Two-Stage Primary Arthroplasty in the Infected Native Knee: A Systematic Review and Pooled Analysis

Abstract

Introduction

The knee is the commonest native joint to develop an infection. A two-stage primary knee replacement, with an interim stage of debridement and cement spacer application, modelled after two-stage revision for periprosthetic joint infections (PJI) has been reported for the management of chronic infections.

Aims

To systematically review the literature to find the infection-free survival and outcomes of this operation and explore its indications.

Methods

PRISMA guidelines were followed for this review. A systematic search of 4 online databases was conducted on 9/8/2020. After reviewing 226 abstracts and applying our selection criteria, 10 papers were selected for full-text review, and 9 included in the final synthesis.

Results

On pooled analysis, an infection-free survival of 95.6% (CI 94.7–96.4) was found at 2 years in 139 knees, which was unchanged over the remainder of the follow-up (Mean 3.9 years). The complication rate after final implantation was 6% in those that did not develop reinfection. The mean pooled Knee Society Score (KSS) and KSS Function score among 70 patients (4 papers) was 83.4 (80.1–89.0) and 76.8 (71.5–78.0), respectively. The mean range of motion among 82 patients (6 papers) was more than 100°.

Conclusions

Two-stage primary knee replacement is a safe, effective and reliable procedure with good results in the short to medium term. Further studies are required to lay down precise indications and cost-effectiveness of this procedure, in comparison to other strategies for chronic infection. All joint registries should develop methods to identify patients undergoing two-stage procedures, to understand their long-term survival and outcomes.

Introduction

The knee is the commonest joint to develop septic arthritis (SA) (54%) [1]. The annual incidence is reported to be 6–10 cases per 100,000 population [2]. A vast majority can be treated successfully by arthroscopic washouts and antibiotics, with a recent review reporting a 97% cure rate, though half the patients required repeated irrigations [3]. In those who fail to respond, the infection can fester, leading to a downward spiral of joint destruction, pain, stiffness and poor function. Many require further surgery within the year in the form of arthrodesis, amputations and arthroplasty [4]. In these patients, a two-stage Primary Arthroplasty modelled on the same principles as the two-stage revision arthroplasty for prosthetic joint infections (PJI) has been attempted successfully, though there are sparse reports in the literature. We conducted this systematic review to find the outcomes of 2-stage primary arthroplasty in knee infections, and to explore its indications.

Methods

The study was conducted in accordance with PRISMA guidelines [5]. A study protocol was made but not registered on a database.

Research question What are the outcomes of, and indications for two-stage primary knee replacements for infection?

Primary outcome Infection-free survival after two-stage primary knee replacement.

Secondary Outcomes

• Outcomes using standardized outcome scores

• Indications/Criteria used for selecting 2-stage replacement

Time frame We did not keep any limits and accepted all papers with electronic prints of the full-text available.

Study selection criteria These criteria were decided before undertaking the literature search.

Inclusion criteria: All papers reporting 2-stage primary knee replacement.

• In English

• Reporting more than 1 patient

• Reporting a minimum follow-up of at least 2 years

• Implanting cement spacers at the 1st stage

Exclusion criteria: papers reporting exclusively on cases.

• With previous joint replacement in the same knee

• Of single-stage replacement for infection

The criterion of minimum follow-up of 2 years was used as literature regarding two-stage revisions for PJI shows that 50–100% of all reinfections occurred within 2 years of final implantation in many series. We presumed a similar likelihood in two-stage primaries for infection and aimed to minimise bias in the pooled infection-free survival arising from a shorter follow-up [6–11].

Search Strategy AM conducted a systematic search of four online databases (Medline, CINAHL, EMBASE, PubMed) on 9th August 2020. The search strategy for Medline is presented in Table 1. AM, HS and IP then independently screened the titles and abstracts. Any differences were resolved by consultation among all authors. The full text of selected articles was reviewed, and their references were also screened to find any additional articles not captured in the searches.

Table 1.

