Patient selection and toxicity predictors in proton reirradiation for locoregional breast cancer recurrence Authors: Molly A. Chakraborty, Charlie White, Atif J. Khan, Siyu Yang, Zhigang Zhang, Francis Yu, Haibo Lin, Lior Z. Braunstein, Diana Roth O'Brien, John J. Cuaron, David M. Guttmann, Quincey LaPlant, Simon N. Powell, Oren Cahlon, J. Isabelle Choi Published: 2026 ABSTRACT PurposePatients with breast cancer isolated local-regional recurrence (iLRR) can benefit from salvage reirradiation (reRT) to provide durable control and prevent further progression. Conformal radiation techniques like pencil beam scanning (PBS) proton beam therapy (PBT) may improve the reRT therapeutic ratio by limiting cumulative dose and improving target region dose delivery.Materials and methodsConsecutive patients between 8/2019–9/2025 treated with PBS-PBT reRT for breast cancer iLRR with prior dose overlap were identified. Clinicopathologic, treatment, and outcomes details were collected. Associations between clinical/treatment factors and time to recurrence and grade (G) ≥2 adverse events (AEs) were evaluated with Fine-Gray models and logistic regression, respectively. Correlations between DVH metrics and late AEs were assessed using Kruskal-Wallis and Wilcoxon rank sum tests.ResultsSeventy patients were identified with median first course (C1), reRT course (C2), and cumulative doses of 53, 50, and 106Gy(RBE), respectively. During C2, 82.2% received comprehensive nodal irradiation. Median follow-up was 30 months. Two- and three-year rates of any recurrence were 14% and 17%, and OS were 100% and 95%, respectively. Maximum acute G2 and G3 AEs occurred in 50% and 7%, and late G2 and G3 AEs in 34% and 10%, respectively. Rib fractures developed in 21.1% and were associated with C1 and C2 chest wall D200cc and C1 rib D10cc.ConclusionIn the largest experience of PBT reRT for breast cancer iLRR to date, we report promising disease control and limited severe AEs despite a high-risk patient cohort. We provide dose metrics that can serve as the foundation for reRT dose constraints and be further refined in future reRT investigations. Introduction With continued advances in multimodality therapy for breast cancer, long-term survival has significantly improved in recent decades [1]. However, approximately 10–15% of patients with invasive breast cancer will still develop an isolated locoregional recurrence (iLRR) with modern multidisciplinary treatment, impacting a high absolute number of women given the overall incidence of breast cancer [2]. [3] Thus, there is an urgent need to optimize salvage treatment approaches. Historical treatment approaches for iLRR have included salvage surgery and systemic therapy. Radiotherapy (RT) is commonly delivered for primary breast cancer, with up to two-thirds of patients with early-stage breast cancer receiving breast conserving therapy (BCT), bringing the question of the clinical implications and optimal delivery approach of reirradiation (reRT) to the forefront [4]. ReRT has been used infrequently for breast cancer iLRR due to concern for severe adverse events (AEs) from cumulative doses to organs-of-interest (OOI), such as the heart, lungs, chest wall (CW), ribs, brachial plexus, skin, and soft tissue. However, without additional RT, repeat disease recurrence after surgery alone is common; up to one-third of patients develop another ipsilateral breast tumor recurrence (IBTR) and up to 50–70% develop uncontrolled disease [5], [6], [7], [8]. For limited IBTR, repeat BCT with partial breast reRT using brachytherapy or external beam RT has demonstrated promising results and is being increasingly applied in clinical practice [9], [10], [11], [12]. For iLRR with high risk features and in the post-mastectomy setting, more comprehensive reRT may be needed due to aberrant lymphatic drainage from multiple prior interventions, including surgeries and RT. Larger reRT target volumes pose inherent risks to surrounding OOI and previously-irradiated tissues, for which more conformal RT modalities such as proton beam therapy (PBT) may be beneficial to reduce dose exposure to previously-irradiated non-target tissues and optimize dose delivery to regions at risk for harboring microscopic disease [13], [14]. Modern pencil beam scanning (PBS)-PBT with soft tissue-based image guidance is particularly well-suited for reRT delivery to achieve a higher level of dose conformality. In this study, we assess patients who were treated with PBS-PBT reRT for breast cancer iLRR to characterize disease control and acute and late AE rates using our institutional standard approach. We explore potential treatment and DVH parameters that may be associated with severe AEs to provide a foundation for future study and development of reRT plannning guidelines. Methods and materials Patients from XXXX treated with PBS-PBT at XXXX between August 2019 and September 2025 for recurrent breast cancer with dose overlap from prior breast cancer RT were retrospectively identified on an IRB-approved protocol. Patients who previously received RT for a non-breast cancer diagnosis and those with distant metastatic disease were excluded. Patient demographics, disease characteristics, treatment details and clinical outcomes were collected from institutional electronic health records. AEs were graded using CTCAE v5.0. Prior RT details and DICOM-RT files were obtained and reviewed when available. Patients typically underwent clinical follow-up at 3 months, 6 months, and 1 year after RT completion, and annually thereafter. Treatment site photographs were obtained throughout the treatment course and at each follow-up visit if the patient was amenable to collection. AEs were collected in a standard template at every visit. All patients were treated on a Varian ProBeam system (Varian Medical Systems, Palo Alto, CA, USA) using PBS and were immobilized with a wing board. Both arms were raised overhead regardless of unilateral or bilateral breast treatment. A free-breathing approach was used for all patients unless a high degree of overlap at the CW-heart interface was noted, in which