The theory of reference values: an unfinished symphony

The beginning of the reference values

Ralph Gräsbeck

The Minerva Foundation Institute for Medical Research, Biomedicum U2, 00290 Helsinki, Finland

My school participated in an exchange of teachers between Sweden and Finland. We had an excellent Swede who taught biology, including genetics. I especially remember that he taught that biometric data, e.g., the size of the leaves of a tree, were distributed in the Gaussian manner. Later, as a medical doctor, I absorbed the prevailing belief that laboratory results from normal people were distributed in this way and that the usual normal values were calculated by taking the mean ±2 SD. Then, starting my scientific career, I participated in a population study involving 1345 persons whose serum vitamin B₁₂ concentrations were assayed [17]. Our team used an excellent statistician who presented our data as histograms and curves fitted to them. The ordinary Gauss curve did not fit, so he took the logarithm of the observed values, and hocus-pocus, the curves fitted excellently. I asked the statistician why, but only got the response that this was the customary thing to do. I was not satisfied and began to dig into the literature and found quite interesting articles (quoted in [18]) casting doubt on the usual Gaussian distribution.

In the late 1960s I became editor of Scandinavian Journal of Clinical and Laboratory Medicine. The editors were supposed to write editorials, so I wrote one together with a statistician Fellman [18]. It criticized the use of the term normal and calculation of normal ranges assuming Gaussian distribution. At about the same time, I was chosen President of the next Scandinavian Congress of Clinical Chemistry and Clinical Physiology outside Helsinki in 1969. The organizers of the Congress had to devise programs for the sessions, and being sensitized to the topic I suggested ‘normal values‘ as a theme which was accepted.

In planning the session and my own presentation, I realized the analogous character of the normal values and the controls in the experimental sciences. I also thought that a less ambiguous term than normal values ought to be used (actually [18] already mentions reference population). My schoolmate Nils-Erik Saris was also one of the organizers of the Congress and experienced in producing IFCC recommendations, especially on quantities and units. I contacted him, and we decided to spend an October weekend in my island in the Baltic planning the session and especially clarifying the hazy concept of normal values and trying to produce a paper with suggested solutions of the problems. (A humorous incident occurred during our two-man sauna symposium in the naked state and has been reported) [19]. During the winter season we produced a short paper [20] suggesting the introduction of the term reference values, outlining a number of branches of the field (nomenclature, reference populations, prenalytical factors, etc.) and calling for international collaboration.

Our suggestions clearly interested the audience, which at the end of the Congress formed the General Assembly of the Scandinavian Society sponsoring the Congress. The Assembly decided to activate the rather dormant function of the Society by nominating committees to produce recommendations in several fields of current interest: quantities and units, enzyme assays and normal values. I became chairman of the last-mentioned committee, which soon began to function and changed its name to the Committee on Reference Values. The other members of the Committee were R. Dybkaer from Denmark, S. Skandsen from Norway and M. Hjelm from Sweden. The early history of the Scandinavian Committee is found in [21].

At about the same time an international congress took place in Geneva and Evian. Afterwords I learned that I had become member of an analogous international Committee on Reference Values of IFCC. When it met, T.P. Whitehead was the chairman and the rest of us members of ‘The expert panel on the theory of reference values’. Its other members, except myself, were G. Siest (F), G. Z. Williams (USA) and P. Wilding (UK); Siest and Wilding entered the Panel somewhat later than the others.

The work of the Panel was slow, nothing was published, which evoked criticism. The cause was mainly disagreement on many points, e.g., the exact meanings of the terms reference value and health, but also on practical details such as specimen collection, some favoring strict rules, while others are lax. To provide some guidelines, a fairly long series of educational articles on reference values appeared in the IFCC Newsletter No. 16, February 1977 [22]. A constructive international symposium on Reference Values in Laboratory Medicine was held outside Helsinki in May 1980 [23]. IFCC at that time had rules preventing a member of a panel to serve for more than a fixed number of years. Whitehead therefore resigned in 1976, and I was appointed chairman of the Panel. The panel now produced its first recommendation, albeit only a preliminary one ‘Provisional recommendation on the theory of reference values’ dated 1978 [24]. Then, I was in turn to resign, but continued to be an unofficial adviser and a link to the Scandinavian Committee, which decided to stick to the practical rather than the theoretical aspects of reference values.

