Influence of Sex on the Survival of Patients With Esophageal Cancer

Abstract

Purpose

The prognostic value of sex for esophageal cancer survival is currently unclear, and growing data suggest that hormonal influences may account for incidence disparities between men and women. Therefore, moving from the hypothesis that hormones could affect the prognosis of patients with esophageal cancer, we investigated the primary hypothesis that sex is associated with survival and the secondary hypotheses that the relationship between sex and survival depends, at least in part, on age, histology, and race/ethnicity.

Patients and Methods

By using the SEER databases from 1973 to 2007, we identified 13,603 patients (34%) with metastatic esophageal cancer (MEC) and 26,848 patients (66%) with locoregional esophageal cancer (LEC). Cox proportional hazards model for competing risks were used for analyses.

Results

In the multivariate analysis, women had longer esophageal cancer-specific survival (ECSS) than men in both MEC (hazard ratio [HR], 0.949; 95% CI, 0.905 to 0.995; P = .029) and LEC (HR, 0.920; 95% CI, 0.886 to 0.955; P < .001) cohorts. When age and histology were accounted for, there was no difference for ECSS between men and women with adenocarcinoma. In contrast, women younger than age 55 years (HR, 0.896; 95% CI, 0.792 to 1.014; P = .081) and those age 55 years or older (HR, 0.905; 95% CI, 0.862 to 0.950; P < .001) with squamous cell LEC had longer ECSS than men. In the squamous cell MEC cohort, only women younger than age 55 years had longer ECSS (HR, 0.823; 95% CI, 0.708 to 0.957; P = .011) than men.

Conclusion

Sex is an independent prognostic factor for patients with LEC or MEC. As secondary hypotheses, in comparison with men, women age 55 years or older with squamous cell LEC and women younger than age 55 years with squamous cell MEC have a significantly better outcome. These last two findings need further validation.

INTRODUCTION

Esophageal cancer is the eighth most common cancer worldwide, with 482,000 new cases (representing 3.8% of all new cancers) estimated in 2008, and the sixth most common cause of death from cancer with 407,000 deaths (representing 5.4% of all new cancers). Its incidence rates vary internationally more than 15-fold in men and almost 20-fold in women.¹ In the United States, it was estimated that 16,640 new cases of esophageal cancer were diagnosed in 2010 and 14,500 deaths occurred. Esophageal cancer is highly lethal with 11,650 (88.7%) estimated deaths among men and 2,850 (81.2%) among women.² Taken together with previous population studies,^3–9 the latter suggests a survival benefit for women when compared with men.

The prevalence of the two main histologic subtypes—adenocarcinoma and squamous cell carcinoma—differs depending on geographic location. Squamous cell carcinoma of the esophagus (SCCE) predominates in the Middle East, Africa, Asia, and parts of Europe. In contrast, adenocarcinoma of the esophagus (ACE) is prevalent in Western countries.¹⁰ In the United States, the incidence of SCCE has steadily decreased in all ethnicities in the past three decades, with a concurrent increase in the incidence of ACE. In the white population, SCCE represents 27% of esophageal cancers. In contrast, SCCE remains a frequent malignancy in Hispanic, African American, and Asian populations (41%, 81%, and 70% of esophageal cancers, respectively).¹¹

In the United States, both ACE and SCCE are more frequent in men than in women, mirroring parts of the world where SCCE largely predominates.¹ Although this may represent various tumor-specific environmental exposures between sexes (eg, alcohol, tobacco), growing data suggest hormonal influences.^12–14 Sex differences affect esophageal cancer incidence, yet the significance of sex as an independent prognostic marker is unclear. A major limitation of previous studies that examined the prognostic value of sex is the lack of adequate adjustment for other relevant clinical prognostic factors. Therefore, we used the SEER database to assess the influence of sex on the esophageal cancer–specific survival (ECSS) in locoregional esophageal cancer (LEC) and metastatic esophageal cancer (MEC). We evaluated metastatic diseases separately from locoregional diseases, because clinicopathologic prognostic factors and treatments may not have the same influence throughout the evolution of the malignancy. On the basis of our previous data,¹⁵ we hypothesized that hormonal status would influence survival in patients with esophageal cancer and that this influence might vary by histology and tumor stage.

