Articles

Hypofractionated breast radiotherapy for 1 week versus

3 weeks (FAST-Forward): 5-year efficacy and late normal
tissue effects results from a multicentre, non-inferiority,
randomised, phase 3 trial
Adrian Murray Brunt*, Joanne S Haviland*, Duncan A Wheatley, Mark A Sydenham, Abdulla Alhasso, David J Bloomfield, Charlie Chan,
Mark Churn, Susan Cleator, Charlotte E Coles, Andrew Goodman, Adrian Harnett, Penelope Hopwood, Anna M Kirby, Cliona C Kirwan,
Carolyn Morris, Zohal Nabi, Elinor Sawyer, Navita Somaiah, Liba Stones, Isabel Syndikus, Judith M Bliss†, John R Yarnold†, on behalf of the
FAST-Forward Trial Management Group

Summary
Background We aimed to identify a five-fraction schedule of adjuvant radiotherapy (radiation therapy) delivered in Lancet 2020; 395: 1613–26
1 week that is non-inferior in terms of local cancer control and is as safe as an international standard 15-fraction Published Online
regimen after primary surgery for early breast cancer. Here, we present 5-year results of the FAST-Forward trial. April 28, 2020
https://doi.org/10.1016/
S0140-6736(20)30932-6
Methods FAST-Forward is a multicentre, phase 3, randomised, non-inferiority trial done at 97 hospitals (47 radiotherapy
See Comment page 1588
centres and 50 referring hospitals) in the UK. Patients aged at least 18 years with invasive carcinoma of the breast
*Joint first authors
(pT1–3, pN0–1, M0) after breast conservation surgery or mastectomy were eligible. We randomly allocated patients to
†Joint senior authors
either 40 Gy in 15 fractions (over 3 weeks), 27 Gy in five fractions (over 1 week), or 26 Gy in five fractions (over 1 week)
University Hospitals of
to the whole breast or chest wall. Allocation was not masked because of the nature of the intervention. The primary
North Midlands and University
endpoint was ipsilateral breast tumour relapse; assuming a 2% 5-year incidence for 40 Gy, non-inferiority was of Keele, Stoke on Trent, UK
predefined as ≤1·6% excess for five-fraction schedules (critical hazard ratio [HR] of 1·81). Normal tissue effects were (Prof A Murray Brunt FRCR);
assessed by clinicians, patients, and from photographs. This trial is registered at isrctn.com, ISRCTN19906132. Clinical Trials and Statistics
Unit, The Institute of Cancer
Research, Sutton, London, UK
Findings Between Nov 24, 2011, and June 19, 2014, we recruited and obtained consent from 4096 patients from 97 UK (Prof A Murray Brunt,
centres, of whom 1361 were assigned to the 40 Gy schedule, 1367 to the 27 Gy schedule, and 1368 to the 26 Gy J S Haviland MSc,
schedule. At a median follow-up of 71·5 months (IQR 71·3 to 71·7), the primary endpoint event occurred in 79 patients M A Sydenham BSc,
Prof P Hopwood MD,
(31 in the 40 Gy group, 27 in the 27 Gy group, and 21 in the 26 Gy group); HRs versus 40 Gy in 15 fractions were 0·86 L Stones BA, Prof J M Bliss MSc);
(95% CI 0·51 to 1·44) for 27 Gy in five fractions and 0·67 (0·38 to 1·16) for 26 Gy in five fractions. 5-year incidence of Royal Cornwall Hospital,
ipsilateral breast tumour relapse after 40 Gy was 2·1% (1·4 to 3·1); estimated absolute differences versus 40 Gy in Treliske, Truro, UK
15 fractions were –0·3% (–1·0 to 0·9) for 27 Gy in five fractions (probability of incorrectly accepting an inferior five- (D A Wheatley FRCR); Beatson
West of Scotland Cancer
fraction schedule: p=0·0022 vs 40 Gy in 15 fractions) and –0·7% (–1·3 to 0·3) for 26 Gy in five fractions (p=0·00019 vs Centre, Glasgow, UK
40 Gy in 15 fractions). At 5 years, any moderate or marked clinician-assessed normal tissue effects in the breast or (A Alhasso FRCR); Brighton and
chest wall was reported for 98 of 986 (9·9%) 40 Gy patients, 155 (15·4%) of 1005 27 Gy patients, and 121 of 1020 (11·9%) Sussex University Hospitals,
26 Gy patients. Across all clinician assessments from 1–5 years, odds ratios versus 40 Gy in 15 fractions were 1·55 Brighton, UK
(D J Bloomfield FRCR); Nuffield
(95% CI 1·32 to 1·83, p<0·0001) for 27 Gy in five fractions and 1·12 (0·94 to 1·34, p=0·20) for 26 Gy in five fractions. Health Cheltenham Hospital,
Patient and photographic assessments showed higher normal tissue effect risk for 27 Gy versus 40 Gy but not for Cheltenham, UK (C Chan FRCS);
26 Gy versus 40 Gy. Worcestershire Acute
Hospitals NHS Trust,
Worcester, UK (M Churn FRCR);
Interpretation 26 Gy in five fractions over 1 week is non-inferior to the standard of 40 Gy in 15 fractions over 3 weeks Imperial Healthcare NHS Trust,
for local tumour control, and is as safe in terms of normal tissue effects up to 5 years for patients prescribed adjuvant London, UK (S Cleator FRCR);
local radiotherapy after primary surgery for early-stage breast cancer. University of Cambridge,
Cambridge, UK
(Prof C E Coles FRCR); Royal
Funding National Institute for Health Research Health Technology Assessment Programme. Devon and Exeter NHS
Foundation Trust, Exeter, UK
Copyright © 2020 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. (A Goodman FRCR); Torbay
Hospital NHS Foundation
Trust, Torquay, UK
Introduction decades, schedules of adjuvant radiotherapy for these (A Goodman); Norfolk and
The Early Breast Cancer Trialists’ Collaborative Group patients delivered 25 fractions of 2 Gy over 5 weeks. Norwich University Hospital,
systematic overview confirms that radiotherapy after Randomised controlled trials with long-term follow-up Norwich, UK (A Harnett FRCR);
The Royal Marsden NHS
primary surgery in women with early-stage cancers have since confirmed that fewer, larger fractions giving a
Foundation Trust and The
reduces locoregional cancer recurrence and breast cancer lower total dose are at least as safe and effective as the Institute of Cancer Research,
deaths, including patients with positive lymph nodes previously used international standard.3–10 Specifically, London, UK (A M Kirby MD,
treated by mastectomy and axillary clearance.1,2 For many mature data confirm the safety and non-inferiority of 15 or N Somaiah FRCR,

www.thelancet.com Vol 395 May 23, 2020 1613

 Articles

Prof J R Yarnold FRCR);

University of Manchester, Research in context
Manchester, UK
(Prof C C Kirwan FRCS); Evidence before this study simplifying curative radiotherapy for women with early breast
Independent Cancer Patients’ We searched PubMed on April 22, 2020, using the search terms cancer.
Voice, London, UK (C Morris); “breast cancer”, “adjuvant radiotherapy”, “hypofractionation”,
Mount Vernon Cancer Centre, Added value of this study
and “randomised clinical trials”. We searched for primary
Northwood, UK (Z Nabi BSc); 15 or 16 fractions over 3·0–3·2 weeks are unlikely to represent
King’s College London, London, research and reviews published in English between Jan 1, 1980,
the limits of this approach, called hypofractionation.
UK (Prof E Sawyer FRCR); and and April 22, 2020. We found 13 randomised studies testing
Clatterbridge Cancer Centre, The FAST-Forward trial shows that 26 Gy in five fractions of
adjuvant breast hypofractionated radiotherapy regimens
Bebington, Wirral, UK 5·2 Gy to the conserved breast or post-mastectomy chest wall
(I Syndikus FRCR)
against standard fractionation ranging in sample size from 30 to
after primary surgery is non-inferior in terms of 5-year
2236 patients. All offered consistent support for the
Correspondence to: ipsilateral local tumour relapse to 40 Gy in 15 fractions over
Prof A Murray Brunt, c/o Clinical experimental approach.
3 weeks within an absolute 1·6% non-inferiority margin
Trials and Statistics Unit,
Radiotherapy (radiation therapy) after primary surgery for early compared with 2% incidence with 40 Gy. The 5-day schedule
The Institute of Cancer Research,
Sutton, London SM2 5NG, UK breast cancer has historically been delivered in five daily doses causes milder early skin reaction and similar rates of late
[email protected] (fractions) of 1·8–2·0 Gy per week over at least 5 weeks, adverse effects. When mature, a randomised FAST-Forward
but randomised phase 3 clinical trials done in Canada, the UK, substudy will report the safety of the five-fraction regimen for
and subsequently, China and Denmark have confirmed the patients prescribed radiotherapy to breast or chest wall
safety and efficacy of 15-fraction or 16-fraction schedules using combined with axilla or supraclavicular fossa.
daily fractions of 2·7 Gy. Four of these trials published 10-year
follow-up data on a total of 7000 patients, and 3-week Implications of all the available evidence
schedules have replaced traditional regimens in many countries FAST-Forward results confirm that 26 Gy in five fractions is as
over the past decade for most, if not all, patients prescribed effective and safe as an international standard 15-fraction
local or locoregional radiation therapy after breast conservation regimen after primary surgery for early breast cancer.
surgery or mastectomy. Most recently, long-term outcome data The 1-week schedule has major benefits over the 3-week or
for a five-fraction schedule delivered by once-weekly 5-week regimens in terms of convenience and cost for patients
treatments has been reported, which suggests further scope for and for health services globally.

