Introduction

In 2020, over 2.3 million patients worldwide were diagnosed with breast cancer1, and improving outcomes mean that many will be long-term survivors2. Current treatments for breast cancer include a combination of locoregional and systemic therapies with most patients having surgery ± radiotherapy together with chemotherapy, targeted treatments and endocrine therapy according to tumour subtype and disease stage.

There is increasing evidence that not all patients, particularly those with low or intermediate risk disease, benefit from all treatments that they currently receive. Most, if not all, however, will experience complications, side effects and other treatment-related toxicities, which can have profound and long-lasting impacts on their well-being and quality of life3. In addition to the personal costs, continued recommendation of treatments with marginal or no benefit has significant societal and global implications, diverting scarce healthcare resources from patients who would benefit from more intensive treatments4,5. Patients are strongly supportive of more personalised treatment approaches but emphasise that high quality research is needed to underpin any change in practice, and to demonstrate that proposed reductions in treatment are ‘safe’ and do not compromise oncological outcomes3. Well-designed, efficient and timely trials evaluating how to safely reduce or ‘de-escalate’ components of breast cancer treatment in appropriately selected patient groups are therefore a global research priority6,7,8.

Locoregional therapy with surgery and/or radiotherapy is the mainstay of treatment for early breast cancer and optimising these treatments will have the greatest potential benefit for the increasing numbers of patients who have low- or intermediate-risk disease. Understanding the current landscape of locoregional therapy de-escalation trials is essential to ensure that future studies are designed to meet the needs of patients and the breast cancer community, address research gaps, minimise duplication of effort and research waste and promote efficient and timely generation of high-quality data that can be rapidly translated into practice. This systematic review therefore aims to map the current landscape of ongoing and recently completed de-escalation trials to inform future research.

Results

The search identified 3991 results. After removal of duplicates, 3142 studies were screened for inclusion of which 129 were reviewed in full. 32 were excluded following discussion (Supplementary Table 1). A total of 97 studies (48 (49.5%) RCTs9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56 and 49 (50.5%) cohorts57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105) with 94,866 planned participants were included in the review (Fig. 1).

Fig. 1
figure 1

PRISMA diagram for the review.

Full size image

Characteristics of included studies are summarised in Table 1. Overall, most were multicentre (n = 67, 69.1%) and based in Europe or North America (n = 73, 75.3%). Of those reporting planned and actual recruitment numbers (n = 44), 61.4% (n = 27/44) achieved their accrual target, with only 8 (18.2%) closing due to recruitment issues. Recruitment periods ranged from 12 months to 12 years (median 5 years) with overall study durations of up to 27 years78 (median 9 years). Most studies (n = 77, 79.4%) had an oncological primary endpoint evaluated at a median of 60 months, but few studies (n = 9, 9.3%) selected participants based on biomarker stratification19,22,26,52,66,75,79,90,91. Patient-reported outcomes were only included as either a primary17 or co-primary11,21 endpoint in three (3.1%) studies and only 6 (6.2%)11,48,51,53,55,91 explicitly stated that patient advocates had been involved in study design. Details of included studies are summarised in Supplementary Table 2.

Table 1 Characteristics of included studies by study design and locoregional treatment modality de-escalated
Full size table

Compared with cohort studies, RCTs were more likely to be multicentre, international, government-funded studies with larger planned and actual recruitment numbers, and longer study durations. Cohort studies were significantly more likely to measure their primary outcome at an earlier timepoint, achieve their target accrual and receive industry funding than RCTs (Table 1).

Of the 97 studies, 51 (52.3%)20,27,31,32,34,40,73,74,76,80,81,89,93,95,96,105 and 46 (47.4%)12,33,41,42,46,57,72,75,82,88,94,97,101,102,103,104 evaluated de-escalation of surgery and radiotherapy respectively.

Compared with radiotherapy studies, surgical de-escalation trials were more likely to focus on reduction/omission of treatment to regional nodes (vs. breast) and treatment after NST. Studies de-escalating radiotherapy (RT), by contrast were more likely to involve international recruitment and be biomarker stratified (Table 1).

