Multi-institutional Outcomes after Stereotactic Radiosurgery for Gastrointestinal Brain Metastases

Abstract

Purpose

To compare outcomes between gastrointestinal and nongastrointestinal patients with brain metastases after radiosurgery.

Methods and Materials

Retrospective cohort study identifying patients completing an initial course of radiosurgery between January 2015 and December 2020, with follow-up data collected through November 2022. Multi-institutional, academic referral centers. The primary outcomes were overall survival and intracranial progression-free survival, calculated by the Kaplan–Meier method. Progression was defined as concern on postradiosurgery imaging for recurrence determined by clinical multidisciplinary consensus. Cox proportional hazard models were used to assess associations between outcomes and covariates.

Results

This study included 1281 nongastrointestinal patients and 102 gastrointestinal patients, of which 45.1% were colorectal, 33.3% esophageal, and the remaining 21.6% comprising other sites. Gastrointestinal patients were more likely to be younger (mean 59.1 vs 63.5 years, P = .001), male (56.9% vs 44.3%, P = 0.014), have received systemic therapy (73.5% vs 63.9%, P = .049), and have resection of brain metastases (45.1% vs 25.0%, P < .001) prior to radiosurgery. Median overall survival was lower for gastrointestinal patients at 5.4 months (95% CI, 3.8-7.7) versus nongastrointestinal patients at 10.6 months (95% CI, 9.3-11.6, P < 0.0001). In a multivariate model, gastrointestinal patients had worse overall survival compared to nongastrointestinal patients (hazard ratio, 1.92; P < .0001; 95% CI, 1.53-2.41). Median intracranial progression-free survival was lower for gastrointestinal patients at 6.2 months (95% CI, 4.0-9.6) versus nongastrointestinal patients at 12.3 months (95% CI, 10.8-13.9; P = 0.0002). In a multivariate model, gastrointestinal patients had worse intracranial progression-free survival compared to nongastrointestinal patients (hazard ratio, 1.60; 95% CI, 1.20-2.14; P = 0.0013). There were no significant differences between colorectal primary patient or esophageal primary patient outcomes compared to all other gastrointestinal primary patients.

Conclusions

Across a multi-institutional stereotactic radiosurgery cohort, brain metastases of gastrointestinal origin demonstrated inferior overall survival and intracranial progression-free survival to those of nongastrointestinal origin. These data may help inform treatment decisions and postradiosurgery surveillance.

Introduction

Up to 40% of all cancer patients will develop brain metastases (BMs).¹ Standard of care typically includes stereotactic radiosurgery (SRS), with or without surgical resection.² Rates of treated tumor control after SRS are typically very high, with known increase in the risk of distant intracranial progression (ICP) compared to whole brain radiation therapy. Typical treatment and surveillance recommendations tend to be relatively uniform irrespective of patient specific features, and although tools have been developed for patient specific care, little is known on how these apply to patients with BMs from gastrointestinal (GI) primaries.3, 4, 5

Whereas GI cancers represent about 10% of new cancer diagnoses in the United States, the rate of BM in this population is relatively low, with most studies reporting rates of < 10%.⁶ Although few data exist to describe the population of patients with GI BM, prognosis is generally thought to be relatively poor and is influenced by performance status, age, extracranial metastases, and number of BMs.7, 8, 9 Additionally, little is known regarding outcomes of these patients following standard therapy, like SRS, and appropriate posttreatment surveillance.

To better understand the prognosis and outcomes of patients with GI BMs we conducted an analysis of our large, multi-institutional cohort of patients treated with radiosurgery. We sought to understand rates of treated tumor control, posttreatment toxicity, and rates of distant intracranial progression. Herein we present our findings.

Methods and Materials

With permission of our institutional IRB (00108434), we retrospectively identified all patients completing an initial SRS course across 2 institutions from January 2015 through December 2020. Patients treated with prior resection of BMs, prior whole brain radiation therapy (WBRT), single-fraction SRS, and multifraction SRS were included. All SRS in this series was delivered via a frameless, linear accelerator technique. Dose selection was at the discretion of the treating radiation oncologist. Demographic variables, clinical variables, and dates were manually obtained via chart review. Data were collected and managed using REDCap database tools. Demographic and clinical variables were manually collected from pre-SRS records and included the following: age, sex (biological assigned at birth), race, Karnofsky performance status (KPS), number of BMs, intracranial and extracranial metastatic burden, interval between development of BMs and primary diagnosis, number of extracranial sites involved, control of extracranial disease, systemic therapy prior to SRS, and resection of brain lesions prior to SRS WBRT prior to SRS, planning target volume (PTV) of all BMs, and systemic therapy after SRS. Note that initial diagnoses were obtained from pathology reports, post-SRS ICP, post-SRS extracranial progression, and SRS associated toxicity were determined by multidisciplinary clinical consensus after review of recorded symptoms, exam findings, radiology reports, and/or any available pathology reports. Patients with missing data were excluded. The patient population and methods of variable collection have been described previously.³

