Arthrodesis Should Be Strongly Considered After Failed Two-stage Reimplantation TKA

Abstract

Background

A two-stage reimplantation procedure is a well-accepted procedure for management of first-time infected total knee arthroplasty (TKA). However, there is a lack of consensus on the treatment of subsequent reinfections.

Questions/purposes

The purpose of this study was to perform a decision analysis to determine the treatment method likely to yield the highest quality of life for a patient after a failed two-stage reimplantation.

Methods

We performed a systematic review to estimate the expected success rates of a two-stage reimplantation procedure, chronic suppression, arthrodesis, and amputation for treatment of infected TKA. To determine utility values of the various possible health states that could arise after two-stage revision, we used previously published values and methods to determine the utility and disutility tolls for each treatment option and performed a decision tree analysis using the TreeAgePro 2012 software suite (Williamstown, MA, USA). These values were subsequently varied to perform sensitivity analyses, determining thresholds at which different treatment options prevailed.

Results

Overall, the composite success rate for two-stage reimplantation was 79.1% (range, 33.3%–100%). The utility (successful outcome) and disutility toll (cost for treatment) for two-stage reimplantation were determined to be 0.473 and 0.20, respectively; the toll for undergoing chronic suppression was set at 0.05; the utility for arthrodesis was 0.740 and for amputation 0.423. We set the utilities for subsequent two-stage revision and other surgical procedures by subtracting the disutility toll from the utility each time another procedure was performed. The two-way sensitivity analysis varied the utility status after an additional two-stage reimplantation (0.47–0.99) and chance of a successful two-stage reimplantation (45%–95%). The model was then extended to a three-way sensitivity analysis twice: once by setting the variable arthrodesis utility at a value of 0.47 and once more by setting utility of two-stage reimplantation at 0.05 over the same range of values on both axes. Knee arthrodesis emerged as the treatment most likely to yield the highest expected utility (quality of life) after initially failing a two-stage revision. For a repeat two-stage revision to be favored, the utility of that second two-stage revision had to substantially exceed the published utility of primary TKA of 0.84 and the probability of achieving infection control had to exceed 90%.

Conclusions

Based on best available evidence, knee arthrodesis should be strongly considered as the treatment of choice for patients who have persistent infected TKA after a failed two-stage reimplantation procedure. We recognize that particular circumstances such as severe bone loss can preclude or limit the applicability of fusion as an option and that individual clinical circumstances must always dictate the best treatment, but where arthrodesis is practical, our model supports it as the best approach.

Introduction

Infection after TKA is a devastating complication that causes severe morbidity to patients and generates tremendous costs to the healthcare system [29]. The frequency of infection after primary TKA has been reported to be approximately 1% in most large series but ranges from 0.5% to 3% [11–13, 19]. For the chronically infected TKA, a two-stage reimplantation procedure has been shown to be effective at controlling infection with success rates as high as 90% [11, 13].

However, recent studies have reported lower success rates of two-stage reimplantation for management of the infected TKA. Mahmud et al. [24], in a series of 253 consecutive two-stage revisions for infection, reported 85% and 78% infection-free survivorship at 5 and 10 years, respectively. Furthermore, the increasing incidence of resistant organisms and increasing patient comorbidities have also contributed to the decreasing success rates of treatment [32]. Mittal et al. [25] reported a reinfection rate of 24% in a series of patients infected with either methicillin-resistant Staphylococcus aureus or methicillin-resistant Staphylococcus epidermidis. Consequently, a substantial number of patients with infected TKA treated with a two-stage revision can fail treatment and present with recurrent infection in their knees [27].

Options for treatment for the reinfected TKA include repeat staged revision, prosthesis retention and chronic antibiotic suppression, knee arthrodesis, and amputation. Although a repeat two-stage reimplantation can preserve both the joint and the limb, it can be associated with significant morbidity to the patient, higher rates of reinfection, and increased burden to the healthcare system [23]. Furthermore, even with successful infection control, patients after two-stage reimplantation often report residual pain, stiffness, and dysfunction after surgery [18]. Thus, the question remains: what is the optimum treatment for the reinfected TKA after a failed two-stage reimplantation procedure?

Therefore, the purpose of this study is to perform a systematic review of the literature and a decision analysis to determine the treatment method likely to yield the highest quality of life for a patient after a failed two-stage reimplantation procedure of an infected TKA.

