Periprosthetic fracture of the femur is an uncommon complication after total hip replacement, but appears to be increasing. We undertook a nationwide observational study to determine the risk factors for failure after treatment of these fractures, examining patient- and implant-related factors, the classification of the fractures and the outcome.
Between 1979 and 2000, 1049 periprosthetic fractures of the femur were reported to the Swedish National Hip Arthroplasty Register. Of these, 245 had a further operation after failure of their initial management. Data were collected from the Register and hospital records. The material was analysed by the use of Poisson regression models.
It was found that the risk of failure of treatment was reduced for Vancouver type B2 injuries (p = 0.0053) if revision of the implant was undertaken (p = 0.0033) or revision and open reduction and internal fixation (p = 0.0039) were performed. Fractures classified as Vancouver type B1 had a significantly higher risk of failure (p = 0.0001). The strongest negative factor was the use of a single plate for fixation (p = 0.001). The most common reasons for failure in this group were loosening of the femoral prosthesis, nonunion and re-fracture.
It is probable that many fractures classified as Vancouver type B1 (n = 304), were in reality type B2 fractures with a loose stem which were not recognised. Plate fixation was inadequate in these cases. The difficulty in separating type B1 from type B2 fractures suggests that the prosthesis should be considered as loose until proven otherwise.
Classification systems for fractures are often based on the radiological appearance and should guide the surgeon to the most appropriate reconstructive treatment. For periprosthetic fracture of the femur the most widely used classification is the Vancouver system.3,4 which takes into account the location of the fracture, the stability of the implant and the quality of the surrounding bone stock. The fractures are divided into three major categories depending on their location (Table I⇓). These factors have been included in a treatment algorithm by many authors.2,5,6
Fracture of the femur in association with THR is a severe complication.7–10 Because of the high morbidity,7 treatment has changed from traction and bracing in a cast to open reduction and internal fixation (ORIF), a revision procedure or a combination of both. There is a general consensus that most fractures associated with a well-fixed stem (type B1) can be treated by ORIF alone.11 In the presence of a loose stem (type B2), the recommended treatment involves revision with or without ORIF.8 In fractures associated with poor bone quality and bone loss (type B3) the recommended treatment is revision and augmentation by structural allograft.5,10,12 We have evaluated the results of treatment of these fractures and attempted to identify the risk factors leading to the failure of treatment.
Patients and Methods
In the Swedish National Hip Arthroplasty Register, it is possible to identify patients with a periprosthetic fracture of the femur. A total of 1049 fractures was reported between 1979 and 2000. From the Register and the medical records we determined the patients’ age, gender, diagnosis at the initial operation and the number of previous revisions and implants used. Information about bone grafting was also obtained. The treatment was classified as revision, revision with ORIF and ORIF only. Plate fixation was the main method of treatment in the last group. In some of the earlier cases the fractures were secured by multiple cerclage wires or bands. From the records containing the pre-operative interpretation of the radiographs and the operation notes, we classified each periprosthetic fracture of the femur according to the Vancouver system (Table I⇑).
We also obtained information from the Swedish National Hip Arthroplasty Register about any additional operation after the initial surgery. Patients were sent a questionnaire to validate the registry information.
Failure was defined as a new surgical intervention because of any type of problem which could be attributed to the earlier operation for the periprosthetic fracture of the femur. Closed reduction of a dislocation was not included as a failure since this complication has not been reported in the Swedish National Hip Arthroplasty Register since 1999. Problems not requiring a further operation were not recorded as failures. Any re-operation before December 31, 2002 was included in the study. A Kaplan-Meier survival analysis was performed.
The hazard function of re-operation was estimated as a function of the current age of the patient and the time since the fracture. Then each variable was tested using the Poisson regression model.13 The function was continuous as a function of all continuous variables and was in the form of an exponential equation (β0 + β1x1 + β2x2 … + βn.xn). In order to make the function continuous over time but with different changes at different intervals, the following functions of time were considered: minimum (time, 2) and maximum (time −2, 0). Any linear combination of those two functions was continuous. When h represented the hazard function of re-operation and d the hazard function of death, the five-year probability of re-operation was calculated as follows:
Factors which were significant in the first analysis described above were then used in a second Poisson regression analysis. The aim was to test the strength of the different factors which had been significant in the first test. For all analyses a p value ≤ 0.05 was considered to be significant.
Of the 1049 patients with a periprosthetic fracture, 47 (4%) were classified as type A, 304 (29%) as type B1, 555 (53%) as type B2, 43 (4%) as B3 and 100 (10%) as type C (Table I⇑). By December 31, 2002, 245 (23%) patients had undergone further operations. Of these, 109 (44.5%) have been re-operated on during the first post-operative year. The mean time from the initial operation to re-operation was 22 months (1 to 144). The survival rate at ten years (end-point, any re-operation) was 69.9% (95% confidence interval (CI) 66.1 to 73.7) (Fig. 1⇓). In the re-operation group there were 108 (44%) men and 137 (56%) women. The mean age in the total cohort of patients with a periprosthetic fracture of the femur and in those with failure of treatment was 74 years.
Factors which did not predispose to the risk of failure included gender, primary diagnosis, whether the periprosthetic fracture occurred in a primary or revision hip replacement and the design of the implant used at the primary procedure (Table II⇓).
