8.8 Revision total knee replacement

No aspect of knee arthroplasty surgery has advanced more in the last decade than revision surgery—although the comparatively few revisions performed have left much of this information unappreciated in the literature. Repeat surgery was initially not considered feasible, and was then performed (and reported) identically to primary replacements. Eventually, special implant techniques and prosthetic systems emerged. A strong argument can be made that (with the exception of cases of infection and extensor mechanism allograft) the results of contemporary first revision arthroplasty, with the techniques described in this chapter will prove at least as, if not more, durable than current primary replacement. The same cannot be said for a second knee revision, emphasizing the importance of the initial opportunity. Given current population demographics and primary arthroplasty rates, a significant increase is expected in the need for revision knee arthroplasty.

A detailed understanding of how knee replacements fail is central to planning a successful revision. This means that a clear diagnosis and a deep appreciation of pathophysiology and biomechanics are essential. This information comes from a comprehensive literature, which cannot be quoted in depth in this format. However, the reader is directed to a few pertinent review articles which are a broad guide to the published data.

New appreciation of the rotational positioning of tibial and femoral components, quantifiable by computed tomography (CT), has implications for all patellar complications and not just tracking, instability, stiffness, and pain. Revision implant systems differ fundamentally from those that work well in primary surgery, with stem extensions, modularity, porous metals, and constrained articulations that enable surgeons equipped with the correct information, to help patients who were once consigned to pain clinics, or worse, arthrodesis and amputation.

The role of revision surgery is not simply to treat the older patient with a ‘worn out’ arthroplasty. Rather, it is the key piece in a strategy to ensure that the younger active patient, who may now be receiving a unicompartmental replacement, can be assured that a functional knee joint will be present through a long and active life.

Several principles underpin revision knee arthroplasty surgery:

There is a coherent argument for eight modes of failure (see Figure 8.8.1) that are amenable to revision arthroplasty and a ninth category (unexplained pain) that is an indication for further investigation and temporizing, but not surgery. All the cases within one diagnostic category will present with similar problems and respond to similar techniques. All can be managed with the ‘3-step technique’ described in this chapter.

Early challenges and catastrophes resulting from difficult surgical exposures led to the development of special, more involved techniques to avoid damage to the extensor mechanism and yet provide full access to the joint. Some of these divide the soft tissue of the extensor proximally and others detach the tuberosity distally. Initially, the very old ‘Coonse–Adams’ quadriceps tendon ‘turn-down’ was adapted to knee arthroplasty surgery (Figure 8.8.3A). While sparing the patellar tendon attachment, it is aggressive and now largely obsolete, except for the most extraordinary circumstances. More limited manoeuvres, such as the ‘quadriceps snip’ or quadriceps ‘split’ often suffice (Figure 8.8.3B). An oblique (inferomedial to proximal lateral) arthrotomy of the quadriceps tendon has been described, but does not differ substantively from the ‘snip’. Any transection of the quadriceps tendon close to the patella should be avoided.

Some surgeons prefer proximal soft tissue procedures and others the tubercle osteotomies. The osteotomy, while providing superb exposure, is best employed where adequate tibial bone quality ensures union. This is often not the case in the difficult revision, with intramedullary stem extensions and methacrylate cement (Figure 8.8.3C). Calamity may ensue when an osteotomy fails, as the extensor mechanism is in jeopardy.

The danger of forcing exposure is primarily to the extensor mechanism, where avulsion of the patellar tendon from either the tubercle or the patella is irreparable. The common practice of placing a pin through the tendon and into the tubercle does not reduce the force applied by an aggressive approach and risks tearing the tendon when enough force is applied. ‘Peeling’ of the insertion, though not universally accepted, is preferable to a transverse tear.

Collateral ligaments may suffer from forceful exposure where they are torn or ripped from the condyles. Though inelegant, this can be corrected with constrained implants and indeed a purposeful removal of the collaterals from femoral condyles has been described for arthroplasty in the stiff knee, where it is referred to as a ‘femoral peel’.

The best new information regarding surgical exposure in revision knee arthroplasty emphasizes the importance of patience and attention to detail. No aggressive manoeuvre should be invoked until there has been close attention to detail. The vast majority of cases, including stiff arthroplasties and reimplantations for infection can be exposed fully after:

It may not be necessary to fully evert the patella for revision surgery (Figure 8.8.4).

Numerous techniques and ‘tricks’ have been described for the removal of well-fixed arthroplasty components. A cautious approach that preserves bone is sound, but time must also be conserved as infection rates increase with longer surgery.

In general, the former sequence of component removal from cement followed by cement removal from bone can be expedited, in favour of sawing between the cement and bone. Simple techniques with a minimum of specialized equipment are useful. Very few components cannot be removed with standard reciprocating and oscillating saws, a narrow curved osteotome, half a dozen old flat osteotomes, mallet, punch, and perhaps one offset chisel.

