Abstract
Background/Aim: Stereotactic radiotherapy (SRT), including stereotactic radiosurgery (SRS) and fractionated stereotactic radiotherapy (FSRT), is the current standard adjuvant treatment after resection of brain metastases (BM). In the era of immunotherapy and targeted systemic therapies enabling effective extracranial control, achieving durable intracranial outcomes has become increasingly important. Intraoperative radiotherapy (IORT) is an emerging alternative to Gamma Knife-based irradiation. We aimed to assess the clinical outcome of BM resection cavities treated with Gamma Knife SRT (GK-SRT), for retrospective comparison to a cohort treated with IORT.
Patients and Methods: This retrospective single-center analysis included all patients who received GK-SRT to the resection cavity after complete surgical removal of BM. A total of 41 cavities from 37 patients were evaluated. Local control (LC), distant intracranial control (DIC), and overall survival (OS) were calculated, and the potential influence of whole-brain radiotherapy (WBRT) as part of the adjuvant treatment on OS was analyzed.
Results: With a median follow-up of 19.3 months, the 1-year LC was 93.3%. The 1-year DIC was 54.1%, and the median OS was 27.3 months. Median cavity volume was 5.89 cm3; the median interval from surgery to GK-SRT initiation was 24.5 days, and to subsequent systemic treatment 59.0 days. DIC, but not LC, was significantly associated with OS, with better DIC correlating with longer survival. WBRT as part of the adjuvant regimen did not confer a significant survival benefit.
Conclusion: Adjuvant GK-SRT to the resection cavity is an effective treatment with excellent local control. Nonetheless, challenges remain regarding distant intracranial progression and overall survival.
- Brain metastases
- Gamma Knife
- resection cavity
- stereotactic radiosurgery
- fractionated stereotactic radiotherapy
Introduction
Brain metastases (BM) represent a significant challenge in neuro-oncology, with up to 40% of cancer patients developing BM over the course of their disease (1). Advanced diagnostic techniques and new systemic therapies are prolonging overall survival (OS) but are also increasing the incidence of BM (2, 3). Although BM may not directly impact overall survival (4), effective local treatment with long-lasting tumor control is crucial to prevent neurological deterioration and maintain quality of life (5).
Current evidence is supporting local radiotherapy of the BM resection cavity (6, 7), as even after complete surgical resection, the recurrence rate without any adjuvant radiation therapy (RT) is approximately 50% (8). Therefore, the standard of care for patients with BM involves surgical resection followed by adjuvant RT, particularly Stereotactic Radiotherapy (SRT), either as Stereotactic Radiosurgery (SRS) or Fractionated Stereotactic Radiation Therapy (FSRT), which deliver high doses of radiation to the resection cavity, while minimizing radiation exposure to the surrounding healthy parenchyma (9).
Currently, Intraoperative Radiotherapy (IORT) has emerged as a potential alternative for the cavity treatment of BM. Prospective data have been published by our group, promising excellent control with a favorable safety profile (10-13). Through a retrospective study of a cohort managed by the same team of neurosurgeons, neuroradiologists and radiation oncologists as the INTRAMET clinical trial (NCT03226483), we aimed to evaluate the local control (LC) of BM resection cavities treated via Gamma Knife SRT (GK-SRT).
Patients and Methods
Study design and patient selection. We examined all resection cavities of patients with BM treated with GK-SRT at the Department of Radiation Oncology of the University Medical Centre Mannheim from March 2017 through January 2022, parallel to the INTRAMET trial. In total, 51 brain metastases were screened. Inclusion criteria comprised age ≥18 years, Karnofsky Performance Status (KPS) ≥50, prior surgery with complete surgical resection of BM as confirmed by contrast-enhanced T1-weighted magnetic resonance imaging (MRI), histologically confirmed BM via frozen section, and subsequent GK-SRT treatment within the resected cavity. Exclusion criteria included the inability to complete the planned GK-SRT treatment, as well as intra-cavity or regional progression observed before or during the GK-SRT. The final number of included postoperative cavities was 41, from 37 patients. For analyzing long-term follow-up variables such as LC and distant intracranial control (DIC), exclusion criteria involved the absence of MRI follow-up. Thus, for the analysis of these endpoints, 36 postoperative cavities from 32 patients were included. Local institutional review board approval (approval number 2018-882R-MA, Ethikkommission II, Mannheim Medical Faculty, Heidelberg University) was obtained before data collection.
