OBJECTIVE--To assess the performance and impact of a two tier neonatal screening programme for cystic fibrosis based on an initial estimation of immunoreactive trypsinogen followed by direct gene analysis. DESIGN--Four year prospective study of two tier screening strategy. First tier: immunoreactive trypsinogen measured in dried blood spot samples from neonates aged 3-5 days. Second tier: direct gene analysis of cystic fibrosis mutations (delta F508, delta I506, G551D, G542X, and R553X) in samples with immunoreactive trypsinogen concentrations in highest 1% and in all neonates with meconium ileus or family history of cystic fibrosis. SETTING--South Australian Neonatal Screening Programme, Adelaide. SUBJECTS--All 88,752 neonates born in South Australia between December 1989 and December 1993. INTERVENTIONS--Neonates with two identifiable mutations were referred directly for clinical assessment and confirmatory sweat test; infants with only one identifiable mutation were recalled for sweat test at age 3-4 weeks. Parents of neonates identified as carriers of cystic fibrosis mutation were counselled and offered genetic testing. MAIN OUTCOME MEASURES--Identification of all children with cystic fibrosis in the screened population. RESULTS--Of 1004 (1.13%) neonates with immunoreactive trypsinogen > or = 99th centile, 912 (90.8%) had no identifiable mutation. 23 neonates were homozygotes or compound heterozygotes; 69 carried one identifiable mutation, of whom six had positive sweat tests. Median age at clinical assessment for the 29 neonates with cystic fibrosis was 3 weeks; six had meconium ileus and two had affected siblings. 63 neonates were identified as carriers of a cystic fibrosis mutation. Extra laboratory costs for measuring immunoreactive trypsinogen and direct gene analysis were $A1.50 per neonate screened. CONCLUSION--This strategy results in early and accurate diagnosis of cystic fibrosis and performs better than screening strategies based on immunoreactive trypsinogen measurement alone.