Literature searches, evidence synthesis and grading of evidence

A literature search was undertaken by each subgroup with predefined search terms, electronic databases used (PubMed, Embase, Cochrane and Medline) and the time period covered. Where appropriate, conference abstracts were used to help formulate recommendations, provided that these were of sufficient scientific rigour.

The PICOs were used to guide the search for evidence and the highest quality studies were prioritised. Systematic reviews or meta-analyses were selected initially. Where systematic reviews, meta-analyses or critically appraised articles were unavailable, lower levels of evidence were selected. Where reviews use poor-quality studies such as that by Spence et al,13 for example, where the review was based on small, often single-centre cohort studies resulting in high levels of heterogenicity, this is discussed in the evidence review and highlighted in the evidence tables. In this instance, more weight is given to larger studies using large databases, which are likely to be more accurate. This hierarchy of searching was performed in a pyramidal sequence from top (high quality) to bottom (low quality), cascading from recently published systematic reviews of randomised controlled trials (RCTs) or observational studies, to RCTs, and subsequently to observational studies providing that no evidence is available from the higher quality categories. The selected evidence was tabulated in an evidence table categorising each study into the study design, intervention/clinical question, participants/population, reference standard, results and conclusions/comments.

Although randomised controlled studies are deemed the most appropriate type of study to assess the effectiveness of an intervention, other types of studies were included to assess types of effectiveness, such as ways of delivering service or outcomes from registries in clinical practice. If evidence from studies was weak or contradictory, searches for alternative sources were undertaken to see if the evidence concurred or contradicted (triangulation of searches). All available negative studies were included in the searches. The subgroups discussed the external validity of the studies and whether the study outcomes are applicable to the target population for the guidelines. Literature searches were transparent and reproducible to reduce ‘dissemination biases’.

All evidence tables and references were downloaded to a shared reference manager (F1000Workspace), to which all GDG members had access. Shortly before the statements were finalised, a further literature search was undertaken. The quality of the evidence was assessed following the GRADE terminology using table 2.14 Evidence is graded as high, moderate, low and very low quality (table 2).

Recommendations

The body of evidence for many questions was of low quality. Thus many statements were based on the consensus of the GDG given the limitations of the evidence.

Recommendations were made by each subgroup pertaining to the PICOs and graded as above with the strength of each recommendation. Recommendation strength is based on four factors:

1. Balance between desirable and undesirable effects (not considering cost).

2. Quality of the evidence across critical/important outcomes.

3. Patients’ values and preferences.

4. Costs (resource use).

There are two grades of recommendations: strong, where the benefits clearly outweigh the risks and burden (‘We recommend’); and weak, where the benefits were closely balanced with risks/burden (‘We suggest’).

Once consensus was reached within the subgroup, the level of agreement for the recommendations was obtained from the whole GDG by anonymised voting. Level of agreement was subdivided into five categories: strongly agree, agree, undecided, disagree and strongly disagree. All comments were used to amend recommendations where appropriate, and a second round of voting was undertaken for any modifications. All results and comments were anonymously sent to the GDG Chairman (MB). Where the GDG level of agreement was consistently below 90% for strongly agree and agree after three rounds of voting, the recommendation was excluded.

Pathogenesis, diagnosis and epidemiology of premalignant and early malignant gastric lesions

GA, GIM and dysplasia

For the purposes of this guideline, CAG collectively includes GA and GIM. It is important to define the histopathology of the premalignant stomach in order to understand progression to cancer and its endoscopic appearances. The normal gastric mucosa is divided into two compartments, and includes the gastrin and mucus-secreting glands of the antrum and the acid and pepsinogen-secreting oxyntic glands of the corpus. GA is defined as the loss of pre-existent glands native to the gastric compartment. The two phenotypic features of GA include the loss of glandular mass with fibrosis of the lamina propria and replacement of the native gastric glands by metaplastic or pseudopyloric glands.

The Correa cascade describes the stepwise progression of precursor lesions towards intestinal-type gastric cancer.9 H. pylori infection initiates the cascade through non-atrophic chronic gastritis, GA, GIM and finally, dysplasia.15

Gastric atrophy

GA is diagnosed histopathologically by two specific features: the presence of chronic inflammatory cells, including lymphocytes and plasma cells that expand the lamina propria, and the loss of the pre-existent gastric glands.

There is inconsistency in the histopathological diagnosis and severity of GA between pathologists, resulting in low interobserver agreement by histopathologists when staging GA using the OLGA (Operative Link on Gastritis Assessment) staging system, based on biopsies taken using the Sydney protocol.16 17 Recent studies, however, have demonstrated that accurate endoscopic staging of the severity of GA are strongly linked to gastric adenocarcinoma risk and that the interobserver and intraobserver agreement of endoscopic severity assessment, in experienced hands, is moderate to excellent.16 18

The prevalence of CAG (including GA) worldwide correlates with the prevalence of H. pylori-associated gastritis, increases with age and tends to be slightly more common in men. Prevalence is typically determined using gastroscopy and serum pepsinogen. In Western populations, the prevalence varies from 0% to 8.3%, depending on age.4 19–21 Studies performed in high-incidence areas such as Japan and China showed a prevalence of CAG of between 33% and 84%.20 22 23

Studies exploring the risk of progressing from CAG to gastric adenocarcinoma report a range of between 0% and 10%, with an annual incidence (person-year) of <1% (range 0–1.2%). This is regardless of whether the study population is from a high-risk or low-risk area.13 24 This is roughly comparable with other premalignant conditions of the digestive tract, such as Barrett’s oesophagus and colonic adenomatous polyps, where there are established guidelines on surveillance. A Swedish observational, population-based cohort study, reviewing biopsy samples of 405 172 patients from 1979 to 2011, demonstrated that 1 in 50 patients with GA would develop gastric adenocarcinoma within 20 years (an annual risk of progression of 0.1%).15 A second Dutch study exploring the follow-up data on 22 365 patients diagnosed with CAG found the overall annual incidence for the development of gastric cancer in patients with GA was 0.1%.25 This increased to 0.25% for GIM, 0.6% for LGD and 6% for HGD within 5 years after diagnosis. A recent systematic review found that the annual incidence in most studies varied from 0.1% to 0.5%, but a pooled analysis was not undertaken as there was significant heterogeneity between studies (I² statistic of 94%).13 This is partly explained by the poor quality of evidence in this systematic review derived from small cohorts of patients from single centres.

Using endoscopic grading, Japanese investigators found the cumulative 5-year incidence of gastric adenocarcinoma to be 0.7% in those with no or mild GA on endoscopic assessment, 1.9% with GA and 10% in severe endoscopic GA.26

Gastric intestinal metaplasia

GIM is a common finding in studies of patients undergoing diagnostic upper GI endoscopy—in particular, in those with a current or past H. pylori infection. GIM prevalence also increases with H. pylori infection, patient age, smokers and also with a first-degree relative with gastric cancer. The overall prevalence of GIM in those undergoing routine endoscopy varies from 13.8% to 19% in Europe.27 28 It is important to point out that the European population is not uniformly representative of that in the UK. For example the population studied by Olmez et al is from a high prevalence area of Eastern Turkey. A Dutch study found GIM to be present in 25.3% of patients undergoing endoscopy for dyspepsia.29 The prevalence in those infected with H. pylori was 33.9% compared with 15.2% of those who were not infected. This study also noted that GIM was present in 55% of patients with gastric ulcer and 100% of patients with intestinal-type gastric adenocarcinoma. A multicentre European study found the prevalence of GIM to be 31.4% in patients infected with H. pylori.30 In high-incidence areas such as Japan and China, the prevalence of GIM in H. pylori-infected individuals was 37% and 29.3%, respectively.31 GIM was found in only 2% of those not infected with H. pylori.15 32

The risk of gastric adenocarcinoma is increased in those found to have GIM. One in 39 patients with GIM develops gastric adenocarcinoma within 20 years,15 with similar rates found by De Vries et al with an annual incidence of cancer of 0.25% at 5 years.25 A meta-analysis carried out by Zullo et al in 2000 found that the risk of gastric adenocarcinoma progression in those with GIM ranged from 0% to 10%, with this range thought to be due to differing study sample sizes and follow-up periods.32 Similarly, a systematic review by Spence et al found the annual incidence in most studies ranged from 0.15% to 0.4%.13

The risk of gastric adenocarcinoma varies with the type and extent of GIM. There are three histological types of GIM, with type I or ‘intestinal’ being termed ‘complete IM’ and types II and III or ‘colonic’ termed ‘incomplete IM’. ‘Incomplete IM’ has been suggested to carry an increased cancer risk compared with ‘complete IM’.32–34 A Portuguese study showed that 31% and 6.9% of those with ‘incomplete IM’ developed LGD and HGD, respectively, compared with only 8% of those with ‘complete IM’, who developed LGD only.35 Therefore, histological subtyping may have a role in establishing gastric cancer risk, although it should be noted that only a minority of patients with invasive gastric cancer seem to have incomplete IM. Additionally, it should be noted that the traditional diagnosis of ‘complete IM’ or ‘incomplete IM’ is made using enzyme-histochemical staining methods that are highly dependent on the person evaluating them and thus are not reproducible. Therefore, the GDG did not consider histological subtyping for the guidelines.

The extent of the distribution of GIM appears to be of key importance. Four patterns of GIM distribution have been described.36 The first, ‘focal’ GIM, consists of scattered foci, mostly in the lesser curvature and incisura. The second, ‘antrum-predominant’ GIM, involves most of the antrum and incisura angularis. These two patterns with less extensive involvement of the gastric mucosa consist almost exclusively of complete type IM. The third, ‘magenstrasse’ GIM spreads throughout the lesser curvature from the cardia to the pylorus, also involving the greater curvature of the prepyloric antrum. The fourth, ‘diffuse’ GIM involves the entire gastric mucosa, with the exception of the fundic areas. These more extensive types had a greater predominance of ‘incomplete’ GIM.

Several studies have demonstrated that more extensive GIM correlates with increased gastric adenocarcinoma risk.32 36–38 Of note, a Columbian study found that in comparison with focal or antral-predominant GIM, those with magenstrasse GIM had a 5.7-fold increased risk of gastric adenocarcinoma, while those with a diffuse pattern (antrum and gastric body) had a 12.2-fold increased risk. Alongside this, an Italian study showed that a >20% extension of GIM identified those at increased risk.32 A Japanese study found the cumulative 5-year incidence of gastric adenocarcinoma to be 1.5% in those without GIM, compared with 5.3% in those with GIM in the antrum only and 9.8% in those with GIM in the antrum and corpus.26

Unlike the OLGA staging system, the OLGIM (Operative Link on Gastric Intestinal Metaplasia) staging system, using the presence of GIM, has proved to be a reproducible marker of risk with high interobserver agreement and strong association with OLGA stage.39 40 The GDG did not agree that OLGIM should be routinely used in clinical practice, although it has practical applicability in research.

Intestinal metaplasia of the gastric cardia has been reported to vary from 5% to 25% in those having endoscopy41 42 and may confer an increased risk of dysplasia and cancer, although the incidence is not clear. Sharma and colleagues found 1 of 76 patients with prevalent LGD in cardia intestinal metaplasia, defined as that just below the gastro-oesophageal junction, compared with 20 of 177 having low-grade and high-grade dysplasia with short segment Barrett’s oesophagus (<3 cm).43

Gastric dysplasia

The endoscopic prevalence of gastric dysplasia varies from 0.5% to 3.7% in Western countries and from 9% to 20% in areas with a high incidence of gastric adenocarcinoma.44–47 The identification of gastric dysplasia should also alert the endoscopist to the possibility of synchronous gastric cancer. Studies have demonstrated the incidence of synchronous gastric adenocarcinoma in those with gastric dysplasia to be up to up to 30%.44

A review of the natural history of gastric dysplasia showed that patients with HGD had a rate of malignant progression or synchronous malignant lesions of 60–85% over a median interval period of 4–48 months.44 Song et al, in their observational study, found that 1 in 19 patients with dysplasia progressed to gastric adenocarcinoma within 20 years, although no differentiation was made between those with LGD or HGD.15 de Vries et al noted the annual incidence of gastric cancer was 6% in patients with HGD within 5 years.25 Of note, it is difficult on histopathology to distinguish between HGD and gastric adenocarcinoma from small biopsy samples.

The risk of progression in individuals with LGD is less clear. There is evidence to show that LGD will regress in 38–75% of patients and persist in 19–50%. In the LGD lesions that persist, the risk of malignant progression ranges from 0% to 23% in the published literature over 10–48 months.44 de Vries et al reported that the annual incidence of gastric adenocarcinoma in those with LGD was 0.6% within 5 years after diagnosis.25

Visible LGD following resection is upstaged in 25–35% of lesions, including those of <1 cm, with an adenocarcinoma rate of 6.9%.44 We have therefore suggested that the risk of prevalent HGD or gastric adenocarcinoma is greater in visible LGD. Although there is uncertainty about the natural history of non-visible LGD, the evidence suggests that there is an increased rate of progression, but the magnitude is unclear.

A summary of the risks of gastric cancer is shown in table 3.

Biomarkers and gastric cancer

What biomarkers are useful in the management of these lesions?

Can they be applied to population screening, monitoring those at risk or those with known lesions?

Measurement of serum pepsinogen I and serum pepsinogen I/II ratio alone or in combination with H. pylori serology, and/or gastrin-17 can identify individuals with extensive atrophic gastritis (evidence level: low quality; grade of recommendation: weak; level of agreement: 93%).

We do not recommend the use of biomarkers as a screening tool in areas with a low incidence of gastric adenocarcinoma, such as the UK (evidence level: low quality; grade of recommendation: weak; level of agreement: 100%).

CAG, dysplasia and gastric adenocarcinoma

Pepsinogen I (PGI) is mainly secreted by chief and mucous neck cells in the fundic mucosa, while pepsinogen II (PGII) is also secreted by pyloric and duodenal Brunner’s glands. Approximately 1% of pepsinogens are found in the serum, with their serum level accepted as a marker for the morphological and functional status of the gastric mucosa.10 112 113 Serum PGI and PGII levels both increase in gastric mucosal inflammation; however, as GA develops and specialised cells are lost, PGI and PGII levels decrease, usually more marked in PGI, resulting in low serum PGI and a low PGI/II ratio.114

The combined use of serum PGI and PGI/II ratio measurements is an accepted useful biomarker for premalignant and malignant gastric lesions.112 Studies exploring their use as a population screening tool, where those with a positive pepsinogen result progress to endoscopic examination, have been shown to be acceptable in screening asymptomatic populations, with a good uptake of invitations for endoscopy—over 60%, in those with a positive pepsinogen result.115 116

Biomarker detection of CAG, dysplasia and gastric cancer

Numerous studies across many different countries and populations have explored the use of serum pepsinogen testing for detection of CAG, dysplasia and gastric adenocarcinoma. The majority of these studies are from countries with a higher incidence of these lesions than the UK. For countries with an incidence of gastric adenocarcinoma similar to that of the UK, a study by Broutet et al in 2003 assessed serum pepsinogen testing across 14 European nations and determined that the PGI/II ratio may be of use as a screening test.112 117 118 However, in the majority of studies, the values used to define a positive pepsinogen test result and the study outcomes are reflective of populations that differ from the UK population. Additionally, many of these studies use differing cut-off values, which makes comparison difficult. However, the most frequently used values for these studies are a PGI <70 ng/mL and a PGI/II ratio ≤3.112 Values can be affected by laboratory methodologies and population settings, and therefore may require adjustment if applied to the UK population.

A meta-analysis by Huang et al found that serum pepsinogen testing had a sensitivity and specificity of 69% and 73% for gastric cancer diagnosis and 69% and 88% for CAG diagnosis, respectively.112 This analysis included a study population of over 30 000 individuals, across 13 different countries and diagnosis confirmed with gastroscopy and biopsy. This is comparable with the 2004 meta-analysis performed by Dinis-Ribeiro et al, using a cut-off value of PGI <50 and PGI/II ratio ≤3 for dysplasia detection, where sensitivity and specificity were 65% and 74–85%, respectively.117 This included a study population of approximately 300 000 patients. Finally, in a 2014 meta-analysis by Terasawa et al, a study population of approximately 32 000 patients, in which individuals were prospectively followed up for between 3.9 and 14 years found that a positive pepsinogen test had a sensitivity of 57% and specificity of 76% for the development of gastric adenocarcinoma.119

Biomarker population screening

Although pepsinogen testing has been demonstrated as a useful tool for population screening in high-risk areas, its use in low-risk areas such as the UK has not been explored.116 119 120 In particular, one must consider whether the moderately effective sensitivities and specificities of pepsinogen testing are cost-effective for screening an asymptomatic population where the incidence of the disease is low. A 2015 study by Yeh et al used a mathematical simulation model to calculate the cost-effectiveness of population screening strategies based on biomarker and endoscopic technologies in the low-risk US population (defined as an age-standardised rate (ASR) <10 per 100 000).113 This study found that although one-time serum pepsinogen testing at the age of 50 could prevent one in four cases of gastric adenocarcinoma among men, it was not of high value in improving cancer outcomes. However, targeting the high-risk group of male smokers over 50 years old could be a cost-effective way to reduce mortality from gastric adenocarcinoma.113 This screening model requires further exploration.

H. pylori serology has been studied extensively both as a population screening tool alone, as part of a ‘test and treat’ strategy and in conjunction with pepsinogen testing. While its use as a screening tool or in the ‘test and treat’ strategy may have advantages in high-incidence areas (defined as an ASR >20 per 100 000), the disadvantages, including low specificity and antibiotic resistance in low-risk populations, render its use redundant.121 The use of H. pylori serology in combination with pepsinogen is more accurate. In particular, the finding of negative serum anti-H. pylori IgG antibody and positive pepsinogen measurements suggests extensive GA, and thus these individuals are at highest risk of progression to cancer.10 121 The ABCD method for the detection of individuals with high gastric adenocarcinoma risk has been extensively investigated in high-risk populations. This method categorises patients tested for H. pylori serology (HP) and the ratios of serum PGI and PGII (sPG) into low risk (A: HP−, sPG−), moderate risk (B: HP+ and sPG−) and high risk (C: HP+ and sPG+; D: HP−, sPG+). Although in a 20-year prospective study the HRs for developing gastric adenocarcinoma were 15 for group D when compared with group A,122 further evidence is required to support the use of these approaches in a low-risk population.

GastroPanel combines PGI, PGII, gastrin-17 and H. pylori serology. A recent meta-analysis assessed the performance of this serum panel test for the diagnosis of CAG in 4241 subjects. The sensitivity for CAG was 74.7% (95% CI 62.0% to 84.3%) and the specificity was 95.6% (95% CI 92.6% to 97.4%). With a prevalence of CAG of 27% (median prevalence across the studies), the negative predictive value was 91% and the positive predictive value was 86%.123 Thus, although studies to date have shown promise, there is little evidence to support its use, with conflicting data on its efficacy.124–126

Finally, note is made of studies exploring the detection of volatile organic compounds in exhaled breath that are associated with the detection of gastric adenocarcinoma.127

Endoscopic diagnosis of premalignant or early malignant lesions of the stomach

Ensuring high-quality endoscopic evaluation

Outside the recommendations within the UK guidelines for quality endoscopy, are there further processes that are suggested for the detection and diagnosis of premalignant or early malignant lesions of the stomach? Is a station-based approach beneficial?

We recommend that patients at higher risk for gastric adenocarcinoma, including GA and GIM, should undergo a full systematic endoscopy protocol of the stomach with clear photographic documentation of gastric regions and pathology. We suggest a minimum examination time of 7 min (evidence level: moderate quality; grade of recommendation: strong; level of agreement: 100%).

Endoscopy needs to be of high quality in order to detect dysplasia and early cancers, particularly in light of postendoscopy gastric cancer rates of 11.3%.128 Recognising and targeting high-risk patients with GA and GIM may be the most effective means to improve gastric cancer detection, and possibly survival, in the UK. There are three basic principles, which are part of routine practice: cleaning of the gastric mucosa, adequate distention of the gastric wall by air insufflation and mapping the entire stomach.

Although there is little available evidence on the use of smooth muscle relaxants and mucosal cleaning techniques, the new BSG position statement on quality standards in upper gastrointestinal endoscopy recommends ‘Adequate mucosal visualisation should be achieved by a combination of adequate air insufflation, aspiration and the use of mucosal cleansing techniques’.129 We recommend that this guidance is also applied to inspection of GA and GIM.

There is also a paucity of data on the assessment of a complete endoscopic procedure. In particular, no clinical trials have directly clarified key performance indicators or quality assurance (QA) for improving gastric cancer detection, and no studies have explored the outcomes of systematic screening protocols for the stomach. The European Society of Gastrointestinal Endoscopy (ESGE) performance measures for upper GI endoscopy have suggested that the inspection of the oesophagus, stomach and duodenum should last at least 7 min from intubation to extubation.130 This statement was based on a retrospective cohort study by Teh et al, which aimed to determine the diagnostic yield for early neoplastic lesions in the stomach.131 After evaluating 837 endoscopies of symptomatic patients with no history of gastric cancer, they found that a ‘slow’ endoscopist (>7 min examination) was twice as likely to detect high-risk gastric lesions, defined as biopsy evidence of GA, GIM, gastric dysplasia or cancer, and three times as likely to detect a case of dysplasia or cancer compared with a ‘fast’ endoscopist (<7 min examination).131 132

Photographic documentation might be an indirect quality indicator. Endoscopists with longer procedure times, who take more than four pictures, detect more pathology.132 The ESGE has recommended five areas in the stomach should be photographically documented, including the cardia and fundus in inversion, corpus in forward view including lesser curvature, corpus in retroflex view including greater curvature, angulus in partial inversion, and antrum. Images may be used in case discussions, patient management and compared with histology to aid learning.

In Japan, a systematic screening protocol for the upper GI tract has been developed, although this is considered too complex for routine clinical practice.133 This was revised by the Japanese Society of Gastroenterological Cancer Screening to a simplified, but still elaborate protocol.134 Yao has more recently simplified this further to propose as a minimum required quality standard a ‘systematic screening protocol for the stomach’.135 This is a station-based approach whereby each area of the stomach is viewed and photographed in either a clockwise or counterclockwise manner. The 22 pictures are arranged according to the order of the procedure. Additional pictures are taken of lesions (Figure 2). A recent study from China found that training including a systematic inspection protocol with 20 photos increased the detection of early gastric cancer from 0.2% to 2.3%.136

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Systematic screening protocol for the stomach. This is a station-based approach whereby each area of the stomach is viewed and photographed in either a clockwise or counterclockwise manner. The 22 pictures are arranged according to the order of the procedure. Q, quadrant; L, lesser curvature; A, anterior wall; G, greater curvature; P, posterior wall; SSS, systematic screening protocol for the somach.

An e-learning module has been developed to teach endoscopists how to diagnose early gastric cancer based on the characterisation of gastritis-like lesions, ulcerative lesions and polypoid lesions, the so-called GUP system.137 138 The GUP system has been evaluated in an RCT involving 332 endoscopists in 27 countries, with a higher mean improvement rate in the e-learning group than that of the non-e-learning group.138 A further study clearly demonstrated the efficacy of an e-learning system in improving endoscopists’ capabilities to diagnose early gastric cancer using magnification-narrow band imaging.139 Such validated training modules may be incorporated into any future quality improvement (QI) programmes aimed at improving diagnosis of gastric cancer.

We recommend that when either GA or GIM is recognised on WLE, a full systematic endoscopic examination of the whole stomach is performed, taking no less than 7 min, with full photographic documentation of antrum, pylorus, incisura, lesser curve, greater curve, fundus and cardia. For patients without known risk factors for gastric cancer, we recommend a standardised high-quality endoscopy as defined in the UK quality in upper gastrointestinal endoscopy position statement.

Normal gastric appearances

White light endoscopy

Typically, the surface of the normal corpus, when the stomach is empty and not distended, is almost invariably in folds, also called rugae, which vary in size depending on the degree of insufflation during the endoscopic assessment. In contrast, the surface of the normal fundus and antrum is smooth. The colour of the normal gastric mucosa, as indeed of the whole GI tract, is velvety, glossy dark rose or red with a regular arrangement of the collecting venules, usually visible as red spidery vessels in the normal corpus.140–143 The presence of these collecting venules is characteristic of a normal stomach without H. pylori (sensitivity 93%, specificity 48%).140 144 145 With current white light, high-resolution endoscopes, the round ‘pit patterns’ of the corpus and elongated ‘pit patterns’ of the antrum can be seen without magnification or enhancement (Figure 3Di).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Normal gastric corpus and antrum mucosal surface pattern with white light, enhanced and magnification endoscopy. The round ‘pit patterns’ of the corpus (Ai) and elongated ‘pit patterns’ of the antrum (Di) can be seen without magnification or enhancement. In the corpus (Ai), the red collecting venules (CV) are evident as well as the round dark red crypt openings (CO). The vascular anatomy becomes more pronounced with NBI (Aii & Dii). The visible anatomical components seen on magnification NBI include the dark brown ‘crypt openings’ (CO), the dark brown sub-epithelial capillary network (SECN), and the light brown marginal crypt epithelia (MCE). The corpus mucosa has dark round ‘crypt openings’, surrounded by the lighter MCE, followed by the darker circular SECN (Aii & Aiii). In contrast, the dark, oblique ‘crypt openings’ in the antrum are grooved so the light coloured ridged or villiform epithelium (MCE) surrounds the dark SECN, termed the ‘groove type pattern’ (Dii & Diii). The corresponding histopathological architecture can be seen in the corpus (B and C) and in the antrum (E and F).

Chronic H. pylori gastritis

A number of endoscopic features are suggestive of chronic H. pylori gastritis, including the absence of collecting venules, antral nodularity, enlarged gastric folds, enlargement and destruction of the gastric glands, sticky tenacious adherent mucous, turbid gastric juice, and xanthomas.140–142 Loss of collecting venules and capillary vascular structure was correlated with chronic inflammation and activity. With progression of mucosal atrophy, irregular collecting venules become visible.151

Gastric atrophy

Four principal endoscopic features of GA are described by Nakayama et al, Uedo et al and Yao et al 152–154: pallor, loss of gastric folds, prominence of the vessels and the atrophic border (Figure 4). Increased visibility of the vascular network showed a sensitivity of 48% and a specificity of 87%, while the loss of gastric folds has a sensitivity of 67% and specificity of 85%.155

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Chronic atrophic gastritis (CAG) and the atrophic border on white light and image enhanced endoscopy. There are four principle endoscopic features of CAG: palor (A, B, C, D and E), loss of gastric folds (A, B, C, D and E), prominence of the vessels (A, B, C and D) and the atrophic border (A and B). The paler areas of atrophy are also clear on image enhancement (F).

Long-term cohort studies suggest that the Kimura-Takemoto classification is a useful risk stratification assessment tool to predict gastric adenocarcinoma development.26 31 156 Essentially this tool uses the extent of the atrophic border (the border between the pale atrophy and normal red-coloured stomach) to stage the extent of GA (Figure 4). In a cross-sectional cluster sampling historical study between the UK and Japan, endoscopic grading was shown to be comparable to histopathology and correctly predict histopathological atrophy with few false-negative results.157 158 This work needs further confirmation in a larger setting outside of Japan. The Kimura-Takemoto classification has been simplified to a modified staging system involving the antrum only (antral), antrum to incisura (antral dominant), antrum to lesser curve (corpus dominant), and antrum, lesser curve and greater curve (pan-atrophy). This staging system integrates the Sydney biopsy system discussed later (Figure 5). The GDG agreed that the extent of GA and GIM should be stratified as low risk (involving the antrum and incisura) or high risk (involving the corpus and the antrum/incisura or the corpus alone).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

The Integrated and Modified Kimura & Sydney Biopsy System. The modified Kimura staging system divides the extent of atrophy into antrum only (antral), antrum to incisura (antral dominant), antrum to lesser curve (corpus dominant), and antrum, lesser curve and greater curve (pan-atrophy). This system integrates Sydney protocol biopsies which should be taken from the antrum (site 1 and 2), incisura (site 3), lesser curve (site 4) and greater curve (site 5). The anatomical CAG boundaries and biopsy sites can be seen in the splayed (A) and cross-sectional cartoon (B) of the stomach. The biopsy sites defined in the endoscopic retroflexed (C) and forward view (D).

Gastric intestinal metaplasia

White light endoscopy

GIM typically appears as small grey-white slightly elevated plaques surrounded by mixed patchy pink and pale areas of mucosa causing an irregular uneven surface (figure 6A). Mottled patchy erythema has also been positively associated with GIM.159 However, diagnosis using standard endoscopy alone (without high-resolution or enhanced imaging) is unreliable.160–165

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Gastric intestinal metaplasia under white light, image enhanced and magnification endoscopy. Intestinal metaplasia typically appears as small grey-white slightly elevated plaques surrounded by mixed patchy pink and pale areas of mucosa causing an irregular uneven surface (A). These appearances are more evident with image enhancement (B). Corpus GIM can be distinguished from the normal straight/tubular glands of the corpus by a ‘groove type pattern’ similar to that of the antrum or villiform pattern of the intestine and may be appreciated with higher resolution technology on white light endoscopy (C). GIM in the antrum is more difficult to characterize as the normal glands are oblique. Additional features of GIM to aid diagnosis in the antrum include the light blue crest (LBC) and the marginal turbid band (MTB) (D). The LBC is a fine, blue-white line on the crest of the epithelial surface seen with NBI enhancement (Fine arrows in D). The MTD can be seen between the broad arrows. The numerous goblet cells characterise GIM (E & F).

Magnification and IEE

As patches of GIM expand, the straight/tubular glands of the corpus elongate to a ‘groove-type pattern’ similar to that of the antrum or villiform pattern of the intestine (figure 6C D). Although these changes can easily be distinguished from the normal mucosal background in the corpus both with high-resolution white light and enhanced endoscopy, GIM in the antrum is more difficult to characterise as the pre-existent mucosal architecture is quite similar and also appears oblique and grooved.148 166 Additional features of GIM include the light blue crest (LBC) and the marginal turbid band (MTB)167 and the white opaque substance (lipid droplets) obscuring the subepithelial capillaries.168 The LBC is a fine, blue-white line on the crest of the mucosal surface seen with NBI enhancement and is a highly specific sign of the presence of GIM166 169 (figure 6E F). The MTB is a white turbid band on the mucosal surface. It is suggested that the MTB may represent a sign of early GIM, whereas the LBC appears with progression to severe GIM, although whether this is reproducible remains uncertain.

Optimal techniques for the detection and classification of GA and GIM

Most studies have demonstrated improved accuracy of enhanced and magnification imaging in the classification and detection of GA, GIM, dysplasia and cancer, in comparison with WLE.150 162 163 165 170 171 An overview of enhanced imaging studies with various modalities and their performance characteristics is provided in table 4.

A simplified classification system using NBI without magnification by Pimentel-Nunes et al has been shown to be accurate and reliable for the diagnosis of GIM and dysplasia.150 In the validation study a tubulovillous mucosal pattern was associated with GIM (accuracy 84%). Irregular vessels and mucosal pattern were associated with dysplasia (accuracy 95%). The LBC finding was moderately reliable (k=0.49), but very specific (96%) for GIM. In a recent study by Kanemitsu et al,168 the sensitivity and specificity of LBC for histologically diagnosed intestinal metaplasia were 62.5% and 93.8%, respectively. The sensitivity and specificity of white opaque substance were 50.0% and 100.0% (95% CI 85.0% to 100.0%), respectively. The combination of LBC and white opaque substance improved the overall sensitivity up to 87% and 93.8%.

A second prospective multicentre study compared WLE with WLE plus NBI.162 NBI demonstrated a high concordance with histopathological diagnosis, superior to standard WLE. However, it is important to note that this study assessed WLE plus NBI, rather than NBI alone, and although high-resolution WLE had a good overall sensitivity of 85% for all pathology, this decreased to 53% for the detection of GIM. NBI versus WLE increased the sensitivity for the diagnosis of intestinal metaplasia significantly (87% vs 53%; p<0.001) and for the diagnosis of dysplasia (92% vs 74%). This study suggests that WLE alone is therefore not sufficiently accurate for GIM detection. Similar comparative results have been demonstrated in a recent prospective blinded trial.165

The sensitivity and specificity of WLE for the histological diagnosis of GA were reported to be 61.5% and 57.7%, respectively, in the antrum, and 46.8% and 76.4%, respectively, in the body of the stomach.170

Compared with WLE, NBI combined with magnifying endoscopy can also effectively diagnose early gastric adenocarcinoma.171

Thus, in summary, the GDG agreed that WLE alone was not sufficiently accurate to reliably diagnose GA or GIM, and enhanced optical techniques should be used for diagnosis and staging.

A scale for endoscopic staging of GIM using NBI was created and returned an area under the curve (AUC) of 0.98 for WLE followed by NBI for diffuse GIM.162 This was externally validated by the same group, and for a diagnosis of OLGIM III/VI the AUC was 0.96 (95% CI 0.93 to 0.98). This endoscopic grading of GIM was 89% sensitive and 95% specific for a risk stratification of moderate to severe GIM if a cut-off score of >4 was used.172 On this basis, it could be argued that endoscopic staging with high-resolution WLE plus NBI is sufficiently accurate for diagnosis and staging. This is an area of active future research.

Gastric dysplasia and early gastric cancer

Detection of gastric dysplasia and early gastric cancer is notoriously difficult due to the often only subtle findings and the lack of well-defined endoscopic appearances under white light inspection. Features commonly described, but not exhaustive, include differences in colour (ie, more red or pale), loss of vascularity, slight elevation or depression, nodularity, thickening, and abnormal convergence or flattening of folds.173 174

Therefore, optimal clearing of mucous and secretions is essential to allow for continuous and meticulous search of areas with features different from the surrounding mucosa.175 176

Recourse to chromoendoscopy with indigo carmine solution (0.2%) or virtual chromoendoscopy (NBI, flexible spectral imaging colour enhancement (FICE), i-Scan, blue laser imaging) is commonly advocated to enhance contrast and visualisation of areas of concern or mucosal abnormalities. Areas of dysplasia may present throughout the stomach, with a slight predominance in the antrum and along the lesser curvature, and can vary in size from a few millimetres to several centimetres.

Gastric dysplasia can be morphologically classified into adenomatous (intestinal), which includes adenomatous polyps; foveolar (gastric); and hybrid type.47 177 Compared with the adenomatous type, the foveolar type appears to be more commonly associated with HGD.47 178 Endoscopically, these lesions are usually detected as 0–Is, 0–IIa or 0–IIc types according to the Paris classification of superficial neoplastic lesions.175

Adenomatous dysplasia is more likely to occur in the gastric body and lesser curvature of the stomach, whereas foveolar dysplasia is more typically located in the gastric antrum and incisura angularis. In addition, foveolar-type lesions are smaller, often reddish in colour and present as flat or depressed areas more frequently than the adenomatous type.47

Surveillance of CAG

As the neoplastic cascade follows a multistep process from H. pylori-associated gastritis through GA and GIM to dysplasia,179 it follows that surveillance of a high-risk population may lead to the detection of early gastric cancer. Furthermore, the evolution of endoscopic techniques such as ESD, with 5-year disease-free survival rates of 99%,180 further supports the detection of early gastric cancer through surveillance.

Several factors influence the risk of progression to cancer, including the extent of atrophy and GIM and a family history of gastric cancer. A strong family history is defined as those with a first-degree relative or two or more second-degree family members with histologically proven gastric adenocarcinoma. Those at the highest risk with GIM in the antrum and corpus have a 9.8% cumulative 5-year incidence of cancer.26

Summary

Although there is a lack of randomised data on the survival benefits of surveillance for GIM, there is moderate evidence demonstrating progression of premalignant conditions, particularly extensive GIM, to gastric adenocarcinoma, and evolving data from surveillance studies. The cost-effectiveness data are also compelling. We do accept, however, as with surveillance of Barrett’s oesophagus, that surveillance of GA and GIM will remain controversial. We recommend a surveillance interval of 3 years for those patients with extensive GA or GIM defined as that affecting the antrum and body. However, surveillance may not be appropriate for all patients with extensive atrophy and GIM, particularly the very elderly and those with multiple comorbidities where the benefit of surveillance may be offset by the risks of diagnostic endoscopy. For those with GA or GIM limited to the antrum, but with additional risk factors such as a family history of gastric adenocarcinoma and persistent H. pylori infection, we also recommend a surveillance interval of 3 years. Persistent H. pylori infection is defined as that refractory to treatment. We also suggest, where possible, as with Barrett’s oesophagus surveillance, that endoscopy is undertaken on a dedicated screening list. The remit of these guidelines do not cover the management of HDGC.63

Treatment: endoscopic therapy

What lesions are amenable to endoscopic removal?

How should these lesions be removed?

Are there criteria on histopathological assessment that determine prognosis and follow-up?

We recommend that all gastric dysplasia and early gastric adenocarcinoma should be resected en bloc (an EMR technique can achieve en bloc excision for lesions ≤10 mm in size, but only an ESD technique can ensure en bloc excision for lesions >10 mm in size) (evidence level: high quality; grade of recommendation: strong; level of agreement: 100%).

We recommend that complete (R0) endoscopic resection of gastric dysplasia and early gastric adenocarcinoma with the following features should be considered as curative:

1. LGD.

2. HGD.

3. Well or moderately differentiated intramucosal adenocarcinoma, irrespective of size and without ulceration.

4. Well or moderately differentiated intramucosal adenocarcinoma, <3.0 cm in size if ulcerated.

5. Well or moderately differentiated submucosal adenocarcinoma, <3.0 cm in size, with superficial submucosal invasion (Sm1; <500 μm submucosal invasion as measured in a straight line from the deepest fibre of the muscularis mucosae).

6. Poorly differentiated intramucosal adenocarcinoma, ≤2.0 cm in size (evidence level: moderate quality; grade of recommendation: strong; level of agreement: 93%).

The histopathological features of early gastric adenocarcinoma associated with a higher risk of LNM after endoscopic resection include the following:

1. Poorly differentiated submucosal cancer, irrespective of invasion depth below muscularis mucosae.

2. Signet ring cancer.

3. Lymphovascular invasion.

4. Depth of submucosal invasion ≥500 μm as measured in a straight line from the deepest fibre of the muscularis mucosae (evidence level: moderate quality; grade of recommendation: strong; level of agreement: 93%).

We suggest that, where possible, all cases considered for resection should be discussed in an MDT with the appropriate expertise, including pathologists and therapeutic endoscopists. When there is no local expertise, patients should be referred to an expert centre. Before any therapeutic procedure is undertaken, the risks and benefits of endoscopic resection and surgery should be discussed with the patient to aid their decision-making.

These recommendations apply to the intestinal type of gastric cancer as defined by the Lauren classification.214 215 Diffuse-type adenocarcinoma carries a worse prognosis than the intestinal type, which appears to be independent of the T and N stage.215

The risk of LNM underpins endoluminal therapy for early gastric adenocarcinoma. This risk has to be weighed against the significant risk of morbidity and mortality following surgical resection. Endoscopic resection has become the preferred organ-preserving treatment for superficial gastric neoplastic lesions because of the low risk of LNM that these lesions portend. In a large series from the Far East, Gotoda et al reviewed the prevalence of LNM in 5265 gastrectomy specimens. On multivariate analysis they found that none of the 979 non-ulcerated lesions had LNM. Additionally, they found that none of the 145 well or moderately differentiated adenocarcinomas measuring <30 mm, those with submucosal invasion of <500 μm (Sm1), and those without lymphovascular invasion showed LNM.216 This supported the initial Japanese guidelines on the indication criteria for endoscopic resection of early gastric adenocarcinoma, which included intestinal-type adenocarcinoma, endoscopically diagnosed intramucosal cancer, lesion size of ≤20 mm and non-ulcerated lesions.212 In a second large series, Hirasawa et al reviewed 3843 patients who underwent gastrectomy with lymph node dissection for poorly differentiated adenocarcinoma. On multivariate analysis they found that lesion size of >20 mm, lymphovascular invasion and submucosal involvement were independent risk factors for LNM.217

Overall the two large series on surgically resected early gastric cancer demonstrate that the risk of LNM of superficial lesions is small (<1%) if the following criteria are met:

1. Well or moderately differentiated intramucosal adenocarcinoma, irrespective of size and without ulceration.

2. Well or moderately differentiated intramucosal adenocarcinoma, ≤3.0 cm in size if ulcerated.

3. Well or moderately differentiated submucosal adenocarcinoma, <3.0 cm in size, with superficial submucosal invasion (Sm1; <500 μm submucosal invasion as measured in a straight line from the deepest fibre of the muscularis mucosae).

4. Poorly differentiated intramucosal adenocarcinoma, <2.0 cm in size.

This led to the adoption of the expanded indications for ESD of early gastric cancer. The expanded criteria, however, should not be taken as absolute, particularly where the balance between risk and benefit of surgery is less clear. The original definition of early gastric cancer was defined in 1971 by the Japanese Society of Gastroenterology and Endoscopy as a carcinoma limited to the mucosa and/or submucosa regardless of the lymph node status. This has recently fuelled much controversy as the survival of early gastric cancer is closely associated with the risk of LNM. We have therefore adopted the definition of adenocarcinoma limited to the mucosa and superficial submucosa (Sm1; <500 μm), as that amenable to endoscopic resection.

Two recent series, one from the Far East and one from the West, assessed the outcomes of the extended indications. The series by Hasuike et al included 470 lesions, of which 466 (99.1%) were resected en bloc and the curative resection rate was 67.4%.218 A similar Western series reported that en bloc resection was achieved in 81 of 91 lesions (89.0%) and curative resection was achieved in 67 of 91 lesions (73.6%).219 The risk of perforation in both series was reported to be up to 2.6%. However, the risk of delayed bleeding was lower in the Western series (2.2% compared with 6.2%). The higher curative resection and lower delayed bleeding rate in the Western series are encouraging, although the number of treated lesions was smaller.

Patient selection is key to achieving favourable outcomes with ESD for early gastric cancer. Patients who do not meet the expanded criteria for a curative outcome following gastric ESD are referred for radical surgery. In a multicentre retrospective study, Hatta et al developed a risk scoring system using multivariate logistic regression analysis of 1101 patients who had undergone radical surgery after failing to meet the criteria for curative endoscopic resection of early gastric cancer.220 They then validated the scoring system in a further 905 patients. They showed that the scoring system known as ‘eCura system’ (table 5) predicted cancer-specific survival in this cohort of patients. This scoring system is promising but will require further validation in other centres.

EMR (cap-assisted) was the initial technique used to resect superficial gastric neoplasia. However, this technique is unable to effectively resect lesions larger than 10 mm en bloc. In a recent meta-analysis comparing the efficacy of gastric ESD and EMR, the en bloc and R0 resection rates of EMR were found to be 51.7% and 42.4%, respectively.221 This, in turn, is associated with local recurrence rates as high as 30%.180 203 207 222–225 In a subgroup analysis of lesions smaller than 10–15 mm, it was noted that there was no difference in survival regardless of the endoscopic resection technique.

ESD was a technique developed to overcome the shortcomings of gastric EMR, enabling the en bloc resection of lesions >10 mm. In a large series from the Far East of 1033 early gastric cancer lesions, Oda et al reported an en bloc resection and R0 resection rate of 98% and 93%, respectively.226 Three meta-analyses comparing the outcomes of EMR and ESD showed that ESD achieved higher en bloc resection rates (92% vs 52%; OR=9.69, 95% CI 7.74 to 12.13), histopathologically complete resection rates (82% vs 42%; OR=5.66, 95% CI 2.92 to 10.96) and lower recurrence rates (1% vs 6%; OR=0.10, 95% CI 0.06 to 0.18).221 227 228

The current Japanese Gastroenterological Endoscopy Society and ESGE guidelines recommend ESD as the preferred treatment for most superficial gastric neoplastic lesions.211 229 However, ESD is a technique that is in its infancy in the West, and the complication rates during early adoption can be high. In one European series of 75 patients who underwent gastric ESD, en bloc resection was achieved in 85.3% and R0 resection in 84.0%. However, the complication rate which included delayed bleeding and perforation was as high as 24%.230 Finally, it should be noted that signet ring cancer is not currently recommended for endoscopic resection. However, prospective data will soon be published on the treatment of these lesions with ESD if they are <20 mm, which may alter this management strategy.

In summary we recommend, where appropriate, endoscopic resection as first-line treatment for all early gastric neoplasia in line with the Japanese extended indications, favouring ESD over EMR for larger lesions owing to the superior R0 resection rate. Surgery should be undertaken only when endoscopic resection is not considered curative or is the preferred patient option.

Treatment: pharmacological

Is there a role for other pharmacological therapies for example, COX inhibitors and antioxidants?

We do not recommend the use of NSAIDs or COX-2 inhibitors to reduce the risk of progression of premalignant lesions of the stomach (evidence level: moderate; grade of recommendation: strong; level of agreement: 100%).

We do not recommend the use of antioxidants as a means to reduce the prevalence of premalignant gastric lesions (evidence level: moderate; grade of recommendation: strong; level of agreement: 100%).

Endoscopic screening for gastric adenocarcinoma

Is there evidence to support the introduction of a population screening programme for glandular gastric cancer?

We suggest endoscopic screening should be considered in individuals aged ≥50 years with multiple risk factors for gastric adenocarcinoma (male, smokers, pernicious anaemia)—in particular, in those with a first-degree relative with gastric cancer (evidence level: low quality; grade of recommendation: weak; level of agreement: 100%).

We do not recommend endoscopic screening for gastric adenocarcinoma in the UK population (evidence level: low quality; grade of recommendation: strong; level of agreement: 100%).

Evidence for the effectiveness of endoscopic screening for prevention of gastric cancer has been gathered from studies conducted in high-risk populations (defined as an ASR >20 per 100 000); in Japan and Korea (29.9 and 41.3, respectively).244 These include five cohort studies and three case–control studies. Although there are no randomised control trials, the results of the available studies suggest a reduction in mortality from gastric cancer in screened populations. In the more recent regional Japanese cohort studies, there was a reduction in gastric cancer mortality (calculated as the standardised mortality ratio or adjusted RR) of 57% and 67% after a 5-year and 6-year follow-up.245–247 The earlier cohort studies from Japan and China were less convincing but were limited by a broad age distribution and poorly matched cohorts.248 249 The results of the case–control studies were equally variable with a reduction in gastric cancer mortality of between 20% and 80% in high-risk populations in Japan and South Korea.250 251 Based on these results, the authors recommended endoscopic screening in regions with a high incidence of gastric cancer.

The test characteristics of endoscopic screening have been described in four studies. Hosokawa et al found the sensitivity of endoscopy to be 78% after comparing the gastric cancer incidence in those screened from a cancer registry.252 Similar sensitivities of 69% and 89% have been found across other cancer registries in Japan and South Korea.253–255 In the South Korean study the gastric cancer detection rate was 2.61 per 1000 screening endoscopies, with a specificity of 96%.253

There have been two studies in low-incidence regions (defined as an ASR <10 per 100 000), such as the USA (3.9 per 100 000), assessing the cost-effectiveness of endoscopic screening. The cost of a single-screening endoscopy at the age of 50 in the general population was US$115 664 per QALY, suggesting that endoscopic screening was not cost-effective.184 256 Cost-effectiveness analyses in higher risk areas have demonstrated an ICER of US$44 098 and US$25 949 per QALY for annual and biannual screening endoscopy, respectively.186 In a study in Taiwan where patients at the age of 50 who had low levels of pepsinogen-I (<30 ng/mL) were offered endoscopy, the ICER was US$29 741 per life-year gained.257

A Markov model of screening in an intermediate-risk population (ASR of >10 and <20 per 100 000 population) for ages 50–75 years found that upper endoscopy combined with screening colonoscopy (every 10 or 5 years) had an ICER of €15 407/QALY and €30 908/QALY, respectively.258 Standalone endoscopic screening (every 5 years) had an ICER of €70 693/QALY and pepsinogen screening an ICER of €143 344/QALY. This work suggests that endoscopic gastric cancer screening in conjunction with a scheduled colonoscopy may be cost-effective in countries with intermediate gastric adenocarcinoma risk such as in Eastern Europe or Portugal. These results imply that resources allocated to endoscopic colorectal cancer screening programmes could be used to provide gastric cancer screening, both for detection of high-risk individuals with extensive premalignant conditions and patients with early gastric cancer.

Although there is insufficient evidence to support screening in low-risk populations, a recent study by Shawihdi and colleagues demonstrated wide variations in rates of elective gastroscopy within general practice populations. They showed that patients with oesophagogastric cancer belonging to practices with low rates of gastroscopy were at increased risk of poor outcome. However, if the low referral practices increased to the mid-referral range, the crude cost per life-year saved for a hypothetical scenario is £140 000, well above a suggested threshold of £25 000. Therefore despite the poorer outcomes, given the high costs, primary care physicians should follow a restrictive referral practice.259

A recent retrospective nationwide Taiwanese propensity-matched cohort study evaluating the impact of non-screening gastroscopy on gastric cancer-related mortality found that patients with gastric cancer who had undergone gastroscopy in the 5 years before the diagnosis of gastric cancer had a better survival than patients who had never undergone a gastroscopy or whose last gastroscopy was more than 5 years before the diagnosis. The authors found that gastric cancer was detected at an earlier stage in patients who had recently undergone endoscopy. The risk of gastric cancer in this study population is low to moderate. This is the first study showing a significant survival advantage of recent endoscopy in patients with gastric cancer in a region of low to moderate gastric cancer incidence.260

There are factors apart from the premalignant stomach that confer a greater risk of gastric adenocarcinoma as described in the ’Risk factors for gastric adenocarcinoma' section. These include family history, particularly those with first-degree relatives, pernicious anaemia with an annual incidence of 0.27% in unsubstantiated cases, older age, male and smoking. Ethnicity is also related to an increased risk, but this may be due to a higher prevalence of H. pylori. In individuals within low-risk populations who have additional risk factors as described above, screening endoscopy may be of value.

In summary, we suggest that only those with multiple risk factors for gastric cancer are considered for screening gastroscopy from the age of 50. If the gastroscopy results are normal, then we would not recommend any further screening. Where CAG is diagnosed, this guideline should be followed.

Diagnosis and management of epithelial gastric polyps

What are epithelial gastric polyps and how should they be managed ?

We recommend that the number of gastric polyps (or estimated number), location of polyps and size of largest polyp should be clearly documented (evidence level: low quality; grade of recommendation: strong; level of agreement: 100%).

We recommend that gastric polyps other than FGPs should be biopsied for histopathological assessment (evidence level: low quality; grade of recommendation: strong; level of agreement: 100%).

We recommend that photographic documentation should be undertaken for all polyps or representative polyps, if numerous (evidence level: low quality; grade of recommendation: strong; level of agreement: 100%).

We recommend that if adenomas or hyperplastic polyps are present, the background mucosa should be endoscopically assessed for GA, GIM, H. pylori and synchronous neoplasia (evidence level: moderate quality; grade of recommendation: strong; level of agreement: 100%).

We recommend that all adenomas should be resected when clinically appropriate and safe to do so (evidence level: low quality; grade of recommendation: strong; level of agreement: 100%).

We recommend that a follow-up gastroscopy should be performed at 12 months after complete endoscopic excision of adenomas, then ongoing surveillance gastroscopy annually thereafter, when appropriate (evidence level: low quality; grade of recommendation: strong; level of agreement: 93%).

We suggest that hyperplastic polyps >1 cm, pedunculated morphology and those causing symptoms (obstruction, bleeding) should be resected. If present, H. pylori should be eradicated before re-evaluation for endoscopic therapy (evidence level: low quality; grade of recommendation: weak; level of agreement: 100%).

We suggest that enhanced endoscopic imaging is used to aid characterisation of gastric polyps when there is diagnostic uncertainty following white light examination (evidence level: low quality; grade of recommendation: weak; level of agreement: 93%).

The scope of the guidelines is restricted to epithelial polyps, and thus neuroendocrine tumours and subepithelial polyps have been excluded. Gastric epithelial polyps can be mainly classified as three types107: FGPs, hyperplastic polyps and adenomatous polyps.

Fundic gland polyps

FGPs are the most prevalent type of gastric polyps (13–77%).261 262 They are typically multiple, small (<1 cm) and located in the fundus and corpus. At endoscopy they appear pale, smooth, glassy, and transparent or translucent (figure 7). Their colour is either lighter or the same colour as the surrounding mucosa. Lacy blood vessels are seen through the translucent surface and the surface shows a pattern of fine grey dots. On enhanced imaging such as NBI, FICE or i-Scan, the surface architecture becomes more prominent. FGPs are usually not associated with an increased risk of cancer, unless in the context of FAP syndrome. However, larger FGPs (>1 cm) have been shown to be dysplastic in 1.9% and contain focal cancer also in 1.9%. FGPs are associated with long-term PPI use and can spontaneously regress when PPIs are stopped.263 There is no association with background H. pylori infection or gastritis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Fundic glandular and hyperplastic polyps. A) Fundic glandular polyps seen in the corpus and body. They are either lighter or the same colour as the surrounding mucosa. B) On near view, with image enhancement, lacy blood vessels are seen through the translucent surface and the surface shows a pattern of fine grey dots. C) Hyperplastic polyps are smooth, red buttered with whitish exudates (fibrin) and are dome shaped. The surface vascular pattern is more prominent on image enhancement (D).

Management of FGPs

Number, location, morphology and the size of the largest polyp should be documented. Representative pictures of the polyps should be taken. Diagnosis is easily made from the endoscopic appearance, as described above, but biopsy confirmation should be sought when in doubt. Large numbers of polyps (>20), young age (<40 years), dysplastic appearing polyps (where the typical surface and vascular architecture alter particularly when irregular) and the presence of duodenal adenomas should lead to exclusion of FAP.107 FGPs do not require excision unless they have atypical features. Size of >1 cm, antral location, ulceration or an unusual appearance should question the diagnosis of FGP and lead to excision.264 Targeted biopsies should be taken where excision is not undertaken. Patients receiving long-term PPIs should be re-evaluated for appropriateness of the PPI, dose of PPI and alternative treatments.263 There is no role for surveillance gastroscopy for FGPs, except in the setting of FAP.107

Hyperplastic polyps

Hyperplastic polyps constitute 18–70% of all gastric polyps, are usually single or few in number and are more frequently observed in the antrum or adjacent to ulcers, stomas and gastrectomy sites. They appear as smooth, red buttered with whitish exudates (fibrin) and are dome-shaped (figure 7). They are usually small (0.5–1.5 cm), but may be larger and present as lobulated and pedunculated masses covered with superficial erosions. They are typically associated with H. pylori gastritis (25%), GA and GIM. Regression generally occurs after eradication of H. pylori (up to 70%).108

Gastric hyperplastic polyps can reveal dysplasia (1.9–19%) and malignant transformation (0.6–2.1%),265–267 especially when >1 cm and in the postgastrectomy stomach.267–270 A dysplastic hyperplastic polyp is associated with an increased risk of synchronous neoplastic lesions in the surrounding mucosa of approximately 6% of cases.271–277

Small, white and flat plaques in the fundus have the appearance of hyperplastic polyps with a foveolar pit pattern, but are areas of focal foveolar hyperplasia or more specifically, hyperplasia of the foveolar epithelium.278 These have been described as multiple white flat lesions (MWFLs) and appear to be more prevalent in those taking PPIs. Histologically, the biopsy specimens from the MWFLs included fundic gland parietal cell protrusions and oxyntic gland dilatations. There was no evidence of intestinal metaplasia.279

Audit and benchmarks

The prevalence of GIM in patients undergoing endoscopy for dyspepsia is as high as 25% in European studies. The prevalence of GIM is influenced by ethnicity, infection rates with H. pylori, age and family history of gastric cancer. GIM is present in 100% of intestinal-type gastric cancer. Detection of GIM is therefore crucial as a first step in order to identify those at risk of gastric adenocarcinoma and may be an obvious benchmark as a quality standard for upper GI endoscopy. The prevalence of GA in Western populations is lower than GIM and varies from 0% to 8%, as previously described. We suggest a benchmark of 10% detection rate for GIM and/or GA in those patients undergoing investigation for upper GI symptoms.

Education

The miss rate for gastric cancer on endoscopy is high, and awareness of the endoscopic features of its precursors, GA and GIM, is low. We suggest that knowledge of these pathologies is incorporated into the new national gastroenterology curriculum for higher training for gastroenterologists, surgeons and pathologists.

Service and cost implications of the guidelines

An extensive service evaluation has not been conducted for this guideline. We have made an estimate of the likely additional work generated by their adoption. As discussed, in Western populations (Europe and USA) the overall prevalence of CAG in young men and women (<55 years) was 0–8.3%. In older age groups (>55 years), the prevalence was reported to be up to 13%20 and in the EUROGAST study <5%19 and 5.3% in those aged 55–64 years. Of the 1.7 million endoscopies performed each year in the UK, approximately 40% are upper GI procedures.288 Thus, in a unit performing a total of 10 000 procedures a year, 4000 will be upper GI endoscopies, of which approximately 200 patients will have CAG. Assuming endoscopic diagnostic accuracy of 100% and a CAG prevalence of 5%, this will give rise to 200 additional sets of Sydney biopsies, assuming the unit was not previously taking biopsy samples for CAG. This is likely to result in a repeat procedure in a proportion of cases where IEE was not available during the initial endoscopy. There are no clear data to define the proportion of CAG that extensively affects the stomach and therefore requires surveillance, but we have estimated this to be 40% of all CAG diagnosed based on our own experience. We estimate that there will be 130 patients per unit requiring surveillance endoscopy between 1 and 3 yearly.

Future research

This guideline aims to improve the standardisation of practice in the management of patients at risk of gastric adenocarcinoma. We envisage that improved endoscopic quality, and consequently the detection of CAG and early gastric neoplasia, with targeted surveillance will improve the outcomes of gastric cancer. However, the effect of the guidelines on gastric cancer survival nationally is likely to be small, principally because only a small proportion of patients with gastric cancer will be detected by endoscopy at an early stage. Therefore, studies of cost-effective, non-invasive, population-based screening should be a research priority of the next 5–7 years. The other main priority is a measurable improvement in the quality of gastroscopy.

Quality standards (QA) and improvement (QI)

• We suggest that a QI bundle for upper GI endoscopy requires derivation and assessment as a priority for research in order to improve diagnostic rates of early gastric neoplasia and its precursors (GA and GIM).

• We suggest that quality indicators are required for systematic gastric surveillance endoscopy and photographic documentation.

Screening and surveillance

• We suggest pilot studies for non-invasive, population-based screening strategies for gastric cancer are a research priority.

• We suggest that a pilot study is required to assess the cost-effectiveness of endoscopic gastric cancer screening when combined with a screening colonoscopy.

• We suggest that further research is required to investigate the optimal surveillance strategy for CAG,

Diagnosis and staging

• We suggest that the accuracy and reproducibility of optical diagnosis and staging of GA and GIM need to be investigated in a multicentre study.

• We suggest research to determine whether the severity and distribution of CAG categorised by OLGA and OLGIM accurately reflect cancer risk during follow-up.

• Cytosponge has been shown to detect cardia intestinal metaplasia,289 but more research is needed to see if this might be a tool to help identify cardia IM as a triage to endoscopy.

• We suggest that further research is required to investigate the risk of progression of histologically and serologically confirmed pernicious anaemia.

• We suggest that the natural history and risk of progression to cancer of visible and non-visible LGD requires further research. 

Prevalence

• We suggest that further research is required to quantify more accurately the prevalence and extent of GIM in European endoscopy practice.

Planned review date

The guidelines should be considered for review in 5 years from the date of submission for publication, estimated to be November 2023.