Published Ahead of Print on January 20, 2015 as 10.1200/JCO.2014.57.

9029
The latest version is at http://jco.ascopubs.org/cgi/doi/10.1200/JCO.2014.57.9029

JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T

RTOG 9804: A Prospective Randomized Trial for

Good-Risk Ductal Carcinoma In Situ Comparing
Radiotherapy With Observation
Beryl McCormick, Kathryn Winter, Clifford Hudis, Henry Mark Kuerer, Eileen Rakovitch, Barbara L. Smith,
Nour Sneige, Jennifer Moughan, Amit Shah, Isabelle Germain, Alan C. Hartford, Afshin Rashtian,
Eleanor M. Walker, Albert Yuen, Eric A. Strom, Jeannette L. Wilcox, Laura A. Vallow, William Small Jr,
Anthony T. Pu, Kevin Kerlin, and Julia White
See accompanying article doi: 10.1200/JCO.2014.59.4259
Author affiliations appear at the end of
this article. A B S T R A C T

Published online ahead of print at Purpose

www.jco.org on January 20, 2015. The Radiation Therapy Oncology Group 9804 study identified good-risk patients with ductal
Written on behalf of the Radiation Ther- carcinoma in situ (DCIS), a breast cancer diagnosis found frequently in mammographically
apy Oncology Group.
detected cancers, to test the benefit of radiotherapy (RT) after breast-conserving surgery
Supported by Radiation Therapy Oncol- compared with observation.
ogy Group Grant No. U10 CA 21661
and Community Clinical Oncology Patients and Methods
Program Grant No. U10 CA 37422 from This prospective randomized trial (1998 to 2006) in women with mammographically detected low-
the National Cancer Institute. or intermediate-grade DCIS, measuring less than 2.5 cm with margins ⱖ 3 mm, compared RT with
Presented in part at the 48th Annual observation after surgery. The study was designed for 1,790 patients but was closed early
Meeting of the American Society of because of lower than projected accrual. Six hundred thirty-six patients from the United States and
Clinical Oncology, Chicago, IL, June Canada were entered; tamoxifen use (62%) was optional. Ipsilateral local failure (LF) was the
1-5, 2012, and the 54th Annual Meet- primary end point; LF and contralateral failure were estimated using cumulative incidence, and
ing of the American Society of Radia-
overall and disease-free survival were estimated using the Kaplan-Meier method.
tion Oncology, Boston, MA, October
28-31, 2012. Results
This article’s contents are the responsi- Median follow-up time was 7.17 years (range, 0.01 to 11.33 years). Two LFs occurred in the RT
bility of the authors and do not neces- arm, and 19 occurred in the observation arm. At 7 years, the LF rate was 0.9% (95% CI, 0.0% to
sarily represent the official views of the 2.2%) in the RT arm versus 6.7% (95% CI, 3.2% to 9.6%) in the observation arm (hazard ratio,
National Cancer Institute.
0.11; 95% CI, 0.03 to 0.47; P ⬍ .001). Grade 1 to 2 acute toxicities occurred in 30% and 76% of
Authors’ disclosures of potential patients in the observation and RT arms, respectively; grade 3 or 4 toxicities occurred in 4.0% and
conflicts of interest are found in the 4.2% of patients, respectively. Late RT toxicity was grade 1 in 30%, grade 2 in 4.6%, and grade
article online at www.jco.org. Author
3 in 0.7% of patients.
contributions are found at the end of
this article. Conclusion
Clinical trial information: NCT00003857. In this good-risk subset of patients with DCIS, with a median follow-up of 7 years, the LF rate was
Corresponding author: Beryl McCor-
low with observation but was decreased significantly with the addition of RT. Longer follow-up is
mick, MD, Memorial Sloan-Kettering planned because the timeline for LF in this setting seems protracted.
Cancer Center, 1275 York Ave, New
York, NY 10065; e-mail: mccormib@ J Clin Oncol 33. © 2015 by American Society of Clinical Oncology
mskcc.org.

© 2015 by American Society of Clinical Bowel Project (NSABP) and the Milan Cancer
Oncology INTRODUCTION
Institute trials for early invasive breast cancers,2,3
0732-183X/15/3399-1/$20.00
Ductal carcinoma in situ (DCIS) of the breast is stage four large prospective trials were designed to ad-
DOI: 10.1200/JCO.2014.57.9029 dress the effectiveness of BCS and RT for women
0 cancer, considered non–life threatening, and the
fourth most common cancer diagnosed in women, with DCIS compared with BCS alone.4-7 The
after invasive breast, lung, and colon cancers. In trials produced strikingly similar results; the
screening programs in the United States, DCIS addition of RT resulted in a reduction in the risk of
accounts for approximately 25% of all new breast local failure (LF) in the breast of ⱖ 50%. Of the
cancers.1 After the demonstration that breast- LFs, approximately half were DCIS and half
conserving surgery (BCS) and radiation (RT) were invasive breast cancer. However, survival
produced results equivalent to mastectomy by is excellent and seems to be independent of
both the National Surgical Adjuvant Breast and local treatment.