MRI follow-up protocol and data analysis. After complete surgical resection of the BM, all patients underwent MRI scans within 72 hours postoperatively. Subsequently, patients underwent quarterly imaging follow-up. MRI reports for each case from the initial postoperative MRI to the last imaging follow-up available were comprehensively analyzed, recording local progression and regional progression. The termination of follow-up was determined by the date of the last MRI scan. Additionally, from the radiotherapy planning MRI scan, we calculated the cavity volume (cm3) and the number of remaining additional brain metastases. To standardize the GK-SRT approach, considering the variability in dose prescription and fractionated schemes applied, we calculated the equivalent dose in 2 Gy fractions (EQD2) for each resection cavity, incorporating the radiation dosage from both SRT and whole-brain radiotherapy (WBRT), if WBRT was included in the adjuvant treatment plan. EQD2 was computed based on the linear-quadratic cell survival model considering an α/β ratio of 6 [EQD2=n×D×((d+6)/(2+6))]. Finally, for each patient, we performed an individual analysis using the electronic patient database of University Medical Centre Mannheim, to determine: sex, age at treatment, KPS (assessed at baseline visit), tumor histopathology, time from surgical procedure to GK-SRT onset date, administration of WBRT either as part of the adjuvant treatment or as salvage treatment, initiation and time interval between surgery and onset of systemic treatment, type of salvage treatment prescribed, and date of death.
SRT treatment. Patients were treated with the Leksell Gamma Knife Icon (Elekta AB, Stockholm, Sweden), either with frame fixation (FF) or thermoplastic mask fixation (MF). For FF, the planning target volume (PTV) consisted of the cavity and any adjacent T1 contrast-enhancing lesion, whereas for MF, an additional margin of 1 mm was added to these structures to define the PTV. Treatment plans were optimized for conformity and coverage using Leksell GammaPlan (Elekta AB, Stockholm, Sweden). If WBRT was part of the adjuvant treatment, it was delivered using a linear accelerator (Elekta Synergy, Elekta AB) as 3D-conformal radiotherapy.
Primary and secondary endpoints. The primary endpoint was the 1-year and median LC, defined as the absence of any marginal MRI-enhancing lesion suspicious for recurrence within the resection cavity, according to the RANO criteria, identified during MRI follow-up scans. LC was calculated from the date of surgical resection.
Secondary endpoints included the 1-year and median DIC, defined as the absence of new enhancing lesions in the cerebral parenchyma suggestive of new brain metastases, according to the RANO criteria, identified during MRI follow-up scans by a board-certified neuroradiologist and subsequent confirmation in a multidisciplinary tumor panel. DIC was also calculated from the date of surgical resection. Further secondary endpoints were: 1-year and median overall survival (OS), defined as the time interval between the date of surgery and the date of death; the influence of LC and DIC on OS; time from surgical resection to initiation of GK-SRT; the impact of WBRT as part of the adjuvant treatment plan on OS; and time from surgery to the initiation of systemic treatment. The outcomes (LC, DIC, and OS) were calculated up to the last patient at risk. The maximum follow-up time was defined by the patient with the longest observation period.
Statistical analysis. Patient and tumor characteristics were summarized using standard descriptive statistics. Survival outcomes were estimated with the Kaplan-Meier method and compared with the log-rank test; 95% confidence intervals (CIs) were calculated. LC was analyzed in 36 resection cavities, and DIC in 32 patients with MRI follow-up. OS analysis included 29 patients with survival status data. Censoring was applied at the date of last imaging follow-up, initiation of salvage WBRT, or death; for OS, patients alive at the data cutoff were censored. Associations of LC, DIC, and OS with clinical factors were assessed with univariable Cox proportional-hazards models; for OS, we additionally fitted Cox models with time-dependent covariates. Results are reported as hazard ratios (HRs) with 95% CI. Proportional-hazards assumptions were evaluated using Schoenfeld residuals.
Statistical analyses were conducted using SPSS v.28.0 (IBM Corp., Armonk, NY, USA) and R, version 4.5.1, making use of several key packages: readxl for importing Excel files, dplyr for data manipulation, tibble for table formatting, and stringr for text handling. Survival analyses, including Kaplan-Meier estimates and Cox proportional-hazards models, were conducted with the survival package, while survminer was used for visualization of Kaplan-Meier curves. The graphical output was based on the ggplot2 engine. Group comparisons of overall survival were performed using the exact permutation log-rank test implemented in the coin package. The statistical significance level was defined as p<0.05.
