MAJOR ARTICLE

Gram-Negative Prosthetic Joint Infections:

Risk Factors and Outcome of Treatment
Pang-Hsin Hsieh,1,2 Mel S. Lee,1,2 Kuo-Yao Hsu,1,2 Yu-Han Chang,1,2 Hsin-Nung Shih,1,2 and Steve W. Ueng1, 2
1
Department of Orthopedic Surgery, Chang Gung Memorial Hospital, and 2College of Medicine, Chang Gung University, Taoyuan, Taiwan

Downloaded from http://cid.oxfordjournals.org/ at University of Western Ontario on April 11, 2015

Background. Little information is available regarding the demographic characteristics and outcomes of patients
with prosthetic joint infection (PJI) resulting from gram-negative (GN) organisms, compared with patients with
PJI resulting from gram-positive (GP) organisms.
Methods. We performed a retrospective cohort analysis of all cases of PJI that were treated at our institution
during the period from 2000 through 2006.
Results. GN microorganisms were involved in 53 (15%) of 346 first-time episodes of PJI, and Pseudomonas
aeruginosa was the most commonly isolated pathogen (21 [40%] of the 53 episodes). Patients with GN PJI were
older (median age, 68 vs. 59 years; P ! .001 ) and developed infection earlier (median joint age, 74 vs. 109 days;
P ! .001) than those with GP PJI. Of the 53 episodes of GN PJI, 27 (51%) were treated with debridement, 16
(30%) with 2-stage exchange arthroplasty, and 10 (19%) with resection arthroplasty. Treating GN PJI with de-
bridement was associated with a lower 2-year cumulative probability of success than treating GP PJI with debri-
dement (27% vs. 47% of episodes were successfully treated; P p .002 ); no difference was found when a PJI was
treated with 2-stage exchange or resection arthroplasty. A longer duration of symptoms before treatment with
debridement was associated with treatment failure for GN PJI, compared with for GP PJI (median duration of
symptoms, 11 vs. 5 days; P p .02).
Conclusions. GN PJI represents a substantial proportion of all occurrences of PJI. Debridement alone has a
high failure rate and should not be attempted when the duration of symptoms is long. Resection of the prosthesis,
with or without subsequent reimplantation, as a result of GN PJI is associated with a favorable outcome rate that
is comparable to that associated with PJI due to GP pathogens.

Infection, one of the most devastating complications (1) prosthesis removal with or without subsequent
after total joint arthroplasty, is often associated with reimplantation and (2) debridement and implant re-
significant morbidity and increased medical costs [1]. tention using long-term antibiotics. Removal of all
Improvements in operative techniques, aseptic proce- prosthesis components (ie, resection arthroplasty) has
dures, and the use of perioperative antibiotic prophy- a higher chance of eradicating infection but requires
laxis have reduced the risk of prosthetic joint infection extensive surgery and often prolonged immobilization
(PJI) to 1%–2% of patients at most medical centers [3]. Debridement and retention of the prosthesis is an
[2]. However, despite these developments, the increas- attractive alternative, which may be attempted for se-
ing number of joint replacement surgical procedures lected patients to salvage the joint prosthesis [4]. This
being performed has led to an increase in the number less-extensive surgery is thought to be associated with
of cases of infection that need to be treated. a lower probability of procedure-related morbidity, less
Treatment for PJI can be divided into 1 main groups: immobilization, and, consequently, less need for re-
habilitation. The main problem with debridement and
retention, however, is that a substantial number of pa-
Received 9 March 2009; accepted 26 May 2009; electronically published 19
tients will ultimately experience a relapse of infection
August 2009. after this less-aggressive procedure, necessitating ex-
Reprints or correspondence: Dr. Pang-Hsin Hsieh, Dept. of Orthopedic Surgery,
change or resection arthroplasty [3, 4]. The risk of fail-
Chang Gung Memorial Hospital, No. 5, Fu-Hsing St., 333 Kweishian, Taoyuan,
Taiwan ([email protected] or [email protected]). ure for this type of treatment depends on a number of
Clinical Infectious Diseases 2009; 49:1036–43 factors, including the causative organism(s) [5–8].
 2009 by the Infectious Diseases Society of America. All rights reserved.
1058-4838/2009/4907-0008$15.00
Most PJIs are caused by gram-positive (GP) patho-
DOI: 10.1086/605593 gens. Staphylococci (including Staphylococcus aureus