PATIENTS AND METHODS

Study Design

The SEER public use database 1973 to 2007 (Version April 2010) was used for this analysis. The SEER Program, sponsored by the National Cancer Institute, collects information on cancer incidence and survival from 17 population-based cancer registries covering approximately 28% of the United States population.¹⁶

Study Population

The criteria defined for inclusion in this study were primary histologically confirmed esophageal cancer and age at diagnosis of 18 years or older. We excluded a total of 14,169 patients (26%) from those diagnosed with esophageal cancer in the SEER database (n = 54,620) mainly because of unstaged or in situ tumors (n = 11,687), diagnosis not microscopically confirmed or unknown confirmation (n = 2,392), or no follow-up records (n = 1,547). A total of 40,451 patients with esophageal cancer matching the specified criteria were included in the final sample for this analysis (Appendix Figure A1, online only).

Statistical Analysis

The primary end point in this study was ECSS, defined as the period from diagnosis to death from esophageal or gastric cancer. ECSS was censored at the last follow-up, December 31, 2007, or 5 years after diagnosis, whichever came first. Of 40,451 patients, 27,414 died from esophageal or gastric cancer, and 5,776 died from other causes within the first 5 years after diagnosis. The median follow-up time for patients (n = 7,261) who were censored was 39 months, and 39% of them had been observed for at least 5 years. Proportional hazards regression model for competing risks, according to the method of Fine and Gray,¹⁷ was used to adjust for comorbidities that competed with death from esophageal cancer. The variables included in the model were sex, age, ethnicity, marital status, histology, tumor primary site, tumor grade, tumor stage, use of esophagectomy, use of radiation therapy, SEER registries, and year of diagnosis (Appendix Table A1, online only). Because detailed information on TNM staging in the early SEER database was not provided, stage was defined as locoregional (localized or regional in the SEER record description), or metastatic (distant). Because of differences in clinicopathologic factors and in treatments and the potentially different underlying hazard of (ECSS) failure, the analyses were conducted separately for patients with LEC or MEC. Pairwise interactions were examined by using stratified models and were tested by comparing corresponding likelihood ratio statistics between the baseline and nested Cox proportional hazards models, which included the multiplicative product terms. Departures of the proportional hazards assumption for the model were examined graphically by using smoothed plots of weighted Schoenfeld residuals. The primary hypothesis of this study was that sex was an independent prognostic factor for ECSS in both the LEC and MEC cohorts (ie, regardless of extent of disease). The secondary hypothesis tested was that the effect of sex on ECSS varied by age, histology, and/or race/ethnicity. All analyses were performed by using SAS software, version 9.2 (SAS Institute, Cary, NC) and library(cmprsk) in S-PLUS, version 3.3 (Statistical Sciences, Seattle, WA; S-PLUS 7.0 Enterprise Developer for Windows; S-PLUS, 1988, 2005, Insightful Corp.). Results were considered significant if a two-sided P < .05 was obtained. P values were not adjusted for multiple testing of secondary hypotheses.¹⁸

RESULTS

LEC

Patient characteristics.

This study included 26,848 patients with LEC diagnosed from 1973 to 2007. The proportion of patients diagnosed in 1973 to 1982, 1983 to 1995, and 1996 to 2007 was 12.2%, 25.4%, and 62.4%, respectively (Appendix Table A2, online only). ACE and SCCE were more commonly diagnosed in men than in women (ratio of men to women was 6.2:1 and 1.9:1, respectively).

Sex and LEC.

The median age for men was 67 years (range, 19 to 103 years), and the median age for women was 70 years (range, 19 to 107 years). When comparing by histology, women were older than men in both ACE and SCCE cohorts (ACE: women, age 72 years [range, 19 to 103 years]; men, age 67 years [range, 19 to 99 years]; SCCE: women, age 69 years [range, 24 to 107 years]; men, age 66 years [range, 20 to 103 years]). The demographic and clinicopathologic characteristics of men and women with LEC are provided in Table 1.

Table 1.

