Between January 1982 and December 1990, forty-seven patients who had a failed total elbow arthroplasty had revision with use of a non-custom semiconstrained Mayo-modified Coonrad implant (Coonrad-Morrey; Zimmer, Warsaw, Indiana) inserted with cement. All patients were followed with use of the Mayo Clinic Total Joint Registry. Preoperative data were obtained from the records, and the patients were contacted at two, five, and ten years postoperatively with the request to complete a questionnaire and to have a radiographic and clinical evaluation. The preoperative and postoperative Mayo Clinic elbow-assessment form was completed by an orthopaedic resident or a physician assistant (R. A. A.), and the questionnaire was completed by the patient. The radiographs were reviewed by two of us (G. J. W. K. and R. A. A.).
Of the original cohort of forty-seven patients, four were lost to follow-up, one died from unrelated causes ten months postoperatively, and one died from metastatic renal-cell carcinoma thirteen months postoperatively. Forty-one patients were followed for an average of six years (range, two to thirteen years). Sixteen patients returned for a physical and radiographic examination and completed a questionnaire, twenty-three mailed the completed questionnaire and a recent radiograph, and two completed the questionnaire and were examined and had radiographs made by their local orthopaedic surgeon. There were twenty-seven women and fourteen men; the average age was fifty-six years (range, twenty-eight to seventy-seven years). Twenty patients had rheumatoid arthritis, twenty had post-traumatic osteoarthrosis, and one had metastatic renal-cell carcinoma.
Before the primary elbow arthroplasty, several patients had had more than one procedure on the elbow, including synovectomy, resection of the radial head, open reduction and fixation of a fracture around the elbow, release of a contracture and excision of heterotopic bone, transposition of the ulnar nerve, ligament reconstruction, tendon transfer, excisional arthroplasty, and arthrodesis. The patients had had one, two, or three total elbow arthroplasties, with use of a variety of implants, before the index procedure: thirty-two patients had had one such procedure, seven had had two, and two had had three. Twenty-two patients had had the earlier arthroplasty at the Mayo Clinic, and nineteen had been operated on elsewhere.
The Mayo elbow performance score14 was used to determine the result at the latest follow-up evaluation. With this system, a score is assigned for pain, motion, stability, and function (Table I). Pain was graded as none, mild (no limitation of activity and occasional use of analgesics), moderate (limitation of activity and regular use of analgesics), or severe (constant pain and regular use of analgesics). The preoperative pain score was an average (and standard deviation) of 12 ± 7 of a possible 45 points. Preoperatively, extension averaged 33 ± 19 degrees and flexion averaged 120 ± 25 degrees. The average preoperative score for motion was 17 ± 5 of a possible 20 points. The elbow joint was graded as stable, mildly unstable (with an end point), or unstable (no end point), and the average preoperative score was 5 ± 4 of a possible 10 points. The average functional score, determined on the basis of the patient's ability to perform normal activities of daily living, was 10 ± 9 of a possible 25 points, indicating that the patients generally had difficulties with simple tasks. Preoperatively, the patients were quite disabled and scored an average of only 44 ± 17 of a possible 100 points.
The indication for the revision was aseptic loosening in thirty-one patients, instability of the joint in seven, fracture of the ulnar component in one, replacement of an incorrectly placed ulnar component in one, and pathological fracture of the olecranon in one patient who had renal-cell carcinoma. Preoperatively, aspiration, technetium bone-scanning, and (in the earlier portion of the study) indium white blood-cell scanning were performed to rule out infection.
Preoperative bone loss in the distal aspect of the humerus was classified as grade I (no patient) when the subchondral architecture was intact, grade II (nineteen patients) when the medial and lateral supracondylar columns had been preserved, grade III (nine patients) when either the medial or the lateral supracondylar column was absent, and grade IV (thirteen patients) when the entire distal aspect of the humerus to or proximal to the level of the olecranon fossa was absent19. The absence of the olecranon process for the osseous attachment of the triceps tendon was recorded as ulnar bone loss. Twelve patients had no olecranon process in the involved extremity.
