A 3-Dimensional Mapping Analysis of Regional Nodal Recurrences in Breast Cancer

International Journal of

Radiation Oncology biology

physics

www.redjournal.org

Scientific Letter

A 3-Dimensional Mapping Analysis of Regional Nodal Recurrences in Breast Cancer Carl DeSelm, MD, PhD,* T. Jonathan Yang, MD, PhD,* Oren Cahlon, MD,* Jamie Tisnado, MD,y Atif Khan, MD,* Erin Gillespie, MD,* Simon Powell, MD, PhD,* and Alice Ho, MD* Departments of *Radiation Oncology and yRadiology, Memorial Sloan Kettering Cancer Center, New York, New York Received May 8, 2018. Accepted for publication Oct 17, 2018.

Summary We analyzed the anatomic pattern of nodal recurrence among a large number of patients with breast cancer. We provide evidence-based contouring guidelines and analysis of pathologic factors specifically associated with recurrence in the supraclavicular, axillary, and internal mammary regions.

Purpose: The primary goal was to map the anatomic pattern of isolated nodal recurrences (NR) in the supraclavicular (SCV), axillary, and internal mammary nodes (IMNs) in patients with breast cancer treated with curative-intent surgery with or without radiation therapy (RT). Secondary objectives were to assess clinical and pathologic factors associated with patterns of NR and survival rates. Methods and Materials: Patients with NR after treatment at a single cancer center during 1998 to 2013 were identified. Patients with prior distant metastases or NR without correlative imaging were excluded. All NRs were overlaid onto representative axial computed tomographic images. Multivariable analysis was performed to identify clinical and pathologic characteristics associated with NR. Kaplan-Meier curves were generated to assess the rate of relapse by nodal region according to pathologic feature or radiation treatment status. Results: The locations of 243 NRs among 153 eligible patients were mapped. The majority of NR occurred in the axilla (42%; 102/243), followed by the IMN (32.5%; 79/ 243) and the SCV (25.5%; 62/243). Radiation Therapy Oncology Group (RTOG) or European Society for Radiation therapy and Oncology (ESTRO) clinical target volume encompassed 82% (198/243) of NRs. The majority of out-of-field NRs were located in the lateral and posterior SCV region for both RTOG (67%; 30/45) and ESTRO (89%; 49/55) guidelines. The high-risk patients who received regional RT to the SCV relapsed at a similar rate in the medial, but a higher rate in lateral SCV (P Z .009), compared with low-risk patients who received no nodal RT. Lymphovascular invasion most strongly associated with IMN NR (P Z .001); grade 3 disease highly associated with both IMN (P Z .001) and SCV NR (P Z .02). The presence of an IMN NR portended for significantly inferior overall survival (OS),

Reprint requests to: Carl DeSelm, MD, PhD, Washington University, Department of Radiation Oncology, 1250 1st Ave, New York, NY 63110. Tel: (314) 273-2931; E-mail: [email protected] or deselmc@wustl. edu Int J Radiation Oncol Biol Phys, Vol. 103, No. 3, pp. 583e591, 2019 0360-3016/$ - see front matter Ó 2018 Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.ijrobp.2018.10.021

Conflict of interest: none. Supplementary material for this article can be found at https://doi.org/ 10.1016/j.ijrobp.2018.10.021.

584

International Journal of Radiation Oncology  Biology  Physics

DeSelm et al.

compared with an axillary NR, with a 5-year OS of 59% versus 72%, respectively (P Z .03). Conclusions: In this 3-dimensional image-based analysis of NR patterns in breast cancer patients treated with contemporary therapies, the lateral and posterior SCV represented a distinct site of NR that is not routinely included within current breast cancer contouring atlases. Grade 3 breast cancer and LVI were most commonly associated with the development of NR in the SCV. Modifying the CTV to encompass the lateral and posterior SCV in patients with breast cancer with these features might be justified. Ó 2018 Elsevier Inc. All rights reserved.