Demographic and Clinicopathologic Characteristics by Sex in Patients With LEC, SEER Data 1973-2007

Characteristic	LEC (N = 26,848)
	Males (n = 19,957)		Females (n = 6,891)		P
	No	%	No	%
Age, years					< .001
18-44	614	3.1	139	2
45-54	2,573	12.9	626	9.1
55-64	5,422	27.2	1,563	22.7
65-74	6,430	32.2	2,129	30.9
≥ 75	4,918	24.6	2,434	35.3
Race					< .001
White	15,026	75.3	5,196	75.4
African American	2,811	14.1	1,094	15.9
Asian	1,002	5	270	3.9
Hispanic	1,040	5.2	301	4.4
Native American	78	0.4	30	0.4
Histology					< .001
Squamous	9,066	45.4	4,869	70.7
Adenocarcinoma	8,736	43.8	1,410	20.5
Other	2,155	10.8	612	8.9

Abbreviation: LEC, locoregional esophageal cancer.

Sex, age, and survival for LEC.

In the multivariate model (Table 2 and Appendix Table A3, online only), women had significantly longer ECSS than men (4% absolute difference for 5-year ECSS rate). On the basis of our previous data in metastatic colorectal cancer, we examined the relationships between sex and ECSS by age, with patients being dichotomized below or above age 55 years for both men and women.¹⁵ The ECSS was longer for women than for men and did not differ significantly between age groups (P_interaction = 0.32; 4% absolute difference for 5-year ECSS rate in both age groups; Appendix Table A4, online only).

Table 2.

Multivariate Analysis for ECSS in Patients With LEC, SEER Data 1973-2007

Characteristic	LEC (N = 26,848)
	No. of Patients	EC Death	5-Year ECSS ± SE (%)	HR	95% CI*	P*
Sex						< .001
Male	19,957	12,081	24 ± 0.3	1 (reference)
Female	6,891	4,275	28 ± 0.5	0.920	0.886 to 0.955
Age, years						< .001
18-44	753	438	31 ± 1.5	0.774	0.701 to 0.855
45-54	3,199	1,889	29 ± 0.8	0.807	0.764 to 0.852
55-64	6,985	4,124	29 ± 0.5	0.794	0.760 to 0.831
65-74	8,559	5,169	26 ± 0.5	0.848	0.813 to 0.883
≥ 75	7,352	4,736	19 ± 0.5	1 (reference)
Race						.031
White	20,222	12,043	26 ± 0.3	1 (reference)
African American	3,905	2,624	22 ± 0.6	1.068	1.016 to 1.123
Asian	1,272	836	23 ± 1.0	1.089	0.999 to 1.186
Hispanic	1,341	780	26 ± 1.1	0.989	0.916 to 1.067
Native American	108	73	22 ± 3.3	1.133	0.838 to 1.533
Histology						.003
Squamous	13,935	9,185	24 ± 0.4	1 (reference)
Adenocarcinoma	10,146	5,506	27 ± 0.5	0.927	0.889 to 0.968
Other	2,767	1,665	24 ± 0.7	0.965	0.911 to 1.022

Abbreviations: EC, esophageal cancer; ECSS, esophageal cancer–specific survival; LEC, locoregional esophageal cancer; HR, hazard ratio.

^*Based on ECSS in competing risks regression model, including all variables in the table and marital status, radiation sequence (no radiation/surgery, neoadjuvant radiation, adjuvant radiation, intraoperative radiation or unknown), stage (localized v regional), SEER registries, and year of diagnosis (1973-1982, 1983-1995, 1996-2007).

Sex, age, histology, ethnicity, and survival for LEC.

We found a significant difference in ECSS between sexes in adenocarcinoma and squamous cell carcinoma (p_interaction = 0.036, Appendix Table A4). ECSS was significantly longer for women when compared with men with squamous cell tumors (5% absolute difference for 5-year ECSS rate; Fig 1). In adenocarcinoma, this sex difference was not significant (Appendix Table A4).

Fig 1.

Adjusted survival curves of esophageal cancer–specific survival (ECSS) in patients with locoregional squamous cell carcinoma of the esophagus.