At the time of presentation, twelve patients had at least one complication associated with either a previous operative procedure or an injury. Three patients had marked restriction of motion secondary to a previous injury or to juvenile rheumatoid arthritis: two had a complete proximal radio-ulnar synostosis and one, heterotopic ossification and an ankylosed elbow. Seven patients had an ulnar neuropathy of varying severity, two had a radial nerve palsy, one had a median nerve palsy, and one had a remote injury of the brachial plexus. Four of the patients who had a nerve injury had tendon transfers before the revision arthroplasty. Seven patients had loosening of the implant in association with a fracture of the medial condyle (two patients), the ulnar shaft (two patients), or the olecranon (three patients). Three patients who had a grossly loose humeral component had a perforation of the humeral shaft by the tip of the prosthesis. One patient who had a pathological fracture of the olecranon because of metastatic renal-cell carcinoma had a revision arthroplasty primarily to treat the fracture.
The semiconstrained Mayo-modified Coonrad implant used in this study was modified in 1981 by adding an extracortical anterior extension or flange to the humeral component14. The flange, in combination with anteriorly placed bone graft, was designed to resist posterior displacement and axial rotation stresses on the implant and to improve load transfer to the supporting bone. The articulation provides approximately 8 degrees of valgus-varus laxity and 8 degrees of rotational laxity21. Initially, a titanium coating was plasma-sprayed onto the proximal portion of the humeral and ulnar components to improve fixation of the implant within the humerus. Beginning in 1987, a porous-ingrowth titanium-beaded surface was used instead of the plasma-spraying.
The design and operative implantation of the Mayo-modified Coonrad implant has been described in previous reports14,15,19. The operative approach was modified depending on the locations of the previous incisions. An extended posteromedial approach, described by Bryan and Morrey, was used in twenty-five elbows; an extended Kocher approach along the line of a previous lateral incision, in eight; a triceps-splitting approach (to allow exposure of the humeral shaft and the radial nerve in patients who had a fracture or severe cortical thinning of the humeral shaft), in four; a trans-olecranon approach, in two patients who had an olecranon fracture; and a triceps-preserving approach, in two patients who had deficient humeral condyles13.
The ulnar nerve was identified and transposed if it had not already been placed in an anterior location at the time of the previous arthroplasty. The radial nerve was routinely exposed only later in the study, after it became clear that injury of the radial nerve was not uncommon during revision elbow arthroplasty. Revision of unstable resurfacing devices was technically challenging because of the difficulty in removing the prosthetic components and the cement from the humerus and the ulna. A cortical window was cut over the tip of the stem to remove the cement and to avoid a fracture or perforation of the cortex. Attempts were made to remove all loose bone cement; however, firmly adherent cement was left in situ if it did not interfere with placement of the stem of the revision component. Loose implants were easier to revise, but a synovectomy was often needed to debride the area of synovitis secondary to particulate debris. A long-stem component (200 millimeters) was used to bypass any fracture (Figs. 1-A and 1-B). In five patients who had an unstable fracture, supplemental internal fixation, such as a plate or cerclage wires, was used to augment the intramedullary fixation provided by the implant. Two patients who had a fracture of the olecranon process had internal fixation with a tension band.
Bone graft was used in all patients and was placed behind the anterior flange of the implant and around any cortical defect or fracture. In two patients, the olecranon was reconstructed with use of an allogenic femoral head graft. Autogenous bone graft was obtained from the iliac crest in twenty-nine patients. Three patients had a staged revision: in one, the revision was postponed in order to rule out a deep infection; in another, a malunited ulnar fracture was treated with use of a Rush rod and the revision was performed nine months later after the fracture had healed; and in the third patient, the prosthesis was contaminated intraoperatively and a second procedure was performed two days later to place the sterile components.
