Papers

β lactam monotherapy versus β lactam-aminoglycoside combination therapy for sepsis in immunocompetent patients: systematic review and meta-analysis of randomised trials

BMJ 2004; 328 doi: http://dx.doi.org/10.1136/bmj.38028.520995.63 (Published 18 March 2004) Cite this as: BMJ 2004;328:668

This article has a correction. Please see:

  1. Mical Paul, consultant (mica{at}zahav.net.il)1,
  2. Ishay Benuri-Silbiger, researcher2,
  3. Karla Soares-Weiser, coordinator of clinical research2,
  4. Leonard Leibovici, associate professor2
  1. 1Department of Medicine E and Infectious Diseases Unit, Rabin Medical Centre, Beilinson Campus, Petah-Tikva 49100, Israel
  2. 2Department of Medicine E, Rabin Medical Centre, Beilinson Campus, Petah-Tikva
  1. Correspondence to: M Paul
  • Accepted 22 December 2003

Abstract

Objective To compare β lactam monotherapy with β lactam-aminoglycoside combination therapy for severe infections.

Data sources Medline, Embase, Lilacs, Cochrane Library, and conference proceedings, to 2003; references of included studies; contact with all authors. No restrictions, such as language, year of publication, or publication status.

Study selection All randomised trials of β lactam monotherapy compared with β lactam-aminoglycoside combination therapy for patients without neutropenia who fulfilled criteria for sepsis.

Data selection Two reviewers independently applied selection criteria, performed quality assessment, and extracted the data. The primary outcome assessed was all cause fatality by intention to treat. Relative risks were pooled with the random effect model (relative risk < 1 favours monotherapy).

Results 64 trials with 7586 patients were included. There was no difference in all cause fatality (relative risk 0.90, 95% confidence interval 0.77 to 1.06). 12 studies compared the same β lactam (1.02, 0.76 to 1.38), and 31 studies compared different β lactams (0.85, 0.69 to 1.05). Clinical failure was more common with combination treatment overall (0.87, 0.78 to 0.97) and among studies comparing different β lactams (0.76, 0.68 to 0.86). There was no advantage to combination therapy among patients with Gram negative infections (1835 patients) or Pseudomonas aeruginosa infections (426 patients). There was no difference in the rate of development of resistance. Nephrotoxicity was significantly more common with combination therapy (0.36, 0.28 to 0.47). Heterogeneity was not significant for these comparisons.

Conclusions In the treatment of sepsis the addition of an aminoglycoside to β lactams should be discouraged. Fatality remains unchanged, while the risk for adverse events is increased.

Introduction

Treatment with a combination of β lactam and an aminoglycoside is purported to be superior to β lactam monotherapy for sepsis on the basis of potential advantages such as in vitro synergism and prevention of development of resistance.17 Textbooks and guidelines advise the combination for specific pathogens, such as Pseudomonas aeruginosa and other Gram negative bacteria, and for infections commonly caused by these pathogens.8 9 In aiming for optimal antibiotic treatment of severe infections, hospital clinicians tend to use combination therapy despite the lack of direct evidence for its effectiveness. Observational studies show that 25-30% of patients with bacteraemia,10 11 surgical infections,12 or pneumonia,13 14 50% of those with klebsiella bacteraemia,15 and 56% of patients with septic shock in the intensive care unit16 are given β lactam-aminoglycoside combination therapy.

We performed a systematic review and meta-analysis of randomised trials comparing β lactam-aminoglycoside combination therapy with β lactam monotherapy for severe infections in patients without neutropenia.

Methods

We searched Medline, Embase, Lilacs, the Cochrane Library (all up to March 2003), conference proceedings of the Interscience Conference on Antimicrobial Agents and Chemotherapy (1995-2002), and citations of included trials with the terms: (aminoglycoside* OR specific aminoglycosides) AND ((infect* OR sepsis OR bacter* OR septicemia OR specific infections/pathogens) OR combi*)). We included studies regardless of date, language, or publication status, and we contacted authors for complementary information.