The femoral component is immediately more accessible and once off makes it simpler to remove the tibia. A ‘reciprocating saw’ fits easily into the anterior, distal, and chamfered interfaces of both cemented and uncemented femoral components (Figure 8.8.5). A narrow curved osteotome reaches between the posterior condyle and bone, where the saw cannot go. A mallet blow via a ‘punch’ placed on the most superior edge of the femoral flange will easily dislodge the component once the interface is disrupted. Gigli saws, once favoured by some, may be difficult to control—excess bone is removed if the saw strays. ‘Axial slap hammer extractors’ once considered essential are not always available and will be difficult to attach to (right and left) asymmetric femoral components. They tempt the surgeon to remove the component (and consequently bone) with force before fixation has been disrupted.

Special situations arise where femoral components must be removed (revision of a revision) that are attached to stem extensions. In general, if the modular junction can be unlocked by the removal of locking screws, it will be advantageous to remove the component itself first. This exposes cement around the stem extension. Uncemented stems may come out with the component but if not, removal of the component enables the surgeon to work on the stem extension fixation directly. Once this has been done, the component may be reattached to the stem, providing an excellent grip. Alternately, only the axial screw may be reattached to the stem and a ‘slap hammer extractor’ applied to the screw and stem. Ultimately, a fully cemented long stem (which is not recommended for precisely this reason) may require an extensive anterior femoral window to remove cement and extricate the stem.

Tibial components may be approached with the large oscillating saw first, cutting between the bone and the cement to save time. Access will be limited by appurtenances (keels, pegs, spikes) on the underside. At that point the reciprocating saw, with its narrow blade, helps but often cannot access the posterolateral corner, where an offset chisel works well, introduced from the medial side. The keel, though still fixed, will usually yield to ‘stacked osteotomes’ that gradually lift the component out of the residual cement mantle (Figure 8.8.5D). Special circumstances, such as extrication of the fully cemented porous-coated stem extension—now rarely encountered—may require creation of windows in the metal tibial surface to access the fixation interface.

Patellar components, if made exclusively of polyethylene are easily removed. An oscillating saw cuts between the cement and the bone, severing whatever posts penetrate into the patella. These are quickly dispatched with a burr, which can also be used to enlarge the diameter of the post hole and so remove all methacrylate, important in the treatment of the infected arthroplasty. Uncemented metal-backed patellar implants may be more problematic. It is worth hammering a narrow (¼-inch) osteotome into the interface. The thicker portion of the instrument will disrupt fibrous ingrowth. True bone ingrowth may require a ‘diamond wheel’ cutting device on a ‘high speed, low torque’ hand piece to disrupt the interface and sever the ingrown posts. These can then be removed with a ‘pencil’ tipped cutting head on the same burr.

The key to revision knee arthroplasty surgery is simple: ‘the femoral component controls the soft tissues’. This means that knee kinematics, the essential balance of stability and motion, can only be managed if the femoral component is revised. Furthermore, the flexion gap is controlled by the size and to a lesser extent, the position of the femoral component, while the extension gap is the product of the proximal distal position of the femoral component. Conventional soft tissue releases, integral to correction of deformity in primary arthroplasty, have a modest role in revision surgery—useful only when releases should have been performed in the primary, but were not.

The ‘3-step technique’ that follows is based on these principles. The technique is really only practicable if diaphyseal engaging stem extensions are used, preferably for the entire technique but certainly for the trial reduction. These stems establish and maintain trial component position independent of bone deficits. There are few primary instruments that work well in revision knee arthroplasty. It is almost dangerous to expect primary instruments to guide the reconstruction as the osseous landmarks that they reference from are no longer present. If the surgeon prefers fully cemented, shorter, narrower stems, they can be selected at the time of implantation—the longer tighter fitting trial components will have guided position and selection of augments.

Cementless fixation has not supplanted methacrylate or the combination of methacrylate and ‘press-fit’ intramedullary stem extensions in general usage, as has been the case for revision total hip arthroplasty. A few committed practitioners have described techniques for revision knee arthroplasty without cement fixation.

Fully cemented stem extensions undoubtedly provide superb fixation (Figure 8.8.6A). They are, however, difficult to remove and by neither ‘fitting’ nor ‘filling’ the canal, they do not accurately guide position of the components. The arguments and literature against uncemented stem extensions (Figure 8.8.6B) is based on studies that have compared fully cemented stems with metaphyseal length uncemented stems. Clearly, the metaphyseal length stem is neither long enough to guide implant position nor improve fixation. Similarly, one would not expect the ‘dangling’ stem or one that does not contact the endosteum to aid fixation significantly.

If uncemented stem extensions are selected, cement that is limited strictly to the cut bone surface is probably inadequate as it does not provide fixation comparable to what is typically used in primary arthroplasty. The strategy that provides excellent fixation and alignment is the combination of asymmetric (to accommodate anatomy) diaphyseal length, modular stems with cement fixation extended a few centimetres, into the metaphysis (Figure 8.8.6C).

