Why Residence Time Must Be Translated Carefully
A regional mean residence time is a time scale; a first-order transit rate is a rate of transfer per unit time. They are related, but they are not interchangeable. In a segmental gut model, assigning a rate to each segment determines not only the expected time spent there, but also the spread of possible transfer times. The distinction matters when comparing regions or predicting how much drug has moved downstream by a particular time.
The introductory material on State Variables for Luminal Drug in Segmental Models establishes that each gut segment can be represented by a state describing the drug present in its lumen. Here the focus is narrower: given a segment mean residence time, what first-order transit rate corresponds to it, and what timing does that rate imply? This conversion is conditional on a particular residence-time model; a mean alone does not uniquely determine the full timing distribution.
From a Constant Exit Rate to a Mean Residence Time
A constant first-order rate means that, conditional on the drug still being in segment \(i\), its instantaneous chance of leaving during a short interval is proportional to the interval length and does not depend on how long it has already been there. Equivalently, the probability of remaining in the segment for at least \(t\) time units is \(S_i(t)=e^{-k_{\mathrm{tr},i}t}\), for \(t\geq 0\). The probability of having left by time \(t\) is therefore \(1-e^{-k_{\mathrm{tr},i}t}\).
For a nonnegative residence time, its mean equals the area under its survival function. Applying that identity gives the rate–mean relationship directly.
Under constant first-order exit, \(P(T_i>t)=e^{-k_{\mathrm{tr},i}t}\).
\(E[T_i]=\int_0^\infty P(T_i>t)\,dt=\int_0^\infty e^{-k_{\mathrm{tr},i}t}\,dt=1/k_{\mathrm{tr},i}\), provided \(k_{\mathrm{tr},i}>0\).
If the assumed-model mean residence time is \(\tau_i\), then \(k_{\mathrm{tr},i}=1/\tau_i\). The units are inverse time, such as \(\mathrm{h}^{-1}\) when \(\tau_i\) is in hours.
This derivation also identifies what the conversion does not establish. A measured or specified mean residence time does not, by itself, prove that residence times are exponentially distributed. The reciprocal is the first-order rate if the constant-hazard model is being used. That assumption implies a particular spread as well as a mean: an exponential residence time has variance \(1/k_{\mathrm{tr},i}^2\), equal to the square of its mean.
Worked Example: Converting One Segment’s Mean
Suppose a hypothetical small-intestinal segment is assigned a mean residence time of \(1.5\ \mathrm{h}\). Under a constant-hazard first-order model, calculate its transit rate and the probability that an amount has left the segment by \(1.5\ \mathrm{h}\).
The rate is the reciprocal of the mean:
The fraction transferred by time \(t\) is \(1-e^{-k_{\mathrm{tr}}t}\). At \(t=1.5\ \mathrm{h}\), the rate–time product is \(0.6667\ \mathrm{h}^{-1}\times1.5\ \mathrm{h}\approx1.000\), so:
Thus approximately 63.2% has left by the segment’s mean residence time under this model. The mean is not the time by which every molecule, or even exactly half of the amount, must have transferred.
Rates Across Sequential Segments
For segments passed through in sequence, each segment has its own mean \(\tau_i\) and rate \(k_{\mathrm{tr},i}=1/\tau_i\). If transit times in successive segments are represented as independent exponential waiting times, the expected time to pass through several segments is the sum of their means. For two segments, \(E[T_1+T_2]=\tau_1+\tau_2\). This sum is not generally the reciprocal of the sum of the rates: adding rates does not produce the overall mean time through the sequence.
The timing distribution for the sequence also differs from the distribution for one exponential residence with the same total mean. For two successive segments with distinct rates \(k_1\) and \(k_2\), the probability of reaching the segment after both by time \(t\) is:
This expression follows by integrating over the possible time \(u\) spent in the first segment: the first residence has density \(k_1e^{-k_1u}\), and the probability that the second residence is at most \(t-u\) is \(1-e^{-k_2(t-u)}\). Hence the probability is \(\int_0^t k_1e^{-k_1u}[1-e^{-k_2(t-u)}]\,du\), which evaluates to the displayed result. When the two rates are equal to \(k\), the limiting expression is \(1-e^{-kt}(1+kt)\).
Worked Example: Two Segments With Different Mean Times
Consider two hypothetical sequential gut segments. Let their mean residence times be \(1.5\ \mathrm{h}\) and \(3.0\ \mathrm{h}\), respectively. Find each transit rate, the expected time through both segments, and the fraction expected to have passed through both by \(3.0\ \mathrm{h}\).
