High Rate of Fusion but High Complication Rate After Knee Arthrodesis for Infected Revision Total Knee Replacement

Abstract

Background

Knee arthrodesis is a demanding technique regarding difficulties to achieve bone fusion, control of infection and its associated complications. The purpose of this study was to evaluate if knee arthrodesis should still be indicated after failed revision total knee replacement.

Methods

This was a retrospective study of 45 patients. Age, gender, follow-up, pathogens, Charlson comorbidity index, time from primary arthroplasty to arthrodesis, number of previous procedures, surgical technique, functional capability assessed by the SF-12 score, limb-length discrepancy after arthrodesis, presence of radiographic knee fusion, and complications were recorded.

Results

The mean age at the time of operation was 72 years. 29% of patients were men, 71% were women, and the mean follow-up was 8.5 years. The average Charlson comorbidity index was 4.5. The most common microorganisms isolated were Staphylococcus epidermidis (29%), Staphylococcus aureus (22%), and Pseudomonas aeruginosa (15%). In 20% of patients, no microorganism was identified, and in 37% of patients the infection was polymicrobial. Time from primary total knee arthroplasty to arthrodesis was 55 months, and patients underwent a mean of 3.9 previous surgeries. The surgical technique used was an intramedullary long nail in 95.5%. Functionally, 93.3% of patients walked with weight bearing. The SF-12 was higher after arthrodesis (p < 0.05). Mean limb length discrepancy was 2.4 cm. Among the group treated with long intramedullary nailing, 91.1% obtained tibiofemoral fusion. Complications occurred in 37.6% of patients.

Conclusions

Knee arthrodesis with a long intramedullary nail after failed infected revision total knee replacement has a high rate of fusion, but the complication rate is high.

Introduction

Periprosthetic joint infection (PJI) is one of the most challenging and devastating complications after total knee replacement (TKR). Risk factors for a failed staged revision arthroplasty for recurrent infection include polymicrobial infections, culture-negative infections, infections caused by methicillin-resistant Staphylococcus aureus species, nicotine abuse, a higher Charlson comorbidity index score, and previously failed revision attempts [1]. Mortazavi et al. [2] observed that the failure rate after two-stage reimplantation for infected TKR was relatively high (28%) while taking into account reinfection and/or recurrence as a failure. If the infection cannot be eradicated or if multiple revision TKRs lead to loss of soft tissue, large bone defects, instability, or deficiency of the extensor mechanism, control of infection using revision TKR may be no longer possible. In these cases, knee arthrodesis, above-knee amputation (AKA), or resection arthroplasty are often the only treatment options [3, 4]. The reported incidence of arthrodesis following failed knee arthroplasty ranges from 0.21 to 1.11% [5].

Wu et al. [6] performed a systematic review on treatment options for persistent infection after failed revision TKR, and concluded that arthrodesis should strongly be considered to control infection and maximize function. In contrast, Röhner et al. [7] reported persistence of infection in 50% of patients, substantially impaired quality of life, and pain after knee arthrodesis. They concluded that bone fusion following the septic failure of revision TKR should be regarded with skepticism.

Knee arthrodesis is a demanding technique associated with a complication rate of nearly 40%, including persistent infection, pain, leg-length discrepancy, and rotational malalignment. To date, the success rate for the eradication of infection remains unclear [8].

In the context described above, should we still be suggesting knee arthrodesis after failed septic revision TKR? The aim of this retrospective review was to evaluate the results of knee arthrodesis after failed infected revision TKR at our institution, to confirm if this procedure should still be considered.

Materials and Methods

We performed a retrospective study of 63 patients who underwent a knee arthrodesis over the last 18 years. 47 patients who underwent an arthrodesis after a failed infected revision TKR were studied. 16 patients who underwent knee arthrodesis for other indications were excluded (malignant bone tumors, severe painful stiffness following TKR, periprosthetic fracture, neuropathic Charcot joint, chronic poliomyelitis syndrome). Additionally, two patients were excluded (one treated with external fixation, and one patient with a definitive static spacer). Finally, 45 patients treated with long (43 patients) or short (2 patients) intramedullary nailing were included (Fig. 1). Age, gender, follow-up, pathogens, Charlson comorbidity index [9], time from primary total knee arthroplasty to arthrodesis, number of previous procedures, surgical technique, functional capability assessed by the SF-12 [10], limb-length discrepancy after arthrodesis, presence of radiographic knee fusion, and complications were recorded. Anteroposterior and lateral radiographs of the femur, tibia, and entire lower limbs were obtained. Fusion was defined as a visible bony bridge between the femur and tibia (Fig. 2) with good bone contact (> 3/4 surface contact) according to Klinger et al. [11]. Leg-length discrepancy was measured clinically and radiographically in digital long-leg radiographs with a 20% magnification factor included (RIS/IMPAX Agfa Healthcare, Mortsel, Belgium). Radiographs were evaluated by two of the authors (GA and JO), blinded to the patient data.

