ABSTRACT
Vancomycin and teicoplanin are the glycopeptides currently in use for the treatment of infections caused by invasive beta-lactam-resistant gram-positive organisms. We conducted a systematic review and meta-analysis of randomized controlled trials that have compared vancomycin and teicoplanin administered systemically for the treatment of suspected or proven infections. A comprehensive search of trials without year, language, or publication status restrictions was performed. The primary outcome was all-cause mortality. Two reviewers independently extracted the data. Risk ratios (RRs) with 95% confidence intervals (CIs) were pooled by using the fixed-effect model (RRs of >1 favor vancomycin). Twenty-four trials were included. All-cause mortality was similar overall (RR, 0.95; 95% CI, 0.74 to 1.21), and there was no significant heterogeneity. In trials that used adequate allocation concealment, the results favored teicoplanin (RR, 0.82; 95% CI, 0.63 to 1.06), while in trials with unknown methods or inadequate concealment, the results favored vancomycin (RR, 3.61; 95% CI, 1.27 to 10.30). The latter trials might have recruited more severely ill patients. No other variable affected the RRs for mortality, including the assessment of glycopeptides administered empirically or for proven infections, neutropenia, the participant's age, and drug dosing. There were no significant differences between teicoplanin and vancomycin with regard to clinical failure (RR, 0.92; 95% CI, 0.81 to 1.05), microbiological failure (RR, 1.24; 95% CI, 0.93 to 1.65), and other efficacy outcomes. Lower RRs (in favor of teicoplanin) for clinical failure were observed with a lower risk of bias and when treatment was initiated for infections caused by gram-positive organisms rather than empirically. Total adverse events (RR, 0.61; 95% CI, 0.50 to 0.74), nephrotoxicity (RR, 0.44; 95% CI, 0.32 to 0.61), and red man syndrome were significantly less frequent with teicoplanin. Teicoplanin is not inferior to vancomycin with regard to efficacy and is associated with a lower adverse event rate than vancomycin.
Methicillin (meticillin)-resistant Staphylococcus aureus (MRSA) infections are a serious and constantly growing public health concern. The incidence of invasive MRSA infections in the United States was estimated to be 31.8 per 100,000 population in the general population in 2005, with a fatality rate of 6.3/100,000 population (32). The percentage of MRSA isolates among all S. aureus bloodstream isolates in hospitals was 49% in United States hospitals (1998 to 2005), with little variability in that proportion occurring between regions (68). In Europe, the proportions ranged from less than 1% in northern countries to >50% in southern countries (1999 to 2007) (17). Community-acquired MRSA is now of growing concern, reaching rates of more than 80% of all community-acquired S. aureus infections in certain locations in the United States (5, 31, 37).
The first-line treatment of choice for invasive MRSA infections is a glycopeptide antibiotic (43). Vancomycin (a glycopeptide) and teicoplanin (a lipoglycopeptide) are naturally occurring substances whose bactericidal activity is mediated mainly by the inhibition of peptidoglycan synthesis of the bacterial cell wall. Their spectrum of coverage is similar except for VanB vancomycin-resistant enterococci that are susceptible to teicoplanin (19, 30, 47). Teicoplanin is not approved for use in the United States, while in Europe it is as commonly used as vancomycin (2, 3, 69).
The comparative clinical efficacies and toxicity profiles of vancomycin and teicoplanin are not established. In a previous review, vancomycin and teicoplanin were found to be equally efficacious, with teicoplanin causing fewer adverse effects than vancomycin (76). Since then, the findings of more trials comparing vancomycin and teicoplanin have been published. We performed a systematic review and meta-analysis of randomized controlled trials that compared vancomycin to teicoplanin. The objectives of our review were to compare the efficacy and safety of these glycopeptides.
MATERIALS AND METHODS
Inclusion criteria.We included randomized or quasirandomized controlled trials that compared systemic treatment with vancomycin versus teicoplanin for suspected or proven infections in adults and children. We included both neutropenic and nonneutropenic patients. Additional antibiotic treatment was permitted, provided that the same antibiotic and dose or the same rules regarding additional antibiotics were applied in both study arms.
