TL;DR

Statistics and epidemiology are among the most predictable, high-yield areas in MRCP Part 1. If you understand core definitions, know which measure fits which study design, and can spot classic examiner traps, this topic becomes a reliable source of marks. This guide distils 50 essential facts, the most tested subtopics, a worked mini-MCQ, and a practical revision checklist.

Why statistics & epidemiology matter in MRCP Part 1

Many candidates approach statistics with anxiety, yet MRCP examiners ask remarkably consistent questions year after year. The maths is basic; the challenge lies in interpretation. A clear grasp of principles—rather than memorising formulas in isolation—can quickly lift your score.

According to the official MRCP(UK) syllabus, statistics and epidemiology underpin evidence-based medicine and clinical decision-making, making them core knowledge for all physicians (https://www.mrcpuk.org).

Scope of the topic

In MRCP Part 1, statistics and epidemiology questions usually fall into five overlapping domains:

1. Measures of disease frequency

2. Measures of association

3. Diagnostic test accuracy

4. Bias, confounding, and study design

5. Screening principles and public health interpretation

The 5 most tested subtopics (and what examiners want)

1. Measures of frequency

Examiners test whether you understand what is being counted and over what time frame.

• Incidence: new cases over time

• Prevalence: existing cases at a point or period

• Relationship: prevalence ≈ incidence × duration (steady state)

2. Measures of association

These test interpretation rather than calculation.

• Relative risk (RR): cohort studies

• Odds ratio (OR): case–control studies

• Hazard ratio (HR): survival/time-to-event data

3. Diagnostic test accuracy

Questions often hinge on denominators and prevalence effects.

• Sensitivity and specificity are test properties

• Predictive values depend on disease prevalence

• Likelihood ratios combine sensitivity and specificity

4. Bias and confounding

Expect recognition questions rather than definitions alone.

• Selection bias

• Recall bias

• Observer bias

• Confounding vs effect modification

5. Screening principles

Frequently tested via interpretation of outcomes.

• Lead-time bias

• Length-time bias

• Overdiagnosis

50 high-yield facts you should know

A. Measures of disease frequency (1–10)

1. Incidence measures new cases only.

2. Prevalence includes all existing cases.

3. Point prevalence refers to a single time-point.

4. Period prevalence includes anyone with disease during the interval.

5. Incidence rate uses person-time.

6. Cumulative incidence assumes a closed population.

7. Prevalence increases with longer survival.

8. Screening increases prevalence without changing incidence.

9. Rare diseases can have high incidence but low prevalence.

10. Chronic diseases usually have high prevalence.

B. Measures of association (11–20)

11. RR compares risk between exposed and unexposed groups.

12. RR = 1 indicates no association.

13. OR approximates RR when outcomes are rare.

14. OR overestimates risk when outcomes are common.

15. Case–control studies cannot calculate incidence.

16. Cohort studies can calculate incidence and RR.

17. HR compares instantaneous risk over time.

18. HR assumes proportional hazards.

19. Absolute risk reduction is clinically meaningful.

20. Number needed to treat (NNT) = 1 / ARR.

C. Diagnostic tests (21–32)

21. Sensitivity = true positives / all with disease.

22. Specificity = true negatives / all without disease.

23. High sensitivity helps rule out disease (SnNout).

24. High specificity helps rule in disease (SpPin).

25. PPV increases as prevalence increases.

26. NPV increases as prevalence decreases.

27. Likelihood ratios are prevalence-independent.

28. LR+ >10 strongly supports disease.

29. LR− <0.1 strongly excludes disease.

30. ROC curves plot sensitivity vs 1 − specificity.

31. Area under the curve reflects overall accuracy.

32. Changing cut-offs trades sensitivity for specificity.

D. Bias and confounding (33–42)

33. Selection bias affects who enters a study.

34. Recall bias affects retrospective data collection.

35. Observer bias is reduced by blinding.

36. Measurement bias is systematic error.

37. A confounder is related to exposure and outcome.

38. Randomisation reduces confounding.

39. Stratification can control confounding.

40. Loss to follow-up biases cohort studies.

41. Intention-to-treat preserves randomisation.

42. Regression allows adjustment for multiple confounders.

E. Screening and public health (43–50)

43. Screening targets asymptomatic populations.

44. Disease must have a detectable latent phase.

45. Screening tests must be acceptable and affordable.

46. Lead-time bias inflates survival time.

47. Length-time bias favours indolent disease.

48. Overdiagnosis detects disease that would not cause harm.

49. Mortality reduction is the key screening outcome.

50. Screening programmes require quality assurance.

RR vs OR vs HR — quick comparison

Mini-MCQ (exam-style)

Question: In a cohort study, 12 of 300 exposed patients develop disease, compared with 6 of 600 unexposed patients. What is the relative risk?

Answer: Risk exposed = 12/300 = 0.04Risk unexposed = 6/600 = 0.01Relative risk = 4.0

Key exam point: Always identify the study design first—this tells you whether RR or OR is appropriate.

Common examiner traps (5 you must avoid)

• Using predictive values without considering prevalence

• Calculating RR from a case–control study

• Ignoring person-time when incidence rate is asked

• Missing lead-time or length-time bias in screening questions

• Confusing absolute and relative risk reductions

Practical study-tip checklist

• Learn definitions before formulas

• Identify study design before calculations

• Practise 2×2 tables until automatic

• Read denominators carefully

• Use question banks to spot wording traps

• Review mistakes weekly

For structured revision, see the MRCP Part 1 overview at👉 https://www.crackmedicine.com/mrcp-part-1/

Test these concepts using free MRCP MCQs:👉 https://www.crackmedicine.com/qbank/

And consolidate under exam conditions with mock tests:👉 https://www.crackmedicine.com/mock-tests/

FAQs

Is statistics heavily tested in MRCP Part 1?

Yes. It appears consistently and is one of the most predictable scoring areas.

Do I need advanced maths?

No. The exam focuses on interpretation and correct application, not complex calculations.

When should I use odds ratio instead of relative risk?

Use OR in case–control studies where incidence cannot be calculated.

Are likelihood ratios important for MRCP?

Yes. They are prevalence-independent and increasingly tested.

Ready to start?

You don’t master statistics by reading alone—you master it by applying concepts under exam pressure.

👉 Revise the full syllabus with the MRCP Part 1 hub:https://www.crackmedicine.com/mrcp-part-1/

👉 Practise high-yield statistics & epidemiology MCQs (with exam-style explanations):https://www.crackmedicine.com/qbank/

👉 Simulate the real exam and identify weak areas fast:https://www.crackmedicine.com/mock-tests/

👉 Strengthen core concepts with focused video teaching:https://www.crackmedicine.com/lectures/

Sources

• MRCP(UK) Examination Syllabus and Guidance – https://www.mrcpuk.org

• BMJ Statistics Notes – https://www.bmj.com/specialties/statistics-notes

• Altman DG. Practical Statistics for Medical Research. Chapman & Hall