Cancer Causes & Control

, Volume 20, Issue 8, pp 1459–1467

The influence of deprivation and ethnicity on the incidence of esophageal cancer in England


  • Sheldon C. Cooper
    • Department of GastroenterologySandwell General Hospital
  • Rosie Day
    • West Midlands Cancer Intelligence Unit, Public Health BuildingThe University of Birmingham
  • Colin Brooks
    • West Midlands Cancer Intelligence Unit, Public Health BuildingThe University of Birmingham
  • Cheryl Livings
    • West Midlands Cancer Intelligence Unit, Public Health BuildingThe University of Birmingham
  • Catherine S. Thomson
    • West Midlands Cancer Intelligence Unit, Public Health BuildingThe University of Birmingham
    • Cancer Research UK
    • Department of GastroenterologySandwell General Hospital
Original paper

DOI: 10.1007/s10552-009-9372-5

Cite this article as:
Cooper, S.C., Day, R., Brooks, C. et al. Cancer Causes Control (2009) 20: 1459. doi:10.1007/s10552-009-9372-5


The incidence of esophageal cancer (EC), particularly esophageal adenocarcinoma (EAC), has been rising dramatically. In the USA, esophageal squamous cell carcinoma (ESCC) is associated with deprivation and black ethnicity, while EAC is more common among whites. The influence of social deprivation and ethnicity has not been studied in England. West Midlands Cancer Intelligence Unit data were used to study the incidence of ESCC and EAC, and the influence of age, sex, socioeconomic status (Townsend quintiles by postcode) and ethnicity (to individual patients from Hospital Episode Statistics). From 1977 to 2004, a total of 15,138 EC were identified. Five-year directly age standardized incidence rates per 100,000 (95% CI) for men increased from 8.6 (8.0–9.1) in 1977–1981 to 13.7 (13.1–14.3) in 2000–2004 and for women from 5.0 (4.7–5.4) to 6.3 (5.9–6.6). ESCC incidence did not alter, but EAC incidence rose rapidly in males [2.1 (1.9–2.4) to 8.5 (8.1–9.0)] and in females [0.5 (0.4–0.6) to 1.7 (1.5–1.9)]. ESCC was strongly associated with the most socially deprived quintile. EAC was not associated with differences in socioeconomic status. EAC was significantly more common in white men 7.3 (6.9–7.7) and women 1.5 (1.3–1.6) when compared with black and Asian populations. In England the incidence of EAC has rapidly risen, particularly in men over the last three decades. ESCC was strongly associated with social deprivation. EAC was more common in white populations, but no association with the socioeconomic status was found.


Esophageal cancerAdenocarcinomaSquamous cell carcinomaEpidemiologyEthnicitySocioeconomic status


Over the last 30 years, the incidence of esophageal cancer (EC) has been reported to be increasing in Europe and the USA [110]. Esophageal squamous cell carcinoma (ESCC) rates have been stable, or declining, while the incidence of esophageal adenocarcinoma (EAC) has been rising rapidly. This rapid rise has been noted to be greater than for any other cancer among men in the USA [11]. Examination of cancer registry data across Europe reveals that the highest incidence of EAC is in the United Kingdom [1].

An association between ethnic origin and the different morphologies of EC has been described in the USA, with ESCC being more common among the black population while EAC is seen most frequently among white males [2, 12]. The influence of ethnicity has not been examined in the United Kingdom.

Esophageal squamous cell carcinoma has been reported to be associated with deprivation [13]; however, reports of the influence of socioeconomic status on EAC are inconsistent, some describing an association with deprivation [14] and others with affluence [15]. A direct comparison of ESCC and EAC in Australia suggests ESCC is more clearly associated with deprivation [16]. The influence of socioeconomic status upon the incidence of EAC has yet to be examined in England, although no association was found in Scotland [17].

We have examined the incidence of EC in the West Midlands, England, between 1977 and 2004 to identify changes in the incidence of its morphological subtypes and to examine the role of sex, ethnicity, and affluence and deprivation.

Materials and methods

The West Midlands Cancer Intelligence Unit (WMCIU) covers 10% of the population of England and Wales, whose borders have remained unchanged during the time period examined. Notification of cancer in the United Kingdom is not a statutory requirement although hospitals within England are required to send details of cancer patients admitted as either in-patients or day cases to their local registry. The WMCIU also receives information from pathology laboratories, screening centers, general practitioners and private hospitals and obtains death certificates on all patients who die with cancer recorded on the death certificate. Data completeness is estimated to be well over 95% across England. Esophageal cancers diagnosed in residents of the West Midlands were included in the study, identified from their postcode. Esophageal cancer data (ICD10 code C15) from 1977 to 2004 were extracted from the WMCIU database and grouped by morphology code classified as adenocarcinoma (ICD-O-2 codes: 814*–838*, 848*–849*, 8550, 8570), squamous cell carcinoma (ICD-O-2 codes: 805*–808*) or other and unknown morphologies. This period was selected as all case notes had already been abstracted onto the computerized cancer registration database, or photocopies of them were available for further examination. To classify the morphology of the initially high proportion of esophageal cancers that were incompletely abstracted and registered as unknown or carcinoma, plus all those coded as cardia in the first 10 years, case notes were examined in addition to the small number of incompletely abstracted notes from later years to improve the data quality. Case notes were reviewed by an experienced clinical gastroenterologist, and randomly selected files were verified by a member of the WMCIU. The Siewert classification was employed to differentiate esophageal and cardia cancers [18].

