Number Needed to Treat: What Common Treatments Actually Do

The Mental Model

NNT = 1 / absolute risk reduction

If untreated risk is 6% and treated risk is 4%, the absolute risk reduction is 2 percentage points. The NNT is 1 / 0.02 = 50. So 50 similar people need the treatment for one additional person to avoid that outcome.

Relative risk can sound larger than the real-world change. A fall from 6% to 4% is a 33% relative reduction, but only a 2-point absolute reduction. Both are true; only the absolute version answers "how many people like me?"

NNH uses the same math for harm. If a side effect rises from 1% to 3%, the absolute increase is 2 points and the NNH is 50.

ARR Absolute risk reduction: treated risk subtracted from untreated risk.

NNT How many need treatment for one extra person to benefit.

NNH How many treated before one extra person is harmed.

Worked statin example

TheNNT estimates five years of statins in primary prevention as NNT 104 to prevent one nonfatal heart attack and NNT 154 to prevent one stroke, with no identifiable mortality benefit and NNH 50 for new diabetes. A "~30% reduction" headline can still mean only about 1-2 people helped per 100 treated over five years.

Quick Reference: Common NNT and NNH Examples

Rows use patient-centered outcomes where available. "None identified" means the cited review did not find a statistically identifiable benefit or harm for that endpoint, not that no individual can ever benefit or be harmed.

Evidence grades: TheNNT color ratings or USPSTF grade where shown.

Best NNTs

Acute asthma steroids, sore-throat corticosteroids, cyclobenzaprine, flu vaccine in children, and peripheral nerve block for hip fracture all have NNTs in the 3-14 range for selected outcomes.

Baseline risk dominates

Warfarin for atrial fibrillation and BP medicines look much better in high-risk patients than low-risk patients. Always ask for your untreated risk.

Screening is not treatment

Mammography and PSA have number-needed-to-screen, false positives, and overdiagnosis. Do not compare screening NNS directly with treatment NNT.

Timeframe is part of the number

NNT 50 over 10 years is not the same as NNT 50 over 10 days. An NNT without timeframe is incomplete.

