*3.5. Population Pharmacokinetic Model Development*

By a one-compartment disposition model with first order elimination and absorption with an absorption lag time, the plasma concentrations of tiropramide were best expressed. Considering the lag time, the numerical values (such as -2LL and AIC) of the model evaluation and graphical data fitting were more improved than otherwise. Although we tried a two-compartment model, it did not show an improved fit when compared with the one-compartment model. In other words, there was a problem (fail to model fit) in fitting the two-compartment model for some individuals, but in the one-compartment model, all individuals were fitted properly. In addition, the -2LL and AIC values significantly decreased in the multiple (two) transit absorption compartment model in which the additional absorption phase was added to the one-compartment model. However, from

two or more transit phases of absorption, there was no significant decrease in the -2LL and AIC values (with increasing number of parameters). As a result, the multiple (two) transit absorption phase-one compartment model with an absorption lag time was selected as the base structure model for tiropramide. The model was parameterized in terms of V/F, Tlag, CL/F, Ka1, and Ka2. When the lag-time or model transit time was given to Ka2 in the parameterization of the model, no significant model improvement was found with the increasing number of parameters. Therefore, significant model improvement was found when parameterization with lag-time was carried out only at Ka1, and the rate constant of substance transit in the absorption compartment. Figure 2 shows a schematic of the disposition compartment model presented for tiropramide.

**Figure 2.** Schematic of tiropramide population pharmacokinetic (PPK) model (two transit absorption phase-one compartment model with an absorption lag time). Depot 1 and 2 represent transit compartments in the absorption phase. Ka1 refers to the rate constant at which the drug is moved from depot 1 to depot 2. Tlag refers to the delay time for the drug to move from depot 1 to depot 2. Ka2 is the rate constant at which the drug moves from depot 2 to the central compartment. V means the volume of drug distribution in the central compartment, and CL means removal of the drug from the central compartment.

An exponential model was used to describe the IIVs on parameters of Tlag, V/F, CL/F, Ka1, and Ka2. By applying an additive error model on log-transformed data, the residual variability was explained. Table 4 summarizes the steps for developing a basic structural model of tiropramide.

In order to find the covariates affecting the PK parameters of tiropramide, we analyzed the effects of each covariate on the PK parameters. The final potential covariates were selected on the basis of the graphical exploration between candidate covariates and PK parameters. The influence of each selected candidate covariate on the PK parameters of tiropramide was assessed by incorporating the covariates into an established basic structural model. The evaluation was based on OFV, which means model improvement. In this regard, the covariate selection process (according to OFV) to be reflected in the final model of tiropramide is summarized in Table 5.


**Table 4.** Basic structural model building steps.

\* Selected model.

**Table 5.** Stepwise search for covariates. −


\* Selected final model.

There was a significant correlation between the total protein and tiropramide V/F as well as the total protein and tiropramide CL/F. Figure 3 shows the correlation between the final selected covariates and CL/F of tiropramide.

**Figure 3.** Relationship between subjects' characteristics and individual predicted pharmacokinetic parameters. Clearance (CL/F) of tiropramide according to total protein (**A**). Volume of distribution (V/F) of tiropramide according to total protein (**B**).

Δ Δ When the correlation between the total protein and CL/F was reflected in the PPK model of tiropramide, the ∆OFV was significantly reduced to −9.25 (*p* < 0.05). Furthermore, the addition of total protein and V/F correlation to tiropramide PPK model significantly reduced the ∆OFV to −9.16

