Comments
Abstract
Rationale: Optimization of β-lactam antibiotic dosing for critically ill patients is an intervention that may improve outcomes in severe sepsis.
Objectives: In this individual patient data meta-analysis of critically ill patients with severe sepsis, we aimed to compare clinical outcomes of those treated with continuous versus intermittent infusion of β-lactam antibiotics.
Methods: We identified relevant randomized controlled trials comparing continuous versus intermittent infusion of β-lactam antibiotics in critically ill patients with severe sepsis. We assessed the quality of the studies according to four criteria. We combined individual patient data from studies and assessed data integrity for common baseline demographics and study endpoints, including hospital mortality censored at 30 days and clinical cure. We then determined the pooled estimates of effect and investigated factors associated with hospital mortality in multivariable analysis.
Measurements and Main Results: We identified three randomized controlled trials in which researchers recruited a total of 632 patients with severe sepsis. The two groups were well balanced in terms of age, sex, and illness severity. The rates of hospital mortality and clinical cure for the continuous versus intermittent infusion groups were 19.6% versus 26.3% (relative risk, 0.74; 95% confidence interval, 0.56–1.00; P = 0.045) and 55.4% versus 46.3% (relative risk, 1.20; 95% confidence interval, 1.03–1.40; P = 0.021), respectively. In a multivariable model, intermittent β-lactam administration, higher Acute Physiology and Chronic Health Evaluation II score, use of renal replacement therapy, and infection by nonfermenting gram-negative bacilli were significantly associated with hospital mortality. Continuous β-lactam administration was not independently associated with clinical cure.
Conclusions: Compared with intermittent dosing, administration of β-lactam antibiotics by continuous infusion in critically ill patients with severe sepsis is associated with decreased hospital mortality.
There is a mechanistic rationale for continuous infusion of β-lactam antibiotics to provide improved patient outcomes compared with intermittent dosing. However, all available studies have been statistically underpowered to test mortality benefits.
In this individual patient data meta-analysis of a large critical care patient population with severe sepsis, we found that administration of β-lactam antibiotics by continuous infusion was associated with decreased hospital mortality compared with intermittent dosing.
Serious infections in critically ill patients are associated with high morbidity and mortality (1, 2). Early recognition and prompt administration of appropriate antibiotic therapy is a standard of care in such patients (3–5). Choice of antibiotic is important to ensure that the antibiotic spectrum includes known or potential pathogens. However, there is uncertainty about the most effective method of drug dosing and administration, despite recognized associations between the achievement of target antibiotic concentrations and improved patient-centered outcomes (5–8).
β-Lactam antibiotics are widely used to treat patients in critical care units (9, 10). They are time-dependent antibiotics with which maintenance of concentrations above the minimum inhibitory concentration (MIC) of the pathogen is associated with maximal bacterial killing (11, 12). Pathophysiological perturbations in critically ill patients may reduce the time that β-lactam antibiotic concentrations are maintained above the MIC when these drugs are administered by intermittent dosing (5, 13). Administering β-lactam antibiotics by continuous infusion results in sustained concentrations throughout the dosing interval, increased time above the MIC, and enhanced bacterial killing (14, 15). Despite some clinical uptake, the efficacy of continuous infusion of β-lactam antibiotics on patient-centered outcomes, such as interval mortality, has not been demonstrated in high-quality randomized controlled trials (RCTs) to date. In a recent aggregated data meta-analysis of RCTs comparing continuous and intermittent infusions of β-lactam antibiotics in hospitalized patients, researchers reported no significant differences in mortality (16). However, that meta-analysis included patients with nonsevere sepsis and had significant heterogeneity.
We aimed to combine patient-level data from RCTs to compare interval mortality and clinical cure rates of continuous versus intermittent infusion of β-lactam antibiotics in critically ill patients with severe sepsis. In addition, we sought to identify patient subgroups that may benefit from administration by continuous infusion.
This meta-analysis followed the Preferred Reporting Items for Systematic reviews and Meta-Analyses guidelines (17). We conducted a literature review in PubMed in November 2015 to identify all completed prospective RCTs in which investigators reported patient outcomes with continuous and intermittent infusion of β-lactam antibiotics in critically ill patients with severe sepsis. Studies were identified using keywords including “β-lactam,” “penicillin,” “cephalosporin,” “carbapenem,” or “monobactam”; “severe sepsis,” “septic shock,” “ICU,” or “critically ill”; and “continuous infusion,” “prolonged infusion,” or “extended infusion.” We hand searched reference lists of the identified papers for additional studies. There were no exclusions made for language of publication. We manually checked all reference sources containing any mention of continuous infusion of β-lactam antibiotics to determine if they reported prospective RCTs. Retrospective studies, pharmacokinetic studies, studies of patients not meeting criteria for severe sepsis, studies with nonequivalent dosing between treatment groups, review articles, editorials, surveys, and protocol papers were excluded. We then undertook full-text reviews of the remaining studies, and study methodology details were documented. Studies were included in this meta-analysis if they (1) were prospective, (2) enrolled patients with severe sepsis or septic shock (18), (3) randomized patients to receive intermittent or continuous administration of one or more β-lactam antibiotics with equivalent (or blinded clinician) dosing in each treatment arm, and (4) reported assessment of outcomes by a clinician blinded to treatment allocation. We invited authors of identified studies to participate in this analysis and provide original, deidentified individual patient data from their trials.
We used the Cochrane Collaboration tool for evaluation of the risk of bias to assess study quality (19). Specifically, the tool was used to assess whether there was adequate generation of randomization sequences, concealment of treatment allocation, masking of assessors, and appropriate methods for addressing missing data. Two researchers (J.A.R. and M.-H.A.-A.) independently conducted quality assessments and agreed on the final categorization. We tested for publication bias by examining a funnel plot produced by using the trim-and-fill procedure with the Review Manager version 5.3 program (The Nordic Cochrane Centre, Copenhagen, Denmark).
All included studies received ethical approval from their local ethical review boards. The Royal Brisbane and Women’s Hospital Human Research Ethics Committee waived the requirement for a full ethical review for this meta-analysis.