Materials and Methods

A systematic review of the English literature was performed to estimate the reinfection rate after a two-stage reimplantation procedure and to establish utility values for the health states that could follow a two-stage revision, including repeat staged replacement, knee arthrodesis, prosthesis retention and chronic suppression, and amputation. Three authors (CHW, CFG, G-CL) independently carried out a comprehensive search of MEDLINE, EMBASE, and Cochrane computerized literature databases for studies relating to treatment of infected TKA with a two-stage reimplantation procedure. The medical subject headings (MeSH) terms used were “knee replacement” AND “infection”, “total knee arthroplasty” AND “infection”, “2 stage procedure”, “2 stage reimplantation”, “2 stage revision” AND “knee replacement”, and “2 stage revision” AND “total knee arthroplasty”.

The inclusion criteria for this study included the papers published in the English language, studies on knee arthroplasties that resulted in infection, outcomes reporting of secondary procedures, reoperation rates reported at the time of final followup, studies including series greater than four patients, and papers with levels of evidence IV or higher. Papers were excluded if they were procedural or technique papers, meta-analyses, or studies not focused on outcomes of treatment of infected TKA. The bibliographies of studies matching the inclusion criteria, including review articles, were traced for citations missed by the electronic search.

All authors independently reviewed the titles and abstracts of the initial literature search. If any abstract was deemed relevant by any reviewer, the full text of the article was reviewed. Only papers deemed relevant by all three reviewers were included for the final analysis. The reviewers resolved disagreement by discussion and consensus (Fig. 1). Two authors (CFG, G-CL) independently extracted the data including number of patients treated, number of infection-free patients at final followup, number of patients who underwent repeat two-stage reimplantation, knee arthrodesis, prosthesis retention and chronic suppression, amputation, death, and any failures of repeat two-stage reimplantation. The reinfection rate after a two-stage reimplantation procedure for infected TKA was determined by pooling the aggregate data and dividing the number of patients with persistent infection by the total number of patients treated with a two-stage revision.

Fig. 1.

Systematic review of the literature on the two-stage reimplantation procedure for management of infected TKA. The search strategy including search criteria and subsequent exclusions is depicted in this flow diagram.

To determine the utility and disutility of the treatment options after a failed two-stage revision procedure for infection, another systematic review on the clinical outcomes of two-stage reimplantation procedures, knee arthrodesis, chronic suppression, and amputation was conducted. Three authors (CHW, CFG, G-CL) again independently carried out a comprehensive computerized search of the MEDLINE, EMBASE, and Cochrane databases. The MeSH terms used to determine the utility and tolls for knee arthrodesis were “TKA” AND “arthrodesis”, “TKA” AND “fusion”, “TKA” AND “arthrodesis” AND “infection”, and “TKA” AND “fusion” AND “infection”. Similarly, for amputation after failed treatment of infected TKA, the MeSH terms used included “amputation” AND “knee infection”, “amputation” AND “infected TKA”, “AKA” and “TKA”, and “above knee amputation” AND “TKA”. Finally, for chronic suppression, the search terms used were “knee infection” AND “suppression”, “infected TKA” AND “suppression”, and “chronic antibiotic suppression” AND “infected knee”.

The inclusion criteria for this study included the papers published in the English language, studies on infected knee arthroplasties that resulted in above-knee amputation (AKA), knee arthrodesis, and débridement and chronic suppression. The studies needed to include outcomes reporting of validated clinical rating scales (ie, Knee Society scores, WOMAC, SF-12 or SF-36, or EQ-5D) or specific utilities or tolls for each procedure, reoperation rates, studies including series greater than four patients, and papers with levels of evidence IV or higher. Papers were excluded if they were procedural or technique papers, meta-analyses, or studies not focused on outcomes of treatment of infected TKA. The bibliographies of studies matching the inclusion criteria, including review articles, were traced for citations missed by the electronic search.

Like with the previous systematic review, all authors independently reviewed the titles and abstracts of the original search. Only papers deemed relevant by all three reviewers were included for final analysis (Figs. 2–4). The reviewers resolved any disagreement by discussion and consensus. Two authors (CFG, G-CL) subsequently extracted the data including the number of patients treated, success and failure rates of a particular treatment, any clinical ratings including Knee Society scores, SF-12 and SF-36 scores, and EQ-5D scores, and utility or disutility (tolls) for each procedure. When unavailable, the utility or tolls were derived using a systematic approach.