The humeral component was usually 150 or 200 millimeters long; in three patients, we used a 100-millimeter-long component. A small-sized or regular-sized ulnar component was used except in one patient in whom a commercially available extra-long ulnar component was used. Thirty-one patients had revision of both components. Ten patients had revision of only one component: eight had aseptic loosening of the humeral component and a well fixed ulnar component, one had a fracture of the ulnar component, and in one the ulnar component had been placed incorrectly. An intramedullary injection system was used to insert the bone cement; since 1983, one gram of tobramycin has routinely been mixed with each package of cement.
Fourteen patients had seventeen intraoperative complications, primarily consisting of a perforation or a fracture of the cortex sustained during removal of the cement and the prosthetic components. The ulnar shaft was perforated in seven patients, and supplemental internal fixation was necessary in three. One patient had a partial injury (mainly sensory) of the ulnar nerve; the injury did not fully resolve, and the patient was left with residual paresthesias and mild muscle weakness in the ulnar distribution of the forearm and the hand. The humeral shaft was perforated in three patients, injuring the radial nerve. In two of these patients, the nerve injury was caused by extravasation of cement from the cortical defect; in one, it was caused by the instrument used for removal of the cement. In one of the three patients, the nerve was repaired immediately with use of nerve grafts; in the other two, simultaneous tendon transfers with nerve repair were performed at a later date to restore function. One of the three patients continued to have pain and loss of function; the other two ultimately had good restoration of function.
The Mayo elbow performance score14 was used to evaluate the outcome of the revision arthroplasty. Information on the patients who could not return for a follow-up evaluation was obtained from the most recent examination. This was done because we found that, with time, there was no improvement or deterioration in the range of motion, provided that the implant was stable clinically and radiographically. The stability of the elbow was considered adequate if radiographs showed the humeral and ulnar components to be linked and there was no sign of loosening.
The lateral radiograph was used to grade the bone-cement interface. Type 0 indicated no radiolucent line or one less than one millimeter wide and involving less than 50 per cent of the interface; type I, a radiolucent line one millimeter wide and involving less than 50 per cent of the interface; type II, a line more than one millimeter wide and involving more than 50 per cent of the interface; type III, a line more than two millimeters wide and traversing the entire interface; and type IV, gross loosening of the implant14.
Statistical analysis was performed with use of a paired t test to compare the preoperative elbow score and range of motion of the elbow with the values after the revision arthroplasty and to compare the elbow scores for the patients who had rheumatoid arthritis with those for the patients who had post-traumatic osteoarthrosis. A chi-square test was used to compare the rates of failure between these two groups.
Clinical Assessment
The forty-one patients were followed for an average of six years (range, two to thirteen years). All of the patients completed a questionnaire and had recent radiographs. During this period, three patients had a revision of the arthroplasty: two, who had post-traumatic osteoarthrosis, had replacement of the worn polyethylene axis bushing six years after the revision arthroplasty, and one, who had rheumatoid arthritis, had revision of the ulnar component because of a periprosthetic fracture fourteen months postoperatively. A fourth patient, who had post-traumatic osteoarthrosis, had removal of the implant four years postoperatively because of recurrent aseptic loosening.
Twenty-two patients had complete relief of pain, sixteen had mild discomfort, and three remained functionally disabled because of continued pain caused by loosening of the component (one), a pre-existing nerve injury (one), or a nerve injury sustained during the revision arthroplasty (one). The arc of flexion of the elbow improved only slightly after the revision (from an average of 87 ± 31 degrees preoperatively to an average of 101 ± 28 degrees postoperatively; p = 0.01) and remained in the functional range. The average postoperative extension was 27 ± 22 degrees, and the average flexion was 128 ± 14 degrees. Forty patients had a stable elbow at the latest follow-up evaluation; the remaining patient, who had had removal of the implant, had instability that necessitated the use of a brace. Thirty-eight patients were able to perform normal activities of daily living; one had a stiff elbow because of heterotopic ossification; one had weakness of the extremity secondary to an injury of the radial nerve; and one, as just mentioned, had an unstable elbow after removal of the prosthesis.