We included all randomised and quasi-randomised trials that compared any β lactam monotherapy with any combination of a β lactam and an aminoglycoside for severe infections. Severe infection was defined as clinical evidence of infection, plus evidence of a systemic response to infection.17 We excluded studies with a dropout rate above 30%, unless intention to treat analysis was given for mortality or failure outcomes, and studies with more than 15% of patients with neutropenia, neonates, and preterm babies.

The primary outcome assessed was all cause fatality by the end of study follow up and up to 30 days. Secondary outcomes included treatment failure, defined as death, non-resolving primary infection, any modification to allocated antibiotics, or any therapeutic invasive intervention not defined by protocol; bacteriological failure, defined as persistence of primary pathogen; bacterial and fungal superinfections and colonisation; adverse events; and length of hospital stay. We separated studies that compared the same β lactam from studies that compared different β lactams. We performed subgroup analyses for P aeruginosa infections, any Gram negative infection, bacteraemia, and specific sources of infection.

Two reviewers independently applied inclusion and exclusion criteria and extracted the data. We extracted outcomes by intention to treat, unless the reasons for exclusions were not presented. In this case, we used the presented results (per protocol analysis) in the main analysis and compared them with results using all randomised patients and assuming failure for drop outs. Heterogeneity was assessed with a χ2 test and the I2 measure.18 We expected heterogeneity with respect to outcomes and used the random effects model, comparing it to a fixed effect model.19 We calculated relative risks with 95% confidence intervals and numbers needed to treat. Study quality measures extracted were allocation generation and concealment, blinding, intention to treat or per protocol analysis, designation of drop outs to treatment arms, number of drop outs, follow up and outcome predefinitions, and publication status.20 The effect of these measures was examined through sensitivity analysis.

We examined a funnel plot of the log of the relative risk against the weight to estimate potential selection bias (such as publication bias) and to assess whether effect estimates were associated with study size.

Results

We evaluated 144 eligible randomised trials and included 64 in the review (fig 1). The trials included 7586 patients, nearly all adults, and were performed between the years 1968-2001. The median number of patients per trial was 87 (range 20-580). Trials differed by the population targeted, type of infection, and antibiotics compared (table 1). The major conditions were severe sepsis, pneumonia, or Gram negative infections (41 trials), abdominal infections (11 trials), urinary tract infections (7 trials), and Gram positive infections (5 trials). Allocation to antibiotics was empirical in 56 trials. The same β lactam was compared in 20 trials, while all other trials compared one β lactam to a different, narrower spectrum β lactam combined with an aminoglycoside.

Fig 1
Fig 1

Detail of trial selection. The list of excluded references (w1-w80) can be found on bmj.com>

Table 1

Characteristics of included studies: patients and intervention

View this table:

All cause fatality—Forty three trials including 5527 patients reported all cause fatality. There was no significant difference between monotherapy and combination therapy when we combined these studies (relative risk 0.90, 95% confidence interval 0.77 to 1.06, fig 2). There was no difference among the 12 studies with 1381 patients that compared the same β lactam (1.02, 0.76 to 1.38) or among studies that compared different β lactams (0.85, 0.69 to 1.05). The heterogeneity for this comparison was low (I2 = 7.7%).

Fig 2
Fig 2

All cause fatality in comparison of β lactam monotherapy v β lactam-aminoglycoside combination therapy for treatment of sepsis. Log scale of relative risks (95% confidence intervals), random effect model. Studies ordered by weight

Treatment failure—We compared clinical and bacteriological failures in 63 and 43 trials, respectively (figs 3 and 4). For both comparisons, monotherapy was not significantly different from combination therapy among studies that compared the same β lactam. Monotherapy was significantly superior to combination therapy among studies that compared different β lactams. The overall comparison favoured monotherapy for clinical failure (0.87, 0.78 to 0.97; 6616 patients; number needed to treat 34, 20 to 147) and for bacteriological failure (0.86, 0.72 to 1.02; 3511 patients).