In the work of the Scandinavian Committee, Solberg proved himself very competent, so I succeeded in getting him appointed as my successor. Under his guidance the Expert Panel completed an impressive number of recommendations. He and the members of the Panel (including myself) also spread the gospel of reference values in personal articles, one of the most noted ones being a humorous ‘Miss Manhattan’ study correcting the view that the requirements for collecting reference values were almost impossible to satisfy [25].

This is what I experienced as the beginnings of the IFCC activities in the reference value field. Accounts have previously appeared in [21, 23, 26].

At the second Pont-à-Mousson meeting in 1972, we described the first results on our family population and we published the first book on reference values [27]. We were in contact with all the scientists involved in the field and in IFCC but also with Tom Whitehead and Peter Wilding who participated in the Pont-à-Mousson meeting. They were working in a comparable health screening organization BUPA in the UK. Production of reference values required large healthy populations, military recruits, blood donors, medical students, and laboratory workers etc… populations are poor representatives of healthy populations. The very comprehensive studies have been published by health screening centers (Table 1).

The recruitment of healthy subjects is crucial for defining better laboratory test biological variability.

We were in a good position with Peter Wilding to develop the first diagrams on the definition which were at the origin of the IFCC recommendations.

Memories of IFCC Expert Panel activities in 1974

Peter Wilding, PhD, F.R.C.Path., FACB

Professor Emeritus, University of Pennsylvania, Philadelphia, PA, USA

In 1974, I was invited by Prof. Tom Whitehead to join the IFCC Expert Panel on the Theory of Reference Values. The panel at that time included Prof. Ralph Gräsbeck, Prof. George Z. Williams, Prof. Gerard Siest and myself. It is only in recent years that I have come to truly appreciate the privilege I enjoyed being part of a group that would establish the principles and terminology that governs the topic of reference value use, and assignment, today and in future years. Moreover, I gained so much from the interaction with four outstanding clinical chemists who worked energetically to improve our profession.

For several years the Expert Panel had strived to define its purpose, and determine a strategy, for proposing a clear and definitive nomenclature for the topic. Much of the problem lay in finding a way to present the subject in a simple and unambiguous manner that would facilitate understanding of why we needed reference values, how to create them and how to interpret them.

In 1976 Gerard Siest and I were asked to find a structure for presentation. We met one evening at a London hotel and after a lengthy, and animated, discussion formulated a diagram that we believed would meet the Panel’s needs. The diagram was presented to the Panel, where it was accepted unanimously, and was presented to the profession in several journals in 1978. The diagram (see below) has stood the test of time and laid the basis of numerous derivative articles from 1978 to 1991 and Approved Recommendations from IFCC. None of the five individuals who formulated, and presented, this time-tested proposal are included in the authorship of the Approved Recommendations, but whenever I see, or read, articles using terms such as ‘reference individuals’, ‘reference values’ and ‘reference intervals’, my memories of a long evening in a London Hotel conference room arguing, discussing and debating with my friend Gerard Siest are revived with a tremendous sense of accomplishment over what we created.

This first document corresponds to the first IFCC paper published 13 years later [28].