The tibial component, comprising one side of both flexion and extension gaps, is not useful for manipulating the dimensions of one gap relative to the other. It is, however, a ‘platform’ or ‘foundation’ from which the dimensions of both flexion and extension gaps can be established. Component position and fixation in the revision is not predicated on precise bone cuts as in the primary arthroplasty, but rather on a tight and accurate fit of an offset stem extension in the tibial canal, plus component contact with a majority of host bone. Small incongruities will be filled with cement. It will be the technique of inserting the stem extensions that determines component position. As a result, the sequence for preparing the tibial surface is:

Unless the primary knee has failed from varus alignment and loosening, most tibial components will, in this era of sound instrumentation, have been well aligned. As a result, the saw cut to remove the component should reproduce a reasonable alignment, including posterior slope, even without an instrument, by following the interface. Some surgeons will prefer a cut with minimal or no posterior slope for simplicity, but this is more likely to resect anterior bone and jeopardize the tibial tubercle. As is the case with much revision surgery, instruments that work well in primary surgery may not at revision. In the case of the tibia, extramedullary cutting guides cannot be pinned, between the existing tibial surface and the tibial tubercle and still enable resection of some bone. Pins hold poorly as the existing bone is deficient, soft, or sclerotic. Intramedullary guides are more promising in this situation, relating as they do, the position of the stem and that of the component.

The purpose of reaming is not to remove much tibial bone, but rather to measure the endosteal diameter as an indication of the best choice of stem extension. It paves the way for the insertion of a trial tibial stem and many systems may allow attachment of a tibial cutting guide to the reamer for the fine cut described earlier. The position of the reamer will ultimately be the position of a tightly fitting stem extension and as a result, the alignment of the component.

The correct reaming technique will vary with the design of the implant. However, two anatomical points are almost universally true: 1) the centre of the tibial diaphysis (on an AP radiograph) is medial to the centre of the cut bone surface and 2) the canal is anterior (on a lateral radiograph) to the centre of the cut tibial surface. For the reamer and stem extension to sit centrally in the diaphysis, the entry hole must be anterior and medial. The natural asymmetry of the tibia can be replicated expeditiously by placing an offset tibial stem extension in the canal medially, with the component oriented laterally on the cut bone surface above. When this trial component is inserted, the stem should fit nicely in the canal, the component will be centred on the cut bone surface, and neutral alignment preserved (Figure 8.8.7).

Applying the trial component upside down to the proximal tibia is a quick way to assess component sizing. Tibial templates that position punches and drill guides often do not fit on the irregular bone surface or cannot be held by pins in bad bone. They can be dispensed with in many cases, and the proximal tibia shaped with a rongeur or broach.

Maximizing tibial coverage is far less important than achieving correct rotational positioning and soft tissue coverage. Surgeons intent on maximizing tibial coverage will be tempted to rotate the component internally, to cover the larger posteromedial corner of bone. Revisions of the stiff knee may be complicated by an uncomfortably tight closure of the medial capsule—a situation that can be improved by undersizing the tibial component slightly and removing a small amount of medial tibial bone.

We can build the flexion gap, with the tibia as a foundation, and then match the extension gap to it. With the knee flexed to 90 degrees, we will: a) establish the rotational position of the femoral component; b) select the size of femoral component (anteroposterior dimension) to restore stability through the collateral ligaments when possible; and c) evaluate how this femoral component and a tibial insert affect the relationship between the patella and the joint line.

Ideally, the medial to lateral axis of the distal femoral component will be placed parallel to the transepicondylar line. The epicondyles, obscure in ideal circumstances in the primary replacement, will be more difficult to locate in the revision. Nonetheless, they are the best indicators of correct rotational positioning, especially as there will be no posterior articulation and no trochlear groove or ‘Whitesides line’ to reference from. Preoperative CT will prove reassuring at revision surgery: the rotational position of the failed femoral component will have been quantified preoperatively and either replicated if acceptable or corrected if not.

Another, somewhat vague guide to rotation will be the residual posterior condylar bone, palpable in the back of the femur. Once the internally rotated component is removed and assuming symmetric bone loss, there will be more residual bone on the posteromedial side. This finding indicates a need for a posterolateral augmentation block to correct internal rotation. Finally, if rotation is off, the patella will track laterally; another indication for a corrective posterolateral femoral augment, assuming the tibial component has been oriented to the tibial tubercle.

