The Epidemiology of Antibiotic-Related Adverse Events in the Treatment of Diabetic Foot Infections: A Narrative Review of the Literature

The use of antibiotics for the treatment of diabetic foot infections (DFIs) over an extended period of time has been shown to be associated with adverse events (AEs), whereas interactions with concomitant patient medications must also be considered. The objective of this narrative review was to summarize the most frequent and most severe AEs reported in prospective trials and observational studies at the global level in DFI. Gastrointestinal intolerances were the most frequent AEs, from 5% to 22% among all therapies; this was more common when prolonged antibiotic administration was combined with oral beta-lactam or clindamycin or a higher dose of tetracyclines. The proportion of symptomatic colitis due to Clostridium difficile was variable depending on the antibiotic used (0.5% to 8%). Noteworthy serious AEs included hepatotoxicity due to beta-lactams (5% to 17%) or quinolones (3%); cytopenia’s related to linezolid (5%) and beta-lactams (6%); nausea under rifampicin, and renal failure under cotrimoxazole. Skin rash was found to rarely occur and was commonly associated with the use of penicillins or cotrimoxazole. AEs from prolonged antibiotic use in patients with DFI are costly in terms of longer hospitalization or additional monitoring care and can trigger additional investigations. The best way to prevent AEs is to keep the duration of antibiotic treatment short and with the lowest dose clinically necessary.


Introduction
Diabetic foot infections (DFIs) are the major cause of non-traumatic amputations globally and are only expected to increase in absolute numbers [1]. The current management of DFIs is multidisciplinary and encompasses elements of surgery, professional wound care, off-loading, revascularization, and antibiotic administration. For many clinicians and patients and their families, the use of antibiotics for DFI represents a 'last resort' option before imminent amputation or surgical resection of the infection. As a result, DFIs are likely a major reason for antibiotic misuse all over the world. Even if antibiotic therapy is successful for a given DFI episode, a subsequent infection, often with a new pathogen, may appear if the underlying local (or general) cause of the former infection has not been corrected [2]. However, the clinical reliance on systemic antibiotic therapy on long-term patient outcomes may not be supported by current epidemiological evidence.
The international IDSA and IWGDF practice guidelines provide helpful recommendations for clinicians to improve the diagnosis of infectious disease, to classify the severity of infections, to promote antibiotic stewardship in the field of DFI, and to guide antibiotic therapy [3][4][5]. However, for complete antibiotic stewardship [6], it is important to understand the epidemiology of the adverse events (AEs) of the most frequent antimicrobial agents in DFIs and the associated costs, the general economic diabetic foot costs, and the specific AE profile of an antimicrobial agent, which can contribute significantly to increased costs in higher monitoring care and additional days of hospitalization [7,8]. This is of particular importance when clinicians are not entirely convinced of the benefits of long-lasting antibiotic treatment and when the longer antibiotics are associated with the increase in multidrug-resistant microorganisms [9]. Moreover, we are aware of the vulnerable nature of comorbid DFI patients to AEs and interactions with concomitant medications [10].
In order to fill these clinical knowledge gaps, we performed a narrative review with an epidemiological focus, with the main aim to summarize the most frequent and most severe AEs reported in prospective trials and observational studies at the global level in DFI.

General Principles for the Use of Antibiotics in Diabetic Foot Infection
Systemic antibiotics administered orally or intravenously have been shown to be effective in the treatment of DFIs; however, none have been shown to be superior to others [11]. The choice and duration of antibiotic therapy depends on several clinical considerations, such as the severity of the infection, soft tissue versus bone infection, and surgical versus conservative treatment of the infection [5]. According to practice guidelines, for mild soft-tissue DFIs an antibiotic duration of 7-10 days is appropriate [12]. However, this can be prolonged for up to 3-4 weeks in cases of severe infection. For example, diabetic foot osteomyelitis (DFO) typically requires a prolonged duration of treatment. In cases with residual infected bone after (partial) amputation, the post-surgical duration is set at 2-4 weeks. In strictly conservative treatment without surgery, the current recommended duration is 4-6 weeks [5]. Interestingly, recommendations that describe the duration of antibiotic prescription before surgical debridement are censored, and the recommendations of duration that apply for the postoperative period are under debate. Likewise, the concomitant use of topical or of intraosseous antibiotic administrations does not alter the recommendations for the simultaneous systemic antibiotic duration in the DFO setting.

