Journal of Infection (2010) 61, 125e132

www.elsevierhealth.com/journals/jinf

Six weeks of antibiotic treatment is sufficientfollowing surgery for septic arthroplasty*

Louis Bernard a,d, Laurence Legout a, Line Zurcher-Pfund a, Richard Stern a,Peter Rohner b, Robin Peter a, Mathieu Assal a, Daniel Lew c,Pierre Hoffmeyer a, Ilker Uckay a,c,*

aOrthopaedic Surgery Service, Geneva University Hospitals & Medical School, Geneva, Switzerlandb Laboratory of Bacteriology, Geneva University Hospitals & Medical School, Geneva, Switzerlandc Service of Infectious Diseases, Geneva University Hospitals & Medical School, Geneva, SwitzerlanddDivision of Infectious Diseases, Bretionneau Hospital, University Hospitals of Tours, France

Accepted 29 May 2010Available online 9 June 2010

KEYWORDSOral antibiotic;Duration;Prosthetic joint;Infection;Surgery

* There are no grants, financial suppinterest. All authors state that they haation for publication by another journDays of Infectiology, in Nice 2005 and* Corresponding author. Orthopaedic

Gentil, 1211 Geneva 14, Switzerland.E-mail address: ilker.uckay@hcuge

0163-4453/$36 ª 2010 The British Infedoi:10.1016/j.jinf.2010.05.005

Summary Objectives: In the treatment of prosthetic joint infections (PJI), the benefit of an-tibiotic therapy for more than 6 weeks after surgery is uncertain. We compared PJI cure ratesaccording to the duration of antibiotics, 6 versus 12 weeks.Methods: A prospective observational non-randomized study in Geneva University Hospitals1996e2007.Results: A total of 144 PJI (62 hip arthroplasties, 62 knee arthroplasties, and 20 hip hemiar-throplasties) were included with a prolonged follow-up ranging from 26 to 65 months. Surgicaltreatment included 60 debridements with implant retention, 10 one-stage exchanges of theprosthesis, 57 two-stage exchanges, and 17 Girdlestone procedures or knee arthrodeses. Se-venty episodes (49%) received 6 weeks antibiotic therapy and 74 episodes, 12 weeks. Curewas achieved in 115 episodes (80%). Cure rate did not change according to the duration of in-travenous antibiotics (>8 days, 8e21 days, >21 days) (KruskaleWallis-test; p Z 0.37). In mul-tivariate analysis, none of the following parameters was statistically significantly associatedwith cure: two-stage exchange (odds ratio 1.1,95%CI 0.2e4.8); number of debridements(0.9, 0.4e1.9); six weeks antibiotherapy (2.7, 0.96e8.3); duration of intravenous course(1.0, 0.96e1.03); sinus tract (0.6, 0.2e1.7); or MRSA infection (0.5, 0.2e1.5), although implantretention showed a tendency for less cure (0.3, 0.1e1.1).

ort, financial interests, consultancy, commercial or other associations that could lead to a conflict ofve read and approved the manuscript. It has not been published elsewhere nor is it under consider-al. Different parts of the manuscript have been presented orally at the 6th and 10th French NationalLyon 2009, respectively, and the Swiss Orthopaedic Congress in Geneva, June 2009.Surgery Service, Service of Infectious Diseases, Geneva University Hospitals, 4, Rue Gabrielle Perret-Tel.: þ41 22 372 3311; fax: þ41 22 372 3987..ch (I. Uckay).

ction Society. Published by Elsevier Ltd. All rights reserved.

126 L. Bernard et al.

Conclusions: Following surgery for treatmentof PJI, antibiotic therapyappears able tobe limitedto a 6-week course, with oneweek of intravenous administration. This approach needs confirma-tion in randomized trials.ª 2010 The British Infection Society. Published by Elsevier Ltd. All rights reserved.

Introduction

Although the infection risk following primary hip and kneearthroplasties is less than 2%,1,2 prosthetic joint infections(PJI) have become more frequent due to the increasingnumber of patients undergoing surgery. Each PJI episoderepresents substantial morbidity,1 with a 5.3e7.2-foldhigher cost than for the initial arthroplasty.3

The optimal surgical and antibiotic treatment of PJI isunclear. The foreign body nature of the implant makes theseinfections difficult to treat, even for pathogens not usuallyconsidered virulent.1,4e6 Expert opinions5 have been summa-rized in algorithms.1 Many recommend antibiotic therapy for3 months for hip arthroplasties (6 months for total knee ar-throplasties) for PJI with arthroplasty retention, and 6weeksfor two-stage exchanges.1,5e9 According to the landmark re-view of Zimmerli et al. in 2004, re-implantation during a two-stage exchange can be performed after a short interval ofonly 2e4 weeks for not-complicated pathogens, with conse-quent extension of antibiotic treatment beyond 6 weeks.1

Sometimes, excessive antibiotic treatment durations areperformed that could become systemic in different thera-peutic schools all over the world, because recommendationsare not evidence-based sensu strictu and only based on ex-perts’ opinion.

