Daily Fractionation of External Beam Accelerated Partial Breast Irradiation to 40Gy is Well Tolerated and Locally Effective

Introduction:

A series of groundbreaking studies have established breast conservation as a safe and effective alternative to mastectomy for early stage breast cancer. As a result, many patients now receive adjuvant whole-breast radiotherapy (RT), which improves long term locoregional-control and survival following lumpectomy ^1–3. However, whole-breast RT can take up to six weeks to administer, often at considerable inconvenience⁴ and it remains unclear whether targeting the entire breast with RT is necessary in many cases.

Several technological advances have enabled postoperative localization of the surgical bed prompting the investigation of non-invasive, external beam accelerated partial breast irradiation (APBI) ^5,6. This approach to breast RT was initially thought to limit therapeutic toxicity and allow for more rapid treatment by targeting only a portion of the breast volume. The foundation of APBI lies in the observation that most local recurrences arise proximate to the original surgical cavity, and that distant parts of the breast may be at lower risk^7–9.

While an array of APBI techniques have been reported, the optimal dose regimens have yet to be defined ^10–12. Initial studies employed brachytherapy with prescription doses based on radiobiologic modeling. Contemporary external beam APBI approaches extrapolate from this experience and largely entail once- or twice-daily treatment for a period of one to three weeks, using a variety of prescription doses. Since the toxicity and efficacy of RT depend largely on the dose per fraction, total dose, and timing of delivery, there is a need to define these parameters for optimizing the overall effectiveness of APBI. Based on preliminary data about the tolerability of various twice-daily regimens¹³, we developed a once-daily regimen of 40Gy delivered over two weeks which may offer a convenient alternative to BID treatment, perhaps with less local toxicity.

Methods:

Patient eligibility:

We enrolled patients with low/intermediate grade ductal carcinoma in situ (DCIS) or invasive breast cancer ≤2cm in size. All patients were node-negative, although lymph node deposits ≤0.2mm (i.e. N0[i+]) were allowed. Margins of ≥2mm were required based on surgical consensus guidelines at the time of study accrual, and all patients had an ECOG performance status of 0 or 1.

Trial design and outcomes:

We conducted a phase II single-arm study of external beam APBI in which a select group of low-risk patients (as defined above) received radiation to the partial breast. Our main objective was to assess the safety and feasibility of a 40 Gy regimen administered in ten once-daily fractions over two weeks. The acceptable safety threshold was defined as a <10% rate at two years of one or more of the following events: (1) grade 3 or higher cutaneous, subcutaneous or pulmonary toxicity, (2) fat necrosis, (3) poor cosmesis, or (4) rib fracture. Secondary objectives included efficacy as defined by local, regional and distant disease control. Toxicity, cosmesis and recurrence outcomes were assessed at 4-9 weeks following RT and annually thereafter using standardized grading criteria (supplementary table 1). Physicians were required to complete standardized toxicity and cosmesis evaluations at each follow-up and to report adverse effects along with their probability of attribution to RT.

Treatment technique:

Patients were treated to a total dose of 40 Gy in 10 once-daily fractions given over a 2-week period. We permitted any combination of photon beams of energy 6MV or higher, with or without the addition of electrons of any energy, provided that dosimetric requirements could be met. The clinical target volume (CTV) was defined as the tumor bed or seroma, demarcated with the aid of 3D imaging and surgical clips. The planning target volume (PTV) added an isotropic margin of 1.5 – 2.0 cm to the CTV, limited anteriorly to within 5mm below the skin surface and posteriorly to the anterior surface of the chest wall/ribs. The PTV was limited to <35% of the whole-breast volume, and 50% of the non-target breast tissue was limited to receiving <50% of the prescribed dose. Dose-volume goals included an ipsilateral lung V20 < 3%, and heart maximum dose < 90% of the prescription dose.

Statistical analyses and investigation of metrics predictive of local toxicity:

Reported metrics found to predict skin toxicity after RT for breast cancer were explored:

1. The relative skin volume receiving at least 35Gy (Skin V₃₅)¹⁴

2. The relative treated volume receiving at least 110% of the prescription dose (TV≥110%)^15,16

3. The relative body dose-surface area receiving at least 30Gy (Body A₃₀)¹⁷

4. Body mass index (BMI)^18,19

5. Smoking status^19,20

Further details pertaining to the specific skin/subcutaneous toxicities studied, as well as model parameters are given in supplementary table 2. Each metric was subject to exploration using logistic regression (equation: 1+(1+exp^{(−(β0+β1*x1…))})). Thus, if model coefficients (the intercept β₀ and the regression coefficient β₁) were provided in the associated publication, these were applied to the corresponding metrics in the studied cohort. However, if they were not, the relevant metrics were re-fitted. For consistency, re-fits were also performed for the publications in which the regression coefficients were provided. The re-fitting process was performed with bootstrap resampling (i.e., repeated 1000 times). Measures of performance included discrimination being represented by the area under the receiver-operating characteristics curve (AUC) with associated p-values, and calibration represented by p-values from a Hosmer-Lemeshow test (p_HL)²¹. For any re-fitted model, these performance measures were all given as the median value across the 1000 samples. A published metric was considered predictive if yielding a p-value ≤ 0.01 (accounting for multiple hypothesis testing with the five investigated metrics) and presenting with a p_HL~0.50. Fractionation correction was typically disregarded among the published reports, and was, therefore, not considered here. Five additional exploratory metrics of interest were similarly analyzed: breast V₂₀, total breast volume, PTV volume, PTV/breast volume ratio, and PTV max dose.

Results:

Beginning in 2010, we enrolled 106 patients (89 with invasive breast cancer, 17 with low/intermediate grade DCIS), all of whom underwent lumpectomy to achieve the required margins followed by 40 Gy APBI in ten fractions over two weeks. The median age was 62, and median tumor size was 9mm (Table 1). The majority of patients had estrogen receptor-positive, HER2-negative disease (91.5%). Due to the low-risk nature of the cohort, few patients received chemotherapy (13.2%), although most received endocrine therapy (84.9%).

Table 1:

Baseline patient and treatment characteristics

	Overall	DCIS (N=17)	Invasive (N=89)
Median age at diagnosis	62	62	62
Median tumor size (cm)	0.9 (0.1 - 2)	0.6 (0.3 - 1.6)	0.9 (0.1 - 2)
Ethnicity (n, %)
Asian	7 (6.6)	2 (11.8)	5 (5.6)
Black	11 (10.4)	3 (17.6)	8 (9)
Hispanic	6 (5.7)	1 (5.9)	5 (5.6)
Multiple	1 (0.9)	1 (5.9)	0 (0)
White	81 (76.4)	10 (58.8)	71 (79.8)
Laterality (n, %)
Left	50 (47.2)	5 (29.4)	45 (50.6)
Right	56 (52.8)	12 (70.6)	44 (49.4)
Nodal evaluation
No pathologic evaluation	15 (14.2)	14 (82.4)	1 (1.1)
Sentinel node biopsy	91 (85.8)	3 (17.6)	88 (98.9)
Histologic Grade (n, %)
I-Well Differentiated	21 (19.8)	3 (17.6)	18 (20.2)
II-Moderately Differentiated	37 (34.9)	13 (76.5)	24 (27)
III-Poorly Differentiated	40 (37.7)	0 (0)	40 (44.9)
ER Positive (n, %)	97 (91.5)	11 (64.7)	86 (96.6)
PR Positive (n, %)	80 (75.5)	2 (11.8)	78 (87.6)
HER2 Positive	4 (3.8)	0 (0)	4 (4.5)
Chemotherapy (n, %)	14 (13.2)	0 (0)	14 (15.7)
Hormone therapy (n, %)	90 (84.9)	7 (41.2)	83 (93.3)
Median BMI (range)	28.5 (19-44)
Smoker (n, %)	45 (42.4)

At a median follow-up of 58 months, ipsilateral breast cancer recurrence arose in two patients (1.8%), both of whom subsequently underwent successful salvage (via repeat lumpectomy and whole-breast radiation to 45 Gy in 25 fractions) and remain without evidence of disease at last follow-up. In addition, one patient (0.9%) developed distant disease 2 years following definitive treatment and succumbed to breast cancer 1.5 years later. There were no other recurrences, deaths or breast-cancer-specific events among the cohort.

Toxicities of grade 2 or higher are tabulated in Table 2. Of 106 patients, 16 (15%) experienced ≥grade 2 skin toxicity. The most common significant toxicities were acute cutaneous changes at 4-9 weeks following RT, including grade 2 erythema in two patients (1.8%) and skin color changes in four patients (3.8%). Only two instances of grade 3 toxicity were reported, including one patient with acute moist desquamation following RT, and another with fibrosis at 2 years. Late fat necrosis arose in one patient and was confirmed both clinically and radiographically, and rib fracture was reported in a single patient at 5 years. No other grade 3 toxicities and no grade 4 toxicities were reported. Of 11 patients who exhibited acute ≥ grade 2 skin toxicity, only 6 manifested a late toxicity. Of 11 patients who exhibited late ≥ grade 2 toxicity, 5 had no preceding acute toxicity that would have predicted for a late adverse event.

