Management of breast cancer has evolved significantly over the past decades, moving away from radical procedures towards less aggressive surgery. Breast-conserving surgery (BCS), when combined with radiotherapy (RT), has been shown to confer equivalent oncological outcomes compared with mastectomy1,2,3 and has been established as standard of care, when technically feasible, especially for patients with early-stage disease.

Advances in the multimodality management of breast cancer have led to improved oncological outcomes and reduced local recurrence (LR) rates.4 However, despite these advances, 5–15%5,6,7 of patients treated with BCS and RT may still experience ipsilateral breast cancer recurrence (IBCR). The surgical management of IBCR has traditionally been mastectomy. This has been supported by international recommendations, including the National Comprehensive Cancer Network (NCCN) guidelines.8 However, a number of studies have suggested that repeat BCS (rBCS) with or without repeat RT (rRT) may be an alternative.9,10,11,12 In one of the first reports, Kurtz et al.9 showed that rBCS without rRT in a selected cohort of patients was associated with acceptable oncological outcomes, as demonstrated by overall survival (OS). Similar results in terms of OS and breast cancer-specific survival (BCSS) have also been shown in more recent studies,13,14,15,16 although there are also publications reporting opposite results.17,18 In addition, the reported LR rates after rBCS have been variable.11,15,18,19,20 However, despite the conflicting data, there has been a trend towards increasing utilization of rBCS15,21 and recently the St. Gallen International Consensus guidelines also supported rBCS as an option, no longer considering mastectomy as absolutely obligatory for the management of IBCR.22

The aim of this study was to perform a systematic review of the literature and meta-analysis of the oncological outcomes in patients treated with rBCS with or without rRT for the management of IBCR following previous BCS and RT.

Methods

Search Strategy and Inclusion Criteria

A systematic review of the literature was conducted in the MEDLINE and EMBASE databases, using the search terms ‘ipsilateral breast tumour recurrence’, ‘ipsilateral breast cancer recurrence’, ‘ipsilateral breast tumor recurrence’, ‘ipsilateral recurrent breast cancer’, ‘IBTR’, ‘local recurrence + breast cancer + breast conserving surgery + mastectomy’. No chronological limitations were stipulated. In the absence of dedicated randomized controlled trials, prospective and retrospective comparative and non-comparative cohort studies, cross-sectional studies reporting on second LR and/or survival after rBCS for IBCR following previous BCS and RT were considered eligible. Studies that did not clearly specify whether the reference population had initially been treated for only ductal carcinoma in situ (DCIS) or both DCIS and invasive breast cancer (IBC) were included in the primary analysis. Respectively, we registered whether data regarding the type of in-breast recurrence (IBC or DCIS) was reported separately or cumulatively. If more than one report on the same patients was available, only the most recent was included.

Data Extraction

Data extraction was performed independently by two authors (CJT and EP) in a preformed Microsoft Excel© (Microsoft Corporation, Redmond, WA, USA) working sheet. The data extraction procedure for the whole dataset (including all eligible studies) was standardized during two training sessions with the senior authors (AK and MKT) using a random sample of five studies. Disagreement was resolved by group consensus. The study methodology was registered with PROSPERO, International Prospective Register of Systematic Reviews (CRD42021286123, https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42021286123).

Quality Assessment

The Newcastle–Ottawa Scale (NOS)23 for observational studies, as assessed by two authors (EP, AK), was used to evaluate the quality of the included studies. Publication bias was assessed with funnel plots and the Egger’s test for small studies. Following analyses and critical appraisal, the Grading of Recommendations, Assessment, Development and Evaluations [GRADE] approach24 was used to assess the strength of evidence and recommendations by two authors (AV and AK). Knowledge gaps and research priorities were subsequently defined.

Statistical Analyses and Reporting

Rates of a second LR and OS at 5 years for rBCS and salvage mastectomy were calculated separately by pooling the outcomes from single-arm and comparative studies. Subgroup analyses were performed depending on whether the reference population had initially been treated for only DCIS, both DCIS and IBC, or IBC only. Subgroup analyses were also undertaken to define the effect of study design (comparative or single-arm), propensity score matching and the effect of RT, regardless of the technique that was utilized. The median follow-up was also extracted. Meta-analyses of comparative studies were also performed. Additionally, leave-one-out meta-analyses of comparative studies were performed to allow for the identification of studies with exaggerated effect sizes and to guide further subgroup and meta-regression analyses. As the literature search was expected to retrieve observational studies, the use of a random-effects model using the DerSimonian Laird method was decided a priori. For source studies directly reporting odds ratio (OR), risk ratio (RR), or hazard ratio (HR), the adjusted analyses and Kaplan–Meier curves were considered for data extraction and calculation of 5-year second LR and OS.25,26 Effect sizes were reported with 95% confidence intervals (CIs). Study heterogeneity was assessed using the I2 statistic.

