Clin Chem Lab Med 2023; 61(5): 741–750

Review

Sverre Sandberg*, Anna Carobene, Bill Bartlett, Abdurrahman Coskun, Pilar Fernandez-Calle,
Niels Jonker, Jorge Díaz-Garzón and Aasne K. Aarsand

Biological variation: recent development and

future challenges
https://doi.org/10.1515/cclm-2022-1255 other standards for deriving and reporting BV data, the
Received December 10, 2022; accepted December 12, 2022; EFLM Biological Variation Database and new applications of
published online December 20, 2022
BV data including personalized reference intervals and
measurement uncertainty.
Abstract: Biological variation (BV) data have many appli-
cations in laboratory medicine. However, these depend on Keywords: biological variation; BIVAC; EuBIVAS; personal-
the availability of relevant and robust BV data fit for pur- ized reference intervals (prRI); reference change value.
pose. BV data can be obtained through different study
designs, both by experimental studies and studies utilizing
previously analysed routine results derived from laboratory Background
databases. The different BV applications include using
BV data for setting analytical performance specifications, Biological variation (BV) describes the variation observed
to calculate reference change values, to define the index in the concentration or activity of different constituents in a
of individuality and to establish personalized reference person, reflecting the regulation by homeostatic processes
intervals. In this review, major achievements in the area of in the body [1]. In a steady state setting, the concentration
BV from last decade will be presented and discussed. These of most measurands is characterized by random variation
range from new models and approaches to derive BV data, around a homeostatic set point, whereas the concentration
the delivery of high-quality BV data by the highly powered of some measurands is also influenced by different life
European Biological Variation Study (EuBIVAS), the Biolog- phases or predictable cyclic variation. The within-subject
ical Variation Data Critical Appraisal Checklist (BIVAC) and variation BV (CVI) denotes the variation of the concentra-
tion/activity of a measurand around a homeostatic set point
within a single individual in steady state, whereas the
*Corresponding author: Sverre Sandberg, Norwegian Organization for
Quality Improvement of Laboratory Examinations (Noklus), Haraldsplass
between-subject BV (CVG) denotes the variation between the
Deaconess Hospital, Bergen, Norway; Department of Medical Biochemistry homeostatic set points of different individuals.
and Pharmacology, Norwegian Porphyria Centre, Haukeland University BV data have many different applications in laboratory
Hospital, Bergen, Norway; and Department of Global Public Health and medicine. A major use of BV data is for setting analytical
Primary Care, University of Bergen, Bergen, Norway, performance specifications (APS) for imprecision, bias, total
E-mail: sverre.sandberg@noklus.no
error and measurement uncertainty. These and other
Anna Carobene, Laboratory Medicine, IRCCS San Raffaele Scientific
Institute, Milan, Italy characteristics can be established on estimates of within-
Bill Bartlett, School of Science and Engineering, University of Dundee, and between-subject BV, utilizing different formulae. The
Dundee, Scotland utility of the conventional population-based reference
Abdurrahman Coskun, Acibadem Mehmet Ali Aydınlar University, School intervals can be assessed by the index of individuality (II),
of Medicine, Istanbul, Türkiye
which usually is calculated as the ratio between CVI and
Pilar Fernandez-Calle and Jorge Díaz-Garzón, Hospital Universitario La
Paz, Quality Analytical Commission of Spanish Society of Clinical Chemistry
CVG. Estimates of CVI and analytical imprecision can also be
(SEQC), Madrid, Spain. https://orcid.org/0000-0002-3171-1505 (J. Díaz- used to calculate reference change values (RCVs) to assess
Garzón) the probability that a difference between two consecutive
Niels Jonker, Certe, Wilhelmina Ziekenhuis Assen, Assen, The Netherlands results in an individual can be explained by analytical and
Aasne K. Aarsand, Norwegian Organization for Quality Improvement of within-subject biological variation [1]. Also, a model for
Laboratory Examinations (Noklus), Haraldsplass Deaconess Hospital,
calculating personalized reference intervals (prRI) based on
Bergen, Norway; and Department of Medical Biochemistry and
Pharmacology, Norwegian Porphyria Centre, Haukeland University BV data has recently been published [2]. This utilizes previ-
Hospital, Bergen, Norway ous test results of a subject in a steady-state condition and
742 Sandberg et al.: Biological variation: recent development and future challenges

