High-Dose Carbon Ion Reirradiation for Recurrent Pelvic Sarcomas: A Retrospective Analysis of Oncologic Outcomes in a Referral Center

Abstract

Purpose

This study aims to evaluate the safety and efficacy of high-dose carbon ion reirradiation (CIR) in patients with recurrent pelvic sarcomas.

Methods and Materials

This retrospective study included 46 patients treated with high-dose CIR for locoregional recurrences of pelvic sarcomas between 2013 and 2023 at the National Center for Oncological Hadrontherapy, Pavia, Italy. Clinical outcomes, toxicity, and cumulative doses were analyzed.

Results

The median prescribed total dose of CIR was 60 GyRBE (range, 45-73.6 GyRBE). The maximum cumulative equivalent dose in 2 Gy fractions to the target volume reached 284 Gy, with a median of 153 Gy. The median follow-up was 32.5 months. The 1-, 2-, and 3-year locoregional progression-free survival rates were 79.1%, 43.9%, and 25.1%, respectively, with a mean time to locoregional relapse of 18.9 months. The 1-, 2-, and 3-year overall survival rates were 100%, 93.9%, and 83.1%, respectively. At last follow-up, grade 3 neuropathy occurred in 6.5% of the patients, grade 3 anorectal dysfunction in 8.7%, and grade 3 skin toxicity in 6.5%.

Conclusions

High-dose CIR is a feasible salvage strategy for recurrent pelvic sarcoma, offering promising outcomes and manageable toxicity in carefully selected patients. Prospective studies are warranted to refine dose-response models and validate long-term efficacy.

Introduction

Pelvic sarcomas are rare and heterogeneous malignancies representing approximately 10% to 15% of all soft tissue and bone sarcomas. Their management is challenging because of their usually large size, complex anatomic location, and frequent involvement of critical structures.1, 2, 3

To achieve optimal results, a multidisciplinary approach in specialized centers is required. Treatment depends on histology, with surgery as the cornerstone when feasible. However, pelvic anatomy and proximity to vital organs often limit complete resection and increase morbidity and functional impairment.3, 4, 5, 6, 7

In light of these challenges, advanced radiation therapy (RT) techniques have become valuable alternatives, especially when surgery is not feasible, would lead to inadequate margins, or is no longer an option because of disease recurrence.^4,8, 9, 10, 11, 12 Furthermore, pelvic sarcomas are often radioresistant and require high doses for effective local control.^13,14 Despite modern treatments, local recurrence remains common, with 5-year rates of 20% to 30%.15, 16, 17 Management of recurrent disease is therefore complex and must be tailored based on prior treatments and institutional expertise.

For isolated locoregional recurrence, high-dose reirradiation is a viable option.18, 19, 20 In this clinical setting, carbon ion RT (CIRT) represents a promising strategy to improve radioresistant tumor control while limiting morbidity, leveraging high-dose conformality, elevated relative biological effectiveness (RBE), and high linear energy transfer (LET) to spare normal tissues.21, 22, 23 Nonetheless, pelvic reirradiation requires accurate cumulative organ-at-risk (OAR) dose assessment to minimize toxicity.²⁴

This retrospective cohort study aims to evaluate the role of carbon ion reirradiation (CIR) in recurrent pelvic sarcomas, exploring whether its physical and biological advantages translate into better local control with acceptable toxicity in this difficult clinical scenario.

Methods and Materials

Patient population

We retrospectively analyzed 46 consecutive patients who underwent high-dose CIR for locoregional recurrent pelvic sarcomas between December 2013 and October 2023 at the National Center for Oncological Hadrontherapy, using data from the institutional longitudinal registry (REGAL, NCT05203250).

Inclusion criteria were as follows: (1) histologically and/or radiologically proven diagnosis of recurrent pelvic sarcoma after a prior course of RT; (2) unresectable relapse or refusal of surgery; (3) an interval of ≥6 months between RT courses; (4) age >18 years; (5) Karnofsky Performance Status ≥60; and (6) a minimum follow-up of 6 months.

The study was conducted in accordance with the Declaration of Helsinki; all patients provided written informed consent for treatment and use of their anonymized data for research. The study was approved by the institutional Ethics Committee (CNAO OSS 77 2025).

