We compare two protocols for newborn screening for cystic fibrosis (CF). The first uses the immunoreactive trypsinogen (IRT) assay with a cutoff of > or = 180 ng/ml and a sweat test to identify CF patients. The second uses the IRT assay with a 100 ng/ml cutoff in conjunction with direct analysis for the delta F508 CF transmembrane conductance regulator (CFTR) mutation in a two-tiered (i.e., IRT/DNA) protocol, followed by a sweat test. We screened 220,865 newborns from Wisconsin for CF, using the IRT protocol identifying 369 infants with an elevated IRT, of whom 46 were found to have CF. Another 7 CF patients were identified who had a false-negative IRT level. The CF incidence in the white population was 1 in 3,431 (carrier incidence of 1 in 30). The IRT protocol had a sensitivity of 87% and a positive predictive value of 12.5%. We subsequently used the IRT/DNA protocol to screen 21,258 infants. Of 518 infants with an IRT level > or = 100 ng/ml, 24 carried at least one copy of the delta F508 CFTR mutation, and 4 of these infants were found to have CF, yielding a positive predictive value for this protocol of 16.7%. Direct comparison of the positive predictive value of the two protocols is not valid, because of the different populations screened. However, had the IRT protocol been used on the IRT/DNA cohort, 50 infants, including the 4 with CF, would have received sweat tests, yielding a positive predictive value of 8%. Because of the small sample size, this positive predictive value is not significantly different from that obtained for the IRT/DNA test. However, from a practical point of view the IRT/DNA approach does decrease considerably the number of sweat tests that must be undertaken. The number of false positives for the IRT protocol (46 in 21,258) is increased significantly compared with that for the IRT/DNA approach (20 in 21,258; P < .001). The incidence of delta F508 carriers detected in cohorts with an elevated IRT level was increased compared with the incidence in the general population. The direct costs for the IRT/DNA approach (100 ng/ml) were $11,374 per CF patient detected, compared with $10,187 per CF patient detected for the IRT protocol. Therefore, we conclude that the IRT/DNA approach to CF newborn screening decreases the number of false-positive subjects contacted, without a significant increase in cost.