# What is drug half-life?

> Drug half-life is the time for a drug's plasma concentration to fall by half. US label values for 15 common drugs run from about 1 hour for amoxicillin to 58 days for amiodarone, and one drug can carry a terminal and an effective half-life that differ fourfold.

## What is drug half-life?

Drug half-life is the time it takes for the concentration of a drug in blood plasma to fall by half. In the US prescribing information for 15 widely used drugs, it runs from 61 minutes for amoxicillin through 1.5 to 2 days for digoxin to about 2 weeks for adalimumab and a mean of **58 days** for amiodarone. After about 5 half-lives a drug is mostly gone: amoxicillin within hours, amiodarone after most of a year.

The single number hides a choice of definition. Warfarin's label gives a terminal half-life of about 1 week after a single dose and an effective half-life of 20 to 60 hours, with a mean near 40, that tracks how levels actually rise and fall on repeated dosing. Both are correct. They measure different things, and a reader who takes the wrong one is off by a factor of four.

Two properties of the body set every half-life, clearance and volume of distribution. The rest of the spread between published values comes from which half-life a label reports and in whom it was measured. Every drug value below is a label's own, for its study population, and none is dosing guidance.

## How clearance and volume set half-life

Terminal plasma half-life is the time for the plasma concentration to halve once the drug has finished distributing through the body and levels are simply declining. It is not the time to eliminate half the dose, a common misreading. Most drugs leave the blood in at least two stages: a fast first phase moves drug into tissues, and a slower terminal phase reflects the body removing it.

Clearance (CL) is the volume of plasma wholly cleared of drug per unit time, in liters per hour. Volume of distribution (Vd) is the apparent volume needed to hold all the drug in the body at the plasma concentration, in liters. A drug that is cleared slowly or spreads into a large volume of tissue lingers, which is why Toutain and Bousquet-Mélou describe half-life as a hybrid of the two:

$$
t_{1/2} = \frac{\ln 2 \cdot V_d}{CL} \approx \frac{0.693\, V_d}{CL}
$$

The [semaglutide](/guides/semaglutide) label allows a check. It gives an apparent volume of distribution of about 12.5 L, an apparent clearance of about 0.05 L/h and an elimination half-life of about 1 week. The formula gives

$$
t_{1/2} = \frac{0.693 \times 12.5\ \text{L}}{0.05\ \text{L/h}} \approx 173\ \text{h} \approx 7.2\ \text{days},
$$

which matches the label's week. Semaglutide is more than 99% bound to plasma albumin, which fits its small clearance.

## How half-life is measured

The standard measurement gives a dose, usually into a vein so that absorption does not complicate the result, and samples blood over time. Plotted as the logarithm of concentration against time, the curve bends and then settles into a straight line, the terminal phase (Figure 1). The half-life follows from that slope, $\lambda_z$:

$$
t_{1/2} = \frac{\ln 2}{\lambda_z}
$$

![Schematic plasma concentration curve on a logarithmic axis: a steep distribution phase that bends into a straight terminal phase, with a dashed line extending the terminal phase back to the start and marks showing the concentration halving over one half-life](/images/guides/drug-half-life/terminal-phase.webp "**Figure 1. The terminal phase and one half-life.** Schematic of an intravenous dose, not data for any drug. On a log axis the terminal phase is a straight line with slope −λz, and the time for the concentration to fall from C to C/2 along it is the half-life, ln 2 divided by λz.")

Obach and colleagues compiled intravenous clearance, volume of distribution and terminal half-life for 670 drugs this way, drawing each value from original studies that dosed into a vein. The route matters because slow absorption can fool the measurement. When absorption is the slow step, the terminal half-life reflects how fast the drug is absorbed rather than eliminated, a situation called flip-flop kinetics.

Labels also use more than one kind of half-life. The terminal value describes the tail of the curve. An effective half-life describes how fast concentrations rise and fall over repeated dosing, which is the gap in the warfarin label. A functional half-life tracks the effect or active metabolites: atorvastatin has a mean plasma half-life of about 14 hours but inhibits its target enzyme for 20 to 30 hours because of active metabolites.