Table 1 defines the common clinical endpoints available for testing in this analysis. Hospital or intensive care unit (ICU) mortality was defined as the proportion of patients who died before discharge from the hospital or from the ICU, respectively. Hospital mortality was pragmatically censored at 30 days. “Clinical cure” was defined as the disappearance of all infection-related symptoms and signs, as assessed by a blinded clinician 7–14 days after the cessation of study antibiotics (20).
| Endpoint | Definition and Description |
|---|---|
| Hospital mortality censored at Day 30 | Proportion of patients who died before hospital discharge censored at 30 d postrandomization. |
| ICU mortality | Proportion of patients who died before discharge from the ICU. |
| Clinical cure | Clinical cure was evaluated 7–14 d after cessation of study antibiotic and was defined as disappearance of all signs and symptoms, including SIRS criteria*, related to infection. Clinical cure was evaluated by a blinded clinician if the participant was still in the ICU or by blinded review of the patient record if the participant was discharged from the ICU. |
| ICU-free days at Day 28 | The number of days the participant was ICU free after successful transfer to a general ward in the first 28 d postrandomization. ICU-free days were assigned as 0 if a patient died or stayed in the ICU for ≥28 d. |
We defined “continuous infusion” as constant intravenous administration throughout a 24-hour period and “intermittent dosing” as administration of an intravenous infusion for less than or equal to 30 minutes. We defined nonfermenting gram-negative bacilli to be the following organisms: Acinetobacter baumannii, Bulkholderia cepacia, Bulkholderia pseudomallei, Pseudomonas aeruginosa, and Stenotrophomonas maltophilia. “Multidrug-resistant” gram-negative organisms were defined as those nonsusceptible to at least one agent in three or more antimicrobial categories (21). “Nonsusceptible organisms” were defined in relationship to the randomized β-lactam antibiotic. Illness severity was estimated using Acute Physiology and Chronic Health Evaluation (APACHE) II scores (ranging from 0 to 71, with higher scores indicating an increased risk of death) (22), and they were calculated using the most abnormal data for the first 24 hours after ICU admission.
Researchers with access to the primary data (J.A.R., M.-H.A.-A., and J.M.D.) reviewed individual study variables and extracted relevant common variables into a single dataset. Data on hospital mortality and 90-day mortality were recoded into a single variable (“hospital mortality censored at 30 days”). There were no duplicate participants in the included studies, and missing data and logic testing were performed on the combined dataset.
We used intention-to-treat principles to conduct the efficacy analysis. Basic characteristics of study participants were presented using number (percentage) and median (interquartile range [IQR]), as appropriate. Heterogeneity between studies was assessed using a χ2 test and was considered significant if the P value was less than 0.10 or the I2 statistic was greater than 50%. Depending on the observed heterogeneity, pooled estimates of effect and the 95% confidence interval (CI) were reported for the combined population using a fixed or random effects model; the random effects model was selected when the I2 statistic was greater than or equal to 50%. Relative risk (RR) was used for dichotomous outcomes (mortality and clinical cure), and median difference was used for continuous outcomes (ICU-free days at Day 28). We constructed a Kaplan–Meier survival curve to compare survival trends censored at hospital discharge or Day 30, whichever was sooner, and comparison between the two treatment groups was performed using a log-rank test with the hazard ratio and 95% CI reported.
We constructed a multivariable logistic regression model and a Cox proportional hazards model to identify significant predictors associated with hospital survival censored at Day 30 and clinical cure, respectively, with the odds ratio and 95% CI reported. In both models, baseline variables considered to be biologically plausible and with a P value less than or equal to 0.15 in univariate analysis were entered into the model. In all models, a study variable was entered as a fixed covariate and interactions between covariates and treatment effect were explored. Goodness of fit for the multivariable logistic regression model was assessed using the Hosmer–Lemeshow test. We also explored the association of three a priori variables—culture-positive infection, infection by a gram-negative pathogen, and requirement for acute renal replacement therapy during ICU admission—with the outcomes of interest. A Classification and Regression Tree analysis was used to delineate patient subgroups associated with greatest separation in the outcome of interest for each treatment group (23).
Statistical analysis was performed using IBM SPSS Statistics version 22 software (IBM Corp., Armonk, NY). A two-sided P value less than 0.05 was considered statistically significant.
A Preferred Reporting Items for Systematic reviews and Meta-Analyses flowchart for identification of included studies is presented in Figure 1. We identified three RCTs of critically ill patients with severe sepsis. These RCTs recruited a total of 632 patients, for all of whom the trial authors provided the requested data (20, 24, 25). Of these, 312 patients were assigned to receive β-lactam antibiotic therapy by continuous infusion and 320 patients were assigned to receive β-lactam antibiotic intermittent dosing. The three studies contributed 432 (20), 140 (24), and 60 (25) patients from 26, 2, and 5 centers, respectively, and were conducted across Australia, New Zealand, Hong Kong, and Malaysia. All studies met the four quality assessment criteria (see Table E1 in the online supplement). There were no missing data in any of the trials for hospital mortality, clinical cure, ICU-free days at Day 28, or ICU mortality. We could find no significant evidence of heterogeneity between the included studies (see Figure E1).
Figure 1. Preferred Reporting Items for Systematic reviews and Meta-Analyses flowchart for identification of included studies.
[More] [Minimize]The baseline characteristics of the continuous and intermittent dosing groups for the population as a whole and by study are reported in Table 2 and Table E2, respectively. There were no significant baseline differences between the treatment groups. A pathogen was isolated from blood in 31 participants (9.7%) in the intermittent dosing group and 28 participants (9.0%) in the continuous dosing group (Table 2). The median durations of study drug administration were 4 days (IQR, 2–7) in the intermittent dosing group and 5 days (IQR, 2–7) in the continuous dosing group (P = 0.57). Randomization occurred within 1 day (IQR, 0–4) of ICU admission in the continuous dosing group and also within 1 day (IQR, 1–4) in the intermittent dosing group. The three most commonly identified causative organisms were Escherichia coli, A. baumannii, and Staphylococcus aureus (Table 3).