Results
Patients and tumor characteristics. Of the total 41 BM cases, the majority occurred in female patients (56.1%). The median age at treatment was 65 years (range=38-82), and the median Karnofsky Performance Status (KPS) was 80 (range=50-100), with over 60% of patients having a KPS ≥80. Most histopathology results corresponded to non-small-cell lung cancer (51.2%). Multiple metastases at the time of GK-SRT were present in 48.8% of cases, with 1 to 4 additional intracranial lesions.
The median cavity volume was 5.89 cm3 (SD±5.4), and the median prescribed dose to the resection cavity was 59.95 Gy EQD2 (SD±13.03). The median follow-up period from surgical resection to the last MRI scan was 19.3 months (range=3.1-94.2). At the time of data collection (June 2025), 5 patients remained under ongoing follow-up. Systemic treatment was administered to 18 of the 32 MRI-followed patients (56%), with a median interval from surgery to systemic treatment of 59.0 days (range=15-102). Patient and tumor characteristics are summarized in Table I.
Patient, tumor and Gamma Knife-SRT characteristics.
Radiotherapy schemes included SRS alone, SRS boost in combination with WBRT, FSRT in 3-7 fractions, and FSRT boost in 2 fractions combined with WBRT (Table II). The median time from surgical resection to GK-SRT initiation was 24.5 days (range=13-59). Salvage treatment was prescribed for 20 of the 32 MRI-followed patients (62.5%), including salvage SRT, salvage WBRT, and salvage surgery (Table III). Salvage WBRT was conducted in 7 out of the 32 patients (21.9%) with MRI follow-up, with a median time from surgery to onset of salvage WBRT of 190 days (range=90-748).
Characteristics of adjuvant resection cavity radiotherapy.
Salvage therapies prescribed for local and regional progression.
Primary and secondary endpoints. The 1-year LC was 93.3% (95% CI=0.85-1.00), while the median time to local failure was not reached. Three local recurrences (in 3 patients) were observed at 5.8, 6.1, and 41.1 months after surgery (Figure 1). No statistical differences were identified regarding cavity volume (HR=0.93; 95% CI=0.70-1.24; p=0.639) or dose prescribed to the cavity (HR=1.11; 95% CI=0.97-1.28; p=0.135). In addition, neither the time from surgery to initiation of GK-SRT (HR=1.04; 95% CI=0.96-1.13; p=0.367) nor the administration of systemic treatment (HR=0.59; 95% CI=0.05-6.55; p=0.669) showed statistical differences between patients with and without local recurrence. However, we observed a longer waiting time to start GK-SRT in the subgroup with local progression (35.3 days, SD±20.8) compared to the subgroup without local progression (28.0 days, SD±11.4). The three patients experiencing local progression received salvage treatment; one patient underwent multiple salvage SRS sessions, and 2 patients received salvage WBRT. The median time from local progression to initiation of salvage treatment was 15.0 days (range=5-55). No significant differences in LC were observed by tumor histopathology when comparing non-small-cell lung cancer (NSCLC; n=17 cases) with breast cancer (n=5 cases; HR=3.98; 95% CI=0.24-65.30; p=0.333) or with other malignancies (n=14 cases; HR=1.72; 95% CI=0.10-28.50; p=0.707). The PH assumption was met for all LC models. The potential impact of WBRT as part of the adjuvant treatment plan on LC could not be addressed, as all three patients with local progression received only GK-SRT as adjuvant radiotherapy.
Kaplan-Meier estimates of local control (LC) after radiotherapy to the resection cavity. Local control was defined as the absence of marginal recurrence within the resection cavity. Tick marks indicate censoring at the date of last imaging follow-up, initiation of salvage whole-brain radiotherapy, or death. Shaded areas indicate 95% confidence intervals.
The 1-year DIC was 54.1% (95% CI=0.38-0.76), and the median time to distant intracranial failure was 12.9% (95% CI=0.09-NA), with new out-of-field metastases occurring between 2.6 and 55.4 months after surgery (Figure 2). No statistical differences were identified in regard to receiving WBRT as part of the adjuvant treatment plan (HR=0.34; 95% CI=0.07-1.63; p=0.177) or receiving systemic treatment (HR=0.97; 95% CI=0.39-2.42; p=0.907). Salvage treatment was prescribed for all patients with regional progression (n=20), with a median time from regional progression to onset of salvage treatment of 9.0 days (range=1-34). No significant differences in DIC were observed by tumor histopathology when comparing NSCLC (n=15 patients) with breast cancer (n=4 patients; HR=2.15; 95% CI=0.64-7.31; p=0.218) or with other malignancies (n=13 patients; HR=1.66; 95% CI=0.62-4.46; p=0.311). The PH assumption was met for all DIC models.