Search strategy using medline

String	Search phrase	No. of papers
1	(two-stage).ti,ab	26,986
2	(2-stage).ti,ab	4867
3	(reconstruction).ti,ab	202,757
4	(arthroplasty).ti,ab	62,280
5	(joint ADJ replacement).ti,ab	6760
6	(1 OR 2)	31,467
7	(3 OR 4 OR 5)	266,545
8	(septic arthritis).ti,ab	6198
9	(infected ADJ2 joint).ti,ab	337
10	(infected ADJ2 knee).ti,ab	596
11	(8 OR 9 OR 10)	7017
12	(6 AND 7)	2899
13	(11 AND 12)	252
14	(revision).ti	22,088
15	(exchange).ti	51,229
16	(periprosthetic).ti	3283
17	(14 OR 15 OR 16)	76,277
18	13 NOT 17	159

Data extraction A standard data capture form was used to extract demographic details, follow-up duration, indications and pre-operative workup, microbiological profile, details of previous treatment for infection, technical details of each stage of the procedure, interim management between stages, associated antibiotic treatment regime and finally, outcomes and complications. The methodological quality and risk of bias were evaluated independently by AM and HS using the Modified Coleman Methodology Score (MCMS), described by Cowan et al. [12]. The score ranges from 0 to 100 (85–100, excellent; 70–84, good; 55–69, fair; < 55, poor). Any discrepancies were addressed in consultation with SK.

Results (Tables 2, 3, 4, 5)

Table 2.

Overview of included papers and demographics

	Supreeth et al. [18]	Kunze et al. [14]	Seo et al. [17]	Xu et al. [15]	Yi et al. [20]	Shaikh et al. [22]	Bauer et al. [16]	Kirpalani et al. [21]	Nazarian et al. [19]
Year	2020	2020	2020	2019	2015	2014	2010	2005	2003
Country	Oman	USA	Korea	China	China	Korea	France	Korea	USA
Type of study	Retrospective Cohort	Retrospective cohort	Retrospective cohort	Retrospective cohort	Retrospective cohort	Retrospective cohort	Retrospective cohort	Retrospective cohort	Retrospective cohort
No. of patients	4	30	21 (14)a	19	16	13	17	5	14
Age (mean/ range)	62.75 (56–66)	57.4	70.2 (57–83)a	59.8	63.7	65 (39–81)	57 (31–82)b	71.8 (63–77)	62 (45–68)
Gender (m/f)	All males	19 (63.3%)	8/14 (57.1%)a	6 (32%)	6/17 (35%)c	5 (36%)	NM	0 (all F)	9 (64%)
Diabetes/ other immunocompromise	2 DM, 1 BA	NM; CCI: 2.1d	NM	NM; 6 ASA 3 +	NM	4 DM, 1 BA, 1 hypopituitarism	NM	5 DM	2 DM, 2 RA
Modified coleman methodology scoree	42	49	45	45	45	45	45	45	45

DM diabetes mellitus; BA bronchial asthma; CCI Charlson Comorbidity Index; ASA American Society of Anaesthesiologists grade; NM not mentioned

^aParameters reported for 14 patients, 7 further patients underwent the procedure after failing washout- details not provided

^bReported for all knees, including those undergoing single-stage replacement for healed infection

^cIncluding 1 patient who did not have 2nd stage surgery

^dIncludes 12 hips

^eThe score ranges from 0 to 100 (85–100, excellent; 70–84, good; 55–69, fair; < 55, poor)

Table 3.

Details of knee infection

	Supreeth et al. [18]	Kunze et al. [14]	Seo et al. [17]	Xu et al. [15]	Yi et al. [20]	Shaikh et al. [22]	Bauer et al. [16]	Kirpalani et al. [21]	Nazarian et al. [19]
Past interventions
Previous surgery	2	2	NM	NM	6	2	NM	0	6
Previous trauma	1	2	NM	8 (Open trauma)	2	2 (ACL tears)	NM	0	0
Intra-articular injections	2	NM	NM	5	8	NM	6	4	NM
Infection and treatment
Duration of infection (at 1st stage)	6 (3–13) m	50 (± 109) d	NM	NM	NM	11 (7–14) d for 4; 7 (3–18) m for 9	NM	6.6 (4–11) weeks	2.1 y (3 w-7..8 y)
Surgery for infection	NM	NM	7 (of 21)	NM	NM	7 (of 9)	17	5	7
osteomyelitis	1	2	NM	NM	NM	4	NM	0	3
Microbiological profile
Culture negative	0	11	7	7	6	7	Excluded	2	6
Staph aureus (mrsa)	1	10 (6)	3	4	3 (1)	3 (3)	10	3 (2)	3
Coagulase neg staph	1	10	2	0	4	0	8	0	2
Others	2	6	1	8	2	1	13	0	3
Fungal/mycobacterial	0	0	1	0	0	2 (1)a	0	0	0
Unknown	0	5b	0	0	2c	0	Excluded	0	0
Total	4	37d	14	19	15	13	31e	5	14
Source of specimen	A + T1f	T1g	Ah	A	A	A	A	A	A