Fig. 1.

Flowchart illustrating selection of patients for this study

Fig. 2.

Visible bony bridge between the femur and tibia after a knee arthrodesis with good bone contact (> 3/4 surface contact)

This study was approved by the local institutional review board (IRB)/ethics committee.

Operative Technique

A long intramedullary (IM) nail was used in 43 patients (95.5%): a T2 long IM nail (Stryker Trauma GmbH, Schönkirchen, Germany) in 26 patients, a Küntscher nail in 14 patients (Biomet Inc, Warsaw, IN, USA), and a Trigen knee fusion nail in three patients (Smith & Nephew, Memphis, TN, USA). Short modular nails were used in two patients (4.5%), one Mutars RS arthrodesis nail (Implancast GmbH, Lüneburger, Germany) and one Endomodel knee fusion nail (Waldemar Link, GmbH & Co. KG, Hamburg, Germany). The procedure was performed with a two-stage protocol: a static antibiotic-loaded polymethylmethacrylate cement spacer was used for 8–12 weeks before the definitive IM nailing was performed. Normal values of inflammatory markers, namely erythrocyte sedimentation rate and C-reactive protein along with improvement in clinical signs and absence of pus discharge, determined the ideal time for the second stage (arthrodesis procedure). Reinfection or persistent infection was diagnosed on the basis of recommendations from the American Academy of Orthopaedic Surgeons (AAOS) and the Musculoskeletal Infection Society [12, 13]. According to these recommendations, infection is confirmed by the presence of a fistula and/or the positive identification of a pathogen by culture from at least two separate tissue or fluid samples obtained from the affected joint (major criteria). If major criteria are absent, four of the following six criteria must be met to diagnose the presence of infection: elevated erythrocyte sedimentation rate (ESR) and serum C-reactive protein (CRP); elevated synovial leukocyte count; elevated synovial neutrophil percentage (PMN%); presence of purulence in the affected joint; isolation of a microorganism in only one culture of periprosthetic tissue or fluid, or greater than 5 neutrophils per high—power field in 5 high power fields observed from histologic analysis of periprosthetic tissue at 400 × magnification. Multiple intraoperative tissue samples were obtained for microscopy, culture, and sensitivities, which included joint aspirate and tissues from the surfaces and canals of the tibia and femur (minimum three samples according to the above-mentioned recommendations) [12, 13]. The choice of antibiotics for each individual was based upon preoperative microbiological investigation and sampling. All patients received a short course (5–14 days) of intravenous antibiotics, followed by 6 weeks of dual oral antibiotics.

In four patients the patella was added as an autologous graft to the arthrodesis. In two cases, two interfragmentary tibiofemoral screws were used in addition to a long IM nail. In two cases, arthrodesis was performed using external fixation in the first stage until control of infection was obtained, and then a long IM nail was used for definitive fixation. 2 cases required a gastrocnemius flap over the arthrodesis. In six patients, a distal femoral sliding graft was performed to minimize limb shortening (Fig. 3). Postoperatively, partial weight bearing was allowed.

Fig. 3.

Knee arthrodesis with an intramedullary nail and distal femoral sliding graft. On the right, 15 years later

Statistical Analysis

Statistical analysis was carried out using IBM SPSS^® Statistics for Windows, version 21.0 (IBM Corp., Armonk, NY, USA). Statistical significance was taken to be a p value of ≤ 0.05. Qualitative data were expressed as frequency and percentage for descriptive statistics. Quantitative data were expressed as mean and standard deviation (SD) if they followed a normal distribution, or as median and interquartile range if the distribution was not normal.

The Student’s t test was used to compare quantitative variables of the groups if the distribution of data was normal, and the Mann–Whitney U test if the variables were not normally distributed.

Results

The mean age at the time of operation was 72 years (range, 57–90 years). Mean follow-up was 8.5 years (range 2–18 years). 11 patients died for their medical comorbidities, not as a consequence of infection, but their follow-up until death was included.

