Nephrol Dial Transplant (2005) 20: 1951–1955
doi:10.1093/ndt/gfh926
Advance Access publication 5 July 2005
Original Article
Online measurement of haemoglobin concentration
Lindsay Chesterton, Stewart H. Lambie, Lisa J. Hulme, Maarten Taal,
Richard J. Fluck and Christopher W. McIntyre
Department of Renal Medicine, Derby City General Hospital, Derby, UK
Abstract
Background. Regular monitoring of haemoglobin in
chronic haemodialysis patients is essential to ensure
that targets for anaemia management are consistently
achieved. Repeated blood sampling can be time
consuming, invasive and, for pragmatic reasons, only
infrequently performed, often delaying therapeutic
change. On-line optical continuous monitoring of the
haemoglobin concentration would allow non-invasive
assessment of haemoglobin, and immediate therapeutic
changes could be implemented, thereby improving the
efficiency of anaemia management. This study aimed
to evaluate the use of on-line haemoglobin concentra-
tion measurement.
Methods. Eleven dialysis monitors (IntegraÕ Hospal)
were calibrated using at least five haemoglobin samples
spread over at least 4 g/dl. Optical measurement of
haemoglobin concentration is already incorporated
into the dialysis monitor to allow the study of relative
blood volume. Fifteen patients were studied with
paired haemoglobin measurements (i.e. dialysis moni-
tor value and conventional laboratory assessment)
taken at intervals over 7 months (mean 11.0±0.28 g/dl,
range 7.5–14.8).
Results. Haemoglobin measured by HemoscanÕ cor-
related well with the laboratory measurements
(r2 ¼ 0.83, P<0.0001), indicating that the machine
values are broadly comparable with laboratory figures.
There was a mean underestimate of haemoglobin
by HemoscanÕ of 0.34%. There was no significant
deterioration in the quality of this correlation over the
study period (r2>0.8).
Conclusion. The ability of the dialysis monitor to
measure the optical concentration of haemoglobin
compared with conventional laboratory assessment is
both precise and accurate. Regular on-line assessment
of haemoglobin may allow more proactive microman-
agement of renal anaemia, with a reduction in the time
Correspondence and offprint requests to: Dr S. H. Lambie,
Department of Renal Medicine, Derby City General Hospital,
Uttoxeter Road, Derby DE22 3NE, UK.
Email: slambie@doctors.org.uk
ß The Author [2005]. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved.
For Permissions, please email: journals.permissions@oupjournals.org
Downloaded from https://academic.oup.com/ndt/article/20/9/1951/1850196 by guest on 06 May 2022
taken to achieve clinically important targets and give
early warning of suboptimal response to treatment.
Keywords: haemoglobin concentration;
on-line monitoring
Introduction
The measurement of haemoglobin (Hb) concentration
is of great importance in the haemodialysis (HD)
population [1], with considerable resources of time,
effort and finance being expended to ensure that the
patients gain the greatest benefit from maintenance
of their Hb at normal or near normal levels [2,3]. Hb
measurements are not routinely measured at <4 weekly
intervals, as this is the NKF-K/DOQI recommended
frequency for checking Kt/V [4]. This inevitably
introduces a delay between any change in the patient’s
Hb concentration and institution of an appropriate
therapeutic response. The greatest individual compo-
nent of this delay is likely to be the time between the
change in Hb occurring and the next routine blood
round, but other components are also involved. These
include the time between taking the blood sample
and processing it, the time between the result being
available and the appropriate physician being aware
of it, and the time from then until the appropriate
therapeutic response is actioned. All of these factors
impair the efficient detection and correction of anaemia
in HD patients.
Potentially, more frequent venesection could help
to improve this situation. There would, however, be a
significantly increased cost in laboratory consumables,
and inconvenience to nursing staff and patients.
Furthermore, increased numbers of blood tests can
result in worsening anaemia if a sufficient volume
of blood is withdrawn [5]. Optical measurement of Hb
(Hbopt) would avoid these problems.
Although optical monitoring of relative change
in Hb concentration has been integrated into many
1952
dialysis monitors for some time, this has not generally
been used to monitor absolute Hb concentrations [6,7].
The HemoscanÕ (Hospal, Mirandola, Italy) system
allows calibration of Hbopt against local laboratory-
based measurement. If HemoscanÕ is to be used for
Hb monitoring in the clinical arena, it is important to
assess its accuracy and reliability. As Hb concentration
changes over the course of a dialysis session, it would
also be important to check that the initial measurement
of Hbopt correlates closely with the pre-dialysis Hb
measured in the laboratory on a sample taken at the
initiation of dialysis.
The accuracy and reliability of HemoscanÕ have
not been tested in routine clinical use. The aim of this
study was to make an initial assessment of the clinical
utility of on-line measurement of Hbopt.
Methods
Õ
Eleven Integra (Hospal, Mirandola, Italy) dialysis machines
were assessed. Relative blood volume (RBV) monitoring by
