Regarding single primary entities, the risk of having intracranial tumor involvement over the clinical course is 37% for malignant melanoma, 50% for non-small-cell lung cancer (NSCLC), >50% for metastatic human epidermal growth factor receptor-positive (HER+) breast cancer or triple-negative breast cancer (TNBC), and 80% for small-cell lung cancer (SCLC)
[4,5,6,7][4][5][6][7]. The most frequent primaries are SCLC or NSCLC, breast cancer, malignant melanoma, renal cell carcinoma (RCC), and colorectal cancer (CRC); the former three primaries constitute 80% due to the high incidences
[8,9][8][9]. Therefore, BM screening with cranial magnetic resonance imaging (MRI) is recommended in the primary staging of SCLC/NSCLC or higher-stage malignant melanoma, but not yet for other high-risk entities such as TNBC
[2]. In childhood and adolescence, BMs are rare; in most cases, the primary tumors are neuroblastoma, sarcoma, nephroblastoma, melanoma, and germline tumors
[10,11][10][11]. All brain regions can be affected, with three-quarters showing parenchymal growth. Out of this, 80% are located supra-tentorial, 15% in the cerebellum, and 5% in the brain stem
[12,13][12][13]. BM can often be detected posterior to the Sylvian fissure at the junction of the temporal, parietal, and occipital lobes, representing the supply zone of the median cerebral artery
[13]. BM are singular/solitary in 72% of cases; SCLC/NSCLC and melanoma show more often multiple lesions compared to RCC and CRC
[14,15][14][15]. Overall, the incidence of BM is increasing because of better systemic treatment options, leading to improved survival for many cancer types and hence a higher risk for BM over the course of the disease
[16]. These new drugs consist of small molecules and antibodies targeting tyrosine kinases, molecular alterations such as epidermal growth factor receptor (EGFR) mutations or anaplastic lymphoma-kinase (ALK) rearrangements, or immune checkpoint inhibitors, which contrary to classic chemotherapies, can better cross the blood–brain barrier for intrathecal anti-tumor effects and therefore protracted development of BM
[17,18,19,20][17][18][19][20]. BM screening in NSCLC/SCLC, or melanoma patients often detects small and clinically unapparent BM that can be easily treated with definitive stereotactic radiosurgery (SRS). But not all BMs are detected at an early stage where they are amenable to radiotherapy (RT). Large or symptomatic lesions in eloquent brain areas can cause disability and even life-threatening conditions that require rapid decompression for fast and persistent symptom relief
[21,22][21][22]. Furthermore, resection supports the histopathologic diagnosis in patients with cancer of unknown primary or when changes in molecular profiles compared to the primary tumor are suspected
[23,24][23][24]. However, local recurrence rates after the resection of a single BM are up to 50%
[25,26,27][25][26][27]. Therefore, RT is indicated to improve local control (LC) at the site of resection. Up to date, several techniques are available, such as whole brain radiation therapy (WBRT), local SRS or hypo-fractionated stereotactic RT (hFSRT) to the cavity, neo-adjuvant stereotactic radiosurgery (N-SRS), and low-energy X-ray intraoperative RT (IORT).
2. Surgery
The indications for BM surgery have recently been updated in the ASCO-SNO-ASTRO guideline and the EANO-ESMO clinical practice guidelines
[23,24][23][24]. BMs affect an exceptionally high number of cancer patients and thereby represent a common challenge in medical care for the population
[28,29][28][29]. In 20 to 25% of cases, a BM may be the primary cause of the diagnosis of an underlying cancerous disease
[28,30,31][28][30][31]. Due to significant improvements in cancer therapies and imaging techniques as well as an aging population, a growing number of patients are eligible for and in need of neurosurgical treatment of brain tumors, a very heterogeneous group of which BMs represent the most prevalent subgroup by a large margin. The antique assumption that BMs are easily removed due to being well-circumscribed and having non-invasive growth has been long disproven
[32,33,34][32][33][34]. BM of SCLC or melanoma can infiltrate the brain by several millimeters
[35]. This might contribute to the lower-than-expected complete resection rates and higher local recurrence rates after surgery for BMs
[34]. Being diagnosed with a BM entails drastic consequences for patients’ quality of life (QoL), functional autonomy, and socioeconomic status. In addition to being diagnosed with cancer of any origin, BMs are often viewed as terminal and incurable by patients, which heavily impedes their functional autonomy and well-being. It must be emphasized that cerebral metastases may produce a range of neurological symptoms leading to severe disability and impairment of QoL
[30,31,36][30][31][36].
