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.

Keywords

Percentile methods Normalization Generalized lab norms Standardization Concordance

Get full access to this article

View all access options for this article.

References

Chuang-Stein

Laboratory data in clinical trials: A statistician's perspective. Drug Inf J. 1998;19:167–177.

Chuang-Stein

Summarizing laboratory data with different reference ranges in multicenter clinical trials

Drug. Inf J. 1992;26:77–84.

Sogliero-Gilbert

Mosher

Zubkoff

A procedure for the simplification and assessment of lab parameters in clinical trials. Drug Inf J. 1986; 20:279–296.

Huang

Brunelle

A nonparametric method for combining multilaboratory data. Drug Inf J. 2002;36:395–406.

Chuang-Stein

Some issues concerning the normalization of laboratory data based on reference ranges. Drug Inf J. 2001;35:153–156.

Herrera

The precision of percentiles in establishing normal limits in medicine. J Lab Clin Med. 1958;52:34–42.

Brunden

Clark

Sutter

A general method of determining normal ranges applied to blood values for dogs. Am J Clin Pathol. 1970;53: 332–339.

Trost

Brailey

Hoffman

Probability-based construction of reference ranges for ratios of Log-Gaussian analytes: An example from automated leukocyte counts. Am J Clin Pathol. 2002;117:851–856.