Soluble l-selectin (sCD62L) plasma concentrations at diagnosis and outcome were investigated in 193 children at first relapse of acute lymphoblastic leukaemia (ALL) after treatment according to the Berlin-Frankfurt-Münster relapsed ALL multicentre trials, ALL-REZ BFM 95 and 96. sCD62L was low (< fifth paediatric reference percentile) in 63 (33%) and high (> 95th percentile) in 36 (19%) children, and was independent of remission duration, sex, BCR-ABL fusion or extramedullary disease. High sCD62L was associated with circulating blasts and T-cell phenotype. More initial adverse events occurred in children with high and low levels of sCD62L (23 out of 99) than in those with normal levels (9 out of 94, P = 0.018). Among 75 worst-prognosis patients (risk groups S3/S4, isolated bone marrow relapse occurring less than 6 months after elective cessation of front-line therapy, or T-cell phenotype with bone marrow involvement), 27 had low sCD62L and decreased event-free survival (EFS) probability (PEFS5 = 0.09 at 5 years) and duration (219 d) compared with normal sCD62L (29 out of 75, PEFS5 = 0.24, 640 d, P = 0.01). Low (44 out of 118), normal (72 out of 118), and high (19 out of 118) sCD62L non-S3/S4 patients fared similarly (average PEFS5 = 0.45, 1369 d; P = 0.5). Low sCD62L may be a marker of malignant blasts replacing normal sCD62L-producing haematopoietic cells. In children with first relapse of ALL and worst prognosis, plasma sCD62L may be useful for risk-adapted stratification.