This was a phase 3, multicenter, double-blind (third-party unblinded) trial of adult patients who were candidates for or had undergone a laparotomy, laparoscopy, or percutaneous drainage of an intra-abdominal abscess and had a known or suspected diagnosis of complicated intra-abdominal infection. All patients were hospitalized at the time of study entry. Before screening of the first patient, the protocol was reviewed and approved by the institutional review board or ethical review committee at each participating center. Written informed consent was obtained from each patient or his or her legal representative before the start of any study procedures. The trial was conducted in accordance with the Declaration of Helsinki.

Patients were stratified at randomization into 2 groups based on their scores on APACHE II: ≤15, or >15 but <31. Using a 1:1 ratio, patients were randomly assigned to receive either tigecycline (initial 100-mg dose given by intravenous [IV] infusion over a 30-minute period, followed by 50 mg IV every 12 hours) or IV imipenem/cilastatin (500 mg/500 mg every 6 hours or dose-adjusted based on weight and creatinine clearance). Patients randomized to tigecycline received a 100 mL normal saline intravenous infusion 6 hours after active drug each day in order to maintain the blind. Unless the patient was a clinical failure (see definition below), the duration of study drug therapy ranged from 5 to 14 days.

Study drug was administered only when there was a strong suspicion (ie, elevated white blood cell count, elevated band cell counts [ie, evidence of a "shift to the left"], fever, or highly suggestive radiographic findings) or a confirmed diagnosis of an intra-abdominal infection (presence of pus within the abdominal cavity), and a baseline intra-abdominal culture was obtained from the site of infection. Patients could be enrolled before drainage of the intra-abdominal infection and may have received up to 2 doses of study drug before the baseline cultures were obtained. Patients did not receive more than 1 dose (or combination) of parenteral nonstudy antibacterial drugs after the baseline intra-abdominal cultures were obtained. However, wound irrigation solutions of sterile water or normal saline and topical antiseptics were permitted throughout the course of the study.

The clinical status of the intra-abdominal infection was assessed at serial visits throughout the study by the presence or absence of the following signs and symptoms: fever; localized or diffuse abdominal wall rigidity or involuntary guarding; abdominal tenderness or pain; ileus or hypoactive bowel sounds; nausea or vomiting. The clinical response to study drug was determined by the investigator. At the test-of-cure visit (14–35 days after therapy), each patient's response was categorized as one of the following: Cure – the course of study drug and the initial intervention (operative and/or radiologically guided drainage procedure) resolved the intra-abdominal infectious process; Failure – the patient required additional antibacterial therapy other than the study drug, the patient required additional surgical or radiologic intervention to cure the infection, death due to infection occurred after 48 hours of therapy, the patient received an extended course of study drug (ie, >120% of the planned number of doses), or the patient was prematurely discontinued from study drug due to an adverse event (after receiving at least 8 doses in 5 days) and required additional antibiotic therapy or surgical intervention; and Indeterminate – the patients was lost to follow-up, or died within 48 hours after the first dose of study drug for any reason, or died after 48 hours because of noninfectious-related reasons (as judged by the investigator).

Baseline aerobic and anaerobic cultures from the primary intra-abdominal site of infection and two sets of blood cultures were obtained within 24 hours of the first dose of study drug. All aerobic and anaerobic bacterial isolates, regardless of the source of cultured material, were identified and tested at a central laboratory (Covance Central Laboratory Services, Inc., Indianapolis, IN, or Geneva, Switzerland) by using a standard procedure approved by the National Committee of Clinical Laboratory Standards (NCCLS) Subcommittee on Antimicrobial Susceptibility Testing. For tigecycline, provisional minimum inhibitory concentration (MIC) breakpoints were used (susceptible ≤2 mg/L; intermediate 4 mg/L; resistant ≥8 mg/L).

