All 179 patients (204 ankles) who had had a primary Mayo total ankle replacement at our institution from 1974 through 1988 were initially entered into the study. The patients had been requested to return for a physical and radiographic examination at one and two years postoperatively and every five years thereafter. Patients who had not had a routine follow-up examination within one year before the time of the study were requested to return for a physical and radiographic examination. Those who were unable to do so were contacted by letter or telephone and were asked to evaluate the clinical result with regard to the level of pain, the maximum distance (in blocks) that they were able to walk, whether they used walking aids, and whether they considered the ankle to be much improved, improved, not improved, or worse when compared with the preoperative status. Radiographs of the ankle were made locally and then sent to us for analysis. Preoperative clinical information was obtained from the total-ankle-replacement data forms that had been completed by each patient before the arthroplasty. Thirty-six patients (forty-four ankles) could not be evaluated: eleven patients (fifteen ankles) had died and twenty-five patients (twenty-nine ankles) had inadequate clinical data recorded or could not be located. The remaining 143 patients (160 ankles) formed the basis of this study.
Ninety-one (57 per cent) of the arthroplasties were done on the right ankle and sixty-nine (43 per cent) were done on the left. Seventeen patients had an arthroplasty on both ankles. There were forty-six men and ninety-seven women. The mean duration of follow-up was nine years (range, two to seventeen years). The mean age of the patients at the time of the arthroplasty was fifty-four years (range, twenty to eighty-four years), the mean height was 1.66 ± 0.92 meters (range, 1.45 to 1.95 meters), and the mean weight was sixty-eight kilograms (range, thirty-four to 114 kilograms).
The diagnosis was rheumatoid arthritis for ninety-six ankles (60 per cent), post-traumatic osteoarthrosis for fifty-six (35 per cent), and osteoarthrosis for eight (5 per cent).
A previous operation had been done on the ipsilateral ankle or foot in thirty-nine limbs (24 per cent). An open reduction and internal fixation of a fracture of the ankle had been done in twenty-two limbs; an arthrodesis of the ankle, in three; a triple arthrodesis, in six; a talonavicular arthrodesis, in one; a synovectomy of the ankle, in two; and a subtalar arthrodesis, lengthening of the Achilles tendon, débridement of the ankle, resection of an osteophyte in the ankle, and an osteotomy through the midfoot, in one each.
Pain was the chief symptom for all of the patients; it was severe (excruciating) and almost always present in 102 ankles and moderate and present daily in twenty-eight. Pain was present in the remaining thirty ankles, but its magnitude was not recorded specifically. The mean duration of the pain preoperatively was eight years (range, one to fifty years). The range of motion of the foot relative to the leg in the sagittal plane was measured with a goniometer, as has been described14. Before the arthroplasty, the mean (and standard deviation) dorsiflexion was 4 ± 9 degrees (range, 20 degrees of plantar flexion to 25 degrees of dorsiflexion) and the mean plantar flexion was 22 ± 11 degrees (range, 0 to 50 degrees).
All of the arthroplasties were performed by or under the supervision of an orthopaedic consultant in our department. The design of the implant (Fig. 1) and the operative technique have been described previously22. All of the procedures were performed through an anterior incision, and both the stainless-steel talar component and the polyethylene tibial component were fixed with methylmethacrylate.
Forty patients (forty-six ankles) returned for evaluation at the Mayo Clinic, twenty-seven patients (thirty ankles) sent back the answers to a questionnaire, and twenty-two patients (twenty-seven ankles) were evaluated on the basis of a telephone interview and radiographs made locally and sent to us. Follow-up information on fifty-four patients (fifty-seven ankles) from whom the implant had been removed was obtained by a review of the medical records; the arthroplasty was considered a failure in these patients. One patient who had had a bilateral arthroplasty had a different type of follow-up for each ankle: one ankle was evaluated by examination and the other, by review of the medical records because an arthrodesis had been done (and thus the result was considered a failure).
Adequate radiographs that had been made before the arthroplasty and at the time of the follow-up were available for 101 (63 per cent) of the 160 ankles. We excluded radiographs that had not been properly exposed or made with proper alignment.