While the usual treatment is by the parenteral route forthe first 2e4 weeks,1,2,6 this attitude is not evidence-basedeither. For economic reasons, as well as patient and nursecomfort, parenteral administration should be kept to a min-imum.10 Up to one-third of patients with PJI may experi-ence antibiotic-related or catheter-related problemsduring parenteral treatment.11

Our study evaluates the risk factors for failure in thetreatment of PJI, with emphasis on surgical interventionand duration of concomitant parenteral and oral antibiotictreatment.

Methods

Geneva University Hospitals is a tertiary center. The Ortho-paedic Service has a dedicated Infectious Disease (ID)specialist12 and a prospective cohort for elective arthroplas-ties (primary and revisions) with active post-discharge sur-veillance.13 This cohort has been approved by the hospitalEthics Committee (No. 08-017). The follow-up consists onregular clinical examination, anamnesis and occasionallylaboratory or radiological controls. The follow-up durationis infinite and patients are censored only in case of death, de-nial to participate in the study or lost to follow-up.

Study population, design and definitions

This prospective observational non-randomized study in-cluded all PJI of our cohort from January 1996 to December

2007. PJI episodes of patients referred from other centerswere excluded, since they were not prospectively followedup after treatment. Three ID specialists collected 37 vari-ables for each PJI including data about patient co-morbid-ities, type of arthroplasty, infection, surgery and antibiotictherapy.

The definition of PJI required the documentation ofintra-articular pus around the infected arthroplasty, and inaddition one of the following conditions: two bacterialcultures of joint aspirates or intraoperative specimensyielding the same pathogen (more than three identicalintraoperative specimens if the pathogen was a skin com-mensal such as coagulase-negative Staphylococcus or Pro-pionibacterium acnes, presence of a sinus tract, orradiological signs of infection such as loosening. Microbio-logical procedures were unchanged during the study periodand based upon Clinical and Laboratory Standards Instituterecommendations.14

PJI were classified as early (infection less than 3 monthsafter arthroplasty), delayed (3e12 months) and late (morethan 12 months).1 Fungal or mycobacterial infections wereexcluded. Cure was defined as the absence of clinical, radio-logical and biological signs of infection in the area of thearthroplasty after a minimum follow-up of 24 months post-surgery. A treatment failure was defined as a re-apparitionof infection according to the criteria of PJI cited above.

Surgical treatment

Surgical options to treat PJI included retention of theprosthesis with debridement alone, one-or two-stage re-placement, and Girdlestone resection arthroplasty of thehip or arthrodesis of the knee.5 Debridements were per-formed with open arthrotomy. Re-implantation of prosthe-sis was performed when there were no clinical, radiologicaland laboratory signs of infection for a minimum of 14 dayswithout antimicrobial treatment.5 When cement was used,it contained gentamicin.

Antibiotic treatment

Empirical antimicrobial treatment was begun immediatelyafter intraoperative samples were obtained, and wasfurther adapted according to microbiological results. Spe-cific antibiotic therapy was continued for 6 or 12 weeksaccording to the recommendations of the orthopaedicsurgeon (mostly 12 weeks) or the ID consultant (mostly 6weeks). In detail, there were several factors influencingthe decision of the total antibiotic duration. For instancea patient with private insurance was more often managedby a chief surgeon, whereas patients with standard in-surances were more often managed by surgeon assistantsand the ID physician. Second, there were several surgeonsand one ID physician at any given period. In the absence ofthe only ID physician (ID rounds, vacations, congresses),

Six weeks’ antibiotic for PJI 127

the initial decision was more often made by surgeons (andnot reversed thereafter by the returning ID physician).Third, the decision for 12 and 6 weeks’ course was alsoinfluenced by removal or retention of the implant; the lat-ter by the co-morbidities, local tissue damage, the patho-gen, the presence of a sinus tract, or the duration ofsigns or symptoms.1 The duration of intravenous versusoral administration depended upon local conditions duringtreatment and organizational issues of case managementsuch as the lack of venous access, recurrent phlebitis, dif-ficulties mobilizing the patient postoperatively, early pa-tient discharge or intolerance to oral medication.