Choice of Antibiotic Agents
Clinicians should individualize the choice of antibiotic agent based on the pathogen, its susceptibilities, patient comorbidities, potential drug-drug interactions, and contraindications. Although it is beyond the scope of this focused manuscript to report on all possible drug-drug interactions, Supplementary Table S1 provides an overview of the most relevant interactions in the therapy of patients with DFI [13].
Beta-lactams are the most frequently used antibiotic class in DFIs for parenteral and/or empirical therapy [14]. The polymicrobial etiology of DFI, with Gram-positive cocci as the main protagonists and others including enterococci [15] and anaerobes, can be covered with the use of amoxicillin/clavulanate (A/C) and piperacillin/tazobactam (P/T), the latter also treated against non-fermenting rods such as Pseudomonas aeruginosa. However, these antibiotics do not cover bacteria stemming from skin commensals in DFO or methicillinresistant Staphylococcus aureus (MRSA) [16]. In acute and severe infections, the use of P/T and carbapenems may cover almost all bacteria. Regarding the less virulent skin commensal Gram-positive bacteria in chronic DFO, including for MRSA and vancomycin-resistant enterococci, clinicians typically refer to glycopeptides, lipopeptides, and lipoglycopeptides (e.g., vancomycin, teicoplanin, daptomycin, dalbavancin, oritavancin) or the oxazolidinone family of antibiotics (linezolid and tedizolid) for treatment.

General Findings
We first performed a literature search in the English language on the PubMed database for relevant prospective trials and observational studies related to DFIs. During the first search, the terms "diabetic foot" and "antibiotics" were used. This search retrieved 144 articles, with 22 studies selected for full text review. A second search was conducted using terms related to "soft tissue infection" or "skin-structure infection and antibiotics", which resulted 1151 potentially relevant articles. Of these, 264 articles underwent full text analysis and 29 were included in this review. Two additional articles were found through nonelectronic methods and included in this review. In total, we included 53 studies. Details of the articles are shown in Table 1 with information on AEs expressed in percentages during and/or after systemic antibiotic therapy. The percentage was calculated as a proportion of the number of adverse reactions associated with the study antibiotics in relation to the total number of patients in the study group.
From these included articles, we found the incidence of AEs varied between 2% and 30% for patients with DFI. The most common AEs were gastrointestinal in nature, covering 5% to 22% of all reported AEs. In the following text, we summarize the most common AEs in patients with DFI stratified against a variety of antibiotic classes through the accurate revision of the selected articles.

Penicillins
AEs ranged from the lowest 2.8% (95% confidence interval, 95% CI 2-4%) [17] to the highest 22% (95% CI 17-29%) [18] from the penicillin or penicillin derivatives group. Skin reactions for various (amino)penicillins were described at low frequencies (1-3%), whereas anaphylaxis or systemic reactions were shown to be rare [12,19] among DFI patients. Gastrointestinal AEs were shown to be the most common and range from 4.5% with amoxicillin/clavulanate (A/C) [20] to between 4.1% and 6.5% with ten days of ticarcillin/clavulanate (T/C) or P/T treatment [21]. A higher proportion of AEs (14%) was found in studies with ampicillin/sulbactam (A/S) with prolonged administration of 10 to 22 days and with P/T for 17 days [22,23]. Additionally, we found 22% for ticarcillin/clavulanate (T/C) followed by A/C for 14 days [18]. In three studies, symptomatic Clostridium difficile colitis (CDI) was reported more frequently during beta-lactam therapy compared with other groups of antibiotic classes (0.5% for non-beta-lactam antibiotics [24] and 2% to 8-9% for beta-lactams such as A/S or I/C, respectively [25,26]). Other AEs reported under penicillin treatment are cytopenia, leucopenia, and thrombocytopenia, which range from 0.6% to 6.7% [27,28]. A symptomatic hepatotoxicity was described between 3% and 16.7%, with higher proportions reported during P/T treatment [28,29]. The high sodium content within intravenous solutions of some beta-lactams such as P/T can initiate decompensated cardiac failure [12]. In addition, penicillins used in combination with fosfomycin may increase the risk for gastrointestinal AEs [30].