estimates of the individual’s BV, derived either for the CV-ANOVA [9] and the Bayesian approach [10], are based on
relevant population or for the individual. data derived from experimental studies, in principle using
Different sources of BV data have been available in the the study design of Fraser and Harris [8] to estimate BV
last decades [1]. In the 1st European Federation of Clinical components. The third model uses retrospectively collected
Chemistry and Laboratory Medicine (EFLM) Strategic Con- data, extracted from laboratory databases, for data mining
ference in 2014, it was, however, recognised that much of the studies [11–13].
available BV data were compromised because of uncertainty
around the estimates or other factors affecting fitness for
purpose. Thus, this limited the utility of these data for the The CV-ANOVA approach to calculate BV
many different BV applications. Furthermore, it highlighted estimates
the need for critical appraisal of existing BV data as well as
for new studies to generate high-quality data [3]. As a result Different ANOVA methods have over time been used to
of the Strategic Conference, different EFLM Working Groups derive BV data, historically mostly standard ANOVA and
and Task Groups set in motion initiatives to improve on the ln-ANOVA. The standard ANOVA, based on a nested ANOVA
availability of quality-assessed BV data and their applica- approach, is performed on the raw data, after exclusion of
tions [4]. In this review, an update on the achievements and outliers, ensuring a steady state situation by trend analysis
developments driven by the EFLM Working Group for Bio- and assessment of variance homogeneity. The resulting
logical Variation (WG-BV) [5], the EFLM Task Group for the within-subject standard deviation (SDI) estimate is divided
Biological Variation Database (TG-BVD) [6] and other key by the total mean to derive the CVI estimate. In the ln-ANOVA
players in the field of BV the last 10 years will be presented, method, data are first ln-transformed and the nested
with suggestions for future developments, focussing on the ANOVA is performed on the transformed results. The esti-
following: mated components from the ANOVA are then transformed
1. New mathematical models to calculate BV estimates. back and become CV values on the original scale. A new
2. Delivery of high-quality BV data from the highly powered ANOVA-method, a CV-ANOVA method – the Røraas
European Biological Variation Study (EuBIVAS). method – was published in 2016 [9] and has since then
3. The Biological Variation Data Critical Appraisal Checklist been widely used. It is based on CV transformation with
(BIVAC), a standard for evaluating BV publications. normalization/standardisation of each person’s data by
4. The EFLM Biological Variation Database. dividing by that person’s mean value, and then performing
5. Standards for reporting of BV data and BV terminology. ANOVA on these results, which provides estimates of
6. New applications of BV data. analytical and within-subject variation in the form of CV
values. However, this does not provide an estimate of the
CVG, since each individual has a mean value of 1. Thus, by
this approach, the CVG must be estimated by the standard
New mathematical models to or alternatively by the ln-ANOVA approach, if data are
calculate BV estimates not normally distributed. These three different methods,
standard ANOVA, ln-ANOVA and CV-ANOVA have been
BV studies have historically been undertaken as prospective compared in computer simulations for different distribu-
experimental studies. The methodology for this was first tions of raw data [9], i.e. normal distribution, ln-normal
described by Fraser and Harris [7]. In essence, a group of distribution and mirrored ln-normal distribution. The
reference individuals is selected and sampled at regular CV-ANOVA method performed well for all types of distri-
time intervals, with a strictly controlled preanalytical phase butions, as opposed to the other ANOVA methods.
and then analysed under standardized conditions. Duplicate
analysis of the samples is recommended to estimate the CVA
component directly. The resulting data are examined for The Bayesian approach of calculating BV
clinical events, trends and statistical outliers as well as estimates
homogeneity. The CVA, CVI, and CVG estimates are thereafter
derived by traditional statistical approaches such as e.g. An approach utilizing Bayesian statistics to derive BV data,
ANOVA. More details about this study design and traditional based on the same experimental study design as that rec-
calculations can be found in [8]. ommended by Fraser and Harris [7], was published by
The last decade, three new approaches have been Røraas in 2019 [10]. This Bayesian model disregards the
developed to estimate BV components. The two first models, assumption of normality and is more robust to extreme
 Sandberg et al.: Biological variation: recent development and future challenges 743

observations by using adaptive Student-t distributions Studies detailing utilizing data derived from laboratory
