A prospective observational study involving all patients undergoing revision of an arthrodesis to total ankle arthroplasty was carried out at our institution. The institutional review board of the University of Basel approved the study. All participants provided informed consent prior to surgery. Between August 1999 and December 2004, the senior author (B.H.) performed thirty procedures in twenty-eight consecutive patients who had a painful malunion (or nonunion) or osteoarthritis of the adjacent joints. Total ankle replacement was performed with use of the HINTEGRA prosthesis (Newdeal SA, Lyon, France) (Fig. 1).
The study group included thirteen women (fourteen ankles) and fifteen men (sixteen ankles) with a mean age of 57.8 years (range, thirty-one to eighty-four years). The mean time from the ankle arthrodesis to the index procedure was 13.2 years (range, 0.8 to fifty-seven years) for the twenty-five ankles that had a fusion, and the mean time from an attempted ankle arthrodesis to the index procedure was 4.5 years (range, 1.4 to 12.3 years) for the five ankles that had a nonunion. The original diagnosis leading to ankle arthrodesis was posttraumatic arthritis in twenty-three ankles (with the arthritis having developed after an ankle fracture in twenty cases, after a pilon fracture in two cases, and after a tibial fracture in one case), primary osteoarthritis in four ankles, and secondary inflammatory joint disease in three ankles (including two ankles in patients with hemophilia and one ankle in a patient with tuberculosis). Following the arthrodesis procedure, the patients had undergone an average of 3.3 operations (range, zero to nine procedures). A pantalar arthrodesis of the ipsilateral hindfoot was done in four patients, and a subtalar arthrodesis was attempted in two patients but resulted in subtalar nonunion. The most common reasons for taking down the arthrodesis and performing a total ankle replacement were painful malunion and painful osteoarthritis of adjacent joints (Table I). One patient (Case 9; see Appendix) was excluded from the study because she had continuous pain and underwent revision to a tibiocalcaneal fusion at another institution.
The Prosthesis and Surgical Technique
The HINTEGRA is an unconstrained three-component total ankle arthroplasty system that provides intrinsic stability in the coronal plane (e.g., against eversion-inversion)11. While primary stability of the tibial component is obtained by means of screw fixation and six pyramidal peaks, primary stability of the talar component is obtained by means of press-fit and screw fixation (Cases 1 to 14; see Appendix). Since April 2003, two pegs, instead of screws, have been used for fixation of the talar component (Cases 15 to 30; see Appendix).
In the case of malalignment and concomitant osteoarthritis of the adjacent joints, additional surgical procedures (one-stage procedures) were performed before prosthetic implantation (see Appendix). The surgical technique is illustrated in Figures 2-A through 2-D. Typically, a standard anterior longitudinal incision was used to approach the fused ankle between the extensor hallucis longus and the anterior tibial tendons. When it was not possible to achieve a minimum distance of at least 4 cm between a previous scar and the intended standard approach, an adapted surgical approach through the previous incision was used. The original ankle joint was marked with four Kirschner wires as planned on the basis of the preoperative radiographs (Fig. 2-A). The tibial cutting block was then positioned with use of the Kirschner wires on the anteromedial and anterolateral corners of the ankle and with use of the sagittal plane alignment rod as a reference (Fig. 2-B), and the tibial cut was made. The medial and lateral cuts were then made from distal to proximal, parallel to the other two Kirschner wires, with use of a reciprocating saw. Finally, the tibial resection block was moved distally to the desired position for the second transverse tibial cut and was positioned to remove 4 to 8 mm of bone (Fig. 2-C). After the second transverse cut was made with the saw, the bone segment was removed and the medial and lateral gutters were mobilized with a chisel until the created talus became mobile within the mortise (Fig. 2-D). A special distractor (Hintermann distractor; Newdeal SA) was positioned on both the medial and lateral sides to allow for controlled distraction of the ankle and exposure of the posterior soft tissues. This distractor has two pins, which are slid into the appropriate holes in the spreading forceps and are then introduced at the appropriate location into the bone fragments to be spread open. The scarred posterior capsule was then removed, with care being taken not to damage the flexor hallucis longus tendon and the tibial nerve. The spacer was inserted into the created joint space to allow for the assessment of hindfoot alignment and ligamentous stability of the ankle. In the case of ligament insufficiency, ligament repair was performed following implantation of the prosthesis (the medial ligament was reconstructed in two cases, and the lateral ligament was reconstructed in one complex case). The talar component was inserted first, followed by the tibial component. The thickness of the polyethylene mobile bearing was determined with use of trial implants, and the appropriately sized polyethylene mobile bearing was inserted. The position of the implants was checked with fluoroscopy, and sagittal plane motion and stability were checked clinically.
