Selecting a Dosing Interval for a Long-Acting Formulation

Why the Interval Is a Model-Based Decision

A long-acting formulation can reduce dosing frequency, but a longer interval may also lower concentration near the end of each cycle. The decision is not simply whether the formulation releases drug slowly: it is whether a proposed dose and interval maintain the exposure needed across the interval, with acceptable variability and without an unacceptable increase in risk of inadequate response.

The case below compares three intervals for a hypothetical depot formulation. It uses a linear, one-compartment disposition model with first-order absorption from a depot. The numerical results are illustrative, not findings from a real study. A trough below a prespecified concentration is treated as an exposure-based risk indicator; it is not automatically equivalent to clinical breakthrough.

Model and assumptions: Each injection deposits dose \(D\) into a depot. A fraction \(F\) reaches the systemic circulation, absorption occurs with first-order rate constant \(k_a\), and drug is eliminated from one well-mixed compartment with first-order rate constant \(k\). Assume \(k_a>k>0\), linear and time-invariant PK, the same dose at every administration, no missed doses, and dosing every \(\tau\) days. The model describes concentrations after repeated dosing has reached steady state. It does not include target-mediated disposition, nonlinear absorption, changing clearance, or a direct model of clinical response.

Repeated-Dose Concentration and Trough

For a single depot dose, the concentration at time \(t\geq 0\) is

C_{\mathrm{single}}(t) = \frac{FDk_a}{V(k_a-k)} \left(e^{-kt}-e^{-k_at}\right) = B\left(e^{-kt}-e^{-k_at}\right), \qquad B=\frac{FDk_a}{V(k_a-k)}.

Here \(V\) is the apparent distribution volume and \(B\) has concentration units. At steady state, concentrations from the current and all previous injections add under the linearity assumption. For time \(t\) after an injection, where \(0\leq t<\tau\), summing the geometric series for the elimination and absorption terms gives:

C_{\mathrm{ss}}(t) = B\left[ \frac{e^{-kt}}{1-e^{-k\tau}} - \frac{e^{-k_at}}{1-e^{-k_a\tau}} \right].

The trough is the concentration immediately before the next injection, \(C_{\min,\mathrm{ss}}=C_{\mathrm{ss}}(\tau^-)\). By periodic steady state, it is also the concentration at the start of the cycle, before the new depot dose has had time to contribute. Substitution gives:

C_{\min,\mathrm{ss}} = B\left[ \frac{e^{-k\tau}}{1-e^{-k\tau}} - \frac{e^{-k_a\tau}}{1-e^{-k_a\tau}} \right].

This is a trough prediction, not a prediction of the minimum concentration at every point in the cycle under every possible absorption process. With the stated depot model, concentration rises after administration and then falls; the cycle-boundary trough is the minimum for the parameter values used here. In applications, the full concentration-time profile should be checked rather than assuming that a named “trough” is necessarily the global minimum.

Worked Example: Computing Troughs for Three Intervals

Suppose a hypothetical model has \(k=0.0400\) per day and \(k_a=0.100\) per day. The dose and scale are summarized by \(B=21.82\) mg/L. Compare intervals of 28, 42, and 56 days. These are model inputs for the example; the calculations assume the same dose and linear PK for all three choices.

For 42 days, \(e^{-k\tau}=e^{-1.68}=0.1864\) and \(e^{-k_a\tau}=e^{-4.2}=0.014996\). Therefore:

C_{\min,\mathrm{ss}} = 21.82\left[ \frac{e^{-1.68}}{1-e^{-1.68}} - \frac{e^{-4.2}}{1-e^{-4.2}} \right] \approx 4.666\ \mathrm{mg/L}.

The same formula gives approximately \(9.155\) mg/L at 28 days and \(2.519\) mg/L at 56 days. The longer interval lowers the modeled trough. Because the model is linear and the dose is held fixed, these differences are not evidence of a change in clearance; they result from allowing more time between administrations.

Accumulation and Average Exposure Are Different Questions

Accumulation describes how concentrations from earlier doses contribute to the current profile. For this absorption model, the steady-state trough can be compared with the concentration at the same time after a single dose. The ratio is:

R_{\mathrm{trough}} = \frac{C_{\min,\mathrm{ss}}} {C_{\mathrm{single}}(\tau)} = \frac{ \frac{e^{-k\tau}}{1-e^{-k\tau}} - \frac{e^{-k_a\tau}}{1-e^{-k_a\tau}} }{ e^{-k\tau}-e^{-k_a\tau} }.