16 fractions of about 2·7 Gy to total doses of 40·0 Gy or schedules of adjuvant radiotherapy to the whole breast
42·5 Gy.5,8 A 3-week schedule of 15 fractions has been or chest wall delivered in 1 week compared with the UK
the UK standard of care for adjuvant locoregional radio­ standard 15-fraction 3-week schedule. Substudies
therapy for early breast cancer since 2009 and is now an included a published acute toxicity study,15 photo­­
international standard for adjuvant local radiotherapy.11,12 graphic assess­ments of late adverse effects, and patient-
There is no reason to assume that 15 fractions represent reported outcomes; not all centres participated in the
the lower limits of this hypofractionated and accelerated substudies. Following recruitment into the main trial a
approach. We report outcomes of the FAST-Forward further substudy opened, testing the same fractionation
randomised phase 3 trial testing two dose levels of a five- schedules for patients requiring radiotherapy to the
fraction regimen delivered in 1 week against 40 Gy in axilla or supraclavicular fossa lymph nodes after sentinel
15 fractions over 3 weeks for patients prescribed local node biopsy or supraclavicular fossa only (levels 3–4)
radiotherapy after breast conservation surgery or after axillary dissection with a primary endpoint
mastectomy for early breast cancer. The objectives are to focusing on safety. Patients and results from this
identify a 1-week schedule non-inferior to a standard substudy are not reported here because follow-up is not
3-week regimen for 5-year local tumour control and yet mature. FAST-Forward was approved by the national
similar in terms of late adverse effects. FAST-Forward was South East Coast Kent research ethics committee
informed by the FAST trial that tested two dose levels of (11/LO/0958) and local research and development
five once-weekly fractions;13,14 FAST trial results to 10-year offices of all participating centres. The trial protocol is
See Online for appendix follow-up are to be published soon. The trial design used in the appendix (pp 15–78).
dose levels estimated to be the upper and lower bounds
that are isoeffective with the control schedule in terms of Patients
tumour control and normal tissue effects. Eligible patients were women or men aged at least 18 years
with invasive carcinoma of the breast (pT1–3, pN0–1, M0)
Methods following complete microscopic excision of the primary
Study design tumour by breast conservation surgery or mastectomy
FAST-Forward is a multicentre, non-blinded, phase 3, (reconstruction allowed), recruited in the UK from
randomised, non-inferiority trial, done at 97 hospitals 47 radiotherapy centres and 50 referral centres. A protocol
(47 radiotherapy centres and 50 referring hospitals) in amendment on Feb 15, 2013, excluded the lowest-risk
the UK, testing the safety and efficacy of five-fraction patients (aged ≥65 years, pT1, grade 1 or 2, oestrogen

1614 www.thelancet.com Vol 395 May 23, 2020

 Articles

receptor [ER] positive, HER2 negative, pN0, M0) to breast or chest wall clinical target volume accounting for
increase the overall primary event rate. All patients set-up error, breast swelling, and breathing to create a
had axillary surgery (sentinel node biopsy or axillary planning target volume (PTV). For all patients, a full
dissection); nodal radiotherapy was not allowed in the 3D CT set of outlines covering the whole breast and
main study. Concurrent endocrine therapy or trastu­ organs at risk was collected with a slice separation up to
zumab, or both, were permitted but not concurrent 5 mm, and organs at risk were outlined prospectively. A
chemotherapy. For the patient-reported outcomes sub­ tangential opposing pair beam arrangement encompassed
study all patients at participating centres were eligible. the whole breast or chest wall PTV, minimising the
All patients who had breast conservation surgery were ipsilateral lung and heart exposure. The treatment plan
eligible for the photographic substudy at participating was optimised with 3D dose compensation to achieve the
centres. A small number of patients who had had
mastectomy were recruited into the photographic
4110 patients enrolled and
substudy to validate the scoring method in patients who randomly assigned
had chest wall radiotherapy, but are not reported here
because photographs were only available for 76 patients.
All patients provided written informed consent.
1368 patients allocated to 1370 patients allocated to 1372 patients allocated to
40 Gy in 15 fractions 27 Gy in five fractions 26 Gy in five fractions
Randomisation and masking (3 weeks) (1 week) (1 week)
Patients were randomly assigned (1:1:1) to receive either
40 Gy in 15 fractions of 2·67 Gy; 27 Gy in five fractions of
5·4 Gy; or 26 Gy in five fractions of 5·2 Gy. A sequential 7 withdrew consent 3 withdrew consent 4 withdrew consent
for use of their data for use of their data for use of their data
tumour bed radiotherapy boost to the conserved breast
was allowed, with centres required to specify boost
intention and dose (10 Gy or 16 Gy in 2-Gy fractions) 1361 included in the 1367 included in the 1368 included in the
intention-to-treat intention-to-treat intention-to-treat
before randomisation. Randomisation was done by population population population
telephone or fax from the recruiting centre to the
Institute of Cancer Research-Clinical Trials and Statistics
Unit (ICR-CTSU), Sutton, London, UK, and used an in- 7 did not receive 12 did not receive 21 did not receive
allocated therapy allocated therapy allocated therapy
house bespoke trial-specific randomisation system set- 2 patient choice 2 ineligible 4 ineligible
up by the ICR-CTSU IT team. Computer-generated 2 treatment 4 patient choice 6 patient choice
random permuted blocks were used (block sizes 6 and prolonged 4 investigator 9 investigator
because of patient decision* decision†
9), stratified by radiotherapy centre and risk group (high illness 1 withdrawal of 2 treatment stopped
[age <50 years or grade 3] vs low [age ≥50 years and 3 treatment stopped consent early
early 1 incorrect dose
grade 1 or 2]). Treatment allocation was not masked to given
clinicians or patients.
Test dose levels were informed by START8 and FAST15
trials generating α/β values for late normal tissue effects. 1354 patients received 1355 patients received 1347 patients received
allocated therapy allocated therapy allocated therapy
Assuming an α/β value of 3 Gy and no effect of overall (per-protocol population) (per-protocol population) (per-protocol population)
time on outcomes, 27 Gy in five fractions of 5·4 Gy was
predicted to match late normal tissue effects of 40 Gy in
143 with no 5-year visit 136 with no 5-year visit 136 with no 5-year visit
15 fractions of 2·7 Gy or 46 Gy in 23 fractions of 2 Gy. form available form available form available
Allowance for a possible effect of treatment time informed 72 died 93 died 75 died
the choice of the slightly lower 26 Gy dose level. 15 withdrew 3 withdrew 4 withdrew
1 lost to 0 lost to 0 lost to follow-up
follow-up follow-up 57 forms not
Radiotherapy 55 forms not 40 forms not received
received received
The whole breast clinical target volume, including the
soft tissues from 5 mm below the skin surface to the deep
fascia, was either established from field-based tangential 1218 patients with 5-year visit 1231 patients with 5-year visit 1232 patients with 5-year visit
fields or the volume was contoured prospectively. Post- form available form available form available

mastectomy chest wall clinical target volume encom­

Figure 1: FAST-Forward trial profile
passed post-surgical skin flaps and underlying soft tissues
*One patient had no radiotherapy as they were unable to get into a stable position; three were given 40 Gy in
to the deep fascia; both excluded underlying muscle and 15 fractions (one because of concern for brachial plexus, one decided on a different treatment plan, and one
rib cage. Surgeons were strongly encouraged to mark the because of constraints of treatment planning). †One patient had no radiotherapy because they were diagnosed
tumour cavity walls with titanium clips or gold seeds at with pemphigoid and eight were given 40 Gy in 15 fractions (one because dose constraints were not met, one was
unable to plan within protocol constraints because of tumour bed position, one had poor planning target volume
the time of breast conservation surgery in order to aid
coverage, one had technical difficulties in planning, one was transferred to direct electron field, one had a simulator
placement of tangential fields and delineation of tumour plan because 3D images were not possible, one had a small pericardial effusion found at planning, and one gave no
bed. A typical margin of 10 mm was added around the reason).