Intervention	Population / use case	NNT or benefit	NNH or harm	Timeframe	Grade	Context
Statins, primary preventionNo prior heart attack or stroke	Adults without known cardiovascular disease; baseline risk varies widely.	104 / 154 NNT 104 for nonfatal MI; 154 for stroke; no mortality benefit identified.	NNH 50 diabetes; NNH 10 muscle symptoms.	5 years	Red / contested	Read note The key lesson is not "never use statins"; it is that a relative-risk reduction can conceal a small absolute benefit in low-risk people. Source: TheNNT statins primary prevention.
Statins, secondary preventionKnown heart disease or prior stroke	People with established cardiovascular disease.	39-125 NNT about 83 death prevented, 39 MI prevented, 125 stroke prevented.	Harms not well quantified in that summary.	5 years	Green	Read note Same drug, different baseline risk: secondary prevention has larger absolute benefit than primary prevention. Source: TheNNT statins known heart disease.
Aspirin, primary preventionFirst heart attack or stroke prevention	People without known cardiovascular disease.	333 NNT 333 to avoid one nonfatal heart attack; no death benefit.	NNH 250 major bleeding.	5-7 years	Red; USPSTF C/D by age	Read note USPSTF says initiation is individual for ages 40-59 with 10%+ 10-year CVD risk and recommends against initiating at age 60+. Sources: TheNNT aspirin primary prevention; USPSTF aspirin 2022.
Aspirin, secondary preventionAfter prior MI or stroke	People with prior heart attack or stroke.	~1-50 Benefits outweigh harms in established disease; TheNNT reports benefits more common than harms.	Bleeding and GI harms exist; lower frequency than benefits in the cited high-risk group.	About 2 years in TheNNT summary	Green	Read note This contrast with primary prevention is a core NNT lesson: prior disease raises baseline risk and improves absolute benefit. Source: TheNNT aspirin after prior MI/stroke.
Blood pressure medicinesFirst-line antihypertensives	Adults treated to prevent death, MI, and stroke.	67-125 NNT 125 death prevented; 67 stroke; 100 heart attack.	NNH 10 medication side effects leading to stopping.	5 years	Benefit depends on risk	Read note Baseline cardiovascular risk changes the personal NNT. A high-risk patient can have a far lower NNT than a low-risk patient. Source: TheNNT blood pressure medicines.
Warfarin for atrial fibrillationPrimary stroke prevention	Non-valvular atrial fibrillation, no prior stroke.	25 / 42 NNT 25 stroke prevented; 42 death prevented.	NNH 25 bleeding; 384 intracranial hemorrhage.	Trial follow-up varied	Green	Read note Modern anticoagulants differ from warfarin, but the row teaches the high baseline stroke risk of atrial fibrillation. Source: TheNNT warfarin AF.
SGLT-2 inhibitorsType 2 diabetes	People with type 2 diabetes, stratified by baseline cardiovascular risk.	38-143 NNT 38 death prevented in high-risk people; 100 in low-risk; heart attack NNT 71 high-risk, 143 low-risk.	NNH 7 genital infection.	Varied across trials	Green	Read note This is a good example of why the table cannot replace clinical context: kidney disease, heart failure, cost, and tolerance all matter. Source: TheNNT SGLT-2/GLP-1 diabetes.
GLP-1 receptor agonistsType 2 diabetes	Type 2 diabetes; patient-centered outcomes and weight change.	Varies TheNNT reports benefit on several patient-centered outcomes compared with many diabetes medications.	NNH 17 severe GI event.	Varied across trials	Green	Read note Effect-size interpretation should separate diabetes outcomes from weight-loss outcomes. Source: TheNNT SGLT-2/GLP-1 diabetes.
MetforminType 2 diabetes	17,470 patients with type 2 diabetes from 14 studies.	Unknown TheNNT lists patient-centered benefit NNT as unknown.	NNH 6 GI events: nausea, vomiting, diarrhea, or abdominal discomfort.	Varied	Unclear magnitude	Read note Common first-line use does not mean a clean NNT exists for every endpoint. Source: TheNNT metformin.
SSRIs for depressionPrimary-care depression trials	Adults with unipolar depression in high-income primary-care settings.	7-8 Cochrane reports SSRI NNT 7-8 for clinical improvement versus placebo.	NNH 20-90 for withdrawal due to side effects.	Acute treatment trials; response may take up to 4 weeks.	Moderate evidence	Read note Severity, diagnosis, suicidality, therapy access, and side-effect tolerance matter. The same NNT can describe a modest average effect and still be very important to selected patients. Source: Cochrane antidepressants vs placebo.