The segment-specific rates are:
The expected total time is the sum of the segment means:
For timing by \(3.0\ \mathrm{h}\), use the two-segment expression. Substituting the unrounded rates, or their displayed approximations, gives:
Approximately 40.0% has passed through both segments by \(3.0\ \mathrm{h}\) under the stated assumptions. Notice that \(k_1+k_2=1.000\ \mathrm{h}^{-1}\), but its reciprocal, \(1.000\ \mathrm{h}\), is not the expected time through both segments; that expected time is \(4.5\ \mathrm{h}\).
Segment-Specific Rates Shape Transfer Timing
A larger \(k_{\mathrm{tr},i}\) means a shorter mean residence time in that segment, and a smaller \(k_{\mathrm{tr},i}\) means a longer one. The rate also controls the proportion transferred by any chosen time. For a single segment, the median residence time solves \(1-e^{-k_{\mathrm{tr},i}t}=0.5\), so it is \(\ln(2)/k_{\mathrm{tr},i}\), which is less than the mean \(1/k_{\mathrm{tr},i}\). Thus the rate affects timing throughout the distribution, not merely its average.
Worked Example: Comparing Transfer by a Fixed Time
Two hypothetical segments have mean residence times of \(0.5\ \mathrm{h}\) and \(2.0\ \mathrm{h}\). Under the same first-order assumption, compare the fractions transferred from each segment by \(0.5\ \mathrm{h}\).
The rates are \(k_A=1/(0.5\ \mathrm{h})=2.000\ \mathrm{h}^{-1}\) and \(k_B=1/(2.0\ \mathrm{h})=0.5000\ \mathrm{h}^{-1}\). Using \(1-e^{-kt}\):
The model therefore predicts about 63.2% transfer from segment A but 22.1% from segment B by \(0.5\ \mathrm{h}\). This comparison reflects the different assumed residence-time distributions; it is not a claim that the two regions have identical anatomy or that a rate alone describes absorption.
Assumptions, Diagnostics, and Alternatives
The reciprocal conversion is appropriate when the segment is represented by a well-mixed state with a constant first-order exit hazard, and when the supplied \(\tau_i\) refers to the mean residence time for that same segment and transit process. The simple sequential timing calculation additionally assumes that the drug passes through the segments in order, that the segment residence times are independent, and that there are no additional delays or routes included in the stated times. These assumptions clarify what the model’s rate represents; they do not guarantee that real transit follows this pattern.
A warning sign is a mismatch between the model-implied exponential timing and the available physiological information. For example, if observed residence times are tightly clustered around a typical value, strongly skewed in a way not compatible with an exponential distribution, or show distinct subpopulations, a single constant rate may reproduce the mean yet misrepresent transfer at early or late times. Also check units before inversion: a mean in minutes produces a rate in \(\mathrm{min}^{-1}\), not \(\mathrm{h}^{-1}\), unless it is converted first.
A fixed-delay representation places transfer at one specified time and has no residence-time spread; a first-order model spreads transfer continuously over time. More general residence-time distributions can represent other shapes, but require information beyond a mean if their shape is not fixed by assumption. These approaches are alternatives when the exponential implication is not suitable; the reciprocal rule should not be treated as a distribution-free way to recover a rate from any transit-time summary.
Check Your Understanding
Answer each question using the constant-hazard first-order model where specified. Solutions are included below.
- A segment has a mean residence time of \(4.0\ \mathrm{h}\). What is its first-order transit rate?
- For a segment with \(k_{\mathrm{tr}}=0.25\ \mathrm{h}^{-1}\), what fraction is predicted to leave by \(2.0\ \mathrm{h}\)?
- Two sequential segments have mean residence times of \(2.0\ \mathrm{h}\) and \(5.0\ \mathrm{h}\). What is the expected total time through both, and why is it not \(1/(1/2+1/5)\ \mathrm{h}\)?
- Give one reason why the reciprocal of a measured mean might fail to provide an adequate transit model.
Solutions. (1) \(k_{\mathrm{tr}}=1/(4.0\ \mathrm{h})=0.2500\ \mathrm{h}^{-1}\). (2) \(1-e^{-(0.25)(2.0)}=1-e^{-0.5}\approx0.3935\), or 39.35%. (3) The expected time is \(2.0+5.0=7.0\ \mathrm{h}\); sequential waiting times add in expectation, whereas adding their reciprocal means does not give the rate of a single equivalent exponential residence. (4) The actual residence-time distribution may not have a constant hazard or an exponential shape, so matching its mean alone may misrepresent transfer timing.