The Implant
The Agility ankle prosthesis (DePuy, Warsaw, Indiana) is a semiconstrained implant with a titanium tibial component and a cobalt-chromium talar component (Fig. 1). A polyethylene component is secured within the tibial component. The semiconstrained design allows some axial rotation as well as some medial-lateral translation of the talar component within the polyethylene component. The implant permits a flexion-extension arc of 60 degrees. The components are placed without cement; the implant is fixed by means of bone ingrowth. There are right and left prostheses, and, during the study period, the components were available in small, medium, and large sizes.
This implant is unique because of its semiconstrained two-component design, because it has a relatively large surface area for bone ingrowth, and because it takes advantage of arthrodesis of the tibiofibular syndesmosis to allow transfer of weight to the fibula. The tibial component has a porous coating on its entire superior surface as well as along the medial and lateral sides that are in contact with the medial and lateral malleoli. The talar component has a porous coating on the inferior surface that is exposed to the talar bone. The talar component is slightly wider anteriorly than posteriorly, which theoretically makes it more stable in dorsiflexion than in plantar flexion.
Initially, the tibial and talar components were both titanium, but beginning in 1989, and with the twenty-first treated ankle, a cobalt-chromium talar component was used because two talar components had loosened within three years after the initial procedure. Also starting with the twenty-first ankle, the tibial component was made thicker because two tibial components had broken (in the seventh ankle, two and one-half years after the procedure, and in the eighth ankle, three and one-half years after the procedure). The minimum thickness of the polyethylene liner varies (from 3.73 to 4.70 millimeters) depending on the size of the implant.
Operative Technique
The patient is positioned supine with a sandbag under the affected side, and a tourniquet is placed on the proximal part of the thigh. An external fixator (EBI Medical Systems, Parsippany, New Jersey) that permits controlled distraction of the ankle is used to facilitate insertion of the components. The fixator is applied medially, with one pin in the talar neck, one in the calcaneus, and two in the tibia. The external fixator is used to distract and align the joint, thus aiding visualization as well as helping to minimize the amount of bone that must be resected.
Under thigh-tourniquet control, the ankle joint is approached anteriorly, between the tibialis anterior and extensor hallucis longus tendons. The anterior aspect of the capsule is incised longitudinally, and the distal part of the tibia is exposed subperiosteally both medially and laterally. The ankle joint is gently distracted five to ten millimeters, and valgus or varus malalignment is corrected with the external fixator.
A separate anterolateral approach to the syndesmosis is then performed. The anterior tibiofibular ligament is reflected, the tibiofibular syndesmosis is distracted, and its soft tissues are removed. The alignment jig is secured to the tibia with a unicortical half-pin, and the ankle is slightly distracted in the neutral position. The correct size of tibial jig is chosen to remove as little bone as possible while ensuring that all cartilaginous surfaces, including the cartilage from the lateral malleolus, are removed. Usually five or six millimeters of bone is removed from the distal aspect of the tibia and an additional five to six millimeters is removed from the talar dome. Once the components are in place, graft consisting of bone obtained from the resected articular surfaces is placed at the site of the syndesmosis, after which the syndesmosis is stabilized with lag screws. Only one screw was used initially, but, starting with the forty-first treated ankle, two screws were used. The fixator is then removed, and the skin is closed. After dressings are applied, a posterior splint is placed with the foot in neutral position. The patient remains non-weight-bearing for six weeks but is encouraged to remove the splint and to move the ankle several times a day. During the first six weeks, the splint is worn, on the average, for more than twenty-three hours each day.
Patient Demographics
The indication for total ankle arthroplasty was severe pain in the ankle joint that was refractory to non-operative measures. One hundred ankle arthroplasties were performed by the senior one of us (F. G. A.) in ninety-five patients (five patients had a bilateral arthroplasty) from July 18, 1984, to December 29, 1993. Of the ninety-five patients, fifty were male and forty-five were female. The preoperative diagnoses included posttraumatic osteoarthrosis in forty-five ankles, rheumatoid arthritis in twenty-six, primary osteoarthrosis in twenty-six, septic arthritis in two, and psoriatic arthritis in one. No patient was receiving Workers' Compensation or had a claim pending. One patient had been injured in a motor-vehicle accident, and the case was under litigation.
