Randomized, Double-Blind, Placebo- and Positive-Controlled Crossover Study of the Effects of Omadacycline on QT/QTc Intervals in Healthy Subjects

Omadacycline, an aminomethylcycline, is an antibiotic that is approved in the United States for once-daily intravenous (i.v.) and oral use for treatment of adults with acute bacterial skin and skin structure infections and community-acquired bacterial pneumonia.

720 ms-with a mean range of 375 ms. The observed range of RR intervals used to calculate the QTcI covered on-treatment RR ranges in 14 (22.2%) and 15 (24.6%) subjects in the omadacycline 100-and 300-mg i.v. groups, respectively. When day Ϫ1 baseline QT/RR data were used for derivation of optimized QTcI, the RR range by subject was substantially wider-between 800 and 2,995 ms-with a mean range of 1,175 ms. This RR range covered the on-treatment RR ranges in all subjects for both doses.
The abilities to remove heart rate dependence were compared among correction methods ( Table 2). QTcS and QTcI resulted in the lowest mean sum of squared slopes (SSS), which was 0.0012 under drug-free conditions (placebo). Since QTcS resulted in lower mean SSS values on active doses, this correction method was selected as the primary endpoint.
Single doses of 100 and 300 mg omadacycline did not have an effect on the QTc interval. The pattern of mean change-from-baseline QTcS (ΔQTcS) closely followed that   A linear exposure-response model provided an acceptable fit to the observed concentration and QT data. The relationship between the individual observed omadacycline concentrations and placebo-corrected ΔQTcS is visualized in the top panel in Fig. 4. The goodness-of-fit plot (Fig. 4, bottom panel) shows the mean placebo-corrected ΔQTcS (90% CI) within each omadacycline concentration decile and the modelpredicted mean placebo-corrected ΔQTcS with 90% CI. From the plot, it can be seen that the predicted placebo-corrected ΔQTcS values are relatively close to the observed values, indicating that the proposed linear model provides an acceptable representa- tion of the relationship between placebo-corrected ΔQTcS and omadacycline concentrations without accounting for hysteresis. The estimated slope of the concentration-ΔQTcS relationship was very shallow at 0.0007 ms per ng/ml (90% CI, 0.0000 to 0.0014), with a treatment effect-specific intercept of Ϫ0.6 ms (90% CI, Ϫ1.2 to 0.11). It can also be seen from the plot that the possibility of an effect on placebo-corrected ΔQTcS exceeding 10 ms can be excluded at omadacycline plasma levels up to ϳ8 g/ml. Safety/tolerability. Adverse events were generally mild: more subjects had events of infusion-site pain following treatment with omadacycline 300 mg (11%) than with omadacycline 100 mg (6%), with rates for placebo and moxifloxacin of 0% and 2%, respectively. Infusion-site rash was reported in 5% of subjects following omadacycline 300 mg; no such events followed the other treatments. Headache was reported with an incidence of 3% to 7% following each treatment. There were no serious adverse events. One subject discontinued the study due to hives after receiving omadacycline 300 mg. This event was considered of moderate intensity and treated with oral diphenhydramine; it resolved ϳ1 h after onset. No abnormal laboratory findings were reported.

DISCUSSION
Prolongation of the QT interval is a well-documented effect of certain antibiotics, including fluoroquinolones, ketolides, and macrolides (10,13,14). For some of these drugs (e.g., telithromycin and azithromycin), postapproval warnings about potential cardiac toxicity were added to the product labeling (15). In addition, based on results of a thorough QTc study (16), telavancin carries a warning about QT interval prolongation in its product labeling. Thus, it is important to assess new antibiotics for their effects on ventricular repolarization.
Omadacycline is approved in the United States for treatment of adults with ABSSSI and CABP. For these indications, following an initial "loading" dose of 200 mg i.v. over 60 min or 100 mg i.v. over 30 min twice daily on day 1, omadacycline is administered once daily (either i.v. or orally), and the therapeutic i.v. dose is 100 mg (7,8,10). In healthy volunteers given a single 100-mg i.v. dose of omadacycline, mean C max was 1.8 to 1.9 g/ml (standard deviation, 0.4 to 0.7) (6,17).
