Logistic Tumor Growth and Carrying Capacity

Why Growth May Slow as Tumor Burden Increases

Exponential growth assumes that the tumor’s per-capita growth rate stays constant: each unit of tumor burden contributes the same expected amount of additional burden per unit time, regardless of total size. That assumption can be useful over a limited range. But when resources, space, or other conditions constrain net growth, a model may need growth to slow as burden increases. The logistic model represents that slowdown with a simple, explicit dependence on current burden.

Following the earlier material on mapping viable tumor cell burden to measured tumor volume, let \(B(t)\) denote the modeled tumor burden at time \(t\). The burden might be viable cell number or a volume measure used as a proxy for it, provided that the relationship between the modeled burden and the measurement is appropriate for the analysis. The logistic model describes the rate of change of this burden; it does not, by itself, specify how a measured image becomes a burden value.

Definition: The logistic tumor-growth model is

\frac{dB}{dt}=rB\left(1-\frac{B}{K}\right),

where \(B(t)\) is tumor burden, \(r\) is the low-burden per-capita net growth rate, and \(K\) is the carrying capacity. For burden measured in \(\mathrm{mm}^3\) and time in days, \(r\) has units \(\mathrm{day}^{-1}\), and \(K\) has units \(\mathrm{mm}^3\). The factor \(1-B/K\) is dimensionless.

The model is usually applied with \(r>0\), \(K>0\), and a positive initial burden. It assumes that \(r\) and \(K\) remain constant over the modeled period, that burden is represented by a single state variable, and that there is no separate treatment input or source of new tumor burden in the equation. It is deterministic and continuous in time. These are modeling assumptions, not claims that tumor biology is truly constant or free of variation.

What Carrying Capacity Means in This Model

The term \(1-B/K\) reduces the net growth rate as \(B\) approaches \(K\). When \(B\) is small relative to \(K\), the factor is near one, so the rate is approximately \(rB\), as in exponential growth. At \(B=K\), the model predicts zero net growth. If \(B\) exceeds \(K\), it predicts a negative rate, which pushes burden back toward \(K\) under the model.

Thus, \(K\) is the positive equilibrium burden implied by the equation. It is not necessarily the largest tumor ever observed, a fixed anatomical limit, or an intrinsic constant that applies across patients and conditions. Its value may summarize several unmodeled influences on net growth. A change in tumor environment, measurement definition, or treatment context may make the same fitted \(K\) unsuitable. The model’s equilibrium interpretation is conditional on the assumptions that \(r\) and \(K\) are constant and that the logistic equation adequately describes the process.

Worked Example: Calculating Growth Below Carrying Capacity

Suppose a hypothetical tumor is modeled with \(r=0.30\ \mathrm{day}^{-1}\) and \(K=1{,}000\ \mathrm{mm}^3\). Calculate its instantaneous absolute and per-capita growth rates when \(B=200\ \mathrm{mm}^3\).

Substitute the values into the logistic rate law:

\frac{dB}{dt} =0.30\ \mathrm{day}^{-1}\times 200\ \mathrm{mm}^3 \left(1-\frac{200\ \mathrm{mm}^3}{1{,}000\ \mathrm{mm}^3}\right) =60\times 0.80 =48\ \mathrm{mm}^3/\mathrm{day}.

The per-capita rate is the absolute rate divided by burden:

\frac{1}{B}\frac{dB}{dt} =\frac{48\ \mathrm{mm}^3/\mathrm{day}}{200\ \mathrm{mm}^3} =0.24\ \mathrm{day}^{-1}.

At this burden, the model predicts an instantaneous increase of \(48\ \mathrm{mm}^3\) per day and a per-capita net growth rate of \(0.24\ \mathrm{day}^{-1}\). These are rates at the specified burden, not a claim that the burden increases by exactly \(48\ \mathrm{mm}^3\) during every subsequent day; the rate changes as \(B\) changes.