*ŋ ŋ ŋ* *ŋ ŋ* (*p* < 0.05). However, other covariates including BMI, *PEPT1* 1287G>C, *OCT2* 808G>T, *ABCB1* 1236C>T, *ABCB1* 2677G>T/A, *ABCB1* 3435C>T, and *CYP2D6* (\*1 and \*10) had no significant effect on model improvement. Even by applying genetic factors (such as *PEPT1* 1287G>C, *OCT2* 808G>T, *ABCB1* 1236C>T, *ABCB1* 2677G>T/A, *ABCB1* 3435C>T, and *CYP2D6* (\*1 and \*10)) alone to the base model, we examined whether they affected Ka1, Ka2, CL/F, and V/F, but no significant associations were identified. The final model of tiropramide (reflecting the effects of covariates) is expressed as follows:

$$\begin{aligned} \text{V/F} &= \text{tvV/F} \cdot (1 + (\text{Totalprotein-7.6}) \cdot \text{dV/F} \text{dT} \text{total} \text{protein}) \cdot \exp(\eta\_{\text{V}})\\ \text{CL/F} &= \text{tvCL/F} \cdot (1 + (\text{Totalprotein-7.6}) \cdot \text{dL1/F} \text{dT} \text{Totalprotein}) \cdot \exp(\eta\_{\text{CL}})\\ &\qquad \text{T}\_{\text{lag}} = \text{tvT}\_{\text{lag}} \cdot \exp(\eta\_{\text{Tlag}})\\ &\qquad \text{K}\_{\text{a1}} = \text{tvK}\_{\text{a1}} \cdot \exp(\eta\_{\text{Ka1}})\\ &\qquad \text{K}\_{\text{a2}} = \text{tvK}\_{\text{a2}} \cdot \exp(\eta\_{\text{Ka2}}) \end{aligned} \tag{1}$$

Population estimates of tiropramide were 466,711 mL/h for CL/F and 1,889,250 mL for V/F. CL/F and V/F values by NCA were 482,567 ± 267,433 mL/h and 2,386,871 ± 1,699,847 mL, respectively. As a result, the CL/F and V/F values estimated in the final model were not significantly different from the NCA values. In the final model, the relative standard error (RSE, %) was 9.53–83.51%. The Eta shrinkage values for the estimated PK parameters were suggested as acceptable at 0.02–0.40%. Compared with the base model, the final model (considering total protein effects) of tiropramide reduced the IIV of V/F from 70.70% to 57.12%, and the IIV of CL/F from 50.55% to 39.95%. Table 6 presents the estimated parameter values in the base model and final PPK model of tiropramide. The AUC0–t, AUC0–∞, Cmax, half-life, and Tmax estimated values by the final PPK model of tiropramide were 242.73 ± 175.28 h· ng/mL, 260.92 ± 179.08 h· ng/mL, 54.89 ± 46.18 ng/mL, 2.73 ± 0.88 h, and 1.65 ± 0.53 h, respectively.



#### *3.6. Population Pharmacokinetic Model Evaluation*

The developed PPK model of the tiropramide was comprehensively evaluated for GOF, bootstrap analysis, and VPC. Figure 4 shows the GOF plots of the base and the final models of tiropramide. As shown in Figure 4B, the observed and predicted concentrations of tiropramide showed a relatively good agreement in the final model. CWRES was well distributed symmetrically with respect to zero, and CWRES was included in ±4 at all points. In addition, the residuals in the final model were more improved than in the base model. In other words, without any specific bias, the CWRES was randomly well distributed, and the residuals in the final model showed a significant decrease when compared with the base model (larger than ±4).

**Figure 4.** *Cont*.

**Figure 4.** Goodness-of-fit (GOF) plots of base model (**A**) and final model (**B**) for tiropramide. (**a**) Population-predicted concentrations (PRED) against observed plasma concentration (DV), (**b**) individual-predicted concentrations (IPRED) against DV, (**c**) PRED against conditional weighted residuals (CWRES), (**d**) time (IVAR) against CWRES, and (**e**) quantile–quantile plot of components of CWRES.

Bootstrap validation was performed to verify the reproducibility and/or robustness of the final PPK model of tiropramide. Table 7 shows the bootstrapping analysis results. The parameter values estimated in the final model were in the 95% CI range of the bootstrap analysis results, and were similar to the median values of the bootstrap (replicates of 1000).


**Table 7.** Estimated population pharmacokinetic parameter values of tiropramide and bootstrap validation (*n* = 1000).

The VPC simulation results of the final PPK model of tiropramide are presented in Figure 5. Most of the observation values of tiropramide were well distributed within the 90% prediction interval of the prediction values. As a result, this suggests that the final model of the tiropramide is precise and explains the data well.