Incidence rates were calculated using Government Office for the Region populations from 1980 to 2002, with 1977–1999 adopting the 1980 population, and 2003–2004 using the population data for 2002. Cancers were grouped into 5-year diagnosis periods, and directly age standardized incidence rates per 100,000 population, using the European Standard Population (EASR), were calculated for each morphological group by sex for the years 1977–2004. A Mann–Whitney U test was employed to detect if any difference in age at diagnosis was significant between men and women in the two time periods 1977–1981 and 2000–2004.

The Townsend Index, derived from the postcode, was used as a marker of socioeconomic status [19]. It is based on four variables—unemployment, overcrowding, non-car ownership and non-home ownership—and is focused on material deprivation. The Townsend Index was produced at the Enumeration District level: small geographical areas used for the output of census data, of which there were 10,864 in the West Midlands in 1991. Five socioeconomic quintiles from one (most affluent) to five (most deprived) were derived by assigning postcodes to equal percentages of the whole West Midlands population in each group. Each patient was then allocated to one of the five groups based on their postcode. EASRs were calculated for each Townsend quintile based on the 1991 census population data. These 1991 populations were then weighted to take into account the change in the size of the West Midlands population from 1977 to 2004. For the period 1999–2004, the income domain of the Index of Multiple Deprivation (IMD) 2004 [20] was used as a second measure of deprivation. The IMD provides a measure of deprivation for each Lower Super Output Area (LSOA) in England (LSOAs are small geographical areas that are fairly homogenous and consist of approximately 1,500 people; there are 3,482 LSOAs in the West Midlands). The income domain consists of several variables, such as the percentage of population in that LSOA claiming various income-based benefits and tax credits, which are then weighted and a score assigned to each LSOA. Patients were again allocated to a quintile (calculated based on the percentage of the English population in each group) and EASRs were calculated for each quintile using the 2001 census population data. This was to check whether the use of West Midlands regional derived deprivation quintiles based on the 1991 Census (Townsend) gave similar results to national quintiles based on a range of measures generally recorded in 2001 and compiled in 2004.

Ethnicity data was compiled via linkage of the data regarding individual patients and their tumors from the cancer registration database with data from the Hospital Episode Statistics (HES). At the time of linkage, the WMCIU had had access to HES data for financial years 1997/1998 through to 2004/2005. Those individuals with multiple ethnicity recordings were classified as other/mixed (including Chinese/Oriental), while the remainder were identified as white, Asian (Indian subcontinent), or black. The ethnic breakdown of the population was derived from 2001 census data [21].



A total of 4,252 patient records (1,731 originally coded as cardia and 2,641 as esophageal) were re-examined to validate coding abstraction or identify morphology (or confirm poorly differentiated carcinoma only) in unclassified cases. A total of 296 (17%) of the cardia cancers were reclassified as esophageal. Of the 138 esophageal cancer morphologies originally coded as carcinoma, 27 (20%) were reclassified as EAC, and 26 (19%) as ESCC. Of the 2,641 esophageal cancers that were either not abstracted or of nonspecific/unknown morphology, 972 were reclassified as EAC, and 946 as ESCC. No other EAC or ESCC morphologies were newly identified in the remaining records examined.


During the 28-year period 1977–2004, 15,138 EC cases were registered with the WMCIU, including 5,517 cases of EAC, and 6,314 ESCC. In 1977, 20% were EAC (n = 78), and 48% ESCC (n = 185), with 32% (n = 124) either another morphology or unclassified. By 2004 the majority of EC was EAC (53%, n = 366), with ESCC 31% (n = 213) and other morphology/unclassified 17% (n = 116). Changes in 5-year EASRs, with 95% confidence intervals (CI) for the different morphologies are shown in Table 1 and Fig. 1. The significant rise observed for all EC in males was due to the marked increase in EAC (greater than fourfold), while there was no significant change for ESCC and a fall in those with other morphology/unclassified. In women, the picture was different, with significant increases being seen for both EAC (greater than threefold) and a smaller but significant rise in ESCC. A small fall in unclassified/other morphology was also observed in women.
Table 1

Five-year EASRs (95% CI) for esophageal cancer by morphology and sex

Cancer morphology








95% CI


95% CI


95% CI


95% CI




































Fig. 1

a, b Incidence of esophageal adenocarcinoma and esophageal squamous cell carcinoma in England (West Midlands) 1977–2004—five-year EASRs (95% CI) for esophageal cancer for men (a) and women (b)

The age at diagnosis of EAC was older in women during 1977–1981 median 72.5 (IQR 63.3–78) than in men 65 (51–71.8), and also in 2000–2004 [78 (69–83) vs. 70 (61–77)] (p < 0.0001 both time periods), and has increased in both men and women during the time period examined.

