Accumulation in a Multiple-Ascending-Dose Study

Why Repeated-Dose Concentrations Need a Time-Based Interpretation

After a dose, concentration declines as drug is eliminated. If another dose is given before the previous dose has been eliminated, the new dose adds to drug remaining in the body. This accumulation changes both peak and trough concentrations over successive dosing intervals. During a multiple-ascending-dose study, dose changes add another feature: concentrations after an increase reflect both the new dose and residual drug from earlier doses.

This tutorial uses a hypothetical study of repeated intravenous bolus doses in a one-compartment model with first-order elimination. The case keeps the calculations transparent; the same analysis principle applies more generally, but predictions must then come from the specified model and dosing history. Peaks below mean concentrations immediately after a dose, and troughs mean concentrations immediately before the next dose.

Model and assumptions: Let \(D\) be the dose, \(V\) the apparent volume of distribution, \(k\) the first-order elimination rate constant, and \(\tau\) the fixed dosing interval. Between doses, concentration declines exponentially, so a concentration \(C_0\) becomes \(C(t)=C_0e^{-kt}\). Doses enter the central compartment instantaneously. The model assumes constant \(V\) and \(k\), linear pharmacokinetics (concentration is proportional to dose), consistent dosing intervals, and no time-varying disease, adherence, or concomitant-treatment effect. Its predictions are conditional on these assumptions.

Accumulation and the Steady-State Profile

Write \(q=e^{-k\tau}\), the fraction of a concentration remaining after one dosing interval. Since \(k>0\) and \(\tau>0\), \(0<q<1\). The first bolus produces an immediate concentration increment \(D/V\). At the next peak, the first dose contributes \((D/V)q\), and the second contributes \(D/V\). At the \(m\)th peak, the contributions sum to a geometric series:

C_{\max,m}=\frac{D}{V}(1+q+\cdots+q^{m-1}) =\frac{D}{V}\frac{1-q^m}{1-q}.

Because \(q^m\) approaches zero as \(m\) increases, the peak approaches a finite steady-state value. The trough is the peak multiplied by the fraction remaining over one interval.

C_{\max,ss}=\frac{D/V}{1-e^{-k\tau}}, \qquad C_{\min,ss}=C_{\max,ss}e^{-k\tau}.

The accumulation ratio for the peak is the steady-state peak divided by the peak after one dose, at the same dose and with no residual drug before that single dose. It is \(R=1/(1-q)\). It describes how much the peak is raised by prior doses in this model; it is not a universal multiplier independent of the dosing interval or elimination rate.

Derivation: For \(m\) doses, each earlier dose contributes its single-dose concentration multiplied by the appropriate power of \(q\). The finite geometric-sum identity gives \(1+q+\cdots+q^{m-1}=(1-q^m)/(1-q)\). Since \(0<q<1\), \(q^m\to0\); therefore the limiting peak is \((D/V)/(1-q)\). Between that peak and the next dose, elimination multiplies concentration by \(q\), giving the steady-state trough \(qC_{\max,ss}\).

Worked Example: Calculating Peak Accumulation

Suppose a hypothetical participant receives 100 mg every 12 hours. The model has \(V=20.0\) L and \(k=0.150\ \text{h}^{-1}\). Calculate the single-dose peak, the accumulation ratio, and the steady-state peak and trough.

The fraction remaining after 12 hours is \(q=e^{-0.150(12)}=e^{-1.80}=0.1653\). The single-dose peak is \(D/V=100\ \text{mg}/20.0\ \text{L}=5.000\ \text{mg/L}\). Thus \(R=1/(1-0.1653)=1.198\), and \(C_{\max,ss}=5.000(1.198)=5.990\ \text{mg/L}\). The trough is \(5.990(0.1653)=0.9902\ \text{mg/L}\). The steady-state peak is about 19.8% greater than the single-dose peak under these assumptions.

Time to Steady State

Steady state is approached rather than reached exactly after a finite number of doses. At a given point in the dosing cycle, the remaining difference between the concentration after \(m\) intervals and its limiting value is proportional to \(q^m=e^{-km\tau}\). More generally, at the same dosing-cycle phase after elapsed time \(t\), the fraction of the initial gap remaining is \(e^{-kt}\). The fraction of the gap closed is therefore \(1-e^{-kt}\).

t_f=\frac{-\ln(1-f)}{k}, \qquad m_f=\frac{-\ln(1-f)}{k\tau},

where \(f\) is the target fraction of steady state. The value \(t_f\) is a continuous-time threshold; concentrations sampled only at scheduled dose times may first exceed that threshold at the next scheduled time. For a linear model, this time depends on \(k\), not on the dose size. A common approximation is that about 4 to 5 half-lives are needed to reach roughly 94% to 97% of steady state, since the half-life is \(t_{1/2}=\ln(2)/k\).

Worked Example: Estimating Time to 95% of Steady State

For the parameters in the previous calculation, estimate the time to 95% of steady state and the fraction reached after two 12-hour intervals.

Here \(k=0.150\ \text{h}^{-1}\), so \(t_{0.95}=-\ln(1-0.95)/0.150=-\ln(0.05)/0.150=19.97\) hours. This is \(19.97/12=1.664\) dosing intervals. At the two-interval checkpoint, the fraction reached is \(1-e^{-0.150(24)}=1-e^{-3.60}=0.9727\), or 97.27%. The 95% threshold is crossed before the second 12-hour checkpoint, so the first scheduled checkpoint beyond it is at 24 hours.

