Introduction
Brain metastases are the most common intracranial tumors in adults, with incidence up to 28% in patients suffering from metastatic cancer [1]. The incidence may be rising due to factors such as improved detection and improved survival [2] with stronger systemic treatments. Brain metastases may be asymptomatic and may as well cause a wide range of neurological symptoms, from focal neurological deficits to seizures and cognitive dysfunction. Prognosis of these patients are historically poor with medial survival traditionally quoted to be in terms of months [3].
There are multiple treatment modalities for brain metastases [4], including pharmacotherapy, radiotherapy, and surgery. Surgery has the unique advantage of obtaining histology e.g., when there is uncertainty regarding diagnosis, or when a change in tumor molecular profile is suspected. It may also be favored in larger tumors, when intracranial pressure is acutely elevated, or when a prolonged course of steroid needs to be avoided. Despite the advantages of neurosurgical excision, more than half of patients [5] experience local recurrence in the surgical cavity. Thus, adjuvant radiotherapy is recommended [6] after resection. Traditionally whole brain radiotherapy (WBRT) has been employed [7,8], but has been shown to increase risk of neurocognitive decline and negatively impact quality of life [9-13] and is falling out of favor.
Stereotactic radiotherapy (SRT) uses highly conformal radiation beams along with custom-made immobilization devices to deliver high doses of radiotherapy to a target, with rapid fall-off to minimize toxicities to nearby normal structures. SRT can be either delivered in a single-fraction (commonly termed stereotactic radiosurgery “SRS”) or in multiple fractions (commonly termed fractionated stereotactic radiotherapy “FSRT”). Two key randomized phase 3 trials [5,13] demonstrated that SRS improved local recurrence rates compared to observation alone [5] and resulted in better neurocognitive outcomes compared to WBRT [13], respectively. As a result of these two studies, post-operative SRS is considered a standard of care in resected brain metastases. FSRT theoretically improves normal tissue tolerance to radiotherapy by using smaller radiation doses per fraction, and several retrospective series demonstrated encouraging outcomes with a variety of FSRT prescription regimens [14-17]. Thus, postoperative SRS/FSRT both represent treatment options in surgically resected brain metastases [4,6,18,19], with FSRT favored in larger resection cavities.
In this study, we review data from two radiotherapy institutions pertaining to case selection, treatment delivery, and outcomes of post-operative SRT, and analyze the impact of radiotherapy dose on tumor control and late-treatment complications.
Materials and Methods
1. Inclusion/exclusion criteria
The institutional radiotherapy databases for the years 2018–2022 of two tertiary oncology centers were reviewed. Patients were eligible if they were aged 18 and above, had undergone surgical resection of at least one brain metastasis, and received SRT to the surgical bed afterwards. If more than one brain metastasis was excised (in the same operation, or in separate surgeries), each surgical cavity was analyzed individually. Patients were excluded if they had incomplete medical records, had prior SRT to the same site, did not complete the full course of prescribed radiotherapy, or did not have follow-up imaging.
2. Patient selection and treatment
The decision for excision was made by the neurosurgeon in charge based on tumor size/symptoms, patient factors and the need for tumor histology. All patients with resected brain metastases were discussed in a multidisciplinary team meeting involving clinical oncologists, neurosurgeons, nurse specialists, and medical physicists, where consensus for treatment plan is obtained. Decision for SRT prescription dose and fractionation (including decision for SRS vs. FSRT) was made by the treating physician in conjunction with the medical physicist in charge.
Patients were immobilized using a frameless system with a custom-made thermoplastic head mask. Patients had contrast-enhanced magnetic resonance imaging (MRI) of the brain performed in treatment position prior to radiotherapy. The MRI is then co-registered with a plain-cut planning computed tomography scan at 1mm slice thickness to assist in target delineation. Target delineation and dose calculations were performed on the Brainlab iPlan (Brainlab, Munich, Germany) platform. The clinical target volume included the surgical cavity and gross residual tumors in the case of subtotal resections, in accordance with international guidelines [20] where applicable. After target delineation, a 0.5mm planning target volume (PTV) margin was given to account for setup variability in all cases. Treatment was prescribed at the 50% to 90% isodose line depending on the size, shape, and location of the cavity. Plans were optimized with the aim that >99% of the PTV received 95% of the prescription dose and >95% of the PTV received 100% of the prescription dose. The target volume was defined as the volume of the PTV, and the treatment volume was defined as the volume receiving the prescription dose. A conformity index of 1.0–2.0 indicates that the plan is per protocol. Normal tissue constraints were taken in reference to the 2018 UK consensus guidelines [21].
