Detection of circulating tumor-derived material in peripheral blood of pediatric sarcoma patients: A systematic review

Abstract

Background

Detection of circulating tumor-derived material (cTM) in the peripheral blood (PB) of cancer patients has been shown to be useful in early diagnosis, prediction of prognosis, and disease monitoring. However, it has not yet been thoroughly evaluated for pediatric sarcoma patients.

Methods

We searched the PubMed and EMBASE databases for studies reporting the detection of circulating tumor cells, circulating tumor DNA, and circulating RNA in PB of pediatric sarcoma patients. Data on performance in identifying cTM and its applicability in diagnosis, and evaluation of tumor characteristics, prognostic factors, and treatment response was extracted from publications.

Results

A total of 79 studies were assigned for the present systematic review, including detection of circulating tumor cells (116 patients), circulating tumor DNA (716 patients), and circulating RNA (2887 patients). Circulating tumor cells were detected in 76% of patients. Circulating DNA was detected in 63% by targeted NGS, 66% by shallow WGS, and 79% by digital droplet PCR. Circulating RNA was detected in 37% of patients.

Conclusion

Of the cTM from Ewing's sarcoma and rhabdomyosarcoma ctDNA proved to be the best target for clinical application including diagnosis, tumor characterization, prognosis, and monitoring of disease progression and treatment response. For osteosarcoma the most promising targets are copy number alterations or patient specific micro RNAs, however, further investigations are needed to obtain consensus on clinical utility.

Introduction

Globally, pediatric sarcomas of bone and soft tissue accounts for approximately 12% of all childhood cancers and 21% of solid malignancies in patients aged 0–19 years [1]. The most common types of pediatric sarcomas include Ewing's sarcoma (ES), osteosarcoma (OS), and rhabdomyosarcoma (RMS). Since the introduction of chemotherapy, the 5-year overall survival has improved from 34 to 49% in the period 1973–1982 to 57–74% in the period 1993–2015 [2], [3], [4], [5], [6]. Pediatric sarcoma patients receive multimodal treatment with a combination of chemotherapy as well as surgery and radiotherapy when applicable [7], [8], [9], [10]. Survival rates are significantly poorer in patients with disseminated metastatic disease and the salvage rate in patient relapse is low [8,9,[11], [12], [13]].

Diagnosis of pediatric sarcomas includes molecular and genetic tests of tumor biopsies and imaging with magnetic resonance imaging (MRI), computed tomography (CT), and/or ¹⁸F-fluorodeoxyglucose positron emission tomography/CT (¹⁸F-FDG-PET/CT), the latter seemingly having best predictive value [14,15]. Therapy response and disease progression are currently assessed using these methods, however, reduction in tumor size following chemotherapy does not significantly correlate with tumor viability [12,[16], [17], [18], [19], [20]]. This implies that medical imaging may not be the optimum method.

Detection of circulating tumor-derived material (cTM) in liquid biopsies based on the distinct genetic profile of the tumor could be a less intrusive and superior method to diagnose, characterize, and especially monitor treatment response and disease progression. cTM detectable in the peripheral blood (PB) of cancer patients include tumor cells (CTCs), tumor DNA (ctDNA), RNA including tumor RNA (ctRNA) and microRNA (miRNA), as well as nucleic acid in extracellular vesicles (EVs). Currently, cTM is used for tumor detection and characterization, disease monitoring and prognosis in adult solid cancers including colorectal cancer, breast cancer and melanoma [21], [22], [23], [24].

This systematic review investigates current methods for detecting tumor-derived cells and nucleic acid in PB from patients with pediatric sarcoma and its applications in diagnosis and monitoring of disease.

Material and methods

Systematic review approach

We performed a systematic literature search in the PubMed and EMBASE databases. The terms used in the literature strategy are listed in Supplementary File 1. Literature was systematically reviewed in accordance with the Preferred Reporting Items of Systematic reviews and Meta-Analyses (PRISMA) guidelines [25,26]. The PRISMA checklist is available in Supplementary File 2. Preceding full text review, initial screening was carried out based on titles and abstracts. Relevant and qualified studies were independently selected by two investigators.

