Impact of Sites of Metastatic Dissemination on Survival in Advanced Gastroesophageal Adenocarcinoma

Abstract

Introduction

Metastatic gastroesophageal adenocarcinoma (GEA) is a heterogeneous disease with an overall poor prognosis. The impact of sites of metastatic dissemination on survival is not well characterized. This study aimed to evaluate whether certain sites of metastatic disease impacts survival.

Methods

A retrospective analysis of 375 patients with metastatic GEA treated at the Princess Margaret Cancer Centre from 2006 to 2016 was performed. Overall survival (OS) and progression-free survival (PFS) were estimated using the Kaplan-Meier method. Cox proportional hazards regression models were used to assess the association between sites of metastases and OS adjusting for baseline patient characteristics.

Results

Median duration of follow-up was 47.8 months. Median OS in this cohort was 11.8 months (95% CI: 10.2–12.9 months). Patients with lymph node only disease, compared to those with other sites of metastases, had the longest median OS (20.4 vs. 10.6 months; p < 0.001) and PFS (11.4 vs. 6.3 months; p < 0.001). On multivariable analysis adjusting for relevant clinical factors including age, sex, and Eastern Cooperative Oncology Group performance status, the presence of lung (HR 1.67, 95% CI: 1.23–2.26; p < 0.001) or bone metastases (HR 1.84, 95% CI: 1.31–2.59; p < 0.001) were independently associated with shorter OS. The majority of patients (68%) were treated with palliative intent first-line platinum-based chemotherapy.

Discussion/Conclusion

Patients with metastatic GEA have an overall poor prognosis. The presence of lung or bone metastases is an independent risk factor for decreased survival. Prognostic models incorporating sites of metastasis should be considered in the clinical evaluation of metastatic GEA.

Introduction

Globally, more than 1,600,000 new cases of gastroesophageal cancer were diagnosed in 2020 [1]. Gastric and esophageal cancers represent the fourth and eighth most common cancers worldwide, respectively [1, 2]. Approximately 50% of these patients present with metastatic disease [3, 4]. Despite the advances in treatment modalities and early diagnosis, prognosis of metastatic gastroesophageal cancer remains poor, with median overall survival (OS) around 1 year [5, 6, 7, 8].

Gastric and gastroesophageal junction tumors are classified in the 8th edition of the Union for International Cancer Control (UICC) staging system. In the metastatic setting, gastric, gastroesophageal, and distal esophageal adenocarcinomas are usually treated in the same manner in trials and clinical practice [5, 7]. Furthermore, recent molecular characterization using The Cancer Genome Atlas (TCGA) database, showed that esophageal and esophagogastric junction adenocarcinomas and gastric adenocarcinomas were similar while esophageal squamous cell cancers clustered more closely with head and neck tumors [9]. As such, we will refer to gastric, gastroesophageal, and distal esophageal adenocarcinomas together as gastroesophageal adenocarcinoma (GEA) based on their similar biological and treatment paradigm.

Current standard of care for first-line therapy for advanced GEA consists of combination chemotherapy with targeted treatments, such as human epidermal growth factor receptor 2 (HER2) monoclonal antibody trastuzumab and vascular endothelial growth factor receptor 2 antagonist ramucirumab. These treatment options, however, result in only modest improvements in survival [10, 11, 12, 13]. Excitingly, immunotherapy may become the new standard of care for patients with advanced GEA [14, 15]. Patients with high mutational burden, such as those with microsatellite instability, may have the greatest benefit [16]. Ongoing research is underway to determine which of the additional subsets of patients will benefit from this treatment modality.

Managing patients with metastatic disease is challenging clinically as the goal of achieving further improvement to long-term survival needs to balance the risks of toxicity. Early identification of patients with poorer prognosis can enable early treatment and consideration for upfront treatment intensification. However, there are no validated clinicopathologic or biomarkers to help risk stratify patients with metastatic disease. Existing data, largely limited to population level registries, have identified several poor prognostic factors, including advanced age, male gender, poor Eastern Cooperative Oncology Group performance status (ECOG PS), primary tumor location, and presence of peritoneal metastasis [3, 17, 18, 19, 20]. Younger age, female gender, good performance status at diagnosis, and treatment with palliative chemotherapy have been associated with improved survival [20, 21].

