FormalPara Key Points
Table 1
Full size table

1 Introduction

Ticagrelor, a direct-acting and reversibly-binding P2Y12 receptor antagonist, is approved to reduce the rate of cardiovascular death, myocardial infarction (MI), and stroke in patients with acute coronary syndrome, or a history of MI, in combination with low-dose aspirin [1, 2]. Ticagrelor is administered as a 180 mg loading dose followed by 90 mg twice daily for 1 year in adults with acute coronary syndrome, and 60 mg twice daily (no loading dose) for adult patients with a history of MI. Ticagrelor inhibits adenosine diphosphate-induced platelet aggregation [3,4,5], and reduces adenosine cellular uptake via nucleoside transporter 1 inhibition [5,6,7]. Ticagrelor pharmacokinetic (PK) and pharmacodynamic (PD) properties in adults are well established [8]. Oral ticagrelor is rapidly absorbed, has a dose- and time-linear PK profile [9, 10], and is not affected by food intake [11]. Ticagrelor is primarily metabolized via the cytochrome P450 (CYP) enzymes CYP3A4/5 to the active metabolite AR-C12490XX [12]. While AR-C12490XX plasma levels are approximately 30–40% of ticagrelor concentrations [13,14,15], the metabolite is equipotent regarding platelet inhibition [5]. Platelet inhibition is rapid, temporary, and directly related to ticagrelor and AR-C12490XX plasma concentrations [5, 9, 10, 14, 16]. Pharmacometric approaches to describe the ticagrelor population PK were developed in adults with acute coronary syndrome [17] or a previous MI [18]. Ticagrelor population PK (adult healthy volunteers) and a population PK/PD model (adults with coronary artery disease [CAD]) were also developed [19].

Ticagrelor is being evaluated in sickle cell disease (SCD). Altered hemoglobin β-chains polymerize when deoxygenated, resulting in rigid, sickle-shaped erythrocytes, which obstruct blood vessels and reduce blood flow, with subsequent pain and organ damage [20, 21]. Pharmacological SCD treatments are available [22], however there is an unmet need for well-tolerated therapies targeting disease pathology, particularly in children [23, 24]. Platelets are activated in SCD [25] and may contribute to vaso-occlusion by cell aggregate formation [26, 27]. Thus, inhibiting platelet activation is a potential SCD therapeutic option. Previous studies have evaluated antiplatelet agents in SCD. Platelet inhibition with ticlopidine reduced the incidence of vaso-occlusive crises by 60% (p = 0.0001) [28]. The DOVE phase III trial in children and adolescents (2–17 years of age) reported numerical differences between prasugrel and placebo in vaso-occlusive crises rates [29] at a modest platelet inhibition level (approximately 20%) [30]. It can therefore be hypothesized that antiplatelet agents with a predictable platelet inhibition aiming for higher platelet inhibition may be useful in SCD.

HESTIA1 was the first study to evaluate ticagrelor PK and PD properties, as well as safety, in pediatric SCD patients [31]. An exploratory population PK/PD modeling and simulation exercise, using adult ticagrelor data, informed the dose selection (AstraZeneca, data on file). The proposed doses were based on three key clinical pharmacology assumptions: (1) that ticagrelor PK in children was similar to adult patients with CAD and could be allometrically scaled by body weight according to common principles; (2) a similar sensitivity between adults and children with SCD (i.e. a similar ticagrelor concentration for half-maximum effect [EC50] on platelet inhibition); and (3) a similar variability in P2Y12 reactivity units (PRU; a measure of platelet reactivity) between adult patients with acute coronary syndrome and pediatric SCD patients.

Population modeling permits quantitative testing of potential covariates on PK and PD parameters, which is important in pediatric studies as patient and sample numbers are limited [32]. The aims of the present analysis were to characterize the time course of ticagrelor and AR-C12490XX plasma concentrations following single and repeated doses using population PK modeling; to assess the potential impact of relevant demographic covariates (particularly body weight) on ticagrelor and AR-C12490XX PK; and to apply the developed population PK model to describe the time course and extent of platelet inhibition as a function of ticagrelor plasma concentrations. Thus, for the first time, we quantitatively characterize the dose-to-exposure-to-platelet inhibition response relationship for ticagrelor in pediatric SCD patients.