instead of normal distributions. The advantage over tradi- databases for estimating e.g. CVI and RCVs for e.g. routine
tional methods such as ANOVA is that laborious statistical chemistry, endocrine and haematology tests have reported
operations, associated with possible subjectivity in data results more or less equivalent to those delivered by stan-
trimming to achieve homogeneity and exclusion of outliers, dard, prospective methods [11, 12, 14, 15]. However, most such
are not required. Furthermore, the Bayesian model delivers studies do not report measures of uncertainty, which is a
individual within-person BV estimates (CVP) that can be used great limitation. However, one recently published study has
to explore heterogeneity of data, asses if the data can be now proposed how this can be done [15]. Further work is
generalised for the whole population, assess relevant sub- required to identify the strengths and limitations of this
groups or to identify individuals not belonging to the group. retrospective, data mining approach to deliver BV data
It is then also possible to assess correlations between the and to meet challenges in measurand data distributions
CVP and e.g. age or homeostatic set points and to calculate (e.g. skewed data), and other aspects allowing to fine-tuning
personalized reference intervals, prRI, directly, as described this approach further.
later. The model utilizes, if available, prior knowledge to
make more precise inference from previous performed
studies. This is particularly valuable if previous data on BV Delivery of high-quality BV
for the measurand, or related measurands, are available.
Some studies that have applied both ANOVA and Bayes
estimates – the European Biological
methods on the same data set have reported obtaining Variation Study (EuBIVAS)
similar BV results with both methods, however, this may
depend on distribution of the data [10, 12]. The EFLM WG-BV decided in 2014 to design and establish the
EuBIVA with the aim of providing updated high-quality BV
estimates for many measurands, derived from a highly
powered and rigorously executed BV study [16]. Briefly,
Using previously analysed data derived from EuBIVAS involved six European laboratories (Milan, Italy;
laboratory information management Bergen, Norway; Madrid, Spain; Padua, Italy; Istanbul,
systems to calculate BV estimates Turkey; Assen, The Netherlands). Following a detailed
screening of enrolled participants, 91 healthy volunteers
This model uses a different approach for the generation (38 males and 53 females; age range, 21–69 years) were
of BV data, unlike the two previous models, as it utilizes included in the study. All involved laboratories followed the
already available data, analysed as part of routine follow-up same protocol for the pre-analytical phase, with all partici-
of patients, held in laboratory databases. This is achieved pants compiling an enrolment questionnaire to verify their
by extracting results from patient cohorts from laboratory health status and to collect information regarding their
information management systems (LIMS) consisting of a lifestyle. Fasting blood samples were drawn for 10 consec-
large number of individuals, where two or more samples utive weeks at each participating laboratory. The samples
routinely have been analysed for the same measurand. were stored at −80 °C until shipped on dry-ice and analysed
Assessment of such data collected for diagnostic or moni- in duplicate for a high number of measurands at the San
toring purposes also provides an opportunity to assess sub- Raffaele Hospital in Milan [16].
groups including different states of health, the effect of time BV estimates for all participants, males, females, and if
between sampling, or other factors, without the efforts of considered relevant for the specific measurands for female
prospectively collecting large data sets. For many analytes subgroups (above and below 50 years, in reflection of
requested for a patient, the concentrations of the analytes menopausal status), were for most EuBIVAS measurands
are not impacted by non-relevant pathology and may estimated by the CV-ANOVA method [9]. The EuBIVAS
represent values obtained for the healthy population. This approach presents a number of benefits. Firstly, it is suffi-
“big data” or “data mining” approach is particularly useful if ciently powered to enable subgroup analysis. This has
the measurand in question is not present in apparently allowed gender specific data and for some measurands also
healthy subjects (e.g. unusual proteins found in myeloma), if menopausal age specific data [17–24] which until this point
the concentration of the measurand is significantly different have been unavailable for many measurands. Secondly, the
from what is found in healthy individuals (e.g. HbA1c in BV estimates have been obtained based on current best-
diabetes mellitus) or if it is unacceptable or unethical to practice recommendations for study design [25, 26]. EuBIVAS
collect specimens from individuals, for example children. has so far delivered BV estimates for 81 different
744 Sandberg et al.: Biological variation: recent development and future challenges