1036 • CID 2009:49 (1 October) • Hsieh et al

and coagulase-negative staphylococci) and streptococci are on examination of periprosthetic tissues, (3) development of a
among the most common organisms, constituting 65%–85% sinus tract, or (4) death related to PJI.
of all isolates. Gram-negative (GN) bacteria, which are less We summarized the collected data (which included patient
commonly associated with PJI, constitute 6%–23% of all epi- demographics, history of orthopedic surgery, comorbidities, pre-
sodes [9–12]. Although GN infections constitute a relatively senting signs of PJI, duration of symptoms, laboratory findings,
minor proportion of all PJIs, they are of significant clinical surgical and medical treatments, and bacteriologic results) at the
importance, because treatment of such infections is considered time of study enrollment. GN and GP PJI data were analyzed
more complicated as a result of the virulence of the organisms, and compared to determine risk factors and prognostic factors
their growing resistance to antimicrobial agents, and the co- predicting outcomes.
morbid conditions of patients [6–8]. Statistical analysis. As appropriate, a x2 analysis or a Fisher
Although there have been numerous reports regarding treat- exact test was used for analyzing categorical data. For numerical
ment of PJI, the comparison between GN and GP PJIs has not, data, an independent t test or the nonparametric Mann-Whit-
to our knowledge, been reported in the English-language lit- ney U test was used for between-group comparisons. The sur-
erature. The purposes of our retrospective comparative study vival rate free of treatment failure was estimated using the

Downloaded from http://cid.oxfordjournals.org/ at University of Western Ontario on April 11, 2015

were, therefore, to evaluate the demographic characteristics and Kaplan-Meier survival method and the log rank test. A P value
outcomes of patients treated for GN PJI and to identify prog- of !.05 was considered to be statistically significant. All statistics
nostic factors that would lead to treatment failure. were 2-sided and were performed using SAS software, version
9.1.3 (SAS).
PATIENTS AND METHODS
RESULTS
Study design. This was a retrospective case cohort study in
which the surgical and medical treatment modalities of the Study population. Fifty-three first-time episodes of GN PJI
patients were not randomized. Decisions of management were occurring among 50 patients from January 1, 2000, through
made by the treating physicians. The patients were observed December 31, 2006, were treated at our institution. This ac-
from the date of diagnosis of PJI until treatment failure, final counted for 53 (15%) of 346 first-time episodes of culture-
clinic visit, death, or loss to follow-up. positive PJI encountered during the study period (table 1).
Study population. Our study included all patients with a Patients with GN PJI were older (median age, 68 vs. 59 years;
diagnosis of PJI of the hip or knee who were treated at our P ! .001) and developed infection earlier after the index joint
institution during the period from January 2000 through De- replacement surgery (median joint age, 74 vs. 109 days; P !
cember 2006. After approval was obtained from the institution- .001) than did patients with GP PJI. Other risk factors for GN
al review board, patients were identified using the electronic PJI that approached statistical significance included fever
database of the hospital, by matching them to the International (P p .09) and the presence of a sinus tract (P p .12). Pa-
Classification of Diseases, Ninth Revision, Clinical Modification, tients with GN PJI did not considerably differ from patients
specific for PJI (code 996.66). Two independent researchers re- with GP PJI with regard to sex, joint location, multiple oper-
viewed and confirmed the medical records. Only cases in which ations, polymicrobial infection, laboratory findings, underly-
the first episode of PJI was treated at our institution were in- ing comorbidity, presence of pus in the joint, bacteremia, or
cluded in the analysis. type of surgical treatment modality.
Definitions. A PJI was diagnosed if the same microorgan- Microbiologic findings. The microbiologic findings of the
ism was recovered from at least 2 joint aspirate or intraopera- 53 episodes of GN PJI are outlined in table 2. A single micro-
tive tissue specimens for culture or if at least one intraoperative organism was isolated in 51 episodes (96%), and multiple mi-
culture was positive for the microorganism, plus if there was croorganisms were isolated in 2 episodes (4%). For patients
evidence of infection at the site of hip or knee prosthesis (ie, with GN PJI, Pseudomonas aeruginosa was the most commonly
presence of a discharging sinus communicating with the joint, isolated pathogen, involved in 21 episodes (19 monomicrobial
operative findings of purulence, or positive laboratory and his- and 2 polymicrobial episodes), followed next by Escherichia coli
topathological test results) [13]. A GN PJI was defined as the in 10 (19%) episodes and then by Klebsiella pneumoniae in 8
presence of at least 1 GN pathogen at the site of infection, (15%) episodes.
regardless of the total number of different strains. Treatment Surgical and medical treatment. The most common initial
failure was defined as the occurrence of the following conditions surgical intervention for the treatment of GN PJI was debride-
at any time after the initial surgical procedure: (1) the presence ment and retention of prosthesis, used in 27 (51%) episodes,
of a PJI due to the original microorganism (relapse of infection) followed by 2-stage exchange in 16 (30%) episodes and resec-
or a different strain (reinfection), (2) surgical findings of pu- tion arthroplasty in 10 (19%) episodes. Debridement typically
rulence around the joint or acute inflammatory histopathology included complete surgical exposure of the joint, removal of

