MABEL from Biomarker Concentration-Response

Why Use a Biomarker Concentration-Response Relationship?

A pharmacologically linked biomarker can connect drug exposure to a measurable biological change. When the biomarker reflects the drug’s intended mechanism, its concentration-response relationship can help identify the exposure at which a minimal biological effect is anticipated. This provides a quantitative pharmacology-based input to first-in-human planning.

The key is to distinguish the effect threshold from a statistical test threshold. The effect threshold is a prespecified magnitude of biomarker change that is considered biologically meaningful or the smallest relevant response for the decision. It is not automatically the smallest change with a p-value below a chosen significance level. Nor does estimating an exposure at that threshold, by itself, establish that a dose is safe.

Definition: A biomarker-based MABEL exposure is the exposure predicted by a pharmacologically linked concentration-response relationship to produce a prespecified minimal biological effect. It is an exposure target, not necessarily a dose. Converting it to a dose requires a suitable exposure-dose relationship and careful alignment of the exposure metric, such as unbound concentration or Cmax, with the concentration-response data.

Specify the Response and the Concentration-Response Model

Let \(C\) denote the relevant drug concentration and let \(\Delta(C)\) denote the drug-related change in a biomarker at that concentration. The change should be defined consistently, for example as a baseline-adjusted change relative to a suitable control. If the biomarker decreases when the drug acts, define the response direction so that a positive \(\Delta(C)\) represents the magnitude of the pharmacological effect. The response and its units must be stated: a change of 6 percentage points is not the same as a 6% relative change.

A common model for a monotonic, saturable stimulatory response is the Hill Emax model:

\Delta(C)=\frac{E_{\max}C^h}{EC_{50}^h+C^h}

Here \(E_{\max}\) is the maximum modeled change, \(EC_{50}\) is the concentration associated with half of \(E_{\max}\), and \(h\) is the positive Hill coefficient, which controls the curve’s steepness. The concentration units for \(C\) and \(EC_{50}\) must match; \(\Delta(C)\) and \(E_{\max}\) share response units. This model assumes a positive concentration, a monotonic response that approaches a finite maximum, and a concentration-response relationship that is adequately represented by this form over the range used.

Before using the curve, establish that the biomarker is linked to the drug’s pharmacology, that the concentration corresponds to the response measurement in a meaningful way, and that the data inform the curve near the chosen effect threshold. A time-dependent biomarker may lag behind plasma concentration; in that case, pairing a single concentration with a response without accounting for the timing may produce a misleading curve. The introductory material on defining MABEL from mechanistic evidence motivates why the pharmacological link matters; here the emphasis is on extracting an exposure threshold from a measured concentration-response relationship.

Deriving the Minimal-Effect Concentration

Let \(\Delta_{\min}\) be the prespecified minimal effect. For the Hill model, a finite concentration that reaches this effect exists only when \(0<\Delta_{\min}<E_{\max}\). Set the modeled response equal to the threshold and solve:

\Delta_{\min}=\frac{E_{\max}C^h}{EC_{50}^h+C^h}

Divide by \(E_{\max}\), and define \(r=\Delta_{\min}/E_{\max}\). Rearranging gives \(r(EC_{50}^h+C^h)=C^h\), so \(rEC_{50}^h=(1-r)C^h\). Since \(0<r<1\), divide by \(1-r\), then take the positive \(h\)th root:

C_{\mathrm{MABEL}}=EC_{50}\left(\frac{r}{1-r}\right)^{1/h} =EC_{50}\left(\frac{\Delta_{\min}}{E_{\max}-\Delta_{\min}}\right)^{1/h}

The result has the units of concentration because the expression in parentheses is dimensionless. If \(\Delta_{\min}\geq E_{\max}\), the model cannot reach the threshold at any finite concentration when equality holds only asymptotically; a threshold above \(E_{\max}\) is unreachable under the model. If the threshold is zero or negative, it does not define a positive minimal-effect concentration in this formulation.

For a threshold below half of \(E_{\max}\), \(r/(1-r)<1\). Increasing \(h\) makes the \(1/h\) power of this ratio larger, moving the corresponding concentration upward toward \(EC_{50}\). Thus a steeper curve does not automatically imply a lower concentration for the same below-half-maximal threshold.