Standard anteroposterior (weight-bearing), lateral (weight-bearing), and mortise radiographs of the ankles were analyzed by one of us (H. B. K.), and the preoperative and follow-up radiographs were compared to identify radiolucent lines around the implant and to grade loosening or migration of the components (Figs. 2-A, 2-B and 2-C). Two types of radiolucent lines were noted: a continuous line extending along at least one-half but less than the entire periphery of the component, and a radiolucent line along the entire surface of the component. Loosening was defined with use of categories similar to those employed by Harris et al. for loosening of the femoral component in the evaluation of total hip replacements inserted with cement. Definite loosening was characterized by obvious migration of one prosthetic component or both; probable loosening, by a continuous radiolucent line surrounding the entire cement-bone interface but with no evidence of migration; and possible loosening, by a radiolucent line extending along at least one-half but less than the entire cement-bone interface. Migration was defined as any discernible change in the position of the component.
The over-all result was rated as good if the patient considered the ankle to be much improved, had no pain, was able to walk more than six blocks, did not need a walking aid, and had no radiographic evidence of loosening. The result was rated as fair if the patient considered the ankle to be improved, had only mild pain occasionally, was able to walk two to six blocks, needed a cane for walking, and had no radiographic evidence of loosening. The result was rated as poor when the patient considered the ankle to be worse or not improved, had moderate pain daily or severe pain that was nearly constant, was able to walk less than two blocks, and needed crutches, a walker, or a brace. The result was rated as a failure if the implant had been removed.
According to the previously defined criteria, the result was good for thirty-one ankles (19 per cent), fair for fifty-five (35 per cent), and poor for seventeen (11 per cent); fifty-seven arthroplasties (35 per cent) were considered to be a failure.
The mean (and standard deviation) dorsiflexion (extension) of the forty-six ankles that were examined clinically was 5 ± 7 degrees (range, 15 degrees of plantar flexion to 20 degrees of dorsiflexion). The mean plantar flexion was 19 ± 11 degrees (range, 0 to 50 degrees).
Complications
There was one complication after eighteen arthroplasties and there were two complications after one. The complications included a superficial infection alone (four ankles), a deep infection (six ankles), delayed wound-healing (two ankles), an acute malleolar fracture alone (three ankles), a superficial infection and a malleolar fracture (one ankle), migration of the wire from the tibial component (one ankle), dislocation of the talar component (one ankle), and ulceration of the medial aspect of the ankle (one ankle).
Superficial infection: In two of the five ankles, the infection was treated successfully with antibiotics but both ankles ultimately had an arthrodesis because of pain and loosening of the prosthetic components. In the third ankle, the superficial infection was also treated successfully with antibiotics, but a decompression was later done because of malleolar impingement and eventually an arthrodesis was performed. The fourth ankle was treated successfully elsewhere with antibiotics and débridement, but eventually an arthrodesis was done because of pain and loosening. The fifth ankle, also treated successfully with antibiotics, had radiographic evidence of definite loosening, but the implant was retained in situ. The result was considered fair.
Deep infection: For one of the six ankles with a deep infection, débridement and eventually a below-the-knee amputation were performed because of soft-tissue necrosis. An arthrodesis was subsequently done in the remaining five ankles because of loosening and pain; two soft-tissue-coverage procedures had been done previously on one ankle in which the infection was associated with definite loosening of the components, débridement and a soft-tissue-coverage procedure had been performed on one ankle, and débridement only had been done on two ankles.
Delayed wound-healing: The two ankles with delayed wound-healing were treated successfully with local wound care, and both healed without the need for additional operative treatment. The result, which was fair for both ankles, was not affected by the delay in healing.
Acute malleolar fracture: Four acute malleolar fractures occurred intraoperatively during placement of a component. Each fracture was fixed with screws or Kirschner wires, and all united. We were unable to determine whether the fracture had an effect on the result of the arthroplasty: the result was good for one ankle, poor for one, and a failure for two.
In six other ankles, deformation of the malleoli with associated new-bone formation developed during the follow-up period. We were unable to determine the time at which this occurred or any relationship between the deformation and either the clinical result or the diagnosis. There was no evidence of distinct fracture lines, and the radiographic appearance suggested that the deformation had resulted from stress fractures or stress remodeling. The importance of these findings in terms of function and durability of the replacement was not apparent in this small group: the result was fair for three ankles, poor for one, and a failure for two.
