Childhood, Adolescent and Young Adult Poor-Prognosis Rhabdomyosarcoma

TL;DR

This review discusses poor-prognosis rhabdomyosarcoma in children and young adults, highlighting the need for better treatments and summarizing current and emerging approaches.

Contribution

The paper focuses on the most challenging subgroups of RMS patients and summarizes novel therapeutic strategies being explored.

Findings

Patients with PAX3(7)::FOXO1 gene fusions and other adverse features have the worst outcomes.

Adolescents and young adults with RMS face poorer prognoses compared to younger children.

Metastatic and relapsed RMS cases remain difficult to treat despite multimodal therapies.

Abstract

Rhabdomyosarcoma (RMS) is an aggressive soft tissue sarcoma which primarily affects children, adolescents and young adults. Multimodality approaches to treatment include chemotherapy, surgery and radiotherapy. Despite attempts to improve outcomes over recent decades, survival rates remain poor for patients with advanced and relapsed/progressive disease and those who have tumours with aggressive biologic features. This review article focuses on patient groups with the poorest outcomes which include those with (1) PAX3(7)::FOXO1 gene fusions and other adverse biological features; (2) RMS in adolescents and young adults; (3) RMS that has already spread to distant organs at the time of first presentation (metastatic disease); (4) RMS that progresses or recurs following first-line treatment. This review highlights the ongoing urgent unmet need for improved treatments in these patients and discusses novel therapeutic approaches currently being investigated.

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma in children and young people. Despite the advances in multimodality treatment over recent decades through successive prospective clinical trials, improved rates of survival for patients are mainly limited to those with localised RMS without adverse biologic features. Current clinicopathologic prognostic factors include PAX3(7)::FOXO1 fusion status, the site of primary disease, the pre-chemotherapy extent of disease (including microscopic vs. macroscopic residual disease, locoregional nodal involvement and metastatic status), tumour size and patient age. These factors are used to stratify patients into prognostic risk groups that guide treatment intensity and duration. Risk stratification algorithms are evolving, supported by advances in molecular biology and cancer genomics. In this review we focus on the poorest prognostic groups of paediatric-type RMS (i.e., Very High Risk or relapsed/progressive disease). These include patients whose tumours harbour poor biological characteristics such as PAX3(7)::FOXO1 fusion-positive tumours with locoregional nodal involvement and tumours harbouring other poor-risk genetic variants (particularly MYOD1 and TP53 variants); adolescent and young adult patients; newly diagnosed patients with metastatic RMS; and patients with relapsed and refractory disease. Here we aim to describe the clinical characteristics of these patients, outline current standard multimodality treatments in the context of sequential international clinical trials across the major cooperative groups and summarise emerging novel diagnostic and therapeutic approaches.

Full text

1. Introduction

Rhabdomyosarcoma (RMS) is a mesenchymal neoplasm with morphologic and immunophenotypic features of skeletal muscle differentiation [1]. It is the most common soft tissue sarcoma (STS) in children and young people (CYP) accounting for about 3% of all childhood cancers and remains a major contributor to cancer-related deaths [2,3]. Despite the advances in multimodality treatment over recent decades through successive prospective clinical trials executed by international paediatric oncology consortia, improved rates of survival for patients are mainly limited to those with localised RMS without adverse biologic features. Current clinicopathologic prognostic factors, which include PAX3(7)::FOXO1 fusion status, the site of primary disease, tumour size, patient age, the extent of pre-chemotherapy disease and locoregional nodal involvement, are well recognised in localised RMS and have been used in clinical trials to stratify patients into prognostic groups that guide treatment intensity and duration. Although these risk factors are similar amongst collaborative groups, they are used slightly differently by different consortia resulting in non-uniform risk stratification algorithms. Risk stratification evolves with an improved understanding of how clinical and biologic features affect disease course; advances in molecular biology and cancer genomics are enabling more sophisticated biologically based risk stratification systems. Table 1 shows the current European paediatric Soft Tissue Sarcoma Group (EpSSG) risk stratification for patients with rhabdomyosarcoma and reports EFS and OS from RMS 2005 and MTS 2008 [4,5,6].

In this review we will primarily focus on the poorest prognostic groups of paediatric-type RMS (subsequently termed poor-risk disease) which include (i) newly diagnosed patients who have biologically poor-risk non-metastatic disease, particularly PAX3(7)::FOXO1 fusion positive (FP) with concomitant lymph node positivity (N1) (5-year overall survival (OS): 45.5%; 5-year event-free survival (EFS): 43%) [7]; newly diagnosed patients with tumours harbouring poor-risk genetic variants (MYOD1 or TP53) or more recently identified fusions (FUS/EWSR1::TFCP2, MEIS1::NCOA2 or ZFP64::NCOA3) [8,9,10]; (ii) adolescent and young adult (AYA) patients; (iii) newly diagnosed patients with metastatic disease (3-year OS: 49.3%; EFS: 35.5%) [11,12]; (iv) patients with relapsed and/or refractory disease (3-year OS: 17%) [12]. Here we aim to describe their clinical characteristics, outline different aspects of current standard multimodality treatments and their historic implementation across cooperative groups and hypothesise how this may evolve to improve outcomes in the future.

2. Biologically Poor-Risk Disease

Histology as a surrogate for adverse biology has long been included in RMS risk stratification systems. Advances in our understanding of tumour biology and the availability of molecular pathology testing techniques and data have enabled us to further recognise underlying molecular risk factors of disease. The current fifth World Health Organization (WHO) histological classification of RMS includes embryonal rhabdomyosarcoma (ERMS), alveolar rhabdomyosarcoma (ARMS), spindle cell/sclerosing rhabdomyosarcoma (ssRMS) and pleomorphic rhabdomyosarcoma (almost exclusively seen in adults and not considered in this review) [13,14]. However, it is now well recognised that certain molecular characteristics are better prognostic biomarkers than histology. Two dominant molecular subtypes exist in RMS: PAX3(7)::FOXO1 FP and PAX3(7)::FOXO1 fusion-negative (FN) RMS, most usually ARMS and ERMS, respectively. Among patients with ARMS, approximately 80% are characterised by the PAX3 or PAX7::FOXO1 fusion [14]. Apart from the characteristic translocations giving rise to these fusions, the burden of somatic aberrations in FP RMS is relatively low [15]. In contrast, FN RMS often harbours aberrancies in key signalling pathways and/or other rarer mutations and gene fusions [14,15,16,17]. An improved understanding of specific molecular landscapes can help inform adaptations to current risk stratification systems and also potentially indicate biologically based novel therapeutic strategies. In addition to PAX3(7)::FOXO1 fusion status, we will briefly discuss other poor prognosis molecular subgroups.

2.1. PAX3(7)::FOXO1 Fusion-Positive RMS

FP RMS harbours either a characteristic t(2;13)(q35;q14) translocation giving rise to the fusion of PAX3 with FOXO1 (60% of cases) or a t(1;13)(p36q14) translocation resulting in the fusion of PAX7 and FOXO1 (20% of cases) [1]. PAX3 and PAX7 code for a family of transcription factors expressed in skeletal muscle progenitor cells and FOXO1 codes for a forkhead transcription factor. The gene fusions encode a novel aberrant transcription factor with a DNA-binding domain at the amino terminus of the PAX3 or PAX7 protein and a transcriptional activation domain at the carboxyl terminus on FOXO1 [1]. Enhanced transcriptional activation is thought to be a functional result of the decreased sensitivity of FOXO1 to inhibitory effects of PAX3 or PAX7 and also a result of higher levels of mRNA and protein expression of PAX3(7)::FOXO1 [1]. PAX3::FOXO1 has been more extensively studied and as a transcription factor increases the expression of downstream target genes by binding to PAX3-binding sites in close proximity to these genes. It also reprogrammes the chromatin landscape, enabling target gene promotor initiation [1]. PAX3::FOXO1 and PAX7::FOXO1 therefore functionally act as oncoproteins through their dysregulation of multiple downstream cellular networks that promote cellular phenotypes such as increased growth, survival and differentiation [1]. Alternate rare fusions with PAX3 have been reported, but there are limited outcome data on patients with these molecular subtypes and, to date, no definitive clinical relevance can be ascertained.

PAX3(7)::FOXO1 fusion positivity is recognised as one of the strongest negative prognostic factors in RMS. Evidence to support the use of fusion status instead of morphology in risk stratification comes from the finding that 20% of ARMS cases that are FN have gene expression profiling, disease course and outcomes more in keeping with ERMS [17,18,19,20,21]. FOXO1 fusion status has replaced favourable or unfavourable histology in current RMS risk stratification systems in the Children’s Oncology Group (COG) and EpSSG [4,22]. This is an important example of the use of molecular diagnostic tools to guide risk group allocation in this new era of molecular disease characterisation in paediatric oncology [23]. The recently completed COG trial ARST1431 and currently ongoing EpSSG FaR-RMS (Frontline and Relapsed RMS) trial prospectively evaluate this modified risk stratification which de-escalates therapy for alveolar, FN patients with localised disease [4,24].

Evidence for whether the FOXO1 fusion partner PAX3 or PAX7 differentially impacts prognosis has thus far been equivocal [25]. Publications from the COG have shown that in metastatic patients, those with PAX3::FOXO1 gene fusions had poorer outcomes and a higher correlation with bone marrow involvement [26]. In a univariate analysis of patients with localised disease, those with PAX7 had an improved prognosis [27]. Currently both translocations are considered unfavourable in clinical risk stratification systems.