measurands; with data for two target measurands (cardiac approach of extracting data from LIMS is a more pragmatic
troponin and serum creatinine) obtained using two different approach to delivery of required data sets, rather than
analytical methods [20, 27]. The results of most of the studies highly powered experimental studies like the EuBIVAS.
are summarized in [28]. For 10 of these measurands, no
previous BV studies had been carried out and further studies
for measurands for where there as of yet are no available
BV data, are also on the way.
The Biological Variation Data
For haematology measurands, given the requirement Critical Appraisal Checklist (BIVAC)
for fresh whole blood, the analytical approach used by
EuBIVAS was not possible. Two different projects, in Italy A high number of BV studies has been published in the last
[29–31] and Turkey [32], have been carried out to deliver four decades. Though historical BV studies were typically
high-quality BV data for complete blood count, utilizing designed according to that time’s standard, many do not to
fresh samples. While the analytical approach necessarily fulfil today’s standard for study design and execution, or
varies from that employed within EuBIVAS, all other ele- they utilized analytical methods that are now considered
ments of these studies were similar. In addition to updating obsolete. Thus, much of the historical BV data may be
the BV estimates of complete blood count parameters compromised with uncertainty or considered unfit for use
currently in use, these studies also provided BV data for today. Following the 1st Strategic Conference of EFLM
some parameters where this was lacking. defining APS in November 2014, the EFLM TG-BVD was
The EuBIVAS study might be described as an exemplar established within the WG-BV, with the objective to appraise
of the classical approach to delivery and reporting for BV the quality of BV data that is publicly available. The result of
data. It has delivered BV estimates for many measurands this work was, among others, the Biological Variation Data
that are lower than those previously reported. This is Critical Appraisal Checklist (BIVAC), a standard for evalu-
probably due to tight control of pre-analytical factors, the ating BV studies [25]. The BIVAC is designed to assess the
use of modern examination methodology, and the critically quality of BV publications by addressing essential elements
important application of correct statistical approach to data that may impact upon veracity and utility of the BV esti-
handling with e.g. assessment of outliers, variance homo- mates. It consists of 14 quality items and focuses on the effect
geneity and trend, according to best practice recommenda- of study design, the measurement procedure and statistical
tions. The absence of clear differences between the subject handling of data on BV estimates. The individual quality
cohorts from Turkey, Norway, The Netherlands, Spain, and items can be awarded scores A, B, C and for some essential
Italy, confirms that the obtained data are transportable items, also D, indicating decreasing compliance with the
across health care systems indicating that they may be used checklist. The lowest score obtained for any quality item
to deliver APS for systems to be used internationally. It is decides the overall grade. A BIVAC grade A indicates that the
reassuring that this consistency was also demonstrated publication is fully compliant with all BIVAC quality items. If
using principal component analysis, an unsupervised the lowest score for any quality item is a B, then the overall
machine learning approach [33]. It should be noted that a grade is a B and similarly C or D if the lowest score is a C or D,
possible disadvantage of the meticulously generated EuBI- respectively. Studies receiving a D grade are not considered
VAS data could be that they are “too good”, having been fit for purpose. Systematic reviews of BV studies for many
delivered under rigorously controlled conditions that do not different measurands have shown that the majority of his-
reflect routine practice. The EuBIVAS estimates may thus be torical studies receive a BIVAC grade C, mostly related to
more useful for setting APS than for RCV calculations or to statistical items such as analysis of variance homogeneity
calculate personalized reference intervals. Instead, data and outliers [23, 40–48]. Since the publication of BIVAC in
mining from results in LIMS might deliver BV data that 2017, however, an increasing number of BV studies fulfil
more accurately reflect the “real life” situation that can be the criteria for a high-quality study and are fully BIVAC
used for e.g. RCV [14] and personalized reference intervals. compliant, as illustrated by the papers included in the Clin-
Furthermore, there is still a great lack of studies on different ical Chemistry and Laboratory Medicine special issue on BV
age groups and states of health, even though high-quality in 2022 [4, 49]. The BIVAC also provides, in combination with
studies for population subgroups such athletes and pregnant the Biological Variation Data Reporting Checklist [26], a
women have also been published in the last 10 years [34–39]. framework that may help those planning BV studies both to
For many states of health, however, it may be that using the perform and publish their study in an appropriate manner.
 Sandberg et al.: Biological variation: recent development and future challenges 745

