Population pharmacokinetics of orally administrated bromopride: Focus on the absorption process

Journal Pre-proof

Population pharmacokinetics of orally administrated bromopride: focus on the absorption process Larissa Lachi-Silva , Aline B. Barth , Gustavo Mendes Lima Santos , Malidi Ahamadi , Marcos Luciano Bruschi , Elza Kimura , ´ Diniz Bibiana Verlindo de Araujo ´ , Andrea PII: DOI: Reference:

S0928-0987(19)30354-9 https://doi.org/10.1016/j.ejps.2019.105081 PHASCI 105081

To appear in:

European Journal of Pharmaceutical Sciences

Received date: Revised date: Accepted date:

14 May 2019 26 August 2019 17 September 2019

Please cite this article as: Larissa Lachi-Silva , Aline B. Barth , Gustavo Mendes Lima Santos , Malidi Ahamadi , Marcos Luciano Bruschi , Elza Kimura , Bibiana Verlindo de Araujo ´ , ´ Andrea Diniz , Population pharmacokinetics of orally administrated bromopride: focus on the absorption process, European Journal of Pharmaceutical Sciences (2019), doi: https://doi.org/10.1016/j.ejps.2019.105081

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier B.V.

Population pharmacokinetics of orally administrated bromopride: focus on the absorption process a

Larissa Lachi-Silva , Aline B. Bartha, Gustavo Mendes Lima Santosb, Malidi Ahamadia, Marcos Luciano Bruschic, Elza Kimura d, Bibiana Verlindo de Araújoe, Andréa Diniza* a

Pharmacokinetics and Biopharmaceutical Laboratory (PKBio), Pharmacy

Departament, State University of Maringa, Maringá-PR, Brazil. b

Brazilian Health Survaillance Agency (ANVISA), Brasilia-DF, Brazil.

c

Laboratory of Research and Development of Drug Delivery System (LABSLiF),

Pharmacy Department, State University of Maringa, Maringá-PR, Brazil. d

Clinical Research and Bioequivalence Center (NPC-BIO), University Hospital,

State University of Maringa, Maringá-PR, Brazil e

Pharmaceutical Sciences Graduate Program, Federal University of Rio Grande

do Sul, Porto Alegre-RS, Brazil. *Corresponding author at: Department of Pharmacy - Universidade Estadual de Maringá, Avenida Colombo 5790, Maringá 87020-900, Paraná, Brazil; Tel.: +55 44 3011 4937, fax: +55 44 3011 4999, E-mail: [email protected]

Abstract Bromopride is a prokinetic and antiemetic drug used to treat nausea and vomiting. Although its prescription is common in Brazil, there is a lack of studies about bromopride pharmacokinetics. Therefore, the aims of this study were to investigate the population pharmacokinetics of bromopride and to evaluate the influence of covariates on its absorption. This study is a retrospective analysis of data collected from bioequivalence studies. The data was modeled using MONOLIX 2018R2. Assuming one-compartment and linear elimination, the absorption phase was evaluated with different structural models. The model of sequential first- and zero-order with combined error and exponential interindividual variability in all parameters best described the atypical absorption profile of bromopride. Population estimates were first-order absorption rate (ka) of 0.08 h-1, fraction of dose absorbed by first-order (Fr) of 32.60%, duration of the zero-order absorption (Tk0) of 0.88 h with latency time (Tlag) of 0.47 h, volume of distribution of 230 l and clearance of 46.80 l h -1. Bodyweight affects Tk0, dosage form was found to correlate with Tk0 and Tlag, while gender affects Tlag. However, simulations evaluating the clinical importance of these covariates on steady-state indicated minimal changes on bromopride exposure. The mixed absorption model was reasonable to describe the absorption process of bromopride because it had the flexibility to fit multiple-peaks profile and shows good agreement with physicochemical properties of drug.

Key words: Bromopride; Monolix; Model selection; Absorption models; Solid dosage forms; Nonlinear mixed-effects modeling.

1. Introduction Bromopride [4-amino-5-bromo-N-(2-(diethylamino)ethyl-2methoxybenzamide] is a dopamine receptor (D2-like receptor family) antagonist, used as prokinetic and antiemetic agent. This drug is used to treat gastric motility dysfunctions, nausea and vomiting (Brodie et al., 1986; Dunne et al., 2013; Roila et al., 1985). Bromopride is the bromo analog of metoclopramide. Extrapyramidal reactions characterized by acute dystonia and tardive dyskinesia are the most serious adverse effects described for bromopride. Furthermore, prolonged use of this drug, can cause hyperprolactinemia associated with gynecomastia and sexual impotence. (Barbieri and Hames, 2011). In the 2000s Food and Drug Administration (FDA) mandated the market withdrawal of the prokinetic drug cisapride, once its use had been associated with reports of severe adverse effects including some deaths. This episode had great impact on the utilization of related drugs, and cisapride use was replaced by other prokinetic drugs. While the United States (US) had an increase in metoclopramide use (Glessner and Heller, 2002), Brazil had an increase in bromopride use (Barreira and Magaldi, 2009). Bromopride is not available in the US, but is marketed in countries like Taiwan, Italy, and Brazil (Dunne et al., 2013). In Brazil, we have around 40 bromopride formulations, approved by ANVISA (Brazilian Health Surveillance Agency), currently marketed in several dosage forms, such as capsules, tablets, oral and injectable solutions, as well as pediatric drops (ANVISA, 2019). Although widely used, there is a lack of clinical evidence to ensure the safety of bromopride, given that most of the information we have was extracted from metoclopramide studies (Barbieri and Hames, 2011). Since oral administration is the most commonly used route of administration, and different oral dosage forms of bromopride are commercially available, the evaluation of its behavior after an oral intake can be useful to develop formulations with enhanced performance and for dose adjustment. The pharmacokinetics of bromopride remains poorly studied and there is a limited number of articles available in the literature. The intravenous pharmacokinetics

