Salvage stereotactic radiotherapy for locally recurrent brain metastases: a systematic review

Abstract

Background

Stereotactic radiotherapy (SRT), including stereotactic radiosurgery (SRS) and hypofractionated stereotactic radiotherapy (hfSRT), plays a key role in the management of brain metastases (BM). As advances in systemic therapies prolong survival, local recurrence of BM has become more frequent, prompting interest in salvage reirradiation strategies.

Objective

This systematic review aimed to evaluate the efficacy and safety of a second course of stereotactic radiotherapy (SRT2) for in-field recurrent brain metastases.

Methods

Data on local control, overall survival, and radionecrosis were extracted and pooled using random-effects models. Eleven retrospective studies published between 2020 and 2025 were included, comprising 914 patients and 2,352 brain metastases, with 389 lesions treated with salvage SRS2/SRT2. Patients who had received prior whole-brain radiotherapy were excluded.

Results

The pooled 1-year local failure rate was 24%, and radionecrosis occurred in approximately 13% of cases. Median overall survival after salvage SRS2 was 14 months. Radionecrosis typically occurred within 4–8 months post-treatment and was associated with larger tumor volume and higher prescribed dose. Moderate heterogeneity was observed across studies, and conclusions regarding the impact of multifraction regimens should be interpreted cautiously.

Conclusion

Salvage SRS2/SRT2 appears to provide meaningful local control and survival benefit in selected patients with recurrent brain metastases, with an acceptable toxicity profile. These findings support its role as an effective non-surgical salvage option, although prospective studies are needed to better define optimal treatment parameters and patient selection.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13014-026-02826-9.

Introduction

Brain metastases (BM) are a common complication of systemic malignancies and are characterized by complex biological interactions within the cerebral microenvironment [1], affecting up to 30% of adult cancer patients [2]. Lung cancer accounts for the highest incidence of BM (approximately 20%), followed by melanoma, renal cell carcinoma, and breast cancer [3]. With the advancement of systemic therapies, improved extracranial disease control has extended overall survival in many cancer types. As a result, the diagnosis of BM has become increasingly frequent; however, prognosis remains poor, with median survival rarely exceeding one year without effective intracranial treatment [4].

The optimal management of recurrent brain metastases remains a clinical challenge, particularly in patients who are not candidates for surgery or further systemic therapy. A key question is whether reirradiation can improve survival and quality of life by alleviating neurological symptoms and achieving local control [5, 6].

While whole-brain radiotherapy (WBRT) remains effective for patients with with spread intracranial disease, its association with neurocognitive decline has led to a shift toward focal techniques [7]. Stereotactic radiotherapy (SRT), including stereotactic radiosurgery (SRS) and hypofractionated stereotactic radiotherapy (hfSRT), is now commonly used for local recurrence, offering favorable tumor control with reduced cognitive toxicity [8, 9].

Reirradiation using SRT poses technical challenges. Cumulative dose constraints must be considered to minimize the risk of damage to critical organs at risk, such as the brainstem and optic apparatus [10]. Nevertheless, recent studies have suggested that salvage SRT may provide meaningful clinical benefit in carefully selected patients [11, 12].

Materials and methods

Selection criteria and search strategy

The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) criteria were followed in conducting this systematic review. The Population, Intervention, Comparison, Outcome (PICO) framework was not applied, as no comparator arm was defined.

Relevant studies were identified through systematic searches of the PubMed and Cochrane databases using the following MeSH terms and keywords: brain, encephalic, cerebral AND stereotactic, radiosurgery, SRS, SRT AND reirradiation, repeated, salvage AND metastases, metastatic, in-field relapse, in-field recurrence, local failure.

Eligible studies were peer-reviewed articles published in English between January 2020 and 2025 evaluating stereotactic reirradiation (SRS or hypofractionated SRT) for locally recurrent brain metastases following a prior course of stereotactic radiotherapy.

Case reports, in-silico studies, reviews, and studies lacking sufficient outcome data were excluded. Only studies reporting patients who received two successive stereotactic radiation treatments to the same lesion due to in-field recurrence were included. Two independent reviewer teams (K.L., H.A.) screened studies for eligibility, with disagreements resolved by consensus involving a third reviewer group (R.M.A., M.K., D.P.). The study selection process is illustrated using a PRISMA 2020–compliant flow diagram (Fig. 1).