Introduction

Methods

Radiation therapy (RT) is an essential component in the management of breast cancer patients, enhancing both locoregional control and survival in multiple randomized trials and large meta-analyses.1 In these trials, the postmastectomy radiation portals included the internal mammary lymph nodes (IMNs), in addition to the chest wall, supraclavicular (SCV) and axillary lymph nodes. More recently, the benefits of regional nodal irradiation (RNI) after breast-conserving surgery was demonstrated by 2 large, randomized controlled trialsdNCIC MA20 and EORTC 22922din which patients who received RNI had a disease-free survival benefit compared with those who had whole breast RT alone.2,3 After the publication of these seminal trials, utilization of RNI and radiation target volumes have expanded in women with node-positive or high-risk node negative breast cancer, magnifying the need for accurate, CT-based delineation of the lymph nodes in radiation treatment planning. Attempts to standardize target delineations in breast cancer began in 2009, when the first consensus definition of the clinical target volume (CTV) was published by the Radiation Therapy Oncology Group (RTOG).4 Subsequently, contouring atlases were published by the European Society for Radiation therapy and Oncology (ESTRO).5-8 More recently, a Radiotherapy Comparative Effectiveness (RADCOMP) atlas was published online in 2016 to guide contouring efforts in an ongoing trial of proton versus photon therapy in breast cancer.9 Although these atlases were guided by expert opinion, large, robust data sets of locoregional patterns of nodal recurrence (NR) were not available at the time to inform the guidelines directly. The primary goal of our study was to conduct a 3-dimensional (3D) analysis of the locations and the patterns of nodal recurrence in patients with nonmetastatic breast cancer who were treated with curative intent surgery with or without adjuvant therapies. Second, we examined these patterns of failure relative to RTOG and ESTRO contouring guidelines. Finally, we explored whether certain clinical, pathologic, and treatment-related factors were associated with locationspecific NRs.

Study design This study was approved by the Memorial Sloan Kettering Cancer Center Institutional Review Board. From our multidisciplinary breast cancer database of consecutive 13,042 patients who received definitive surgery for nonmetastatic, invasive breast cancer between 1998 and 2013, women who had a regional NR as the first relapse and dedicated imaging (magnetic resonance imaging or computed tomography [CT] with or without positron emission tomography of the NR were identified. For patients who received radiation, 3D-conformal methods were used throughout the study period, and all nodal regions were contoured and assessed by dose-volume histogram when evaluating the treatment plan. Patients with initial stage IV breast cancer, history of noncutaneous malignancies (except melanoma), incomplete medical records, contralateral NR or chest wall or breast-only recurrence (without a component of NR) and evidence of distant metastases >1 month from the diagnosis of NR were excluded. In patients lacking a corresponding biopsy of the NR, radiographic features of malignancy such as those nodes >1 cm in short axis, extent of fluorodeoxyglucoseavidity (if positron emission tomography was used), infiltrative borders, and inhomogeneous enhancement were used as criteria to determine an NR.

Mapping of NR All cases of NR were mapped manually using CT imaging of a female patient who had undergone left-sided mastectomy and sentinel node biopsy without reconstruction. Her CT scan of the chest with intravenous contrast acquired with both arms abducted overhead and uploaded onto the Pinnacle planning system (Philips Health Care). The spatial locations of the NRs were transferred to the template CT. The epicenter of each NR was identified as the center of a 5-mm circle. Transferred recurrences were centrally reviewed by the study radiation oncologists (C.D. and A.H.) and radiologist (J.T.). A red circle represented NR cases in patients who received RT as part of primary treatment, and a yellow circle represented those who did not.

Volume 103  Number 3  2019

Nodal recurrences in breast cancer

585

Statistics

Patterns of nodal recurrences

The c2 or Fisher exact test was used to compare patterns of NR. Multivariate Cox proportional hazard regression models were used to correlate clinical and pathologic features to NR. Kaplan-Meier plots were performed for each patient cohort, and 2-year landmark analysis from time of surgery for the primary breast cancer was used to plot survival.

NRs were confirmed by pathology in 73% (112/153) of patients and clinical and radiographic findings for the remaining 26% (40/153) of patients; 44% of patients had isolated NR, and 56% of patients had sites of distant metastasis that were identified at the time of (but not before) the NRs. Lastly, 19% had a synchronous breast or chest wall failure and NR. Nodal regions were defined for the supraclavicular, IMN, and axillary nodal areas (Table E1; available online at https:// doi.org/10.1016/j.ijrobp.2018.10.021). The breakdown of NR was as follows: 42% (102/243) occurred in level 1 to 3 in the axilla, 32.5% (79/243) in the IMN, and 25.5% (62/243) in the SCV. Fig. 1 shows recurrence sites mapped on condensed axial CT slices of a representative patient. Fig. E1 demonstrates, slice-by-slice in axial view, all recurrences and recommended contours (available online at https://doi.org/10. 1016/j.ijrobp.2018.10.021). The majority (82%; 198/234) of NRs were encompassed by either the RTOG-CTV or ESTRO-CTV. Specifically, the RTOG-CTV group included 82% (198/234) and the ESTRO-CTV group included 77% (188/234) of the NRs. The SCV group harbored the bulk of out-of-field recurrences from both RTOG (missing 48% or 30/62 of SCV recurrences) and ESTRO (missing 79% or 49/ 62 of SCV recurrences) guidelines, followed by the IMN (2.5%; 2/79 for both) and axilla (13% or 13/102 for RTOG, and 1% for ESTRO). Given the high rate of out-of-field recurrences in the SCV, we formulated definitions for the medial, lateral, and