Tight glycemic controlType 2 diabetes, intensive targets	Intensive versus conventional glucose targets.	250 NNT 250 limb amputation prevented; no identified death, MI, stroke, or kidney failure benefit.	NNH 6 severe hypoglycemia requiring hospitalization.	About 5 years	Red	Read note Surrogate endpoints such as A1c are not the same as outcomes patients directly feel. Source: TheNNT tight glycemic control.
Systemic steroids for asthma attackED acute asthma	Adults and children treated early during an asthma exacerbation.	8-11 NNT 8 hospital admission prevented; 10 relapse prevented; 11 later admission prevented.	No serious side effects identified in the summary.	ED treatment; relapse follow-up 7-21 days	Green	Read note A spectacular winner in the right setting; effect was larger in sicker asthma populations. Source: TheNNT systemic steroids asthma.
IV magnesium for acute asthmaModerate to severe adult exacerbation	Adults in the ED, usually after standard therapy.	14 NNT 14 hospitalization prevented.	Harms could not be fully aggregated; one large trial suggested NNH 30 for any adverse event.	ED visit / hospitalization decision	Green	Read note Not a home remedy; this is an ED adjunct for selected exacerbations. Source: TheNNT IV magnesium asthma.
Nebulized ipratropium for asthmaAdded to beta-agonist	Adults with acute asthma exacerbation.	11 NNT 11 hospitalization prevented.	No serious side effects identified.	ED exacerbation	Green / cautious	Read note The evidence is useful but heterogeneous; admission decisions were not always primary endpoints. Source: TheNNT ipratropium asthma.
Antibiotics for acute otitis mediaChildren	Children 2 months to 15 years in high-income countries.	7-33 No pain benefit at 24 hours; NNT 7-20 decreased pain after 24h; 33 tympanic membrane perforation avoided; 11 contralateral ear infection avoided.	NNH 14 vomiting, diarrhea, or rash.	2-12 days; 3 months for recurrence/hearing	Selective use	Read note This is why watchful waiting can be reasonable in nonsevere unilateral cases, while severe cases differ. Source: TheNNT otitis media antibiotics.
Corticosteroids for sore throatPresumed infectious sore throat	Adult and pediatric outpatients, often with usual care.	5 NNT 5 complete pain resolution at 24h and 48h; about 6h faster onset and 12h faster complete relief.	No one harmed in the summary; serious complications tracked.	24-48 hours	Green	Read note Strong symptom NNT, but the outcome is pain relief, not prevention of rare complications. Source: TheNNT corticosteroids sore throat.
NSAIDs for acute low back painNon-specific acute pain	Adults with acute non-specific low back pain, not sciatica or dangerous causes.	12-14 NNT 14 pain reduction; 12 disability reduction; 13 global improvement.	No increase in adverse events identified in included trials; trials may be underpowered for rare harms.	1 day to 2 months follow-up	Yellow	Read note The average pain difference was statistically significant but may be below a clinically important threshold. Source: TheNNT NSAIDs low back pain.
Cyclobenzaprine for low back painMuscle relaxant	Adults with uncomplicated low back pain.	3 NNT 3 global improvement by day 10.	NNH 4 drowsiness, dry mouth, dizziness, or nausea.	10 days; more effect in first 4 days	Benefit vs sedation tradeoff	Read note A low NNT and low NNH can both be true. Benefit may be front-loaded while side effects persist. Source: TheNNT cyclobenzaprine.
Gabapentin for neuropathic painPostherpetic neuralgia / diabetic neuropathy	Adults with selected chronic neuropathic pain indications.	6-8 NNT 8 postherpetic neuralgia; 6 diabetic neuropathy for 50% pain reduction.	NNH 8 dizziness; 11 somnolence; 13 ataxia; 21 edema.	Chronic pain trials	Selected indications	Read note Evidence is indication-specific; do not generalize to every pain condition. Source: TheNNT gabapentin.
BisphosphonatesPrior fracture or very low bone density	Post-menopausal women with prior fracture or very low bone density.	20 / 100 NNT 20 vertebral fracture prevented; 100 hip fracture prevented.	A small number harmed; exact NNH not cleanly quantified in TheNNT summary.	Trial-dependent, commonly years	Green in high-risk group	Read note Fracture-prevention NNTs are much better in people with prior fractures than in low-risk populations. Source: TheNNT bisphosphonates.
Vitamin D for fracture preventionCommunity-dwelling older adults	Older adults living in the community.	None No fracture-prevention benefit identified.	NNH 36 kidney stones or kidney damage.	Trial-dependent	Red	Read note The institutionalized frail-elderly case is different; setting changes baseline risk. Source: TheNNT vitamin D community.
