3.2 Dose-Response Relationships

Dose-response curves are the cornerstone of quantitative pharmacology. They quantify how drug concentration relates to biological effect at the individual level (graded) and population level (quantal), enabling precise comparisons of drug potency, efficacy, and safety.

Historical Context

The dose-response concept dates back to Paracelsus (1493-1541): "The dose makes the poison." Quantitative dose-response analysis was formalized by Clark (1933) using mass-action kinetics and extended by Hill (1910) who introduced cooperativity. Trevan (1927) developed the LD_50 concept for standardizing drug toxicity in populations.

Derivation 1: The Hill Equation & Log-Dose Response Curve

The sigmoidal dose-response curve on a log scale arises naturally from the Hill equation, which generalizes simple occupancy to account for cooperativity.

Step 1: Start with Clark's Occupancy

For a single binding site with no cooperativity (n=1):

\( E = E_{max} \cdot \frac{[A]}{EC_{50} + [A]} \)

This produces a hyperbolic curve on a linear scale.

Step 2: Introduce the Hill Coefficient

Hill (1910) studied hemoglobin-oxygen binding and introduced a cooperativity exponent n (Hill coefficient):

\( E = E_{max} \cdot \frac{[A]^n}{EC_{50}^n + [A]^n} \)

When n = 1: standard hyperbolic (no cooperativity). When n > 1: positive cooperativity (steeper curve). When n < 1: negative cooperativity (shallower curve).

Step 3: The Sigmoidal Shape on Log Scale

Let x = log[A]. Then [A] = 10^x. Substituting:

\( E = \frac{E_{max} \cdot 10^{nx}}{EC_{50}^n + 10^{nx}} = \frac{E_{max}}{1 + 10^{n(\log EC_{50} - x)}} \)

This is a logistic (sigmoidal) function in x = log[A]. The curve is symmetric about the inflection point at x = log(EC_50), where E = E_max/2. The steepness is controlled by n.

Step 4: Linearization (Hill Plot)

Taking logs of both sides of E/(E_max - E) = ([A]/EC_50)^n:

\( \log\left(\frac{E}{E_{max} - E}\right) = n \cdot \log[A] - n \cdot \log EC_{50} \)

A Hill plot of log(E/(E_max - E)) vs log[A] yields a straight line with slope n and x-intercept at log(EC_50).

The Hill Equation

\( E = \frac{E_{max} [A]^n}{EC_{50}^n + [A]^n} \)

At [A] = EC_50: E = E_max/2 (by definition). The range from 20% to 80% response spans approximately 1.6 log units when n = 1, and narrows as n increases.

Derivation 2: Therapeutic Index & Therapeutic Window

The therapeutic index quantifies the margin of safety between the effective dose and the toxic dose of a drug.

Definition

The therapeutic index (TI) is the ratio of the dose producing toxicity in 50% of the population to the dose producing the desired effect in 50%:

\( TI = \frac{LD_{50}}{ED_{50}} \quad \text{or} \quad TI = \frac{TD_{50}}{ED_{50}} \)

LD_50 is used in preclinical studies; TD_50 is preferred clinically since lethality is not an acceptable endpoint.

Certain Safety Factor (CSF)

A more conservative measure that accounts for the overlap between therapeutic and toxic dose-response curves:

\( CSF = \frac{TD_1}{ED_{99}} \)

A CSF less than 1 means the dose required for 99% therapeutic efficacy already causes toxicity in 1% of the population, indicating a dangerously narrow margin.

Wide TI Drugs

Penicillin (TI > 100), benzodiazepines (TI approximately 100). Large margin of safety, less monitoring needed.

Narrow TI Drugs

Warfarin (TI approximately 2-3), digoxin, lithium, phenytoin, theophylline. Require therapeutic drug monitoring (TDM).

Derivation 3: Quantal Dose-Response & Probit Analysis

While graded dose-response measures the magnitude of effect in a single subject, quantal dose-response measures the proportion of a population that responds at each dose (all-or-none: either responds or does not).

Step 1: The Frequency Distribution

Individual thresholds for drug response vary across a population. If we plot the fraction of individuals responding at each dose, the resulting distribution is approximately log-normal (Gaussian on a log-dose scale):

\( f(\log D) = \frac{1}{\sigma\sqrt{2\pi}} \exp\left(-\frac{(\log D - \mu)^2}{2\sigma^2}\right) \)

where mu = log(ED_50) is the mean log-threshold and sigma is the standard deviation of the log-threshold distribution.