© 2015 by American Society of Clinical Oncology 1

Information downloaded from jco.ascopubs.org and provided by at Bibliotheque De L Universite Laval on January 22, 2015
Copyright © 2015 Americanfrom
Society of Clinical Oncology. All rights reserved.
132.203.227.62
Copyright 2015 by American Society of Clinical Oncology
 McCormick et al

More recently, an understanding of DCIS as not just one disease

PATIENTS AND METHODS
but a group of related subtypes of cancers has emerged, with a spec-
trum of LF risk. Is there a low-risk DCIS for which the benefit of RT This trial was conducted by the RTOG and reviewed and approved by the
would not be seen? Several groups addressed this intriguing issue by American College of Radiology Institutional Review Board (IRB) and local
assessing clinical and pathologic factors in their DCIS patient data- IRBs from participating sites. All patients gave written informed consent in
bases and modeling for recurrence; examples include the Van Nuys accordance with each center’s IRB guidelines.
Recurrence Score8 and the Memorial Sloan-Kettering Nomogram.9 The primary objective was to assess the role of RT versus observation
The Radiation Therapy Oncology Group (RTOG) 9804 phase III after BCS in decreasing or delaying the appearance of invasive cancer or DCIS
LF and preventing the need for mastectomy. Secondary end points included
trial was conceived to address the question of RT benefit in a
disease-free survival (DFS).
good-risk DCIS subset. The definition for good risk was derived Another objective was to demonstrate a working pathology classifi-
from the best available clinical information at the time and was cation system for DCIS. The study opened shortly after the Consensus
aimed at addressing the smaller size and lower grade lesions de- Conference on the Classification of Ductal Carcinoma In Situ in 1997.10
tected by screening mammography. Definitions for DCIS subtypes developed at that meeting were adopted

A Age
1. < 50
2. ≥ 50
R
S Final Path Margins
1. Negative (re-excision) a
t
2. 3-9 mm n
r 3. ≥ 10 mm Arm 1
d Observation with or without tamoxifen 20 mg per day for 5 years
a Mammographic/Pathologic
Size of Primary o Arm 2
t Radiation therapy* to the whole breast, with or without tamoxifen
1. ≤ 1 cm m
i 2. > 1 cm to ≤ 2.5 cm 20 mg per day for 5 years
i
f Nuclei Grade
1. Low z
y
2. Intermediate e
Tamoxifen Use
1. No
2. Yes

B Randomly allocated
(N = 636)

Observation with or without tamoxifen (n = 322) Radiation therapy to the whole breast with or without tamoxifen (n = 314)
Randomly allocated with follow-up information (n = 317) Randomly allocated with follow-up information (n = 312)

Excluded (n = 24) Excluded (n = 27)

Ineligible (n = 19) Ineligible (n = 24)
Missing DCIS measurement (n = 5) Missing DCIS measurement (n = 6)
Tumor dimension > 2.5 cm (n = 3) Unable to confirm eligibility (n = 4)
Unable to confirm eligibility (n = 2) Palpable lesion (n = 2)
Multicentric disease (n = 2) RT began more than 12 weeks post-surgery (n = 2)
No post-excision mammogram within 12 weeks of (n = 2) Tumor dimension > 2.5 cm (n = 2)
final surgery High-grade DCIS (n = 1)
Microcalcifications not measured on diagnostic (n = 1) Microcalcifications not measured on diagnostic (n = 1)
mammogram mammogram
Retrospective pathologic review reveals LCIS, not DCIS (n = 1) No post-excision mammogram within 12 weeks of (n = 1)
Tumor margin width < 3 mm (n = 1) final surgery
Unknown lesion measurements (n = 1) Patient was not taking tamoxifen at time of study entry (n = 1)
No follow-up and unable to confirm eligibility (n = 1) Retrospective pathologic review reveals LCIS, not DCIS (n = 1)
Withdrew consent (n = 5) Tumor margin width < 3 mm (n = 1)
No follow-up (n = 4) Unable to confirm prebiopsy mammographic findings (n = 1)
No follow-up and missing DCIS measurement (n = 1)
Withdrew consent (n = 3)
No follow-up (n = 1)

Allocated to protocol treatment (n = 298) Allocated to protocol treatment (n = 287)

Received protocol treatment (n = 295) Received protocol treatment (n = 284)
Did not receive protocol treatment due to patient refusal (n = 3) Did not receive protocol treatment due to patient refusal (n = 3)

Eligible and evaluable for efficacy end points (n = 298) Eligible and evaluable for efficacy end points (n = 287)

Fig 1. (A) Schema and (B) CONSORT patient flow diagram of the Radiation Therapy Oncology Group (RTOG) 9804 study. (*) Radiation therapy dose consisted of
one of the following: 1.8 Gy per fraction for 28 fractions, for a total dose of 50.4 Gy; 2.0 Gy per fraction for 25 fractions, for a total dose of 50.0 Gy; or 2.656 Gy per
fraction for 16 fractions, for a total dose of 42.5 Gy. DCIS, ductal carcinoma in situ; LCIS, lobular carcinoma in situ.

