FormalPara Key Points
Table 1
Full size table

1 Introduction

Schizophrenia is a chronic and severe mental disorder, affecting more than 23 million people worldwide in 2018 [1, 2]. Patients suffering from schizophrenia are two to three times more likely to die early than the general population due to cardiovascular or metabolic diseases and infections [1]. Therefore, an effective and individual treatment is important. One treatment option for schizophrenia is risperidone.

Risperidone is an antipsychotic agent chemically classified as a benzisoxazole derivative that has selective antagonistic properties at serotonin 5-HT2 and dopamine D2 receptors [3]. Depending on the dosage of risperidone, the incidence of antipsychotic-induced extrapyramidal symptoms (EPS) occurs more frequently than with other second-generation antipsychotics [4,5,6]. Besides the limiting EPS profile, risperidone therapy is associated with an increased prolactin level, which is plasma concentration dependent [7].

The antipsychotic agent risperidone is mainly metabolized in the liver via the cytochrome P450 (CYP) 2D6 enzymatic pathway [7,8,9]. Furthermore, in vitro studies [8, 9] as well as clinical dug–drug interaction studies [10, 11] have suggested that CYP3A4 is also involved in the enzymatic pathway of risperidone, but plays a minor role compared with CYP2D6 [9]. The contribution of both enzymes was evaluated with specific in vitro inhibitors using quinidine (CYP2D6 inhibitor) and ketoconazole (CYP3A4 inhibitor) [8, 9]. Consequently, CYP2D6 and CYP3A4 are the relevant enzymes for the metabolism of risperidone to its active metabolite 9-hydroxyrisperidone (9-OH-RIS), which has pharmacologic effects equivalent to that of risperidone [8, 12]. Other metabolites such as 7-hydroxyrisperidone, among others, do not exhibit significant pharmacologic activities [7]. The clinical effect of the drug is determined by the sum of the concentration of risperidone and its active metabolite 9-OH-RIS, called active moiety [13]. The therapeutic reference range of the active moiety is 20–60 ng/mL according to the Arbeitsgemeinschaft für Neuropsychopharmakologie und Pharmakopsychiatrie (AGNP) consensus guideline on therapeutic drug monitoring (TDM) published by Hiemke et al. [14].

There is evidence of several differences in risperidone treatment showing large inter-individual variability in risperidone and 9-OH-RIS concentrations [15, 16]. The wide pharmacokinetic variability of many drugs is largely explained by the CYP2D6 genetic polymorphism [17], but can also be influenced by age, renal and hepatic function, disease status, and co-medication [17, 18]. To date, > 100 allelic variants of the CYP2D6 genotype have been proposed in four different phenotypes [19, 20]: extensive metabolizer (EM), intermediate metabolizer (IM), poor metabolizer (PM), and ultra-rapid metabolizer (UM). The frequency of metabolizers is ethnically dependent and was estimated for a white population by Xie and Feng [21] to be 60–85% EM, 10–15% IM, and 1–10% UM. In the white population, approximately 7–10% are PM [21], completely lacking CYP2D6 activity because of two non-functional alleles [19]. Due to the non-functional alleles of the CYP2D6 gene, the PM phenotype appears to be associated with more adverse drug reactions (ADRs) and frequent discontinuation of therapy due to ADRs [22]. Especially for risperidone, Ereshefsky [23] proposed that a risperidone/9-OH-RIS ratio > 1 is an index indicating a CYP2D6 PM in the absence of a potent CYP2D6 inhibitor. CYP2D6 PMs do not have CYP2D6 activity but are able to produce 9-OH-RIS. Wu [24] supposed that in an individual who produces an inactive variant for an enzyme in the major metabolic pathway, the drug can be metabolized through an alternate or minor pathway, assuming that these enzymes are not affected genetically. Therefore, it is likely that the messenger RNA (mRNA) of those genes are up-regulated to compensate for the presence of the deficient enzyme [24]. One possible alternative metabolic pathway in CYP2D6 PMs may be the alicyclic dehydroxylation and oxidative N-dealkylation via CYP3A [25]. All in all, the relative contribution of CYP3A4 in PMs is not yet fully understood.