Step 2: Cumulative Distribution (S-shaped Curve)

The cumulative fraction responding at dose D is the integral of the frequency distribution:

\( P(D) = \frac{1}{2}\left[1 + \text{erf}\left(\frac{\log D - \mu}{\sigma\sqrt{2}}\right)\right] \)

This cumulative normal distribution function produces the classic S-shaped quantal dose-response curve. At D = ED_50, exactly 50% of the population responds.

Step 3: Probit Transformation

Bliss (1934) introduced the probit transformation to linearize the cumulative curve. The probit of P is the inverse normal (z-score) plus 5:

\( \text{probit}(P) = \Phi^{-1}(P) + 5 \)

Plotting probit(P) vs log(D) yields a straight line with:

\( \text{slope} = \frac{1}{\sigma}, \quad \text{probit}(0.5) = 5 \text{ at } \log D = \mu \)

This linearization allows precise estimation of ED_50, LD_50, and confidence intervals using regression.

Derivation 4: Potency vs Efficacy

Two fundamental properties of drugs are often confused. Potency and efficacy are independent parameters that can be formally separated using the Hill equation.

Potency: Horizontal Position

Potency is determined by EC_50. A lower EC_50 means higher potency (less drug needed for 50% effect). On a log-dose plot, higher potency shifts the curve leftward:

\( \text{Potency} \propto \frac{1}{EC_{50}} \)

Drug A is more potent than Drug B if EC_50(A) < EC_50(B). Potency depends on both affinity (K_D) and efficacy.

Efficacy: Vertical Maximum

Efficacy is the maximum response a drug can produce (E_max). A drug with lower E_max has lower efficacy, represented as a lower plateau on the dose-response curve:

\( \text{Efficacy} = E_{max} = \alpha \cdot E_{max,system} \)

Morphine (full agonist) has higher efficacy than buprenorphine (partial agonist) at mu-opioid receptors, regardless of their relative potencies.

Formal Separation

For two drugs acting on the same receptor system, comparing their Hill equations:

\( E_A = \frac{E_{max,A} [A]^{n_A}}{EC_{50,A}^{n_A} + [A]^{n_A}} \quad \text{vs} \quad E_B = \frac{E_{max,B} [B]^{n_B}}{EC_{50,B}^{n_B} + [B]^{n_B}} \)

If E_max,A = E_max,B but EC_50,A < EC_50,B: A is more potent but equally efficacious. If EC_50,A = EC_50,B but E_max,A > E_max,B: A is equally potent but more efficacious.

Graded vs Quantal Dose-Response Curves

Graded (Individual)

Quantal (Population)

Graded: measures magnitude

EC_50, E_max from single subjects

Quantal: measures proportion

ED_50, LD_50 from populations

Python Simulation: Dose-Response Analysis

Dose-Response — Hill Equation, Therapeutic Index, Quantal & Potency/Efficacy

Python

script.py86 lines

import numpy as np

log_dose = np.linspace(-3, 3, 200)
dose = 10**log_dose

# ── Panel 1: Hill Equation with varying n ──
EC50 = 1.0
hill_ns = [0.5, 1.0, 2.0, 4.0]
print("CHART1_TITLE: Hill Equation (varying Hill coefficient n)")
print("CHART1_XLABEL: log[Drug]")
print("CHART1_YLABEL: E / E_max")
for i in range(0, len(dose), 5):
    vals = []
    for n in hill_ns:
        e = dose[i]**n / (EC50**n + dose[i]**n)
        vals.append(f"n_{n:.1f}={e:.4f}")
    print(f"logD={log_dose[i]:.2f} " + " ".join(vals))

# ── Panel 2: Therapeutic Index Visualization ──
print("\nCHART2_TITLE: Therapeutic Index (ED50 vs TD50 vs LD50)")
print("CHART2_XLABEL: log[Dose]")
print("CHART2_YLABEL: % Population Responding")
ED50 = 1.0
TD50 = 10.0
LD50 = 50.0
sigma = 0.4
for i in range(0, len(dose), 5):
    ld = log_dose[i]
    p_eff = 100 * 0.5 * (1 + np.tanh((ld - np.log10(ED50)) / (sigma * 1.4)))
    p_tox = 100 * 0.5 * (1 + np.tanh((ld - np.log10(TD50)) / (sigma * 1.4)))
    p_leth = 100 * 0.5 * (1 + np.tanh((ld - np.log10(LD50)) / (sigma * 1.4)))
    print(f"logD={ld:.2f} Efficacy={p_eff:.2f} Toxicity={p_tox:.2f} Lethality={p_leth:.2f}")