d days; w weeks; m months; NM Not mentioned, asp aspirated, ACL Anterior Cruciate Ligament, Staph Staphylococci, MRSA Methicillin-Resistant S aureus

^a1 patient had concomitant Tuberculosis and Candida infection

^bReported not found in patient notes, excluded from pool

^cNot aspirated pre-op; excluded from pool

^dIncludes Hips

^eIncludes single-stage knee replacements for healed infections

^fPre-op aspirate and tissue from 1st stage yielded the same organisms

^gReported Results from tissue collected at the 1st stage only

^hA = Aspirate (pre-operative); 1st stage tissue results not reported

Table 4.

Details of the procedure

	Supreeth et al. [18]	Kunze et al. [14]	Seo et al. [17]	Xu et al. [15]	Yi et al. [20]	Shaikh et al. [22]	Bauer et al. [16]	Kirpalani et al. [21]	Nazarian et al. [19]
First stage and interim period
Spacer type	Static	27 Articulating 3 Static	Articulating	Articulating	Articulating	Articulating	Not clear	Static + Articulatinga	Static
Antibiotic duration; w	9.1 (5–13.3) of which IV 3.2	6.6 (± 1.8) of which IV 6 min	IV 6 f/b Oral in some	6 min b	IV 1 f/b oral c	(4–12) of which IV 2 min	13.2 (6.4–25.7)b,d	NM	IV 6 f/b oralc
Total duration with spacer(s); m	4.3 (3–7.2)	2.6 (± 1.3)	6 (3–18)	4.9b	3.9 (2.3–6.2)	5.6 (2–29)m	1.4 (0.9–3.7)b	1.7 (1.4–1.8)	(1.4–8.2)
need for repeat debridement; new spacer	0	4	NM	0	1e	1	NM	0	1
Final implantation
Implant used/ need for constraint	PS-1; CCK-1; RHK-2 f	PS-3; CCK-27	PS; CCK if ‘felt necessary’	NM	PS in all	Used CCK in all	NM	NM	PS; 4 had CCK
Post-op antibiotics	Till cultures negative; 3d	NM	7d	Till cultures neg	IV 3d + 4 w Oral	IV 2w f/b oral till lab results normal; 4–12 w	7.2 w g	NM	IV 5 d; f/b 6 m oral. 6 lifelong

NM not mentioned, min minimum, d days, w weeks, m months, f/b followed by, PS posterior stabilised, CCK constrained condylar knee, RHK rotating-hinge knee

^aNot explicitly mentioned, radiographs in the report show both types of spacer

^bIncluding hips

^cContinued till final implantation

^dIncludes antibiotic administration post 2nd stage

^eFailed 3 spacer applications and proceeded to arthrodesis

^f1 knee required augments in addition to RHK

^gEstimated from mean duration after 1st stage and mean duration of spacer

Table 5.

