Flow diagram of results of the literature search for comparative studies of the additive effect of rifampin, including in vitro investigations, animal investigations, and human investigations. The schematic indicates how many hits were found on the search, the number of articles excluded from the systematic review, and the reasons for exclusion from the systematic review. *Review was determined to be unlikely to yield references relating to adjunctive rifampin therapy for Staphylococcus aureus infections based on the abstract (eg, review of treatment of S aureus colonization). †Inappropriate comparisons for the systematic review included studies relating to biofilm diffusion, pharmacokinetics, descriptive ecology, S aureus prophylaxis, treatment of colonization, non–S aureus organisms, clinical microbiology methods, review of medication adverse events, comparison of intravenous and oral therapies, antibiotic effects on virulence factor production, compartmental pharmacokinetics and pharmacodynamics, investigations of antibiotic-impregnated catheters or devices, observational investigations of antibiotic use and antimicrobial resistance, and epidemiologic investigations (including epidemiology of antimicrobial resistance, molecular epidemiologic studies, and descriptive molecular analyses). In addition, noncomparative studies and conference abstracts were excluded. ‡Irrelevant studies for the systematic review included studies or articles relating to treatment of colonization or pharmacokinetics. In addition, errata and conference abstracts were excluded.
Perlroth J, Kuo M, Tan J, Bayer AS, Miller LG. Adjunctive Use of Rifampin for the Treatment of Staphylococcus aureus InfectionsA Systematic Review of the Literature. Arch Intern Med. 2008;168(8):805-819. doi:10.1001/archinte.168.8.805
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
Staphylococcus aureus causes severe life-threatening infections and has become increasingly common, particularly methicillin-resistant strains. Rifampin is often used as adjunctive therapy to treat S aureus infections, but there have been no systematic investigations examining the usefulness of such an approach.
A systematic review of the literature to identify in vitro, animal, and human investigations that compared single antibiotics alone and in combination with rifampin therapy against S aureus.
The methods of in vitro studies varied substantially among investigations. The effect of rifampin therapy was often inconsistent, it did not necessarily correlate with in vivo investigations, and findings seemed heavily dependent on the method used. In addition, the quality of data reporting in these investigations was often suboptimal. Animal studies tended to show a microbiologic benefit of adjunctive rifampin use, particularly in osteomyelitis and infected foreign body infection models; however, many studies failed to show a benefit of adjunctive therapy. Few human studies have addressed the role of adjunctive rifampin therapy. Adjunctive therapy seems most promising for the treatment of osteomyelitis and prosthetic device–related infections, although studies were typically underpowered and benefits were not always seen.
In vitro results of interactions between rifampin and other antibiotics are method dependent and often do not correlate with in vivo findings. Adjunctive rifampin use seems promising in the treatment of clinical hardware infections or osteomyelitis, but more definitive data are lacking. Given the increasing incidence of S aureus infections, further adequately powered investigations are needed.
Staphylococcus aureus infections are common, severe, and associated with significant morbidity and mortality. Staphylococcus aureus is the most common cause of skin and soft-tissue infections and is a frequent cause of serious infections such as health care–associated bloodstream infections,1,2 device-associated infections,3,4 and osteomyelitis.5,6 Worldwide, S aureus is the most common cause of infective endocarditis.7,8
Of concern, the number of infections caused by methicillin-resistant S aureus (MRSA) continues to rise. In intensive care units, the proportion of S aureus infections that are MRSA in the United States has been increasing by 3% per year, and MRSA now constitutes more than 60% of S aureus strains that cause infections.9 Methicillin-resistant S aureus has been recently implicated as the causative agent of life-threatening community-acquired infections such as sepsislike syndromes, necrotizing pneumonia, and necrotizing fasciitis.10- 12 Therapeutic options for MRSA are more limited than those for methicillin-susceptible S aureus strains. Furthermore, treatment failure with standard therapies for MRSA is common.13- 15 There is a need to better understand the efficacy of antimicrobial therapies for MRSA and difficult-to-treat S aureus infections.