Why the Logistic Form Slows Growth

One justification for the logistic form starts with a simple growth assumption: the number of net growth opportunities is proportional to burden, but the fraction of opportunities that succeed declines as burden occupies more of a limiting environment. Represent that available fraction as \(1-B/K\). Multiplying burden by the low-burden per-capita rate \(r\) and by this fraction gives the logistic rate law. This is a compact phenomenological assumption; it does not identify which biological mechanism causes the limitation.

Dividing the rate law by \(B\), for \(B>0\), makes the per-capita assumption explicit:

\frac{1}{B}\frac{dB}{dt} =r\left(1-\frac{B}{K}\right) =r-\frac{r}{K}B.

The per-capita rate therefore declines linearly with burden under this model. The absolute growth rate, \(rB(1-B/K)\), does not decline across the entire range below \(K\): burden is also increasing. In fact, it is greatest at \(B=K/2\). Differentiating the absolute rate with respect to \(B\) gives \(r(1-2B/K)\), which is zero at \(B=K/2\); the rate increases before that point and decreases after it. Confusing declining per-capita growth with declining absolute growth can lead to an incorrect interpretation of tumor measurements.

Worked Example: Comparing Burdens Along the Logistic Curve

Use \(r=0.30\ \mathrm{day}^{-1}\) and \(K=1{,}000\ \mathrm{mm}^3\) to compare the instantaneous absolute growth rate at \(B=200\), \(500\), and \(800\ \mathrm{mm}^3\).

At \(B=200\ \mathrm{mm}^3\), the calculation is \(0.30\times200\times(1-200/1{,}000)=48\ \mathrm{mm}^3/\mathrm{day}\). At \(B=500\ \mathrm{mm}^3\), it is \(0.30\times500\times(1-500/1{,}000)=75\ \mathrm{mm}^3/\mathrm{day}\). At \(B=800\ \mathrm{mm}^3\), it is \(0.30\times800\times(1-800/1{,}000)=48\ \mathrm{mm}^3/\mathrm{day}\).

The absolute rates are \(48\), \(75\), and \(48\ \mathrm{mm}^3/\mathrm{day}\), respectively. The model’s per-capita rates at those burdens are \(0.24\), \(0.15\), and \(0.06\ \mathrm{day}^{-1}\). So per-capita growth slows at each step, while absolute growth first rises and then falls. The equal absolute rates at \(200\) and \(800\ \mathrm{mm}^3\) do not imply equal tumor states or equal future growth trajectories.

Interpreting Parameters and Using Data

The parameter \(r\) is the per-capita growth rate as burden approaches zero in the logistic model. It is not the per-capita rate at every burden, and it is not necessarily a directly measured cell-division rate: the equation describes net growth. The parameter \(K\) marks where the model’s net rate is zero. Their interpretation depends on the chosen burden scale and time unit. A change from days to hours changes the numerical value of \(r\), while changing burden units changes the numerical value of \(K\).

If estimated per-capita rates are available at two burdens, the model gives a way to check whether a linear decline with burden is plausible. For a short observation interval, a common approximation to the per-capita rate is the change in log burden divided by elapsed time. Such estimates can be noisy, especially when measurement error is substantial or intervals are long. They should not be treated as exact instantaneous rates without justification.

Worked Example: Inferring Logistic Parameters From Two Rates

Suppose hypothetical short-interval data suggest per-capita rates of \(0.24\ \mathrm{day}^{-1}\) at \(B=200\ \mathrm{mm}^3\) and \(0.12\ \mathrm{day}^{-1}\) at \(B=600\ \mathrm{mm}^3\). Assume the rates are adequately estimated and the logistic model applies over this burden range. Estimate \(r\) and \(K\).

Under the model, the per-capita rate \(g(B)\) is a straight line: \(g(B)=r-(r/K)B\). The slope based on the two rates is

\frac{0.12-0.24}{600-200} =-0.000300\ \mathrm{day}^{-1}/\mathrm{mm}^3.

