Incidence of Brain Metastasis as First Event in Patients with Esophageal Carcinoma: a Report from Three Prospective Alliance Clinical Trials

Historically, reported incidence of brain metastasis secondary to esophageal carcinoma is low. We sought to determine the incidence of brain metastasis in a contemporary cohort of patients with carcinoma of the esophagus. Data from patients with localized esophageal carcinoma prospectively enrolled on three curative intent Alliance treatment trials (N0044, N0342, N044E) were reviewed including time to diagnosis of first progression event (brain versus other site) and overall survival. Eighty-five patients comprised the cohort of which 85% were male and 86% had adenocarcinoma primary tumor histology. Thirty-nine of the 85 patients had documented progression to any site, and of those, brain metastasis occurred as the first event in 15% (6 of 39). Adenocarcinoma was the primary histology in all 6 patients and tumor grade was high (3 or 4) in 5 of the 6 patients (one not documented). Median time to brain metastasis (9.6 months) versus non-brain metastasis (12.4 months) and median survival after first progression (5.4 months versus 8.1 months, respectively) were not statistically different. In this prospective cohort of patients with esophageal carcinoma, those with high-grade adenocarcinoma appear to have a higher incidence of brain metastasis than historically reported. The pattern of brain metastases corroborates recent findings in terms of incidence, predominance of adenocarcinoma primary tumor histology, timing after diagnosis, and overall survival. Further study to confirm these findings, as well as the value of baseline, restaging and follow-up cranial imaging for brain metastasis is recommended. NCT00022139 (NCCTG N0044), NCT00100945 (NCCTG N0342), and NCT00100945 (NCCTG N044E).


Introduction
The incidence of esophageal cancer continues to increase in the USA, with an expected 20,640 cases and 16,410 deaths in 2022 [1].The increase is largely due to the rising incidence of adenocarcinoma, surpassing squamous cell between 1990 and the early 2000s as the most common histology in the USA, the causes for which are unclear [2][3][4].
Historically, estimates of brain metastasis in patients with esophageal carcinoma have been low, ranging from 0.5 to 4.8% [5][6][7][8][9][10][11][12].However, many of these studies included patients from the 1980s and 1990s when squamous cell was the predominant histology.Due to this shift in histology, altered patterns of metastasis may be expected.
Information on the incidence rate of brain metastasis by histologic type is limited.More recent studies indicate a trend toward association between adenocarcinoma and brain metastasis.Gabrielsen et al. reported a 3.6% rate of brain metastasis in those with adenocarcinoma [12].Welch et al. reported diagnosis of brain metastasis in 22 of 583 patients (3.8%) with stage I-IV esophageal cancer, and 21 of those patients had adenocarcinoma [13].
In a retrospective analysis of a contemporary cohort of patients with esophageal cancer treated in a community setting, Smith and Miller reported a 13% incidence rate of brain metastasis, higher than historically reported [14].Notably, adenocarcinoma was the primary tumor histology in all cases of brain metastasis.Subsequently, a prospective trial of neoadjuvant chemotherapy followed by surgical resection in a contemporary cohort of patients with esophageal carcinoma reported a similar incidence of brain metastasis of 15%, also noting a predominance of adenocarcinoma histology [15].
We sought to further validate and understand the development of brain metastasis secondary to esophageal carcinoma in patients enrolled on contemporary, prospective clinical treatment trials.

Materials and Methods
After obtaining Mayo Clinic Institutional Review Board exemption and Alliance for Clinical Trials in Oncology approval, the datasets of all patients with a primary diagnosis of esophageal carcinoma treated with curative intent on three prospective phase II Alliance for Clinical Trials in Oncology (Alliance) trials (North Central Cancer Treatment Group (NCCTG) N0044, N0342, N044E [16,17] were reviewed (details in Supplement).NCCTG is now part of the Alliance.Each participant had signed an IRB-approved, protocol-specific informed consent document in accordance with federal and institutional guidelines.Patient demographics, clinical characteristics, time to diagnosis of brain metastasis as first event, time to other first progression event, and subsequent overall survival data were collected.Only the first site of disease progression was documented per trial design as location of later sites was not the primary focus of these studies.
These studies actively enrolled patients between 2002 and 2008.Eligibility criteria for enrollment onto these trials included a diagnosis of adenocarcinoma or squamous cell carcinoma of the esophagus, considered to be surgically resectable, including nodal disease if present.N044E excluded those with stage T1 or T2 in the absence of nodal involvement, allowed celiac node involvement, as well as supraclavicular node involvement in the setting of upper esophageal malignancy.N0044 allowed patients with T4 or celiac node involvement.None of the patients had a diagnosis or suspicion of brain metastasis at the time of trial enrollment, but central nervous system imaging was not mandated.
Associations of categorical variables were done via Fisher's exact test due to small numbers.Associations of continuous variables with categorical data were done via the Kruskal-Wallis test or Wilcoxon rank-sum test.Time to first progression and overall survival post-first progression were analyzed using Kaplan-Meier analysis and the stratified log-rank test, where we stratified by the study trial.Time to first progression was defined as the time from study entry to the first disease progression, where this time was compared between patients who progressed to the brain versus those that progressed to non-brain sites.Only those that progressed were included.Survival post-first progression was defined as the time from first progression to death from any cause, where the date of last follow-up was used for those still alive (i.e., censors).Data collection and statistical analyses were conducted by the Alliance Statistics and Data Management Center.All analyses were based on the study database (frozen on December 6, 2021).SAS version 9.4 (SAS Institute, Cary, NC) was used for statistical analysis and 2-sided p-values ≤ 0.05 were considered statistically significant.

