From a Population PK Model to a Regimen Comparison
A regimen that produces suitable exposure in a typical patient can still miss the target in many individuals. Clearance, volume of distribution, covariates, and other sources of between-patient variability spread exposures across a population. Simulated target attainment makes this spread visible: a population PK model generates exposure predictions for virtual patients, and the analysis counts how many meet a target chosen in advance.
The result is a model-based comparison, not a direct measurement of clinical benefit. Its meaning depends on the population represented, the exposure metric and target, and the assumptions used to generate predictions. This tutorial uses a simple linear PK model to make those choices and calculations explicit.
Specify the Model, Target, and Population
For a worked comparison, assume a one-compartment model with instantaneous intravenous bolus doses, first-order elimination, and linear PK. For each virtual patient, clearance \(CL\) and volume of distribution \(V\) are positive. The elimination rate constant is \(k=CL/V\), and concentration declines exponentially between doses. A regimen is defined by dose amount \(D\) and dosing interval \(\tau\). Assume the patient receives every dose and has reached steady state.
These assumptions let us isolate the comparison of dose amount and interval. Real models may include oral absorption, multiple compartments, time-varying parameters, or missed doses; the simulation must use the model appropriate to the intended application. The target also needs a clinical or pharmacological justification. For the examples, the prespecified target is steady-state \(AUC_{24}\) from 100 to 150 mg·h/L, inclusive. The numbers are hypothetical and are not a treatment recommendation.
For a regimen whose interval divides 24 hours, the steady-state exposure over a 24-hour period is the number of doses in that period multiplied by the exposure per dose interval. Under the stated model, exposure per interval is \(D/CL\). Therefore:
Here \(D\) is in mg, \(CL\) in L/h, and \(\tau\) in hours, so \(AUC_{24}\) is in mg·h/L. This calculation assumes \(24/\tau\) is an integer. When an interval does not divide 24 hours, define the target using an explicitly specified rolling 24-hour exposure or another appropriate metric, rather than applying this shortcut without checking it.
Why the Interval Can Change Concentrations Without Changing Daily AUC
The exposure-per-interval result follows from mass balance. At steady state, the amount of drug in the body at the end of a dosing interval equals the amount at its start. Thus the net change in body amount over one interval is zero. With linear elimination, the elimination rate is \(CL\,C(t)\); the total amount eliminated during the interval is \(CL\) multiplied by the concentration-time integral. Since the dose entering the body is \(D\), that integral must be \(D/CL\):
This establishes the interval AUC, not the concentration profile. For a fixed total daily dose, a shorter interval can reduce peak-to-trough fluctuation even though the daily AUC is unchanged. In the same one-compartment bolus model, each dose adds \(D/V\) to concentration. The pre-dose steady-state concentration is the sum of the residual concentrations from all previous doses, a geometric series:
This trough expression depends on both \(CL/V\) and \(\tau\). It provides a second summary for comparing regimens when trough concentration is relevant, but it does not replace the prespecified AUC target.
Worked Example: Calculate Exposure for One Virtual Patient
Suppose one hypothetical patient has \(CL=5.00\) L/h. Compare 300 mg every 12 hours with 600 mg every 24 hours. Both schedules deliver 600 mg per day, and both satisfy the assumption that the interval divides 24 hours.
For 300 mg every 12 hours, there are \(24/12=2\) doses per day:
For 600 mg every 24 hours:
Both predictions are within the hypothetical target of 100–150 mg·h/L. Their troughs need not be equal. With \(V=50.0\) L, \(k=5.00/50.0=0.100\) h\(^{-1}\). The 12-hour regimen gives \(C_{\min,ss}= (300/50.0)/(e^{1.2}-1)=2.59\) mg/L. The 24-hour regimen gives \(C_{\min,ss}= (600/50.0)/(e^{2.4}-1)=1.20\) mg/L. Equal daily AUC therefore does not imply equal trough concentration.
Calculate Attainment Across Simulated Patients
A population simulation applies the same regimen to many virtual patients, each with their own parameter values and, where appropriate, covariates. For every patient, calculate the chosen exposure metric and classify the prediction against the prespecified target. The denominator must include all simulated patients in the analysis population; do not discard patients outside the target range.
Worked Example: Count Target Attainment in a Small Simulated Population
Consider eight hypothetical virtual patients with the following simulated clearances. For clarity, this example varies \(CL\) and holds the model and other conditions fixed. Compare 300 mg every 12 hours, 600 mg every 24 hours, and 500 mg every 24 hours. The target remains 100–150 mg·h/L, inclusive.
| Simulated \(CL\) (L/h) | 300 mg every 12 h \(AUC_{24}\) (mg·h/L) | 600 mg every 24 h \(AUC_{24}\) (mg·h/L) | 500 mg every 24 h \(AUC_{24}\) (mg·h/L) |
|---|---|---|---|
| 3.0 | 200.0 | 200.0 | 166.7 |
| 4.0 | 150.0 | 150.0 | 125.0 |
| 4.5 | 133.3 | 133.3 | 111.1 |
| 5.0 | 120.0 | 120.0 | 100.0 |
| 5.5 | 109.1 | 109.1 | 90.9 |
| 6.0 | 100.0 | 100.0 | 83.3 |
| 7.0 | 85.7 | 85.7 | 71.4 |
| 8.0 | 75.0 | 75.0 | 62.5 |
For example, at \(CL=4.5\) L/h on 300 mg every 12 hours, \(AUC_{24}=2(300)/4.5=133.3\) mg·h/L. For that regimen, five of the eight predictions fall within the target: 150.0, 133.3, 120.0, 109.1, and 100.0 mg·h/L. Its estimated target-attainment proportion in this small simulated set is:
The 600 mg every 24 hours regimen also attains the target in five of eight patients, because its total daily dose is the same and the model makes \(AUC_{24}\) depend on daily dose and clearance, not interval. The 500 mg every 24 hours regimen attains the target in three of eight patients, or \(3/8=37.5\%\). This finite, deliberately small set illustrates the calculation; it should not be treated as a precise estimate for a real population.
