The incidence of esophageal adenocarcinoma continues to rise: analysis of period and birth cohort effects on recent trends

  • A.P. Thrift
    Correspondence
    Correspondence to: Mr A. P. Thrift, Cancer Control Laboratory, Queensland Institute of Medical Research, Locked Bag 2000 Royal Brisbane Hospital, Brisbane, QLD 4029, Australia. Tel: +61-7-33620250; Fax: +61-7-38453502;
    Affiliations
    Population Health Department, Queensland Institute of Medical Research, Brisbane, Queensland

    School of Population Health, The University of Queensland, Brisbane, Queensland, Australia
    Search for articles by this author
  • D.C. Whiteman
    Affiliations
    Population Health Department, Queensland Institute of Medical Research, Brisbane, Queensland
    Search for articles by this author

      ABSTRACT

      Background

      During the past four decades, the incidence of esophageal adenocarcinoma (EAC) has increased markedly in Western populations. Recent reports have suggested that the rate of increase has slowed or plateaued.

      Patients and methods

      Using data from cancer registries in Australia, the United States and Sweden, we examined incidence trends for esophageal and gastric cardia tumors between 1984 and 2008 using joinpoint analyses and age–period–cohort models.

      Results

      EAC incidence continues to undergo statistically significant annual increases in Australia and the United States, although the rate of increase has slowed. Among men, incidence increased annually by 2.2% [95% confidence interval (CI) 1.5% to 2.9%] between 1994 and 2008 in Australia and 1.5% (95% CI 0.2% to 2.8%) between 1998 and 2008 in the United States. EAC incidence among men remained unchanged in Sweden between 2001 and 2008 (P = 0.52). EAC incidence among women showed significant linear increases between 1984 and 2008. Age–period–cohort models suggested strong effects for both period and birth cohort on EAC incidence in Australia and the United States, and a strong period effect for Sweden.

      Conclusions

      EAC incidence continues to increase in Australia and the United States. The continued increases, even among more recent birth cohorts, suggest that EAC incidence will continue to rise during coming decades.

      Keywords

      introduction

      The epidemiology of esophageal cancer has changed over the past four decades, reflecting changes in the incidence of the main histologic subtypes. Previously, esophageal squamous cell carcinoma (ESCC) was the most common subtype in Western populations, attributed to high prevalence of smoking and heavy alcohol use [
      • Lagergren J.
      • Bergstrom R.
      • Lindgren A.
      • et al.
      The role of tobacco, snuff and alcohol use in the aetiology of cancer of the oesophagus and gastric cardia.
      ,
      • Pandeya N.
      • Williams G.
      • Green A.C.
      • et al.
      Alcohol consumption and the risks of adenocarcinoma and squamous cell carcinoma of the esophagus.
      ,
      • Pandeya N.
      • Williams G.M.
      • Sadhegi S.
      • et al.
      Associations of duration, intensity, and quantity of smoking with adenocarcinoma and squamous cell carcinoma of the esophagus.
      ]. More recently, esophageal adenocarcinoma (EAC) incidence has increased rapidly in industrialized countries and EAC is now the most common histologic subtype in these countries. In the United States, Australia and Northern Europe, countries with established population-based cancer registries, the increasing incidence trends for EAC can be traced back to the early 1970s [
      • Pohl H.
      • Welch H.G.
      The role of overdiagnosis and reclassification in the marked increase of esophageal adenocarcinoma incidence.
      ,
      • Newnham A.
      • Quinn M.J.
      • Babb P.
      • et al.
      Trends in the subsite and morphology of oesophageal and gastric cancer in England and Wales 1971–1998.
      ,
      • Bollschweiler E.
      • Wolfgarten E.
      • Gutschow C.
      • et al.
      Demographic variations in the rising incidence of esophageal adenocarcinoma in white males.
      ]. The underlying causes of these trends, and the marked male predominance, remain unclear [
      • Cook M.B.
      • Chow W.H.
      • Devesa S.S.
      Oesophageal cancer incidence in the United States by race, sex, and histologic type, 1977–2005.
      ].
      It was speculated that at least some of the increase may have been due to misclassification of cancers at adjacent sites or to changes in the delivery of health care that may have led to greater detection of these cancers, but detailed investigation of population-based data has essentially discounted these types of period effects as plausible explanations for the observed trends [
      • Pohl H.
      • Welch H.G.
      The role of overdiagnosis and reclassification in the marked increase of esophageal adenocarcinoma incidence.
      ,
      • Lagergren J.
      • Mattsson F.
      No further increase in the incidence of esophageal adenocarcinoma in Sweden.
      ].
      Recent analyses of data from the United States National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program and the Swedish Cancer Registry have suggested that the rate of increase in EAC incidence has slowed or even plateaued. The US study reported that the annual rate of increase has slowed from 8.2% between 1973 and 1996 down to 1.3% since 1996 [
      • Pohl H.
      • Sirovich B.
      • Welch H.G.
      Esophageal adenocarcinoma incidence: are we reaching the peak?.
      ]. The Swedish study reported that EAC incidence rates have been stable since 2001 [
      • Lagergren J.
      • Mattsson F.
      No further increase in the incidence of esophageal adenocarcinoma in Sweden.
      ]. No nationwide EAC incidence studies have been reported for Australia, but analyses of data from several Australian states showed an increase in EAC incidence among men between 1982 and 1993 [
      • Lord R.V.N.
      • Law M.G.
      • Ward R.L.
      • et al.
      Rising incidence of oesophageal adenocarcinoma in men in Australia.
      ]. In the most populous Australian state of New South Wales, EAC incidence increased by over 4% annually between 1988 and 2005 [
      • Stavrou E.P.
      • McElroy H.J.
      • Baker D.F.
      • et al.
      Adenocarcinoma of the oesophagus: incidence and survival rates in New South Wales, 1972–2005.
      ]. It is not known however whether the reported flattening of the incidence curves observed in other high-incidence populations is occurring in Australia.
      To investigate the various contributions of age, period and birth cohort effects on EAC incidence, we report here the results of analyses carried out simultaneously for Australia, the United States and Sweden, countries with population cancer registries and where large epidemiological studies for EAC have been undertaken [
      • Lagergren J.
      • Bergstrom R.
      • Lindgren A.
      • et al.
      Symptomatic gastroesophageal reflux as a risk factor for esophageal adenocarcinoma.
      ,
      • Whiteman D.C.
      • Sadeghi S.
      • Pandeya N.
      • et al.
      Combined effects of obesity, acid reflux and smoking on the risk of adenocarcinomas of the oesophagus.
      ,
      • Vaughan T.L.
      • Davis S.
      • Kristal A.
      • et al.
      Obesity, alcohol, and tobacco as risk factors for cancers of the esophagus and gastric cardia: adenocarcinoma versus squamous cell carcinoma.
      ,
      • Brown L.M.
      • Silverman D.T.
      • Pottern L.M.
      • et al.
      Adenocarcinoma of the esophagus and esophagogastric junction in white men in the United States: alcohol, tobacco, and socioeconomic factors.
      ,
      • Gammon M.D.
      • Schoenberg J.B.
      • Ahsan H.
      • et al.
      Tobacco, alcohol, and socioeconomic status and adenocarcinomas of the esophagus and gastric cardia.
      ]. We have used joinpoint regression analyses to identify periods of uniform incidence trends and age–period–cohort models to distinguish the effects of birth cohort from calendar period on these trends. To determine whether misclassification or detection biases may explain some of the observed trends for EAC, we also conducted analyses of ESCC and adenocarcinoma of the gastric cardia (GCA) for the same time periods.

