Introduction
Gastrointestinal (GI) malignancies represent a significant burden of disease in the United States with colorectal cancer representing nearly 10% of all new cancer diagnoses and ranking as the second leading cause of cancer related death [
1]. These sites rarely metastasize to the brain with reported rates at less than 9% [
2‐
6]. Nevertheless, given the prevalence of GI malignancies, brain metastases from these primaries still represent 4–8% of all diagnosed brain metastases [
3,
7‐
9]. Moreover, their incidence is increasing possibly due to longer survival seen in patients with colorectal primaries as a result of improvements in systemic therapy and earlier diagnosis [
10,
11]. Additionally, wider utilization of brain MRIs has contributed to an increased diagnosis of previously occult brain metastases [
12].
Historically, whole brain radiation therapy (WBRT) has been the standard of care for brain metastases given the lack of accurate targeting coupled with the concern for microscopic disease elsewhere in the brain [
13]. With improving targeting technology however, stereotactic radiosurgery (SRS) has emerged as an alternative treatment modality for many cases of metastatic disease to the brain. As compared to WBRT, SRS has the potential to be completed in less time, achieve increased efficacy against radioresistant tumors, increase the likelihood of re-irradiation, decrease hair loss, and mitigate neurocognitive side effects [
14‐
17].
Literature documenting clinical outcomes of patients with brain metastasis from GI primaries treated with radiation therapy is scarce and focuses mainly on outcomes following Gamma Knife radiosurgery (GKS) and not LINAC-based techniques, which are becoming a more prevalent modality of SRS. The purpose of this retrospective review is to explore the clinical outcomes, local and distant intracranial control rates, and prognostic factors associated with GI brain metastasis treated with WBRT and SRS.
Discussion
Brain metastases are generally considered a late event for patients with gastrointestinal primaries with reported rates at presentation of only 0.2% [
5]. Hence, patients rarely present with only metastases to the brain and more commonly develop liver or lung metastases prior to the development of metastatic brain foci [
5]. This is consistent with what was found in this series with 72.7% of patients presenting with liver and/or lung metastases at the time of brain metastases diagnosis. Though rare, brain metastases are an often fatal and morbid aspect of disease progression, and radiation therapy has become an accepted treatment modality for these metastatic lesions [
9,
10]. However, there is also a high competing risk of death from the significant burden of extracranial disease present in this patient population.
SRS for GI brain metastases has been documented to provide local tumor control rates of 84 to 96% [
9,
10,
20‐
22]. Our results show a crude local control rate of 64.3%. This is much lower than other reported rates, which may be explained by differences in prior radiation therapy, patient population, treatment modality, radiation doses, percentage adenocarcinoma histology, lesion sizes, and local failure definitions. Many studies of the clinical outcomes for GI brain metastases have included patients who previously received RT [
9,
10,
22]. For instance, Da Silva et al. included 55% of patients who had previously received cranial RT, and Trifiletti et al. included 36.1% of patients [
9,
10]. In contrast, Matsunaga et al. included only 5.9% of patients having previously received WBRT similar to our study which included only two lesions (7.14%) with WBRT prior to SRS [
22]. Matsunaga et al. found a crude local control rate of 91%, however they defined local failure as an increase greater than 25% in maximum diameter and had a shorter median radiographic follow-up than the current series [
22]. Additionally, our series treated larger lesions (mean size of 12.1 cc) than the previous studies [
9,
10,
22]. Larger colorectal brain lesions have been shown to have worse local control following RT [
22]. Furthermore, other studies have treated a smaller portion of adenocarcinoma histology than our current series, which treated exclusively adenocarcinoma [
9,
10,
22]. Table
4 summarizes previous studies that have reported crude local control rates after RT for GI brain metastases.
The overall median survival found in this study was 5.1 months and previous studies have shown similar findings with poor outcomes following the development of GI brain metastases. A study performed by Ogawa et al. and Weinberg et al. showed a median OS of 6.7 months and 2.3 months, respectively, after the development of brain metastases from esophageal carcinoma. York et al. reported a 2.3 month median OS from gastric cancer brain metastases, and both Schoeggl et al. and Matsunaga et al. reported median survival of 6.0 months after colorectal brain metastases. This poor overall survival is worse than other published data from non-GI primaries and is indicative of the importance in identifying prognostic factors associated with these poor clinical outcomes to appropriately allocate dose escalation and/or concurrent use of radiosensitizers [
23]. Furthermore, identifying prognostic factors for poor overall survival can help us better advise patients regarding hospice management and palliative care.
