Pharmacokinetic-Pharmacodynamic Modeling of the Effect of Varenicline on Nicotine Craving in Adult Smokers

Abstract

Introduction:

Varenicline has been shown to significantly reduce craving and several aspects of smoking reinforcement in clinical trials, compared with placebo. This is the first report describing the concentration-effect relationship of varenicline on relief of craving.

Methods:

The pharmacokinetics (PK) and pharmacodynamics (PD) of a single 2mg dose of varenicline were investigated in 40 smokers in a randomized, crossover study comparing the effect of varenicline with placebo on ameliorating abstinence-and cue-induced craving and withdrawal symptoms. Subjects were asked to complete self-reported questionnaires (Smoking Urges Scale and Minnesota Nicotine Withdrawal Scale [MNWS]) and blood samples were simultaneously collected for measurement of varenicline concentrations. Only the data from the 4-hr postdose abstinence period (just prior to the cue session) were analyzed. Data were described by a 2-compartment PK model and a linear PD model with first-order onset/offset rate constants describing the placebo response “kinetics.” Response was described as the net effect of the baseline, placebo, and drug responses.

Results:

Varenicline significantly decreased mean craving score when compared with placebo and the magnitude of this response was related to varenicline concentration. The time-course and magnitude of both placebo and varenicline craving response were characterized by a large degree of unexplained variability. Simulations were used to illustrate the expected craving response over time and its associated random variability after chronic dosing.

Conclusions:

Craving reduction is associated with increased varenicline concentrations. The relatively rapid onset of this effect within 4hr postdose suggests that, smokers may experience some craving relief after acute administration of varenicline.

Introduction

Craving or urge to smoke, which can be elicited by both abstinence and exposure to smoking-associated cues, has been identified as a prominent feature of nicotine dependence and a major contributor to the high failure rate of treatment interventions for smoking.¹ Because strong, repeated, and persistent withdrawal-induced cravings present challenges for smokers wanting to quit smoking, the relief of these symptoms would be considered an important step in preventing relapse.^2,3 Varenicline tartrate (Chantix®/Champix®) is currently approved in 110 countries worldwide as an aid to smoking cessation. The recommended adult dosage is 1mg twice daily (b.i.d.) for 12 weeks with an initial up-titration week. The pharmacological profile of varenicline has been described in detail elsewhere and is consistent with that of a potent and selective partial agonist of the α₄β₂ nicotinic acetylcholine receptor, the subtype that plays a key role in the addictive effects of nicotine.^4–6 Because of its dual agonist-antagonist properties, varenicline offers the potential therapeutic benefit of simultaneously relieving symptoms of nicotine withdrawal and cigarette craving during abstinence, while attenuating the reinforcing effects of nicotine and psychological reward associated with smoking.^4,7 Clinical evidence from self-report data have indeed indicated that, unlike other cessation medications, abstinence increased over the first weeks of varenicline use; this suggests that smoking whilst on varenicline is gradually less rewarding, thereby helping to initiate abstinence and perhaps setting the stage for a more stable quit.^8–10 Similar findings have been reported from laboratory investigations with varenicline when self-reporting smoking reward was assessed immediately after smoking.^11–13

The present report is part of a study, evaluating the effect of varenicline on ameliorating abstinence- and cue-induced nicotine craving and withdrawal symptoms in a population of 40 smokers not currently intending to quit.¹⁴ The cue-reactivity model was based on a rigorous laboratory-based assessment paradigm including in vivo exposure to smoking cues, specifically designed to provoke craving and assess the effects of smoking cues on the urge to smoke and withdrawal symptoms. Results showed that a single dose of varenicline reduced tonic craving levels in smokers who refrained from smoking, but not cue-provoked craving. The subject of this paper was to develop suitable population pharmacokinetic (PK)-pharmacodynamic (PD) models to relate the intensity of craving to varenicline plasma concentration and explore any clinical or demographic covariate factors as potential predictors of response. Population analysis (also known as nonlinear mixed effects modeling) is the application of a pharmacostatistical modeling technique to describe the average response in a population as well as identify the sources of variability in response.^15–18 Response refers to either changes in the concentrations (PK) or pharmacological effect (PD) of the drug over time. A pharmacostatistical model consists of three components: a structural model that describes the response for a typical (average) individual; and two distinct statistical models that describe the variability between individuals and the remaining unexplained residual variability. Because data arise from more than one individual, the influence of patient characteristics (demographic or physiological factors) on the PK or PK-PD profile can be quantified. Furthermore, this population-based approach allows borrowing information between individuals which is required in a sparse sampling setting as practiced routinely in Phase 2 and 3 clinical studies.^19–21 To our knowledge, a population PK-PD model has been previously developed to characterize the relationship between nicotine replacement therapy (NRT)-induced nicotine concentration and abstinence-induced craving, as assessed by the Tiffany scale.²² While the modeling approach was able to describe the temporal changes on craving in a period of 72hr after placebo and NRT, our modeling exercise aimed at quantifying the acute effect of varenicline on relief of craving during the 4-hr postdose abstinence period, just prior to the cue reactivity sessions. The final PK-PD model was then utilized to explore the expected craving response under different varenicline dosing scenarios.