Carbapenems
Gastrointestinal perturbances were found as the most common AE occurring in between 3% and 12% of cases for imipenem/cilastatine (I/C) or meropenem treatments. Symptomatic CDI occurred at a rate of 8 to 9% during I/C treatment [25,36]. Hepatotoxicity was described to be up to 10% with ertapenem [37]. Seizures, a severe AE, were described with carbapenems (meropenem and I/C) in 0.2% of patients [38] and were specifically slightly higher with I/C, from 2% to 2.2% [25,26].

Fluoroquinolones
Although gastrointestinal AEs are also the most frequently reported, quinolones show fewer AEs related to diarrhea when compared with beta-lactam therapy (4 to 5%) [39,40]. Rarer AEs include hepatotoxicity, interstitial nephritis, or deliriant central nervous system effects [41,42]. In a clinical trial, ofloxacin showed neurological symptoms in 16% of cases (i.e., headache and insomnia) after a mean duration of 21 days after starting treatment [43]. Pefloxacin yielded a global risk for gastrointestinal AEs of 8.8% [44]. Sparfloxacin presented symptomatic photosensitivity in 11% of cases, whereas this was shown to occur in 2% of cases with clinafloxacin and 0.7% with ciprofloxacin [45,46]. An asymptomatic prolongation of the QTc time was observed for sparfloxacin in 0.5% to 2.4% of the therapeutic episodes [24,47]. An asymptomatic hepatitis due to fluoroquinolones was shown to occur in 0.2% to 3.2% of cases [24,48]. In 2018, the Food and Drug Administration (FDA) warned about possible fluoroquinolone-associated ruptures of pre-existing aortic aneurysms. Since then, other reports have echoed a similar claim [49]. The current recommendation is to avoid quinolones in cases of severe tendinopathies or aortic aneurisms. However, we did not identify any reports on aneurism ruptures related to the use of quinolones specifically in the DFI population. Likewise, despite an abundance of foot and ankle pathologies in adult diabetic populations, no reports of increased Achilles tendon disease were found in the DFI population. Newer quinolones, such as delafloxacin and levonadifloxacin, with broadspectrum antimicrobial activity (including for MRSA strains) showed a low risk for AEs. For example, hepatic alterations occurred in 3.2% of patients and hyper/hypoglycemia was shown to occur in between 0.1% and 0.3% of cases for delafloxacin, and gastrointestinal symptoms occurred in 5.2% of cases for levonadifloxacin [48,50].

Rifampin/Rifampicin
In a landmark randomized clinical trial regarding the duration of antibiotic therapy for the conservative management of DFO, 27 (68%) of the DFI patients received rifampicin and 19 (48%) received rifampicin with levofloxacin. In this study, antibiotic-related adverse events were recorded in 16 patients (40%); 6 patients (30%) at 6 weeks of follow-up and 10 patients (50%) at the 12-week follow-up timepoint. The most common AEs found were gastrointestinal (nausea and vomiting), all attributable to rifampicin at the 12-week followup timepoint (10%) [51]. Rifampin also has potential for significant drug interactions in addition to these AEs (see Supplemental Table S1).

