Colistin Monotherapy versus Combination Therapy for Carbapenem-Resistant Organisms

The OVERCOME trial was a randomized, double-blind, placebo-controlled trial conducted at hospitals in the United States, Thailand, Taiwan, Israel, Greece, Italy, and Bulgaria from October 2012 to August 2020 (ClinicalTrials.gov number, NCT01597973). The institutional review board at each site approved the protocol, and all participants, or their authorized representatives, provided written informed consent. The trial protocol, provided with the full text of this article at evidence.nejm.org, was designed by the principal investigator (K.S.K.) and approved by the National Institutes of Health and the U.S. Food and Drug Administration. The trial was performed in accordance with the principles of the Declaration of Helsinki.¹¹ The data were obtained by the site investigators and coordinators (provided in the Supplementary Appendix, available with the full text of this article at evidence.nejm.org), and analysis was performed by the statistical team (G.W.D., L.S., and K.L.) in collaboration with the first and last authors on the publication (K.S.K. and J.M.P.). The authors assume responsibility for the accuracy of the data and analyses and the trial results presented. The initial draft of the manuscript was prepared by J.M.P. and K.S.K. and reviewed and edited by all coauthors, and all authors agreed to publish the manuscript.

Patients were eligible for enrollment if they were 18 years of age or older and had pneumonia and/or BSI caused by XDR A. baumannii, XDR P. aeruginosa, or CRE, with in vitro susceptibility to colistin (colistin minimum inhibitory concentration [MIC] ≤2 mg/l)¹² (Table 1). Patients were ineligible if they received 72 hours or more of polymyxin treatment within 96 hours of enrollment or if they had a life expectancy of 24 hours or less. The trial protocol lists the full inclusion and exclusion criteria (Section 5.2), as well as detailed definitions of trial pathogens (Section 5.2.1), trial infections (BSI and pneumonia, Section 5.2.3), and microbiologic methods (Section 7.2.3).

Patients were enrolled in the trial based on carbapenem and colistin susceptibility test results performed at the participating site. Confirmatory susceptibility testing by broth microdilution (BMD) for all available isolates was performed either at trial sites or the central laboratory (details are provided in the trial protocol, Sections 7.2.1 and 7.2.2). If isolates were found to be colistin susceptible at the trial site by a method other than BMD and the patient received on-trial treatment but the index pathogen was subsequently found to be colistin resistant by BMD at the central laboratory, the patient was removed from the modified intention-to-treat (mITT) analysis (described in the following section) but was still assessed in the all-treated population.

All participants received intravenous colistin (loading dose of 5 mg/kg followed by 1.67 mg/kg every 8 hours). Participants were randomly assigned to receive an intravenous carbapenem or a matching normal saline placebo. Imipenem (500 mg every 6 hours as a 30-minute infusion) was used initially, as the National Institutes of Health requirements allowed only for the use of generic medications, and four patients were randomly assigned to this treatment. Once meropenem became generically available, the protocol was amended, and in November 2013, imipenem was switched to meropenem (1000 mg every 8 hours as a 30-minute infusion) because of meropenem’s ease of administration and improved safety profile. Trial medications were renally dose adjusted by trial personnel as detailed in the trial protocol (Sections 6.2 and 6.3). The duration of trial drug therapy was 7 to 14 days and determined by the treating physician. If it was determined that treatment was needed beyond 14 days, therapy was continued off-trial at the discretion of the treating physician, and the participant was categorized as a clinical failure.

Eight separate balanced block randomization sequences were used for each participating hospital. Strata were defined according to infection type, organism confirmation status (confirmed or presumptive), and, for patients with pneumonia only, severity of illness (measured by using the Acute Physiology and Chronic Health Evaluation II [APACHE II] score,¹³ with patients stratified according to scores <25 vs. ≥25). APACHE II has a range of scores from 0 to 71, and an APACHE II score of 25 is associated with a predicted in-hospital mortality of 55%. No additional stratification was made for the pneumonia group, but patients with BSI were further stratified according to primary (the BSI is not secondary to other sources) versus secondary (the BSI is secondary to infection at a distal site) bacteremia.¹⁴ Varying block sizes were used within each stratum. The eight randomization sequences for each site were generated by using nQuery Advisor version 4.0, GraphPad Software. A binder with the eight sequences was used by each hospital’s unblinded research pharmacist.