Materials and methods

Full details of the study design and methodology of the clinical trial have already been described elsewhere.¹⁴

Study Design

This study used a randomized, double-blind, placebo-controlled, crossover design comparing a single 2mg (2×1mg tablets) dose of varenicline with placebo for relief of cigarette craving in a population of otherwise healthy smokers, between the ages of 18 and 65, and not currently intending to quit smoking. The study was conducted at a single center in the United States (Center for Behavioral Medicine at the Miriam Hospital) and consisted of a screening visit, two 1-day treatment visits separated by a 7-day washout period, and a closeout (follow-up) visit, 1 week after the second treatment visit. To be included, subjects had to have: (a) smoked at least 20 cigarettes (one pack) per day on average during the past year, or (b) smoked between 11 and 19 (inclusive) cigarettes per day on average during the past year and responded to the first item on the Fagerström Test of Nicotine Dependence (“How soon after you wake do you smoke your first cigarette?”) by indicating “within 30min of arising.” In addition, subjects were required to have had no period of abstinence from smoking greater than three months in the past year, and were not presently intending or attempting to quit. Subjects also had to be willing to refrain from smoking for a specified 8-hr period during each of the two study treatment days while in the Clinical Research Unit (CRU). The study was conducted in accordance with the International Conference on Harmonization Guidelines for Good Clinical Practice (GCP), the Declaration of Helsinki, and Food and Drug Administration regulations. The protocol was approved by the Institutional Review Board at the Miriam Hospital. Written informed consent was obtained from all participants prior to enrolment in the study.

Study Procedures

During the treatment phase, subjects were admitted to the CRU between 8:00 and 10:00 a.m. on the day of each treatment visit, abstinent from smoking and eating from at least midnight the night before. Subjects were instructed to smoke in their usual manner on the day prior to each visit. Overnight abstinence was confirmed biochemically with CO levels < 15 ppm according to the study protocol. Subjects were provided a standard meal after they completed a series of self-report questionnaires—the Smoking Urges Scale^23,24 and the MNWS,^25,26 modified to assess current experience of symptoms—to specifically assess baseline craving and withdrawal symptoms. After completing the questionnaires and between 0.5 and 1.0hr after eating in Period 1, subjects were administered study medication with 240ml of water under the supervision of the study staff. Subjects then completed the set of questionnaires every 30min following dosing, and immediately prior to and at 15 and 45min after a specialized, scripted cue reactivity session beginning 4hr after dosing. Serial blood samples for varenicline content were concomitantly collected at predose and at 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 5.25, 5.75, and 6.25hr postdose. Subjects remained in the clinic throughout the visit day and were required to refrain from cigarette smoking or use of any tobacco product during that time. Once all procedures were performed, subjects were discharged from the CRU. After a 7-day washout period, subjects returned to the CRU for the second treatment visit (Period 2 of the crossover). Procedures during the second treatment period were identical to those in the first period except that subjects were crossed over to the alternate drug treatment (e.g., placebo to varenicline or varenicline to placebo).

Analytical Assay

Plasma samples were analyzed for varenicline concentrations using a validated high-performance liquid chromatography-atmospheric pressure ionization/tandem mass spectrometry detection assay following liquid-liquid extraction, as described previously.^27,28 The dynamic range of the assay was 0.10–50.0ng/ml. Maximum plasma concentration (C _max) and the time of the first occurrence of C _max (T _max) were derived directly from the observed data.

