Detection windows

How long do drugs stay in your system?

There is no single answer: how long a drug stays in your system depends on the substance, the dose, how often you use, and the test itself — blood, saliva and urine each look for something different. Published research does let us put ranges on it. Below is every substance we model: open a row to see how long it remains detectable, and why.

Detection times by drug, in detail

Open a row for the full picture: the window for each use pattern in each matrix, the molecule or metabolite the assay is actually looking for, the cutoff it is anchored to, what shifts it, and the published work behind it.

Cannabinoids 1

The only group here with a fat compartment. That second storage mechanism is why its chronic windows are measured in weeks while everything else is measured in days.

Cannabis (THC)

weed, hash, edibles
Cannabinoid 4–12 h → 2–7 d 1–4 d → 14–30 d

Cannabis is the outlier of the whole table, and for one reason: THC is fat-soluble. Every other substance here leaves the body on a single downward curve, while THC is stored in fat and released back into blood for days after the last joint. That is why its urine window is the only one measured in weeks, and why body fat — not dose — is the strongest personal predictor.

Blood

One-off dose
4–12 h
Occasional (≤ once a month)
6–24 h
Regular / daily
24–72 h
Chronic heavy
2–7 days

What the test measures THC itself. The model reports plasma, which runs roughly twice whole blood — it is a population estimate, not forensic evidence. Cutoff ~1–2 ng/mL (app default 2.0 ng/mL)

Saliva (oral fluid)

One-off dose
4–12 h
Occasional (≤ once a month)
6–24 h
Regular / daily
12–36 h
Chronic heavy
18–48 h

What the test measures THC deposited in the mouth and re-secreted into oral fluid. An edible leaves no oral deposit, so its saliva window is about half as long. Cutoff ~2 µg/L (SAMHSA confirmation / roadside swab)

Urine

One-off dose
1–4 days
Occasional (≤ once a month)
3–7 days
Regular / daily
7–15 days
Chronic heavy
14–30 days

What the test measures THC-COOH, the inactive breakdown product. It is fat-stored and clears far more slowly than THC itself, which is the entire reason urine outlasts blood by weeks. Cutoff 50 ng/mL (standard immunoassay)

What shifts these windows

  • Body fat is the single strongest personal predictor, because THC-COOH is stored in fat rather than simply diluted by it. At the extreme end the model puts a very heavy user with worst-case parameters near 43 days in urine.
  • The terminal half-life during abstinence runs about 3 days for occasional use, 4 for regular and 5 for chronic heavy use — Lowe 2009 reports 5 to 7 days.
  • Daily use builds a baseline that never fully falls away: about 1.2 ng/mL in plasma at 54 mg a day, about 3.0 ng/mL at 162 mg a day (Karschner 2009), with half of chronic users still above 1 ng/mL on day 7.
  • Method changes the peak, not the tail: a joint peaks near 100 ng/mL plasma at 54 mg, a vape at 10–25 mg reaches roughly 7–32 ng/mL, and an edible saturates around 3–6 ng/mL across 10 to 50 mg because most of it is converted on first pass.

The misconception

The most common mistake is reading a blood number as proof of anything legal. This model reports plasma, roughly double whole blood, and it is an estimate of a population curve rather than a measurement of you — while a positive urine test in a daily user says only that a fat-stored metabolite crossed 50 ng/mL, possibly weeks after the last use and with no effect left at all.

Get the Clearhead app — it estimates how long Cannabis (THC) stays in your system from the dose you take, how often you use, and your body. With a live countdown. →

THC driving limits by country →

Sources

Stimulants 3

Water-soluble and cleared in days, not weeks. Their long windows come from repeated dosing building a metabolite tail, never from a single big dose.

Cocaine

coke, crack
Stimulant 12–48 h → 1–3 d 2–4 d → 4–14 d

Cocaine itself is gone within hours — its plasma half-life is about one hour. Everything long about cocaine testing belongs to benzoylecgonine, the metabolite urine screens actually target. And the lever that moves that window is not how much you took but how many days sit between doses: two or three sessions a month behave almost exactly like a single dose.

Blood

One-off dose
12–48 h
Occasional (≤ once a month)
18–48 h
Regular (2–3× a month)
24–60 h
Weekly to daily
24–72 h

What the test measures Parent cocaine, fast (6–24 h), with benzoylecgonine carrying the rest of the window. Cutoff 10 ng/mL — the conservative low edge of the typical 10–50 ng/mL per-se band

Saliva (oral fluid)

One-off dose
4–24 h
Occasional (≤ once a month)
6–28 h
Regular (2–3× a month)
6–31 h
Weekly to daily
12–50 h

What the test measures Cocaine plus benzoylecgonine; BE is the longer signal (cocaine 4–13 h, BE roughly 24–31 h after a single dose). Cutoff 8 ng/mL (SAMHSA oral-fluid confirmation)

Urine

One-off dose
2–4 days
Occasional (≤ once a month)
2–4 days
Regular (2–3× a month)
2–5 days
Weekly to daily
4–14 days

What the test measures Benzoylecgonine, not cocaine. Its primary half-life is only about 6 hours; the long tail is a slow second phase that only daily dosing builds. Cutoff 150 ng/mL (federal immunoassay)

A research-grade assay at 5 ng/mL reaches 17 to 22 days in daily users. That is a warning about how much a cutoff decides, not a figure the app uses by default.