The ITT population consisted of all randomly assigned patients. The mITT population included all randomly assigned patients with a trial pathogen who received at least one dose of trial medication and whose trial pathogens exhibited colistin susceptibility according to BMD (Fig. 1). The all-treated population included all patients in the mITT population as well as patients initially randomly assigned to a trial arm who received more than 48 hours of a trial drug but were subsequently excluded from the mITT population due to identification of a colistin-resistant trial pathogen according to central laboratory BMD testing. For patients with BSI, blood cultures were obtained daily until they were negative for 2 consecutive days. For patients with pneumonia, respiratory cultures were obtained daily (when feasible) until they were negative for 2 consecutive days.

An independent data and safety monitoring committee was established to regularly review data and make recommendations regarding continuation of the trial. A clinical research organization was used to ensure the integrity of data, trial protocol process, and trial medication accountability.

The primary outcome was all-cause 28-day mortality in the mITT population. Secondary efficacy outcomes, including clinical failure and microbiologic cure, were evaluated in patients who survived more than 48 hours after enrollment. Patients who were withdrawn from the study before completing trial therapy for any reason other than meeting clinical failure criteria were deemed indeterminant for clinical failure and were subsequently excluded from this end point (Fig. 1). Clinical failure was a composite end point defined as meeting any of the following: death either while on trial therapy or within 7 days following completion; receipt of rescue therapy for the trial pathogen within 7 days of completion of trial treatment; removal from the trial due to an adverse event considered related to trial treatment; bacteremia more than 5 days after initiation of trial treatment for patients with BSI; or failure to improve or worsening of oxygenation by the end of trial treatment in patients with pneumonia. For patients with pneumonia who were mechanically ventilated at baseline, improvement of oxygenation was defined as any of the following occurring by the end of trial therapy: removal from mechanical ventilation, an increase in the ratio of arterial oxygen partial pressure to the fraction of inspired oxygen (PaO₂/FiO₂) by 100 mm Hg, or an increase in PaO₂/FiO₂ to a value 300 mm Hg or higher. For patients who remained on mechanical ventilation at the end of trial treatment without a recorded PaO₂/FiO₂ ratio on that date, recorded ratios from ±2 days from end of therapy were used to assess oxygenation status at the end of therapy, with preference given to the value closest to the end of therapy. If no values were available from within this time window, patients were indeterminant for this outcome. For patients with pneumonia who were not on mechanical ventilation at baseline, worsening was defined as requiring mechanical ventilation at the end of trial treatment.

Microbiologic cure was only assessed for patients in whom repeat specimens were obtained and was defined as eradication of pathogen from the infection site (i.e., negative cultures) by the end of treatment. For trial patients with pneumonia who completed a full course of trial therapy but did not have a documented negative respiratory culture, microbiologic outcomes were determined by the date of their last recorded positive respiratory sample according to the following criteria. If the patient’s last respiratory culture was positive and was collected 5 or more days after trial enrollment, they were considered a microbiologic failure. If the patient’s last respiratory culture was positive and obtained 4 days or less after enrollment, the patient was classified as indeterminant for microbiologic cure, as no inference could be made with respect to eradication at the end of therapy. Furthermore, if a patient was removed from the trial early due to death or treatment failure before documented eradication of the pathogen from the respiratory sample, they were categorized as a microbiologic failure.

Safety outcomes evaluated included decreases in renal function, as well as neurologic and hypersensitivity reactions. Acute kidney injury (AKI) was defined according to the RIFLE (Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease) criteria.¹⁵ All patients who received a trial drug for 48 hours or more with a baseline creatinine clearance of 30 ml/min or higher were eligible for AKI assessment, regardless of whether they remained in the mITT population. Details of the case definition of other adverse events¹⁶ are provided in the Supplemental Appendix.

The target sample size, a total of at least 422 patients with known mortality status, was selected to give 80% power to detect a 28% relative reduction in mortality with combination therapy, assuming 50% mortality in the monotherapy arm. One interim analysis with a P value of 0.003 required for stopping was planned when one half (n=211) of the target sample size was enrolled. Stopping for futility was planned if the conditional power under the alternative dropped below 5%. The primary mortality analysis used a chi-square test, and associated analyses included Kaplan–Meier survival curve estimation and log-rank testing. Cox regression was used for secondary survival analysis. The t tests or Wilcoxon rank-sum tests were used to assess other continuous or ordinal variables as appropriate.

For secondary outcomes, no prespecified multiple comparison adjustment was included in the statistical analysis plan, and thus no significance tests are presented for those outcomes. Instead, only the estimated percentage-point differences ([colistin plus placebo] − [colistin plus meropenem]) and the associated 95% confidence intervals (CIs) are presented. The widths of those CIs have not been adjusted for multiplicity. Thus, the CIs should not be used to reject or not reject treatment effect hypotheses. No data were missing for the primary outcome or for the secondary outcomes beyond what has been previously described. For clinical failure, patients who were indeterminant in terms of the improvement in oxygenation criteria were assessed based on the other components of the composite outcome. Those who met another component of clinical failure were considered clinical failures, whereas those who did not meet any other failure criteria were considered not to be failures. Patients who were indeterminant for the microbiologic cure of pneumonia outcome were excluded from the analysis of this outcome.