Cohort
A total of 88 patients were enrolled on the three Alliance trials, of which three were excluded from subsequent analysis due to trial ineligibility or no treatment received prior to discontinuing trial participation, resulting in 85 evaluable patients for this analysis.Eighty-four of the 85 patients received neoadjuvant chemoradiotherapy and 66 of the 85 patients underwent subsequent surgical resection (1 patient underwent definitive surgery alone).
At time of last follow-up, 56 patients had died.In the 29 patients still alive, the median (range) follow-up time was 49.6 months (32.4 to 60 months).Thirty-nine of the 85 patients had documented disease progression, and 22 were alive without evidence of progression.
The baseline characteristics of the study cohort are shown in Table 1.The mean age was 62.3 years and 72 (85%) were male.Adenocarcinoma was the primary histology in 73 patients (86%), squamous cell carcinoma in eight patients (9%), adenosquamous carcinoma in two patients (2%), and histologic information was not available for two patients (2%).Histologic grade was high (3 or 4) in 56 (66%), low (1 or 2) in 20 (24%), and unreported in nine patients (11%).Tumor location, as reported, was significantly different between trials (p < 0.0001), while all other variables were similar between trials (see Table 1).

Site of First Progression
Brain was the first site of progression in six (7%) of the 85 evaluable patients.Time to first progression was not significantly different between those that progressed to the brain versus other sites (p = 0.30; Fig. 1).Specifically, the median time to progression for those that progressed to the brain was 9.6 months (95% CI = 4.8-not reached) as compared to a median of 12.4 months (HR 95% CI = 8.4-20.2) in those patients with first progression to non-brain sites.Characteristics for patients with disease progression (N = 39) are shown in Table 2, and in 15% of this group, the brain was the first site of disease progression.

Histologic Type
Seventy-three of the 85 evaluable patients (86%) had primary adenocarcinoma of the esophagus.All six patients with brain metastasis as site of first progression had adenocarcinoma primary tumor histology (8% of the 73 patients with adenocarcinoma) as shown in Table 2. Of the 39 patients with documented disease progression, 35 (90%) had primary adenocarcinoma of the esophagus, three had primary squamous cell carcinoma, and histology was unknown in one patient.Therefore, the brain was the first site of documented progression in six of 35 (17%) patients with adenocarcinoma who experienced disease progression.

Histologic Grade
Of the six patients with brain metastasis as site of first progression, five (83%) had a primary tumor grade that was high (grade 3 or 4), and one had no grade reported (Table 2).First progression to non-brain sites occurred in 33 patients, 23 (70%) with high-grade tumors, eight (24%) with low-grade tumors, and grade was not reported in two (6%) patients.

Overall Survival
Median overall survival (OS) after first progression to the brain was 5.4 months (HR 95% CI = 1.1-not reached), while median OS after first progression to a non-brain site was 8.1 months (HR 95% CI = 2.2-15.3),but the difference was not statistically significant (p = 0.23) (Fig. 2).

Associations with Site of Progression
We tested the association of baseline characteristics with site of first progression (brain vs. other site), including age, gender, Eastern Cooperative Oncology Group (ECOG) Performance Status, tumor location, histologic grade, body mass index (BMI), and histologic type.None of these variables were found to be statistically significant from this analysis (Table 2).

