**1. Introduction**

Tramadol is an orally available, centrally acting, weak-opioid analgesic drug [1]. The anti-nociceptive effect of tramadol is due to two mechanisms: an opioid mechanism, and a non-opioid mechanism [2]. Tramadol acts as a μ-opioid receptor agonist, like traditional opioids. It also has analgesic e ffects by inhibiting reuptake of monoamine neurotransmitters, such as 5-hydroxytryptamine (5-HT) and noradrenaline [3,4]. These mechanisms lead to reduced pain signal conduction and analgesic effects. Tramadol is a racemate, and analgesic mechanisms di ffer depending on the isomer:

(−)-tramadol exhibited ≈ 10-fold higher noradrenaline reuptake inhibitory activity than (+)-tramadol, and (+)-tramadol showed ≈ 4-fold higher 5-HT reuptake inhibitory activity than (−)-tramadol [5].

Tramadol is predominantly metabolized in the liver. Approximately 10%–30% of administered tramadol is excreted, unmetabolized, in the urine. The well-known metabolic pathway of tramadol is divided into three major pathways: *O*-desmethyltramadol (M1) is produced by cytochrome P450 (CYP) 2D6, and *N*-desmethyltramadol (M2) is produced by CYP2B6 and CYP3A4. These metabolites are converted to either *<sup>N</sup>*,*O*-didesmethyltramadol (M5) and other inactive metabolites by CYP2D6, CYP2B6, and CYP3A4, or converted to glucuronides by UDP-glucuronosyltransferase (UGT) 1A8 and UGT2B7 [6]. M1, a major active metabolite of tramadol, has about 700-fold higher affinity for μ-opioid receptors than tramadol [7]. M5 also has 24-fold higher affinity for μ-opioid receptors than tramadol [3]. Tramadol is mainly considered to inhibit monoamine reuptake, while M1 and M5 bind to μ-opioid receptors and exhibit analgesic effects. Thus, genetic polymorphism of CYP2D6 could have an effect on the risk of adverse events during tramadol administration [4].

A physiologically-based pharmacokinetic (PBPK) approach is a bottom-up approach that requires data about the physicochemical properties and pharmacokinetic (PK) parameters (i.e., absorption, distribution, metabolism, and excretion; ADME) of the target drug [8,9]. In addition, the PBPK model considers bodyweight, height, organ volume, blood flow, and inter-individual variation for metabolizing enzymes and transporters [10]. Therefore, the PBPK model can be used to predict plasma concentration–time profiles more closely than traditional compartmental PK models [11]. In this regard, PBPK modeling can predict human PK profiles using in vitro or preclinical study data from drug development. Further, such modeling is also used to investigate the interaction potential with other drugs or food, and to predict PK profiles in special populations, such as pregnan<sup>t</sup> women, geriatric patients, or children [9,12].

Studies predicting the PK profile of tramadol using PK modeling have been reported previously. Many articles used a population PK approach with nonlinear mixed-effects modeling (NONMEM), however a PBPK approach was rarely used to predict the PK profile of tramadol [13–15]. When tramadol was administered, M1 also had impact on efficacy and toxicity [16]. Therefore, M1 should be integrated for PBPK model of tramadol for better interpretation.

The aims of this study were to develop a PBPK model that could predict the concentration–time profiles of tramadol and M1 in Koreans, and to investigate effects of the CYP2D6 genotype on PK profiles at routinely administered doses. The developed PBPK model was applied to predict the effects of CYP2D6 genotype and tramadol dosage on the plasma concentration profiles of tramadol and M1 in a healthy Korean population.

#### **2. Materials and Methods**

#### *2.1. Clinical Study Design*

The clinical study was approved by the Institutional Review Board of Keimyung University (Deagu, Republic of Korea, approval number: 40525-201509-BR-70-02, 23 February 2016) and Kyungpook National University Hospital (Daegu, Republic of Korea, approval number of clinical trial: 2016-08-005, 24 August 2016) and carried out at the Kyungpook National University Hospital Clinical Trial Center (Daegu, Republic of Korea). A total of 23 subjects participated who voluntarily agreed to take part in the clinical study and signed their informed consent. Subject characteristics are presented in Table 1. All subjects received a 100-mg tramadol hydrochloride tablet (Tridol® extended-release (ER); Yuhan Pharmaceutical, Seoul, Korea) five times at 12-h intervals. Whole blood was collected in an anticoagulant tube at pre-dose, and at 0.5, 1, 1.5, 2, 2.5, 4, 6, 8, 10, 12, 24, 48, and 72 h after administration. The obtained whole blood was used for CYP2D6 genotyping, and the plasma was separated for determination of tramadol and M1 [17].


**Table 1.** Demographic characteristics (*n* = 23).

> SD, standard deviation.