Socioeconomic status

Figures 2 and 3 illustrate the effect of socioeconomic status on the incidence of EAC and ESCC, respectively, over the time-period 1977–2004 for both sexes, giving the least and most deprived quintiles. It can be seen that there was no consistent relationship between socioeconomic status and the incidence of EAC in men (Fig. 2a) or women (Fig. 2b) during the 28-year period. In contrast, Fig. 3a illustrates clearly that while ESCC was significantly more common in the most deprived [EASR (95% CI) 6.19 (5.05–7.33)] than the most affluent [1.85 (1.31–2.4)] for men in the earlier years, this association with deprivation was no longer evident in the later years of the study period, with both affluent and deprived men having rates of about four per 100,000 population. A similar pattern was found for women with ESCC [most deprived 3.67 (2.88–4.45)], most affluent 2.03 (1.52–2.54) in the early years of the study period (Fig. 3b). However, the difference by socioeconomic status was less marked and was not apparent after the 1992–1996 period.
Fig. 2

a, b Incidence of esophageal adenocarcinoma in most affluent and most deprived Townsend quintiles in men (a) and women (b) in England (West Midlands)
Fig. 3

a, b Incidence of esophageal squamous cell carcinoma in most affluent and most deprived Townsend quintiles in men (a) and women (b) in England (West Midlands)

When the analyses were repeated using the IMD 2004 for cancers diagnosed in 2000 to 2004, the results agreed with the Townsend Index for EAC, with no association between socioeconomic status and incidence of EAC. However, the incidence of ESCC was positively associated with deprivation, with the most deprived quintile having EASRs of 4.73 (3.96–5.5) in men and 4.96 (4.22–5.69) in women. In contrast, the rates of ESCC in the most affluent quintile were 2.55 (1.93–3.17) in males and 2.53 (1.97–3.1) in females, respectively.


Table 2 shows the EASRs of EAC and ESCC by ethnic group for the period 1998–2004. EAC was significantly more common among the white population than among other ethnic groups for both sexes. Indeed, no cases of EAC were observed among women in any ethnic groups other than white during the 7 years examined. In contrast, ESCC was significantly more common than EAC among black men. ESCC was more common among black men than in white men, although this difference was not statistically significant.
Table 2

Esophageal adenocarcinoma and esophageal squamous cell carcinoma EASRs and absolute numbers by ethnicity for the 7-year period 1998–2004

Ethnic group










95% CI




95% CI




95% CI




95% CI





































































Not known













Not HES matched






















Key: “Not HES matched”—patient ethnicity undetermined as not matched to Hospital Episode Statistics


The rising incidence of EC, specifically EAC, observed in the United Kingdom 10 years ago [1, 36] continues unabated during the past decade. In comparison, the incidence of ESCC has altered little in England, confirming that the rise in EAC is very unlikely to be a result of improved diagnostic or reporting methods, despite there being a fall in the rates of the unclassified cases. Re-examination of 4,252 of the records for this study further strengthens this conclusion with similar percentages reclassified as EAC and ESCC from cases previously unclassified.

The development of EAC is positively associated with the gastro-esophageal reflux [22, 23], obesity [2325] and being male [11]. Limited data suggest an increase in recent years in the endoscopic incidence of the premalignant complication of chronic gastro-esophageal reflux Barrett’s esophagus (BE) [26]. This change may in part be due to increasing use of endoscopy but also due to greater awareness of BE among endoscopists [27]. Obesity among the US population has been increasing for 50 years [28], with the incidence of obesity within the UK population doubling from 1985 to 1992 [29]. This rapid change, with obesity also associated with an increased incidence of gastro-esophageal reflux symptoms [30], may well have contributed to the rise in the incidence EAC. The rise in the incidence of EAC was greatest in men, although a significant but smaller rise was also seen among women. Male pattern obesity with increased abdominal or visceral fat is particularly associated with EAC [31]. Occupational exposure to cement dust and asbestos [32] has been associated with development of EAC. Men develop EAC on an average 8 years younger than women and cases of EAC before the menopause are very unusual. This raises the possibility that female sex hormones may protect against the development of EAC, or EAC oncogenesis may be, in part, driven by androgens.

