Radiotherapy Dose Escalation Failed to Improve Local Control for Intermediate-Risk Rhabdomyosarcoma on ARST1431: A Report from the Children’s Oncology Group

Abstract

PURPOSE:

To evaluate local failure (LF) rates for patients with intermediate-risk rhabdomyosarcoma (IR-RMS) treated on the Children’s Oncology Group (COG) ARST1431 clinical trial, the first and largest international, phase 3 randomized study to use FOXO1 fusion status for risk stratification. To improve local control, radiation therapy (RT) dose was increased to 59.4 Gy for patients with tumors >5cm and residual gross disease at the time of RT.

PATIENTS AND METHODS:

For the 297 patients included, LF was defined as progression or relapse at the primary site. The rate of LF was calculated 3-years after enrollment.

RESULTS:

LF for group III, FOXO1 fusion-positive patients (n=58) compared to fusion-negative patients (n=175) was 10.7% vs. 21.5%, respectively (p=0.08). The LF rate for patients with tumors >5cm at diagnosis (n = 180; 24.4%) was higher than that of patients with tumors ≤5cm at diagnosis (n = 117; 9.8%), p=0.002. The risk of LF for patients who received proton (n=99) vs. photon RT (n=126) was not different (16.1% vs. 15.9%, p=0.8). For the 75 patients with tumors >5cm at diagnosis and gross disease at the time of RT, the boost to 59.4 Gy did not improve the 3-year LF rate compared to that of patients who did not receive the boost (29.7% vs. 16.1%, p=0.6). For patients with group III/IV disease, those who underwent delayed primary excision (DPE) (n=72) had a lower LF rate compared to those who had RT alone (n=151) (5.8% vs. 19.7%, p<0.01).

CONCLUSION:

On ARST1431, tumors >5cm at diagnosis had poor local control despite dose escalation to 59.4 Gy. Proton and photon RT had equivalent local control. For select patients, DPE significantly improved local control.

INTRODUCTION

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma in children and adolescents. Half of patients with RMS have intermediate risk (IR) disease, and nearly two-thirds of children with IR-RMS will become long-term survivors.¹ The prognosis of IR-RMS has improved substantially since the 1970s; indeed, 4-year event-free survival (EFS) rates on the most recent trials for IR-RMS patients have ranged from 59-73%.^2,3

Local failure (LF) is the most common pattern of treatment failure for patients with IR-RMS, and relapsed disease portends a poor prognosis.^3–7 Three recent Children’s Oncology Group (COG) clinical trials for IR-RMS patients, D9803, ARST0531, and ARST1431, focused on systemic therapy approaches combining the backbone of vincristine, dactinomycin, cyclophosphamide (VAC) with other agents to improve EFS. D9803 randomized patients with IR-RMS to VAC (cumulative cyclophosphamide dose of 30.8 g/m²) vs. VAC/VTC (vincristine, topotecan, and cyclophosphamide) (cumulative cyclophosphamide dose of 25.1 g/m²).⁸ Four-year EFS was 73% in the VAC arm and 68% in the VAC/VTC arm (p = 0.3).² Five-year rates of LF for group I/II alveolar RMS and group III RMS were 10% and 19%, respectively.⁴ On ARST0531, patients with IR-RMS were treated with VAC (cumulative cyclophosphamide dose 16.8 g/m²) or VAC/VI (vincristine and irinotecan) (cumulative cyclophosphamide dose 8.4 g/m²).³ Five-year rates of LF for group I/II alveolar RMS, group III alveolar RMS, and group III embryonal RMS were 13.4%, 20.2%, and 28.4%, respectively.¹ Multiple series support the correlation between cyclophosphamide dose and risk of LF. A retrospective analysis of patients with PM RMS showed that poor response to induction chemotherapy is associated with an increased risk of LF.⁹ A secondary analysis of ARST0331 of 66 patients with subset 2, low-risk RMS treated with a total cyclophosphamide dose of 4.8 g/m² found suboptimal EFS rates associated with the reduced cyclophosphamide dose.¹⁰ Two additional series of patients with head and neck RMS¹¹ and PM-RMS¹² showed higher LF rates associated with reduced-dose cyclophosphamide.

