Subjects. The population studied comprised all births and fetal deaths from women living in this area in the period between 1 August 2000 and 31 January 2001. Cases were all perinatal deaths that occurred in the study period in the area identified through a linkage of the Brazilian Live Birth Information System (SINASC) and the Brazilian Mortality Information System (SIM) databases (both available at www.datasus.gov.br). Both systems have very high population coverage for São Paulo (Almeida et al. 2006). Controls were a random sample of all children who survived the 28th day of life, and were obtained through the same databases. We excluded all nonhospital deliveries. The initial sample was defined for a study on perinatal mortality; however, preliminary analysis of results indicated differences between risk factors for early neonatal and fetal deaths. Thus, we decided to analyze these components separately, maintaining the number of controls previously established.