Statistical analysis

The primary outcome parameter was overall cure of PJI;the secondary was the duration of intravenous antibiotictreatment. Due to several important confounding parame-ters during treatment we further stratified cure incidencesinto clinical important subgroups and also analyzed themseparately, including: cure according to 6 vs. 12 weeks ofantibiotic treatment, implant retention vs. implant ex-change, parenteral antibiotics alone vs. oral antibioticsalone, and early vs. late PJI. While the total duration ofantibiotic treatment was binary (6 or 12 weeks), theparenteral part of the antibiotic therapy varied. Age andthe duration of parenteral antibiotic therapy were com-puted as continuous and categorical variables. For all groupcomparisons, we used the Pearson c2-test, Fisher exact-test, Wilcoxon rank sum-test or the KruskaleWallis-test,as appropriate.

In a second step a logistic regression was adjusted forcase-mix. Variables with a p value � 0.2 in univariate anal-ysis were included in a stepwise forward selection processfor multivariate analysis. Key variables were checked forcollinearity and interaction, the latter by ManteleHaenszelestimates. The number of variables in the final model waslimited to the ratio of 1 variable to 8e10 outcome events.15

In a third step, we repeated the regression analysis sepa-rately within the aforementioned strata, always controllingfor cure and the duration of concomitant antibiotic treat-ment. P values � 0.05 (all two-tailed) were significant.STATA� software (9.0, STATA Corp, College Station, USA)was used.

Results

Patients’ demographics

A total of 144 PJI episodes in 135 patients were included inthe study. Seventy-five episodes (75/144, 52%) occurred inwomen, median age was 77 years (interquartile range (IQR),67e82 years). In 62 episodes (62/144, 43%) patients wereimmunocompromised. There were 29 patients with diabetesmellitus, 16 with chronic alcoholism, and 17 with carcinoma,organ transplantation, steroid therapy for autoimmunedisease, andnon-alcoholic cirrhosis CHILDB or C. Themedianfollow-up was 36 months after treatment (IQR, 26e65months). At the end of the study period, 39 patients (39/135, 29%) had died, three secondary to their PJI.

Arthroplasties

There were 62 infected total hip arthroplasties (20/62cemented), 20 infected hip hemiarthroplasties (all ce-mented), and 62 infected total knee arthroplasties (52/62cemented).

Infection

Forty-two PJI occurred early, 31 delayed and 71 occurredlate, with a median delay of 12 months (IQR, 1e36 months)between insertion and infection. In 33 episodes (33/144,23%) the prosthesis was loose and a sinus tract was presentin 35 episodes (24%). PJI was bacteremic in 34 cases (34/144, 24%) and polymicrobial in 14 episodes (10%). Gram-positive pathogens predominated: Staphylococcus aureus(n Z 57; 26/57 hospital-acquired methicillin-resistantstrains (MRSA)), and coagulase-negative Staphylococci(n Z 27). The most frequent Gram-negative pathogen wasEscherichia coli (n Z 6), followed by Pseudomonas aerugi-nosa (n Z 3).

Antibiotic treatment

All PJI were treated with antibiotics for a total of 6 weeksin 70 episodes (70/144, 49%) and 12 weeks in 74 episodes(51%). The patients who were treated for 6 weeks wereequal with those treated for 12 weeks, except that those inthe 12 week group had a longer duration of parenteraltherapy and those in the 6 week group more frequentlyunderwent a two-stage replacement (Table 1). Overall, themedian duration of intravenous administration was 14 days(IQR, 8e42 days); the parenteral part of the entire antibi-otic therapy was during 0e7 days for 44 episodes, 8e21days for 35 episodes, and for more than 21 days for 65PJI episodes. The four most frequently used antibioticswere rifampicin (n Z 58, always in combination and onlyfor staphylococcal infections), ciprofloxacin (n Z 42), van-comycin (n Z 40), and amoxicillin/clavulanic acid(n Z 25). Cotrimoxazole and clindamycin were used in11 and 14 PJI, respectively. Combination therapy was ad-ministered in 80 PJI (80/144, 56%). Among 84 staphylococ-cal PJI, 46 (55%) were treated with a rifampin-basedcombination.

Surgical treatment

All patients underwent surgical intervention for curativepurposes, including one-stage exchange (n Z 10), two-stage exchange (n Z 57) with a median delay of 14 weeks(IQR, 7e21 weeks) between stages, implant retention withdebridement (n Z 60), and Girdlestone procedure or kneearthrodesis (n Z 17). The mean number of surgical de-bridements per patient was 2 (range, 1e6 interventions).Debridement and prosthesis retention were performed sig-nificantly more in cases of early infection and in thosewithout radiological evidence of prosthesis loosening(Pearson c2-tests; p < 0.01). The patient populations ofimplant retention and removal were almost perfectly equal(Table 2).

