This study investigated the incidence and survival of oesophageal and gastric cancers in England using data on 133,804 patients diagnosed between 1998 and 2007. The incidence of lower oesophageal cancer increased until 2002 then remained relatively stable, whereas the incidence of cancers of the cardia, non-cardia, and stomach NOS declined over this period. The incidence was higher in males compared with females for both oesophageal and gastric cancer. This was most evident in lower oesophageal and cardia cancers where the incidence was around four times higher in males. In general the incidence rates of all oesophageal and gastric cancers were higher in the more deprived areas. Overall survival was poor in all sub-groups with 1 year survival ranging from 14.8 to 40.8% and 5 year survival ranging from 3.7 to 15.6%.
Strengths and limitations
This national study included a large number of patients diagnosed with oesophageal or stomach cancer over a 10 year period. It was therefore possible to investigate differences in incidence by six cancer subgroups rather than only the traditional groups of oesophageal (C15) and stomach (C16) cancer, which obscure the unique features of the lower oesophageal and cardia tumours. It is also strengthened by available socioeconomic and survival data.
One limitation of the dataset was the relatively large proportion of patients with an unspecified anatomical subsite, particularly for gastric cancers where over half (52.6%) fell into this group. This meant that these patients could not be assigned to either the cardia or non-cardia subgroup. Defining the oesophageal cancer groups using both the anatomical site and tumour morphology led to a smaller proportion of patients in the not otherwise specified group (12.8%) compared to the groups defined and used for the sensitivity analysis based on morphology alone (17.5%). This sensitivity analysis demonstrated similar patterns in incidence and survival between the oesophageal squamous cell carcinoma group and the upper and middle oesophageal group, the oesophageal adenocarcinoma group and the lower oesophageal group, and the "other and unspecified" and the oesophageal not otherwise specified group.
Finally, another limitation was that 4,990 patients (4%) had to be excluded from the survival analysis as their registrations were based only on data from the death certificate.
Comparison to previous studies
In most developed countries the incidence of oesophageal squamous cell carcinoma, which is more commonly found in the upper and middle oesophagus , has remained constant or declined over the last 30 years  whilst the incidence of oesophageal adenocarcinoma, primarily found in the lower third of the oesophagus [7, 17] has increased, particularly in males [2–12]. In Sweden the increase in oesophageal adenocarcinoma incidence peaked in the mid 2000's and then remained stable . Our study found similar results with stable incidence rates of upper and middle oesophageal cancer over the 10 year period and an initial increase in the incidence of lower oesophageal cancer in males which has slowed and remained relatively stable after 2002.
The increase in the incidence of lower oesophageal cancer is mirrored in our study by a decrease in the incidence of cancers of the cardia. Previous studies have noted an increase in both these cancer groups [2–8], although others have found a similar stable or slightly declining trend in the incidence of gastric cardia cancer since the early 1990's [2, 12, 19]. It is possible that the trends in these two adjacent sites were influenced by changing diagnostic classification. However, after 2002 when the incidence of lower oesophageal cancer remained stable in males the incidence of cardia cancer continued to decline. If changes in diagnostic classification were responsible these trends would be expected to stabilise at a similar time. It is also possible that some of the oesophageal NOS cancers could have been cardia cancers. Reassigning those with a histological diagnosis of adenocarcinoma to the lower oesophageal group may have influenced the trends. A decline in the incidence of non-cardia gastric cancer in more developed countries has been seen in the past century [8, 20–22]. Our study confirms this continued decline for both non-cardia stomach and stomach NOS cancers.
The declining incidence of gastric cancers in this study and other studies [8, 20–22] may be associated with the decreasing prevalence in developed countries of Helicobacter pylori infection, a known risk factor for gastric cancer . However, meta-analyses have found that infection with the most common H pylori strains (CagA+) may protect against the development of oesophageal adenocarcinoma, possibly because infection causes achlorhydria and so reduces gastric acid reflux, one of the main risk factors associated with the development of oesophageal adenocarcinoma [23, 24]. A systematic review did find a lower prevalence of H pylori infection in patients with gastro-oesophageal reflux disease (GORD) . Therefore, the declining prevalence of H pylori infection could contribute to the increasing incidence of lower oesophageal cancer found here.
The increasing incidence of oesophageal adenocarcinoma found in this study may also be associated with the rising prevalence of obesity in England . Other studies have found that increasing body mass index is associated with an increased risk of oesophageal adenocarcinoma and cardia cancer [27, 28].