The resection of BMs is generally not of a curative nature. As the presence of a BM merely mirrors the progression of the causative systemic disease, any cranial surgery is only indicated for symptom control and securing a histopathological diagnosis in cases of a yet unknown primary cancer
[34]. While the currently available evidence does not support neurosurgical resection as a curative intervention, complete resection has recently been shown to play a significant role in survival extension in addition to being a measure of symptom control and preservation of QoL
[28,29,31,32,34,37,38][28][29][31][32][34][37][38].
3. Definitive Radiosurgery (SRS)
The important role of SRS in the treatment of solitary BM was evaluated several decades ago in a multicenter retrospective study in North America
[87][39]. This study included 122 patients from four institutions. The inclusion criteria exactly matched those of two randomized trials published a few years earlier that evaluated the role of surgery alone versus surgery plus WBRT
[43,88][40][41]. It was a surprise to find that radiosurgery plus WBRT achieves absolutely comparable local tumor control to that achieved by surgery plus WBRT. Therefore, this study demonstrated the ablative nature of radiosurgery.
Subsequent randomized studies have shown significantly better local tumor control when WBRT is combined with radiosurgery compared to WBRT alone
[89,90][42][43]. Furthermore, radiosurgery plus WBRT compared to WBRT alone even improved OS in certain subgroups of patients with favorable recursive partitioning analysis/graded prognostic assessment (RPA/GPA)
[91][44].
Given the very good local tumor control of radiosurgery and the possible side effects of WBRT, a number of studies have attempted to answer the question of whether—in the case of 1–4 brain lesions—radiosurgery can be applied alone, without WBRT. In a large prospective randomized multicentric EORTC (European Organization for Research and Treatment of Cancer) trial, Kocher et al.
[92][45] evaluated the role of WBRT in the treatment of patients with 1–3 BM. Four arms were considered: radiosurgery, surgery, radiosurgery plus WBRT, and surgery plus WBRT. The study showed that WBRT does not improve functional independence (the primary endpoint of the study) or OS, although it significantly reduces local and distant cerebral relapses and, moreover, the number of patients who die from neurological disease. Interestingly, local relapse after local therapy (without WBRT) was 59% after surgery alone and 31% after radiosurgery alone. In contrast to the European study, a prospective multicenter randomized Japanese study shows that in a group of patients with NSCLC and favorable prognosis (diagnosis-specific GPA 2.5–4), adding WBRT to radiosurgery (RS) can significantly improve OS
[93,94][46][47]. However, in a systematic review of the Cochrane database evaluating RS alone versus RS plus WBRT, the conclusion is that WBRT has no impact on OS
[95][48]. Therefore, radiosurgery alone remains the indicated treatment in patients with 1–4 non-resected BM.
An important argument in the decision to abandon WBRT is the significant negative effect of WBRT on cognitive function. Chang et al.
[96][49] observed in a prospective randomized trial that patients treated with WBRT plus RS had a significant decline in learning and memory function 4 months after treatment compared to patients who were treated with radiosurgery alone. A few years later, Brown et al. reported similar results in another study, also prospective, multicenter, randomized, and cognitive function as the endpoint
[97][50].