www.thelancet.com Vol 395 May 23, 2020 1615

 Articles

40 Gy in 27 Gy in 26 Gy in following PTV dose distribution: more than 95% of PTV

15 fractions five fractions five fractions
(n=1361) (n=1367) (n=1368)
received 95% of prescribed dose, less than 5% of PTV
received 105% or more, less than 2% of PTV received
Age, years
107% or more, and a global maximum of less than 110%.
Median (IQR) 60 (53–66) 61 (53–67) 61 (52–66)
Dose constraints for the control group were as follows:
Range 29–89 25–90 25–89
volume of ipsilateral lung receiving 12 Gy less than 15%,
<40 12 (0·9%) 16 (1·2%) 28 (2·0%)
and volume of heart receiving 2 Gy less than 30% and
40–49 186 (13·7%) 173 (12·7%) 189 (13·8%)
that receiving 10 Gy less than 5%. Dose constraints for
50–59 440 (32·3%) 423 (30·9%) 414 (30·3%) the five-fraction schedules were as follows: volume of
60–69 506 (37·2%) 511 (37·4%) 524 (38·3%) ipsilateral lung receiving 8 Gy less than 15%, and volume
70–79 175 (12·9%) 197 (14·4%) 172 (12·6%) of heart receiving 1·5 Gy less than 30% and that receiving
≥80 42 (3·1%) 47 (3·4%) 41 (3·0%) 7 Gy less than 5%. X-ray beam energies for treatment
Sex were 6 MV or 10 MV, but a mixture of energies—eg, 6 MV
Female 1355 (99·6%) 1365 (99·9%) 1362 (99·6%) and 10–15 MV—was allowed for larger patients, assessed
Male 6 (0·4%) 2 (0·1%) 4 (0·3%) on a case-by-case basis. Tumour bed boost was delivered
Unknown 0 0 2 (0·1%) via electrons or photons. Verification was done using
Tumour grade electronic portal imaging using MV or kV x-rays. Control
1 315 (23·1%) 315 (23·0%) 300 (21·9%) group treatment verification was required for at least
2 660 (48·5%) 663 (48·5%) 690 (50·4%) three fractions in the first week with correction for any
3 386 (28·4%) 389 (28·5%) 378 (27·6%) systematic error and then once weekly with a tolerance of
Risk group 5 mm. The five-fraction schedules required verification
Low (age ≥50 and grade 1 or 2) 843 (61·9%) 854 (62·5%) 854 (62·4%) imaging for each fraction with recommendations to
High (age <50 or grade 3, or both) 518 (38·1%) 513 (37·5%) 514 (37·6%) correct all measured displacements. A comprehensive
Primary surgery quality assurance programme involved every radiotherapy
Breast conservation surgery 1270 (93·3%) 1278 (93·5%) 1284 (93·9%) centre before trial activation and continued throughout
Breast conservation surgery with 42 (3·1%) 33 (2·4%) 42 (3·1%) trial accrual; this was coordinated by the UK Radiotherapy
oncoplastic technique Trials Quality Assurance team based at Mount Vernon
Mastectomy 91 (6·7%) 89 (6·5%) 84 (6·1%) Hospital, Northwood, UK. The radiotherapy planning
Mastectomy with immediate 8 (0·6%) 11 (0·8%) 7 (0·5%) pack is in the appendix (pp 79–103).
reconstruction
Autologous reconstruction 5/8 (62·5%) 7/11 (63·6%) 3/7 (42·9%) Assessments
Implant-based reconstruction 2/8 (25·0%) 4/11 (27·3%) 4/7 (57·1%) Patients were assessed by clinicians for ipsilateral breast
Reconstruction type not specified 1/8 (12·5%) 0 0 tumour relapse and late normal tissue effects at annual
Side of primary tumour follow-up visits. Starting 12 months after trial entry,
Left 726 (53·3%) 674 (49·3%) 662 (48·4%) late-onset normal tissue effects in ipsilateral breast or
Right 635 (46·7%) 693 (50·7%) 704 (51·5%) chest wall (breast distortion, shrinkage, induration and
Unknown 0 0 2 (0·1%) telangiectasia; and breast or chest wall oedema and
Maximal extent of axillary staging discomfort) were graded by clinicians on a four-point
Sentinel node biopsy or guided 1157 (85·0%) 1184 (86·6%) 1164 (85·1%) scale (none, a little, quite a bit, or very much), interpreted
axillary sampling as none, mild, moderate, or marked. Symptomatic rib
Axillary clearance 200 (14·7%) 181 (13·2%) 201 (14·7%) fracture, symptomatic lung fibrosis, and ischaemic
Other 4 (0·3%) 2 (0·1%) 1 (0·1%) heart disease were recorded. Clinical assessments of
Unknown 0 0 2 (0·1%) acute skin toxicity have been previously reported.15
Pathological node status In the patient-reported outcomes substudy, ques­
Positive 257 (18·9%) 243 (17·8%) 256 (18·7%) tionnaires were administered at baseline (before ran­
Negative 1103 (81·0%) 1124 (82·2%) 1110 (81·1%) domisation) and at 3, 6, 12, 24, and 60 months, including
Unknown 1 (0·1%) 0 2 (0·1%) the European Organisation for Research and Treatment
Histological type of Cancer QLQ-BR23 breast cancer module, body image
Infiltrating ductal 1084 (79·6%) 1096 (80·2%) 1086 (79·4%) scale, and protocol-specific questions relating to changes
Lobular 144 (10·6%) 139 (10·2%) 127 (9·3%) to the affected breast after treatment (including breast
Mixed 51 (3·7%) 63 (4·6%) 65 (4·8%) appearance changed, smaller, harder or firmer, and skin
Other 82 (6·0%) 69 (5·0%) 87 (6·4%) appearance changed). Patient assess­ments used a four-
Unknown 0 0 3 (0·2%) point scale (not at all, a little, quite a bit, and very much).
Pathological tumour size, cm In the photographic substudy, photographs were taken
Median (IQR) 1·6 (1·1–2·2) 1·6 (1–2·2) 1·6 (1·1–2·4) at baseline and at 2 and 5 years after radiotherapy.
(Table 1 continues on next page) Change in photographic breast appearance compared
with baseline (after surgery and before radiotherapy) was

1616 www.thelancet.com Vol 395 May 23, 2020

 Articles

scored on a three-point scale (none, mild, or marked)

40 Gy in 27 Gy in 26 Gy in
based on changes in breast size and shape relative to the 15 fractions five fractions five fractions
contralateral breast. Patients were ineligible for further (n=1361) (n=1367) (n=1368)
photographic assessments after breast reconstruction (Continued from previous page)
surgery and further ipsilateral disease. Digital photo­
Pathological T stage
graphs were scored by three observers who were masked
T1mi 4 (0·3%) 5 (0·4%) 6 (0·4%)
to patient identity and treatment allocation, following
T1a 69 (5·1%) 68 (5·0%) 51 (3·7%)
scoring procedures established in the START trials.16
T1b 258 (19·0%) 270 (19·8%) 256 (18·7%)
Breast size and surgical deficit were assessed from the
T1c 612 (45·0%) 601 (44·0%) 602 (44·0%)
baseline photographs on a three-point scale (small,
T2 394 (28·9%) 389 (28·5%) 424 (31·0%)
medium, and large).
T3 21 (1·5%) 30 (2·2%) 25 (1·8%)
Unknown 3 (0·2%) 4 (0·3%) 4 (0·3%)
Outcomes
ER and HER2 status
The primary endpoint was ipsilateral breast tumour
ER positive HER2 positive 103 (7·6%) 103 (7·5%) 93 (6·8%)
relapse, defined as invasive carcinoma or ductal carci­
ER positive HER2 negative 1108 (81·4%) 1130 (82·7%) 1097 (80·2%)
noma in situ presenting anywhere in the ipsilateral
breast parenchyma or overlying skin or post-mastectomy ER negative HER2 positive 32 (2·4%) 34 (2·5%) 42 (3·1%)

chest wall, whether considered local recurrence or new ER negative HER2 negative 111 (8·2%) 96 (7·0%) 128 (9·4%)

primary tumour. Data on first regional relapse (axilla, Not known 7 (0·5%) 4 (0·3%) 8 (0·6%)
supraclavicular fossa, and internal mammary chain), Progesterone receptor status
distant metastases, new primary cancer, and death were Positive 577 (73·1%)† 541 (70·3%)† 566 (69·8%)†
collected. Key secondary endpoints were late normal Negative 212 (26·9%)† 229 (29·7%)† 245 (30·2%)†
tissue effects assessed by clinicians, patients, and from Not done 571 (42·0%) 596 (43·6%) 555 (40·6%)
photographs, and other disease-related and survival Missing on form 1 (0·1%) 1 (0·1%) 2 (0·1%)
outcomes (locoregional relapse, distant relapse, disease- Lymphovascular invasion
free survival, and overall survival; appendix p 29). Present 186 (13·7%) 178 (13·0%) 202 (14·8%)
Absent 1085 (79·7%) 1084 (79·3%) 1055 (77·1%)
Statistical analysis Uncertain 34 (2·5%) 40 (2·9%) 51 (3·7%)
The target sample size was 4000 patients (balanced Unknown 56 (4·1%) 65 (4·8%) 60 (4·4%)
allocation between groups). This provided 80% power Neoadjuvant chemotherapy received‡
(one-sided α of 0·025 allowing for non-inferiority Yes 48 (3·5%) 56 (4·1%) 43 (3·1%)
hypothesis and a simple Bonferroni correction taking into No 1312 (96·4%) 1311 (95·9%) 1323 (96·7%)
account comparisons between each test schedule and the Unknown 1 (0·1%) 0 2 (0·1%)
control group17) to exclude an absolute increase of 1·6% in Adjuvant therapy received: all patients
5-year ipsilateral breast tumour relapse incidence for a Chemotherapy§ 333/1360 (24·5%) 324/1367 (23·7%) 370/1366 (27·1%)
five-fraction schedule compared with control, assuming Adjuvant therapy received:‡ HER2-positive patients
2% 5-year incidence in the 40 Gy group (START data,7 and Chemotherapy and trastuzumab 84/135 (62·2%) 85/137 (62·0%) 100/135 (74·1%)
allowing for reduced ipsilateral breast tumour relapse due Trastuzumab, no chemotherapy 16/135 (11·9%) 13/137 (9·5%) 13/135 (9·6%)
to evolution of surgical techniques and systemic therapy). Chemotherapy, no trastuzumab 2/135 (1·5%) 2/137 (1·5%) 0
The 1·6% absolute non-inferiority margin was defined at No chemotherapy, no trastuzumab 33/135 (24·4%) 37/137 (27·0%) 22/135 (16·3%)
the trial design stage by the protocol development group, Adjuvant therapy received:‡ ER-positive patients
which included clinicians and patient advocates and was Endocrine therapy 1169/1216 (96·1%) 1186/1237 (95·9%) 1157/1196 (96·7%)
considered to be acceptable and appropriate. Binary Boost given
proportions were used for the sample size calculations Yes 342 (25·1%) 337 (24·7%) 332 (24·3%)
because event rates are so low. Estimates allowed for
No 1017 (74·7%) 1027 (75·1%) 1031 (75·4%)
10% loss to follow-up or unevaluable patients, expected to
Not known 2 (0·1%) 3 (0·2%) 5 (0·4%)
be largely due to development of metastatic disease.
Boost dose
2196 patients (732 per group) was estimated for the
10 Gy in five fractions 260/342 (76·0%) 273/337 (81·0%) 257/332 (77·4%)
photographic and patient-reported outcomes substudies
16 Gy in eight fractions 80/342 (23·4%) 64/337 (19·0%) 75/332 (22·6%)
to provide 80% power to detect an 8% difference in the
Unknown 2/342 (0·6%) 0 0
5-year prevalence of late normal tissue effects between the
five-fraction schedules (assuming 35% with 5-year mild Data are n (%) or n/N (%) unless otherwise stated. ER=oestrogen receptor. *14 patients withdrew consent for any of
their data to be used in analysis. †These percentages are calculated out of total patients who had available results for
or marked change in photographic breast appear­ ance this test. ‡Patients could have more than one type of adjuvant systemic therapy. §Chemotherapy type (for those
from START-B 40 Gy results7), allowing for 10% loss to specified): anthracyclines (n=584), taxane and anthracyclines (n=348), taxane and other—eg, docetaxel, carboplatin,
follow-up or unevaluable patients. and trastuzumab (n=83), and other (n=3).
Kaplan-Meier estimates (with 95% CIs) of 5-year Table 1: Demographic, clinical, and treatment characteristics at randomisation (n=4096)*
ipsilateral breast tumour relapse incidence were

