Ten years ago, we made an incidental flow cytometric observation while immunophenotyping biopsy and marrow samples from children suspected to have leukemia/non-Hodgkin's lymphoma, but were subsequently diagnosed with neuroblastoma. The samples contained neoplastic CD45(-) cells that had an extremely bright CD56(+) (beyond the fourth decade on a four-decade scale) population distinguishable from CD45(+)CD56(usual density+) natural killer lymphocytes as well as other CD45(-)CD56(usual density+) nonhematopoietic tumors such as small cell carcinoma or melanoma. Following the "rare event" philosophy of selecting one negative and two positive antigens, we initially tried a "cocktail" of CD45(-)CD56(very bright+) neuron-specific enolase (NSE)(cytoplasmic+). We later modified the procedure to a more clinically applicable "lysed whole blood" CD45(-)CD56(very bright+) ganglioside GD2(+) cocktail to improve turnaround time (eliminating the cell permeabilization step for cytoplasmic NSE analysis), specificity, and sensitivity of the assay. A total of 123 marrow/tissue/fluid samples were analyzed by the various forms of the assay. Clearly interpretable samples had an 83% specificity and a 100% sensitivity. The three-color GD2 assay has successfully detected cells in marrow samples to a level of 0.002% (1 per 10(5) cells) using patient samples (not artificially "spiked" material). We added CD81 expression of the neuroblastoma cells as a fourth color and now use this rare event clinical test to help stage and monitor all patients with neuroblastoma. Copyright 2002 Wiley-Liss, Inc.