Although risperidone metabolism is regulated by the genetic polymorphism of CYP2D6, several authors detect only little differences between the pharmacokinetics and effects of the active moiety between the different types of metabolizer [26,27,28,29,30]. However, physiological changes according to the different CYP2D6 phenotypes may influence the pharmacokinetics of the administrated drug. So, the aim of the current investigation was to develop a whole-body physiologically based pharmacokinetic (PBPK) model to analyze the relationship between the pharmacokinetics of risperidone and 9-OH-RIS and the CYP2D6 phenotype. Furthermore, the influence of the several types of CYP2D6 metabolizer on the dosage should be investigated to be able to treat patients individually according to their phenotype. In this context, it should also be investigated whether it is possible to draw conclusions on the different phenotypes using the risperidone/9-OH-RIS ratio, to avoid expensive genotyping. To our knowledge, this is the first study examining the effects of CYP2D6 phenotype-related physiological alterations on the pharmacokinetics of risperidone using a PBPK approach.

2 Materials and Methods

2.1 Physiologically Based Pharmacokinetic (PBPK) Modelling

The modelling was performed using PK-Sim® version 7.1.0 build 63 [31] as part of the Open Systems Pharmacology Suite (Bayer Technology Services, Leverkusen, Germany), a freely available ‘whole-body’ PBPK simulation software that allows description of the fate of xenobiotics, especially drugs in humans, as well as in several mammalian organisms [32, 33]. The software implements a generic structure consisting of 17 different compartments, each representing a single tissue or organ [34, 35]. All compartments are connected via organ-specific blood flow to the arterial and venous blood pool to enable inter-compartmental mass transport [32, 34, 35]. Each organ is subdivided into subcompartments, such as vasuclar space containing erythrocytes and plasma, intercellular and tissue interstitial [35]. Based on physicochemical properties (e.g., plasma protein binding or fraction unbound in plasma, lipophilicity and molecular weight), compound-specific clearance information, as well as organ/tissue composition data (e.g., lipid, protein, water), PK-Sim® estimates drug-related model parameters to describe the rate and extent of drug disposition in each compartment [34]. For a detailed description of the generic model structure of PK-Sim® and the input/output parameters, refer to the user software manual [33], Willmann et al. [32, 34], or Kuepfer et al. [36].

2.2 General Workflow

The workflow for the development, evaluation, and transmission of the risperidone and 9-OH-RIS PBPK model is described in Fig. 1. First, an initial single-dose model was created based on a systematic literature survey to detect relevant anatomical, physicochemical, and drug-specific parameters of oral risperidone and its active metabolite. Additionally, properties relevant to absorption, distribution, metabolism, and elimination for risperidone and 9-OH-RIS were integrated. No difference was made between different CYP2D6 phenotypes. Based on the initial PBPK model, the model was evaluated using study data, supplied by two bioequivalence trials [37]. Plasma concentrations of risperidone and 9-OH-RIS were used to develop different PBPK models, according to each CYP2D6 phenotype: EM, IM, PM, and UM. Finally, simulations were performed in four different populations during steady-state. All PBPK simulations for risperidone and 9-OH-RIS were completed using PK-Sim® software [31].

Fig. 1
figure 1

Schematic workflow for the development of a physiologically based pharmacokinetic model for healthy individuals. The initial model includes absorption, distribution, metabolism, and elimination (ADME), physicochemical, and physiological data on literature values only. For the evaluated model the literature-based adaptions were modified by 71 genotyped healthy individuals from the study. Based on the evaluated single-dose model, the model was first scaled from risperidone 1–6 mg and subsequently simulated for steady-state. EM extensive metabolizer, IM intermediate metabolizer, PM poor metabolizer, RIS risperidone, UM ultra-rapid metabolizer

