Assessment of QT_c-prolonging potential of BX471 in healthy volunteers. A ‘thorough QT_c study’ following ICH E14 using various QT correction methods

Abstract

AIMS

Within the framework of the clinical development of BX471, this study was intended to provide experience in conducting ‘thorough QT_c studies’ according to ICH E14. A broad range of QT correction methods and analysis strategies was employed.

METHODS

A double-blind, placebo- and positive-controlled, single-centre, three-way cross-over study was conducted in 74 healthy volunteers. Electrocardiograms were read by blinded experts. QT correction methods included Bazett's (QT_cB), Fridericia's (QT_cF) and several regression-based corrections.

RESULTS

There was a significant QT_cF prolongation of 10.26 ms by the positive control compared with placebo [95% confidence interval (7.83, 12.70)]. BX471 at therapeutic doses did not cause substantial QT_c prolongation [QT_cF estimate 2.93 ms, 95% confidence interval (1.00, 4.86); QT_cB estimate 3.30 ms, 95% confidence interval (0.85, 5.74)]. Regression-based QT correction methods yielded similar results to Fridericia's correction [e.g. using a linear regression across the study population, QT_c estimate 2.39 ms, 95% confidence interval (0.55, 4.23)]. Differences between the various regression-based correction methods were small. Results were not affected by whether the QT corrections were performed per ECG or per beat.

CONCLUSIONS

BX471 does not cause meaningful QT_c prolongation. Three QT correction methods may be sufficient in future studies: Bazett's (required by regulatory authorities), Fridericia's (as the most reliable fixed formula) and a regression-based correction (individually or population-based), each performed per ECG (i.e. applied to the means of several beats of one ECG recording).

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT

• A prolongation of the QT interval in the ECG is a sign of delayed cardiac repolarization, a condition that increases the risk of potentially fatal arrhythmias.

• Since the QT interval depends on heart rate, some correction is indicated to obtain a less heart rate dependent ‘QT_c interval’.

• Building on the recently issued ICH E14 guideline, the most appropriate methods to evaluate the QT/QT_c prolonging potential of novel drugs are yet to be found.

WHAT THIS STUDY ADDS

• This study gives a detailed account of different analysis strategies and QT correction methods consistent with ICH E14.

• Regression-based QT correction methods, applied to average intervals of each ECG recording, yielded results comparable with those obtained by Fridericia's formula; using one representative of such methods may be sufficient in practice.

• BX471 does not cause meaningful QT_c prolongation.

Introduction

BX471 is a potent and selective antagonist of C-C chemokine receptor 1 [1–3], and as such it is in early-phase clinical development as a candidate treatment for endometriosis. In view of this potential application it is being tested thoroughly for safety in humans, and such tests focus inter alia upon the assessment of any possible pro-arrhythmic effects. The absence of such effects is considered an important criterion for the safety of new drugs for diseases that are not life-threatening.

The study described here was therefore conducted with the primary aim of determining the safety of BX471 with respect to QT prolongation and possible arrhythmia. QT prolongation must be assessed because it is considered to increase the risk of arrhythmia. Recent attention to this in the clinical research sector has resulted in the publication of guidelines and recommendations, in particular the ICH guideline E14 on the clinical evaluation of prolongation of the QT/QT_c interval and its relation to the pro-arrhythmic potential of new drugs [4].

QT is dependent upon heart rate (or equivalently, RR). This requires a correction of QT to allow QT intervals from various heart rates to be compared directly. At present, the optimal choice of the QT correction method is under debate. By convention, the basis of all corrections is an RR interval of 1 s as a reference point (i.e. QT_c≡ QT for RR = 1 s), and QT is corrected for RR values that differ from 1 s. A simple formula, already in use for many decades, is that of Bazett [5]: (RR in seconds). The use of this formula is required by many regulatory authorities. However, the formula has been criticized for overcorrection, especially when RR values lie relatively far from 1 s. Several other approaches have been proposed as being more suitable. The simplest of these is a modification proposed by Fridericia [6], in which the square root of RR is replaced by its cube root, but there is no shortage of alternative suggestions. A collation of those published up to 2002 has been given in which 31 proposed correction formulae are classified according to their mathematical form (linear, logarithmic, power law, etc.) [7].