of bromopride was previously described by Brodie et al. (1986) and reported as a one-compartmental model. Although these authors had both intravenous and oral administration data, they did not describe the absorption process, limiting the findings to the low bromopride bioavailability. All Brazilians bromopride products present in their drug labeling Brodie's results as pharmacokinetic information for patients. Considering the lack of information of bromopride pharmacokinetics evaluation, the aim of this study was to describe bromopride population pharmacokinetics following a single 10 mg oral dose administration of commercially available capsules or tablets to healthy volunteers. To our knowledge, this is the first time a population pharmacokinetic model was developed to characterize the absorption process of bromopride, and also to assess the potential effects of intrinsic and extrinsic factors on bromopride pharmacokinetics and exposure. Furthermore, simulations of the final pharmacokinetic model were performed to evaluate the impact of covariates on the drug exposure at steady-state.

2. Materials and methods 2.1 Dataset This study is a retrospective analysis of data obtained from bioequivalence studies presented to ANVISA by different pharmaceutical companies as a part of drug registration. Several submissions were analyzed, and data from three randomly studies were collected. The data including plasma concentration profiles and volunteer characteristics (demographic and laboratory tests) of six different bromopride formulations (four capsules and two tablets) from different manufacturers, as summarized in table 1. All the selected studies presented approved protocols, conducted in accordance with the ANVISA requirements (RDC 1170/2004 and RDC 899/2003) and the Declaration of Helsinki, all healthy volunteers provided written informed consent. The procedure of different study design has been described in detail in the registration form of each manufacturer, and precisely described the date and time of each blood sample collected for drug concentration analyzes. Samples below the limit of quantification (BLQ)

represented less than 2% of the data, and they were excluded and treated as missing in the analysis. Study BE1 was an open-label, randomized, single-dose, two-period, and two-sequence crossover design with an adequate washout period between the two treatment on bioequivalence study. The participants took a single dose of 10 mg (as one 10 mg capsule) of either bromopride as the reference or test formulations, here called capsule C1 and capsule C2. Study BE2 presented a similar protocol design as BE1, with the difference, that the dosage form was a 10 mg tablet, named tablet T1 and tablet T2. Study BE3 also showed a similar design and volunteers took 10 mg bromopride capsules, called capsule C3 and capsule C4, respectively. We have collected all data from each study. We used R to randomly select volunteers to obtain the pharmacokinetic profile for only one treatment for each subject. Therefore, inter-occasion variability was not evaluated. Studies, formulations and number of subjects are described in Table 1 2.2 Population pharmacokinetic analysis Individual plasma concentration profile data were analyzed using stochastic approximation expectation maximization algorithm (SAEM) for nonlinear mixed effects followed by importance sampling methods implemented on Monolix 2018R2 (Lixoft, France). The model selection was based on goodness-of-fit criteria and visual inspection of the diagnostics plots, comparisons between the relative standard deviations of parameters estimation, precision of the estimations and by changes in Bayesian information criteria (BIC) and log-likelihood (-2xLL). The model was qualified by visual predictive check and bootstrap. Graphical and statistical evaluation used during the model development were generated in R 3.5.1 (R Foundation for Statistical Computing, Austria) and Monolix.

2.3 Structural pharmacokinetic model A staged approach was used to develop the final population pharmacokinetic model. First, a stable base model was developed. The base model describes the bromopride plasma concentration–time data in healthy

volunteer without considering covariate effects. Second, a full covariate model was developed by incorporating the effect of all prespecified intrinsic and extrinsic covariate parameter relationships of interest into the base model. Lastly, the final model was developed from the full covariate model by backward elimination of all covariates that are not of potential clinical relevance. Single- and multiple-compartmental models were evaluated to describe the disposition of bromopride. We assumed first-order for all pharmacokinetic parameters other than absorption. Several absorption models with components describing delayed and variable absorption were tested as initial structural models. These included zero- or first-order models with lag time, simultaneous and sequential mixed zero- and first-order absorption (with or without one or more lag time), transit compartmental model alone or combined with zero or first order, as well double absorption site model (Godfrey et al., 2011; Lee et al., 2015; Lipka et al., 1995; Metsugi et al., 2008; Schuck et al., 2002; Zhou, 2003). The between subject-variability (BSV) was incorporated in the model, assuming a log-normal distribution for all parameters, except for absorbed fraction (Fr) that appears in all mixed absorption models, which was logit transformed, to limit its value between 0 and 1. An exponential model was used to describe BSV. For intraindividual variability (residual random error - ε) a combined proportional and additive model was assumed. A zero-centered mean and variance 1 were also assumed for the models’ variability. The basic structural model, which includes key parameters and does not incorporate any covariates, was selected based on goodness-of-fit plots, precision of estimates, and the BIC criteria within Monolix. 2.4 Covariate selection All demographic variables, some laboratory tests and dosage form presented in the dataset collected from ANVISA, could be evaluated as potential covariates during the pharmacokinetic model building. Body surface area, creatinine clearance and glomerular filtration rate were calculated according to Dubois & Dubois, Cockcroft-Gault and Levey formulae, respectively (Cockcroft and Gault, 1976; Du Bois and Du Bois, 1989; Levey et al., 2009). Each covariate was screened using either graphical and statistical

(Wald test) methods. For visual screening, a parameter versus covariate scatter plot was used for continuous variables and a boxplot was used for categorical covariates. From the basic structural model, the continuous covariates were incorporated as a power model as follows (Eq. 1): (