| Characteristic | Continuous Infusion (n = 312) | Intermittent Dosing (n = 320) |
|---|---|---|
| Age, yr | 61 (49–70) | 63 (49–72) |
| Male sex | 198 (63.5) | 204 (63.8) |
| APACHE II score | 21 (16–26) | 20 (16–25) |
| Organism identified | 97 (31.1) | 114 (35.6) |
| Study antibiotic | ||
| Piperacillin-tazobactam | 203 (65.1) | 221 (69.1) |
| Meropenem | 94 (30.1) | 93 (29.1) |
| Cefepime | 11 (3.5) | 2 (0.6) |
| Ticarcillin-clavulanate | 4 (1.3) | 4 (1.2) |
| Antibiotic 24-h dose, g | ||
| Piperacillin-tazobactam | 13.5 (13.5–18.0) | 13.5 (13.5–18.0) |
| Meropenem | 3.0 (2.0–3.0) | 3.0 (1.7–3.0) |
| Cefepime | 6.0 (6.0–6.0) | 6.0 |
| Ticarcillin-clavulanate | 12.4 (12.4–13.2) | 12.4 |
| Duration from ICU admission to randomization, d | 1 (0–4) | 1 (1–4) |
| Duration of randomized treatment, d | 5 (2–7) | 4 (2–7) |
| Postrandomization length of ICU stay, d | 7 (4–12) | 6 (3–12) |
| Organ dysfunction | ||
| Cardiovascular | 214 (68.6) | 217 (67.8) |
| Respiratory | 207 (66.3) | 208 (65.0) |
| Renal | 74 (23.7) | 82 (25.6) |
| Metabolic acidosis | 71 (25.2) | 73 (25.2) |
| Hematological | 45 (14.4) | 32 (10.0) |
| Primary infection site | ||
| Lung | 175 (56.1) | 172 (53.8) |
| Intraabdominal | 70 (22.4) | 79 (24.7) |
| Blood | 28 (9.0) | 31 (9.7) |
| Skin or skin structure | 22 (7.1) | 28 (8.8) |
| Urinary tract | 21 (6.7) | 23 (7.2) |
| Central nervous system | 4 (1.3) | 7 (2.2) |
| Ear, nose, and throat | 4 (1.3) | 2 (0.6) |
| Indwelling vascular catheter | 4 (1.3) | 1 (0.3) |
| Pleural | 2 (0.6) | 3 (0.9) |
| Bone and joint | 4 (1.3) | 0 (0.0) |
| Cardiac | 3 (1.0) | 1 (0.3) |
| Gynecological | 1 (0.3) | 0 (0.0) |
| Others | 10 (3.2) | 4 (1.3) |
| Characteristic | Continuous Infusion (n = 312) | Intermittent Dosing (n = 320) |
|---|---|---|
| Organism identified | 96 (30.8) | 112 (35.0) |
| Gram-positive | 25 (26.0) | 32 (28.6) |
| Methicillin-susceptible Staphylococcus aureus | 9 (36.0) | 5 (15.6) |
| Methicillin-resistant Staphylococcus aureus | 2 (8.0) | 9 (28.1) |
| Streptococcus pneumoniae | 4 (16.0) | 3 (9.4) |
| Streptococcus milleri group | 2 (8.0) | 5 (15.6) |
| Enterococcus faecalis | 3 (12.0) | 2 (6.3) |
| Staphylococcus epidermidis | 2 (8.0) | 2 (6.3) |
| Streptococcus pyogenes | 1 (4.0) | 3 (9.4) |
| Enterococcus faecium | 0 (0.0) | 1 (3.1) |
| Granulicatella adiacens | 0 (0.0) | 1 (3.1) |
| Mycoplasma pneumoniae | 0 (0.0) | 1 (3.1) |
| Streptococcus mitis | 1 (4.0) | 0 (0.0) |
| Streptococcus dysgalactiae | 1 (4.0) | 0 (0.0) |
| Gram-negative | 71 (74.0) | 80 (71.4) |
| Escherichia coli | 19 (26.8)* | 15 (18.8) |
| Acinetobacter baumannii | 16 (22.5)† | 16 (20.0)‡ |
| Klebsiella pneumoniae | 14 (19.7)§ | 14 (17.5) |
| Pseudomonas aeruginosa | 14 (19.7) | 12 (15.0) |
| Enterobacter cloacae | 1 (1.4) | 5 (6.3)‖ |
| Serratia marcescens | 1 (1.4) | 5 (6.3)¶ |
| Haemophilus influenzae | 1 (1.4) | 1 (1.3) |
| Klebsiella oxytoca | 0 (0.0) | 2 (2.5) |
| Morganella morganii | 0 (0.0) | 2 (2.5)¶ |
| Proteus mirabilis | 2 (2.8) | 0 (0.0) |
| Stenotrophomonas maltophilia | 0 (0.0) | 2 (2.5) |
| Bulkholderia cepacia | 0 (0.0) | 1 (1.3) |
| Bulkholderia pseudomallei | 1 (1.4) | 0 (0.0) |
| Chlamydophila pneumoniae | 0 (0.0) | 1 (1.3) |
| Citrobacter koseri | 1 (1.4) | 0 (0.0) |
| Coxiella burnetii | 0 (0.0) | 1 (1.3) |
| Enterobacter aerogenes | 0 (0.0) | 1 (1.3) |
| Raoultella planticola | 0 (0.0) | 1 (1.3) |
| Salmonella typhimurium | 1 (1.4) | 0 (0.0) |
| Vibrio vulnificus | 0 (0.0) | 1 (1.3) |
| Polymicrobial infection, n (%) | 25 (25.8) | 29 (25.4) |
In the combined study population, hospital mortality was significantly lower (RR, 0.74; 95% CI, 0.56–1.00; P = 0.045) and clinical cure was significantly higher (RR, 1.20; 95% CI, 1.03–1.40; P = 0.021) in the continuous dosing group than in the intermittent dosing group (Figure 2). The numbers needed to treat with continuous infusion to improve patient survival and clinical cure were 15 and 11 patients, respectively. The Kaplan–Meier 30-day survival curve for the combined study population is shown in Figure 3. There was no difference between the groups in terms of ICU-free days at Day 28 (RR, 0; 95% CI, −3 to 3; P = 0.90) or ICU mortality (RR, 0.82; 95% CI, 0.58–1.16; P = 0.26).