What shifts these windows

  • Low or atypical butyrylcholinesterase roughly doubles the parent half-life, and can push it towards threefold. Pregnancy, hormonal contraception, liver disease and malnutrition all lower this enzyme. In the app it changes the concentration curve, not the published windows.
  • Alcohol taken alongside cocaine forms cocaethylene in the liver, a separate metabolite with its own 2.5-hour half-life. The app models it as a distinct compartment and flags it as a cardiac danger; it does not carry its own detection window.
  • Body fat is deliberately not modelled as a reservoir here. Cocaine is not lipophilic enough for it to matter, and the app is explicit that the cannabis fat model must not be ported across.
  • Sex makes almost no difference: Mendelson 1999 found no pharmacokinetic difference in peak, half-life or exposure, with women showing roughly 5% lower butyrylcholinesterase. The model nudges the curve by that much and no more; the published windows are the same either way.

The misconception

"I took a lot, so it will show for longer." The size of a single dose barely moves the urine window at all. What moves it is dosing again within a few days, which lets benzoylecgonine accumulate into its slow phase — and that is why the app asks about days between sessions rather than about grams.

Get the Clearhead app — it estimates how long Cocaine stays in your system from the dose you take, how often you use, and your body. With a live countdown. →

Full page: how long cocaine stays in your system →

Sources

  • Ellefsen 2016 — oral fluid: cocaine 4–13 h, benzoylecgonine up to ~24–31 h after a single dose
  • Jufer 2000 — oral-fluid benzoylecgonine tail ~50 h in daily users; urine accumulation in chronic use
  • Preston 2002 — urine positive 10–14 days in chronic daily users (detox cohort, 0.5–2 g/day)
  • Nickley 2017 — research-grade 5 ng/mL assay reaches 17–22 days in daily users
  • Mendelson 1999 — no PK sex difference in Cmax, half-life or AUC; ~5% lower butyrylcholinesterase in women
  • Jeffcoat — intranasal absorption, absorption half-life ~11.7 min

Amphetamine

speed, pep, prescription dexamfetamine
Stimulant 12–48 h → 1–4 d 1–4 d → 2–7.5 d

Amphetamine is the clearest example of frequency mattering and dose barely mattering. The app does not grade it by "how often a month" at all — it flips between a single-dose window and a daily-use window, and then interpolates between them by typical daily dose. Someone who takes speed at the weekend is on the single-dose curve, not on the chronic one.

Blood

Single or episodic
12–48 h
Daily / dependent
1–4 days

What the test measures Parent amphetamine. Plasma half-life about 10 hours at neutral urine pH. Cutoff 20 ng/mL

Saliva (oral fluid)

Single or episodic
20–60 h
Daily / dependent
1.5–7 days

What the test measures Parent amphetamine, which concentrates in oral fluid at roughly 2.76 times plasma. Cutoff 25 ng/mL in the model (DRUID forensic 25 ng/mL, SAMHSA screening 50 ng/mL)

Urine

Single or episodic
1–4 days
Daily / dependent
2–7.5 days

What the test measures The amphetamine class as a whole — parent plus metabolites — on a standard immunoassay strip. Unlike cannabis these metabolites are water-soluble, so the window stays in days. Cutoff 500 ng/mL (immunoassay screen)

What shifts these windows

  • Urine pH is the biggest swing factor and the least known: the half-life runs from about 7 hours in acidic urine to about 34 hours in alkaline urine. The app publishes its windows at neutral pH and warns about the spread rather than modelling it numerically.
  • A heavier daily habit lengthens the window. Between the single-dose and the daily figures the model scales with the typical daily dose rather than jumping straight to the chronic number, so someone taking a small amount every day sits well below the 2–7.5 day chronic urine window.
  • Route changes the peak, not the clearance: insufflation raises absorption and peak, smoking raises them further, but neither touches the half-life or the window.

The misconception

"I use most weekends, so I am a chronic user." Not by this model. Amphetamine is water-soluble and does not accumulate the way THC does, so anything short of genuinely daily or dependent use sits on the single-dose window. The long saliva tail of up to seven days belongs to daily high-dose use and nothing else.