A detailed investigation of individuals has suggested that there may indeed be no generally applicable correction, i.e. the relation between QT and RR (determined by plotting QT as a function of RR, whereby each point is provided by the QT and RR values measured for a single heartbeat or for several, averaged over a short period) may vary substantially from one individual to another [8]. Furthermore, even in a given individual, diurnal [9] and longer-term [10] intra-individual differences in the relationship between QT and RR have been found. Nevertheless, the QT–RR relationship in healthy subjects has been found to display substantially greater inter-subject than intra-subject variability [11]. QT corrections derived from regression on the pooled data for all subjects (sometimes referred to as ‘population based’ corrections) provide a compromise between the one-size-fits-all approaches, using a standard formula, and the individual approach (various approaches to correction have been based on individual subjects; for details see below), which is desirable but not always practicable.

Because of the importance attached to the absence of pro-arrhythmic properties in new and established drugs, and of the present flux in opinion regarding appropriate correction of QT, the issue of QT/QT_c prolongation was the subject of deliberations by ICH that led to the recommendations set out in its guideline of 2005 [4]. This guideline deals with the clinical evaluation of QT/QT_c prolongation in the assessment of pro-arrhythmic potential. Central to these recommendations is the concept of the ‘thorough QT/QT_c study’. Such a study would generally be carried out with healthy individuals, would include a negative control (usually placebo) and an adequate positive control, and would, when feasible, have a cross-over design in order to minimize the blurring of its result by inter-individual variability. A clinically-based criterion of potential physiological significance, such as a predefined prolongation threshold should be applied. ECG recording should be timed to reflect the pharmacokinetic profile of the drug being tested. Recommendations concerning the analysis of ECG data are also made, including the use of several correction formulae, the use of potentially clinically relevant categories of QT_c prolongation and the description of any morphological abnormalities observed.

Accordingly, the present study was designed to adhere to the recommendations made in the draft versions of the ICH guideline that were available at the time when the study was designed [12, 13]. In these drafts, a prolongation of 5 or 7.5 ms, respectively, was considered of potential physiological significance. The study is also compliant with the final, adopted guideline [4] that uses a threshold of 10 ms. Subjects were treated with a negative control (placebo), a positive control (moxifloxacin) and the test drug BX471. Moxifloxacin was considered a suitable positive control because of its known effect upon QT [14]; the product information for Avalox® states that a 400 mg dose increases QT_c by 6 ± 26 ms [15].

A six-arm cross-over design was adopted, which allowed all possible permutations of order of treatment. The number of subjects was based upon a statistical rationale. ECG recording took place at time points known to be close to the maximum serum concentration of each substance, and this was confirmed by parallel pharmacokinetic measurements. Finally, QT correction was applied by taking three approaches: (i) standard correction formulae, (ii) population-based corrections and (iii) individual corrections. Furthermore, BX471 was administered not only at the putative therapeutic dose of 600 mg, but also in supratherapeutic doses up to 1200 mg following repeated dosing. The study was approved by the Ethics Committee of the Chamber of Physicians in Berlin, Germany.

This investigation demonstrates the feasibility of the ‘thorough QT/QT_c study’ and allows a suggestion to be made regarding the selection of QT correction methods. Although there has been extensive discussion of the theory and principles of the ‘thorough QT/QT_c study’, this is to our knowledge one of the first detailed reports of such a study that also discusses a variety of different QT correction methods, since the appearance of the final ICH guideline on this subject. A study with comparable design features, but also including the acquisition of ECG data by the Holter method appeared earlier [16].

Methods

Design

This GCP-compliant study followed a randomized, double-blind, double-dummy, cross-over design with three treatments (BX471, moxifloxacin and placebo) and three treatment periods. Healthy volunteers were randomized into six treatment-sequence groups; each group followed one of the six possible treatment sequences. Pharmacokinetic analysis (non-blinded) was restricted to active drug samples; the results were not disclosed until release of the database. The ECG core laboratory remained blinded with regard to treatment, time and volunteer identifier.

Treatment

Treatment with BX471 took place as follows. The treatment period started with a drug-free baseline day. From the morning of the next day (day 1) BX471 (manufactured by Schering AG, now Bayer Schering Pharma) was administered thrice daily at 8 h intervals. The first dose of BX471 on day 1 was 600 mg, and the 8 h between this and the second administration (also 600 mg) were used for the analyses of the putative therapeutic dose. The third BX471 dose on day 1 and the first on day 2 were 900 mg; the remaining two doses on day 2 and the morning dose on day 3 were 1200 mg. Measurements on day 3 were used for the analyses of the supratherapeutic dose. A placebo (‘moxifloxacin placebo’) was administered 2 h after the last dose of BX471, as explained below. Thus, the total intake of BX471 was 2 × 600 mg plus 2 × 900 plus 3 × 1200 mg. The tapering-in of BX471 was performed in order to avoid initial adverse reactions. Between treatment periods a wash-out interval of at least 4 days was instituted.

