Classifications in Brief: Pires Classification of Interprosthetic Femur Fractures

History

More than 1 million total joint arthroplasties (TJAs) are performed annually in the United States, with projections suggesting this may increase to approximately 4 million by 2030 [2, 8]. As TJA becomes more frequently performed, the number of revision procedures will necessarily increase, including those done for periprosthetic and interprosthetic femur fractures [1, 8, 9]. An interprosthetic femur fracture (IFF) is defined as a femur fracture between an ipsilateral THA prosthesis and TKA prosthesis [15]. An estimated 19,200 Americans have ipsilateral hip and knee prostheses, with this number likely to increase as more TJAs are performed [4, 7]. This will necessarily increase the number of interprosthetic femur fractures, warranting a streamlined, reproducible classification system with an associated treatment algorithm.

Interprosthetic femur fractures can cause severe morbidity; a high proportion of patients who are treated surgically for IFFs undergo subsequent reoperations, and some eventually have high amputation [1, 7]. Patients who sustain these injuries are typically older, and many have multiple medical comorbidities and poor bone quality, which may be attributed in part to the presence of the prosthesis [1, 9, 15].

Periprosthetic femur fractures have been extensively studied, and the Vancouver classification has proven to be a durable and well-validated classification system to guide the treatment of these injuries [1]. In contrast, partly because of the relative rarity of IFFs, there was no accepted classification scheme until recently, and there have been few reports about the surgical treatment of interprosthetic femur fractures [5, 7, 15]. Prior classification systems were incomplete in their assessment of IFFs because they did not emphasize the most important prognostic factors, nor did they offer a treatment algorithm guided by these factors [13, 14, 17].

The Pires classification system sought to remedy those problems [12]. Before the Pires classification, the Vancouver classification [1] was used to describe interprosthetic femur fractures in conjunction with a periprosthetic distal femoral fracture classification such as those of Su et al. [17], Rorabeck et al. [14] or the Société Française de Chirurgie Ortopédique et Traumatologique [12]. In 2011, Platzer et al. [13] stratified IFFs according to their fracture site and proximity to each prosthesis using a modified Vancouver classification. A subsequent report [10] described a treatment algorithm based on Platzer’s modified Vancouver classification. Similarly, Soenen et al. [16] proposed additional grades for the Société Française de Chirurgie Ortopédique et Traumatologique and Vancouver classifications dedicated to IFFs, emphasizing those involving a stemmed knee implant. In response to a lack of consensus on a treatment algorithm, no prior dedicated classification for IFFs, and an increasing prevalence of IFFs, Pires et al. [12] created their classification.

Purpose

The original description of the IFF classification system of Pires et al. [12] provided both a descriptive classification and a treatment algorithm based on multiple factors, including the fracture site, the stability of hip and knee prostheses, and femoral bone quality. These factors are believed to be important in guiding treatment decisions for patients with IFFs [7, 10, 12, 13, 15]. The creators of the Pires classification hoped to create a uniform treatment algorithm specific to IFFs, one that would theoretically result in improved consistency in the treatment of these patients. Among the chief determinants of a classification scheme’s value are standardization and the creation of a common language and framework of understanding [3]. As the Pires classification has demonstrated inadequate intra- and interobserver reproducibility [5, 11], we recommend it not be used to guide treatment, estimate prognosis, facilitate communication among providers, or stratify patients in research studies.

Description

In this classification, fractures are first stratified by their location. Type I denotes a fracture closer to the femoral stem of a THA than to the femoral component of an unstemmed TKA, whereas Type II denotes a fracture closer to the femoral component of an unstemmed TKA than to the femoral stem of a THA. For Type I and II fractures, a further division is made based on whether the hip or knee prosthesis is well fixed. Type A denotes well-fixed hip and knee prostheses, Type B a loose hip prosthesis and well-fixed knee prosthesis, Type C a well-fixed hip prosthesis and loose knee prosthesis, and Type D loose hip and knee prostheses (Figs. 1 and 2).

