Cardiopulmonary bypass surgery induces an inflammatory reaction among others by activation of granulocytes. Leukocyte filtration has been shown to reduce the postoperative morbidity mediated by activated granulocytes. However, little is known about the mechanism of filter-leukocyte interaction. This study examines whether a leukocyte filter removes activated granulocytes or a general leukocyte population. Eleven patients undergoing cardiopulmonary bypass surgery were included in this study. Leukocyte filtration was achieved before the reperfusion phase with a Pall non-woven polyester filter located at the venous side of the heart-lung machine. After filtration, the trapped granulocytes inside the filter were examined morphologically with light and scanning electron microscopy and immunologically by CD45RO antigen binding to the filter material. Furthermore, leukocyte release markers were measured to determine whether cells were activated during filtration. Microscopic evaluation revealed 84% granulocytes and 14% lymphocytes trapped in the filter, compared with 78% granulocytes and 22% lymphocytes in the blood before filtration. Granulocytes were trapped significantly more in the first blood contact layer of the filter material than in the middle layer and last layer, whereas lymphocytes trapped slightly more in the middle layer. The near maximum level of CD45RO expression was measured on granulocytes trapped inside the filter material, whereas CD2 and CD19 measured on lymphocytes were bound to a minor extent. Beta-glucuronidase concentration did not increase after filtration, suggesting the absence of activation of granulocytes by filtration. A leukocyte filter made of non-woven polyester material removes the activated granulocytes rather than leukocytes at random. This implies that this particular type of leukocyte removal filter is suitable for use in cardiopulmonary bypass patients whose granulocytes in the circulation are activated. Furthermore, measurement of activated granulocytes instead of total leukocyte count is likely preferable for functional assessment of leukocyte removal devices.