www.thelancet.com Vol 395 May 23, 2020 1617

 Articles

relapse, with death from any cause as a competing event

Cumulative Estimated Hazard ratio (95% CI); Estimated absolute
number of cumulative p value difference vs 40 Gy in a Fine–Gray competing risks regression model.
events incidence by at 5 years (95% CI) Clinician and patient assessments of late normal tissue
5 years (95% CI) effects were analysed as follows: (1) 5-year cross-sectional
Ipsilateral breast tumour (local) relapse* analyses compared prevalence of moderate or marked
40 Gy (n=1361) 31 (2·3%) 2·1% (1·4 to 3·1) 1 (ref) ·· effects versus none or mild effects between groups using
27 Gy (n=1367) 27 (2·0%) 1·7% (1·2 to 2·6) 0·86 (0·51 to 1·44); –0·3% (–1·0 to 0·9) risk ratios and risk differences and Fisher’s exact test; and
0·56 (2) longitudinal analyses of moderate or marked effects
26 Gy (n=1368) 21 (1·5%) 1·4% (0·9 to 2·2) 0·67 (0·38 to 1·16); –0·7% (–1·3 to 0·3) (vs none or mild) using generalised estimating equations19
0·15
including all assessments, comparing groups across the
Locoregional relapse† whole follow-up period using odds ratios (ORs) and the
40 Gy (n=1361) 43 (3·2%) 2·8% (2·0 to 3·9) 1 (ref) ·· Wald test. Generalised estimating equations models
27 Gy (n=1367) 35 (2·6%) 2·3% (1·6 to 3·3) 0·80 (0·51 to 1·25); –0·5% (–1·4 to 0·7) included a term for years of follow-up, enabling time
0·33
trends to be modelled. Additionally, to enable comparison
26 Gy (n=1368) 29 (2·1%) 1·8% (1·2 to 2·7) 0·66 (0·41 to 1·06); –0·9% (–1·6 to 0·2)
0·083
of clinician-assessed normal tissue effects with results
Distant relapse
reported from other trials, survival analysis methods
40 Gy (n=1361) 59 (4·3%) 3·8% (2·9 to 5·0) 1 (ref) ··
analysed time to first moderate or marked event, including
27 Gy (n=1367) 69 (5·0%) 4·7% (3·7 to 6·0) 1·16 (0·82 to 1·64); 0·6% (–0·7 to 2·3)
Kaplan-Meier estimates of cumulative incidence, and
0·41 groups compared using HRs from Cox proportional
26 Gy (n=1368) 76 (5·6%) 5·1% (4·0 to 6·4) 1·27 (0·90 to 1·79); 1·0% (–0·4 to 2·9) hazards regression and the pairwise log-rank test.
0·17 Scores for change in photographic breast appearance at
Any breast cancer-related event‡ 2 and 5 years were modelled using generalised estimating
40 Gy (n=1361) 119 (8·7%) 7·8% (6·5 to 9·4) 1 (ref) ·· equations. Categories of mild and marked change in
27 Gy (n=1367) 112 (8·2%) 7·2% (5·9 to 8·7) 0·93 (0·71 to 1·20); –0·6% (–2·2 to 1·5) photographic breast appearance were combined for
0·56 analysis because very few had marked changes. Pairwise
26 Gy (n=1368) 114 (8·3%) 7·5% (6·2 to 9·0) 0·94 (0·73 to 1·22); –0·4% (–2·1 to 1·6) comparisons of mild or marked change at 2 or 5 years
0·65
between groups were described by ORs obtained from
All-cause mortality
the generalised estimating equations models and the
40 Gy (n=1361) 92 (6·8%) 5·4% (4·3 to 6·8) 1 (ref) ··
Wald test. Because of multiple testing, a significance
27 Gy (n=1367) 105 (7·7%) 6·9% (5·7 to 8·4) 1·12 (0·85 to 1·48); 0·6% (–0·8 to 2·5) level of 0·005 was used for the clinician and patient
0·42
normal tissue effects assessments; all hypotheses for the
26 Gy (n=1368) 90 (6·6%) 5·6% (4·5 to 7·0) 0·96 (0·72 to 1·28); –0·2% (–1·5 to 1·5)
0·78 normal tissue effects endpoints were two-sided.
Estimates of fractionation sensitivity (α/β values) in
Hazard ratios less than 1 favour five-fraction schedules. p values were calculated by log-rank test (two-sided). *Includes
three patients with angiosarcoma in ipsilateral breast (one in the 40 Gy group and two in the 26 Gy group). †Defined FAST-Forward were obtained for the primary endpoint of
as ipsilateral breast tumour relapse or regional relapse (axilla, supraclavicular fossa, and internal mammary chain). ipsilateral breast tumour relapse and late normal tissue
‡Includes local, regional, or distant relapse, breast cancer death, or contralateral breast cancer (disease-free survival). effects as per methods in the START and FAST trials.3
Table 2: Relapse and mortality by fractionation schedule: time-to-event analysis (n=4096) The α/β estimate for breast cancer was obtained from a
Cox proportional hazards regression model of time to
first ipsilateral breast tumour relapse, and for late normal
calculated, and hazard ratios (HRs; with 95% CIs) tissue effects from generalised estimating equations
comparing fractionation schedules obtained from Cox models including all follow-up assessments (separate
proportional hazards regression, censoring patients at models for photographic and clinician assessments).
date of death or last follow-up. Absolute differences (with Each model included terms for total dose and total dose
95% CIs) in 5-year ipsilateral breast tumour relapse multiplied by fraction size; the α/β ratio was calculated by
incidence were estimated by applying the HRs (and dividing the two parameter estimates respectively, with a
95% CIs) to the control group 5-year event-free estimate.18 95% CI estimated from the model using the covariance of
Primary assessment of non-inferiority was based on the two estimates (lower confidence limits were truncated
whether the upper limit of the two-sided 95% CI at zero). Isoeffect doses in 2 Gy equivalents (EQD2) were
(corresponding to one-sided 97·5% CI) for the absolute calculated for the five-fraction schedules, together with an
difference in 5-year ipsilateral breast tumour relapse was estimate of the five-fraction schedule that would be
less than 1·6%. Non-inferiority of each five-fraction isoeffective with 40 Gy in 15 fractions in terms of local
schedule versus control was also tested using the a priori tumour control and late normal tissue effects. No
critical HR of 1·81 (ln0·964/ln0·98, from protocol- correction was made for difference in treatment time.
specified incidence); p<0·025 was deemed statistically No formal interim analyses were done; accumulating
signifi­cant (probability of incorrectly accepting an infer­ data were monitored annually by the independent data
ior five-fraction schedule). An exploratory competing monitoring committee. All analyses were performed on
risks analysis was done for ipsilateral breast tumour an intention-to-treat basis that included all patients

1618 www.thelancet.com Vol 395 May 23, 2020

 Articles

100

3
40 Gy in 15 fractions
27 Gy in five fractions
26 Gy in five fractions
Ipsilateral breast tumour relapse (%)

27 Gy vs 40 Gy: hazard ratio 0·86 (95% CI 0·51 to 1·44);

5-year difference –0·3% (95% CI –1·0 to 0·9); non-inferiority p=0·0022

26 Gy vs 40 Gy: hazard ratio 0·67 (95% CI 0·38 to 1·16);

5-year difference –0·7% (95% CI –1·3 to 0·3); non-inferiority p=0·00019
0
0 1 2 3 4 5 6 7
Time since randomisation (years)
40 Gy
Number at risk 1361 1347 1307 1281 1230 1045 486 91
Censored 0 13 46 65 109 289 844 1239
Events 0 1 8 15 22 27 31 31
27 Gy
Number at risk 1367 1352 1328 1303 1255 1066 508 90
Censored 0 11 27 48 90 278 833 1250
Events 0 4 12 16 22 23 26 27
26 Gy
Number at risk 1368 1347 1325 1302 1257 1070 524 89
Censored 0 17 34 54 95 280 824 1258
Events 0 4 9 12 16 18 20 21

Figure 2: Cumulative risk of ipsilateral breast tumour relapse by fractionation schedule

according to their allocated treatment regardless of what 40 Gy group, 12 in the 27 Gy group, and 21 in the 26 Gy
was actually received. Because the main hypothesis was group did not receive the allocated therapy and were not
non-inferiority, the primary endpoint was also tested in included in the per-protocol popu­lation. Compliance with
the per-protocol population, which excluded patients for allocated treatment was 99%. Demographic and clinical
whom a major deviation was reported. The database charac­teristics at baseline were well balanced between
snapshot was taken on Nov 22, 2019; Stata, version 15 groups (table 1). 5-year visit forms were available for
(StataCorp), was used for analyses. The trial is registered 3681 (96%) patients of 3833 still in follow-up (not died,
at isrctn.com, ISRCTN19906132. withdrawn, or lost).
After a median follow-up of 71·5 months (IQR 71·3 to
Role of the funding source 71·7), ipsilateral breast tumour relapse was recorded in
The funder of the study had no role in study design, data 79 patients (31 in the 40 Gy group, 27 in the 27 Gy group,
collection, data analysis, data interpretation, or writing of and 21 in the 26 Gy group); HRs versus 40 Gy in
the report. The corresponding author had full access to 15 fractions were 0·86 (95% CI 0·51 to 1·44) for 27 Gy in
all data in the study and had final responsibility for the five fractions and 0·67 (0·38 to 1·16) for 26 Gy in five
decision to submit for publication. fractions. Estimated cumulative incidence of ipsilateral
breast tumour relapse up to 5 years was 2·1% (95% CI
Results 1·4 to 3·1) for 40 Gy (expected incidence 2%), 1·7%
Between Nov 24, 2011, and June 19, 2014, 4110 patients (1·2 to 2·6) for 27 Gy and 1·4% (0·9 to 2·2) for 26 Gy
were enrolled in the FAST-Forward trial. After ran­ (table 2, figure 2). Estimated absolute differences in
domisation, 14 patients withdrew consent for use of ipsilateral breast tumour relapse versus 40 Gy were –0·3%
data and were removed from the intention-to-treat popula­ (–1·0 to 0·9) for 27 Gy and –0·7% (–1·3 to 0·3) for 26 Gy.
tion; thus, 4096 patients were included in the intention- The upper confidence limits excluded an increase in
to-treat anlaysis (1361 assigned to 40 Gy in 15 fractions; ipsilateral breast tumour relapse of 1·6% or more so non-
1367 assigned to 27 Gy in five frac­tions; and 1368 assigned inferiority can be claimed for both five-fraction schedules
to 26 Gy in five fractions; figure 1). Seven patients in the compared with 40 Gy in 15 fractions. A test against the

