Upload
r-thomas
View
214
Download
0
Embed Size (px)
Citation preview
FORUM
The in vitro effects of crystalloids and colloids on
coagulation
K. Ekseth,1 L. Abildgaard,2 M. Vegfors,3 J. Berg-Johnsen1 and O. Engdahl2
1 Consultant Neurosurgeon, Department of Neurosurgery, The National Hospital, 0027 Oslo, Norway,
2 Consultant Anaesthesiologist, and 3 Head of Neuroanaesthesia, Department of Anaesthesiology, University Hospital,
58085 Linkoping, Sweden
Summary
Classically haemodilution is regarded as causing coagulopathy. However, haemodilution with
saline seems to cause a hypercoagulable state both in vivo and in vitro. The aim of the present study
was to measure the effect of mild to severe haemodilution using thrombelastography. Blood
samples were taken in 12 healthy volunteers and divided into seven aliquots. One aliquot was
undiluted and acted as control. The other six were diluted with normal saline, Ringer Acetate,
4% albumin, Dextran 70, 6% and 10% hydroxyethylstarch to 10%, 20%, 40%, 50% and 60%
dilution. The dilution was checked by measuring the haemoglobin concentration. Each aliquot was
placed in a temperature-controlled thrombelastography channel. Increased coagulation activity, as
measured by thrombelastography changes, was detected at low and medium levels of dilution with
all the tested solutions. At more than 40% dilution, coagulation returned to normal while in the
case of dextran and hydroxyethylstarch coagulopathy developed. For crystalloids and albumin,
dilution had to exceed 50% before coagulation was impaired. If these findings can be reproduced
in vivo, they may have implications for transfusion practice and prophylaxis against thrombosis.
Keywords Coagulation: haemodilution; thrombelastography.
........................................................................................................
Correspondence to: Dr K. Ekseth
E-mail: [email protected]
Accepted: 14 May 2002
Haemodilution is common following blood loss. Tradi-
tionally the focus during haemodilution has been on the
oxygen carrying capacity of blood [1], and haemodilution
has generally been regarded as a cause of hypocoagulation.
However, in 1950 Tocantins et al. reported that moderate
blood dilution with normal saline caused accelerated
coagulation. Numerous reports describing alterations of
coagulation during haemodilution followed [2–6]. Dur-
ing normovolaemic haemodilution in patients under
anaesthesia, abnormal haemostasis develops before
impaired global tissue oxygenation [7].
Thrombelastography was developed as a research tool
[8] but has been developed into a clinically useful
coagulation monitor [9]. Thrombelastography can be
used at the bedside and can detect impaired coagulation as
well as hypercoagulable states. By using thrombelasto-
graphy, Tuman et al. found increased coagulability with
progressive blood loss, even when losses were replaced by
crystalloids and packed red cells [3].
Therefore the aim of the present investigation was to
use thrombelastography in vitro to find the level of
dilution required to cause hypercoagulation with com-
monly used crystalloids and colloids. We also wanted to
determine whether the degree of dilution influenced
coagulation.
Methods
The thrombelastograph (Haemoscope Corp., 5693 West
Howard Street ⁄Niles, IL 60714, USA) consists of a
heated (37 �C) cuvette containing 0.36 ml of whole
blood. A pin suspended by a torsion wire is lowered into
the blood. The cuvette oscillates on its vertical axis
through 4.0–4.5 degrees. While the blood remains liquid,
Anaesthesia, 2002, 57, pages 1102–1133.....................................................................................................................................................................................................................
1102 � 2002 Blackwell Publishing Ltd
the motion of the cup does not affect the pin. However,
when the clot starts to form, the motion of the cup is
transmitted to the pin and then amplified to give the
thrombelastography trace. The technique evaluates both
clot initiation and the structural characteristics of the clot
formed together with its stability [9].
The reaction time (r: normal range 40–55) measures
the time from when the blood is placed in the cuvette
until an amplitude of 2 mm is reached on the thromb-
elastography trace (Fig. 1). The reaction time represents
the time necessary for initial fibrin formation. The
coagulation time (k: normal value 19–24 mm) is the
time between 2 and 20 mm divergence on the
thrombelastography tracing. The k-value is a measure
of the rapidity of fibrin build-up and fibrin cross-linking.
The maximum amplitude (normal value 46–60 mm) is
the width of the thrombelastography trace at its maxi-
mum divergence and reflects the absolute strength of the
fibrin clot and depends on platelet number and function
as well as fibrinogen levels. The a-angle (normal value
18–25 degrees) is the slope of the outside divergence of
the tracing from the r to k-values. It denotes the speed at
which clot is formed and cross-linked [3].
Twelve healthy volunteers who were not taking
salicylates or non-steroid anti-inflammatory drugs were
enrolled. Approval was obtained from the local ethics
committee, and all the participants gave informed con-
sent. Blood samples (20 ml) were taken from a large
forearm vein and divided into seven 2 ml portions. One
portion was undiluted and acted as control. The other six
2 ml portions were diluted with normal saline, Ringer
Acetate, 4% albumin, Dextran 70, 6% or 10% hydroxy-
ethylstarch (Fresenius Kabi Norge AS, Svinesundv 80,
1750 Halden, Norway). New samples from successive
vein punctures were diluted to 10%, 20%, 40%, 50% and
60% into each of the different crystalloid or colloid
solutions. After obtaining the appropriate dilution,
0.36 ml of solution was placed in parallel-connected
thrombelastography channels. Before analysing samples
from a new individual all channels were calibrated and
the first blood sample in each individual was used in all
channels to check calibration. After dilution all samples
were inverted in the same manner to ensure an equal
activation of the clotting system. Each control and diluted
sample was checked for accuracy of dilution by measuring
the haemoglobin concentration.
The r and k-values, a-angle and maximal amplitude
were obtained from the thromobelastograph. Each dilu-
ted sample was compared to the corresponding control
sample. Analysis of differences between the test solutions
was performed using one- and two-way analysis of
variance (ANOVA) with multiple measurement correction.
The means and standard deviations (SDs) are presented as
summary statistics.
Results
Reaction time
The r-values for the different solutions and different
dilutions are presented in Table 1 and Fig. 2. The analysis
of variance p-value for the solutions (F ¼ 24.0), dilution
(F ¼ 55.3) and their interaction (F ¼ 3.9) was < 0.0001.
Ringer Acetate, normal saline and 4% albumin had a
reduced r-value until 50% dilution after which further
dilution caused an increased r-value for normal saline and
4% albumin, while the reaction time remained reduced
for Ringer Acetate at 60% dilution.
Six percent and 10% hydroxyethylstarch initially also
produced a slightly reduced r-value, but from 40%
dilution the r-value was increased. At 60% dilution 6%
and 10% hydroxyethylstarch had much increased r-values;
however, Dextran 70 seems to have an intermediate
effect, turning from hypercoagulation to hypocoagulation
at 50% dilution.
K-value
The k-values for the different solutions and different
dilutions are presented in Table 2 and Fig. 3.
Figure 1 The thromboelastographtracing is composed of four principalparameters, reaction time (r), a – angle,k-value and maximum amplitude.
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1103
The analysis of variance p-value for the solutions
(F ¼ 31.095), dilution (F ¼ 43.811) and their interaction
(F ¼ 4.375) was < 0.0001. Both crystalloids and albumin
caused hypercoagulation at low dilution. The synthetic
colloids initially showed only minor changes but changed
into hypocoagulation at 40% dilution. The k-value record-
ing of 6% hydroxyethylstarch, however, needed 50%
dilution until hypocoagulation was seen. For crystalloids
and albumin, dilution had to be carried until 60% before
any sign of hypocoagulation was seen. For Ringer Acetate
at 60% dilution, the k-value was still reduced.
a-angle
The a-angle values for the different solutions and
different dilutions are presented in Table 3 and Fig. 4.
The analysis of variance p-value for the solutions
(F ¼ 35.863), dilution (F ¼ 37.02) was < 0,0001. The
p-value for their interaction (F ¼ 1.82) was 0.0178.
All diluents initially increase the a-angle as a sign of
hypercoagulation. For the synthetic colloids this is
returned to baseline at 40% dilution and present hypo-
coagulative values with increasing dilution. Crystalloids
and albumin on the other hand show 50% increase in
the a-angle value between 20% and 40% dilution and
maintain hypercoagulation until 60% dilution.
Maximal amplitude
The maximal amplitude values for the different solutions
and different dilutions are presented in Table 4 and
Fig. 5. The analysis of variance p-value for the solutions
(F ¼ 63.165), dilution (F ¼ 75.125) and their interaction
(F ¼ 2.197) was < 0.0001. Maximal amplitude was the
thrombelastography parameter that was least affected by
dilution. The crystalloids and albumin showed signs of
increased clotting with increased maximal amplitude
values from 10% to 40% dilution, but from 50% dilution
a slightly impaired coagulation was found. The maximal
amplitude values for 10% hydroxyethylstarch and dextran
were unaffected by 10% and 20% dilution, but from 40%
dilution hypocoagulation was present. Ringer Acetate
and albumin showed signs of increased clotting until 50%
dilution. The maximal amplitude value was most affected
Table 1 The r-values (means and (standard deviations)) for the various test solutions and dilutions. The superscripted numerals showwhich solutions differed significantly between each other. Comparisons in this table are not significant unless the correspondingp-value is less than 0.0033.
Dilution (1) NaCl 4,5,6 (2) Ringer4,5,6 (3) Albumin4,5 (4) 6% HES1,2,3 (5) 10% HES1,2,3,6 (6) Dextran1,2,5
10% 39.5 (6.9) 40.8 (7.1) 42.3 (9.8) 42.1 (7.6) 45.1 (6.6) 46.5 (10.9)20% 36.6 (6.5) 37.8 (10.4) 36.4 (7.5) 40.6 (12.5) 44.8 (9.8) 42.8 (8.0)40% 36.3 (12.0) 36.0 (9.8) 37.0 (11.5) 48.7 (18.7) 47.5 (15.1) 40.5 (10.5)50% 41.9 (12.0) 35.4 (4.1) 42.0 (8.4) 62.4 (25.9) 59.6 (16.9) 47.1 (16.4)60% 52.4 (15.7) 40.0 (6.6) 64.4 (20.6) 74.6 (19.8) 75.4 (32.0) 70.4 (18.5)
Figure 2 Mean r-values of the solutions,and their reaction to increasing haemo-dilution; shortening of r suggests hyper-coagulation. Albumin, shaded diamond;10% hydroxyethylstartch, open square;Dextran 70, solid triangle; Normalsaline, open circle; Ringer Acetate, star;6% hydroxyethylstartch, solid circle.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1104 � 2002 Blackwell Publishing Ltd
by normal saline showing hypercoagulative effect until
60% dilution.
Discussion
Dilution of blood reduces its viscosity, but the assumption
that this also causes reduced coagulative ability is not
necessarily true. Haemodilution causes a dilution of
coagulation factors and has generally been thought to
attenuate clot formation. However, Tocantins reported
that �regardless of the activating agent used, the rate of
coagulation of haemophilic plasma can be made equal to
that of normal plasma by appropriate dilution.� [2].
Furthermore, he showed that the nature of the crystalloid
was not crucial, since all crystalloids had a similar effect.
Recently these findings have been re-evaluated and
several studies have reported hypercoagulation following
haemodilution. In 1979 Tsunehiro performed an in vitro
study using thrombelastography with colloid dilution of
blood and found an initial hypercoagulation changing
Table 2 The k-values (means and (standard deviations)) for the various test solutions and dilutions. The superscripted numerals showwhich solutions differed significantly between each other. Comparisons in this table are not significant unless the correspondingp-value is less than 0.0033.
Dilution (1) NaCl 4,5,6 (2) Ringer4,5,6 (3) Albumin4,5 (4) 6% HES1,2,3 (5) 10% HES1,2,3,6 (6) Dextran1,2,5
10% 16.5 (3.9) 17.3 (4.3) 19.0 (5.7) 18.3 (4.8) 19.3 (5.3) 21.1 (5.9)20% 13.2 (3.5) 14.8 (4.2) 15.2 (5.0) 16.7 (5.9) 19.5 (6.3) 20.0 (6.7)40% 13.9 (4.9) 14.8 (8.3) 13.4 (4.5) 19.7 (7.1) 22.6 (6.0) 26.3 (10.3)50% 14.9 (5.6) 14.8 (3.7) 16.8 (6.0) 26.2 (7.7) 31.6 (8.9) 30.5 (15.1)60% 23.3 (8.1) 17.8 (4.3) 24.0 (9.1) 36.0 (13.0) 66.7 (59.4) 55.3 (32.6)
Figure 3 Mean k-values of the solu-tions, and their reaction to increasinghaemodilution; shortening of k suggestshypercoagulation. Albumin, shadeddiamond; 10% hydroxyethylstarch,open square; Dextran 70, solid triangle;Normal saline, open circle; RingerAcetate, star; 6% hydroxyethylstarch,solid circle.
Table 3 The a-angle values in degrees (means and (standard deviations)) for the various test solutions and dilutions. The superscriptednumerals show which solutions differed significantly between each other. Comparisons in this table are not significant unless thecorresponding p-value is less than 0.0033.