Based on the results of the baseline intra-abdominal culture, the susceptibilities of identified organisms, and the clinical outcome of the patient, the investigator also determined the microbiologic response at the patient level and at the isolate level. Microbiologic response by patient was categorized at the test-of-cure visit as eradication, persistence, superinfection (ie, the emergence of a new isolate was documented at the site of infection with worsening signs and symptoms of infection). The microbiologic response for each baseline isolate at the test-of-cure visit was described according to the following definitions: eradication, persistence, or indeterminate. Because many patients did not have follow-up cultures, many microbiologic responses both at the patient and isolate level were categorized as either presumed eradication or presumed persistence.

All patients who received at least one dose of study drug were evaluated for safety (modified intent-to-treat [mITT] population). Safety was assessed from serial medical history and physical examinations, reports of clinical adverse events, and findings from routine electrocardiograms (ECGs), and serum chemistry, hematology, coagulation, and urinalysis tests. Adverse events were recorded throughout the study period, up to and including the test-of-cure visit. Before unblinding, the investigator categorized the severity of each adverse event and the potential for relationship to study drug. Serious adverse events (ie, those that were life-threatening, led to prolongation of the existing hospitalization, caused persistent or significant disability or incapacity, or death) were also recorded.

Several subpopulations of patients were assessed for safety, clinical, and bacteriologic outcomes. Patients who satisfied the inclusion/exclusion criteria were included in the intent-to-treat (ITT) population, whereas the subset of patients who received at least 1 dose of study drug made up the mITT population. Those patients in the mITT population who had clinical evidence of a complicated intra-abdominal infection, by meeting the minimal disease criteria, and had a confirmed baseline isolate made up the microbiological-modified (m-mITT) population. From this latter group, the microbiologically evaluable (ME) population was defined as those who met all inclusion/exclusion criteria; had at least 5 days of therapy; did not receive concomitant antibiotics after the baseline intra-abdominal culture was obtained through the test-of-cure visit; had a test-of-cure visit 14 to 35 days after the first dose of study drug; and had a baseline intra-abdominal culture containing at least one causative isolate that was susceptible to both study drugs. If these criteria were not met at any time during the study, the patient was declared non-evaluable and the outcome of cure/failure/indeterminate was analyzed within the m-mITT population. Patients were considered nonevaluable for inclusion in the ME population if death occurred or if they withdrew from the study <48 hours after the first dose of study drug.

The primary endpoints of the study were clinical response at the test-of-cure visit (14–35 days after therapy) for the m-mITT and ME populations. Secondary analyses included bacteriologic response at the test-of cure visit by patient and isolate, as well as clinical response rates stratified as monomicrobial versus polymicrobial, and by isolate.

Statistical analysis was performed by the Clinical Biostatistics department of Wyeth Research, Collegeville, PA. Categorical baseline demographic and medical variables were analyzed using the Fisher exact test. Continuous variables were compared using a one-way analysis of variance (ANOVA) model with treatment as a factor. Between-group comparisons of adverse events were analyzed by using the Fisher exact test. For laboratory tests, vital signs, and ECG results, within-group changes from baseline were analyzed by using a paired t-test and between-group comparisons were made by using the analysis of covariance, adjusting for baseline value. The difference between treatment groups in the percentage of premature withdrawal from study drug was evaluated by using a 2-sided Fisher exact test.

The noninferiority efficacy of tigecycline compared with imipenem/cilastatin was evaluated for clinical and microbiologic responses by using a 2-sided 95% confidence interval (CI) for the true difference in efficacy (tigecycline minus imipenem/cilastatin) adjusted for the stratification variable APACHE II score and corrected for continuity. Noninferiority was concluded if the lower limit of the 2-sided 95% CI was greater than or equal to -15%. For all subpopulation analyses (eg, monomicrobial versus polymicrobial infection), an adjusted difference between treatment groups with its 95% CI was calculated from a generalized linear model with a binomial probability function and an identity link (SAS^® Proc GENMOD). Interaction effects were tested at the 0.10 level of significance. With the planned sample size (n = 788) and an evaluability rate of 50%, the trial had a power of at least 90% to determine the noninferiority of tigecycline compared with imipenem/cilastatin.