Gram-Negative Prosthetic Joint Infections • CID 2009:49 (1 October) • 1037

 Table 1. Characteristics of 343 Patients with 346 Episodes of Gram-Negative (GN) or Gram-Positive
(GP) Prosthetic Joint Infection (PJI)

GN PJI GP PJI
Characteristic (n p 53 episodes) (n p 293 episodes) P
a
Age, median years (range) 68 (35–85) 59 (32–79) !.001
Male sex 33 (62) 200 (68) .48
Osteoarthritis 29 (55) 177 (60) .43
Total hip arthroplasty 36 (68) 206 (70) .75
a
Prosthesis age, median days (range) 74 (8–296) 109 (6–976) !.001
Multiple operations (⭓2 times) before PJI 20 (38) 94 (32) .42
Polymicrobial PJI 2 (4) 16 (5) .75
Laboratory data, median (range)
C-reactive protein level,b mg/L 39 (9–312) 34 (7–320) .32
White blood cell count, 109 cells/L 1.2 (0.3–2.2) 1.1 (0.4–2.1) .27

Downloaded from http://cid.oxfordjournals.org/ at University of Western Ontario on April 11, 2015

Underlying comorbidity
Diabetes mellitus 21 (40) 107 (37) .67
Rheumatoid arthritis 6 (11) 39 (13) .69
Malignancy 3 (6) 12 (4) .71
Liver cirrhosis 6 (11) 36 (12) .84
Use of steroid 8 (15) 36 (12) .57
Presentation of infection
Discharging sinus 23 (43) 95 (32) .12
Purulent fluid or pus in the joint 17 (32) 108 (37) .50
Fever (temperature, ⭓38.3C) 10 (19) 31 (11) .09
Bacteremia 5 (9) 14 (5) 0.19
Type of surgery
Debridement 27 (51) 127 (43) .31
2-stage exchange 16 (30) 118 (40) .17
Resection arthroplasty 10 (19) 48 (16) .66

NOTE. Data are no. (%) of episodes, unless otherwise indicated. There were 53 episodes of GN PJI that occurred
among 50 patients, and there were 293 episodes of GP PJI that occurred among 293 patients.
a
A P value of !.05 was considered to be statistically significant.
b
There were 9 episodes with missing data.

