Prenatal-toxic risk estimation for the alkylating model compound methylnitrosourea (MNU) was performed using different procedures. Risk of low doses was estimated using linear extrapolation to zero (estimated ED0.1%: 0.1 mg/kg body wt MNU) as well as extrapolation by probit analysis based on a dose-response study (estimated ED0.1%: 1.6 mg/kg body wt). Furthermore, a "virtually safe dose" was established by means of the NOAEL risk factor approach (e.g., factor 30:0.03 mg MNU per kg body wt). In previous studies in murine embryos using MNU, we combined dose-response data and DNA adduct rate measurements and deduced that O6-methylguanine is a suitable variable for molecular dosimetry. In a tentative approach, we estimated the teratogenic risk of low doses based on the adduct rates of O6-methylguanine in the DNA of the embryos. It is concluded that in the case of steep dose-response relationships, which are typical for the majority of teratogenic effects, the NOAEL risk factor approach is more conservative than extrapolation based on probit analysis. Risk estimation using dosimetry with this model compound yields estimated incidences similar to linear extrapolation.