¶ Preferred empiric agents:
¶ Which empiric antimicrobial agents are preferred for osteomyelitis?
¶ Executive Summary:
Clinical Review (insufficient quality of evidence to enable a Clear Recommendation):
Based on observational and randomized controlled studies, aerobic gram-positive cocci, primarily S. aureus, have been the organisms most frequently isolated from culture in patients with osteomyelitis, including DFO. Enterobacterales have been the predominant group of gram-negative pathogens, with E. coli the most common. Thus, when treating osteomyelitis, it is reasonable to empirically cover gram-positive cocci, primarily Staphylococcus spp., and gram-negative bacilli if therapy cannot be delayed until culture availability (Table 2). For DFO, many physicians add anaerobic activity; however, data are not available to determine the benefit or harm of this approach. Pseudomonal activity is generally not necessary in treating osteomyelitis unless patients have been exposed to multiple prior courses of antibiotics, the wound is gangrenous, the organism has been previously cultured, the patient underwent a recent (e.g., < 3 months) surgical procedure in a healthcare setting, or the patient has a specific site of infection particularly prone to P. aeruginosa (e.g., malignant otitis externa).
For early, late, and hematogenous PJI, S. aureus and coagulase-negative Staphylococcus have been the most commonly isolated organisms. Gram-negative bacilli, most commonly Enterobacterales, have also been regularly isolated. Thus, reasonable empiric therapy for PJI of all stages generally includes coverage for gram-positive cocci and Enterobacterales. Antibiotic regimens to treat early (< 3 months since procedure), but not later, PJI may include coverage for P. aeruginosa, although some authors feel this is not routinely necessary depending on local microbiology. Anaerobes, such as Peptostreptococcus and C. acnes, are isolated infrequently. C. acnes is more often isolated in shoulder PJI compared to other joints, and thus would warrant empiric coverage for shoulder PJI; however, this is usually accomplished with anti-staphylococcal coverage. In all cases, local susceptibility profiles inform empiric therapy.
¶ Overall Summary:
¶ Osteomyelitis without Implant
Observational and randomized controlled studies have concluded that aerobic gram-positive cocci, primarily S. aureus, are the most frequently cultured organisms in patients with osteomyelitis.(Mylona, Samarkos et al. 2009, Bhavan, Marschall et al. 2010, Marschall, Bhavan et al. 2011, Kim, Song et al. 2012, Bernard, Dinh et al. 2015, Agarwal, Wo et al. 2016, Lopes Floro, Munckhof et al. 2018, Li, Rombach et al. 2019, Park, Kim et al. 2019) The rate of MRSA strains has varied by study, ranging from 0%(Mylona, Samarkos et al. 2009) to 46%,(Marschall, Bhavan et al. 2011) depending on geography. Other common aerobic gram-positive cocci isolates included Streptococcus spp. and Enterococcus spp. Gram-negative bacilli were identified from cultures approximately a quarter of the time, varying by study. The Enterobacterales were the most predominant group of gram-negative pathogens, of which E. coli has been the most frequently identified species. P. aeruginosa was isolated in cultures at rates of 10% or less in most studies. For pyogenic vertebral osteomyelitis, cultures were typically monomicrobial (86%), with blood cultures positive a little over half the time (58%).(Mylona, Samarkos et al. 2009) Microbiologic etiology may be influenced by patient specific elements such as prior environmental or community exposures, certain risk factors such as intravenous drug use,(Lopes Floro, Munckhof et al. 2018) recent healthcare exposure, or recent antibiotic treatment. Known MRSA colonization is the largest individual risk factor for MRSA infection.(Butler-Laporte, Cheng et al. 2018)
In an observational cohort study of 358 patients with hematogenous vertebral osteomyelitis conducted in five tertiary care hospitals in the Republic of Korea,(Park, Kim et al. 2019) the most frequently isolated organisms were methicillin-susceptible S. aureus (MSSA) (33.5%), followed by MRSA (24.9%) and Enterobacterales (19.3%). P. aeruginosa was isolated in only five specimens (1.4%). Moreover, the authors found differences in the proportion of pathogens isolated between community-acquired and healthcare-associated hematogenous vertebral osteomyelitis. MRSA was more frequent in healthcare-associated hematogenous vertebral osteomyelitis (43.6% vs. 13.8%; p < 0.001), whereas MSSA and Streptococcus spp. were more commonly found in community-acquired hematogenous vertebral osteomyelitis (44% vs. 13.8%; p < 0.001, 16% vs. 4.5%; p = 0.001 respectively).
Thus, although RCTs assessing empiric antibiotic choice are not available, based on observational data, selecting an empiric antibiotic therapy with coverage of S. aureus (including MRSA), Streptococcus spp., and Enterobacterales is reasonable for osteomyelitis in the absence of an implant. A reasonable regimen is a third-generation cephalosporin lacking pseudomonal coverage, such as ceftriaxone, with or without addition of vancomycin for MRSA coverage. Where MRSA is uncommon, some experts prefer to replace vancomycin with a b lactam with anti-staphylococcal activity (e.g., oxacillin, flucoxacillin). In those with cephalosporin allergy, TMP-SMX, or a fluoroquinolone, such as levofloxacin or ciprofloxacin, could be used as alternatives, provided that local antibiogram data are favorable. In cases where there is high suspicion of more resistant pathogens, such as extended spectrum β-lactamase (ESBL)-producing gram-negative bacilli, or P. aeruginosa, using a carbapenem or cefepime may be reasonable. Consideration of patient specific factors, such as comorbidities, prior healthcare exposure including procedures, known colonization with antibiotic-resistant organisms, severity of illness including sepsis or septic shock, and local epidemiology and resistance patterns is important when selecting an empiric regimen.
¶ Diabetic Foot Osteomyelitis
Overall, S. aureus has been the most frequently isolated organism from bone biopsy results in patients with DFO; MRSA has varied in isolation between 0%-20% of cultures.(Senneville, Melliez et al. 2006, Aragon-Sanchez, Cabrera-Galvan et al. 2008, Senneville, Lombart et al. 2008, Senneville, Morant et al. 2009, Elamurugan, Jagdish et al. 2011, Lesens, Desbiez et al. 2011, Lazaro-Martinez, Aragon-Sanchez et al. 2014, Tone, Nguyen et al. 2015, Gariani, Pham et al. 2020) Coagulase-negative Staphylococcus has been the next most frequently isolated organism, although it is difficult to determine if the organism is pathogenic when isolated, as it likely reflects surface colonization/specimen contamination. Other gram-positive organisms, such as Streptococcus spp., Enterococcus spp., and Corynebacterium spp., have also been described from culture results. Isolation of organisms typically considered normal skin flora, such as coagulase-negative Staphylococcus or Corynebacterium spp., can be of unclear significance, although pairing histology findings with culture results can help delineate whether these organisms are pathogenic. Of aerobic gram-negative bacilli, Enterobacterales were reported in 12%-50% of cultures, with the most common species being E. coli, Klebsiella spp., and Proteus spp. Typically, when aerobic gram-negative bacilli were isolated in culture, the culture was polymicrobial. Rates of polymicrobial cultures ranged from 26%-85%. Obligate anaerobes, such as Peptococcus, Peptostreptococcus, Prevotella spp., Clostridium spp., or Bacteroides spp., were isolated in higher frequency in older studies,(Wheat, Allen et al. 1986, Bamberger, Daus et al. 1987, Lavery, Sariaya et al. 1995, Tan, Friedman et al. 1996) while more contemporary studies report lower rates (3%-12%).(Senneville, Melliez et al. 2006, Senneville, Morant et al. 2009, Lesens, Desbiez et al. 2015, Tone, Nguyen et al. 2015) The differences in rates of anaerobic isolation may be due to differences in sampling technique, transport time to the lab, and microbiology lab handling of the samples.(Charles, Uckay et al. 2015)
In a single center retrospective study over 10 years in Spain, patients with biopsy proven DFO who had gram-negative bacilli isolated in bone culture (n = 150) were compared to those with other organisms or sterile cultures (n=191).(Aragon-Sanchez, Lipsky et al. 2013) Overall, 58.3% (224/384) of bone specimens isolated gram-positive cocci, 40.6% (156/384) gram-negative bacilli, and 1.1% (4/384) fungi. The most frequent gram-negatives isolated overall were E. coli (21.2%), P. aeruginosa (15.4%), and Enterobacter cloacae (12.8%). Patients whose cultures isolated gram-negative organisms more frequently had fetid odor, necrosis, soft tissue infection accompanying osteomyelitis, and clinically severe infection compared to those without gram-negative organisms. By multivariate analysis, having a glycosylated hemoglobin <7% (OR 2; 95% CI, 1.1-3.5) and a wound caused by traumatic injury (OR 2; 95% CI, 1-3.9) were found to be the most significant predictors to isolate gram-negative bacilli from bone samples. Duration of the foot wound did not affect the likelihood of isolating a gram-negative pathogen.
Thirty-four bone samples from US patients who were hospitalized with moderate-to-severe DFI with a high suspicion of DFO were evaluated by 16S ribosomal ribonucleic acid (rRNA) gene sequencing.(van Asten, La Fontaine et al. 2016) S. aureus was the most common pathogen isolated, at 50% (13/26) by conventional culture technique and 86.9% (20/23) by sequencing methods. The distribution of other gram-positive organisms identified by 16S sequencing technique included: Streptococcus spp. 56.5%, (13/23), unknown Dermabacteriae 34.8% (8/23), and Corynebacterium spp. 78.3% (18/23). Corynebacterium spp. appeared to have a lower contribution to the total bacterial population compared to Staphylococcus spp., and its pathogenic role in DFO is not well described. Gram-negative pathogens that were sequenced included: Pseudomonas spp. 21.7% (5/23) and Enterobacter spp. 26.1% (6/23). Facultative anaerobes isolated by sequencing included Actinomyces 26.1% (6/23) and Helcococcus spp. 21.7% (5/23). Obligate anaerobes included Peptoniphilus 73.9% (17/23), Finegoldia 65.2% (15/23), Anaerococcus 52.2% (12/23), Clostridium 39.1% (9/23), Porphyromonas 30.4% (7/23), and Prevotella 21.7% (5/23). Of the three samples that did not sequence, Stenotrophomonas maltophilia, S. aureus, and Enterobacter cloacae were isolated by conventional culture methods. Compared to standard culture methods, 16S rRNA sequencing found significantly more anaerobic pathogens (86.9% vs. 23.1%, p = 0.001), more polymicrobial cultures (91% vs. 64% p = 0.02), and more gram-positive bacilli (78.3% vs. 3.8%, p < 0.001). The clinical significance of these anaerobes remains unclear.
Although aerobic gram-positive cocci remain the predominant pathogens in many studies, one may also take geographic location of the patient into consideration when determining empiric antibiotics. In a meta-analysis by Zenelaj et al., studies of DFI conducted in countries with a warm climate, such as desert or tropical, tended to have a relatively lower percentage of infections caused by Staphylococcus spp.(Zenelaj, Bouvet et al. 2014) Instead, gram-negative bacilli were isolated in higher frequency compared to rates published in European countries or the US.(Elamurugan, Jagdish et al. 2011, Parvez, Dutta et al. 2012, Widatalla, Mahadi et al. 2012)
In cases of less severe DFO that did not have associated complications (e.g., necrotizing soft tissue infections or peripheral artery disease), oral therapy with either ciprofloxacin, amoxicillin-clavulanic acid, or trimethoprim-sulfamethoxazole (TMP-SMX) has been used.(Lazaro-Martinez, Aragon-Sanchez et al. 2014) If IV therapy is needed initially, then monotherapy of ampicillin-sulbactam, ceftriaxone, or the fluoroquinolones for patients with b lactam allergies may be reasonable options. If broader gram-positive coverage (i.e., MRSA) is needed, addition of clindamycin (PO or IV), linezolid (PO or IV), doxycycline (PO or IV), or vancomycin (IV) may be also reasonable.
