Brain Metastases and Place of Antiangiogenic Therapies in Alveolar Soft Part Sarcoma: A Retrospective Analysis of the French Sarcoma Group

The effect of anti‐angiogenic therapies in alveolar soft part sarcoma patients with metastatic disease remains to be clarified. This article analyzes clinical factors associated with patient outcomes with a focus on the natural history of brain metastasis and patient response to anti‐angiogenic therapies.

Abstract

Background.

Alveolar soft part sarcoma (ASPS) is a rare sarcoma characterized by a slow evolution, brain metastasis (BM), and resistance to doxorubicin. Antiangiogenic therapies (AAT) have shown clinical activity, but little is known about the optimal therapeutic strategy, specifically considering BM.

Subjects, Materials, and Methods.

We performed a retrospective analysis of all patients with ASPS treated in three referral centers of the French Sarcoma Group. We aimed to describe factors associated with overall survival (OS) and the impact of BM on outcome of patients treated by AAT.

Results.

We identified 75 patients between 1971 and 2012 (median age = 23, range: 5–96 years). Median follow‐up was 74 months. Patients with localized (n = 44, 59%) and metastatic (n = 31, 41%) diseases had a 10‐year OS of 69% and 25%, respectively. Only surgical incomplete resection was associated with shorter OS in localized disease (hazard ratio [HR] = 5.2, 95% confidence interval [CI] 1.2–22.4, p = .02). Fifty‐two (69%) patients developed lung metastasis (LM; baseline: n = 31, [41%]; de novo: n = 21, [28%]). Thirteen patients developed BM, all occurring after LM. Tumor size ≥5 cm was associated with poorer BM‐free survival (HR = 8.4, 95% CI 2.1–33.9, p = .002). Median OS post‐BM was 17 months (95% CI 15 to not assessable). Overall, 12 patients were treated with AAT (sunitinib n = 10): 5 patients had BM and achieved poor outcomes compared with patients without, with median progression‐free‐survivals of 2 versus 11 months, respectively.

Conclusion.

Baseline larger tumors were associated with increased risk of brain metastasis in patients with ASPS. Patients with BM seem to have little benefit from AAT, suggesting the need to develop antineoplastic agents with high central nervous system penetrance in this setting.

Implications for Practice.

Alveolar soft part sarcoma (ASPS) is an extremely rare subtype of sarcoma that is particularly resistant to conventional therapies. Antiangiogenic therapies (AAT) have shown promising results. However, patients with ASPS still die of tumor evolution. This study highlights the prognostic shift induced by brain metastasis (BM), identifying this event as a major contributor to the death of patients with ASPS, and observes a striking lack of effectiveness of AAT in patients who had previously developed BM. This observation is of interest for the therapeutic development in ASPS, highlighting the need to develop strategies dedicated to BM, such as radiosurgery or high‐central nervous system penetrance tyrosine kinase inhibitors.

Introduction

Alveolar soft part sarcoma (ASPS) is a rare tumor subtype in the sarcoma field, accounting for 0.3% of all soft tissue sarcoma [1]. Since its first description in 1952 [2], a handful of reports depicted the natural history of the disease, through retrospective analyses [3], [4], [5]. Thus, ASPS is currently thought to represent a sarcoma subtype with an indolent evolution and primary resistance to chemotherapy (anthracycline‐based). In addition, brain metastasis has also been described in the natural course of the disease [6]. In the last decade, an improvement in the management of patients with metastatic disease was observed, possibly in part because of antiangiogenic therapies (AAT). These agents showed efficacy in different reported phase II trials and more importantly in a randomized phase III (PALETTE [7]) and a randomized phase II study (REGOSARC [8]). Notably, patients’ overall response rate (ORR) ranged from 35% for cediranib [9], 28%–55% for sunitinib [10], [11], and 20% for pazopanib [12]. Thus, AAT appear among the most efficient agents in this setting.