Eligibility criteria

Studies were considered eligible if they met all the following criteria: (1) analyzed cTM in human PB specimens, including serum, plasma, and exosome vesicles; (2) compared patients with definitive, histopathologically verified diagnoses of OS, ES, or RMS with healthy control subjects; (3) evaluated pediatric patients and/or young adults defined as ages 0–39 years; and (4) written in English language. Studies were excluded if they met at least one of the following criteria: (1) studies not investigasting cTM in human PB; (2) studies evaluating cell lines, tissues and/or xenograft animal models only; (3) studies with insufficient or unqualified data including data not specific to either OS, ES or RMS patients; (4) non-original articles or articles published in the form of reviews and editorials or (5) duplicated articles.

Data Extraction

Data extraction was carried out to collect information on study design, patient characteristics, methodology, and findings. We retrieved the following data from all included studies; article year and author(s), study design including total number of cases and controls, patient characteristics, methodology as well as statistical tests and findings of the studies. Patient characteristics extracted included age, gender, country of residence, sarcoma subtype (OS, ES, or RMS), tumor stage, and treatment regimen. Methodology information included sample type (total blood, plasma, serum, or exosome vesicles), time(s) of sampling (e.g., at diagnosis, during treatment, after remission), assay used to investigate nucleic acid in PB and, if applicable, the targets of the assay.

Risk of bias assessment

The risk of bias (RoB) in the individual included studies were assessed using a combination of the Cochrane risk-of-bias tool for randomized trials (RoB 2) and the Risk of Bias in Non-Randomized Studies of Exposures (ROBINS-E) [27,28]. Further, the RoB was evaluated across studies in each methodology category.

Data synthesis and meta-analysis

The studies were grouped by sarcoma subtype and methodology in order to carry out data synthesis. Within each group, detection percentages of cTM across studies were calculated and the methodology was evaluated for specific sarcoma subtypes. In addition, data from two studies investigating CTCs in RMS patients were combined [29,30]. Unpaired t-tests were carried out to compare mean CTC level of RMS patients with localized vs metastasized disease using in R version 4.1.0 [31].

Results

A total of 981 articles were acquired from the literature search on PubMed and Embase and an additional two articles were identified through other sources as shown in Fig. 1. After the removal of 34 duplicates, 949 articles were screened of which 834 studies were deemed irrelevant based on the eligibility criteria. Of the 34 duplicates, 21 studies were included in this systematic review [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52]. Full text screening of the remaining 115 articles resulted in 79 studies assessed eligible for qualitative synthesis and listed in Table 1. For a detailed summary of the findings of all included studies refer to Supplementary Table 1 and the results of the individual RoB analysis is presented in Supplementary File 3. Overall, the studies that investigated CTCs presented the lowest risk of bias, whereas studies investigating ctDNA by NGS posed some concern for risk of bias.

Fig. 1.

PRISMA flowchart.

Table 1.

Characteristics of included studies grouped by technology and target.

β3-AR: β3-adrenergic receptor; circRNA: circular RNA; CK: Cytokeratin; CSV: Cell surface vimentin; CTC: Circulating tumor cell; ctDNA: Circulating tumor DNA; ctRNA: Circulating tumor RNA; DAPI: 4′,6-diamidino-2-phenylindole; ddPCR: digital droplet polymerase chain reaction; ES: Ewing's sarcoma; EV: Extracellular vesicle; FACS: Fluorescence-activated cell sorting; fPCR: Fluorescence polymerase chain reaction; lncRNA: Long noncoding RNA; miRNA: Micro RNA; mRNA; Messenger RNA; NGS: Next generation sequencing; OS: Osteosarcoma; PBMC: peripheral blood mononuclear cell; PBSC: peripheral blood stem cell; PCR: Polymerase chain reaction; qPCR: Quantitative polymerase chain reaction; qRT-PCR: Quantitative reverse transcription polymerase chain reaction; RMS: Rhabdomyosarcoma; RT-PCR: Reverse transcription polymerase chain reaction; sEV: Small extracellular vesicle; sqRT-PCR: Semiuantitative reverse transcription polymerase chain reaction; sWGS: Shallow whole genome sequencing; TLDA: TaqMan low density array; Ø: Diameter.

