| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
METHODS |
From Daiichi Medical Research, New Jersey, and the Department of Cardiology, Karolinska University Hospital, Stockholm, Sweden (Dr Darpo); and Clinical Development, Pfizer Inc, United Kingdom and United States (Dr Agin, Dr Kazierad, Mr Layton, Mr Muirhead, Dr Gray, Dr Jorkasky).
Address for reprints: Gary Layton, MSc, Clinical Statistics, Clinical R & D, Pfizer Global Research and Development, Ramsgate Road, Sandwich, Kent, CT13 9NJ, United Kingdom; e-mail: gary.layton{at}pfizer.com
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Key Words: QT prolongation • ECG methodology • manual methods • automated methods
Many factors, such as autonomic tone, time of the day, food intake, smoking status, age, and gender affect the QT interval.10-12 In addition, methodology-related factors, such as choice of electrocardiogram (ECG) lead(s) and the method by which the interval is measured, also have an impact on variability.13,14 Manual methods have historically been preferred in drug development15-18 and have clearly established advantages when applied to ECGs recorded in patients with altered T wave morphology due to cardiovascular disease13 or treatment with potent, pure IKr blockers.19-22 Any technique for QT interval measurements will have inherent variability and measurement error, which for manual techniques include day-to-day, within- and between-reader variability.14 The latter may be eliminated by using computer-derived, automatic QT algorithms,23,24 which are widely available and often contained in modern ECG machines.
We recently observed unexpected QT prolongation in 2 phase I studies in healthy volunteers. The studies were not thorough QT studies per se, but careful QT assessment was included to allow early detection of any QT effect and to support the decision of further clinical development. Both of the 2 ECG core laboratories involved in these 2 studies had quality assurance procedures in place that were deemed appropriate for that time and had reported interreader and intrareader variability within acceptable limits. The first study had a parallel group design and was conducted at 1 site with 4 groups of 6 subjects each (age, mean ± SD: 37 ± 9 years; 15 men and 9 women) who received either placebo or an investigational substance at 3 dose levels (1x, twice daily; 2x, once daily; or 2x, twice daily). Subjects were dosed for 14 days, and ECGs were recorded at baseline and after dosing. Electrocardiograms were recorded on days 1 through 16 with a Burdick E550 machine (Burdick Inc, Deerfield, Wis), and paper printouts were sent to and analyzed at a central ECG laboratory in 2 separate batches (days 1-9 and 13-16). Because of the unexpected results from this analysis (Figure 1), all intervals were remeasured by the same ECG laboratory, with the provision that the same technician should analyze all ECGs from a subject and therefore ECGs from both batches. In addition, all values obtained by the automated algorithm in the ECG machine were analyzed. The unexpected QTc increase with placebo and investigational substance on days 13 through 16 with the first manual reading could not be reproduced with these latter 2 methodologies, which were consistent with each other. After careful scrutiny of all study-related procedures, we concluded that other things being equal, the effect observed in the first manual analysis was caused by 2 completely separate groups of technicians each reading only 1 batch of ECGs. The second study enrolled 18 healthy volunteers (age, 32 ± 8 years; 10 men and 8 women) at 1 site who underwent 2 treatment periods. In the first period, subjects were dosed with clarithromycin 500 mg twice daily for 3 days. After 14 days' washout, subjects were dosed with an investigational drug once daily for 21 days in the second period. Starting on day 8 of this period, clarithromycin 500 mg twice daily was added during the last 14 days. Electrocardiograms were recorded at baseline (day 0 in each period) and after dosing on days 1 through 4 in period 1 and on days 1 through 22 in period 2, using a Hewlett-Packard M1700A Pagewriter XLi Cardiograph ECG machine (Hewlett-Packard, Palo Alto, Calif). The initial measurements were performed manually on paper printouts sent in 3 batches to a central ECG laboratory (period 1; period 2, days 0-7; and period 2, days 8-22). A clear increase in QTcB of about 17 milliseconds was observed with the first manual readings on the morning of day 8, before the first dose of clarithromycin, which could not be explained by a corresponding change in drug exposure or heart rate or by discrepancies in study conduct or data analysis. It was subsequently found that the manual measurements had been performed by 3 generally discrete sets of technicians, 1 set per batch of ECGs. Based on this unexpected finding of QT prolongation, ECGs were reanalyzed by a single technician at a different central ECG laboratory using a different manual method but the same lead (lead II). In addition, all values obtained by the automated algorithm in the ECG machine were analyzed. An increase of about 6 to 10 milliseconds was observed with the machine-measured intervals and with those from the second central ECG laboratory (Figure 2).
|
|
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Informed consent was obtained from all subjects.
