Cancer Causes & Control

, Volume 23, Issue 4, pp 555–564

Predictors of survival among patients diagnosed with adenocarcinoma of the esophagus and gastroesophageal junction

Authors

  • Aaron P. Thrift
    • Population Health DepartmentQueensland Institute of Medical Research
    • School of Population HealthThe University of Queensland
  • Christina M. Nagle
    • Population Health DepartmentQueensland Institute of Medical Research
  • Paul P. Fahey
    • Population Health DepartmentQueensland Institute of Medical Research
  • Bernard M. Smithers
    • Division of Surgery, Upper GI and Soft Tissue UnitPrincess Alexandra Hospital
  • David I. Watson
    • Department of SurgeryFlinders University
    • Population Health DepartmentQueensland Institute of Medical Research
    • Cancer Control Laboratory, Queensland Institute of Medical ResearchRoyal Brisbane Hospital
Original paper

DOI: 10.1007/s10552-012-9913-1

Cite this article as:
Thrift, A.P., Nagle, C.M., Fahey, P.P. et al. Cancer Causes Control (2012) 23: 555. doi:10.1007/s10552-012-9913-1

Abstract

Purpose

Patients diagnosed with esophageal adenocarcinoma (EAC) or gastroesophageal junction adenocarcinoma (GEJAC) have poor survival. We investigated the possible influence of pre-morbid lifestyle factors on survival for these lethal cancers.

Methods

This study included a population-based cohort of patients with EAC (n = 362) and GEJAC (n = 421) tumors. Detailed information about demographic and lifestyle factors was obtained around the time of diagnosis, and deaths were identified using the National Death Index. Hazard ratios (HR) and 95% confidence intervals (95% CI) were calculated from Cox proportional hazards models, adjusted for age, sex, pre-treatment American Joint Committee on Cancer tumor stage, treatment and presence of comorbidities.

Results

Median follow-up for mortality was 6.4 years. Five-year survival rates were 27 and 33% for EAC and GEJAC, respectively. As expected, tumor and treatment characteristics were the strongest predictors of survival for both cancer sites. Among patients diagnosed with GEJAC tumors, those who were older (≥70 years, adjusted HR = 1.70, 95% CI 1.24–2.32) and those who reported being current smokers (adjusted HR = 1.45, 95% CI 1.02–2.06) fared worse. Other lifestyle factors putatively associated with risk of developing GEJAC including body mass index, gastroesophageal reflux symptoms, alcohol, and use of non-steroidal anti-inflammatory drugs were not associated with survival. Likewise, after adjusting for stage and treatment, no clear associations were detected between lifestyle factors and survival among patients with EAC tumors. We found similar results for analyses restricted to patients treated surgically.

Conclusions

Overall, our data suggest that lifestyle factors do not appear to unduly influence survival for these cancers.

Keywords

Esophageal adenocarcinomaGastroesophageal junctionSurvivalLifestyle factors

Introduction

There are two main subtypes of esophageal cancer: squamous cell carcinoma and adenocarcinoma. While incidence rates for esophageal squamous cell carcinoma (ESCC) have been decreasing, the increase in incidence of esophageal adenocarcinoma (EAC) since the 1970s has been among the highest for any cancer in Western populations [13]. A similar pattern, albeit at a lower rate, has also been reported for gastroesophageal junction adenocarcinoma (GEJAC). The principal causes of the increase in incidence of EAC, and to a lesser extent GEJAC, are thought to be the increasing prevalences of both gastroesophageal reflux (GER) and central obesity in white men [48]. Tobacco smoking is also an important risk factor for these cancers, whereas Helicobacter pylori infection and the frequent use of non-steroidal anti-inflammatory drugs (NSAIDs) have been associated with reduced risks, although the magnitude of their effects varies across the two tumor sites [911].

