BACKGROUND: Distinguishing mild and moderate epithelial dysplasia is difficult and controversial. We demonstrated earlier that silver-stainable nucleolar organizer region (AgNOR) is a simple, inexpensive and quantifiable method which can standardize this distinction reliably. OBJECTIVE: To establish a cut-point mean AgNOR count for epithelial dysplasia validated by bootstrapping. METHODS: One hundred oral leukoplakia biopsies at Dr R. Ahmed Dental College, Kolkata, India, examined for epithelial dysplasia using standard method (gold standard) were assessed for their mean AgNOR counts. A cut-point was selected employing receiver operating characteristic (ROC) curve. This estimate was stabilized by 10 000 resamples (with replacement) using parametric and non-parametric bootstrap, a Monte Carlo statistical method, corrected for bias to estimate standard errors (SE) of cut-point. RESULTS: Area under the curve (AUC) was 74%; non-parametric method suggested mean AgNOR cut-point = 2.42 AgNORs/nucleus; parametric method suggested cut-point = 2.57. The cut-point 95% confidence intervals (CIs) (bias corrected and accelerated) from parametric method (2.31; 2.66) was lower than non-parametric (2.43; 2.82). CONCLUSION: Applying a conservative paradigm, taking the lowest of all bootstrap 95% CIs, we suggest that mean AgNOR count of 2.3 be used as a cut-point for distinguishing between mild and moderate dysplasia. This strategy will provide an inexpensive, meaningful, definitive, reproducible and consistent diagnostic test for epithelial dysplasia.