Population pharmacokinetic analysis of the multiple peaks phenomenon in sumatriptan

Sumatriptan, a selective agonist acting on vascular 5-hydroxytryptamine (5-HT1B/1D) receptors, has been found to be a safe and effective treatment for migraine attacks.[1, 2] It is currently marketed as oral, subcutaneous, intranasal, and suppository (limited distribution) formulations.[3]

After oral administration, sumatriptan is rapidly absorbed and eliminated with a half-life of about 2 h.[3, 4, 5] The median C_max (the maximum plasma drug concentration) following oral dosing with 50 mg of sumatriptan is 29 ng/mL and the t_max (the time to reach maximum plasma concentration) generally ranges from 0.5 to 3 h after administration of 25-, 50-, and 100-mg oral doses.[3] The bioavailability is 100% for the subcutaneous route and 14% for the oral route.[4] This lower bioavailability following oral administration is primarily due to pre-systemic metabolism in the gut wall and in the liver, and partly due to incomplete absorption.[2] It is metabolized by monoamine oxidase (MAO), predominantly the MAO-A isoenzyme, and inhibitors of that enzyme may alter sumatriptan pharmacokinetics (PK) and increase systemic exposure.[3] Sumatriptan and its metabolites are largely renally excreted (about ~60%), with about 40% found in the feces; only 3% of an oral dose can be recovered as unchanged sumatriptan.[3]

Sumatriptan frequently displays a particular absorption profile with multiple peaks in the plasma concentration.[2, 4] These multiple peaks produce considerable between-subject variability (BSV) in the t_max (0.5-5 h) and influence clinical response by altering the rate and extent of absorption in the early absorption phase following oral administration of sumatriptan.[1, 4, 6] The BSV in plasma concentrations is much greater with the oral route than the parenteral route.[2] Given this complexity, conventional parameters, such as C_max and AUC_t (the area under the plasma concentration-time curve from zero until the last sampling time), may be inadequate for PK assessments.

Despite the relatively high BSV of sumatriptan PK, only a few reports have addressed the possible causes of this variability using population PK modeling. Thus, a precise PK assessment using a population approach with a mixed-effects model may be necessary for the evaluation and application of sumatriptan. In this study, we sought to develop a population PK model in healthy Korean male subjects for sumatriptan, which frequently shows an atypical absorption profile with multiple peaks.

The present study developed population PK models for the multiple peaks phenomenon after oral administration of sumatriptan in healthy male subjects. The structural model included one- and two-compartment models with first-order elimination, and the following approaches were tested for the absorption process, which showed the multiple peaks phenomenon in many subjects: first-order absorption followed by zero-order absorption with lag time (M1), and zero-order absorption followed by first-order absorption with lag time (M2). Also, nonlinear elimination was joined to M1 and M2 by the Michaelis-Menten equation, which considered potential saturable elimination (M3 and M4).

The structure of M1 is similar to a model described by Cosson et al.[6] However, the estimated PK parameter values have considerable differences between our model and that of Cosson et al. The clearance and the lag time for first-order absorption estimated here were similar to that of Cosson et al., but the absorption constant (4.05 h^-1 vs. 0.667 h^-1), the lag-time (1.71 h vs. 0.48 h) and the duration (1.59 h vs. 3.97 h) for zero-order absorption were distinct between the models.

We also evaluated a model that included zero-order absorption followed by first absorption (M2). The absorption structure of M2 corresponded to that reported by Christensen et al.[11], except for the distribution structure, which used a two-compartment model in our study versus a one-compartment model in the Christensen et al. study. The volume of distribution in the steady state was higher in our study than the values reported from Christensen et al. Clearance was estimated to be lower than the value estimated by Christensen et al. The value (1.91 h^-1) of the absorption constant here was lower than that (7.03 h^-1) described in Christensen et al. There were also significant differences between the two models in lag time (0.21 h vs. 0.41 h) and duration (0.42 h vs. 1.29 h) for zero-order absorption. The estimated fraction of the dose absorbed via first-order absorption was higher (0.75) than the value reported by Christensen et al. (0.33). These differences in results may be related substantially to the structural PK model and the characteristics (including populations) of the subjects. The M1 model decreased the objective function value (OFV) significantly (P < 0.05) and yielded a lower AIC value when compared with the M2 model.

Sumatriptan is metabolized primarily (80%) by the MAO-A isozyme.[6, 12] Furthermore, sumatriptan and its metabolites are predominantly renally excreted. Modeling was performed under the assumption that saturation may occur in the metabolism or elimination phases.[12, 13] Consequently, elimination was expressed by parallel linear and nonlinear pathways. Nonlinear elimination was incorporated using the Michaelis-Menten equation described by the maximal elimination rate of the saturable pathway in terms of amount per unit time denoted as V_max (mg/h) and K_m (mg/L), the concentration at which the elimination rate is half the maximal rate. Therefore, nonlinear elimination was incorporated into M1 and M2 using the Michaelis-Menten equation (M3 and M4). With this model (M3), a marked decrease in OFV was obtained as compared to that in M1 (Δ = -24.801, P < 0.001, df = 2), indicating improved fit of the model to the data. However, M4 did not improve the model to a greater extent than did M3. Among the four models (M1-M4), the data were best described using a two-compartment model with first-order absorption followed by zero-order absorption, and parallel elimination of the linear and nonlinear pathway (M3).

The best structure of the PK model available to explain the time-concentration profiles of sumatriptan is the one-compartment model with an atypical absorption process, characterized by multiple peaks, described by the combined transit compartment model and first-order absorption (M5). A schematic overview of the proposed PK model is shown in Figure 2.