13 patients were men (29%) and 32 were women (71%). The average Charlson comorbidity index was 4.5 (range 1–13). Hypertension, diabetes mellitus, obesity, and cardiac disease were the most common comorbidities (Table 1). In 6 patients, an extensor mechanism rupture occurred. Time from primary total knee arthroplasty to arthrodesis was 55 months (range 12–240 months). Time from primary TKR to the first stage was 42 months (range 10–228 months), and 4 months (2–12 months) from the first to second stage (arthrodesis). Time from the second stage to union was 5 months (3–10 months).Patients underwent a mean of 3.9 previous surgeries (range 2–9 previous surgical procedures). The most common microorganisms were Staphylococcus epidermidis (29%), Staphylococcus aureus (22%), and Pseudomonas aeruginosa (15%), either in isolation or combination. In nine cases (20%), no microorganism was identified, and other pathogens were found in 13% of cases (gram-negative rods such as Klebsiella pneumoniae, Enterobacter cloacae; gram-positive rods and other gram-positive cocci). In 17 patients (37.7%), the infection was polymicrobial. The Charlson comorbidity index was higher in patients infected with Pseudomonas aeruginosa compared with those infected with other pathogens, although this difference was not statistically significant (p = 0.559, Mann–Whitney U test). Forty-two patients (93.3%) walked with full weight bearing. The SF-12 physical component was higher (p < 0.05) after arthrodesis (39.5 ± 5.7) than before (27.5 ± 4.8). The SF-12 mental component was also higher (p < 0.05) after (53 ± 9.2) compared with before the procedure (27.1 ± 10.3). Among the group treated with long IM nailing, 91.1% achieved tibiofemoral fusion between the tibia and femur on radiographs. The mean limb-length discrepancy was 2.4 cm (range 0–6 cm). Complications after arthrodesis were present in 17 patients (37.6%) (Table 2). Nail breakage occurred in three patients. In one patient, the broken nail was removed and a new nail was inserted, while the other two patients required above-the-knee amputation to eradicate concomitant infection and control sepsis. In two more elderly, frail patients, infection of the arthrodesis occurred, and chronic antibiotic suppression was indicated. In three patients, fractures took place around the nail, including two intraoperative femur fractures and one hip fracture above the nail (in this case, a tibia fracture below the nail also occurred).

Table 1.

Medical comorbidities

Comorbidity (isolated or in combination)	Number of patients
Hypertension	21
Diabetes mellitus	11
Cardiac disease (atrial fibrillation, ischemic heart disease, valvular heart disease or congestive heart failure)	9
Obesity	9
Rheumatoid arthritis	4
Chronic obstructive pulmonary disease	4
Chronic renal failure	2
Ovarian cancer	1
Urothelial carcinoma	1

Table 2.

Complications

Complication	Proportion of patients
Aseptic non-union	4 patients (8.8%)
Infection of the arthrodesis	4 patients (8.8%)
Nail breakage	3 patients (6.6%)
Fractures around the nail	3 patients (6.6%)
Peroneal nerve paralysis	1 patient (2.2%)
Delayed union	1 patient (2.2%)
Gastrocnemius flap revision	1 patient (2.2%)
Contralateral TKR infection	2 patients ( 4.4%)

Discussion

PJI is a particular devastating complication after TKR. Carr et al. [1] found a trend toward performing more AKA for persistent TKR infection in a national database. AKA does not rely on achieving bony union at the knee joint, whereas arthrodesis may be difficult in cases of severe bone loss. There is also a risk of reinfection after arthrodesis due to the presence of hardware in a knee that may be still infected. 4 patients in the present series presented with reinfection after arthrodesis: two patients required AKA to control infection and sepsis, and the other 2, who were frail patients, were treated with chronic antibiotic suppression. In contrast, Hungerer et al. [4] showed a comparable functional outcome, quality of life, and postoperative complication rate between patients treated with AKA or knee arthrodesis after the septic failure of revision, and recommended that younger amputee patients in a proper physical and mental state may benefit from an AKA with proper orthotics, whereas in physically compromised older patients, arthrodesis may be the superior treatment. Chen et al. [14] found that patients undergoing fusion function mentally and physically better on the SF-12 than patients undergoing AKA. More than 90% of patients in the present series walked with weight bearing and functioned better after arthrodesis. Several scores have been used to report functional outcomes (Table 3).

Table 3.