Outside this expert panel, different groups of scientists were involved in studies on reference values between the years 1970 and 2000. We could classify them into four international groups:

1. SFBC (Société Française de Biologie Clinique) and French speaking group:

• G. Siest, J. Henny, R. Guegen, Ch. Sachs,

  L. Guize, B. Gouget, G. Blin, J.P. Bretaudière,

  M. Drosdowsky, H. Favre, E. Daubrosse-Panek,

  A. Aellig, J.M. Queralto, F. Ramon, A. Albert,

  C. Heusghem, C. Petitclerc, L. Munan, R. Zender…

2. Scandinavian group:

• R. Gräsbeck, R. Dybkaer, S. Skandsen, M. Hjelm,

  H. Solberg, P. Hyltoft-Petersen, N. Lahti,

  M. Jungner, A. Kallner

3. American and British group:

• G.Z. Williams, E.K. Harris, D. Young, J. Boyd,

  P. Wilding, J. Lott, C. Fraser

4. Spanish group:

• J.M. Queralto, F. Antoja, M. Cortes, M.V. Domenech,

  J. Fuentes, M.J. Llagostera, J. Ordonez, F. Ramon,

  I. Rojo, J. Fuentes-Arderiu, J. Horno, P. Ferrer

Due to the activity of the same people in IFCC and the French speaking group, the recommendations have been prepared simultaneously by both groups perhaps with more details in the SFBC one where the meetings were more frequent. Some countries, e.g., Spain also had a reference values committee with good input from our Spanish colleagues.

We can compare the papers published by the four main groups, they are not equivalent. Often papers in French were published before and used as reference documents later for IFCC recommendations that came much too long after. IFCC Expert Panel on theory of reference values was created in 1972 and the last approved recommendations came 19 years later! (Table 2).

The SFBC Reference Values Working group was created in 1973 and recommendations were published in the early 1980s; a last recommendation on the use of reference values [48] was not addressed by the IFCC expert panel. The Spanish group was also very active, publishing six articles of recommendations. Two of them were additional documents to those of the IFCC and SFBC: one on the transferability of reference values, the second one on the intra-individual reference values.

Simultaneously, the Scandinavian group, published in 1993, on behalf of the Committee on Reference Values of the Scandinavian Society for Clinical Chemistry the recommendation for the preanalytical step including management of the biological samples [55].

The works of the different committees were presented in IFCC conferences or other specific meetings. They were occasions to meet our American and Canadian colleagues. That was the case in 1975 in Toronto, where we presented the concept of reference state and met Claude Petitclerc and Louis Munan. They have developed an ‘atlas of blood data’ and it was the start of a fruitful collaboration.

Claude Petitclerc, MD, PhD, F.R.C.Path.

Department of Biochemistry, CHUM-Hôpital-Notre-Dame, Montreal, QC, Canada

In the 1970s, I was involved in a long and fruitful collaboration with epidemiologists Louis Munan and Anthea Kelly at Sherbrooke University. The Sherbrooke data obtained from a probability sample of families of a large freestanding population was the subject of many publications on reference data and correlates [56–59]. In 1978 I took sabbatical leave to join Gerard Siest’s team.

CMP was the mecca. Not only was the staff remarkable in number and quality but the study population was much larger, family-based and very well-documented. Reference values were systematically produced and published.

Similarities of both sampled populations and laboratory technologies prompted a comparative study that revealed not only population differences but also analytical biases [60]. The problem of transferability of laboratory results and hence reference values had to be addressed. Reference data produced with the technology of the 1970s could not be used decades later. However, the enormous contribution of Siest et al. at CMP was the systematic study of factors of biological variations and their stratification for each substance. This information, independent of technology, was and remains transferable.

More complex was the issue of analytical variability imbedded in the massive amount of population data over the years, especially at the time of data mining. It was then a concern and is still one. A challenge for study of long-term longitudinal variations. A long and fruitful collaboration of 35 years was the outcome of this sabbatical leave.

During my stay, I had the honor to serve on Commission des valeurs de référence de la Société Française de Biologie Clinique (SFBC). Inspired by Gerard Siest‘s leadership SFBC was very proactive in publishing guidelines for the production of reference values; being part of such a dynamic group opened the way to relay Gerard Siest on the Expert Panel on Theory of Reference Values of IFCC. I keep a vivid memory of the members of the panel with whom I enjoyed to exchange with for some 10 years under chairman Erik Solberg. Document 2 and 3 were spinoffs of both SFBC and Scandinavian Society for Clinical Chemistry recommendations.

