Post-metastasis survival in extremity soft tissue sarcoma: A recursive partitioning analysis of prognostic factors

European Journal of Cancer (2014) 50, 1649– 1656

Available at www.sciencedirect.com

ScienceDirect journal homepage: www.ejcancer.com

Post-metastasis survival in extremity soft tissue sarcoma: A recursive partitioning analysis of prognostic factors Seungcheol Kang a,b, Han-Soo Kim a,b, SungJu Kim c, Wanlim Kim a,b, Ilkyu Han a,b,⇑ a

Department of Orthopaedic Surgery, Seoul National University Hospital, 101 Daehak-ro Jongno-gu, Seoul 110-744, Republic of Korea Musculoskeletal Tumor Center, Seoul National University Cancer Hospital, 101 Daehak-ro Jongno-gu, Seoul 110-744, Republic of Korea c Department of Statistics, Korea University, 145 Anam-ro, Seongbuk-gu, Seoul 136-701, Republic of Korea b

Received 25 January 2014; received in revised form 26 February 2014; accepted 1 March 2014 Available online 3 April 2014

KEYWORDS Soft tissue sarcoma Extremity Metastasis Survival Recursive partitioning analysis Grade Metastasectomy Disease-free interval

Abstract Background: Recursive partitioning analysis (RPA) enables grouping of patients into homogeneous prognostic groups in a visually intuitive form and has the capacity to account for complex interactions among prognostic variables. In this study, we employed RPA to generate a prognostic model for extremity soft tissue sarcoma (STS) patients with metastatic disease. Methods: A retrospective review was conducted on 135 patients with metastatic STS who had undergone surgical removal of their primary tumours. Patient and tumour variables along with the performance of metastasectomy were analysed for possible prognostic effect on post-metastatic survival. Significant prognostic factors on multivariate analysis were incorporated into RPA to build regression trees for the prediction of post-metastatic survival. Results: RPA identified six terminal nodes based on histological grade, performance of metastasectomy and disease-free interval (DFI). Based on the median survival time of the terminal nodes, four prognostic groups with significantly different post-metastatic survival were generated: (1) group A: low grade/metastasectomy; (2) group B: low grade/no metastasectomy/ DFI P 12 months or high grade/metastasectomy; (3) group C: low grade/no metastasectomy/DFI < 12 months or high grade/no metastasectomy/DFI P 12 months; and (4) group D: high grade/no metastasectomy/DFI < 12 months. The 3-year survival rates for each group were: group A, 76.1 ± 9.6%; group B, 42.3 ± 10.3%; group C, 18.8 ± 8.0%; and group D, 0.0 ± 0.0%. Conclusion: Our prognostic model using RPA successfully divides STS patients with metastasis into groups that can be easily implemented using standard clinical parameters. Ó 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: Department of Orthopaedic Surgery, Seoul National University Hospital, 101 Daehak-ro Jongno-gu, Seoul 110-744, Republic of Korea. Tel.: +82 2 2072 0682; fax: +82 2 764 2718. E-mail address: [email protected] (I. Han).

http://dx.doi.org/10.1016/j.ejca.2014.03.003 0959-8049/Ó 2014 Elsevier Ltd. All rights reserved.