Kaplan-Meier estimates of distant intracranial control (DIC) per patient. DIC was defined as the absence of new brain metastases on a per-patient basis. Tick marks indicate censoring at the date of last imaging follow-up, initiation of salvage whole-brain radiotherapy, or death. Shaded areas indicate 95% confidence intervals.
Concerning OS, survival status was available for 29 patients at the time of data collection, among which 21 (72.4%) had passed away, while 8 (27.6%) were still alive. The median survival was 27.3 months (95% CI=10.9-53.6 months), with a 1-year OS of 62.1% (95% CI=0.47-0.83) (Figure 3). No significant differences in OS were identified in regard to the number of additional metastases (HR=1.04; 95% CI=0.69-1.59; p=0.843). There were no statistically significant differences in OS regarding either the time from surgery to onset of GK-SRT (HR=0.97; 95% CI=0.93-1.02; p=0.202) or the reception of systemic treatment (HR=0.72; 95% CI=0.28-1.83; p=0.490). Regarding WBRT, among the total 29 patients with survival status data, 5 patients (17.2%) received WBRT combined with GK-SRT as part of the adjuvant treatment concept. Including WBRT in the adjuvant treatment plan did not have a statistically significant impact on OS (exact permutation log-rank; p=0.708), compared to the subgroup of patients treated with adjuvant GK-SRT only (Figure 4). No significant differences in OS were observed by tumor histopathology when comparing NSCLC (n=14 patients) with breast cancer (n=4 patients; HR=1.49; 95% CI=0.44-4.99; p=0.519) or with other malignancies (n=11 patients; HR=1.25; 95% CI=0.48-3.26; p=0.649). In the histology model, the PH assumption was violated (Schoenfeld test, global p=0.014). The Cox model with time-dependent covariates showed a significant association between distant intracranial failure and OS (HR=2.87; 95% CI=1.08-7.61; p=0.034) but not between local failure and OS (HR=2.75; 95% CI=0.60-12.60; p=0.193). No significant associations with LC, DIC, or OS were observed for sex (HR=1.34; 95% CI=0.57-3.17; p=0.502), age at treatment (HR=1.01; 95% CI=0.97-1.05; p=0.622), or KPS (HR=0.77; 95% CI=0.52-1.14; p=0.193).
Kaplan-Meier estimate of overall survival (OS) per patient. Tick marks indicate censoring for patients alive at the time of data cutoff. Shaded areas indicate 95% confidence intervals.
Kaplan-Meier estimates of overall survival (OS) per patient stratified by adjuvant whole brain radiotherapy (WBRT). Tick marks indicate censoring for patients alive at the time of data cutoff. Shaded areas indicate 95% confidence intervals for each curve.
Discussion
Stereotactic radiotherapy has changed the paradigm of BM management by offering precise targeted irradiation to the lesion. Over the past few decades, the use of SRS and FSRT for treating BM resection cavities has significantly evolved to become the standard adjuvant treatment following surgical resection (9, 20). In our retrospective study, we evaluated the effectiveness of GK-SRT as an adjuvant treatment for BM resection cavities in a cohort parallel to INTRAMET and managed by the same department conducting the clinical trial. We analyzed 41 BM resection cavities from 37 patients over a period of nearly five years, achieving a substantial median follow-up of 19.3 months.
Our findings demonstrate excellent LC (1-year LC=93.3%) in line with previous studies, where percentages generally range from 80 to 90% (14-17). Furthermore, novel alternatives such as IORT show similar LC as our cohort (11, 12). The INTRAMET cohort (n=35, median follow-up of 25.7 months), managed by the same department as our GK-SRT cohort, is comparable to our MRI-followed collective (n=36, median follow-up of 19.3 months) (13). With comparable 1-year LC (94.3% vs. 93.3%), these results suggest that adjuvant GK-SRT and IORT may represent equivalent treatment options. It is worth noting that this SRT cohort comprises an average patient population at a German university hospital, not a carefully chosen population for a prospective trial. Tumor histopathology has not been identified as a prognostic factor for local control in previous studies, which is consistent with our findings (9, 18-21). Comparing SRS and FSRT, one might expect that the LC of resection cavities treated with SRS could differ from that of an FSRT approach. However, despite the tendency for larger cavities to receive FSRT, Lehrer et al. found no significant differences in the estimated 1-year LC between SRS and FSRT (22). In addition, consistent with findings from other studies, systemic treatment was not associated with LC (18-20).
