Local Antibiotic Delivery Ceramic Bone Substitutes for the Treatment of Infected Bone Cavities and Bone Regeneration: A Systematic Review on What We Have Learned from Animal Models

Aims: the focus of this study is to evaluate if the combination of an antibiotic with a ceramic biomaterial is effective in treating osteomyelitis in an infected animal model and to define which model and protocol are best suited for in vivo experiments of local bone infection treatment. Methods: a systematic review was carried out based on PRISMA statement guidelines. A PubMed search was conducted to find original papers on animal models of bone infections using local antibiotic delivery systems with the characteristics of bone substitutes. Articles without a control group, differing from the experimental group only by the addition of antibiotics to the bone substitute, were excluded. Results: a total of 1185 records were retrieved, and after a three-step selection, 34 papers were included. Six manuscripts studied the effect of antibiotic-loaded biomaterials on bone infection prevention. Five articles studied infection in the presence of foreign bodies. In all but one, the combination of an antibiotic with bioceramic bone substitutes tended to prevent or cure bone infection while promoting biomaterial osteointegration. Conclusions: this systematic review shows that the combination of antibiotics with bioceramic bone substitutes may be appropriate to treat bone infection when applied locally. The variability of the animal models, time to develop an infection, antibiotic used, way of carrying and releasing antibiotics, type of ceramic material, and endpoints limits the conclusions on the ideal therapy, enhancing the need for consistent models and guidelines to develop an adequate combination of material and antimicrobial agent leading to an effective human application.


Introduction
The current protocols to treat chronic osteomyelitis consist of the intravenous and oral administration of drugs for long periods and surgical debridement of all devitalized bone fragments [1]. Adequate debridement may leave a bone defect ("dead space")

Materials and Methods
This systematic review was based on PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) statement guidelines [25]. Literature search: to identify all in vivo studies with original data on local antibiotic delivery ceramic bone substitutes to treat infected bone cavities, a PubMed search was performed using a combination of the following terms or equivalents: "ceramic bone substitute", "antibiotic", "osteomyelitis", and "animal" (Figure 1). Article selection: study selection was conducted in three steps. In step 1, two researchers screened titles and abstracts independently (N.A. and S.R.S.). In step 2, full-text articles were analyzed independently, and disagreements were discussed between reviewers. When a consensus was not reached, a third researcher was involved in the discussion (F.J.M.). In step 3, data were extracted and analyzed. The exclusion criteria were (1) articles not written in English, Portuguese, or Spanish; (2) article type: editorial, comment, guidelines, case report, abstract, review, or letter; (3) studies not involving animals; (4) studies not dealing with infected bone cavities; (5) studies not using antibiotics; (6) studies where the biomaterial was not ceramic; (7) absence of a control group that received the same biomaterial without antibiotic; (8) non-extractable data; and (9) data duplicated from another article. After the first selection, all cited references of the selected papers were cross-checked, and the screening procedure was repeated. The general results on implant effect were retrieved from individual papers, with data presented according to the causative agent of bone infection, type and composition of each biomaterial, type and amount of antibiotic, time between material implantation and analysis, and presence of concomitant conditions. The reported biomaterial effects on infection and bone remodeling were converted into graphic summary tables. The selected papers' quality was evaluated by the application of the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines for reporting animal research [26]. biomaterial, type and amount of antibiotic, time between material implantation and analysis, and presence of concomitant conditions. The reported biomaterial effects on infection and bone remodeling were converted into graphic summary tables. The selected papers' quality was evaluated by the application of the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines for reporting animal research [26]. Article selection process. * The truncation symbol (asterisk) was used to search for multiple variants of the word (singular/plural/conjugations, etc.) all at once, e.g., infected, infection, infectious, according to the PubMed search engine guide.

Systematic Review
Five hundred sixteen references were retrieved after a PubMed literature search. After applying the exclusion criteria, a total of 32 articles were included in this review [19,. Two additional papers [58,59] were added to the final literature review after backward citation tracking, making a total of 34 papers. Figure 1 summarizes the article selection process [25].