Figure 2. Differences in mortality and clinical cure, along with 95% confidence intervals (95% CIs), for continuous infusion (CI) versus intermittent infusion (II). (A) Hospital mortality censored at Day 30. (B) Intensive care unit mortality. (C) Intensive care unit–free days at Day 28. (D) Clinical cure. df = degrees of freedom; IB = intermittent bolus; IV = inverse variance; M-H = Mantel–Haenszel test.
[More] [Minimize]
Figure 3. Kaplan–Meier 30-day survival curves for combined study population. CI = confidence interval; HR = hazard ratio.
[More] [Minimize]The factors predicting hospital mortality and clinical cure are summarized in Table 4. Higher APACHE II score, receipt of acute renal replacement therapy, respiratory and cardiovascular dysfunction on admission, β-lactam antibiotic administration by intermittent dosing, infection by nonfermenting gram-negative bacilli, and older age were all statistically significant predictors of hospital mortality. The absence of acute renal replacement therapy during ICU admission, lower APACHE II score, younger age, and nonbloodstream infection were all statistically significant for clinical cure. β-Lactam antibiotic administration by continuous infusion was not independently associated with clinical cure.
| Variable | All Factors Included in the Model | Final Model | ||
|---|---|---|---|---|
| OR (95% CI) | P Value | OR (95% CI) | P Value | |
| Factors predicting hospital mortality censored at 30 d | ||||
| APACHE II score, per 1-point increase | 1.08 (1.05–1.11) | <0.001 | 1.08 (1.05–1.11) | <0.001 |
| Renal replacement therapy | 2.48 (1.40–4.39) | 0.002 | 3.02 (1.76–5.18) | <0.001 |
| Study | 0.01 | 0.003 | ||
| Dulhunty and colleagues (25) | 0.70 (0.23–2.15) | 0.54 | 0.66 (0.26–1.67) | 0.38 |
| Abdul-Aziz and colleagues (24) | 2.89 (1.33–6.24) | 0.01 | 2.55 (1.35–4.79) | 0.01 |
| Dulhunty and colleagues (20) (reference group) | 1.00 | 1.00 | ||
| Respiratory dysfunction on admission | 1.86 (1.15–3.00) | 0.01 | 1.71 (1.07–2.75) | 0.03 |
| Infection by NFGNB* | 2.85 (1.35–6.04) | 0.01 | 2.72 (1.32–5.62) | 0.01 |
| Age, per 1-yr increase | 1.02 (1.00–1.03) | 0.02 | 1.02 (1.00–1.03) | 0.03 |
| Cardiovascular dysfunction on admission | 1.68 (1.03–2.75) | 0.03 | 1.72 (1.06–2.80) | 0.02 |
| Hematological dysfunction on admission | 1.78 (0.97–3.25) | 0.06 | — | — |
| Renal dysfunction on admission | 1.41 (0.87–2.28) | 0.17 | — | — |
| Continuous infusion | 0.52 (0.08–3.21) | 0.48 | 0.62 (0.41–0.94) | 0.03 |
| Continuous infusion × study interaction | 0.69 | — | — | |
| Dulhunty and colleagues (25) compared with Dulhunty and colleagues (20) × continuous infusion | 0.77 (0.12–5.10) | 0.79 | — | — |
| Abdul-Aziz and colleagues (24) compared with Dulhunty and colleagues (20) × continuous infusion | 0.67 (0.26–1.71) | 0.89 | — | — |
| Causative organism identified | 0.76 (0.18–3.17) | 0.70 | — | — |
| Goodness of fit | ||||
| Hosmer–Lemeshow test | χ2 = 3.26, df = 8 | 0.917 | χ2 = 3.21, df = 8 | 0.921 |
| Factors predicting clinical cure | ||||
| Renal replacement therapy | 0.41 (0.25–0.69) | 0.001 | 0.38 (0.24–0.62) | <0.001 |
| APACHE II score, per 1-point increase | 0.97 (0.94–0.99) | 0.01 | 0.97 (0.94–0.99) | 0.003 |
| Age, per 1-yr increase | 0.99 (0.98–0.99) | 0.03 | 0.98 (0.97–0.99) | 0.03 |
| Bacteremia | 0.54 (0.29–1.00) | 0.05 | 0.51 (0.28–0.92) | 0.03 |
| Study | 0.05 | 0.01 | ||
| Dulhunty and colleagues (25) | 1.02 (0.45–2.30) | 0.97 | 0.87 (0.40–1.89) | 0.73 |
| Abdul-Aziz and colleagues (24) | 0.45 (0.23–0.87) | 0.02 | 0.36 (0.20–0.62) | 0.001 |
| Dulhunty and colleagues (20) (reference group) | 1.00 | 1.00 | ||
| Continuous infusion | 1.17 (0.78–1.74) | 0.45 | — | — |
| Continuous infusion × study interaction | 0.08 | 0.01 | ||
| Dulhunty and colleagues (25) compared with Dulhunty and colleagues (20) × continuous infusion | 2.50 (0.75–8.37) | 0.14 | 2.45 (1.22–4.92) | 0.06 |
| Abdul-Aziz and colleagues (24) compared with Dulhunty and colleagues (20) × continuous infusion | 2.20 (0.97–4.97) | 0.06 | 3.02 (0.97–9.37) | 0.01 |
| Urinary tract infections | 1.82 (0.92–3.63) | 0.09 | — | — |
| Skin or skin structure infections | 0.64 (0.34–1.21) | 0.17 | — | — |
| Hematological dysfunction on admission | 0.77 (0.45–1.34) | 0.36 | — | — |
| Causative organism identified | 1.85 (0.47–7.38) | 0.38 | — | — |
| Infection by NFGNB* | 0.77 (0.45–1.34) | 0.46 | — | — |
| Renal dysfunction on admission | 1.11 (0.72–1.71) | 0.63 | — | — |
| Goodness-of-fit | ||||
| Hosmer–Lemeshow test | χ2 = 8.01, df = 8 | 0.433 | χ2 = 10.2, df = 8 | 0.252 |
Using a Classification and Regression Tree analysis, APACHE II scores less than 22 were associated with improved hospital survival (P = 0.024) and clinical cure (P = 0.019). The patient outcome data for APACHE II and renal replacement therapy subgroups are presented in Table 5. In these groups, statistically significant advantages were observed for clinical cure only. The Cox regression 30-day hospital survival curve for the combined study population is shown in Figure 4.
