Ordinary MCQ practice cannot, on its own, prepare you for three things in ESEGH: reading endoscopy images accurately enough to drive a management decision, interpreting liver data as a moving trend rather than a single value, and navigating the points where European and UK guidance diverge. A question bank measures whether you can pick the best of five options. It does not, by itself, build the visual fluency, the pattern-reading of laboratory and imaging data, or the jurisdictional awareness that a broad, timed, best-of-five specialty paper quietly rewards. This exam-level hub names those gaps and gives each one a concrete training route.
Banks are still essential for ESEGH — the exam is 200 best-of-five questions and you cannot pass it without large volumes of practice. The point is the ceiling of a bank used alone, and the specific skills that live above that ceiling.
The ESEGH format map
According to the Federation of Royal Colleges of Physicians (thefederation.uk, verified 21 July 2026), ESEGH is delivered as two papers, each three hours, each containing 100 best-of-five questions — 200 in total, five options per question, one mark per correct answer and no negative marking. It is computer-based, administered by Surpass at a test centre. UK candidates are examined against the JRCPTB Specialty Training Curriculum for Gastroenterology; European candidates against the UEMS "Blue Book" (2017). The 2026 fees were £700 (UK), €800 (ESBGH member countries) and £875 (elsewhere), and the exam is delivered jointly by the Federation and the European gastroenterology and hepatology bodies. The blueprint spans luminal and functional GI, hepatology, pancreaticobiliary disease, nutrition, endoscopy and relevant general medicine.
That format is single-best-answer throughout, which is precisely why the three skills below are under-tested by the format yet decisive at the margin: the exam builds image and data interpretation into text stems, and it uses guideline divergence as a discriminator between candidates who have merely memorised and those who understand. It is also an endurance test — 200 questions across two three-hour papers at roughly 1.8 minutes each — so an interpretive skill that is merely present is not enough; it has to be fast and reliable deep into the second paper, which is a further reason description-only practice falls short.
What a correct answer proves — and what it does not
A correct best-of-five answer proves you could recognise the best option among five, on this stem, often with the diagnosis already signposted. It does not prove you could have read the endoscopy image yourself if the finding had not been described in words, that you could interpret a liver panel as a trajectory rather than a snapshot, or that you know which guideline framework applies to your candidate cohort when European and UK practice differ. Recognition is not interpretation, and a single right answer on a described finding is not the same as the fluent reading the specialty demands. The three gaps below are where "I scored well in the bank" and "I can do this cold" separate.
The three modality gaps
Endoscopy images. ESEGH questions lean on endoscopic findings — a bleeding ulcer and its stigmata, the appearance and grade of oesophageal varices, dysplasia in Barrett's oesophagus, the mucosal pattern of inflammatory bowel disease, a polyp's morphology. Much of this is not free-text recall but graded interpretation: a peptic ulcer's Forrest classification changes whether you offer endoscopic therapy; the Los Angeles grade of oesophagitis changes follow-up; a polyp's Paris morphology changes resection technique. A text bank that writes "an actively bleeding vessel is seen" has done the grading for you and collapsed the decision. The exam may show the image and expect you to grade it yourself and then choose the intervention. The gap is visual recognition converted into action, and it is built only by volume of real images seen, interpreted and checked — not by reading descriptions of images.
Liver data. Hepatology is data-dense: liver-function panels, synthetic-function markers such as albumin and the prothrombin time or INR, non-invasive fibrosis scores, ascitic-fluid analysis with the serum-ascites albumin gradient, and composite scores such as MELD and Child-Pugh. The skill the exam rewards is reading these as a pattern and a trajectory — is synthetic function deteriorating over serial results, does this ascitic tap with a raised neutrophil count indicate spontaneous bacterial peritonitis, does a cholestatic versus hepatocellular pattern point you toward a different work-up — not reciting a single threshold in isolation. Banks often present the tidy, unambiguous version with one abnormal value flagged; real stems and real patients present several values that must be weighed together, and the decisive one is rarely the one in bold.
European versus UK guideline differences. ESEGH examines two candidate cohorts against two curricula, and gastroenterology guidance genuinely diverges — surveillance intervals for Barrett's and colorectal polyps, treatment thresholds in viral hepatitis and inflammatory bowel disease, screening approaches, and even terminology, such as the shift from NAFLD to MASLD/MASH. The gap is knowing which framework applies to your candidate route and where European (UEMS/ESGE/EASL-informed) and UK (NICE/BSG) practice part company. A candidate who has learned one framework as though it were universal will answer a threshold question confidently and wrongly, because the item was written precisely to separate rote memorisation from genuine understanding of why the guidance differs.