HemoscanÕ is fitted as standard to these machines. This
module allows continuous monitoring of RBV, the display
of signals in numerical and graphical mode, off-line data
retrieval and periodic on-board alignment to the laboratory
Hb values, thereby accounting for any relative drift between
the two devices. Alignment with the local laboratory involves
the simultaneous measurement of Hb by both HemoscanÕ
and standard laboratory assessment, and consists of three
parts. First, the HemoscanÕ module collects data during a
dialysis session, laboratory data are then inserted and finally
standard linear regression is performed. Five separate pairs
of samples are required to align the machines correctly
(not necessarily from the same patients). Sufficient spread
between the results is also required (at least 4 g/dl between
the maximum and minimum recorded values). All blood
samples were taken directly from the arterial line while a
useable HemoscanÕ trace was available. Hb was subsequently
measured by the hospital laboratory Sysmex XE 2100Õ
analyser (Sysmex, Kobe, Japan), which uses the sodium lauryl
sulfate-haemoglobin method, and has a coefficient of varia-
tion of <1% for all red blood cell (RBC) parameters [8].
The coefficient of variation is higher at lower Hb levels, rising
to 3.9% for samples with a mean Hb of 6.9±0.27 g/dl. Once
sufficient numbers and spread of samples are available,
the IntegraÕ monitor internally aligns Hbopt against the
laboratory samples with standard linear regression. After it
has been performed, the displayed values of Hbopt and blood
volume are computed by applying the regression coefficients.
HemoscanÕ has a range of measurement from 7 to 14 g/dl, an
accuracy of ±0.5 g/dl if aligned and a resolution of ±0.1 g/dl.
In the first phase of the study, further paired measure-
ments were made from a range of dialysis patients, using all
11 machines, at regular intervals over the course of 7 months,
in order to ascertain the accuracy and reliability of the
HemoscanÕ . Blood samples for laboratory analysis were
taken mid-dialysis while a useable HemoscanÕ trace was
available, and a simultaneous note of Hbopt was made.
Ten further paired observations were taken in the middle
of the study.
In the second phase of the study, 16 blood samples were
taken at dialysis initiation for laboratory analysis, as per
L. Chesterton et al.
routine clinical blood sampling guidelines. The laboratory
values obtained were then compared with Hbopt recorded
by DialmasterÕ (Hospal, Mirandola, Italy) at the start of
dialysis. DialmasterÕ is a central computer system with
the facility to record data from dialysis monitors linked to
the system. Patients are identified by a card system, allowing
for subsequent off-line analysis and interpretation of a large
number of monitored variables. In this instance, a record
of Hbopt was studied, and the initial value was taken from
the first Hbopt recorded after dialysis initiation.
A total of 46 data points were collected (11 at initial
alignment of the dialysis monitors, 10 from the middle of
the study, nine at completion at the study and 16 from a
comparison of recorded data with pre-dialysis samples).
Autocalibration sufficient to analyse RBV is performed
prior to every dialysis session. There are no manufacturer’s
Downloaded from https://academic.oup.com/ndt/article/20/9/1951/1850196 by guest on 06 May 2022
guidelines as to the frequency with which further laboratory
alignment should be performed.
Patients
Fifteen patients were recruited from our chronic HD
population, mean age 65 years. All patients were undergoing
standard HD. Patients were dialysed using bicarbonate
buffering, a dialysate sodium concentration of 140 mmol/l
and a dialysate flow rate of 500 ml/min. HD used either
low-flux haemophan polycarbonate membranes (Hospal
HG 500–700) or mid-flux cellulose diacetate membranes
(Hospal Diacepal 20); no dialysers were reused, and a linearly
decreasing ultrafiltration profile was used throughout. Those
known to have haemoglobinaemia or myoglobinaemia or
those prescribed rifampicin were excluded. These substances
absorb light at the same wavelength as Hb and can interfere
with the accurate measurement of Hbopt.
Appropriate ethical approval was obtained from the
local research ethics committee.
Statistical analysis
All data were analysed using GraphPad Prism version
3.00 for Windows (GraphPad Software, San Diego CA,
www.graphpad.com). Data are expressed as mean±SD
unless otherwise stated. Bland–Altman plots were created
by plotting the difference between Hbopt and laboratory-
measured Hb, expressed as a percentage of the mean
of the two measurements, against the mean of the two
measurements.
Results
Hb measured by HemoscanÕ shows a strong correla-
tion with that measured by SysmexÕ analyser
(r2 ¼ 0.89, Figure 1) and is accurate as demonstrated
in the Bland–Altman plot (Figure 2). Mean over-
estimate of Hb by HemoscanÕ was 1.3±2.4%.
Seven months later, precision and accuracy appear
undiminished (Figures 3 and 4) (r2 ¼ 0.97, P<0.0001).
At this stage, there is a mean underestimate by
HemoscanÕ of 1.0±1.5%. Two machines had had
adjustment of their software or other problems, and
were no longer aligned. These monitors were removed
from further analysis.
Online measurement of haemoglobin concentration 1953
15.0 15.0

