Distribution of Locoregional Breast Cancer Recurrence in Relation to Postoperative Radiation Fields and Biological Subtypes

International Journal of

Radiation Oncology biology

physics

www.redjournal.org

Clinical Investigation

Distribution of Locoregional Breast Cancer Recurrence in Relation to Postoperative Radiation Fields and Biological Subtypes Jamila Adra, MD,*,1 Dan Lundstedt, MD, PhD,*,1 Fredrika Killander, MD, PhD,y Erik Holmberg, PhD,* Mahnaz Haghanegi,y Elisabeth Kjelle´n, MD,y Per Karlsson, MD,* and Sara Alkner, MD, PhDy *Department of Oncology, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Sahgrenska University Hospital, Gothenburg; and yDivision of Oncology and Pathology, Institute of Clinical Sciences, Lund University, Skane University Hospital, Lund, Sweden Received Mar 27, 2019. Accepted for publication Jun 1, 2019.

Summary The pattern of locoregional recurrence (LRR) in relation to postoperative locoregional radiation therapy and primary breast cancer biology has been analyzed. The 10year cumulative incidence of LRR was 7.1%. Most LRR developed within the irradiated volume, whereas the most common location for out-of-field LRR was the cranial supraclavicular fossa. A higher LRR frequency was

Purpose: To investigate incidence and location of locoregional recurrence (LRR) in patients who have received postoperative locoregional radiation therapy (LRRT) for primary breast cancer. LRR-position in relation to applied radiotherapy and the primary tumor biological subtype were analyzed with the aim of evaluating current target guidelines and radiation therapy techniques in relation to tumor biology. Methods and Materials: Medical records were reviewed for all patients who received postoperative LRRT for primary breast cancer in southwestern Sweden from 2004 to 2008 (N Z 923). Patients with LRR as a first event were identified (n Z 57; distant failure and death were considered competing risks). Computed tomographic images identifying LRR were used to compare LRR locations with postoperative LRRT fields. LRR risk and distribution were then related to the primary breast cancer biologic subtype and to current target guidelines. Results: Cumulative LRR incidence after 10 years was 7.1% (95% confidence interval [CI], 5.5-9.1). Fifty-seven of the 923 patients in the cohort developed LRR (30 local recurrences and 30 regional recurrences, of which 3 cases were simultaneous local and

Corresponding author: Sara Alkner, MD, PhD; E-mail: Sara.Alkner@ med.lu.se 1 Co-first authors. This work was supported by Skane County Council’s Research and Development Foundation with Governmental Funding for Clinical Research from the National Health Service; by the Swedish state under an agreement between the Swedish government and county councils (the ALF-agreement); by the King Gustav the V Jubilee Clinic Foundation; and by the Percy Falk Foundation; and by the Swedish Cancer Society. Disclosures: P.K. declares a research contract with PFS Genomics. None of the funding sources involved had any impact on the study’s Int J Radiation Oncol Biol Phys, Vol. -, No. -, pp. 1e11, 2019 0360-3016/$ - see front matter Ó 2019 Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.ijrobp.2019.06.013

design; the collection, analysis, and interpretation of data; the writing of the report; or the decision to submit the article for publication. Supplementary material for this article can be found at https://doi.org/ 10.1016/j.ijrobp.2019.06.013. AcknowledgmentsdThe authors thank the staff of the Western and Southern breast radiotherapy group for the development of breast radiotherapy guidelines with special thanks to Per Malmstro¨m, Lena Wittgren, Gudrun Bankvall, Go¨ran Bjelkengren, Kalifatidis Dimitiros, Ha˚kan Griph, Sonny Jonsborg, Andrej Tomaszewicz, and Karl-Axel Johansson.

2

International Journal of Radiation Oncology  Biology  Physics

Adra et al.

observed in ER- or HER2þ breast cancer. In addition, ER- or HER2þ tumors more frequently recurred in field (68%) rather than marginally or out-of-field (32%).

regional recurrence). Most cases of LRR developed fully (56%) or partially (26%) within postoperatively irradiated areas. The most common location for out-of-field regional recurrence was cranial to radiation therapy fields in the supraclavicular fossa. Patients with an estrogen receptor negative (ERe) (hazard ratio [HR], 4.6; P < .001; 95% CI, 2.5-8.4) or HER2þ (HR, 2.4; P Z .007; 95% CI, 1.3-4.7) primary breast cancer presented higher risks of LRR compared with those with ERþ tumors. ER-/ HER2þ tumors more frequently recurred in-field (68%) rather than marginally or out-of-field (32%). In addition, 75% of in-field recurrences derived from an ER- or HERþ tumor, compared with 45% of marginal or out-of-field recurrences. A complete pathologic response in the axilla after neoadjuvant treatment was associated with a lower degree of LRR risk (P Z .022). Conclusions: Incidence and location of LRR seem to be related to the primary breast cancer biologic subtype. Individualized LRRT according to tumor biology may be applied to improve outcomes. Ó 2019 Elsevier Inc. All rights reserved.

Introduction Adjuvant locoregional radiation therapy (LRRT; ie, to the remaining breast or thoracic wall and to regional lymph nodes) has been shown to prevent locoregional recurrence (LRR) and to improve survival in women with node-positive breast cancer (BC).1 However, treatment presents risk of toxicities such as lymphedema and clinical pneumonitis.2,3 In addition, an increased long-term risk of lung cancer and cardiac-related mortality has been demonstrated, especially among smokers.4 Over the last few decades, radiation therapy (RT) techniques have developed from simple beam arrangements to 3-dimensional techniques with definition of target volumes based on soft tissue anatomy shown on computed tomographic imaging. Recent technological advances, such as intensity modulated radiation therapy and proton therapy, allow for an even higher degree of conformity.5,6 Although this advantage can reduce doses given to normal tissues, it renders target volume definition and knowledge of LRR patterns even more important to prevent risking underdosing tissue at high risk of microscopic disease. Although systemic adjuvant BC treatment is currently individualized according to patient and tumor characteristics, LRRT is generally not modified by BC subtype or histologic grade. There are, however, indications that triple negative and HER2þ BC might be more radioresistant and thus present a higher risk of recurrence within irradiated areas.7,8 Customized target volumes and RT doses according to tumor biology could lead to improved outcomes with fewer adverse effects. In 2003, Swedish radiation oncologists in western and southern Sweden defined guidelines for target contouring and treatment planning for LRRT after BC surgery (modified radical mastectomy [MRM] or breast-conserving surgery [BCS]). We analyzed outcomes for patients with BC receiving LRRT in southwestern Sweden between 2004 and 2008 to identify frequency and location of LRR in relation to delivered RT with the aim to evaluate target definition guidelines. A few recent publications have investigated LRR patterns in relation to a fictive target volume according to ESTRO/RTOG (European Society for

Radiotherapy and Oncology / Radiation Therapy Oncology Group) contouring guidelines.9-12 Instead, we focus on LRR development and locations in relation to applied RT and tumor biology.