| Continuous Infusion | Intermittent Dosing | RR (95% CI) | P Value | |
|---|---|---|---|---|
| Hospital mortality censored at 30 d | ||||
| APACHE II score ≥22* | 44 (29.7) | 57 (40.4) | 0.74 (0.53–1.01) | 0.06 |
| APACHE II score <22† | 17 (10.4) | 27 (15.1) | 0.69 (0.39–1.21) | 0.19 |
| RRT‡ | 21 (38.2) | 27 (45.8) | 0.83 (0.54–1.29) | 0.41 |
| Without RRT§ | 40 (15.6) | 57 (21.8) | 0.71 (0.49–1.03) | 0.07 |
| Culture-positive infection‖ | 32 (33.3) | 30 (26.8) | 1.24 (0.82–1.88) | 0.30 |
| Gram-positive organism¶ | 7 (28.0) | 6 (18.8) | 1.49 (0.57–3.89) | 0.41 |
| Gram-negative organism** | 25 (35.2) | 24 (30.0) | 1.17 (0.74–1.86) | 0.50 |
| Culture-negative infection†† | 29 (13.4) | 54 (26.0) | 0.52 (0.34–0.78) | 0.001 |
| Susceptible microorganism‡‡ | 34 (29.8) | 32 (28.1) | 1.06 (0.71–1.60) | 0.77 |
| Piperacillin-tazobactam§§ | 33 (16.3) | 57 (25.8) | 0.63 (0.43–0.93) | 0.02 |
| Meropenem‖‖ | 20 (21.3) | 27 (29.0) | 0.73 (0.44–1.21) | 0.22 |
| Lung infection¶¶ | 40 (22.9) | 43 (25.0) | 0.91 (0.63–1.33) | 0.64 |
| Intraabdominal infection*** | 19 (27.1) | 21 (26.6) | 1.02 (0.60–1.74) | 0.94 |
| ICU mortality | ||||
| APACHE II score ≥22* | 34 (23.0) | 41 (29.1) | 0.79 (0.53–1.17) | 0.24 |
| APACHE II score <22† | 13 (7.9) | 18 (10.1) | 0.79 (0.40–1.56) | 0.49 |
| RRT‡ | 18 (32.7) | 23 (39.0) | 0.84 (0.51–1.38) | 0.49 |
| Without RRT§ | 29 (11.3) | 36 (13.8) | 0.82 (0.52–1.29) | 0.39 |
| Culture-positive infection‖ | 22 (22.9) | 18 (16.1) | 1.43 (0.81–2.50) | 0.21 |
| Gram-positive organism¶ | 5 (20.0) | 5 (15.6) | 1.28 (0.42–3.94) | 0.67 |
| Gram-negative organism** | 17 (23.9) | 13 (16.3) | 1.47 (0.77–2.82) | 0.24 |
| Culture-negative infection†† | 25 (11.6) | 41 (19.7) | 0.59 (0.37–0.93) | 0.02 |
| Susceptible microorganism‡‡ | 23 (20.2) | 22 (19.3) | 1.05 (0.62–1.77) | 0.87 |
| Piperacillin-tazobactam§§ | 28 (13.8) | 39 (17.6) | 0.78 (0.50–1.22) | 0.28 |
| Meropenem‖‖ | 13 (13.8) | 20 (21.5) | 0.64 (0.34–1.22) | 0.17 |
| Lung infection¶¶ | 32 (18.3) | 30 (17.4) | 1.04 (0.67–1.65) | 0.84 |
| Intraabdominal infection*** | 13 (18.6) | 13 (16.5) | 1.13 (0.56–2.27) | 0.73 |
| Clinical cure | ||||
| APACHE II score ≥22* | 69 (46.6) | 47 (33.3) | 1.40 (1.05–1.87) | 0.02 |
| APACHE II score <22† | 104 (63.4) | 101 (56.4) | 1.12 (0.94–1.34) | 0.19 |
| RRT‡ | 15 (27.3) | 19 (32.2) | 0.85 (0.48–1.50) | 0.57 |
| Without RRT§ | 158 (61.5) | 129 (49.4) | 1.24 (1.06–1.45) | 0.01 |
| Culture-positive infection‖ | 44 (45.8) | 49 (43.8) | 1.05 (0.77–1.42) | 0.76 |
| Gram-positive organism¶ | 10 (40.0) | 12 (37.5) | 1.07 (0.55–2.06) | 0.85 |
| Gram-negative organism** | 34 (47.9) | 37 (46.3) | 1.04 (0.74–1.45) | 0.84 |
| Culture-negative infection†† | 129 (59.7) | 99 (47.6) | 1.25 (1.05–1.50) | 0.01 |
| Susceptible microorganism‡‡ | 60 (52.6) | 49 (43.0) | 1.22 (0.93–1.61) | 0.15 |
| Piperacillin-tazobactam§§ | 113 (55.7) | 93 (42.1) | 1.32 (1.09–1.61) | 0.01 |
| Meropenem‖‖ | 53 (56.4) | 50 (53.8) | 1.05 (0.81–1.36) | 0.72 |
| Lung infection¶¶ | 91 (52.0) | 81 (47.1) | 1.10 (0.89–1.37) | 0.36 |
| Intraabdominal infection*** | 35 (50.0) | 39 (49.4) | 1.01 (0.73–1.40) | 0.94 |
| ICU-free days censored at Day 28 | ||||
| APACHE II score ≥22* | 15 (0–22) | 12 (0–22) | 3 (−4 to 10) | 0.47 |
| APACHE II score <22† | 21 (15–24) | 22 (11–25) | −1 (−3 to 1) | 0.78 |
| RRT‡ | 2 (0–18) | 3 (0–20) | −1 (−9.76 to 7.76) | 0.98 |
| Without RRT§ | 20 (12–24) | 21 (6–24) | −1 (−2.77 to 0.77) | 0.92 |
| Culture-positive infection‖ | 18 (0–22) | 19 (4–24) | −1 (−6 to 3) | 0.22 |
| Gram-positive organism¶ | 17 (1–21) | 19 (12–24) | −2 (−13 to 10) | 0.10 |
| Gram-negative organism** | 18 (0–23) | 19 (1–23) | −1 (−6 to 4) | 0.69 |
| Culture-negative infection†† | 20 (6–24) | 19 (0–24) | 1 (−2 to 4) | 0.36 |
| Susceptible microorganism‡‡ | 19 (2–23) | 19 (0–23) | 0 (−3 to 4) | 0.94 |
| Piperacillin-tazobactam§§ | 19 (6–24) | 20 (3–24) | −1 (−4 to 2) | 0.96 |
| Meropenem‖‖ | 18 (4–23) | 17 (0–23) | 1 (−6 to 7) | 0.25 |
| Lung infection¶¶ | 18 (2–22) | 17 (0–23) | 1 (−2 to 4) | 0.69 |
| Intraabdominal infection*** | 17 (0–24) | 21 (4–24) | −4 (−11 to 2) | 0.46 |
Figure 4. Cox regression 30-day survival curves for combined study population. APACHE = Acute Physiology and Chronic Health Evaluation; CI = confidence interval; HR = hazard ratio; NFGNB = nonfermenting gram-negative bacilli (Acinetobacter baumannii, Bulkholderia cepacia, Bulkholderia pseudomallei, Pseudomonas aeruginosa, and Stenotrophomonas maltophilia); RRT = renal replacement therapy.
