Eltrombopag does not affect cardiac repolarization: results from a definitive QT_c study in healthy subjects

Abstract

AIM

To evaluate the effect of eltrombopag on cardiac repolarization and to characterize the relationship between plasma eltrombopag concentrations and change in QT_c.

METHODS

This was a double-blind, placebo- and active-controlled, randomized, balanced four-period, crossover study in healthy men and women. Subjects were randomized to receive eltrombopag 50 mg and 150 mg, moxifloxacin 400 mg (positive control) and placebo in one of four sequences.

RESULTS

Eighty-seven subjects entered the study and 48 completed. There was no prolongation of QT_c (Fridericia) following eltrombopag treatment, as the upper limit of the 90% confidence interval (CI) for the time-matched change from baseline in QT_cF between drug and placebo (ddQT_cF) did not exceed 10 ms for eltrombopag at either dose. Maximum observed mean treatment difference was 2.29 ms (90% CI 0.34, 4.24) for eltrombopag 150 mg at 1 h post-dose and 11.64 ms (90% CI 9.64, 13.64) for moxifloxacin 400 mg at 4 h. Eltrombopag C_max and AUC(0,24 h) increased in a dose proportional manner between 50 mg and 150 mg after 5 days' dosing. Proportions of subjects with adverse events were similar across treatments (52–66% of subjects). Most withdrawals (26/39 subjects) were due to elevated platelets. Three subjects were withdrawn for ventricular premature beats (one following each active treatment) reported as related to the study drug.

CONCLUSIONS

No clinically significant QT_c prolongation was observed for eltrombopag at therapeutic and supratherapeutic doses.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT

Some non-anti-arrhythmic drugs delay cardiac repolarization, which can be measured as an increase in the QT interval. Delays in cardiac repolarization create an electrophysiological environment that favours the development of cardiac arrhythmias, which may lead to torsade de pointes, which can be fatal. As part of the clinical development of eltrombopag, a thorough QT_c study was conducted to evaluate the effects of eltrombopag on cardiac repolarization at both therapeutic and supratherapeutic doses and to characterize the relationship between plasma eltrombopag concentrations and change in QT_c.

WHAT THIS STUDY ADDS

This study found no clinically significant QT prolongation for eltrombopag when administered as 50 mg or 150 mg every day for 5 days. There were no safety or tolerability signals of clinical concern. A small incidence of ventricular premature beats was observed, but this was consistent with previously reported incidences in healthy volunteers without apparent heart disease.

Introduction

Eltrombopag (Promacta®; SB-497115) is the first oral, small molecule, non-peptide thrombopoietin receptor agonist to be approved for the treatment of chronic idiopathic thrombocytopenic purpura. Eltrombopag selectively interacts with the thrombopoietin receptor (a transmembrane receptor), activating signal transduction pathways that induce proliferation and differentiation of cells in the megakaryocytic lineage resulting in increased platelet counts [1]. Previous studies of eltrombopag have shown it to increase platelet counts in a dose-dependent manner in healthy volunteers [2] and in patients with relapsed or refractory chronic idiopathic thrombocytopenic purpura [3]. Eltrombopag has also been shown to increase platelet counts in patients with thrombocytopenia associated with chronic hepatitis C, allowing the initiation of antiviral therapy [4].

Some non-anti-arrhythmic drugs delay cardiac repolarization [5, 6]. This can be measured as an increase in the QT interval, which represents the duration of ventricular depolarization and subsequent repolarization. Delays in cardiac repolarization create an electrophysiological environment that favours the development of cardiac arrhythmias, which may lead to torsade de pointes, which can be fatal. Many drugs have been withdrawn from the market or undergone labelling changes because of their tendency to prolong QT intervals [5, 6]. Consequently, the emphasis placed on the assessment of QT and heart rate corrected QT (QT_c) effects has increased in recent years for drugs in development.

In an in vitro study to measure the effect of eltrombopag on human ether-a-go-go related gene (hERG) currents recorded from human embryonic kidney (HEK-293) cells stably transfected with hERG-1 cDNA, eltrombopag was found to inhibit hERG channel tail current in a concentration-dependent manner with an estimated IC₅₀ of 0.69 µm (0.31 µg ml⁻¹) (data on file). Although in vitro electrophysiology findings suggested potential for cardiac conduction abnormality, including QT_c prolongation, there was no in vivo evidence of electrocardiogram (ECG) abnormalities in non-clinical studies. Therefore, as part of the clinical development of eltrombopag, a thorough QT_c study in healthy volunteers was conducted to evaluate the effects of eltrombopag on cardiac repolarization at both therapeutic and supratherapeutic doses and to characterize the relationship between plasma eltrombopag concentrations and change in QT_c. The study was designed and conducted according to Food and Drug Administration guidance for industry on the clinical evaluation of QT_c interval prolongation and pro-arrhythmic potential for non-arrhythmic drugs [7] and utilized a placebo and active control.

