A global multicenter study on reference values: 1. Assessment of methods for derivation and comparison of reference intervals

- A global multicenter study on reference values (RVs) was conducted targeting 50 analytes using a common protocol.
- A panel of sera were measured in common for alignment of RVs across 12 countries.
- Need for secondary exclusion of RVs was obvious in analytes which reflect nutritional abnormalities.
- Validity of parametric method was confirmed by almost invariable success in Gaussian transformation
- Nearly half of the analytes showed variable degrees of between-country differences.

ObjectivesThe IFCC Committee on Reference Intervals and Decision Limits coordinated a global multicenter study on reference values (RVs) to explore rational and harmonizable procedures for derivation of reference intervals (RIs) and investigate the feasibility of sharing RIs through evaluation of sources of variation of RVs on a global scale.MethodsFor the common protocol, rather lenient criteria for reference individuals were adopted to facilitate harmonized recruitment with planned use of the latent abnormal values exclusion (LAVE) method. As of July 2015, 12 countries had completed their study with total recruitment of 13,386 healthy adults. 25 analytes were measured chemically and 25 immunologically. A serum panel with assigned values was measured by all laboratories. RIs were derived by parametric and nonparametric methods.ResultsThe effect of LAVE methods is prominent in analytes which reflect nutritional status, inflammation and muscular exertion, indicating that inappropriate results are frequent in any country. The validity of the parametric method was confirmed by the presence of analyte-specific distribution patterns and successful Gaussian transformation using the modified Box-Cox formula in all countries. After successful alignment of RVs based on the panel test results, nearly half the analytes showed variable degrees of between-country differences. This finding, however, requires confirmation after adjusting for BMI and other sources of variation. The results are reported in the second part of this paper.ConclusionThe collaborative study enabled us to evaluate rational methods for deriving RIs and comparing the RVs based on real-world datasets obtained in a harmonized manner.