2 Methods

Data underlying the findings described in this article may be obtained in accordance with AstraZeneca’s data sharing policy, available at https://astrazenecagrouptrials.pharmacm.com/ST/Submission/Disclosure.

2.1 Study Design, Patients, and Dosing

HESTIA1 (NCT02214121) was a two-part, randomized, multicenter, phase II study with an open-label, dose-ranging phase (Part A) and an optional, double-blind, placebo-controlled, extension phase (Part B). Patients were ≥2 to <18 years of age, with either homozygous sickle cell or sickle beta-zero thalassemia [31]. Concomitant drugs considered to be strong CYP3A4 inhibitors or inducers were not permitted in HESTIA1.

In Part A, 46 patients were randomized to different dosing schedules, including two single doses of ticagrelor (0.125–2.25 mg/kg) ≥ 7 days apart, then twice-daily dosing for 7 days (0.125, 0.563, or 0.75 mg/kg). In Part B, patients received ticagrelor (0.125, 0.563, or 0.75 mg/kg twice daily; n = 16) or placebo (n = 7) for 4 weeks. Ticagrelor was administered as an oral suspension in water and administered according to individual screening body weight (mg/kg) [31].

2.2 Sample Collection and Bioanalyses

Time-matched blood samples were collected (Online Resource Fig. 1).

Ticagrelor and AR-C12490XX plasma concentrations were measured using a fully validated liquid chromatography tandem mass spectrometry method [33]. Calibration ranges were 1.9–3827 nmol/L (ticagrelor) and 5.2–2090 nmol/L (AR-C12490XX), with a 100 µL sample volume. The lower limits of quantification (LLOQ) were 0.96 nmol/L (ticagrelor) and 2.61 nmol/L (AR-C12490XX); the lower limits of detection were not defined for the bioanalysis [33].

Measurement of PRU was performed using VerifyNow® (Accriva Diagnostics, San Diego, CA, USA) according to the manufacturer’s instructions. The number of PRU samples was limited by the blood sample volume (4 mL; the first 2 mL were discarded from the blood line, and the next 2 mL were used for the assay) needed for the assay in relation to the total blood volume in children.

2.3 Datasets

A dataset structured for NONMEM analysis was created based on demographic, medical, and surgical history data, as well as individual PK and PD data from all blood samples collected from HESTIA1 patients, if dosing and/or sampling history records were complete for the samples, even if the actual time post-dosing was > 12 h. Two PK samples were excluded as time records were missing. Six patients had missing dose intake time information after 1 week of twice-daily dosing, and the dose time was imputed to 8:00 pm prior to this visit. PK samples below the LLOQ were included in the dataset, and were set to 50% of the LLOQ, to facilitate testing of inclusion of such data in the model as outlined by Beal [34]. In the prespecified analysis plan, if < 10% of PK samples were below the LLOQ, these data were to be excluded from the modeling (M1 method in the article by Beal [34]) as the impact of this exclusion on parameter bias is considered low [35, 36]. Sex, age, race, dose, and medical and surgical history information [37] were included in the model to assess the impact (or lack of impact) of relevant covariates on ticagrelor PK and PD.

2.4 Model Development

Analyses were based on established principles [38, 39]. All population analyses were performed using NONMEM® software (version 7.3, Globomax, Hanover, ND, USA) on a Linux-based Intel cluster (CentOS 5). Estimation of the maximum likelihood inference was performed using first-order conditional estimations with interaction. Model fitting was performed in the Linux environment using a GFortran FORTRAN compiler (version 4.7.3, Gnu Compiler Collection). Perl-speaks-NONMEN (PsN; version 4.4.8, Uppsala University, Uppsala, Sweden) was used for NONMEN runs, visual predictive checks (VPCs), and log-likelihood profiling.

Data processing, graphical analyses, model diagnostics, and statistical summaries were conducted using R version 3.2.4 (R Development Core Team, Vienna, Austria). The ancillary program Xpose4 (version 4.5.3; Uppsala University, Uppsala, Sweden) generated VPCs and modeling result summaries [40].

