The Stat That Answers What Actually Matters
Number Needed to Treat (NNT) is elegantly simple: it's the number of patients you must treat to prevent one bad outcome or achieve one good outcome. NNT = 1 / absolute risk reduction. Understanding NNT transforms how you evaluate treatments.
Consider acute urinary tract infection treated with antibiotics. Studies show antibiotics resolve UTI symptoms in 80% of patients versus 20% spontaneous resolution—an absolute difference of 60%. NNT = 1/0.60 = 1.67, rounded to 2. Treat two UTI patients with antibiotics and one additional person (beyond natural resolution) recovers. This sounds fantastic because UTI treatment carries minimal harm and massive symptom relief.
Contrast this with primary prevention statins. Large trials (like WOSCOPS) show statins reduce heart attack risk from roughly 7.5% to 5% in low-risk men over five years—a 2.5% absolute reduction. NNT = 1/0.025 = 40. Treat 40 asymptomatic people for five years so one person avoids a heart attack. That person's identity remains unknown; you cannot predict who benefits. Some of the 40 experience muscle pain, liver enzyme elevation, or diabetes-like symptoms.
NNT contextualizes benefit. An NNT of 2 suggests strong benefit-to-harm ratio; NNT of 100 suggests the intervention suits only select populations. But NNT is half the story. You must also calculate Number Needed to Harm (NNH)—how many patients experience an adverse effect.
Consider a probiotic study: 200 subjects receive probiotic or placebo. Bloating resolves in 70% of probiotic group, 40% of placebo—a 30% absolute difference. NNT = 1/0.30 = 3.33, rounded to 3 or 4. But suppose adverse effects (diarrhea, gas) occur in 15% of probiotic group versus 5% placebo—a 10% absolute harm rate. NNH = 1/0.10 = 10. Now you're treating 3-4 people to benefit one, while harming 1 in 10. Different practitioners weigh this risk-benefit differently.
Microbiome-modifying interventions rarely report NNT/NNH comprehensively. Studies measure microbiota diversity changes, Firmicutes-to-Bacteroidetes ratios, or short-chain fatty acid production—all surrogate endpoints. But patient-centered outcomes (symptom relief, quality-of-life improvement, hospitalizations prevented) are what NNT addresses. As microbiome research matures toward clinical outcomes, NNT calculations will become essential.
Different conditions demand different thresholds. For infectious disease with high mortality, an NNT of 50 justifies treatment. For cosmetic conditions or symptom reduction, patients may tolerate only NNT of 5-10. Shared decision-making requires transparent NNT discussion.
Visualization helps. Imagine 10 stick figures. With NNT = 2, two figures light up (benefit achieved). With NNT = 50, only one figure out of 50 lights up. This visual clarity transcends percentages. Modern health communication increasingly uses these figures to improve understanding.
Forest plots in meta-analyses sometimes display NNT directly. When evaluating summaries of evidence, look for explicit NNT reporting. If absent, you can calculate it from absolute risk numbers provided.