Outcomes.The primary outcome assessed was all-cause mortality and was preferentially extracted at day 30. Secondary outcomes included clinical failure, defined as a nonresolved infection, treatment modification, or death due to the infection; microbiological failure, defined as the persistence or the reappearance of the initiating pathogen during treatment, as defined in the study (after day 3); relapse, defined as the reisolation of the initiating pathogen after the completion of treatment; superinfection, defined as the development of a different infection during treatment or within 1 week of the discontinuation of treatment; resistance development, defined as the isolation of glycopeptide-resistant gram-positive bacteria; and adverse events, including adverse events requiring the discontinuation of treatment, nephrotoxicity (as defined in the study), red man syndrome, and rash.
Search strategy.We searched the Cochrane Register of Controlled Trials, PubMed, and LILACS databases. Unpublished trials were sought in the references of all selected studies; the conference proceedings of the Interscience Conference on Antimicrobial Agents and Chemotherapy, the European Congress of Clinical Microbiology of Infectious Diseases, and the Infectious Diseases Society of America; trial registries; ongoing trial databases; and personal contacts with the investigators of the trials included. No language or date restrictions were imposed. The last search of all sources was performed in June 2008. The terms “glycopeptides” and “chemical” and the generic and trade names of vancomycin and teicoplanin were searched for, in combination with the use of the Cochrane filter for randomized controlled trials in PubMed (28).
Review methods.Two reviewers (S.S. and M.P.) independently applied the inclusion criteria and extracted the data. We identified the intention-to-treat population of each trial (all randomized participants), all patients treated per protocol, and microbiologically evaluable patients. Outcomes were extracted preferentially or imputed for the intention-to-treat population. Whenever data were missing, we contacted the authors and primarily requested data on the trial methods used and the primary outcome. We assessed the risk of bias in the studies included using domain-based evaluation. The domains assessed included sequence generation, allocation concealment, blinding, early stopping, selective reporting, patient attrition, the number of patients excluded from the analysis, and unit-of-analysis errors (recruitment of patients more than once or reporting of more than one outcome per patient without adjustment for multiple testing). Each domain was scored as adequate, unclear, or inadequate by using the criteria suggested in the Cochrane handbook (28). We judged that allocation concealment for all outcomes and double blinding for outcomes other than mortality carried the highest risk for bias and thus report on sensitivity analyses for these domains.
Statistical analysis.Risk ratios (RRs) for individual studies were calculated with 95% confidence intervals (CIs). RRs of >1 favor vancomycin. Heterogeneity was assessed by the chi-square test (P < 0.1) and the I2 measure of inconsistency (I2 > 50%). Meta-analyses were conducted by using the fixed-effect model, unless significant heterogeneity was present, in which case the random-effects model was used. The effects of risk of bias assessment are reported as subgroup differences on the basis of a fixed-effect inverse variance meta-analysis (15). Small-study effects were assessed through visual inspection of funnel plots. Analyses were conducted by using the RevMan (version 5) program (Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark, 2008).
RESULTS
The trial flow is depicted in Fig. 1. The search yielded 269 publications, 61 of which were potentially eligible; and 24 individual randomized controlled trials comparing vancomycin versus teicoplanin fulfilled inclusion criteria (1, 4, 8, 10, 11, 14, 18, 20, 22, 33, 34, 40, 42, 45, 46, 48, 49, 51, 60, 62, 71-72, 77).
The trials were conducted between 1986 and 2007 and included a total of 2,332 patients (Table 1). Twelve trials addressed patients with febrile neutropenia, and the assigned treatment was mainly empirical (Fig. 1). Twelve studies recruited patients without neutropenia, and treatment was initiated mostly for suspected or documented infections with gram-positive bacteria. Three trials recruited patients with MRSA infections only or mostly, one trial recruited patients with methicillin-sensitive S. aureus (MSSA) infections only, and the median prevalence of MRSA infections in the other studies reporting on MRSA was 2% (range, 0 to 26%). MRSA isolates were reported to be susceptible to glycopeptides, but the studies used various breakpoint definitions, depending on the year of the study. The median rate of bacteremia in 17 trials reporting on the number of patients with bacteremia was 50% (range, 12.5 to 100%). Children alone were assessed in four trials, and adults and children were assessed in one trial; all other trials addressed patients >12 years of age. Patients with renal failure were excluded from 17 trials.