After surgery, a well-padded short-leg splint held the foot in a neutral position. After two days, a short-leg walking cast was applied. Full weight-bearing was then allowed in the cast. The cast was removed after six weeks when only the ankle had been replaced and after eight weeks when additional procedures such as realignment, ligament reconstruction, and/or adjacent joint fusion had been performed. After cast removal, patients performed intensive walking exercises and received training in ankle motion, balance, and proprioception.
Clinical Evaluation
All patients were seen postoperatively in our outpatient clinic by two independent reviewers who had not taken part in any of the operations. The clinical examination involved careful assessment of ankle alignment and range of motion with the patient standing (passive range of motion) and ankle stability with the patient sitting. The range of motion was determined clinically with a goniometer that was placed along the lateral border of the leg and foot. This examination was performed preoperatively and postoperatively by the same independent reviewer who had not taken part in any of the operations. The patients rated the pain on a visual analog scale ranging from 0 points (no pain) to 10 points (maximum pain). The patients also indicated their level of function in daily activities (walking and climbing stairs, for example) and their satisfaction with the procedure. The hindfoot score was then calculated according to the system of the American Orthopaedic Foot and Ankle Society (AOFAS), which is an unvalidated scoring system12. The gait of patients was observed clinically and then was analyzed with use of pedobarography.
Radiographic Measurements
All postoperative radiographic examinations were performed with use of fluoroscopy to ensure the achievement of standardized and true anteroposterior and lateral views of both components. All radiographs were made with the patient bearing weight.
Angular and linear values were defined to delineate alignment and component migration11,13. These values were measured digitally with a special metric software system (ImageAccess; PIC Systems AG, Glattbrugg, Switzerland).
Loosening of the tibial component was defined as a change in the position of the flat base of the component by >2° relative to the long axis of the tibia and/or as a progressive radiolucent line of >2 mm in thickness on the anteroposterior or lateral radiograph.
Loosening of the talar component, as seen on the lateral radiograph, was defined as subsidence into the talar bone by >5 mm or a change in position of >5° relative to the line drawn from the top of the talonavicular joint to the tuberosity of the calcaneus11. Evaluation of any minor change in position of the talar component on the anteroposterior radiograph was very difficult, and it was not possible to evaluate radiolucent lines beneath the talar component on either view.
In all cases, true foot and ankle motion was measured on lateral radiographs that were made under fluoroscopy while the patient was standing on a footplate. The footplate was plantar flexed and dorsiflexed as much as possible until the tibia started to follow foot motion.
All radiographs were evaluated by three individuals (A.B., M.K., V.V.), and decisions were based on consensus.
Statistical Methods
All experimental data are presented as the mean and the standard deviation. A Kolmogorov-Smirnov normality test was performed to verify whether our data followed a Gaussian distribution. The Kolmogorov-Smirnov distance was calculated. As most of the data were not normally distributed, we compared data in two unpaired groups with use of the Mann-Whitney rank-sum test. The level of significance was set at p < 0.05. To demonstrate a significant correlation between single parameters, we calculated the Spearman rho correlation coefficient according to a "non-normal" distribution of the data.