Get the Clearhead app — it estimates how long Amphetamine stays in your system from the dose you take, how often you use, and your body. With a live countdown. →

Full page: how long amphetamine stays in your system →

Sources

  • Dolder PC & Liechti ME 2017, Front Pharmacol 8:617 — 40 mg d-amphetamine, Cmax ~120 ng/mL, Tmax ~3 h
  • de la Torre R et al. 2004 — volume of distribution ~4 L/kg, half-life 6–12 h, urine-pH dependence (~7 h acidic to ~34 h alkaline)
  • Cone EJ & Huestis MA 2007 — oral-fluid pharmacokinetics, saliva/plasma ratio ~2.76
  • Verstraete AG 2004 — detection times across matrices
  • Andås 2016, J Anal Toxicol — chronic high-dose use extended oral fluid to ~8 days; the amphetamine model tops out at 7.5 days
  • SAMHSA / DHHS and DRUID — oral-fluid and urine cutoffs

Methamphetamine

crystal, meth, pervitin
Stimulant 24–48 h → 1.5–4 d 1–4 d → 3–8 d

Methamphetamine behaves like amphetamine with two twists. About 7% of it converts to amphetamine in the body, so it trips both lines on a test panel and the amphetamine signal appears later and lingers. And in daily users its saliva window runs up to eight days, matching urine — the opposite of what almost everyone expects from a swab.

Blood

Single or episodic
24–48 h
Daily / dependent
1.5–4 days

What the test measures Parent methamphetamine. Plasma half-life about 10 hours, oral bioavailability about 0.67. Cutoff 20 ng/mL

Saliva (oral fluid)

Single or episodic
24–72 h
Daily / dependent
1.5–8 days

What the test measures Parent methamphetamine, at roughly twice plasma concentration. Cutoff 25 ng/mL in the model (DRUID forensic 25 ng/mL, SAMHSA screening 50 ng/mL)

The eight-day figure comes from a detox cohort using 0.5–2 g a day. It is the daily-user number, and it runs as long as urine in that same cohort — a daily user should not read "a couple of days" as cleared.

Urine

Single or episodic
1–4 days
Daily / dependent
3–8 days

What the test measures Parent plus metabolites on a standard strip. Because roughly 7% converts to amphetamine, methamphetamine use trips both the mAMP and the AMP line. Cutoff 500 ng/mL (immunoassay screen)

What shifts these windows

  • Half-life is about 10 hours at neutral pH, but the literature range is 5 to 30 hours once urine pH is taken into account — the same acid-to-alkaline swing that governs amphetamine.
  • The same dose scaling applies as for amphetamine: the urine estimate climbs with the typical daily dose, and only a genuinely heavy daily habit reaches the full 3–8 day window.
  • Other metabolites, such as p-hydroxymethamphetamine, are not modelled — only the 7% conversion to amphetamine is.

The misconception

Almost everyone assumes urine is always the longest window. In daily methamphetamine use it is not the only one: oral fluid can stay positive for about eight days — as long as urine, and far longer than anyone expects from a swab. The reason the app models this explicitly is precisely so a daily user is not falsely reassured.

Get the Clearhead app — it estimates how long Methamphetamine stays in your system from the dose you take, how often you use, and your body. With a live countdown. →

Full page: how long meth stays in your system →

Sources

  • Schepers RJF et al. 2003 — controlled human dosing; still the gold-standard data for single-dose windows
  • Cook CE 1993 — half-life ~10–11 h, volume of distribution ~3.7 L/kg, oral bioavailability ~0.67
  • Cody JT 2002 — methamphetamine to amphetamine conversion (~7%)
  • Andås 2016, J Anal Toxicol — detox cohort 0.5–2 g/day, oral fluid to ~8 days, "never reported previously"
  • Cruickshank CC & Dyer KR 2009 review; StatPearls 2025 — re-validation of the older controlled-dose data
  • DRUID (forensic 25 ng/mL) and SAMHSA / DHHS (screening 50 ng/mL) — oral-fluid cutoffs
Empathogens 1

An empathogen is a drug whose main effect is emotional openness rather than stimulation. MDMA sits on its own here: dose moves its windows more than frequency does, because the enzyme that clears it saturates itself.

MDMA

ecstasy, molly, pills
Empathogen 8–24 h → 18–48 h 1–3 d → 1.5–4 d

MDMA is the one substance here where dose moves the window more than frequency does. The enzyme that clears it inhibits itself, so a bigger pill is disproportionately slower to leave — the app scales its windows by total dose and then caps them, because MDMA is bound by its own kinetics rather than by a fat reservoir. Frequency only nudges the upper edge.

Blood

One-off dose
8–24 h
Occasional (≤ once a month)
10–28 h
Regular (2–3× a month)
12–32 h
Weekly to daily
18–48 h

What the test measures Parent MDMA. Plasma half-life about 8 hours, peak around 2 hours. Cutoff 10 ng/mL

These are the windows at a 100 mg total dose. A stronger pill lengthens them and a weak one shortens them, but only within limits: MDMA is bound by its own saturating kinetics rather than by a fat reservoir, so dose cannot run the window away.