In the negative-control period, treatment was identical, except that placebo was administered throughout. In the positive-control period, placebo was administered as for the negative control, except for a single dose of moxifloxacin (commercially available Avalox® 400 mg; Bayer Vital GmbH) on the morning of day 3 two h after the last intake of ‘BX471 placebo’ (see below). A summary of the treatment schedule is given in Figure 1.

Figure 1.

Summary of the treatment scheme

All intake of study medication was oral. Moxifloxacin and moxifloxacin placebo were administered in capsules; BX471 and BX471 placebo were administered as coated tablets. By means of this double-dummy blinding method, the treatments and their respective placebos were made indistinguishable for the subjects and for the investigation team. Intake took place under supervision. Since moxifloxacin reaches peak plasma concentrations substantially more quickly than BX471, it was administered 2 h after the last dose of BX471 placebo on day 3, so that peak plasma concentrations of BX471 and moxifloxacin could be reached at approximately the same time of day.

Subjects

A total of 74 healthy, adult, non-smoking volunteers were treated. Structural heart disease and electrophysiological abnormalities were assessed at screening by standard ECG, Holter ECG, echocardiography and ergometry. Volunteers were excluded if they had any relevant cardiac history, past or present long QT syndrome, any risk factors for torsades de pointes, any recent severe disease, contra-indication or hypersensitivity to, or planned treatment with, fluoroquinolones, any one of several specified ECG, blood-pressure or heart-rate abnormalities, pregnancy, breast-feeding or inadequate contraception or any other specified relevant medical or life-style factors preventing participation.

Figure 2 displays the number of subjects screened, randomized, treated, and completing the study. Table 1 shows the demographic characteristics of the 74 treated subjects. Table 2 summarizes any abnormal findings in these subjects from cardiac evaluations at screening. At study entry, all treated subjects had normal blood pressure (systolic >100 and <150 mmHg, diastolic >60 and <95 mmHg) and heart rates [>50 and <100 beats min⁻¹, except for one subject with a heart rate of 49 beats min⁻¹ (bradycardia with a normal PQ interval)]. There were no relevant events in their medical history or their laboratory values at study entry.

Figure 2.

Study flowchart

Table 1.

Subject characteristics at screening

		Mean (range)
Age (years)		37	(20, 55)
Height (m)		1.74	(1.55, 1.93)
Body weight (kg)		72.3	(50.5, 101.5)
Body mass index (kg m−2)		23.7	(20.0, 28.5)
		n (%), total = 74
Gender:	Male	36	(49%)
	Female	38	(51%)
Race	Caucasian	73	(99%)
	Asian	1	(1%)
Smoking habits	Never smoked	58	(78%)
	Former smokers	16	(22%)
Alcohol consumption	Never	12	(16%)
	Seldom	31	(42%)
	Occasionally	30	(41%)
	Regularly	1*	(1%)

^*One beer daily.

Table 2.

Number of subjects with abnormal findings in cardiac assessments at screening

	n (%), total = 74		Most frequent abnormal findings
ECG	58	(78%)	Sinus bradycardia (in 30 subjects), inverted T waves (22), left ventricular preponderance (14), conduction disturbance (10), and prolonged or inverted P waves (9)
Holter ECG	2	(3%)	Ventricular ectopy (1), AV block degree 2 (1)
Echocardiography	2	(3%)	First-degree mitral regurgitation (1), minor mitral-valve insufficiency (1)
Ergometry	14	(19%)	ST segment descent (11)

All findings were clinically not relevant.

Assessments

Pharmacokinetic sampling was conducted at time points designed to check that the serum concentration profiles of BX471 and moxifloxacin were as expected (viz: immediately before first dosing and then hourly up to 8 h after dosing, thereafter at increasing intervals up to 48 h; then hourly until 56 h and at increasing intervals up to 96 h, with a final sampling at 120 h). Plasma concentrations of BX471 and moxifloxacin were measured by standard, validated methods (LC-MS/MS and HPLC). Pharmacokinetic parameters for each analyte were calculated in a model-independent manner. Safety assessments included standard laboratory variables, vital signs, adverse events and safety review of the ECG data. Samples were retained for pharmacogenetic analysis in case of need, but as no safety concerns were raised, and no QT-relevant effects of BX471 were observed, this analysis did not take place.

