Divergent Metastatic Patterns Between Esophageal Squamous-cell Carcinoma and Esophageal Adenocarcinoma: a Propensity-matched Analysis

Background: To explore the different metastatic patterns between esophageal squamous-cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC). Methods: For this propensity-matched analysis, we used data from the latest iteration of the Surveillance, Epidemiology, and End Results (SEER) database and included patients diagnosed with esophageal cancer from 2010 to 2017. Results: A total of 20,189 patients were identied, including 6,610 ESCC and 13,579 EAC. After propensity score matching, 4597 pairs were selected. Compared with ESCC, EAC had a higher rate of liver metastasis (P < 0.001) and brain metastasis (P < 0.001), and a lower rate of lung metastasis (P < 0.001), with no signi ﬁ cant di ﬀ erence in bone metastasis (P = 0.255). The liver preferentially co-metastasized with lung in both cohorts. Brain metastasis was commonly observed in combination with other organ metastases in EAC. Conclusions: There are major differences in metastatic patterns between ESCC and EAC. The patterns identied may reect the underlying biology of metastatic esophageal cancer and have potential to inuence future monitoring strategies depending on clinical settings.


Background
Esophageal cancer is the seven most common cancer worldwide and ranks sixth in terms of mortality among all types of cancer [1]. The detection and clinical management of esophageal cancer metastasis poses a critical global health challenge, since approximately 40% of patients with esophageal cancer are diagnosed with metastatic disease at the time of presentation [2]. As a disease entity, esophageal cancer principally comprises two major subtypes: esophageal squamous-cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC). In spite of sharing an anatomical site, there are many differences between ESCC and EAC in terms of the etiology and clinical features [2,3]. Meanwhile, according to the study conducted by the Cancer Genome Atlas Research Network, ESCC and EAC might belong to distinct molecular entities in view of their divergent genetic characterizations [4].
The process of tumor metastasis is fraught with complexity [5][6][7] and the metastatic patterns might be affected by cancer subtypes [8]. Two previous Surveillance, Epidemiology, and End Results (SEER) based studies had suggested some differences in metastatic patterns between ESCC and EAC [9,10], but their ndings were inconsistent. Moreover, patterns of combined metastases had not been fully illustrated in their study and propensity score matching was not performed to minimize potential selection biases. In addition, as the incidences of ESCC and EAC have been changed over the past several decades with an increasing of EAC and a decreasing of ESCC [2], this transformation should also be taken into consideration when assessing the characteristics of distant metastasis in esophageal cancer.
In the present study, we examined the metastatic patterns between ESCC or EAC, including both singleorgan and multi-organ metastases, by utilizing the data extracted from the latest iteration of the SEER database.

Patients
The SEER program is a nationwide database that captures approximately 28% of the United States population. We searched the SEER for data on all patients who were primary histologically con rmed with ESCC (

Statistical analyses
All characteristics of selected patients were included as categorical variables and determined using Pearson's χ2 test or Fisher's exact test. A nearest-neighbor 1:1 propensity score matching analysis with a 0.1-caliper width was used to reduce selection biases. Multivariable logistic regression analyses were done on the propensity-matched sample and were performed for each organ-speci c metastasis. Results were considered signi cant if a two-sided p value of less than 0.05 was obtained. All analyses were conducted using Stata/SE 14.0 (Stata Corp., College Station, TX, USA).

Results
A total of 20,189 patients with esophageal cancer were enrolled, including 6,610 patients with ESCC and 13,579 patients with EAC. Compared with patients in the EAC group, those with ESCC tended to have older age, a higher rate of male, a lower incidence of white race, an earlier T stage, and a greater proportion of choice of surgery. Distant metastatic disease was present in 6502 (32.2%) patients and was found in 25.5% and 35.5% of patients with ESCC and EAC, respectively. Based on metastasis data extracted from the SEER database, the four metastatic lesions (bone, brain, liver, and lung) accounted for 78.1% (5,075/6,502) of all metastatic cases. After propensity score matching, 4597 pairs were selected.
The distribution of baseline patient characteristics was well balanced with all values of standardized differences smaller than 10% after matching. The detailed clinicopathological data and standardized differences of baseline variables before and after propensity score matching were shown in Table 1 and  Distribution of metastasis according to histology There were considerable differences in metastatic patterns between the two groups. As shown in Table 2, liver and lung were the primary sites of distant metastasis and brain was the least frequent metastatic lesion for both ESCC and EAC. The most common site of distant metastasis was lung for ESCC, whereas EAC markedly metastasized to liver more frequently. The metastatic rates of liver, bone, and brain in EAC were much higher than those in ESCC and the frequency of lung metastasis in ESCC was higher than that of EAC. To further validate this nding, multivariable analyses were performed to adjust for confounding variables including age, sex, race, T and N descriptors, and surgery in the matched cohorts. It was demonstrated that compared to ESCC, patients with EAC tended to have more liver metastasis (P < 0.001; 95% con dence interval (CI): 1.52-1.99) and brain metastasis (P < 0.001; 95% CI: 1.43-3.08), and a lower rate of lung metastasis (P < 0.001; 95% CI: 0.47-0.64), with no signi cant difference in bone metastasis (P = 0.255; 95% CI: 0.76-1.07) ( Table 3).

