Assessment of different mathematical models for calculating and expressing the results of coagulation test procedures.

Coagulation activity, expressed as percentage of normal and as clotting time ratio, was estimated in 220 specimens from patients on long-term anticoagulant treatment by 3 different coagulation test procedures, i.e. Thrombotest, Simplastin-A and Normotest. The estimates were calculated from the same determinations. The distribution of percentage values showed a fairly pronounced deviation from normality. After logarithmic transformation, the distribution was normalized, the regression lines between Thrombotest and other tests became parallel, and that between PIVKA-insensitive systems was shifted to a 45 degree line. Logarithmic transformation also stabilized the residual variance. These features make percentage values accessible for treatment according to the standard methods of bioassay statistics. Attempts to normalize the distribution of ratio values by various transformations were unsuccessful. Formal analysis of data revealed a variation in the proportionality of ratio values with the level of estimated coagulation defect. This may restrict the usefulness of the ratio approach. Logarithmic transformation partly reduced the discrepancy.