Results

	Supreeth et al. [18]	Kunze et al. [14]	Seo et al. [17]	Xu et al. [15]	Yi et al. [20]	Shaikh et al. [22]	Bauer et al. [16]	Kirpalani et al. [21]	Nazarian et al. [19]
Follow up duration (years)	2.9 (2.2–3.4)	3.3 (2–10)	3.2a	4.7 (2.2–10.8)b	3.8 (2–8)	4 (2–7)	5 (2–13)c	3.1 (2.4–3.8)	4.5 (2–7)
outcome scores; mean post-op (mean pre-op)	OKS 41 (23)	KSS 80.1 (35.9) KSSF 71.5 (38.0)	KOOS- Data not given	NM	HSS 83.9 (37)	KSS 85 (41) KSSF 83 (43) WOMAC 18 (51)	KSS 83 KSSF 80	HSS Pain 82.8 Function 73.2d	KSS 89 (46) KSSF 78
Post-op rom (degrees)	Arc 112.5	Arc 100.5	Data not presented	NM	E 1.6 (0–5) F 107.5 (95–125)	Arc 115	NM	E 5 (5–5) F 104 (90–120)	E < 5 F > 110 in 10
Infection free survivorship at 2 years	4/4 (100%)	28/30 (93.3%)	21/21 (100%)	17/19 (89%)	16/16 (100%)	13/13 (100%)	15/17 (88%)	5/5 (100%)	14/14 (100%)
complications	NM	3 arthrofibrosis requiring MUA; 2 Reopse,f	NM	NM separatelyg	NM	NM	NM	1 HO	3 Reopsh

OKS Oxford Knee Score; KSS Knee Society Score; KOOS Knee Injury and Osteoarthritis Outcome Score; HSS Hospital for Special surgery Score; ROM Range of Motion; E Extension; F Flexion; NM Not mentioned

^aEstimated from mean from 1st stage 3.6 (3–5.3) years and spacer duration- reported for 14 knees

^bIncluding hips

^cIncluding single-stage Knees and Hips

^dHave reported scores as such and not a single figure as described in the standard/modified HSS scoring systems

^eReoperations: 1 lateral release and poly exchange for patellar instability; 1 arthroscopic adhesiolysis for arthrofibrosis that failed MUA

^fAlso report 2 complications of 1st stage procedure-1 arthrofibrosis requiring MUA and 1 spacer dislocation

^gComplications of both Hip and Knee reported together

^hReoperations: 1 lateral release for patellar subluxation; 1 hematoma evacuation; 1 skin grafting for wound problems

Included Papers

A total of 226 articles were obtained for a title and abstract review after deduplication. Of these, we found 10 which reported on our procedure of interest. After applying our selection criteria, 1 article was rejected as it had a smaller duration of follow-up [13] and 9 were deemed suitable for synthesis (Fig. 1). Of these, 3 reported on Hips as well [14–16] with one [16] including single-stage operations for infections considered healed. 1 paper presented a comparison with patients undergoing arthroscopic washout [17].

Fig. 1.

Prisma flow chart

Description of Studies Included

All included studies are retrospective cohort studies. The mean MCMS was 45.1 (42–49), and all papers were graded as poor.

Demographic Parameters

The studies include 139 (4–30) patients. The average age estimated across the series is 62.7 (31–83) years. In the 9 studies [14, 15, 17–22] reporting gender distribution, women constitute 53.4% (62/116) of all patients.

Systemic Comorbidities

More than a third (13/34) of the patients are diabetic among four studies [18, 19, 21, 22] which note this, with one reporting it in all patients [21]. A further 17% (5/30) were on long term steroids [18, 19, 22]. Xu et al. report significant comorbidities (ASA 3 or higher) in a third of their patients [15].

Pre-existing Knee Problems

All authors have proceeded to a two stage only if there were advanced radiographic/arthroscopic features of degenerative joint conditions. Knee injections have been noted in 5 series [15, 16, 18, 20, 21] as the source of septic arthritis (34%–25/74), which suggests that many patients had a background of Osteoarthritis being managed non-operatively. Previous surgeries to the knee are noted by 6 authors [14, 15, 18–20, 22] of which 3 [15, 19, 20] mention it in more than 40% of patients. Surgery for previous injuries is mentioned by 4 authors [14, 18, 20, 22] in 7 cases, including 2 who had ACL reconstruction. 1 paper [15] has reported infections as a result of open fractures, and this accounts for all 8 of their post-operative cases (Table 5).

