Pooling of Laboratory Safety Data in Multicenter Studies

The pooling of laboratory data obtained in multicenter studies is highly desirable in order to increase the efficiency of drug safety monitoring. Since, in general, different centers use different “normal” ranges for their data, it is proposed that the pooling is accomplished by transformation of the data of each center into percent-scaled values with 0% for the lower and 100% for the upper “normal” bound specific for each variable and each center. The data thus transformed will then be used to construct an all-center multivariate non-parametric tolerance region based on the percent values “before treatment.” Each multivariate patient value “after treatment” is checked by a distance function whether it is still tolerable or not. The non-parametric multivariate method yields an objective decision instrument for the safety of the drug(s) under evaluation.