# ── Panel 3: Quantal Dose-Response (frequency + cumulative) ──
print("\nCHART3_TITLE: Quantal Dose-Response (frequency & cumulative)")
print("CHART3_XLABEL: log[Dose]")
print("CHART3_YLABEL: Fraction")
mu = 0.0
sig = 0.5
for i in range(0, len(dose), 5):
    ld = log_dose[i]
    freq = np.exp(-0.5 * ((ld - mu) / sig)**2) / (sig * np.sqrt(2 * np.pi))
    cumul = 0.5 * (1 + np.tanh((ld - mu) / (sig * 1.4)))
    print(f"logD={ld:.2f} frequency={freq:.4f} cumulative={cumul:.4f}")

# ── Panel 4: Potency vs Efficacy ──
print("\nCHART4_TITLE: Potency vs Efficacy Comparison")
print("CHART4_XLABEL: log[Drug]")
print("CHART4_YLABEL: Response (%)")
drugs = [
    ("DrugA_high_pot_high_eff", 0.1, 100),
    ("DrugB_low_pot_high_eff", 10.0, 100),
    ("DrugC_high_pot_low_eff", 0.1, 60),
    ("DrugD_low_pot_low_eff", 10.0, 60),
]
for i in range(0, len(dose), 5):
    vals = []
    for name, ec, emax in drugs:
        e = emax * dose[i] / (ec + dose[i])
        vals.append(f"{name}={e:.2f}")
    print(f"logD={log_dose[i]:.2f} " + " ".join(vals))

# ── Panel 5: Probit Analysis ──
print("\nCHART5_TITLE: Probit Transformation (Linearized Quantal)")
print("CHART5_XLABEL: log[Dose]")
print("CHART5_YLABEL: Probit Value")
for i in range(0, len(dose), 5):
    ld = log_dose[i]
    p = 0.5 * (1 + np.tanh((ld - mu) / (sig * 1.4)))
    if 0.01 < p < 0.99:
        z = np.sqrt(2) * np.arctanh(2 * p - 1)
        probit = z + 5
        print(f"logD={ld:.2f} probit={probit:.3f}")

# ── Panel 6: Hill Plot Linearization ──
print("\nCHART6_TITLE: Hill Plot (Linearization)")
print("CHART6_XLABEL: log[A]")
print("CHART6_YLABEL: log(E/(Emax-E))")
for n in [1.0, 2.0]:
    for i in range(0, len(dose), 5):
        e = dose[i]**n / (EC50**n + dose[i]**n)
        if 0.01 < e < 0.99:
            hill_val = np.log10(e / (1 - e))
            print(f"logA={log_dose[i]:.2f} n_{n:.0f}={hill_val:.3f}")

print("\nSimulation complete: 6 panels generated")

Click Run to execute the Python code

Code will be executed with Python 3 on the server

Clinical Applications

Warfarin Dosing

Warfarin has a narrow TI (approximately 2-3). The quantal dose-response curves for therapeutic anticoagulation and bleeding overlap significantly, requiring INR monitoring and dose individualization based on CYP2C9/VKORC1 genotype.

Opioid Dose-Response

The Hill coefficient for opioid analgesia is approximately 2-3, producing a steep dose-response. This explains why small dose increases near the ED_50 produce large changes in effect, contributing to overdose risk.

Drug Potency Comparison

Hydromorphone is approximately 5x more potent than morphine (lower EC_50) but has similar efficacy (same E_max). Potency differences require equianalgesic dose conversion tables when switching opioids.

LD_50 in Toxicology

Probit analysis of quantal mortality data gives LD_50 with 95% confidence intervals. Modern approaches use the up-and-down method (OECD 425) requiring fewer animals than classical LD_50 determination.

Key Takeaways

1.
The Hill equation E = E_max[A]^n/(EC_50^n + [A]^n) produces a sigmoidal curve on a log-dose scale, with the Hill coefficient n controlling steepness.
2.
The therapeutic index TI = TD_50/ED_50 quantifies drug safety; narrow TI drugs require therapeutic drug monitoring.
3.
Quantal dose-response measures population-level all-or-none responses; the cumulative distribution gives ED_50 and the probit transformation linearizes the curve.
4.
Potency (EC_50, horizontal position) and efficacy (E_max, vertical maximum) are independent parameters that must both be considered in drug comparison.

← 3.1 Drug-Receptor 3.3 Agonists & Antagonists →

Share:X Reddit LinkedIn