Duration and Nature of Joint Infection

The average duration of infection at the time of the first stage procedure in 62 patients across 5 papers [14, 18, 19, 21, 22] which note this is 37.7 weeks (Range 3 weeks to 7.8 year) [19]. Thus, a two-stage procedure appears to be generally selected for long-standing infections. However, it was also performed in more acute settings, with one of the papers [22] reporting 4 patients who have undergone the first stage after 1–2 weeks of the start of infection. Yi et al. [20] also seem to have performed two-stage surgery for patients presenting with acute septic arthritis, though they have not provided the duration of infection to confirm this.

Most of the patients had at least 1 procedure for joint lavage and debridement, either open or arthroscopic, in the 4 papers which note this. One paper [17] reports that a majority of patients (14/21) did not have an arthroscopic washout before their first stage, though it is not clear if this was attempted during previous treatment. 3 patients in Kirpalani’s [21] series had 2 washouts. 1 patient in Shaikh’s series had ACL graft removal in addition to synovectomy and lavage. Surgery for removal of implants is not mentioned by any papers dealing with infection secondary to trauma or its treatment.

2 papers [21, 22] note the duration of antibiotic treatment and report at least 4 weeks in patients with chronic (pyogenic) infection, but this varied up to 20 weeks. Finally, 3 papers [14, 19, 21] have treated 9 patients with clinical and/or imaging evidence of underlying osteomyelitis, which includes 4 patients with a draining sinus in one of the papers [21].

Workup and Decision for Staged Joint Replacement

All papers have performed the operation in patients with imaging/arthroscopic evidence of advanced degenerative arthritis. In one of the papers [20], it seems the procedure was performed in patients presenting with the first episode of SA without attempting conventional treatment with washouts and antibiotics. Other authors have reserved the procedure for those where conventional treatment failed to control infections [16, 21] recurrent infections and chronic infections [17, 19], with some using the terms ‘evolutive’ [15, 16] and ‘recalcitrant’ infection [21]. However, only 5 have presented clear criteria. Seo et al. have performed arthroscopic washouts in patients who fit their criteria but refused two-stage surgery [17]. One of the papers [22] mentions three cases of elective primary TKR with intra-operative detection of infection which they treated with two-stage procedures.

Lastly, 4 papers [14, 18, 19, 22] mention evidence of osteomyelitis (OM) either clinically or on imaging in 10 patients (7.1%). 3 of these papers mention performing advanced imaging studies like MRI scans [18, 22] and Bone scans [19] to evaluate patients with OM, but they have not used them routinely for others.

Microbiology

Of papers where data could be extracted [15, 17–22], 41.7% (35/84) patients were culture negative before the 1st stage. The commonest organisms found were Staph aureus (24%; of which 31% MRSA) and Coagulase negative Staphylococci (10%). 3 patients with Fungal infection (Candida) were reported in 2 papers [17, 22] with 1 having concomitant Tuberculosis. It is not clear in any of these studies what the source of the material was (aspiration or tissue samples).

Two papers [14, 18] have reported the microbiology from tissue samples taken at spacer implantation and the pooled culture-negative rate is 26.8% (11/41). However, one of the papers [14] includes both hips and knees. Further, comparison with the yield from pre-op aspirates was not statistically significant.

Only one author [18] has reported both pre-operative and intra-operative (1st stage) culture results (4 patients) and found no change in the isolated organism.

First Stage (Debridement and Spacer Application)

While two papers [18, 19] have used static spacers alone, articulating spacers have been used commonly in other studies. Overall, among papers which clearly state it [14, 15, 17–20, 22], 85% (96/113) patients received articulating spacers. All except one [16] have mixed antibiotics with the cement, in some cases adding to premixed antibiotics. One author [22] has added an anti-tubercular (streptomycin) as a default and used Amphotericin B when dealing with fungal infections. Some authors [18–20] state that they avoided intramedullary cutting guides to avoid contamination of the space, using extramedullary devices only.

Interim Period and Diagnosis of Eradication of Infection

Most authors have used at least 6 weeks of antibiotics, with three [14, 17, 19] specifically using the intravenous route for this period. Some authors have used shorter IV regimens converting to orals at 1 week [20] or 2 weeks [22]. Three authors [16, 19, 20] have continued oral antibiotics till final implantation. Comparing the mean antibiotic duration with spacer duration, it appears that at least two [14, 17] have stopped at some stage, with the latter clarifying that oral antibiotics were continued selectively, based on clinical assessment and lab results.