Worked Example: Inverting a Hill Curve for a Minimal Effect

Suppose a hypothetical biomarker response is described by \(E_{\max}=60\) percentage points, \(EC_{50}=3\) nM, and \(h=1\). The prespecified minimal effect is 6 percentage points. Assume the biomarker is pharmacologically linked, the concentration and response are appropriately paired, and the observed data support the model near this threshold.

First calculate the response fraction, \(r=6/60=0.1\). Applying the inverse:

C_{\mathrm{MABEL}}=3\left(\frac{0.1}{1-0.1}\right)^{1/1} =3\left(\frac{1}{9}\right) =\frac{1}{3}\ \mathrm{nM}

The predicted minimal-effect concentration is therefore \(1/3\) nM. Substituting the unrounded value back into the response model gives:

\Delta\left(\frac{1}{3}\right) =\frac{60(1/3)}{3+1/3} =\frac{20}{10/3} =6\ \text{percentage points}

This verifies the inversion. Reporting a rounded concentration for presentation is acceptable, but the substitution check uses the unrounded value so rounding does not create an apparent discrepancy.

Assumptions, Diagnostics, and Failure Modes

A calculated concentration is useful only to the extent that the model and evidence support it. Review the following before interpreting it as a MABEL exposure:

Pharmacological relevance: The biomarker should reflect a response linked to the drug’s mechanism, rather than merely changing alongside drug exposure for an unrelated reason.
Response definition: Baseline, control adjustment, response direction, measurement time, and units should be consistent. An inconsistent baseline or assay scale changes the threshold being modeled.
Concentration metric: Use the concentration measure that is relevant to the assay and biological response. Do not silently substitute total for unbound concentration or a plasma peak for a different exposure measure.
Curve support: The data should cover concentrations informative about the threshold. If the threshold lies far below the lowest tested concentration, its estimate is an extrapolation and can depend strongly on the model form.
Model adequacy: Plot the observations and fitted curve. Look for systematic departures, nonmonotonicity, unexplained time dependence, or influential observations. A visually poor fit weakens the meaning of the inverted threshold.
Parameter uncertainty: \(E_{\max}\), \(EC_{50}\), and \(h\) are estimated quantities. Uncertainty in them propagates to \(C_{\mathrm{MABEL}}\); a single point estimate should not be mistaken for a precisely known biological boundary.

A particularly important failure mode is choosing a threshold that is smaller than the assay can reliably distinguish from background. The model may return a precise-looking concentration even when the underlying response data cannot resolve such a small effect. Another is using a single concentration-response curve when the biomarker has a delayed response or marked hysteresis: response may then depend on exposure history, not just the concentration at the sampling time.

Worked Example: How the Hill Coefficient Changes the Result

Consider a hypothetical response with \(E_{\max}=80\) percentage points, \(EC_{50}=4\) nM, and a minimal effect of 8 percentage points. The response fraction is \(r=8/80=0.1\), which is below one-half. Compare \(h=1\) with \(h=2\), assuming the same response definition and otherwise adequate model support.

For \(h=1\):

C_{\mathrm{MABEL},1} =4\left(\frac{0.1}{0.9}\right) =4\left(\frac{1}{9}\right) =\frac{4}{9} \approx 0.4444\ \mathrm{nM}

For \(h=2\):

C_{\mathrm{MABEL},2} =4\left(\frac{1}{9}\right)^{1/2} =\frac{4}{3} \approx 1.333\ \mathrm{nM}

With the same below-half-maximal effect threshold and \(EC_{50}\), the larger Hill coefficient gives a higher concentration, closer to \(EC_{50}=4\) nM. This follows because \(1/9\) is below 1, and its square root is greater than \(1/9\). A steepness estimate therefore matters to the inferred minimal-effect concentration; it should be supported by data rather than chosen to obtain a preferred result.

From Exposure Target to a Dose: A Separate Step

The concentration-response model identifies an exposure associated with the chosen effect. It does not, by itself, state what administered dose produces that exposure in humans. Dose translation needs an exposure-dose relationship appropriate to the route, regimen, and metric used. If pharmacokinetics are nonlinear, or if the concentration-response assay uses a different concentration metric, a simple proportional conversion may not be valid.