2 © 2015 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Information downloaded from jco.ascopubs.org and provided by at Bibliotheque De L Universite Laval on January 22, 2015
Copyright © 2015 Americanfrom
Society of Clinical Oncology. All rights reserved.
132.203.227.62
 Phase III Trial in Low-Risk DCIS: Radiotherapy v Observation

for this study and were used in developing College of American Pathol- Statistical Methods
ogy guidelines.11 All analyses were performed using SAS version 9.2 statistical software
The study developed a pathology teaching Web site to assist pathologists (SAS Institute, Cary, NC). Secondary end points included overall survival
in identifying low-risk patients, because funding for central pathology review (OS), defined as death from any cause; contralateral breast failure (CBF);
was not available. Another objective included a planned review by a breast distant failure (DF), defined as progression of disease beyond the treated breast
pathologist of 10% of all patients, recently reported separately.12 and regional nodes; and salvage mastectomy failure (MF), defined as removal
of the study breast. Death, LF, CBF, DF, and MF were considered failures for
DFS. Secondary end points were measured from the date of random assign-
Eligibility
ment to date of failure or last follow-up for censored patients.
Women with DCIS detected by mammogram or incidentally found in
Cumulative incidence methods17 were used to estimate the rates of LF,
tissue of an otherwise benign biopsy were eligible. The DCIS was unicentric,
CBF, DF, and MF, with death as the competing risk. OS and DFS rates were
low or intermediate nuclear grade, and less than 2.5 cm, on pathology or
estimated using the Kaplan-Meier method.18 Tests of the statistical signifi-
imaging, based on the work of Lagios et al.13 Minimal margin width was 3 mm
cance of the effect of RT on all end points were carried out using the log-rank
to ink. All patients had a negative postexcision mammogram, and women in
test.19 For LF, CBF, DF, and MF, Gray’s test comparing treatment arms is also
the RT arm began treatment within 12 weeks of final surgery.
shown.20 Univariable Cox proportional hazards models21 were used to test for
Patients were not eligible if they were less than age 26 years, had active
treatment differences (observation v RT) and are coded such that HR less than
connective tissue disorders, or were pregnant or lactating. The usual exclusions
of women with prior cancers were applied, except for those with basal or
squamous skin cancers or in situ carcinoma of the cervix.
Women taking hormone therapy at the time of study entry were ex- Table 1. Patient Demographics and Clinical Characteristics
cluded. Tamoxifen was allowed but only if started within 4 weeks from the
diagnosis of DCIS. Observation ⫾ Radiotherapy ⫾
Tamoxifen (n ⫽ 298) Tamoxifen (n ⫽ 287)
Demographic or Clinical
Characteristic No. of Patients % No. of Patients %
Treatment
Tamoxifen was required in both arms when the study opened, based on Age, years
information from the NSABP B24 study comparing RT with or without Median 58 58
tamoxifen.14 RTOG 9804 was amended in 2001 to make tamoxifen optional, Range 35-84 34-85
based on the conflicting data regarding tamoxifen in DCIS from the United Q1-Q3 51-68 51-67
Kingdom, Australia, and New Zealand trial results.7 Tamoxifen dose was 20 Largest dimension of DCIS
lesion on pathology
mg daily for 5 years and was added as a stratification factor based on intent to slide, cm
use in 2001. Median 0.5 0.5
For women randomly assigned to whole-breast RT, the dose was 50 Gy in Range 0.1-2.5 0.1-2.5
25 fractions or 50.4 Gy in 28 fractions. The study was amended in 2001 to Q1-Q3 0.3-0.8 0.3-0.8
include 42.5 Gy in 16 fractions; no boost was allowed in the study. Standard Race
whole-breast tangent fields were used, planned with either conventional or White 246 82.6 227 79.1
computed tomography simulation. Dose was prescribed to the lung– chest African American 21 7.0 22 7.7
wall interface at the isocenter, and maximum dose to the planning target Hispanic 12 4.0 17 5.9
volume allowed was 15%. Compensators, wedges, and dynamic therapy were American Indian 5 1.7 2 0.7
techniques used to achieve this constraint. All patients were observed every 3 Asian 13 4.4 14 4.9
months for the first year, then every 6 months for 2 years, and then yearly. Native Hawaiian or other
Mammograms were required yearly. Pacific Islander 1 0.3 4 1.4
Other 0 0.0 1 0.3
Age, years
Design
⬍ 50 61 20.5 54 18.8
The primary end point was LF, which was defined as biopsy-proven
ⱖ 50 237 79.5 233 81.2
invasive or DCIS recurrence measured from date of random assignment to
Final microscopic margins,
date of failure or last follow-up for censored patients. Patients were stratified by mm
age (⬍ v ⱖ 50 years), DCIS size (ⱕ v ⬎ 1 cm), and pathology margins (negative ⱖ 3-9 106 35.6 104 36.2
re-excision v 3 to 9 mm v ⱖ 10 mm). When tamoxifen was made optional, ⱖ 10 48 16.1 45 15.7
tamoxifen use (yes v no) and nuclear grade (low v intermediate) were added as Negative by negative
stratification factors. Patients were randomly assigned 1:1 to RT versus obser- re-excision 144 48.3 138 48.1
vation as described by Zelen15 (Fig 1A) using a permuted block random Mammographic size of
assignment scheme. The revised design, after the 2001 amendment, was based primary tumor, cm
on the assumption that two thirds of the patients would receive tamoxifen and ⱕ1 217 72.8 207 72.1
that low- and intermediate-grade distribution would be equally frequent. In ⬎1 81 27.2 80 27.9
addition, the LF rates for the control arm based on tamoxifen use and nuclear Nuclear grade
grade were assumed to be 3%, 5%, 7%, and 10% for yes/low, no/low, yes/ 1 131 44.0 121 42.2
intermediate, and no/intermediate, respectively, with an overall assumed 2 167 56.0 166 57.8
LF rate of 6% for the observation arm. The primary hypothesis was that the Tumor location
addition of RT would reduce the LF rate from 6% to 3.5% (hazard ratio Left breast 148 49.7 142 49.5
[HR], 0.58). Right breast 150 50.3 145 50.5
On the basis of these assumptions and the sequential design approach of Intention to use tamoxifen
Kim and Tsiatis16 with three treatment comparisons (two interim analyses and No 91 30.5 90 31.4
one final analysis) using two-sided log-rank test statistics, a maximum of 129 Yes 207 69.5 197 68.6
local recurrence events was required to detect the hypothesized reduction in
Abbreviations: DCIS, ductal carcinoma in situ; Q1, first quartile; Q3, third
the LF cumulative incidence with a statistical power of 80% and significance quartile.
level of P ⫽ .05. The target sample size was 1,790 patients.