CIR procedures

All patients underwent simulation computer tomography (CT) in treatment position, supine or prone, depending on tumor location, with immobilization using a thermoplastic mask and a personalized cushion. Planning CT scans, acquired with a 2-mm slice thickness without contrast enhancement, were completed by 3-Tesla magnetic resonance imaging (MRI) with gadolinium contrast using the same setup. The CT and MRI images were rigidly registered in the Treatment Planning System (TPS) to improve target and OARs delineation.

The gross tumor volume was defined as the visible tumor on MRI. A 2 to 5 mm safety margin was added to create the clinical target volume (CTV), adapted to anatomy, tumor spread pathways, natural barriers, and previously irradiated regions to minimize OAR exposure.

For 37 patients, previous RT plans were digitally available. The planning CT of the first course (CT-1) was rigidly registered to the planning CT from the second course (CT-2), followed by deformable image registration. Only gross tumor volume and CTV contours were transferred from CT-1 to CT-2; first-course dose distributions were deformed onto CT-2 to estimate cumulative OAR dose. When applicable, an overlap volume was generated by identifying the intersection between the CTVs of the 2 courses, providing the basis for categorizing recurrences as in-field, marginal, or out-of-field according to their spatial relationship with the initial CTV. To account for differences in dose and fractionation, the doses from the first and second RT courses were converted to equivalent doses in 2 Gy fractions (EQD2) using the linear-quadratic model with an α/β = 3, acknowledging that this approach does not account for variations in radiation type. Cumulative dose distributions were evaluated on the CT-2 by summing the EQD2 plans from both RT courses (EQD2 TOTAL) to assess the total dose delivered to targets and OARs (Fig. 1). For the remaining 9 patients, the original digital treatment plans were unavailable; therefore, automated 3-dimensional dose summation was not feasible. Cumulative doses to targets and OARs were manually estimated from available printed beam arrangements and dose-volume histogram printouts, allowing careful definition of dose prescriptions and OAR constraints.

Figure 1.

Sacral chordoma. (A, B) Axial and coronal views of the initial CIRT treatment plan delivered in prone position (prescription dose: 64 GyRBE), yellow line = CTV1, red zone = 72 GyEQD2. (C, D) Axial and coronal views of the reirradiation CIRT plan delivered in prone position (prescription dose: 67.2 GyRBE), green line = CTV2, red zone = 79 GyEQD2. (E, F) Axial and coronal cumulative dose distribution, yellow line = CTV1, green line = CTV2, pink line = CTV OVERLAP, red zone = 150 GyEQD2, yellow zone = 75 GyEQD2.

Abbreviations: CIRT = carbon ion radiation therapy; CTV = clinical target volume; RBE = relative biological effectiveness.

Treatment plans were performed using Syngo PT TPS (Siemens AG Healthcare) until October 2019 and Raystation TPS (Raysearch) thereafter; older plans were recalculated in Raystation before summation. Plans were optimized using intensity modulated particle therapy, based on the Local Effect Model version 1 for RBE calculation.²⁵ The aim was CTV D95 (the dose received by 95% of the target volume) ≥95% of the prescription for the first RT and, when feasible, for CIR, with intentional underdosage when needed to respect OAR constraints.

The following OAR dose constraints were applied for the initial treatment: rectum maximum dose (Dmax) <60 GyRBE, sigmoid colon Dmax <52 GyRBE, small bowel Dmax <46 GyRBE, skin D5 cm² <65 GyRBE, and cauda equina <105% of the prescription dose. The bladder was excluded from primary constraints because of its geographic distance from the target volumes in this specific cohort. During CIR, these constraints were also considered, and whenever spatial margins allowed, the maximum tolerable dose or more was delivered to OARs. Cumulative dose allowances were assessed on a case-by-case basis, and constraints were adjusted considering the dose previously delivered and the expected tissue recovery over time.^26,27

Follow-up

Patients were assessed at baseline, during RT, and at treatment completion. Follow-up was conducted every 4 months during the first 2 years and every 6 months thereafter. Each visit included clinical examination and MRI of the pelvis and/or lumbosacral spine, with additional staging as per histology-specific guidelines. Response to reirradiation was assessed using the Response Evaluation Criteria in Solid Tumors, version 1.1. Toxicity was graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0. For each patient, the maximum grade of acute and late toxicities was recorded, focusing on neuropathy, pain, bladder dysfunction, rectal dysfunction, skin toxicity, and bone toxicity.