Standard 12-lead ECGs (Hellige® Marquette PC-ECG) were recorded at time points matching the pharmacokinetic sampling. To reduce variability, three ECGs were recorded at each time point, each consisting of five consecutive beats and 2 min apart from its neighboring ECGs. A series of such ECGs was recorded on the baseline day as a reference, enabling the calculation of changes from time-matched baseline on dosing days. Lead II was used as the standard lead for QT measurements. If the quality of lead II was not adequate for the measurement, then another lead was used (e.g. lead V5) according to standard procedures at the ECG core laboratory. All measurements for a given volunteer had to be made using the same lead. Manual evaluation of the electronic data was performed on screen applying an automatic calliper. The intervals that were measured and the exact start and end of the interval were documented electronically by a skilled reader. The same skilled reader evaluated all intervals from all subjects. Of the three ECGs recorded at each time point, one was assessed qualitatively by a cardiologist. This was considered appropriate, as a clinically significant change within 4 min is unlikely to occur in healthy volunteers. The same hardware and software was used for all ECGs, and the same cardiologist evaluated all ECGs from all subjects. At all time points in each of the three recorded ECGs in five complexes, the following measurements were made: RR interval, QT interval, PQ (PR) interval, QRS duration. In addition, the following assessments were performed: heart rate (derived from the beats used for QT measurement), rhythm/arrhythmia, number and origin of ectopic beats, conduction disorders, morphology of the T wave (normal/abnormal; if abnormal, to be specified), and the presence or absence of U waves.

Statistics

The primary target variable was the greatest of the QT/QTc interval changes from time-matched baseline within the 8 h time window of interest: after the first dose on day 1 for the putative therapeutic dose, and after the (last) dose on day 3 for the putative supratherapeutic dose. It was derived in four steps. First, averages of QT and RR were computed over five complexes for each ECG recording separately. The average RR values thus obtained were then used to correct their respective average QT intervals by one of various methods (see below). Subsequently, the resulting QT_c values for the three ECG repetitions taken at each time point were averaged. Finally, for each subject, period and time point in the time window of interest, the change in each QT_c value relative to its corresponding time-matched baseline value was calculated. The resulting set of values was a temporal succession of QT_c differences (treatment minus baseline) for each subject and period, over the respective time window. This set was then analyzed by considering i) its greatest value, ii) its arithmetic mean and iii) each difference separately. Option i) was chosen as the primary target variable in this study; option iii) was included post hoc and is not presented here. Comparisons between the various methods for correcting the raw QT interval were made possible by performing the above procedure, from the second step on, for each correction method separately.

The analysis was performed by mixed-model analyses of variance, including the factors ‘treatment-sequence group’, ‘period’, ‘treatment’, and ‘subject’ (as a random factor), and using the baseline values as continuous covariates (cf. CPMP [17]). Carry-over was not included in the model, since the wash-out time was considered sufficient. The differences in treatment effect were estimated for BX471 (both dose levels) vs placebo and for moxifloxacin vs placebo. The two time windows, one for the therapeutic dose, the other for the supratherapeutic dose and moxifloxacin, were evaluated separately.

The main aim of the study was to show that on average, the administration of BX471 does not prolong QT/QT_c to a clinically meaningful degree, compared with placebo. This conclusion was to be drawn if the upper limit of the two-sided 95% confidence interval for the mean difference in the target variable between BX471 and placebo was smaller than the non-inferiority margin Δ (in ms). As at the time of protocol finalization, the discussion about an appropriate choice of Δ was ongoing in the scientific community, it was decided to use Δ= 7.5 ms (following ICH E14, draft 3 [13]) and Δ= 5 ms (following its draft 2 [12]) in a hierarchical procedure that does not inflate the overall type I error rate. Note that the final version of ICH E14 [4], that was approved after the start of this study, is less stringent in the sense that it recommends Δ= 10 ms and one-sided instead of two-sided testing. The sample size had been calculated to achieve a satisfactory power for the whole hierarchical procedure. Assay sensitivity was considered to be established if the lower limit of the two-sided 95% confidence interval for the mean difference in the target variable between moxifloxacin and placebo was greater than 0.

QT corrections

Six versions of QT/QT_c were used:

• QT uncorrected

• QT_cF (corrected according to Fridericia; see above)

• QT_cB (corrected according to Bazett; see above)

• QT_cp (population-corrected, performed for the entire study population). All drug-free QT measurements were pooled and regressed onto the corresponding RR intervals. The estimated regression equation was used to calculate QT_cp for all ECGs

• QT_cs (population-corrected as above, but performed separately for each sex)

• QT_ci (corrected individually: as above, but performed separately for each subject)

For QT_cp, QT_cs and QT_ci, three separate regression equations were used:

i)	linear:	QT =a× RR +b
ii)	linear on the logarithmic scale:	QT =b× RRa
iii)	linear on the exponential scale:	QT = ln(a× RR +b)

Fridericia's correction was selected as the primary correction method; other methods were compared with it in secondary analyses. The choice of Fridericia's correction was based upon its wide use and its reasonable performance compared with all other methods in common use. The regression methods are generally considered superior, but are less well established. Bazett's correction was included, in spite of its known poor performance, on account of its widespread earlier use.

