A computer-assisted review of 22,540 consecutive total knee arthroplasties
that had been performed at our institution between 1969 and 1997
was conducted to identify patients who had died within thirty days
after the operation. Data on all patients undergoing arthroplasty
are entered prospectively into the Total Joint Registry at our institution,
and the information is updated at regular intervals. We chose a
thirty-day perioperative time-limit on the assumption that a shorter
time-period might have led to the exclusion of deaths that were directly
related to the knee surgery whereas a longer time-period might have
led to the inclusion of deaths that were not attributable to the
knee surgery.
Forty-seven patients died within thirty days after total
knee arthroplasty. The records of these forty-seven patients
were reviewed with regard to age, gender, medical history, preoperative
diagnosis, details of the operation (including whether or not cement
was used), and postoperative events. The pathology department records
were reviewed, and the findings at autopsy (if performed) were compared
with the clinical cause of death that had been recorded prior to
autopsy. The data on the forty-seven patients who died
were compared with the data on all of the patients in the Total
Joint Registry in order to calculate thirty-day mortality
for the entire group as well as for specific subgroups that were
stratified according to diagnosis, gender, type of operation (primary
or revision, unilateral or bilateral), method of implant fixation
(with or without cement), and age (more or less than seventy years).
Statistical Analysis
Single risk factors were analyzed with either the chi-square test
or the Fisher exact test. Multivariate analysis was performed with
use of a logistic regression model. Risk factors were retained in
the model if they were either significant (p < 0.05) or
of borderline significance (p < 0.10). During the modeling,
two-way interactions were studied for all combinations
of risk factors. After completion of the modeling, the relative
risk of death for each combination of risk factors was estimated
with 95% confidence intervals.
Demographic Data
Of the 22,540 patients who had a knee arthroplasty between 1969
and 1997, forty-seven died within thirty days after the operation.
Twenty-seven of these patients were men, and twenty were
women. The average age at the time of the procedure was significantly
higher for those who died (seventy-five years; range, sixty
to ninety years) than for those who did not die (65.5 years; range,
sixteen to ninety-eight years) (p < 0.005). The average
age of the patients who died after bilateral total knee arthroplasty
(seventy-four years; range, seventy to eighty-eight years)
was not significantly different from that of the patients who underwent
unilateral total knee arthroplasty (seventy-five years; range, sixty
to ninety years) (p = 0.7). The patients who underwent
revision knee arthroplasty were younger than those who underwent
primary knee arthroplasty (average age, 66.2 compared with 75.8
years) (p < 0.01). The average age of the 18,810 patients
in whom the procedure was performed with cement (65.9 years; range, thirty-eight
to ninety-six years) was significantly higher than that of the 3730
patients in whom the procedure was performed without cement (49.2
years; range, twenty-nine to sixty years) (p < 0.004).
The indication for knee arthroplasty was osteoarthritis in thirty-two
of the patients who died, rheumatoid arthritis in twelve, pathological
fracture in one, ankylosing spondylitis in one, and avascular necrosis
in one. Forty-three patients died after primary knee arthroplasty,
and four died after revision arthroplasty. Thirteen (27.7%)
of the forty-seven patients died after a one-stage
bilateral knee arthroplasty performed during the same anesthetic
session. Eight of the patients who died had had a previous knee
arthroplasty on the contralateral side.
Mortality Rates
Three patients (0.01%) died intraoperatively. Two of
these patients died after sudden cardiorespiratory collapse during cementing
of the tibial component (one patient) or deflation of a tourniquet
(one patient), and the third patient died during débridement
of infected tissue and removal of a knee prosthesis. The overall
thirty-day mortality after knee arthroplasty was 0.21% (forty-seven
of 22,540), including the three intraoperative deaths. The thirty-day
mortality after revision knee surgery (0.09%; four of 4375)
was significantly lower than that after primary knee surgery (0.24%;
forty-three of 18,165) (p < 0.0003). All four
patients who died after revision surgery had received a long-stem,
cemented prosthesis. There were no deaths among the 3730 patients
in whom the procedure was performed without cement. The thirty-day
mortality rate after procedures performed with cement (0.25%;
forty-seven of 18,810) was significantly higher than that after
procedures performed without cement (0%; zero of 3730)
(p < 0.0001). Thirteen patients died after bilateral knee
arthroplasty. Eleven of these thirteen patients had undergone a
simultaneous bilateral knee arthroplasty, and the other two had
undergone a staged bilateral knee arthroplasty during the same anesthetic session.