Methods and Materials Study cohort The initial cohort includes all patients registered as having received postoperative LRRT in southwestern Sweden from 2004 to 2008 (N Z 1260). This 5-year interval was chosen because Swedish national guidelines for LRRT target definition and treatment planning were accepted in 2003 (described as follows and in the supplementary material, radiotherapy techniques and dose fractionation schedules recommended by the Swedish breast cancer group during the study period - Appendix 1, found online at https://doi. org/10.1016/j.ijrobp.2019.06.013). The cohort includes patients operated either with BCS or MRM, and all patients received LRRT to the remaining breast and thoracic wall, axilla level IIþIII, interpectoral lymph nodes (LNs), and the supraclavicular fossa (SCV). Ninety-nine percent of patients had an axillary lymph node dissection with 4 LNs removed. Reasons for exclusion are given in Figure 1. Three hundred thirty-seven women were excluded, leaving 923 women in the study cohort. The study was approved by the Regional Ethical Review Boards (LU 2013/742, Gbg 030-15) and was completed in accordance with The Code of Ethics of the World Medical Association.

Radiotherapy In 2003, radiation oncologists in western and southern Sweden defined guidelines for LRRT target definition and treatment planning after BC surgery. These guidelines were presented at the meeting of the Nordic Radiotherapy Society in Bergen of 2004,13 and they have since been used by Swedish oncological departments and are listed on the Swedish Breast Cancer Group’s Homepage (www.swebcg.

Volume -  Number -  2019

Pattern of locoregional breast cancer relapse

3

1260 patients included (registred as having received LRRT for breastcancer in southwestern Sweden 2004-2008)

337 patient excluded 147 RT for generalized disease/recurrence 104 not registred correctly 62 prevous breast cancer 2 did not complete LRRT 3 neoadjuvant LRRT 19 charts not found

923 patients in the study cohort

30 patients with regional recurrences* 27 patients with local recurrences*

9 axillary 18 supraclavicular 1 parasternal 1 parasternal + SCV 1 parasternal + SCV + axilla

Fig. 1. Inclusion versus exclusion flowchart for the study cohort. Reasons for exclusion included the following: radiation therapy (RT) given as treatment for locoregional recurrence or disseminated disease (n Z 147), incorrect registration (n Z 104), (only RT toward the breast or thoracic wall (n Z 96), RT during 2003 (n Z 4), treated for plasmocytoma (n Z 1), “test person” (n Z 1), treated for prostatic cancer (n Z 1), adjusted treatment owing to former RT for Hodgkin disease (n Z 1)), breast cancer previous to the tumor treated in 2004 to 2008 (n Z 62), did not complete an RT course (n Z 2), neoadjuvant RT (n Z 3), and no chart available (n Z 19). *Three patients presented simultaneous local and regional recurrence. They are here included among the 30 patients with regional recurrences. Abbreviations: LRRT Z locoregional radiotherapy; SCV Z supraclavicular fossa. se; supplementary material, found online at https://doi.org/ 10.1016/j.ijrobp.2019.06.013). The planning target volume includes the breast and thoracic wall, axilla levels IIþIII, interpectoral lymph LNs, and the SCV. To reduce doses given to the heart and lung, level I was not included. However, approximately 80% of level I is still estimated to be covered by the tangential breast and thoracic wall fields.14 In concordance with Swedish national guidelines of the time, internal mammary nodes (IMNs) were not included in the target volume. LRRT during the study period generally was given to patients with 4 LN metastases, after neoadjuvant chemotherapy, and to those with advanced tumors (T3-T4). Toward the end of the study period, local treatment guidelines started to recommend LRRT for patients with 13 LN metastases as well. The recommended dose was set to 50 Gy given in 25 fractions with an additional increase of 16 Gy given in 8 fractions either to the tumor site for patients age 40 years or younger and who underwent BCS or to the positive surgical margins in rare cases in which additional surgery was not possible. Additional details regarding radiation techniques are available in the supplementary material (found online at https://doi.org/10.1016/j. ijrobp.2019.06.013).

Definition and classification of LRR LRR was defined as a recurrent disease of the ipsilateral breast or thoracic wall (local recurrence [LR]) or of the ipsilateral axillary, supraclavicular, infraclavicular, or IMNs (regional recurrence [RR]). Computed tomography (CT) or positron emission tomography (PET)-CT images identifying recurrence were used by 2 radiation oncologists to delineate local and regional recurrences on the CT that had been used for RT planning (blinded for the RT plan). The RT plan used for the postoperative LRRT was then compared with the LRR location. In a few cases, no CT results demonstrating LRR could be retrieved. If it from the description given on the chart and from mammographic images was clear whether LRR fell within postoperative LRRT fields or not (ie, local recurrences clearly within the remaining breast parenchyma), the LRR was classified accordingly. For 7 patients, however, LRR positioning in relation to the given RT could not be identified. These patients were classified as missing in the analyses of LRR locations in relation to previous RT procedures. LR taking the form of multiple cutaneous metastases was generally considered as one recurrence; however, 2 distinctly separate areas of local metastasis (one of which might fall within the

4

Adra et al.

previous RT field while the other is positioned outside) were considered 2 separate cases of LR. Five patients had LN metastases in the contralateral axilla at the time of LRR diagnosis. Although contralateral LN metastases in some cases could represent an RR rather than a disseminated disease,15 we evaluated LRR location in relation to postoperative LRRT. Hence, only ipsilateral LRRs were considered in the study, and patients with isolated contralateral recurrences were censored at this event. LRR cases were classified as in field when recurrence developed within postoperative RT fields and as out of field when not covered by the RT-plan. All LRR cases falling partially outside or on the field edge were defined as marginal.