Glycopeptides, Lipopeptides, and Lipoglycopeptides
Glycopeptides such as vancomycin are commonly associated with hypersensitivity reactions such as infusion reaction and rash, known as red-man syndrome, which is an expected histamine release with immediate skin rash (3-40%) [52,53]. Infusion reactions have also less been described with telavancin, dalbavancin, and oritavancin. Other true AEs may include vancomycin nephrotoxicity [54,55] or elevated serum creatin-phosphokinase or rhabdomyolysis with daptomycin (2.8-3%) [56,57]. Data on newer lipoglycopeptide drugs indicate that dalbavancin showed reduced AEs in soft-tissue infections [58], with less nausea, diarrhea, and pruritus [59,60]. Many of these relatively new drugs are not commonly used in DFI; however, experts expect that their use will increase in the coming decades. Of note, teicoplanin is not available in every country.

Oxazolidinones: Linezolid and Tedizolid
These agents, both with nearly universal anti-Gram-positive activity, may cause potentially substantial toxicities (transient myelosuppression or lactic acidosis and optic or peripheral neuropathy) due to impaired mitochondrial protein synthesis and consecutive mitochondrial dysfunction. A common feature is the expected drop in serum platelet and hemoglobulin counts, which usually occurs two weeks after the start of treatment [52,54]. Linezolid, compared with other oxazolidinones, generally results in anemia and thrombocytopenia that is entirely related to the duration of treatment. Patients who developed anemia after linezolid typically did so when treatment lasted more than 14 days [61]. Additionally, clinical trials comparing tedizolid with linezolid revealed reduced thrombocytopenia (0% vs. 4.9%) and neutropenia-pancytopenia (0% vs. 2.4%) from tedizolid after an average duration of 10 days [62].

Aminoglycosides
Our search revealed that no clinical trials or multicenter studies with the use of an aminoglycoside as a systemic antibiotic for DFIs were published. However, gentamicin was the most frequently used topical antibiotic agent for infected diabetic foot ulcers and also takes place as intraosseous deposits for chronic osteomyelitis in the diabetic foot and other orthopedic infections. When used as gentamicin sponges for infected diabetic foot ulcers, gentamycin showed no systemic AE and was very well tolerated [63,64]. However, the use of any topical antibiotic treatment is not yet recommended in international guidelines, at least not as a combination with systemic antimicrobial agents [5].

Tetracyclines
Gastrointestinal symptoms are the most common AE in this antibiotic group. According to reports captured through our search, nausea and vomiting may occur up to 25% (95% CI, 19-32%), 35% (95% CI, 31-39%), or 40% (95% CI. 36-45%) of the time [65][66][67]. Furthermore, the newer tetracyclines (omadacycline) showed greater gastrointestinal tolerance at standard doses. However, a higher loading dose was still associated with nausea and vomiting [68]. Few studies showed lower percentages of gastrointestinal events (less than 2%) when the duration of treatment and the prescription of other antibiotics was at the discretion of the physician [69].

Cotrimoxazole
As widely experienced by many clinicians, some articles confirmed the potential for a worsening of transient kidney function in the laboratory with the combination of cotrimoxazole and cephalexin [70]. Other risks are seldom, such as skin rash or hepatitis.

Clindamycin
The most common AEs are gastrointestinal in nature, with few reported cases of pseudomembranous colitis due to Clostridium difficile. We found that the relation of CDI with clindamycin in monotherapy was around 1.7% for an average duration of 14 days [71]. -Overall, AEs were 18% (14-22%) in A vs. 16% (13-20%) in B.
-The most common AEs leading to withdrawal from the study in A were 4% for GI and cutaneous manifestation.

