A la Dirección de Recursos Humanos le corresponde las siguientes atribuciones:

A four-stage strategy was employed for model development and validation.

Figure 2.4: Work flow process of model development for the simulation of AL pharmacokinetics in adults and children in the presence and absence of DDIs .

The workflow was implemented for the prediction of the pharmacokinetics of artemether and lumefantrine in adults and paediatrics. The workflow also showed process involved in DDI prediction in adults and children, followed by dosage optimisation of artemether and lumefantrine in the presence of DDI with a rifampicin-based anti-tuberculosis regimen.

2.3.1.1 Step 1: Base model development and validation

This step focussed on the development of a base model on Simcyp®. The compound physicochemical properties of artemether and lumefantrine obtained from literature were used to develop a model that recovered the absorption, distribution and the elimination phases of the concentration time profiles observed from clinical studies and to simulate the key pharmacokinetic properties of artemether and lumefantrine such as the maximum concentration (Cmax), areas under the curve (AUCs) and the lumefantrine day 7 concentration

where required at different dose regimens as reported in the clinical studies.

When required, model parameters were optimised to produce parameters with better certainty, and which best recovered published clinical drug profiles. Parameter optimisation was done with the Weighted Least Square (WLS) approach and the Nelder-Mead minimisation method on Simcyp®. For artemether, literature- reported isozyme specific hepatic intrinsic clearances were utilised for the description of drug metabolism (Table 2.3). For lumefantrine, the isozyme specific hepatic intrinsic clearance (CLint) was back-calculated

using the Simcyp® retrograde calculator from the oral clearance and assuming CYP3A4 was the predominant isozyme for lumefantrine metabolism (208). The Simcyp® retrograde calculator is a feature on Simcyp® which permits the reverse calculation of intrinsic clearance of drug from either oral or intravenous clearance data. In certain cases, in vitro hepatic intrinsic clearance (CLint)of a drug molecule is not available, since PBPK modelling

is dependent on this parameter to mechanistically predict the elimination of the drug, this

parameter may be calculated using the data from adult human in vivo intravenous (CLiv) or

oral (CLpo) clearance as long as the elimination pathway and hepatic uptake factor for the

drug is known. The accuracy of such prediction is dependent on the accuracy of the in vivo intravenous (CLiv) or oral (CLpo) clearance reported for the drug. The reverse calculation

method back-calculates the CLint values from the oral or IV clearance using information on

additional clearance (for example, the biliary clearance), ideal renal clearance for a 20-30 year old healthy male subject, hepatocyte uptake, unbound fraction of drug in plasma and blood to plasma ratio. This model also factors in the impact of fraction of drug absorbed (fa),

Table 2.3: Input base model parameter values and predicted PBPK values for use in the simulation of artemether, lumefantrine.

a _{Simcyp® mechanistic prediction;} b _{Default values in Simcyp®;} c _{Parameter estimated;} d

Simcyp® retrograde calculation from population estimates of CLpo followed by parameter

estimation (final optimised value: 0.85 µL/min/pmol for CYP3A4); MW: Molecular weight; Peff: human effective permeability; B/P: blood-to-plasma ratio; CLint: in vitro intrinsic clearance; Vss: Steady state volume of distribution; ISEF: Intersystem extrapolation factor for scaling CYP in vitro kinetic data; Ki: concentration of inhibitor supporting half-maximal inhibition; Kinact: inactivation rate of the enzyme; Kapp: concentration of mechanism based inhibitor associated with half-maximal inactivation rate.

Parameters Artemether Lumefantrine

Compound type Monoprotic base Diprotic base

Molecular weight (g/mol) 298.4(209) 528.94(209)

Log P 3.53(210) 8.70(211) Fu 0.05(212) 0.003 (212) pKa 1 3.9(209) _14.1(209) pKa 2 - 9.80(209) B/P 1.09a 0.80(213) Vss (L/kg) 1.77a 0.70a Peff (10-4 cm/s) 2.74a 8.3a Kp scalar 1b ₁b Solubility (mg/mL) 0.012(214) 0.002(215) CLpo (L/min) - 0.25(172) CLint3A4 (µL/min/pmol) 1.47(216) 2.61c,d CLint2B6 (µL/min/pmol) 9.31(216) - ISEF CYP3A4 1b - ISEF CYP2B6 1b -

Absorption model ADAM ADAM

Due to unavailability of published data for the human jejunal effective permeability

(Peff), the molecular descriptors (PSA and HBD) of AL, were predicted and optimised for

artemether and lumefantrine respectively. The Peff was predicted using the equation below

(217):

𝑙𝑜𝑔𝑃_𝑒𝑓𝑓 = 4 − 2.546 − 0.011 × 𝑃𝑆𝐴 − 0.278 × 𝐻𝐵𝐷 ( 43 )

The Kp scalar, which is a scaling factor used to scale the tissue to plasma partitioning coefficient (Kp) of the drug to the expected steady state volume of distribution (Vss) (218) of the drug were further optimised for AL using a parameter estimate method within Simcyp® to yield optimal estimates of tissue distribution and Vss prediction.