Table 1 Group comparison of 6 vs. 12 week’s antimicrobial therapy.

Six weeksn Z 70

Twelve weeksn Z 74

p value(two-tailed)

Patients characteristicsFemale sex 38 (51%) 37 (53%)Median age 76 years 78 yearsDiabetes mellitus 14 (20%) 15 (20%)Chronic alcoholism 8 (11%) 8 (11%)Active neurologic disease 11 (16%) 11 (15%)Other immunosuppression 8 (11%) 9 (12%)

Initial arthroplastyHip joint 40 (57%) 42 (58%)Due to osteoarthritis 43 (61%) 52 (70%)Due to fracture 17 (24%) 14 (19%)

Median no. of prior surgeries before infection 1 surgery 1 surgery

InfectionMedian time delay arthroplastyeinfection 11 weeks 12 weeksMedian delay symptomatic diseaseesurgery 1 week 1 weekEarly infection (�3 months) 22 (31%) 21 (28%)Late infection (>12 months) 35 (50%) 36 (49%)Median total leukocyte count before antibiotics 7 G/L 6 G/LMedian C-reactive protein level before antibiotics 76 mg/L 100 mg/LRadiological loosening of implant 24 (38%) 19 (28%)Sinus tract 17 (24%) 18 (24%)Positive blood cultures 16 (23%) 18 (25%)Infection with a single pathogen 65 (93%) 65 (88%)Infection due to methicillin-susceptible S. aureus 16 (23%) 15 (20%)Infection due to methicillin-resistant S. aureus 11 (16%) 15 (20%)Infection due to coagulase-negative staphylococci 15 (21%) 12 (16%)Infection due to S. agalactiae 2 (3%) 9 (12%)Infection due to other Streptococcus spp 11 (16%) 9 (12%)Infection due to E. coli 1 (1%) 5 (6%)Infection due to P. aeruginosa 2 (3%) 1 (1%)Infection due to other Gram-negative organisms 4 (6%) 7 (9%)Co-infection with anaerobes 6 (9%) 3 (4%)

TreatmentMedian no of surgical interventions for cure 1 1No. of episodes treated with retention and debridement 20 (29%) 40 (54%) 0.01a

One-stage exchange 4 (6%) 6 (8%)Two-stage exchange 36 (51%) 20 (27%) <0.01a

Median delay between the exchange surgeries 13 weeks 16 weeksGirdlestone or arthrodesis 10 (14%) 8 (11%)

Intravenous antibiotic treatment only 24 (35%) 27 (37%)Median duration of intravenous treatment 10 days 15 days <0.01b

Combination antibiotic treatment 41 (59%) 39 (53%)Use of penicillin G 5 (7%) 8 (11%)Use of amoxicillin 13 (19%) 12 (16%)Use of ceftriaxone 7 (10%) 9 (12%)Use of ciprofloxacin 21 (30%) 21 (28%)Use of ofloxacin 6 (9%) 4 (5%)Use of levofloxacin 1 (1%) 3 (4%)Use of clindamycin 7 (10%) 7 (9%)Use of rifampicin 32 (44%) 27 (37%)Use of cotrimoxazole 5 (7%) 6 (8%)Use of vancomycin 19 (27%) 21 (28%)

128 L. Bernard et al.

Table 1 (continued )

Six weeksn Z 70

Twelve weeksn Z 74

p value(two-tailed)

OutcomeMedian time delay begin of treatmentefailure 3 weeks 3 weeks

Persistence of infection 6 (85%) 18 (82%)New infection 1 (14%) 5 (23%)

Death of all causes during follow-up 15 (21%) 24 (32%)Death due to prosthetic joint infection 1 (1%) 2 (3%)

Percentages indicate the proportion of the variables inside the six or twelve week groups. Only significant p values � 0.05 are displayed.a c2-test.b Wilcoxon rank sum-test.

Six weeks’ antibiotic for PJI 129

Cure

Overall cure

A total of 115 PJI (115/144, 80%) were cured. Among the 29treatment failures, 24 were due to persistent infection and5 to a new infection by another pathogen. Recurrenceswere encountered after a median of 3 months post-surgery(IQR, 2e4 months).