Secondary analyses of both D9803 and ARST0531 revealed that group III patients with tumors >5cm at diagnosis were significantly more likely experience LF than those with tumors ≤5cm at diagnosis.^1,4 On D9803, the 5-year rate of LF was 25% for tumors ≥5cm at diagnosis, compared to 10% for tumors <5cm at diagnosis (p=0.004).⁴ On ARST0531, the 5-year cumulative incidence of LF was 32.3% for tumors >5cm at diagnosis and 16.7% for tumors ≤5cm at diagnosis (p < 0.001).¹ Risk of LF did not differ by histology, nodal status, or primary site.^1,4

On ARST1431 (NCT02567435), patients with IR-RMS were randomized to 1 of 2 regimens, VAC/VI or VAC/VI plus temsirolimus, an mTOR inhibitor, and the primary outcome was EFS.¹³ The addition of temsirolimus to VAC/VI did not improve EFS relative to VAC/VI alone.¹³ The protocol incorporated both delayed primary excision (DPE) and a radiotherapy (RT) boost for select tumors. RT was administered after week 12 of chemotherapy, and both protons and photons were allowed.

RT dose escalation has been explored previously for IR-RMS patients. IRS-IV randomized clinical group III patients to conventionally fractionated RT (CFRT; 50.4 Gy in 1.8 Gy daily fractions) or to a hyperfractionated RT (HFRT; 59.4 Gy in 1.1 Gy twice daily (BID) fractions).^14,15 After randomization, there were no significant differences in the EFS, OS, and local control rates between the groups.¹⁴ Nevertheless, the 5-year local control rate for group III patients was 85-87%, the highest rate ever reached on IRS studies. Given the prior experience using HFRT on IRS-IV, coupled with the poor local control outcomes for patients with tumors >5cm at diagnosis on D9803 and ARST0531, ARST1431 mandated the treatment of patients with tumors >5cm at diagnosis and gross residual disease at the time of RT with a CFRT boost to 59.4 Gy in 33 fractions.

Regarding the surgical approach to RMS, typically patients undergo primary excision prior to additional therapy if the surgery is anticipated to result in a complete resection without significant functional or cosmetic deficit.¹⁶ DPE can assess for tumor viability after induction chemotherapy.¹⁷ Various studies have shown that patients with unresected disease (group III) can undergo DPE, and DPE allows for reduced-dose of radiation for definitive local control.^16,18,19 On ARST1431, patients who underwent DPE were treated to a lower dose of RT.

The goal of this secondary analysis was to define the LF rate on ARST1431 and examine variables associated with LF, including fusion status, RT modality, size at diagnosis, receipt of RT boost, and receipt of DPE.

METHODS

Eligibility and Systemic Therapy

Between June 2016 and January 2022, ARST1431 accrued 325 patients with IR-RMS, of which 297 patients were eligible and evaluable, and randomized them to either Arm A (VAC/VI) or Arm B (VAC/VI plus temsirolimus) (see CONSORT Diagram). The study protocol was approved by independent ethics committees and all relevant institutional review boards for each site, and all patients (or guardians) provided written informed consent. Clinical group I and II tumors are defined as having primary tumor resection before initiation of chemotherapy.^4,20 Group III patients have incomplete resection with gross residual disease or biopsy only prior to starting chemotherapy, while group IV patients have metastatic disease at diagnosis.²⁰ Stage 1 disease includes primary tumors at any favorable site (biliary, orbit, non-PM head and neck, and non-bladder/non-prostate genitourinary (GU)) without metastases, regardless of size at diagnosis or nodal status. Stage 2 and 3 disease includes primary tumors at any unfavorable site (Stage 3 includes size >5cm at diagnosis or node- positive (N1)), while stage 4 is metastatic.²⁰ IR-RMS patients fit one of the following categories: 1) fusion-positive (FP) RMS, clinical groups I-III, stages 1-3, 2) fusion-negative (FN) RMS, clinical group III, stage 2-3, 3) FN RMS, stage 1, clinical group III (non-orbit), 4) FN RMS, clinical group I-II, stage 3, or 5) age <10 years old with FN RMS, clinical group IV, stage 4.²⁰ The 297 patients randomized on the trial were included in this secondary analysis.