Study 1. Eighty healthy male (n = 40) and female subjects (mean age, 38.8 years; range, 18-63 years) were included at 4 sites. It was a single-dose study, with a 5-way crossover design with placebo, ketoconazole (Nizoral) 800 mg, and 3 doses of an investigational drug. Only data from the placebo (n = 77) and the ketoconazole (n = 76) periods were analyzed for this investigation. Electrocardiograms were obtained as single recordings 0, 1, 2, 3, 4, 6, 8, 10, 12, 16, and 24 hours after drug administration and at corresponding clock times at baseline (day 0 of each period), with either a Hewlett-Packard 708 and 709 (2 sites) or a GE Marquette Mac5000 or MacVu (GE Medical Systems, Milwaukee, Wis) ECG machine (2 sites). Manual measurements were performed at a central ECG laboratory, as described below. There were no protocol specifications with regard to the number of technicians involved in the measurements. QT values were rate corrected with the Fridericia formula.25
Study 2. Sixty-one healthy male (n = 30) and female subjects (mean age, 29.9 years; range, 19-44 years) were enrolled in a single dose, single-site, 5-way cross-over study with placebo, moxifloxacin 400 mg (Avelox) and 3 doses (1x, 3x, and 9x) of an investigational substance. Data from all treatment periods (placebo, moxifloxacin, and investigational drug; n = 57) are included in the Bland-Altman analysis, and data from the placebo (n = 59) and moxifloxacin periods (n = 58) are included in all other analyses. Electrocardiograms were recorded before (3 separate recordings, 30 minutes apart) and 1, 2, 3, 4, 8, and 12 hours after dosing in each period, and the QT intervals were corrected with the Fridericia formula. Electrocardiograms were recorded with a Hewlett-Packard 709 machine, and manual measurements were performed as described below. There were no protocol specifications as to the number of technicians involved in the measurements.
Study 3. Study 3 was a multiple dose, single-site, 2-way crossover study in which 60 healthy male subjects (mean age, 36.4 years; range, 19-64 years) were enrolled. Placebo and 3 escalating doses of an investigational drug were given, and moxifloxacin 400 mg (Avelox) was added as a single-dose active control at the end of the placebo study period. Data from all treatment periods (placebo, moxifloxacin, and investigational drug; n = 60) are included in the Bland-Altman analysis, and data from the placebo and moxifloxacin are included in all other analyses. Single ECGs were recorded before (3 separate recordings, 30 minutes apart) dosing on day 1 and at -30 minutes, 30 minutes, 1, 2, 4, 8, and 12 hours after drug administration on days 3, 6, and 9, relating to the 3 escalating doses of the investigational drug. Electrocardiograms were also recorded on day 10 at the same times after moxifloxacin administration. Electrocardiograms were recorded with a Hewlett-Packard 709 machine, and manual measurements were performed, as described below. The protocol did not prespecify the number of technicians who were to measure the ECG intervals manually.
ECG Recordings and QT Measurements
Electrocardiograms were recorded after 30 minutes semirecumbent rest as a 10-second strip at a paper speed of 25 mm/s and amplification of 10 mm/mV. For automated ECG measurements, the proprietary QT algorithms contained in the ECG machines were used. These algorithms are based on proprietary technology that is only partially accessible. Almost all automated measurements were accepted by the investigators, and only 10 of 9500 recordings were manually read by the investigator because of technical errors or obvious miscalculations. These manually read intervals were databased and included in this analysis. Electrocardiograms were printed on paper and transferred to a central ECG laboratory for manual assessment. Uncorrected QT and RR values were entered into a database, rate corrected, and analyzed by the sponsor.
ECG Machines and QT Algorithms
Hewlett-Packard (HP) M1700A 708 and 709 (studies 1, 2, and 3). The proprietary algorithm HP ECG Measurements Adult Analysis Program (HPAP) determines a global QT interval from the median of the individual QT intervals found on the 8 most stable leads. A stable lead is defined as one with a low standard deviation of the QT interval determined on every beat in each lead.
GE Marquette (GE) Mac5000 and MacVU (study 1). The QT algorithm 12SL ECG Analysis software (12SL) in this recorder constructs a super lead from a median beat of all 12 leads, and the end of the T wave is defined as the point at which the slope of the downward portion of the T wave is less than 25% of maximum slope. This QT interval essentially corresponds to earliest onset of Q to latest offset of T in any lead, and the algorithm therefore generates relatively large QT values compared to other algorithms.
Manual Method for QT Measurement
The same central ECG laboratory performed all manual measurements in studies 1, 2, and 3. Paper ECGs were amplified and displayed on a screen using a videoscope. Intervals were measured in lead II with calipers on the screen. The asymptotic corner method26 was used, whereby a tangential line was drawn along the steepest portion of the T wave downslope, and the end of the T wave was defined at the point where this tangent crossed the baseline. Measurements were performed by trained technicians and reviewed and potentially overread by a cardiologist.