Table 2 Comparison of subgroups with removal and retention

Remova

Patients characteristicsFemale sex 45 (54%Median age 76 yearDiabetes mellitus 17 (20%Chronic alcoholism 11 (13%Other immunosuppression 12 (14%Active neurologic disease 10 (12%

Initial arthroplastyHip joint 45 (54%

Due to arthritis 56 (67%Due to fracture 14 (17%

Cemented arthroplasty 51 (61%

InfectionEarly infection (�3 months) 11 (13%Late infection (>12 months) 56 (67%Sinus tract 17 (20%Positive blood cultures 21 (26%Infection with a single pathogen 79 (94%Infection due to methicillin-susceptible S. aureus 15 (18%Infection due to methicillin-resistant S. aureus 14 (17%Infection due to coagulase-negative staphylococci 15 (18%Infection due to P. aeruginosa 1 (1%)

TreatmentSix weeks of antibiotic treatment 50 (60%Intravenous antibiotic treatment only 29 (35%

Median duration of intravenous treatment 10 days

Percentages indicate the proportion of the variables inside the remdisplayed.a c2-test.b Wilcoxon rank sum-test.

Stratified group comparisons

Role of parenteral medicationSeven days of intravenous administration had the same curerate thanmore than 3 weeks of parenteral antibiotics (37/44vs. 50/65,c2-test,pZ 0.36; (odds ratio 1.6, 95%CI 0.5e5.0)).The duration of intravenous antibiotics (>8 days, 8e21 days,>21days) didnot changecure rateswhencomparedbetweeneach other (KruskaleWallis-test; pZ 0.37).

of the arthroplasty.

l n Z 84 Retention n Z 60 p value (two-tailed)

) 30 (50%)s 78 years) 12 (20%)) 5 (8%)) 5 (8%)) 12 (20%)

) 38 (63%)) 39 (65%)) 17 (28%)) 39 (68%)

) 32 (53%) <0.01a

) 15 (25%) <0.01a

) 18 (30%)) 13 (22%)) 51 (85%)) 16 (26%)) 12 (20%)) 12 (20%)

2 (3%)

) 20 (33%) <0.01a

) 22 (37%)15 days 0.04b

oval or retention groups. Only significant p values � 0.05 are

130 L. Bernard et al.

Parenteral versus oral antibiotic medicationForty-nine PJI (49/144, 34%) were treated only with in-travenous antibiotics for the entire period. Conversely, sixepisodes were treated with only oral antibiotics from thebeginning of therapy. Their outcome did not differ in termsof cure (37/49 vs. 5/6, Fisher exact-test, p Z 1 (odds ratio0.6, 95%CI 0.1e6.4)). When stratifying the comparison ofonly oral vs. only parenteral treatment for the patientswith a total six weeks’ course of antibiotic therapy, therewas no difference in the cure rate either (3/4 vs. 20/21,Fisher exact-test, p Z 0.30 (odds ratio 0.2, 95%CI0.1e15.7)).

Other detailed group comparisonsDetailed analyses are summarized in Table 3. Among the 60PJI treated with implant retention and debridement alone,20 (20/60, 33%) received 6 weeks of antibiotic treatment.Eighteen episodes of them (18/20, 90%) resulted in cure.This incidence was equal to the patient population who un-derwent six weeks of antibiotic treatment but with implantremoval (18/20 vs. 18/20, Fisher exact-test, pZ 1 (odds ra-tio 1.0, 95%CI 0.1e15.2)).

Adjustments in multivariate analysis

Overall logistic regressionIn multivariate analysis (Table 4), there was no single vari-able which was formally significantly associated with over-all cure: Retention with debridement (odds ratio 0.3, 95%CI0.1e1.1); two-stage exchange (OR 1.1, 0.2e4.8); number ofsurgical interventions (OR 0.9, 0.4e1.9); or total durationof intravenous antibiotics (OR 1.0, 0.98e1.02). However,there were also strong tendencies in favor (six weeks’ anti-biotic treatment 2.7, 0.96e7.8) or in disfavor of cure (im-plant retention 0.3, 0.1e1.1). The patient group with sixweeks’ antibiotic course had significantly more a two-stageexchange or the implant removed than the 12 weeks’ group(Table 1). Both variables showed significant interaction(ManteleHaenszel estimate, p Z 0.015).

Logistic regression stratifiedTo better adjust for case-mix, regression analysis for curewas repeated stratified according to the following groups:(a) PJI with retention and debridement vs. removal; (b) PJIwith two-stage replacement vs. the rest of the studypopulation; (c) Early onset PJI vs. the rest of the popula-tion; and (d) Late onset PJI vs. the rest of the population.

In univariate analysis for subgroup “a,” the followingvariables were significantly associated with cure: age, timedelay between onset of symptoms and debridement, six

Table 3 Cure incidences stratified according key parameters (F

Parenteral antibiotic treatment For � 8 da37/44

Removal vs. retention of arthroplasty Removal75/84

Time of onset of infection Early infec38/42

weeks of antibiotic treatment, and duration of intravenousantibiotics. In multivariate analysis, a six week course ofantibiotherapy (OR 7.0, 1.2e44.2) remained associatedwith enhanced cure, while MRSA infection (OR 0.3,0.1e1.2), total parenteral duration of antibiotics (OR 1.0,0.97e1.03), a two to three weeks’ IV course (OR 1.8,0.3e9.7), a more than three weeks’ IV course (OR 1.0,0.1e25.6), or total number of surgical interventions (OR0.8, 0.2e2.6) were not.

