The accurate identification of small numbers of leukemic cells in the cerebrospinal fluid (CSF) presents a diagnostic problem in the treatment of acute lymphoblastic leukemia (ALL). We demonstrated that soluble CD9 antigen was shed into CSF obtained from children with ALL, using enzyme-linked immunosorbent assay (ELISA), which used the activity of CD9 antigen to bind the Ricinus communis agglutinin (RCA1) and a monoclonal antibody, SJ-9A4, simultaneously. Using RCA1/SJ-9A4 ELISA, CD9 antigen was detectable in CSF but not in plasma from 12 cases of CD9+ ALL in central nervous system (CNS) relapse. However, CD9 antigen was not released into CSF from 11 cases of CD9- ALL with CNS involvement, 136 cases of CD9+ ALL in complete remission (CR), 29 cases of CD9- ALL in CR, or 21 cases of aseptic meningitis. Interestingly, the levels of CD9 antigen were elevated in CSF from 7 of 10 CD9+ ALL patients without cytologically proven CNS involvement at diagnosis, with subsequent return to undetectable levels after initial induction chemotherapy was begun. In addition, sequential analysis of CSF from a 5-year-old boy with CD9+ ALL in CNS relapse showed that levels of CD9 antigen correlated well with the number of leukemic cells in CSF. Serial quantitative analysis of CD9 antigen in CSF could be useful to detect the proliferation of residual leukemic cells before the clinical manifestation.