Statistical analysis

The cohort was described by specific patients and treatment-related characteristics. Absolute and relative frequencies were reported for qualitative variables, whereas the median, minimum, and maximum were reported for quantitative variables. Overall survival (OS) and locoregional progression-free survival (LRPFS) were defined as right-censored time-to-event endpoints measured from the end of reirradiation to death from any cause and locoregional progression, respectively. OS and LRPFS curves were estimated using the Kaplan-Meier method, and differences among groups were assessed using the Long-Rank test. The association between each explanatory variable and the hazard of locoregional progression was evaluated using an univariable Cox proportional hazards model, with results reported as hazard ratios (HRs) and corresponding 95% CI. Diagnostics for the proportional hazards assumption were conducted for all models. A subgroup analysis was performed to evaluate the oncologic outcomes of chordoma, which represented the largest homogeneous subcohort and a biologically and clinically distinct entity from other pelvic sarcomas.

Detailed definitions of endpoints, censoring, and additional methodological aspects, including covariates, handling of missing data, and sensitivity analyses, are provided in the E Supplementary Materials (Sections E1.1–E1.4). Supplementary Materials (SM) (Sections 1.1-1.4). To estimate the associations between cumulative dose and toxicity grades at baseline and follow-up, a multinomial model for ordinal data with proportional odds was used. All statistical analyses were performed using the R Studio 4.5.2 software and MATLAB software (MathWorks Inc, 2024b).

Results

Participant attrition

Out of 46 participants enrolled, none withdrew consent, and all were included in the final survival analysis. No missing data were observed for the outcome or key variables. During follow-up, 17 participants were censored. Among these, 5 (3 patients with chordoma and 2 patients without chordoma) were lost to follow-up because of distant progression with subsequent referral to oncology colleagues; for these patients, neither the exact timing of distant progression nor the date of transfer to another hospital was available. The remaining 12 cases were censored at the end of the available follow-up period.

Patient and treatment characteristics

Patient and treatment details are shown in Table 1. The most common histologic subtype was chordoma (80.4%), followed by high-grade chondrosarcoma (13%). Other 3 sarcomas were recorded: 2 leiomyosarcomas and 1 spindle cell sarcoma. Median age at reirradiation was 60 years. Median time from the end of primary RT to recurrence was 24 months (range, 3-152 months), and from the end of the first RT to the start of reirradiation was 31.5 months (range, 6-157 months). Recurrence occurred in-field in 24 patients, marginal in 17, and out-of-field in 5.

Table 1.

Patient and treatment characteristics

Characteristic	Value
Total patients	46
Sex, n (%)	33 male (72%) 13 female (28%)
Age at initial treatment (y)	Min: 35, max: 82; median: 56
Age at reirradiation (y)	Min: 36, max: 83; median: 60
Histology, n (%)	Chordoma: 37 (80.4%) High-grade chondrosarcoma: 6 (13%) Leiomyosarcoma: 2 (4.3%) Spindle cell sarcoma: 1 (2.2%)
Tumor localization at diagnosis and initial treatment, n (%)	Sacrum: 40 (87%) Left hemipelvis: 2 (4.3%) Left iliac wing: 1 (2.2%) Right gluteus: 1 (2.2%) Right acetabulum: 1 (2.2%) Pelvis: 1 (2.2%)
Surgery prior to first RT, n (%)	No: 24 (52.2%) Yes: 22 (47.8%)
Neoadjuvant chemotherapy prior to first RT, n (%)	No: 42 (92%) Yes: 4 (8%)
Indication for first RT, n (%)	At diagnosis with curative intent 24 (52.2%) Adjuvant RT after surgery 6 (13%) RT for local recurrence after surgery 16 (34.8%)
First RT modalities, n (%)	Carbon ions: 35 (76.1%) Protons: 5 (10.9%) Photons: 5 (10.9%) Photons/carbon ions: 1 (2.2%)
Total dose first RT (GyRBE)	Min: 30, max: 74; median: 70.4
Dose per fraction first RT (GyRBE)	Min: 1.8, max: 4.6; median: 4.4
Number of fractions first RT	Min: 10, max: 37; median: 16
GTV first RT (cm3)	Min: 6.51, max: 5085.09; median: 151.83; missing: 11 (23.9%)
Site of recurrence (vs initial CTV)	In-field: 24 Marginal: 17 Out-of-field: 5
Total dose re-RT (GyRBE)	Min: 45, max: 73.6; median: 60
Dose per fraction re-RT (GyRBE)	Min: 3, max: 4.8; median: 4
Number of fractions re-RT	Min: 12, max: 22; median: 16
GTV re-RT (cc)	Min: 7.37, max: 1115.26; median: 87.18; missing: 2 (4.3%)
Median EQD2 D95 cumulative dose (CTV)	130 Gy (range, 27-244)
Median EQD2 D1 cumulative dose (CTV)	153 Gy (max: 284 Gy)