The intercept is the rate at zero burden. Using the first point, it is

r=0.24-(-0.000300)(200) =0.300\ \mathrm{day}^{-1}.

Since the slope is \(-r/K\), the implied carrying capacity is

K=\frac{r}{0.000300\ \mathrm{day}^{-1}/\mathrm{mm}^3} =\frac{0.300\ \mathrm{day}^{-1}}{0.000300\ \mathrm{day}^{-1}/\mathrm{mm}^3} =1{,}000\ \mathrm{mm}^3.

These calculations recover the parameters implied by the two supplied rates. They do not establish that the logistic model is biologically correct. With only two points, the straight-line relationship is determined exactly and cannot be checked against additional observations; uncertainty in either rate would also affect the estimates.

Model Checks, Alternatives, and Limitations

The exponential model is a useful comparison: it has \(dB/dt=rB\), so per-capita growth does not depend on burden. Logistic growth adds a specific form of size-dependent feedback, making per-capita growth decline linearly with burden. More mechanistic approaches can represent particular constraints or processes explicitly, but require suitable state variables and data. The logistic equation is comparatively compact; that convenience comes at the cost of not identifying the biological source of the slowing.

In practice, a fitted carrying capacity can be poorly determined when all observed burdens are small relative to \(K\). In that range, \(1-B/K\) is close to one, so many large values of \(K\) may produce predictions close to exponential growth. A narrow burden range, sparse measurements, noisy volumes, or changes in treatment and tumor composition can also make parameter estimates unstable or undermine the constant-parameter assumptions.

Useful checks include plotting observed burden over time against model predictions, inspecting whether residuals show systematic curvature, and checking whether the implied per-capita rate actually falls approximately linearly with burden. Repeatedly underpredicting high-burden observations or overpredicting low-burden observations may signal a poor fit. Also check that fitted \(r\) and \(K\) are positive and have the expected units. A numerical estimate of \(K\) is not strong evidence for a meaningful plateau if the data never approach the range where the fitted feedback matters.

A common mistake is to call \(K\) the tumor’s known maximum size because the fitted curve levels off there. The correction is that \(K\) is the equilibrium implied by a particular model under its assumptions. Data may not identify it precisely, and its value need not represent a universal physical limit. A second mistake is to conclude that absolute growth must slow whenever per-capita growth slows. The worked comparison shows why that is false: absolute growth depends on both burden and per-capita rate.

Key takeaway: Logistic growth is exponential-like at low burden and has declining per-capita growth as burden rises. Its carrying capacity \(K\) is the model’s condition-dependent zero-growth equilibrium, not automatically a measured or universal tumor-size limit.

Check Your Understanding

Answer each question, then compare your reasoning with the solutions below.

For \(r=0.20\ \mathrm{day}^{-1}\), \(K=800\ \mathrm{mm}^3\), and \(B=400\ \mathrm{mm}^3\), calculate the instantaneous absolute growth rate and the per-capita rate.
In the logistic model, what does the rate law predict at \(B=K\), and what does that prediction mean?
At \(B=0.1K\), is the per-capita rate greater than or less than \(r\)? Calculate it as a fraction of \(r\).
Why might data collected only at burdens far below an estimated \(K\) provide weak evidence about the value of \(K\)?

Solutions. (1) \(dB/dt=0.20\times400\times(1-400/800)=40\ \mathrm{mm}^3/\mathrm{day}\); the per-capita rate is \(40/400=0.10\ \mathrm{day}^{-1}\). (2) At \(B=K\), \(dB/dt=rK(1-1)=0\), so the model predicts zero net growth at that burden, not that a measured tumor must stop growing there. (3) The rate is \(r(1-0.1)=0.9r\), which is less than \(r\). (4) When \(B\ll K\), the factor \(1-B/K\) is close to one, so a wide range of large \(K\) values can yield nearly exponential predictions over the observed range.