Are the common reference intervals truly common? Case studies on stratifying biochemical reference data by countries using two partitioning methods

• Published in: Scandinavian journal of clinical and laboratory investigation Vol. 64; no. 4; pp. 407 - 430
• Main Author: Lahti, A.
• Format: Journal Article
• Language: English
• Published: England Informa UK Ltd 01-06-2004; Taylor & Francis
• Subjects: Alkaline phosphatase (ALP); Chemistry, Clinical - standards; Chemistry, Clinical - statistics & numerical data; Clinical Medicine - standards; Clinical Medicine - statistics & numerical data; creatine kinase (CK); Data Interpretation, Statistical; Europe; gamma-glutamyl transpeptidase (GGT); Humans; International Cooperation; Laboratories, Hospital - standards; partitioning into several subgroups; prevalence; reference interval; reference limit; reference value; Reference Values; Statistical Distributions; stratification; tied observation; Europe
• ISSN: 0036-5513; 1502-7686
• DOI: 10.1080/00365510410006027

The Harris-Boyd method, recommended for partitioning biochemical reference data into subgroups by the NCCLS, and a recently proposed new method for partitioning were compared in three case studies concerning stratification by countries (Denmark, Finland, Norway, and Sweden) of reference data collected in the Nordic Reference Interval Project (NORIP) for the enzymes alkaline phosphatase (ALP), creatine kinase (CK), and -glutamyl transpeptidase (GGT). The new method is based on direct estimation of the proportions of two subgroups outside the reference limits of the combined distribution, while the Harris-Boyd method uses easy-to-calculate test parameters as correlates for these proportions. The decisions on partitioning suggested by the Harris-Boyd method deviated from those obtained by using the new method for each of the three enzymes when considering pair-wise partitioning tests. The reasons for the poor performance, as it seems to be, of the Harris-Boyd method were discussed. Stratification of reference data into more than two subgroups was considered as both a theoretical problem and a practical one, using the four country-specific distributions for each enzyme as illustration. Neither the Harris-Boyd method nor the new method seems ideal to solve the partitioning problem in the case of several subgroups. The results obtained by using prevalence-adjusted values for the proportions seemed, however, to warrant the conclusion to be made that there are no major differences in terms of the partitioning criteria between the levels of each of the three enzymes in the four countries. Because these three enzymes include those two tests (CK, GGT), which in the preliminary analyses of the project data had shown largest variation between countries, the tentative conclusion was drawn that application of common reference intervals in the Nordic countries is feasible, not only for the three enzymes examined in the present study but for all of the tests involved in the NORIP project.