Abbreviations: CTV = clinical target volume; GTV = gross tumor volume; RBE = relative biological effectiveness; RT = radiation therapy; re-RT = reirradiation.

Regarding initial RT, 35 patients received CIRT, 5 received proton therapy, 5 were treated with photons, and 1 patient received a mixed-beam treatment with photons and carbon ions. Median doses were 70.4 GyRBE (range, 64-73.6 GyRBE) for CIRT, 74 GyRBE (range, 60-74 GyRBE) for proton therapy, and 70.4 Gy (range, 30-74 Gy) for photons. The patient treated with the mixed-beam approach received 73.2 GyRBE of total dose (49.2 Gy with photons + 24 GyRBE with carbon ions). Twenty-four patients were treated with curative intent at diagnosis. Twenty-two patients underwent surgery prior to RT; of these, 6 received adjuvant RT, whereas the remaining 16 underwent RT for local relapse.

All patients underwent reirradiation with carbon ions, using 1 of 2 different fractionation schedules: most patients (65%) were treated with ≥4 GyRBE per fraction, delivered 4 times per week, whereas the remaining 35% received 3 GyRBE per fraction, 5 times per week. The total dose was individualized, based on the expected risk-benefit ratio between tumor control and normal tissue toxicity. Median reirradiation doses were 64 GyRBE (range, 48-73.6 GyRBE) for the ≥4 GyRBE/fraction group and 54 GyRBE (range, 45-66 GyRBE) for those treated with 3 GyRBE/fractionation.

Regarding cumulative dose calculations, a median EQD2 D95 to the CTV was 130 Gy (range, 27-244 Gy). The lowest cumulative EQD2 D95 values were observed in patients with out-of-field recurrences, where the reirradiation CTV did not overlap with the high-dose region of the initial treatment. In these cases, the cumulative dose mainly reflected the reirradiation course, and lower D95 values were primarily because of intentional target undercoverage to meet OAR constraints. The maximum EQD2 D1 (the dose received by 1% of the contoured volume) to the CTV reached 284 Gy, with a median of 153 Gy. Cumulative EQD2 doses to OARs are reported in Table 2. A summary of the initial irradiation modality and the corresponding EQD2 D95, together with the EQD2 D95 delivered during reirradiation, is reported in SM (Table E14).

Table 2.

Cumulative EQD2 doses to organs at risk at 1% of the volume (D1)

Organ-at-risk	Median EQD2 D1 (Gy)	Range (Gy)
Rectum	41	3-108
Sigmoid colon	40	1-107
Small bowel	20	0-96
Skin	67	0-154
Cauda equina	71	0-171

Abbreviation: EQD2 = equivalent doses in 2 Gy fractions.

Clinical outcomes

Median follow-up was 32.5 months (range, 6-107.2 months). After CIR, 14 patients (30%) maintained stable disease, and 3 patients (6%) had a partial response. Locoregional progression occurred in 29 patients (63%): 25 (54%) had local progression (in-field or marginal), and 4 (8%) had regional failures (out-of-field). Among those progressing, 9 (19.5%) underwent a second reirradiation, 10 (21.7%) started treatment with a tyrosine kinase inhibitor, and 1 patient (2%) received a third reirradiation. At the last follow-up, 8 patients (17.4%) died, with 5 deaths (62.5%) because of disease progression.