Dilution (1) NaCl 4,5,6 (2) Ringer4,5,6 (3) Albumin4,5 (4) 6% HES1,2,3 (5) 10% HES1,2,3,6 (6) Dextran1,2,5
10% 28.0 (7.4) 26.3 (8.1) 24.5 (9.1) 23.7 (6.4) 23.7 (6.8) 21.9 (7.6)20% 32.9 (6.4) 32.6 (11.2) 29.5 (10.3) 27.3 (10.1) 24.4 (8.0) 23.5 (8.5)40% 32.0 (9.4) 33.2 (8.0) 32.4 (7.3) 23.3 (6.8) 22.6 (8.5) 20.8 (7.1)50% 30.9 (10.6) 29.6 (7.4) 28.2 (9.4) 17.1 (5.6) 16.2 (4.6) 18.2 (7.4)60% 21.0 (7.5) 24.6 (5.6) 19.6 (5.6) 13.2 (5.0) 10.5 (4.5) 11.5 (5.0)
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1105
Figure 4 Mean a-angle values of thesolutions, and their reaction to increas-ing haemodilution; increasing a-anglesuggests hypercoagulation. Albumin,shaded diamond; 10% hydroxy-ethylstarch, open square; Dextran 70,solid triangle; Normal saline, open circle;Ringer Acetate, star; 6% hydroxy-ethylstarch, solid circle.
Table 4 The maximum amplitude values in mm (means and (standard deviations)) for the various test solutions and dilutions. Thesuperscripted numerals show which solutions differed significantly between each other. Comparisons in this table are not significantunless the corresponding p-value is less than 0.0033.
Dilution (1) NaCl 4,5,6 (2) Ringer4,5,6 (3) Albumin4,5 (4) 6% HES1,2,3 (5) 10% HES1,2,3,6 (6) Dextran1,2,5
10% 51.9 (8.0) 49.0 (6.5) 48.5 (5.9) 48.9 (7.5) 45.5 ( ⁄ .6) 43.3 (6.7)20% 52.5 (6.2) 48.5 (6.2) 52.6 (9.0) 46.4 (8.4) 43.4 (6.5) 42.9 (6.6)40% 51.8 (5.9) 47.6 (6.2) 49.5 (6.3) 45.4 (4.5) 36.1 (5.9) 36.7 (5.9)50% 48.9 (7.2) 43.0 (6.8) 45.2 (7.2) 39.9 (3.6) 31.5 (5.2) 34.0 (6.4)60% 42.3 (6.8) 37.6 (5.8) 41.8 (5.9) 35.9 (3.8) 27.1 (5.5) 29.1 (4.0)
Figure 5 Mean maximal amplitude val-ues of the solutions, and their reaction toincreasing haemodilution. An increasingmaximal amplitude suggests hypercoag-ulation. Albumin, shaded diamond; 10%hydroxyethylstarch, open square; Dex-tran 70, solid triangle; Normal saline,open circle; Ringer Acetate, star; 6%hydroxyethylstarch, solid circle.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1106 � 2002 Blackwell Publishing Ltd
into hypocoagulation when the dilution exceeded 20%
[10]. The fact that small to moderate amounts of
intravenous normal saline can cause hypercoagulation
was also reported by Janvrin et al. [11]. They found an
increased frequency of deep venous thrombosis in patients
receiving intravenous normal saline during surgery com-
pared to patients who were not administered intravenous
fluids. Others have also observed that a hypercoagulable
rather than a hypocoagulable state developing in associ-
ation with mild to moderate degrees of haemodilution
[12] and with a moderate degree of blood loss and
haemodilution in vivo [3]. Ruttman also concluded that
dilution of blood with both saline and Hemaccel caused
hypercoagulation measured by thrombelastography [6].
Clotting is a dynamic process, and measurements using
static end-points provide no information about the
quality of the clot or the dynamics of its formation [9].
Thrombelastography in contrast to conventional coagu-
lation laboratory tests is capable of comparing different
levels of activity throughout the integrated coagulation
process and is especially helpful in demonstrating hyper-
coagulability. We have shown that mild to moderate
haemodilution using different solutions causes increased
coagulability. The speed of clot formation was increased
significantly for all solutions at mild to moderate
dilution. The clot strength, represented by the maximal
amplitude, was moderately, but significantly increased,
indicating enhanced fibrin and platelet activity. Acceler-
ated clotting was seen until approximately 40% dilution
for the synthetic colloids and until 50% dilution for the
crystalloids. As blood becomes increasingly diluted the
initial hypercoagulation changes into normo- or hypo-
coagulation. However, with the crystalloids and albumin,
dilution has to be extreme before this change is reached.
Many hospitals have adopted protocols for substitution
of blood loss with plasma when intra-operative blood
loss exceeds 50% of the estimated blood volume. Some
authors using thrombelastography to assess coagulation
have challenged this notion and found that blood loss in
many instances can exceed 50% before administration of
plasma or platelets is required [3]. In our study the level of
dilution that transforms hypercoagulation into normo- or
hypocoagulation with crystalloids or albumin would
represent a blood loss of 70–90% of the circulating blood
volume in vivo. Tuman reported four patients with an
estimated blood volume loss of 80% who had trombo-
elastographic signs of increased coagulant activity [3]. If
these results represent the normal effects of haemodilution
with crystalloids or albumin in patients undergoing
surgery, the protocol for substituting blood loss with
plasma or coagulation factors may need re-evaluation.
Synthetic colloids have been implicated as a cause of
coagulopathy [13]. Experimental evidence indicates that
dextran decreases platelet function and the activity of
clotting factors [14]. According to our results, the effect of
dextran as anticoagulant appears to be dose-dependent.
Dextran in low concentrations activates the coagulation
cascade. Haemodilution above 20% was required to
achieve a hypocoagulative effect in-vitro.
Hydroxyethylstarch is considered to have haemostatic
profile similar to that of dextran [15, 16] although
conflicting evidence have been reported [17–20]. Based
on the results from the present investigation, hydroxy-
ethylstarch causes less hypercoagulation than the crystal-
loids and Albumin. Initially, we tested only the 10%
hydroxyethylstarch solution. This hydroxyethylstarch is a
hyperoncotic solution that acts by extracting fluid from
the interstitium diluting itself to iso-oncotic levels. The
6% solution may be better suited in these experiments to
represent the effects of hydroxyethylstarch on coagulation
in vivo. The importance of the molecular weight of
hydroxyethylstarch is often neglected. The high molecu-
lar weight starches interfere more with the factor VIII
complex than do the medium molecular weight startch
[20, 21]. Trumble et al. [18] reported coagulopathy with
hydroxyethylstarch of 480 kDalton while Davies found
that a higher molecular weight decelerated clotting [13].
A study on the effect of hydroxyethylstarch 40 (molecular
weight 56–61 kDalton) using conventional coagulation
tests discovered no changes beyond simple dilution [22].
We used hydroxyethylstarch with an average weight of
200 kDalton, but nevertheless found significantly attenu-
ated coagulation with dilution above 40% for 10%
hydroxyethylstarch and above 50% for 6% hydroxyeth-
ylstarch. At lower degrees of dilution normal or even
accelerated clotting was observed.
The mechanism by which haemodilution causes
enhanced coagulation has not been elucidated. Several
authors have attributed their observation of hypercoag-
ulation to surgical stress, tissue trauma and elevation in
serum catecholamine levels [3, 9, 23, 24]. Platelet adhesion
increases significantly after catecholamine and angiotensin
administration [23, 25]. The in vitro setting of our
experiments has eliminated the impact of drugs, endo-
thelium, tissue damage and stress, and the described
reaction to dilution is thus predominantly initiated
exclusively by the intrinsic coagulation system. Several
other factors including other drugs or ASA status and
physical factors such as hypothermia may modify the
effect of dilution and so alter the coagulation effects of
intravenous solutions in clinical practice.
In summary this study demonstrates an increased
coagulative activity measured by thrombelastography
in vitro occurring at low and medium levels of dilution
with all tested crystalloids and colloids. At more than 40%
dilution the hypercoagulation decreases and in the case of
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1107
dextran and hydroxyethylstarch turns into hypocoagula-
tion. Dilution with crystalloids and albumin has to be
more than 50%; this represents a substituted blood loss of
approximately 70%. If these findings can be reproduced
in vivo, they may have implications on transfusion practice
and prophylaxis against thrombosis.
References
1 Consensus conference: perioperative red blood cell transfu-
sion. Journal of American Medical Association 1988; 260: 2700–3.
2 Tocantins L, Carrel T, Holburn RH. The clot accelerating
effect of dilution on blood and plasma. Relation to the
mechanism of coagulation of normal and hemophilic blood.
Blood 1951; 6: 720–39.
3 Tuman KJ, Speiss BD, McCarthy RJ, Ivankovich AD.
Effects of progressive blood loss on coagulation as measured
by thrombelastography. Anesthesia and Analgesia 1987; 66:
856–63.
4 Howland WS. A comparison of intraoperative measurements
of coagulation. Anesthesia and Analgesia 1974; 53: 657–63.
5 Ng KF, Lo JW. The development of hypercoagulability
state, as measured by thrombelastography, associated with
intraoperative surgical blood loss. Anaesthesia and Intensive
Care 1996; 24: 20–5.
6 Ruttmann TG. Haemodilution induces a hypercoagulable
state. British Journal of Anaesthesia 1996; 76: 412–14.
7 McLoughlin TM. Profound normovolemic hemodilution:
hemostatic effects in patients and in a porcine model.
Anesthesia and Analgesia 1996; 83: 459–65.
8 Hartert H. Blutgerinnugsstudien mit der Thrombelasto-
graphie. Einen neuen Untersuchungsverfahren. Klinische
Wochenschrift 1948; 16: 257.
9 Mallett S, Cox D. Thrombelastography. British Journal of
Anaesthesia 1992; 69: 307–13.
10 Tsunehiro N, Isao M, Teruo K, Etsutaro I. An experimental
study on the effect of plasma expanders on blood coagula-
bility. Bulletin Tokyo Medical Dental University 1979; 26:
25–32.
11 Janvrin S, Davies G, Greenhalgh R. Postoperative deep vein
thrombosis caused by intravenous fluids during surgery.
British Journal of Surgery 1980; 67: 690–3.
12 Heather B, Jennings S, Greenhalgh M. The saline dilution
test – a preoperative predictor of DVT. British Journal of
Surgery 1980; 67: 63–5.
13 Davies M. The role of colloids in blood conservation.
International Anesthesiology Clinics 1990; 28: 205–9.
14 Aberg M, Bergentz S, Hedner U. Effect of dextran on factor
VIII (antihemophilic factor) and platelet function. Annals of
Surgery 1979; 189: 243–7.
15 Mischler JM. Synthetic plasma volume expanders: their
pharmacology, safety and clinical efficacy. Clinical Haema-
tology 1984; 13: 75–92.
16 London MJ, Ho JS, Triedman JK, et al. A randomized
clinical trial of 10% pentastarch (low molecular weight
hydroxyethyl starch) versus 5% albumin for plasma Volume
expansion after cardiac operations. Journal of Thoracic
Cardiovascular Surgery 1989; 97: 785–97.
17 Damon L, Adams M, Stricker RB, Ries C. Intracranial
bleeding during treatment with hydroxyethylstarch. New
England Journal of Medicine 1987; 317: 964–5.
18 Trumble ER, Muizelaar JP, Myseros JS, Choi SC, Warren
BB. Coagulopathy with the use of hetastarch in the treat-
ment of vasospasm. Journal of Neurosurgery 1995; 82: 44–7.
19 Treib J, Haass A, Pindur G, et al. HES, 200 ⁄ 0.5 is not HES
200 ⁄ 0.5. Influence of the C2 ⁄ C6 hydroxyethylation ratio of
hydroxyethyl starch (HES) on haemorrheology, coagulation
and elimination kinetics. Thrombosis and Haemostasis 1995;
74: 1452–6.
20 Mortier E, Ongenae M, De BL, et al. In vitro evaluation of
the effect of profound haemodilution with hydroxyethyl
starch 6%, modified fluid gelatin 4% and dextran 40, 10% on
coagulation profile measured by thromboelastography.
Anaesthesia 1997, 52: 1061–4.
21 Strauss RG. Pentastarch may cause fewer effects on coagu-
lation than hetastarch. Transfusion 1988; 28: 257–60.
22 Treib J. All medium starches are not the same. Influence of
the degree of hydroxyethyl substitution of hydroxyethyl
starch on plasma volume, haemorrheologic conditions, and
coagulation. Transfusion 1996; 36: 450–5.
23 Barrer MJ. Platelet function. Anesthesiology 1977; 46: 202–
11.
24 Gibbs NM, Crawford GP, Michalopoulos N. Postoperative
changes in coagulant and anticoagulant factors after
abdominal aortic surgery. Journal of Cardiothoracic Vascular
Anesthesia 1992; 6: 680–5.
