Distinguishing true precursor lesions on the basis of clinical or histological features alone is unreliable but is important so that appropriate intervention can be instigated. Preliminary studies have shown that a microsatellite assay may provide important new prognostic information. To build on these observations, we have performed a case-control study to establish whether we can be confident about incorporating this new information into clinical practice. We have determined the frequency of allelic imbalance (AI) within key chromosomal regions, by matching 39 cases with dysplastic oral lesions that developed a tumor on the same side of the mouth, for as many variables as possible, with controls presenting with similar lesions that did not progress to malignancy when followed for the same period. Our findings confirm that the group that developed tumor had precursor lesions that harbor AI at more loci (P = 0.002). However, no consistent patterns of AI were associated with the three grades of dysplasia: mild, moderate, and severe. One-third of the tumors developed at the same site as the dysplastic lesion and two-thirds at a different site, which revealed that the presence of these aberrations in a dysplastic lesion provided information about the risk of malignant change within a larger field. This suggests that the process of field cancerization is more widespread than previously recognized. On the basis of these findings, we advocate complete excision of all suspicious areas that show AI at two or more key loci, regardless of the degree of dysplasia. However, because the remaining mucosa is also "at risk," these cases should also be targeted to receive dietary advice and chemoprevention, to minimize their risk of tumor formation at a distant site.