Other studies analyzing GI brain metastases have found improved local control with margin dose ≥ 20 Gy and improved survival in luminal primaries (esophagus, small bowel, stomach, colon and rectum) versus non-luminal primaries (pancreas and liver) [
10]. A study analyzing esophageal carcinoma brain metastases found that a higher KPS was associated with an increase in survival [
24]. Furthermore, a study analyzing colorectal brain metastases found unfavorable local control with larger tumor volume and lower margin dose, while poor OS was associated with lower KPS and the presence of extracranial metastases [
22]. Our study did not identify significant differences on UVA and MVA in the variables analyzed for overall survival and local control. Prior studies have demonstrated improved local control with SRS as compared to whole brain [
25,
26]; this study showed a trend towards improved local control with SRS, but did not attain statistical significance possibly due to the short radiographic follow-up periods and the small patient cohort.
Radiosensitizers are being studied for brain metastases for a variety of tumor types to improve intracranial disease control rates. Gadolinium-based nanoparticles and BRAF inhibitors concurrent with RT are being explored for melanoma brain metastases [
27,
28]. Small tyrosine kinase inhibitors, small molecule HER2 targeting agents, temozolomide, and topotecan are being investigated with breast cancer [
29,
30]. Furthermore in non-small cell lung cancer brain metastases, cisplatin and pemetrexed with concurrent cranial RT is actively being investigated [
31]. For metastatic GI malignancies, there has been promising research done into radiosensitization with irinotecan, oxaliplatin, and biological agents for non-cranial sites [
32‐
34]. Drawing upon this research, there is potential for concurrent novel therapies to improve the efficacy of cranial irradiation [
32‐
35]. While dose escalation has been shown to improve tumor control in radioresistant histologies, large tumor size would limit this strategy for many cases [
15,
36].
The role of WBRT and SRS in the management of brain metastases is evolving in the era of new systemic therapies [
37]. In our study, overall survival was not significantly different between patients who received upfront SRS versus WBRT, however we did observe higher distant brain failure rates with SRS alone. Additionally, the association with ECD suggests tumor intracranial seeding as an important factor. A study by Aoyama et al. found no improvement in survival for the use of WBRT plus SRS, but did find the rate of intracranial relapse to be increased in those who received SRS alone similar to our series [
25]. Sneed et al. also found no difference in survival when comparing SRS vs. WBRT and SRS [
23]. A study by Chang et al. demonstrated an increased decline in learning and memory function in patients treated with SRS and WBRT vs. SRS, and an improved 4-month OS in the SRS alone group; however, the local and distant control was improved in the SRS plus WBRT group [
14]. It is important to note that these studies were analyzing brain metastasis in general and not brain metastases from GI primaries as in the present study. Current research is exploring the utilization of SRS for greater than five brain metastases where historically WBRT was the preferred treatment modality. Yamamoto et al. demonstrated promising results when comparing patients with two to four brain metastases versus five to ten brain metastases treated with SRS and found non-inferiority in overall survival [
15]. Compared to WBRT, SRS is associated with a shorter treatment duration, higher BED, and reduced neurocognitive deficits [
14,
15].
Limitations of this study include its retrospective nature and the small sample size due to the relative rarity of this patient population. Additionally, given the short radiographic follow-up and the inclusion of cavitary lesions, it is often challenging to differentiate between radionecrosis and progression. Nevertheless, due to the relatively poor prognosis associated with brain metastases from GI primaries, future research should continue to assess prognostic and predictive factors, optimal dose fractionation schedules, treatment paradigms, and adverse effects. Future research should also attempt to discern the effect of radiation therapy on patient reported outcomes and quality of life endpoints. As survival improves for metastatic GI cancer with advances in systemic therapies, the incidence of brain metastasis will likely rise necessitating further improvement in treatments.
Acknowledgements
The authors would like to acknowledge the assistance of the staff at the Department of Radiation Oncology at Georgetown University for their assistance in obtaining necessary charts, and of Huiqi Zhuang for her assistance with data analysis.