Details of trial design were available for 73/97 (75.3%) trials: (43/48 (89.6%) RCTs and 30/49 (61.2%) cohorts). Most RCTs (38/43, 88.4%) were non-inferiority trials with oncological primary endpoints (34/38, 89.5%). Reported non-inferiority margins (n = 27) ranged from 1.25% to 10% (median 3.6%) with no differences between locoregional recurrence (LRR) vs survival endpoints or surgery and radiotherapy trials (data not shown) (Supplementary Table 3). When reported, most cohort studies (29/30, 96.7%) stated basing their sample size calculation on a pre-defined acceptable threshold often for LRR (17/29, 58.6%) or a composite (6/29, 20.7%) endpoint including survival (e.g. disease-free or event-free survival). Reported acceptable LRR thresholds (14/17) were consistently <10% (median 4.5%, range 2% to >10%) whereas thresholds for composite survival-based outcomes were more variable (range >98.5% to >84%) with reductions of up to 10% considered acceptable in some studies (Supplementary Table 3). Only 21 studies17,21,26,40,42,46,48,51,53,55,64,67,71,79,84,91,96, however, reported how non-inferiority margins or (un)acceptable thresholds were selected. Over half (n = 11) reported basing estimates on the published literature or those used in previous/ongoing trials and over a quarter (n = 6) stated the decision had been made exclusively by study team. Few studies (n = 4) referred to engagement with the wider breast cancer community or explicitly stated that patient advocates had been involved in these decisions (n = 4). Statistical power was reported for 41/97 (42.3%) trials and was mostly 80% (31/41 75.6%) with only one quarter (10/41, 24.4%) reporting 90% power or higher.

Locoregional therapy de-escalation research questions could be broadly divided into two groups: i) evaluation of response-adjusted locoregional treatment in patients having NST and ii) omission/reduction of surgery or radiotherapy in patients with low- or intermediate-risk disease (Table 2, Supplementary Table 2). In both settings, multiple concurrent studies, often in different geographical locations, addressing identical or similar research questions in different or overlapping patient populations were identified (Fig. 2).

Fig. 2: Concurrent locoregional treatment de-escalation studies including dates of publication of first pivotal study.
figure 2

a Trials of omission of SLNB in cN0 patients. b Trials of omission of ALND in SLNB+ patients. c Trials of omission of radiotherapy in low/very low risk early breast cancer. d Trials of partial breast irradiation (PBI) in early breast cancer.

Full size image
Table 2 Current landscape of locoregional de-escalation studies by treatment modality and research question
Full size table

Just over a quarter of trials (n = 26, 26.8%), predominantly early-phase and cohort studies (n = 18/26, 69.2%), evaluated de-escalation of surgery (n = 22)15,20,32,54,71,80,87,89,98,100 or radiotherapy (n = 6)11,20,35,64,72,94 in patients who responded well to NST (Table 2, Supplementary Table 2). This included one study de-escalating ‘axillary treatment’ (surgery or radiotherapy)11 and an RCT for cN1 patients that was withdrawn due to lack of funding20.

Five studies evaluated omission of breast surgery66,68,85,87,89 in patients with a complete response to NST. Of these, two85,89 were specifically early-phase feasibility studies while the remaining three aimed to determine oncological safety at a minimum of 3 years66,68,87. A further four omission of axillary surgery studies, aimed to assess the oncological safety of this approach following complete response to NST67.

Six studies (RCTs n = 211,32, cohorts n = 471,83,98,100) aimed to evaluate omission of axillary lymph node dissection (ALND) in patients converting from cN1 to cN0/ypN0 following NST. Cohort studies were either small single-centre trials98,100 or feasibility studies71,83, with two large RCTs11,32 investigating the impact of omitting ALND on five-year disease free survival (DFS) in a planned total of 4216 patients. Fewer studies (three RCTs9,10,54 and one cohort84) investigated omission of ALND in patients with ypN+ disease, and these aimed to assess oncological safety at a minimum of 5 years.

In many of the studies using response to NST as a biomarker for patient selection, definitions and inclusion criteria were not standardised. Definitions of ‘complete response’ ranged from clinical assessment alone, through to inclusion of imaging ± demonstration of a pathological complete response (pCR) in the breast. Studies omitting ALND in ypN+ disease varied in the burden of residual axillary disease permitted for study inclusion.