* Diagnostic value is determined by creating receiver operating characteristics (ROC) curves and calculating the sensitivity (Sen), specificity (Spe), and the area under the curve (AUC).

** Percentage successfully distinguished from other pediatric cancer types.

*** Percentage of patients with consensus between CNA of tumor tissue and cfDNA.

† Duplicated studies.

Detection of tumor-derived cells and nucleic acid in peripheral blood

Circulating tumor cells

CTCs of pediatric sarcoma patients were investigated by microfiltration and flow cytometry using varying approaches in five studies [29,30,[53], [54], [55]]. In total these studies included 116 patients with active ES (n=39), OS (n=39) or RMS (n=38). CTCs were detected in 76% of patients (Table 1).

All five studies defined CTCs as CD45- PB cells of varying sizes and/or carrying varying surface markers. Through negative staining for the lymphocyte biomarker CD45, regular white blood cells were eliminated, and of the remaining cells, a larger fraction are likely to be tumor derived. Three studies used cell surface vimentin (CSV) and cell diameters of >7, >10 and ∼14 μm, respectively, to distinguish CTCs from other CD45- cells [29,54,55]. By this method, Dao et al. showed a sensitivity of 75% and specificity of 85.3% of CTCs (≥0.17 CTCs/mL PB) to detect sarcoma [54]. Calvani et al. defined CTCs as CD99+ CD45- BP cells, and further evaluated the expression of β3-adrenergic receptor (β3-AR) on the surface of these cells to investigate its potential as a prognostic marker for ES [53]. Tombolan et al. defined CTCs as DAPI+ CK+ EpCam+ CD45- BP cells with a diameter > 4 μm, and further implemented capturing based on desmin, finding this increased the identification of CTCs [30].

In addition to CTC enumeration, the molecular characteristics of the CTCs were investigated by analysis of RNA extracted from CTCs by reverse transcription polymerase reaction (RT-PCR) and single cell sequencing [29,55]. Based on single cell sequencing of cells isolated with CellSieve from one patient, Hayashi et al. found that 10% of these cells represented CTCs [29].

Circulating tumor DNA

Overall, 23 studies investigated circulating tumor DNA (ctDNA) from a total of 716 patients with ES (n=442), OS (n=180) or RMS (n=66) (Table 1) [30,32,33,[56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73]].

The ctDNA was identified and analyzed in cell-free DNA (cfDNA) by next generation sequencing (NGS) in ten studies including 412 patients in total [32,56,57,[67], [68], [69], [70], [71], [72], [73]]. The NGS methods included targeted NGS (tNGS) panels in seven studies [56,57,67,69,[71], [72], [73]], and shallow whole genome sequencing (sWGS) in three studies [32,68,70]. cfDNA extraction was carried out using the QIAamp Circulating Nucleic Acid Kit (Qiagen, Germany), Avenio ctDNA expanded kit (Roche, Switzerland) or the Maxwell RSC LV cfDNA kit (Promega, USA) (Table 1). To detect copy number alterations (CNAs), one study carried out low-pass whole genome sequencing (LP-WGS) and two studies carried out ultralow passage whole genome sequencing (ULP-WGS) in addition to tNGS [56,67,73]. Overall, the seven studies that carried out tNGS identified ctDNA in 63% of the patients and found >80% concordance between alterations in tumor DNA identified in PB and tumor tissue. The three studies that carried out sWGS identified ctDNA in 66% of 155 unique patients through subsequent CNA analysis as wells as analysis of fragmentation patterns in one study. High concordance was seen between tumor tissue and cfDNA and misclassifications were mainly associated with sample quality rather than disease extent. Henson et al. investigated ctDNA through a C-circle DNA assay including 7 patients, and were able to detect high levels of DNA C-circles in four patients [66].

