Professional Documents
Culture Documents
Biological Variability HbA1C
Biological Variability HbA1C
a r t i c l e i n f o a b s t r a c t
Article history: Background: The measurement of glycated hemoglobin (HbA1c) has a pivotal role in monitoring glycemic
Received 28 June 2010 state in diabetic patients. Furthermore, the American Diabetes Association has recently recommended the
Received in revised form 23 July 2010 use of HbA1c for diabetes diagnosis, but a clear definition of the clinically allowable measurement error is still
Accepted 23 July 2010 lacking. Information on biological variability of the analyte can be used to achieve this goal.
Available online 2 August 2010
Methods: We systematically reviewed the published studies on the biological variation of HbA1c to check
consistency of available data in order to accurately define analytical goals.
Keywords:
Results: The nine recruited studies were limited by choice of analytic methodology, population selection,
Biological variability
Glycated hemoglobin
protocol application and statistical analyses.
Diabetes mellitus Conclusions: There is an urgent need to determine biological variability of HbA1c using a specific and
traceable assay, appropriate protocol and appropriate statistical evaluation of data.
© 2010 Elsevier B.V. All rights reserved.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1606
2. Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1607
3. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1607
3.1. Selected studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1607
3.2. Definition of the measurand and influence of method specificity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1607
3.3. Type of studied subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1607
3.4. Study duration and sampling frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1608
3.5. Sample analysis and assessment of analytical variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1608
3.6. Statistical analysis of data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1609
4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1609
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1610
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1610
1. Introduction Considering the average life span of red cells, the HbA1c value should
mimic the mean glycemic value of the previous 2–3 months. The
Glycated hemoglobin or βN-1-deoxyfructosyl-hemoglobin American Diabetes Association (ADA) recommends HbA1c determi-
(HbA1c)1 is the product of a nonenzymatic reaction of glycation, nation in patients with diabetes mellitus on therapy in order to
namely a condensation between the aldeidic group of glucose and the monitor the glyco-metabolic status in the medium–long term and
amino group of the terminal valine in the β-chain of hemoglobin A0. thus reduce the risk of vascular complications [1]. Studies performed
The amount of HbA1c is strictly related to blood glucose concentration. on non-diabetic subjects showed that even slight elevations of HbA1c
concentration in blood correlated with an increased cardiovascular
risk [2,3]. The “Standards of medical care in diabetes” recently
⁎ Corresponding author. Laboratorio Analisi Chimico-Cliniche, Azienda Ospedaliera published by ADA, recommends the use of HbA1c as a diagnostic
‘Luigi Sacco’, Via GB Grassi 74, 20157 Milano, Italy. Fax: + 39 02 503 19835. criteria for diabetes, a finding which may further increase the clinical
E-mail address: federica.braga@studenti.unimi.it (F. Braga).
1
relevance of HbA1c testing [1].
Nonstandard abbreviations: HbA1c, glycated hemoglobin; ADA, American Diabetes
Association; NGSP, National Glycohemoglobin Standardization Programme; SI,
Many analytical methods are available for HbA1c determination
‘Sistème Internationale de Measure’; CVI, intra-individual biological variation; CVG, [4]. Techniques based on the difference in isoelectric point of HbA1c
inter-individual biological variation. and other hemoglobins, and immunochemical methods specifically
0009-8981/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.cca.2010.07.030
F. Braga et al. / Clinica Chimica Acta 411 (2010) 1606–1610 1607
3. Results and discussion The analytical (σA2), intra-individual (σ2I), and inter-individual (σG2) variations
are calculated by analysis of variance:
—The analytical variation is estimated from the duplicate results for each specimen,
3.1. Selected studies —The intra-individual variation is estimated from the serial results for each subject,
—The inter-individual variation is estimated from the total variance (σ2total) of data
Table 3 shows the main characteristics of selected studies on HbA1c minus the analytical and intra-individual components.
biological variability [16–24]. While all nine studies estimated HbA1c All the components of variation are then transformed to the relevant CVs using the
CVI, only five did HbA1c CVG. corresponding means.
