Salvage resection of recurrent previously irradiated brain metastases: tumor control and radiation necrosis dependency on adjuvant re-irradiation

Abstract

Purpose

The efficacy of salvage resection (SR) of recurrent brain metastases (rBrM) following stereotactic radiosurgery (SRS) is undefined. We sought to describe local recurrence (LR) and radiation necrosis (RN) rates in patients undergoing SR, with or without adjuvant post-salvage radiation therapy (PSRT).

Methods

A retrospective cohort study evaluated patients undergoing SR of post-SRS rBrM between 3/2003–2/2020 at an NCI-designated cancer center. Cases with histologically-viable malignancy were stratified by receipt of adjuvant PSRT within 60 days of SR. Clinical outcomes were described using cumulative incidences in the clustered competing-risks setting, competing risks regression, and Kaplan–Meier methodology.

Results

One-hundred fifty-five rBrM in 135 patients were evaluated. The overall rate of LR was 40.2% (95% CI 34.3–47.2%) at 12 months. Thirty-nine (25.2%) rBrM treated with SR + PSRT trended towards lower 12-month LR versus SR alone [28.8% (95% CI 17.0–48.8%) versus 43.9% (95% CI 36.2–53.4%), p = .07 by multivariate analysis]. SR as re-operation (p = .03) and subtotal resection (p = .01) were independently associated with higher rates of LR. On univariate analysis, tumor size (p = .48), primary malignancy (p = .35), and PSRT technique (p = .43) bore no influence on LR. SR + PSRT was associated with an increased risk of radiographic RN at 12 months versus SR alone [13.4% (95% CI 5.5–32.7%) versus 3.5% (95% CI 1.5–8.0%), p = .02], though the percentage with symptomatic RN remained low (5.1% versus 0.9%, respectively). Median overall survival from SR was 13.4 months (95% CI 10.5–17.7).

Conclusion

In this largest-known series evaluating SR outcomes in histopathologically-confirmed rBrM, we identify a significant LR risk that may be reduced with adjuvant PSRT and with minimal symptomatic RN. Prospective analysis is warranted.

Introduction

As long-term cancer outcomes improve, the need to achieve durable brain metastasis (BrM) control is increasing [1, 2]. Stereotactic radiosurgery (SRS) is the preferred treatment modality for small BrM not amenable to resection, with resection plus SRS preferred for larger, symptomatic and accessible tumors [3-6]. However, SRS fails in 10–20% [7-9] and induces radiation necrosis (RN) in up to 24% of cases [10, 11]. As such, the CNS is increasingly a site of refractory, recurrent disease as patients live longer with cancer [12-15], with significant implications for quality of life, mortality, and healthcare expenditure [16-18].

Post-SRS recurrent BrM (rBrM) treatment options include re-irradiation, salvage resection (SR), or CNS-active therapeutics, depending on several tumor- and patient-specific characteristics. Re-irradiation confers 54–79% 1 year local control, the benefit of which must be balanced with the risk of compounded radiotoxicity [19-24]. Alternatively, resection allows for palliation in cases with larger, symptomatic recurrences as well as histopathologic and molecular review. However, SR outcomes, local recurrence (LR) rates, and predictors of such are poorly characterized, in part owing to historically short survival on the order of 1 year [25]. Furthermore, while post-operative irradiation considerably reduces the high rate of post-resection recurrence in the upfront setting [26-29], the role of post-SR adjuvant re-irradiation is unclear given the potential risk of cumulative radiation toxicity.

We sought to define rates of LR and RN following SR of previously SRS-treated rBrM, and determine the impact of post-salvage radiation therapy (PSRT) on clinical outcomes. We further sought to compare the risks of recurrence and necrosis by tumor viability and treatment histories.

Methods

Patient selection

A retrospective, single institution, cohort analysis identified patients with previously SRS-treated rBrM that underwent SR at an NCI-designated cancer center between March 2003 and February 2020, with histologically-confirmed recurrent tumor on pathology (Fig. 1). Patients with prior resection plus adjuvant SRS were included given the ablative nature of this modality. Patients who only received prior whole-brain radiation therapy (WBRT) were excluded due to the relatively lower risk of RN with re-irradiation. Our institutional practices are generally to reserve SR for those with oligometastatic active CNS disease and for the primary indication(s) of palliation, local control, and/or tissue diagnosis.