An expedient way to visually and mentally commit to the desired rotational position for the femoral component is to create the intercondylar box cut at right angles to the epicondylar axis before reconstructing the flexion gap. Virtually every revision knee component will require some type of box cut and a minimum of instrumentation is required. The width of the box can be marked on the distal bone and a modest depth created. By the technique described here, we will not yet have decided where to seat the femoral component, proximally to distally until Step 3. The box bone, once removed, can be used as graft, whereas if the canal is drilled or reamed early, this bone will be destroyed. The notch, once created, is an unmistakable reminder of the desired rotation.

The first of our ‘keys to revision surgery’ has not yet been implemented: the flexion gap will be ‘controlled’ by the (anteroposterior) size of the femoral component and to a lesser extent by its position. The largest component that can be sensibly implanted will be the one that extends from medial to lateral across the distal femur—any larger and it would overhang unacceptably. The maximum size might be necessary in the revision of a knee with flexion instability, where the flexion gap needs to be reduced. In most cases a size similar to the failed component will suffice. Sometimes (especially in revising the stiff knee with poor flexion) one or two sizes smaller may be appropriate. An easy way to assess the largest usable component is to place the trial component, articulation against distal bone, to compare medial lateral dimensions. It should be clear that the femoral component is never selected primarily for its medial–lateral fit, but rather for the effect that the anteroposterior dimension has on the collateral ligaments.

When the femoral canal is reamed, there is a tendency to enter the bone posteriorly and to lower the arm holding the reamer. As a direct result, the reamer assumes a ‘flexed’ angle relative to the medullary canal. This leads to impingement of the tip of a long press-fit stem on the anterior endosteum, which may give the impression of a tight press-fit, but which actually limits the diameter of the stem extension to something smaller than would be accomplished with parallel implantation. Most revision knees, if the femur is reamed parallel to the anterior cortex, will require some form of anterior augmentation on the femoral component (Figure 8.8.8).

Press-fit stem extensions ultimately become ‘implantable instruments’. By lodging in the diaphysis, they ascertain alignment relative to the (distal half) of the femoral canal. Even if a surgeon prefers shorter, smaller diameter stems that will be fully cemented, using longer (i.e. 200mm for the femur) trial stems at this stage, as a means of determining alignment, position, and accordingly, the necessary augmentation, will help achieve the desired component position. The reaming technique can be manipulated (described later), to alter intramedullary stem position and arthroplasty alignment. If greater modification is required, then smaller diameter fully cemented stems will be required. At this point the stems in the femoral canal position the femoral trial reproducibly. A trial tibial articular polyethylene is required to stabilize the joint in flexion.

If the knee cannot be stabilized in flexion despite the use of the largest possible (extends from medial to lateral) femoral component, a very thick articular polyethylene and perhaps even block augments on the tibial baseplate, it is clear that the collateral ligaments are absent or have suffered plastic deformation. This is the first of two, specific indications for a constrained implant. If the extent of instability in flexion exceeds the ‘jump distance’ of a non-linked constrained articulation, this is an indication for a ‘linked constrained’ device or ‘hinge’.

With the femoral component size and anteroposterior position established, the tibial articular surface that provides stability will also create a ‘joint line’ in flexion. There should be no confusion about the concept of joint line—it is not simply a two-dimensional ‘line’, but rather a three-dimensional ‘shape’ corresponding to where the tibia and femur articulate—from full extension to maximal flexion. For the purpose of revision knee arthroplasty and assuming a relatively ‘normal’ patellar tendon length, the joint line can be evaluated relative to the inferior pole of the patella.

A sound revision can be created from more than one size of femoral component, even if positioned identically in the anteroposterior plane (as viewed on a lateral radiograph). The flexion gap can be stabilized with a larger femoral component and a thinner articular polyethylene or alternately with a smaller femoral component and a thicker polyethylene. Both will be stable, but in the latter scenario, additional distal femoral bone would need to be resected to accommodate the thicker polyethylene in extension. The sequence of events becomes: smaller femoral component, more spacious flexion gap, thicker polyethylene, higher joint line in flexion, additional distal femur removed to accommodate thicker polyethylene, and so higher joint line in extension. The preferred combination of femoral and tibial components will be that with the best ‘joint line’. Our recommendation is that this choice be made relative to patellar height and not a landmark on the femur or tibia (Figure 8.8.9).

This final step is simple and invokes the second key to revision arthroplasty surgery: the extension gap is controlled by the proximal–distal position of the femoral component. At this point all the components have been selected and positioned on press-fit stems. The tibia cannot move, but by slowly and gently extending the knee, the femoral component will be pushed into the distal femur to precisely the point that it needs to be in order to make the extension gap equal to the flexion gap. If we push too far and place the knee in recurvatum, the femoral component will have migrated too far proximally into the femur. If we fail to eliminate a flexion deformity, the femur will be ‘too proud’ on the distal femur and the deformity will persist. Trial femoral components that include cutting slots facilitate this step. Small cuts on the residual, irregular bone once the knee is fully extended will create, millimetre for millimetre, the correct space to accommodate an augment that guarantees the position we require (Figure 8.8.10).

