SummaryDuring the collection of samples for plasma β-thromboglobulin (β-TG) determination, it is well established that artificially high values can be observed due to in-vitro release. To estimate the reliability of a single β-TG measurement, blood samples were collected simultaneously from both arms on two separate occasions in 56 diabetic patients selected for a clinical trial. From each arm, blood was taken into two tubes containing an anticoagulant mixture with (tube A) and without (tube B) PGE!. The overall mean value of B-TG in tube B was 1.14 times higher than in tube A (p <0.01). The markedly large between-arms variation accounted for the most part of within-subject variation in both tubes and was significantly greater in tube B than in tube A. Based on the difference between B-TG values from both arms, the number of subjects with artifically high B-TG values was significantly higher in tube B than in tube A on each occasion (overall rate: 28% and 14% respectively). Estimate of between-occasions variation showed that B-TG levels were relatively stable for each subject between two occasions in each tube. It is concluded that the use of PGEi decreases falsely high B-TG levels, but a single measurement of B-TG does not provide a reliable estimate of the true B-TG value in vivo.