Accompanying genetic aberrations in FP tumours most commonly include focal CDK4 (13%) and MYCN (10%) amplifications while variants in BCOR, NF1, TP53 and PIK3CA are less frequent [15]. The additive impact of these secondary aberrations on clinical course remains to be elucidated. The amplification of 1p36 and 13q13, which contain PAX7 and FOXO1 genes, respectively, are seen in 90% of ARMS with a PAX7::FOXO1 fusion [28]. PAX7 amplification in PAX7::FOXO1 ARMS has been associated with an improved prognosis, which has not been confirmed by multivariate analysis [28]. The amplifications of 2p24 (containing MYCN) in approximately 10–15% of ARMS and 12q13-q14 (containing CDK4), most commonly seen in PAX3::FOXO1 fusions, may be associated with a worse prognosis, but this needs to be confirmed in larger studies [29]. Somatic TP53 mutations are found in <4% of ARMS and are associated with a poor prognosis [10].

Non-metastatic ARMS with regional nodal involvement accounts for about 10% of all newly diagnosed patients with RMS [7,30]. In the EpSSG RMS 2005 trial, the 5-year EFS of patients with FP ARMS N1 tumours was significantly inferior to that of patients with FN ARMS N1 tumours (43% vs. 74%, p = 0.01) underscoring the prognostic superiority of molecular subtype over histologic subtype [7]. Irrespective of the primary site or the chemotherapy regimen used to treat FP N1 RMS, patient outcomes are similar [31]. The poor EFS and OS, which are only marginally better than those for metastatic RMS, have influenced the current EpSSG approach to stratify N1 FP RMS into the Very High-Risk (VHR) group which also includes patients with metastatic disease (M1), resulting in more intensive treatment [4]. The ability to discern between PAX3(7)::FOXO1 FP and FN disease has enabled a firm embedding of this molecular variable into current risk stratification systems both in the EpSSG as well as COG and allows for improved accuracy in identifying RMS subtypes of clinical significance. Fusion status however is only the first of several molecular biomarkers in RMS that are, or in the near future will be, incorporated into risk stratification systems and exemplifies the dynamic process by which prognostic factors are considered.

2.2. PAX3(7)::FOXO1 Fusion-Negative Poor-Risk RMS Subtypes

Fusion-negative RMS tumours most often have an embryonal histology, are aneuploid and can harbour molecular aberrations such as a loss of heterozygosity (LoH) at 11p15, with chromosomal gains (2, 8, 7q, 11q and 13), losses (10 and 15) and mutations of TP53, PIK3CA, FGFR4, FBXW7, ALK, BCOR, NF1, CTNNB1, ARID1A, MYOD1, CDKN2A and CDKN2B [10,23,32,33,34]. Mutations in the RAS pathway are the commonest aberration in FN RMS, present in >50% of cases, with an even higher predominance in infants [10,23]. RAS mutations currently are not considered on their own to convey poorer outcomes [10,23] although previous smaller retrospective molecular studies had suggested this [35,36]. The clinical and prognostic significance of several candidate mutations is currently under investigation. How RMS patients with specific molecular features should be treated remains under debate, with some proposing the “upstaging” of certain biological subgroups to justify more aggressive treatment [4,22,23].

2.2.1. TP53 Variants

TP53 induces the transcription of genes associated with various cellular activities including cell cycle arrest, apoptosis and cellular metabolism [37]. TP53 also has transcription-independent properties such as a role in DNA repair [38]. Molecular variants that are functionally tumour suppressors, loss-of-function mutations, deletions or gain-of-function mutations in TP53 have been found in several cancers, including RMS, often resulting in the loss of tumour suppressor function and/or gain of oncogenic properties [39,40,41,42]. TP53 is the key germline cancer predisposition gene implicated in Li–Fraumeni Syndrome (LFS). Broadly there exist two types of TP53 mutations: the first in which mutations occur at amino acids directly interacting with DNA, and the second in which mutations change the conformational structure of TP53, resulting in a loss of DNA binding activity [40,41,43,44,45]. In a recent analysis of various solid tumours (including RMS) from patients with LFS, somatic mutations implicating Wnt and PI3K/AKT pathways, epigenetic modifications, as well as homologous recombination and “prior chemotherapy” specific mutational signatures were observed [46]. Furthermore, an earlier LoH of TP53 and gain of the mutant allele may distinguish germline from somatic TP53 mutations [46].

Clinically, patients with RMS harbouring TP53 mutations comprise a heterogenous group. Mutations may be germline or somatic and occur in FN and FP RMS. In all cases, the aggressive clinical course of patients with TP53 mutations has been established. Shern, Selfe and colleagues performed an international consortium study designed to determine the incidence of driver mutations and their association with the outcome in RMS. Tumour sample DNA was collected from 631 patients enrolled on COG, International Society of Pediatric Oncology (SIOP) Malignant Mesenchymal Tumour (MMT) and EpSSG (RMS2005) studies between 1995 and 2017. Through custom capture sequencing, deep deletions, truncating mutations and point mutations were found throughout the mutant TP53 genes with some enrichment seen within the DNA-binding domain. Although we do not know yet whether specific mutations in the TP53 gene confer different prognoses, this analysis revealed the incidence of TP53 mutations to be higher than previously reported at 13% (69/515) in FN RMS and 4% (5/126) in FP RMS [10]. Due to a lack of matched germline samples, the authors were unable to ascertain whether the lesions represented somatic or germline events. TP53 mutations were associated with an inferior OS and EFS in both COG and SIOP-MMT/EpSSG cohorts (in both univariate and risk-stratified analyses) with authors nominating alterations in TP53 as a poor prognostic biomarker in both FN and FP RMS [10]. Although there is accumulating evidence that TP53 variants may be an independent adverse prognostic factor, TP53 status has not yet been uniformly adopted into risk stratification schema across collaborative groups. Whereas the EpSSG has not implemented change based on TP53 status, the COG has proposed that patients with very low-risk (VLR) or low-risk (LR) disease must have wild-type TP53 (and MYOD1) [47].

In a study by Shenoy et al., tumour mutational data (based on targeted panel sequencing) and central pathology review was ascertained for 146 patients (in whom TP53 status was available) enrolled on a series of COG studies between 1997 and 2013 [48]. TP53 mutations were found to be present in 9% of tumours (13/146). Among the 38 cases of anaplastic RMS in this patient cohort, 24% (9/38) had TP53 mutations. The majority of TP53 mutant tumours were noted to have an anaplastic histology (9/13; 69%) [48]. The study refuted previous reports that anaplasia was an independent poor prognostic factor but the authors concluded that in future, the presence of anaplasia could be used to identify tumours with the highest likelihood of harbouring a TP53 mutation, highlighting again that molecular subgrouping outperforms histologic subtype as a prognostic biomarker in RMS [48].

Hettmer et al. looked at RMS tissue from a small number of patients with germline TP53 mutations to determine whether this predisposed patients to develop RMS with anaplastic histology [49]. Eight cases with known TP53 germline mutations and RMS had their tumour histology retrospectively reviewed and seven cases of anaplastic RMS with unknown germline TP53 status underwent germline testing. Eleven TP53 germline mutation carriers with a median age of 40 months at diagnosis all exhibited an non-alveolar anaplastic histology. Among patients with LFS, anaplastic RMS may soon be recognised as one of the sentinel cancers indicating the need for germline TP53 mutation screening irrespective of age or family history [49]. Referral to genetics for germline mutation testing of several potential cancer predisposition genes is recommended for children with RMS who are younger than 5 years of age at the time of initial diagnosis, so that early identification can result in surveillance screening and the earlier detection of second malignancies for at-risk populations [50,51].

2.2.2. MYOD1 Variants

RMS harbouring a MYOD1^L122R^ variant can be found in patients of all ages although it is slightly more common in adults and older children [52]. MYOD1-mutant tumours frequently involve the head and neck extremities and trunk and have a slight female preponderance [52]. A retrospective analysis of pooled paediatric RMS cases from the US and UK showed a frequency of 3% among PAX3(7)::FOXO1 FN tumours (17/515) and these patients had universally poor outcomes, independent of clinical risk stratification [10]. The subtype of ssRMS in the current WHO classification includes RMS with MYOD1 mutations, although the mutation has been identified in ERMS also (less frequently). Since the initial recognition of ssRMS as an entity in the 2013 WHO classification, tumours harbouring MYOD1 mutations have been identified as a distinct poor prognostic subset [53]. Studies interrogating both spindle cell and sclerosing RMS in children and adults have found homozygous or heterozygous mutations in exon 1 of MYOD1, almost exclusively p.L122R, further justifying their shared grouping in the WHO classification as a single entity [54]. Tumours harbouring MYOD1 mutations have a more aggressive disease course and a less favourable prognosis, frequently progressing or recurring (often with metastases) despite the use of intense multimodal therapies [33,55]. In a recent review of 72 published cases of MYOD1-mutant RMS, only 12 of 37 patients (33%) with at least one year of follow-up were alive without disease and 43% had died [8]. A detailed analysis of 30 cases of MYOD1^L122R^-variant RMS showed that the median time to event in 25 patients with adequate follow-up data was 9 months and the majority (80%) of events included the failure of local control [52].