1650

S. Kang et al. / European Journal of Cancer 50 (2014) 1649–1656

1. Introduction Soft tissue sarcoma (STS) represents a group of cancers that exhibit mesenchymal differentiation, accounting for approximately 1% of all adult malignancies [1–3]. STSs tend to metastasise in an early stage, mainly hematogenously with a predilection for the lungs and less frequently metastasise to liver and bone [2,4] Lymphogenic spread is relatively uncommon in STS [5], except for certain histological types such as the rhabdomyosarcomas, synovial sarcomas and epithelioid sarcomas [6,7]. About 10% of patients presents with metastatic disease [2,5,8], and almost one-quarter of patients with localised disease develop metastases in due course [2,9,10]. In general, prognoses of patients with metastatic STS remain poor, with the 3-year survival rate of 20–30% [11,12]. However, prolonged survival has been demonstrated in some patients, particularly in those with metastases amenable to resection. Although several clinicopathological parameters, such as histological grade, tumour size, disease-free interval and resectability have been suggested to be predictive of survival in metastatic STS, a better prognostication of metastatic STS is needed to guide decisions regarding adjuvant therapy and surveillance [13–16]. Estimating the aggregate risk based on the presence or absence of multiple factors is warranted, preferably using readily available clinicopathological parameters. In this regard, we sought to generate a prognostic model for STS patients with metastatic disease using recursive partitioning analysis (RPA). RPA enables grouping of patients into homogeneous prognostic groups based on multiple variables and provides an easily interpretable method for classifying patients [17,18]. In this study, we employ RPA to divide STS patients with metastatic disease into clinically useful prognostic groups. 2. Patients and methods 2.1. Patients From the prospectively collected database of our institute, 476 consecutive patients who had undergone surgical removal of extremity STS from February 1995 to May 2011 were reviewed. Among these 476 patients, we identified 153 (32.1%) patients with metastatic disease, who either presented with (n = 55) or developed (n = 98) metastases during follow-up. Of the 153 patients, patients with a follow-up duration shorter than 6 months (n = 8) and patients with isolated lymph node metastasis (n = 10) were excluded, which left 135 patients for analysis. In patients with systemic metastasis and lymph node metastasis, lymph node metastasis was regarded as locoregional disease, not as a metastasis

[19,20]. The mean follow-up duration was 18.9 ± 18.9 months. The institutional review board of our institute approved this study. 2.2. Prognostic variables Medical records were reviewed for the potential clinicopathological factors that might influence post-metastatic survival in STS: (1) patient demographics, (2) factors related to the primary tumour, (3) pattern of metastasis and (4) treatment of metastasis. For demographic data, patients’ gender, age and period of diagnosis were investigated. There were 85 males (63%) and 50 females (37%). The mean age at the time of metastasis was 49 years (range, 11–92 years). The patients’ ages were dichotomised as <50 years and P50 years for analysis. Forty-six patients (34%) presented after an unplanned removal of a STS before the correct histological diagnosis was made, without regard for the necessity to remove a margin of normal tissue surrounding the tumour. For factors related to the primary tumour, anatomical site, histological diagnosis, histological grade, tumour size and tumour depth were investigated. Anatomical site of primary tumour was classified as upper extremity (n = 37, 27%) or lower extremity (n = 98, 73%), and as proximal extremity (n = 93, 69%) or distal extremity (n = 42, 31%). Most common histological diagnoses were undifferentiated pleomorphic sarcoma (UPS, n = 34), synovial sarcoma (n = 23), liposarcoma (n = 15) and leiomyosarcoma (n = 12). As for histological grading of the primary tumour, there were four grade 1 (3%), 56 grade 2 (42%) and 70 grade 3 (52%) tumours according to the Federation Nationale des Centres de Lutte Contre le Cancer (FNCLCC) classification system [21,22]. Size of the primary tumour, defined as the largest diameter on the pathological examination report or preoperative magnetic resonance imaging (MRI), was 9.8 cm (range, 1.1–35.0 cm). For the purpose of analysis, tumour size was dichotomised by a cut-off value of 5 cm. Tumours located exclusively above the superficial fascia were defined as superficial. Tumour depth could be determined in 128 patients (95%) with nine superficial and 119 deep tumours. All primary tumours were resected and pathologically negative surgical margins were achieved in 112 patients (83%). Postoperative radiation therapy was administered in 79 patients (59%), all of whom received external beam radiation with the median dose of 60 Gy (range, 50–65 Gy). Postoperative chemotherapy was administered in 21 patients (23%). For surveillance for distant metastasis, chest imaging and bone scans were performed every 3–4 months for 2 years, then every 6 months for the next 3 years, and then annually. Imaging of the primary site was done with MRI or ultrasound based on the risk of local recurrence.

