Toxicity of Hypofractionated Whole Breast Radiotherapy Without Boost in a Large “Real-Life” Cohort of Consecutive Early-Stage Breast Cancer Patients

To report toxicity of hypofractionated whole-breast radiotherapy (HWBRT) in a large cohort of early-stage breast cancer (BCA) patients. and From 02/2009-05/2017, 1325 consecutive BCA patients were treated with 40.05 Gy/15 fractions, without boost with manually optimized IMRT. Median age: 62 (IQR:51.1-70.5) years. Chemotherapy was prescribed for 28% of patients, hormonal therapy for 80.3%, monoclonal antibodies for 8.2%. Ten% of patients <45 years, 23.8% obese, 5.9% diabetic, 34% hypertensive. Results Median follow-up: 72.4 (IQR: 44.6-104.1) Acute RTOG toxicity was: 69.8% and 1.7% G3. Late SOMA-LENT toxicities were: edema-hyperpigmentation(E-H): G1 G2 4.41%, G3 0.15%; brosis-atrophy-telangiectasia-pain(F-A-T-P): G1 14.6%, G2 3.2%, G3 0.8%, G4 0.1%. Median time to rst occurrence was 6 and 18 months, respectively. Aesthetic result excellent in 20.3% and poor in 9.5% of Change in breast appearance after was mild in in 2.3% and marked in 1.3% of and late toxicity in a large cohort of consecutive patients. Timing of changes in breast appearance after radiotherapy was registered and patient, therapy and dosimetric predictive toxicity factors were identied.


Introduction
Hypofractionated whole breast radiotherapy (HWBRT) becomes the standard adjuvant treatment in BCT over the last decade, demonstrating similar risk of loco-regional failure, breast cancer death, no differences in late toxicity and decreased risk of acute toxicity, edema and telangiectasia with standard fractionation [1,2]. All patients of the Ontario trial and more than half of those in START-B were treated with HWBRT without boost, with good results [3,4]. Early in 2009 our department adopted the HWBRT of the START B trial, without boost, to de-escalate the treatment and improve cosmetic outcomes, while maintaining good local control and here we report toxicity results and clinical and dosimetric factors associated with increased risk of acute and late toxicity.

Material And Methods
Study details have already been published elsewhere [5]. In short, from February 2009 to May 2017, 1325 consecutive patients who underwent BCT for early-stage breast cancer (BCa) were treated with HWBRT. All patients were treated with manually optimized intensity-modulated with/without physical/dynamic wedges at 40.0 Gy/ 15 fr without boost to the tumor bed independently of age, obesity, breast volume, side, chemotherapy/trastuzumab prescription, comorbidities, and were included in this analysis. All patients signed an informed consent and the institutional ethics committee approved the study.
Signi cant patient and treatment characteristics for this toxicity analysis are summarized in Table 1. Simulation, contouring, treatment planning, patient monitoring: Whole breast was the clinical target volume (CTV) de ned on 5-mm axial chest CT scan images, then isotropically expanded in all directions except lung, and a 5 mm crop made to the body to obtain the planning target volume (PTV).
Critical organs (OAR) had speci c constraints [5]. Multiple segments were used within a tangential two eld irradiation technique to obtain a homogeneous dose, with hotspots Dmax ≤ 108%. Photon(X) 6 megavolt(MV) beams with/without wedges were used; in selected cases, X18 MV segments were added to improve dose distribution. A 1-cm water-equivalent plastic bolus was admitted in the rst years to improve super cial dose distribution and was used in 313 (23.6%) patients, mostly for 5-7 of 15 fractions. It was abandoned after an unpublished analysis in which an increase in toxicity was observed. Set-up errors > 5 mm were corrected. Patients were visited twice during the treatment, and acute toxicity was reported using the RTOG scale [9]. Follow-up visits were scheduled at 6-18-30-42-54-66 months after RT end. The SOMA-LENT toxicity scale was used to assess late collateral effects [10,11]. Frontal and lateral photographic images were acquired at beginning and/or end of radiotherapy and during follow-up visits. Breast size, applying a visual 4-point graded scale considering breast volume and its proportion with the patient's BMI, was scored by two observers as small, medium, large and very-large. Post-surgical de cits (asymmetry, edema, scar and/or nipple retraction) were de ned at baseline. Aesthetic results (photographic) were reviewed and scored by two observers and the nal score (using the Harvard 4-point scale): excellent, good, acceptable and poor [12], reached by consensus.