Although anaerobic pathogens are isolated in bone biopsy cultures, there are currently no available data regarding whether empiric anaerobic therapy affects outcomes, either with improved cure rates or potentially higher adverse event rates. Many of the monotherapy options we list have varying degrees of anaerobic coverage. Thus, routine addition of broad anaerobic coverage with drugs like metronidazole for empiric therapy may not be required. Addition of metronidazole is of particular concern for patients with underlying neuropathy, as prolonged therapy can result in drug-induced neuropathy.
¶ Osteomyelitis with Retained Implant (including PJI)
For PJI, the frequency in which the gram-positive, gram-negative, or anaerobic pathogens are isolated varies by the timing of onset of PJI from the placement of the prosthetic implant. Early PJI (definitions in the literature vary but range from 1-3 months post-op) is acquired due to contamination intraoperatively and is typically caused by virulent organisms. Delayed-onset PJI (definitions used range from 1-12 months post-op) are acquired during time of surgery but are caused by less virulent organisms where the infectious presentation may not present within the immediate postoperative period. Late onset PJI is usually caused by hematogenous route or direct inoculation from other infectious foci.
Overall, S. aureus has been the most frequent cause of PJIs, regardless of whether they are early, delayed, late onset, or hematogenous PJI, contributing to approximately one third of cases (range 9%(Crockarell, Hanssen et al. 1998)-62%(Berdal, Skramm et al. 2005)). Rates of MRSA PJI have been low, ranging from 0%(Holmberg, Thorhallsdottir et al. 2015) (Sweden) to 17%(Rao, Crossett et al. 2003) (US) and vary by country. Coagulase-negative Staphylococcus has also been isolated in high frequency, ranging from 7%(Geurts, Janssen et al. 2013) to 50%,(Rao, Crossett et al. 2003) with S. epidermidis being the most frequently identified species within this group.(Harris, El-Bouri et al. 2010) Other gram-positives, such as Streptococcus spp., Enterococcus spp., and Corynebacteirum spp., have been isolated in decreasing frequency. Of the aerobic gram-negative bacilli, Enterobacterales have been the most common, ranging from 3%(Rao, Crossett et al. 2003) to 33%.(Crockarell, Hanssen et al. 1998) P. aeruginosa (~<10%) and Acinetobacter spp. (<3%) were infrequently isolated. Obligate anaerobes, such as Peptostreptococci and Cutibacterium acnes, were also isolated infrequently. Cutibacterium acnes was more often isolated in shoulder PJI compared to other joints, likely due to the organism being common flora in the axilla.(Sperling, Kozak et al. 2001, Coste, Reig et al. 2004, Cuff, Virani et al. 2008, Piper, Jacobson et al. 2009, Dodson, Craig et al. 2010, Sabesan, Ho et al. 2011, Verhelst, Stuyck et al. 2011, Klatte, Junghans et al. 2013) Cultures may be polymicrobial up to 39% of the time and were more likely to be a cause of early PJI (<3 months) vs. late PJI.(Moran, Masters et al. 2007, Geurts, Janssen et al. 2013, Tornero and Soriano 2016)
In a multicenter, retrospective study of PJIs from Spain (n = 2,524), the four most common organisms found in early postop PJI (<1 month from procedure) were S. aureus (35.6%), S. epidermidis (15.5%), E. coli (15.4%), and P. aeruginosa (15.3%).(Benito, Franco et al. 2016) For chronic PJI (>1 month post procedure and symptoms persisting >3 weeks), S. epidermidis (33.2%), S. aureus (20%), coagulase-negative Staphylococcus not identified to species level (16.7%), and C. acnes (5.2%) were most commonly isolated. In acute hematogenous infections (symptoms <3 weeks after uneventful procedure), S. aureus (39.2%), E. coli (12.5%), S. agalactiae (10.9%), and viridans group streptococci (4.5%) were the most common. The proportion of PJIs caused by multidrug resistant bacteria increased from 9.3% in 2003-2004 to 15.8% in 2011-2012 (p = 0.008). The increase was primarily due to multidrug resistant gram-negative bacilli (5.3% to 8.2%, p = 0.032) during the time period, rather than MRSA (4.7% to 7.6%, p = 0.2).
In a retrospective study of 112 patients with elbow, ankle, and shoulder PJI in the UK, gam-negative bacilli were more frequently isolated in early PJI (<3 months from prosthesis placement) and late chronic PJI (>12 months from prosthesis placement).(Moran, Masters et al. 2007) Gram-negative bacilli, including Enterobacterales, Pseudomonas, and Acinetobacter spp., as well as anaerobes were more likely to be isolated in early PJI. However, after three months, the frequency of both decreased. No gram-negative bacilli or anaerobes were isolated between 3-12 months after prosthetic joint placement. Enterobacterales and anaerobes were isolated in 4.2% (1/24) and 8.3% (2/24) of cases occurring >12 months after surgery. The rate of polymicrobial samples also declined, with the highest rate of polymicrobial samples within the first three months of implantation (47%), compared to 9.1% and 20.8% at 3-12 months and >12 months after implantation, respectively.
In a recent observational study from Australia, among 607 patients with prosthetic joint infection, the microbiology differed among patients with early PJI vs. other types.(Davis, Metcalf et al. 2022) S. aureus was the most common pathogen overall, but patients with early PJI had twice the frequency of Gram negative bacterial infections, and including P. aeruginosa as compared to later PJI.
Fungal prosthetic joint infections occur infrequently (1%) compared to bacterial causes.(Azzam, Parvizi et al. 2009) Patients with fungal PJI have different risk factors compared to those with bacterial causes including immunosuppression, overuse of antibacterials, presence of indwelling catheters, multiple revision surgeries, and complex reconstructions. Candida spp. were the etiology in the majority of fungal PJI. Aspergillus spp. and Rhodotorula spp. were rare causes.(Cutrona, Shah et al. 2002, Savini, Sozio et al. 2008, Azzam, Parvizi et al. 2009)
Thus, in order to cover the most likely pathogens for PJI, a reasonable empiric therapy for early PJI (<3 months) could include a combination of a third- or fourth-generation cephalosporin with antipseudomonal activity, or piperacillin-tazobactam, with or without IV vancomycin for MRSA coverage. However, some authors believe that anti-pseudomonal therapy is not routinely needed for early PJI depending on local microbiology of infection. Indeed, as some of the results mentioned above were derived from single center studies and some with small sample sizes, local epidemiology and resistance patterns will generally dictate the need or not for broader coverage for multidrug resistant gram-negative bacilli.
Alternatives to vancomycin can include daptomycin or linezolid,(Byren, Rege et al. 2012, Benkabouche, Racloz et al. 2019) while fluoroquinolones may be options for patients with significant penicillin or cephalosporin allergies. For delayed onset PJI, non-pseudomonal gram negative coverage combined with vancomycin IV with may be reasonable, although gram-negative bacilli are less frequently isolated in this group. Empiric therapy for late onset PJI (>12 months) may include vancomycin IV with a third-generation cephalosporin. As late onset PJI is unlikely to include Pseudomonas spp., empiric antipseudomonal therapy is not routinelynecessary in the absence of other risk factors. Empiric coverage of fungal etiologies in PJI is likely not needed unless the patient has had a prior PJI with isolation of a fungal pathogen.
¶ MRSA Coverage
¶ When should antimicrobial coverage against MRSA be included?
¶ Executive summary:
Based on observational and RCT data, rates of MRSA bone and joint infections vary by country. In areas with low MRSA prevalence, and for patients who are not known to be colonized by MRSA, it may be reasonable to hold MRSA coverage, and focus on MSSA coverage. In patients known to be colonized by MRSA (the largest individual risk factor for MRSA infection), or at centers with higher rates of MRSA among their S. aureus isolates, it is reasonable to initiate an anti-MRSA agent empirically while waiting for culture results, particularly for clinically unstable patients.
¶ Overall summary:
¶ Vertebral Osteomyelitis
Several studies reported that MRSA rates fell dramatically across multiple hospitals between the early 2000s and 2010-2016.(David, Daum et al. 2014, Kourtis, Hatfield et al. 2019, King, Castellucci-Garza et al. 2020) Despite decreases in MRSA incidence, MRSA infection remains more frequently observed in healthcare-associated settings than in community settings. Indeed, a meta-analysis found that there was a surge of community-acquired MRSA infections between the mid-1990s and 2005 in the US, but with substantially declining rates since then.(Dukic, Lauderdale et al. 2013)
As far as risk factors for MRSA in osteomyelitis, an observational cohort study performed in the Republic of Korea has shown that patients who are infected with MRSA hematogenous vertebral osteomyelitis were more likely to be older, have diabetes or a malignancy, and their infections were more frequently hospital-onset.(Park, Chong et al. 2013) Similarly, a more recent publication in 2019 by Park et al. concluded that MRSA was more frequent in healthcare-associated hematogenous vertebral osteomyelitis than in community-acquired hematogenous vertebral osteomyelitis (43.6% vs. 13.8%; p < 0.001).(Park, Kim et al. 2019) In these studies, healthcare-associated hematogenous vertebral osteomyelitis was defined as onset of symptoms after one month of hospitalization with no evidence of vertebral osteomyelitis at admission, hospital admission within six months before symptoms onset, or ambulatory diagnostic or therapeutic manipulations within six months before symptom onset. Furthermore, a retrospective study which included 586 patients with pyogenic vertebral osteomyelitis suggested MRSA may be more common in patients with chronic kidney disease regardless of being on dialysis (34.4% vs. 14.7%, p < 0.05).(Kim, Lydecker et al. 2020)
In one large study, colonization by MRSA in the nose or rectum far surpassed other factors in predicting MRSA as the etiologic pathogen for bacteremia.(Butler-Laporte, Cheng et al. 2018) These data suggest that colonization may similarly be a predominant predictor of MRSA of osteomyelitis as well.
¶ Diabetic Foot Osteomyelitis
Ashong et al. conducted a single-center retrospective review of 131 patients with an initial episode of probable or definite foot osteomyelitis.(Ashong, Raheem et al. 2017) Bone cultures were collected intraoperatively, percutaneously or with image-guided bone biopsies. Significantly more patients who received insulin therapy were in the MRSA group than non-MRSA group (68.8% vs. 61.6%, p = 0.02). MRSA, MSSA, and other staphylococcal species were isolated in 31 (23%), 27 (20%), and 14 (10.4%) bone cultures, respectively. The studies showed MRSA bone isolates were not associated with a greater risk of treatment failure. Of note, patients who had MRSA isolated in bone culture but did not receive antibiotic therapy targeting it were not at higher risk for treatment failure of DFO. However, vancomycin was part of the study's empiric treatment algorithm.