the inflamed soft tissues and bone, and debridement of pros- therapy of ciprofloxacin, sulfamethoxazole, and trimethoprim
thetic-bone interfaces. Implantation of antibiotic-loaded ce- for 3 patients.
ment beads within the joint was performed at the time of Outcomes and prognostic factors. The 2-year survival rate
removal of the prosthesis and in 6 (22%) of the 27 episodes free of treatment failure for patients with GN PJI was 27%
treated with debridement and retention. (95% confidence interval [CI], 16%–34%) for debridement and
All patients with GN PJI received appropriate intravenous retention, 87% (95% CI, 80%–99%) for 2-stage exchange, and
antimicrobial therapy that was based on the antimicrobial sus- 69% (95% CI, 59%–84%) for resection arthroplasty (figure 1).
ceptibility pattern of the bacteria isolated from the intraoper- Patients with GN PJI who underwent debridement alone were
ative cultures (table 3). The median duration of effective par- significantly more likely to experience treatment failure than
enteral antibiotic therapy was 38 days (range, 24–52 days) for those who underwent 2-stage exchange (P ! .001) and resection
GN PJI episodes treated with debridement and retention of the arthroplasty (P p .008).
components, 23 days (range, 6–36 days) for 2-stage exchange, The outcomes of patients treated for GN PJI with 2-stage
and 17 days (range, 10–39 days) for episodes treated with re- exchange and resection arthroplasty were similar to those of
section arthroplasty. Patients with GN PJI who received de- patients treated for GP PJI with similar surgical modalities: the
bridement and retention were given additional oral antibiotic 2-year estimates of the survival rate free of treatment failure
suppression for a median of 49 days (range, 28–92 days). Var- for patients with GP PJI were 94% (95% CI, 86%–100%;
ious regimens of oral antibiotic treatment were prescribed, in- P p .39) for 2-stage exchange and 78% (95% CI, 70%–91%;
cluding ciprofloxacin for 15 patients, clindamycin for 2 pa- P p .30) for resection arthroplasty. However, treating GN PJI
tients, a combination of both for 7 patients, and a combined with debridement and retention of components was associat-

1038 • CID 2009:49 (1 October) • Hsieh et al

Table 2. Microbiologic Findings of 53 Episodes of Gram-Neg- patients), underlying malignancy (6 patients), myocardial in-
ative (GN) Prosthetic Joint Infection (PJI) Occurring among Pa- farction (2 patients), and a traffic accident (1 patient).
tients Treated during 2000–2006
DISCUSSION
GN PJI
Infecting microorganism(s) (n p 53 episodes) The number of total joint replacement surgical procedures is
Monomicrobial increasing rapidly. It is estimated that, from 2005 to 2030, the
Pseudomonas aeruginosa 19 (36) number of primary total hip arthroplasties in the United States
Escherichia coli 10 (19) will increase by 174%, to 572,000 procedures per year, and new
Klebsiella pneumoniae 8 (15) cases of total knee replacement will increase by 673%, to 3.48
Enterobacter cloacae 3 (6) million procedures per year [14]. Although the rate of infection
Acinetobacter baumannii 2 (4)
after prosthetic joint replacement remains relatively constant,
Salmonella enterica 2 (4)
these large projected increases in demand for hip and knee
Hemophilus influenzae 2 (4)
replacements imply that the demand for treating patients with
Proteus mirabilis 1 (2)

Downloaded from http://cid.oxfordjournals.org/ at University of Western Ontario on April 11, 2015