case a deep-inspiration breast hold technique was used. Target delineation was performed using the RadComp contouring atlas, with modifications applied in select cases according to individual patient anatomy and physician clinical judgment [15]. The clinical target volume (CTV) was cropped from the skin as clinically indicated (5 mm post-lumpectomy, 3 mm post-mastectomy, 2 mm post-mastectomy with high risk for dermal/skin involvement, 0 mm for inflammatory disease). Treatments were delivered using two to four en face fields with 5 cm water-equivalent range shifters and employed a combination of single-field optimization (SFO) and multifield optimization (MFO), with robustness criteria of 5 mm and 3.5% range uncertainty. Plans were generated in Eclipse (Varian Medical Systems, Palo Alto, CA) prior to 9/2023 and with RayStation (RaySearch Laboratories, Stockholm, Sweden) thereafter. Plans were evaluated using robust optimization to ensure V95 > 95% across all CTV permutations. OOI dose constraints were individualized based on prior treatments and current clinical presentation. Patient setup utilized surface tracking, daily orthogonal radiographs, and cone beam computed tomography. Mid-treatment repeat simulations were performed to verify reproducibility and assess for significant anatomic changes; adaptive plans were generated if the dose distribution was impacted. When prior RT plans with DICOM-RT files were available, deformable image registration was performed between the current PBT and prior CT simulation images using MIM (MIM Software, Cleveland, OH, USA) for cumulative plan generation. Cumulative doses were summed across all RT courses in the treatment planning system. Degree of overlap (none, partial, complete) of the breast, CW, axilla, supraclavicular region (SCV), and internal mammary lymph nodes (IMN) were defined based on prior and current treatment volumes. Patients treated with partial breast irradiation (PBI) were defined as receiving RT to partial breast; whole breast (WB) RT in the prone position = WB; WB RT in the supine position = WB and partial axillary LN, regional nodal irradiation (RNI) = whole axilla, SCV, IMN (unless otherwise specified), post-mastectomy RT = whole CW, axilla, SCV, IMN. For example, if C1 was WB RT in the supine position and C2 was post-mastectomy radiotherapy, this was defined as complete CW overlap and partial axillary overlap. The rib structure was auto-contoured in the treatment planning system. The CW structure was delineated for all patients on the PBT reRT planning scan and on all available prior RT planning scans, and was defined as ribs, pectoralis major, pectoralis minor, and intercostal muscles of the ipsilateral hemithorax. Categorical and continuous variables were described using frequency and percentage, median and interquartile range (IQR), respectively. Time to recurrence was estimated using the Fine-Gray method and was defined as time from reRT completion to recurrence Death without recurrence was treated as a competing risk. Patients who were alive and recurrence-free were censored at last follow-up. Acute AEs were defined as occurring ≤3 months following RT completion, and late AEs as extending beyond or starting >3 months after RT completion. Competing risks regression and logistic regression were used to evaluate associations between clinical/treatment factors and time to recurrence and Grade (G) ≥2 AEs, respectively. If models did not converge due to sparse numbers, Gray's test or Fisher's exact test were utilized. Correlations between dose-volume histogram (DVH) metrics and late AEs were assessed using Kruskal-Wallis rank sum test and Wilcoxon rank sum test. Median follow-up was estimated using the Reverse Kaplan-Meier method. P-value <0.05 was considered statistically significant. Statistical analysis was performed using R version 4.4.3. Results In total, 70 consecutive patients were identified who met study criteria. The median age at reRT was 63 years (52–68); median time between RT courses was 89 months (38–159). Patient and disease characteristics during the first (C1) and reRT (C2) course are listed in Table 1. At C1, 30% had G3 disease, and 22.9% had lymphovascular invasion (LVI) present. Approximately one-quarter had node-positive disease. Forty-seven percent of patients underwent chemotherapy, and 70% received hormone therapy. At C2, 31.4% had stage 3 disease; high-risk features were present in the majority, including nodal involvement (50.0%), LVI (28.6%), dermal invasion (28.6%), extracapsular extension (ECE) (21.4%), nipple stroma invasion (11.4%), and skeletal muscle invasion (11.4%). At recurrence, surgical management with mastectomy +/− nodal evaluation occurred in 38.6%, lumpectomy +/− nodal evaluation in 18.6%, and CW excision in 14.3%. Approximately 64% of patients underwent chemotherapy at C2, and 57% received hormone therapy.Table 1Initial (C1) and reRT (C2) Course Baseline Characteristics.CourseCharacteristicC1 N = 701,2C2 N = 701,2LateralityBilateral0 (0%)1 (1.4%)Left34 (49%)34 (49%)Right36 (51%)35 (50%) HistologyDCIS8 (11.4%)1 (1.4%)IDC48 (68.6%)43 (61%)ILC7 (10%)9 (13%)Invasive Carcinoma5 (7.1%)17 (24%) Grade14 (5.7%)0 (0%)219 (27.4%)31 (44.3%)321 (30.0%)26 (37.1%) LVINo26 (37.1%)27 (38.6%)Yes16 (22.9%)20 (28.6%) Receptor Subtype+/+/−40 (57.1%)39 (55.7%)+/−/−1 (1.4%)13 (18.6%)−/+/−1 (1.4%)0 (0%)+/−/+0 (0%)1 (1.4%)+/+/+2 (2.9%)1 (1.4%)−/−/+1 (1.4%)1 (1.4%)−/−/−11 (15.7%)12 (17.1%) MarginsNegative49 (70%)40 (57.1%)Positive2 (2.9%)8 (11.4%) pT StageT0/Tis13 (18.6%)24 (34.3%)T1-T249 (70%)35 (50%)T3-T42 (2.9%)7 (10%)pN StageN0/Nx45 (64.3%)19 (27.1%)N115 (21.4%)15 (21.4%)N23 (4.3%)14 (20%)N30 (0%)6 (8.6%) Overall Stage09 (12.9%)3 (4.3%)I49 (70%)37 (52.9%)III3 (4.3%)22 (31.4%)ECENo13 (18.6%)18 (25.7%)Yes5 (7.1%)15 (21.4%) Dermal InvasionNo8 (11.4%)21 (30%)Yes4 (5.7%)20 (28.6%) Nipple Stroma InvasionNo9 (12.9%)26 (37.1%)Yes1 (1.4%)8 (11.4%) Skeletal M InvasionNo10 (14.3%)15 (21.4%)Yes0 (0%)8 (11.4%) Surgery TypeALND0 (0%)12 (17%)Axillary Excision0 (0%)1 (1.4%)Biopsy only0 (0%)5 (7.1%)CW Excision1 (1.4%)10 (14%)IMN Dissection0 (0%)1 (1.4%)Lumpectomy11 (15.7%)8 (11%)Lumpectomy/ALND10 (14.3%)2 (2.9%)Lumpectomy/SLNB37 (52.9%)3 (4.3%)Mastectomy3 (4.3%)9 (13%)Mastectomy/ALND2 (2.9%)8 (11%)Mastectomy/IMN dissection0 (0%)1 (1.4%)Mastectomy/SLNB5 (7.1%)9 (13%)Video-assisted Thorascopic Surgery0 (0%)1 (1.4%) ChemoNo37 (53%)24 (34.2%)Yes33 (47%)45 (64.3%) HTNo21 (30%)30 (43%)Yes49 (70%)40 (57%)1n (%).