Systematic Review
Five hundred sixteen references were retrieved after a PubMed literature search. After applying the exclusion criteria, a total of 32 articles were included in this review [19,. Two additional papers [58,59] were added to the final literature review after backward citation tracking, making a total of 34 papers. Figure 1 summarizes the article selection process [25].

Article Quality
The ARRIVE score varied from 5 to 18.5, and it was possible to identify a trend of higher scores in more recent papers ( Figure 2). Five out of thirty-four articles received a negative appreciation [29,35,36,46,59].

Article Quality
The ARRIVE score varied from 5 to 18.5, and it was possible to identify a trend of higher scores in more recent papers ( Figure 2). Five out of thirty-four articles received a negative appreciation [29,35,36,46,59].

Article Quality
The ARRIVE score varied from 5 to 18.5, and it was possible to identify a trend of higher scores in more recent papers (Figure 2). Five out of thirty-four articles received a negative appreciation [29,35,36,46,59].

Radiological Evaluation
Most selected articles (37 out of 46 studies, representing 80.4%) reported using radiology to evaluate the appearance and evolution of osteomyelitis. The Norden score [60] was used to quantify the status of osteomyelitis in 16 series [29,30,33,35,[37][38][39]50,52,53,57], the Aktekin score was used in 2 series [43], and the Beenken score [27], Odekerken score [47], and Dahners score [58] were used in 1 series each; however, in general, there was a subjective appreciation of the radiological bone characteristics. In all but two papers [27,44], there was a radiographic improvement in osteomyelitis in the experimental group. In all the control groups, there was maintenance or increased bone destruction, suggesting osteomyelitis progression. In three series, there was no radiological difference between the experimental and control groups [27,44].

Histological Evaluation
Only seven studies (15.2%) did not undertake a histological evaluation [28,32,35,48,49,51]. The evolution of the histological characteristics of osteomyelitis and the integration of the ceramic material as well as the filling of the bone cavity by new woven bone, increase in collagen, the appearance of new vessels, and signs of inflammation were assessed. The Smeltzer score [61] was used to quantify the histologic pattern and to measure the changes in the osteomyelitis signs in 17 series [29,30,33,[37][38][39][40][41]44,50,52,53]. Except for three experiments [27,50], there were differences in the microscopic evaluations between the experimental and control groups, suggesting a favorable evolution after the implantation of the biomaterial with antibiotics.

Discussion
This is the first systematic review that summarizes the in vivo effect of adding antibiotics to ceramic bone substitutes to treat experimental osteomyelitis.

Quality of Selected Papers
There was a trend of higher ARRIVE scores in the more recent papers, reflecting the increased care used with animal experiments, namely regarding the number, welfare, and experiment design.

Infection Model
Different animal models were used in the included studies. This may influence the results as, among species, there are differences in bone regeneration and architecture, which poses problems when comparing the results from different studies [63,64]. The New Zealand White rabbit (NZWR) was the most frequently used animal. Rabbits are used in about one-third of all animal musculoskeletal studies [64] due to their relatively low cost, ease of handling, availability, and minimal phylogenetic development [65]. The main drawback is related to size, as rabbits do not allow for large implants [66]. Only two studies used larger animals: mongrel dogs [59] and Spanish goats [51]. Although other animals, such as non-human primates or sheep, should be used because they represent more reliable models, they may pose more ethical issues and limitations in terms of availability, housing, and handling [63,64].
Even for the same species, the location and size of bone defects are variable, leading to different results. The proximal tibia was chosen in most experiments. The resemblance to clinical practice (osteomyelitis occurs mainly around the knee) and easy access justify why it is the first choice. In most series, the infection protocol was based on Norden's model [67], and the time between inoculation and treatment was three or four weeks. In some series, the antibiotic-loaded biomaterial was implanted simultaneously with bacteria inoculation when the objective was to prevent, rather than cure, bone infection.
There are many infection models, and this may be confusing when comparing different series [68]. Variations between species, the bone defect size and location, the time to introduce the biomaterial, and the existence of a foreign body may explain some of the differences found between studies.