A Japanese multicenter prospective observational study evaluated the role of radiosurgery alone in the treatment of patients with multiple BM: 5–10 cerebral lesions with a maximum total metastatic volume of less than 15 mL
[98][51]. In the 1194 patients with 1–10 unresectable BM included in the study, the authors find that, in terms of OS, there is no difference between patients with 2–4 metastases and those with 5–10 lesions. Patients with solitary lesions had better survival. The authors conclude that, given the side effects of WBRT, it can be abandoned in favor of radiosurgery, even if the latter should be repeated in case of relapse. Moreover, in a more recent cohort study, the authors advocated radiosurgery alone in patients with up to 15 lesions
[99][52].
4. Radiation Therapy in Combination with Surgery
4.1. Local Adjuvant Radiotherapy of the Resection Cavity
According to international guidelines, surgical removal of BMs is recommended in patients with single, large (>3 cm) and symptomatic lesions in whom definitive radiosurgery is deemed ineligible or in rare cases that require histological confirmation
[23,100][23][53]. Because 1- and 2-year LC rates following surgery were reported to range around 40% only
[92[45][54],
101], post-operative radiotherapy (PORT) of the surgical cavities is generally recommended.
Historically, PORT was conducted via WBRT, administering a variety of dose fractionations with single doses of 1.8–4 Gy and total doses of 20–50.4 Gy. Across several rather small studies, WBRT significantly reduced the risk of both local and distant intracerebral relapse. Also, death from neurological causes occurred less frequently after post-operative WBRT; however, there was no consistent effect on OS
[25].
According to the aforementioned pivotal EORTC 22952-26001 trial, post-operative WBRT following radio-surgical treatment was not found to improve OS, but it significantly decreased QoL
[92,102][45][55]. However, as reported before, post-operative WBRT did improve LC and reduce the number of neurological deaths. To make use of this local effect and simultaneously spare the healthy brain, several studies have investigated the effect of either immediate or salvage SRS instead of post-operative WBRT or watch-and-scan strategies:
Postoperative SRS instead of observation strategies did not alter OS rates but improved LC rates from 43 to 72%
[101][54] at modest toxicity rates. Postoperative SRS instead of WBRT was superior in preserving neurocognition and QoL
[6]. However, after 6 months, surgical bed control following WBRT was surprisingly significantly superior to SRS (80% vs. 87%), which may be related to challenges in target volume delineation following surgery
[103][56]. Reserving post-operative SRS for patients with either residual or recurrent BMs following resection was analyzed in a large Japanese study and randomized against elective post-operative WBRT. The authors found comparable OS and less toxicity with salvage SRS vs. elective WBRT
[104][57].
In conclusion, post-operative radiotherapy is still considered mandatory after surgical removal of BMs, but it should not be conducted as WBRT. Secondary to the aforementioned neuro-cognitive sequelae affecting the patient´s QoL, there is a trend toward focal radiotherapy to the resected BM site to omit or delay WBRT
[25,105,106,107,108,109][25][58][59][60][61][62]. Local post-operative RT to the resection cavity spares the healthy brain and can therefore preserve neuro-cognition, and is recommended in established guidelines
[6,23,24,101,110][6][23][24][54][63]. Normo-fractionated or modestly hypo-fractionated adjuvant RT has been investigated by several groups demonstrating satisfying control rates
[100[53][64][65][66][67],
111,112,113,114], For example, Byrne et al. applied 30–42 Gy in single-fractions of 3 Gy to the post-operative cavity in 54 patients. They achieved a local control rate of 97.0% and 88.2% at 6 and 12 months, respectively
[112][65]. Another study demonstrated a local control rate of 83% at 1 year and 78% at 2 years in 57 patients treated postoperatively with local RT; the median dose was 48 Gy (30.0–50.4 Gy) in 25 (10–28) fractions