[2]Data not shown denotes missing information.Abbreviations: DCIS = ductal carcinoma in situ, IDC = invasive ductal carcinoma, ILC = invasive lobular carcinoma, LVI = lymphovascular invasion, ER = estrogen receptor, PR = progesterone receptor, HER2 = human epidermal growth factorreceptor 2, ECE = extracapsular extension, ALND = axillary lymph node dissection, CW = chest wall, IMN = internal mammary lymph node, SLNB = sentinel lymph node biopsy. Table 1Initial (C1) and reRT (C2) Course Baseline Characteristics.CourseCharacteristicC1 N = 701,2C2 N = 701,2LateralityBilateral0 (0%)1 (1.4%)Left34 (49%)34 (49%)Right36 (51%)35 (50%) HistologyDCIS8 (11.4%)1 (1.4%)IDC48 (68.6%)43 (61%)ILC7 (10%)9 (13%)Invasive Carcinoma5 (7.1%)17 (24%) Grade14 (5.7%)0 (0%)219 (27.4%)31 (44.3%)321 (30.0%)26 (37.1%) LVINo26 (37.1%)27 (38.6%)Yes16 (22.9%)20 (28.6%) Receptor Subtype+/+/−40 (57.1%)39 (55.7%)+/−/−1 (1.4%)13 (18.6%)−/+/−1 (1.4%)0 (0%)+/−/+0 (0%)1 (1.4%)+/+/+2 (2.9%)1 (1.4%)−/−/+1 (1.4%)1 (1.4%)−/−/−11 (15.7%)12 (17.1%) MarginsNegative49 (70%)40 (57.1%)Positive2 (2.9%)8 (11.4%) pT StageT0/Tis13 (18.6%)24 (34.3%)T1-T249 (70%)35 (50%)T3-T42 (2.9%)7 (10%)pN StageN0/Nx45 (64.3%)19 (27.1%)N115 (21.4%)15 (21.4%)N23 (4.3%)14 (20%)N30 (0%)6 (8.6%) Overall Stage09 (12.9%)3 (4.3%)I49 (70%)37 (52.9%)III3 (4.3%)22 (31.4%)ECENo13 (18.6%)18 (25.7%)Yes5 (7.1%)15 (21.4%) Dermal InvasionNo8 (11.4%)21 (30%)Yes4 (5.7%)20 (28.6%) Nipple Stroma InvasionNo9 (12.9%)26 (37.1%)Yes1 (1.4%)8 (11.4%) Skeletal M InvasionNo10 (14.3%)15 (21.4%)Yes0 (0%)8 (11.4%) Surgery TypeALND0 (0%)12 (17%)Axillary Excision0 (0%)1 (1.4%)Biopsy only0 (0%)5 (7.1%)CW Excision1 (1.4%)10 (14%)IMN Dissection0 (0%)1 (1.4%)Lumpectomy11 (15.7%)8 (11%)Lumpectomy/ALND10 (14.3%)2 (2.9%)Lumpectomy/SLNB37 (52.9%)3 (4.3%)Mastectomy3 (4.3%)9 (13%)Mastectomy/ALND2 (2.9%)8 (11%)Mastectomy/IMN dissection0 (0%)1 (1.4%)Mastectomy/SLNB5 (7.1%)9 (13%)Video-assisted Thorascopic Surgery0 (0%)1 (1.4%) ChemoNo37 (53%)24 (34.2%)Yes33 (47%)45 (64.3%) HTNo21 (30%)30 (43%)Yes49 (70%)40 (57%)1n (%).[2]Data not shown denotes missing information.Abbreviations: DCIS = ductal carcinoma in situ, IDC = invasive ductal carcinoma, ILC = invasive lobular carcinoma, LVI = lymphovascular invasion, ER = estrogen receptor, PR = progesterone receptor, HER2 = human epidermal growth factorreceptor 2, ECE = extracapsular extension, ALND = axillary lymph node dissection, CW = chest wall, IMN = internal mammary lymph node, SLNB = sentinel lymph node biopsy. Initial (C1) and reRT (C2) Course Baseline Characteristics. 1n (%). [2]Data not shown denotes missing information. Abbreviations: DCIS = ductal carcinoma in situ, IDC = invasive ductal carcinoma, ILC = invasive lobular carcinoma, LVI = lymphovascular invasion, ER = estrogen receptor, PR = progesterone receptor, HER2 = human epidermal growth factor receptor 2, ECE = extracapsular extension, ALND = axillary lymph node dissection, CW = chest wall, IMN = internal mammary lymph node, SLNB = sentinel lymph node biopsy. C1 and C2 RT details are described in Table 2. Thirty-one patients (44.3%) had C1 RT plans available for cumulative dose evaluation. Median C1 and C2 total dose including boost was 53Gy and 50Gy (relative biological effectiveness (RBE)), respectively. Median cumulative dose was 107Gy(RBE). During C1, 61% received WB RT, 11.4% RNI with WB or CW, 11.5% partial nodal RT with WB/CW, and 5.7% CW only. Only 3 patients (4.3%) received PBI. At C2, most (82.9%) received RNI +/− WB/CW, 4.3% WB/CW only, 8.6 PBI, and 4.3% partial node RT +/− WB/CW. C1, C2, and cumulative DVH and OOI metrics are summarized in Supp Table 1. Median cumulative heart mean dose was 1.7Gy(RBE) (IQR: 0.7–3.6). Median cumulative ipsilateral lung receiving 20Gy (V20Gy) was 21% (IQR: 15.0–28.0) and V5Gy was 49.0% (IQR: 40.0–53.0). The median cumulative maximum brachial plexus dose was 53.6Gy(RBE) (IQR: 51.5–76.8).Table 2Initial (C1) and reRT (C2) radiation treatment characteristics.CourseCharacteristicC1 N = 701,2C2 N = 701,2RT Dose Initial (Gy(RBE))45.0 (42.4, 50.0)50.4 (45.0, 50.4)Boost Dose (Gy(RBE))10.0 (0.0, 14.0)14.0 (10.0, 14.4)Total Dose (Gy(RBE))53 (50, 60)50 (50, 59)Cumulative Dose (Gy(RBE))107 (100,115)Time from end of C1 to C2 (months)89 (38–159)RT VolumesIORT1 (1.4%)0 (0%)PBI3 (4.3%)6 (8.6%)WB/WB (Recon)45 (64.9%)0 (0%)WB/WB (Recon) + RNI4 (5.8%)7 (24.3%)WB + partial nodes3 (4.3%)0 (0%)Partial CW0 (0%)1 (1.4%)CW3 (4.3%)0 (0%)CW (Recon)1 (1.4%)2 (2.9%)CW (Recon) + partial nodes1 (1.4%)1 (1.4%)CW (Recon) + RNI2 (2.9%)18 (26%)CW + partial nodes4 (5.8%)0 (0%)CW + RNI2 (2.9%)22 (31%)Partial nodes (Recon)0 (0%)2 (2.9%)RNI0 (0%)1 (1.4%)Boost VolumeLumpectomy cavity42 (88%)1 (4.2%)Lumpectomy cavity and grossly involved LNs0 (0%)1 (4.2%)Tumor bed0 (0%)7 (29%)Scar6 (13%)0 (0%)CW0 (0%)3 (13%)CW/Node0 (0%)1 (4.2%)Axilla0 (0%)2 (8.3%)Axilla, IMN, SCV0 (0%)1 (4.2%)RNI + Lumpectomy Cavity0 (0%)1 (4.2%)IMN0 (0%)5 (21%)Sternum + IMN0 (0%)1 (4.2%)[1]Median (IQR); n (%).[2]Data not shown denotes missing information.Abbreviations: Gy = Gray, RBE = relative biological effectiveness, IORT = intraoperative radiotherapy, PBI = partial breast irradiation; WB = whole breast, recon = reconstruction, CW = chest wall, RNI = regional nodal irradiation, LN = lymph node, IMN = internal mammary lymph node, SCV = supraclavicular region. Table 2Initial (C1) and reRT (C2) radiation treatment characteristics.CourseCharacteristicC1 N = 701,2C2 N = 701,2RT Dose Initial (Gy(RBE))45.0 (42.4, 50.0)50.4 (45.0, 50.4)Boost Dose (Gy(RBE))10.0 (0.0, 14.0)14.0 (10.0, 14.4)Total Dose (Gy(RBE))53 (50, 60)50 (50, 59)Cumulative Dose (Gy(RBE))107 (100,115)Time from end of C1 to C2 (months)89 (38–159)RT VolumesIORT1 (1.4%)0 (0%)PBI3 (4.3%)6 (8.6%)WB/WB (Recon)45 (64.9%)0 (0%)WB/WB (Recon) + RNI4 (5.8%)7 (24.3%)WB + partial nodes3 (4.3%)0 (0%)Partial CW0 (0%)1 (1.4%)CW3 (4.3%)0 (0%)CW (Recon)1 (1.4%)2 (2.9%)CW (Recon) + partial nodes1 (1.4%)1 (1.4%)CW (Recon) + RNI2 (2.9%)18 (26%)CW + partial nodes4 (5.8%)0 (0%)CW + RNI2 (2.9%)22 (31%)Partial nodes (Recon)0 (0%)2 (2.9%)RNI0 (0%)1 (1.4%)Boost VolumeLumpectomy cavity42 (88%)1 (4.2%)Lumpectomy cavity and grossly involved LNs0 (0%)1 (4.2%)Tumor bed0 (0%)7 (29%)Scar6 (13%)0 (0%)CW0 (0%)3 (13%)CW/Node0 (0%)1 (4.2%)Axilla0 (0%)2 (8.3%)Axilla, IMN, SCV0 (0%)1 (4.2%)RNI + Lumpectomy Cavity0 (0%)1 (4.2%)IMN0 (0%)5 (21%)Sternum + IMN0 (0%)1 (4.2%)[1]Median (IQR); n (%).