Among 5 papers [14, 17–19, 22] that describe their post-op rehab protocol, the knee was splinted in extension in case of static spacers. Range of motion was allowed in the case of articulating spacers with one author using Continuous Passive Motion and a posterior stabilising brace [22]. One author has also used Partial Weight Bearing for 2 weeks graduating to Full-weight bearing [17].

Most papers [14–18, 20–22] have used blood results [C-Reactive Protein (CRP) levels/Erythrocyte Sedimentation Rates (ESR)] to establish eradication of infection, with some [16, 17, 20] using a criterion of normal results over 2 consecutive weeks. Three papers [15, 17, 18] have mentioned stopping antibiotics for at least 2 weeks before using lab parameters to confirm eradication. In one paper [14], all patients underwent a diagnostic aspiration for cell-counts and cultures, in addition to bloods to establish eradication.

The mean duration between spacer implantation and final implant application across series was 16.5 weeks, with a minimum of 6 weeks. 6 patients (4.4%) have been reported to need a second operation for debridement and spacer application, before proceeding to their 2nd stage [14, 19, 22]. This excludes one patient [20] where the knee had to be fused after failure to control infection after 3 reoperations. 1 paper mentions a case needing manipulation for arthrofibrosis with a spacer implanted as well as a case of spacer dislocation [14]. One paper [22] also reports 2 patients who elected not to proceed with a 2nd stage as they were satisfied with knee function after the 1st stage.

Second Stage (Final Implantation)

3 papers [14, 17, 22] mention the use of peri-operative sampling for frozen sections in all patients, while 3 [15, 19, 20] reserved its use only if there was intra-operative suspicion of continued infection. The number of samples specified in 2 papers [15, 17] varies from 3–6. The cut-off mentioned in 2 papers (14,17) was 10 WBCs/High power field, though one of them recommends a lower figure of 5 if CRP/ESR were also elevated. Three of the authors [14, 18] mention routine sampling for microbiology at 2nd stage, with 2 [19, 20] only doing so if there was suspicion of infection.

While one author has used semi-constrained implants routinely [22], others [14, 17, 19, 21] have used it as needed with one [14] using it in 90% (27/30) of patients. 2 patients in one series [18] have been treated with rotating hinge implants, with one needing the use of augments.

Post-operatively, no papers mention any change from their routine protocols for primary knee replacements. There is considerable variation in the post-op antibiotic regime in the 6 papers reporting this, varying from a week [17] or till cultures are reported negative [15, 18], to 6 months. Nazarian et al. [19] have continued lifelong suppression in nearly half their patients after discussion.

Results and Complications

At 2 years the infection-free survival is 95.6% (133/139, CI 94.7–96.4) pooled across series. 1 patient with reinfection was successfully treated by a poly exchange with antibiotics and did not require revision of other components [14]. 1 patient also had a femoral condyle fracture in addition [14]. 1 patient had a polyethylene exchange when reoperated for patellar instability [14]. No further infections or revision for other causes were mentioned over the follow-up period (mean 3.9 years; range 2–13 years). Over the pooled exposure of 543 person years, the estimated rate of re-infection is 1.1 per 100 person years. Mean time to reinfection in the 6 cases could not be calculated, but 2 developed it within 6 months to a year [16], and in 4 [14, 15] it occurred between 1 and 2 years of final implantation.

Non-infectious complications mentioned are arthrofibrosis requiring manipulation (3 patients) and arthroscopic adhesiolysis (1 patient) [14], 1 case of Heterotrophic Ossification [21], 2 cases of patellar instability needing further operation [14, 19], 1 case of hematoma formation requiring washout [19], and one skin grating for wound-related problems [19]. These constitute a pooled complication rate of 6% among those without reinfection after the second stage. However, reporting was inconsistent with five papers [16–18, 20, 22] mentioning no complications while two papers reported rates as high as 16.7% [14] and 21.4% [19].

8 papers [14–20, 22] mention the use of standard outcome measures, with the commonest being the Knee Society Score (KSS) used in 4 papers [14, 16, 19, 22]. The pooled KSS and KSS Function scores in the included 70 patients, at last review (which was after 2 years of the final implantation) was 83.4 (range of means 80.1–89) and 76.8 (range of means 71.5–83), respectively.