Hb by Sysmex analyser (g/dl)

Hb by Sysmex analyser (g/dl)

12.5
12.5

10.0
10.0
n=9

n=11 7.5
7.5 10.0 12.5 15.0
7.5
Hb by Hemoscan (g/dl)

Downloaded from https://academic.oup.com/ndt/article/20/9/1951/1850196 by guest on 06 May 2022

7.5 10.0 12.5
Hb by Hemoscan (g/dl)
Fig. 1. Haemoglobin measured by SysmexÕ analyser vs HemoscanÕ
immediately after alignment, showing a high degree of correlation
between the two measures (r2 ¼ 0.90, P<0.0001).
15.0
Fig. 3. Linear regression plot of haemoglobin measured by the
SysmexÕ analyser vs HemoscanÕ 7 months after alignment,
showing a significant correlation between the two measures
(r2 ¼ 0.97, P<0.0001).

20 20

Difference between Hemoscan

Difference between Hemoscan

percentage of the mean (%)

percentage of the mean (%)

and Sysmex analyser, as a

and Sysmex analyser, as a

15 15
10 10
5 5
0 0

−5 −5

−10 −10

−15 n=11 −15 n=9

−20 −20
8 9 10 11 12 13 14 15 8 9 10 11 12 13 14 15 16
Mean of Hemoscan and Sysmex
Mean of Hemoscan and Sysmex analyser (g/dl)
analyser (g/dl)
Fig. 4. Bland–Altman plot demonstrating the difference between
Fig. 2. Bland–Altman plot demonstrating the difference values detected by the HemoscanÕ and SysmexÕ analyser
between values detected by the HemoscanÕ and SysmexÕ analyser expressed as a percentage of the mean against the mean of Hb
expressed as a percentage of the mean against the mean of Hb measured by the HemoscanÕ and SysmexÕ analyser 7 months
measured by the HemoscanÕ and SysmexÕ analyser after post-alignment. There is a mean underestimate by HemoscanÕ of
alignment. There is a mean overestimate by HemoscanÕ of 1.0±1.5%.
1.3±2.4%.