SRT was delivered using a Varian Clinac iX (Varian, Palo Alto, CA, USA) linear accelerator using step-and-shoot intensity modulated radiotherapy or conformal arc therapy. SRS was delivered in a single fraction and FSRT was delivered in 3 or 5 fractions in alternate days. All patients were treated using an image-guided system with pre-treatment verification using the ExacTrac (Brainlab) system and online correction performed by a 6 degree-of-freedom treatment couch.
All patients were followed up jointly by oncology and neurosurgery units, with regular imaging (most commonly MRI) arranged to detect recurrences and treatment complications.
3. Statistical analysis
Data was retrieved from the institutional databases and electronic medical records. We extracted demographic data (e.g., patient sex, age, performance status), information regarding the primary malignancy (date of diagnosis, histology, extracranial control, systemic treatment), information relating to the excised brain metastasis (date and completeness of resection, tumor number/size/location, dural involvement, symptoms from tumor, use of steroids), information relating to radiotherapy (prior radiotherapy, dose prescription, target/treatment volumes, conformity indices) and information relating to treatment outcomes (local control, distant brain control, survival, development of radiation necrosis [RN]).
Descriptive data was summarized in mean/median/range. Tumour progression or development of RN was defined by the reporting neuro-radiologist. MRI findings suggestive of RN include lack of restricted diffusion or lack of increased relative cerebral blood volume on perfusion imaging. Patients were also considered to have RN if excision was performed showing pathological features of RN and no viable tumour cells. Local control rate (LCR), distant brain control rate (DBCR), and overall survival (OS) were defined by the start date of radiotherapy to date of respective event or last follow-up. For statistical inference, categorical variables are analyzed by chi-square test or Fisher’s exact test as appropriate. LCR and DBCR were analyzed using cumulative incidence function to account for competing risk (patient death). OS was analyzed by Kaplan-Meier method. Significant prognosticators were identified by univariate analysis and multivariate analysis. Predictive factors with p < 0.2 in univariate analysis were included in multivariate analysis. Cox regression with forward likelihood ratio selection was used. A p-value of <0.05 was considered statistically significant. SPSS version 27 (IBM Corp., Armonk, NY, USA) was used for data entry and analysis.
Results
1. Patient demographics
Seventy-six cavities in 69 patients fulfilled the inclusion criteria. After application of the exclusion criteria, 65 cavities were available for analysis. Reasons for exclusion included: no available follow-up imaging (five cavities), prior SRT to that region (three cavities), the neurosurgical procedure being a biopsy (two cavities), and treatment prescribed but not given due to clinical deterioration (one cavity). All cavities were analyzed individually.
The median follow-up period was 45 months. Patient characteristics are summarized in Table 1. The cases comprised 34 men and 31 women; the mean patient age at treatment was 60.7 years. In most cases (n = 52, 80.0%), patients had Karnofsky performance score of 70 or above. The cohort mainly consisted of lung (n = 38, 58.4%) and breast (n = 14, 21.6%) cancers, while colorectal and urological cancers were also represented.
In most cases, patients presented with a symptomatic brain metastasis (n = 59, 90.8%) and were prescribed steroids for symptoms (n = 39, 60.0%). The metastasis was most commonly supratentorial (n = 50, 76.9%). The mean size of tumor in pre-operative imaging was 3.51 cm.
2. Treatment characteristics
Treatment characteristics are summarized in Table 2. Gross total excision was achieved in most cases (n = 57, 87.7%). The mean time from surgery to radiotherapy was 70.6 days and the mean time from planning MRI to radiotherapy was 20.6 days. There was prior radiotherapy to the brain in four cases (6.1%). In most cases, the surgical cavity was the only lesion being treated (n = 43, 66.2%). Three patients had excision of two brain metastases, and both cavities were treated with SRT. In the remaining cases, there was at least one other non-excised metastasis on pre-treatment MRI, and all such lesions were treated with SRT (these lesions were not included in analysis). The maximum number of lesions treated in a single patient was nine.