S. Kang et al. / European Journal of Cancer 50 (2014) 1649–1656

Forty-two patients presented with metastasis at the time of diagnosis (synchronous group, 31%) and 93 patients developed metastasis during follow-up after resection of the primary tumour (metachronous group, 69%). There were no significant differences in characteristics between the two groups except for the tumour size >5 cm (87% in the synchronous group versus 69% in the metachronous group, p = 0.027) and positive pathological margin at primary surgery (29% in the synchronous group versus 12% in the metachronous group, p = 0.017). In the metachronous group, metastasis-free interval (MFI) was recorded as the time from resection of the primary tumour until the development of metastasis. The median MFI was 16 months. Local recurrence was identified in 51 patients (38%), with 24 (47%) occurring prior to the diagnosis of metastasis. In the metachronous group, disease-free interval (DFI) was defined as the time from resection of the primary tumour until the first detection of metastasis or local recurrence. The median DFI was 13 months. The most common sites of initial metastases were lung (n = 97, 72%) and bone (n = 19, 14%). Four patients (3%) had multiples sites of metastases while 131 patients (97%) presented with single site of metastasis. The number of lung metastasis was counted using the chest CT scans. A cut-off value of 5 was used to dichotomise the number of lung metastasis [23]. Among the 64 patients with lung only metastasis, the number of lesions were >5 in 23 patients (n = 36%) and 65 in 41 patients (64%). Regarding the treatment of metastasis, the performance of metastasectomy, the administration of chemotherapy and the first-line chemotherapy regimen were investigated. When diagnosed with metastases, patients underwent metastasectomy unless a contraindication existed. Contraindications included unresectable diseases and the presence of significant comorbid disease to tolerate surgery. In all, 44 patients (33%) underwent metastasectomy (30 patients for lung, four for bone, 17 for lymph node and six for other sites, note: duplicated in some cases). Metastasectomy was performed with the intent to resect all clinically evident metastatic diseases. For the patients with lung metastasis, 16 patients underwent thoracotomy once, eight patients twice, four patients three times and two patients four times. Of the 44 patients who underwent metastasectomy, 37 patients had chemotherapy. Patients were given chemotherapy based upon several factors such as histological grade, histological type and age [11,24], but no prospectively selected criteria was used. In all, 91 patients (67%) underwent chemotherapy for metastasis. In consideration of the heterogeneity of chemotherapy regimens used, three sets of comparison were made; ifosfamide-containing (n = 75, 82%) versus non-ifosfamide-containing, doxorubicin-containing (n = 51, 56%) versus non-doxorubicin-containing and ifosfamide and doxorubicin combined (n = 39, 43%) versus others.

1651

For the selection of treatment modality, a specialist team for musculoskeletal sarcoma, composed of orthopaedic oncologists, medical oncologists, radiation oncologists and musculoskeletal radiologists, regularly discuss and decide the treatment modalities. However, no prospectively selected criteria are used for treatment selection. 2.3. Statistical analysis Disease-specific survival was used as the endpoint of the study. Post-metastasis survival was measured from the date of any first evidence of metastasis to the date of the patient’s death. Patients who were alive were classified as censored observations at the time of the last follow-up. Various clinicopathological factors were analysed for possible prognostic effect on post-metastatic survival. Survival was estimated using Kaplan–Meier survival curves and the log-rank test for univariate analysis. To eliminate bias due to confounders among the variables [25], multivariate analysis was performed using the variables with p-values of <0.05 in univariate analysis. Multivariate analysis was performed using the Cox proportional hazards model. A p-value of less than 0.05 was considered significant. Statistical analyses were performed using the SPSS software (Version 21.0; IBM Co., Armonk, NY, United States of America (USA)). For the variables showing significance in multivariate analysis, recursive partitioning analysis (RPA) was done with free software (R version 3.0.1; rpart package version 3.1-53, http://cran.r-project.org/). RPA is a statistical method that groups patients into distinct cohorts based on maximising the value of log-rank tests for the clinical end-point of interest [26]. The first two cohorts are defined by assessing all possible dichotomisations

Fig. 1. Kaplan–Meier curve of post-metastatic survival. The median survival for all patients was 20 months. The 2-year and 3-year survival rates were 47.1% and 34.4% respectively.

1652

S. Kang et al. / European Journal of Cancer 50 (2014) 1649–1656

Table 1 Univariate analysis of prognostic factors for post-metastasis survival. Characteristics (n)

Number (%)

Median overall survival (months)

Demographics Gender (135) Male Female

85 (63.0%) 50 (37.0%)

26 ± 4.3 13 ± 3.3

Age (135) <50 years P50 years

67 (49.6%) 68 (50.4%)

20 ± 5.8 18 ± 5.2

Primary tumour Anatomical site (135) Upper extremity Lower extremity

37 (27.4%) 98 (72.6%)

26 ± 4.2 17 ± 3.2

Histological diagnosis (135) UPS Others

34 (25.2%) 101 (74.8%)

11 ± 1.3 26 ± 7.5

Histological grade (FNCLCC) (130) Grade 1 or 2 Grade 3 Tumour size (125) 65 cm >5 cm