This manuscript was prepared according to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines.27 Stata v17 (StataCorp. 2021. Stata Statistical Software: Release 17. College Station, TX: StataCorp LLC.) was used for all statistical analyses.

Results

Study Selection and Characteristics

After the removal of duplicates, the literature search retrieved 42 studies, with 24 examining outcomes after a primary IBC, 17 reporting on both IBC and DCIS, and 1 on DCIS only (MOOSE flowchart is presented in Fig. 1). Twenty-eight studies examined outcomes on both LR and OS, 9 on OS only, and 5 on LR only. Study characteristics and NOS scores are shown in Table 1. On two occasions, it was not explicitly reported by the authors if the study population was the same as in another publication by the same group.28,29 Therefore, all studies were included in Table 1 but only the most recent studies providing data following propensity score matching were included in the meta-analysis.19,30

Fig. 1
figure 1

Flowchart of systematic review and meta-analysis of observational studies in epidemiology. * Two studies were not explicitly described by the authors if they represented the same population as other publications

Full size image
Table 1 Characteristics of the included studies
Full size table

Second Local Recurrence

Source studies reporting on a second LR had a median follow-up ranging from 24.5 to 165.6 months (median of medians 70 months, interquartile range [IQR] 52–73). The overall pooled incidence of a second LR after rBCS was 15.7% (95% CI 12.1–19.7), and 10.3% (95% CI 6.9–14.3) after salvage mastectomy. Despite the fact these were separately pooled outcomes without comparison, the confidence intervals were numerically overlapping, suggesting that the difference may not be significant, but study heterogeneity was high. The results of the subgroup analyses across all included studies are summarized in Table 2. Overall, among patients treated with rBCS, those who received rRT had the lowest pooled second LR rate compared with the other subgroups (9.6%, 95% CI 5.0–15.3).

Table 2 Pooled rates of second local recurrence with separate subgroup analyses across all studies (single-arm and comparative)
Full size table

A total of 17 studies provided comparative data on second LR after rBCS and salvage mastectomy. The median follow-up ranged from 30 to 165.5 months (median of medians 72 months, IQR 52–79). In comparative studies, the pooled second LR rate was higher after rBCS (19.6%, 95% CI 15.5–24.0) versus after salvage mastectomy (9.6%, 95% CI 6.3–13.5) [Table 2]. On meta-analysis, rBCS was associated with a significantly increased risk of second LR (RR 2.103, 95% CI 1.535–2.883; p < 0.001, I2 = 55.1%), as shown in Fig. 2. Leave-one-out meta-analysis (electronic supplementary Fig. S1) did not demonstrate any differences. Only concomitant RT retained a protective effect in meta-regression analysis (coefficient − 0.317, 95% CI − 0.596 to − 0.038; p = 0.026, I2 = 40.4%). No publication bias or small-studies effect was detected (Egger’s test beta-1 1.540; p = 0.103).

Fig. 2
figure 2

Forest plot of studies comparing repeat breast-conserving surgery versus salvage mastectomy for second local recurrence. *Study by Kurtz et al.49

Full size image

Overall Survival

Pooled OS rates and subgroup analyses for patients treated with rBCS or salvage mastectomy are presented in Table 3. Overall, at a median follow-up ranging from 30 to 168 months (median of medians 66 months, IQR 55–79), the pooled 5-year OS rate was 86.8% (95% CI 83.4–90.0) after rBCS, and 79.8% (95% CI 74.7–84.5) after salvage mastectomy. Subgroup analyses (Table 3) did not demonstrate any factor that correlated with difference in outcomes for each group (rBCS or salvage mastectomy). Meta-analysis of comparative studies (n = 20) showed a small OS benefit in favor of rBCS (RR 1.040, 95% CI 1.003–1.079; p = 0.032, I2 = 70.8%) [Fig. 3]. The median follow-up in these studies ranged from 42 to 168 months (median of medians 72 months, IQR 59–126.6). Leave-one-out meta-analysis (electronic supplementary Fig. S2) showed that the omission of four studies (one at a time) would result in a difference, despite that the numeric value of the RR was not significantly affected. Subsequent subgroup and meta-regression analysis was performed (electronic supplementary Table S1). RT did not affect the outcome on meta-regression analysis (coefficient 0.0019, 95% CI − 0.0274–0.0312; p = 0.898, I2 = 70.8%). With regard to primary tumor, studies reporting on both DCIS and IBC reported survival benefit for rBCS (RR 1.119, 95% CI 1.019–1.230; p = 0.019), but this effect was not retained on meta-regression analysis (coefficient 0.0721, 95% CI − 0.0017–0.1458; p = 0.056). When looking into publication bias, the Egger’s test detected a small-studies effect (Egger’s test beta-1 0.93; p = 0.041).