## Half-lives of common drugs

Figure 2 plots the label values of 15 drugs on a log scale, and Table 1 lists them from shortest to longest. We converted each reported value to hours, using 24 hours per day and 168 per week. The last column records what each label's number refers to, because the numbers are not all the same kind of quantity.

![Horizontal bar chart on a logarithmic hours axis of the label half-life of 15 drugs, from aspirin at about 1 hour to amiodarone at about 1,400 hours](/images/charts/drug-half-life-label-values-by-drug.webp "**Figure 2. Label half-lives of 15 common drugs, on a log scale.** Each bar ends at the label's headline value in hours: the stated mean or average where the label gives one, otherwise the upper end of the stated range, with the label's own wording printed beside it. Aspirin is the parent molecule only, diazepam and fluoxetine are the parent drugs (their active metabolites last longer), and warfarin is the terminal value, not the 20 to 60 hour effective half-life. Values are from the Table 1 sources.")

  <caption><strong>Table 1. Label half-lives of 15 common drugs, shortest first.</strong> Values are as printed in US prescribing information on DailyMed, accessed November 11, 2026; label versions are listed in the sources. The hours column is our conversion of the reported value. Labels differ in route, formulation and study population, so rows are comparable in order of magnitude only. Sources: DailyMed labels for Durlaza, amoxicillin, ibuprofen, Combogesic IV, caffeine citrate, metformin, Dilantin Infatabs, Lipitor, Valium, Lanoxin, Jantoven, Prozac, Ozempic, Humira and Pacerone (2015 to 2026).</caption>
  <thead>
    <tr>
      <th scope="col">Drug</th>
      <th scope="col">Label half-life</th>
      <th scope="col">Approximate hours</th>
      <th scope="col">What the value describes</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>Aspirin</td>
      <td>20 to 60 minutes</td>
      <td>0.3 to 1</td>
      <td>Mean plasma half-life of aspirin itself, before its conversion to salicylate; extended-release capsule</td>
    </tr>
    <tr>
      <td>Amoxicillin</td>
      <td>61.3 minutes</td>
      <td>1.0</td>
      <td>Oral chewable tablet; no population stated</td>
    </tr>
    <tr>
      <td>Ibuprofen</td>
      <td>1.8 to 2.0 hours</td>
      <td>1.8 to 2.0</td>
      <td>Serum half-life; oral tablet</td>
    </tr>
    <tr>
      <td>Acetaminophen (paracetamol)</td>
      <td>2 to 3 hours</td>
      <td>2 to 3</td>
      <td>Adults; shorter in children, longer in neonates and cirrhosis; injection label</td>
    </tr>
    <tr>
      <td>Caffeine</td>
      <td>5 hours</td>
      <td>5</td>
      <td>Adult-like metabolism, reached by about 9 months of age; neonates 3 to 4 days</td>
    </tr>
    <tr>
      <td>Metformin</td>
      <td>6.2 hours</td>
      <td>6.2</td>
      <td>Plasma elimination half-life after an oral dose; blood half-life 17.6 hours</td>
    </tr>
    <tr>
      <td>Phenytoin</td>
      <td>14 hours (range 7 to 29)</td>
      <td>14</td>
      <td>Average plasma half-life; saturable metabolism makes it dose dependent</td>
    </tr>
    <tr>
      <td>Atorvastatin</td>
      <td>about 14 hours</td>
      <td>14</td>
      <td>Mean plasma half-life; inhibitory activity lasts 20 to 30 hours</td>
    </tr>
    <tr>
      <td>Diazepam</td>
      <td>up to 48 hours</td>
      <td>up to 48</td>
      <td>Terminal phase; the active metabolite N-desmethyldiazepam up to 100 hours</td>
    </tr>
    <tr>
      <td>Digoxin</td>
      <td>1.5 to 2 days</td>
      <td>36 to 48</td>
      <td>Healthy volunteers with normal renal function; 3.5 to 5 days in anuric patients</td>
    </tr>
    <tr>
      <td>Warfarin</td>
      <td>about 1 week terminal; 20 to 60 hours effective</td>
      <td>168 terminal; 20 to 60 effective</td>
      <td>Single dose; the effective value (mean about 40 hours) tracks the anticoagulant effect</td>
    </tr>
    <tr>
      <td>Fluoxetine</td>
      <td>1 to 3 days acute; 4 to 6 days chronic</td>
      <td>24 to 72 acute; 96 to 144 chronic</td>
      <td>Parent drug; the active metabolite norfluoxetine 4 to 16 days (96 to 384 hours)</td>
    </tr>
    <tr>
      <td>Semaglutide</td>
      <td>about 1 week</td>
      <td>168</td>
      <td>Subcutaneous injection, adults with type 2 diabetes</td>
    </tr>
    <tr>
      <td>Adalimumab</td>
      <td>about 2 weeks (10 to 20 days)</td>
      <td>336 (240 to 480)</td>
      <td>Mean terminal half-life across studies; antibody given by injection</td>
    </tr>
    <tr>
      <td>Amiodarone</td>
      <td>58 days (range 15 to 142)</td>
      <td>1,392 (360 to 3,408)</td>
      <td>Mean terminal half-life after a single oral dose in 12 healthy subjects; active metabolite 36 days</td>
    </tr>
  </tbody>