The correct femoral component proximal–distal position in this step is actually determined by the posterior soft tissues—capsule, hamstrings, and gastrocnemius. So, even if there has been collateral ligament failure, we may have the impression of full, stable extension because the fully extended knee is stabilized by the posterior structures. This is also the second point at which it may be necessary to select a constrained prosthesis—if varus–valgus instability persists as a result of collateral ligament failure. It may be necessary to assess the knee in a few degrees of flexion to eliminate the confounding effect of the posterior structures.

The earliest revision knee arthroplasties were performed with primary implants under challenging circumstances. With neither stem extensions nor augments in that era, surgeons were almost obligated to select the femoral component size that would fit the residual bone, once the failed implant had been removed. As we know from Step 2c of the revision technique described in this chapter, a smaller femoral component would necessitate a thicker polyethylene insert and inevitably an elevated joint line in flexion. To stabilize this knee, distal femoral bone loss either would have to be ignored to accommodate the tibial insert or, perhaps, distal femoral bone might be resected to achieve full extension. The results of early revisions were poor and the most glaring abnormality on postoperative anteroposterior radiographs was bizarrely thick polyethylene and an elevated joint line. The first efforts to remedy this situation led to a surgical technique that began by establishing the distal position of the femoral component, i.e. the (distal) joint line (or the joint line at full extension) as a preliminary reference point for the entire reconstruction. This was usually a calculated distance from an osseous landmark—often the fibular head. This strategy, though an improvement, did not address stability of the knee in flexion or recognize that there is a joint ‘line’ for every position of the knee from full extension to maximal flexion, because of the state of the soft tissues.

An illustration of the challenges of revision knee arthroplasty is to visualize a very badly failed knee replacement—perhaps a second revision or even a resection arthroplasty about to be reimplanted. The cruciate ligaments will have been long gone and the soft tissue balance is likely to have suffered plastic deformation or contracture. The hypothetical questions then become: if all normal bone and articular cartilage were restored in such a knee, and the compromised soft tissues left as they are, would this knee function well? Would it be both stable and mobile? After all, the original anatomical joint line would have been restored perfectly in all three dimensions. The reality is that the state of the soft tissues limits the ability of such an anatomical restoration of the joint line to work well in every case.

In contrast, a surgical technique that accepts that the envelope of soft tissue around the knee joint is likely to have changed, and that is based on the selection of femoral component (controlling as it does soft tissue tension) is much more likely to be succeed. Revision arthroplasty then becomes very much like pitching an unfamiliar circus tent, where poles are brought inside the collapsed structure and the outer fabric is erected.

Simply stated, the joint ‘line’ is not a line but rather a complex three-dimensional construct corresponding to the contact of the tibia with the femur through an entire arc of motion. It corresponds to the shape of the femoral component articular surface. There will be a joint line in flexion and extension and everywhere in between. Even restoration of a joint line that is normal in three dimensions may not produce a good joint. Pragmatism with respect to the state of the soft tissues is needed.

The patella is best reserved for the end of the case because once the old patellar implant is removed the patella itself is more vulnerable to fracture. A treatment algorithm can be established (Figure 8.8.11) initially based on whether the implant is loose. A well-fixed patellar component that articulates compatibly with the revision femoral component should be left in place, unless the revision is being performed for stiffness and the entire construct is inordinately thick. Patellae with loose components may have adequate bone that would make revision feasible. If the residual bone is thin and perhaps scaphoid in shape (if only the anterior cortex is left) then reconstitution with particulate bone graft and a tensor fascia lata cover can restore thickness. Alternately the residual patella can be split from superior to inferior and both halves of the shell ‘cracked up’ to recreate a more conforming shape—the so-called gull-wing osteotomy. Newer porous metal implants, though more expensive and time consuming, produce excellent results and have largely supplanted the first two options. If the patella is too small, it may be left as is or reshaped slightly to enhance tracking. The patient with an extremely small or no patella, who suffers an extensor lag postoperatively, may benefit from an extensor mechanism allograft.

Infection continues to be one of the most perplexing calamities to complicate knee arthroplasty surgery. The infection rate should be no greater than 2% and in most centres 0.5–1.0%. An aggressive and organized approach to the management of wound problems is essential to mitigate the risk of deep sepsis. The surgeon should also surrender early to the idea that the arthroplasty might be infected and act decisively so that a diagnosis will be established and not simply obscured by undisciplined use of antibiotics.

This means that all wound problems be taken seriously, as many cases of deep sepsis originate with superficial problems. Whilst many surgeons hesitate to aspirate an arthroplasty for fear of introducing infection, it is only by knowing what is inside the joint that therapy is likely to prevail. There is a strong rationale to aspirate any arthroplasty with late pain or an early wound problem before any antibiotic is administered. Introducing antibiotic without an aspirate compromises the prompt diagnosis of sepsis.