2.4.1 Model Selection Criteria

Model selection was based on goodness-of-fit (GOF) plots, the objective function value (OFV; a global measure of fit in NONMEN), parameter estimates, standard error, and correlation of parameter estimates. If adding a parameter resulted in a fall in the OFV of > 3.84 (α = 0.05), this addition was considered a significant improvement in model fit.

2.4.2 Pharmacokinetic Model Development

The structural PK model (Fig. 1) was parameterized in terms of clearance and volume parameters. These parameters were allometrically scaled using fixed exponents for weight (clearance: 0.75; volume: 1) [41, 42]. The fraction of ticagrelor metabolized to AR-C124910XX was set to 0.22 (Astra Zeneca, data on file).

Fig. 1
figure 1

Structural pharmacokinetic model for the simultaneous modeling of ticagrelor and AR-C124910XX plasma concentration and time data. CL/F apparent ticagrelor clearance, CLm/F apparent AR-C124910XX clearance, Fm fraction of ticagrelor metabolized into AR-C124910XX (fixed to 0.22), KTR first-order absorption rate constant, parameterized as the mean transit time in the pharmacokinetics, Q/F ticagrelor intercompartmental clearance, Qm/F AR-C124910XX intercompartment clearance, Vc/F apparent central ticagrelor volume of distribution, Vcm/F apparent central AR-C124910XX volume of distribution, Vp/F apparent peripheral ticagrelor volume of distribution, Vpm/F apparent peripheral AR-C124910XX volume of distribution

Full size image

2.4.3 Pharmacokinetic/Pharmacodynamic Model Development

A sequential PK/PD analysis method was utilized. Empirical Bayes estimates of individual PK parameters were used to predict individual ticagrelor concentrations, which were then linked to the observed absolute scale PRU time course. A suitable baseline PRU model was established using PRU data at randomization.

The PK/PD model was a proportional direct effect model where ticagrelor concentrations drive platelet aggregation inhibition in a sigmoid Emax-type model:

$$ {\text{PRU}}_{\text{t}} = {\text{PRU}}_{\text{baseline}} \times \left( {1 - \frac{{E_{\hbox{max} } \times C_{\text{t}}^{\gamma } }}{{{\text{EC}}_{50}^{\gamma } + C_{\text{t}}^{\gamma } }}} \right) $$

where PRUt = platelet inhibition at a given time point; PRUbaseline = PRU value at time zero; EC50 = concentration at half-maximum effect; Emax = maximum effect; γ = sigmoid slope factor; and Ct = individual predicated ticagrelor plasma concentration.

Between-subject variability (BSV) of baseline PRU and EC50, and the correlation between these random effects, were estimated. Only ticagrelor concentrations were considered in the PK/PD model; it was assumed that the parent : metabolite ratio was similar and that the metabolite contribution to the overall effect was the same in all patients.

2.4.4 Covariate Relationships

A stepwise forward inclusion and backward exclusion process (implemented in PsN) [α = 0.05 for inclusion and α = 0.01 for exclusion] was used to evaluate covariates as part of the final PK and PK/PD model building. The covariate that resulted in the greatest drop in OFV was used as the base to develop the model. Any remaining covariates that significantly decreased OFV as a function of one of the primary model parameters were then sequentially added to the base covariate model, and the process repeated until all significant covariates were included.

2.5 Model Evaluation

The VPC is a simulation-based diagnostic tool in which the distribution of simulated outputs from the PK and PK/PD models are compared with the observed data, which informs on predictive properties of the model. The 5th, 50th, and 95th percentiles of the original data were binned on appropriate time intervals. Simulations (n = 1000) from the final PK model were used to calculate 95% confidence intervals (CIs) for the n simulated 5th, 50th, and 95th percentiles. The 5th, 50th, and 95th percentiles of the original data should fall within or close to these predicted limits. The prediction-corrected VPC was stratified by sampling intensity (intense/sparse) and age [43]. Parameter uncertainty was assessed through the relative standard error (%RSE).

3 Results

3.1 Demographics and Datasets

Mean patient age was 11 years, 47% were male, and 78% were Black (Table 1). Body weight range was wide (17–82 kg, quartile range 26–44 kg). The median (range) body mass index was 16.92 kg/m2 (12.1–28.4).