Saliva (oral fluid)

One-off dose
12–24 h
Occasional (≤ once a month)
12–28 h
Regular (2–3× a month)
12–32 h
Weekly to daily
18–48 h

What the test measures Parent MDMA, which concentrates strongly in oral fluid — a median of roughly 5.6 times plasma and a peak ratio far higher, because a basic drug gets stuck in acidic saliva and concentrates there — ion trapping. Cutoff 25 ng/mL confirmation (SAMHSA screening 50 ng/mL)

Urine

One-off dose
24–72 h
Occasional (≤ once a month)
1–3.3 days
Regular (2–3× a month)
1.3–3.8 days
Weekly to daily
1.5–4 days

What the test measures MDMA plus MDA on a workplace panel. A lab LC-MS confirmation can also read HMMA, which stretches the result towards five or six days — that longer figure is a laboratory scenario, not the screening window. Cutoff 500 ng/mL (SAMHSA immunoassay)

What shifts these windows

  • Dose is the dominant lever, and not proportionally: MDMA saturates the enzyme that clears it, so exposure rises faster than the dose does. A second dose within 24 hours raises the peak by around 29% and total exposure by around 77%.
  • Body weight is the only covariate a modern population PK analysis finds clinically relevant, and it scales on total weight rather than lean mass. Women are given about 10% longer half-life; age adds up to 15% but only above 50.
  • MDMA does not accumulate in fat the way THC does, which is why its ladder from occasional to heavy use is compressed rather than spread across weeks. Episodic use clears between sessions.
  • European pills average 130–180 mg but vary enormously, and often contain PMA, PMMA, 2C-B or methamphetamine instead. An unknown pill makes any window estimate an estimate about a substance you have not confirmed.

The misconception

A cheap "amphetamines" dipstick is widely assumed to catch ecstasy. Cross-reactivity is unreliable, and a basic strip can miss MDMA entirely — what actually flags it is a dedicated MDMA pad or a laboratory test. The reverse mistake is just as common: methamphetamine sold as ecstasy converts partly to amphetamine, so one pill can light up lines the person never expected.

Get the Clearhead app — it estimates how long MDMA stays in your system from the dose you take, how often you use, and your body. With a live countdown. →

Full page: how long MDMA stays in your system →

Sources

  • de la Torre R et al. 2004, Ther Drug Monit 26(2):137–144 — MDMA pharmacokinetics, non-linear above ~100 mg (CYP2D6 saturation)
  • de la Torre R et al. 2000, Br J Clin Pharmacol (N=14) — non-linear kinetics, CYP2D6 autoinhibition
  • Kolbrich EA et al. 2008, Ther Drug Monit (N=17) — controlled dosing, Cmax 162.9 at 77 mg and 291.8 at 123 mg, half-life 6.9–8.1 h
  • Farré M et al. 2004, Psychopharmacology (N=9) — second dose within 24 h: +29% peak, +77% exposure
  • Desrosiers NA & Huestis MA 2013, Anal Bioanal Chem (N=26); Navarro 2001 — oral fluid / plasma ratio, median ~5.6, peak far higher
  • Huestis MA 2024/25 population PK (N=65) — body weight the only clinically relevant covariate
  • Verstraete AG 2004 and SAMHSA — detection times and cutoffs
  • EMCDDA European Drug Report 2024 and DrugsData — European pill strength and adulteration
Depressants 1

One class name, wildly different molecules. A benzodiazepine window is meaningless until you name which benzodiazepine — the spread runs from hours to six weeks.

Benzodiazepines

diazepam, alprazolam, clonazepam, lorazepam and others
Depressant 6–24 h (clonazepam 1–2 d, diazepam 1–4 d) 1–4 d → 3–7 d (clonazepam 14–21 d, diazepam 14–42 d)

"Benzodiazepine" is a class name covering molecules that behave nothing alike. Five of the seven the app models clear in days. Diazepam and clonazepam do not: they are long-acting, they accumulate on daily dosing, and diazepam carries an active metabolite with a 90-hour half-life that drags its urine window out to six weeks. Quoting a single benzodiazepine window without naming the molecule is meaningless.

Blood

Short-acting (alprazolam, bromazepam, lorazepam, oxazepam, etizolam)
6–24 h
Clonazepam
24–48 h
Diazepam
1–4 days

What the test measures The parent drug. Blood is the only benzodiazepine matrix the app scales with dose; the ranges below are the floor, which the engine can extend but never shorten. Cutoff Per molecule, 5–50 ng/mL (alprazolam, clonazepam and etizolam 5; lorazepam 10; bromazepam 20; diazepam and oxazepam 50)

Saliva (oral fluid)

Short-acting, single or daily
2–24 h
Clonazepam — episodic
6–60 h
Clonazepam — daily
1–6 days
Diazepam — episodic
0.3–5 days
Diazepam — daily
2–7 days

What the test measures Parent drug, and for diazepam also nordiazepam. Benzodiazepines are heavily protein-bound and cross into saliva poorly — diazepam's oral fluid to plasma ratio is about 0.035 — so these windows are short relative to urine. Cutoff Per molecule, 2–10 ng/mL

The short-acting row is a range across all five short-acting molecules: alprazolam and lorazepam sit at 2–18 h, bromazepam, oxazepam and etizolam at 4–24 h. The per-molecule detail below carries the exact figures.