Characteristics of included trials
Four trials were performed prior to Eli Lilly's launch of purified vancomycin in 1990 (8, 11, 62, 72), and in one study the purified preparation was introduced during the trial (71). The standard dosing of teicoplanin for adults was 400 mg/day for adults and 10 to 12 mg/kg of body weight/day for children, following a loading dose (Table 1); lower doses (200 mg/day) were used for adults during the first part of the study in three trials (45, 62, 71). The intramuscular administration of teicoplanin was permitted in four trials. Monitoring of serum vancomycin and teicoplanin levels was explicitly stated in 12 and 11 trials, respectively. None of the trials stated the target serum level. The serum drug levels achieved during the trial were reported in seven trials; median trough values for teicoplanin and vancomycin were 8.8 μg/ml (range, 4.7 to 15.9 μg/ml) and 8.2 μg/ml (range, 8 to 14.3 μg/ml), respectively.
Adequate allocation generation and concealment were described in 11 trials. Three additional trials described only adequate allocation generation, and four trials described only adequate allocation concealment. Two trials were triple blinded, three were double blinded, and the remaining trials were open label. In eight trials, unit-of-analysis errors were detected (Table 1). The full data for one trial were provided by the authors (1). Additional data on methods or outcomes were obtained for 11 trials.
All-cause mortality.Eighteen trials reported all-cause mortality at the end of follow-up (14 to 50 days). Mortality was similar for teicoplanin and vancomycin (RR, 0.95; 95% CI, 0.74 to 1.21) (Fig. 2). No significant heterogeneity was detected for the overall comparison (P = 0.57, I2 = 0%). There were no statistically significant differences between teicoplanin and vancomycin among neutropenic and nonneutropenic patients, by empirical treatment and treatment of suspected or proven infections caused by gram-positive bacteria, for adults and children, for different teicoplanin dosing regimens, for different types of vancomycin preparations, and with or without drug level monitoring (Table 2). There was no statistically significant difference in mortality in the subgroups of patients with bacteremia caused by gram-positive bacteria (RR, 1.84; 95% CI, 0.65 to 5.18; six studies and 158 patients).
All-cause mortality for trials. Trials are subgrouped by the methods of allocation concealment used in the trial. M-H, Mantel-Haenszel.
Subgroup analyses for vancomycin versus teicoplanin
When the results were stratified according to the methodological quality of the studies, divergent results were observed (Fig. 2). With adequate methods for concealment of the allocation sequence, the RR for mortality was 0.82 (95% CI, 0.63 to 1.06) in favor of teicoplanin, while unknown methods or inadequate concealment resulted in an RR of 3.61 (95% CI, 1.27 to 10.30) significantly in favor of vancomycin (P = 0.01 for difference between subgroups). However, the studies with a higher risk of bias included more patients with S. aureus bacteremia (median, 100%; range, 62 to 100%) than the low-risk studies did (median, 45.5%; range, 20 to 100%). In four double-blind trials, the RR was 0.48 (95% CI, 0.22 to 1.07), while in open trials it was 1.04 (95% CI, 0.80 to 1.36). There was no evidence of the small-study effect overall (Fig. 3).
Funnel plots. (A) All-cause mortality; (B) nephrotoxicity. RRs are plotted against standard errors, as a measure of the study's precision.