Table 1 Patient demographics and key covariate information
Full size table

Overall, 878 PK samples above the LLOQ were available from 45 pediatric patients following single and multiple doses of ticagrelor. One randomized patient had inadequate venous access and did not receive ticagrelor. Individual observed plasma concentration and time profiles following the two single doses are shown in Online Resource Figs. 2 and 3. Overall, 8.2% (78/956 samples) of concentration measurements were below the LLOQ, and were excluded from the population PK analysis (method 1 in the article by Beal [34]).

The PD data set comprised 341 PRU samples. Individual observed platelet inhibition profiles over time are depicted in Online Resource Fig. 4.

3.2 Final Population PK Model

The final developed population PK model (Table 2) could adequately describe the ticagrelor and AR-C124910XX plasma concentration and time data from HESTIA1 following single and repeated ticagrelor administration.

Table 2 Final PK and PK/PD model parameter estimates
Full size table

The developed population PK model for ticagrelor and AR-C124910XX was based on a previous population PK model in adults [19], which was further refined to describe HESTIA1 data. The structural model used (Fig. 1) consisted of a two-compartment, disposition model for ticagrelor and AR-C124910XX with linear elimination. An Erlang-type absorption model with a fixed number of five absorption transit compartments (added in a stepwise fashion) was utilized, and the first-order transfer-rate constant was estimated [44]. Absorption was rapid, with a mean transit time of 0.68 h. An allometric body weight relationship to ticagrelor and AR-C124910XX clearance and volume of distribution was used. Attempts to estimate the allometric exponent did not significantly improve the model fit. The exponents were estimated at 0.48 (95% CI 0.20–0.77) and 0.80 (95% CI 0.50–1.11) for clearance and volume, respectively (CIs were generated using log-likelihood profiling). However, the exponent values were fixed to 0.75 (clearance) and 1 (volume), in the final PK model, which is consistent with general principles [41, 42]. Estimated apparent oral clearances for ticagrelor (CL/F) and AR-C124910XX (CLm/F) were 22.8 and 9.97 L/h, respectively, for a 35-kg patient (median bodyweight in the study) (Table 2), and 12.1 and 5.28 L/h, respectively, for a 15-kg patient. Figure 2 depicts the allometric relationship between ticagrelor CL/F and body weight, while Fig. 3 shows the derived area under the concentration-time curve at steady state (AUCss) for ticagrelor and AR-C124910XX.

Fig. 2
figure 2

Individual estimated ticagrelor oral clearance (CL/F) versus body weight (final PK model). Open circles represent individual estimated oral clearance values with the final PK model in relation to individual body weight. The dashed line represents the population mean clearance prediction across body weights by the allometric function. PK pharmacokinetic

Full size image
Fig. 3
figure 3

Individual a ticagrelor and b AR-C124910XX AUCss versus body weight. The AUCss for ticagrelor and AR-C124910XX was derived by dividing the actual (mg) ticagrelor dose with the individual estimated ticagrelor apparent oral clearance. The graph includes more than one observation per patient as the actual dose administered on separate occasions was different. For AR-C124910XX, AUCss was derived by multiplying the ticagrelor dose by the fraction metabolized into AR-C124910XX (0.22) and dividing by the individual estimated AR-C124910XX apparent oral clearance. The outlying value in panel (a) was due to a low ticagrelor oral clearance in one patient who had taken a strong cytochrome P450 3A4 inhibitor (a protocol violation). AUCss area under the concentration-time curve at steady state

Full size image

Covariate searches identified sex and cholecystectomy as significant predictors for relative ticagrelor bioavailability (25.9% decrease in males) and central ticagrelor volume of distribution (29.6% increase in cholecystectomized patients), respectively (Table 2). However, the overall impact of both covariates on these parameters is low. None of the other investigated covariates (dose, age, race) significantly improved model fit.

The typical model parameters were, in general, estimated with good precision (Table 2). However, the larger %RSE (>30%) observed for BSV on oral AR-C124910XX clearance (CLm/F) and the covariate effect of cholecystectomy on ticagrelor central volume of distribution (Vc/F) indicate that these parameters were relatively uncertain, which may be attributed to the relatively sparse HESTIA1 PK data.