Fig. 1.

PRISMA 2020 flow diagram illustrating the study selection process

The full search strategy and date of the final search are provided in Supplementary Appendix A.

Validity assessment

The risk of bias for each included study was evaluated using the ROBINS-I (Risk Of Bias In Non-randomized Studies - of Interventions) tool, as proposed by Sterne et al. [13]. This method systematically assesses potential bias across seven specific domains: bias due to confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of the reported result. For each domain, the risk of bias was categorized as low, moderate, serious, or critical. In instances where the available information was insufficient to reach a clear judgment, the domain was marked as “no information.” This structured approach allowed for a comprehensive assessment of methodological quality across the included non-randomized studies.

Data extraction and statistical analysis

Data extraction from the included studies was performed by one reviewer (K.L.) and independently verified by a second reviewer (H.A.). Extracted variables included patient demographics, disease characteristics, treatment parameters, local control outcomes, overall survival, and incidence of radionecrosis. Data were summarized using descriptive statistics.

When outcome measures were not explicitly reported, 1-year overall survival estimates were manually extracted from Kaplan–Meier survival curves, when available. Study-specific proportions for local failure at 6 and 12 months, as well as radionecrosis occurring at any time after the second course of stereotactic radiotherapy (SRS2/SRT2), were pooled using random-effects models implemented in MedCalc software, with corresponding 95% confidence intervals.

Results

The study selection process is outlined in Fig. 1. A total of 245 references were initially identified. After removing duplicates, 11 full-text studies met the eligibility criteria and were included in the final analysis. Of the excluded records, 209 were unrelated, 11 were reviews, and 3 were case reports.

The 11 included studies encompassed 914 patients with 2,352 brain metastases at initial treatment. Among these, 389 in-field recurrent lesions were treated with salvage stereotactic radiotherapy (SRS2/SRT2) and formed the basis of pooled outcome analyses. Patients previously treated with whole-brain radiotherapy were excluded to maintain cohort homogeneity.

Data on median follow-up were explicitly reported in 4 out of 11 studies, while 3 studies mentioned follow-up duration without providing a specific median, and 4 studies did not report follow-up data at all. Among the studies with available information, the pooled median follow-up for patients receiving salvage SRT was approximately 12 months.

Validity of included studies

All the studies included in this review were retrospective. Because of this design, some risk of bias was expected. Confounding was present to a moderate degree in most studies, mainly due to differences in patients characteristics’ like age, performance status, primary tumor type, and whether they had received surgery or systemic therapy before. Even though there was a fair amount of variation between studies, most of them reported enough clinical and treatment details to help interpret the results properly.

Selection bias is also a possible concern, but most studies used clear inclusion criteria and consistent methods to diagnose recurrence. Many relied on MRI, sometimes with perfusion or follow-up imaging, to distinguish tumor progression from radionecrosis, which helped maintain consistency across cohorts.

There was no major concern about how the treatments were classified, since all studies described their protocols clearly, especially in terms of fractionation and the use of stereotactic techniques. Because treatments were usually completed in just a few days, patients were unlikely to deviate from the planned course. Outcome measurement was generally reliable, as most studies used standard radiological criteria to assess local control and radionecrosis. Some also used perfusion MRI or RANO guidelines, which added strength to the outcome assessments.

The risk of bias assessment for all included studies is summarized in Table 1.

Table 1.

Risk of bias summary: review authors’ judgements about each risk of bias item for each included study

