Arthrodesis of the knee

The Journal of Arthroplasty Vol. 15 No. 7 2000

Arthrodesis of the Knee Experience With Intramedullary Nailing Stephen J. Incavo, MD, Jacob W. Lilly, BS, Craig S. Bartlett, MD, and David L. Churchill, PhD

Abstract: Knee arthrodesis using an intramedullary nail has gained acceptance as treatment in difficult cases such as infection after total knee arthroplasty (TKA), neuropathic joint, and obesity. A retrospective review of 22 cases treated at our institution using an intramedullary nail for knee arthrodesis was performed. Deep infection after primary (11) or revision (6) TKA was the most common indication for this procedure. A long intramedullary nail was used in 3 cases, a long nail with a proximal interlocking screw was used in 6 cases, and a customized nail with a valgus bend and a proximal interlocking screw was used in 11 cases. A modular knee fusion nail was used in 1 case. Successful fusion occurred in all cases, although 4 patients required additional surgery. Average operative blood loss was 748 mL, and average time to union was 7 months. Shortening of the extremity averaged 3.2 cm. Tibiofemoral alignment was improved by using a customized valgus nail (average, 3.1 valgus; range, 1–5) when compared with a straight nail (average, 0.2 valgus; range, 3 varus to 3 valgus). No patient developed infection in the hip or ankle region as a result of the long intramedullary nail. Intramedullary nailing is an excellent technique for knee arthrodesis in difficult cases. A customized proximal interlocking nail with 5° to 7° of valgus and 5° of anterior angulation improves tibiofemoral alignment and is straightforward to insert or extract should it be necessary. Stability and pain relief are rapid, and the fusion rate is maximized. Key words: knee, fusion, arthrodesis, intramedullary railing.

Successful knee arthrodesis is difficult to achieve, especially after failed arthroplasty. Intramedullary nailing has emerged as the technique of choice in this difficult situation [1]. Compared with external fixation or plating, intramedullary nailing has higher union rates [2–5]. The use of a long intramedullary nail has been criticized as technically demanding, as

evidenced by long operative times, large blood loss, and nail migration [6,7]. Other potential drawbacks of the long nail technique include loss of anatomic valgus knee alignment and lack of tibial rotational control. The long nail technique is not applicable when an ipsilateral hip arthroplasty or bone deformity is present. For these reasons, the use of a short nail with multiple interlocking screws has been advocated for knee arthrodesis in these situations. We have used intramedullary nail fixation for knee arthrodesis since 1990 at our institution. More recently, we have used a proximal interlocking screw and a custom-made nail incorporating flexion and valgus angulation. This article describes our experience, outcomes, and complications using intramedullary nail fixation for knee arthrodesis.

From the Department of Orthopaedics and Rehabilitation, The University of Vermont, College of Medicine, Burlington, Vermont. Submitted December 30, 1999; accepted May 10, 2000. No benefits or funds were received in support of this study. Reprint requests: Stephen J. Incavo, MD, Department of Orthopaedics and Rehabilitation, McClure Musculoskeletal Research Center, Stafford Hall 428 A, Burlington, VT 05405-0084. Copyright r 2000 by Churchill Livingstonet 0883-5403/00/1507-0008$10.00/0 doi:10.1054/arth.2000.9060

871

872 The Journal of Arthroplasty Vol. 15 No. 7 October 2000

Materials and Methods A retrospective review of 22 patients that were managed at our institution between 1990 and 1998 was undertaken. There were 13 men and 9 women. The mean age of patients was 65.8 years, and the median age was 72.5 years (range, 24–91 years). No patients were lost to follow-up, the average being 30 months (range, 12–91 months). One patient died of pseudomembranocolitis 3 months after her operation and was not included in the final analysis, leaving 21 patients in this study. One patient was treated at our institution with a 2-stage reimplantation after deep infection of a total knee arthroplasty (TKA). The infection recurred, and the patient was treated with intramedullary fusion elsewhere. The patient returned for subsequent treatment because of persistent osteomyelitis. All other patients had knee fusion performed at our institution. The indications for arthrodesis were as follows: deep infection after primary TKA in 11 cases, deep infection after revision TKA in 6 cases, degenerative arthritis with obesity in 1 case, neuropathic arthropathy in 2 cases, septic arthritis in 1 case, and posttraumatic arthritis in 1 case. The number of previous surgeries were 0 in 1 case, 2 in 11 cases, and ⱖ3 in 10 cases. There were 18 cases associated with deep infection: 11, Staphylococcus aureus; 3, Staphylococcus epidermis; 1, Klebsiella pneumoniae; 2, gramnegative infections; and 1, mixed gram-positive and gram-negative infection. Of the 21 cases, 3 used a standard long Ku¨ntscher rod; 6 used a proximal interlocking Ku¨ntscher rod; and 11 used a custommade, proximal interlocking nail with a 5° to 7° valgus bend and a 5° to 10° flexion angle. The non–custom-made nails had a prepared flexion bend of approximately 5°. In 1 case, an intramedullary modular nail was used, which was placed retrograde into the femur and antegrade into the tibia per manufacturer recommendations. Union was defined radiographically. Bridging osseous trabecula across the fusion site as seen on anteroposterior and lateral radiographs was considered union, even though the fusion continued to remodel and mature after this point. Clinical evaluation was considered and noted to confirm radiographic evidence. Radiographs also were used to determine varus and valgus and flexion angles in the fused leg. Shortening of the entire leg was defined as the measured difference between the preoperative and postoperative radiographs. In cases in which sequential measurements could not be performed, bilateral radiographs were used to determine the loss of bone and shortening.