Fig. 1.

Retrospective study design in categorizing outcomes for rBrM undergoing SR. Flowchart outlining the retrospective, single institution analysis of 135 patients with 155 recurrent, previously irradiated BrM undergoing SR. Resection sites were characterized by the administration of adjuvant PSRT within 60 days of SR, and followed longitudinally until the time of LR, RN, or continued site-specific stability at death or last follow up. BrM brain metastases; rBrM recurrent brain metastases; SRS stereotactic radiosurgery; PSRT post-salvage radiation therapy; RANO-BM response assessment in neuro-oncology brain metastases; LR local recurrence; RN radiation necrosis; SR salvage resection

Clinical information

Patient demographics, treatment histories, and clinical, pathologic and radiographic courses were collected. Tumor dimensions were measured from gadolinium-enhanced MRI. Extracranial disease status was determined using last-available body CT or PET scan prior to SR, time of death and/or last follow-up.

The proportion of histologically-preserved, viable-appearing tumor at SR was estimated by a board-certified neuropathologist. Based on this, rBrM were assigned to two groups: (1) tumor-prevalent, and (2) tumor-rare lesions with extensive treatment effect and rare viable-appearing disease (≤ 10% of total evaluable material). Extent of SR (EOR) was classified as gross-total (GTR), near-total (> 90%, NTR), or subtotal (STR) based on post-operative neuroradiologist interpretation.

Adjuvant PSRT was defined as any radiation delivered ≤ 60 days post-SR. This cutoff was selected to complement published adjuvant radiation practices [26]. PSRT was categorized as single-fraction SRS (SF-SRS), hypofractionated SRS (HF-SRS), partial-brain radiotherapy (PBRT), WBRT, and WBRT + resection-cavity boost. The PSRT technique employed was determined contemporaneously by the treating radiation oncologist.

Recurrent local enhancement at the post-operative site was evaluated using RANO-BM [30]. The immediate post-operative MRI, or that which demonstrated a nadir in post-operative contrast-enhancement to account for postsurgical changes, served as the baseline post-SR scan. Progression was defined as ≥ 20% increase in a locally-enhancing nodule plus ≥ 5 mm expansion of the longest diameter.

Enhancing processes were further categorized as viable-appearing LR, including potential treatment effect admixed with recurrent tumor, versus pure RN using 3 modalities in order of decreasing value: (1) tissue confirmation, (2) advanced neuroimaging incorporating perfusion and/or metabolic characteristics, and (3) longitudinal contrast-enhanced MRI with clinical correlation. Histopathologic confirmation was the gold standard among the subset that underwent sampling. Advanced neuroimaging consisted of plasma volume maps calculated from dynamic contrast-enhanced T1 MRI perfusion, relative cerebral blood volume maps calculated from dynamic susceptibility contrast-enhanced T2* MR perfusion (sensitivity 74–95%, specificity 78–97%) [31], and/or FDG-PET brain (sensitivity 81–86%, specificity 40–94%) [32]. If neither tissue nor advanced imaging were available, conventional MRI in combination with expert consensus and clinical trajectory were prioritized [33]. In ambiguous cases due to imaging discordance or atypical radiographic evolution, classification was adjudicated by a blinded neuroradiologist.

Statistical analysis

The cohort and lesions under study were characterized using simple, descriptive characteristics such as frequencies, medians, and ranges. Rates of LR, RN, and adjuvant PSRT following SR were estimated using cumulative incidence in the clustered competing risks setting where follow-up time was calculated from SR until event of interest, death, or last follow-up. Univariable and multivariable competing risks regression with sandwich-based variance adjustment for multiple lesions per patient were performed to estimate hazard ratios (HR) and 95% confidence intervals (CIs) for the association of variables of interest with LR and RN. Univariable logistic regression was performed to determine which variables were associated with receiving early adjuvant PSRT. Kaplan–Meier methodology was utilized to characterize overall survival (OS), and to investigate OS stratified by adjuvant PSRT and by the histopathologic degree of viable-appearing tumor. Follow-up time for OS was calculated from SR until death for events or last follow-up for those who were censored. Univariable Cox proportional hazards regression was performed to estimate HRs and 95% CIs for variables of interest and their association with OS. Statistical analysis was performed using R version 4.0.1 and SAS version 9.4.