Two patients (two ankles) who had died less than two years after the procedure were excluded from the study. The remaining ninety-eight ankles were evaluated radiographically. The radiographs were made preoperatively, early postoperatively, at six months, at two years, and at the time of the most recent follow-up (average, 4.8 years postoperatively). A total of twelve patients (fourteen ankles), including the two who were followed for less than two years, died from reasons unrelated to the ankle arthroplasty. In addition, one patient had a resection of the implant with subsequent arthrodesis. The remaining eighty-five ankles (eighty-two patients) were the basis for the clinical evaluation in this study. Of the eighty-two patients, forty-three (52 per cent) were male and thirty-nine (48 per cent) were female. The average age at the time of the operation was sixty-three years (range, twenty-seven to eighty-one years), and the patients were followed for an average of 4.8 years (range, 2.8 to 12.3 years). The preoperative diagnoses for the eighty-five ankles were posttraumatic osteoarthrosis (forty-four ankles; 52 per cent), rheumatoid arthritis (nineteen ankles; 22 per cent), primary osteoarthrosis (nineteen ankles; 22 per cent), septic arthritis (two ankles; 2 per cent), and psoriatic arthritis (one ankle; 1 per cent). Two of the three patients with bilateral disease had posttraumatic osteoarthrosis, and the third had severe polyarticular rheumatoid arthritis.
Clinical Evaluation
Follow-up evaluation was carried out independent of the surgeon who had performed the operative procedures. One of us (M. T. P.) conducted an interview and a clinical examination, and three of us (M. T. P., J. J. C., and C. L. S.) performed the radiographic assessment. Fifty-four patients (fifty-six ankles) returned for clinical and radiographic examination. The remaining twenty-eight patients (twenty-nine ankles) completed detailed written and telephone questionnaires and sent radiographs that had been made locally.
The questionnaire related to the level of pain, function (as compared with preoperative function) in activities of daily living as well as in specific activities (sports and climbing stairs, for example), and satisfaction with the procedure. The patients were asked to rate the pain on a scale of 0 to 10 points. No pain was given 0 points; mild pain, 1, 2, or 3 points; moderate pain, 4, 5, or 6 points; and severe pain, 7 points or more. The clinical examination that was performed on the fifty-four patients (fifty-six ankles) involved a careful assessment of the neurovascular status, the alignment of the ankle with the patient standing, and the range of motion of the ankle. When the patient had had a unilateral arthroplasty, the alignment of the ankle was compared with that on the uninvolved side; when the patient had had a bilateral arthroplasty, the alignment was compared with the average value for the uninvolved ankles of the patients who had had a unilateral arthroplasty. The range of motion was determined with use of a goniometer along the lateral border of the leg and foot and was compared, when the patient had had a unilateral arthroplasty, with that on the uninvolved side. An ankle-hindfoot score, according to the system of the American Orthopaedic Foot and Ankle Society24, was then calculated from the data derived from the examination and the questionnaire for the fifty-six ankles that had had a physical examination by one of us, independent of the surgeon who had performed the operation. The ankle-hindfoot score is determined with use of a 100-point rating system based on symptoms, level of activity, need for walking aids, deformity, and motion.
Radiographic Evaluation
The follow-up images included non-weight-bearing anteroposterior, mortise, and lateral radiographs of the ankle made early postoperatively, at six months, at two years, and at the time of the most recent follow-up. The positions of the tibial and talar components were measured on both the anteroposterior and the lateral radiograph. We defined three angular values and one linear value to delineate migration of the component (Figs. 2-A and 2-B). On the lateral radiograph, angle A is formed by the longitudinal axis of the tibia, represented by a line drawn along the posterior border of the tibia, and a line drawn parallel to the superior surface of the tibial component. To assess the position of the talar component on the lateral radiograph, we drew a line connecting the posterior-inferior and anterior-superior aspects of the talus and determined the angle between that line and a line drawn along the inferior edge of the talar component (angle B). On the mortise radiograph, we measured the angle subtended by the longitudinal axis of the tibia, represented by a line drawn along the lateral border of the tibia, and the horizontal surface of the tibial component (angle D). We also measured the vertical distance from the horizontal surface of the tibial component to the tip of the medial malleolus (the medial malleolar line [MML]). In some patients, we were able to measure true ankle motion on lateral flexion-extension radiographs by measuring angle C, which is formed by the posterior border of the tibia and the inferior aspect of the talar component (Figs. 2-A and 3).
Migration of a component was defined as a shift of a minimum of 5 degrees in the measured angle indicating the position of the component. The 5-degree criterion was based on reliability testing and the rotational effects of radiographic positioning. We assessed the accuracy of the measurements by having the same observer (M. T. P.) repeat the measurements on sixty sets of radiographs with no markings on two separate occasions. We found a high level of intra-rater reliability with relatively small errors of the mean. For angle A, the standard error of the mean was 0.7 degree, the range was 0 to 2 degrees, and the 95 per cent bound was 2 degrees. For angle B, the standard error of the mean was 1.2 degrees, the range was 0 to 4 degrees, and the 95 per cent bound was 3 degrees. For angle D, the standard error of the mean was 0.7 degree, the range was 0 to 2 degrees, and the 95 per cent bound was 2 degrees. Although the radiographs were made in the same radiographic suite by the same technologists, rotation of the limb was not absolutely controlled between serial examinations. In view of this and the modest errors of the mean of the repeat angular measurements, we defined a 5-degree change in an angular measurement as radiographically relevant.
