This audit is for gastroenterology and hepatology trainees using BMJ OnExamination to revise for the ESEGH — the European Specialty Examination in Gastroenterology & Hepatology, which is the Specialty Certificate Examination (SCE) for the specialty. It addresses the written best-of-five papers only; there is no OSCE or viva to reproduce. The principal limitation to watch is that a self-selected difficulty feed and a raw home-screen percentage can drift a long way from the official 200-question blueprint, and can quietly reward recognition rather than recall.
What BMJ OnExamination offers for ESEGH right now
The figures below are vendor-reported and were last checked on 20 July 2026. Confirm the live numbers on the product page before you rely on any of them, because BMJ OnExamination changes access windows and pricing more often than it changes the bank.
- Bank size: 360+ ESEGH questions (vendor-reported).
- Question type: best-of-five, mapped to the curriculum areas published by the Royal College of Physicians, and written by gastroenterology and hepatology consultants (vendor-reported).
- Price and access: at the time of checking the ESEGH product was listed as free during an editorial review period, with subscription tiers from 1 to 12 months shown at £0. This is unusual and clearly temporary — treat any free window as a snapshot and verify the current price before planning around it.
- Features the vendor confirms: a "Select Questions" mode that lets you filter by difficulty and review flagged items; themed "Mock Tests" pitched at recent exam patterns; feedback with peer comparison; revision plans that target weaker areas; a "Group Learning" leaderboard mode; and an offline mobile app.
- Features the vendor does not clearly confirm for this product: a genuine adaptive-difficulty algorithm, a numeric percentile, a single "predicted score", a per-domain coverage tracker, or per-item timing analytics. If your readiness plan depends on any of these, open the live dashboard and confirm it exists before trusting it.
That last line matters. BMJ OnExamination is best described here as a consultant-written bank with difficulty selection, themed mocks and peer feedback — not as a proven adaptive engine. This audit treats it that way and makes no claim about a proprietary algorithm.
The ESEGH exam anchor
The ESEGH is delivered jointly by the Federation of the Royal Colleges of Physicians of the UK, the British Society of Gastroenterology (BSG) and the European Section and Board of Gastroenterology and Hepatology, aligned to the European curriculum (the "Blue Book", 2017 edition) for European candidates and the JRCPTB Specialty Training Curriculum for Gastroenterology for UK trainees. The structure is the standard SCE format: two papers of 100 best-of-five questions each, 200 questions in total, three hours per paper, sat on the same day. One mark is awarded per correct answer and there is no negative marking, so on a blueprint of this size a blank is simply a mark you chose not to try for. Delivery is computer-based through Surpass at a test centre, and from 2025 candidates have also been able to sit online-proctored from home or office (vendor/exam-body reported; verify on thefederation.uk).
The published blueprint is the target every analytics screen should be measured against:
| ESEGH domain | Questions | Share |
|---|---|---|
| Inflammatory bowel disease and colonic disorders | 40 | 20% |
| Liver disorders | 40 | 20% |
| Small intestinal disorders | 20 | 10% |
| Stomach and duodenal disorders | 20 | 10% |
| Biliary tree | 16 | 8% |
| Oesophageal disorders | 16 | 8% |
| Pancreatic disorders | 16 | 8% |
| GI haemorrhage | 10 | 5% |
| Nutrition | 10 | 5% |
| Other (mouth and salivary gland, endoscopy, GI physiology, investigations, symptoms and signs, anal disorders) | 12 | 6% |
Two domains — liver and IBD/colonic — carry 40% of the paper between them. Any revision pattern that under-weights hepatology or IBD is mis-aligned before you start, regardless of what the home screen average says.
Every metric on the screen, defined
An analytics-led bank shows a family of numbers, and each one answers a different question. Define them before you act on them.
First-attempt accuracy is the percentage correct on questions you have never answered before. It is the closest in-app proxy for unseen performance, and the only accuracy figure worth quoting to yourself. Repeat accuracy is the percentage correct on questions you have already seen; it rises with exposure and mostly measures memory of the item, not mastery of the topic. Percentile or peer comparison ranks you against other users of the same bank — a self-selected, motivated sub-population, not the ESEGH cohort, so read it as a coarse relative signal rather than a pass gauge. A predicted score, if the product shows one, is a model built on that same population and on repeated items; never convert it into a pass probability. Coverage is the proportion of the bank, or of a blueprint domain, you have attempted at least once — completion, not competence. Difficulty is the item's historic percentage-correct across users, which is why a "hard" filter changes who you are being compared against. Time per item is your pace; the ESEGH gives you roughly 108 seconds per question (100 questions in three hours), so a per-item timer is a genuine readiness input.
The discipline is to separate the two accuracies. If first-attempt accuracy is 58% and repeat accuracy is 84%, the honest reading is "58% on unseen material, and I remember items I have seen before" — not "improving to 84%".
