As you might have noticed, I often use the Number Needed to Treat (NNT) to communicate research results. While I strive to write evidence-based but easy-to-understand articles, I believe the NNT is a research concept every GP should know.
- What is the NNT? It’s the number of patients you need to treat to prevent one bad outcome.
- How to calculate it? 1 divided by the “absolute risk reduction” (= the risk in the control group minus the risk in the intervention group).
For further explanations, see the University of Oxford or simply Wikipedia. Alternatively, you can watch this great and easy to understand 9-min-Video by Dr Roger Seheult from MedCram.
Now, some context: The NNT was first introduced in the NEJM in 1988 as a useful measure to communicate and prioritize medical treatments. However, a 2020 analysis of 875 clinical trials showed that only 9% reported an NNT.
Key points about NNT
- Time-wise variation: NNT can vary significantly when applied to different time periods (e.g., 1 year vs. 10 years). A shorter time frame usually leads to a higher NNT, and vice versa.
- Baseline risk: NNT depends on the average baseline risk of study participants, which can differ in real life patients. A lower baseline risk usually leads to a higher NNT, and vice versa.
- Comparison group: NNT is influenced by the specific comparator used, like a placebo or another treatment, which may strongly affect its magnitude.
- Confidence interval: NNT also has a confidence interval (e.g., NNT=10, 95% CI 5 to 15), which is often not reported.
- Similar Concepts: Number Needed to Harm (NNH) and Number Needed to Screen (NNS) are the equivalents for adverse effects and screenings.
My thoughts
NNT is a valuable tool for realistically assessing treatment effects but has limitations.