25 Uza G, Crisnic I. Effect of angiotensin II upon platelet
adhesiveness and the thrombelastogram in patients with
essential arterial hypertension. Pathology Europe 1975; 10:
327–32.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1108 � 2002 Blackwell Publishing Ltd
FORUM
Estimating unmeasured anions in critically ill patients:
Anion-gap, base-deficit, and strong-ion-gap
D. A. Story,1 S. Poustie2 and R. Bellomo3
1 Joint Coordinator of Research, 2 Trial Coordinator and 3 Director of Research, Department of Intensive Care, Austin
and Repatriation Medical Centre, Studley Road, Heidelberg, Victoria 3084, Australia
Summary
We used 100 routine blood samples from critically ill patients to establish whether correcting the
anion-gap and base-deficit for decreased plasma albumin improves agreement with the strong-ion-
gap for estimating unmeasured anions and whether the modifications increase the proportion of
samples with levels of anion-gap or base-deficit above the reference ranges. We used Bland)Altman analyses to compare the methods of estimating unmeasured ions. Compared with the
strong-ion-gap, modification reduced the limits of agreement for both the anion-gap and the base-
deficit. The bias for the base-deficit was also reduced but the bias for the anion-gap was increased.
The proportion of samples with an anion-gap > 22 meq.l)1 increased from 4 to 29% (p < 0.001),
and the proportion with a base-deficit > 5 meq.l)1 in creased from 8 to 42% (p < 0.001). Con-
sequently, metabolic acidosis from unmeasured ions in critically ill patients maybe more frequent
than often recognised.
Keywords Acid)base measurement.
........................................................................................................
Correspondence to: Dr D. A. Story
Accepted: 29 June 2002
We have previously examined the acid-base status of
routine blood samples from critically ill patients. We
found decreased plasma albumin concentration in all the
samples and an overall metabolic alkalosis [1]. Several
groups have suggested that decreased plasma albumin can
conceal other acid)base disturbances [2–4]. One such
disturbance is increased unmeasured ions. In the Stewart
approach to acid)base disorders [5], the total concentra-
tion of unmeasured anions can be estimated using the
strong-ion-gap equation [3, 6]. Fencl and colleagues [3]
suggested improving the accuracy of the anion-gap
approach for detecting unmeasured ions in the critically
ill by modifying for decreased plasma albumin.
Balasubramanyan et al. [2] used a similar modification
of the base-deficit calculation for altered plasma albumin
concentration. This modification is another way of
incorporating the Stewart approach [4, 7]. Neither Fencl
et al. [3] nor Balasubramanyan et al. [2] compared the
modified anion-gap and modified base-deficit measure-
ments to the strong-ion-gap.
Extending our previous data [1], we examined wheth-
er: (i) modifying the anion-gap and base-deficit improves
agreement with the strong-ion-gap for estimating
unmeasured anions, and (ii) whether such modifications
increase the proportion of samples with levels of anion-
gap or base-deficit well above the reference ranges.
Methods
Data were prospectively collected from intensive care unit
records at the Austin and Repatriation Medical Centre, a
tertiary referral hospital affiliated to the University of
Melbourne. All samples were routine morning samples
taken from arterial lines in patients requiring intensive
care management. No additional sampling was required
and collected data did not identify patients. The Austin
and Repatriation Medical Centre Human Research
Ethics Committee waived the need for informed consent.
Arterial blood samples were collected in heparinised
blood-gas syringes (Rapidlyte, Chiron Diagnostics, East
Walpole, MA, USA) and analysed in the intensive care
unit (ICU) blood gas analyser (Ciba Corning 865, Ciba
Corning Diagnostics, Medfield, MA, USA). The analyser
measured the samples at 37 �C and ICU nursing staff who
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1109
had been taught to use the machine performed the
analyses. Samples were not stored on ice. We recorded
lactate and ionised calcium concentrations and the
calculated base-deficit and bicarbonate concentrations.
For each data set, another routine sample was drawn at
the same time from the same arterial sampling point using
a vacuum technique with clot activating tubes for serum or
lithium heparin tubes for plasma (Vacuette, Greiner
labourtechnik, Kremsmunster, A-4550, Austria). Our
laboratory gives an identical reference range for serum
and plasma levels for the variables measured in this study.
We included the data if the nurse recorded that the two
samples were taken simultaneously. These samples were
sent to the hospital core laboratory in the Division of
Laboratory Medicine. Plasma underwent a multicompo-
nent analysis (Hitachi 747, Roche Diagnostics, Sydney,
Australia). Scientific staff from the hospital clinical chem-
istry department analysed samples. Samples were not
stored on ice. We recorded sodium, potassium, magne-
sium, chloride, phosphate and albumin concentrations.
We calculated the anion-gap, modified-anion-gap, and
the modified-base-deficit.
Anion-gap
The anion-gap was estimated as
Anion-gap (meq.l-1) ¼ ½Naþ� þ ½Kþ� � ½HCO�3 �
� ½Cl��
Figge et al. [8] used regression analysis to modify the
anion-gap for changes in plasma albumin concentration.
We used this correction factor to modify the anion-gap
for changes in albumin in which 42 g.l)1 was the median
value of the reference range for albumin in our hospital
and 0.25 the correction factor:
Modified anion-gap (meq.l�1Þ ¼ anion gap
þ 0:25 � ½42 � measured albumin concentration�:
Base-deficit
The base-deficit was calculated as the negative of the
standard base-excess reported from the blood gas ma-
chine. The contribution to the measured base-deficit by
unmeasured ions was calculated by taking the negative
values of base-excess corrected for the effects of changes
in the strong-ion-difference and total weak acid concen-
tration [2, 4, 7].
We corrected the base-deficit (base excess) using the
approach developed by Fencl and modified by others
[2, 4, 7].
The major determinants of the strong-ion-difference
are the sodium and chloride concentrations. The base-
excess is modified for the effect of changes in free water
on the sodium concentration and for changes in the
chloride concentration [2]:
sodium modification (meq.l�1Þ ¼ 0:3� ð½Naþ� � 140Þ;
chloride modification (meq.l�1Þ¼ 102�ð½Cl�� � 140=½Naþ�Þ:
Albumin is the major determinant of total weak-acid
concentration. Gilfix et al. [4] used regression analysis
from Figge’s group [9] to calculate the base-excess effect
of changes in plasma albumin. This calculation was
simplified by Balasubramanyan et al. [2]:
albumin modification (meq.l�1Þ¼ 0:34 � ð45 � ½albumin�Þ:
Therefore, the base-excess due to unmeasured ions
equals calculated base-excess from blood gas analysis
minus sodium modification minus chloride modification
minus albumin modification [2].
For comparison with the strong-ion-gap we used the
negative of the base-excess, the base-deficit. Both the
base-deficit and the strong-ion-gap become more positive
as the concentration of unmeasured-anions increases.
Strong-ion-gap
We calculated the strong-ion-gap with the all the ions
currently measured at our hospital [10]
strong-ion-gap (meq.l�1Þ¼ ½Naþ� þ ½Kþ� þ ½Mg2þ� þ ½Ca2þ� � ½Cl��� ½lactate anions� � ½albumin anions�� ½phosphate anions� � ½bicarbonate�:
Hydrogen, hydroxyl and carbonate ions were excluded
because of their micromolar concentrations [9].
The anions formed from the weak acids albumin and
phosphate were calculated from Figge et al.’s equations [9]:
albumin anions (meq.l�1Þ¼ ½albumin� g.l�1 � ð0:123 � pH � 0:631Þ
phosphate anions (meq.l�1Þ¼ ½phosphate� mmol.l�1 � ð0:309 � pH � 0:469Þ:
Statistical analysis
Data were collected from patient charts and the hospital
computer system and were stored on a computer
spreadsheet (Excel, Microsoft, Seattle, WA, USA). All
statistical calculations were performed using STATVIEW
software (Abacus Concepts, Berkeley, CA, USA). We
used the limits of agreement method of Bland & Altman
[11, 12] to determine the agreement between the strong-
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1110 � 2002 Blackwell Publishing Ltd
ion-gap and each of the anion-gap, modified-anion-gap,
base-deficit and modified-base-deficit. We proposed that
acceptable agreement was a bias of up to ± 1 meq.l)1 and
limits of agreement less than bias ± 2 meq.l)1 so making a
total difference of up to 3 meq.l)1.
At our hospital the reference range for the anion gap is
10–20 meq.l)1 and for the base-deficit is )3 to 3 meq.l)1.
We examined the number of samples that, after
modification, exceeded thresholds that were above the
reference ranges for anion-gap (22 meq.l)1) [3] and base-
deficit (5 meq.l)1) [2]. Chi-squared analysis was used to
analyse the number of patients who exceeded these
thresholds for anion-gap and base-deficit before and after
modification. A p-value < 0.05 was considered statistically
significant.
Results
One hundred pairs of data were collected from 100
patients. Data collection was limited by sample pairs not
being drawn at the same time or failure to measure the
required variables.
Overall, the sodium and chloride concentrations were
within the normal range, which led to the strong-ion-
difference having little effect on the modified base-deficit
(Table 1). All samples had a plasma albumin below the
lower limit of our normal range (36 g.l)1) (Table 1).
These low plasma albumin levels had considerable effects
on the modified anion-gap and modified base-deficit.
Modification reduced the bias between the strong-ion-
gap and the base-deficit but increased the bias between
the strong-ion-gap and the anion-gap (Figs 1–4). Mod-
ification improved the precision of the agreement
between the strong-ion-gap and both the anion-gap and
the base-deficit (Figs 2 and 4). All four estimates had
limits of agreement well outside our proposed acceptable
limits of a bias of up to ± 1 meq.l)1 and limits of
agreement within ± 2 meq.l)1.
Fewer than 10% of samples had an anion-gap
> 22 meq.l)1 or a base-deficit > 5 meq.l)1 before mod-
ification (Table 2). Modification revealed an anion-gap
> 22 meq.l)1 in a quarter of the samples and base-deficit
> 5 meq.l)1 in a third of the samples (Table 2).
Discussion
Main findings
We found that modifying the anion-gap and base-deficit
for decreased plasma albumin improved agreement with
the strong-ion-gap. Both estimates, however, were out-
side our proposed acceptable limits of a bias (± 1 meq.l)1)
and limits of agreement (± 2 meq.l)1). The agreement
was better with the base-deficit than the anion-gap. The
unmeasured ions did not include lactate because lactate
was measured and used in the strong-ion-gap calculation.
Figure 1 Bland)Altman plot of thedifferences in unmeasured anion con-centration between the strong-ion-gapand the base deficit and the average ofthe two measures [(strong-ion-gap +anion-gap) ⁄ 2]. The full lines are thelimits of agreement and the dashed lineis the bias. In meq.l)1 the summarystatistics are bias )6.7, lower limit )11.4and upper limit )2.0.
Table 1 Acid)base variables for 100 samples.
Variable Median Interquartile range
pH 7.43 7.38 to 7.48CO2; kPa 5.5 4.8 to 6.3HCO3; mmol.l)1 25.4 23.7 to 30.6Sodium; mmol.l)1 140 138 to 144Chloride; mmol.l)1 103 99 to 106Lactate; mmol.l)1 1.6 1.1 to 2.4Albumin; g.l)1 23 19 to 27Strong-ion-gap; meq.l)1 8.5 6.0 to 11.4Anion-gap; meq l)1 14.9 12.4 to 17.5Modified anion-gap; meq.l)1 19.2 16.3 to 22.8Base-deficit; meq.l)1 ) 1.7 ) 6.6 to 1.3Modified base-deficit; meq.l)1 4.4 2.0 to 8.3
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1111
Further, in a quarter of the samples, modification
increased levels of anion-gap or base-deficit to above
the normal range.
Comparison with other studies
As in previous studies, we found that modifying the
anion-gap [6] or base-deficit [4] for decreased plasma
albumin improved the agreement with the strong-ion-
gap. We used Bland)Altman analyses rather than corre-
lation analyses that were used in the other studies [4, 6].
Correlation statistics are likely to overestimate the
relationship between estimates and cannot quantify
agreement [11].
In quantifying the effect of falling plasma albumin
concentration Figge et al. found that a 10 g.l)1 decrease in
plasma albumin required a 2.5 meq.l)1 correction in the
anion-gap [8]. In response to criticism from Reilly &
Anderson [13] and as part of a further study of their data,
Fencl et al. [3] examined a subgroup with base-deficit and
anion-gap within the reference range. After adjusting for
decreased plasma albumin the proportion of patients with
an anion-gap of > 22 meq.l)1 increased from 3 of 20 to
13 of 20. In Balasubramyan et al.’s study of 255 critically
ill children [2], after modifying for both decreases in
plasma albumin and changes in strong-ion-difference, the
proportion of patients with a base-deficit > 5 meq.l)1
increased from 27 to 54%. We also found important
increases in the number of samples that exceeded an anion
gap of 22 meq.l)1 [3] or a base-deficit of > 5 meq.l)1 [2]
or both after modification.
Figure 2 Bland)Altman plot of thedifferences in unmeasured anion con-centration between the strong-ion-gapand the modified anion-gap, and theaverage of the two measures [(strong-ion-gap + modified anion-gap) ⁄ 2]. Thefull lines are the limits of agreement andthe dashed line is the bias. In meq.l)1
the summary statistics are bias )11.3,lower limit )14.7 and upper limit )7.9.
