Presented as an alternative to fusion for salvage of the infected TKA by Kaufer and Matthews in 1986 [21], permanent resection or excision arthroplasty of the knee involves the removal of any prosthetic material and thorough debridement of the synovium, with no subsequent joint reconstruction (Fig. 20.2). The absence of any metalwork makes biofilm formation less likely. Patients are fitted with a cast or brace post-operatively, as a period of stability is required for the soft tissues to settle, prior to commencing flexion and extension as comfort allows. Whilst resection arthroplasty will generally allow a limited range of movement and is more likely to be comfortable when sitting, the development of a fibrous ankylosis is unsurprisingly less stable and has less load-carrying capacity than that of a solid fusion, making ambulation more difficult.

In 1987, Falahee et al. reported that infection was cleared in 25 out of 28 knees which underwent resection arthroplasty (89%) for infected TKA but that the subjective rate of dissatisfaction was 39% [22]. An important finding was that those with lower functional demands tolerated the procedure better. Lettin et al. reported better results in a smaller cohort of 15 patients: an eradication rate of 100% and subjective dissatisfaction at 20% [23]. More recently, Mine et al. found that there was no infection recurrence in nine patients following permanent resection arthroplasty combined with a pedicled muscle flap [24]. The subjective rate of dissatisfaction was 33%, and there was no apparent correlation between pain, limb length discrepancy, and patient satisfaction. Although it should be reserved in the first instance for patients whose functional demands are low, a resection arthroplasty may be converted to an arthrodesis at a later stage [21, 22].

AKA is generally considered the last option for the infected TKA but may be necessary in the presence of certain comorbidities, such as peripheral vascular disease. Although AKA will be discussed separately in the following chapter, there are two contrasting perspectives on choosing between knee arthrodesis and AKA following PJI: the first is that arthrodesis is preferable for younger patients with high functional demands and AKA should be reserved for older patients who are less likely to tolerate multiple limb reconstruction or staged salvage procedures; and the second is that younger patients can adjust far more easily to using an above-knee prosthesis and limb preservation should be attempted in older candidates, who are less likely to achieve functional independence.

Although each case must be considered individually, there is some evidence that appears to support the second theory. The energy expenditure for walking is 25% greater for AKA compared with arthrodesis, measured at 0.20 mL/kg/min and 0.16 mL/kg/min of oxygen, respectively [19]. Comparative studies have concluded that knee arthrodesis is associated with better function and ambulatory status than AKA following recurrent PJI [25–27], whilst indicating that AKA represents a greater psychological burden for patients. However, a recent investigation into the outcomes of 2634 patients who were arthrodesed and 5001 patients who underwent AKA for infected TKA suggested that, conversely, the first philosophy is perhaps being adopted in clinical practice [28]. There was a significant increasing trend towards AKA rather than arthrodesis over the 7-year study period, and patients who underwent AKA tended to be older. Systemic complications, in-hospital mortality, and length of stay were significantly higher in the AKA cohort, but without case matching, this may simply reflect the increased age and comorbidities of those patients.

Radical debridement is the imperative determinant of successful infection clearance and should not be compromised in favour of subsequent reconstruction, regardless of whether a one- or two-stage procedure for arthrodesis is undertaken. We recommend that the knee joint and IM canals are debrided according to a cyclical protocol, using principles developed by Lautenbach, as it may be difficult to achieve sufficient clearance of infection in a single pass [34–36].

All surfaces are surgically debrided of membrane, biofilm, and avascular tissue with curettage and rongeurs, and the IM canals are reamed successively under power to remove all persistent necrotic and infected tissue. Multiple intraoperative tissue samples should be taken with non-contaminated instruments. Once the bone surfaces and canals appear visibly debrided, normal saline is used as powered pulse lavage to wash out residual debris and to make any remaining membrane and biofilm oedematous. A second pass of curettage and mechanical debridement is performed to remove the residual oedematous membrane and biofilm, with successive cycles being undertaken as required. Chemical debridement may also be performed using antimicrobial agents such as 3% acetic acid, chlorhexidine, or povidone-iodine solution [34, 37].

Bone contact is a fundamental aspect of knee arthrodesis, and although debridement takes precedence, the minimal necessary bone resection should be performed where possible to allow vascular cancellous apposition and prevent excessive limb shortening. Interdigitation of the bone ends may improve stability, but multiply revised knees will inevitably have less bone stock, and cortical apposition might only be achievable in such cases. Bone loss can be assessed according to the Anderson Orthopaedic Research Institute (AORI) classification system [38, 39], in which tibial and femoral defects are scored separately, based on radiographic evaluation of metaphyseal integrity:

More recently, Klinger et al. have proposed an intraoperative classification system of bone loss, specifically for knee arthrodesis [40]:

Charnley emphasised the importance of cancellous apposition and compared this principle ‘to that of a fracture without displacement under ideal conditions’ [4, 5]. However, his series comprised native joints that were arthrodesed before the development of TKA, and improvements in revision techniques have meant that metaphyseal bone is now commonly deficient in salvage procedures. Fusion rates are lower in those who have undergone higher numbers of failed previous TKA revisions, as demonstrated by Hak et al. in a series of 36 patients who underwent arthrodesis with an external fixator [41] and Razii et al. with 12 patients who received a long IM nail [42].