Twelve studies analyzed ctDNA by polymerase chain reaction (PCR) methods including 310 patients in total [57,68,73,33,[58], [59], [60], [61], [62], [63], [64], [65]]. The PCR methods included digital droplet PCR (ddPCR), quantitative PCR (qPCR) and quantitative reverse transcription PCR (qRT-PCR). Ten studies including PB from 284 patients in total used ddPCR to detect ctDNA in 79%, and when excluding one study investigating OS, ctDNA was detected by ddPCR in 90% of 212 patients with ES or RMS. In order to quantify ctDNA using ddPCR or qRT-PCR, nine studies identified the individual genomic profiles of the tumor tissue, e.g. the EWSR1 fusion sites of ES patients or the PAX3/PAX7 fusion sites of RMS patients, before carrying out the PCR-based method [33,57,58,[61], [62], [63], [64],68,73]. Lyskjaer et al. detected ctDNA in OS patients through methylation-specific ddPCR, selecting and validating four CpG markers [59]. Van Zogchel et al. used ddPCR to detect hypermethylated RASF1A ctDNA in RMS patients [60]. Bodlak et al. report that ddPCR-based ctDNA detection is superior to ctRNA detection, detecting ctDNA in 100% of included patients and ctRNA in only 80% [33]. Two studies used other methods than ddPCR. Yu et al. used qPCR to determine the copy number of cell-free mitochondrial DNA (cf mtDNA) in order to detect cancer based on low levels [65]. Eguchi-Ishimae et al. detected EWSR1 fusions by qRT-PCR in one ES patient [62]. Four of the studies that used PCR-based methods compared ctDNA findings to medical imaging including MRI, CT, and ¹⁸F-FDG-PET/CT [58,61,62].

Circulating tumor RNA

By PCR-based methods, a total of 55 studies listed in Table 1 investigated RNA in PB from 2887 patients with ES (n=360), OS (n=2444) or RMS (n=83). The studies investigated circulating tumor RNA (ctRNA) by targeting specific gene transcription products, expression levels of specific genes as well as microRNA (miRNA), long non-coding RNA (lncRNA), and RNA in extracellular vessels (EVs).

In total, sixteen studies investigated the expression of tumor-derived or tumor-associated genes in 337 patients. The majority of studies used RT-PCR, however some used semiquantitative RT-PCR (sqRT-PCR), digital PCR (dPCR) and ddPCR [33,74,75]. Wong et al. distinguished patients from controls by investigating the expression levels of specific genes including COLL by sqRT-PCR [74]. The remaining fifteen studies detected ctRNA in 37% of 326 patients (62 known case duplicates). The ctRNA was detected by targeting fusion transcription products related to ES and RMS, i.e., EWSR1-FLI1, EWSR1-ERG, PAX3-FKHR, PAX7-FKHR, and EWS-WT1^33,75–87, and/or transcripts of genes associated with RMS including myogenin, MyoD1 and AChR [78,79,88]. Five of the studies investigated stem cells, mononuclear cells, or progenitor cells in PB^8485777479. Bodlak et al. compared ctRNA findings to medical imaging with ¹⁸F-FDG-PET/CT³³.

Another approach is the use of quantitative RT-PCR (qRT-PCR) to investigate the differential expression and/or release of miRNAs by solid tumors. 33 studies investigated miRNA in the serum or plasma of 2293 patients (miRNAs are listed Supplementary Table 2) [[34], [35], [36], [37], [38], [39], [40], [41], [42],44,[46], [47], [48], [49], [50], [51], [52],[89], [90], [91], [92], [93], [94], [95], [96], [97], [98], [99], [100], [101], [102], [103], [104]]. In addition to qRT-PCR, one study used ddPCR [42] and three studies used TaqMan low density array (TLDA) [38,48,51]. The diagnostic value of one or more miRNAs were often evaluated by calculating area under the curve (AUC) at a predefined cutoff level of receiver operating characteristics (ROC) curves differentiating miRNA expression levels of patients and controls. Six studies used medical imaging including MRI, CT, and ¹⁸F-FDG-PET/CT to determine tumor size, treatment response, and/or relapse [34,89,93,96,97,103].

Finally, five studies explored RNA in circulating EVs and one study investigated lncRNA by qRT-PCR including a total of 257 patients [43,45,[105], [106], [107], [108]].

Applications of liquid biopsies in pediatric sarcoma

The various methodologies investigating cTM including CTCs, DNA and RNA proved to have potential of application in a clinical setting. As overviewed in Fig. 2, the included studies suggest that current technologies can be used to: (1) diagnose pediatric sarcoma; (2) distinguish patients from controls and patients with other cancers; (3) determine tumor characteristics including tumor location and histological subtypes; (4) differentiate patients based on prognostic factors such as tumor size and metastatic status; and (5) monitor disease and treatment response.