Further analyses

The continuous measure of the QT(c) interval was additionally evaluated in a dichotomized fashion: volunteers were counted who experienced a QT(c) interval (or an interval prolongation over baseline) exceeding a certain threshold. Numbers of outliers and of subjects with wave abnormalities were likewise analyzed. QT_c values found at the respective pharmacokinetic t_max were also compared; for details see Results.

Results

Pharmacokinetics

A complete pharmacokinetic analysis was conducted for both BX471 and moxifloxacin; only a brief summary description is given here. For the purpose of the ECG investigation in which the timing of ECG acquisition has to be related to the plasma concentration of the respective drug investigated, the most important value was t_max. For BX471 this was found to be 5.0 h after the first administration (600 mg BX471; range 0.9–8.0 h) and also 5.0 h after the final administration (1200 mg BX471; range 0.0–8.0 h) following repeated dosing according to the schedule described above. After repeated dosing at the supratherapeutic dose, C_max and the AUC increased by a factor of ∼3 and ∼4, respectively. For moxifloxacin (400 mg, single dose) t_max was found to be 2.0 h (range 1.0–4.0 h); this lies well within the range of values determined by others [18]. Mean moxifloxacin concentration peaked at approximately the same time (relative to BX471 administration) as mean BX471 concentration did, as indeed had been intended by the employed dosing schedule.

Of importance for the present study is the fact that the times of the ECG recordings (1, 2, … 8 h after administration) did cover the t_max times of both moxifloxacin and of BX471, as intended. For all pharmacokinetic parameters of moxifloxacin, good correspondence was found between the values measured in this study and those in the literature [14, 15, 19]. For BX471 equally good correspondence was found with earlier results (data on file, to be published in detail elsewhere).

Safety

Table 3 summarizes the most frequent adverse events observed in this study. Overall, no safety concerns emerged, and no unexpected or previously unknown adverse events were observed. Application site erythema (from the ECG electrode patches) was observed in 71 of 74 subjects, evenly distributed across all treatments. Nausea, vomiting, diarrhoea and abdominal discomfort were observed mainly after BX471. The adverse event profiles of the active treatments corresponded to those expected for BX471 [2] and for moxifloxacin [15]. Of particular importance in the context of the present report was the fact that there were no clinically relevant safety-ECG abnormalities after treatment for the volunteers who completed the study. No major or consistent differences between treatments were seen in the number of volunteers with abnormal safety ECG findings for the time points investigated, and no abnormal blood pressure or heart rate values were documented for any volunteer.

Table 3.

Number of subjects with most frequent adverse events (>=10%)

	Placebo		Moxifloxacin		BX471		Overall
Adverse event	n (%), total = 73		n (%), total = 73		n (%), total = 74		n (%), total = 74
Application site erythema	37	(50.7%)	39	(53.4%)	38	(51.4%)	71	(95.9%)
Nausea	6	(8.2%)	1	(1.4%)	36	(48.6%)	38	(51.4%)
Vomiting	1	(1.4%)	0	(0.0%)	17	(23.0%)	18	(24.3%)
Diarrhoea	1	(1.4%)	1	(1.4%)	12	(16.2%)	12	(16.2%)
Abdominal discomfort	2	(2.7%)	1	(1.4%)	8	(10.8%)	9	(12.2%)

Following the draft of ICH E14 current at the time of design of this study [12], particular attention was directed toward torsade de pointes, ventricular tachycardia, ventricular fibrillation, ventricular flutter, cardiac arrest, sudden death, syncope, dizziness, palpitations and seizures; seizures were not distinguished from convulsions. Of these events, only dizziness (two subjects receiving placebo, four receiving BX471) and palpitations (one subject receiving BX471) occurred. (Note that in the final version of the ICH guideline E14 [4], dizziness and palpitations no longer appear in the list of adverse events possibly indicative of a drug-related pro-arrhythmic potential.)

Two volunteers were withdrawn from the study because of adverse events that in view of their timing were rated as being possibly drug-related: one withdrawal was due to mild ventricular extrasystoles on day 1 of period 3 (treatment with BX471; total dose 600 mg), and the other to mild frequent supraventricular extrasystoles on day 2 of period 1 (treatment with BX471, total dose 2100 mg). One adverse event that was recorded for 23% of subjects who had taken BX471 was vomiting. This is not likely to have influenced the uptake of BX471, as these subjects showed a high plasma concentration at the time of emesis.

