How to Calculate AUC in Pharmacokinetics

AUC in pharmacokinetics is calculated by plotting a drug’s measured blood concentrations over time and then computing the total area enclosed by that curve. The most common method is the trapezoidal rule, which breaks the curve into a series of trapezoids between each pair of consecutive blood samples and sums their areas. The concept is straightforward, but the details matter: which flavor of the trapezoidal rule you pick, how many blood samples you collect, and whether you extrapolate past the last measured point all change the final number. Because AUC is the foundation for decisions about drug dosing, safety, and approval, getting the calculation right has real consequences.

What AUC Actually Tells You

AUC stands for “area under the concentration-time curve.” After someone takes a drug, you can draw blood at set intervals and measure how much drug is circulating. Plot those concentrations on the vertical axis against time on the horizontal axis, connect the dots, and the total area under that line represents cumulative drug exposure. A larger AUC means the body saw more drug overall; a smaller AUC means less. Regulatory toxicity protocols tie measures of systemic exposure, including AUC and peak concentration, directly to observed toxic effects, which is why the number must be reliable.¹

AUC is not the same as peak concentration. A drug could spike high and vanish quickly, giving a modest AUC, or it could reach a moderate peak but linger for hours, producing a large AUC. Both numbers matter for different reasons, but AUC captures the total picture of how much drug the body was exposed to over the full dosing window.

The Linear Trapezoidal Rule

The workhorse method is the linear trapezoidal rule. Imagine you have blood-concentration measurements at several time points. Between any two consecutive points, draw a straight line. That line and the time axis form a trapezoid. The area of each trapezoid is easy to compute: take the average of the two concentration values and multiply by the time between them. Add up all the trapezoids and you have an estimate of total AUC.

This approach is considered the reference method for non-compartmental pharmacokinetic analysis. A comparison of several numerical integration methods found that the trapezoidal rule remains an especially reliable option when the concentration-time profile does not follow a strict log-linear decline, and its accuracy improves with more sampling points.² In practice, this means the more blood draws you schedule, the tighter the trapezoids fit the actual curve, and the closer you get to the true AUC.

The weakness shows up during the elimination phase. After a drug peaks, concentrations usually fall in a curved, roughly exponential fashion. Connecting those falling points with straight lines overestimates the area because the straight line sits above the actual curve. When concentrations are dropping steeply between two samples spaced far apart, this overestimation can be meaningful.

Log-Linear and Linear-Up/Log-Down Variants

To handle that curved decline more accurately, pharmacokineticists developed the logarithmic trapezoidal rule. Instead of assuming a straight line between two points, it assumes concentrations decline log-linearly, which is closer to how drugs actually leave the body during the elimination phase. A classic evaluation of the potential error from using only straight-line trapezoids recommended a hybrid approach: use the linear trapezoidal method for data before and around the peak, and switch to the logarithmic trapezoidal method for the post-peak decline.³

This hybrid is now commonly called the “linear-up/log-down” method. While concentrations are rising (during absorption), the linear trapezoidal rule is used. Once concentrations start falling, the logarithmic version takes over.⁴ Most modern pharmacokinetic software defaults to this method or offers it as a standard option, and it is widely considered the best general-purpose choice for non-compartmental analysis.

A broader comparison of numerical integration methods, including spline methods, Lagrange interpolation, and parabolic fits, found that some fancier techniques suffered from high variance in their estimates, making them unreliable despite sounding more sophisticated. A simpler hybrid of parabolas near the origin followed by the log-trapezoidal rule performed especially well across simulated oral, intravenous bolus, and infusion profiles.

Extrapolating Beyond the Last Sample

Blood sampling always has to stop at some point, but the drug is usually still in the body when you collect that final sample. If you only add up trapezoids through the last measured concentration, you get AUC from time zero to the last observation, often written AUC₀₋ₜ. That captures most of the exposure, but it misses the tail.

To estimate total exposure from time zero to infinity (AUC₀₋∞), you need to extrapolate the tail. The standard approach takes the last measured concentration and divides it by the terminal elimination rate constant, a value estimated from the slope of the log-concentration-versus-time line during the final phase of decline. That slope is typically fitted using the last few measurable concentrations that fall on a straight line when plotted on a log scale, ideally with a correlation coefficient above 0.98 to ensure a good fit.⁵

A rule of thumb in regulatory submissions is that the extrapolated portion should not exceed about 20% of the total AUC₀₋∞. If it does, sampling probably stopped too early, and the estimate of the tail becomes unreliable. When the extrapolated fraction is small, it mainly captures a low-level residual amount of drug, and modest errors in the tail barely affect the overall number.