For subgroup “b,” no single variable was associated withcure in univariate or multivariate analysis.

For subgroup “c”, a six week course of antibioticsshowed tendency for cure (OR 4.3, 0.4e47.3) while theduration of intravenous antibiotics (OR 1.0, 0.99e1.02) didnot.

This constellation was the same for subgroup “d”: sixweek course (OR 1.5, 0.3e7.6), total parenteral antibiotics(OR 1.0, 0.98e1.01), a more than three weeks’ IV course(OR 1.6, 0.1e46.4), and number of surgical interventions(OR 0.8, 0.2e3.0).

Discussion

We report our experience with PJI in our 12-year pro-spective arthroplasty cohort with a very long post-dischargesurveillance protocol. Our 80% cure incidence comparesfavorably to incidences published in the literature1,5,8,16,17

or is even higher.18e20

Besides a difference in the proportion of removedimplants, the groups of 6 weeks vs. 12 weeks were equalin terms of patient characteristics, arthroplasty type,inflammatory markers, fever, bacteremia, implant loosen-ing, pathogens, sinus tract, type of surgical treatment, andchoice of antibiotic agents. But because of the large case-mix and the bias introduced by the fact that 6 weeks’patients were more frequently managed by two-stagereplacement, we performed overall and stratified multivar-iate regression analyses for factors revealing differences incrude group comparisons.

Albeit not formally significant, implant retention re-vealed as a risk factor for recurrent PJI in the overallanalysis. This was not the same, when analyses wereperformed for the therapeutic strata of two-stage ex-changes or late infections, for which implants were re-moved. Here, total post-surgical antibiotic treatment for 6weeks was equivalent to 12 weeks in terms of cure. Finally,the 12 week’s course of total antibiotic treatment nevershowed a tendency to increased cure rates, independentlyof the stratification of the analysis. Hence, we conclude,that 6 week’s of antibiotic therapy following implant

isher exact or c2 e tests).

p value

ys For � 21 days50/65 0.47Retention40/60 <0.01

tion Late infection56/71 0.13

Table 4 Logistic regression with outcome “overall cure of prosthetic joint infection”.

Associations Univariate analysis Multivariate analysis

Odds ratio with 95%confidence intervals

Odds ratio with 95%confidence intervals

Patient characteristicsFemale sex 1.4 (0.6e3.2) 1.3 (0.5e3.4)Age 1.0 (0.9e1.0) 1.0 (0.9e1.0)Other co-morbidities 1

Diabetes mellitus 1.0 (0.3e2.6)Chronic alcoholism 1.9 (0.4e8.7)Other immunosuppression 1.2 (0.3e4.5)

ArthroplastyCemented arthroplasty 0.6 (0.3e1.5)

InfectionEarly infection (<3 months) 2.4 (0.8e6.7)All pathogens 1

Methicillin-resistant S. aureus 0.4 (0.2e1.0) 0.5 (0.2e1.5)Monomicrobial infection 1

Polymicrobial infection 1.0 (0.3e3.5)Negative blood cultures 1

Positive blood cultures 0.7 (0.3e1.8)No sinus tract 1

Presence of sinus tract 0.5 (0.2e1.3) 0.6 (0.2e1.7)

TreatmentNo. of surgeries for infection 1.3 (0.7e2.4) 0.9 (0.4e1.9)Implant removal 1

Two-stage exchange 3.8 (1.4e10.7) 1.1 (0.3e4.8)One-stage exchange 1.0 (0.2e5.0)Retention with debridement 0.2 (0.1e0.6) 0.3 (0.1e1.1)

Twelve weeks of antibiotic therapy 1Six weeks of antibiotic therapy 3.8 (1.5e9.6) 2.7 (0.9e8.3)

Intravenous & oral antibiotic therapyIntravenous antibiotics only 0.7 (0.3e1.7)

Duration of intravenous therapy (days) 1.0 (0.99e1.01) 1.0 (0.99e1.01)Less than 8 days of intravenous therapy 0.8 (0.2e2.4)8e21 days of intravenous therapya 0.6 (0.2e1.7) 1.3 (0.4e4.5)

Antibiotic monotherapy 1Combination therapy 1.4 (0.6e3.3)

No rifampicin use (staphylococcal infections) 1Use of rifampicin 1.4 (0.6e3.2)

a Compared to less than 8 days of intravenous antibiotic therapy.