The majority of patients with EAC and GEJAC present with metastatic disease, precluding them from definitive treatments, and this results in poor overall survival. In a recent Australian study, 5-year survival rates for patients diagnosed with EAC were 22% for localized cancer, 12% for cancers with regional spread, and 2% for disseminated cancers [1]. There is currently no comparable information available for patients with GEJAC tumors. Resection with or without neoadjuvant therapy is considered gold standard when therapy is aiming for cure; however, the long-term prognosis following resection also remains poor [12]. The strongest prognostic factors for EAC are the pathologic stage and tumor grade at diagnosis [1, 13, 14]. It is important therefore to identify if any additional factors, not presently measured at the time of patient presentation, might predict clinical outcomes and inform treatment decisions.

There is some evidence that lifestyle factors that are important in the etiology of these tumors might also affect survival, independent of known clinico-pathologic predictors. In a recent Australian study, we found that smoking and alcohol consumption were associated with poor survival for ESCC, independent of the effects of other prognostic factors [15]. Only two previous population-based studies have investigated whether lifestyle factors influence survival for EAC [12, 16]. While one study reported a better prognosis for EAC in obese patients [12], there was no association in the other study [16], and it remains unclear whether the inverse association was confounded by tumor stage [17, 18]. Other lifestyle factors, including smoking and alcohol, did not influence EAC survival [12, 16]. To our knowledge, no previous population-based studies have investigated the effect of lifestyle factors on survival among patients diagnosed with GEJAC tumors.

Given the universally poor outcomes of patients with these cancers, we sought to evaluate the role of lifestyle factors in predicting survival from EAC, and for the first time, GEJAC using data from a large, population-based cohort of patients followed for a minimum of 5 years using a national death registration system.

Methods

The study cohort included 783 patients aged 18–79 years with histologically confirmed primary EAC (n = 362) or GEJAC (n = 421), diagnosed between July 1 2001 and June 30 2005, and identified through major treatment centers and state-based cancer registries throughout Australia. Full details of the original study have been reported previously [19]. Briefly, 1,577 patients with esophageal cancer received an invitation to participate in the study, of whom 1,102 returned a completed questionnaire (70% of all invited). Ten patients were subsequently excluded following pathology review. Ethics approval for the research was received from the Queensland Institute of Medical Research, and all the hospitals from where the participants were recruited. Written informed consent was obtained from all participants in the original study, which included granting study investigators full access to medical records.

Baseline data collection

Upon enrollment, participants completed a health and lifestyle questionnaire asking about their education, occupation, general health, height, weight (1 year prior to diagnosis and heaviest ever), smoking history, alcohol consumption, physical activity, frequency of symptoms of GER, and frequency of use of aspirin and non-aspirin NSAIDs in the 5 years prior to diagnosis. Body mass index (BMI) was calculated by dividing weight (kg) by height (m2), and standard categories were used for analyses (<25 healthy weight; 25–29.9 overweight; 30–34.9 obese; ≥35 morbidly obese). Participants were asked whether, over their whole life, they had ever smoked more than 100 cigarettes, cigars or pipes; positive responses elicited further questions about consumption and duration of smoking. Current smokers and ex-smokers were defined by their smoking status at 1 year prior to their diagnosis with EAC or GEJAC [20]. Cumulative exposure to cigarettes in pack-years was calculated by multiplying the average number of cigarettes smoked per day by the number of years smoked and dividing by 20. We asked participants to report the frequency with which they consumed alcohol (light beer, regular beer, white wine, red wine, port/sherry, and spirits/liqueurs) at ages 20–29, 30–49, and ≥50 years, as applicable [19]. For these analyses, total alcohol consumption was summed across all the age groups for all types of alcohol, from which we calculated a weighted average number of standard drinks (10 g ethanol) consumed per week between age 20 and age at diagnosis. We derived a three-level physical activity index (low, medium, and high) based on frequency and intensity of physical activity at work and in leisure time [21]. A history of GER was elicited by asking about experience of heartburn (“a burning pain behind the breastbone after eating”) or acid reflux (“a sour taste from acid or bile rising up into the mouth or throat”) at ages 10–19, 20–29, 30–49, and 50–79 years [5]. For analysis, we used the highest reported frequency for either symptom during the age interval coinciding with 10 years before diagnosis. Participants were also asked to report the presence of any medical conditions (from a predefined list) and/or any other medical conditions requiring regular medical care. These data were classified using the comorbidity categories defined by Charlson et al. [22], and dichotomized (none or ≥1 comorbidities).

