OBJECTIVE To evaluate prospectively the relative risks and accuracy of first-trimester chorionic villus sampling (CVS) and second-trimester amniocentesis in the genetic evaluation of twin gestations. METHODS Between March 1984 and August 1990, patients presenting for prenatal diagnosis of a twin gestation of less than 12 weeks were offered sampling by either first-trimester CVS or amniocentesis at 16-18 weeks' gestation. Selection was based solely on patient preference and was obtained before ultrasound identification of placental position. Women presenting beyond 12 weeks' gestation were sampled by amniocentesis. Clinical and laboratory outcomes were evaluated. RESULTS Eighty-one women had amniocentesis (nine of whom also had CVS), and 161 women had CVS. All fetuses in both groups were successfully sampled and karyotyped; 85.3% of the amniocentesis patients and 75.8% of the CVS patients were sampled in two or fewer passes (P = not significant). There were three cases of twin-twin villus contamination following CVS; one of these led to incorrect gender assignment because of erroneous laboratory interpretation. Loss of the entire pregnancy from the time of sampling until the 28th week of gestation followed amniocentesis in 2.9% of the cass and CVS in 3.2%. The total fetal loss rates were 9.3% for amniocentesis and 4.9% for CVS (P = not significant). When pregnancies having mosaic or abnormal karyotype results are excluded, the total amniocentesis loss rate remained 9.3% and the CVS loss rate became 3.9% (P < .05). CONCLUSIONS In the hands of experienced operators, CVS is at least as safe and effective as amniocentesis for the prenatal diagnosis of twin gestations.