Table 3 Pooled overall 5-year survival rates with separate subgroup analyses across all studies (single-arm and comparative)
Full size table
Fig. 3
figure 3

Forest plot of studies comparing repeat breast-conserving surgery versus salvage mastectomy for overall survival

Full size image

Study Quality and Strength of Recommendations

The median NOS score was 8.5 (IQR 7–9). No correlation was identified between the timing of the study publication and the median NOS, suggesting that study quality has not improved over the years.

The GRADE recommendations from the meta-analysis are summarized in Table 4. The certainty of evidence was very low due to serious risk of bias (mainly selection), inconsistency, and imprecision. The main reasons for that were deemed to be the design of available studies (retrospective single-arm and comparative, mostly without matching or consecutive patients), the fact that most studies reported outcomes in the form of rates, rather than effect sizes such as HRs that are much more appropriate for time-to-event outcomes, and, finally, that most source studies did not accurately report on primary and recurrent tumor biology as well as adjuvant therapy, for example use of RT after BCS for the management of the initial cancer, or RT for the management of the recurrence, which may play a pivotal role in oncological outcomes. These factors constituted the main knowledge gaps and thus research priorities for future studies.

Table 4 GRADE assessment and recommendations
Full size table

Discussion

Mastectomy has traditionally been considered as the standard of care for the management of IBCR. This has been recommended by national and international guidelines, including the NCCN guidelines.8 Reasons for this practice include the concerns about rRT and also the fact that IBCR has been associated with poor prognosis,6,31 potentially supporting the argument for more aggressive local treatment. However, salvage mastectomy does not eliminate the risk of local or distant recurrence32,33 and there is increasing data supporting the feasibility of rRT.16,34 In addition, advances in multidisciplinary management of breast cancer, including systemic therapy and RT options, as well as a general trend towards surgical de-escalation, have likely contributed to the increasing use of rBCS as part of an individualized, tailored approach.15,21 This is also now supported by the St. Gallen International Consensus Guidelines.22 Avoidance of mastectomy, if oncologically safe, could be associated with improved patient satisfaction in terms of cosmetic outcome and quality of life35,36 apart from cost and resource implications for healthcare providers. However, the existing data do not conclusively support rBCS or salvage mastectomy in terms of oncological outcomes, with a number of studies reporting opposite results.9,10,11,12,13,17,18,19,20,29,37,38

The present systematic literature review showed variable second LR rates after rBCS. The overall pooled second LR rate was found to be 15.7% after rBCS compared with 10.3% after salvage mastectomy. However, it should be noted that the included studies are markedly heterogeneous and there was no standardized multidisciplinary treatment protocol for the management of IBCR. In addition, it is important to highlight that in a number of studies, a proportion of patients did not receive RT for the management of the primary cancer, with insufficient data provided to allow stratification for this in the analysis. On meta-analysis, rBCS was associated with a significantly higher RR for second LR (RR 2.103), albeit with moderate study heterogeneity. This RR is similar to that reported in a recent meta-analysis (RR 1.87).39 The small observed difference may be explained by the fact that the present meta-analysis included 17 studies providing data on second LR compared with 13 studies in the meta-analysis by Mo et al.39.

On subgroup analysis, the lowest second LR rate among patients treated with rBCS was observed in those receiving rRT (9.6%). The protective effect of rRT was also demonstrated in meta-regression analysis. This finding is in line with previous reports highlighting the potentially important role of rRT in improving local control after rBCS for IBCR.34,39 This is an important consideration when individualizing the management plan, especially as a number of rRT options, for example brachytherapy,40,41,42 intraoperative RT43,44 and external beam RT,16 have been shown to be associated with an acceptable toxicity profile. In the RTOG 1014 prospective phase II clinical trial, three-dimensional conformal external beam partial breast rRT after rBCS for IBCR in patients previously treated with BCS and RT was associated with low risk of second LR (5%) and late Grade 3 adverse events in only 7% of the cases, while there were no Grade 4 or higher reported adverse events.16 Tolerability of rRT has also been supported by the results from a recent meta-analysis.34