The average preoperative Mayo elbow performance score was 44 ± 17 points, and the average postoperative score was 87 ± 16 points (p < 0.0001) (Table I). Twenty-two patients had an excellent result; thirteen, a good result; three, a fair result; and three, a poor result. The preoperative (p = 0.3) and postoperative (p = 0.7) elbow scores for the patients who had rheumatoid arthritis were similar to those for the patients who had post-traumatic osteoarthrosis. The patient who had had revision of the ulnar component because of a periprosthetic fracture still had good function of the elbow three years later, with an elbow score of 85 points. Of the two patients who had had replacement of the axis bushing, one had a score of 85 points one year after the second revision and the other had a score of 80 points three years after the second revision. The patient who had had removal of the implant continued to have moderate pain, decreased function, and instability and had a score of 50 points three years after the removal.
Radiographic Assessment
At the time of the latest follow-up examination, thirty-one patients had no radiolucent lines around the humeral component, four had type-I lines, and three had type-II lines. Type-IV lines were seen in two patients who had post-traumatic osteoarthrosis; they had synovitis at the time of the revision arthroplasty, which was thought to be secondary to shedding from the titanium plasma-sprayed components. At the latest follow-up examination, thirty-four patients had no radiolucent lines around the ulnar component, five had type-I lines, and one had type-III lines.
Postoperative Complications
Nine patients had thirteen postoperative complications. At the latest follow-up examination, three of these patients had an unsatisfactory result. Recurrent synovitis secondary to particulate wear debris from the plasma-spray-coated implant developed in three patients. One of them had two synovectomies, and the plasma-spray-coated flange was covered with bone cement to prevent additional shedding of particulate debris. There was no evidence of aseptic loosening; however, the axis bushing was replaced six years after the revision arthroplasty. The second patient had removal of the implant because of persistent synovitis with recurrent aseptic loosening. The third patient, who had an elbow score of 70 points, had aseptic loosening. Despite continued pain and decreased function, the patient refused additional treatment.
One patient had ulnar neuritis nine years after the revision arthroplasty. This patient, unlike those managed later in the series, did not have an ulnar nerve transposition at the time of either the primary or the revision arthroplasty. This patient did not need additional operative treatment. Another patient had bursitis over an old non-union of the olecranon; the site became infected and needed drainage. Two years later, the patient had an avulsion of the tip of the olecranon and was managed with repeat open reduction and internal fixation with tension-band wiring. One other patient who had a non-union of the olecranon after reconstruction with allograft had occasional mild discomfort that was alleviated with use of analgesics. One patient who had had preoperative ankylosis secondary to heterotopic ossification again had heterotopic ossification after the revision arthroplasty. The patient initially had had a crush injury to the extremity, which had been treated with multiple operative procedures to obtain soft-tissue coverage and with tendon transfers. Postoperatively, the latissimus dorsi graft became detached from the olecranon, necessitating reattachment. Despite some improvement in the symptoms, the result was a failure because of persistent pain and stiffness (elbow score, 30 points). One patient with post-traumatic osteoarthrosis had a revision of the axis bushing six years after the revision arthroplasty. The symptoms of squeaking and instability were relieved.
The twenty patients who had rheumatoid arthritis did not need a revision arthroplasty for prosthetic loosening; however, one patient had a revision because of an ulnar fracture through an old external fixation pin track that had not united after immobilization in a cast. Three of the forty-one patients had another revision arthroplasty, and a fourth had removal of the prosthesis.
In recent years, the results of primary total elbow arthroplasty have improved because of better selection of patients, design of the implants, and operative technique3,10,14,15,19,22-24. However, loosening with or without infection, periprosthetic fracture, and recurrent dislocation of the prosthesis continue to occur. These complications often necessitate secondary reconstruction that has commonly been either an arthrodesis or a resection arthroplasty, particularly for loosening with infection16,18.
