A multicenter retrospective cohort study was conducted to review the cases
of patients with a periprosthetic infection after total knee arthroplasty at
the OrthoCarolina Hip and Knee Center from 1987 to 2003 as well as those
treated at the Mayo Clinic from 1995 to 1999. Data from the joint registries
at each site and from the medical records were retrospectively reviewed. The
majority of these patients were referred to our tertiary centers for revision
surgery. Of the 263 knees that were treated because of an infection around a
total knee replacement during these periods, thirty-seven (14%) had an
infection with methicillin-resistant Staphylococcus aureus or
methicillin-resistant Staphylococcus epidermidis.
Of the thirty-seven knees, twenty-five had an infection with
methicillin-resistant Staphylococcus aureus and twelve had an
infection with methicillin-resistant Staphylococcus epidermidis.
Twenty of the thirty-seven patients were men, and seventeen were women. The
median age was seventy years (interquartile range, 11.5 years) at the time of
the index treatment. Twenty-five of these primary total joint arthroplasties
had been performed for osteoarthritis; eight, for rheumatoid arthritis; and
four, for posttraumatic arthritis. The median duration of follow-up was
fifty-one months (range, twenty-four to 111 months).
In this cohort, management decisions were made individually by the treating
physicians. All patients had joint cultures and histopathologic evaluation of
synovial tissue at the time of resection arthroplasty and reimplantation. The
following variables were evaluated as potential risk factors for failure:
comorbidities, prior surgical procedures, and the duration of intravenous
antimicrobial therapy. The comorbidities reviewed in this study included
diabetes mellitus, tobacco use, exogenous steroid dependency, rheumatoid
arthritis, and renal insufficiency. Failure of the two-stage reimplantation
was defined as the need for an arthrodesis or an amputation or a reinfection
with the same organism. Additional surgery secondary to instability or aseptic
loosening was noted but was not considered a failure of treatment.
Surgical Technique
Each of these patients who had a documented infection with
methicillin-resistant Staphylococcus aureus or methicillin-resistant
Staphylococcus epidermidis underwent a two-stage reimplantation
procedure. Previously healed skin incisions were utilized and extended as
needed to perform the arthrotomy. Once the joint was exposed, multiple
cultures were taken and sent for aerobic and anaerobic cultures and tissue was
obtained for frozen and permanent pathologic section. Antibiotics were
withheld until cultures were obtained.
The prosthesis and all cement were removed with use of bone-conserving
techniques. Each joint then underwent débridement of the synovial
membrane, pseudomembrane, necrotic tissue, and bone, if necessary. Regions of
osteomyelitis were curetted until healthy bone was encountered. The resected
bone ends and joint space were thoroughly irrigated with use of pulsatile
lavage, and a polymethylmethacrylate cement spacer laden with antibiotics was
placed. Depending on the operating surgeon, an articulating or static spacer
was used along with intramedullary rods or beads. If the infecting organism
was known at the time of surgery, an organism-specific antibiotic was chosen
for use in the cement. If the infecting organism was not known at the time of
resection arthroplasty or multiple organisms were suspected, broad coverage
with vancomycin and an aminoglycoside was used. The actual doses per batch of
cement (40 g) varied among the operating surgeons. Thirty-five (95%) of the
thirty-seven patients were treated with a spacer made of antibiotic-laden
cement, with use of an aminoglycoside and/or vancomycin. No antibiotic was
used in one other patient, and the type of cement used in the remaining
patient was not available by chart review (see Appendix).
Intravenous antimicrobial therapy was administered on the basis of the
recommendations of an infectious disease consultant. Patients also received
oral antibiotics during or after the course of intravenous antibiotics
according to the discretion of the infectious disease consultant. Before
reimplantation, the erythrocyte sedimentation rate and/or C-reactive protein
levels were obtained to help to determine the presence of persistent
infection. Normalization or substantial improvement in these serologic
markers, as determined by the operating surgeon, helped to guide the timing of
reimplantation. In patients with systemic inflammatory disease, these
serologic markers do not consistently return to normal. In these patients, the
wound appearance and the results of the aspirate before reimplantation were
used to guide the decision to reimplant.