CMP is no longer what it once was, but the people who made it what it was are still around, with us or in our memory. I cherish them all.

The importance of Eugene K. Harris contribution

Josep M. Queraltó

Hospital de la Santa Creu i Sant Pau, Servei de Bioquimica, Av. Sant Antoni M. Claret, 167, 08025 Barcelona, Spain

I was very happy to work with Eugene K. Harris during a sabbatical period. I am indebted to him sharing his great scientific culture with me. Although reference values theory clearly is a part of the so-called post-analytical phase, the development of reference values theory is without a doubt a decided drive to refine preanalytical and analytical concepts. The central idea to create a space where a presumptive patient can be tested against non-diseased individuals raised the need for rigorous, practical tools to manage physiological pathological knowledge. Uncertainty, variability, sampling, confidence, tolerance, robustness, expectancy… are concepts with an accurate meaning in mathematical statistics which soon were part of the reference values theory. Accurate use of these terms deserved the special attention of all the groups which conceptualized reference values. Among the statisticians who played a key role to translate statistical science into laboratory medicine was Eugene K. Harris (1927–1997). Gene Harris was a wise man with a broad culture and scientific interest. Gene had an exceptional ability in grasping what those professionals which consult him really needed. At the end of the 1960s, Gene begun to collaborate with Prof. George Z. Williams, who had already published some articles about laboratory data process [61]. Their insights on the components of medical laboratory data variation, biological and analytic, were published in a series of influent articles [2–6]. He realized the relevance of within-person variability in the assessment of biological variation of a number of analytes [62–65]. Eugene Harris contributed significantly in virtually all the aspects of the reference values theory where statistics are required: gaussianity requirements [66], analytical goals [67, 68], analytical variation in monitoring reference individual over time [69, 70], the relative influence of analytical to biological variance [71], the relevance of intra-individual component of biological variance [72, 73], the interpretation of the practical significance of a change in laboratory results [74–78], multivariate regions of reference [79] and the requirement of partitioning reference populations [80, 81], as well as comprehensive reviews [82] and books on statistical aspects of reference values production and use [83, 84]. Moreover, he was persuaded of the need to convey to both clinical chemists and doctors the clearest and sharper statistical concepts together with the straightforward and practical procedures and methods available. At that time, when personal computers did not enjoy today’s facilities, he provided also useful and simple code to help in statistical calculations [84]. The vision and personality of Eugene Harris inspired the work from national and international committees and scientists [32, 52] and today still remains as a reference.

During this period, we started the recruitment of 1000 families of the Stanislas cohort [85] on a more specific research group in CMP focused on individual specific reference values in subject under medication, e.g., contraceptive pills. We tried also to propose new presentation of reference limits particularly for a better use in prevention on health maintenance including for and by the patients themselves.

The data produced were not limited to biochemical components but also to weight, blood pressure etc… The material collected in the Stanislas cohort project comprised a lot of interesting longitudinal familial data and of biological variability and many publications came out from this tool [85] and particularly the effects of genetics [86], e.g., Apolipoprotein E.

We tried many times particularly as IFCC president to recreate an IFCC reference values committee but the IFCC scientific committee opposed propositions made by a group of scientists. The objectives of this committee would comprise the following:

1. Update IFCC recommendations;

2. Produce practical guidelines;

3. Increase the transferability;

4. Improve the proper and extended use of reference values by clinicians, laboratory scientists and patients.

We proposed, without success, a joint CCLM IFCC committee for a better and quicker diffusion of the work and the recommendations.

The founding texts of various scientific societies over the past 20 years are true reference knowledge. They had the merit to disseminate the concept of reference values. Thanks to them, the RIs have become a basic tool for the interpretation of the quantitative results of laboratory tests. However, the evolution of medical practice led to question the current relevance of this concept.

