This phase 2 study was a dose-ranging, randomized, double-blind, noncontrolled study conducted at 12 sites (8 enrolled patients) in the United States. The study was approved by institutional review boards, and informed consent was given to all patients before enrollment. The study was conducted in accordance with International Conference on Harmonization (ICH) and FDA guidelines, good clinical practice (GCP), and the Declaration of Helsinki.
Male or female in-hospital patients or outpatients 18 to 75 years old diagnosed with cSSSI caused by a suspected or confirmed Gram-positive pathogen were eligible for enrollment. Infections under study were abscesses (with at least 2 cm of surrounding induration or requiring incision and drainage), surgical or posttraumatic wounds, and deep extensive cellulitis. All patients were required to have at least two of the following signs and symptoms: purulent or seropurulent drainage, erythema, fluctuance, heat or warmth, pain or tenderness, swelling or induration, or requirement for surgical drainage. Additionally, at least one sign of systemic infection was required: oral temperature of >38°C, white blood cell count of >10,000 cells/mm3, or >10% immature neutrophils. Systemic signs were not required in patients with lesions greater than 5 cm in diameter.
Excluded diagnoses were diabetic foot infection, gangrene, perirectal abscess, burns, decubitus or ischemic ulcers, necrotizing infection, infection at a central catheter site or near a prosthetic device, or presence of metastatic infection such as septic arthritis, endocarditis, or osteomyelitis. Patients were excluded for the following reasons: creatinine clearance of <52 ml/min estimated by Cockroft-Gault formula, hepatic disease (aspartate transaminase [AST] or alanine aminotransferase [ALT] of >3 times the upper limit of normal [ULN] or bilirubin of >1.5 times the ULN or alkaline phosphatase of >3 times the ULN), human immunodeficiency virus (HIV) infection with a CD4 count of <200 cells/mm3, neutropenia with absolute neutrophil counts of <1,000 cells/mm3, Bazett-corrected QT interval (QTCB) of >450 ms in males or 470 ms in females, or body mass index of >35 kg/m2. More than 24 h of antibiotic therapy within 96 h prior to randomization was prohibited unless the patient was considered a failure after at least 48 h of therapy. The use of serotonergic agents, sympathomimetic amine derivatives, chronic systemic corticosteroids the equivalent of ≥10 mg/day of prednisone, for more than 14 days, or a high-tyramine diet was prohibited. Patients with a history of hypertensive crises, uncontrolled hypertension, migraine headaches, gastrointestinal resection, advanced alcohol-related disease, uncontrolled diabetes, chronic systemic immunosuppressive therapy, known or suspected bacteremia, or any life-threatening condition were also excluded from enrollment.
The study drug was supplied as 200 mg of torezolid phosphate disodium salt capsules and could be taken with or without food. Molecular weights are 370.34 for torezolid, 450.32 for torezolid phosphate-free acid, and 494.28 for torezolid phosphate disodium salt. (Torezolid phosphate-free acid is the formulation to be used in future clinical trials.)
Patients were randomized 1:1:1 using a central interactive voice response system (IVRS) to receive 200, 300, or 400 mg of oral torezolid phosphate disodium salt (equivalent to 149.9, 224.8, and 299.7 mg of torezolid) once daily for at least 5 but not more than 7 days. Participants and investigators were blind to treatment assignment. Duration of therapy was determined by the investigator based on the patient's response, including resolution or improvement of signs and symptoms and improvement of abnormal inflammatory markers. No adjunctive antimicrobial therapy was allowed. Patients enrolled before culture results were available and found to require Gram-negative antimicrobial coverage were discontinued from the study.
The primary objective of the study was to determine the clinical response rate of each dose group at the test-of-cure (TOC) visit in the clinically evaluable (CE) and clinical modified-intent-to-treat (cMITT) populations. Secondary objectives included cure rates at the end-of-therapy (EOT) visit, microbiological response rates, the safety profile of each dose group, and the characterization of absorption and disposition of torezolid using population PK modeling.
Directly observed therapy and clinical assessments were performed as follows: on screening/day 1; days 2, 3, and 5; at an EOT visit; at a TOC visit (7 to 14 days posttreatment); and at a late follow-up (LFU) visit (21 to 28 days posttreatment, in office or by telephone). Patients were allowed to take the study drug at home on day 4 and day 6. At each visit, a thorough examination of the cSSSI site, vital signs assessment, and physical examination were performed; samples for chemistry and hematology laboratory testing were sent to the central reference laboratory (Eurofins Medinet, Chantilly, VA), and assessment of concomitant medications and adverse events were completed. An electrocardiogram (ECG) was performed at screening and repeated at the EOT visit. Samples for microbiology testing were evaluated by a local laboratory, with confirmatory identification and susceptibility testing performed at the central laboratory.
The investigators assessed the clinical outcome at both the EOT and TOC visits. Clinical cure was defined as resolution of the infection or improvement of signs and symptoms of the cSSSI such that no further treatment was required. A clinical failure was defined as either persistence or incomplete resolution of cSSSI or development of new signs and symptoms such that further antibiotic therapy was required, unplanned surgical intervention was necessary after 48 h on therapy, a new diagnosis of osteomyelitis was made, a treatment-limiting adverse event leading to discontinuation of the study drug was identified, or death due to cSSSI was determined. An indeterminate outcome was assessed under the following circumstances: a treatment change before at least two doses of study medication, death unrelated to cSSSI, osteomyelitis at baseline (diagnosed after enrollment), or isolation of a Gram-negative organism at baseline that required treatment or when the patient was lost to follow-up before the TOC visit.
All patients receiving the study drug were included in the modified intent-to-treat (MITT) population and evaluated for safety. Safety assessments included reviews of vital signs, physical examinations, ECG findings, laboratory evaluations, and adverse events. An independent cardiologist performed a blinded overread of all ECGs.