Results

Patient demographics

Four hundred two post-SRS surgical BrM cases were screened. Following exclusion for pure RN, incomplete history, or lack of post-SR imaging, 155 rBrM in 135 patients were included.

Metastases derived from non-small-cell lung (37.0%), melanoma (25.9%), breast (20.7%), renal (4.4%), colorectal (2.2%) and other (9.6%) primary malignancies (Table 1). Twenty-five (16.1%) BrM received prior upfront resection, and all BrM received ≥ 1 course of SRS prior to SR. Prior site-directed treatments consisted of 1 SRS course (74.8%), 2 SRS courses (1.3%), SRS + WBRT (7.7%), upfront resection + SRS (12.9%), and upfront resection + SRS + WBRT (3.2%) (Table 1, Supplementary Table 1).

Table 1.

Demographics of patients undergoing sr of previously-irradiated rBrM

Demographics	Total cohort, No. (%)	Adjuvant PSRT, No. (%)	Observation, No. (%)	p-valuea
Total patients	135 (100)	33 (24)	102 (76)	0.72
Age, median (range)	58 (20–92)	56 (33–81)	59 (20–92)
Male	49 (36.3)	17 (43.6)	40 (33.9)	0.31
Female	86 (63.7)	22 (56.4)	76 (64.4)
Primary malignancy
NSCLC	50 (37.0)	11 (33.3)	39 (38.2)	0.58
Melanoma	35 (25.9)	10 (30.3)	25 (24.5)
Breast	28 (20.7)	8 (24.2)	20 (19.6)
RCC	6 (4.4)	0 (0.0)	6 (5.9)
CRC	3 (2.2)	2 (6.1)	1 (1.0)
Other	13 (9.6)	2 (6.1)	11 (10.8)
Total rBrM	155 (100.0)	39 (25.2)	116 (74.8)
Laterality
Left	83 (53.5)	20 (51.3)	63 (54.3)	0.73
Right	70 (45.2)	18 (46.1)	52 (44.8)
Midline	2 (1.3)	1 (2.6)	1 (0.1)
Location
Frontal	48 (31.0)	11 (28.2)	37 (31.9)	0.81
Parietal	35 (22.6)	10 (25.6)	25 (21.6)
Occipital	20 (12.9)	5 (12.8)	15 (12.9)
Temporal	23 (14.8)	4 (10.3)	19 (16.4)
Cerebellar	29 (18.7)	9 (23.1)	20 (17.2)
Supratentorial	126 (81.3)	30 (76.9)	96 (82.8)	0.43
Infratentorial	29 (18.7)	9 (23.1)	20 (17.2)
Parenchymal	143 (92.3)	36 (92.3)	107 (92.2)	0.99
Dural	12 (7.7)	3 (7.7)	9 (7.8)
Pre-SR Treatment
SRS, one course	116 (74.8)	30 (76.9)	86 (74.1)	0.47
SRS, two courses	2 (1.3)	0 (0.0)	2 (1.7)
SRS + WBRT	12 (7.7)	1 (2.6)	11 (9.5)
Surgery + SRS	20 (12.9)	8 (20.5)	12 (10.3)
Surgery + SRS + WBRT	5 (3.2)	0 (0.0)	5 (4.3)

SR salvage resection; rBrM recurrent brain metastasis; NSCLC non-small cell lung cancer; RCC renal cell carcinoma; CRC colorectal carcinoma; SRS stereotactic radiosurgery; WBRT whole-brain radiation therapy

^aP-values from logistic regression estimating odds of adjuvant PSRT. Type III p-values are provided for primary malignancy, laterality, location and pre-SR treatment covariates

Salvage resection

Median age at SR was 58 years (range 20–92) with a female predominance (63.7%) and median KPS of 80 (range 50–100). The median number of BrM at SR was 2 (range 1–20) and extracranial metastatic disease was at least stable in 72.9% of cases. The median time from first irradiation course to SR was 9.8 months (range 0.4–56.2). SR indications were attempted local control (100%), palliation (79.4%), and diagnostic confirmation (7.1%). Sixty-nine percent of rBrM were tumor-prevalent and 31.0% were tumor-rare; 83.9% underwent GTR or NTR.

