Article Text

Download PDFPDF

Quality control in upper gastrointestinal endoscopy: detection rates of gastric cancer in Oxford 2005–2008
  1. S Vradelis1,2,
  2. N Maynard3,
  3. B F Warren4,
  4. S Keshav1,
  5. S P L Travis1
  1. 1Translational Gastroenterology Unit, John Radcliffe Hospital, Oxford, UK
  2. 2Department of Upper Gastrointestinal Surgery, Nuffield Department of Surgery, John Radcliffe Hospital, Oxford, UK
  3. 3Department of Cellular Pathology, John Radcliffe Hospital, Oxford, UK
  4. 4Department of Internal Medicine, General University Hospital of Alexandroupolis, Democritus University of Thrace, Region Dragana, Alexandroupolis, Greece
  1. Correspondence to Dr S P L Travis, Gastroenterology Unit, John Radcliffe Hospital, Oxford OX3 9DU, UK; simon.travis{at}


Background Gastric cancer (GC) represents the sum of advanced gastric cancer (AGC) and early gastric cancer (EGC). Endoscopy (with biopsies) is the gold standard for detection of GC, but a false-negative rate of up to 19% is reported.

Aim To determine whether patients with GC had had an oesophagogastroduodenoscopy (OGD) in the year preceding diagnosis that might reasonably have been expected to detect the cancer, as a measure of quality assurance of endoscopic practice.

Methods Patients with histologically proven GC were identified from pathology records. Endoscopy reports and case notes were examined to identify any OGD before diagnosis, the interval and endoscopic findings. A false-negative OGD was defined as one where GC was neither suspected nor shown at pathology, but where a diagnosis of GC was made within 12 months.

Results Between January 2005 and February 2008, 9764 OGDs were performed. GC was diagnosed in 74 patients (male/female ratio 2.89; median age 76, range 38–95). Nine (12%) patients had EGC. There were no differences in age, sex or symptoms between the EGC and AGC group. Sixty-eight of the 74 patients with GC (92%) presented with alarm symptoms. Ten of the 74 had had an OGD within 12 months before definitive diagnosis; all these were planned because of suspicious lesions. Significantly fewer biopsies were performed at OGDs preceding definitive diagnosis (median 2 (0–10) vs 6 (2–12); p=0.002).

Conclusion False-negative rates of 0% (within 12 months) and 8% (within 3 years) for diagnosis of GC are reassuring, but an inadequate number of biopsies compromises the quality assurance of endoscopy. GC presents without alarm symptoms in <10%.

  • Adult gastroenterology
  • endoscopy
  • gastrointestinal tumours
  • clinical audit
  • quality in healthcare

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Gastric cancer (GC) represents the sum of advanced gastric cancer (AGC) and early gastric cancer (EGC). It is defined as AGC when adenocarcinoma extends into or beyond the muscularis propria, and as EGC when confined to the mucosa or submucosa. This distinction is maintained regardless of the presence or absence of lymph node metastasis and derives from the fact that the most important prognostic factor is the depth of invasion.1 2 The term ‘early’ is used to highlight the limited invasion of gastric wall, which provides curability of the disease. EGC has an excellent 5-year survival rate of more than 90% following resection, which is in contrast with the dismal 5-year survival rate of 10–20% for AGC.3 4 It should be understood that the term ‘early’ does not imply that the tumour is present for a short time, because it has been estimated that the doubling time of EGC is roughly 1.5–9.5 years, in contrast with AGC, which has a doubling time of 2–10 months.5 6 Almost half a century has passed since the definition of EGC was coined by the Japanese (Tasaka, 1962),1 but there is still no answer to the question of whether EGC and AGC are different conditions or the same condition at different time points, as the data are consistent with either interpretation.

Prompt endoscopic diagnosis is the single most effective measure for reducing GC-associated mortality. This concept was one of the drivers for open-access endoscopy, which has created a considerable endoscopic burden in many countries. Nevertheless, a global false-negative rate of around 20% for the diagnosis of GC is reported. This is because the lesions of EGC are often small, so can be overlooked, and because EGC can only be diagnosed definitely from a surgical specimen, when the full thickness of the stomach wall is available.6 In the UK and other Western countries, EGC accounts for ∼15% of all GC and is rare.3 4 7 Consequently, the detection rate can be used as a quality control measure of endoscopic practice. Benchmarks for detection can be either the proportion of all oesophagogastroduodenoscopies (OGDs), the proportion of cases of GC, or the rate of detection of EGC. Endoscopic practice in need of improvement might be inferred if patients with established GC had had a negative OGD in preceding years. The aim therefore of this study was to determine whether patients with GC had had an OGD in the preceding months that might reasonably have been expected to detect the cancer, as a measure of quality assurance of endoscopic practice in Oxford.



Patients with histologically confirmed GC diagnosed at the John Radcliffe Hospital between January 2005 and March 2008 were identified from histopathology and endoscopic records. The notes of patients who had undergone at least one OGD before the definitive diagnosis were examined. The interval between the date the patient first underwent OGD (index endoscopy) to that of the OGD that established the diagnosis of GC was measured. A ‘false-negative OGD’ was defined as one where GC was neither suspected nor established by histopathology, but where a diagnosis of GC was made within 12 months. A ‘true-positive OGD’ was defined as one where the endoscopist either diagnosed GC endoscopically, which was later confirmed histologically or suspected GC that was later confirmed histologically.

Data analysis

Endoscopic and pathology reports were available from computerised databases. Information records included demographic details, mode of presentation, previous endoscopies, histological examination, operative procedures and postoperative course. Endoscopists (both gastroenterologists and surgeons) were divided, according to their grade, into consultants and middle grade. Cases referred from other hospitals were excluded. EGC was defined after resection2; all other GC was defined as AGC.

Tumour location was categorised by dividing the stomach into antrum (distal), gastric body (middle) and fundus-gastric cardia (proximal).2 Localisation of proximal tumours used Siewert's classification,8 which applies to adenocarcinomas arising in the vicinity of the oesophagogastric junction, separating those arising in the distal oesophagus (type I) from those arising at or below the cardia (type II, type III). Type II and type III adenocarcinomas are considered as GC.9 This morphological classification, based on the anatomical location of the tumour in relation to the endoscopic cardia, makes it easy to classify the tumour by endoscopy in clinical practice.9 The endoscopic cardia represents the proximal end of the gastric folds.

Adenocarcinomas were graded on the degree of glandular differentiation into well, moderately and poorly differentiated subtypes, because there is evidence that histological grading is a prognostic factor for survival from GC.10 11

Alarm symptoms were classified according to guidelines from the UK National Institute of Health and Clinical Excellence: chronic gastrointestinal bleeding, dysphagia, progressive unintentional weight loss, persistent vomiting, iron deficiency anaemia, epigastric mass, or a suspicious barium meal result.12 Dyspepsia was defined as recurrent epigastric pain, heartburn, or acid regurgitation with or without bloating, nausea or vomiting.13 Patients without any alarm symptoms, but with dyspepsia, were classified as presenting with uncomplicated dyspepsia.

A false-negative endoscopy within 1 year of definitive diagnosis of gastric cancer was considered a clinically relevant measure of poor quality endoscopy, so endoscopy reports preceding diagnosis and the notes of all patients were scrutinised to audit quality and determine factors associated with failure to establish the diagnosis at the initial endoscopy.

Statistical analysis

The Wilcoxon Mann–Whitney U test was used to evaluate continuous, non-parametric variables, and the Wilcoxon signed rank test was used when these were paired observations. The χ2 test was used to compare proportions of categorical variables, and the McNemar test for paired categorical data. SPSS V.15.0 was used for data analysis. Statistical significance was considered as p≤0.05, and a Bonferroni correction was made for multiple comparisons.



From January 2005 to March 2008, a total of 74 (0.8%) patients were identified with GC (median age 76 years (range 35–95)). The proportion with EGC was 9/74 (12%) (median age 77 years (range 68–88); seven were male and two female, ns). The remaining 65 (88%) had AGC (median age 76 years (range 35–95); 48 were male and 17 female). There was no significant difference between EGC and AGC with regard to age at diagnosis (p=0.48), but AGC was significantly more common in men (male/female ratio 55/19=2.82; p=0.0008). Sixty-nine of the 74 (93%) patients were of British origin.

Symptoms at the time of diagnosis

Alarm symptoms were present in 68/74 (92%) of all GC cases. Uncomplicated dyspepsia was the presenting symptom in 6/74 (8%), with no difference between AGC and EGC. The dominant alarm symptoms were anaemia (34/68; 50%), dysphagia (16/68; 24%), melaena or haematemesis (8/68; 12%), weight loss (8/68; 12%) and persistent nausea or vomiting (2/68; 2%).

Number of OGDs in all patients with GC

A total of 103 OGDs were performed in the 74 patients (88 in the 65 patients with AGC and 15 in nine with EGC), with a median 1 (range 1–4) for AGC and 2 (range 1–3) for EGC (ns). During the period of study, 9764 OGDs were performed at the endoscopy unit.

Number of OGDs before the date of diagnosis

A prior OGD had been carried out in 19 (26%) of the 74 patients with GC (total 29 OGDs, 23 in 14 patients with AGC and six in five patients with EGC). Ten (53%) of the 19 had undergone an endoscopy ≤1 year before definitive diagnosis, 5/19 (27%) within 1–2 years, 1/19 2–3 years before, 2/19 >3 years before diagnosis, and 1/19 under endoscopic surveillance. For those who had had an endoscopy within a year of diagnosis, all were planned follow-up OGDs (and all were performed within 6 months) for suspicious lesions (7/14 AGC and 3/5 EGC). No patient had an endoscopy between 6 and 18 months before diagnosis. Examining the time period up to 2 years before the diagnostic OGD, 5/19 patients with GC (4/14 AGC and 1/5 EGC) had an OGD. In two of the four who had AGC diagnosed within the subsequent 2 years, the OGD reported ‘no findings’, but a ‘polyp’ or ‘minor abnormalities’ were reported in the other two, as well as the patient subsequently diagnosed with EGC. Before 2 years before the diagnostic OGD, three patients had an OGD (all with AGC and 2/3 more than 3 years before diagnosis) that reported an ulcer, erosion or minor abnormalities. The remaining patient, who was known to have EGC, was under endoscopic surveillance.

This means that no patient had an OGD other than for a suspicious lesion within 18 months of diagnosis. On the other hand, 6/74 (8%; 5/65 with AGC and 1/9 with EGC) of all GC had an endoscopy that was reported as normal, an ulcer, erosion or ‘minor abnormalities’ within 3 years. These could be termed ‘missed cases’ if the period up to 3 years before diagnostic OGD is examined. The median survival of these patients was 15.5 months (range 1–26), in contrast with 8.5 months (range 5–26) for those with AGC who had not had a prior endoscopy. Consequently the sensitivity of OGD can be estimated as 68/74 (92%, 95% CI 85.7% to 98.1%).

Interval between index endoscopy and subsequent diagnosis of gastric cancer

The median interval between the index OGD and the date of diagnosis of GC in the 19/74 patients with a previous OGD was 227 days (range 7–2160). For patients with AGC, the median interval was 717 days (7–2160), and for EGC this was 89 days (range 49–150) (ns). There was also an ‘exceptional case’, which reflects the natural history of EGC: this patient had a small (<1 cm) gastric ulcer initially diagnosed in 1999. Because of his age at the time of diagnosis (87 years), it was left untreated, and by January 2005 the ulcer was 6 cm in diameter. The patient died in June 2005 from other causes and was captured in this cohort 1 January 2005 to 29 February 2008.

Diagnostic efficacy at the time of endoscopic diagnosis

To evaluate the diagnostic efficacy of the endoscopist, true-positive cases were examined. Of the 74 cases of GC, 68 were diagnosed at the index endoscopy, at presentation; the other six were potentially ‘missed’, although none within 18 months of diagnosis. Considering AGC, 50/65 were diagnosed by the endoscopist (endoscopic diagnosis before the confirmation of the histology report) and 12 suspected. As a result, 62/65 (95%) represent the true-positive cases. The true-positive case rate for EGC was 4/9 (45%).

Grade of endoscopist

Twenty-seven consultants and 47 middle-grade endoscopists performed the diagnostic OGD in the 74 patients with GC (24/41 consultants/middle grade for AGC and 3/6 for EGC). Fifteen consultants and 14 middle-grade endoscopists performed prior OGDs; there was no association between grade at index endoscopy and subsequent correct diagnosis (figure 1).

Figure 1

Patients with early gastric cancer initially missed. Comparison of grade at prior endoscopy and grade at diagnosis (McNemar test). There is no association between training grade and correct diagnosis.

Tumour distribution in the stomach

Altogether, 28/74 (38%, 16 Siewert type II) GCs were located in the proximal, 27/74 (36%) in the middle, and 17/74 (23%) in the distal stomach; the remaining 2/74 (3%) were in the gastric remnant (post-gastrectomy). Similar patterns of distribution were found in EGC and AGC.

Number of biopsies

The median number of biopsies performed during diagnostic OGD for GC without a previous OGD was 6 (range 3–12). There was no significant difference between AGC (range 3–12) and EGC (range 4–9). On the other hand, the median number of biopsies performed at a previous OGD in those who subsequently turned out to have GC was 2 (range 0–10) (p=0.002). The median was 0 (range 0–10) for AGC and 4 (range 2–10) for EGC (figure 2).

Figure 2

Number of biopsies taken at prior endoscopy (before) and at diagnosis (after); the difference is significant (p=0.002).

Histological characteristics

Of the 65 patients with AGC, 42 (65%) had poorly differentiated adenocarcinoma, 13 (20%) moderately differentiated, three well differentiated, and seven unclassified. In EGC, >60% had well or moderately differentiated adenocarcinoma and the remaining were intramucosal.


All patients with EGC (except the untreated case under surveillance) underwent surgical or endoscopic therapy and were alive at the time of case note review (median follow-up 16 months, range 5–39). The median survival of the 14 patients with AGC who had had a prior OGD was 5.5 months (range 1–27) from the diagnostic OGD, in contrast with 8.5 months (range 5–26, ns) for the 51 patients with AGC who had not had prior endoscopy.


The incidence of GC in this study was just 0.8% of 9764 upper gastrointestinal endoscopies over 3 years. This is consistent with a detection rate of <2% in other reports,7 14 15 and patient demographics were similar, although EGC was detected in only 12%.3 16 Numbers in the present study are small, but EGC is also uncommon in other European countries: between 3 and 9 a year were reported in five hospitals in Germany and Austria.17 More importantly for clinical practice, our study found a similar pattern of presentation of EGC and AGC. This is notable because there is a widely held assumption that the presentation of EGC resembles that of peptic ulcer disease, rather than AGC.3 6 16 This concept needs to be re-examined.

Uncomplicated dyspepsia occurs in up to 40% of the population, yet in those who come to endoscopy, the cause is GC in ≤2%.7 So what is going wrong with open-access endoscopy? There is no evidence that GC is characterised by uncomplicated dyspepsia.18 Two audits from the UK suggest that the large majority of patients (>92%) with upper gastrointestinal malignancy who present with uncomplicated dyspepsia have inoperable disease.19 20 Consequently uncomplicated dyspepsia is not so much an ‘early sign’ of GC, but a symptom in a minority of patients with GC. Most present with alarm symptoms (>90% of our patients).

Nevertheless, missing GC during OGD is quite common. The false-negative rate for the diagnosis of GC is a useful measure of quality for OGD.14 15 21 A 3-year period before the definitive diagnostic endoscopy appears to be relevant, based on a prospective study from Japan that reported a median duration of 37 months between endoscopic diagnosis of EGC and progression to AGC.22 Detection of GC up to 3 years after a negative OGD could be construed as a missed lesion.21 When 305 patients with oesophageal or GC who had had previous OGDs up to 3 years before their diagnosis (1994–2001) were reviewed, nearly 10% of patients with upper gastrointestinal malignancy had had at least one OGD within the defined period, and the overall miss rate for diagnosis was 7.2%.23 Another study from the UK examined 81 patients with GC and concluded that the false-negative rate was 16.6%.24 An Edinburgh study had a false-negative rate of 14% (18/129), when patients with GC within 2 years of their diagnostic OGD were reviewed.25 A study from Finland reported a false-negative rate of 4.6% (13/284) within 3.5 years before diagnosis.26 In a review of endoscopic screening in Japan, 6/26 patients with GC (23%) had had a negative endoscopy within the preceding 2 years.27 Similarly, when the diagnostic accuracy of 37 094 OGDs for the detection of GC between 1984 and 1989 was reviewed, the overall diagnostic accuracy of endoscopy was 99.6%, but the false-negative rate (155/814=19%) was notable. For these false-negative cases, the interval between the endoscopy and the definitive diagnosis of GC was <12 months for 38 examinations (38/155=25%), 1–2 years for 67 (67/155=43%), and 2–3 years for 50 (50/155=32%). The conclusion from this study was that 19% of GC was overlooked for up to 3 years after an initial endoscopy (sensitivity 81%).14 The same authors conducted another study in which 8364 patients had a negative OGD and 237 were GC. OGD was repeated in 3672/8364 patients 1–3 years from the initial OGD, and GC was diagnosed in 32/3672, which means that 73 cases of GC could be expected in the 8364 OGDs originally classified as negative. This gives a false-negative rate of up to 23% (73/310).15 21 In our study, there was a 0% false-negative rate in the 18 months before the diagnostic OGD, and 8% if the period was extended up to 3 years, which is relatively reassuring.

In this study the interval between initial and diagnostic OGD was longer for AGC than for EGC. A plausible explanation is that a longer period results in advanced stage disease. European studies have attributed the miss rate to the fact that middle-grade endoscopists assume most of the endoscopic burden,25 28 but this was not the case in our study. The question of whether early detection can change the outcome is intriguing. All patients with EGC in the present series had a favourable outcome (alive at last review), whereas patients with AGC had a dismal survival rate. It is questionable if earlier endoscopy could change this. The problem is that symptom pattern and duration did not differ between the two groups. Poorly differentiated adenocarcinoma dominated AGC in our patients, while well-differentiated morphology was most commonly seen in EGC. This is in accordance with other reports.29

Improvement of GC survival depends on early detection of the disease with OGD. As an operator-dependent procedure, room for improvement should include an adequate number of biopsies, even for minor mucosal abnormalities.30 Six specimens result in 100% accuracy of diagnosis,30 and this was the median number at the diagnostic OGD in the present study. A notable deficiency of the index OGD that failed to diagnose GC in the present series was an inadequate number of biopsies. Objective assessment of quality in endoscopy through a robust marker such as missed diagnosis of GC acts as a catalyst for improving practice. This audit of endoscopy outcomes in Oxford shows that 12 and 36 month false-negative rates for GC are a practical measure of quality that is directly relevant to patient care.

Main messages

  • Insufficient biopsies (fewer than six) of a lesion are the only factor associated with a gastric cancer potentially being missed by an initial endoscopy.

  • No patient had a false-negative endoscopy within 12 months of definitive diagnosis of cancer.

  • The clinical picture of early gastric cancer resembles that of advanced gastric cancer.

  • Gastric cancer almost always (>90%) presents with dyspepsia and alarm symptoms.

  • 12 and 36 month false-negative endoscopy rates for gastric cancer are a practical quality measure for audit purposes.

Current research questions

  • Is early gastric cancer a different entity from advanced gastric cancer, or a continuum of the same condition?

  • Does early endoscopic diagnosis of advanced gastric cancer alter the outcome?

  • Do patients with uncomplicated dyspepsia need an endoscopy?



  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.