Six authors [14, 17–20, 22] reported remarkable improvement in knee function (as measured by knee range of motion, outcome score performance or pain measures) and quality of life as a result of the two-stage replacement. Two of them [14, 20] have shown the improvement in outcome measure performance and ROM to be statistically significant.

Predictors of Failure

Two papers [14, 15] have looked for risk factors associated with re-infection, albeit analysing two-stage primary replacements in hips and knees together. Kunze et al. found raised ESR and CRP levels before 2nd stage to be independently predictive of complications (including septic recurrence). Xu et al. found higher CRP levels before 1st stage, older age and resistant organisms to be associated with infection recurrence. Unlike the former, they did not find the ESR or CRP levels before the second stage to be predictive. The latter explicitly mention all the risk factors which were analysed for an association—they did not find any predictive value of gender, BMI, comorbidities, previous surgery or other microbiological diagnoses.

Discussion

In the setting of two-stage revisions for infected knee replacements, Kunutsor et al. [23] in their meta-analysis, found an 8.8% infection rate within 2 years of final implantation (n = 5129). However, this included patients with a follow-up of less than 2 years. Nagra et al. [24] reported a reinfection rate of 13.5%, including patients with a minimum follow-up of 2 years (n = 185) only. Results of both reports compared favorably with the results for two-stage primary knee replacement in this review, with a reinfection rate of 4.4.% at 2 years (n = 139).

We also found good clinical outcomes of the procedure. The average Knee Society Scores in 4 papers grade as excellent [14, 16, 19, 22]. 6 papers [14, 18–22] report the mean range of knee flexion to be beyond 100 degrees in 82 patients, with 3 [19–21] also noting no extension lag and no flexion contractures > 5°. Statistically significant improvements from pre-operative scores as well as ROM were noted in 2 papers [14, 20].

It is well established that appropriate treatment of acute septic arthritis (SA) consists of early detection, joint aspiration to confirm the diagnosis and isolate an organism, joint debridement and washout (open or arthroscopic, evidence favors the latter [3]) and appropriate antibiotics. Only 1 author in our review has exclusively reported patients presenting to them with the first episode of acute septic arthritis [21]. They proceeded with two-stage arthroplasty only after failing to control infection with conventional methods (1–2 arthroscopic washouts and at least 4 weeks of antibiotic treatment) in 5 patients. These constitute 3.7% of the 136 knees with acute septic arthritis they treated over 3 years, which is comparable to the 3% failure rate with arthroscopic washouts in 123 knees reported by Johns et al. [3].

One of the authors [20] seems to have directly proceeded with a two-stage procedure in knees with advanced degeneration presenting with the first acute episode of SA. While they have obtained good results, such a strategy is unlikely to be cost-effective when applied generally, given that the vast majority of these infections resolve with a single arthroscopic washout. Further, an arthroscopic washout can aid in confirming joint degeneration, and deep-tissue samples can be collected to improve microbiological yield. This can provide valuable information if a two-stage replacement is eventually required.

All other papers have performed the procedure for a recurrent or ongoing (chronic) knee infection. In this context, the effectiveness of arthroscopic debridement for curing infection is likely to be low. However, it may still be useful to diagnose the extent of joint involvement using the Gächter classification [25], decrease infection load, provide deep samples for pre-op microbiological yield as well as histopathologic examination which can clarify other causes of a similar presentation such as rheumatological disease. According to Elsissy et al. [26] the Gächter classification is predictive of the need for more extensive debridement. Seo et al. [17] performed arthroscopic washout in patients who refused two-stage replacement, and it was effective in eradicating infection in 81% (31/38) patients, though 39% (12/31) of these were later treated by single-stage knee replacements. However, comparing patients undergoing arthroscopic washout (n = 38) with those undergoing 2-stage procedure (n = 21 including 7 patients coming from arthroscopy group after a recurrence of infection), they found that failure to control infection was statistically higher in the arthroscopic washout group. Thus, the evidence remains equivocal and further studies, especially looking at Quality of Life parameters and costs, are required to clarify this.