Fig. 3.

Systematic review of the literature on AKA after infected TKA. The search strategy including search criteria and subsequent exclusions is depicted in this flow diagram.

Fig. 2.

Systematic review of the literature on knee arthrodesis after failed TKA. The search strategy including search criteria and subsequent exclusions is depicted in this flow diagram.

Fig. 4.

Systematic review of the literature on chronic suppression in the setting of infected TKA. The search strategy including search criteria and subsequent exclusions is depicted in this flow diagram.

A utility is a patient-centric measure of preference for choosing one treatment option over another that serves as a proxy for clinical outcome. In this context, it is a linear scale in which a value of 1 indicates normal health (function) with no morbidity and a value of 0 indicates death (no function). When a utility value was unavailable from the literature, the utility for the particular procedure was determined by normalizing the average of the published health-related quality measures (ie, SF-12, SF-36, EQ-5D scores) on a scale from 0 to 1. For example, for the utility value for a two-stage revision procedure for infection, Baker et al. reported a mean EQ-5D score of 0.473 (95% confidence interval, 0.397–0.548) [2]. Therefore, the utility input for a two-stage revision in the decision tree was 0.47. Conversely, for AKAs after infected TKA, the utility value was derived by adapting the SF-12 data published by Fedorka and associates [9]. The utility input was estimated by combining the physical component score (maximum 50 points) and mental component score (maximum 50 points) and dividing the aggregate by 100 to yield a number between 0 and 1. Additionally, the utility tolls (cost of performing a procedure to the patient and healthcare system) were estimated from the literature and subtracted from the original utility value each time an additional procedure was performed in accordance with previously described methodology by Bozic et al. [7]. For example, the toll for a two-stage revision was abstracted from previous work by Bedair et al. and the value of 0.2 was used as the input for the decision tree [3]. The toll for prosthesis retention was assumed to be 0.05 as a result of a lack of published cost data and using the 0.2 value of a two-stage reimplantation procedure as a benchmark for patient morbidity and cost. Important input values for the model are summarized (Table 1).

Table 1.

Summary of important values for decision tree

Variable name	Utility	Probability of success	Utility toll
Amputation	0.423	1.00
Fusion	0.74	0.786
Two-stage reimplantation	0.473	0.791
Chronic suppression		0.52
Revision toll			0.2
Chronic suppression toll			0.05

These variables were then inputted into a decision tree analysis modeling software (TreeAge Pro 2012; TreeAge Software, Williamstown, MA, USA). Using probabilities of success, failure, clinical outcome (utility), and cost (utility tolls), every possible subsequent treatment option at a specific point in the management of a reinfected TKA can be graphically represented as lines that are connected by nodes that represent the decision to choose a particular treatment or chance result of a previous decision. These lines and nodes are combined to form a tree of possible outcomes allowing the software to mathematically produce a preferred choice (Fig. 5). In our model, successful arthrodesis, amputation, and death were considered terminal nodes at which point infection was considered to be eliminated. Both prosthesis retention and chronic suppression and repeat two-stage revision were considered nonterminal nodes because reinfection was still a possible clinical scenario. For those who continued to fail treatment, the decision tree was iterated until a third and final two-stage reimplantation procedure.

Fig. 5.

Graphic representation of care decision-making in a simulated patient status after failed two-stage revision. The boxed values represent starting values and their assigned respective variable names. The blue square node is the decision node. The green circle node is the probability node, where literature values on the probability of future outcomes are computed. The red node is the terminal node, where infection is resolved. Texts above each line represent the status of the patient at that specific stage. The numbers and texts below each line represent the formula used to calculate the probability of entering that branch of the tree. The formula to the right of the terminal node calculates the utility gained from having traversed down that specific branch.

Finally, given the element of uncertainty, variable study quality, and incomplete data, a two-way and three-way sensitivity analysis was performed. The input variables such as utility and success rates of repeat two-stage revision, AKA, and chronic suppression were varied in head-to-head comparisons (two-way) and in group comparisons (three-way) to assess whether the model continued to prefer one treatment over another when these variables were changed over a range of plausible values derived from the published literature.