Another study performed by Aragon-Sanchez et al. compared the outcome of surgical treatment between DFO caused by MRSA vs. MSSA in Spain.(Aragon-Sanchez, Lazaro-Martinez et al. 2009) The number of surgeries performed in patients with DFO caused by MRSA was significantly greater. However, there were no significant differences in the final outcome of surgical treatment or mortality between the two groups. Similarly, in a French, multicenter, RCT of DFO, MRSA isolation in bone biopsy culture was not associated with patient-level outcomes.(Tone, Nguyen et al. 2015)
Overall, these results indicate considerable variation in MRSA rates geographically, although they also demonstrate declining rates in many parts of the world. Local rates of MRSA, combined with information regarding colonization status of the patient (via MRSA nasal or perineal swab, or prior culture results), are reasonable to guide the choice of empiric MRSA selection.(Butler-Laporte, Cheng et al. 2018)
¶ Pseudomonal coverage
¶ When should antimicrobial coverage against P. aeruginosa be included?
¶ Executive Summary
Observational studies demonstrate that P. aeruginosa is an uncommon cause of osteomyelitis outside of patients with specific risk factors. Thus, empiric therapy including antipseudomonal agents can be limited to patients with such risk factors. For example, Pseudomonas spp. are more prevalent in patients residing in subtropical and tropical climates than in temperate climates. Other risk factors include the presence of chronic wound infections with multiple prior antibiotic courses, gangrene, a history of positive culture with Pseudomonas spp. in the past, a recent (e.g., < 3 months) surgical procedure in a healthcare setting (e.g., early PJI), or specific sites of infection (e.g., malignant otitis externa).
¶ Overall Summary
¶ Osteomyelitis (Including DFO) without a Prosthetic Implant
Multiple studies have shown P. aeruginosa is a relatively uncommon isolate in patients with vertebral osteomyelitis or DFO.(Wheat, Allen et al. 1986, Bamberger, Daus et al. 1987, Lavery, Sariaya et al. 1995, Tan, Friedman et al. 1996, Senneville, Lombart et al. 2008, Bhavan, Marschall et al. 2010, Kim, Song et al. 2012, Lazaro-Martinez, Aragon-Sanchez et al. 2014, Bernard, Dinh et al. 2015, Lesens, Desbiez et al. 2015, Tone, Nguyen et al. 2015, Agarwal, Wo et al. 2016, Lopes Floro, Munckhof et al. 2018, Li, Rombach et al. 2019, Park, Kim et al. 2019, Gariani, Pham et al. 2020) However, some observational studies suggest gram-negative bacilli may be more prevalent in patients who reside in Asian countries with warm and humid climates, have chronic or trauma-related wound infection, have contiguous wounds complicated by gangrene,(Farkas, Lin et al. 2019, Kim, Lydecker et al. 2020) or are suffering peripheral vascular disease.
In studies of patients with DFI from India, Malaysia, Turkey, and Kuwait, gram-negative bacilli were frequently isolated (e.g., 76% from the study in India, 52% from Malaysia, 50% from Kuwait), and P. aeruginosa was the most common isolate, causing 22%, 20%, 25%, 30%, and 17% of infections, respectively.(Raja 2007, Bansal, Garg et al. 2008, Ramakant, Verma et al. 2011, Turhan, Mutluoglu et al. 2013)
In contrast, a retrospective study from Spain did not find a positive correlation between gram-negative organism isolation, warm climate, and duration of the foot infection.(Aragon-Sanchez, Lipsky et al. 2013) The study included 341 patients with DFO. Bone cultures were obtained intraoperatively. The study suggested gram-negative organisms were more frequently isolated from patients with wounds that developed after trauma (p = 0.045).
In a retrospective review of 103 combat veterans with a diagnosis of osteomyelitis, gram-negative organisms were isolated in 91% of cultures of bone and deep wounds taken during initial debridement from patients with combat-related wounds.(Yun, Branstetter et al. 2008) Twenty-four percent of the specimens grew P. aeruginosa. In another retrospective study conducted in a trauma center in Brazil, Cordeiro de Carvalho et al. reviewed the clinical and microbiological profiles of 101 patients with gram-negative osteomyelitis associated with open fracture.(Carvalho, Oliveira et al. 2012) P. aeruginosa was isolated in 19.8% of these bone cultures.
In a retrospective study of 302 patients, King et al. found that those with peripheral vascular disease had a higher incidence of foot and ankle osteomyelitis caused by gram-negative organisms (OR 2.2; 95% CI, 1.3-3.6; p = 0.004).(King, Castellucci-Garza et al. 2020) Pseudomonas spp. were the most frequently isolated gram-negative pathogens. The author concluded that longer wound duration and differences in wound environment led to overall incidence of gram-negative organisms isolated in patients with peripheral vascular disease. This conclusion is supported by observational studies reporting an increased risk of Pseudomonas spp. infection in diabetic foot ulcers complicated by gangrene.(Farkas, Lin et al. 2019, Kim, Lydecker et al. 2020)
There are also specific sites of infection that have been associated with a particularly high risk of P. aeruginosa as a cause of osteomyelitis, generally indicating empiric anti-pseudomonal therapy. For example, in multiple observational studies of patients with malignant otitis externa, P. aeruginosa has been a leading cause of infection.(Dibb 1991, Roland and Stroman 2002, Ninkovic, Dullo et al. 2008) Nail puncture wounds of the feet may also be associated with pseudomonal infection, although this observation is more anecdotal.(Fitzgerald and Cowan 1975, Miller and Semian 1975, Lavery, Harkless et al. 1994)
¶ Prosthetic Joint Infection
Based on epidemiological studies, isolation of P. aeruginosa is more common in early PJI (<3 months) or acute hematogenous PJI than in late PJI.(Ceriotti, Marino et al. 2017) In a large multicenter, retrospective study from Spain (n = 2,524), gram-negative bacilli were seldom isolated in chronic (>1 month post arthroplasty and symptoms >3 weeks in duration) or acute hematogenous PJI (symptoms <3 weeks after an uneventful procedure).(Benito, Franco et al. 2016) In contrast, Pseudomonas spp. were the fourth most common organism identified (15.3%) in those with early postoperative PJI <1 month post arthroplasty. Among those with pseudomonal infections, multidrug resistance increased significantly from 0.7% from 2003-2004 to 1.8% in 2011-2012, p = 0.044. Thus, contemporary, local resistance patterns will need to be evaluated to determine the most appropriate empiric choice for anti-pseudomonal coverage, if empiric coverage is needed. In some centers with low pseudomonal rates of infection, empiric coverage may not be routinely indicated for early PJI.
In a US cohort of hip or knee arthroplasty infections seen at a single center, 91 patients had PJI caused by Pseudomonas spp. between 1969 and 2012.(Shah, Osmon et al. 2016) A little over half (57%) had hip PJI, 22% had a history of diabetes, 22% had history of GI or GU surgery, and 16% were on immunosuppressive medications. Fifteen of 102 PJI episodes (15%) were early PJI (<3 months), 44% were late infections (3 months-2 years), and 41% were delayed onset PJI (>2 years after implantation procedure). Five patients had a history of renal disease and four patients had recurrent UTIs, two of which were due to P. aeruginosa. The authors hypothesize that acquisition of Pseudomonas spp. as a colonizer in their patients occurred during prior surgeries or procedures. Thus, while acquisition of colonization from Pseudomonas spp. from the initial arthroplasty may be the cause of late or delayed onset chronic infections, other surgeries or procedures, patient comorbidities, other infections, or antibiotic exposure should be considered when determining if empiric antipseudomonal antibiotic therapy is necessary in delayed or late onset PJI.
¶ Bone Penetration of Antibiotics
¶ Does “bone penetration” of an antimicrobial agent matter clinically, and should it be used to select therapy?
¶ Executive Summary
Bone penetration of antibiotic agents for the treatment of osteomyelitis is a frequently discussed yet poorly studied drug property (Table 3). There are numerous limitations that need to be considered when evaluating bone penetration studies. While it is intuitive that antibiotics cannot successfully treat an infection if they do not reach the site at a concentration sufficient to inhibit microbial growth, there are limited outcomes data for osteomyelitis to support this concept.
¶ Overall Summary
¶ Understanding the limitations
The main limitation of most bone penetration studies results from the measurement of total antibiotic concentration in tissue homogenates. This technique disrupts the various compartments within bone and mixes the organic and inorganic bone matrices. Additionally, measuring total concentration does not provide a measurement of unbound drug which is the theoretical concentration of the drug that is available to exert an effect.(Drusano 2004) Uncertainty around these points could create problems from both an antimicrobial and pathogen perspective. Because therapeutic agents do not distribute within each compartment in an identical manner, total homogenate concentration does not represent the available concentration at the actual site of infection.(Mouton, Theuretzbacher et al. 2008) Likewise, pathogens such as S. aureus can survive differentially in various compartments within the bone, which again makes interpreting studies that used whole tissue homogenate problematic, as the concentration is not measured at the site of the invading pathogen.(Ahmed, Meghji et al. 2001)
A second significant limitation of many bone penetration studies is the reporting of a single rather than multiple concentrations over time, making the measured concentration highly dependent on the sampling time. By failing to capture the dynamic nature between bone and serum concentration, any single value of a ratio of tissue vs. serum concentration is theoretically possible.(Mouton, Theuretzbacher et al. 2008) Lastly, taking this single value and comparing it to the minimum inhibitory concentration (MIC) of a pathogen to derive an ‘inhibitory quotient’ may lead to erroneous conclusions, as pharmacokinetics/pharmacodynamics (PK/PD) index values for antimicrobial agents are derived from serum rather than tissue concentrations that are obtained without regard to the time course of drug exposure.(Mouton, Theuretzbacher et al. 2008)
Additional limitations of bone penetration studies include variability in what the reported concentrations represent (i.e., mg/kg of total bone mass, organic mass, dry bone mass, total bone volume), small sample size compromising mostly of healthy patients with uninfected bone who receive a single dose of antibiotics prior to undergoing joint replacement, and the conduct of many studies before advances in sample preparation and bioanalytic methodology.(Thabit, Fatani et al. 2019)
Finally, an important point about interpreting bone penetration studies is that for an antimicrobial effect to occur, an absolute amount of drug must be present to inhibit microbial growth; relative ratios of drug in bone vs. blood do not necessarily translate to achieving necessary absolute levels in bone or not. For example, a drug with very low blood levels that achieves a high ratio of reported bone:blood concentrations (e.g., tigecycline) may still not achieve adequate absolute levels to inhibit microbial growth in bone. Conversely, a drug with very high blood levels but low ratio of bone:blood concentrations (e.g., ceftriaxone) may still achieve absolute levels in bone adequate to inhibit microbial growth in bone. All of these limitations should lead to the cautious interpretation of bone penetration literature.
¶ Bone Penetration Studies
Four systematic reviews on antibiotic bone penetration have been published spanning the period of 1978-2018.(Boselli and Allaouchiche 1999, Landersdorfer, Bulitta et al. 2009, Spellberg and Lipsky 2012, Thabit, Fatani et al. 2019) An additional literature search from 11/1/2018-4/30/2021 was conducted. Methodology for inclusion was similar to Landersdorfer and colleagues(Landersdorfer, Bulitta et al. 2009) in that only human studies were included, a minimum of five patients was required, bone:serum concentration ratios were calculated from reported mean concentrations if not otherwise calculated, and a bone density of 1 kg/L was utilized unless otherwise stated by the authors of the study. Table 3 summarizes the mean bone:serum concentration ratio for available agents which are reported as bone concentration (mg/kg) divided by serum concentration (mg/L). Where possible, the absolute levels in bone are listed as mg/g.