Rifampin, a broad-spectrum antimicrobial agent that is bactericidal against S aureus, achieves high intracellular levels and is one of the few antimicrobial agents that can penetrate biofilms and kill organisms in the sessile phase of growth.16- 18 Its use as monotherapy has been abandoned because of the rapid development of resistance, which is prevented by combination with another active antibiotic. Combination therapy with rifampin has been used to treat S aureus infections.19,20 Nevertheless, it has been commonly used adjunctively to treat S aureus infections.19,20 However, data to support this practice are limited and are typically based on small clinical studies or basic in vitro investigations. To our knowledge, there has been no attempt to synthesize the literature that has examined the efficacy of rifampin therapy against S aureus. To this aim, we conducted a systematic review of the use of rifampin as adjunctive therapy to treat S aureus infections.
To identify in vitro, animal, and human subject data regarding the efficacy of rifampin as adjunctive therapy for the treatment of S aureus, 2 independent reviewers (J.P. and M.K.) searched PubMed, Cochrane Library, and EMBASE for publications containing the text phrases Staph* AND rifamp*. These terms were used to avoid ignoring articles with permutations of the words Staphylococcus and rifampin, (eg, Staph aureus or rifampicin). All abstracts were printed for review. The search was limited to English-language articles published between January 1, 1966, and January 31, 2006. We reviewed only manuscripts relating to S aureus and not coagulase-negative staphylococci. We also contacted several experts in the field of adjunctive rifampin therapy to determine if there were any pertinent recently published meeting abstracts or published articles that we missed by our systematic review.
An investigation was included in our systematic analysis if it met each of the following criteria: (1) the organism under study was S aureus; (2) the study design compared the efficacy of 1 or more antibiotics alone and in combination with rifampin; (3) the study outcome assessed quantitative bacterial measurements, cure rates (or eradication of colonization), or staphylococcal-related mortality; and (4) outcome data were explicitly reported. By the nature of our methods, we excluded studies reporting the efficacy of rifampin therapy alone compared with other antibiotics, the efficacy of rifampin as prophylaxis to prevent infections in uninfected hosts, or the use of rifampin-impregnated devices or catheters. Also excluded were articles not containing original research (eg, reviews, editorials, case reports, abstracts, and letters). We also examined the bibliographies of selected review articles for original research articles that may have contained references of articles that were missed by our search criteria.
Each investigator was blinded to the other investigator's data extraction. The 2 reviewers independently rejected or accepted each abstract based on the inclusion and exclusion criteria. Article texts of selected abstracts were reviewed, as were article texts of abstracts that could not be excluded based on abstract review alone. All disagreements between the abstractors as to whether the article should be included were settled by a third independent reviewer (L.G.M.). Data from each trial were entered onto a standardized form, verified for accuracy, and input into a computerized database. Information extracted included study design, antibiotics tested, number of subjects, year of publication, duration of follow-up, clinical setting (in vitro, animal, or human), and intervention (dosage, frequency, and duration of therapy or exposure). Abstracted data included the outcome (eg, mortality, clinical failure, and colony count after treatment) and the time of evaluation of treatment outcome. Discrepancies in data between abstractors were identified and resolved via discussion among the investigators. For human studies, 2 independent reviewers (M.K. and J.T.) reviewed the 8 human subject trials and assigned them Jadad scores; disagreements between the reviewers were settled by a third independent reviewer (L.G.M.).
For investigations that did not perform statistical analyses but reported results, we attempted to perform statistical analyses. For investigations comparing dichotomous outcomes, we performed χ2 or Fisher exact test. For studies comparing means, if standard deviations and group sample sizes were available, we performed Wilcoxon rank sum test. We did not perform analyses from investigations that reported results of statistical tests or contained inadequate information for us to perform tests of significance. We initially planned to perform a meta-analysis of results but abandoned this method because study outcome heterogeneity was substantial (eg, in vivo studies variably used outcomes of cure rates, proportion of sterile cultures, decrease in colony-forming units [CFUs], and others). In addition, performing a meta-analysis was problematic because of the large number of strata in disease studied (eg, bacteremia and abscess) and in treatment (β-lactam antibiotics, glycopeptides, fluoroquinolones, and others).