Bone grafting can be performed for cases with extensive bone loss as an alternative to endoprosthetic implants. Autologous cancellous bone grafting is generally preferable to allograft, and it should also be placed around the periphery of the arthrodesis site to promote revascularization from the surrounding soft tissues. With multiple previous revisions, IM circulation of the bone is usually compromised following explanation of prior implants and the cement mantle.

Bone grafting techniques include vascularised fibular grafts – which can be free transfers or rotated on the vascular pedicle of the peroneal artery – and grafts from the iliac crest or patella [19, 43, 44]. Lee et al. reported using mixed autologous iliac graft and deep-frozen femoral head allograft in a series of eight infected arthroplasty patients who received an IM nail, and all went on to union [45]. More recently, the application of free vascularised fibular grafts has been discussed for revision limb salvage procedures following explantation of infected IM nails in failed knee arthrodesis [46], although the literature on these grafts has traditionally tended to focus on the management of bone loss following tumour resection.

The ideal position for limb alignment is with the knee in approximately 5–15° of flexion and 5–7° of anatomical valgus [15, 19], to allow sufficient ground clearance during the swing phase, with the initial contact in stance phase in slight external rotation to match normal gait. Slight knee flexion improves gait efficiency, due to the push-off direction of gastrocnemius [47], but with considerable bone loss, a position closer to full extension maintains as much length as possible. The knee should not be fixed in more than 20° of flexion, as the patient will have a short-legged gait which causes significant pelvic tilt and strain on the lower back.

Various configurations of external fixators have been developed since the Charnley clamp [4, 5]. Although the popularity of monoplanar external fixators has declined since the main indication for knee arthrodesis shifted towards the infected TKA, some recent papers have reported good results with this technique (Fig. 20.3 ). Roy et al. used the dynamic axial fixator (Orthofix SRL, Verona, Italy) in 24 patients for a variety of causes, including infected TKA, previous tuberculosis, and complex trauma [48]. All went on to union, although there were wound complications in six patients (25%). Balci et al. achieved union in 16 out of 17 patients (94%) following infected TKA [49], with a different unilateral external fixator (LRS; Tasarim Med, Istanbul, Turkey). Complications in addition to the one nonunion included pin-tract infections in 11 patients (65%), three fractures (18%), and one infection recurrence (6%).

Biplanar fixators provide greater stability than their monoplanar counterparts [14, 41] and gained popularity in the 1980s [50, 51], but they have largely been overlooked in the literature in recent years in favour of circular frames. Whilst a monoplanar fixator is less cumbersome for the patient and permits easier mobilisation, surgeons who would consider sacrificing this benefit to obtain greater stability might well prefer to use a circular frame rather than a biplanar fixator.

The majority of series describing knee arthrodesis with external fixation in the last two decades have used circular frames. Different options exist, such as the Sheffield Ring Fixator [52] or the Ilizarov frame (Fig. 20.4), which allows both rigidity and alignment of fixation to be adjusted, whilst also enabling limb lengthening if required. Circular frames have been reported as an improvement over previous methods of external fixation, allowing for accurate alignment in both the coronal and sagittal planes, but the operation is technically challenging [14]. Compared with IM nails and compression plating, however, removal of the hardware is relatively easy.

Reddy et al. achieved union in 15 out of 16 cases (93.75%) of infected revision TKA treated with the Ilizarov method [53], although pin-tract infections occurred in five patients (31%) and pin loosening in three patients (19%). Spina et al. reported union in 13 out of 17 (76.5%) infected TKA patients (12 primary and 5 revision prostheses) treated with an Ilizarov frame [54]. Oostenbroek and Roermund have suggested that this method of arthrodesis can be performed in the presence of active infection, having achieved primary fusion in 14 out of 15 cases (93%), but they did report an 80% incidence of complications, including pin-tract osteomyelitis in three patients [55].

Studies of knee arthrodesis with a long IM nail (Fig. 20.6) following PJI have reported high fusion rates and infection clearance of around 80–100%, even in the presence of severe bone loss [31, 42, 60–63]. The majority of these case series have described the results of a two-stage procedure, but it is possible to perform single-stage arthrodesis with an IM nail for selected patients [42, 64], if the recommendations of the International Consensus on Periprosthetic Joint Infection relating to revision TKA are met [13] and the principles of debridement are strictly adhered to [34]. The Biomet Arthrodesis Nail (Biomet Inc., Warsaw, Indiana, USA) has shown good results in the literature for more than two decades [42, 60, 62] but is no longer manufactured. Similar devices, however, such as the TRIGEN Knee Fusion Nail (Smith & Nephew plc, London, UK) and the T2 Knee Arthrodesis Nail (Stryker Corp, Kalamazoo, Michigan, USA), are currently available [31, 63].

The surgical technique for a long IM nail involves reaming of the tibia from within the knee wound, with the tibial canal diameter determining the IM nail diameter. The femur is reamed retrograde and then antegrade through a separate proximal incision over the piriformis fossa. Femoral and tibial canal lengths are measured separately, with the sum of the canal lengths used to assess the required nail length, although surgeons can also estimate this pre-operatively from long-leg radiographs. The nail should span from the level of the greater trochanter to 2 cm above the ankle and locked proximally to prevent migration, a recognised complication in unsecured nails [60]. Distal locking screws are generally unnecessary, as three-point fixation over the length of the nail avoids rotational instability whilst still allowing dynamic axial compression [42, 61].