Number Needed to Treat (NNT): A Deep Dive for MRCP Part 1

TL;DR;

Number Needed to Treat (NNT) tells you how many patients must receive a treatment to prevent one additional adverse outcome. For MRCP Part 1, examiners test not just the formula but your understanding of baseline risk, timeframes, and common misinterpretations. If you can move confidently between absolute risk reduction and NNT—and spot when NNT is being misused—you’ll pick up reliable marks across multiple specialties.

Why NNT matters in MRCP Part 1

Evidence-based medicine is woven throughout the MRCP Part 1 syllabus, and NNT is one of the most frequently examined concepts within it. Unlike relative measures, NNT converts trial data into a clinically intuitive number, making it ideal for testing judgement, counselling, and real-world applicability.

The exam expects candidates to understand what NNT means, how it is calculated, and—crucially—when it can mislead. This is why NNT appears not only in statistics questions, but also in cardiology, endocrinology, infectious diseases, and public health stems.

Authoritative guidance on assessment standards comes from MRCP(UK), which emphasises interpretation of evidence rather than rote calculation.

What exactly is Number Needed to Treat?

Number Needed to Treat (NNT) is the number of patients who need to be treated for a defined period to prevent one additional adverse outcome compared with a control group.

Formula (core knowledge):

• Absolute Risk Reduction (ARR) = Control Event Rate (CER) − Experimental Event Rate (EER)

• NNT = 1 ÷ ARR

Key rules:

• ARR must be expressed as a decimal, not a percentage.

• NNT is always rounded up to the nearest whole number.

• The timeframe of treatment must be stated.

High-yield points examiners love (read this twice)

1. NNT depends on baseline risk – identical relative risk reductions can produce very different NNTs in high- vs low-risk populations.

2. Lower NNT = greater absolute benefit, but only for the same outcome and timeframe.

3. Time matters – NNT over 10 years cannot be compared with NNT over 6 months.

4. Outcome-specific – NNT for mortality is not the same as NNT for symptom control.

5. Derived from ARR, not RRR – a classic MRCP Part 1 trap.

6. Always round up, even if the decimal is 1.01.

7. Confidence intervals can cross infinity, signalling uncertainty.

8. Large NNTs may still matter in population health.

9. Do not compare NNTs across diseases or different endpoints.

10. Harm has its own metric – Number Needed to Harm (NNH).

The five most tested NNT subtopics in MRCP Part 1

1. NNT vs Relative Risk Reduction (RRR)

Relative risk reduction often sounds impressive (“a 50% reduction”), but without baseline risk it can exaggerate benefit. MRCP questions frequently provide RRR alongside baseline risk to force you to calculate ARR and then NNT.

2. Baseline risk and population context

A statin may have an excellent NNT in secondary prevention but a much higher NNT in primary prevention. Expect stems contrasting post-MI patients with low-risk community cohorts.

3. Time-dependent NNT

An NNT quoted “over 5 years” reflects cumulative benefit. Shorter follow-up almost always increases NNT. Examiners may test whether you recognise that longer follow-up ≠ stronger drug effect.

4. NNT versus Number Needed to Harm (NNH)

Balanced judgement questions may ask which treatment offers net benefit. You are expected to consider both NNT and NNH, not one in isolation.

5. Patient counselling and shared decision-making

Some questions frame NNT as a communication tool. The best answers emphasise absolute benefit, not dramatic relative statistics.

A simple comparison table (exam-friendly)

Measure	What it shows	Why MRCP tests it
Relative Risk (RR)	Proportional change in risk	Can hide baseline risk
Relative Risk Reduction (RRR)	Percentage reduction vs control	Common source of overstatement
Absolute Risk Reduction (ARR)	Actual difference in event rates	Essential calculation step
Number Needed to Treat (NNT)	Patients needed to treat to prevent one event	Clinically intuitive
Number Needed to Harm (NNH)	Patients treated to cause one harm	Risk–benefit judgement

Practical example (MRCP-style mini-case)

Trial data:

• Control event rate (CER): 12%

• Experimental event rate (EER): 8%

• Follow-up: 4 years

Step 1 – Calculate ARRARR = 12% − 8% = 4% (0.04)

Step 2 – Calculate NNTNNT = 1 ÷ 0.04 = 25

Interpretation: You must treat 25 patients for 4 years to prevent one event.

Why this matters for MRCP Part 1:Many candidates stop at “33% relative risk reduction” and miss that the absolute benefit is modest. The exam rewards candidates who interpret the result clinically, not just mathematically.

Five common MRCP Part 1 traps

• Calculating NNT directly from RRR instead of ARR

• Forgetting to round up

• Ignoring the timeframe of treatment

• Comparing NNTs from different populations

• Assuming a low NNT automatically means “best treatment” without considering harm

A practical study-tip checklist

• □ Memorise ARR → NNT, not RR → NNT

• □ Always ask: over what time period?

• □ Look for baseline risk clues in the stem

• □ Check whether the question is about benefit, harm, or balance

• □ Use order-of-magnitude logic to eliminate wrong options quickly

FAQs

What is a good NNT?

There is no universal “good” NNT. It depends on baseline risk, outcome severity, timeframe, and associated harms.

Is NNT always rounded up?

Yes. Even an NNT of 1.1 becomes 2—partial patients do not exist.

Can NNT be negative?

No. If treatment causes harm rather than benefit, you calculate Number Needed to Harm (NNH) instead.

Why does MRCP Part 1 prefer NNT over relative risk?

Because NNT reflects absolute clinical benefit and avoids the exaggeration seen with relative measures.

Ready to start?

Ready to lock in Number Needed to Treat (NNT) for exam day?👉 Practise real MRCP-style questions now with our curated question bank and timed tests:https://crackmedicine.com/qbank/

For full exam simulation and performance tracking, start a complete mock here:https://crackmedicine.com/mock-tests/

Sources

• MRCP(UK) official site: https://www.mrcpuk.org

• Guyatt G et al. Users’ Guides to the Medical Literature. JAMA: https://jamanetwork.com

• Oxford Centre for Evidence-Based Medicine – Measures of treatment effect: https://www.cebm.ox.ac.uk/resources/ebm-tools/number-needed-to-treat-nnt