The EFLM Biological Variation for assessing the quality of such kind of studies is lacking.
Such a standard will be developed, prior to including these
Database types of studies in the meta-analysis that provides global BV
estimates in the database.
Historically, different sources of BV data have been avail-
able, also online. After the Strategic Conference, one of the
main objectives of the WG-BV and TG-BVD was to establish a
new database with quality-assessed BV data, available to
Reporting of BV data and BV
users worldwide. The “EFLM Biological Variation Database” terminology
was launched during the EuroMedLab in Barcelona in May
2019 and is available at www.biologicalvariation.eu. To The Biological Variation Data Reporting
populate the database, systematic literature searches for BV Checklist
studies for relevant measurands have been performed and
identified publications appraised by the BIVAC. Both the There have historically been no internationally recognized
BIVAC scores as well as a BV minimum data set, encom- standards for production, reporting and transmission of
passing around 30 descriptive items are published in the BV data. The WG-BV published in 2015 the Biological Varia-
database for all included measurands, thus offering a tion Data Reporting Checklist [26], which identifies key ele-
detailed and updated source of quality-assessed data. Also, a ments required in published BV studies. The reporting
meta-analysis approach to pool estimates from BIVAC checklist is based on the same structure as the STARD [50]
compliant studies with similar study design to provide global and identifies six main items for focus with a number of
BV estimates was developed [25]. In the EFLM Biological sub items including 1) title/abstract/keywords, 2) introduc-
Variation Database, meta-analyses are automatically per- tion, 3) methods, 4) data analysis, 5) results and 6) discussion.
formed for studies of acceptable study design (BIVAC grade The sub items have been additionally mapped to a minimum
A–C), when performed in healthy adults with more than two data set domains previously identified by the WG.
samples collected per individual and with biweekly to Studies complying with this checklist and the BIVAC can
monthly samplings. The meta-analysis report estimates for be considered fit for purpose, include essential statistical
CVI and CVG with confidence intervals (CI), based on a analyses such as outlier and variance homogeneity testing,
weighted median approach where the study design as well use recommended terminology and report BV estimates
as the BIVAC grade is taken into account [25]. As of December accompanied by key metadata. This is essential as these BV
2022, global BV results derived by meta-analysis have been data are reference data; they need to be applied with care,
published for 139 measurands, as well as several thousand with understanding of their provenance and intrinsic
detailed BV data sets and more than 560 BV studies (Figure 1). characteristics if they are to be transported safely and
In addition to presenting BV estimates for each evaluated effectively into clinical practice across health care systems.
paper and the summary BV estimates derived by meta- The EFLM BV groups have developed further materials
analysis, the database also provides automatic calculation of to support the delivery and reporting of studies. This work
RCV and the most common APS used in laboratory practice. has initially focused on the “classical” prospective experi-
These applications are available when accessing the global mental approach to delivery of BV data. Access is available
meta-analysis BV estimates (Figure 1) [8]. online in the form of an interactive mind map [51] and an
In today’s database, global BV estimates are published overview of the structure is presented in Figure 2. The con-
based on data derived from healthy adults with biweekly to tent includes embedded links and documents that can be
monthly sampling. However, data from studies performed in used to assess the veracity of data from existing BV studies
different states of health, other age groups and sampling and to enable the design of new BV experiments. It reflects
intervals are also included in the database. In the future, what is believed to be the current state of the art and is in a
meta-analyses will also be provided for these different set- format that draws parallels with that of STARD to be used
tings and states of health. As of yet, BV data derived from as a possible publication structure to guide researchers
data mining of retrospectively collected data have not been (Figure 2). This represents an initial iteration as there is a
included in the meta-analysis, as these are, with one very degree of complexity now arising in the subject area as a
recent exception [15], not accompanied by CI, which is one of consequence of the emergence of new approaches to the
the prerequisites to be included. Furthermore, a standard generation of data sets as described in this review. However,
746 Sandberg et al.: Biological variation: recent development and future challenges

Figure 1: Screenshot of the front page of the online EFLM Biological Variation Database [8]. In this database, detailed information for biological variation
data sets derived from critically appraised biological variation studies, as well global CVI and CVG from meta-analysis are published and automatically
updated whenever new data are added. For each measurand analytical performance specifications (APS) and reference change values (RCV) are also
automatically calculated.

Figure 2: Screenshot of the online proposed publication structure for studies of biological variation data. The published version [51] which is a further
development of [26] allows the topics in the topic boxes to be expanded to reveal further subtopics with notes key reference sources embedded as links.
In addition to the The Biological Data Critical Appraisal Checklist (BIVAC) is also embedded with scoring criteria for the different quality items [25].
 Sandberg et al.: Biological variation: recent development and future challenges 747

