OBJECTIVE Patients with hyperleukocytic leukaemia were graded according to the severity of symptoms possibly caused by leukostasis to evaluate the effectiveness of therapy and to test the relative contribution of blast type and count of blasts and promyelocytes in the development of leukostasis syndrome. METHODS Ninety-five patients (59 male, 36 female, median age 52 yr) with hyperleukocytic leukaemia [leukocytes above 50 x 10(9)/L, 48 acute myeloid leukaemia (AML), 31 chronic myeloid leukaemia (CML), 13 acute lymphoblastic leukaemia (ALL), three chronic myelomonocytic leukaemia (CMML)] were grouped according to the presence or absence and severity of neurologic, pulmonary and other symptoms into four categories (no, possible, probable and highly probable leukostasis syndrome). Age, white blood count (WBC), haemoglobin, blast count and total of blasts plus promyelocytes of these groups were compared by Mann-Whitney U-test. RESULTS Patients with myeloid leukaemia (AML M1/M2, CML) which scored as highly probable leukostasis showed significantly higher WBC (P = 0.011), lower haemoglobin (P = 0.004), higher peripheral blast counts (P = 0.004) and higher total of peripheral blasts plus promyelocytes (P < 0.001) compared with the lower probability groups. In leukaemia involving the monocytic lineage (AML M4/M5, CMML) no significant differences were found in any of these factors between patients with highly probable leukostasis and the other patients. CONCLUSIONS Our results show that a four-stage clinical grading scale is a valuable tool for analysing hyperleukocytic patient populations and evaluate the effectiveness of therapy more precisely. We further demonstrate that the mechanisms of leukostasis are different in myeloid leukaemia as compared with leukaemia with involvement of the monocytic lineage.