Bacteroides fragilis 1 (2)
PJI is likely to increase. Although many authors have reported
Unidentified 3 (6) on the treatment of PJI of the hip or knee [4, 6, 9, 11, 12],
Polymicrobial most of them dealt with infections resulting from a mixed group
P. aeruginosa, methicillin-resistant S. of microorganisms, with GP bacteria being the most common
aureus 1 (2) pathogens involved in these studies.
P. aeruginosa, coagulase-negative Staphy- A number of difficulties occur during the treatment of bone
lococcus species 1 (2)
and joint infections caused by GN bacteria. These difficulties
NOTE. Data are no. (%) of episodes. Because of rounding, percentages occur primarily affect immunocompromised patients and are
may not add up to 100%.
associated with treatment failure [6–8]; only a limited number
of experimental models have been described [15], and ran-
domized, controlled clinical trials are hampered by the fact that
ed with a lower 2-year cumulative survival rate free of treatment
most institutions do not have a sufficient number of patients
failure than treating GP PJI with debridement and retention of
for such studies. We sought to report the demographic char-
components (27% vs. 47% of episodes were successfully treated;
P p .002).
The potential risk factors leading to the differences between Table 3. Intravenous Antimicrobial Therapy Used for 53 Epi-
sodes of Gram-Negative (GN) Prosthetic Joint Infection (PJI) Oc-
the subgroups of patients receiving debridement and retention
curring among Patients Treated during 2000–2006
are outlined in table 4. No risk factor was identified in terms
of age, sex, joint location, duration of symptoms prior to sur- GN PJI
gery, multiple operations before PJI, polymicrobial infection, Type of antimicrobial therapy, drug(s) (n p 53 episodes)
laboratory findings, durations of intravenous and oral antibi- Monotherapy
otic therapies, single antimicrobial medication, use of local an- Ciprofloxacin 4 (8)
tibiotic beads, underlying comorbidity, or presentation of PJI. Clindamycin 3 (6)
Imipenem 3 (6)
Of the 27 episodes of GN PJI occurring among 27 patients
Metronidazole 2 (4)
treated with debridement and retention, treatment failure was
Ceftriaxone 2 (4)
observed in 20 (74%) episodes. A total of 21 (78%) patients had Aztreonam 1 (2)
their serum C-reactive protein levels decrease to half their pre- Amikacin 1 (2)
treatment levels during antimicrobial therapy. However, of these Gentamicin 1 (2)
21 episodes, only 7 (33%) showed no evidence of recurrent in- Piperacillin 1 (2)
fection, and 14 (67%) ultimately involved treatment failure dur- Combination therapy
ing follow-up. A shorter duration of symptoms before treatment Ceftazidime and ciprofloxacin 10 (19)
Cefepime and gentamicin 9 (17)
was associated with successful debridement in episodes of GN
Ceftazidime and amikacin 6 (11)
PJI (median duration of symptoms, 11 vs. 5 days; P p .02) (table
Ceftriaxone and clindamycin 5 (9)
5). No other risk factor for treatment failure was identified. Ceftriaxone and metronidazole 3 (16)
Mortality. Twenty-six patients had died during the course Vancomycin and aztreonam 1 (2)
of writing this paper. Seven (5 patients with GP PJI and 2 with Vancomycin and ciprofloxacin 1 (2)
GN PJI) deaths were related to the infection. Other causes of NOTE. Data are no. (%) of episodes. Because of rounding, percentages
death included complications secondary to liver cirrhosis (10 may not add to 100%.

Gram-Negative Prosthetic Joint Infections • CID 2009:49 (1 October) • 1039

 Downloaded from http://cid.oxfordjournals.org/ at University of Western Ontario on April 11, 2015
Figure 1. Survival rate free of treatment failure in 346 episodes of prosthetic joint infection (PJI) treated with debridement, 2-stage exchange, or
resection arthroplasty during the period from 2000 through 2006. GN, gram-negative; GP, gram-positive.

acteristics and treatment outcomes of patients with GN PJI and decisions (eg, regarding the selection of a broader empiric an-
the prognostic factors associated with GN PJI. Data provided timicrobial coverage after the surgery). These data might serve
by the present study may help treating physicians identify pa- as a guide for predicting the outcome of such infections.
tients at risk of acquiring GN infection and then make informed Our data suggest that GN PJIs represent a substantial pro-

Table 4. Comparison of Episodes of Gram-Negative (GN) and Gram-Positive (GP) Prosthetic Joint
Infection (PJI) Occurring among Patients Who Underwent Debridement and Retention of the Prosthesis

GN PJI GP PJI
Variable (n p 27 episodes) (n p 127 episodes) P
Age, median years (range) 63 (36–80) 59 (35–73) .17
Male sex 16 (59) 84 (66) .50
Total hip arthroplasty 17 (63) 84 (66) .75
Duration of symptoms, median days (range) 8 (2–19) 10 (3–35) .34
Multiple operations (⭓2 times) before PJI 11 (41) 36 (28) .20
Polymicrobial PJI 0 (0) 4 (3) .35
Laboratory data, median (range)
C-reactive protein level,a mg/L 34 (9–134) 39 (11–216) .44
White blood cell count, 109 cells/L 1.1 (0.5–1.6) 1.1 (0.6–1.4) .38
Intravenous antibiotic therapy, median days
(range) 38 (24–52) 37 (12–54) .12
Oral antibiotic suppression, median days
(range) 49 (28–92) 58 (21–112) .22
Local antibiotic therapy 6 (22) 37 (29) .47
Underlying comorbidity
Diabetes mellitus 11 (41) 48 (38) .77
Rheumatoid arthritis 4 (15) 17 (13) .77
Malignancy 2 (7) 4 (3) .28
Liver cirrhosis 3 (11) 6 (5) .19
Use of steroid 5 (19) 16 (13) .54
Presentation of infection
Discharging sinus 6 (22) 35 (28) .57
Purulent fluid or pus in the joint 8 (30) 39 (31) .91
Fever (temperature, ⭓38.3C) 3 (11) 16 (13) .83
Bacteremia 2 (7) 8 (6) .63

NOTE. Data are no. (%) of episodes, unless otherwise indicated.

a
There were 2 episodes with missing data.