̂

)

(Eq.1)

where ppop is the population typical value, covj and côv represent individual covariate values and median covariate value for the population respectively, βcov is the covariate effect coefficient and ηj refers to BSV, whereas the categorical covariates gender and products (capsules C1-C4 or tablets T1-T2) were incorporated as an exponential model, such that (Eq. 2): (

)

( )

(Eq. 2)

The covariates were added one by one and were selected to be fixed on the structural model if their addition was able to cause a decrease of 3.84 (pvalue < 0.05) on log-likelihood ratio (LRT). Once all covariates were tested by univariate analysis, the significant ones were added to the model, and a stepwise backward elimination was fulfilled. Covariates were kept in the model if their elimination could result in increase of at least 6.63 (p-value < 0.01), of the LRT, as well as reductions in BSV (Hutmacher and Kowalski, 2015). Only the variables having potential physiological correlation to the parameters were included in the final model. 2.5 Model evaluation The selection of the final model was based on a series of model evaluation. For instance, models which presented the lowest value of BIC were accepted if the value of parameters uncertainty (relative standard error – RSE) for estimated parameters was lower than 25% and the RSE for random effects were lower than 35% (Ette et al., 1998). The shrinkage of parameters was also evaluated to avoid values higher than 20-30% (Savic and Karlsson, 2009). Goodness-of-fit (scatter plots of observed vs. predicted and weighted residuals) were used for single run-based diagnostics during model development. For the final model, the stability, robustness and the predictive performance were evaluated using bootstrapping and the visual predictive check (VPC).

The bootstrap procedure was conducted using the function confintmlx of Rsmlx package from R, a total of 1,000 bootstrap-resampled datasets were simulated. The model robustness was evaluated comparing the closeness between the final parameters estimated by the population model developed and the bootstrap's median and 95% confidence intervals (Cis, 5th and 95th percentiles). 2.6. Simulation The potential clinical impact of statistically significant covariates was evaluated by comparing the steady- state area under the curve (AUCss) and peak plasma concentration (Cmax) on steady-state values from 300 simulated profiles from virtual volunteers receiving 10 mg of bromopride three times a day (every 8 hours), using the final pharmacokinetic model. The simulated profiles were created using the simulx function, and the AUCss and Cmax were obtained using the exposure function, both available on mlxR package for R. These simulations of AUCss and Cmax were performed for each significant covariate, and then, normalized by the average of the reference population. Forest plots were created to compare the simulated parameter values for each candidate covariate to a reference population simulated using median and 95% CI covariate values. Since there is no publicly available information about bromopride efficacy pharmacokinetic target and safety, we had to make some assumptions considering that these are approved products. We assumed that an AUCss and Cmax equal or higher than the estimated for capsule C1 would be needed for efficacy. Meanwhile AUCss e Cmax lower than presented by tablet T1 would be considered safe. These choices were made based in the results from noncompartmental analysis (data not shown), where capsule C1 and tablet T1 presented respectively, the lowest (0.18 h mg l-1, 0.02 mg l-1) and highest (0.28 h mg l-1, 0.04 mg l-1) AUC0-inf and Cmax values. Since all products evaluated in our study were approved by ANVISA and could be considered safe and efficacious, we intended to further evaluate them using the population pharmacokinetics model and simulations. 3. Results

3.1 Dataset and volunteers characteristics The dataset included data from 89 volunteers from three different studies of bromopride oral solid dosage forms presented to ANVISA, as aforementioned. All subjects took a single oral dose of 10 mg of bromopride. Overall, a total of 1692 bromopride concentrations were used for pharmacokinetic model building. Typical bromopride plasma concentration-time curves for each product are shown in figure 1. We can notice that bromopride presents an atypical absorption profile, almost 70% of volunteers had multiple peaks or shouldering display of absorption. After the oral administration of bromopride, the time for absorption peak (tmax) values ranging 0.75-4 h. Demographic characteristics and laboratory tests of this population were summarized in table 2. All volunteers were healthy adults with a median age of 30 years and 51.69% were male, the body weight values ranging 48.8-91 kg, and all results of laboratory tests were within the established reference ranges. 3.2 Selection and validation of structural model This study developed a population pharmacokinetics model for the multiple peaks phenomenon observed after oral administration of bromopride in healthy volunteers. The model selection was based on visual inspection of diagnostic plots ,decreases on BIC values and other criteria, like RSE and shrinkage. Fitting a two-compartment model was not successful and resulted in a non-parsimonious fit with greater residual errors as compared to fitting with a one-compartment model. Hence, we decided to use a one-compartment model with linear elimination. Once many subjects showed atypical absorption profiles, represented by multiple or shouldering peaks, the following approaches were tested for absorption process: first- and zero- order with lag time, transit compartment, parallel first order, transit compartment combined with first and zero-order, sequential zero- and first-order, sequential first- and zero-order, double absorption site, simultaneous zero- and first-order with lag time. The number of parameters and the BIC values for each model are presented in table 3.