Preoperative Planning Proper preoperative planning is imperative when considering a custom fusion rod. Radiographs of the entire femur and tibia (ie, anteroposterior and lateral) are necessary. A long ruler is placed to provide accurate length and canal dimensions. In general, the rod should extend from the tip of the greater trochanter to the distal 8 cm of the tibia. If a single long-leg cassette is used for the anteroposterior view, any distraction between the distal femur and proximal tibia must be subtracted from the total length. The custom rods were designed by a corporate engineer and reviewed by the operating surgeon. Design templates were agreed on before rod manufacture. The goal in these cases was to have the valgus and anterior angulation occur gradually over the length of the rod with the apex being at the level of the tibiofemoral contact point or 2 to 3 cm proximal to this, depending on the anatomy present. This goal was achieved in most cases. Surgical Technique The procedure was carried out in the semisupine position with the patient, rolled approximately 45°, using a beanbag positioning device. A radiolucent operating room table was used with a single fluoroscopic unit, which was moved for visualization in multiple planes. Two incisions were made: a buttock incision for identification of the greater trochanter and piriformis fossa and an anterior knee incision. Through the proximal incision, a guide wire is introduced into the femoral canal and retrieved at the knee. The femur and tibia are reamed in an antegrade fashion. The proximal femur is overreamed by 1 mm for a straight nail or 2 mm if a valgus bend has been placed in the fusion rod. The tibia is reamed only to the diameter of the nail. Once reaming is performed, the nail is introduced over the guide wire. When the nail reaches the level of the knee joint, the tibia is positioned under the femur so that the rod enters the tibial canal for about 3 to 4 cm. Using fluoroscopy, the intramedullary nail is advanced distally into the tibial canal. Just before obtaining a tight medullary fit in the tibia, rotational position is selected. This position is held in place for final seating of the nail. The nail is positioned such that the proximal tip is flush with the greater trochanter for placement of the proximal interlocking screw. Because the guide wire was significantly shorter than the custom nail, it was removed through the knee once the nail advanced to the level of the distal femur. The single case of placement outside of the

Knee Arthrodesis With Intramedullary Nailing ●

tibia occurred late in this series. This problem would have been avoided if an extra-long guide wire were available for use. When this case was reoperated on, an extra-long guide wire was used and added to the precision of the surgical procedure. The guide wire should be sufficiently flexible to allow removal through the extra-long angulated rod, and we now prefer to use a guide wire of suitable length. No attempt to recut the bone of the distal femur or proximal tibia was made, and no vigorous impaction of the tibia and femur was performed. It was common that a relatively large gap between the tibia and femur was present at the completion of the procedure (Fig. 1). The patient was encouraged to bear weight as soon as possible postoperatively, and a cast or brace was not used. Bone grafting was not performed at the time of initial arthrodesis. For the infected cases, no intraoperative frozen section analysis was performed. No case of infected TKA was treated with arthrodesis performed as a singlestage procedure. One patient had attempted fusion 2 weeks after initial infected TKA removal (see Results section). All other patients with an infected TKA had at least 6 weeks of intravenous antibiotic treatment as well as antibiotic-impregnated cement spacer placement.

Incavo et al.

873

Results After intramedullary arthrodesis, 17 of 21 patients achieved union with no further surgery. The average time to fusion was 7.0 months (range, 3–14 months). Four cases required further surgery to achieve fusion as a result of various causes, although fusion ultimately was accomplished in all cases. The first case remained infected because of inadequate antibiotic coverage. Methicillin-resistant S. aureus was not recognized initially, and the patient received only cefazolin (Kefzol). Repeated irrigation and de´bridement was performed, in addition to rod removal and subsequent adequate antibiotic coverage. With rerodding, the patient was asymptomatic in 6 weeks, and fusion was confirmed at 6 months. Delayed healing occurred in 1 patient who developed a deep soft tissue calf abscess 3 months after rod placement. Treatment included irrigation, de´bridement, and a delayed gastrocnemius muscle flap. Nine months after the initial procedure, radiographs indicated fusion, and the rod was removed to eliminate a possible source of infection. Increased pain and flexion angulation were observed, despite bridging of medial and lateral bone. No change in

Fig. 1. (A) Immediate postoperative radiograph shows poor bone apposition. (B) One-year postoperative radiograph shows compression of the fusion site and union and valgus tibiofemoral alignment. (C) Two-year postoperative radiograph shows maturation of fusion.