Using AUC to Measure Bioavailability

One of the most common uses of AUC is calculating bioavailability, or the fraction of an administered dose that actually reaches the bloodstream. If you give a drug intravenously, 100% enters the blood by definition. Give the same drug as a pill, and some fraction gets lost to incomplete absorption or metabolism in the gut and liver before it reaches the general circulation. Absolute bioavailability is the ratio of AUC after oral dosing to AUC after intravenous dosing, adjusted for any dose difference.

For example, a study of the blood-pressure drug valsartan found that about 23% of a capsule dose and 39% of an oral solution dose reached systemic circulation, based on comparing oral AUC to intravenous AUC.⁶ A similar study of the HIV drug maraviroc calculated absolute bioavailability at about 23% for a tablet compared with an intravenous dose.⁷ These numbers tell formulators and clinicians how much drug you actually need to swallow to get a therapeutically useful amount into the blood.

AUC in Bioequivalence Testing

When a generic drug seeks approval, regulators require proof that it delivers essentially the same systemic exposure as the brand-name product. The core comparison is between the AUC values (and peak concentrations) of the two products in a crossover study. If the 90% confidence interval for the ratio of generic-to-brand AUC falls within a pre-specified acceptance window, the products are deemed bioequivalent.

An analysis of generic anti-seizure medications approved in Europe found that for 99% of the products assessed, the confidence intervals for the AUC ratio fell within the tighter acceptance range applied to narrow-therapeutic-index drugs, and the variability in AUC between doses within the same person was under 10% for most products.⁸ That kind of tight agreement is what regulators look for. The calculation method itself (which trapezoidal variant, how the terminal slope is estimated) must be consistent between the two arms of the study, because even small methodological differences could shift the ratio enough to matter when the acceptance window is narrow.

Interestingly, studies have compared AUC ratios obtained from non-compartmental analysis against those from compartmental modeling and population approaches, and the final bioequivalence conclusions were consistent regardless of the method used.⁹ For regulatory submissions, non-compartmental analysis remains the default because it makes fewer assumptions about how the drug behaves in the body.

AUC-Guided Dosing in Clinical Practice

AUC is not just for drug development. Clinicians use it at the bedside to individualize dosing for drugs with narrow safety margins. The antibiotic vancomycin is a prominent example. Current Japanese consensus guidelines recommend targeting an AUC of 400 to 600 µg·h/mL, with dosing software predicting a regimen that maximizes the chance of hitting that target based on one or two measured blood levels from an individual patient.¹⁰ In practice, a pharmacist enters the patient’s weight, kidney function, and measured vancomycin levels into Bayesian software, which fits a pharmacokinetic model and estimates AUC. If the estimated AUC is too low, the dose goes up; too high, it comes down.

Whether AUC-guided vancomycin dosing actually reduces kidney injury compared to older trough-only monitoring remains an active question. A recent retrospective study found that AUC-guided dosing did not reduce the incidence of acute kidney injury compared to trough-only monitoring, though patients managed with AUC targets tended to have lower average trough concentrations.¹¹ The clinical community is still sorting out when AUC targeting is clearly better versus when simpler monitoring suffices.

Oncology is another area where AUC-guided dosing has proven valuable. A multicenter study of the chemotherapy drug 5-fluorouracil in metastatic colorectal cancer adjusted doses to keep AUC between 20 and 30 mg·h/L. By the fourth treatment cycle, about 54% of patients were within that target range, and the rates of severe side effects like diarrhea and nausea were lower than historical benchmarks, even though more than half the patients ended up receiving higher doses than they started with.¹² This is the appeal of AUC-based dosing: you can push the dose up when the patient’s body is clearing the drug faster than average, and pull it back when clearance is slow, instead of giving everyone the same milligrams and hoping for the best.

Dealing with Limited Blood Samples

Collecting a full pharmacokinetic profile with 10 or 15 blood draws is feasible in a controlled clinical study, but it is not practical in routine patient care or in certain preclinical settings. Limited sampling strategies address this by identifying a small number of optimally timed samples that, when plugged into a regression equation or Bayesian model, can predict the full AUC with acceptable accuracy.