Secondary outcomes.No significant difference between teicoplanin and vancomycin with regard to clinical failure was shown overall (RR, 0.92; 95% CI, 0.81 to 1.05; 21 trials and 1,729 patients). RRs tended to be in favor of teicoplanin for the nonneutropenic subgroup of patients, when treatment was initiated for a suspected or a proven infection with gram-positive bacteria, with higher doses of teicoplanin, and with adequate allocation concealment and double blinding (Table 2). A modified intention-to-treat analysis, with the assumption of treatment failure for all dropouts (excluding studies in which the number randomized was unclear), yielded an RR of 0.94 (95% CI, 0.84 to 1.05; 17 trials and 1,679 patients). Clinical failure could also be assessed in the subgroups of patients with documented S. aureus infections (RR, 0.97; 95% CI, 0.63 to 1.51; 13 trials and 289 patients), MRSA infections (RR, 0.67; 95% CI, 0.26 to 1.72; 5 trials and 73 patients), and bacteremia (RR, 0.95; 95% CI, 0.73 to 1.24; 15 trials and 718 patients).
The microbiological failure rates were not significantly different for teicoplanin treatment and vancomycin treatment (RR, 1.24; 95% CI, 0.93 to 1.65; 12 trials and 716 patients). No significant differences in the rates of relapse (RR, 0.61; 95% CI, 0.29 to 1.30; 8 trials and 330 patients) and superinfection (RR, 1.02; 95% CI, 0.77 to 1.37; 10 trials and 1,083 patients) were found. None of the trials reported quantitatively on the changes in the glycopeptide MICs for persisting gram-positive bacteria. Seven trials reported qualitatively that no resistant isolates were detected, except in 3/68 patients in one trial (18) who acquired teicoplanin-resistant isolates following treatment with teicoplanin.
Adverse events.The overall number of adverse events was reported in all trials except the unpublished trial (1). There were significantly fewer total adverse events for teicoplanin than for vancomycin, when they were reported as the number of adverse event episodes per patient episode (RR, 0.61; 95% CI, 0.50 to 0.74; 23 trials and 2,046 patient episodes) and when they were reported as the number of patients experiencing at least one adverse event per patient (RR, 0.57; 95% CI, 0.45 to 0.72; 19 trials and 1,641 patients). There were significantly fewer adverse events requiring the discontinuation of treatment with teicoplanin (RR of 0.54 and 95% CI of 0.33 to 0.87 for 13 trials and 904 patient episodes; RR of 0.59 and 95% CI of 0.36 to 0.97 for 12 trials and 829 patients). There was significantly less nephrotoxicity reported for teicoplanin (Fig. 4) (RR of 0.44 and 95% CI of 0.32 to 0.61 for 21 trials and 1,992 patient episodes; RR of 0.33 and 95% CI of 0.22 to 0.50 for 17 trials and 1,587 patients). Nephrotoxicity was defined heterogeneously as creatinine levels above the normal range (1.1 to 1.5 mg/dl), by an absolute increase of 0.5 mg/dl or as a 50% to 100% increase from the baseline level. Similar RRs were observed in studies that recruited children, although none of the differences were statistically significant among children alone (RR for any adverse event, 0.53 [95% CI, 0.24 and 1.18; four trials and 216 children]; RR for adverse events requiring discontinuation, 0.66 [95% CI, 0.11 and 3.9; two trials and 147 children]; RR for nephrotoxicity, 0.31 [95% CI, 0.09 to 1.07; three trials and 197 children]). Severe nephrotoxicity, defined as the need for hemodialysis, was reported only with vancomycin and among adults (RR, 0.16; 95% CI, 0.03 to 0.86; eight trials and 341 patients [events were reported in three trials]). There were no cases of red man syndrome in the teicoplanin group, whereas there was a 5% incidence of red man syndrome in the vancomycin group (RR of 0.21 and 95% CI of 0.08 to 0.54 for 10 trials and 756 episodes and RR of 0.24 and 95% CI of 0.08 to 0.76 for 7 trials and 500 patients); this outcome was assessed only among adults. The occurrence of other rash was not significantly different between the groups (RR of 0.74 and 95% CI of 0.49 to 1.12 for 16 trials and 1,707 episodes; RR of 0.65 and 95% CI of 0.41 to 1.03 for 12 trials and 1,398 patients).
