Resistance From Altered Target Sensitivity

Representing Altered Target Sensitivity

A tumor may respond less to a drug even when the exposure is unchanged. One compact representation is to keep a single tumor population but alter the parameters linking concentration to effect. A shift in potency means that a higher concentration is needed to produce a given effect. A reduction in maximal effect means that even very high concentrations cannot produce the same maximum effect. These representations describe different changes in the concentration–effect relationship; neither, by itself, specifies the biological cause of resistance.

To make the distinction concrete, let \(C\) be the effective drug concentration at the target, measured in nM, and let \(E(C)\) be the drug-induced tumor-cell killing rate, measured in day\(^{-1}\). Consider a saturable Emax function:

Here \(E_{\max}\) is the maximum killing rate and \(EC_{50}\) is the concentration producing half of that maximum. A potency shift changes \(EC_{50}\); a reduction in maximal effect changes \(E_{\max}\). For a less-sensitive target, the potency shift is typically upward, so \(EC_{50,R}>EC_{50,S}\), where \(S\) and \(R\) denote sensitive and less-sensitive parameter values. Alternatively, a reduced maximum has \(E_{\max,R}<E_{\max,S}\).

For the tumor dynamics, suppose untreated tumor growth follows a Gompertz law and drug-induced killing is proportional to tumor burden:

In this expression, \(N\) is tumor burden, \(r\) is a growth-rate parameter in day\(^{-1}\), and \(K\) is the Gompertz asymptotic size. The per-capita growth rate at the current burden is \(r\ln(K/N)\), while \(E(C)\) is the per-capita killing rate. The net per-capita rate is therefore \(r\ln(K/N)-E(C)\): a negative value predicts shrinkage at that instant, and a positive value predicts growth. This is a local prediction; it does not guarantee the same sign at later times if burden, concentration, or model parameters change.

The model assumes that concentration is known and adequately represents exposure at the target, the effect is immediate, and the same tumor growth law applies before and during treatment. It also assumes a single effective population, with killing proportional to burden. These are modeling choices, not universal biological facts. The earlier material on exposure-effect links and tumor-growth laws supplies the relevant building blocks; here the focus is how altered sensitivity changes their predictions.

How Potency and Maximum Effect Change Predictions

For a potency shift, write the sensitive and less-sensitive effects as:

Both curves have the same upper limit, \(E_{\max}\), but the less-sensitive curve produces less effect at any finite positive concentration. The difference becomes small at concentrations far above both \(EC_{50}\) values, where both effects approach the common maximum. By contrast, if sensitivity changes through a reduced maximum, then increasing concentration cannot restore the lost effect.

Deriving the Concentration Needed for a Target Effect

The Emax equation also makes the distinction useful for dose or exposure reasoning. If a desired killing rate \(E^*\) is below \(E_{\max}\), solve the Emax equation for the required concentration:

The algebra requires \(E_{\max}-E^*>0\). Thus, a finite positive concentration can achieve a positive target effect only when \(0<E^*<E_{\max}\). If \(E^*\geq E_{\max}\), the target is unattainable at finite concentration. Under a potency shift with unchanged \(E_{\max}\), the concentration required for any attainable target increases in direct proportion to \(EC_{50}\). Under a lower \(E_{\max}\), some targets that were previously attainable may become impossible.

These calculations assume the concentration can be increased as specified and that the fitted effect function remains valid over that range. They do not establish that the required concentration is safe, clinically achievable, or appropriate. A concentration–effect model describes a response relationship; it is not, by itself, a dosing recommendation.

Why a Resistant Clone Is a Different Representation

A distinct resistant clone represents two populations with separate burdens, rather than one population with a shifted effect curve. To isolate the difference, suppose both clones share the same per-capita growth rate \(g(N)\), but have different killing rates \(E_S(C)\) and \(E_R(C)\). Their equations are:

Adding the equations gives the total-burden rate. If \(f=N_R/N\) is the resistant fraction, then \(1-f=N_S/N\), and the instantaneous total per-capita killing rate is:

Thus, a clone mixture has a weighted effect at a particular concentration and composition. That weighted value can match a single altered-sensitivity model at one instant. But it need not match the same single Emax curve across concentrations, and the fraction \(f\) can change as the clones respond differently to treatment. A fixed shifted \(EC_{50}\) does not explicitly track that changing composition.

Model Checks, Comparisons, and Limitations

The choice between altered sensitivity and a distinct clone depends on the question and the available data. A single altered-sensitivity model is compact when the goal is to represent a changed concentration–effect relationship without tracking subpopulations. An explicit clone model is useful when clone fractions, clone-specific responses, or treatment-driven changes in composition are central to the prediction. The two representations can yield similar predictions over a limited exposure range, so a good fit alone does not necessarily identify the underlying mechanism.

Useful checks include plotting observed or simulated effect against concentration, checking whether the response approaches a plausible plateau, and verifying that units are consistent. If an estimated \(EC_{50}\) is far outside the observed concentration range, potency may be poorly identified. If data show a plateau below the expected maximum, a reduced \(E_{\max}\) may be a candidate explanation, but incomplete exposure coverage or other model misspecification can produce similar patterns. If the apparent concentration–effect relationship changes over time, a constant single-population parameter may fail to describe the data; a changing clone composition is one possible explanation, not the only one.

A further diagnostic is to compare predictions at multiple concentrations, times, and tumor burdens rather than judging the model from one endpoint. A model can match one measured tumor change while giving a different prediction under another exposure or after treatment continues. Inspect residual patterns and parameter uncertainty, and check whether the estimated maximum is supported by observations near saturation. Where the data do not cover a broad concentration range or do not distinguish populations, report that limitation rather than assigning a unique mechanism.

• Mistake: treating potency and maximum effect as interchangeable. Both can lower effect at a particular concentration, but a potency shift preserves the upper limit whereas a reduced maximum changes it.
• Mistake: inferring shrinkage from drug effect alone. Compare \(E(C)\) with the current per-capita growth rate; tumor burden shrinks instantaneously only if \(E(C)>r\ln(K/N)\) in the stated model.
• Mistake: interpreting a fitted shifted curve as proof of a resistant clone or a target alteration. A concentration–effect fit describes model behavior; biological mechanism requires supporting evidence.
• Mistake: assuming greater concentration can always overcome resistance. That may follow from a potency shift within the model, but it does not follow when the target effect is at or above a reduced \(E_{\max}\), and model calculations do not address safety.