www.thelancet.com Vol 395 May 23, 2020 1619

 Articles

were similar between groups, with no statistically

40 Gy in 27 Gy in 26 Gy in
15 fractions five fractions five fractions significant differences (table 2; appendix pp 3–4). No
(n=1361) (n=1367) (n=1368) formal subgroup analyses were done because of the low
Local tumour control event (primary 31 (2·3%) 27 (2·0%) 21 (1·5%) number of primary endpoint events, but frequencies of
endpoint)* ipsilateral breast tumour relapse, regional relapse, and
Local relapse 23 (1·7%) 22 (1·6%) 17 (1·2%) distant relapse were tabulated according to age, grade,
Ipsilateral breast, new primary 6 (0·4%) 3 (0·2%) 4 (0·3%) and ER and HER2 status for descriptive purposes; as
Cannot differentiate 2 (0·1%) 2 (0·1%) 0 expected, ipsilateral breast tumour relapse occurred in
Regional relapse 13 (1·0%) 11 (0·8%) 10 (0·7%) more of the patients with higher-grade primary tumours
Distant relapse 59 (4·3%) 69 (5·0%) 76 (5·5%) (appendix p 7). Invasive contralateral breast cancer
Contralateral breast, second primary 23 (1·7%) 20 (1·5%) 23 (1·7%) was reported for 55 (1·3%) of 4096 patients (18 [1·3%] in
Invasive 18 (1·3%) 17 (1·2%) 20 (1·5%) the 40 Gy group, 17 [1·2%] in the 27 Gy group, and
Ductal carcinoma in situ 5 (0·4%) 3 (0·2%) 2 (0·1%) 20 [1·5%] in the 26 Gy group; table 3), and non-breast
Unknown 0 0 1 (0·1%) second primary cancers were reported for 123 (3·0%) of
Non-breast, second primary 42 (3·1%) 37 (2·7%) 44 (3·2%)
4096 patients (42 [3·1%] in the 40 Gy group, 37 [2·7%] in
Death 92 (6·8%) 105 (7·7%) 90 (6·6%)
the 27 Gy group, and 44 [3·2%] in the 26 Gy group;
Breast cancer† 47 (3·5%) 51 (3·7%) 53 (3·9%)
table 3), the most common being colorectal cancer
Second cancer 12 (0·9%) 16 (1·2%) 10 (0·7%)
(25 cases).
287 (7·0%) of 4096 patients died, 151 (3·7%) from breast
Cardiac 10 (0·7%) 9 (0·7%) 8 (0·6%)
cancer, 125 (3·1%) from other causes, and 11 (0·3%) with
Other cause 17 (1·7%) 27 (2·0%) 16 (1·2%)
unknown cause of death and no evidence of disease
Unknown 6 (1·2%) 2 (0·1%) 3 (0·2%)
relapse (table 3). Of 27 patients with a cardiac-related death,
Data are n (%). Patients reporting events of more than one type are included in each relevant row. *Includes angiosarcoma 15 had a history of cardiac disease reported at random­
in ipsilateral breast (one in the 40 Gy group and two in the 26 Gy group) and six patients with ductal carcinoma in situ
(three in the 40 Gy group, two in the 27 Gy group, and one in the 26 Gy group). †Includes 13 patients with distant relapse isation or were a current or ex-smoker in the past year.
before death from other causes (four in the 40 Gy group, four in the 27 Gy group, and five in the 26 Gy group). At least one annual clinical assessment of normal
tissue effects was available for 3975 (97·0%) of
Table 3: Relapses, second primary cancers, and deaths by fractionation schedule (n=4096)
4096 patients. At 5 years, any moderate or marked
clinician-assessed normal tissue effects in the breast or
critical HR greater than 1·81 confirmed the result, with a chest wall was reported for 98 of 986 (9·9%) 40 Gy
p value of 0·0022 for 27 Gy and 0·00019 for 26 Gy patients, 155 (15·4%) of 1005 27 Gy patients, and 121 of
compared with 40 Gy. Analyses in the per-protocol 1020 (11·9%) 26 Gy patients (appendix pp 5–6, 8–9), with
population were consistent (estimated absolute difference a significant difference between 40 Gy and 27 Gy
vs 40 Gy –0·4% [–1·0 to 0·8], p=0·0017 for 27 Gy and (p=0·0003) but not between 40 Gy and 26 Gy (p=0·17).
–0·6% [–1·2 to 0·4], p=0·00037 for 26 Gy; full data Breast shrinkage was the most prevalent moderate or
for per-protocol analyses not shown because treatment marked effect at 5 years, reported in 50 (5·5%) of
compliance was 99%). Comparing the five-fraction 916 40 Gy patients, 78 (8·2%) of 948 27 Gy patients, and
schedules, the estimated absolute difference in ipsilateral 65 (6·8%) of 954 26 Gy patients (appendix pp 8–9).
breast tumour relapse cumulative incidence up to 5 years Longitudinal analysis of all annual clinical assessments
was –0·4% (–1·0 to 0·6) for 26 Gy versus 27 Gy. The of normal tissue effects over follow-up showed a
unadjusted α/β estimate for ipsilateral breast tumour significantly increased risk of any moderate or marked
relapse was 3·7 Gy (0·3 to 7·1), with EQD2 estimates of effect in the breast or chest wall for the 27 Gy group
44·7 Gy for 40 Gy, 43·1 Gy for 27 Gy, and 40·6 Gy for 26 Gy compared with 40 Gy (OR 1·55 [95% CI 1·32 to 1·83],
with no correction for treatment time. Adjusting for risk p<0·0001), with no significant difference between 26 Gy
group and ER and HER2 status made minimal difference and 40 Gy (1·12 [0·94 to 1·34], p=0·20; table 4). This
(adjusted α/β estimate 3·7 Gy [95% CI 0·4 to 6·9]). HRs pattern was similar for the individual effects of breast
obtained from a competing risks analysis of ipsilateral distortion, shrinkage, induration, and breast or chest
breast tumour relapse with death from any cause as a wall oedema, with significantly higher risk for 27 Gy
competing event were almost identical to those from the than 40 Gy but not for 26 Gy (table 4; appendix pp 5–6).
primary analysis (HRs from competing risks model were Comparing the two five-fraction schedules, 26 Gy had
0·85 [95% CI 0·51 to 1·43] for 27 Gy vs 40 Gy; and 0·67 significantly lower risk of any moderate or marked
[0·38 to 1·16] for 26 Gy vs 40 Gy). breast or chest wall normal tissue effects (p=0·0001) and
Regional relapses occurred in 34 (0·8%) of 4096 patients breast shrinkage (p=0·0018) compared with 27 Gy. Esti­
(13 [1·0%] of 1361 in the 40 Gy group, 11 [0·8%] in the mates of 5-year cumulative incidence of any mode­rate or
27 Gy group, and ten [0·7%] in the 26 Gy group; marked clinician-assessed normal tissue effects in the
table 3), six of which were concurrent with ipsilateral breast or chest wall were 26·8% (95% CI 24·4 to 29·4)
breast tumour relapse. Incidence of locoregional relapse, for 40 Gy, 35·1% (32·4 to 37·9) for 27 Gy, and 28·5%
distant relapse, dis­ease-free survival, and overall survival (26·0 to 31·1) for 26 Gy (appendix p 10). Results for

1620 www.thelancet.com Vol 395 May 23, 2020

 Articles

Number of moderate or Odds ratio for schedule p value for comparison p value for Odds ratio for years of
marked events/total (95% CI) with 40 Gy comparison follow-up (95% CI); p value
number of assessments between 27 Gy
over follow-up and 26 Gy
Any adverse event in the ·· ·· ·· ·· 0·98 (0·96–1·00); 0·055
breast or chest wall*
40 Gy 651/6121 (10·6%) 1 (ref) ·· ·· ··
27 Gy 1004/6303 (15·9%) 1·55 (1·32–1·83) <0·0001 ·· ··
26 Gy 774/6327 (12·2%) 1·12 (0·94–1·34) 0·20 0·0001 ··
Breast distortion† ·· ·· ·· ·· 0·99 (0·95–1·02); 0·38
40 Gy 232/5724 (4·0%) 1 (ref) ·· ·· ··
27 Gy 363/5953 (6·1%) 1·51 (1·15–1·97) 0·0028 ·· ··
26 Gy 299/5945 (5·0%) 1·20 (0·91–1·60) 0·19 0·083 ··
Breast shrinkage† ·· ·· ·· ·· 1·03 (1·00–1·06); 0·023
40 Gy 330/5728 (5·8%) 1 (ref) ·· ·· ··
27 Gy 503/5944 (8·5%) 1·50 (1·20–1·88) 0·0004 ·· ··
26 Gy 369/5943 (6·2%) 1·05 (0·82–1·33) 0·71 0·0018 ··
Breast induration ·· ·· ·· ·· 1·00 (0·96–1·04); 0·95
(tumour bed)†
40 Gy 185/5713 (3·2%) 1 (ref) ·· ·· ··
27 Gy 304/5948 (5·1%) 1·56 (1·19–2·05) 0·0013 ·· ··
26 Gy 236/5937 (4·0%) 1·19 (0·90–1·59) 0·23 0·047 ··
Breast induration ·· ·· ·· ·· 0·96 (0·90–1·02); 0·17
(outside tumour bed)†
40 Gy 45/5712 (0·8%) 1 (ref) ·· ·· ··
27 Gy 137/5943 (2·3%) 2·79 (1·74–4·50) <0·0001 ·· ··
26 Gy 97/5930 (1·6%) 1·90 (1·15–3·14) 0·013 0·059 ··
Telangiectasia ·· ·· ·· ·· 1·21 (1·14–1·29); <0·0001
40 Gy 63/6087 (1·0%) 1 (ref) ·· ·· ··
27 Gy 100/6272 (1·6%) 1·68 (1·07–2·65) 0·025 ·· ··
26 Gy 102/6300 (1·6%) 1·53 (0·96–2·43) 0·070 0·65
Breast or chest wall ·· ·· ·· ·· 0·73 (0·69–0·78); <0·0001
oedema
40 Gy 89/6097 (1·5%) 1 (ref) ·· ·· ··
27 Gy 217/6287 (3·4%) 2·18 (1·57–3·03) <0·0001 ·· ··
26 Gy 155/6318 (2·4%) 1·47 (1·03–2·09) 0·032 0·0097 ··
Breast or chest wall ·· ·· ·· ·· 0·93 (0·89–0·97); 0·0003
discomfort
40 Gy 234/6086 (3·8%) 1 (ref) ·· ·· ··
27 Gy 269/6285 (4·3%) 1·10 (0·86–1·40) 0·44 ·· ··
26 Gy 250/6309 (4·0%) 0·98 (0·76–1·26) 0·86 0·35 ··

Results for years of follow-up show trend in normal tissue effects over follow-up across all fractionation schedules. p values are calculated by Wald test; odds ratios are
estimated from the generalised estimating equations model including all follow-up data and show relative odds of moderate or marked adverse event (vs none or mild) for
each pairwise comparison of fractionation schedules across all follow-up assessments. *Includes shrinkage, induration, telangiectasia, or oedema. †Patients who had breast
conservation surgery or mastectomy with reconstruction.