Comparison of functional outcomes within literature

Study	Number of patients	Arthrodesis technique	Functional outcome
Friedrich et al. [8]	37	Modular nail	38 (mean Oxford knee score)
Mabry et al. [20]	85	IM nail (28%)	Not reported
		EF (72%)
Rao et al. [27]	7	Modular nail	40.5 (Mean Oxford knee score)
			SF-12: PCS 49.1 MCS 51.1
Klinger et al. [11]	20	IM nail (10%) EF (90%)	KOOS: 33% no difficulty, 50% mild or moderate difficulty SF-36 scores: “similar to those for TKR patients”
Corona et al. [18]	21	EF	SF-12 (age 64–74: PCS 42.8/MCS 59.3)
			SF-12 (age > 75: PCS 34.9/MCS 47.1)
Leroux et al. [22]	17	IM nail	WOMAC (26/88)
			Lequesne (10.5/24)
Kuchinad et al. [21]	21	EF (76%)	Ambulation:
		IM nail (24%)	EF group:93%
			IM nail group: 80%
Present study	45	IM nail (95.5%)	Overall:
		Modular nail (4.5%)	93.3% ambulatory status
			SF-12: PCS 39.5
			MCS 53

SF-12 questionnaire, PCS physical component score, MCS mental component score, WOMAC Western Ontario and McMaster Universities Osteoarthritis Index, KOOS Knee Injury and Osteoarthritis outcome score, IM intramedullary, EF external fixation

According to a National Database [15], clinical factors associated with arthrodesis included acute renal failure, obesity, and having additional infection-related revisions. A higher Charlson comorbidity score (5 or more), obesity, deep vein thrombosis, and additional revisions were factors associated with AKA. In another nationwide register-based study [4], 64% of patients who underwent a knee arthrodesis had one or more comorbid medical conditions at the time of primary knee arthroplasty (range 1–4), most commonly hypertension (26%), heart disease (16%), neurologic disease (15%), and endocrine disease (13%). In our series, the average Charlson comorbidity index was 4.5, with hypertension, diabetes mellitus, obesity, rheumatoid arthritis, and chronic obstructive pulmonary disease being the most common comorbidities.

In a systematic review, Wu et al. [6] found the two-stage approach resulted in infection control in 79%. Conversely, 12% of patients undergoing so-called “aseptic” revision had evidence of PJI that had either not been appropriately evaluated or had eluded the available diagnostic modalities. Bacteriological cultures should be observed for 10–14 days because of slow-growing organisms or small colony variants. It has been shown, for example, that Cutibacterium (formerly Propionibacterium) acnes and Peptostreptococci may be detected on routine culture plates only after incubation for between 10 and 12 days [16, 17]. In our study, 20% of PJI were culture-negative, 37% were polymicrobial, and 4.1% were infected with methicillin-resistant Staphylococcus aureus. In these patients, external fixation should be considered [3, 11, 18, 19]. In a National Register [5], the predominant microorganism isolated preoperatively or intraoperatively was Staphylococcus. Nine percent of infections were polymicrobial. Friedrich et al. [8] found 41% resistant pathogens, 49% polymicrobial infections, and no pathogen in 27% of patients. Spina et al. [18] observed a 23.5% rate of culture–negative PJI even when there were local conditions, laboratory (elevated inflammatory markers), or radiological (locally increased uptake in total body scintigraphy with marked granulocytes) results that were positive for infection. Finally, Mabry et al. [20] reported a 21% rate of polymicrobial infection, with Staphylococcus species being predominant alone or in combination. In our series, Staphylococcus species were also most common, but the incidence of polymicrobial infections was higher (37%). The culture-negative PJI rate in our series could have been minimized using sonication of the retrieved prosthesis [17] (application of ultrasound on retrieved implants to dislodge bacterial cells from the biofilm, followed by culture of the resultant sonicated fluid).

Gottfriedsen et al. [5] described that the mean number of surgical procedures prior to arthrodesis was 2.4 (range 0–9). The mean number of previous surgical procedures in the present study was 3.9 (range 2–9). Corona et al. [18] reported a mean of 3.4 procedures (range 1–15), and Kuchinad et al. [21] reported 2.33 previous procedures. We delayed performing knee arthrodesis, and considered this procedure to be a salvage treatment option.