The results of the literature search and the reasons for exclusion from the systematic review are summarized in the Figure. To facilitate review of our findings, results are summarized herein as in vitro investigations, animal investigations, and human investigations.
We identified 72 publications comparing antibiotic efficacy with and without rifampin using in vitro models (Table 1A and Table 1B). Of 164 individual antibiotic trials in these 72 publications, 41 trials tested both methicillin-susceptible S aureus and MRSA strains. We found that methods used to determine the nature of the antibiotic interactions were heterogeneous and included E test, time-kill curves, checkerboard assays, serum bactericidal activity, and ex vivo intracellular bactericidal activity. Even within-method differences (eg, among time-kill studies) were substantial in terms of the media used, the inoculum (from 104 to 109 bacteria), the growth phase (ie, stationary vs log phase), the temperature of inoculation (from 36°C to 38°C), the time the CFU outcomes were quantified (from 6 to 48 hours after inoculation), and the concentration of the antibiotics used (0.25 to many times the mean inhibitory concentration).
Methodological variability was reflected in the inconsistent interactions reported. Many studies using time-kill assays failed to report results in terms of synergy or antagonism, presenting the data only graphically and without statistical analysis.21- 25 Definitions of synergy, antagonism, or indifference in individual investigations did not necessarily correlate with those of other in vitro studies26- 30 or were not provided.31 Some investigations used categories of “additive” interactions26,32- 34; others did not specify the interaction other than as “not synergistic,”35 “not antagonistic,”36 “improved,”37,38 “reduced,”39 or “enhanced.”40 Some did not report each specific interaction.41,42 As shown in Table 1A and Table 1B, many of the antibiotics tested with rifampin had inconsistent outcomes. Some authors noted that different concentrations of antibiotics changed the nature of the interaction.30,36,43,44 Several investigations revealed different interactions depending on the method used (time-kill vs checkerboard).30,33,37,45,46 Some investigators attempted to validate in vitro findings with in vivo models or clinical outcomes.18,25,35- 37,47,48 Because outcome definitions were heterogeneous and often missing, we thought it was hazardous and likely inappropriate to summarize data or to state general conclusions from the in vitro investigations.
In the animal models, between-study differences included the dosing route, dosing frequency, antibiotic dosages, S aureus strain used, animal model investigated, timing of outcome assessment, end point studied (eg, microbiologic or cure), and duration between bacterial inoculation and therapy initiation (Table 2). Many studies did not report statistical analyses. To facilitate description of the animal investigations, we first stratified models by disease and then by primary antibiotic treatment.
In peritonitis models, we identified 3 studies. Of 2 investigations of fluoroquinolones, one noted that combination therapy resulted in a larger decrease in CFUs per milliliter and sterilization of fluid but did not report a statistical analysis.47 The other investigation of fluoroquinolones used the same end points and found the fluoroquinolone-rifampin combination to be superior.49 A third investigation stated that there was indifference or a simple additive effect when oxazolidinones were combined with rifampin.50
Two investigations examined skin and soft-tissue infections (without hardware). One abscess model demonstrated that adjunctive rifampin with teicoplanin resulted in diminished CFU decrease compared with teicoplanin monotherapy.51 The other abscess model showed that ciprofloxacin plus rifampin was no more effective as either antibiotic alone.89 A third study of cloxacillin in a mastitis model found that dual therapy resulted in statistically significant CFU decreases in 3 experiments but no significant difference in a fourth.25
Of 16 antibiotics used in 12 publications pertaining to endocarditis, 7 showed superiority of combination therapy, 4 showed superiority of monotherapy, and 5 showed no difference. Dual therapy was superior in terms of decreases in CFUs or valve sterilization in trials using vancomycin,48 cloxacillin,52 ciprofloxacin,23 quinupristin-dalfopristin,52 teicoplanin,51 and daptomycin.53 However, in other investigations, dual therapy with vancomycin was indifferent,31,54 and dual therapy with ciprofloxacin was worse than monotherapy.23 Dual therapy with oxazolidinones had inconsistent results, with decreases in CFUs reported only with low doses of linezolid.21,50 Taken in sum, investigations of dual therapy with β-lactam antibiotics tended to show improved microbiological outcomes as opposed to survival or valve sterilization (Table 2).