The demographic characteristics for the 502 ME patients were comparable between the two treatment groups (Table 1). The study population was of mixed racial/ethnic background with whites (41.8%) and Hispanics (19.5%) represented most often. There was a predominance of men (67.5%) and the mean age of enrolled patients was 43 years old. Complicated appendicitis (59%) was the most common intra-abdominal infection diagnosis, followed by perforated intestine (8.8%) and gastric/duodenal ulcer (4.6%). No significant differences between the treatment groups were observed in the number or types of infections diagnosed at baseline. The severity of intra-abdominal illness was similar in each treatment group (mean APACHE II score was ~5.7).

For the ME population, clinical cure rates were 80.6% for tigecycline and 82.4% for imipenem/cilastatin (95% CI -9.0, 5.4; Table 2). Corresponding clinical cure rates for the m-mITT population were 73.5% and 78.2% (95% CI -11.8, 2.3), respectively. For both the ME and m-mITT populations, tigecycline was efficacious and statistically noninferior to imipenem/cilastatin. Multiple subgroup analyses of clinical responses (eg, age, sex, race, geographic location) found consistently efficacious clinical responses between the treatment groups. No significant treatment differences in clinical response were observed between the two treatment groups when patients were stratified by the number of isolated baseline isolates (Table 2). For the ME population, tigecycline had a 89.8% clinical cure rate at the test-of-cure visit for monomicrobial infections and a 75.3% clinical cure rate for polymicrobial infections. Similar rates were observed for recipients of imipenem/cilastatin (88.5% and 78.1%, respectively).

For complicated appendicitis, the most frequent diagnosis, clinical cure rates at the test-of-cure visit was 84.2% for tigecycline and 86.2% for imipenem/cilastatin (Table 3). In both treatment groups, lower clinical cure rates (≤72%) were observed in patients who had intra-abdominal abscess, complicated diverticulitis, or intestinal perforation (Table 3). Overall, there were no significant differences in clinical cure rates between tigecycline and imipenem/cilastatin based on primary intra-abdominal diagnosis. A total of 14 tigecycline- and 27 imipenem/cilastatin-treated patients in the ME population had a positive pretherapy blood culture. Clinical cure in patients with bacteremia was reported for 71.4% of tigecycline and 74.1% of imipenen/cilastatin recipients.

For the ME population, eradication of intra-abdominal isolates at the patient level was reported for 80.6% of tigecycline- and 82.4% of imipenem/cilastatin-treated patients (95% CI -9.0, 5.4), indicating that tigecycline was efficacious and statistically noninferior to imipenem/cilastatin (Table 4). No significant differences between the treatment groups were found when eradication rates were stratified by monomicrobial versus polymicrobial infection (Table 4).

Generally, eradication rates at the test-of-cure visit for the most commonly isolated intra-abdominal isolates were similar between the two treatment groups (Table 5). For E. coli, the most commonly isolated aerobe, eradication rates were 80.4% for tigecycline versus 83.5% for imipenem/cilastatin. Corresponding eradication rates for Klebsiella spp, the second most frequently isolated gram-negative aerobe, were 87.1% and 85.7%, respectively. A total of 6 ESBL-producing E. coli and 7 ESBL-producing K. pneumoniae isolates were identified pretherapy. The majority of these isolates were eradicated by tigecycline: 83% (5/6) and 71% (5/7), respectively. Eradication rates for Bacteroides fragilis were 69.8% for tigecycline and 72.5% for imipenem/cilastatin.

Pretherapy in vitro activity against baseline isolates for tigecycline and imipenem/cilastatin are shown in Table 6. The mean MIC₉₀ for tigecycline against the most commonly isolated aerobes and anaerobes was ≤2.0 mg/L. No pretherapy isolates displayed resistance to tigecycline based on the provisional breakpoints used. Bacterial susceptibilities to tigecycline appeared to be consistent with clinical responses.