Results Patient and disease characteristics The study population included 153 patients with 243 NRs, whose key clinical characteristics at the time of diagnosis for their primary breast cancer are described in Table 1. Relative to the index population of 13,042 women with nonmetastatic breast cancer, the NR study population was younger (median age, 51 years; range, 23-79 years), and the majority had grade 3 disease (68% vs 25%; P Z .02). More than half (51.3%) of the NR population had lymphovascular invasion (LVI; 51% vs 31%; P Z .02). NR patients were less frequently ER positive (46% vs 75%; P Z .02). Extracapsular extension (ECE) was more common in the NR group (22% vs 14%; P Z .03). The majority (52%) of the NR study population presented with 0 positive lymph nodes, followed by 31.2% receiving 1 to 3 nodes and 17% receiving 4 positive nodes. Overall, 60% and 84% of the study population received some form of adjuvant RT or chemotherapy, or both.

Table 1

Characteristics of all patients in the registry, and those identified to have regional nodal recurrences as a site of first failure Patient Characteristics

All (N Z 13,042)

Regional NR (n Z 153)

Median age (range), y Laterality Tumor size, cm Grade 3 LVI ER positive HER2 positive Triple negative ECE No. of pathologically involved LNs 0 nodes 1 node 2 nodes 3 nodes 4 or more nodes No. of LNs sampled Surgery type Receipt of adjuvant breast or chest wall RT Receipt of chemotherapy

54 (20-95) 50% left, 50% right 1.7 46% 31% 75% 13% 18% 14% Mean Z 0.7 (range, 0-54); median Z 1 60% 17.3% 7.3% 3.6% 12% Mean Z 9 (range, 1-65); median Z 6 46% mastectomy, 54% BCS 65% (33% of mastectomy, 90% of lumpectomy) 68%

51 (23-79) 55% left, 45% right 3.6* 68%* 51%* 46%* 14% 29%* 22%* Mean Z 4.2* (range, 0-31); median Z 1 52%* 17.5% 8.7% 5% 17%* Mean Z 14 (range, 1-54); median Z 14 63% mastectomy, 37% BCS 60% (50% of mastectomy, 82% of lumpectomy) 84%*

Abbreviations: BCS Z Breast Conserving Therapy; ECE Z extracapsular extension; ER Z estrogen receptor; LN Z lymph node; LVI Z lymphovascular invasion. * P < .05.

586

DeSelm et al.

International Journal of Radiation Oncology  Biology  Physics

Fig. 1. Condensed axial views of nodal recurrences (NR), with suggested contours that maximally encompass the recurrences in each axial plane while minimally targeting surrounding tissues not involved in regional tumor spread. After radiation therapy, NRs are in red and NRs in patients without previous RT are in yellow. A full view of recurrences and recommended contours is shown in Fig. E1. posterior regions of the SCV. We defined the medial SCV as posterior to the SCM muscle, lateral SCV as the fat space lateral to the SCM muscle, and posterior SCV as the fat space posterior to the patient’s transverse process (midspinal canal; Fig. E1; available online at https://doi.org/10. 1016/j.ijrobp.2018.10.021). The RTOG SCV definition correlates to our medial SCV. Nearly half (48%; 28/58) of SCV NRs occurred in the medial SCV, followed by 40% (22/58) in the lateral SCV and 14% (8/58) in the posterior SCV. All 8 of the patients with posterior SCV NR had grade 3 disease; 38% of them (3/8) had LVI and only 12.5% of them (1/8) had ECE. The average size of the primary tumor size was 1.8 cm, and 60% of tumors (6/10) were node negative at the time of initial treatment. Despite not clearly having high-risk disease at baseline by current clinicopathologic factors, nearly all (7/8) patients with posterior