Biological subtype The biological subtype of the primary BC was categorized according to the following: Luminal A-like (estrogen receptor [ER]þ, HER2e and Nottingham histological grade [NHG] 1, or NHG 2 if Progesterone receptor [PR]þ), Luminal B-like (ERþ, HER2e, and NHG 3 or NHG 2 if PR-), HER2þ (all HER2þ tumors), and Triple Negative (ERePReHER2e). This approach echoes the 13th St Gallen BC expert recommendations.16 However, we could not use Ki67 over NHG to separate Luminal A-like and Blike because Ki67 was not analyzed at the time of diagnosis for the majority of these tumors. In total, 136 tumors lacked data on HER2 and 13 on NHG or PR. In addition, 7 tumors were ERePRþHER2e. These tumors were classified as ERþ other (i.e., ERþ with unknown HER2 or unknown for Luminal A/B) or ERe other (i.e., ERe with unknown HER2 or ERePRþHER2e). ER, PR, and HER2 status was abstracted from pathology records and regional data registers. In line with Swedish clinical standards during this period, tumors with 10% stained nuclei or greater were considered ERþ or PRþ. HER2 was initially investigated by immunohistochemistry. Tumors scored 2þ or 3þ with immunohistochemistry (HercepTest) were analyzed further with fluorescence in situ hybridization, and only amplified tumors were considered HER2þ.

Statistical analysis To investigate patterns of LRR in relation to postoperative LRRT, the primary endpoint chosen was LRR, and competing events were defined as disseminated disease or death without BC recurrence. Survival was measured from the start of LRRT, and patients were followed until the first of the following events: LRR, disseminated disease, death, or the last follow-up list on the chart. The presence of contralateral BC or another malignancy did not exclude the patients from follow-up. Isolated recurrences in the contralateral axilla were censored at this event. For statistical calculations, the Stata 13.1/15.1 (StataCorp) software package was used. Cumulative LRR incidence was estimated using a competing risk method and

International Journal of Radiation Oncology  Biology  Physics

was summarized graphically. A cause-specific Cox regression treating competing events as censoring was used to estimate hazard ratios (HRs). Associations between LRR and other variables were evaluated with a c2 for trend. Assumptions of proportional hazards were checked graphically. Ninety-five percent confidence intervals (CIs) corresponding to a P value threshold of .05 were used to summarize the variability of estimated effects.

Results From 2004 to 2008, 923 patients received LRRT for primary BC at the study sites. The median follow-up was 9.3 years (range, 0.06-14.2 years) for patients without an event. Patient and tumor characteristics are described in Table 1. Eighty-one percent of patients received chemotherapy (Table 1). In 17% of patients receiving chemotherapy (127 of 733), it was given neoadjuvantly, in 74% adjuvantly (541 of 733), and 9% (65 of 733) received both neoadjuvant and adjuvant chemotherapy. If no contraindications were present, a combination therapy including taxanes was used, with the most common combination being fluorouracil, epirubicin, and cyclophosphamide followed by docetaxel. Seventy-six percent of all patients and 99% of ERþ patients received endocrine treatment, and 64% of HER2þ patients received trastuzumab. During follow-up periods, 39 patients developed an isolated LRR as a first event, 18 patients developed LRR plus simultaneous disseminated disease (within 3 months after LRR diagnosis), 329 patients developed disseminated disease, and 64 patients died of causes not related to BC. The cumulative incidence of a first event after 10 years was 7.1% for LRR (with or without simultaneous disseminated disease), 37.9% for distant recurrence, and 8.2% for death without BC recurrence (Table 2). Four hundred seventy patients were alive without BC recurrence by the end of the follow-up period.

LRR in relation to patient and tumor characteristics The median time to LRR was 2.5 years (range, 0.27-10.2 years). Of the 57 LRR patients, 30 had LR and 30 had RR (3 patients with simultaneous LR and RR; Table 2, Fig. 1). Of the cases of RR, 9 (30%) were axillary, 18 (60%) were supraclavicular, and 1 (3%) was parasternal (IMN). One patient (3%) experienced simultaneous RR in SCV and IMN, and 1 patient (3%) experienced simultaneous RR in SCV, axilla, and IMN (Table E1). The majority of patients with an LRR diagnosis (n Z 36) received a diagnosis of disseminated BC within 3 months. Among the remaining 21 patients, 8 later developed distant BC metastases, 1 died of causes not related to BC, and 12 remained BC-free at the last follow-up 0.33 to 12.5 years after LRR. Figure E1 shows survival after diagnosis of an isolated LRR.

Volume -  Number -  2019 Table 1

Pattern of locoregional breast cancer relapse

Patient and tumor characteristics Characteristics

Age at diagnosis <50 years 50 years Median (range) Type of surgery MRM BCS Missing Neoadjuvant treatment No Chemotherapy Endocrine Chemotherapy plus endocrine Missing Node status* Without neoadjuvant treatment, median (range) 0 LN-met 1-3 LN-met 4-9 LN-met 10 LN-met After neoadjuvant treatment, median (range)y 0 LN-met 1-3 LN-met 4-9 LN-met 10 LN-met Missing Tumor size (mm) Without neoadjuvant treatment, median (range) 20 mm >20 mm After neoadjuvant treatment, median (range)y 20 mm >20 mm Missing Multifocal tumor No Yes Missing Histology Ductal Lobular Other Missing ER status <10% 10% Missing PR status <10% 10% Missing HER2 status Negative

5

Table 1 (continued )

Number (%) (N Z 923) 293 (32) 630 (68) 57 (27-91) 763 (83) 158 (17) 2 705 190 13 1

(78) (21) (1) (0) 14

6 (0-41) 7 84 398 213 4 40 49 71 41

(1) (12) (57) (30) (0-22) (20) (24) (35) (20) 20

30 (4-150) 206 (29) 493 (71) 30 (0-180) 62 (34) 122 (66) 40 680 (80) 169 (20) 74 665 (75) 198 (22) 26 (3) 34 233 (26) 674 (74) 16 430 (48) 472 (52) 21 597 (78) (continued)