Fundamental works: evolution of the concept and proposals for new procedures

In the early 2000s The Catalan Association of Clinical Laboratory Sciences organized two successive symposia for opening dialogue between professionals and industry representatives. It approached the legal and normative alternatives to the production of specific reference values and the transferability of reference values and the methodological aspects (how define a homogenous starting population, how to define acceptable health status for reference individuals, methodological and statistical aspects, diffusion and tracking of the information) [91].

A year later, in 2004, after the second Catalan Symposium, a special issue of CCLM was published in order to share the basic concepts and new ideas on the field reference values [92]. The expressed opinion were very diverse and do not reflect systematically the official practices. It was the choice of editors to open debates and initiate discussions. Among the 25 contributions it is impossible to say which are the most determining and most helpful. At the risk of being wrong, here are the themes that seem, 10 years later, the most significant: is there a concept of health and normality?, statistical description of reference values, interpretation of reference limits, the concept of common and multicenter reference intervals, and analytical aspects in relation to reference intervals.

Per Hyltoft Petersen

NOKLUS, Norwegian Quality Improvement of Primary Care Laboratories, Division for General Practice, University of Bergen, Bergen, Norway

I have been involved in application of the concept in creation of common reference intervals for plasma proteins in the Nordic countries in the 80th and 90th, mostly published as NORDKEM reports and in Upsala Journal of Medical Sciences [93, 94]. NORIP, the other Nordic project on common reference intervals, was chaired by Pål Rustad in Oslo, Norway.

The development of the concept on reference values was a great intellectual conquest which solved many problems in relation to interpretation of clinical laboratory data. The concept, however, is developed for single laboratories, so there may still be differences in measurement results and in the interpretation for the same samples analyzed in different laboratories. Consequently, the next aim was to harmonize measurement results and interpretation of data among laboratories, so that results were interchangeable and clinical explanation would be independent of which laboratory performed the analysis.

Analytical quality: The question on analytical quality of measurements is technical analytical and relates to many elements in the traceability chain from reference method values or international certified reference preparations to the single measurement result. Even these efforts are essential; the final analytical quality can only be proved by control with genuine human materials with target values measured by the reference method. The documentation procedure is considerable and rules for acceptable analytical quality specifications are outlined.

With the analytical quality assured and documented, the next question is whether the reference intervals are dependent on gender, genetic and other biological characteristics.

Models for treatment of the biological information from reference values

• Even the idea of sampling of reference individuals for estimation of reference intervals is clear in the sense of defining healthy individuals through rule-in and rule-out criteria, the concept of health is more complicated than supposed for varying criteria, and a common question is whether the reference interval should mirror the current apparently healthy population (without any known diagnosis) or be an ideal for all, e.g., based on teetotallers.

• Further, differences in rule-in and rule-out criteria may be needed for varying components. A reference interval for, e.g., serum-thyroid stimulating hormone (TSH) will need criteria for absence of thyroid antibodies in all reference individuals, and when, e.g., reference values are to be collected for plasma-glucose and HbA_1c in relation to diabetes diagnostic, the criteria will be very strict regarding, e.g., first-generation diabetes relatives. Moreover, the biological concept has been overruled by the international clinical guidelines which recommend diagnostic decision limits for the diabetes diagnosis, neglecting biological variation and other individualities.

• When two or more subgroups of reference individuals are disclosing deviating reference intervals, it is necessary to decide whether the groups can be combined or should be separated. Varying partitioning criteria have been proposed, but the dominating model is to separate if the result of using the common reference interval results in percentages of reference individuals for lower as well as upper reference limits for both groups are beyond 0.9%–4.1%.

The idea of common reference intervals is extremely attractive and several attempts have been made. For serum electrolytes and some metabolites it seems possible to establish common reference intervals within ethnic groups and selected geographical areas, but for most other components it is only possible within smaller well-defined groups, and it is necessary to test whether common reference intervals can be established.