Statistical analysis
Median and interquartile range (IQR) were reported for continuous variables, while frequency distribution and percentages were reported for categorical variables. Ordinal logistic regression model was tted to identify risk factors for acute toxicity (outcome variable with three toxicity thresholds: G0, G1, G2 or G3). Age, neoadjuvant and concomitant chemotherapy, hormonal therapy, monoclonal antibodies, obesity, diabetes, hypertension, smoking, alcohol consumption, presence of bolus and PTV volume were entered as independent variables in the initial model. A cubic root transformation was applied to the PTV volume to normalize its skewed distribution and, to avoid collinearity issues, V105 and V95 dosimetric parameters were not included as covariates in the model. Backward elimination method was applied to select a smaller set of relevant covariates. Binary logistic regression models were estimated to identify predictors associated with the development of breast liponecrosis and those associated with the development of ≥ grade (G)2 late edema/hyperpigmentation (E-H) toxicity within six months. Cox regression model for interval censored data was applied to identify risk factors for late brosis-atrophy-telangiectasia-pain (F-A-T-P) toxicity ≥ G2.
For the latter three outcomes (breast liponecrosis, E-H and F-A-T-P) the same set of independent variables were considered as for acute toxicity outcome. We wish to highlight that for both late E-H and late F-A-T-P, the occurrence of ≥ G2 toxicity was considered the event of interest without distinguishing among different skin reactions/manifestations.
Due to the small number of events, nonparametric tests such as Mann-Whitney test and Fisher's exact test were performed, depending on the type (quantitative/qualitative) of variable tested, to examine differences between patients with and without late cardiac or with and without lung toxicities with respect to patient and treatment characteristics, D5%, V35 Gy, V20 Gy and V95% or mean lung doses, D20%, V22 Gy, V17 Gy, and mean heart doses (for left sided patients), respectively. All the analyses were performed using R statistical software (version 3.5.2, https://cran.rproject.org/index.html). Signi cance level was set at 0.05.

Results
Median follow-up was 72.4 (IQR:44.6-104.1) months. A median number of 4 (2-11) segments were used. While for 90.7% of patients up to 4 elds were necessary to reach the constraints, a more complex construction was required in the remaining patients. Median PTV and OAR volumes and selected dosimetry parameters were included in Table 2. switched to two years of AI after three years of Tamoxifen. AI are associated with increased cardiovascular mortality and ischemic stroke [13]. Another patient, treated with AI died of ischemic stroke.
Although pulmonary brosis was not systematically investigated, and was reported for only symptomatic patients or those where it was observed on CT or chest X-ray performed for other reasons, 19 patients (1.4%) were diagnosed with a pulmonary event.
Late skin toxicity evolution E-H were present mainly at six-month follow-up (in approximately 33% of patients) and decreased over time.
Hence, considering only patients with complete follow-up up to 42 months (n = 998) the proportion of patients with G0 toxicity increased from 68.1% at six months to 96.8% at 42 months; the proportion of patients with G1 toxicity decreased from 28-3%, and G2 toxicity from 3.8-0.2%. G3 toxicity disappeared within 18 months (see Fig. 1).
The evolution of F-A-T-P toxicity was opposite to that of E-H toxicity. Even though improvements over time were registered for some patients, an overall worsening trend for toxicity over time was observed. For those patients showing at least G1 F-A-T-P toxicity overall, the rst occurrence of the toxicity was registered at 6 months for 37.4% of the patients. For the same patients the median time to rst occurrence of toxicity was 18 months.
Focusing on the 998 patients with complete follow-up up to 42 months, the proportion of patients with G0 toxicity decreased from 94.4% at 6 months to 88.3% at 42 months, while the proportion of patients with G1 toxicity increased from 4.9-8.7%, and G2 toxicity from 0.7-2.4%. G3 toxicity was registered for the rst time at 18 months after radiotherapy, increasing from 0.3 to 0.5% at 42 months, while the sole G4 toxicity registered at 18 months follow-up remained unchanged at 42 months (see Fig. 2). Few patients worsened thereafter, but one new G1toxicity was registered even at the 9th-year of follow-up.
Aesthetic result after surgery, evaluated on the Harvard 4-point scale at the time of radiation treatment, was excellent in 28.7%, good in 41.5%, acceptable in 20.3% and poor in 9.5% of patients.
Change in breast appearance after radiotherapy was separately evaluated (patient-reported, clinically and with photographs) and at the end of follow-up was mild in 6.9%, moderate in 2.3% and marked in 1.3% of patients.
Other primary tumors:

Factors in uencing acute toxicity
In the nal selected ordinal regression model, concomitant chemotherapy, obesity, smoking, presence of bolus and PTV volume were signi cantly associated with higher acute toxicity (G1 or G2/G3 versus G0), while patients ≥ 55 years seem less likely to experience a higher acute toxicity (see Table 3). For liponecrosis, again referring to the nal logistic regression model selected, PTV volume and smoking were signi cantly associated with increased toxicity risk, while HT was a protective factor (see Table 4).

Discussion
Despite the bene ts of the shorter treatment schedule HWBRT has not been widely adopted [14]. Concerns were related to the risk of relapse but also to long term toxicity, especially in patients treated with chemotherapy, trastuzumab, left sided breast, large breast. Compared to the Ontario trial, whose patients were treated without boost, but which employed a slightly higher biologically effective dose than ours, more than 40% of START and 23% of DCGB HYPO trial patients received boost [3,4,[15][16][17]; thus our study represents the largest "real-life" cohort of non-selected patients treated with 40 Gy/15 fr, without boost.
Acute toxicity is less described in the randomized prospective trials. Jagsi et al., in a real-world observational study described the acute toxicity in 578 patients treated with hypofractionation: G ≥ 2 dermatitis was registered in 27.4%, while moist desquamation (G3 toxicity) in 6.6% [18]. De Santis et al., analyzed the toxicity outcome of 537 BCa patients treated with hypofractionation (42.4 Gy/16 fr), of whom 27% received sequential boost, reporting 61.3% G1 toxicity and 20.5% G2/G3 brosis [19]. Only the boost was found to increase the risk of acute toxicity at multivariate analysis. With G2 RTOG toxicity of 14.3% and G3 of 1.7% our results are favorable compared to the cited studies; concomitant chemotherapy, obesity, smoke, presence of bolus and PTV were signi cantly associated with higher acute toxicity, while older age was protective.
Radiation associated cardiac disease ( ve patients died of cardiac disease, three left-sided) was rare in the DBCG HYPO trial, with a similar rate (7.6%) of patients with trastuzumab prescription [17]. In the 10-year result publication of the Start B trial, ischemic heart disease was reported in 1.5% of patients, but con rmed in only 0.7% of patients after imaging and other investigation (0.4% left sided), symptomatic lung brosis in 1.7% of patients, con rmed in 0.7%, and symptomatic rib fracture in 2.2% of patients, con rmed in 0.3% (others were traumatic or due to metastases) [16]. With 6-year median follow-up the incidence of contralateral breast cancer was 1.5% for hypofractionated arm and 2.3% of other primary cancers (lung, endometrial, ovarian, colon) [4]. Only two cases of radiation pneumonitis and one rib fracture were reported by Whelan et al [3,15]. Incidence of heart disease and lung pneumonitis was higher in our cohort, likely due to the higher proportion of patients with chemotherapy and AI prescription, as well as to the trastuzumab therapy in Her2 + patients, not used in the older series of Ontario and START trials. Radiotherapy may be expected to increase cardiac and lung toxicities, as well as second tumor incidence, and the association of synergic systemic therapies from this point of view increases the risk [22][23][24]. Recent reports, however, demonstrate a positive impact of new schedules and technological improvements, lowering risk [25,26]. The short follow-up for cardiac side effects, but also the low heart Dmean (2.49 Gy) obtained in our left-sided patients probably contributed to the low cardiac toxicity observed, with no signi cant difference between patients for right and left breast cancer. No rib fractures were registered in our cohort. Other primary tumors were largely discussed in results.