Clinical data collection

Clinical and pathologic information was abstracted from the medical records by trained abstracters, entered on standardized data collection sheets, cleaned and checked. Information was collected regarding stage of disease at diagnosis (pre-treatment), tumor grade, treatment and various other clinical and pathologic prognostic variables [23]. Pre-treatment tumor stage (hereafter referred to as stage) was defined according to the American Joint Committee on Cancer (AJCC) stage groups for esophageal cancer [24]. As previously described [23], for patients who did not have AJCC stage recorded in their medical notes, we imputed AJCC stage using tumor-node-metastasis (TNM) codes, fluorodeoxyglucose positron emission tomography scan results for M status and endoscopic ultrasound for T and N status. However, for about half of the patients in this population sample, we were unable to impute AJCC stage with sufficient precision and so this group was classified separately as “AJCC stage undetermined”. Tumor grade was defined as well differentiated, moderately differentiated, or poorly/undifferentiated as per the pathologist report. Finally, we defined curative treatment intent as attempted resection with or without neoadjuvant therapy and/or definitive chemoradiotherapy (i.e., combined chemotherapy and radiotherapy where radiotherapy was targeted at the esophagus and dosage was 50 Gray or more).

Outcomes

Personal identifiers (full name, date of birth, sex, date, and state of residence of last known contact) were used to link the cohort to the National Death Index, which contains records of all deaths that have occurred in Australia since 1980. Person-time was accumulated from a patient’s date of consent to their date of death or to the date of last follow-up (September 15 2010). Median time from date of diagnosis to date of consent was 91 days. We had complete follow-up on mortality through the National Death Index, with the cohort followed for between 4.8 and 9.0 years from date of consent (median 6.4 years). Because the National Death Index has a lag time of more than 18 months for coded cause of death [25], the ICD-coded cause of death information was available for 356 patients who had died (63%); of these, 84% had died from esophageal cancer. We used all-cause mortality as the primary endpoint for follow-up.

Statistical analysis

Overall and stratified survival distributions were estimated and plotted using the Kaplan–Meier technique, and log-rank tests were used to assess any heterogeneity in survival curves. Estimates of hazard ratios (HR) and 95% confidence intervals (95% CI) were obtained from Cox proportional hazards regression analysis. Terms for potential confounders were retained in the final models if they changed the β coefficient by 10% or more or improved the fit of the models. Analyses shown are adjusted for age, sex, AJCC stage (I, II, III, IV, and undetermined), treatment intent, and presence of comorbid conditions. To test for linear trends across categories, we modeled the median of each category as a continuous variable. For variables in which the lowest category was “unexposed” (e.g., pack-years of smoking), trend tests were restricted to the “exposed” categories. We used stratified analyses to assess effect modification and evaluated significance of interactions by assessing the p value for the type III analysis of effects for the interaction term. Statistical significance was determined at α = 0.05, and all tests for statistical significance were two-sided. All analyses were performed by using SAS version 9.2 (SAS Institute, Inc, Cary, NC).

Results

The majority of patients with EAC and GEJAC were male, and the median age was 64 years for both tumor sites. Of those with tumor stage recorded, 59% of patients with EAC and 48% of patients with GEJAC had advanced stage disease (stages III–IV) (Table 1). Most patients had high-grade (poorly or moderately differentiated) tumors (93 and 89% for EAC and GEJAC, respectively). Patients with GEJAC were more likely to have undergone surgery than patients with EAC tumors (85% vs. 53%). As expected, factors known to be associated with the development of these cancers were overrepresented in the cohort (EAC, 75% had GER symptoms, 80% were overweight or obese, and 74% were current or ex-smokers; GEJAC, 72% had GER symptoms, 74% were overweight or obese, and 77% were current or ex-smokers).
Table 1

Baseline characteristics and the proportion of patients with EAC and GEJAC who survived 5 years

