Hepatoblastoma, the commonest primary malignant liver tumor in infants and children, is usually associated with elevated serum alpha-fetoprotein (AFP) levels. The authors sought to determine if AFP levels can be used to modify treatment, thereby avoiding the wait for formal imaging studies and prolonged suboptimal treatment and limiting the use of effective but toxic chemotherapy. From April 1984 to December 1997, 8 children were diagnosed with AFP-secreting hepatoblastoma. Serum AFP levels were measured weekly. If AFP levels failed to improve, or increased on at least 2 successive examinations, the chemotherapy protocol was changed. When an excellent response was achieved, less toxic chemotherapy was substituted. Six patients (75%) were disease-free for at least 2 years, some with high-risk or metastatic disease. Two patients died. Six of the 7 nonmetastatic patients (86%) remain disease-free (only one had a resectable tumor). Chemotherapy changes resulted in reduced AFP levels in 7 patients. This study supports the use of AFP monitoring to modify treatment in hepatoblastoma responding to therapy with less toxic drugs and the use of nonstandard therapy when suboptimal responses are obtained.