Absence of need for amniocentesis in patients with elevated levels of maternal serum alpha-fetoprotein and normal ultrasonographic examinations.

Background: Recent improvements in the accuracy of sonographic diagnosis of neural-tube and ventral-wall defects have raised a question about the wisdom of routinely offering amniocentesis to women who have elevated levels of maternal serum alpha-fetoprotein with a structurally normal fetus as determined by ultrasonography.

Methods: We reviewed the ultrasound findings in 51 consecutive fetuses with spina bifida, encephalocele, gastroschisis, or omphalocele that were delivered or aborted at a single hospital, to estimate the sensitivity of ultrasonography for these diagnoses. In all cases, the mothers had undergone prenatal sonography at one facility between 16 and 24 weeks after the last menstrual period. We used these data to calculate the probability of an affected fetus in a women with a given level of maternal serum alpha-fetoprotein and a normal sonogram.

Results: These four types of anomalies were correctly identified in all 51 cases, yielding a sensitivity of 100 percent (95 percent confidence interval, 94 to 100 percent). Using the lower limit of this confidence interval, we calculated that the probability of an affected fetus ranges from 0.01 to 0.15 percent for maternal serum alpha-fetoprotein levels ranging from 2.0 to 3.5 times the median, respectively.

Conclusions: This level of risk is less than the reported risk of abortion due to amniocentesis and may lead some women with elevated levels of alpha-fetoprotein to decide not to proceed with amniocentesis.