Before undertaking the fitting using the mixed absorption models fitting two constants (models M4 to M10), a preliminary fit was performed with a single absorption (models M1 to M3). This provided a benchmark against which the complex absorption models fitting could be compared. Transit compartment model (M3) is an example of typical absorption process, with first-order constant, but is considered more physiological than models with lag time (Savic et al., 2007). Sometimes it is misleading to determine the nature of the absorption process by typical models, since that oral drug absorption is a very complex process, which involves various interactions with a host of physicochemical and physiological variables. Mixed absorption models can provide flexibility to fit the atypical profile presented by bromopride (Figure 1). All models evaluated had good performance, but according to BIC values and fit improvements, the data were best described by M9 and M8. Among the complex models of absorption, the model M9 (double absorption site), presented the lower value of BIC and it would be the best one. However, this model is considered overparameterized, and could lead to poor accuracy of the estimated parameters (Godfrey et al., 2011). The residual standard error of model M9 for many parameters was very high. The second best model, based on BIC value, is the model M8, where the absorption process is explained by a mixed model with two constants. A fraction of the dose (Fr), is absorbed by a first-order (ka) mechanism and the absorption rate is proportional to the amount of the drug in gut, while the remaining fraction (1-Fr) is then absorbed, after a time interval (Tlag) in a rate constant for a period of time (Tk0), in other words, in a zero-order process. The sequential first- and zero-order absorption model is shown schematically in Figure 2. A pharmacokinetic model of bromopride following an intravenous administration was studied by Brodie et al. In this study we also find that the one-compartment with first-order elimination is the best model, as described by them, and we used it to evaluate various absorption models of bromopride following oral administration. As expected, the two analysis led to similar values of pharmacokinetic parameters. For instance, 1) volume of distribution (Vd) estimated by the sequential first- and zero-order absorption model (M8) was 230 l, very similar to that value reported by Brodie et al. 215±16 l and 2)

clearance (CL) values from our model was 46.8 l h-1 while Brodie et al. was 53.94 l h-1. The one-compartment model with an atypical absorption process, characterized by sequential first- and zero-order absorption was defined as the best structural pharmacokinetic model to explain the time-concentration profiles of bromopride. With this model the BSV for all parameters were estimated successfully, and the major inter subject variability was observed with the absorption parameters (ka, Tlag and Tk0). No random effects (Ƞs) had high shrinkage or low precision. Combined residual errors were estimated in the final model. These results were confirmed by the goodness-of-fit plot (Figure 3), which demonstrated the reasonable agreement amongst the observed and predicted plasma concentrations. No systematic trends were noticed in population and individual conditional weighted residuals (PWRES and IWRES) plotted against time or predictions, and values were centered at ~ 0 (supplemental material figure S1). Figure 3 showed that most of the data lie within ±2 units of the zero ordinate. In this study, twelve intrinsic and extrinsic factors, including demographic characteristics, laboratory measurements and bromopride products, could be evaluated as covariate. Given that the absorption parameters were the large source of variability and there is no physiological correlation between them and the laboratory measurements, only the demographic characteristics and products, were evaluated using Wald test to verify whether they could explain any parameter variability in the final model. At first moment we evaluated the six different bromopride products as a categorical covariate, but after some screening, we noticed that what had some impact in the model was the dosage form and not the product. Therefore, the dosage form (capsules or tablets) was assumed as a categorical covariate candidate. The covariate dosage form, i.e. capsule or tablet, could explain part of variability in all absorption parameters except on fraction of the dose absorbed by first-order (Fr). Although dosage form was identified as being significant for ka during the initial covariate evaluation, it was not retained in the final model because of statistically insignificant increases in LRT (LRT= 6.3, P<0.01) during stepwise backward elimination. Dosage affects Tk0 (LRT= 8.48, p=2.52e-3) and

Tlag (LRT= 31.5, p=5.67e-10), with negative βcov for tablets, which means that the dosage form tablet presents lower time duration of zero-order process and latency time. The inclusion of bodyweight positively affect Tk0 (LRT= 14.08, p=7.54e-9), as well as, the covariate gender significantly affects Tlag (LRT= 9.87, p=0.00922), with positive βcov for female, suggesting a higher latency time compared with male, the reference gender. The inclusion of dosage form and bodyweight reduces the BSV on Tk0 from 58.8% to 49.6%, while the inclusion of formulation and gender on Tlag explained 11.3% of variability. The estimated parameters by the population pharmacokinetic model with covariates addition were showed on table 4. The η-shrinkages for fixed effects were 0.31%, 3.87%, 11.9%, 4.9%, 3.17% and 2.19% for ka, Tk0, Fr, Tlag, V and CL respectively, all within the range considered ideal for informativeness. The uncertainty of fixed and random effects had acceptable precision, with RSE between 1.9 and 43.8% for parameters and amongst 7.58 and 21.8% for variability. The model and parameter estimates showed robustness in the bootstrap procedure. All parameter estimates from the final model were within the 95% bootstrap CI as presented on table 4. The predictive performance was also sufficient, according to the VPC result (Figure 4). 3.2 Clinical impact of significant covariates In order to assess the clinical impact of the significant covariates in the absorption process of bromopride, we simulated exposure on steady-state for specified populations. Firstly, a reference population was generated by setting some characteristic, such as male sex, dosage form capsule and the median value of bodyweight (67.2 kg). To measure the clinical relevance of gender, dosage form and bodyweight, their impacts on bromopride Cmax and AUCss were evaluated by simulation 300 virtual subjects in the following subpopulations: (1) gender female; (2) tablets as dosage form; and for the bodyweight groups (based on quartiles), (3) <60 kg; (4) 60-75 kg; (5) >75 kg. Forest plots of the impact of gender, dosage form and bodyweight on bromopride exposure were shown on figure 5. The plots suggesting that exposure was not impacted by any of the covariates, once the 95% CI of each