[More] [Minimize]We also examined the effect of infection caused by a gram-negative pathogen and the presence of acute renal replacement therapy on outcomes. We found no differences between the continuous infusion and intermittent dosing groups for hospital mortality (RR, 1.17; 95% CI, 0.74–1.86; P = 0.50), clinical cure (RR, 1.04; 95% CI, 0.74–1.45; P = 0.84), or ICU mortality (RR, 1.47; 95% CI, 0.77–2.82; P = 0.24) in patients in whom the infection was caused by a gram-negative pathogen.
In this individual patient data meta-analysis including data from 632 randomized patients, we found that, compared with intermittent infusion, continuous infusion of β-lactam antibiotics was associated with lower hospital mortality censored at 30 days. Clinical cure was higher in the continuous dosing group, although this was not statistically significant after adjusting for between-study heterogeneity and in multivariable analysis. We also identified patient subgroups in whom further prospective RCTs should be considered, namely patients with higher APACHE II scores and those not requiring renal replacement therapy (26, 27).
We focused on patients with severe sepsis rather than sepsis overall, as these patients are more likely to have pathophysiological changes leading to subtherapeutic drug concentrations (5, 13, 28). Researchers in pharmacokinetic studies of critically ill patients with sepsis have concluded that administration by continuous infusion increases the achievement of target concentrations, both in plasma and in tissues, compared with intermittent dosing (5, 14, 15, 25, 27, 29).
Several RCTs comparing continuous infusion and intermittent dosing of β-lactam antibiotics have been conducted in critically ill patients, although none have had adequate statistical power to detect a difference in mortality (20, 24, 25, 30–35). In contrast to the present study, authors of previous aggregated data metaanalyses of this question found no significant difference in mortality between the continuous and intermittent dosing groups (16, 36–40). These metaanalyses were less selective than the present analysis in their inclusion criteria, included data from both critically ill and non–critically ill patients, and allowed differing antibiotic doses in the two treatment groups, all of which may have diluted any advantage of continuous infusion.
The results of our meta-analysis imply that administration of β-lactam antibiotics by continuous infusion compared with intermittent infusion in critically ill patients with severe sepsis may decrease hospital mortality, although the relationship with clinical cure was more complex and may have been influenced by the subjective limitations of this endpoint. Moreover, our findings imply that patients receiving renal replacement therapy may not derive a significant benefit from continuous infusion. In addition, they imply that the beneficial impact of continuous infusion is likely greatest in infections with nonfermenting gram-negative bacilli (e.g., A. baumannii and P. aeruginosa).
These implications are biologically plausible because patients receiving renal replacement therapy are likely to have reduced drug clearance and hence higher serum antibiotic concentrations, regardless of which antibiotic administration method is used (41). Similarly, nonfermenting gram-negative bacilli tend to have higher MICs than β-lactam antibiotics (42), supporting data from pharmacokinetic and dosing simulation studies indicating that continuous infusion of β-lactam antibiotics is more likely than intermittent dosing to achieve therapeutic targets for less-susceptible pathogens (14, 15).
Owing to the stringency of our inclusion criteria, only three studies were included in this analysis. However, these inclusion criteria were all selected a priori, had mechanistic precedents, and were applied equally to all studies identified in our systematic literature search. The fact that the included patients were enrolled across only four countries means that the results may not be generalizable to all treatment settings. We did observe that study was independently associated with mortality and clinical cure, suggesting some degree of between-study heterogeneity. This was particularly true for one study (24) when compared with the other two (20, 25), highlighting potential differences in baseline factors by geographical region. However, despite this, a statistically significant interaction effect between study and treatment was observed only for clinical cure and not for mortality. This leads us to conclude that the treatment effect observed was consistent across studies for mortality, while study variability in a (nonsignificant) treatment effect on clinical cure may have been influenced by subjective differences between studies in how this endpoint was assessed. The relatively low proportion of patients with identified pathogens and the lack of MIC data across each of the studies prevented us from testing the importance of pathogen MIC on patient outcomes. Although such an analysis would provide useful mechanistic data, it would be of secondary importance to the patient-centered outcomes reported in the present study.