Three cohort studies64,72,94 aimed to investigate the oncological safety of omitting RT following breast conserving surgery (BCS) in cN0 patients achieving pCR, with one terminated prematurely due to recruitment issues72. One large RCT focused on omission of postmastectomy radiotherapy (PMRT) in patients with good treatment response35 and one RCT aimed to address omission of axillary radiotherapy in cN1 patients who had converted to ypN0 following treatment11.

The majority of studies in the portfolio (71/97, 73.2%) evaluated locoregional treatment de-escalation in patients having primary surgery, most commonly the omission of surgery or radiotherapy in patients with very low-risk disease.

In total, 12 concurrent, geographically diverse studies (3 RCTs16,48,55 and 9 cohorts62,69,70,73,74,81,93,95,105) commenced between 2014 and 2022 evaluating minimally-invasive alternatives to lumpectomy in patients with small, unifocal, node negative cancers. Most (n = 7/12, 58.3%) focused on cryoablation16,69,70,73,74,95,105, with the remainder studying a range of techniques including vacuum-assisted excision48,81, laser62 and radiofrequency ablation93. Many were early-phase and over half (7/12) included completeness of excision (a technical, feasibility outcome) post-procedure as a primary55,62,69,70,74,81 or co-primary48 endpoint.

Eight studies (5 RCTs14,27,34,49,56 and 3 cohorts76,86,99) aimed to recruit a total of 14,958 patients to assess the 5-year oncological safety of omitting SLNB. Most studies recruited clinically node negative patients aged 18 or over with T1-2 disease14,27,34,56,99, and all but one study76 required confirmation of node negativity with one14,27,34,49,56 or more86,99 imaging modality. Only one study restricted recruitment to patients over 6576 and only two limited their inclusion criteria to smaller tumours49,86. Most studies required participants to undergo breast conserving surgery followed by radiotherapy and any additional recommended adjuvant treatments14,27,34,56,76,99 with only one study permitting mastectomy as a surgical treatment option56. Of the five RCTs, four involving 10,437 patients have completed recruitment14,27,34 and the earliest trial published results in 202349 (Fig. 2a).

A total of 12 trials (6 RCTs19,21,22,24,33,41 and 6 cohort studies75,79,91,92,103,106) with a combined recruitment target of 13,145 patients aimed to evaluate the oncological safety of omitting breast radiotherapy following BCS in patients at very low risk of disease recurrence. Almost all (11/12, 91.7%) had a LRR primary endpoint assessed between 2.5 and 10 years (median 5 years). Studies recruited patients with ‘low risk’ T1N0 breast cancer but varied regarding inclusion criteria and how recurrence risk was assessed: several exclusively recruited older women21,33,41,103 and/or those with specific disease subtypes19,21,22,24,41. Several specifically recruited patients with luminal A disease21,22,79 determined either by assessment of ki6779,91 in combination with clinicopathological features (e.g. St Gallen Criteria) or using gene expression profiling with Prosigna/PAM5022,106. Other studies used Oncotype DX19,75 scores for risk stratification in similar patient populations. A single study used clinicopathological features and imaging with MRI to identify a low-risk group for omission for radiotherapy92. Most studies required patients to take adjuvant endocrine therapy for at least 5 years, but one study randomised participants to receive either partial breast radiotherapy or endocrine therapy21. Although the first radiotherapy omission trial opened in 2003 and has now published both 5-107 and 10-year41 outcomes, five large, multicentre RCTs addressing similar research questions19,21,22,24,33 and involving up to 5,989 patients are currently recruiting (Fig. 2c).