[111][64]. However, most data are available for adjuvant stereotactic RT, which can be applied with single-session SRS or hypo-fractionated (3–7 treatments) stereotactic RT (hFSRT)
[109,115,116,117,118,119,120,121,122][62][68][69][70][71][72][73][74][75]. The ideal stereotactic scheme is still under debate; cavity size and dosimetric aspects have to be taken into consideration
[123][76]. Up to date, there are no randomized controlled trials available to determine which of the two techniques is superior in terms of efficacy (LC) and safety, namely the risk of radiation necrosis (RN). Currently, two prospective randomized trials are comparing post-operative SRS and hFSRT (NCT04114981 and NCT05160818) to answer this crucial question. Data from several predominantly retrospective studies are available that used different dose, fractionation, and contouring schemes (e.g., surgical tract included or not included, margin vs. no margin). Two trials on SRS reported 1-year LC rates of 72% and 60%, respectively
[6,101][6][54]. In SRS, often no or a very small (1 mm) safety margin is applied to minimize toxic effects on the normal brain, which could in part explain the inferior LC rate. A recently published series on >500 cavities treated with stereotactic RT (SRS 15.2%, hFSRT 84.8%) demonstrated excellent LC rates of 93% with 1-year overall adverse radiation effects of 9%; in regard to the imbalanced groups, the authors could not find inferiority in SRS compared to hFSRT
[124][77]. Another meta-analysis looking at 24 studies for definitive and post-operative RT of large BMs reported for hFSRT (405 patients, heterogeneous fractionation schemes) a LC of 87% compared to 68% in the SRS group (183 patients)
[125][78]. Furthermore, in the majority of studies, the definitions of LC and RN were not clearly defined and depended predominantly on the institutional standards, mostly relying on radiologic findings. A recent multicenter analysis of 558 cavities treated with hFSRT (median total dose 30 Gy, median dose per fraction 6 Gy) demonstrated a 1-year LC rate of 84% and an RN rate of 4.1%
[122][75]. In general, adjuvant hFSRT seems favorable over single-fraction radiosurgery in terms of higher LC and lower incidences of RN, most likely due to breaks between fractions and a higher biologic effective dose (BED)
[122,125,126,127,128][75][78][79][80][81]. For hFSRT dose (BEDα/β = 10 > 48 Gy), cavity/PTV volume (<23 mL), single BM, extent of resection, and a controlled primary tumor are associated with LC
[108,109,115,122,129,130,131][61][62][68][75][82][83][84]. Risk factors for RN are treating volume, fractional dose, and total RT dose
[132,133][85][86]. Some studies described constraints for fractional (30 Gy in 5 fractions, BED10 = 48 Gy, BED2 = 120 Gy) irradiated normal brain volume receiving 20 Gy or more of 20 cm
3 and 25 cm
3 (V20Gy > 20 cm
3 and V20Gy > 25 cm
3) for the development of symptomatic RN
[121,129,134,135][74][82][87][88]. RN rates for 5–7 fractions (BED2 87.5–138.1 Gy) are 0–10% with a 1-year LC of 68 to 93%
[109,121,122,128,135,136][62][74][75][81][88][89]. Whereas three-fraction hFSRT (BED2 120–214.5 Gy) shows higher LC rates (88–98.9%) but a much higher incidence of RN (14–25.7%)
[118,137,138,139,140][71][90][91][92][93]. Therefore, hFSRT with 5 to 7 fractions can be considered effective and safe.
Despite the increasing use of local adjuvant RT, there is no definite standard for target delineation. A recent consensus guideline recommended enclosing the surgical tract and, if necessary, a margin for the bone flap, venous sinus, and pia mater in cases where the BM abuts the appropriate structures
[103][56]. Based on several studies, adding a 2–3 mm margin to the resection cavity for the planned target volume (PTV) seems beneficial for LC
[116,141,142][69][94][95]. However, a safety margin can be discussed controversial. A meta-analysis from 2013 including 629 patients did not find an effect of an extra margin on LC, and in the aforementioned study (n = 500, SRS 15.2%, hFSRT 84.8%), no difference was detected for margin (1–3 mm; 76% of cases) vs. no margin
[124,143][77][96].