This ratio is specific to the trough and this model; it is not a universal accumulation factor for every point on the profile. For comparison, with instantaneous input and first-order elimination the familiar accumulation factor is \(1/(1-e^{-k\tau})\). That simpler expression does not account for continuing depot absorption and should not be substituted for the full model when estimating a depot trough.

Under complete absorption into a linear system, the area under the concentration-time curve during one dosing interval is \(AUC_\tau=FD/CL\), where \(CL=kV\). With the same dose at each administration, this area is the same across intervals, but the average concentration over an interval is \(AUC_\tau/\tau\). Thus a 28-day interval has 1.5 times the average concentration of a 42-day interval, while a 56-day interval has 0.75 times its average concentration. Equal dose per injection does not mean equal average exposure when the interval changes.

Worked Example: Quantifying Trough Accumulation

For the 42-day interval, the single-dose concentration at day 42 is:

C_{\mathrm{single}}(42) = 21.82(0.1864-0.014996) \approx 3.740\ \mathrm{mg/L}.

The steady-state trough calculated above is about \(4.666\) mg/L, so the trough accumulation ratio is \(4.666/3.740\approx 1.248\). In this model, prior doses raise the day-42 concentration by about 24.8% relative to the day-42 concentration after one dose. Using the same calculation, the ratios are approximately 1.580 for 28 days and 1.123 for 56 days. The shorter interval allows more residual contribution at the cycle boundary.

This comparison is internally consistent with the assumptions: dose and PK parameters are unchanged, and all prior doses contribute additively. It does not imply that the peak concentration has the same accumulation ratio; peaks occur at another time and require evaluating the full profile.

From Trough Prediction to Below-Target Probability

A population decision often needs more than a typical trough. Suppose individual steady-state troughs at a given interval are approximated by a lognormal distribution, so \(\ln(C_{\min,\mathrm{ss}})\) is normally distributed. If the geometric mean trough is \(G\) and the standard deviation on the log scale is \(\omega\), then:

P(C_{\min,\mathrm{ss}}<C_{\mathrm{target}}) = \Phi\left( \frac{\ln(C_{\mathrm{target}}/G)}{\omega} \right),

where \(\Phi\) is the standard normal cumulative distribution function. This calculation assumes the lognormal model adequately represents the modeled population distribution and that \(G\) and \(\omega\) apply to the interval being assessed. It gives the fraction predicted to have a trough below the chosen concentration, not the probability that a particular patient will experience breakthrough, and not the probability that the model is correct.

Worked Example: Estimating 42-Day Below-Target Risk

For the hypothetical 42-day interval, take the geometric mean trough as \(G=4.666\) mg/L and the log-scale standard deviation as \(\omega=0.2936\). Let the prespecified exposure threshold be \(C_{\mathrm{target}}=4.0\) mg/L. Assume the lognormal distribution and these variability inputs are appropriate for this example.

z = \frac{\ln(4.0/4.666)}{0.2936} \approx -0.5245.

Therefore:

P(C_{\min,\mathrm{ss}}<4.0) = \Phi(-0.5245) \approx 0.3000.

About 30.0% of the modeled population is predicted to have a steady-state trough below 4.0 mg/L under these assumptions. This is a model-based exposure shortfall estimate. To call it a breakthrough probability would require a justified exposure-response relationship linking trough exposure to clinical events, plus any relevant sources of variability in response.

Making and Qualifying the Interval Recommendation

For this comparison, suppose 4.0 mg/L is a prespecified minimum exposure target and the development team regards a below-target fraction above 10% as unacceptable. To extend the same population-variability assumption to other intervals, assume the individual troughs scale proportionally with the calculated typical trough and retain \(\omega=0.2936\). This is an explicit simplifying assumption, not an automatic property of every population PK model.

At 28 days, the modeled geometric mean trough is about 9.155 mg/L. The standardized log threshold is \(\ln(4.0/9.155)/0.2936\approx -2.82\), corresponding to a below-target fraction of about 0.24%. At 42 days, the worked calculation gives about 30.0%. At 56 days, the standardized value is \(\ln(4.0/2.519)/0.2936\approx 1.58\), corresponding to a below-target fraction of about 94.2%. Given the stated target and decision criterion, 28 days is the only candidate that meets the criterion in this example. This recommendation concerns modeled exposure under specified assumptions; it does not establish comparative clinical efficacy or safety.