Methods

Subjects

The study population consisted of healthy men and women aged 18–50 years with a body mass index (BMI) ≥19 and ≤30 kg m⁻². Body weight requirement was ≥50 kg for men and ≥45 kg for women. Women were of non-child-bearing potential or agreed to protocol-specified methods of contraception. All subjects provided written, informed consent. Subjects were excluded if they had protocol-specified ECG abnormalities, a history of protocol-specified cardiovascular abnormalities, abnormalities in platelet number and function or abnormalities in clotting mechanisms. There were also standard exclusion criteria concerning alcohol/drug abuse (including tobacco) and prior blood donation, recent use of aspirin, non-steroidal anti-inflammatory drugs and antacids, and concomitant medications that were known to prolong the QT_c interval. Subjects with a history of sensitivity to moxifloxacin or any quinolone were ineligible for the study, as were subjects with abnormal laboratory parameters.

The study protocol and amendments were reviewed and approved by the Research Consultants' Review Committee Institutional Review Board (Austin Texas, USA), the Aspire Independent Review Board (San Diego, CA, USA) and the Covance Clinical Research Unit Institutional Review Board (Madison, WI, USA). The study was conducted in accordance with good clinical practice and according to the guiding principles of the Declaration of Helsinki.

Study design

This was a double-blind, placebo and active (moxifloxacin) controlled, randomized, balanced crossover study to evaluate the effect of eltrombopag on cardiac repolarization. Four treatment regimens were administered: (i) eltrombopag 50 mg once daily for 5 days with a dose of placebo to match moxifloxacin on the final day of dosing (day 5), (ii) eltrombopag 150 mg once daily for 5 days with placebo for moxifloxacin on day 5, (iii) placebo to match eltrombopag once daily for 5 days with placebo for moxifloxacin on day 5, and (iv) placebo for eltrombopag once daily for 5 days with moxifloxacin 400 mg on day 5. Subjects were randomized to one of four sequences to receive each of the four regimens in a crossover fashion, using a Williams square design (Figure 1). All study drugs were taken with 240 ml of water and at least 2 h before or after food intake; on day 5 of each period, subjects fasted overnight prior to dosing and for an additional 4 h after dosing. Subjects stayed at the clinical research unit from day −2 to the morning of day 6 of each treatment period. There were washout periods of at least 14 days between each treatment period. The total duration of each subject's participation in the study, from screening to follow-up, was a maximum of 18 weeks.

Figure 1.

Subject disposition diagram. A = eltrombopag 50 mg once daily for 5 days plus moxifloxacin placebo on day 5; B = eltrombopag 150 mg once daily for 5 days plus moxifloxacin placebo on day 5; C = eltrombopag placebo once daily for 5 days plus moxifloxacin placebo on day 5; D = eltrombopag placebo once daily for 5 days plus moxifloxacin 400 mg on day 5

Electrocardiographic methods

Holter monitoring was used for QT_c analysis. Time-matched ECGs were collected in triplicate at the following time points, using the planned dosing time on day −1 and the actual dosing time on day 5 as starting points: 0.5 h pre-dose, and 0.5, 1, 2, 3, 4, 6, 12 and 23.25 h post-dose. In addition to Holter monitoring, 12-lead ECGs for standard assessment of safety were measured pre-dose, at 1, 2, 3, 4, 6 h, and 24 h post-dose on days 1, 2, 3, 4 and 5. Telemetry was conducted from at least 6 h pre-dose on day 1 until at least 24 h after the last dose on day 5.

A central laboratory was employed to measure the QT interval. Measurers were blinded to the treatment and QT intervals were measured automatically with manual adjustment. Lead II was the primary lead for measurement with lead V5 as the secondary lead for measurement. In instances of quality issues or unstable heart rate a tertiary lead was used for measurement. It was the central laboratory's standard practice to measure the QT interval on a consistent lead within a subject across multiple visit days except when instances of quality or unstable heart rate existed. Three consecutive ECG complexes were measured. To assess intra- or inter-observer variation for QT interval measurements, 2% inter-reader was performed. The central laboratory standard intra-reader is an annual program where skilled technicians are required to re-measure an identical set of 50 ECGs previously measured for routine studies.