Osteomyelitis Agent
Staphylococcus aureus is the most common pathogen isolated in osteomyelitis, which is why it was the chosen agent to develop the experimental infection. Although the incidence of methicillin-resistant S. aureus (MRSA) bacteremia has decreased over the past decade [69], MRSA remains associated with poorer clinical outcomes compared with methicillin-sensitive S. aureus (MSSA) [70]. Most recent studies tend to use MRSA, reflecting the concern with this agent.

Antibiotics
Gentamicin [32,33,35,40,46,50,56,58] was the antibiotic most often used in earlier studies, but there was a clear shift to vancomycin in most recent research [29][30][31]34,39,41,43,49,52,57]. The use of other antibiotics was scarce, but two papers showed an attempt to develop new substances. In one study, antimicrobial peptides (AMP) were used, originally derived from the venom of the wild bee [48], which seemed capable of avoiding microbial resistance and bypassing the biofilm barrier. In two other studies, human lactoferrin1-11 was used [33,50], a broad-spectrum antibiotic with in vitro activity against both bacteria and fungi that has proven efficacy against MRSA after systemic administration in a mouse model of thigh infection [71]. Although biofilm formation, namely around metallic implants, is responsible for the difficulty in treating implant-related osteomyelitis [72,73], with rifampicin being the antibiotic of choice to overcome its effect [74], only in two series [55] was it used in association with vancomycin. In five series, the antibiotic was encapsulated in poly(lacticco-glycolic acid) (PLGA) microspheres [29,30,53], which has the advantage of changing the rate and speed of its degradation by adjusting the proportion of microspheres [75]. In one series [31], the antibiotic, together with sodium bicarbonate (NaHCO 3 ), was placed in PLGA shells. According to its authors, in the inflammation-induced acidic environment of osteomyelitis, the NaHCO 3 that was encapsulated in hollow PLGA microspheres rapidly reacted with acid to generate CO 2 bubbles, disrupting their PLGA shell and thereby promptly releasing the antibiotic. In one series [43], the antibiotic-loaded ceramic material was involved with a poly-l-lactic acid (PLLA) coating, which attempted to slow the antibiotic release [76]. Liposomes were also used as antibiotic carriers in one series [35] based on their beneficial characteristics of being antibiotic and antineoplastic carriers [77]. The dose of the antibiotic added to the biomaterial was very variable between series. In the case of vancomycin, the dose varied between 5% and 20% of the weight. The dose of gentamicin varied between 3.2% and 5% of the weight, but in older series, the value of the antibiotic concentration in the material was not extractable [32,35,46,50,56]. None of the analyzed articles presented results on the effect of different antibiotic concentrations on bone regeneration or considered the eventual toxic effect of antibiotics on osteoblasts under experimental conditions. Many in vitro studies, preliminary to the in vivo studies that were analyzed, presented results on the way the antibiotic was released. Only one article [57] presented the antibiotic release profile in the animal model with infection, an aspect that is crucial not only for the treatment of infection but also, above all, for preventing the emergence of bacterial resistance.

Bioactive Ceramics
Bioactive ceramics are osteoconductive materials (sometimes they may have osteoinductive properties) that are used to fill bone cavities. With time, these ceramics are replaced by new oven bone, behaving as excellent bone substitutes with many advantages over an autologous bone graft [78]. Calcium sulfate and hydroxyapatite were the first ceramics tested. Until 2010, all reviewed papers reported studies on the addition of antibiotics to ceramic material. Most recent papers associate other substances with the ceramic base to promote controlled antibiotic release or enhance new bone formation. Collagen was added in two series [54], benefitting from having a sponge-like elasticity, more easily fitting into bone defects, improving adsorption, and having a long-term release of substances while enhancing bone formation [79].