The knee presenting with pain months to years after the primary, whose history does not specifically arouse suspicion of infection and whose peripheral blood test for erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) is normal, may not require aspiration. When fluid is aspirated it should be sent not only for culture and sensitivity, but also for a cell count and differential. Solid evidence exists in the chronically painful knee, that white blood cell (WBC) counts in excess of 2500/mL, especially where there are more than 50% polymorphonucleocytes are consistent with deep sepsis. This may not pertain in the first few weeks after surgery.

There is almost universal agreement that a two-stage reimplantation protocol cures more infected knee replacements. Nonetheless, the appeal of ‘single-stage’ reimplantations is undeniable and is the practice of some surgeons. The debate lies in the relative success rates of both procedures. Common to both procedures is the importance of a precise bacteriological diagnosis and antibiotic sensitivity as well as an aggressive and complete debridement. It is good practice to obtain a radiograph (that images to the ankle) in the operating theatre with the patient still anaesthetized, once the surgeon feels that all foreign material has been removed. This often reveals more cement, frequently in the patella lug holes or the tibial metaphysis. Ultimately, and judicially, a radiograph that demonstrates ‘no foreign material’ is compelling evidence that the surgeon has understood and complied with a principle of treating infected arthroplasties: that all foreign material be removed. In addition, this radiograph will show bone deficiency in a way that is not feasible once a spacer has been implanted, and facilitates planning the reimplantation.

Two-stage reimplantation protocols represent the standard of care in most countries. Eradication of sepsis and acceptable knee function have been achieved with, in general, a 6-week period after the infected arthroplasty was removed during which the patient receives antibiotic therapy. Most surgeons use some type of antibiotic-impregnated methacrylate spacer and good results have been described with solid (non-articulating) spacers as well as those that resemble an arthroplasty and permit some motion. Some surgeons endorse autoclaving and reimplanting one or both of the implants that have been removed using limited fixation and antibiotic-impregnated cement. The rationale is to maintain motion and facilitate the reimplantation. To date, no compelling evidence exists that would suggest that any particular approach to ‘spacers’ has either better function or increased survival. While early reports on two-stage protocols included reimplantation at 6 weeks, immediately after antibiotic therapy was concluded, recent practice has included waiting, with the patient off of therapy for durations of weeks to months and reaspirating prior to reimplantation. No data demonstrate superiority of that concept to date.

Reimplantation surgery will challenge even the most experienced surgeon. Careful attention to detail in the surgical exposure (see earlier) will pay dividends. Tubercle osteotomies in this situation may be problematic—especially if performed for both the removal and reimplantation. Bone quality is compromised and any hardware, even wires, left to hold the osteotomy runs against the principle of removal of all foreign material. The tubercle, ultimately implanted against stem extensions may not heal solidly, representing then a case of prior sepsis and extensor mechanism rupture.

The aggressive debridement, necessarily including curetting of the canal and wholesale removal of suspicious bone often leave the surgeon with challenges of bone defects and fixation. Porous metal augments advantageously address both issues and are very useful in reimplantation surgery. Fully cemented stem extensions have the advantage of excellent fixation and can include antibiotic, though less of a dose than can be used in spacers, because of the compromise to the mechanical properties of the cement. If the reimplantation suffers persistent or recurrent infection, removal of fully cemented stems may be a destructive prelude to amputation.

My preferred technique is a disciplined approach to establishing a bacteriological diagnosis, aggressive debridement on the removal, reimplantation at completion of the course of antibiotics without cessation of antibiotic therapy, long uncemented stem extensions and porous metal conical augments to enhance fixation whether large defects are present or not.

Acute or intraoperative ruptures of the extensor mechanism, often become chronic extensor ruptures until some tissue substitute, either an allograft (complete tibial tubercle, patellar tendon, patella, and quadriceps tendon construct) or autograft (semitendinosis) has been introduced. Ruptures anywhere from the tibial tubercle to the superior pole of the patella suffer similar results and this includes transverse, displaced patellar fractures with an extensor lag. This last entity should be removed conceptually from the category of fracture and treated as a chronic extensor rupture.

This is generally regarded as one of the most difficult problems to treat. The extensor rupture often occurs in an arthroplasty that is in some ways unstable. Then, when the extensor fails, global instability or posterior tibial dislocation (in flexion) often ensues. At other times, the patient with a chronically ruptured extensor will hyperextend the knee to prevent buckling. With time, this results in a recurvatum deformity. Maltracking is a common accompaniment to any patellar problem. Accordingly, revision knee arthroplasty is usually required when the extensor mechanism is reconstructed. Extensor mechanism allografts are used more frequently than other techniques and are generally successful in the short term if the graft is implanted under considerable tension. This is most likely due, not to stretching of the graft itself, but chronic retraction of the quadriceps up into the thigh during the period of rupture. If the graft is placed loosely, when the knee bends ands stretches the muscle, the lag recurs.