Adjuvant post-salvage radiation therapy

Thirty-nine (25.2%) post-operative cavities underwent adjuvant PSRT within 60 days (median 41, range 23–58). PSRT consisted of SF-SRS (5.1%), HF-SRS (41.1%), PBRT (17.9%), WBRT (33.3%), and WBRT + resection-cavity boost (2.6%) (Supplementary Table 2).

Those treated with adjuvant PSRT were more likely to have received only 1 prior RT course (OR 6.98, 95% CI 1.37–128, p = .015), have tumor-prevalent histology at SR (OR 4.01, 95% CI 1.57–12.4, p = .003), and trended in favor of patients with KPS ≥ 80 (OR 2.12, 95% CI 0.97–4.95, p = .06). There were no differences in age (p = .72), primary malignancy (p = .58), or EOR at the time of SR (p = .40) among those who received adjuvant PSRT.

Predictors of local recurrence following salvage resection

Post-SR local enhancement developed in 78 resection cavities (50.3%). Cases were categorized as LR (n = 68) and RN (n = 10). This stratification was supported primarily by tissue confirmation in 25.6%, advanced neuroimaging in 52.6%, and conventional MRI in 21.8%.

The 12-month rate of LR for the entire cohort was 40.2% (95% CI 34.3–47.2%). At 6 and 12 months, LR occurred in 35.9% and 43.9% of rBrM in the observation group and in 18.0% and 28.8% in the adjuvant PSRT group, respectively (Fig. 2a, Supplementary Table 3). On multivariate analysis controlling for the degree of viable tumor, EOR, tumor size, and prior resection, this improvement in local control trended towards, but did not reach, statistical significance (HR 0.60, 95% CI 0.34–1.04, p = .07, Table 2). The PSRT modality of choice (WBRT/PBRT versus SRS) bore no influence on subsequent local control (p = .43).

Fig. 2.

Cumulative Incidence Rates of LR versus RN Following SR of Previously-Irradiated rBrM. Cumulative incidence curves showing the rates of LR (a) versus RN (b) within the resection cavities following SR of previously-irradiated rBrM. Resection cavities were stratified by whether they were monitored conservatively (Obs, solid lines) or received adjuvant PSRT within 60 days of SR (Adjuvant PSRT, dotted lines). LR local recurrence; RN radiation necrosis; rBrM recurrent brain metastases; Obs observation; PSRT post-salvage radiation therapy; SR salvage resection

Table 2.

Predictors of LR and RN following SR

Treatment-specific factors	HR	95% CI	p-value
Univariate analysis
Risk of local recurrence
Adjuvant PSRT versus observation	0.71	0.41–1.22	0.21
SR as re-operation versus initial resection	1.66	0.98–2.81	0.06
GTR/NTR versus STR	0.44	0.24–0.80	0.007
Tumor-prevalent versus tumor-rare histology	1.33	0.77–2.28	0.31
Tumor size ≥ 3 cm versus < 3 cm	0.84	0.52–1.37	0.48
Risk of radiation necrosisa
Adjuvant PSRT versus observation	4.55	1.26–16.39	0.02
SR as re-operation versus initial resection	3.84	1.10–13.39	0.035
Tumor-prevalent versus tumor-rare histology	0.67	0.19–2.37	0.53
Tumor size ≥ 3 cm versus < 3 cm	1.26	0.37–4.33	0.71
Multivariate analysis
Risk of local recurrence
Adjuvant PSRT versus observation	0.60	0.34–1.04	0.07
Tumor-prevalent versus tumor-rare histology	1.31	0.75–2.28	0.34
SR as re-operation versus initial resection	1.79	1.07–2.99	0.03
GTR/NTR versus STR	0.43	0.24–0.78	0.01
Tumor size ≥ 3 cm versus < 3 cm	0.78	0.47–1.29	0.33