subpopulation overlapped considerably and the median of each scenario explored was close to that of reference. The evaluation of the impact of different solid dosage forms on bromopride exposure was based on a limited sample size, only 32.58% of volunteers in the dataset took tablets as dosage form, maybe additional data are needed to further assess whether was a true impact of dosage form on bromopride pharmacokinetics and the potential clinical implication for bromopride safety and efficacy. In addition, all the collected data was from healthy volunteers, rather than in the target patient population, which could present different pharmacokinetics and different covariate influence on the pharmacokinetics. Subjects who took the capsule formulation had an estimated 0.99-fold lower median AUCss and 1.1-fold higher median Cmax, when compared to those subjects that took the tablet formulation. Therefore, no significant difference on the exposure was observed between capsules and tablets. The exploratory comparison between the different bromopride products, assuming capsule C1 and tablet T1 as efficacy and safety endpoints, respectively, confirmed that all products involved in this population pharmacokinetics analysis could be considered safe and efficient. Using the final model to predict exposure at steady-state the values of AUCss ranged from 0.202 to 0.265 h mg l-1 and Cmax within 0.04 mg l-1 limit. Both higher than capsule C1 and lower than tablet T1 NCA results. 4. Discussion According to literature bromopride presents a monoexponential decay after an intravenous administration which is explained by the equation of onecompartment model (Brodie et al., 1986). Visual inspection of the semilogarithmic plasma concentration profile of the majority of volunteers enrolled in this study (data not show) demonstrate the same pattern. The absorption process of bromopride has never been described before. Furthermore, the plasma concentration-time profile of bromopride exhibits multiple-peaks, or at least a shouldering peak, after the oral administration of capsules or tablets. This phenomenon is not new in the pharmacokinetic field. Multiple-peaks phenomenon is described for many drugs including ranitidine

(Schuck et al., 2002), cimetidine (Oberle and Amidon, 1987), sorafenib (Jain and Woo, 2011), sumatriptan (Cosson and Fuseau, 1999; Lee et al., 2015), imatinib (Golabchifar et al., 2016) and others. This happen because oral absorption behavior is often complex, once various factors are involved, including mainly physicochemical and physiological ones. The multiple-peak phenomenon is a clear example of the complex kinetics involved in drug absorption after oral administration. Literature postulate a series of hypothesis to explain this phenomenon, where the enterohepatic recirculation (EHC) is the most common. Considering bromopride, as Brodie et al. do not observe a second peak in their intravenous analyses, we could discard the EHC hypothesis (Brodie et al., 1986; Godfrey et al., 2011; Zhou, 2003). Other possible reasons proposed by literature includes drug interaction with biliary salts, different sites of absorption, irregular gastric emptying, residence time of the drug in the stomach and along the bowel, physicochemical and biopharmaceutics properties of drugs (Holford et al., 1992; Metsugi et al., 2008). However, there is some criticism to accept all hypothesis, and according to Godfrey et al. the multiple-peaks phenomenon has been the subject of considerable interest. It is also important to take into account that the sample collection times could alter the observation of double-peaks (Oberle and Amidon, 1987). Bromopride is classified as class II (low solubility, high permeability) compound in biopharmaceutical classification system (BCS), and therefore the dissolution in gastrointestinal fluids have markedly impact on the oral absorption of this class of drugs (Takano et al., 2008). We can categorize the oral absorption of class II compounds into rate-limited and solubility-limited, based on their solubility, permeability and dissolution-rate. For absorption limited by solubility, the amount of drug absorbed reaches saturation, that could be better explained by zero-order kinetics (Holford et al., 1992; Takano et al., 2008). The solubility of bromopride is pH-dependent, increases in pH results in lower solubility, since this drug exhibits pKa value of 9.35 (Silva et al., 2015). Commonly basic drugs have a decrease in solubility transiting down the gastrointestinal tract (Zhou, 2003). As noted for imatinib (Peng et al., 2005) and sorafenib (Jain and Woo, 2011). The high functional specialization presented in each segment of the gastrointestinal tract may limit the absorption of drugs to

certain regions, named “windows of absorption”. This pH-dependent solubility could be considered as a window of drug input (Holford et al., 1992) and may explain the multiple-peak profile of drugs with this characteristic. Typical pharmacokinetic models for oral administration tends to assume either a first or a zero-order rate of absorption. We have developed a combined first- followed by zero-order model to describe absorption of bromopride after a single oral dose of 10 mg. This model is in accordance with physicochemical properties of bromopride, specially solubility, and with the physiological aspects of the intestinal absorption. In addition, the model has enough flexibility to explain the multiple-peaks profile observed. We hypothesize that bromopride stay soluble in the stomach and when it is delivered to intestine, where it solubility decrease, a part of this amount could precipitate or form micelles, and just a fraction of this amount (Fr) is readily absorbed by a first-order process (under the assumption that the rate of absorption is limited by bromopride solubility in the gut fluid and the volume of this fluid is assumed to be constant). The remaining fraction (1-Fr) of bromopride would just be absorbed after a short time (Tlag) by a constant rate (zero-order absorption) until the solubility is reached again, as discussed by Holford et al. for cefetamet pivoxil. Previous studies have discussed the closer relation between low bioavailability and high inter-individual variability (Hellriegel et al., 1996). The absolute bioavailability (BA) of bromopride capsules is 50% approximately (Brodie et al., 1986) so a large BSV could be expected for this drug. Unfortunately, intravenous data were not available in our database, so we were unable to estimate BA. However, higher variability in drug exposure can be noticed, especially in the absorption parameters. As far as we know, this is the first population model describing the population pharmacokinetics of bromopride characterizing also the absorption process. The clearance was estimated as 46.80±2.03 l h-1. This is in accordance with the findings of Brodie et al. who had reported 53.94 l h-1, and as they discuss, this is a high value for systemic clearance and indicate a probable participation of hepatic metabolism on elimination process (Brodie et al., 1986). Corroborating with that hypothesis, Dunne et al, found at least twenty metabolites of bromopride using in vitro hepatocytes assay (Dunne et al., 2013).