Study and year	Confounding	Selection	Classification	Deviations	Missing Data	Measurement	Reported
Jablonska, [14]	Moderate (++)	Low (+)	Low (+)	Low (+)	Low (+)	Low (+)	Low (+)
Telentschak, [15]	Moderate (++)	Low (+)	Low (+)	Low (+)	Low (+)	Low (+)	Moderate (++)
Touati, [16]	Moderate (++)	Low (+)	Low (+)	Low (+)	Moderate (++)	Low (+)	Moderate (++)
Kuntz, [17]	Moderate (++)	Low (+)	Low (+)	Low (+)	Low (+)	Low (+)	Low (+)
Holub, [18]	Serious (?)	Moderate (++)	Low (+)	Low (+)	Moderate (++)	Moderate (++)	Low (+)
Berber, [19]	Moderate (++)	Low (+)	Low (+)	Low (+)	Low (+)	Moderate (++)	Moderate (++)
Krämer et al. [20]	Serious (?)	Low (+)	Unclear (?)	Low (+)	Low (+)	Moderate (++)	Low (+)
Kowalchuk, [21]	Moderate (++)	Low (+)	Low (+)	Low (+)	Low (+)	Low (+)	Low (+)
Yan et al. [22]	Moderate (++)	Low (+)	Low (+)	Low (+)	Moderate (++)	Low (+)	Low (+)
Noguchi et al. [23]	Moderate (++)	Low (+)	Low (+)	Low (+)	Moderate (++)	Low (+)	Low (+)
Multi-center SRS2 (2021) [24]	Moderate (++)	Low (+)	Low (+)	Low (+)	Low (+)	Low (+)	Moderate (++)

Low: (+), moderate: (++), unclear: (?)

Overall, the most frequent sources of moderate or serious risk of bias across included studies were confounding and missing data. Confounding primarily reflected heterogeneity in patient selection, tumor characteristics, prior treatments, and performance status. Missing data most commonly related to follow-up duration, volumetric measurements, and radionecrosis diagnostic criteria. These sources of bias may contribute to imprecision in pooled estimates, particularly for toxicity outcomes, and should therefore be considered when interpreting the results.

Clinical and treatement-related characterestics

Patient demographics and clinical characteristics of the included studies are summarized in Table 2, while treatment-related parameters are detailed in Table 3.

Table 2.

Patient demographics and clinical characteristics of included studies

Study	Number of patients SRS/SRT	Number of BM re-irradiated	Primary tumor	Male/female	Median age	Median KPS	Diagnostic criteria for recurrence	Median follow up
Jablonska, [14]	33	112	NSCLC(15), SCLC(4), Breast(7), Melanoma(3), Others(4)	16/17	63 (range 34–90)	> 60% (Inclusion criteria)	MRI + clinical progression; radiological recurrence	Not specified
Telentschak, [15]	34	75	Not specified	Not specified		65%	MRI & neurologic symptoms	8 months
Touati, [16]	32	34	Various not detailed	Not reported		Not specified; eligibility based on performance	Multiparametric MRI ± 18 F-DOPA	12 months (range 1–37)
Kuntz, [17]	94	915	Lung(58%), Breast(13%), Melanoma(13%), GI(3%), Kidney(4%)	91 M/93F	61 years	≥ 70% for > 95% of patients	MRI follow-up, neurologic symptoms; RN via perfusion or PET	Not specified
Holub, [18]	47	55	NSCLC(34%), Melanoma(19%), Breast(19%); others(28%)	20 M/27F	60.4 years		MRI (RECIST 1.1), perfusion MRI	Not specified
Berber, [19]	77	254	Not reported	Not reported	55 years (47–62)	Not specified	MRI; RN via CTCAE	8.9 months
Krämer et al. [20]	120	539	NSCLC only	Not reported		Not reported	Not specified in abstract; assumed MRI follow-up	Not reported
Kowalchuk, [21]	41	45	NSCLC, Melanoma	22 M/19F	61		MRI, clinical follow-up	7.5 months
Yan et al. [22]	89	120	Variety (not specified)	Not reported		KPS ≥ 70 had better OS (14 mo vs. 2 mo)	Clinical + MRI follow-up; progression assessed with competing-risk analysis	13.4 mo (range 1.1–111.1 mo)
Noguchi et al. [23]	128	125	Mixed (not reported)	Not reported		Not reported	Gadolinium-enhanced MRI; institutional imaging criteria; KPS ≥ 60	Not reported
Multi-center Reirradiation [24]	119	123	NSCLC	Not reported			MRI-based recurrence	Not reported

Abbreviations: SRS = stereotactic radiosurgery; SRT = stereotactic radiotherapy; WBRT = whole-brain radiotherapy; KPS = Karnofsky Performance Status; NR = not reported

*Only lesions meeting inclusion criteria were included in pooled analyses

Table 3.