The design features of the Mayo-modified Coonrad implant make it possible to reconstruct elbows with severe loss of humeral or proximal ulnar bone stock19. Only two patients during the eight-year period of this study needed a revision with custom-designed components. Ferlic and Clayton reported the advantages of use of this implant in the treatment of failed elbow arthroplasty. Dent et al. used custom-designed implants in twelve of twenty-six elbows.
Loss of the humeral condyles and the supporting collateral ligaments did not compromise the result of the revision arthroplasty in the current series; however, the absence of a functioning triceps mechanism due to loss of bone stock from the olecranon was an impediment to a successful reconstruction. The results of allografting to reconstruct an olecranon with deficient bone have proved unpredictable because of resorption and non-union; in our experience, the results have been more predictable when autogenous bone is used. Revision elbow arthroplasty is technically challenging because of the small caliber of the medullary canals and the thin cortical walls, which often have defects due to erosion and osteolysis. Because of the difficulty in removing the components, damage to the supporting bone stock was a particular problem in patients who had had a revision for recurrent dislocation of the prosthesis4,5,8.
In the early years of our study, the radial nerve was injured by extravasation of the cement through a humeral cortical defect in two patients, and the radial nerve was transected in another patient by the instrument used for removal of the cement. We subsequently have avoided this complication by the routine exposure and protection of the radial nerve through a triceps-splitting approach when removing cement from the humerus. Similarly, the ulnar nerve also is routinely identified and transposed to avoid intraoperative injury or subsequent irritation.
Synovitis secondary to the shedding of titanium particulate debris was associated with the initial plasma-spray-coated implant. Since 1987, when the plasma spray was replaced with a beaded surface on the proximal part of the implant, this has not been a problem. Long-term follow-up will be needed to ensure adequate biocompatibility of this modification.
There were no documented postoperative infections in this series of patients, in contrast with the findings in other reports6-8,11,16,17. One of our patients who had persistent synovitis secondary to the shedding of titanium particulate debris was thought to have an infection; however, repeated cultures of specimens obtained during open biopsy and aspiration were normal. The absence of infection in our series may be due to more rigid criteria for the selection of patients, detection of subclinical infection with use of aspiration and technetium and indium white blood-cell scans, meticulous handling of the soft tissues, and the addition of antibiotics to the bone cement.
It is well known that revision arthroplasty is associated with an increased rate of complications in patients who have rheumatoid arthritis or post-traumatic osteoarthrosis. Twelve of our forty-one patients had residual complications from the initial injury or a previous operation. Eleven patients sustained a complication that either influenced the end result (three) or necessitated additional operative treatment (eight). This rate of complications is similar to that reported in our early experience with primary total elbow arthroplasty16,18 and to that reported by others4,5. Even more important is that eight patients needed additional operative procedures and three needed another revision total elbow arthroplasty within six years after the index arthroplasty. One patient who had a loose total elbow prosthesis elected not to have another revision. This rate is markedly better than the 24 per cent rate of revision reported in our initial experience with primary total elbow arthroplasty16,18. The lower functional demands of patients who have systemic arthritis may allow excessive loading of the prosthesis to be avoided and may explain the improved reliability of the implant in these patients. The rate of failure in the present study compares favorably with that reported by Dent et al., who found recurrent loosening in four of twenty-six elbows at an average of thirty-five months postoperatively, and with the results of Ferlic and Clayton, who performed two of nine revisions for either a failure of the bearings or recurrent loosening at an average of twenty-nine months postoperatively. Over-all, the results in the current series were similar to the initial19 and subsequent15 results after treatment of post-traumatic osteoarthrosis.
It must be recognized that revision total elbow arthroplasty is a salvage procedure. When it is successful, the functional results are markedly superior to those of arthrodesis or resection arthroplasty of the elbow. Therefore, an attempt should be made to revise a failed total elbow prosthesis in patients who have sufficient bone stock, an adequate soft-tissue envelope, and no documented infection. Alternative reconstructive procedures continue to be recommended for younger patients who have post-traumatic osteoarthrosis, to avoid the need for premature revision arthroplasty.