Major limitations of our study are the absence of randomization and its retrospective nature. The greatest strength is the homogeneity of our hypofractionated treatment, from the point of view of fractionation, dose, volumes and technical modalities, resulting in the largest cohort of "real-life" consecutive BCa patients without boost irradiation, systematically monitored for a "su ciently long" period of time after the treatment. We observe that, despite the use of chemotherapy and trastuzumab in all " t" patients, molecular subtypes were the most signi cant factors predicting relapse [5]. In these consecutive, unselected patients, treated with anthracyclines, taxanes, trastuzumab, as required by their illness, we could also evaluate acute and late toxicities and the factors in uencing them. Thus we consider that, even with its limitations, this study offers "real-life" information of outcomes of patients treated with a three-week hypofractionated schedule that could be complementary to the randomized trials in which some of these treatments were less prescribed or unavailable, and could favor the wider adoption of a still underused hypofractionation [14].
Another relative limitation is the forward planned-IMRT technique. However, this approach is still widely used for breast cancer, and provides, as reported here, very good results in terms of dose homogeneity and OAR sparing. Of note, this approach was recently updated with the implementation of an automatic planning technique "in-house" developed, mimicking and slightly improving the performance, and reducing dependence on the human factor (critical in forward IMRT optimization) [27].
Results of FAST and FAST-Forward trials were recently published [28,29] and, combined with recent international recommendations to reduce hospital visits during Sars-Cov2 pandemic [30], could move treatments towards only 5 fractions. Thus the three-week schedule could become obsolete, at least for a fraction of patients. However, for the FAST-Forward trial, long-term control and toxicity data, vital for nal clinical acceptance (due to the 18.8% EQD2 dose reduction and the substantially higher dose per fraction [17]), are missing. Moreover, the boost proposed in FAST Forward trial is only sequential, reducing the overall treatment time gain, while for moderate hypofractionation a simultaneous boost can be easily integrated.

Conclusion
The HWBRT, without boost, after BCS demonstrated limited acute and late toxicity in a large real-life cohort of 1325 consecutive patients. Predictive factors for toxicity were identi ed in these patients treated with modern chemotherapy, trastuzumab, and a wide use of AI for a longer period than HT in the Ontario and START trials. Changes in breast appearance after radiotherapy, assessed on photographs separately from changes produced by surgery, involved 10.5% of patients, but were marked in only 1.3%.

Declarations
Con ict of interests: The authors report no con ict of interests.
Funding: This research did not receive any speci c grant from funding agencies in the public, commercial, or not-forpro t sectors.
Clinical trial information: The study is registered with ClinicalTrials.gov (NCT03077191) together with an ongoing prospective study of quality of life.
Data Sharing Statement: The study protocol did not include a sharing plan. However, anonymized individual participant data will be available following the publication of the article until the closure of the study upon request on a case-by-case basis to researchers who provide a methodologically sound proposal. Requests made to the corresponding author will be evaluated by the IRCCS San Raffaele Scienti c Institute Ethics Committee. Figure 1 The decrease of late edema-hyperpigmentation over time: the alluvial plot shows the changes in degrees toxicity over time; streams are represented by the waves in shades of gray between the columns (passage from G1, G2, and G3 to G0, from G0, G2 and G3 to G1, etc.), where the toxicity frequency distribution is shown.