Statistical Methods
For each study, the effect of the positive control (ketoconazole or moxifloxacin) compared to placebo was estimated at each time of assessment. For this analysis, the QTcF interval at each time point on placebo (eg, 4 hours after dosing in the placebo treatment period) was subtracted from the corresponding QTcF interval on drug (in this example, 4 hours after dosing in the positive control treatment period). The mean difference between treatments at each time point and the corresponding 2-sided 90% confidence intervals were calculated.
In study 1, a full 24-hour baseline day was also included in each treatment period, which allowed for a comparison of the QTcF intervals from 2 drug-free days from different treatment periods. For this analysis, each value during the baseline day of the placebo period (eg, at 10:00 AM) was subtracted from the value at the same clock time on the baseline day in the ketoconazole period and displayed as described above.
An outlier analysis was also performed, tabulating the frequency of ECG recordings with absolute QTcF values 
450 milliseconds, >450 to 480 milliseconds, and >480 milliseconds at any time point or increases from baseline of QTcF <30 milliseconds, 30 to <60 milliseconds, and 60 milliseconds for each treatment (categories consistent with those mentioned in ICH E14). For the latter, a single baseline value was used in study 1 (hour 0 on the dosing day) and in study 3 (hour -0.5 on the dosing day), and the mean of 3 recordings (hour -1.25, -0.75, and 0 on dosing day) was used in study 2. The same analysis was also performed based on the number of subjects who exhibited the same outlier values within each treatment. In study 1, change from baseline was also analyzed using a time-matched baseline from a full 24-hour day assessment recorded on the day before dosing in each treatment period.
To compare uncorrected QT intervals generated by manual and machine methods in a pairwise fashion, the method proposed by Bland and Altman was employed.27,28 In these plots, the difference between uncorrected QT intervals obtained on the same ECG by machine and manual measurements (machine minus manual) was plotted versus the mean of the 2 values. Limits of agreements (LOA) were defined as the mean ± 1.96 x SD of these differences. For studies 2 and 3, data from all treatment periods were used, whereas for study 1, only data from the placebo and positive control period (ketoconazole) were used. The same analysis was performed for the change from baseline in uncorrected QT. The baseline values for this analysis were the same as described above.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
Variability of the QTcF Interval During a Drug-Free 24-Hour Baseline Assessment
The drug-free time-matched variation of the QTcF interval between baseline days in 2 separate periods was studied in study 1, in which there was a full 24-hour baseline assessment in each period (Table II). The point estimate of the differences between measurements on each baseline day was small with all methods and varied from -3.9 to 1.9 milliseconds (manual) and -3.5 to 1.7 milliseconds (machine) for recordings from the HP machine and from -2.5 to 6.4 milliseconds (manual) and -7.5 to 1.9 milliseconds (machine) for recordings from the 12SL machine. With one technique (manual readings on paper recordings from 12SL machines), the difference between drug-free days approached or exceeded the threshold of about 5 milliseconds, with upper 2-sided 90% confidence bounds greater than 10 milliseconds at 2 of 11 time points, which could be interpreted as a false positive clinical signal: 4.9 milliseconds at 1 hour (upper bound = 9.7 milliseconds) and 6.4 milliseconds at 10 hours (upper bound = 10.1 milliseconds). With either the machine readings or the manual readings from the HP recordings, no false positives were observed. The largest QTcF decrease of -7.5 milliseconds between the 2 different days at the same time points on the 2 different days was seen with the machine measurement from the 12SL machine at 12 hours. The difference between methods (machine minus manual) was otherwise small over the 11 time points.
|
Time-Matched Effect on the QTcF Interval Induced by the Positive Control
The placebo-corrected time-matched effect of a positive control on QTcF is shown at various time points after dosing in Table III. In study 1, 800-mg ketoconazole showed mean increases postdose compared to placebo, exceeding 5 milliseconds, that were statistically significant (lower bound of 90% confidence interval greater than 0, no adjustments were made for multiple comparisons), 1 to 4 hours postdose with all methods, except the machine measurements from the 12SL machine. These statistically significant differences were observed despite the data having been split between the 2 machines. Mean increases ranged from 3.6 to 9.8 milliseconds, across the methods 6, 8, and 10 hours postdose (data not shown for 6 and 10 hours), with statistical significance observed with all methods except the manual measurements from the Hewlett-Packard machine and 12SL at 8 hours. At 12, 16, and 24 hours after dosing, the mean effect was below 5 milliseconds with all methods (data not shown for 16 and 24 hours).
|
In studies 2 and 3, 400 mg moxifloxacin showed mean increases postdose compared to placebo that were statistically significant for all time points postdose (measured to 12 hours). The maximum mean increases from placebo were 11 to 14 milliseconds, which occurred at either 1 or 2 hours postdose. In study 3, mean increases exceeded 5 milliseconds for all time points postdose, with exception for hour 12 with both methods and for hour 8 with the HPAP. The difference in point estimates between methods (machine minus manual) at any given time point ranged from -3.4 to 3.6 milliseconds in study 2 and from -3.3 to 5.5 milliseconds in study 3, the largest differences being observed at baseline in both studies.