Characteristics Positive Missing Nottingham histologic grade 1 2 3 Missing Biologic subtype ER positive Luminal A-like Luminal B-like ERþ other HER2þ ERe Triple negative ERe other Missing Endocrine treatment Yes No Missing Chemotherapy Yes No Missing Trastuzumab if HER2þ Yes No Missing

Number (%) (N Z 923) 169 (22) 157 81 (9) 404 (47) 382 (44) 56 594 240 226 128 170 144 110 34

(65) (26) (25) (14) (19) (16) (12) (4) 15

688 (76) 221 (24) 14 733 (81) 176 (19) 14 106 (64) 59 (36) 5

Abbreviations: BCS Z breast conserving surgery; ER Z estrogen receptor; LN-met Z lymph node metastases; MRM Z modified radical mastectomy; PR Z progesterone receptor. * Median number of lymph nodes extracted was for all patients, 15 (range, 1-43); without neoadjuvant treatment, 15 (range, 1-43); with neoadjuvant treatment, 13 (range, 2-34). Ninety-nine percent of patients had an axillary lymph node dissection with 4 lymph nodes removed. y At surgery after neoadjuvant treatment.

Table 2 and Figure 2 show the cumulative incidence of first endpoint or LRR in relation to patient and tumor characteristics. Neoadjuvant treatment (P Z .048) and biologic subtypes (P < .001) were significantly associated with the cumulative incidence of LRR. Additional analyses using Cox regressions show neoadjuvant treatment, a large number of LN metastases, high histologic grade (grade 3), and triple-negative or HER2þ biologic subtypes corresponding to a higher degree of LRR risk (Tables 2 and 3). Because only 6 patients (6 of 204 Z 3%) were treated with BCS after neoadjuvant treatment, the increased level of LRR risk observed after neoadjuvant treatment did not appear to depend on the use of less surgery. Interestingly, none of the 40 patients with benign lymph nodes at the point of surgery after neoadjuvant treatment developed LRR compared with 12% (19 of 161) of patients presenting remaining LN metastases after treatment (c2 test, P Z .022).

6

International Journal of Radiation Oncology  Biology  Physics

Adra et al.

Table 2 Patient and tumor characteristics in relation to the first end-point of the follow-up period: Cumulative incidence and 95% confidence interval for the first event after 10 years N Z 912*

Locoregional recurrence, Distant recurrence, Death without % (95% CI) % (95% CI) recurrence, % (95% CI)

Total Age (year) <50, n Z 290 (32%) 50, n Z 622 (68%) Type of surgery MRM, n Z 754 (83%) BCS, n Z 156 (17%) Missing, n Z 2 Neoadjuvant treatment No, n Z 696 (78%) Yes, n Z 202 (22%) Missing, n Z 14 Lymph-node metastases (number)y 0, n Z 47 (5%) 1-3, n Z 134 (15%) 4-9, n Z 471 (52%) 10, n Z 251 (28) Missing, n Z 9 Tumor size (mm)y 20, n Z 266 (30%) >20, n Z 612 (70%) Missing, n Z 34 Nottingham histologic grade 1, n Z 80 (9%) 2, n Z 399 (47%) 3, n Z 377 (44%) Missing, n Z 56 Biological subtype ERþ, n Z 590 (66%) Luminal A, n Z 237 (26%) Luminal B, n Z 226 (25%) ERþ other, n Z 127 (14%) ERe, n Z 139 (16%) Tripple Negative, n Z 106 (12%) ERe other, n Z 33 (4%) HER2þ, n Z 168 (19%) Missing n Z 15

7.1 (5.5-9.1) P Z .832 7.9 (5.0-12.3) 6.8 (5.0-9.2) P Z .138 6.4 (4.8-8.5) 10.8 (6.4-17.7)

37.9 (34.5-41.5) P Z .923 37.3 (31.4-43.9) 38.2 (34.1-42.5) P < .001 41.3 (37.5-45.3) 20.3 (14.3-28.6)

8.2 (6.3-10.5) P < .001 0.7 (0.1-5.2) 11.4 (8.8-14.7) P Z .537 8.4 (6.4-11.1) 7.0 (3.5-13.7)

P Z .048 6.2 (4.6-8.5) 10.1 (6.5-15.6)

P Z .008 36.1 (32.2-40.2) 43.7 (36.9-51.2)

P Z .004 9.4 (7.1-12.3) 3.0 (1.1-7.9)

P Z .117 0 5.8 (2.8-11.8) 7.0 (4.8-10.1) 9.3 (6.2-13.8)

P < .001 10.8 (4.6-24.0) 26.2 (19.2-35.2) 32.4 (28.0-37.5) 59.0 (52.4-65.8)

P Z .275 11.7 (4.5-28.7) 11.2 (6.3-19.7) 8.6 (6.1-12.2) 4.6 (2.5-8.4)

P Z .487 8.3 (5.3-12.8) 6.6 (4.8-9.1)

P < .001 24.5 (19.3-30.7) 42.8 (38.6-47.3)

P Z .740 9.7 (6.2-15.2) 7.9 (5.8-10.7)

P Z .073 1.3 (0.2-9.0) 7.0 (4.7-10.2) 9.2 (6.5-12.9)

P < .001 25.8 (16.1-39.9) 35.3 (30.4-40.8) 41.7 (36.5-47.3)

P Z .321 15.2 (7.7-28.9) 8.7 (6.0-12.4) 7.0 (4.6-10.8)

P < .001 4.8 (3.2-7.1) 3.0 (1.4-6.6) 7.4 (4.3-12.3) 15.7 (10.3-23.4) 19.5 (12.8-29.2)

P Z .029 38.4 (34.2-42.9) 32.9 (26.7-40.2) 43.0 (36.1-50.6) 37.4 (29.7-46.2) 36.9 (28.3-47.1)

P Z .528 10.0 (7.5-13.2) 8.4 (5.1-13.6) 6.9 (3.9-12.0) 6.7 (3.1-14.0) 6.9 (2.8-16.5)

8.8 (5.3-14.4)

33.7 (26.7-42.0)

3.5 (1.5-8.3)

Abbreviations: BCS Z breast conserving surgery; CI Z confidence interval; ER Z estrogen receptor; MRM Z modified radical mastectomy; NHG Z Nottingham histological grade. * Recurrence status missing for 1 patient, and 8 patients with missing follow-up data. In addition, 2 patients with isolated recurrence in the contralateral axilla excluded. y These analyses include all patients; however, separate analyses were done for patients receiving neoadjuvant versus no neoadjuvant treatment with similar results.

