Levodopa in Parkinson’s Disease: A Review of Population Pharmacokinetics/Pharmacodynamics Analysis

- Background : Parkinson’s disease is the second most common neurodegenerative disorder after Alzheimer’s disease. Although levodopa remains the single effective agent in the management of Parkinson’s disease, the accurate determination of this optimal dosage is complicated by marked between-subject and between-occasion variability in this population. This review presents a synthesis of the population pharmacokinetic and pharmacodynamic models of levodopa described in Parkinson’s disease. METHODS: A literature search was conducted from the PubMed database, from their inception through April 2016, using the following terms: levodopa, pharmacokinetic(s), pharmacodynamic(s) population, model(ling) and nonlinear mixed effect. Articles were excluded if they were not pertinent. References of all selected articles were also evaluated. RESULTS : A total of 12 articles were finally retained. The following covariates were selected as interindividual variability factors: body weight, age, sex, creatinine clearance and levodopa dose. The clinical response versus effect site concentration relationship was described with different sigmoidal E max models. Different pharmacodynamic effects were described: UPDRS, Tapping, Dyskinesia, CURSΣ and treatment response scale. DISCUSSION: This review allows us to realize interpretation of a patient’s clinical picture and confirmed the appropriateness of the pharmacokinetic-pharmacodynamic modeling for levodopa. External evaluation of previous published models should be also continued to evaluate these previous studies. New pharmacokinetic and/or pharmacodynamic population modelling studies could be consider to improve future models and decrease variability, to better understand the evolution of patients with Parkinson’s disease treated by levodopa.