874 The Journal of Arthroplasty Vol. 15 No. 7 October 2000 valgus alignment was observed. Rerodding was performed, and the patient was pain free in 4 weeks. Radiographic union was observed at 5 months. In 1 case (Fig. 2), impaction of the femur and tibia did not occur because of an excessively anteriorly bowed rod contour. In response to the persistent gap, bone grafting (autogenous and coralline hydroxyapatite) was performed 4 months after the initial operation. Radiographic union was evident 10 months later. The significant result of this procedure was the absence of any measurable shortening of the affected leg. The final delayed union occurred because of technical reasons. The patient suffered from chronic renal failure, requiring renal dialysis, in addition to a long-standing S. epidermis infection of a revised TKA. During the arthrodesis procedure, tibial reaming was uneventful. Final rod placement was outside the tibia because of an unrecognized tibial fracture, however. Twelve months postoperatively, the patient developed knee pain, and rod loosening was suspected. The patient was treated with rod removal and grafting (coralline hydroxyapatite) of the tibial cortical defect. Two months later, rerodding with a custom nail was performed, and radiographic union was observed at 6 months. Average operative blood loss was 748 mL (range, 200–1,600 mL). The average flexion angle was 5.8° (range, 3°–13°). The average tibiofemoral angle of

the 12 procedures using a custom rod was 3.1° valgus (range, 1°–5° valgus). In contrast, the average alignment of the 10 procedures not using a custom rod was 0.2° valgus (range, 3° varus to 3° valgus). Measurements indicated a mean shortening of 3.2 cm (range, 0.4–7.5 cm). Severe bone loss and shortening (⬎5.0 cm) was observed in 2 cases. Moderate shortening (1.5–4.9 cm) was evident in 15 cases, whereas mild shortening (⬍1.5 cm) was found in 4 cases. Average shortening was 2.4 cm for the custom-prepared rods and 3.4 cm for the remaining rods. The single case of no bone shortening represents the aforementioned case in which grafting was necessary to bridge distraction of the fusion site. No radiographic or clinical rotational change occurred during the postoperative period. Rod removal was performed in 4 cases. Three proximally interlocked nails were removed because of various complications, as described earlier. The fourth case of rod removal was performed for the single case of infection of the modular fusion nail. De´bridement of a drainage sinus, performed at another institution, revealed communication through the anterior femoral cortex into the intramedullary region. After referral to our institution, rod removal with local irrigation and de´bridement was performed.

Fig. 2. (A) Immediate postoperative radiograph shows excessive flexion angulation of the nail, which prevented collapse of the fusion site. (B) Bone grafting was performed 4 months postoperatively resulting in fusion, best seen at the posterior tibiofemoral region on this lateral radiograph.

Knee Arthrodesis With Intramedullary Nailing ●

To remove the rod, the anterior cortex of the femur and proximal tibia was elevated in a single piece. The tibia had been covered previously with a medial gastrocnemius flap so that surgical exposure of the femur was selected preferentially over the tibia. The rod was transected below the modular connection, which allowed the proximal piece to be extracted through the cortical opening. The distal piece was then proximally and removed. The anterior femoral cortex was cerclaged back into position, using multiple wraps of absorbable suture. Healing was uneventful, and no recurrence of infection occurred 18 months after this procedure.

Discussion Intramedullary fixation is the technique of choice for knee arthrodesis in difficult cases. The main indication in this series was infected primary or revised TKA. It is well accepted that infections tend to complicate or decrease fusion rates. Patient death may occur in severe or repeat infections. An analysis of reinfection treatments after total hip arthroplasty includes 1 death in 31 cases [8]. In the present study, 3 patients suffered from rheumatoid arthritis: one patient developed pseudomembranocolitis, leading to her death 15 weeks later, and the other 2 patients experienced prolonged morbidity. We believe the technique outlined in this article is the optimal approach for difficult knee arthrodesis. Blood loss averaged 748 mL, and operative time was generally ⬍90 minutes. These values compare favorably with other studies [2]. The use of a proximally interlocking nail has several advantages. Proximal nail migration is prevented because the tibia is allowed to slide on the nail to promote compression of the fusion site. Custom preparation of an intramedullary fusion nail allows proper diameter selection. Different diameters are used to accommodate the varying dimensions of the femur and tibia. Resultant anatomic alignment of the knee is an observed advantage. The implementation of a custom valgus-bent rod promotes improvement of postoperative coronal plane alignment. The lack of any cases with postoperative varus alignment when the rod was prepared with a valgus bend best illustrates the effectiveness of a custom rod; this was true even in cases of severe bone loss after infected revisions. Rod removal is an important consideration in arthrodesis, particularly when dealing with persistent and recurrent infections. Complete eradication is not always achieved with successful fusion, and rod removal may be required. Lai et al [9] have