Fig. 2.

Overview of circulating tumor-derived material and its clinical application

CTC: Circulating tumor cells; ctDNA: Circulating tumor DNA; ctRNA: Circulating tumor RNA; ES: Ewing's sarcoma; EV: Extracellular vesicle; ddPCR: Digital droplet polymerase chain reaction; FCM: Flow cytometry; MF: Microfiltration; miRNA: MicroRNA; NGS: Next generation sequencing; qRT-PCR: Quantitative reverse transcription polymerase chain reaction; RT-PCR: Reverse transcription polymerase chain reaction; TLDA: TaqMan low density array. Created with BioRender.com.

Ewing sarcoma

The analyses of cTM have primarily been focusing on detection of EWSR1-FLI1 fusions or fusion transcripts which can aid the diagnosis of ES. CTCs have been detected at time of diagnosis in 79% (n=80) of the analyzed patients in three studies [29,53,54]. Further, ctDNA was detected at time of diagnosis by tNGS in 63% (n=115) of patients included in six studies [56,57,67,69,71,73], by sWGS in 77% (n=110) of patients included in three studies [32,68,70], and in 94% (n=190) in eight studies [33,57,58,61,63,64,68,73]. The NGS-based methods did not require somatic mutations or CNAs and included detection of TP53 mutations and CNAs of TP53 and STAG3. The use of ctRNA by RT-PCR only detected fusion transcripts in 32% (n=285) of patients included in eleven studies [33,[75], [76], [77],80,[82], [83], [84], [85], [86], [87]]. However, Samuel et al. investigating RNA in EVs detected cTM in 70% (n=10) of the patients [106]. Patients were distinguished from controls by CTC enumeration, ddPCR targeting ctDNA and differential expression profiling of miR-125b [54,65,96]. Further, ES could be distinguished from other cancer types including OS and RMS based on epigenetics and differential expression of miRNAs [49,69]. The genetic abnormalities detected in ctDNA correlated well with genetic alterations observed in tumor tissue [67,69].

The absence of ctDNA at diagnosis was associated with improved overall survival, and increased ctDNA levels correlated with increased risk of death [56,58,68]. Further, many studies were able to distinguish ES patients based on prognostic factors including tumor size, metastatic status, and risk of relapse. The presence and/or increased levels of ctDNA correlated with larger tumor size [33,58,61,68], and ctRNA was detected significantly more frequently in patients with larger tumors of >200 ml compared to those with small tumors [86]. Metastatic status correlated with CTC levels and elevated expression of β3-AR on CTCs [29,53]. Higher levels of ctDNA were detected among patients with metastatic disease [33,57,58]. Krumbholz et al. detected ctDNA in all but two patients who had small, localized tumors that had been resected prior to plasma sampling [64]. Pfleiderer et al. investigated ctRNA in five patients with metastatic disease and five patients with localized disease, and found that only one patient with metastatic disease at time of diagnosis had detectable fusion transcripts [80]. The detection of ctDNA could be used to estimate relapse probability at diagnosis [68], and the relapse rate was greater in the group with detected ctRNA at last follow up than the group with no detection at last follow-up [87]. It was possible to predict the course of disease based on analysis of cTM as aggressive course of disease correlated with elevated β3-AR expression on CTCs [53], and high ctRNA levels [75].

Compatibility between samples of bone marrow and PB was observed in 12 of 16 collection time points during follow-up [87]. This suggest that disease progression correlates with cTM levels, and it has been proved possible to monitor disease progression after diagnosis by analysis of BP. Calvani et al. reported elevated expression of β3-AR observed on CTCs derived from relapsed patients [53]. Yaniv et al. showed a correlation between levels of tumor cells in the blood (harvested PBSC) and relapse after transplantation [77]. In addition, increasing ctDNA levels in cases with disease relapse have been observed in multiple studies [61,64,71].