Sites of metastatic dissemination have become important in selecting treatment strategies in other gastrointestinal cancers. Select patients with metastatic colorectal cancer may benefit from metastasectomy in a multimodality approach with potential for long-term survival [22]. Whether this paradigm may apply to GEA patients is currently unknown and requires a better understanding of the impact of location of GEA metastases on OS. The aim of our study was to characterize the association between sites of metastatic dissemination of GEA and survival.

Methods

Study Population and Data Collection

Consecutive patients diagnosed with histologically confirmed advanced GEA, who were treated at the Princess Margaret Cancer Centre between 2006 and 2016 were retrospectively reviewed. Adenocarcinoma of the distal esophagus and adenocarcinoma of esophagogastric junction types I and II (Siewert classification) were included as esophageal cancer, and type III was defined as gastric cancer. Patients with other histologies such as squamous cell carcinoma, adenosquamous carcinoma, undifferentiated carcinoma, and small-cell carcinoma were excluded. Patient characteristics including age at diagnosis, sex, Asian ethnicity, body mass index, alcohol consumption, smoking history, and ECOG PS were recorded. Tumor characteristics, including date of diagnosis, clinical staging, tumor grade, HER2 status, as well as sites of metastases at presentation were also collected. All patients included had baseline radiological staging including computed tomography scans of the thorax, abdomen, and pelvis. Distant metastases were biopsied if clinically warranted. Clinical staging was determined using the 6th edition of the American Joint Committee on Cancer (AJCC 6) staging manual given the retrospective nature of this cohort [23]. Treatment details, including intent of treatment, type and date of palliative surgery, chemotherapy regimen, and radiation dose and fractionation were collected. This study was approved by the University Health Network Research Ethics Board (CAPCR ID 14-8075).

Statistical Analysis

Patient characteristics were summarized using descriptive statistics. OS was defined as the time from diagnosis of metastatic disease to death from any cause. Progression-free survival (PFS) was defined as the time from diagnosis of metastatic disease to either progression on imaging or death from any cause. The Kaplan-Meier method was used for time to event analyses. Patients without documented evidence of an event were censored at the date of last follow-up.

The log-rank test was used to compare outcomes between treatment groups. Cox proportional hazards regression model was used to assess the association between patient characteristics and OS. The model adjusted for the following variables determined a priori based on existing literature: age group, sex, ethnicity, body mass index, alcohol consumption, smoking history, ECOG PS, clinical stage, tumor grade, HER2 status, sites of metastasis, and treatment modality received (palliative surgery, chemotherapy, radiation). Median follow-up was calculated using the reverse Kaplan-Meier estimator. A statistical significance level of 5% (p < 0.05) was used.

Results

Patient and Treatment Characteristics

Between 2006 and 2016, 375 patients with metastatic GEA were identified. Median age was 60 years (range: 20.5–91.6 years), most were male (n = 253, 67%), and non-Asian (n = 317, 85%) (shown in Table 1). At initial presentation, 81% of the patients (n = 303) had an ECOG PS of 0 or 1. Primary sites of disease were gastroesophageal junction (n = 136, 36%) and stomach (n = 239, 64%). In our cohort, 8% of the patients (n = 29) underwent palliative surgical debulking for symptom management, 41% (n = 155) received radiation, and 68% (n = 254) received chemotherapy. The chemotherapy regimen included platinum doublet (12%), platinum triplet (74%), and other (14%). Sites of metastatic disease (shown in Fig. 1) included distant lymph node(s) only (n = 45, 12%), lung (n = 71, 19%), liver (n = 169, 45%), peritoneal (n = 154, 41%), bone (n = 53, 14%), brain (n = 4, 1%), ovary (n = 29, 8%), adrenal (n = 21, 6%), or other (n = 21, 6%). Among this cohort, 38% had more than one site of metastases. Patients with bone and lung metastases had more sites of metastatic disease, with 26% and 36% presenting with more than 3 sites, respectively (data not shown). Patients with lymph node only metastasis, and those with lung or bone metastasis had similar ECOG PS at the start of treatment (shown in online suppl. Fig. 1a–c; see www.karger.com/doi/10.1159/000525616 for all online suppl. material).