For a transparent illustration, suppose a hypothetical pharmacokinetic prediction relates dose \(D\) in mg to the relevant peak concentration by \(C_{\max}=0.010D\) nM, over the dose range considered. This is a stated linear exposure-dose assumption, not a general rule. The predicted dose corresponding to a biomarker-derived peak concentration can then be found by dividing the target concentration by the proportionality constant.

Worked Example: Translating a Biomarker Exposure to a Dose

Suppose the fitted biomarker model has \(E_{\max}=40\) percentage points, \(EC_{50}=2\) nM, \(h=1\), and a prespecified minimum effect of 4 percentage points. Assume the pharmacokinetic model predicts the relevant peak concentration as \(C_{\max}=0.010D\) nM, where \(D\) is dose in mg. For this hypothetical calculation, assume that peak concentration is the appropriate exposure metric for the biomarker response and that proportionality holds over the range used.

The response fraction is \(r=4/40=0.1\). Thus:

C_{\mathrm{MABEL}} =2\left(\frac{0.1}{0.9}\right) =\frac{2}{9} \approx 0.2222\ \mathrm{nM}

Set the predicted peak concentration equal to this exposure:

0.010D=\frac{2}{9} \qquad\Longrightarrow\qquad D=\frac{2/9}{0.010} =22.22\ \mathrm{mg}

Under the stated assumptions, 22.22 mg is the dose predicted to produce the biomarker-derived peak concentration. It is not, on this calculation alone, a recommended starting dose or a claim of safety. If the exposure-dose relationship, peak metric, or concentration-response timing is inappropriate, this dose translation is not justified.

How This Approach Fits with Other Pharmacology Evidence

A biomarker concentration-response curve is most useful when the biomarker is measurable and meaningfully connected to the intended pharmacology. It can provide an integrated biological readout rather than relying only on binding or receptor occupancy. By contrast, receptor-occupancy and in vitro potency approaches use different measurements and assumptions; they can provide complementary evidence, but their numerical thresholds should not be treated as interchangeable. Agreement among independent lines of pharmacological evidence can strengthen confidence, while disagreement calls for examining what each assay measures, its concentration metric, and its biological context.

The approach also has limitations. A biomarker may be an indirect or insensitive readout, the selected effect threshold may be uncertain, and a fitted curve can be poorly identified if the data do not span the relevant response range. A concentration-response relationship established in one experimental setting may not directly predict human response if exposure, assay conditions, or biological context differ. Therefore, report the response definition, model, threshold, concentration metric, and uncertainty alongside the result.

Common mistake: “The curve says 0.22 nM, so the first-in-human dose is 0.22 mg.” This confuses concentration with dose; they have different units and are related only through pharmacokinetics. First calculate the biomarker-based exposure, then use an appropriate exposure-dose model to translate it. The result remains conditional on the assumptions and does not alone establish a safe starting dose.

Check Your Understanding

Answer each question using the model and distinctions developed above. Solutions follow the questions.

For \(E_{\max}=50\) percentage points, \(EC_{50}=5\) nM, \(h=1\), and \(\Delta_{\min}=5\) percentage points, calculate the minimal-effect concentration.
For a threshold below half of \(E_{\max}\), does increasing \(h\) increase or decrease the concentration predicted by the inverse Hill model? Briefly justify.
Why is a statistically significant biomarker change not automatically the appropriate minimal biological effect?
A concentration-response model uses unbound concentration, but the exposure-dose model predicts total plasma concentration. What must be addressed before translating the target exposure into a dose?

Solutions. (1) \(r=5/50=0.1\), so \(C_{\mathrm{MABEL}}=5(0.1/0.9)=5/9\approx0.5556\) nM. (2) It increases the predicted concentration toward \(EC_{50}\): when \(r<0.5\), \(r/(1-r)<1\), and increasing \(h\) makes its \(1/h\) power larger. (3) Statistical significance depends on variability, sample size, and the test; it does not by itself define the smallest biologically relevant effect. (4) The relationship between total and unbound concentration, and whether the relevant exposure metric is preserved, must be established before dose translation.