Compare the Trade-Offs, Not Only One Percentage
A regimen comparison should place target attainment alongside practical and pharmacological considerations. One regimen may reach the AUC target in more simulated patients, while another involves fewer daily administrations or produces a different trough profile. The preferred option cannot be selected from the attainment percentage alone; the target, relevant safety constraints, dosing burden, and clinical context matter.
For example, consider a larger hypothetical simulation of 1,000 virtual patients under the same fixed model. Suppose 724 patients on 300 mg every 12 hours and 724 on 600 mg every 24 hours fall within the AUC target, while 498 on 500 mg every 24 hours do. These are illustrative simulation outputs. The calculations are \(724/1000=72.4\%\) for either 600 mg/day regimen and \(498/1000=49.8\%\) for 500 mg/day. The difference between 72.4% and 49.8% is 22.6 percentage points, not a 22.6% relative increase.
If trough concentration also matters, suppose the same simulation indicates that 612 patients on 300 mg every 12 hours and 284 on 600 mg every 24 hours meet a separately prespecified trough criterion. These counts imply 61.2% and 28.4%, respectively. The pattern is compatible with the model: the two regimens have equal daily AUC, but the longer interval can produce a lower trough. This does not establish that the more frequent regimen is clinically superior; the importance of the trough criterion and any other relevant outcomes must be justified separately.
Worked Example: Summarize a Simulated Regimen Comparison
Using the hypothetical 1,000-patient outputs above, compare the three regimens on AUC target attainment and daily administrations. The first two deliver 600 mg/day; the third delivers 500 mg/day.
| Regimen | Patients in AUC target | AUC target attainment | Administrations per day |
|---|---|---|---|
| 300 mg every 12 h | 724 of 1,000 | 72.4% | 2 |
| 600 mg every 24 h | 724 of 1,000 | 72.4% | 1 |
| 500 mg every 24 h | 498 of 1,000 | 49.8% | 1 |
The arithmetic is direct: for 300 mg every 12 hours, \(724/1000=0.724\), or 72.4%. The 600 mg every 24 hours regimen has the same AUC target-attainment count in this example, but one rather than two administrations per day. The 500 mg daily regimen has lower attainment in this simulated population. If trough is an important criterion, the separate trough results also favor the more frequent schedule in this hypothetical comparison. A sound summary reports these dimensions separately rather than collapsing them into an unsupported single “best regimen” score.
Reproducibility, Diagnostics, and Failure Modes
A reproducible analysis records the population definition, model version, parameter-generation method, dose and interval, exposure metric, target limits, simulation size, and random seed. If candidate regimens are applied to the same virtual patients, comparisons are paired: each regimen is evaluated for the same simulated patient characteristics. This helps make differences reflect regimen changes rather than unrelated samples from the population.
Check that the simulated population has plausible parameter and covariate distributions, that all calculated exposures have the expected units and range, and that the target is applied consistently at its boundaries. Confirm that the simulation represents the intended regimen and steady-state conditions. For a linear model with the same daily dose, the \(AUC_{24}\) predictions should agree across intervals when the interval divides 24 hours; a discrepancy suggests an implementation, unit, or dosing-schedule error.
Inspect more than the final proportion. Exposure distributions and summaries such as median, spread, and the fraction below and above a target window can reveal whether attainment is being driven by one tail. If the population simulation omits an important covariate or uses an inappropriate variability structure, a precise-looking percentage can still be misleading. Target attainment is also sensitive to the target definition: changing a boundary or switching from AUC to trough changes the question being answered.
A typical-patient calculation is a useful check, but it does not estimate the fraction of a variable population that meets a target. A population simulation addresses that distributional question directly. Conversely, an attainment percentage does not explain why patients miss the target, and it is only as credible as the model and population inputs. The percentage is a conditional prediction under the specified model, not a guarantee about future patients or proof of efficacy or safety.
Check Your Understanding
Answer each question using the model and target-attainment definitions in this tutorial. Solutions follow the questions.
- For a patient with \(CL=6.0\) L/h, calculate \(AUC_{24}\) for 400 mg every 12 hours. Is the result in the 100–150 mg·h/L target?
- In a simulation of 500 virtual patients, 365 meet the target. Calculate and interpret the target-attainment proportion.
- Under the stated linear model, what happens to \(AUC_{24}\) if a 600 mg/day regimen changes from 300 mg every 12 hours to 600 mg every 24 hours? Name one quantity that can change.
- Why must the target be chosen before comparing candidate regimens?
Solutions. (1) \(AUC_{24}=(24/12)(400/6.0)=133.3\) mg·h/L, which is within the inclusive target. (2) \(365/500=0.730\), or 73.0%; under the specified model and simulated population, 73.0% of the virtual patients meet the target. (3) Both schedules give \(600/CL\) mg·h/L over 24 hours for the same patient, so \(AUC_{24}\) is unchanged. The trough concentration can change because it depends on the dosing interval. (4) Choosing the target after seeing the results can favor a regimen through a selectively chosen criterion; prespecification makes the comparison answer a defined question rather than one tailored to the output.