However, regarding the indolent evolution of the disease, few studies with sufficiently long‐term follow‐up have reported patterns of metastasis in patients with ASPS. Indeed, in the largest report from the Surveillance, Epidemiology, and End Results database [5], which described baseline factors associated with overall survival (OS), the prognostic impact of resection margins was unknown; likewise, little was known about the long‐term evolution of the disease, where relapses have been reported decades after initial treatment [13], [14]. More particularly, there are a lack of data regarding factors associated with brain metastasis‐free survival (BMFS). Finally, to our knowledge, apart from a recent study reporting the therapeutic effect of pazopanib with a median progression‐free survival (PFS) on therapy of 13.6 months [12], long‐term follow‐up of patients was not reported in phase II studies. Overall, the exact impact of AAT in ASPS patients with metastatic disease remains to be further clarified [5].

In this study, we aimed to analyze clinical factors associated with patients’ outcome with a focus on the natural history of brain metastasis and patients’ response to AAT.

Subjects, Materials, and Methods

Patients

This multicenter retrospective cohort included all patients with localized or metastatic ASPS, treated and followed in three referral centers of the French Sarcoma Group (Gustave Roussy Cancer Campus, Villejuif; Bergonie Institute, Bordeaux; Leon Berard Cancer Center, Lyon). All patients registered in the three databases were screened. Follow‐up was considered from the date of diagnosis to the date of death or date of last assessment (last visit before the date of collection of data). All patients were followed according to standard of care in Medical Oncology in France.

ASPS Diagnostic

All diagnostics were confirmed by centralized histological evaluation, in the French Sarcoma Cooperative Group (GSF) setting. ASPL‐TFE3 translocation was confirmed for each sample available using TFE3 staining by immunochemistry.

Collection of Data

For Gustave Roussy Cancer Campus patients, data from patients’ medical charts were retrospectively reviewed and collected. For Bergonie Institute and Leon Berard Institute, prospectively actualized databases were provided (J‐M.C., J‐Y.B.) at the time of data collection on January 2013.

Data collected were as follows: patient's baseline characteristics (date of birth, gender, personal history of cancer or radiotherapy, familial history of cancer, Eastern Cooperative Oncology Group Performance Status), clinical presentation (mass tumor syndrome/pain/limb dysfunction/trauma/metastasis, date of first reported symptom, date of histologically confirmed diagnosis, primary tumor localization, metastatic sites, and primary tumor size), first therapeutic strategy (surgery, and, if applicable, resection margins: positive [R1] or negative [R0]; antineoplastic drugs administrated; radiotherapy), and, if applicable, tumor response, as stated in the medical records (progression or stable disease or tumor response). If RECIST 1.1 response was available in the medical record, these data were collected. Follow‐up data collected were as follows: date of relapse (for patient with localized disease at baseline) and consecutive progressions (for patients with metastatic disease), and date of death or date of last assessment before data collection. De novo lung metastasis refers to metastasis diagnosed during the course of the disease, as opposed to baseline lung metastasis, which refers to metastasis diagnosed at the baseline radiological evaluation. Details regarding local treatments applied for specific site of relapse, such as brain metastasis, could not be systematically recorded because of the extended interval of diagnosis, and the heterogeneity of follow‐up.

OS and BMFS were estimated from the date of diagnosis to the date of the considered event. Events of interest for OS and BMFS included the following: death (regardless of the cause) and first radiological documentation of brain metastasis, respectively [15] (censored by date of death or last assessment). PFS under AAT was estimated from the date of AAT initiation to the date of progression or death (censored by date of last assessment).

Baseline continuous variables were initial primary tumor size and age. Tumor size was managed in binary analysis (T1: tumor ≤50 mm; T2: tumor >50 mm; according to the American Joint Committee on Cancer [AJCC] 2010 classification) and in quantitative analyses.

The data collection was in accordance with French recommendations regarding ethics and protection of patients’ personal data in the setting of a noninterventional retrospective study and approved by each Institutional board in the three centers. Anonymized data were registered on a GSF‐Groupe d'Etude des Tumeurs Osseuses (GETO) scientific group database, approved by the French “National Committee of data processing for data protection.”

Objectives

Our objectives were to describe initial characteristics that correlated with a poorer OS and to define the frequency and significance of brain metastasis and AAT in the course of the disease.