Table 4: Longitudinal analysis of moderate or marked clinician-assessed late normal tissue effects for patients with at least one annual clinical assessment
(n=3975)

comparison of schedules from the analyses of time to were 1771 (99·6%) of 1778 at baseline, 1668 (96·2%) of
first moderate or marked effect were similar to those 1733 at 3 months, 1622 (94·2%) of 1722 at 6 months,
from the long­itudinal modelling of all annual clinical 1599 (93·7%) of 1707 at 1 year, 1531 (91·7%) of 1669 at
assessments (appendix p 10). 2 years, and 1334 (84·0%) of 1589 at 5 years. Of the
1796 patients consented to the patient-reported 1774 patients with at least one completed questionnaire,
outcomes substudy, 18 of whom withdrew consent 1634 had breast conservation surgery and 140 had
immediately after randomisation or were not given the mastectomy. Change in breast appearance had the
baseline booklet. Questionnaires returned from those highest 5-year prevalence, with moderate or marked
expected (patients alive and well, not withdrawn) change reported in 140 (32·4%) of 432 for 40 Gy,

www.thelancet.com Vol 395 May 23, 2020 1621

 Articles

158 (35·9%) of 440 for 27 Gy, and 136 (31·7%) of 429 for moderate or marked breast hardness or firmness at
26 Gy. 5-year prevalence of patient-reported adverse 5 years was not significantly increased for 27 Gy
effects were not significantly different between the compared with 40 Gy and breast swelling was not more
schedules (appendix pp 5–6, 11–12). Patient-reported prevalent in both five-fraction schedules than the 40 Gy

Number of patients Number of moderate or Odds ratio for p value for p value for Odds ratio for years of
reporting moderate marked events/total schedule (95% CI) comparison comparison follow-up (95% CI);
or marked event at number of assessments with 40 Gy between p value
baseline/total* over 3–60 months of 27 Gy and
follow-up 26 Gy
Protocol-specific items
Breast appearance ·· ·· ·· ·· ·· 1·03 (1·01–1·05);
changed 0·0010
40 Gy 170/573 (29·7%) 778/2480 (31·4%) 1 (ref) ·· ·· ··
27 Gy 177/583 (30·4%) 929/2550 (36·4%) 1·22 (1·02–1·46) 0·033 ·· ··
26 Gy 155/581 (26·7%) 770/2563 (30·0%) 0·91 (0·75–1·10) 0·33 0·0018 ··
Breast smaller ·· ·· ·· ·· ·· 1·11 (1·09–1·13);
<0·0001
40 Gy 96/560 (17·1%) 585/2445 (23·9%) 1 (ref) ·· ·· ··
27 Gy 106/576 (18·4%) 606/2520 (24·0%) 1·05 (0·85–1·29) 0·67 ·· ··
26 Gy 90/574 (15·7%) 515/2542 (20·3%) 0·81 (0·65–1·00) 0·053 0·017 ··
Breast harder or firmer ·· ·· ·· ·· 0·95 (0·93–0·97);
<0·0001
40 Gy 94/558 (16·8%) 499/2446 (20·4%) 1 (ref) ·· ·· ··
27 Gy 105/572 (18·4%) 690/2512 (27·5%) 1·42 (1·17–1·72) 0·0003 ·· ··
26 Gy 95/566 (16·8%) 626/2534 (24·7%) 1·22 (1·00–1·48) 0·048 0·1007 ··
Skin appearance ·· ·· ·· ·· ·· 0·96 (0·93–0·99);
changed 0·0080
40 Gy 78/577 (13·5%) 345/2505 (13·8%) 1 (ref) ·· ·· ··
27 Gy 61/586 (10·4%) 392/2571 (15·2%) 1·03 (0·83–1·28) 0·77 ·· ··
26 Gy 67/580 (11·5%) 338/2576 (13·1%) 0·90 (0·72–1·13) 0·37 0·23 ··
European Organisation for Research and Treatment of Cancer QLQ-BR23 items
Breast pain ·· ·· ·· ·· ·· 0·96 (0·94–0·99);
0·011
40 Gy 53/583 (9·1%) 338/2538 (13·3%) 1 (ref) ·· ·· ··
27 Gy 42/590 (7·1%) 428/2601 (16·5%) 1·23 (0·98–1·54) 0·068 ·· ··
26 Gy 53/588 (9·0%) 417/2597 (16·1%) 1·23 (0·98–1·53) 0·074 0·96 ··
Breast swollen ·· ·· ·· ·· ·· 0·84 (0·80–0·89);
<0·0001
40 Gy 56/583 (9·6%) 122/2538 (4·8%) 1 (ref) ·· ·· ··
27 Gy 43/589 (7·3%) 236/2597 (9·1%) 1·46 (1·10–1·94) 0·0080 ·· ··
26 Gy 47/589 (8·0%) 192/2599 (7·4%) 1·27 (0·95–1·69) 0·11 0·22 ··
Breast oversensitive ·· ·· ·· ·· ·· 0·96 (0·93–0·99);
0·0097
40 Gy 57/579 (9·8%) 283/2528 (11·2%) 1 (ref) ·· ·· ··
27 Gy 42/584 (7·2%) 334/2596 (12·9%) 1·10 (0·87–1·40) 0·43 ·· ··
26 Gy 62/586 (10·6%) 319/2587 (12·3%) 1·11 (0·88–1·41) 0·37 0·91 ··
Skin problems in breast ·· ·· ·· ·· ·· 0·96 (0·92–1·01);
0·11
40 Gy 26/582 (4·5%) 156/2539 (6·1%) 1 (ref) ·· ·· ··
27 Gy 24/290 (4·1%) 209/2596 (8·0%) 1·25 (0·95–1·65) 0·11 ·· ··
26 Gy 18/590 (3·0%) 164/2592 (6·3%) 0·98 (0·73–1·31) 0·90 0·084 ··
Arm or shoulder pain ·· ·· ·· ·· ·· 1·00 (0·97–1·03);
>0·99
40 Gy 66/582 (11·3%) 401/2537 (15·8%) 1 (ref) ·· ·· ··
27 Gy 78/591 (13·2%) 441/2601 (17·0%) 1·12 (0·91–1·37) 0·29 ·· ··
26 Gy 81/589 (13·7%) 455/2599 (17·5%) 1·14 (0·93–1·40) 0·2006 0·83 ··
(Table 5 continues on next page)

1622 www.thelancet.com Vol 395 May 23, 2020

 Articles

Number of patients Number of moderate or Odds ratio for p value for p value for Odds ratio for years of
reporting moderate marked events/total schedule (95% CI) comparison comparison follow-up (95% CI);
or marked event at number of assessments with 40 Gy between p value
baseline/total* over 3–60 months 27 Gy and
follow-up 26 Gy
(Continued from previous page)
Arm or hand swollen ·· ·· ·· ·· ·· 1·06 (1·00–1·11);
0·031
40 Gy 24/582 (4·1%) 101/2536 (4·0%) 1 (ref) ·· ·· ··
27 Gy 17/588 (2·9%) 103/2600 (4·0%) 0·95 (0·66–1·36) 0·77 ·· ··
26 Gy 22/590 (3·7%) 124/2592 (4·8%) 1·14 (0·80–1·62) 0·46 0·31 ··
Difficulty raising arm ·· ·· ·· ·· ·· 1·04 (0·99–1·08);
0·089
40 Gy 27/582 (4·6%) 171/2533 (6·7%) 1 (ref) ·· ·· ··
27 Gy 36/589 (6·1%) 209/2599 (8·0%) 1·24 (0·94–1·63) 0·12 ·· ··
26 Gy 37/587 (6·3%) 188/2596 (7·2%) 1·12 (0·85–1·48) 0·42 0·46 ··
Results for years of follow-up show trend in normal tissue effects over follow-up across all fractionation schedules. p values are calculated by Wald test; odds ratios are
estimated from the generalised estimating equations model including all questionnaires (baseline to 5 years) and show relative odds of moderate or marked adverse events
(vs none or mild) for each pairwise comparison of fractionation schedules across all questionnaires. *Total is those who completed the corresponding question.