[2]Data not shown denotes missing information.Abbreviations: Gy = Gray, RBE = relative biological effectiveness, IORT = intraoperative radiotherapy, PBI = partial breast irradiation; WB = whole breast, recon = reconstruction, CW = chest wall, RNI = regional nodal irradiation, LN = lymph node, IMN = internal mammary lymph node, SCV = supraclavicular region. Initial (C1) and reRT (C2) radiation treatment characteristics. [1]Median (IQR); n (%). [2]Data not shown denotes missing information. Abbreviations: Gy = Gray, RBE = relative biological effectiveness, IORT = intraoperative radiotherapy, PBI = partial breast irradiation; WB = whole breast, recon = reconstruction, CW = chest wall, RNI = regional nodal irradiation, LN = lymph node, IMN = internal mammary lymph node, SCV = supraclavicular region. The median follow-up was 30 months (range: 1–73). Overall survival (OS) at 2 and 3 years was 100% (95.0%CI: 100%–100%) and 95% (95%CI: 89.0%–100%), respectively. The 2- and 3-year any recurrence rate was 14.0% (95%CI: 6.5%–25.0%) and 17.0% (95%CI: 8.0%–29.0%), respectively. (Supp Fig. 1) 7.1% experienced local recurrence (LR), 1.4% regional recurrence (RR), 8.6% distant recurrence (DR), and 13% any recurrence. (Table 3, Supp Table 2). Most recurrences developed within 2 years after C2. (Supp Fig. 1) Dermal invasion at the time of reRT was associated with an increased risk of any recurrence (p = 0.004).Table 3Clinicopathologic and treatment characteristics for patients who developed a locoregional recurrence after reirradiation.Patient 3Patient 37Patient 50Patient 60Patient 64Age (years)6559794371Stage at RecurrenceyIIAIIICIIICIIIBIIALocation of iLRR•Axillary LN•Breast•Axillary LN•Breast•Axillary LN•IMN•Breast•Axillary LN•BreastHigh Risk Features•Dermal invasion•Dermal invasion•Positive surgical margin•ECE•Dermal invasion•Skeletal muscle invasion•Positive surgical margin•Grade 3•LVI•Dermal invasion•Triple negative•BRCA2+•Grade 3•Dermal invasion•Triple negativeSurgical Intervention for RecurrenceALNDMastectomy/ALNDLumpectomyMastectomy/SLNBLumpectomySystemic Therapy for RecurrenceYesYesYesYesYesC1 Treatment VolumesNo recordWB + cavity boostWB + cavity boostWB + SCV + cavity boostWB + cavity boostC1 Treatment Dose (Gy)No record42.4 (WB) +10 (cavity boost)46 (WB) +14 (cavity boost)50 (WB + SCV) + 10 (cavity boost)42.56 (WB) +10 (cavity boost)Time between RT courses267.8 months136.0 months158.7 months26.9 months28.9 monthsC2 Treatment VolumesWB + RNICW + RNIWB + RNICW + RNIWB + RNIC2 Treatment Dose (Gy(RBE))50.450.450 (WB) +10 (cavity and grossly involved nodes)48.645 (WB) +10.4 (cavity and regional LN)Location of Recurrence after reRT•Breast•Chest wall•Mediastinal LN•Bone•Sternum•Mediastinal LN•Lung•Chest Wall•BreastRecurrence In or Out of C2 Volume•Within C2 Volume•Within C2 Volume•Out of C2 Volume•Distant•Out of C2 Volume•Distant•Within C2 Volume•Within C2 VolumeTime to Recurrence22 months21 months15 months7 months4 monthsSalvage TreatmentMastectomyXelodaFulvestrant PalbociclibMastectomySacituzumabMastectomyAbbreviations: iLRR = isolated locoregional recurrence, C = course, Gy = Gray), mo = months, RBE = relative biological effectiveness, reRT = reirradiation, LN = lymph node, ALND = axillary lymph node dissection, WB = whole breast, SCV = supraclavicular region, IMN = internal mammary lymph nodes, RNI = regional nodal irradiation, SLNB = sentinel lymph node biopsy, ECE = extracapsular extension, LVI = lymphovascular invasion. Table 3Clinicopathologic and treatment characteristics for patients who developed a locoregional recurrence after reirradiation.Patient 3Patient 37Patient 50Patient 60Patient 64Age (years)6559794371Stage at RecurrenceyIIAIIICIIICIIIBIIALocation of iLRR•Axillary LN•Breast•Axillary LN•Breast•Axillary LN•IMN•Breast•Axillary LN•BreastHigh Risk Features•Dermal invasion•Dermal invasion•Positive surgical margin•ECE•Dermal invasion•Skeletal muscle invasion•Positive surgical margin•Grade 3•LVI•Dermal invasion•Triple negative•BRCA2+•Grade 3•Dermal invasion•Triple negativeSurgical Intervention for RecurrenceALNDMastectomy/ALNDLumpectomyMastectomy/SLNBLumpectomySystemic Therapy for RecurrenceYesYesYesYesYesC1 Treatment VolumesNo recordWB + cavity boostWB + cavity boostWB + SCV + cavity boostWB + cavity boostC1 Treatment Dose (Gy)No record42.4 (WB) +10 (cavity boost)46 (WB) +14 (cavity boost)50 (WB + SCV) + 10 (cavity boost)42.56 (WB) +10 (cavity boost)Time between RT courses267.8 months136.0 months158.7 months26.9 months28.9 monthsC2 Treatment VolumesWB + RNICW + RNIWB + RNICW + RNIWB + RNIC2 Treatment Dose (Gy(RBE))50.450.450 (WB) +10 (cavity and grossly involved nodes)48.645 (WB) +10.4 (cavity and regional LN)Location of Recurrence after reRT•Breast•Chest wall•Mediastinal LN•Bone•Sternum•Mediastinal LN•Lung•Chest Wall•BreastRecurrence In or Out of C2 Volume•Within C2 Volume•Within C2 Volume•Out of C2 Volume•Distant•Out of C2 Volume•Distant•Within C2 Volume•Within C2 VolumeTime to Recurrence22 months21 months15 months7 months4 monthsSalvage TreatmentMastectomyXelodaFulvestrant PalbociclibMastectomySacituzumabMastectomyAbbreviations: iLRR = isolated locoregional recurrence, C = course, Gy = Gray), mo = months, RBE = relative biological effectiveness, reRT = reirradiation, LN = lymph node, ALND = axillary lymph node dissection, WB = whole breast, SCV = supraclavicular region, IMN = internal mammary lymph nodes, RNI = regional nodal irradiation, SLNB = sentinel lymph node biopsy, ECE = extracapsular extension, LVI = lymphovascular