      methods

      For Australia, the United States and Sweden, we obtained aggregate data on primary tumor site and histology, sex, age at diagnosis (5-year age groups), race (SEER only) and year of diagnosis for all persons diagnosed with esophageal or gastric cardia cancers for the 25-year period from 1984 to 2008. We identified incident cases for Australia from the Australian Institute of Health and Welfare Australian Cancer Database [
      • Australian Institute of Health and Welfare (AIHW)
      Australian Cancer Incidence and Mortality Books (ACIM).
      ]; notification of invasive cancer is a statutory requirement for all hospitals and pathology services in Australia and the data represent the entire Australian population. Incidence data for the United States were ascertained from the nine population-based cancer registries in the SEER program [

      Surveillance Epidemiology and End Results (SEER) ProgramSEER*Stat Database: Incidence—SEER 9 Regs Research Data, Nov 2010 Sub (1973–2008) <Katrina/Rita Population Adjustment>—Linked To County Attributes—Total U.S., 1969–2009 Counties, National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, released April 2011 (updated 10/28/2011), based on the November 2010 submissionwww.seer.cancer.gov

      ]. These nine registries cover ∼10% of the United States' population and the SEER data are broadly representative of cancer incidence in the United States as a whole. Incident cases for Sweden were identified from the Swedish Cancer Registry [

      Swedish Cancer Registry, The National Board of Health and Welfarewww.socialstyrelsen.se/statistics (27 January 2012, date last accessed)

      ], which covers the entire Swedish population; it is compulsory for health-care providers in Sweden to report incident cancers. Population denominators for each year between 1984 and 2008 were obtained from the Australian Bureau of Statistics, the United States Census Bureau and Statistics Sweden.
      For Australia and the United States, we used anatomic site and histologic codes of the International Classification of Diseases for Oncology, Third Edition (ICD-O-3), to define invasive cancers: esophageal (site codes, C15.0–C15.9) and gastric cardia (C16.0). We included incident data on all patients with histologically confirmed adenocarcinoma (histologic codes, M8140–8573) and squamous cell carcinoma (M8050–8082). For Sweden, we used the seventh revision of the International Classification of Diseases (ICD-7) and histology codes (pathological anatomic diagnosis, PAD) to define EAC (ICD-7 150; PAD 096), ESCC (ICD-7 150; PAD 146) and GCA (ICD-7 151.1; PAD 096).

       statistical analysis

      Age-standardized (2000 United States standard population) incidence rates were calculated according to country, sex, race, age group and year of diagnosis. We estimated annual percentage change in incidence rates by fitting a least-squares regression line to the natural logarithm of the rate, using calendar year as a regressor variable. This method uses a statistical algorithm to determine whether there are any significant changes in the magnitude or direction of trends over time. A maximum of three joinpoints were allowed, and a minimum of four observations were required between two joinpoints [
      • Yu B.
      • Barrett M.J.
      • Kim H.-J.
      • et al.
      Estimating joinpoints in continuous time scale for multiple change-point models.
      ]. Monte Carlo permutation tests were used to examine trends for each combination of joinpoints and we selected the trend line that provided the best fit to the data [
      • Kim H.J.
      • Fay M.P.
      • Feuer E.J.
      • et al.
      Permutation tests for joinpoint regression with applications to cancer rates.
      ]. We used Joinpoint Software version 3.5.2 (http://surveillance.cancer.gov/joinpoint/) and a significance level of α = 0.05.
      Age, period and birth cohort effects were estimated by fitting a log-linear Poisson regression model, assuming that the number of cases followed a Poisson distribution. We sequentially fitted five models (age, age–drift, age–period, age–cohort and age–period–cohort) [
      • Clayton D.
      • Schifflers E.
      Models for temporal variation in cancer rates. II. Age-period-cohort models.
      ] and estimated the parameters of the model using the maximum-likelihood method. To attempt to address the nonidentifiability problem (i.e. parameters for age, period and cohort are not uniquely identifiable due to their linear dependence), we constrained the regression parameters for the first and last periods (when estimating cohort effects) and the first and last cohorts (when estimating period effects) to zero [
      • Clayton D.
      • Schifflers E.
      Models for temporal variation in cancer rates. II. Age-period-cohort models.
      ]. Age–period–cohort analyses were carried out using 5-year calendar periods (1984–1988 to 2004–2008), 5-year age intervals (40–44 to 80–84 years) and a total of 13 overlapping 10-year birth cohorts (1900–1909 to 1960–1969, identified by the central year of birth from 1905 to 1965). Period and birth cohort effects were estimated as relative risks, using the 1984–1988 period and 1905 birth cohort as the respective reference categories. The goodness-of-fit was evaluated by the Pearson statistics and likelihood ratio tests, and age–period–cohort analyses were carried out using R software version 2.12.2 (http://www.r-project.org/).