Finally, the relationship between sex, histology, and age was analyzed. In patients with adenocarcinoma, there was no significant difference in ECSS between men and women in patients younger than age 55 years or age 55 years or older. In contrast, women (n = 4,300) with squamous cell LEC age ≥ 55 years old had significantly longer ECSS than men (n = 7,660) of similar age (HR, 0.905; 95% CI, 0.862 to 0.950; P < .001; 5% absolute difference for 5-year ECSS rate). There was a trend in the association between sex (1,406 men and 569 women) and ECSS in squamous cell LEC for those younger than age 55 years (HR, 0.896; 95% CI, 0.792 to 1.014; P = .081; 4% absolute difference for 5-year ECSS rate). The effect of sex on ECSS was consistent across all ethnicities (P_interaction > .05; Appendix Table A4).

MEC

Patient characteristics.

This study included 13,603 patients with MEC diagnosed from 1973 to 2007. The proportion of patients with MEC diagnosed in 1973 to 1982, 1983 to 1995, and 1996 to 2007 was 12.1%, 23.1%, and 64.8%, respectively (Appendix Table A2). ACE and SCCE were more commonly diagnosed in men than in women (ratio of men to women was 6.1:1 and 3.6:1, respectively).

Sex and MEC.

The median age for men was 64 years (range, 22 to 97 years), and the median age for women was 67 years (range, 21 to 95 years). When analyzing by histology, women were older in both ACE and SCCE cohorts (ACE: women, age 69 years [range, 21 to 95 years]; men, age 64 years [range, 22 to 97 years]; SCCE: women, age 66 years [range, 26 to 95 years]; men, age 64 years [range, 23 to 95 years]). The demographic and clinicopathologic characteristics of men and women with MEC are provided in Table 3.

Table 3.

Demographic and Clinicopathologic Characteristics by Sex in Patients With MEC, SEER Data 1973-2007

Characteristic	MEC (N= 13,603)				P
	Males (n = 10,752)		Females (n = 2,851)
	No.	%	No.	%
Age, years					< .001
18-44	444	4.1	104	3.6
45-54	1,776	16.5	377	13.2
55-64	3,258	30.3	737	25.9
65-74	3,202	29.8	889	31.2
≥ 75	2,072	19.3	744	26.1
Race					< .001
White	7,996	74.4	2,012	70.6
African American	1,521	14.1	550	19.3
Asian	555	5.2	128	4.5
Hispanic	628	5.8	140	4.9
Native American	52	0.5	21	0.7
Histology					< .001
Squamous	4,162	38.7	1,660	58.2
Adenocarcinoma	5,171	48.1	841	29.5
Other	1,419	13.2	350	12.3

Abbreviation: MEC, metastatic esophageal cancer.

Sex, age, and survival for MEC.

In the multivariate model (Table 4 and Appendix Table A3), women had significantly longer ECSS than men (2% absolute difference for 2-year ECSS rate). The difference in ECSS between sexes varied significantly by age in patients with metastatic disease (P_interaction = .048; Appendix Table A4). Women with MEC younger than age 55 years had an ECSS better than men of similar age (5% absolute difference for 2-year ECSS rate). In contrast, this sex difference did not persist in patients ≥ 55 years old (2% absolute difference for 2-year ECSS rate).

Table 4.

Multivariate Analysis for ECSS in Patients With MEC, SEER Data 1973-2007

Characteristic	MEC (N = 13,603)
	No.	EC Death	2-Year ECSS ± SE (%)	HR	95% CI*	P*
Sex						.029
Male	10,752	8,769	9 ± 0.3	1 (reference)
Female	2,851	2,289	11 ± 0.5	0.949	0.905 to 0.995
Age, years						< .001
18-44	548	439	14 ± 1.3	0.816	0.742 to 0.898
45-54	2,153	1,771	11 ± 0.6	0.910	0.856 to 0.968
55-64	3,995	3,274	11 ± 0.4	0.921	0.872 to 0.973
65-74	4,091	3,297	10 ± 0.4	0.907	0.858 to 0.958
≥ 75	2,816	2,277	7 ± 0.4	1 (reference)
Race						.20
White	10,008	8,166	10 ± 0.3	1 (reference)
African American	2,071	1,683	9 ± 0.5	1.035	0.974 to 1.101
Asian	683	551	11 ± 1.0	0.942	0.854 to 1.039
Hispanic	768	597	11 ± 0.9	0.929	0.854 to 1.010
Native American	73	61	7 ± 2.0	1.087	0.788 to 1.499
Histology						.027
Squamous	5,822	4,695	10 ± 0.4	1 (reference)
Adenocarcinoma	6,012	4,929	10 ± 0.4	1.071	1.019 to 1.125
Other	1,769	1,434	9 ± 0.5	1.051	0.985 to 1.121

Abbreviations: EC, esophageal cancer; ECSS, esophageal cancer–specific survival; HR, hazard ratio; MEC, metastatic esophageal cancer.