The VPC model diagnostic plots showed that the median and percentiles predicted from this final PK model were within the respective CIs calculated from the observed data following a single ticagrelor dose and at steady-state dosing (Fig. 4). GOF plots for the final PK model are shown in Online Resource Fig. 5. The GOF plots were, in general, featureless and random with a normal distribution, mean of zero, and a small standard deviation, with the absolute weighted residuals < 3. Collectively, these observations confirm that the developed population PK model adequately described ticagrelor and AR-C124910XX plasma concentration and time data in pediatric SCD patients.

Fig. 4
figure 4

Visual predictive check plots (single dose) showing suitability of the final population pharmacokinetic model for single-dose a ticagrelor and b AR-C124910XX, and for steady-state dosing for c ticagrelor and d AR-C124910XX. The crosses (+) represent observed data normalized to the population prediction in the VPC (prediction correction). The black solid line represents the median observed data and the dashed lines represent the observed 10th and 90th percentiles. The grey-shaded area represents the 95% confidence interval around the median and respective percentiles. VPC visual predictive check

Full size image
Fig. 5
figure 5

Visual predictive check plots of the final PK/PD model for a absolute PRU response, b relative PRU response, and for the final model age and study stratified for c single dose and d steady state. The crosses (+) represent observed data normalized to the population prediction in the VPC (prediction correction). The black solid line represents the median observed data and the dashed lines represent the observed 10th and 90th percentiles. The grey-shaded area represents the 95% confidence interval around the median and respective percentiles. PK/PD pharmacokinetic/pharmacodynamic, PRU platelet reaction unit, VPC visual predictive check

Full size image

3.3 Final Population PK/PD Model

The individual predicted ticagrelor plasma concentrations from the final population PK model were used to assess the effect on platelet inhibition using a population PK/PD model previously developed for adult CAD patients [19]. This model was further refined for the available pediatric data (Table 2). Baseline PRU was assumed to be normally distributed, with an estimated baseline PRU of 283. The ticagrelor concentration at half-maximum PRU effect (EC50) was estimated at 233 nmol/L. Emax was fixed in this model as it was assumed that full platelet inhibition was achievable, as supported by previous findings [5, 9, 10, 14, 16] and the present data. No significant covariates were identified during the population PK/PD model covariate search. Simulation-based diagnostics (VPCs) showed that the final population PK/PD model adequately described the time course and extent of PRU inhibition by ticagrelor across the observed dose range (Fig. 5). There were no observed differences in model performance after single or repeated ticagrelor dosing (Fig. 5). GOF plots for the final PK/PD model are shown in Online Resource Fig. 6. The exposure versus PRU response relationship for ticagrelor appeared to be predictable (Fig. 6).

Fig. 6
figure 6

Ticagrelor exposure versus a absolute and b relative platelet inhibition. Dashed line represents the typical value prediction for the PK/PD model. In panel b any relative PRU values < 0 were included as 0. PK/PD pharmacokinetic/pharmacodynamic, PRU platelet reaction unit

Full size image

4 Discussion

HESTIA1 was the first study to report the association between ticagrelor dose, observed exposure, and inhibition of platelet aggregation in pediatric SCD patients [31]. Our population modeling is the first quantitative approach, using HESTIA1 data, to characterize ticagrelor and AR-C124910XX PK, thus enabling an assessment of the dose-exposure-response relationship. The presented PK and PK/PD models could adequately describe the time course of ticagrelor and AR-C12490XX plasma concentrations in pediatric SCD patients, with an allometric scaling of clearance and volume parameters to body weight. Additionally, potential PK-model relevant covariate relationships were investigated (dose, age, sex, race, and medical history of cholecystectomy). The final population PK/PD model also adequately captured the quantitative relationship of ticagrelor plasma exposure with platelet inhibition in these patients.

Previous ticagrelor population PK approaches in adults were developed using a wealth of data from patients with acute coronary syndrome [17], those with prior MI [18], healthy volunteers, and CAD patients [19]. Encouragingly, our pediatric ticagrelor population PK model is in overall agreement with findings in adults with regard to, for example, identified covariates (body weight and sex [18]), as well as ticagrelor absorption and distribution [17].