      results

      Between 1984 and 2008, there were 7714 men and 1658 women diagnosed with EAC in Australia, 10 330 men and 1783 women diagnosed with EAC in the SEER 9 registries (hereafter ‘United States’), and 2393 men and 577 women diagnosed with EAC in Sweden (supplementary Table S1, available at Annals of Oncology online). The proportion of all esophageal cancers that were EAC increased from 20% to 52% in Australia, from 18% to 63% in the United States and from 19% to 50% in Sweden.
      The age-standardized incidences of EAC increased in all three countries during the study period and are plotted in Figure 1. Rates were typically twofold higher in Australia and the United States than in Sweden. The best fitting models for incidence trends and the inflection points identified by joinpoint regression are shown in Table 1. EAC incidence was approximately sixfold higher in men than in women in Australia and Sweden and eightfold higher in men than in women in the United States, and this marked male predominance remained throughout the study period (Figure 1). Among men in Australia and the United States, joinpoint regression identified one significant inflection point (1994 and 1998, respectively) and thus two linear segments (trends). Age-standardized incidence rates for EAC rose by 7.7% (CI 5.8% to 9.6%) annually between 1984 and 1994 in Australia and by 7.1% (CI 5.8% to 8.4%) annually between 1984 and 1998 in the United States. The rates of increase slowed in subsequent years; however, we found statistically significant annual increases in EAC incidence in Australia between 1994 and 2008 (2.2%; CI 1.5% to 2.9%) and the United States between 1998 and 2008 (1.5%; CI 0.2% to 2.8%). The patterns were similar in the Unites States when we restricted our analyses to Hispanic and non-Hispanic whites (Table 1). In Sweden, the incidence pattern was different. EAC incidence in men was stable between 1984 and 1993 (nonsignificant trend, P = 0.47), then increased by 12.2% (CI 7.1% to 17.4%) annually between 1993 and 2001, and thereafter remained unchanged (P = 0.52).
      Figure 1
      Figure 1Time trends of age-standardized incidence rates for esophageal adenocarcinoma by sex (A, men; B, women) in Australia (solid triangles), the United States (open squares) and Sweden (crosses), 1984–2008. Fitted dashed lines were derived from joinpoint regression.
      Table 1Annual percentage change in incidence of esophageal adenocarcinoma, esophageal squamous cell carcinoma and gastric cardia adenocarcinoma, 1984–2008
      Cancer subtype and countryPersonsMenWomen
      YearsAPC
      a APC derived from joinpoint regression. The age-standardized rates used in the joinpoint regression are standardized to the 2000 US standard population. Bold indicates that the APC is statistically significant from zero.
      (95% CI)
      YearsAPC
      a APC derived from joinpoint regression. The age-standardized rates used in the joinpoint regression are standardized to the 2000 US standard population. Bold indicates that the APC is statistically significant from zero.
      (95% CI)
      YearsAPC
      a APC derived from joinpoint regression. The age-standardized rates used in the joinpoint regression are standardized to the 2000 US standard population. Bold indicates that the APC is statistically significant from zero.
      (95% CI)
      Esophageal adenocarcinoma
       Australia1984–19947.4 (5.7–9.1)1984–19947.7 (5.8–9.6)1984–20083.0 (2.4–3.7)
      1994–20082.4 (1.8–3.0)1994–20082.2 (1.5–2.9)
       United States1984–19987.1 (5.9–8.2)1984–19987.1 (5.8–8.4)1984–20084.4 (3.5–5.3)
      1998–20081.7 (0.5–2.9)1998–20081.5 (0.2–2.8)
       United States (Whites)
      b ‘Whites’ refers to the age-standardized incidence rates for total Whites (Hispanic and non-Hispanic whites) extracted from SEER 9.
      1984–19997.2 (6.2–8.2)1984–19987.4 (6.1–8.7)1984–20084.8 (3.8–5.7)
      1999–20081.5 (0.2–2.8)1998–20081.6 (0.4–2.9)
       Sweden1984–1990-2.1 (-9.1 to 5.4)1984–19931.5 (-2.6 to 5.7)1984–20085.3 (4.3–6.4)
      1990–20029.9 (7.5–12.3)1993–200112.2 (7.1–17.4)
      2002–20080.8 (-3.3 to 5.0)2001–20081.1 (-2.3 to 4.6)
      Esophageal squamous cell carcinoma
       Australia1984–1994-0.1 (-1.3 to 1.1)1984–19910.1 (-1.6 to 1.9)1984–19950.4 (-1.1 to 1.9)
      1994–2008-2.9 (-3.5 to -2.2)1991–2008-2.5 (-2.9 to -2.1)1995–2008-3.4 (-4.5 to -2.3)
       United States1984–2008-3.2 (-3.5 to -2.9)1984–2008-3.6 (-3.9 to -3.3)1984–2008-2.7 (-3.1 to -2.2)
       United States (Whites)
      b ‘Whites’ refers to the age-standardized incidence rates for total Whites (Hispanic and non-Hispanic whites) extracted from SEER 9.
      1984–2008-3.0 (-3.3 to -2.7)1984–2008-3.4 (-3.8 to -3.1)1984–2008-2.6 (-3.1 to -2.1)
       Sweden1984–2008-2.3 (-2.7 to -1.8)1984–2008-3.0 (-3.5 to -2.4)1984–2008-1.0 (-1.9 to 0.0)
      Gastric cardia adenocarcinoma
       Australia1984–19982.1 (1.2–3.0)1984–19981.7 (0.8–2.7)1984–20032.2 (1.2–3.3)
      1998–2008-1.0 (-2.2 to 0.2)1998–2008-1.3 (-2.6 to -0.1)2003–2006-12.5 (-35.1 to 18.0)
      2006–200822.0 (-7.4 to 60.7)
       United States1984–2008-0.1 (-0.4 to 0.2)1984–2008-0.3 (-0.6 to 0.0)1984–20080.5 (-0.1 to 1.0)