Results

The initial literature search returned 204 papers on two-stage reimplantation and infected TKA. After elimination of errata, there were 79 papers that underwent abstract review resulting in 53 other exclusions. Twenty-six full-text papers were reviewed and 18 papers reporting on 494 patients who underwent two-stage reimplantation for infected TKA with information on outcomes and reoperation rates were available for final analysis (Appendix 1). Overall, the aggregate success rate of a first two-stage reimplantation procedure was 79.1% (391 of 494 patients). Of the 103 failures, reported salvage procedures included: repeat two-stage reimplantation (n = 44 [8.9%]), knee arthrodesis (n = 26 [5.3%]), prosthesis retention and chronic suppression (n = 17 [3.4%]), and AKA (n = 6 [1.2%]). Ten patients (2%) who underwent two-stage reimplantation for infected TKA had died at the final followup.

The estimated utilities for a successful repeated two-stage reimplantation procedure, knee arthrodesis, and AKA were 0.47, 0.74, and 0.42, respectively [Table 1]. The success rates for two-stage reimplantation for infected TKA averaged 79% (range, 33%–100%). The failure rates after AKA and successful knee arthrodesis were assumed to be minimal. Interestingly, the utility estimated from available published data on patients with a successful knee arthrodesis was higher than the utility of a patient after revision TKA (0.74 versus 0.47) [2, 5]. Although this finding can initially appear inconsistent, it is important to remember that utility is a patient centric measure derived from self-reported quality measures (SF-12, SF-36, EQ-D5), which takes into account mental states and overall sense of well-being rather than individual joint function. In this situation, patients with a successfully fused knee without pain can therefore have a significantly higher utility rating compared with a group of patients after a two-stage revision for infection with persistent pain, stiffness, and possibly less than expected knee function. For patients undergoing repeat two-stage revision for treatment of a reinfected TKA, the utility toll reported by Bedair et al. [3] was equal to 0.20. This factor was calculated to simulate the additive morbidity of additional treatments as a cost for each procedure. Finally, because there was no published information on the utility for prosthesis retention and chronic suppression, it was arbitrarily assigned a toll value of 0.05 using the 0.2 toll for repeat two-stage revision as a benchmark. The rationale was that prosthesis retention was significantly less morbid than undergoing multiple repeat operations but included the morbidity of chronic antibiotic use, cost, side effect profile, and resistance risk.

The decision tree model favored knee arthrodesis as the treatment modality that optimizes infection control and maximizes function as represented by utility after a failed two-stage reimplantation for infected TKA. Two-way sensitivity analysis varies utility of two-stage revision from 0.47 to 0.99 and success rate of revision from 45% to 95% simultaneously. This was then extended to include a third variable: utility of fusion or utility toll, hence also known as three-way sensitivity analysis. First, the utility of fusion is lowered to 0.47 to match the utility of two-stage reimplantation. Second, the utility toll of a failed revision is lowered from 0.2 to 0.05. Each of these cases was run independently over the same range of values as the two-way sensitivity analysis to see if the preference for arthrodesis would alter. In nearly all scenarios, knee arthrodesis remained the preferred treatment arm (Fig. 6A–B). For a repeat two-stage reimplantation procedure to be favored, the utility of the subsequent two-stage reimplantation had to exceed the published utility of primary TKA of 0.84 [28], and the probability of achieving infection control had to exceed 90%.

Fig. 6A–B.

(A) Two-way sensitivity analysis. Color represents the preferred treatment choice for the patient to maximize utility. Fusion is the dominant strategy, except in regions where success rate of two-stage reimplantation is expected to be >90% and revision utility >0.95. (B) Three-way sensitivity analysis. Color represents the preferred treatment choice for the patient to maximize utility. Fusion is the dominant strategy unless revision utility is at least greater than 0.80. Above revision utility of 0.80, >65% chance of revision success rate will favor two-stage reimplantation, whereas a >0.90 revision utility but <65% chance of revision success rate will favor chronic suppression as the treatment choice for maximizing utility. Rv = revision; 2SR (2) = two-stage reimplantation.

Discussion

The two-stage reimplantation procedure has been shown to be an effective treatment strategy for infected TKA [11–13]. However, recurrent infection has been reported to be as high as 22% at 10 years [24], although other studies suggest it may be much lower [15, 33, 34]. Although joint and limb preservation after a prior failed treatment is appealing to both the patient and surgeon, a repeat two-stage reimplantation procedure can be associated with additional risks and morbidity to the patient, decreasing success rates for infection control and added economic costs [29]. Additionally, a failed repeat two-stage reimplantation can lead to further bone loss and damage to the soft tissue envelope that can potentially compromise the success of other salvage options such as knee arthrodesis. Thus, the question remains: what is the strategy that optimizes infection control and maximizes function after a failed two-stage reimplantation for infected TKA? Unfortunately, the literature provides few direct answers to this question. Therefore, the purpose of this study was to perform decision analysis to establish the management strategy for the recurrent infected TKA most likely to result in the highest quality of life for a hypothetical patient.