MYOD1 is necessary for myogenic differentiation via the control of the MAPK and PI3K-AKT pathways. It encodes a nuclear protein that transcriptionally regulates the differentiation of muscle cells by inducing cell cycle arrest, which is required for myogenic initiation. Mutant MYOD1 competes with wild-type MYOD1 for binding sites conferring decreased transcriptional activation and cellular differentiation and also binds to sites regulated by MYC [56]. For a detailed discussion on the mechanism of mutant MYOD1, please see Di Carlo et al. 2023 [8].

Tumours with MYOD1 mutations may also harbour additional co-mutations (HRAS, NRAS, FGFR4, PIK3CA), amplifications (MYOD1, FGFR4, IGF2, MDM2) or deep deletions (PTEN, GATA3, CDKN2A/B) and these too may impact prognosis although larger scale prospective studies are needed to explore this further [55,56,57].

In current clinical practice, the detection of somatic MYOD1 mutation now leads to “upstaging” within the risk stratification schema both in the COG and EpSSG [8,47].

2.3. PAX3(7)::FOXO1 Fusion-Negative Newly Emerging RMS Subtypes of Uncertain Prognostic Significance

2.3.1. TFCP2- and MEIS1-Associated Gene Fusions

RMS with TFCP2-associated gene fusions and less commonly MEIS1::NCOA2 fusions are an aggressive, extremely rare and newly discovered entity of primary intraosseous RMS [58,59,60]. These tumours mostly affect young adult patients (median age at diagnosis of 31 years), have a slight female preponderance and are occasionally found in the soft tissue but are most often intraosseous and have a strong predilection for the craniofacial bones where they can often destroy the cortical bone and invade surrounding soft tissue. The median survival of patients with TFCP2 fusion-positive RMS is poor (8 months in the largest published series) [9]. These tumours have a hybrid spindle cell and epithelioid histology and display nuclear pleomorphism and frequent mitoses. Underpinning this subtype is the TFCP2 (12q13.12) gene rearrangement wherein the 5’ portion of the gene is fused to an FET partner (EWSR1 or FUS) resulting in oncogenesis via transcriptional dysregulation through novel fusion proteins, the disruption of normal cellular functions and genomic instability [9,61,62,63,64,65].

TFCP2 codes for an evolutionarily conserved transcription factor targeting thymidylate synthase involved in the enzymatic regulation of DNA synthesis and contains an N-terminus DNA-binding domain and a C-terminus sterile alpha motif domain that plays a role in protein dimerization [9]. Furthermore, these tumours have complex genetic profiles but have homogenous transcriptomes. In the largest published series of 14 cases of FET::TFCP2-positive tumours, all cases that underwent genomic profiling showed homozygous deletions of the CDKN2A tumour suppressor gene, while other homozygous deletions were also displayed in a small number of cases [9]. Alterations of ALK (without underlying ALK fusions) resulting in ALK overexpression at both the transcriptional and protein levels have also been noted in a number of cases [9,64]. The upregulation of ALK does not result from translocation or amplification but rather correlates with ALK genomic deletion and therefore may involve alternative transcription initiation [9]. ALK expression related to intragenic deletions and aberrant splicing results in oncogenic ALK variants [64]. Functional studies show that FUS::TFCP2 blocks myogenic differentiation, induces the transcription of ALK and truncated TERT, and inhibits DNA repair [58]. TFCP2-rearranged tumours exhibit genomic instability and DNA methylation profiling demonstrates a close relationship with undifferentiated sarcomas [65]. Reports of sarcomas preceded by benign lesions carrying FUS::TFCP2 indicate stepwise sarcomagenesis [59]. Le Loarer and colleagues suggest that ALK upregulation in their cohort was as high or higher than that found in ALK-rearranged tumours such as ALK fusion-positive inflammatory myofibroblastic tumours. Whether there is a subset of these patients in whom ALK inhibition could potentially be utilised as a component of therapy remains to be seen and further information is needed to understand how to better manage these patients [9].

2.3.2. ZFP64::NCOA3 Gene Fusions

Han and colleagues have recently proposed a novel rare and aggressive molecular entity among spindle cell RMS which harbours a ZFP64::NCOA3 gene fusion [66]. The group describes a retrospective archival analysis of tissue, yielding five cases all occurring in adult males (aged 28–71 yrs). In the reported cases, all tumours arose from deep soft tissues and three out of five developed distant metastases, two of whom died of disease within 24 months from initial diagnosis. Only one of the five patients received systemic chemotherapy with curative intent in the frontline setting. Histologically, these tumours lack rhabdomyoblasts and are made of spindle cells arranged in a herringbone pattern amidst collagenous to myxoid stroma with monomorphic nuclei and variable mitotic activity [66]. In two of the five cases identified, the foci of metaplastic bone or dystrophic calcification were present. Most tumours retained immunohistochemistry in keeping with rhabdomyoblastic differentiation (patchy staining for desmin and focal staining for MyoD1 but negative for myogenin) [66]. The authors propose this as a new entity among spindle cell RMS but admit that in the absence of prototypic rhabdomyoblasts, this molecular subgroup could also represent an aggressive myofibroblastic derivative and additional studies are needed to further characterise this subgroup of tumours. Nuclear receptor coactivator 3 (NCOA3) is a member of the steroid receptor coactivator p160/SRC family and contributes to the pathophysiology of several cancers through gene expression modulation [67,68,69]. Zinc finger (ZF) proteins as a group are implicated in cell proliferation, apoptosis, immune function and tumorigenesis. ZFP64 is a coactivator of Notch1 signalling and may mediate differentiation in mesenchymal cells [70].

With respect to novel fusions such as PAX3::NCOA2 which have recently identified, it is not yet clear what the prognostic significance of these entities are. A recent publication suggests that this entity clusters with ARMS and PAX3::MAML on methylation profiling and therefore may be regarded like ARMS [71].

As we continue to identify poor-risk biologic candidates, the further refinement of EpSSG risk stratification is highly likely and follows a similar refinement within the COG [47].

3. Clinically Poor-Risk Disease

3.1. Rhabdomyosarcoma in Adolescents and Young Adults

RMS is typically a tumour of childhood but can occur at any age. Among the different prognostic factors, age in itself has been recognised as independently influencing survival, and as a consequence paediatric protocols include age as one of the variables used for risk stratification [72].

The survival rates of adolescent patients are inferior to those of children, but the outcome is even worse in adults, with historical published series reporting 5-year OS rates in the 20–40% range [73,74,75]. The reasons for this survival gap are likely multifactorial [75]. However, it has been suggested that the adverse outcome of AYA with RMS may be correlated, at least in part, to differences in clinical management (that could include patient referral patterns, time to diagnosis, enrolment into clinical trials and the adequacy and intensity of treatment). Various studies have suggested that AYA patients with RMS may have better outcomes when treated with multidisciplinary treatment in line with the paediatric approach, in particular with regard to the use of intensive multidrug chemotherapy [76,77,78,79,80,81]. However, outcomes seem to be inferior in AYA patients when compared with children, even when they were treated in the same way. A recent analysis from the EpSSG (on 1977 patients registered on two concurrent clinical protocols, the EpSSG RMS 2005 for localised RMS patients and the EpSSG MTS 2008 protocol for metastatic patients) reported a 5-year OS of 57.1% for patients aged 15–21 years and 77.9% for those aged 0–14 years old enrolled in the same trials (p-value < 0.0001) [80]. These data suggest that the more recent outcomes for AYA patients with RMS are better than those reported in epidemiological studies (for example, the EUROCARE-5 study which covered 2000–2007 showed a 39.6% 5-year OS in 15–19 year old patients), supporting the idea that these patients should be included in paediatric RMS trials to offer them the best chances of a cure. Poor AYA outcomes may suggest that part of the prognostic gap between children and AYA patients with RMS could be related to differences in tumour biology and the intrinsic aggressiveness of the disease. We discussed earlier in this manuscript the prognostic impact of RMS with MYOD1 mutation and TFPC2 gene fusions, both of which may have a higher incidence in the AYA age group when compared with younger age groups. In addition, an alveolar histology (as a surrogate for PAX3(7)::FOXO 1 fusion status) forms a higher proportion of RMS cases in adolescent patients aged 15–21 years than in patients < 15 years (46% vs. 26%) [78]. Age-independent scientific collaboration and the development of more integrated genomic approaches focused on AYA patients are needed. While these are awaited, we estimate that RMS in AYA patients should be considered “poor-risk” in itself.

3.2. Patients with Metastatic (M1) Disease at Presentation

Approximately 15–20% of paediatric patients have at least one site of distant metastatic disease detected radiographically or clinically at the time of initial presentation. Table 2 summarises potential sites of metastatic disease and modalities which can be used in detection as part of staging. Metastatic spread occurs via both the haematogenous and lymphatic routes. The most common metastatic site is the lung (involved approximately 50% of the time) [11,81,82]. Bone marrow, bone (each involved around 30% of the time) and distant nodes are also common sites of metastatic involvement with rarer sites including pleura, peritoneum and cerebrospinal fluid (CSF) [82]. Outcomes for patients with metastatic disease remain dismal and although most patients achieve an objective response with systemic induction chemotherapy (92% of patients achieved ≥33% volume reduction in the primary tumour) and local control measures, the majority of patients relapse, resulting in a historic 3-year EFS and OS of only 27% and 34%, respectively [6,82]. As for adolescents and adults there is an overlap between clinical and biological risk factors: in recent European studies 57–60% of patients with metastatic RMS have an alveolar histology compared with 24% of patients with localised RMS [5,80,83]. For this group of patients, frontline treatment has not advanced significantly over the last several decades in spite of attempts at therapy intensification; new approaches are desperately needed.