All patients have shortening of the affected extremity following knee arthrodesis [22, 23]^. If symptomatic limb-length discrepancy cannot be satisfactorily treated with a shoe lift, the patient may undergo tibial lengthening over the nail with a monolateral fixator or exchange nailing with a femoral internal lengthening device [24] or with the Ilizarov method [21]. Reconstructive techniques that preserve length may include the use of distal femoral and proximal tibial sliding grafts (modified Putti–Juvara) [25] or metal cones with intramedullary autograft [26]. In the present series, mean limb length discrepancy was 2.4 cm (range 0–6 cm). In six patients, a distal femoral sliding graft (modified Juvara technique) was performed to minimize limb shortening (Fig. 3). However, a slightly shortened, fused leg allows for easier clearance during the swing phase of gait [17]. Modular nails have been advocated to avoid excessive limb-length discrepancy [8], as was performed in two patients.

Complication rates following knee arthrodesis are high, ranging from 20 to 84% [3, 20, 22, 28]. Our complication rate after arthrodesis was 37.6%, consistent with previous studies of knee arthrodesis using long IM or modular nails [8, 20, 22, 27, 29].

Röhner et al. [7] reported persistent infections in 50% of patients undergoing arthrodesis. In our series, 4 patients (8.8%) presented with persistent infection after arthrodesis. Mabry et al. [19] observed a rate of 8.3%, similar to our results, and Friedrich et al. [8] described an infection recurrence rate after arthrodesis of 13.5%. We agree that the success rate of arthrodesis for the eradication of infection remains unclear.

Counseling prior to arthrodesis and appropriate patient selection are essential to prevent unrealistic postoperative expectations and to minimize postoperative depression [3, 23]. Patients can expect a stable, painless extremity with difficulty climbing stairs and sitting. In the present study, more than 90% of patients walked with full weight bearing, which is a similar but slightly better result than in other series [21, 22, 28]. The success rate in achieving control of infection and fusion has been reported to be between 75 and 94%, being higher when IM nailing was used [16, 20] compared with external fixation [11, 18, 19] (Table 4). Our rate of fusion using a long IM nail was 91.1%. Internal fixation allows patients to more easily mobilize postoperatively, and the strength of the construct allows early weight bearing, but persistent infection following nailing is a potential hazard, as we observed in four patients.

Table 4.

Comparison of fusion rates within literature

Study	Number of patients	Arthrodesis technique	Fusion rate
Mabry et al. [20]	85	IM nail (28%)	95.8%
		EF (72%)	67.2%
Klinger et al. [11]	20	IM nail (0.5%)	100% (1 case)
		Modular nail (0.5%)	–
		EF (90%)	89%
Corona et al. [18]	21	EF	81%
Leroux et al. [22]	17	IM nail	94%
Kuchinad et al. [21]	21	IM nail (24%)	100%
		EF (76%)	93.3%
Gottfriedsen et al. [5]	152 (nationwide study)	IM nail (21%)	84%
		EF (73%)	61%
		Plate (6%)	85.7%
Spina et al. [19]	17	EF	76.4%
Schwarzkopf [28]	43 knees	IM nail (69.8%	81.5%
	(42 patients)	EF (4.7%)	0%
		Plate (20.9%)	77.8%
Present study	45	IM nail (95.5%)	91.1%
		Modular nail (4.5%)	–

IM intramedullary, EF external fixation

Finally, patients should understand preoperatively that, in most cases, a knee arthrodesis cannot be converted to a TKR at a later date [30].

Limitations of this study are the lack of comparative groups of patients and the retrospective design. The low rate of arthrodesis (2.7 cases per year at our institution) made it difficult to acquire a greater number of patients. When dealing with infrequently performed techniques it is difficult to perform a prospective study or a clinical trial. Thus, only retrospective studies or national databases are available.

Conclusion

Knee arthrodesis after a failed infected revision TKR has a high rate of fusion and postoperative ambulation, but the complication rate is high and persistent infection is a potential hazard. Based on our study, we believe knee arthrodesis should still be indicated after multiple revision procedures, if TKR infection cannot be eradicated or if multiple revision TKRs lead to loss of soft tissue or deficiency of the extensor mechanism.

Patients undergoing fusion function better after arthrodesis than before.

Joint resection should be indicated in nonambulatory patients. AKA should be considered in ambulatory patients when the infection persists after arthrodesis or in younger patients with proper orthotics.

Author contributions

GA made substantial contributions to the conception and design of this study, including collection of data,literature review, analysis, and drafting the manuscript. JO and SB participated in the design of this study sufficiently to take responsibility for the content.

Funding

None.

Compliance with Ethical Standards

Conflict of interest

The authors declare no conflicts of interests. None of the authors received or will receive direct or indirect benefits from third parties related to this study.

Ethical standard statement

This study was approved by the local Ethics committee or Intitutional Review Board (IRB).

Informed consent

Informed consent was obtained.