Our population PK model incorporated an allometric relationship of body weight with ticagrelor and AR-C124910XX clearance and volume of distribution in pediatric SCD patients, with no additional improvement in the model data description by estimating allometric exponents. The presence of an allometric relationship was a key assumption in ticagrelor dose selection for HESTIA1. A prerequisite for estimating the shape of the allometric relationship is a sufficient distribution of observed body weights over a wide range. The body weight data in HESTIA1 (range 17–82 kg; quartile range 26–44 kg) was unable to support separating fixed from estimated allometric exponents. Hence, fixed exponents of 0.75 (clearance) and 1 (volume), in line with general PK scaling principles, were used and offered an adequate description of the data [42, 43]), although future ticagrelor studies with potentially more data in younger and smaller children will contribute to further refining the shape of these relationships. The allometric body weight relationship translates to lower ticagrelor clearance in patients with low body weight and vice versa. The estimated oral ticagrelor clearance was 22.8 L/h at 35 kg, the median body weight in HESTIA1; the extrapolated adult-equivalent (70 kg) oral ticagrelor clearance would be 38.4 L/h. While this extrapolated oral ticagrelor clearance is higher than the clearance in patients with acute coronary syndrome, i.e. 14 L/h [17], it is in line with that reported in healthy adult volunteers, i.e. 40.7 L/h after a single 180-mg ticagrelor dose [45]. A body weight dependency of ticagrelor and AR-C124910XX oral clearance has been confirmed in larger populations of adult post-MI patients with population PK modeling [18].

Estimated individual oral clearances for ticagrelor and AR-C124910XX were used to derive AUCss in HESTIA1. Despite the established body weight-dependent CL/F, and ticagrelor dosing by body weight (mg/kg) in HESTIA1, there was a trend toward a lower AUCss in lighter patients. This trend may be caused by the linear milligram/kilogram dosing concept being an overcompensation for the derived allometric body weight relationship on CL/F.

HESTIA1 patients were predominantly Black or African American as SCD disproportionately affects people of African descent versus other races [46]. This race imbalance may have limited the possibility to detect any covariate influence of race in the current population PK modeling. The adult population PK model in patients with acute coronary syndrome showed a 20% lower relative ticagrelor bioavailability in Black versus White patients [17]. However, this difference would only result in a minor increase in oral clearance in Black patients and was considered to be of limited clinical significance [17]. The current assessment of HESTIA1 data identified sex as a covariate for relative ticagrelor bioavailability, which appeared lower in males versus females. This difference indicates a higher oral clearance (approximately 26%) in boys versus girls. The relationship between sex and relative bioavailability identified in pediatric SCD patients is in line with adult data [18]. The exact mechanism for the small differences in ticagrelor oral clearance by sex is currently unknown. However, the influence of sex on ticagrelor exposure is also considered to be of limited clinical significance in pediatric SCD patients and the adult post-MI population.

We also explored covariates connected with the medical and surgical history of patients for potential impact on ticagrelor PK, such as cholecystectomy, which is commonly performed in SCD patients due to cholecystitis, and was conducted in 22% of children in HESTIA1. Ticagrelor and AR-C124910XX are excreted via the bile following metabolism in the liver [15]; hence, cholecystectomy as a potential covariate was deemed plausible. Cholecystectomy was identified in the covariate search for the ticagrelor central volume of distribution, which was increased by approximately 30% in patients with a cholecystectomy. However, this increase is not expected to have any significant clinical impact on the overall ticagrelor exposure in pediatric SCD patients. Given the predominant hepatic metabolism of ticagrelor [15], a history of severe liver sequestration, which may occur in SCD patients due to sequestration of sickled erythrocytes [20, 21], could be expected as a potential covariate impacting ticagrelor PK. However, as only one patient with this condition was enrolled in HESTIA1, this factor could not be included in the covariate search. Overall, the identified PK-model relevant covariates, even though significantly improving the model description of the data, are not expected to significantly impact the overall ticagrelor exposure in children with SCD to an extent that warrants dose adjustments.

Ticagrelor is mainly metabolized by the CYP3A4 enzyme [12, 15], which is expected to be fully mature in children ≥2 years of age [47]. Furthermore, no dose-dependent effect on oral ticagrelor clearance was observed in pediatric SCD patients over the relatively wide ticagrelor dose range (2–139.5 mg) in HESTIA1. This finding is in agreement with the linear PK of ticagrelor in adults [9, 10].