Figure 3 Bland)Altman plot of thedifferences in unmeasured anion con-centration between the strong-ion-gapand the base deficit, and the average ofthe two measures [(strong-ion-gap +base-deficit) ⁄ 2]. The full lines are thelimits of agreement and the dashed lineis the bias. In meq.l)1 the summarystatistics are bias 11.0, lower limit 2.0,and upper limit 20.0.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1112 � 2002 Blackwell Publishing Ltd
Limitations
There is currently no ‘gold standard’ for detecting the
presence or concentration of unmeasured ions [3]. One
limitation of using the strong-ion-gap is that the
measurement errors for each of the components of the
calculation are added. We found, even with this limita-
tion, improved agreement particularly between the
strong-ion-gap and the base-deficit.
Clinical importance
Low plasma albumin is almost universal in critically ill
patients. Modifying the anion-gap for decreased plasma
albumin increases the proportion of patients with anion-
gap concentration well above the reference range. Even
with modification, however, the anion-gap agrees poorly
with the strong-ion-gap, particularly due to a large bias.
Modifying the base-deficit for decreased plasma albu-
min increases the number of samples with metabolic
acidoses. Further, modifying the base-deficit improves
agreement with the strong-ion-gap. The strong-ion-gap
is one method to estimate unmeasured ions but is limited
by having many variables and therefore many combined
measurement errors. The measurement error of the
strong-ion-gap maybe up to 10 meq.l)1 [14]. Although
the agreement between the strong-ion-gap and the
modified base-deficit was greater than our defined
acceptable limits (3 meq.l)1), measurement error in the
strong-ion-gap may have contributed substantially.
Because the modified base-deficit requires fewer mea-
surements and has less measurement error, we believe that
modifying the base-deficit to estimate the acid-base effect
of unmeasured ions is a reasonable alternative to using the
strong-ion-gap. This is particularly so if the clinically
important threshold for the unmeasured ion effect on
base-deficit is 5 meq.l)1 [2].
We support combining the base-deficit with a ‘Stewart’
style analysis of a patient’s acid)base status [2, 4, 7]. The
base-deficit can be used to examine the overall metabolic
component [15] as well as the effects of the strong-ion-
difference, weak acids (mainly albumin) and unmeasured
ions. Therefore, one variable, base-deficit, can be used to
analyse the components of a patient’s acid)base status
with simple calculations. The analysis of these base-deficit
effects, particularly the effects of the strong-ion-differ-
ence, will probably need further work.
We conclude that of the four estimates we tested the
modified base-deficit agrees most closely with the strong-
ion-gap. Further, metabolic acidosis due to unmeasured
ions in critically ill patients maybe more frequent than
often recognised. The identities of the unmeasured ions
and the importance of these findings to physiological
Figure 4 Bland)Altman plot of thedifferences in unmeasured anion con-centration between the strong-ion-gapand the modified base-deficit, and theaverage of the two measures [(strong-ion-gap + modified base-deficit) ⁄ 2].The full lines are the limits of agreementand the dashed line is the bias.In meq.l)1 the summary statistics arebias 3.5, lower limit )1.6 and upperlimit 8.5.
Table 2 Effect of modification on the anion-gap and base-deficit.
VariableAnion-gap> 22 meq.l)1 (%)
Base-deficit> 5 meq.l)1 (%)
Before modification 4 8After modification 29 42Absolute increase 25 3495% confidence interval
of increase18)34 26)44
p-Value (before vs. aftermodification)
< 0.001 < 0.001
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1113
homeostasis and outcome for patients need to be explored
further.
Acknowledgements
We wish to thank the ICU nursing staff; and Mr Sam
Tobgui, Senior Scientist from the Division of Laboratory
Medicine, for advice on the assays.
References
1 Story DA, Poustie S, Bellomo R. Quantitative physical
chemistry analysis of acid)base disorders in critically ill
patients. Anaesthesia 2001; 56: 530–3.
2 Balasubramanyan N, Havens PL, Hoffman GM. Unmea-
sured anions identified by the Fencl)Stewart method predict
mortality better than base excess, anion gap, and lactate in
patients in the pediatric intensive care unit. Critical Care
Medicine 1999; 27: 1577–81.
3 Fencl V, Jabor A, Kazda A, Figge J. Diagnosis of metabolic
acid)base disturbances in critically ill patients. American
Journal of Respiratory and Critical Care Medicine 2000; 162:
2246–51.
4 Gilfix BM, Bique M, Magder S. A physical chemical
approach to the analysis of acid)base balance in the clinical
setting. Journal of Critical Care 1993; 8: 187–97.
5 Stewart PA. Modern quantitative acid)base chemistry.
Canadian Journal of Physiology and Pharmacology 1983; 61:
1444–61.
6 Kellum JA, Kramer DJ, Pinsky MR. Strong ion gap: a
methodology for exploring unexplained anions. Journal of
Critical Care 1995; 10: 51–5.
7 Fencl V, Leith DE. Stewart’s quantitative acid)base chem-
istry: applications in biology and medicine. Respiratory
Physiology 1993; 91: 1–16.
8 Figge J, Jabor A, Kazda A, Fencl V. Anion gap and hypo-
albuminemia. Critical Care Medicine 1998; 26: 1807–10.
9 Figge J, Mydosh T, Fencl V. Serum proteins and acid)base
equilibria: a follow-up. Journal of Laboratory and Clinical
Medicine 1992; 120: 713–9.
10 Hayhoe M, Bellomo R, Liu G, McNicol L, Buxton B. The
aetiology and pathogenesis of cardiopulmonary bypass-as-
sociated metabolic acidosis using polygeline pump prime.
Intensive Care Medicine 1999; 25: 680–5.
11 Bland JM, Altman DG. Statistical methods for assessing
agreement between two methods of clinical measurement.
Lancet 1986; 1: 307–10.
12 Bland J, Altman D. Measuring methods in medical research.
Statistical Methods in Medical Research 1999; 8: 135–60.
13 Reilly R, Anderson R. Interpreting the anion gap. Critical
Care Medicine 1998; 26: 1771.
14 Story D, Poustie S, Bellomo R. Comparison of three
methods to estimate plasma bicarbonate in critically ill pa-
tients. Henderson)Hasselbach, enzymatic, and strong-ion-
gap. Anaesthesia and Intensive Care 2001; 29: 585–90.
15 Siggard-Anderson O. The Acid-Base Status of the Blood, 4th
edn. Copenhagen: Munksgaard, 1974.
FORUM
Anaesthesia for cardioversion: a comparison
of sevoflurane and propofol*
S. Karthikeyan,1� S. Balachandran,1� J. Cort,2 M. H. Cross3 and M. Parsloe4
1 Clinical Fellow in Cardiothoracic Anaesthesia, 2 Specialist Registrar in Anaesthesia, and 3 Consultant Anaesthetist,
Yorkshire Heart Centre, Leeds General Infirmary, Leeds, UK, 4 Consultant Anaesthetist, Conquest Hospital, The Ridge,
St Leonards-on-Sea TN37 7RD, UK
Summary
This study compared the induction time, haemodynamic changes, recovery characteristics and patient
satisfaction for sevoflurane and propofol when used as the main anaesthetic agents for cardioversion.
Sixty-one unpremedicated patients scheduled for elective cardioversion were anaesthetised with
either inhaled sevoflurane 8% or an intravenous propofol target-controlled infusion set at 6 lg.ml)1.
There was no significant difference in induction time between the two groups: mean
(SD) ¼ 90.1(40) s in the sevoflurane group vs. 83.7(35) s in the propofol group. Mean (SD) time to
recovery was significantly shorter in the sevoflurane group than in the propofol group: 318 (127) s vs.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1114 � 2002 Blackwell Publishing Ltd
738 (355) s, respectively, p < 0.001. At recovery, the patients in the propofol group had significantly
lower systolic and diastolic blood pressures than those in the sevoflurane group, p < 0.001. The
incidence of complications was low in both groups, with similar patient satisfaction expressed after the
procedure. We conclude that sevoflurane is a suitable choice for anaesthesia for cardioversion and may
provide greater haemodynamic stability than a target-controlled infusion of propofol.
Keywords Anaesthetics, inhalational; sevoflurane. Anaesthetics, intravenous; propofol. Heart;
cardioversion.
........................................................................................................
Correspondence to: Dr M. Parsloe
*This study was presented as a poster presentation with oral discussions
in the 9th ESA Annual Meeting, Gothenburg, Sweden 7–10 April
2001.
�Present address: Consultant Anaesthetists, University Hospital of
Wales, Cardiff CF14 4XW, UK.
Accepted: 3 July 2002
Cardioversion has been used in clinical practice to
convert abnormal cardiac rhythms to sinus rhythm since
1962 [1]. Patients undergoing cardioversion require a
short anaesthetic for which various agents have been
used. The ideal agent should produce rapid loss of
consciousness, cardiovascular stability and prompt recov-
ery with few side-effects. Thiopental, etomidate and
benzodiazepines have been widely used for cardiover-
sion, although they can be associated with prolonged
recovery [2], haemodynamic instability [3] and myo-
clonic activity [4]. Studies by Valtonen et al. [2] and
Hullander et al. [4] considered propofol to be an
ideal agent for anaesthesia for cardioversion. The use
of inhalational anaesthetic agents in anaesthesia for
cardioversion has not previously been studied. Sevoflu-
rane and propofol both have properties such as rapid
induction and quick recovery that are suitable for day-
case anaesthesia [5, 6]. The purpose of our study was to
compare sevoflurane and a target-controlled infusion
(TCI) of propofol for induction and maintenance of
anaesthesia for cardioversion.
Methods
After local research ethics committee approval and writ-
ten, informed consent, 61 patients who were scheduled
for elective cardioversion were included in this rando-
mised trial. All were given an information sheet one
week before the procedure. Patient were not studied if
they were aged < 18 years, ASA physical status IV,
considered to be at risk of regurgitation or with a
potentially difficult airway or a body mass index
(BMI) > 35 kg.m)2. Patients were randomly allocated to
either of the treatment groups using computer-generated
random numbers.
No premedication was given. All patients had an
intravenous cannula sited in their non-dominant hand
and were given intravenous glycopyrronium 200 lg
during a 3 min pre-oxygenation period. On induction
of anaesthesia, both groups breathed 50% oxygen in
nitrous oxide from a Mapleson D breathing system that
had a 5 l reservoir bag. Patients in the propofol group
received a propofol TCI with the target concentration set
at 6 lg.ml)1. This concentration was continued through-
out the procedure. The total amount of propofol used
was recorded. The sevoflurane group received 8%
sevoflurane in 50% oxygen in nitrous oxide, with the
patient being asked to breathe normally during induction.
Anaesthesia was maintained with sevoflurane during the
procedure.
Just before induction, patients were asked to tap the
index finger of their dominant hand repeatedly on their
chest. Cessation of finger tapping was taken as the time
of loss of consciousness. One minute after this, the DC
cardioversion shock was applied. Induction time, dura-
tion of anaesthesia, number of shocks, energy used in
joules and successful conversion to sinus rhythm were
recorded. At the end of the procedure, the patient was
transferred to the recovery area. A recovery nurse, who
was unaware of the anaesthetic technique used, called the
patient’s name every 30 s, and the time of eye opening
was taken as the time to wakening.
Heart rate, non-invasive blood pressure, SpO2 and
respiratory rate were recorded before and after induction,
just after the application of the DC shock and on
awakening. Complications such as apnoea, cough,
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1115
abnormal movements and postoperative nausea and
vomiting (PONV) were recorded. According to local
guidelines, patients were discharged when their systolic
blood pressure was within 20% of baseline values, when
they were steady in gait and had minimal nausea and
vomiting. Before discharge, patients completed a ques-
tionnaire regarding their satisfaction with the anaesthetic.
Data were analysed using a computer statistical package
(STATVIEW, Version 5.0.1, SAS Institute Inc, Cary, NC,
USA). Student’s t-test was used for continuous variables
and the Chi-squared test was used for nominal data. One
way analysis of variance with Fisher’s post-hoc least
significant difference test was applied to the haemodya-
namic variables. A p-value of < 0.05 was considered
significant.
Results
The two groups were comparable with regard to age,
gender, BMI, as well as initial systolic and diastolic blood
pressure (Table 1). Atrial fibrillation was the most com-
mon arrhythmia in both groups (93% of the sevoflurane
group, 87% of the propofol group), with atrial flutter in
the remaining patients.
Induction and recovery characteristics are shown in
Table 2. Induction time was similar in the two groups.
Time to eye opening was quicker in the sevoflurane
group than in the propofol group, mean (SD) ¼ 318
(127) s vs. 738 (355) s, respectively, p < 0.001. The
mean (SD) propofol dose used for the procedure was
2.7 (0.49) mg.kg)1. There were no significant differ-
ences in heart rate (Fig. 1) or return to sinus rhythm,
which occurred in 94% of the propofol group and 84%
of the sevoflurane patients. Both groups showed a
decrease in systolic and diastolic blood pressure after
induction and in the period after cardioversion (Fig. 2).
