OBJECTIVE--To determine the feasibility of establishing a system of screening for neuroblastoma. DESIGN--Prospective study of mass screening in four clearly defined geographical areas. SETTING--Four health districts of the Northern region of England. SUBJECTS--20,829 babies aged 6 months, 92% of target population. INTERVENTIONS--Collection of urine on filter paper for analysis of content of homovanillic and vanillylmandelic acid in relation to urinary creatinine concentrations. MAIN OUTCOME MEASURES--Derivation of reference range. Identification of babies with homovanillic or vanillylmandelic acid > 3 SD above the mean (positive cases). Investigation of positive cases for evidence of neuroblastoma. RESULTS--The upper limit of normal (3 SD above the mean) for vanillylmandelic acid was 15 mumol/mmol creatinine and for homovanillic acid 24 mumol/mmol creatinine. Of the 20,829 babies screened, 2537 (12.2%) required a second sample to be taken because the first sample was inadequate. Of these, 527 (2.5%) provided a liquid urine specimen and 10 (0.04%) had positive results for neuroblastoma. Two of them had neuroblastoma (true positives) and eight did not (false positives). A further three children from the cohort were subsequently found to have neuroblastoma; they had raised homovanillic acid or vanillylmandelic acid values, or both, but screened negative at 6 months. CONCLUSIONS--Screening for neuroblastoma is possible in the health care system of the United Kingdom. Evaluation of the efficacy of screening in reducing the mortality from neuroblastoma requires a controlled trial.