Statistical Analysis

Follow‐up was estimated with the reverse Kaplan‐Meier method, with absolute range and interquartile range. Survivals were estimated using the Kaplan‐Meier method, described using 2‐, 5‐, or 10‐year survival, depending on the clinically relevant threshold. Survival curves analyses used two‐sided log‐rank test (p < .05). Associations between patients/tumor baseline characteristics and survival outcomes were assessed in univariable and, when applicable, multivariable analyses, with Cox regression model, stratified on patients’ center (Gustave Roussy Cancer Campus vs. others). Variables were included in multivariable models if significant in univariable analysis (p < .05). Proportional hazards assumption was tested for each analysis. Patients with missing data for a variable were not included in the analysis of the considered variable or in the multivariable model. Statistical analyses were performed using the R software (v 3.3.3).

Results

Patient Characteristics

Eighty patients with ASPS were identified between 1971 and 2012. Among those, five patients consulting for a second medical opinion had no available follow‐up data and were therefore excluded from further analyses. Thus, the data of 75 patients with available clinico‐pathological data were analyzed in this report. The median duration of follow‐up was 74 months (range: 3–326 months, interquartile range: 33–192 months). At the date of last follow‐up, 22 (29%) patients had died of the disease.

Median age at diagnostic was 23 years (range: 5–96). We observed a normal distribution of age at diagnosis with a peak at 20 years, and a 2:3 sex ratio (29 males and 46 females; Table 1). At baseline, the 31 (41%) patients with metastatic disease had radiologically documented lung metastasis. In addition, four (12.9%) patients had associated bone (n = 3) or liver (n = 1) metastasis. None had brain metastasis at diagnosis.

Table 1. Patients’ baseline characteristics.

^aAccording to the American Joint Committee on Cancer v.2010 classification.

Abbreviation: ECOG PS, Eastern Cooperative Oncology Group Performance Status.

Initial Therapeutic Management

Surgical resection was the cornerstone of the therapeutic strategy, either in localized (n = 43, 98%) or in metastatic settings (n = 21, 68%; Table 2). Among patients with localized diseases, 29 (67%) achieved a complete resection (R0), and local radiotherapy was more commonly performed in patients with T2‐stage tumors (n = 10, 63%) versus T1‐stage tumors (n = 8, 30%; Chi‐square test: p = .035). There was no difference in local radiotherapy according to resection margins status (R0 vs. R1; Chi‐square test: p = .6).

Table 2. Initial therapeutic strategies.

^aAccording to the American Joint Committee on Cancer v.2010 classification.

Abbreviation: Pts, patients.

Twenty‐eight patients received front‐line chemotherapy, either in the neoadjuvant (n = 6) or in the metastatic settings (n = 22). Anthracycline was used in most regimens (n = 26, 93%). As a result, 2 (7%) patients achieved partial responses, 17 (61%) patients had a stable disease, and 8 (29%) patients experienced tumor progression.

Factors Associated with Overall Survival

We analyzed associations between patients’ baseline characteristics/treatments and OS. Presence of metastasis at diagnosis was the only factor significantly associated with a poorer OS (Fig. 1). Hence, the 10‐year OS for T1 M0, T2 M0, and metastatic diseases was 70% (SD = 10%), 65% (SD = 17%), and 25% (SD = 14%), respectively. Age and gender were not associated with OS. In patients with localized disease, only R1 resection margin was associated with a poorer OS (hazard ratio [HR] = 4.4, 95% confidence interval [CI] 1.0–19.0, p = .04; Table 3). The 10‐year OS in R1 and R0 patients was 40% (SD = 18%) and 84% (SD = 9%), respectively.

Figure 1.

Overall survival (OS) according to baseline tumor stage. Kaplan‐Meier curve. Note that median OS in patients with T1 M0 and T2 M0 was not reached as compared with median OS of 74 months in patients with metastatic diseases: 95% confidence interval 60 to not assessable. p value indicates survival curves comparison using log‐rank test.

Abbreviations: M1, metastasis; T1 and T2, T stage according to the American Joint Committee on Cancer v.2010 classification.

Table 3. Univariable analysis of clinical and pathological features associated with patients’ overall survival.

^aEstimated using Cox regression model, stratified on inclusion center (Gustave Roussy Cancer Campus vs. others).