The decrease in blood pressure was not significantly
different between the groups, except at the time of
wakening, when it was significantly lower in the
propofol group (p < 0.0001). No patient was consid-
ered to require haemodynamic support to treat this
decrease in blood pressure.
Respiratory rate and SpO2 are shown in Table 3. One
patient in the sevoflurane group desaturated to 88% after
cardioversion. Neither laryngospasm nor bronchospasm
were observed in any patient and there were no
significant differences in the incidence of other compli-
cations (Fig. 3). Patient responses to the postoperative
questionnaire are shown in Table 4. In both groups,
50% of patients thought the anaesthetic was pleasant,
with only 5% of each group finding it unpleasant. The
majority of patients would opt for the same anaesthetic for
future procedures.
Discussion
There have been no studies comparing propofol with
an inhalational agent for anaesthesia for cardioversion. In
contrast to comparative studies in day-case anaesthesia [5,
7], there was no significant difference in time to loss of
consciousness as assessed by finger tapping. We believe
this is because we used a propofol infusion rather than a
bolus technique [7]. Recovery times in our study suggest
that sevoflurane produces significantly quicker recovery
than propofol TCI at a target concentration of 6 lg.ml)1.
Although we did not perform sophisticated tests to assess
recovery and �street fitness�, our results confirmed the
findings of previous studies that recovery time is more
rapid with sevoflurane in day-case anaesthesia patients [5,
7]. Studies comparing sevoflurane and propofol in day-
case anaesthesia suggest that the two agents have similar
characteristics [6, 8]. A quicker return to consciousness
could be an advantage in decreasing the time spent in the
recovery ward, but this was not found to be statistically
significant in our study.
In order to standardise the injection technique, we
chose TCI for the propofol group. Target induction
plasma concentration with propofol TCI is usually
4–6 lg.ml)1 [9]. Standard clinical practice with TCI is
to induce the patient with 6 lg.ml)1 and then to reduce
to 4 lg.ml)1 after loss of consciousness. In our study, we
chose to maintain a TCI concentration of 6 lg.ml)1
because of the significant stimulus of cardioversion in an
opioid-free anaesthetic technique. The mean (SD) prop-
ofol dose used for the procedure was 2.7 (0.49) mg.kg)1,
Table 1 Demographic data. Values are mean (SD) or number.
Propofol group(n = 31)
Sevoflurane group(n = 30)
Age; years 67 (11) 68 (8.7)Gender; male: female 21 : 10 22 : 8Body mass index; kg.m)2 28 (3.9) 29 (3.4)
Table 2 Induction and recovery characteristics of patientsreceiving sevoflurane or propofol. Values are mean (SD).
Propofol group(n = 31)
Sevoflurane group(n = 30)
Induction time; s 83.7 (35) 90.9 (40)Time to eye opening; s 738 (355) 318 (127)*Duration of procedures; s 253 (85) 232 (83)Hospital stay; h: min 2 : 43 (0.55) 2 : 08 (0.43)
*Significantly different from propofol group, p < 0.001.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1116 � 2002 Blackwell Publishing Ltd
which is similar to that given in a previous study that used
an intermittent bolus technique [2].
Both agents were associated with decreased blood
pressure after induction, but the patients in the propofol
group had significantly lower systolic and diastolic
pressures in the recovery area. Decreased blood pressure
after cardioversion may be a disadvantage in patients who
have a compromised myocardium [10] and may also delay
discharge to the ward. Some volatile anaesthetic agents
such as halothane sensitise the ventricle to catecholamines
[11], but others may have a stabilising effect similar to that
of calcium channel-blocking drugs [12]. Volatile agents
may also depress the sinus node leading to an increased
atrial refractory period or decreased atrioventricular
conduction [13]. In this respect, sevoflurane seems to
have properties similar to isoflurane and similar dose-
dependent effects on blood pressure, but a lower
Table 3 Peripheral oxygen saturation and respiratory rate.Values are mean (SD).
Propofol group(n = 31)
Sevoflurane group(n = 30)
SpO2 before induction; % 98 (2) 98 (1.7)SpO2 after induction; % 99 (1) 99 (1)SpO2 after cardioversion; % 98 (1.7) 98 (3)SpO2 in recovery area; % 98 (1.4) 98 (2)Respiratory rate before
induction; breath.min)121 (5) 20 (6)
Respiratory rate afterinduction; breath.min)1
19 (10) 23 (8)
Respiratory rate aftercardioversion; breath.min)1
26 (7) 27 (7)
Respiratory rate in recoveryarea; breath.min)1
24 (4) 25 (4)
Figure 2 Changes in blood pressure.Error bars indicate SD. *Significantlydifferent from sevoflurane group,p < 0.001.
Figure 1 Changes in heart rate. Errorbars indicate SD.
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1117
incidence of tachycardia and coronary vasodilatation [14].
In our study, there was no difference between the groups
in the success rate of cardioversion, and new arrhythmias
were not observed in either group. Propofol is not known
to have any myocardial stabilising properties but can
produce more hypotension after bolus dosing when
compared with sevoflurane inhalation induction [7]. It
has been proposed that this is due to a peripheral
vasodilator effect rather than direct cardiac depression
[15]. Other workers have suggested that hypotension can
be minimised with a slow induction by infusion [16], but
even a slow propofol induction may be associated with a
lower mean arterial pressure compared with sevoflurane
inhaled induction [17]. Our study appears to demonstrate
an increased haemodyanamic effect of propofol TCI that
extends into the period after awakening when compared
with sevoflurane for induction and maintenance.
In contrast to previous studies [5, 7], the incidence of
PONV was similar but low in both groups. This may be
due to the opioid-free anaesthetic technique. Patient
satisfaction was similar for both agents, with pain on
injection being the main reason for complaint in 16% of
the propofol group and the unpleasant smell being the
main complaint in 18% of the sevoflurane patients.
Our findings suggest that induction and maintenance
of anaesthesia with sevoflurane has advantages for
cardioversion when compared with propofol TCI at
6 lg.ml)1. There is more rapid awakening and better
maintenance of blood pressure in recovery. We believe
that sevoflurane is a suitable agent for induction and
maintenance of anaesthesia for elective cardioversion.
References
1 Lown B, Amarsingham R, Newman J. New method for
terminating cardiac arrhythmias. Journal of the American
Medical Association 1962; 182: 548–55.
2 Valtonen M, Kanto J, Klossner J. Anaesthesia for cardio-
version: a comparison of propofol and thiopentone. Cana-
dian Journal of Anaesthesia 1988; 35: 479–83.
3 Canessa R, Lerma G, Urzua J. Anesthesia for elective
cardioversion: a comparison of four anesthetic agents. Journal
of Cardiothoracic and Vascular Anesthesia 1991; 5: 566–8.
4 Hullander RM, Leivers D, Wingler K. A comparison of
propofol and etomidate for cardioversion. Anesthesia and
Analgesia 1993; 77: 690–4.
5 Smith I, Thwaites AJ. Target-controlled propofol vs sevo-
flurane: a double-blind, randomised comparison in day case
anaesthesia. Anaesthesia 1999; 54: 745–52.
Figure 3 Incidence of complications.PONV ¼ postoperative nausea andvomiting.
Table 4 Results of patient questionnaire. Values are number (%)of patients.
Propofol group(n = 31)
Sevoflurane group(n = 30)
Mask – unpleasant 2 (8) 2 (9)Smell – unpleasant 0 4 (18)Pain on injection 4 (16) 0
Opinion of anaestheticPleasant 11 (44) 10 (45)Indifferent 13 (52) 11 (50)Unpleasant 1 (4) 1 (5)
Choice of anaesthetic for similar procedure in the futureSame 16 (64) 15 (68)Different 2 (8) 1 (5)No preference 7 (28) 6 (27)
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1118 � 2002 Blackwell Publishing Ltd
6 Fish WH, Hobbs AJ, Daniels MV. Comparison of sevoflu-
rane and total intravenous anaesthesia for daycase urological
surgery. Anaesthesia 1999; 54: 999–1006.
7 Thwaites A, Edmends S, Smith I. Inhalation induction
with sevoflurane: a double-blind comparison with
propofol. British Journal of Anaesthesia 1997; 78:
356–61.
8 Dashfield AK, Birt DJ, Thurlow J, Kestin IG, Langton JA.
Recovery characteristics using single-breath 8% sevoflurane
or propofol for induction of anaesthesia in day-case
arthroscopy patients. Anaesthesia 1998; 53: 1062–6.
9 Struys M, Versichen L, Rolly G. Influence of pre-
anaesthetic medication on target propofol concentration
using a �Diprifusor� TCI system during ambulatory surgery.
Anaesthesia 1998; 53 (Suppl. 1): 68–71.
10 Gupta A, Lennmarken C, Vegfors M, Tyden H. Anaesthesia
for cardioversion. Anaesthesia 1990; 45: 872–5.
11 Johnston RR, Eger EI, Wilson C. A comparative interaction
of epinephrine with enflurane, isoflurane and halothane in
man. Anesthesia and Analgesia 1976; 55: 709–12.
12 Kroll DA, Knight PR. Antifibrillatory effects of volatile
anesthetics in acute occlusion ⁄ reperfusion arrhythmias.
Anesthesiology 1984; 61: 657–61.
13 Bosnjak ZJ, Kampine JP. Effects of halothane, enflurane and
isoflurane on the SA node. Anesthesiology 1983; 58: 314–21.
14 Ebert TJ, Harkin CP, Muzi M. Cardiovascular responses to
sevoflurane: a review. Anesthesia and Analgesia 1995; 81:
S11–S22.
15 Schmidt C, Roosens C, Struys M et al. Contractility in
humans after coronary artery surgery. Anesthesiology 1999;
91: 58–70.
16 Peacock JE, Lewis RP, Reilly CS, Nimmo WS. Effect of
different rates of infusion of propofol for induction of
anaesthesia in elderly patients. British Journal of Anaesthesia
1990; 65: 346–52.
17 Kirkbride DA, Parker JL, Williams GD, Buggy DJ. Induc-
tion of anesthesia in the elderly ambulatory patient: a
comparison of propofol and sevoflurane. Anesthesia and
Analgesia 2001; 93: 1185–7.
FORUM
The value of risk scores for predicting postoperative
nausea and vomiting when used to compare patient
groups in a randomised controlled trial
R. Thomas1, N. A. Jones1 and P. Strike2
1 Specialist Registrar in Anaesthetics, Department of Anaesthetics, North Hampshire Hospital, Aldermaston Road,
Basingstoke, RG24 9NA, UK
2 Medical Statistician, Research and Development Support Unit, Salisbury District Hospital, Salisbury, UK
Summary
Whilst conducting a randomised controlled trial into the effects of combination anti-emetics, we
endeavoured to confirm that our patient groups were matched using the predictive scoring systems
for postoperative nausea and vomiting (PONV) and postoperative vomiting (POV) reported in the
literature. One hundred and seventy-seven female patients attending for day case gynaecological
surgery were studied and their individual risks of PONV and POV were calculated using four
predictive models for PONV and two predictive models for POV. The scoring systems were then
evaluated to see if agreement existed between them using the method described by Bland and
Altman. Bias and 95% limits of agreement were calculated for each combination. Agreement
between scoring systems was poor. As the scoring systems gave widely divergent predictions, we
concluded that the predictive risk for PONV or POV would be dependent upon the scoring
system chosen, thus limiting their usefulness in this role.
Keywords Postoperative period: nausea and vomiting. Scoring systems.
........................................................................................................
Correspondence to: R. Thomas
Accepted: 10 July 2002
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1119
Over the last 10–15 years there has been little progress in
reducing the incidence of postoperative nausea and
vomiting (PONV) in spite of research in the area and
the introduction of new classes of anti-emetics [1]. The
idea that PONV has a multifactorial aetiology has led to
an increased interest in the use of combination anti-
emesis. Research in this field is burgeoning in the hope of
elucidating the most efficacious anti-emetic drugs and
dose combinations [2].
The use of combination anti-emesis has potential
implications for cost of treatment, both in financial terms
and in terms of clinical risk)benefit ratios. Local policy
and pharmacy committee decisions are based on such cost
analyses, and recommendations exist for the use of both
monotherapy and combination anti-emesis in those
patients considered at high risk of PONV [3, 4]. It is
therefore of great advantage to the clinician to be able to
identify those patients most at risk of PONV in order to
make clinical judgements about the use of anti-emetic
therapy.
The prediction of PONV also has important implica-
tions for research. Numerous patient, surgical and anaes-
thetic technique-related factors have been proposed in the
aetiology of PONV [5, 6] and controlling these factors
makes clinical trials difficult to conduct. Apfel et al. [7]
recommend the use of validated PONV risk scores for
group comparisons in randomised controlled trials of anti-
emetic strategies.
Several scoring systems exist to predict the risk of PONV
or postoperative vomiting (POV). These are developed
using logistic regression modelling (a model used for
relating binary outcome variables to multiple independent
explanatory variables) and all have been validated for their
discriminatory power by calculating the area under the
receiver operating characteristics (ROC) curve for each
model [8–12]. The scoring systems described by Apfel et al.
