Mean Transit Time and Transit-Compartment Number

Why Mean and Shape Both Matter

A delayed or spread-out absorption profile cannot be described fully by a single timing quantity. The mean transit time (MTT) describes the average time for drug to pass through the transit process before entering the systemic circulation. The number of transit compartments, \(N\), helps describe how broadly those individual transit times are distributed around their mean. Two models can therefore have the same MTT but very different absorption-time profiles.

Here, \(N\) follows a common convention: there are \(N\) transit compartments before the final absorption step, so a molecule passes through \(N+1\) sequential stages in total. The special case \(N=0\) has one exponential waiting stage and no intermediate transit compartments. Stating this convention matters because alternative conventions can attach a different stage count to the same symbol.

Definition: The mean transit time, MTT, is the expected time from the start of the transit process until drug enters the systemic circulation. The integer \(N\geq 0\) is the number of intermediate transit compartments under the convention above. MTT sets the mean of the transit-time distribution; \(N\) affects its shape and dispersion.

The transit-time input distribution is the distribution of times at which drug becomes available for systemic input. As established in the earlier material on deriving the transit-time input distribution, for identical sequential first-order stages with rate \(k_{\mathrm{tr}}\), its density is an Erlang distribution. Under the stated convention, it has shape \(N+1\). The density is normalized to integrate to one; multiplying it by the systemically available dose gives the corresponding input rate.

g(t)=\frac{k_{\mathrm{tr}}^{N+1}t^N e^{-k_{\mathrm{tr}}t}}{N!}, \qquad t\geq 0

The rate \(k_{\mathrm{tr}}\) has units of inverse time, while MTT and the standard deviation of transit time have units of time. The distribution begins at time zero, although its density may initially be zero when \(N>0\). Thus, MTT is an average, not a fixed delay at which absorption suddenly begins.

How MTT and N Determine the Distribution

The mean and variance follow directly by viewing the total transit time as a sum of waiting times. Let \(X_i\) be the waiting time in stage \(i\), for \(i=1,\ldots,N+1\). Under the model assumptions, these are independent exponential random variables with rate \(k_{\mathrm{tr}}\). Each has mean \(1/k_{\mathrm{tr}}\) and variance \(1/k_{\mathrm{tr}}^2\). The total transit time is \(T=\sum_{i=1}^{N+1}X_i\).

E(T)=\sum_{i=1}^{N+1}E(X_i)=\frac{N+1}{k_{\mathrm{tr}}}

\operatorname{Var}(T)=\sum_{i=1}^{N+1}\operatorname{Var}(X_i) =\frac{N+1}{k_{\mathrm{tr}}^2}

The variance sum uses independence: the covariance terms are zero. Since \(E(T)=\mathrm{MTT}\), the standard deviation and coefficient of variation (standard deviation divided by the mean) can also be expressed in terms of MTT and \(N\).

\mathrm{MTT}=\frac{N+1}{k_{\mathrm{tr}}}, \qquad \operatorname{Var}(T)=\frac{\mathrm{MTT}^2}{N+1}, \qquad \operatorname{SD}(T)=\frac{\mathrm{MTT}}{\sqrt{N+1}}, \qquad \operatorname{CV}(T)=\frac{1}{\sqrt{N+1}}

These relationships give the central interpretation. Increasing MTT while holding \(N\) fixed shifts transit times later and increases their absolute spread in proportion to MTT. Increasing \(N\) while holding MTT fixed leaves the mean unchanged but reduces the relative spread. The timing becomes more concentrated around the mean, rather than being delayed by an additional fixed amount.

For \(N\geq1\), the density reaches its maximum at the mode \(t_{\mathrm{mode}}=N/k_{\mathrm{tr}}\), or \(N/(N+1)\) times MTT. For \(N=0\), the exponential density is greatest at time zero and then declines. The mode, mean, and time of peak concentration are distinct quantities: peak concentration also depends on drug disposition after systemic input.

