Computerised tomography features of giant cell tumour of the knee are associated with local recurrence after extended curettage

Giant cell tumours of bone (GCTBs) are locally aggressive and intermediate (rarely metastasizing) bone tumours that usually occur in young individuals aged 20 to 45 years [1]. GCTBs account for approximately 20% of all musculoskeletal tumours in Asian patients [2]. GCTBs are typically found in the epi-metaphyseal region of long bones, with 50–65% of the most often localizing around the knee [3]. Extended curettage has increasingly become the preferred treatment for GCTBs, but the high recurrence rate (20–30%) after curettage poses a major challenge for orthopaedic surgeons [4‐6].

Some studies have found pre-operative computed tomography (CT) to be valuable in evaluating bone involvement of GCTBs and predicting their recurrence [7, 8]. Puthoor et al. [9] found that patients having prior CT classification had a significantly lower long-term recurrence rate compared to those without pre-operative CT (i.e., 12.9% CT classified versus 30% non-CT classified). Certain features of GCTBs on CT images have been selected to develop a prediction radiomics model in spinal GCTBs, achieving 89% ability to predict local recurrence [10]. However, while GCTBs predominantly arise around the knee joint, only a few studies have specifically focused on this region using CT [8, 11]. Those CT studies found that patients showed a higher recurrence rate with cortical-bone involvement or with short tumour-articular distances, but these rates did not reach statistical significance [12, 13]. Residual tumor tissue related to GCTBs post-surgery was usually considered responsible for local recurrence in these studies, but they did not consider possible surgical factors as causative [12]. Thus, in the present study, by using appropriate statistical methodologies and taking surgical approach factors into consideration, we adopted a more refined method to analyze pre-operative CT images of primary GCTBs around the knee, and we included other risk factors to predict recurrence.

The aims of this study were (1) to analyze the local recurrence rate for primary GCTBs occurring around the knee treated after extended curettage; (2) to identify and analyze the features of GCTBs around the knee on CT images related to recurrence.

For the period between November 2010 and June 2019, 585 patients diagnosed with giant cell tumour of bone were treated in our institution. We retrospectively evaluated records of all patients (n = 188) who underwent extended intralesional curettage for GCTB located in the distal femur or proximal tibia and who had available CT imaging records. A total of 161 patients received a histopathological diagnosis of primary benign GCTB, and 23 patients had a recurrence of GCTB; four patients with malignant GCTB were excluded. Of the 161 patients, 23% (37 patients) were lost to follow-up in the last year; we followed up the remaining patients for at least 24 months. Among the included patients, the average follow-up time was 69.5 months (range: 24–127 months). Finally, a total of 124 patients (64 women, 60 men) were included for analysis; the detailed flow chart of the study is shown in Fig. 1. The average patient age at surgery was 36.4 years (range: 14–72 years). Table 1 summarizes demographic, clinical, and tumour characteristics of the included patients at baseline.

The overall local recurrence rate after extended curettage was 21.0% (26/124), with an average surgery-recurrence interval of 20.7 ± 13.9 months (range, 3–63 months). Local recurrence-free survival at one year was 5.6% and at 2 years was 16.1%. The average age in the recurrence group was significantly younger compared to that in the non-recurrence group (28.50 versus 38.50 years, p < 0.001). The average Ki-67 proliferation index (recurrence: 18.27%; non-recurrence, 15.32%, p = 0.12) and the average tumour size (recurrence: 57.72 mm; non-recurrence: 53.47 mm, p = 0.20) of the two groups were not significantly different. GCTBs with DTAs < 2 mm were significantly related to recurrence (Pearson’s chi-square, p = 0.006). Similarly, DPC was significantly related to recurrence (Pearson’s chi-square, p = 0.001) (Table 1). No other demographic or clinical characteristics of patients were significantly related to recurrence. Recurrence was seen in 12.5% (1 patient) of cement only group, 21.0% (21 patients) of bone graft group, and 25.0% (4 patients) of cement + bone graft group. In the group with adjuvant treatment, the recurrence rate was 13.0% (three of twenty-three) in patients who received pre-operative denosumab, 10.0% (two of twenty) in those who received post-operative denosumab, and 14.3% (two of fourteen) in those who received post-operative bisphosphonate (Table 1). In the subgroup analyses, the surgical approach used for extended curettage had a significant risk difference (RD) for tumour recurrence (posterior approach: RD =  − 50% [95% CI: − 31.48 to − 68.52%]) (see Supplementary Table S1).

To determine which possible risk factors were associated with local recurrence, we performed multivariate logistic regression analysis based on the minimal AIC. The risk factors associated with GCTB recurrence were patient age at surgery, DTA, and DPC (AIC = 95.01). Age was entered as a covariate in the logistics regression analysis. Table 2 shows that DTA (< 2 mm) and DPC were independent risk factors for GCTB recurrence after extended curettage (p < 0.05). The C-index of the final model was 0.79 (95% CI: 0.71 to 0.88). Values greater than 0.7 indicate a good model fit, and indices greater than 0.8 are considered strong models [20]. The Kaplan–Meier survival curves (Fig. 5a and 5b) showed a worse prognosis for recurrence-free survival in patients when the posterior cortical bone was destructed (HR, 3.50 (1.62 7.57); p = 0.001) and the distance between tumour edge and the articular surface was less than 2 cm (HR, 3.36 (1.35 8.37); p = 0.006). In sensitivity analyses (excluding patients receiving denosumab or bisphosphonate), there showed no clear difference in trends (see Supplementary Tables S2 and S3).