1608 F. Braga et al. / Clinica Chimica Acta 411 (2010) 1606–1610
Table 3 tions and the study duration. Such duration should be neither too
Studies on biological variability of HbA1c. short (a few days) nor too long (years), so that the variability study
Study Author, No. and No. of Analytical CVI, CVG, may not be related to periods different from those used in clinical
no. year type of samples method % % practice for the measurements of specific analyte nor influenced by
subjects (frequency) additional causes (e.g. seasonal variation) [21,30]. It seems, therefore,
1 Godsland, 10 healthy 11–27 Ion-exchange 1.8 – reasonable to adjust the duration of studies evaluating biological
1985 [16] volunteers (7 days) chromatography variability of HbA1c between 1 and 6 months, which corresponds to
(6 M, 4 F)
the advised frequency of the assay in different clinical situations [1]. If
2 Howey et 8 diabetics 6 (3–4 days) Electroendosmosis 7.3 10.8
al., 1989 the majority of studies apply this criterion, the choice by the other
[17] authors to extend the duration beyond 6 months appears question-
3 Phillipou 73 4 (1 month) Affinity 4.2a; – able [18,19,22,24]. It is indeed likely that an evaluation lasting for
et al., diabeticsa,b chromatography 5.1b more than 6 months would include variability sources beside the
1993 [18] 4 7.1a; –
biological one. For instance, Kolatkar et al. [19] attributed to an
(3 months) 9.8b
4 Kolatkar 29 NS HPLC (ion- 2.4 – intensive treatment regimen the decrease in variability of HbA1c
et al., diabetics (3 months) exchange) concentrations in diabetic patients found along the wide study period
1994 [19] NS 4.1 – (6 years), wrongly interpreting this result as a reduction of the HbA1c
(12 months)
CVI. The variation observed in these subjects may undoubtedly be
5 Kilpatrick 12 healthy 10 (15 days) HPLC (ion- 1.9 6.8
et al., volunteers exchange) utilized for clinical decision making, but certainly not to estimate the
1998 [20] (7 M, 5 F) biological variability and the reference change value of HbA1c, which
6 Garde et 11 healthy 5 (7 days) HPLC (ion- b 0.7 3.3 must be assessed on a healthy population for a suitable time period to
al., 2000 volunteers exchange) be adopted as the basic criterion for interpretation of HbA1c results
[21] (F only)
and evaluation of glycemic control in diabetic patients over time.
7 Trapé et 47 4–7 Immunoturbidimetry 7.9c; –
al., 2000 diabeticsc,d,e (6 months) 5.4d;
[22] 3.9e 3.5. Sample analysis and assessment of analytical variability
8 Rohlfing 45 healthy 12 (7 days) HPLC (affinity) 1.7f 4.0
et al., volunteers
Apart from the method employed, in the retrieved studies two
2002 (M only)
[23]
further limitations can be found when compared to the ideal protocol
9 Desmeules 38 5 (1 year) HPLC (ion- 4.8 12.8 mentioned in Table 1. First, HbA1c measurements were often not
et al., pediatric exchange) performed in duplicate. Duplicate measurement permits a direct
2010 [24] non- evaluation of the within-run analytical variability, and that is better
diabetics,
than deriving it from the results of the laboratory internal quality
but with
cystic control programme, sometimes obtained on non-commutable mate-
fibrosis rials displaying a different behaviour from patient samples. Second,
(24 M, 14 F) HbA1c measurements along the study were often performed on fresh
CVI, intra-individual variability; CVG, inter-individual variability; M, males; F, females; samples the same day of blood collection, spreading them on several
NS, not specified. analytical runs, instead of performing measurements for all appro-
a
39 on stable metabolic control (16 on short-term, 23 on long-term), b34 on variable priately stored samples only in a single batch and, consequently,
metabolic control (13 on short-term, 21 on long-term); c13 on stable glycemic control,
d exclude from the total variability calculation the effect of between-run
10 on acceptable diabetic control, e24 on poor glycemic control; fanalytical variability
included. analytical variation, so difficult to be correctly assessed.
To perform all the determinations in duplicate on the same analytical
concentrations in a biological fluid around its homeostatic set point [15]. run represents the ideal experimental model for a biological variability
The presence of disease, mainly if unstable and not well controlled, may study, limiting to within-run variability the influence of the analytical
markedly modify the obtained information by amplifying such fluctua- variability on total variability of results, with less chance of error in the
tion. Fig. 1 clearly shows the difference between CVI values obtained in subsequent estimate of biological variability, obtained by subtraction
healthy volunteers (from b0.7% to 1.9%) and in diabetic patients (from 2.4 (Table 2). Unfortunately, the only study which correctly included sample
to 9.8%) with no overlap between the two groups. The same is also freezing in order to perform all HbA1c determinations on the same batch
evident for CVG in Fig. 2. Some authors [29] have argued that the at the end of blood collection phase, did not assay samples in duplicate
biological variability of HbA1c could be more efficiently evaluated in and got the information on analytical variability from the results obtained
diabetic patients, because the reference change value in clinical practice is on control materials [21]. This did not allow an accurate definition of CVI,
used to monitor HbA1c concentrations in such patients rather than in as total variability of experimental results was less than the analytical
non-diabetic subjects. This approach, however, does not take into account variability alone obtained from quality control data, that was obviously
that the disease itself may act as an extra-component of the variability overestimated [21].