Variables

EAC

GEJAC

Baseline na

5-year survival n (% survived)

pb

Baseline na

5-year survival n (% survived)

pb

Total

362

98 (27)

 

421

138 (33)

 

Sex

 Men

327

89 (27)

 

366

113 (31)

 

 Women

35

9 (26)

0.83

55

25 (45)

0.04

Age group

 <60 years

129

39 (30)

 

156

63 (40)

 

 60–69 years

133

29 (22)

 

147

52 (35)

 

 ≥70 years

100

30 (30)

0.26

118

23 (19)

0.001

Highest level of education

 No tertiary qualification

170

48 (28)

 

172

49 (28)

 

 Technical college/diploma

170

44 (26)

 

205

71 (35)

 

 University degree

22

6 (27)

0.86

44

18 (41)

0.17

AJCC stagec

 I

21

14 (67)

 

25

17 (68)

 

 II

53

22 (42)

 

81

36 (44)

 

 III

43

7 (16)

 

71

16 (23)

 

 IV

65

3 (5)

<.001

26

1 (4)

<.001

Tumor grade

 Well differentiated

24

12 (50)

 

42

16 (38)

 

 Moderately differentiated

135

44 (33)

 

136

55 (40)

 

 Poorly/undifferentiated

169

32 (19)

<.001

221

60 (27)

0.02

Treatment intent

 Curative intent

  Attempted re-section and chemoradiotherapy

190

75 (39)

 

355

132 (37)

 

  Chemoradiotherapy only

29

6 (21)

 

19

3 (16)

 

 Non-curative

137

15 (11)

<.001

45

1 (2)

<.001

Number of comorbidities

 0

168

38 (23)

 

207

69 (33)

 

 ≥1

194

60 (31)

0.08

214

69 (32)

0.86

Average lifetime alcohol consumptiond (standard drinks/week)

 <1e

37

9 (24)

 

49

16 (33)

 

 1–6

83

27 (33)

 

115

42 (37)

 

 7–20

131

27 (21)

 

143

47 (33)

 

 ≥21

109

33 (30)

0.19

112

33 (29)

0.75

Smoking status

 Never smoker

92

29 (32)

 

97

40 (41)

 

 Ex-smoker

195

48 (25)

 

208

60 (29)

 

 Current smoker

73

19 (26)

0.49

113

37 (33)

0.10

Cumulative smoking history

 Never smoker

92

29 (32)

 

97

40 (41)

 

 0–14.9 pack-years

70

23 (33)

 

86

25 (29)

 

 15.0–29.9 pack-years

69

21 (30)

 

91

28 (31)

 

 ≥30 pack-years

131

25 (19)

0.07

147

45 (31)

0.24

BMI last year (kg/m2)

 <25.0

71

17 (24)

 

106

30 (28)

 

 25.0–29.9

147

37 (25)

 

165

55 (33)

 

 30.0–34.9

89

35 (39)

 

98

34 (35)

 

 ≥35.0

40

6 (15)

0.02

36

13 (36)

0.69

Frequency of GER symptomsf

 Never

79

13 (16)

 

118

32 (27)

 

 Less than weekly

128

36 (28)

 

148

45 (30)

 

 At least weekly

151

47 (31)

0.05

152

61 (40)

0.06

Physical activity index

 Low

92

26 (28)

 

92

33 (36)

 

 Moderate

125

33 (26)

 

151

45 (30)

 

 High

144

39 (27)

0.96

176

60 (34)

0.58

Frequency of use of NSAIDs in the 5 years prior to diagnosis

 Never

172

42 (24)

 

212

62 (29)

 

 Less than weekly

118

35 (30)

 

147

54 (37)

 

 At least weekly

69

20 (29)

0.56

57

21 (37)