3.2.1. Risk Stratification in Metastatic Disease

The presence of distant metastases at diagnosis is the single most powerful predictor of survival in patients with RMS. An analysis of 788 patients enrolled on European and COG studies between 1984 and 2000 aimed to define pretreatment clinical characteristics that could stratify patients into risk groups to guide treatment decisions, such that patients with the highest risk of relapse could be identified for frontline innovative therapies [82]. Four independent adverse risk factors (Oberlin factors) were identified: age < 1 year or ≥10 yrs, site (extremity, other or unknown primary), bone or bone marrow involvement and ≥3 metastatic organ sites. The 3-year EFS was 50% for patients without any risk factors and 42%, 18%, 12% and 5%, respectively, in patients with one, two, three or four risk factors [82]. Two distinctly different prognostic subgroups were identified through this analysis: patients with 0–1 Oberlin risk factors (3-year EFS of 44%) and those with 2–4 Oberlin risk factors (3-year EFS of 14%) delineating favourable and unfavourable sub-strata within metastatic RMS [82]. However, outside the trial setting, no cooperative group is currently uniformly using this approach to risk stratify patients or inform treatment decisions in patients with metastatic RMS. Within the current EpSSG FaR-RMS study, Oberlin risk factors are taken into account to establish eligibility for radiotherapy (RT) randomisations [4]. Notably, histology was not found to be a significant prognostic factor in the Oberlin analysis and PAX3(7)::FOXO1 fusion status was not available for the patients included in the analysis. In the future, a similar analysis, taking into account adverse biologic features such as fusion status may provide novel insights.

The recently published EpSSG MTS 2008 study in metastatic RMS analysed outcomes by Oberlin risk factors and showed that patients with 0–2 Oberlin risk factors have an improved 3-year OS (60%) compared with the historical Oberlin cohort [6]. This may be the result of the gradual introduction of more sensitive staging diagnostics, potentially adding, for instance, a site to the cumulative Oberlin risk factors and thereby allowing for the more accurate calculation of risk for some patients [6].

In a recently published COG paper, pooled data from six COG trials (three of which included metastatic patients; n = 187) was used for survival tree regression for EFS to recursively select prognostic factors that would result in the branching and split of the data [84]. In this analysis, the clinical factors included were age, FOXO1 status, clinical group, histology, nodal status, the number of metastatic organ sites, primary site, sex, tumour size and the presence of metastases in specific sites. In patients with metastatic RMS, FOXO1 FP was the strongest negative prognostic factor [84]. Five-year EFS and OS were 46% and 58%, respectively, for FN patients and only 6% and 19%, respectively, for FP patients [84]. Their model identified metastatic FN patients with a single site of metastatic disease as a better prognostic group with improved 5-year EFS (54%) and OS (70%) compared with FN patients with >one site of metastatic disease (5-year EFS of 34% and OS of 40%). The authors proposed a high-risk group which would include all metastatic patients with FP RMS and metastatic patients with FN RMS who had more than one metastatic site [84]. The latter group, i.e., single-organ site metastatic disease has been reported in multiple studies to have higher rates of survival [82,84,85,86,87].

The COG study ARST0431 showed that among metastatic patients, the group (n = 20) that appeared to have an improved OS were patients with embryonal RMS who were younger than 10 years old at the time of initial diagnosis, carrying a 3-year EFS and OS of 60% and 79%, respectively [88]. It is on this basis that within the COG, metastatic patients younger than 10 years with FN RMS are considered a more favourable prognostic subgroup and are treated as Intermediate Risk (IR). By contrast, within the EpSSG and Cooperative Weichteilsarkom Studiengruppe (CWS), fusion status or age does not currently impact how metastatic patients are risk stratified or treated, as all metastatic patients are stratified as VHR within the current FaR-RMS study.

Patients with distal extremity RMS who present with proximal nodal involvement (without distant metastatic disease) have a prognosis comparable to M1 patients [89,90,91]. In contrast, patients with distal extremity primaries with only popliteal or epitrochlear nodal involvement have comparatively better outcomes [91]. An interesting point of discussion was broached by Terwisscha et al. exploring the idea of considering patients with distal extremity RMS and proximal nodal involvement as having the equivalent of metastatic disease given their poor prognosis [91]. Although this specific approach has not yet been adopted, it is important to note that the majority of patients with distal extremity RMS were found to have an alveolar histology (and when tumours were tested, were FP) and this subgroup (FP N1) is known for poor outcomes and has been grouped together with metastatic disease in EpSSG’s current VHR group in the FaR-RMS trial. Perhaps the small subgroup of FN cases in this cohort requires prospective validation to understand this further.

3.2.2. Single-Organ Metastatic Involvement

The commonest single metastatic site in RMS is the lung followed by the lymph nodes [82,86]. A number of studies suggest improved outcomes in lung-only metastatic RMS [86,87]. By definition, single-organ metastatic disease involving a site other than bone or bone marrow can at most have two Oberlin risk factors which may contribute to improved subgroup outcomes.

A COG study recently analysed 428 patients with metastatic ERMS and ARMS. The 55 patients with lung-only metastases were more likely to be <10 years of age or have ERMS and were less likely to have locoregional nodal disease or primary extremity tumours [86]. The lung-only cohort had significantly better survival outcomes than patients with all other sites of metastases with a 5-year EFS of 48.1 versus 18.8% and a 5-year OS of 64.1 versus 26.9% [86]. In patients with ERMS and lung-only metastases, there was no significant difference in survival between patients ≥ 10 years and 1–9 years, raising the question of whether all patients irrespective of age with FN RMS and lung-only metastases will be stratified differently in future COG studies [86].

An EpSSG analysis of 270 M1 patients registered on the MTS 2008 study reported that among 59 patients with lung-only metastases, the 3-year EFS and OS were 40% and 60%, respectively, compared with 28% and 35% in the lung and other sites and 36% and 49%, respectively, in other single organ sites [87]. The improved outcomes in lung-only metastatic RMS could be accounted for by these patients having fewer Oberlin risk factors than other metastatic patients [87].

The CWS performed a similar analysis, although their patient cohort was restricted to 53 patients with ERMS and lung-only metastases and the authors found that the 5-year EFS and OS were 41% and 52%, respectively. This improved outcome is expected given that the analysis was restricted to favourable histology/fusion-negative patients with single-organ metastases [92].

Mercolini et al. investigated the clinical characteristics and outcomes of M1 patients on EpSSG protocols in whom distant lymph nodes were the only site of metastatic involvement, accounting for about 7% of all metastatic patients. In 22 eligible patients the primary tumour was often ARMS (n = 15) and/or >5 cm in size and/or was located in the distal extremity. Patients with metastatic nodal involvement almost always had locoregional nodal involvement (21/22) [85]. Of the 18 tumours for which the fusion status was available, 12 were FP. Following the completion of induction chemotherapy and local control (surgery in 50% and RT in 91%) most patients went on to receive maintenance chemotherapy. The median follow-up was 54 months and the 3-year EFS and OS were 67.1% and 71.9%, respectively [85]. In the cohort of patients for whom fusion status was known, no events occurred in the six patients with FN RMS and the 3-year EFS and OS for FN vs. FP patients were 100% and 100% vs. 46.6% and 57.1%, respectively [85]. The results of this study suggest that this patient subgroup, and especially the FN patients, may represent a better prognostic subgroup among metastatic patients. The 3-year EFS and OS were 72% and 79.4%, respectively, for patients with zero or one Oberlin risk factors and 57.1% and 57.1% for those with two factors, which are markedly improved according to what would have been predicted by Oberlin et al. from the 2008 retrospective analysis [82]. The authors hypothesise that lymphatic dissemination is likely to be biologically distinct from haematogenous spread and this may impact improved survival in this patient subgroup. They went on to further postulate that the comparative ease with which local control can be delivered (in particular with the use of RT) to lymph nodes may also positively impact survival rates [85].

The result of this study raises the question of whether RMS patients with distant node-only metastatic disease should continue to be treated in the same way as other metastatic patients. Given the limited sample size, future initiatives aimed at addressing this question through the analysis of pooled data in the International Soft Tissue Sarcoma Consortium (INSTRuCT) database are currently ongoing [93].

3.2.3. Systemic Therapy for Metastatic RMS

Attempts at improving outcomes through the intensification of systemic therapy via the addition of active agents in metastatic RMS have been largely unsuccessful [94]. Although outcomes over time are marginally improved in comparison with historical cohorts, there have been no statistically significant improvements seen in EFS or OS that are attributable to the addition of new agents to the standard backbone regimen. The marginal improvements in outcome are likely to be multifactorial, for example, secondary to improved supportive care and improvements in local control techniques. Over the years and across paediatric oncology consortia, if a chemotherapeutic agent (or combination) was found to be active in preclinical RMS models, it has been tested in a window study. Therapeutic window studies have been conducted with the goal of confirming the utility of chemotherapeutics in high-risk patients.