Pharmacokinetic methods

Serial blood samples were collected via the forearm vein into ethylenediaminetetraacetic acid (EDTA) anti-coagulation tubes for the determination of drug concentrations in plasma at 0.5 h pre-dose and at 0.5, 1, 1.5, 2, 2.5, 3, 4, 6, 8, 12, 24 h post-dose on day 5 of each period. Collected samples were immediately chilled on crushed ice water. Plasma was separated within 1 h of sample collection by centrifugation (at 1500 g) for 10 min at 4 °C. Plasma samples were transferred to polypropylene tubes and stored frozen at −20°C until analysis. Plasma eltrombopag and moxifloxacin concentrations were determined by GlaxoSmithKline (King of Prussia, PA, USA) using validated analysis methods. Eltrombopag and moxifloxacin were extracted from human plasma by protein precipitation. Extracts were analyzed by high-performance liquid chromatography with tandem mass spectrometry detection. The assays were validated over the eltrombopag concentration range of 100–50 000 ng ml⁻¹ and over the moxifloxacin concentration range of 25–5000 ng ml⁻¹ in human plasma.

Pharmacokinetic parameters were calculated from the eltrombopag and moxifloxacin plasma concentrations, separately, using the non-compartmental Model 200 of WinNonlin Professional Edition version 4.1. Calculations were based on actual collection times recorded during the study. The following repeat-dose eltrombopag pharmacokinetic parameters were estimated on day 5 of the eltrombopag 50 mg and 150 mg once daily treatments: maximum concentration (C_max), time to C_max (t_max), area under the plasma concentration curve over the dosing interval (AUC(0,24 h)), and the concentration at the end of the dosing interval (C_τ). The following single dose moxifloxacin pharmacokinetic parameters were estimated for the moxifloxacin 400 mg treatment: C_max, t_max, and AUC(0,24 h).

Safety methods

Adverse events were recorded from screening to the final follow-up. Standard clinical laboratory and vital signs assessments were conducted at pre-specified time points throughout the study.

Elevated platelets were not captured as adverse events because this was an expected result of eltrombopag treatment. To avoid thrombocytosis, subjects were discontinued from the study if they achieved platelet counts >800 × 10⁹ µl⁻¹ or if platelet counts went >400 × 10⁹ µl⁻¹ following dosing and remained >350 × 10⁹ µl⁻¹ on day −2 of the next period. It was planned that any subject with a QT_c value >500 ms during the study was to be withdrawn (although no subject was actually withdrawn for this reason).

Statistical methods

Power calculation and study population

The primary endpoint in this study was change from baseline in QT_c interval adjusted according to Fridericia's correction (dQT_cF) between drug and placebo (ddQT_cF) at each time point (average of triplicate ECGs). A sufficient number of subjects were enrolled to ensure at least 40 subjects completed the study. Based on an assumed true mean difference of 0 ms in dQT_cF between drug and placebo and a within-subject variability estimate of 10 ms, 40 subjects would provide at least 99% power to demonstrate a lack of effect on the QT_c interval at each time point. A lack of effect on the QT_cF interval was pre-specified as an upper 90% confidence interval (CI) value of ddQT_cF was less than 10 ms [7, 8]. The power calculation was based on a non-inferiority testing procedure in 2 × 2 crossover design and PASS 2005 was used to get an appropriate sample size. A power of 99% at each time point allowed the overall power to be retained at least 90% [∼ (0.99)⁹]. Subjects who received at least one dose of any or all of the four treatment regimens were included in the summary of safety data. Subjects who provided pharmacokinetic data for at least one treatment were included in the summary of the pharmacokinetic data. Subjects who completed the placebo treatment and at least one of the other three treatment regimens were included in the QT_c analysis; of these subjects, those who received eltrombopag were included in the characterization of the relationship between plasma eltrombopag concentrations and change in QT_c.

Electrocardiographic analysis

The primary endpoint was analyzed by mixed effects analysis of covariance (ancova) fitting terms appropriate to the study design, including sequence, period, regimen, time, and time-by-regimen interaction as fixed effects and subject within sequence as a random effect. Baseline QT_cF was included in the model as a covariate. Point estimates and 90% CIs were constructed for the difference, active vs. placebo, for eltrombopag 50 mg, eltrombopag 150 mg, and moxifloxacin 400 mg at each time point using the residual variance.

Secondary endpoints (change from baseline in QT_c adjusted using Bazett's correction [QT_cB] and change from baseline in QT_c adjusted using the individualized correction [QT_cI]) were analyzed in a similar way. Distributional assumptions underlying the analyses were assessed by residual plots. Homogeneity of variance was assessed by plotting the studentized residuals against the predicted values from the model, whilst normality was examined by normal probability plots. If assumptions were grossly violated, alternative approaches were to be considered. An outlier analysis was done to determine the number and percentage of subjects in which an increase from baseline in QT_c >30 ms and >60 ms occurred for each regimen.