through time the map can be adapted dynamically to address Analytical performance specifications
emergent developments, thus enabling assurance of quality
of newer data sets. One of the challenges faced by those In the consensus document from the 1st Strategic Conference
assessing (e.g. BIVAC scoring) publications of BV data is that of the EFLM in 2014 it was agreed that APS can be established
key metadata required to enable the process are missing or using three different models [54]. Model 1 is based on the
poorly described. In practice, this means well executed effect of analytical performance on clinical outcomes. Model
studies deliver data that are devalued and non-translatable 2 is based on components of BV and Model 3 is based on state
as a consequence of deficiencies in reporting. The mind of the art. In the Biological Variation Database, model 2
map will enable users to check that existing and new based APS utilizing the global BV estimates are automatically
studies are compliant with a proposed structure, reporting calculated for imprecision, bias, total error and measure-
a considered minimum data set that includes the embedded ment uncertainty, presented as minimum, desirable and
BIVAC. Further development needs to include consider- optimal [8]. These first three are presented in the database
ation of data mining approaches and differing data [6] with the pro and con for each [8] and are also described
management. in detail elsewhere [1]. In addition, APS for maximum
allowable standard measurement uncertainty (MAu) is
included. When using MAu bias should, in principle, be
Terminology for components of BV eliminated, and all the remaining sources of variation
added linearly as variances. Accordingly, the MAu can be
Over years, a wide range of terms and symbols have been set as 0.5 × CVI, and the maximum expanded allowable
used to describe the components of BV in published litera- measurement uncertainty (MAU) as k × 0.5 × CVI. The “k” is
ture. In 2015, a recommendation on harmonized use of the coverage factor, for example, 2 or 3, to obtain a certain
terminology in the field of BV was published [52] and these confidence level (95 or 99). The most used coverage factor is
definitions and abbreviations, as used in this review, are 2. Thus, MAU can be calculated as MAU<2 × 0.5 * CVI.
recommended the EFLM WG-BV and TG-BVD as well as by However, when calculating APS, we also must consider
Clinical Chemistry and Laboratory Medicine [53]. Addition- for which situations or scenario we are going to use them.
ally, in 2016, the EFLM WG-BV proposed and got accepted Different APS will probably be relevant if these are applied
the establishment of a Medical Subject Heading (MeSH) term in e.g. internal quality control, in External Quality Assurance
for “Biological Variation” in the National Library of Medi- (EQA) schemes, to evaluate lot to lot variations, in clinical
cine. This MeSH term has been available from December guidelines. To set correct APS is an ongoing and difficult
2017, thus facilitating systematic searches for BV publica- process. The Cutting Edge of Laboratory Medicine (CELME)
tions for current publications. conference, to be arranged in 2023 [55], will have the title
“Analytical performance specifications: Moving from
models to practical recommendations” and aims to take
New application of BV data this area forward.

Reference change values and index of

individuality Personalized reference intervals

Most BV application have been published several decades When using laboratory tests for the diagnosis and moni-
ago, as reviewed in [1]. This includes the use of the index of toring of patients, a reliable reference to which the results
individuality, which is used to assess the value of using can be compared is required. Today, most reference data are
population-based reference interval or RCV for monitoring derived from the population, derived either by experimental
of an individual. The RCV enumerates the value that the reference interval studies or by utilizing laboratory test
difference between two test results in the same individual results stored in LIMS. However, such reference data have
can be, with a certain probability, explained by analytical limitations when used as the reference for an individual,
and within-subject biological variation. The RCV has earlier especially for measurands with a low index of individuality.
been calculated as a symmetrical value, whereas it should in Thus, patients’ test results preferably should be compared
fact be calculated with asymmetrical limits derived from a ln with their own, individualized reference intervals, i.e. a
RCV method [1]. In the EFLM Biological Variation Database personalized RI (prRI). Recently, a new model for calculating
[8], the published RCVs are based on the asymmetrical prRI has been developed, utilizing BV data [2, 56–60]. The
approach. model is based on the homeostatic set point (HSP) and the
748 Sandberg et al.: Biological variation: recent development and future challenges

total variation around the HSP (TVset) of the analytes. To Concluding remarks
derive the prRI, firstly the HSP of the measurand for the
individual being assessed, must be estimated. This is ach- In the later years, there has been a great progress in
ieved by calculating the mean of previous test results the methodology of deriving, calculating, estimating and
obtained in a steady state situation. Thereafter, the prRI reporting BV data and derived parameters. The results of
can be constructed by prediction intervals in two ways: 1) many of these initiatives have been incorporated in the
using the person’s own within-person biological variation EFLM Biological Variation Database, which provides
estimate (SDP) or 2) using the within-subject biological quality-assessed BV data, with automatically calculated
variation estimate (SDI) derived from a population similar to BV applications, freely available for users worldwide.
the person being monitored. Furthermore, new applications for use of BV data, such as
The prRI can thereafter be calculated as prRI and MAU have been developed. Clinical Chemistry and
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
(n + 1) Laboratory Medicine has been a key partner for the EFLM
prRI = HSP ± k × (SD2A + SD2B ) BV groups and others in driving the area of BV forward,
n
including by publishing a Special Issue on Biological Vari-
where HSP is the homeostatic set point; k is a constant
ation in 2022. This special issue contains 21 articles on
depending on the type of distribution (normal distribution
BV and related aspects [4, 49]. We indeed congratulate
(z) in case of population-based SDI and t distribution in case of
Clinical Chemistry and Laboratory Medicine with its 60th
SDP) and the probability, n is the number of previous test
anniversary.
results and the SDB refers to either the SDP or the SDI and the
For the future, it is important to take into account the
SDA the analytical variation. To get a reliable estimate of the
fact that there is a lack of robust and high-quality BV for
SDP, more than five previous test results from a steady state
many measurands, population groups and settings, and
situation for an individual is required. This will often be a
furthermore, it is important to be aware that some BV data
limitation, and it may therefore be easier to construct the prRI
may be associated with a large uncertainty that should be
using the SDI. In this setting, at least three samples are required
considered whenever these data are used for different BV
to estimate the HSP [2]. The two models will give rather similar
application. Importantly, we must in the future address
prRIs if the presuppositions of the models are fulfilled.
how BV data can best benefit screening, diagnosing and
It is well known that clinicians use the limits of a
monitoring of patients and thereby improve patient
population-based reference intervals to act on patients’
outcomes.
results, although these limits are not the same as clinical
decision/action limits [61]. It is possible to estimate prRIs for Research funding: None declared.
all measurands where there are repeated measurements Author contributions: All authors have accepted respon-
from a steady state situation and as such, these may repre- sibility for the entire content of this manuscript and
sent personalized action limits. The calculation of prRI can approved its submission.
be integrated in laboratory information systems and has the Competing interests: Authors state no conflict of interest.
potential to be useful in diagnosing and follow up of patients. Informed consent: Not applicable.
However, there are unresolved questions considering the Ethical approval: Not applicable.
clinical use of prRIs e.g., no studies have demonstrated their
actual benefit to patients. Furthermore, clinical decision/
action limits are set independently of reference intervals, References
and they are, unlike reference limits, similar for all kinds of
measurement procedures. A further question that remains is 1. Sandberg S, Røraas T, Aarsand AK. Biological variation and analytical
if clinical decision/action limits should be personalized, such performance specifications (Internet). In: Rifai N, Chiu RWK, Young I,
as for example each person having their own diagnostic cut Burnham CAD, Wittver CT, editors. Tietz textbook of Laboratory
medicine, 7th ed. St Lous: Elsevier; 2022:335–56 pp.
off for a diabetes mellitus diagnosis or for being treated for
2. Coşkun A, Sandberg S, Unsal I, Cavusoglu C, Serteser M, Kilercik M,
hypercholesterolemia. This will probably be dependent if et al. Personalized reference intervals in laboratory medicine: a new
the measurand itself is part of the pathology or if it is only a model based on within-subject biological variation. Clin Chem 2021;67:
marker of pathology. 374–84.
 Sandberg et al.: Biological variation: recent development and future challenges 749