^*Based on ECSS in competing risks regression model, including all variables in the table and marital status, radiation sequence (no radiation/surgery, neoadjuvant radiation, adjuvant radiation, intraoperative radiation or unknown), SEER registries, and year of diagnosis (1973-1982, 1983-1995, 1996-2007).

Sex, age, histology, ethnicity, and survival for MEC.

There was no significant difference for ECSS between women and men in both ACE and SCCE (Fig 2A and Appendix Table A4). However, when the relationship between sex, histology, and age was analyzed, women younger than age 55 years (n = 284) had significantly longer ECSS than men of the same age (n = 791; HR, 0.823; 95% CI, 0.708 to 0.957; P = .011; Fig 2B; 5% absolute difference for 2-year ECSS rate). In contrast, women with SCCE age ≥ 55 years (n = 1,376) had no significantly different ECSS than men of same age (n = 3,371; HR, 0.967; 95% CI, 0.902 to 1.037; P = .35). In patients with ACE, there was no significant difference in ECSS between men and women in age groups younger than age 55 years or age 55 years or older. The effect of sex on ECSS was consistent across all ethnicities (P_interaction > .05; Appendix Table A4).

Fig 2.

Adjusted survival curves of esophageal cancer-specific survival (ECSS) in (A) all patients with metastatic squamous cell carcinoma of the esophagus and (B) in patients with metastatic squamous cell carcinoma of the esophagus younger than age 55 years.

DISCUSSION

The important burden of esophageal cancer in the United Stated is linked with its poor outcome (eighth position for cancer deaths), with both squamous cell and adenocarcinoma histology having poor prognosis.² In the last 25 years, there has been a shift in esophageal histology with a decrease of SCCE and an increase in ACE. This shift has been attributed to the change in commonly reported risk factors.^19–21 Although women experienced a shift in esophageal cancer histology similar to that of men, their rates of SCCE decrease and ACE increase were lower.²² Importantly, despite this gradual shift, SCCE remains a frequent malignancy representing 35% of all carcinomas of the esophagus.¹¹

We demonstrate in one of the largest cohorts of patients with esophageal cancer that sex is an independent prognostic marker for patients with either LEC or MEC. Furthermore, our data suggest that sex may be prognostic only in squamous cell esophageal tumors. Previous studies have reported a prognostic value for sex in Asian patients with stage I to III SCCE.^7,23,24 We hypothesized that hormonal influences, and thus menopause, could explain this prognostic difference. Age is commonly considered as a surrogate for menopause. On the basis of our previous study,¹⁵ we selected age 55 years as a surrogate for menopause. This predetermined cutoff was chosen because menopause occurs in only 5% of women after age 55 years.²⁵ We found that in metastatic SCCE, only women younger than age 55 years had a lower risk of dying of esophageal cancer when compared with men of similar age. In contrast, women younger than or older than age 55 years with locoregional SCCE had a better prognosis, with the latter being statistically significant.

Estrogen receptors (ERs) are expressed in SCCE, and early preclinical studies^26,27 have demonstrated that estrogens could inhibit squamous cell tumor growth. The influence of sex hormones was further demonstrated in in vivo studies.^26,28 These antiproliferative functions are likely to occur through ER beta (ERβ), which is the predominant ER expressed in SCCE.²⁹ In contrast to ER alpha (ERα), ERβ has antiproliferative functions.³⁰ Unfortunately, the link between ER and esophageal cancer is currently poorly understood, mostly because of the confounding role of circulating estrogen and the concomitant expression of tumor androgen receptors (ARs). Current literature supports a biologic role of estrogens in early SCCE.^12,31 These data contrast with our study, showing no survival difference between women younger than age 55 years and those age 55 years or older with localized SCCE. This apparent contradiction may be reconciled by taking into account the prosurvival signals induced by androgens. ARs are expressed in SCCE and have been shown to enhance tumor growth.^26,27,32 Therefore, it is not surprising in our study that men with localized SCCE have an overall worse outcome. Our data show that the impact of age is different in localized and metastatic SCCE, suggesting that ERβ pathway signaling may be dynamic. Metastatic squamous cell tumors may lose their inhibitory ERβ capacity, which would reduce the survival benefit for women, particularly in postmenopausal women with low levels of ERβ ligand. Further studies are needed to clarify the mechanism and interaction of sex steroids in esophageal squamous cell tumors.