For the immunosuppressant mycophenolate, for instance, researchers derived a five-sample schedule (at 2, 2.5, 3, 5, and 6 hours after infusion start) that could precisely estimate the 12-hour AUC in transplant recipients.¹³ Similar limited sampling models have been developed for carboplatin chemotherapy, where even a one- or two-sample model can give a useful AUC estimate for dosing decisions.¹⁴

In preclinical toxicology work with small animals, the situation is even more constrained. You may only be able to draw blood once from each animal, meaning a full curve for a single subject does not exist. Statistical methods such as the Bailer-Satterthwaite approach pool single observations from different animals taken at different time points to construct a composite AUC estimate with confidence intervals.¹⁵ Bootstrap resampling techniques offer an alternative that works even when the concentration data are not normally distributed and can also provide standard errors for secondary parameters beyond AUC itself.¹⁶

Population Models and Compartmental AUC

Everything described so far falls under non-compartmental analysis, where you let the data speak for themselves without fitting a pharmacokinetic model. The alternative is compartmental modeling, where you define the body as one, two, or more theoretical compartments, fit a mathematical model to the concentration data, and then calculate AUC from the model’s parameters rather than from trapezoids.

Population pharmacokinetic modeling takes this further by analyzing data from many patients simultaneously, estimating typical parameter values and the variability around them. A population model of voriconazole in pediatric patients used a two-compartment model with nonlinear elimination to describe the data, and the resulting AUC distributions were used to compare pediatric and adult exposure levels.¹⁷ The AUC derived from a fitted compartmental model is not inherently more or less “correct” than a trapezoidal AUC; it is a different approach that can be more powerful when data are sparse or when you need to predict AUC under dosing scenarios you have not directly observed.

When AUC Does Not Scale with Dose

For many drugs, doubling the dose roughly doubles the AUC. This is called dose proportionality, and it makes life easy because you can predict exposure at new doses from existing data. But some drugs do not behave this way. If the enzymes responsible for eliminating a drug start to saturate at higher doses, the body cannot clear the extra drug proportionally, and AUC rises disproportionately. The degree of nonlinearity can be quantified by dividing the ratio of AUC values by the ratio of doses; a value greater than one indicates the drug’s exposure is climbing faster than the dose.¹⁸

Ethanol is a textbook example. The blood-alcohol AUC shows a markedly nonlinear relationship with dose, meaning a moderate increase in the amount consumed can produce a much larger increase in total exposure.¹⁹ Prescription drugs like verapamil and propranolol can also show nonlinear behavior because of saturable first-pass metabolism in the liver, where higher doses or faster absorption overwhelm the enzymes that break the drug down before it reaches the bloodstream.²⁰ For these drugs, calculating AUC at multiple dose levels is essential because you cannot simply scale from one dose to another.

Software for AUC Calculation

Hardly anyone calculates AUC by hand anymore. The commercial standard for decades has been Phoenix WinNonlin, and most regulatory submissions reference it. But free, validated alternatives now exist. In R alone, at least four packages handle non-compartmental analysis: NonCompart, ncar, PKNCA, and ncappc. All four use the trapezoidal method, and their results have been benchmarked against WinNonlin during development to ensure accuracy. Several of them can format output in the Study Data Tabulation Model format required for regulatory submission to agencies like the FDA.²¹

Online platforms have also emerged. One recent tool, CPhaMAS, was benchmarked against WinNonlin for bioequivalence calculations involving conventional, high-variability, and narrow-therapeutic-index drugs, showing a mean relative error of under 0.01% for AUC and peak concentration parameters.²² The practical upside is that researchers no longer need an expensive software license to produce regulatory-quality AUC calculations, though they still need to understand what the software is doing under the hood so they can catch errors in data input or parameter selection.

AUC in Tissue Versus Blood

Most AUC calculations are based on drug concentrations measured in blood plasma, but the drug’s therapeutic target is usually somewhere else: an infected joint, a tumor, the lining of the abdomen. Plasma AUC may overestimate or underestimate what the tissue actually sees. Microdialysis is a technique that allows researchers to measure drug concentrations directly in tissue fluid over time, producing a tissue-specific AUC.

In a study of the antibiotic ertapenem in morbidly obese patients undergoing abdominal surgery, researchers placed microdialysis probes in subcutaneous fat and peritoneal fluid. The unbound-drug AUC in subcutaneous tissue was about 49% lower than the unbound AUC in plasma, and in peritoneal fluid it was about 25% lower.²³ Differences this large matter when you are deciding whether a standard dose delivers enough antibiotic to the infection site. Plasma AUC is the standard metric because it is the easiest to measure and the most reproducible, but it should be understood as a surrogate for what is happening at the tissue level, not a direct measurement of it.