In this individual patient data meta-analysis of a large critical care patient population with severe sepsis, we found that administration of β-lactam antibiotics by continuous infusion is associated with decreased hospital mortality and a higher rate of clinical cure compared with intermittent dosing. Given these findings, we recommend that a definitive RCT be conducted in patients with a higher level of sickness severity who are not receiving renal replacement therapy and are at risk of infection by less-susceptible pathogens such as nonfermenting gram-negative bacilli.
| 1. | Dombrovskiy VY, Martin AA, Sunderram J, Paz HL. Rapid increase in hospitalization and mortality rates for severe sepsis in the United States: a trend analysis from 1993 to 2003. Crit Care Med 2007;35:1244–1250. |
| 2. | Kaukonen KM, Bailey M, Suzuki S, Pilcher D, Bellomo R. Mortality related to severe sepsis and septic shock among critically ill patients in Australia and New Zealand, 2000-2012. JAMA 2014;311:1308–1316. |
| 3. | Ranieri VM, Thompson BT, Barie PS, Dhainaut JF, Douglas IS, Finfer S, Gårdlund B, Marshall JC, Rhodes A, Artigas A, et al.; PROWESS-SHOCK Study Group. Drotrecogin alfa (activated) in adults with septic shock. N Engl J Med 2012;366:2055–2064. |
| 4. | Kumar A, Roberts D, Wood KE, Light B, Parrillo JE, Sharma S, Suppes R, Feinstein D, Zanotti S, Taiberg L, et al. Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock. Crit Care Med 2006;34:1589–1596. |
| 5. | Roberts JA, Paul SK, Akova M, Bassetti M, De Waele JJ, Dimopoulos G, Kaukonen KM, Koulenti D, Martin C, Montravers P, et al.; DALI Study. DALI: Defining antibiotic levels in intensive care unit patients: are current β-lactam antibiotic doses sufficient for critically ill patients? Clin Infect Dis 2014;58:1072–1083. |
| 6. | McKinnon PS, Paladino JA, Schentag JJ. Evaluation of area under the inhibitory curve (AUIC) and time above the minimum inhibitory concentration (T>MIC) as predictors of outcome for cefepime and ceftazidime in serious bacterial infections. Int J Antimicrob Agents 2008;31:345–351. |
| 7. | Li C, Du X, Kuti JL, Nicolau DP. Clinical pharmacodynamics of meropenem in patients with lower respiratory tract infections. Antimicrob Agents Chemother 2007;51:1725–1730. |
| 8. | Rayner CR, Forrest A, Meagher AK, Birmingham MC, Schentag JJ. Clinical pharmacodynamics of linezolid in seriously ill patients treated in a compassionate use programme. Clin Pharmacokinet 2003;42:1411–1423. |
| 9. | Dulhunty JM, Paterson D, Webb SA, Lipman J. Antimicrobial utilisation in 37 Australian and New Zealand intensive care units. Anaesth Intensive Care 2011;39:231–237. |
| 10. | Corona A, Bertolini G, Lipman J, Wilson AP, Singer M. Antibiotic use and impact on outcome from bacteraemic critical illness: the BActeraemia Study in Intensive Care (BASIC). J Antimicrob Chemother 2010;65:1276–1285. |
| 11. | Drusano GL. Antimicrobial pharmacodynamics: critical interactions of ‘bug and drug’. Nat Rev Microbiol 2004;2:289–300. |
| 12. | Craig WA. Pharmacokinetic/pharmacodynamic parameters: rationale for antibacterial dosing of mice and men. Clin Infect Dis 1998;26:1–10, quiz 11–12. |
| 13. | Roberts JA, Abdul-Aziz MH, Lipman J, Mouton JW, Vinks AA, Felton TW, Hope WW, Farkas A, Neely MN, Schentag JJ, et al.; International Society of Anti-Infective Pharmacology and the Pharmacokinetics and Pharmacodynamics Study Group of the European Society of Clinical Microbiology and Infectious Diseases. Individualised antibiotic dosing for patients who are critically ill: challenges and potential solutions. Lancet Infect Dis 2014;14:498–509. |
| 14. | Roberts JA, Kirkpatrick CM, Roberts MS, Robertson TA, Dalley AJ, Lipman J. Meropenem dosing in critically ill patients with sepsis and without renal dysfunction: intermittent bolus versus continuous administration? Monte Carlo dosing simulations and subcutaneous tissue distribution. J Antimicrob Chemother 2009;64:142–150. |
| 15. | Roberts JA, Roberts MS, Robertson TA, Dalley AJ, Lipman J. Piperacillin penetration into tissue of critically ill patients with sepsis—bolus versus continuous administration? Crit Care Med 2009;37:926–933. |
| 16. | Roberts JA, Webb S, Paterson D, Ho KM, Lipman J. A systematic review on clinical benefits of continuous administration of β-lactam antibiotics. Crit Care Med 2009;37:2071–2078. |
| 17. | Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009;6:e1000097. |
| 18. | Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, Cook D, Cohen J, Opal SM, Vincent JL, Ramsay G; International Sepsis Definitions Conference. 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Intensive Care Med 2003;29:530–538. |
| 19. | Higgins JPT, Green S, editors. Cochrane handbook for systematic reviews of interventions. Chichester, UK: Wiley-Blackwell; 2011. |
| 20. | Dulhunty JM, Roberts JA, Davis JS, Webb SA, Bellomo R, Gomersall C, Shirwadkar C, Eastwood GM, Myburgh J, Paterson DL, et al.; BLING II Investigators for the ANZICS Clinical Trials Group. A multicenter randomized trial of continuous versus intermittent β-lactam infusion in severe sepsis. Am J Respir Crit Care Med 2015;192:1298–1305. |
| 21. | Magiorakos AP, Srinivasan A, Carey RB, Carmeli Y, Falagas ME, Giske CG, Harbarth S, Hindler JF, Kahlmeter G, Olsson-Liljequist B, et al. Multidrug-resistant, extensively drug-resistant and pandrug-resistant bacteria: an international expert proposal for interim standard definitions for acquired resistance. Clin Microbiol Infect 2012;18:268–281. |
| 22. | Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med 1985;13:818–829. |
| 23. | Breiman L, Friedman J, Olshen RA, Stone CJ. Classification and regression trees. Boca Raton, FL: Chapman & Hall/CRC Press; 1984. |
