Biological variation – reliable data is essential

Aasne K. Aarsand; Thomas Røraas; Sverre Sandberg

doi:10.1515/cclm-2014-1141

Article Publicly Available

Biological variation – reliable data is essential

Aasne K. Aarsand , Thomas Røraas and Sverre Sandberg

Published/Copyright: January 7, 2015

Published by

Become an author with De Gruyter Brill

Submit Manuscript Author Information Explore this Subject

From the journal Clinical Chemistry and Laboratory Medicine (CCLM) Volume 53 Issue 2

Biological variation (BV) data is a cornerstone in the interpretation of laboratory test results, being the basis for many of the decisions we make every day both in the laboratory and in clinical practise. Among the many applications is its use in diagnosis and monitoring of disease. Most typically this occurs when comparing a person’s level of the analyte of interest against a reference interval, based on the between-subject variation, CV_G, or when comparing a change against the reference change value, based on the within-subject variation, CV_I. Furthermore, BV data is probably the most commonly used approach for setting analytical quality specifications for bias, imprecision and total error for many laboratory constituents. General assumptions for the uses of BV data are that estimates are reliable, i.e., adequately collected and calculated, and that the estimates are representative for the specific population and setting for which they will be applied.

At the Stockholm conference in 1999, Carmen Ricos and the Analytical Quality Commission of the Spanish Society of Clinical Chemistry presented an overview of estimates of BV for a large number of constituents [1]. As a continuation of this work, estimates of CV_I and CV_G, identified by literature searches, and the associated analytical quality specifications for bias, imprecision and total error have been presented on the Westgard webpage [2], being updated every 2 years. This data has been used as the main source of estimates of BV in our community. The impact of this information is, and has been, huge. In the present issue of Clinical Chemistry and Laboratory Medicine, Perich et al. present in more detail than has previously been published, how they select and judge the papers they use as basis for their BV estimates [3]. It is an impressive amount of work involved, with a total of 358 analytes included in the 8th edition, and we commend them for their continued dedication to this important topic. However, the wide variation in CV_I estimates published for different constituents such as, e.g., urinary albumin [4] and enzymes [5] is clearly of concern. Some of this variation can probably be explained by different time intervals used for data collection; but we think it far more likely that it is caused by improper methodology for sampling and inadequate statistical analysis. The laboratory profession has for years been very concerned with diagnostic accuracy of biochemical markers, but a critical view on the evidence base of the available BV data is less evident.

The criteria applied by Perich et al. to select publications from which they base their BV estimates include that the studies have been designed for such a purpose, that a performance index score, defined as CV_A/0.5*CV_I, ≤2 must be obtained, and that the type of statistical model used is described, scored in the following order: 1) ANOVA; 2) as described by Fraser and Harris (also an ANOVA) [6, 7]; and 3) unclear models. We agree that ANOVA, typically with a nested random design in two levels, is a robust method for estimating BV. However, homoscedasticity of the individual CV’s has to be fulfilled for the estimates to be representative for the population it describes. If this requirement is not examined for and/or not met, the obtained estimates of BV may be of no value. To assure that estimates are sound, consensus on mandatory statistical elements is necessary, including, e.g., testing for outliers and examination of homoscedasticity of variances. Furthermore, the reliability of BV estimates greatly depends on the study design, e.g., number of individuals, samples and replicates included and the level of analytical imprecision. It is also of great concern that measures of uncertainty on BV data have been lacking until recently, and this should be required when reporting BV [8].

On the webpage produced by the Spanish group [2], BV estimates derived from studies on apparently healthy adults and from subjects with documented pathology are presented in two different databases. The number of studies included for different constituents in these databases varies from one to 45. For more than half of the constituents only one publication is available. For constituents with more than one publication, the estimates for CV_I and CV_G presented on the webpage are generated using the median of the compiled data. We suggest that prior to estimating a common value for the CV_I, it has to be examined if the varying estimates for one constituent originating from different studies can be assumed to be the same, considering how factors, such as population, sex, age, timing of samples and the statistical model applied, may have impact. This could be achieved by generating, in the same fashion as for diagnostic accuracy studies [9], an instrument for examining homogeneity of BV studies and their methodological quality. Thereafter, statistical models for meta-analysis of the BV results must be developed to decide how common estimates are to be produced.