[7], Koivuranta et al. [8] and Palazzo & Evans [9] have been
subject to external evaluation and comparison and each is
reported as having moderate but acceptable accuracy as
measured by the area under the curve (AUC) [13].
Because the various scoring systems have been validated
for their discriminating power and are based on similar
risk factors (Tables 1 and 2), the researcher or clinician
should be able to use any number of the scoring systems
and expect reasonable agreement among the scores.
Whilst conducting a randomised controlled trial into
the effects of combination anti-emesis [14], we endeav-
oured to confirm that our patient groups were matched,
as recommended by Apfel et al. [7], by using predictive
scoring systems. We chose to use two scoring systems [10,
11] to ensure the validity of the result and as a quality
control measure. We were surprised that the two scoring
systems chosen did not agree. We conducted a literature
search and identified a total of four models for the
prediction of PONV and two for the prediction of POV.
The aim of this study was to investigate whether
agreement existed between scoring systems cited in the
literature and therefore whether scoring systems in their
current form are a reliable utility for the comparison of
study groups in randomised controlled trials.
Methods
Having obtained local ethics research committee approval
and written informed consent we investigated 177 ASA I
or II female patients, aged 19–53 years, attending for day
case gynaecological surgery.
The patients were questioned individually, by one of
the two investigators, prior to surgery using a series of
closed questions such that the answers solely represented
the subjective experience of the patient and were not
open to interpretation by the interviewer.
Patients were asked whether they smoked, whether
they had had an episode of nausea or vomiting after a
previous anaesthetic, and whether they suffered from
travel or motion sickness.
Table 1 Risk factors for postoperative nausea and vomiting (PONV) for each scoring system.
Risk factor Sinclair et al. [10] Palazzo & Evans [9] Koivuranta et al. [8] Apfel et al. [11]
Gender * * * *Previous PONV * * *Travel sickness (TS) * *PONV or TS *Non smoker ⁄ smoker * * *Post-operative opioids * *Female and PONV *Duration * *Age *General anaesthesia *Type of surgery *
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1120 � 2002 Blackwell Publishing Ltd
The age of the patient and duration of surgery were
recorded for each patient.
All patients were prescribed postoperative opioid
analgesia, either in the form of paracetamol 325 mg
and dextropropoxyphene 32.5 mg, or paracetamol
500 mg and codeine phosphate 30 mg. The patients
were then interviewed by telephone 24 h postoperatively
to ascertain whether they had taken the prescribed
medication.
The data were analysed using a standard spreadsheet
(EXCEL 2000, Microsoft USA). From the collated data we
calculated the predictive risk of postoperative nausea and
vomiting for each patient using each one of four scoring
systems: Koivuranta [8], Palazzo [9], Sinclair [10] and
Apfel (Simplified Risk Score) [11]. We also scored the
patients individually for their risk of postoperative
vomiting using two scoring systems: Koivuranta [8] and
Apfel [12]. (See Appendix for details.)
The scoring systems were evaluated to see if agreement
existed between them using the method for testing the
agreement of measurement methods as described by
Bland & Altman [15].
Scatter diagrams including the line of equality and
Bland)Altman plots were constructed to assess the level
of agreement for each combination of scoring systems.
The bias (mean of the difference between scores) and the
95% limits of agreement [mean difference (1.96 SD)]
were calculated for each combination of scoring system
using EXCEL 2000 (Microsoft) and MINITAB v12 (Minitab
Inc., State College, PA, USA).
Results
Of the 177 patients studied, 68 (38.4%) had experienced
previous PONV associated with general anaesthesia, 58
(32.8%) suffered with travel sickness, 113 (63.8%) were
non-smokers and 141 (79.7%) received postoperative
opioids. All had gynaecological surgery that did not
involve dilatation and curettage. The mean (SD) age was
35.7 (7.3) years.
Bland–Altman plots (Fig. 1) show the difference in
predicted risk for each individual patient plotted against
the average risk for each patient using each pair of
scoring systems. Summary statistics from the Bland–
Altman plots (Table 3) show the average difference
between scoring systems (bias), together with the stand-
ard deviation of the difference (precision) and the 95%
limits of agreement.
The best agreement existed when the scores of Sinclair
were compared with those of Palazzo & Evans. Other
pairs of scoring systems showed less agreement.
The results of the anti-emetic study are reported
elsewhere [14].
Discussion
Agreement existed when the scores of Sinclair and
Palazzo were compared (bias )2.5%, precision 5.6%).
However, both scoring systems greatly underestimated
our observed prevalence of PONV in spite of the fact that
all our patients received anti-emetics. Previous studies
have demonstrated an offset in the calibration curves
(correlation between predicted and actual incidences of
PONV) for both Sinclair and Palazzo & Evans of 30%
such that both scores underestimate the true incidence [7,
16, 17]. Our results corroborate these findings [14]. Apfel
et al. suggested that the difference might be explained by
the high odds ratio of 53.0 for a positive history of PONV
in the scoring system proposed by Palazzo & Evans and
that the odds ratios ascribed by Sinclair et al. to different
surgical specialties may differ significantly between centres
[7].
Evaluation of the risk factors for the prediction of
PONV and POV is complex owing to the multifactorial
aetiology of the condition [5, 6]. Nausea is also a highly
subjective experience making it difficult to evaluate as an
endpoint [12]. The scoring systems that we analysed are
designed to give a quantitative risk of PONV or POV,
using complex and precise equations. However, the
data used for the development of many of these
scoring systems rely on subjective qualitative information
obtained from the patient. How the patient actually
reports is therefore dependent upon what an individual
considers significant. This has profound implications for
the reliability of the derived risk (by reliability we mean
the extent to which the instrument delivers the same
results in different hands or at different locations).
Sinclair et al. [10], Koivuranta et al. [8] and Apfel et al.
[11] all identified non-smoking as an independent risk
factor for PONV. However, the extent to which a patient
smokes may have a bearing on the way they report,
Table 2 Risk factors for postoperative vomiting (POV) for eachscoring system.
Risk factor Koivuranta et al. [8] Apfel et al. [12]
Gender * *Previous PONV *Travel Sickness (TS) *PONV or TS *Non-smoker ⁄ smoker * *Post-operative opioidsFemale and PONVDuration * *Age *General anaesthesiaType of surgery
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1121
Figure 1b Sinclair v Koivuranta for postoperative nausea and vomiting.
Figure 1a Sinclair v Palazzo for postoperative nausea and vomiting.
Figure 1a–g Bland–Altman plots showing the difference in predicated risk for each individual patient plotted against the average riskfor each patient, using each pair of scoring systems. The bias and 95% limits of agreement are also shown.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1122 � 2002 Blackwell Publishing Ltd
especially to health care personnel since occasional smokers
may report as non-smokers if they consider the amount
which they smoke to be insignificant. Indeed, even regular
smokers may report as non-smokers either so as not to be
stigmatised or because they have been issued health or life
insurance on non-smoking risk terms.
Figure 1d Palazzo v Koivuranta for postoperative nausea and vomiting. Data points represent identical results from many subjects.
Figure 1c Sinclair v Apfel for postoperative nausea and vomiting.
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1123
Similarly, Palazzo & Evans [9], Koivuranta et al. [8] and
Apfel et al. [11] found a previous history of motion
sickness to be a predictor for PONV. This, too, relies on
subjective interpretation by the patient, and is neither
precise nor reliable. On applying their score to a different
population, Palazzo & Evans found that travel sickness no
Figure 1e Palazzo v Apfel for postoperative nausea and vomiting. Data points represent identical results from many subjects.
Figure 1f Koivuranta v Apfel postoperative nausea and vomiting. Data points represent identical results from many subjects.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1124 � 2002 Blackwell Publishing Ltd
longer became important and that only female sex,
previous PONV and postoperative opioids were sig-
nificant predictors of PONV [16].
Sinclair et al. [10], Palazzo & Evans [9], Apfel et al. [11]
and Koivuranta et al. [8] all identified a previous history of
PONV as being a major contributor to the likelihood of
future PONV. Patients who have not previously had an
operation cannot have a history of PONV. The propor-
tion of patients in each of their studies who have never
previously had an operation will therefore influence the
weighting of each score, as they will be considered as
being at low risk, when in fact their true risk of PONV is
not known.
We also question the utility of a predictive test that is
dependent upon prospective data, such as duration of
operation and the use of postoperative opioids. Apfel et al.
[11] defend this, suggesting that, with many patients, it is
possible to predict whether opioids will be required, and
in most circumstances the duration of surgery can also be
predicted.
Such subjective data limit the usefulness of scoring
systems designed to give accurate quantitative informa-
tion, as it is hard to extrapolate binary outcomes from
continuous data. This may help to explain why the scoring
systems only had moderate accuracy when evaluated by
the area under the receiver operator characteristic curves.
Indeed Apfel et al. [11] and Koivuranta et al. [8] both
simplified their scores, making the weighting for each risk
factor unity, and found that this did not significantly
reduce the discriminatory power of the score.
Table 3 Results of Bland)Altman plots for each pair of scoring systems.
Comparativetests
Mean differencein risk (bias)
SD of differencein risk (precision)
95% Limits ofagreement (1.96SD)
PONVSinclair v Palazzo ) 2.5 5.6 ) 13.8 to 8.4Sinclair v Koivuranta ) 35.1 12.5 ) 59.7 to ) 10.5Sinclair v Apfel ) 50.3 13.0 ) 75.8 to ) 24.8Palazzo v Koivuranta ) 32.6 11.3 ) 54.7 to )10.5Palazzo v Apfel ) 47.8 13.5 ) 74.3 to ) 21.3Koivuranta v Apfel ) 15.2 10.5 ) 35.8 to 5.4
POVApfel v Koivuranta POV 19.8 7.4 5.2 to 34.4
Figure 1g Apfel v Koivuranta postoperative vomiting.
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1125
It is important to remember that failure of agreement
does not imply that all the systems are poor at predicting
PONV or POV. Our type of analysis could be criticised
as being limited by the fact that all our patients received
anti-emetics on ethical grounds. Hence, we are unable to
specify which of the reported scoring systems most
closely reflected the truth. This, however, should be
incidental when such risk scores are used in demographic
tables for group comparisons in randomised control trials.
For this purpose, an investigator should be able to use
any one of the risk scores at random and be confident
that the prediction obtained will be similar to any of the
others.
The true test of these scoring systems is in the clinical
or research field when applied to the role for which they
have been developed. We were perplexed that different
nausea and vomiting scoring systems gave very different
predictions of risk. As a result, we decided not to use
scoring systems in our demographic table for the com-
parison of our study groups, as recommended by Apfel
et al. [7], as we believed the accuracy of such a test could
not be guaranteed. It is of interest that, in a subsequent
publication, Apfel et al. [18] have used their simplified
predictive risk score to identify those patients who
meet the inclusion criteria for their study but have
not used them in their demographic tables for group
comparisons.
Although outside our study remit, we question the
application of these scoring systems in order to make
clinical risk)benefit decisions. We are unaware of any
randomised clinical trial comparing the incidence of
PONV in a population prescribed prophylactic anti-
emetics on the basis of identified risk using the above
scoring systems with a population prescribed anti-emetic
prophylaxis using any other criteria or at random. Because
the numbers needed to treat, with the currently available
prophylactic anti-emetics, remains at between five and six
[19] and because the best predictive models give an
accuracy AUC of 0.7, we feel that it is unlikely that a
clinically significant difference would exist.
Apfel accepts that AUCs of only 0.7 are not ideal.
However, from work carried out on a virtual model [20],
he believes that the current scoring systems cannot be
improved by the addition or subtraction of further
independent explanatory variables and therefore the
current scoring systems should act as an acceptable model.
Whilst the principle of a model for the prediction of
PONV remains an attractive concept we do not believe
that the above systems currently fulfil the role for which
they have been designed.
We conclude that it would not be possible to use any
one of these risk scores in isolation for demographic
purposes in research, as the predictive risk obtained
would be dependent upon the scoring system chosen.
We accept, however, that such a demographic table
could be constructed using all available risk scores and
allowing the reader to compare all scores (purely as a
similarity of the numbers obtained) between groups
without necessarily assigning predictive risk. In addition,
the �number� that a PONV risk score generates will
invariably lead readers to make judgements about the
effectiveness (or otherwise) of the investigated anti-
emetic strategy that may not necessarily be relevant to the
method analysis. We believe this would produce a
complicated and possibly misleading demographic analy-
sis that may well detract from the main thesis of the anti-
emetic strategy under investigation. Indeed, this is what
we discovered to be the case in our parallel study and
ultimately we decided to restrict our demographic table
to raw data.
References
1 Scholz J, Steinfath M, Tonner PH. Postoperative nausea
and vomiting. Current Opinion in Anesthesiology 1999; 12:
657–61.
2 Hefferman AM, Rowebottom DJ. Postoperative nausea and
vomiting ) time for balanced antiemesis? British Journal of
Anaesthesia 2000; 85: 675–7.
3 Mehernoor F, Watcha MD. The cost effective management
of postoperative nausea and vomiting. Anesthesiology 2000;
92: 931–3.