Statistically significant p-values are bolded

^aGTR/NTR versus NTR did not converge in the radiation necrosis model and therefore is not shown in the table above LR local recurrence; RN radiation necrosis; HR hazard ratio; CI confidence interval; STR subtotal resection; GTR gross-total resection; NTR near-total resection; PSRT post-salvage radiation therapy

Additional LR predictors after SR are outlined in Table 2. GTR/NTR was associated with lower LR versus STR (HR 0.44, 95% CI 0.24–0.80, p = .007). Tumor size (p = .48), primary malignancy type (p = .35), and location (p = .94) were not predictive of LR by univariate analysis. On multivariate analysis, SR as a re-operation (HR 1.79, 95% CI 1.07–2.99, p = .03) and GTR/NTR (HR 0.43, 95% CI 0.24–0.78, p = .01) were independently predictive of LR. Tumor-prevalent histology trended toward a higher LR rate on multivariate analysis (HR 1.31, 95% CI 0.75–2.28, p = .34) but did not reach significance.

Predictors of radiation necrosis following salvage resection

The 6- and 12-month rates of RN were 3.5% (95% CI 1.5–8.0%) and 3.5% (95% CI 1.5–8.0%) in the observation group and 2.6% (95% CI 0.4–18.6%) and 13.4% (95% CI 5.5–32.7%) in the adjuvant PSRT group, respectively (Fig. 2b, Supplementary Table 3). Between 6 and 12 months post-SR, no additional RN events occurred in the observation group, however a significant rise in RN occurred in the PSRT group compared to those observed (HR 4.55, 95% CI 1.26–16.39, p = 0.02). Of these, RN was symptomatic in 2 patients treated with adjuvant PSRT (5.1% of the PSRT cohort) versus 1 that was observed (0.9% of the observation cohort).

SR as a re-operation was also a risk factor for subsequent RN on univariate analysis (HR 3.84, 95% CI 1.10–13.39, p = .035, Table 2). Neither tumor-prevalence (p = .53) nor tumor size at SR (p = .71) were predictive of RN. All lesions that developed RN in the PSRT cohort had received either adjuvant SRS or PBRT. However, the relative risk of specific PSRT modalities on its development cannot be reliably estimated given the rarity of RN.

To account for potential ambiguity in resolving LR and RN, the cumulative incidence and risk factors for any local enhancing changes are provided in Supplementary Fig. 1 and Supplementary Tables 4 and 5.

Neurologic and survival outcomes

The management of local progressive enhancement following SR included additional index-site surgery and/or focal radiation (n = 29), WBRT (n = 10), and systemic therapy change (n = 28). Despite these interventions, refractory progression persisted in 20 lesions (30.3%).

Including the 24.5% (95% CI 18.9–31.7%) of surgical cavities that received early adjuvant PSRT, the cumulative incidence of any site-directed irradiation was 31.1% at 6 months (95% CI 24.9–38.9%) and 37.0% at 12 months (95% CI 30.0–45.7%; Supplementary Fig. 2a). All radiation delivered after the 60-day adjuvant PSRT window was performed in the setting of progressive index-site disease, with or without distant CNS failure. Tumor-prevalent rBrM, in addition to receiving higher adjuvant PSRT rates, also required more site-directed radiation treatments over time than tumor-rare lesions (HR 2.72, 95% CI 1.37–5.41, p = .004, Supplementary Fig. 2b). The 6- and 12-month rates of any site-directed irradiation were 38.2% (95% CI 30.1–48.5%) and 44.9% (95% CI 37.1–54.5%) for tumor-prevalent, and 14.7% (95% CI 8.4–25.9%) and 18.9% (10.8–33.3%) for tumor-rare rBrM.