Full size image

2.3 Development of a Single-Dose PBPK Model

Risperidone- and 9-OH-RIS-relevant anatomical, physicochemical, and drug-specific parameters were taken from the literature and integrated into the initial PBPK model. After oral administration, the absolute bioavailability of risperidone amounts to 70% [38] after rapid absorption. Risperidone is excreted in the urine as metabolites (65%) and as unchanged drug (5%) [39]. Additionally, 14% of risperidone is excreted as metabolites in the feces [39]. For the initial PBPK model, risperidone clearance were determined to be 5 ± 1 mL/min/kg [40] regarding all types of CYP2D6 metabolizer. Besides the metabolizing enzymes CYP2D6 and CYP3A4, the drug transporter P-glycoprotein (P-gp) was incorporated in the simulation, based on P-gp adenosine triphosphatase activity measurement [41]. The PK-Sim® Standard method [42, 43] was chosen for the estimation of partition coefficient and cellular permeabilities.

2.4 Evaluation of a Single-Dose PBPK Model

2.4.1 Study Data

For the PBPK model evaluation of risperidone and 9-OH-RIS, data were supplied by two bioequivalence trials [37] using Risperdal® 1 mg (Janssen-Cilag, Madrid, Spain) as reference formulation and one test formulation. As described previously [37], a single-dose, randomized, two-period, crossover study was conducted. Between May and July 2003, blood samples of 71 healthy volunteers (36 women and 35 men) were collected at the following times after dosing: 0, 0.25, 0.5, 1.0, 1.5, 2.0, 3.0, 4.0, 5.0, 6.0, 8.0, 10, 12, 24, 48, 72, and 96 h. All subjects were Caucasian, expect for three Hispanic persons (one from Mexico and two from Argentina). Inclusion criteria were 18–35 years of age, non-smoking, body mass index between 19 and 30 kg/m2, and results within the reference range of laboratory values for hematology, as well as other criteria [37]. Demographic data and patient characteristics are summarized in Table 1.

Table 1 Characteristics of genotyped subjects in the study
Full size table

Genotyping of CYP2D6 was conducted by standard methods previously described in Cabaleiro et al. [44]. According to the activity score (AS) of the different CYP2D6 alleles, four phenotypic classes were determined: EM, IM, PM, and UM [20, 45]. A comparison of the genes and AS is provided in Table 2.

Table 2 Comparison of gene activity score of different genes between literature data and study appearance according to each phenotype
Full size table

2.4.2 PBPK Model Evaluation

For model evaluation, plasma concentrations of risperidone and 9-OH-RIS were used to predict the pharmacokinetics of a mean population of healthy volunteers. A differentiation between the various phenotypes was not performed. Only plasma concentrations relating to the Risperdal® formulation were used. Timepoints with very few measurements above the lower limit of quantification (LLOQ: 0.1 ng/mL) are not representative for the whole population, and taking these data into account would lead to a bias. Therefore, only those timepoints in which at least 25% of all subjects of the respective phenotype had plasma concentrations above the LLOQ were deemed eligible for analysis. A detailed overview of data selection is provided in Electronic Supplementary Material Figure S1. Measured data were used as observed values. In comparison to the initial model, lipophilicity (logP) for risperidone and 9-OH-RIS was modified according to predicted values from the literature [46,47,48]. Risperidone itself undergoes massive hepatic metabolism (65 % urine changed and 14% feces changed) mediated by the integration of CYP enzymes [39]. Additionally, the specific hepatic and renal clearances of 9-OH-RIS were fitted to the plasma concentration–time profile. Specific clearance represents the intrinsic clearance normalized to the volume where the process occurs [33] and amounts to 0.04/min (hepatic) and 0.67/min (renal). The model parameter partition coefficient and cellular permeability were adjusted according to the observed data. All other model-building parameters were kept from the initial model. Final PBPK parameters are listed in Table 3.