SCV NR not only developed synchronous recurrences in the axilla, but 75% (6/8) also developed synchronous distant metastases. We next examined the pattern of SCV failure among patients who received RT to the SCV versus those who did not, hypothesizing that if a region of the SCV were repeatedly undercovered using conventional techniques, the undercovered region might exhibit more failures over time. We identified 1103 patients who underwent RT to the regional nodes, which included the SCV in all of these patients, in addition to the IMNs in 202 (18%). Patients who received RNI were significantly higher risk at baseline, with a 7.7-cm versus 2.2-cm average primary tumor, a median of 4 versus 0 positive nodes, significantly more LVI, ECE, higher grade, and lower ER positivity (Table 2). Despite these findings, the rate of relapse in the medial

Volume 103  Number 3  2019 Table 2

Nodal recurrences in breast cancer

587

Characteristics of patients identified to have had RNI and those who had no RNI

Patient characteristics

RNI (n Z 1103)

No RNI (n Z 5979)

Age, y Laterality Tumor size, cm Grade 3 LVI ER positive HER2 positive Triple negative ECE No. of pathologically involved LNs 0 nodes 1 node 2 nodes 3 nodes 4 or more nodes Number of LNs sampled Surgery type Receipt of breast or chest wall RT Receipt of chemotherapy

51* 50% left, 50% right 7.6* 70%* 53%* 56%* 23% 27% 46%* Mean Z 5.8* (range, 0-42); median Z 4* 2%* 1%* 1%* 9%* 50%* Mean Z 20* (range, 0-66); median Z 18 71% mastectomy, 29% BCS 100%* 98%*

57 50% left, 50% right 2.2 49% 4.2% 69% 22% 24% 5% Mean Z 0.67 (range, 0-40); median Z 0 68% 19% 7% 2% 3% Mean Z 7 (range, 0-48); median Z 4 61% mastectomy, 39% BCS 29% 63%

Abbreviations: BCS Z breast conserving therapy; ECE Z extracapsular extension; ER Z estrogen receptor; HER2 Z human epidermal growth factor receptor 2; LN Z lymph node; LVI Z lymphovascular invasion; RNI Z regional nodal irradiation; RT Z radiation therapy. * P < .05 between RNI and No RNI groups.

SCV was not significantly different among the high-risk patients who received RNI. However, the rate of relapse in the lateral SCV was significantly higher in the patients who received RNI (P Z .009; Fig. 2). Only 2/58 SCV NRs occurred above the cricoid cartilage. Both NRs were at the level of the hyoid bone and posterior to the SCM in patients who were treated previously with adjuvant RT and were not encompassed within the RTOG or ESTRO CTV groups. Among the 79 NRs in the IMN chain, 63% (50/79), 18% (14/79), and 13% (10/79) were located within the first, second, and third intercostal spaces, respectively. In addition, 6% (5/79) of IMN NRs were located between the

Medial SCV

superior border of the clavicle and first rib, whereas none occurred above the superior border of the clavicle or below the third intercostal space. There were 2 IMN NRs abutting the sternum >5 mm medial to the internal mammary vessels (outside the ESTRO guidelines), suggesting that the majority of IMNs were included within CTV coverage that extended all the way to the sternal border. The axilla represented the most common site of NR (102/243). Few were outside of RTOG (14/102) or ESTRO (5/102) guidelines. Among those not encompassed by RTOG, most (7/14) occurred in the level 1 region before the axillary vessels formally cross the lateral border of pectoralis minor or within the interpectoral region (5/14),

Lateral SCV

Post RNI No RNI

Posterior SCV

1

Nodal RFS

0.8

ns

0.6

P = .009

ns

0.4 0.2 0 0

50

100

150

200

0

50

100

150

200

0

50

100

150

200

Time (months)

Fig. 2. Nodal relapse-free survival (RFS) among high-risk patients who underwent regional nodal irradiation (RNI; red ) and low-risk patients who did not receive RNI (blue). Log rank (Mantel-Cox) analysis was performed to assess for significant nodal RFS differences between curves (medial SCV Z 0.10; lateral SCV Z 0.009).

588

International Journal of Radiation Oncology  Biology  Physics

DeSelm et al.

(P Z .001 and P Z .02, respectively). ER negativity was independently associated with NR in all regions (axilla P Z .015; IMN P Z .001; SCV P Z .023).