Pharmacokinetic analysis

Dose proportionality was assessed for eltrombopag 50 mg and 150 mg using a mixed effect analysis of variance model to give point estimates and 90% CIs for the ratios of the geometric means for the comparisons of interest.

Analysis of relationship between plasma eltrombopag exposure and change in QT_cF and subsequent simulations

A correlation analysis was conducted between individual maximum change from baseline in QTcF and plasma eltrombopag AUC(0,τ) and C_max.

In addition to the statistical analysis, a population pharmacokinetic/pharmacodynamic analysis was completed to evaluate the relationship between ddQT_cF and plasma eltrombopag concentration. Exploratory graphical displays assessed the shape of the potential relationship between plasma eltrombopag concentrations (Cp) and ddQT_cF, and a linear relationship with no delay in the effect of concentration on ddQT_cF was selected as the base model. The analysis was performed using a mixed-effects modelling approach with the first-order conditional method (FOCE-INTER) of NONMEM Version 5.

The base model was: ddQT_cF = θ₁ + η₁ + (θ₂ + η₂) × Cp + ε₁; where θ₁ was the typical value of pre-dose ddQT_c on day 5 (intercept) and θ₂ was the typical value of slope relating plasma drug concentration to ddQT_c, η₁ was the inter-subject variability for pre-dose ddQT_c and η₂ was the inter-subject variability for the slope, Cp was the plasma drug concentration and ε₁ was within-subject residual variability. Because subjects received two doses of eltrombopag in separate study periods, inter-occasion variability was added for both intercept and slope and significantly improved the model fit. Following acceptance of the final model, covariates were tested. Demographic factors, including sex, ethnicity, race, age, weight and BMI were plotted against η to assess potential covariates for inclusion in the Cp-ddQT_cF analysis.

The final Cp-ddQT_cF model was evaluated using the non-parametric bootstrapping procedure, in which bootstrap samples were generated by random sampling with replacement from the original dataset. A total of 500 bootstrap runs successfully converged. Median parameter estimates and their 5th and 95th percentiles were calculated and compared with those obtained from the original dataset. This was done to affirm that the estimated model parameters were consistent with the data.

Based on the final Cp-ddQT_cF model, simulations were performed to predict the mean (90% CI) for ddQT_cF at C_max for therapeutic and supratherapeutic eltrombopag doses of 50 mg once daily, 150 mg once daily and 300 mg once daily. Dose proportionality and constant coefficient of variation for C_max were assumed for extrapolation to doses that were not studied (eltrombopag 300 mg once daily). Simulations of 1000 replicate studies were generated for each dose level. The mean ddQT_cF was calculated for each study. The 5th and 95th percentiles of the distribution of study means were used to estimate the 90% CI.

Safety analysis

No formal statistical comparison was conducted for safety assessment. Adverse events were coded using the Medical Dictionary for Regulatory Activities.

Results

Subject disposition and demographics

Of the 87 subjects enrolled, 64 received at least one dose of placebo, 62 received at least one dose of eltrombopag 50 mg, 77 received at least one dose of eltrombopag 150 mg and 63 received a single dose of moxifloxacin 400 mg. Forty-eight (55%) subjects completed the study (Figure 1). Twenty-six subjects were withdrawn for elevated platelet counts. Thirteen other subjects were withdrawn, including three for non-cardiac adverse events (increased eosinophil count and two cases of gingival pain), three for cardiac adverse events (ventricular tachycardia and two cases of ventricular extrasystoles), one for nonadherence and six for protocol violations. The study population comprised 62 men and 25 women. Mean (SD) age was 29.9 (9.17) years and median (range) BMI was 25.8 (20.2–30.0) kg m⁻².

Electrocardiographic results

Three correction methods were compared using a correlation analysis using QT_c and inter-beat interval (RR) values to assess the appropriateness of QT_c for each correction method. The Fridericia correction (QT_cF) demonstrated slight overcorrection at lower RR values and slight undercorrection at higher RR values. The Bazett's correction (QT_cB) and individualized correction (QT_ci) demonstrated more striking correction effect than the Fridericia's correction. The Bazett's correction demonstrated undercorrection at lower RR values and overcorrection at higher RR values, while the individualized correction demonstrated the opposite effect. The Pearson's correlations (P value) between QT_cF and RR by treatment are presented in (Table 1). The correlations from QT_cF were all statistically non-significant, while those from QT_cB showed slightly negative and those from QT_ci showed slightly positive correlations for each treatment. Although these three corrections showed different correction profiles, all three corrections were adequate because the linear relationship between QT_c and RR intervals became negligible after the RR correction, as reflected by the low Pearson correlation coefficients (Table 1).

Table 1.

