Modelling prior reproductive history to improve prediction of risk for very preterm birth
Article first published online: 22 JUN 2010
© 2010 Blackwell Publishing Ltd.
Paediatric and Perinatal Epidemiology
Volume 24, Issue 5, pages 402–415, September 2010
How to Cite
Watson, L. F., Rayner, J.-A., King, J., Jolley, D., Forster, D. and Lumley, J. (2010), Modelling prior reproductive history to improve prediction of risk for very preterm birth. Paediatric and Perinatal Epidemiology, 24: 402–415. doi: 10.1111/j.1365-3016.2010.01134.x
- Issue published online: 29 JUL 2010
- Article first published online: 22 JUN 2010
- very preterm birth;
- past obstetric history;
- statistical methodology
Watson LF, Rayner J-A, King J, Jolley D, Forster D, Lumley J. Modelling prior reproductive history to improve prediction of risk for very preterm birth. Paediatric and Perinatal Epidemiology 2010.
In published studies of preterm birth, analyses have usually been centred on individual reproductive events and do not account for the joint distributions of these events. In particular, spontaneous and induced abortions have often been studied separately and have been variously reported as having no increased risk, increased risk or different risks for subsequent preterm birth. In order to address this inconsistency, we categorised women into mutually exclusive groups according to their reproductive history, and explored the range of risks associated with different reproductive histories and assessed similarities of risks between different pregnancy histories.
The data were from a population-based case–control study, conducted in Victoria, Australia. The study recruited women giving birth between April 2002 and April 2004 from 73 maternity hospitals. Detailed reproductive histories were collected by interview a few weeks after the birth. The cases were 603 women who had had a singleton birth between 20 and less than 32 weeks gestation (very preterm births including terminations of pregnancy) and the controls were 796 randomly selected women from the population who had had a singleton birth of at least 37 completed weeks gestation. All birth outcomes were included. Unconditional logistic regression was used to assess the association of very preterm birth with type and number of prior abortions, prior preterm births and sociodemographic factors.
Using the complex combinations of prior pregnancy experiences of women (including nulligravidity), we showed that a history of prior childbirth (at term) with no preterm births gave the lowest risk of very preterm birth. With this group as the reference category, odds ratios of more than two were associated with all other prior reproductive histories. There was no evidence of difference in risk between types of abortion (i.e. spontaneous or induced) although the risk increased if a prior preterm birth had also occurred. There was an increasing risk of very preterm birth associated with increasing numbers of abortions.
This method of data analysis reveals consistent and similar risks for very preterm birth following spontaneous or induced abortions. The findings point to the need to explore commonalities rather than differences in regard to the impact of abortion on subsequent births.