Source of Funding
No external funding was received in support of the present study. The royalties that the first author (B.H.) received from Integra were given to the research fund at the clinic where the work was performed.
Clinical Results and Patient Satisfaction
In twenty-nine ankles in twenty-seven patients, the AOFAS hindfoot score increased from 34.1 points (range, 12 to 60 points) preoperatively to 70.6 points (range, 42 to 94 points) at the time of the latest follow-up at 55.6 months (range, thirty-six to ninety months) (p < 0.001) (see Appendix). The postoperative score did not differ significantly according to sex, time since the ankle arthrodesis, or patient age. However, the scores were slightly higher for patients with an age of sixty years or more as compared with those with an age of less than sixty years (mean AOFAS score, 74.8 ± 13.8 compared with 68.2 ± 15.1 points; p = 0.08). Pain and function values were also lower for patients who had had more than four previous operations as compared with the other patients (mean AOFAS score, 58.9 ± 16.1 compared with 75.5 ± 10.7 points; p = 0.008). Previous surgery within the year prior to conversion to total ankle arthroplasty also led to a worse result (mean AOFAS score, 57.4 ± 14.8 compared with 77.2 ± 11.8 points; p = 0.04).
Fourteen patients (fifteen ankles; 51.7%) were very satisfied, nine patients (nine ankles) were satisfied, four patients (four ankles) were partially satisfied, and one patient (one ankle) was not satisfied with the result of the total ankle replacement, which correlated well with the clinical results.
While five ankles (17.2%; four patients) were completely pain-free, twenty-one ankles (72.4%; seventeen patients) were moderately painful, with an average visual analog score of 1.7 points (range, 1 to 3 points). The remaining three ankles (10.3%; three patients) had a visual analog score of 5.2 points (range, 3.5 to 7 points), and these patients were partially satisfied or not satisfied with the result. The overall visual analog score was 1.8 points, which is significantly lower than the preoperative visual analog score of 7.5 points (p < 0.001). The location of the pain was periarticular in twelve ankles (41.4%; twelve patients), posterior (Achilles tendon) in nine ankles (31.0%; eight patients), anterior in eight ankles (27.6%; seven patients), medial in seven ankles (24.1%; seven patients), and lateral in three ankles (10.3%; three patients).
The average clinically measured range of motion (24.3°; range, 10° to 40°) amounted to 55.1% of that of the contralateral, unaffected ankle (44.1°; range, 20° to 65°) (n = 25, with bilateral cases excluded). Under fluoroscopy, the average range of true ankle motion (25.5°; range, 8° to 40°) amounted to 56.7% of that of the contralateral, unaffected ankle (45.0°; range, 20° to 62°) (n = 25, with bilateral cases excluded) (see Appendix) (Figs. 3-A, 3-B, and 3-C).
Radiographic Findings
Radiographically, the tibial component was stable in all ankles but one (Case 3), in which the talar component showed subsidence. Although the patient was not pain-free and was only partially satisfied with the result, he considered his situation better than it had been before he had had the surgery.
On the talar side, the bone-component interface could not be seen radiographically and loosening could only be inferred from migration of the component. Talar migration was observed in four ankles but was symptomatic in only two of them. One patient (Case 3; see Appendix) refused surgical revision, and the other patient (Case 9; see Appendix) underwent revision to a tibiocalcaneal arthrodesis after 2.5 years. In the other two ankles, migration occurred within the first four months and no further migration of the component was seen thereafter. These ankles may fail, although the patients were nearly pain-free and were very satisfied at the time of the latest follow-up. Eight ankles (27.6%) had evidence of mild heterotopic periarticular bone formation (see Appendix).
Complications and Revisions
At the time of fusion takedown, intraoperative complications included five fractures of the malleoli (including the medial malleolus in three ankles, the lateral malleolus in one ankle, and a bimalleolar fracture in one ankle) and one transection of the flexor hallucis longus tendon (see Appendix). The fractured malleoli were reduced and were fixed with cannulated screws in four cases. In one case, the fractured lateral malleolus was fixed with a plate. The injured tendon was repaired.