Summary of Pearson's correlation between QT_c and RR by treatment (all data)

			QT correction method
Treatment		n	QTcF (Fridericia)	QTcB (Bazett)	QTci
Eltrombopag 50 mg once daily for 5 days	Coefficient	912	0.03998	−0.40321	0.05863
	P value		0.2278	<0001	0.0768
Eltrombopag 150 mg once daily for 5 days	Coefficient	1031	0.01860	−0.49434	0.03846
	P value		0.5507	<0001	0.2172
Moxifloxacin 400 mg single dose	Coefficient	943	0.03461	−0.39537	0.07866
	P value		0.2884	<0001	0.0157
Placebo	Coefficient	1107	0.00570	−0.048764	0.04674
	P value		0.8498	<0001	0.1202

Eltrombopag did not cause any clinically significant prolongation of the primary endpoint QT_cF, as the upper limit for the 90% CI of ddQT_cF did not exceed 10 ms at any time point from dosing to 23.25 h post-dose (Figure 2A). The greatest treatment difference from placebo for mean QT_cF was at 2 h post-dose for the eltrombopag 50 mg dose (difference: 1.54 ms; 90% CI −0.47, 3.55) and at 1 h post-dose for the eltrombopag 150 mg dose (difference: 2.29 ms; 90% CI 0.34, 4.24). The greatest difference from placebo for moxifloxacin 400 mg was 11.64 ms (90% CI 9.64, 13.64) at 4 h post-dose. There was no clinically significant prolongation of the QT_cB (Figure 2B) and QT_cI (Figure 2C) intervals at either dose of eltrombopag. The lower limit of the 90% CI of ddQT_cF (Figure 2A), ddQT_cB (Figure 2B) and ddQT_cI (Figure 2C) exceeded 5 ms for moxifloxacin 400 mg at multiple time points.

Figure 2.

A) Plot of model-adjusted ddQT_cF vs. time for eltrombopag 50 mg (n = 52) (▵), eltrombopag 150 mg (n = 59) (◊) and moxifloxacin 400 mg (n = 53) (□). B) Plot of model-adjusted ddQT_cB vs. time for eltrombopag 50 mg (n = 52) (▵), eltrombopag 150 mg (n = 59) (◊) and moxifloxacin 400 mg (n = 53) (□). C) Plot of model-adjusted ddQT_cI vs. time for eltrombopag 50 mg (n = 52) (▵), eltrombopag 150 mg (n = 59) (◊) and moxifloxacin 400 mg (n = 53) (□)

No subject experienced a QT_cF or _QTcI value that exceeded 450 ms. A total of eight subjects had QT_cB values ranging from 450 to 480 ms: three subjects each for placebo and moxifloxacin 400 mg and one subject each for eltrombopag 50 mg and 150 mg. Individual changes from baseline in QT_cF were less than 30 ms for all four treatments. For QT_cB and QT_cI, most changes from baseline were less than 30 ms and all were less than 60 ms. Fourteen subjects who received moxifloxacin 400 mg, four subjects who received eltrombopag (50 mg or 150 mg) and two subjects who received placebo experienced QT_cB changes from baseline of 30–60 ms. One subject who received moxifloxacin 400 mg experienced a QT_cI change from baseline of 30–60 ms. None of the subjects experienced a change from baseline greater than 60 ms for any QT_c.

Pharmacokinetics results

Plasma eltrombopag concentrations were quantifiable over the entire dosing interval for all doses after repeat dose administration for 5 days. Moxifloxacin concentrations were quantifiable within 0.5 h and remained quantifiable through the 24 h sampling period after single dose administration.

Plasma eltrombopag AUC(0,24 h) and C_max increased in a dose proportional manner between the 50 mg and 150 mg dose levels after 5 days of dosing, with values of AUC(0,24 h) and C_max approximately three-fold higher for eltrombopag 150 mg compared with 50 mg (Table 2). Estimated dose proportionality ratios (90% CI) were 1.04 (0.987, 1.09) for AUC(0,24 h) and 1.01 (0.942, 1.08) for C_max, where a ratio of 1.0 represents dose proportionality.

Table 2.

Derived pharmacokinetic parameters following 5 days of repeat dosing of eltrombopag 50 mg and 150 mg once daily and following single dose administration of moxifloxacin 400 mg

	Geometric mean (95% confidence interval)
	Eltrombopag 50 mg	Eltrombopag 150 mg	Moxifloxacin 400 mg
Parameter	(n = 60)	(n = 73)	(n = 60)
AUC(0,24 h) (µg ml−1 h)*	65.4 (59.7, 71.6)	204 (186, 223)	22.6 (21.4, 23.9)
Cmax; (µg ml−1)	6.40 (5.87, 6.97)	19.0 (17.4, 20.6)	2.05 (1.93, 2.18)
Cτ (µg ml−1)	1.19 (1.05, 1.34)	4.07 (3.64, 4.55)	–
tmax (h)†	3.19 [2.17–6.22]	2.67 [1.67–6.20]	2.17 [0.63–6.17]

^*Presented as area under the plasma concentration curve from time 0 to 24 h post-dose for single dose moxifloxacin 400 mg.