Overall, bone concentrations approach or exceed 50% of the serum concentration for the fluoroquinolones, azithromycin, tigecycline, clindamycin, linezolid, fusidic acid, and rifampin. The concentrations achieved in bone generally exceed the MICs of susceptible organisms except for tigecycline.(Spellberg and Lipsky 2012, Thabit, Fatani et al. 2019) However, significant variability exists among specific agents and within classes. For example, the range in bone:serum and bone concentrations for doxycycline across two studies was 0.02-0.7 and 0.1-2.6 mg/g, respectively.(Gnarpe, Dornbusch et al. 1976, Bystedt, A et al. 1978) The low end of the range would not exceed the MIC90 of S. aureus. Landersdorfer and colleagues offered disruption of circulation in fractures or slow equilibration between plasma and bone as possible explanations, although the precise reason for the discordance is not fully understood.(Landersdorfer, Bulitta et al. 2009) Slow equilibration time between bone and plasma may also account for variability observed with ciprofloxacin, although study results showing an increase in bone:serum ratio over time are conflicting.(Massias, Buffe et al. 1994, Leone, Sampol-Manos et al. 2002, Landersdorfer, Bulitta et al. 2009)
Cephalosporins, penicillins, and carbapenems generally achieve bone concentrations of 5%-25% of serum. While these agents have a low reported bone:serum ratio, serum levels are high and, as a result, the actual concentrations in bone are quite high, and intravenous agents among these classes are likely to achieve concentrations in excess of the MIC for most susceptible pathogens. Bone concentrations with oral b lactams are more variable and may be less likely to exceed the MIC of specific organisms. Oral flucloxacillin, for example, has demonstrated adequate concentrations to exceed the MIC of S. aureus in one study while failing to achieve measurable bone concentrations in a second study.(Kondell, Nord et al. 1982, Alvarez Ferrero, Vree et al. 1993) Likewise, oral administration of cefuroxime did not result in measured concentrations while intravenous delivery achieves acceptable bone concentrations.(Leigh, Marriner et al. 1982, Cain, Jones et al. 1987, Alvarez Ferrero, Vree et al. 1993, Ketterl and Wittwer 1993, Renneberg, Christensen et al. 1993, Lovering, Perez et al. 1997)
Vancomycin and daptomycin are generally thought to penetrate bone poorly with serum to bone concentrations of ~5%-30% and <10%, respectively. Bone concentrations for both agents would, however, be expected to exceed the MIC90 for S. aureus. Concentrations of vancomycin may not exceed the MICs for individual enterococcal strains.(Boselli and Allaouchiche 1999)
Table 3 also presents bone:serum concentration ratios for cortical and cancellous bone. Landersdorfer et al. reported a non-significant, numerically higher median bone:serum concentration in cancellous vs. cortical bone (bone:serum ratio 0.25 and 0.16 for cancellous and cortical bone, respectively; p = 0.06).(Landersdorfer, Bulitta et al. 2009) Again, there was significant variability in ratios within classes and agents. Among b lactam agents, cefepime, ceftriaxone, ertapenem, and piperacillin-tazobactam had higher penetration and concentrations in cancellous compared to cortical bone.(Boselli and Allaouchiche 1999, Landersdorfer, Bulitta et al. 2009)
A minority of agents have been investigated in the setting of ischemia. Compared to studies in non-ischemic bone, the reported penetration and concentrations into ischemic bone are generally decreased.(Kitzes-Coehn, Erde et al. 1990, Raymakers, Schaper et al. 1998, Raymakers, Houben et al. 2001) However, the effect of ischemia is not consistent across all agents. Lozano-Alonso et al. studied 46 patients who had received at least four doses of antibiotics for an infection in the setting of limb ischemia necessitating major amputation.(Lozano-Alonso, Linares-Palomino et al. 2016) Four measurements of transcutaneous pressure of oxygen were conducted ranging from the thigh, which had the best perfusion, to the distal foot, which had the worst perfusion, with a measurement in the chest being the control. A serum sample as well as bone biopsies at each of the three lower limb sites were obtained. Clindamycin, vancomycin, and meropenem showed decreased bone:serum ratios as ischemia worsened, while linezolid, levofloxacin, and ceftazidime did not show decreasing ratios. Except for clindamycin, all agents would have achieved bone concentrations in excess of the MIC for typical target pathogens.
Three studies across four agents were identified that utilized microdialysis techniques to obtain multiple bone concentrations over 24 hours that were then paired with serum concentrations.(Traunmuller, Schintler et al. 2010, Andreas, Zeitlinger et al. 2015, Bue, Tottrup et al. 2018) This allowed for a comparison of area under the curve (AUC) concentration from bone to that of plasma. As mentioned previously, inclusion of concentration over time is a more robust measure as it accounts for distribution between compartments and provides concentration at the site of infection.
For example, Traunmuller et al. measured daptomycin bone concentrations in 10 patients with DFI who had received multiple doses of 6 mg/kg.(Traunmuller, Schintler et al. 2010) The 24-hour fAUCbone/AUCplasma was 1.2 and equilibration between plasma and bone occurred within three hours of the infusion start. Cmax in the metatarsal bone was 4.7 mg/mL. A second study by Andreas et al. measured sternal bone concentrations in nine patients who received 6,000 mg cefazolin and 1,200 mg linezolid over a 24-hour period during which they underwent coronary artery bypass grafting with left mammary artery harvesting.(Andreas, Zeitlinger et al. 2015) Mean bone concentrations of cefazolin were 112 mg/ml and 159 mg/ml while linezolid were 10.9 mg/ml and 12.6 mg/ml on the left and right, respectively. Mean cefazolin AUCbone/AUCplasma was 0.7 on the left and 1.0 on the right while linezolid penetration was 0.8 and 1.0 on the left and right, respectively. Lastly, Bue et al. measured vancomycin concentrations over 24 hours in 10 male patients undergoing total knee revision whom had received 1,000 mg vancomycin as antibiotic prophylaxis.(Bue, Tottrup et al. 2018) The AUCbone/AUCplasma ratio for vancomycin was higher in the cancellous bone, 0.5, compared to cortical bone, 0.2.
A mean concentration of 2 mg/mL but not 4 mg/mL was able to be achieved in cortical bone. In cancellous bone it took < 1 hour to achieve a mean concentration of 4 mg/mL with a Cmax of 10.6 mg/mL noted. Clinical outcomes were not provided in any of the studies.
¶ Clinical Outcomes
¶ Comparative Studies
Four trials of patients with chronic osteomyelitis have compared clinical outcomes of treatment with agents that have high vs. low bone penetration.(Greenberg, Tice et al. 1987, Gentry and Rodriguez 1990, Mader, Cantrell et al. 1990, Gentry and Rodriguez-Gomez 1991) In these trials, an oral fluoroquinolone (ofloxacin or ciprofloxacin) was compared to either parenteral cephalosporins (cefazolin or ceftazidime) or antistaphylococcal penicillins with or without an aminoglycoside or clindamycin. Nonsignificant differences in clinical cures were observed in the fluoroquinolone group for all four trials (77%, 74%, 79%, 50% for fluoroquinolone vs. 79%, 86%, 83%, 68% for alternative therapy).(Greenberg, Tice et al. 1987, Gentry and Rodriguez 1990, Mader, Cantrell et al. 1990, Gentry and Rodriguez-Gomez 1991)
Three additional sub-studies of DFO patients from larger DFI studies have compared agents with high bone penetration to those with lower bone penetration.(Lipsky, Baker et al. 1997, Lipsky, Itani et al. 2004, Lauf, Ozsvar et al. 2014) Two of three studies again showed lower but non-significantly different clinical cure rates in the groups with high bone penetration. In contrast, Lauf et al. reported a very low clinical response rate to tigecycline, which has low bone penetration, compared to ertapenem (32% vs. 54%).(Lauf, Ozsvar et al. 2014) Based on the range of MICs and MIC90 data presented in the paper for organisms such as, E. faecalis and MSSA, and bone concentrations reported in a prior PK study, tigecycline concentrations in bone were too low to exert an antimicrobial effect.(Lauf, Ozsvar et al. 2014)
In a second open label RCT, Lipsky and colleagues compared linezolid (oral or parenteral) to an aminopenicillin and b lactamase combination (ampicillin-sulbactam or amoxicillin-clavulanate) in patients with DFO, with a cure rate of 61% vs. 69%, respectively, the difference of which was not statistically significant (Lipsky, Itani et al. 2004). Most patients were started on oral therapy and the predominant organisms were Staphylococcus spp. Given the low serum concentrations (Cmax 3.5-4.5 mg/L) and bone penetration, it is unlikely that bone concentrations of amoxicillin-clavulanate would have exceeded the MIC90 of S. aureus and coagulase negative staphylococci, yet clinical cure was high.(MacGregor and Graziani 1997)
High cure rates with predominantly amoxicillin-clavulanate in DFO were also shown in another recent RCT.(Gariani, Pham et al. 2020) Finally, a trial comparing parenteral followed by oral ofloxacin to a combination of aminopenicillin and b lactamase inhibitor (ampicillin-sulbactam followed by amoxicillin-clavulanate) demonstrated a higher rate of cure/improvement in the ofloxacin group (75% vs. 60%, respectively).(Lipsky, Baker et al. 1997) The number of patients in both groups is small with only five patients in the aminopenicillin group. More patients in the ofloxacin group underwent bone debridement, although overall there did not appear to be a difference in cure/improvement between patients who underwent bone debridement vs. those that did not (73% vs. 67%, respectively).
While there are limited data on oral administration of b lactam antibiotics other than amoxicillin-clavulanate, several case series in the 1970s and early 1980s were published on the use of cephalexin in chronic osteomyelitis. Cephalexin has low bone penetration (0.2) and concentrations in bone (1.3-3.1 mg/L), which would not be expected to exceed the MIC90 of most organisms.(Jalava, Saarimaa et al. 1977, Akimoto, Uda et al. 1990) Nonetheless, satisfactory clinical response was noted in these case series, ranging from 79%-85%.(Hernandez, Bravo et al. 1970, Evrard 1973, Endler and Hackel 1979, Hughes, Nixon et al. 1981)
In summary, there are insufficient trial data to determine whether measured bone concentrations are sufficient to predict antibiotic activity. There are certainly examples, such as tigecycline, where low concentrations may have contributed to excess failures. However, data for amoxicillin-clavulanate and cephalexin would seemingly argue against the notion that low predicted bone levels result in failure, as success was achieved in several studies despite low predicted bone concentrations. Ultimately, it is treatment success in clinical trials that should be prioritized for selecting antimicrobial regimens. It may be reasonable to consider bone concentrations in choosing antibiotics after first considering drugs with established efficacy in clinical studies.
¶ Adjunctive Rifampin
¶ Does adjunctive rifampin alter osteomyelitis treatment outcomes; for which organisms is rifampin therapy potentially useful, and if it is used, is there a preferred dosing?