Minor skin necrosis occurred in the central part of the wound in one patient (Case 16; see Appendix) who had undergone five operations since the ankle fusion nine years earlier. The patient achieved uneventful healing. In another patient (Case 28; see Appendix), wound-healing was delayed but occurred uneventfully within eight weeks.
An important consideration is the extent of ligament stability that could be achieved after the arthrodesis, particularly in cases in which the fibula has been osteotomized and used to provide stability for the arthrodesis. Although seven ankles had undergone extensive fibular resection, persistent lateral ligament instability was found in only one ankle (Case 24; see Appendix). In this ankle, the tibiotalar fusion had been accomplished by shortening the distal part of the tibia by 2.5 cm. At the time of revision, the peroneal tendons were found to be unstable and unable to provide lateral stability. The residual portion of the fibula was lengthened, and tenodesis of the peroneal tendons to the fibula was performed. The ankle then became stable and remained stable over time. On the medial side, two ankles had incompetence of the deltoid ligament after the arthroplasty. One ankle (Case 2), in which the deltoid ligament had been damaged when the ankle arthrodesis was done from the medial side, became fully stable after medial ankle ligament reconstruction at the time of the arthroplasty. In the second ankle (Case 22), in which the ankle arthrodesis was done for the treatment of stage-4 posterior tibial dysfunction, some medial ankle instability persisted after the arthroplasty, which had been combined with a triple arthrodesis. Following reconstruction of the medial ankle ligaments with use of a free tendon graft, the ankle became stable. Although the ankle ligaments may have been damaged during the original arthrodesis or may have atrophied after fusion, the remaining soft tissues provided sufficient stability against eversion-inversion forces when combined with the intrinsic coronal plane stability provided by the ankle prosthesis design that we used.
All complications resolved uneventfully and did not influence the outcome. The outcome for ankles that underwent additional surgical procedures was not different from that for ankles that did not have additional surgical procedures (see Appendix).
One talar component (size 1) showed subsidence and loosening (Case 4; see Appendix). A revision was performed after eight months with use of a custom-made implant to achieve more extended osseous support. Although the implant was stable, the patient still experienced severe pain and stiffness at the time of the latest follow-up. However, the patient refused revision to a tibiocalcaneal arthrodesis.
One patient (Case 6; see Appendix) was lost to follow-up. At the time of the latest follow-up, at four years, he reported persistent chronic pain. Although he was only partially satisfied with the result, he felt better than he had prior to total ankle replacement.
Another patient (Case 9; see Appendix) experienced continuous pain and stiffness after surgery. She had undergone an ankle arthrodesis 2.2 years earlier to treat ankle pain resulting from a clubfoot deformity. Because of symptomatic osteoarthritis of the subtalar joint, she underwent subtalar fusion and total ankle replacement. Postoperatively, she gained no motion and, at the time of the latest follow-up, she had 8° of plantar flexion and no dorsiflexion. After 2.5 years, she underwent a tibiocalcaneal fusion at another institution, and she apparently was satisfied with the results of that procedure. This patient was excluded from the present study, as already mentioned.
The treatment options for debilitating end-stage osteoarthritis of the ankle are arthrodesis and arthroplasty. Although numerous reports have described the benefits of ankle arthrodesis2-7, this procedure has been associated with a number of problems, such as pseudarthrosis and malunion14; increased demands on other joints, resulting in degenerative disease1,8; gait abnormalities15; and a long period of convalescence1,9,16,17. Revision arthrodesis of the ankle, arthrodesis of additional joints, and osteotomies may be treatment options for patients who are unsatisfied with an ankle arthrodesis and who have pain. However, following an ankle arthrodesis, stiffness, hindered coupled movement between the ankle and hindfoot joints, and pain may persist. A solution designed to restore some motion might better meet patient expectations. Greisberg et al., in a retrospective study, appear to have been the first to report on the takedown of problematic ankle fusion and successful conversion to total ankle replacement with use of a semiconstrained two-component ankle design10. However, they encountered specific problems such as progressive tilting of the talus within the mortise and subsidence of the implants. Those problems could have been partly explained by the prosthetic design. To our knowledge, we are the first to report on the use of an unconstrained three-component total ankle design.