Pharmacodynamics: change in QT_cF from baseline

For BX471 and for moxifloxacin, the greatest change in QT_cF relative to time-matched baseline was compared with the corresponding change for placebo. Results are summarized in the first rows of Table 4. The data are interpreted as follows.

Table 4.

Difference from placebo for the greatest QT_c change from time-matched baseline within the first 8 h after administration of BX471, using different QT correction methods

Correction method	Regression type	Comparison (see footnote)	Point estimate	95% confidence interval
QTcF	–	1	2.93	1.00, 4.86
		2	6.24	3.80, 8.68
		3	10.26	7.83, 12.70
QTcB	–	1	3.30	0.85, 5.74
		2	6.11	3.21, 9.00
		3	10.33	7.45, 13.21
QTci	Linear	1	1.96	0.06, 3.87
		2	5.87	3.38, 8.36
		3	9.64	7.16, 12.12
	Logarithmic	1	2.09	0.15, 4.02
		2	5.95	3.45, 8.45
		3	9.68	7.18, 12.17
	Exponential	1	1.97	0.08, 3.85
		2	5.83	3.35, 8.31
		3	9.56	7.09, 12.03
QTcp	Linear	1	2.39	0.55, 4.23
		2	6.18	3.75, 8.61
		3	9.96	7.53, 12.39
	Logarithmic	1	2.58	0.71, 4.44
		2	6.27	3.84, 8.69
		3	10.14	7.72, 12.56
	Exponential	1	2.37	0.55, 4.19
		2	6.16	3.74, 8.59
		3	9.88	7.45, 12.30
QTcs	Linear	1	2.48	0.58, 4.38
		2	5.89	3.51, 8.27
		3	9.93	7.56, 12.30
	Logarithmic	1	2.85	0.85, 4.86
		2	6.08	3.66, 8.51
		3	9.96	7.54, 12.39
	Exponential	1	2.50	0.62, 4.38
		2	5.82	3.47, 8.16
		3	9.82	7.48, 12.16

Abbreviations for the various QT_c are defined in the text (‘QT corrections’). 1, 600 mg BX471 vs placebo (single dose); 2, 1200 mg BX471 vs placebo (after repeated dosing); 3, 400 mg moxifloxacin vs placebo (single dose).

In the BX471–placebo comparison on day 1 (top row), the greatest QT_cF change from time-matched baseline within the first 8 h after administration was 2.93 ms greater following a single dose of 600 mg BX471 than following placebo (95% confidence limits 1.00 and 4.86 ms). The upper confidence limit for this prolongation was smaller than 5 ms. Therefore, the null hypothesis of a difference of at least 5 ms between BX471 and placebo could be rejected for this dose. Thus, the primary objective of this study was achieved for the anticipated therapeutic dose of 600 mg.

On day 3, the greatest QT_cF change from time-matched baseline within the first 8 h after administration was 6.24 ms greater following the last dose of 1200 mg BX471 than following placebo (second row; 95% confidence limits 3.80 and 8.68 ms). In particular, the upper confidence limit for this prolongation was not smaller than 7.5 ms. Thus, the primary objective of this study was not achieved for the anticipated supratherapeutic dose. However, the upper confidence limit was smaller than 10 ms, which is the recommended choice of the non-inferiority margin according to the final version of ICH E14 [4], which was not available when this study was designed.

Finally, the single dose of 400 mg moxifloxacin substantially prolonged the greatest QT_cF change from time-matched baseline within the first 8 h after administration, compared with placebo (third row). In particular, the lower confidence limit was greater than 0 ms. Thus, assay sensitivity was established.

Qualitatively similar results were obtained for the arithmetic mean of the eight QT_cF changes from time-matched baseline within the first 8 h after administration (first rows of Table 5).

Table 5.

Difference from placebo for the mean QT_c change from time-matched baseline within the first 8 h after administration of BX471, for QT_cF and QT_cB

Correction method	Comparison (see footnote)	Point estimate	95% confidence interval
QTcF	1	1.82	0.60, 3.05
	2	4.41	2.60, 6.22
	3	6.97	5.16, 8.79
QTcB	1	2.03	0.45, 3.62
	2	4.32	2.26, 6.37
	3	7.47	5.41, 9.53

Abbreviations for the various QT_c are defined in the text (‘QT corrections’). 1, 600 mg BX471 vs placebo (single dose); 2, 1200 mg BX471 vs placebo (after repeated dosing); 3, 400 mg moxifloxacin vs placebo (single dose).