www.jco.org © 2015 by American Society of Clinical Oncology 3

Information downloaded from jco.ascopubs.org and provided by at Bibliotheque De L Universite Laval on January 22, 2015
Copyright © 2015 Americanfrom
Society of Clinical Oncology. All rights reserved.
132.203.227.62
 McCormick et al

1 indicates a decreased risk of failure for the RT arm. All eligible patients are interim analyses were performed. The Data Monitoring Committee
included in the intent-to-treat analysis, based on the treatment arm to which recommended that all patients be observed for a minimum of 5 years
they were randomly assigned. The data were also analyzed with the inclusion of and then the study results be reported. No protocol-specified primary
patients found to be ineligible, because the most common reasons for ineligi-
end point interim analyses occurred before this reporting. All analyses
bility included missing DCIS size, inability to confirm all eligibility criteria, or
tumor reportedly larger than 2.5 cm, and patients who withdrew consent but are based on data through March 2012. Forty-three women were
had follow-up information before withdrawal (Fig 1B). ineligible on review, of whom 41 had follow-up information, and
Acute toxicities (ⱕ 90 days from treatment start) were scored using the eight women withdrew consent, of whom three had follow-up
National Cancer Institute Common Toxicity Criteria version 2.0. Late RT information. Reasons for ineligibility are shown in the CONSORT
toxicities (⬎ 90 days from treatment start) were scored using the RTOG/ diagram (Fig 1B).
European Organisation for Research and Treatment of Cancer Late Radiation Median age was 58 years. Table 1 lists the patient and tumor
Morbidity Scheme.
characteristics in the two randomized arms. Median follow-up for all
patients is 7.17 years (range, 0.01 to 11.33 years; 25th percentile, 5.92
RESULTS years; 75th percentile, 8.87 years), and median follow-up for alive
patients is 7.35 years (range, 0.03 to 11.33 years; 25th percentile, 5.99
Administrative Data years; 75th percentile, 8.93 years).
This study opened in December 1999, and as a result of not
meeting targeted accrual, it closed in July 2006, with a total of 636 Study End Points
patients of the 1,790 planned, entered from almost 200 institutions. At Cumulative rates of LF in the ipsilateral breast at 5 and 7 years
the time the trial closed, the required number of events for the first were 0.4% and 0.9% in the RT arm versus 3.5% and 6.7% in the
protocol-specified interim analysis of LF had not been reached, so no observation arm, respectively (Fig 2A; Table 2; log-rank and Gray’s

A 30 Failed Total B 30 Failed Total

Observation 19 298 Observation 21 317
RT 2 287 RT 2 312
25 25
Local Failure (%)

Local Failure (%)

20 20
Gray's test P < .001 Gray's test P < .001
HR = 0.11 (0.03 to 0.47) HR = 0.10 (0.02 to 0.41)
15 15

10 10
6.7% 7.2%

5 3.5% 5 4.0%
0.9% 0.8%
0.4% 0.3%

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7

Time Since Random Allocation (years) Time Since Random Allocation (years)
No. at risk No. at risk
Observation 298 287 272 257 240 225 182 126 Observation 317 302 285 269 251 232 187 128
RT 287 278 265 250 235 211 174 128 RT 312 301 285 269 252 225 183 135

C 100 D 100
Disease-Free Survival (%)
Overall Survival (%)

80 80

60 Failed Total 60 Failed Total

Observation 15 298 Observation 41 298
RT 22 287 RT 33 287
40 40

20 Log-rank P = .18 20 Log-rank P = .44

HR = 1.56 (0.81 to 3.01) HR = 1.56 0.84 (0.53 to 1.32)

0 1 2 3 3 5 6 7 0 1 2 3 3 5 6 7

Time Since Random Allocation (years) Time Since Random Allocation (years)
No. at risk No. at risk
Observation 298 295 292 287 273 259 214 161 Observation 298 291 281 274 262 249 202 147
RT 287 282 275 268 257 240 207 153 RT 287 280 270 261 250 231 197 145

Fig 2. (A) Local failure in ipsilateral breast for all eligible patients (n ⫽ 585). (B) Local failure in ipsilateral breast for all accrued patients with follow-up (n ⫽ 629). (C)
Disease-free survival (n ⫽ 585). (D) Overall survival (n ⫽ 585). HR, hazard ratio; RT, radiation therapy.