**Figure 5.** Visual predictive check (VPC) of the final model for tiropramide. Observed concentrations were depicted by the dots. The 95th, 50th, and 5th percentiles of the predicted concentrations are represented by black dashed lines. The 95% confidence intervals (CI) for the predicted 5th and 95th percentiles are represented by the blue shaded regions. The 95% CI for the predicted 50th percentiles are represented by the red shaded regions. The values on the *y*-axis are logarithms.

The NPDE distribution and histogram are presented in Figure S1. The assumption of a normal distribution for the differences between predictions and observations was acceptable. The quantile–quantile plots and histogram also confirmed the normality of the NPDE (Figure S1).

#### **4. Discussion**

The mechanism of action of tiropramide, as mentioned in the Introduction section, has been studied relatively well and has been reported in the past. However, studies on the in vivo PK characteristics (including metabolism and excretion) of tiropramide are still insufficient. Therefore, little data was available regarding the dosage and usage of tiropramide in clinical as well as formulation development. According to Lee et al., despite the frequent use of tiropramide in clinical practice, studies on safety and efficacy are very poor [4]. We conducted a PPK model development study of tiropramide to explore the effective covariates related to PK diversity of tiropramide and to investigate the characteristics of PK in the population. This study was new and was expected to be useful in the evaluation of the safety and efficacy of tiropramide in clinical use. As mentioned in the Abstract section, although tiropramide has a (relatively) broad margin of safety, this study involving healthy subjects was very important because it could find new covariates in healthy subjects that had not been reported before and/or be used to predict PPK for patients in the clinic by establishing PPK in healthy adults. In addition to this, in patients with abdominal pain and irritable bowel syndrome, it is very likely that the absorption process of tiropramide will change. Therefore, if clinical trials are conducted for patients in the future, it is thought that the PK variation of tiropramide between individuals can be explained more specifically through the application of our PPK model.

In this study, the PK of tiropramide was modelled as a two transit absorption phase-one compartment model with an absorption lag time. Various errors (including residual error and IIV) models and covariate effects were evaluated to establish factors that significantly influence the PK parameters of tiropramide and to explain the PK diversity of the tiropramide in the population. As a result of evaluating the model, the final tiropramide PPK model showed relatively good GOF plots, suggesting that the final PPK model had an acceptable predictive power. In addition, all CWRES values over time or predicted concentrations were in the range of -4 and 4, suggesting that the model is relatively stable. In addition, the bootstrap and VPC simulation results suggested that the final tiropramide PPK model was accurate, stable, and precise. We compared the estimated parameter values (of AUC0–t, AUC0–∞, Cmax, half-life, and Tmax) by tiropramide's final PPK model with these values by NCA analysis. As a result, there were no significant differences (*p* > 0.05) between the parameter values predicted by the final PPK model of tiropramide and those calculated by NCA analysis. These results suggest that the final PPK model of tiropramide established in this study explains the experimental data relatively well.

The 24 healthy Korean male PK data used to establish the tiropramide PPK model were similar to the previously reported PK results. In other words, the previously reported PK parameter values of tiropramide were similar to our PK results obtained by NCA analysis. After oral administration of 100 mg of tiropramide in humans, the obtained NCA PK parameters (as previously reported values) were 2.34-6.99 h for t1/2, 0.66-1.6 h for Tmax, 77.4-111 ng/mL for Cmax, and 267.7-812.7 h·ng/mL for AUC [6,12,18–21]. On the other hand, the NCA PK parameters in this study were 3.41 ± 1.99 h for t1/2, 1.74 ± 0.63 h for Tmax, 69.07 ± 59.74 ng/mL for Cmax, and 280.34 ± 199.96 h·ng/mL for AUC, which were similar to the previously reported values. Table 8 summarizes these results.