Patient characteristics were summarized using median for continuous variables and with frequencies and percentages for categorical descriptors. ICP was defined as any progressive and/or new lesions concerning for recurrent BMs after multidisciplinary review of brain magnetic resonance imaging (MRI). Survival analyses began at time of SRS completion and used the Kaplan–Meier method. In the ICP analyses, patients were censored at date of death. In the intracranial progression-free survival (iPFS) analyses, any ICP or death was recorded as an event. Cox proportional hazard models were used to assess associations between overall survival (OS), iPFS, and covariates. Multivariate models were constructed in a stepwise fashion using candidate predictors and an alpha < 0.05. All analyses were performed using JMP Pro software (version 17).

Results

We identified 1383 patients who completed SRS between January 2015 and December 2020 with a median follow up of 9.93 months (95% CI, 9.08-10.99). Out of the entire cohort, 102 (7.4%) had BM from GI primaries. Of the GI primaries, 46 (45.1%) were of colorectal (CRC) origin, 34 (33.3%) were of esophageal origin, and the remaining 22 (21.6%) were from other GI sites, including anal (6.9%), pancreatic (5.9%), gastric (4.9%), GI of unknown origin (2.9%), and hepatocellular carcinoma (1.0%). Of the 1201 non-GI primaries, 54.7% (n = 757) were lung, 14.7% (n = 203) were breast, 7.9% (n = 109) were GU, 7.2% (n = 100) were skin/melanoma.

Table 1 summarizes the demographic factors and clinical parameters in the GI and non-GI cohorts. GI BM patients were more likely to be younger (mean 59.1 vs 63.5 years, P = .001), male (56.9% vs 44.3%, P = .014), to develop BM >1 year from primary diagnosis (62.7% vs 47.5%, P = .016), have more extracranial metastases (mean 1.9 vs 1.6, P = .003), have received systemic therapy (73.5% vs 63.9%, P = .049) or resection of BM (45.1% vs 25.0%, P < .001) prior to SRS, have larger PTV of all BMs (mean 20.3 cm³ vs 15.0 cm³, P = .013), and were less likely to receive WBRT prior to SRS (3.9% vs 10.8%, P = .028) or systemic therapy after SRS (54.9% vs 68.9%, P = .004).

Table 1.

Patient demographics

Variables	GI	Non-GI	P value
Total no. (%)	1281	102
Median age	65	61	.001
Sex, female (%)	714 (55.7%)	44 (43.1%)	.014
KPS ≥ 80 (%)	928 (72.4%)	74 (72.5%)	.982
Race, White (%)	950 (74.2%)	82 (80.4%)	.379
>1 BM at SRS (%)	689 (53.8%)	49 (48.0%)	.263
>5 total metastatic sites (%)	672 (52.5%)	63 (61.8%)	.070
BM developed >1 year after primary diagnosis	609 (47.5%)	64 (62.7%)	.016
Median number of extracranial sites	1	2	.003
Extracranial disease controlled, yes	439 (34.3%)	31 (30.4%)	.426
Systemic therapy prior to SRS (%)	818 (63.9%)	75 (73.5%)	.049
Resection prior to SRS (%)	315 (24.6%)	46 (45.1%)	.001
WBRT prior to SRS (%)	138 (10.8%)	4 (3.9%)	.028
Median PTV of all BM (cc)	6.1	15.3	.013
Systemic therapy after SRS (%)	882 (68.9%)	56 (54.9%)	.004

Abbreviations: BM = brain metastasis; KPS = Karnofsky performance status; PTV = planning target volume; SRS = stereotactic radiosurgery; WBRT = whole brain radiation therapy.

Demographic and disease details for gastrointestinal and nongastrointestinal primary patients.

Median OS was significantly lower for GI patients at 5.4 months (95% CI 3.8-7.7) compared to non-GI patients at 10.6 months (95% CI, 9.3-11.6; P < .0001, Fig. 1). 12-month survival for GI patients was 22.6% and it was 46.2% for non-GI patients. In a stepwise multivariate model using candidate predictors of age, sex, race, KPS, metastatic burden, systemic therapy prior to SRS, systemic therapy prior to SRS, systemic therapy after SRS, PTV of BM, resection prior to SRS, WBRT prior to SRS, and number of BMs, GI patients had significantly worse OS compared to non-GI patients (hazard ratio, 1.92; P < .0001, 95% CI, 1.53-2.41; see Table 2).