Table 2:

Grade 2 or higher toxicity.

	4-9 wks (n = 106)	12 mo (n = 99)	24 mo (n = 93)	36 mo (n = 92)	48 mo (n= 90)	60mo (n = 50)
Breast Atrophy		1		1
Breast Edema		1
Breast Pain		1			1
Dry desquamation	1
Erythema	2		1
Fatigue	2
Fibrosis or retraction			1*	2
Moist desquamation	1*
Pneumonitis	1
Skin color changes	4		1
Fat necrosis				1
Rib fracture						1

^*Grade 3 toxicity arose in two patients (one with fibrosis, another with moist desquamation). There were no grade 4 toxicities.

The study reached its primary safety threshold with no patients experiencing poor overall cosmesis and 99.1% achieving excellent/good cosmesis (fair cosmesis was noted in a single patient; table 3). Among the other pre-specified safety criteria, significant toxicity was rare including total grade 3 skin or subcutaneous toxicity (n=2), fat necrosis (n=1), and rib fracture (n=1).

Table 3:

Overall Cosmesis

	Overall
Final Cosmesis
Excellent	59 (55.7)
Good	45 (42.5)
Fair	1 (0.9)
Poor	0

To explore potentially predictive metrics of RT-associated toxicity, dose-volume histogram data were analyzed for 88 evaluable patients. Overall, the performance of five previously-reported metrics was modest with only BMI and Skin V₃₅ approaching significance (AUC=0.61, 0.63; p=0.03 for both; Table 4). Re-fitting Skin V₃₅ did not improve model performance (AUC=0.61; p=0.03), nor did re-fitting improve performance of another model for which the regression coefficients were available, i.e., Body A₃₀ (AUC=0.58 for both; p=0.38 vs. 0.37).

Table 4.

Exploration of published metrics in the current cohort.

Ref	Published metric	No re-fit β0	β1	AUC	p	pHL	Re-fit β0	β1	AUC	p	pHL
Lee [14]	Skin V35 [%]	−1.44	0.05	0.61	0.03	0.37	−2.44	0.06	0.61	0.03	0.68
Chen [15]	TV≥110% [%]	-	0.98	-	-	-	−1.88	999	0.57	0.07	0.68
Lazzari [16]	TV≥110% [%]	-	0.98	-	-	-	−1.88	999	0.57	0.07	0.68
Pastore [17]	Body A30 [%]	−15.54	0.18	0.43	0.36	1.00	−1.17	−3.14	0.58	0.38	0.68
Twardella [18]	BMI	-	-	-	-	-	−4.38	0.09	0.63	0.03	0.70
De Langhe [19]	BMI BMI Smoking (Yes vs. No)	- - -	0.04 0.07 0.43	-	- - -	-	−4.38 −4.38 −1.93	0.09 0.09 0.37	0.63 0.63 0.56	0.03 0.03 0.41	0.70 0.70 0.69
Lilla [20]	Smoking (for 30y vs. ≤30y)*	-	0.36	-	-	-	−1.93	0.37	0.56	0.41	0.69

Abbreviations: AUC: Area under the receiver-operating characteristics curve; β_0: Intercept of regression model; β₁: Regression coefficient; p_HL: P-value from a Hosmer-Lemeshow test.

^*only Smoking (Yes vs. No) investigated in the current cohort

Among an additional set of exploratory metrics, both breast V₂₀ and absolute PTV volume were significantly predictive of skin toxicity (AUC=0.69, 0.77; p=0.01, 0.0003; p_HL=0.68, 0.69; Table 5). Including these two metrics in a multivariate analysis generated a model with slightly improved performance to either alone (AUC=0.78; p=9.6e-5; p_HL=0.72). The associated univariate and multivariate dose-response curves based on these models can be used to assess and further minimize the expected rate of skin toxicity given a certain PTV, and/or Breast V₂₀, as illustrated in Figure 1 and defined by the following equations:

1. PTV: Skin tox=1/(1+exp^{(−(−3.95+0.01*PTV))})

2. Breast V₂₀: Skin tox=1/(1+exp^{(−(−5.30+7.23*Breast V20))})

3. PTV and Breast V₂₀: Skin tox=1/(1+exp^{(−(−5.59+0.01*PTV+3.96*Breast V20))})

Table 5.

Analysis of clinical planning metrics of interest with regard to prediction of skin toxicity. The upper rows show univariate results; the last row shows multivariate results for the two significant predictors.