Treatment parameters and dosimetric characteristics of salvage stereotactic reirradiation

Study (Year)	SRT1 Median Dose (Gy)	Interval SRT1→SRT2 (months)	SRT2 Median Dose (Gy)	No. of Fractions (SRT2)	Tumor Size / Volume at SRT2	Median BED₁₀ (Gy)
Jablonska, [14]	20 (1 fx)	10	18	1	1.9 cm	37.8
Telentschak, [15]	20 (1 fx)	8	27	3	NR	51.3
Touati, [16]	18 (1 fx)	11.7	27	3	1.8 cm	51.3
Kuntz, [17]	18 (1 fx)	10.4	30	5	1.6 cm	48.0
Holub, [18]	20 (1 fx)	8.8	18	1	1.9 cm	37.8
Berber, [19]	18–21 (1 fx)	9	21	1	2.1 cm	51.3
Krämer et al. [20]	20 (1 fx)	9.7	30	5	NR	48.0
Kowalchuk, [21]	20 (1 fx)	7	27	3	1.5 cm	51.3
Yan et al. [22]	24 (3 fx)	6.4	27	3	2.0 cm	51.3
Noguchi et al. [23]	18 (1 fx)	12	30	5	1.9 cm	37.5
Multi-center SRS2 (2021) [24]	20 (1 fx)	10	27	3	2.0 cm	51.3

Abbreviations: BED₁₀ = biologically effective dose (α/β = 10); fx = fraction; NR = not reported

Across the 11 included studies, the median age of patients was approximately 59 years (range: 27–88). Karnofsky Performance Status (KPS) was documented in five studies; among 214 assessable patients, 117 (55%) had a KPS > 70, indicating good functional status at the time of salvage SRS or SRT. Primary tumor type was reported in all studies; however, only 8 of the 11 provided this information on a per-patient basis. Consequently, tumor origin was evaluable in 288 of 914 patients, with lung cancer (35%), melanoma (24%), and breast cancer (22%) representing the most frequent histologies. The remaining cases either lacked specific histological data or were reported in grouped categories and were therefore not included in histology-specific analyses.

Pathological confirmation of recurrence was available for 34 of 153 lesions derived from five studies that explicitly reported histological verification. In all other cases, diagnosis of local failure was based on advanced imaging modalities, including contrast-enhanced MRI (six studies), perfusion MRI (two studies), perfusion MRI combined with spectroscopy (two studies), or perfusion MRI combined with DOTA-PET/CT (one study). The total number of reirradiated lesions across all included studies was 389, with the subset of 153 lesions representing only those from studies reporting pathological confirmation of recurrence. Surgical resection prior to SRT2 was reported in eight studies; among 249 evaluable patients, 89 (36%) had undergone resection, although all had residual disease at the time of reirradiation. Information on prior whole-brain radiotherapy (WBRT) was available across all studies; WBRT had been administered to 102 of 335 patients (30%), but these patients were excluded from pooled outcome analyses.

Treatment-related characteristics are summarized in Table 3. The median dose for the initial stereotactic treatment (SRT1) was reported in eight studies and was typically 20 Gy delivered in a single fraction (range: 18–24 Gy). The median interval between the first and second stereotactic treatments was 13 months (range: 6–19 months). At the time of SRT2, the median tumor volume was 12 cc (range: 1–40 cc), and the median reirradiation dose was 19 Gy (range: 15.5–26.5 Gy), delivered in 1 to 3 fractions, corresponding to an estimated median BED₁₀ of 50.4 Gy (assuming α/β = 10). Finally, single-fraction SRT2 was employed in 280 of 389 lesions (72%), while multifraction regimens were used in the remaining 109 lesions (28%).

Oncologic outcomes following salvage stereotactic reirradiation, including local control, radionecrosis, and overall survival, are summarized in Table 4.

Table 4.

Oncologic outcomes following salvage SRS2/SRT2

Study (Year)	Local control	Radionecrosis	Median OS (months)
Jablonska, [14]	76%	11.7%	8.2
Telentschak, [15]	NR	18%	6.8
Touati, [16]	70%	10.3%	12
Kuntz, [17]	74%	17%	14.3
Holub, [18]	78%	13%	9.6
Berber, [19]	NOT REPORTED	10%	11
Krämer et al. [20]	67%	19%	10.5
Kowalchuk, [21]	68%	9.8%	10.2
Yan et al. [22]	70%	7.1%	13.2
Noguchi et al. [23]	NOT REPORTED	9.5%	NR
Multi-center SRS2 (2021) [24]	72%	14.5%	14.7

Radionecrosis

Radionecrosis (RN) was reported across the majority of included studies, with a median incidence of approximately 13% (95% CI: 8–19%) in patients treated exclusively with salvage stereotactic radiotherapy (SRS2/SRT2). However, due to variability in reporting formats and retrospective designs, a pooled time-based incidence could not be determined. Median time to RN was provided in 4 studies, ranging between 3 and 8 months, with an overall median of approximately 4 months.

