Salvage therapy for relapsed large-cell lymphoma (LCL) is more effective in patients with minimal disease, suggesting that early detection of relapse might increase the chance of long-term survival. To determine whether current follow-up procedures are effective in identifying preclinical disease, we analyzed patterns of relapse in 139 LCL patients who achieved a complete remission (CR) with high/moderate-dose methotrexate with leucovorin, bleomycin, doxorubicin, cyclophosphamide, vincristine, and dexamethasone (M/m-BACOD). The timing and results of all posttreatment follow-up tests were examined in the 36 patients who relapsed from complete remission (CR) and 46 controls who remain in CR. Despite conscientious posttreatment follow-up, only two of the 36 relapses (6%) were detected before the development of symptoms. Sixty-seven percent of patients relapsed in new disease sites (42% in new and old sites, and 25% in new sites only). Consistent with this observation, the tests most sensitive to clinical relapse were those not targeted to specific sites of disease: gallium scan (sensitivity, 90%), physical examination (80%), and lactate dehydrogenase (LDH) (65%). Of screening tests performed, only LDH was successful in detecting preclinical relapse, with a sensitivity of 42% and specificity of 85% for impending symptomatic relapse. These results indicate that conventional screening was ineffective in detecting preclinical relapse in LCL patients. We recommend prospective evaluation of a strategy that (1) screens with a frequency appropriate to a patient's risk of relapse, (2) uses sensitive test(s) not targeted to specific sites, and (3) limits aggressive screening to those high-risk patients eligible for potentially curative salvage therapy.