Vitamin D for hip fracture preventionFrail institutionalized elderly	Frail elderly people in nursing-home/institutional settings.	36 NNT about 36 hip fracture prevented.	NNH about 36 kidney problems.	Trial-dependent	Context-specific	Read note Same supplement, different baseline risk. Source: TheNNT vitamin D institutionalized elderly.
Influenza vaccineHealthy adults, elderly, children	One influenza season; effectiveness varies by year, strain, and population.	5-71 NNV 71 healthy adults; 29 elderly; 5 children for preventing influenza.	NNH 125 fever in adults; serious neurologic harms not increased in cited review.	One season	Green; season-specific	Read note CDC emphasizes that flu vaccine effectiveness varies every season and by study design. Sources: TheNNT flu vaccines; CDC flu vaccine effectiveness studies.
Oseltamivir / zanamivirInfluenza treatment	Adults and children with suspected or confirmed influenza in trials.	Hours No hospitalization reduction; symptoms shortened by about 16.8h for adult oseltamivir and 14.4h for adult zanamivir.	Oseltamivir NNH 28 nausea, 22 adult vomiting, 19 child vomiting.	Acute illness	Modest symptom effect	Read note This is a good row for spotting surrogate-like outcomes: hours of symptom relief is not the same as preventing hospitalization. Source: TheNNT neuraminidase inhibitors.
Mammography screeningBreast cancer screening	Average-risk women; US recommendations differ from some non-US bodies.	NNS ~2000 Cochrane estimate: inviting 2000 women for 10 years prevents one breast-cancer death if 15% mortality reduction is assumed.	About 10 overdiagnosed/overtreated per 2000 invited in Cochrane estimate.	10 years invitation; USPSTF current recommendation biennial age 40-74	USPSTF B, age 40-74	Read note Screening math includes false positives, overdiagnosis, and lead-time bias. Sources: Cochrane mammography; USPSTF breast cancer screening.
PSA screeningProstate cancer screening	Men aged 55-69 considering PSA screening.	~769 USPSTF: about 1.3 prostate-cancer deaths prevented per 1000 men screened over about 13 years; about 3 metastatic cases prevented per 1000.	False positives, overdiagnosis, biopsy harms, and treatment harms; no all-cause mortality reduction shown.	About 13 years	USPSTF C age 55-69; D 70+	Read note USPSTF recommends individualized decision-making for 55-69 and against PSA screening at 70+. Source: USPSTF prostate cancer screening.
Routine annual health checksGeneral health checks	Adults invited to routine health checks.	None No mortality benefit identified.	Unknown number harmed by unnecessary testing and treatment side effects.	Varied	Red	Read note This does not mean targeted screening or symptom-driven visits are useless. It means broad routine checks did not show mortality benefit. Source: TheNNT routine health checks.
Vitamin D to prevent respiratory infectionsARI prevention	Broad trial populations; updated evidence as of May 2025 in TheNNT summary.	None No meaningful prevention benefit identified after newer studies.	Harms may be underreported; high-dose vitamin D linked to falls/dizziness concerns in older adults.	Varied	Red	Read note The row is dated because the evidence summary explicitly references May 2025 guideline status. Source: TheNNT vitamin D for ARI.
Peripheral nerve block for hip fractureInpatient peri-operative care	Adults admitted with acute hip fracture awaiting repair.	7 / 14 NNT 14 delirium prevented; 7 chest infection prevented; pain on movement reduced by 2.5/10.	Harms not reported in the summary.	Peri-operative hospitalization	Green	Read note A strong example of a procedural intervention with patient-centered outcomes beyond pain score. Source: TheNNT peripheral nerve block.
tPA for acute ischemic strokeThrombolysis	10,431 patients in 26 randomized trials in TheNNT summary.	Uncertain TheNNT reports good outcome, symptomatic intracranial hemorrhage, and mortality as uncertain/not reported NNT.	Uncertain/not reported NNH in that summary.	Acute stroke window; trial-dependent	Debated	Read note Urgency and eligibility dominate this decision; this row is for understanding uncertainty, not for delaying emergency care. Source: TheNNT tPA acute ischemic stroke.
Topical NSAIDsAcute musculoskeletal pain	Adults with acute sprains, strains, and similar musculoskeletal pain.	~4-5 Good pain-relief NNT in TheNNT/Cochrane topical NSAID summary; exact value differs by formulation and endpoint.	Mostly local skin reactions; systemic harms less common than oral NSAIDs in trials.	About 1 week in many trials	Green	Read note Use this row to distinguish local pain relief from systemic disease modification. Source: TheNNT topical NSAIDs.