This study was not statistically powered to determine differences between dose groups. The sample size chosen was to provide clinically meaningful information on safety, tolerability, and efficacy as well as the PK profile of each of three torezolid phosphate dose levels. All safety and efficacy data for the study were summarized using descriptive statistics for each study population. Five populations were defined for analysis: MITT, randomized patients receiving at least one dose of study medication; cMITT, patients in the MITT population with a diagnosis of cSSSI; microbiological modified intent-to-treat (mMITT), patients in the cMITT with a Gram-positive bacterial pathogen isolated at baseline; CE, patients receiving the minimum requirement of study drug, having a clinical assessment of success or failure at the TOC visit, and having no other confounding events or factors preventing assessment of outcome; and microbiologically evaluable (ME), patients in both the CE and mMITT populations.
Blood sampling and drug assay.
Serial blood samples for population PK analysis were obtained at two time points. For the first sampling, a median of two samples (range, one to four) per patient was obtained between 19 to 71 h after the first dose. For the second sampling, a median of five samples (range, one to six) per patient was obtained between 41 to 123 h after the first dose. A previous study (3
) demonstrated that torezolid phosphate was rapidly and completely converted in vivo
to torezolid, the prodrug active metabolite; therefore, only the latter was considered in the PK analysis. Torezolid was extracted from plasma by acetonitrile (ACN)-methanol-0.1% formic acid protein precipitation (5:4:1, vol/vol/vol). The supernatant was diluted 1:3 with high-performance liquid chromatography (HPLC)-grade deionized water; 50 μl of the resulting solution was analyzed using a validated liquid chromatographic method consisting of an isocratic step of ACN-0.05% trifluoroacetic acid (25:75, vol/vol) on a Higgins Analytical Targa C18
column and positive-ion heated nebulizer mass spectrometric detection. The analytical method was selective for torezolid and was linear (using 1/x
fit with mean r2
values of 0.9996 ± 0.0002) over the concentration range of 0.16 to 1,000 ng/ml, with a lower limit of quantification (LLQ) of 0.16 ng/ml. Mean analyte recovery values ranged from 110 to 117% (standard deviation, ± 4 to 18%). Intra-assay accuracy values for the quality controls ranged from 90.4 to 98.0% with precision (coefficient of variation [CV]) values between 1.46 and 6.90%. Interassay accuracy values for the quality controls ranged from 94.9 to 102%, with CV values between 3.29 to 5.50%.
Models with one, two, or three disposition compartments were assessed. Absorption of the prodrug torezolid phosphate and conversion of torezolid phosphate to torezolid were modeled by a first-order process with or without a lag compartment. Additionally, a previously developed semiphysiological absorption model (8
) was evaluated. As torezolid phosphate did not achieve quantifiable concentrations in plasma, we simplified the structural model in that we did not include the plasma concentrations of the prodrug as a dependent variable.
To explore a potential saturation of torezolid elimination, we considered various models for drug elimination. These included models with first-order elimination, mixed-order (Michaelis-Menten) elimination, parallel mixed-order and first-order elimination, and autoinhibition of clearance. Similar to the model with penetration of drug into an inhibition compartment proposed by Plock et al. (31
) for linezolid, the model we assessed included an inhibition compartment. The plasma concentration of torezolid was assumed to stimulate the input into the inhibition compartment, and this compartment subsequently inhibited the clearance of torezolid.
Parameter variability model.
An exponential parameter variability model was used to describe the between-subject variability (BSV) and between-occasion variability (BOV) of PK parameters. The BSV was modeled by a block-diagonal variance-covariance matrix with one block for the disposition parameters and one block for the absorption parameters. We included BOV for the absorption but not for the disposition parameters.
Residual error model.
Residual unidentified variability was explained by an additive plus proportional error model.
Population PK modeling was conducted in S-ADAPT (version 1.56) using an importance sampling method (pmethod=4 in S-ADAPT) of the Monte Carlo parametric expectation maximization (MC-PEM) algorithm. To bridge between an intensively sampled phase 1 study in healthy volunteers (data not shown here) and the present phase 2 study in patients but with fewer samples, we employed a three-stage hierarchical approach (hprior=1 option in S-ADAPT). Compared to a standard population PK analysis, the three-stage hierarchical approach has the advantage of borrowing information from healthy volunteers for the purposes of analyzing patient data and accounting for uncertainty in the prior means and prior variability estimates. The ability to incorporate uncertainty in the prior information is the most important advantage of the three-stage hierarchical method compared to a maximum a posteriori (MAP) Bayesian approach. The MAP Bayesian method assumes that the prior means and prior variability estimates are known with certainty and are identical between the prior data (from healthy volunteers in this case) and the patient data to be analyzed. These assumptions are not made by the three-stage hierarchical approach.
Sensitivity analysis for priors.
We assessed the impact of the extent of uncertainty for the prior means and prior variability estimates on the final parameter estimates. As the ratio of the area under the concentration-time curve for free drug to the MIC (f
AUC/MIC) is the most predictive PK/pharmacodynamic index for torezolid (23
), this sensitivity analysis focused on clearance (CL). We evaluated the following cases: case A, with the uncertainty taken directly from the estimates for healthy volunteers; case B, which is the same as case A but with uncertainty for CL and for BSV of CL increased by 16-fold (on variance scale); case C, which is the same as case B but with uninformative priors for the residual error parameters; case D, which is the same as case C but with uninformative priors on all absorption parameters; and case E, no priors (i.e., standard population PK analysis).
Model selection and qualification.
The average objective function (−2 × log-likelihood) during the last 10 iterations, individual curve fits, and the standard diagnostic plots were used for model selection. Predictive performance was ensured via visual predictive checks as described previously (7