Nephrotoxicity. Trials are subgrouped by the type of vancomycin preparation used in the trial: trials that used purified vancomycin and those reporting on the use of unpurified vancomycin for part or all of the trial. Similar results were obtained when the results for one trial published in 1991 were removed from the unpurified subgroup. M-H, Mantel-Haenszel.
Nephrotoxicity was less frequent with teicoplanin in trials with both purified and unpurified vancomycin (Fig. 4) and both in studies that reported routine monitoring of vancomycin levels (RR, 0.31; 95% CI, 0.18 to 0.51; 9 studies and 663 episodes) and in those that did not (RR, 0.60; 95% CI, 0.38 to 0.95; 12 studies and 1,329 episodes). In a further sensitivity analysis that excluded all trials published prior to 1994, the results were similar (RR for nephrotoxicity, 0.48; 95% CI, 0.32 to 0.72; 15 trials and 1,662 episodes), and there was no significant association by meta-regression analysis between the year of the trial's start and the total number of adverse events or nephrotoxicity (P = 0.28 and P = 0.73, respectively). Similarly, there was no association between trial size and effects (Fig. 3).
DISCUSSION
We compared the efficacy and safety of the glycopeptides currently in use, teicoplanin and vancomycin. All-cause mortality was the primary outcome, since this outcome is the most objective and its prevention is the primary motivation for the treatment of severe infections.
There was no significant difference in all-cause mortality between teicoplanin and vancomycin overall. In studies reporting adequate allocation concealment (thus ensuring appropriate randomization), there was a trend in favor of teicoplanin, while studies with unclear or inadequate allocation concealment showed a significant advantage for vancomycin. The latter studies recruited more patients with bacteremia. There were no significant differences between teicoplanin and vancomycin with regard to clinical failure, microbiological failure, and other secondary efficacy outcomes. There were significantly fewer adverse events with teicoplanin than with vancomycin, including events requiring the discontinuation of treatment, nephrotoxicity, and red man syndrome. The effect estimate for nephrotoxicity was an RR of 0.33 (95% CI, 0.22 to 0.50), which translates into a number needed to harm (NNH) of 14 (95% CI, 11 to 25) patients. This effect was not associated with the year of the study or drug level monitoring. The effect was similar when the analysis was limited to trials that used the currently available purified preparation of vancomycin (RR, 0.44 [95% CI, 0.30 to 0.65]; NNH, 20 [95% CI, 14 to 33]).
Comparative assessment for the development of resistance is of interest in an era of glycopeptide resistance, especially when a policy of using one or another glycopeptide is being considered. It is an important untoward effect of antibiotic therapy with implications for the individual treated and future patients. Randomized controlled trials provide a well-controlled framework for its assessment. However, resistance development was poorly reported in all trials, both old and new. None of the trials compared quantitatively the changes in MICs of the persisting S. aureus isolates. Qualitatively, several trials comparing teicoplanin to vancomycin reported that resistance development was null. However, the definitions of “resistance” have changed since the years that those studies were conducted, following CLSI breakpoint definitions (70). While susceptibility to vancomycin is currently defined as an MIC of ≤2 μg/ml, the breakpoints used in the older studies ranged from 4 to 8 μg/ml. Thus, we do not have information on the emergence of S. aureus strains with MICs of >2 μg/ml from randomized trials comparing vancomycin to teicoplanin. Resistance to teicoplanin emerges sooner than resistance to vancomycin in S. aureus strains in observational studies and case reports (38, 39, 50); however, the culprit antibiotic inducing glycopeptide resistance in clinical isolates was usually vancomycin (61).
The main limitation of the available evidence is the paucity of patients with MRSA infections in existing trials. This limited the ability of the primary studies and our analysis to assess clinical and microbiological efficacy for the treatment of MRSA infections (the rate of mortality for patients infected with MRSA was reported in a single trial [72]), to stratify the analyses by the baseline isolates' susceptibility, and to adequately assess the development of resistance. However, the trials recruited mostly septic patients, who are commonly given glycopeptides (at least empirically) in clinical practice. Data on mortality were incomplete and were available for only 17/24 trials; the authors could not supply further data since the trials were old. We noted a divergence in the results by randomization method that can be explained by a true effect (the better efficacy of teicoplanin) or by the association between studies with a higher risk for bias and sicker patients, as reflected by the higher percentage of patients with bacteremia in higher-risk studies (in this case, indicating the better efficacy of vancomycin). Unbalanced randomization (allocating less seriously ill patients to teicoplanin with inadequate methods for concealment) is another possibility, although we did not find a difference in baseline patient characteristics (data not shown).
