Epidemiologic data were analyzed for a total of 2,693 infants with esophageal atresia registered in nine congenital malformation registries around the world. The average recorded prevalence at birth was 2.6 per 10,000 births, with a significant variability among programs--and sometimes within a program--and a maximum prevalence of above 3 per 10,000 births. Clusters of infants with esophageal atresia were observed but may be random. An increasing rate was seen during the period 1965 to 1975 (Norway, South America, Sweden). The type of esophageal atresia was specified in only 439 cases, but no major differences were seen in the epidemiologic characteristics of infants with the most common type (distal fistula) and infants with other types. There was an excess of low birth weight and preterm birth, and infants with esophageal atresia had a birth weight 500 to 1,000 g less than normal infants in each gestational week. There was an excess of twins, apparently mainly or exclusively due to monozygotic twinning, but in only two pairs did both twins have esophageal atresia. There was no effect seen of maternal age, but low parity, irrespective of maternal age, was associated with an increased risk for esophageal atresia. Infant survival varied among programs and depended heavily on associated malformations. Among 1,107 sibs born before the proband and 385 born after the proband, only 25 (1.7%) had a serious malformation; three had esophageal atresia. In 57.3% of the infants with esophageal atresia, no other malformations were present, in 36.4% other major malformations were recorded, and in 6.3% there were chromosomal anomalies. The malformations present associated with esophageal atresia were analyzed: a large proportion entered the constellation sometimes called "caudal mesoderm spectrum of malformations": VATER, Potter, and caudal regression sequences.