Fig 3
Fig 3

Clinical failure in comparison of β lactam monotherapy v β lactam-aminoglycoside combination therapy for treatment of sepsis. Log scale of relative risks (95% confidence intervals), random effect model. Studies ordered by weight

Fig 4
Fig 4

Bacteriological failure in comparison of β lactam monotherapy v β lactam-aminoglycoside combination therapy for treatment of sepsis. Log scale of relative risks (95% confidence intervals), random effect model. Studies ordered by weight

Subgroup analysis—Major effectiveness outcomes were compared within the defined patient subgroups expected to benefit most from combination therapy (tables 2 and 3). We did not detect an advantage to combination therapy with any subgroup tested. Mortality was higher among patients with P aeruginosa (21%), Gram negative infections (13%), and bacteraemia (15%), and outcomes were similar with combination versus monotherapy. Patients with infections outside the urinary tract (mainly pneumonia) had significantly fewer failures with monotherapy. Five trials specifically assessed Gram positive infections, endocarditis in four (table 1).21 32 48 63 69 Combined relative risks for fatality and failure favoured monotherapy, although differences were non-significant.

Table 2

All cause fatality in comparison of β lactam monotherapy v β lactam-aminoglycoside combination therapy for treatment of sepsis: subgroup analyses

View this table:
Table 3

Clinical failure in comparison of β lactam monotherapy v β lactam-aminoglycoside combination therapy for treatment of sepsis: subgroup analyses

View this table:

Development of resistance—Combination therapy did not lower bacterial superinfection or colonisation rates, which we would have expected if combination therapy prevented the development of resistance (fig 5). Relative risks tended in favour of monotherapy for bacterial superinfections (0.79, 0.59 to 1.06). Rates of fungal superinfection were similar. Six studies performed routine surveillance cultures, and nine assessed the development of resistance among pretreatment isolates. In these also we found no advantage with combination therapy. Twenty six studies reported coverage rates of the allocated treatment, although outcomes were not related to coverage. Among studies with different β lactams, the monotherapy β lactam provided broader coverage than the combination β lactam in 13 studies, the opposite occurring in two studies. Combined coverage of the β lactam and the aminoglycoside equalled monotherapy in these studies.

Fig 5
Fig 5

Summary relative risks for outcome relating to resistance development in comparison of β lactam monotherapy v β lactam-aminoglycoside combination therapy for treatment of sepsis. Log scale of relative risks (95% confidence intervals), random effect model. Studies ordered by weight

Drop outs and adverse events—The dropout rate was 12.6% and similar in both study groups (1.01, 0.85 to 1.20, 24 studies, 3631 patients). Few patients (2%) discontinued treatment because of adverse events with no difference between study groups (0.89, 0.52 to 1.52, 15 studies, 3042 patient). Nephrotoxicity was more common with combination therapy in nearly all studies, and the combined relative risk was 0.36 (0.28 to 0.47, fig 6), corresponding to a number needed to harm of 15 (14 to 17) for combination therapy.

Fig 6
Fig 6

Adverse events: nephrotoxicity in comparison of β lactam monotherapy v β lactam-aminoglycoside combination therapy for treatment of sepsis. Log scale of relative risks (95% confidence intervals), random effect model. Studies ordered by weight

Sensitivity analysisFigure 7 shows sensitivity analyses for measures of study quality. Two studies were quasi-randomised as they used patient identifications numbers for allocation (table 4).34 49 Concealment of allocation was adequate in 33% (21/64) of studies, and generation of allocation was adequate in 53% (34/64). Seven studies used some type of blinding, most commonly of outcome assessors only. Extraction of data by intention to treat was possible in 46% (20/43) of studies for fatality and in 21% (13/63) for failure (table 4).4 All sensitivity comparisons were non-significant. Adequate concealment and generation of allocation were associated with relative risks closer to 1 for fatality. The advantage of monotherapy was more significant in trials that used some type of blinding. Smaller trials showed larger effect estimates regarding failure. Analysis per protocol and by the fixed effect model did not affect results. The funnel plot for treatment failure generated a nearly symmetrical “funnel distribution.”