Location of LRR in relation to postoperative radiotherapy Of the 30 LR cases, 19 (63%) developed after MRM and 11 (37%) developed after BCS. The corresponding values for RR were 27 (90%) after MRM and 3 (10%) after BCS; this was due to MRM being more common than BCS in the cohort, and no significant difference in LRR frequency was seen in relation to operation technique (Table 2). Most LRR cases developed fully (56%) or partially (26%) within postoperatively irradiated areas, and this was observed after MRM and BCS (Table E2). Only 3 patients developed LRR

fully out of postoperative RT fields, and all did so after MRM (Tables E1 and E2). Figure 3 shows LRR locations in relation to postoperative LRRT and biologic subtype (red denotes infield LRR, blue denotes out-of-field LRR, and green denotes marginal LRR. Circle denotes ERþHER2e subtype, square ER-HER2e, and triangle HER2þ). Axial CT images of LRR locations are available as Figure E2D. In the images, LRR after MRM is shown on the right side of the patient, whereas LRR after BCS is shown on the left side of the patient. However, in reality, 76% of the LRR cases were left sided, whereas

Volume -  Number -  2019

A

B 0.4

Cumulative incidence

0.20

Cumulative incidence

First event Distant recurrence LRR Death

0.3

7

Pattern of locoregional breast cancer relapse

0.2

Number of LN−metastases 0 1−3 4−9 ≥ 10

0.15

0.10

0.05

P CIF = .117

0.00

0.1

0

1

2

3

4

5

6

7

8

9

10

37 64 212 80

37 57 187 73

34 47 164 59

20 33 111 35

Years since start of treatment At risk 0 47 1 − 3 134 4 − 9 471 ≥ 10 251

0.0 0 At risk 912

1

2

833

732

3 4 5 6 7 Years since start of treatment 643

582

504

431

396

8

9

10

355

305

199

C

44 115 386 181

43 101 348 146

39 81 275 105

37 71 233 87

Subtype

Cumulative incidence

0.25

0.15

0.10

0.05 P CIF < .001 0

1

2

3

4

5

6

7

8

9

Luminal A Luminal B HER2 + TN

0.20 0.15

P CIF < .001

0.10 0.05 0.00

0.00

0

10

1

2

At risk ER + 590 ER − 139 HER2 + 168

558 112 153

510 82 130

448 74 113

400 68 107

349 57 91

299 48 78

3

4

5

6

7

8

9

10

131 92 72 32

118 80 67 29

103 68 60 26

67 41 39 17

Years since start of treatment

Years since start of treatment 275 44 72

244 40 67

207 36 60

At risk Luminal A Luminal B HER2 + TN

132 27 39

E

237 226 168 106

228 208 153 85

212 187 130 58

195 158 113 54

178 142 107 50

156 123 91 41

139 103 78 35

F 0.20

0.20

Nottingham histological grade

Neoadjuvant treatment No Yes

NHG 1 NHG 2 NHG 3

0.15

0.10

0.05 P CIF = .073

Cumulative incidence

0.15 Cumulative incidence

41 93 318 126

D ER HER2 ER + ER − HER2 +

0.20 Cumulative incidence

46 124 435 221

0.10

0.05 P CIF = .043

0.00 0

1

2

3

4

5

6

7

8

9

10

0

Years since start of treatment At risk NHG 1 80 NHG 2 399 NHG 3 377

0.00 1

2

3

4

5

6

7

8

9

10

306 86

273 79

233 70

155 43

Years since start of treatment 76 380 330

74 342 272

69 309 229

64 282 206

53 247 176

47 208 150

44 194 134

41 174 117

34 151 100

21 102 62

At risk No 696 Yes 202

641 181

572 149

514 120

465 110

397 101

333 93

Fig. 2. Cumulative incidence of first events during the follow-up period. (A) Locoregional recurrence, distant recurrence, or death from causes other than breast cancer. (B) Cumulative incidence of locoregional recurrence in relation to the number of lymph-node metastases, (C) ER/HER2 status, (D) Biologic subtype, (E) Nottingham histologic grade, and (F) Neoadjuvant treatment. Abbreviations: CIF Z cumulative incidence function; ERþ Z estrogen receptor positive; ERe Z estrogen receptor negative; HER2þ Z HER2 positive; LN Z lymph node; LRR locoregional recurrence; NHG Z Nottingham histological grade; TN Z triple negative. 24% were right sided (Figure E3). Marginal or out-offield LRs were found on the lateral thoracic wall (n Z 5), anterior to the sternum (n Z 3), on the upper arm (n Z 1), and below the breast region (n Z 1). Marginal or out-of-field RRs were located in cranial,

medial, or dorsal locations in the SCV (n Z 7), in level 1 of the axilla (n Z 3), and in the IMN (n Z 3). All cases of LRR affecting patients who did not have disseminated disease within 3 months of LRR were fully or marginally covered by the given RT fields.