Acrylic cement laden with vancomycin and/or an aminoglycoside was used in
thirty-three of the thirty-seven patients at the time of reimplantation. The
remaining four knee implants were cemented without the inclusion of
antibiotics in the cement. Antimicrobial therapy was continued postoperatively
until all cultures and pathology reports were final and reported as
negative.
Statistical Analysis
The Shapiro-Wilk test was used to determine the non-normal distribution of
the data. Standard descriptive statistics, including median and interquartile
range, were reported. The interquartile range is a more stable statistic than
the range. Interquartile range is defined as the difference between the 75th
percentile (third quartile) and the 25th percentile (first quartile) and is a
descriptive measure of variability. Categorical variables were compared with
use of the chi-square Fisher exact test. Two-by-two contingency tables were
designed to assess the effect of the length of intravenous administration of
antibiotics (greater than six weeks and less than six weeks) on the rate of
reinfection (reinfection or no reinfection); the time to reimplantation
(greater than eight weeks and less than eight weeks) on the rate of
reinfection; previous total joint surgery (previous surgery and no previous
surgery) on the rate of reinfection; and comorbidity (presence of
comorbidities and no comorbidity) on the rate of reinfection. Additionally,
the relative risk of comorbidities was calculated. For all statistical tests,
type-I error was set at 0.05. Post hoc power analyses were performed with use
of observed proportions and sample sizes from the Fisher exact tests.
Four (11%) of the thirty-seven patients had a reinfection with the same
organism: methicillin-resistant Staphylococcus epidermidis in three
patients and methicillin-resistant Staphylococcus aureus in one. Of
these four failures, two patients were lost to follow-up and two were unable
to tolerate repeat surgery because of poor health and were treated with
chronic suppressive antibiotics. Five patients (14%) had a reinfection with a
different organism (see Appendix). Of the five patients, one underwent another
two-stage reimplantation. The remaining four patients underwent irrigation and
débridement with exchange of the tibial insert only and retention of
the components. At a minimum of three years of follow-up, infection had not
recurred in these patients.
The duration of antimicrobial therapy did not appear to be related to
reinfection. All but one of the patients were treated with a course of
intravenous vancomycin between the resection arthroplasty and the
reimplantation. The remaining patient was treated with oral trovafloxacin. The
median duration of the intravenous antimicrobial therapy was six weeks
(interquartile range, twelve weeks). Of the thirty-seven patients, twenty-two
were treated with intravenous antibiotics for at least six weeks. Of these
twenty-two patients, two had a reinfection while twenty (91%) had no
reinfection. Fifteen patients were treated with intravenous antibiotics for
less than six weeks. Of these fifteen patients, two had a reinfection while
thirteen had no reinfection. In addition to the intravenous administration of
antibiotics after the first stage, oral antibiotics were used to treat eight
patients. Rifampin was used to treat six patients, and the remaining two
patients were treated with trimethoprim-sulfamethoxazole (see Appendix). The
remaining twenty-nine patients were not treated with oral antibiotics between
stages. No significant association was detected, on the basis of the numbers,
between the length of intravenous antibiotic therapy and reinfection (p =
0.7). Of the thirty-seven patients, four did not receive antib iotics in the
cement at the time of reimplantation. None of the four patients had a
reinfection.
The timing of reimplantation did not seem to affect the rate of
reinfection. The average time to reimplantation following resection was 12.9
weeks (range, four to fifty-nine weeks). In the sixteen patients undergoing
reimplantation before eight weeks, there were two reinfections. In the
twenty-one patients undergoing reimplantation after eight weeks, there were
also two reinfections. This difference was not significant (p = 0.77), given
the small numbers.
All cultures obtained at the time of reimplantation were negative. Six
patients had a final histopathological finding consistent with acute
inflammation, defined as an overall histologic picture of acute inflammation
(that is, multiple foci of polymorphonuclear leukocytes throughout the samples
studied)11. Two
patients with negative cultures but positive histopathological findings were
placed on chronic suppressive antibiotics, and neither had a reinfection. Four
patients with negative cultures and final histopathological findings
consistent with infection at the time of reimplantation were not given
suppressive antibiotics. Two of those four patients went on to have a
reinfection (see Appendix).
