British Journal zyxwvuts of zyxwvutsrq Obstetrics and Gynaecology zyxwvut September 1991, zyxwvutsr Vol. 98, pp. 898-904 Mode of delivery and neonatal death in 17 587 infants presenting by the breech JOHN L. KIELY Abstract Objective-To study thc effects of caesarean section on neonatal mortality in infants presenting by the breech. Design-Population-based non-experimental comparison of infants prcscnting by the breech born vaginally with those born by cacsarcan scction. Neonatal mortality rates were calculated for 250 g birthweight intervals. Weight-spccific relative risks (RRs) were further adjusted for birthweight in zyx 50 g catcgorics. Setting-New York City, 1978-1983. Data came from the Departrncnt of Health’s computerized vital records on livebirths and infant dcaths. Subjects-17 587 singleton breech livebirths 2500 g birthwcight, with congeni- tal anomaly deaths excluded. 6178 were born vaginally and zyxw 11409 were born by caesarean section. Main outcome measures-Birthweight-specific and birthweight-adjusted neo- natal mortality. Results-At birthwcights of 501 to 1750 g, the risk of neonatal dcath for breech infants born vaginally was significantly higher than the risk for those born by caesarean section (.weight-adjusted RR= 1.7). For breech infants with birth- weights over 3000 g, the weight-adjustcd risk was 5.6 times greater for a vaginal birth compared with caesarean section. Thc addition of 16 additional control variables in multiple logistic rcgrcssion analyses did not change these RRs. Conclusion-Population-based studies indicatc that an increase in the cae- sarean section rate among breech singletons may be associatcd with increased neonatal survival, but a large multicentrc randomized trial of managemcnt of brccch presentation would answcr the question much more dcfinitivcly. This is a report of the results of an analysis of whether babies presenting by the breech who are born vaginally arc at higher risk of neonatal mor- tality than those who are born by caesarean sec- tion. The data were non-expcrimental, that is, pregnant women were not randomly allocatcd to Gertrude H. Sergievsky Center, Faculty of Medicine, Columbia University J. L. KIELY Correspondcnce: Dr John Kiely, Gcrtrudc H. Scrgi- cvsky Center, Faculty of Medicinc, Columbia Univer- sity, 630 West 168th Strcct. New York 10032, USA. 898 vaginal or cacsarean ‘treatment’ groups. In non- experimental studies of therapeutic intcrvcn- tions, sclcction bias can be a major problem (Chalmcrs 1983; Feinstein 1985; Miettincn 1983). In assessing perinatal interventions, where the decision to intervene is madc by the clinician, selcction bias operates as follows: women at different risks of an adverse outcome havc different probabilities of recciving the intervention bcing studied (Greenland & Neutra 1980). In this case, a comparison of outcomes in caesarean and vaginal births is confounded by the fact that the two groups have different inher- ent risks (Lumlcy et al. 1985).