Table 1.

Baseline characteristics

Age	n = 375
Median (min, max)	60 (20.5, 91.6)
Sex, n (%)
Male	253 (67)
Female	122 (33)
Ethnicity, n (%)
Asian	58 (15)
Non-Asian	317 (85)
BMI, n (%)
0 Underweight	18 (7)
1 Normal	141 (52)
2 Overweight	72 (27)
3 Obese	39 (14)
Missing	105
Alcohol, n (%)
Frequent/past	74 (20)
Rarely/never	255 (68)
Unknown	46 (12)
Smoking, n (%)
Current smoker	46 (12)
Ex-smoker	130 (35)
Never smoker	165 (44)
Unknown	34 (9)
ECOG, n (%)
0	78 (21)
1	225 (60)
2	47 (13)
3	24 (6)
4	1 (0)
Siewart, n (%)
AEG1	53 (14)
AEG2	83 (22)
AEG3	27 (7)
Gastric	212 (57)
Site of metastases,a n (%)
Lung
No	304 (81)
Yes	71 (19)
Liver
No	206 (55)
Yes	169 (45)
Peritoneal
No	221 (59)
Yes	154 (41)
Bone
No	322 (86)
Yes	53 (14)
Brain
No	371 (99)
Yes	4 (1)
Ovary
No	346 (92)
Yes	29 (8)
Adrenal
No	354 (94)
Yes	21 (6)
Other
No	353 (94)
Yes	21 (6)
Lymph node only
No	330 (88)
Yes	45 (12)
>1 Metastases
No	231 (62)
Yes	144 (38)
Grade, n (%)
GX: undetermined	101 (27)
G1: well-differentiated	7 (2)
G2: moderately differentiated	77 (21)
G3: poorly differentiated	190 (51)
HER2, n (%)
Positive	93 (25)
Negative	87 (23)
Unknown	195 (52)
Surgery, n (%)
No	343 (92)
Yes	29 (8)
Missing	3
Chemotherapy, n (%)
No	120 (32)
Yes	254 (68)
Missing	1
Radiation, n (%)
No	220 (59)
Yes	155 (41)

BMI, body mass index; ECOG PS, Eastern Cooperative Oncology Group performance score; LN, lymph node.

^aPatients with site of metastasis, other than lymph node only disease, may have multiple sites of involvement.

Fig. 1.

Breakdown of anatomic sites of metastatic disease.

Survival Outcomes

Median follow-up was 9.6 months (interquartile range 5.1–16.6 months). The median OS for our cohort is 11.8 months (shown in Fig. 2; 95% CI: 10.2–12.9 months), and the 1-year OS rate was 49% (95% CI: 44–55%). At the time of analysis, 284 (76%) of 375 patients had died. Univariable analysis showed ECOG PS (p < 0.001) and tumor grade (p = 0.033) were the only patient factors associated with OS (shown in Table 2). With respect to sites of metastatic dissemination, median OS was significantly longer in patients who had lymph node metastases only (shown in Fig. 3a; 20.4 months vs. 10.6 months; hazard ratio: 0.45; 95% CI: 0.31–0.67; p < 0.001). Sites of metastases associated with a worse OS include lung (shown in Fig. 3b; 7.7 months vs. 12.8 months; hazard ratio: 1.78; 95% CI: 1.34–2.36; p < 0.001), peritoneal (10.6 months vs. 12.4 months; hazard ratio 1.27; 95% CI: 1.1–1.61; p = 0.049), and bone (shown in Fig. 3c; 7.3 months vs. 12.4 months; hazard ratio 1.95; 95% CI: 1.41–2.7; p < 0.001).

Fig. 2.

Kaplan-Meier OS curve for retrospective study cohort.

Table 2.