Discussion
Recurrent misuse of systemic antibiotics is high in the adult population with DFIs [6]. With long antibiotic treatments, the risk for antibiotic-related AEs may increase. Most systemic AEs occur during the first three weeks of treatment [19,63]. We have limited the review to the analysis of 53 selected articles looking for AEs documented in the published literature on DFI. Therefore, some AEs could be underrepresented if no articles were documented.
Overall, gastrointestinal intolerances were the most reported AE, ranging from 5% to 22% among all antibiotic therapies, although this occurred most frequently during oral betalactam or clindamycin use and at the increasing doses related to tetracyclines. Symptomatic colitis due to Clostridium difficile may be a severe complication and is higher for beta-lactams (8-9%) than other antibiotics. More serious AEs were reported: hepatotoxicity due to betalactams may not be as rare (5% to 17%) and it is important to be aware of that for monitoring care. The hepatotoxicity of fluoroquinolones is generally lower (3%), and there is cytopenia related to linezolid (5%) or beta-lactams (5% to 7%), nausea and multiple drug interactions under rifampicin, and renal insufficiency under cotrimoxazole.
Kidney function may be altered by a wide range of antibiotic agents. However, vancomycin (compared with other agents) showed a higher potential for reduced glomerular filtration, and monitoring levels for vancomycin could help to avoid toxicity for which a potential overdose may lead to dialysis. Furthermore, and despite that the selected studies could not prove this toxicity, the regular use of cotrimoxazole at high doses is associated with transient nephrotoxicity and metabolic acidosis [13] in serum laboratory controls.
Hematologic disturbances occur regularly with linezolid when prescribed for more than 10-14 days, with fewer cytopenia cases with tedizolid compared with linezolid. This mechanism of inhibition of bacterial ribosomes also has an effect on the mitochondrial protein synthesis of human cells and causes mitochondrial dysfunction with a decrease in ATP in the bone marrow precursor cells that leads to myelosuppression. This is important to know to fix the data to request a control hemogram blood test if treatment is to be continued for a longer period of time, i.e., more than 14 days. Since tedizolid is administered once a day, it allows longer recovery periods from mitochondrial dysfunction and can mitigate the higher intrinsic toxicity compared with linezolid [72].
Skin rash was shown to seldom occur and was typically related to use of penicillins or cotrimoxazole. We believe inguinal mycosis (intertrigo) is mostly recognized in only hospitalized patients and is therefore underreported in the literature. Severe cutaneous and systemic toxic reactions have been described, such as eosinophilic rash (DRESS syndrome) together with a systemic inflammatory response syndrome (SIRS) with beta-lactam drugs, clindamycin, and vancomycin. In addition, cotrimoxazole may also be a potential cause of SIRS. The Stevens-Johnson syndrome of severe skin toxicity is not described in our focused review on DFIs, but clinicians should also be aware of this severe complication of antibiotic therapy, it being more frequent with sulfonamides and beta-lactams.
All systemic antibiotic agents in diabetic patients with a poor glycemic control and classically with macerated inguinal regions due to overlapping abdominal fat tissues may produce fungal skin infections during or shortly after the antibiotic therapy (e.g., intertrigo). However, we found data regarding this clinically frequent observation to be sparce. One article attributed an overall incidence of 2% and 4% concerning all antibiotic classes [12,19], and in our Balgrist orthopedic cohort experience from 2018 to 2022, which included at least one AE in 250 patients from 2312 orthopedic infections including diabetic patients (737, 32%), skin mycosis affected 51 patients (2.2%) (personal communication IU and LSB; not published before).
The musculoskeletal system was an uncommonly described AE, despite the fact that the risk of tendinopathies is a hallmark of fluoroquinolones; however, the published literature was not related to that of the selected DFI articles. Regarding the risk of quinolone-related tendon pathology, it is reported in the literature as a rare adverse event in 0.5-2% of cases treated with fluoroquinolones [73].
With regards to neurological AEs, headache or insomnia is frequently described for many antibiotics. The incidences of such conditions are higher with quinolones, whereas seizures, confusions, or hallucinations are commonly associated with beta-lactam agents. Seizures are related to the beta-lactam ring, which binds to GABA receptors and alters ionic conduction. The most frequent neurological disturbances (i.e., somnolence or confusion) occur with 3rd-4th generation cephalosporins (high with cefepime) and imipenem. A switch to early oral therapy by other agents or a therapeutic drug monitoring of the serum cefepime level in elderly patients with impaired renal function can be an option to reduce neurological AEs in patients with DFI. In our review, within the low frequency of seizures due to carbapenems, imipenem was the most common causative agent. Of note, monobactams (aztreonam) are less involved in neurological events [74].
The main limitation of this review is the restriction to scientific publications in the English language only and to articles specifically targeting DFI antibiotic therapy. It is clear that the same antibiotic agents might represent other AE in other patient populations with fewer co-morbidities inherent to the DFI patient population. Such an expanded review would be beyond the scope of this article. Likewise, we excluded the problem of additional costs specifically related to the antibiotic-related AEs during the treatment of DFI (for which there are no specific data available) and we purposely skip the discussion of antimicrobial resistances in the management of DFI, which many clinicians would indicate as the most important adverse problem. However, the resistance problem is not a classical or random 'adverse event' of antibiotic therapy. It is a systematic flipside of each prolonged antibiotic administration in human and veterinary medicine. Articles on antimicrobial resistances in DFI treatment already exist and would complicate the content and structure of this manuscript.