These findings are supported by box plots (Figures 3 and 4) showing the greatest and the mean QT_cF changes from time-matched baseline by day and treatment. As moxifloxacin was not administered until day 3, this treatment behaved like a placebo on day 1. On day 3, after the last dose of 1200 mg BX471, a marginal prolongation of the target variable was apparent. The strongest prolongation of QT_cF was observed after the administration of moxifloxacin on day 3.

Figure 3.

Greatest QT_cF change from time-matched baseline

Figure 4.

Mean QT_cF change from time-matched baseline

The time-dependent course of the changes from baseline was also analyzed. It is plotted in Figures 5 and 6 from 1 h to 8 h after administration. On day 1 no substantial differences between the treatment arms were seen; an initial QT_cF shortening was normalized after a few hours in all arms. On day 3 there was a slight rise for BX471, and a strong rise for moxifloxacin around its t_max.

Figure 5.

Average QT_cF changes from time-matched baseline, day 1. PL (); BX (); MX ()

Figure 6.

Average QTcF changes from time-matched baseline, day 3. PL (); BX (); MX ()

Pharmacodynamics: change in QT_cB from baseline

Subsequent rows of Tables 4 and 5 show the corresponding results for QT_cB. Some differences exist with respect to the hierarchical testing procedure. For the primary analysis (greatest changes from baseline) the upper confidence limit was smaller than 7.5 ms, but not smaller than 5 ms. Therefore, had Bazett's correction been chosen for the primary analysis, the primary objective of this study would have been achieved for the anticipated therapeutic dose of 600 mg only for the non-inferiority margin of 7.5 ms. This is in contrast to the result using Fridericia's correction (above), where the margin 5 ms could also be excluded. For the other changes in QT_cB, the qualitative result is the same as that for QT_cF, but the upper confidence limits are greater throughout. All confidence intervals are wider for QT_cB than for QT_cF.

Comparison of QT corrections per beat and per ECG

The results were re-calculated by applying the different QT correction methods to single beats (rather than to ECG recordings consisting of averages of five beats) and then averaging QT_c over all 3 × 5 beats per time point. Results for greatest changes from baseline in QT_cF and QT_cB calculated by the per-beat method were very similar to those obtained by the per-ECG method. Results for mean changes and for individual time points were virtually identical, irrespective of whether they were calculated on a per-beat or a per-ECG basis.

Comparison of QT correction methods

Table 4 also shows the primary analysis for different QT correction methods (each method applied to per-ECG means; the target variable was the greatest of the changes from time-matched baseline within the first 8 h after administration).

The results from the regression-based correction methods were comparable to those from Fridericia's method. The estimates of QT_c prolongation tended to be slightly smaller for the individual corrections than for the population-based corrections, with the corrections per gender in between them. The corrections based on the logarithmic method tended to yield slightly larger point estimates than those based on the linear or exponential method. Generally, the differences were small. The regression-based point estimates ranged from 1.96 ms (linear, individual) to 2.85 ms (logarithmic, per gender) for 600 mg BX471. The widths of the confidence intervals of the regression-based correction methods were comparable with those of Fridericia's method, and smaller than those of Bazett's method.

Further ECG-related analyses

Several further analyses were conducted, with findings that did not differ in any noticeable way from those described above. These are summarized here.

Threshold analyses: Thresholds chosen prospectively were 450, 480 and 500 ms for QT(c) and 30 and 60 ms for changes in QT(c). These thresholds were exceeded only very rarely. The comparison (McNemar test) of moxifloxacin with placebo on day 3 showed an effect in the greatest QT_cB changes from baseline (threshold: 30 ms), and in the maximum of the QT_cB values (threshold: 450 ms). This effect was not reflected by the respective means.

Outliers: One subject, under treatment with BX471, revealed a QT_cB value of 504 ms (first ECG at this time point, 520 ms; second, 509 ms; third, 484 ms). All other QT_c values of all volunteers were below 500 ms during the whole study, irrespective of QT correction method.

Wave characteristics and overall interpretation: Only few volunteers showed T-wave abnormalities or abnormal ST segments, while U waves were present in a substantial number of volunteers. No systematic differences between the treatments could be detected for either of these. Also with regard to the overall interpretation of the ECG, which was abnormal in a moderate number of subjects, no treatment difference could be detected.