Figure 1.

Overall survival, GI versus non-GI. Overall survival of GI versus non-GI patients with brain metastases treated with stereotactic radiosurgery. Overall survival was poorer in the GI group.

Abbreviation: GI = gastrointestinal; SRS = stereotactic radiosurgery.

Table 2.

Multivariate regression, overall survival

Predictor	Hazard ratio	P value	Lower 95%	Upper 95%
GI vs non-GI primary (GI)	1.92	<.001	1.53	2.41
Age at SRS ≥ 65	1.17	.019	1.03	1.34
Sex (Male)	1.25	.001	1.1	1.42
Race (White)	1.29	.002	1.1	1.53
KPS < 80	1.73	<.001	1.5	1.99
Metastatic burden (>5 lesions)	2.05	<.001	1.78	2.36
Systemic therapy after SRS (No)	2.76	<.001	2.39	3.19
PTV of all brain metastases ≥ 10 cc	1.3	<.001	1.13	1.51
Resection prior to SRS (No)	1.76	<.001	1.47	2.1
WBRT prior to SRS (Yes)	1.51	<.001	1.23	1.84
>1 BM at SRS	1.29	<.001	1.13	1.48

Abbreviations: BM = brain metastasis; GI = gastrointestinal; KPS = Karnofsky performance status; PTV = planning target volume; SRS = stereotactic radiosurgery; WBRT = whole brain radiation therapy.

Multivariate logistic regression model built in stepwise fashion identifying predictors of overall survival.

Patients with BMs from GI primary tumors also had worse iPFS. Only 2 GI patients (2.0%) had ICP at site of prior SRS, whereas 141 (11.0%) non-GI patients had ICP at site of prior SRS (odds ratio, 0.162; P < .002). Median iPFS was significantly lower for GI patients at 6.2mos (95% CI: 4.0-9.6) compared to non-GI patients at 12.3 months (95% CI, 10.8-13.9; P = .0002, Fig. 2). Twelve-month iPFS was 18.5% in the GI group and 42.0% in the non-GI group. In a stepwise multivariate model controlling for the same candidate predictors listed previously, GI patients had a significantly worse iPFS compared to non-GI patients (hazard ratio, 1.60; 95% CI, 1.20-2.14; P = .0013, Table 3). There were no significant differences in iPFS between colorectal primary patients or esophageal primary patients compared to all other GI primary patients (see Figs. E1, E2). There appeared to be a trend for improved iPFS in colorectal cancer BMs patients (P = .082) and worse iPFS in esophageal cancer patients with BMs compared to other GI BMs patients.

Figure 2.

Intracranial progression-free survival, GI versus non-GI. Intracranial progression-free survival of GI versus non-GI patients with brain metastases treated with stereotactic radiosurgery. Intracranial progression-free survival was poorer in the GI group.

Abbreviation: GI = gastrointestinal; SRS = stereotactic radiosurgery.

Table 3.

Multivariate regression, intracranial progression

Predictor	Hazard ratio	P value	Lower 95%	Upper 95%
GI vs non-GI primary (GI)	1.60	.001	1.20	2.14
Age at SRS ≥ 65	1.24	.002	1.10	1.49
>1 BM at SRS	1.81	<.001	1.55	2.11
Systemic therapy prior to SRS (Yes)	1.37	<.001	1.17	1.60

Abbreviations: BM = brain metastasis; GI = gastrointestinal; SRS = stereotactic radiosurgery.

Multivariate logistic regression model built in stepwise fashion identifying predictors of intracranial progression-free survival.

Discussion

We performed a retrospective multi-institutional analysis of patients receiving SRS for BM from GI and non-GI primaries and found that patients with BM with GI origins had worse OS and ICP as compared with other primary types. We also found that ICP was nearly always because of distant IC failure, not local failure. To our knowledge, this is the only study to compare outcomes after SRS specifically between patients with BM from GI and non-GI primaries.

There is a very limited body of work investigating outcomes after SRS for patients with BM because of GI primaries. These studies, in general, found median OS was between 5 and 7 months,10, 11, 12, 13, 14 which is similar to our result of 5.4 months for GI patients. Note that this was almost half of the median OS for all other primary sites of 10.6 months. Additionally, on multivariable analysis, we found that this OS difference persisted. Other significant predictors of poorer OS included age ≥ 65, KPS < 80, total metastatic burden > 5 lesions, and >1 BM at time of SRS. In an update of the original seminal analysis identifying prognostic factors for patients with BM treated with SRS, Sperduto et al⁹ found a median survival time of about 8 months in a cohort of 792 patients of GI patients with BM. They found KPS < 80, age ≥ 60, presence of extracranial metastases, and >3 BMs at time of SRS were associated with worse prognosis. Our median OS time is 3 months shorter for GI patients; however, our significant predictors of poorer survival are very similar to those put forward in Sperduto et al’s work.