Six weeks’ antibiotic for PJI 131

removal are sufficient for the treatment of PJI and confirmtwo recent studies: Farhad et al. indicated that six weeksantibiotic therapy seems to be sufficient for all orthopaedicinfections; those with implants included21 and Byren et al.showed that the duration of antibiotic therapy did notpredict treatment failure, but that prolongation simplypostponed, rather than prevented failure.17

A second message is in the feasibility of oral antibiotictreatment for PJI, independently if the implant was re-moved or retained. PJI treated with exclusively parenteralantibiotics alone had similar cure incidences as thosetreated by oral antibiotics from the start. Equally, theduration of intravenous part as categorized variables didnot show and significant differences in terms of cure inmultivariate analysis or when compared between eachother. The total duration of parenteral antibiotic therapy

interpreted as a continuous variable in showed odds ratiosof 1.0 in univariate or multivariate analyses with a verynarrow confidence interval, suggesting that the duration ofthe intravenous part does not seem to play a significant rolein terms of cure. This was equally statistically true whencomparing only the patients with only a six weeks course ofantibiotic treatment.

The current routine of parenteral antibiotics for the first2e4 weeks of treatment for practically all severe osteo-articular infections1,2,5,6 remains difficult to challenge dueto the lack of prospective randomized trials. Withoutdoubt, bone penetration during parenteral administrationis good,22 but recent data suggest that an early switch tooral antibiotics is as effective as prolonged parenteral reg-imens.23 For the treatment of bone infections there aresome antibiotics that have already proven to be effective

132 L. Bernard et al.

in oral form. Quinolones, linezolid (alone or combined withrifampin2,5,6,24 or clindamycin in monotherapy,2,5e7,24,25

have oral bio-availability of more than 90%.25 Byren et al.demonstrated that in treating PJI an intravenous courseof antibiotics for longer than 4 weeks did not enhancecure.17 Zimmerli et al. summarized observational studiesthat showed the same failure rates of PJI treatment despitea prolonged (4e6 weeks) period of intravenous treatment.2

Even in cases of diabetic foot osteomyelitis, there are nodata indicating the superiority of any particular route of de-livery of systemic antibiotics.26 Finally, literature even sup-ports the option of oral antibiotic regimens alone in right-sided endocarditis.27

Our study has several limitations: (a) It is a small samplesize not-randomized retrospective study, thus limiting theability to generalize the findings; (b) The influence ofeventual small-colony variants was not investigated28; (c)All antibiotic treatment in our study was set in a binary dis-tribution to either exactly 6 or 12 weeks, and thus we areunable to estimate the effect of other durations such as 4or 8 weeks.

In conclusion, following surgery for treatment of PJI,antibiotic therapy might be able to be limited to a 6-weekcourse, with only a few days of intravenous administration.This approach needs confirmation in randomized trials.

Acknowledgements

We are indebted to Christophe Barea and Dr Anne Lubbekefor retrieving data. We thank to the teams of the Ortho-paedic Service and the Laboratory of Bacteriology for theirsupport.

References

1. Zimmerli W, Trampuz A, Ochsner PE. Prosthetic-joint infec-tions. N Eng J Med 2004;351:1645e54.

2. Zimmerli W, Widmer AF, Blatter M, Frei R, Ochsner PE. Role ofrifampin for treatment of orthopedic implant-related staphylo-coccal infections: a randomized controlled trial. JAMA 1998;279:1537e41.

3. Bengtson S. Prosthetic osteomyelitis with special reference tothe knee: risks, treatment and costs. Ann Med 1993;25:523e9.

4. Uckay I, Pittet D, Vaudaux P, Sax H, Lew D, Waldvogel F. For-eign body infections due to Staphylococcus epidermidis. AnnMed 2009;41:109e19.

5. Bernard L, Hoffmeyer P, Assal M, Vaudaux P, Schrenzel J,Lew D. Trends in the treatment of orthopaedic prosthetic in-fections. J Antimicrob Chemother 2004;53:127e9.

6. Trampuz A, Zimmerli W. Antimicrobial agents in orthopaedicsurgery: Prophylaxis and treatment. Drugs 2006;66:1089e105.

7. Kosters K, van Crevel R, Sturm PD, Willem Schreurs B, de WaalMalefijt MC, van Kampen A, et al. Treatment of knee prosthesisinfections: evaluation of 15 patients over a 5-year period. IntOrthop 2008;33:1249e54.

8. Laffer RR, Graber P, Ochsner PE, Zimmerli W. Outcome of pros-thetic knee-associated infection: evaluation of 40 consecutiveepisodes at a single centre. Clin Microbiol Infect 2006;12:433e9.