4 Scuderi PE, James RL, Harris L, et al. Multimodal antiemetic
management prevents early postoperative vomiting after
outpatient laparoscopy. Anesthesia and Analgesia 2000; 91:
1408–14.
5 Palazzo MGA, Strunin L. Anaesthesia and emesis I. Etiology.
Canadian Society of Anaesthetists Journal 1984; 31: 178–87.
6 Watcha M, White P. Postoperative nausea and vomiting. Its
etiology, treatment and prevention. Anesthesiology 1992; 77:
162–84.
7 Apfel CC, Kranke P, Eberhart LHJ, Roos A, Roewer N.
Comparison of predictive models for postoperative nausea
and vomiting. British Journal of Anaesthesia 2002; 88: 234–40.
8 Koivuranta M, Laara L, Alahuhta S. A survey of postoper-
ative nausea and vomiting. Anaesthesia 1997; 52: 443–9.
9 Palazzo M, Evans R. Logistic regression analysis of fixed
patient factors for postoperative sickness: a model for
risk assessment. British Journal of Anaesthesia 1993; 70:
135–40.
10 Sinclair DR, Chung F, Mezei G. Can postoperative nausea
and vomiting be predicted? Anesthesiology 1999; 91: 109–18.
11 Apfel CC, Laara E, Koivuranta M, Greim CA, Roewer N.
A simplified risk score for predicting postoperative nausea
and vomiting. Anesthesiology 1999; 91: 693–700.
12 Apfel CC, Greim CA, Haubitz I, et al. A risk score to predict
the probability of postoperative vomiting in adults. Acta
Anaesthesiologica Scandinavica 1998; 42: 495–501.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1126 � 2002 Blackwell Publishing Ltd
13 Eberhart LHJ, Hogel J, Seeling W, Staack AM, Geldner G,
Georgieff M. Evaluation of three risk scores to predict
postoperative nausea and vomiting. Acta Anaesthesiologica
Scandinavica 2000; 44: 480–8.
14 Thomas R, Jones N. Prospective randomised double-blind
comparative study of dexamethasone, ondansetron and
ondansetron plus dexamethasone as prophylactic anti-emetic
therapy in patients undergoing day-case gynaecological
surgery. British Journal of Anaesthesia 2001; 87: 588–92.
15 Bland JM, Altman DG. Statistical methods for assessing
agreement between two methods of clinical measurement.
Lancet 1986; 8 (8476): 307–10.
16 Toner CC, Broomhead CJ, Littlejohn IH, et al. Prediction of
postoperative nausea and vomiting using a logistic regression
model. British Journal of Anaesthesia 1996; 76: 347–51.
17 Pierre S, Bernais H, Pouymayou J. A comparison of two risk
scores for predicting postoperative nausea and vomiting.
European Journal of Anaesthesiology 2001; 18: 8–28.
18 Apfel CC, Kranke P, Katz MH, et al. Volatile anaesthetics
may be the main cause of early but not delayed postoperative
vomiting: a randomised controlled trail of factorial design.
British Journal of Anaesthesia 2002; 88: 659–68.
19 Tramer MR, Reynolds DJ, Moore RA, McQuay HJ.
Efficacy, dose)response, and safety of ondansetron in pre-
vention of postoperative nausea and vomiting. A quantita-
tive systematic review of randomized placebo-controlled
trials. Anesthesiology 1997; 87: 1277–89.
20 Apfel CC, Kranke P, Greim C-A, Roewer N. What can
be expected from risk scores for predicting postoperative
nausea and vomiting? British Journal of Anaesthesia 2001; 86:
822–7.
Appendix
The scoring systems proposed by Palazzo & Evans,
Koivuranta, Apfel, Sinclair and their colleagues were
produced by logistic regression analysis (a model for
relating independent explanatory variables to a binary
outcome variable) using the following equation:
Loge ðp=1 � pÞ ¼ z
where p ¼ proportion of patients with sickness (PONV
or POV) and (1 ) p) ¼ proportion of patients without
sickness (PONV or POV) and
z ¼ b0 þ b1X1 þ b2v2 þ bnXn
where b represents the parameter estimates for the
variable (i.e. the weighting) and Xn ¼ the value (presence
or absence) of the nth independent variable.
The z-value for each of the predictive tests is given by
the equations that are listed below.
The probability of an event (postoperative nausea and
vomiting or postoperative vomiting) is related to the odds
ratio such that:
Probability ¼ Odds ratio=ð1 þ Odds ratioÞ
The odds ratio ¼ (p ⁄ 1 ) p), and as Loge(p ⁄ 1 ) p) ¼ z
and it can be seen that the odds ratio ¼ ez.
Therefore the probability of PONV or POV is
estimated from the equation
p ¼ ez=ð1 þ ezÞ
which simplified equals 1 ⁄ (1 + e–z).
Sinclair score for PONV [10]
Z ¼� 5:97 þ ð�0:14 � age=10Þþ ð�1:03 � SexÞþ ð�0:42 � SmokeÞþ ð1:14 � PONV historyÞþ ð0:46 � DurationÞþ ð2:36 � GAÞþ ð1:20 � GynNon DCÞ:
(Sex ¼ 0 for female, smoke ¼ 1 if smoker, PONV history
¼ 1 if previous history of PONV, duration ¼ duration of
surgery in 30-min increments, general anaesthetic ¼ 1,
GynNon DC ¼ 1 if gynaecological surgery and non-
dilatation and curettage procedure).
Palazzo & Evans score for PONV [9]
Z¼�5:03þð2:24�postoperative opioidsÞþð3:97�previous sickness historyÞþð2:4�genderÞþð0:78�history of motion sicknessÞ�ð3:2�gender�previous sickness historyÞ:
[Postoperative opioids ¼ 1, previous sickness history
(PONV) ¼ 1, gender ¼ 1 if female, history of motion
sickness ¼ 1].
Apfel et al. score for POV [12]
Z ¼ ð1:28 � female genderÞ � ð0:029 � ageÞ� ð0:74 � smokingÞþ ð0:63 � history of motion sickness or PONVÞþ ð0:26 � durationÞ � 0:92
(Female ¼ 1, age ¼ in years, smoking ¼ 1 if smoker,
history of motion sickness and ⁄ or PONV ¼ 1, duration
of anaesthesia ¼ hours).
Simplified Apfel Score for PONV [11]
Using a logistic regression model, Apfel et al. identified
four risk factors for the prediction of PONV. These were
female gender, prior history of PONV or motion
sickness, non-smoking and the use of postoperative
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1127
opioids. These had equal weighting. The probability of
PONV was 10, 21%, 39, 61 and 78 if no, one, two, three
or four risk factors were present.
Simplified Koivuranta scores for PONV
and POV [8]
Based on logistic regression analysis Koivuranta identified
the five strongest predictors for PONV and POV, as
female sex, previous PONV, duration of operation
> 60 min, history of motion sickness and non-smoking.
The score was simplified by giving each predictor equal
weight. The risk for PONV was related to the number of
factors present such that the probability of nausea was 17%
with no factors present, increasing to 18, 42, 54, 74 and
87 for one to five factors present. Correspondingly the
risk for vomiting was 7, 7, 17, 25, 38 and 61 for no factors
to all five present.
The probability of PONV and POV was calculated for
each patient using the above equations such that for each
patient we generated four predictive values for PONV
and two for POV.
FORUM
Measurement of liver tissue oxygenation after orthotopic
liver transplantation using a multiparameter sensor
A pilot study
T. S. Leary,1 J. R. Klinck,2 G. Hayman,3 P. Friend,4 N. V. Jamieson5 and A. K. Gupta2
1 Consultant, Intensive Care, Norfolk and Norwich University Hospital, UK
2 Consultant, Department of Anaesthesia, 3 Operating Department Practitioner, and 5 Consultant, Department
of Surgery, Addenbrooke’s Hospital, Cambridge CB2 2QQ, UK
4 Professor of Transplant Surgery and Oxford Transplant Centre, Churchill Hospital, Oxford, UK
Summary
The currently used methods of monitoring liver perfusion and oxygenation after liver transplan-
tation have major limitations in clinical use. We describe the use of a multiparameter sensor to
enable continuous monitoring of liver tissue oxygen tension, carbon dioxide tension and hydrogen
ion concentration in the early postoperative period in 12 patients after liver transplantation. The
sensor was inserted under direct vision via the falciform ligament into the liver before skin closure.
Tissue oxygen tension values decreased in the first 24 h and subsequently increased to a mean
(SD) ¼ 7.3 (2.8) kPa at 48 h after surgery. This was associated with a decrease in the degree of
acidosis. There were no complications attributable to the sensor. This study demonstrates that
continuous measurement of liver oxygen tension, carbon dioxide tension and pH is possible. This
technique may be useful as a continuous monitor to help identify grafts at risk of ischaemia.
Keywords Liver transplantation. Monitoring: physiologic.
........................................................................................................
Correspondence to: Dr A. K. Gupta
E-mail: [email protected]
Accepted: 14 July 2002
The use of orthotopic liver transplantation as a treatment
for end-stage liver disease and fulminant hepatic failure is
increasing. Whilst 5-year survival rates continue to
improve [1], catastrophic primary non-function as a result
of ischaemia still has a reported incidence of 6% [2].
Injury during organ preservation and transfer, or pro-
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1128 � 2002 Blackwell Publishing Ltd
longed systemic hypotension after transplantation, may be
aetiological factors in hepatic infarction. However, such
injury is usually caused by hepatic artery or portal vein
thrombosis (incidences 3–12% and 1–7%, respectively
[3]). Continuous and reliable monitoring of graft oxy-
genation might provide early detection of ischaemia and
allow timely surgical re-exploration and correction of
vascular insufficiencies. Without a functioning graft,
patient mortality approaches 100% [1].
Methods of monitoring liver organ perfusion during
the early postoperative period include Doppler ultr-
asonography [3], laser flowmetry [4] and near-infrared
spectroscopy [5]. These methods all have significant
limitations that decrease their usefulness as clinical
monitors of perfusion and oxygenation. In our unit,
we currently rely on biochemical markers or meas-
urement of prothrombin time to assess hepatic func-
tion, from which adequacy of perfusion may be
inferred.
The ‘Paratrend 7’ (Diametrics Medical, High
Wycombe, UK) is a multiparameter sensor that was
originally designed for the continuous intra-arterial mon-
itoring of PaO2, PaCO2, hydrogen ion concentration and
temperature. This sensor has been used as a monitor of
brain tissue oxygenation in animals [6] and humans [7].
We aimed to evaluate a Paratrend 7 sensor placed within
the liver tissue as a continuous monitor of allograft
oxygenation and metabolism during the first 48 h after
surgery.
Methods
After Local Research Ethics Committee approval and
written, informed consent, 12 adult patients admitted for
liver transplantation were studied. Surgery comprised a
conventional orthotopic approach with caval anasto-
moses. A standardised anaesthetic technique including
fentanyl, atracurium and isoflurane in oxygen and air was
used. A low-dose dopamine infusion was given through-
out the peri-operative period.
Two Paratrend 7 sensors were calibrated against
standard concentrations of gases before insertion. One
sensor was inserted into the patient’s brachial artery for
continuous blood gas monitoring before surgery. After
revascularisation of the donor liver in the abdomen of the
recipient, the second sensor was passed aseptically to the
surgeon who inserted it into liver tissue via the falciform
ligament. The proximal end of the sensor was brought
out from the abdominal wall about 5 cm from the main
incision. Readings were taken for tissue and arterial PO2,
PCO2 and hydrogen ion concentration, and were recor-
ded initially at 30-min intervals for the first 2 h and then
four hourly.
Both sensors were removed from the patient approxi-
mately 48 h after surgery. A simple dressing was applied
to the entry site of the liver probe. Pressure was applied to
the antecubital fossa. A note of any complications
attributable to the catheter or any observed impairment
in graft function was made.
The data obtained from the liver probe were analysed
using repeated measures analysis of variance. Significance
was set at p < 0.05.
Results
Of the 12 patients recruited, complete data were collected
from five patients and partial data from a further four.
Data could not be collected when irreparable damage to
the fibres within the sensor occurred due to kinking
during insertion, at the time of skin closure, or whilst the
patient was nursed on the ward. The oxygen electrode,
which is a Clarke electrode, appeared to be far more
Figure 1 Liver tissue oxygen tension(PlO2) in nine subjects measured withthe Paratrend probe. Different symbolsand lines indicate values in individualpatients. Times are given in hours afterinsertion of the probe.
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1129
robust than the fibreoptic hydrogen ion and carbon
dioxide sensors.
The hepatic tissue measurements of oxygen tension,
carbon dioxide tension and hydrogen ion concentration
(PlO2, PlCO2 and [Hl+]) are shown in Figs 1–3. The values
of PlO2, PlCO2 and [Hl+] taken at 6-h intervals are
presented in Table 1 along with arterial values for
comparison. The mean (SD) PlO2 decreased in the first
24 h from 6.1 (5.8) to 4.7 (2.0) kPa, and then increased to
7.3 (2.8) kPa (p ¼ 0.97). These changes were associated
with a decrease in tissue acidosis, as the mean (SD)
hydrogen ion concentration decreased from 66.2 (23.7) to
46.5 (3.2) nmol.l)1 (p ¼ 0.54).
