The Modied West China Hospital Radiographic Classication for Fibrous Dysplasia in Femur: A Retrospective Analysis of 238 Patients

Background: To investigate the reliability and clinical outcome of a newly developed classication system for patients, who had brous dysplasia (FD) in the femur. Methods: A total of 238 patients with FD in the femur were included in this retrospective study. All affected femurs were measured and treated based on our classication. The intraobserver and interobserver reproducibility were assessed using the Cohen kappa statistic. The clinical outcome was evaluated using the criteria of Guille. Results: At a median follow-up of 60 months, 238 patients were categorized into the following ve types: type I to V. The interobserver and intraobserver kappa scores were excellent. For clinical outcomes, there was no signicant difference in the postoperative Guille score between type I (mean 9.01 ± 1.22), II (mean 8.40 ± 1.38), and V (mean 8.47 ± 1.69). Type III and IV had signicantly lower postoperative Guille scores than type I, II, and V. Moreover, type III had a signicantly higher Guille score (mean 7.81 ± 0.96) than type IV (mean 6.57 ± 2.09). Conclusion: Our classication is reproducible and provides a one-to-one correspondence between diagnosis and treatment. Therefore, we recommend this classication for the diagnosis and treatment of the FD in the femur. loss of proximal femur (focal thinning of cortical bone and/or involvement of the calcar). Type III deformity included coxa vara and femoral shaft deformity, either alone or in combination. Additionally, hip internal or external rotation was measuring. Type IV deformity was characterized by severe genu valgum. And, type IV was to be alone or in combination with type III. Type V deformity was dened as any type associated with severe arthritis of the hip. For polyostotic FD patients involving the bilateral lower limbs, we record the severer side. To eliminate LLD, preoperative planning set the intact side (monostotic cases) or relatively short side (polyostotic cases) as The opening- and closing-wedge osteotomy were provided for preventing LLD intraoperatively. analyzed FD 2009 and 2019 our institution. Radiological diagnosis features, grayish “ground-glass” appearance, endosteal scalloping, shepherd’s crook deformity, and intramedullary expansible lesion with a smooth sclerotic margin[13]. Biopsy with histological evaluation in questionable cases and patients with high suspicion malignancy. type II-V cases histologically FD postoperatively. study study Association


Introduction
Fibrous dysplasia (FD) is a common skeletal disorder (monostotic or polyostotic) that results in pathological fractures, deformity, limping, and pain. FD is a benign intramedullary bro-osseous lesion, which was rst identi ed by Lichtenstein in 1938 [1]. The mutation of the GNAS-gene decreases the GTPase activity of the stimulatory G-protein, which increased the intracellular levels of cyclic adenosine monophosphate (cAMP) and interleukin-6 (IL-6) secretion [2].
The increased intracellular cAMP content and increased IL-6 secretion result in the increased numbers of osteoclasts and bone resorption. Primitive bone has failed to transform into the mature lamellar bone and realign in response to mechanical stress [3]. The lower extremity is frequently affected by deformity, fracture, leg length discrepancy (LLD), and limping [4][5][6][7]. Given the numerous surgical options and complicated deformities in the femur, it can be di cult for orthopedic surgeons to evaluate the severity of deformities and select an appropriate treatment strategy. Ippolito et al [8] have developed a classi cation system that characterizes femoral deformities of FD in six patterns, which only serve to guide the prediction of progression.
The surgical strategy is aimed at bone pain relief [9], restoring normal femoral alignment [5], gait normalization [6], LLD [5], and preventing pathological fracture [4,10] for FD patients. Based on our early experience in assessing deformities and pairing treatment strategies [6,11,12], we modi ed our previous classi cation into ve types following corresponding treatment options, which were as follows: including proximal femur bone loss, neck-shaft angle, femoral shaft deformity, genu valgum, and arthritis of hip. This study aimed to investigate the interobserver reliability and intraobserver reliability of our system and to evaluate the outcomes of surgical procedures based on our classi cation system that pair surgical strategy.

Development of our classi cation system
For radiographic analysis, the following features were determined in the lower extremity radiographs and axial computed tomography (CT) scans complex deformities based on standing limb alignment [13]: focal thinning of cortical bone and involvement of the calcar in the proximal femur, which measured on axial CT [14]; coxa vara, in which the neck-shaft angle was ≤ 120°; metaphyseal and femoral shaft deformity, in which varus or valgus malalignment was detected; genu valgum, in which the mechanical femorotibial angle was ≥ 10°[15]; hip arthritis, in which was detected according to the Kellgren Lawrence grading system[16].