Overall, our developed population PK model sufficiently described ticagrelor PK in pediatric SCD patients, thus enabling its further use to characterize the exposure-response relationship for ticagrelor with platelet inhibition. The population PK/PD model used individual predicted ticagrelor concentrations to evaluate the time course and extent of PRU inhibition. The individual contribution to PRU inhibition by AR-C124910XX was not included in the population PK/PD model as the relative contribution of this metabolite to PRU inhibition is assumed to be constant. This assumption is supported by the observed parent-to-metabolite ratio (3:1), which appears to be similar in children/adolescents in HESTIA1, and in adults [14, 15].

Platelet inhibition in pediatric SCD patients occurred rapidly and was directly related to ticagrelor plasma concentrations, similar to findings in adults [5, 9, 10, 14, 16]. Observed PRU values returned toward baseline values at 6–8 h post-dose, in keeping with decreasing ticagrelor plasma concentrations, and reversible P2Y12 inhibiton as ticagrelor is excreted [5, 9, 10, 14, 16]. The wide ticagrelor dose range in HESTIA1 enabled exploration of the full platelet inhibition profile from baseline to near full PRU inhibition, thus facilitating the characterization of the full exposure-response relationship. We demonstrated that the developed population PK/PD model was able to adequately describe the time course and extent of platelet inhibition after both single and repeated ticagrelor doses. Moreover, the exposure-response relationship in pediatric SCD patients appears similar to that in adult healthy volunteers and CAD patients [8]. Our results are supported by in vitro results of platelet reactivity, suggesting that, at equivalent levels of ticagrelor exposure, children and adults would have a comparable antiplatelet response [48]. Thus, it appears that platelets in children with SCD do not react differently to ticagrelor compared with adults.

The maximum PRU effect (Emax) was fixed to 1 in our base population PK/PD model, assuming that ticagrelor can achieve full platelet inhibition. No significant covariates were identified in the population PK/PD model development. Baseline PRU was estimated at 283, and the ticagrelor EC50 was 233 nmol/L. Results suggest that ticagrelor doses from 0.75 mg/kg twice daily and higher would provide >50% PRU inhibition during the full 12-h dosing interval in pediatric SCD patients.

A relationship between high levels of PRU inhibition (absolute PRU values <85) and an increased bleeding risk has been reported in CAD patients receiving dual antiplatelet treatment [49]. Our analyses intended to further investigate this relationship in children/adolescents with SCD; however, since no bleeding events were reported in HESTIA1 [31], a further analysis of ticagrelor exposure and/or platelet inhibition to any bleeding events in pediatric SCD patients was not possible. The long-term safety profile of ticagrelor has yet to be determined in this patient population. A population PK model for prasugrel was recently developed in pediatric SCD patients, showing no relationship between exposure to the active metabolite of prasugrel and efficacy and safety endpoints [50]; however, a positive trend was observed between vaso-occlusive crises in the DOVE study at a modest PRU inhibition [29].

Inhibition of platelet activation has a clear physiological rationale as a potential SCD treatment, as platelet activation is increased in SCD [25, 26]. Several studies have evaluated antiplatelet agents, such as ticlopidine [28], prasugrel [29], and ticagrelor [51], and the P-selectin (a factor released from activated platelets) monoclonal antibody crizanlizumab [52] in SCD patients, focusing on clinical endpoints. However, to date, no antiplatelet agent has significantly impacted any clinical endpoint evaluated in SCD, possibly due to the complexity of this disease. HESTIA1 was designed to explore the PK/PD relationship across a dose range of ticagrelor, and the limited patient numbers and treatment duration did not allow for evaluation of the potential effects on clinical endpoints [31].

5 Conclusions

Our pharmacometric approach to describe the ticagrelor population PK and population PK/PD relationship in pediatric SCD patients is a valuable addition to the limited data in this clinical setting, and is supported by previous observations in adults. The presented modeling exercise offers the first quantitative characterization of the ticagrelor dose-exposure-response relationship, and included an allometric relationship to body weight for oral clearance and volume of distribution. Ultimately, we hope that these data can improve the understanding of ticagrelor use in the pediatric SCD setting, thus enabling predictive and well-informed study design and dose selection for potential future studies to establish the safety and efficacy of ticagrelor in this patient population.