Urine

Short-acting — single
1–4 days
Short-acting — daily
3–7 days
Clonazepam or diazepam — single
3–7 days
Clonazepam — daily
14–21 days
Diazepam — daily
14–42 days

What the test measures A class immunoassay — and this is where the class name does real damage. The screen reflects diazepam and oxazepam reliably, is unreliable for alprazolam, clonazepam and lorazepam, is variable for bromazepam, and does not detect etizolam at all.

The app carries no numeric urine cutoff for benzodiazepines: these windows are fixed constants tied to standard screening practice rather than computed from a threshold, so we publish them without a cutoff instead of inventing one. The short-acting rows are likewise a range across five molecules: a single dose is 24–72 h for alprazolam, lorazepam and oxazepam, 48–96 h for bromazepam and 48–72 h for etizolam.

Per-molecule detail

  • Diazepam — Half-life 46 h, plus nordiazepam at 90 h. Blood 24–96 h, saliva 6–120 h single and 48–168 h daily, urine 3–7 days single and 14–42 days daily. The longest windows of the class.
  • Clonazepam — Half-life 35 h. Blood 24–48 h, saliva 6–60 h episodic and 24–144 h daily, urine 3–7 days single and 14–21 days daily. Accumulates, and a standard screen catches it unreliably.
  • Alprazolam — Half-life 11.2 h. Blood 6–24 h, saliva 2–18 h, urine 24–72 h single and 3–7 days daily. Does not accumulate — and a standard urine screen often misses it.
  • Lorazepam — Half-life 14 h, cleared by glucuronidation rather than by CYP enzymes, so age barely moves it. Blood 6–24 h, saliva 2–18 h, urine 24–72 h single and 4–7 days daily.
  • Oxazepam — Half-life 7 h, the shortest of the group after etizolam, also glucuronidated. Blood 6–24 h, saliva 4–24 h, urine 24–72 h single and 4–7 days daily. One of only two the class screen reflects reliably.
  • Bromazepam — Half-life 17 h — short enough that the model does not treat it as accumulating. Blood 6–24 h, saliva 4–24 h, urine 48–96 h single and 4–7 days daily.
  • Etizolam — A thienodiazepine sold on the designer market. Half-life 3.4 h with an active metabolite at 8.2 h. Blood 6–24 h, saliva 4–24 h, urine 48–72 h single and 4–7 days daily. A standard benzodiazepine screen does not detect it at all, and the app widens its uncertainty band accordingly.

What shifts these windows

  • Age matters for the CYP-metabolised benzodiazepines and barely at all for the glucuronidated ones. From middle age onwards diazepam and clonazepam lengthen markedly (Klotz 1975), alprazolam, bromazepam and etizolam less so, while lorazepam and oxazepam — cleared by glucuronidation, a simpler route that ageing barely slows — are almost unaffected.
  • Body fat lengthens the lipophilic ones. It weighs heaviest on etizolam and diazepam, and matters most in practice for diazepam, whose long metabolite tail it stretches further still — Abernethy 1982 measured desmethyldiazepam at 154 h in obese subjects against 57 h in lean ones. The glucuronidated pair barely moves.
  • Accumulation is not a matter of degree: between two daily doses a benzodiazepine either clears or it does not. Only diazepam and clonazepam are long-acting enough to build up; the rest are gone before the next dose. On daily dosing nordiazepam builds up nearly sixfold.
  • Renal and hepatic function are not modelled for benzodiazepines at all, unlike for opioids. Where the app has no lever, we say so rather than publish a number.

The misconception

A negative benzodiazepine screen is routinely read as "it has cleared". It very often is not. The standard class immunoassay catches alprazolam, clonazepam and lorazepam only unreliably and does not detect etizolam at all, so a negative result is not proof that nothing is there. In the other direction, diazepam's saliva window is often quoted as 48 hours; the app deliberately uses about five days for a single dose, because the assay also reads nordiazepam, whose 90-hour half-life has been measured at a median of 132 hours in oral fluid. A 48-hour figure would falsely reassure.