Further, I have been involved in evaluation of analytical quality specifications, together with Elizabeth Gowans, Callum Fraser, and others, and in proposing criteria for partitioning together with Ari Lahti.

Together with Callum Fraser, Anders Kallner and Desmond Kenny, I arranged the Stockholm conference on analytical quality specifications, published in Scand J Clin Lab Invest 1999;59:475–585, and together with Joseph Henny I edited the CCLM issue on reference values [92]. Moreover, I wrote a paper together with Esther Jensen and Ivan Brandslund: Analytical performance, reference values and decision limits. There is a need to differentiate between reference intervals and decision limits and to define analytical quality specifications which shows many of the restrictions of use of reference intervals and of decision limits [95].

Analytical aspects

The analytical bias affects the measurement results and can change RIs, then this may can change the interpretation of results. Determining RIs with analytical procedures traceable to a reference system allows transferring results from one laboratory to another [129]. The use of this strategy should be promoted and favored by IDV manufacturers to provide more effective information to clinicians. However, it has no universal value for many years; it concerns only the analytes traceable to a reference system (about 60 to this day). For others the issue of transferability remains. C-RIDL Committee is working hard on this issue and should propose solutions soon. Alternatives could also be considered, such as the use of a common calibrator, but also with other limitations [130].

The role of the IFCC-LM: creation of C-RIDL (Committee for Reference Intervals and Decision Limits)

Until 2004 the Scientific Division of the IFCC-LM was no longer interested in the reference values issues. It must be recognized that the guidelines proposed by the IFCC and the NCCLS (now CLSI) were less and less adapted to the needs of the professionals: IVD’s manufacturers and laboratorians. Directive 98/79/EC of the European Community requires manufacturers to provide reference limits in the package inserts of laboratory reagents kits [87]. Other international organizations, such as ISO also recommends the use of RIs in the laboratory reports [88]. To meet the growing demands of international regulations and, largely on the insistence of IVD manufacturers, IFCC-LM decided in 2005 to create the Committee on Reference Intervals and Limits decision (C-RIDL). It was composed of Ceriotti F. (Chairman), Boyd J., Henny J., Kairisto V., Klein G., and Queralto J.M. Almost simultaneously, a working group of the CLSI was created with similar objectives. A joint Working-Group IFCC-LM and CLSI was created in 2005 during the XIX International Congress of the IFCC-LM in Orlando. The purpose of this joint Working-Group was to review the document CLSI (C 28–A) published in its original version in 1995 [131] and revised in 2000 [132] in order:

1. To provide clinical laboratories and diagnostic test manufacturers with updated guidelines for determining reference intervals for quantitative analytes;

2. To provide recommendations regarding procedures that can be used to verify reliable reference intervals for use in laboratory medicine; and

3. To publish a document common to IFCC and CLSI.

Updated document was published in 2008 [133]. The major guidelines of the original recommendations were preserved in the updated document. Three main innovations were introduced: the introduction of a robust method for analyzing references values (especially if the number of data is <100), the transference of RIs and RIs verification, and the reference intervals multicenter studies [134, 135].

Independently, in parallel, C-RIDL published from 2005 to 2010 a series of works. The first seeks to identify universal reference intervals for creatinine. A careful study of the literature showed that very few articles had taken into account the requirements of metrological traceability, thus few data from the literature are transferable from one laboratory to another. The C-RIDL established that for laboratory using a method traceable to IDMS the proposed RIs can be used [136].

The second job of the C-RIDL was devoted to the opportunity to determine Common Reference Intervals. In a preliminary article Ceriotti recalled the position of the IFCC-LM and prerequisites to follow [137]. These are related to the design of the study and the analytical conditions (traceability). Ceriotti also recalls that the populations (that of laboratory and that of the study) must be comparable. An application’s work was made for the measurement of three enzymes (AST, ALT and GGT) measured with commercial analytical systems, but according to the standard methods recommended by the IFCC. Patient’s sera from four regions (Italy, China, Turkey and the Nordic countries) were analyzed. For AST and ALT the use of common RIs appears possible. In contrast, GGT shows significant differences between populations, worldwide RIs do not seem applicable [138].