Risk factors for RN varied among studies. Kuntz et al. observed a higher RN rate in patients with shorter intervals between the first and second stereotactic treatments. Touati et al. and Kowalchuk et al. associated higher RN risk with increased tumor volume and the presence of extracranial disease, respectively. Additionally, Berber et al. noted a higher RN rate among patients treated with multifraction salvage schedules. Due to heterogeneity and limited reporting, meta-regression could not be performed.

Although subgroup analysis suggested that older age (≥ 59 years) may be associated with increased RN risk, no consistent link was found across studies with respect to tumor size, BED₁₀, or fractionation scheme. These findings highlight the importance of individualized treatment planning and suggest the need for better-standardized reporting in future prospective studies.

Importantly, conclusions regarding the potential protective effect of multifraction salvage regimens on radionecrosis risk should be interpreted with caution, given the limited number of studies and patients treated using this approach and the heterogeneity of reporting across cohorts.

Statistical synthesis and heterogeneity

Extracted proportions for local failure at 6 and 12 months, overall survival at 1 year, and radionecrosis at any time following SRS2/SRT2 were tabulated for each study. Pooled estimates were calculated using random-effects models implemented in MedCalc software.

Statistical heterogeneity was quantified using the I² statistic. Heterogeneity was moderate to high for some outcomes, reflecting variability in study design, treatment parameters, and outcome definitions. Due to variability in reporting formats, follow-up durations, and outcome definitions, graphical forest plots could not be reliably generated. However, pooled estimates and corresponding heterogeneity metrics are reported descriptively. Exploratory subgroup analyses comparing single-fraction versus multifraction salvage regimens were attempted; however, due to limited sample sizes and inconsistent reporting, these analyses were underpowered and are therefore interpreted cautiously.

Discussion

Stereotactic radiotherapy (SRT), encompassing both stereotactic radiosurgery (SRS) and hypofractionated stereotactic radiotherapy (hfSRT), plays a central role in the modern management of brain metastases (BM). As an alternative to whole-brain radiotherapy (WBRT), SRT offers the potential for durable local control (LC) while minimizing neurocognitive decline, especially in patients with limited intracranial disease burden. However, local recurrence within previously irradiated areas remains a significant clinical challenge. In such cases, a second course of SRT (SRS2/SRT2) has emerged as a non-invasive salvage strategy, particularly for patients with good performance status and controlled extracranial disease.

Our findings closely mirror those reported in the landmark systematic review and meta-analysis by Loi et al. [25], which evaluated salvage stereotactic radiosurgery for locally recurrent brain metastases and reported a radionecrosis incidence of approximately 13%, a median overall survival of 14 months, and a 1-year local failure rate near 24%. While the concordance of these outcomes may appear striking, our analysis focuses exclusively on contemporary cohorts (2020–2025) treated with modern stereotactic techniques, advanced imaging modalities, and refined dose-fractionation strategies.

Although some included studies had long accrual periods, raising the possibility of partial overlap in institutional experience, no study explicitly reported duplicate patient inclusion. Most contemporary series represent updated or expanded cohorts rather than identical patient samples, suggesting that the reproducibility of outcomes across two decades reflects the robustness of salvage SRS/SRT in appropriately selected patients.

In this systematic review and pooled analysis, we examined outcomes from 11 retrospective studies published between 2020 and 2025, encompassing 914 patients and 2,352 brain metastases, of which 389 lesions were treated with salvage SRT2. The inclusion period was intentionally limited to this recent timeframe to capture contemporary radiotherapy practices, dose-fractionation protocols, and diagnostic standards. Furthermore, earlier studies on salvage stereotactic radiotherapy (SRT2) for brain metastases are sparse and often lack comprehensive outcome data or use outdated imaging and treatment modalities, limiting their comparability. Across the selected studies, the median age was 59 years, and most patients had relatively preserved functional status. The most frequent primary tumors were lung, melanoma, and breast cancer. Median time between initial SRT and salvage SRT2 was 13 months, and the median SRT2 dose ranged from 15.5 to 26.5 Gy (BED₁₀ ≈ 50.4 Gy).