Patients with metastatic disease and especially those with >one Oberlin risk factor remain a challenging population to treat. Therapy intensification has included adding active or targeted agents to the standard ifosfamide, vincristine, actinomycin D (IVA) or vincristine, actinomycin D, cyclophosphamide (VAC) backbone chemotherapies, dose escalation, the delivery of chemotherapy on an interval-compressed schedule, administering high-dose chemotherapy with stem cell rescue (SCR) and/or prolonging therapy with known active agents in the form of maintenance. Table 3 lists completed international clinical trials which have recruited patients with metastatic RMS over the last few decades. In the following sections we detail the conclusions drawn from various cooperative group studies over time that are relevant to the treatment of metastatic RMS.

Adding Active Agents to Standard Backbone

Despite numerous modifications to systemic chemotherapy regimens, outcomes for patients with metastatic RMS have not improved significantly over time. Early trials such as IRS I and II showed no improvement in outcome in patients with stage IV RMS following the addition of doxorubicin to a VAC backbone [95,96,113,114]. IRS III introduced cisplatin and etoposide to select patients, showing modest improvements in progression-free survival (PFS) and OS for certain primary tumour sites [97]. Subsequent IRS IV and other IRSG/COG studies (as shown in Table 3) identified active agents through phase II window trials, but integrating these into standard regimens failed to improve survival. Notably, the D9802 study demonstrated that vincristine–irinotecan (VIr) was highly active in newly diagnosed metastatic RMS [100]. The ARST0431 trial, which used an intensified multi-agent regimen including VIr, showed improved outcomes (3-year EFS 69%), but only in patients with favourable risk profiles (0–1 Oberlin risk factors) [88].

European trials like RMS79 and CWS-81/86 also did not yield major advances. RMS79 showed no benefit from different actinomycin-D schedules [104]. CWS-81 had dismal survival outcomes (11% 5-year DFS), while CWS-86 slightly improved outcomes (19% 5-year EFS) with a shortened VAIA regimen [109,110].

The MMT 98 study confirmed the safety but limited efficacy of carboplatin in RMS [107]. However, a doxorubicin window within MMT 98 showed a promising 60% partial response rate, reinforcing its activity in RMS, though without definitive survival benefit [107].

The EpSSG subsequently adopted doxorubicin combined with standard IVA chemotherapy (IVADo) the first four cycles of induction treatment to a total doxorubicin dose of 240 mg/m^2^ for metastatic RMS within the BERNIE and MTS 2008 studies. Patients in both studies also received 12 months of vinorelbine and cyclophosphamide maintenance chemotherapy as the standard of care (see Maintenance Chemotherapy Section). In a pooled analysis of patients from both studies, 372 RMS patients with a median follow-up of 55.2 months had a 3-year EFS and OS of 35.5% and 49.3%, respectively [6]. Patients with <two Oberlin risk factors had improved outcomes compared with patients with three or four risk factors (3-year EFS of 46.1% versus 12.5% and 3-year OS of 60.0% versus 26.0%) [6]. Patients with metastatic RMS treated as per CWS guidelines with CEVAIE during a similar era had companiable outcomes [11].

Adding Targeted Agents to Standard Backbone

Several recent trials were designed to incorporate targeted agents with standard backbone chemotherapy in frontline metastatic RMS. The BERNIE phase II study randomised patients with metastatic STS (n = 103 with RMS) to standard IVADo/IVA chemotherapy with or without bevacizumab, a monoclonal antibody targeting vascular endothelial growth factor (VEGF) [108]. Since STS significantly overexpresses angiogenic VEGF, the rationale was to inhibit tumour angiogenesis. The addition of bevacizumab to backbone chemotherapy was tolerable but there was no significant difference in median EFS between the STS patients who did or did not receive bevacizumab and the 2-year EFS was 41% for RMS patients within both groups. The study was unable to establish prognostic value for bevacizumab-related plasma or tumour biomarkers [108].

A COG study (ARST08P1) evaluated the feasibility and efficacy of adding cixutumumab (insulin-like growth factor-1 monoclonal antibody) or temozolomide to the ARST0431 intensive chemotherapy backbone (VDC/IE/VIr/VAC) in patients with metastatic RMS. No additive toxicities were noted but the addition of neither cixutumumab (3-year EFS 16%) nor temozolomide (3-year EFS 18%) showed a survival benefit [103].

Dose Intensification with Interval-Compressed Chemotherapy

The high-risk COG study ARST0431 for patients with metastatic disease trialled a multipronged approach to improve outcomes in this patient subset [88]. Not only were active agents (irinotecan, doxorubicin, etoposide and ifosfamide) in RMS added to VAC backbone chemotherapy but irinotecan also doubled as a radiosensitizer and VDC/IE chemotherapy was delivered according to an interval-compressed schedule with chemotherapy cycles being delivered every 14 days with GCSF support. All M1 patients included in ARST0431 (n = 109) were reported to have a 3-year EFS and OS of 38% and 56%, respectively [88]. This was improved compared with the preceding survival reported in the IRSG/COG window studies (best reported outcomes were with the combinations of ifosfamide/doxorubicin with a 3-year FFS of 25% and OS of 39%and ifosfamide/etoposide with a 3-year FFS of 33% and OS of 55%) and was very similar to the EpSSG’s MTS 2008 outcome [81,115]. Patients with 0–1 Oberlin risk factors had a 3-year EFS of 67% which was improved compared with historical cohorts and those with 2 Oberlin risk factors had a 3-year EFS of 19% [88]. One of the challenges of the design of ARST0431 was the inability to discern which of the three interventions (the use of irinotecan as a radiosensitizer; the use of interval-compressed cycles; or the use of multiple known active agents) provided the most patient benefit and the results were confounded by the finding that the FN < 10 year cohort (subsequently treated as IR in COG studies) accounted for the improved EFS and OS for the entire patient cohort. Dose intensification through the addition of irinotecan on days 8–12 of a standard 21-day IVA cycle (IrIVA) is currently being investigated in the FaR-RMS study [4].

High-Dose Chemotherapy with Stem Cell Rescue

Several clinical trials (MMT4 91, RMS 4.99 and MMT 98) have investigated the role of myeloablative chemotherapy with SCR in an effort to intensify systemic therapy for the highest risk patients with metastatic disease. There was no significant improvement in survival outcomes but an increased risk of treatment-related morbidity was found [105,107,112]. A Cochrane systematic review of high-dose chemotherapy for children and young adults with stage IV RMS concluded that the use of high-dose chemotherapy (HDC) with SCR as a standard therapy in metastatic RMS is not justified by the existing literature but that selection bias could have contributed to the poor outcome for patients with adverse risk factors who were treated with HDC and SCR, resulting in an underestimation of efficacy [116]. No randomised controlled trial to address the efficacy of HDC and SCR has been conducted to date in patients with RMS. The toxicity associated with HDC currently excludes this as a standard option in patients with metastatic RMS [116,117].

Maintenance Chemotherapy

A prolonged duration of treatment and metronomic dosing may both contribute to a maintenance-type approach in RMS [118]. Minimal residual disease may be present even when disease shows complete resolution on imaging. Although not typically thought of as maintenance chemotherapy, patients on IRS I and II had prolonged treatment periods for up to 2 years, during which time cyclophosphamide was given orally. CWS 96 allocated metastatic patients (by physician choice) into two groups following carboplatin, epirubicin, vincristine, actinomycin D, ifosfamide and etoposide (CEVAIE) to either receive high-dose chemotherapy with thiotepa, cyclophosphamide, melphalan and etoposide followed by autologous stem cell return or low-dose oral metronomic maintenance therapy with trofosfamide and idarubicin (TI) alternating with trofosfamide and etoposide (TE) [11,111]. Oral maintenance therapy appeared to show an improved outcome, but selection bias could not be eliminated. This was one of the first studies to pave the way towards the further exploration of this approach. A more recent analysis of patients from the CWS-IV2002 and DOKIV 2004 trials concluded that maintenance chemotherapy was superior to allogeneic haematopoietic stem cell transplant or high-dose chemotherapy in metastatic RMS. However, in addition to inherent selection bias in this non-randomised comparison, treatment arms were significantly imbalanced, bringing into question the strength of this evidence [119].

Vinorelbine and cyclophosphamide chemotherapy is an active combination both in the relapse setting and in the frontline setting in High-Risk (HR) and VHR patients [120,121]. Following the completion of 25 weeks of IVADo, patients in the BERNIE and EpSSG MTS 2008 studies received 12 months of oral maintenance chemotherapy with vinorelbine and cyclophosphamide in a non-randomised fashion. Outcomes in metastatic RMS were modestly improved compared with previous studies but the contribution of maintenance chemotherapy could not be clearly ascertained [6]. Of 270 patients, 125 (46%) developed progressive disease, had an insufficient response, relapsed or died either during/at the completion of induction (n = 65) or during maintenance treatment (n = 60) [6]. The FaR-RMS study is exploring prolonging maintenance treatment, with metastatic patients being randomised between 12 and 24 months of maintenance chemotherapy [4].

Outcomes for patients with metastatic disease remain suboptimal despite the intensification of systemic therapy via the addition of known active and targeted agents, the use of HDC and SCR and the addition of maintenance chemotherapy. New approaches are urgently required to address this ongoing unmet need.

3.2.4. Local Control in Metastatic RMS

In the EpSSG protocols, disease reassessment occurs after six cycles of induction chemotherapy in patients with metastatic RMS, with an aim to initiate the delivery of local control in the form of surgery and/or RT from week 22 (i.e., following cycle seven) in most cases.