Table 3 Multivariable analysis of factors associated with recurrence by site Axilla

SCV

IMN

P P P OR value OR value OR value

Factor LVI ER positive Nuclear grade 3 vs 1-2 Age Number of (þ) LNs

1.9 0.5 1.8 1 0.97

.026 2.1 .015 0.4 .097 3.6 .785 1 .311 0.98

Recurrence and survival outcomes

.033 2.5 .001 .023 0.3 .001 .008 4.4 .004 .778 0.98 .063 .397 1 .95

To assess the time-dependent association of clinicopathologic features with specific regions of NRs, we performed Kaplan-Meier analysis of time to failure among those who experienced NR, broken down by cohorts that were most significant on MVA. Relative to ER-positive patients, ER patients recurred more quickly in the SCV, axillary, and IMN regions. The 5-year IMN RFS is 92% for ER positive versus 70% for ER negative, but the 12-year IMN RFS was 69% for both ER positive and ER negative. In contrast, patients with grade 3 versus grade 1 or 2 disease (n Z 94) had a statistically lower rate of RFS in the SCV, axillary, and IMN regions across all time points (Fig. 3). A 2-year landmark analysis from the time of surgery demonstrated that patients with NR (n Z 153) had inferior OS compared with those without NR (n Z 8769), with a 5year OS of 74% versus 92% (P < .001, 3 years from the 2year landmark date; Fig. 4A). When examined by the location of NR, the presence of IMN NR (n Z 55) was associated with significantly inferior OS compared with axillary NR (n Z 83) or SCV NR (n Z 62), with a 5-year OS of 66% versus 80% (P Z .03; Fig. 4B), likely because

Abbreviations: ER Z estrogen receptor; IMN Z internal mammary node; LN Z lymph node; LVI Z lymphovascular invasion; OR Z odds ratio; SCV Z supraclavicular.

medial to the axillary vessels. One recurrence under the latissimus dorsi muscle and one over the latissimus dorsi were not encompassed by either guideline.

Predictors of nodal recurrence Multivariable analysis of the effect of ER status, age, LVI, total number of positive nodes, and grade on NR-free survival in each region of NR was performed (Table 3). LVI was most strongly associated with IMN NR (P Z .001) and grade 3 was associated with both IMN and SCV NR Axilla

SCV

80

100

80

*

*

80

60

60

60

40

40

40

20

20

20

0

0 0

Nodal Recurrence Free Survival

IMN

100

100

50

100

150

200

0 0

50

100

150

0

200

100

100

100

80

80

80

60

60

60

40

40

40

20

20

20

0

0

0 0

50

100

150

200

0

100

100

80

80

50

100

150

*

40

40

20

20

20

0

0 100

150

200

50

80

40

50

100

150

200

ECE No ECE

100

150

200

100

60

0

50

*

0

200

60

*

60

ER + ER -

*

G1 G2 G3

*

0 0

50

100 150

200

0

50

100

150

200

Months to Recur

Fig. 3. Time to nodal recurrence by region, grade, extracapsular extension (ECE), and estrogen receptor (ER) status. Abbreviations: ECE Z extracapsular extension; ER Z estrogen receptor; NR Z nodal recurrence. *P < .05.

Volume 103  Number 3  2019

Nodal recurrences in breast cancer

No lsolated Regional Recurrence lsolated regional Recurrence

Percent survival

100

50 P < .001

100

Axilla SCV IMN

50 P = .03

0

0 0

50

100

150

200

Months From 2 Year Landmark Number at risk: No NR: NR:

C

Survival by Site of Recurrence

B Percent survival

Overall Survival

A

589

3423 82

874 21

28 2

0

50

100

150

200

Months From 2 Year Landmark Axilla: SCV: IMN:

49 25 25

13 9 5

P = .02

0.8

ns

0.6 0.4 0.2 0.0 V IM N

SC

Ax

ill

a

Concurrent DM Rate

P = .002

1.0

Fig. 4. Overall survival by recurrence site. (A) Two-year landmark analysis from time of surgery, showing inferior overall survival among patients with nodal recurrence (NR; n Z 158) versus all patients (N Z 13,062). (B) Overall survival is inferior among patients with internal mammary node (IMN) NR, compared to patients with axillary NR and supraclavicular (SCV) NR. (C) Patients with IMN NR present with higher rates of concurrent distant metastasis (DM) relative to axillary NR and SCV NR groups. of the increased presence of concurrent distant metastasis (DM) in patients with IMN recurrence compared with either axillary (P Z .002) or SCV (P Z .02; Fig. 4C).