The thirty-day mortality after bilateral knee arthroplasty (0.49%;
thirteen of 2679) was significantly higher than that after unilateral
knee arthroplasty (0.17%; thirty-four of 19,861)
(p < 0.002). There was no difference in the mortality rates
associated with different diagnoses. Specifically, the perioperative
mortality among patients who had rheumatoid arthritis (0.20%;
twelve of 6102) was not significantly different from that among
patients who had osteoarthritis (0.20%; thirty-two of 16,091)
(p = 0.82).
Surgical Data
Examination of the anesthetic records revealed that an intraoperative
hypotensive event, defined as a drop in systolic blood pressure
of more than 20 mm Hg, was observed in eighteen of the forty-seven
patients who died. Hypotension occurred at the time of tourniquet
deflation in seven patients, cementing of the tibial component in
five, cementing of the femoral component in three, insertion of
a femoral intramedullary guide-rod in two, and closure of the fascia
in one. The hypotensive event was accompanied by a recorded arrhythmia in
nine patients.
Cause and Time of Death
An autopsy was performed on thirteen of the forty-seven patients
who died. Routine macroscopic pathological examination of vital
organs was performed, and the site of the knee surgery was also
examined in detail. The major findings were limited to the lungs
and the heart. Comparison of the cause of death that had been recorded
prior to autopsy with the actual cause of death that was identified
at the time of the postmortem examination revealed a close association
between the clinical and autopsy reports for eleven of thirteen
patients. The cause of death was respiratory failure secondary to
acute pulmonary edema in ten of the thirteen patients and acute myocardial
infarction in three.
No autopsy was performed for the other thirty-four patients who
died; the presumed cause of death was sudden respiratory failure
in twenty-one, cardiac arrest in eight, sepsis from a source
other than the knee in two, and cerebrovascular accident in two.
The remaining patient committed suicide twenty-seven days
after the index procedure.
Nineteen of the forty-seven patients died in the hospital.
Medical History
Forty-three of the forty-seven patients who died had a strong history
of preexisting cardiovascular and/or pulmonary disease.
Thirty-eight patients had a history of preexisting cardiovascular
disease, and twenty-nine had a history of pulmonary disease (Table II). Twenty-three
patients had coexistent cardiovascular and pulmonary disease, and
twenty had at least two coexistent cardiovascular conditions.
In 1983, our institution began routine recording of the American
Society of Anesthesiologists score13 for
all patients undergoing operative treatment. The average score was
calculated for the patients who died, and we were able to obtain
an accurate score for 7522 patients who did not die. The average
score for the patients who died (3.2; range, 2 to 4) was significantly
higher than that for the patients who did not die (2.4; range, 1
to 4) (p < 0.0001).
Analysis of Risk Factors
Univariate analysis revealed that several risk factors were associated
with a significant increase in thirty-day mortality. These
included use of a cemented prosthesis, one-stage simultaneous bilateral
arthroplasty, an age of more than seventy years, primary arthroplasty,
and a history of cardiorespiratory disease (Table I). Factors that
were not significant included gender, operating surgeon, preoperative
diagnosis (osteoarthritis or rheumatoid arthritis), and side of
the operation.
The present study provides data on perioperative mortality associated
with total knee arthroplasty from a large consecutive cohort of
patients who were operated on at a single institution. Primary arthroplasty
was associated with significantly greater perioperative mortality
compared with revision arthroplasty (p < 0.0003), despite
the fact that revision procedures are longer and involve greater
blood loss. The adverse effects of embolization of fat and marrow
contents from a previously unviolated medullary canal may partly
explain this difference. By using ultrasonography to detect emboli
in the femoral vein and by analyzing femoral vein blood samples,
Herndon et al.14 showed a markedly
decreased amount of embolic material in patients who had had previous
insertion of a cemented femoral hip prosthesis. However,
selection bias may also have contributed to our results. For example,
surgeons may have been more willing to subject older and less healthy
patients to primary knee arthroplasty than to revision arthroplasty.
In addition, the patients who underwent revision knee arthroplasty were
younger than those who underwent primary knee arthroplasty (average
age, 66.2 compared with 75.8 years; p < 0.01).
An age of more than seventy years appeared to be a risk factor for
perioperative mortality. Although we were unable to document the
preoperative medical condition of the 22,493 patients who did not
die after total knee surgery, it is reasonable to assume that older
patients were more likely to have medical comorbidities that contributed
to the risk of perioperative mortality.