On each radiograph, we determined whether the syndesmosis had fused, had had a delayed union (defined as a union that took longer than six months), or had never fused (by the time of the most recent follow-up). Radiolucent lines at the interface between the prosthesis and the bone were also evaluated. On the anteroposterior and mortise radiographs, the tibial component was divided into six zones (Fig. 4-A), and the presence or absence of space between the component and the bone was assessed. Lucency was defined as a radiolucent line of two millimeters in width or less, and ballooning lysis was defined as a radiolucent area of greater than two millimeters. Any change over time was noted. Lucency and lysis were also assessed on the lateral radiograph, but only three zones were identified (Fig. 4-B). The lateral radiograph was also used to evaluate other joints of the foot (in particular, the talonavicular, calcaneocuboid, and subtalar joints) for the progression of osteoarthrosis.
Statistical Methods
The Fisher exact test was used to evaluate the association between pain and the preoperative diagnosis, the rate of union of the syndesmosis, or postoperative plantar-flexion contracture. The Fisher exact test was also used to determine whether there was an association between circumferential lucency or ballooning lysis and the preoperative diagnosis, migration of the component, or the rate of union of the syndesmosis.
The McNemar test was used to evaluate any change in the presence of circumferential lucency or ballooning lysis over time, and the Student t test was used to assess whether age or weight was associated with the presence of circumferential lucency or ballooning lysis.
To examine the association between pain and the age and weight of the patient at the time of the operation, the range of motion of the ankle, and the initial postoperative position of the component as seen radiographically, one-way analysis of variance followed by the Tukey test for multiple comparison of measures was used. This analysis involved a comparison of the average ages, weights, ranges of motion, and initial postoperative positions of the components associated with no pain, mild pain, moderate pain, and severe pain.
Complications
Because of persistent pain, one ankle implant (in a patient who had rheumatoid arthritis) was removed and a tibiotalar arthrodesis was performed two and one-half years after the initial arthroplasty. Of the remaining eighty-five ankles, four were revised. One revision was done because of a fracture of the tibial component; two, because of loosening of the talar component; and one, because of malpositioning of the talar component during the initial procedure. One other tibial component fractured, but the patient was not symptomatic and had no additional treatment. There were no deep infections. There were two superficial wound infections in the immediate postoperative period. These resolved after changes of the dressing and oral administration of cephalexin (500 milligrams four times a day for ten days). Six patients had isolated decreased sensation in the distribution of the superficial peroneal nerve, but it did not cause a functional problem.
Clinical Results
Pain relief was our main criterion of success. Of the eighty-five ankles that were available for follow-up, forty-seven (55 per cent) caused no pain according to the patient; twenty-four (28 per cent), mild pain; fourteen (16 per cent), moderate pain; and none, severe pain. Of the eighty-five ankle arthroplasties, eighty-three (98 per cent) were considered to have provided pain relief and two were not considered to have provided relief. Sixty-five (79 per cent) of the eighty-two patients stated that they were extremely satisfied with the result of the procedure, eleven (13 per cent) stated that they were satisfied, three (4 per cent) were indifferent, and three (4 per cent) were either disappointed or very unhappy with the result. Seventy-eight (95 per cent) of the eighty-two patients stated that they would have the operation again, and seventy-nine (96 per cent) said that they would recommend the operation to a friend. Sixty (73 per cent) of the eighty-two patients reported an increase in the functional level because of the ankle replacement. Sixty-eight (83 per cent) of the eighty-two patients did not take any pain medication for the ankle, whereas fourteen (17 per cent) regularly did so. None of the patients who had had a bilateral ankle arthroplasty took pain medication because of the ankles.