This study has several limitations. First, this is a systematic review of the literature and therefore the data are an aggregate of multiple studies with potentially varying selection criteria and treatment techniques (ie, antibiotic dosage in spacers, timing of reimplantation, patient population, and followup). Pooling of data of studies of varying quality and levels of evidence can result in both selection and treatment biases. However, because the studies selected for final analysis reported relatively standard protocols of two-stage revision, treatment bias is minimized. Additionally, because our model included looking at scenarios of varying success rates of two-stage revision procedures from 45% to 95% during the sensitivity analysis, the impact of these imprecisions is minimized. Second, not all studies reported on all of the variables required deriving the probabilities of success of subsequent procedures (ie, repeat two-stage reimplantation, arthrodesis, or chronic suppression) and, thus, this can lead to the studied variables being underpowered and result in Type II error. To address this possibility, a robust two- and three-way sensitivity analysis with a wide range varying utilities, probabilities of success for each treatment alternative, was performed to minimize the risk of arriving at a wrong conclusion. However, even when sensitivity analysis is performed, if the range of success rates and utilities is poorly defined, the decision tree can yield erroneous results. Nevertheless, our model shows that a repeat two-stage reimplantation would have to be equally functionally and successful as a primary TKA and with a 90% chance of infection control to be favored over other treatments. This is an unlikely scenario based on published outcomes of two-stage reimplantation procedures for infected TKA.

Our model considers all infected TKAs to be the same and does not stratify patients and infections based on host factors (ie, aging patient), bone loss, failed amputations requiring additional surgeries and antimicrobial profiles, and, therefore, it may fail to identify circumstances under which a repeat two-stage revision may be favored over other treatments. Although other tools for decision modeling such as time cost-tradeoff modeling or Markov Chain Monte Carlo analysis can in theory take into account some of these recursive analyses, they also required significantly more accurate and complete input data. A major limitation in this study is the relative lack of published data and the wide range of methods of published outcomes without standardization. This required assumptions and a systematic approach for deriving the missing inputs into the decision tree. A decision tree analysis has advantages over other statistical tools because it allows one to simulate situations over a larger range of values without having to know the true value. Our literature review serves only as a starting point to provide us with a plausible range of values and does not intend to be a true meta-analysis because it is not subjected to the methodological rigors of one. Additionally, this tool also allows for more than one correct interpretation for any given situation. Depending on what the surgeon/patient believes the true utility or success rates for a specific approach is for that particular scenario, they will be able to find the utility-maximizing decision for that particular scenario on the sensitivity analysis graph. Consequently, unlike conventional statistical methods, it cannot precisely show that one treatment is better than another treatment with a specific confidence interval, but rather it can suggest that arthrodesis is the utility-maximizing treatment over the range of utility and success rate values simulated. Decision trees, therefore, overcome relatively poorer data inputs by running a larger number of simulations. Our hope is that this study serves as a starting point to large, prospective multicenter clinical studies dealing with this very complicated clinical situation. Finally, we did not review or consider a single-stage exchange as a treatment option for management of a chronically reinfected knee. Single-stage exchanges can minimize morbidity of treatment but may come at the cost of lower rates of infection control [16, 35]. Although single-stage exchange is an acceptable treatment option for a first-time infected TKA, the role of these procedures in the setting of a failed prior two-stage revision is not well defined.