Why difficulty-selected feeds distort your raw percentage
Because you choose the difficulty band and the topics in a Select Questions feed, your headline percentage is a product of your filters as much as your knowledge. A fortnight drilling only "hard" liver items will lower your average — not because you are getting worse, but because you have concentrated on the most self-selected corner of one domain; a fortnight re-doing flagged questions will raise it for the opposite, equally misleading, reason. Neither is comparable with a mixed, unseen, full-blueprint block, the only condition that resembles the exam. This is the trap described in your q-bank percentage is not your exam score: a within-bank average is an artefact of how you sampled the bank, not the quantity the examiner will measure.
A blueprint audit beats the home-screen average
Instead of reading the dashboard average, rebuild your attempts against the published weighting. Export or note your attempted count per domain and compare the share you have actually practised with the official share above. A worked snapshot — illustrative numbers only — shows how quickly the gap appears:
| Domain (target share) | Your attempts | Your share | Gap vs blueprint |
|---|---|---|---|
| Liver (20%) | 46 | 14% | under by 6 points |
| IBD and colonic (20%) | 78 | 24% | over by 4 points |
| Small intestine (10%) | 20 | 6% | under by 4 points |
| Biliary (8%) | 12 | 4% | under by 4 points |
| Nutrition (5%) | 6 | 2% | under by 3 points |
| Other incl. endoscopy (6%) | 4 | 1% | under by 5 points |
The home screen might report a comfortable 71% here, yet the candidate has under-practised the single biggest paper (liver) and barely touched nutrition and the endoscopy/investigations "Other" group. The average conceals the two things most likely to cost marks. Building this matrix once a week is the highest-value ten minutes in the whole workflow; the method generalises to any exam and is set out in question-bank completion is not coverage.
What a credible readiness signal requires
A number is only a readiness signal if it was generated under exam-like conditions. For ESEGH that means five things at once: the questions are unseen (first attempts, not repeats); the block is timed at roughly 108 seconds per item; the block is mixed across domains rather than filtered to one topic; you take it with no assistance — no notes, no looking up the answer mid-item; and the sample is large enough to be stable, which in practice means blocks of at least 50 and ideally a full 100-question paper before you read anything into the percentage. A 68% on a self-chosen easy single-topic set tells you almost nothing. A 68% first-attempt on a mixed, timed, unseen 100 is a signal you can act on.
Override rules: what to force into your feed
Adaptive-style or self-selected feeds systematically under-serve small, awkward domains, because there are fewer items and you avoid the topics you dislike. Force them in on a schedule rather than waiting for the feed to surface them. For ESEGH the recurring blind spots are: the endoscopy and investigations items hiding in the 6% "Other" band; nutrition (parenteral and enteral feeding, refeeding, micronutrient deficiency) at 5%; biliary and pancreatic management sequencing; and any question built on an image or a data set — liver function trends, ascitic fluid analysis, iron studies, a histology or endoscopy picture. If your bank lets you filter by these tags, set a standing weekly quota for each. If it does not expose them, write them on a checklist and self-audit. Guideline-sensitive topics — IBD biologics, hepatitis B and C treatment, variceal management — should be checked against current BSG, EASL and NICE/CKS positions rather than the explanation alone, and medicines detail confirmed against the SmPC/eMC.
Worked dashboard: turning analytics into next week's quotas
Take a hypothetical mid-revision candidate, six weeks out, with a mixed timed 100 giving 64% first-attempt. Domain first-attempt accuracy reads: liver 52%, IBD/colonic 70%, oesophageal 60%, nutrition 45%, biliary 58%, pancreatic 55%, everything else broadly at target. The instinct is to grind the highest-volume topic; the better move is to weight next week's quota by marks at risk, which is blueprint share multiplied by the accuracy gap.
- Liver carries 20% of the paper at 52% — the largest single pool of recoverable marks. Assign the biggest quota: 40 fresh items across cirrhosis complications, autoimmune and metabolic liver disease, and viral hepatitis.
- Nutrition is only 5% but sits at 45%; a small, cheap win. Assign 12 items and read one guideline summary.
- Oesophageal, biliary and pancreatic (8% each) at 55–60% get 15 items apiece.
- IBD/colonic is high-volume but already at 70%; hold it at a maintenance 15 items to keep it warm.
That is roughly 100 fresh questions for the week, weighted to marks at risk rather than to comfort — and deliberately with no pass prediction attached, because a single mixed block does not license one.
A seven-day pattern around clinical work
This loop uses BMJ OnExamination for one defined job — structured, blueprint-weighted practice with feedback — and iatroX for a second, different job: measuring whether that knowledge transfers to questions you have never seen. iatroX is not a gastroenterology-specific bank; its role is the free UK/MRCP-level knowledge and the unseen-question measurement layer that sits alongside a specialty SCE bank. Nothing here claims access to any proprietary algorithm.
- Monday: 25 BMJ items in the week's priority domain (liver), untimed, reading every explanation.
- Tuesday: 25 BMJ items in two smaller domains (nutrition, biliary).
- Wednesday: rest day or a short 10-item review of Monday's flags between clinics.