Fig 7
Fig 7

Sensitivity analyses Randomisation methods were classified as A=adequate; B=unknown; C=inadequate.85 Central randomisation, inaccessible computer randomisation, and sealed opaque envelopes were considered adequate for allocation concealment. Table of random numbers, computer generated lists, and consecutive selection were considered adequate for allocation generation. *Fatality comparison includes studies that reported results for all randomised patients (ITT=intention to treat) v studies reporting results for evaluable patients only (PP=per protocol). Studies that did not state method of analysis and did not refer to drop outs are not included. Failure comparison includes studies that reported results or drop outs for all randomised patients (drop outs counted as failures, ITT) v studies performed per protocol that did not state number of drop outs per study arm (PP). Results with all studies combined in this graph differ from those attained in main comparison because drop outs are counted as failures (relative risk 0.92, 0.82 to 1.03). † Comparison for studies comparing same β lactam was not performed as only one study used blinding

Table 4

Characteristics of included studies: methods

View this table:

Discussion

Main findings

In this systematic review of all randomised trials we have shown that β lactam-aminoglycoside combination therapy and β lactam monotherapy for the treatment of sepsis have similar effects in patients without neutropenia.

Twenty trials compared the same β lactam. All cause fatality, the most significant and objective outcome, was not reduced by the addition of aminoglycosides. Clinical and bacteriological failure, which may be prone to bias with non-blinded trials and are of much lesser relevance to patients, were not significantly different. However, rates of adverse event increased with the aminoglycoside. Nephrotoxicity was much more common with combination therapy, while vestibular damage and ototoxicity, other important morbidities associated with aminoglycosides, were not routinely examined.

Forty four trials compared a broad spectrum, usually novel, β lactam with a “routine” combination regimen. Rates of appropriate antibiotic treatment with combination therapy and monotherapy were similar when reported. Fatality was not significantly different. Failures were significantly more common with combination therapy. Among all trials, we found no evidence for any potential prevention of infection by resistant isolates with combination therapy.

How should these findings be interpreted?

It can be debated which design appropriately examines the clinical interpretation of synergism, studies comparing same or different β lactams. Synergism has been defined as a 2 log10 or greater reduction in bacterial count with the combination versus that with each of the agents alone.86 In studies comparing the same β lactam this is directly tested, but the effect of increasing the antibiotic spectrum cannot be separated from a synergistic effect. In studies comparing different β lactams the spectrum of coverage was similar in both arms. However, synergism can be examined only indirectly. If we assume that the aminoglycoside offers more than its additional coverage, the combination arm should perform as well, or better, than the broader spectrum β lactam monotherapy. With the former design we did not detect an advantage to the combination, while with the latter we found an advantage to monotherapy.

Weaknesses of the study

The quality of included studies was poor overall. We did not detect bias induced by any of the measures assessed. We could not obtain data on all cause fatality for 33% of studies. It is unlikely that missing results would shift the results for studies comparing the same β lactam (relative risk 1.02, 0.76 to 1.38), but it is of concern that studies comparing different β lactams (0.85, 0.69 to 1.05) may not detect important harm to patients.

Our assessment of treatment effects for patients with P aeruginosa, Gram negative, and blood infections relies on subgroup analysis. We did not detect an advantage for combination therapy among these patients. Only few patients with documented P aeruginosa infections could be evaluated. The types of infections addressed by included studies—severe infections acquired in the hospital or pneumonia acquired in intensive care units—suggest that further infections were caused by this pathogen.

Does further evidence support our findings?