Discussion
The current study was undertaken to better understand the development of brain metastasis in a prospectively followed cohort of patients with non-metastatic esophageal cancer treated on pre-operative chemoradiotherapy trials.Our results revealed brain to be the first site of metastasis in 7% (6/85) of all patients, 8% (6/73) in those with adenocarcinoma as the primary tumor histology, and 15% (6/39) of those who experienced disease progression after primary therapy.The median time to brain progression as a first event was 9.6 months.Adenocarcinoma was the primary tumor histology in all six patients with brain as the first site of progression.However, as only the first site of disease progression was documented, the overall incidence of brain metastasis may be higher than identified in this study.
In a prospective phase II trial of neoadjuvant chemotherapy followed by surgical resection in a contemporary group of 40 patients with esophageal cancer (85% with adenocarcinoma), Cleary et al. reported that six patients (15%) developed brain metastasis at a median of 13.9 months after surgery [15].In five of the six patients, adenocarcinoma was the primary histology and brain was the first site of metastasis in four of 40 patients (10%).
Our results and the study reported by Cleary et al. [15], two prospectively followed esophageal cancer patient cohorts, demonstrated similar rate of brain as first site of progression, median time to development of brain metastasis, and predominance of adenocarcinoma as the primary tumor histology.Additionally, the 15% overall incidence of brain metastasis observed in the prospective study by Cleary et al. [15] is consistent with the 13% incidence reported in the earlier retrospective study by Smith and Miller [14].
A correlation between adjuvant therapy and incidence of brain metastasis was reported in a large series of patients with esophageal carcinoma who underwent esophagectomy with or without adjuvant therapy from 1985 to 2002 [9].The overall incidence of brain metastasis was 3.8%; however, the rate at 5 years posttreatment was 2.5% with surgery alone and as high as 18.4% in patients who underwent surgery with both preoperative and postoperative therapy.All cases of brain metastasis occurred in patients with adenocarcinoma, consistent with our current and prior study, but treatment type by histology was not reported.In a large series of patients who underwent R0 esophagectomy between 1995 and 2016, 85% with primary adenocarcinoma and 67% who underwent neoadjuvant therapy, Nobel et al. reported a 2.3% incidence of isolated brain metastasis with median time to diagnosis of 12 months [18].Ninety percent of these patients had adenocarcinoma primary tumors, and all received neoadjuvant therapy.Differences in populations between the Nobel study and ours include the proportion who underwent neoadjuvant therapy, 67% (1181/1760) vs. 98% (84 of 85), respectively, and the proportion with locally advanced disease, 71% vs. 100%, respectively.These differences could be factors in the discrepancy of the reported incidence of brain metastasis between the two studies.
In a Mayo Clinic surgical series of 708 patients with primary esophageal/gastroesophageal junction adenocarcinoma who underwent definitive resection, none of whom received neoadjuvant therapy, the 5-year central nervous system (CNS) relapse as first site of recurrence was 5.8% in HER-2 positive primary tumors versus 1.2% in HER-2 negative primary tumors (anytime CNS relapse 8.3% versus 2.4%, respectively) [19].The similar incidence rates of the brain as first site of metastasis in HER-2 positive esophageal adenocarcinoma patients without neoadjuvant therapy (5.8%) and our series of predominately adenocarcinoma patients treated with neoadjuvant therapy (7%) suggest that histology, particularly HER-2 overexpressing adenocarcinoma, rather than neoadjuvant therapy, is the driver of CNS metastatic potential.
Tumor location was found to correlate with incidence of brain metastasis in a study by Harada et al. [20].Tumor location in our patient cohort was non-descript for "esophagus" and "esophagus or GE junction" which is a limitation of this study.Tumor location in 4 of the 6 patients in our cohort that developed brain metastasis was described as GE junction, 1 as esophagus or GE junction and 1 as esophagus.Therefore, in at least 4 of the 6 patients with brain as first site of progression in our study, the location of GE junction would correlate to location of the increased incidence of brain metastasis described by Harada et al. (Siewert type I, 5.9%), approaching the 7% incidence in our study.Coexisting variable with tumor location is the propensity for adenocarcinomas to arise in the distal esophagus and Harada et al. concluded that "the highest risk of developing brain metastasis was with adenocarcinoma histology (proximal esophageal adenocarcinoma 4.8%) than with squamous cell carcinoma histology" (1.2%) [20].Whether treatment type influences the natural history of metastatic spread or is a surrogate for other characteristics such as primary tumor histology, tumor location, and/or molecular characteristics such as HER-2 overexpression, deserves further investigation.