Turning each gap into a trainable skill
Give each gap an observable behaviour, a deliberate-practice task, a feedback source and an exit standard, so a vague weakness becomes something you can train and verify.
| Gap | Observable behaviour | Deliberate-practice task | Feedback source | Exit standard |
|---|---|---|---|---|
| Endoscopy images | Grade an unlabelled finding and select the intervention | Timed image sets with captions hidden; commit before revealing | Endoscopy atlases, ESGE guidance, a supervising gastroenterologist | Correct interpretation-plus-action on unseen images within time |
| Liver data | Read a panel/score as a trend and act on it | Serial data sets where you state the trajectory and next step | Hepatology references, EASL/UK guidance, clinician review | Reliable trend-reading and correct action on unseen data |
| Guideline differences | State which framework applies and the correct threshold | Side-by-side European/UK comparison cards for divergent topics | Current NICE/BSG and UEMS/ESGE/EASL sources, dated | Correctly attributes and applies the right framework cold |
The unifying rule is to hide the label and force interpretation. As soon as a stem or a bank hands you the reading, you are practising recognition again, not the skill that fails candidates.
A four-week modality ladder
- Week 1 — isolated skill. Drill each gap alone: captioned-hidden endoscopy image sets; serial liver-data exercises; European-versus-UK comparison cards for the highest-yield divergent topics. Method before speed.
- Week 2 — coached case. Integrated cases reviewed with a colleague or trainer — an endoscopic finding that drives management, a decompensating-liver trajectory, a surveillance-interval decision — with feedback on the reasoning.
- Week 3 — timed integrated case. Add the clock: mixed best-of-five blocks at exam pace (roughly 1.8 minutes per item) combining images, data and guideline-difference stems, all error-coded.
- Week 4 — unseen simulation. Full, timed, unseen blocks under exam-like conditions, then three-axis error coding by domain, cognitive process and format. This is the readiness read.
When to trust AI feedback — and when not
AI feedback is useful for the cheap, fast parts of the loop: explaining why a distractor is wrong, generating extra recall items on a threshold you keep missing, and drafting European-versus-UK comparison summaries you then verify against primary guidance. It is unreliable as the sole arbiter of endoscopy-image interpretation, where a confident wrong reading is the exact hazard you are trying to eliminate, and it should not be trusted to settle which guideline applies when frameworks diverge — that requires a dated, primary source. For image interpretation and for any judgement that turns on current jurisdiction-specific guidance, a gastroenterologist or the primary guideline is the required feedback source. The method for calibrating any automated score before trusting it is set out in how to calibrate AI-graded feedback. Let AI accelerate; let a clinician and the guideline adjudicate.
Build a balanced task matrix
Candidates drift toward the domains they enjoy, which is how a strong performer arrives at the exam with an untrained hepatology-data blind spot. Force balance: list the blueprint domains down one axis (luminal GI, hepatology, pancreaticobiliary, nutrition, endoscopy, malignancy/screening, general medicine) and the three modality gaps across the top, and require a minimum of recent, first-attempt practice in every cell. An empty cell is a planned weakness. Review it weekly and let the gaps, not your preferences, set the agenda — the same coverage discipline as the ESEGH content-gap checklist.
Red flags your practice has gone stale
- Memorised scripts. You recognise items rather than reason them; reviewed accuracy far exceeds fresh accuracy.
- Repeated cases. The same stems recur across sources; coverage has stopped growing.
- Generic feedback. Explanations no longer name a specific score, threshold or guideline.
- Uncalibrated scoring. You know your overall percentage but not your first-attempt accuracy by domain and by format.
- No official-rubric check. You have never sat official-format material under exam conditions, so you do not know whether your practice matches the live standard.
Two or more of these together mean the input needs changing: fresh unseen items, hidden-caption image work, serial-data exercises, or a clinician's eye on your reasoning.
Three mistakes this approach is designed to stop
The first mistake is treating a described finding as if you had interpreted it. When a stem hands you "grade 2 varices on a background of cirrhosis", you practise the management step but never the recognition that precedes it, and on exam day the image arrives without the caption. Hiding labels during practice forces you to generate the reading yourself, which is the skill actually tested.