As an indicator of the quality of the BV estimates they publish, Perich et al. use the ratio of CV_{I MAX}/CV_{I MIN}, and consider that a ratio below seven indicates robustness. This covers 86% of the components included in the database. We would, however, suggest that a CV_{I MAX}/CV_{I MIN} >2 is likely to indicate a significant difference between the BV estimates [8]. Considering how we use this data as basis for our analytical quality specifications and for diagnosing and monitoring patients, the effect of using a CV_I estimate of, e.g., 10% instead of, e.g., 20%, a CV_{I MAX}/CV_{I MIN} ratio of only 2, is substantial, with direct implications for the quality specifications and control rules we use in the laboratory. Thus, the effect of unreliable BV estimates on the quality of our analytical work and on the care of patients is both immediate and large. We, and others, call for an international standard for performing and reporting of studies on BV, similar to what has been achieved for diagnostic studies through the STARD initiative [10]. An initiative is already underway by the Biological Variation Working Group of the European Federation of Laboratory Medicine (EFLM) [11, 12]. We applaud the Analytical Quality Commission of the Spanish Society of Clinical Chemistry for their readiness and willingness to continue their work and to collaborate with such initiatives, which will further improve the quality and usefulness of the BV database.

Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission.

Financial support: None declared.

Employment or leadership: None declared.

Honorarium: None declared.

Competing interests: The funding organization(s) played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.

Corresponding author: Aasne K. Aarsand, Laboratory of Clinical Biochemistry, Haukeland University Hospital, 5021 Bergen, Norway, Phone: +47 55 973127/+47 55 973170, Fax: +47 55973115, E-mail: aasne.aarsand@helse-bergen.no; and Norwegian Quality Improvement of Primary Care Laboratories (NOKLUS), Haraldsplass Deaconess Hospital, Bergen, Norway

References

1. Ricos C, Alvarez V, Cava F, Garcia-Lario JV, Hernandez A, Jimenez CV, et al. Current databases on biological variation: pros, cons and progress. Scand J Clin Lab Invest 1999;59:491–500.10.1080/00365519950185229Search in Google Scholar PubMed

2. Ricos C, Alvarez V, Cava F, Garcia-Lario JV, Hernandez A, Jimenez CV, et al. Desirable specifications for total error, imprecision, and bias, derived from intra- and inter-individual biologic variation. The 2014 update. Available from: www.westgard.com/biodatabase1.htm#1. Accessed November, 2014.Search in Google Scholar

3. Perich C, Minchinela J, Ricos C, Fernandez-Calle P, Alvarez V, Doménech MV, et al. Biological variation database: structure and criteria used for generation and update. Clin Chem Lab Med 2015;53:299–305.10.1515/cclm-2014-0739Search in Google Scholar PubMed

4. Miller WG, Bruns DE, Hortin GL, Sandberg S, Aakre KM, McQueen MJ, et al. Current issues in measurement and reporting of urinary albumin excretion. Clin Chem 2009;55:24–38.10.1373/clinchem.2008.106567Search in Google Scholar PubMed

5. Carobene A, Braga F, Roraas T, Sandberg S, Bartlett WA. A systematic review of data on biological variation for alanine aminotransferase, aspartate aminotransferase and gamma-glutamyl transferase. Clin Chem Lab Med 2013;51:1997–2007.10.1515/cclm-2013-0096Search in Google Scholar PubMed

6. Fraser CG. Biological variation: from principles to practice. Washington DC: AACC Press, 2001.Search in Google Scholar

7. Fraser CG, Harris EK. Generation and application of data on biological variation in clinical chemistry. Crit Rev Clin Lab Sci 1989;27:409–37.10.3109/10408368909106595Search in Google Scholar PubMed

8. Roraas T, Petersen PH, Sandberg S. Confidence intervals and power calculations for within-person biological variation: effect of analytical imprecision, number of replicates, number of samples, and number of individuals. Clin Chem 2012;58:1306–13.10.1373/clinchem.2012.187781Search in Google Scholar PubMed

9. Reitsma JB, Moons KG, Bossuyt PM, Linnet K. Systematic reviews of studies quantifying the accuracy of diagnostic tests and markers. Clin Chem 2012;58:1534–45.10.1373/clinchem.2012.182568Search in Google Scholar PubMed

10. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig LM, et al. Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. Clin Chem Lab Med 2003;41:68–73.10.1515/CCLM.2003.012Search in Google Scholar PubMed

11. Bartlett WA, Braga F, Carobene A, Coskun A, Prusa R, Fernandez-Calle P, et al. Identification of key meta data to enable safe accurate and effective transferability of biological variation data. American Association of Clinical Chemistry Annual Meeting 2014, Chicago. 2014. Available from: www.aacc.org/~/media/files/annual-meeting/2014/abstracts/aacc_14_abstractbook_1_combined.pdf?la=u. Accessed November, 2014.Search in Google Scholar

12. European Federation of Laboratory Medicine Biological Variation Working Group. Available from: www.efcclm.eu/index.php/wg-biological-variation.html. Accessed November, 2014.Search in Google Scholar

Published Online: 2015-1-7

Published in Print: 2015-2-1

Articles in the same Issue

https://doi.org/10.1515/cclm-2014-1141