¶ Executive Summary
Numerous observational studies and three small RCTs found that patients with osteomyelitis, with or without a retained implant, had improved clinical success rates, due to reduced relapse, when treated with adjunctive rifampin (rifampin monotherapy is never advisable due to concerns about emergence of resistance on therapy). However, other observational studies and one small RCT did not find a benefit of adjunctive rifampin. Meta-analysis of the four RCTs suggests a benefit of rifampin therapy (Figures 1-2). However, given the small size of these studies and the heterogeneity in results, patient populations, rifampin dosing, and background antibiotic therapy, these data remain hypothesis-generating, and a Clear Recommendation cannot be made for or against such therapy. A large RCT is necessary to clarify or disprove efficacy. In the meantime, it may be reasonable to consider adjunctive rifampin therapy for osteomyelitis caused by gram-positive cocci or non-fermenting gram-negative bacilli, with or without a retained implant, in individual patients based on risk:benefit assessment. Such assessment should include the uncertainty of the efficacy data balanced against potential drug interactions and adverse events of rifampin. If used, the dosing of rifampin has varied widely in studies. However, 450-600 mg per dose likely increases PD target attainment and adherence, and hence may be preferred, compared to 300 mg multiple daily dosing. Whether dose escalation to 900 mg once daily or 600 twice daily improves efficacy and/or worsens safety for treating osteomyelitis is unknown. To minimize emergence of resistance and treatment failure, it may be prudent to initiate rifampin only after bacteremia is cleared and surgical source control is achieved if it is necessary.
¶ Overall Summary
¶ Potential Role for Adjunctive Rifampin
The primary potential role for rifampin in the treatment of osteomyelitis, with or without foreign body/implants, is as adjunctive therapy with another antibiotic to reduce the risk of relapse/long term clinical failure. Rifampin should not be used as monotherapy due to its low barrier to resistance.
Relapse is a common cause of long-term clinical failure of osteomyelitis treatment. Even with appropriate treatment, osteomyelitis has a long-term relapse rate of 10%-30%.(Waldvogel and Papageorgiou 1980, Norden 1988, McHenry, Easley et al. 2002, Haidar, Der Boghossian et al. 2010, Wang, Fang et al. 2020) Observational studies have described even higher rates of failure, possibly up to 50% with long-term follow up, for infections caused by S. aureus treated with vancomycin or monotherapy fluoroquinolones, or for infections caused by non-fermenting gram-negative bacilli, such as P. aeruginosa.(MacGregor and Gentry 1985, Bach and Cocchetto 1987, Conrad, Williams et al. 1991, Zimmerli, Widmer et al. 1998, Tice, Hoaglund et al. 2003, Tice, Hoaglund et al. 2003, Dombrowski and Winston 2008, Spellberg and Lipsky 2012, Liang, Khair et al. 2014)
Although the precise pathophysiology of this high relapse rate is unknown, several lines of evidence suggest that slowly or non-replicating bacterial persister/small colony variants play a role.(Ciampolini and Harding 2000, McHenry, Easley et al. 2002) First, relapses after monomicrobial osteomyelitis are well described after multiple decades, with several reports occurring even 50 to 80 years after the original infection (often caused by S. aureus).(Gallie 1951, Korovessis, Fortis et al. 1991, Donati, Quadri et al. 1999, Al-Maiyah, Hemmady et al. 2007, Libraty, Patkar et al. 2012, Clerc, Zeller et al. 2020) It is difficult to conceive of bacteria actively replicating in bone for multiple decades with no resulting inflammatory response or signs or symptoms of infection. Such cases strongly suggest pathogenesis involving prolonged periods of a very slowly or non-replicating bacterial metabolic state in bone.
Second, with the exception of infections caused by S. aureus specifically treated with monotherapy quinolones,(MacGregor and Gentry 1985, Bach and Cocchetto 1987, Conrad, Williams et al. 1991, Zimmerli, Widmer et al. 1998, Spellberg and Lipsky 2012) relapsing strains have been reported to remain susceptible to the antibiotics with which the patient was originally treated.(Ciampolini and Harding 2000, McHenry, Easley et al. 2002) Failure to develop resistance after exposure to antibiotics is a hallmark of non-replicating persisters, as these bacteria do not express the biochemical targets of the antibiotic, and thus experience no selective pressure from the drugs.(Gollan, Grabe et al. 2019) Finally, studies of animal models and patients increasingly describe the role of small colony variants, which adopt a slowly or non-replicating phenotype, in S. aureus persistence during osteomyelitis.(Proctor, van Langevelde et al. 1995, Kahl 2014, Kahl, Becker et al. 2016, Tuchscherr, Kreis et al. 2016, Tuchscherr, Geraci et al. 2017, Yang, Wijenayaka et al. 2018, Gimza and Cassat 2021)
Rifampin is one of the few antibiotics that possesses the ability to reliably kill non-replicating persister bacteria.(Drusano, Sgambati et al. 2010, Keren, Mulcahy et al. 2012, Lechner, Patra et al. 2012, Conlon, Nakayasu et al. 2013, Tuchscherr, Kreis et al. 2016, Alexander, Guru et al. 2020) Thus, there is a potential, biologically plausible basis for the hypothesis that adjunctive rifampin could help reduce the relapse rate for osteomyelitis, even in the absence of prosthetic material.
¶ Preclinical Data Suggesting Adjunctive Rifampin May Be of Benefit
Consistent with this hypothesis, rifampin has been repeatedly shown to be more effective than a wide array of other antibiotics at eradicating bacteria from bone in preclinical models of infection, despite having less impressive activity than these other drugs during log phase, planktonic growth in vitro.(Norden, Fierer et al. 1983, Norden and Shaffer 1983, Dworkin, Modin et al. 1990, O'Reilly, Kunz et al. 1992, Jorgensen, Skovdal et al. 2016, Albac, Labrousse et al. 2020)
¶ Observational Clinical Data Regarding Adjunctive Rifampin
The preclinical data are mirrored by numerous observational or retrospective studies in patients. For example, among a cohort of patients who had had multiple relapses of osteomyelitis over 15 years or more, use of regimens that included adjunctive rifampin led to cessation of relapses in most patients.(Norden, Fierer et al. 1983) The only relapses observed with adjunctive rifampin treatment occurred in patients infected with gram-negative bacilli (primarily Enterobacterales) that were resistant to the non-rifampin agent. Similarly, in a retrospective review of 35 patients with vertebral osteomyelitis, relapses occurred in 0/15 patients treated with adjunctive rifampin vs. 5/20 patients not treated with rifampin (p = 0.048).(Livorsi, Daver et al. 2008) More recently, a large retrospective cohort study from the Veteran’s Health Administration found that patients with DFO treated with adjunctive rifampin had a significant reduction in long-term amputation and death compared to patients not treated with rifampin.(Wilson, Bessesen et al. 2019)
Multiple studies have also found that patients with PJI had reduced relapses when treated with adjunctive rifampin treatment vs. not.(El Helou, Berbari et al. 2010, Senneville, Joulie et al. 2011, Lora-Tamayo, Murillo et al. 2013, Ascione, Pagliano et al. 2014, Holmberg, Thorhallsdottir et al. 2015, Fiaux, Titecat et al. 2016, Lora-Tamayo, Senneville et al. 2017, Lesens, Ferry et al. 2018, Becker, Kreitmann et al. 2020, Munoz-Gallego, Viedma et al. 2020, Beldman, Lowik et al. 2021) In each of the three largest of these retrospective studies, totaling more than 1,500 patients, by multivariate analysis, patients treated with adjunctive rifampin had significant reductions in relapse/late failure compared to patients not treated with rifampin.(Lora-Tamayo, Murillo et al. 2013, Lora-Tamayo, Senneville et al. 2017, Beldman, Lowik et al. 2021)
However, the data are mixed, as other observational studies have not reported significant differences in relapse rates in patients treated with adjunctive rifampin.(Morata, Senneville et al. 2014, Akgun, Trampuz et al. 2017, Mahieu, Dubee et al. 2019, Achermann, Kusejko et al. 2020, Vilchez, Escudero-Sanchez et al. 2021, Davis, Nsengiyumva et al. 2022) One meta-analysis of 13 observational studies of adjunctive rifampin therapy for PJI found no clear benefit, and emphasized that the individual studies were highly subject to selection bias.(Aydin, Ergen et al. 2021) A more recent meta-analysis of rifampin therapy for the treatment of staphylococcal PJI included one RCT which did not show benefit (discussed below) and 63 observational studies.(Scheper, Gerritsen et al. 2021) They reported that adjunctive rifampin use was associated with a relatively small but significant benefit, with a pooled risk ratio for effectiveness of 1.10 (95% CI, 1.00–1.22). However, the analysis was subject to the same concerns about significant heterogeneity, and several types of bias. Thus, even meta-analyses of rifampin effect based on observational studies have conflicted in their conclusions.
¶ RCTs of Adjunctive Rifampin Therapy
Three small RCTs have demonstrated potential therapeutic benefit of adjunctive rifampin therapy for osteomyelitis with or without PJI.(Van der Auwera, Klastersky et al. 1985, Norden, Bryant et al. 1986, Zimmerli, Widmer et al. 1998) In the first, double-blinded, placebo-controlled trial by Van der Auwera et al., 101 patients with invasive S. aureus infection, of which 23 had biopsy-confirmed osteomyelitis, were randomized to treatment with oxacillin plus rifampin (600 mg twice daily) vs. oxacillin plus placebo.(Van der Auwera, Klastersky et al. 1985) Clinical success among the osteomyelitis patients occurred in 90% (9/10) treated with rifampin vs. 62% (8/13) treated with placebo (Figures 1-2). In the second, open-label trial by Norden et al., 18 patients with S. aureus osteomyelitis were randomized to nafcillin plus weight-based rifampin (300 twice daily, 300 thrice daily, or 600 twice daily for <50 kg, 50-74 kg, >74 kg, respectively) or nafcillin alone (no placebo).(Norden, Bryant et al. 1986) Treatment success rates were 80% (8/10) and 50% (4/8) for the rifampin vs. control group. Finally, in the third, double-blinded trial by Zimmerli et al., patients with S. aureus osteomyelitis in the setting of PJI were randomized to receive ciprofloxacin plus either rifampin (450 mg twice daily) or placebo.(Zimmerli, Widmer et al. 1998) In the per-protocol population, cure rates were 100% (12/12) for rifampin-treated vs. 58% (7/12) for placebo-treated patients (p < 0.02). Of particular importance, four of the five treatment failures in the placebo arm were caused by relapse associated with the development of resistance to fluoroquinolones while on therapy (i.e., ciprofloxacin monotherapy). In contrast, no relapses, and no resistance, were detected in the adjunctive rifampin group.(Zimmerli, Widmer et al. 1998)
However, a fourth, more recent, open-label RCT of rifampin for PJI by Karlsen et al. had twice the sample size of the prior PJI RCT and showed no benefit of adjunctive rifampin therapy.(Karlsen, Borgen et al. 2020) Forty-eight evaluable patients were randomized to adjunctive rifampin (dosed at 300 mg thrice daily) therapy or not. At a median of two years of follow-up, treatment success rates were 74% (17/23) in the rifampin arm and 72% (18/25) in the no rifampin arm. The Kaplan-Meier curve of time to failure did separate initially, but the difference waned as follow-up time elapsed.