4 © 2015 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Information downloaded from jco.ascopubs.org and provided by at Bibliotheque De L Universite Laval on January 22, 2015
Copyright © 2015 Americanfrom
Society of Clinical Oncology. All rights reserved.
132.203.227.62
 Phase III Trial in Low-Risk DCIS: Radiotherapy v Observation

Table 2. Efficacy End Points

LF at 5 Years (%) LF at 7 Years (%)

No. of No. of Estimated Estimated Log-Rank/Gray’s Hazard

End Point and Arm Patients Failures Rate 95% CI Rate 95% CI Test P Ratio 95% CI
Local failure in treated breast
All eligible patients 585
Observation ⫾ tamoxifen 298 19 3.5 1.4 to 5.7 6.7 3.4 to 10.0 ⬍ .001 0.11 0.03 to 0.47
Radiotherapy ⫾ tamoxifen 287 2 0.4 0.0 to 1.1 0.9 0.0 to 2.2
All accrued patients with follow-upⴱ 629
Observation ⫾ tamoxifen 317 21 4.0 1.8 to 6.3 7.2 3.9 to 10.5 ⬍ .001 0.10 0.02 to 0.41
Radiotherapy ⫾ tamoxifen 312 2 0.3 0.0 to 1.0 0.8 0.0 to 2.0
Contralateral breast failure 585
Observation ⫾ tamoxifen 298 12 2.2 0.4 to 3.9 4.8 2.0 to 7.7 .86/.88 1.07 0.48 to 2.39
Radiotherapy ⫾ tamoxifen 287 12 3.4 1.2 to 5.6 3.9 1.5 to 6.4
Overall survival 585
Observation ⫾ tamoxifen 298 15 97.5 94.8 to 98.8 95.1 91.4 to 97.2 .18 1.56 0.81 to 3.01
Radiotherapy ⫾ tamoxifen 287 22 96.7 93.7 to 98.3 91.7 87.2 to 94.7
Disease-free survival 585
Observation ⫾ tamoxifen 298 41 92.7 89.1 to 95.2 85.6 80.4 to 89.4 .44 0.84 0.53 to 1.32
Radiotherapy ⫾ tamoxifen 287 33 93.4 89.7 to 95.8 88.0 83.1 to 91.6
ⴱ
Ineligible patients with follow-up and patients who withdrew consent but had follow-up information before withdrawal are included.

test, P ⬍ .001; HR, 0.11; 95% CI, 0.03 to 0.47). The histology of the LFs mographically detected good-risk DCIS. Other groups have also
was invasive in 42.1% and noninvasive in 57.9% of patients in the reported lower rates of in-breast recurrence after excision alone when
observation arm; in the RT arm, one patient each experienced invasive selecting for good-risk DCIS. One other multi-institutional prospec-
and noninvasive LF. Results were essentially the same when including tive study has reported similar results to RTOG 9804 in this specific
the ineligible patients with follow-up data (n ⫽ 41) or who withdrew population of DCIS. Hughes et al22 recently reported a 7-year LF rate
consent (n ⫽ 3); the LF rates at 5 and 7 years were 0.3% and 0.8% in of 10.5% in a low-risk DCIS cohort, with observation only, in women
the RT arm versus 4.0% and 7.2% in the observation arm, respectively after lumpectomy on the Eastern Cooperative Oncology Group
(Table 2; log-rank and Gray’s test, P ⬍ .001; HR, 0.10; 95% CI, 0.02 to (ECOG) 5194 DCIS single-arm prospective trial. The low- or
0.41; Fig 2B). LF rates by tamoxifen use and nuclear grade were 1.2%, intermediate-grade patients in this trial matched the eligibility for
5.3%, 2%, and 8.4% for yes/low, no/low, yes/intermediate, and no/ RTOG 9804.
intermediate, respectively. The risk of CBF at 7 years was 3.9% in the Another prospective study by the Dana-Farber group of obser-
RT arm and 4.8% in the observation arm (Table 2; log-rank and Gray’s vation for low-risk DCIS was recently updated. That study was
test, P ⫽ .86 and P ⫽ .88, respectively; HR, 1.07; 95% CI, 0.48 to 2.39). planned for 200 women, but it closed early because of LF events
Mastectomy rates were low; in the observation arm, eight women exceeding protocol stopping rules. Good-risk criteria for this study
underwent subsequent mastectomy, including four unilateral mastec- included mammographically detected DCIS, measuring ⱕ 2.5 cm,
tomies for an ipsilateral LF, three bilateral mastectomies for ipsilateral with a predominant nuclear grade of 1 or 2, and a margin of ⱖ 1 cm or
LF (n ⫽ 2) and bilateral LF (n ⫽ 1), and one elective bilateral mastec- a negative re-excision if re-excised. With a median follow-up time of
tomy. In the RT arm, four women had mastectomies, all bilateral. 11 years in 143 women, the cumulative incidence of LF at 10 years
Indications included ipsilateral failure (n ⫽ 1), contralateral failure was 15.5%. Of note, unlike our study and the ECOG study, the
(n ⫽ 1), bilateral failure (n ⫽ 1), and one elective surgery. The 7-year presence of high nuclear grade did not exclude patients, as long as
cumulative incidence of mastectomy was 2.8% (95% CI, 0.7% to
this was not the dominant grade. In the updated analysis, the
4.9%) in the observation arm and 1.5% (95% CI, 0% to 3.0%) in the
presence of any high nuclear grade was associated with a relative
RT arm. As expected, the OS (Fig 2C) and DFS (Fig 2D) rates were
risk ratio of 14.0 in their model.23
excellent and not different between the two arms (Table 2).
Our results, together with ECOG 5194, imply that clinical
Patients assigned to the RT arm had a higher rate of grade 1 and 2
pathologic criteria can be used to define a cohort of patients with
acute toxicities than patients in the observation arm (76% v 30%,
DCIS who can be expected to have a much lower rate of in-breast
respectively; P ⬍ .001). The rate of toxicities ⱖ grade 3 was 4% in both
recurrence without RT in the first 7 years after lumpectomy than
arms. Late RT toxicity was grade 1 in 30.0%, grade 2 in 4.6%, and grade
previously reported in past randomized trials. For comparison, the
3 in 0.7% of patients. There was no late grade 4 or 5 RT toxicity.
LF rates at 5 years in NSABP B17,4 in patients not selected for good
risk, were 20.9% with no RT and 9.6% with RT. For many
DISCUSSION women, these new data may support their decision for omission
of adjuvant RT after BCS, particularly because the rate of mas-
To our knowledge, this is the first prospective randomized trial com- tectomy in RTOG 9804 in both treatment arms was low. How-
paring RT with observation in a subset of women with defined, mam- ever, RTOG 9804 confirms that RT provides significant benefit

