Patients were placed in a supine position as for standard TKA, and an anterior midline skin incision was made through the old scar. All previous prosthetic components were removed and cement was thoroughly excised using a small osteotome. Samples were taken for microbiological culture and pathological analysis, including intraoperative frozen sections to establish the extent of infection. Meticulous excision was performed to remove all remaining cement, granulation and necrotic devitalized tissue in a stepwise manner, including in the suprapatellar pouch, medial and lateral gutter, and posterior compartment of the knee. The sclerotic bony surfaces positioned by femoral and tibial components were cut back to the normal bone using a high-speed tip burr. Through the previous insertion site of the femur and tibial stem components, the medullary canal was debrided by retrograde (femur) and antegrade (tibia) intramedullary reaming in 0.5- or 1.0-mm increments to avoid possible infection. After thorough debridement, we determined tissue viability by observing blood oozing from the soft tissue and remaining bone stock using a 1-minute deflation of the tourniquet. Under fluoroscopy guidance, flexible intramedullary rods (Synthes Inc., West Chester, PA, USA) were inserted retrogradely in the femoral side with the knee in flexion. After the first rod was driven across the narrowest portion of the femoral canal (approximately 5 cm), second, third, fourth, fifth or more rods were inserted one by one as permitted by the intramedullary canal size. After filling the intramedullary canal, the inserted rods were tapped into the medullary canal for more secure positioning, and the end of each rod was cut at the level of the articular bony defect of the proximal tibia (Fig. 1A). The rods were then driven back securely into the proximal femur. We aimed for as much rod length as possible to cross the isthmus of the femoral canal to maximize stability. The rods on the tibial side were then antegradely inserted into the tibia using the same method as used on the femoral side. The tibial rods were trimmed to the level of the area of the distal anterior femoral cortex bone loss to maximize overlap with femoral-side rods (Fig. 1B). The rod ends from the femoral and tibial sides were arranged such that they overlapped and interdigitated (Fig. 1C) to maximize mechanical strength. The interdigitating rod ends were tightly fixed using two (or three) cerclage wires (Zimmer, Warsaw, IN, USA). Antibiotic-loaded cement was made by mixing two 40-g packs of Palacos bone cement (Biomet Merk, Warsaw, IN, USA) with 4 g vancomycin and 4.8 g tobramycin per cement pack. At the dough time period, usually 2 to 3 minutes from the beginning of the cement mixing, the cement was placed within the knee and polymerized so that the cement conformed to the irregular contour of the femur and tibia that resulted from the severe bone loss. The cement macrointerdigitated into bone defects and the intercondylar notch, extended into the medullary canal and suprapatellar pouch, and surrounded the rod bundle to create a cement arthrodesis of the knee joint (Fig. 1D). Postoperatively, patients were allowed to weight bear as tolerated. Patients were administered intravenous antibiotics until the wound dried, and were then administered oral antibiotics for 6 weeks.

The first patient was an 80-year-old woman who underwent a staged bilateral TKA in 2004 for degenerative osteoarthritis. Because of periprosthetic infection in the left TKA at postoperative 2 months, the patient then underwent a two-stage revision TKA comprising removal of the prosthesis and insertion of antibiotic cement followed by a 6-week antibiotic infusion and insertion of a stemmed-revision TKA component (Nexgen® LCCK, Zimmer; Legacy® knee constrained condylar knee, Zimmer). However, clinical signs of infection including pain, swelling, and drainage at the anterior aspect of proximal tibia continued for 9 months after this secondary surgery. The infecting organism was identified as methicillin-resistant Staphylococcus aureus (MRSA). Hematology findings included erythrocyte sedimentation rate (ESR: 76 mm/h) and C-reactive protein (CRP: 44 mg/dl) concentration. Radiology indicated stem loosening in the femur with a severe bone defect especially in the anterior tibia and around the anterior femur flange. The skin defect (3×2 cm) due to persistent drainage overlying the tibial tuberosity was prominent and expanding. We diagnosed a persistent prosthetic infection. The patient had infection recurrence, poor bone stock and was physico-psychologically compromised following multiple failed surgical interventions and old age. We decided to perform a one-stage artificial fusion with a bundle of flexible rods and an antibiotic-loaded cement spacer. The procedure was uneventful. At 8 weeks post-operatively, the joint showed no sign of infection, the patient could walk with a cane without pain. The most recent radiographs, at 44 months after surgery, showed no sign of rod or cement failure (Fig. 2).

A 75-year-old man who had renal insufficiency underwent a right-side TKA in 2003 due to posttraumatic arthritis resulting from a tibial plateau fracture 10 years ago. Postoperative infection developed 6 months after TKA. A two-stage revision arthroplasty was performed but failed due to a recurrent MRSA infection. The patient refused a further revision TKA and was very concerned about reinfection. Therefore, we performed a knee arthrodesis. It was not possible to use a long intramedullary nail because the patient had a dynamic hip screw on the ipsilateral (right) hip for a previous intertrochanteric fracture. Therefore, knee arthrodesis using dual plate fixation was performed. However, pain, swelling and elevated ESR and CRP levels persisted for 6 months after arthrodesis. Therefore, we performed artificial fusion using a flexible rod bundle and antibiotic-loaded bone cement. The previous plates were removed and a thorough debridement performed, after which the rods and cement were applied (Fig. 3). Partial weight-bearing was permitted 1 week after surgery, and full weight-bearing was allowed at 3 weeks postoperatively. At a 56-month follow-up, the patient could walk independently without any support and without pain. There was no evidence of infection according to physical, laboratory or radiograph examinations. Similarly, there was no evidence of rod or cement failure.