Collectively across these four small RCTs, treatment success occurred in 84% (46/55) of patients treated with rifampin vs. 64% (37/58) not. By meta-analysis, the adjusted difference in success rate is 20% (95% CI, 4%-36%), p = 0.01, suggesting benefit (Figure 1). Subgroup analyses focusing just on osteomyelitis without PJI demonstrated treatment success rates of 85% (17/20) vs. 50% (12/21), with an adjusted difference in cure of 29% (95% CI, 3%-55%). For the PJI subgroup analysis, a random effects meta-analysis model was used due to significant heterogeneity across the two available RCTs. Composite treatment success rates were 83% (29/35) vs. 68% (25/37), for an adjusted difference in success rate of 21% (95% CI, -19% to +61%).
It must be emphasized that the dosing of rifampin varied across the four studies, the primary antibiotic varied across the two trials involving PJI (b lactam in one study, ciprofloxacin in the other), and that two studies were double-blinded whereas two were open-label. Hence variations in trials results could be due to small sample sizes, resulting in overlapping confidence intervals, or to drug selection and dosing variances, or other patient- or provider-assessment variations.
Cumulatively, these RCTs provide some support for the hypothesis that adjunctive rifampin therapy may be of benefit for both for osteomyelitis and PJI. However, all four RCTs were small, with different patient populations, antibiotic regimens, and designs, and one was discordant. Ultimately, a large RCT is needed to provide a hypothesis-confirming level of evidence either for or against rifampin benefit for osteomyelitis.
¶ Dosing Considerations
If rifampin is administered to patients with osteomyelitis, there are no clinical outcomes data to guide optimal dosing. However, at doses of approximately 450 mg, rifampin biliary clearance becomes saturated, such that greater than proportionate increases in serum levels occur above that individual dose.(Acocella 1983, Peloquin, Jaresko et al. 1997) Since the best predictor of rifampin efficacy is thought to be AUC24 divided by the MIC (AUC24/MIC),(Jayaram, Gaonkar et al. 2003, Hirai, Hagihara et al. 2016) increasing serum levels has the potential to increase the chance of reaching target attainment to optimize outcomes. Indeed, for S. aureus infections in mice, an AUC24/MIC ratio of >950 optimized outcomes.(Hirai, Hagihara et al. 2016) This target was predicted to be difficult to achieve in a modeling study of human dosing, underscoring the need to optimize serum levels to treat bone infections.(Marsot, Menard et al. 2020) Higher doses also result in higher peak levels, which may improve bone penetration. It may be advisable, therefore, to administer rifampin at individual doses above 300 mg to improve peak levels, AUC24/MIC target attainment, and antimicrobial effects. However, no clinical data are available to confirm this hypothesis. Nevertheless, once daily dosing is also easier for patients. For these reasons, 600 mg once per day may be preferred to 300 mg twice or thrice daily, although, again, clinical data are not available to validate this assertion.
Whether or not dose escalation to 900 or 1200 mg per day (in divided doses) might be of benefit or result in excess toxicity, compared to a 600 mg total daily dose, remains uncertain. In three RCTs of dose escalation rifampin for the treatment of tuberculosis, the impact of dosing on microbicidal effects varied, but no excess toxicity was seen with higher dosing. Specifically, two RCTs found that doses of 900 mg once per day or 35 mg/kg per day resulted in more rapid declines in bacterial density than 600 mg once per day or 10 mg/kg/day.(Boeree, Heinrich et al. 2017, Velasquez, Brooks et al. 2018) However, in none of the three trials did higher doses (900 mg per day, 1200 mg per day, or 35 mg/kg per day) result in higher rates of ultimate microbiological eradication, nor improve clinical cure.(Aarnoutse, Kibiki et al. 2017, Boeree, Heinrich et al. 2017, Velasquez, Brooks et al. 2018) Nor were higher doses associated with a higher rate of adverse events in any of the studies.
Whether these results translate to treatment of osteomyelitis is unclear. However, in a large retrospective study of PJI, 450 or 600 mg twice daily of adjunctive rifampin was not associated with superior cure rates by multivariate analysis compared to 600 mg once daily.(Beldman, Lowik et al. 2021) Furthermore, in contrast to the trials of patients with tuberculosis, in the retrospective studies of PJI, doses above 10 mg/kg/day were associated with progressively higher adverse event rates.(Becker, Kreitmann et al. 2020, Beldman, Lowik et al. 2021, Tonnelier, Bouras et al. 2021) Thus, a dose of 600 mg once per day may be a reasonable balance between safety and efficacy, although dosing to 8 to 10 mg/kg may be considered in heavier individuals.
When to initiate the rifampin and duration of rifampin therapy are also not certain. However, one observational study found that patients treated with adjunctive rifampin for >14 days resulted in reduced relapse rates compared <14 days.(Becker, Kreitmann et al. 2020) It is reasonable to administer rifampin during the total duration of antimicrobial therapy, provided the drug is tolerated. Such a strategy may limit patient confusion about changing regimens. It may also be reasonable to hold initiation of adjunctive rifampin until bacteremia is cleared (if present) and source control is achieved to reduce the potential for emergence of resistance to rifampin. Indeed, in one large observational study in which rifampin use was associated with increased treatment success in the treatment of PJI, initiation of rifampin within the first 5 days of surgical debridement was independently associated with treatment failure as compared to starting the rifampin later.(Beldman, Lowik et al. 2021)
Clinicians should be cautious of co-administering linezolid and rifampin due to a pharmacological interaction that lowers linezolid bioavailability and resulting blood levels via a variety of mechanisms.(Gandelman, Zhu et al. 2011, Okazaki, Tsuji et al. 2019) This interaction has been associated with increased failures in the treatment of PJI compared to treatment with either monotherapy linezolid or other drug combinations including rifampin.(Tornero, Morata et al. 2016) Similarly, rifampin appears to lower clindamycin blood levels when the latter drug is co-administered orally (but not intravenously),(Bernard, Kermarrec et al. 2015, Curis, Pestre et al. 2015, Zeller, Magreault et al. 2021) and also may lower fusidic acid levels.(Pushkin, Iglesias-Ussel et al. 2016) Nevertheless, two studies from a single center in Australia have reported high levels of treatment success with rifampin plus fusidic acid for PJI treated with DAIR.(Aboltins, Page et al. 2007, Peel, Buising et al. 2013) Similarly, rifampin may lower trimethoprim levels, but without compelling evidence that the combination decreases clinical efficacy.(Ribera, Pou et al. 2001, Harbarth, von Dach et al. 2015)
¶ Organism Considerations
Most data for adjunctive rifampin therapy come from the treatment of S. aureus infections. However, rifampin is broadly active, observational studies have described superior outcomes in patients with streptococcal osteomyelitis and PJI treated with adjunctive rifampin,(Fiaux, Titecat et al. 2016, Lora-Tamayo, Senneville et al. 2017) and there is a biologically plausible basis for its use as adjunctive therapy for other gram-positive organisms and for non-fermenting gram-negative bacilli (e.g., Pseudomonas and Acinetobacter), the latter of which have particularly high relapse rates.(Korvick, Peacock et al. 1992, Aydemir, Akduman et al. 2012, Durante-Mangoni, Signoriello et al. 2013) It is less clear that rifampin would be appropriate as an adjunctive agent for Enterobacterales, which tend to have substantially higher rifampin MICs, and given relapses of Enterobacterales despite adjunctive rifampin therapy in one small, uncontrolled study.(Norden, Fierer et al. 1983)
¶ Final Summary
There is a biologically plausible mechanism by which adjunctive rifampin may be of advantage for reducing relapse for osteomyelitis with or without foreign body implants: killing of slowly replicating or non-replicating/small colony variant strains in bone. Multiple preclinical models by many groups over several decades have found that rifampin is a more effective antimicrobial at sterilizing bone infections than any other antibiotic tested. Some observational clinical studies further support the potential for adjunctive rifampin therapy to reduce relapse rates. Finally, three small RCTs, each of which were individually under-powered and had different trial populations and designs, were concordant, suggesting that patients randomized to receive rifampin therapy had improved treatment success rates/reduced relapse rates. However, the data are mixed, as other retrospective studies and one small RCT of patients with PJIs are discordant. When these disparate trials were meta-analyzed, the results suggest potential benefit of rifampin, although not to the level of hypothesis-confirmation.
If adjunctive rifampin is to be used, its potential benefit should be balanced against its known toxicities and drug interactions when making a risk:benefit decision in individual patients. A large RCT is necessary and desired to provide more definitive, confirmatory evidence. In the meantime, it is reasonable to consider the use of rifampin for this purpose in individual patients, with careful risk:benefit considerations, but its use should not be considered standard of care.
Rifampin should not be used in patients with concomitant medications that would pose risks for serious drug interactions, and hence medication reconciliation/rationalization and assessment for drug interactions should always be conducted before initiation of rifampin therapy. Similarly, patients with active liver disease may not be appropriate candidates for adjunctive rifampin therapy. Given the remaining equipoise on risk:benefit for rifampin in this setting, involving the patient in shared decision-making, documentation of the reasons supporting its use, and its potentially favorable risk:benefit ratio in individual cases, is a prudent step.
¶ Long-acting glycopeptides
¶ What is the role of long-acting glycopeptide antibiotics?
¶ Executive Summary:
The two long-acting glycopeptides available on the market, dalbavancin and oritavancin, are not licensed for the treatment of osteomyelitis, but are licensed for the treatment of acute bacterial skin and soft tissue infection (ABSSI). One RCT of dalbavancin (n = 70 patients) vs. standard of care, which was largely vancomycin (n = 10), showed similar cure rates for non-vertebral osteomyelitis without prosthetic material, and a shorter length of hospital stay in the dalbavancin arm. No other randomized trial data are available for long-acting glycopeptides and osteomyelitis. Multiple, small, single-center, observational studies (all n < 50) have reported similar outcomes with both dalbavancin and oritavancin and comparator regimens. Few safety concerns were raised in these studies, and the glycopeptides were rarely stopped due to adverse events. There are currently no data to suggest that long-acting glycopeptides would have superiority over other regimens, including oral therapy options. Thus, based on available evidence, the most likely role for long-acting glycopeptides in osteomyelitis is for patients with non-vertebral osteomyelitis: a) who are unlikely/unable to take an oral regimen, or b) where an oral regimen is contraindicated (e.g., due to resistance patterns). There is minimal evidence of long-acting glycopeptide therapy for osteomyelitis in the presence of prosthetic material and for vertebral osteomyelitis, so caution is warranted in these settings.
¶ Overall Summary:
There are two long-acting glycopeptides currently available for routine use: dalbavancin and oritavancin.(Krsak, Morrisette et al. 2020) Both of these are licensed by the US Food and Drug Administration (FDA) and European Medicines Agency (EMA) for the treatment of ABSSSI, but not for other indications at present. Both agents maintain serum and tissue concentrations for prolonged periods of time (days to weeks) due to the half-lives of the drugs (10-14 days). PK data from a single study of dalbavancin suggested therapeutic levels of drug (against susceptible gram-positive pathogens) would be maintained for up to eight weeks in bone and plasma after two doses one week apart.(Dunne, Puttagunta et al. 2015)
Due to their convenient dosing and favorable PK, there is significant interest in the use of long-acting glycopeptides in osteomyelitis, with relevant studies summarized below.