1.Surgical Resection

Primary surgical excision is rarely indicated in the setting of metastatic disease. The delayed resection of the primary site following neoadjuvant chemotherapy is the preferred approach, where appropriate. Surgery to metastatic sites is sometimes considered in individual cases, especially in patients with oligometastatic disease, though the mainstay of local therapy for metastatic sites is RT.

2.Radiotherapy

RT is recommended for the primary tumour site, any involved locoregional lymph nodes and as far as possible to all metastatic disease sites [4,6,122]. The timing for RT in metastatic disease is similar in the COG (week 20) protocols. The RT dose is dependent on the post induction chemotherapy tumour volume and for patients with multiple sites of metastatic disease, some sites may be prioritised, with RT to other sites being deferred. Dávila Fajardo et al. recently published an overview on the details and specificities of RT for RMS [122].

Despite the recommendation for all sites of disease to receive RT, data from MTS 2008 and BERNIE revealed that only 22% of all patients received RT to all metastatic sites. A landmark analysis of RMS patients from the BERNIE study reported better outcomes for patients receiving RT to some or all disease sites [123]. The analysis of MTS 2008 data confirmed that the 3-year PFS was impacted by the extent of RT and was highest in patients receiving irradiation to all disease sites (radical RT; 3-year PFS 62%), lower in those receiving partial RT (3-year PFS 39.5%) and lowest in patients receiving no RT (3-year PFS 30.1%) [123,124]. Ferrari and colleagues retrospectively reviewed a cohort of 80 paediatric and AYA patients with M1 RMS treated between 1990 and 2020 and found that patients who had RT to all metastatic sites had an improved 5-year EFS and OS of 70.6% and 76%, respectively [125]. These retrospective findings cautiously suggest that RT to all disease sites may influence the outcome but must be interpreted cautiously in view of the retrospective nature of the analyses and the risk of bias.

Although RT to all feasible metastatic sites is generally considered in the paediatric setting, this is not routinely the case in adults. Whether this influences the differential age-related outcome in RMS remains unclear. The delivery of RT in metastatic patients may have several attendant toxicities including myelosuppression, limiting the timely delivery of systemic therapy and increasing the susceptibility to infections. Furthermore, RT to multiple sites of disease can be a logistical challenge. In the case of oligometastatic disease, and when technically feasible, stereotactic ablative body radiotherapy (SABR) may be considered. Local treatment in the form of RT to all or some metastatic sites has been a topic of debate. The question of whether the benefit of complex, multi-site RT outweighs the burden of this intense treatment on these patients with a very poor prognosis is being prospectively addressed in a randomised study within the FaR-RMS trial [4,122].

3.Cytoreductive Surgery and Hyperthermic Intraperitoneal Chemotherapy

Peritoneal sarcomatosis can occur in children and young people with abdomino-pelvic RMS and presents a therapeutic challenge: an R1 resection is often not possible and limitations to whole abdomen RT exist due to the tolerance of the overlying bowel [126]. In this setting, CYP with RMS have been included in trials assessing the feasibility and safety of hyperthermic intraperitoneal chemotherapy (HIPEC) in specialised centres where the procedure has been well tolerated [127,128]. A recent review summarising reported cases in the literature included some apparent success in RMS but concluded that the impact of HIPEC versus systemic chemotherapy and debulking surgery on OS in paediatric malignancies is uncertain due to the lack of clinical trials and very small sample size across tumour subsets [129].

3.2.5. Biologic Challenges in Metastatic RMS

An in-depth discussion of why metastatic disease has been challenging to treat for the last several decades is beyond the scope of this review, but it is indisputable that better treatment strategies specific for metastatic patients are urgently needed. PAX3(7)::FOXO1 gene fusions encode potent aberrant transcription factors which impact growth, apoptosis and myogenic differentiation and stimulate metastasis [14,18]. Research on human primary RMS cell lines and animal models has shown that PAX3(7)::FOXO1 expression levels at the cellular level are heterogeneous [130]. To what degree this heterogeneity exists among primary and metastatic sites would be interesting to further understand. Variable expression levels may enable FP RMS cells to exploit their host on multiple levels and the selective targeting of PAX3(7)::FOXO1 high expressors may not prove to be curative. It may also be important to understand the pathways which influence fluctuating PAX3(7)::FOXO1 expression to explore treatments which may overcome plasticity [130]. Thus, the as-yet unsuccessful attempts to target the PAX3(7)::FOXO1 fusion gene may not be the anticipated holy grail.

An analysis of the integrin expression profiles of circulating plasma exosomes isolated from cancer patients may be able to predict sites of future metastasis in patients, which would revolutionise how we risk stratify metastatic patients and help develop inhibitors of metastasis before it has occurred in patients [131,132].

The tumour microenvironment (TME) including the extracellular matrix and vascular, stromal and immune cells may have an impact on the interplay between RMS cells and their metastatic fate. The influences of hypoxia, metabolites and cytokines on tumour cells and the TME is likely to influence not only the proliferative abilities of cells but also their metastatic potential [133]. For example, HIF activation is known to promote angiogenesis, metabolic reprogramming and stem cell-ness, which all have a combined impact on invasion and metastases [134,135].

Radio-resistant RMS cell lines studied in the lab are characteristic of an aggressive and pro-metastatic phenotype that can resist the harmful effects on the cell of reactive oxygen species, engage DNA damage repair pathways, counteract RT-induced cell cycle arrest and express cancer stem cell-ness [136]. Whether these features of radio resistance are intrinsic, acquired or a combination of both and the underlying mechanisms whereby this occurs will be pivotal in understanding the pathophysiology, especially for patients with in-field or field-edge local failure following RT. The problem of radio resistance also emphasises the importance of exploring potential radio-sensitising approaches.

Achieving objective radiologic responses is possible for the majority of patients with metastatic RMS, although in the MTS 2008 study only 103/259 patients completed both induction and maintenance chemotherapy with a median time to event of 11.6 months [6]. Despite the multimodality approaches and attempts to intensify treatments, there has been relatively little progress towards improving outcomes over the last four decades; the main ongoing challenge is achieving sustained response/remission. The development of liquid biopsies as a tool in RMS may enable the non-invasive detection of residual/relapsed disease and the genetic evolution of the tumour to facilitate future treatment strategies [137,138].

3.3. Relapsed RMS

Recurrent RMS remains very challenging to treat, with poor outcomes for most patients. In the EpSSG RMS 2005 study, 29% of all patients with M0 RMS had an event within 5 years, and in the MTS 2008 study 66% of patients with M1 RMS had an event within 3 years (median time to event of 11 months) [5,6]. The three-year OS following relapse is <20% underlining the ongoing need to develop novel approaches to treatment and ultimately more effective treatments for these patients [139,140].

3.3.1. Relapse Risk

There is a distinction between the relapse or progression of disease which occurs following an initial response to chemotherapy and the rarer primary refractory disease in which patients progress through treatment without an initial period of response. Refractory disease carries a very poor prognosis and makes up ~10% of the patients included in the VIT0910 study [140]. The majority of relapses occur after either a complete or partial response and the poorest outcomes among this group are those with relapse or progression while on treatment [141]. For patients who have completed frontline therapy, earlier relapses (<18 months from diagnosis) are associated with inferior survival [142].

There are several factors used in clinical risk stratification in newly diagnosed patients which correlate with an increased risk of relapse. These include age > 10 years or <1 year, histology (alveolar now replaced by PAX3(7)::FOXO1 fusion-positive status), site (unfavourable), IRS group (higher group), size (>5 cm) and nodal involvement (N1). In multiple studies, an alveolar histology has been identified both as a risk factor for recurrence (both localised and metastatic) but also as a harbinger of poor outcome following treatment for relapsed disease [7,19]. After metastatic status, fusion status is currently recognised as the most important factor associated with the outcome in patients with RMS and currently ongoing studies will prospectively study this [4]. Over time, the definitions of an unfavourable site have evolved as we continue to refine the prognostic relevance of this clinical feature. Currently within EpSSG, unfavourable sites of localised disease include extremities, head and neck parameningeal and “other” sites, and for metastatic disease, extremities and “other” sites [4].

3.3.2. Patterns of Relapse

Table 4 depicts the timing and pattern of relapse published by various cooperative groups over the last five decades. A reduction in both relapse rate and OS rates over time are attributed to the improved delivery of systemic treatment, which includes advancements in supportive care, as well as an improved ability to deliver RT, particularly in initially non-metastatic disease. An analysis of 695 patients with IRS group II RMS with or without resected positive lymph nodes treated on the IRS I-IV studies showed that distant failure rate (14%) exceeded local (8%) and regional (4%) failure rates. Over time, between 1972 and 1997, the distant failure rate decreased particularly in ERMS [143].

The pattern of relapse in RMS may vary and may include local, locoregional, oligometastatic or widely metastatic disease alone or in combination. For patients with initially localised RMS, most (64–76%) relapses involve the primary site of disease and draining nodal basin(s) [146,147,148]. Of these, 81–87% are local relapses involving only the primary site of disease and the remaining relapses occur in the locoregional lymph nodes with or without the involvement of the primary site [148]. Certain groups such as infants with incompletely resected IRS group III tumours have been noted to have higher rates of local failure which is attributable to the avoidance of aggressive local control approaches [149,150,151,152,153]. Most patients who have a metastatic recurrence have multiple synchronous metastases [145].