Furthermore, for artemether, where necessary, the in vitro metabolic clearance was optimised through the parameter estimation of the Inter System Extrapolation Factor (ISEF) (Table 2.3). The ISEF is a scaling factor used for prediction of the human drug clearance (CLint) from in vitro CLint obtained from recombinantly expressed human cytochrome P450s

(rhCYP). This is important because the intrinsic activity per unit CYP between rhCYP and human hepatic enzymes differs, therefore, the ISEF allows for IVIVE of human drug clearance from rhCYPs (219).

The published clinical studies used for the validation of artemether and lumefantrine included a study conducted in 120 adult subjects who were orally dosed the branded AL

combination called Coartem®,and studies conducted in 16 subjects who were orally dosed

the branded AL combination called Riamet® _{(220). For lumefantrine an additional study} included a 6-dose study conducted in 17 subjects (221).

2.3.1.2 Step 2: Simulation and validation of drug-drug (DDI) interactions in adults This step focussed on the validation of the adult DDI predictions. CYP3A4 inhibition and induction mechanisms were simulated using ketoconazole and rifampicin respectively.

the validation was conducted against, and primarily included matching age ranges, male-to- female ratios and identical dose/dosing intervals. In order to validate the capability of the model to predict a broad range of DDIs, the pre-validated Simcyp® in-built compounds ketoconazole and rifampicin were directly utilised in simulations as candidates to simulate CYP3A4 inhibition DDIs (ketoconazole) and CYP3A4 induction DDIs (rifampicin). Rifampicin and ketoconazole compounds were used in simulations without modification from the library of pre-validated drug molecules within the Simcyp® simulator, using a 1st- order absorption model and assuming dosing in solution form.

Where the Simcyp® ADAM (Advanced Dissolution Absorption Model) was used, an immediate release formulation with an applied diffusion layer model was utilised for modelling with literature-reported solubility parameters included. Where simulations were performed in paediatrics, all APIs were assumed to be dosed in solution form, mimicking the dispersible/crushed application of AL in paediatric subjects (26).

A previously validated isoniazid compound file (222) was used for all rifampicin DDI simulations to account for the impact of isoniazid mediated CYP3A4-inhibition associated with TB chemotherapy. All simulations included both rifampicin (as the primary perpetrator) and isoniazid (as the secondary perpetrators), however results are presented for the key interactions between AL and rifampicin only, and reflects the clinical net effect of CYP3A4 induction with the clinical use of the combination of rifampicin and isoniazid in DDI- focussed studies (223-225).

Clinical studies demonstrating such a DDI were obtained from Lefèvre et al (2002) who studied AL with ketoconazole (220) (single dose of 80/480 mg of AL and 5 day treatment with ketoconazole) and Lamorde et al (2013) (226), who studied AL DDI with rifampicin where rifampicin was dosed at 10 mg/kg for the duration of the study with AL dosed as six 80/480 mg doses (12 hourly) on days 8, 9 and 10.

2.3.1.3 Step 3: Simulation and validation of base model in paediatric population In this step the previously validated adult base model of artemether and lumefantrine were scaled to paediatric groups reported in clinical trials. This step focussed on the validation of artemether and lumefantrine model predictions in paediatrics. In these studies, weight bandings were simulated based on dosing strategies for AL if the clinical study did not use a weight normalised dosing method. Dosing boundaries were set at 1 tablet for 5- 14.9 kg, 2 tablets for 15-24.9 kg and 3 tablets for 25-34.9 kg and trials were run to ensure, where possible, an equal proportion of subjects were included into each distribution banding based on the total number of subjects recruited within each reported trial. Simulated profiles were body weight stratified according to the corresponding clinical trial and analysed consequently.

2.3.1.4 Step 4: Prediction of drug-drug interaction between AL and rifampicin-based drug in the paediatric population

This step focussed on simulations to predict the impact of rifampicin-mediated DDIs on artemether and lumefantrine pharmacokinetics in children of 2-5 years of age over a weight boundary of 5-14.9 kg or 15-24.9 kg. In these simulations, trials of 100 subjects were simulated and analysed with appropriate weight-based dosing (see above) and under treatment of rifampicin with AL. A 100-subject simulation was run in a 10x10 trial (10 subjects per trial with 10 trials) to ensure that reasonable inter-/intra individual variability was captured within the model simulations. However, as simulations are not possible with defined age and weight ranges, pooling and post-processing of output data was conducted to match individuals to the required age-weight boundary conditions for the study.

For all validation steps, unless otherwise stated, all observed data sets were obtained from ‘supervised’ administration groups in reported clinical studies and simulated under ‘fed’ conditions. Furthermore, unless otherwise stated all simulations included subjects of ≥ 5 years.