Our systematic review found the two-stage approach resulted in infection control in 79% of patients treated with it. This result is inferior to previously reported success rates of treatment approaching 90% [11–13, 24, 33]. A reason for this discrepancy is that there is great heterogeneity in the published studies dealing with outcomes after a two-stage reimplantation procedure for treatment of infected TKA. Because of this variability, an aggregated result of the success and utility of this procedure can be imprecise as a result of both selection and treatment bias. This limitation is hopefully mitigated by the sensitivity analysis. Furthermore, although some authors have recently reported continual high success of two-stage reimplantation protocols even in the presence of resistant organisms provided minimal bactericidal titers are achieved [34], recent studies have confirmed the declining effectiveness of two-stage reimplantation contributed by the emergence of drug-resistant organisms, increased patient comorbidities, and atypical infections[11, 26, 36]. Mortazavi et al. [26] reported a series of 117 infected TKAs treated with two-stage reimplantation and found that 28% required reoperation for infection at an average of only 3.4 years. In this series, culture-negative infections, methicillin resistance, and increased reimplantation time were predictors of failure. Additionally, Zmistowski et al. [36] reported even lower success rates in TKAs infected with gram-negative and polymicrobial organisms. In their series, success rates of 69% were achieved when sensitive gram-positive organisms were the infecting agent compared with rates of infection control approximating 50% in gram-negative and methicillin-resistant infections. Polymicrobial infections were associated with universally unsuccessful treatment. Consequently, patients should be counseled that successful control of infection is dependent on a combination of factors including host factors and microbial profile and antibiotic resistance.

Although the wish for a functional knee may impact clinical decision-making, the data do not support repeat two-stage revision after one failed two-stage reimplantation procedure, because the success rate and knee function are likely to be poor. There is wide variability in the reported rates of infection control after a second two-stage reimplantation procedure. Azzam et al. [1] reported in a series of 18 patients undergoing a repeat two-stage reimplantation for recurrent infection a 22% rate of persistent infection after repeat two-stage exchange. However, Hanssen et al. [14] reported poor results in a series of 24 reinfected knees with failure to control infection in 23 knees (96%). At last followup, there were 10 successful arthrodeses, five chronically infected TKA, four AKAs, three failed arthrodesis, and one resection arthroplasty. Furthermore, although Azzam et al. [1] reported reasonable pain relief and function after repeat two-stage revision procedures, Kim et al. [18] reported only modest return to activities of daily living with significant residual pain and stiffness after a successful two-stage reimplantation procedure. Additionally, Bengston et al. [4] also reported unpredictable functional outcomes after resection and reimplantation. The authors reported achieving infection control with an eventual functional prosthesis in only seven of 13 (54%) patients. Finally, although amputation can provide advantages of surgical infection control, return to function can be unpredictable and dependent on if prosthesis fitting and training are successfully achieved [22]. Sierra et al. [30] and Fedorka et al. [9] reported fewer than 50% of patients were eventually fitted for a prosthesis. Thus, patients and surgeons must take into consideration the amount of bone loss, host status, and antimicrobial profile as part of the shared decision-making process to maximize the likelihood of successful treatment.

In our decision tree modeling, knee arthrodesis is the treatment strategy most reliable at achieving infection control and maximizing function in patients with infected TKA after a failed two-stage revision. Knee fusion has been shown to be successful at achieving infection control in patients with chronically infected knees [22]. Mabry et al. [21] compared intramedullary nailing and external fixation arthrodesis for infected TKA and reported a 95.8% (23 of 24) fusion rate in the intramedullary group with a 8% reinfection rate compared with 64% (24 of 41) union using external fixation with only a 4.9% reinfection rate. More recently, use of circular frame fixators has improved union rates to rates comparable to intramedullary fixation [22, 31]. However, the degree of bone loss and residual limb length can affect the success and outcome for achieving solid fusion and good clinical function. Kutscha Lissberg et al. [20] reported in their series successful arthrodesis in 88% of patients with severe infections managed with hybrid external fixators but also warned against bone loss and significant limb shortening as contraindications.

Therefore, although repeating another two-stage reimplantation may be appealing to the patient, considerations must be given to the possibility that another failed two-stage reimplantation may jeopardize the chances at a successful knee arthrodesis in the future. Finally, knee arthrodesis, when successful, has been shown to result in pain relief and reasonable clinical function. Johnson and Bannister [17] reported in a small series of 25 knee infections that arthrodesis was the most successful treatment modality achieving pain relief and infection control in 11 of 12 (92%) patients at final followup. Additionally, others have shown comparable, if not improved, function compared with amputation or repeat two-stage reimplantation [30]. Blom et al. [6] reported comparable outcomes in a group of patients undergoing knee arthrodesis compared with those undergoing two-stage reimplantation in terms of Oxford knee scores. Lastly, Fuchs and Mersmann [10] evaluated a group of 19 patients with infected TKA treated with knee arthrodesis and reported no significant deterioration of their quality of life after successful arthrodesis compared with having a bending knee. Consequently, the findings of our decision analysis are consistent with available evidence, including those presented by Chen et al. [8], that knee arthrodesis can achieve infection control and return reasonable function in patients with an infected TKA. We recognize that particular circumstances such as severe bone loss can preclude the applicability of fusion as an option and that individual clinical circumstances must always dictate the best treatment.