Information about the severity of infection can also be obtained from MRI scans, which can detect early evidence of bone involvement and deep soft tissue foci that are unlikely to be eliminated by washouts and need more extensive debridement. This can be reliably provided with bone excision in a first-stage procedure, with an articulating spacer implantation to maintain function. Further studies are required to elucidate the role of MRI scans in this setting.

10 cases of Osteomyelitis about the knee were treated by 4 papers [14, 18, 19, 22] in this review. The possibility of treatment with a washout alone is remote with bone involvement. A two-stage procedure allows removal of necrotic tissue and bony foci, facilitating antibiotic penetration to heal the infection. The spacer can also allow time for soft tissue healing, and if there are any issues with soft-tissue coverage, allow plastic surgical procedures to address it before implantation of the final prosthesis. Most authors have used advanced imaging (MRI/ Bone scan) to confirm the extent of bone involvement.

In the appropriate settings, a high index of suspicion should be maintained for tuberculosis or fungal infections in case of chronic joint infections, and appropriate specimens for cultures and histopathological examination should be sent. In the developing world, tuberculosis should feature in the differential diagnosis for all chronic knee infections. In this review, 3 patients had a fungal infection [17, 22], one of whom also had a concomitant tuberculous infection [22].

One of the authors [14] has treated advanced degeneration with remote SA considered to have healed with no signs of continued infection by a 2-stage arthroplasty while another [16] has performed single-stage arthroplasty for the same. While results for this group are not presented by the former, the latter have reported no infective recurrences or failures in their 23 patients. Previous papers reporting a single-stage approach report a relatively low deep reinfection rate ranging from 3–7.7% [27–29]. As such the evidence is supportive of a single-stage replacement in this scenario, though Kunze et al. recommend the two-stage procedure as the safer option while advocating shared decision-making with due consideration for the stakes involved. Further studies are required to clarify the best approach.

The procedural details of the two-stage protocol, including antibiotic treatment regimes, selection of diagnostic tests and their timing, surgical choices regarding cement, spacers and implants, and rehabilitation varied between the papers. This likely reflects the differences in routine protocols in two-stage revision arthroplasty for periprosthetic joint infections. Indeed, it is sound practice to follow the evidence from this far commoner surgery with a much wider evidence base. We were unable to find any difference in outcomes based on these differences due to small numbers in the papers.

On scrutiny of the reporting forms of primary knee replacement for three nation-wide registries (the British National Joint Register (NJR), the Swedish Knee Arthroplasty Register (SKAR) and the Australian Orthopedic Association National Joint Replacement Registry (AOANJRR) we found ‘infection’ listed among the data entry fields for indications in two (AOANJRR and NJR), but there was no data entry field to register a ‘two-stage’ operation. We believe this should be incorporated to follow these patients on a larger scale and obtain reliable information on their long-term survival and outcomes.

This review is the first to look at two-stage primary knee replacements for infection. We have included papers with a follow-up of at least 2 years after final implantation to provide a reliable picture of the primary outcome (infection-free survival). We have looked at the indications as well as outcomes using standardized scoring systems. We have also summarized pertinent technical details.

There are many limitations of this review. All the studies were retrospective cohort studies. The mean MCMS was 45.1 (42–49). All were graded poor on quality assessment and the evidence generated from them is weak. There is considerable heterogeneity among the included patients in the papers, with differences in the duration and nature of the joint infection. Finally, as the exact duration of follow-up was not reported for all included patients, it was not possible to construct Kaplan–Meier survival curves.

In conclusion, two-stage primary knee replacement is a safe, effective and reliable procedure for the treatment of recurrent or chronic knee infections on a background of advanced joint degeneration. It can treat the infection as well as provide good outcomes in the short to medium-term. Further studies are required to clarify the exact indications and cost-effectiveness in comparison to other strategies. All joint registries should introduce dedicated data entry fields in their primary knee arthroplasty data capture forms to identify patients undergoing two-stage procedures to understand their long-term survival and outcomes.

Declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical standard statement

This article does not contain any studies with human or animal subjects performed by any of the authors.

Informed consent

Informed Consent is not required for this type of study.