Research Criteria
In October 2022 we searched PubMed using two distinct MeSH terms for studies relating to antibiotic-related AEs in DFI. We searched for relevant titles and abstracts with the search terms "antibiotics" AND "diabetic foot". Our second search included the search terms "antibiotics" AND ("skin-structure infection" OR "soft tissue infection") NOT ("skin grafting" OR "psoriasis" OR "tumor" OR "hematologic disease"). In both searches we included clinical studies, clinical trials, clinical trial protocols, multicenter studies, and randomized controlled trials that were published in the English language with adult patients 19 years and above between January 1990 and September 2022. Figure S1 shows our study selection flow diagram. Purposely, we excluded the problem of antimicrobial resistances linked to the (over)use of antibiotics, for which other and broader reviews are available.
We calculated 95% confidence intervals for proportions following the Clopper-Pearson approach.
In preparing this narrative review, we adhered to key applicable items for narrative reviews from the PRISMA statement [75].

Criteria for Considering Studies for This Review
We included randomized controlled trials (RCTs), nonrandomized or quasi-randomized controlled trials, and observational studies (prospective or retrospective observational epidemiological studies) from January 1990 to September 2022 in the English language. We included studies of people with diabetes mellitus (type 1 or 2) with any type of foot wound (ulcers of neuropathic or ischemic etiology) and with DFI (osteomyelitis or soft-tissue infections). All patients included in the selected studies had to be treated with systemic antibiotics (any type oral or parenteral) with well-described reports of adverse effects.

Exclusion Criteria
We excluded studies that evaluated infections other than soft-tissue infection or osteomyelitis at the same time as DFI, studies that lacked data on antibiotic-related AEs, studies that included a case series with less than 30 diabetic patients, and studies that reported on topical antibiotics.

Conclusions
There is a clear need for greater antibiotic stewardship in the clinical management of DFIs given the increasing worldwide incidence and the increased use of antibiotic therapies that lead to an increase the days of hospitalization and higher complications for patients. Epidemiological understanding of antibiotic-related AEs is considered the cornerstone of every stewardship program and cannot be limited to only the acquisition of resistances and costs. The medical complications that arise from some AEs are not negligible.
The proportion of AEs in patients during or shortly after antibiotic therapy for DFI was shown to be as high as 30%. The risks and benefits of prolonged antibiotic use must be considered in each clinical situation, and knowledge of the most common and most severe AEs is useful for the clinician to be aware of the presentation of these AEs. A proven method to protect against such AEs is to reduce the duration of antibiotic treatment with lowered dosages as deemed clinically necessary. As mentioned by the stewardship program, clinicians should have an understanding of possible AEs and inform patients in cases where longer antibiotic therapy is deemed necessary.

Conflicts of Interest:
The authors declare no conflict of interest.