Pharmacokinetic and Pharmacodynamic Analyses

Population PK and PK-PD models were developed using nonlinear mixed-effects modeling (NONMEM 7.1.2; ICON Development Solutions),²⁹ to explore the relationship between craving response and plasma concentration of varenicline. Only the data from the abstinence period between time 0 (just before dosing) and 4hr postdose (just prior to the cue reactivity session) were included in these analyses because of possible confounding of the PD data by the intervening cue exposures. Although two subjects did not complete both periods of the crossover design, any incomplete PK-PD data they contributed were included in the mixed-effects model-based analysis. The Smoking Urges Scale was the primary endpoint to assess cigarette craving. The scale consists of five questions each scored as an integer on a 0 to 100 scale, where 0 represents “Not at all” and 100 represents “The strongest feeling possible.” Craving was defined as the average of the five items’ scores (mean Q1–5) for each subject and scheduled assessment. A secondary endpoint was derived from the MNWS, which consists of nine questions. Each question was rated from 0 to 4 where 0 = “Not at all,” 1 = “Slight,” 2 = “Moderate,” 3 = “Quite a bit,” and 4 = “Extreme.” Based on Cappelleri et al., the scale consists of two multi-item domains (negative affect [including depressed mood, irritability, anxiety, and difficulty concentrating]: Q2–Q5; and Insomnia: Q8–Q9) and three individual items (urge to smoke: Q1; restlessness: Q6; and increased appetite: Q7), which should be analyzed separately rather than as part of a composite score. Ratings for each domain were averaged over the combined items to yield a composite score (mean Q2–Q9) at each assessment. The categorical endpoint, Item 1 (Q1) was examined separately to assess craving. The PK-PD data were best characterized by a two-compartment PK model and a linear PD model with first-order onset/offset rate constants to describe the placebo response “kinetics” (Figure 1). The final PK model and population parameters previously obtained in a varenicline pooled analysis were used as prior information in a maximum a posteriori (MAP) Bayesian analysis (POSTHOC option) of the new PK data from this study.³⁰ Concentrations below the lower limit of quantification (<2%) were excluded. Individual PK parameters for the two-compartment model with first-order absorption and elimination were estimated, given the population priors, and individual varenicline concentrations were predicted for all actual PD observation times. The time-varying behavior of the placebo response was modeled as a separate hypothetical kinetic system with a “dummy” unit dose introduced at times of placebo or varenicline dosing. The concentrations of varenicline and placebo in the central compartment of each kinetic system were linked to effect by a linear PD model with slope parameters, DSLP and PSLP, respectively. Response was described as the net effect of the baseline (E ₀), placebo (PSLP*C _PBO), and drug treatment (DSLP*C _DRUG). Because linear models are simplistic empirical characterizations of the “true” concentration-response relationships, a more clinically relevant PD model, which is naturally constrained between 0 and 100, was developed using logistic transformation. Inter-individual and residual random effects were characterized by an additive error model. All PK-PD models were run using the general linear model subroutine (ADVAN 5) and First-Order Conditional Estimation method.²⁹ Goodness-of-fit and model performance were assessed by inspection of diagnostic plots and evaluation of final Minimum Objective Function (MOF) value, parameter estimates, standard errors, and visual predictive checks (VPC). Typical diagnostic plots included model predictions versus observed data, weighted residuals versus time and model predictions, and the correlation of interindividual random effects.³¹ Changes in MOF guided the model building process and parameter estimates were assessed for their clinical plausibility and precision. VPC are simulation-based diagnostics used to assess the model’s ability to reproduce the distributional characteristics (e.g., central tendency, 5^th and 95^th percentiles) of the observed data.^32,33 Potential covariate effects were investigated by graphical inspection of covariate-parameter relationships.^31,34 Further exploratory evaluation of covariates selected from the graphical screen was conducted using the nonlinear mixed-effects model. The final models were evaluated via visual predictive checks.

Figure 1.

Pharmacokinetic-pharmacodynamic model for the effect of varenicline on relief of cigarette craving, as assessed by the Smoking Urges Scale.

Results

Subject Disposition and Smoking History

The study randomized 21 male and 19 female smokers, contributing a total of 305 concentration-time data, 387 placebo-craving responses, and 390 varenicline craving responses. The mean age was 36 years (range 18–63). Subject’s body weight was on average 77kg (range 59–95) and 72kg (range 58–90) for males and females, respectively. All but one subject (97.5%) were White. On average, subjects smoked approximately 21 cigarettes per day (range 16–40) and had been smoking since the age of 17 (range 12–31). Subjects tried to quit three times, ranging from 0 to 9 quit attempts. Seven of the 40 subjects reported using Zyban® in past quit attempts. On average, subjects scored 5.6 (SE ± 0.31) of a possible 10 points (total score) on the Fagerström Test for Nicotine Dependence.