The Vd reported by Brodie et al. is 215 l, also similar to the Vd estimated by our population pharmacokinetics model (230±10 l). These relatively large Vd according to Brodie et al. implies an extensive tissues penetration. Furthermore, less than 40% of the bromopride present in plasma are bound to protein (Brodie et al., 1986). As the volunteers enrolled in these bioequivalence studies were required to be healthy, it was expected that the laboratory measurements would not result on significant effects. The biggest source of variability were the absorption parameters, which could be correlated with demographics characteristics. Dosage form had a significant effect on Tlag and Tk0. Given that bromopride is a BCS class II and that, oral absorption is closely related to the capacity of solid dosage forms dissolve, this correlation was expected. As gastric acid secretion is lower in female sex than in male, this physiological change can alter the absorption of drugs (Zhou, 2003). Especially whether this drug had pH-solubility dependent as bromopride. As we can observe the covariate gender effects Tlag, exactly the parameter that indicates the time for zero-order absorption rate starts. We also can observe the effect of bodyweight on Tk0. The final population pharmacokinetic model allowed the characterization of the individual plasma concentration-time curve for bromopride and enabled subsequent simulation of the effects of different characteristics of the population. The clinical impact of the observed covariate effects was assessed through simulations of covariate scenarios compared to a reference. Gender, dosage form and ranges of bodyweight related to absorption parameters had negligent effects on bromopride exposure on steady-state. Consequently, it suggests that no dose adjustment is required in this subpopulation. Many studies with other drugs also demonstrate that the covariates had minimal effects on exposure at steady-state (Sy et al., 2018; Zhang et al., 2015). Once there is no data about bromopride efficacy or safety available on literature and considering that all products evaluated by our model were approved by ANVISA, we just compared the exposure within the six products at steady-state to assess whether they were efficient and safe. Additional data are required to

adequately assess the true magnitude of the impact of dosage form on bromopride safety and efficacy. 5. Conclusion This is the first population model describing bromopride pharmacokinetics based on its known disposition characteristics such as solubility. The parameters predicted by the model were in agreement with the physicochemical properties of bromopride and with the physiology of intestinal absorption process. The model allowed us to identify some important covariates related to the variability in this atypical absorption profile. In addition, the population pharmacokinetic model could also be used to simulate the exposure profile at steady-state for different categories and identify that no covariates were deemed to have clinical impacts. Acknowledgements Araucária Foundation (number 029/17 prot. 48027) and Paraná State Government for grants. Lachi-Silva L. thanks the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001 for the scholarship. Barth, A. B. thanks CNPq for PhD scholarship. References ANVISA, 2019. Brazilian Health Survaillance Agency Product verification [WWW Document]. Consult. - Agência Nac. Vigilância Sanitária. URL https://consultas.anvisa.gov.br/#/medicamentos/q/?substancia=1455&situa caoRegistro=V (accessed 2.19.19). Barbieri, D., Hames, R.L., 2011. Bromoprida e suas aplicações em pediatria. Pediatr. Mod. 47, 84–88. Barreira, E.R., Magaldi, R.B., 2009. Distonia aguda relacionada ao uso de bromoprida em pacientes pediátricos. Rev. Paul. Pediatr. 27, 110–114. https://doi.org/10.1590/S0103-05822009000100017 Brodie, R.R., Chasseaud, L.F., Rooney, L., Taylor, T., 1986. Pharmacokinetics and Bioavailability of the Anti-Emetic Agent Bromopride. Biopharm. e Drug Dispos. 7, 215–222. Cockcroft, D.W., Gault, H., 1976. Prediction of Creatinine Clearance from Serum Creatinine. Nephron 16, 31–41. https://doi.org/10.1159/000180580 Cosson, V.F., Fuseau, E., 1999. Mixed effect modeling of sumatriptan pharmacokinetics during drug development: II. From healthy subjects to phase 2 dose ranging in patients. J. Pharmacokinet. Biopharm. 27, 149– 171. https://doi.org/10.1023/A:1020601906027