1040 • CID 2009:49 (1 October) • Hsieh et al

 Table 5. Data on 27 Episodes of Gram-Negative Prosthetic Joint Infection Occurring among Patients
Who Underwent Debridement and Retention of the Prosthesis

Successful
debridement Persistent infection
Variable (n p 7 episodes) (n p 20 episodes) P
Age, median years (range) 62 (47–79) 63.5 (37–80) .59
Male sex 4 (57) 12 (60) .89
Total hip arthroplasty 2 (29) 8 (40) .68
Prosthesis age, median days (range) 59 (14–123) 73 (8–121) .53
a
Duration of symptoms, median days
(range) 5 (3–8) 11 (2–19) .02
Multiple operations (⭓2 times) before PJI 3 (43) 8 (40) NS
Polymicrobial PJI 0 (0) 0 (0) NA
Laboratory data, median (range)

Downloaded from http://cid.oxfordjournals.org/ at University of Western Ontario on April 11, 2015

C-reactive protein level, mg/L 31 (9–134) 36 (15–111) .65
White blood cell count, 109 cells/L 1.1 (0.5–1.6) 1.1 (0.6–1.4) .38
Intravenous antibiotic therapy
Duration, median days (range) 46 (29–52) 37 (24–44) .10
Monotherapy 1 (14) 6 (30) .63
Oral antibiotic suppression
Duration, median days (range) 45 (28–87) 53 (36–92) .51
Monotherapy 6 (86) 11 (55) .2
Local antibiotic therapy 1 (14) 5 (25) .66
Underlying comorbidity
Diabetes mellitus 4 (57) 7 (35) .39
Rheumatoid arthritis 1 (14) 3 (15) NS
Malignancy 0 (0) 2 (10) .60
Liver cirrhosis 2 (29) 1 (5) .16
Use of steroid 2 (29) 3 (15) .58
Presentation of infection
Discharging sinus 0 (0) 6 (30) .15
Purulent fluid or pus in the joint 2 (29) 6 (30) NS
Fever (temperature, ⭓38.3C) 1 (14) 2 (10) NS
Bacteremia 1 (14) 1 (5) .46

NOTE. Data are no. (%) of episodes, unless otherwise indicated. NA, not applicable; NS, not significant.
a
A P value of !.05 was considered to be statistically significant.

portion of all occurrences of PJI. The proportion of episodes compared the subgroups of patients with GN and GP PJI who
(53 [15%] of 346 episodes of PJI) that occurred in the pres- underwent debridement surgery on the basis of a number of
ent study is in conformity with those reported in the literature factors. However, with the numbers of patients available, no
[9–12]. We found that GN PJIs tend to occur more frequent- factors could be identified in terms of patients’ demographic
ly among elderly patients and during the early postoperative characteristics, surgical and medical treatment history, com-
period. Although not statistically significant, patients with GN orbidity, laboratory data, or presentation of infection. This find-
PJI are more likely to have fever and develop a discharging si- ing mirrored the fact that the development of PJI as a result
nus. We also found that patient with GN PJI who were treated of a GN pathogen, in itself, might be an important factor con-
with 2-stage exchange and resection arthroplasty had a favor- tributing to the poor outcome of debridement and retention.
able outcome comparable to that of patients with GP PJI who In an analysis of 99 episodes of PJI among patients who
are treated with similar surgical modalities. However, in our underwent debridement and retention, Marculescu et al [4] did
study, the attempt to retain the prosthesis with debridement not find GN infection to be associated with treatment failure.
alone was associated with a significantly less favorable outcome One explanation for this might be that only a small number
for patients with GN PJI than for patients with GP PJI. of episodes of GN PJI (ie, 6 [6%] of 99 episodes) were includ-
In an attempt to find the reason for the worse outcome of ed in their study; thus, their study lacked sufficient power to
debridement and retention associated with GN PJI, we have detect a difference.