The patients who had posttraumatic osteoarthrosis reported significantly more pain at the time of follow-up than did either those who had primary osteoarthrosis or those who had rheumatoid arthritis (p < 0.05). Seventeen (40 per cent) of the forty-two patients who had posttraumatic osteoarthrosis were pain-free, whereas fifteen of the eighteen who had rheumatoid arthritis and thirteen of the nineteen who had primary osteoarthrosis were pain-free. Conversely, nine (21 per cent) of the forty-two patients who had posttraumatic osteoarthrosis, two of the eighteen who had rheumatoid arthritis, and one of the nineteen who had primary osteoarthrosis had moderate pain. Both patients who had septic arthritis also reported moderate pain. Of the patients who had had a bilateral arthroplasty, the one with rheumatoid arthritis had no pain in either ankle, one with posttraumatic osteoarthrosis had no pain in either ankle, and the other patient with posttraumatic osteoarthrosis had mild pain bilaterally.
With the numbers available for study, we did not find an association between the age or weight of the patient at the time of the operation and postoperative pain.
Few patients needed walking aids postoperatively, although fifty-six (68 per cent) of the eighty-two patients reported on the questionnaire that they felt more comfortable walking in shoes, especially ones with a small, elevated heel. Four patients used a cane on a regular basis, and three wore a brace because of a valgus deformity. One of the patients who had a deformity considered the ankle to be painless, and the other two rated the pain as moderate. One of the patients who reported moderate pain despite the use of an ankle-foot orthosis had had six pain-free years after the ankle replacement.
As stated, fifty-six ankles were examined at the time of follow-up. The patients who had had a unilateral ankle arthroplasty usually had less motion on the side of the arthroplasty. The average range of motion in the sagittal plane was 36 degrees (range, 10 to 64 degrees). Twenty-eight (50 per cent) of the fifty-six ankles had a plantar-flexion contracture. The contractures averaged 7 degrees (range, 1 to 20 degrees). Six (11 per cent) of the ankles dorsiflexed to neutral, and twenty-two (39 per cent) were able to dorsiflex beyond neutral (average, 4.6 degrees [range, 1 to 10 degrees] beyond neutral). With the numbers available, we were unable to detect an association between pain and a postoperative plantar-flexion contracture. However, thirty-seven (45 per cent) of the eighty-two patients stated that they were not able to climb stairs as well as they would like; ten (27 per cent) of the thirty-seven blamed the ankle, whereas twenty-seven (73 per cent) blamed other problems (age, other joints, and so on).
An ankle-hindfoot score was calculated, with use of the system of the American Orthopaedic Foot and Ankle Society24, for all patients who had had a complete follow-up examination. The average overall score was 85 points (range, 40 to 100 points).
Radiographic Results
A radiographic assessment was performed on ninety-eight of the 100 ankles in the initial consecutive series. The other two ankles were excluded from the analysis because the patients had died less than two years (0.2 and 1.5 years) after the operation. All other patients, including the other ten who had died, had been followed for at least two years.
Of the ninety-eight ankles (ninety-three patients) available for radiographic evaluation, sixty-one (62 per cent) had a successful fusion of the syndesmosis (Fig. 5), twenty-eight (29 per cent) had a delayed union (more than six months were required for radiographic union), and nine (9 per cent)—including the ankle that had removal of the implant and tibiotalar arthrodesis—had a non-union.
Of the eighty-five ankles in the patients who had completed the pain questionnaire, fifty-four (64 per cent) had a solid union. Of these fifty-four ankles, thirty were not painful, fourteen were mildly painful, and ten were moderately painful. Of the twenty-three ankles (27 per cent) with a delayed union, eleven were not painful, eight were mildly painful, and four were moderately painful. Of the eight ankles (9 per cent) with a non-union of the syndesmosis, six were not painful, two were mildly painful, and none were moderately or severely painful.
There was at least a 5-degree change in the radiographic position of nineteen components (seven talar and twelve tibial). None of the seven migrations of the talar components were associated with a delayed union or a non-union of the syndesmosis, whereas eight of the twelve migrations of the tibial components were associated with a delayed union or a non-union. Migration of the tibial component was found in five of the nine ankles that had a non-union of the syndesmosis, three (11 per cent) of the twenty-eight ankles that had a delayed union, and four (7 per cent) of the sixty-one ankles that had a successful fusion. The risk ratio for migration of the tibial component for ankles with a non-union compared with those with a solid fusion was 8.5 (95 per cent confidence interval, 2.8 to 25.8; p < 0.001).
Non-union was also clearly associated with ballooning lysis at the interface between the bone and the tibial component (Fig. 6). By six months after the procedure seven of the nine ankles with a non-union had radiographic evidence of ballooning lysis, and by two years all of the ankles with a non-union had evidence of ballooning lysis. At the time of the most recent follow-up, ballooning lysis was seen in eighteen (64 per cent) of the twenty-eight ankles with a delayed union and in seventeen (28 per cent) of the sixty-one with a solid union. The prevalence of ballooning lysis in the ankles with a non-union was significantly higher than that in the ankles with a delayed or solid union (p < 0.0001). Ballooning lysis was found only in zones 5 and 6. Similarly, the prevalence of ballooning lysis in the ankles with a delayed union was significantly higher than that in the ankles with a solid union (p < 0.05).