Du Bois, D., Du Bois, E.F., 1989. A formula to estimate the approximate surface area if height and weight be known. 1916. Nutrition 5, 303–11; discussion 312-3. Dunne, C.E., Bushee, J.L., Argikar, U.A., 2013. Metabolism of bromopride in mouse, rat, rabbit, dog, monkey, and human hepatocytes. Drug Metab Pharmacokinet. 28, 453–461. https://doi.org/10.2133/dmpk.DMPK-13-RG010 Ette, E.I., Sun, H., Ludden, T.M., 1998. Balanced Designs in Longitudinal Population Pharmacokinetic Studies. J. Clin. Pharmacol. 38, 417–423. https://doi.org/10.1002/j.1552-4604.1998.tb04446.x Glessner, M.R., Heller, D.A., 2002. Changes in related drug class utilization after market withdrawal of cisapride. Am. J. Manag. Care 8, 243–50. Godfrey, K.R., Arundel, P.A., Dong, Z., Bryant, R., 2011. Modelling the Double Peak Phenomenon in pharmacokinetics. Comput. Methods Programs Biomed. 104, 62–69. https://doi.org/10.1016/j.cmpb.2010.03.007 Golabchifar, A.-A., Rezaee, S., Dinan, N.M., Kebriaeezadeh, A., Rouini, M.-R., 2016. Population Pharmacokinetic Analysis of the Oral Absorption Process and Explaining Intra-Subject Variability in Plasma Exposures of Imatinib in Healthy Volunteers. Eur. J. Drug Metab. Pharmacokinet. 41, 527–539. https://doi.org/10.1007/s13318-015-0292-3 Hellriegel, E.T., Bjornsson, T.D., Hauck, W.W., 1996. Interpatient variability in bioavailability is related to the extent of absorption: Implications for bioavailability and bioequivalence studies. Clin. Pharmacol. Ther. 60, 601– 607. https://doi.org/10.1016/S0009-9236(96)90208-8 Holford, N.H.G., Ambros, R.J., Stoeckel, K., 1992. Models for describing absorption rate and estimating extent of bioavailability: Application to cefetamet pivoxil. J. Pharmacokinet. Biopharm. 20, 421–442. https://doi.org/10.1007/BF01061464 Hutmacher, M.M., Kowalski, K.G., 2015. Covariate selection in pharmacometric analyses: a review of methods. Br. J. Clin. Pharmacol. 79, 132–147. https://doi.org/10.1111/bcp.12451 Jain, L., Woo, S., 2011. Population pharmacokinetic analysis of sorafenib in patients with solid tumours. Br. J. Clin. Pharmacol. 72, 294–305. https://doi.org/10.1111/j.1365-2125.2011.03963.x Lee, J., Lim, M., Seong, S.J., Park, S., Gwon, M., Han, S., Lee, S.M., Kim, W., Yoon, Y., Yoo, H., 2015. multiple peaks phenomenon in sumatriptan 23, 66–74. Levey, A.S., Stevens, L.A., Schmid, C.H., Zhang, Y.L., Castro, A.F., Feldman, H.I., Kusek, J.W., Eggers, P., Van Lente, F., Greene, T., Coresh, J., 2009. A New Equation to Estimate Glomerular Filtration Rate. Ann. Intern. Med. 150, 604. https://doi.org/10.7326/0003-4819-150-9-200905050-00006 Lipka, E., Langguth, P., Spahn-, H., Mutschler, E., Amidon, G.L., 1995. Celiprolol Double-Peak Occurrence and Gastric Motility: Nonlinear Mixed Effects Modeling of Bioavailability Data Obtained in Dogs 23, 267–286. Metsugi, Y., Miyaji, Y., Ogawara, K.I., Higaki, K., Kimura, T., 2008. Appearance of double peaks in plasma concentration-time profile after oral administration depends on gastric emptying profile and weight function. Pharm. Res. 25, 886–895. https://doi.org/10.1007/s11095-007-9469-z Oberle, R.L., Amidon, G.L., 1987. The influence of variable gastric emptying and intestinal transit rates on the plasma level curve of cimetidine; an

explanation for the double peak phenomenon. J. Pharmacokinet. Biopharm. 15, 529–544. https://doi.org/10.1007/BF01061761 Peng, B., Lloyd, P., Schran, H., 2005. Clinical Pharmacokinetics of Imatinib. Clin. Pharmacokinet. 44, 879–894. https://doi.org/10.2165/00003088200544090-00001 Roila, F., Minotti, V., Ballatori, E., Basurto, C., Tonato, M., 1985. Evaluation of the antiemetic activity of bromopride in cancer patients treated with i.v. CMF. Tumori 71, 455–8. Savic, R.M., Jonker, D.M., Kerbusch, T., Karlsson, M.O., 2007. Implementation of a transit compartment model for describing drug absorption in pharmacokinetic studies. J. Pharmacokinet. Pharmacodyn. 34, 711–726. https://doi.org/10.1007/s10928-007-9066-0 Savic, R.M., Karlsson, M.O., 2009. Importance of Shrinkage in Empirical Bayes Estimates for Diagnostics: Problems and Solutions. AAPS J. 11, 558–569. https://doi.org/10.1208/s12248-009-9133-0 Schuck, V.J.A., Costa, T.D., Barros, S.G.S. De, Gruber, C., Schapoval, E.E.S., 2002. Compartmental analysis of ranitidine doubled peak plasma profile after oral administration to healthy volunteers. Rev. Bras. Ciência do Solo 38. https://doi.org/10.1590/S1516-93322002000200007 Silva, M.G.R., Volpato, N.M., Pinto, E.C., Cabral, L.M., de Sousa, V.P., 2015. Development of a Dissolution Test for Extended-Release Bromopride Pellets with In Vivo–In Vitro Correlation. Dissolution Technol. 22, 24–33. https://doi.org/10.14227/DT220215P24 Sy, S.K.B., Chia, Y.L., Gordi, T., Hoch, U., Eldon, M.A., 2018. Integrated population pharmacokinetics of etirinotecan pegol and its four metabolites in cancer patients with solid tumors. Cancer Chemother. Pharmacol. 81, 897–909. https://doi.org/10.1007/s00280-018-3562-3 Takano, R., Furumoto, K., Shiraki, K., Takata, N., Hayashi, Y., Aso, Y., Yamashita, S., 2008. Rate-limiting steps of oral absorption for poorly watersoluble drugs in dogs; prediction from a miniscale dissolution test and a physiologically-based computer simulation. Pharm. Res. 25, 2334–2344. https://doi.org/10.1007/s11095-008-9637-9 Zhang, J., Hayes, S., Sadler, B.M., Minto, I., Brandt, J., Piscitelli, S., Min, S., Song, I.H., 2015. Population pharmacokinetics of dolutegravir in HIVinfected treatment-naive patients. Br. J. Clin. Pharmacol. 80, 502–514. https://doi.org/10.1111/bcp.12639 Zhou, H., 2003. Pharmacokinetic strategies in deciphering atypical drug absorption profiles. J. Clin. Pharmacol. 43, 211–227. https://doi.org/10.1177/0091270002250613

Graphical abstract

Table 1. Overview of bioequivalence studies included in bromopride population pharmacokinetic analysis. Study (total of Number of volunteers Formulation volunteers) allocated per formulation Capsule (C1) 15 BE1 (30) Capsule (C2) 15 Tablet (T1) 17 BE2 (30) Tablet (T2) 12 Capsule (C3) 15 BE3 (29) Capsule (C4) 15