8

International Journal of Radiation Oncology  Biology  Physics

Adra et al. Table 3

Cox-regression analysis of LRR risk in relation to primary breast cancer characteristics Univariate analysis HR (95% CI), P value

Neoadjuvant treatment No Yes Number of LN metastasesy 0 1-3 4-9 10 Nottingham histologic grade 1 2 3 Biological subtype of primary BC ERþ ERe HER2þ HER2þ without transtuzumabz HER2þ with transtuzumabz

Multivariate analysis* HR (95% CI), P value

Ref. 1.9 (1.1-3.3), P Z .025

Ref. 2.7 (1.5-5.0), P Z .001

Ref. 1.2 (0.5-2.7), P Z .715 2.1 (0.9-4.8), P Z .097

Ref. 1.6 (0.65-3.7), P Z .305 3.0 (1.2-7.3), P Z .016

Ref. 5.3 (0.72-39.3), P Z .101 8.2 (1.1-60.5), P Z .038

Ref. 4.0 (0.5-30.0), P Z .175 4.1 (0.5-31.2), P Z .168

4.6 2.4 3.0 2.2

Ref. (2.5-8.4), P (1.3-4.7), P (1.2-7.5), P (1.0-4.8), P

< .001 Z .007 Z .015 Z .042

5.4 2.5 3.2 2.3

Ref. (2.7-10.6), P < .001 (1.3-5.0), P Z .009 (1.3-8.0), P Z .013 (1.0-5.1), P Z .046

Abbreviations: BC Z breast cancer; CI Z confidence interval; ER Z estrogen receptor; HR Z hazard ratio; LRR Z locoregional recurrence; LN Z lymph node. * Neoadjuvant treatment, number of LN metastases, Nottingham histologic grade, and biological subtype included in the multivariable analysis. y The group with 0 LN metastases not possible to have as reference group or analyze alone because of its small size. These analyses include all patients. However, separate analyses were done for patients receiving neoadjuvant versus no neoadjuvant treatment with similar results. z HER2þ subgroup separated in relation to trastuzumab treatment.

Biologic subtype of primary BC in relation to LRR frequency and locations Patients with ERe or HER2þ primary BC presented a significantly higher risk of developing LRR (Tables 2 and 3, and E3; Fig. 3). A Cox-regression analysis showed an HR of 4.6 for LRR after ERe BC and a value of 2.4 for LRR after HER2þ BC (ERþ BC reference group; Table 3). In patients with HER2þ primary BCs, a nonsignificant trend was found for higher LRR risk in patients not primarily treated with neoadjuvant or adjuvant trastuzumab (Table 3). In addition, a higher percentage of ERþ primary BC cases were luminal B-like compared with luminal A-like among the LRR patients than for the whole cohort (c2 test, P Z .03; Tables 2 and E3). Table E2 shows LRR locations in relation to postoperative RT fields and biologic subtypes. In-field LRR more often derived from ERe/HER2þ tumors than did marginal or out-of-field recurrence (c2 test for trend P Z .02).

Discussion We have investigated LRR patterns in relation to delivered postoperative radiotherapy for a modern high-risk BC cohort with detailed information on tumor biology, treatment, and outcomes. Although LRR-locations in relation to a fictive target volume according to ESTRO/RTOG-

guidelines have been investigated previously,9-12 to our knowledge this is the largest study to be conducted on LRR patterns in relation to applied RT procedures and tumor biology. From 2004 to 2008, 923 patients received LRRT for primary BC in southwestern Sweden. The cumulative incidence for LRR after 10 years was 7.1%, which is lower than the level of LRR incidence reported in the latest Early Breast Cancer Trialists’ Collaborative Group metaanalyses, according to which postoperative RT reduced LRR incidence after MRM from 26.0% to 8.1% and after BCS from 43.0% to 12.4% within 10 years.1,17 However, these meta-analyses included patients randomized through adjuvant RT-trials before 2000, whereas our cohort was treated from 2004 to 2008. Hence, improved pharmacologic adjuvant treatments (eg, an increased use of taxane-based chemotherapy, aromatase inhibitors, and HER2-targeted therapy) might have reduced expected levels of LRR for this study.18,19 More recent studies have shown lower levels of LRR incidence, echoing our findings.2,8,20-22 However, these studies also cover patients with an intrinsically lower degree of LRR risk than the high-risk patients of our cohort; therefore, the results cannot be compared directly. Hence, we believe that a cumulative 10-year LRR incidence rate of 7.1% for this high-risk cohort denotes the use of a well-functioning target and RT technique. Interestingly, significantly higher levels of LRR frequency were observed for patients with ERe/HER2þ primary BC. In addition, primary BC causing LRR within postoperative RT fields was more often of the ERe/

Volume -  Number -  2019

Pattern of locoregional breast cancer relapse

9

Fig. 3. Delineation of locoregional recurrence patterns in computed tomographic (CT) topogram views and axial views. LRR on the right side after MRM, and LRR on the left side after BCS. In-field recurrences are marked in red, marginal recurrences are marked in green, and out-of-field LRRs are marked in blue. A circle represents LRR after ERþ breast cancer, a square LRR after ERe BC, and a triangle LRR after HER2þ BC. (A) Coronal view. (B) Sagittal view after BCS. (C) Sagittal view after MRM. Axial CT sections in Figure E2 (available in the supplements found online at https://doi.org/10. 1016/j.ijrobp.2019.06.013). Abbreviations: BCS Z breast conserving surgery; CT Z computed tomography; ERþ Z estrogen receptor positive; ERe Z estrogen receptor negative; HER2þ Z HER2 positive; LRR Z locoregional recurrence; MRM Z modified radical mastectomy HER2þ subtypes than was BC causing marginal or out-offield recurrence. This finding is in accordance with earlier studies showing that these subtypes are associated with a higher degree of LRR risk and with resistance to RT.7,11,18,23-25 LRR risk associated with HER2 overexpression has been suggested to decline with trastuzumab treatment,7,18 and we found similar indications in this study. Nevertheless, although trastuzumab reduced LRR risk for HER2þ tumors, the difference was not significant, and an increased risk relative to that observed for ERþ tumors remained after treatment. The idea of adjusted RT according to tumor biology is interesting. Adjusted target volumes, fractionation schedules, or the use of radiosensitizing agents could be explored as the means to improve outcomes in high-risk groups (eg, those with ERe/HER2þ tumors). Several ongoing trials are also exploring the possibility of reducing or abstaining adjuvant RT for ERþ low-risk tumors,18,26-29 although a subgroup for which RT could be abstained without increasing LRR-risk has been difficult to identify.8,30 In addition, although we investigate cumulative incidence at 10 years, ERþ BC cases are at risk of very late recurrence, and an even longer follow-up period could change patterns of LRR in relation to the biologic subtypes seen here. In addition to ER/HER2 status, we found other characteristics of a more aggressive tumor phenotype (high histologic grades or the number of LN metastases) to be associated with increased LRR-risk.11,31 Previous studies have shown a high degree of LRR risk after neoadjuvant chemotherapy followed by BCS.31,32 However, we also found neoadjuvant treatment to be associated with increased LRR risk in this cohort in which 97% of neoadjuvantly treated patients were treated with an MRM, possibly because more aggressive tumors were selected for neoadjuvant treatment during the study period. Echoing our