Evaluation Methods
The general observation of animals showed that the changes in behavior, weight, and fur characteristics, which appeared after osteomyelitis induction, returned to normal in the experimental groups during the first weeks. The number of fractures and deaths in the control groups was superior to that in the experimental groups. Although the evaluated parameters were different between studies, there was an increase in specificity in most recent papers.
The WBC count returned to normal after treatment in all experimental groups and was maintained above normal in the control groups. This was the most reliable laboratory test to reflect the positive evolution after treatment. Other blood tests were not useful due to their low ability to discriminate between the experimental and control groups.
The radiographic evaluation of infection was mainly subjective, qualifying the evolution of the infected cavity, periosteal new bone formation, sequestrum formation, and extent of the disease. In some studies, there was an attempt to quantify and score this evolution. The Norden score was the most-used classification system [60], but other classification systems, as proposed by Beenken [62], Aktekin [80], Odekerken [81], Mader and Wilson [82], and Lane and Sandhu [83], were also used, suggesting that different parameters were being evaluated and scored. These differences in interpretation were a drawback when trying to compare the results between different studies. Other image methodologies were scarcely used (micro-CT in three of the most recent reports [54][55][56] and 18 F-FDG PET imaging, pQCT imaging, and SEM in one article [45]).
A histological evaluation was used in most studies, using hematoxylin and eosin staining and applying the Smeltzer score [61] based on signs of intraosseous acute inflammation, intraosseous chronic inflammation, periosteal inflammation, and bone necrosis. This score was described to evaluate the development of a murine osteomyelitis model, and, at its origin, there was no introduction of a bone substitute in the infected cavity. The use of this score to evaluate bone behavior in the presence of a bioactive ceramic biomaterial may be inadequate, so new quantification systems are needed.
The microbiological analysis of bone samples was expressed as the presence or absence of bacteria, although some articles gave semi-quantitative results based on the amount of growth in the agar plate. The quantification of CFU on a specific volume of collected bone was made in 20 of the 44 series [27,28,32,33,36,37,42,44,46,47,[49][50][51]53,55]. However, it was not possible to conduct a meta-analysis of the quantification of CFU due to the heterogeneity (approximately 80%) in the counting methods and results. Nevertheless, there was an improvement in the experimental groups compared to the control groups.

