Seventy-four patients from a prospective randomized trial comparing autologous bone marrow (ABM) versus blood stem cell (BSC) transplantation after high-dose chemotherapy for intermediate and high grade non-Hodgkin's lymphoma (NHL) were studied for the presence of residual lymphoma prior to transplantation. Pre-transplant bone marrow (BM), peripheral blood (PB) and the ABM or BSC harvest were studied by molecular assays immediately after collection and at weekly intervals after the initiation of in vitro cultures. B-NHLs with t(14:18) at the major breakpoint region (mbr) were monitored by detecting cells with the translocation. Other B-NHLs were monitored with tumour-specific primers and probes to the immunoglobulin heavy chain (IgH) gene complementary determining region (CDR) III. T-NHLs were similarly monitored using the T-cell receptor gamma chain gene V-J junctional region as the tumour-specific marker. Of the 74 patients, seven did not have adequate tumour biopsies for molecular characterization. Of the remaining 67 cases, 35 had identifiable markers for follow-up studies and 20/35 cases (52%) had tumour cells detected in either the pretransplant BM/PB samples or the ABM/BSC harvest. Residual tumours were detected at a high frequency in T-NHL (100%) and t(14;18)+ B-NHL (86%) but at a lower frequency in B-NHLs without t(14,18) (44%). In five cases, one or more of the samples were initially negative for residual lymphoma but became positive after a period of culture; additional studies confirmed that in vitro culture enhanced the sensitivity of tumour detection in about half of these samples. Molecular assay for minimal residual disease can be performed in the setting of multicentre prospective clinical trials. The substantial frequency of failure of obtaining tumour-specific IgH CDRIII sequences in paraffin-embedded B-NHLs argues for the storage of frozen tumour samples for possible molecular studies.