Analysis of Outliers
The number and frequency of ECG recordings, which exceeded prospectively defined values for absolute QTcF and for change from baseline QTcF, are shown in Table IV for study 1 and in Table V for studies 2 and 3. There was no QTcF value, regardless of measurement technique, above 500 milliseconds and only 1 postdose value above 480 milliseconds (1 in 2171; 0.05% with the HPAP), and 1 change from baseline above 60 milliseconds (1 in 2171; 0.05% with HPAP). In fact, these 2 outliers relate to one ECG that was recognized as aberrant by the investigator but was not changed in the database. The proportion of QTcF measurements with values between 451 and 480 milliseconds ranged from 0% to 2.6% across techniques and studies and was somewhat higher with the 12SL algorithm than with the HPAP algorithm in study 1. In study 2, the manual technique generated a slightly higher proportion of recordings of both QTcF values above 450 milliseconds and changes from baseline of 30 milliseconds or more. The studies were not powered to detect small differences in the proportion of outliers, however, and no statistical analysis was performed.
|
|
In study 1, in which a full baseline day was also included, the change-from-baseline analysis was also calculated using the time-matched value at the corresponding clock time as baseline (eg, hour 8 at baseline was compared with hour 8 postdose to yield a change from baseline value, data not shown). This finding did not substantially affect the frequency of outliers (data not shown).
Bland-Altman Analysis
Uncorrected QT values derived from the HPAP QT algorithm in the Hewlett-Packard ECG machine were consistently smaller on average than the values derived from the same recording with a manual technique (mean difference for pairs of machine minus manual: -16, -19, and -19 milliseconds in studies 1, 2 and 3, respectively; Table VI). In contrast, the QT intervals as measured by the GE Marquette algorithm (12SL) generated, on average, larger values than the intervals measured manually (mean difference of 7 milliseconds in study 1).
|
The Bland-Altman plots (Figures 3, 4, and 5) show greater agreement between the HPAP and manual readings than between the 12SL and manual readings, which is demonstrated by the narrower limits of agreement (approximately 20 milliseconds with the HPAP and 30 milliseconds with the 12SL). It was also apparent that the difference between the HPAP and manual readings was consistent over the range of QT intervals. With the 12SL, the difference increased with increasing QT interval, with more machine QT intervals substantially longer than the manually generated ones. The one very noticeable outlier with the HPAP algorithm in Figure 3 was again the QT interval identified by the investigator as being aberrant but was not changed in the database.
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Obviously, studies have to be appropriately powered, which, dependent on the variability of the end point and the study design, may mean approximately 60 subjects for a crossover study or up to 80 subjects per treatment group for a parallel study.32 Because several doses of the investigational drug should be studied, this may result in a 4- or 5-way crossover design (placebo, positive control, and 2 to 3 doses of the drug). An important feature of the study is also to account adequately for the individual's diurnal variation of the QTc interval or potential diurnal patterns in the study. In a parallel study, this can be achieved through inclusion of a full 24-hour predose day with time-matched ECG recordings (at corresponding clock times as after dosing) and in a crossover study by comparison with the time-matched effect during the placebo period. With either approach, the number of ECG recordings is substantial, often ranging between 3300 and 6000, if single-dosing is appropriate (60 subjects, 1-10 ECGs at baseline and, for example, 10 after dosing in each of 5 periods = 11-20 ECGs/subject x 5 x 60). It has been convincingly demonstrated that averaging replicate ECG recordings from individual time points significantly decreases the intrasubject variability when either machine or manual techniques are employed. The use of replicates increases the number of recordings per subject but decreases the total sample size.13,32 With this number of ECG recordings, it is obvious that any technique that could be used in healthy volunteers, which would provide an automated measure of the QT interval with sustained quality and reliability, would offer a substantial advantage compared to more resource-intensive and potentially unreliable manual techniques.
