A total of 2931 women randomized to either transabdominal CVS, transcervical CVS, or amniocentesis were studied. Unless intended or unintended abortion had occurred, they had completed up to 28 weeks of pregnancy. No significant difference was seen between total fetal loss in the transabdominal CVS group and the amniocentesis group (6.5 and 6.8 per cent, respectively, SE difference = 0.92 per cent, p = 0.01). The total fetal loss in the transcervical CVS group was 10.1 per cent. After pooling our data with data from the Canadian randomized study and the American non-randomized study, the difference in risk between transcervical CVS and amniocentesis was 1.8 per cent (SE difference = 0.64 per cent, p = 0.8). When the number of failed procedures and those cases evaluated as unfeasible for the assigned method--for anatomical reasons--are compared, the overall sampling efficacy is poorer transcervically than transabdominally.