Pharmacokinetics

Varenicline plasma concentrations were characterized by a mean (SD) C _max of 8.27 (1.47) ng/ml and a median T _max of 3.00hr (range 1.50–6.25). The previously established varenicline population PK model, which included the covariate effects of renal function on clearance (CL/F) and body weight on volume of distribution (V/F) as predictors of intersubject variability in exposure, was used to predict the individual concentration-time profiles based on the subjects’ characteristics in this study, which were then utilized to drive the concentration-effect relationships. The MAP Bayesian estimation of the individual compartmental parameters resulted in good fits to the individual observations (data not shown), thus supporting the use of empirical Bayes estimates for subsequent PK-PD modeling.

Relationship Between Varenicline Concentration and Mean (1–5) Craving Score

The individual PK-PD data were highly variable (Figure 2) and when viewed as pooled data, no apparent population concentration-response relationship was observed (not shown). Although the individual-specific nature of the population PK-PD model allowed controlling for response variability due to PK differences, a great deal of variability in PD (craving score) still remained. The proposed PK-PD model described the mean craving response as a sum of responses for baseline craving, placebo effect, and varenicline treatment effect. When examining the individual craving response patterns observed during the placebo period, it was necessary to account for a placebo response that changed over time. Since this was a crossover study, the placebo response for a given individual was assumed to be the same on both treatment periods. The placebo-effect model was first fitted to placebo craving score data and then, in a second step, the linear PD model was simultaneously fitted to both the placebo and varenicline craving scores, having fixed the placebo “kinetic” parameters to the individual specific values estimated previously. The model building process showed a significant varenicline treatment effect. The base model used for comparison was a PK-PD model including baseline and placebo responses only. The addition of a fixed-effect parameter for drug effect resulted in an improved goodness-of-fit, as evidenced by the substantial drop (−71 points) in the minimum objective function value (Bonferroni-adjusted p-value, based on likelihood ratio test adjusted for number of model comparisons, p < .025). When an interindividual random effect was added to the drug slope parameter, the change in objective function was nearly 725 points (Bonferroni-adjusted p-value < .025) compared with the base (no drug effect) model. It should be noted that although the large variability precluded detection of a concentration-response relationship when individual PK-PD data were pooled, the model-based analyses were able to resolve the underlying subject-specific concentration-response relationships.

Figure 2.

Mean (±SD) time-course profiles of observed varenicline plasma concentrations and mean (1–5) craving scores, as assessed by the Smoking Urges Scale (first 4hr postdose).

Several covariate factors, including study (not treatment period) baseline nicotine concentration (mean 12.3ng/ml; range 0–29), study (not treatment period) baseline cotinine concentration (mean 235ng/ml; range 74–430), study (not treatment period) baseline carbon monoxide (mean 20 ppm; range 10–38), years of smoking, daily amount smoked, Fagerström question 1 score (<5 min: 9; 6–30 min: 23; 31–60 min: 6; and >60 min: 2), treatment sequence, and sex were available for inclusion in the population PK-PD model. A graphical inspection of potential covariate-parameter relationships was not indicative of any obvious trends. Further attempts to explain random variability by univariate inclusion of covariate factors directly into the nonlinear mixed-effects model revealed no improvements in goodness-of-fit as assessed by MOF-based Likelihood Ratio Test, and diagnostic plots. No further covariate model building was conducted. There was no apparent relationship between baseline craving and the magnitude of drug response. An attempt was made to incorporate a mixture model to explore the individual probability of belonging to a subpopulation of responders versus non-responders, but this effort was not supported by the data. Scatter plots of observed versus predicted mean craving scores for each subject together with the identity line revealed the good agreement between model predictions and observations. The goodness-of-fit diagnostics for the final population PK-PD craving model revealed little bias and a large amount of unexplained random variability. When individual random effects were estimated, given the individual data and population priors, the estimation of response was clearly improved. All the estimated PK-PD model parameters demonstrated large unexplained interindividual variability (Supplementary Table S1), which is reflected both in the heterogeneity of the relationship between varenicline concentration and the changes on the Smoking Urges Scale (Figure 3), and in the high interindividual variability in the response-time profiles for placebo and varenicline treatments (Supplementary Figures S1–S2).