Discussion Randomized trials have demonstrated survival benefits of RNI in women with node positive breast cancer, justifying its selective use in this population. Data from the National Cancer Database demonstrated an increase in PMRT among women with 1 to 3 Nþ breast cancer from 19.1% in 2003 to 30.3% in 2012.10 Coincidently, the use of IMRT and proton therapy for breast cancer has increased, emphasizing the need for accurate delineation of nodal regions. These highly conformal modalities result in fall-offs in the high dose region and little or no dose to regions that historically received therapeutic levels of dose with 2-dimensional and 3D conformal techniques. Such information is best obtained by detailed analyses of NR in patients treated for primary breast cancer, in whom correlative imaging is explicitly available at the time of the NR. The primary objective of our study was to perform a mapping analysis of 243 posttreatment NRs from 153 patients whose treatment spanned over a decade. Secondarily, we validated ESTRO and ASTRO contouring guidelines for CTV delineation in breast cancer patients. Given the large sample size and NR events in our study, we also examined the association of clinical and pathologic variables on patterns of NR.

Risk-tailored axillary treatment has long been a topic of significance in breast cancer. In the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-04 trial, axillary RT was as effective as axillary lymph node dissection in breast cancer patients with clinically negative nodes, none of whom received systemic therapy.11 Decades later, the AMAROS trial showed similar results in contemporary patients treated with sentinel lymph node biopsy (SLNB) and systemic therapy, achieving lower rates of lymphedema with axillary RT.12 The axilla represented the most common site of NR (42%) in our study, reflecting predominant patterns of regional failure observed in large clinical trials of early-stage breast cancer treated with either SLNB or axillary dissections. We found that current contouring guidelines encompassed the majority of axillary NRs in our study. However, it is possible women with clinically positive nodes will require different CTVs from those in whom RT is used electively to prevent recurrence.13 Accurate delineation of the SCV is also essential in women receiving RNI, with the underlying premise that isolated SCV recurrences, whereas infrequent occurrences are preventable events. In our study, approximately half (52%) of SCV NRs occurred within the lateral or posterior SCV outside of either the RTOG or ESTRO guidelines, consistent with results of a large, Korean mapping study of 162 NRs among 234 patients in which the level 4 (medial SCV) lymph node region represented 33.8% of all regional recurrences, whereas 20% of recurrences were located

590

DeSelm et al.

posterolateral to the anterior scalene muscles.14,15 Among the 20% of NRs that occurred in the lateral or posterior SCV, 11 of 16 had multiple other NRs. In another series from MD Anderson Cancer Center, 21% of SCV NRs were located in the posterior or lateral SCV, although a caveat of this study was that most patients had metastatic disease on presentation.16 A report from China described the patterns of SCV nodal spread in 55 breast cancer patients, 17 of whom exhibited SCV involvement at initial presentation and 38 did so upon relapse.17 More than 80% of patients had SCV NR specifically in the posteriorelateral portion of the SCV. Finally, in a Mayo Clinic study in which more than half of the study patients exhibited metastases, 39% of patients had NR outside of RTOG volumes, nearly all of which were within the posterior or lateral SCV.18 Our study results and those from the aforementioned studies must be interpreted in context of fundamental limitations. First, the cohorts were identified retrospectively and therefore subject to sampling bias. Inconsistencies in the follow-up and detection further lent themselves to inaccuracies in reporting of NRs. Manual registration of NRs can compromise reproducibility given interpatient differences in anatomy, but potentially can be overcome with deformable registration in future studies. Patients with recurrent or metastatic tumor could manifest disease differently than those with micrometastases at presentation.13,17 It is striking that 60% of our study cohort initially exhibited node negative disease at presentation, suggesting factors in addition to nodal status are important in determining disease recurrence risk. We attempted to identify which clinical and pathologic characteristics are associated with NRs by region, as such findings would enable personalization of contouring regional nodes for breast cancer. We found that ER-negative status, LVI, and grade 3 disease independently predicted SCV and IMN recurrence. At the time of NR, but not upon initial presentation, positive lymph nodes in adjacent axillary levels were also associated with posterior SCV NR. Ultimately, no clinicopathologic features were identified that specifically differentiated the patient who would recur in the posterior SCV, relative to elsewhere in the SCV. Nevertheless, we believe it is reasonable to include the entirety of the SCV within the CTV in patients receiving RNI, including the clavicle and surrounding fat spaces but respecting muscular boundaries. Our recommendations differ from those specified by Jing et al.,17 who did not include the clavicle within the CTV. We observed a large proportion of recurrences within millimeters of the clavicle, and because clavicle position is highly sensitive to arm movement and thus potentially more prone to setup error than other targeted regions, the risk for underdosing the area directly surrounding the clavicle might be greater than the benefit of sparing this noneweight-bearing bone. Furthermore, we acknowledge there are overlap in the definition of the SCV between various atlases. For example, our definition of the “lateral SCV” was included within the level 2/3 axillary nodal contours in the RADCOMP atlas. Subtleties in differences between