Metric	β0	β1	AUC	p	pHL	Median (range) With gr ≥2 tox	Without gr≥2 tox
Breast volume [cm3]	−2.95	8.00e-4	0.67	0.12	0.66	1700 (801-2850)	1140 (540-3400)
Breast V20 [%]*	−5.30	7.23	0.69	0.01	0.68	0.51 (0.29-0.89)	0.47 (0.19-0.67)
PTV [cm3]*	−3.95	0.01	0.77	3e-4	0.69	243 (68.2-537)	166 (55.8-513)
PTV/Breast volume	−2.51	4.52	0.59	0.41	0.68	0.16 (0.08-0.26)	0.15 (0.04-0.32)
PTV max dose [Gy]	−22.6	0.50	0.71	0.06	0.61	42.3 (41.5-44.5)	41.5 (39.5-45.7)
Breast V20 [%] PTV [cm3]	−5.59	0.01 3.96	0.78	9.58e-5	0.72	-	-

Abbreviations: AUC: Area under the receiver-operating characteristics curve; β_0: Intercept of regression model; β₁: Regression coefficient; p_HL: P-value from a Hosmer-Lemeshow test; PTV: Planning target volume, tox: toxicity.

^*Candidate predictor.

Fig 1.

Dose-response curves for the two treatment planning metrics found to significantly predict skin toxicity (PTV: upper left; Breast V₂₀: lower left), and the dose-response curve combining these metrics (right). Various parameterizations of PTV, Breast V₂₀, or the combination thereof and the resulting rate of skin toxicity according to the predictions are shown in the upper left insets; these may be used as example thresholds to further reduce the rate of skin toxicity upfront. Note: Associated AUCs are inserted in the upper right corners; Predictions: solid line; Observations: (error bars: 95% Binomial confidence intervals).

Notably, 62% of those who developed skin toxicity also exhibited a breast V₂₀ > 50%, whereas many of the remaining patients with skin toxicity had a V₂₀ ~ 45%, suggesting that limiting the breast V₂₀ below 45% may further improve treatment tolerability.

Discussion:

In this five-year report of a phase II single-arm study of 40 Gy APBI in ten daily fractions, this convenient regimen exhibits minimal toxicity with excellent disease control among appropriately selected patients. Our findings expand on a body of literature that has largely focused on twice-daily APBI regimens administered over a single week, with mixed results as discussed below. In addition, analysis of relevant dosimetric parameters among our cohort suggests that breast V₂₀ and absolute PTV volume are significantly predictive of grade ≥2 skin toxicity.

Ever since a series of landmark studies demonstrated that the combination of breast conserving surgery and whole-breast irradiation is at least as effective as mastectomy at establishing long-term disease control, the optimal radiation target volume has been subject to further investigation. Several lines of evidence suggest that targeting the entire breast may be avoidable in a subset of patients, many of whom can likely be identified based on clinicopathologic features.

Anatomic data, for example, have demonstrated that for most patients the majority of malignant cells in the breast are found proximate to the primary tumor. A review of 217 mastectomy specimens found that among lesions without an extensive intraductal component (EIC), it was rare for cancer foci to be found more than 2 cm beyond the edge of the tumor mass ⁷. Similarly, patients who undergo re-excision of the primary site because of initially inadequate margins are more likely to have residual tumor in the re-excision specimen if the initial excision shows an EIC-positive tumor^8,9,22,23.

The possibility of multiple synchronous cancers in the breast has also been used to justify giving whole-breast irradiation; however, in a study of 183 mastectomy specimens, only 3 patients (1.6%) had multiple lesions without any histopathologic continuity between them ²⁴.

Patterns-of-failure studies further characterize areas at risk within the breast. In the first 5-10 years after BCT, most recurrences are at or near the original tumor bed, although with increasing follow-up, a larger proportion of recurrences arise in other quadrants ²⁵. This observation suggests that two processes are actually reflected in the aggregate incidence of breast “recurrence”: 1) the early regrowth of residual tumor cells following initial therapy, and 2) the later development of metachronous primaries unrelated to the original lesion.

A number of randomized trials have compared partial- to whole-breast irradiation, informing appropriate patient selection. An early British study randomized 708 patients to breast and regional nodal RT versus treatment to the affected quadrant alone (40-42.5 Gy in 8 fractions delivered over 10 days) ^26,27. At a median follow-up of 65 months, the 7-year LR rates were 11% and 20%, respectively. Strikingly, this difference was largely driven by tumor histology, as 7-year LR for invasive ductal carcinoma (IDC) was 11% with whole-breast RT versus 15% with APBI, in contrast to invasive lobular carcinoma (ILC) which exhibited a 7-year LR of 8% versus 34%. Moreover, recurrences arose in the same quadrant as the index lesion for 64% of IDCs versus only 38% of ILCs.

