Generalized Lab Norms for Standardizing Data from Multiple Laboratories

Abstract

Data from multiple laboratories utilizing different units of measurements and different normal reference ranges present challenges when pooling the data to obtain population summary statistics. This manuscript provides an algorithm for obtaining statistical summaries. First, standardization of measurement units by conversion to SI units is proposed. Conversion factors for key clinical analytes focused on oncology are provided. Second, it is proposed that generalized lab norms (GLN) be created for a “phantom” laboratory on which to standardize multiple sources of laboratory data. The percentile method is proposed for developing GLN. Phantom laboratory GLN estimates are provided for standardization. The GLN were applied to data from an oncology trial that utilized 30 local laboratories. Proportions of samples misclassified normal/abnormal were mostly ≤10% (Range: 0% to 14.6%) for different analytes. The National Cancer Institute Common Toxicity Criteria grades based on GLN were generally consistent with grades utilizing individual laboratory normal reference ranges. The Gamma values for concordance applying GLN versus utilizing individual lab normal reference values ranged from 0.98 to 1.0.