The Childrens Cancer Study Group conducted a case-control study designed to assess in utero and postnatal exposures in children with acute nonlymphoblastic leukemia (ANLL). Analyses were performed for reported maternal use of medications and drugs in the year preceding and during the index pregnancy of the 204 case-control pairs. An 11-fold risk (P = 0.003) was found for maternal use of mind-altering drugs just prior to or during the index pregnancy. Compared with ANLL cases not exposed to marijuana, exposed cases were significantly younger at diagnosis of ANLL (P less than 0.01) and were more often of the myelomonocytic and monocytic subtypes (P less than 0.01). Use of antinausea medication for more than 11 weeks was also associated with a significantly elevated relative risk of 2.81 and a dose-response relationship was noted (P = 0.05 for trend). These results suggest that maternal drug use of marijuana may have an etiologic role in childhood ANLL and may be specific for morphologically defined subgroups.