Despite excellent LC, regional disease control remained challenging, with a 1-year DIC of 54.1%, a pattern also reported in prior studies (15, 23). In our cohort, better DIC correlated with longer OS, in contrast to most published reports (6, 8), but underscoring the need for further investigation of strategies to improve regional disease control, whether by optimized local or systemic approaches. In our analysis, SRS and FSRT as salvage treatment of regional recurrence were feasible in the majority of cases, warranting regular surveillance imaging following GK-SRT to enable early treatment of regional progression. Our 1-year OS (62.1%) and median survival time (27.3 months) are consistent with findings from previous studies, generally suggesting modest survival benefits associated with adjuvant SRT (15, 24). It is crucial to acknowledge the inherent limitations of retrospective analyses in evaluating survival outcomes, particularly considering incomplete data. While receiving WBRT as part of the adjuvant treatment plan did not show a significant impact on OS compared to the subgroup treated only with GK-SRT, these results could be confounded due to the small size of patients that received adjuvant WBRT. Yet, the proven negative impact of WBRT on neurocognitive function and quality of life question whether it should be incorporated into the adjuvant plan (25). Furthermore, the lack of significant differences based on tumor histopathology or the number of additional metastases suggests a complexity of factors influencing prognosis, highlighting the need for research into novel therapeutics to achieve long-term survival (9).
Typically, the interval between surgery and the initiation of adjuvant GK-SRT spans several weeks to prevent impairments in surgical wound healing. Yaghi et al. (n=176) reported that a postoperative delay of more than 22 days was associated with a decreased risk of all-cause mortality; however, waiting more than 40 days doubled the risk of local tumor progression (26). In our cohort, the median interval from surgery to the onset of GK-SRT was 24.5 days, aligning with the recommended 22-40-day interval. Similar findings have been observed in other studies (24, 26). To fully leverage the benefits of novel targeted therapies and immunotherapy, a prompt initiation or seamless continuation of systemic therapy is essential. Therefore, the safe yet early initiation of GK-SRT appears critical for achieving optimal local control and facilitating the rapid commencement of systemic therapies. Notably, IORT permits the immediate initiation or continuation of systemic therapy post-wound healing and has been demonstrated to reduce the interval between resection and systemic therapy (11).
Study limitations. Firstly, its retrospective design introduces inherent biases typical of observational studies. Secondly, the smaller sample size and single-center setting may limit the generalizability of our findings. Thirdly, variability in dose prescription and fractionation schemes may have influenced the results, despite our efforts to standardize reporting by calculating the EQD2 for each resection cavity. On the other hand, the diversity of radiation schemes reflects evolving clinical practice and thus constitutes “real-world” data. Of note, SRT is now delivered on multiple platforms, including linear accelerators with advanced techniques (27, 28). Prospective studies such as the upcoming LEXIMATE trial, which directly compares SRT on multiple platforms to IORT as an alternative treatment strategy, may yield more robust evidence (29).
Conclusion
Adjuvant GK-SRT to the resection cavity provides excellent local control in patients with brain metastases. IORT may represent a comparable alternative, with confirmation from prospective trials awaited. Despite effective local strategies, distant intracranial progression remains a major limitation and, in this retrospective study, appeared to be associated with poorer overall survival. Salvage options such as additional SRT are feasible. Continued research is needed to integrate local therapies with systemic and multimodal approaches to improve both intracranial control and survival outcomes.
Acknowledgements
None.
Footnotes
Authors’ Contributions
AGJ: Data curation, Formal analysis, Writing – original draft. GW: Visualization, Formal analysis, Writing – review & editing. SB: Data curation, Writing – review & editing. AYA: Data curation. FS: Data curation. FAG: Conceptualization, Supervision, Writing – review & editing. AMR: Conceptualization, Methodology, Supervision, Writing – original draft, Writing – review & editing.
Conflicts of Interest
The Authors declare no conflicts of interest.
Funding
The Authors received no specific funding for this work.
Artificial Intelligence (AI) Disclosure
No artificial intelligence (AI) tools, including large language models or machine learning software, were used in the preparation, analysis, or presentation of this manuscript.
- Received August 4, 2025.
- Revision received September 2, 2025.
- Accepted September 4, 2025.
- Copyright © 2025 The Author(s). Published by the International Institute of Anticancer Research.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.