Although hormonal influences are strong candidates for explaining this sex difference in outcome, other prognostic factors should be considered. One candidate is the human papilloma virus (HPV). Oncogenic types of HPV have been reported to play a significant role in SCCE carcinogenesis in high-risk geographic areas (eg, East Asia). In contrast, its role in the United States is less clear, with low detection rates being reported.^33,34 Moreover, in contrast to head and neck cancer in which HPV infection predominates in males and is correlated with better prognosis,³⁵ no sex imbalance or prognostic value has been shown in SCCE.^36,37 Other potential prognostic factors are alcohol or smoking status. Because the SEER database provides no information on smoking status or alcohol consumption, we could not investigate their potential prognostic influence.

In agreement with a recent SEER-based study,³⁸ women with LEC or MEC and adenocarcinoma histology did not have a different outcome from men, irrespective of their menopausal status. ERβ and AR protein expression have been documented in adenocarcinoma of the esophagus.^29,39,40 However to date, convincing preclinical data showing a direct influence of sex hormones on adenocarcinoma cells or xenografts are lacking. Altogether, although hormonal influences may play a role in the carcinogenesis of esophageal adenocarcinoma,^14,41,42 ER or AR pathways may drive established adenocarcinoma of the esophagus only marginally, if at all.

Our study has several limitations related to the lack of details provided by the SEER database on the TNM tumor stage (only recently reported), on treatment information for the sequence of radiation therapy, on systemic treatment (no information), and on socioeconomic status. Despite these limitations, our results remained significant in the LEC cohort after adjusting for localized or regional stage. Our analyses also remained significant after adjusting for the year of diagnosis as a surrogate for type of therapy as follows: the study population was divided into three subsets based on year of diagnosis (1973 to 1982, 1983 to 1995, and 1996 to 2007). These cutoffs were selected on the basis of relevant publications likely to have driven a change of practice.^43–45 Adjustment for the year of diagnosis also allowed controlling for the quality of the elder SEER database. Although details on socioeconomic status are not provided, our results remain significant after adjusting for marital status, previously shown to be prognostic in colorectal cancer.⁴⁶

Because of the lack of details on comorbidities in the SEER database, we restricted our analyses to ECSS, which is influenced to a lesser degree than overall survival. However, we cannot exclude that our data reflect an imbalance between comorbidities between sexes. It is possible that misclassification of cardia adenocarcinoma might have obscured the prognostic influence of sex steroids in ACE.⁴⁷ Finally, our findings derived from our secondary hypotheses have not been adjusted for multiple testing and therefore should be confirmed in futures studies. Despite these limitations, studies performed in population-based registries provide unique opportunities to evaluate a large number of patients. This approach is particularly useful in diseases such as esophageal cancer for which few large prospective studies are available.

Despite aggressive multimodality treatment strategies for localized SCCE, the expected 2-year survival rate remains between 20% and 50%.⁴⁸ To set a benchmark, it has been estimated that adding concomitant chemotherapy to radiotherapy in localized SCCE provides a 7% reduction to 2-year mortality.⁴⁹ Therefore, although the magnitude of the effect of sex in localized SCCE can be seen as small (5% absolute difference for 5-year ECSS), this may translate to the development of clinically meaningful treatment strategies in a disease in which there are a limited number of active agents. Similarly, in the metastatic setting in which no phase III trials have ever been reported, our data showing a 5% absolute difference for 2-year ECSS rate between men and women younger than age 55 years is intriguing. Nevertheless, when considering the relatively low prevalence of ECSS, the latter may have a modest impact in the United States. These data may be of critical importance for other parts of the world with a higher prevalence of SCCE.