In the present series, the HINTEGRA total ankle prosthesis was found to be a successful alternative to extended fusion in twenty-nine of the thirty ankles. In seventeen cases there was substantial peripheral osteoarthritis, and in thirteen cases total ankle replacement was supplemented by fusion of the subtalar and/or the talonavicular joint. Five other patients had fusion of the subtalar joint (combined with a talonavicular fusion in all cases but one) at the time of conversion, and five patients had a painful nonunion at the site of the ankle arthrodesis. All of these patients had an improvement in the AOFAS hindfoot score, in part because of an increase in ankle motion (which makes up only 8% of the score), but mainly because of improvements in pain level and function (which make up more than half of the score). The mean AOFAS hindfoot score of 70.9 points was lower than the scores ranging from 74 to 85 points reported in the literature in association with other total ankle prostheses11,18-22. Overall, patient satisfaction was high and was correlated with pain relief and functional outcome, including the resulting range of motion. All patients but one stated that they would accept the same treatment again.
For the surgeon, it is extremely difficult to determine the original center of rotation of the tibiotalar joint following a fusion. When using a two-component design, such as the AGILITY ankle10, the surgeon will have to determine the axis of rotation in this "new ankle," whereas with a three-component design, such as the HINTEGRA ankle, soft-tissue tension can be used to predict the axis of rotation of this "new ankle." Furthermore, the translational freedom at the interface between the polyethylene insert and the flat tibial component will even allow a changing axis of rotation when the foot is moved from dorsiflexion through plantar flexion. The lack of coronal plane stability provided by the AGILITY ankle may explain the malalignment and tilting problems encountered by Greisberg et al.10, and those authors concluded that an ankle with a resected lateral malleolus might be a relative contraindication to this procedure. In the present series, this problem did not occur. We believe that an ankle prosthesis design such as the HINTEGRA ankle that provides intrinsic stability in the coronal plane is essential for the success of this procedure as it is usually extremely difficult to determine the competence of the deltoid and lateral ligaments preoperatively in a fused ankle. In addition, ankle arthrodeses often have been performed with an attempt to achieve some hindfoot valgus. In these ankles, the use of an unconstrained three-component prosthetic design that requires appropriate ligament balancing does not always allow the full correction of valgus malalignment. This results in translational forces of the talus toward the medial malleolus, as seen in three cases (Cases 8, 16, and 23; see Appendix). After a medial sliding osteotomy of the calcaneus, medial impingement was no longer observed, and we believe that any remaining valgus malalignment should be corrected at the time of total ankle replacement to avoid this particular problem.
Of concern was the extent to which ankle motion could be regained after an arthrodesis. Clinically, a mean of 24.3° of ankle motion was found at the time of the latest follow-up. This value corresponded to 55.1% of the value for the contralateral ankle. Maximum plantar flexion (19.0°) was substantially greater than dorsiflexion (5.3°). Similar results were found under fluoroscopy in the assessment of "true" ankle motion. The six ankles in which heel-cord lengthening was performed at the time of surgery did not have consistently better dorsiflexion. Apparently, contracture of the posterior soft-tissue structures is a specific problem after a long-standing ankle fusion.