Table 2. Characteristics of healthy volunteers whose data were included in the development of the pharmacokinetic model Characteristics, (n=89)

Volunteer demographic Age (years) Weight (kg) Height (m) Body mass index (kg/m2) Body surface area (m2) Gender Male (%) Female (%) Laboratory measurements Protein (g/dl) Bilirubin (mg/ml) AST (U/l) ALT (U/l) Creatinine clearance (ml/min) Glomerular filtration rate (ml/min)

Median [range] or % of volunteers (n)

30 [18, 50] 67.20 [48.80, 91] 1.67 [1.50, 1.86] 24.93 [18.91, 29.71] 1.77 [1.42, 2.07] 51.69 (46) 48.31 (43) 7.10 [6.10, 8.0] 0.65 [0.23, 2.72] 17 [11, 34] 16 [4, 46] 103.42 [59.76, 169.01] 90.85 [50.64, 154.28]

n: number of volunteers, AST: aspartate aminotransferase, ALT: alanine aminotransferase

Table 3. Structural absorption models evaluated for oral solid dosage forms of bromopride. Absorption process (number of parameters to Model BIC be estimated) M1 First-order with a lag time (4) -15187.00 M2 Zero-order with a lag time (4) -15539.94 M3 Transit compartment (5) -15580.88 M4 Parallel first-order (6) -15604.99 M5 Combined transit compartment and first-order (8) -15683.32 M6 Combined transit compartment and zero-order (8) -15484.87 M7 Sequential zero- and first-order (5) -15730.94 M8 Sequential first- and zero-order (6) -15766.31 M9 Double absorption site (8) -15824.06 M10 Simultaneous zero- and first-order with lag time (7) -15704.88 BIC: Bayesian information criteria

Table 4. Final model population parameters and bootstrap result Estimate %BSV Bootstrap estimates* Parameter (units) (%RSE) (%RSE) Median 95% CI -1 ka (h ) 0.08 (9.45) 49.40 (13.40) 0.08 0.06 – 0.10 Tk0 (h) 0.88 (6.92) 49.60 (8.38) 0.89 0.07 – 1.01 Fr (%) 32.60 (1.9) 36.50 (21.80) 31.00 26.00 – 36.00 Tlag (h) 0.47 (6.88) 40.10 (7.90) 0.46 0.42 – 0.52 Vd (l) 230.00 (4.67) 41.90 (8.22) 232.00 207.50 – 257.40 CL (l h-1) 46.80 (4.33) 40.20 (7.58) 46.89 43.25 – 50.74 βFORMonTk0 -0.27 (43.8) -0.29 -0.48 – -0.10 βWTonTk0 1.58 (25.3) 1.54 0.85 – 2.19 βFORMonTlag -0.58 (16.1) -0.56 -0.71 – -0.40 βSEXonTlag 0.23 (38.4) 0.23 0.07 – 0.38 Error model parameters a 3.72e-4 (2.3e-5) 3.75e-4 3.09e-4 – 4.46 e-4 b 0.11 (2.7e-3) 0.11 0.11 – 0.120 Log likelihood estimation -2xLL -15927.85 BIC -15847.06 %RSE: percent relative standard error of the estimate, BSV: between subject variability, 95% CI: 95% confidence interval on the parameter, ka: first-order absorption rate, Tk0: duration of the zeroorder absorption, Fr: fraction of dose absorbed by first-order, Tlag: latency time, Vd: volume of distribution, CL: apparent clearance, a: additive component of the error model, b: proportional component of the error model. The reference population for pharmacokinetic parameters are 67.6 kg male and the dosage form are capsule. β is the covariate effect coefficient. Reference groups for categorial covariates form and sex were capsule and male, respectively. * From 1000 completed bootstrap runs.

Figure 1.

Individual plasma concentration vs. time plots of six bromopride products involved in this study. The bold black line is the mean value. Bromopride shows atypical absorption profile with multiple peaks or shouldering. C1-C4: capsule formulations, T1-T2: tablet formulations.

Figure 2. Scheme of the one-compartment open model with sequential firstzero order of absorption. Where ka: absorption rate constant, Fr: fraction of the dose absorbed by a first-order process, Tk0: duration of the zero-order process, (1-Fr): remaining fraction of the dose absorber by zero-order mechanism, Tlag: time for the second fraction of the drug to begin to be absorbed.

Figure 3. Goodness of fit plots for the final model: scatter plots of the population-weighted residuals (PWRES) vs. time (upper left panel) and population model predictions (upper right panel), scatter plots of the individualweighted residuals (IWRES) vs. time (middle left panel) and individual model predictions (middle right panel), and scatter plots of the observed plasma concentration of bromopride vs. the populational model predictions (lower left panel) and the individual model predictions (lower right panel).

Figure 4. Visual predictive check plot of the final model after a single oral administration of 10 mg bromopride, 95% prediction interval. The 10th, 50th and 90th empirical percentiles of the observed data are also demonstrated. A total of 1000 datasets were simulated using the final pharmacokinetic parameter estimates.

Figure 5. Predicted fold changes in steady-state exposure (AUCss on the top panel and Cmax on the bottom panel) of bromopride (BRO) over 10 mg oral doses three times a day (every 8 hours) to reference covariate categories (fold change in median and 95% confidence interval. The reference population categories: male, dosage form capsule and bodyweight median 67.2 kg. The bodyweight used in each category was based on quartiles of the original dataset. AUCss: steady-state area under the plasma concentration-time curve, Cmax: maximum plasma concentration on steady-state.