Finally, detection of cTM has been used to monitor treatment response. During initial chemotherapy, levels of ctDNA have been shown to rapidly decline, often to undetectable levels [63,64,69,73], and at end of therapy ctDNA is rarely detected [29,69,73]. Shah et al. reported that previously undetectable ctDNA increased considerably after radiotherapy, however, ctDNA could not be detected at end of therapy [69]. Levels of ctDNA decreased with good response to treatment determined by medical imaging including ¹⁸F-FDG-PET/CT, MRI and CT [58,61]. Further, ctDNA has been used to assess treatment induced toxicity and organ damage [68]. Zoubek et al. proposed that tumor cells are mobilized during surgical biopsy procedures. They detected fusion transcripts in PB during open surgery, however not at time of diagnosis and six days after surgery [82]. Grohs et al. detected ctRNA before and after surgical biopsy suggesting that dissemination of tumor cells was not prevented by preoperative treatment with chemotherapy [76]. In a study suggesting miR-125b as a diagnostic marker for ES due to observed low levels, Nie et al. observed significant downregulation in patients who presented poor response to chemotherapy [96].

In conclusion, monitoring of ctDNA is the most promising biomarker and should be considered as an add-on tool in the clinical care of ES patients. Additionally, ddPCR was best at detecting ctDNA at diagnosis (94%, n=190), and both ddPCR and tNGS proved useful in the monitoring of disease progression and treatment response.

Osteosarcoma

Detection of cTM at diagnosis in OS patients has been based on SNVs or CNAs of genes including TP53, ATRX, FUS, and STAG2, as well as differential expression of miRNAs. CTCs were detected at diagnosis in 62% (n=34) of patients included in three studies [29,54,55]. Detection of ctDNA was successful by tNGS in 65% (n=96) of patients included in five studies [56,69,[71], [72], [73]], by sWGS in 91% (n=22) of patients included in three studies [32,68,70], and by ddPCR in 44% (n=72) of patients included in one study [66]. ctRNA was detected by RT-PCR in 91% (n=11) of patients included in one study [74]. Differential expression of various miRNAs listed in Supplementary Table 2 has been widely used to validate specific miRNAs as diagnostic markers of OS and to distinguish OS from other cancers.

Characterization of OS tumors involves determining histological subtypes which has been possible through qRT-PCR analysis of specific miRNAs levels.

The prognosis of OS patients correlates with ctDNA, as increasing levels has been shown to increase the risk of death [56]. Further, detection of ctDNA defined as methylations of ≥2/4 carefully selected CpG sites independently correlated with overall survival [66]. Moreover, metastatic status which is a known prognostic factor of cancer correlates with the levels of CTCs and ctRNA. Hayashi et al. observed a higher number of CTCs in patients with metastatic disease [29]. Wong et al. showed that substantially raised levels of COLL mRNA levels in CTCs increased the risk of metastasis and that the levels correlated with the number of months for a secondary tumor to develop [74].

One study detected ctDNA in patients at relapse suggesting that this may be useful in monitoring disease progression [71]. An assay of circulating DNA C-circles proved to be specific to alternative lengthening telomeres and responsive to the activity of these [29]. Further, ctDNA levels has been shown to decrease at initiation of chemotherapy, and treatment response assessed by extent of tumor necrosis correlates with ctDNA levels during treatment [69,73].

Finally, analysis of differential expression of miRNAs has proved to be useful to determine prognostic factors including metastatic status and tumor size, predict patient survival, and monitor disease progression and treatment response.

In conclusion, detection of cTM at diagnosis was most frequent when targeting either ctDNA by sWGS or ctRNA by RT-PCR. Prognostic factors correlated with both CTC, ctDNA and ctRNA levels. Both ctDNA and miRNA proved useful in the monitoring of disease progression and treatment response.

Rhabdomyosarcoma

The analysis of cTM can be used as a diagnostic tool for RMS and has been based primarily on detection of PAX fusions or fusion transcripts as well as genetic abnormalities of genes including TP53, PIK3CA, RASF1A, MyoD1, and Myogenin.

