isolation, but the colour of the main bag differs from the colour of the segments that remain sterile, presumably because of the tiny initial inoculum per unit volume.12 Whether this sign will turn out to be useful in screening blood remains to be seen. The Centers for Disease Control in Atlanta have suggested testing each unit over 25 days old for the presence ofbacteria before transfusion.7 This would entail breaching the closed system and conventional culture would result in unacceptable delays. The Food and Drug Administration has not found currently available rapid screening tests for bacteria such as Gram and acridine orange staining and endotoxin assays to be reliable in tests on "spiked" units.8 A rapid test based on the presence of bacterial 16S ribosomal RNA is currently undergoing trials with encouraging results (K Piper et al, 92nd general meeting of American Society for Microbiology, New Orleans, 26-30 May 1992). In the meantime, it should be emphasised that the reaction to contaminated blood may clinically resemble that to incompatible blood. Standard microbiological investigation with Gram staining and culture of donor blood (or platelets) at 20°C and 37°C is indicated after any severe transfusion reaction with no obvious serological cause. After transfusion has been stopped fever and hypotension should prompt antimicrobial treatment with drugs likely to act against most Gram negative contaminating organisms (for example, iprofloxacin or ceftazidime). Treatment with anti-endotoxin seems logical in cases of strongly suspected or confirmed bacterial contamination of blood or platelet transfusion, although there have been no reports of this application. To enable the delineation of the extent of the problem any suspected cases should be fully investigated and reported to the