The influence of previous surgery was reviewed and found not to predispose
the patient to reinfection. Ten (27%) of thirty-seven patients underwent one
or more previous surgeries prior to total knee arthroplasty. Of those ten
patients, one had a reinfection. Of the twenty-seven patients who had had no
previous surgery, three (11%) had a reinfection. No significant association (p
= 0.9) was found between previous surgery and reinfection, on the basis of the
numbers. Comorbidities also did not influence the reinfection rate. Sixteen
(43%) of the thirty-seven patients were found to have specific comorbidities.
Of the twenty-one patients with no comorbidity, twenty had no reinfection and
one had a reinfection. Of the sixteen patients with comorbidities, thirteen
had no reinfection and three had a reinfection. No significant association (p
= 0.17) was found between comorbidities and reinfection, with the small
numbers available. However, patients with comorbidities were nearly five times
more likely to have a reinfection (relative risk, 4.84).
Without adequate power for any of the analyses (1 — ß =
<0.2), it is difficult to draw definitive conclusions about the effect of
intravenous antibiotics, previous surgery, or comorbidity on the rate of
reinfection. The sample size was simply inadequate relative to the variability
of the data and the difference in proportions between groups.
The two-stage treatment protocol is the most consistently effective method
of treatment for an infection around a total knee replacement when both
function and eradication of the infection are considered as criteria for a
successful
outcome12. Multiple
studies reviewing the results of two-stage reimplantation have been published
through the
years2-8.
The combined success rate in those studies was 83.2%. This success rate,
however, was an overall average and was not stratified on the basis of the
infecting organism.
While the overall success rate in our series was only 76% (twenty-eight of
thirty-seven knees), most (five) of the nine knees that had failure of the
reimplantation were reinfected with a different organism. Therefore, we were
encouraged that we were successful in eradicating the resistant organisms in
this group of patients nearly 90% of the time. This success rate is similar to
those in our previously reported series of infections at the site of total
knee replacements that were treated in a two-stage fashion at our respective
institutions13,14.
In the series from the OrthoCarolina Hip and Knee Center, 88% (twenty-two) of
twenty-five patients with static spacers and 93% (twenty-eight) of thirty
patients treated with articulating spacers remained free of infection at the
time of
follow-up13. In a
recent series from the Mayo Clinic, 91% of ninety-six knees that had two-stage
reimplantation because of infection at the site of a total knee replacement
were free of infection at the time of
follow-up14. The
first series from OrthoCarolina was not stratified for organism type. In the
Mayo Clinic series, there was no difference in recurrence rates between
methicillin-susceptible and methicillin-resistant staphylococcal infections,
although the sample size was small.
In a series of fifty-five knees that underwent a two-stage procedure,
reimplantation was successful in 83% (twenty) of twenty-four knees with
coagulase-negative staphylococci, streptococcus, and proteus; twelve of
eighteen knees with Staphylococcus aureus and enterococcus species;
five of seven knees with polymicrobial infection; and four of six knees with
methicillin-resistant Staphylococcus aureus. It is difficult to
explain the difference between our series and the limited reports with lower
success
rates9-10.
Lower success rates have been explained by the mechanism of resistance among
methicillin-resistant staphylococci. When dealing with these organisms, there
is a loss of affinity for targeted penicillin-binding proteins. This means
that the target site for the entire class of beta-lactam antibiotics is lost.
Thus, methicillin-resistant staphylococci have been traditionally treated with
vancomycin, a drug with inferior antistaphylococcal activity compared with
beta-lactams active against
staphylococci15.
Although several published series have advocated the use of at least six
weeks of intravenous antibiotics after the first stage, the duration of
antibiotic administration in our series ranged from zero to twelve
weeks2,3,16.
Different infectious disease consultants guided antibiotic treatment at each
institution, leading to this nonuniform approach. We noted no significant
association, on the basis of the numbers, between the length of antibiotic
administration and the rate of reinfection in our study.