Manual techniques have historically been preferred in studies of drug-induced QT effects14,17,29,35,36 and have clear benefits when used in patients with cardiovascular disease. However, it was unexpected QT increases observed with a standard manual technique,36 most likely caused by poorly controlled manual methods and between-reader bias and variability, that triggered our interest to look further into various methodologies. Subsequently, this analysis demonstrated that this bias introduced by reader variability essentially disappeared with automated techniques, with sustained capability to detect small QTc effects. In addition, we have previously shown substantial time-dependent within-reader variability when the same technicians performed the QT measurement 5 (11 milliseconds mean difference between readings) and 6 months (6 milliseconds difference) apart.14 Because the QT interval is inherently measured with error, regardless of the method of measurement, there can be no gold standard to identify any true QT interval. It is therefore difficult to demonstrate the accuracy of any technique or argue for 1 technique over another.
This analysis was based on ECG data using manual and machine-based techniques for QT interval measurements from 3 studies in healthy volunteers, which were all designed to detect small QT changes. The results demonstrate that different measurement techniques yield different QTc values but, more important, that automated techniques are capable of demonstrating small effects on the QTcF interval similar to those with manual readings and with the same clinical conclusions. The placebo-corrected, time-matched effect induced by the positive control (ketoconazole or moxifloxacin) determined as the point estimate and the 2-sided 90% confidence interval was similar across methods, as was the duration of the effect (Table III). The clinical conclusion that the positive control caused a small QTcF prolongation, exceeding around 5 milliseconds31 was the same regardless of the method.
There seem to be consistent differences between automated QT algorithms, which result in different absolute QT and QTcF intervals. In this analysis, Hewlett-Packard generated approximately 16 to 19 milliseconds shorter QT intervals than manually measured values, and 12SL generated approximately 7 milliseconds longer QT intervals than the manual technique (Table I and Figures 3, 4, and 5). The longer baseline QT values in healthy volunteers with the 12SL algorithm, in the range from 394 to 401 milliseconds, has previously been observed in a study using ibutilide as a positive control.37 The proportion of absolute QTcF greater than 450 milliseconds seemed to be somewhat higher with the 12SL algorithm compared with either manual or the HPAP algorithm, calculated on the number of recordings or the number of subjects. This finding is expected, as 12SL generated somewhat higher baseline values. It seems reasonable to assume that differences will also be observed between different manual techniques, which should be taken into consideration when comparing QT results across studies when different methods of measurement are used and when performing studies that are not placebo controlled.
The pairwise comparison between methods (Table VI and Figures 3, 4, and 5) demonstrated that the LOA between the manual technique and QT algorithm in the HPAP were between ±18 and ±21 milliseconds, which can be regarded as an acceptable difference.37 Furthermore, this difference did not seem to depend on the magnitude of the QT interval, as evidenced by a visual inspection of the Bland-Altman plots (Figure 3A).27,28 As the difference was consistent across the range of QT intervals, the mean difference between methods was negligible for the changes from baseline. Contrary to this finding, the comparison between manual and the GE Marquette QT algorithm, 12SL, showed a clearly wider interval between the LOAs (±31 milliseconds, Table VI and Figure 3B), and the difference seemed to increase with increasing QT interval, with more machine QT intervals substantially longer than the manually generated ones. This property of the algorithm is troublesome because the precision of the measurement decreases when the QT interval approaches ranges of clinical concern. Based on this observation and on comparisons with other studies,14,37 it is evident that automated ECG algorithms differ and must be validated individually. Furthermore, proprietary-automated ECG algorithms often are quite substantially changed, and it seems advisable to validate any such major change. Validation in this context should mean an approach similar to that used in this study, including the assessment of the effect of a mildly QT prolonging agent.
The use of automated algorithms may be potentially problematic when T wave morphologic abnormalities exist, with noisy signals, and when there are difficulties in differentiating between a U wave and the T wave. It has therefore been suggested that these techniques are not "sufficiently precise and robust to satisfy the precision required in the assessment of drug cardiac safety."13(p333) The current study is, as far as we are aware, the first of its kind. It compared a mild QT prolongation induced by a positive control in healthy volunteers, measured by manual techniques by specialized ECG laboratories and by fully automated QT algorithms included as standard software in ECG machines. Based on our results, we believe there is no firm basis for avoiding automated measurements in thorough QT studies performed in healthy volunteers with drugs that do not cause more marked T wave morphologic changes, provided validation is provided, as discussed above. A number of previously published studies comparing manual with automated ECG methods have showed a poor agreement between methods. These studies have, however, in many cases, been limited to patients with cardiovascular disease,38 to patients with congenital long QT syndrome,39 or to children.40 In a study in 10 healthy volunteers, Hoon and coworkers found a mean bias between an automated QT algorithm and manual measurements of 4 milliseconds, without intervention,41 which seems consistent with our current and previous14,37 findings.