^†Data presented as median [range]. AUC(0,24 h), area under the plasma concentration curve over the dosing interval; C_max, maximum plasma concentration, C_τ, concentration at the end of the dosing interval; t_max, time to C_max.

Results of assessment of relationship between plasma eltrombopag exposure and change in QT_cF and subsequent simulations

Based on statistical analysis, there was no correlation between individual maximum change from baseline in QT_cF and plasma eltrombopag AUC(0,τ) or C_max, where the correlation coefficients were −0.0161 (P = 0.8544) and −0.0080 (P = 0.9273), respectively.

In addition to the statistical analysis, a population pharmacokinetic/pharmacodynamic analysis was completed to evaluate the relationship between ddQT_cF and plasma eltrombopag concentration. For the eltrombopag treatment, ddQT_cF values ranged from −44 ms to +37 ms over the 24 h sampling on day 5 and plasma eltrombopag concentrations included in the analysis ranged from 0.224 to 12.2 µg ml⁻¹ for the 50 mg dose and from 0.841 to 31.4 µg ml⁻¹ for the 150 mg dose (Figure 3).

Figure 3.

Plot of ddQT_cF vs. plasma concentration following repeat dose administration of eltrombopag 50 mg once daily and 150 mg once daily for 5 days

The final model was a linear model with no delay in effect of Cp on ddQT_cF; fixed effects for pre-dose ddQT_cF on day 5 (intercept, θ1) and the slope relating plasma eltrombopag concentration to ddQT_cF (θ2) were included, along with inter-individual variability and inter-occasion variability (TRT1 = 50 mg and TRT2 = 150 mg) for both fixed effects, and additive random residual variability (ε1), as defined by the following equation:

The intercept (θ₁) estimate was close to zero (i.e. no effect on ddQT_cF when eltrombopag concentration is equal to zero) with a mean estimate of −0.792 ms (relative standard error (RSE) = 138%), with an inter-occasion variability estimate of 12.1 (RSE = 52%), much less than the estimated inter-individual variability of 32.6 (RSE = 24%). The linear relationship between ddQT_cF and plasma eltrombopag concentration was flat, with a model estimated slope (θ₂) value of 0.120 ms µg⁻¹ ml⁻¹ (RSE = 63%). This is in agreement with the 90% CI obtained from the bootstrap analysis for the slope estimate (−0.014, 0.244 ms µg⁻¹ ml⁻¹) that contained zero. The inter-occasion variability estimate for the slope (0.0392, RSE = 263%) was much greater than that of the inter-individual variability estimate (0.0065, RSE = 1718%), implying that the randomness of the ddQT_cF was largely independent of the change in plasma eltrombopag concentrations. Based on the graphical displays, sex and BMI were added as covariates in the model, but neither was significant and therefore they were not retained.

Based on the final Cp-ddQT_cF model, simulations were performed to predict ddQT_cF at C_max for therapeutic and supratherapeutic eltrombopag doses. Simulations of ddQT_cF at C_max for eltrombopag 50 mg and 150 mg once daily were consistent with the observed data. For example, for the eltrombopag 150 mg once daily regimen, the predicted mean (90% CI) ddQT_cF of 1.60 ms (−0.50, 4.03 ms) (Table 3) was consistent with the observed mean ddQT_cF values of −1.64 ms to 2.29 ms across the time points, and all upper limits of the 90% CIs were <5 ms (Figure 2A). Simulations of ddQT_cF at C_max predicted that eltrombopag would not have a clinically significant effect on ddQT_cF at concentrations predicted for a dose up to 300 mg once daily (Table 3).

Table 3.

Simulation results for ddQT_cF at C_max for therapeutic and supratherapeutic eltrombopag doses

	Plasma eltrombopag Cmax (µg ml−1)	Predicted ddQTcF (ms)
Eltrombopag dose	Mean (95% confidence interval)	Mean (90% confidence interval)*
50 mg once daily	6.72 (6.35, 7.10)	0.02 (−1.92, 2.42)
150 mg once daily	20.2 (19.0, 21.3)	1.60 (−0.50, 4.03)
300 mg once daily†	40.3 (38.1, 42.6)	4.03 (1.55, 6.79)

^*Based on 1000 study simulations per dose level (n = 60 subjects for 50 mg, n = 73 subjects for 150 mg, n = 81 subjects for 300 mg per simulation).