INTRODUCTION
Parkinson's disease is the second most common neurodegenerative disorder after Alzheimer's disease [1]. The disease is characterized by progressive degeneration of the dopaminergic nigrostriatal system and depletion of dopamine, which results in the core motor symptoms of bradykinesia, rigidity, tremor, and postural instability [2]. Parkinson's disease is a progressive neurodegenerative disease that affects approximately 1-2% of the population above 60 years of age [3]. The cardinal clinical manifestations of Parkinson's disease are resting tremor, rigidity, bradykinesia, and gait dysfunction. During the early stages of the disease, about 70% of patients may experience a slight tremor. Bradykinesia is described as a general reduction in spontaneous movement, and can cause difficulty with repetitive movements, such as finger tapping. Rigidity may cause stiffness of the limbs, neck, and trunk. In the 80s, the Movement Disorders Society has developed the Unified Parkinson's Disease Rating Scale (UPDRS) and in 2001, the MDS sponsored a critique of the UPDRS. The summary conclusions recommended the development of a new version of the UPDRS that would retain the strengths of the original scale, but especially incorporate a number of clinically pertinent PD-related problems poorly captured in the original version. Based on this critique, the MDS commissioned a revision of the scale, resulting in a new version: The Movement Disorders Society-Unified Parkinson's Disease Rating Scale (MDS-UPDRS) [4]. This scale provides an efficient and flexible means of monitoring Parkinson's disease-related disability and impairment, and has been used in studies of early, mild, moderate, and advanced disease with motor fluctuations. The MDS-UPDRS rates 65 items in comparison to 55 on the original UPDRS, 48 that had 0 to 4 options and 7 with yes/no responses. The total MDS-UPDRS scale comprises four components: Part I, Mentation; Part II, Activities of daily living; Part III, Motor; Part IV, Complication of therapy.
Facilitation of dopamine biosynthesis by administration of the precursor levodopa is one, and until now the most important, therapeutic principle in drug therapy of Parkinson's disease [5,6,7,8,9,10]. Levodopa is one of the effective agents in the management of Parkinson's disease [8,11,12] and reaches its effect site by crossing the blood-brain barrier. In brain tissues, levodopa is decarboylated to dopamine, which normally stored in presynaptic terminals of striatal neurons [5,13]. Unfortunately, after several years of treatment motor complications such as motor fluctuations and dyskinesias can arise [14,15]. Risk and time to emergence of these motor complications vary substantially among patients for complex reasons, including both disease-and drug-related factors, particularly treatment with levodopa [16,17] . A previous review of studies of motor complications estimated that the risk of developing motor fluctuations and dyskinesias were both about 40% after 4-6 years of levodopa treatment [18]. However, most previous studies have been based on unrepresentative samples, with attendant selection biases [19,20]. Only two representatives, community-based incidence studies have examined the development of motor complications over time, both of which were small, only reported dyskinesias and one was retrospective [21,22]. An increased understanding of the levodopa plasma concentration-effect relationship could be valuable in the assessment of Parkinson's disease management [23,24,25,26].
This has created a need to examine more carefully the factors which influence the variability in levodopa pharmacokinetics and pharmacodynamics. After oral administration, levodopa is completely absorbed in the proximal small bowel but undergoes marked presystemic decarboxylation to dopamine. The plasma elimination half-life of levodopa is about 2 hours. With coadministration of a peripheral decarboxylase inhibitor, degradation to 3-Omethyl-dopa (3-OMD) by catechol-Omethyltransferase (COMT) is the major route of levodopa metabolism [5,6,7,8,9,10,27].
Many pharmacokinetic, pharmacodynamic or pharmacokinetic/pharmacodynamic studies in Parkinson's disease patients have been performed.
In particular, nonlinear mixed-effect modelling, a commonly used population-based modelling approach, have been used to identify covariates that could influence the dose-concentration or dose-effect relationship. Population pharmacokinetic/pharmacodynamic approach allowed realizing bayesian dose estimation and adaptation according to population pharmacokinetic parameters and estimated variability in a specific population. Population pharmacokinetic modeling was first introduced in 1972 by Sheiner et al. and thirty years after their introduction, population pharmacokinetic / pharmacodynamic approaches become a reference method for drug evaluation and dose adaptation [28].
Understanding the relationship between pharmacokinetics and pharmacodynamics of levodopa has been at the center of the most discussion of effective management of Parkinson's disease patients experiencing motor fluctuations. The degree of correlation between variations in plasma levodopa concentrations and motor performance has been described by four types of responses to levodopa: the early disease duration of response, late disease duration of response, negative response and dyskinesias. The early disease duration of response assessed by the response to a single dose of levodopa is characterized by: (i) a plasma compartment and central-effect compartment threshold concentration of levodopa required to obtain a clinical response, (ii) a lag-time between peak plasma levodopa concentration and clinical response, (iii) the magnitude of the clinical response follows a dichotomic response after the threshold concentration is reached and (iv) the duration of the clinical response is linearly related to the plasma concentration, there is improvement in motor function that roughly parallels plasma levodopa concentrations [29]. The late disease duration of response is common to most antiparkinsonian agents and is found in mildly and severely affected parkisonian patients. The late disease duration of response is inversely related to disease severity and decays more rapidly in more severely affected patients. It is estimated that the late disease duration of response contributes about 30 to 50 % of the total levodopa response in response fluctuators [30]. The negative response is the deterioration of motor function, which can last from minutes up to an hour and is most frequently observed as the early disease duration of response of levodopa disappears. Dyskinesias are linked to the early disease duration of response, "off" phenomenon occur when plasma levodopa concentrations are low and the early disease duration of response disappears. Similar to the early disease duration of response, the duration of dyskinesia is proportional to the plasma levodopa concentration and the severity of dyskinesia is an "all or none" response. While much has been achieved in attempts to decipher the pharmacokinetic-pharmacodynamic relationships, the results from several studies still remain equivocal and a clear-cut delineation of the pharmacokinetic-pharmacodynamic relationship of levodopa continues to be a subject of further investigations.
This review presents a synthesis of the population pharmacokinetic, pharmacodynamic or pharmacokinetic/pharmacodynamic analyses performed for levodopa in Parkinson's disease patients. The objective was to describe the different published pharmacokinetic and pharmacodynamic models to determine if there was a consensus on a structural pharmacokinetic (dose-concentration relationship) or pharmacodynamic (dose-effect relationship) model and which motor responses were investigated.

PRISMA
We have followed the principles of the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) statement to guide assessment of quality of this review. The PRISMA statement helps us to improve the reporting of systematic review with checklist [31].