Radiographic Management
Type I lesion was de ned as none of the ve features mentioned above. Type II lesion was de ned as extensive bone loss of proximal femur (focal thinning of cortical bone and/or involvement of the calcar). Type III deformity included coxa vara and femoral shaft deformity, either alone or in combination. Additionally, hip internal or external rotation was measuring. Type IV deformity was characterized by severe genu valgum. And, type IV was to be alone or in combination with type III. Type V deformity was de ned as any type associated with severe arthritis of the hip. For polyostotic FD patients involving the bilateral lower limbs, we record the severer side. To eliminate LLD, preoperative planning set the intact side (monostotic cases) or relatively short side (polyostotic cases) as standard. The opening-and closing-wedge osteotomy were provided for preventing LLD intraoperatively.
We retrospectively analyzed FD patients who were treated between January 2009 and January 2019 in our institution. Radiological diagnosis was made according to the features, including a grayish "ground-glass" appearance, endosteal scalloping, shepherd's crook deformity, and intramedullary expansible lesion with a smooth sclerotic margin [13]. Biopsy with histological evaluation was required in questionable cases and patients with a high suspicion of malignancy. All type II-V cases were histologically proven FD postoperatively. The study was approved by the Institutional Ethics Committee of West China Hospital (Chengdu, China), and the study protocol adhered to the guidelines stipulated in the World Medical Association Declaration of Helsinki.
For correspondence treatment of each type, type I was monitoring every 6 months [10]. And the indication for surgical treatment in type II-V patients included the following: mechanical/ weight-bearing bone pain, hip and/or knee stiff, walking with a limp, fracture, and severe LLD [10]. Therefore, type II was treated with internal xation (IF) following simple curettage [4]. Type III was treated with the IF following single or multiple level valgus osteotomies and simple curettage [5,6]. Type IV was treated with a high tibial osteotomy (HTO) [17] or distal femoral osteotomy (DFO)[18], following femoral malalignment correction.
Type V was treated with total hip arthroplasty (THA) and lesion curettage [9], if necessary single-level valgus osteotomy was performed to t femur stem [12] ( Table 1)  THA, total hip arthroplasty Type I patients followed up with a semi-annual assessment. Early weight-bearing protocols varied in type II-IV. Type II patients were allowed full weight-bearing immediately after surgery. Type III and IV patients were allowed toe-touch weight-bearing within 6 weeks, and partial weight-bearing at 6-12 weeks postoperatively. Then progressive weight-bearing was permitted thereafter [19]. For type V patients who underwent valgus osteotomy, the weight-bearing protocol was the same as that of type III and IV. Without valgus osteotomy, the protocol of type V patients was the same as that for type II patients [12].

Results
Classi cation of our population A total of 238 patients with FD in the femur were enrolled in our institution. Of the included 238 patients, 168 had monostotic FD and 70 had polyostotic FD.
All femurs could be categorized by our radiographic classi cation. The classi cation including ve types of FD was noted ( Table 2).

Clinical Outcome
At a median follow-up of 60 months (range 6-120 months), all patients were evaluated using the criteria of Guille et al [6,21]. With regards to the variations of functional outcomes before and after surgery, there was a signi cant increase in the postoperative Guille score in types II-V, compared to the preoperative values. In type I, there was no signi cant difference between the initial and latest assessments. As for the postoperative Guille score, there was no signi cant difference among type I (mean 9.01 ± 1.22), II (mean 8.40 ± 1.38), and V (mean 8.47 ± 1.69) cases. Type III and IV cases were signi cantly lower postoperatively Guille score than type I, II, and V. Moreover, type III cases had signi cantly higher Guille score (mean 7.81 ± 0.96) than type IV cases (mean 6.57 ± 2.09) ( Table 3). *U = unsatisfactory, A = average, and S = satisfactory. Clinical outcomes were scored as 0 (unsatisfactory), 1 (average), or 2(satisfactory). For a potential max points, >9 points were de ned as excellent, 7 or 8 points as good, 5 or 6 points as fair, and <5 points as poor.
For complications, two type III patients still complained of pain, one type III patients had mild-to-moderate Trendelenburg gait. One type IV patients had a mild Trendelenburg gait. And two type V patients still had mild limping.

