Assessing Pregnancy Risk

August 2001; Volume 3; 61-62

By Anthony R. Scialli, MD

The assessment of pregnancy risk is made using two general kinds of information. The most appealing kind of information is follow-up data on pregnancies in women exposed to the agent of interest. There are important problems with this kind of information. First, there may not be many pregnancy exposures, resulting in inadequate power to identify an increased risk of an adverse outcome. The incidence of congenital malformations among newborns is 3-5%. Identifying an exposure that increases that incidence by 0.5% would require follow-up on hundreds of exposed pregnancies.

Exposures known to increase the incidence of birth defects in humans affect relatively few structures. For example, valproic acid, an anticonvulsant, increases the incidence of spina bifida from about 0.1% to about 1%.1 The first assessment of valproic acid effects in pregnancy included 12 exposed pregnancies, all of which resulted in normal babies.2 Of course, the problem is that too few exposures were evaluated. Even in collections of cases with several dozen exposed pregnancies, normal outcomes, while reassuring, do not prove safety.

The second potential problem with follow-up studies lies in the quality of the assessment of the child. Trained dysmorphologists find more birth defects than do general pediatricians, and pediatricians find more birth defects than do obstetricians, who may restrict their examination of the baby to counting fingers and toes. Birth certificates, which usually are completed by obstetricians, who are notoriously inaccurate in identifying children with birth defects. Some birth defects also are more likely to be diagnosed as children get older and some maybe diagnosed with different levels of completeness depending on the access of the child to diagnostic equipment such as echocardiography.

The other kind of information used in assessing pregnancy risk comes from experimental animal studies. Pregnant animals, usually rats and rabbits, but sometimes mice, hamsters, guinea pigs, or other species, are given high doses of the compound of interest. A high dose is used to bring out a tendency of the compound to produce abnormal development. It is common, however, for a high dose of any compound to produce illness in the mother animal. When embryo development or survival is impaired under these conditions, it may be difficult to tell if the abnormality was due to a direct effect of the test compound on the embryo, or due to maternal illness. Teratologists have criteria by which they evaluate these possibilities. For example, if the test compound is given at three different doses, the incidence or severity of the abnormality in the embryos would be expected to increase as the dose is increased. Although the presence of abnormal outcome in experimental animal studies may not give rise to a clear estimate of risk to human pregnancy, the lack of adverse pregnancy outcome when animals are given high doses of test compounds generally is reassuring.


1. Bjerkedal T, et al. Valproic acid and spina bifida. Lancet 1982;1:1096.

2. Hiilesmaa VK, et al. Valproic acid during pregnancy. Lancet 1980;1:883.