Inclusion criteria
We included all described pharmacokinetic population models of levodopa. The articles were accepted if they met the following inclusion criteria: -Studied populations: Idiopathic Parkinson's disease.
-Pharmacokinetic analysis: nonlinear mixed effect model by a population approach

Exclusion criteria
The articles were excluded if they are reviews, methodology articles or if the analysis did not used a population pharmacokinetic modelling and population studies not involving a mixed-effects models analysis.

Search strategy
A literature search was conducted from the PubMed database, from their inception through April 2016, using the following terms: (levodopa AND pharmacokinetic) OR (levodopa AND pharmacokinetic AND population) OR (levodopa AND pharmacokinetic AND model(ling)) OR (levodopa AND pharmacokinetic AND nonlinear mixed effect model) OR (levodopa AND pharmacodynamic) OR (levodopa AND pharmacodynamic c AND population) OR (levodopa AND pharmacodynamic AND model(ling)) OR (levodopa AND pharmacodynamic AND nonlinear mixed effect model). Moreover, additional studies were identified from the reference list of selected papers. The search was additionally limited to "English language" and "clinical data".

Data extraction
The results of these investigations were closely evaluated and articles were retained if they met the inclusion criteria. Pertinent articles were assessed and the following data were extracted: year of publication, number of patients, number of samples, software, structural model, value and expression of pharmacokinetic parameters, value and expression of pharmacodynamic parameters, included covariates, between-subject and betweenoccasion variability, residual error and validation method.
Following Brendel et al.
[33], the evaluation methods were divided into three categories according to increasing order of quality: basic internal methods (goodness-of-fit plots), advanced internal methods (bootstrap, cross-validation, Monte Carlo simulations…) and external model evaluation.
Concerning extracted parameters, main pharmacokinetic and pharmacodynamic parameters with clinical link were detailed. Main pharmacokinetic parameters were clearance and volume of distribution to describe the evolution of concentration versus time. Main pharmacodynamic parameters were Emax (the amplitude between baseline and maximal effect), EC50 (the concentration producing 50% of Emax, and the Hill coefficient (the slope of concentrationeffect curve) to describe the effect versus concentration and time.

Trial flow
A total of 18 studies were identified through Pubmed database searching. These 18 articles were screened and a total of 16 articles were first selected to have their full-text versions assessed for eligibility. Among these 4 were excluded regarding to the inclusion and exclusion criteria.  Figure 1). Absorption time was modelled by lag time, transit compartment or simple oral absorption. In all the publications, several covariates were tested; five covariates were retained in these final models. The following covariates were selected as interindividual variability factors for clearance (CL or Q): body weight (CL and Q), age (CL), sex (CL), creatinine clearance (CL) and levodopa dose (CL). The following covariates were selected as interindividual variability factors for volume of distribution (V/V1 or V2): body weight (V, V1 and V2) and sex (V). Age was also selected as covariate on ka. Table 2 summarizes mean values of pharmacokinetic parameters for one-and twocompartment models described in Parkinson's disease patients. The range estimate of mean value of clearance, intercompartmental clearance, central volume of distribution and peripheral volume of distribution were between 17.0 and 36.6 L/h (n=6), 6.8 and 38.7 L/h, 11.0 and 124.0 L, 23.4 and 72.9 L, respectively.
All models were evaluated with internal or external method ( Table 2). Concerning the basic internal evaluation, three authors used this method. This evaluation most frequently used was calculation of indicators of the performance of prediction (bias and precision). Two authors used advanced internal evaluation, with visual predictive check (500 simulations) and bootstrap (1000 simulations). In case of external evaluation, two authors chose to test their model on a prospective group (between 16 and 311 patients).  Table 3). The clinical response versus effect site concentration relationship was described with different sigmoidal Emax models (Table 3 and Figure 1). Concerning UPDRS, the mean values of E0, Emax (score), EC50 (µg/ml) and γ (Hill coefficient) were ranged from 0 to 54.1, 0.72 to 63.0, 0.812 to 1.41, and 0.503 to 6.2, respectively. The mean values of Emax for tapping were between 35 and 146 Taps/min. Concerning dyskinesia, the mean values of Emax (score), EC50 (µg/ml) and γ were ranged from 1 to 17.9, 0.601 to 6.28, and 2.1 to 30, respectively. The mean values of interindividual variability of E0, Emax, EC50 and γ were ranged from 19 to 53% (n=5), 39 to 90% (n=4), 22 to 101% (n=6), 15 to 130% (n=5), respectively. . None of these covariates had a major influence; the model was optimized only until the concentration data were reasonably well described. The presence of a sex-related effect on levodopa pharmacokinetics has been evaluated in several others studies using non-compartmental approaches and conflicting evidence has been reported [47,48,49,50]. It has been suggested that the apparent sex-related difference in exposure was partly explained by incorrectly normalizing body weight difference between men and women in some analyses [49,50].