Most cavities received fractionated treatment (n = 58, 89.2%). The most common dose prescription was 30 Gy/5 fractions (Fr) (n = 49, 75.4%), followed by 18 Gy/1 Fr (n = 5, 7.7%), 25 Gy/5 Fr (n = 3, 4.6%), and 27.5 Gy/5 Fr (n = 3, 4.6%). The median target/treatment volume was 2.80 cm3/3.76 cm3 for single-fraction treatment, and 7.88 cm3/10.3 cm3 for fractionated treatment. The median conformity index was 1.36 (range, 1.08 to 1.82).
3. Outcomes
At the time of analysis, there were 25 local failure events, 36 distant failure events, and 40 patient deaths. Estimates of 1-year LCR, DBCR, and OS were 69.0%, 51.9%, and 71.0%, respectively (Fig. 1A–C). There were 10 cases (15.4%) of leptomeningeal (LM) relapse. There were four cases of RN (6.2%). Patient and treatment information for these cases are presented in Table 3.
4. Prognostic factors
We performed univariate analysis to look for patient, tumor, and treatment factors of prognostic significance with regards to local control, OS (Table 4), distant brain control, and LM relapse (Supplementary Table S1). Modeled as a continuous variable, biological effective dose using α/β ratio of 10 (BED10) was a significant positive prognostic factor for local control (hazard ratio [HR], 0.862; 95% confidence interval [CI], 0.787 to 0.944; p = 0.001). Controlled extracranial disease (defined as a controlled primary tumor plus absence of extracranial metastasis) and BED10 were significant positive prognostic factors for OS. Negative prognostic factors for OS include target volume (HR, 1.044; 95% CI, 1.001 to 1.089) and multiple lesions treated in the same session (HR, 1.907; 95% CI, 1.009 to 3.607; p = 0.047).
We performed multivariate analysis to further investigate prognosticators for local control. Pre-operative dural involvement, target volume, and BED10 met the thresholds for inclusion in multivariate analysis for local control (Table 5). Multivariate modeling with Cox proportional hazard model showed that BED10 remains to be a significant prognosticator (HR, 0.890; 95% CI, 0.803 to 0.986; p = 0.026) for improved local control after adjusting for pre-operative dural involvement and target volume.
We compared local control between cavities that received BED10 ≥45 Gy and those receiving BED10 <45 Gy using Fine-Gray model to account for the competing risk of death without local failure (Fig. 2). Cavities that received BED10 ≥45 Gy had significantly better local control than cavities that received BED10 <45 Gy (p = 0.002).
5. Effect of fractionation on local control and RN
We hypothesized that fractionation would reduce the risk of RN and that local control was maintained as long as BED10 ≥45 Gy. We excluded cases involving re-irradiation (either whole-brain radiotherapy or stereotactic radiotherapy to the same site) in this analysis. There was a significant difference in the risk of RN between SRS and FSRT subgroups (SRS, 28.6%; FSRT, 1.8%; p = 0.03, Fisher’s exact test). For cavities within this cohort that received BED10 ≥45 Gy, there no significant difference in local control between single-fraction and fractionated subgroups (p = 0.542) (Fig. 3).
Discussion and Conclusion
Our study represents a real-world cohort of brain metastasis patients managed in a public sector, tertiary hospital setting. The fact that most patients had a symptomatic, sizable tumor reflects case selection criterion for neurosurgical excision. While cavities receiving fractionated treatment tended to have larger target volume, the mean target volume for fractionated treatment was still relatively low, reflecting physician preference in our locality.
Studies have previously investigated the relationship between radiation dose and local control in stereotactic cavity irradiation. Prospective evidence is only available for SRS but not FSRT. In the prospective trial reported by Mahajan et al. [5], prescription dose was determined by SRS target volume (16 Gy for <10 cm3, 14 Gy for 10.1-15 cm3, and 12 Gy for >15 cm3). In the cohort treated with SRS, patients with tumors up to 2.5cm had 1-year local control of 100% and patients with tumors >3.5 cm had 1-year local control of 72%. Additionally, Luther et al. [22] and Iorio-Morin et al. [23] showed retrospectively that SRS margin dose and maximum dose, respectively, were correlated with risk of local failure.