0.20

AJCC American Joint Committee on Cancer; BMI body mass index; EAC esophageal adenocarcinoma; GEJAC adenocarcinoma of the gastroesophageal junction; GER gastroesophageal reflux; NSAIDs non-steroidal anti-inflammatory drugs

aColumn numbers may not sum to total because of missing values

bp value for χ2 test for heterogeneity for comparing survivors to non-survivors for the distribution of each categorical variable

cEAC, 180 patients had undetermined pre-treatment AJCC tumor stage (50%); GEJAC, 218 patients had undetermined pre-treatment AJCC tumor stage (52%)

dOne standard drink is equivalent to 10 g of ethanol

eIncludes life-long non-drinkers and patients consuming <1 drink/week on average of alcohol

fFrequency of heartburn or acid reflux in 10-year period before diagnosis

Two hundred and seventy-five (76%) of the patients with EAC and 292 (69%) of the patients with GEJAC died during the follow-up period. The overall 5-year survival rates for EAC and GEJAC were 27% (95% CI 23–32) and 33% (95% CI 29–38), respectively (Table 1). EAC and GEJAC patients with early stage or well-differentiated cancers and those treated with curative intent were significantly more likely to survive. For EAC and GEJAC tumors, the crude Kaplan–Meier survival curves by AJCC tumor stage show that survival outcomes were poor among patients with late-stage disease (both log-rankp < 0.001) (Fig. 1). However, among patients with AJCC stage “undetermined”, the survival outcomes were similar to patients with AJCC stage II/III disease. In univariate analyses (Table 1), surviving patients with EAC were significantly more likely to be obese and have GER symptoms. For GEJAC, surviving patients were more likely female and aged less than 70 years.
https://static-content.springer.com/image/art%3A10.1007%2Fs10552-012-9913-1/MediaObjects/10552_2012_9913_Fig1_HTML.gif
Fig. 1

Kaplan-Meier survival curves by pre-treatment AJCC tumor stage for a patients with EAC and b patients with GEJAC

Clinico-pathologic factors and survival

In multivariate analyses, more advanced AJCC tumor stage, high tumor grade and non-curative treatment intent had clear adverse effects on survival for both tumor sites (Table 2). Age also influenced survival for GEJAC, with older patients faring worse and this remained significant after adjusting for the effects of other prognostic factors (≥70 years, HR = 1.70, 95% CI 1.24–2.32).
Table 2

Adjusted hazard ratios and 95% CI for the influence of sex, age, and clinical factors on risk of death among patients with EAC and GEJAC

Variables

EAC

GEJAC

Adjusted HRa

(95% CI)

Adjusted HRa

(95% CI)

Sex

 Men

1.00

Ref

1.00

Ref

 Women

0.97

(0.62–1.50)

0.82

(0.56–1.21)

Age in years (continuous)

1.00

(0.98–1.01)

1.03

(1.01–1.04)

Age group

 <60 years

1.00

Ref

1.00

Ref

 60–69 years

1.20

(0.89–1.63)

1.19

(0.88–1.60)

 ≥70 years

0.89

(0.63–1.25)

1.70

(1.24–2.32)

AJCC stage

 I

1.00

Ref

1.00

Ref

 II

2.68

(1.03–6.98)

2.74

(1.23–6.09)

 III

3.99

(1.55–10.3)

4.81

(2.17–10.6)

 IV

5.86

(2.26–15.2)

7.50

(3.07–18.3)

Undeterminedb

3.64

(1.46–9.06)

3.72

(1.73–7.99)

Tumor grade

 Well differentiated

1.00

Ref

1.00

Ref

 Moderately differentiated

1.64

(0.88–3.08)

1.26

(0.79–1.99)

 Poorly/undifferentiated

2.50

(1.33–4.67)

1.55

(1.01–2.38)

Treatment intent

 Curative intent

  Attempt re-section and chemoradiotherapy

1.00

Ref

1.00

Ref

  Chemoradiotherapy only

2.03

(1.28–3.24)

1.31

(0.74–2.34)

 Non-curative

2.80

(2.04–3.86)

3.34

(2.29–4.86)

Number of comorbidities

 0

1.00

Ref

1.00

Ref

 ≥1

0.79

(0.61–1.03)

0.93

(0.73–1.19)