Discussion
This study provides a framework for the systematic evaluation and management of the FD-induced-deformity. The spectrum of femoral deformities in fulllength was classi ed into ve categories. Furthermore, their corresponding management was recommended. Currently, there have been only two prior systematic categories [8,11], which classify a variety of proximal femoral deformities in FD. However, both classi cations were failed to guide the most appropriate strategy for treatment. In Ippolito et al's [8] study, three orthopedic surgeons and one pathologist evaluated FD femurs on two occasions with an interval of 6 weeks. The intraobserver (0.855) and interobserver (range, 0.833-0.871) agreement were both excellent. The highest percentage of mistakes was made when distinguishing mild shepherd's crook deformity from severe shepherd's crook deformity. In our primary classi cation system [11], two senior orthopedic surgeons evaluated the cases for two rounds with a 6-week interval. The intraobserver and interobserver agreements were both excellent. In this study, the mean interobserver and intraobserver kappa scores were 0.85 (range 0.77-0.89) and 0.85 (0.79-0.92), respectively. Although the agreement was excellent, the dispute was focused on the genu valgum between types III and IV. In some of the type IV lesions, genu valgum secondary to the restoration of the normal femoral alignment was remarkable after valgus osteotomy. Therefore, the surgeon must be prepared to manage the genu valgum in some type IV patients. Moreover, we recommend classifying these patients into type IV for correspondence surgical options (Fig. 1).
Type I lesions have no focal thinning of cortical bone and involvement of the calcar in the proximal femur, and we recommend conservative treatment with monitoring every 6 months (Fig. 2). Bone pain in FD should be discreetly assessed. Focal or weight-bearing pain may indicate an imminent or impending fracture [22]. Physiotherapy and pain medication can be administered, including opioids and non-steroidal anti-in ammatory drugs [10]. Besides, intravenous bisphosphonate is proposed for persistent, moderate to severe pain, even in children and adolescents [23,24]. Moreover, denosumab, targeting RANKL that is expressed by osteogenic cells, maybe a potential treatment for bone pain caused by FD [25]. During follow-up, no severe complications were detected in type I patients.
Type II lesions are characterized by focal thinning of cortical bone and/or involvement of the calcar, without other femoral deformities. Type II patients often have mechanical or weight-bearing bone pain, which is a signal of stress or an impending fracture [10,26]. Therefore, curettage, bone graft, and internal IF are recommended (Fig. 3). The e cacy and complications of bone graft are still controversial [10,27]. However, cortical allografts were still recommended for the nal and slow internal replacement by the host bone, especially in monostotic FD patients [3,5]. The intramedullary lesion should be adequately bridged by IF including dynamic hip screw (DHS), anatomical plate, or intramedullary nail [4,28,29]. In our study, only one type II patients had mild-to-moderate pain postoperatively.
Type III patients refer to the deformity of coxa vara and/or femoral shaft deformity associated with bone pain. For the single-level osteotomy site, the subtrochanteric region [5,6,30] (Fig. 4) and the dome of the deformity [6,31,32] (Fig. 5) were recommended. However, the double-level osteotomy is strongly considered inadequate for correcting severe deformity [6,31,33]. After osteotomy, orderly curettage, massive impaction allograft, and IF are performed [4,34,35]. The choice of IF is still controversial. Previously, some authors suggested the longer DHS rather than intramedullary nailing, because of its ability in correcting varus and rotational deformities of the femoral neck and simplify procedure [5,30,32]. However, some studies report that intramedullary nails can provide good biomechanical support [6,34]. In general, we recommend intramedullary nail for the following reasons. Firstly, it provides su cient stability that prevents stress fracture and screw loosening or pullout, especially in polyostotic FD patients [5,29]. Secondly, it accommodates multiple-level osteotomy [31]. To increase the initial stability, the transversal surface of the femur after osteotomy should be entirely matched for locking each other. Additionally, the intramedullary lesions with sclerotic rim have su cient bone mass, which provides adequate stability for nail xation. Therefore, the sclerotic bone should be discreetly preserved when curettage and reaming canal. Moreover, rotational deformities of the femur could be gradually corrected. In our study, two type III cases still complained of pain, remained pain, owing to mild hip joint degeneration. One type III patient had mild-to-moderate Trendelenburg gait. Preoperatively, the neck-shaft angle of this patient was only 75°. Over-tensioning of the gluteus medius was inevitable postoperatively after the correction of coxa vara [36].
Type IV are detected in patients with severe genu valgum. When combined with type III deformity, a two-stage treatment was recommended. Complex femoral deformities and lesion curettage were corrected rstly. After six months of rehabilitation, the second-stage procedure was performed for patients, who still complained of typical symptoms. We suggest HTO or DFO, for achieving a satisfactory appearance, correcting limb alignment, and relieving pain [18,37]. However, for polyostotic FD patients, genu valgum is mainly caused by the proximal tibia and distal femur; thus, DFO and HTO are both recommended (Figs. 6   and 7). Type IV lesions are uncommon and the most challenging type and patients with this type of deformity have a lower Guille score, compared to other types of patients. In our study, only one type IV patient had mild Trendelenburg gait, because of the over-tensioning of the gluteus medius.
Type V lesions are found in FD patients with severe hip arthritis, combined with other types (II/III/IV). The deformities of the femoral shaft, coxa vara, and genu valgum, are signi cantly associated with the degeneration of the hip and knee. Additionally, polyostotic FD is found to be more prevalent in hip arthritis than monostotic FD [38,39]. In our study, type V patients had signi cantly higher Guille scores than type III and IV, but no signi cant difference was found between patients with type V and those types I and II. Although a small fraction of FD patients with mild deformity may be classi ed into type V, the relatively high Guille score was preliminary evidence to guide treatment. Sierra et al. [9] rstly reported THA in patients with FD. When hip arthritis has been severe enough, total hip arthroplasty is suggested. Furthermore, a cemented stem was suggested to have a lower revision rate than an uncemented stem. However, our institution reported that the long uncemented stem showed reliable xation at mid-term follow up [12], using Mimics V17.0 Software (Materialise Corp. Belgium), precise preoperative planning, and simulation of the osteotomy. Also, implantation of the prosthesis stem in a three-dimensional reconstructive model is essential. For the femoral component, we recommend fully coated stems, which engage in normal diaphyseal bone bypassing the lesion areas at least 2 femoral canal diameters to decrease the risk of postoperative fractures [9] (Figs. 8 and 9). The stem implantation is a crucial point. First, precision osteotomy was assisted by a patient-speci c instrument, using a micropendulum saw. Second, the femoral cavity was temporarily xed after thoroughly evacuation and bone graft. Third, the femur was reamed following addressing the greater trochanter. Two type V patients had mild limping, because of leg length discrepancy after THA.
There were several limitations to our study. Firstly, the classi cation was retrospective and con ned to radiographic images. Secondly, the follow-up time was signi cantly different among the ve groups. Thus, multicenter studies with a longer follow-up time are needed to make su cient conclusions on our classi cation, especially for type V. Thirdly, surgical strategies and classi cations are appropriate only for adult patients since most patients in our department were adults (median age 29 years, range 7.1-73 years). Finally, although many patients rstly visited the surgical department due to a pathologic fracture, we have treated some patients with conditions that may not be classi ed by our system.