Reduction in target volume with external beam partial breast irradiation (PBI) in women with low-risk disease having BCS was evaluated in 13 studies (10 RCTs12,17,18,25,28,30,36,37,42,46 and 3 cohorts65,78,88) which opened between 2005 and 2019 with a combined recruitment target of 16,249 women (Fig. 2d). There were small but significant differences in inclusion/exclusion criteria across all studies. Almost all restricted recruitment by age, although the age criteria varied, with patients aged >4042,78, >5030,36,37,46, >5525, >6018,88 eligible for inclusion in different trials. Similarly, most trials restricted inclusion by tumour size, most commonly <3 cm12,28,30,37,42,88, but some studies only recruited patients with T1 disease25,46 or tumours <2.5 cm36,78. Consistent with identifying a ‘low risk’ population, almost all studies only included patients with node negative disease, but two also recruited N1 patients12,28. There was significant variability in the tumour grade and biological subtypes of cancer included in different studies and in the surgical resection margin required for study entry. This was most commonly >2 mm28,30,46,78 but ranged from ‘no tumour on ink’37 to >5mm25 in different trials.

Of the 13 studies, most have completed; two RCTs25,46 were terminated early due to recruitment issues and two remain open17,30. Most (n = 8/13, 61.5%) focused on oncological safety with LRR endpoints at between 5 and 20 years12,28,42,46,65,78,88 but radiotherapy-related toxicity18,36,37, cosmesis25,36 or quality of life17 were also evaluated at variable timepoints.

Nine trials (2 RCTs and 7 cohorts) opened between 2000 and 2015 to evaluate PBI using intraoperative radiotherapy53,63,77,82,101,102 or brachytherapy23,97,104. These have all completed recruitment.

A total of 13 trials (12 RCTs and 1 cohort study) de-escalating either surgery13,40,43,44,47,54,96 or radiotherapy13,26,29,50,51,52 in patients with intermediate-risk pN1 disease were identified.

Five large European RCTs13,40,43,44,47 aimed to assess the oncological safety of omitting ALND in SLNB+ patients in 11,930 patients (Fig. 2b) including two studies13,40 evaluating the omission of axillary treatment (either ALND or nodal radiotherapy). Most trials recruited patients aged 18 or above13,40,43,44 with clinically node negative T1-2 breast cancer13,40,44,47. Only one study included T3 disease and a further study restricted recruitment to patients aged between 40 and 75 years47. Three studies required US assessment of the axilla as part of the entry criteria13,43,47. Most permitted either BCS or mastectomy, but one study included only patients undergoing mastectomy13. Inclusion criteria regarding the degree of sentinel node involvement were also variable. Some studies mandated patients to have one or two sentinel node macro-metastases40,43 whereas others permitted inclusion of patients with micro-metastases13 or isolated tumour cells in their sentinel nodes44.

Of the five trials, three did not meet their recruitment targets, with two closed due to recruitment issues13,47. One of these later re-opened as single arm mastectomy cohort108. Only one is currently recruiting44.

Only one large RCT recruiting cN1 patients having primary surgery was identified and is currently open with participants randomised to either targeted axillary surgery and radiotherapy or ALND54.

Two RCTs aimed to assess the impact of omitting axillary and/or regional radiotherapy on 5-year survival in patients with limited pN1(sn) disease38,51, one of which was terminated prematurely due to poor recruitment38.

Four further RCTs aimed to assess the oncological safety of omitting regional nodal irradiation (RNI)29,39,52 or internal mammary chain (IMC) radiotherapy26 in intermediate-risk N1 patients. Two studies26,52 used biomarker stratification to assess risk and two included omission of PMRT39,52 in mastectomy patients randomised to omission of RNI. Omission of PMRT alone in intermediate-risk patients was also assessed in a further RCT50. One study closed due to poor recruitment29, one is in follow-up50 and the remainder are currently recruiting.

Discussion

This systematic review provides a comprehensive overview of the current landscape of locoregional therapy de-escalation trials in early breast cancer, which is focused on two broad areas: i) response-adjusted locoregional treatment in patients having NST and ii) omission/reduction of locoregional treatment in patients with low to intermediate-risk disease.

There is, however, significant duplication of effort across the portfolio, with multiple studies including several concurrent large-scale RCTs aiming to address identical or similar research questions. This was particularly evident in low-risk patients, with nine SLNB omission studies, 12 radiotherapy omission studies and 13 PBI studies with planned accrual of 12,919, 19,134 and 16,249 patients respectively identified. There were also five concurrent European RCTs evaluating omission of ALND in the SLNB+ population.