International Journal of Radiation Oncology  Biology  Physics

definitions can be highlighted by the technique used. For example, whereas the RADCOMP definitions extended inferiorly to the acromial-clavicular joint, we recommend avoidance of dose to this structure, because no recurrences occurred there, to minimize the possibility of range-ofmotion or shoulder dysfunction that might occur after radiation.19 Conventional 4-field arrangements including the SCV and axilla extend inferiorly to the acromial-clavicular joint and medially to the pedicle of the vertebral bodies.20 The regions that we refer to as the lateral and posterior SCV are variably covered with conventional SCV fields and are dependent on where the lateral-most edge of the SCV field is placed and the depth to which the SCV prescription dose is prescribed, respectively. Depending on whether 1 or 2 SCV fields are used, our posterior SCV is expected to be covered by the 70% to 90% isodose line during conventional treatment.21 When we analyzed the pattern of SCV recurrence among patients who received RNI versus those who did not, we found that the pattern and rate of relapse was not significantly different among high-risk patients who received RNI versus low-risk patients who did not, except in the area of the lateral SCV. Although SCV RT effectively brought the rate of relapse among high-risk patients down to the rate of low-risk patients in the medial SCV, the lateral SCV did not exhibit the same benefit, and there was a significantly higher relapse rate among patients who received RNI compared with those who did not. Because both regions were equally powered to detect a difference (22 recurrences in the lateral vs 28 in the medial SCV), this further supports the need for improved coverage of the lateral SCV. The posterior SCV had too few events for statistical power, but it trended toward higher relapses among the high-risk patients who received RNI. Although the concept of contour-driven nodal definitions presents the opportunity to optimize coverage of target structures, it must be balanced against the concern for potential toxicities such as brachial plexopathy, lymphedema, shoulder stiffness, and esophagitis. Although several NRs occurred in the space between the internal jugular vein and carotid artery, no NRs were observed medial to the carotid artery, suggesting that the medial border of the carotid artery can serve as the medial-most border of the SCV and therefore minimize undesired dose to the esophagus. The development of an NR and DM are often inextricably linked, making it difficult to discern which event comes first.16 The most compelling rationale for preventing regional nodal failures was provided by the NCIC MA20 and EORTC 22922 trials, in which RNI was associated with an absolute reduction in DM of 4%.2,3 Our study excluded patients with DM identified before NR, but 56% of patients still had DM detected concurrently (within 1 month of diagnosis of NR). Furthermore, among 30 patients with lateral or posterior SCV NR, 26 patients (86%) had sites of concurrent DM, suggesting that SCV NR in these locations might be markers of other regional nodal and distant

Volume 103  Number 3  2019

involvement. Larger-scale, independent data sets are obviously required to confirm this hypothesis. Patients with isolated IMN NR had inferior survival compared with SCV or axillary NR and were also more likely to develop DM. These survival outcomes should be interpreted in context of the potential selection bias introduced by the inclusion of patients with isolated NR or concurrent DM, or both. Consistent with other reports, the vast majority (94%) of IMN NR were located within the first 1 to 3 intercostal spaces, with no IMN NRs occurring caudal to the fourth rib.22 Unlike other studies that have examined patterns of IMN NRs, we did not examine the association between IMN NR events and the quadrant of the primary lesion,2,3,15,23 although we did observe IMN NRs abutting the sternum >5 mm medial to the internal mammary vessels (2.5%; 2/79), suggesting CTV coverage should extend all the way to the sternal border. In conclusion, in this large, 3D imageebased analysis of NR patterns in patients with breast cancer treated with contemporary therapies, the lateral and posterior SCVs represented a distinct site of NR not routinely included within current breast cancer contouring atlases. Grade 3 breast cancer and LVI were most commonly associated with the development of NR in these regions. Modifying the CTV to encompass the lateral and posterior SCV in patients with breast cancer patients with these features might be justified.