around the HbA1c set point in the individual. This pathophysiologic In some studies, blood samples were not frozen and were,
change should actually be highlighted by the reference change value, not therefore, assayed in different runs, estimating analytical variability
used to derive it. For similar reasons, it is also questionable to carry out a from results obtained on control materials [18,20,24]. Other studies
biological variability study on a pediatric cohort affected by cystic fibrosis obtained the analytical variability from literature data [19]. The
instead of on ostensibly healthy children [24]. extreme case of such approximation can be found in the study by
Four studies correctly performed the evaluation on healthy Rohlfing et al. [23], in which, in addition to the lack of mentioning
volunteers, but two of those enrolled only subjects of one gender, sample storage conditions, singlicate measurements of samples made
females [21] or males [23], possibly because the samples did not it necessary to subtract to total experimental variability of results the
belong to specifically enrolled population. “within-day” analytical variability, obtained from quality control data,
so overestimated that only an approximate and possibly under-
3.4. Study duration and sampling frequency estimated CVI was reported as “probably b1%”.
Homogeneity of collection and in the number of samples drawn
It is well known that a biological variability estimate is strictly from each individual is another critical procedural point; the availability
dependent on the time interval between consecutive sample collec- of too many samples from the same subject may imply the technical
F. Braga et al. / Clinica Chimica Acta 411 (2010) 1606–1610 1609
Fig. 1. Comparison between intra-individual biological variability (expressed as CVI) obtained in studies enrolling apparently healthy individuals and in studies evaluating diabetic
subjects. Study numbers as in Table 3. Dashed line shows CVI value as reported by Ricòs et al. (www.westgard.com/biodatabase1.htm).
impossibility to perform all measurements in the same run. A range reported in the present paper, it should, however, be noted that the
from a minimum of 11 to a maximum of 27 weekly blood samples information currently available on HbA1c biological variability is
obtained in subjects during the same study means to estimate the generally taken as definitive in most scientific and professional circles.
biological variability in a period from b3 to N6 months, making The most popular informations on biological variability of biochemical
incorrect to approximate individual data to a mean CVI value [16]. and hematological analytes, employed in daily practice worldwide,
are those compiled by Ricòs et al., freely available at www.westgard.
3.6. Statistical analysis of data com/biodatabase1.htm. Values of biological variability components
for HbA1c listed in this database (CVI 3.4%, CVG 5.1%) are apparently
The statistical method used to assess variability components was obtained from the (weighed?) mean of results available in the
not specified (totally or partially) in three papers [16,19,22]; the literature, taken as valid by the authors of the compilation. It is,
remaining studies correctly employed analysis of variance (ANOVA). however, impossible to understand the criterion (if one) of study
CV G estimate was added to CV I estimate in five studies evaluation and selection of data, so that some doubts appear justified.
[17,20,21,23,24], three of them carried out in groups of apparently The observation reported in the present study and, particularly, the
healthy subjects (CVG values from 3.3% to 6.8%) [20,21,23]. finding of a substantial difference of variability results obtained in
healthy individuals and diabetic patients (Figs. 1 and 2) show the
4. Conclusions fallacy of determining a single value of CVI and CVG simply using the
mean of available results. Despite all reported limitations, keeping
A clear message from this systematic review of available literature into account the available data obtained in healthy subjects [16,20],
is the obvious lack of robust data on biological variability of HbA1c the CVI of HbA1c should approximately be 1.8–1.9%, much lower than
concentrations in blood (Table 4). Despite all limitations and remarks that reported by Ricòs et al. in the mentioned database. It is, however,
Fig. 2. Comparison between inter-individual biological variability (expressed as CVG) obtained in studies enrolling apparently healthy individuals and in studies evaluating diseased
subjects. Study numbers as in Table 3. Dashed line shows CVG value as reported by Ricòs et al. (www.westgard.com/biodatabase1.htm).
1610 F. Braga et al. / Clinica Chimica Acta 411 (2010) 1606–1610