Another important aspect of target volume delineation is the fact that the cavity volume can vary over time after surgery. One study on 68 cavities demonstrated, in most cases, a median percent volume change of minus 29% compared to preoperative BM volume within 1–3 days after resection, but without significant further changes over the next 6 weeks
[144][97]. Other series stated cavity dynamics within 4 weeks after resection, and larger cavities show larger shrinkage than smaller ones
[145,146,147][98][99][100]. This might imply a protracted start of adjuvant stereotactic RT. However, several studies have shown that a delay has a negative impact on LC
[148,149][101][102]. Taken together, due to variable changes in the cavity volume within weeks after surgery, a recent MRI not older than 7 days should be obtained for RT planning
[142,144,150][95][97][103]. Consequently, the right timing for post-surgical RT is not standardized, but besides surgical-related factors such as the patient´s performance and wound healing, the urge for a systemic treatment depending on the tumor staging should be taken into consideration
[122,135,146][75][88][99].
Regarding the challenges of optimal target delineation and the dynamics of the resection cavity, further modalities of local RT for resectable BM are gaining increasing interest
[151,152][104][105].
4.2. Neoadjuvant Radiation Therapy
As an alternative to adjuvant RT to the resection cavity, neo-adjuvant stereotactic radiosurgery (N-SRS) of the BM prior to resection seems to be another option. First of all, the concept of N-SRS of intact BM offers improved target volume contouring since, with adequate imaging, exact delineation of intact BM is fairly straightforward. Additional safety margins are not necessarily needed for optimal coverage, and no surgical tract has to be included in the target volume, which enables better sparing of the surrounding healthy brain. A dosimetric analysis comparing neoadjuvant RT with adjuvant hFSRT could demonstrate a significantly reduced dose exposure to the brain and favorable conformity indices
[153][106]. Overall, irradiated target volume appears lower for pre-operative RT compared to post-operative RT, especially for smaller metastases; in addition, the adjacent brain volume receiving the prescription dose is removed during surgery, which could lead to less frequent RN and wound healing complications
[154,155][107][108]. Another major aspect is that during resection, tumor cells might be disseminated, with a significant risk of leptomeningeal disease (LMD)
[156,157][109][110]. Therefore, N-SRS might sterilize viable tumor cells, which are then less competent for replication when later spilled during tumor resection, potentially translating into lower rates of LMD. Also, the much-needed systemic therapy or rehabilitation could be initiated more rapidly because there is no 2–3 week delay to ensure adequate wound healing and perform adjuvant hFSRT.
Over the last few years, there has been growing evidence for the effectiveness and safety of neoadjuvant SRS for BM
[158,159,160,161,162,163,164,165][111][112][113][114][115][116][117][118]. The first larger series investigated the role of N-SRS in 47 patients with 51 BM (median diameter 3.0 cm, range: 1.3–5.2 cm) undergoing resection in median within 1 day (range: 0–7 days) after N-SRS with a median dose of 14.0 Gy (range: 11.6–18 Gy) prescribed to the 80% isodose level
[158][111]. LC was 97.8%, 85.6%, and 71.8% at 6, 12, and 24 months, respectively. Eight patients with local failure were re-operated and proved to have recurrence without RN. Local failure was more likely for lesions larger than 3.4 cm (
p = 0.014). One trial compared pre- and post-operative stereotactic RT in 180 patients, with 189 BMs being resected. Of note, in the N-SRS group (n = 66), the marginal dose was reduced by 20% (median dose 14.5 Gy vs. 18 Gy) with no extra margin added to the gross BM volume compared to the post-surgical cohort with a safety margin of 2 mm
[160][113]. Patients underwent BM resection within 48 h after SRS. Outcomes were similar regarding local recurrence, distant brain recurrence, and overall survival, but with significantly lower rates of symptomatic RN and leptomeningeal spread in favor of the pre-operative group, with 4.9% vs. 16.4% (
p = 0.1) and 16.6% vs. 3.2% (
p = 0.1) at 2 years, respectively
[160][113]. In another study by this group, pre-operative SRS (66 patients with 71 lesions) was compared to post-operative WBRT (36 patients with 42 lesions). In analogy to the aforementioned study, the dose was reduced by 20% with no extra margin for the PTV with surgery within 48 h after neoadjuvant RT
[161][114]. Again, results were similar for local recurrence, distant brain failure, and LMD recurrence. Rates for symptomatic RN were higher for pre-operative SRS (5.6% vs. 0.0%), and the cavity size was significantly smaller in this group
[161][114]. Comparably, a study on 117 patients with 125 lesions treated with neoadjuvant SRS showed LC rates of 74.9% at 24 months and RN in 4.8% of cases
[163][116]. Lately, pooled data from a multicenter cohort of 253 index lesions with N-SRS (median dose 15 Gy) were published, describing local recurrence rates of 15.0% and 17.9% at 1 and 2 years, respectively; LMD occurred in 7.6% after 24 months
[162][115].