Table 3 Input parameters for the physiologically based pharmacokinetic model of risperidone and 9-hydroxyrisperidone
Full size table

2.5 Cytochrome P450 2D6 Phenotype-Related Scaling of PBPK Model Parameters

Next, all 71 healthy volunteers were divided into different subgroups according to their CYP2D6 phenotype: EM (n = 33), IM (n = 26), PM (n = 6), and UM (n = 6). Following that, four separate populations were created for each type of metabolizer to allow a comparison among the different CYP2D6 phenotypes. All chemical and physicochemical parameters of risperidone and 9-OH-RIS, as well as anatomical and physiological information on the evaluated model, were transferred unchanged to all four models, as they did not vary between the several types of metabolizers. In addition to the compound properties, the administration protocol and the formulation were also adopted as they do not differ. Due to the genetic polymorphism of CYP2D6, the catalytic rate constant (kcat) (/min), also known as the turnover frequency, was changed during model adaption because of the different enzyme activities. kcat is defined as the catalytic efficiency of enzymes, or rather the number of substrate molecules, each enzyme site converts to product per unit of time, and in which the enzyme is working at maximum efficiency [31, 33, 36]. The extent of the model adjustments was based on the views of Gaedigk et al. [45] using the AS, describing the respective enzyme activity of the different metabolizers. Subjects with an AS of 0 were considered as PMs, an AS value of 0.5–1.0 as IMs, a value of 1.5–2.0 as EMs, and an AS > 2 as UMs [45]. Based on the different AS ratios in the literature, all kcat values were first estimated through parameter optimization by PK-Sim® and subsequently set into ratio according to their status (Table 4). kcat values for CYP2D6 9-hydroxylation, 7-hydroxylation, and N-dealkylation were set to 0/min for PMs, because this type of metabolizer has two non-functional alleles, which prevents the enzyme from functioning [19, 49]. In contrast, the CYP3A4 kcat value for N-dealkylation and 7-hydroxylation was increased from 0.15/min (EM, IM, and UM) to 2.5/min (PM), as this is more active in PMs [25]. UMs, who have at least three copies of functional alleles, CYP2D6 kcat values were set at 4.0/min (9-hydroxylation) and 2.0/min (7-hydroxylation and N-dealkylation).

Table 4 Summary of phenotype-related input parameter for cytochrome P450 2D6 and 3A4
Full size table

2.6 Steady-State Simulation

Based on the single-dose PBPK models for EMs, IMs, PMs, and UMs, the simulations were conducted for steady-state conditions using PK-Sim®. This is possible because the concentrations of risperidone and its active metabolite 9-OH-RIS increase dose proportionally over the dosing range of 1–16 mg daily (0.5–8 mg bi-daily) [38], indicating that risperidone exhibits linear elimination pharmacokinetics [50, 51]. As risperidone has a half-life of 3.2 h and 9-OH-RIS of 24.7 h [52], steady-state is reached within 1 day for risperidone and within 5 days for 9-OH-RIS [50]. In addition, CYP2D6 PMs show a much longer risperidone half-life than EMs [7]. The dose of risperidone was fixed to 6 mg/day (3 mg twice daily), which is the recommended dose for the management of schizophrenia in adults [53]. All simulations were conducted for a treatment period of 420 h. According to the product information [51], the dose was escalated starting with 1 mg twice a day followed by 2 mg twice daily and 3 mg twice a day from the third treatment day [51]. The daily target dose of the active moiety corresponds to the therapeutic reference range, which was set to a range of 20–60 ng/mL, as published by Hiemke et al. [14]. For plasma concentrations outside the therapeutic range, the daily dose was adjusted until it was within the range. Risperidone/9-OH-RIS ratios for EMs, IMs, PMs, and UMs were calculated and used to compare each metabolizer type.

2.7 Model Evaluation and Statistical Analysis

The evaluation of the PBPK models was carried out in accordance with the Guideline on the Qualification and Reporting of Physiologically Based Pharmacokinetic Modelling and Simulation [54]. Assessment of the plasma concentration–time profiles of the PBPK models was initially based on visual inspections of the observed and predicted plasma concentration. Simulations were carried out with the aid of the computer-assisted Monte-Carlo study on populations with n = 1000. PBPK model quality was analyzed using goodness-of-fit (GOF) plots and weighted residuals versus observed concentration. To describe model accuracy and precision, prediction error (PE), mean prediction error (MPE) and mean absolute prediction error (MAPE) were calculated according to Eqs. 1–3:

$$ \text{PE} \, [\%] = \frac{{\text{predicted}} - {\text{observed}}}{\text{observed}} \times 100 \%$$
(1)
$$ {\text{MPE }}\left[ {\text{\% }} \right] = \frac{1}{n} x \mathop \sum \limits_{i = 1}^{n} {\text{PEi}} $$
(2)
$$ {\text{MAPE }}\left[ {\text{\% }} \right] = \frac{1}{n} x \mathop \sum \limits_{i = 1}^{n} \left| {\text{PEi}} \right| $$
(3)