HemoscanÕ was also accurate in the second phase

of the study. Laboratory Hb measured in the standard
pre-dialysis manner was comparable with the value
for Hbopt recorded by DialmasterÕ at the start of
dialysis (Figure 5). In this case, HemoscanÕ under-
estimated Hb by a mean of 0.7±4.1% compared with
the SysmexÕ analyser, though precision was slightly
reduced (r2 ¼ 0.75, compared with r2 ¼ 0.90 for all
paired samples).
Finally, in an overall analysis (n ¼ 46), HemoscanÕ
shows good precision and accuracy (Figure 6).
Mean overestimate by HemoscanÕ was 0.1±3.3%.
This is despite combining results from alignment,
the middle and the end of the study, or from the
comparison of initial Dialmaster records of Hbopt,
for blood samples and HemoscanÕ results taken
simultaneously.
Discussion
Initial results immediately post-alignment comparing
Hbopt and Hb measured by the SysmexÕ XE 2100
analyser show good precision and accuracy. This
precision and accuracy are maintained in at least
the medium term (over the 7 month period of this
study). Overall results suggest that precision is
increased for higher values. Some of the scatter seen
is due to inherent variability within the Sysmex XE
2100Õ analyser used as a comparator. While there
appears to be less agreement between the two methods
at lower Hb levels, the documented precision of the
SysmexÕ analyser is also reduced for these concentra-
tions [8]. Therefore, the scatter seen at lower Hb
concentrations may be due to the SysmexÕ analyser
1954 L. Chesterton et al.
Difference between Hemoscan
20 Our data suggest that while some precision is lost,
percentage of the mean (%)
and Sysmex Analyser, as a
15 this is still a clinically useful measure of Hb. Therefore,
10 off-line results will be available at the physician’s
5 convenience.
0 Measurement of Hbopt may also lead to further
−5 refinement of anaemia management. Measurement of
−10
Hbopt at every dialysis session would lead to early
n=16 detection of anaemia, and therefore should reduce the
−15
response time prior to appropriate clinical therapeutic
−20 action. This in turn should lead to an overall increase
8 9 10 11 12 13 14 15
in the total time that a patient spends with a satisfactory
Mean of Hemoscan and Sysmex
analyser (g/dl) Hb. Secondly, when considering the response to
changes in Hb and erythropoietin doses, currently
Fig. 5. Bland–Altman plot demonstrating the difference between available algorithms [9] depend on considering both
values detected by the HemoscanÕ and SysmexÕ analyser expressed
the absolute Hb concentration and the rate of change

Downloaded from https://academic.oup.com/ndt/article/20/9/1951/1850196 by guest on 06 May 2022

as a percentage of the mean against the mean of Hb measured by
the HemoscanÕ and SysmexÕ analyser using data recorded by
DialmasterÕ . There is a mean underestimate by HemoscanÕ of
0.7±4.1%.
Difference between Hemoscan
of Hb concentration. Similar algorithms could be
constructed responding to smaller rates of change in
Hbopt, potentially delivering a much more individually
tailored and more efficient erythropoietin regimen.
The full value of anaemia management using Hbopt

20 compared with monthly laboratory-based measure-

percentage of the mean (%)
and Sysmex analyser, as a

15 ment would require evaluation by further study.