Consistent with previous studies [3–8, 12] lower oesophageal and cardia cancer incidence rates were much higher in males compared with females (M:F 4:1). The reasons for this is not clear, but an abdominal distribution of body fat, which is more common in males, may lead to higher levels of GORD and therefore to an increased risk of developing these cancers [27, 29]. Barrett's oesophagus, secondary to chronic GORD, is another risk factor which occurs more commonly in males  and has been linked to abdominal obesity [27, 31]. Differing patterns of past smoking behaviours in males and females could also partly explain the differing incidence of these cancers. The risk of developing squamous cell carcinoma of the oesophagus declines steadily following smoking cessation, although the risk of both oesophageal adenocarcinoma and cardia cancer does not decline until 30 years after cessation .
The finding that lower oesophageal and cardia cancer have a higher incidence in the more socioeconomically deprived groups contradicts other studies which have found no association . Squamous cell carcinoma  and gastric cancer  have been associated with deprivation in previous studies, which our findings support. The known lifestyle risk factors already discussed are likely to be more common in those living in deprived areas and so explain the higher incidence found in our study.
Implications for policy and practice
The poor prognosis of both oesophageal and gastric cancer highlights the need to concentrate efforts on primary prevention. Smoking and high alcohol consumption are risk factors for gastric cancer and squamous cell carcinoma of the oesophagus [33–35]. Smoking is also a risk factor for oesophageal adenocarcinoma, but the role of alcohol consumption is less certain . Public health initiatives aimed at reducing smoking and encouraging sensible alcohol consumption would help reduce the incidence of these cancers. A systematic review found that reducing weight may improve symptoms of GORD although not all studies have found this association . Other public health initiatives aimed at reducing obesity therefore may help to decrease the prevalence of chronic GORD which is one of the main risk factors for developing oesophageal adenocarcinoma.
The particularly high incidence of lower oesophageal and cardia tumours in males may have implications for earlier diagnosis. Current guidelines for referral and investigation of upper gastrointestinal symptoms do not specify this increased risk in males, but advise a similar threshold for males and females . Raising awareness in primary care of the differing incidence should be considered, and a lower threshold for referral in males investigated. The poor prognosis of all patients suggests that evaluation of a national programme of earlier investigation of non-specific UGI symptoms may be warranted, and new tools such as the cytosponge for identifying Barretts' epithelium may have a role to play in the future [33, 37].
Since oesophageal and gastric tumours are relatively uncommon and difficult to diagnose population-wide screening is unlikely to be cost effective. Efforts to identify high risk groups such as those with regular chronic reflux (oesophageal cancer) could perhaps be considered in developing focused screening efforts in the future, but evidence on the effectiveness of screening these groups will be needed. At present endoscopic screening is not considered feasible . However, an American study in 2010 did suggest that the incidence in White males over 60 with weekly GORD or over 55 with daily GORD was high enough to investigate the effectiveness of screening in these groups .
The poor prognosis of these cancers also suggests the need for greater focus on earlier diagnosis. Raising public awareness and knowledge of symptoms, particularly in more deprived areas and in males, will be important. A recent study found that 21% of oesophageal and 32% of stomach cancers were diagnosed as a result of emergency admissions and that emergency admissions were associated with poorer one-year survival . Therefore, greater awareness of these cancers and improved knowledge of symptoms could help to identify earlier stage tumours and consequently improve the prognosis of these cancers.
Unanswered questions and future research
These data highlight the need for further aetiological research to understand the links between sex and oesophageal and gastric cancers. Of particular importance is to fully understand why the incidence rates of lower oesophageal and cardia cancer are so much higher in males compared with females, why the incidence of these cancers are higher in the UK compared with other developed countries and why the incidence has increased over time. This study lacked information on several factors that may affect the incidence of these cancers such as smoking, alcohol, obesity, GORD and H pylori infection which, if available, could have helped in the interpretation of the analysis. Future work could record how these factors vary within populations in England. An earlier English study found significantly higher postoperative mortality following operations for oesophageal cancer in patients living in more deprived areas . Future studies could also investigate differences in survival for the six subgroups in this study taking into account other factors that influence survival such as age, socioeconomic deprivation and stage of disease.
Finally, these data also show that better classification of gastric tumours by site is needed to understand outcomes. It is important that both the cancer site and the morphology of these cancers are identified and recorded in clinical practice where possible. This information needs to be passed to the cancer registries to ensure that further studies can investigate these groups with more accuracy.