0.289

0.392

0.345

0.030a

<0.001a 60 (46.2%) 70 (53.8%)

40 ± 13.9 13 ± 3.7

32 (25.6%) 93 (74.4%)

40 ± 13.7 18 ± 2.8

Tumour depth (128) Superficial Deep

9 (7.0%) 119 (93.0%)

34 ± 8.6 19 ± 4.2

Pattern of metastasis Initial metastatic site (135)b Lung Bone Other sites

97 (71.9%) 19 (14.1%) 23 (17.0%)

26 ± 4.4 12 13 ± 2.1

Number of initial metastatic sites (135) Multiple Single

4 (3.0%) 131 (97.0%)

14 ± 18.5 20 ± 3.4

0.030a

0.442

0.088 0.827 0.075 0.954

Number of lung metastasis in patients with lung metastasis only (n = 64) 65 41 (64.1%) >5 23 (35.9%)

23 ± 5.7 26 ± 9.2

Disease course Local recurrence before metastasis (135) Present Absent

24 (17.8%) 111 (82.2%)

10 ± 2.6 26 ± 4.0

Metastasis-free interval (145) MFI < 12 months MFI P 12 months or synchronus group

32 (23.7%) 103 (76.3%)

11 ± 2.4 26 ± 4.3

Disease-free interval (135) DFI < 12 months DFI P 12 months or synchronus group

42 (31.1%) 93 (69.9%)

11 ± 2.1 26 ± 5.3

Treatment of metastasis Metastasectomy (135) Done Not done

44 (32.6%) 91 (67.4%)

40 ± 5.5 13 ± 2.7

91 (67.4%) 44 (32.6%)

26 ± 6.0 13 ± 4.4

75 (82.4%) 16 (17.6%)

26 ± 4.0 33 ± 1.3

Chemotherapy for metastasis (135) Done Not done Chemotherapy regimen – after metastasis (91) Ifosfamide Containing Not containing

p

0.709

0.062

0.051

0.002a

<0.001a

0.058

0.771

S. Kang et al. / European Journal of Cancer 50 (2014) 1649–1656

1653

Table 1 (continued ) Characteristics (n) Doxorubicin Containing Not containing Combination ifosfamide and doxorubicin Yes No

Number (%)

Median overall survival (months)

51 (56.0%) 40 (44.0%)

26 ± 4.3 23 ± 9.9

39 (42.9%) 52 (57.1%)

26 ± 4.6 32 ± 8.8

p 0.969

0.924

UPS, undifferentiated pleomorphic sarcoma; FNCLCC, Federation Nationale des Centres de Lutte Contre le Cancer; MFI, metastasis-free interval; DFI, disease-free interval. a p-Value < 0.05. b Duplicated in some cases.

of all predictor variables to find the one dichotomisation that produces the largest log-rank test statistic. The method then repeats this assessment within each of these two cohorts so that one of these two cohorts is further split into two smaller subgroups. In this study, a recursive decision tree was made by the dichotomisation proceeding with the split criteria of p < 0.05 in the logrank test. Then, the final nodes were grouped according to their median survival time, and the Kaplan–Meier graphs with log-rank tests are presented as the final set of prognostic groups. 3. Results 3.1. Predictors of post-metastatic survival Among the 135 patients, 76 (56%) died during the follow-up. The median post-metastatic survival of whole patients was 20 ± 3.7 months. The 2-year and 3-year survival rates were 47.1 ± 4.8% and 34.4 ± 5.2% respectively (Fig. 1). Univariate analysis of various factors regarding patient demographics, primary tumour, pattern of metastasis and treatment of metastasis is shown in Table 1. In multivariate analysis using the Cox proportional hazards model, four factors remained significant (Table 2): histological diagnosis of UPS (p = 0.002), high histological grade (p = 0.006), no carrying-out of metastasectomy (p = 0.009) and DFI < 12 months (p = 0.007). 3.2. Recursive partitioning analysis With these independent prognostic factors, a recursive decision tree was created producing six terminal nodes (Fig. 2). RPA identified six terminal nodes based on histological grade, performance of metastasectomy and DFI. Based on the median survival time of the terminal nodes, four prognostic groups with significantly different post-metastatic survival were generated: (1) group A: low grade/metastasectomy; (2) group B: low grade/no metastasectomy/DFI P 12 months or high grade/metastasectomy; (3) group C: low grade/no

Table 2 Multivariate analysis of prognostic factors for post-metastasis survival. HR (95% CI) 2.235 (1.336–3.739) 1 (Ref.)