invasion. Clinicopathologic and treatment characteristics for patients who developed a locoregional recurrence after reirradiation. •Axillary LN Axillary LN •Breast Breast •Axillary LN Axillary LN •Breast Breast •Axillary LN Axillary LN •IMN IMN •Breast Breast •Axillary LN Axillary LN •Breast Breast •Dermal invasion Dermal invasion •Dermal invasion Dermal invasion •Positive surgical margin Positive surgical margin •ECE ECE •Dermal invasion Dermal invasion •Skeletal muscle invasion Skeletal muscle invasion •Positive surgical margin Positive surgical margin •Grade 3 Grade 3 •LVI LVI •Dermal invasion Dermal invasion •Triple negative Triple negative •BRCA2+ BRCA2+ •Grade 3 Grade 3 •Dermal invasion Dermal invasion •Triple negative Triple negative •Breast Breast •Chest wall Chest wall •Mediastinal LN Mediastinal LN •Bone Bone •Sternum Sternum •Mediastinal LN Mediastinal LN •Lung Lung •Chest Wall Chest Wall •Breast Breast •Within C2 Volume Within C2 Volume •Within C2 Volume Within C2 Volume •Out of C2 Volume Out of C2 Volume •Distant Distant •Out of C2 Volume Out of C2 Volume •Distant Distant •Within C2 Volume Within C2 Volume •Within C2 Volume Within C2 Volume Abbreviations: iLRR = isolated locoregional recurrence, C = course, Gy = Gray), mo = months, RBE = relative biological effectiveness, reRT = reirradiation, LN = lymph node, ALND = axillary lymph node dissection, WB = whole breast, SCV = supraclavicular region, IMN = internal mammary lymph nodes, RNI = regional nodal irradiation, SLNB = sentinel lymph node biopsy, ECE = extracapsular extension, LVI = lymphovascular invasion. Acute and late AE reRT details are described in Table 4. Maximum acute AE G1 was observed in 43%, G2 in 50% (radiation dermatitis 47%, CW pain 5.7%, hyperpigmentation 5.7%, skin pain: 5.7%, fatigue 5.7%, decreased shoulder range of motion 2.9%), and G3 in 7.1% (radiation dermatitis 5.7%, skin pain 1.4%). Maximum late AE G1 was observed in 51%, G2 in 34% (fibrosis 21%, CW pain 17%, shoulder stiffness 16%, arm lymphedema 10%, rib fracture 7.1%, wound complications 4.3%, breast pain 1.4%,) and G3 in 10% (fibrosis 5.7%, shoulder stiffness 4.3%, arm lymphedema 1.4%, breast/CW lymphedema 1.4%, wound complication 1.4%). Late rib fracture occurred in 21.1% (G1, n = 10, 14%; G2, n = 5, 7.1%, no ≥G3); 10 were diagnosed after imaging workup for related symptoms, whereas 4 were incidentally found on imaging. Median time to rib fracture diagnosis was 22 months after reRT (6–35). There were no cardiac or pulmonary acute or late AEs.Table 4Acute and late toxicities (CTCAE v5.0) for reRT (C2) course.Grade1Characteristic02122232Acute ToxicityBreast Edema68 (97%)2 (2.9%)0 (0%)0 (0%)Breast Pain66 (94%)4 (5.7%)0 (0%)0 (0%)CW Pain54 (77%)12 (17%)4 (5.7%)0 (0%)Decreased Shoulder ROM46 (66%)21 (30%)2 (2.9%)0 (0%)Dermatitis2 (2.9%)31 (44%)33 (47%)4 (5.7%)Esophagitis66 (94%)4 (5.7%)0 (0%)0 (0%)Fatigue16 (23%)50 (71%)4 (5.7%)0 (0%)Hyperpigmentation14 (20%)52 (74%)4 (5.7%)0 (0%)Pruritus47 (67%)23 (33%)0 (0%)0 (0%)Rib Fracture70 (100%)0 (0%)0 (0%)0 (0%)Skin Pain44 (63%)21 (30%)4 (5.7%)1 (1.4%)Late ToxicityArm Lymphedema37 (53%)23 (33%)7 (10%)1 (1.4%)Brachial Plexopathy65 (93%)3 (4.3%)0 (0%)0 (0%)Breast/CW Lymphedema49 (70%)18 (26%)0 (0%)1 (1.4%)Breast Pain60 (86%)7 (10%)1 (1.4%)0 (0%)CW Pain37 (53%)19 (27%)12 (17%)0 (0%)Soft Tissue Fibrosis23 (33%)26 (37%)15 (21%)4 (5.7%)Hyperpigmentation20 (29%)48 (69%)0 (0%)0 (0%)Osteonecrosis66 (94%)2 (2.9%)0 (0%)0 (0%)Rib Fracture53 (76%)10 (14%)5 (7.1%)0 (0%)Decreased Shoulder Range of Motion19 (27%)35 (50%)11 (16%)3 (4.3%)Telangiectasia60 (86%)8 (11%)0 (0%)0 (0%)Wound Complications61 (87%)1 (1.4%)3 (4.3%)1 (1.4%)[1]Data not shown denotes missing information.2n (%).Abbreviations: CW = chest wall, ROM = range of motion. Table 4Acute and late toxicities (CTCAE v5.0) for reRT (C2) course.Grade1Characteristic02122232Acute ToxicityBreast Edema68 (97%)2 (2.9%)0 (0%)0 (0%)Breast Pain66 (94%)4 (5.7%)0 (0%)0 (0%)CW Pain54 (77%)12 (17%)4 (5.7%)0 (0%)Decreased Shoulder ROM46 (66%)21 (30%)2 (2.9%)0 (0%)Dermatitis2 (2.9%)31 (44%)33 (47%)4 (5.7%)Esophagitis66 (94%)4 (5.7%)0 (0%)0 (0%)Fatigue16 (23%)50 (71%)4 (5.7%)0 (0%)Hyperpigmentation14 (20%)52 (74%)4 (5.7%)0 (0%)Pruritus47 (67%)23 (33%)0 (0%)0 (0%)Rib Fracture70 (100%)0 (0%)0 (0%)0 (0%)Skin Pain44 (63%)21 (30%)4 (5.7%)1 (1.4%)Late ToxicityArm Lymphedema37 (53%)23 (33%)7 (10%)1 (1.4%)Brachial Plexopathy65 (93%)3 (4.3%)0 (0%)0 (0%)Breast/CW Lymphedema49 (70%)18 (26%)0 (0%)1 (1.4%)Breast Pain60 (86%)7 (10%)1 (1.4%)0 (0%)CW Pain37 (53%)19 (27%)12 (17%)0 (0%)Soft Tissue Fibrosis23 (33%)26 (37%)15 (21%)4 (5.7%)Hyperpigmentation20 (29%)48 (69%)0 (0%)0 (0%)Osteonecrosis66 (94%)2 (2.9%)0 (0%)0 (0%)Rib Fracture53 (76%)10 (14%)5 (7.1%)0 (0%)Decreased Shoulder Range of Motion19 (27%)35 (50%)11 (16%)3 (4.3%)Telangiectasia60 (86%)8 (11%)0 (0%)0 (0%)Wound Complications61 (87%)1 (1.4%)3 (4.3%)1 (1.4%)[1]Data not shown denotes missing information.2n (%).Abbreviations: CW = chest wall, ROM = range of motion. Acute and late toxicities (CTCAE v5.0) for reRT (C2) course. [1]Data not shown denotes missing information. 2n (%). Abbreviations: CW = chest wall, ROM = range of motion. Acute reRT ≥G2 AE was associated with higher body mass index (BMI) (p = 0.028) and presence of ECE during C2 (p = 0.019). (Supp Table 3) Late ≥G2 AE was associated with CW and RNI during either RT course (p = 0.027 C1; p = 0.03 C2) and full breast (p = 0.005), CW (p = 0.003), axilla (p = 0.002), SCV (p = 0.006), and IMN (p = 0.008) overlap. (Table 5) Late ≥G2 CW AEs (CW pain, breast pain, telangiectasia, hyperpigmentation, breast/CW lymphedema, rib fracture, osteonecrosis, wound complications, fibrosis) were associated with full breast (p = 0.009), CW (p = 0.007), axillary (p = 0.009), SCV (p = 0.034), and IMN overlap (p = 0.022). (Supp Table 4) Late ≥G2 joint AEs (shoulder stiffness, arm lymphedema) were associated with lymph node inclusion during C1 (p = 0.022), full axillary overlap (p = 0.001), and full SCV overlap (p = 0.002). (Supp Table 5) Interval between courses was not associated with acute or late AEs.Table 5Univariable logistic regression