       United States (Whites)
      b ‘Whites’ refers to the age-standardized incidence rates for total Whites (Hispanic and non-Hispanic whites) extracted from SEER 9.
      1984–20080.0 (-0.3 to 0.4)1984–2008-0.3 (-0.6 to 0.0)1984–20080.6 (-0.1 to 1.3)
       Sweden1984–19941.9 (0.3–3.4)1984–19971.4 (0.1–2.8)1984–2008-0.2 (-0.9 to 0.5)
      1994–2008-1.2 (-2.0 to -0.3)1997–2008-2.0 (-3.5 to -0.4)
      APC, annual percentage change; CI, confidence interval.
      a APC derived from joinpoint regression. The age-standardized rates used in the joinpoint regression are standardized to the 2000 US standard population. Bold indicates that the APC is statistically significant from zero.
      b ‘Whites’ refers to the age-standardized incidence rates for total Whites (Hispanic and non-Hispanic whites) extracted from SEER 9.
      Among women, the models had one segment, with EAC incidence increasing annually by 3.0% (CI 2.4% to 3.7%), 4.4% (CI 3.5% to 5.3%) and 5.3% (CI 4.3% to 6.4%) in Australia, the United States and Sweden, respectively, between 1984 and 2008.
      ESCC incidence showed significant linear declines in age-standardized incidence in all three populations over the study period (Table 1 and Figure 2). In contrast, the incidence of GCA increased significantly in Australia and Sweden until the mid-1990s and then appeared to decline slightly. In the United States, GCA incidence has remained largely unchanged.
      Figure 2
      Figure 2Time trends of age-standardized incidence rates for esophageal squamous cell carcinoma (A) and gastric cardia adenocarcinoma (B) in Australia (solid triangles), the United States (open squares) and Sweden (crosses), 1984–2008. Fitted dashed lines were derived from joinpoint regression.
      The results of the age–period–cohort analyses for EAC incidence are shown in supplementary Table S1 (available at Annals of Oncology online). In Australia and the United States, the full age–period–cohort models provided a significant improvement over the age–period and age–cohort models. In contrast, there was no statistically significant cohort effect in Sweden after adjusting for age and period (P = 0.68). The results of age–period–cohort analyses when carried out separately for men and women were similar to the overall results reported here (data not shown).
      Figure 3 shows the effects of age, period and cohort for each country. We calculated relative risks for EAC by period and birth cohort, with constraints on birth cohort and period in the age–period–cohort models where necessary (Table 2). These analyses suggested that the period effects in Sweden, where we observed a 228% excess risk of EAC in 2004–2008 compared with 1984–1988, were stronger than in Australia (151%) and the United States (145%). The risks of EAC increased with successive birth cohorts and, although we found no significant cohort effect for Sweden, the relative risks were similar in all three countries. We found similar relative risks and EAC incidence trends by age group, calendar period and birth cohort when we examined men and women separately (data not shown).
      Figure 3
      Figure 3Age, period and cohort effects (and corresponding 95% confidence intervals) of esophageal adenocarcinoma incidence in Australia (A), the United States (B) and Sweden (C).
      Table 2Relative risks for esophageal adenocarcinoma incidence by calendar period and birth cohort
      AustraliaUnited StatesSweden
      RR (95% CI)RR (95% CI)RR (95% CI)
      Calendar period
      a Based on the full age–period–cohort model with cohort constraints (parameters for the first and last cohorts set to zero).
       1984–19881.00 (Ref.)1.00 (Ref.)1.00 (Ref.)
       1989–19931.42 (1.29–1.56)1.48 (1.36–1.61)1.13 (0.95–1.36)
       1994–19981.91 (1.72–2.12)1.83 (1.67–2.02)1.81 (1.48–2.22)
       1999–20032.17 (1.92–2.46)2.20 (1.96–2.47)2.81 (2.21–3.57)
       2004–20082.51 (2.18–2.89)2.45 (2.14–2.80)3.28 (2.48–4.35)
      Birth cohort
      b Based on the full age–period–cohort model with period constraints (parameters for the first and last periods set to zero).
       19051.00 (Ref.)1.00 (Ref.)1.00 (Ref.)
       19101.21 (0.94–1.57)1.08 (0.83–1.40)1.00 (0.68–1.48)
       19151.41 (1.11–1.80)1.74 (1.37–2.22)1.31 (0.92–1.87)
       19201.64 (1.30–2.08)2.25 (1.78–2.84)1.71 (1.21–2.41)
       19252.22 (1.76–2.79)3.01 (2.39–3.80)2.36 (1.69–3.31)
       19302.70 (2.13–3.42)3.76 (2.96–4.76)2.99 (2.11–4.26)
       19353.06 (2.41–3.90)4.80 (3.77–6.10)3.96 (2.76–5.68)
       19403.66 (2.85–4.69)5.89 (4.61–7.52)5.42 (3.73–7.87)
       19454.78 (3.70–6.18)7.79 (6.07–10.0)7.76 (5.26–11.4)
       19507.06 (5.40–9.24)9.63 (7.43–12.5)10.8 (7.11–16.5)
       19558.78 (6.57–11.7)11.8 (9.01–15.6)17.6 (11.0–28.3)
       196011.2 (8.03–15.6)13.2 (9.77–17.9)21.2 (11.6–38.7)
       196515.9 (10.4–24.2)14.7 (9.84–22.0)35.4 (15.2–82.2)
      CI, confidence interval; RR, relative risk.
      a Based on the full age–period–cohort model with cohort constraints (parameters for the first and last cohorts set to zero).
      b Based on the full age–period–cohort model with period constraints (parameters for the first and last periods set to zero).