Heterotopic Bone Formation
Heterotopic bone (abundant ectopic new bone) developed adjacent to the ankle replacement in two patients, one who had rheumatoid arthritis and one who had post-traumatic osteoarthrosis (Fig. 3). We were unable to determine whether the ectopic bone had a definite effect on function or caused any symptoms: the result was fair for one ankle and poor for the other. Heterotopic bone formation did not appear to be related to subclinical infection or stress remodeling.
Reoperations
Ninety-four additional operations were performed on sixty-six (41 per cent) of the 160 ankles. Seven débridements were done because of infected or necrotic tissue, two below-the-knee amputations were done because of soft-tissue necrosis or infection and non-union at the site of the arthrodesis, ten revision total ankle arthroplasties were done because of pain, thirteen decompressions were done because of malleolar impingement, and four procedures (in three ankles) were done to provide soft-tissue coverage. In addition, fifty arthrodeses were done on forty-nine ankles because of pain. A subset of the patients who had an arthrodesis was reported on previously14. Eight other operations were performed on eight ankles. These consisted of removal of the marker wire from the tibial component, partial resection of the lateral malleolus, lengthening of the Achilles tendon, tendon transfer, resection arthroplasty of the ankle, tendon repair, removal of screws, and nerve decompression.
Failures
During the seventeen years of the study period, fifty-seven total ankle arthroplasties (36 per cent) failed in fifty-four patients, at an average of 4.4 years (range, 0.1 to 13.1 years) postoperatively. Failure was defined as the need to remove the implant. The implant was removed and fifty arthrodeses were done in forty-six patients (forty-nine ankles), revision total ankle arthroplasty was done in ten patients (ten ankles), and below-the-knee amputation was necessary in two patients (two ankles). All of these reoperations were performed because of persistent pain. Three patients had revision total ankle arthroplasty followed by arthrodesis, and one patient had two attempts at arthrodesis and then amputation. The rate of failure was influenced by the age of the patient: the rate was 43 per cent (thirty-nine of ninety-one ankles) for the seventy-nine patients who were fifty-seven years old or less and 26 per cent (eighteen of sixty-nine ankles) for the sixty-four patients who were more than fifty-seven years old.
The rate of failure did not vary according to the diagnosis. The rate was 32 per cent (thirty-one of ninety-six ankles) for the eighty patients who had rheumatoid arthritis and 41 per cent (twenty-six of sixty-four ankles) for the sixty-three patients who had osteoarthrosis or post-traumatic osteoarthrosis.
Radiographic evidence of loosening was not invariably associated with a clinical failure. Of the eight ankles in which there was definite migration of the tibial component, one had a good result, one had a poor result, and six had a failure. Of the forty-nine ankles with a radiolucent line along the entire surface of the tibial component, two had a good result, fourteen had a fair result, eight had a poor result, and twenty-five had a failure. Seven of these forty-nine ankles also had a radiolucent line along the entire surface of the talar component: one had a fair result and six had a failure. Of the fifty-eight ankles in which there was definite migration of the talar component, eight had a good result, twenty had a fair result, nine had a poor result, and twenty-one had a failure.
Radiographic Analysis
There was no radiolucent line around the tibial component in twenty-two (22 per cent) of the 101 ankles for which follow-up radiographs were available. A radiolucent line extended along at least one-half but less than the entire surface of the tibial component in thirty ankles (30 per cent) and along the entire surface in forty-nine (49 per cent). There was no radiolucent line around the talar component in seventy-seven ankles (76 per cent). A radiolucent line extended along at least one-half but less than the entire surface in seventeen (17 per cent) and along the entire surface in seven (7 per cent).
No migration of the tibial component was seen in eighty-eight (87 per cent) of the 101 ankles. There was probable migration9 of this component in five ankles (5 per cent) and definite migration in eight (8 per cent). No migration of the talar component was seen in twenty-nine ankles (29 per cent); probable migration, in fourteen (14 per cent); and definite migration, in fifty-eight (57 per cent) (Figs. 4-A and 4-B).
There was no radiographic evidence of loosening of the tibial or talar component in twelve ankles (12 per cent), possible loosening in thirteen (13 per cent), probable loosening in fifteen (15 per cent), and definite loosening in sixty-one (60 per cent). Most of the ankles with radiographic evidence of definite loosening had an unsuccessful clinical result: the result was good for ten (16 per cent) of the sixty-one ankles, fair for twenty-one (34 per cent), poor for nine (15 per cent), and a failure for twenty-one (34 per cent).