We also found that perioperative mortality after simultaneous bilateral
total knee arthroplasty was significantly higher than that after
unilateral arthroplasty (p < 0.002). The reasons for this
finding are likely multiple and intertwined. The patients who died
after bilateral arthroplasty were slightly sicker preoperatively
than those who did not die after bilateral arthroplasty (average
preoperative American Society of Anesthesiologists score, 2.6 compared
with 2.3), but this difference was not significant (p = 0.7).
Another possible explanation may be that simultaneous bilateral
knee arthroplasty performed during the same anesthetic session may
result in more blood loss, leading to a greater degree of hypotension and
necessitating greater colloid and vasopressor administration. Finally,
simultaneous bilateral knee arthroplasty may produce a higher "embolic
load" of fat and marrow elements to the heart and lungs,
with adverse physiological consequences leading to more pronounced
cardiorespiratory disturbances in some patients. Lynch et al.5, in a study evaluating the risks
associated with bilateral knee arthroplasty in elderly patients,
observed that more cardiac, respiratory, and neurological complications
developed in patients managed with bilateral arthroplasty than in
those managed with unilateral arthroplasty.
An interesting finding of this study was the higher mortality observed
among patients who had cement fixation. This difference may have
resulted in part from biased patient selection. Uncemented components
were more often used for younger patients (average age, 49.2 years),
during the years of peak use of these devices, whereas cemented
components were more often used for elderly patients (average age,
65.9 years). In some patients, however, the difference may have been
caused by the marrow fat and debris embolization process that has
been shown to occur during total joint arthroplasty1,4,8,12,15-19. Orsini et al.20, in an experimental study in dogs,
showed that increased intramedullary pressures were generated during
cementation of femoral rods. Marked cardiorespiratory changes, including decreased
arterial oxygen tension, increased intrapulmonary shunt fraction,
and increased pulmonary arterial pressures, were observed. We believe
that the rare intraoperative deaths reported previously8,16,21,22 and in the current study
likely represent an uncommon fulminant manifestation of a very common
physiologic insult associated with joint arthroplasty in an at-risk
host. Frequently, patients undergoing knee replacement have concomitant
cardiorespiratory diseases and may have limited cardiorespiratory
reserve9,23. It has been postulated
that such patients who have a cardiorespiratory crisis during cementation
of components may go on to persistent pulmonary dysfunction, resulting
in perioperative mortality9,20,23,24.
Acute hypotension occurs commonly during cementing and has been
associated with elevated pulmonary artery pressure and transient
hypoxemia9,23,25-27. These physiological
changes have been demonstrated at the time of cement and prosthesis
insertion both in patients13,22,27 and
in animals28. Although the pathophysiology
of hypotension associated with implant insertion remains incompletely
understood, release of particulate fat and marrow emboli has been
clearly implicated24,28. The adverse
effects of this embolic load can be moderated by decompression of
the femoral canal by overdrilling the femoral entry hole, use of
suction devices, use of fluted or cannulated rods, irrigation of
the canal, and slow insertion of the instruments16,18,19.
Another striking finding of this study was the presence of severe
underlying cardiovascular and pulmonary disease in forty-three
of the forty-seven patients who died. We were able to confirm
that the prevalence of preexisting comorbidities, as measured by
American Society of Anesthesiologists score, was significantly higher
in patients who died compared with patients who did not die (p < 0.0001).
In conclusion, the factors that were associated with increased mortality
after knee arthroplasty included an age of more than seventy years,
primary arthroplasty, the use of cement, preexisting cardiopulmonary
disease, and simultaneous bilateral total knee arthroplasty. The
overwhelming majority of the patients who died (forty-three
of forty-seven) had preexisting cardiopulmonary disease.
Recognizing these risk factors may enable the orthopaedic surgeon
to reduce perioperative mortality after knee arthroplasty. Measures
that may be worthy of consideration include avoiding bilateral one-stage
total knee arthroplasty in patients who are ill or elderly; optimizing
the medical condition of patients who have a history of cardiac
or pulmonary problems; vigilant anesthetic monitoring, especially
around the times of surgical measures that are known to be associated
with marrow and fat embolization; and liberal use of vasopressor
agents during episodes of hypotension. Modifications in surgical
technique and implant choice to reduce marrow and fat embolization
may also be appropriate in some high-risk patients. Awareness of
the risk factors associated with perioperative mortality and use
of measures to minimize the insult of the surgical procedure should
help to make this already safe and effective operation even safer.