In S aureus bacteremia models, an investigation showed a benefit to adding rifampin to methicillin or trimethoprim.55 In a survival study using single-dose penicillin or trimethoprim, the same authors reported a similar benefit (although the therapeutic groups were directly compared with control groups instead of with each other but nonetheless can be compared in a post hoc manner).55
In osteomyelitis models, 16 antibiotic trials were reported in 8 publications. Statistical analyses revealed significant reductions in positive bone cultures,56 increases in sterile bone cultures,35,37,56- 58 or reductions in CFUs per gram.58,59 In 2 treatment groups from 1 publication, no discrete numbers were reported in the text of the publication.60 No trial demonstrated worse clinical or microbiological results in dual-therapy arms.
Nine treatment comparisons presented in 4 publications of device- or hardware-associated infections with adjunctive rifampin therapy were identified. In an abscess model with implanted foreign material, dual therapy with rifampin and either vancomycin, ciprofloxacin, fleroxacin, or teicoplanin showed superiority in terms of cure rate.18 Fleroxacin-rifampin was superior in another trial measuring sterilization of abscess fluid.61 A study of vancomycin and fleroxacin showed decreased CFUs in abscess fluid and on foreign material with adjunctive rifampin.62 An infected partial knee replacement model testing vancomycin or quinupristin-dalfopristin showed a significant reduction of CFUs in bone with adjunctive rifampin and a higher proportion of animals with sterile bone cultures.63 Finally, in a meningitis model, rifampin added to nafcillin or vancomycin added modest decreases in colony counts but no improvement with sterilization of cerebrospinal fluid cultures.64
Some investigators attempted to correlate in vitro and in vivo effects of rifampin therapy. Commonly, a contradiction or poor correlation was found between the 2 methods.18,36,37,47,48 We should note, while some animal investigations’ outcomes achived statistical significance, the clinical significance of differences are unclear. For example, in an investigation of mastitis in a mouse model, adjunctive rifampin when added to cloxacillin was associated with a significant (P <.05) decreases in mean log10 bacteria/gland (5.7 ± 0.2 vs 5.0 ± 0.2).25 Some other significant differences were likewise of unclear clinical significance (data not shown).
Of 7 identified trials in humans that compared antibacterial therapy with or without rifampin, 6 were prospective randomized trials, 2 were placebo controlled, and 1 was retrospective (some were of combined designs) (Table 3). Antibiotics used in the investigations (in varying dosages), were vancomycin, pefloxacin, ciprofloxacin, oxacillin, fleroxacin, and nafcillin. Rifampin doses varied (typically 600-1200 mg/d). Study populations were small (15-65 patients) and diverse in terms of comorbidities (malignant neoplasms, trauma, and previous surgery), sites of infection (wound, osteomyelitis, and bacteremia), and acuity of infection (acute to many years in the case of osteomyelitis). End points were heterogeneous and included cure, clinical improvement, and persistence of bacteremia. Duration of follow-up varied from several days to more than 3 years. The mean Jadad score of the studies was 2 (median, 1 [range 0-5]), One study65 did not separate S aureus from coagulase-negative staphylococcal infections in its intent-to-treat analysis, although the differences between groups in the as-treated analysis could be determined.