Pharmacokinetic parameters, such as the median area under the systemic drug concentration–time curve from time zero to infinity (AUC∞), as well as the maximum concentration of a drug (Cmax), and the time it takes until the drug reach the Cmax (tmax), was calculated by PK-Sim®. All statistical and graphical pre-analytics were performed in R (version 3.4.2; R Foundation for Statistical Computing, Vienna, Austria) [55].

3 Results

3.1 Single-Dose PBPK Model

After incorporation of the known physicochemical properties of risperidone in the initial PBPK model, risperidone exposure after a single oral dose accurately described the plasma concentrations for a population of healthy volunteers (Fig. 2). Risperidone is rapidly absorbed and shows a Cmax between 4.2 and 8.8 µg/L within 1–2 h after dosing, according to the CYP2D6 phenotype (Table 5). In PMs the Cmax of risperidone tended to be higher and the elimination much slower than in EMs, IMs, and UMs. Due to the lack of CYP2D6 activity, PMs convert less risperidone into its active metabolite, so that the Cmax of the metabolite is significantly lower than in the other metabolizers (Fig. 2, Table 5). All PBPK models can accurately predict the influence of the CYP2D6 metabolizer status, as can be seen by the respective MPE and MAPE values shown in Table 6. Especially in the concentration range ≥ 1 µg/L, the models show a high accuracy illustrated by a minimal bias (MPE range − 13.8% to + 9.07%) and a good precision (MAPE range 9.70–20.9%) in their prediction. In accordance with this, 78.9% (EMs), 81.8% (IMs), 53.8% (PMs), and 82.4% (UMs) of the plasma concentration values in the concentration range ≥ 1 µg/L were in the 1.25-fold error range in the GOF plot (Fig. 3a). About 90.3% of all weighted residuals versus observed plasma concentrations ≥ 1 µg/L were in the ± 30% range (Fig. 3c), indicating that the distribution of risperidone and 9-OH-RIS in the early phase of administration is predicted more accurately. Especially for plasma concentrations lower than 1 µg/L, a poorer prediction can be seen for all types of metabolizer, suggesting that the clearance of risperidone is under-predicted. This in turn leads to an overestimation of risperidone and an underestimation of 9-OH-RIS plasma concentrations in the lower, subtherapeutic concentration range. The respective values for MPE and MAPE (Table 6) as well as the pharmacokinetic parameters (Table 5) support these observations.

Fig. 2
figure 2

Plasma concentration–time curve profiles (mean and standard deviation) of risperidone and 9-hydroxyrisperidone after the administration of a single 1 mg dose of risperidone in healthy adults according to their cytochrome P450 (CYP) 2D6 phenotype. Solid lines indicate the mean and the colored area indicates the standard deviation of the physiologically based pharmacokinetic model. Circles indicate observed plasma concentrations. a Extensive metabolizers (n = 33); b intermediate metabolizers (n = 26); c poor metabolizers (n = 6); and d ultra-rapid metabolizers (n = 6)

Full size image
Table 5 Mean predicted and observed pharmacokinetic parameters of risperidone and 9-hydroxyrisperidone after single-dose oral risperidone 1 mg
Full size table
Table 6 Bias (mean prediction error) and precision (mean absolute prediction error) for different model approaches depending on different risperidone and 9-hydroxyrisperidone concentration ranges
Full size table
Fig. 3: a
figure 3