Study 1 allowed an assessment of the incidence of false positive clinical conclusions, as 2 drug-free days were compared using a time-matched approach. Notably, using the manual technique, an effect approaching or even exceeding 10 milliseconds, could not be excluded at 2 time points (1 and 10 hours), despite narrow limits of the 95% confidence interval (±3-4 milliseconds). This spontaneous, drug-free variability challenges the 10-millisecond criteria outlined in the ICH E14 document to conclusively exclude a drug-induced effect on the QTcF interval.
Thorough QT studies are usually performed in healthy volunteers in a highly controlled setting such as a phase I unit, with continuous telemetry. Experienced and well-trained investigators, physicians, and other reviewers at these sites can provide an initial assessment of cardiac safety based on machine-read intervals. This can also serve as an assessment of T wave changes, provided the reviewer is sufficiently trained. The combination of automated interval measurement and investigator overread of T wave alterations at the study site seems to be a clinically and scientifically valid and cost-efficient approach to detect drug-induced effects on the QTc. In this study, only 0.11% (10 of 9500) of the recordings were overread and corrected by the investigator. The question remains as to whether any intervals measured by the machine should be overread at all, as the correction of only extremely large outlier values invariably would introduce a bias in postdose QTc intervals and changes from baseline. In our view, rereads should only be performed when the investigator clearly believes that the machine miscalculates the ECG interval, and an audit trace should be kept to show how and why these few intervals were changed.
There was no specific requirement in the protocol in any of the 3 studies with regard to the number of readers per subject for the manual measurements at the core ECG laboratory. As the degree and duration of the QT effect induced by the positive control (with limited published data for ketoconazole) are consistent with what we and others have observed in similar studies,34,42 it seems reasonable to believe that the data would not have been substantially altered by inclusion of such a requirement. Based on the examples given previously in this article, it seems prudent to recommend that the same primary reader and the same cardiologist should read all ECGs from one subject. In a crossover study, any bias or variability in the interval measurements or the clinical interpretations will thereby be distributed across all treatment groups. In a parallel study, the same primary reader and the same cardiologist should still read all ECGs from each subject to make consistent clinical conclusions for that subject and to guard against potential bias in case a subject receives more than 1 treatment (eg, a placebo run-in) or if the study drug is titrated within the same person. A reader by treatment interaction is less unlikely in a standard parallel study in which each subject receives only 1 treatment, even if the ECGs are sent to the core laboratory in batches because each batch is very likely to contain ECGs from every treatment group. If the dose is titrated up or if during the course of the study the treatments are discrete, it is important that if the ECGs are sent in batches, then the same set of readers be used throughout the study.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The studies were approved by East Kent Local Research Ethics Committee, Kent & Canterbury Hospital, United Kingdom; Independent Ethical Committee, Torlands, Devon, United Kingdom; Ravenscourt Ethics Committee, Ashford Hospital, United Kingdom; Clinical Research Ethic Committee, Kortenberg, Belgium; and Clinical Research Ethic Committee, Brussels, Belgium. The authors acknowledge the contributions of the following colleagues for the conduct and analysis of the studies: Richard Anziano, MS, Anna Crossland, MSc, John Davis, PhD, Frances Hackman, MSc, Virginie Herben, PhD, Dik W. H. Ng, PhD, Yuanjun Shi, PhD, George Weissberger, MD, and Nolan Wood, PhD, of Clinical Research & Development/Development Operations, Pfizer Inc, UK and USA; Rob Berman, MD, of Bristol-Myers Squibb; and Mike Stevens, BSc, of EMStat Ltd.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
1. Darpo B. Spectrum of drugs prolonging the QT interval and the incidence of torsades de pointes. Eur Heart J. 2001;3(suppl. K): 70-80.
2. Fenichel RR, Malik M, Antzelevitch C, et al. Drug-induced torsades de pointes and implications for drug development. J Cardiovasc Electrophysiol. 2004;15: 475-495.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
3. Shah RR. The significance of QT interval in drug development. Br J Clin Pharmacol. 2002;54: 188-202.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
4. Shah RR. Drug-induced prolongation of the QT interval: why the regulatory concern? Fundam Clin Pharmacol. 2002;16: 119-124.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
5. Shah RR. Drug-induced prolongation of the QT interval: regulatory dilemmas and implications for approval and labelling of a new chemical entity. Fundam Clin Pharmacol. 2002;16: 147-156.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
6. Yap YG, Camm AJ. Drug induced QT prolongation and torsades de pointes. Heart. 2003;89: 1363-1372.
7. Committee for Proprietary Medicinal Products. Points to Consider: The Assessment of the Potential for QT Interval Prolongation by Non-Cardiovascular Medicinal Products. 17 December 1997. London, United Kingdom: European Agency for the Evaluation of Medicinal Products; 1997.