Histological grade (FNCLCC) Grade 3 Grade 1 or 2

2.039 (1.225–3.394) 1 (Ref.)

Tumour size >5 cm 65 cm

1.154 (0.597–2.231) 1 (Ref.)

Metastasectomy Done Not done

0.448 (0.245–0.819) 1 (Ref.)

Disease-free interval DFI < 12 months DFI P 12 months or synchronus group

p 0.002a

Histological diagnosis UPS Others

0.006a

0.670

0.009a

0.007a 1.983 (1.201–3.273) 1 (Ref.)

HR, hazard ratio; CI, confidence interval; UPS, undifferentiated pleomorphic sarcoma; FNCLCC, Federation Nationale des Centres de Lutte Contre le Cancer; DFI, disease-free interval. a p-Value < 0.05.

metastasectomy/DFI < 12 months or high grade/no metastasectomy/DFI P 12 months; (4) group D: high grade/no metastasectomy/DFI < 12 months. The 3-year survival rates for each group were: group A, 76.1 ± 9.6%; group B, 42.3 ± 10.3%; group C, 18.8 ± 8.0%; and group D, 0.0 ± 0.0%. This grouping led to a significant separation of the Kaplan–Meier survival curves among the four groups (Fig. 3).

4. Discussion This study sought to generate a prognostic model for STS patients with metastases using clinicopathological variables. We estimated the distribution of post-metastatic survival, performed multivariate analyses to assess prognostic factors for post-metastatic survival, and finally performed RPA to generate a prognostic model for post-metastatic survival in a defined cohort of STS

1654

S. Kang et al. / European Journal of Cancer 50 (2014) 1649–1656

Fig. 2. Recursive partitioning analysis tress. Recursive decision tree was created with the independent prognostic factors, and produced six terminal nodes. These terminal nodes were categorised into four prognostic groups according to the median survival time: group A, low grade/ metastasectomy (node 1); group B, high grade/metastasectomy (node 2) or low grade/no metastasectomy/disease-free interval (DFI) P 12 months (node 3); group C, low grade/no metastasectomy/DFI < 12 months (node 4) or high grade/no metastasectomy/DFI P 12 months (node 5) and group D, high grade/no metastasectomy/DFI < 12 months (node 6). NA indicates not applicable, as more than 50% were still alive at the end of the study period. DFI, disease-free interval.

Fig. 3. Kaplan–Meier curves of four prognostic groups derived from recursive partitioning analysis. Log-rank test revealed significant differences among the four groups.

patients. Few prognostic models have been reported regarding metastatic STS in the literature. To our knowledge, this is the first study to utilise RPA and group the patients with metastatic STSs. Accurate estimates of the likelihood of survival are essential for patient counselling and informed decision making in metastatic cancer [27]. Several prognostic models based on statistical tools, such as nomograms, artificial neural networks and RPA, have been utilised in several cancer types. RPA has the capacity to account for complex relationships, and provides a simple and

intuitive method for classifying patients in a clinically useful form [17]. In RPA, there is progressive splitting of the population into groups based on the independent prognostic variables. RPA allows the clinician to simply follow the paths of the tree that best describe the characteristics of the patient and arrive at the prediction of the outcome for the particular patient [17,18,28]. Our prognostic model identified three factors for prognostic grouping: histological grade, performance of metastasectomy and DFI. The prognostic significance of these factors is well documented in previous studies [11,12,15,29–31]. This prognostic model based on purely clinical factors may serve as a backbone for the future incorporation of molecular prognostic markers. A prognostic model for metastatic STS would be more useful if only the factors that can be assessed at the time of diagnosis of metastasis are included. However, our model included the performance of metastasectomy in our model. In line with previous studies, patients with resectable diseases survived longer than those with unresectable diseases [15]. The survival advantage of resectable diseases may, in part, reflect the difference in tumour biology apart from the benefit of the resection itself [15]. Local recurrence rate of 38% in this study seems relatively high compared to the rates of general STS. Less aggressive treatment of the primary tumours with known metastatic disease, as reflected by the 29% positive surgical margin in the synchronous group, might have contributed to this finding. In addition, the aggressiveness of the tumour itself might also have played a role, as previously reported [32].

