Scandinavian Journal of Clinical & Laboratory Investigation, 2012; 72: 169–172

ORIGINAL ARTICLE

Significance of a change between two consecutive measured values

ARIADNA PADRÓ-MIQUEL, RAÜL RIGO-BONNIN & XAVIER FUENTES-ARDERIU

Laboratori Clínic, Hospital Universitari de Bellvitge, L’Hospitalet de Llobregat, Catalonia, Spain

Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 05/13/15

Abstract
Background. The interpretation of a change between two consecutive measured values of a given quantity in the same
patient is still controversial and it is not commonly found in clinical laboratory reports. We present here a manageable and
affordable approach for all clinical laboratories to help physicians to interpret these differences. Methods. We selected all
pairs of two consecutive measured values of serum albumin concentration and serum thyrotropin concentration, both
within the biological reference interval, delivered by our clinical laboratory in a 2-year period (2008–2009). We calculated
the relative difference between pairs from which we estimated percentiles 2.5 and 97.5 in order to set an interval defining
the significance of a change. We verified these intervals with data from the year 2010. Results. During the 2-year period,
we found 122 626 consecutive pairs of serum albumin concentration and 9 374 pairs of serum thyrotropin concentration,
with both measured values within our biological reference interval. The intervals defining the significance of a change
between two consecutive measured values were: (⫺14.0 to 17.6)% for the relative differences of albumin concentration,
and (⫺61.4 to 177.8)% for the differences of thyrotropin concentration, which determined 12.7% and 18.4% of signifi-
For personal use only.

cant relative differences, respectively. Similar results were obtained from data for the year 2010. Conclusions. The inter-
val that defines the significance of a change could easily be estimated with the patients’ measured values obtained daily
in the laboratory. Our proposal is more appropriate to each specific population than the conventional intra-individual
biological variation approach.

Key Words: Albumin, critical difference, data interpretation, significant change, thyrotropin

Introduction
The problem of objectively interpreting a change
between two consecutive measured values of a given
quantity in the same patient is still unsolved. While
biological reference limits or discrimination (deci-
sion) values are very important tools in disease diag-
nosis, physicians’ experience and intuition is still the
basis in disease monitoring, as clinical laboratories
seldom report the significance of a change between
successive measurements.
Both the so-called ‘delta check’ and the significance
of a change have been commonly based in the intra-
individual biological variation, but the purpose of them
are quite different: the first one is used as a quality
control tool, whereas the aim of the second one is to
be included in the laboratory report to help physicians
to interpret changes observed in the same patient.
Among the different authors studying how to esti-
mate the intra-individual biological variation of a quan-
tity and what are the applications of this information,
Eugene K. Harris has probably been the most influ-
ential one. First, he published the statistical basis to
estimate the intra-individual biological coefficient of
variation of a group of volunteers [1,2]; afterwards,
he made a proposal for setting ‘analytical goals’ from
biological variation data [3,4]; finally (in this field),
he was the main author of various co-authored
articles on the application of intra-individual bio-
logical variation, combined with the metrological
variation, to detect significant changes when measur-
ing on different days the same quantity in the same
patient [5,6].
However, Harris’ proposal for significant change
detection has two major pitfalls: for many biochemi-
cal quantities, there is not interchangeability of esti-
mates of day-to-day imprecision between commercial
control materials and serum [7], and, generally, the
intra-individual biological coefficients of variation
are very different between individuals [8]. In addi-
tion, there is a lack of agreement in the different

Correspondence: Ariadna Padró-Miquel, Laboratori Clínic, Hospital Universitari de Bellvitge, Feixa Llarga s/n, 08907 L’Hospitalet de Llobregat, Catalonia,
Spain. Tel: ⫹ 34 93 260 75 43. Fax: ⫹ 34 93 260 75 46. E-mail: apadro@bellvitgehospital.cat

(Received 22 March 2011; accepted 30 August 2011)