Half-lives span about four orders of magnitude, from the 20 minutes of aspirin's parent molecule to 3,400 hours at the long end of amiodarone's range. Small molecules that the liver or kidney clear quickly sit at the short end. Adalimumab, [a monoclonal antibody](/guides/what-are-monoclonal-antibodies), and semaglutide, tightly bound to albumin, sit at the long end with small clearances. Amiodarone shows the widest spread between subjects, from 15 to 142 days.

Some rows also hide a second molecule. The aspirin value belongs to the parent only: the Durlaza label shows the dose leaving mainly as salicylic acid and its conjugates. Diazepam and fluoxetine are the reverse, with an active metabolite that outlasts the parent. Fluoxetine's label says that because of these long half-lives, changes in dose are not fully reflected in plasma for several weeks.

## Half-lives needed to clear a drug

Each half-life removes half of what remains, so the amount left after $n$ half-lives is a simple power:

$$
\text{fraction remaining} = \left(\tfrac{1}{2}\right)^{n}
$$

After one to five half-lives, 50%, 25%, 12.5%, 6.25% and 3.1% remains. That is the origin of the rule of thumb that a drug is practically eliminated after 4 to 5 half-lives. The same percentages describe how close a regularly dosed drug is to its steady-state level, and both readings assume a single first-order half-life.

Repeated dosing runs the arithmetic in reverse. Levels climb toward a plateau, the steady state, and the terminal half-life controls how long that takes and how much the drug accumulates. The amiodarone label applies the rule to its own values: with a terminal half-life of 26 to 107 days in chronic therapy (mean about 53), steady state without a loading period would arrive in 130 to 535 days, averaging 265. Each figure is 5 half-lives.

Digoxin gives a gentler example. For a half-life of 1.5 to 2 days, 5 half-lives take 7.5 to 10 days. In renal impairment the half-life lengthens and, as the label notes, a new steady state takes longer to reach. The rule assumes first-order kinetics, in which a fixed fraction is cleared per unit time, and for some drugs that assumption fails.

## Why half-lives differ between people

A label value is an average over a study population, and the half-life equation explains most departures from it: anything that changes clearance or volume of distribution changes the half-life. Age is the clearest case. The caffeine label gives a half-life of 3 to 4 days in neonates and 5 hours once metabolism approaches the adult pattern at about 9 months, a 14 to 19-fold difference.

The diazepam label reports a half-life of 20 hours at 20 years of age that increases by about 1 hour for each further year, attributed to a rise in volume of distribution and a fall in clearance. In children aged 3 to 8 the mean is 18 hours.

