The heterogeneity of STS has rarely been taken into account in the design of clinical trials to investigate systemic therapies in STS patients. Our results indicated that LMS clearly represented a distinct STS subgroup with a significantly better outcome in the advanced setting. Previous studies have shown worse outcomes for LMS than the results obtained in our current analysis. The largest study published to date was a retrospective analysis of 2185 patients with advanced STS treated in the first-line studies of EORTC-STBSG; these patients showed no significant differences in terms of OS between LMS (492 cases) and the other histological subtypes, with a median OS of approximately 12 months [6]. However, this study, which focused only on first-line treatment, included patients diagnosed before the identification of the KIT mutation in gastrointestinal stromal tumors [4]. Therefore, a significant proportion of gastrointestinal stromal tumors, which are chemorefractory, were likely included in the LMS group. The better outcome of LMS may be explained by a specific biology but also by the potentially higher sensitivity to some anti-cancer agents such as gemcitabine, dacarbazine, or trabectedin. For instance, in a recent phase II randomized trial, patients with leiomyosarcomas of any origin benefited significantly from the combination of gemcitabine with dacarbazine, achieving a median progression-free survival (PFS) and OS of 4.9 and 13.8 months, respectively, versus 2.1 and 7.8 months, respectively, for the non-leiomyosarcoma subtypes [7]. Moreover, a large worldwide expanded access program for trabectedin showed a median OS of 16.2 months in 321 heavily pre-treated leiomyosarcoma patients versus a median survival time of 11.9 months for the whole cohort of 903 patients [8].
We report here the first study assessing the outcomes of patients with advanced UPS. Some past reports included patients with malignant fibrous histiocytomas (MFHs). However, a significant subset of tumors initially diagnosed as MFH showed a specific line of differentiation (lipogenic, neurogenic, myogenic, or non-sarcomatous) [9–12]. “MFH” is now considered an obsolete terminology and has been replaced by the term UPS, which is a diagnosis of exclusion. We found that patients with advanced UPS had the worst outcome with the shortest TNT and a median OS of only 11 months. These results illustrate the particular resistance to chemotherapy of this histological subset and an intrinsically more aggressive biology. Further investigations are needed to better understand the mechanisms of their tumorigenesis and to define more appropriate therapeutic strategies.
Approximately 45% of the 1575 patients who underwent systemic therapy received a combination chemotherapy regimen in the first-line setting. The first-line chemotherapy for advanced, metastatic, or non-resectable STS is typically based on single-agent doxorubicin [13]. Indeed, the majority of clinical studies comparing single agents with combinations failed to show an OS advantage but consistently showed improvement in the response rates and PFS [14, 15]. Interestingly, our analysis showed a significant impact of the use of combination chemotherapy on OS, with a hazard ratio of 0.822 (0.724–0.932) and P = 0.0003. Judson et al. [14] recently published the results of a randomized clinical trial evaluating doxorubicin as a single agent in the control arm versus doxorubicin-ifosfamide in the experimental arm as a first-line treatment for advanced or metastatic STS. Although the Kaplan–Meier curves presented in the publication highlighted a difference between the two treatment arms in favor of polychemotherapy, the trial failed to detect a significant effect of polychemotherapy on OS, which was in contrast to our results. Our results suggest that the negative outcome of this study may simply be due to a lack of power as already suggested by Benjamin and Lee [16]. Indeed, by including 450 patients and observing at least 366 events, the trial was designed to detect a maximum HR of 0.737. Due to the large size of our dataset, we were able to observe an HR of 0.822. Based on their hypotheses, a total of 827 events would be required to detect a similar treatment effect in a randomized clinical trial. Although our study suggests a benefit in terms of OS, clinicians should also be aware that randomized trials have clearly demonstrated that combination chemotherapy is more toxic than single-agent doxorubicin with a potential significant impact on the quality of life [14, 15]. Therefore, a combination of doxorubicin with a second drug such as ifosfamide should be used only after a careful discussion with the patient on the benefit/risk ratio of this approach, particularly when tumor shrinkage is expected to improve the symptoms or clinical benefits.