OS by patient characteristics

Variable, n (events)	HR (95% CI)	p valuea
Age 375 (284)
≤60	Reference	0.31
>60	1.13 (0.89, 1.42)
Sex 375 (284)
Male	Reference	0.43
Female	1.11 (0.86, 1.42)
Race 375 (284)
Asian	Reference	0.55
Non-Asian	0.9 (0.64, 1.27)
BMI 270 (212)
0 Underweight	Reference
1 Normal	0.89 (0.52, 1.54)	0.37
2 Overweight	0.88 (0.5, 1.54)
3 Obese	0.64 (0.34, 1.2)
Alcohol 375 (284)
Frequent/past	Reference
Rarely/never	0.86 (0.65, 1.15)	0.6
Unknown	0.9 (0.58, 1.4)
Smoking 375 (284)
Current smoker	Reference
Ex-smoker	0.89 (0.62, 1.29)	0.94
Never smoker	0.93 (0.65, 1.34)
Unknown	0.89 (0.52, 1.51)
ECOG 375 (284)
0	Reference	<0.001
1 +	1.68 (1.26, 2.24)
Clinical T 375 (284)
T0–2	Reference
T3–4	1.32 (0.57, 3.06)	0.59
TX	1.15 (0.51, 2.6)
Clinical N 375 (284)
N0	Reference
N+	0.99 (0.7, 1.41)	0.22
NX	0.79 (0.54, 1.16)
Lymph node only 375 (284)
No	Reference	<0.001
Yes	0.45 (0.31, 0.67)
Lung 375 (284)
No	Reference	<0.001
Yes	1.78 (1.34, 2.36)
Liver 375 (284)
No	Reference	0.11
Yes	1.21 (0.96, 1.53)
Peritoneal 375 (284)
No	Reference	0.049
Yes	1.27 (1, 1.61)
Bone 375 (284)
No	Reference	<0.001
Yes	1.95 (1.41, 2.7)
Brain 375 (284)
No	Reference	0.75
Yes	1.2 (0.39, 3.76)
Ovary 375 (284)
No	Reference	0.95
Yes	1.01 (0.64, 1.62)
Adrenal 375 (284)
No	Reference	0.24
Yes	1.37 (0.81, 2.3)
Other 374 (283)
No	Reference	0.16
Yes	1.42 (0.87, 2.32)
Grade 375 (284)
G1–2	Reference	0.033
G3	1.48 (1.1, 1.99)
GX	1.31 (0.94, 1.83)
HER2 375 (284)
Positive	Reference
Negative	1.15 (0.83, 1.6)	0.15
Unknown	1.32 (1, 1.76)
Surgery 372 (283)
No	Reference	<0.001
Yes	0.25 (0.14, 0.42)
Chemotherapy 374 (283)
No	Reference	<0.001
Yes	0.54 (0.42, 0.69)
Radiation 375 (284)
No	Reference	<0.001
Yes	1.66 (1.31, 2.1)
>1 Metastases 375 (284)
No	Reference	<0.001
Yes	1.84 (1.44, 2.35)

OS, overall survival; CI, confidence interval; HR, hazard ratio; BMI, body mass index; ECOG PS, Eastern Cooperative Oncology Group performance score; LN, lymph node. p values were calculated based on continuous values for age and excluded patients the missing values for each characteristic.

^aSignificant value(s) appear in boldface type.

Fig. 3.

Kaplan-Meier OS curves among patients with metastases to lymph node only (a), lung (b), and bone (c).

Median PFS for the entire cohort was 6.6 months (95% CI: 6.1–7.2 months). Median PFS was significantly longer in patients with lymph node metastases only (shown in online suppl. Fig. 2a; hazard ratio: 0.46; 95% CI: 0.36–0.61; p < 0.001). Sites of metastases associated with a worse PFS include lung (shown in online suppl. Fig. 2b; 5.8 months vs. 6.8 months; hazard ratio: 1.38; 95% CI: 1.02–1.87; p = 0.038), peritoneal (6.1 months vs. 7.1 months; hazard ratio 1.29; 95% CI: 1.03–1.62; p = 0.031) and bone (shown in online suppl. Fig. 2c; 5.2 months vs. 6.8 months; hazard ratio 1.86; 95% CI: 1.27–2.73; p = 0.003).