Data from all patients in the mITT population (n = 825) were analyzed for safety. The m-ITT population received a median of 7 days of tigecycline or imipenem/cilastatin treatment. Regardless of study drug causality, the frequency and distribution of treatment-emergent adverse events occurring in at least 3% of patients in either treatment group were similar to those observed in the imipenem/cilastatin treatment group. The majority of these adverse events were related to study medication (56%) and were mild to moderate in intensity (94%). Digestive system (56.9% vs 49.8%, P = 0.043), nausea (31.0% tigecycline, 24.8% imipenem/cilastatin; P = 0.052), vomiting (25.7% tigecycline, 19.4% imipenem/cilastatin; P = 0.037), and diarrhea (21.3% tigecycline, 18.9% imipenem/cilastatin; P = 0.435) were the most frequently reported adverse events in both treatment groups. The majority of patients in both treatment groups experienced mild to moderate nausea and/or vomiting (94%). There was no significant difference between the treatment groups in the number of patients who required antiemetic therapy for nausea and/or vomiting. No tigecycline-treated patient has a positive Clostridium difficile toxin assay, nor developed C. difficile associated diarrhea.

In the tigecycline group, infections (13.6% vs 7.5%, P = 0.006), hypoproteinemia (8.0% vs 4.1%, P = 0.028), and dyspnea (6.8% vs 2.9%, P = 0.014) were statistically higher than in the imipenem/cilastatin treatment group. The difference in infection rates between the treatment groups was primarily due to the development of secondary wound infections. No apparent trends or risk factors were identified in the development of secondary wound infections in either treatment group.

One hundred forty-five (145) patients had one or more serious adverse events during the study (81 [19.6%] tigecycline, 64 [15.5%] imipenem/cilastatin) (P = 0.143). The most frequently reported serious adverse events were abnormal healing (14 tigecycline, 6 imipenem/cilastatin), abscess (10 tigecycline, 8 imipenem/cilastatin), and infection (10 tigecycline, 9 imipenem/cilastatin). Significantly more patients treated with tigecycline (6 [1.5%]) versus none treated with imipenem/cilastatin reported pneumonia as a serious adverse event (P = 0.031).

Adverse events were the primary reason for early withdrawal of study drug. A total of 27 (6.5%) tigecycline- and 15 (3.6%) imipenem/cilastatin-treated patients discontinued treatment prematurely because of an adverse event (P = 0.080). A total of 10 (2.5%) tigecycline- and 4 (1.0%) imipenem/cilastatin-treated patients stopped therapy prematurely secondary to either nausea (6 tigecycline, 2 imipenem/cilastatin) and/or vomiting (4 tigecycline, 2 imipenem/cilastatin). There were no significant differences between treatment groups in any single adverse event leading to the discontinuation of study drug.

Twenty-nine (29) patients died during the study: 17 patients in the tigecycline group and 12 patients in the imipenem/cilastatin treatment group. Only two of the deaths, both in the tigecycline group, were considered by the investigators to be possibly related to study drug secondary to treatment failure. The first patient was a 78 year old female who received tigecycline for one week. Two days following discontinuation of therapy the patient developed septic shock; she died one day later. The second patient, a 23 year old female, presented with sepsis and received 3 days of tigecycline therapy. On day 3 she was found to have pneumonia and progressed to multiple organ failure with sepsis and died the same day.

Few clinically important or unexpected changes in any routine hematologic or serum chemistry tests, vital signs, or ECG data were associated with the use of tigecycline or imipenem/cilastatin. However, significantly more patients treated with imipenem/cilastatin (312/410, 76.1%) than those treated with tigecycline (275/408, 67.4%) had 1 or more laboratory findings of potential clinical importance (P = 0.007). Imipenem/cilastatin-treated patients had significantly lower serum potassium (≤3 mmol/L; P = 0.004), phosphorus values (≤ 0.8 mmol/L; P < 0.001), and lymphocytes values (≤0.6 cells × 10⁹/L; P < 0.001). Yet, significantly more patients treated with tigecycline than those treated with imipenem/cilastatin had clinically significant hypoproteinemia (≤35 g/L; P = 0.001). No significant changes in QTc interval were observed in either treatment group at any time point.