Suggestions for combination treatment for P aeruginosa rely mostly on a prospective observational study of 200 patients with P aeruginosa bacteraemia, in which combination therapy was associated with improved survival and in which synergistic combinations were associated with a trend for improved survival compared with non-synergistic combinations.87 A similar study focusing on Klebsiella bacteraemia found an advantage for combination therapy only among patients with hypotension,15 while other studies have not found such an advantage.1012 88

Immunocompromised patients are the most likely to gain from enhanced bactericidal activity possibly offered by β lactam-aminoglycoside combination therapy.9 In a comparison of β lactam monotherapy with β lactam-aminoglycoside combination therapy restricted to patients with neutropenia we found no advantage to combination treatment.89 Although the approach to the management of patients with and without neutropenia is separated in clinical practice, this similarity supports a biological basis underlying our results.

Implications for practice

Antibiotic treatment is nearly always instituted empirically and is often continued with no isolate to direct specific treatment. Most trials assessed this scenario and do not support a benefit for combination therapy. Clinicians may still opt for combination empirical treatment to increase the probability of appropriate empirical treatment, which has indeed been shown to improve survival.90 91 Current evidence suggests that aminoglycoside monotherapy may be inadequate for infections outside the urinary tract.10 92 93 Thus, for the purpose of enhancing antimicrobial spectrum, aminoglycosides may constitute a poor choice. Combination treatment is considered for patients with severe infections. However, these are the patients most prone to harm by the addition of an aminoglycoside. With no proved survival benefit, combination therapy may be unjustifiable. Several studies, included in the overall and subgroup analyses, directly assessed semiempirical combination versus monotherapy. These, similarly, do not support combination therapy for specific pathogens, when detected.

Implications for further research

Should further research be conducted to assess combination versus monotherapy? Novel β lactams should not be compared with older generation β lactams or penicillins combined with aminoglycosides. The reason for further trials assessing the addition of an aminglycoside to a β lactam seems dubious as well. The relative risks and confidence intervals available with all current evidence do not point to a potential benefit overall or in specific subgroups of patients. Furthermore, assessment of efficacy among subgroups such as patients with P aeruginosa infections probably requires an unachievable number of patients treated empirically at the time benefit of antibiotic treatment is most evident.

We included in our review a small subset of trials that assessed the value of addition of an aminoglycoside in Gram positive infections. Three studies assessed staphylococcal endocarditis,21 48 63 one study assessed any staphylococcal infection,32 and one assessed streptococcal endocarditis.69 β lactam-aminoglycoside treatment is well ingrained in existing guidelines and clinical practice with these infections,94 yet our results do not point to a clinical benefit with combination therapy. With these infections, further studies should assess whether the addition of an aminoglycoside is justified.

What is already known on this topic

Early appropriate antibiotic treatment for severe infections decreases mortality

In vitro studies have shown that the bactericidal activity of a β lactam may be enhanced by the addition of an aminogoycoside

Prospective studies have suggested that the combination also has a clinical advantage

What this study adds

There is no difference in mortality when β lactam-aminoglycoside combination therapy is compared with β lactam monotherapy

Clinical failure and renal toxicity are more common with combination therapy

β lactam-aminoglycoside combination therapy does not improve clinical outcomes in patients with severe infections

Acknowledgments

We thank the Cochrane Infectious Diseases Group for their support, review process, and help in obtaining papers; all the authors who responded for our requests for additional data; and Rika Fujiya, who translated the Japanese studies. The protocol for this review with the detailed search strategy and methods is published in the Cochrane Library.95 The complete review will be published in the forthcoming issue of the Cochrane Library.

Footnotes

  • Contributors MP and LL performed the search. All authors selected trials for inclusion, performed data extraction and quality assessment of the trials, and analysed the data. MP and LL contacted authors and requested missing data. All authors participated drafting the manuscript for the Cochrane review and for the journal article. MP is guarantor for the article.

  • Funding This work was supported by an EU 5th framework grant (TREAT project, grant No 1999-11459).

  • Competing interests None declared.

  • Ethical approval Not required.

  • Embedded Image A full list of references to excluded studies can be found on bmj.com>

References

View Abstract