      discussion

      Using data from cancer registries in Australia, the United States and Sweden, we confirmed that EAC incidence has increased over the past 25 years. While the rates of increase in incidence have slowed in each country in the past decade, these have continued to be significant increases in Australia and the United States. Further, age–period–cohort modeling showed that period and birth cohort effects were responsible for the secular trends in EAC rates in Australia and the United States. In contrast, we found no significant increase in the incidence of EAC in Sweden between 2002 and 2008, and although the birth cohort curves showed increased risks with successive generations, the effect of birth cohort on EAC incidence was not statistically significant after adjusting for age and calendar period.
      In the United States, previous studies analyzing SEER registry data have reported that the rate of increase in EAC incidence between 1975 and 2001 was greater than any other solid tumor [
      • Pohl H.
      • Welch H.G.
      The role of overdiagnosis and reclassification in the marked increase of esophageal adenocarcinoma incidence.
      ]. In a more recent paper, the same authors used joinpoint regression and SEER data to examine patterns of EAC incidence over time and reported that the rate of increase had slowed to 1.3% annually from 1996 to 2006 [
      • Pohl H.
      • Sirovich B.
      • Welch H.G.
      Esophageal adenocarcinoma incidence: are we reaching the peak?.
      ]. Adding two more years of incidence data involving 1601 cases of EAC, we have shown that EAC incidence had not plateaued in the United States and was still increasing by 1.7% annually. In contrast, we and others have found that the incidence of EAC in Sweden has been stable since 2002 [
      • Lagergren J.
      • Mattsson F.
      No further increase in the incidence of esophageal adenocarcinoma in Sweden.
      ].
      Changes in the pattern of disease incidence over time can be due to a number of influences operating dynamically on the population. We used age–period–cohort models to estimate the effects of age, period and birth cohort on EAC incidence trends. In general, analyses reporting statistically significant effects for calendar period are typically interpreted as being due to changes in the environment (e.g. population exposures such as contagion and radiation) that affect all age groups equally, as well as to changes in diagnostic methods and disease classification. On the other hand, analyses identifying birth cohort effects are interpreted as being due to changes in the prevalence of exposure to causal factors which differ across successive generations [
      • Clayton D.
      • Schifflers E.
      Models for temporal variation in cancer rates. I. Age-period and age-cohort models.
      ].
      We found a strong and statistically significant effect of birth cohort on EAC incidence in the United States population, as have others [
      • Zheng T.Z.
      • Mayne S.T.
      • Holford T.R.
      • et al.
      Time trend and age-period-cohort effects on incidence of esophageal cancer in Connecticut, 1935–89.
      ]. Age-cohort models also significantly explained EAC incidence in Australia, but not in Sweden (although cohort curves for EAC diagnosed after 1990 show a strong birth cohort effect). These birth cohort effects are likely to reflect changes in exposure to lifestyle and environmental factors that commence early in life.
      Obesity, gastroesophageal reflux and, to a lesser extent, tobacco smoking are the principal factors associated with increased risks for EAC [
      • Olsen C.M.
      • Pandeya N.
      • Green A.C.
      • et al.
      Population attributable fractions of adenocarcinoma of the esophagus and gastroesophageal Junction.
      ], and there are data to suggest that they may act synergistically when present together [
      • Whiteman D.C.
      • Sadeghi S.
      • Pandeya N.
      • et al.
      Combined effects of obesity, acid reflux and smoking on the risk of adenocarcinomas of the oesophagus.
      ,
      • Lagergren J.
      • Ye W.M.
      • Bergstrom R.
      • et al.
      Utility of endoscopic screening for upper gastrointestinal adenocarcinoma.
      ]. The decline in smoking in Australia, the United States and Sweden is thought to be the primary explanation for the declining rates of ESCC. Given that smoking confers an approximate twofold increased risk of EAC [
      • Pandeya N.
      • Williams G.M.
      • Sadhegi S.
      • et al.
      Associations of duration, intensity, and quantity of smoking with adenocarcinoma and squamous cell carcinoma of the esophagus.
      ,
      • Cook M.B.
      • Kamangar F.
      • Whiteman D.C.
      • et al.
      Cigarette smoking and adenocarcinomas of the esophagus and esophagogastric junction: a pooled analysis from the International BEACON Consortium.
      ], it is anticipated that the decline in smoking may also have removed one of the drivers of EAC in the population. Since the late 1970s, prevalence of obesity has increased in Australia [
      • Walls H.L.
      • Wolfe R.
      • Haby M.M.
      • et al.
      Trends in BMI of urban Australian adults, 1980–2000.
      ,
      • National Health Survey
      ], the United States [
      • Kuczmarski R.J.
      • Flegal K.M.
      • Campbell S.M.
      • et al.
      Increasing prevalence of overweight among US adults. The National Health and Nutrition Examination Surveys, 1960 to 1991.
      ,
      • Flegal K.M.
      • Carroll M.D.
      • Kuczmarski R.J.
      • et al.
      Overweight and obesity in the United States: prevalence and trends, 1960–1994.
      ,
      • Flegal K.M.
      • Carroll M.D.
      • Ogden C.L.
      • et al.
      Prevalence and trends in obesity among US adults, 1999–2008.
      ] and Sweden [
      • Lissner L.
      • Johansson S.E.
      • Qvist J.
      • et al.
      Social mapping of the obesity epidemic in Sweden.
      ,
      • Sundquist K.
      • Qvist J.
      • Johansson S.E.
      • et al.
      Increasing trends of obesity in Sweden between 1996/97 and 2000/01.
      ,
      • Neovius M.
      • Teixeira-Pinto A.
      • Rasmussen F.
      Shift in the composition of obesity in young adult men in Sweden over a third of a century.
      ], with successive generations in the United States becoming heavier at younger ages [
      • Ogden C.L.
      • Carroll M.D.
      • Curtin L.R.
      • et al.
      Prevalence of overweight and obesity in the United States, 1999–2004.
      ]. Given the presumed long latency between exposure to causal factors and onset of EAC, it is reasonable to speculate that one mechanism through which the effects of birth cohort might be mediated is the increasing population prevalence of obesity. Trends in obesity have increased at a similar rate in both sexes. While data are limited, it appears that obesity is a more potent risk factor for EAC in men than women [
      • Whiteman D.C.
      • Sadeghi S.
      • Pandeya N.
      • et al.
      Combined effects of obesity, acid reflux and smoking on the risk of adenocarcinomas of the oesophagus.
      ,
      • Kubo A.
      • Corley D.A.
      Body mass index and adenocarcinomas of the esophagus or gastric cardia: a systematic review and meta-analysis.
      ] and that this is thought to be related to visceral (abdominal) adiposity rather than subcutaneous adiposity. These two distributions are quite different in men and women, and this may explain the lower incidence trends seen among women. A recent study has argued against a dominant explanatory role for obesity, citing a lack of temporality [
      • Abrams J.A.
      • Sharaiha R.Z.
      • Gonsalves L.
      • et al.
      Dating the rise of esophageal adenocarcinoma: analysis of Connecticut Tumor Registry Data, 1940–2007.
      ]. However, epidemiological data suggest that the risks of EAC rise continuously, and perhaps nonlinearly, with increasing body mass, even among people with body mass close to ‘healthy’ [
      • Abnet C.C.
      • Freedman N.D.
      • Hollenbeck A.R.
      • et al.
      A prospective study of BMI and risk of oesophageal and gastric adenocarcinoma.
      ]. If so, then at the population level, it is not the change in the proportion of people classified as ‘obese’ that should be considered as the driver of EAC incidence, but rather the incremental increases in body fat across the population with successive generations. The increasing prevalence of reflux [
      • Ness-Jensen E.
      • Lindam A.
      • Lagergren J.
      • et al.
      Changes in prevalence, incidence and spontaneous loss of gastro-oesophageal reflux symptoms: a prospective population-based cohort study, the HUNT study.
      ], independent of the rise in obesity rates, may explain part of the continued rise in EAC incidence.
      Another factor associated with risk of EAC, and for which marked cohort effects have been observed [
      • Banatvala N.
      • Mayo K.
      • Megraud F.
      • et al.
      The cohort effect and Helicobacter pylori.
      ], is infection with Helicobacter pylori. There are consistent data to show that infection with this organism confers markedly reduced risks of EAC and its precursor [
      • Whiteman D.C.
      • Parmar P.
      • Fahey P.
      • et al.
      Association of Helicobacter pylori infection with reduced risk for esophageal cancer is independent of environmental and genetic modifiers.
      ,
      • Thrift A.P.
      • Pandeya N.
      • Smith K.J.
      • et al.
      Helicobacter pylori infection and the risks of Barrett's oesophagus: a population-based case-control study.
      ]. The rates of H. pylori infection appear to be declining with successive birth cohorts in Western populations [
      • Grad Y.H.
      • Lipsitch M.
      • Aiello A.E.
      Secular trends in Helicobacter pylori seroprevalence in adults in the United States: evidence for sustained race/ethnic disparities.
      ]; thus, more recent birth cohorts have experienced lower infection rates and would be expected to have higher rates of EAC assuming a causal association.
      In addition to birth cohort effects, we found strong and statistically significant period effects in all three countries; indeed, this was the dominant feature in the Swedish data. Like others [
      • Pohl H.
      • Welch H.G.
      The role of overdiagnosis and reclassification in the marked increase of esophageal adenocarcinoma incidence.
      ], we found no evidence to suggest that the period effects were due to anatomic reclassification of gastric cardia cancers as EAC. While rates of GCA decreased in Sweden at the same time that EAC rates increased, the extra 1023 EAC diagnoses over 10 years were 10-fold higher than the concurrent decrease in GCA diagnoses; thus, reclassification does not explain the period effect. No discussion of period effects for EAC can overlook the profound changes in the availability of pharmacological agents to suppress gastric acid production, starting with H2-receptor antagonists in the 1980s, and the proton-pump inhibitors in the 1990s and 2000s [