^bAccording to the American Joint Committee on Cancer v.2010 classification.

Abbreviations: —, not applicable; 95% CI, 95% confidence interval; CI, confidence interval; HR, hazard ratio; OS, overall survival; SD, standard deviation.

Kinetics of Brain Metastasis Development

None of the patients had documented brain metastasis at diagnosis. Overall, 13 (17%) patients developed brain metastasis during the course of the disease (Fig. 2A). All of these latter patients had previously developed lung metastasis (n = 52, 69%—among whom 31 patients had baseline lung metastasis and 21 patients had de novo metastasis): median interval time between lung and brain metastasis was 35.4 months (range: 0–127 months; interquartile range: 17–48 months; Fig. 2B). In the whole cohort, 22 (29%) patients had died at the date of last assessment; among them, 8 (11%) had prior developed brain metastasis and the 14 (19%) others died without evidence of brain metastasis. The five (6%) last patients who developed brain metastasis were still alive at the date of last assessment (Fig. 3C).

Figure 2.

Brain metastasis in the course of the disease. (A): Brain metastasis‐free survival according to baseline tumor stage. Kaplan‐Meier curves. Median survival in patients with T1 tumors was not reached, compared with patients with T2 tumors: 112 months, 95% confidence interval (CI) 70.5 to not assessable (NA); and with patients with baseline lung metastasis: 127 months, 95% CI 92.8 to NA. p value indicates survival curves comparison using log‐rank test. (B): Brain metastasis‐free survival after lung metastasis diagnostic. Kaplan‐Meier curve with 95% CI (gray area). Patients considered are all patients who developed lung metastasis in the course of the disease (n = 52 patients), either at baseline (n = 31) or de novo (n = 21). Median: 127 months, 95% CI 92.8 to NA. (C): Overall survival after brain metastasis. Kaplan‐Meier curve with 95% CI. Patients included: 13 patients. Median: 16.6 months, 95% CI 14.8 to NA. (D): Correlation between initial tumor size and time to brain metastasis diagnostic. Patients included: 13 patients. R² and p value were estimated using a linear regression model with 95% CI (gray area).

Abbreviations: M1, metastasis; T1 and T2, T stage according to the American Joint Committee on Cancer v.2010 classification.

Figure 3.

Therapeutic effect of antiangiogenic therapies. (A): Progression‐free survival (PFS). Kaplan‐Meier curve with 95% confidence interval (CI; gray area). Median PFS: 9 months, 95% CI 2 to not assessable (NA). (B): RECIST 1.1 response to antiangiogenic therapies. *Patients with brain metastasis at initiation of AAT. RECIST 1.1 response was not available in two patients: one patient because of early neurologic event related to brain metastasis, and data missing in another patient. All patients were treated with sunitinib, except patients 5 investigational new drug (ING), and 6 (sorafenib). (C): Antiangiogenic therapy therapeutic interval related to the duration of metastatic disease. Patients 11 and 12 are patients without available data regarding the RECIST 1.1 response rate, and with brain metastasis at AAT initiation. (D): Progression‐free survival according to the presence of brain metastasis at antiangiogenic therapy initiation. Median PFS for patients with brain metastasis: 2 months, 95% CI 1.9 to NA. Median PFS for patients without brain metastasis: 11 months, 95% CI 9.0 to NA. Kaplan‐Meier curve comparison: p value indicates survival curves comparison using log‐rank test.

Abbreviation: AAT, antiangiogenic therapies.

We then explored baseline factors associated with poorer BMFS. In univariable analysis, only T2 tumors (T2 stage; HR = 8.4, 95% CI 2.1–33.9, p = .002), or larger tumors in continuous variable (HR = 1.03, 95% CI 1.01–1.04, p < .001), and lung metastasis at diagnosis (HR = 4.0, 95% CI 1.2–13.2, p = .02) were associated with poorer BMFS. In multivariable analysis (including tumor size and metastatic status—72 patients with available data), only larger tumor size remained independently associated with a poorer BMFS (HR = 1.03, 95% CI 1.0–1.05, p = .01; Fig. 2A). We then estimated median BMFS after lung metastasis diagnosis of 127 months (95% CI 93 to not assessable [NA]), with a 10‐year BMFS of 58% (SD = 10%; Fig. 2B). Median OS after brain metastasis was 17 months (95% CI 15 to NA), with a 2‐year OS of 26% (SD = 14%; Fig. 2C). Exploratory analysis suggests a linear chronological correlation between initial tumor size and time to brain metastasis (Fig. 2D).