Table 5: Longitudinal analysis of moderate or marked patient-assessed late normal tissue effects from baseline to 5 years for patients with at least one
completed questionnaire (n=1774)

schedule, at the prespecified cutoff of p=0·005. The unadjusted α/β estimate for any moderate or
Longitudinal analyses of all patient assessments from marked clinician-assessed normal tissue effects in the
baseline to 5 years showed a significantly higher risk of breast or chest wall was 1·7 Gy (95% CI 1·2 to 2·3),
moderate or marked breast hardness or firmness for giving EQD2 estimates of 47·1 Gy for 40 Gy in 15 frac­tions,
27 Gy compared with 40 Gy (OR 1·42, 1·17, 1·72, 51·6 Gy for 27 Gy in 5 fractions, and 48·3 Gy for 26 Gy
p=0·0003), and a lower risk of change in breast in 5 fractions; adjusting for prognostic factors (age,
appearance for 26 Gy compared with 27 Gy (p=0·0018), boost, and whole-breast planning treatment volume as
but no significant differences between schedules for the a proxy for breast size) made very little difference.
other normal tissue effects (table 5; appendix pp 5–6). The α/β estimated from the photographic endpoint
Of the 1737 patients who consented to the photographic (adjusting for breast size and surgical deficit assessed
substudy, baseline photographs were received for from the baseline photographs) was very similar (1·8 Gy
1634 (94·1%), and 2-year or 5-year photo­ graphs were [1·1 to 2·4]). The unadjusted α/β estimate for patient-
available for 1385 (79·7%). 1309 (75·4%) were patients who reported change in breast appearance was 2·3 Gy
had breast conservation surgery; for these patients, 2-year (1·8 to 2·9), resulting in EQD2 estimates of 46·1 Gy for
photographs were assessed in 1267 and 5-year photographs 40 Gy, 48·2 Gy for 27 Gy, and 45·2 Gy for 26 Gy; adjusting
were assessed in 875 (appendix p 13). 226 patients died or for covariates made minimal difference.
withdrew from the photographic substudy by year 5; for The most common specialist referral for radiotherapy-
the remainder, the most common reasons for photographs related adverse effects during follow-up was to lympho­
not being taken were appointments not made because edema clinics (appendix p 14). Incidence of ischaemic
of clerical errors at the centres, patients not attending heart disease, symptomatic rib fracture, and symptomatic
clinic visits, and patients withdrawing consent from the lung fibrosis was very low at this stage of follow-up
substudy. At 2 years, mild or marked change in photo­ (appendix p 14).
graphic breast appearance was reported in 35 (8·5%) of
411 for 40 Gy, 67 (15·6%) of 429 for 27 Gy, and 46 (10·8%) of Discussion
427 for 26 Gy; corresponding results at 5 years were We demonstrated non-inferiority, measured in terms of
34 (12·0%) of 283 for 40 Gy, 83 (26·9%) of 308 for 27 Gy, ipsilateral breast tumour relapse, of 27 Gy and 26 Gy five-
and 37 (13·0%) of 284 for 26 Gy (appendix p 13). Modelling fraction schedules compared with 40 Gy in 15 fractions at
2-year and 5-year photographic assessments together, 5 years’ follow-up for patients with early breast cancer,
27 Gy had a significantly increased risk of mild or marked most of whom were treated by local tumour excision and
change in breast appearance compared with 40 Gy sentinel node biopsy for node-negative disease. Normal
(OR 2·29 [95% CI 1·60 to 3·27], p<0·0001), with no tissue effects up to 5 years for the 26 Gy in five fractions
significant difference between 26 Gy and 40 Gy (OR 1·26 schedule were similar to those with the 40 Gy in
[0·85 to 1·86], p=0·24; appendix p 13). 26 Gy had a 15 fractions schedule. Low rates of ipsilateral breast
significantly lower risk of change in photographic breast tumour relapse and of moderate or marked late normal
appearance than 27 Gy (p=0·0006). tissue effects can be attributed to improvements in all

www.thelancet.com Vol 395 May 23, 2020 1623

 Articles

diagnostic and treat­ ment modalities and to the com­ about biological processes, which include a time factor
mitment of patients to early diagnosis and randomised in FAST-Forward, does not interfere with clinical evalu­
trials.20 ation and decisions on implementation of FAST-Forward
The 10-year analyses of ipsilateral breast tumour results in similar patient groups.
relapse and normal tissue effects reported by earlier The five-fraction regimen is relevant to partial-breast
Canadian5 and UK trials3,4,6–8 confirm that although radiotherapy, the preferred alternative to whole-breast
normal tissue effects continue to accumulate beyond radiotherapy for many women after recent phase 3
5 years, there is evidence that relative differences between trials.22–25 Beyond its safety and effectiveness, the 26 Gy
test and control groups change very little over time.21 In FAST-Forward schedule is convenient and substantially
the START-B trial, the HR for clinician-assessed breast less expensive for patients and for health services. It is
shrinkage after 40 Gy in 15 fractions compared with also likely to be safe for patients requiring regional radio­
50 Gy in 25 fractions was 0·83 (95% CI 0·66–1·04) at therapy, an approach that is under formal assessment in
5 years and 0·80 (0·67–0·96) at 10 years, by which a randomised FAST-Forward substudy comparing 40 Gy
time the proportion of patients with breast shrinkage in 15 fractions and 26 Gy in five fractions. Assuming no
increased from 11·4% (9·5–13·6) at 5 years to 26·2% effect of time, 26 Gy in five fractions is equivalent to
(23·2–29·6) at 10 years.8 The findings of FAST-Forward 46·8 Gy and 53·7 Gy in 2-Gy fractions assuming
can be applied to different prognostic groups in view of α/β values of 2 Gy and 1 Gy, respectively, dose intensities
the very low overall ipsilateral breast tumour relapse well within the limits of tolerance for these structures.26,27
incidence, a conclusion consistent with a meta-analysis In terms of limitations of our study, there is no reason to
of the 5861 patients entered into the three START trials, consider the heart more sensitive to fraction size than
which identified no inconsistency of effect in terms of most other soft tissues. It is undoubtedly sensitive to
normal tissue effects or recurrence risk across any of the total dose but the tiny number of cardiac events in FAST-
prognostic or treatment subgroups investigated.8 Forward prevents meaningful analysis.28 Any heart
The absence of a detectable dose response for local exposure is potentially harmful even after 2 Gy fractions,
tumour control between 26 Gy and 27 Gy in five frac­ so the priority is to exclude the heart from the treatment
tions is a potential limit to precision, but this feature volume as far as possible using deep inspiration breath
reflects the shallowness of the dose response curve for hold or a similar technique,29,30 or partial breast
subclinical breast cancer that is around the 98% local radiotherapy.23 The size of the trial prevents reliable
tumour control level, so the –0·4% estimated difference subgroup analyses by patient age, tumour grade, receptor
in absolute levels of ipsilateral breast tumour relapse status, and systemic therapies, but consistent with our
between 27 Gy and 26 Gy probably reflects random 10-year analyses of almost 6000 patients in the START
sampling variability in the ipsilateral breast tumour hypofractionation trials, there is no suggestion of
relapse rate or chance imbalances in unmeasured heterogeneity.8 Finally, synchronous boost regimens
prognostic factors between test groups. For late normal were avoided despite current interest in this application
tissue effects, the dose response is much steeper, of hypofractionation. It is safer to consider boost as an
enabling detection of clinically and statistically signifi­ independent treatment variable such as tumour grade or
cant differ­ences in event rates between 26 Gy and 27 Gy adjuvant systemic therapy whose impacts are randomly
in five fractions. The five-fraction schedule isoeffective distributed across treatment groups. This allows a pure
with 40 Gy in 15 fractions allows direct estimation of assessment of whole-breast hypofractionation without
α/β for late normal tissue effects, which is consistent having to con­sider the partial volume effects of different
with values generated from our other trials. The breast dose levels. Routine implementation of 26 Gy in
α/β value of 3·7 Gy (95% CI 0·3–7·1) for tumour control five fractions can more naturally incorporate appropriate
in FAST-Forward is similar to the 3·5 Gy (1·2–5·7) five-fraction synchronous boost regimens.
estimated from the START pilot and START-A trials.8 In conclusion, 5-year ipsilateral breast tumour relapse
Point estimates of α/β values, assuming no effect of incidence after a 1-week course of adjuvant breast
time, for late normal tissue effects in FAST-Forward radiotherapy delivered in five fractions is non-inferior to
scored by clinicians, patients, and photographic assess­ the standard 3-week schedule according to the predefined
ments are closer to 2 Gy than the 3 Gy estimated in the inferiority threshold. The 26 Gy dose level is similar to
earlier START8 and FAST trials,14 but 95% CIs overlap 40 Gy in 15 fractions in terms of patient-assessed normal
for each endpoint in all trials. In FAST, 915 women were tissue effects, clinician-assessed normal tissue effects,
randomly assigned after breast conservation surgery for and photographic change in breast appearance, and is
node negative disease to 50 Gy in 25 fractions versus similar to normal tissue effects expected after 46–48 Gy in
two dose levels of a five-fraction regimen delivered once 2 Gy fractions. The consistency of FAST-Forward results
weekly, thereby ensuring complete repair between with earlier hypofractionation trials supports the adoption
fractions and controlling for overall treatment time.13,14 of 26 Gy in five daily fractions as a new standard for
The α/β value for change in photographic breast women with operable breast cancer requiring adjuvant
appearance in FAST was 2·6 Gy (1·4–3·7). Uncertainty radiotherapy to partial or whole breast.

1624 www.thelancet.com Vol 395 May 23, 2020

 Articles

Contributors expressed are those of the authors and not necessarily those of the NIHR
AMB and JRY are the current and previous chief investigators or the Department of Health and Social Care.
respectively, and DAW is the chief clinical coordinator for the trial.
References
JMB is the trials methodology lead within the Institute of Cancer 1 Darby S, McGale P, Correa C, et al. Effect of radiotherapy after
Research-Clinical Trials and Statistics Unit (ICR-CTSU) and provided breast-conserving surgery on 10-year recurrence and 15-year breast
oversight and guidance for trial management throughout the trial. cancer death: meta-analysis of individual patient data for
JRY, JMB, and JSH were responsible for the study design. AMB, JRY, 10,801 women in 17 randomised trials. Lancet 2011; 378: 1707–16.
and JSH wrote the first draft of the manuscript. JSH was responsible for 2 McGale P, Taylor C, Correa C, et al. Effect of radiotherapy after
statistical analyses and contributed to data interpretation. AA, DJB, CC, mastectomy and axillary surgery on 10-year recurrence and 20-year
MC, SC, CEC, AG, AH, PH, AMK, CCK, CM, ZN, ES, NS, and IS are breast cancer mortality: meta-analysis of individual patient data for
members of the FAST-Forward trial management group (TMG), 8135 women in 22 randomised trials. Lancet 2014; 383: 2127–35.
which contributed to study design, was responsible for oversight 3 Yarnold J, Ashton A, Bliss J, et al. Fractionation sensitivity and dose
throughout the trial and contributed to data interpretation and response of late adverse effects in the breast after radiotherapy for
manuscript preparation. MAS and LS managed the study and data early breast cancer: long-term results of a randomised trial.
collection at ICR-CTSU. CM is a patient advocate member of the TMG Radiother Oncol 2005; 75: 9–17.
and provided guidance for study documentation and reports. PH was 4 Owen JR, Ashton A, Bliss JM, et al. Effect of radiotherapy fraction
the lead for the patient-reported outcomes substudy. ZN was size on tumour control in patients with early-stage breast cancer
responsible for radiotherapy quality assurance. All authors reviewed after local tumour excision: long-term results of a randomised trial.
Lancet Oncol 2006; 7: 467–71.
and approved the manuscript.
5 Whelan TJ, Pignol JP, Levine MN, et al. Long-term results of
Declaration of interests hypofractionated radiation therapy for breast cancer. N Engl J Med
JMB, JSH, MAS, and LS report grants from National Institute for Health 2010; 362: 513–20.
Research Health Technology Assessment and Cancer Research UK during 6 Bentzen SM, Agrawal RK, Aird EG, et al. The UK Standardisation
the conduct of the study. JMB reports grants and non-financial support of Breast Radiotherapy (START) trial A of radiotherapy
from Novartis (previously GlaxoSmithKline), AstraZeneca, Clovis hypofractionation for treatment of early breast cancer: a randomised
Oncology, Janssen-Cilag, Merck Sharpe and Dohme, Puma Biotechnology, trial. Lancet Oncol 2008; 9: 331–41.
Pfizer, and Roche; and grants from Medivation, outside the submitted 7 Bentzen SM, Agrawal RK, Aird EG, et al. The UK Standardisation
work. DAW reports travel grants from Roche Pharmaceuticals outside the of Breast Radiotherapy (START) trial B of radiotherapy
submitted work. CCK reports personal fees from Roche Pharmaceutical hypofractionation for treatment of early breast cancer: a randomised
outside the submitted work. All other authors declare no competing trial. Lancet 2008; 371: 1098–107.
interests. 8 Haviland JS, Owen JR, Dewar JA, et al. The UK Standardisation of
Breast Radiotherapy (START) trials of radiotherapy
Data sharing hypofractionation for treatment of early breast cancer: 10-year
Deidentified individual participant data, together with a data dictionary follow-up results of two randomised controlled trials. Lancet Oncol
defining each field in the set, will be made available to other researchers 2013; 14: 1086–94.
on request. Trial documentation including the protocol are available 9 Offersen B, Nielsen HM, Jacobsen EH, et al. Hypo- vs
online. The Institute of Cancer Research-Clinical Trials and Statistics normofractionated radiation of early breast cancer in the For the trial documentation
Unit (ICR-CTSU) supports wider dissemination of information from the randomized DBCG HYPO trial. Radiother Oncol 2018; 127: S312. and protocol see https://www.
research it conducts and increased cooperation between investigators. 10 Wang SL, Fang H, Song YW, et al. Hypofractionated versus icr.ac.uk/fastforward
Trial data are obtained, managed, stored, shared, and archived according conventional fractionated postmastectomy radiotherapy for patients
to ICR-CTSU standard operating procedures to ensure the enduring with high-risk breast cancer: a randomised, non-inferiority,
open-label, phase 3 trial. Lancet Oncol 2019; 20: 352–60.
quality, integrity, and utility of the data. Formal requests for data sharing
are considered in line with ICR-CTSU procedures, with due regard given 11 National Institute for Health and Care Excellence. Early and locally
advanced breast cancer: diagnosis and management. July 18, 2018.
to funder and sponsor guidelines. Requests are via a standard proforma
https://www.nice.org.uk/guidance/ng101 (accessed April 22, 2020).
describing the nature of the proposed research and extent of data
12 Smith BD, Bellon JR, Blitzblau R, et al. Radiation therapy for the
requirements. Data recipients are required to enter a formal data sharing
whole breast: executive summary of an American Society for
agreement, which describes the conditions for release and requirements
Radiation Oncology (ASTRO) evidence-based guideline.
for data transfer, storage, archiving, publication, and intellectual property. Pract Radiat Oncol 2018; 8: 145–52.
Requests are reviewed by the trial management group in terms of 13 Brunt AM, Haviland J, Sydenham M, et al. FAST phase III RCT of
scientific merit and ethical considerations, including patients’ consent. radiotherapy hypofractionation for treatment of early breast cancer:
Data sharing is undertaken if proposed projects have a sound scientific or 10-year results (CRUKE/04/015). Int J Radiat Oncol 2018;
patients’ benefit rationale, as agreed by the trial management group and 102: 1603–04.
approved by the independent data monitoring and steering committee, 14 Agrawal RK, Alhasso A, Barrett-Lee PJ, et al. First results of the
as required. Restrictions relating to patients’ confidentiality and consent randomised UK FAST trial of radiotherapy hypofractionation for
will be limited by aggregating and anonymising identifiable patients’ treatment of early breast cancer (CRUKE/04/015). Radiother Oncol
data. Additionally, all indirect identifiers that could lead to deductive 2011; 100: 93–100.
disclosures will be removed in line with ICR-CTSU data sharing 15 Brunt AM, Wheatley D, Yarnold J, et al. Acute skin toxicity
guidelines. associated with a 1-week schedule of whole breast radiotherapy
compared with a standard 3-week regimen delivered in the UK
Acknowledgments FAST-Forward Trial. Radiother Oncol 2016; 120: 114–18.
We thank the patients who participated in this trial and staff at the 16 Haviland JS, Ashton A, Broad B, et al. Evaluation of a method for
participating centres and at the Institute of Cancer Research-Clinical Trials grading late photographic change in breast appearance after
and Statistics Unit (ICR-CTSU); we also thank FAST-Forward trial radiotherapy for early breast cancer. Clin Oncol (R Coll Radiol) 2008;
management group members past and present, and the independent data 20: 497–501.
monitoring committee and trial steering committee for overseeing the trial 17 Howard DR, Brown JM, Todd S, Gregory WM. Recommendations
(appendix pp 1–2). The trial is sponsored by The Institute of Cancer on multiple testing adjustment in multi-arm trials with a shared
Research (London, UK). The FAST-Forward trial is funded by the National control group. Stat Methods Med Res 2018; 27: 1513–30.
Institute for Health Research (NIHR) Health Technology Assessment 18 Altman DG, Andersen PK. Calculating the number needed to treat
programme (UK; 09/01/47), with programme grants from Cancer for trials where the outcome is time to an event. BMJ 1999;
Research UK to support the work of ICR-CTSU (C1491/A15955; 319: 1492–95.
C1491/A25351). We acknowledge NHS funding to the NIHR Biomedical 19 Hanley JA, Negassa A, Edwardes MD, Forrester JE. Statistical
Research Centre at the Royal Marsden NHS Foundation Trust (London, analysis of correlated data using generalized estimating equations:
UK) and The Institute of Cancer Research (London, UK). The views an orientation. Am J Epidemiol 2003; 157: 364–75.

www.thelancet.com Vol 395 May 23, 2020 1625

 Articles

20 Mannino M, Yarnold JR. Local relapse rates are falling after breast 26 Pierce SM, Recht A, Lingos TI, et al. Long-term radiation
conserving surgery and systemic therapy for early breast cancer: complications following conservative surgery (CS) and radiation
can radiotherapy ever be safely withheld? Radiother Oncol 2009; therapy (RT) in patients with early stage breast cancer.
90: 14–22. Int J Radiat Oncol Biol Phys 1992; 23: 915–23.
21 Whelan T, MacKenzie R, Julian J, et al. Randomized trial of breast 27 Gałecki J, Hicer-Grzenkowicz J, Grudzień-Kowalska M, Michalska T,
irradiation schedules after lumpectomy for women with lymph Załucki W. Radiation-induced brachial plexopathy and
node-negative breast cancer. J Natl Cancer Inst 2002; 94: 1143–50. hypofractionated regimens in adjuvant irradiation of patients with
22 Polgár C, Ott OJ, Hildebrandt G, et al. Late side-effects and breast cancer—a review. Acta Oncol 2006; 45: 280–84.
cosmetic results of accelerated partial breast irradiation with 28 Darby SC, Ewertz M, McGale P, et al. Risk of ischemic heart disease
interstitial brachytherapy versus whole-breast irradiation after in women after radiotherapy for breast cancer. N Engl J Med 2013;
breast-conserving surgery for low-risk invasive and in-situ 368: 987–98.
carcinoma of the female breast: 5-year results of a randomised, 29 Bartlett FR, Colgan RM, Carr K, et al. The UK HeartSpare study:
controlled, phase 3 trial. Lancet Oncol 2017; 18: 259–68. randomised evaluation of voluntary deep-inspiratory breath-hold in
23 Coles CE, Griffin CL, Kirby AM, et al. Partial-breast radiotherapy women undergoing breast radiotherapy. Radiother Oncol 2013;
after breast conservation surgery for patients with early breast 108: 242–47.
cancer (UK IMPORT LOW trial): 5-year results from a multicentre, 30 Bartlett FR, Donovan EM, McNair HA, et al. The UK HeartSpare
randomised, controlled, phase 3, non-inferiority trial. Lancet 2017; study (stage II): multicentre evaluation of a voluntary breath-hold
390: 1048–60. technique in patients receiving breast radiotherapy. Clin Oncol 2017;
24 Vicini FA, Cecchini RS, White JR, et al. Long-term primary results 29: e51–56.
of accelerated partial breast irradiation after breast-conserving
surgery for early-stage breast cancer: a randomised, phase 3,
equivalence trial. Lancet 2019; 394: 2155–64.
25 Whelan TJ, Julian JA, Berrang TS, et al. External beam accelerated
partial breast irradiation versus whole breast irradiation after breast
conserving surgery in women with ductal carcinoma in situ and
node-negative breast cancer (RAPID): a randomised controlled trial.
Lancet 2019; 394: 2165–72.

1626 www.thelancet.com Vol 395 May 23, 2020