Why Is Screening Different?

Number Needed to Screen

Screening does not directly treat anyone. It invites many people to testing so a smaller number can be diagnosed earlier and a still smaller number can avoid a serious outcome.

Example: Cochrane's mammography framing estimates 2000 women invited over 10 years to avoid one breast-cancer death, with about 10 overdiagnosed/overtreated.

Lead-time bias

Finding disease earlier can make survival time from diagnosis look longer even if death is not delayed.

Example: A cancer found 3 years earlier and dying at the same age creates 3 extra "survival" years on paper without extending life.

Overdiagnosis

Screening can find disease that would never have caused symptoms or death in that person's lifetime.

Example: PSA screening may prevent about 1.3 prostate-cancer deaths per 1000 men screened over 13 years, but it also creates false positives, biopsies, and treatment harms.

What Do I Ask My Doctor?

What is my absolute risk without this?
"Out of 100 people like me, how many have the outcome in 5 years if we do nothing?"

What is my absolute risk with this?
The difference between those two numbers is the treatment's absolute benefit.

What is the NNT for someone my age and risk?
A 45-year-old low-risk patient and a 72-year-old high-risk patient can have very different NNTs for the same drug.

What is the NNH and what harm counts?
Mild nausea, major bleeding, hospitalization, and death should not be treated as interchangeable harms.

What happens if we wait a month?
Some decisions are urgent; others are preference-sensitive and can tolerate a repeat measurement or lifestyle trial.

Is this treating a number, a symptom, or an outcome I care about?
Lowering A1c, LDL, or PSA is not automatically the same as preventing death, disability, pain, or hospitalization.

How to Read a Study Headline

Relative-risk spotting drill

Headline: "Drug reduces events by 30%."

Ask: 30% of what baseline? If untreated risk is 10%, treated risk is 7%, ARR is 3 points and NNT is 34. If untreated risk is 1%, treated risk is 0.7%, ARR is 0.3 points and NNT is 334.

Screening headline drill

Headline: "Screening reduces disease-specific mortality."

Ask: What is the all-cause mortality result, number needed to screen, false-positive rate, biopsy/procedure rate, and overdiagnosis estimate?

Common Mistakes and Anti-Patterns

Relative-risk seduction

A 30% relative reduction can be life-changing or trivial depending on baseline risk. Always convert to absolute risk.

"NNT 50 means it won't work for me"

NNT is population math. You cannot know in advance whether you are the one helped; the decision is about expected value and stakes.

Comparing unlike timeframes

NNT 10 over 48 hours, NNT 10 over 5 years, and NNT 10 over a lifetime are different claims.

Ignoring baseline risk

The same intervention may be compelling for high-risk secondary prevention and marginal for low-risk primary prevention.

Surrogate endpoint substitution

Better lab numbers matter only insofar as they translate into outcomes: symptoms, hospitalization, disability, cancer death, heart attack, stroke, or survival.

Forgetting harm severity

NNH 10 for mild muscle symptoms and NNH 250 for major bleeding should not be weighed as equal harms.

Treating screening as diagnosis

Screening starts a pathway. False positives and overdiagnosis are pathway harms, not footnotes.

Assuming "no evidence" equals "no effect"

A review may be underpowered, too heterogeneous, or aimed at the wrong population. Note the uncertainty instead of overclaiming.

Using averages for preference-sensitive choices

One person may strongly value avoiding a nonfatal MI; another may strongly value avoiding daily side effects. NNT informs the tradeoff; it does not choose the preference.

ER Triage - when the real question is urgency, not long-run efficacy.
Weight-Loss Levers - effect-size honesty for lifestyle interventions.

The spec also names future companion pages for risk dashboards and hospital-stay survival; link them here when those pages exist.

Primary Sources Used

TheNNT explained and per-intervention TheNNT summaries for NNT/NNH values.
Cochrane mammography screening review.
USPSTF breast cancer screening, prostate cancer screening, and aspirin primary prevention recommendations.
CDC seasonal flu vaccine effectiveness studies.
Oxford CEBM NNT definition.