Tools, Technologies and Training for Healthcare Laboratories

Guest Essay

Questions on the BioDatabase 2014

While we celebrate 15 years of evidence-based quality goals, we also have a few persistent questions. Why are some goals listed as desirable, optimal, or minimal? Find out why...

Questions about the 2014 update of the Biological variation database, and quality specifications for imprecision, bias and total error.

February 2014

Joana Minchinela[1,2], Carmen Ricós[1], Carmen Perich* [1,3], Pilar Fernández-Calle[1,4], Virtudes Alvarez[1,5], Mariví Domenech[1,6], Margarita Simón[1,7], Carmen Biosca[1,8], Beatriz Boned[1,9], Fernando Cava[1,10], José-Vicente García-Lario[1,11], Mª Pilar Fernández-Fernández[1]

1. Analytical Quality Commission, Spanish Society of Clinical Biochemistry and Molecular Pathology (SEQC)
2. Laboratori Clínic del Barcelonès Nord i Vallès Oriental Badalona)
3. Laboratori Clínic Bon Pastor (Barcelona)
4. Servicio Análisis Clínicos Hospital La Paz (Madrid)
5. Laboratori Clínic de l’Hospitalet (Barcelona)
6. Laboratori Clínic Manso (Barcelona)
7. Consorci del Laboratori Intercomarcal de l’Alt Penedès, l’Anoia i el Garraf (Barcelona)
8. Servei de Bioquímica Hospital GermansTrias i Pujol (Badalona)
9. Hospital Royo Villanova (Zaragoza)
10. Laboratorio Central BR Salud-Hospital Universitario Infanta Sofía (Madrid)
11. Laboratorio de Análisis Clínicos, Hospital Universitario Virgen de las Nieves (Granada).

* President of the Analytical Quality Committee, Spanish Society of Clinical Chemistry and Molecular Pathology (SEQC).

1. Why are some analytes on the minimum list and some are on the optimum list? Is there any rule for including or defining an analyte that should be on those lists?

Selection of the analytes to be included in the list of desirable, optimum and minimum specifications is based on the results obtained by Spanish laboratories participating in the SEQC external quality assurance programs. <p >For each analyte, percentage deviations of results with respect to the peer-group mean are depicted in ascending order, percentiles 10, 20, 30, etc, up to 90, 95,100 are described and compared with the limits derived from biological variation (VB). Analytes in which 100 to 95 percentiles fell within the optimum VB limit, are included in the optimum list; analytes with 94 to 81 percentiles fell within the desirable VB limit are included in the desirable list and analytes with less than 80 percentiles felling within the desirable limit are included in the minimum list.<p >This criterion is known by laboratories participating in our EQA programs because it has been published in a technical note (January 2013) and also in our website (1).

2. Are there some analytes that haven't yet been studied, that are on your "wish list"?

In principle, all analytes without data on BV or with few data are in our “wish list”. So, we are continuously looking for papers dealing with BV studies.

3. Why do you think we see so few of these biologic variation studies performed in any given year? Is it too expensive and time-consuming to conduct a biologic variation study? Are diagnostic companies and other funding sources unlikely to support this type of study?

We suspect that the reason for the low number of new BV papers per year is the expensiveness and time consuming of this kind of study. Also may be the IVD companies are poorly interested to support these studies.

4. Are there some easy basics you could describe for laboratories that are interested in contributing to this database? For instance, can such a biologic variation study be performed retrospectively, on patients that have already been through the system?

Yes, we believe this method must be developed in depth, on the basis of Cembrowsky paper published in 2010 (2).

5. I realize that studies of drugs and non-natural substances are not included. There was a recent paper that proposed a new way to estimate the variation of drug assays [Graham Jones, 2008 AACC poster]. But would it be possible to measure "stable patients" using the traditional biological variation study approach, and get an estimate of expected within-subject variation on TDMs, etc.?

They was a paper from Callum Fraser on this subject (3); however, because of our feeble understanding of it and the fact that not further studies have appeared for many years, we decided not include data on biological variation on TDM in our compilation.<p >The work from Graham Jones presented at the 2008 AACC congress had not been known by us since now; so, we are very grateful for your information. We contacted Graham and read his presentation, which seemed to us a very interesting approach, although it does not allow estimating the between-subject component of variation. So, we will consider learning more on this experience for the future development of our biological variation database.

References

  1. Perich C, Alvarez V, Biosca C, Boned B et al. Recomendación para el uso de las especificaciones de la calidad analítica. http://www.contcal.org/qcweb/Documents/10%20Informacio%20general/35%20Especificacions%20de%20la%20Qualitat%20Analítica/CAS/Recomendaciones%20para%20el%20uso%20de%20las%20especificaciones%20de%20la%20Calidad%20Analítica.pdf
  2. Cembrowsky G, Tran DV and Higgins TN. The use of serial patient blood gas, electrolyte and glucose results to derive biological variation: a new tool to assess the acceptability of intensive care unit testing. Clin Chem Lab Med 2010; 48:1447-1454.
  3. Fraser CG. Desirable goals in therapeutic drug monitoring. Clin Chem 1987;33:387–9.

 

-          Annex I, showing quality specifications, has a new column with the number of articles compiled for each analyte. This information also appears in Annex II; however, the number of consultations received dealing with this item seemed us to indicate that many readers did not realize about annex II.

This column highlights the robustness of CVI and CVG described (median of values compiled for each analyte). For example, there are 58 cholesterol CVI values coming from 46 papers, which results to a median of CVI=5.9% and a confidence interval (p<0.05) of 0.43, with values ranging from 2.4% to 9.7%.  AST has 21 data coming from de 13 papers, with a median of   CVI = 12.3%,   a confidence interval of 2.02% and values ranging between 4.7 and 24.2%. There are many analytes with a reduced number of papers published, thus the CVI and CVG estimated are quite poor and further studies should be made.