9. Volin SJ, Hinrichs SH, Garvin KL. Two-stage reimplantation oftotal joint infections: a comparison of resistant and non-resis-tant organisms. Clin Orthop Relat Res 2004;427:94e100.

10. Mermel LA, Farr BM, Sherertz RJ, Raad II, O’Grady N, Harris JS,et al.Guidelines for themanagementof intravascularcatheter-re-lated infections. Infect Control Hosp Epidemiol 2001;22:222e42.

11. Pulcini C, Couadau T, Bernard E, Lorthat-Jacob A, Bauer T,Cua E, et al. Adverse effects of parenteral antimicrobial ther-apy for chronic bone infections. Eur J Clin Microbiol Infect Dis2008;27:1227e32.

12. Uckay I, Vernaz-Hegi N, Harbarth S, Stern R, Legout L,Vauthey L, et al. Activity and impact on antibiotic use andcosts of a dedicated infectious diseases consultant on a septicorthopaedic unit. J Infect 2009;58:205e12.

13. Lubbeke A, Stern R, Garavaglia G, Zurcher L, Hoffmeyer P. Dif-ferences in outcomes of obese women and men undergoing pri-mary total hip arthroplasty. Arthritis Rheum 2007;57:327e34.

14. Wayne P. Performance standards for antimicrobial susceptibil-ity testing. Clinical and Laboratory Standards Institute; 2007.17th informational supplement (Standard M100-S17).

15. Vittinghoff E, McCulloch CE. Relaxing the rule of ten events pervariable in logistic and Cox regression. Am J Epidemiol 2007;165:710e8.

16. Kilgus DJ, Howe DJ, Strang A. Results of periprosthetic hip andknee infections caused by resistant bacteria. Clin Orthop RelatRes 2002;404:116e24.

17. Byren I, Bejon P, Atkins BL, Angus B, Masters S, McLardy-Smith P, et al. One hundred and twelve infected arthroplastiestreated with ’DAIR’ (debridement, antibiotics and implant re-tention): antibiotic duration and outcome. J Antimicrob Che-mother 2009;63:1264e71.

18. Betsch BY, Eggli S, Siebenrock KA, Tauber MG, Muhlemann K.Treatment of joint prosthesis infection in accordance with cur-rent recommendations improves outcome. Clin Infect Dis 2008;46:1221e6.

19. Marculescu CE, Berbari EF, Hanssen AD, Steckelberg JM,Harmsen SW, Mandrekar JN, et al. Outcome of prosthetic jointinfections treated with debridement and retention of compo-nents. Clin Infect Dis 2006;42:471e8.

20. Brandt CM, Sistrunk WW, Duffy MC, Hanssen AD,Steckelberg JM, Ilstrup DM, et al. Staphylococcus aureus pros-thetic joint infection treated with debridement and prosthesisretention. Clin Infect Dis 1997;24:914e9.

21. Farhad R, Roger PM, Albert C, Pelligri C, Touati C,Dellamonica P, et al. Six weeks antibiotic therapy for allbone infections: results of a cohort study. Eur J Clin MicrobiolInfect Dis 2009;29:217e22.

22. Smilack JD, Flittie WH, Williams Jr TW. Bone concentrations ofantimicrobial agents after parenteral administration. Antimi-crob Agents Chemother 1976;9:169e71.

23. Daver NG, Shelburne SA, Atmar RL, Giordano TP, Stager CE,Reitman CA, et al. Oral step-down therapy is comparable to in-travenous therapy for Staphylococcus aureus osteomyelitis. JInfect 2007;54:539e44.

24. Darley ES, MacGowan AP. Antibiotic treatment of gram-positivebone and joint infections. J Antimicrob Chemother 2004;53:928e35.

25. Toma MB, Smith KM, Martin CA, Rapp RP. Pharmacokinetic con-siderations in the treatment of methicillin-resistant Staphylo-coccus aureus osteomyelitis. Orthopedics 2006;29:497e501.

26. Berendt AR, Peters EJ, Bakker K, Embil JM, Eneroth M,Hinchliffe RJ, et al. Specific guidelines for treatment of diabeticfoot osteomyelitis. Diabetes Metab Res Rev 2008;24:190e1.

27. Heldman AW, Hartert TV, Ray SC, Daoud EG, Kowalski TE,Pompili VJ, et al. Oral antibiotic treatment of right-sidedstaphylococcal endocarditis in injection drug users: prospec-tive randomized comparison with parenteral therapy. Am JMed 1996;101:68e76.

28. Vaudaux P, Kelley WL, Lew DP. Staphylococcus aureus smallcolony variants: difficult to diagnose and difficult to treat.Clin Infect Dis 2006;43:968e70.