Conclusions
We developed a new classi cation system for FD in the femur which was built on a review of the literature and clinical outcomes. Agreement analysis of the classi cation of our population showed that our classi cation system is reproducible, and clinically directed for standardizing the surgical treatment for these deformities. Moreover, follow-up evaluations showed pain relief and gait improvement in most patients. We believe that our classi cation system provides a one-to-one correspondence between diagnosis and treatment. Therefore, we recommend this classi cation for the diagnosis and treatment of the FD in the femur.

Availability of data and materials
The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
This study was approved and monitored by the Ethical Committee of West China Hospital, Sichuan University in China (No.2019342). All patients signed the informed consent.

Consent for publication
Written informed consent was obtained from all patients for publication of this study and any accompanying images.

Figure 1
The classi cation is indicated by Roman numerals, starting with I and ending with V. Type III, IV, and V have three subtypes (A, B, and C).   A Type III lesion with coxa vara. Fig. 4-B 7 month after surgery. Fig. 4-C The radiograph showed no evidence of recurrence of lesion and re-progress of deformity at 24 months.  A Type IV lesion was associated with genu valgum. Fig 6-B Tibial valgus osteotomy is performed 3 months after the femoral valgus osteotomy.

Figure 7
A Type IV lesion was associated with coxa vara, femoral shaft deformity, and genu valgum.