Dopamine Buffering Capacity Imaging: A Pharmacodynamic fMRI Method for Staging Parkinson Disease

Abstract

We propose a novel pharmacological fMRI (phMRI) method for objectively quantifying disease severity in Parkinson disease (PD). It is based on the clinical observation that the benefit from a dose of levodopa wears off more quickly as PD progresses. Biologically this has been thought to represent decreased buffering capacity for dopamine as nigrostriatal cells die. Buffering capacity has been modeled based on clinical effects, but clinical measurements are influenced by confounding factors. The new method proposes to measure the effect objectively based on the timing of the known response of several brain regions to exogenous levodopa. Such responses are robust and can be quantified using perfusion MRI. Here we present simulation studies based on published clinical dose-response data and an intravenous levodopa infusion. Standard pharmacokinetic-pharmacodynamic methods were used to model the response. Then the effect site rate constant k e was estimated from simulated response data plus Gaussian noise. Predicted time - effect curves sampled at times consistent with phMRI differ substantially based on clinical severity. Estimated k e from noisy input data was recovered with good accuracy. These simulation results support the feasibility of levodopa phMRI hysteresis mapping to measure the severity of dopamine denervation objectively and simultaneously in all brain regions with a robust imaging response to exogenous levodopa.

Keywords: ASL; drug discovery and development; hysteresis; levodopa; phMRI; pharmacodynamics; pharmacokinetic-pharmacodynamic modeling; pharmacological biomarkers.

Copyright © 2020 Black, Acevedo and Koller.

Figures

Figure 1

Across groups of PD patients, ke is a surrogate for disease duration (r = 0.95). Data redrawn from Harder and Baas (8) and Contin et al. (9).

Figure 2

Plasma levodopa concentrations in PD patients following the “final dose” intravenous infusion method in Black et al. (26). The 3 lines mark the mean, 90th and 10th percentile for samples collected in the corresponding intervals. Redrawn from data reported in Black et al. (26).

Figure 3

Predicted levodopa concentration in the effect compartment at various disease severity levels. Curves are labeled by t½e = ln 2/ke from more severe PD (t½e = 5 min.) to milder PD (t½e = 277 min.).

Figure 4

Predicted time:effect curves at various disease severity levels assuming (A) mean, (B) high, and (C) low Cp(t) in response to the levodopa infusion.

Figure 5

Estimated ke (vertical axis) across 100 sets of noise added to the time: effect curve computed for the ke, EC50, and n for various severities of PD as reported in Contin et al. (9) (horizontal axis), assuming (A) mean, (B) high and (C) low Cp(t) in response to the levodopa infusion. Noise CoV = 12.9%. Width of plot is proportional to frequency of output of the given magnitude. Filled circle: input n. Horizontal lines note the 5, 50, and 95th percentiles. At right, similar results are shown for noise CoV = 5% for (D) mean, (E) high, and (F) low Cp(t) responses to the LD infusion.

Figure 6

Estimated ke (vertical axis) across 100 sets of noise added to the time:effect curve computed for the ke, EC50, and n for various severities of PD as reported in Contin et al. (9) (horizontal axis), with Cp(t) estimated for an levodopa infusion twice as long (at half the rate, so that the total infused dose is equivalent). (A) noise CoV = 12.9%; (B) noise CoV = 5%.