Efficacy of Antiangiogenic Therapies in Patients with Brain Metastasis

Between 2001 and 2012, 12 (16%) patients were treated with AAT in a first‐ (n = 2), second‐ (n = 8), and third‐line setting (n = 2). All pretreated patients were exposed to anthracycline and ifosfamide except 2 (3%) patients (patient ID 2 and 10). Ten (13%) patients were treated with sunitinib, one with sorafenib, and one with a vascular endothelial growth factor tyrosine kinase inhibitor as an investigational drug. Patients experienced a median PFS of 9.0 months (95% CI 2.3 to NA) under AAT (Fig. 3A). Among the 10 patients who had RECIST 1.1 available data, only 1 patient had an objective response (Fig. 3B). Additionally, we noticed that the ratio of AAT therapeutic time interval relative to the duration of the metastatic phase did not exceed 9% (range: 1%–9%) for patients with brain metastasis but was up to 57% (range: 7%–57%; mean: 23%) in patients without (Fig. 3C). Following that, we investigated factors that may be associated with PFS under AAT (e.g., age, gender, and the presence of brain metastasis). Only brain metastasis was associated with poorer PFS (Fig. 3D): All patients with brain metastasis discontinued AAT because of progression or death within the first 3 months, whereas the others continued until 8.9–18.4 months (Fig. 3C, 3D). When focusing on patients with brain metastasis, two patients died of early brain metastasis progression (patient ID 11 and 12), one patient experienced a brain metastasis hemorrhagic progression (patient ID 10), one patient was still treated at the date of last assessment (patient ID 8), and one patient with a slow brain metastasis progression discontinued sunitinib because of toxicity and was monitored in a watchful waiting strategy (patient ID 9).

Discussion

We report a large retrospective cohort of ASPS patients with a long follow‐up and describe the kinetics of brain metastasis in the course of ASPS as well as patients’ response to AAT. Despite the indolent evolution of the disease, we highlighted herein that brain metastasis constitutes a major shift in the natural history of the disease with no efficacy of AAT in this setting.

We observed a chronological sequence in the metastasis development, in which lung metastasis seemed to be a preliminary required step before further dissemination. Moreover, we observed a significant correlation between initial larger tumor size and shorter BMFS suggesting that brain metastasis could be a nonstochastic time‐dependent event. Although patients without brain metastasis experienced an overall similar PFS under AAT as previously described in other cohorts [9], [10], [11], [12], it is striking here to note that efficacy of AAT was dismal in patients with brain metastasis. Objective partial response was observed in only one patient without brain metastasis, possibly due to a later use of AAT in the course of the disease.

Of note, median OS for patients with metastatic diseases at diagnosis was twice as long as previously reported: OS in the large series published in 1989 (inclusions: 1923–1986) [3], 2001 (inclusions: 1959–1998) [4], and 2016 (inclusions: 1973–2012) [5] was 36 months, 40 months, and 36 months, respectively, as compared with 74 months in our cohort. This could be related to the retrospective nature of the study or to the centralization of care for sarcomas in France in specialized referral cancer centers. For instance, in our population, the extensive use of surgery (68%) by trained teams could have influenced the good survival outcomes of our cohort. However, we could not rule out other survival‐influencing factors related to tumor heterogeneity. Recently, the centralization of care for sarcomas in France has been suggested to provide better outcomes [16]. The NETSARC database analysis of 12,528 patients confirmed this better outcome tendency [17].

The putative time‐dependent brain metastasis development suggests that patients should be carefully monitored for brain metastasis after occurrence of lung metastasis, in order to propose local treatments for brain metastasis. Indeed, stereotactic radiosurgery [18] has been reported as a potential local therapeutic strategy even for the treatment of tumors considered resistant to radiotherapy [19].