| 24. | Abdul-Aziz MH, Sulaiman H, Mat-Nor MB, Rai V, Wong KK, Hasan MS, Abd Rahman AN, Jamal JA, Wallis SC, Lipman J, et al. β-Lactam Infusion in Severe Sepsis (BLISS): a prospective, two-centre, open-labelled randomised controlled trial of continuous versus intermittent β-lactam infusion in critically ill patients with severe sepsis. Intensive Care Med[online ahead of print] 11 Jan 2016; DOI: 10.1007/s00134-015-4188-0. |
| 25. | Dulhunty JM, Roberts JA, Davis JS, Webb SA, Bellomo R, Gomersall C, Shirwadkar C, Eastwood GM, Myburgh J, Paterson DL, et al. Continuous infusion of β-lactam antibiotics in severe sepsis: a multicenter double-blind, randomized controlled trial. Clin Infect Dis 2013;56:236–244. |
| 26. | Abdul-Aziz MH, Dulhunty JM, Bellomo R, Lipman J, Roberts JA. Continuous β-lactam infusion in critically ill patients: the clinical evidence. Ann Intensive Care 2012;2:37. |
| 27. | Abdul-Aziz MH, Lipman J, Akova M, Bassetti M, De Waele JJ, Dimopoulos G, Dulhunty J, Kaukonen KM, Koulenti D, Martin C, et al.; DALI Study Group. Is prolonged infusion of piperacillin/tazobactam and meropenem in critically ill patients associated with improved pharmacokinetic/pharmacodynamic and patient outcomes? An observation from the Defining Antibiotic Levels in Intensive care unit patients (DALI) cohort. J Antimicrob Chemother 2016;71:196–207. |
| 28. | Roberts JA, Udy AA, Jarrett P, Wallis SC, Hope WW, Sharma R, Kirkpatrick CM, Kruger PS, Roberts MS, Lipman J. Plasma and target-site subcutaneous tissue population pharmacokinetics and dosing simulations of cefazolin in post-trauma critically ill patients. J Antimicrob Chemother 2015;70:1495–1502. |
| 29. | Mouton JW, Vinks ATMM. Is continuous infusion of β-lactam antibiotics worthwhile?—efficacy and pharmacokinetic considerations. J Antimicrob Chemother 1996;38:5–15. |
| 30. | Angus BJ, Smith MD, Suputtamongkol Y, Mattie H, Walsh AL, Wuthiekanun V, Chaowagul W, White NJ. Pharmacokinetic-pharmacodynamic evaluation of ceftazidime continuous infusion vs intermittent bolus injection in septicaemic melioidosis. Br J Clin Pharmacol 2000;50:184–191. |
| 31. | Cousson J, Floch T, Vernet-Garnier V, Appriou M, Petit JS, Jovenin N, Lamiable D, Hoizey G. Pharmacodynamic interest of ceftazidime continuous infusion vs intermittent bolus administration in patients with severe nosocomial pneumonia [in French]. Pathol Biol (Paris) 2005;53:546–550. |
| 32. | Georges B, Conil JM, Cougot P, Decun JF, Archambaud M, Seguin T, Chabanon G, Virenque C, Houin G, Saivin S. Cefepime in critically ill patients: continuous infusion vs. an intermittent dosing regimen. Int J Clin Pharmacol Ther 2005;43:360–369. |
| 33. | Nicolau DP, McNabb J, Lacy MK, Quintiliani R, Nightingale CH. Continuous versus intermittent administration of ceftazidime in intensive care unit patients with nosocomial pneumonia. Int J Antimicrob Agents 2001;17:497–504. |
| 34. | Rafati MR, Rouini MR, Mojtahedzadeh M, Najafi A, Tavakoli H, Gholami K, Fazeli MR. Clinical efficacy of continuous infusion of piperacillin compared with intermittent dosing in septic critically ill patients. Int J Antimicrob Agents 2006;28:122–127. |
| 35. | Roberts JA, Boots R, Rickard CM, Thomas P, Quinn J, Roberts DM, Richards B, Lipman J. Is continuous infusion ceftriaxone better than once-a-day dosing in intensive care? A randomized controlled pilot study. J Antimicrob Chemother 2007;59:285–291. |
| 36. | Kasiakou SK, Sermaides GJ, Michalopoulos A, Soteriades ES, Falagas ME. Continuous versus intermittent intravenous administration of antibiotics: a meta-analysis of randomised controlled trials. Lancet Infect Dis 2005;5:581–589. |
| 37. | Falagas ME, Tansarli GS, Ikawa K, Vardakas KZ. Clinical outcomes with extended or continuous versus short-term intravenous infusion of carbapenems and piperacillin/tazobactam: a systematic review and meta-analysis. Clin Infect Dis 2013;56:272–282. |
| 38. | Teo J, Liew Y, Lee W, Kwa AL. Prolonged infusion versus intermittent boluses of β-lactam antibiotics for treatment of acute infections: a meta-analysis. Int J Antimicrob Agents 2014;43:403–411. |
| 39. | Yusuf E, Spapen H, Piérard D. Prolonged vs intermittent infusion of piperacillin/tazobactam in critically ill patients: a narrative and systematic review. J Crit Care 2014;29:1089–1095. |
| 40. | Shiu J, Wang E, Tejani AM, Wasdell M. Continuous versus intermittent infusions of antibiotics for the treatment of severe acute infections. Cochrane Database Syst Rev 2013;3:CD008481. |
| 41. | Jamal JA, Mat-Nor MB, Mohamad-Nor FS, Udy AA, Wallis SC, Lipman J, Roberts JA. Pharmacokinetics of meropenem in critically ill patients receiving continuous venovenous haemofiltration: a randomised controlled trial of continuous infusion versus intermittent bolus administration. Int J Antimicrob Agents 2015;45:41–45. |
| 42. | Valenza G, Seifert H, Decker-Burgard S, Laeuffer J, Morrissey I, Mutters R; COMPACT Germany Study Group. Comparative Activity of Carbapenem Testing (COMPACT) study in Germany. Int J Antimicrob Agents 2012;39:255–258. |
J.A.R. and J.S.D. received salary funding from the National Health and Medical Research Council of Australia: Career Development Fellowship APP1048652 (J.A.R.) and Early Career Fellowship APP1013411 (J.S.D.). Funding was also received from the Australian National Health and Medical Research Council for Centre of Research Excellence (APP1099452).
Author Contributions: J.A.R., M.-H.A.-A., and J.M.D. had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis; J.A.R., J.S.D., J.M.D., J.M., R.B., M.O.C., and J.L. designed the study; M.-H.A.-A. and J.M.D. performed the statistical analysis; and J.A.R., J.S.D., J.M.D., and J.L. drafted the initial manuscript. All authors critically revised the manuscript for important intellectual content and read and approved the final version.
This article has an online supplement, which is accessible from this issue’s table of contents at www.atsjournals.org
Originally Published in Press as 10.1164/rccm.201601-0024OC on March 15, 2016
Author disclosures are available with the text of this article at www.atsjournals.org.