¶ Randomized trial data
In a phase 2 RCT, Rappo et al. randomized patients (n = 80) with a clinical-radiological diagnosis of osteomyelitis at a single center in Ukraine to dalbavancin or standard of care at a 7:1 ratio, meaning 70 patients received dalbavancin and 10 standard of care.(Rappo, Puttagunta et al. 2019) All patients had baseline debridement and open biopsy, with histology supportive of chronic osteomyelitis in around 60% of patients. Patients with vertebral osteomyelitis or infections associated with prosthetic material were not included. More patients had a diagnosis of diabetes in the standard of care arm (50%) compared to the dalbavancin arm (14%). Standard of care was largely IV vancomycin followed by oral linezolid or levofloxacin. The major pathogen was MSSA (43/80, 54%), with the next most common pathogen being coagulase-negative staphylococci (16/80, 20%), followed by a variety of other pathogens. Importantly, in the dalbavancin arm, 23/70 cultured pathogens were either gram-negative, anaerobes, or mixed pathogens in which we would not expect any reliable activity of dalbavancin (given that its spectrum of activity is limited to gram-positive organisms). Only 3/11 patients with gram-negative infection in the dalbavancin group received adjunctive aztreonam. Notably, vancomycin was used in the standard of care rather than b-lactam therapy even though MSSA was the most common pathogen.
Multiple population endpoints were used in the trial, including a modified intention-to-treat (mITT) population consisting of those who had known or suspected gram-positive osteomyelitis, a clinically evaluable population (the subset that could be evaluated at day 42), and a microbiological mITT population that only included those that grew gram-positive pathogens from bone and/or blood. In all populations, dalbavancin was non-inferior to standard of care, with a 97% cure rate by day 42 vs 88% in the standard of care arm. Similar results were seen in other analyses. Length of stay was significantly reduced in those receiving dalbavancin (15.8 vs. 33.3 days).
In summary, dalbavancin appeared as safe and effective as a standard of care arm largely comprised of vancomycin as the primary treatment for gram-positive osteomyelitis without prosthetic material and excluding infections of the spine. Limitations include that the study was single-centered, small in size, and so vulnerable to baseline imbalances in patient characteristics, and the directed standard of care therapy could be considered suboptimal. Also, all patients received source control prior to enrollment, limiting generalizability outside of this setting.
We found no published RCTs of oritavancin for the treatment of osteomyelitis.
¶ Observational data
There are a significant number of studies evaluating real world experience of dalbavacin for osteomyelitis(Almangour, Fletcher et al. 2017, Bouza, Valerio et al. 2018, Almangour, Perry et al. 2019, Bork, Heil et al. 2019, Bryson-Cahn, Beieler et al. 2019, Morata, Cobo et al. 2019, Morrisette, Miller et al. 2019, Streifel, Sikka et al. 2019, Tobudic, Forstner et al. 2019, Wunsch, Krause et al. 2019, Almangour, Perry et al. 2020, Cain, Bremmer et al. 2022) (reviewed in (Almangour and Alhifany 2020)). The majority of these were small (all n < 50), single center experiences from the US or Europe of dalbavancin in which a proportion of patients with osteomyelitis were included. They generally described similar clinical success of dalbavancin to standard of care regimes. In two studies, lower clinical success was identified, although both of these were small (n = 7(Bryson-Cahn, Beieler et al. 2019) and n = 11(Bork, Heil et al. 2019)) and included some patients who did not receive the full treatment course. In summary, the limited real world published experience supports dalbavancin having similar efficacy to other agents in the treatment of osteomyelitis.
Data are even more limited with oritavancin. Although there is a growing literature in skin and soft tissue infection, there was initial concern about higher rates of progression to osteomyelitis in patients with skin and soft tissue infections when treated with oritavancin in early clinical trials (0.6%, 6/796 with oritavancin vs. 0.1%, 1/983 with vancomycin; OR for osteomyelitis 7.4, 95% CI, 0.9-61.6; p = 0.06), leading the FDA to issue a package insert warning about the risk of osteomyelitis when treating skin and soft tissue infection.(Corey, Kabler et al. 2014, Corey, Good et al. 2015) However, all diagnoses were made within nine days of initiation of therapy, suggesting these patients may have had pre-existing, occult osteomyelitis rather than development of osteomyelitis on therapy.
A few small studies have been published on real-world experience of oritavancin in osteomyelitis.(Chastain and Davis 2019, Morrisette, Miller et al. 2019, Brownell, Adamsick et al. 2020, Scoble, Reilly et al. 2020, Van Hise, Chundi et al. 2020) Similar to the dalbavancin data, most patients achieved clinical cure, and in the matched studies, at similar rates to those treated with standard of care. The largest study included 134 patients across 20 different centers in six US states, with an 88% clinical success at the end of dose evaluation and with a similar proportion achieving longer term cure.(Van Hise, Chundi et al. 2020)
¶ Adverse events
In all studies, adverse events were relatively mild and rarely required treatment discontinuation.
¶ Final Summary
The two long-acting glycopeptides on the market, oritavancin and dalbavancin, have a small amount of real-world evidence in the treatment of osteomyelitis. Nearly all studies were single-centered and retrospective, with a large number of varied comparator agents, definitions of disease, and definitions of cure. Adverse events with these agents were limited, and in the one RCT of dalbavancin, length of stay was much shorter in the dalbavancin group. There is no current evidence to suggest long-acting glycopeptides are more effective than other agents available for the treatment of osteomyelitis.
¶ Tables and Figures
¶ Table 2: Reasonable Empiric Antimicrobial Therapy Options with Published Data*
|
Types of Osteomyelitis |
Empiric IV Antibiotics† |
Alternative Empiric IV Antibiotics |
Empiric Oral AntibioticsY |
|
Osteomyelitis without a Retained Implant |
ceftriaxone ± vancomycin |
Alternative to b lactam: fluoroquinolone Alternative to vancomycin: linezolid, daptomycin, or clindamycin |
TMP-SMX or clindamycinj or linezolidj or fluoroquinolone or doxycycline¥ ± rifampin |
|
Diabetic Foot Osteomyelitis (DFO) |
ampicillin-sulbactam or amoxicillin-clavulanate or ceftriaxone‡ ± metronidazole ± vancomycin |
Alternative to b lactam: fluoroquinolone‡ ± metronidazole Alternative to vancomycin: linezolid, daptomycin, or clindamycin |
amoxicillin-clavulanate or TMP-SMX or clindamycinj or linezolidj or fluoroquinolone or doxycycline¥ ± metronidazole‡ ± rifampin |
|
Osteomyelitis with a Retained Implant (including PJI) |
|||
|
< 3 months since procedure (early) |
(anti-pseudomonal b lactam or ceftriaxone) + vancomycinx |
Alternative to b lactam: fluoroquinolone Alternative to vancomycin: linezolid, daptomycin, or clindamycin
|
fluoroquinolone ± rifampin or If gram-positive confirmed: TMP-SMX or clindamycinj or linezolidj or doxycycline¥ ± rifampin |
|
≥ 3 months after procedure (later onset) |
ceftriaxone + vancomycinx |
Alternative to b lactam: fluoroquinolone Alternative to vancomycin: linezolid, daptomycin, or clindamycin |
TMP-SMX or clindamycinj or linezolidj or fluoroquinolone or doxycycline¥ ± rifampin |
|
* This table addresses reasonable therapies with published data to be administered in the absence of available Gram stain, culture, histopathology, or other guiding information that enable targeted therapy. Biopsies should be obtained for such information prior to initiation of therapy when the risk:benefit ratio is favorable, see question 3 for a thorough discussion of initiation of empiric therapy vs. waiting for biopsy information to target therapy. In all cases, antibiotic selection should be adjusted based on local sensitivities for likely target pathogens. This table is not meant to indicate that other therapeutic options cannot be considered for specific patients based on clinical circumstances. †Add empiric anti-MRSA coverage (e.g., vancomycin) and/or replace ceftriaxone with an anti-pseudomonal b lactam (e.g., cefepime, piperacillin-tazobactam, etc.) if specific risk factors for MRSA (e.g., colonization, prior MRSA infection, healthcare exposure with endemic MRSA) and/or P. aeruginosa (exposed to prior courses of antibiotics, prior cultures with P. aeruginosa, gangrenous wounds, recent surgical procedures, specific sites of infection such as malignant otitis externa) are present, respectively, see question 4, sections b and c. When such risk factors are present, the authors unanimously prefer the use of non-carbapenem anti-pseudomonal options for stewardship reasons, unless there is a specific concern for ESBL pathogens. Similarly anti-anaerobic coverage is not routinely needed, but if the wound is gangrenous or there is specific concern for anaerobic infection, metronidazole may be added, or ceftriaxone replaced with ampicillin-sulbactam or amoxicillin-clavulanate. Finally, for patients in whom an MRSA active agent is deemed unnecessary, some authors prefer to add an anti-staphyloccocal b-lactam (e.g., oxacillin, cloxacillin, nafcillin, cefazolin) to ceftriaxone. ‡Anaerobic coverage is routinely added by many practitioners; however, data are not available to demonstrate whether it adds clinical benefit or not. xWhile many authors would initiate empiric anti-pseudomonal therapy, some authors do not believe that anti-pseudomonal coverage is routinely needed for early PJI infection, based on the frequency with which the organism is locally encountered. Most authors who would initiate rifampin prefer to wait until oral transition but some authors would consider initiating empiric IV rifampin. If rifampin use is being considered, it may be prudent to wait until bacteremia is cleared (if present) and surgical source control is achieved (if necessary), to reduce the risk of treatment failure.