The 1-, 2-, and 3-year LRPFS rates were 79.1% (SE, 6.21%; 95% CI, 67.8%-92.3%), 43.9% (SE, 8.15%; 95% CI, 30.5%-63.1%), and 25.1% (SE, 7.44%; 95% CI, 14.0%-44.9%), respectively. The 1-, 2-, and 3-year OS rates were 100%, 93.9% (SE, 4.15%; 95% CI, 86.1%-100%), and 83.1% (SE, 6.95%; 95% CI, 70.6%-97.9%), respectively. After 4 years, OS plateaued at 62.21% (SE, 11.90%; 95% CI, 42.6%-90.4%). Among the 8 deceased patients, half died within 30.9 months from reirradiation (Fig. 2).

Figure 2.

Kaplan-Meier curves for (A) locoregional progression-free survival and (B) overall survival following reirradiation.

No significant survival differences were found by recurrence site (P = .34), initial RT intent (P = .80), reirradiation dose ≥60 GyRBE (P = .70), or fractionation schedule (P = .20) (Fig. 3). Despite the limited sample size for subgroup comparisons, some trends emerged. The global median LRPFS was 23.5 months: 24.5 months for in-field recurrences, 17.3 months for marginal, and 12.2 months for out-of-field. Patients receiving ≥60 GyRBE had a median LRPFS of 23.5 months, versus 17.3 months for <60 GyRBE. Regarding dose per fraction, treatment with 3 GyRBE was associated with a median LRPFS of 24.4 months, compared with 17.3 months for ≥4 GyRBE.

Figure 3.

Kaplan-Meier survival analyses for locoregional progression-free survival according to (A) recurrence site (log-rank P value = .34), (B) intent of the first RT (log-rank P value = .80), (C) total reirradiation dose of ≥60 GyRBE ≥ (log-rank P value = .70), and (D) fractionation schedule (log-rank P value = .20).

Abbreviations: RT = radiation therapy; RBE = relative biological effectiveness.

Results of Kaplan-Meier analyses for the chordoma subcohort are reported in SM (Section E2.1). Results from univariable and multivariable Cox regression models are reported in SM (Section E2.2).

Sensitivity analysis on LRPFS curves, assuming a worst-case informative censoring scenario, showed results substantially consistent with the primary analysis. However, other results showed discrepancies, and the details are provided in SM (Sections E2.3 and E2.4).

Toxicity

At the last follow-up, grade 3 (G3) neuropathy occurred in 3 patients (6.5%), with cumulative EQD2 D1 doses to the cauda equina of 88.79 Gy, 89.52 Gy, and 78.11 Gy. Four patients (8.7%) experienced G3 anorectal dysfunction related to neuropathy. Of these, in 2 affected patients, cumulative EQD2 D1 doses to the cauda equina were 171.44 Gy and 102.61 Gy; dose reconstruction was not feasible in the remaining 2 because of unavailable primary-treatment plans. Notably, 5 patients (10.9%) had pre-existing anorectal neuropathy at baseline. Of these, 1 patient actually showed symptomatic improvement after reirradiation. For patients with pre-existing anorectal neuropathy not reporting severe symptoms or G4 toxicity, the minor clinical variations after reirradiation could not be conclusively distinguished from their baseline condition. G3 skin toxicity was observed in 3 patients (6.5%). One patient had a cumulative EQD2 D1 skin dose of 82.29 Gy, while a cumulative dose assessment was not possible in the other 2 because the primary-treatment plans were unavailable. The incidence of G2 pain increased from 2 to 10 patients (21.7%), and G2 bladder dysfunction rose from 6 to 16 patients (34.8%). Bone fractures within the irradiated volume were reported in 14 patients (30%): 10 (21.7%) were pre-existing at baseline because of the previous treatment, and 4 (8.7%) new bone fractures developed post-CIR. None of these patients required any intervention. At last follow-up, the most frequent adverse events were grade 2 neuropathy (43.5%), grade 1 pain (39.1%), and grade 2 skin toxicity (34.8%). Complete functional preservation (grade 0) was achieved for the urinary (37.0%) and anorectal (43.5%) function, and 60.9% of patients remained free of bone fractures. Notably, no grade 4 toxicities were recorded.