In the multivariable model (shown in Table 3), ECOG PS of 0 remained the only clinical characteristic associated with a better OS (p < 0.001). Sites of metastatic dissemination to lung and bone were independent poor prognostic factors (p < 0.001 and p < 0.001, respectively). The presence of lymph node only metastasis was not associated with OS in multivariable analysis (p = 0.1). To minimize the effects of selection bias, we performed a sub-analysis examining only patients who received chemotherapy. The median OS for patients receiving chemotherapy was 13.1 months compared to 6.0 months without and the 1-year OS rate was 59% (95% CI: 53–65%). Among this treated cohort, the negative prognostic effect of lung and bone metastasis remained (shown in online suppl. Fig. 3a–c). Majority of patients received platinum-based doublet or triplet chemotherapy. Chemotherapy regimen used in the metastatic setting was similar regardless of sites of metastatic disease (shown in online suppl. Fig. 4a–c).

Table 3.

Multivariable model of OS

Covariate	HR (95% CI)	p valuea
Age group
≤60	Reference	0.34
>60	1.13 (0.88, 1.44)
Sex
Male	Reference	0.51
Female	1.1 (0.83, 1.46)
ECOG gp
0	Reference	<0.001
1 +	1.65 (1.24, 2.2)
Lymph node only
No	Reference	0.1
Yes	0.67 (0.41, 1.08)
Lung
No	Reference	<0.001
Yes	1.67 (1.23, 2.26)
Liver
No	Reference	0.55
Yes	1.1 (0.81, 1.49)
Peritoneal
No	Reference	0.13
Yes	1.28 (0.93, 1.75)
Bone
No	Reference	<0.001
Yes	1.84 (1.31, 2.59)
Brain
No	Reference	0.78
Yes	1.23 (0.29, 5.14)
Ovary
No	Reference	0.51
Yes	0.84 (0.49, 1.42)
Adrenal
No	Reference	0.32
Yes	1.31 (0.77, 2.23)
Other
No	Reference	0.32
Yes	1.29 (0.78, 2.16)

HR, hazard ratio; CI, confidence interval; BMI, body mass index; ECOG PS, Eastern Cooperative Oncology Group performance score; LN, lymph node. p values were calculated based on continuous values for age and excluded patients the missing values for each characteristic.

^aSignificant values appear in boldface type.

Discussion/Conclusion

Despite current standard of care, patients with metastatic GEA have a dismal OS [24, 25, 26, 27, 28, 29]. Although new chemotherapeutic regimens in heavily pretreated patients are currently being investigated, analysis to date indicates that the impact on survival is modest at best [30, 31]. Second and third line approaches using vascular endothelial growth factor receptor 2 monoclonal antibody ramucirumab [11, 32], or more recently with checkpoint inhibitors [33, 34, 35, 36], only marginally improved OS. As such, there exists a dire need to identify patients who would potentially benefit from treatment intensification or earlier palliation. Our study supports the notion that metastatic GEA is a heterogeneous disease with significant differences in outcomes associated with sites of metastases [37]. Lung and bone are common sites of metastatic dissemination. In a large Swedish registry of gastric cancer, lung, and bone metastases are involved in approximately 15% and 12% of patients with metastatic disease, respectively, ranking just behind liver and peritoneum [18]. We show that lung and bone metastasis are independent negative prognostic factors for OS. It is important to note that this observation was obtained from retrospective analysis and will need to be validated in prospective studies.

For patients with bone metastases, several features including pain, immobilization, and skeletal related complications may contribute to decreased ECOG PS and OS [38, 39]. Among these study patients, higher levels of serum lactate dehydrogenase and carcinoembryonic antigen are poor prognostic markers [38, 39]. Patients with pleural and lymphangitic metastasis are known to have worse outcome compared to hematogenous sites of metastases [40]. In our study, patients with distant lymph node metastasis had an improved outcome in univariable analysis; however, this did not hold true in multivariable analysis. The median survival for patients with lymph node metastasis only was 20.4 months versus 10.6 months for other metastatic sites. Whether this is due to earlier identification and thus lower disease burden, or if this is in fact a reflection of underlying biology or response to treatment, remains to be explored.

Previous population studies have shown similar OS among patients with different sites of metastatic dissemination in advanced gastric cancer, averaging around 4 months [18]. These studies use cancer registries which often omit important site-specific details apart from M stage; furthermore, they are hampered by lack of clinical details including treatment [8, 18, 20, 21, 41]. Other research has suggested that patients with peritoneal disease may have better outcome as compared to those with distant metastases [41]. In select patients, palliative gastrectomy has shown to have modest survival benefit, although randomized data are lacking [42, 43]. It is intriguing to hypothesize that site of metastatic dissemination may reflect differing underlying tumor biology, tumor microenvironment or response to treatment. Further research is needed to contextualize this clinical information by using molecular subgroups as well as biomarkers to develop better patient prognostication tools and enhance our understanding of the tumor pathobiology.