Routine measurements of prothrombin time, alanine
transaminase and serum bilirubin concentrations are
documented in Table 2. There were no cases of hepatic
impairment requiring urgent re-exploration in our series.
Patients numbers 3 and 9 had presumed episodes of acute
rejection with increased biochemical markers that subse-
quently returned to normal over the ensuing 5 days.
Figure 3 Liver tissue hydrogen ionconcentration [Hl
+] in nine subjectsmeasured with the Paratrend probe.
Figure 2 Liver tissue carbon dioxidetension (PlCO2) in nine subjects meas-ured with the Paratrend probe.
Table 1 Mean (SD) values of arterial and liver tissue PO2, PCO2 and hydrogen ion concentration as measured by the Paratrend probeafter liver transplantation.
Time afterprobe insertion; h
Arterial[H+]
Liver[H+]
ArterialPO2; kPa
LiverPO2; kPa
ArterialPCO2; kPa
LiverPCO2; kPa
0 65.2 (37.3) 66.2 (23.7) 20.6 (11.2) 6.1 (5.8) 5.5 (1.0) 7.3 (3.0)6 41.8 (6.1) 75.7 (40.4) 15.6 (5.1) 5.5 (5.1) 5.3 (0.7) 7.1 (1.1)
12 40.4 (6.4) 54.0 (11.5) 16.7 (7.1) 5.0 (3.2) 5.3 (0.5) 6.8 (0.9)18 39.5 (5.9) 55.5 (15.0) 15.2 (3.9) 4.7 (2.0) 5.5 (0.5) 6.5 (0.4)24 38.4 (3.4) 58.7 (10.4) 13.2 (2.87) 4.8 (2.2) 5.1 (0.7) 7.0 (0.8)36 38.3 (2.42) 55.1 (11.0) 16.2 (3.6) 6.1 (4.0) 5.8 (0.6) 7.3 (0.4)48 38.5 (2.4) 46.5 (6.8) 12.8 (4.0) 7.3 (2.8) 5.45 (0.3) 6.3 (0.8)
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1130 � 2002 Blackwell Publishing Ltd
Patient number 6 developed acute cellular rejection on
the fourth postoperative day and required re-transplanta-
tion at 10 days due to graft failure secondary to non-
thrombotic infarction. Patient number 8 initially showed
poor hepatic function associated with increasing trans-
aminase levels. By day 4, hepatic function was clearly
recovering. However, on day 5, the patient died of acute
respiratory distress syndrome and sepsis. Post-mortem
examination confirmed the biochemical findings, dem-
onstrating vascular anastomotic patency.
No complications occurred in any of the patients that
could be attributed to the sensors.
Discussion
This study attempted to monitor intrahepatic changes in
graft PO2, PCO2 and hydrogen ion concentration con-
tinuously after liver transplantation. The mean PlO2
appears to decrease initially and subsequently to plateau.
This does not appear to be a function of systemic
oxygenation, which did not change significantly. The
initial mean tissue PlO2 of 6.1 kPa is similar to values
measured when samples are taken from the hepatic vein
after liver graft reperfusion [8]. The initial decrease in
PlO2 may be due to a change in the metabolic function of
the new graft in the first 24 h. Hydrogen ion concen-
tration and PlCO2 do not change significantly, confirming
the adequacy of perfusion and excluding this as a cause for
changes in PlO2.
A continuous method of monitoring the new graft is
required so that deteriorating function can be detected
early, allowing rapid intervention to rescue the new
organ. The methods currently used have significant
limitations. Biochemical markers (serum transaminases,
hypoglycaemia, acidosis and hyperkalaemia), prothrom-
bin time, indocyanine green [9] clearance and mono-
ethylglycinxylid production following lidocaine bolus
[10] have been used to monitor hepatic graft function.
The disadvantage of these methods is that there may be a
long lag time between graft ischaemia and the detection
of malfunction. Furthermore, the cause of malfunction
may not be identified, as the results are often non-specific
[11, 12].
A low gastric mucosal pH (pHi) has been associated
with impaired liver function after transplantation. How-
ever, no predictive effects for patient morbidity or
mortality have been shown using this method [13].
Therefore, pHi monitoring as an indicator of gastric
mucosal perfusion either before [14] or after [11] surgery
should not be regarded as a reliable indicator of poor graft
function.
Although its sensitivity remains unknown, the speci-
ficity of transabdominal colour Doppler ultrasonography
in the diagnosis of hepatic artery and portal vein
thrombosis has been shown to be 100% [3], hence
avoiding time-consuming angiography. However, such
examinations require significant operator skill and are
intermittent. Thus, in most cases, manifestation of graft
dysfunction is seen before ultrasonographic diagnosis is
made. Implantable Doppler probes have been successfully
attached to the relevant vessels [15], allowing continuous
monitoring of blood flow. However, risks of vessel
damage and thrombus initiation exist [16]. Laser Doppler
flowmetry may be used to monitor hepatic microcircu-
lation. However, this method is non-quantitative and its
suitability as a postoperative technique remains untested.
The method of tissue monitoring using the Paratrend 7
is similar to that of near-infrared spectroscopy, in that
both methods are estimates of tissue oxygenation in a
small area of tissue. Near-infrared spectroscopy also
measures Caa3, the terminal enzyme in the respiratory
chain, allowing defects in oxygen supply and utilisation to
be distinguished. The Paratrend 7 monitor continuously
measures tissue PCO2 and pH in addition to PO2. The
advantage of measuring PlCO2 and [Hl+] as well as PlO2 is
that an indication of changes in tissue metabolism is
Table 2 Biochemical markers of graft function after liver transplantation in the nine subjects studied.
Time afterprobe insertion; h
Prothrombin time; s Serum bilirubin; lmol.l)1 Serum alanine transaminase; lmol.l)1
0 24 48 0 24 48 0 24 48
Patient no.1 15.0 18.7 17.8 25 90 103 58 726 7522 23.0 20.4 17.2 103 72 133 632 471 3273 32.8 28.5 41.6 54 81 125 1437 706 29554 17.3 18.6 17.3 104 153 156 694 645 7635 23.7 21.4 17.5 65 39 33 320 318 2526 22.8 23.2 22.7 62 77 95 619 701 7377 25.0 23.7 22.6 129 194 160 311 279 1328 14.6 30.5 30.6 9 84 112 41 2924 43419 17.2 18.6 17.4 70 40 51 401 439 1100
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1131
obtained. An organ failing because of impaired perfusion
would be expected to show a decrease in PO2 and pH,
and an increase in PCO2. Kitai et al. [5] have demonstrated
the application of near-infrared spectroscopy for monit-
oring hepatic allografts. Graft oxygen saturations as low as
20% were recorded, with a successful outcome. These
were well below that measured in a peripheral artery. At
48 h after surgery, we recorded a mean PaO2 of 12.8 kPa
and PlO2 of 7.3 kPa. However, Kitai also describes a
significant wound infection rate. Although near-infrared
spectroscopy is non-invasive, it does require a large skin
port for the probe, whereas the Paratrend 7 sensor is only
0.5 mm in diameter and can be passed through the skin
through a 20G intravenous cannula.
The ability of the Paratrend 7 probe to monitor tissue
oxygen tensions reliably has recently been validated in
brain tissue using positron emission tomography [17]. We
did not address the issue of catheter calibration drift in our
study. Valedka has recently shown that when used as a
monitor of brain tissue oxygen for at least 40 h, the
Paratrend 7 sensor can be expected to return mean (SD)
values of 0.9 (0.2) kPa in zero oxygen calibration solution
and of 18.4 (2.7) kPa in room air [18]. We felt that this
degree of accuracy was sufficient to identify trends of
deterioration in organ oxygenation.
Although the fibreoptic technology incorporated into
the Paratrend 7 sensor has led to a significant advance in
the ability to measure tissue gas tension continuously,
further modifications are required before this system can
be used as a reliable monitor. Data were lost in four
patients due to the fragility of the fibreoptic CO2 and
pH sensors within the system. These parameters are
particularly useful in detecting deteriorating tissue per-
fusion. A simple modification could stiffen the sensor
system to preserve all the variables so as to ensure
continuous monitoring over a prolonged period of time,
either during surgery or afterwards in the intensive care
unit.
In this study, we have established that the Paratrend 7
multiparameter sensor can be used as a method of
continuously monitoring tissue oxygenation and meta-
bolism in the post-transplantation organ. Furthermore,
baseline values of PlO2, PlCO2 and [Hl+] have been
described in nine patients. Further studies are required
to confirm these values and to determine threshold
levels below which primary non-function can be
predicted.
References
1 Hickman PE, Potter JM, Pesce AJ. Clinical chemistry and
post-liver–transplant monitoring. Clinical Chemistry 1997;
43: 1546–54.
2 Ploeg RJ, D’Alessandro AM, Knechtle SJ, et al. Risk factors
for primary dysfunction after liver transplantation – a mul-
tivariate analysis. Transplantation 1993; 55: 807–13.
3 Hellinger A, Roll C, Stracke A, Erhard J, Eigler FW. Impact
of colour Doppler sonography on the hepatic artery and the
portal vein after liver transplanation. Langenbecks Archiv Fur
Chirurgie 1996; 381: 182–3.
4 Steifalian AM, Mallet SV, Rolles K, Davidson BR. Hepatic
microcirculation during human orthotopic liver transplan-
tation. British Journal of Surgery 1997; 84: 1391–5.
5 Kitai T, Shinohara H, et al. Postoperative monitoring of the
oxygenation state of the graft liver in cases with hepato-
pulmonary syndrome. Transplantation 1996; 62: 1676–8.
6 Zauner A, Bullock MR, Di X, Young HF. Brain oxygen,
CO2, pH and temperature monitoring: evaluation in the
feline brain. Neurosurgery 1995; 37: 1168–77.
7 Gupta AK, Hutchinson PJ, et al. Measuring brain tissue
oxygenation compared with jugular venous oxygen satura-
tion for monitoring cerebral oxygenation after traumatic
brain injury. Anesthesia and Analgesia 1999; 88: 549–53.
8 Tallgren M, Makisalo H, Hockerstedt K, Lindgren L.
Hepatic and splanchnic oxygenation during liver transplan-
tation. Critical Care Medicine 1999; 27: 2383–8.
9 Krenn CG, Schafer B, Berlakovich GA, Steininger R,
Steltzer H, Spiss CK. Detection of graft nonfunction after
liver transplantation by assessment of indocyanine green
kinetics. Anesthesia and Analgesia 1998; 87: 34–6.
10 Schutz E, Luy-Kaltefleiter M, Kaltefleiter M, Burdeslski M,
Ringe B, Armstrong VW, Oellerich M. The value of serial
determination of MEGX and hyaluronic acid early after
orthotopic liver transplantation. European Journal of Clinical
Investigation 1996; 26: 907–16.
11 Muraca M, Kohlhaw K, Vilei MT. Serum bile acids and
esterified biirubin in early detection and differential diagnosis
of hepatic dysfunction following orthotpic liver transplan-
tation. Journal of Hepatology 1993; 17: 141–5.
12 Sankary HN, Williams JW, Foster PF. Can serum liver
function tests differentiate rejection from other causes of
liver dysfunction after hepatic transplantation? Transplantation
Proceedings 1988; 20: 669–70.
13 Maring JK, Klompmaker IJ, Zwaveling JH, Verwer R,
Slooff MJ. Gastric mucosal pH is associated with initial graft
function but is not a predictor of major morbidity after liver
transplantation. Liver Transplant Surgery 1997; 3: 611–6.
14 Welte M, Pichler B, Groh J, Anthuber M, Jauch KW,
Pratschke E, Lenhart FP, Haller M, Frey L, Peter K. Peri-
operative mucosal pH and splanchnic endotoxin concen-
tration in orthotopic liver transplantation. British Journal of
Anaesthesia 1996; 77: 560–1.
15 Yokoyama I, Tabuchi Y, Negita M et al. Measurement of
portal venous flow velocity with an implantable miniature
Doppler probe in pig liver transplantation. Transplantation
International 1997; 10: 116–20.
16 Zurbrugg HRP, Bachmann S, Liebold A, Behr R, Philipp
A, Birnbaum DE, Neuhaus P. Continuous blood flow
measurement after liver transplantation: first clinical experi-
ence. Transplantation Proceedings 1994: 2218–26.
Forum Anaesthesia, 2002, 57, pages 1102–1133......................................................................................................................................................................................................................
1132 � 2002 Blackwell Publishing Ltd
17 Gupta AK, Hutchinson PJ, Fryer T. Measurement of brain
tissue oxygenation performed using Positron Emission To-
mography scanning to validate a novel monitoring method.
Journal of Neurosurgery 2002; 96: 263–8.
18 Valedka AB, Gopinath SP, Constant CF, Uzura M,
Robertson CS. Relationship of brain tissue PO2 to
outcome after severe head injury. Critical Care Medicine
1998; 26: 1576–81.
Anaesthesia, 2002, 57, pages 1102–1133 Forum......................................................................................................................................................................................................................
� 2002 Blackwell Publishing Ltd 1133