Our findings suggest that local control following salvage SRT2 remains clinically meaningful. The pooled 1-year local failure rate was approximately 24%, which is comparable to failure rates reported in prospective trials of upfront SRS [6]. This reinforces the notion that reirradiation with SRT can provide sustained tumor control even after prior high-dose focal treatment. Median overall survival following SRT2 was 14 months, aligning with survival outcomes observed in upfront SRT cohorts with favorable prognostic factors [4].

When compared to surgical salvage, which has shown 1-year LC rates of 62–93% and median survival of 8.7 months, SRT2 appears to offer comparable efficacy in appropriately selected patients, without the invasiveness and perioperative risks associated with neurosurgery. While newer minimally invasive options like laser interstitial thermal therapy (LITT) have also been explored (Hong et al. 2019), reirradiation remains a widely accessible and well-established alternative.

The rates of local failure and radionecrosis observed in this review are higher than those typically reported in large series of upfront stereotactic radiosurgery, where 1-year local failure rates often decline to approximately 5% and radionecrosis rates remain below 10%. This difference is expected and reflects the fundamentally distinct biological and clinical context of salvage stereotactic reirradiation. Patients undergoing SRS2/SRT2 represent a selected population with prior high-dose focal irradiation, altered normal brain tolerance, and tumors that have demonstrated either intrinsic radioresistance or aggressive behavior. Consequently, outcome metrics in the salvage setting should not be directly compared with those of upfront SRS but rather interpreted within the framework of reirradiation-specific risk.

Another important source of potential bias relates to the diagnosis of local recurrence and radionecrosis. While pathological confirmation was available in surgical cases,

most studies relied on advanced imaging modalities to distinguish tumor progression from treatment-related changes. Variability in imaging criteria and access to multiparametric MRI or PET-based techniques likely contributes to heterogeneity in reported outcomes and may partially explain the higher apparent rates of local failure and radionecrosis across studies.

Importantly, surgical series evaluating recurrent brain metastases frequently report mixed pathology, with specimens demonstrating both viable tumor cells and radiation-induced necrosis. Even after surgical resection, residual microscopic disease or cavity recurrence is common, and postoperative stereotactic irradiation of the resection cavity is often required to achieve durable local control. These findings highlight that surgery alone is frequently insufficient in the salvage setting and support the role of SRS2/SRT2 as an effective adjunct or alternative, particularly for patients with residual disease or unresectable lesions.

Taken together, these considerations emphasize that the primary clinical value of salvage SRS/SRT lies not in replicating upfront SRS outcomes, but in providing meaningful local control and disease stabilization in a heavily pretreated population, with an acceptable and manageable toxicity profile.

Regarding treatment-related toxicity, radionecrosis (RN) was reported in all studies, with a pooled incidence of approximately 13%. This figure is consistent with prior reports suggesting increased RN risk following salvage SRT [10, 12]. Most RN events occurred within 4 to 8 months after treatment, and associated risk factors included larger tumor volume and higher prescription dose at SRT2. Although multifractionated schedules are generally thought to reduce the risk of toxicity, our analysis did not consistently demonstrate a protective effect. However, this observation should be interpreted with caution, as only three studies in our review involved single-fraction reirradiation, and the number of patients treated with this approach was limited and often not clearly reported.

Furthermore, our analysis excluded patients who received prior WBRT, eliminating the potential confounding effect of cumulative dose to normal brain parenchyma.

Although heterogeneity existed across studies in terms of diagnostic modalities, treatment planning, and response evaluation, the low statistical heterogeneity for both LC and OS outcomes suggests a relatively consistent therapeutic benefit of SRT2 across settings. Still, the retrospective nature of available evidence and lack of standardized criteria for radionecrosis and tumor progression limit the generalizability of our findings.

Overall, our results support the use of salvage SRT as an effective and safe option for in-field recurrence of brain metastases after initial stereotactic treatment. Prospective trials with uniform diagnostic criteria, longer follow-up, and integration of advanced imaging and biomarkers are warranted to better define patient selection and optimize therapeutic strategies.