A less successful outcome in patients undergoing two-stage reimplantation
has been noted in patients who underwent surgery prior to the knee replacement
or had a diagnosis of rheumatoid
arthritis17. Our
series, however, showed no increase in reinfection rates in these groups of
patients.
Multiple previous studies have noted that diabetes, obesity, and smoking
predispose patients to a number of complications, including wound
infections18-21.
A statistical analysis was performed to determine the association of comorbid
conditions with the rate of reinfection. While this association was not
significant, patients with comorbidities were nearly five times more likely to
have a reinfection (relative risk, 4.84). Again, the small sample size in our
study limits our ability to make a firm conclusion regarding this
association.
We noted, however, that of the nine reinfections in our thirty-seven
patients, five were caused by a different organism, methicillin-sensitive
Staphylococcus aureus. This finding raises the possibility that host
compromise may play a role in the recurrence of periprosthetic infection. In
these instances, the primary organism was eradicated with use of the standard
two-stage technique, yet the host remained susceptible to reinfection
following reimplantation. While none of the comorbidities examined in this
study were specifically isolated as increasing the risk of infection, other
factors such as malnutrition may have played a role. An albumin level of
<3.5 g/dL, a total lymphocyte count of <1500, or a serum transferrin
level of <262 mg/dL have all been found to increase the risk of infection,
length of hospital stay, and mortality in elderly
patients22-24.
Although it was not studied in this group of patients, nutritional
supplementation between stages and a nutritional evaluation before
reimplantation may have a role in the treatment of high-risk patients who have
an infection with resistant organisms.
Since the results in the present study are similar to those in reports on
two-stage reimplantation without organism stratification, we believe that
two-stage reimplantation remains a viable option for patients who have an
infection with a resistant organism. With regard to the management of
periprosthetic knee infection, we offer the following recommendations.
A thorough débridement of all hardware and cement must be performed at
the initial stage. Intraoperative radiographs should be made to ensure
complete cement removal at the time of resection. The proper timing of
reimplantation after parenteral antibiotic therapy should be based on the
clinical appearance of the wound and improvement in serologic markers of
infection, such as the C-reactive protein level and the erythrocyte
sedimentation rate.
At the time of reimplantation, tissue specimens should be sent for culture
and histopathologic evaluation. If all preoperative and intraoperative indices
are acceptable, then reimplantation ensues. If the frozen section suggests
continued acute inflammation, the wound should be débrided again, a
cement spacer reapplied, and the wound closed. If after reimplantation,
intraoperative cultures or final histologic findings are positive for
infection or acute inflammation, the decision to institute suppressive
antibiotic therapy needs to be discussed with an infectious disease
consultant. In our study, six patients had a final histologic finding
consistent with persistent acute inflammation. The two patients who received
chronic suppressive antibiotics remained free of infection while two of the
remaining four patients who did not receive chronic antibiotic treatment had a
reinfection with the same organism. Chronic suppressive antibiotic therapy may
have benefited those
patients24.
Although therapy for methicillin-resistant staphylococci is limited to a few
intravenous antibiotics (e.g., the standard is vancomycin), many of these
isolates remain susceptible to oral agents such as
trimethoprimsulfamethoxazole and even tetracyclines, such as minocycline,
which may be used as suppressive
agents15.
This study has limitations. The median length of follow-up was fifty-one
months. Multiple surgeons and infectious disease physicians using varying
lengths of antibiotic therapy treated these patients. In addition, one patient
did not receive antibiotics in the interim spacer and four patients did not
receive antibiotics in the cement at the time of reimplantation. Fortunately,
none of those patients had a reinfection because of this break in standard
protocol. This study also had a limited number of thirty-seven patients who
were seen with methicillin-resistant Staphylococcus aureus or
methicillin-resistant Staphylococcus epidermidis infections, although
it is, as far as we know, the largest series of patients with periprosthetic
knee infections involving these resistant bacteria.
Tables showing antibiotic cement dosages, parenteral and oral antibiotic
regimens, and a summary of the clinical outcome for all thirty-seven patients
are available with the electronic versions of this article, on our web site at
(go to
the article citation and click on "Supplementary Material") and on
our quarterly CD-ROM (call our subscription department, at 781-449-9780, to
order the CD-ROM). ?