DISCUSSION
The drug concentrations in plasma or blood are correlated with simultaneously measured clinical effects by a suitable pharmacodynamic model which usually assumes a linear or sigmoidal (Emax) concentration-effect relationship. The parameters of a sigmoidal model are Emax (the amplitude between baseline and maximal effect), EC50 (the concentration producing 50% of Emax, which might be interpreted as a target concentration necessary to obtain a clinical considerable effect) and the Hill coefficient (above 5 units indicates a very steep concentration-effect curve and suggests a dichotomic response (all-or-nothing) [51]. Despite the heterogeneity of the methodological approaches used (different score systems for the pharmacodynamic response: TRS, UPDRS, CURSΣ, Dyskinesia, Tapping), these studies on levodopa have demonstrated almost consistent comparable findings. All studies prove found the Emax model to be suitable for describing the concentration-effect relationship of levodopa [34,35,38,40,41,42,43,44,45]. The steepness of the concentration-effect curves, characterized by the Hill coefficient, was relatively high (between 3 and 5) for one study and very high (>5) for three studies, indicating an almost immediate onset of the pharmacodynamic response to levodopa when the EC50 is reached and the appearance of "off" reactions when concentrations falls below this level [34,35,42,45]. Furthermore, the different estimates of EC50 indicate that the effect of levodopa is not dependent on the height of the plasma concentrations once the EC50 value has been exceeded. Indeed, Kempster et al. showed that the amplitude of the motor response depends on the pretreatment; who describe an increase in the amplitude of the motor response when patients with Parkinson's disease were grouped according to duration of disease [9]. The duration of disease of patients included in these studies is very wide (0.5 and 25 years) and this data was not included as a covariate except in one study. . According to their results, the progression of disease was estimated to be an increase of 0.7 points score per year (UPDRS Part III). Indeed, disease progression is characterized by a number of pharmacodynamic modifications. The baseline motor function in the absence of any drug (E0), which correlates with disease duration, decreases but with no decrease in the maximal therapeutic response (Emax), which means that the intrinsic activity remains preserved [40,52]. The levodopa concentration required to obtain 50% of Emax (EC50) progressively increases over the years and is at least 2.5 times higher in fluctuating responders compared with stable responders. In addition, it is well known that factors other than levodopa concentration or duration of disease may influence parkinsonian motor symptoms. These factors include stress level, food intake, time of day, physical activity, intake of other pharmaceuticals affecting dopaminergic, or other receptors. None of these factors are included in these studies, and these will therefore seem as unexplained between-subject variability or residual error.
In addition, the role of 3-O-metyldopa (3OMD) in the pharmacokineticpharmacodynamic relationship of levodopa has been investigated. This metabolite is a large neutronal amino acid with a long half-life. It crosses the blood-brain barrier but does not bind to the dopamine receptor and has no recognised intrinsic antiparkinsonian activity. Nevertheless, 3OMD can competitively inhibit transport of levodopa [53]. The role of 3OMD in the pharmacokinetic-pharmacodynamic of levodopa has been investigated with concomitant administration of specific catechol-0methyltransferase (entacapone, tolcapone), and may be helpful in elucidating the relationship between plasma concentrations of levodopa and its effects . Others studies have postulated that 3OMD could inhibit the uptake of levodopa across the brain-blood thus suggesting a relationship between the central and peripheral levels of levodopa [54]. The role of 3OMD was still controversial in the pharmacokineticpharmacodynamic relationship of levodopa. Nevertheless, the possibility of an antagonistic effect of 3OMD on levodopa response could be reconsidered.
Evaluation of between-occasion variability was also an important aspect of the care of patients with Parkinson's disease because levodopa is administered as long-term therapy. Ignoring between-occasion variability when it is present can lead to model misspecifications [55]. Betweenoccasion variability was included in three studies. In the study of Westin et al., the data set was too small to separately distinguish between-subject and between-occasion variability in parameters [34]. Therefore, each occasion was treated as a separate patient, and the resulting parameter variability will approximate the sum of betweensubject and between-occasion variability. Chan et al. showed that the most of the overall variability in parameters is due to between-occasion variation [37]. Their results showed that ignoring betweenoccasion variability inflates both between-subject variability and residual error as demonstrated by Karlsson and Sheiner [56]. In more, their results showed that ignoring between-occasion variability may cause imprecision in parameter estimation and an underestimate of total population parameter variability. Regarding study of Troconiz et al., introducing between-occasion variability in their model did not lead to different covariate models but decreased the objective function value [35]. In more, the important estimated between-occasion variability on ke0 (40%) suggests that the events taking place in the biophase are the chief determinants of the unpredictability of the response to levodopa.
Several models were evaluated and confirmed by robust method. . These models allow to minimize problems due to temporal delay between Cmax and clinical response and to obtain pharmacokinetic-pharmacodynamic parameters. These different studies allow us to conclude that modifications in stables and fluctuating patients are a reflection of the progression of the disease since the threshold concentrations needed to obtain an appropriate antiparkinsonian effect are increased and the duration of the effect is shortened (EC50 and Hill coefficient). These studies allow us to realize interpretation of a patient's clinical picture.