      Medicare Benefits Schedule (MBS) Item Statistics Reportswww.medicareaustralia.gov.au/statistics/mbs_item.shtml (20 March 2012, date last accessed)

      ,
      • Guda N.M.
      • Vakil N.
      Proton pump inhibitors and the time trends for esophageal dilation.
      ,
      • Hermansson M.
      • Ekedahl A.
      • Ranstam J.
      • et al.
      Decreasing incidence of peptic ulcer complications after the introduction of the proton pump inhibitors, a study of the Swedish population from 1974–2002.
      ]. Some have speculated that long-term use of these drugs may alter the stomach chemistry and promote EAC by allowing bile reflux, which produces fewer symptoms than acid reflux, into the lower esophagus [
      • Nason K.S.
      • Wichienkuer P.P.
      • Awais O.
      • et al.
      Gastroesophageal reflux disease symptom severity, proton pump inhibitor use, and esophageal carcinogenesis.
      ]. If such causal pathways exist, then it remains possible that the widespread uptake of these drugs in recent decades may at least partly explain the observed period effect. While proton-pump inhibitor use was negligible in the mid-1990s and cannot explain the marked increase in EAC incidence since the 1970s, we cannot exclude its potential role in more recent increases.
      Strengths of our study include the complete enumeration of all persons diagnosed with invasive esophageal and gastric cardia cancers between 1984 and 2008 in Australia, the United States SEER 9 registries and Sweden. We used common diagnoses and histology codes, adjacent tumor sites to explore misdiagnoses and uniform analytical methods to examine secular trends. Furthermore, as the registries collected data prospectively and independently of our study hypotheses, our results cannot be influenced by systematic recall or information bias. The main limitation of our study is that it was based on cancer registration data and that no information on individual risk factors was available. As such, we can only speculate as to the drivers of the period and cohort effects we observed for the EAC incidence trends. Although we suggest that generational changes in body fat and H. pylori infection are the most likely explanations for the observed effects of birth cohort, it is clearly possible that other factors were responsible. Finally, as the trends for the most recent birth cohorts were based on few cases, the relative risk estimates for these birth cohorts should be interpreted cautiously.
      In summary, EAC incidence continues to rise among men and women, and while the rate of increase appears to have slowed recently, our observation that age-specific rates continue to increase in successive birth cohorts would suggest that the ‘epidemic’ may be expected to continue for some time yet.

      funding

      A.P.T. was supported by an Australian Postgraduate Award (University of Queensland) and the Cancer Council NSW STREP grant 08-04 . D.C.W. was supported by a Future Fellowship from the Australian Research Council ( FT0990987 ).

      disclosure

      The authors have declared no conflicts of interest.