QT_c at t_max: The treatment effect estimates were comparable with those in the primary analysis (Table 4, QT_cF). However, in the case of the anticipated therapeutic dose of 600 mg, the upper 95% confidence limit was below 7.5 ms but not below 5 ms (in contrast to the primary analysis, where this limit was below 5 ms). The effect of moxifloxacin was slightly stronger at t_max than in the primary analysis (13.43 ms compared with 10.26 ms, Table 4). Thus, overall conclusions were the same as in the primary analysis.

Discussion

The two aims of this study were to assess the QT prolongation potential of BX471 and, within this framework, to provide experience of the conduct of ‘thorough QT_c’ studies based on the ICH E14 guideline, while at the same time obtaining more information on the various methods of correcting QT for heart rate.

The study analysis was based upon the then current draft ICH guideline [12] for studies to detect any prolongation and pro-arrhythmic potential of non-anti-arrhythmic drugs. During the planning and conduct of this study a new draft was issued [13]. For this reason, a hierarchical evaluation was chosen, with thresholds of clinical significance set at 7.5 and 5 ms. The drafts of the ICH guideline were followed by the finally adopted version [4], in which a thresholds of 10 ms for clinical significance was recommended. In a study designed today, only the latter value would be considered relevant. The final guideline additionally recommends one-sided testing instead of the two-sided testing implemented in this study; however, the two-sided test is more rigorous, and therefore its use does not compromise the conclusion of the study.

The controls functioned as expected. The negative control (placebo) showed no effect, as shown not only by the analysis in the placebo period, but also, incidentally, by the results from the initial part of the moxifloxacin period: moxifloxacin was administered only at the end of the moxifloxacin period, so that earlier administrations in the same period consisted of placebo, as required by the double-dummy design. The positive control, a single dose of 400 mg moxifloxacin, led to a prolongation of the mean QT_c change from baseline of ∼7 ms (greatest QT_c change from baseline, ∼10 ms), the effect varying somewhat according to the correction method chosen. This agrees well with literature values: for example the product specifications for Avalox® 400 mg state a QTc prolongation of 6 ± 26 ms [15], and Morganroth et al.[19] observed that a single dose of 400 mg moxifloxacin increased QT_ci by an average of 7 ms, while Noel et al.[14] found a prolongation of QT_cB by ∼17 ms for 800 mg moxifloxacin.

Thus, the negative and positive controls confirmed the ability of the study design to yield firm results. For the therapeutic dose of BX471, it could be ruled out that the estimated QT_c prolongation reached the threshold of clinical relevance (a Δ value of 10 ms) as defined by the finally adopted ICH guideline. In fact, the prolongation did not even reach the stricter, and no longer applied, Δ value of 5 ms.

The supratherapeutic dose of BX471 (1200 mg thrice daily) did reach the 5 ms threshold – that is to say, an increase in QT_c of 5 ms could not be ruled out. However, even this high dose would not be considered problematical according to the final guideline (Δ= 10 ms).

In the prospective analysis based on the draft ICH guidelines [12, 13], the greatest and the mean changes from baseline of QT_c were considered. The final ICH guideline [4] uses the wording ‘largest time-matched mean effect’, which may be interpreted as analyzing all time points separately; Δ would have to be ruled out at each of these time points. The target here is the greatest of several time-matched treatment differences, rather than the treatment difference of the greatest change from time-matched baseline. This analysis was added post hoc (results not reported here); it may be chosen as primary analysis of a study designed today.

In the comparison of correction methods for QT, regression-based methods yielded results similar to those given by Fridericia's correction. Differences between the various regression-based correction methods were small. Analysis results were almost identical irrespective of whether the QT corrections were performed per ECG or per beat.

This conclusion is supported by observations of Desai et al.[20] who used a power law of the form QT_c= QT / RR^α, where α would be 0.50 for Bazett's correction and 0.33 for Fridericia's. Desai et al. determined α values on an individual basis. They found that α varied from 0.23 to 0.38 in 16 healthy subjects; Fridericia's correction falls within this range, while Bazett's does not.

Further evidence for a better performance of Fridericia's correction compared with Bazett's is cited by Funck-Brentano et al.[21] and ICH guideline E 14 [4].

In summary, BX471 does not cause meaningful QT_c prolongation. The results of this study suggest that it may be sufficient in future studies to perform only three different QT corrections: Bazett's (required by regulatory authorities), Fridericia's (as the most reliable fixed-formula correction) and a representative of a regression-based correction (e.g. using linear regression, individually or population based), all of these per ECG (i.e. applied to the means of several beats of one ECG recording).

Competing interests

All authors are or were employed by the study sponsor. All authors hold or held shares in the study sponsor.

We would like to thank Cosima Klein for valuable statistical programming support.