The recommendation should be revisited if the target was not clinically justified, if the variability is misspecified, or if the chosen regimen produces tolerability concerns at peak exposure. Useful checks include comparing predicted profiles with observed concentration-time data, confirming that the late-interval decline and absorption timing are represented, and checking whether the estimated variability supports the tail probability being used. A good typical fit does not by itself validate the predicted fraction in the low-exposure tail.

Define the decision criterion.
Specify the exposure target and an acceptable below-target fraction before comparing intervals. State whether the target is a validated response threshold or only a modeling benchmark.

Predict full profiles and troughs.
Use the repeated-dose model with the candidate dose and interval, and evaluate the entire cycle as well as its boundary concentration.

Assess variability and accumulation.
Quantify the population distribution of troughs, check its assumptions, and distinguish trough accumulation from average exposure and peak accumulation.

Choose and qualify a regimen.
Select only among regimens meeting the prespecified exposure criterion, then state what the model does not establish about clinical outcomes.

Alternatives, Diagnostics, and Common Pitfalls

A full repeated-dose model is preferable when the formulation has a delayed absorption phase or when interval-end concentrations drive the decision. A simpler terminal half-life calculation can be a useful screening tool, but it omits the depot input and can misrepresent both the time course and trough. Direct comparison of observed profiles can inform the model, but observed data at only one interval or sparse late-cycle sampling may not identify the tail behavior needed for interval selection. A population simulation can represent parameter variability and covariates more fully than a single typical profile, but its results remain conditional on the model and parameter distributions supplied.

Potential warning signs include systematic late-cycle prediction errors, observed delayed peaks not captured by first-order absorption, concentrations inconsistent with the assumed elimination rate, or sensitivity of the below-target fraction to modest changes in \(\omega\). Also examine whether the predicted trough is truly the minimum over the cycle and whether the formulation has a lag, release failure, or dose-dependent absorption not represented in the model. These findings call for model revision or a narrower claim, not merely a different numerical cutoff.

Wrong reasoning: “The typical trough exceeds the target, so nearly everyone is protected.”
Correction: A typical value does not describe the low-exposure tail. Use a justified population distribution and report the below-target fraction with its assumptions.
Wrong reasoning: “The below-target fraction is the probability of clinical breakthrough.”
Correction: It is an exposure-based probability. A clinical risk interpretation requires an appropriate exposure-response link and outcome model.
Wrong reasoning: “The instantaneous-input accumulation factor gives the depot trough.”
Correction: That factor ignores ongoing absorption. Sum the absorption and elimination contributions using the repeated-dose model.
Wrong reasoning: “The same injection dose gives the same average exposure for every interval.”
Correction: With linear PK, AUC per dose is unchanged, but average exposure is \(AUC_\tau/\tau\), so it depends on the interval.

Key takeaway: Choose a long-acting dosing interval by combining the repeated-dose concentration profile, trough distribution, and a prespecified exposure decision criterion. State explicitly where the model ends: a predicted trough shortfall is not, on its own, a predicted clinical failure.

Check Your Understanding

Answer each question, then compare with the brief solutions.

Why does the steady-state trough formula contain separate terms for \(k\) and \(k_a\)?
For a 42-day interval, the geometric mean trough is 4.666 mg/L and \(\omega=0.2936\). What is the predicted fraction below 4.0 mg/L?
If the same dose is given every 56 days rather than every 28 days, what happens to average exposure under linear PK? State the ratio.
Why is \(1/(1-e^{-k\tau})\) not generally the correct depot trough accumulation ratio?
What additional evidence is needed before calling a below-target fraction a clinical breakthrough probability?

Solutions. (1) Elimination and depot absorption each contribute a geometric series across prior doses, with different rate constants. (2) \(z=\ln(4.0/4.666)/0.2936\approx-0.5245\), so \(\Phi(z)\approx0.3000\), or 30.0%. (3) Average exposure is inversely proportional to \(\tau\) at fixed dose, so the 56-day regimen has \(28/56=0.50\) times the average exposure of the 28-day regimen. (4) That expression assumes instantaneous input and accounts only for elimination; depot absorption continues during the interval. (5) A justified exposure-response relationship linking the relevant exposure measure to clinical outcomes is needed, along with appropriate characterization of response variability and other determinants of clinical events.