Organ function matters when that organ does the clearing. Digoxin is eliminated mainly by the kidneys, and its half-life rises from 1.5 to 2 days to 3.5 to 5 days in anuric patients, who produce no urine. Metformin's plasma half-life is also prolonged as renal function declines. Liver disease acts on drugs cleared by metabolism: in cirrhosis, fluoxetine and norfluoxetine clearances fall and their half-lives lengthen.

Genetics acts through the enzymes that metabolize drugs. The warfarin label notes that the CYP2C9\*2 and CYP2C9\*3 variants reduce metabolism of S-warfarin, with allele frequencies of about 11% and 7% in Caucasians. The fluoxetine label reports that about 7% of the population has reduced CYP2D6 activity, and that these poor metabolizers cleared S-fluoxetine more slowly.

Other drugs act through the same enzymes. The fluoxetine label notes that the half-life of diazepam given at the same time may be prolonged in some patients, one of the [CYP inhibition interactions](/guides/cyp-inhibition-and-drug-interactions) that work through clearance.

A fixed half-life breaks down when the clearing enzyme saturates. Phenytoin's liver enzyme is saturable at high concentrations, so small dose increments can lengthen the half-life and sharply raise serum levels.

Ethanol is the textbook case. Its main liver enzyme is saturated after the first few drinks, so blood alcohol falls at a roughly constant rate instead of by a constant fraction. Jones puts the rate at 10 to 15 mg per 100 mL per hour on an empty stomach, 15 to 20 after food and 25 to 35 in alcoholics in detoxification. With a constant rate, the time to fall by half depends on the starting level, so ethanol has no single half-life.

## How half-life is predicted for a new molecule

For a molecule that exists only on screen, half-life cannot be looked up. Smith and colleagues, writing from Pfizer, note that medicinal chemists change half-life mainly by changing clearance through structural modification, and to a lesser extent volume of distribution, so predicting half-life means predicting those two quantities. They describe 12 to 48 hours as generally ideal for once-daily oral drugs.

Clearance is predicted from lab measurements and from animals. In the in vitro half-life approach, human liver microsomes are incubated with the compound and the rate at which it disappears is scaled up to a human clearance. Obach tested this method on 29 drugs and found that the treatment of binding to microsomes and plasma proteins changed the result, and that acidic compounds were hardest to predict.

Scaling from rat, dog and monkey is the other route. Lombardo and colleagues compared 37 methods on about 400 compounds. The best, based on monkey clearance or on rat clearance adjusted for protein-binding differences, put roughly 60% of compounds within 2-fold of the human value. With volume of distribution right, a 2-fold error in clearance is a 2-fold error in half-life: a predicted 12 hours can really be 6 or 24.

Machine-learning models skip the experiment and predict half-life directly from structure. [Admetica](/app/admetica) has a half-life endpoint in hours, alongside hepatocyte and microsome clearance and volume of distribution (VDss). Its half-life model was trained on the Obach compilation of 667 compounds, and its developers report a mean absolute error of 29.2 hours on 24 test structures, against 32.9 hours for the ADMETLab web service.

An error of about 29 hours is larger than the label half-life of eight of the 15 drugs in Table 1. When we ran [ADMET-AI](/app/admet-ai) and Admetica on 12 approved drugs to compare [ADMET prediction](/guides/admet-prediction) tools, fluconazole came out at 0.6 hours from one model and 30.4 from the other, against a label value of about 30 hours. Drugs this well known probably sit in the training data, so one model landing near the label proves little.

Predicted half-life is therefore a coarse filter, good for separating hours from days within a series of related compounds. The clearance and volume predictions it is built from say more about why a molecule would last.

## Reading a half-life number

The 15 labels show that half-life belongs to the body as much as to the molecule. The relation behind adalimumab's two weeks also gives digoxin five days in a patient without kidney function and caffeine three to four days in a newborn. Before comparing two half-lives, check they are the same kind, for the same molecule, in comparable people. The fourfold gap between warfarin's two values is a definition, not a difference in the drug.
