No abstract

A database with reliable information to derive definitive analytical quality specifications for a large number of clinical laboratory tests was prepared in this work. This was achieved by comparing and correlating descriptive data and relevant observations with the biological variation information, an approach that had not been used in the previous efforts of this type. The material compiled in the database was obtained from published articles referenced in BIOS, CURRENT CONTENTS, EMBASE and MEDLINE using "biological variation & laboratory medicine" as key words, as well as books and doctoral theses provided by their authors. The database covers 316 quantities and reviews 191 articles, fewer than 10 of which had to be rejected. The within- and between-subject coefficients of variation and the subsequent desirable quality specifications for precision, bias and total error for all the quantities accepted are presented. Sex-related stratification of results was justified for only four quantities and, in these cases, quality specifications were derived from the group with lower within-subject variation. For certain quantities, biological variation in pathological states was higher than in the healthy state. In these cases, quality specifications were derived only from the healthy population (most stringent). Several quantities (particularly hormones) have been treated in very few articles and the results found are highly discrepant. Therefore, professionals in laboratory medicine should be strongly encouraged to study the quantities for which results are discrepant, the 90 quantities described in only one paper and the numerous quantities that have not been the subject of study.

Quantitative data on the components of biological variation (BV) are used for several purposes, including calculating the reference change value (RCV) required for the assessment of the significance of changes in serial results in an individual. Pathology may modify the set point in diseased patients and, more importantly, the variation around that set-point. Our aim was to collate all published BV data in situations other than health. We report the within-subject coefficient of variation (CV(I)) for 66 quantities in 34 disease states. We compared the results with the CV(I) determined in healthy individuals and examined whether the data derived in specific diseases could be useful for clinical applications. For the majority of quantities studied, CV(I) values are of the same order in disease and health: thus the use of RCV derived from healthy subjects for monitoring patients would be reasonable. However, for a small number of quantities considered to be disease specific markers, the CV(I) differed from those in health. This could mean that RCV derived from healthy CV(I) may be inappropriate for monitoring patients in certain diseases. Hence, disease-specific RCVs may be clinically useful.

Application of BIVAC to BV publications identified deficiencies in required study detail and delivery, especially for statistical analysis. Those deficiencies impact the veracity of BV estimates. BV data from BIVAC-compliant studies can be combined to deliver robust global estimates for safe clinical application.

The RCV concept is an approach that can be offered by laboratories to assess changes in serial results. The RCV data in this study are presented as a point of departure for a widely applicable objective guide to interpret changes in serial results.