It has also been shown that the development of EAC is negatively associated with Helicobacter pylori [33] and increasing consumption of fruit, vegetables, and fiber [34]. The incidence of H. pylori has fallen by 25% between the 1970s and 1990s [35, 36]. A significant minority of patients with H. pylori develop atrophic gastritis and are rendered achlorhydric [37, 38] and therefore not at risk of acid gastro-esophageal reflux and its complications BE [39, 40] and EAC [39]. The fall in incidence of H. pylori may therefore, have, increased the numbers of individuals in the population at risk of EAC. The consumption of fruit has changed little, while that of green vegetables has fallen in the United Kingdom during the study time-period [41]. However, it has been hypothesized that the use of nitrogenous fertilizers on vegetables may have contributed to the rise in the incidence of EAC [42, 43].

Previous studies have reported both associations between EAC and deprivation [14], and between EAC and affluence [15]. The present study used two measures of socioeconomic status, the Townsend score based on the 1991 census using regional quintiles and the IMD 2004 using national quintiles based on 2001 data, with almost identical results, showing there was no association between socioeconomic status and EAC. Thus it is unlikely that deprivation or affluence contributes to the development of EAC in England, in keeping with findings in Scotland from 1987 to 1996 [17].

In contrast, deprivation was strongly associated with ESCC as has been previously noted [13]. Unexpectedly, this association appeared to become weaker in the 1990s using the Townsend index, due to a declining incidence in the deprived in addition to an increasing incidence in the affluent. However, the IMD 2004 measure showed that the association between deprivation and ESCC remained strong at the end of the study period, with a higher incidence in the deprived group than in the affluent group. It is not clear why the two measures of socioeconomic status are discordant. The Townsend index focuses more on material deprivation rather than income and benefit allowances identified by the IMD, potentially leading to discrepancies between the scoring systems. The deprivation level of an area can change considerably, for example with the introduction of new housing and redevelopment schemes, which may have altered over the study period. It is possible that because the Townsend index utilized was based on the 1991 census, that for cases diagnosed in the late 1990s and 2000s this is a less reliable measure than the IMD. In our study, the Townsend Index was preferred, as the data used to construct this Index were taken from around the midpoint of the study period 1991. It is known that ESCC is associated with a diet poor in fruit and vegetables [44], smoking, and excess alcohol [25], all of which are more common among those from deprived backgrounds [4548].

The higher rates of EAC in both men and women with white ethnicity than in other ethnic groups confirm the findings of studies in other countries [2, 12]. Although GORD symptoms appear equally common among nonwhite subjects, endoscopic esophagitis and Barrett’s esophagus are much less common in these groups [49]. The higher incidence of H. pylori among nonwhite subjects [50] may partly explain these differences, but further research into this striking difference in incidence is clearly needed. It must be noted that this analysis is limited by the fact that 13.3% of the population (523 out of 3,909) were either unknown or not HES matched. However, as HES data are derived directly from the patient, interpretation of ethnicity by data collectors will not have been an issue. The creation of the National Cancer Intelligence Network should serve to improve the quality of ethnicity data, enabling application to the whole of the population of England.

Through examination of those cases incompletely abstracted with unknown morphology and those coded as gastric cardia, between 1977 and 1986, the validity of the data has been strengthened. The proportionally high number of other or unknown cancer morphologies, particularly early in the time period examined, may mean an under-representation of the incidence of EAC or ESCC, but there is no reason to suspect that this would be preferential for either morphology. During the time period examined, although the number of unknown/other morphologies did not change as an absolute figure (124 cancers in 1977 to 116 in 2004), there was a significant fall in the rates of cancers without a specific known morphology. Thus, there has been an improvement in either reporting trends or classification by histopathologists, but this is similar for both EAC and ESCC. The data presented here are only for patients who live within the unchanging boundaries of the WMCIU and do not include individuals referred from outside of the region. Thus, referral patterns for treatment do not explain any of the incidence patterns.

This analysis of the incidence of esophageal cancer covers one of the longest periods examined in the literature (28 years), includes a novel examination of the influence of socioeconomic status in a representative population of England, and is the first to report on the influence of ethnicity in the United Kingdom. The incidence of esophageal cancer continues to rise in England due to a marked rise in EAC, which is strongly associated with male gender and white ethnicity. Socioeconomic status does not appear to influence the incidence of EAC. The incidence of ESCC has not changed, and is the most common tumor morphology in black men and white women. ESCC is strongly associated with measures of deprivation.


We wish to acknowledge the assistance in data analysis from V. Madurasinghe, West Midlands Cancer Intelligence Unit.

Conflict of interest

No author has any conflict of interest to declare.

Copyright information

© Springer Science+Business Media B.V. 2009