Statistics and evidence-based medicine questions on MRCP Part 1 are paradoxical: they are among the easiest to get right with specific preparation and among the most commonly lost without it. The concepts are finite, predictable, and learnable in a few focused sessions. Yet candidates who spend weeks on cardiology spend hours — or minutes — on statistics.
This is a mistake. Statistics questions appear in every sitting (typically 5-10 questions), the concepts do not change between sittings, and the marks are achievable with modest investment. These are the definition of "easy marks lost."
The Concepts That Appear Every Sitting
Sensitivity. The proportion of people with the disease who test positive. "If you have the disease, what is the probability the test catches it?" High sensitivity = good for ruling OUT disease (SnOUT — Sensitivity rules Out). Formula: True Positives / (True Positives + False Negatives).
Specificity. The proportion of people without the disease who test negative. "If you do not have the disease, what is the probability the test correctly says so?" High specificity = good for ruling IN disease (SpIN — Specificity rules In). Formula: True Negatives / (True Negatives + False Positives).
Positive Predictive Value (PPV). The probability that a person with a positive test actually has the disease. Critically, PPV depends on prevalence — the same test has a higher PPV in a high-prevalence population. Formula: True Positives / (True Positives + False Positives).
Negative Predictive Value (NPV). The probability that a person with a negative test genuinely does not have the disease. NPV increases with decreasing prevalence. Formula: True Negatives / (True Negatives + False Negatives).
Number Needed to Treat (NNT). The number of patients you need to treat for one additional patient to benefit compared to control. Calculated as 1 / Absolute Risk Reduction. Lower NNT = more effective treatment. NNT of 1 = every patient benefits. NNT of 100 = treat 100 patients for one to benefit.
Number Needed to Harm (NNH). Same concept for adverse effects. NNH = 1 / Absolute Risk Increase.
Absolute Risk Reduction (ARR). The difference in event rates between control and treatment groups. ARR = Control Event Rate - Treatment Event Rate.
Relative Risk Reduction (RRR). The proportional reduction in risk. RRR = ARR / Control Event Rate. Beware: RRR sounds more impressive than ARR. "50% relative risk reduction" might mean an absolute reduction from 2% to 1% — impressive-sounding but modest in absolute terms.
Odds Ratio (OR). Used in case-control studies. Approximates relative risk when event rates are low. OR > 1 = increased risk; OR < 1 = decreased risk; OR = 1 = no association. The confidence interval must not cross 1 for statistical significance.
Hazard Ratio (HR). Used in survival analysis. Interpretation similar to relative risk but accounts for time-to-event data.
Study Design Hierarchy
MRCP Part 1 tests whether you can identify study types and their appropriate applications.
Randomised controlled trial (RCT): Gold standard for intervention studies. Tests whether treatment A is better than treatment B. Minimises bias through randomisation and blinding.
Cohort study: Follows exposed and unexposed groups forward in time. Calculates relative risk. Prospective or retrospective.
Case-control study: Compares people with disease (cases) to those without (controls), looking backward for exposure. Calculates odds ratio. Efficient for rare diseases.
Cross-sectional study: Snapshot of a population at one point in time. Measures prevalence. Cannot determine causation.
Meta-analysis/systematic review: Synthesises data from multiple studies. Highest level of evidence when done well. Susceptible to publication bias and heterogeneity.
Bias Types
Selection bias: The sample is not representative of the population. Includes volunteer bias, referral bias, and Berkson's bias (hospital-based selection).
Information bias: Systematic error in measurement. Includes recall bias (cases remember exposures differently), observer bias, and detection bias.
Confounding: A third variable associated with both the exposure and outcome creates a spurious association. Managed through randomisation, matching, stratification, or multivariate analysis.
Lead-time bias: Screening appears to extend survival by detecting disease earlier, even if outcome is unchanged. The patient lives longer from diagnosis because diagnosis was earlier, not because they survived longer.
Length-time bias: Screening preferentially detects slow-growing, less aggressive disease (which has a longer detectable preclinical phase), making screening appear more beneficial than it is.
How to Prepare
These concepts are finite and learnable. Spend 2-3 focused sessions (total 6-8 hours) working through the definitions, formulae, and worked examples. Test yourself with Q-bank questions specifically filtered to statistics/EBM. The iatroX Q-Bank includes EBM questions with guideline-grounded explanations.
Once learned, use spaced repetition to retain them. The concepts are simple but easily confused under exam pressure (sensitivity vs specificity, PPV vs NPV). Regular brief revision — even 5-10 questions per week — keeps them fresh.
Do not waste marks on statistics. The concepts are learnable. The questions are predictable. The marks are free. Take them.