Most IMGs know their cardiology and their pharmacology cold. Then a question about sensitivity and specificity appears, and the marks quietly disappear. That is frustrating, because these questions follow predictable patterns — and once you understand the logic, they reward you reliably.
Why Statistics Questions Feel Harder Than They Are
The honest answer is that medical statistics feels abstract after years of clinical training where you ordered the test, read the result, and made a decision. Nobody asked you to recalculate a likelihood ratio mid-ward round.
But PLAB 1 is not testing whether you can run a clinical trial. It is testing whether you can read research critically and apply evidence — a core GMC expectation under the principles of Good Medical Practice. The questions are therefore narrow and repeatable. A small cluster of concepts accounts for nearly all the statistics and evidence-based medicine marks in the exam: sensitivity and specificity, positive and negative predictive values, number needed to treat (NNT), number needed to harm (NNH), absolute and relative risk reduction, confidence intervals, and p-values. Master those eight ideas and you have covered the territory.
Sensitivity and Specificity: Get the Words Exactly Right
These two terms trip people up because they sound interchangeable until you anchor them properly.
Sensitivity is a property of the test itself, measured in people who have the disease. A highly sensitive test rarely misses true cases — a negative result is reassuring ("SnNout": a sensitive test, when Negative, rules Out). Think of a screening test such as D-dimer for pulmonary embolism: it is designed to be sensitive so that few cases are missed.
Specificity is also a property of the test, measured in people who do not have the disease. A highly specific test rarely flags false positives — a positive result is meaningful ("SpPin": a specific test, when Positive, rules In). A confirmatory test such as the VDRL for syphilis is designed to be specific.
Positive predictive value (PPV) and negative predictive value (NPV) depend on prevalence. This is the concept most often tested in a clinical vignette: the same test has a lower PPV in a low-prevalence population than in a high-risk clinic. If a question tells you the test has 95% sensitivity and 95% specificity but is used in a population where only 1 in 1,000 people have the disease, the PPV will still be low. Work through that logic once with real numbers and it stays with you.
Absolute Risk, Relative Risk, and NNT
Evidence-based medicine questions in PLAB 1 almost always pivot on the difference between relative and absolute risk reduction, because that difference matters enormously in clinical practice and in appraising drug trials.
- Absolute risk reduction (ARR) = risk in control group − risk in treatment group.
- Relative risk reduction (RRR) = ARR ÷ risk in control group, expressed as a percentage.
- Number needed to treat (NNT) = 1 ÷ ARR (where ARR is expressed as a decimal).
- Number needed to harm (NNH) uses the same formula but applied to adverse events.
A worked example: if a drug reduces the risk of a stroke from 4% to 2%, the ARR is 2% (0.02), so the NNT is 1 ÷ 0.02 = 50. The RRR is 50% — the same data, but a figure that sounds far more impressive. Drug advertisements tend to quote RRR; appraisal questions in PLAB 1 expect you to calculate NNT. A lower NNT is better (you need to treat fewer patients for one to benefit).
P-Values and Confidence Intervals Without the Maths Degree
You do not need to calculate a p-value in PLAB 1. You need to interpret one correctly.
A p-value below the conventional threshold of 0.05 means the result is statistically significant — that is, unlikely to have occurred by chance alone, assuming the null hypothesis were true. It does not mean the effect is clinically important, large, or causal. A massive trial can return p = 0.001 for an effect that is too small to matter in practice.
A confidence interval (CI) tells you the range within which the true value probably lies. For a ratio (relative risk, odds ratio, hazard ratio), a 95% CI that crosses 1.0 means the result is not statistically significant — the effect could be zero. For an absolute value (like a mean difference), a CI crossing zero has the same meaning. The question will usually ask you to identify whether a result is significant, or to pick the study whose CI suggests a genuine effect.
Study Design: Knowing Which Study Answers Which Question
PLAB 1 occasionally asks you to choose the most appropriate study design for a given clinical question. The short reference list:
- Randomised controlled trial (RCT) — best for treatment efficacy; reduces confounding.
- Cohort study — best for examining risk factors prospectively; gives relative risk.
- Case-control study — efficient for rare diseases; gives odds ratio, not relative risk.
- Cross-sectional study — prevalence at one point in time; cannot establish causation.
- Systematic review / meta-analysis — highest level of evidence for clinical decision-making, when well conducted.
If a question asks about the best evidence for a new drug's effectiveness, the answer is almost always an RCT or a systematic review of RCTs. If it asks about a rare cancer and its potential occupational exposure, a case-control study is usually correct.
How to Drill These Concepts Before Exam Day
Understanding the theory is half the work; applying it under timed conditions is the other half. These topics reward active practice over passive re-reading. Working through single-best-answer questions that present a 2×2 table and ask you to calculate sensitivity, or that give you trial data and ask for NNT, builds the pattern recognition that the exam requires. The Ant PLAB question bank includes a dedicated evidence and statistics cluster with worked explanations so you can see exactly where the logic goes wrong when you pick the incorrect option — and the performance analytics will flag if this blueprint area is consistently costing you marks.
The statistics cluster in PLAB 1 is small. Across a full sitting it may account for only a handful of questions. But those are questions with clearly correct answers, bounded by a limited set of concepts, that many candidates surrender without a fight. You have put too much work into this exam to leave marks on the table over a formula you can learn in an afternoon.
FAQ
What is the difference between sensitivity and specificity in simple terms? Sensitivity measures how well a test detects people who have a disease (true positive rate), while specificity measures how well it identifies people who do not have it (true negative rate). A highly sensitive test is good for ruling out a diagnosis; a highly specific test is good for ruling it in.
How do I calculate NNT from a clinical trial result in PLAB 1? Subtract the event rate in the treatment group from the event rate in the control group to get the absolute risk reduction (ARR). NNT = 1 ÷ ARR (with ARR as a decimal). For example, an ARR of 5% (0.05) gives an NNT of 20.
Does a p-value below 0.05 mean a treatment is clinically useful? Not necessarily. Statistical significance (p < 0.05) means the result is unlikely to be due to chance, but it says nothing about the size or clinical importance of the effect. A very large study can detect a statistically significant difference that is too small to benefit any individual patient.