β-Lactam antibiotics (oxacillin or nafcillin) were studied in 4 trials. Oxacillin was used in 2 trials of heterogeneous infection types and resulted in significantly improved cure rates in one66 but not another.67 Nafcillin was studied in 2 osteomyelitis trials demonstrating that dual therapy was equivalent to monotherapy in one study68 but superior in another.69
Vancomycin was used in 2 studies66,67 (in which oxacillin therapy was also analyzed) and as the primary study drug in an endocarditis trial of MRSA.70 In the first 2 trials, vancomycin-rifampin dual therapy was superior in terms of cure or improvement66 and for bacteriologic success.67 For the endocarditis investigation, there was no difference in outcome with adjunctive rifampin.70
In a trial of treating hardware infections with fluoroquinolones, clinical cure was achieved more often with dual therapy in a study (P =.002).65 Cure rates were superior in an as-treated analysis for those receiving dual therapy compared with monotherapy (P =.04).65 The investigation also showed a trend toward a benefit for the rifampin-containing therapy (16 of 18 cured vs 9 of 15 cured, P =.10) in the intent-to-treat analysis of patients who did not have their hardware removed, although this analysis included some patients with coagulase-negative Staphylococcus infections. In another trial using pefloxacin for osteomyelitis or septic arthritis infections, there was no advantage to adjunctive rifampin therapy, although the cure rates in the monotherapy arm were high, creating a ceiling effect.71 In summary, human trials investigating adjunctive rifampin use have occasionally demonstrated a beneficial result in terms of clinical or bacteriologic cure rates.
Rifampin is an antibiotic of great interest in the face of rising incidence, morbidity, and mortality of S aureus infections. There is a strong theoretic foundation as to why rifampin may provide important clinical advantages. Specifically, rifampin has bactericidal activity, concentrates well intracellularly,38 and penetrates biofilms, killing S aureus in sessile and planktonic (log) growth phases.18,32,47,72 To examine the clinical benefit of rifampin, we systematically identified data obtained from in vitro studies, animal models, and human trials that examined the efficacy of adjunctive use of rifampin for the treatment of S aureus infections. We found that investigations using in vitro methods tested rifampin combined with many antibiotic classes. In addition, we found that methods were heterogeneous, although the time-kill or checkerboard dilution assays were most commonly used. Findings among method types (time-kill, checkerboard, and serum bactericidal activity) often correlated poorly, an observation previously noted.30,37,45,46,48,73,74 Methods often differed at other levels such as inoculum used, outcome studied, and experiment duration.26- 30 Formal statistical findings were frequently unreported.21- 23 Therefore, it seems that for in vitro investigations results are heavily method dependent. This raises a serious question as to whether in vitro models of the efficacy of combination antibiotic therapy with rifampin against S aureus have relevance in the treatment of clinical infections.
Likewise, studies18,36,37,47,48 that examined both in vitro and in vivo effects of rifampin commonly had contradictory results or were poorly correlated. One group examining MRSA found vancomycin-rifampin antagonism by checkerboard assay and then in a subsequent animal endocarditis model found that rifampin-treated animals had decreased bacterial burden on valves and higher cure rates. The investigators then downplayed concerns that in vitro antagonism might predict similar in vivo interactions.48 More recent studies have not tied in vivo to in vitro investigations, perhaps acknowledging that discrepant findings are common. Until there are clear data as to which in vitro models have the most relevance for specific types of clinical infections (and such studies are probably challenging to perform), the role may be little for in vitro models of adjunctive rifampin therapy for S aureus other than to ensure that the study strain is rifampin susceptible.
The animal models reviewed included the following 5 principal types of infection: peritonitis, endocarditis, osteomyelitis, bacteremia, and device- or hardware-related infections. Most investigations examined rifampin used with fluoroquinolones, glycopeptides, and β-lactam antibiotics. Dual therapy was significantly better than monotherapy in some investigations in terms of outcomes such as bone sterilization, bacterial counts, and cure rates, regardless of the disease or animal model used. More specifically, the fluoroquinolone-rifampin combination was generally efficacious in peritonitis and device-related infections, with inconsistent results (indifference or benefit) in the osteomyelitis and endocarditis models. On the other hand, the glycopeptides were not more effective when combined with rifampin in osteomyelitis and endocarditis studies but generally resulted in better microbiological and clinical outcomes in foreign body–related (abscess or prosthetic device) infections.