A total of 467 patients were enrolled, 464 were randomly assigned to the trial, and 423 met inclusion criteria and comprised the mITT population (Fig. 1). Twenty-six patients who were randomly assigned to a trial arm were treated for 48 hours or more based on non-BMD colistin susceptibility testing performed at the site that was later determined to be colistin resistant when tested via BMD by the central laboratory or at the site. These patients were subsequently removed from the mITT group. Results between treatment arms, including these patients (i.e., the all-treated population; n=449), are provided in Table S2. Other reasons for removal of randomly assigned patients from the mITT are detailed in the Supplementary Appendix, and described in Fig. 1.

The mITT population included 213 patients who received colistin in combination with placebo and 210 who received colistin in combination with a carbapenem (n=206 meropenem and n=4 imipenem [hereafter referred to as colistin plus meropenem]). Baseline characteristics of patients in the two treatment arms are listed in Table 2 and Table S1. The mean age of the cohort was 68.0±16.0 years; 158 (37%) were female, 216 (51%) were White, and 173 (41%) were Asian. The most frequent comorbidities were hypertension (n=262 [62%]), diabetes (n=167 [39%]), and chronic lung disease (n=163 [39%]). Most patients were critically ill, with 291 (69%) in an intensive care unit at the time of enrollment. The trial population was highly reflective of the population most affected by these infections globally (Table S4).

Pneumonia was the most common type of index infection (n=298 [70%]), which included 23 patients with a concomitant BSI (Table 2 and Table S1). The remaining 125 patients were enrolled with a BSI. The most common causative pathogen was A. baumannii (n=329 [78%]) followed by CRE (n=69 [16%]) and P. aeruginosa (n=43 [10%]). Seventeen patients (4%) in the cohort had multiple trial pathogens as the cause of their index infection, and 93 (22%) were coinfected with a nontrial pathogen. The median time from onset of index infection to administration of appropriate therapy (i.e., therapy with in vitro activity against the trial pathogen) was 3 days (interquartile range, 2 to 4 days). Before enrollment, 290 patients (69%) received a median of 1 day of treatment with a drug active against the trial pathogen. When a patient received a nontrial drug active against the trial pathogen before enrollment, the drug in almost all cases was colistin or polymyxin B (n=288 [99%]), and 52 (18%) patients who received an active, nontrial drug before enrollment received colistin in combination with a carbapenem for a median duration of 1 day before enrollment (Table S1).

There was no significant difference in 28-day mortality between the colistin monotherapy and colistin plus meropenem combination therapy trial groups (43 vs. 37%; P=0.17) (Table 3), and there was no difference in time to mortality (Fig. 2). Similarly, there was no difference in mortality between treatment arms in any infection type or pathogen subgroup. When assessing 28-day mortality according to time to appropriate therapy, or when excluding patients who received pre-enrollment colistin plus carbapenem combination therapy or who were coinfected with nontrial pathogens, there remained no difference between treatment arms (Table S2).

A total of 374 patients were eligible for analysis of the clinical failure end point (Fig. 1). There was no difference in clinical failure between the monotherapy and combination therapy trial groups (65 vs. 58%; difference, 6.8 percentage points; 95% CI, −3.1 to 16.6), nor were there any differences between treatment arms in any infection type or pathogen subgroup (Table 4 and Table S3). In analyses accounting for the impact of time to appropriate therapy or pre-enrollment receipt of colistin plus carbapenem combination therapy on clinical failure, there remained no difference in clinical failure rate between treatment arms.

A total of 334 patients were eligible for analysis of the microbiologic end point (Fig. 1). Microbiologic cure was shown in 106 (65%) of 163 patients receiving monotherapy compared with 103 (60%) of 171 receiving combination therapy (difference, 4.8 percentage points; 95% CI, −5.6 to 15.2). There were no differences in microbiologic cure between treatment arms in terms of infection type or infecting pathogen (Table 4).

A similar number of patients receiving monotherapy and combination therapy were removed from the trial due to adverse events (12 of 213 [6%] vs. 21 of 210 [10%], respectively; P=0.094). Most of these events (70%) were AKI. Overall, AKI rates were similar between patients receiving monotherapy versus those receiving combination therapy (88 of 170 [52%] vs. 85 of 174 [49%]; P=0.55). Neurotoxicity (4%) and hypersensitivity reactions (1%) were infrequent and occurred at similar rates in both treatment arms (Table 4).