3. Carobene A. Reliability of biological variation data available in an online 20. Ceriotti F, Marco JDG, Fernández-Calle P, Maregnani A, Aarsand AK,
database: need for improvement. Clin Chem Lab Med 2015;53:871–7. Coskun A, et al. The European Biological Variation Study (EuBIVAS):
4. Sandberg S, Carobene A, Aarsand AK. Biological variation – eight years weekly biological variation of cardiac troponin I estimated by the use of
after the 1st Strategic Conference of EFLM. Clin Chem Lab Med 2022;60: two different high-sensitivity cardiac troponin I assays. Clin Chem Lab
465–8. Med 2020;58:1741–7.
5. EFLM – working group: biological variation. Available from: https:// 21. Cavalier E, Lukas P, Bottani M, Aarsand AK, Ceriotti F, Coşkun A, et al.
www.eflm.eu/site/page/a/1148 [Accessed 10 Dec 2022]. European biological variation study (EuBIVAS): within- and between-
6. EFLM – task group biological variation database. Available from: subject biological variation estimates of β-isomerized C-terminal
https://www.eflm.eu/site/page/a/1394 [Accessed 10 Dec 2022]. telopeptide of type I collagen (β-CTX), N-terminal propeptide of type I
7. Fraser CG, Harris EK. Generation and application of data on biological collagen (PINP), osteocalcin, intact fibroblast growth factor 23 and
variation in clinical chemistry. Crit Rev Clin Lab Sci 1989;27:409–37. uncarboxylated-unphosphorylated matrix-gla protein—a cooperation
8. Aarsand AK, Webster C, Coskun A, Gonzales-Lao E, Diaz-Garzon J, between the EFLM working group on biological variation and the
Roraas T, et al. EFLM biological variation database. Available from: international osteoporosis foundation-international federation of
https://biologicalvariation.eu [Accessed 10 Dec 2022]. clinical chemistry committee on bone metabolism. Osteoporos Int
9. Røraas T, Støve B, Petersen PH, Sandberg S. Biological variation: the 2020;31:1461–70.
effect of different distributions on estimated within-person variation 22. Clouet-Foraison N, Marcovina SM, Guerra E, Aarsand AK, Coşkun A,
and reference change values. Clin Chem 2016;62:725–36. Díaz-Garzón J, et al. Analytical performance specifications for
10. Røraas T, Sandberg S, Aarsand AK, Støve B. A Bayesian approach to lipoprotein(a), apolipoprotein B-100, and apolipoprotein A-I using the
biological variation analysis. Clin Chem 2019;65:995–1005. biological variation model in the EuBIVAS population. Clin Chem 2020;
11. Jones GRD. Estimates of within-subject biological variation derived 66:727–36.
from pathology databases: an approach to allow assessment of the 23. Carobene A, Guerra E, Marqués-García F, Boned B, Locatelli M,
effects of age, sex, time between sample collections, and analyte Coşkun A, et al. Biological variation of morning serum cortisol: updated
concentration on reference change values. Clin Chem 2019;65: estimates from the European Biological Variation Study (EuBIVAS) and
579–88. meta-analysis. Clin Chim Acta 2020;509:268–72.
12. Røys EÅ, Guldhaug NA, Viste K, Jones GD, Alaour B, Sylte MS, et al. Sex 24. Carobene A, Aarsand AK, Coşkun A, Díaz-Garzón J, Locatelli M,
hormones and adrenal steroids: biological variation estimated using Fernandez-Calle P, et al. Biological variation of serum iron from the
direct and indirect methods. Clin Chem 2023;69:100–9. European biological variation study (EuBIVAS). Clin Chem Lab Med
13. Marques-Garcia F, Boned B, González-Lao E, Braga F, Carobene A, 2023;61:e57–60.
Coskun A, et al. Critical review and meta-analysis of biological 25. Aarsand AK, Røraas T, Fernandez-Calle P, Ricos C, Díaz-Garzón J,
variation estimates for tumor markers. Clin Chem Lab Med 2022;60: Jonker N, et al. The biological variation data critical appraisal checklist:
494–504. a standard for evaluating studies on biological variation. Clin Chem
14. Cembrowski GS, Lyon AW, McCudden C, Qiu Y, Xu Q, Mei J, et al. 2018;64:501–14.
Transformation of sequential hospital and outpatient laboratory data 26. Bartlett WA, Braga F, Carobene A, Coskun A, Prusa R, Fernandez-Calle P,
into between-day reference change values. Clin Chem 2022;68: et al. A checklist for critical appraisal of studies of biological variation.
595–603. Clin Chem Lab Med 2015;53:879–85.
15. Marqués-García F, Nieto-Librero A, González-García N, Galindo- 27. Carobene A, Marino I, Coskun A, Serteser M, Unsal I, Guerra E, et al. The
Villardón P, Martínez-Sánchez LM, Tejedor-Ganduxé X, et al. Within- EuBIVAS project: within-and between-subject biological variation data
subject biological variation estimates using an indirect data mining for serum creatinine using enzymatic and alkaline picrate methods and
strategy. Spanish multicenter pilot study (BiVaBiDa). Clin Chem Lab implications for monitoring. Clin Chem 2017;63:1527–36.
Med 2022;60:1804–12. 28. Carobene A, Aarsand AK, Bartlett WA, Coskun A, Diaz-Garzon J,
16. Carobene A, Strollo M, Jonker N, Barla G, Bartlett WA, Sandberg S, et al. Fernandez-Calle P, et al. The European Biological Variation Study
Sample collections from healthy volunteers for biological variation (EuBIVAS): a summary report. Clin Chem Lab Med 2022;60:505–17.
estimates’ update: a new project undertaken by the Working Group on 29. Buoro S, Carobene A, Seghezzi M, Manenti B, Dominoni P, Pacioni A,
Biological Variation established by the European Federation of Clinical et al. Short- and medium-term biological variation estimates of red
Chemistry and Laboratory Medicine. Clin Chem Lab Med 2016;54: blood cell and reticulocyte parameters in healthy subjects. Clin Chem
1599–608. Lab Med 2018;56:954–63.
17. Aarsand AK, Díaz-Garzón J, Fernandez-Calle P, Guerra E, Locatelli M, 30. Buoro S, Seghezzi M, Manenti B, Pacioni A, Carobene A, Ceriotti F, et al.
Bartlett WA, et al. The EuBIVAS: within- and between-subject Biological variation of platelet parameters determined by the Sysmex
biological variation data for electrolytes, lipids, urea, uric acid, total XN hematology analyzer. Clin Chim Acta 2017;470:125–32.
protein, total bilirubin, direct bilirubin, and glucose. Clin Chem 2018; 31. Buoro S, Carobene A, Seghezzi M, Manenti B, Pacioni A, Ceriotti F, et al.
64:1380–93. Short- and medium-term biological variation estimates of leukocytes
18. Carobene A, Aarsand AK, Guerra E, Bartlett WA, Coskun A, Díaz- extended to differential count and morphology-structural parameters
Garzón J, et al. European biological variation study (EuBIVAS): within- (cell population data) in blood samples obtained from healthy people.
and between-subject biological variation data for 15 frequently Clin Chim Acta 2017;473:147–56.
measured proteins. Clin Chem 2019;65:1031–41. 32. Coskun A, Carobene A, Kilercik M, Serteser M, Sandberg S, Aarsand AK,
19. Bottani M, Banfi G, Guerra E, Locatelli M, Aarsand AK, Coşkun A, et al. et al. Within-subject and between-subject biological variation
European Biological Variation Study (EuBIVAS): within- and between- estimates of 21 hematological parameters in 30 healthy subjects. Clin
subject biological variation estimates for serum biointact parathyroid Chem Lab Med 2018;58:618–28.
hormone based on weekly samplings from 91 healthy participants. Ann 33. Carobene A, Campagner A, Uccheddu C, Banfi G, Vidali M, Cabitza F.
Transl Med 2020;8:855. The multicenter European Biological Variation Study (EuBIVAS): a new
750 Sandberg et al.: Biological variation: recent development and future challenges