www.jco.org © 2015 by American Society of Clinical Oncology 5

Information downloaded from jco.ascopubs.org and provided by at Bibliotheque De L Universite Laval on January 22, 2015
Copyright © 2015 Americanfrom
Society of Clinical Oncology. All rights reserved.
132.203.227.62
 McCormick et al

in further reducing in-breast recurrence for women who opt to of the planned events; therefore, this study has significantly reduced
receive it. statistical power for the original hypothesis. Given that, although these
Despite the similarity of the patient populations in RTOG 9804 results are not definitive, in the original hypothesis, it was estimated
and the ECOG 5194 low-risk cohort, there was some difference in the that the addition of RT would reduce the risk of LF from 6% to 3.5%
LF rates reported at a similar median follow-up of 7 years (6.6% v (HR, 0.58). To date, the LF rate in the RT arm is less than 1%, yielding
10.5%, respectively). This may reflect different utilization rates of a statistically significant difference between the two arms as a result of
tamoxifen, which was taken by 31.3% of women accrued to ECOG the much larger than anticipated hazard reduction (HR, 0.11). Be-
5194 compared with 62% of women in the RTOG trial. The NSABP cause, to our knowledge, this is the first prospective randomized DCIS
B24 study of women with DCIS treated with BCS and then randomly study for good-risk patients, these data suggest the original estimate
assigned to RT versus RT and tamoxifen demonstrated the interaction based on prior studies that included high-risk DCIS was too high. The
between RT and tamoxifen in further lowering the risk of LF.14 Al- magnitude of proportional reduction in local recurrence in this study
though the NSABP B24 study did not have a treatment arm with from RT (HR, 0.10) is much larger than that seen in the four prior
tamoxifen and no RT, a phase III randomized controlled trial from the trials comparing RT with observation (HR, ⬃0.46) after lumpectomy
United Kingdom, Australia, and New Zealand in DCIS did contain for DCIS.25 At this time in follow-up, RTOG 9804 has few events and
such an arm. In a recent follow-up report, women assigned to tamox- wide CIs. Follow-up is continuing, and future analyses with longer
ifen only had a significant reduction in new breast events, and tamox- follow-up will determine the reliability of this proportional reduction.
ifen specifically was noted to reduce the risk of DCIS LF in the In conclusion, the RTOG 9804 trial in DCIS successfully identi-
ipsilateral breast, compared with the no tamoxifen/no RT arm.24 In fied a subset of women with good-risk DCIS based on standard pa-
RTOG 9804, there are so few events at this time that the data cannot thology features including nuclear grade, size, and margin width.
support or refute the role of tamoxifen in the treatment of low-risk Although the addition of RT significantly decreased the LF rate for the
DCIS. Data on tamoxifen use in study follow-up are being collected, patients accrued to this study, the full clinical implications of these
and with more events expected in the future, a meaningful analysis results will require further follow-up, given the historic patterns of LF
may be able to be performed at a later date. over 10 to 15 years from diagnosis of good-risk DCIS.
The rates of LF in RTOG 9804 at 5 and 7 years, even in the
observation arm, are low. But what is likely to happen with longer
follow-up time? The Early Breast Cancer Trialists’ Collaborative AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS
Group overview demonstrated that although the relative benefit of RT OF INTEREST
remained stable, the absolute benefit increased with time; at 5 years
after random assignment, the absolute reduction in risk was 10.5% Disclosures provided by the authors are available with this article at
(SE, 1.2%; LF rate: 7.6% with RT v 18.1% with observation), whereas www.jco.org.
at 10 years, it was 15.2% (SE, 1.6%; LF rate: 12.9% with RT v 28.1%
with observation).25 Solin et al26 retrospectively looked at 15-year
AUTHOR CONTRIBUTIONS
results for 1,003 women with DCIS, detected mammographically and
treated with BCS and RT, in a collaborative study of 10 institutions. At
Conception and design: Beryl McCormick, Kathryn Winter, Clifford
5 years, the LF rate for the high-grade lesions with comedonecrosis was
Hudis, Henry Mark Kuerer, Barbara L. Smith, Nour Sneige, Eric A.
12%, and for those without these features, the LF rate was 3%. How- Strom, Julia White
ever, at 10 years, the lower grade DCIS group essentially caught up Provision of study materials or patients: Eileen Rakovitch, Eric A.
with the higher risk group, with an LF rate of 15% in the low-grade Strom, Julia White
group compared with 18% in the high-grade group. These data sug- Collection and assembly of data: Kathryn Winter, Eileen Rakovitch,
gest that lower risk DCIS may have a low rate of LF in the first 5 years Amit Shah, Isabelle Germain, Alan C. Hartford, Afshin Rashtian, Eleanor
that continues to increase over time. The RTOG 9804 data will need M. Walker, Albert Yuen, Jeannette L. Wilcox, Laura A. Vallow, William
Small Jr, Anthony T. Pu, Kevin Kerlin
continued follow-up to fully realize the rates of LF in this cohort of Data analysis and interpretation: Beryl McCormick, Kathryn Winter,
low-risk patients. Jennifer Moughan, William Small Jr
It is recognized that this study closed early, with 636 patients Manuscript writing: All authors
accrued of the original planned 1,790 women and approximately 20% Final approval of manuscript: All authors