Taking down an ankle arthrodesis and performing a total ankle replacement is technically demanding. First, it is sometimes very difficult for the surgeon to determine the landmarks for the osteotomies. Extended analysis of the lateral radiographs of the affected ankle and the unaffected, contralateral ankle may be necessary to estimate the previous joint architecture and to determine the best level of resection. Fluoroscopy should always be used when performing the resection cuts as planned preoperatively. Second, it may be extremely difficult to estimate the amount of bone to resect in order to provide enough, but not too much, tension on the surrounding soft tissues when the prosthesis is inserted. As a rule, we resected 4 to 8 mm of bone between the tibia and the talus, which allowed the 12-mm-thick ankle prosthesis to be "squeezed in" while creating an overall distraction of 4 to 8 mm. If, after resection of the bone and the posterior capsule, tension in the soft-tissue structures is too high, additional bone can be resected as required. Third, a major concern after taking down an ankle arthrodesis is the quality of bone stock, particularly the osseous strength of the tibial and talar resection surfaces. On the tibial side, with the prosthesis that we used, we did not see any subsidence or tilting of components during the follow-up period, perhaps because this prosthesis design fully covers the tibial resection surface. On the talar side, we found the quality of bone stock to be much more critical to resist the compressive forces applied to the implant. This was particularly true for the smaller components in the first part of this series (Cases 1 through 14; see Appendix). These components did not have additional pegs for primary stability. When the mortise had been narrowed at the time of the arthrodesis (almost always in this series because of osteotomy or resection of the fibula), a smaller implant had to be used. This increased the forces transmitted between the implant and the bone. This became even more critical as the anterior border of a smaller implant was located more posteriorly on the talar neck. Despite these concerns, the design of the HINTEGRA ankle prosthesis provided reliable results with regard to component stability over the intermediate-term follow-up.
At the time of writing (after a minimum duration of follow-up of three years), only one tibial component had been revised, but three ankles had radiographic evidence of subsidence of the talar component and one ankle had loosening of the talar component. This finding supports our opinion that the talar component is at higher risk for failure because of general undersizing due to narrowing of the mortise after ankle arthrodesis. With inclusion of only the ankles with symptomatic component loosening, the rate of component loosening was 6.7%. With inclusion of the two ankles that showed loosening radiographically but were clinically stable, the failure rate was 13.3%. This rate is better than the failure rates of 16% to 49% that have been reported in recent studies of primary ankle replacements after intermediate to long-term follow-up. The failure rate in the present study was particularly better than the 42.1% failure rate reported by Greisberg et al.10, demonstrating that the use of an unconstrained three-component ankle design may be a key feature in creating motion in ankles that were previously partially destroyed and fused for a long period of time.
Although this procedure did relieve pain in most patients, only five patients were completely pain-free at the time of the latest follow-up. However, all patients but one experienced substantial improvement and were satisfied with the result. These results support our belief that the main advantage of a three-component arthroplasty design is its free adaptation to the individual geometry of the ankle and tension patterns of the soft tissues when it is implanted. This is of particular importance in a fused ankle. In addition, this prosthetic design provides coronal plane stability to overcome the incompetence of the medial and lateral ankle ligaments that can result from the arthrodesis. The favorable results in this small series of thirty ankles support our hypothesis that takedown of fusion may be a viable option in patients with a problematic ankle arthrodesis, particularly those who have a painful malunion or painful osteoarthritis of adjacent joints. Our current contraindications are listed in Table II.
The present study had some limitations. First, the surgeon who performed all of the procedures (B.H.) is one of the designers of the prosthesis system, which raises a concern about possible conflict of interests. However, the clinical evaluation was performed by two independent reviewers, and the radiographic evaluation was performed by three of the authors (A.B., M.K., and V.V.) who had not taken part in any of the operations. Furthermore, the statistical analysis was performed by an independent statistician (A.B.). Second, the AOFAS score was used for the clinical evaluation and it is not validated. Third, the study should have included more patients to be more conclusive. However, we are not advocating that every painful arthrodesis should be converted to an arthroplasty. Finally, this patient cohort will need to be evaluated at five to ten years of follow-up for valid conclusions to be drawn regarding the long-term viability of this surgical procedure.
In conclusion, the results following the takedown of problematic ankle arthrodeses and conversion to total ankle arthroplasty with use of a three-component prosthesis are promising after intermediate-term follow-up.