(Beldman, Lowik et al. 2021) See question 4, section c for a discussion of empiric pseudomonal therapy, and section e for a full discussion of the potential benefits:risks of adjunctive rifampin therapy. YSee question 5 for full discussion of oral therapy, including which agents, and timing of initiation. TMP-SMX = trimethoprim-sulfamethoxazole. Rifampin may be important to add to fluoroquinolones when treating S. aureus infections, and possibly when treating Pseudomonas or Acinetobacter infections, to reduce emergence of resistance. Other uses of rifampin are discussed in question 4, section e. jAs clindamycin and linezolid have no reliable gram-negative coverage, they should only be used when the clinician is confident that the infection is not likely caused by a gram-negative pathogen, or they should be administered with gram-negative coverage. ¥There are less published data for doxycycline, however it has been used with anecdotal success and was used in a minority of patients in the OVIVA trial,10 so it may be an alternative agent in individual patients. |
|||
¶ Table 3: Antibiotic Concentrations in Bone
| Antibiotic |
Time after Last Dose |
Mean Bone Concentration (mg/g) |
Overall Bone:Serum Concentration Ratio (range) |
Bone:Serum Concentration Ratio (range) |
|
|
Cortical |
Cancellous |
||||
|
Levofloxacin (von Baum, Bottcher et al. 2001, Rimmele, Boselli et al. 2004, Metallidis, Topsis et al. 2007) Ischemic bone(Lozano-Alonso, Linares-Palomino et al. 2016) |
0.7-2 h NR |
3-7.4 (von Baum, Bottcher et al. 2001, Rimmele, Boselli et al. 2004) 4.1-6.4 |
0.4-1 0.3-0.4 |
0.36-1 (von Baum, Bottcher et al. 2001, Rimmele, Boselli et al. 2004, Metallidis, Topsis et al. 2007) |
0.5-0.9 (von Baum, Bottcher et al. 2001, Rimmele, Boselli et al. 2004, Metallidis, Topsis et al. 2007) |
|
Ciprofloxacin (Fong, Ledbetter et al. 1986, Massias, Buffe et al. 1994, Leone, Sampol-Manos et al. 2002) Ischemic bone(Kitzes-Coehn, Erde et al. 1990) Osteomyelitis(Fong, Ledbetter et al. 1986) |
0.5-13 h 1 h 2-4.5 h |
1.1-2.9 (Fong, Ledbetter et al. 1986, Massias, Buffe et al. 1994) NA 1.4 |
0.3-1.2 0.2-0.3 0.4 |
|
|
| Ofloxacin (Meissner, Borner et al. 1990, Tolsdorff 1993, Tolsdorff 1993) |
0.5-12 |
0.3-1.1 |
0.09-1.0 |
|
|
| Moxifloxacin (Landersdorfer, Holzgrabe et al. 2004, Malincarne, Ghebregzabher et al. 2006, Metallidis, Topsis et al. 2007, Landersdorfer, Kinzig et al. 2009) |
1.5 h |
1.3-1.9 |
0.3-1.1 |
0.4-1.1 |
0.5-0.9 |
| Azithromycin (Malizia, Tejada et al. 1997, Malizia, Batoni et al. 2001) |
0.5-6.5 d |
1.6-1.9 |
2.5-6.3 |
|
|
|
Clindamycin (Nicholas, Meyers et al. 1975, Schurman, Johnson et al. 1975, Dornbusch, Carlstrom et al. 1977, Bystedt, A et al. 1978) Ischemic bone (Lozano-Alonso, Linares-Palomino et al. 2016) |
1-2 h NR |
0.6-3.8 0.8-1.2 |
0.2-0.5 0.2-0.3 |
|
|
|
Rifampicin (Sirot, Lopitaux et al. 1977, Sirot, Prive et al. 1983, Cluzel, Lopitaux et al. 1984, Roth 1984) Osteomyelitis (Roth 1984) |
2-14 h 3.5-4.5 h |
0.7-5 5 |
0.08-0.6 0.6 |
0.2 |
0.2-0.4 |
| Tigecycline (MacGregor and Graziani 1997, Rodvold, Gotfried et al. 2006) |
4-24 h (Rodvold, Gotfried et al. 2006) 4-24 h (MacGregor and Graziani 1997) |
0.08 0.4 |
0.4-2.0 NR |
|
|
| Doxycycline (Gnarpe, Dornbusch et al. 1976, Bystedt, A et al. 1978) |
3 h |
0.1-2.6 |
0.02-0.7 |
|
|
|
Vancomycin (Graziani, Lawson et al. 1988, Borner, Hahn et al. 1989, Massias, Dubois et al. 1992, Martin, Alaya et al. 1994, Kitzes-Cohen, Farin et al. 2000, Vuorisalo, Pokela et al. 2000)
Osteomyelitis (Graziani, Lawson et al. 1988, Bue, Tottrup et al. 2018)
Ischemic bone (Lozano-Alonso, Linares-Palomino et al. 2016) |
0.7-6 h
1-7 h
0-8 h (Bue, Tottrup et al. 2018)
NR |
1.1-10 (Graziani, Lawson et al. 1988, Martin, Alaya et al. 1994, Kitzes-Cohen, Farin et al. 2000) 3.6-8.4
NA
4.3-7.2 |
0.05-0.7
0.2-0.3 (Graziani, Lawson et al. 1988)
0.3-0.4 |
0.07 (Graziani, Lawson et al. 1988) 0.2(Graziani, Lawson et al. 1988)
0.2*(Bue, Tottrup et al. 2018) |
0.1 (Graziani, Lawson et al. 1988)
0.3 (Graziani, Lawson et al. 1988)
0.5*(Bue, Tottrup et al. 2018) |
| Teicoplanin(Wilson, Taylor et al. 1988, de Lalla, Novelli et al. 1993, Nehrer, Thalhammer et al. 1998, Wilson 2000) |
0.5-3.2 h 4-16 h |
1.3-7.1 7 |
0.2-0.9 0.5-0.6 |
|
|
|
Daptomycin (Traunmuller, Schintler et al. 2010, Montange, Berthier et al. 2014)
|
8 h 0-16 h 0-24 h |
3.3 NA NA |
0.09 1.1* 1.2* |
|
0.09 |
|
Linezolid (Lovering, Zhang et al. 2002, Rana, Butcher et al. 2002) Infected bone (Kutscha-Lissberg, Hebler et al. 2003) Osteoarticular tuberculosis (Li, Huang et al. 2019, Wen, Zhang et al. 2021) Ischemic bone (Lozano-Alonso, Linares-Palomino et al. 2016) |
0.5-16 h 2.5-24 0.9 h 1.7-24 h NR |
8.5-9
NA 4
0.6-3.9 10.5-21 |
0.4-0.5 0.8-1.0* 0.2 0.4-0.5
0.2-0.3 |
|
0.2 (Kutscha-Lissberg, Hebler et al. 2003) 0.8-1.0*
|
| Dalbavancin (Dunne, Puttagunta et al. 2015) |
0.5 & 14 d |
6.3 & 4.1 |
0.07 & 0.3 |
|
|
|
Fusidic acid (Hierholzer, Knothe et al. 1970) Osteomyelitis (Hierholzer, Knothe et al. 1966, Chater, Flynn et al. 1972) |
NR 1-13 h |
12-25 NA |
0.5-0.9 0.1-0.3 |
|
|
| Fosfomycin (Plaue, Muller et al. 1980, Wittmann 1980, Sirot, Lopitaux et al. 1983) |
0.5-7 h |
NA |
0.1-0.5 |
|
|
|
Trimethoprim- Sulfamethoxazole (Saux, Le Rebeller et al. 1982) |
1-1.5 h |
3.7/19 |
0.5/0.2 |
|
|
| Amoxicillin- clavulanic acid(Akimoto, Kaneko et al. 1982, Grimer, Karpinski et al. 1986, Adam, Heilmann et al. 1987, Weismeier, Adam et al. 1989) |
2 h 0.5-6 h
0.8-2.8 h
0.6 h 1 h |
NA NA
5.9-26 / 0.7-2.5
NA NA / 17.5-32.5 |
0.2-0.3 / NR 0.03-0.07 / 0.01-0.09 0.08-0.2 / 0.04-0.08 0.2 / 0.1 NR / 1.1-1.8 |
0.1-1.8 (clavulanic acid) |
0.1-1.1 (clavulanic acid) |
| Ampicillin-sulbactam (Wildfeuer, Mallwitz et al. 1997, Warnke, Wildfeuer et al. 1998, Dehne, Muhling et al. 2001) |
0.25-4 h |
12-20 / 5-7 |
0.1-0.7 / 0.2-0.7 |
|
|
| Piperacillin-tazobactam (Incavo, Ronchetti et al. 1994, Boselli, Breilh et al. 2002, Al-Nawas, Kinzig-Schippers et al. 2008) |
1 h 1.5 h 3 h |
21.3 / 3.8 15.1-18.9 / 2 9 / 1.2 |
0.2 / 0.2-0.3 0.2-0.3 / 0.3 0.2 / 0.1 |
0.2 / 0.2-0.3
|
0.2-0.3 / 0.3
|
| Flucloxacillin (Unsworth, Heatley et al. 1978, Kondell, Nord et al. 1982, Wilson, Taylor et al. 1988, Torkington, Davison et al. 2017) |
0.3-3 h NR |
2 NA |
0.1-1.2 0.05-0.08 |
|
0.6 |
| Cloxacillin (Kondell, Nord et al. 1982, Sirot, Lopitaux et al. 1982) |
1-3 h |
2 |
0.1-0.6 |
0.1 |
0.2 |
| Oxacillin (Fitzgerald, Kelly et al. 1978) |
1 h |
2.1 |
0.11 |
0.1 |
|
| Methicillin (Schurman, Johnson et al. 1975, Fitzgerald, Kelly et al. 1978, Sorensen, Colding et al. 1978) |
1-2 h |
3.1 |
0.04-0.2 |
0.2 |
|
| Ertapenem (Boselli, Breilh et al. 2007) |
1.6-23.8 h |
0.3-13.2 |
0.1-0.2 |
0.1 |
0.2 |
|
Meropenem Ischemic bone(Lozano-Alonso, Linares-Palomino et al. 2016) |
NR |
19.2-34 |
0.7-1.2 |
|
|
| Cefazolin(Polk, Hume et al. 1983, Williams, Gustilo et al. 1983, Cunha, Gossling et al. 1984, Yamada, Matsumoto et al. 2011) |
0.25-1.1 h 2.5-24 h (Andreas, Zeitlinger et al. 2015) |
4.7-32.3
NA |
0.06-0.4
0.7-1.0* |
|
|
| Cephalexin (Akimoto, Uda et al. 1990) |
1.5-2 h |
2.1 |
0.2 |
|
|
|
Cefuroxime (Leigh, Marriner et al. 1982, Cain, Jones et al. 1987, Ketterl and Wittwer 1993, Renneberg, Christensen et al. 1993, Lovering, Perez et al. 1997, Vuorisalo, Pokela et al. 2000, Dehne, Muhling et al. 2001) Osteomyelitis (Ketterl and Wittwer 1993) |
0.2-6.5 h 0.5-0.75 h 1 h |
2-36 NA 15-28 |
0.09-0.6 0.01-0.1 0.04-0.08 |
|
|
| Cefadroxil (Quintiliani 1982) |
2-5 h |
NA |
0.2-0.4 |
|
|
| Cefotaxime (Braga, Pozzato et al. 1982) |
0.75-4 h |
2.1-5.4 |
0.02-0.3 |
|
|
|
Ceftriaxone (Soudry, Sirot et al. 1986, Scaglione, De Martini et al. 1997, Lovering, Walsh et al. 2001) Osteomyelitis (Garazzino, Aprato et al. 2011) |
0.2-24 h 1.5-8 h |
2.2-20.9 9.6-30.8 |
0.07-0.2
|
0.05-0.08 0.08* 100% T>MIC for 24h |
0.1-0.2 0.2* 100% T>MIC for 24h |
|
Ceftazidime(Adam, Reichart et al. 1983) Ischemic bone (Raymakers, Schaper et al. 1998, Raymakers, Houben et al. 2001) Ischemic bone (Lozano-Alonso, Linares-Palomino et al. 2016) |
2 h 1-2 h NR |
20 3.1 2.6-3.7 |
0.5 0.04-0.08 0.1-0.2 |
|
|
| Cefepime (Breilh, Boselli et al. 2003) |
1-2 h |
35.6-52.5 |
|
0.5 |
0.8 |
| Tobramycin (Wilson, Taylor et al. 1988, Boselli and Allaouchiche 1999) |
0.3 14.3 |
NA NA |
0.1 0.09 |
|
|
| Gentamicin (Torkington, Davison et al. 2017) |
NR |
NA |
0.1 |
|
|
|
AUC = area under the curve serum level * = AUCbone/AUCplasma rather than serum |
|||||
¶ Figure 1: RCTs comparing rifampin success rates S. aureus
¶ Figure 2: RCTs comparing rifampin vs. no rifampin in OM and PJI