The final structural model was parameterized using the identical transfer rate constant for transit compartment absorption (k_tr), the rate constant of first-order absorption (k_a1), the absorption rate constant from the final transit compartment to the central compartment (k_a2), the fraction of the dose absorbed by the transit (f) and the first-order input (1-f) processes, mean transit time (MTT), the number of transit compartments placed before the central compartment (n), the lag time of first-order absorption (ALAG1), the apparent clearance (CL/F) and the apparent central volume of distribution (V/F), where F is the bioavailability. The BSV for CL/F, V/F, k_a2 and MTT were estimated successfully. Combined residual errors were estimated in the final model. In addition, screening for the effects of covariates on the PK parameters suggested that the inclusion of creatinine clearance (CrCL) had a significant effect (ΔOFV=7.42, P < 0.01) on the absorption fraction (f) (M6). Other covariates including age, weight and BMI did not appear to affect any of the PK parameters. Final parameter estimates with corresponding coefficients of variation (CV) and bootstrap results are summarized in Table 3. The basic goodness-of-fit plots for the final model are presented in Figure 3.

Figure 3
Final model diagnosis plot produced using the final pharmacokinetic model. (A) Observations (DV) vs. population predictions (PRED) (B) DV vs. individual predictions (IPRED) (C) Conditional weighted residuals (CWRES) vs. PRED and (D) CWRES vs. time (TIME).

• Click for larger image

Table 3
Final parameter estimates and bootstrap results

• Click for larger image

The model and parameter estimates were adequately robust in the bootstrap procedure. All parameter estimates from the final model were within the 95% bootstrap confidence interval (CI). The predictive performance was also sufficient, according to the VPC result (Fig. 4). The trend of the predicted time-concentration profile accorded well with the raw data.

Figure 4
Visual predictive check plot of the final model 0 and 12 h after a single oral administration of 50 mg sumatriptan. A total of 1,000 datasets were simulated using the final PK parameter estimates. Circles represent the observed sumatriptan plasma concentrations: the 90% confidence interval of the simulated concentrations (gray area), and observed concentration (solid line) of the 5^th, median, and 95^th percentiles.

• Click for larger image

In this study, we presented the results of PK modeling of the time-concentration profiles of sumatriptan after oral administration in healthy subjects. The population PK of sumatriptan is well described by a one-compartment with combined transit compartment model and first-order absorption. This combined absorption model successfully fits the multiple peaks observed in the absorption phase of sumatriptan (Fig. 5).

Figure 5
PK curves from representative individuals showing multiple peaks. Circle, observed value; solid red line, individual predicted value.

• Click for larger image

We do occasionally encounter the multiple peaks phenomenon in pharmacokinetics. The multiple peaks phenomenon can occur due to a number of different mechanisms, including those related to physicochemical and formulation factors (solubility-limited absorption, complexation: formation of poorly absorbable bile salt micelles and modified-release formulation) and physiological factors (enterohepatic recycling, gastric emptying and intestinal transit time, site-specific absorption, gastric secretion-enteral reabsorption, anesthesia and surgery).[14] Davies et al. and Aurora et al. have reported that gastric motor activity and gastric emptying play important roles in the multiple peaks phenomenon following oral administration of sumatriptan.[14, 15] A non-linear mixed effect modeling analysis with a two-process absorption model was described for the observed multiple peaks phenomenon of sumatriptan concentration-time data.[6, 11] Although the two-process absorption model resulted in an improved model compared with typical absorption models, the multiple peaks of individual subjects were well not captured visually by these models. Thus, we developed a model using a transit compartment model to pursue more mechanistic PK modeling approaches.

The use of a transit compartment model to describe a delay in the onset of absorption was developed by Radojka Savic et al.[8] They reported that a transit compartment model describes drug absorption delay as a multi-step process, represented by a chain of identical pre-systemic compartments that are linked to the central compartment by a first-order absorption process, without assigning a physical correlate to each transit compartment.[8] They suggested that one of the complexities of the absorption process is incomplete gastric emptying. The combined transit compartment model and first-order absorption had a reduced AIC value compared with the previous conventional models (Table 2), and captured the multiple peaks in the absorption phase of sumatriptan well.

We also explored the incorporation of significant covariates in the base model. Although we included only healthy adult male volunteers, who were evaluated under well-controlled experimental conditions, CrCL as covariate had a significant effect on the absorption fraction (f) (P < 0.01) (M6 in Table 2). One possible explanation for this is that as the elimination of sumatriptan is increased, sumatriptan is relatively well absorbed by first-order absorption. Studies from a more diverse population may allow for more meaningful results and accurate PK modeling of sumatriptan. There was good agreement between the observed and predicted concentrations, and no obvious trends in the weighted residuals vs. time (Fig. 3). The PK parameters were generally well-estimated, with the standard error for the estimation being under 40% of the estimated population parameter values. Additionally, when nonlinear elimination was incorporated into the base model (M5) using the Michaelis-Menten equation, the OFV of M7 decreased significantly compared to that of M5 (Δ = -10.17, P < 0.01, df = 2). These results indicate that saturation may have occurred in the metabolism or elimination processes. The saturation of the MAO-A isozyme in sumatriptan was not assessed in this study, but would be a reasonable factor to assess in future studies.

In this study, a population PK model for multiple peaks was developed and reasonable parameters were obtained from the data of healthy Korean male subjects. We could find that the estimated parameters were altered by the adjusted models in the PK showing multiple peaks. This is the first study to model the PK of sumatriptan, which frequently exhibits an atypical absorption profile with multiple peaks, using a combined transit compartment model and first-order absorption using NONMEM. This model may facilitate the determination of basic sumatriptan PK characteristics. Furthermore, our absorption model may facilitate PK modeling of multiple peaks during the absorption phase.