No significant correlations were observed between cumulative OAR doses and toxicity. A trend was noted for skin dose: overall skin toxicity showed a marginal association (odds ratio [OR], 0.97; 95% CI, 0.95-1.01; P = .067), with G1 (OR, 0.969; 95% CI, 0.94-1.00; P = .0686) and G2 (OR, 0.99; 95% CI, 0.93-1.05; P = .7046) skin toxicity showing increasing risk. Dose-effect analysis suggested a threshold effect: for cumulative skin doses ≤58 Gy, toxicity remained limited, whereas doses >58 Gy were associated with a marked increase in higher-grade events (G2-G3) (Fig. 4).

Figure 4.

Association between cumulative skin dose and late skin toxicity. A dose-effect relationship was observed. Toxicity remains low below 58 Gy, whereas the probability of ≥G2 events increases markedly above this threshold.

Discussion

Although advanced RT improves retreatment feasibility, data regarding reirradiation for rare, heterogeneous sarcomas remain limited. This retrospective study demonstrates that high-dose CIR is a feasible and effective salvage strategy for patients with recurrent pelvic sarcomas. In our cohort of 46 patients, the treatment provided encouraging oncologic outcomes, including a 3-year OS of 83.1% and a median survival time of 30.9 months from reirradiation. Although LRPFS declined rapidly after the first year, CIR achieved manageable toxicity with no reported grade 4 adverse events.

Furthermore, the 3-year OS exceeding 80% is highly notable, considering that all patients had at least 1 prior locoregional relapse and is comparable with primary-treatment series.^10,28 These outcomes emphasize the value of CIR as a possible strategy for patients with limited therapeutic alternatives.

This study is primarily limited by its retrospective design, small sample size, and histologic heterogeneity. In addition, a methodological weakness of this study relates to the 5 patients who were lost to follow-up because of the onset of distant metastasis and were therefore referred to oncology services. Given that distant progression may be associated with a worse prognosis and an increased risk of death, the censoring mechanism could be informative.

The marked decline in LRPFS between the first and second years—from 79.1% to 43.9%—reflects the biological aggressiveness of these tumors and the challenges of achieving durable control after salvage therapy. These findings align with previous reports showing that most local recurrences occur within the first 24 months after treatment; similarly, the time of first recurrence observed in our cohort reflects this trend. Moreover, our 2-year LRPFS of 43.9% and median locoregional survival time of 23.5 months are consistent with prior reirradiation studies, further supporting the potential utility of reirradiation in this setting.^20,29, 30, 31 This early and rapid decline in control may also reflect tumor evolution over time, as recurrent disease often acquires additional genomic alterations and exhibits increased biological complexity, potentially contributing to greater treatment resistance and more aggressive clinical behavior.^32,33

Interestingly, although not statistically significant, a clinically relevant trend emerged by recurrence pattern: in-field relapses achieved a longer median time to progression than out-of-field failures (24.5 vs 12.2 months). This pattern suggests that in-field recurrences represent locally persistent disease with low metastatic potential. Consequently, geographically confined, locally aggressive tumors may derive greater benefit from high-dose salvage therapies, such as CIR, to secure durable local control. This observation is supported by biological hallmarks of chordoma, the most representative histology in our study (80.4%), which demonstrated 1-, 2-, and 3-year LRPFS rates of 77.2%, 40.5%, and 22.1%, respectively, whereas the 1-, 2-, and 3-year OS rates were 100%, 92.3%, and 88.1%, respectively. Given the high local aggressiveness and low propensity for fast systemic spread, chordoma appear ideal candidates for high-dose salvage CIR, a hypothesis warranting further investigation.

We did not observe significant correlations between cumulative OAR doses and toxicity, except for skin, where a threshold effect at 58 Gy emerged. Beyond this level, the risk of G2 to G3 events increased significantly. This finding is clinically important and underscores the importance of careful dosimetric optimization in pelvic reirradiation.

To our knowledge, this is the first study specifically investigating high-dose CIR for recurrent pelvic sarcomas—a patient population with limited treatment options and generally poor prognosis. However, comparative data from other pelvic reirradiation studies provide useful context. For instance, Yamada et al³⁴ reported a 3-year OS of 61% and locoregional control of 69% in recurrent rectal cancer treated with CIR, with 20% G3 late toxicity and no grade ≥4 events. Despite differences in histology and dose (70.4 GyRBE in their study vs a median of 64 GyRBE in ours), their results support the role of CIR as an effective and well-tolerable salvage approach for pelvic recurrences.