Previous investigators have described malleolar impingement as a complication of total ankle arthroplasty1,24. On the radiographs that we had available, we were unable to assess quantitatively how much medial or lateral impingement (or both) was present, and we could not establish an association between either the malleolar fractures or the deformities of the malleoli and the presence of tibiotalar impingement.
The preoperative radiographs of the 101 ankles showed rheumatoid arthritis or osteoarthrosis of the subtalar joints to be absent or mild in forty-six ankles, moderate in fourteen, and severe in thirteen17. In the other twenty-eight ankles, the subtalar joints were fused. At the time of follow-up, there was radiographic evidence that rheumatoid arthritis or osteoarthrosis of the subtalar joints was absent or mild in thirty-three ankles, moderate in eleven, and severe in twenty-four. The subtalar joints were fused in thirty-three ankles.
Little information has been published about the long-term results of the Mayo total ankle replacement. In a previous report on 102 patients seen at our institution, Stauffer and Segal23 noted a 72 per cent rate of success after an average duration of follow-up of two years. In the present series, approximately one-third (36 per cent; fifty-seven) of the 160 arthroplasties had failed after a mean of nine years of follow-up; this demonstrates the importance of long-term follow-up when determining the ultimate value of total ankle replacement. Demottaz et al. reported the results after twenty-one arthroplasties (in nineteen patients) in which six different types of implants had been used; the Mayo design had been used in ten. Of these ten Mayo replacements, one had a good-to-excellent result, seven had a fair result, and two had a poor result after a mean duration of follow-up of twelve months (range, three to twenty-two months). Lachiewicz et al. reported that, of fourteen ankles on which a Mayo arthroplasty had been performed, seven had an excellent result and seven had a good result at an average of thirty-nine months. Unger et al. reported the results of twenty-three arthroplasties with a Mayo implant that had been done in patients who had rheumatoid arthritis. After a mean of 5.6 years, the result was excellent for two, good for thirteen, fair for four, and poor for four. In a study by Bolton-Maggs et al., forty-one arthroplasties had been done with the ICLH implant; after a mean of 5.5 years, the result was satisfactory in only thirteen ankles. Takakura et al. reported the results of twenty-five arthroplasties. At eight years, the result was good for six ankles, fair for nine, and poor for ten; no result was excellent.
Other, relatively constrained implants, such as the ICLH design, have also been associated with problems such as loosening of the components1,6,10,11,20. Unconstrained implants that allow for more axial rotation may decrease the risk of loosening, but they have been reported to be less stable. In addition, they may lead to other problems such as malleolar impingement1,24. There has been no appreciable difference in the clinical results after the use of an unconstrained or a constrained device3.
In our study, the mean motion of the ankle in the sagittal plane postoperatively was 5 degrees of dorsiflexion and 19 degrees of plantar flexion, for a total arc of motion of 24 degrees. This range of motion was similar to the mean total arc of motion preoperatively and was less than the 30-degree mean that was found by Stauffer and Segal23. The combined range of dorsiflexion and plantar flexion that is measured during a clinical examination actually represents the combination of motion in the sagittal plane of the ankle and the joints of the midfoot. Therefore, the actual tibiotalar motion is even less than the motion that is measured. However, after ankle replacement, the range of motion of the ankle is adequate for satisfactory function. In our series, no consistent association was found between the range of motion in the sagittal plane and the clinical result. The range of motion in the sagittal plane postoperatively in our patients did not differ substantially from the 30 degrees of motion that has been found for normal subjects walking on a level surface18,23.
According to the criteria of Harris et al., the rate of definite loosening was 8 per cent (eight of 101) for the tibial components and 57 per cent (fifty-eight of 101) for the talar components. However, we could not demonstrate any association between the clinical result and these radiographic findings. Despite this lack of association, the radiographic findings are worrisome because they may be harbingers of trouble.
In summary, primary total ankle arthroplasty with a Mayo implant yielded a good clinical result for thirty-one, a fair result for fifty-five, a poor result for seventeen, and a failure for fifty-seven of the 160 ankles that had been followed for at least two years. The relatively high rates of complications and reoperations are of concern. For these reasons, we no longer recommend ankle arthroplasty with the constrained Mayo implant for rheumatoid arthritis or osteoarthrosis of the ankle.