CTCs has been detected in 71% (n=38) of patients included in three studies [29,30,54]. By single CTC RNA sequencing, one of these studies found that 10% of cells isolated with microfiltration represented CTCs [29]. At time of diagnosis, ctDNA was detected by tNGS in 82% (n=17) of patients included in three studies [67,69,73], by sWGS 74% (n=39) of patients included in three studies [32,68,70], and by ddPCR 74% (n=21) of patients included in four studies [60,62,68,73]. One of the studies using ddPCR were able to show higher levels of cfDNA in RMS patients compared to other childhood cancers such as neuroblastoma, renal tumors, and CNS tumors [60]. RT-PCR was used to detect ctRNA at diagnosis in 50% (n=42) of patients included in five studies [78,79,81,85,88]. Four studies observed good diagnostic value of varying miRNAs [42,49,89,90]. Further, when comparing analyses of tumor tissue and PB, concordance was observed with both ctDNA and ctRNA [67,69,88].

Characterization of RMS tumors is based on the genetic profile including fusion status and histological subtypes. Peneder et al. were able to distinguish alveolar from embryonal RMS by observed reduction of cfDNA fragment coverage at specific DNase I hypersensitive sites in alveolar RMS [68]. Further, principal component analysis based on differential expression of miRNAs showed a clear separation between RMS subtypes [42].

RMS patients has been distinguished based on prognostic factors including tumor size and metastatic status. Eguchi-Ishimae et al. reported that the ctDNA levels reflected tumor size determined by ¹⁸F-FDG-PET/CT imaging [62]. Tombolan et al. found significant associations between miR-30b/-30c levels and clinicopathological features including clinical stage and tumor size [42]. Metastatic status could be determined by enumeration of CTCs, as higher levels were observed patients with metastatic disease than in those with localized disease [29,30]. In meta-analysis grouping the patients included in these two studies no significant difference was observed in CTC level of RMS patients with localized and metastatic disease, respectively (unpaired t-test; p=0.87). Also, the presence of ctRNA correlates with worse survival as Krsková et al. found that all patients with detected ctRNA died of tumor progression [79].

Detection of ctDNA, ctRNA and circulating miRNA has proved to be useful in the monitoring of disease progression during treatment. Rapid decline of ctDNA levels has been observed during initial chemotherapy and rising levels detected in the same patients at disease progression or relapse and in one patient while off-therapy [62,69,73]. Detectable ctRNA at end of treatment worsened the prognosis as risk of metastasis development and death increased [78]. Finally, miR-206 levels has been shown to decrease with tumor reduction determined by ¹⁸F-FDG-PET/CT imaging performed during treatment to determine treatment response [89].

In conclusion, detection of CTCs, ctDNA, and ctRNA were all shown to be plausible targets for application in the clinical care of RMS patients. For diagnostic purposes, tNGS was best at detecting ctDNA at diagnosis (82%, n=17). Prognostic factors were mostly based on CTC and ctDNA levels, and all three methods proved useful in the monitoring of disease progression and treatment response.

Discussion

In pediatric sarcoma patients, cTM has been detected in the form of CTCs, ctDNA, ctRNA, miRNA and EV-associated RNA and DNA. Based on studies investigating CTCs, the CSV marker constituted the most efficient method for CTC detection. The analysis of ctDNA and circulating RNA was primarily carried out using NGS-based methods including tNGS and sWGS as well as PCR-based methods including ddPCR, RT-PCR, and qRT-PCR. PCR-based methods were more sensitive than NGS-based methods. However, PCR require knowledge of genetic characteristics of the tumor, whereas NGS-based methods do not require prior molecular analysis of tumor tissue to detect genomic alterations. Of the NGS-based methods, sWGS outperformed tNGS, however, CNA analysis is only usable in tumors with CNAs. Klega et al. investigated ctDNA and detected ESWR1 fusions in 91% of ES patients by tNGS, outperforming ULP-WGS that only identified fusions in 55%. However, ULP-WGS identified ctDNA in 90% of OS patients for whom tNGS was not carried out [73]. The use of sWGS and tNGS in combination could be an ideal approach to increase detection ctDNA. The results of the included articles showed that detection of ctDNA is superior to ctRNA since 63-94% (n=400) of ES patients had detectable ctDNA compared to 32% (n=285) that had detectable ctRNA. A reason for this could be that many of the studies investigating ctRNAs by RT-PCR date back to the 1990s, and there seems to be improvement in performance beginning from the mid 2000s (Table 1). In addition, ctRNA is more unstable than ctDNA and pre-analytical factors such as blood sampling and storage might impact RNA integrity [109].