Worked Example: Interpreting MTT and N Together

Suppose a hypothetical transit model has MTT \(=3\) hours and \(N=2\). There are \(N+1=3\) sequential stages. The rate consistent with this mean is \(k_{\mathrm{tr}}=(N+1)/\mathrm{MTT}=3/3=1\) per hour. The transit-time standard deviation is \(3/\sqrt{3}=1.732\) hours, and its coefficient of variation is \(1/\sqrt{3}=0.5774\).

The density is \(g(t)=t^2e^{-t}/2\), with time measured in hours and density in inverse hours. Its mode is \(N/k_{\mathrm{tr}}=2/1=2\) hours, which is earlier than the 3-hour mean. The model therefore describes a distribution of absorption times, not absorption occurring exactly at 3 hours.

For this three-stage distribution, the fraction of the total input that has occurred by time \(t\) is \(1-e^{-k_{\mathrm{tr}}t}\{1+k_{\mathrm{tr}}t+(k_{\mathrm{tr}}t)^2/2\}\). At 3 hours, this is \(1-e^{-3}(1+3+4.5)=1-(0.049787)(8.5)=0.5768\). Thus, about 57.7% of the modeled input has occurred by the mean time; the mean does not imply that exactly half has occurred by then.

Same Mean, Different Dispersion

A useful way to isolate the role of \(N\) is to compare distributions with the same MTT. The following example uses an MTT of 4 hours. The input fractions by 4 hours are computed from the Erlang cumulative distribution function. For integer \(N\), that function is \(1-e^{-k_{\mathrm{tr}}t}\sum_{j=0}^{N}(k_{\mathrm{tr}}t)^j/j!\).

Worked Example: Holding MTT Fixed While Changing N

Consider two hypothetical models, both with MTT \(=4\) hours. For \(N=0\), \(k_{\mathrm{tr}}=1/4=0.25\) per hour. The distribution is exponential, its standard deviation is \(4/\sqrt{1}=4\) hours, and its fraction of input by 4 hours is \(1-e^{-1}=0.6321\).

For \(N=3\), \(k_{\mathrm{tr}}=(3+1)/4=1\) per hour. The standard deviation is \(4/\sqrt{4}=2\) hours. The fraction of input by 4 hours is \(1-e^{-4}(1+4+4^2/2+4^3/6)=1-(0.018316)(23.6667)=0.5665\).

Both models have the same 4-hour mean, but the \(N=3\) model has half the standard deviation and a more concentrated distribution. Its fraction by the mean is not required to be greater: in this comparison it is 56.7%, versus 63.2% for \(N=0\). A larger \(N\) describes less relative dispersion, not a universal increase in the fraction absorbed at every time point.

Changing the Time Scale and Interpreting Input Rates

When \(N\) is fixed, changing MTT scales the transit-time distribution in time without changing its relative shape. For example, doubling MTT doubles the mean and standard deviation, but leaves the coefficient of variation unchanged. This follows from \(\operatorname{SD}(T)=\mathrm{MTT}/\sqrt{N+1}\). The next example connects the distribution to the rate at which a hypothetical dose enters the systemic circulation.

Worked Example: Peak Input Rate for a Hypothetical Dose

Suppose 100 mg is systemically available and the transit process has MTT \(=3\) hours and \(N=2\), as in the first example. The normalized input density is \(g(t)=t^2e^{-t}/2\) per hour. The systemic input rate is the available dose multiplied by this density: \(100g(t)\) mg per hour.

For \(N\geq1\), the mode is found by differentiating the log density: \(\log g(t)=(N+1)\log k_{\mathrm{tr}}+N\log t-k_{\mathrm{tr}}t-\log(N!)\). Setting its derivative \(N/t-k_{\mathrm{tr}}\) to zero gives \(t=N/k_{\mathrm{tr}}\). Here the mode is 2 hours. At that time, \(g(2)=2^2e^{-2}/2=2e^{-2}=0.2707\) per hour, so the peak systemic input rate is \(100(0.2707)=27.07\) mg per hour.