Infection Treatment and Biomaterial Osteointegration
In all 34 articles, there was some form of an evaluation of the infection evolution. Most of them reported positive results in the experimental group when compared to the control group, suggesting the benefit of adding antibiotics to ceramic materials. The definitions of "good result" and "cure of infection" are variable. Some authors consider the absence of bacterial growth in bone samples [30,32,34,35,[37][38][39][40][41]43,[45][46][47]51,52,[54][55][56], while others value a difference in the CFU count between the experimental and control groups [28,33,36,44,49,50,53]. Of the 46 series, bone swabs or bacteria counts were used in 37 series to evaluate the progression or cure of the infection. Only in 15 experimental groups was there no identification of the bacteria in the bone samples. In the other 22 series, the rate of cure was not total, but it was considered positive if the rate of the cured animals in the experimental group was superior to that of the control group. Only in one article were the results between the experimental and control groups not different [27]. Egawa et al. [54] reported the absence of MSSA in the week 1 bone samples when vancomycin was used and in the week 2 bone samples when cefazolin was added to the bone substitute. In the week 4 bone samples, they reported a decrease in the CFU number in the control group, suggesting that the animal studied (Wistar rat) could cure the infection, even without antibiotics. This report raises the concern that other articles using a similar model may have achieved good results due to the disease's natural evolution rather than the treatment efficacy. Some authors report good results for the infection evolution, compared to the control groups when the histological or radiographic interpretation reveals a better score. For others, an osteomyelitis cure is considered when no evidence of bacteria is found in the bone samples, and a good result is achieved when the number of animals cured in the experimental group is proportionally higher than that in the control group. Other authors tried to quantify the number of CFUs in a defined bone volume and consider the difference between the groups.
The choice of the area of interest is subjective, and the counting methods are different between studies.
Although all papers deal with ceramic materials with the final goal of developing a biomaterial that may be capable of treating infection while promoting bone formation and osteointegration, this was evaluated in only 14 out of 46 series. Cao et al. [29] reported that the scaffold and bone were almost integrated with one another, along with the complete healing of all bone defects in the experimental groups, 12 weeks after implantation. For the same implantation time, Yan et al. [53] reported a large number of type I collagen fibers around the materials, with most of the material degraded and new trabecular bone and cartilage formed. Jia et al. [38], also 12 weeks after implantation, reported that, in the experimental group, newly formed bone was remodeled and restored to its original structural integrity. Xie et al. [52] showed that, 8 weeks after implantation, borate glass was mostly reabsorbed and replaced by new bone. Jiang et al. [39], using hydroxyapatite pellets, suggested that the infected bone became normal bone after 6 weeks, exhibiting reduced periosteal action and a well-shaped trabecular bone structure; however, Koort et al. [45], for the same endpoint, using bioactive glass, could not reach the same conclusion, suggesting the need for longer follow-up time points. Nelson et al. [19] used calcium sulfate pellets and reported that, at four weeks, tobramycin-loaded material showed 96% of the pellets as resorbed and 51% bone formation in the original defect compared to the control group, which showed 71% pellet resorption and 30% bone formation.
Most articles rely on imaging evaluation (histological or radiographic), which is mainly qualitative and observer-dependent, being subjective and introducing bias to the results [84]. Some papers use quantification scores, trying to bring some objectivity to the subjective evaluation, but even those scores, such as the most-used Smeltzer score, were described to evaluate bone infection in the absence of local treatment, which is not the case for the selected articles, and even the selection and definition of the region of interest for analysis are dependent on the observer's interpretation and judgment [84].

Conclusions
The addition of an antibiotic to a ceramic biomaterial seemed to be sufficient to make it effective in both the treatment of osteomyelitis and in preventing the evolution of a contaminated bone cavity to osteomyelitis while promoting bone formation and osteointegration for all of the animal models used.
Longer follow-up studies are required to observe the natural evolution of bone infection in animal models: some may have the ability to cure the infection by themselves even in the absence of antibiotics; others, which showed lower bacteria counts for a short follow-up time (but not zero) and were interpreted as a "good result", may evolve to chronic infection after the total release of the antibiotic from the biomaterial if the bacteria are still viable.
Appropriate protocols and a standard method of creating bone defects and osteomyelitis are recommended. The New Zealand White rabbit is a very adequate model for this purpose, but the use of a large animal model may be advisable to approximate the results to human pathology. Due to its similarity to human infection, the ideal place to develop a local infection, which was successfully used in most series, is the proximal tibial or distal femoral metaphysis. The use of a local adjuvant agent to create infection is not needed, as the injection of bacterial suspension is enough to develop osteomyelitis. Three weeks is the most adequate time to develop an infection and create a bone cavity without increasing the morbidity and mortality of the animals.
Future studies must evaluate the releasing profile of antibiotics in vivo and under infection. Although the releasing profile of most of the tested biomaterials has been studied in vitro or in healthy animals, there is a lack of knowledge of the drug release profile under conditions closer to reality. This is important data to ensure the absence of low concentrations of antibiotics for long periods in the infected area, which would facilitate the creation of bacterial resistance.
In conclusion, the addition of antibiotics to bioactive ceramic bone substitutes is, apparently, a good solution to treat infected bone cavities while allowing bone regeneration; however, it is not possible to say, as of now, what is the most effective biomaterial for this double purpose. The best way to promote a controlled release, which allows for a concentration above the minimum inhibitory concentration (MIC) for a long enough time to locally eradicate bacteria, is not yet defined. The preferred first-line antibiotic and the ideal ceramic vehicle (pure or in association with other substances) are still under investigation. It is important to follow consistent guidelines and develop appropriate models in order to shorten the amount of time between animal investigation and human application.