In many cases of ES and RMS it is possible to detect ctDNA analysis without prior molecular analysis of tumor tissue due to the rapid occurrence of fusions that drive the development and relapse of these cancer types. The cases of RMS and ES in which fusions cannot be detected pose a greater challenge to the clinical application of cTM detection, which is why biopsy is required in all cases. In contrast, OS is not driven by gene fusions and is characterized by CNAs rather than SNVs as proven by sWGS outperforming ddPCR in ctDNA analysis. For these reasons, determining a molecular target in OS is more difficult. Researchers has mainly relied on other approaches using large tumor panels to catch genetic abnormalities and carrying out analysis of differential expression of miRNAs. Almost all studies find miRNAs to have important clinical relevance in pediatric OS, however, when it comes to specific miRNAs a very muddled picture reveals itself. Many miRNAs are suggested as diagnostic and prognostic markers; however, the studies rarely find the same miRNAs and often discrepancies between studies occur (Supplementary Table 2). The discrepancies could be partially explained by different primers used in the qRT-PCR assay.

Based on the review of all included studies, we suggest that detection of cTM is candidate for clinical implementation. However, cost and accessibility of the different technologies as well as pediatric blood sample volume limits must be taken into consideration [110]. The described methods seem to be especially relevant to implement in the current evaluation of disease burden and prognostic factors as well as monitoring disease progression and treatment response in pediatric sarcoma patients. It is likely that detection of cTM is more sensitive to predict prognostic factors than current methods. Prognostically, the tumor size evaluated by ¹⁸F-FDG-PET/CT and response to chemotherapy evaluated by tumor necrosis correlates with the prognosis of bone cancers ES and OS. However, the response to chemotherapy based on tumor size and necrosis does not correlate with the prognosis of RMS patients, and since current options to predict risk and monitor disease in the clinic is scarce, clinical implementation of cTM detection seem to be especially relevant in RMS [111], [112], [113]. In the monitoring of disease progression and treatment response of pediatric sarcoma, detection of cTM could be implemented in the clinic as a supplement to the current methods that includes multiple biopsies and medical imaging with MRI and ¹⁸F-FDG-PET/CT [114,115]. Levels of cTM may be a more sensitive marker, and in contrast to current methods, could be used to monitor treatment response early and during chemotherapy as it can be measured more frequently. Further, implementation could guide clinicians in deciding dosage of chemotherapy and frequency of medical imaging initially and during follow-up based on levels in PB.

For future research, it would be of great interest to align the methods used to investigate cTM for each disease in common clinical trials. This would ensure more comparable data to investigate topics such as the sensitivity of PB levels, the impact of collected blood volumes and the discovery of specific time-points during treatment that has especially important predictive value. Based on the findings, a consensus method for detecting ctDNA could be ddPCR for ES and RMS. However, NGS could answer questions regarding tumor biology and new treatment targets.

This systematic review was limited by varying study designs such as different sampling time-points; different follow-up periods; different age intervals; varying information on outcome and detection levels; and different technologies and molecular targets. For these reasons, it was not possible to directly compare all studies. Of note, it was not possible to investigate the differences between the occurrence of cTM in patients of different age groups, which could have implications when looking at PB volume analyzed.

Of the circulating cTM derived from ES and RMS ctDNA proved to be the best target for clinical application including diagnosis, tumor characterization, prognosis and monitoring of disease progression and treatment response. For OS further investigation is especially needed, however the most promising targets are CNAs or patient-specific miRNAs.

Funding

This work was supported by Childhood Oncology Network Targeting Research, Organisation & Life expectancy (CONTROL) funded by Danish Cancer Society (R-257-A14720) and the Danish Childhood Cancer Foundation (2019-5934).

CRediT authorship contribution statement

Eva Kristine Ruud Kjær: Conceptualization, Project administration, Methodology, Data curation, Writing – original draft. Christian Bach Vase: Methodology. Maria Rossing: Writing – review & editing. Lise Barlebo Ahlborn: Conceptualization, Project administration, Writing – review & editing. Lisa Lyngsie Hjalgrim: Conceptualization, Project administration, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper