Introduction Oesophagogastroduodenoscopy (OGD) is the investigation of choice for excluding upper gastrointestinal cancer (UGIC). However, published studies suggest 14% of UGIC subjects had an OGD that failed to diagnose cancer in the 3 years prior to diagnosis (post gastroscopy UGIC (PGUGIC)). We have investigated the rate and risk factors for PGUGIC in a national data set in England.

Methods Hospital Episode Statistics (HES) collate information on all NHS hospital attendances in England. Subjects undergoing OGD without an UGIC diagnosis 6–36 months before subsequent diagnosis were identified as PGUGIC cases (definitely missed – OGD without UGIC diagnosis 6–12 months prior to UGIC diagnosis; probably missed – OGD without UGIC diagnosis 12–36 months prior to UGIC diagnosis) and subjects with no OGD 6–36 months before diagnosis served as controls. The influence of personal and institutional variables on PGUGIC were examined by multivariate logistic regression.

Results HES records from 2001–2012 were analysed including 5826932 OGD in 4163023 subjects. 132075 subjects were diagnosed with UGIC. 5659 (4.3%) definitely missed PGUGIC cases and 8518 (6.4%) probably missed PGUGIC cases were found. Gastritis/duodenitis (2512 subjects, 17.7%) and gastric ulcer (2117 subjects, 15.0%) were the most common coded findings in PGUGIC cases. Emergency OGD was negatively associated with PGUGIC compared with day case OGD (OR 0.70 (95% CI 0.67–0.73), p < 0.001). Female gender (1.19 (1.1–1.2), p < 0.001), South Asian (1.32 (1.2–1.6), p < 0.001) and Afro-Caribbean (1.26 (1.1–1.5), p < 0.001) ethnicity and comorbidities (liver disease (3.05, (2.3–4.1), p < 0.001), severe liver disease (3.01 (2.1–4.2), p < 0.001), peptic ulcer (1.98 (1.9–2.1), p < 0.001), pulmonary disease (1.17 (1.1–1.3), p < 0.001)) were associated with PGUGIC. Subjects with PGUGIC were less likely to undergo surgery (0.76 (0.7–0.8), p < 0.001) or chemotherapy (0.49 (0.47–0.51), p < 0.001) than controls, however, this did not affect overall survival, which was similar to controls. There was a fourfold variation in PGUGIC rates between units. Unit volume did not affect the rate of PGUGIC (lowest tertile volume compared with highest tertile 0.97 (0.9–1.1), p = 0.5). The annual rate of PGUGIC did not change over the study period.

Conclusion The rate of PGUGIC up to 3 year prior to UGIC diagnosis was 10.7% in England between 2001 and 2012. PGUGIC was associated with an elective procedure, female gender, ethnicity and comorbidities. PGUGIC subjects were less likely to have surgery or chemotherapy, although there was no effect on overall survival. There were large variations in PGUGIC rates between units but no evidence of a volume effect.

Disclosure of Interest None Declared.