Goodness-of-fit plot for model prediction of risperidone and 9-hydroxyrisperidone plasma concentrations in healthy adults. b Weighted residuals vs. observed risperidone and 9-hydroxyrisperidone concentration for healthy adult volunteers over the entire concentration range. The solid black line indicates the line of identity, filled (risperidone) and open (9-hydroxyrisperidone) circles indicate observed data. c Weighted residuals vs. observed risperidone and 9-hydroxyrisperidone concentration for healthy adult volunteers for plasma concentrations ≥ 1 µ/L. The solid black line indicates the line of identity, dashed lines indicate ± 30% range, dotted lines indicate ± 50%, filled (risperidone) and open (9-hydroxyrisperidone) circles indicate observed data. a, b, c Black circles indicate extensive metbaolizer, blue intermediate metabolizer, red poor metaboizer, and green ultra-rapid metabolizer

Full size image

3.2 Steady-State PBPK Model

Using the evaluated PBPK models, steady-state conditions with administration of risperidone 3 mg twice a day were simulated (Fig. 4). All predicted trough plasma concentrations for EMs, IMs, PMs, and UMs during steady-state were in the therapeutic reference range of 20–60 µg/L determined by Hiemke et al. [14]. However, the predicted Cmax values of the active moiety for PMs and IMs were outside this range. Especially in PMs, 57.1% of all predicted plasma concentrations were higher than the therapeutic reference range during steady-state. In the case of IMs, only 26.5% of all plasma concentrations were above the therapeutic reference range.

Fig. 4
figure 4

Mean predicted concentration–time profiles of total active moiety (sum of risperidone and 9-hydroxyrisperidone) after administration of multiple 6 mg daily doses (3 mg twice a day) of risperidone according to their cytochrome P450 (CYP) 2D6 phenotype. The shaded boxes represent the therapeutic reference range between 20 and 60 µg/L. a Extensive metabolizers; b intermediate metabolizers; c poor metabolizers; and d ultra-rapid metabolizers

Full size image

3.2.1 Risperidone Dose Adjustment for Poor Metabolizers (PMs) and Intermediate Metabolizers (IMs)

Due to the high plasma concentrations of total active moiety in PMs and IMs during steady-state, the dose was adjusted (Fig. 5). The steady-state trough plasma concentrations of PMs were + 106% higher than those of EMs (25.3 µg/L in EMs and 52.4 µg/L in PMs) at the same dose. Even the Cmax of PM was significantly higher (+ 53.1%) in PMs than in EMs (50.5 µg/L in EMs and 77.4 µg/L in PMs). This was also evident in the IMs; the trough plasma concentration was + 62.3% higher than that of EMs (25.3 µg/L in EMs and 41.1 µg/L in IMs), and the Cmax was + 38.5% higher in IMs than in EMs (50.5 µg/L in EMs and 70.0 µg/L in IMs). Consequently, the original dose of risperidone 6 mg (3 mg twice a day) was reduced by − 10% to 5.4 mg for IMs and by − 25% to 4.5 mg for PMs. After dose adjustment trough plasma concentrations and Cmax values of the active moiety were within the therapeutic reference range and showed similar plasma concentrations to those of EMs and UMs with risperidone 6 mg (3 mg twice a day).

Fig. 5
figure 5

Dose adjustment by – 10% for intermediate metabolizers and − 25% for poor metabolizers. Mean predicted plasma concentration–time profiles of total active moiety (sum of risperidone and 9-hydroxyrisperidone) after administration of multiple doses. The shaded boxes represent the therapeutic reference range between 20 and 60 µg/L. a Multiple daily doses of risperidone 5.4 mg for intermediate metabolizers; and b multiple daily doses of risperidone 4.5 mg for poor metabolizers

Full size image

3.2.2 Risperidone/9-Hydroxyrisperidone (9-OH-RIS) Ratio

Table 7 describes how the risperidone/9-OH-RIS ratio increases progressively according to reduced CYP2D6 activity. For this purpose, steady-state schedules with 4 mg (2 mg twice a day) and 6 mg (3 mg twice a day) were simulated. The smallest risperidone/9-OH-RIS ratio was observed for UMs and ranged between 0.10 and 0.40 for risperidone 6 mg/day and 0.1 and 0.39 for 4 mg/day depending on the time of the measurement. In contrast, PMs showed greater ratios, ranging between 4.38 and 6.29 for risperidone 4 and 6 mg/day during steady-state. Thus, the mean risperidone/9-OH-RIS ratio increases progressively according to the number of functional CYP2D6 alleles. This resulted in the following gradation of the mean risperidone/9-OH-RIS ratio: 0.16 (UMs), 0.27 (EMs), 0.72 (IMs), and 5.00 (PMs).