The stiff knee replacement is often accepted, with disappointment, by both surgeon and patient. Indeed, the patient who had very poor flexion prior to the arthroplasty probably has an extensor that is chronically scarred and tight. This cannot easily be completely overcome with revision surgery or therapy. In addition, the patient who truly has ‘arthrofibrosis’ and mounts an aggressive scar response to any surgery is unlikely to respond well to revision arthroplasty. Fortunately this is a very small minority of individuals. The majority of patients with good preoperative flexion who either bend less then 105 degrees or develop a flexion contracture exceeding 10 degrees will probably benefit from revision surgery, in particular if there is also pain. The scar that the surgeon confronts at surgery is then the result and not the cause of stiffness.

Revision is demanding for surgeon, patient, and the rehabilitation team. Once stiffness is chronic, very little will succeed short of a complete revision operation, with aggressive soft tissue release, and alteration of every feature of the knee away from stiffness towards mobility. For example, femoral components usually need to be reduced in size and implanted in neutral position (as viewed on the lateral radiograph), not flexed where they encroach on the flexion gap. Internal rotation of femoral and tibial components will be observed in the majority of stiff knees arthroplasties. The resultant tendency to patellar dislocation inhibits many patients from flexing past about 60 degrees. Those who succeed in regaining motion suffer patellar dislocations—those that don’t become stiff. CT scanning, prior to revision, to quantify rotational position is urged. Patients who undergo expert and complete revision arthroplasty, with constructive alteration of component position and size may expect welcome pain relief, full extension and flexion on average to about 100 degrees.

The unstable knee arthroplasty is at once the most intellectually challenging and, if approached successfully the most gratifying. Central to the problem of instability are four questions:

The frequent and unfortunate strategy to instability is often only: ‘What constrained implant should I use?’ If the deforming forces are not reduced, instability will recur, often with dire consequences.

The patient who reports ‘that their knee arthroplasty is unstable is suffering from one or more of the following:

Some have postulated that a linked constrained device (hinge) will be necessary if the collateral ligaments have failed completely and that a non-linked device will suffice if some remnant of soft tissue envelope is present. While this may be a useful guide to some, it will fail if all the factors listed here are not taken considered.

Fractures requiring surgery are more common in the supracondylar region of the femur than any part of the tibia. Patellar fractures, if transverse, represent a chronic rupture of the extensor mechanism and belong in ‘Mode of failure 2’. Vertical patellar fractures can be treated non-surgically at first, to allow swelling and acute pain to subside. If a lag persists, or if the patellar component of a resurfaced knee is loose, surgery will be required. In most cases of vertical fracture, there will be a tendency to maltracking and these cases belong in ‘Mode of failure 7’.

Supracondyar fractures usually occur in the patient with: a) osteoporosis; b) a stress riser (in the form of a supracondylar notch), and c) limited flexion so that during a fall the knee reaches its limit of motion and then the porotic bone yields. These fractures have been treated successfully in the past with both blade plates and plates with sliding screws. This was superseded by supracondylar locking nails and most recently have been replaced by locking plates with or without minimally invasive approaches.

Some low-demand patients with comminution and poor quality bone may require a revision arthroplasty either with a distal femoral allograft or more typically a tumour-style distal femoral replacement implant. Tibial fractures are generally the late sequel to extensive osteolysis and tibial component loosening. These require revision arthroplasty, not fracture fixation.

Aseptic loosening was once thought to result from ‘micromotion’ at the interface that gradually increased in amplitude until the component came loose. Alternately, it has been attributed to bone overload and collapse. It has been correctly associated with varus alignment and the results of medial compartment overload since 1977. While some uncemented or partially cemented tibial components loosen early from failed ingrowth or inadequate fixation, and some femoral components loosen from gaps that are left between the posterior femoral bone and the femoral flange, virtually all late loosening will result from wear, particle generation, osteolysis, bone loss, and subsidence.

While partial revision with bone grafting may be appealing, the process of osteolysis is likely to progress through the joint and loosen the other components. Accordingly, complete revision is almost universally recommended. As with the unstable arthroplasty, causative factors must be identified and corrected specifically. Varus alignment is problematic, inducing inordinate forces on the medial component. A full-length radiograph showing hips, knees, and ankles is important; not only to appreciate the mechanical axis, but to plan how much valgus alignment can be tolerated to unload the medial compartment at the revision. Wear results not only from the articular surface, but the modular interface. A substandard modular locking mechanism and poor quality polyethylene may be implicated and should be replaced. CT scans depict the extent of osteolytic lesions with distressing accuracy.