AJCC American Joint Committee on Cancer; CI confidence interval; EAC esophageal adenocarcinoma; GEJAC adenocarcinoma of the gastroesophageal junction; HR hazard ratio

aMutually adjusted for all variables in the table

bPatients with AJCC stage undetermined were retained in multivariate models as a separate category

Lifestyle factors and survival

Table 3 presents adjusted HRs for the associations between lifestyle factors and survival. Overall, after adjusting for clinico-pathologic prognostic factors, we found no evidence that any lifestyle factors predicted survival among patients with EAC tumors. In particular, education, frequency of symptoms of GER, BMI, tobacco smoking, alcohol, and frequency of use of NSAIDs were not associated with survival in this cohort. There was some indication of a linear trend of increasing risk with increasing levels of physical activity (p-trend = 0.10) for EAC, but the point estimates were non-significant. Among patients with GEJAC tumors, current smokers had a moderately increased risk of early death, independent of known clinico-pathologic prognostic factors (HR = 1.45, 95% CI 1.02–2.06). However, there was no trend toward increased risk of early death with increasing number of pack-years smoked (p-trend = 0.93). Other lifestyle factors putatively associated with the risk of developing GEJAC were not associated with survival. When the same analyses were repeated in patients who had undergone esophageal resection for EAC or GEJAC, these results were unchanged.
Table 3

Adjusted hazard ratios and 95% CI for the influence of demographic and lifestyle factors on risk of death among patients with EAC and GEJAC

Variables

EAC

GEJAC

Adjusted HRa

(95% CI)

Adjusted HRa

(95% CI)

Highest level of education

 No tertiary qualification

1.00

Ref

1.00

Ref

 Technical college/diploma

0.97

(0.75–1.25)

0.83

(0.65–1.06)

 University degree

1.00

(0.60–1.68)

0.71

(0.47–1.08)

Frequency of GER symptomsb

 Never

1.00

Ref

1.00

Ref

 Less than weekly

0.98

(0.71–1.36)

0.82

(0.61–1.09)

 At least weekly

0.87

(0.63–1.21)

0.76

(0.57–1.03)

BMI last year (kg/m2)

 <25.0

1.00

Ref

1.00

Ref

 25.0–29.9

1.05

(0.76–1.47)

0.90

(0.67–1.20)

 30.0–34.9

0.85

(0.58–1.24)

0.94

(0.67–1.32)

 ≥35.0

1.48

(0.94–2.33)

0.64

(0.39–1.05)

Smoking statusc

 Never smoker

1.00

Ref

1.00

Ref

 Ex-smoker

0.87

(0.63–1.21)

1.27

(0.92–1.75)

 Current smoker

0.99

(0.67–1.47)

1.45

(1.02–2.06)

Cumulative smoking historyc

 Never smoker

1.00

Ref

1.00

Ref

 0–14.9 pack-years

0.79

(0.54–1.15)

1.46

(1.01–2.10)

 15.0–29.9 pack-years

0.80

(0.54–1.20)

1.13

(0.78–1.63)

 ≥30 pack-years

1.07

(0.75–1.52)

1.38

(0.99–1.94)

Average lifetime alcohol consumptiond,e (standard drinks/week)

 <1f

1.00

Ref

1.00

Ref

 1–6

0.90

(0.57–1.44)

0.77

(0.51–1.17)

 7–20

1.52

(0.98–2.37)

0.73

(0.48–1.10)

 ≥21

1.02

(0.64–1.64)

0.92

(0.59–1.44)

Physical activity index

 Low

1.00

Ref

1.00

Ref

 Moderate

1.22

(0.88–1.68)

1.08

(0.79–1.48)

 High

1.32

(0.96–1.80)

1.16

(0.84–1.59)

Frequency of use of NSAIDs in the 5 years prior to diagnosis

 Never

1.00

Ref

1.00

Ref

 Less than weekly

1.00

(0.75–1.32)

0.98

(0.75–1.27)

 At least weekly

0.90

(0.64–1.26)

0.80

(0.54–1.16)