glance provided by the Principal Component Analysis (PCA), a machine 46. Coskun A, Braga F, Carobene A, Ganduxe XT, Aarsand AK, Fernandez-
learning unsupervised algorithms, based on the basic metabolic panel Calle P, et al. Systematic review and meta-analysis of within-subject and
linked measurands. Clin Chem Lab Med 2022;60:556–68. between-subject biological variation estimates of 20 haematological
34. Diaz-Garzon J, Fernandez-Calle P, Aarsand AK, Sandberg S, Coskun A, parameters. Clin Chem Lab Med 2020;58:25–32.
Carobene A, et al. Long-term within- and between-subject biological 47. Coşkun A, Aarsand AK, Braga F, Carobene A, Díaz-Garzón J, Fernandez-
variation of 29 routine laboratory measurands in athletes. Clin Chem Calle P, et al. Systematic review and meta-analysis of within-subject and
Lab Med 2022;60:618–28. between-subject biological variation estimates of serum Zinc, Copper
35. Diaz–Garzon J, Fernandez-Calle P, Aarsand AK, Sandberg S, Buno A. and Selenium. Clin Chem Lab Med 2022;60:479–82.
Biological variation of venous acid-base status measurands in athletes. 48. González-Lao E, Corte Z, Simón M, Ricos C, Coskun A, Braga F, et al.
Clin Chim Acta 2021;523:497–503. Systematic review of the biological variation data for diabetes related
36. Kristoffersen AH, Petersen PH, Sandberg S. A model for calculating the analytes. Clin Chim Acta 2019;488:61–7.
within-subject biological variation and likelihood ratios for analytes 49. Sandberg S, Carobene A, Aarsand AK, Issue editors. Biological variation
with a time-dependent change in concentrations; exemplified with the – 8 years after the 1st Strategic Conference of EFLM (special issue). Clin
use of D-dimer in suspected venous thromboembolism in healthy Chem Lab Med 2022;60:462–644.
pregnant women. Ann Clin Biochem 2012;49:561–9. 50. STARD guidelines. Available from: https://www.equator-network.org/
37. Kristoffersen AH, Petersen PH, Bjørge L, Røraas T, Sandberg S. reporting-guidelines/stard/ [Accessed 10 Dec 2022].
Concentration of fibrin monomer in pregnancy and during the 51. Biological Variation Data Reporting Checklist. Available from:
postpartum period. Ann Clin Biochem 2019;89:73–9. https://www.wabthings.co.uk/biological-variation [Accessed 2 Dec 2022].
38. Kristoffersen AH, Petersen PH, Røraas T, Sandberg S. Estimates of 52. Simundic AM, Kackov S, Miler M, Fraser CG, Petersen PH. Terms and
within-subject biological variation of protein C, antithrombin, protein S symbols used in studies on biological variation: the need for
free, protein S activity, and activated protein C resistance in pregnant harmonization. Clin Chem 2015;61:438–9.
women. Clin Chem 2017;63:898–907. 53. Plebani M, Padoan A, Lippi G. Biological variation: back to basics. Clin
39. Kristoffersen AH, Petersen PH, Bjørge L, Røraas T, Sandberg S. Within- Chem Lab Med 2015;53:155–6.
subject biological variation of activated partial thromboplastin time, 54. Sandberg S, Fraser CG, Horvath AR, Jansen R, Jones G, Oosterhuis W,
prothrombin time, fibrinogen, factor VIII and von Willebrand factor in et al. Defining analytical performance specifications: consensus
pregnant women. Clin Chem Lab Med 2018;56:1297–308. statement from the 1st Strategic Conference of the European
40. Diaz-Garzón J, Fernandez-Calle P, Minchinela J, Aarsand AK, Bartlett WA, Federation of Clinical Chemistry and Laboratory Medicine. Clin Chem
Aslan B, et al. Biological variation data for lipid cardiovascular risk Lab Med 2015;53:833–5.
assessment biomarkers. A systematic review applying the biological 55. 5th Symposium CELME 2023. Available from: http://www.celme2023.cz
variation data critical appraisal checklist (BIVAC). Clin Chim Acta 2019; [Accessed 2 Dec 2022].
495:467–75. 56. Coskun A, Sandberg S, Unsal I, Cavusoglu C, Serteser M, Kilercik M,
41. Fernández-Calle P, Díaz-Garzón J, Bartlett W, Sandberg S, Braga F, et al. Personalized reference intervals: using estimates of within-
Beatriz B, et al. Biological variation estimates of thyroid related subject or within-person biological variation requires different
measurands – meta-analysis of BIVAC compliant studies. Clin Chem statistical approaches. Clin Chim Acta 2021;524:201–2.
Lab Med 2022;60:483–93. 57. Coskun A, Theodorsson E, Oosterhuis WP, Sandberg S. Measurement
42. Diaz-Garzon J, Fernandez-Calle P, Sandberg S, Özcürümez M, uncertainty for practical use. Clin Chim Acta 2022;531:352–60.
Bartlett WA, Coskun A, et al. Biological variation of cardiac troponins in 58. Coskun A, Sandberg S, Unsal I, Yavuz FG, Cavusoglu C, Serteser M, et al.
health and disease: a systematic review and meta-analysis. Clin Chem Personalized reference intervals – statistical approaches and
2021;67:256–64. considerations. Clin Chem Lab Med 2022;60:629–35.
43. Carobene A, Lao EG, Simon M, Locatelli M, Coşkun A, Díaz-Garzón J, 59. Coskun A, Sandberg S, Unsal I, Serteser M, Aarsand AK. Personalized
et al. Biological variation of serum insulin: updated estimates from the reference intervals: from theory to practice. Crit Rev Clin Lab Sci 2022;
European Biological Variation Study (EuBIVAS) and meta-analysis. Clin 59:1–16.
Chem Lab Med 2022;60:479–82. 60. Carobene A, Banfi G, Locatelli M, Vidali M. Personalized Reference
44. Jonker N, Aslan B, Boned B, Marqués-García F, Ricós C, Alvarez V, et al. Intervals: from the statistical significance to the clinical usefulness. Clin
Critical appraisal and meta-analysis of biological variation estimates for Chim Acta 2021;524:203–4.
kidney related analytes. Clin Chem Lab Med 2022;60:469–78. 61. Ozarda Y, Sikaris K, Streichert T, Macri J,
45. Ricos C, Fernandez-Calle P, Gonzales-Lao E, Simon M, Diaz-Garzon J, (C-RIDL) IC on R intervals and DL. Distinguishing reference intervals
Boned B, et al. Critical appraisal and meta-analysis of BV studies on and clinical decision limits – a review by the IFCC Committee on
glycosylated albumin, glucose, and HbA1c. Adv Lab Med 2020;1: Reference Intervals and Decision Limits. Crit Rev Clin Lab Sci 2018;55:
23–9. 420–31.