We did a prospective study of women with singleton viable pregnancies at 10-13 weeks' gestation who requested first-trimester fetal karyotyping because of advanced maternal age, parental anxiety, or family history of chromosomal abnormality. Women were counselled as to the available options of non-invasive screening or invasive testing by mid-trimester amniocentesis, early amniocentesis (EA), or chorionic villus sampling (CVS), or randomisation to EA or CVS at 10-13 weeks. EA was done in 731 patients (493 by choice and 238 by randomisation) and CVS in 570 (320 by choice and 250 by randomisation). Both procedures were done by transabdominal ultrasound-guided insertion of a 20-gauge needle. The rate of successful sampling was the same for both procedures (97.5%). Spontaneous loss (intrauterine or neonatal death) was significantly higher after EA (total group mean = 5.3%, 95% CI 3.8-7.2; randomised subgroup mean = 5.9%, 3.3-9.7) than after CVS (total group: mean = 2.3%, 1.2-3.9; randomised subgroup: mean = 1.2%, 0.3-3.5). The gestation at delivery and birthweight of the infants after EA and CVS were similar. In the EA group the incidence of talipes equinovarus (1.63%), was higher than in the CVS group (0.56%), but this difference was not significant.