Median OS from SR was 13.4 months (95% CI 10.5–17.7; Fig. 3a). No OS difference was observed based on receipt of adjuvant PSRT (p = .80, Fig. 3b), EOR (p = .20), or primary malignancy (p = .20). Histopathologic classification demonstrated a divergent survival trend, with a median OS of 11.1 months (95% CI 8.65–14.6) for tumor-prevalent rBrM and 17.7 months (95% CI 12.83–36.9) for tumor-rare rBrM (Fig. 3c), but this did not reach statistical significance (HR 1.33, 95% CI 0.89–2.0, p = .20).

Fig. 3.

Survival Probability Curves Among Patients Undergoing SR of Previously-Irradiated rBrM. Kaplan–Meier curves demonstrating OS statistics from the time of SR of previously-irradiated rBrM. a Median OS of the entire cohort was 13.4 months. b The administration of early adjuvant PSRT within 60 days of SR was not associated with any survival benefit, with median OS of 13.4 months in the observation cohort and 13.1 months in the adjuvant PSRT cohort (p = .80). c Patients with highly necrotic tumor-rare histology at SR had a longer median OS of 17.7 months, compared to 11.1 months in the tumor-prevalent histopathologic cohort, but the difference did not reach statistical significance (p = .20). SR salvage resection; rBrM recurrent brain metastases; OS overall survival; Obs observation; PSRT post-salvage radiation therapy

At last follow-up, 107 surgical sites (69.0%) were radiographically stable or better. Despite the aforementioned local control rates, neurologic death occurred in 51 patients (37.8% of total cohort), with 16 deaths (11.9%) attributable to the index rBrM. Neurologic causes of death independent of salvage treatment strategy included CNS metastatic progression (86.3%), focal or systemic treatment complications (5.9%), CNS infection (3.9%), and intratumoral hemorrhage (3.9%). Twenty-two patients (16.3%) developed leptomeningeal carcinomatosis, the rate of which was not associated with adjuvant PSRT exposure (p = .99). Sixty-seven patients (49.6%) had stable or responsive extracranial disease at death or last follow-up.

Discussion

BrM are increasingly treated with SRS given its safety, convenience and efficacy. However, when SRS fails, salvage commonly hinges on further local measures, with accessible BrM often undergoing SR. Only limited data informs long-term surgical outcomes and determinants in these heavily pre-treated sites.

We show that the 12-month rate of LR following SR of rBrM is 40.2% and approaches that of BrM resection in the upfront setting. The addition of adjuvant PSRT resulted in a trend towards improved 6- and 12-month local control compared to SR alone. Adjuvant irradiation modality did not correlate with recurrence, with SRS equally effective to PBRT and WBRT. Primary malignancy and metastasis size, which influence recurrence in the upfront setting [26, 27], were not correlated with progression in the salvage setting within the powering of this study. It is possible that a more robust sample size or prospective analysis might result in statistically significant effects, however we are unable to determine if our results are due to a lack of power or a true null result. When compared with the 46–50% recurrence rate for upfront BrM resection [26, 34] and the 10–28% seen with the addition of post-operative irradiation [26, 27, 34], the control rates herein described are reassuring. These data further suggest that local salvage therapies for BrM with refractory, recurrent biology is reasonable and carries lasting benefit. Indeed, our finding of substantially improved local control with GTR or NTR supports the approach of maximal safe resection. This conclusion is also supported by the 13.4 month survival in this cohort, comparable to that of the upfront BrM resection literature.

Importantly, adjuvant PSRT was well-tolerated. The feared risk of compounded radiation toxicity leads to clinical uncertainty in the salvage post-SRS setting, and in fact the utilization of adjuvant re-irradiation in our series was limited to only 25% despite identified viable malignancy in all cases. While patients who received adjuvant PSRT were significantly more likely to develop RN with a 13.4% 1 year incidence, this was rarely symptomatic. This necrosis rate is on par with or less than repeat salvage radiosurgery alone, with reported toxicity rates of 11–24% in smaller metastases than those treated herein [19-21].