8. Anderson ME, Al-Khatib SM, Roden DM, Califf RM. Cardiac repolarization: current knowledge, critical gaps, and new approaches to drug development and patient management. Am Heart J. 2002;144: 769-781.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
9. Haverkamp W, Breithardt G, Camm AJ, et al. The potential for QT prolongation and proarrhythmia by non-antiarrhythmic drugs: clinical and regulatory implications: report on a policy conference of the European Society of Cardiology. Eur Heart J. 2000;21: 1216-1231.
10. Andrassy G, Szabo A, Dunai A, et al. Acute effects of cigarette smoking on the QT interval in healthy smokers. Am J Cardiol. 2003;92: 489-492.[Medline] [Order article via Infotrieve]
11. Bednar MM, Harrigan EP, Anziano RJ, Camm AJ, Ruskin JN. The QT interval. Prog Cardiovasc Dis. 2001;43: 1-45.[Web of Science][Medline] [Order article via Infotrieve]
12. Frederiks J, Swenne CA, Kors JA, et al. Within-subject electrocardiographic differences at equal heart rates: role of the autonomic nervous system. Pflugers Arch. 2001;441: 717-724.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
13. Malik M, Camm AJ. Evaluation of drug-induced QT interval prolongation: implications for drug approval and labelling. Drug Saf. 2001;24: 323-351.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
14. Sarapa N, Morganroth J, Couderc JP, et al. Electrocardiographic identification of drug-induced QT prolongation: assessment by different recording and measurement methods. Ann Noninvasive Electrocardiol. 2004;9: 48-57.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
15. Honig PK, Wortham DC, Zamani K, Conner DP, Mullin JC, Cantilena LR. Terfenadine-ketoconazole interaction: pharmacokinetic and electrocardiographic consequences. JAMA. 1993;269: 1513-1518.
16. Morganroth J, Brown AM, Critz S, et al. Variability of the QTc interval: impact on defining drug effect and low-frequency cardiac event. Am J Cardiol. 1993;72: 26B-31B.[CrossRef][Medline] [Order article via Infotrieve]
17. Morganroth J, Silber SS. How to obtain and analyze electrocardiograms in clinical trials: focus on issues in measuring and interpreting changes in the QTc interval duration. Ann Noninvasive Electrocardiol. 1999;4: 425-433.[CrossRef]
18. Morganroth J. Focus on issues in measuring and interpreting changes in the QTc interval duration. Eur Heart J. 2001;3(suppl K): 105-111.
19. Choy AM, Darbar D, Dell'Orto S, Roden DM. Exaggerated QT prolongation after cardioversion of atrial fibrillation. J Am Coll Cardiol. 1999;34: 396-401.
20. Houltz B, Darpo B, Edvardsson N, et al. Electrocardiographic and clinical predictors of torsades de pointes induced by almokalant infusion in patients with chronic atrial fibrillation or flutter: a prospective study. Pacing Clin Electrophysiol. 1998;21: 1044-1057.[CrossRef][Medline] [Order article via Infotrieve]
21. Houltz B, Darpo B, Swedberg K, et al. Comparison of QT dispersion during atrial fibrillation and sinus rhythm in the same patients, at normal and prolonged ventricular repolarization. Europace. 2000; 2: 20-31.
22. Roden DM. Taking the "idio" out of "idiosyncratic": predicting torsades de pointes. Pacing Clin Electrophysiol. 1998;21: 1029-1034.[CrossRef][Medline] [Order article via Infotrieve]
23. Dota C, Skallefell B, Edvardsson N, Fager G. Computer-based analysis of dynamic QT changes: toward high precision and individual rate correction. Ann Noninvasive Electrocardiol. 2002;7: 289-301.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
24. Willems JL, Abreu-Lima C, Arnaud P, et al. The diagnostic performance of computer programs for the interpretation of electrocardiograms. N Engl J Med. 1991;325: 1767-1773.[Abstract]
25. Fridericia LS. Die Systolendauer im Elektrokardiogramm bei normalen Menchen und bei Herzkranken. Acta Med Scand. 1920; 53: 469-486.
26. Prineas R, Crow RS, Blackburn H. The Minnesota Code Manual of Electrocardiographic Findings: Standard and Procedures for Measurement and Classification. London: John Wright; 1982.
27. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986; 1: 307-310.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
28. Bland JM, Altman DG. Comparing methods of measurement: why plotting difference against standard method is misleading. Lancet. 1995;346: 1085-1087.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
29. Fleishaker JC, Francom SF, Herman BD, Knuth DW, Azie NE. Lack of effect of reboxetine on cardiac repolarization. Clin Pharmacol Ther. 2001;70: 261-269.[Medline] [Order article via Infotrieve]
30. Redfern WS, Carlsson L, Davis AS, et al. Relationships between preclinical cardiac electrophysiology, clinical QT interval prolongation and torsade de pointes for a broad range of drugs: evidence for a provisional safety margin in drug development. Cardiovasc Res. 2003;58: 32-45.