This is a peak input rate, not a prediction of peak plasma concentration. Concentration depends on how the body distributes and eliminates drug as input occurs.

Assumptions, Diagnostics, and Alternatives

The moment interpretation above relies on a specific structure: each stage has the same constant first-order transfer rate; successive stage waiting times are independent; the pathway is sequential rather than parallel; and the input distribution is time-invariant. It describes the timing of the modeled systemic input. Interpreting it as the timing of the full dose additionally assumes that the dose fraction assigned to that input pathway is known and that no unmodeled loss or competing absorption pathway changes the amount entering.

In practice, MTT and \(N\) can be difficult to distinguish from concentration data, especially if samples do not capture the absorption phase well. A long MTT can shift input later, while a change in \(N\) can alter the spread and peakiness without changing the mean. Drug disposition also affects observed concentrations, so the absorption density cannot generally be read directly from the concentration-time curve. Look for unstable estimates, strong dependence on initial values, or marked changes in MTT and \(N\) when a few early observations are removed; these can indicate that the data do not distinguish timing from dispersion well.

A fixed lag-time model represents a deterministic delay in its ideal form, rather than a spread of transit times. A single first-order absorption step corresponds to the \(N=0\) exponential case, with coefficient of variation 1. A transit model with \(N>0\) can represent a narrower relative distribution through sequential stages. These alternatives answer different structural questions; no one model is universally preferable. Model choice should reflect the dosing process, available observations, and whether the added shape flexibility is supported by data.

The integer-stage interpretation is most literal when \(N\) is a nonnegative integer. Some modeling implementations allow a continuous shape parameter to be estimated, which can represent an effective distribution shape but is not a literal count of physical compartments. Also, changing \(N\) while holding MTT fixed changes the rate associated with each stage. It is therefore incorrect to treat a change in \(N\) as adding compartments while leaving all other transit properties unchanged.

Common pitfall: “MTT is 4 hours, so the drug begins absorbing at 4 hours,” is incorrect. MTT is the average transit time; the distribution generally has input before and after its mean. Likewise, a larger \(N\) at fixed MTT does not mean a longer mean delay. It means a smaller relative spread under the stated sequential-stage model.

Key takeaway: MTT sets the average timing of the transit input. Under the convention used here, \(N\) determines the number of sequential stages beyond the single-stage case and controls relative dispersion: \(\operatorname{CV}(T)=1/\sqrt{N+1}\). Interpret both parameters through the input distribution, not as direct substitutes for observed concentration peak time.

Check Your Understanding

Answer each question, then compare with the concise solutions below.

For MTT \(=6\) hours and \(N=3\), what are \(k_{\mathrm{tr}}\), the standard deviation, and the coefficient of variation?
Two models have MTT \(=5\) hours, one with \(N=0\) and one with \(N=8\). Which has the smaller relative dispersion, and why?
For MTT \(=4\) hours and \(N=1\), find the mode of the transit-time density.
Why is MTT not generally equal to the time of peak plasma concentration?

Solutions. (1) \(k_{\mathrm{tr}}=(3+1)/6=0.6667\) per hour; \(\operatorname{SD}(T)=6/\sqrt{4}=3\) hours; \(\operatorname{CV}=1/\sqrt{4}=0.5\). (2) \(N=8\) has smaller relative dispersion because its coefficient of variation is \(1/\sqrt{9}=1/3\), compared with 1 for \(N=0\). (3) \(k_{\mathrm{tr}}=(1+1)/4=0.5\) per hour, so the mode is \(N/k_{\mathrm{tr}}=1/0.5=2\) hours. (4) MTT is the mean time of systemic input, whereas peak concentration also depends on disposition, including distribution and elimination.