Gram-Negative Prosthetic Joint Infections • CID 2009:49 (1 October) • 1041

 In choosing between retention and removal of the prosthesis given the small number of successful debridements for pa-
for the management of PJI, clinicians and patients are often tients with GN PJI, we were unable to determine whether the
forced to face a trade-off between short-term surgical morbid- presence of a sinus tract was a significantly poor prognostic fac-
ity and mortality and long-term relapse of infection. Patients tor (P p .15). Nevertheless, it is important to note that none
sometimes ask to retain the prosthesis, under the misappre- of the GN PJI episodes that were managed successfully involved
hension that a failed attempt at debridement causes little harm. patients who had a sinus tract.
This is a common clinical scenario. In fact, a failed attempt at Our study has several limitations. First, this is a retrospective
debridement and retention of the prosthesis can result in the study with all the potential drawbacks implicit in such a study
kind of morbidity associated with an orthopedic surgical pro- design. Our institution is a tertiary care referral center with an
cedure. It also consumes valuable healthcare resources, because established protocol for the treatment of PJI. Furthermore, in
further surgery is often required. We are not aware of any data all cases, decisions were made in consultation with the infection
from prospective clinical trials comparing prosthesis retention specialists. Nonetheless, there remains the potential for uncon-
and removal for the treatment of PJI. Fisman et al [3], using trolled selection bias among treating physicians. In addition,
a Markov model to compare the cost-effectiveness of debride- the number of GN PJIs included during the study period was

Downloaded from http://cid.oxfordjournals.org/ at University of Western Ontario on April 11, 2015

ment and staged exchange in simulated patients, concluded that relatively small, and thus our study may have lacked enough
debridement and retention would provide greater quality-ad- power to detect slight differences among subsets of patients.
justed life expectancy gains than would 2-stage exchange ar- Finally, the quality of debridement among patients is likely to
throplasty among patients with infected hip after arthroplasty. affect the outcome. However, it was very difficult to evaluate
However, this model was based on published studies in which the quality of surgery, and thus we did not specifically address
GP microorganisms were the most common pathogens. The this variable in the present study.
authors found that their results were most sensitive to the an- In summary, GN PJI accounts for a substantial percentage
nual rate of relapse after debridement. Thus, their conclusion of all PJI episodes. We believe that the choice of antimicrobial
might not be applied properly to the episodes of GN PJI re- therapy for PJI should take into account the possibility of GN
ported in our study, which resulted in a significantly higher infection, especially in suspected patients who are older and
rate of treatment failure. who develop PJI early in the postoperative period. In com-
Despite the dismal figure, debridement and retention of parison with PJI due to GP pathogens, GN PJI treated with 2-
prosthesis was successfully attempted in 7 (26%) of 27 epi- stage exchange and resection arthroplasty is associated with a
sodes of GN PJI. A short duration of symptoms before sur- comparatively favorable outcome. However, debridement and
gery was the only identified factor contributing to the success retention for GN PJI carries a higher rate of treatment fail-
of treatment. In fact, in all of these episodes, the surgery was ure and should be used sparingly among patients within the
performed within 8 days after the onset of symptoms. These first few days after the onset of symptoms. Further data from
randomized clinical trials are required to substantiate these
observations are consistent with outcome data from other co-
findings.
horts of patients infected mainly by GP pathogens. Brandt et
al [16], in a study of 33 staphylococcal PJIs, reported that
prostheses that were debrided more than 2 days after onset of Acknowledgments
symptoms were associated with a higher probability of treat- We thank Zoe Chen for her assistance with statistical analysis.
ment failure than were those debrided within 2 days of onset. Potential conflicts of interest. All authors: no conflicts.
For 34 patient in a retrospective study by Tattevin et al [12],
there were 13 patients (38%) who had a significantly short References
interval (!5 days) from onset of symptoms to debridement and 1. Sculco TP. The economic impact of infected joint arthroplasty. Or-
were successfully treated, and there were 21 (62%) patients who thopedics 1995; 18:871–3.
had a much longer median interval (54 days) and experienced 2. Huo MH, Gilbert NF, Parvizi J. What’s new in total hip arthroplasty.
J Bone Joint Surg Am 2007; 89:1874–85.
a failed debridement. Marculescu et al [4] reported that a du- 3. Fisman DN, Reilly DT, Karchmer AW, Goldie SJ. Clinical effectiveness
ration of symptoms of ⭓8 days was associated with a greater and cost-effectiveness of 2 management strategies for infected total hip
risk of treatment failure, higher by a factor of 2, after debri- arthroplasty in the elderly. Clin Infect Dis 2001; 32: 419–30.
4. Marculescu CE, Berbari EF, Hanssen AD, et al. Outcome of prosthetic
dement and prosthesis retention. On the basis of the results of
joint infections treated with debridement and retention of components.
previous studies and the present study, we emphasize the im- Clin Infect Dis 2006; 42:471–8.
portance of expedient debridement when attempting to sal- 5. Moran E, Masters S, Berendt AR, et al. Guiding empirical antibiotic
vage the prosthesis. therapy in orthopaedics: the microbiology of prosthetic joint infection
managed by debridement, irrigation and prosthesis retention. J Infect
The presence of a sinus tract has been associated with treat- 2007; 55:1–7.
ment failure in previous reports [4, 17]. In our current study, 6. McDonald DJ, Fitzgerald RH Jr, Ilstrup DM. Two-stage reconstruction