Figure 3.

Observed and individual model-predicted (IPRED) concentration-response relationships for the effect of varenicline on relief of cigarette craving (placebo-adjusted mean craving score, as assessed by the Smoking Urges Scale). No PD data were available for Subject ID #20 who contributed only PK information to the PK/PD analyses.

Simulation of the Reduction in Cigarette Craving Under Various Varenicline Dosing Regimens

Given that the population PK/PD model adequately characterized the effect of a single dose of varenicline on craving, as measured by the Smoking Urge Scale, it was decided to explore the time-course of the craving response following repeat administration of varenicline at therapeutic dose regimens. Based on the final model and parameter estimates, 500 replicate trials (40 subjects each) for each dosing scenario of varenicline administered 1mg once or twice (titrated and non-titrated) daily were generated using Monte Carlo simulation. Simulations were conducted to describe the population mean craving response-time profile and included the uncertainty (derived from the variance-covariance matrix of estimates) associated with the final PK-PD model parameters. Figure 4 illustrates the expected reduction in craving at the usual adult dosage of 1mg b.i.d. varenicline after an initial titration week; simulation results for the non-titrated regimens are presented in Supplementary Figure S3. Overall, relief from craving is gradually achieved as dose progressively increases from 0.5mg to 2mg daily. Consistent with the proposed PK-PD model, the response pattern shows a reduction in craving scores as varenicline appears in plasma followed by a slow return to baseline score as varenicline concentrations decline. After steady-state attainment of varenicline 0.5mg b.i.d., mean craving response is expected to be substantial and sustained by day 7. The magnitude of this effect is slightly further pronounced after a week at the target dose of 1mg b.i.d.

Figure 4.

Simulations of the expected time-course of the craving response (as assessed by the Smoking Urges Scale) during varenicline treatment at the usual adult dosage (0.5mg q.d. for three days, followed by 0.5mg b.i.d. for four days, and then 1mg b.i.d.). The black line is the median response; the gray lines are the 90% prediction interval; and the gray shaded area represents “extrapolation”.

Relationship Between Varenicline Concentration and MNWS (Urge to Smoke)

In addition to the Smoking Urges Scale scores, a concentration-effect relationship for Question 1 (Urge to Smoke) and mean Q2–Q9 of the MNWS was explored using population PK-PD modeling. A preliminary assessment was conducted by treating these MNWS endpoints as continuous data using the same structural model as was employed for the mean craving (1–5) endpoint. Model convergence was problematic for both of these secondary endpoints; an ordered categorical cumulative logit model was subsequently implemented to more adequately describe the individual longitudinal MNWS Q1 data up to 4hr postdose (Supplementary Figures S4–S5). No further model building was however attempted with the other single- or multiple-item domains as no trends toward a significant drug treatment effect were evident from graphical inspection.

Discussion

The finding of a significant effect of varenicline on reducing abstinence-induced craving, as measured by the Smoking Urges Scale, prompted the exploration of a possible relationship between the temporal changes of craving scores and observed varenicline plasma concentrations.¹⁴ The laboratory study employing a validated cue reactivity model showed that a single dose of varenicline compared with placebo ameliorated self-reported craving and withdrawal symptoms during the 4-hr postdose abstinence period that preceded cue reactivity sessions. Cue-stimulated craving was, however, not blunted. The results are consistent with varenicline’s effect on tonic craving but not phasic response to a provocative stimulus in smokers who refrained from smoking. The PK-PD modeling approach developed in this study was able to describe variations in craving by characterizing the onset and extent of the pharmacodynamic response as the sum of responses for baseline craving, placebo effect, and varenicline effect. Varenicline significantly decreased craving when compared with placebo treatment and the intensity of this response was related to varenicline concentration. This was further supported by a significant concentration-response relationship that was identified using a categorical logistic regression model for the secondary endpoint, Item 1 (Smoking Urge) of the MNWS. It was also evident that the time-course and magnitude of both placebo and drug responses were characterized by a large amount of unexplained interindividual and possibly intra-individual variability. The range of variability due to unexplained sources (i.e., distribution of residuals) was noticeably greater than the predicted variability due to fixed effects (i.e., distribution of differences between the model predicted values and the mean of the observed data). This indicates that the PK-PD model structure and parameters describe the typical population response reasonably well, but the fixed-effects model lacks individual-specific predictive ability to explain interindividual variability in response. Although this study tested volunteers after a single 2mg dose, plasma peak concentrations were consistent with those typically observed following chronic administration of varenicline 1mg b.i.d.³⁵ These results also are consistent with the patient-reported data from the Urge to Smoke item of the MNWS, the Brief Questionnaire of Smoking Urges, and the Modified Cigarette Evaluation Questionnaire in clinical trials of varenicline in smokers who make an attempt to quit smoking, supporting that varenicline reduces the craving and reinforcing effects of smoking.^8,9,10,36