Few human trials to date have directly compared outcomes with and without rifampin. Among the trials we identified, the median Jadad score was 1, and the range was 0 to 5, suggesting that the quality of published trials varies considerably. Cure rates were often higher in the adjunctive rifampin arms, but studies were typically underpowered to detect differences between groups. However, no study indicated a trend toward worse outcomes with adjunctive rifampin therapy.
Because of the limitations noted, it is challenging to draw conclusions from studies of rifampin in terms of its role as adjunctive therapy in infected patients. Nevertheless, our review identified clinical scenarios in which rifampin therapy seems promising. For example, rifampin seems beneficial in the treatment of prosthetic device infections and bone infections in human studies and animal models. In other disease states, data are less promising or are not well explored in human investigations. Further clinical studies may choose to build on promising in vivo data or clinical studies noted in our review.
Although our findings demonstrate that adjunctive rifampin use is not strongly supported with clinical or high-quality relevant animal or basic clinical data, there are 2 important observations worth noting. First, rifampin use does not seem antagonistic to other antibiotics in human studies. Second, although rifampin seems to be well tolerated in most patients with S aureus infections, some degree of intolerance occurs. In the study by Zimmerli et al,65 3 of 18 patients stopped rifampin therapy temporarily because of nausea, although rifampin was successfully reintroduced at a lower dosage. Two subjects discontinued the study because of exanthems, although nausea prompted discontinuation in a subject treated with monotherapy. Subjectively, we conclude that rifampin therapy may be reasonable in infections in which cure rates are not high, assuming patients are at low risk for toxic effects from rifampin or significant drug-drug interactions (eg, with anticoagulants and immunosuppressive medications). In cases in which rifampin treatment may compromise patient safety, the use of this medication is questionable given that the benefit of rifampin remains poorly defined.
In summary, we found that investigations of rifampin adjunctive therapy for S aureus infection are plagued by numerous limitations. There are situations in which adjunctive rifampin therapy seems promising, but none in which benefit is definitively established. We also found that in vitro models seem to contribute little to our understanding of the role of rifampin in vivo given that results are heavily method dependent. Adequately powered clinical studies need to be performed to assess outcomes with or without rifampin in the clinical scenarios in which poor outcomes are common. These include osteomyelitis, hardware-associated infections, and perhaps infections caused by MRSA strains. Given the rising global incidence of MRSA infections, there is an urgent need to better define the role of rifampin for the treatment of clinical S aureus infections.
Correspondence: Loren G. Miller, MD, MPH, Division of Infectious Diseases, Harbor-UCLA Medical Center, 1000 W Carson St, Bin 466, Torrance, CA 90509 (firstname.lastname@example.org).
Submitted for Publication: March 26, 2007; final revision received September 6, 2007; accepted September 7, 2007.
Author Contributions:Study concept and design: Perlroth and Miller. Acquisition of data: Perlroth, Kuo, Tan, and Miller. Analysis and interpretation of data: Perlroth, Kuo, Tan, Bayer, and Miller. Drafting of the manuscript: Perlroth, Kuo, Tan, Bayer, and Miller. Critical revision of the manuscript for important intellectual content: Perlroth, Kuo, Bayer, and Miller. Statistical analysis: Miller. Obtained funding: Miller. Administrative, technical, and material support: Perlroth, Tan, and Miller. Study supervision: Miller.
Financial Disclosure: None reported.
Funding/Support: This study was supported in part by grant R01 CCR923419 from the Centers for Disease Control and Prevention (Dr Miller).
Additional Contributions: James Steckelberg, MD, Donald Levine, MD, and Paul Holtom, MD, assisted with identifying abstracts and articles that could have been missed by our systematic review. Amy J. Chatfield, MLS, and Penny Coppernoll-Blach, MLS, assisted with the EMBASE database.