Table 7 Calculated risperidone/9-hydroxyrisperidone ratios taken from the physiologically based pharmacokinetic simulations for steady-state plasma concentrations of risperidone and 9-hydroxyrisperidone
Full size table

4 Discussion

In the present study, we analyzed the influence of CYP2D6- and CYP3A4-related physiological changes on the pharmacokinetics of healthy subjects treated with risperidone as a single-dose and in a steady-state condition using a PBPK approach. The transmission of single-dose models to steady-state was possible due to linear pharmacokinetics of the active moiety [38, 50, 51]. The two phenotypes IM and PM were the focus of the investigations as they metabolize risperidone significantly slower (Fig. 2) due to a reduced number (IM) or lack (PM) of CYP2D6 active alleles [19, 49].

4.1 PBPK Model Development

For the first time, various PBPK models for the prediction of the plasma concentrations of risperidone and 9-OH-RIS were successfully developed in different CYP2D6 phenotypes (Fig. 2). Overall, the absorption, distribution, and elimination phases of the single-dose concentration–time courses were accurately predicted by the models. Especially in the plasma concentration range ≥ 1 µg/L, the models showed a high predictive power supported by minimal MPEs and good MAPE values (Table 6). Even the pharmacokinetic parameters Cmax and tmax were predicted very well. In contrast, plasma concentrations below 1 µg/L showed larger deviations between the observed and predicted values. However, this concentration range was considered less important during model development, which was for multiple reasons. First, we developed single-dose models instead of multiple doses as used in practice. During steady-state, risperidone and 9-OH-RIS accumulate in the body, which leads to higher plasma concentrations, recognizable by the therapeutic reference range of 20–60 µg/L [14]. Secondly, the clinical and pharmacological relevance of such low concentrations is questionable, as the therapeutic reference range of the active moiety is 20–60 times higher. For this purpose, plasma concentrations below 1 µg/L were neglected during model development.

4.2 Limitations

Potential limitations should be taken into account when interpreting the results of PBPK model development. First, data from healthy volunteers was used rather than patients, which may be limiting when using the results for clinical applications. The applicability of the model needs to be evaluated further with clinical pharmacokinetic data including elderly patients. Furthermore, it is apparent that not only genetic but also pharmacological (e.g., co-medications such as inhibitors or inducers) and other variables such as age or co-morbidity should be considered to personalize dosing in the clinical environment [56].

4.3 Dose Adjustment for IMs and PMs

As reported by de Leon et al. [22], the CYP2D6 PM phenotype appears to be associated with risperidone’s ADRs and discontinuation due to ADRs. According to the authors’ experience, CYP2D6 PMs show a three to six times higher risk of experiencing ADRs related to risperidone or discontinuations because of ADRs [22, 57]. PMs and IMs consistently showed significantly higher plasma concentrations in our PBPK models than EMs and UMs at a dose of risperidone 3 mg twice a day (Fig. 4), which leads to an active moiety within the supratherapeutic range (> 60 µg/L). As a consequence, drug plasma concentrations within the supratherapeutic range can lead to an increased occurrence of ADRs [14]. So, an optimal dose setting for PMs is very important to reduce the rate of ADRs and discontinuation due to ADRs.