CONSORT Diagram

Local Therapy

Local control for RMS can be attained via surgery, definitive RT, or both. On ARST1431, RT began at week 13. RT dosing is as per Supplemental Table 1. Patients with clinical group IV disease were treated as per the clinical group of the primary tumor. DPE was permitted for patients with non-parameningeal (PM), group III disease after 9 weeks of induction chemotherapy if gross total resection (GTR) was achievable without causing significant functional impairment.

Various types of RT were allowed per protocol, including brachytherapy, proton RT, and photon RT. For external beam RT, two gross tumor volumes (GTVs) are defined per protocol. GTV1 is gross disease prior to surgery or chemotherapy, while GTV2 is residual tumor following induction chemotherapy and DPE, if performed. Clinical target volume (CTV) 1 and CTV2 are defined as a 1 cm expansion on the respective GTV and include any involved lymph nodes. GTV1/CTV1 were treated to 36 Gy, whereas GTV2/CTV2 were treated to 50.4 or 59.4 Gy (Supplemental Table 1).

To maximize protocol compliance, the Quality Assurance Review Center performed a review of all radiation planning materials at the start of RT. Prescription doses that were <90% or >110% of protocol recommendations and cases with missing tumor volume were considered major deviations.

Endpoints and Statistical Methods

LF was defined as progression or relapse in the primary tumor site occurring as a first event (with or without concurrent regional and/or distant failure). Time to LF was defined from the date of enrollment. Gray’s test was used to compare incidence of LF across clinical variables, with regional and/or distant failures and death treated as competing events.²¹ Statistical analyses were conducted using SAS version 9.4.

RESULTS

Patient Population

For the 297 patients randomized on ARST1431, median follow up time was 3.8 years (interquartile range 3.0-4.8 years). The most common primary tumor sites were PM (n = 78; 26.3%), non-PM head and neck (n = 45; 15.2%), extremity (n = 44; 14.8%), retroperitoneal (n = 38; 12.8%), and GU bladder/prostate (n = 35; 11.8%). Most patients had group III FN RMS (n = 175; 58.9%) and were N0 (n = 182; 61.3%). Of the 235 (79%) patients who received RT, 99 (42%) received protons (Table 1). There were 14 major RT deviations (6%). A comparison of LF rates on D9803, ARST0531, and ARST1431 can be found in Table 3.

Table 1.

Baseline Patient and Tumor Characteristics on ARST1431

Characteristic	Total (N=297)
Primary Site Category, n (%)
Favorable	61 (20.5%)
Unfavorable	236 (79.5%)
Parameningeal Primary Site, n (%)
No	219 (73.7%)
Yes	78 (26.3%)
Nodal Stage*, n (%)
N0	182 (61.3%)
N1	110 (37.0%)
Not evaluated	5 (12.7%)
Tumor Size at Diagnosis, n (%)
≤5 cm	117 (39.4%)
>5 cm	180 (60.6%)
DPE, n (%)
No	224 (75.4%)
Yes	73 (24.6%)
RT Modality, n (%)
No RT	62 (20.9%)
Protons	99 (33.3%)
Photons	126 (42.4%)
Brachytherapy	8 (2.7%)
Missing	2 (0.7%)

^*By clinical or imaging nodal assessment

Table 3.

Comparison of Local Failure Rates on D9803, ARST0531, and ARST1431

Patient Subset	D98031 (5-year local failure rate) (CPMa: 25.1-30.8 g/m2)	ARST05312 (5-year local failure rate) (CPM: 8.4-16.8 g/m2)	ARST1431 (3-year local failure rate) (CPM: 12.6 g/m2)
Alveolar/Fusion-Positive, Group III	17%	20%	11%
Embryonal/Fusion-Negative, Group III	20%	28%	22%
>5cm	25%	32%	23%
≤5cm	10%	17%	11%

^aCPM = cyclophosphamide

Clinical Group I/II FP RMS

There were 15 patients (5%) who were clinical group I or II with FP disease. The 3-year cumulative incidence of LF for these patients was 7.2% (95% confidence interval (CI) 0.4-28.7%).

Clinical Group III RMS

There were 235 patients with clinical group III disease on ARST1431. Of these 235 patients, 175 (74%) were FN, and 58 (25%) were FP (2 patients had missing fusion status). For all group III patients, the 3-year cumulative incidence of LF was 18.6% (95% CI 13.8-24.0%). The 3-year LF rate for FP patients was 10.7% (95% CI 4.3%-20.4%), compared to 21.5% (95% CI 15.6%-28.0) for FN patients (p=0.08; Figure 1A).