Challenges were highlighted with current de-escalation study design. It is increasingly stated that tumour biology rather than anatomy is key to determining outcomes but currently less than 10% of trials in the portfolio use biomarker stratification to inform participant selection and/or treatment. Although response to NST is a dynamic biomarker which is being utilised in a proportion of trials, the ongoing use of traditional methods of risk-stratification such as pathological stage in studies de-escalating locoregional treatment in patients with low- and intermediate-risk disease is a concern. Furthermore, future studies should consider not only the use of prognostic biomarkers to determine level of risk, but also whether predictive biomarkers can identify patients who will not benefit from specific treatments, enabling de-escalation based on biology as well as risk of recurrence.

As expected, most studies have an oncological primary endpoint and are designed to assess either non-inferiority or ‘acceptable’ levels of oncological safety at a minimum of 5-years. This is problematic as most patients have low risk disease with few recurrences. This results in large, long, and expensive studies with few events that may be underpowered, often struggle to recruit and risk practice changing in the interim without evidence, which may result in patient harm.

Justification for multiple concurrent large-scale RCTs addressing similar or identical research questions is challenging but may relate to lack of engagement of patients and the wider breast cancer community highlighted by this review. Patients are extremely supportive of research to improve personalisation of care3, but less than 10% of trials explicitly stated involving patient advocates in the study design. Similarly, few trials engaged with clinicians outside the immediate study team. Failure to engage the broader breast cancer community in both the choice of research question and study design can jeopardise future implementation. Lack of engagement and ownership increases the risk that trials will be criticised and reduces the likelihood that the findings will be implemented. Results of the ACOSOG Z0011 trial109,110, for example, were not accepted in Europe due to concerns about inclusion criteria, event rate and lack of radiotherapy quality assurance111. Five large-scale RCTs13,40,43,44,108 were subsequently developed to validate the results, delaying patient benefit and creating an opportunity cost to the breast cancer community through inefficient use of research funding.

This review highlights a lack of a co-ordinated global research across the entire locoregional de-escalation portfolio but similar issues also exist in the systemic therapy context112. There is therefore a need for the international breast cancer community to work together with patients to develop efficient, credible, well-designed studies that can and will be implemented globally.

There are limitations to this review which require consideration. Firstly, it has relied on publicly available data sources for information on included trials. While many trials have published either protocols or preliminary results, for many ongoing trials, only the trial registry record was available. This information is limited and may not include details such as patient involvement in study design or information on sample size calculations such that some of these parameters were underestimated. It is also possible that the search strategy may have failed to identify all potentially relevant trials. A combination of medical database and trials registry search was chosen to minimise this risk. Trial registration is mandatory, so it is unlikely that any potentially practice changing studies were missed. This review is therefore likely to present a comprehensive picture of the current locoregional de-escalation landscape and the areas where improvement is required.

There is an urgent need to address the current slow, inefficient and siloed approach to the design and delivery of locoregional de-escalation studies to optimise the value of future research and prevent research waste. This will involve international collaboration and engagement of the global breast cancer community, not only to agree key research priorities but also to ensure that future trials are well-designed and credible so that the results are implemented into practice globally. Patients want research to demonstrate oncological safety3, but recurrence in this patient population is now rare, so traditional RCTs powered to demonstrate non-inferiority for oncological endpoints are increasingly numerically unfeasible. Large sample sizes could be addressed by co-ordinated international recruitment or harmonisation of trial protocols particularly inclusion criteria and endpoint selection with agreed core outcome sets for locoregional de-escalation studies to facilitate data pooling and pre-planned individual patient level meta-analysis, which are currently hampered by the heterogeneity described above. This would require global collaboration and consensus, but should be achievable. While both strategies are undoubtedly part of the future solution, however, more innovative and efficient study designs such as trials within cohorts113 or platform studies114, facilitating exploration of multiple research hypotheses simultaneously, will be essential to generate timely results to inform practice. A platform approach, for example, could evaluate multiple potential approaches to risk stratification for radiotherapy omission, increasing efficiency and reducing costs by obviating the need for multiple separate trials. This, together with increased use of biomarkers to appropriately stratify participant risk and promote greater personalisation of care will create an efficient, responsive de-escalation portfolio. Finally, and most importantly, engagement with patient partners will be essential in the design and development of all future treatment de-escalation studies. Only then can we ensure such studies answer clinically relevant questions for the breast cancer community and are feasible, acceptable and deliverable.