GCTBs are intermediate bone tumours that have a tendency to recur locally despite surgical removal [18]. It is believed that residual tumour tissue left after surgery is responsible for local recurrence [21]. We hypothesized that, in GCTBs, destruction of posterior cortical bone and invasion of the articular surface of the knee are two factors that may be associated with a higher risk of missing residual tumour tissue following extended curettage via a typical surgical approach. In the present study, multivariate logistic regression analyses identified DTA (< 2 mm) and DPC, in addition to patient age, as factors significantly related to recurrence. Thus, to reduce local recurrence, our results suggest that accurately determining DTA and DPC in GCTBs prior to extended curettage may represent an important step. In this regard, CT imaging is superior to traditional two-dimensional radiographs for assessing bone invasion and destruction of posterior cortical and subchondral bone; it is more difficult to clearly identify this kind of invasion and destruction in traditional 2-D radiographs because structures overlap on the radiographs [22].

The main limitation of our study was that it is retrospectively conducted at a single center. However, to reduce subject-selection bias, we did include all consecutive patients treated at our hospital who met the study criteria. As a national leading orthopaedic institution, we accept a large number of referral patients from across the country, making our study subject to some degree of referral bias. Also, as adjuvant therapy (denosumab, bisphosphonates) has been extensively used for the treatment of GCTBs and denosumab may increase the rate of local recurrence after curettage [4, 23], the details (dosages and frequencies) of adjuvant treatments were not available in our study. Because denosumab was not approved in Mainland China during the study period, all patients had to obtain denosumab in other countries and regions. It was occasionally not available, and some patients were unable to receive denosumab therapy according to guideline recommendations. Thus, we performed sensitivity analyses by excluding each group of patients receiving adjuvant treatments and the results were stable. Lastly, although GCTBs generally recur within two years [24], six cases had recurrence after two years (23.1% of all recurrent patients). Thus, the possibility of recurrence might still exist for patients who lacked follow-up beyond 24 months.

In the current study, we found the local recurrence rate was consistent with previous studies (Table 3) after extended curettage combined with adjuvant therapy [13, 25‐29]. Although with a lower risk of tumour recurrence, en bloc resection has a greater functional impairment and higher incidence of complications [30]. Joint-preserving intralesional curettage has increasingly become the first choice to retain maximum limb function and achieve a high quality of life, especially common in young patients [31]. As with other authors, the univariate analysis results showed that the recurrence rate did not correlate with sex, pathological fracture, or Campanacci classification [4, 32, 33]. Multiple early studies in the 1990s revealed that cementation might reduce the risk of GCTB recurrence [34, 35]. However, according to recent proceedings, PMMA was no longer considered to have a locally anti-tumour effect and, instead, only a method of mechanical enforcement of the tumoral cavity [36]. Our results also showed no significant difference between the groups of cavity reconstruction. Puthoor et al. [9] classified GCTB into three classes based on CT findings, and they only used it for surgical determination but not for prognostic analysis. In our study, a large absolute but statistically insignificant difference was found in local recurrence rate among groups according to the previous CT classification (class 1, 14.5% vs. class 2, 26.8% vs. class 3, 33.3%), and its clinical significance needs further investigation with more large-scale research. Young people have a higher risk of recurrence, and our study is coincides with the report by Klenke et al. [21] and Kivioja et al. [37].

Other clinical factors may also increase the risk of recurrence. McGough et al. [38] observed that subchondral GCTBs commonly recur, and Suzuki et al. [39] identified an inverse relationship between subchondral bone thickness and recurrence on the articular-surface side. These conclusions were confirmed by our present finding that DTA is a risk factor for recurrence in patients with GCTB of the knee joint. GCTBs occur mostly in the epi-metaphysis of long bones and usually extend to subchondral bone [29]. When subchondral bone is extensively invaded by the tumour, balancing complete tumor removal and preserving articular cartilage can present a surgical dilemma [29]. In such cases, surgeons have limited options, and thus can scrape away only a small amount of affected bone on the joint side [8]. Thus, it is imperative to accurately evaluate the condition of subchondral bone, as failure to do so might lead to insufficient removal of a lesion [38].

GCTBs usually appear as eccentrically developed lesions [7]. Thus, surgeons generally use a medial or lateral approach depending on the side more affected by the tumour, and rarely use a posterior approach, unless the tumour only invades posterior cortical bone [9]. Critical structures, such as nerves and blood vessels, can adhere to adjacent tumour-affected cortical bone, it can be especially difficult for surgeons to scrape and grind the posterior cortical bone and completely remove soft tumour tissue extending to the posterior side. According to Puthoor et al. [9], removal of GCTBs is better approached for curettage through the site of the cortical break, based on CT. When GCTBs invade only the posterior cortical bone, a posterior approach might be advantageous, as it allows sufficient exposure of the operative field. In addition, through a posterior approach, surgeons could avoid deleterious neurovascular complications caused by non-contact thermal necrosis of PMMA [40]. In our series, seven patients received extended curettage through a posterior approach (2 femora, 5 tibias), and none of them had a local recurrence with a minimum follow-up of 24 months. Given the small number of patients, further large-scale prospective studies would be needed to draw a firm conclusion about this issue.

To our knowledge, this is the first study of demographic, clinical, and CT-determined risk factors associated with recurrence in patients with primary GCTB around the knee. We found a higher local recurrence rate in younger patients, greater destruction of posterior cortical bone, and a shorter distance between the tumor edge and articular surface on CT. For the contemporary treatment of GCTB, it is vital to stratify patients by the risk of recurrence. High-risk patients require more careful pre-operative assessment and more rigorous post-operative monitoring.