for Grade ≥ 2 late adverse events.CharacteristicNEvent NOR95% CIp-valueC2 Receptor Subtype+/+/−3713––0.0431+/−/−139+/−/+10+/+/+11−/−/+11−/−/−123C1 RT VolumesWB4415––CW421.930.21, 17.50.5CW + LN976.771.43, 49.40.027IORT/PBI310.970.04, 10.9>0.9WB + LN742.580.51, 14.60.3C2 RT VolumesCW + RNI2114––0.0301CW (Recon) + RNI189CW +/− partial LN41Partial nodes (Recon)20PBI/WB50RNI10WB (Recon) + RNI11WB + RNI165Breast OverlapFull145––0.0051Partial90None4424Chest Wall OverlapFull5629––0.0031Partial90None20Axilla OverlapFull1310––Partial39180.260.05, 0.990.064None1510.020.00, 0.170.002SVC OverlapFull1411––None53180.140.03, 0.510.006IMN OverlapFull98––0.0081Partial10None5721[1]Fisher's Exact Test Bold values denote P-value <0.05 suggesting statistical significance.Abbreviations: CI = Confidence Interval, OR = Odds Ratio, RT = radiation, RNI = regional nodal irradiation, LN = lymph nodes, IORT = intraoperative radiotherapy, PBI = partial breast irradiation, WB = whole breast, Recon = reconstruction, SCV = supraclavicular region, IMN = internal mammary lymph nodes. Table 5Univariable logistic regression for Grade ≥ 2 late adverse events.CharacteristicNEvent NOR95% CIp-valueC2 Receptor Subtype+/+/−3713––0.0431+/−/−139+/−/+10+/+/+11−/−/+11−/−/−123C1 RT VolumesWB4415––CW421.930.21, 17.50.5CW + LN976.771.43, 49.40.027IORT/PBI310.970.04, 10.9>0.9WB + LN742.580.51, 14.60.3C2 RT VolumesCW + RNI2114––0.0301CW (Recon) + RNI189CW +/− partial LN41Partial nodes (Recon)20PBI/WB50RNI10WB (Recon) + RNI11WB + RNI165Breast OverlapFull145––0.0051Partial90None4424Chest Wall OverlapFull5629––0.0031Partial90None20Axilla OverlapFull1310––Partial39180.260.05, 0.990.064None1510.020.00, 0.170.002SVC OverlapFull1411––None53180.140.03, 0.510.006IMN OverlapFull98––0.0081Partial10None5721[1]Fisher's Exact Test Bold values denote P-value <0.05 suggesting statistical significance.Abbreviations: CI = Confidence Interval, OR = Odds Ratio, RT = radiation, RNI = regional nodal irradiation, LN = lymph nodes, IORT = intraoperative radiotherapy, PBI = partial breast irradiation, WB = whole breast, Recon = reconstruction, SCV = supraclavicular region, IMN = internal mammary lymph nodes. Univariable logistic regression for Grade ≥ 2 late adverse events. [1]Fisher's Exact Test Bold values denote P-value <0.05 suggesting statistical significance. Abbreviations: CI = Confidence Interval, OR = Odds Ratio, RT = radiation, RNI = regional nodal irradiation, LN = lymph nodes, IORT = intraoperative radiotherapy, PBI = partial breast irradiation, WB = whole breast, Recon = reconstruction, SCV = supraclavicular region, IMN = internal mammary lymph nodes. Increasing dose to 100 cc of the CW (D100cc) (p = 0.007) and D200cc (p = 0.004) during C2 were associated with late CW pain. (Table 6) Increasing CW Dmax (p = 0.042) was associated with breast or CW lymphedema. Higher CW D200cc during both C1 (p = 0.013) and C2 (p = 0.021) and increasing rib D10cc (p = 0.03) during C1 were associated with rib fracture.Table 6Association of late chest wall adverse events with dose volume histogram parameters.Late Chest Wall Pain GradeCharacteristic0 N = 3711 N = 1912 N = 121p-value [2]Cumulative Chest Wall D100cc6703 (5178, 9114)9804 (7291, 10,262)8196 (8035, 8624)0.2Cumulative Chest Wall D200cc5214 (4877, 6610)9062 (6631, 9631)5406 (5212, 5415)0.4C2 Chest Wall D100cc4594 (4135, 5129)5136 (5108, 5174)5118 (4571, 5194)0.007C2 Chest Wall D200cc4437 (1455, 5020)5038 (4766, 5124)4862 (4506, 5109)0.004C2 Chest Wall Dmax5278 (4776, 5522)5428 (5348, 5500)5463 (4973, 5562)0.3C1 Chest Wall D100cc3768 (446, 4297)3231 (1004, 5105)4223 (4106, 4232)0.6C1 Chest Wall D200cc500 (36, 3966)2709 (402, 4886)2196 (332, 4060)0.3C1 Chest Wall Dmax4503 (2723, 5204)4947 (2580, 5473)4557 (4514, 4616)0.7Late Breast/Chest Wall Lymphedema GradeCharacteristic0 N = 491≥1 N = 191p-value [2]Cumulative Chest Wall D100cc7511 (5235, 9122)8624 (5197, 9639)0.8Cumulative Chest Wall D200cc5253 (4908, 6610)5406 (4437, 8604)>0.9C2 Chest Wall D100cc5091 (4552, 5165)5126 (4566, 5197)0.2C2 Chest Wall D200cc4679 (4165, 5060)4960 (4347, 5122)0.4C2 Chest Wall Dmax5350 (4782, 5495)5501 (5329, 5554)0.042C1 Chest Wall D100cc3837 (537, 4282)4236 (1898, 4946)0.2C1 Chest Wall D200cc519 (59, 4014)3762 (197, 4057)0.5C1 Chest Wall Dmax4498 (1913, 4856)4654 (4479, 5290)0.2Late Rib Fracture GradeCharacteristic0 N = 531≥1+ N = 151p-value [2]Cumulative Chest Wall D100cc7343 (5216, 9114)9782 (8035, 10,224)0.11Cumulative Chest Wall D200cc5253 (4877, 6610)9300 (5212, 9422)0.3C2 Chest Wall D100cc5042 (4530, 5159)5165 (5088, 5189)0.068C2 Chest Wall D200cc4563 (4034, 5060)5034 (4739, 5124)0.021C2 Chest Wall Dmax5350 (4783, 5522)5463 (5348, 5554)0.2C1 Chest Wall D100cc3768 (627, 4297)4623 (4232, 5045)0.062C1 Chest Wall D200cc334 (81, 3762)4437 (4060, 4858)0.013C1 Chest Wall Dmax4492 (2723, 5124)4927 (4557, 5346)0.12C2 Rib D5cc4659 (4129, 5055)5004 (4769, 5101)0.084C2 Rib D10cc4471 (3960, 4907)4781 (4524, 4977)0.080C2 Rib Dmax5065 (4580, 5410)5333 (5260, 5388)0.3C1 Rib D5cc3667 (1406, 42284635 (4168, 5100)0.070C1 Rib 10 cc2494 (513, 4154)4564 (4,08, 5048)0.030C1 Rib Dmax4419 (2710, 4782)4830 (4362, 5336(0.2[1]Median (IQR).[2]Wilcoxon rank sum test; Wilcoxon rank sum exact test. Abbreviations: D = dose, cc = cubic centimeters. Table 6Association of late chest wall adverse events with dose volume histogram parameters.Late Chest Wall Pain GradeCharacteristic0 N = 3711 N = 1912 N = 121p-value [2]Cumulative Chest Wall D100cc6703 (5178, 9114)9804 (7291, 10,262)8196 (8035, 8624)0.2Cumulative Chest Wall D200cc5214 (4877, 6610)9062 (6631, 9631)5406 (5212, 5415)0.4C2 Chest Wall D100cc4594 (4135, 5129)5136 (5108, 5174)5118 (4571, 5194)0.007C2 Chest Wall D200cc4437 (1455, 5020)5038 (4766, 5124)4862 (4506, 5109)0.004C2 Chest Wall Dmax5278 (4776, 5522)5428 (5348, 5500)5463 (4973, 5562)0.3C1 Chest Wall D100cc3768 (446, 4297)3231 (1004, 5105)4223 (4106, 4232)0.6C1 Chest Wall D200cc500 (36, 3966)2709 (402, 4886)2196 (332, 4060)0.3C1 Chest Wall Dmax4503 (2723, 5204)4947 (2580, 5473)4557 (4514, 4616)0.7Late Breast/Chest Wall Lymphedema GradeCharacteristic0 N = 491≥1 N = 191p-value [2]Cumulative