31. International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use. The Clinical Evaluation of QT/QTc Interval Prolongation and Proarrhythmic Potential for Non-Cardiovascular Drugs: E14. Brussels, Belgium: International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use; 2005.
32. Patterson S, Agin M, Anziano R, et al. Investigating drug-induced QT and QTc prolongation in the clinic: a review of statistical design and analysis considerations. Report from Pharmaceutical Research and Manufacturer of America QT statistics expert team. Drug Inf J. 2005;39: 243-265.[Web of Science]
33. Yao JA, Du X, Lu D, Baker RL, Daharsh E, Atterson P. Estimation of potency of HERG channel blockers: impact of voltage protocol and temperature. J Pharmacol Toxicol Methods. 2005;52: 146-153.[CrossRef][Medline] [Order article via Infotrieve]
34. Chaikin P, Gillen MS, Malik M, Pentikis H, Rhodes GR, Roberts DJ. Co-administration of ketoconazole with H1-antagonists ebastine and loratadine in healthy subjects: pharmacokinetic and pharmacodynamic effects. Br J Clin Pharmacol. 2005;59: 346-354.[CrossRef][Medline] [Order article via Infotrieve]
35. Morganroth J, Lyness WH, Perhach JL, et al. Lack of effect of azelastine and ketoconazole coadministration on electrocardiographic parameters in healthy volunteers. J Clin Pharmacol. 1997; 37: 1065-1072.[Abstract]
36. Morganroth J, Hunt T, Dorr MB, Magner D, Talbot GH. The effect of terfenadine on the cardiac pharmacodynamics of sparfloxacin. Clin Ther. 1999;21: 1514-1524.[Medline] [Order article via Infotrieve]
37. Azie NE, Adams G, Darpo B, et al. Comparing methods of measurement for detecting drug-induced changes in the QT interval: implications for thoroughly conducted ECG studies. Ann Noninvasive Electrocardiol. 2004;9: 166-174.[CrossRef][Web of Science][Medline] [Order article via Infotrieve]
38. Eskola MJ, Nikus KC, Voipio-Pulkki LM, et al. Comparative accuracy of manual versus computerized electrocardiographic measurement of J-, ST- and T-wave deviations in patients with acute coronary syndrome. Am J Cardiol. 2005;96: 1584-1588.[Medline] [Order article via Infotrieve]
39. Miller MD, Porter C, Ackerman MJ. Diagnostic accuracy of screening electrocardiograms in long QT syndrome I. Pediatrics. 2001;108: 8-12.
40. Blair J, Scahill L, State M, Martin A. Electrocardiographic changes in children and adolescents treated with ziprasidone: a prospective study. J Am Acad Child Adolesc Psychiatry. 2005;44: 73-79.[Medline] [Order article via Infotrieve]
41. Hoon TJ. Performance of an electrocardiographic analysis system: implications for pharmacodynamic studies. Pharmacotherapy. 1996;16: 230-236.[Medline] [Order article via Infotrieve]
42. Sanofi-synthelabo SL770499 (alfuzosin) briefing package. Food and Drug Administration Web site. Available at: http://www.fda.gov/ohrms/dockets/ac/03/briefing/3956B1_03_Sanofi-synthelabo.pdf.
CiteULike 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  N. Sarapa, S. F. Francom, S. M. Azzam, P. K. Wickremasingha, and B. Tyl Detection of Clinically Significant QTc Prolongation in Phase I Studies in Healthy Participants: Comparison of 2 Semiautomated QT Measurement Methods J. Clin. Pharmacol., January 1, 2009; 49(1): 103 - 108. [Full Text] [PDF]
|
|
|
|
|
|
M. Malik, K. Hnatkova, J. Ford, and D. Madge Near-Thorough QT Study as Part of a First-In-Man Study J. Clin. Pharmacol., October 1, 2008; 48(10): 1146 - 1157. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. van Noord, W. M van der Deure, M. C J M Sturkenboom, S. M J M Straus, A. Hofman, T. J Visser, J. A Kors, J. C M Witteman, and B. H C. Stricker High free thyroxine levels are associated with QTc prolongation in males J. Endocrinol., July 1, 2008; 198(1): 253 - 260. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Sarapa, J. L. Mortara, B. D. Brown, L. Isola, and F. Badilini Quantitative Performance of E-Scribe Warehouse in Detecting Quality Issues With Digital Annotated ECG Data From Healthy Subjects J. Clin. Pharmacol., May 1, 2008; 48(5): 538 - 546. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Sen Does the Choice of QTc Formula Matter? Anesth. Analg., August 1, 2007; 105(2): 536 - 536. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