## *2.2. CYP2D6 Genotyping*

The determination of CYP2D6 genotype was performed for *CYP2D6\*2* (normal function), *CYP2D6\*5* (no function), and *CYP2D6\*10* (decreased function). Genotyping for *CYP2D6\*2* and *CYP2D6\*10* was performed using pyrosequencing. *CYP2D6\*5* was sequenced by long polymerase chain reaction (PCR) because of deletion of a specific sequence. Pyrosequencing was performed using Pyromark Q96 ID and Pyromark Gold Q96 reagents (Qiagen, Hilden, Germany). The conditions of PCR (total 35 cycles) were: denaturation (94 ◦C for 30 s), annealing (56 ◦C for 30 s), and polymerization (72 ◦C for 30 s). The processes were finished by extension at 72 ◦C for 5 min. *CYP2D6\*5* and duplication were determined by long-PCR, as previously described [18,19]. The *CYP2D6* phenotype was determined based on genotype and activity score [20–23]. *CYP2D6*\*5 was a non-functional allele and homozygous *CYP2D6*\*5 was classified as a poor metabolizer (PM). Homozygous *CYP2D6*\*10 was classified as an intermediate metabolizer (IM).

#### *2.3. Determination of Tramadol and O-Desmethyltramadol (M1) Using LC-MS*/*MS*

The obtained whole blood samples were immediately centrifuged at 4 ◦C, 3000 rpm for 10 min. The isolated plasma samples were transferred to a new microcentrifuge tube and kept at −70 ◦C until analysis. The plasma samples were completely thawed, then 10 μL of internal standard (tramadol 13C-d3 for tramadol, and M1-d6 for M1) were added to 100 μL samples and mixed briefly. A total of 300 μL of acetonitrile was added, and then samples were mixed thoroughly for 30 s prior to centrifugation at 2500 rpm for 10 min. The organic solvent layer was transferred to a new polypropylene tube and evaporated under nitrogen gas at 40 ◦C. Methanol 200 μL was added into tubes containing pellets for reconstitution, and 5 μL of reconstituted sample were analyzed. Analyses were carried out on API3200 tandem mass spectrometry system (AB SCIEX, Framingham, MA, USA) equipped with an Agilent 1260 series HPLC system (Agilent Technologies, Santa Clara, CA, USA). Separation of tramadol and M1 was conducted using a Luna C18 column (5.0 μm, 2.0 × 50 mm; Phenomenex, Torrance, CA, USA). Five millimoles ammonium formate and 0.1% formic acid in methanol (A), and 5 mM ammonium formate solution (B), were used for the mobile phase. The used gradient method was as follows: 0–2 min (97%–5% B), 2–4 min (5% B), 4–5 min (5%–97% B), and 5–8 min (97% B). Electrospray ionization-positive ion mode was used for mass detection. The mass transitions (m/z) used were 264.2 →58.1 for tramadol, 268.2 →58.1 for tramadol internal standard (IS), 250.2 →58.2 for M1, and 256.2 →64.1 for M1 IS. To obtain pharmacokinetic parameters, non-compartmental analysis (NCA) was performed using Phoenix (Certara Inc., Princeton, NJ, USA).

#### *2.4. Parallel Artificial Membrane Permeability Assay (PAMPA)*

To determine the permeability of tramadol, a parallel artificial membrane permeability assay (PAMPA) was performed [24]. Gentest ™ Pre-coated PAMPA Plate System (Corning, Tewksbury, MA, USA) was used for the permeability assay. All the processes were followed according to the manufacturer's protocol. Tramadol hydrochloride, dimethyl sulfoxide (DMSO), phosphate-bu ffered saline (PBS), and acetonitrile were purchased from Sigma-Aldrich (St. Louis, MO, USA). Tramadol

stock solution (1 mM) was prepared using 100% DMSO and diluted to 15 μM using PBS (pH 7.4). The PAMPA plate was equilibrated for 30 min at room temperature before performing the permeability assay. PBS 200 μL was dispensed on the acceptor side and 300 μL of working solution was dispensed on the donor side. Incubation was carried out at room temperature for 5 h, and the acceptor and donor side bu ffers were analyzed using liquid chromatography-tandem mass spectrometry (LC-MS/MS). A total of 12 replicated samples were assayed and mean permeability was calculated and applied to the model.