Methods

This systematic review was prospectively registered on the PROSPERO International Register of Systematic Reviews (CRD42023487777) and is reported according to PRISMA guidelines115.

A comprehensive search of online databases (OVID, Medline, Embase and the Cochrane Library) and English language trial registries (clinicaltrials.gov, ISRCTN, Dutch trials registry, and European clinical trials registry platform) was undertaken on the 16/01/2024 to identify ongoing or recently completed trials evaluating the de-escalation of locoregional therapy in patients with early breast cancer. The search strategy was developed and iteratively refined using terms for ‘breast cancer’, ‘surgery’, ‘radiotherapy’ and ‘de-escalation’ and adapted for use in the trial registries. Details of the searches performed are summarised in Supplementary Table 4. The search was limited to studies published on or after 01/01/2019 to focus on the contemporary de-escalation portfolio.

All phase 2 and 3 randomised controlled trials (RCTs) and non-randomised interventional cohort studies evaluating de-escalation of locoregional treatment in adult patients >18 years with early invasive breast cancer that were in set-up, ongoing, had recently terminated or been published in English between 01/01/2019 and 31/12/2023 were eligible for inclusion. Excluded were observational (non-interventional) studies, reviews, letters and editorials. Studies investigating de-escalation of systemic breast cancer therapy were also excluded, but those evaluating de-escalation of locoregional treatments following neoadjuvant systemic therapy (NST) were eligible for inclusion.

Citations were uploaded to Zotero® software and duplicates removed. Remaining studies were screened for inclusion using pre-specified inclusion criteria by one reviewer (ADMcC). All potentially relevant studies and those in which eligibility was unclear were discussed by the study team at regular meetings and inclusion of all studies agreed by consensus.

For the purposes of this review, early breast cancer was defined as Stage 1–3 disease treated with curative intent. Locoregional treatment was defined as surgery or radiotherapy to the breast and/or regional nodal areas.

De-escalation was defined as either i) a reduction in the extent or ii) complete omission of surgery or radiotherapy. For radiotherapy, this included any reduction in target volume (e.g. partial breast irradiation) but not method of delivery, dose or fractionation.

Biomarker stratification was defined as selection of patients for treatment based on any biological, molecular or genetic information over and above that used for routine breast cancer diagnostics (i.e. ER/PR/HER2). This included Ki67, gene expression profiling and other genomic tests.

A data extraction tool was developed, iteratively refined and piloted in Microsoft Excel® by the study team. Data extracted included study name, registry number, number of sites, geographical location; design; modality of locoregional treatment de-escalated, site at which treatment was reduced/omitted (breast or nodal areas), use of biomarker stratification, primary outcome, source of funding and current status. Study start/end dates; planned/actual duration of recruitment and planned/actual recruitment numbers were also extracted. Finally, specific details related to the research question, study design, hypothesis, sample size calculation and patient and public involvement were extracted verbatim. All publicly available data sources including study protocols, published conference proceedings/abstracts and registry records were used to optimise data completeness. Data extraction was performed by one reviewer (ADMcC); the study team met regularly throughout the data extraction process to review the data and discuss and resolve uncertainties. A proportion (approximately 10%) of studies were independently reviewed by senior members of the team (SP, SAMcI) to ensure consistency and methodological rigor.

Simple descriptive statistics were used to summarise the results. Categorical data were summarised by counts and percentages and continuous data by median, interquartile range (IQR) and range. Characteristics of included studies were compared by i) study design (RCT vs cohort) and ii) treatment modality de-escalated (surgery vs. radiotherapy) using chi-squared and Kruskal-Wallis tests for categorical and continuous variables respectively. Extracted verbatim data regarding study design, research question and hypothesis testing were explored using content analysis116 and a narrative synthesis was performed. No risk of bias assessment or pooling of the data was undertaken as the aim of the review was to describe the landscape of locoregional breast cancer de-escalation studies.