Given this trade-off of countervailing risks and the importance of identifying patients who might benefit from adjuvant PSRT, we sought to identify correlates of LR and RN. Despite the common diagnosis of admixed treatment effect at SR, the relationships between the degree of viable malignancy within the resected tumor and subsequent recurrence or necrosis have not been previously described. We hypothesized that highly-viable disease would carry a higher recurrence risk, and that adjuvant PSRT would be reserved for these cases given the theoretical compounded risk of radiation injury and potential lessened recurrence risk in more necrotic cases. Indeed, tumor-prevalent metastases trended towards higher recurrence on multivariate analysis when controlling for adjuvant PSRT, however this did not reach significance. Furthermore, tumor-prevalent cases were significantly more likely to require delayed site-directed irradiation than tumor-rare cases. These findings indicate that highly-viable resected rBrM require more focal therapies and may confer a more aggressive phenotype than highly-necrotic resected rBrM, an observation worthy of prospective analysis with quantitative histopathologic evaluations. Tumor-rare histology conversely trended towards longer OS, potentially reflecting more radiosensitive biology. The risks versus benefits of post-operative irradiation to these lesions remain unclear since they were under-represented in our adjuvant PSRT cohort. Finally, salvage as a re-operation was significantly associated with both LR and RN, a finding that may be reflective of the aggressiveness and size of such metastases. Surgically-resected BrM are typically larger than those managed with SRS alone, which is a known risk factor for both subsequent recurrence and for development of RN [35, 36]. The devitalization of normal brain has also been posited to contribute to RN [37].

A high proportion of neurologic death implicates the CNS as the rate-limiting site of disease for many patients. While extrapolation of these figures to the brain metastasis population is challenging given the inherent biases of a pre-treated population at a specialized center, this finding underscores the tenacity of CNS-refractory disease and supports a role for continued innovation in multimodality palliative approaches. As CNS metastases are increasingly common as patients live longer with cancer [2], the development of effective, minimally-toxic salvage strategies will become more important. For example, cesium-131 brachytherapy has been shown to confer promising control for newly diagnosed and large BrM [38, 39] and with minimal toxicity [40]. Ongoing studies are evaluating this prospectively (NCT04690348) [41]. Percutaneous laser-induced thermal therapy allows concurrent biopsy and has been studied for treating both tumor- and necrosis-predominant recurrences [42, 43], although retrospective series suggest relatively high complication and tract recurrence rates [44, 45].

This study has several limitations. Only one-quarter of the resected rBrM in our series underwent adjuvant PSRT, and this small sample size might have influenced LR and RN outcomes. A larger cohort with comparable rates of available radiation modalities would better inform the impact on long-term control. Our surgical patient cohort was relatively young and high-functioning, in keeping with retrospective surgical series with inherent selection biases. To minimize the potential for inter-pathologist variability in histopathologic review, we limited our analysis to broad qualitative categorization of “tumor-rare” versus “tumor-prevalent” lesions; a more quantitative approach may have allowed for more robust analysis. Similarly, categorization of radiographic enhancing changes as tumor versus treatment effect remains imprecise even with best-available advanced neuroimaging and careful discrimination of pure RN from tumorcontaining lesions, given the ramifications of CNS progression in such patients. Finally, cancer treatments have evolved significantly in the period covered by this analysis. In the latter years of this dataset, WBRT utilization fell by half and modern CNS-active systemic therapies (bevacizumab, immune checkpoint blockade, tyrosine kinase inhibition) began to rise, reflecting advances in cancer therapeutics and also institutional treatment bias [46, 47]. Given this high degree of therapeutic heterogeneity encompassing both local and systemic treatments, the influence of cancer-directed therapies on site-specific outcomes and survival cannot be reliably determined. Despite these limitations, the outcomes from this analysis are informative, underscore the complex interplay of approaches, and highlight the need for better management strategies for the growing problem of rBrM.

Conclusions

Management of previously-irradiated rBrM is challenging. In this largest-known retrospective series of pathologically-confirmed rBrM, the overall recurrence rate after post-SRS SR approached that of resection in the upfront setting. A trend towards improved local control was observed in those that received SR plus adjuvant re-irradiation compared to SR alone; the risk of symptomatic RN was also low among these patients. Prospective analysis is needed to define the role of adjuvant re-irradiation and to optimize patient selection for this technique. Tumor-prevalent histology correlated with long-term re-irradiation requirements and may serve as a biomarker for more aggressive clinical trajectory.