      References

        • Lagergren J.
        • Bergstrom R.
        • Lindgren A.
        • et al.
        The role of tobacco, snuff and alcohol use in the aetiology of cancer of the oesophagus and gastric cardia.
        Int J Cancer. 2000; 85: 340-346
        • Pandeya N.
        • Williams G.
        • Green A.C.
        • et al.
        Alcohol consumption and the risks of adenocarcinoma and squamous cell carcinoma of the esophagus.
        Gastroenterology. 2009; 136: 1215-1224
        • Pandeya N.
        • Williams G.M.
        • Sadhegi S.
        • et al.
        Associations of duration, intensity, and quantity of smoking with adenocarcinoma and squamous cell carcinoma of the esophagus.
        Am J Epidemiol. 2008; 168: 105-114
        • Pohl H.
        • Welch H.G.
        The role of overdiagnosis and reclassification in the marked increase of esophageal adenocarcinoma incidence.
        J Natl Cancer Inst. 2005; 97: 142-146
        • Newnham A.
        • Quinn M.J.
        • Babb P.
        • et al.
        Trends in the subsite and morphology of oesophageal and gastric cancer in England and Wales 1971–1998.
        Aliment Pharmacol Ther. 2003; 17: 665-676
        • Bollschweiler E.
        • Wolfgarten E.
        • Gutschow C.
        • et al.
        Demographic variations in the rising incidence of esophageal adenocarcinoma in white males.
        Cancer. 2001; 92: 549-555
        • Cook M.B.
        • Chow W.H.
        • Devesa S.S.
        Oesophageal cancer incidence in the United States by race, sex, and histologic type, 1977–2005.
        Br J Cancer. 2009; 101: 855-859
        • Lagergren J.
        • Mattsson F.
        No further increase in the incidence of esophageal adenocarcinoma in Sweden.
        Int J Cancer. 2011; 129: 513-516
        • Pohl H.
        • Sirovich B.
        • Welch H.G.
        Esophageal adenocarcinoma incidence: are we reaching the peak?.
        Cancer Epidemiol Biomarkers Prev. 2010; 19: 1468-1470
        • Lord R.V.N.
        • Law M.G.
        • Ward R.L.
        • et al.
        Rising incidence of oesophageal adenocarcinoma in men in Australia.
        J Gastroenterol Hepatol. 1998; 13: 356-362
        • Stavrou E.P.
        • McElroy H.J.
        • Baker D.F.
        • et al.
        Adenocarcinoma of the oesophagus: incidence and survival rates in New South Wales, 1972–2005.
        Med J Aust. 2009; 191: 310-314
        • Lagergren J.
        • Bergstrom R.
        • Lindgren A.
        • et al.
        Symptomatic gastroesophageal reflux as a risk factor for esophageal adenocarcinoma.
        N Engl J Med. 1999; 340: 825-831
        • Whiteman D.C.
        • Sadeghi S.
        • Pandeya N.
        • et al.
        Combined effects of obesity, acid reflux and smoking on the risk of adenocarcinomas of the oesophagus.
        Gut. 2008; 57: 173-180
        • Vaughan T.L.
        • Davis S.
        • Kristal A.
        • et al.
        Obesity, alcohol, and tobacco as risk factors for cancers of the esophagus and gastric cardia: adenocarcinoma versus squamous cell carcinoma.
        Cancer Epidemiol Biomarkers Prev. 1995; 4: 85-92
        • Brown L.M.
        • Silverman D.T.
        • Pottern L.M.
        • et al.
        Adenocarcinoma of the esophagus and esophagogastric junction in white men in the United States: alcohol, tobacco, and socioeconomic factors.
        Cancer Causes Control. 1994; 5: 333-340
        • Gammon M.D.
        • Schoenberg J.B.
        • Ahsan H.
        • et al.
        Tobacco, alcohol, and socioeconomic status and adenocarcinomas of the esophagus and gastric cardia.
        J Natl Cancer Inst. 1997; 89: 1277-1284
        • Australian Institute of Health and Welfare (AIHW)
        Australian Cancer Incidence and Mortality Books (ACIM).
        AIHW, Canberra, Australia2012
      1. Surveillance Epidemiology and End Results (SEER) ProgramSEER*Stat Database: Incidence—SEER 9 Regs Research Data, Nov 2010 Sub (1973–2008) <Katrina/Rita Population Adjustment>—Linked To County Attributes—Total U.S., 1969–2009 Counties, National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, released April 2011 (updated 10/28/2011), based on the November 2010 submissionwww.seer.cancer.gov

      2. Swedish Cancer Registry, The National Board of Health and Welfarewww.socialstyrelsen.se/statistics (27 January 2012, date last accessed)