Reading an Ascending-Dose Profile

A dose increase does not reset the concentration to zero. For example, if a participant has reached steady state at one dose and the dose is increased at the next scheduled administration, the subsequent profile includes drug remaining from the prior regimen. A model comparison should therefore use the actual dose history, not treat each dose level as an isolated single-dose period. If the purpose is to assess steady-state exposure at each level, the sampling schedule must allow sufficient time after each change, or the predictions must explicitly account for carryover.

Dose proportionality is a separate question from accumulation. Under linear pharmacokinetics, with the same dosing interval and the same \(V\) and \(k\), steady-state peaks and troughs scale directly with dose. Equivalently, dose-normalized concentrations should be similar across dose levels. Similarity in a small data set is descriptive evidence, not by itself a formal statistical demonstration of proportionality.

Worked Example: Comparing Observed Peaks and Troughs With Predictions

Consider hypothetical steady-state data from three dose levels given every 12 hours. The model predictions use \(V=20.0\) L and \(k=0.150\ \text{h}^{-1}\). Small differences between the displayed predicted values reflect rounding. Signed percent prediction error is \(100(\text{observed}-\text{predicted})/\text{predicted}\).

Dose (mg)	Peak observed (mg/L)	Peak predicted (mg/L)	Trough observed (mg/L)	Trough predicted (mg/L)
100	5.90	5.990	0.960	0.9902
200	12.2	11.98	2.05	1.980
300	17.4	17.97	2.88	2.971

For the 200 mg peak, the error is \(100(12.2-11.98)/11.98=1.836\%\). For the 300 mg peak, it is \(100(17.4-17.97)/17.97=-3.172\%\). The 100 mg trough error is \(100(0.960-0.9902)/0.9902=-3.050\%\). For the 200 mg trough it is \(100(2.05-1.980)/1.980=3.535\%\), and for the 300 mg trough it is \(100(2.88-2.971)/2.971=-3.063\%\). Positive values indicate observations above predictions; negative values indicate observations below predictions.

The observed peaks per 100 mg are \(0.0590\), \(0.0610\), and \(0.0580\ \text{mg/L per mg}\) at 100, 200, and 300 mg, respectively. These are fairly similar, as are the troughs per 100 mg: \(0.00960\), \(0.01025\), and \(0.00960\ \text{mg/L per mg}\). The pattern is descriptively compatible with dose proportionality, but the modest number of observations does not establish that proportionality holds in the study population. The prediction errors likewise describe model fit at these measurements; they are not p-values or measures of uncertainty.

Diagnostics, Alternatives, and Limitations

Before interpreting accumulation estimates, verify that dose amounts, administration times, and sample times are correct. A nominal “peak” collected well after administration is not the immediate post-bolus peak, and a sample taken before the scheduled next dose is not necessarily a comparable trough. Check that the predicted peak-to-trough pattern is compatible with the elimination rate and dosing interval, and inspect errors across dose levels and time rather than relying on a single summary. A trend in signed errors—for example, increasingly negative peak errors as dose rises—may indicate nonlinearity, a changing parameter, incorrect dose history, or a measurement or timing problem.

The simple equations are useful when the model assumptions are credible and sampling captures interpretable points in the profile. If measurements are sparse, or dosing is not at steady state, fitting or simulating the full concentration-time model with each participant’s actual dosing history is generally preferable to calculating accumulation ratios from isolated peaks. Comparing observed dose-normalized peak and trough values is a simpler descriptive alternative, but it does not distinguish random variability from a systematic dose trend. AUC-based summaries offer a different view of exposure; they do not replace checking peak and trough predictions when those are the study targets.

This one-compartment bolus model does not represent absorption delay, multiple distribution phases, or time-varying clearance. It can also mislead when dose changes alter bioavailability, clearance, or distribution, or when a participant misses doses. In such cases, apparent lack of dose proportionality may reflect model misspecification or study conduct rather than a direct dose effect on pharmacokinetics. Interpret discrepancies in context and revise the model only when supported by data and a scientifically plausible mechanism.

Common mistake: declaring steady state after one half-life because half the drug has been eliminated. Correction: with repeated dosing, contributions from many prior doses accumulate; the remaining gap to steady state falls as \(e^{-kt}\), so several half-lives are usually needed.
Common mistake: comparing an observed trough at one dose with a steady-state prediction when the sample was taken soon after a dose increase. Correction: account for the dose history and time since escalation, or compare only measurements collected under the intended steady-state conditions.
Common mistake: treating a small prediction error as proof that the model is correct. Correction: examine timing, assumptions, patterns across doses and participants, and uncertainty; a close match at a few points cannot validate every model feature.

Key takeaway: Accumulation and time to steady state follow from the elimination rate and dosing interval in a linear model. Dose proportionality concerns how exposure scales with dose. Keep those questions distinct, use the complete dose history, and interpret peak and trough discrepancies alongside the model assumptions and sampling times.

Check Your Understanding

Answer each question, then compare with the brief solutions.

If \(k=0.10\ \text{h}^{-1}\) and \(\tau=8\) hours, what fraction of concentration remains after one interval?
Under the same conditions, what is the peak accumulation ratio?
How long does this model take to reach 90% of steady state?
Why should a trough collected shortly after dose escalation not automatically be compared with the new dose’s steady-state trough?
What pattern in dose-normalized steady-state peaks would be expected under dose-proportional pharmacokinetics?

Solutions. (1) \(q=e^{-0.10(8)}=e^{-0.8}=0.4493\), so about 44.93% remains. (2) \(R=1/(1-0.4493)=1.816\). (3) \(t_{0.90}=-\ln(0.10)/0.10=23.03\) hours. (4) Residual drug from the prior regimen means the concentration has not necessarily reached the new regimen’s steady-state profile; compare at the appropriate time or model the full dose history. (5) The peak concentration divided by dose should be approximately constant across dose levels, subject to variability and measurement error.