10 It is important to note that optical measurement
5 of Hb does have some limitations. Its use is restricted
0 in situations where substances that absorb light at
−5
the same wavelength as Hb may be present. These
include treatment with rifampicin, myoglobinaemia
−10
n=46
and haemoglobinaemia. Our data would indicate that
−15 repeated alignment of the HemoscanÕ module with
−20 the local laboratory should perhaps be performed every
7 8 9 10 11 12 13 14 15 16 3 months to ensure that drift does not occur.
Mean of Hemoscan and Sysmex Overall, this study suggests that HemoscanÕ is a
analyser (g/dl) reliable system for producing an accurate measure
Fig. 6. Bland–Altman plot demonstrating the difference between not only of RBV, but also of absolute Hb concen-
values detected by the HemoscanÕ and SysmexÕ analyser expressed tration. This has the potential to be a useful tool
as a percentage of the mean against the mean of Hb measured by
the HemoscanÕ and SysmexÕ analyser for all results taken during
with which to adjust erythropoietin dosages, enabling a
the study (alignment, middle, end and post hoc recorded results). more rapid response to any change in Hb, and hence
Overall, there is a mean overestimate by HemoscanÕ of 0.1±3.3%. more efficient use of a scarce resource.

rather than the HemoscanÕ module alone. In either
case, the magnitude of the fluctuations is sufficiently
small that they would have little impact on clinical
decision making.
Results from the comparison of off-line analysis
of the HemoscanÕ results with pre-dialysis blood
samples demonstrated slightly reduced precision and
accuracy compared with the initial alignment samples
(r2 ¼ 0.75, as compared with r2 ¼ 0.90). This is still
sufficiently reliable for the vast majority of clinical
situations, and is further supported by the ability
of HemoscanÕ to measure Hbopt at every dialysis
session. Any inaccuracy would be automatically
re-checked after 2 or 3 days, and there will be less
urgency to respond to single results if trends in Hb
can be observed. Recorded HemoscanÕ results by
Dialmaster are suitably comparable with pre-dialysis
Hb results measured in the conventional manner.
Conflict of interest statement. C.W.M. has received an unrestricted
educational grant from Gambro Hospal. No other authors have
any conflict of interest to declare.
References
1. Locatelli F, Conte F, Marcelli D. The impact of haematocrit
levels and erythropoietin treatment on overall and cardiovas-
cular mortality and morbidity—the experience of the
Lombardy Dialysis Registry. Nephrol Dial Transplant 1998; 13:
1642–1644
2. Burton C, Ansell D, Taylor H, Dunn E, Feest T. Management
of anaemia in United Kingdom renal units: a report from the
UK Renal Registry. Nephrol Dial Transplant 2000; 15:
1022–1028
3. Working Party for European Best Practice Guidelines for the
Management of Anaemia in Patients with Chronic Renal
Failure. European best practice guidelines for the management
of anaemia in patients with chronic renal failure. Nephrol Dial
Transplant 1999; 14 [Suppl 5]: 1–50
Online measurement of haemoglobin concentration
4. I. NKF-K/DOQI Clinical Practice Guidelines for Hemodialysis
Adequacy: update 2000. Am J Kidney Dis 2001; 37 [Suppl 1]:
S7–S64
5. Napolitano LM. Scope of the problem: epidemiology of anemia
and use of blood transfusions in critical care. Crit Care 2004; 8
[Suppl 2]: S1–S8
6. Tonelli M, Astephen P, Andreou P et al. Blood volume
monitoring in intermittent hemodialysis for acute renal failure.
Kidney Int 2002; 62: 1075–1080
1955
7. Moret K, Hassell D, Kooman JP et al. Ionic mass balance and
blood volume preservation during a high, standard, and
individualized dialysate sodium concentration. Nephrol Dial
Transplant 2002; 17: 1463–1469
8. Walters J, Garrity P. Performance evaluation of the Sysmex
XE-2100Õ hematology analyzer. Lab Hematol 2000; 6: 83–92
9. Richardson D, Bartlett C, Will EJ. Optimizing erythropoietin
therapy in hemodialysis patients. Am J Kidney Dis 2001; 38:
109–117
Received for publication: 14.12.04
Accepted in revised form: 4.5.05
Downloaded from https://academic.oup.com/ndt/article/20/9/1951/1850196 by guest on 06 May 2022