The reason for the absence of therapeutic effect in patients with brain metastasis is for now unsettled. This could be due to a switch in the tumor biology in the brain microenvironment, or to a poor diffusion of sunitinib in the central nervous system (CNS). The comparison of response rates between cediranib [9], [20], sunitinib [10], [11], and pazopanib [12], as well as with MET inhibitors such as tivantinib [21], appears complex because of both the heterogeneity of patients’ tumor characteristics and the timing of AAT introduction. Our observations, in line with recent studies [12], support the use of AAT as soon as possible in patients with evolutive metastatic diseases, before the development of brain metastasis. AAT with high CNS penetrance might therefore be evaluated in this indication.

Our analysis presents several limitations. Despite the long‐term follow‐up, our observations regarding OS were limited by the number of events related to the indolent nature of the disease. More particularly, our study was underpowered to explore the added value of surgery in patients with metastatic disease. Other limitations include the multicenter retrospective design of data collection over an extended period and the small number of patients treated with AAT. Furthermore, the important heterogeneity of follow‐up length could have also biased the observation of early events in patients with short follow‐up. However, this bias appeared limited when considering the external validity of our observations for survival analysis. Another major issue relies on the absence of systematic brain imaging on the entire period, and most particularly at baseline: Brain metastasis incidence may have been underestimated in patients without lung metastasis. Data regarding brain metastasis local treatments could not be accurately collected. Finally, brain metastasis occurence could have been censored by competing events leading to death, or by date of last assessment, because of follow‐up heterogeneity.

Conclusion

Despite these limitations, our data describe the kinetics of brain metastasis within the course of ASPS and suggest the need for brain screening in patients who develop lung metastasis. The limited efficacy observed for AAT in ASPS patients with documented brain metastasis suggests the need to assess agents with high CNS penetrance in this setting, or specific multimodal therapeutic strategies.

Author Contributions

Conception/design: Gabriel G. Malouf, Guillaume Beinse, Axel Le Cesne

Provision of study material or patients: Jean‐Michel Coindre, Jean‐Yves Blay, Axel Le Cesne

Collection and/or assembly of data: Gabriel G. Malouf, Guillaume Beinse

Data analysis and interpretation: Gabriel G. Malouf, Guillaume Beinse

Manuscript writing: Gabriel G. Malouf, Guillaume Beinse, Julien Adam, Olivier Mir, Ali N. Chamseddine, Philippe Terrier, Charles Honore, Jean‐Philippe Spano, Antoine Italiano, Jean‐Emmanuel Kurtz, Jean‐Michel Coindre, Jean‐Yves Blay, Axel Le Cesne

Final approval of manuscript: Gabriel G. Malouf, Guillaume Beinse, Julien Adam, Olivier Mir, Ali N. Chamseddine, Philippe Terrier, Charles Honore, Jean‐Philippe Spano, Antoine Italiano, Jean‐Emmanuel Kurtz, Jean‐Michel Coindre, Jean‐Yves Blay, Axel Le Cesne

Disclosures

Gabriel G. Malouf: Pfizer, Novartis, Bristol‐Meyers Squibb, Astellas (C/A), Pfizer, Novartis (RF); Guillaume Beinse: Novartis, Pharmamar (Other: travel expenses); Olivier Mir: Amgen, AstraZeneca, Bayer, Blueprint, Bristol‐Meyers Squibb, Eli Lilly & Co., Novartis, Pfizer, Roche, Servier (C/A), Eli Lilly & Co., Roche (H); Ali N. Chamseddine: Amgen (C/A), Eli Lilly & Co., Roche (Other: travel expenses); Jean‐Philippe Spano: Roche, MSD (C/A), BMS, Novartis, Pfizer, AZ, PFO, Gilead, Lilly, Leopharma, Myriads (H, SAB); Jean‐Emmanuel Kurtz: Tesaro, AstraZeneca (C/A), Roche, Pharmamar (Other: travel expenses); Jean‐Yves Blay: Novartis, Roche, GlaxoSmithKline, Bayer, Pfizer (RF); Axel Le Cesne: Amgen, Novartis, Pharmamar, Pzizer, Eli Lilly & Co. The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board