Get the Clearhead app — it estimates how long Benzodiazepines stays in your system from the dose you take, how often you use, and your body. With a live countdown. →

Full page: how long benzodiazepines stay in your system →

Sources

  • Greenblatt DJ & Abernethy DR — classic benzodiazepine pharmacokinetics; the source of the age and obesity levers in the model
  • Vindenes V et al. 2019 — nordiazepam in oral fluid, median 132 h at a 1 ng/mL limit of quantification after a single diazepam dose
  • Klotz U et al. 1975 — age and diazepam disposition
  • Abernethy DR et al. 1982, J Pharm Sci — prolongation of drug half-life due to obesity, studied on desmethyldiazepam (154 h obese vs 57 h lean) →
  • Friedman H et al. 1992; Jongen S & Vermeeren A 2018; Gorenstein C et al. 1994 — psychomotor and on-road impairment, and tolerance in dependent users
  • Front Toxicol 2025, DOI 10.3389/ftox.2025.1639890 — benzodiazepine immunoassay cross-reactivity
  • FDA product labels; ASAM 2025 tapering guidance; EUDA reporting on designer benzodiazepines
Opioids 1

The class where the test matters more than the drug. A standard opiate panel reflects morphine-like opioids and misses most of the rest entirely.

Opioids

heroin, morphine, codeine, oxycodone, fentanyl, methadone, buprenorphine
Opioid 6–24 h (buprenorphine 1–3 d, methadone up to 5 d) 1–3 d → 2–4 d (methadone, buprenorphine 3–7 d → 7–14 d; fentanyl to 14–28 d)

With opioids the question "how long does it stay" is less important than "will the test see it at all". A standard opiate panel reflects morphine-like drugs and misses fentanyl, methadone, buprenorphine, tramadol, tapentadol and nitazenes entirely, while catching oxycodone only unreliably. On top of that the class spans four orders of magnitude in potency, and three of its members — methadone, buprenorphine and fentanyl — have windows several times longer than the rest.

Blood

Codeine, dihydrocodeine, oxycodone, morphine, hydromorphone, heroin
6–12 h
Fentanyl bolus, tapentadol, tilidine, nitazenes
6–24 h
Tramadol
12–24 h
Methadone — single
12–48 h
Methadone — daily
2–5 days
Buprenorphine
24–72 h
Fentanyl — transdermal patch
2.5–5 days

What the test measures The active species rather than always the parent: codeine is measured through the morphine it produces, heroin through 6-MAM and then morphine, tramadol as tramadol plus O-desmethyltramadol, tilidine as nortilidine. Cutoff Per molecule, 0.1–50 ng/mL (nitazenes 0.1; fentanyl 0.5; buprenorphine 1; hydromorphone 5; codeine, morphine, oxycodone and heroin 10; dihydrocodeine and tilidine 20; tramadol, methadone and tapentadol 50)

Saliva (oral fluid)

Most opioids
24–48 h
Heroin and nitazenes
6–24 h

What the test measures The same active species as blood. The app models no separate window for daily use in oral fluid, so these figures apply to episodic and daily use alike. Cutoff Per molecule, the same 0.1–50 ng/mL band as blood

Urine

Most opioids — single
24–72 h
Most opioids — daily
2–4 days
Methadone, buprenorphine — single
3–7 days
Methadone, buprenorphine — daily
7–14 days
Fentanyl — daily or patch
14–28 days

What the test measures A standard opiate immunoassay reflects codeine, dihydrocodeine, morphine and heroin. It is unreliable for oxycodone and hydromorphone, and it does not detect fentanyl, methadone, buprenorphine, tramadol, tapentadol, tilidine or nitazenes — those need dedicated or targeted laboratory tests.

As with benzodiazepines, the app carries no numeric urine cutoff for opioids — the urine windows are fixed constants anchored to standard screening practice, so we publish them without inventing a threshold.

Per-molecule detail

  • Heroin — Never proven by the parent drug, which is gone in minutes. 6-MAM is the unique heroin marker and has a very short window; downstream morphine is what stretches urine to days. Saliva 6–24 h, blood 6–12 h, urine 24–72 h single and 2–4 days daily.
  • Fentanyl — A bolus clears in 6–24 h, but a transdermal patch keeps blood positive for 60–120 h because of a skin depot with an apparent 20-hour off half-life. Daily or patch use pushes urine to 14–28 days — the biggest single-versus-chronic gap in the class. A standard opiate panel does not see it.
  • Methadone — The only opioid the model treats as accumulating. A single dose clears in 12–48 h, but daily dosing builds a tissue reservoir that keeps blood positive for 2–5 days and urine for 7–14. Its half-life far exceeds its effect, so plasma can keep rising on a fixed dose and toxicity can appear on day 2 to 5; steady state takes about five days.
  • Buprenorphine — Half-life 38 h and a very low cutoff of 1 ng/mL. Blood 24–72 h, urine 3–7 days on a single dose and 7–14 days on daily use. Not seen by a standard opiate panel.
  • Codeine — A prodrug: what the model tracks is the morphine your CYP2D6 enzyme makes from it. A poor metaboliser produces almost none, an ultra-rapid metaboliser around 45% more than normal — which is why codeine has been fatal in children.
  • Nitazenes — Synthetic opioids from the new-substance market, modelled at a 0.1 ng/mL cutoff and the widest uncertainty band in the app. Saliva 6–24 h. Neither a standard opiate test nor a fentanyl test reflects them; they need targeted laboratory analysis.
  • Tramadol — Modelled as tramadol plus its O-desmethyl metabolite, half-life 6.7 h. Blood 12–24 h, urine 24–72 h single and 2–4 days daily. Not detected by a standard opiate panel.