In summary, this study shows that sex is an independent prognostic factor regardless of the extent of the disease. Several findings derived from our secondary hypotheses should be confirmed in another independent cohort: the better outcome seen in women with squamous cell MEC may be restricted to women younger than age 55 years; women younger than or older than age 55 years with squamous cell LEC have a better prognosis, with the latter being statistically significant; sex is not a significant prognostic factor in established esophageal adenocarcinoma. We suggest that these associations are possibly related to both estrogen and androgen exposition. Taken together, these data are important because SCCE remains an important burden despite the shift in histology occurring in the Western world. Moreover, in areas of the world where the SCCE incidence is high, this shift has yet not been observed with relatively stable SCCE incidences.⁵⁰ If our secondary hypotheses are validated in other databases, future correlative studies in SCCE should focus on the pathways of sex steroids to elucidate the molecular mechanisms linked with poor outcome. With better understanding, the activation of these pathways may be modified, creating new approaches in SCCE treatment.

Appendix

Table A1.

Definitions of Variables and Cut Points

Variable	Definitions and Cut Points
Sex	Male and female
Age	18-44, 45-54, 55-64, 65-74, and ≥ 75 years
Race/ethnicity	White, African American, Asian/Pacific Islander, Hispanic (identified as having Spanish/Hispanic surname or of Spanish origin regardless of the category of race/ethnicity), and Native American
Histology	Adenocarcinoma, squamous cell carcinoma, and other (other histology or not specified)
Tumor primary site	Lower third (C152, C155), middle third (C151, C154), upper third (C150, C153), and overlapping or not specified (C158, C159)
Tumor grade/differentiation	Well differentiated (grade 1), moderately differentiated (grade 2), poorly differentiated or undifferentiated (grade 3 or 4), and not determined
Treatment	Esophagectomy (partial or total esophagectomy) and radiation therapy (beam radiation or radiation; method or source not specified). Detailed information on surgery of the primary site was not available prior to 1983. Records on chemotherapy were not available in the SEER database.
Radiation sequence	No radiation/surgery, neoadjuvant radiation, adjuvant radiation, and intraoperative radiation or unknown
Marital status at diagnosis	Not married (including never married, separated, divorced, widowed, and unknown) and married (including common law)
Stage	Metastatic (SEER historic stage A code 4) and locoregional (localized or regional corresponding to SEER historic stage A code 1 or 2).
SEER registry	17 SEER registry sites (http://seer.cancer.gov/registries/list.html)
Year of diagnosis	Three intervals: 1973-1982, 1983-1995, and 1996-2007

Table A2.

No. of Patients by Three Intervals of Year of Diagnosis and Inclusion and Exclusion Criteria

Variable	Year of Diagnosis
	1973-1982		1983-1995		1996-2007
	No.	%	No.	%	No.	%
Total No. of patients diagnosed with esophageal cancer in the SEER database	7,400		14,112		33,108
Total No. of patients with in situ or unstaged esophageal cancer	2,008	27	3,667	26	6,012	18
Total No. excluded from the analysis	2,466	33	4,159	29	7,544	23
Total No. included in the analysis	4,934	67	9,953	71	25,564	77
Stage in patients included in the analysis
Locoregional	3,284	67	6,808	68	16,756	66
Metastatic	1,650	33	3,145	32	8,808	34

Table A3.