Finally, two members of the Committee attempted to clarify the position of Reference Intervals vs. Decision Limits. In an article in the eJICC, Ceriotti and Henny recalled that the reference limits should not be confused with the decision limits. They recalled their definitions and characteristics (Conditions Influencing them, Information gathered, Statistical methodologies …). This is one of the few articles on the role of reference intervals in the process of the interpretation of laboratory results [139]. Incidentally, this topic was raised again a few years later by Haeckel [140] and Hyltoft Petersen [95].

In 2011, Ichihara K. became the new chairman of the C-RIDL renewed (members: D. Armbruster, Barth J., Klee G., Ozarda Y., Pekelharing M.). Now, it aims to derive reliable country specific RIs through multicenter studies at a global level. A protocol was developed by the Committee. It will be applied to markers traceable to a reference system and non-standardized biochemical markers. The ongoing project will demonstrate the feasibility of the protocol and deriving RIs applicable to a global scale if there are no regional and/or ethnic groups’ differences. In 2012 the study is in progress and the results should be published in 2013.

Deriving works of reference limits

Besides these fundamental works extensively devoted to the derivation of reference intervals, numerous articles have been published in various scientific journals:

The influence of ethnicity as a partition factor has never been really studied systematically. Horn [141] and Johnson [142] observed for some analytes differences between populations that require separate reference intervals. In this way a large number of works from emerging countries have been published most often in scientific regional journals (they are too numerous to be mentioned here). Most of them follow (more or less) the traditional protocol of the IFCC, applied according to local opportunities and constraints. Each article covers a limited number of analytes and is applied to a group of well-defined populations.

Another group of works relates more specifically to certain periods of life: pediatric age, elderly and pregnancy.

1. Pediatrics reference intervals determination is one of the most difficult tasks to carry out, primarily because of ethical limitations related to blood drawing in very young children (or in neonates). Apart from a few specific derivation RIs studies for highly specialized analytes [143–146], the Canadian multicenter study CALIPER, aimed to establish a data base of age- and gender-specific pediatric intervals for 40 serum biochemical markers [147, 148]. As emphasized, Ceriotti [149] the authors have followed the recommendations of the IFCC and CLSI including selection of young children in good health (on the basis of clinical data). However, traceability of analytical data is not truly established. Also transferability of produced RIs may not be always transferable to other analytical systems. The Canadian group, as a service to major clinical laboratories, published a series of articles showing RIs related to various specific analytical systems [150–153].

2. Elderly RIs determination is even more challenging than for young children. The main difficulty is to select healthy individuals: most seniors do not meet the criteria of the IFCC/CLSI. Taking medication(s), unrecognized subclinical diseases can alter the width of the reference range. Hence it is very difficult to differentiate the effect of age, aging or a pathological condition. During this period 2000–2012 some authors are interested in this issue [154–156].

3. Pregnancy is a factor of biological variation due to physiological changes during pregnancy and postpartum. Although changes during pregnancy are well-known some articles are still published [157–159].

4. Genetic is now a well-recognized determinant for many biochemical markers. Use of genetic information for partitioning reference intervals could reduce the misidentification of unusual test results caused by non-disease associated genetic variation, as pointed by Shirts [160]. Siest and co-workers started to be more precise in using genetic information for subgroups stratification, e.g., for ApoE [86] and more recently for haptoglobin [161]. However, the part of biological variability induced by genetic variants is often low and the knowledge of the genetic status of the reference individuals is often lacking. In these conditions the partitioning of RIs by genetic information remain for the moment (and for many long years) only a possible way for the future.

During these years, the theory of reference values was unexpectedly applied to the veterinary laboratory medicine! A French team from the Veterinary School of Toulouse published several theoretical and applied articles, noticeably on reference intervals derived for several animals [162–166].