Fig. 1.

A-D These illustrations show Pires Type I fractures. (A) The fracture is closer to the THA than the TKA and both the THA and the TKA are well fixed. (B) The fracture is closer to the THA than the TKA, with a loose THA and a well-fixed TKA. (C) The fracture is closer to the THA than the TKA, with a well-fixed THA and a loose TKA. (D) The fracture is closer to the THA than the TKA, and both the THA and the TKA are loose.

Fig. 2.

A-D These illustrations show Pires Type II fractures. (A) The fracture is closer to the TKA than the THA with a well-fixed THA and TKA. (B) The fracture is closer to the TKA than the THA; the THA is loose and the TKA is well fixed. (C) The fracture is closer to the TKA than the THA, with a well-fixed THA and a loose TKA. (D) The fracture is closer to the TKA than the THA and both the THA and TKA are loose.

Type III fractures involve a stemmed TKA femoral component, and these are further categorized by the quality of implant fixation (loose or well fixed) and the viability of interprosthetic bone stock. The classification defines viability as “at least 5 centimeters with no cement and prosthesis components in the fracture site” [12]. Type IIIA represents well-fixed hip and knee prostheses with viable interprosthetic bone. Type IIIB represents well-fixed prostheses with a nonviable interprosthetic bony interval. Type IIIC represents loose prostheses (hip, knee, or both) with viable interprosthetic bone. Type IIID represents loose prostheses (hip, knee, or both) with a nonviable interprosthetic bony interval (Fig. 3) [12, 15]. The authors identified the two main factors believed to be responsible for complications resulting in reoperation in this specific scenario: little or no viable bone between the prostheses, defined as the presence of cement in remaining bone that may impair blood flow to the fracture site, resulting in a decreased likelihood that the fracture will heal, and a short interval of bone in between prostheses, limiting fixation [12, 16]. The authors of this classification stratified these fractures such that each type and subtype have a specific treatment modality (Fig. 4) [12].

Fig. 3.

A-D These illustrations show Pires Type III fractures. (A) The fracture is closer to the stemmed femoral TKA implant than the THA; both the THA and the TKA are well fixed with viable bone between the prostheses. (B) The fracture is closer to the stemmed femoral TKA implant than the THA; both the THA and the TKA are well fixed, with nonviable bone between prostheses. (C) The fracture is closer to the stemmed femoral TKA implant than the THA, with loose implants (hip, knee, or both) and viable bone between the prostheses. (D) The fracture is closer to the stemmed femoral TKA implant than the THA, with loose implants (hip, knee, or both) and nonviable bone between prostheses. Shading is representative of nonviable bone, defined as “at least 5 centimeters with no cement and prosthesis components in the fracture site”. Reprinted from Injury, volume 45. Pires RES, Lourenço PRBDT, Labronici PJ, da Rocha LR, Balbachevsky D, Cavalcante FR, de Andrade MAP. Interprosthetic femoral fractures: proposed new classification system and treatment algorithm, copyright 2014, with permission from Elsevier.

Fig. 4.

This illustration shows the treatment algorithm based on the Pires classification, which depends on prosthetic stability, viability of interprosthetic bone, and the presence of a TKA with a stemmed femoral component. Reprinted with permission from Scolaro JA, Schwarzkopf R. Management of interprosthetic femur fractures. Journal of the American Academy of Orthopaedic Surgeons. 2017;25(4):e63-e69.

Validity

To date, there have been two studies assessing the validity of the Pires classification; one by Pires et al. [11] and another by Jennison et al. [5]. In both of those studies, the κ values (which denote intra- or interobserver agreement) suggested that a high percentage of these fractures could not be classified reliably. As such, we recommend that this classification not be used in clinical practice or research until or unless future studies demonstrate better reliability.