A high proportion of patients received more than two lines of systemic treatment. With the exception of leiomyosarcomas, our results indicate that the benefit of a greater than third-line regimen is very limited, with the median TNT and OS ranging between 2.3 and 3.7 months and 5.4 and 8.5 months, respectively. This result is consistent with the data from the PALETTE study, which led to the approval of pazopanib in advanced STS [17]. In that study, the number of previous lines of chemotherapy was a significant prognostic factor in the multivariate analysis for PFS with a significantly worse outcome in patients receiving pazopanib in the third- or fourth-line settings versus the first- or second-line settings. Given the potential toxicity and the moderate benefit of systemic therapy after failure of the second-line treatment, best supportive care should be considered as a reasonable option, particularly in patients with non-leiomyosarcoma histology and a poor performance status or patients who were not eligible to participate in a clinical trial. Notably, 50% of patients received an off-label drug during their treatment disease course. This result reflects the increasing evidence for the use of other drugs besides doxorubicin and ifosfamide in the sarcoma field. The most frequently prescribed off-label drug in this study was gemcitabine. Indeed, gemcitabine with or without docetaxel is commonly used in some specific sarcoma subsets, particularly in leiomyosarcomas and angiosarcomas [18–21], although neither of these drugs is approved for this indication. Another not yet approved drug that is frequently used in the sarcoma field is paclitaxel, which shows activity particularly in angiosarcomas [22, 23].
A significant proportion of patients with metastatic STS (27%) did not receive any systemic therapy. An age > 75 years was significantly associated with a lower probability of receiving any systemic treatment. Aging is associated with progressive functional declines, an increased prevalence of comorbidities, and a higher risk of cardiac and hematological toxicities related to anthracyclines [24–26]. These data may explain the reluctance of oncologists to use chemotherapy in elderly patients with STS and raises the question of the development of adapted chemotherapy regimens for elderly patients with advanced STS, such as low-dose cyclophosphamide [27] or liposomal doxorubicin [28].
A total of 49% of the patients received a loco-regional treatment of the metastasis, the most frequent of which were surgery followed by radiotherapy and radiofrequency ablation. The majority of these patients (71%) had lung metastases. The published evidence on the role of locoregional treatments, such as pulmonary metastasectomy, is derived from a small number of studies with limited sample sizes [29]. Primary bone sarcomas, which may represent a distinct disease, are often included in these analyses. Our present study differed from previous publications because we used a larger database cohort, which increased the power of the multivariate analysis; additionally, we focused on STS exclusively to enhance the homogeneity of the study population. As suggested by previous studies, patients who underwent a locoregional metastasis treatment had improved survival in the multivariate analysis. Arguments have suggested that an observational study may not provide evidence that a difference in survival is attributable to the locoregional treatment and that only a randomized trial can answer the question. However, we observed that more than 80% of metastatic patients alive 5 years after the diagnosis of metastasis had received a locoregional treatment, versus 50% in the general population, and this parameter was most significantly associated with the probability of being alive at 5 years in the logistic regression model. Precisely, the descriptive analyses of the patients alive after 5 years suggest that surgery, radiofrequency, and a combination of different modalities are particularly beneficial in terms of survival. This hypothesis was confirmed by our sensibility analysis, since we found that the positive effect on the probability of 5-year survival was significant for these three treatment modalities only.
No data are available from randomized clinical trials to define how best to integrate the locoregional treatment of metastases in the management of patients with advanced disease. The most recent attempts were made by the European Organisation for Research and Treatment of Cancer (EORTC-Protocol 62933) with a randomized multicenter trial to assess metastasectomy alone versus induction chemotherapy followed by metastasectomy in a targeted sample size of 340 patients. Started in 1996, this trial was closed due to poor accrual in November of 2000. Notably, we report here the first large series of patients who received non-surgical locoregional treatment of metastases, including 254 patients treated with radiotherapy, 42 with radiofrequency ablation, and 320 with a combination of surgery plus radiotherapy or surgery plus radiofrequency ablation of metastases.
The gold standard endpoint in randomized clinical trials in oncology is OS. However, the use of a surrogate endpoint at an earlier stage in clinical trials would speed up the assessment of treatments and might reduce the cost of drug development. Studies that assess the use of alternative outcome measures, such as the response rate or PFS, as surrogate endpoints for OS in sarcoma patients showed only a modest if any correlation with PFS and OS [30, 31]. This issue was recently illustrated with the pivotal trial that led to eribulin approval in patients with liposarcomas that showed a benefit in OS but not in PFS [32]. TNT is an established endpoint that is mostly applied in hematological malignancies and has recently been used in breast, colon, and prostate cancer [33–35]. The use of this parameter is predicated on the concept that a change in treatment usually occurs in response to a real change in the patient status by integrating the efficacy and toxicity components. In our study, we found a strong correlation between TNT and OS. The prospective validation of this endpoint as a surrogate for OS should be done in future studies.