tectomy for early breast cancer. N Engl J Med ductal carcinoma in situ of the breast—Results of
REFERENCES 347:1227-1232, 2002 a randomized trial in a population offered mam-
4. Fisher B, Dignam J, Wolmark N, et al: mography screening. Acta Oncol 45:536-543,
1. Siegel R, NaishadhamD, Jemal A: Cancer Lumpectomy and radiation therapy for the treat- 2006
statistics 2013. CA Cancer J Clin 63:11-30, ment of intraductal breast cancer: Findings from the 7. Houghton J, George WD, Cuzick J, et al:
2013 National Surgical Adjuvant Breast and Bowel Project Radiotherapy and tamoxifen in women with com-
2. Fisher B, Anderson S, Bryant J, et al: B-17. J Clin Oncol 16:441-452, 1998 pletely excised ductal carcinoma in situ of the breast
Twenty-year follow-up of a randomized trial 5. Julien JP, Bijker N, Fentimen IS, et al: Radio- in the UK, Australia and New Zealand: Randomized
comparing total mastectomy, lumpectomy, and therapy in breast conserving treatment for ductal controlled trial. Lancet 362:95-102, 2003
lumpectomy plus irradiation for the treatment of carcinoma in situ: First results of the EORTC ran- 8. Silverstein MJ, Lagios MD, Groshen S, et al:
invasive breast cancer. N Engl J Med 347:1233- domized phase III trial 10853—EORTC Breast Can- The influence of margin width on local control of
1241, 2002 cer Cooperative Group and EORTC Radiotherapy ductal carcinoma in situ of the breast. N Engl J Med
3. Veronesi U, Cascinelli N, Mariani L, et al: Group. Lancet 355:528-533, 2000 340:1455-1461, 1999
Twenty-year follow-up of a randomized study com- 6. Emdin SO, Granstrand B, Ringberg A, et al: 9. Rudloff U, Jacks LM, Goldberg JI, et al: No-
paring breast-conserving surgery with radical mas- SweDCIS: Radiotherapy after sector resection for mogram for predicting the risk of local recurrence

6 © 2015 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Information downloaded from jco.ascopubs.org and provided by at Bibliotheque De L Universite Laval on January 22, 2015
Copyright © 2015 Americanfrom
Society of Clinical Oncology. All rights reserved.
132.203.227.62
 Phase III Trial in Low-Risk DCIS: Radiotherapy v Observation

after breast-conserving surgery for ductal carcinoma 15. Zelen M: The randomization and stratification of the breast: A trial of the Eastern Cooperative Oncology
in situ. J Clin Oncol 28:3762-3769, 2010 of patients to clinical trials. J Chronic Dis 27:365- Group. J Clin Oncol 27:5319-5324, 2009
10. The Consensus Conference Committee: Con- 375, 1974 23. Wong JS, Chen YH, Gadd MA, et al: Eight-
sensus Conference on the classification of ductal 16. Kim K, Tsiatis AA: Study duration for clinical year update of a prospective study of wide excision
carcinoma in situ. Cancer 80:1798-1802, 1997 trials with survival response and early stopping rule. alone for small low- or intermediate-grade ductal
11. Lester SC, Connolly JL, Amin MB: College of Biometrics 46:81-92, 1990 carcinoma in situ (DCIS). Breast Cancer Res Treat
American Pathologists protocol for the reporting of 17. Kalbfleisch J, Prentice R: The Statistical Anal- 143:343-350, 2014
ductal carcinoma in situ. Arch Pathol Lab Med ysis of Failure Time Data. New York, NY, John Wiley 24. Cuzick J, Sestak I, Pinder SE, et al: Effect of
133:13-14, 2009 & Sons, 1980, pp 167-169 tamoxifen and radiotherapy in women with locally
12. Woodward W, Sneige N, Winter K, et al: Web 18. Kaplan EL, Meier P: Nonparameteric estima- excised ductal carcinoma in situ: Long-term results
based pathology assessment in RTOG 98-04. J Clin tion from incomplete observations. J Am Statist from the UK/ANZ DCIS trial. Lancet Oncol 12:21-29,
Pathol 67:777-780, 2014 Assoc 53:457-481, 1958 2011
13. Lagios MD, Margolin FR, Westdahl PR, et al: 19. Mantel N: Evaluation of survival data and two 25. Early Breast Cancer Trialists’ Collaborative
Mammographically detected duct carcinoma in situ: new rank order statistics arising in its consideration. Group, Correa C, McGale P, et al: Overview of the
Frequency of local recurrence following tylectomy Cancer Chemother Rep 50:163-170, 1966 randomized trials of radiotherapy in ductal carci-
and prognostic effect of nuclear grade on local 20. Gray R: A class of K-sample tests for compar- noma in situ of the breast. J Natl Cancer Inst
recurrence. Cancer 63:618-624, 1989 ing the cumulative incidence of a competing risk. Monogr 2010:162-177, 2010
14. Fisher B, Dignam J, Wolmark N, et al: Tamox- Ann Stat 16:1141-1154, 1988 26. Solin LJ, Kurtz J, Fourquet A, et al: Fifteen-
ifen in treatment of intraductal breast cancer: Na- 21. Cox DR: Regression models and life tables. year results of breast-conserving surgery and defin-
tional Surgical Adjuvant Breast and Bowel Project J R Stat Soc B 34:187-220, 1972 itive breast irradiation for the treatment of ductal
B-24 randomised controlled trial. Lancet 353:1993- 22. Hughes LL, Wang M, Page DL, et al: Local exci- carcinoma in situ of the breast. J Clin Oncol 14:754-
2000, 1999 sion alone without irradiation for ductal carcinoma in situ 763, 1996