As mentioned above, few studies have been done on the metabolic ratio (including pathway) and excretion of tiropramide in humans (especially patient groups), making it difficult to predict candidate covariates. On the basis of previous reports (as drug information provided by the manufacturer) that tiropramide is metabolized in the liver and excreted into urine (about 10–20% of administered dose), the covariate effects were tested in this study by obtaining the physicochemical information of liver function indicators (such as AST, ALT, and ALP) and genotyping the *CYP2D6* gene related to metabolism in the body. The creatinine clearance and the functional indicator of the kidney were collected for each subject, and the covariate effects related to CL/F were tested. In addition, genotyping of genes (such as *ABCB1*, *OCT2*, and *PEPT1*) associated with various transporters known to be widely involved in the distribution, absorption, excretion, and metabolism of drugs in the body has been performed to identify the effects of the covariate associated with the PK parameters. Despite these efforts, only total protein was found to have a significant effect on V/F and CL/F of tiropramide. As shown in Figure 3, the total protein, and the V/F and CL/F showed a significant negative correlation of 45.72% (*r* <sup>2</sup> = 0.209) and 45.28% (*r* <sup>2</sup> = 0.205), respectively. The correlation values of the total protein to CL/F and V/F were the largest of all the covariates we collected in this study.


**Table 8.** Previously reported pharmacokinetic (PK) parameter values of tiropramide obtained by non-compartmental analysis (NCA) analysis.

Although the plasma protein binding ratio of tiropramide in humans has not been reported accurately, it has been reported that the plasma protein binding ratio of tiropramide in rats is about 48–51% [22]. This suggests that the amount of plasma protein may affect the in vivo PK properties of tiropramide by binding to the plasma proteins in the blood. According to our tiropramide PPK model, higher total protein levels in the blood mean a smaller distribution of tiropramide in the body and a lower excretion. This can be explained by the fact that tiropramide binds to proteins in the blood and affects the distribution and excretion of substances from the body. Tiropramide combined with proteins in the blood will make it difficult to filter the glomerulus of the kidney and distribute from blood to many organs. The reflection of the total protein covariates in tiropramide PPK model reduced V/F IIV and CL/F IIV by 13.58% and 10.60%, respectively. These results suggested that the variabilities of tiropramide plasma concentrations could be partly explained by individual variances of total protein level related with V/F and CL/F of tiropramide. On the other hand, other candidate covariates (such as AST, ALT, ALP, and creatinine clearance) had no significant effect on the PK parameter values and IIV improvement of tiropramide. Although tiropramide is metabolized in the liver and excreted in the kidney, our results suggest that tiropramide is a drug that does not require dose control depending on the liver and renal function. However, because our PPK model was based on the data from healthy men, further studies (for patient groups) will be needed for further clarification. That is, if PK data significantly different from the normal group (like our study) was obtained from the patient groups, and if the PPK analysis was conducted in the same manner as in this study with our model (from patient groups), other significant covariates may be identified. Therefore, PK studies or PPK analysis of tiropramide in patient groups will need to be performed in the future. Nevertheless, this study was important because it is a PPK model study of tiropramide that has not been previously conducted, and other related studies (such as clinical dose setting, formulation development, and PK comparison with certain other groups) on tiropramide may be possible in the future, on the basis of our findings. In addition, another limitation of our study was that PK analysis and PPK model studies were conducted for limited ages (between 19 and 29 years old). In the future, PK analysis and/or PPK model studies of tiropramide for more diverse age groups will need to be conducted in this regard.

As shown in Table 6, unexplained variability (as IIV) still exists in K<sup>a</sup> (74.63% for Ka1 and 74.57% for Ka2) and Tlag (32.50%). These results meant that multi-complex gastrointestinal (GI) tract absorption processes of tiropramide including the variabilities of each individual gastric emptying time, GI tract transit time, and other transporters, among others, could considerably affect the variabilities of tiropramide plasma concentrations. This study could not find any significant covariates that could explain K<sup>a</sup> IIV and Tlag IIV. Perhaps, despite collecting various genetic and demographic information, there was still not enough information to explain the K<sup>a</sup> and Tlag IIVs of tiropramide. Therefore, further studies are needed to explore significant covariates related to the absorption of tiropramide in the body.