^†Simulations extrapolated beyond range of observed data; dose proportionality and constant coefficient of variation assumed. C_max, maximum plasma concentration; ddQT_cF, time-matched change from baseline in QT_cF between drug and placebo.

Safety results

There were no deaths or serious adverse events. Six subjects were withdrawn from the study because of adverse events. One of these subjects was withdrawn after experiencing a 4 beat non-sustained ventricular tachycardia and a 5 beat non-sustained ventricular tachycardia in the third treatment period (after 4 days of eltrombopag 150 mg dosing). The subject had previously completed treatment with eltrombopag 50 mg and moxifloxacin 400 mg in periods 1 and 2, respectively. The subject was asymptomatic. Both events were reported as drug-related. No abnormal findings were recorded on ECG (taken at 1, 2, 3, 4, 6, and 24 h post-dose on days 4 and 5 of treatment period 3) for this subject on or around the time these events were observed on telemetry. The subject had no prior history of cardiac arrhythmia. The maximum change from baseline QT_cF in period 3 for this subject was 13 ms.

Two subjects experienced premature ventricular complexes/ventricular extrasystoles after receiving eltrombopag. Neither of these subjects had a prior history of arrhythmia. The first subject had completed treatment period 1 (eltrombopag 150 mg) and had received one dose in treatment period 2 (eltrombopag 50 mg), when he experienced 138 multifocal premature ventricular complexes/ventricular extrasystoles on telemetry over a 68 h time period. The subject was asymptomatic and the events were reported as related to study drug. The subject was withdrawn from the study because of this adverse event. In this subject, an abnormal ECG had shown flat T-waves 4 h after study drug administration in treatment period 1 (eltrombopag 150 mg). An ECG also showed sinus bradycardia on the first day of dosing (pre-dose and 1 h post-dose) in treatment period 2 (eltrombopag 50 mg) and on the following day. The maximum change from baseline QT_cF in period 1 for this subject was 2 ms. The second subject had completed treatment period 1 with eltrombopag 50 mg. In period 2 (moxifloxacin 400 mg), prior to receiving the third dose of eltrombopag placebo, a total of 87 premature ventricular complexes/ventricular extrasystoles were observed on telemetry over a 66 h period. These were also reported as related to study drug, and the subject was withdrawn from the study because of the adverse event. There were no abnormal ECG findings (taken at 1, 2, 3, 4, 6, and 24 h post-dose on day 3 of period 2) or telemetry findings around the time of these events. The maximum change from baseline QT_cF in period 1 for this subject was 1 ms.

Two subjects withdrew because of mild gingival pain accompanied by tooth abscess, and gingivitis and oral discharge (one reported to be drug-related and one not related). Both subjects were withdrawn after a single dose of eltrombopag (50 mg and 150 mg, respectively) during treatment period 3, but had previously received moxifloxacin 400 mg (in periods 1 and 2, respectively). One subject was withdrawn at the end of the final treatment period (having received all four study treatments) for a moderate increase in eosinophil count reported as drug related. The adverse event onset was prior to dosing in period 4 (eltrombopag 50 mg once daily for 5 days was the previous regimen that the subject had received, in period 3).

In addition to the six subjects withdrawn for adverse events, 26 subjects were withdrawn for the pre-defined stopping criterion of elevated platelet counts (>400 × 10⁹ l⁻¹): 24 subjects after receiving eltrombopag 150 mg (range 401–670 × 10⁹ l⁻¹) and one each after eltrombopag 50 mg (438 × 10⁹ l⁻¹) and moxifloxacin 400 mg (451 × 10⁹ l⁻¹).

The proportions of subjects who experienced adverse events were similar across treatments (range 52%–66% of subjects) and most were mild in intensity. The most frequently reported events were related to placement of electrodes followed by headache, which occurred at a similar frequency in all treatment groups (11–13%). Other than subjects with elevated platelet counts, there were no notable treatment-related changes in laboratory parameters over the course of the study. There were also no notable changes in vital signs. The incidence of abnormalities in ECG findings was similar for both eltrombopag doses (21–22%), placebo (23%) and moxifloxacin (16%). Among these abnormal ECG findings, a second degree atrioventricular block was observed in a total of seven subjects after receiving eltrombopag 50 mg (four, 6%), eltrombopag 150 mg (one, 1%) and moxifloxacin 400 mg (two, 3%). All subjects were asymptomatic although it was reported as related to study drug.