S. Kang et al. / European Journal of Cancer 50 (2014) 1649–1656

There are some considerations when interpreting the results of this study. First, only the patients with metastatic STS who had the primary tumours resected were included in this study. As patients with advanced STS or poor performance status not suitable for resection of the primary tumour are not included in the study, the results of this study may not be applicable to the whole population of metastatic STS. In line with this observation, the age of the patients was relatively younger than previous studies [12]. Second, our model was generated based on all histological subtypes of STS. As current treatment strategies for STSs are increasingly adapted to a specific histological subtype, a prognostic model for specific subtypes of STS will be needed. Third, treatment for STS metastases may not have been standardised as this study was performed over a relatively long period of time. Of note, there was a trend towards better survival in the later periods of the study. Fourth, validation of our model in independent data sets is needed. In conclusion, we have developed a RPA-derived prognostic model for extremity STS patients with metastatic disease. Histological grade, performance of metastasectomy and DFI were the significant factors for postmetastasis survival. This model can be implemented easily in the clinical practice and may provide a backbone for future prognostic models of post-metastasis survival.

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

Role of the funding source There is no funding source. Conflict of interest statement None declared.

[17]

[18]

[19]

References [20] [1] Lewis JJ, Brennan MF. Soft tissue sarcomas. Curr Probl Surg 1996;33(10):817–72. [2] Komdeur R, Hoekstra HJ, van den Berg E, et al. Metastasis in soft tissue sarcomas: prognostic criteria and treatment perspectives. Cancer Metastasis Rev 2002;21(2):167–83. [3] Jemal A, Thomas A, Murray T, Thun M. Cancer statistics, 2002. CA Cancer J Clin 2002;52(1):23–47. [4] Gadd MA, Casper ES, Woodruff JM, McCormack PM, Brennan MF. Development and treatment of pulmonary metastases in adult patients with extremity soft tissue sarcoma. Ann Surg 1993;218(6):705–12. [5] Coindre JM, Terrier P, Guillou L, et al. Predictive value of grade for metastasis development in the main histologic types of adult soft tissue sarcomas: a study of 1240 patients from the French Federation of Cancer Centers Sarcoma Group. Cancer 2001;91(10):1914–26. [6] Weingrad DN, Rosenberg SA. Early lymphatic spread of osteogenic and soft-tissue sarcomas. Surgery 1978;84(2):231–40. [7] Lawrence Jr W, Hays DM, Heyn R, et al. Lymphatic metastases with childhood rhabdomyosarcoma. A report from the Intergroup Rhabdomyosarcoma Study. Cancer 1987;60(4):910–5. [8] Bauer HC, Trovik CS, Alvegard TA, et al. Monitoring referral and treatment in soft tissue sarcoma: study based on 1,851

[21]

[22]

[23]

[24] [25]

[26]