Chest Wall D100cc7511 (5235, 9122)8624 (5197, 9639)0.8Cumulative Chest Wall D200cc5253 (4908, 6610)5406 (4437, 8604)>0.9C2 Chest Wall D100cc5091 (4552, 5165)5126 (4566, 5197)0.2C2 Chest Wall D200cc4679 (4165, 5060)4960 (4347, 5122)0.4C2 Chest Wall Dmax5350 (4782, 5495)5501 (5329, 5554)0.042C1 Chest Wall D100cc3837 (537, 4282)4236 (1898, 4946)0.2C1 Chest Wall D200cc519 (59, 4014)3762 (197, 4057)0.5C1 Chest Wall Dmax4498 (1913, 4856)4654 (4479, 5290)0.2Late Rib Fracture GradeCharacteristic0 N = 531≥1+ N = 151p-value [2]Cumulative Chest Wall D100cc7343 (5216, 9114)9782 (8035, 10,224)0.11Cumulative Chest Wall D200cc5253 (4877, 6610)9300 (5212, 9422)0.3C2 Chest Wall D100cc5042 (4530, 5159)5165 (5088, 5189)0.068C2 Chest Wall D200cc4563 (4034, 5060)5034 (4739, 5124)0.021C2 Chest Wall Dmax5350 (4783, 5522)5463 (5348, 5554)0.2C1 Chest Wall D100cc3768 (627, 4297)4623 (4232, 5045)0.062C1 Chest Wall D200cc334 (81, 3762)4437 (4060, 4858)0.013C1 Chest Wall Dmax4492 (2723, 5124)4927 (4557, 5346)0.12C2 Rib D5cc4659 (4129, 5055)5004 (4769, 5101)0.084C2 Rib D10cc4471 (3960, 4907)4781 (4524, 4977)0.080C2 Rib Dmax5065 (4580, 5410)5333 (5260, 5388)0.3C1 Rib D5cc3667 (1406, 42284635 (4168, 5100)0.070C1 Rib 10 cc2494 (513, 4154)4564 (4,08, 5048)0.030C1 Rib Dmax4419 (2710, 4782)4830 (4362, 5336(0.2[1]Median (IQR).[2]Wilcoxon rank sum test; Wilcoxon rank sum exact test. Abbreviations: D = dose, cc = cubic centimeters. Association of late chest wall adverse events with dose volume histogram parameters. [1]Median (IQR). [2]Wilcoxon rank sum test; Wilcoxon rank sum exact test. Abbreviations: D = dose, cc = cubic centimeters. Discussion To our knowledge, this is the largest study to date of PBT reRT for breast cancer iLRR. Despite the high-risk nature of our patient cohort, local and regional control rates were promising, with a primary distant recurrence pattern. Severe acute and late AEs were limited, and no ≥G4 AEs were observed, suggesting overall safety of this approach, although continued follow-up is needed. Greater treatment overlap of the breast/CW and regional lymph nodes was associated with an increased risk of acute and late AEs, whereas increased CW dose was associated with CW pain, lymphedema and rib fracture. Few other studies have reported outcomes of PBT reRT for breast cancer recurrence [16], [17], [18], [19]. A series from Memorial Sloan Kettering Cancer Center/ProCure included 46 patients treated with uniform scanning (US) (70%) or PBS (30%) PBT reRT for breast cancer. Median first, reRT, and cumulative doses were 60, 50.4 and 110 Gy(RBE). At a median follow-up of 21 months, there were no LR or RR, but 17% had DR. The two-year OS was 93.6%. Maximum acute G2 and G3 AEs were observed in 52% and 30%, and late G2 and G3 AEs in 17.4% and 8.7%, respectively. No ≥G4 AEs were observed. Rates of AEs were higher in this study compared with our study, possibly attributed to the use of US PBT in the majority of patients, versus PBS in our study [16]. A study from the prospective Proton Collaborative Group (PCG) included 50 patients treated with either PBS or US PBT reRT for breast cancer. Of note, 26% of patients did not undergo surgical intervention prior to radiotherapy. The median first, second, and cumulative doses were 60, 55.1, and 110.6 Gy(RBE), respectively. At a median follow-up of 12.7 months, 6% experienced LR and 8% developed metastatic disease. G3 acute and late AEs rates were 10% and 8%, respectively, similar to those observed in our study, albeit with shorter follow-up in the PCG analysis [19]. The University of Pennyslvania conducted a study including 27 patients who underwent PBT reRT for breast cancer, two-thirds of whom (67%) received PBS PBT and the remainder receiving double scattering PBT. Patients were most commonly treated to 49.5–51 Gy in 1.5 Gy twice daily fractions for their reRT course, unlike our study in which all patients were treated once daily. At 16.6 month median follow-up, 3.7% developed LR, 40.7% DR, and OS was 74%. Acute G3 radiation dermatitis and breast pain were observed in 7.4% of patients each. Late AEs included 3.7% G3 dermatitis, 3.7% G3 breast pain, and 3.7% G4 dermatitis. Similar to our study, the rate of observed rib fractures was 22.2% [18]. In a more dated series, Wahl et al. published a multi-institutional study in 2008 on photon reRT for breast cancer. That study included 81 patients at 8 institutions with median first and second course doses of 60 and 48 Gy, respectively. One-year local recurrence-free survival (LRFS) was 53% but was 100% for patients without gross disease, and one-year OS was 64%. Late G3–4 AEs developed in 4.9%; however, median follow-up was limited at only 12 months. The total rate of acute G3–4 AEs is unclear, but at least 8.6% required a one-week treatment break due to severe acute AEs [20]. Longer follow-up of photon reRT for breast cancer will be of interest to compare clinical outcomes with patients receiving proton reRT. The rate of rib fractures in our study, 21.1%, is higher compared with historical event rates in first-course/primary breast radiation reports, across both series assessing photon therapy and proton therapy [21], [22], [23]. However, this rate is similar to findings in previous proton reRT reports [18]. This is likely attributed to the cumulative radiation dose sustained by the ribs across multiple curative courses of breast radiotherapy, in which the chest wall, including the ribs, are within or adjacent to the target region receiving full prescription dose. Consideration can also be given to the increased relative biological effectiveness (RBE) at the distal end of the proton beam range, which may also contribute to the risk of rib fractures. Given the predominant distant recurrence pattern (8.6%) found in our study, which is consistent with prior reports, further study on optimal combinations of systemic therapies with local-regional therapies in the recurrent setting is needed, particularly with significant recent advances in breast cancer drug therapy including immunotherapy and targeted agents [16], [18], [24], [25], [26], [27]. A current