To evaluate accurately the current performance of filtration, the French Produits Sanguins Labiles study group, composed of 21 transfusion teams, conducted a large-scale 6-month study involving over 1400 filtrations and 3000 controls. Some 745 standard red cell concentrates (RBC concentrates) and 690 concentrates previously white cell (WBC)-reduced by removal of buffy coat (BC-poor RBC concentrates) were filtered using six commercially available filters: at least 170 results were collected per filter, spread among a minimum of three teams. Prefiltration controls show that the removal (manual and automated) of the buffy coat results in an initial WBC reduction of approximately 63 percent, along with a hemoglobin loss of 4 g (7%). After filtration, residual WBCs were counted in the Nageotte manual counting chamber. The reliability of this counting method, which is simple and adapted to low WBC concentrations, was characterized in this study by a 25-percent coefficient of variation (CV) for a concentration of 2.5 WBCs per microL (i.e, 0.6 x 10(6) WBCs/filtered unit). The analysis of the results shows that, for five of six filters (1 filter was excluded), the postfiltration median value of residual WBCs was 1.1 x 10(6) in filtered RBC concentrates (n = 590), whereas it was 0.34 x 10(6) in filtered BC-poor RBC concentrates (n = 581). The difference is significant (p less than 10(-8), Wilcoxon test). Hemoglobin loss due to filtration varies according to the filter, from 5.7 +/- 2.2 to 17.3 +/- 2.5 g.(ABSTRACT TRUNCATED AT 250 WORDS)