Limitations
Assessing the risk of bias should be part of the conduct and reporting of any systematic review. For systematic reviewers, understanding the risk of bias on the results of studies is often difficult, because the report is only a surrogate of the actual conduct of the study. There are three main ways to assess risk of bias: individual components, checklists, and scales. The new Cochrane risk of bias tool [60.] is one such component approach. According to the Cochrane risk of bias tool, we have highlighted the various biases and limitations of this review (Table 4). This review is only an image, at a given time, in a particular indication and population of the published pk/pd models of levodopa.

CONCLUSION
The importance of individualizing therapy in Parkinson's disease has been well established [61]. The large number of variables which affect final dopamine concentration at receptor sites in the striatum necessitate to move to an individualized dosing approach. The population approach allows the pharmacokinetic and pharmacodynamic characterization of drugs in a target population, the evaluation of the associated within subject and within individual variability, and the identification of covariates affecting such variability. Understanding the variability associated with the pharmacokinetics and pharmacodynamic, and identifying subpopulations with special features can provide clinicians with relevant information regarding dose individualization.
Given the relative low number of studies on population pharmacokinetic and pharmacodynamic modelling of levodopa in Parkinson's disease, news pharmacokinetic and/or pharmacodynamic population modelling studies could be consider. External evaluation of previous published models should be also continued to evaluate these previous studies. It would be interesting to continue research on pharmacokinetic and pharmacodynamic of levodopa to improve future models and decrease variability, to upgrade the care of patients with Parkinson's disease treated by levodopa.
Knowledge of individual patients' levodopa pharmacokinetics and pharmacodynamics variables, particularly duration of motor effect and matched EC50 values after administration of levodopa oral dose, can help clinicians to objectively assess where Parkinson's disease patients stand in the disease process. This information can also contribute to adapt drug treatment from the early stages and to modify it according disease progression. These knowledges aim to simplify pharmacological treatment schedules as far as possible, to reduce the risks of adverse effects (acute and chronic) and possibly to delay the development of a severe disability.
In a clinical perspective, this review advances relevant information for clinicians and researcher about the pharmacokinetics and pharmacodynamics of levodopa. To optimize levodopa dosage, this review point out the relevant information according to the target population.