What shifts these windows

  • Kidney function is the largest lever, and it acts through metabolites rather than the drug. Renal failure raises the half-life of the active morphine metabolite M6G about ten- to sixteen-fold (Osborne/Hanna 1993; Dean 2004), which is also the mechanism behind delayed respiratory depression. The model follows that literature: morphine, codeine and heroin are affected most, tramadol, dihydrocodeine and hydromorphone far less.
  • Body fat only matters for the lipophilic members — fentanyl, buprenorphine, methadone and nitazenes. Age slows clearance too: from about 50 the windows drift toward the top of their range.
  • CYP2D6 genotype changes how much active opioid your body makes from codeine, tramadol and dihydrocodeine. CPIC records roughly 96% less morphine from codeine in a poor metaboliser and about 45% more in an ultra-rapid one (Crews 2021). That moves the concentration curve, not the published detection window.
  • Route changes absorption and peak sharply — smoked and injected reach the highest peaks — but it does not change the detection window. It does change risk: the app weights respiratory risk by route precisely so that injecting does not read as the mildest option.

The misconception

The dangerous belief here is that a negative standard panel means nothing is present. It does not. Fentanyl and nitazenes require dedicated tests, methadone, buprenorphine, tramadol and tapentadol are not on an opiate panel at all, and oxycodone is caught only a fraction of the time compared with laboratory confirmation. The second misconception is about heroin: it is never proven by the parent drug, only by 6-MAM within a short window, and after that by morphine — which is indistinguishable from other sources.

Get the Clearhead app — it estimates how long Opioids stays in your system from the dose you take, how often you use, and your body. With a live countdown. →

Full page: how long opioids stay in your system →

Sources

What makes a detection window longer

1. Your pattern of use — the dominant factor

Every long window in that table is a frequency story, not a dose story. Elimination is broadly first-order — a fixed percentage falls away per unit of time — so doubling a single dose starts you higher on the same curve and adds hours. What adds days is dosing again before the previous dose has cleared, which lets metabolites accumulate into a slow secondary phase.

That is why the app asks whether use is episodic or daily rather than asking how many grams. Someone who uses cocaine two or three times a month behaves close to a single user — around 2 to 5 days in urine — while a daily user reaches 10 to 14 days. Cannabis breaks this rule in two directions at once: THC is stored in fat and released slowly, so it stacks a second, much longer mechanism on top of frequency.

2. Metabolite versus parent drug

Most of the surprise in these numbers disappears once you know that many tests do not look for the drug at all. Urine screens target metabolites: benzoylecgonine for cocaine, THC-COOH for cannabis, morphine and 6-MAM for heroin, nordiazepam for diazepam. Metabolites are usually more abundant and always slower to clear, which is the whole reason urine outlasts blood by days.

It also means a positive result carries less information than people assume. It says a breakdown product crossed a threshold. It does not say when, how much, or whether anything is still acting.

3. The cutoff of the test

A detection window is a statement about a threshold, not about your body. The same urine sample can be negative at 150 ng/mL and positive at 5 ng/mL. In a daily cocaine user that is the gap between 4 to 14 days on the standard 150 ng/mL screen and 17 to 22 days on a research-grade 5 ng/mL assay. Whenever you see two sources disagree wildly, an unstated cutoff is almost always the reason.

4. Individual biology

Enzyme activity, kidney function and urine pH all shift the curve. Low or atypical butyrylcholinesterase roughly doubles cocaine's parent half-life. Impaired renal function is the single largest lever in the model for morphine-type opioids; for other substances it matters pharmacologically, but Clearhead does not model it, so it is not in the table. Alkaline urine markedly extends amphetamine windows — its half-life runs from about 7 hours in acidic urine to around 34 hours in alkaline urine. Body composition matters mainly for cannabis, where fat is a storage compartment rather than just a volume.

How long each test can see it: blood, saliva, urine

  • Blood — the closest proxy for what is currently circulating, which is why it is the confirmatory sample in roadside and forensic settings. Short windows, and the only matrix where per se legal thresholds — where any result above a set number is the offence, impairment or not — are normally defined.
  • Saliva (oral fluid) — the practical roadside screen. It tracks plasma reasonably well, though for some drugs it concentrates well above plasma: MDMA sits at roughly 5 to 6 times plasma, cocaine at about twice, amphetamine at about 2.8 times. For daily methamphetamine use it runs as long as urine.
  • Urine — the workplace and clinical standard, and a history test rather than a recency test. Longest windows, metabolite-based, and blind to timing.