Failures from osteolysis and loosening have constituted the majority of revision arthroplasties. Loosening of the revision is likely unless realignment reduces destructive forces, and fixation is supplemented by stem extensions that are either fully cemented or uncemented and fill the diaphyseal canal. The interface itself can be a problem—methacrylate cement against poor quality, sclerotic bone enjoys no interlock. By contrast, porous metal augments enhance fixation by providing a transition of interfaces from poor bone (that will yet grow into porous metal) and porous metal that on its other side accepts bone cement which provides an excellent bond to the implant. The porous implants can be impacted into the medullary canal of the tibia or femur without concern for rotational, flexion–extension or varus valgus position, so long as the augment does not impede the positioning of the component. Structural allografts, useful for massive bone defects are required less frequently as porous metal is commonly available.

There is an ineluctable relationship between patellar maltracking and internal rotation positioning of the femoral and or tibial components. The relationship is additive, not compensatory. Furthermore, the effect of maltracking is responsible for virtually every patellar complication: as the patellar bone tries to slide off laterally and the polyethylene component stays in the trochlear groove, they may inevitably dissociate, yielding patellar fractures and component loosening or breakage. Maltracking invariably requires complete revision—the only way to correct rotational positioning of the components.

The tibial component will be correctly rotated when positioned directly behind the tibial tubercle. As the conforming articular polyethylene necessarily aligns itself under the femoral condyles, the only way that the tubercle and in turn the patellar tendon and patella itself can be positioned in the femoral trochlear groove, is to line the tibial component with the tubercle. Femoral components are most reliably aligned parallel to the transepicondylar axis—this being the only landmark available at revision surgery. A posterolateral augment will generally ensure more external rotation. These conclusions are not only logical but confirmed by the published data on CT scanning of knees with maltracking.

Component breakage is far less common than previously. Some cases of tibial baseplate fracture were in fact cases of wear, osteolysis, and bone loss such that the component was unsupported and fractured. Posterior femoral condyles that have broken probably did so because of a gap between the posterior bone and an unsupported posterior condyle. Catastrophic polyethylene breakage is seen far less commonly as a result of better production methods, conforming articulations, and avoidance of thin modular inserts. Complete revision to improve both is mandatory, except perhaps in one of the few indications for partial revision—late wear of a tibial polyethylene insert in an elderly, low-demand patient.

Fractures have been reported of posterior stabilized tibial spines. Closer evaluation confirms that the mechanism is generally hyperextension of the knee so that the anterior femoral flange strikes sharply against the base of the spine. In these cases the cause of recurvatum is usually (femoral) component subsidence. These cases require a complete revision with considerations appropriate to ‘loosening’ above.

Breakage of the more prominent (and vulnerable) tibial spine on non-linked constrained implants results either from alignment leading to potent varus or valgus forces, or malrotation that can ‘twist’ the spine off the baseplate. In each case complete, corrective revision is required.

The problem arthroplasty without a coherent diagnosis should not undergo revision surgery as the results are likely to be poor. This situation is a strong indication to evaluate the patient more thoroughly. Failure to identify a problem in the knee should be a strong reminder to revisit the hip and spine, perhaps even with a bone scan. Recent ESR and CRP, as well as an aspiration for cell count and differential are appropriate. A CT scan should be obtained to quantify tibial and femoral component position.

Radiographs that predate the primary, to answer whether the patient actually had sufficiently severe cartilage loss to expect that arthroplasty would be useful, are in order. Methodical palpation of anatomical landmarks on the knee may identify neuromas or tendinitis that will more likely respond to cryoablation or physical therapy. While always dangerous to attribute physical suffering to mental anguish, there is nonetheless a literature that correlates lower satisfaction with depression.

The conceptual appreciation of how knee arthroplasties fail, the surgical technique, and the implants themselves have improved dramatically over two decades. Any one surgeon performs relatively few revisions and so many of these advances have not been incorporated into their practice. The understandable, though at times misguided, desire to perform less surgery at revision has led to worse rather than better results. Diagnosis, and the coherent surgical plan that flows from it, is absolutely essential. Revision knee arthroplasty is not one procedure but rather about eight different operations depending on the cause of failure. The ideas, technique, and implants that work well for primary arthroplasty generally do not apply to revisions. New ideas and a methodical approach to surgical exposure, even for stiff arthroplasties has made the need for quadriceps snips infrequent and both patellar turn-downs and tibial tubercle osteotomies almost obsolete. While attention to the level of the joint line is appropriate, using it as the guiding principle in revision surgery complicates the technique unnecessarily and should be abandoned, remembering that this ‘line’ is actually a three-dimensional concept corresponding to the shape of the femoral articular surface.

The unsolved problems in revision knee arthroplasty are still infection and extensor mechanism ruptures. While strategies exist for these problems, the long-term results are less good.

By establishing a tibial platform, stabilizing the knee in 90 degrees of flexion and then selecting the proximal–distal position of the femoral component to balance the knee in extension, with a judicious use of constrained implants, the results of first-time revision knee arthroplasties may exceed those of primary knee replacement.