In this thorough QTc study, a therapeutic (100 mg i.v.) dose and a supratherapeutic (300 mg i.v.) dose of omadacycline had no effect on the QT interval. Since the possibility of an effect on placebo-corrected ΔQTcS above 10 ms could be excluded at all postdosing time points with both doses, the findings clearly represent negative QT study results as defined in the International Conference on Harmonisation E14 guidance (12). Using exposure-response analysis, as suggested in the recently revised E14 guidelines, the slope of the relationship between omadacycline concentrations in plasma and placebo-corrected ΔQTcS was very shallow-only 0.0007 ms per ng/ml. Consequently, estimating the effect throughout the full range of observed plasma levels, the possibility of an effect on placebo-corrected ΔQTcF exceeding 10 ms can be excluded up to omadacycline concentrations in plasma of ϳ8 g/ml, which is ϳ4.4-fold above the mean C max (ϳ1.8 g/ml) in patients given the therapeutic dose of 100 mg i.v. over 30 min. In patients receiving the 200-mg i.v. dose, mean C max levels would be somewhat higher (ϳ2.2 g/ml), but the safety margin is still severalfold greater than that seen with concentrations that may lead to clinically relevant QT prolongation. It therefore seems highly unlikely that omadacycline will cause clinically concerning QT prolongation.
A clear and quite marked effect on heart rate was observed with mean peak effects of 16.8 and 21.6 bpm after the i.v. infusion of 100 mg and 300 mg, respectively. In such cases, it is important to evaluate which method for correction represents the best way to remove the heart rate dependence of the QTc interval. Tested in the manner proposed by a team from the U.S. Food and Drug Administration (18), the differences as shown by the SSS across the five methods applied to the data (QTcF, QTcS, QTcI, optimized QTcI, and QTcP) were small. As a consequence, the result of the QT evaluation was negative (i.e., upper bound of 90% CI of placebo-corrected ΔQTc of Ͻ10 ms) with all correction methods. Interestingly, QTcS and QTcI removed the heart rate dependence of the QT interval somewhat better than other methods using placebo data, but not consistently on omadacycline treatment. At the highest dose, 300 mg i.v., population-based methods (QTcF and QTcP), optimized QTcI and QTcS all worked better than QTcI. In our view, this observation argues for the importance of testing the ability of heart rate correction methods to remove the heart rate dependence of the QTc interval (19) and goes against the view that an individualized QTc method is by default always the best way to evaluate QT effect with drugs that have a pronounced heart rate effect (20).
When a heart rate effect of this magnitude is observed in healthy subjects, it becomes important to evaluate the effect in patients, to place the finding into a clinical context. It is not obvious that patients with infections who are febrile and experiencing stress would react the same way as healthy subjects when exposed to a drug that provides effective treatment for the infection. In a phase 3 study in adult patients with ABSSSI (7), no clinically significant heart rate changes were observed with omadacycline. In patients with CABP, the population was older and preexisting cardiovascular disease more frequent than were seen with the population in the infected-skin studies (21). Baseline heart rate was higher, and it seems reasonable to assume that the patients who were older with CABP-and who were, at times, hypoxic-were in greater distress than those with infected skin and skin structure. ECGs were recorded before and 30 to 90 min after the start of the 30-min infusion of the first and third doses. A small ΔHR (4.3 bpm) was observed after the first dose of omadacycline. With continued effective antibacterial therapy, no such effect was seen before or after the third dose, but the reduction of mean ΔHR was somewhat smaller than in the moxifloxacin group (1.8 and -1.1 versus -5.4 and -6.8 bpm, respectively). Importantly, throughout the omadacycline development program, very few outliers in terms of pronounced heart rate effects (patients with heart rate of Ͼ120 bpm and ΔHR of Ͼ15 bpm) have been observed, with no notable differences compared to active comparators. The observed increase in heart rate was small and transient and would not be expected to lead to adverse cardiac events. Such events, e.g., myocardial ischemia and episodes of decompensated heart failure, were few (0% to 0.3% and 0% to 0.8%, respectively) and occurred at rates similar to the observed incidence for comparators (0.3% to 0.5% for linezolid and moxifloxacin) (7,8).
Nonclinical studies performed with omadacycline have demonstrated that the drug inhibits the M 2 subtype of the muscarinic acetylcholine receptor, resulting in a nonadrenergic, vagolytic effect (22). In isolated rabbit sinus node preparations, omadacycline did not affect the intrinsic rate of the sinus node and, interestingly, reversed the bradycardia caused by cholinergic stimulation with carbamylcholine. These nonclinical results are consistent with the clinical observation that the heart rate effect of omadacycline is more pronounced in subjects with greater vagal tone and relatively low resting heart rate (i.e., in healthy subjects than in patients suffering from disease). A possible explanation for the difference observed to date between heart rate effects in healthy subjects and in patients with infections might be that in patients with active infections, the cholinergic impact on heart rate is attenuated through the effects of fever, dehydration, and systemic hypermetabolic manifestations of inflammation and distress with increased sympathetic tone.