Overall, N-SRS followed by resection appears to be effective and safe, but there are some disadvantages. First of all, N-SRS is performed without prior histopathologic proof for BM. MRI offers a high positive predictive value for BM; however, there is still a low risk of falsely irradiating high-grade glioma, lymphoma, or even an inflammatory process with N-SRS. Furthermore, urgent surgery for space-occupying lesions should not be delayed by N-SRS. Therefore, patients with cancer of unknown primary and acute symptoms necessitating rapid decompression might not be the best candidates for N-SRS. Furthermore, more data are needed on dose and fractionation schemes as well as with regard to immunogenetic effects in interplay with checkpoint inhibitors. So far, the number of studies on N-SRS is limited compared to post-operative RT, but further prospective studies on effectiveness, safety, and QoL are underway
[166,167,168][119][120][121].
4.3. Intraoperative Radiation Therapy (IORT) with Low-Energy X-rays
Among the vast therapeutic arsenal for treating BM, IORT poses an interesting option in the adjuvant spectrum. The introduction of modern IORT for treating cancer traces back to the 1960s, when some initial reports announced promising outcomes in various anatomical locations using electron-based IORT (IOERT)
[169][122]. Falling within the frame of brachytherapy (BT), this approach allows delivering radiation by placing an emitting source in direct contact with the target tissue during a single surgical act. The evolution of this technique over the past fifty years, since the early IOERT or BT-based IORT reports, has yielded a robust and convenient modality to still consider nowadays, despite a persistent lack of large prospective data.
4.4. Re-Irradiation of Recurrent Brain Metastases
The treatment options of immunotherapy and targeted therapy led to the prolonged survival of many stage IV cancer patients. This intensified the focus on optimal treatment strategies for patients with recurrent BM (rBM) after initial RT. So far, published data comprises only retrospective, mostly monocentric case series, and recommendations from guidelines are lacking as well. From a radiation oncological perspective, it is appropriate to distinguish between patients with recurrences after WBRT and patients with local failures after focal radiotherapy.
Despite missing prospective data, SRS is the treatment of choice for oligometastatic recurrences (1–4 metastases) after WBRT due to its high efficiency and minimal toxicity
[188,189][123][124]. The focus of the published data lies in appropriate patient selection. KPS, the interval from WBRT, and controlled systemic disease are common selection criteria in this situation. As RN is rare after WBRT without focal RT, there is only a limited diagnostic dilemma in this situation.
The situation of rBM after focal RT is definitively more challenging due to the limited sensitivity and specificity of available non-invasive diagnostic procedures to distinguish it from RN
[190,191,192,193][125][126][127][128]. As combinations of RN and rBM occur, even a biopsy may be subject to sample error. This emphasizes the role of salvage surgery in this setting
[190][125]. After surgery, focal re-radiation of the cavity increases LC
[194][129]. Focal RT of suspected rBM holds the risk of unintentional treatment of RN with possible aggravation of neurologic symptoms. Thus, several published series show favorable LC rates with only moderately increased RN rates. Despite increasing numbers of patients with suspected rBM (versus RN), the optimal diagnostic workup in this situation remains unclear
[195,196][130][131].