References 1. Early Breast Cancer Trialists’ Collaborative G, Darby S, McGale P, Correa C, 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. 2. Poortmans PM, Colette S, Kirkove C, et al. Internal mammary and medial supraclavicular irradiation in breast cancer. N Engl J Med 2015;373:317-327. 3. Whelan TJ, Olivotto IA, Parulekar WR, et al. Regional nodal irradiation in early-stage breast cancer. N Engl J Med 2015;373:307-316. 4. White J, Tai A, Arthur D, et al. RTOG breast cancer atlas; 2009. Available at: https://www.rtog.org/LinkClick.aspx?fileticketZvzJFh PaBipE%3d&tabidZ236. 5. Kirova YM, Castro Pena P, Dendale R, et al. Simplified rules for everyday delineation of lymph node areas for breast cancer radiotherapy. Br J Radiol 2010;83:683-686. 6. Nielsen MH, Berg M, Pedersen AN, et al. Delineation of target volumes and organs at risk in adjuvant radiotherapy of early breast cancer: National guidelines and contouring atlas by the danish breast cancer cooperative group. Acta Oncol 2013;52:703-710. 7. Offersen BV, Boersma LJ, Kirkove C, et al. Estro consensus guideline on target volume delineation for elective radiation therapy of early stage breast cancer. Radiother Oncol 2015;114:3-10.

Nodal recurrences in breast cancer

591

8. Verhoeven K, Weltens C, Remouchamps V, et al. Vessel based delineation guidelines for the elective lymph node regions in breast cancer radiation therapyeprocab guidelines. Radiother Oncol 2015; 114:11-16. 9. MacDonald SaC O. Breast contouring radcomp consortium I; 2016. Available at: https://www.rtog.org/LinkClick.aspx?fileticketZeVB 451KQ83M%3d&tabidZ429. 10. Ohri N, Sittig MP, Tsai CJ, et al. Trends and variations in postmastectomy radiation therapy for breast cancer in patients with 1 to 3 positive lymph nodes: A national cancer data base analysis. Cancer 2018;124:482-490. 11. Fisher B, Jeong JH, Anderson S, et al. Twenty-five-year follow-up of a randomized trial comparing radical mastectomy, total mastectomy, and total mastectomy followed by irradiation. N Engl J Med 2002;347: 567-575. 12. Donker M, van Tienhoven G, Straver ME, et al. Radiotherapy or surgery of the axilla after a positive sentinel node in breast cancer (eortc 10981-22023 amaros): A randomised, multicentre, open-label, phase 3 non-inferiority trial. Lancet Oncol 2014;15: 1303-1310. 13. White J. Defining target volumes in breast cancer radiation therapy for the future: Back to basics. Int J Radiat Oncol Biol Phys 2015;93:277280. 14. 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. 15. 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. 16. Reed VK, Cavalcanti JL, Strom EA, et al. Risk of subclinical micrometastatic disease in the supraclavicular nodal bed according to the anatomic distribution in patients with advanced breast cancer. Int J Radiat Oncol Biol Phys 2008;71:435-440. 17. Jing H, Wang SL, Li J, et al. Mapping patterns of ipsilateral supraclavicular nodal metastases in breast cancer: Rethinking the clinical target volume for high-risk patients. Int J Radiat Oncol Biol Phys 2015;93:268-276. 18. Brown LC, Diehn FE, Boughey JC, et al. Delineation of supraclavicular target volumes in breast cancer radiation therapy. Int J Radiat Oncol Biol Phys 2015;92:642-649. 19. Hidding JT, Beurskens CH, van der Wees PJ, et al. Treatment related impairments in arm and shoulder in patients with breast cancer: A systematic review. PLoS One 2014;9:e96748. 20. Halperin EC, Perez CA, Brady LW. Perez and Brady’s Principles and Practice of Radiation Oncology. Philadelphia, PA: Wolters Kluwer HealtheLippincott Williams and Wilkins; 2008. 21. Jabbari K, Azarmahd N, Babazade S, et al. Optimizing of the tangential technique and supraclavicular fields in 3 dimensional conformal radiation therapy for breast cancer. J Med Signals Sens 2013;3:107-116. 22. Cheon H, Kim HJ, Lee SW, et al. Internal mammary node adenopathy on breast MRI and PET/CT for initial staging in patients with operable breast cancer: Prevalence and associated factors. Breast Cancer Res Treat 2016;160:523-530. 23. Jethwa KR, Kahila MM, Hunt KN, et al. Delineation of internal mammary nodal target volumes in breast cancer radiation therapy. Int J Radiat Oncol Biol Phys 2017;97:762-769.