The second mistake is reading liver data as isolated numbers rather than as a trajectory. A candidate who has drilled thresholds — this MELD, that neutrophil count — can still miss the story a series of results tells, because the decisive signal is often the direction of travel across several values, not a single flagged result. Serial-data exercises, where you state the trend before the answer, train the reading the exam expects.
The third mistake is learning one guideline framework as though it were universal. Because ESEGH examines against both European and UK curricula, a candidate who never noticed the divergence will apply a familiar threshold with false confidence. Building dated, side-by-side comparison cards for the divergent topics turns that hidden trap into a known, answerable question.
Frequently asked questions
How do I know whether I have covered the full ESEGH blueprint? You have covered it when a domain-by-modality matrix shows adequate, recent, first-attempt practice in every cell — each blueprint domain crossed with endoscopy images, liver data and guideline-difference reasoning — and when your accuracy on unseen, timed, mixed blocks is stable. Confirm the matrix against the official curriculum, not a bank's contents, and check at least once against official-format material under exam conditions. Coverage is an auditable matrix, not a finished question bank; a high overall percentage with an empty image or data cell is not coverage.
Can one question bank be enough for ESEGH? For core recall a single strong specialty bank can carry much of the load, but no text bank fully trains endoscopy-image fluency, liver-data trend-reading and European-versus-UK guideline discrimination at once. Most candidates pair the bank with image-and-data work and current primary guidance, and add a source of unseen questions late on. Decide by your matrix: if every cell is filled at exam standard from one source, one may suffice; if images, data or guideline nuance are thin, supplement rather than repeat the same bank.
What should I measure instead of my overall Q-bank percentage for ESEGH? Measure first-attempt accuracy on unseen, timed, mixed blocks, broken down by blueprint domain and by modality (endoscopy image, liver data, guideline-difference), plus your high-confidence-error rate and your retention after three to four weeks. The headline percentage blends fresh and reviewed questions and disguises the interpretive gaps that actually fail candidates. The breakdown on fresh questions is the number that predicts exam-day performance, as argued in Your Q-Bank Percentage Is Not Your Exam Score.
When should I stop doing new ESEGH questions? Stop adding new questions in a domain once its first-attempt accuracy on fresh items is stable and high and it has been reviewed recently. Stop new questions altogether only when every domain-and-modality cell is adequately covered, unseen timed accuracy is stable, and the binding constraint is stamina and pacing across two three-hour papers rather than knowledge. At that stage, switch to full timed simulation and to targeted image, data and guideline practice, rather than pouring more recall items into domains you have already mastered.
Which ESEGH resource should I use for my weakest component? Match the resource to the failure. For an endoscopy-image weakness, use image atlases and ESGE-informed material with captions hidden, checked by a gastroenterologist. For a liver-data weakness, drill serial data sets and scores against EASL/UK hepatology guidance. For a guideline-difference weakness, build dated European-versus-UK comparison cards from primary NICE/BSG and UEMS/ESGE/EASL sources. For transfer and retention across everything, use an unseen-MCQ layer with spaced retrieval — the job iatroX does as a general UK/MRCP-level knowledge layer rather than a specialty bank — to confirm fixed concepts hold on new stems. Choose by the coded weakness.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 21 July 2026. Exam-format facts are from the Federation of Royal Colleges of Physicians; any figures attributed to StudyPRN, RevisionPro SCE, Licence Medical or other providers are vendor-reported and were correct to the best of our checking on this date — verify current details on the provider's own page. Domain weightings reflect a published blueprint rather than a fixed per-topic count; confirm the current blueprint on thefederation.uk, and note that the applicable framework (European or UK) depends on your candidate route. Medicines facts should be taken from the SmPC/eMC (the electronic medicines compendium). Disclosure: iatroX operates a competing question bank; in this hub its role is confined to unseen-MCQ measurement and spaced retrieval, and it is explicitly not presented as a specialty-specific ESEGH bank, an endoscopy-image trainer or a clinical simulator. Corrections are welcome via the feedback route on iatrox.com.
References: the Federation ESEGH page (thefederation.uk); UEMS "Blue Book" gastroenterology curriculum (2017); ESGE guidelines and e-learning (esge.com); the ESEGH content-gap checklist; Your Q-Bank Percentage Is Not Your Exam Score; the iatroX comparison hub.