3.3.3. Systemic Therapy for Relapsed RMS

Successful efforts to bring new therapies to the clinic for relapsed RMS patients have been limited. There are few randomised trials in the relapsed setting and some of these trials have been designed to identify new agents for use in the frontline setting rather than to assess treatments specific to the relapse setting. Outcomes are limited by reduced efficacy, fewer opportunities for effective local therapy, heavy pretreatment and possibly acquired resistance mechanisms.

The COG conducted a trial comparing response rates of two schedules of irinotecan and vincristine in patients with relapsed or refractory RMS and reported outcomes from 89 assessable patients. One schedule was a total dose per cycle of 200 mg/m^2^ of irinotecan given on days 1–5 and 8–12 of a 21 day cycle (response rate, 26%; 1-year FFS 37%; 1-year OS 55%) and the second was 250 mg/m^2^ given on days 1–5 of a 21 day cycle (response rate, 37%; 1-year FFS 38%; 1-year OS 60%). The study concluded that as there was no difference in response rates between the two irinotecan schedules, the shorter, more convenient regimen would be taken forward [154].

A phase II COG study randomising patients at first relapse to vinorelbine and cyclophosphamide with either bevacizumab or temsirolimus reported a 6-month EFS of 69.1% for temsirolimus vs. 54.6% for bevacizumab and objective response rates of 47% and 28%, respectively. The study had a “pick the winner” design with no standard arm. The temsirolimus arm was superior to the bevacizumab arm and superior to historical controls [155]. Temsirolimus was taken forward into the ARST1431 frontline IR study but did not improve outcomes when combined with standard backbone chemotherapy (3-year EFS 64.8% in the VAC/VIr group vs. 66.8% in the VAC/VI plus temsirolimus group) [24].

Within EpSSG, a randomised phase II trial for relapsed (first or subsequent relapse) or refractory RMS patients tested the 5-day VI schedule with or without the addition of temozolomide and reported an overall response rate following two cycles of 44% for the vincristine, irinotecan and temozolomide (VIT) arm, which was superior to 31% for the VI arm [140]. Both PFS and OS were superior for the VIT arm (2-year OS 32% vs. 21%, p < 0.05; PFS 21% vs. 13%) in relapsed and refractory patients, although only OS was statistically significant [140]. VIT has been taken forward as the European standard of care in relapse for patients who received alkylating agents during first-line treatment. In the current FaR-RMS trial, relapsed RMS patients were randomised between VIT and VIr + regorafenib, a multi-tyrosine kinase inhibitor (TKI), following promising results with this novel combination in RMS in a phase Ib setting [4,156].

Patients have also been treated in non-randomised phase II studies. A recent COG publication described the outcome of 175 RMS patients treated in 13 COG single-agent phase II studies. Although not explicitly stated, this cohort was likely primarily comprised ofpatients who had had multiple relapses or whose disease was multi-refractory. In most cases, patients in first relapse are treated with multiple agents rather than single-agents. The six-month EFS was estimated at 16.8% and only trials of vinorelbine and rebeccamycin achieved their response-based primary endpoint [157].

Patients in the relapse setting who relapse/progress on more established phase II combinations are generally offered access to biomarker-driven phase I/II studies or targeted agents available through compassionate or extended use programmes based on the molecular profiling of the relapsed tumour. The COG MATCH trial (Molecular Analysis for Therapy choice; NCT 03155620) has included patients in phase II treatment arms of molecularly targeted therapies based on identified molecular variants in tumour samples [158]. A recent systematic review of early-phase studies that included children and young people with relapsed and refractory RMS found an objective response rate of only 22.6% and poor outcomes; but the study quality was limited by poor and inconsistent reporting [159].

Outside of the clinical trial setting, the choice of systemic and local therapy options may be made based on several considerations which have been recently outlined in an EpSSG summary of treatment recommendations in relapsed RMS [160].

3.3.4. Local Control for Relapsed RMS

Several factors influence the decisions around local control at the time of relapse, at the forefront of which is curative vs. palliative intent, the location and extent of relapsed disease, whether or not a recurrence is in the field of previous RT and the organ/tissue tolerance in the region potentially requiring re-irradiation [139]. In the setting of the relapse of localised RMS, the absence of RT during primary treatment (for example in very young patients) not only makes this a favoured local control option where feasible but is also an independent positive prognostic indicator [142]. The current thinking on local control for relapsed RMS is described in the recently published EpSSG treatment recommendations for RMS [160].

3.3.5. Outcomes of Relapsed RMS

Several features of relapsed RMS may have an impact on survival outcome. In an analysis of localised (M0) patients from the European SIOP-MMT trials who had completed treatment, had CR/stable residual for at least 6 months from the end of treatment and at least 3 years of follow-up from the last event for alive patients, 37% of patients were alive at >3 years from their first relapse [142]. Outcomes following the relapse of metastatic RMS are extremely poor [82,161]. Patients whose disease recurs at distant metastatic sites have an abysmal prognosis with a reported 5-year post relapse OS of 0–10% [12,82,113,142,144,145]. A better prognostic group among relapsed patients includes those with local or locoregional recurrences [141]. Patients with a large tumour (>5 cm in maximal diameter) at relapse have also been identified as a poorer prognostic group [146]. The time to progression has been associated with time outcome in relapsed patients. Like several of the preceding features, this can also be described as a spectrum [139]. Patients whose disease recurs while on treatment or those whose disease progresses through treatment carry the worst prognosis (5-year OS 2–8%), while prognosis for those who relapse following a period of clinical remission varies according to the time to recurrence [157]. The four-year post relapse survival was 12%, 21%,and 41%, respectively, for those who relapsed <6 months, 6–12 months and >12 months, respectively, after the completion of initial therapy [162].

1.Relapse of Initially Localised RMS

Several groups have studied outcomes in patients who relapse after initial presentation with localised RMS [12,82,113,142,144,145]. Although in the early IRSG studies patients with both localised and metastatic disease were included in the analyses, given that the prognoses are so different, relapsed patients with initially metastatic disease should be separately analysed to determine the best prognostic features to risk stratify this patient group.

An analysis of pooled data from the Italian studies RMS 79, 88 and 96 identified alveolar subtype, parameningeal or other site, metastatic recurrence and on-treatment recurrence as independent risk factors for poor survival following relapse [144]. On the basis of this analysis, the authors developed a risk stratification model predicting post relapse 5-year OS which was 16% in patients with two risk factors, while no patients with three or more risk factors survived [144].

In the setting of initially localised RMS, tumour-related factors (metastatic relapse, large size, unfavourable site, nodal involvement at initial diagnosis, alveolar histology, time to relapse < 18 months) and treatment-related factors (prior RT and chemotherapy agents used) adversely influence the chance of a cure with second-line treatment [142]. An analysis completed by Chisholm et al. used a weighted scoring system to generate a nomogram, allowing an estimation of the chance of salvage following the first relapse of localised RMS [142].

The COG pooled data from IRS III-IV identified histologic subtype as strongly correlating with post relapse survival [12]. Among patients with alveolar RMS, the only good prognostic factor was favourable site (orbital, paratesticular or vaginal primary) while for patients with embryonal RMS, stage, pattern of relapse and chemotherapy received were found to have prognostic significance [12].

Pooled data from multiple CWS studies (CWS-81, CWS-86, CWS-91 and CWS-96) found age, histology, tumour size, tumour site, post surgical stage and the omission of RT to be factors associated with an increased risk or recurrence [145].

2.Relapse of Initially Metastatic RMS

Although few patients with metastatic RMS appear to be salvaged following relapse, a small proportion do survive. A three-year OS of 10% was reported by the CWS group in 129 patients with the relapse of metastatic RMS, and the COG reported 12% 5-year OS in relapsed metastatic embryonal RMS and 3% in patients with group II, III and IV alveolar RMS [12,161]. An analysis of pooled data within INSTRuCT looking at survival after first recurrence in initially metastatic patients has recently been completed (In press).. Randomised trials of relapsed RMS have historically included the relapse of both metastatic and localised disease: in the future, the further refinement of risk factors for patient stratification may be considered.

4. Looking Forward—Advances in Developmental Diagnostics and Therapeutics

4.1. Diagnostics

Diagnostic radiology is a constantly developing field with the implementation of techniques such as whole-body diffusion-weighted MRI with background suppression (DWIBS) and FDG-PET/MRI. In recent years several potential game changers have been introduced. First, there was the advent of photon counting computed tomography, a technique with a higher contrast-to-noise ratio, improved spatial resolution and optimised spectral imaging compared with conventional CT [163]. This can impact diagnostics due to a higher resolution, e.g., in head and neck RMS where osseous involvement may be detected earlier. Perhaps more exciting is the implementation of novel contrast agents and their potential use in molecular imaging [164]. A development in the field of nuclear medicine, Long Axial Field of View (LAFOV) PET-CT, makes it possible to image the whole body in a single field of view with higher resolution images, decreased radiation exposure and shorter scan times [165,166]. In a study of 10 children with neuroblastoma, scan times were as short as 10 min and none of the children required sedation for the scan [167]. Finally, the introduction of ^68^Ga-FAPI PET/CT is based on evidence that the stroma of most sarcomas overexpresses Fibroblast Activation Protein (FAP), a type II transmembrane serine protease [168,169]. Using FAP Inhibitors (FAPIs) as a radiotracer may lead to a higher diagnostic sensitivity of PET imaging. Preliminary studies with a small number of patients have shown that in a head-to-head comparison ^68^Ga-FAPI PET/CT outperformed 18FDG PET/CT in adult patients [170]. If indeed it is as selective as it seems to be, this new tracer could also be used for FAP-targeted radioligand therapy in the future. The classic radiologic size response to neoadjuvant chemotherapy has not shown to be prognostic for outcome [171,172]. Currently there is no early surrogate outcome measure in RMS. Therefore, in the FaR-RMS study, diffusion-weighted MRI response (both in the primary and relapsed setting) and ^18^FDG-PET/CT response (in the primary setting) are prospectively being investigated as potential early surrogate outcome measures [4].

