The techniques of total joint arthroplasty have often been advanced by the study of failed procedures. The rate and results of reoperation after total knee arthroplasty are frequently mentioned in scientific reports, but there has been little discussion of the specific problems encountered, their origin, and how best to address them. I performed 1000 posterior cruciate retaining primary total knee arthroplasties using the Press-fit Condylar prosthesis (Johnson and Johnson, Raynham, Massachusetts, USA) during a nine year period beginning in 1984. At an average follow-up of six years, I reviewed the complications encountered, the details of specific cases requiring reoperation, and the techniques used in order to help identify and understand the most common reasons for reoperation after total knee arthroplasty.

Failure of the Femoral Component

          Among the 1000 total knee arthroplasties reviewed, 309 (31%) of the femoral components were inserted with cement, and 691 (69%) were inserted without cement. Only four of the 1000 femoral components (0.4%) failed within the study period: one by loosening and three by fracture of the medial femoral condyle of the prosthesis.

Loosening of the Femoral Component
          Only one of the 1000 femoral components has loosened. This occurred in a patient with a femoral component inserted without cement and was associated with a 15¼ valgus deformity of the mechanical axis of the lower extremity. The preoperative deformity was corrected at the knee joint but unrecognized diaphyseal bowing added eight degrees of valgus to the mechanical axis. During the four-year period immediately after the arthroplasty, the femoral component gradually subsided into a position of increased valgus. The component was revised with a cemented component to treat the patientÕs deformity and pain.

Breakage of the Femoral Component
           Stress fracture of the medial femoral condyle occurred at the angled junction between the distal condyle and the posterior chamfered surface in three patients. (Figure 1) All three were large, male patients, each weighing greater than 100 kg. Relatively large porous-coated prostheses were used (one size 5 and two size 6 femoral components), all of them inserted without cement. Apart from the fractured condyle, each component was well fixed with bony ingrowth. The mechanism of fracture appeared to be a cyclic tension applied to the cintered surface during weight bearing due to the expansion force transmitted to the trochlea and posterior condyles. This failure mechanism is similar to that described by Whiteside and colleagues for the Ortholoc prosthesis.1

Failure of the Tibial Component

          Tibial components were inserted in one of three ways: the plateau and keel were fully cemented in 875 knees (87%); the plateau was cemented, but the keel was not in 87 knees (9%); and no cement was used in 38 knees (4%). Only four of the 1000 tibial components (0.4%) failed within the study period: one of the partially cemented (1%) and three of the uncemented prostheses (8%). The single partially cemented failure was a patient with systemic lupus erythematosus who had recurrent aseptic knee synovitis and increase activity around the tibial component on bone scan. She underwent a fully cemented revision two years after the initial procedure. The three cementless tibias were revised for pain. Two of these knees were painful from the time of the surgery - one of these was revised at one year and the other at eight years from the time of surgery. The remaining knee was initially painless, but gradually subsided, became painful and was revised four years after the initial operation.

Failure of the Tibial Insert

          Before 1987 a flat tibial insert was used. Between 1987 and 1992, an insert with an anterior and posterior lip was implanted routinely. A so-called curved insert with greater conformity to the shape of the femoral condyles was used only in selected knees with severe deformity and instability between 1987 and 1992, but was used exclusively thereafter. (Figure 2)

           Only seven inserts have been exchanged. Of the 162 flat inserts implanted with minimum nine-year follow-up, six have been exchanged due to radiographic evidence of polyethylene wear. Two of these patients had symptomatic effusions and four were asymptomatic. In two patients with varus malalignment, the worn tibial insert was exchanged for a custom insert that added either three or five degrees of valgus. The only patient with a lipped insert requiring exchange was an active asymptomatic 62-year-old with asymmetric wear of a six-millimeter insert noted on radiographs. No curved inserts have been exchanged.

Failure of the Patellar Component

          One hundred and seventy-four patellae were not resurfaced and 11 had prior patellectomy. A metal-backed patella was used for the first 87 of the 815 resurfaced patellae. The metal backed patella was discontinued in 1986 and a three-pegged all polyethylene component was used thereafter.

Wear-Through of the Polyethylene
          Five of the 87 metal-backed patellas (6%) were revised for wear through of the polyethylene leading to metal debris and synovitis.

Patella Fracture
          Traumatic patellar fracture occurred in three knees with all polyethylene components. In each case, the fracture was minimally displaced and the polyethylene component was not loose. Treatment consisted of splint immobilization for one month.

Painful Unresurfaced Patellae
          Nine percent of patients in whom the patella is not resurfaced at the time of total knee arthroplasty will develop mild anterior discomfort that is tolerable and not disabling. Only two of the 174 unresurfaced patellas (1%) required reoperation for secondary resurfacing.