Jennison et al. [5] identified 19 interprosthetic fractures, and their orthogonal radiographs were analyzed by four reviewers (three consultant orthopaedic surgeons specializing in arthroplasty and one specialty registrar in orthopaedic surgery) at two timepoints 4 weeks apart. To assess the value of this classification in guiding surgical treatment, the four reviewers were then asked to make treatment recommendations [5]. That study compared the interobserver and intraobserver reliability of three proposed classifications for IFFs: Pires [11], Soenen [16], and Platzer [13]. The Pires classification demonstrated interobserver reliability of κ = 0.499 and intraobserver reliability κ = 0.636, leaving a high proportion of fractures classified differently by different observers and by the same observer when performed at different times [5]. The authors did not account for or note variability in the agreement of interpretation between the consultant surgeons and the specialty registrar, and they noted that the Platzer classification also had similar degrees of intraobserver and interobserver reliability, at κ = 0.767 and κ = 0.586, respectively [5]. The Soenen classification demonstrated intraobserver and interobserver reliability at κ = 0.318 and κ = 0.489, respectively, which are insufficiently high to recommend for clinical or research use [5].

In a similar study, Pires et al. [11] identified 23 interprosthetic fractures, and orthogonal radiographs were analyzed by five reviewers (three fellowship-trained orthopaedic surgeons—one knee surgeon, one hip surgeon, one traumatologist—and two senior orthopaedic residents). Radiographs were used to classify prostheses according to the Pires classification at two separate timepoints, 30 days apart. This study assessed intraobserver and interobserver variability but did not compare treatment recommendations from the reviewers. Interobserver reliability was assessed among fellowship-trained surgeons at both timepoints, with κ coefficients ranging from 0.424 to 0.559. Between residents and fellowship-trained surgeons, interobserver reliability was assessed at one of two evaluations, with κ coefficients ranging from 0.240 to 0.628. Among residents, agreement was demonstrated at κ = 0.448 for the first evaluation and at κ = 0.370 for the second evaluation [11]. The authors found that intraobserver reliability was too low to be measured with the numbers available [11]. This study, performed by the authors of the original classification, demonstrated similar results to those of Jennison et al. [5]; taken together, these results suggest that providers would disagree about the interpretation of a high proportion of fractures, emphasizing this classification scheme’s lack of utility in clinical practice.

Limitations

As noted, the most important limitation of the Pires classification is that its intra- and interobserver reliability are too low to recommend its use in clinical practice or research, and accordingly, until or unless future studies find better intra- and interobserver agreement, we believe it should not be used.

In addition, as with the Vancouver classification, implant loosening, or lack thereof, plays a crucial role in stratification in the Pires classification [12]. In a criticism that also applies to the Vancouver classification, plain radiographs may be inadequate in some clinical scenarios to make the diagnosis of loosening, making the distinctions among subtypes A through D difficult to distinguish. This emphasizes the importance of intraoperative testing of implant fixation and appropriate preoperative preparation for the possibility of a loose prothesis. Likewise, it can sometimes be difficult on plain radiographs to ascertain whether the bone around some Type III fractures is viable. With this in mind, when the quality of remaining bone stock is in question, the surgeon should be prepared with all available implants for the worst-possible scenario.

Conclusion

Although IFFs are still relatively uncommon, their incidence will continue to rise as higher numbers of ipsilateral THAs and TKAs are performed in older patients [6, 11, 15]. These injuries are associated with substantial morbidity and a high risk of reoperations, and before the Pires classification, there was no streamlined treatment algorithm. The Pires classification sought to fill this void with a treatment algorithm married to a classification scheme [11]. Unfortunately, because of low demonstrated interobserver and intraobserver reliability, we recommend against its use in clinical practice or for the stratification of patients in research studies. Until or unless this classification can be independently validated and shown to be reliable, other approaches to describing these fractures—perhaps simple descriptions—should be used instead.