We conducted a retrospective cohort study to assess the risk of amniocentesis in twin pregnancy for adverse outcomes. The study base consisted of women who had an amniocentesis performed during twin pregnancy and a comparison representative sample of women who carried a twin pregnancy, but did not have invasive prenatal diagnosis. The 227 women in each of the exposed and non-exposed groups were residents of the state of New South Wales, Australia, over the period 1980-92, and were matched on maternal age and period of the infant's birth. Nearly 10% of twin pregnancies among the women having an amniocentesis were affected by a stillbirth, and the stillbirth rate among exposed fetuses (5.3%) was nearly twice as high as among non-exposed fetuses (3.1%). After adjustment for confounding and excluding abnormalities, there was a non-significant elevated relative risk of stillbirth after exposure to amniocentesis. The analysis by type of amniocentesis (with and without methylene blue dye) was limited by small numbers, but the burden of risk was primarily among women who had dye exposure during amniocentesis (relative risk = 3.64, 95% confidence interval = 1.15, 11.48). This increase remained after adjusting for confounding, although the confidence interval was wide. In conclusion, we were unable to establish with certainty whether an increased risk of stillbirth could be ruled out among women who had any type of amniocentesis in twin pregnancy.