The critical review section of MRCPsych Paper B is the most feared — and the most learnable — part of the exam. Candidates who dismiss it until the last week tend to lose 20–30% of the paper. Candidates who spend 3 focused weeks on it often find it becomes their highest-scoring section.
This guide strips out the mathematical complexity and focuses on what you actually need to know — the concepts, the vocabulary, and the pattern recognition to answer exam questions confidently.
What Critical Review Actually Tests
The exam does not require you to perform complex calculations from scratch. It tests whether you can interpret research findings and apply statistical concepts to clinical scenarios. The key areas are:
✓ Study design hierarchy (RCT, cohort, case-control, cross-sectional)
✓ Bias and confounding — how to identify and minimise
✓ Sensitivity, specificity, PPV, NPV
✓ 2×2 table calculations
✓ p-values and what they mean (and don't mean)
✓ Confidence intervals — interpretation and width
✓ NNT and NNH — calculating from ARR
✓ Forest plots — how to read them
✓ Funnel plots — publication bias
✓ Common statistical tests (t-test, ANOVA, chi-squared, Mann-Whitney)
✓ Reliability and validity types
✓ Health economic concepts (QALY, cost-effectiveness)
The Core Concepts Explained Simply
Sensitivity vs Specificity
The two most tested concepts in the entire critical review section.
Sensitivity (SnOut)
How good the test is at catching true positives. High sensitivity = few missed cases. Used to rule OUT a condition.
Specificity (SpIn)
How good the test is at identifying true negatives. High specificity = few false positives. Used to rule IN a condition.
Mnemonic: SnOut — a highly Sensitive test, when Negative, rules it Out. SpIn — a highly Specific test, when Positive, rules it In.
p-values — What They Mean
A p-value is the probability of observing your result (or a more extreme one) if the null hypothesis were true.
- p < 0.05 means: this result would occur by chance less than 5% of the time — we call it "statistically significant"
- p < 0.05 does NOT mean: the treatment is clinically important, the effect is large, or the result is true
- p > 0.05 does NOT mean: the treatment does not work — it may mean the study was underpowered
NNT — Number Needed to Treat
NNT = 1 ÷ ARR (absolute risk reduction). It tells you how many patients need the treatment for one extra patient to benefit.
Example calculation:
Control event rate (CER) = 40% = 0.4
Experimental event rate (EER) = 25% = 0.25
ARR = CER − EER = 0.4 − 0.25 = 0.15
NNT = 1 ÷ 0.15 = 6.7 ≈ 7
Smaller NNT = better treatment. NNH (number needed to harm) uses the same formula but for adverse events.
Reading a Forest Plot
Forest plots appear in meta-analyses and are increasingly common in Paper B critical review questions.
✓ Each horizontal line represents one study — the width of the line shows the confidence interval
✓ The box in the centre of each line is the point estimate (the measured effect)
✓ The diamond at the bottom is the pooled result from all studies
✓ If the line (or diamond) crosses 1.0 for ORs/RRs, or 0 for mean differences — the result is NOT statistically significant
✓ A narrow diamond = precise estimate. A wide diamond = high heterogeneity
Study Design Hierarchy
Questions often ask you to identify the strongest or most appropriate study design for a given research question.
[Highest] Systematic review / Meta-analysis
Synthesises multiple RCTs. Strongest evidence for treatment efficacy.
[High] Randomised controlled trial (RCT)
Gold standard for causality. Randomisation controls for confounding.
[Medium] Cohort study
Follows people over time. Good for incidence, relative risk, and prognosis.
[Medium] Case-control study
Looks backwards from disease to exposure. Good for rare conditions.
[Lower] Cross-sectional study
Snapshot in time. Good for prevalence. Cannot establish causality.
[Lowest] Case report / series
Descriptive only. Useful for novel presentations, not causal inference.
How to Prepare for Critical Review
✓ Start at least 6 weeks before the exam — do not leave it to the last 2 weeks
✓ Learn the 2×2 table by heart: TP, FP, FN, TN and all four formulas from it
✓ Practice calculating NNT/NNH from abstract data — takes 30 seconds once automatic
✓ Do a forest plot and funnel plot interpretation question every day for 2 weeks
✓ Memorise the paired/unpaired test decision tree: parametric vs non-parametric, two groups vs multiple groups
✓ Use full critical review mocks (Psychgateway Mock 21 has 150 dedicated statistics questions)
► Psychgateway's Critical Review Mock (Mock 21) has 150 statistics questions with images — forest plots, funnel plots, Fagan nomograms, and ROC curves. Visit psychgateway.com