Figure 1.

A) Cumulative incidence of local failure for clinical group III patients stratified by fusion status. B) Cumulative incidence of local failure for all clinical groups, stratified by RT modality (protons or photons). C) Cumulative incidence of local failure for all patients stratified by tumor size at diagnosis, ≤5 cm or >5cm at diagnosis.

RT Modality, Size, RT Boost

There was no difference between the cumulative incidence of LF for patients who received proton RT (n= 99) (16.1%, 95% CI 10.4%-26.1%) compared to photon RT (n=126) (15.9%, 95% CI 10%-23.1%), p=0.8 (Figure 1B).

There was no significant difference in the LF rate for patients by favorable or unfavorable primary site (12 vs. 20.2%, p = 0.2), PM or non-PM site (26.2 vs. 16.0%, p = 0.1), treatment with or without temsirolimus (17.3% vs. 20.2%, p = 0.9), or N1 vs N0 nodal status (17.6% vs. 19.2%, p = 0.6).

Patients with tumors >5cm at diagnosis had a 3-year LF rate of 24.4% (95% CI 18.2%-31.1%), compared to 9.8% (95% CI 5.2%-16.2%) for those with tumors ≤5cm at diagnosis (p=0.002) (Figure 1C). For group III patients, the 3-year rates of LF for patients with tumors ≤5 cm at diagnosis and >5cm at diagnosis were 11.2% (95% CI 5.7-18.7%) and 23.4% (95% CI 16.6-30.8%), respectively (p=0.03; Supplemental Figure 1; Table 2).

Table 2.

Characteristics of Patients Meeting Criteria for RT Boost

Patients Meeting Criteria for RT Boost
	Boost (N=59)	No Boost (N=16)	P-value
Primary Site, n (%)			0.971
Extremity	3 (5.1%)	1 (6.3%)
GU Bladder/Prostate	10 (16.9%)	2 (12.5%)
GU Non-Bladder/Prostate	1 (1.7%)	0 (0.0%)
Head & Neck	9 (15.3%)	2 (12.5%)
Intrathoracic	1 (1.7%)	0 (0.0%)
Parameningeal	17 (28.8%)	6 (37.5%)
Perineum-Anus	1 (1.7%)	1 (6.3%)
Retroperitoneum	12 (20.3%)	4 (25.0%)
Trunk	2 (3.4%)	0 (0.0%)
Other	3 (5.1%)	0 (0.0%)
Tumor Size, median (min-max) in cm	7 (5.1-19)	7.3 (5.5-14.4)	0.531
RT Modality, n (%)			0.702
Photons	30 (50.8%)	9 (56.2%)
Protons	29 (49.2%)	7 (43.8%)
Group, n (%)			0.401
III	53 (89.8%)	13 (81.3%)
IV	6 (10.2%)	3 (18.8%)
Fusion, n (%)			0.461
FOXO1+	9 (15.3%)	4 (25.0%)
FOXO1−	50 (84.7%)	12 (75.0%)

¹Fisher Exact p-value;

²Chi-Square p-value

There were 238 patients in groups III/IV with available information regarding complete response (CR) in the primary site to chemotherapy at week 9. Twenty patients (8.4%) had a CR at the primary site at week 9. Fourteen of these patients (70%) had FN disease. The three-year LF rate for those patients who had a CR at week 9 was 10.9% (95% CI 0.3-22%), compared to 18.2% (95% CI 13.2-23.8%) for the 218 group III/IV patients who did not have a CR at week 9 (p=0.4).

A total of 89 patients with tumors >5cm at diagnosis in groups III or IV had evidence gross disease after induction chemotherapy (without DPE) and were therefore mandated to receive a boost to 59.4 Gy. Of the 89 patients, 75 (84%) received RT, and 59 (66%) patients received the full, protocol-specified boost. Nine out of the 14 (64%) patients who did not receive RT had progressive disease during induction chemotherapy and subsequently went off trial. The remaining 5 patients went off trial for other reasons. For the 16 patients who did not receive the full boost, reasons for deviation were primarily due to concern for toxicity. Specifically cited reasons included patient age between 0-2 (4, 25%), poor response to chemotherapy and RT (1, 6%), negative fusion status (2, 12.5%), concerns about adjacent organs toxicity (1, 6%), and very large tumor size (1, 6%).