1042 • CID 2009:49 (1 October) • Hsieh et al

 of a total hip arthroplasty because of infection. J Bone Joint Surg Am 12. Tattevin P, Cremieux AC, Pottier P, Huten D, Carbon C. Prosthetic
1989; 71:828–34. joint infection: when can prosthesis salvage be considered? Clin Infect
7. Legout L, Senneville E, Stern R, et al. Treatment of bone and joint Dis 1999; 29:292–5.
infections caused by gram-negative bacilli with a cefepime-fluoroquin- 13. Mirra JM, Amstutz HC, Matos M, Gold R. The pathology of the joint
olone combination. Clin Microbiol Infect 2006; 12:1030–3. tissues and its clinical relevance in prosthesis failure. Clin Orthop Relat
8. Schurman DJ, Wheeler R. Gram negative bone and joint infection: Res 1976; 117:221–40.
sixty patients treated with amikacin. Clin Orthop Relat Res 1978; 134: 14. Kurtz S, Ong K, Lau E, Mowat F, Halpern M. Projections of primary
268–74. and revision hip and knee arthroplasty in the United States from 2005
9. Zimmerli W, Ochsner PE. Management of infection associated with to 2030. J Bone Joint Surg Am 2007; 89:780–5.
prosthetic joints. Infection 2003; 31:99–108.
15. Norden CW, Shinners E. Ciprofloxacin as therapy for experimental os-
10. Kaandorp CJ, Dinant HJ, van de Laar MA, Moens HJ, Prins AP, Dijk-
teomyelitis caused by Pseudomonas aeruginosa. J Infect Dis 1985; 151:
mans BA. Incidence and sources of native and prosthetic joint infec-
291–4.
tion: a community based prospective survey. Ann Rheum Dis 1997;
56:470–5. 16. Brandt CM, Sistrunk WW, Duffy MC, et al. Staphylococcus aureus
11. Hsieh PH, Shih CH, Chang YH, Lee MS, Shih HN, Yang WE. Two- prosthetic joint infection treated with debridement and prosthesis re-
stage revision hip arthroplasty for infection: comparison between the tention. Clin Infect Dis 1997; 24:914–9.
interim use of antibiotic-loaded cement beads and a spacer prosthesis. 17. Burger RR, Basch T, Hopson CN. Implant salvage in infected total
J Bone Joint Surg Am 2004; 86:1989–97. knee arthroplasty. Clin Orthop Relat Res 1991; 273:105–12.

Downloaded from http://cid.oxfordjournals.org/ at University of Western Ontario on April 11, 2015

Gram-Negative Prosthetic Joint Infections • CID 2009:49 (1 October) • 1043