findings, axillary responses to neoadjuvant therapy have been shown to affect recurrence risk.33 Whether LRRT is still needed after observing a pathologically complete response at surgery is currently being investigated.34 Interestingly, 76% of the LRR cases diagnosed in this study were left-sided. Although the reason for this is unclear, one explanation could be that in nonlobular nonmultifocal cases, a reduced coverage to the breast and thoracic wall target outside the tumor bed was allowed to reduce dose to organs at risk. It’s possible that this option was more often used in left-sided cases to reduce heart dose, which might to some extent explain why more leftsided LRRs were seen. The target volumes defined by the Swedish contouring atlas have similarities to the recently published ESTRO consensus guidelines.35 However, under the Swedish guidelines the target volume is generally extended somewhat more cranially in the medial part of the SCV. Interestingly, the most common location for out-of-field RR was cranial for RT fields in the medial SCV. Chang et al9 found similar results when retrospectively investigating whether established LRRs were covered by the clinical target volume (CTV) when radiotherapy is applied according to ESTRO guidelines [9]. Although they found most LRR cases to be covered by the ESTRO CTV, cranial in SCV was one of the most common locations for LRR outside a simulated ESTRO CTV. In addition, other recent studies confirm SCV to be a common site for LRR developing outside of recommended ESTRO/RTOG target volumes.11,12 However, LRR incidence in our study was low, and the majority of LRR cases that occurred did so within postoperative RT fields. Hence, we would not recommend a general cranial extension of CTV in SCV. This approach can, however, be considered for high-risk patients.

10

Adra et al.

Other locations for LRR out-of-field include lateral positions on the thoracic wall, areas anterior to the sternum, areas under the breast, medial or dorsal SCV, level I between the breast and axillary region, and the IMN. Interestingly, for patients free of disseminated BC disease within 3 months of LRR (i.e., those with the highest expected benefits derived from preventing or treating LRR), all recurrences were fully or marginally contained within postoperative RT fields. This finding again suggests the effectiveness of the RT technique and thus that treatment resistance is the main cause of LRR despite prior RT application. A debated issue in Sweden concerns whether LRRT should be applied to the IMN. Although a survival benefit has been shown for IMN-RT (especially for central or medial tumors or 4 LN metastases),36 inclusion generally leads to higher RT doses given to the heart and lungs. We found few IMN recurrences for this high-risk cohort, for which IMN had not been included in the target volume. Although we viewed CT scans from all patients noted to have mediastinal metastases to find IMN recurrences, these are probably more difficult to identify than clinically visible or palpable LRRs. In addition, survival benefits derived from LRRT could mainly depend on the eradication of subclinical metastases and on the prevention of further spread rather than on the prevention of clinically detectable recurrences.

Conclusion In our high-risk BC cohort receiving LRRT during 2004 to 2008, cumulative LRR incidence after 10 years was 7.1%. LRR outside of postoperative RT fields occurred laterally on the thoracic wall, anterior to the sternum, under the breast, cranial, medial or dorsal in SCV, in axilla level I, and in the IMN. However, the majority of LRR cases developed within previous RT fields. Primary BC causing LRR was more frequently observed in cases of ERe/ HER2þ tumors. In addition, ERe/HER2þ primary tumors presented a higher risk of causing LRR within previous RT fields. This warrants further investigation to personalize adjuvant radiotherapy for each patient’s individual needs.

References 1. Early Breast Cancer Trialists’ Collaborative Group, McGale P, Taylor C, et al. Effect of radiotherapy after mastectomy and axillary surgery on 10-year recurrence and 20-year breast cancer mortality: meta-analysis of individual patient data for 8135 women in 22 randomised trials. Lancet 2014;383:2127-2135. 2. Whelan TJ, Olivotto IA, Parulekar WR, et al. Regional nodal irradiation in early-stage breast cancer. New Engl J Med 2015;373:307-316. 3. Gregorowitsch ML, Verkooijen HM, Houweling A, et al. Impact of modern-day axillary treatment on patient reported arm morbidity and physical functioning in breast cancer patients. Radiother Oncol 2019; 131:221-228.