Only retrospective evidence is available for FSRT. Kumar et al. [24] analyzed 43 surgical cavities that received Cyberknife FSRT of 3 or 5 fractions and found that a higher BED10 significantly improved local control; a threshold BED10 of 48Gy was identified using partition analysis. Garimall et al. [25] analyzed 144 cavities that received linear accelerator FSRT in 2–10 fractions and found EQD210 (equivalent doses in daily 2 Gy fractionation assuming an α/β ratio of 10) to be significant prognosticator of local control. Traylor et al. [26] analyzed 67 patients that received cavity Gamma Knife FSRT in 3 or 5 fractions and found BED10 to be a significant factor in local control and demonstrated that BED10 >48 Gy led to better outcomes. Musunuru et al. [27] analyzed 63 cavities and showed that 25 Gy/5 Fr was inferior to higher doses (30–36 Gy in 5–6 Fr).
In our study, we analyzed a similar number of cavities and were able to show that BED10, modeled as a continuous variable, was associated with superior local control in both univariate and multivariate analysis. We also showed that a BED10 ≥45 Gy was associated with significantly better local control.
Additionally, our cohort had a low incidence of RN. The 1-year risk of RN can be up to 28% [28] in large post-operative cavities. Risk of RN is likely a function of radiation dose, volume, and fractionation. For example, Minniti et al. [29] showed that FSRT of 27 Gy/3 Fr had reduced risk of RN (and better local control) compared to single-fraction SRS in the treatment of intact brain metastases >2 cm. While recommendations have been made [30,31] for dose constraints to normal brain tissue (e.g., D10cc <12 Gy for single-fraction treatment), surgical cavities are unique in that the target volumes mainly include non-brain tissue (e.g., fluid). We were able to show that cavities receiving single-fraction treatment were at higher risk of RN. The cavities in our cohort were relatively small in patients receiving fractionated treatment (median, 7.88 cm3), suggesting that the benefit of fractionation may not be limited to larger target volumes.
Overall, prospective data on post-operative FSRT is lacking. The ongoing Alliance A071801 (NCT04114981) [32] phase 3 study is comparing post-operative SRS to FSRT and may provide further clarity in this field. While awaiting the results of this trial, our study suggests a fractionated radiotherapy regimen with BED10 ≥45 Gy (such as 30 Gy/5 Fr) may be preferred in the setting of excised brain metastases. Our findings are consistent with International Stereotactic Radiosurgery Society guidelines [20] which state that a prescription dose of EQD210 30–50 Gy (equivalent to BED10 36–60 Gy) is associated with acceptable rates of local control. Additionally, a meta-analysis [33] of 50 studies found that patients treated with FSRT had better local control than those treated with SRS. However, fractionated regimens are less cost-effective and more cumbersome for patients due to the need for multiple sessions of treatment. There is likely a cavity size in which single-fraction treatments provide adequate local control while keeping RN rates acceptably low, and further studies may help to find this balance.
Our study had several limitations. Firstly, the 1-year LCR in our cohort was relatively low and may be related to the relatively long time between surgery, planning MRI, and start of treatment. While the optimal timing of radiotherapy is not well-defined, an expert review [34] suggested treatment within 4 weeks. Time between surgery and radiotherapy generally ranged from 30–50 days in other studies [5,24,25]. Secondly, local failure and RN were defined radiologically. Differentiating tumor progression from RN or pseudoprogression is notoriously challenging due to similar conventional radiographic features. Thirdly, single-fraction SRS represented a relatively small proportion of our cohort, and most cases received a prescription of 30 Gy/5 Fr. Finally, although an effort was made to correct for exposure to any effective systemic treatment, patients with metastatic cancer typically receive multiple lines of systemic therapy for variable durations, and it remains difficult to quantify the relationship between systemic therapy and intracranial disease control separately.
Our study adds to the existing body of evidence that a prescription dose of BED10 ≥45 Gy, such as 30 Gy/5 Fr, is associated with an acceptable LCR and a potentially lower incidence of RN compared to single-fraction treatment. Adequate BED10 has to be maintained even in larger cavities after carefully balancing the risk of local treatment failure against the risk of RN.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation 
Print 
