A retrospective study involving 603 total or partial hip and total
knee arthroplasties, performed between 1994 and 1998 in 529 patients,
was carried out. The study was approved by our Institutional Review
Board, and all patients provided informed consent. Patients were
excluded from the study if they were undergoing surgery for a hip
fracture; were undergoing revision surgery; had a hematological
disease; were being treated with chronic dicumarol therapy for anticoagulation;
had a chronic liver impairment; had had any type of hemorrhagic
complication, such as gastrointestinal hemorrhage, during the postoperative
period; or received a transfusion even though the pretransfusion
hemoglobin level was >85 g/L.
A total of 370 procedures (296 patients) met the study criteria and
were included in the study group. There were 203 women (68.6%)
and ninety-three men (31.4%). The mean age was 66.9 years
(range, thirty-five to eighty-nine years), the mean weight was 74.5
kg (range, 53 to 105 kg), and the mean height was 156 cm (range,
136 to 174 cm). There were 209 total hip arthroplasties and 161
total knee arthroplasties. The preoperative diagnosis was degenerative
arthritis in 337 cases and rheumatoid arthritis in thirty-three.
The average duration of the surgical procedures was 90.2 minutes
(range, fifty to 180 minutes). The mean preoperative hemoglobin
level was 144 g/L (range, 101 to 185 g/L), and
the mean preoperative hematocrit was 43% (range, 30.4% to
53.2%). Although sixty-seven patients had more than one
operation (sixty had two and seven had three), none had the procedures
performed simultaneously. The mean interval between the procedures
was 25.7 months (range, six to sixty months). The age and weight
of the patient at the time of each procedure were used for the statistical
calculation.
The parameters that were evaluated included the preoperative hemoglobin
level, gender, age, type of surgical procedure, weight, height,
need for transfusion, and number of units of packed red blood cells
transfused.
All patients were given antithromboembolic prophylaxis with low-molecular-weight
heparin (nadroparin calcium; Sanofi Winthrop, Paris, France). Patients
who weighed <70 kg received 0.3 mL (3075 IU) subcutaneously
twelve hours before surgery, followed by 0.3 mL/day for
the first three days after surgery and 0.4 mL (4100 IU)/day
from the fourth day to a maximum of two months after surgery. Patients
who weighed >=70 kg received 0.4 mL preoperatively,
followed by the same daily dose for three days and then a dose of
0.6 mL (6150 IU)/day up to two months after surgery. (This
schedule has not been approved by the United States Food and Drug Administration.)
The statistical analysis consisted of a univariate analysis of
all independent variables (hemoglobin level, age, gender, weight, height,
and type and duration of surgery) to establish their relationship
with the need for postoperative transfusion. Variables that were
seen to have a significant relationship were included in a multivariate
analysis by logistic regression.
The adjustment was performed backward by the likelihood of model —2LL0
(procedure RV of SPSS, version 9; Hispanoportuguesa SPSS, Madrid,
Spain).
Transfusion was required in 117 patients (128 procedures). Allogenic
blood was used in all instances. An average of 2.5 units (range,
one to six units) of packed red blood cells, with each unit equivalent
to 300 mL of red-blood-cell concentrate, was given to the patients
who received a transfusion. The average duration of the hip arthroplasties
was 89.5 minutes (range, fifty to 180 minutes), and the patients
undergoing those procedures had a mean preoperative hemoglobin level of
145 g/L (range, 101 to 179 g/L). Transfusion was
required after 39.2% (eighty-two) of the hip arthroplasties
(in seventy-four patients). The average duration of the knee arthroplasties was
91.1 minutes (range, fifty to 180 minutes), and the patients undergoing
those procedures had an average preoperative hemoglobin level of
142 g/L (range, 114 to 185 g/L). Transfusion was
required after 28.6% (forty-six) of the knee arthroplasties
(in forty-three patients). The multivariate analysis revealed no
relationship between the type of procedure (hip or knee) and the
need for transfusion (p = 0.73)
A chi-square test showed no significant difference in the need for
transfusion between patients who had had more than one joint replacement
and those who had been operated on once (p = 0.76). Therefore,
the patients who had a reoperation were considered new patients.
A univariate analysis including all of the independent variables
(Table I) established
a significant relationship between the need for postoperative transfusion
and the preoperative hemoglobin level (p = 0.0001), duration
of surgery (p = 0.0001), weight (p = 0.002), height
(p = 0.019), and gender (p = 0.0056). However,
no relationship was found between the age of the patient and the
need for transfusion (p = 0.74).
Although the duration of surgery had a significant relationship with
the need for transfusion in the univariate analysis, it was not
included in the multivariate analysis because it was a postoperative
variable and did not have predictive value.
The multivariate analysis revealed two significant variables: preoperative
hemoglobin level (p = 0.0001) and weight (p = 0.011),
with odds ratios of 2.51 (95% confidence interval, 1.83
to 3.44) and 1.05 (95% confidence interval, 1.01 to 1.09), respectively.
The logistic regression equation for the probability (p) of no transfusion
being required with use of these variables was p = 1/1 + e-(-16.04 + 0.092 ¥ hemoglobin
level + 0.05 ¥ weight), where e = 2.72.