#### *2.5. Assessment of Intrinsic Clearance of M1*

For the experiments, *O*-desmethyltramadol HCl (M1), glucose 6-phosphate, glucose 6-phosphate dehydrogenase, MgCl2, β-nicotinamide adenine dinucleotide phosphate (NADP), chlorpropamide, Trizma ® base, Trizma ® hydrochloride, DMSO and formic acid were obtained from Sigma-Aldrich (St. Louis, MO, USA). To evaluate the intrinsic clearance of M1 by CYPs, metabolic stability studies under NADPH system were conducted in human liver microsoms (HLM) [25,26]. For details, NADPH-generating system (1.3 mM NADP<sup>+</sup>, 3.3 mM glucose 6-phosphate, 3.3 mM MgCl2, and 0.4 unit/mL glucose-6-phosphate dehydrogenase) and HLM 0.25 mg/mL were added and preincubated at 37 ◦C for 5 min. Then, 20 μM M1 was added and reacted at 37 ◦C for 0, 1, 5, 10, 20, 30, and 40 min, respectively. The total volume of the reaction mixture was 100 μL. After each reaction, the reaction was terminated by addition of 150 μL of acetonitrile containing an internal standard (100 ng/mL chlorpropamide). All experiments were repeated duplicated and the samples were vortexed for 5 min and centrifuged (13,000 rpm, 4 ◦C) for 15 min. Then, supernatants were injected into LC-MS/MS and M1 were analyzed. The concentration of M1 (20 μM) at 0 min was used to evaluate the metabolism of CYP through the change of drug concentration over time.

#### *2.6. Qunatitaion Methods of M1 in In Vitro Experiments Using LC-QTOF*

High-performance liquid chromatography (HPLC)-grade acetonitrile and deionized water were obtained from Berdick and Jackson (Muskegon, MI, USA). HLM (50 donors pooled) were purchased from Corning. Analyses were carried out on Sciex QTOF 5600 plus (AB SCIEX, Framingham, MA, USA) equipped with an Agilent 1260 series HPLC system (Agilent Technologies, Santa Clara, CA, USA). For quantitation of M1, the compounds were separated on a Poroshell 120 (4.6 × 50 mm, 2.7 μm; Agilent Technologies) with an isocratic mobile phase consisting of acetonitrile and 0.1% aqueous formic acid (70:30 *v*/*v*) at a flow rate of 0.5 mL/min. The overall run time was 5 min per sample. The column and autosampler temperatures were maintained at 40 ◦C and 4 ◦C, respectively. Time-of-flight mass spectrometry analysis was selected in positive ion mode for the sample analysis. The quantitative analytical data were processed using PeakView ® (Version 2.2.0; AB SCIEX, Framingham, MA, USA) and MultiQuant ® (Version 3.0.2; Framingham, MA, USA), and the formulas C15 H23NO2 (M1), and C10 H13ClN2O3S (chlorpropamide) were used for quantitation.

#### *2.7. Development of PBPK Model for Tramadol and M1*

PBPK model development was performed using SimCYP ® simulator version 18 (Certara, She ffield, UK). Most of the parameters for tramadol and M1 were entered with reference to the literature. According to previous reports, tramadol was not substrate for P-glycoprotein (P-gp) (ABCB1) and the role of proton-based pumps such as OATP for tramadol permeability was unclear [27,28]. Therefore, in this study, the permeability of tramadol was determined by PAMPA assay. The advanced drug absorption and metabolism model was used to consider the ER formulation, and the dissolution profiles of Tridol ® ER 100 mg (Yuhan Pharmaceutical, Seoul, Korea) were applied. The elimination profile for M1 applied the intrinsic clearance by HLM. Kp scalar of tramadol was set to match the observed Vss and the predicted Vss value in the model, and Kp scalar of M1 was obtained from parameter estimation. In clinical study, the Vss of tramadol was calculated to 2.6 L/kg by non-compartmental analysis. That of M1 was not calculated because the exact dose of M1 is unknown. Intrinsic clearance involved in tramadol metabolism was estimated using retrograde model option, and human liver microsomal intrinsic clearance was applied to M1 library. The renal clearance of tramadol and M1 were applied for the predicted value which is the closest to the observed blood concentration–time profile by parameter estimation. The PBPK model was evaluated so that it could e ffectively predict PK profiles for tramadol and M1 when observed mean plasma concentrations fitted to the predicted plasma concentration–time profile and its 90% confidence interval (CI). The other evaluation criteria were geometric mean ratio for peak plasma concentration at steady-state (Cmax,ss), and area under the plasma concentration–time curve at last observation at steady-state (AUClast,ss), and the 90% CI for these values. If the geometric mean ratio and its 90% CI were within the range 0.7–1.43, the model was considered to fit well.

#### *2.8. Prediction of Changes in Concentration–Time Profiles for Tramadol and M1 According to CYP2D6 Genotype and Dosing Regimen*

The therapeutic range and toxic range of tramadol and M1 were determined by reference to the literature. Because the manufacturer's recommended acceptable maximum dose of Tridol ® ER was 400 mg per day, the tramadol ER tablet was administered to a virtual healthy Korean population at 100 and 200 mg (5 times at 12-h intervals) to simulate the change of concentration–time profiles for tramadol and M1. This simulation assumes linear PK properties for multiple doses of tramadol 100 mg and 200 mg [29]. The e ffect of CYP2D6 genotype was also simulated for tramadol 100 mg and 200 mg for populations consisting of CYP2D6 poor metabolizers (PM), intermediate metabolizers (IM), extensive metabolizers (EM), and ultra-rapid metabolizers (UM).