Affiliations
Beryl McCormick and Clifford Hudis, Memorial Sloan-Kettering Cancer Center, New York, NY; Kathryn Winter and Jennifer
Moughan, Radiation Therapy Oncology Group Statistical Center, Philadelphia; Amit Shah, York Cancer Center, York; Albert Yuen, Reading
Hospital and Medical Center, Reading, PA; Henry Mark Kuerer, Nour Sneige, and Eric A. Strom, The University of Texas MD Anderson
Cancer Center, Houston, TX; Barbara L. Smith, Massachusetts General Hospital, Boston, MA; Alan C. Hartford, Dartmouth Hitchcock
Medical Center, Lebanon, NH; Afshin Rashtian, University of Southern California, Los Angeles, Los Angeles; Anthony T. Pu, Radiological
Associates of Sacramento, Sacramento, CA; Eleanor M. Walker, Henry Ford Hospital, Detroit, MI; Jeannette L. Wilcox, Greenville SC
Community Clinical Oncology Program (CCOP), Greenville, SC; Laura A. Vallow, Mayo Clinic, Jacksonville, FL; William Small Jr,
Northwestern University Feinberg School of Medicine, Chicago, IL; Kevin Kerlin, Southeast Cancer Control Consortium CCOP, Winston-
Salem, NC; Julia White, James Cancer Hospital, Ohio State University, Columbus, OH; Eileen Rakovitch, Toronto-Sunnybrook Regional
Cancer Centre, Toronto, Ontario; and Isabelle Germain, L’Hotel-Dieu de Quebec, Quebec, Quebec, Canada.
■ ■ ■

www.jco.org © 2015 by American Society of Clinical Oncology 7

Information downloaded from jco.ascopubs.org and provided by at Bibliotheque De L Universite Laval on January 22, 2015
Copyright © 2015 Americanfrom
Society of Clinical Oncology. All rights reserved.
132.203.227.62
 McCormick et al

AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

RTOG 9804: A Prospective Randomized Trial for Good-Risk Ductal Carcinoma In Situ Comparing Radiotherapy With Observation
The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are
self-held unless noted. I ⫽ Immediate Family Member, Inst ⫽ My Institution. Relationships may not relate to the subject matter of this manuscript. For more
information about ASCO’s conflict of interest policy, please refer to www.asco.org/rwc or jco.ascopubs.org/site/ifc.
Beryl McCormick Afshin Rashtian
No relationship to disclose No relationship to disclose
Kathryn Winter Eleanor M. Walker
Consulting or Advisory Role: Roche No relationship to disclose
Clifford Hudis Albert Yuen
Other Relationship: Breast Cancer Research Foundation No relationship to disclose
Henry Mark Kuerer Eric A. Strom
Consulting or Advisory Role: Gerson Lehrman Group, Lightpoint No relationship to disclose
Medical
Research Funding: Genomic Health (Inst) Jeannette L. Wilcox
Patents, Royalties, Other Intellectual Property: McGraw-Hill No relationship to disclose
Publishing
Laura A. Vallow
Eileen Rakovitch No relationship to disclose
Research Funding: Genomic Health
William Small Jr
Barbara L. Smith Honoraria: Zeiss
No relationship to disclose Speakers’ Bureau: Zeiss
Nour Sneige Travel, Accommodations, Expenses: Zeiss
No relationship to disclose Anthony T. Pu
Jennifer Moughan Leadership: Radiological Associates of Sacramento
No relationship to disclose Stock or Other Ownership: Radiological Associates of Sacramento
Amit Shah Kevin Kerlin
No relationship to disclose Employment: 21st Century Oncology
Isabelle Germain Julia White
No relationship to disclose Honoraria: Genomic Health, Bayer Healthcare, Qfix, Varian
Alan C. Hartford
No relationship to disclose

© 2015 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Information downloaded from jco.ascopubs.org and provided by at Bibliotheque De L Universite Laval on January 22, 2015
Copyright © 2015 Americanfrom
Society of Clinical Oncology. All rights reserved.
132.203.227.62