The relatively short interval between diagnosis of esophageal carcinoma and development of brain metastasis as site of first progression in our study (median of 9.6 months), is similar to that reported by others as ranging from 5.6 to 15.9 months [5, 7, 13-15, 18, 19, 21].Of the 29 patients with brain metastasis in the study by Rice et al.,20 (69%) were diagnosed within 1 year following esophagectomy [9].Therefore, progression to the brain appears to be a relatively early occurring event following a diagnosis of esophageal carcinoma, particularly adenocarcinoma.
Should pre-operative brain imaging be routinely obtained in patients with adenocarcinoma of the esophagus?In a study by Gabrielsen et al., in an era when squamous cell carcinoma was the predominantly histology, 12 of the 16 patients diagnosed with brain metastasis displayed neurologic symptoms, with the conclusion that screening for brain metastasis in patients with esophageal carcinoma was not cost-effective [12].Wadhwa et al. studied a contemporary group with esophageal adenocarcinoma receiving trimodality therapy and also concluded that the low incidence of brain metastasis (3.9%) did not justify screening and/or surveillance studies [22].
Given the increasing predominance of adenocarcinoma histology and its potential association with brain metastasis, we advocate close attention to neurologic signs and symptoms in patients with esophageal cancer and recommend brain imaging as a component of initial and follow-up evaluations when clinically indicated.Further studies designed to assess variables associated with brain versus non-brain progression are needed before brain MRI can be recommended for screening in asymptomatic patients.Furthermore, given the potential association between histology and brain as first site of progression, research involving potential molecular drivers of site-specific metastasis or biomarkers, such as circulating tumor DNA that may signal metastatic potential, is also recommended.
This study analyzed prospectively collected data from a contemporary cohort of patients enrolled on pre-operative chemoradiotherapy trials; this prospective design is a major strength.One limitation of the data collection was that only the first site of progression was documented according to these protocols, which may have led to an underestimate of the overall incidence of brain metastasis and resulted in lack of adequate comparison to past literature.The small sample size also limited the detection of potentially significant associations with the risk of brain metastasis including histology, grade, treatment modality, or other characteristics.Other limitations include an absence of information on how the brain metastases were treated, an absence of data on how many brain metastases were detected, and an absence of information about whether brain metastases were synchronous or not with other metastatic sites.Nonetheless, research should focus on identifying factors associated with risk to develop brain metastasis, the appropriate surveillance strategies for these patients, and how to best to treat those who develop brain metastasis.Further understanding of the molecular drivers of metastatic progression to the brain may lead to therapeutic strategies to reduce the incidence of this event.The ongoing evolution of current standards of care, most notably the integration of nivolumab as maintenance therapy in the treatment of subgroups of esophageal cancer patients with locally advanced disease, may impact the incidence of progression to the brain; however, the descriptive nature of our findings is important, as our results call attention to the risk of brain metastasis in esophageal cancer patients and illustrate the natural history of this cancer.
In conclusion, this study of patients with nonmetastatic esophageal cancer prospectively enrolled on contemporary treatment trials revealed a higher incidence of brain metastasis than historically reported.Our study adds evidence to trends seen in other studies of brain metastasis occurring relatively early in the course of disease and predominantly in those with esophageal adenocarcinoma [12-15, 19, 20].Therefore, brain metastasis secondary to esophageal cancer may be expected to occur more often and within a relatively early timeframe following diagnosis, than in the era when squamous cell carcinoma was the predominant histology.

Fig. 1
Fig. 1 Time to first progression.No significant difference in time for first progression based on site of first progression (brain versus not) in this cohort (stratified log-rank p = 0.30).Hazard ratio analysis stratified by trial to take into account possible differences between trials

Fig. 2
Fig. 2 Overall survival after first progression.No significant difference in overall survival after first progression to brain or to other sites (stratified log-rank p = 0.23).Hazard ratio analysis stratified by trial to take into account possible differences between trials

Table 1
Patient characteristicsThe table excluded 3 patients who were either ineligible or cancelled prior to receiving treatment * Unable to calculate p-value due to no patients having metastases at baseline ** Post-op adjuvant trial a Kruskal Wallis b Fisher exact c Eastern Cooperative Oncology Group Performance Status d Gastroesophageal e Body mass index

Table 2
Associations of variables with site of progressionStatistical testing for predictors of subsequent brain metastasis as first event did not reach levels of significance based on the small number of events and subset classes a Wilcoxon rank-sum b Fisher Exact c Eastern Cooperative Oncology Group Performance Status d Body mass index e Gastroesophageal