Summary
In order to evaluate the comparability of data obtained with various available kits
for the immunological determination of PAI-1 antigen in plasma and in order to investigate
the underlying cause of observed differences, e. g. problems of specificity or of
proper calibration of the provided standard, a multicenter study was organised in
the framework of the Subcommittee of Fibrinolysis of the Scientific and Standardization
Committee.
Eight different plasma samples were distributed among 16 laboratories: a pooled normal
plasma, NIBSC 87/512, PAI-1 antigen depleted plasma, PAI-1 depleted plasma supplemented
with 59 ng/ml active PAI-1 and four different individual plasma samples. A considerable
variation in absolute values is observed between the various kits, e.g. in pooled
normal plasma a value is found ranging between 7.4 and 28 ng/ml. Harmonization of
all data relative to the PAI-l-depleted plasma supplemented with an exact amount of
active PAI-1 (59 ng/ml), followed by a statistical analysis using a two way analysis
of variance, revealed that 6 out of 7 kits yielded values that were not significantly
different with coefficients of variation around 30%. Correlations between the values
obtained with these kits yielded slopes between 0.75 and 1.44 with correlation coefficients
between 0.973 and 0.999. Values obtained with one kit appeared to be significantly
different (even after harmonization) from the other kits (p <0.001 to p <0.05). Comparison of PAI-1 antigen with the PAI activity values in the analysed
samples suggests that one kit may deal with a problem of a difference in reactivity
between active and latent PAI-1.
In conclusion 6 different kits yield results that, only after harmonization, are comparable
and do correlate very well. Thus proper calibration of the provided standards may
solve the majority of the problems observed with PAI-1 antigen kits. Differential
specificity towards different conformational forms of PAI-1, as a consequence of the
use of different monoclonal antibodies, is a minor, but a potential problem, that
needs to be considered when comparing data obtained with various methods.