There was no significant difference in the primary site, tumor size, RT modality (protons vs photons), clinical group, or fusion status between the patients with tumors >5cm at diagnosis and residual gross disease at the time of RT who received and did not receive the full boost (Table 2). Patients who received the full 59.4 Gy had a 3-year cumulative incidence of LF of 29.7% (95% CI 18%-42.3%), while patients who did not receive the full boost had a 3-year cumulative incidence of LF of 16.1% (95% CI 3.8%-36.2%), p=0.6 (Figure 2). Only one patient out of 59 (1.7%) who received the boost to 59.4 Gy had a major protocol deviation. This patient had low prescription dose coverage of the CTV due to overlap with the rectum.

Figure 2.

Cumulative incidence of local failure for group III/IV patients with tumors >5cm at diagnosis with gross disease after induction chemotherapy, stratified by receipt of RT boost to 59.4 Gy.

DPE

There were 72 patients with group III/IV disease who underwent DPE (1 patient who received DPE had disease group recorded IIa, microscopic residual disease after primary surgery). Those patients who underwent DPE had a lower 3-year LF rate compared to those who underwent RT alone (n=151) (5.8% vs 19.7%, p < 0.01) (Figure 3). There were more patients with extremity tumors who received DPE (20/72; 27.8%) compared to those who did not receive DPE (9/151; 6%), p < 0.001 (Supplemental Table 2). Four patients (5.4%) with PM disease underwent DPE. A total of 55/73 (75%) of the instances of DPE resulted in a GTR. For group III patients specifically, the 3-year rates of LF for DPE ± RT compared to RT alone were 5.1% (95% CI 1.3-12.8%) vs. 20.6% (95% CI 14-28.1%) (p < 0.01; Supplemental Figure 2).

Figure 3.

Cumulative incidence of local failure for group III/IV patients who received RT alone or DPE, with or without RT.

DISCUSSION

Failure at the primary site is the most common pattern of failure for all patients with RMS.²² To guide future treatment of IR-RMS, it is critical to understand the factors associated with LF, as well as the interventions that might lower the risk of LF for these patients

The fact that 16/75 (21%) of patients with group III/IV tumors, disease >5cm at diagnosis, and residual gross disease at the time of RT did not receive the full boost is notable. On IRS-IV, the protocol compliance to the HFRT arm was 57% for patients <5 years of age compared to 88% for patients ≥5 years of age.¹⁴ These data underscore the difficulty in in dose escalating across a wide variety of primary sites in a patient population where balance of maximizing local control while minimizing the risk of toxicity is paramount. Even when the full boost dose of 59.4 Gy was delivered, the 5-year LF rate was 29.7%, whereas for all patients with group III/IV disease >5cm at diagnosis, the 5-year cumulative incidence of LF was 23.4%. In comparison, on ARST0531 and D9803, the 5-year incidence of LF for group III patients with tumors >5cm at diagnosis was 32.3% and 25.0%, respectively (Table 3).^1,4 Thus, our analysis shows that the CFRT boost to 59.4 Gy failed to improve local control for group III/IV patients with >5cm tumors at diagnosis.

For all group III patients, the 3-year cumulative incidence of LF on ARST1431 was 18.6%, and there was a trend toward increased LF in FN patients compared to FP disease, albeit not statistically significant. Many previous analyses have shown that FP disease is associated with decreased EFS and OS compared to FN disease.^23–25 ARST1431 was the first COG trial to use FOXO1 fusion status in stratifying patients with RMS. Previously, on ARST0531 and D9803 the 5-year rates of LF were lower for group III alveolar patients compared to embryonal patients, but neither difference reached statistical significance (Table 3).^1,4 While fusion status cannot serve as a direct comparison to histology, these data support a lower LF rate for alveolar/FP disease compared to embryonal/FN disease. Reconciling these data with the previous analyses on worse outcomes associated with FP tumors suggests that the poorer outcomes associated with FP disease are likely due to distant—rather than local—failure.