It is important to note which factors are likely affecting the OS of GI patients. Prior work has shown that patients with BM of GI origin do not survive as long as patients with BM from other primaries, and that this is likely because of the burden of systemic disease.15, 16, 17, 18 In our study, GI patients did have a higher burden of extracranial metastases and were more likely to have already received systemic therapy prior to SRS. This history of systemic disease and extensive prior treatment may all contribute to worse OS after SRS. It has also been shown that a longer interval, defined as >1 year, from diagnosis of GI primary cancer to development of BM is associated with poorer survival in this population.¹⁹ We attempted to control for this interval in multivariate analysis and it was not significant in our model, possibly because we also accounted for both prior treatment history and systemic disease burden.

We found that GI patients had a median iPFS of 6.2 months, which was again nearly half of the iPFS of non-GI patients at 12.3 months. In prior studies, there were high local control rates of treated lesions, mostly between 80% and 95%, which is similar to the reported control in treated lesions in this cohort of about 80%, which was previously reported.^5,7,9,11, 12, 13^,20 Notably, only 2% of GI patients had progression of the treated lesion in this study, so the majority of ICP was distant in the brain. Many of the prior investigations did not evaluate for distant progression in the brain outside of the treated lesions; 2 studies that did so showed a similar iPFS to our study.^10,20 We found that age ≥ 65, >1 BM at time of SRS, and receiving systemic therapy prior to SRS were associated with increased risk of iPFS post-SRS. The literature on prognostic factors for ICP in this population is sparse; further work is needed to validate our findings.

It is worth noting the fact that patients with multiple BMs have worse OS in both our and Sperduto et al’s work, as well as worse iPFS. At primary diagnosis, GI patients are not routinely staged with brain MRI unless they present with neurologic symptoms. It is possible that patients presenting with subclinical BM may be missed at diagnosis. It is also possible that, given the relatively low rates of BM in GI primary cancers relative to other cancer types, providers’ thresholds for obtaining a brain MRI are somewhat higher in this population, leading to delayed scans. Both of these scenarios could lead to a higher BM burden at time of SRS, and thus poorer iPFS and OS. Further work is needed to optimize the timing of intracranial staging in this population given the poorer prognosis.

We did not find significant differences in ICP when we compared both CRC and esophageal primary patients to all other GI primaries. There was a suggestion that CRC patients tended toward better ICP while esophageal patients may have worse ICP, although neither of these met statistical significance. There is limited evidence from prior studies that luminal primaries may have better outcomes.¹²

There are several strengths of this investigation worth noting. One is that our patient cohort represents a real-world, multi-institutional sample of patients from different regions of United States. Another strength is that this is the only study specifically comparing survival and iPFS outcomes between GI and non-GI patients with BM treated with SRS to our knowledge. As such, it represents a novel contribution to the body of work investigating treatment of BM in GI patients.

There are several limitations in our work. One is that the majority of patients (54%.7) in our non-GI cohort were lung primary patients. Novel systemic therapies, such as immunotherapies and targeted therapies, have shown large benefits in certain metastatic populations, for example in lung cancer patients.²¹ This may influence our results and our work may be less generalizable outside of these populations. We also found that GI patients were more likely to have extracranial disease at time of SRS and to have been previously treated with systemic therapy. These have been shown in prior work to be associated with poorer survival and likely influence our OS results despite attempts to control for them in a multivariate model.^7,9,14, 15, 16, 17, 18 Additionally, in our iPFS analysis, death was treated as a censored event rather than a competing risk; as such, it may be contributing to the increased iPFS rate of GI patients. Finally, larger and prospective studies are needed to validate our results.

Conclusions

In conclusion, this is the first investigation specifically comparing outcomes after initial SRS course between patients with BM from GI and non-GI primaries. We found that GI patients had worse OS and iPFS after SRS. Additionally, GI patients with an older age, poorer performance status, and multiple BMs had worse survival, which supports previously published work. Our data suggest that GI patients with BM are at a higher risk of poor outcomes after SRS, and may benefit from enhanced surveillance and oncologic care. We would recommend all appropriate GI patients with BM be referred to appropriate centers for thoughtful multidisciplinary management.

Disclosures

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.