In summary, these results demonstrate that omadacycline, at doses of 100 and 300 mg i.v., does not prolong the QTc interval. The observed heart rate effect in healthy subjects seems much less pronounced in patients with infections and is therefore not clinically concerning. The ability to place omadacycline in a category of low cardiac risk will be an important component to the overall benefit-risk assessment for this new antibiotic.

Study design.
This was a single-dose, double-blind, randomized, four-way crossover study with an enrollment period extending from September 2008 to January 2009. The study was conducted in accordance with the ethical principles relating to biomedical research involving human subjects as adopted by the 18th World Medical Association General Assembly, Helsinki (1964) and subsequently amended. It was also conducted in accordance with local laws and regulations for the use of investigational therapeutic agents. The study protocol and all amendments were reviewed by the Institutional Review Board for the study center (PRACS Institute, Ltd.). The study investigator reviewed the study, and written information was provided to eligible subjects. Informed consent was obtained from each subject in writing before randomization.
Because the omadacycline i.v. solution had a yellow tint and the infusions were of different duration (30 or 60 min), all infusion bags were covered for anonymization and infusions were administered by personnel who did not participate in any other study procedures or assessments.
Subject selection. Healthy men and women aged 18 to 45 years were eligible, using standard criteria for clinical pharmacology studies. Women of childbearing potential had to use an effective birth control method from the time of screening until 30 days after the study; they also had to have had a negative pregnancy test at screening and on the day before dosing. Subjects were excluded for the following reasons: a history of allergy to any tetracycline or quinolone antibiotic; body weight of Ͻ45 or Ͼ100 kg; body mass index value of Ͻ18 or Ͼ30 kg/m 2 ; systolic or diastolic blood pressure of Ͼ140 or Ͼ90 mm Hg, respectively; use of any investigational drug within 1 month prior to enrollment; clinically significant electrolyte abnormality; a history of cardiovascular disease, family history of QT prolongation, or any other medical condition or ECG abnormality that could interfere with the conducting of the study. Treatment groups. Subjects were randomized to 1 of 4 treatment sequences, with the same treatments delivered in different orders. Subjects received an omadacycline therapeutic dose (100 mg i.v., infused over 30 min), an omadacycline supratherapeutic dose (300 mg i.v., infused over 60 min), a placebo infusion (negative control), and moxifloxacin (400 mg orally; active control). Moxifloxacin 400 mg or placebo capsules were given orally at the start of a 60-min placebo infusion. The omadacycline therapeutic dose corresponded to the proposed therapeutic dose for the phase 3 clinical trial of omadacycline in ABSSSI; the supratherapeutic dose was chosen because of good tolerability at this dose in phase 1 studies of omadacycline in healthy, young males.
Study procedures. The study was performed on an inpatient basis, and in each treatment period, subjects were admitted to the clinical site on day Ϫ2, 2 days before the day of dosing (day 1), and were discharged after completion of safety procedures on the day after dosing (day 2). The washout between consecutive study periods was Ն7 days. Study treatment was administered with subjects in the fasting state in the morning of day 1.
In treatment period 1, continuous 12-lead ECG data were recorded for 24 h on day Ϫ1, and subjects underwent a graded exercise test using a modified Bruce protocol with a target heart rate of Ͼ90 bpm 24 h and 25 min before dosing. On day 1 in all treatment periods, a continuous 12-lead ECG recording procedure was performed from 1.5 h before dosing to 24 h after dosing. The 12-lead ECGs were extracted in triplicate at three time points (1.5, 1.0, and 0.5 h) before dosing and at 20, 35, 50, and 65 min and 1.5, 2, 4, 6, 12, 18, and 22 h after dosing. ECG intervals were measured by a central ECG laboratory with a semiautomated technique.
Blood samples were obtained before dosing, at time points similar to those used for the postdose ECG extractions, and at 48, 72, and 96 h after the dose to measure omadacycline plasma levels and determine pharmacokinetic parameters, including C max , time to maximum concentration, and AUC 0 -24h . Blood samples were obtained 5 min after each time point indicated for ECG measurement.
Statistical analysis. Sample size was calculated based on the following assumptions: (i) the intrasubject standard deviation for change-from-baseline QTcF (ΔQTcF) would be 7 ms; (ii) the underlying QT effect for omadacycline would be 4 ms; (iii) ΔQTcF would be determined at 11 ECG time points per dose and at two doses. Based on these assumptions, a sample size of 50 subjects would provide more than 90% overall power to demonstrate that the upper bound of each one-sided 95% CI falls below 10 ms for up to 11 time points. Allowing for potential dropouts, a total of 64 subjects were to be randomized.