Multivariate Analysis for ECSS in Patients With EC by Stage, SEER Data 1973-2007

Characteristic	LEC (n = 26,848)*				MEC (n = 13,603)*
	No. of Patients	HR†	95% CI	P†	No. of Patients	HR†	95% CI	P†
Site
Lower third	13,597	1 (reference)			7,598	1 (reference)
Middle third	7,462	1.074	1.031 to 1.119		2,908	1.021	0.969 to 1.075
Upper third	2,868	0.920	0.868 to 0.975		964	0.958	0.884 to 1.039
Overlapping or NOS	2,921	1.094	1.037 to 1.154	< .001	2,133	0.986	0.932 to 1.043	.44
Tumor grade
1	1,695	0.695	0.647 to 0.747		475	0.829	0.751 to 0.915
2	9,108	0.847	0.818 to 0.878		3,679	0.867	0.831 to 0.905
3 or 4	10,944	1 (reference)			6,707	1 (reference)
Not determined	5,101	0.785	0.751 to 0.820	< .001	2,742	0.868	0.825 to 0.912	< .001
Radiation
No or unknown	10,283	1 (reference)			6,226	1 (reference)
Yes	16,565	0.993	0.952 to 1.035	.73	7,377	0.806	0.775 to 0.838	< .001
Esophagectomy
No or unknown	18,793	1 (reference)			12,746	1 (reference)
Yes	8,055	0.593	0.566 to 0.622	< .001	857	0.651	0.600 to 0.707	< .001
Radiation sequence
No radiation/surgery	21,807	1 (reference)			12,577	1 (reference)
Neoadjuvant radiation	2,886	0.873	0.818 to 0.932		346	0.700	0.619 to 0.793
Adjuvant radiation	1,967	0.930	0.875 to 0.989		641	0.936	0.866 to 1.012
Intraoperative radiation or unknown	188	0.914	0.777 to 1.074	< .001	39	0.966	0.735 to 1.269	< .001
Marital status at diagnosis
Not married	11,383	1 (reference)			5,698	1 (reference)
Married	15,465	0.893	0.864 to 0.923	< .001	7,905	0.934	0.899 to 0.971	< .001

Abbreviations: EC, esophageal cancer; ECSS, esophageal cancer–specific survival; HR, hazard ratio; LEC, locoregional esophageal cancer; MEC, metastatic esophageal cancer; NOS, not otherwise specified.

^*In all, 16,356 ECSS deaths and 4,154 non-ECSS deaths occurred among the LEC cohort, and 11,058 ECSS deaths and 1,622 non-ECSS deaths occurred among the MEC cohort.

^†Based on disease-specific survival in competing risks regression model, including all variables in this table plus sex, age, race, and histology, SEER registries, and year of diagnosis (1973-1982, 1983-1995, 1996-2007). Stage (localized v regional) was adjusted in patients with LEC.

Table A4.

Summarizing the Tests for Interaction: Testing for the Consistency for HR Comparing ECSS Between Women and Men, According to Age, Histology, and Race

Characteristic	LEC			MEC
	HR	95% CI	P for Interaction*	HR	95% CI	P for Interaction*
Age, years			.32			.048
< 55	0.893	0.798 to 0.998		0.812	0.727 to 0.906
≥ 55	0.922	0.886 to 0.960		0.975	0.926 to 1.028
Histology			.036			.18
Squamous	0.907	0.867 to 0.948		0.945	0.887 to 1.006
Adenocarcinoma	1.002	0.926 to 1.084		1.018	0.937 to 1.106
Race			.80			.70
White	0.917	0.877 to 0.959		0.963	0.910 to 1.019
African American	0.901	0.825 to 0.984		0.896	0.807 to 0.996
Asian	0.909	0.764 to 1.080		0.889	0.718 to 1.100
Hispanic	1.035	0.864 to 1.239		1.058	0.859 to 1.304

Abbreviations: ECSS, esophageal cancer–specific survival; HR, hazard ratio; LEC, locoregional esophageal cancer; MEC, metastatic esophageal cancer.

^*.Patients with other histology were excluded for testing interaction between sex and histology. Native American patients were excluded for testing interaction between sex and race because of their small number.

Fig A1.

A flowchart of patient selection from the SEER database.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The author(s) indicated no potential conflicts of interest.

AUTHOR CONTRIBUTIONS

Conception and design: Pierre Bohanes, Dongyun Yang, Ruchika S. Chhibar, Melissa J. Labonte, Thomas Winder, Yan Ning, Wu Zhang, Heinz-Josef Lenz

Financial support: Heinz-Josef Lenz

Administrative support: Rita El-Khoueiry

Collection and assembly of data: Dongyun Yang, Ruchika S. Chhibar, David Paez, Rita El-Khoueiry

Data analysis and interpretation: Pierre Bohanes, Dongyun Yang, Ruchika S. Chhibar, Melissa J. Labonte, Armin Gerger, Léonor Benhaim, Takeru Wakatsuki, Fotios Loupakis, Wu Zhang, Heinz-Josef Lenz

Manuscript writing: All authors

Final approval of manuscript: All authors