Miscellaneous Causes of Re-operation

Recurrent Hemarthrosis
          Three knees required open synovectomy to treat recurrent hemarthrosis. The cause of such recurrent hemarthroses may be synovial entrapment.2

Recurrent Rheumatoid Synovitis
          Three knees in rheumatoid patients required synovectomy for recurrent rheumatoid synovitis unresponsive to medical therapy or intra-articular corticosteroid injection. In each case the synovitis was associated with a generalized flare of the disease and there was no evidence of infection.

Infection
          Five knees in four patients became infected, all of them at a time remote from the surgery. The source of the infection was related to an intestinal condition in one patient, cellulitis about the ankle in another, and a lung infection in two knees in a single immunocompromised patient. The source was uncertain in one patient. All patients were treated initially with debridement, exchange of the tibial insert, and intravenous antibiotics, but two eventually required a delayed exchange of all components.

Traumatic Instability
          Two patients suffered a fall that led to insufficiency of the medial collateral ligament and valgus instability. Both were treated with liner exchange to restore tension in the medial collateral ligament. One patient had a syringomyelia and has become non-ambulatory and the other has a quadriceps deficient knee after patelletomy. The later patient fell again, injured her posterior cruciate ligament, and was revised to a posterior stabilized prosthesis.

 

Discussion

          My current indications for implantation of a femoral component without cement are far narrower than in the past. Cemented components are favored in most patients as they are less expensive and they have been successful well into the second decade in most long-term reports.3-5 I will consider an uncemented prosthesis in younger patients who have a life expectancy greater than 15 to 20 years.

           The 8% failure rate of uncemented tibial components is unacceptable in light of the outstanding results of fully cemented components. While the addition of screw fixation to the tibial tray might improve the initial results beyond what I observed, the potential for screw migration into the tibial insert leading to polyethylene wear, fretting of the screw on the tray, and subsequent osteolysis make screw fixation unappealing in the long-term.

          Polyethylene wear can be the result of many factors including the manufacturing and sterilization process, the thickness of the polyethylene, the contact area of the articulation, and dynamic forces encountered during gait related to limb malalignment and instability. Over time is has become apparent that a more conforming tibial topography leads to increased contact area and lower polyethylene stresses.6 In my experience, the flat tibial insert intended to allow the retained posterior cruciate ligament to control the kinematics of the articulation has been more likely to produce abnormal wear and osteolysis. I now use the more conforming curved inserts exclusively.

          There is no apparent advantage of metal backed patellae over the all polyethylene design.7 Considering my unacceptably high 6% failure rate, I see no role for metal backed patellae.
I forgo patellar resurfacing in a select group of younger patients with noninflammatory osteoarthritis, congruent tracking of the patella on the pre-operative skyline radiograph, congruent tracking with the prosthetic trochlear flange noted intra-operatively, and the absence of eburnated patellar bone.8 Reoperation for a painful unresurfaced patellae is straightforward, but rarely necessary. As a result of the success of the three-pegged all polyethylene patellar component, I routinely resurface the patella in older patients.

          One important element of successful posterior cruciate retaining total knee arthroplasty is careful cruciate balancing. We use the so-called POLO test to assess the balance of the posterior cruciate.9 (Figure 3) PO stands for pull-out and LO stands for lift off. The posterior cruciate may be too loose in flexion if the trial tibial tray can be pulled out while the knee is flexed 90¼. A thicker insert should be inserted until the tray cannot be pulled out. The posterior cruciate ligament may be too tight in flexion if the anterior part of the tibial tray lifts off the anterior tibial cortex as the knee is flexed to between 80¼ and 100¼. To relieve this, the cruciate is initially recessed to the level of the bone resection on the tibial side. If this is insufficient, it is released from its femoral attachment until lift off disappears. Care should be taken to eliminate posterior femoral osteophytes or uncapped posterior condylar bone, both of which can cause abnormal lift off.

          In summary, I have experienced a 3.5% reoperation rate at six years in the treatment of 1000 knees with total knee arthroplasty. The most common reasons for reoperation were wear of a flat polyethylene tibial insert, wear of a metal backed patellar component, and late metastatic infection. Lessons learned from this experience have reinforced changes in prosthetic design, surgical technique, and patient selection which should contribute to a decreased re-operation rate in the future.

Richard D. Scott, MD is an Attending Surgeon at Brigham and Women's Hospital, and Associate Clinical Professor of Orthopaedic Surgery at Harvard Medical School

Address correspondence to:
Dr. Richard D. Scott; Department of Orthopaedics; Brigham and Women's Hospital; 75 Francis St.; Boston, MA 02115