This means that for each ten grams per deciliter of increase in
the preoperative hemoglobin level, the probability that the patient
will not need a transfusion increases 2.5 times (Fig. 1). In addition,
for each kilogram of increase in the weight of the patient, the
probability of no transfusion increases by 1.05 times. Although
weight is significant, preoperative hemoglobin levels have a greater
influence on the need for transfusion. All procedures were classified
with use of the above equation, and prediction was accurate for
89.3% (216) of the procedures without transfusion, 53.1% (sixty-eight)
of those with transfusion, and 76.8% (284) of all procedures.
Greater than 75% is considered to be good prediction.
Depending on the preoperative hemoglobin values, the sample was
divided into three groups: Group 1 (<130 g/L),
Group 2 (130 to 150 g/L), and Group 3 (>150 g/L).
Transfusion was required after 69% (forty-five) of the
procedures performed in patients with a hemoglobin level of <130
g/L, 36% (sixty-eight) of the procedures performed
in patients with a level between 130 and 150 g/L, and 13% (fifteen)
of the procedures performed in patients with a level of >150
g/L (Fig. 2). The risk of transfusion in Group
1 was four times greater than that in Group 2 and 15.3 times greater
that that in Group 3.
Blood transfusion does not effectively restore hemoglobin levels
to preoperative levels17. Also,
it entails a risk of transmitting diseases such as hepatitis and
AIDS2,3, which has led to the
use of preoperative self-donation as an alternative18. Furthermore, identification errors
can occur during the handling of blood units, which can be a source
of medical problems and legal liability. In addition, a patient
may refuse a transfusion, particularly for religious reasons. Hence,
it is important to predict a priori the target population at a higher risk
of requiring blood transfusion and to establish the appropriate
prophylactic measures.
Various studies have demonstrated a relationship between preoperative
hemoglobin levels and the need for blood transfusion10,14,16,19-22. It seems clear that
the most important factor in the prediction of the need for blood
transfusion is the preoperative level of hemoglobin1,10,14-16,19,21,22, although other
factors, such as weight11 and
volume of blood loss19(which
are directly related to each other), are involved.
The criterion for determining whether a transfusion is indicated
is not a static value; the clinical status of the patient must be
taken into account2,16,23. A hemoglobin
level of approximately 80 to 90 g/L is accepted by most
as a relative indication for transfusion2,16.
For this reason and to standardize our sample, a hemoglobin level
of <85 g/L was established as the criterion for
transfusion, despite the fact that we are aware of the importance
of clinical factors2.
The univariate analysis showed a significant relationship between
postoperative transfusion and the preoperative hemoglobin level
(p = 0.0001), weight (p = 0.002), height (p = 0.019),
gender (p = 0.0056), and duration of surgery (p = 0.0001).
However, the multivariate analysis revealed a significant relationship
only with the preoperative hemoglobin level (p = 0.0001)
and weight (p = 0.011), with odds ratios of 2.51 and 1.05,
respectively.
On the basis of the results obtained in this study, with multivariate
analysis showing no significant differences regarding the need for
transfusion between hip and knee surgery, it was possible to establish
an equation, based on the preoperative hemoglobin level (in grams
per liter) and weight (in kilograms), for calculating the probability
of a patient not requiring a transfusion: p = 1/1 + e-(-16.04 + 0.092 ¥ hemoglobin
level + 0.05 ¥ weight). This formula
should be applicable to patients with a hemoglobin level and weight
within the range of our study group (hemoglobin range, 101 to 185
g/L, and weight range, 53 to 105 kg). The formula is not
valid for patients who meet any of the exclusion criteria stated
in the Materials and Methods section of this paper. Furthermore,
it is not valid for patients who receive a transfusion because of
complications that occur after the surgical procedure (for example,
gastrointestinal bleeding). The values obtained with use of this
formula fall between 0 and 1, with 0 corresponding to 0% and
1 corresponding to 100% (the percentages represent the
probability of each particular patient not requiring a transfusion).
In addition to its clinical importance, we believe that this information could
also be useful in blood bank forecasts.
The hemoglobin level, the most important prognostic factor, could
be classified into three distinct groups: <130 g/L (Group
1), 130 to 150 g/L (Group 2), and >150 g/L
(Group 3). Blood transfusion was required after 69% (forty-five)
of the procedures performed in Group-1 patients, and this risk was four
times greater than the risk for Group-2 patients and 15.3 times
greater than the risk for Group-3 patients. These results are consistent
with those of Faris et al.19,
who stated that patients with a preoperative hemoglobin level between
100 and 130 g/L had an increased risk of requiring blood
transfusion. Sculco and Gallina21 reported
that preoperative hemoglobin levels were inversely proportional
to the need for a transfusion. We found, as did Bierbaum et al.1, that the prevalence of blood transfusion
after hip surgery is higher than that after knee surgery, although
this difference was not significant. The prevalence of transfusions
in patients with a preoperative hemoglobin level of >130
g/L in our series was lower than that in the study by Sculco
and Gallina. This difference may be explained by the fact that patients
who received a transfusion even though the pretransfusion hemoglobin
level was >85 g/L were excluded from our study
but were not excluded from the study by Sculco and Gallina.