Combination of metastases
Some patients showed metastatic disease at more than one site at the time of diagnosis. The frequencies of all possible combinations of the four metastatic lesions were listed in Table 2. The most frequent bi-organ metastasis was the liver and lung (ESCC: 2.3%, EAC: 3.9%) and the most frequent triorgan metastasis was the liver, lung, and bone (ESCC: 0.7%, EAC: 1.2%). The potential interactions among these metastatic sites were further analyzed (Fig. 3, A-D). Liver preferentially co-metastasized with lung in both cohorts. Brain metastasis was commonly observed in combination with other organ metastases in EAC. In patients with bone metastasis, ESCC had a higher rate of co-metastasis to the lung and liver than brain while EAC was prone to co-metastasis with liver than lung and brain.

Discussion
This study is a comprehensive research for metastatic patterns in the two predominant subtypes of esophageal cancer (ESCC and EAC). Major differences between ESCC and EAC were observed, such as frequencies of single-site metastasis, patterns of combination of metastatic sites. These ndings may help physicians implement better tailored screening modalities and follow-up strategies in future clinical settings.
Compared with ESCC, EAC had a higher rate of liver or brain metastasis and a lower rate of lung metastasis, with a similar trend in metastasis to bone. These results are partially consistent with ndings from two previous SEER studies. Ai et al. found EAC was more likely to have liver, bone, or brain metastasis but had a lower rate of lung metastasis compared with ESCC [10]. The other study of SEER data from 2010 through 2014 reported that patients with EAC had a higher rate of liver or brain metastasis and those with ESCC were more likely to have lung metastasis, with similar proportions of bone metastasis among the two groups [9]. Diverse outcomes might partly be attributable to different sample sizes and confounding clinical variables adjusted in the analysis. In addition, this discrepancy could also be the result of the transition of esophageal histology in the recent decades, which is an increase in the incidence of EAC because of the growing prevalence of obesity and the decline in ESCC incidence after promotion of smoking cessation [2,11,12].
Of note, the proportion of liver metastasis in EAC far outweighs that of in ESCC. This may in part be explained by differences in histological type and tumor location as well as pattern of lymphatic spread, as most cases of EAC are present in the lower esophagus where lymphatic ow tends to be downward, while ESCC is more common in the upper two thirds of the esophagus where lymphatic ow is inclined to be upward [13]. Furthermore, the relative short-distance from lower esophagus to liver might also facilitate the spread of esophageal cancer cells to this organ through the venous drainage of the gastrointestinal tract to the portal vein.
A concerning observation was that a large number of patients with metastatic disease exhibited multiorgan metastases. Among all combined metastases, the most frequent multi-organ metastatic pattern was the liver and lung in both ESCC and EAC groups. In concert with some published literatures, liver and lung are the two main drainage regions frequently colonized by a variety of cancers, targeting these sites of high vascular ow [7,14]. Another observation is that liver or brain metastasis was commonly observed in combination with other organ metastases in EAC. These ndings highlight the essential of regular imaging of chest and liver for all esophageal cancer patients and early PET-CT assessment of patients with EAC, especially in cases suspected with liver or brain metastasis. The mechanism of the propensity for a metastatic cell to spread to several anatomically distinct locations, either sequentially or synchronously, has not been demonstrated explicitly in previous literatures. In some extent, this phenomenon could be explained by the 'seed and soil' hypothesis that the microenvironment of two host organs bear a resemblance to each other thereby facilitating metastasis to the same degree [15].
To our best knowledge, this is the largest population-based study summarizing the metastatic patterns in ESCC and EAC by analyzing the recently released US SEER-18 cancer registry data. Nevertheless, our study is subject to a number of limitations. The rst limitation is the retrospective nature of this study. Second, even though several studies have described molecular changes contributing tumor cells of esophageal cancer to a more aggressive biological behavior [16][17][18][19], the underlying mechanism for differences in metastatic patterns between histological subtypes remains unclear. Additionally, most metastatic lesions were clinically diagnosed without histologically con rmed, which might have resulted in misclassi cation.

Conclusion
In summary, this retrospective study showed that metastatic patterns were different between ESCC and EAC, which have potential to in uence future monitoring strategies depending on clinical settings. Figure 1 Standardized differences of baseline variables between patients with ESCC and patients with EAC before and after propensity score matching. ESCC, esophageal squamous cell carcinoma; EAC, esophageal adenocarcinoma.