Tumor Shrinkage and Regrowth During Intermittent Therapy

Why Separate Treatment and Drug-Free Time?

An intermittent regimen alternates periods when drug is present with periods when it is absent or has negligible effect. Tumor burden can shrink during treatment and then regrow during a drug-free interval. Consequently, observing shrinkage during a dosing period does not by itself establish that the regimen reduces tumor burden over a complete cycle. The timing and duration of both intervals matter.

Here, \(V(t)\) denotes modeled tumor burden, expressed in a consistent unit such as cubic centimeters. The model treats the net tumor growth rate and the drug-induced kill rate as constant within each interval. It therefore provides a transparent way to calculate trajectories under specified rates and schedules. It does not, by itself, explain how drug exposure produces a kill rate; that exposure-effect link is taken as given.

Model and assumptions: During treatment, tumor burden follows \(dV/dt=(g-k)V\), where \(g\) is the untreated fractional growth rate and \(k\) is the effective fractional kill rate. During a drug-free interval, \(dV/dt=gV\). The rates have units of inverse time, \(V(0)>0\), and burden remains positive. Within each interval, rates are constant; the rate changes instantaneously at the interval boundary. This model assumes exponential dynamics within each phase, no treatment delay, no persistent drug effect during the drug-free interval, and no change in tumor sensitivity or growth characteristics over the calculation.

Deriving the Trajectory Across One Cycle

Let \(\tau\) be the treatment duration and \(\rho\) the subsequent drug-free duration. For a constant rate \(r\), the differential equation \(dV/dt=rV\) has solution \(V(t)=V(0)e^{rt}\). Applying this result separately to the two intervals gives the treatment endpoint and the next cycle endpoint.

V_{\mathrm{on}}=V_0e^{(g-k)\tau}, \qquad V_{\mathrm{next}}=V_{\mathrm{on}}e^{g\rho} =V_0e^{(g-k)\tau+g\rho}.

The cycle multiplication factor is \(e^{(g-k)\tau+g\rho}\). It is below 1 when the exponent is negative, equal to 1 when the exponent is zero, and above 1 when the exponent is positive. Thus, the direction of change across a cycle depends on the accumulated growth and kill over both phases, not only on the treatment-phase rate.

(g-k)\tau+g\rho<0 \quad\Longleftrightarrow\quad k\tau>g(\tau+\rho).

When treatment duration is positive, this condition can also be written \(k>g(1+\rho/\tau)\). It is more demanding than \(k>g\), the condition for shrinkage during treatment alone: kill must offset growth during both the treatment period and the subsequent drug-free period. With the same \(\tau\) and \(\rho\) in every cycle, the burden after \(m\) cycles is \(V_m=V_0e^{m[(g-k)\tau+g\rho]}\). The formula follows by multiplying the same cycle factor \(m\) times.

Duration result: If the drug-free duration \(\rho\) is fixed and \(k>g\), a cycle declines when \(\tau>g\rho/(k-g)\). Increasing treatment duration adds time at the net rate \(g-k\), which is negative under this condition. This threshold does not apply when \(k\leq g\): in that case, extending treatment cannot make the treatment phase shrink under this constant-rate model.

Worked Example: Calculating Shrinkage and Regrowth in One Cycle

Suppose a hypothetical tumor has initial modeled burden \(V_0=1.00\) cm³, untreated growth rate \(g=0.018\) day⁻¹, and treatment kill rate \(k=0.030\) day⁻¹. Treatment lasts 14 days, followed by 14 drug-free days. Assume the stated rates apply throughout their respective intervals.

During treatment, the net rate is \(g-k=0.018-0.030=-0.012\) day⁻¹. The burden at day 14 is

V_{\mathrm{on}} =1.00e^{(-0.012)(14)} =1.00e^{-0.168} \approx 0.8454\ \text{cm}^3.

Thus, the modeled burden decreases by about \(0.1546\) cm³ during treatment. Over the next 14 days without drug effect, the burden grows by a factor of \(e^{(0.018)(14)}=e^{0.252}\approx1.287\). The end-of-cycle burden is

V_{\mathrm{next}} =0.8454e^{0.252} =1.00e^{0.084} \approx1.088\ \text{cm}^3.