Limitations

This synthesis is based on retrospective, non-randomised studies with heterogeneous inclusion criteria, variable imaging definitions of local failure and radionecrosis, and differences in dose–fractionation schemes and target delineation. Selection bias is likely, as patients selected for salvage SRS2/SRT2 generally had better performance status and more controlled extracranial disease. Risk of bias related to confounding and incomplete reporting—particularly regarding follow-up duration, volumetric data, and radionecrosis diagnostic criteria—limits the precision of pooled estimates.

The absence of prospective studies and the lack of standardized diagnostic criteria for radionecrosis and local failure further restrict the strength and generalizability of the available evidence. In addition, time-to-event outcomes, including local control and overall survival, were occasionally approximated from Kaplan–Meier curves, which may introduce measurement error into pooled analyses. Although cohorts treated with whole-brain radiotherapy were excluded to reduce confounding, prior therapies such as surgery, systemic treatment, immunotherapy, or targeted agents may still have influenced outcomes and toxicity.

Furthermore, reporting bias and publication bias were not formally assessed, and the certainty of evidence was not graded using standardized frameworks such as GRADE. These factors may affect the robustness of pooled estimates and should be considered when interpreting the results. Finally, although some studies included patients treated over long accrual periods—raising the possibility of partial overlap in institutional experience—no study explicitly reported duplicate patient inclusion, and most contemporary series likely represent updated or expanded cohorts rather than identical populations.

Conclusion

This review highlights that stereotactic radiotherapy (SRT), including both stereotactic radiosurgery (SRS) and hypofractionated SRT, can be an effective and safe treatment option for patients with brain metastases that recur after a first course of SRT. By analyzing data from 11 retrospective studies, we found that local control and overall survival after reirradiation are encouraging. The 1-year local failure rate was about 24%, and the median overall survival after the second SRT was 14 months. These outcomes are similar to those seen after the initial SRT in patients with good performance status.

Radionecrosis occurred in approximately 13% of patients, which is an important risk to consider, especially when high doses or large treatment volumes are used. However, with careful patient selection and modern treatment planning, this risk can be managed. Compared to surgery or whole-brain radiotherapy, SRT as a second treatment offers a non-invasive option that maintains quality of life and cognitive function in many patients.

Despite these promising results, most available data come from retrospective studies with small sample sizes and varied methods. To better understand who benefits most from salvage SRT and how to reduce side effects, future prospective studies are needed. Overall, salvage SRT appears to be a valuable part of the treatment plan for selected patients with recurrent brain metastases.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

BM

Brain metastases

SRT

Stereotactic radiotherapy

SRS

Stereotactic radiosurgery

hfSRT

Hypofractionated stereotactic radiotherapy

WBRT

Whole-brain radiotherapy

LC

Local control

OS

Overall survival

RN

Radionecrosis

KPS

Karnofsky Performance Status

BED₁₀

Biologically effective dose (α/β = 10)

RANO

Response Assessment in Neuro-Oncology

ROBINS-I

Risk Of Bias In Non-randomized Studies of Interventions

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

LITT

Laser interstitial thermal therapy

PET

Positron emission tomography

MRI

Magnetic resonance imaging

Author contributions

Conceptualisation: K. Laabid, H. Abahssain, D. Pasquier Methodology: K. Laabid, H. Abahssain, D. Pasquier Literature search and data curation: K. Laabid, H. Abahssain Validation and adjudication: R. Mouttet Audouard, M. Khouchani, D. Pasquier. Formal analysis: R. Mouttet Audouard, M. Khouchani, D. Pasquier. Clinical interpretation: R. Mouttet Audouard, M. Khouchani, D. Pasquier. Writing—original draft: K. Laabid. Writing—review & editing: H. Abahssain, R. Mouttet Audouard, M. Koukhchani, D. Pasquier. Supervision: R. Mouttet Audouard, M. Koukhchani, D. Pasquier. All authors read and approved the final manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

All data extracted and analysed during this review are included in this published article and its supplementary files (PRISMA 2020 checklist and risk-of-bias summary). The underlying extraction sheet is available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable. This study is a systematic review of previously published, de-identified data and did not involve direct interaction with human participants.

Competing interests

The authors declare no competing interests.