Of 2882 women allocated to either transabdominal CVS (TA) or transcervical CVS (TC) at two large obstetric centres in Denmark, 2707 had blood samples drawn before and 30 min after CVS for maternal serum-alpha-fetoprotein (MSAFP) measurement. 2535 of these women had cytogenetically normal pregnancies and 2091 of them went on to have samples drawn at the 18-20 week follow-up. Post-procedure MSAFP values were correlated to the biopsy method used, with mean MSAFP values significantly higher after TA than TC, 33 and 15 kU/l, respectively (P < 0.001). Following TA procedures, 18 per cent of cases had feto-maternal transfusion higher than 0.1 ml; this occurred in only 5 per cent of TC cases. MSAFP levels were associated with spontaneous fetal loss in the TA group but not in the TC group. TC, however, was followed by more losses than TA. The post-CVS MSAFP value was positively correlated with the amount of villi aspirated. The difference in post-procedure elevation in MSAFP 30 min later (average 18 kU/l higher for TA than for TC) was not reflected in raised levels at the 18-20 week follow-up. Study medians at mid-trimester did not differ from reference group medians established from a group of singleton pregnancies with sonographically determined gestational age who did not experience invasive procedures and delivered normal infants. Our findings suggest that CVS does not compromise mid-trimester MSAFP for screening for neural tube defects (NTDs). Extremely high mid-trimester MSAFP values in the TC group could predict imminent loss.