- Thursday: 25 BMJ items, timed, mixed across the week's domains.
- Friday: a fresh, timed, mixed unseen block in iatroX to check transfer — first-attempt accuracy is the only number you record.
- Saturday: a themed BMJ mock, timed, to rehearse pace and stamina.
- Sunday: rebuild the blueprint matrix, set next week's quotas, and re-space the questions you missed rather than the ones you got right.
The split matters because doing all your measuring inside the same bank you revise from lets recognition inflate the score. Rotating an unseen block through a second bank is the two-q-bank rule: a small volume of genuinely unseen items, used for measurement, not a second full syllabus to complete.
Decision checklist: continue, supplement, switch or stop
Base each decision on a measurable gap, not on novelty or on how much you have already paid.
- Continue BMJ OnExamination if first-attempt accuracy on mixed unseen blocks is rising and your blueprint matrix is filling evenly.
- Supplement with a second, unseen source when repeat accuracy has pulled well ahead of first-attempt accuracy — the classic sign you are memorising items — or when specific domains stay flat despite practice.
- Switch your primary bank only if a structured audit shows a domain the bank genuinely does not cover well, not because a competitor looks new.
- Stop adding volume when mixed, timed, unseen blocks have plateaued at a comfortable margin and your remaining errors are careless rather than knowledge gaps; at that point rest and light maintenance beat more questions.
Bottom line
BMJ OnExamination is a credible, consultant-written ESEGH bank with useful difficulty selection, themed mocks and peer feedback, and at the time of writing an unusually generous free window worth using. What it is not — on the public evidence — is a proven adaptive engine that manages your blueprint for you. Treat the home-screen average as a starting point, separate first-attempt from repeat accuracy, rebuild your attempts against the official 200-question weighting, and reserve your readiness judgement for mixed, timed, unseen blocks.
Frequently asked questions
Is BMJ OnExamination enough for ESEGH on its own? For many candidates a single well-covered bank plus the official curriculum can be enough to pass, and BMJ OnExamination's 360+ consultant-written items (vendor-reported, 20 July 2026) are a reasonable core. The risk is not volume but measurement: if you only ever practise and measure inside one bank, repeat accuracy flatters you. Add a small amount of unseen, mixed, timed practice from a second source to check transfer, and confirm your attempts cover the whole blueprint rather than your favourite domains.
Which ESEGH component does BMJ OnExamination not reproduce well? The ESEGH is entirely written best-of-five, so there is no OSCE or viva to reproduce — that is not the gap. The gap is within the written paper: image- and data-heavy items (endoscopy pictures, histology, liver-function and iron-study trends, ascitic-fluid interpretation) and the small "Other" and nutrition bands are easy to under-practise, and BMJ OnExamination does not publicly confirm a per-domain coverage tracker to catch that. You have to audit coverage yourself.
How many BMJ OnExamination questions should I complete per day for ESEGH? There is no single number, but a sustainable target for a trainee working clinically is 25 to 40 questions on active days, with the explanations read in full, rather than a larger number skimmed. Quality of review — why the four wrong options are wrong — predicts transfer far better than raw volume. Build the week around one timed mixed block and one blueprint audit rather than a flat daily quota.
When should I stop using BMJ OnExamination and move to mixed mocks? Move to predominantly mixed, timed, full-length mocks once your blueprint matrix is broadly even and your first-attempt accuracy on 50–100-question unseen blocks has stabilised. In the final two to three weeks, mixed full mocks that rehearse the roughly 108-second pace and the two-paper stamina are more useful than more single-topic drilling. Keep using BMJ's themed sets only to patch the specific domains a mock exposes.
How should I combine BMJ OnExamination with iatroX without duplicating practice? Give each tool one job. Use BMJ OnExamination as your primary ESEGH bank for structured, blueprint-weighted practice with feedback; use iatroX once or twice a week for a fresh, timed, unseen block whose only purpose is to measure whether that knowledge transfers to items you have never seen. Because iatroX is a general UK/MRCP-level bank rather than an ESEGH-specific one, it will not re-serve you the same specialty items, which is precisely what keeps the measurement honest.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 20 July 2026; all BMJ OnExamination figures (bank size, pricing, free window and features) are vendor-reported and change without notice — verify the current numbers on the product page before relying on them. Disclosure: iatroX operates a question bank and clinical-knowledge platform that competes with BMJ OnExamination; this audit confines iatroX's role to jobs BMJ OnExamination does not claim — general UK/MRCP-level knowledge and unseen-question measurement — and iatroX is not a gastroenterology-specific ESEGH bank. Corrections are welcome via the feedback route on iatrox.com.
References: the Federation of the Royal Colleges of Physicians of the UK ESEGH pages and ESEGH blueprint (thefederation.uk); the European Section and Board of Gastroenterology and Hepatology (eubogh.org); BMJ OnExamination ESEGH product page (onexamination.com); and internally, your q-bank percentage is not your exam score, question-bank completion is not coverage, the two-q-bank rule and the iatroX comparison hub.