In conclusion, based on current published reports, knee arthrodesis should be strongly considered when patients present with failed two-stage revision for infected TKA. Based on our decision analysis, fusion is most likely to provide infection control while maximizing patient function when there is sufficient residual bone stock and when repeat a two-stage reimplantation procedure has low likelihood of success (ie, resistant organisms, poor host, and inadequate soft tissue envelope). Future studies will need to risk-stratify prosthetic knee infections and focus on the effects of host factors, microbiology, antimicrobial profile, and residual bone stock and their effects on clinical outcome.

Appendix 1

Papers reporting on patients who underwent two-stage reimplantation for infected TKA

1. Kalore NV, Maheshwari, A, Sharma A, Cheng E, Gioe TJ. Is there a preferred articulating spacer technique for infected total knee arthroplasty? A preliminary study. Clin Orthop Relat Res. 2012;470:228–235.

2. Macheras GA, Kateros K, Galanakos SP, Koutsostathis SD, Kontou E, Papadakis SA. The long term results of a two stage protocol for revision of an infected total knee replacement. J Bone Joint Surg Br. 2011;93:1487–1492.

3. Choi HR, von Knoch F, Zurakowski D, Nelson SB, Malchau H. Can implant retention be recommended for treatment of infected TKA? Clin Orthop Relat Res. 2011;469:961–969.

4. Shen H, Zhang X, Jiang Y, Wang Q, Chen Y, Wang Q, Shao J. Intraoperative-made cement on cement antibiotic loaded articulating spacer for infected total knee arthroplasty. Knee. 2010;17:407–411.

5. Mittal Y, Fehring TK, Hanssen AD, Marculescu C, Odum SM, Osmon D. Two stage reimplantation for periprosthetic knee infection involving resistant organisms. J Bone Joint Surg Am. 2007;89:1227–1231.

6. Jamsen E, Sheng P, Halonen P, Lehto MU, Moilanen T, Pajamaki J, Puolakka T, Kontinnen YT. Spacer prosthesis in two stage revision of infected knee arthroplasty. Int Orthop. 2006; 30:257–261.

7. Ip D, Yam SK, Chen CK. Implications of the changing pattern of bacterial infections following total joint replacements. J Orthop Surg. 2005;13:125–130.

8. Hofmann AA, Goldberg T, Tanner AM, Kurstin SM. Treatment of infected knee arthroplasty using an articulating spacer: 2 to 12 year experience. Clin Orthop Relat Res. 2005;430:125–131.

9. Blom AW, Brown J, Taylor AH, Pattison G, Whitehouse S, Bannister GC. Infection after total knee arthroplasty. J Bone Joint Surg Br. 2004;86:688–691.

10. Hirakawa K, Stulberg BN, Wilde AH, Bauer TW, Secic M. Results of 2 stage reimplantation for infected total knee arthroplasty. J Arthroplasty. 1998;13:22–28.

11. Whiteside LA. Treatment of infected total knee arthroplasty. Clin Orthop Relat Res. 1994;299:169–172.

12. Teeny SM, Dorr L, Murata G, Conaty P. Treatment of infected total knee arthroplasty. Irrigation and débridement versus two-stage reimplantation. J Arthroplasty. 1990;5:35–39.

13. Bengston S, Knutson K, Lidgren L. Treatment of infected knee arthroplasty. Clin Orthop Relat Res. 1989;245:173–178.

14. Wilde AH, Ruth JT. Two stage reimplantation in infected total knee arthroplasty. Clin Orthop Relat Res. 1988;236:23–35.

15. Bengtson S, Knutson K, Lidgren L. Revision of infected knee arthroplasty. Acta Orthop Scand. 1986;57:489–494.

16. Walker RH, Schurman DJ. Management of infected total knee arthroplasties. Clin Orthop Relat Res. 1984;86:81–89.

17. Wasielewski RC, Barden RM, Rosenberg AG. Results of different surgical procedures on total knee arthroplasty infections. J Arthroplasty. 1996;11:931–938.

18. Goldman RT, Scuderi GR, Insall JN. 2 stage reimplantation for infected total knee replacement. Clin Orthop Relat Res. 1996;331:118–124.