International Journal of Radiation Oncology  Biology  Physics 4. Taylor C, Correa C, Duane FK, et al. Estimating the risks of breast cancer radiotherapy: evidence from modern radiation doses to the lungs and heart and from previous randomized trials. J Clin Oncol 2017;35:1641-1649. 5. Arsene-Henry A, Fourquet A, Kirova YM. Evolution of radiation techniques in the treatment of breast cancer (BC) patients: From 3D conformal radiotherapy (3D CRT) to intensity-modulated RT (IMRT) using Helical Tomotherapy (HT). Radiother Oncol 2017;124:333-334. 6. Verma V, Iftekaruddin Z, Badar N, et al. Proton beam radiotherapy as part of comprehensive regional nodal irradiation for locally advanced breast cancer. Radiother Oncol 2017;123:294-298. 7. Tseng YD, Uno H, Hughes ME, et al. Biological subtype predicts risk of locoregional recurrence after mastectomy and impact of postmastectomy radiation in a large national database. Int J Radiat Oncol Biol Phys 2015;93:622-630. 8. Sjostrom M, Lundstedt D, Hartman L, et al. Response to radiotherapy after breast-conserving surgery in different breast cancer subtypes in the Swedish Breast Cancer Group 91 Radiotherapy Randomized Clinical Trial. J Clin Oncol 2017;35:3222-3229. 9. Chang JS, Lee J, Chun M, et al. Mapping patterns of locoregional recurrence following contemporary treatment with radiation therapy for breast cancer: A multi-institutional validation study of the ESTRO consensus guideline on clinical target volume. Radiother Oncol 2018; 126:139-147. 10. Chang JS, Byun HK, Kim JW, et al. Three-dimensional analysis of patterns of locoregional recurrence after treatment in breast cancer patients: validation of the ESTRO consensus guideline on target volume. Radiother Oncol 2017;122:24-29. 11. DeSelm C, Yang TJ, Cahlon O, et al. A 3-dimensional mapping analysis of regional nodal recurrences in breast cancer. Int J Radiat Oncol Biol Phys 2019;103:583-591. 12. Borm KJ, Voppichler J, Dusberg M, et al. FDG/PET-CT-based lymph node atlas in breast cancer patients. Int J Radiat Oncol Biol Phys 2019;103:574-582. 13. Karlsson P, Bankvall G, Bjelkengren G, et al., West and South Swedish Breast Cancer Radiotherapy Group. Common guidelines of target definitions and treatment techniques in breast cancer radiotherapy in West and South of Sweden. Bergen, Norway: Meeting of the Nordic Radiotherapy Society; 2004. 14. Reznik J, Cicchetti MG, Degaspe B, Fitzgerald TJ. Analysis of axillary coverage during tangential radiation therapy to the breast. Int J Radiat Oncol Biol Phys 2005;61:163-168. 15. Moossdorff M, Vugts G, Maaskant-Braat AJ, et al. Contralateral lymph node recurrence in breast cancer: Regional event rather than distant metastatic disease. A systematic review of the literature. Eur J Surg Oncol 2015;41:1128-1136. 16. Goldhirsch A, Winer EP, Coates AS, et al. Personalizing the treatment of women with early breast cancer: Highlights of the St Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2013. Ann Oncol 2013;24:2206-2223. 17. Early Breast Cancer Trialists’ Collaborative Group, Darby S, McGale P, et al. Effect of radiotherapy after breast-conserving surgery on 10-year recurrence and 15-year breast cancer death: meta-analysis of individual patient data for 10,801 women in 17 randomised trials. Lancet 2011;378:1707-1716. 18. Horton JK, Jagsi R, Woodward WA, Ho A. Breast cancer biology: Clinical implications for breast radiation therapy. Int J Radiat Oncol Biol Phys 2018;100:23-37. 19. Zeidan YH, Habib JG, Ameye L, et al. Postmastectomy radiation therapy in women with T1-T2 tumors and 1 to 3 positive lymph nodes: Analysis of the Breast International Group 02-98 Trial. Int J Radiat Oncol Biol Phys 2018;101:316-324. 20. Haviland JS, Owen JR, Dewar JA, et al. The UK Standardisation of Breast Radiotherapy (START) trials of radiotherapy hypofractionation for treatment of early breast cancer: 10-year follow-up results of two randomised controlled trials. Lancet Oncol 2013;14:1086-1094.

Volume -  Number -  2019 21. Poortmans PM, Collette S, Kirkove C, et al. Internal mammary and medial supraclavicular irradiation in breast cancer. New Engl J Med 2015;373:317-327. 22. Haviland JS, Mannino M, Griffin C, et al. Late normal tissue effects in the arm and shoulder following lymphatic radiotherapy: Results from the UK START (Standardisation of Breast Radiotherapy) trials. Radiother Oncol 2018;126:155-162. 23. Karlsson P, Cole BF, Chua BH, et al. Patterns and risk factors for locoregional failures after mastectomy for breast cancer: An International Breast Cancer Study Group report. Ann Oncol 2012;23:28522858. 24. Langlands FE, Horgan K, Dodwell DD, Smith L. Breast cancer subtypes: response to radiotherapy and potential radiosensitisation. Br J Radiol 2013;86:20120601. 25. Panoff JE, Hurley J, Takita C, et al. Risk of locoregional recurrence by receptor status in breast cancer patients receiving modern systemic therapy and post-mastectomy radiation. Breast Cancer Res Treat 2011;128:899-906. 26. A prospective cohort study evaluating risk of local recurrence following breast conserving surgery and endocrine therapy in low risk luminal A breast cancer (LUMINA). Avaliable at: https://clinicaltrials. gov/ct2/show/NCT01791829. 27. The PRECISION Trial (profiling early breast cancer for radiotherapy omission): A phase II study of breast-conserving surgery without adjuvant radiotherapy for favorable-risk breast cancer. Available at: https://clinicaltrials.gov/ct2/show/NCT02653755. 28. EXPERT. Examining Personalized Radiation Therapy for Low-risk Early Breast Cancer. Available at: https://clinicaltrials.gov/ct2/show/ NCT02889874.

Pattern of locoregional breast cancer relapse

11

29. Kirwan CC, Coles CE, Bliss J, et al. It’s PRIMETIME. Postoperative avoidance of radiotherapy: Biomarker selection of women at very low risk of local recurrence. Clin Oncol 2016;28:594-596. 30. Laurberg T, Tramm T, Nielsen T, et al. Intrinsic subtypes and benefit from postmastectomy radiotherapy in node-positive premenopausal breast cancer patients who received adjuvant chemotherapy - results from two independent randomized trials. Acta Oncol 2018;57:38-43. 31. Valachis A, Mamounas EP, Mittendorf EA, et al. Risk factors for locoregional disease recurrence after breast-conserving therapy in patients with breast cancer treated with neoadjuvant chemotherapy: An international collaboration and individual patient meta-analysis. Cancer 2018;124:2923-2930. 32. Early Breast Cancer Trialists’ Collaborative Group. Long-term outcomes for neoadjuvant versus adjuvant chemotherapy in early breast cancer: Meta-analysis of individual patient data from ten randomised trials. Lancet Oncol 2018;19:27-39. 33. Stecklein SR, Park M, Liu DD, et al. Long-term impact of regional nodal irradiation in patients with node-positive breast cancer treated with neoadjuvant systemic therapy. Int J Radiat Oncol Biol Phys 2018; 102:568-577. 34. ClinicalTrials.gov. NSABPB-51/RTOG-1304 trial (NCT01872975). 2019. 35. Offersen BV, Boersma LJ, Kirkove C, et al. ESTRO consensus guideline on target volume delineation for elective radiation therapy of early stage breast cancer, version 1.1. Radiother Oncol 2016;118:205208. 36. Thorsen LB, Offersen BV, Dano H, et al. DBCG-IMN: A populationbased cohort study on the effect of internal mammary node irradiation in early node-positive breast cancer. J Clin Oncol 2016;34:314-320.