Increasing MICs for vancomycin have been observed in S. aureus isolates over the last decade and is termed “MIC creep” (64). Increasing MICs, in turn, have been associated with vancomycin treatment failures, even within the currently defined susceptible range (MIC ≤ 2 μg/ml). MICs of 2 μg/ml and even between 1 and 2 μg/ml were significant and independent predictors of treatment failure in observational studies (27, 35, 41, 58, 63). This has led to recommendations regarding vancomycin level monitoring and dosing. For complicated infections (bacteremia, endocarditis, osteomyelitis, meningitis, and hospital-acquired pneumonia) and for infections caused by strains with MICs of >1 μg/ml, trough levels of 15 to 20 μg/ml are recommended (56). The usual dosing strategies are insufficient to achieve these concentrations in patients with normal renal function, and doses of vancomycin up to 4 g/day are necessary. As these doses are associated with increased nephrotoxicity (36), monitoring of trough serum levels is recommended (56). Similar data for teicoplanin are largely lacking. A mean daily dose of 4 mg/kg was associated with treatment failure when compared to a mean daily dose of 6 mg/kg, but even the higher dose resulted in low mean trough levels of 7.8 ± 4.8 μg/ml (24). Doses of 6 mg/kg twice daily achieved serum concentrations of >10 μg/ml by the second day of treatment. Thus, the administration of teicoplanin loading doses of 6 mg/kg twice daily for 48 h has been recommended for all infections, and for complicated infections, continuation of the high-dose regimen is recommended (6). Notably, in our review demonstrating the similar efficacies of teicoplanin and vancomycin, once-daily dosing of teicoplanin was used in all studies.
To resolve the remaining questions on the efficacy and safety of vancomycin versus those of teicoplanin, a contemporary trial that uses adequate randomization methods and that recruits patients with bacteremia (preferably MRSA bacteremia) is needed. Such a trial is important, given the common use of these agents and the implication of a policy to use one or another drug. New lipoglycopeptides targeting improved pharmacokinetics, antibacterial activity, and spectrum of coverage against vancomycin-resistant enterococci and staphylococci have been developed. Telvancin, dalbavancin, and oritavancin are currently undergoing clinical testing (Fig. 1). Trials assessing both old and new glycopeptides should target patients who will be given glycopeptides in clinical practice. Skin and soft tissue infections, which are frequently targeted for the assessment of new antibiotics, constitute a poor choice for the assessment of antibiotic efficacy, since the outcomes do not necessarily depend on the antibiotic treatment. Studies should address resistance development using sensitive methods for the detection of glycopeptide resistance (39, 44, 50, 74).
Given the similar efficacies teicoplanin and vancomycin and the lower rate of adverse events with teicoplanin, including serious adverse events, the use of teicoplanin for the treatment of infections caused by MRSA and other resistant gram-positive organisms should be considered. Uncomplicated infections permit once-daily dosing of teicoplanin and the possibility of intramuscular injection. Complicated infections probably mandate higher dosing strategies for both vancomycin and teicoplanin that should be accompanied by monitoring for renal toxicity, especially with vancomycin.
ACKNOWLEDGMENTS
We thank the authors who responded to our mail and provided additional data.
We have no conflicts of interest to declare.
This work was performed in partial fulfillment of the M.D. thesis requirements of the Sackler Faculty of Medicine, Tel Aviv University (S.S.).
FOOTNOTES
- Received 12 March 2009.
- Returned for modification 6 April 2009.
- Accepted 6 July 2009.
- Copyright © 2009 American Society for Microbiology