BMI body mass index; CI confidence interval; EAC esophageal adenocarcinoma; GEJAC adenocarcinoma of the gastroesophageal junction; GER gastroesophageal reflux; HR hazard ratio; NSAIDs non-steroidal anti-inflammatory drugs

aAdjusted for age, sex, AJCC stage (I, II, III, IV, and undetermined), treatment intent and number of comorbidities

bFrequency of heartburn or acid reflux in 10-year period before diagnosis

cAdditionally adjusted for average lifetime alcohol consumption

dOne standard drink is equivalent to 10 g of ethanol

eAdditionally adjusted for smoking status

fReferent group includes life-long non-drinkers and patients consuming <1 drink/week on average of alcohol

Discussion

We investigated the association between lifestyle factors and survival among a large population-based cohort of patients with EAC and GEJAC, with comprehensive measures of known etiologic and prognostic factors relevant to these tumors. While we have confirmed that AJCC tumor stage, histologic grade, and treatment intent were strong independent prognostic factors for both EAC and GEJAC tumors, of interest was our finding that tobacco smoking and older age at diagnosis were associated with poorer survival for GEJAC, independent of the effects of other prognostic factors. Other lifestyle factors important in the etiology of these cancers were not associated with survival in this cohort.

Overall survival rates and surgical rates in our cohort were higher than national [1] and international figures [12, 16], potentially due to a “healthy cohort” effect (i.e., patients had to survive long enough to participate in the original study), and the method used to identify and recruit patients. As patients were predominantly identified through major treatment centers and were aged less than 80 years, participation was likely to be biased toward patients who were more likely to be able to undergo surgical treatment with curative intent. The apparently more favorable, although not statistically significant, survival among patients with GEJAC compared with EAC tumors may also be the result of information bias. As we have reported previously [23], because the location of a tumor can be identified more precisely from surgical resection specimens than from endoscopy, patients who have undergone surgery are more likely to have the anatomical location of their tumor classified as gastroesophageal junction than patients who have not received surgery. Nonetheless, we have no reason to believe that this would have affected the internal validity of our study since estimates were based on comparisons between patients for whom we had complete follow-up. The strong effects of AJCC tumor stage and treatment intent on EAC survival in our cohort confirm results from earlier studies [12, 14, 16]. We found similar effects for stage and treatment on GEJAC survival, but we also found that age was associated with survival for patients with GEJAC, independent of AJCC stage and treatment.

Consistent with overall results from previous population-based studies, we found little indication that lifestyle factors, including education, GER, tobacco smoking, alcohol, and NSAIDs influenced survival for EAC [12, 16]. However, in their population-based study, Sundelöf et al. [12] reported better prognosis for obese patients (BMI ≥ 30) compared to patients with a normal weight (BMI 22–24.9) 20 years before diagnosis. Similarly, a surgical series conducted in Canada reported better survival for obese patients (preoperative BMI ≥ 30) who had undergone surgical resection for EAC [17]. In contrast, surgical series in Ireland [26] and Houston, USA [18] found that obesity (preoperative BMI ≥ 30) did not independently influence survival for EAC. In univariate analyses, we found that obesity (BMI ≥ 30 1 year prior to diagnosis) was associated with better survival for patients with EAC tumors. However, in agreement with these latter studies, we found no association with survival for EAC after adjusting for key clinico-pathologic prognostic factors. Additionally, there was no statistically significant variation in the risk estimates across strata of tumor stage, and when we restricted analyses to patients who had undergone surgery, we again found no association. While it is possible that BMI 1 year prior to diagnosis may underestimate a patient’s BMI prior to onset of cancer symptoms, we found HRs of similar magnitude associated with maximum lifetime BMI. Thus, while we cannot completely rule out misclassification as an explanation for the lack of association with BMI in our data, we consider the likelihood to be low. It appears that BMI has no influence on survival for EAC, and any survival benefit associated with obesity is likely due to obese patients being diagnosed at an early stage.

