Calibration Guidelines Challenge Outlier Practices

Abstract

Thromboplastin, a biological agent important to safe clinical management of patients who receive anticoagulant therapy, is obtained by extraction from mammalian tissues. Because of this natural origin, there is wide variation between successive batches. Therefore, a laboratory experiment must be conducted in order to calibrate each new batch of the material against a standard. In its advice on the statistical estimation of the calibration factor, but without presenting reasons or scientific justification, the World Health Organization uses a simple though crude rule on detecting outliers among the experimental data and deleting them. After a general survey of the nature and cause of outliers, this article offers a critical discussion of the principles of identifying and handling them. It suggests the need for extensive empirical studies before accepting a questionable practice that has ethical implications for patient welfare and scientific integrity.