Effect of prothrombin time outlier exclusion rules on ISI calibration in multicentre studies.

International Sensitivity Index (ISI) calibration is based on prothrombin time (PT) determinations in fresh plasma samples of healthy (normal) individuals and patients treated with vitamin K-antagonists (VKA). The ISI is calculated from the slope of the orthogonal regression line of the log(PT) results. The ISI calibration model is based on the assumption that the mean logarithms of the PT's of the normals are found on the orthogonal regression line derived using patients' samples. According to World Health Organization (WHO) guidelines, patients' samples with International Normalized Ratio (INR) beyond the 1.5 to 4.5 interval shall be excluded for ISI calibration. According to the WHO guidelines, outlier samples are defined as those at a perpendicular distance from the orthogonal regression line greater than 3 residual standard deviations, and shall be excluded as well. The purpose of the present study was to assess the effect of sample exclusion on ISI calibration, using the data base of three historic multicenter studies performed in 1990, 1995 and 2005, respectively. Various rules for sample exclusion were tried. In comparison to calibration without any exclusion, between-laboratory variation of the ISI was slightly reduced by sample exclusion using the WHO rule. Furthermore, the adequacy of the ISI calibration model was improved. The effect of the WHO sample exclusion rule on the mean value of the ISI of the current International Standards for thromboplastins was not greater than approximately 1%. It is concluded that the WHO rule for sample exclusion is appropriate for reliable ISI calibration.