To our knowledge, no previous population-based study has investigated whether etiological lifestyle factors influence survival for GEJAC. In our study, tobacco smoking was the only independent prognostic factor for GEJAC among all the lifestyle factors we considered. We found that current smokers had significantly increased risk of early death from GEJAC, independent of clinico-pathologic prognostic factors. Analogous studies have reported that smoking is an independent prognostic factor for upper digestive and oral cancers [27], and we have recently reported an adverse association with ESCC survival [15]. Assuming that the association with GEJAC cancer is real, it is possible that tobacco may indirectly influence survival for these cancers by suppressing the host defenses against cancer [28]. However, the higher mortality rate may also be the result of delayed diagnosis and treatment. In our data, we did observe that late-stage patients with GEJAC tumors were more likely to report being current smokers and, although we adjusted for AJCC stage, it is difficult to fully control for its effect in a retrospective study, therefore the association we observed may be the result of residual confounding. With the lack of dose–effect and the apparent discrepancies between cancer sites, we also cannot completely exclude the possibility that this positive association was the result of chance.

The main strength of our study was the large population-based prospective design, combining data from the original study with subsequent retrieval of clinical characteristics from medical records, and complete follow-up on mortality through the National Death Index. A limitation of this study was the number of patients for whom an accurate AJCC stage was unable to be determined (51%). This was unfortunate but not surprising given that patients were ascertained from across the population, including major treatment centers, some provincial hospitals, outpatient clinics, and private consulting rooms. Investigations were undertaken according to clinical indications and to assist with treatment planning. This meant that not all patients underwent the same level of investigations prior to treatment. Regardless, the lack of accurate staging data for some patients does not invalidate these findings because for the analyses presented here, tumor stage was simply a potential confounder of the associations between the exposure of interest (pre-morbid lifestyle factors) and the primary outcome (all-cause mortality). Although we were unable to accurately determine AJCC stage for some patients, we did have complete treatment and follow-up information for all patients, and our data show that survival for patients in the AJCC stage “undetermined” group was similar to those with stage II/III disease. Moreover, to control for potential confounding by AJCC stage, we included all patients in the main models and retained those with undetermined status as a separate category. The impact of competing risks is another potential source of bias; however, in our study, esophageal cancer was the cause of death for 84% of patients for whom we had ICD-coded cause of death information, and we adjusted for presence of comorbidities. Misclassification of exposure from self-reported data cannot be excluded, although this information was obtained before the outcome of interest (death). Therefore, differential recall between patients with short or long survival is unlikely. It is also possible that other lifestyle factors, including diet, may play a role in prognosis and survival; however, currently there are no firm data. Finally, because of the small number of non-white patients in this study, and exclusion of people over the age of 79 years, our findings are limited to younger, white populations.

In summary, we have assessed the potential impact of lifestyle factors on survival separately among patients presenting with EAC and GEJAC in one of the largest studies to date, with adjustment for key treatment and clinico-pathologic predictors and complete follow-up through the National Death Index. Overall, our findings suggest that lifestyle factors do not appear to unduly influence survival. Improvements in survival from these diseases will occur when we are able to select which will be the optimal treatment for individual patients. Given that lifestyle factors do not allow prediction of outcome from these cancers, it is clear that the focus for future research will relate to improvement in preoperative staging as well as methods that will predict patient response to therapies that will need to be obtained from the molecular characteristic of a patient’s cancer.

Acknowledgments

The authors acknowledge assistance from Dr. Shahram Sadeghi and Dr. Harish Babu for pathology abstractions, and Ms Anne Russell and Dr. Nirmala Pandeya for data cleaning and programming. This study was supported by the Cancer Council Queensland and the National Health and Medical Research Council of Australia (NHMRC) (Program no. 199600; Project 389820). APT is supported by an Australian Postgraduate Award (University of Queensland) and the Cancer Council NSW STREP grant 08-04. CMN is supported by a Research Fellowship from the NHMRC, and DCW is supported by a Future Fellowship from the Australian Research Council. The funding bodies played no role in the design or conduct of the study; in the collection, management, analysis, or interpretation of the data; or in the preparation, review, or approval of the manuscript.

Conflict of interest

The authors have no conflicts of interest to disclose.

Copyright information

© Springer Science+Business Media B.V. 2012