Circumferential lucency around the tibial component was seen on the mortise radiographs of twenty-five (26 per cent) of the ninety-eight ankles at the time of follow-up. The lucent lines were rarely progressive after two years (Figs. 7-A and 7-B). Circumferential lucency around the tibial component on the lateral radiograph was associated with non-union of the syndesmosis (p < 0.01). A circumferential line was found in four of the nine ankles with a non-union, two (7 per cent) of the twenty-eight with a delayed union, and one (2 per cent) of the sixty-one with a solid union.
The position of the tibial component had an effect on the outcome. Tibial components that had been placed in more than 4 degrees of valgus, as seen on the mortise radiograph, were associated with a significantly higher rate of pain (p < 0.05). No appreciable progression of degenerative changes over time was observed in the subtalar, talonavicular, or calcaneocuboid joint. We could detect no relationship between age, weight, or preoperative diagnosis and the development of lysis, lucency, or migration of the component.
The early clinical results with the Agility ankle implant are encouraging, although the radiographic findings remain a cause for concern. At the time of follow-up, however, the radiographic findings were not found to be related to the clinical findings. Non-union and delayed union of the syndesmosis were associated with higher rates of ballooning lysis at the interface between the bone and the tibial component. In the present series, fusion of the syndesmosis was tantamount to stability of the tibial component. Despite the somewhat worrisome radiographic appearance of the bone-implant interface, we found no association between pain and the rate of union of the syndesmosis. Eighty (98 per cent) of the eighty-two patients reported that they had obtained some pain relief and seventy-eight (95 per cent) stated that they would have the operation again.
In this study, we defined lucency and lysis according to the width of the radiolucent line around the component. The time-course and appearance of these radiographic lines suggest that, in most patients, they were not due to the same biological phenomenon that occurs after other arthroplasties as a result of particles of debris. Lucent lines around these ankle implants were always apparent within two years after the procedure and were typically non-progressive. Similarly, after a few years most areas of lysis (defined as a radiolucent line that was wider than two millimeters) were non-progressive and probably indicated a lack of fixation of the implant. The ballooning lysis almost always occurred along the lateral margin of the implant, between the fibula and the tibial component. The development of lysis in this area was more frequent in the ankles that had a delayed union or a non-union of the syndesmosis. This finding indicates that these radiographic findings may, in most patients, represent osseous resorption in an area of high interfacial shear stresses between the implant and the residual lateral malleolus.
At the time of the study, there had been five revisions after the first 100 ankle arthroplasties. Two of the first, titanium talar components loosened and were revised. Loosening of the talar component has not been a problem since cobalt-chromium talar components began to be used early in the series. Another talar component was revised in the first three months because it had been placed in an improper position at the time of insertion. Two tibial base-plates fractured, but only one was revised; the other one was not causing symptoms. No tibial base-plate has fractured since 1989, when the base-plate was made thicker. Finally, one patient had removal of both components and arthrodesis of the ankle secondary to persistent pain. This patient had a non-union of the syndesmosis and no bone ingrowth.
Several features of the implant's design are unique and may be responsible for the relative clinical success found in this follow-up study. First, the semiconstrained, two-component design allows slight medial-lateral translation as well as axial rotation of the talar component within the tibial component with motion. Constraint is therefore somewhere in between that of the more restrictive constrained designs and that of the less restrictive non-constrained (usually three-component) designs3,6,25. The reported high rates of loosening of constrained components may have been secondary to increased forces at the bone-prosthesis interface, while non-constrained designs have been plagued with problems such as altered kinematics, instability, and impingement3. Moreover, the Agility implant relies on bone ingrowth, rather than methylmethacrylate, for fixation. Most of the older total ankle implants were inserted with cement. The results from more recent reports of implants inserted without cement have demonstrated overall better results6,25,41,47. Because of its unique design, the Agility ankle implant has a relatively large surface area for bone growth into the tibial component and it allows stress transfer by incorporating fibular weight-bearing through an arthrodesis of the syndesmosis.
The early experience with the Agility ankle implant is encouraging; the results appear to be better than those associated with most other ankle prostheses. In the present series, most patients were satisfied with the outcome of the operation and were functioning well with the implant. Additional clinical and radiographic follow-up is needed to determine the long-term efficacy of this design and to clarify whether delayed union or non-union of the syndesmosis is associated with an increased rate of clinically important problems.