Hair testing extends to months but is not covered here or in the app: Clearhead does not model hair, and we would rather leave a gap than publish a number we cannot derive.

Detection is not impairment

These two timelines are frequently treated as one, and they are not. They can run in either direction. A stimulant crash impairs attention and reaction time for hours after blood levels fall below any cutoff. A daily cannabis user can test positive in urine for weeks with no effect whatsoever. A benzodiazepine can leave residual sedation into the next morning while the parent drug is already low.

So a window on this page answers the question "would a test show something". It never answers "am I fit to drive" — and in most EU countries the legal question for drugs other than cannabis is zero tolerance anyway, meaning any confirmed presence is the offence. A few, such as Romania since 2025, also require proof of impairment. The numeric-threshold approach exists mainly for THC.

Where these numbers come from

Every window on this page is read directly out of the Clearhead engine, so the web and the app cannot drift apart. Each substance model carries its own pharmacokinetic parameters and its own published sources, and each window is anchored to a named cutoff wherever the app defines one. Where it does not — the urine cutoffs for benzodiazepines and opioids, for instance — we say so in the row rather than borrow a threshold from elsewhere.

The same applies to matrices and molecules the app does not model: hair, 7-aminoclonazepam, EDDP. They are absent here because they are absent there, and a gap is more honest than a plausible-looking figure. Cocaethylene is a third case: the app does model it, but as a risk compartment with no detection window of its own, so it has no row.

Questions people actually ask

How long do drugs stay in your system?

There is no single number, because each test measures something different. Blood and saliva run from hours to a couple of days; urine runs 1 to 4 days after occasional use, and 2 weeks or more for cannabis and for daily cocaine, benzodiazepine or methadone use. The matrix, the cutoff and your frequency decide.

Which matters more, the dose or how often I use?

How often, by a wide margin. A bigger single dose starts higher on the same curve and buys hours, not days; dosing again within a few days lets metabolites accumulate, which turns a 2-day window into a 2-week one. Cannabis and MDMA are the two exceptions — fat storage for one, saturating kinetics for the other.

Why do urine tests detect drugs for so much longer?

Because most urine assays target metabolites, not the parent drug. Cocaine is measured as benzoylecgonine, cannabis as THC-COOH, heroin as morphine and 6-MAM, diazepam partly as nordiazepam. These breakdown products are produced in larger quantities and cleared more slowly than the drug itself, so the urine window outlasts the blood window by days, sometimes weeks.

Does a positive test mean I am impaired?

No, and this is the most consequential misunderstanding on the subject. Detection and impairment are separate timelines that only partly overlap. A urine test can be positive days after every effect has ended, while a stimulant crash can impair attention and reaction time when blood levels have already fallen below any cutoff. A test result is never a statement about fitness to drive.

Can drinking water, exercise or detox products shorten the window?

No. Elimination is governed by enzyme activity and renal excretion, and none of those are changed by water, sweat, coffee or commercial detox products. Drinking large volumes before a urine test dilutes the sample, which laboratories screen for using creatinine and specific gravity — a diluted specimen is generally flagged and rejected.

Why do published detection windows differ so much between sources?

Usually because they quietly assume different cutoffs and different use patterns. A urine window at a 150 ng/mL immunoassay cutoff and one at a 5 ng/mL research assay describe the same person and disagree by weeks. Any figure that does not name its matrix, its cutoff and its use pattern cannot be compared with any other.

Does a standard drug test detect every benzodiazepine and every opioid?

No, and this is where class names mislead badly. A standard benzodiazepine immunoassay reflects diazepam and oxazepam reliably, catches alprazolam, clonazepam and lorazepam unreliably, and does not detect etizolam at all. A standard opiate panel reflects codeine, dihydrocodeine, morphine and heroin, is unreliable for oxycodone and hydromorphone, and does not detect fentanyl, methadone, buprenorphine, tramadol, tapentadol or nitazenes. Those need dedicated or targeted laboratory tests.

Which test am I most likely to face?

Roadside checks in Europe are usually an oral-fluid screen backed up by a blood sample for confirmation, because blood is the matrix legal thresholds are written in. Workplace and clinical testing is overwhelmingly urine, which is a history test rather than a recency test. The two answer different questions, so the same person can be negative on one and positive on the other at the same moment.

Every figure on this page is read out of the Clearhead engine and reflects population ranges, not a measurement of any individual. It is not medical or legal advice, and it is never permission to drive. Reviewed 2 September 2026.

Estimate your own timing

Clearhead uses this same pharmacokinetic model to estimate — for your pattern, dose and body — when the effects are likely to have worn off. Private, offline, no account.

Get the app