Background: We compared the international flow reference method (IRM) platelet counts with those obtained from CellDyn Sapphire (impedance and optical counts), LH750 (impedance counts), and the flowcytometry based ReaPanThrombo Immunoplatelet method (ReaMetrix). We further evaluated the degree of agreement of above methods with the IRM at the transfusion thresholds of 10 x 10(9) l(-1) and 20 x 10(9) l(-1).
Methods: A total of 104 thrombocytopenic blood samples with platelet count of <50 x 10(9) l(-1) were selected for the study. All samples were tested in parallel by various methods within 6 h of blood collection.
Results: For bias estimation, a Bland-Altman analysis was done by taking the IRM as the standard method. The bias for CDS-I counts was +6.505 x 10(9) l(-1) (95% LA -2.110 to +15.122), for CDS-O counts the bias was -3.779 x 10(9) l(-1) (95% LA -8.950 to +1.392), for LH750 the bias was +0.111 x 10(9) l(-1) (95% LA -5.862 to +6.084) and that for ReaMetrix was -1.602 x 10(9) l(-1) (95% LA -7.400 to +4.194). The LH750 had the least average bias and it overestimated platelet counts marginally. The ReaMetrix method showed the highest degree of agreement with the IRM, at both the threshold points with a K value of 0.960 (threshold < or = 10 x 10(9) l(-1)) and 0.923 (threshold < or = 20 x 10(9) l(-1)).
Conclusions: Impedance platelet counts from LH750 were more accurate than optical methods in thrombocytopenic patients. ReaMetrix immunoplatelet counts show the maximum degree of agreement with the IRM at clinically relevant transfusion thresholds. We conclude that as current platelet transfusion thresholds are based on results of automated hematology analyzer methods, the true thresholds may be determined using the IRM and CD41/61 based single-platform immunoplatelet methods.
(c) 2010 Clinical Cytometry Society.