Baseline for each period was defined as the average of the measured QTc intervals from the three ECG time points (Ϫ1.5, Ϫ1.0, and Ϫ0.5 h) recorded before dosing in that period on day 1. The primary endpoint was the placebo-corrected change-from-baseline QTc (ΔQTc) determined using Fridericia's correction (i.e., the placebo-corrected ΔQTcF) with QTcF ϭ QT/RR 1/3 , unless a substantial peak heart rate (i.e., a mean placebo-corrected change-from-baseline HR [ΔHR] of Ͼ10 bpm in the "by time point" analysis) was observed in either of the two omadacycline treatment groups. In such cases, the following correction methods for QTc were to be explored and tested for their ability to remove the heart rate dependence.
Method I. Method I employed an individualized HR-corrected QT interval (QTcS) calculated from QT/RR data obtained at supine resting time points on day Ϫ1 in the first treatment period. Based on QT/RR pairs from all subjects, the QTcS correction coefficient was derived from a linear mixed-effects model as follows: log(QT) ϭ log(a) ϩ b ϫ log(RR) (with gender included as a fixed effect and subject included as a random effect for both intercept and slope). The coefficient of log(RR) for each subject, b i , was then used to calculate QTcS for each subject as follows: QTcS ϭ QT⁄RR b i .
Method II. Method II employed an individualized HR-corrected QT interval (QTcI) derived from QT/RR data obtained from a broader range of heart rates by also using ECG data extracted during time points of a graded exercise, in addition to the time points described above for day Ϫ1. Based on QT/RR pairs, the correction coefficient was derived, separately for each subject, from a linear regression model as follows: log(QT) ϭ log(a) ϩ b ϫ log(RR). The coefficient of log(RR) for each subject, b i , was then used to calculate QTcI for that subject as follows: Method III. Method III employed a population HR-corrected QT interval (QTcP) derived from the same data as QTcS from all subjects. Based on QT/RR pairs from all subjects, QTcP was derived from a linear regression model as follows: log(QT) ϭ log(a) ϩ b ϫ log(RR). The coefficient of log(RR), b, was then used to calculate QTcP for each subject as follows: QTcP ϭ QT/RR b .
Method IV. Method IV employed an optimized HR-corrected QT interval (optimized QTcI) derived from a broader range of HRs by using all QT/RR data from day Ϫ1 of the first treatment period, i.e., QT/RR pairs from the full baseline 24-h recording. The QT/RR pairs from each subject were used for that subject's individual correction coefficient, which was derived from a linear regression model as follows: log(QT) ϭ log(a) ϩ b ϫ log(RR). The coefficient of log(RR) for each subject, b i , was then used to calculate QTcI2 for that subject as follows: QTcI2 ϭ QT⁄RR b i .
For each method, the relationship between QTc and RR interval was then investigated using on-treatment data (omadacycline, moxifloxacin, and placebo) and a linear regression model as follows: QTc ϭ c ϩ d ϫ RR. Mean QTc and RR values from all nominal time points (including predose) were used. The RR coefficient for each subject, d i , was then used to calculate the average SSS for each of the different QT-RR correction methods. The QTc method for which the average on-treatment slope value was closest to zero (i.e., the lowest average SSS value) for omadacycline and placebo was then selected as the primary endpoint (18).
The by-time-point analysis for QTc was based on a linear mixed-effects model with change-frombaseline QTc for the selected primary endpoint as the dependent variable; period, sequence, time (categorical), treatment (omadacycline, moxifloxacin, and placebo), and time-by-treatment interaction as fixed effects; and baseline QTc as a covariate. Subject was included as a random effect for the intercept. An unstructured covariance matrix was specified for the repeated measures at postdose time points for subject within the treatment period. From this analysis, the least-squares mean and two-sided 90% CIs were calculated for the contrast "omadacycline versus placebo" at each dose of omadacycline and each postdose time point, separately. For secondary endpoints (assay sensitivity and other ECG parameters), the same model was used. Assay sensitivity requirements were deemed to have been met if the placebo-corrected ΔQTc value was significantly above 5 ms at 1.5, 2, and 4 h postdose, controlling for multiplicity (23).
The relationship between the omadacycline concentrations in plasma and ΔQTc was investigated based on a linear mixed-effects model with ΔQTcF as the dependent variable, time-matched concentration of omadacycline as a continuous covariate (i.e., 0 for placebo), centered baseline QTc as an additional covariate, treatment (active ϭ 1 or placebo ϭ 0) and time as categorical factors, and a random intercept and slope per subject. The appropriateness of a linear model and, in cases of a QT effect in the by-time-point analysis, the presence of hysteresis (time delay between peak drug concentrations in plasma and peak QT effect) were tested.