Limitations of the present work, however, should be noted. The current study design—single-dose with rising concentrations only—provided useful information for exploration of the PK-PD relationship, but was not optimal. Future data should include rising and falling concentrations and multiple-dosing designs for a more robust characterization of the time-course of the mean craving response as it relates to varenicline plasma concentration over the dosing interval. Although the simulations exhibit a consistent, decreasing trend in the craving intensity over time, the assumption of a direct inhibitory pharmacodynamic effect may or may not be accurate. In addition, the simulations do not take into account smoking behavior during pre-quit use of varenicline, since this study was conducted as a Phase 2 study prior to the varenicline Phase 3 pivotal studies and pre-quit use of varenicline was not considered at that time. The simulated craving response profiles should be interpreted with this limitation in mind. Another evident model deficiency was an inability to describe the observed response variability as a function of any predictive covariates or independent variables. It is possible that in a larger study with a more heterogeneous population, a wider range and larger number of covariates might lead to some significant predictors. In particular, studies should include individuals with a wider range of nicotine dependence; the current study focused primarily on heavier smokers. A third complicating factor in the development of the PK-PD model was the possibility of inter-occasion variability in placebo response. The predicted typical drug effect on craving was consistent, but small. With the current study design (crossover with two treatment periods), it was not possible to estimate the inter-occasion variability in the placebo (or varenicline) response and, therefore, it was assumed that the response was consistent across both treatment periods. It is possible, however, that the placebo (or drug) response was not consistent across treatment periods. This dilemma could be overcome by measuring both drug and placebo treatment responses on multiple, distinct occasions. Finally, these models were tested among moderately dependent smokers following overnight nicotine deprivation. Future studies are required to determine whether similar results would hold under different conditions, for example, when smokers are nicotine satiated.

In conclusion, this is the first PK-PD investigation to describe the concentration-effect relationship for the effect of varenicline on measures of craving in smokers who refrained from smoking. The PK-PD model revealed that varenicline significantly decreased craving scores when compared with placebo and the magnitude of this response was related to plasma varenicline concentration. This analysis was specifically focused on the onset of craving reduction, and clinical evidence of a concentration-dependent change on craving over the first 4hr after oral administration of varenicline is a useful finding with respect to the characterization of the pharmacological effects of varenicline. The findings are consistent with reports from multiple clinical trials in adult smokers who make an attempt to quit smoking. Given the sparse nature of the observations available, the current PK-PD model should however be viewed as a descriptive tool. The model could be improved with the collection of additional, sparse PK-PD data from future studies specifically designed for this purpose. Of note, varenicline’s reduction in tonic craving observed within the first 4hr after dosing may be of clinical importance in mediating initial abstinence. This relatively fast-acting “rescue” medication effect could potentially have a role in relapse prevention and would require further study.

Supplementary Material

Supplementary Figures S1–S5 and Table S1 can be found online at http://www.ntr.oxfordjournals.org

Funding

The study was funded by Pfizer Inc (study no. A3051005).

Declaration of Interests

TCL is a full-time employee of Pfizer Inc. RN was the principal investigator of the primary study, sponsored by Pfizer and conducted while at The Miriam Hospital. He has consulted for Pfizer and has served on the Pfizer speaker bureau. MRG, President and CEO of Metrum Research Group, made significant scientific contributions to the PK/PD modeling analyses reported in this article. Metrum Research Group served as paid consultants to Pfizer for their technical expertise in connection with the data analyses and development of this article. PR is a current employee of Boehringer Ingelheim, Pharmaceuticals Inc. HMF is a current employee of Takeda Pharmaceuticals International Co. Both PR and HMF were full-time employees of Pfizer at the time of completion of these analyses and writing of this manuscript.