Existing literature largely focuses on the biology of the primary tumor [6, 44]. Metastatic invasion is a complex multi-step process involving tumor migration, vascular invasion, and extravasation, and recruitment or adaptation of the metastatic microenvironment [45, 46]. Recently, integrated genomic analyses of peritoneal carcinomatosis have shown two distinct molecular subtypes with different genetic signatures and varying responses to chemotherapy [47]. One of the barriers of further understanding the underpinning molecular pathways of metastases is limited availability of tissue biopsies of metastases, as they are infrequently done due to potential complications and delay in treatment. Furthermore, a subtype of tumors with elevated expression of TGF-B1 (transforming growth factor-β), immune checkpoint TIM-3 (T-cell immunoglobulin and mucin domain-containing protein 3), its ligand galectin-9 and VISTA (V-domain immunoglobulin suppressor of T-cell activation) may be uniquely positioned for immuno-oncology therapies. This detailed characterization in other metastatic compartments is currently unavailable. Knowing the genetics and biology of the metastatic disease will facilitate targeted therapeutic approaches of metastatic GEA in future clinical trials. For example, HER2-targeted therapies are currently the standard of care for patients with metastatic GEA overexpressing this receptor [10]. Future work will need to characterize if these driver events are also playing a role in the metastatic compartment.

There are several limitations to our study. It is a retrospective review of a single high-volume center experience. As such, potential confounders exist that could potentially explain the worse OS in patients with lung and bone metastases. The results of this study will require confirmation from larger prospective trials, with multicenter collaboration, to increase the sample size as well as to reduce bias. Furthermore, given the pace of advance in the field, the standard of care has evolved over the past decade. Even though HER2 is a known prognostic marker, 52% of our cohort had an unknown status as this only became standard of care after 2010 following the publication of ToGA [10]. The impact of newer lines of treatment on patient survival taking into consideration sites of metastatic disease will need to be further explored. Despite these shortcomings, this study represents a large, curated dataset of metastatic GEA with detailed clinical information including sites of metastases and treatments to date. The strength of our analysis is that it includes existing clinical, treatment, and pathologic data with detailed characterization of sites of metastases. Our data suggest the need to consider sites of metastatic disease when considering prognosis and treatment strategies.

In conclusion, among patients with metastatic GEA, those with metastases to lung and bone may be independent negative prognostic markers, although this will need prospective validation. This observation provides new prognostic information for patients and clinicians and should be considered as part of clinical risk stratification. The role of strategic selection of these patients for intensified, targeted treatments requires further study.

Statement of Ethics

This study was approved by the University Health Network Research Ethics Board, approval number CAPCR ID 14-8075. Waiver of informed consent was obtained from the Research Ethics Board due to the minimal risk to study subjects as well as the retrospective nature of this study.

Conflict of Interest Statement

Elena Elimova discloses advisory role and research funding received from Bristol-Myers Squibb. All other authors disclose no financial interests or conflicts of interest.

Funding Sources

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Author Contributions

Xin Wang: conceptualization, data curation, methodology, formal analysis, writing − original draft. Osvaldo Espin-Garcia: formal analysis, methodology, visualization. Di Maria Jiang: data curation, writing − review and editing. Michael James Allen, Lucy Xiaolu Ma, Eric Xueyu Chen, Gail Elizabeth Darling, Johnathan Chi-Wai Yeung, Rebecca Wong, Patrick Veit-Haibach, Sangeetha Kalimuthu, Raymond Woo-Jun Jang: data curation, writing − review and editing. Yvonne Bach, Chihiro Suzuki, Marta Honorio: data curation. Elena Elimova: supervision, data curation, funding acquisition, conceptualization, writing − review and editing.

Data Availability Statement

All data generated or analyzed during this study are included in this article and its online supplementary material. Further inquiries can be directed to the corresponding author.

Supplementary Material

Supplementary data

Funding Statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.