Discussion

This study found no clinically significant QT prolongation for eltrombopag when administered as 50 mg or 150 mg once daily for 5 days. The finding of no clinically significant QT prolongation can be concluded, as values of the upper limit of the 90% CI for ddQT_cF did not exceed the pre-specified limit of 10 ms. The greatest observed mean treatment difference from placebo in QT prolongation was 2.29 ms for eltrombopag 150 mg once daily, and the upper bound of the 90% CI was 4.24 ms, which is lower than the pre-specified 10 ms. The study was sensitive enough to detect the effect of moxifloxacin.

Of the 87 enrolled subjects, 26 subjects (30%) withdrew due to protocol-defined stopping criteria for elevated platelets, and the majority of these subjects (24 of 26) withdrew following eltrombopag 150 mg once daily. Because platelet count elevations are delayed relative to the start of dosing (with peak platelet response occurring approximately 2 weeks after initiating dosing), subjects completed the treatment period assessments prior to withdrawing from the study. Therefore, the high withdrawal rate due to elevated platelet counts was not expected to bias the overall study assessment of QT_c. This lack of bias was supported by the similar change from baseline in QT_cF for the eltrombopag 150 mg treatment for 18 subjects who prematurely withdrew from the study prior to receiving placebo (mean change from baseline QT_cF ranged from −1.2 to −7.5 across the time points) and for 59 subjects who completed both the eltrombopag 150 mg and placebo treatments (mean change from baseline QT_cF ranged from −1.1 to −7.1 across the time points).

The population pharmacokinetic/pharmacodynamic analysis evaluating the relationship between ddQT_cF and plasma eltrombopag concentration was developed in order to support the statistical ddQT_cF analysis and to allow simulation of QT_cF effect of eltrombopag at higher than studied doses. A slight slope of 0.120 ms µg⁻¹ ml⁻¹ was estimated by the model; however, the 90% CI obtained from the bootstrap analysis for the slope estimate contained zero, suggesting that there is no significant relationship between plasma eltrombopag concentration and ddQT_cF.

Based on the final Cp-ddQT_cF model, simulations were performed to predict ddQT_cF at C_max for 50, 150, and 300 mg once daily regimens. Simulations of ddQT_cF at C_max for eltrombopag 50 mg and 150 mg once daily were consistent with the observed data, thus providing evidence that the model was predictive. The simulated mean (90% CI) ddQT_cF of 4.03 ms (1.55, 6.79 ms) for eltrombopag 300 mg once daily suggested that eltrombopag would not have a clinically significant effect on ddQT_c at concentrations much higher than studied. The mean C_max of 40.3 µg ml⁻¹ predicted for the 300 mg once daily regimen in healthy subjects was approximately three-fold the mean C_max value observed in patients with idiopathic thrombocytopenic purpura receiving eltrombopag 75 mg once daily, which is currently the highest approved dose for this population [9]. Patients with hepatitis C virus (HCV) achieve plasma eltrombopag C_max values approximately 50% higher and AUC(0,τ) values approximately 2–2.5-fold higher than those achieved in patients with idiopathic thrombocytopenic purpura and subjects with liver disease are more susceptible to QT_c prolongation. Assessment of the potential impact of eltrombopag on the QT_c interval in HCV patients is incorporated into the ongoing clinical trials in this population.

Both dose concentrations of eltrombopag had safety profiles similar to placebo. Three of the six discontinuations caused by adverse events were notable as they occurred after subjects experienced ventricular premature beats. All three subjects had received eltrombopag. Previous studies have been published on the incidence of electrocardiographic abnormalities in young adult volunteers without apparent heart disease who underwent continuous 24-h electrocardiographic monitoring. One study in women [10] and one in men [11] found ventricular tachycardia in 2% of volunteers: one subject in each study experienced episodes of 3 beats and 5 beats, respectively. The ventricular tachycardia observed in this study was a 5 beat episode in one of the 87 enrolled subjects (1%). Reports in the literature indicate that ventricular premature beats (>50 beats in 24 h) were observed in 6% of women [9] and 2% of men [10]. The incidence of ventricular premature beats in the current study (3/87 subjects; 3.4%) was consistent with the reported values seen in healthy subjects without apparent heart disease, suggesting that the reported ventricular premature beats may not have been due to study drug. In addition, a second degree atrioventricular (AV) block was also observed in seven subjects across different treatment groups. The incidence of AV block in this study (1% to 6%) is also consistent with the reported observation in healthy subjects [12].

In conclusion, this study showed no clinically significant QT_c prolongation by eltrombopag at therapeutic and supratherapeutic doses. There were no safety or tolerability signals of clinical concern. A small incidence of ventricular premature beats was observed, but this was consistent with previously reported incidences in healthy volunteers without apparent heart disease.

Competing interests

GM, JWP, SMM, MBW, JZ, CB and JMJ are employees of GlaxoSmithKline (GSK) and hold stock and shares in GSK. BP was an employee of GSK and holds stocks in GSK.