A subsequent Hungarian study randomized patients with up to pT1N1a non-lobular breast cancers to whole-breast (50 Gy) or partial-breast RT (via either high-dose rate brachytherapy using 7.2Gy × 5 fractions or 50Gy partial breast electron beam RT) ²⁸. At a median follow-up of 10.2 years, the ten-year actuarial rate of LR was 5.1% in the whole-breast RT arm versus 5.9% in the PBI arm (p=0.77), supporting this stringent patient selection.

The European GEC-ESTRO group conducted an APBI study randomizing 551 lumpectomy patients to whole-breast RT or APBI using multi-catheter brachytherapy ²⁹. At 5-years, the incidence of LR was 1.44% with APBI versus 0.52% with whole-breast RT (p=0.42). Notably, the risk of grade 2-3 late adverse effects was 3.2% with APBI versus 5.7% with whole-breast RT (p=0.08), demonstrating the tolerability of multi-catheter brachytherapy.

IMPORT-LOW, a major phase 3 study randomized 2,018 women in 1:1:1 fashion to receive 40Gy whole-breast RT, 36Gy whole-breast RT and 40Gy to the partial breast, or 40Gy to the partial breast only ³⁰. At a median follow-up of 72.2 months, 5-year LR estimates were 1.1%, 0.2% and 0.5%, respectively, demonstrating non-inferiority of partial-breast RT in terms of LR. Moreover, the partial breast approach yielded significantly lower rates of changes in breast appearance or firmness, suggesting improved tolerability.

This growing literature, in addition to our findings, contrasts with preliminary toxicity reports from the RAPID trial which randomized 2,135 women with ≤3cm invasive or in situ breast cancer to 38.5Gy APBI in 10 fractions twice daily or whole-breast irradiation ³¹. At a median 3 years, adverse cosmesis was notably worse in the APBI arm (29%) versus the standard whole-breast RT arm (17%; p<0.001) prompting caution in the use of external beam techniques.

The final report of a landmark NSABP study (B-39) will shed further light on this topic. B39 randomized 4,216 patients to whole-breast RT versus 4-5 days of APBI (via 3.4-3.85 Gy given twice daily with either brachytherapy or external beam approaches). While the initial report showed that PBI did not meet prespecified criteria for equivalence with whole-breast RT, the absolute difference in the 10-year rate of in-breast recurrence was small (4.8% APBI versus 4.1% whole-breast RT).³² Furthermore, an unplanned subset analysis showed that the rate of recurrence was potentially higher among those receiving brachytherapy than external-beam RT. Publication of this study promises to inform the future of APBI, but does not address the feasibility of a once-per-day fractionation scheme, which is preferable for some patients.

Our results must be interpreted in the context of the study design which was a single-arm trial, utilizing pragmatic toxicity reporting instruments and questionnaires. Given the overall low event rate, comparison to historical reports suggests that this regimen is feasible for further development. A possible limitation is that first follow-up was conducted at 4-6 weeks following RT per our routine standard, potentially missing an increase in shorter-term peak toxicity, although we found no clinical evidence to suggest this was the case. Moreover, strengths of this study include a uniformity of radiotherapeutic technique across the cohort, strict dosimetric evaluation, robust patient follow-up during the period under investigation, and a unique fractionation scheme that remains hitherto unexplored.

Thus, there is a large and growing body of literature that supports both the biologic rationale and clinical feasibility of APBI via multiple approaches including various brachytherapy and external beam techniques. Here, we demonstrate that 40Gy external beam APBI administered once-daily over ten days represents a convenient, well-tolerated and effective regimen among appropriately selected patients. The findings of this study are in marked contrast to the results of the APBI arm of the RAPID trial, with low rates of acute and late toxicity observed. Although difficult to quantify, the quality of surgery is another major variable determining the cosmetic outcome in the breast. The increased use of sentinel lymph node biopsies in this period also assists in improved tolerance of radiotherapy in the breast. Longer-term follow-up of this cohort and others will ultimately determine the optimal dose, fractionation and overall treatment time of partial-breast RT approaches.

Whereas external beam accelerated partial breast irradiation has traditionally been administered using twice-daily fractionation, we sought to develop a more convenient once-daily regimen (40 Gy administered in ten daily fractions over two weeks). At a median follow-up of nearly 5 years, we identified low rates of local toxicity among >100 accrued patients and noted only three local recurrences, suggesting that this regimen is both well-tolerated and effective on this time scale, and merits further study.