Despite shrinkage during treatment, the burden at the end of the full 28-day cycle is about \(8.8\%\) above its starting value. The cycle exponent is \((-0.012)(14)+(0.018)(14)=0.084>0\), confirming that result.

How Treatment Duration Changes the Cycle Outcome

If the drug-free duration is held fixed, extending treatment changes the cycle exponent by \((g-k)\) times the additional treatment duration. When \(k>g\), this change is negative: longer treatment lowers the cycle endpoint relative to the same starting burden. It may change a growing cycle into a declining one, but does not guarantee tumor elimination. The model predicts positive burden at every finite time when it starts positive.

Worked Example: Finding the Required Treatment Duration

Consider a hypothetical tumor with \(g=0.020\) day⁻¹ and treatment kill rate \(k=0.035\) day⁻¹. The drug-free interval is fixed at \(\rho=14\) days. First consider \(\tau=10\) treatment days.

The cycle exponent is

(g-k)\tau+g\rho =(-0.015)(10)+(0.020)(14) =-0.150+0.280 =0.130.

The cycle factor is \(e^{0.130}\approx1.139\): the tumor shrinks during treatment, since \(g-k=-0.015\) day⁻¹, but the modeled burden increases about \(13.9\%\) over the full cycle.

To find the duration at which the cycle just stops increasing, set the exponent to zero and solve for \(\tau\):

(g-k)\tau+g\rho=0, \qquad \tau=\frac{g\rho}{k-g} =\frac{(0.020)(14)}{0.035-0.020} =\frac{0.280}{0.015} \approx18.67\ \text{days}.

Thus, with a fixed 14-day drug-free interval, treatment must last longer than about 18.67 days for the cycle factor to be below 1. For a 20-day treatment, the exponent is \((-0.015)(20)+(0.020)(14)=-0.020\), giving a factor of \(e^{-0.020}\approx0.9802\), or about a \(2.0\%\) decline per cycle. This comparison assumes the kill rate remains constant for the entire extended treatment.

The fixed drug-free interval matters to this result. If the total cycle length is fixed instead, increasing \(\tau\) necessarily decreases \(\rho\). In that case, the schedule comparison must use the actual durations of both phases; one should not apply the fixed-\(\rho\) threshold while silently changing the off-treatment time.

Worked Example: Following Several Repeated Cycles

For a hypothetical tumor, let \(V_0=2.00\) cm³, \(g=0.025\) day⁻¹, and \(k=0.045\) day⁻¹. Use 14 treatment days followed by 7 drug-free days, repeating the same schedule three times. The model assumes these rates stay unchanged over all three cycles.

The treatment-phase exponent is \((0.025-0.045)(14)=-0.280\), so each treatment period multiplies burden by \(e^{-0.280}\approx0.7558\). The drug-free exponent is \((0.025)(7)=0.175\), giving a regrowth factor of \(e^{0.175}\approx1.191\). The full-cycle exponent and factor are

(-0.280)+0.175=-0.105, \qquad e^{-0.105}\approx0.9003.

The cycle-end burden falls by about \(10.0\%\) per cycle. After three complete cycles, or 63 days, it is

V_3=2.00e^{3(-0.105)} =2.00e^{-0.315} \approx1.460\ \text{cm}^3.

The cycle endpoint is not the minimum burden in each cycle. At the end of the third treatment period, day 56, the cumulative exponent is two full cycles plus the third treatment period: \(2(-0.105)+(-0.280)=-0.490\). The burden is \(2.00e^{-0.490}\approx1.225\) cm³, then it regrows to about \(1.460\) cm³ by day 63. Reporting only cycle endpoints would miss this within-cycle trough.

Assumptions, Diagnostics, and Alternatives

The piecewise calculation is appropriate when the question supplies effective rates and interval durations and asks for the resulting trajectory under those assumptions. Before interpreting its predictions, check that the rate units match the time units, that the stated \(k\) is the fractional kill rate for the treatment interval, and that drug effect is genuinely assumed to be absent during the off interval. If rates are given per week, for example, durations must be converted to weeks before using the exponent.