In order to reduce the proportion of the plasma concentrations of the active moiety in the supratherapeutic area, a dose adjustment, especially for PMs and IMs, should be considered. The PBPK analysis results in a recommendation for a dose adjustment (Fig. 5) in subjects with a reduction in the number (IMs) or lack (PMs) of CYP2D6 active alleles of − 10% for IMs and − 25% for PMs. This observation is consistent with the results in the literature. Mas et al. [58] and Kirchheiner et al. [59] postulated a strong correlation between CYP2D6 status and risperidone dosage. Both authors recommended a dose adjustment for PMs of − 20% [58] and − 13% [59] and for IMs of − 13% [58] and – 5% [59]. Dose adjustment in PMs is also suggested in the clinical guideline for using CYP2D6 genotypes in patients taking antipsychotics published by de Leon et al. [57]. Williams [60] recommends that CYP2D6 PMs take half or less of the upper range in normal circumstances. However, no adjustment is necessary in UMs.

4.4 Risperidone/9-OH-RIS Ratio

Steady-state plasma concentration of risperidone and 9-OH-RIS can be used to calculate the risperidone/9-OH-RIS ratio, which is a marker of CYP2D6 activity. In patients taking oral risperidone, a risperidone/9-OH-RIS ratio > 1 indicates the patient is a genetic CYP2D6 PM or is taking a potent CYYP2D6 inhibitor that completely inhibits CYP2D6 activity [61]. Our calculated risperidone/9-OH-RIS ratio for PM ranged from 4.38 to 6.29 (Table 7) and was therefore above the ratio of 1. In comparison, the remaining phenotypes showed lower ratios, which ranged from 0.19 to 0.59 (EMs), 0.59 to 1.15 (IMs), and 0.10 to 0.40 (UMs). Thereby, the risperidone/9-OH-RIS ratio seems to be an inverse measurement of CYP2D6 activity. The increase in the number of active CYP2D6 alleles indicates a decreased ratio due to the metabolism of risperidone. As a result, UMs, who have at least three copies of functional alleles [45], show lower ratios because UMs metabolize risperidone faster and the proportion of the active metabolite 9-OH-RIS is higher. In contrast, PMs, who have two non-functional alleles [45], show high risperidone/9-OH-RIS ratios because only a small proportion of risperidone is converted into the active metabolite due to the lack of CYP2D6 enzymes. CYP2D6 PMs do not have CYP2D6 activity but are able to produce some 9-OH-RIS using CYP3A4 [9, 24].

4.5 Phenotyping Using the Risperidone/9-OH-RIS Ratio

Due to the significant differences in the ratio, only the risperidone/9-OH-RIS ratio needs to be calculated to determine whether the patient is a PM or not. For assessment of the PM, the risperidone/9-OH-RIS ratio is an equivalent alternative to genotyping. Therefore, determination of the plasma concentrations of risperidone and 9-OH-RIS would be sufficient to exclude a PM. If the observed plasma concentration value exceeds 4.38 (Table 7), the patient can be classified as a PM. Such a classification into the several types of metabolizers would make sense during TDM, which is recommended for risperidone by many experts [14]. With respect to the genetic polymorphism in CYP2D6, TDM of risperidone and its main metabolite is an important tool to improve patient safety. So, when TDM is available, phenotyping could be used to classify a patient as a PM.

5 Conclusion

The present study generated and evaluated a PBPK model using PK-Sim® for risperidone and its active metabolite 9-OH-RIS for CYP2D6 EMs, IMs, PMs, and UMs taking CYP3A4 into account. Especially in the concentration range ≥ 1 µg/L, the models show high accuracy which is illustrated by a minimal bias and good precision. PMs and IMs showed plasma concentrations of the active moiety above the therapeutic reference range. To reduce the proportion of plasma concentrations in the supratherapeutic range, the dose was adjusted by – 10% for IMs and – 25% for PMs when using risperidone 3 mg twice a day [53]. As a marker of CYP2D6, the risperidone/9-OH-RIS ratio during steady-state was calculated showing that PMs have a mean ratio of 4.96 whereas EMs show a mean ratio of 0.27.

In the future, calculation of the risperidone/9-OH-RIS ratio from TDM measurements could be used to phenotype each patient. To ensure the best therapy for each subject, dose adjustment, especially for PM and IMs, must be considered to reduce ADRs and discontinuation due to ADRs in different CYP2D6 phenotypes.