The potential of new diagnostic techniques, some of which are still experimental, must be evaluated in prospective studies, where important questions around the potential of the upstaging of patient, and the subsequent risk of overtreatment, should be explored.

4.2. Minimally Invasive Treatments/Interventional Oncology

In the adult oncology setting, the minimally invasive treatment of metastatic disease using techniques such as trans-arterial chemoembolization (TACE), thermal ablation, cryotherapy and irreversible electroporation (IRE) are widely used in the palliative phase of care [173,174,175]. The recent advances in interventional oncology, a sub-discipline of interventional radiology, necessitate its inclusion in multimodality treatment and has been coined the fourth pillar of oncology [176]. Even though there have been early adopters of some of these approaches, the general uptake of therapeutic interventions in paediatric oncology has been relatively slow [177,178,179]. The implementation of interventional radiological techniques within RMS treatment will require collaboration with paediatric interventional radiologists and the inclusion of new techniques in prospective clinical trials like FaR-RMS.

4.3. Detecting Relapse: Current and Future Paradigms

Routine surveillance with a combination of the cross-sectional imaging of the primary site and/or metastatic sites and chest x-ray is recommended for up to 5 years following the completion of treatment, with most frequent surveillance in the initial two years after treatment. A comprehensive European imaging guideline for RMS has recently been published [171,180]. A study of surveillance imaging in localised RMS showed that the majority (61%) of relapses were detected at a time of clinical symptoms and there was no survival advantage for patients whose asymptomatic relapse was detected by routine surveillance [147]. At present, there is no role for FDG-PET imaging in the setting of routine surveillance imaging and it is only performed as part of re-staging once relapse has been confirmed.

Cell-free DNA (cfDNA), small DNA fragments derived from dead cells found in the blood and other body fluids, may arise from tumour shedding and have potential future utility as biomarkers in solid tumour patients. A surrogate for repeat tumour biopsies, these liquid biopsies allow for the detection of PAX3(7)::FOXO1 in FP cases and other variants of interest which can be detected in FN RMS. The use of minimally invasive techniques to detect recurrence earlier (routine blood sampling to assess for circulating tumour DNA (ctDNA) or Circulating Tumour Cells (CTCs) for example) is currently under investigation [181,182]. ctDNA levels correlate with the clinical and radiologic response in pilot studies and will be investigated further in a cohort within the FaR-RMS study [4]. If this approach proves feasible and laboratory assays can be harmonised, it holds the possibility of the earlier detection of relapsed disease and may identify patients with the most potential for benefit from enrolment in early-phase trials at an earlier time in the patient journey. The prospective collection of ctDNA for the purposes of monitoring treatment response, assessing utility as an earlier biomarker of recurrence than conventional imaging, and providing information about the genetic makeup of a tumour and how this evolves over time may have implications on how disease monitoring occurs in the future [181,183,184,185,186]. The prognostic value of liquid biopsies in an IR RMS study has recently been published by the COG and both patients with FN and FP RMS with detectable ctDNA at diagnosis had poorer outcomes than those without [138].

4.4. Novel Therapies

As more genomic and molecular data from patients with RMS becomes available, particularly from patients enrolled on trials, we will be able to couple this data with patient outcomes and develop a more in-depth understanding of biologic subtypes. Re-biopsy at the time of relapse can potentially provide information on novel genomic alterations and tumour evolution. There is also great potential in the use of liquid biopsies to monitor disease burden and identify molecular biomarkers of predictive or prognostic value. The timely identification of therapeutic targets is not only important to determine eligibility for ongoing early-phase trials but also has the promise of providing insight into disease resistance mechanisms that must be better understood as we aim for improved patient outcomes in relapsed patients. Tumour-agnostic trials investigating the targeting of ALK, HDAC, RAS/PI3K, FGFR4, CXCR4, IGFR1, MET and PDGFRA have so far not given rise to data that would currently support the use of a targeted agent as monotherapy at second or subsequent relapse so combined testing with traditional systemic chemotherapy is needed in future studies as preclinical work to support this continues [187,188]. Table 5 lists completed phase I and II trials in which patients with relapsed RMS were eligible.

4.5. Immune Checkpoint Inhibition

Immunotherapy in the form of immune checkpoint inhibition (ICI) has thus far not shown impressive responses in paediatric embryonal tumours. Whilst the reasons for this are incompletely understood, several factors probably contribute. TME in paediatric sarcomas is highly immunosuppressive, attributed to regulatory T cells and myeloid-derived suppressor cells, and this may curtail the efficacy of ICI [189]. The lack of a high mutational burden in many paediatric and TYA cancers, which is particularly relevant to FP RMS, suggests many of these patients lack natural adaptive immunity. It remains to be seen however if there is a differential response between FP and FN RMS (typically with a relatively higher mutational burden than FP RMS) when treated with ICI. Finally, it is possible that a more incompletely developed immune system in younger children may play a contributing role.

4.6. Engineered T Cell Therapy

Due to the lack of an endogenous immune response to a tumour, an alternate strategy is engineering patient T cells to be redirected to the cancer. This can be in the form of viral transductions with a transgenic T cell receptor or with a chimeric antigen receptor to develop chimeric antigen receptor-T (CAR-T) cells. Tumour antigens may be products of mutated genes, over expressed or aberrantly expressed cellular proteins, altered cell surface glycolipids or glycoproteins and cell-type-specific differentiation antigens [190]. These antigens may be present on both tumour cells as well as normal cells, albeit to a lesser extent, causing “on-target off-tumour toxicity” when treated with CAR-T cells. A biochemical and metabolic milieu unfavourable to T cell effector function combined with immunosuppressive signals from cancer and stromal cells are important factors that likely impact engineered T cell efficacy for solid tumours such as RMS. T cell activation without exhaustion, expansion, persistence and memory cell development has the potential to provide long-lasting defence against recurrent disease. Although there have been several well described barriers to the use of CAR-T therapy in solid tumours, the engraftment of CAR-T cells can be improved via the use of lymphodepleting conditioning regimens and improvements in CAR-T cell persistence are being investigated through additional engineering to enhance persistence [191].

There are several currently ongoing clinical trials with CAR-T cell therapy in solid tumours in children and young adults and although neuroblastoma and CNS tumours are leading in this frontier, CAR-T cells targeting HER2, GD2, B7H3, PSMA, CD171, GPC3, IL13Ra2, EGFR806, CD55V6 and CD276 have been trialled in children and young people and future targets in RMS may include FGFR4, SLC19A1, EPHB4 and ACVR2A. The knowledge of target antigens, co-stimulatory molecules, safety mechanisms and safety profiles will be important in understanding the utility of CAR-T cells for relapsed patients with RMS.

4.7. Antibody Drug Conjugates

Antibody drug conjugates (ADCs) are a class of therapeutics, which although administered systemically at a high concentration of drug, have their efficacy enhanced and toxicity diminished through the targeting of highly expressed tumour specific antigens which allow a specific antibody to deliver a covalently linked drug payload [192]. Many of the targets identified for CAR-T development are also attractive ADC candidates for RMS. As this class of agents expands, the aim of ADC delivery to tumour/metastatic sites in a selective fashion is paramount and holds the promise of an improved therapeutic index [192]. Factors which will need to be overcome in the setting of relapsed sarcomas in order for ADCs to advance in the clinic include the identification of tumour-specific target antigens and the optimisation of delivery to heterogenous tumours impacted by a myriad of variables within the TME.

4.8. Therapeutic Cancer Vaccines

Therapeutic cancer vaccines have thus far failed to show benefits in most paediatric malignancies [193] and RMS is no exception, specifically with attempts to therapeutically exploit the PAX3(7)::FOXO1 breakpoint region., Further attempts in this field are impeded by limited HLA class I or class II antigen–peptide binding [194]. The currently ongoing PerVision study (NCT06094101) is a multicentre, open-label, phase I/II feasibility and early proof of concept study open to the recruitment of patients with metastatic sarcomas that have a fusion. The trial enables the delivery of a bespoke sarcoma-specific individualised cancer peptide vaccine to patients.

5. Conclusions

We have described the clinical and biological features of the poorest risk subgroups in patients with RMS and summarised the attempts to date to improve outcomes. Further research is needed to identify new robust biomarkers that can help define the poorest risk RMS subgroups and to progress our understanding of factors contributing to aggressive biology. It will be essential to explore how to pharmacologically and immunologically exploit vulnerabilities, validate predictive models of relapse risk and develop mechanisms to counter resistance. Ultimately there is a need to make significant advances in systemic treatments for these patients while simultaneously continuing to improve local therapies.