1655

patients from the Scandinavian Sarcoma Group Register. Acta Orthop Scand 2001;72(2):150–9. Pisters PW, Leung DH, Woodruff J, Shi W, Brennan MF. Analysis of prognostic factors in 1,041 patients with localized soft tissue sarcomas of the extremities. J Clin Oncol 1996;14(5):1679–89. Stojadinovic A, Leung DH, Hoos A, Jaques DP, Lewis JJ, Brennan MF. Analysis of the prognostic significance of microscopic margins in 2,084 localized primary adult soft tissue sarcomas. Ann Surg 2002;235(3):424–34. Van Glabbeke M, van Oosterom AT, Oosterhuis JW, et al. Prognostic factors for the outcome of chemotherapy in advanced soft tissue sarcoma: an analysis of 2,185 patients treated with anthracycline-containing first-line regimens – a European Organization for Research and Treatment of Cancer Soft Tissue and Bone Sarcoma Group Study. J Clin Oncol 1999;17(1):150–7. Italiano A, Mathoulin-Pelissier S, Cesne AL, et al. Trends in survival for patients with metastatic soft-tissue sarcoma. Cancer 2011;117(5):1049–54. Pisters PW, Harrison LB, Leung DH, Woodruff JM, Casper ES, Brennan MF. Long-term results of a prospective randomized trial of adjuvant brachytherapy in soft tissue sarcoma. J Clin Oncol 1996;14(3):859–68. van Geel AN, van Coevorden F, Blankensteijn JD, et al. Surgical treatment of pulmonary metastases from soft tissue sarcomas: a retrospective study in The Netherlands. J Surg Oncol 1994;56(3):172–7. Billingsley KG, Burt ME, Jara E, et al. Pulmonary metastases from soft tissue sarcoma: analysis of patterns of diseases and postmetastasis survival. Ann Surg 1999;229(5):602–10 [discussion 610-2]. Casson AG, Putnam JB, Natarajan G, Johnston DA, Mountain C, McMurtrey M, et al. Five-year survival after pulmonary metastasectomy for adult soft tissue sarcoma. Cancer 1992;69(3):662–8. Cook EF, Goldman L. Empiric comparison of multivariate analytic techniques: advantages and disadvantages of recursive partitioning analysis. J Chronic Dis 1984;37(9–10):721–31. Lee JW, Um SH, Lee JB, Mun J, Cho H. Scoring and staging systems using cox linear regression modeling and recursive partitioning. Methods Inf Med 2006;45(1):37–43. Riad S, Griffin AM, Liberman B, Blackstein ME, Catton CN, Kandel RA, et al. Lymph node metastasis in soft tissue sarcoma in an extremity. Clin Orthop Relat Res 2004;426:129–34. Al-Refaie WB, Andtbacka RH, Ensor J, Pisters PW, Ellis TL, Shrout A, et al. Lymphadenectomy for isolated lymph node metastasis from extremity soft-tissue sarcomas. Cancer 2008;112(8):1821–6. Trojani M, Contesso G, Coindre JM, et al. Soft-tissue sarcomas of adults; study of pathological prognostic variables and definition of a histopathological grading system. Int J Cancer 1984;33(1):37–42. Coindre JM, Trojani M, Contesso G, et al. Reproducibility of a histopathologic grading system for adult soft tissue sarcoma. Cancer 1986;58(2):306–9. Bedi M, King DM, Charlson J, Whitfield R, Hackbarth DA, Zambrano EV, et al. Multimodality management of metastatic patients with soft tissue sarcomas may prolong survival. Am J Clin Oncol 2012. Available from: http://journals.lww. com/amjclinicaloncology/Abstract/publishahead/Multimodality_ Management_of_Metastatic_Patients.99460.aspx. Clark MA, Fisher C, Judson I, Thomas JM. Soft-tissue sarcomas in adults. N Engl J Med 2005;353(7):701–11. Klausner JD, Pollack LM, Wong W, Katz MH. Same-sex domestic partnerships and lower-risk behaviors for STDs, including HIV infection. J Homosex 2006;51(4):137–44. Zhan H. Recursive partitioning and Tree-based methods. Papers/ Humboldt-Universitat Berlin, Center for Applied Statistics and

1656

S. Kang et al. / European Journal of Cancer 50 (2014) 1649–1656

Economics (CASE), No. 2004,30. Available from: http://hdl.handle.net/10419/22203 [accessed Jan 19, 2014]. [27] Miles BJ, Giesler B, Kattan MW. Recall and attitudes in patients with prostate cancer. Urology 1999;53(1):169–74. [28] Shariat SF, Karakiewicz PI, Suardi N, Kattan MW. Comparison of nomograms with other methods for predicting outcomes in prostate cancer: a critical analysis of the literature. Clin Cancer Res 2008;14(14):4400–7. [29] Billingsley KG, Lewis JJ, Leung DH, Casper ES, Woodruff JM, Brennan MF. Multifactorial analysis of the survival of patients with distant metastasis arising from primary extremity sarcoma. Cancer 1999;85(2):389–95.

[30] Karavasilis V, Seddon BM, Ashley S, Al-Muderis O, Fisher C, Judson I. Significant clinical benefit of first-line palliative chemotherapy in advanced soft-tissue sarcoma: retrospective analysis and identification of prognostic factors in 488 patients. Cancer 2008;112(7):1585–91. [31] Jablons D, Steinberg SM, Roth J, Pittaluga S, Rosenberg SA, Pass HI. Metastasectomy for soft tissue sarcoma. Further evidence for efficacy and prognostic indicators. J Thorac Cardiovasc Surg 1989;97(5):695–705. [32] Choong PF, Gustafson P, Rydholm A. Size and timing of local recurrence predicts metastasis in soft tissue sarcoma. Growth rate index retrospectively analyzed in 134 patients. Acta Orthop Scand 1995;66(2):147–52.