy was, however, associated with an earlier onset of spontaneous labour (hazard ratio, 1.38; 95% CI 1.03-1.84). This study provides a basis for predicting the probability of labour at a given gestational age at term.