ISSN 0036-5513 print/ISSN 1502-7686 online © 2012 Informa Healthcare
DOI: 10.3109/00365513.2011.637575
 170 A. Padró-Miquel et al.
published studies of intra-individual biological varia-
tion, showing that the robustness of this tool is seri-
ously questionable [8]. These arguments may be two
of the reasons why clinical laboratories have still not
achieved an agreement on how to decide when a
difference between sequential measured values
should be regarded as significant.
Keeping in mind the essence of the proposal of
Harris, we present here a manageable and affordable
approach to interpret the change between two con-
secutive measured values of a given quantity in the
same patient, which can be easily implemented in
any clinical laboratory. To illustrate our proposal, we
chose serum albumin and thyrotropin concentra-
tions, because these two quantities represent two dif-
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 05/13/15
ferent facts: patients may suffer abrupt changes of
serum thyrotropin concentration as a result of a
treatment, while serum albumin concentration is
known to be more stable.
Material and methods
We selected all patients’ measured values of serum
albumin and thyrotropin concentration delivered by
our clinical laboratory in a 2-year period (2008–
For personal use only.
2009). Among them, we selected pairs of two con-
secutive measured values from the same patient
which were within our corresponding biological ref-
erence interval. We calculated the relative difference
between the pairs of measured values according to
the following equation:
x ⫺x
d (%) ⫽ i i⫺1 ⫻100
xi⫺1
where d (%) is the relative difference between two
consecutive measured values, xi is a measured value
and xi⫺1 is its preceding one.
As in the case of biological reference limits [9],
we estimated non-parametrically percentiles 2.5 and
97.5 of these relative differences in order to set an
interval defining the significance of a change between
two consecutive measured values. All relative differ-
ences that lie outside this interval (changes either up
or down) will be considered significant.
Afterwards, we applied this interval to data from
January to October, 2010. We calculated the propor-
tion (percentage) of significant differences between
consecutive measured values and we used the χ2-test
to compare the proportion obtained with that
observed in the period 2008–2009.
To study the possible relationship between the
relative differences and the time elapsed between
them, we used the Pearson correlation coefficient.
To compare our proposal with the ‘reference
change’ approach from Harris and Yasaka [6], for
each quantity under study, we estimated two ‘refer-
ence change values’ using two intra-individual bio-
logical coefficients of variation [10–13] and our
metrological coefficients of variation.
All statistical analyses were performed with
Microsoft Excel 2003 (Microsoft Corporation, US)
and SPSS statistical software (Version 14.0 for
Windows; SPSS, Chicago, US). Statistical signifi-
cance was set at p ⬍ 0.05.
Serum albumin concentration was measured in
a Modular Hitachi analyzer (Roche Diagnostics,
Mannheim, Germany) using bromocresol green and
molecular absorption spectrometry, whereas serum
thyrotropin concentration was measured in a Modu-
lar E170 analyzer (Roche Diagnostics, Mannheim,
Germany) using an immuno-electrochemilumines-
cent method.
Results
On the first hand, for both quantities, we estimated
the Pearson correlation coefficient that correlates the
time elapsed between two consecutive measured
values with the relative difference between pairs. No
statistical relationship was obtained for serum albu-
min concentration and time (r ⫽ ⫺0.003; p ⫽ 0.177).
On the contrary, the relative difference between pairs
of serum thyrotropin concentration was slightly
related with time (r ⫽ ⫺0.025; p ⫽ 0.001); however,
we assumed that this slight relationship was practi-
cally negligible based on the fact that the Pearson
coefficient was close to zero.
During the 2-year period (2008–2009), 248 717
measured values of serum albumin concentration
and 43 819 of serum thyrotropin concentration were
delivered by our clinical laboratory. We found
122 626 and 9 374 consecutive pairs, respectively,
with both measured values within our biological
reference interval. Representation of the distribution
of d (%) (relative difference) of this set of pairs can
be found in Figure 1.
The estimated intervals (limited by percentiles 2.5
and 97.5) that define the significance of a change
between two consecutive measured values were: (⫺14.0
to 17.6)% for the relative differences of serum albumin
concentration, and (⫺61.4 to 177.8)% for the relative
differences of serum thyrotropin concentration.
On the other hand, the estimated intervals using
the conventional ‘reference change’ approach from
Harris and Yasaka [6] using two different intra-
individual biological coefficients of variation for
each quantity were: (0 ⫾ 7.9)% and (0 ⫾ 19.7)% for
serum albumin concentration, and (0 ⫾ 46.3)% and
(0 ⫾ 70.5)% for serum thyrotropin concentration
(Table I).
Finally, we applied the intervals that define the
significance of a change according to both proposals,
to the measured values obtained in our clinical labo-
ratory from January to October 2010. Table II shows
the comparison between the period when the interval
was set (2008–2009) and the following year. We used
the χ2-test to look for differences in the percentages
of both periods using our approach. We did not find

Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 05/13/15
Figure 1. (A) Distribution of the relative difference obtained with all patients having two consecutive measured values within the biological
reference intervals of serum albumin. (B) Distribution of the relative difference obtained with all patients having two consecutive measured
values within the biological reference intervals of serum thyrotropin concentration.
any difference regarding the serum thyrotropin con-
centration (p ⫽ 0.288), but unexpectedly, the test was
statistically significant for the serum albumin concen-
tration (p ⬍ 0.001). However, this fact could be explained
by the effect size [14] due to the huge number of pairs
taken into account of serum albumin concentration
For personal use only.
compared to those of serum thyrotropin concentra-
tion. In fact, it can be seen at a glance in Table II, that
the percentages were almost the same in both periods
with regard to albumin concentration.
Discussion
Physicians need to know the significance of the differ-
ence between two consecutive measured values of a
given biological quantity in the same patient, as it is
clear from some recommendations made by medical
societies that propose particular medical actions based
on significant changes of quantity values [15,16]. Thus,
ideally, clinical laboratories should provide physicians
with this information, taking into account both the
intra-individual biological variation and the metro-
logical variation.
In the present proposal, after selecting the pairs
of patients’ measured values that lay within the bio-
logical reference interval, we estimated another refer-
ence interval that define the significance of a change
Table I. Estimations of reference change values according to the
‘reference change value’ (RCV1 and RCV2) proposal [6] using
different coefficients of variation (CVBw1 and CVBw2) corresponding
to intra-individual biological variation [10–13].
Biochemical quantity CVM CVBw 1 CVBw 2 RCV1 RCV 2
P – Albumin; mass c. 1.7% 2.3% 6.9% 7.9% 19.7%
P – Thyrotropin; arb. 4.1% 16.2% 25.1% 46.3% 70.5%
subst.c.
Biochemical quantities are described according to the IUPAC and
IFCC recommended syntax [18].
CVM ⫽ day-to-day metrological coefficient of variation.
Significance of a change 171
according to the percentiles 2.5 and 97.5 of the
distribution of their relative differences, as usual in
the reference values theory [9].
Our proposed method was applied at a maxi-
mum elapse of time of 2 years, as follows from the
experimental design, and it may be applicable as
long as the metrological characteristics of the mea-
suring systems do not change.
Other approaches to deciding the significance
of a change use formulas that take into account
published estimations of the intra-individual bio-
logical variation and their metrological variation.
However, there is notorious variability among bio-
logical intra-individual variability studies [8].
In our proposal, the interval defining the sig-
nificance of a change is estimated taking into
account the biological intra-individual variation
corresponding to the population under consider-
ation, whereas the conventional ‘reference change
value’ approach uses coefficients of variation cor-
responding to biological intra-individual variation
from other populations (obtained from different
publications); these different coefficients of varia-
tion may originate very different interpretations of
a change (Tables I and II).
On the other hand, non-SI traceable biological
quantities, as serum thyrotropin concentration, that
are measured with different immunochemical mea-
suring systems, lead to measured values that are not
comparable. Biological variation studies of these
quantities should not be pooled, as means or medi-
ans do not represent any biological reality [17].
It is clear in the measured values – illustrated
by the example of thyrotropin concentration – that
the percentage of increase is not always symmetric
to the percentage of decrease; hence the importance
of maintaining the sign of the difference: in our case,
more than 177.8% of increase in serum thyrotropin
concentration is needed to decide significance,
 172 A. Padró-Miquel et al.
Table II. Percentage of pairs of consecutive measured values that represent a significant change during two periods (years 2008–2009 and
year 2010) according to our proposal (% S (PP)), and according to the conventional ‘reference change value’ proposal [6] using two
intra-individual biological coefficients of variation for each biochemical quantity (% S (RCV1) and % S (RCV2)).
Biochemical quantity Period n % S (PP)
P – Albumin; mass c. 2008–2009 184 614
2010 53 501
P – Thyrotropin; arb.subst.c. 2008–2009 16 774