knowledge gap is the unclear interaction and possible exaggeration of radiation AEs with concurrent delivery of newer systemic agents with RT, which is of heightened concern with reRT [28], [29]. A strength of our study is that patients were treated at a single institution with group discussion for most patients and using standardized treatment guidance. In addition, patients were predominantly treated by one attending physician, thus reducing heterogeneity in treatment design and short- and long-term AE assessment. All patients were treated with a single type of proton technology, de novo PBS-PBT (versus retrofitted PBS-PBT), which also represents the most advanced technology available to date. Patients received daily on-board soft tissue-based image guidance. All patients received prior RT for breast cancer, representing a homogeneous cohort, versus inclusion of patients who received prior RT for non-breast cancer histologies such as lymphoma. AEs were collected prospectively in a standardized and robust manner for all patients at all timepoints, allowing for a more reliable reporting of outcomes, despite the retrospective nature of our study. Accessibility of C1 RT plans for cumulative dose evaluation is a known challenge in the study of reRT, especially in disease sites such as breast cancer in which recurrences can occur many years after the first diagnosis, precluding prior RT plan availability and review for some patients. Our institution has rigorous standard clinical workflows that require obtaining prior RT plans whenever possible. Despite these efforts, only 44.3% of patients had C1 RT plans available, which may restrict analysis of potential correlation between cumulative doses and DVH parameters with increased toxicity. Clinical workflows to obtain prior treatment records, including DICOM-RT or treatment plan summary, are prioritized at our institution. When DICOM-RT are not available, recreation of volumes with isodose distribution should be attempted for evaluation of areas that may be dose-reduced or further avoided based on risk level and to apply reRT dose constraints as they are developed. Limitations of this evaluation include patient anatomy and positioning differences between courses, addressed to some degree using deformable registration. Refinement of reliable dose summation approaches for overlapping tissue will be needed. In addition, patients included in this study either live in a region with access to specialized services such as PBT, or have the financial resources to seek care far from home, thus limiting generalizability to patients in regions without access to PBT. The median follow-up in this study of 30 months may be insufficient for fully assessing the rate of late toxicities, particularly cardiac adverse events. Thus, the rates of these toxicities should be interpreted with caution. Each case of breast cancer iLRR is unique, with many factors to consider, including interval surgeries, systemic therapy regimens, prior RT dose and volumes, time between RT courses, treatment-related complications, age, and comorbid status. Patient goals of care should be strongly considered, and an open dialogue with the patient should take place to allow for informed treatment decision-making. Discussion with the multidisciplinary team to ensure alignment of treatment timing and overall plan of care is essential. ReRT for breast cancer iLRR is a necessary component of the multidisciplinary salvage treatment paradigm that is overall well-tolerated but is associated with heightened risk for severe AEs compared with first course RT. A method of predicting interval recovery between RT courses for different tissue types may be helpful in dose selection [30]. More study of cumulative and second course DVH parameters associated with AEs will be helpful to improve guidance for reRT delivery [31]. A comparison of PBT and photon breast reRT outcomes will be of interest to assess differences and determine the impact of each on enhancing the therapeutic ratio of reRT. Optimization of systemic therapies in the recurrent setting will be essential to further improve salvage outcomes. While reRT will remain a higher risk treatment, harmonization of patient and treatment goals, incorporation of late AE prevention and management strategies, and refinement of reRT guidance will allow for more informed and ultimately safer reRT delivery. Conclusion In the largest experience of PBT reRT for breast cancer iLRR to date, we report low rates of locoregional recurrence despite the high-risk nature of the patient cohort. Severe AEs occurred, as expected in the setting of reRT, but at a limited rate and notably with no ≥G4 AEs observed. These promising results warrant additional study on the continued optimization and clinical impact of PBT reRT for breast cancer iLRR, and comparison to outcomes of photon reRT. Further, we provide dose metrics that may serve as the foundation for reRT dose constraints to be applied and further refined in future reRT investigations. CRediT authorship contribution statement Molly A. Chakraborty: Writing – original draft, Formal analysis, Data curation. Charlie White: Writing – original draft, Formal analysis, Data curation. Atif J. Khan: Writing – review & editing. Siyu Yang: Writing – review & editing, Data curation. Zhigang Zhang: Writing – original draft, Formal analysis, Data curation. Francis Yu: Writing – review & editing, Data curation. Haibo Lin: Writing – review & editing, Data curation. Lior Z. Braunstein: Writing – review & editing. Diana Roth O'Brien: Writing – review & editing. John J. Cuaron: Writing – review & editing. David M. Guttmann: Writing – review & editing. Quincey LaPlant: Writing – review & editing. Simon N. Powell: Writing – review & editing. Oren Cahlon: Writing – review & editing. J. Isabelle Choi: Writing – original draft, Supervision, Methodology, Formal analysis, Data curation, Conceptualization. Funding This research was supported by NIH/NCI Memorial Sloan Kettering Cancer Center Support Grant/Core Grant No. P30-CA008748. 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.