Despite the finding that rBCS may be associated with a higher risk of second LR, which was two-fold higher based on the results of the present meta-analysis, it may not have a negative impact on survival. A number of retrospective studies have shown that OS was not inferior, or was even improved, in patients treated with rBCS with or without rRT compared with those treated with salvage mastectomy.13,15,19,29,30,42,45 An analysis of the Surveillance, Epidemiology, and End Results (SEER) database including data from 1998 to 2013 showed no significant difference in terms of OS and BCSS in patients treated with rBCS or salvage mastectomy.14 However, another analysis of the SEER database looking into data from 1973 to 2003 showed different results.17 In that study, the authors found that rBCS was associated with worse OS and BCSS and that rRT had a protective effect in terms of OS. Although, there is no clear explanation for the discordant findings, a potential reason may be the different time periods, as multidisciplinary breast cancer management has significantly evolved over the past decades. The recent meta-analysis by Mo et al. also supports the findings that rBCS may not be associated with worse OS.39 The results of the present meta-analysis showed a marginal benefit in OS in favor of rBCS (RR 1.040). The difference between the two meta-analyses may be explained by the different number of included studies (8 vs. 20 in the present analysis). The median NOS of the studies10,11,12,33,37,42,45,46 included in the meta-analysis by Mo et al.39 was 9 (IQR 7–9) and the median NOS of the studies in the present meta-analysis was also 9 (IQR 8–9), with the additional 12 studies having a median NOS of 9 (IQR 9–9). It has to be noted though that a small-study effect was found, underlining potential publication bias. While such an effect was not detected in the meta-analysis by Mo et al.39, cautiousness is required due to the small number of included studies.

Although rRT was found to have a protective effect in terms of local control and has previously been shown to have a role in improving OS,17,45 in the present meta-analysis OS was not affected by rRT on meta-regression analysis. However, these results should be interpreted with caution as the included studies were substantially heterogeneous and the effect size had marginal significance.

The findings of this meta-analysis suggest that although rBCS may be associated with higher risk of subsequent LR, this may not have a negative impact on OS. This suggests that rBCS may be an alternative option in the context of individualized management of IBCR in line with the St. Gallen International Consensus Guidelines,22 especially for women who want to preserve their breast, following careful consultation about the currently accepted standard recommendation of salvage mastectomy as per NCCN8 guidelines. However, appropriate patient selection for such an approach would be of paramount importance. In the first report of rBCS for IBCR, Kurtz et al. suggested an algorithm for patient selection including tumor size <2 cm, no fixation of the cancer on the skin or chest wall, clinically node-negative status, and no significant RT changes.9 Other important parameters included disease-free interval and the size and histopathology of the recurrence as these have been shown to be independent prognostic factors of OS.45 Gentilini et al. have suggested that patients with small (≤2 cm), late (>48 months) IBCR would be the ideal candidates for rBCS.47 Similar selection criteria have been proposed by the German Society of Radiation Oncology (DEGRO) expert panel, suggesting that rBCS can be considered in patients ≥50 years of age with unifocal, small (<2–3 cm) IBCR, ≥48 months after primary treatment who are willing to undergo rBCS and this is technically feasible.48 The St. Gallen International Panel suggests that rBCS can be considered for low-risk recurrent cancers with favorable tumor biology (small, Luminal A) for which rRT may not be required or for IBCR >5 years after primary treatment.22 The common denominator of these suggested algorithms for patient selection is an individualized approach mainly based on tumor biology and anatomical stage. The role of multidisciplinary management of IBCR, with systemic therapy (endocrine therapy, chemotherapy, or targeted therapy for example anti-HER2) with or without rRT cannot be overemphasized. The potential effect of such recommendations could not be assessed in this meta-analysis due to lack of studies providing data that would allow such an analysis.

Although, rBCS is increasingly being used for the management of IBCR,15,21 and de-escalated tailored therapeutic approaches are favored within modern multidisciplinary working, the quality of the studies providing data on oncological outcomes of rBCS does not appear to improve over time, as demonstrated by the NOS assessment of the studies included in this meta-analysis. The low quality of available source studies constitutes the limitation of this meta-analysis, as potentially uncontrolled biases, lack of standardized reports of treatment modalities, and outcomes of interest increase heterogeneity and mandate a careful interpretation of the results. This fact was illustrated in the outcomes of the GRADE approach and highlights the importance of collaboration across different specialties to set-up prospective research studies designed to address the knowledge gaps highlighted.

Conclusions

rBCS may have a role in the management of IBCR in patients previously treated with BCS and RT. This should be based on individualized assessment of tumor and patient factors and multidisciplinary working to develop a tailored management plan. Further research in this field is warranted to allow optimal patient selection and address existing knowledge gaps.