The use of proton RT for the treatment of pediatric malignancies has been increasing in recent decades.²⁶ Previously, there had been concerns about local control rates for patients with RMS treated with proton RT given the steep dose falloff.²⁷ In an analysis of patients with RMS of the head and neck treated with proton or photon RT, proton RT was associated with worse local control rates, but this trend did not reach statistical significance. The authors identified that patients who received proton RT were significantly more likely to have received reduced-dose cyclophosphamide that patients who received photon RT.¹¹ Multiple studies have shown that proton RT affords similar local control rates to photon RT for the treatment of RMS.^27–29 On ARST0531, 47 patients (12% of those that received RT) were treated with proton RT. On ARST1431, 99 patients (42% of those that received RT) were treated with protons. Thus, the utilization of proton RT has increased. Encouragingly, we observed no difference in the cumulative incidence of LF for patients receiving photon versus proton radiotherapy on ARST1431. While this was not a randomized comparison, these data support prior analyses which have shown that proton RT provides similar local control to photon RT while potentially decreasing damage to adjacent normal tissue.

The relatively high local control rates seen on IRS-IV and D9803 posed questions about how to optimally achieve local control in future trials for IR-RMS patients. Induction chemotherapy is another consideration for the attainment of local control. For many decades now, the backbone of induction chemotherapy for the treatment of RMS in the United States has been VAC. After ARST0531 showed that there were no statistically significant differences in 4-year EFS and OS between VAC (cumulative cyclophosphamide dose 16.8 g/m²) compared with VAC/VI (cyclophosphamide dose 8.4 g/m²) with higher toxicity rate in the VAC arm, ARST1431 used a modest dose of cyclophosphamide, a cumulative of 12.6 g/m² including maintenance therapy.^3,8 Multiple studies have suggested that reduced-dose cyclophosphamide is associated with increased risk of LF.^1,10–12 These studies, in combination with the data presented here, support the argument to evaluate the optimal cyclophosphamide dose and schedule in future IR-RMS trials, especially for patients in whom DPE is not indicated (i.e. PM disease).On ARST1431, patients with group III/IV disease who received DPE had significantly lower rates of LF than those who had RT alone, consistent with prior analyses.¹⁶ It is important to acknowledge potential selection bias in this group of patients. For example, more patients who received DPE had extremity tumors compared to those who did not receive DPE. However, the evidence would suggest that in the context of this chemotherapy regimen, when a GTR of the primary tumor can be achieved with minimal morbidity, and the patient receives post-DPE RT, local control is excellent. These data are further supported in an analysis of 101 children with metastatic RMS showing an overall survival benefit for patients treated with both surgery and RT (44% ± 8%), compared to either treatment alone (18.8% ± 15% for surgery alone and 16.1% ± 7% for RT alone).³⁰ A comprehensive analysis of the outcomes of patients who received DPE on ARST1431 is forthcoming.

In Europe, the Frontline and Relapsed Rhabdomyosarcoma (FaR-RMS) trial is randomizing patients with 1) unresected disease, 2) an incomplete response, and 3) at a high risk LF (patients with disease at an unfavorable primary site including all sites other than genitourinary, head and neck non-PM, orbit, and biliary, or adult patients) to receive either standard dose of 50.4 Gy or dose-escalated radiotherapy to 59.4 Gy.³¹ The trial is currently ongoing and will provide additional data on the efficacy of local control for dose-escalated RT. Further work is also ongoing to better risk stratify patients with information such as genetic mutations in MYOD1 and TP53.³² Consideration should also be given to potential radiosensitizers such as inhibition of the DNA damage response pathway proteins PARP1 or ATR, for example.³³

Conclusions

In summary, for patients with IR-RMS, we confirmed that tumor size at diagnosis is the most important risk factor for LF. On ARST1431, there was no difference in local control by RT modality (photons or protons). For clinical group III or IV patients with tumors >5cm at diagnosis and macroscopic residual disease after induction chemotherapy, receipt of an RT boost to 59.4 Gy did not improve local control. For a select group of patients with tumors amenable to GTR without significant loss of function or morbidity, DPE with post-operative RT is an effective approach to decreased risk of LF. Further study is needed to identify the optimal strategy for achieving local control in IR-RMS patients.

Data Sharing Statement:

Data will not be shared.