        • Yu B.
        • Barrett M.J.
        • Kim H.-J.
        • et al.
        Estimating joinpoints in continuous time scale for multiple change-point models.
        Comput Stat Data Anal. 2007; 51: 2420-2427
        • Kim H.J.
        • Fay M.P.
        • Feuer E.J.
        • et al.
        Permutation tests for joinpoint regression with applications to cancer rates.
        Stat Med. 2000; 19: 335-351
        • Clayton D.
        • Schifflers E.
        Models for temporal variation in cancer rates. II. Age-period-cohort models.
        Stat Med. 1987; 6: 469-481
        • Clayton D.
        • Schifflers E.
        Models for temporal variation in cancer rates. I. Age-period and age-cohort models.
        Stat Med. 1987; 6: 449-467
        • Zheng T.Z.
        • Mayne S.T.
        • Holford T.R.
        • et al.
        Time trend and age-period-cohort effects on incidence of esophageal cancer in Connecticut, 1935–89.
        Cancer Causes Control. 1992; 3: 481-492
        • Olsen C.M.
        • Pandeya N.
        • Green A.C.
        • et al.
        Population attributable fractions of adenocarcinoma of the esophagus and gastroesophageal Junction.
        Am J Epidemiol. 2011; 174: 582-590
        • Lagergren J.
        • Ye W.M.
        • Bergstrom R.
        • et al.
        Utility of endoscopic screening for upper gastrointestinal adenocarcinoma.
        J Am Med Assoc. 2000; 284: 961-962
        • Cook M.B.
        • Kamangar F.
        • Whiteman D.C.
        • et al.
        Cigarette smoking and adenocarcinomas of the esophagus and esophagogastric junction: a pooled analysis from the International BEACON Consortium.
        J Natl Cancer Inst. 2010; 102: 1344-1353
        • Walls H.L.
        • Wolfe R.
        • Haby M.M.
        • et al.
        Trends in BMI of urban Australian adults, 1980–2000.
        Public Health Nutr. 2010; 13: 631-638
        • National Health Survey
        Summary of Results, 2007–2008. Australian Bureau of Statistics, Canberra, Australia2009 (Cat. no. 4364.0)
        • Kuczmarski R.J.
        • Flegal K.M.
        • Campbell S.M.
        • et al.
        Increasing prevalence of overweight among US adults. The National Health and Nutrition Examination Surveys, 1960 to 1991.
        J Am Med Assoc. 1994; 272: 205-211
        • Flegal K.M.
        • Carroll M.D.
        • Kuczmarski R.J.
        • et al.
        Overweight and obesity in the United States: prevalence and trends, 1960–1994.
        Int J Obes. 1998; 22: 39-47
        • Flegal K.M.
        • Carroll M.D.
        • Ogden C.L.
        • et al.
        Prevalence and trends in obesity among US adults, 1999–2008.
        J Am Med Assoc. 2010; 303: 235-241
        • Lissner L.
        • Johansson S.E.
        • Qvist J.
        • et al.
        Social mapping of the obesity epidemic in Sweden.
        Int J Obes. 2000; 24: 801-805
        • Sundquist K.
        • Qvist J.
        • Johansson S.E.
        • et al.
        Increasing trends of obesity in Sweden between 1996/97 and 2000/01.
        Int J Obes. 2004; 28: 254-261
        • Neovius M.
        • Teixeira-Pinto A.
        • Rasmussen F.
        Shift in the composition of obesity in young adult men in Sweden over a third of a century.
        Int J Obes. 2008; 32: 832-836
        • Ogden C.L.
        • Carroll M.D.
        • Curtin L.R.
        • et al.
        Prevalence of overweight and obesity in the United States, 1999–2004.
        J Am Med Assoc. 2006; 295: 1549-1555
        • Kubo A.
        • Corley D.A.
        Body mass index and adenocarcinomas of the esophagus or gastric cardia: a systematic review and meta-analysis.
        Cancer Epidemiol Biomarkers Prev. 2006; 15: 872-878
        • Abrams J.A.
        • Sharaiha R.Z.
        • Gonsalves L.
        • et al.
        Dating the rise of esophageal adenocarcinoma: analysis of Connecticut Tumor Registry Data, 1940–2007.
        Cancer Epidemiol Biomarkers Prev. 2011; 20: 183-186
        • Abnet C.C.
        • Freedman N.D.
        • Hollenbeck A.R.
        • et al.
        A prospective study of BMI and risk of oesophageal and gastric adenocarcinoma.
        Eur J Cancer. 2008; 44: 465-471
        • Ness-Jensen E.
        • Lindam A.
        • Lagergren J.
        • et al.
        Changes in prevalence, incidence and spontaneous loss of gastro-oesophageal reflux symptoms: a prospective population-based cohort study, the HUNT study.
        Gut. 2011;
        • Banatvala N.
        • Mayo K.
        • Megraud F.
        • et al.
        The cohort effect and Helicobacter pylori.
        J Infect Dis. 1993; 168: 219-221
        • Whiteman D.C.
        • Parmar P.
        • Fahey P.
        • et al.
        Association of Helicobacter pylori infection with reduced risk for esophageal cancer is independent of environmental and genetic modifiers.
        Gastroenterology. 2010; 139: 73-83
        • Thrift A.P.
        • Pandeya N.
        • Smith K.J.
        • et al.
        Helicobacter pylori infection and the risks of Barrett's oesophagus: a population-based case-control study.
        Int J Cancer. 2012; 130: 2407-2416
        • Grad Y.H.
        • Lipsitch M.
        • Aiello A.E.
        Secular trends in Helicobacter pylori seroprevalence in adults in the United States: evidence for sustained race/ethnic disparities.
        Am J Epidemiol. 2012; 175: 54-59
      3. Medicare Benefits Schedule (MBS) Item Statistics Reportswww.medicareaustralia.gov.au/statistics/mbs_item.shtml (20 March 2012, date last accessed)

        • Guda N.M.
        • Vakil N.
        Proton pump inhibitors and the time trends for esophageal dilation.
        Am J Gastroenterol. 2004; 99: 797-800
        • Hermansson M.
        • Ekedahl A.
        • Ranstam J.
        • et al.
        Decreasing incidence of peptic ulcer complications after the introduction of the proton pump inhibitors, a study of the Swedish population from 1974–2002.
        BMC Gastroenterol. 2009; 9: 25
        • Nason K.S.
        • Wichienkuer P.P.
        • Awais O.
        • et al.
        Gastroesophageal reflux disease symptom severity, proton pump inhibitor use, and esophageal carcinogenesis.
        Arch Surg. 2011; 146: 851-858