Factors which contributed significantly to a lower risk of failure are presented in Table III⇓ and those associated with a higher risk in Table IV⇓. There was a significantly lower risk of failure if a long cemented Lubinus prosthesis (Waldemar Link GmbH & Co., Hamburg, Germany) or a long uncemented Bi-Metric stem (Biomet, Warsaw, Indiana) was used. The long distally-fixed prostheses (Wagner (Zimmer, Winterthur, Switzerland) and Lubinus MP) and the Charnley (DePuy, Warsaw, Indiana) and the Exeter prostheses (Stryker, Kalamazoo, Michigan) did not affect the risk. Periprosthetic fractures classified as types A, B3 or C had no increased risk of failure. However, those classified as types B1 or B2 were different. The type-B1 periprosthetic fracture had a high risk of later failure, compared with the other types of fracture. By contrast, a type-B2 fracture had a reduced risk of failure. If an implant was considered to be loose by the surgeon the risk of later failure was significantly decreased.
If the treatment of the periprosthetic fracture was undertaken as revision or revision with ORIF, there was a significant lower risk of re-operation. However, if the operation was done with plate fixation or cerclage wiring (ORIF only) the risk increased. In the revision group, the rate of failure was 17.1% (n = 48) and the corresponding figures for revision and ORIF were 18.7% (n = 83). With plate fixation alone the rate of failure was 33.9% (n = 87) and with cerclage 43.9% (n = 18).
The principle reasons for failure (Table V⇓) were loosening, re-fracture and nonunion. In the plate fixation group, the main problems were re-fracture and loosening. Subsequently, 59% (n = 51) of this group underwent revision. Patients requiring further surgery because of recurrent dislocation were most commonly found in the revision group. The total rate of infection was 2.3% (n = 24) and was twice as common in the plate fixation group than in the revision groups.
All the significant factors from the univariate model were tested together in the Poisson model. Three factors appeared to be the most important. These were the age at the time of fracture, the time since operation on the fracture and plate fixation as the method used. These findings are illustrated in Figure 2⇓, in which two different categories of age were chosen, namely, 60 and 70 years. The risk of failure was increased in the two first years after the operation for the fracture and then reached almost a steady state in both age groups. The risk was higher if there was plate fixation as the only treatment compared with other methods. The risk of failure increased with age, but was not as pronounced as the other two factors.
Our study examined the rate of failure and identifiable factors causing failure in a large series of patients receiving operative treatment for periprosthetic fracture of the femur. The total rate of failure was high in our study. Tower and Beals20 reviewed 86 patients after a periprosthetic fracture of the femur. Of these, 45 (52%) had a poor outcome which was defined as nonunion, sepsis, new fracture or severe deformity. The frequency of re-operation was not mentioned. Springer, Berry and Lewallen21 presented a follow-up of 118 hips in 116 patients who had a revision procedure. At the latest mean follow-up of 5.4 years, 17% had undergone re-operation because of loosening, nonunion, recurrent dislocation, re-fracture or infection. The corresponding rate of failure after a revision procedure in our study was 12%.
There were no differences in age, gender and primary diagnosis in the failure group compared with the series as a whole. It might have been expected that a revision before fracture would be detrimental to the outcome, but this was not found. Three factors in our study were significantly associated with a lower risk of failure. They included the stability of the implant, the Vancouver classification and the selected treatment. If, at the time of fracture, the implant was considered to be loose, categorised as a type B2 fracture, and a revision or revision with ORIF performed, the risk of failure was decreased. The implant used at revision of the periprosthetic fracture of the femur did have some influence on success. A Lubinus long-stemmed cemented prosthesis or an uncemented revision prosthesis was associated with a decreased risk of failure. Experience at the Mayo Clinic has shown that cemented long-stemmed implants perform well in elderly patients with poor bone quality and a simple fracture pattern.21 With proximally porous-coated stems the results were less successful than with uncemented extensively porous-coated implants.
A fracture categorised as type B1 and operated on by plate fixation or cerclage wiring had a significantly increased risk of failure. Many authors2,10,11,22–25 have suggested that type-B1 fractures can be treated by plate fixation (ORIF only), but Tower and Beals20 described a rate of loosening, fixation failure and re-fracture of 13% each in patients treated by plate fixation. Our study has almost identifical findings. The main reason for failure was when plate fixation alone was performed. In the second Poisson analysis, plate fixation was the strongest negative predictor.
The finding that type B1 periprosthetic fractures were associated with an increased risk of failure appears to be contradictory. It would be expected that a fracture around a truly stable stem should unite after adequate ORIF. We suspect that the reason for the poor outcome was that the surgeon, in many cases, misinterpreted the stability of the stem and classified a type B2 fracture as type B1, and subsequently undertook treatment with plate fixation without revision of the stem. The high percentage (59%, n = 51) of revisions performed after a failure in the plate group lends support to our assertion. When the surgeon classified the fracture as type B2, loosening of the stem would have been obvious and revision of the stem performed. Thus the more ‘difficult’ type-B2 fractures were treated by a more appropriate approach with a paradoxical decreased risk of failure. The higher rate of infection among patients with plate fixation compared with those with revision lacked clear explanation. It may be that all patients with revision were given antibiotics and this may have been omitted in isolated plate fixation. Alternatively, plate fixation may have been delegated to a less experienced surgeon.
We suggest that if there is any doubt about the status of the implant, it should be considered to be loose and treated as such. Plate fixation should be reserved for fractures in which there is no doubt about the stability of the implant.
No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.
- Received August 1, 2005.
- Accepted September 14, 2005.
- © 2006 British Editorial Society of Bone and Joint Surgery