2010 2 763
Biochemical quantities are described according to the IUPAC and IFCC recommended syntax [18].
n, total number of pairs.
whereas less than a 61.4% of decrease is needed for
the same decision.
The advantage of our approach is that the
Scand J Clin Lab Invest Downloaded from informahealthcare.com by Nyu Medical Center on 05/13/15
interval that defines the decision whether a change
is significant or not is calculated with the measured
values contained in the laboratory reports of the
same population that attend the hospital and that
each laboratory could easily calculate its own.
The interval that defines the significance of a
change could easily be estimated with the patients’
measured values obtained every day in the labora-
tory, circumventing, in this way, two problems: the
reliability of the intra-individual biological coefficient
of variation, and the reliability of the metrological
For personal use only.
coefficient of variation used. Thus, our proposal is
more appropriate to each specific population than
the conventional ‘reference change value’ based on a
‘universal’ intra-individual biological coefficient of
variation.
Declaration of interest: The authors report no
conflict of interest. The authors alone are responsible
for the content and writing of the article.
References
[1] Harris EK. Distinguishing physiologic variation from
analytic variation. J Chron Dis 1970;23:469–80.
[2] Harris EK, Kanofsky P, Shakarji G, Cotlove E. Biological and
analytical components of variation in long-term studies of
serum constituents in normal subjects. II. Estimating biological
components of variation. Clin Chem 1970;16:1022–7.
[3] Harris EK. Statistical principles underlying analytical goal-
setting in clinical chemistry. Am J Clin Pathol 1979;72:374–82.
[4] Harris EK. Proposed goals for analytical precision and accu-
racy in single-point diagnostic testing. Arch Pathol Lab Med
1988;112:416–20.
[5] Harris EK, Brown SS. Temporal changes in the concentra-
tion of serum constituents in healthy men. Distributions of
within-person variances and their relevance to the interpre-
tation of differences between successive measurements. Ann
Clin Biochem 1979;16:169–76.
% S (RCV1) % S (RCV2)
12.7% 56.3% 9.7%
12.2% 64.8% 9.3%
18.4% 65.8% 38.1%
19.3% 71.2% 41.1%
[6] Harris EK,Yasaka T. On the calculation of ‘reference change’
for comparing two consecutive measurements. Clin Chem
1983;29:25–30.
[7] Fuentes-Arderiu X, González-de-la-Presa. Interchangeabil-
ity of estimates of day-to-day imprecision between commer-
cial control materials and serum pools. Clin Chem
2002;48:573–4.
[8] Fuentes-Arderiu X. Variability of the biological variation.
Scand J Clin Lab Invest 2002;62:561–4.
[9] International Federation of Clinical Chemistry, Interna-
tional Committee for Standardization in Hæematology.
Approved recommendation (1987) on the theory of refer-
ence values. Part 5. Statistical treatment of collected refer-
ence values. Determination of reference limits. Clin Chim
Acta 1987;170:S13–S32; also published in J Clin Chem
Clin Biochem 1987;25:645–56.
[10] Van Steirteghem AC, Robertson EA, Young DS. Variance
components of serum constituents in healthy individuals.
Clin Chem 1978;24:212–22.
[11] Gallagher SK, Johnson LK, Milne DB. Short- and long-
term variability of selected indices related to nutritional
status. II. Vitamins, Lipids, and proteins indices. Clin Chem
1992;38:1449–53.
[12] Browning MCK, Ford RP, Callaghan SJ, Fraser CG. Intra-
and interindividual biological variation of five analytes used
in assessing thyroid function: implications for necessary
standards of performance and the interpretation of results.
Clin Chem 1986;32:962–6.
[13] Ankrah-Tetteh T, Wijeratne S, Swaminathan R. Intraindi-
vidual variation in serum thyroid hormones, parathyroid
hormone and insulin-like grow factor-1. Ann Clin Biochem
2008;45:167–9.
[14] Nakagawa S, Cuthill IC. Effect size, confidence interval and
statistical significance: a practical guide for biologists. Biol
Rev 2007;82:591–605.
[15] Smellie WS. What is a significant difference between sequen-
tial laboratory results? J Clin Pathol 2008;61:419–25.
[16] Vasile VC, Saenger AK, Kroning JM, Jaffe AS. Biological
and analytical variability of a novel high-sensitivity cardiac
troponin T assay. Clin Chem 2010;56:1086–90.
[17] Fuentes-Arderiu X. Biological variation of non-SI traceable
biological quantities: example of proteins. Chem Lab Med
2006;44:1497.
[18] International Union of Pure and Applied Chemistry, Inter-
national Federation of Clinical Chemistry. Properties and
units in the clinical laboratory sciences-I. Syntax and seman-
tic rules (Recommendations 1995). Pure Appl Chem
1995;67:1563–74.