Check the phase boundary. At the end of treatment, burden should be multiplied by \(e^{(g-k)\tau}\); during the drug-free interval, it should be multiplied by \(e^{g\rho}\). Reversing the rates or durations changes the predicted trajectory.
Check signs and limiting cases. If \(k=0\), both phases should give untreated growth. If \(g=k\), burden is constant during treatment. If \(\rho=0\), the cycle exponent should reduce to \((g-k)\tau\).
Compare within-cycle and cycle-end trends. A declining treatment phase does not imply a declining full cycle. Calculate the total exponent, and inspect the treatment endpoint as well as the cycle endpoint when troughs or peaks matter.
Check the schedule and model against the intended use. Changing rates across cycles, residual drug effect after dosing stops, or a changing tumor growth rate would violate the constant-rate assumptions. A numerical prediction can be computed correctly and still be unsuitable if these assumptions are implausible.

A continuous constant-rate approximation can summarize a repeating schedule using the average net rate over a cycle, \(\{(g-k)\tau+g\rho\}/(\tau+\rho)\). Its exponential prediction at cycle endpoints agrees with the piecewise model when the same rates and durations repeat. However, the average rate does not describe the within-cycle shrinkage and regrowth. A more detailed time-varying model or numerical integration may be needed when rates vary continuously, drug effect persists after dosing, or a more complex growth law is required. Such models require additional information and assumptions; they cannot be recovered from the two constant rates alone.

The model also has biological limitations. It assumes a single burden with a shared growth and kill rate, does not represent changes in sensitivity, and extrapolates exponential growth without a size-dependent limit. Tumor volume may be used as the modeled burden, but the calculation does not establish that measured volume is a direct count of viable cells. These limitations matter especially when projecting across many cycles or beyond the time range supporting the assumed rates.

A Common Mistake

A common error is to see \(k>g\), conclude that the tumor must decline over the regimen, and ignore time off treatment. That inequality establishes only that \(g-k<0\) during treatment. For example, the cycle exponent also includes \(g\rho\), which can outweigh treatment-phase shrinkage. The correction is to calculate \((g-k)\tau+g\rho\) for the full cycle and interpret its sign. Another mistake is to treat a longer treatment duration as automatically beneficial without stating what happens to the drug-free duration; a schedule comparison must specify both.

Check Your Understanding

Use the piecewise constant-rate model above. Assume each stated rate applies throughout its named interval, and give rates and durations in matching time units.

A tumor has \(g=0.010\) day⁻¹, \(k=0.025\) day⁻¹, \(\tau=8\) days, and \(\rho=12\) days. What is the cycle factor, and does the burden decline over the full cycle?
For \(g=0.030\) day^{−1}\), \(k=0.020\) day⁻¹, \(\tau=15\) days, and \(\rho=5\) days, does the burden shrink during treatment? Does it decline over a full cycle?}
With \(g=0.015\) day⁻¹, \(k=0.027\) day⁻¹, and fixed \(\rho=14\) days, find the treatment duration that makes the cycle factor equal to 1.
A tumor begins at \(2.40\) cm³ and has \(g=0.020\) day⁻¹, \(k=0.050\) day⁻¹, \(\tau=10\) days, and \(\rho=5\) days. Calculate its burden at the end of treatment and at the end of the cycle.

Solutions. (1) The exponent is \((0.010-0.025)(8)+(0.010)(12)=-0.120+0.120=0\), so the cycle factor is \(e^0=1\); the cycle-end burden is unchanged. (2) The treatment net rate is \(0.030-0.020=0.010\) day⁻¹, so burden increases during treatment. The cycle exponent is \((0.010)(15)+(0.030)(5)=0.300\), giving factor \(e^{0.300}\approx1.350\); the burden also increases over the full cycle. (3) Set the cycle exponent to zero: \(\tau=g\rho/(k-g)=(0.015)(14)/(0.027-0.015)=0.210/0.012=17.50\) days. (4) At treatment end, \(V_{\mathrm{on}}=2.40e^{(0.020-0.050)(10)}=2.40e^{-0.300}\approx1.778\) cm³. The drug-free interval multiplies this by \(e^{(0.020)(5)}=e^{0.100}\); equivalently, the cycle exponent is \(-0.300+0.100=-0.200\). Therefore \(V_{\mathrm{next}}=2.40e^{-0.200}\approx1.965\) cm³.