Additional insight comes from Hayashi et al³⁵, who reported outcomes of 48 patients reirradiated with CIRT for recurrent head and neck cancers, including sarcomas (12.5%). The median dose for initial CIRT and reirradiation was 57.6 Gy and 54.0 GyRBE, respectively. They reported 2-year OS and locoregional control rates of 59.6% and 33.5%, respectively, but a higher rate of grade ≥3 toxicity (37.5%), including 1 G5 event.³⁵ Compared with our lower toxicity (10.8% G3 events, no ≥G4), this discrepancy may reflect differences in anatomic complexity, as the pelvis region allows for greater flexibility in dose distribution and organ sparing. Similarly, their median cumulative EQD2 dose was 136.3 Gy (range, 115.3-159.1 Gy), compared with our 130 Gy (range, 27-244 Gy). Moreover, multivariate analysis also revealed that the site of failure at reirradiation was a significant prognostic factor of progression-free survival (P = .042), with 2-year progression-free survival of 38.6% for in-field and 19.1% for marginal recurrences—closely mirroring our trends, although without statistical significance in our series.

The Heidelberg experience in head and neck chordomas and chondrosarcomas also supports our findings. In their retrospective series, 25 patients were treated with CIR (including 12 after prior CIRT), receiving a median dose of 51 GyRBE (range, 45-60 GyRBE). EQD2 was used to assess the cumulative Dmax. The 2-year local control rate was 79.3%, slightly higher than that observed in our cohort, possibly because of stricter patient selection or a shorter interval to reirradiation. Similar to our study, they investigated whether a total dose of up to 51 GyRBE was associated with improved local control rates, but did not find a statistically significant correlation.³⁶

Furthermore, some reports lack essential information on image registration, target delineation, and the cumulative doses delivered to both the target and OARs. Currently, there are no standardized protocols or cumulative dose constraints for pelvic reirradiation, and the integration of different modalities (photons, protons, carbon ions) further complicates interpretation. Although EQD2 or biologically effective dose calculations are often used to harmonize different radiation types, these models were developed for low-LET radiation and may not fully reflect the biological complexity of carbon ions. Although EQD2 offers a practical approximation for cumulative dose estimation, biological modeling for carbon ions is complex and remains incompletely represented by conventional linear-quadratic models.³⁷ An additional critical concern is the uncertainty in cumulative dose to pelvic OARs, particularly mobile organs such as the small bowel, sigmoid colon, and rectum. Organ motion and variable filling can result in substantial deviations from planned dose distributions, making the assessment of cumulative dose and associated toxicity challenging.

Surgical outcomes for recurrent pelvic sarcomas remain suboptimal. Stacchiotti et al¹⁷ reported on 82 patients with recurrent chordoma of the mobile spine and sacrum, where 5-year OS after surgery dropped to 50% following progression, and 5-year local relapse-free survival was 47% with a median local relapse-free survival of 57 months. Similarly, Laitinen et al³⁸ studied 126 patients undergoing surgery for local recurrence of chondrosarcoma of the pelvis and limb bone, with 1- and 5-year disease-free survival rates of 62.5% and 45.5%, respectively. However, neither study reported on treatment-related toxicity. Notably, in our cohort, no toxicities exceeded G3.

Our findings could be particularly relevant for specialized referral centers managing rare and radioresistant pelvic malignancies. Because the majority of our cohort consisted of chordomas, the outcomes may be most representative for this specific histology. However, the biological advantages and highly conformal dose distribution suggest that these findings can be cautiously generalized to other radioresistant pelvic sarcomas requiring aggressive local treatment. Although the limited global availability of CIRT facilities remains a barrier to widespread implementation, our data provide a critical evidence-based benchmark for multidisciplinary teams worldwide when considering high-dose salvage options for complex pelvic recurrences. The physical and biological properties of carbon ions—particularly high LET and enhanced RBE—make them especially suitable for reirradiating radioresistant histologies, where conventional approaches are often limited.

Conclusions

Our results support high-dose CIR as a feasible and effective salvage option for recurrent pelvic sarcomas, providing encouraging oncologic outcomes with manageable toxicity in a heavily pretreated population. These data support integrating CIR into multidisciplinary care for patients with limited alternatives and underscore the need for prospective, multicenter studies to refine dose-response models and establish evidence-based cumulative dose constraints.

Declaration of interests

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.