Incavo et al.

875

reported this disadvantage for the Huckstep nail, stating that retrieval was the major drawback of their technique. In the present study, 4 cases required rod removal. This removal was accomplished easily in the 3 cases in which a proximal interlocking nail was used, whereas technical difficulty was encountered only in the modular nail removal. The main disadvantage of the intramedullary technique is the shortening of the leg. Despite this shortening, functional outcome analyses have not cited this concern when compared with other methods [10]. Because successful fusion depends on the impaction of the tibia and femur, the shortening obtained here is not uncharacteristic or unexpected. Structural allograft may offset this bone shortening; however, the infection and circumstances more generally should be considered. We prefer not to use distally interlocking screws, primarily because of the gradual impaction of the fusion site that occurs with weight bearing. Rotational control is not a significant concern with nails having a proper fit; each nail was designed with a 5° to 10° flexion (including the noncustom nails). An overall average flexion angle of 5.8° indicates that a prepared nail flexion facilitates suitable postoperative flexion. In this technique, no attempt was made to recut the ends of the bones after TKA removal. Because leg shortening is a significant problem after this technique (because of bone loss after failed TKA), any recutting of the femur and tibia leads to further shortening. It is our experience that gradual compression occurs until enough bone contact is present to produce union. This compression is a function of tibiofemoral angulation as well as any lateral subluxation caused by already present bone loss. The principal reason why fusion occurs is the constant, gradual compression of the fusion site because the tibia is not rigidly fixed. Structural allografts, autogenous grafting, or bone–graft substitutes, in addition to distally interlocking screws, are possible alternatives. Potential and persistent local infections advise against these options. Alternatively, with an intramedullary nail in place, limb lengthening over the existing nail is a consideration once arthodesis has been achieved. Intramedullary rodding, using a proximally interlocking nail, is preferred because of its relative surgical ease and high fusion rate. Distal or proximal infection spread was not observed in any of the aforementioned cases, and proximal migration of the nail did not occur. The procedure is straightforward, and significant pain relief generally occurs within 4 weeks. The observed influence on postop-

876 The Journal of Arthroplasty Vol. 15 No. 7 October 2000 erative alignment and the facilitation of rod removal are additional factors that recommend this procedure.

References 1. Damron TA, McBeath AA: Arthrodesis following failed total knee arthroplasty: comprehensive review and meta-analysis of recent literature. Orthopedics 18:361, 1995 2. Ellingsen DE, Rand JA: Intramedullary arthrodesis of the knee after failed total knee arthroplasty. J Bone Joint Surg Am 76:870, 1994 3. Nichols SJ, Landon GC, Tullos HS: Arthrodesis with dual plates after failed total knee arthroplasty. J Bone Joint Surg Am 73:1020, 1991 4. Rand JA: Instructional Course Lecture, The American Academy of Orthopaedic Surgeons. Alternatives to reimplantation for salvage of the total knee arthro-

5.

6.

7.

8.

9.

10.

plasty complicated by infection. J Bone Joint Surg Am 75:282, 1993 Wilde AH, Stearns KL: Intramedullary fixation for arthrodesis of the knee after infected total knee arthroplasty. Clin Orthop 248:87, 1989 Donley BG, Matthews LS, Kaufer H: Arthrodesis of the knee with an I intramedullary nail. J Bone Joint Surg Am 73:907, 1991 Puranen J, Kortelainen P, Jalovaara P: Arthrodesis of the knee with intramedullary nail fixation. J Bone Joint Surg Am 72:433, 1990 Pagnano MW, Trousdale RT, Hanssen AD: Outcome after reinfection following reimplantation hip arthroplasty. Clin Orthop 338:192, 1997 Lai K, Shen W, Yang C: Arthrodesis with a short Huckstep nail as a salvage procedure for failed total knee arthroplasty. J Bone Joint Surg Am 80:380, 1998 Benson ER, Resine ST, Lewis CG: Functional outcome of arthrodesis for failed total knee arthroplasty. Orthopedics 21:875, 1998