A sensitive and specific quantitative real-time polymerase chain reaction method, involving three rounds of amplification with two allele-specific oligonucleotide primers directed against an rearrangement, was developed to quantify minimal residual disease (MRD) in B-lineage acute lymphoblastic leukemia (ALL). For a single sample containing 10 microg of good quality DNA, MRD was quantifiable down to approximately 10(-6), which is at least 1 log more sensitive than current methods. Nonspecific amplification was rarely observed. The standard deviation of laboratory estimations was 0.32 log units at moderate or high levels of MRD, but increased markedly as the level of MRD and the number of intact marker gene rearrangements in the sample fell. In 23 children with ALL studied after induction therapy, the mean MRD level was 1.6 x 10(-5) and levels ranged from 1.5 x 10(-2) to less than 10(-7). Comparisons with the conventional one-round quantitative polymerase chain reaction method on 29 samples from another 24 children who received treatment resulted in concordant results for 22 samples and discordant results for seven samples. The sensitivity and specificity of the method are due to the use of nested polymerase chain reaction, one segment-specific and two allele-specific oligonucleotide primers, and the use of a large amount of good quality DNA. This method may improve MRD-based decisions on treatment for ALL patients, and the principles should be applicable to DNA-based MRD measurements in other disorders.